summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMaarten Lankhorst <maarten.lankhorst@linux.intel.com>2024-11-04 14:42:05 +0100
committerMaarten Lankhorst <maarten.lankhorst@linux.intel.com>2024-11-04 14:42:05 +0100
commitf9eea2e05e38aca32a1dda21906d035f10024530 (patch)
treecbe1c015b9278c2f65056ded67ac8e6f0dda1550
parenta886d61dd26710c8165866acc4f2718a7130cabc (diff)
2024y-11m-04d-13h-35m-52s UTC: drm-tip rerere cache update
git version 2.45.2
-rw-r--r--rr-cache/08258303c29ea30c680903186b0af4c79e46654f/preimage442
-rw-r--r--rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.42234
-rw-r--r--rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.52234
-rw-r--r--rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.22248
-rw-r--r--rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.32248
-rw-r--r--rr-cache/66926cdbe314a619c9f8586b1353ca1d3570e1e8/preimage429
-rw-r--r--rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.6982
-rw-r--r--rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.7982
-rw-r--r--rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/postimage451
-rw-r--r--rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/preimage480
-rw-r--r--rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage423
-rw-r--r--rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.1423
-rw-r--r--rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.2423
-rw-r--r--rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.102244
-rw-r--r--rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.92244
15 files changed, 0 insertions, 18487 deletions
diff --git a/rr-cache/08258303c29ea30c680903186b0af4c79e46654f/preimage b/rr-cache/08258303c29ea30c680903186b0af4c79e46654f/preimage
deleted file mode 100644
index 59f2b277d105..000000000000
--- a/rr-cache/08258303c29ea30c680903186b0af4c79e46654f/preimage
+++ /dev/null
@@ -1,442 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.enable_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.enable_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(xe);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(xe);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(xe);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.enable_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(xe);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_disable(&xe->drm);
-<<<<<<<
-=======
- if (!runtime)
- intel_display_driver_disable_user_access(xe);
- }
->>>>>>>
-
- if (!runtime)
- intel_display_driver_suspend(xe);
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
-<<<<<<<
-=======
- if (!runtime && has_display(xe))
- intel_display_driver_suspend_access(xe);
-
->>>>>>>
- intel_encoder_suspend_all(&xe->display);
-
- intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
-<<<<<<<
-=======
- if (!runtime && has_display(xe))
- intel_display_driver_resume_access(xe);
-
->>>>>>>
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime)
- intel_display_driver_resume(xe);
-
-<<<<<<<
-=======
- if (has_display(xe)) {
- drm_kms_helper_poll_enable(&xe->drm);
- if (!runtime)
- intel_display_driver_enable_user_access(xe);
- }
->>>>>>>
- intel_hpd_poll_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_enable(&xe->drm);
-
- intel_opregion_resume(xe);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.enable_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.4 b/rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.4
deleted file mode 100644
index 232d29e9a561..000000000000
--- a/rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.4
+++ /dev/null
@@ -1,2234 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2022 Intel Corporation
- */
-
-#include <linux/dma-fence-chain.h>
-
-#include "xe_pt.h"
-
-#include "regs/xe_gtt_defs.h"
-#include "xe_bo.h"
-#include "xe_device.h"
-#include "xe_drm_client.h"
-#include "xe_gt.h"
-#include "xe_gt_tlb_invalidation.h"
-#include "xe_migrate.h"
-#include "xe_pt_types.h"
-#include "xe_pt_walk.h"
-#include "xe_res_cursor.h"
-#include "xe_trace.h"
-#include "xe_ttm_stolen_mgr.h"
-#include "xe_vm.h"
-
-struct xe_pt_dir {
- struct xe_pt pt;
- /** @children: Array of page-table child nodes */
- struct xe_ptw *children[XE_PDES];
-};
-
-#if IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM)
-#define xe_pt_set_addr(__xe_pt, __addr) ((__xe_pt)->addr = (__addr))
-#define xe_pt_addr(__xe_pt) ((__xe_pt)->addr)
-#else
-#define xe_pt_set_addr(__xe_pt, __addr)
-#define xe_pt_addr(__xe_pt) 0ull
-#endif
-
-static const u64 xe_normal_pt_shifts[] = {12, 21, 30, 39, 48};
-static const u64 xe_compact_pt_shifts[] = {16, 21, 30, 39, 48};
-
-#define XE_PT_HIGHEST_LEVEL (ARRAY_SIZE(xe_normal_pt_shifts) - 1)
-
-static struct xe_pt_dir *as_xe_pt_dir(struct xe_pt *pt)
-{
- return container_of(pt, struct xe_pt_dir, pt);
-}
-
-static struct xe_pt *xe_pt_entry(struct xe_pt_dir *pt_dir, unsigned int index)
-{
- return container_of(pt_dir->children[index], struct xe_pt, base);
-}
-
-static u64 __xe_pt_empty_pte(struct xe_tile *tile, struct xe_vm *vm,
- unsigned int level)
-{
- struct xe_device *xe = tile_to_xe(tile);
- u16 pat_index = xe->pat.idx[XE_CACHE_WB];
- u8 id = tile->id;
-
- if (!xe_vm_has_scratch(vm))
- return 0;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return vm->pt_ops->pde_encode_bo(vm->scratch_pt[id][level - 1]->bo,
- 0, pat_index);
-
- return vm->pt_ops->pte_encode_addr(xe, 0, pat_index, level, IS_DGFX(xe), 0) |
- XE_PTE_NULL;
-}
-
-static void xe_pt_free(struct xe_pt *pt)
-{
- if (pt->level)
- kfree(as_xe_pt_dir(pt));
- else
- kfree(pt);
-}
-
-/**
- * xe_pt_create() - Create a page-table.
- * @vm: The vm to create for.
- * @tile: The tile to create for.
- * @level: The page-table level.
- *
- * Allocate and initialize a single struct xe_pt metadata structure. Also
- * create the corresponding page-table bo, but don't initialize it. If the
- * level is grater than zero, then it's assumed to be a directory page-
- * table and the directory structure is also allocated and initialized to
- * NULL pointers.
- *
- * Return: A valid struct xe_pt pointer on success, Pointer error code on
- * error.
- */
-struct xe_pt *xe_pt_create(struct xe_vm *vm, struct xe_tile *tile,
- unsigned int level)
-{
- struct xe_pt *pt;
- struct xe_bo *bo;
- int err;
-
- if (level) {
- struct xe_pt_dir *dir = kzalloc(sizeof(*dir), GFP_KERNEL);
-
- pt = (dir) ? &dir->pt : NULL;
- } else {
- pt = kzalloc(sizeof(*pt), GFP_KERNEL);
- }
- if (!pt)
- return ERR_PTR(-ENOMEM);
-
- pt->level = level;
- bo = xe_bo_create_pin_map(vm->xe, tile, vm, SZ_4K,
- ttm_bo_type_kernel,
- XE_BO_FLAG_VRAM_IF_DGFX(tile) |
- XE_BO_FLAG_IGNORE_MIN_PAGE_SIZE |
- XE_BO_FLAG_PINNED |
- XE_BO_FLAG_NO_RESV_EVICT |
- XE_BO_FLAG_PAGETABLE);
- if (IS_ERR(bo)) {
- err = PTR_ERR(bo);
- goto err_kfree;
- }
- pt->bo = bo;
- pt->base.children = level ? as_xe_pt_dir(pt)->children : NULL;
-
- if (vm->xef)
- xe_drm_client_add_bo(vm->xef->client, pt->bo);
- xe_tile_assert(tile, level <= XE_VM_MAX_LEVEL);
-
- return pt;
-
-err_kfree:
- xe_pt_free(pt);
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_populate_empty() - Populate a page-table bo with scratch- or zero
- * entries.
- * @tile: The tile the scratch pagetable of which to use.
- * @vm: The vm we populate for.
- * @pt: The pagetable the bo of which to initialize.
- *
- * Populate the page-table bo of @pt with entries pointing into the tile's
- * scratch page-table tree if any. Otherwise populate with zeros.
- */
-void xe_pt_populate_empty(struct xe_tile *tile, struct xe_vm *vm,
- struct xe_pt *pt)
-{
- struct iosys_map *map = &pt->bo->vmap;
- u64 empty;
- int i;
-
- if (!xe_vm_has_scratch(vm)) {
- /*
- * FIXME: Some memory is allocated already allocated to zero?
- * Find out which memory that is and avoid this memset...
- */
- xe_map_memset(vm->xe, map, 0, 0, SZ_4K);
- } else {
- empty = __xe_pt_empty_pte(tile, vm, pt->level);
- for (i = 0; i < XE_PDES; i++)
- xe_pt_write(vm->xe, map, i, empty);
- }
-}
-
-/**
- * xe_pt_shift() - Return the ilog2 value of the size of the address range of
- * a page-table at a certain level.
- * @level: The level.
- *
- * Return: The ilog2 value of the size of the address range of a page-table
- * at level @level.
- */
-unsigned int xe_pt_shift(unsigned int level)
-{
- return XE_PTE_SHIFT + XE_PDE_SHIFT * level;
-}
-
-/**
- * xe_pt_destroy() - Destroy a page-table tree.
- * @pt: The root of the page-table tree to destroy.
- * @flags: vm flags. Currently unused.
- * @deferred: List head of lockless list for deferred putting. NULL for
- * immediate putting.
- *
- * Puts the page-table bo, recursively calls xe_pt_destroy on all children
- * and finally frees @pt. TODO: Can we remove the @flags argument?
- */
-void xe_pt_destroy(struct xe_pt *pt, u32 flags, struct llist_head *deferred)
-{
- int i;
-
- if (!pt)
- return;
-
- XE_WARN_ON(!list_empty(&pt->bo->ttm.base.gpuva.list));
- xe_bo_unpin(pt->bo);
- xe_bo_put_deferred(pt->bo, deferred);
-
- if (pt->level > 0 && pt->num_live) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
-
- for (i = 0; i < XE_PDES; i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i), flags,
- deferred);
- }
- }
- xe_pt_free(pt);
-}
-
-/**
- * DOC: Pagetable building
- *
- * Below we use the term "page-table" for both page-directories, containing
- * pointers to lower level page-directories or page-tables, and level 0
- * page-tables that contain only page-table-entries pointing to memory pages.
- *
- * When inserting an address range in an already existing page-table tree
- * there will typically be a set of page-tables that are shared with other
- * address ranges, and a set that are private to this address range.
- * The set of shared page-tables can be at most two per level,
- * and those can't be updated immediately because the entries of those
- * page-tables may still be in use by the gpu for other mappings. Therefore
- * when inserting entries into those, we instead stage those insertions by
- * adding insertion data into struct xe_vm_pgtable_update structures. This
- * data, (subtrees for the cpu and page-table-entries for the gpu) is then
- * added in a separate commit step. CPU-data is committed while still under the
- * vm lock, the object lock and for userptr, the notifier lock in read mode.
- * The GPU async data is committed either by the GPU or CPU after fulfilling
- * relevant dependencies.
- * For non-shared page-tables (and, in fact, for shared ones that aren't
- * existing at the time of staging), we add the data in-place without the
- * special update structures. This private part of the page-table tree will
- * remain disconnected from the vm page-table tree until data is committed to
- * the shared page tables of the vm tree in the commit phase.
- */
-
-struct xe_pt_update {
- /** @update: The update structure we're building for this parent. */
- struct xe_vm_pgtable_update *update;
- /** @parent: The parent. Used to detect a parent change. */
- struct xe_pt *parent;
- /** @preexisting: Whether the parent was pre-existing or allocated */
- bool preexisting;
-};
-
-struct xe_pt_stage_bind_walk {
- /** base: The base class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @vm: The vm we're building for. */
- struct xe_vm *vm;
- /** @tile: The tile we're building for. */
- struct xe_tile *tile;
- /** @default_pte: PTE flag only template. No address is associated */
- u64 default_pte;
- /** @dma_offset: DMA offset to add to the PTE. */
- u64 dma_offset;
- /**
- * @needs_64k: This address range enforces 64K alignment and
- * granularity.
- */
- bool needs_64K;
- /**
- * @vma: VMA being mapped
- */
- struct xe_vma *vma;
-
- /* Also input, but is updated during the walk*/
- /** @curs: The DMA address cursor. */
- struct xe_res_cursor *curs;
- /** @va_curs_start: The Virtual address coresponding to @curs->start */
- u64 va_curs_start;
-
- /* Output */
- struct xe_walk_update {
- /** @wupd.entries: Caller provided storage. */
- struct xe_vm_pgtable_update *entries;
- /** @wupd.num_used_entries: Number of update @entries used. */
- unsigned int num_used_entries;
- /** @wupd.updates: Tracks the update entry at a given level */
- struct xe_pt_update updates[XE_VM_MAX_LEVEL + 1];
- } wupd;
-
- /* Walk state */
- /**
- * @l0_end_addr: The end address of the current l0 leaf. Used for
- * 64K granularity detection.
- */
- u64 l0_end_addr;
- /** @addr_64K: The start address of the current 64K chunk. */
- u64 addr_64K;
- /** @found_64: Whether @add_64K actually points to a 64K chunk. */
- bool found_64K;
-};
-
-static int
-xe_pt_new_shared(struct xe_walk_update *wupd, struct xe_pt *parent,
- pgoff_t offset, bool alloc_entries)
-{
- struct xe_pt_update *upd = &wupd->updates[parent->level];
- struct xe_vm_pgtable_update *entry;
-
- /*
- * For *each level*, we could only have one active
- * struct xt_pt_update at any one time. Once we move on to a
- * new parent and page-directory, the old one is complete, and
- * updates are either already stored in the build tree or in
- * @wupd->entries
- */
- if (likely(upd->parent == parent))
- return 0;
-
- upd->parent = parent;
- upd->preexisting = true;
-
- if (wupd->num_used_entries == XE_VM_MAX_LEVEL * 2 + 1)
- return -EINVAL;
-
- entry = wupd->entries + wupd->num_used_entries++;
- upd->update = entry;
- entry->ofs = offset;
- entry->pt_bo = parent->bo;
- entry->pt = parent;
- entry->flags = 0;
- entry->qwords = 0;
-
- if (alloc_entries) {
- entry->pt_entries = kmalloc_array(XE_PDES,
- sizeof(*entry->pt_entries),
- GFP_KERNEL);
- if (!entry->pt_entries)
- return -ENOMEM;
- }
-
- return 0;
-}
-
-/*
- * NOTE: This is a very frequently called function so we allow ourselves
- * to annotate (using branch prediction hints) the fastpath of updating a
- * non-pre-existing pagetable with leaf ptes.
- */
-static int
-xe_pt_insert_entry(struct xe_pt_stage_bind_walk *xe_walk, struct xe_pt *parent,
- pgoff_t offset, struct xe_pt *xe_child, u64 pte)
-{
- struct xe_pt_update *upd = &xe_walk->wupd.updates[parent->level];
- struct xe_pt_update *child_upd = xe_child ?
- &xe_walk->wupd.updates[xe_child->level] : NULL;
- int ret;
-
- ret = xe_pt_new_shared(&xe_walk->wupd, parent, offset, true);
- if (unlikely(ret))
- return ret;
-
- /*
- * Register this new pagetable so that it won't be recognized as
- * a shared pagetable by a subsequent insertion.
- */
- if (unlikely(child_upd)) {
- child_upd->update = NULL;
- child_upd->parent = xe_child;
- child_upd->preexisting = false;
- }
-
- if (likely(!upd->preexisting)) {
- /* Continue building a non-connected subtree. */
- struct iosys_map *map = &parent->bo->vmap;
-
- if (unlikely(xe_child))
- parent->base.children[offset] = &xe_child->base;
-
- xe_pt_write(xe_walk->vm->xe, map, offset, pte);
- parent->num_live++;
- } else {
- /* Shared pt. Stage update. */
- unsigned int idx;
- struct xe_vm_pgtable_update *entry = upd->update;
-
- idx = offset - entry->ofs;
- entry->pt_entries[idx].pt = xe_child;
- entry->pt_entries[idx].pte = pte;
- entry->qwords++;
- }
-
- return 0;
-}
-
-static bool xe_pt_hugepte_possible(u64 addr, u64 next, unsigned int level,
- struct xe_pt_stage_bind_walk *xe_walk)
-{
- u64 size, dma;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return false;
-
- /* Does the virtual range requested cover a huge pte? */
- if (!xe_pt_covers(addr, next, level, &xe_walk->base))
- return false;
-
- /* Does the DMA segment cover the whole pte? */
- if (next - xe_walk->va_curs_start > xe_walk->curs->size)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- /* Is the DMA address huge PTE size aligned? */
- size = next - addr;
- dma = addr - xe_walk->va_curs_start + xe_res_dma(xe_walk->curs);
-
- return IS_ALIGNED(dma, size);
-}
-
-/*
- * Scan the requested mapping to check whether it can be done entirely
- * with 64K PTEs.
- */
-static bool
-xe_pt_scan_64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- struct xe_res_cursor curs = *xe_walk->curs;
-
- if (!IS_ALIGNED(addr, SZ_64K))
- return false;
-
- if (next > xe_walk->l0_end_addr)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- xe_res_next(&curs, addr - xe_walk->va_curs_start);
- for (; addr < next; addr += SZ_64K) {
- if (!IS_ALIGNED(xe_res_dma(&curs), SZ_64K) || curs.size < SZ_64K)
- return false;
-
- xe_res_next(&curs, SZ_64K);
- }
-
- return addr == next;
-}
-
-/*
- * For non-compact "normal" 4K level-0 pagetables, we want to try to group
- * addresses together in 64K-contigous regions to add a 64K TLB hint for the
- * device to the PTE.
- * This function determines whether the address is part of such a
- * segment. For VRAM in normal pagetables, this is strictly necessary on
- * some devices.
- */
-static bool
-xe_pt_is_pte_ps64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- /* Address is within an already found 64k region */
- if (xe_walk->found_64K && addr - xe_walk->addr_64K < SZ_64K)
- return true;
-
- xe_walk->found_64K = xe_pt_scan_64K(addr, addr + SZ_64K, xe_walk);
- xe_walk->addr_64K = addr;
-
- return xe_walk->found_64K;
-}
-
-static int
-xe_pt_stage_bind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_bind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- u16 pat_index = xe_walk->vma->pat_index;
- struct xe_pt *xe_parent = container_of(parent, typeof(*xe_parent), base);
- struct xe_vm *vm = xe_walk->vm;
- struct xe_pt *xe_child;
- bool covers;
- int ret = 0;
- u64 pte;
-
- /* Is this a leaf entry ?*/
- if (level == 0 || xe_pt_hugepte_possible(addr, next, level, xe_walk)) {
- struct xe_res_cursor *curs = xe_walk->curs;
- bool is_null = xe_vma_is_null(xe_walk->vma);
-
- XE_WARN_ON(xe_walk->va_curs_start != addr);
-
- pte = vm->pt_ops->pte_encode_vma(is_null ? 0 :
- xe_res_dma(curs) + xe_walk->dma_offset,
- xe_walk->vma, pat_index, level);
- pte |= xe_walk->default_pte;
-
- /*
- * Set the XE_PTE_PS64 hint if possible, otherwise if
- * this device *requires* 64K PTE size for VRAM, fail.
- */
- if (level == 0 && !xe_parent->is_compact) {
- if (xe_pt_is_pte_ps64K(addr, next, xe_walk)) {
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_64K;
- pte |= XE_PTE_PS64;
- } else if (XE_WARN_ON(xe_walk->needs_64K)) {
- return -EINVAL;
- }
- }
-
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, NULL, pte);
- if (unlikely(ret))
- return ret;
-
- if (!is_null)
- xe_res_next(curs, next - addr);
- xe_walk->va_curs_start = next;
- xe_walk->vma->gpuva.flags |= (XE_VMA_PTE_4K << level);
- *action = ACTION_CONTINUE;
-
- return ret;
- }
-
- /*
- * Descending to lower level. Determine if we need to allocate a
- * new page table or -directory, which we do if there is no
- * previous one or there is one we can completely replace.
- */
- if (level == 1) {
- walk->shifts = xe_normal_pt_shifts;
- xe_walk->l0_end_addr = next;
- }
-
- covers = xe_pt_covers(addr, next, level, &xe_walk->base);
- if (covers || !*child) {
- u64 flags = 0;
-
- xe_child = xe_pt_create(xe_walk->vm, xe_walk->tile, level - 1);
- if (IS_ERR(xe_child))
- return PTR_ERR(xe_child);
-
- xe_pt_set_addr(xe_child,
- round_down(addr, 1ull << walk->shifts[level]));
-
- if (!covers)
- xe_pt_populate_empty(xe_walk->tile, xe_walk->vm, xe_child);
-
- *child = &xe_child->base;
-
- /*
- * Prefer the compact pagetable layout for L0 if possible. Only
- * possible if VMA covers entire 2MB region as compact 64k and
- * 4k pages cannot be mixed within a 2MB region.
- * TODO: Suballocate the pt bo to avoid wasting a lot of
- * memory.
- */
- if (GRAPHICS_VERx100(tile_to_xe(xe_walk->tile)) >= 1250 && level == 1 &&
- covers && xe_pt_scan_64K(addr, next, xe_walk)) {
- walk->shifts = xe_compact_pt_shifts;
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_COMPACT;
- flags |= XE_PDE_64K;
- xe_child->is_compact = true;
- }
-
- pte = vm->pt_ops->pde_encode_bo(xe_child->bo, 0, pat_index) | flags;
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, xe_child,
- pte);
- }
-
- *action = ACTION_SUBTREE;
- return ret;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_bind_ops = {
- .pt_entry = xe_pt_stage_bind_entry,
-};
-
-/**
- * xe_pt_stage_bind() - Build a disconnected page-table tree for a given address
- * range.
- * @tile: The tile we're building for.
- * @vma: The vma indicating the address range.
- * @entries: Storage for the update entries used for connecting the tree to
- * the main tree at commit time.
- * @num_entries: On output contains the number of @entries used.
- *
- * This function builds a disconnected page-table tree for a given address
- * range. The tree is connected to the main vm tree for the gpu using
- * xe_migrate_update_pgtables() and for the cpu using xe_pt_commit_bind().
- * The function builds xe_vm_pgtable_update structures for already existing
- * shared page-tables, and non-existing shared and non-shared page-tables
- * are built and populated directly.
- *
- * Return 0 on success, negative error code on error.
- */
-static int
-xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- struct xe_device *xe = tile_to_xe(tile);
- struct xe_bo *bo = xe_vma_bo(vma);
- bool is_devmem = !xe_vma_is_userptr(vma) && bo &&
- (xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo));
- struct xe_res_cursor curs;
- struct xe_pt_stage_bind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_bind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .vm = xe_vma_vm(vma),
- .tile = tile,
- .curs = &curs,
- .va_curs_start = xe_vma_start(vma),
- .vma = vma,
- .wupd.entries = entries,
- .needs_64K = (xe_vma_vm(vma)->flags & XE_VM_FLAG_64K) && is_devmem,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- int ret;
-
- /**
- * Default atomic expectations for different allocation scenarios are as follows:
- *
- * 1. Traditional API: When the VM is not in LR mode:
- * - Device atomics are expected to function with all allocations.
- *
- * 2. Compute/SVM API: When the VM is in LR mode:
- * - Device atomics are the default behavior when the bo is placed in a single region.
- * - In all other cases device atomics will be disabled with AE=0 until an application
- * request differently using a ioctl like madvise.
- */
- if (vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT) {
- if (xe_vm_in_lr_mode(xe_vma_vm(vma))) {
- if (bo && xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- /**
- * If a SMEM+LMEM allocation is backed by SMEM, a device
- * atomics will cause a gpu page fault and which then
- * gets migrated to LMEM, bind such allocations with
- * device atomics enabled.
- */
- else if (is_devmem && !xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- } else {
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- }
-
- /**
- * Unset AE if the platform(PVC) doesn't support it on an
- * allocation
- */
- if (!xe->info.has_device_atomics_on_smem && !is_devmem)
- xe_walk.default_pte &= ~XE_USM_PPGTT_PTE_AE;
- }
-
- if (is_devmem) {
- xe_walk.default_pte |= XE_PPGTT_PTE_DM;
- xe_walk.dma_offset = vram_region_gpu_offset(bo->ttm.resource);
- }
-
- if (!xe_vma_has_no_bo(vma) && xe_bo_is_stolen(bo))
- xe_walk.dma_offset = xe_ttm_stolen_gpu_offset(xe_bo_device(bo));
-
- xe_bo_assert_held(bo);
-
- if (!xe_vma_is_null(vma)) {
- if (xe_vma_is_userptr(vma))
- xe_res_first_sg(to_userptr_vma(vma)->userptr.sg, 0,
- xe_vma_size(vma), &curs);
- else if (xe_bo_is_vram(bo) || xe_bo_is_stolen(bo))
- xe_res_first(bo->ttm.resource, xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- else
- xe_res_first_sg(xe_bo_sg(bo), xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- } else {
- curs.size = xe_vma_size(vma);
- }
-
- ret = xe_pt_walk_range(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- *num_entries = xe_walk.wupd.num_used_entries;
- return ret;
-}
-
-/**
- * xe_pt_nonshared_offsets() - Determine the non-shared entry offsets of a
- * shared pagetable.
- * @addr: The start address within the non-shared pagetable.
- * @end: The end address within the non-shared pagetable.
- * @level: The level of the non-shared pagetable.
- * @walk: Walk info. The function adjusts the walk action.
- * @action: next action to perform (see enum page_walk_action)
- * @offset: Ignored on input, First non-shared entry on output.
- * @end_offset: Ignored on input, Last non-shared entry + 1 on output.
- *
- * A non-shared page-table has some entries that belong to the address range
- * and others that don't. This function determines the entries that belong
- * fully to the address range. Depending on level, some entries may
- * partially belong to the address range (that can't happen at level 0).
- * The function detects that and adjust those offsets to not include those
- * partial entries. Iff it does detect partial entries, we know that there must
- * be shared page tables also at lower levels, so it adjusts the walk action
- * accordingly.
- *
- * Return: true if there were non-shared entries, false otherwise.
- */
-static bool xe_pt_nonshared_offsets(u64 addr, u64 end, unsigned int level,
- struct xe_pt_walk *walk,
- enum page_walk_action *action,
- pgoff_t *offset, pgoff_t *end_offset)
-{
- u64 size = 1ull << walk->shifts[level];
-
- *offset = xe_pt_offset(addr, level, walk);
- *end_offset = xe_pt_num_entries(addr, end, level, walk) + *offset;
-
- if (!level)
- return true;
-
- /*
- * If addr or next are not size aligned, there are shared pts at lower
- * level, so in that case traverse down the subtree
- */
- *action = ACTION_CONTINUE;
- if (!IS_ALIGNED(addr, size)) {
- *action = ACTION_SUBTREE;
- (*offset)++;
- }
-
- if (!IS_ALIGNED(end, size)) {
- *action = ACTION_SUBTREE;
- (*end_offset)--;
- }
-
- return *end_offset > *offset;
-}
-
-struct xe_pt_zap_ptes_walk {
- /** @base: The walk base-class */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're building for */
- struct xe_tile *tile;
-
- /* Output */
- /** @needs_invalidate: Whether we need to invalidate TLB*/
- bool needs_invalidate;
-};
-
-static int xe_pt_zap_ptes_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_zap_ptes_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- /*
- * Note that we're called from an entry callback, and we're dealing
- * with the child of that entry rather than the parent, so need to
- * adjust level down.
- */
- if (xe_pt_nonshared_offsets(addr, next, --level, walk, action, &offset,
- &end_offset)) {
- xe_map_memset(tile_to_xe(xe_walk->tile), &xe_child->bo->vmap,
- offset * sizeof(u64), 0,
- (end_offset - offset) * sizeof(u64));
- xe_walk->needs_invalidate = true;
- }
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_zap_ptes_ops = {
- .pt_entry = xe_pt_zap_ptes_entry,
-};
-
-/**
- * xe_pt_zap_ptes() - Zap (zero) gpu ptes of an address range
- * @tile: The tile we're zapping for.
- * @vma: GPU VMA detailing address range.
- *
- * Eviction and Userptr invalidation needs to be able to zap the
- * gpu ptes of a given address range in pagefaulting mode.
- * In order to be able to do that, that function needs access to the shared
- * page-table entrieaso it can either clear the leaf PTEs or
- * clear the pointers to lower-level page-tables. The caller is required
- * to hold the necessary locks to ensure neither the page-table connectivity
- * nor the page-table entries of the range is updated from under us.
- *
- * Return: Whether ptes were actually updated and a TLB invalidation is
- * required.
- */
-bool xe_pt_zap_ptes(struct xe_tile *tile, struct xe_vma *vma)
-{
- struct xe_pt_zap_ptes_walk xe_walk = {
- .base = {
- .ops = &xe_pt_zap_ptes_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- u8 pt_mask = (vma->tile_present & ~vma->tile_invalidated);
-
- if (!(pt_mask & BIT(tile->id)))
- return false;
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.needs_invalidate;
-}
-
-static void
-xe_vm_populate_pgtable(struct xe_migrate_pt_update *pt_update, struct xe_tile *tile,
- struct iosys_map *map, void *data,
- u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_pt_entry *ptes = update->pt_entries;
- u64 *ptr = data;
- u32 i;
-
- for (i = 0; i < num_qwords; i++) {
- if (map)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, ptes[i].pte);
- else
- ptr[i] = ptes[i].pte;
- }
-}
-
-static void xe_pt_abort_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- u32 i, j;
-
- for (i = 0; i < num_entries; i++) {
- if (!entries[i].pt_entries)
- continue;
-
- for (j = 0; j < entries[i].qwords; j++)
- xe_pt_destroy(entries[i].pt_entries[j].pt, xe_vma_vm(vma)->flags, NULL);
- kfree(entries[i].pt_entries);
- }
-}
-
-static void xe_pt_commit_locks_assert(struct xe_vma *vma)
-{
- struct xe_vm *vm = xe_vma_vm(vma);
-
- lockdep_assert_held(&vm->lock);
-
- if (xe_vma_is_userptr(vma))
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- else if (!xe_vma_is_null(vma))
- dma_resv_assert_held(xe_vma_bo(vma)->ttm.base.resv);
-
- xe_vm_assert_held(vm);
-}
-
-static void xe_pt_commit_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries, bool rebind,
- struct llist_head *deferred)
-{
- u32 i, j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (i = 0; i < num_entries; i++) {
- struct xe_pt *pt = entries[i].pt;
- struct xe_pt_dir *pt_dir;
-
- if (!rebind)
- pt->num_live += entries[i].qwords;
-
- if (!pt->level) {
- kfree(entries[i].pt_entries);
- continue;
- }
-
- pt_dir = as_xe_pt_dir(pt);
- for (j = 0; j < entries[i].qwords; j++) {
- u32 j_ = j + entries[i].ofs;
- struct xe_pt *newpte = entries[i].pt_entries[j].pt;
-
- if (xe_pt_entry(pt_dir, j_))
- xe_pt_destroy(xe_pt_entry(pt_dir, j_),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[j_] = &newpte->base;
- }
- kfree(entries[i].pt_entries);
- }
-}
-
-static int
-xe_pt_prepare_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- int err;
-
- *num_entries = 0;
- err = xe_pt_stage_bind(tile, vma, entries, num_entries);
- if (!err)
- xe_tile_assert(tile, *num_entries);
- else /* abort! */
- xe_pt_abort_bind(vma, entries, *num_entries);
-
- return err;
-}
-
-static void xe_vm_dbg_print_entries(struct xe_device *xe,
- const struct xe_vm_pgtable_update *entries,
- unsigned int num_entries)
-#if (IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM))
-{
- unsigned int i;
-
- vm_dbg(&xe->drm, "%u entries to update\n", num_entries);
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
- struct xe_pt *xe_pt = entry->pt;
- u64 page_size = 1ull << xe_pt_shift(xe_pt->level);
- u64 end;
- u64 start;
-
- xe_assert(xe, !entry->pt->is_compact);
- start = entry->ofs * page_size;
- end = start + page_size * entry->qwords;
- vm_dbg(&xe->drm,
- "\t%u: Update level %u at (%u + %u) [%llx...%llx) f:%x\n",
- i, xe_pt->level, entry->ofs, entry->qwords,
- xe_pt_addr(xe_pt) + start, xe_pt_addr(xe_pt) + end, 0);
- }
-}
-#else
-{}
-#endif
-
-#ifdef CONFIG_DRM_XE_USERPTR_INVAL_INJECT
-
-static int xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- u32 divisor = uvma->userptr.divisor ? uvma->userptr.divisor : 2;
- static u32 count;
-
- if (count++ % divisor == divisor - 1) {
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
-
- uvma->userptr.divisor = divisor << 1;
- spin_lock(&vm->userptr.invalidated_lock);
- list_move_tail(&uvma->userptr.invalidate_link,
- &vm->userptr.invalidated);
- spin_unlock(&vm->userptr.invalidated_lock);
- return true;
- }
-
- return false;
-}
-
-#else
-
-static bool xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- return false;
-}
-
-#endif
-
-/**
- * struct xe_pt_migrate_pt_update - Callback argument for pre-commit callbacks
- * @base: Base we derive from.
- * @bind: Whether this is a bind or an unbind operation. A bind operation
- * makes the pre-commit callback error with -EAGAIN if it detects a
- * pending invalidation.
- * @locked: Whether the pre-commit callback locked the userptr notifier lock
- * and it needs unlocking.
- */
-struct xe_pt_migrate_pt_update {
- struct xe_migrate_pt_update base;
- bool bind;
- bool locked;
-};
-
-/*
- * This function adds the needed dependencies to a page-table update job
- * to make sure racing jobs for separate bind engines don't race writing
- * to the same page-table range, wreaking havoc. Initially use a single
- * fence for the entire VM. An optimization would use smaller granularity.
- */
-static int xe_pt_vm_dependencies(struct xe_sched_job *job,
- struct xe_range_fence_tree *rftree,
- u64 start, u64 last)
-{
- struct xe_range_fence *rtfence;
- struct dma_fence *fence;
- int err;
-
- rtfence = xe_range_fence_tree_first(rftree, start, last);
- while (rtfence) {
- fence = rtfence->fence;
-
- if (!dma_fence_is_signaled(fence)) {
- /*
- * Is this a CPU update? GPU is busy updating, so return
- * an error
- */
- if (!job)
- return -ETIME;
-
- dma_fence_get(fence);
- err = drm_sched_job_add_dependency(&job->drm, fence);
- if (err)
- return err;
- }
-
- rtfence = xe_range_fence_tree_next(rtfence, start, last);
- }
-
- return 0;
-}
-
-static int xe_pt_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_range_fence_tree *rftree =
- &xe_vma_vm(pt_update->vma)->rftree[pt_update->tile_id];
-
- return xe_pt_vm_dependencies(pt_update->job, rftree,
- pt_update->start, pt_update->last);
-}
-
-static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_pt_migrate_pt_update *userptr_update =
- container_of(pt_update, typeof(*userptr_update), base);
- struct xe_userptr_vma *uvma = to_userptr_vma(pt_update->vma);
- unsigned long notifier_seq = uvma->userptr.notifier_seq;
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
- int err = xe_pt_vm_dependencies(pt_update->job,
- &vm->rftree[pt_update->tile_id],
- pt_update->start,
- pt_update->last);
-
- if (err)
- return err;
-
- userptr_update->locked = false;
-
- /*
- * Wait until nobody is running the invalidation notifier, and
- * since we're exiting the loop holding the notifier lock,
- * nobody can proceed invalidating either.
- *
- * Note that we don't update the vma->userptr.notifier_seq since
- * we don't update the userptr pages.
- */
- do {
- down_read(&vm->userptr.notifier_lock);
- if (!mmu_interval_read_retry(&uvma->userptr.notifier,
- notifier_seq))
- break;
-
- up_read(&vm->userptr.notifier_lock);
-
- if (userptr_update->bind)
- return -EAGAIN;
-
- notifier_seq = mmu_interval_read_begin(&uvma->userptr.notifier);
- } while (true);
-
- /* Inject errors to test_whether they are handled correctly */
- if (userptr_update->bind && xe_pt_userptr_inject_eagain(uvma)) {
- up_read(&vm->userptr.notifier_lock);
- return -EAGAIN;
- }
-
- userptr_update->locked = true;
-
- return 0;
-}
-
-static const struct xe_migrate_pt_update_ops bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-struct invalidation_fence {
- struct xe_gt_tlb_invalidation_fence base;
- struct xe_gt *gt;
- struct dma_fence *fence;
- struct dma_fence_cb cb;
- struct work_struct work;
- u64 start;
- u64 end;
- u32 asid;
-};
-
-static void invalidation_fence_cb(struct dma_fence *fence,
- struct dma_fence_cb *cb)
-{
- struct invalidation_fence *ifence =
- container_of(cb, struct invalidation_fence, cb);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_cb(xe, &ifence->base);
- if (!ifence->fence->error) {
- queue_work(system_wq, &ifence->work);
- } else {
- ifence->base.base.error = ifence->fence->error;
- dma_fence_signal(&ifence->base.base);
- dma_fence_put(&ifence->base.base);
- }
- dma_fence_put(ifence->fence);
-}
-
-static void invalidation_fence_work_func(struct work_struct *w)
-{
- struct invalidation_fence *ifence =
- container_of(w, struct invalidation_fence, work);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_work_func(xe, &ifence->base);
- xe_gt_tlb_invalidation_range(ifence->gt, &ifence->base, ifence->start,
- ifence->end, ifence->asid);
-}
-
-static int invalidation_fence_init(struct xe_gt *gt,
- struct invalidation_fence *ifence,
- struct dma_fence *fence,
- u64 start, u64 end, u32 asid)
-{
- int ret;
-
- trace_xe_gt_tlb_invalidation_fence_create(gt_to_xe(gt), &ifence->base);
-
- xe_gt_tlb_invalidation_fence_init(gt, &ifence->base, false);
-
- ifence->fence = fence;
- ifence->gt = gt;
- ifence->start = start;
- ifence->end = end;
- ifence->asid = asid;
-
- INIT_WORK(&ifence->work, invalidation_fence_work_func);
- ret = dma_fence_add_callback(fence, &ifence->cb, invalidation_fence_cb);
- if (ret == -ENOENT) {
- dma_fence_put(ifence->fence); /* Usually dropped in CB */
- invalidation_fence_work_func(&ifence->work);
- } else if (ret) {
- dma_fence_put(&ifence->base.base); /* Caller ref */
- dma_fence_put(&ifence->base.base); /* Creation ref */
- }
-
- xe_gt_assert(gt, !ret || ret == -ENOENT);
-
- return ret && ret != -ENOENT ? ret : 0;
-}
-
-static void xe_pt_calc_rfence_interval(struct xe_vma *vma,
- struct xe_pt_migrate_pt_update *update,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- int i, level = 0;
-
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- update->base.start = ALIGN_DOWN(xe_vma_start(vma),
- 0x1ull << xe_pt_shift(level));
- update->base.last = ALIGN(xe_vma_end(vma),
- 0x1ull << xe_pt_shift(level)) - 1;
-}
-
-/**
- * __xe_pt_bind_vma() - Build and connect a page-table tree for the vma
- * address range.
- * @tile: The tile to bind for.
- * @vma: The vma to bind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before binding the built tree to the live vm tree.
- * @num_syncs: Number of @sync entries.
- * @rebind: Whether we're rebinding this vma to the same address range without
- * an unbind in-between.
- *
- * This function builds a page-table tree (see xe_pt_stage_bind() for more
- * information on page-table building), and the xe_vm_pgtable_update entries
- * abstracting the operations needed to attach it to the main vm tree. It
- * then takes the relevant locks and updates the metadata side of the main
- * vm tree and submits the operations for pipelined attachment of the
- * gpu page-table to the vm main tree, (which can be done either by the
- * cpu and the GPU).
- *
- * Return: A valid dma-fence representing the pipelined attachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs,
- bool rebind)
-{
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update bind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_bind_ops : &bind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
- .bind = true,
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL;
- struct xe_range_fence *rfence;
- int err;
-
- bind_pt_update.locked = false;
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- err = xe_pt_prepare_bind(tile, vma, entries, &num_entries);
- if (err)
- goto err;
-
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- goto err;
-
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &bind_pt_update, entries,
- num_entries);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. on LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * on !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!rebind && xe_vm_has_scratch(vm) && xe_vm_in_preempt_fence_mode(vm))) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
- } else if (rebind && !xe_vm_in_lr_mode(vm)) {
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
- }
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, xe_vma_bo(vma), q,
- entries, num_entries,
- syncs, num_syncs,
- &bind_pt_update.base);
- if (!IS_ERR(fence)) {
- bool last_munmap_rebind = vma->gpuva.flags & XE_VMA_LAST_REBIND;
- LLIST_HEAD(deferred);
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- bind_pt_update.base.start,
- bind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
- /* TLB invalidation must be done before signaling rebind */
- if (ifence) {
- int err = invalidation_fence_init(tile->primary_gt,
- ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
- }
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence, rebind ||
- last_munmap_rebind ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_bind(vma, entries, num_entries, rebind,
- bind_pt_update.locked ? &deferred : NULL);
-
- /* This vma is live (again?) now */
- vma->tile_present |= BIT(tile->id);
-
- if (bind_pt_update.locked) {
- to_userptr_vma(vma)->userptr.initial_bind = true;
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
- if (!rebind && last_munmap_rebind &&
- xe_vm_in_preempt_fence_mode(vm))
- xe_vm_queue_rebind_worker(vm);
- } else {
- kfree(rfence);
- kfree(ifence);
- if (bind_pt_update.locked)
- up_read(&vm->userptr.notifier_lock);
- xe_pt_abort_bind(vma, entries, num_entries);
- }
-
- return fence;
-
-err:
- return ERR_PTR(err);
-}
-
-struct xe_pt_stage_unbind_walk {
- /** @base: The pagewalk base-class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're unbinding from. */
- struct xe_tile *tile;
-
- /**
- * @modified_start: Walk range start, modified to include any
- * shared pagetables that we're the only user of and can thus
- * treat as private.
- */
- u64 modified_start;
- /** @modified_end: Walk range start, modified like @modified_start. */
- u64 modified_end;
-
- /* Output */
- /* @wupd: Structure to track the page-table updates we're building */
- struct xe_walk_update wupd;
-};
-
-/*
- * Check whether this range is the only one populating this pagetable,
- * and in that case, update the walk range checks so that higher levels don't
- * view us as a shared pagetable.
- */
-static bool xe_pt_check_kill(u64 addr, u64 next, unsigned int level,
- const struct xe_pt *child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- unsigned int shift = walk->shifts[level];
- u64 size = 1ull << shift;
-
- if (IS_ALIGNED(addr, size) && IS_ALIGNED(next, size) &&
- ((next - addr) >> shift) == child->num_live) {
- u64 size = 1ull << walk->shifts[level + 1];
-
- *action = ACTION_CONTINUE;
-
- if (xe_walk->modified_start >= addr)
- xe_walk->modified_start = round_down(addr, size);
- if (xe_walk->modified_end <= next)
- xe_walk->modified_end = round_up(next, size);
-
- return true;
- }
-
- return false;
-}
-
-static int xe_pt_stage_unbind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- xe_pt_check_kill(addr, next, level - 1, xe_child, action, walk);
-
- return 0;
-}
-
-static int
-xe_pt_stage_unbind_post_descend(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
- u64 size = 1ull << walk->shifts[--level];
-
- if (!IS_ALIGNED(addr, size))
- addr = xe_walk->modified_start;
- if (!IS_ALIGNED(next, size))
- next = xe_walk->modified_end;
-
- /* Parent == *child is the root pt. Don't kill it. */
- if (parent != *child &&
- xe_pt_check_kill(addr, next, level, xe_child, action, walk))
- return 0;
-
- if (!xe_pt_nonshared_offsets(addr, next, level, walk, action, &offset,
- &end_offset))
- return 0;
-
- (void)xe_pt_new_shared(&xe_walk->wupd, xe_child, offset, false);
- xe_walk->wupd.updates[level].update->qwords = end_offset - offset;
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_unbind_ops = {
- .pt_entry = xe_pt_stage_unbind_entry,
- .pt_post_descend = xe_pt_stage_unbind_post_descend,
-};
-
-/**
- * xe_pt_stage_unbind() - Build page-table update structures for an unbind
- * operation
- * @tile: The tile we're unbinding for.
- * @vma: The vma we're unbinding.
- * @entries: Caller-provided storage for the update structures.
- *
- * Builds page-table update structures for an unbind operation. The function
- * will attempt to remove all page-tables that we're the only user
- * of, and for that to work, the unbind operation must be committed in the
- * same critical section that blocks racing binds to the same page-table tree.
- *
- * Return: The number of entries used.
- */
-static unsigned int xe_pt_stage_unbind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries)
-{
- struct xe_pt_stage_unbind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_unbind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- .modified_start = xe_vma_start(vma),
- .modified_end = xe_vma_end(vma),
- .wupd.entries = entries,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.wupd.num_used_entries;
-}
-
-static void
-xe_migrate_clear_pgtable_callback(struct xe_migrate_pt_update *pt_update,
- struct xe_tile *tile, struct iosys_map *map,
- void *ptr, u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_vma *vma = pt_update->vma;
- u64 empty = __xe_pt_empty_pte(tile, xe_vma_vm(vma), update->pt->level);
- int i;
-
- if (map && map->is_iomem)
- for (i = 0; i < num_qwords; ++i)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, empty);
- else if (map)
- memset64(map->vaddr + qword_ofs * sizeof(u64), empty,
- num_qwords);
- else
- memset64(ptr, empty, num_qwords);
-}
-
-static void
-xe_pt_commit_unbind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 num_entries,
- struct llist_head *deferred)
-{
- u32 j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (j = 0; j < num_entries; ++j) {
- struct xe_vm_pgtable_update *entry = &entries[j];
- struct xe_pt *pt = entry->pt;
-
- pt->num_live -= entry->qwords;
- if (pt->level) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
- u32 i;
-
- for (i = entry->ofs; i < entry->ofs + entry->qwords;
- i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[i] = NULL;
- }
- }
- }
-}
-
-<<<<<<<
-static const struct xe_migrate_pt_update_ops unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
-=======
-static void
-xe_pt_update_ops_rfence_interval(struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int i, level = 0;
- u64 start, last;
-
- for (i = 0; i < pt_op->num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &pt_op->entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- start = ALIGN_DOWN(xe_vma_start(vma), 0x1ull << xe_pt_shift(level));
- last = ALIGN(xe_vma_end(vma), 0x1ull << xe_pt_shift(level)) - 1;
-
- if (start < pt_update_ops->start)
- pt_update_ops->start = start;
- if (last > pt_update_ops->last)
- pt_update_ops->last = last;
-}
-
-static int vma_reserve_fences(struct xe_device *xe, struct xe_vma *vma)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- return dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv,
- xe->info.tile_count);
-
- return 0;
-}
-
-static int bind_op_prepare(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- pt_op->vma = NULL;
- pt_op->bind = true;
- pt_op->rebind = BIT(tile->id) & vma->tile_present;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- err = xe_pt_prepare_bind(tile, vma, pt_op->entries,
- &pt_op->num_entries);
- if (!err) {
- xe_tile_assert(tile, pt_op->num_entries <=
- ARRAY_SIZE(pt_op->entries));
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, true);
-
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. On LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * On !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!pt_op->rebind && xe_vm_has_scratch(vm) &&
- xe_vm_in_preempt_fence_mode(vm)))
- pt_update_ops->needs_invalidation = true;
- else if (pt_op->rebind && !xe_vm_in_lr_mode(vm))
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
-
- vma->tile_staged |= BIT(tile->id);
- pt_op->vma = vma;
- xe_pt_commit_prepare_bind(vma, pt_op->entries,
- pt_op->num_entries, pt_op->rebind);
- } else {
- xe_pt_cancel_bind(vma, pt_op->entries, pt_op->num_entries);
- }
-
- return err;
-}
-
-static int unbind_op_prepare(struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- if (!((vma->tile_present | vma->tile_staged) & BIT(tile->id)))
- return 0;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- /*
- * Wait for invalidation to complete. Can corrupt internal page table
- * state if an invalidation is running while preparing an unbind.
- */
- if (xe_vma_is_userptr(vma) && xe_vm_in_fault_mode(xe_vma_vm(vma)))
- mmu_interval_read_begin(&to_userptr_vma(vma)->userptr.notifier);
-
- pt_op->vma = vma;
- pt_op->bind = false;
- pt_op->rebind = false;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- pt_op->num_entries = xe_pt_stage_unbind(tile, vma, pt_op->entries);
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, false);
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
- pt_update_ops->needs_invalidation = true;
-
- xe_pt_commit_prepare_unbind(vma, pt_op->entries, pt_op->num_entries);
-
- return 0;
-}
-
-static int op_prepare(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op)
-{
- int err = 0;
-
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- err = bind_op_prepare(vm, tile, pt_update_ops, op->map.vma);
- pt_update_ops->wait_vm_kernel = true;
- break;
- case DRM_GPUVA_OP_REMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va));
-
- if (!err && op->remap.prev) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.prev);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- if (!err && op->remap.next) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.next);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- break;
- case DRM_GPUVA_OP_UNMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va));
- break;
- case DRM_GPUVA_OP_PREFETCH:
- err = bind_op_prepare(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va));
- pt_update_ops->wait_vm_kernel = true;
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-
- return err;
-}
-
-static void
-xe_pt_update_ops_init(struct xe_vm_pgtable_update_ops *pt_update_ops)
-{
- init_llist_head(&pt_update_ops->deferred);
- pt_update_ops->start = ~0x0ull;
- pt_update_ops->last = 0x0ull;
-}
-
-/**
- * xe_pt_update_ops_prepare() - Prepare PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Prepare PT update operations which includes updating internal PT state,
- * allocate memory for page tables, populate page table being pruned in, and
- * create PT update operations for leaf insertion / removal.
- *
- * Return: 0 on success, negative error code on error.
- */
-int xe_pt_update_ops_prepare(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct xe_vma_op *op;
- int err;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- xe_pt_update_ops_init(pt_update_ops);
-
- err = dma_resv_reserve_fences(xe_vm_resv(vops->vm),
- tile_to_xe(tile)->info.tile_count);
- if (err)
- return err;
-
- list_for_each_entry(op, &vops->list, link) {
- err = op_prepare(vops->vm, tile, pt_update_ops, op);
-
- if (err)
- return err;
- }
-
- xe_tile_assert(tile, pt_update_ops->current_op <=
- pt_update_ops->num_ops);
-
-#ifdef TEST_VM_OPS_ERROR
- if (vops->inject_error &&
- vops->vm->xe->vm_inject_error_position == FORCE_OP_ERROR_PREPARE)
- return -ENOSPC;
-#endif
-
- return 0;
-}
-
-static void bind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present |= BIT(tile->id);
- vma->tile_staged &= ~BIT(tile->id);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- to_userptr_vma(vma)->userptr.initial_bind = true;
- }
-
- /*
- * Kick rebind worker if this bind triggers preempt fences and not in
- * the rebind worker
- */
- if (pt_update_ops->wait_vm_bookkeep &&
- xe_vm_in_preempt_fence_mode(vm) &&
- !current->mm)
- xe_vm_queue_rebind_worker(vm);
-}
-
-static void unbind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present &= ~BIT(tile->id);
- if (!vma->tile_present) {
- list_del_init(&vma->combined_links.rebind);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
-
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- }
-}
-
-static void op_commit(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- bind_op_commit(vm, tile, pt_update_ops, op->map.vma, fence,
- fence2);
- break;
- case DRM_GPUVA_OP_REMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va), fence,
- fence2);
-
- if (op->remap.prev)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.prev,
- fence, fence2);
- if (op->remap.next)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.next,
- fence, fence2);
- break;
- case DRM_GPUVA_OP_UNMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va), fence, fence2);
- break;
- case DRM_GPUVA_OP_PREFETCH:
- bind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va), fence, fence2);
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-}
-
-static const struct xe_migrate_pt_update_ops migrate_ops = {
- .populate = xe_vm_populate_pgtable,
- .clear = xe_migrate_clear_pgtable_callback,
->>>>>>>
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-/**
- * __xe_pt_unbind_vma() - Disconnect and free a page-table tree for the vma
- * address range.
- * @tile: The tile to unbind for.
- * @vma: The vma to unbind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before disconnecting the tree to be destroyed.
- * @num_syncs: Number of @sync entries.
- *
- * This function builds a the xe_vm_pgtable_update entries abstracting the
- * operations needed to detach the page-table tree to be destroyed from the
- * man vm tree.
- * It then takes the relevant locks and submits the operations for
- * pipelined detachment of the gpu page-table from the vm main tree,
- * (which can be done either by the cpu and the GPU), Finally it frees the
- * detached page-table tree.
- *
- * Return: A valid dma-fence representing the pipelined detachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_unbind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs)
-{
-<<<<<<<
- struct xe_vm *vm = vops->vm;
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL, *mfence = NULL;
- struct dma_fence_chain *chain_fence = NULL;
- struct xe_range_fence *rfence;
- struct xe_vma_op *op;
- int err = 0, i;
- struct xe_migrate_pt_update update = {
- .ops = pt_update_ops->needs_userptr_lock ?
- &userptr_migrate_ops :
- &migrate_ops,
- .vops = vops,
- .tile_id = tile->id,
-=======
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update unbind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_unbind_ops :
- &unbind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
->>>>>>>
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence = NULL;
- struct invalidation_fence *ifence;
- struct xe_range_fence *rfence;
- int err;
-
- LLIST_HEAD(deferred);
-
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- num_entries = xe_pt_stage_unbind(tile, vma, entries);
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &unbind_pt_update, entries,
- num_entries);
-
-<<<<<<<
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- return ERR_PTR(err);
-
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
-=======
- if (pt_update_ops->needs_invalidation) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence) {
- err = -ENOMEM;
- goto kill_vm_tile1;
- }
- if (tile->media_gt) {
- mfence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!mfence) {
- err = -ENOMEM;
- goto free_ifence;
- }
- chain_fence = dma_fence_chain_alloc();
- if (!chain_fence) {
- err = -ENOMEM;
- goto free_ifence;
- }
- }
- }
->>>>>>>
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- /*
- * Even if we were already evicted and unbind to destroy, we need to
- * clear again here. The eviction may have updated pagetables at a
- * lower level, because it needs to be more conservative.
- */
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, NULL, q ? q :
- vm->q[tile->id],
- entries, num_entries,
- syncs, num_syncs,
- &unbind_pt_update.base);
- if (!IS_ERR(fence)) {
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- unbind_pt_update.base.start,
- unbind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
-<<<<<<<
- /* TLB invalidation must be done before signaling unbind */
- err = invalidation_fence_init(tile->primary_gt, ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- DMA_RESV_USAGE_BOOKKEEP);
-
- /* This fence will be installed by caller when doing eviction */
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_unbind(vma, entries, num_entries,
- unbind_pt_update.locked ? &deferred : NULL);
- vma->tile_present &= ~BIT(tile->id);
- } else {
- kfree(rfence);
- kfree(ifence);
- }
-
- if (!vma->tile_present)
- list_del_init(&vma->combined_links.rebind);
-
- if (unbind_pt_update.locked) {
- xe_tile_assert(tile, xe_vma_is_userptr(vma));
-=======
- xe_pt_commit(pt_op->vma, pt_op->entries,
- pt_op->num_entries, &pt_update_ops->deferred);
- pt_op->vma = NULL; /* skip in xe_pt_update_ops_abort */
- }
-
- if (xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- pt_update_ops->start,
- pt_update_ops->last, fence))
- dma_fence_wait(fence, false);
-
- /* tlb invalidation must be done before signaling rebind */
- if (ifence) {
- if (mfence)
- dma_fence_get(fence);
- invalidation_fence_init(tile->primary_gt, ifence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- if (mfence) {
- invalidation_fence_init(tile->media_gt, mfence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- dma_fence_chain_init(chain_fence, &ifence->base.base,
- &mfence->base.base, 0);
- fence = &chain_fence->base;
- } else {
- fence = &ifence->base.base;
- }
- }
-
- if (!mfence) {
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op, fence, NULL);
- } else {
- dma_resv_add_fence(xe_vm_resv(vm), &ifence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- dma_resv_add_fence(xe_vm_resv(vm), &mfence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op,
- &ifence->base.base, &mfence->base.base);
- }
->>>>>>>
-
- if (!vma->tile_present) {
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
-
- return fence;
-<<<<<<<
-=======
-
-free_rfence:
- kfree(rfence);
-free_ifence:
- dma_fence_chain_free(chain_fence);
- kfree(mfence);
- kfree(ifence);
-kill_vm_tile1:
- if (err != -EAGAIN && tile->id)
- xe_vm_kill(vops->vm, false);
-
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_update_ops_fini() - Finish PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operations
- *
- * Finish PT update operations by committing to destroy page table memory
- */
-void xe_pt_update_ops_fini(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = 0; i < pt_update_ops->current_op; ++i) {
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[i];
-
- xe_pt_free_bind(pt_op->entries, pt_op->num_entries);
- }
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
-}
-
-/**
- * xe_pt_update_ops_abort() - Abort PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Abort PT update operations by unwinding internal PT state
- */
-void xe_pt_update_ops_abort(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = pt_update_ops->num_ops - 1; i >= 0; --i) {
- struct xe_vm_pgtable_update_op *pt_op =
- &pt_update_ops->ops[i];
-
- if (!pt_op->vma || i >= pt_update_ops->current_op)
- continue;
-
- if (pt_op->bind)
- xe_pt_abort_bind(pt_op->vma, pt_op->entries,
- pt_op->num_entries,
- pt_op->rebind);
- else
- xe_pt_abort_unbind(pt_op->vma, pt_op->entries,
- pt_op->num_entries);
- }
-
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
->>>>>>>
-}
diff --git a/rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.5 b/rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.5
deleted file mode 100644
index 232d29e9a561..000000000000
--- a/rr-cache/41549dd6cc337627199acbe6749c5685c7e927d3/preimage.5
+++ /dev/null
@@ -1,2234 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2022 Intel Corporation
- */
-
-#include <linux/dma-fence-chain.h>
-
-#include "xe_pt.h"
-
-#include "regs/xe_gtt_defs.h"
-#include "xe_bo.h"
-#include "xe_device.h"
-#include "xe_drm_client.h"
-#include "xe_gt.h"
-#include "xe_gt_tlb_invalidation.h"
-#include "xe_migrate.h"
-#include "xe_pt_types.h"
-#include "xe_pt_walk.h"
-#include "xe_res_cursor.h"
-#include "xe_trace.h"
-#include "xe_ttm_stolen_mgr.h"
-#include "xe_vm.h"
-
-struct xe_pt_dir {
- struct xe_pt pt;
- /** @children: Array of page-table child nodes */
- struct xe_ptw *children[XE_PDES];
-};
-
-#if IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM)
-#define xe_pt_set_addr(__xe_pt, __addr) ((__xe_pt)->addr = (__addr))
-#define xe_pt_addr(__xe_pt) ((__xe_pt)->addr)
-#else
-#define xe_pt_set_addr(__xe_pt, __addr)
-#define xe_pt_addr(__xe_pt) 0ull
-#endif
-
-static const u64 xe_normal_pt_shifts[] = {12, 21, 30, 39, 48};
-static const u64 xe_compact_pt_shifts[] = {16, 21, 30, 39, 48};
-
-#define XE_PT_HIGHEST_LEVEL (ARRAY_SIZE(xe_normal_pt_shifts) - 1)
-
-static struct xe_pt_dir *as_xe_pt_dir(struct xe_pt *pt)
-{
- return container_of(pt, struct xe_pt_dir, pt);
-}
-
-static struct xe_pt *xe_pt_entry(struct xe_pt_dir *pt_dir, unsigned int index)
-{
- return container_of(pt_dir->children[index], struct xe_pt, base);
-}
-
-static u64 __xe_pt_empty_pte(struct xe_tile *tile, struct xe_vm *vm,
- unsigned int level)
-{
- struct xe_device *xe = tile_to_xe(tile);
- u16 pat_index = xe->pat.idx[XE_CACHE_WB];
- u8 id = tile->id;
-
- if (!xe_vm_has_scratch(vm))
- return 0;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return vm->pt_ops->pde_encode_bo(vm->scratch_pt[id][level - 1]->bo,
- 0, pat_index);
-
- return vm->pt_ops->pte_encode_addr(xe, 0, pat_index, level, IS_DGFX(xe), 0) |
- XE_PTE_NULL;
-}
-
-static void xe_pt_free(struct xe_pt *pt)
-{
- if (pt->level)
- kfree(as_xe_pt_dir(pt));
- else
- kfree(pt);
-}
-
-/**
- * xe_pt_create() - Create a page-table.
- * @vm: The vm to create for.
- * @tile: The tile to create for.
- * @level: The page-table level.
- *
- * Allocate and initialize a single struct xe_pt metadata structure. Also
- * create the corresponding page-table bo, but don't initialize it. If the
- * level is grater than zero, then it's assumed to be a directory page-
- * table and the directory structure is also allocated and initialized to
- * NULL pointers.
- *
- * Return: A valid struct xe_pt pointer on success, Pointer error code on
- * error.
- */
-struct xe_pt *xe_pt_create(struct xe_vm *vm, struct xe_tile *tile,
- unsigned int level)
-{
- struct xe_pt *pt;
- struct xe_bo *bo;
- int err;
-
- if (level) {
- struct xe_pt_dir *dir = kzalloc(sizeof(*dir), GFP_KERNEL);
-
- pt = (dir) ? &dir->pt : NULL;
- } else {
- pt = kzalloc(sizeof(*pt), GFP_KERNEL);
- }
- if (!pt)
- return ERR_PTR(-ENOMEM);
-
- pt->level = level;
- bo = xe_bo_create_pin_map(vm->xe, tile, vm, SZ_4K,
- ttm_bo_type_kernel,
- XE_BO_FLAG_VRAM_IF_DGFX(tile) |
- XE_BO_FLAG_IGNORE_MIN_PAGE_SIZE |
- XE_BO_FLAG_PINNED |
- XE_BO_FLAG_NO_RESV_EVICT |
- XE_BO_FLAG_PAGETABLE);
- if (IS_ERR(bo)) {
- err = PTR_ERR(bo);
- goto err_kfree;
- }
- pt->bo = bo;
- pt->base.children = level ? as_xe_pt_dir(pt)->children : NULL;
-
- if (vm->xef)
- xe_drm_client_add_bo(vm->xef->client, pt->bo);
- xe_tile_assert(tile, level <= XE_VM_MAX_LEVEL);
-
- return pt;
-
-err_kfree:
- xe_pt_free(pt);
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_populate_empty() - Populate a page-table bo with scratch- or zero
- * entries.
- * @tile: The tile the scratch pagetable of which to use.
- * @vm: The vm we populate for.
- * @pt: The pagetable the bo of which to initialize.
- *
- * Populate the page-table bo of @pt with entries pointing into the tile's
- * scratch page-table tree if any. Otherwise populate with zeros.
- */
-void xe_pt_populate_empty(struct xe_tile *tile, struct xe_vm *vm,
- struct xe_pt *pt)
-{
- struct iosys_map *map = &pt->bo->vmap;
- u64 empty;
- int i;
-
- if (!xe_vm_has_scratch(vm)) {
- /*
- * FIXME: Some memory is allocated already allocated to zero?
- * Find out which memory that is and avoid this memset...
- */
- xe_map_memset(vm->xe, map, 0, 0, SZ_4K);
- } else {
- empty = __xe_pt_empty_pte(tile, vm, pt->level);
- for (i = 0; i < XE_PDES; i++)
- xe_pt_write(vm->xe, map, i, empty);
- }
-}
-
-/**
- * xe_pt_shift() - Return the ilog2 value of the size of the address range of
- * a page-table at a certain level.
- * @level: The level.
- *
- * Return: The ilog2 value of the size of the address range of a page-table
- * at level @level.
- */
-unsigned int xe_pt_shift(unsigned int level)
-{
- return XE_PTE_SHIFT + XE_PDE_SHIFT * level;
-}
-
-/**
- * xe_pt_destroy() - Destroy a page-table tree.
- * @pt: The root of the page-table tree to destroy.
- * @flags: vm flags. Currently unused.
- * @deferred: List head of lockless list for deferred putting. NULL for
- * immediate putting.
- *
- * Puts the page-table bo, recursively calls xe_pt_destroy on all children
- * and finally frees @pt. TODO: Can we remove the @flags argument?
- */
-void xe_pt_destroy(struct xe_pt *pt, u32 flags, struct llist_head *deferred)
-{
- int i;
-
- if (!pt)
- return;
-
- XE_WARN_ON(!list_empty(&pt->bo->ttm.base.gpuva.list));
- xe_bo_unpin(pt->bo);
- xe_bo_put_deferred(pt->bo, deferred);
-
- if (pt->level > 0 && pt->num_live) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
-
- for (i = 0; i < XE_PDES; i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i), flags,
- deferred);
- }
- }
- xe_pt_free(pt);
-}
-
-/**
- * DOC: Pagetable building
- *
- * Below we use the term "page-table" for both page-directories, containing
- * pointers to lower level page-directories or page-tables, and level 0
- * page-tables that contain only page-table-entries pointing to memory pages.
- *
- * When inserting an address range in an already existing page-table tree
- * there will typically be a set of page-tables that are shared with other
- * address ranges, and a set that are private to this address range.
- * The set of shared page-tables can be at most two per level,
- * and those can't be updated immediately because the entries of those
- * page-tables may still be in use by the gpu for other mappings. Therefore
- * when inserting entries into those, we instead stage those insertions by
- * adding insertion data into struct xe_vm_pgtable_update structures. This
- * data, (subtrees for the cpu and page-table-entries for the gpu) is then
- * added in a separate commit step. CPU-data is committed while still under the
- * vm lock, the object lock and for userptr, the notifier lock in read mode.
- * The GPU async data is committed either by the GPU or CPU after fulfilling
- * relevant dependencies.
- * For non-shared page-tables (and, in fact, for shared ones that aren't
- * existing at the time of staging), we add the data in-place without the
- * special update structures. This private part of the page-table tree will
- * remain disconnected from the vm page-table tree until data is committed to
- * the shared page tables of the vm tree in the commit phase.
- */
-
-struct xe_pt_update {
- /** @update: The update structure we're building for this parent. */
- struct xe_vm_pgtable_update *update;
- /** @parent: The parent. Used to detect a parent change. */
- struct xe_pt *parent;
- /** @preexisting: Whether the parent was pre-existing or allocated */
- bool preexisting;
-};
-
-struct xe_pt_stage_bind_walk {
- /** base: The base class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @vm: The vm we're building for. */
- struct xe_vm *vm;
- /** @tile: The tile we're building for. */
- struct xe_tile *tile;
- /** @default_pte: PTE flag only template. No address is associated */
- u64 default_pte;
- /** @dma_offset: DMA offset to add to the PTE. */
- u64 dma_offset;
- /**
- * @needs_64k: This address range enforces 64K alignment and
- * granularity.
- */
- bool needs_64K;
- /**
- * @vma: VMA being mapped
- */
- struct xe_vma *vma;
-
- /* Also input, but is updated during the walk*/
- /** @curs: The DMA address cursor. */
- struct xe_res_cursor *curs;
- /** @va_curs_start: The Virtual address coresponding to @curs->start */
- u64 va_curs_start;
-
- /* Output */
- struct xe_walk_update {
- /** @wupd.entries: Caller provided storage. */
- struct xe_vm_pgtable_update *entries;
- /** @wupd.num_used_entries: Number of update @entries used. */
- unsigned int num_used_entries;
- /** @wupd.updates: Tracks the update entry at a given level */
- struct xe_pt_update updates[XE_VM_MAX_LEVEL + 1];
- } wupd;
-
- /* Walk state */
- /**
- * @l0_end_addr: The end address of the current l0 leaf. Used for
- * 64K granularity detection.
- */
- u64 l0_end_addr;
- /** @addr_64K: The start address of the current 64K chunk. */
- u64 addr_64K;
- /** @found_64: Whether @add_64K actually points to a 64K chunk. */
- bool found_64K;
-};
-
-static int
-xe_pt_new_shared(struct xe_walk_update *wupd, struct xe_pt *parent,
- pgoff_t offset, bool alloc_entries)
-{
- struct xe_pt_update *upd = &wupd->updates[parent->level];
- struct xe_vm_pgtable_update *entry;
-
- /*
- * For *each level*, we could only have one active
- * struct xt_pt_update at any one time. Once we move on to a
- * new parent and page-directory, the old one is complete, and
- * updates are either already stored in the build tree or in
- * @wupd->entries
- */
- if (likely(upd->parent == parent))
- return 0;
-
- upd->parent = parent;
- upd->preexisting = true;
-
- if (wupd->num_used_entries == XE_VM_MAX_LEVEL * 2 + 1)
- return -EINVAL;
-
- entry = wupd->entries + wupd->num_used_entries++;
- upd->update = entry;
- entry->ofs = offset;
- entry->pt_bo = parent->bo;
- entry->pt = parent;
- entry->flags = 0;
- entry->qwords = 0;
-
- if (alloc_entries) {
- entry->pt_entries = kmalloc_array(XE_PDES,
- sizeof(*entry->pt_entries),
- GFP_KERNEL);
- if (!entry->pt_entries)
- return -ENOMEM;
- }
-
- return 0;
-}
-
-/*
- * NOTE: This is a very frequently called function so we allow ourselves
- * to annotate (using branch prediction hints) the fastpath of updating a
- * non-pre-existing pagetable with leaf ptes.
- */
-static int
-xe_pt_insert_entry(struct xe_pt_stage_bind_walk *xe_walk, struct xe_pt *parent,
- pgoff_t offset, struct xe_pt *xe_child, u64 pte)
-{
- struct xe_pt_update *upd = &xe_walk->wupd.updates[parent->level];
- struct xe_pt_update *child_upd = xe_child ?
- &xe_walk->wupd.updates[xe_child->level] : NULL;
- int ret;
-
- ret = xe_pt_new_shared(&xe_walk->wupd, parent, offset, true);
- if (unlikely(ret))
- return ret;
-
- /*
- * Register this new pagetable so that it won't be recognized as
- * a shared pagetable by a subsequent insertion.
- */
- if (unlikely(child_upd)) {
- child_upd->update = NULL;
- child_upd->parent = xe_child;
- child_upd->preexisting = false;
- }
-
- if (likely(!upd->preexisting)) {
- /* Continue building a non-connected subtree. */
- struct iosys_map *map = &parent->bo->vmap;
-
- if (unlikely(xe_child))
- parent->base.children[offset] = &xe_child->base;
-
- xe_pt_write(xe_walk->vm->xe, map, offset, pte);
- parent->num_live++;
- } else {
- /* Shared pt. Stage update. */
- unsigned int idx;
- struct xe_vm_pgtable_update *entry = upd->update;
-
- idx = offset - entry->ofs;
- entry->pt_entries[idx].pt = xe_child;
- entry->pt_entries[idx].pte = pte;
- entry->qwords++;
- }
-
- return 0;
-}
-
-static bool xe_pt_hugepte_possible(u64 addr, u64 next, unsigned int level,
- struct xe_pt_stage_bind_walk *xe_walk)
-{
- u64 size, dma;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return false;
-
- /* Does the virtual range requested cover a huge pte? */
- if (!xe_pt_covers(addr, next, level, &xe_walk->base))
- return false;
-
- /* Does the DMA segment cover the whole pte? */
- if (next - xe_walk->va_curs_start > xe_walk->curs->size)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- /* Is the DMA address huge PTE size aligned? */
- size = next - addr;
- dma = addr - xe_walk->va_curs_start + xe_res_dma(xe_walk->curs);
-
- return IS_ALIGNED(dma, size);
-}
-
-/*
- * Scan the requested mapping to check whether it can be done entirely
- * with 64K PTEs.
- */
-static bool
-xe_pt_scan_64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- struct xe_res_cursor curs = *xe_walk->curs;
-
- if (!IS_ALIGNED(addr, SZ_64K))
- return false;
-
- if (next > xe_walk->l0_end_addr)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- xe_res_next(&curs, addr - xe_walk->va_curs_start);
- for (; addr < next; addr += SZ_64K) {
- if (!IS_ALIGNED(xe_res_dma(&curs), SZ_64K) || curs.size < SZ_64K)
- return false;
-
- xe_res_next(&curs, SZ_64K);
- }
-
- return addr == next;
-}
-
-/*
- * For non-compact "normal" 4K level-0 pagetables, we want to try to group
- * addresses together in 64K-contigous regions to add a 64K TLB hint for the
- * device to the PTE.
- * This function determines whether the address is part of such a
- * segment. For VRAM in normal pagetables, this is strictly necessary on
- * some devices.
- */
-static bool
-xe_pt_is_pte_ps64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- /* Address is within an already found 64k region */
- if (xe_walk->found_64K && addr - xe_walk->addr_64K < SZ_64K)
- return true;
-
- xe_walk->found_64K = xe_pt_scan_64K(addr, addr + SZ_64K, xe_walk);
- xe_walk->addr_64K = addr;
-
- return xe_walk->found_64K;
-}
-
-static int
-xe_pt_stage_bind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_bind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- u16 pat_index = xe_walk->vma->pat_index;
- struct xe_pt *xe_parent = container_of(parent, typeof(*xe_parent), base);
- struct xe_vm *vm = xe_walk->vm;
- struct xe_pt *xe_child;
- bool covers;
- int ret = 0;
- u64 pte;
-
- /* Is this a leaf entry ?*/
- if (level == 0 || xe_pt_hugepte_possible(addr, next, level, xe_walk)) {
- struct xe_res_cursor *curs = xe_walk->curs;
- bool is_null = xe_vma_is_null(xe_walk->vma);
-
- XE_WARN_ON(xe_walk->va_curs_start != addr);
-
- pte = vm->pt_ops->pte_encode_vma(is_null ? 0 :
- xe_res_dma(curs) + xe_walk->dma_offset,
- xe_walk->vma, pat_index, level);
- pte |= xe_walk->default_pte;
-
- /*
- * Set the XE_PTE_PS64 hint if possible, otherwise if
- * this device *requires* 64K PTE size for VRAM, fail.
- */
- if (level == 0 && !xe_parent->is_compact) {
- if (xe_pt_is_pte_ps64K(addr, next, xe_walk)) {
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_64K;
- pte |= XE_PTE_PS64;
- } else if (XE_WARN_ON(xe_walk->needs_64K)) {
- return -EINVAL;
- }
- }
-
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, NULL, pte);
- if (unlikely(ret))
- return ret;
-
- if (!is_null)
- xe_res_next(curs, next - addr);
- xe_walk->va_curs_start = next;
- xe_walk->vma->gpuva.flags |= (XE_VMA_PTE_4K << level);
- *action = ACTION_CONTINUE;
-
- return ret;
- }
-
- /*
- * Descending to lower level. Determine if we need to allocate a
- * new page table or -directory, which we do if there is no
- * previous one or there is one we can completely replace.
- */
- if (level == 1) {
- walk->shifts = xe_normal_pt_shifts;
- xe_walk->l0_end_addr = next;
- }
-
- covers = xe_pt_covers(addr, next, level, &xe_walk->base);
- if (covers || !*child) {
- u64 flags = 0;
-
- xe_child = xe_pt_create(xe_walk->vm, xe_walk->tile, level - 1);
- if (IS_ERR(xe_child))
- return PTR_ERR(xe_child);
-
- xe_pt_set_addr(xe_child,
- round_down(addr, 1ull << walk->shifts[level]));
-
- if (!covers)
- xe_pt_populate_empty(xe_walk->tile, xe_walk->vm, xe_child);
-
- *child = &xe_child->base;
-
- /*
- * Prefer the compact pagetable layout for L0 if possible. Only
- * possible if VMA covers entire 2MB region as compact 64k and
- * 4k pages cannot be mixed within a 2MB region.
- * TODO: Suballocate the pt bo to avoid wasting a lot of
- * memory.
- */
- if (GRAPHICS_VERx100(tile_to_xe(xe_walk->tile)) >= 1250 && level == 1 &&
- covers && xe_pt_scan_64K(addr, next, xe_walk)) {
- walk->shifts = xe_compact_pt_shifts;
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_COMPACT;
- flags |= XE_PDE_64K;
- xe_child->is_compact = true;
- }
-
- pte = vm->pt_ops->pde_encode_bo(xe_child->bo, 0, pat_index) | flags;
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, xe_child,
- pte);
- }
-
- *action = ACTION_SUBTREE;
- return ret;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_bind_ops = {
- .pt_entry = xe_pt_stage_bind_entry,
-};
-
-/**
- * xe_pt_stage_bind() - Build a disconnected page-table tree for a given address
- * range.
- * @tile: The tile we're building for.
- * @vma: The vma indicating the address range.
- * @entries: Storage for the update entries used for connecting the tree to
- * the main tree at commit time.
- * @num_entries: On output contains the number of @entries used.
- *
- * This function builds a disconnected page-table tree for a given address
- * range. The tree is connected to the main vm tree for the gpu using
- * xe_migrate_update_pgtables() and for the cpu using xe_pt_commit_bind().
- * The function builds xe_vm_pgtable_update structures for already existing
- * shared page-tables, and non-existing shared and non-shared page-tables
- * are built and populated directly.
- *
- * Return 0 on success, negative error code on error.
- */
-static int
-xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- struct xe_device *xe = tile_to_xe(tile);
- struct xe_bo *bo = xe_vma_bo(vma);
- bool is_devmem = !xe_vma_is_userptr(vma) && bo &&
- (xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo));
- struct xe_res_cursor curs;
- struct xe_pt_stage_bind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_bind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .vm = xe_vma_vm(vma),
- .tile = tile,
- .curs = &curs,
- .va_curs_start = xe_vma_start(vma),
- .vma = vma,
- .wupd.entries = entries,
- .needs_64K = (xe_vma_vm(vma)->flags & XE_VM_FLAG_64K) && is_devmem,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- int ret;
-
- /**
- * Default atomic expectations for different allocation scenarios are as follows:
- *
- * 1. Traditional API: When the VM is not in LR mode:
- * - Device atomics are expected to function with all allocations.
- *
- * 2. Compute/SVM API: When the VM is in LR mode:
- * - Device atomics are the default behavior when the bo is placed in a single region.
- * - In all other cases device atomics will be disabled with AE=0 until an application
- * request differently using a ioctl like madvise.
- */
- if (vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT) {
- if (xe_vm_in_lr_mode(xe_vma_vm(vma))) {
- if (bo && xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- /**
- * If a SMEM+LMEM allocation is backed by SMEM, a device
- * atomics will cause a gpu page fault and which then
- * gets migrated to LMEM, bind such allocations with
- * device atomics enabled.
- */
- else if (is_devmem && !xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- } else {
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- }
-
- /**
- * Unset AE if the platform(PVC) doesn't support it on an
- * allocation
- */
- if (!xe->info.has_device_atomics_on_smem && !is_devmem)
- xe_walk.default_pte &= ~XE_USM_PPGTT_PTE_AE;
- }
-
- if (is_devmem) {
- xe_walk.default_pte |= XE_PPGTT_PTE_DM;
- xe_walk.dma_offset = vram_region_gpu_offset(bo->ttm.resource);
- }
-
- if (!xe_vma_has_no_bo(vma) && xe_bo_is_stolen(bo))
- xe_walk.dma_offset = xe_ttm_stolen_gpu_offset(xe_bo_device(bo));
-
- xe_bo_assert_held(bo);
-
- if (!xe_vma_is_null(vma)) {
- if (xe_vma_is_userptr(vma))
- xe_res_first_sg(to_userptr_vma(vma)->userptr.sg, 0,
- xe_vma_size(vma), &curs);
- else if (xe_bo_is_vram(bo) || xe_bo_is_stolen(bo))
- xe_res_first(bo->ttm.resource, xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- else
- xe_res_first_sg(xe_bo_sg(bo), xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- } else {
- curs.size = xe_vma_size(vma);
- }
-
- ret = xe_pt_walk_range(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- *num_entries = xe_walk.wupd.num_used_entries;
- return ret;
-}
-
-/**
- * xe_pt_nonshared_offsets() - Determine the non-shared entry offsets of a
- * shared pagetable.
- * @addr: The start address within the non-shared pagetable.
- * @end: The end address within the non-shared pagetable.
- * @level: The level of the non-shared pagetable.
- * @walk: Walk info. The function adjusts the walk action.
- * @action: next action to perform (see enum page_walk_action)
- * @offset: Ignored on input, First non-shared entry on output.
- * @end_offset: Ignored on input, Last non-shared entry + 1 on output.
- *
- * A non-shared page-table has some entries that belong to the address range
- * and others that don't. This function determines the entries that belong
- * fully to the address range. Depending on level, some entries may
- * partially belong to the address range (that can't happen at level 0).
- * The function detects that and adjust those offsets to not include those
- * partial entries. Iff it does detect partial entries, we know that there must
- * be shared page tables also at lower levels, so it adjusts the walk action
- * accordingly.
- *
- * Return: true if there were non-shared entries, false otherwise.
- */
-static bool xe_pt_nonshared_offsets(u64 addr, u64 end, unsigned int level,
- struct xe_pt_walk *walk,
- enum page_walk_action *action,
- pgoff_t *offset, pgoff_t *end_offset)
-{
- u64 size = 1ull << walk->shifts[level];
-
- *offset = xe_pt_offset(addr, level, walk);
- *end_offset = xe_pt_num_entries(addr, end, level, walk) + *offset;
-
- if (!level)
- return true;
-
- /*
- * If addr or next are not size aligned, there are shared pts at lower
- * level, so in that case traverse down the subtree
- */
- *action = ACTION_CONTINUE;
- if (!IS_ALIGNED(addr, size)) {
- *action = ACTION_SUBTREE;
- (*offset)++;
- }
-
- if (!IS_ALIGNED(end, size)) {
- *action = ACTION_SUBTREE;
- (*end_offset)--;
- }
-
- return *end_offset > *offset;
-}
-
-struct xe_pt_zap_ptes_walk {
- /** @base: The walk base-class */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're building for */
- struct xe_tile *tile;
-
- /* Output */
- /** @needs_invalidate: Whether we need to invalidate TLB*/
- bool needs_invalidate;
-};
-
-static int xe_pt_zap_ptes_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_zap_ptes_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- /*
- * Note that we're called from an entry callback, and we're dealing
- * with the child of that entry rather than the parent, so need to
- * adjust level down.
- */
- if (xe_pt_nonshared_offsets(addr, next, --level, walk, action, &offset,
- &end_offset)) {
- xe_map_memset(tile_to_xe(xe_walk->tile), &xe_child->bo->vmap,
- offset * sizeof(u64), 0,
- (end_offset - offset) * sizeof(u64));
- xe_walk->needs_invalidate = true;
- }
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_zap_ptes_ops = {
- .pt_entry = xe_pt_zap_ptes_entry,
-};
-
-/**
- * xe_pt_zap_ptes() - Zap (zero) gpu ptes of an address range
- * @tile: The tile we're zapping for.
- * @vma: GPU VMA detailing address range.
- *
- * Eviction and Userptr invalidation needs to be able to zap the
- * gpu ptes of a given address range in pagefaulting mode.
- * In order to be able to do that, that function needs access to the shared
- * page-table entrieaso it can either clear the leaf PTEs or
- * clear the pointers to lower-level page-tables. The caller is required
- * to hold the necessary locks to ensure neither the page-table connectivity
- * nor the page-table entries of the range is updated from under us.
- *
- * Return: Whether ptes were actually updated and a TLB invalidation is
- * required.
- */
-bool xe_pt_zap_ptes(struct xe_tile *tile, struct xe_vma *vma)
-{
- struct xe_pt_zap_ptes_walk xe_walk = {
- .base = {
- .ops = &xe_pt_zap_ptes_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- u8 pt_mask = (vma->tile_present & ~vma->tile_invalidated);
-
- if (!(pt_mask & BIT(tile->id)))
- return false;
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.needs_invalidate;
-}
-
-static void
-xe_vm_populate_pgtable(struct xe_migrate_pt_update *pt_update, struct xe_tile *tile,
- struct iosys_map *map, void *data,
- u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_pt_entry *ptes = update->pt_entries;
- u64 *ptr = data;
- u32 i;
-
- for (i = 0; i < num_qwords; i++) {
- if (map)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, ptes[i].pte);
- else
- ptr[i] = ptes[i].pte;
- }
-}
-
-static void xe_pt_abort_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- u32 i, j;
-
- for (i = 0; i < num_entries; i++) {
- if (!entries[i].pt_entries)
- continue;
-
- for (j = 0; j < entries[i].qwords; j++)
- xe_pt_destroy(entries[i].pt_entries[j].pt, xe_vma_vm(vma)->flags, NULL);
- kfree(entries[i].pt_entries);
- }
-}
-
-static void xe_pt_commit_locks_assert(struct xe_vma *vma)
-{
- struct xe_vm *vm = xe_vma_vm(vma);
-
- lockdep_assert_held(&vm->lock);
-
- if (xe_vma_is_userptr(vma))
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- else if (!xe_vma_is_null(vma))
- dma_resv_assert_held(xe_vma_bo(vma)->ttm.base.resv);
-
- xe_vm_assert_held(vm);
-}
-
-static void xe_pt_commit_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries, bool rebind,
- struct llist_head *deferred)
-{
- u32 i, j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (i = 0; i < num_entries; i++) {
- struct xe_pt *pt = entries[i].pt;
- struct xe_pt_dir *pt_dir;
-
- if (!rebind)
- pt->num_live += entries[i].qwords;
-
- if (!pt->level) {
- kfree(entries[i].pt_entries);
- continue;
- }
-
- pt_dir = as_xe_pt_dir(pt);
- for (j = 0; j < entries[i].qwords; j++) {
- u32 j_ = j + entries[i].ofs;
- struct xe_pt *newpte = entries[i].pt_entries[j].pt;
-
- if (xe_pt_entry(pt_dir, j_))
- xe_pt_destroy(xe_pt_entry(pt_dir, j_),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[j_] = &newpte->base;
- }
- kfree(entries[i].pt_entries);
- }
-}
-
-static int
-xe_pt_prepare_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- int err;
-
- *num_entries = 0;
- err = xe_pt_stage_bind(tile, vma, entries, num_entries);
- if (!err)
- xe_tile_assert(tile, *num_entries);
- else /* abort! */
- xe_pt_abort_bind(vma, entries, *num_entries);
-
- return err;
-}
-
-static void xe_vm_dbg_print_entries(struct xe_device *xe,
- const struct xe_vm_pgtable_update *entries,
- unsigned int num_entries)
-#if (IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM))
-{
- unsigned int i;
-
- vm_dbg(&xe->drm, "%u entries to update\n", num_entries);
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
- struct xe_pt *xe_pt = entry->pt;
- u64 page_size = 1ull << xe_pt_shift(xe_pt->level);
- u64 end;
- u64 start;
-
- xe_assert(xe, !entry->pt->is_compact);
- start = entry->ofs * page_size;
- end = start + page_size * entry->qwords;
- vm_dbg(&xe->drm,
- "\t%u: Update level %u at (%u + %u) [%llx...%llx) f:%x\n",
- i, xe_pt->level, entry->ofs, entry->qwords,
- xe_pt_addr(xe_pt) + start, xe_pt_addr(xe_pt) + end, 0);
- }
-}
-#else
-{}
-#endif
-
-#ifdef CONFIG_DRM_XE_USERPTR_INVAL_INJECT
-
-static int xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- u32 divisor = uvma->userptr.divisor ? uvma->userptr.divisor : 2;
- static u32 count;
-
- if (count++ % divisor == divisor - 1) {
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
-
- uvma->userptr.divisor = divisor << 1;
- spin_lock(&vm->userptr.invalidated_lock);
- list_move_tail(&uvma->userptr.invalidate_link,
- &vm->userptr.invalidated);
- spin_unlock(&vm->userptr.invalidated_lock);
- return true;
- }
-
- return false;
-}
-
-#else
-
-static bool xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- return false;
-}
-
-#endif
-
-/**
- * struct xe_pt_migrate_pt_update - Callback argument for pre-commit callbacks
- * @base: Base we derive from.
- * @bind: Whether this is a bind or an unbind operation. A bind operation
- * makes the pre-commit callback error with -EAGAIN if it detects a
- * pending invalidation.
- * @locked: Whether the pre-commit callback locked the userptr notifier lock
- * and it needs unlocking.
- */
-struct xe_pt_migrate_pt_update {
- struct xe_migrate_pt_update base;
- bool bind;
- bool locked;
-};
-
-/*
- * This function adds the needed dependencies to a page-table update job
- * to make sure racing jobs for separate bind engines don't race writing
- * to the same page-table range, wreaking havoc. Initially use a single
- * fence for the entire VM. An optimization would use smaller granularity.
- */
-static int xe_pt_vm_dependencies(struct xe_sched_job *job,
- struct xe_range_fence_tree *rftree,
- u64 start, u64 last)
-{
- struct xe_range_fence *rtfence;
- struct dma_fence *fence;
- int err;
-
- rtfence = xe_range_fence_tree_first(rftree, start, last);
- while (rtfence) {
- fence = rtfence->fence;
-
- if (!dma_fence_is_signaled(fence)) {
- /*
- * Is this a CPU update? GPU is busy updating, so return
- * an error
- */
- if (!job)
- return -ETIME;
-
- dma_fence_get(fence);
- err = drm_sched_job_add_dependency(&job->drm, fence);
- if (err)
- return err;
- }
-
- rtfence = xe_range_fence_tree_next(rtfence, start, last);
- }
-
- return 0;
-}
-
-static int xe_pt_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_range_fence_tree *rftree =
- &xe_vma_vm(pt_update->vma)->rftree[pt_update->tile_id];
-
- return xe_pt_vm_dependencies(pt_update->job, rftree,
- pt_update->start, pt_update->last);
-}
-
-static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_pt_migrate_pt_update *userptr_update =
- container_of(pt_update, typeof(*userptr_update), base);
- struct xe_userptr_vma *uvma = to_userptr_vma(pt_update->vma);
- unsigned long notifier_seq = uvma->userptr.notifier_seq;
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
- int err = xe_pt_vm_dependencies(pt_update->job,
- &vm->rftree[pt_update->tile_id],
- pt_update->start,
- pt_update->last);
-
- if (err)
- return err;
-
- userptr_update->locked = false;
-
- /*
- * Wait until nobody is running the invalidation notifier, and
- * since we're exiting the loop holding the notifier lock,
- * nobody can proceed invalidating either.
- *
- * Note that we don't update the vma->userptr.notifier_seq since
- * we don't update the userptr pages.
- */
- do {
- down_read(&vm->userptr.notifier_lock);
- if (!mmu_interval_read_retry(&uvma->userptr.notifier,
- notifier_seq))
- break;
-
- up_read(&vm->userptr.notifier_lock);
-
- if (userptr_update->bind)
- return -EAGAIN;
-
- notifier_seq = mmu_interval_read_begin(&uvma->userptr.notifier);
- } while (true);
-
- /* Inject errors to test_whether they are handled correctly */
- if (userptr_update->bind && xe_pt_userptr_inject_eagain(uvma)) {
- up_read(&vm->userptr.notifier_lock);
- return -EAGAIN;
- }
-
- userptr_update->locked = true;
-
- return 0;
-}
-
-static const struct xe_migrate_pt_update_ops bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-struct invalidation_fence {
- struct xe_gt_tlb_invalidation_fence base;
- struct xe_gt *gt;
- struct dma_fence *fence;
- struct dma_fence_cb cb;
- struct work_struct work;
- u64 start;
- u64 end;
- u32 asid;
-};
-
-static void invalidation_fence_cb(struct dma_fence *fence,
- struct dma_fence_cb *cb)
-{
- struct invalidation_fence *ifence =
- container_of(cb, struct invalidation_fence, cb);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_cb(xe, &ifence->base);
- if (!ifence->fence->error) {
- queue_work(system_wq, &ifence->work);
- } else {
- ifence->base.base.error = ifence->fence->error;
- dma_fence_signal(&ifence->base.base);
- dma_fence_put(&ifence->base.base);
- }
- dma_fence_put(ifence->fence);
-}
-
-static void invalidation_fence_work_func(struct work_struct *w)
-{
- struct invalidation_fence *ifence =
- container_of(w, struct invalidation_fence, work);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_work_func(xe, &ifence->base);
- xe_gt_tlb_invalidation_range(ifence->gt, &ifence->base, ifence->start,
- ifence->end, ifence->asid);
-}
-
-static int invalidation_fence_init(struct xe_gt *gt,
- struct invalidation_fence *ifence,
- struct dma_fence *fence,
- u64 start, u64 end, u32 asid)
-{
- int ret;
-
- trace_xe_gt_tlb_invalidation_fence_create(gt_to_xe(gt), &ifence->base);
-
- xe_gt_tlb_invalidation_fence_init(gt, &ifence->base, false);
-
- ifence->fence = fence;
- ifence->gt = gt;
- ifence->start = start;
- ifence->end = end;
- ifence->asid = asid;
-
- INIT_WORK(&ifence->work, invalidation_fence_work_func);
- ret = dma_fence_add_callback(fence, &ifence->cb, invalidation_fence_cb);
- if (ret == -ENOENT) {
- dma_fence_put(ifence->fence); /* Usually dropped in CB */
- invalidation_fence_work_func(&ifence->work);
- } else if (ret) {
- dma_fence_put(&ifence->base.base); /* Caller ref */
- dma_fence_put(&ifence->base.base); /* Creation ref */
- }
-
- xe_gt_assert(gt, !ret || ret == -ENOENT);
-
- return ret && ret != -ENOENT ? ret : 0;
-}
-
-static void xe_pt_calc_rfence_interval(struct xe_vma *vma,
- struct xe_pt_migrate_pt_update *update,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- int i, level = 0;
-
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- update->base.start = ALIGN_DOWN(xe_vma_start(vma),
- 0x1ull << xe_pt_shift(level));
- update->base.last = ALIGN(xe_vma_end(vma),
- 0x1ull << xe_pt_shift(level)) - 1;
-}
-
-/**
- * __xe_pt_bind_vma() - Build and connect a page-table tree for the vma
- * address range.
- * @tile: The tile to bind for.
- * @vma: The vma to bind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before binding the built tree to the live vm tree.
- * @num_syncs: Number of @sync entries.
- * @rebind: Whether we're rebinding this vma to the same address range without
- * an unbind in-between.
- *
- * This function builds a page-table tree (see xe_pt_stage_bind() for more
- * information on page-table building), and the xe_vm_pgtable_update entries
- * abstracting the operations needed to attach it to the main vm tree. It
- * then takes the relevant locks and updates the metadata side of the main
- * vm tree and submits the operations for pipelined attachment of the
- * gpu page-table to the vm main tree, (which can be done either by the
- * cpu and the GPU).
- *
- * Return: A valid dma-fence representing the pipelined attachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs,
- bool rebind)
-{
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update bind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_bind_ops : &bind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
- .bind = true,
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL;
- struct xe_range_fence *rfence;
- int err;
-
- bind_pt_update.locked = false;
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- err = xe_pt_prepare_bind(tile, vma, entries, &num_entries);
- if (err)
- goto err;
-
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- goto err;
-
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &bind_pt_update, entries,
- num_entries);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. on LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * on !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!rebind && xe_vm_has_scratch(vm) && xe_vm_in_preempt_fence_mode(vm))) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
- } else if (rebind && !xe_vm_in_lr_mode(vm)) {
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
- }
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, xe_vma_bo(vma), q,
- entries, num_entries,
- syncs, num_syncs,
- &bind_pt_update.base);
- if (!IS_ERR(fence)) {
- bool last_munmap_rebind = vma->gpuva.flags & XE_VMA_LAST_REBIND;
- LLIST_HEAD(deferred);
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- bind_pt_update.base.start,
- bind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
- /* TLB invalidation must be done before signaling rebind */
- if (ifence) {
- int err = invalidation_fence_init(tile->primary_gt,
- ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
- }
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence, rebind ||
- last_munmap_rebind ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_bind(vma, entries, num_entries, rebind,
- bind_pt_update.locked ? &deferred : NULL);
-
- /* This vma is live (again?) now */
- vma->tile_present |= BIT(tile->id);
-
- if (bind_pt_update.locked) {
- to_userptr_vma(vma)->userptr.initial_bind = true;
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
- if (!rebind && last_munmap_rebind &&
- xe_vm_in_preempt_fence_mode(vm))
- xe_vm_queue_rebind_worker(vm);
- } else {
- kfree(rfence);
- kfree(ifence);
- if (bind_pt_update.locked)
- up_read(&vm->userptr.notifier_lock);
- xe_pt_abort_bind(vma, entries, num_entries);
- }
-
- return fence;
-
-err:
- return ERR_PTR(err);
-}
-
-struct xe_pt_stage_unbind_walk {
- /** @base: The pagewalk base-class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're unbinding from. */
- struct xe_tile *tile;
-
- /**
- * @modified_start: Walk range start, modified to include any
- * shared pagetables that we're the only user of and can thus
- * treat as private.
- */
- u64 modified_start;
- /** @modified_end: Walk range start, modified like @modified_start. */
- u64 modified_end;
-
- /* Output */
- /* @wupd: Structure to track the page-table updates we're building */
- struct xe_walk_update wupd;
-};
-
-/*
- * Check whether this range is the only one populating this pagetable,
- * and in that case, update the walk range checks so that higher levels don't
- * view us as a shared pagetable.
- */
-static bool xe_pt_check_kill(u64 addr, u64 next, unsigned int level,
- const struct xe_pt *child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- unsigned int shift = walk->shifts[level];
- u64 size = 1ull << shift;
-
- if (IS_ALIGNED(addr, size) && IS_ALIGNED(next, size) &&
- ((next - addr) >> shift) == child->num_live) {
- u64 size = 1ull << walk->shifts[level + 1];
-
- *action = ACTION_CONTINUE;
-
- if (xe_walk->modified_start >= addr)
- xe_walk->modified_start = round_down(addr, size);
- if (xe_walk->modified_end <= next)
- xe_walk->modified_end = round_up(next, size);
-
- return true;
- }
-
- return false;
-}
-
-static int xe_pt_stage_unbind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- xe_pt_check_kill(addr, next, level - 1, xe_child, action, walk);
-
- return 0;
-}
-
-static int
-xe_pt_stage_unbind_post_descend(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
- u64 size = 1ull << walk->shifts[--level];
-
- if (!IS_ALIGNED(addr, size))
- addr = xe_walk->modified_start;
- if (!IS_ALIGNED(next, size))
- next = xe_walk->modified_end;
-
- /* Parent == *child is the root pt. Don't kill it. */
- if (parent != *child &&
- xe_pt_check_kill(addr, next, level, xe_child, action, walk))
- return 0;
-
- if (!xe_pt_nonshared_offsets(addr, next, level, walk, action, &offset,
- &end_offset))
- return 0;
-
- (void)xe_pt_new_shared(&xe_walk->wupd, xe_child, offset, false);
- xe_walk->wupd.updates[level].update->qwords = end_offset - offset;
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_unbind_ops = {
- .pt_entry = xe_pt_stage_unbind_entry,
- .pt_post_descend = xe_pt_stage_unbind_post_descend,
-};
-
-/**
- * xe_pt_stage_unbind() - Build page-table update structures for an unbind
- * operation
- * @tile: The tile we're unbinding for.
- * @vma: The vma we're unbinding.
- * @entries: Caller-provided storage for the update structures.
- *
- * Builds page-table update structures for an unbind operation. The function
- * will attempt to remove all page-tables that we're the only user
- * of, and for that to work, the unbind operation must be committed in the
- * same critical section that blocks racing binds to the same page-table tree.
- *
- * Return: The number of entries used.
- */
-static unsigned int xe_pt_stage_unbind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries)
-{
- struct xe_pt_stage_unbind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_unbind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- .modified_start = xe_vma_start(vma),
- .modified_end = xe_vma_end(vma),
- .wupd.entries = entries,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.wupd.num_used_entries;
-}
-
-static void
-xe_migrate_clear_pgtable_callback(struct xe_migrate_pt_update *pt_update,
- struct xe_tile *tile, struct iosys_map *map,
- void *ptr, u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_vma *vma = pt_update->vma;
- u64 empty = __xe_pt_empty_pte(tile, xe_vma_vm(vma), update->pt->level);
- int i;
-
- if (map && map->is_iomem)
- for (i = 0; i < num_qwords; ++i)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, empty);
- else if (map)
- memset64(map->vaddr + qword_ofs * sizeof(u64), empty,
- num_qwords);
- else
- memset64(ptr, empty, num_qwords);
-}
-
-static void
-xe_pt_commit_unbind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 num_entries,
- struct llist_head *deferred)
-{
- u32 j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (j = 0; j < num_entries; ++j) {
- struct xe_vm_pgtable_update *entry = &entries[j];
- struct xe_pt *pt = entry->pt;
-
- pt->num_live -= entry->qwords;
- if (pt->level) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
- u32 i;
-
- for (i = entry->ofs; i < entry->ofs + entry->qwords;
- i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[i] = NULL;
- }
- }
- }
-}
-
-<<<<<<<
-static const struct xe_migrate_pt_update_ops unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
-=======
-static void
-xe_pt_update_ops_rfence_interval(struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int i, level = 0;
- u64 start, last;
-
- for (i = 0; i < pt_op->num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &pt_op->entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- start = ALIGN_DOWN(xe_vma_start(vma), 0x1ull << xe_pt_shift(level));
- last = ALIGN(xe_vma_end(vma), 0x1ull << xe_pt_shift(level)) - 1;
-
- if (start < pt_update_ops->start)
- pt_update_ops->start = start;
- if (last > pt_update_ops->last)
- pt_update_ops->last = last;
-}
-
-static int vma_reserve_fences(struct xe_device *xe, struct xe_vma *vma)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- return dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv,
- xe->info.tile_count);
-
- return 0;
-}
-
-static int bind_op_prepare(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- pt_op->vma = NULL;
- pt_op->bind = true;
- pt_op->rebind = BIT(tile->id) & vma->tile_present;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- err = xe_pt_prepare_bind(tile, vma, pt_op->entries,
- &pt_op->num_entries);
- if (!err) {
- xe_tile_assert(tile, pt_op->num_entries <=
- ARRAY_SIZE(pt_op->entries));
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, true);
-
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. On LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * On !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!pt_op->rebind && xe_vm_has_scratch(vm) &&
- xe_vm_in_preempt_fence_mode(vm)))
- pt_update_ops->needs_invalidation = true;
- else if (pt_op->rebind && !xe_vm_in_lr_mode(vm))
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
-
- vma->tile_staged |= BIT(tile->id);
- pt_op->vma = vma;
- xe_pt_commit_prepare_bind(vma, pt_op->entries,
- pt_op->num_entries, pt_op->rebind);
- } else {
- xe_pt_cancel_bind(vma, pt_op->entries, pt_op->num_entries);
- }
-
- return err;
-}
-
-static int unbind_op_prepare(struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- if (!((vma->tile_present | vma->tile_staged) & BIT(tile->id)))
- return 0;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- /*
- * Wait for invalidation to complete. Can corrupt internal page table
- * state if an invalidation is running while preparing an unbind.
- */
- if (xe_vma_is_userptr(vma) && xe_vm_in_fault_mode(xe_vma_vm(vma)))
- mmu_interval_read_begin(&to_userptr_vma(vma)->userptr.notifier);
-
- pt_op->vma = vma;
- pt_op->bind = false;
- pt_op->rebind = false;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- pt_op->num_entries = xe_pt_stage_unbind(tile, vma, pt_op->entries);
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, false);
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
- pt_update_ops->needs_invalidation = true;
-
- xe_pt_commit_prepare_unbind(vma, pt_op->entries, pt_op->num_entries);
-
- return 0;
-}
-
-static int op_prepare(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op)
-{
- int err = 0;
-
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- err = bind_op_prepare(vm, tile, pt_update_ops, op->map.vma);
- pt_update_ops->wait_vm_kernel = true;
- break;
- case DRM_GPUVA_OP_REMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va));
-
- if (!err && op->remap.prev) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.prev);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- if (!err && op->remap.next) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.next);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- break;
- case DRM_GPUVA_OP_UNMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va));
- break;
- case DRM_GPUVA_OP_PREFETCH:
- err = bind_op_prepare(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va));
- pt_update_ops->wait_vm_kernel = true;
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-
- return err;
-}
-
-static void
-xe_pt_update_ops_init(struct xe_vm_pgtable_update_ops *pt_update_ops)
-{
- init_llist_head(&pt_update_ops->deferred);
- pt_update_ops->start = ~0x0ull;
- pt_update_ops->last = 0x0ull;
-}
-
-/**
- * xe_pt_update_ops_prepare() - Prepare PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Prepare PT update operations which includes updating internal PT state,
- * allocate memory for page tables, populate page table being pruned in, and
- * create PT update operations for leaf insertion / removal.
- *
- * Return: 0 on success, negative error code on error.
- */
-int xe_pt_update_ops_prepare(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct xe_vma_op *op;
- int err;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- xe_pt_update_ops_init(pt_update_ops);
-
- err = dma_resv_reserve_fences(xe_vm_resv(vops->vm),
- tile_to_xe(tile)->info.tile_count);
- if (err)
- return err;
-
- list_for_each_entry(op, &vops->list, link) {
- err = op_prepare(vops->vm, tile, pt_update_ops, op);
-
- if (err)
- return err;
- }
-
- xe_tile_assert(tile, pt_update_ops->current_op <=
- pt_update_ops->num_ops);
-
-#ifdef TEST_VM_OPS_ERROR
- if (vops->inject_error &&
- vops->vm->xe->vm_inject_error_position == FORCE_OP_ERROR_PREPARE)
- return -ENOSPC;
-#endif
-
- return 0;
-}
-
-static void bind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present |= BIT(tile->id);
- vma->tile_staged &= ~BIT(tile->id);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- to_userptr_vma(vma)->userptr.initial_bind = true;
- }
-
- /*
- * Kick rebind worker if this bind triggers preempt fences and not in
- * the rebind worker
- */
- if (pt_update_ops->wait_vm_bookkeep &&
- xe_vm_in_preempt_fence_mode(vm) &&
- !current->mm)
- xe_vm_queue_rebind_worker(vm);
-}
-
-static void unbind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present &= ~BIT(tile->id);
- if (!vma->tile_present) {
- list_del_init(&vma->combined_links.rebind);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
-
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- }
-}
-
-static void op_commit(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- bind_op_commit(vm, tile, pt_update_ops, op->map.vma, fence,
- fence2);
- break;
- case DRM_GPUVA_OP_REMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va), fence,
- fence2);
-
- if (op->remap.prev)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.prev,
- fence, fence2);
- if (op->remap.next)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.next,
- fence, fence2);
- break;
- case DRM_GPUVA_OP_UNMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va), fence, fence2);
- break;
- case DRM_GPUVA_OP_PREFETCH:
- bind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va), fence, fence2);
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-}
-
-static const struct xe_migrate_pt_update_ops migrate_ops = {
- .populate = xe_vm_populate_pgtable,
- .clear = xe_migrate_clear_pgtable_callback,
->>>>>>>
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-/**
- * __xe_pt_unbind_vma() - Disconnect and free a page-table tree for the vma
- * address range.
- * @tile: The tile to unbind for.
- * @vma: The vma to unbind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before disconnecting the tree to be destroyed.
- * @num_syncs: Number of @sync entries.
- *
- * This function builds a the xe_vm_pgtable_update entries abstracting the
- * operations needed to detach the page-table tree to be destroyed from the
- * man vm tree.
- * It then takes the relevant locks and submits the operations for
- * pipelined detachment of the gpu page-table from the vm main tree,
- * (which can be done either by the cpu and the GPU), Finally it frees the
- * detached page-table tree.
- *
- * Return: A valid dma-fence representing the pipelined detachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_unbind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs)
-{
-<<<<<<<
- struct xe_vm *vm = vops->vm;
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL, *mfence = NULL;
- struct dma_fence_chain *chain_fence = NULL;
- struct xe_range_fence *rfence;
- struct xe_vma_op *op;
- int err = 0, i;
- struct xe_migrate_pt_update update = {
- .ops = pt_update_ops->needs_userptr_lock ?
- &userptr_migrate_ops :
- &migrate_ops,
- .vops = vops,
- .tile_id = tile->id,
-=======
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update unbind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_unbind_ops :
- &unbind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
->>>>>>>
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence = NULL;
- struct invalidation_fence *ifence;
- struct xe_range_fence *rfence;
- int err;
-
- LLIST_HEAD(deferred);
-
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- num_entries = xe_pt_stage_unbind(tile, vma, entries);
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &unbind_pt_update, entries,
- num_entries);
-
-<<<<<<<
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- return ERR_PTR(err);
-
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
-=======
- if (pt_update_ops->needs_invalidation) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence) {
- err = -ENOMEM;
- goto kill_vm_tile1;
- }
- if (tile->media_gt) {
- mfence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!mfence) {
- err = -ENOMEM;
- goto free_ifence;
- }
- chain_fence = dma_fence_chain_alloc();
- if (!chain_fence) {
- err = -ENOMEM;
- goto free_ifence;
- }
- }
- }
->>>>>>>
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- /*
- * Even if we were already evicted and unbind to destroy, we need to
- * clear again here. The eviction may have updated pagetables at a
- * lower level, because it needs to be more conservative.
- */
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, NULL, q ? q :
- vm->q[tile->id],
- entries, num_entries,
- syncs, num_syncs,
- &unbind_pt_update.base);
- if (!IS_ERR(fence)) {
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- unbind_pt_update.base.start,
- unbind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
-<<<<<<<
- /* TLB invalidation must be done before signaling unbind */
- err = invalidation_fence_init(tile->primary_gt, ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- DMA_RESV_USAGE_BOOKKEEP);
-
- /* This fence will be installed by caller when doing eviction */
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_unbind(vma, entries, num_entries,
- unbind_pt_update.locked ? &deferred : NULL);
- vma->tile_present &= ~BIT(tile->id);
- } else {
- kfree(rfence);
- kfree(ifence);
- }
-
- if (!vma->tile_present)
- list_del_init(&vma->combined_links.rebind);
-
- if (unbind_pt_update.locked) {
- xe_tile_assert(tile, xe_vma_is_userptr(vma));
-=======
- xe_pt_commit(pt_op->vma, pt_op->entries,
- pt_op->num_entries, &pt_update_ops->deferred);
- pt_op->vma = NULL; /* skip in xe_pt_update_ops_abort */
- }
-
- if (xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- pt_update_ops->start,
- pt_update_ops->last, fence))
- dma_fence_wait(fence, false);
-
- /* tlb invalidation must be done before signaling rebind */
- if (ifence) {
- if (mfence)
- dma_fence_get(fence);
- invalidation_fence_init(tile->primary_gt, ifence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- if (mfence) {
- invalidation_fence_init(tile->media_gt, mfence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- dma_fence_chain_init(chain_fence, &ifence->base.base,
- &mfence->base.base, 0);
- fence = &chain_fence->base;
- } else {
- fence = &ifence->base.base;
- }
- }
-
- if (!mfence) {
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op, fence, NULL);
- } else {
- dma_resv_add_fence(xe_vm_resv(vm), &ifence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- dma_resv_add_fence(xe_vm_resv(vm), &mfence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op,
- &ifence->base.base, &mfence->base.base);
- }
->>>>>>>
-
- if (!vma->tile_present) {
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
-
- return fence;
-<<<<<<<
-=======
-
-free_rfence:
- kfree(rfence);
-free_ifence:
- dma_fence_chain_free(chain_fence);
- kfree(mfence);
- kfree(ifence);
-kill_vm_tile1:
- if (err != -EAGAIN && tile->id)
- xe_vm_kill(vops->vm, false);
-
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_update_ops_fini() - Finish PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operations
- *
- * Finish PT update operations by committing to destroy page table memory
- */
-void xe_pt_update_ops_fini(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = 0; i < pt_update_ops->current_op; ++i) {
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[i];
-
- xe_pt_free_bind(pt_op->entries, pt_op->num_entries);
- }
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
-}
-
-/**
- * xe_pt_update_ops_abort() - Abort PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Abort PT update operations by unwinding internal PT state
- */
-void xe_pt_update_ops_abort(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = pt_update_ops->num_ops - 1; i >= 0; --i) {
- struct xe_vm_pgtable_update_op *pt_op =
- &pt_update_ops->ops[i];
-
- if (!pt_op->vma || i >= pt_update_ops->current_op)
- continue;
-
- if (pt_op->bind)
- xe_pt_abort_bind(pt_op->vma, pt_op->entries,
- pt_op->num_entries,
- pt_op->rebind);
- else
- xe_pt_abort_unbind(pt_op->vma, pt_op->entries,
- pt_op->num_entries);
- }
-
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
->>>>>>>
-}
diff --git a/rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.2 b/rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.2
deleted file mode 100644
index bb6eadffff5e..000000000000
--- a/rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.2
+++ /dev/null
@@ -1,2248 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2022 Intel Corporation
- */
-
-#include <linux/dma-fence-array.h>
-
-#include "xe_pt.h"
-
-#include "regs/xe_gtt_defs.h"
-#include "xe_bo.h"
-#include "xe_device.h"
-#include "xe_drm_client.h"
-#include "xe_gt.h"
-#include "xe_gt_tlb_invalidation.h"
-#include "xe_migrate.h"
-#include "xe_pt_types.h"
-#include "xe_pt_walk.h"
-#include "xe_res_cursor.h"
-#include "xe_trace.h"
-#include "xe_ttm_stolen_mgr.h"
-#include "xe_vm.h"
-
-struct xe_pt_dir {
- struct xe_pt pt;
- /** @children: Array of page-table child nodes */
- struct xe_ptw *children[XE_PDES];
-};
-
-#if IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM)
-#define xe_pt_set_addr(__xe_pt, __addr) ((__xe_pt)->addr = (__addr))
-#define xe_pt_addr(__xe_pt) ((__xe_pt)->addr)
-#else
-#define xe_pt_set_addr(__xe_pt, __addr)
-#define xe_pt_addr(__xe_pt) 0ull
-#endif
-
-static const u64 xe_normal_pt_shifts[] = {12, 21, 30, 39, 48};
-static const u64 xe_compact_pt_shifts[] = {16, 21, 30, 39, 48};
-
-#define XE_PT_HIGHEST_LEVEL (ARRAY_SIZE(xe_normal_pt_shifts) - 1)
-
-static struct xe_pt_dir *as_xe_pt_dir(struct xe_pt *pt)
-{
- return container_of(pt, struct xe_pt_dir, pt);
-}
-
-static struct xe_pt *xe_pt_entry(struct xe_pt_dir *pt_dir, unsigned int index)
-{
- return container_of(pt_dir->children[index], struct xe_pt, base);
-}
-
-static u64 __xe_pt_empty_pte(struct xe_tile *tile, struct xe_vm *vm,
- unsigned int level)
-{
- struct xe_device *xe = tile_to_xe(tile);
- u16 pat_index = xe->pat.idx[XE_CACHE_WB];
- u8 id = tile->id;
-
- if (!xe_vm_has_scratch(vm))
- return 0;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return vm->pt_ops->pde_encode_bo(vm->scratch_pt[id][level - 1]->bo,
- 0, pat_index);
-
- return vm->pt_ops->pte_encode_addr(xe, 0, pat_index, level, IS_DGFX(xe), 0) |
- XE_PTE_NULL;
-}
-
-static void xe_pt_free(struct xe_pt *pt)
-{
- if (pt->level)
- kfree(as_xe_pt_dir(pt));
- else
- kfree(pt);
-}
-
-/**
- * xe_pt_create() - Create a page-table.
- * @vm: The vm to create for.
- * @tile: The tile to create for.
- * @level: The page-table level.
- *
- * Allocate and initialize a single struct xe_pt metadata structure. Also
- * create the corresponding page-table bo, but don't initialize it. If the
- * level is grater than zero, then it's assumed to be a directory page-
- * table and the directory structure is also allocated and initialized to
- * NULL pointers.
- *
- * Return: A valid struct xe_pt pointer on success, Pointer error code on
- * error.
- */
-struct xe_pt *xe_pt_create(struct xe_vm *vm, struct xe_tile *tile,
- unsigned int level)
-{
- struct xe_pt *pt;
- struct xe_bo *bo;
- int err;
-
- if (level) {
- struct xe_pt_dir *dir = kzalloc(sizeof(*dir), GFP_KERNEL);
-
- pt = (dir) ? &dir->pt : NULL;
- } else {
- pt = kzalloc(sizeof(*pt), GFP_KERNEL);
- }
- if (!pt)
- return ERR_PTR(-ENOMEM);
-
- pt->level = level;
- bo = xe_bo_create_pin_map(vm->xe, tile, vm, SZ_4K,
- ttm_bo_type_kernel,
- XE_BO_FLAG_VRAM_IF_DGFX(tile) |
- XE_BO_FLAG_IGNORE_MIN_PAGE_SIZE |
- XE_BO_FLAG_PINNED |
- XE_BO_FLAG_NO_RESV_EVICT |
- XE_BO_FLAG_PAGETABLE);
- if (IS_ERR(bo)) {
- err = PTR_ERR(bo);
- goto err_kfree;
- }
- pt->bo = bo;
- pt->base.children = level ? as_xe_pt_dir(pt)->children : NULL;
-
- if (vm->xef)
- xe_drm_client_add_bo(vm->xef->client, pt->bo);
- xe_tile_assert(tile, level <= XE_VM_MAX_LEVEL);
-
- return pt;
-
-err_kfree:
- xe_pt_free(pt);
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_populate_empty() - Populate a page-table bo with scratch- or zero
- * entries.
- * @tile: The tile the scratch pagetable of which to use.
- * @vm: The vm we populate for.
- * @pt: The pagetable the bo of which to initialize.
- *
- * Populate the page-table bo of @pt with entries pointing into the tile's
- * scratch page-table tree if any. Otherwise populate with zeros.
- */
-void xe_pt_populate_empty(struct xe_tile *tile, struct xe_vm *vm,
- struct xe_pt *pt)
-{
- struct iosys_map *map = &pt->bo->vmap;
- u64 empty;
- int i;
-
- if (!xe_vm_has_scratch(vm)) {
- /*
- * FIXME: Some memory is allocated already allocated to zero?
- * Find out which memory that is and avoid this memset...
- */
- xe_map_memset(vm->xe, map, 0, 0, SZ_4K);
- } else {
- empty = __xe_pt_empty_pte(tile, vm, pt->level);
- for (i = 0; i < XE_PDES; i++)
- xe_pt_write(vm->xe, map, i, empty);
- }
-}
-
-/**
- * xe_pt_shift() - Return the ilog2 value of the size of the address range of
- * a page-table at a certain level.
- * @level: The level.
- *
- * Return: The ilog2 value of the size of the address range of a page-table
- * at level @level.
- */
-unsigned int xe_pt_shift(unsigned int level)
-{
- return XE_PTE_SHIFT + XE_PDE_SHIFT * level;
-}
-
-/**
- * xe_pt_destroy() - Destroy a page-table tree.
- * @pt: The root of the page-table tree to destroy.
- * @flags: vm flags. Currently unused.
- * @deferred: List head of lockless list for deferred putting. NULL for
- * immediate putting.
- *
- * Puts the page-table bo, recursively calls xe_pt_destroy on all children
- * and finally frees @pt. TODO: Can we remove the @flags argument?
- */
-void xe_pt_destroy(struct xe_pt *pt, u32 flags, struct llist_head *deferred)
-{
- int i;
-
- if (!pt)
- return;
-
- XE_WARN_ON(!list_empty(&pt->bo->ttm.base.gpuva.list));
- xe_bo_unpin(pt->bo);
- xe_bo_put_deferred(pt->bo, deferred);
-
- if (pt->level > 0 && pt->num_live) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
-
- for (i = 0; i < XE_PDES; i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i), flags,
- deferred);
- }
- }
- xe_pt_free(pt);
-}
-
-/**
- * DOC: Pagetable building
- *
- * Below we use the term "page-table" for both page-directories, containing
- * pointers to lower level page-directories or page-tables, and level 0
- * page-tables that contain only page-table-entries pointing to memory pages.
- *
- * When inserting an address range in an already existing page-table tree
- * there will typically be a set of page-tables that are shared with other
- * address ranges, and a set that are private to this address range.
- * The set of shared page-tables can be at most two per level,
- * and those can't be updated immediately because the entries of those
- * page-tables may still be in use by the gpu for other mappings. Therefore
- * when inserting entries into those, we instead stage those insertions by
- * adding insertion data into struct xe_vm_pgtable_update structures. This
- * data, (subtrees for the cpu and page-table-entries for the gpu) is then
- * added in a separate commit step. CPU-data is committed while still under the
- * vm lock, the object lock and for userptr, the notifier lock in read mode.
- * The GPU async data is committed either by the GPU or CPU after fulfilling
- * relevant dependencies.
- * For non-shared page-tables (and, in fact, for shared ones that aren't
- * existing at the time of staging), we add the data in-place without the
- * special update structures. This private part of the page-table tree will
- * remain disconnected from the vm page-table tree until data is committed to
- * the shared page tables of the vm tree in the commit phase.
- */
-
-struct xe_pt_update {
- /** @update: The update structure we're building for this parent. */
- struct xe_vm_pgtable_update *update;
- /** @parent: The parent. Used to detect a parent change. */
- struct xe_pt *parent;
- /** @preexisting: Whether the parent was pre-existing or allocated */
- bool preexisting;
-};
-
-struct xe_pt_stage_bind_walk {
- /** base: The base class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @vm: The vm we're building for. */
- struct xe_vm *vm;
- /** @tile: The tile we're building for. */
- struct xe_tile *tile;
- /** @default_pte: PTE flag only template. No address is associated */
- u64 default_pte;
- /** @dma_offset: DMA offset to add to the PTE. */
- u64 dma_offset;
- /**
- * @needs_64k: This address range enforces 64K alignment and
- * granularity.
- */
- bool needs_64K;
- /**
- * @vma: VMA being mapped
- */
- struct xe_vma *vma;
-
- /* Also input, but is updated during the walk*/
- /** @curs: The DMA address cursor. */
- struct xe_res_cursor *curs;
- /** @va_curs_start: The Virtual address coresponding to @curs->start */
- u64 va_curs_start;
-
- /* Output */
- struct xe_walk_update {
- /** @wupd.entries: Caller provided storage. */
- struct xe_vm_pgtable_update *entries;
- /** @wupd.num_used_entries: Number of update @entries used. */
- unsigned int num_used_entries;
- /** @wupd.updates: Tracks the update entry at a given level */
- struct xe_pt_update updates[XE_VM_MAX_LEVEL + 1];
- } wupd;
-
- /* Walk state */
- /**
- * @l0_end_addr: The end address of the current l0 leaf. Used for
- * 64K granularity detection.
- */
- u64 l0_end_addr;
- /** @addr_64K: The start address of the current 64K chunk. */
- u64 addr_64K;
- /** @found_64: Whether @add_64K actually points to a 64K chunk. */
- bool found_64K;
-};
-
-static int
-xe_pt_new_shared(struct xe_walk_update *wupd, struct xe_pt *parent,
- pgoff_t offset, bool alloc_entries)
-{
- struct xe_pt_update *upd = &wupd->updates[parent->level];
- struct xe_vm_pgtable_update *entry;
-
- /*
- * For *each level*, we could only have one active
- * struct xt_pt_update at any one time. Once we move on to a
- * new parent and page-directory, the old one is complete, and
- * updates are either already stored in the build tree or in
- * @wupd->entries
- */
- if (likely(upd->parent == parent))
- return 0;
-
- upd->parent = parent;
- upd->preexisting = true;
-
- if (wupd->num_used_entries == XE_VM_MAX_LEVEL * 2 + 1)
- return -EINVAL;
-
- entry = wupd->entries + wupd->num_used_entries++;
- upd->update = entry;
- entry->ofs = offset;
- entry->pt_bo = parent->bo;
- entry->pt = parent;
- entry->flags = 0;
- entry->qwords = 0;
-
- if (alloc_entries) {
- entry->pt_entries = kmalloc_array(XE_PDES,
- sizeof(*entry->pt_entries),
- GFP_KERNEL);
- if (!entry->pt_entries)
- return -ENOMEM;
- }
-
- return 0;
-}
-
-/*
- * NOTE: This is a very frequently called function so we allow ourselves
- * to annotate (using branch prediction hints) the fastpath of updating a
- * non-pre-existing pagetable with leaf ptes.
- */
-static int
-xe_pt_insert_entry(struct xe_pt_stage_bind_walk *xe_walk, struct xe_pt *parent,
- pgoff_t offset, struct xe_pt *xe_child, u64 pte)
-{
- struct xe_pt_update *upd = &xe_walk->wupd.updates[parent->level];
- struct xe_pt_update *child_upd = xe_child ?
- &xe_walk->wupd.updates[xe_child->level] : NULL;
- int ret;
-
- ret = xe_pt_new_shared(&xe_walk->wupd, parent, offset, true);
- if (unlikely(ret))
- return ret;
-
- /*
- * Register this new pagetable so that it won't be recognized as
- * a shared pagetable by a subsequent insertion.
- */
- if (unlikely(child_upd)) {
- child_upd->update = NULL;
- child_upd->parent = xe_child;
- child_upd->preexisting = false;
- }
-
- if (likely(!upd->preexisting)) {
- /* Continue building a non-connected subtree. */
- struct iosys_map *map = &parent->bo->vmap;
-
- if (unlikely(xe_child))
- parent->base.children[offset] = &xe_child->base;
-
- xe_pt_write(xe_walk->vm->xe, map, offset, pte);
- parent->num_live++;
- } else {
- /* Shared pt. Stage update. */
- unsigned int idx;
- struct xe_vm_pgtable_update *entry = upd->update;
-
- idx = offset - entry->ofs;
- entry->pt_entries[idx].pt = xe_child;
- entry->pt_entries[idx].pte = pte;
- entry->qwords++;
- }
-
- return 0;
-}
-
-static bool xe_pt_hugepte_possible(u64 addr, u64 next, unsigned int level,
- struct xe_pt_stage_bind_walk *xe_walk)
-{
- u64 size, dma;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return false;
-
- /* Does the virtual range requested cover a huge pte? */
- if (!xe_pt_covers(addr, next, level, &xe_walk->base))
- return false;
-
- /* Does the DMA segment cover the whole pte? */
- if (next - xe_walk->va_curs_start > xe_walk->curs->size)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- /* Is the DMA address huge PTE size aligned? */
- size = next - addr;
- dma = addr - xe_walk->va_curs_start + xe_res_dma(xe_walk->curs);
-
- return IS_ALIGNED(dma, size);
-}
-
-/*
- * Scan the requested mapping to check whether it can be done entirely
- * with 64K PTEs.
- */
-static bool
-xe_pt_scan_64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- struct xe_res_cursor curs = *xe_walk->curs;
-
- if (!IS_ALIGNED(addr, SZ_64K))
- return false;
-
- if (next > xe_walk->l0_end_addr)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- xe_res_next(&curs, addr - xe_walk->va_curs_start);
- for (; addr < next; addr += SZ_64K) {
- if (!IS_ALIGNED(xe_res_dma(&curs), SZ_64K) || curs.size < SZ_64K)
- return false;
-
- xe_res_next(&curs, SZ_64K);
- }
-
- return addr == next;
-}
-
-/*
- * For non-compact "normal" 4K level-0 pagetables, we want to try to group
- * addresses together in 64K-contigous regions to add a 64K TLB hint for the
- * device to the PTE.
- * This function determines whether the address is part of such a
- * segment. For VRAM in normal pagetables, this is strictly necessary on
- * some devices.
- */
-static bool
-xe_pt_is_pte_ps64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- /* Address is within an already found 64k region */
- if (xe_walk->found_64K && addr - xe_walk->addr_64K < SZ_64K)
- return true;
-
- xe_walk->found_64K = xe_pt_scan_64K(addr, addr + SZ_64K, xe_walk);
- xe_walk->addr_64K = addr;
-
- return xe_walk->found_64K;
-}
-
-static int
-xe_pt_stage_bind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_bind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- u16 pat_index = xe_walk->vma->pat_index;
- struct xe_pt *xe_parent = container_of(parent, typeof(*xe_parent), base);
- struct xe_vm *vm = xe_walk->vm;
- struct xe_pt *xe_child;
- bool covers;
- int ret = 0;
- u64 pte;
-
- /* Is this a leaf entry ?*/
- if (level == 0 || xe_pt_hugepte_possible(addr, next, level, xe_walk)) {
- struct xe_res_cursor *curs = xe_walk->curs;
- bool is_null = xe_vma_is_null(xe_walk->vma);
-
- XE_WARN_ON(xe_walk->va_curs_start != addr);
-
- pte = vm->pt_ops->pte_encode_vma(is_null ? 0 :
- xe_res_dma(curs) + xe_walk->dma_offset,
- xe_walk->vma, pat_index, level);
- pte |= xe_walk->default_pte;
-
- /*
- * Set the XE_PTE_PS64 hint if possible, otherwise if
- * this device *requires* 64K PTE size for VRAM, fail.
- */
- if (level == 0 && !xe_parent->is_compact) {
- if (xe_pt_is_pte_ps64K(addr, next, xe_walk)) {
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_64K;
- pte |= XE_PTE_PS64;
- } else if (XE_WARN_ON(xe_walk->needs_64K)) {
- return -EINVAL;
- }
- }
-
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, NULL, pte);
- if (unlikely(ret))
- return ret;
-
- if (!is_null)
- xe_res_next(curs, next - addr);
- xe_walk->va_curs_start = next;
- xe_walk->vma->gpuva.flags |= (XE_VMA_PTE_4K << level);
- *action = ACTION_CONTINUE;
-
- return ret;
- }
-
- /*
- * Descending to lower level. Determine if we need to allocate a
- * new page table or -directory, which we do if there is no
- * previous one or there is one we can completely replace.
- */
- if (level == 1) {
- walk->shifts = xe_normal_pt_shifts;
- xe_walk->l0_end_addr = next;
- }
-
- covers = xe_pt_covers(addr, next, level, &xe_walk->base);
- if (covers || !*child) {
- u64 flags = 0;
-
- xe_child = xe_pt_create(xe_walk->vm, xe_walk->tile, level - 1);
- if (IS_ERR(xe_child))
- return PTR_ERR(xe_child);
-
- xe_pt_set_addr(xe_child,
- round_down(addr, 1ull << walk->shifts[level]));
-
- if (!covers)
- xe_pt_populate_empty(xe_walk->tile, xe_walk->vm, xe_child);
-
- *child = &xe_child->base;
-
- /*
- * Prefer the compact pagetable layout for L0 if possible. Only
- * possible if VMA covers entire 2MB region as compact 64k and
- * 4k pages cannot be mixed within a 2MB region.
- * TODO: Suballocate the pt bo to avoid wasting a lot of
- * memory.
- */
- if (GRAPHICS_VERx100(tile_to_xe(xe_walk->tile)) >= 1250 && level == 1 &&
- covers && xe_pt_scan_64K(addr, next, xe_walk)) {
- walk->shifts = xe_compact_pt_shifts;
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_COMPACT;
- flags |= XE_PDE_64K;
- xe_child->is_compact = true;
- }
-
- pte = vm->pt_ops->pde_encode_bo(xe_child->bo, 0, pat_index) | flags;
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, xe_child,
- pte);
- }
-
- *action = ACTION_SUBTREE;
- return ret;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_bind_ops = {
- .pt_entry = xe_pt_stage_bind_entry,
-};
-
-/**
- * xe_pt_stage_bind() - Build a disconnected page-table tree for a given address
- * range.
- * @tile: The tile we're building for.
- * @vma: The vma indicating the address range.
- * @entries: Storage for the update entries used for connecting the tree to
- * the main tree at commit time.
- * @num_entries: On output contains the number of @entries used.
- *
- * This function builds a disconnected page-table tree for a given address
- * range. The tree is connected to the main vm tree for the gpu using
- * xe_migrate_update_pgtables() and for the cpu using xe_pt_commit_bind().
- * The function builds xe_vm_pgtable_update structures for already existing
- * shared page-tables, and non-existing shared and non-shared page-tables
- * are built and populated directly.
- *
- * Return 0 on success, negative error code on error.
- */
-static int
-xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- struct xe_device *xe = tile_to_xe(tile);
- struct xe_bo *bo = xe_vma_bo(vma);
- bool is_devmem = !xe_vma_is_userptr(vma) && bo &&
- (xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo));
- struct xe_res_cursor curs;
- struct xe_pt_stage_bind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_bind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .vm = xe_vma_vm(vma),
- .tile = tile,
- .curs = &curs,
- .va_curs_start = xe_vma_start(vma),
- .vma = vma,
- .wupd.entries = entries,
- .needs_64K = (xe_vma_vm(vma)->flags & XE_VM_FLAG_64K) && is_devmem,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- int ret;
-
- /**
- * Default atomic expectations for different allocation scenarios are as follows:
- *
- * 1. Traditional API: When the VM is not in LR mode:
- * - Device atomics are expected to function with all allocations.
- *
- * 2. Compute/SVM API: When the VM is in LR mode:
- * - Device atomics are the default behavior when the bo is placed in a single region.
- * - In all other cases device atomics will be disabled with AE=0 until an application
- * request differently using a ioctl like madvise.
- */
- if (vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT) {
- if (xe_vm_in_lr_mode(xe_vma_vm(vma))) {
- if (bo && xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- /**
- * If a SMEM+LMEM allocation is backed by SMEM, a device
- * atomics will cause a gpu page fault and which then
- * gets migrated to LMEM, bind such allocations with
- * device atomics enabled.
- */
- else if (is_devmem && !xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- } else {
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- }
-
- /**
- * Unset AE if the platform(PVC) doesn't support it on an
- * allocation
- */
- if (!xe->info.has_device_atomics_on_smem && !is_devmem)
- xe_walk.default_pte &= ~XE_USM_PPGTT_PTE_AE;
- }
-
- if (is_devmem) {
- xe_walk.default_pte |= XE_PPGTT_PTE_DM;
- xe_walk.dma_offset = vram_region_gpu_offset(bo->ttm.resource);
- }
-
- if (!xe_vma_has_no_bo(vma) && xe_bo_is_stolen(bo))
- xe_walk.dma_offset = xe_ttm_stolen_gpu_offset(xe_bo_device(bo));
-
- xe_bo_assert_held(bo);
-
- if (!xe_vma_is_null(vma)) {
- if (xe_vma_is_userptr(vma))
- xe_res_first_sg(to_userptr_vma(vma)->userptr.sg, 0,
- xe_vma_size(vma), &curs);
- else if (xe_bo_is_vram(bo) || xe_bo_is_stolen(bo))
- xe_res_first(bo->ttm.resource, xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- else
- xe_res_first_sg(xe_bo_sg(bo), xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- } else {
- curs.size = xe_vma_size(vma);
- }
-
- ret = xe_pt_walk_range(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- *num_entries = xe_walk.wupd.num_used_entries;
- return ret;
-}
-
-/**
- * xe_pt_nonshared_offsets() - Determine the non-shared entry offsets of a
- * shared pagetable.
- * @addr: The start address within the non-shared pagetable.
- * @end: The end address within the non-shared pagetable.
- * @level: The level of the non-shared pagetable.
- * @walk: Walk info. The function adjusts the walk action.
- * @action: next action to perform (see enum page_walk_action)
- * @offset: Ignored on input, First non-shared entry on output.
- * @end_offset: Ignored on input, Last non-shared entry + 1 on output.
- *
- * A non-shared page-table has some entries that belong to the address range
- * and others that don't. This function determines the entries that belong
- * fully to the address range. Depending on level, some entries may
- * partially belong to the address range (that can't happen at level 0).
- * The function detects that and adjust those offsets to not include those
- * partial entries. Iff it does detect partial entries, we know that there must
- * be shared page tables also at lower levels, so it adjusts the walk action
- * accordingly.
- *
- * Return: true if there were non-shared entries, false otherwise.
- */
-static bool xe_pt_nonshared_offsets(u64 addr, u64 end, unsigned int level,
- struct xe_pt_walk *walk,
- enum page_walk_action *action,
- pgoff_t *offset, pgoff_t *end_offset)
-{
- u64 size = 1ull << walk->shifts[level];
-
- *offset = xe_pt_offset(addr, level, walk);
- *end_offset = xe_pt_num_entries(addr, end, level, walk) + *offset;
-
- if (!level)
- return true;
-
- /*
- * If addr or next are not size aligned, there are shared pts at lower
- * level, so in that case traverse down the subtree
- */
- *action = ACTION_CONTINUE;
- if (!IS_ALIGNED(addr, size)) {
- *action = ACTION_SUBTREE;
- (*offset)++;
- }
-
- if (!IS_ALIGNED(end, size)) {
- *action = ACTION_SUBTREE;
- (*end_offset)--;
- }
-
- return *end_offset > *offset;
-}
-
-struct xe_pt_zap_ptes_walk {
- /** @base: The walk base-class */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're building for */
- struct xe_tile *tile;
-
- /* Output */
- /** @needs_invalidate: Whether we need to invalidate TLB*/
- bool needs_invalidate;
-};
-
-static int xe_pt_zap_ptes_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_zap_ptes_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- /*
- * Note that we're called from an entry callback, and we're dealing
- * with the child of that entry rather than the parent, so need to
- * adjust level down.
- */
- if (xe_pt_nonshared_offsets(addr, next, --level, walk, action, &offset,
- &end_offset)) {
- xe_map_memset(tile_to_xe(xe_walk->tile), &xe_child->bo->vmap,
- offset * sizeof(u64), 0,
- (end_offset - offset) * sizeof(u64));
- xe_walk->needs_invalidate = true;
- }
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_zap_ptes_ops = {
- .pt_entry = xe_pt_zap_ptes_entry,
-};
-
-/**
- * xe_pt_zap_ptes() - Zap (zero) gpu ptes of an address range
- * @tile: The tile we're zapping for.
- * @vma: GPU VMA detailing address range.
- *
- * Eviction and Userptr invalidation needs to be able to zap the
- * gpu ptes of a given address range in pagefaulting mode.
- * In order to be able to do that, that function needs access to the shared
- * page-table entrieaso it can either clear the leaf PTEs or
- * clear the pointers to lower-level page-tables. The caller is required
- * to hold the necessary locks to ensure neither the page-table connectivity
- * nor the page-table entries of the range is updated from under us.
- *
- * Return: Whether ptes were actually updated and a TLB invalidation is
- * required.
- */
-bool xe_pt_zap_ptes(struct xe_tile *tile, struct xe_vma *vma)
-{
- struct xe_pt_zap_ptes_walk xe_walk = {
- .base = {
- .ops = &xe_pt_zap_ptes_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- u8 pt_mask = (vma->tile_present & ~vma->tile_invalidated);
-
- if (!(pt_mask & BIT(tile->id)))
- return false;
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.needs_invalidate;
-}
-
-static void
-xe_vm_populate_pgtable(struct xe_migrate_pt_update *pt_update, struct xe_tile *tile,
- struct iosys_map *map, void *data,
- u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_pt_entry *ptes = update->pt_entries;
- u64 *ptr = data;
- u32 i;
-
- for (i = 0; i < num_qwords; i++) {
- if (map)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, ptes[i].pte);
- else
- ptr[i] = ptes[i].pte;
- }
-}
-
-static void xe_pt_abort_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- u32 i, j;
-
- for (i = 0; i < num_entries; i++) {
- if (!entries[i].pt_entries)
- continue;
-
- for (j = 0; j < entries[i].qwords; j++)
- xe_pt_destroy(entries[i].pt_entries[j].pt, xe_vma_vm(vma)->flags, NULL);
- kfree(entries[i].pt_entries);
- }
-}
-
-static void xe_pt_commit_locks_assert(struct xe_vma *vma)
-{
- struct xe_vm *vm = xe_vma_vm(vma);
-
- lockdep_assert_held(&vm->lock);
-
- if (xe_vma_is_userptr(vma))
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- else if (!xe_vma_is_null(vma))
- dma_resv_assert_held(xe_vma_bo(vma)->ttm.base.resv);
-
- xe_vm_assert_held(vm);
-}
-
-static void xe_pt_commit_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries, bool rebind,
- struct llist_head *deferred)
-{
- u32 i, j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (i = 0; i < num_entries; i++) {
- struct xe_pt *pt = entries[i].pt;
- struct xe_pt_dir *pt_dir;
-
- if (!rebind)
- pt->num_live += entries[i].qwords;
-
- if (!pt->level) {
- kfree(entries[i].pt_entries);
- continue;
- }
-
- pt_dir = as_xe_pt_dir(pt);
- for (j = 0; j < entries[i].qwords; j++) {
- u32 j_ = j + entries[i].ofs;
- struct xe_pt *newpte = entries[i].pt_entries[j].pt;
-
- if (xe_pt_entry(pt_dir, j_))
- xe_pt_destroy(xe_pt_entry(pt_dir, j_),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[j_] = &newpte->base;
- }
- kfree(entries[i].pt_entries);
- }
-}
-
-static int
-xe_pt_prepare_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- int err;
-
- *num_entries = 0;
- err = xe_pt_stage_bind(tile, vma, entries, num_entries);
- if (!err)
- xe_tile_assert(tile, *num_entries);
- else /* abort! */
- xe_pt_abort_bind(vma, entries, *num_entries);
-
- return err;
-}
-
-static void xe_vm_dbg_print_entries(struct xe_device *xe,
- const struct xe_vm_pgtable_update *entries,
- unsigned int num_entries)
-#if (IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM))
-{
- unsigned int i;
-
- vm_dbg(&xe->drm, "%u entries to update\n", num_entries);
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
- struct xe_pt *xe_pt = entry->pt;
- u64 page_size = 1ull << xe_pt_shift(xe_pt->level);
- u64 end;
- u64 start;
-
- xe_assert(xe, !entry->pt->is_compact);
- start = entry->ofs * page_size;
- end = start + page_size * entry->qwords;
- vm_dbg(&xe->drm,
- "\t%u: Update level %u at (%u + %u) [%llx...%llx) f:%x\n",
- i, xe_pt->level, entry->ofs, entry->qwords,
- xe_pt_addr(xe_pt) + start, xe_pt_addr(xe_pt) + end, 0);
- }
-}
-#else
-{}
-#endif
-
-#ifdef CONFIG_DRM_XE_USERPTR_INVAL_INJECT
-
-static int xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- u32 divisor = uvma->userptr.divisor ? uvma->userptr.divisor : 2;
- static u32 count;
-
- if (count++ % divisor == divisor - 1) {
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
-
- uvma->userptr.divisor = divisor << 1;
- spin_lock(&vm->userptr.invalidated_lock);
- list_move_tail(&uvma->userptr.invalidate_link,
- &vm->userptr.invalidated);
- spin_unlock(&vm->userptr.invalidated_lock);
- return true;
- }
-
- return false;
-}
-
-#else
-
-static bool xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- return false;
-}
-
-#endif
-
-/**
- * struct xe_pt_migrate_pt_update - Callback argument for pre-commit callbacks
- * @base: Base we derive from.
- * @bind: Whether this is a bind or an unbind operation. A bind operation
- * makes the pre-commit callback error with -EAGAIN if it detects a
- * pending invalidation.
- * @locked: Whether the pre-commit callback locked the userptr notifier lock
- * and it needs unlocking.
- */
-struct xe_pt_migrate_pt_update {
- struct xe_migrate_pt_update base;
- bool bind;
- bool locked;
-};
-
-/*
- * This function adds the needed dependencies to a page-table update job
- * to make sure racing jobs for separate bind engines don't race writing
- * to the same page-table range, wreaking havoc. Initially use a single
- * fence for the entire VM. An optimization would use smaller granularity.
- */
-static int xe_pt_vm_dependencies(struct xe_sched_job *job,
- struct xe_range_fence_tree *rftree,
- u64 start, u64 last)
-{
- struct xe_range_fence *rtfence;
- struct dma_fence *fence;
- int err;
-
- rtfence = xe_range_fence_tree_first(rftree, start, last);
- while (rtfence) {
- fence = rtfence->fence;
-
- if (!dma_fence_is_signaled(fence)) {
- /*
- * Is this a CPU update? GPU is busy updating, so return
- * an error
- */
- if (!job)
- return -ETIME;
-
- dma_fence_get(fence);
- err = drm_sched_job_add_dependency(&job->drm, fence);
- if (err)
- return err;
- }
-
- rtfence = xe_range_fence_tree_next(rtfence, start, last);
- }
-
- return 0;
-}
-
-static int xe_pt_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_range_fence_tree *rftree =
- &xe_vma_vm(pt_update->vma)->rftree[pt_update->tile_id];
-
- return xe_pt_vm_dependencies(pt_update->job, rftree,
- pt_update->start, pt_update->last);
-}
-
-static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_pt_migrate_pt_update *userptr_update =
- container_of(pt_update, typeof(*userptr_update), base);
- struct xe_userptr_vma *uvma = to_userptr_vma(pt_update->vma);
- unsigned long notifier_seq = uvma->userptr.notifier_seq;
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
- int err = xe_pt_vm_dependencies(pt_update->job,
- &vm->rftree[pt_update->tile_id],
- pt_update->start,
- pt_update->last);
-
- if (err)
- return err;
-
- userptr_update->locked = false;
-
- /*
- * Wait until nobody is running the invalidation notifier, and
- * since we're exiting the loop holding the notifier lock,
- * nobody can proceed invalidating either.
- *
- * Note that we don't update the vma->userptr.notifier_seq since
- * we don't update the userptr pages.
- */
- do {
- down_read(&vm->userptr.notifier_lock);
- if (!mmu_interval_read_retry(&uvma->userptr.notifier,
- notifier_seq))
- break;
-
- up_read(&vm->userptr.notifier_lock);
-
- if (userptr_update->bind)
- return -EAGAIN;
-
- notifier_seq = mmu_interval_read_begin(&uvma->userptr.notifier);
- } while (true);
-
- /* Inject errors to test_whether they are handled correctly */
- if (userptr_update->bind && xe_pt_userptr_inject_eagain(uvma)) {
- up_read(&vm->userptr.notifier_lock);
- return -EAGAIN;
- }
-
- userptr_update->locked = true;
-
- return 0;
-}
-
-static const struct xe_migrate_pt_update_ops bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-struct invalidation_fence {
- struct xe_gt_tlb_invalidation_fence base;
- struct xe_gt *gt;
- struct dma_fence *fence;
- struct dma_fence_cb cb;
- struct work_struct work;
- u64 start;
- u64 end;
- u32 asid;
-};
-
-static void invalidation_fence_cb(struct dma_fence *fence,
- struct dma_fence_cb *cb)
-{
- struct invalidation_fence *ifence =
- container_of(cb, struct invalidation_fence, cb);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_cb(xe, &ifence->base);
- if (!ifence->fence->error) {
- queue_work(system_wq, &ifence->work);
- } else {
- ifence->base.base.error = ifence->fence->error;
- dma_fence_signal(&ifence->base.base);
- dma_fence_put(&ifence->base.base);
- }
- dma_fence_put(ifence->fence);
-}
-
-static void invalidation_fence_work_func(struct work_struct *w)
-{
- struct invalidation_fence *ifence =
- container_of(w, struct invalidation_fence, work);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_work_func(xe, &ifence->base);
- xe_gt_tlb_invalidation_range(ifence->gt, &ifence->base, ifence->start,
- ifence->end, ifence->asid);
-}
-
-static int invalidation_fence_init(struct xe_gt *gt,
- struct invalidation_fence *ifence,
- struct dma_fence *fence,
- u64 start, u64 end, u32 asid)
-{
- int ret;
-
- trace_xe_gt_tlb_invalidation_fence_create(gt_to_xe(gt), &ifence->base);
-
- xe_gt_tlb_invalidation_fence_init(gt, &ifence->base, false);
-
- ifence->fence = fence;
- ifence->gt = gt;
- ifence->start = start;
- ifence->end = end;
- ifence->asid = asid;
-
- INIT_WORK(&ifence->work, invalidation_fence_work_func);
- ret = dma_fence_add_callback(fence, &ifence->cb, invalidation_fence_cb);
- if (ret == -ENOENT) {
- dma_fence_put(ifence->fence); /* Usually dropped in CB */
- invalidation_fence_work_func(&ifence->work);
- } else if (ret) {
- dma_fence_put(&ifence->base.base); /* Caller ref */
- dma_fence_put(&ifence->base.base); /* Creation ref */
- }
-
- xe_gt_assert(gt, !ret || ret == -ENOENT);
-
- return ret && ret != -ENOENT ? ret : 0;
-}
-
-static void xe_pt_calc_rfence_interval(struct xe_vma *vma,
- struct xe_pt_migrate_pt_update *update,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- int i, level = 0;
-
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- update->base.start = ALIGN_DOWN(xe_vma_start(vma),
- 0x1ull << xe_pt_shift(level));
- update->base.last = ALIGN(xe_vma_end(vma),
- 0x1ull << xe_pt_shift(level)) - 1;
-}
-
-/**
- * __xe_pt_bind_vma() - Build and connect a page-table tree for the vma
- * address range.
- * @tile: The tile to bind for.
- * @vma: The vma to bind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before binding the built tree to the live vm tree.
- * @num_syncs: Number of @sync entries.
- * @rebind: Whether we're rebinding this vma to the same address range without
- * an unbind in-between.
- *
- * This function builds a page-table tree (see xe_pt_stage_bind() for more
- * information on page-table building), and the xe_vm_pgtable_update entries
- * abstracting the operations needed to attach it to the main vm tree. It
- * then takes the relevant locks and updates the metadata side of the main
- * vm tree and submits the operations for pipelined attachment of the
- * gpu page-table to the vm main tree, (which can be done either by the
- * cpu and the GPU).
- *
- * Return: A valid dma-fence representing the pipelined attachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs,
- bool rebind)
-{
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update bind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_bind_ops : &bind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
- .bind = true,
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL;
- struct xe_range_fence *rfence;
- int err;
-
- bind_pt_update.locked = false;
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- err = xe_pt_prepare_bind(tile, vma, entries, &num_entries);
- if (err)
- goto err;
-
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- goto err;
-
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &bind_pt_update, entries,
- num_entries);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. on LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * on !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!rebind && xe_vm_has_scratch(vm) && xe_vm_in_preempt_fence_mode(vm))) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
- } else if (rebind && !xe_vm_in_lr_mode(vm)) {
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
- }
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, xe_vma_bo(vma), q,
- entries, num_entries,
- syncs, num_syncs,
- &bind_pt_update.base);
- if (!IS_ERR(fence)) {
- bool last_munmap_rebind = vma->gpuva.flags & XE_VMA_LAST_REBIND;
- LLIST_HEAD(deferred);
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- bind_pt_update.base.start,
- bind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
- /* TLB invalidation must be done before signaling rebind */
- if (ifence) {
- int err = invalidation_fence_init(tile->primary_gt,
- ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
- }
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence, rebind ||
- last_munmap_rebind ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_bind(vma, entries, num_entries, rebind,
- bind_pt_update.locked ? &deferred : NULL);
-
- /* This vma is live (again?) now */
- vma->tile_present |= BIT(tile->id);
-
- if (bind_pt_update.locked) {
- to_userptr_vma(vma)->userptr.initial_bind = true;
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
- if (!rebind && last_munmap_rebind &&
- xe_vm_in_preempt_fence_mode(vm))
- xe_vm_queue_rebind_worker(vm);
- } else {
- kfree(rfence);
- kfree(ifence);
- if (bind_pt_update.locked)
- up_read(&vm->userptr.notifier_lock);
- xe_pt_abort_bind(vma, entries, num_entries);
- }
-
- return fence;
-
-err:
- return ERR_PTR(err);
-}
-
-struct xe_pt_stage_unbind_walk {
- /** @base: The pagewalk base-class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're unbinding from. */
- struct xe_tile *tile;
-
- /**
- * @modified_start: Walk range start, modified to include any
- * shared pagetables that we're the only user of and can thus
- * treat as private.
- */
- u64 modified_start;
- /** @modified_end: Walk range start, modified like @modified_start. */
- u64 modified_end;
-
- /* Output */
- /* @wupd: Structure to track the page-table updates we're building */
- struct xe_walk_update wupd;
-};
-
-/*
- * Check whether this range is the only one populating this pagetable,
- * and in that case, update the walk range checks so that higher levels don't
- * view us as a shared pagetable.
- */
-static bool xe_pt_check_kill(u64 addr, u64 next, unsigned int level,
- const struct xe_pt *child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- unsigned int shift = walk->shifts[level];
- u64 size = 1ull << shift;
-
- if (IS_ALIGNED(addr, size) && IS_ALIGNED(next, size) &&
- ((next - addr) >> shift) == child->num_live) {
- u64 size = 1ull << walk->shifts[level + 1];
-
- *action = ACTION_CONTINUE;
-
- if (xe_walk->modified_start >= addr)
- xe_walk->modified_start = round_down(addr, size);
- if (xe_walk->modified_end <= next)
- xe_walk->modified_end = round_up(next, size);
-
- return true;
- }
-
- return false;
-}
-
-static int xe_pt_stage_unbind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- xe_pt_check_kill(addr, next, level - 1, xe_child, action, walk);
-
- return 0;
-}
-
-static int
-xe_pt_stage_unbind_post_descend(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
- u64 size = 1ull << walk->shifts[--level];
-
- if (!IS_ALIGNED(addr, size))
- addr = xe_walk->modified_start;
- if (!IS_ALIGNED(next, size))
- next = xe_walk->modified_end;
-
- /* Parent == *child is the root pt. Don't kill it. */
- if (parent != *child &&
- xe_pt_check_kill(addr, next, level, xe_child, action, walk))
- return 0;
-
- if (!xe_pt_nonshared_offsets(addr, next, level, walk, action, &offset,
- &end_offset))
- return 0;
-
- (void)xe_pt_new_shared(&xe_walk->wupd, xe_child, offset, false);
- xe_walk->wupd.updates[level].update->qwords = end_offset - offset;
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_unbind_ops = {
- .pt_entry = xe_pt_stage_unbind_entry,
- .pt_post_descend = xe_pt_stage_unbind_post_descend,
-};
-
-/**
- * xe_pt_stage_unbind() - Build page-table update structures for an unbind
- * operation
- * @tile: The tile we're unbinding for.
- * @vma: The vma we're unbinding.
- * @entries: Caller-provided storage for the update structures.
- *
- * Builds page-table update structures for an unbind operation. The function
- * will attempt to remove all page-tables that we're the only user
- * of, and for that to work, the unbind operation must be committed in the
- * same critical section that blocks racing binds to the same page-table tree.
- *
- * Return: The number of entries used.
- */
-static unsigned int xe_pt_stage_unbind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries)
-{
- struct xe_pt_stage_unbind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_unbind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- .modified_start = xe_vma_start(vma),
- .modified_end = xe_vma_end(vma),
- .wupd.entries = entries,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.wupd.num_used_entries;
-}
-
-static void
-xe_migrate_clear_pgtable_callback(struct xe_migrate_pt_update *pt_update,
- struct xe_tile *tile, struct iosys_map *map,
- void *ptr, u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_vma *vma = pt_update->vma;
- u64 empty = __xe_pt_empty_pte(tile, xe_vma_vm(vma), update->pt->level);
- int i;
-
- if (map && map->is_iomem)
- for (i = 0; i < num_qwords; ++i)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, empty);
- else if (map)
- memset64(map->vaddr + qword_ofs * sizeof(u64), empty,
- num_qwords);
- else
- memset64(ptr, empty, num_qwords);
-}
-
-static void
-xe_pt_commit_unbind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 num_entries,
- struct llist_head *deferred)
-{
- u32 j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (j = 0; j < num_entries; ++j) {
- struct xe_vm_pgtable_update *entry = &entries[j];
- struct xe_pt *pt = entry->pt;
-
- pt->num_live -= entry->qwords;
- if (pt->level) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
- u32 i;
-
- for (i = entry->ofs; i < entry->ofs + entry->qwords;
- i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[i] = NULL;
- }
- }
- }
-}
-
-<<<<<<<
-static const struct xe_migrate_pt_update_ops unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
-=======
-static void
-xe_pt_update_ops_rfence_interval(struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int i, level = 0;
- u64 start, last;
-
- for (i = 0; i < pt_op->num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &pt_op->entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- start = ALIGN_DOWN(xe_vma_start(vma), 0x1ull << xe_pt_shift(level));
- last = ALIGN(xe_vma_end(vma), 0x1ull << xe_pt_shift(level)) - 1;
-
- if (start < pt_update_ops->start)
- pt_update_ops->start = start;
- if (last > pt_update_ops->last)
- pt_update_ops->last = last;
-}
-
-static int vma_reserve_fences(struct xe_device *xe, struct xe_vma *vma)
-{
- int shift = xe_device_get_root_tile(xe)->media_gt ? 1 : 0;
-
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- return dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv,
- xe->info.tile_count << shift);
-
- return 0;
-}
-
-static int bind_op_prepare(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- pt_op->vma = NULL;
- pt_op->bind = true;
- pt_op->rebind = BIT(tile->id) & vma->tile_present;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- err = xe_pt_prepare_bind(tile, vma, pt_op->entries,
- &pt_op->num_entries);
- if (!err) {
- xe_tile_assert(tile, pt_op->num_entries <=
- ARRAY_SIZE(pt_op->entries));
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, true);
-
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. On LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * On !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!pt_op->rebind && xe_vm_has_scratch(vm) &&
- xe_vm_in_preempt_fence_mode(vm)))
- pt_update_ops->needs_invalidation = true;
- else if (pt_op->rebind && !xe_vm_in_lr_mode(vm))
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
-
- vma->tile_staged |= BIT(tile->id);
- pt_op->vma = vma;
- xe_pt_commit_prepare_bind(vma, pt_op->entries,
- pt_op->num_entries, pt_op->rebind);
- } else {
- xe_pt_cancel_bind(vma, pt_op->entries, pt_op->num_entries);
- }
-
- return err;
-}
-
-static int unbind_op_prepare(struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- if (!((vma->tile_present | vma->tile_staged) & BIT(tile->id)))
- return 0;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- /*
- * Wait for invalidation to complete. Can corrupt internal page table
- * state if an invalidation is running while preparing an unbind.
- */
- if (xe_vma_is_userptr(vma) && xe_vm_in_fault_mode(xe_vma_vm(vma)))
- mmu_interval_read_begin(&to_userptr_vma(vma)->userptr.notifier);
-
- pt_op->vma = vma;
- pt_op->bind = false;
- pt_op->rebind = false;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- pt_op->num_entries = xe_pt_stage_unbind(tile, vma, pt_op->entries);
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, false);
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
- pt_update_ops->needs_invalidation = true;
-
- xe_pt_commit_prepare_unbind(vma, pt_op->entries, pt_op->num_entries);
-
- return 0;
-}
-
-static int op_prepare(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op)
-{
- int err = 0;
-
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- err = bind_op_prepare(vm, tile, pt_update_ops, op->map.vma);
- pt_update_ops->wait_vm_kernel = true;
- break;
- case DRM_GPUVA_OP_REMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va));
-
- if (!err && op->remap.prev) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.prev);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- if (!err && op->remap.next) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.next);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- break;
- case DRM_GPUVA_OP_UNMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va));
- break;
- case DRM_GPUVA_OP_PREFETCH:
- err = bind_op_prepare(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va));
- pt_update_ops->wait_vm_kernel = true;
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-
- return err;
-}
-
-static void
-xe_pt_update_ops_init(struct xe_vm_pgtable_update_ops *pt_update_ops)
-{
- init_llist_head(&pt_update_ops->deferred);
- pt_update_ops->start = ~0x0ull;
- pt_update_ops->last = 0x0ull;
-}
-
-/**
- * xe_pt_update_ops_prepare() - Prepare PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Prepare PT update operations which includes updating internal PT state,
- * allocate memory for page tables, populate page table being pruned in, and
- * create PT update operations for leaf insertion / removal.
- *
- * Return: 0 on success, negative error code on error.
- */
-int xe_pt_update_ops_prepare(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct xe_vma_op *op;
- int shift = tile->media_gt ? 1 : 0;
- int err;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- xe_pt_update_ops_init(pt_update_ops);
-
- err = dma_resv_reserve_fences(xe_vm_resv(vops->vm),
- tile_to_xe(tile)->info.tile_count << shift);
- if (err)
- return err;
-
- list_for_each_entry(op, &vops->list, link) {
- err = op_prepare(vops->vm, tile, pt_update_ops, op);
-
- if (err)
- return err;
- }
-
- xe_tile_assert(tile, pt_update_ops->current_op <=
- pt_update_ops->num_ops);
-
-#ifdef TEST_VM_OPS_ERROR
- if (vops->inject_error &&
- vops->vm->xe->vm_inject_error_position == FORCE_OP_ERROR_PREPARE)
- return -ENOSPC;
-#endif
-
- return 0;
-}
-
-static void bind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present |= BIT(tile->id);
- vma->tile_staged &= ~BIT(tile->id);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- to_userptr_vma(vma)->userptr.initial_bind = true;
- }
-
- /*
- * Kick rebind worker if this bind triggers preempt fences and not in
- * the rebind worker
- */
- if (pt_update_ops->wait_vm_bookkeep &&
- xe_vm_in_preempt_fence_mode(vm) &&
- !current->mm)
- xe_vm_queue_rebind_worker(vm);
-}
-
-static void unbind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present &= ~BIT(tile->id);
- if (!vma->tile_present) {
- list_del_init(&vma->combined_links.rebind);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
-
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- }
-}
-
-static void op_commit(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- bind_op_commit(vm, tile, pt_update_ops, op->map.vma, fence,
- fence2);
- break;
- case DRM_GPUVA_OP_REMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va), fence,
- fence2);
-
- if (op->remap.prev)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.prev,
- fence, fence2);
- if (op->remap.next)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.next,
- fence, fence2);
- break;
- case DRM_GPUVA_OP_UNMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va), fence, fence2);
- break;
- case DRM_GPUVA_OP_PREFETCH:
- bind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va), fence, fence2);
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-}
-
-static const struct xe_migrate_pt_update_ops migrate_ops = {
- .populate = xe_vm_populate_pgtable,
- .clear = xe_migrate_clear_pgtable_callback,
->>>>>>>
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-/**
- * __xe_pt_unbind_vma() - Disconnect and free a page-table tree for the vma
- * address range.
- * @tile: The tile to unbind for.
- * @vma: The vma to unbind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before disconnecting the tree to be destroyed.
- * @num_syncs: Number of @sync entries.
- *
- * This function builds a the xe_vm_pgtable_update entries abstracting the
- * operations needed to detach the page-table tree to be destroyed from the
- * man vm tree.
- * It then takes the relevant locks and submits the operations for
- * pipelined detachment of the gpu page-table from the vm main tree,
- * (which can be done either by the cpu and the GPU), Finally it frees the
- * detached page-table tree.
- *
- * Return: A valid dma-fence representing the pipelined detachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_unbind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs)
-{
-<<<<<<<
- struct xe_vm *vm = vops->vm;
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL, *mfence = NULL;
- struct dma_fence **fences = NULL;
- struct dma_fence_array *cf = NULL;
- struct xe_range_fence *rfence;
- struct xe_vma_op *op;
- int err = 0, i;
- struct xe_migrate_pt_update update = {
- .ops = pt_update_ops->needs_userptr_lock ?
- &userptr_migrate_ops :
- &migrate_ops,
- .vops = vops,
- .tile_id = tile->id,
-=======
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update unbind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_unbind_ops :
- &unbind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
->>>>>>>
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence = NULL;
- struct invalidation_fence *ifence;
- struct xe_range_fence *rfence;
- int err;
-
- LLIST_HEAD(deferred);
-
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- num_entries = xe_pt_stage_unbind(tile, vma, entries);
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &unbind_pt_update, entries,
- num_entries);
-
-<<<<<<<
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- return ERR_PTR(err);
-
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
-=======
- if (pt_update_ops->needs_invalidation) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence) {
- err = -ENOMEM;
- goto kill_vm_tile1;
- }
- if (tile->media_gt) {
- mfence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!mfence) {
- err = -ENOMEM;
- goto free_ifence;
- }
- fences = kmalloc_array(2, sizeof(*fences), GFP_KERNEL);
- if (!fences) {
- err = -ENOMEM;
- goto free_ifence;
- }
- cf = dma_fence_array_alloc(2);
- if (!cf) {
- err = -ENOMEM;
- goto free_ifence;
- }
- }
- }
->>>>>>>
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- /*
- * Even if we were already evicted and unbind to destroy, we need to
- * clear again here. The eviction may have updated pagetables at a
- * lower level, because it needs to be more conservative.
- */
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, NULL, q ? q :
- vm->q[tile->id],
- entries, num_entries,
- syncs, num_syncs,
- &unbind_pt_update.base);
- if (!IS_ERR(fence)) {
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- unbind_pt_update.base.start,
- unbind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
-<<<<<<<
- /* TLB invalidation must be done before signaling unbind */
- err = invalidation_fence_init(tile->primary_gt, ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- DMA_RESV_USAGE_BOOKKEEP);
-
- /* This fence will be installed by caller when doing eviction */
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_unbind(vma, entries, num_entries,
- unbind_pt_update.locked ? &deferred : NULL);
- vma->tile_present &= ~BIT(tile->id);
- } else {
- kfree(rfence);
- kfree(ifence);
- }
-
- if (!vma->tile_present)
- list_del_init(&vma->combined_links.rebind);
-
- if (unbind_pt_update.locked) {
- xe_tile_assert(tile, xe_vma_is_userptr(vma));
-=======
- xe_pt_commit(pt_op->vma, pt_op->entries,
- pt_op->num_entries, &pt_update_ops->deferred);
- pt_op->vma = NULL; /* skip in xe_pt_update_ops_abort */
- }
-
- if (xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- pt_update_ops->start,
- pt_update_ops->last, fence))
- dma_fence_wait(fence, false);
-
- /* tlb invalidation must be done before signaling rebind */
- if (ifence) {
- if (mfence)
- dma_fence_get(fence);
- invalidation_fence_init(tile->primary_gt, ifence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- if (mfence) {
- invalidation_fence_init(tile->media_gt, mfence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- fences[0] = &ifence->base.base;
- fences[1] = &mfence->base.base;
- dma_fence_array_init(cf, 2, fences,
- vm->composite_fence_ctx,
- vm->composite_fence_seqno++,
- false);
- fence = &cf->base;
- } else {
- fence = &ifence->base.base;
- }
- }
-
- if (!mfence) {
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op, fence, NULL);
- } else {
- dma_resv_add_fence(xe_vm_resv(vm), &ifence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- dma_resv_add_fence(xe_vm_resv(vm), &mfence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op,
- &ifence->base.base, &mfence->base.base);
- }
->>>>>>>
-
- if (!vma->tile_present) {
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
-
- return fence;
-<<<<<<<
-=======
-
-free_rfence:
- kfree(rfence);
-free_ifence:
- kfree(cf);
- kfree(fences);
- kfree(mfence);
- kfree(ifence);
-kill_vm_tile1:
- if (err != -EAGAIN && tile->id)
- xe_vm_kill(vops->vm, false);
-
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_update_ops_fini() - Finish PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operations
- *
- * Finish PT update operations by committing to destroy page table memory
- */
-void xe_pt_update_ops_fini(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = 0; i < pt_update_ops->current_op; ++i) {
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[i];
-
- xe_pt_free_bind(pt_op->entries, pt_op->num_entries);
- }
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
-}
-
-/**
- * xe_pt_update_ops_abort() - Abort PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Abort PT update operations by unwinding internal PT state
- */
-void xe_pt_update_ops_abort(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = pt_update_ops->num_ops - 1; i >= 0; --i) {
- struct xe_vm_pgtable_update_op *pt_op =
- &pt_update_ops->ops[i];
-
- if (!pt_op->vma || i >= pt_update_ops->current_op)
- continue;
-
- if (pt_op->bind)
- xe_pt_abort_bind(pt_op->vma, pt_op->entries,
- pt_op->num_entries,
- pt_op->rebind);
- else
- xe_pt_abort_unbind(pt_op->vma, pt_op->entries,
- pt_op->num_entries);
- }
-
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
->>>>>>>
-}
diff --git a/rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.3 b/rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.3
deleted file mode 100644
index bb6eadffff5e..000000000000
--- a/rr-cache/4951c0e45d299a9570812ec9f1cc27e11aa21d6e/preimage.3
+++ /dev/null
@@ -1,2248 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2022 Intel Corporation
- */
-
-#include <linux/dma-fence-array.h>
-
-#include "xe_pt.h"
-
-#include "regs/xe_gtt_defs.h"
-#include "xe_bo.h"
-#include "xe_device.h"
-#include "xe_drm_client.h"
-#include "xe_gt.h"
-#include "xe_gt_tlb_invalidation.h"
-#include "xe_migrate.h"
-#include "xe_pt_types.h"
-#include "xe_pt_walk.h"
-#include "xe_res_cursor.h"
-#include "xe_trace.h"
-#include "xe_ttm_stolen_mgr.h"
-#include "xe_vm.h"
-
-struct xe_pt_dir {
- struct xe_pt pt;
- /** @children: Array of page-table child nodes */
- struct xe_ptw *children[XE_PDES];
-};
-
-#if IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM)
-#define xe_pt_set_addr(__xe_pt, __addr) ((__xe_pt)->addr = (__addr))
-#define xe_pt_addr(__xe_pt) ((__xe_pt)->addr)
-#else
-#define xe_pt_set_addr(__xe_pt, __addr)
-#define xe_pt_addr(__xe_pt) 0ull
-#endif
-
-static const u64 xe_normal_pt_shifts[] = {12, 21, 30, 39, 48};
-static const u64 xe_compact_pt_shifts[] = {16, 21, 30, 39, 48};
-
-#define XE_PT_HIGHEST_LEVEL (ARRAY_SIZE(xe_normal_pt_shifts) - 1)
-
-static struct xe_pt_dir *as_xe_pt_dir(struct xe_pt *pt)
-{
- return container_of(pt, struct xe_pt_dir, pt);
-}
-
-static struct xe_pt *xe_pt_entry(struct xe_pt_dir *pt_dir, unsigned int index)
-{
- return container_of(pt_dir->children[index], struct xe_pt, base);
-}
-
-static u64 __xe_pt_empty_pte(struct xe_tile *tile, struct xe_vm *vm,
- unsigned int level)
-{
- struct xe_device *xe = tile_to_xe(tile);
- u16 pat_index = xe->pat.idx[XE_CACHE_WB];
- u8 id = tile->id;
-
- if (!xe_vm_has_scratch(vm))
- return 0;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return vm->pt_ops->pde_encode_bo(vm->scratch_pt[id][level - 1]->bo,
- 0, pat_index);
-
- return vm->pt_ops->pte_encode_addr(xe, 0, pat_index, level, IS_DGFX(xe), 0) |
- XE_PTE_NULL;
-}
-
-static void xe_pt_free(struct xe_pt *pt)
-{
- if (pt->level)
- kfree(as_xe_pt_dir(pt));
- else
- kfree(pt);
-}
-
-/**
- * xe_pt_create() - Create a page-table.
- * @vm: The vm to create for.
- * @tile: The tile to create for.
- * @level: The page-table level.
- *
- * Allocate and initialize a single struct xe_pt metadata structure. Also
- * create the corresponding page-table bo, but don't initialize it. If the
- * level is grater than zero, then it's assumed to be a directory page-
- * table and the directory structure is also allocated and initialized to
- * NULL pointers.
- *
- * Return: A valid struct xe_pt pointer on success, Pointer error code on
- * error.
- */
-struct xe_pt *xe_pt_create(struct xe_vm *vm, struct xe_tile *tile,
- unsigned int level)
-{
- struct xe_pt *pt;
- struct xe_bo *bo;
- int err;
-
- if (level) {
- struct xe_pt_dir *dir = kzalloc(sizeof(*dir), GFP_KERNEL);
-
- pt = (dir) ? &dir->pt : NULL;
- } else {
- pt = kzalloc(sizeof(*pt), GFP_KERNEL);
- }
- if (!pt)
- return ERR_PTR(-ENOMEM);
-
- pt->level = level;
- bo = xe_bo_create_pin_map(vm->xe, tile, vm, SZ_4K,
- ttm_bo_type_kernel,
- XE_BO_FLAG_VRAM_IF_DGFX(tile) |
- XE_BO_FLAG_IGNORE_MIN_PAGE_SIZE |
- XE_BO_FLAG_PINNED |
- XE_BO_FLAG_NO_RESV_EVICT |
- XE_BO_FLAG_PAGETABLE);
- if (IS_ERR(bo)) {
- err = PTR_ERR(bo);
- goto err_kfree;
- }
- pt->bo = bo;
- pt->base.children = level ? as_xe_pt_dir(pt)->children : NULL;
-
- if (vm->xef)
- xe_drm_client_add_bo(vm->xef->client, pt->bo);
- xe_tile_assert(tile, level <= XE_VM_MAX_LEVEL);
-
- return pt;
-
-err_kfree:
- xe_pt_free(pt);
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_populate_empty() - Populate a page-table bo with scratch- or zero
- * entries.
- * @tile: The tile the scratch pagetable of which to use.
- * @vm: The vm we populate for.
- * @pt: The pagetable the bo of which to initialize.
- *
- * Populate the page-table bo of @pt with entries pointing into the tile's
- * scratch page-table tree if any. Otherwise populate with zeros.
- */
-void xe_pt_populate_empty(struct xe_tile *tile, struct xe_vm *vm,
- struct xe_pt *pt)
-{
- struct iosys_map *map = &pt->bo->vmap;
- u64 empty;
- int i;
-
- if (!xe_vm_has_scratch(vm)) {
- /*
- * FIXME: Some memory is allocated already allocated to zero?
- * Find out which memory that is and avoid this memset...
- */
- xe_map_memset(vm->xe, map, 0, 0, SZ_4K);
- } else {
- empty = __xe_pt_empty_pte(tile, vm, pt->level);
- for (i = 0; i < XE_PDES; i++)
- xe_pt_write(vm->xe, map, i, empty);
- }
-}
-
-/**
- * xe_pt_shift() - Return the ilog2 value of the size of the address range of
- * a page-table at a certain level.
- * @level: The level.
- *
- * Return: The ilog2 value of the size of the address range of a page-table
- * at level @level.
- */
-unsigned int xe_pt_shift(unsigned int level)
-{
- return XE_PTE_SHIFT + XE_PDE_SHIFT * level;
-}
-
-/**
- * xe_pt_destroy() - Destroy a page-table tree.
- * @pt: The root of the page-table tree to destroy.
- * @flags: vm flags. Currently unused.
- * @deferred: List head of lockless list for deferred putting. NULL for
- * immediate putting.
- *
- * Puts the page-table bo, recursively calls xe_pt_destroy on all children
- * and finally frees @pt. TODO: Can we remove the @flags argument?
- */
-void xe_pt_destroy(struct xe_pt *pt, u32 flags, struct llist_head *deferred)
-{
- int i;
-
- if (!pt)
- return;
-
- XE_WARN_ON(!list_empty(&pt->bo->ttm.base.gpuva.list));
- xe_bo_unpin(pt->bo);
- xe_bo_put_deferred(pt->bo, deferred);
-
- if (pt->level > 0 && pt->num_live) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
-
- for (i = 0; i < XE_PDES; i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i), flags,
- deferred);
- }
- }
- xe_pt_free(pt);
-}
-
-/**
- * DOC: Pagetable building
- *
- * Below we use the term "page-table" for both page-directories, containing
- * pointers to lower level page-directories or page-tables, and level 0
- * page-tables that contain only page-table-entries pointing to memory pages.
- *
- * When inserting an address range in an already existing page-table tree
- * there will typically be a set of page-tables that are shared with other
- * address ranges, and a set that are private to this address range.
- * The set of shared page-tables can be at most two per level,
- * and those can't be updated immediately because the entries of those
- * page-tables may still be in use by the gpu for other mappings. Therefore
- * when inserting entries into those, we instead stage those insertions by
- * adding insertion data into struct xe_vm_pgtable_update structures. This
- * data, (subtrees for the cpu and page-table-entries for the gpu) is then
- * added in a separate commit step. CPU-data is committed while still under the
- * vm lock, the object lock and for userptr, the notifier lock in read mode.
- * The GPU async data is committed either by the GPU or CPU after fulfilling
- * relevant dependencies.
- * For non-shared page-tables (and, in fact, for shared ones that aren't
- * existing at the time of staging), we add the data in-place without the
- * special update structures. This private part of the page-table tree will
- * remain disconnected from the vm page-table tree until data is committed to
- * the shared page tables of the vm tree in the commit phase.
- */
-
-struct xe_pt_update {
- /** @update: The update structure we're building for this parent. */
- struct xe_vm_pgtable_update *update;
- /** @parent: The parent. Used to detect a parent change. */
- struct xe_pt *parent;
- /** @preexisting: Whether the parent was pre-existing or allocated */
- bool preexisting;
-};
-
-struct xe_pt_stage_bind_walk {
- /** base: The base class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @vm: The vm we're building for. */
- struct xe_vm *vm;
- /** @tile: The tile we're building for. */
- struct xe_tile *tile;
- /** @default_pte: PTE flag only template. No address is associated */
- u64 default_pte;
- /** @dma_offset: DMA offset to add to the PTE. */
- u64 dma_offset;
- /**
- * @needs_64k: This address range enforces 64K alignment and
- * granularity.
- */
- bool needs_64K;
- /**
- * @vma: VMA being mapped
- */
- struct xe_vma *vma;
-
- /* Also input, but is updated during the walk*/
- /** @curs: The DMA address cursor. */
- struct xe_res_cursor *curs;
- /** @va_curs_start: The Virtual address coresponding to @curs->start */
- u64 va_curs_start;
-
- /* Output */
- struct xe_walk_update {
- /** @wupd.entries: Caller provided storage. */
- struct xe_vm_pgtable_update *entries;
- /** @wupd.num_used_entries: Number of update @entries used. */
- unsigned int num_used_entries;
- /** @wupd.updates: Tracks the update entry at a given level */
- struct xe_pt_update updates[XE_VM_MAX_LEVEL + 1];
- } wupd;
-
- /* Walk state */
- /**
- * @l0_end_addr: The end address of the current l0 leaf. Used for
- * 64K granularity detection.
- */
- u64 l0_end_addr;
- /** @addr_64K: The start address of the current 64K chunk. */
- u64 addr_64K;
- /** @found_64: Whether @add_64K actually points to a 64K chunk. */
- bool found_64K;
-};
-
-static int
-xe_pt_new_shared(struct xe_walk_update *wupd, struct xe_pt *parent,
- pgoff_t offset, bool alloc_entries)
-{
- struct xe_pt_update *upd = &wupd->updates[parent->level];
- struct xe_vm_pgtable_update *entry;
-
- /*
- * For *each level*, we could only have one active
- * struct xt_pt_update at any one time. Once we move on to a
- * new parent and page-directory, the old one is complete, and
- * updates are either already stored in the build tree or in
- * @wupd->entries
- */
- if (likely(upd->parent == parent))
- return 0;
-
- upd->parent = parent;
- upd->preexisting = true;
-
- if (wupd->num_used_entries == XE_VM_MAX_LEVEL * 2 + 1)
- return -EINVAL;
-
- entry = wupd->entries + wupd->num_used_entries++;
- upd->update = entry;
- entry->ofs = offset;
- entry->pt_bo = parent->bo;
- entry->pt = parent;
- entry->flags = 0;
- entry->qwords = 0;
-
- if (alloc_entries) {
- entry->pt_entries = kmalloc_array(XE_PDES,
- sizeof(*entry->pt_entries),
- GFP_KERNEL);
- if (!entry->pt_entries)
- return -ENOMEM;
- }
-
- return 0;
-}
-
-/*
- * NOTE: This is a very frequently called function so we allow ourselves
- * to annotate (using branch prediction hints) the fastpath of updating a
- * non-pre-existing pagetable with leaf ptes.
- */
-static int
-xe_pt_insert_entry(struct xe_pt_stage_bind_walk *xe_walk, struct xe_pt *parent,
- pgoff_t offset, struct xe_pt *xe_child, u64 pte)
-{
- struct xe_pt_update *upd = &xe_walk->wupd.updates[parent->level];
- struct xe_pt_update *child_upd = xe_child ?
- &xe_walk->wupd.updates[xe_child->level] : NULL;
- int ret;
-
- ret = xe_pt_new_shared(&xe_walk->wupd, parent, offset, true);
- if (unlikely(ret))
- return ret;
-
- /*
- * Register this new pagetable so that it won't be recognized as
- * a shared pagetable by a subsequent insertion.
- */
- if (unlikely(child_upd)) {
- child_upd->update = NULL;
- child_upd->parent = xe_child;
- child_upd->preexisting = false;
- }
-
- if (likely(!upd->preexisting)) {
- /* Continue building a non-connected subtree. */
- struct iosys_map *map = &parent->bo->vmap;
-
- if (unlikely(xe_child))
- parent->base.children[offset] = &xe_child->base;
-
- xe_pt_write(xe_walk->vm->xe, map, offset, pte);
- parent->num_live++;
- } else {
- /* Shared pt. Stage update. */
- unsigned int idx;
- struct xe_vm_pgtable_update *entry = upd->update;
-
- idx = offset - entry->ofs;
- entry->pt_entries[idx].pt = xe_child;
- entry->pt_entries[idx].pte = pte;
- entry->qwords++;
- }
-
- return 0;
-}
-
-static bool xe_pt_hugepte_possible(u64 addr, u64 next, unsigned int level,
- struct xe_pt_stage_bind_walk *xe_walk)
-{
- u64 size, dma;
-
- if (level > MAX_HUGEPTE_LEVEL)
- return false;
-
- /* Does the virtual range requested cover a huge pte? */
- if (!xe_pt_covers(addr, next, level, &xe_walk->base))
- return false;
-
- /* Does the DMA segment cover the whole pte? */
- if (next - xe_walk->va_curs_start > xe_walk->curs->size)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- /* Is the DMA address huge PTE size aligned? */
- size = next - addr;
- dma = addr - xe_walk->va_curs_start + xe_res_dma(xe_walk->curs);
-
- return IS_ALIGNED(dma, size);
-}
-
-/*
- * Scan the requested mapping to check whether it can be done entirely
- * with 64K PTEs.
- */
-static bool
-xe_pt_scan_64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- struct xe_res_cursor curs = *xe_walk->curs;
-
- if (!IS_ALIGNED(addr, SZ_64K))
- return false;
-
- if (next > xe_walk->l0_end_addr)
- return false;
-
- /* null VMA's do not have dma addresses */
- if (xe_vma_is_null(xe_walk->vma))
- return true;
-
- xe_res_next(&curs, addr - xe_walk->va_curs_start);
- for (; addr < next; addr += SZ_64K) {
- if (!IS_ALIGNED(xe_res_dma(&curs), SZ_64K) || curs.size < SZ_64K)
- return false;
-
- xe_res_next(&curs, SZ_64K);
- }
-
- return addr == next;
-}
-
-/*
- * For non-compact "normal" 4K level-0 pagetables, we want to try to group
- * addresses together in 64K-contigous regions to add a 64K TLB hint for the
- * device to the PTE.
- * This function determines whether the address is part of such a
- * segment. For VRAM in normal pagetables, this is strictly necessary on
- * some devices.
- */
-static bool
-xe_pt_is_pte_ps64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
-{
- /* Address is within an already found 64k region */
- if (xe_walk->found_64K && addr - xe_walk->addr_64K < SZ_64K)
- return true;
-
- xe_walk->found_64K = xe_pt_scan_64K(addr, addr + SZ_64K, xe_walk);
- xe_walk->addr_64K = addr;
-
- return xe_walk->found_64K;
-}
-
-static int
-xe_pt_stage_bind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_bind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- u16 pat_index = xe_walk->vma->pat_index;
- struct xe_pt *xe_parent = container_of(parent, typeof(*xe_parent), base);
- struct xe_vm *vm = xe_walk->vm;
- struct xe_pt *xe_child;
- bool covers;
- int ret = 0;
- u64 pte;
-
- /* Is this a leaf entry ?*/
- if (level == 0 || xe_pt_hugepte_possible(addr, next, level, xe_walk)) {
- struct xe_res_cursor *curs = xe_walk->curs;
- bool is_null = xe_vma_is_null(xe_walk->vma);
-
- XE_WARN_ON(xe_walk->va_curs_start != addr);
-
- pte = vm->pt_ops->pte_encode_vma(is_null ? 0 :
- xe_res_dma(curs) + xe_walk->dma_offset,
- xe_walk->vma, pat_index, level);
- pte |= xe_walk->default_pte;
-
- /*
- * Set the XE_PTE_PS64 hint if possible, otherwise if
- * this device *requires* 64K PTE size for VRAM, fail.
- */
- if (level == 0 && !xe_parent->is_compact) {
- if (xe_pt_is_pte_ps64K(addr, next, xe_walk)) {
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_64K;
- pte |= XE_PTE_PS64;
- } else if (XE_WARN_ON(xe_walk->needs_64K)) {
- return -EINVAL;
- }
- }
-
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, NULL, pte);
- if (unlikely(ret))
- return ret;
-
- if (!is_null)
- xe_res_next(curs, next - addr);
- xe_walk->va_curs_start = next;
- xe_walk->vma->gpuva.flags |= (XE_VMA_PTE_4K << level);
- *action = ACTION_CONTINUE;
-
- return ret;
- }
-
- /*
- * Descending to lower level. Determine if we need to allocate a
- * new page table or -directory, which we do if there is no
- * previous one or there is one we can completely replace.
- */
- if (level == 1) {
- walk->shifts = xe_normal_pt_shifts;
- xe_walk->l0_end_addr = next;
- }
-
- covers = xe_pt_covers(addr, next, level, &xe_walk->base);
- if (covers || !*child) {
- u64 flags = 0;
-
- xe_child = xe_pt_create(xe_walk->vm, xe_walk->tile, level - 1);
- if (IS_ERR(xe_child))
- return PTR_ERR(xe_child);
-
- xe_pt_set_addr(xe_child,
- round_down(addr, 1ull << walk->shifts[level]));
-
- if (!covers)
- xe_pt_populate_empty(xe_walk->tile, xe_walk->vm, xe_child);
-
- *child = &xe_child->base;
-
- /*
- * Prefer the compact pagetable layout for L0 if possible. Only
- * possible if VMA covers entire 2MB region as compact 64k and
- * 4k pages cannot be mixed within a 2MB region.
- * TODO: Suballocate the pt bo to avoid wasting a lot of
- * memory.
- */
- if (GRAPHICS_VERx100(tile_to_xe(xe_walk->tile)) >= 1250 && level == 1 &&
- covers && xe_pt_scan_64K(addr, next, xe_walk)) {
- walk->shifts = xe_compact_pt_shifts;
- xe_walk->vma->gpuva.flags |= XE_VMA_PTE_COMPACT;
- flags |= XE_PDE_64K;
- xe_child->is_compact = true;
- }
-
- pte = vm->pt_ops->pde_encode_bo(xe_child->bo, 0, pat_index) | flags;
- ret = xe_pt_insert_entry(xe_walk, xe_parent, offset, xe_child,
- pte);
- }
-
- *action = ACTION_SUBTREE;
- return ret;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_bind_ops = {
- .pt_entry = xe_pt_stage_bind_entry,
-};
-
-/**
- * xe_pt_stage_bind() - Build a disconnected page-table tree for a given address
- * range.
- * @tile: The tile we're building for.
- * @vma: The vma indicating the address range.
- * @entries: Storage for the update entries used for connecting the tree to
- * the main tree at commit time.
- * @num_entries: On output contains the number of @entries used.
- *
- * This function builds a disconnected page-table tree for a given address
- * range. The tree is connected to the main vm tree for the gpu using
- * xe_migrate_update_pgtables() and for the cpu using xe_pt_commit_bind().
- * The function builds xe_vm_pgtable_update structures for already existing
- * shared page-tables, and non-existing shared and non-shared page-tables
- * are built and populated directly.
- *
- * Return 0 on success, negative error code on error.
- */
-static int
-xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- struct xe_device *xe = tile_to_xe(tile);
- struct xe_bo *bo = xe_vma_bo(vma);
- bool is_devmem = !xe_vma_is_userptr(vma) && bo &&
- (xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo));
- struct xe_res_cursor curs;
- struct xe_pt_stage_bind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_bind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .vm = xe_vma_vm(vma),
- .tile = tile,
- .curs = &curs,
- .va_curs_start = xe_vma_start(vma),
- .vma = vma,
- .wupd.entries = entries,
- .needs_64K = (xe_vma_vm(vma)->flags & XE_VM_FLAG_64K) && is_devmem,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- int ret;
-
- /**
- * Default atomic expectations for different allocation scenarios are as follows:
- *
- * 1. Traditional API: When the VM is not in LR mode:
- * - Device atomics are expected to function with all allocations.
- *
- * 2. Compute/SVM API: When the VM is in LR mode:
- * - Device atomics are the default behavior when the bo is placed in a single region.
- * - In all other cases device atomics will be disabled with AE=0 until an application
- * request differently using a ioctl like madvise.
- */
- if (vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT) {
- if (xe_vm_in_lr_mode(xe_vma_vm(vma))) {
- if (bo && xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- /**
- * If a SMEM+LMEM allocation is backed by SMEM, a device
- * atomics will cause a gpu page fault and which then
- * gets migrated to LMEM, bind such allocations with
- * device atomics enabled.
- */
- else if (is_devmem && !xe_bo_has_single_placement(bo))
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- } else {
- xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
- }
-
- /**
- * Unset AE if the platform(PVC) doesn't support it on an
- * allocation
- */
- if (!xe->info.has_device_atomics_on_smem && !is_devmem)
- xe_walk.default_pte &= ~XE_USM_PPGTT_PTE_AE;
- }
-
- if (is_devmem) {
- xe_walk.default_pte |= XE_PPGTT_PTE_DM;
- xe_walk.dma_offset = vram_region_gpu_offset(bo->ttm.resource);
- }
-
- if (!xe_vma_has_no_bo(vma) && xe_bo_is_stolen(bo))
- xe_walk.dma_offset = xe_ttm_stolen_gpu_offset(xe_bo_device(bo));
-
- xe_bo_assert_held(bo);
-
- if (!xe_vma_is_null(vma)) {
- if (xe_vma_is_userptr(vma))
- xe_res_first_sg(to_userptr_vma(vma)->userptr.sg, 0,
- xe_vma_size(vma), &curs);
- else if (xe_bo_is_vram(bo) || xe_bo_is_stolen(bo))
- xe_res_first(bo->ttm.resource, xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- else
- xe_res_first_sg(xe_bo_sg(bo), xe_vma_bo_offset(vma),
- xe_vma_size(vma), &curs);
- } else {
- curs.size = xe_vma_size(vma);
- }
-
- ret = xe_pt_walk_range(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- *num_entries = xe_walk.wupd.num_used_entries;
- return ret;
-}
-
-/**
- * xe_pt_nonshared_offsets() - Determine the non-shared entry offsets of a
- * shared pagetable.
- * @addr: The start address within the non-shared pagetable.
- * @end: The end address within the non-shared pagetable.
- * @level: The level of the non-shared pagetable.
- * @walk: Walk info. The function adjusts the walk action.
- * @action: next action to perform (see enum page_walk_action)
- * @offset: Ignored on input, First non-shared entry on output.
- * @end_offset: Ignored on input, Last non-shared entry + 1 on output.
- *
- * A non-shared page-table has some entries that belong to the address range
- * and others that don't. This function determines the entries that belong
- * fully to the address range. Depending on level, some entries may
- * partially belong to the address range (that can't happen at level 0).
- * The function detects that and adjust those offsets to not include those
- * partial entries. Iff it does detect partial entries, we know that there must
- * be shared page tables also at lower levels, so it adjusts the walk action
- * accordingly.
- *
- * Return: true if there were non-shared entries, false otherwise.
- */
-static bool xe_pt_nonshared_offsets(u64 addr, u64 end, unsigned int level,
- struct xe_pt_walk *walk,
- enum page_walk_action *action,
- pgoff_t *offset, pgoff_t *end_offset)
-{
- u64 size = 1ull << walk->shifts[level];
-
- *offset = xe_pt_offset(addr, level, walk);
- *end_offset = xe_pt_num_entries(addr, end, level, walk) + *offset;
-
- if (!level)
- return true;
-
- /*
- * If addr or next are not size aligned, there are shared pts at lower
- * level, so in that case traverse down the subtree
- */
- *action = ACTION_CONTINUE;
- if (!IS_ALIGNED(addr, size)) {
- *action = ACTION_SUBTREE;
- (*offset)++;
- }
-
- if (!IS_ALIGNED(end, size)) {
- *action = ACTION_SUBTREE;
- (*end_offset)--;
- }
-
- return *end_offset > *offset;
-}
-
-struct xe_pt_zap_ptes_walk {
- /** @base: The walk base-class */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're building for */
- struct xe_tile *tile;
-
- /* Output */
- /** @needs_invalidate: Whether we need to invalidate TLB*/
- bool needs_invalidate;
-};
-
-static int xe_pt_zap_ptes_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_zap_ptes_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- /*
- * Note that we're called from an entry callback, and we're dealing
- * with the child of that entry rather than the parent, so need to
- * adjust level down.
- */
- if (xe_pt_nonshared_offsets(addr, next, --level, walk, action, &offset,
- &end_offset)) {
- xe_map_memset(tile_to_xe(xe_walk->tile), &xe_child->bo->vmap,
- offset * sizeof(u64), 0,
- (end_offset - offset) * sizeof(u64));
- xe_walk->needs_invalidate = true;
- }
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_zap_ptes_ops = {
- .pt_entry = xe_pt_zap_ptes_entry,
-};
-
-/**
- * xe_pt_zap_ptes() - Zap (zero) gpu ptes of an address range
- * @tile: The tile we're zapping for.
- * @vma: GPU VMA detailing address range.
- *
- * Eviction and Userptr invalidation needs to be able to zap the
- * gpu ptes of a given address range in pagefaulting mode.
- * In order to be able to do that, that function needs access to the shared
- * page-table entrieaso it can either clear the leaf PTEs or
- * clear the pointers to lower-level page-tables. The caller is required
- * to hold the necessary locks to ensure neither the page-table connectivity
- * nor the page-table entries of the range is updated from under us.
- *
- * Return: Whether ptes were actually updated and a TLB invalidation is
- * required.
- */
-bool xe_pt_zap_ptes(struct xe_tile *tile, struct xe_vma *vma)
-{
- struct xe_pt_zap_ptes_walk xe_walk = {
- .base = {
- .ops = &xe_pt_zap_ptes_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
- u8 pt_mask = (vma->tile_present & ~vma->tile_invalidated);
-
- if (!(pt_mask & BIT(tile->id)))
- return false;
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.needs_invalidate;
-}
-
-static void
-xe_vm_populate_pgtable(struct xe_migrate_pt_update *pt_update, struct xe_tile *tile,
- struct iosys_map *map, void *data,
- u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_pt_entry *ptes = update->pt_entries;
- u64 *ptr = data;
- u32 i;
-
- for (i = 0; i < num_qwords; i++) {
- if (map)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, ptes[i].pte);
- else
- ptr[i] = ptes[i].pte;
- }
-}
-
-static void xe_pt_abort_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- u32 i, j;
-
- for (i = 0; i < num_entries; i++) {
- if (!entries[i].pt_entries)
- continue;
-
- for (j = 0; j < entries[i].qwords; j++)
- xe_pt_destroy(entries[i].pt_entries[j].pt, xe_vma_vm(vma)->flags, NULL);
- kfree(entries[i].pt_entries);
- }
-}
-
-static void xe_pt_commit_locks_assert(struct xe_vma *vma)
-{
- struct xe_vm *vm = xe_vma_vm(vma);
-
- lockdep_assert_held(&vm->lock);
-
- if (xe_vma_is_userptr(vma))
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- else if (!xe_vma_is_null(vma))
- dma_resv_assert_held(xe_vma_bo(vma)->ttm.base.resv);
-
- xe_vm_assert_held(vm);
-}
-
-static void xe_pt_commit_bind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries, bool rebind,
- struct llist_head *deferred)
-{
- u32 i, j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (i = 0; i < num_entries; i++) {
- struct xe_pt *pt = entries[i].pt;
- struct xe_pt_dir *pt_dir;
-
- if (!rebind)
- pt->num_live += entries[i].qwords;
-
- if (!pt->level) {
- kfree(entries[i].pt_entries);
- continue;
- }
-
- pt_dir = as_xe_pt_dir(pt);
- for (j = 0; j < entries[i].qwords; j++) {
- u32 j_ = j + entries[i].ofs;
- struct xe_pt *newpte = entries[i].pt_entries[j].pt;
-
- if (xe_pt_entry(pt_dir, j_))
- xe_pt_destroy(xe_pt_entry(pt_dir, j_),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[j_] = &newpte->base;
- }
- kfree(entries[i].pt_entries);
- }
-}
-
-static int
-xe_pt_prepare_bind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 *num_entries)
-{
- int err;
-
- *num_entries = 0;
- err = xe_pt_stage_bind(tile, vma, entries, num_entries);
- if (!err)
- xe_tile_assert(tile, *num_entries);
- else /* abort! */
- xe_pt_abort_bind(vma, entries, *num_entries);
-
- return err;
-}
-
-static void xe_vm_dbg_print_entries(struct xe_device *xe,
- const struct xe_vm_pgtable_update *entries,
- unsigned int num_entries)
-#if (IS_ENABLED(CONFIG_DRM_XE_DEBUG_VM))
-{
- unsigned int i;
-
- vm_dbg(&xe->drm, "%u entries to update\n", num_entries);
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
- struct xe_pt *xe_pt = entry->pt;
- u64 page_size = 1ull << xe_pt_shift(xe_pt->level);
- u64 end;
- u64 start;
-
- xe_assert(xe, !entry->pt->is_compact);
- start = entry->ofs * page_size;
- end = start + page_size * entry->qwords;
- vm_dbg(&xe->drm,
- "\t%u: Update level %u at (%u + %u) [%llx...%llx) f:%x\n",
- i, xe_pt->level, entry->ofs, entry->qwords,
- xe_pt_addr(xe_pt) + start, xe_pt_addr(xe_pt) + end, 0);
- }
-}
-#else
-{}
-#endif
-
-#ifdef CONFIG_DRM_XE_USERPTR_INVAL_INJECT
-
-static int xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- u32 divisor = uvma->userptr.divisor ? uvma->userptr.divisor : 2;
- static u32 count;
-
- if (count++ % divisor == divisor - 1) {
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
-
- uvma->userptr.divisor = divisor << 1;
- spin_lock(&vm->userptr.invalidated_lock);
- list_move_tail(&uvma->userptr.invalidate_link,
- &vm->userptr.invalidated);
- spin_unlock(&vm->userptr.invalidated_lock);
- return true;
- }
-
- return false;
-}
-
-#else
-
-static bool xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
-{
- return false;
-}
-
-#endif
-
-/**
- * struct xe_pt_migrate_pt_update - Callback argument for pre-commit callbacks
- * @base: Base we derive from.
- * @bind: Whether this is a bind or an unbind operation. A bind operation
- * makes the pre-commit callback error with -EAGAIN if it detects a
- * pending invalidation.
- * @locked: Whether the pre-commit callback locked the userptr notifier lock
- * and it needs unlocking.
- */
-struct xe_pt_migrate_pt_update {
- struct xe_migrate_pt_update base;
- bool bind;
- bool locked;
-};
-
-/*
- * This function adds the needed dependencies to a page-table update job
- * to make sure racing jobs for separate bind engines don't race writing
- * to the same page-table range, wreaking havoc. Initially use a single
- * fence for the entire VM. An optimization would use smaller granularity.
- */
-static int xe_pt_vm_dependencies(struct xe_sched_job *job,
- struct xe_range_fence_tree *rftree,
- u64 start, u64 last)
-{
- struct xe_range_fence *rtfence;
- struct dma_fence *fence;
- int err;
-
- rtfence = xe_range_fence_tree_first(rftree, start, last);
- while (rtfence) {
- fence = rtfence->fence;
-
- if (!dma_fence_is_signaled(fence)) {
- /*
- * Is this a CPU update? GPU is busy updating, so return
- * an error
- */
- if (!job)
- return -ETIME;
-
- dma_fence_get(fence);
- err = drm_sched_job_add_dependency(&job->drm, fence);
- if (err)
- return err;
- }
-
- rtfence = xe_range_fence_tree_next(rtfence, start, last);
- }
-
- return 0;
-}
-
-static int xe_pt_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_range_fence_tree *rftree =
- &xe_vma_vm(pt_update->vma)->rftree[pt_update->tile_id];
-
- return xe_pt_vm_dependencies(pt_update->job, rftree,
- pt_update->start, pt_update->last);
-}
-
-static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
-{
- struct xe_pt_migrate_pt_update *userptr_update =
- container_of(pt_update, typeof(*userptr_update), base);
- struct xe_userptr_vma *uvma = to_userptr_vma(pt_update->vma);
- unsigned long notifier_seq = uvma->userptr.notifier_seq;
- struct xe_vm *vm = xe_vma_vm(&uvma->vma);
- int err = xe_pt_vm_dependencies(pt_update->job,
- &vm->rftree[pt_update->tile_id],
- pt_update->start,
- pt_update->last);
-
- if (err)
- return err;
-
- userptr_update->locked = false;
-
- /*
- * Wait until nobody is running the invalidation notifier, and
- * since we're exiting the loop holding the notifier lock,
- * nobody can proceed invalidating either.
- *
- * Note that we don't update the vma->userptr.notifier_seq since
- * we don't update the userptr pages.
- */
- do {
- down_read(&vm->userptr.notifier_lock);
- if (!mmu_interval_read_retry(&uvma->userptr.notifier,
- notifier_seq))
- break;
-
- up_read(&vm->userptr.notifier_lock);
-
- if (userptr_update->bind)
- return -EAGAIN;
-
- notifier_seq = mmu_interval_read_begin(&uvma->userptr.notifier);
- } while (true);
-
- /* Inject errors to test_whether they are handled correctly */
- if (userptr_update->bind && xe_pt_userptr_inject_eagain(uvma)) {
- up_read(&vm->userptr.notifier_lock);
- return -EAGAIN;
- }
-
- userptr_update->locked = true;
-
- return 0;
-}
-
-static const struct xe_migrate_pt_update_ops bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_bind_ops = {
- .populate = xe_vm_populate_pgtable,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-struct invalidation_fence {
- struct xe_gt_tlb_invalidation_fence base;
- struct xe_gt *gt;
- struct dma_fence *fence;
- struct dma_fence_cb cb;
- struct work_struct work;
- u64 start;
- u64 end;
- u32 asid;
-};
-
-static void invalidation_fence_cb(struct dma_fence *fence,
- struct dma_fence_cb *cb)
-{
- struct invalidation_fence *ifence =
- container_of(cb, struct invalidation_fence, cb);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_cb(xe, &ifence->base);
- if (!ifence->fence->error) {
- queue_work(system_wq, &ifence->work);
- } else {
- ifence->base.base.error = ifence->fence->error;
- dma_fence_signal(&ifence->base.base);
- dma_fence_put(&ifence->base.base);
- }
- dma_fence_put(ifence->fence);
-}
-
-static void invalidation_fence_work_func(struct work_struct *w)
-{
- struct invalidation_fence *ifence =
- container_of(w, struct invalidation_fence, work);
- struct xe_device *xe = gt_to_xe(ifence->gt);
-
- trace_xe_gt_tlb_invalidation_fence_work_func(xe, &ifence->base);
- xe_gt_tlb_invalidation_range(ifence->gt, &ifence->base, ifence->start,
- ifence->end, ifence->asid);
-}
-
-static int invalidation_fence_init(struct xe_gt *gt,
- struct invalidation_fence *ifence,
- struct dma_fence *fence,
- u64 start, u64 end, u32 asid)
-{
- int ret;
-
- trace_xe_gt_tlb_invalidation_fence_create(gt_to_xe(gt), &ifence->base);
-
- xe_gt_tlb_invalidation_fence_init(gt, &ifence->base, false);
-
- ifence->fence = fence;
- ifence->gt = gt;
- ifence->start = start;
- ifence->end = end;
- ifence->asid = asid;
-
- INIT_WORK(&ifence->work, invalidation_fence_work_func);
- ret = dma_fence_add_callback(fence, &ifence->cb, invalidation_fence_cb);
- if (ret == -ENOENT) {
- dma_fence_put(ifence->fence); /* Usually dropped in CB */
- invalidation_fence_work_func(&ifence->work);
- } else if (ret) {
- dma_fence_put(&ifence->base.base); /* Caller ref */
- dma_fence_put(&ifence->base.base); /* Creation ref */
- }
-
- xe_gt_assert(gt, !ret || ret == -ENOENT);
-
- return ret && ret != -ENOENT ? ret : 0;
-}
-
-static void xe_pt_calc_rfence_interval(struct xe_vma *vma,
- struct xe_pt_migrate_pt_update *update,
- struct xe_vm_pgtable_update *entries,
- u32 num_entries)
-{
- int i, level = 0;
-
- for (i = 0; i < num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- update->base.start = ALIGN_DOWN(xe_vma_start(vma),
- 0x1ull << xe_pt_shift(level));
- update->base.last = ALIGN(xe_vma_end(vma),
- 0x1ull << xe_pt_shift(level)) - 1;
-}
-
-/**
- * __xe_pt_bind_vma() - Build and connect a page-table tree for the vma
- * address range.
- * @tile: The tile to bind for.
- * @vma: The vma to bind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before binding the built tree to the live vm tree.
- * @num_syncs: Number of @sync entries.
- * @rebind: Whether we're rebinding this vma to the same address range without
- * an unbind in-between.
- *
- * This function builds a page-table tree (see xe_pt_stage_bind() for more
- * information on page-table building), and the xe_vm_pgtable_update entries
- * abstracting the operations needed to attach it to the main vm tree. It
- * then takes the relevant locks and updates the metadata side of the main
- * vm tree and submits the operations for pipelined attachment of the
- * gpu page-table to the vm main tree, (which can be done either by the
- * cpu and the GPU).
- *
- * Return: A valid dma-fence representing the pipelined attachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs,
- bool rebind)
-{
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update bind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_bind_ops : &bind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
- .bind = true,
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL;
- struct xe_range_fence *rfence;
- int err;
-
- bind_pt_update.locked = false;
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- err = xe_pt_prepare_bind(tile, vma, entries, &num_entries);
- if (err)
- goto err;
-
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- goto err;
-
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &bind_pt_update, entries,
- num_entries);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. on LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * on !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!rebind && xe_vm_has_scratch(vm) && xe_vm_in_preempt_fence_mode(vm))) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
- } else if (rebind && !xe_vm_in_lr_mode(vm)) {
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
- }
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, xe_vma_bo(vma), q,
- entries, num_entries,
- syncs, num_syncs,
- &bind_pt_update.base);
- if (!IS_ERR(fence)) {
- bool last_munmap_rebind = vma->gpuva.flags & XE_VMA_LAST_REBIND;
- LLIST_HEAD(deferred);
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- bind_pt_update.base.start,
- bind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
- /* TLB invalidation must be done before signaling rebind */
- if (ifence) {
- int err = invalidation_fence_init(tile->primary_gt,
- ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
- }
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence, rebind ||
- last_munmap_rebind ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_bind(vma, entries, num_entries, rebind,
- bind_pt_update.locked ? &deferred : NULL);
-
- /* This vma is live (again?) now */
- vma->tile_present |= BIT(tile->id);
-
- if (bind_pt_update.locked) {
- to_userptr_vma(vma)->userptr.initial_bind = true;
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
- if (!rebind && last_munmap_rebind &&
- xe_vm_in_preempt_fence_mode(vm))
- xe_vm_queue_rebind_worker(vm);
- } else {
- kfree(rfence);
- kfree(ifence);
- if (bind_pt_update.locked)
- up_read(&vm->userptr.notifier_lock);
- xe_pt_abort_bind(vma, entries, num_entries);
- }
-
- return fence;
-
-err:
- return ERR_PTR(err);
-}
-
-struct xe_pt_stage_unbind_walk {
- /** @base: The pagewalk base-class. */
- struct xe_pt_walk base;
-
- /* Input parameters for the walk */
- /** @tile: The tile we're unbinding from. */
- struct xe_tile *tile;
-
- /**
- * @modified_start: Walk range start, modified to include any
- * shared pagetables that we're the only user of and can thus
- * treat as private.
- */
- u64 modified_start;
- /** @modified_end: Walk range start, modified like @modified_start. */
- u64 modified_end;
-
- /* Output */
- /* @wupd: Structure to track the page-table updates we're building */
- struct xe_walk_update wupd;
-};
-
-/*
- * Check whether this range is the only one populating this pagetable,
- * and in that case, update the walk range checks so that higher levels don't
- * view us as a shared pagetable.
- */
-static bool xe_pt_check_kill(u64 addr, u64 next, unsigned int level,
- const struct xe_pt *child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- unsigned int shift = walk->shifts[level];
- u64 size = 1ull << shift;
-
- if (IS_ALIGNED(addr, size) && IS_ALIGNED(next, size) &&
- ((next - addr) >> shift) == child->num_live) {
- u64 size = 1ull << walk->shifts[level + 1];
-
- *action = ACTION_CONTINUE;
-
- if (xe_walk->modified_start >= addr)
- xe_walk->modified_start = round_down(addr, size);
- if (xe_walk->modified_end <= next)
- xe_walk->modified_end = round_up(next, size);
-
- return true;
- }
-
- return false;
-}
-
-static int xe_pt_stage_unbind_entry(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
-
- XE_WARN_ON(!*child);
- XE_WARN_ON(!level);
-
- xe_pt_check_kill(addr, next, level - 1, xe_child, action, walk);
-
- return 0;
-}
-
-static int
-xe_pt_stage_unbind_post_descend(struct xe_ptw *parent, pgoff_t offset,
- unsigned int level, u64 addr, u64 next,
- struct xe_ptw **child,
- enum page_walk_action *action,
- struct xe_pt_walk *walk)
-{
- struct xe_pt_stage_unbind_walk *xe_walk =
- container_of(walk, typeof(*xe_walk), base);
- struct xe_pt *xe_child = container_of(*child, typeof(*xe_child), base);
- pgoff_t end_offset;
- u64 size = 1ull << walk->shifts[--level];
-
- if (!IS_ALIGNED(addr, size))
- addr = xe_walk->modified_start;
- if (!IS_ALIGNED(next, size))
- next = xe_walk->modified_end;
-
- /* Parent == *child is the root pt. Don't kill it. */
- if (parent != *child &&
- xe_pt_check_kill(addr, next, level, xe_child, action, walk))
- return 0;
-
- if (!xe_pt_nonshared_offsets(addr, next, level, walk, action, &offset,
- &end_offset))
- return 0;
-
- (void)xe_pt_new_shared(&xe_walk->wupd, xe_child, offset, false);
- xe_walk->wupd.updates[level].update->qwords = end_offset - offset;
-
- return 0;
-}
-
-static const struct xe_pt_walk_ops xe_pt_stage_unbind_ops = {
- .pt_entry = xe_pt_stage_unbind_entry,
- .pt_post_descend = xe_pt_stage_unbind_post_descend,
-};
-
-/**
- * xe_pt_stage_unbind() - Build page-table update structures for an unbind
- * operation
- * @tile: The tile we're unbinding for.
- * @vma: The vma we're unbinding.
- * @entries: Caller-provided storage for the update structures.
- *
- * Builds page-table update structures for an unbind operation. The function
- * will attempt to remove all page-tables that we're the only user
- * of, and for that to work, the unbind operation must be committed in the
- * same critical section that blocks racing binds to the same page-table tree.
- *
- * Return: The number of entries used.
- */
-static unsigned int xe_pt_stage_unbind(struct xe_tile *tile, struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries)
-{
- struct xe_pt_stage_unbind_walk xe_walk = {
- .base = {
- .ops = &xe_pt_stage_unbind_ops,
- .shifts = xe_normal_pt_shifts,
- .max_level = XE_PT_HIGHEST_LEVEL,
- },
- .tile = tile,
- .modified_start = xe_vma_start(vma),
- .modified_end = xe_vma_end(vma),
- .wupd.entries = entries,
- };
- struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
-
- (void)xe_pt_walk_shared(&pt->base, pt->level, xe_vma_start(vma),
- xe_vma_end(vma), &xe_walk.base);
-
- return xe_walk.wupd.num_used_entries;
-}
-
-static void
-xe_migrate_clear_pgtable_callback(struct xe_migrate_pt_update *pt_update,
- struct xe_tile *tile, struct iosys_map *map,
- void *ptr, u32 qword_ofs, u32 num_qwords,
- const struct xe_vm_pgtable_update *update)
-{
- struct xe_vma *vma = pt_update->vma;
- u64 empty = __xe_pt_empty_pte(tile, xe_vma_vm(vma), update->pt->level);
- int i;
-
- if (map && map->is_iomem)
- for (i = 0; i < num_qwords; ++i)
- xe_map_wr(tile_to_xe(tile), map, (qword_ofs + i) *
- sizeof(u64), u64, empty);
- else if (map)
- memset64(map->vaddr + qword_ofs * sizeof(u64), empty,
- num_qwords);
- else
- memset64(ptr, empty, num_qwords);
-}
-
-static void
-xe_pt_commit_unbind(struct xe_vma *vma,
- struct xe_vm_pgtable_update *entries, u32 num_entries,
- struct llist_head *deferred)
-{
- u32 j;
-
- xe_pt_commit_locks_assert(vma);
-
- for (j = 0; j < num_entries; ++j) {
- struct xe_vm_pgtable_update *entry = &entries[j];
- struct xe_pt *pt = entry->pt;
-
- pt->num_live -= entry->qwords;
- if (pt->level) {
- struct xe_pt_dir *pt_dir = as_xe_pt_dir(pt);
- u32 i;
-
- for (i = entry->ofs; i < entry->ofs + entry->qwords;
- i++) {
- if (xe_pt_entry(pt_dir, i))
- xe_pt_destroy(xe_pt_entry(pt_dir, i),
- xe_vma_vm(vma)->flags, deferred);
-
- pt_dir->children[i] = NULL;
- }
- }
- }
-}
-
-<<<<<<<
-static const struct xe_migrate_pt_update_ops unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
-=======
-static void
-xe_pt_update_ops_rfence_interval(struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int i, level = 0;
- u64 start, last;
-
- for (i = 0; i < pt_op->num_entries; i++) {
- const struct xe_vm_pgtable_update *entry = &pt_op->entries[i];
-
- if (entry->pt->level > level)
- level = entry->pt->level;
- }
-
- /* Greedy (non-optimal) calculation but simple */
- start = ALIGN_DOWN(xe_vma_start(vma), 0x1ull << xe_pt_shift(level));
- last = ALIGN(xe_vma_end(vma), 0x1ull << xe_pt_shift(level)) - 1;
-
- if (start < pt_update_ops->start)
- pt_update_ops->start = start;
- if (last > pt_update_ops->last)
- pt_update_ops->last = last;
-}
-
-static int vma_reserve_fences(struct xe_device *xe, struct xe_vma *vma)
-{
- int shift = xe_device_get_root_tile(xe)->media_gt ? 1 : 0;
-
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- return dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv,
- xe->info.tile_count << shift);
-
- return 0;
-}
-
-static int bind_op_prepare(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing bind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- pt_op->vma = NULL;
- pt_op->bind = true;
- pt_op->rebind = BIT(tile->id) & vma->tile_present;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- err = xe_pt_prepare_bind(tile, vma, pt_op->entries,
- &pt_op->num_entries);
- if (!err) {
- xe_tile_assert(tile, pt_op->num_entries <=
- ARRAY_SIZE(pt_op->entries));
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, true);
-
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
-
- /*
- * If rebind, we have to invalidate TLB on !LR vms to invalidate
- * cached PTEs point to freed memory. On LR vms this is done
- * automatically when the context is re-enabled by the rebind worker,
- * or in fault mode it was invalidated on PTE zapping.
- *
- * If !rebind, and scratch enabled VMs, there is a chance the scratch
- * PTE is already cached in the TLB so it needs to be invalidated.
- * On !LR VMs this is done in the ring ops preceding a batch, but on
- * non-faulting LR, in particular on user-space batch buffer chaining,
- * it needs to be done here.
- */
- if ((!pt_op->rebind && xe_vm_has_scratch(vm) &&
- xe_vm_in_preempt_fence_mode(vm)))
- pt_update_ops->needs_invalidation = true;
- else if (pt_op->rebind && !xe_vm_in_lr_mode(vm))
- /* We bump also if batch_invalidate_tlb is true */
- vm->tlb_flush_seqno++;
-
- vma->tile_staged |= BIT(tile->id);
- pt_op->vma = vma;
- xe_pt_commit_prepare_bind(vma, pt_op->entries,
- pt_op->num_entries, pt_op->rebind);
- } else {
- xe_pt_cancel_bind(vma, pt_op->entries, pt_op->num_entries);
- }
-
- return err;
-}
-
-static int unbind_op_prepare(struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma)
-{
- u32 current_op = pt_update_ops->current_op;
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[current_op];
- int err;
-
- if (!((vma->tile_present | vma->tile_staged) & BIT(tile->id)))
- return 0;
-
- xe_bo_assert_held(xe_vma_bo(vma));
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx)\n",
- xe_vma_start(vma), xe_vma_end(vma) - 1);
-
- /*
- * Wait for invalidation to complete. Can corrupt internal page table
- * state if an invalidation is running while preparing an unbind.
- */
- if (xe_vma_is_userptr(vma) && xe_vm_in_fault_mode(xe_vma_vm(vma)))
- mmu_interval_read_begin(&to_userptr_vma(vma)->userptr.notifier);
-
- pt_op->vma = vma;
- pt_op->bind = false;
- pt_op->rebind = false;
-
- err = vma_reserve_fences(tile_to_xe(tile), vma);
- if (err)
- return err;
-
- pt_op->num_entries = xe_pt_stage_unbind(tile, vma, pt_op->entries);
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
- pt_op->num_entries, false);
- xe_pt_update_ops_rfence_interval(pt_update_ops, vma);
- ++pt_update_ops->current_op;
- pt_update_ops->needs_userptr_lock |= xe_vma_is_userptr(vma);
- pt_update_ops->needs_invalidation = true;
-
- xe_pt_commit_prepare_unbind(vma, pt_op->entries, pt_op->num_entries);
-
- return 0;
-}
-
-static int op_prepare(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op)
-{
- int err = 0;
-
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- err = bind_op_prepare(vm, tile, pt_update_ops, op->map.vma);
- pt_update_ops->wait_vm_kernel = true;
- break;
- case DRM_GPUVA_OP_REMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va));
-
- if (!err && op->remap.prev) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.prev);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- if (!err && op->remap.next) {
- err = bind_op_prepare(vm, tile, pt_update_ops,
- op->remap.next);
- pt_update_ops->wait_vm_bookkeep = true;
- }
- break;
- case DRM_GPUVA_OP_UNMAP:
- err = unbind_op_prepare(tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va));
- break;
- case DRM_GPUVA_OP_PREFETCH:
- err = bind_op_prepare(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va));
- pt_update_ops->wait_vm_kernel = true;
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-
- return err;
-}
-
-static void
-xe_pt_update_ops_init(struct xe_vm_pgtable_update_ops *pt_update_ops)
-{
- init_llist_head(&pt_update_ops->deferred);
- pt_update_ops->start = ~0x0ull;
- pt_update_ops->last = 0x0ull;
-}
-
-/**
- * xe_pt_update_ops_prepare() - Prepare PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Prepare PT update operations which includes updating internal PT state,
- * allocate memory for page tables, populate page table being pruned in, and
- * create PT update operations for leaf insertion / removal.
- *
- * Return: 0 on success, negative error code on error.
- */
-int xe_pt_update_ops_prepare(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct xe_vma_op *op;
- int shift = tile->media_gt ? 1 : 0;
- int err;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- xe_pt_update_ops_init(pt_update_ops);
-
- err = dma_resv_reserve_fences(xe_vm_resv(vops->vm),
- tile_to_xe(tile)->info.tile_count << shift);
- if (err)
- return err;
-
- list_for_each_entry(op, &vops->list, link) {
- err = op_prepare(vops->vm, tile, pt_update_ops, op);
-
- if (err)
- return err;
- }
-
- xe_tile_assert(tile, pt_update_ops->current_op <=
- pt_update_ops->num_ops);
-
-#ifdef TEST_VM_OPS_ERROR
- if (vops->inject_error &&
- vops->vm->xe->vm_inject_error_position == FORCE_OP_ERROR_PREPARE)
- return -ENOSPC;
-#endif
-
- return 0;
-}
-
-static void bind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present |= BIT(tile->id);
- vma->tile_staged &= ~BIT(tile->id);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
- to_userptr_vma(vma)->userptr.initial_bind = true;
- }
-
- /*
- * Kick rebind worker if this bind triggers preempt fences and not in
- * the rebind worker
- */
- if (pt_update_ops->wait_vm_bookkeep &&
- xe_vm_in_preempt_fence_mode(vm) &&
- !current->mm)
- xe_vm_queue_rebind_worker(vm);
-}
-
-static void unbind_op_commit(struct xe_vm *vm, struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma *vma, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm) {
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- if (fence2)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence2,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
- }
- vma->tile_present &= ~BIT(tile->id);
- if (!vma->tile_present) {
- list_del_init(&vma->combined_links.rebind);
- if (xe_vma_is_userptr(vma)) {
- lockdep_assert_held_read(&vm->userptr.notifier_lock);
-
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- }
-}
-
-static void op_commit(struct xe_vm *vm,
- struct xe_tile *tile,
- struct xe_vm_pgtable_update_ops *pt_update_ops,
- struct xe_vma_op *op, struct dma_fence *fence,
- struct dma_fence *fence2)
-{
- xe_vm_assert_held(vm);
-
- switch (op->base.op) {
- case DRM_GPUVA_OP_MAP:
- if (!op->map.immediate && xe_vm_in_fault_mode(vm))
- break;
-
- bind_op_commit(vm, tile, pt_update_ops, op->map.vma, fence,
- fence2);
- break;
- case DRM_GPUVA_OP_REMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.remap.unmap->va), fence,
- fence2);
-
- if (op->remap.prev)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.prev,
- fence, fence2);
- if (op->remap.next)
- bind_op_commit(vm, tile, pt_update_ops, op->remap.next,
- fence, fence2);
- break;
- case DRM_GPUVA_OP_UNMAP:
- unbind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.unmap.va), fence, fence2);
- break;
- case DRM_GPUVA_OP_PREFETCH:
- bind_op_commit(vm, tile, pt_update_ops,
- gpuva_to_vma(op->base.prefetch.va), fence, fence2);
- break;
- default:
- drm_warn(&vm->xe->drm, "NOT POSSIBLE");
- }
-}
-
-static const struct xe_migrate_pt_update_ops migrate_ops = {
- .populate = xe_vm_populate_pgtable,
- .clear = xe_migrate_clear_pgtable_callback,
->>>>>>>
- .pre_commit = xe_pt_pre_commit,
-};
-
-static const struct xe_migrate_pt_update_ops userptr_unbind_ops = {
- .populate = xe_migrate_clear_pgtable_callback,
- .pre_commit = xe_pt_userptr_pre_commit,
-};
-
-/**
- * __xe_pt_unbind_vma() - Disconnect and free a page-table tree for the vma
- * address range.
- * @tile: The tile to unbind for.
- * @vma: The vma to unbind.
- * @q: The exec_queue with which to do pipelined page-table updates.
- * @syncs: Entries to sync on before disconnecting the tree to be destroyed.
- * @num_syncs: Number of @sync entries.
- *
- * This function builds a the xe_vm_pgtable_update entries abstracting the
- * operations needed to detach the page-table tree to be destroyed from the
- * man vm tree.
- * It then takes the relevant locks and submits the operations for
- * pipelined detachment of the gpu page-table from the vm main tree,
- * (which can be done either by the cpu and the GPU), Finally it frees the
- * detached page-table tree.
- *
- * Return: A valid dma-fence representing the pipelined detachment operation
- * on success, an error pointer on error.
- */
-struct dma_fence *
-__xe_pt_unbind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue *q,
- struct xe_sync_entry *syncs, u32 num_syncs)
-{
-<<<<<<<
- struct xe_vm *vm = vops->vm;
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- struct dma_fence *fence;
- struct invalidation_fence *ifence = NULL, *mfence = NULL;
- struct dma_fence **fences = NULL;
- struct dma_fence_array *cf = NULL;
- struct xe_range_fence *rfence;
- struct xe_vma_op *op;
- int err = 0, i;
- struct xe_migrate_pt_update update = {
- .ops = pt_update_ops->needs_userptr_lock ?
- &userptr_migrate_ops :
- &migrate_ops,
- .vops = vops,
- .tile_id = tile->id,
-=======
- struct xe_vm_pgtable_update entries[XE_VM_MAX_LEVEL * 2 + 1];
- struct xe_pt_migrate_pt_update unbind_pt_update = {
- .base = {
- .ops = xe_vma_is_userptr(vma) ? &userptr_unbind_ops :
- &unbind_ops,
- .vma = vma,
- .tile_id = tile->id,
- },
->>>>>>>
- };
- struct xe_vm *vm = xe_vma_vm(vma);
- u32 num_entries;
- struct dma_fence *fence = NULL;
- struct invalidation_fence *ifence;
- struct xe_range_fence *rfence;
- int err;
-
- LLIST_HEAD(deferred);
-
- xe_bo_assert_held(xe_vma_bo(vma));
- xe_vm_assert_held(vm);
-
- vm_dbg(&xe_vma_vm(vma)->xe->drm,
- "Preparing unbind, with range [%llx...%llx) engine %p.\n",
- xe_vma_start(vma), xe_vma_end(vma), q);
-
- num_entries = xe_pt_stage_unbind(tile, vma, entries);
- xe_tile_assert(tile, num_entries <= ARRAY_SIZE(entries));
-
- xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
- xe_pt_calc_rfence_interval(vma, &unbind_pt_update, entries,
- num_entries);
-
-<<<<<<<
- err = dma_resv_reserve_fences(xe_vm_resv(vm), 1);
- if (!err && !xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- err = dma_resv_reserve_fences(xe_vma_bo(vma)->ttm.base.resv, 1);
- if (err)
- return ERR_PTR(err);
-
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence)
- return ERR_PTR(-ENOMEM);
-=======
- if (pt_update_ops->needs_invalidation) {
- ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!ifence) {
- err = -ENOMEM;
- goto kill_vm_tile1;
- }
- if (tile->media_gt) {
- mfence = kzalloc(sizeof(*ifence), GFP_KERNEL);
- if (!mfence) {
- err = -ENOMEM;
- goto free_ifence;
- }
- fences = kmalloc_array(2, sizeof(*fences), GFP_KERNEL);
- if (!fences) {
- err = -ENOMEM;
- goto free_ifence;
- }
- cf = dma_fence_array_alloc(2);
- if (!cf) {
- err = -ENOMEM;
- goto free_ifence;
- }
- }
- }
->>>>>>>
-
- rfence = kzalloc(sizeof(*rfence), GFP_KERNEL);
- if (!rfence) {
- kfree(ifence);
- return ERR_PTR(-ENOMEM);
- }
-
- /*
- * Even if we were already evicted and unbind to destroy, we need to
- * clear again here. The eviction may have updated pagetables at a
- * lower level, because it needs to be more conservative.
- */
- fence = xe_migrate_update_pgtables(tile->migrate,
- vm, NULL, q ? q :
- vm->q[tile->id],
- entries, num_entries,
- syncs, num_syncs,
- &unbind_pt_update.base);
- if (!IS_ERR(fence)) {
- int err;
-
- err = xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- unbind_pt_update.base.start,
- unbind_pt_update.base.last, fence);
- if (err)
- dma_fence_wait(fence, false);
-
-<<<<<<<
- /* TLB invalidation must be done before signaling unbind */
- err = invalidation_fence_init(tile->primary_gt, ifence, fence,
- xe_vma_start(vma),
- xe_vma_end(vma),
- xe_vma_vm(vma)->usm.asid);
- if (err) {
- dma_fence_put(fence);
- kfree(ifence);
- return ERR_PTR(err);
- }
- fence = &ifence->base.base;
-
- /* add shared fence now for pagetable delayed destroy */
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- DMA_RESV_USAGE_BOOKKEEP);
-
- /* This fence will be installed by caller when doing eviction */
- if (!xe_vma_has_no_bo(vma) && !xe_vma_bo(vma)->vm)
- dma_resv_add_fence(xe_vma_bo(vma)->ttm.base.resv, fence,
- DMA_RESV_USAGE_BOOKKEEP);
- xe_pt_commit_unbind(vma, entries, num_entries,
- unbind_pt_update.locked ? &deferred : NULL);
- vma->tile_present &= ~BIT(tile->id);
- } else {
- kfree(rfence);
- kfree(ifence);
- }
-
- if (!vma->tile_present)
- list_del_init(&vma->combined_links.rebind);
-
- if (unbind_pt_update.locked) {
- xe_tile_assert(tile, xe_vma_is_userptr(vma));
-=======
- xe_pt_commit(pt_op->vma, pt_op->entries,
- pt_op->num_entries, &pt_update_ops->deferred);
- pt_op->vma = NULL; /* skip in xe_pt_update_ops_abort */
- }
-
- if (xe_range_fence_insert(&vm->rftree[tile->id], rfence,
- &xe_range_fence_kfree_ops,
- pt_update_ops->start,
- pt_update_ops->last, fence))
- dma_fence_wait(fence, false);
-
- /* tlb invalidation must be done before signaling rebind */
- if (ifence) {
- if (mfence)
- dma_fence_get(fence);
- invalidation_fence_init(tile->primary_gt, ifence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- if (mfence) {
- invalidation_fence_init(tile->media_gt, mfence, fence,
- pt_update_ops->start,
- pt_update_ops->last, vm->usm.asid);
- fences[0] = &ifence->base.base;
- fences[1] = &mfence->base.base;
- dma_fence_array_init(cf, 2, fences,
- vm->composite_fence_ctx,
- vm->composite_fence_seqno++,
- false);
- fence = &cf->base;
- } else {
- fence = &ifence->base.base;
- }
- }
-
- if (!mfence) {
- dma_resv_add_fence(xe_vm_resv(vm), fence,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op, fence, NULL);
- } else {
- dma_resv_add_fence(xe_vm_resv(vm), &ifence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- dma_resv_add_fence(xe_vm_resv(vm), &mfence->base.base,
- pt_update_ops->wait_vm_bookkeep ?
- DMA_RESV_USAGE_KERNEL :
- DMA_RESV_USAGE_BOOKKEEP);
-
- list_for_each_entry(op, &vops->list, link)
- op_commit(vops->vm, tile, pt_update_ops, op,
- &ifence->base.base, &mfence->base.base);
- }
->>>>>>>
-
- if (!vma->tile_present) {
- spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
- spin_unlock(&vm->userptr.invalidated_lock);
- }
- up_read(&vm->userptr.notifier_lock);
- xe_bo_put_commit(&deferred);
- }
-
- return fence;
-<<<<<<<
-=======
-
-free_rfence:
- kfree(rfence);
-free_ifence:
- kfree(cf);
- kfree(fences);
- kfree(mfence);
- kfree(ifence);
-kill_vm_tile1:
- if (err != -EAGAIN && tile->id)
- xe_vm_kill(vops->vm, false);
-
- return ERR_PTR(err);
-}
-
-/**
- * xe_pt_update_ops_fini() - Finish PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operations
- *
- * Finish PT update operations by committing to destroy page table memory
- */
-void xe_pt_update_ops_fini(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = 0; i < pt_update_ops->current_op; ++i) {
- struct xe_vm_pgtable_update_op *pt_op = &pt_update_ops->ops[i];
-
- xe_pt_free_bind(pt_op->entries, pt_op->num_entries);
- }
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
-}
-
-/**
- * xe_pt_update_ops_abort() - Abort PT update operations
- * @tile: Tile of PT update operations
- * @vops: VMA operationa
- *
- * Abort PT update operations by unwinding internal PT state
- */
-void xe_pt_update_ops_abort(struct xe_tile *tile, struct xe_vma_ops *vops)
-{
- struct xe_vm_pgtable_update_ops *pt_update_ops =
- &vops->pt_update_ops[tile->id];
- int i;
-
- lockdep_assert_held(&vops->vm->lock);
- xe_vm_assert_held(vops->vm);
-
- for (i = pt_update_ops->num_ops - 1; i >= 0; --i) {
- struct xe_vm_pgtable_update_op *pt_op =
- &pt_update_ops->ops[i];
-
- if (!pt_op->vma || i >= pt_update_ops->current_op)
- continue;
-
- if (pt_op->bind)
- xe_pt_abort_bind(pt_op->vma, pt_op->entries,
- pt_op->num_entries,
- pt_op->rebind);
- else
- xe_pt_abort_unbind(pt_op->vma, pt_op->entries,
- pt_op->num_entries);
- }
-
- xe_bo_put_commit(&vops->pt_update_ops[tile->id].deferred);
->>>>>>>
-}
diff --git a/rr-cache/66926cdbe314a619c9f8586b1353ca1d3570e1e8/preimage b/rr-cache/66926cdbe314a619c9f8586b1353ca1d3570e1e8/preimage
deleted file mode 100644
index 49ec5bef5f4a..000000000000
--- a/rr-cache/66926cdbe314a619c9f8586b1353ca1d3570e1e8/preimage
+++ /dev/null
@@ -1,429 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.enable_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.enable_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(xe);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(xe);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(xe);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.enable_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(xe);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
- if (has_display(xe)) {
- drm_kms_helper_poll_disable(&xe->drm);
- intel_display_driver_disable_user_access(xe);
- }
-
- if (!runtime)
- intel_display_driver_suspend(xe);
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
-<<<<<<<
- if (!runtime && has_display(xe)) {
-=======
- if (has_display(xe))
->>>>>>>
- intel_display_driver_suspend_access(xe);
- intel_encoder_suspend_all(&xe->display);
- }
-
- intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold);
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
- if (has_display(xe))
- intel_display_driver_resume_access(xe);
-
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime)
- intel_display_driver_resume(xe);
-
- if (has_display(xe)) {
- drm_kms_helper_poll_enable(&xe->drm);
- intel_display_driver_enable_user_access(xe);
- }
- intel_hpd_poll_disable(xe);
-
- intel_opregion_resume(xe);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.enable_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.6 b/rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.6
deleted file mode 100644
index 42ffb0f8cccc..000000000000
--- a/rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.6
+++ /dev/null
@@ -1,982 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2021 Intel Corporation
- */
-
-#include "xe_exec_queue.h"
-
-#include <linux/nospec.h>
-
-#include <drm/drm_device.h>
-#include <drm/drm_file.h>
-#include <drm/xe_drm.h>
-
-#include "xe_device.h"
-#include "xe_gt.h"
-#include "xe_hw_engine_class_sysfs.h"
-#include "xe_hw_fence.h"
-#include "xe_lrc.h"
-#include "xe_macros.h"
-#include "xe_migrate.h"
-#include "xe_pm.h"
-#include "xe_ring_ops_types.h"
-#include "xe_trace.h"
-#include "xe_vm.h"
-
-enum xe_exec_queue_sched_prop {
- XE_EXEC_QUEUE_JOB_TIMEOUT = 0,
- XE_EXEC_QUEUE_TIMESLICE = 1,
- XE_EXEC_QUEUE_PREEMPT_TIMEOUT = 2,
- XE_EXEC_QUEUE_SCHED_PROP_MAX = 3,
-};
-
-static int exec_queue_user_extensions(struct xe_device *xe, struct xe_exec_queue *q,
- u64 extensions, int ext_number);
-
-static void __xe_exec_queue_free(struct xe_exec_queue *q)
-{
- if (q->vm)
- xe_vm_put(q->vm);
-
- if (q->xef)
- xe_file_put(q->xef);
-
- kfree(q);
-}
-
-static struct xe_exec_queue *__xe_exec_queue_alloc(struct xe_device *xe,
- struct xe_vm *vm,
- u32 logical_mask,
- u16 width, struct xe_hw_engine *hwe,
- u32 flags, u64 extensions)
-{
- struct xe_exec_queue *q;
- struct xe_gt *gt = hwe->gt;
- int err;
-
- /* only kernel queues can be permanent */
- XE_WARN_ON((flags & EXEC_QUEUE_FLAG_PERMANENT) && !(flags & EXEC_QUEUE_FLAG_KERNEL));
-
- q = kzalloc(struct_size(q, lrc, width), GFP_KERNEL);
- if (!q)
- return ERR_PTR(-ENOMEM);
-
- kref_init(&q->refcount);
- q->flags = flags;
- q->hwe = hwe;
- q->gt = gt;
- q->class = hwe->class;
- q->width = width;
- q->logical_mask = logical_mask;
- q->fence_irq = &gt->fence_irq[hwe->class];
- q->ring_ops = gt->ring_ops[hwe->class];
- q->ops = gt->exec_queue_ops;
- INIT_LIST_HEAD(&q->lr.link);
- INIT_LIST_HEAD(&q->multi_gt_link);
-
- q->sched_props.timeslice_us = hwe->eclass->sched_props.timeslice_us;
- q->sched_props.preempt_timeout_us =
- hwe->eclass->sched_props.preempt_timeout_us;
- q->sched_props.job_timeout_ms =
- hwe->eclass->sched_props.job_timeout_ms;
- if (q->flags & EXEC_QUEUE_FLAG_KERNEL &&
- q->flags & EXEC_QUEUE_FLAG_HIGH_PRIORITY)
- q->sched_props.priority = XE_EXEC_QUEUE_PRIORITY_KERNEL;
- else
- q->sched_props.priority = XE_EXEC_QUEUE_PRIORITY_NORMAL;
-
- if (vm)
- q->vm = xe_vm_get(vm);
-
- if (extensions) {
- /*
- * may set q->usm, must come before xe_lrc_create(),
- * may overwrite q->sched_props, must come before q->ops->init()
- */
- err = exec_queue_user_extensions(xe, q, extensions, 0);
- if (err) {
- __xe_exec_queue_free(q);
- return ERR_PTR(err);
- }
- }
-
- return q;
-}
-
-static int __xe_exec_queue_init(struct xe_exec_queue *q)
-{
- struct xe_vm *vm = q->vm;
- int i, err;
-
- if (vm) {
- err = xe_vm_lock(vm, true);
- if (err)
- return err;
- }
-
- for (i = 0; i < q->width; ++i) {
- q->lrc[i] = xe_lrc_create(q->hwe, q->vm, SZ_16K);
- if (IS_ERR(q->lrc[i])) {
- err = PTR_ERR(q->lrc[i]);
- goto err_unlock;
- }
- }
-
- if (vm)
- xe_vm_unlock(vm);
-
- err = q->ops->init(q);
- if (err)
- goto err_lrc;
-
- return 0;
-
-err_unlock:
- if (vm)
- xe_vm_unlock(vm);
-err_lrc:
- for (i = i - 1; i >= 0; --i)
- xe_lrc_put(q->lrc[i]);
- return err;
-}
-
-struct xe_exec_queue *xe_exec_queue_create(struct xe_device *xe, struct xe_vm *vm,
- u32 logical_mask, u16 width,
- struct xe_hw_engine *hwe, u32 flags,
- u64 extensions)
-{
- struct xe_exec_queue *q;
- int err;
-
- q = __xe_exec_queue_alloc(xe, vm, logical_mask, width, hwe, flags,
- extensions);
- if (IS_ERR(q))
- return q;
-
- err = __xe_exec_queue_init(q);
- if (err)
- goto err_post_alloc;
-
- return q;
-
-err_post_alloc:
- __xe_exec_queue_free(q);
- return ERR_PTR(err);
-}
-
-struct xe_exec_queue *xe_exec_queue_create_class(struct xe_device *xe, struct xe_gt *gt,
- struct xe_vm *vm,
- enum xe_engine_class class,
- u32 flags, u64 extensions)
-{
- struct xe_hw_engine *hwe, *hwe0 = NULL;
- enum xe_hw_engine_id id;
- u32 logical_mask = 0;
-
- for_each_hw_engine(hwe, gt, id) {
- if (xe_hw_engine_is_reserved(hwe))
- continue;
-
- if (hwe->class == class) {
- logical_mask |= BIT(hwe->logical_instance);
- if (!hwe0)
- hwe0 = hwe;
- }
- }
-
- if (!logical_mask)
- return ERR_PTR(-ENODEV);
-
- return xe_exec_queue_create(xe, vm, logical_mask, 1, hwe0, flags, extensions);
-}
-
-/**
- * xe_exec_queue_create_bind() - Create bind exec queue.
- * @xe: Xe device.
- * @tile: tile which bind exec queue belongs to.
- * @flags: exec queue creation flags
- * @extensions: exec queue creation extensions
- *
- * Normalize bind exec queue creation. Bind exec queue is tied to migration VM
- * for access to physical memory required for page table programming. On a
- * faulting devices the reserved copy engine instance must be used to avoid
- * deadlocking (user binds cannot get stuck behind faults as kernel binds which
- * resolve faults depend on user binds). On non-faulting devices any copy engine
- * can be used.
- *
- * Returns exec queue on success, ERR_PTR on failure
- */
-struct xe_exec_queue *xe_exec_queue_create_bind(struct xe_device *xe,
- struct xe_tile *tile,
- u32 flags, u64 extensions)
-{
- struct xe_gt *gt = tile->primary_gt;
- struct xe_exec_queue *q;
- struct xe_vm *migrate_vm;
-
- migrate_vm = xe_migrate_get_vm(tile->migrate);
- if (xe->info.has_usm) {
- struct xe_hw_engine *hwe = xe_gt_hw_engine(gt,
- XE_ENGINE_CLASS_COPY,
- gt->usm.reserved_bcs_instance,
- false);
-
- if (!hwe)
- return ERR_PTR(-EINVAL);
-
- q = xe_exec_queue_create(xe, migrate_vm,
- BIT(hwe->logical_instance), 1, hwe,
- flags, extensions);
- } else {
- q = xe_exec_queue_create_class(xe, gt, migrate_vm,
- XE_ENGINE_CLASS_COPY, flags,
- extensions);
- }
- xe_vm_put(migrate_vm);
-
- return q;
-}
-
-void xe_exec_queue_destroy(struct kref *ref)
-{
- struct xe_exec_queue *q = container_of(ref, struct xe_exec_queue, refcount);
- struct xe_exec_queue *eq, *next;
-
- xe_exec_queue_last_fence_put_unlocked(q);
- if (!(q->flags & EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD)) {
- list_for_each_entry_safe(eq, next, &q->multi_gt_list,
- multi_gt_link)
- xe_exec_queue_put(eq);
- }
-
- q->ops->fini(q);
-}
-
-void xe_exec_queue_fini(struct xe_exec_queue *q)
-{
- int i;
-
- for (i = 0; i < q->width; ++i)
- xe_lrc_put(q->lrc[i]);
- __xe_exec_queue_free(q);
-}
-
-void xe_exec_queue_assign_name(struct xe_exec_queue *q, u32 instance)
-{
- switch (q->class) {
- case XE_ENGINE_CLASS_RENDER:
- snprintf(q->name, sizeof(q->name), "rcs%d", instance);
- break;
- case XE_ENGINE_CLASS_VIDEO_DECODE:
- snprintf(q->name, sizeof(q->name), "vcs%d", instance);
- break;
- case XE_ENGINE_CLASS_VIDEO_ENHANCE:
- snprintf(q->name, sizeof(q->name), "vecs%d", instance);
- break;
- case XE_ENGINE_CLASS_COPY:
- snprintf(q->name, sizeof(q->name), "bcs%d", instance);
- break;
- case XE_ENGINE_CLASS_COMPUTE:
- snprintf(q->name, sizeof(q->name), "ccs%d", instance);
- break;
- case XE_ENGINE_CLASS_OTHER:
- snprintf(q->name, sizeof(q->name), "gsccs%d", instance);
- break;
- default:
- XE_WARN_ON(q->class);
- }
-}
-
-struct xe_exec_queue *xe_exec_queue_lookup(struct xe_file *xef, u32 id)
-{
- struct xe_exec_queue *q;
-
- mutex_lock(&xef->exec_queue.lock);
- q = xa_load(&xef->exec_queue.xa, id);
- if (q)
- xe_exec_queue_get(q);
- mutex_unlock(&xef->exec_queue.lock);
-
- return q;
-}
-
-enum xe_exec_queue_priority
-xe_exec_queue_device_get_max_priority(struct xe_device *xe)
-{
- return capable(CAP_SYS_NICE) ? XE_EXEC_QUEUE_PRIORITY_HIGH :
- XE_EXEC_QUEUE_PRIORITY_NORMAL;
-}
-
-static int exec_queue_set_priority(struct xe_device *xe, struct xe_exec_queue *q,
- u64 value)
-{
- if (XE_IOCTL_DBG(xe, value > XE_EXEC_QUEUE_PRIORITY_HIGH))
- return -EINVAL;
-
- if (XE_IOCTL_DBG(xe, value > xe_exec_queue_device_get_max_priority(xe)))
- return -EPERM;
-
- q->sched_props.priority = value;
- return 0;
-}
-
-static bool xe_exec_queue_enforce_schedule_limit(void)
-{
-#if IS_ENABLED(CONFIG_DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT)
- return true;
-#else
- return !capable(CAP_SYS_NICE);
-#endif
-}
-
-static void
-xe_exec_queue_get_prop_minmax(struct xe_hw_engine_class_intf *eclass,
- enum xe_exec_queue_sched_prop prop,
- u32 *min, u32 *max)
-{
- switch (prop) {
- case XE_EXEC_QUEUE_JOB_TIMEOUT:
- *min = eclass->sched_props.job_timeout_min;
- *max = eclass->sched_props.job_timeout_max;
- break;
- case XE_EXEC_QUEUE_TIMESLICE:
- *min = eclass->sched_props.timeslice_min;
- *max = eclass->sched_props.timeslice_max;
- break;
- case XE_EXEC_QUEUE_PREEMPT_TIMEOUT:
- *min = eclass->sched_props.preempt_timeout_min;
- *max = eclass->sched_props.preempt_timeout_max;
- break;
- default:
- break;
- }
-#if IS_ENABLED(CONFIG_DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT)
- if (capable(CAP_SYS_NICE)) {
- switch (prop) {
- case XE_EXEC_QUEUE_JOB_TIMEOUT:
- *min = XE_HW_ENGINE_JOB_TIMEOUT_MIN;
- *max = XE_HW_ENGINE_JOB_TIMEOUT_MAX;
- break;
- case XE_EXEC_QUEUE_TIMESLICE:
- *min = XE_HW_ENGINE_TIMESLICE_MIN;
- *max = XE_HW_ENGINE_TIMESLICE_MAX;
- break;
- case XE_EXEC_QUEUE_PREEMPT_TIMEOUT:
- *min = XE_HW_ENGINE_PREEMPT_TIMEOUT_MIN;
- *max = XE_HW_ENGINE_PREEMPT_TIMEOUT_MAX;
- break;
- default:
- break;
- }
- }
-#endif
-}
-
-static int exec_queue_set_timeslice(struct xe_device *xe, struct xe_exec_queue *q,
- u64 value)
-{
- u32 min = 0, max = 0;
-
- xe_exec_queue_get_prop_minmax(q->hwe->eclass,
- XE_EXEC_QUEUE_TIMESLICE, &min, &max);
-
- if (xe_exec_queue_enforce_schedule_limit() &&
- !xe_hw_engine_timeout_in_range(value, min, max))
- return -EINVAL;
-
- q->sched_props.timeslice_us = value;
- return 0;
-}
-
-typedef int (*xe_exec_queue_set_property_fn)(struct xe_device *xe,
- struct xe_exec_queue *q,
- u64 value);
-
-static const xe_exec_queue_set_property_fn exec_queue_set_property_funcs[] = {
- [DRM_XE_EXEC_QUEUE_SET_PROPERTY_PRIORITY] = exec_queue_set_priority,
- [DRM_XE_EXEC_QUEUE_SET_PROPERTY_TIMESLICE] = exec_queue_set_timeslice,
-};
-
-static int exec_queue_user_ext_set_property(struct xe_device *xe,
- struct xe_exec_queue *q,
- u64 extension)
-{
- u64 __user *address = u64_to_user_ptr(extension);
- struct drm_xe_ext_set_property ext;
- int err;
- u32 idx;
-
- err = __copy_from_user(&ext, address, sizeof(ext));
- if (XE_IOCTL_DBG(xe, err))
- return -EFAULT;
-
- if (XE_IOCTL_DBG(xe, ext.property >=
- ARRAY_SIZE(exec_queue_set_property_funcs)) ||
- XE_IOCTL_DBG(xe, ext.pad) ||
- XE_IOCTL_DBG(xe, ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_PRIORITY &&
- ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_TIMESLICE))
- return -EINVAL;
-
- idx = array_index_nospec(ext.property, ARRAY_SIZE(exec_queue_set_property_funcs));
- if (!exec_queue_set_property_funcs[idx])
- return -EINVAL;
-
- return exec_queue_set_property_funcs[idx](xe, q, ext.value);
-}
-
-typedef int (*xe_exec_queue_user_extension_fn)(struct xe_device *xe,
- struct xe_exec_queue *q,
- u64 extension);
-
-static const xe_exec_queue_user_extension_fn exec_queue_user_extension_funcs[] = {
- [DRM_XE_EXEC_QUEUE_EXTENSION_SET_PROPERTY] = exec_queue_user_ext_set_property,
-};
-
-#define MAX_USER_EXTENSIONS 16
-static int exec_queue_user_extensions(struct xe_device *xe, struct xe_exec_queue *q,
- u64 extensions, int ext_number)
-{
- u64 __user *address = u64_to_user_ptr(extensions);
- struct drm_xe_user_extension ext;
- int err;
- u32 idx;
-
- if (XE_IOCTL_DBG(xe, ext_number >= MAX_USER_EXTENSIONS))
- return -E2BIG;
-
- err = __copy_from_user(&ext, address, sizeof(ext));
- if (XE_IOCTL_DBG(xe, err))
- return -EFAULT;
-
- if (XE_IOCTL_DBG(xe, ext.pad) ||
- XE_IOCTL_DBG(xe, ext.name >=
- ARRAY_SIZE(exec_queue_user_extension_funcs)))
- return -EINVAL;
-
- idx = array_index_nospec(ext.name,
- ARRAY_SIZE(exec_queue_user_extension_funcs));
- err = exec_queue_user_extension_funcs[idx](xe, q, extensions);
- if (XE_IOCTL_DBG(xe, err))
- return err;
-
- if (ext.next_extension)
- return exec_queue_user_extensions(xe, q, ext.next_extension,
- ++ext_number);
-
- return 0;
-}
-
-<<<<<<<
-=======
-static const enum xe_engine_class user_to_xe_engine_class[] = {
- [DRM_XE_ENGINE_CLASS_RENDER] = XE_ENGINE_CLASS_RENDER,
- [DRM_XE_ENGINE_CLASS_COPY] = XE_ENGINE_CLASS_COPY,
- [DRM_XE_ENGINE_CLASS_VIDEO_DECODE] = XE_ENGINE_CLASS_VIDEO_DECODE,
- [DRM_XE_ENGINE_CLASS_VIDEO_ENHANCE] = XE_ENGINE_CLASS_VIDEO_ENHANCE,
- [DRM_XE_ENGINE_CLASS_COMPUTE] = XE_ENGINE_CLASS_COMPUTE,
-};
-
-static struct xe_hw_engine *
-find_hw_engine(struct xe_device *xe,
- struct drm_xe_engine_class_instance eci)
-{
- u32 idx;
-
- if (eci.engine_class >= ARRAY_SIZE(user_to_xe_engine_class))
- return NULL;
-
- if (eci.gt_id >= xe->info.gt_count)
- return NULL;
-
- idx = array_index_nospec(eci.engine_class,
- ARRAY_SIZE(user_to_xe_engine_class));
-
- return xe_gt_hw_engine(xe_device_get_gt(xe, eci.gt_id),
- user_to_xe_engine_class[idx],
- eci.engine_instance, true);
-}
-
-static u32 bind_exec_queue_logical_mask(struct xe_device *xe, struct xe_gt *gt,
- struct drm_xe_engine_class_instance *eci,
- u16 width, u16 num_placements)
-{
- struct xe_hw_engine *hwe;
- enum xe_hw_engine_id id;
- u32 logical_mask = 0;
-
- if (XE_IOCTL_DBG(xe, width != 1))
- return 0;
- if (XE_IOCTL_DBG(xe, num_placements != 1))
- return 0;
- if (XE_IOCTL_DBG(xe, eci[0].engine_instance != 0))
- return 0;
-
- eci[0].engine_class = DRM_XE_ENGINE_CLASS_COPY;
-
- for_each_hw_engine(hwe, gt, id) {
- if (xe_hw_engine_is_reserved(hwe))
- continue;
-
- if (hwe->class ==
- user_to_xe_engine_class[DRM_XE_ENGINE_CLASS_COPY])
- logical_mask |= BIT(hwe->logical_instance);
- }
-
- return logical_mask;
-}
-
->>>>>>>
-static u32 calc_validate_logical_mask(struct xe_device *xe, struct xe_gt *gt,
- struct drm_xe_engine_class_instance *eci,
- u16 width, u16 num_placements)
-{
- int len = width * num_placements;
- int i, j, n;
- u16 class;
- u16 gt_id;
- u32 return_mask = 0, prev_mask;
-
- if (XE_IOCTL_DBG(xe, !xe_device_uc_enabled(xe) &&
- len > 1))
- return 0;
-
- for (i = 0; i < width; ++i) {
- u32 current_mask = 0;
-
- for (j = 0; j < num_placements; ++j) {
- struct xe_hw_engine *hwe;
-
- n = j * width + i;
-
- hwe = find_hw_engine(xe, eci[n]);
- if (XE_IOCTL_DBG(xe, !hwe))
- return 0;
-
- if (XE_IOCTL_DBG(xe, xe_hw_engine_is_reserved(hwe)))
- return 0;
-
- if (XE_IOCTL_DBG(xe, n && eci[n].gt_id != gt_id) ||
- XE_IOCTL_DBG(xe, n && eci[n].engine_class != class))
- return 0;
-
- class = eci[n].engine_class;
- gt_id = eci[n].gt_id;
-
- if (width == 1 || !i)
- return_mask |= BIT(eci[n].engine_instance);
- current_mask |= BIT(eci[n].engine_instance);
- }
-
- /* Parallel submissions must be logically contiguous */
- if (i && XE_IOCTL_DBG(xe, current_mask != prev_mask << 1))
- return 0;
-
- prev_mask = current_mask;
- }
-
- return return_mask;
-}
-
-int xe_exec_queue_create_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct xe_device *xe = to_xe_device(dev);
- struct xe_file *xef = to_xe_file(file);
- struct drm_xe_exec_queue_create *args = data;
- struct drm_xe_engine_class_instance eci[XE_HW_ENGINE_MAX_INSTANCE];
- struct drm_xe_engine_class_instance __user *user_eci =
- u64_to_user_ptr(args->instances);
- struct xe_hw_engine *hwe;
- struct xe_vm *vm;
- struct xe_gt *gt;
- struct xe_tile *tile;
- struct xe_exec_queue *q = NULL;
- u32 logical_mask;
- u32 id;
- u32 len;
- int err;
-
- if (XE_IOCTL_DBG(xe, args->flags) ||
- XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
- return -EINVAL;
-
- len = args->width * args->num_placements;
- if (XE_IOCTL_DBG(xe, !len || len > XE_HW_ENGINE_MAX_INSTANCE))
- return -EINVAL;
-
- err = __copy_from_user(eci, user_eci,
- sizeof(struct drm_xe_engine_class_instance) *
- len);
- if (XE_IOCTL_DBG(xe, err))
- return -EFAULT;
-
- if (XE_IOCTL_DBG(xe, eci[0].gt_id >= xe->info.gt_count))
- return -EINVAL;
-
- if (eci[0].engine_class == DRM_XE_ENGINE_CLASS_VM_BIND) {
- if (XE_IOCTL_DBG(xe, args->width != 1) ||
- XE_IOCTL_DBG(xe, args->num_placements != 1) ||
- XE_IOCTL_DBG(xe, eci[0].engine_instance != 0))
- return -EINVAL;
-
- for_each_tile(tile, xe, id) {
- struct xe_exec_queue *new;
- u32 flags = EXEC_QUEUE_FLAG_VM;
-
- if (id)
- flags |= EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD;
-
-<<<<<<<
- eci[0].gt_id = gt->info.id;
- logical_mask = bind_exec_queue_logical_mask(xe, gt, eci,
- args->width,
- args->num_placements);
- if (XE_IOCTL_DBG(xe, !logical_mask))
- return -EINVAL;
-
- hwe = find_hw_engine(xe, eci[0]);
- if (XE_IOCTL_DBG(xe, !hwe))
- return -EINVAL;
-
- /* The migration vm doesn't hold rpm ref */
- xe_pm_runtime_get_noresume(xe);
-
- flags = EXEC_QUEUE_FLAG_VM | (id ? EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD : 0);
-
- migrate_vm = xe_migrate_get_vm(gt_to_tile(gt)->migrate);
- new = xe_exec_queue_create(xe, migrate_vm, logical_mask,
- args->width, hwe, flags,
- args->extensions);
-
- xe_pm_runtime_put(xe); /* now held by engine */
-
- xe_vm_put(migrate_vm);
-=======
- new = xe_exec_queue_create_bind(xe, tile, flags,
- args->extensions);
->>>>>>>
- if (IS_ERR(new)) {
- err = PTR_ERR(new);
- if (q)
- goto put_exec_queue;
- return err;
- }
- if (id == 0)
- q = new;
- else
- list_add_tail(&new->multi_gt_list,
- &q->multi_gt_link);
- }
- } else {
- gt = xe_device_get_gt(xe, eci[0].gt_id);
- logical_mask = calc_validate_logical_mask(xe, gt, eci,
- args->width,
- args->num_placements);
- if (XE_IOCTL_DBG(xe, !logical_mask))
- return -EINVAL;
-
- hwe = find_hw_engine(xe, eci[0]);
- if (XE_IOCTL_DBG(xe, !hwe))
- return -EINVAL;
-
- vm = xe_vm_lookup(xef, args->vm_id);
- if (XE_IOCTL_DBG(xe, !vm))
- return -ENOENT;
-
- err = down_read_interruptible(&vm->lock);
- if (err) {
- xe_vm_put(vm);
- return err;
- }
-
- if (XE_IOCTL_DBG(xe, xe_vm_is_closed_or_banned(vm))) {
- up_read(&vm->lock);
- xe_vm_put(vm);
- return -ENOENT;
- }
-
- q = xe_exec_queue_create(xe, vm, logical_mask,
- args->width, hwe, 0,
- args->extensions);
- up_read(&vm->lock);
- xe_vm_put(vm);
- if (IS_ERR(q))
- return PTR_ERR(q);
-
- if (xe_vm_in_preempt_fence_mode(vm)) {
- q->lr.context = dma_fence_context_alloc(1);
-
- err = xe_vm_add_compute_exec_queue(vm, q);
- if (XE_IOCTL_DBG(xe, err))
- goto put_exec_queue;
- }
- }
-
- mutex_lock(&xef->exec_queue.lock);
- err = xa_alloc(&xef->exec_queue.xa, &id, q, xa_limit_32b, GFP_KERNEL);
- mutex_unlock(&xef->exec_queue.lock);
- if (err)
- goto kill_exec_queue;
-
- args->exec_queue_id = id;
- q->xef = xe_file_get(xef);
-
- return 0;
-
-kill_exec_queue:
- xe_exec_queue_kill(q);
-put_exec_queue:
- xe_exec_queue_put(q);
- return err;
-}
-
-int xe_exec_queue_get_property_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct xe_device *xe = to_xe_device(dev);
- struct xe_file *xef = to_xe_file(file);
- struct drm_xe_exec_queue_get_property *args = data;
- struct xe_exec_queue *q;
- int ret;
-
- if (XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
- return -EINVAL;
-
- q = xe_exec_queue_lookup(xef, args->exec_queue_id);
- if (XE_IOCTL_DBG(xe, !q))
- return -ENOENT;
-
- switch (args->property) {
- case DRM_XE_EXEC_QUEUE_GET_PROPERTY_BAN:
- args->value = q->ops->reset_status(q);
- ret = 0;
- break;
- default:
- ret = -EINVAL;
- }
-
- xe_exec_queue_put(q);
-
- return ret;
-}
-
-/**
- * xe_exec_queue_is_lr() - Whether an exec_queue is long-running
- * @q: The exec_queue
- *
- * Return: True if the exec_queue is long-running, false otherwise.
- */
-bool xe_exec_queue_is_lr(struct xe_exec_queue *q)
-{
- return q->vm && xe_vm_in_lr_mode(q->vm) &&
- !(q->flags & EXEC_QUEUE_FLAG_VM);
-}
-
-static s32 xe_exec_queue_num_job_inflight(struct xe_exec_queue *q)
-{
- return q->lrc[0]->fence_ctx.next_seqno - xe_lrc_seqno(q->lrc[0]) - 1;
-}
-
-/**
- * xe_exec_queue_ring_full() - Whether an exec_queue's ring is full
- * @q: The exec_queue
- *
- * Return: True if the exec_queue's ring is full, false otherwise.
- */
-bool xe_exec_queue_ring_full(struct xe_exec_queue *q)
-{
- struct xe_lrc *lrc = q->lrc[0];
- s32 max_job = lrc->ring.size / MAX_JOB_SIZE_BYTES;
-
- return xe_exec_queue_num_job_inflight(q) >= max_job;
-}
-
-/**
- * xe_exec_queue_is_idle() - Whether an exec_queue is idle.
- * @q: The exec_queue
- *
- * FIXME: Need to determine what to use as the short-lived
- * timeline lock for the exec_queues, so that the return value
- * of this function becomes more than just an advisory
- * snapshot in time. The timeline lock must protect the
- * seqno from racing submissions on the same exec_queue.
- * Typically vm->resv, but user-created timeline locks use the migrate vm
- * and never grabs the migrate vm->resv so we have a race there.
- *
- * Return: True if the exec_queue is idle, false otherwise.
- */
-bool xe_exec_queue_is_idle(struct xe_exec_queue *q)
-{
- if (xe_exec_queue_is_parallel(q)) {
- int i;
-
- for (i = 0; i < q->width; ++i) {
- if (xe_lrc_seqno(q->lrc[i]) !=
- q->lrc[i]->fence_ctx.next_seqno - 1)
- return false;
- }
-
- return true;
- }
-
- return xe_lrc_seqno(q->lrc[0]) ==
- q->lrc[0]->fence_ctx.next_seqno - 1;
-}
-
-/**
- * xe_exec_queue_update_run_ticks() - Update run time in ticks for this exec queue
- * from hw
- * @q: The exec queue
- *
- * Update the timestamp saved by HW for this exec queue and save run ticks
- * calculated by using the delta from last update.
- */
-void xe_exec_queue_update_run_ticks(struct xe_exec_queue *q)
-{
- struct xe_file *xef;
- struct xe_lrc *lrc;
- u32 old_ts, new_ts;
-
- /*
- * Jobs that are run during driver load may use an exec_queue, but are
- * not associated with a user xe file, so avoid accumulating busyness
- * for kernel specific work.
- */
- if (!q->vm || !q->vm->xef)
- return;
-
- xef = q->vm->xef;
-
- /*
- * Only sample the first LRC. For parallel submission, all of them are
- * scheduled together and we compensate that below by multiplying by
- * width - this may introduce errors if that premise is not true and
- * they don't exit 100% aligned. On the other hand, looping through
- * the LRCs and reading them in different time could also introduce
- * errors.
- */
- lrc = q->lrc[0];
- new_ts = xe_lrc_update_timestamp(lrc, &old_ts);
- xef->run_ticks[q->class] += (new_ts - old_ts) * q->width;
-}
-
-void xe_exec_queue_kill(struct xe_exec_queue *q)
-{
- struct xe_exec_queue *eq = q, *next;
-
- list_for_each_entry_safe(eq, next, &eq->multi_gt_list,
- multi_gt_link) {
- q->ops->kill(eq);
- xe_vm_remove_compute_exec_queue(q->vm, eq);
- }
-
- q->ops->kill(q);
- xe_vm_remove_compute_exec_queue(q->vm, q);
-}
-
-int xe_exec_queue_destroy_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct xe_device *xe = to_xe_device(dev);
- struct xe_file *xef = to_xe_file(file);
- struct drm_xe_exec_queue_destroy *args = data;
- struct xe_exec_queue *q;
-
- if (XE_IOCTL_DBG(xe, args->pad) ||
- XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
- return -EINVAL;
-
- mutex_lock(&xef->exec_queue.lock);
- q = xa_erase(&xef->exec_queue.xa, args->exec_queue_id);
- mutex_unlock(&xef->exec_queue.lock);
- if (XE_IOCTL_DBG(xe, !q))
- return -ENOENT;
-
- xe_exec_queue_kill(q);
-
- trace_xe_exec_queue_close(q);
- xe_exec_queue_put(q);
-
- return 0;
-}
-
-static void xe_exec_queue_last_fence_lockdep_assert(struct xe_exec_queue *q,
- struct xe_vm *vm)
-{
- if (q->flags & EXEC_QUEUE_FLAG_VM)
- lockdep_assert_held(&vm->lock);
- else
- xe_vm_assert_held(vm);
-}
-
-/**
- * xe_exec_queue_last_fence_put() - Drop ref to last fence
- * @q: The exec queue
- * @vm: The VM the engine does a bind or exec for
- */
-void xe_exec_queue_last_fence_put(struct xe_exec_queue *q, struct xe_vm *vm)
-{
- xe_exec_queue_last_fence_lockdep_assert(q, vm);
-
- if (q->last_fence) {
- dma_fence_put(q->last_fence);
- q->last_fence = NULL;
- }
-}
-
-/**
- * xe_exec_queue_last_fence_put_unlocked() - Drop ref to last fence unlocked
- * @q: The exec queue
- *
- * Only safe to be called from xe_exec_queue_destroy().
- */
-void xe_exec_queue_last_fence_put_unlocked(struct xe_exec_queue *q)
-{
- if (q->last_fence) {
- dma_fence_put(q->last_fence);
- q->last_fence = NULL;
- }
-}
-
-/**
- * xe_exec_queue_last_fence_get() - Get last fence
- * @q: The exec queue
- * @vm: The VM the engine does a bind or exec for
- *
- * Get last fence, takes a ref
- *
- * Returns: last fence if not signaled, dma fence stub if signaled
- */
-struct dma_fence *xe_exec_queue_last_fence_get(struct xe_exec_queue *q,
- struct xe_vm *vm)
-{
- struct dma_fence *fence;
-
- xe_exec_queue_last_fence_lockdep_assert(q, vm);
-
- if (q->last_fence &&
- test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &q->last_fence->flags))
- xe_exec_queue_last_fence_put(q, vm);
-
- fence = q->last_fence ? q->last_fence : dma_fence_get_stub();
- dma_fence_get(fence);
- return fence;
-}
-
-/**
- * xe_exec_queue_last_fence_set() - Set last fence
- * @q: The exec queue
- * @vm: The VM the engine does a bind or exec for
- * @fence: The fence
- *
- * Set the last fence for the engine. Increases reference count for fence, when
- * closing engine xe_exec_queue_last_fence_put should be called.
- */
-void xe_exec_queue_last_fence_set(struct xe_exec_queue *q, struct xe_vm *vm,
- struct dma_fence *fence)
-{
- xe_exec_queue_last_fence_lockdep_assert(q, vm);
-
- xe_exec_queue_last_fence_put(q, vm);
- q->last_fence = dma_fence_get(fence);
-}
diff --git a/rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.7 b/rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.7
deleted file mode 100644
index 42ffb0f8cccc..000000000000
--- a/rr-cache/6e989852f5454d81ebf331bbd2c55116dc711575/preimage.7
+++ /dev/null
@@ -1,982 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2021 Intel Corporation
- */
-
-#include "xe_exec_queue.h"
-
-#include <linux/nospec.h>
-
-#include <drm/drm_device.h>
-#include <drm/drm_file.h>
-#include <drm/xe_drm.h>
-
-#include "xe_device.h"
-#include "xe_gt.h"
-#include "xe_hw_engine_class_sysfs.h"
-#include "xe_hw_fence.h"
-#include "xe_lrc.h"
-#include "xe_macros.h"
-#include "xe_migrate.h"
-#include "xe_pm.h"
-#include "xe_ring_ops_types.h"
-#include "xe_trace.h"
-#include "xe_vm.h"
-
-enum xe_exec_queue_sched_prop {
- XE_EXEC_QUEUE_JOB_TIMEOUT = 0,
- XE_EXEC_QUEUE_TIMESLICE = 1,
- XE_EXEC_QUEUE_PREEMPT_TIMEOUT = 2,
- XE_EXEC_QUEUE_SCHED_PROP_MAX = 3,
-};
-
-static int exec_queue_user_extensions(struct xe_device *xe, struct xe_exec_queue *q,
- u64 extensions, int ext_number);
-
-static void __xe_exec_queue_free(struct xe_exec_queue *q)
-{
- if (q->vm)
- xe_vm_put(q->vm);
-
- if (q->xef)
- xe_file_put(q->xef);
-
- kfree(q);
-}
-
-static struct xe_exec_queue *__xe_exec_queue_alloc(struct xe_device *xe,
- struct xe_vm *vm,
- u32 logical_mask,
- u16 width, struct xe_hw_engine *hwe,
- u32 flags, u64 extensions)
-{
- struct xe_exec_queue *q;
- struct xe_gt *gt = hwe->gt;
- int err;
-
- /* only kernel queues can be permanent */
- XE_WARN_ON((flags & EXEC_QUEUE_FLAG_PERMANENT) && !(flags & EXEC_QUEUE_FLAG_KERNEL));
-
- q = kzalloc(struct_size(q, lrc, width), GFP_KERNEL);
- if (!q)
- return ERR_PTR(-ENOMEM);
-
- kref_init(&q->refcount);
- q->flags = flags;
- q->hwe = hwe;
- q->gt = gt;
- q->class = hwe->class;
- q->width = width;
- q->logical_mask = logical_mask;
- q->fence_irq = &gt->fence_irq[hwe->class];
- q->ring_ops = gt->ring_ops[hwe->class];
- q->ops = gt->exec_queue_ops;
- INIT_LIST_HEAD(&q->lr.link);
- INIT_LIST_HEAD(&q->multi_gt_link);
-
- q->sched_props.timeslice_us = hwe->eclass->sched_props.timeslice_us;
- q->sched_props.preempt_timeout_us =
- hwe->eclass->sched_props.preempt_timeout_us;
- q->sched_props.job_timeout_ms =
- hwe->eclass->sched_props.job_timeout_ms;
- if (q->flags & EXEC_QUEUE_FLAG_KERNEL &&
- q->flags & EXEC_QUEUE_FLAG_HIGH_PRIORITY)
- q->sched_props.priority = XE_EXEC_QUEUE_PRIORITY_KERNEL;
- else
- q->sched_props.priority = XE_EXEC_QUEUE_PRIORITY_NORMAL;
-
- if (vm)
- q->vm = xe_vm_get(vm);
-
- if (extensions) {
- /*
- * may set q->usm, must come before xe_lrc_create(),
- * may overwrite q->sched_props, must come before q->ops->init()
- */
- err = exec_queue_user_extensions(xe, q, extensions, 0);
- if (err) {
- __xe_exec_queue_free(q);
- return ERR_PTR(err);
- }
- }
-
- return q;
-}
-
-static int __xe_exec_queue_init(struct xe_exec_queue *q)
-{
- struct xe_vm *vm = q->vm;
- int i, err;
-
- if (vm) {
- err = xe_vm_lock(vm, true);
- if (err)
- return err;
- }
-
- for (i = 0; i < q->width; ++i) {
- q->lrc[i] = xe_lrc_create(q->hwe, q->vm, SZ_16K);
- if (IS_ERR(q->lrc[i])) {
- err = PTR_ERR(q->lrc[i]);
- goto err_unlock;
- }
- }
-
- if (vm)
- xe_vm_unlock(vm);
-
- err = q->ops->init(q);
- if (err)
- goto err_lrc;
-
- return 0;
-
-err_unlock:
- if (vm)
- xe_vm_unlock(vm);
-err_lrc:
- for (i = i - 1; i >= 0; --i)
- xe_lrc_put(q->lrc[i]);
- return err;
-}
-
-struct xe_exec_queue *xe_exec_queue_create(struct xe_device *xe, struct xe_vm *vm,
- u32 logical_mask, u16 width,
- struct xe_hw_engine *hwe, u32 flags,
- u64 extensions)
-{
- struct xe_exec_queue *q;
- int err;
-
- q = __xe_exec_queue_alloc(xe, vm, logical_mask, width, hwe, flags,
- extensions);
- if (IS_ERR(q))
- return q;
-
- err = __xe_exec_queue_init(q);
- if (err)
- goto err_post_alloc;
-
- return q;
-
-err_post_alloc:
- __xe_exec_queue_free(q);
- return ERR_PTR(err);
-}
-
-struct xe_exec_queue *xe_exec_queue_create_class(struct xe_device *xe, struct xe_gt *gt,
- struct xe_vm *vm,
- enum xe_engine_class class,
- u32 flags, u64 extensions)
-{
- struct xe_hw_engine *hwe, *hwe0 = NULL;
- enum xe_hw_engine_id id;
- u32 logical_mask = 0;
-
- for_each_hw_engine(hwe, gt, id) {
- if (xe_hw_engine_is_reserved(hwe))
- continue;
-
- if (hwe->class == class) {
- logical_mask |= BIT(hwe->logical_instance);
- if (!hwe0)
- hwe0 = hwe;
- }
- }
-
- if (!logical_mask)
- return ERR_PTR(-ENODEV);
-
- return xe_exec_queue_create(xe, vm, logical_mask, 1, hwe0, flags, extensions);
-}
-
-/**
- * xe_exec_queue_create_bind() - Create bind exec queue.
- * @xe: Xe device.
- * @tile: tile which bind exec queue belongs to.
- * @flags: exec queue creation flags
- * @extensions: exec queue creation extensions
- *
- * Normalize bind exec queue creation. Bind exec queue is tied to migration VM
- * for access to physical memory required for page table programming. On a
- * faulting devices the reserved copy engine instance must be used to avoid
- * deadlocking (user binds cannot get stuck behind faults as kernel binds which
- * resolve faults depend on user binds). On non-faulting devices any copy engine
- * can be used.
- *
- * Returns exec queue on success, ERR_PTR on failure
- */
-struct xe_exec_queue *xe_exec_queue_create_bind(struct xe_device *xe,
- struct xe_tile *tile,
- u32 flags, u64 extensions)
-{
- struct xe_gt *gt = tile->primary_gt;
- struct xe_exec_queue *q;
- struct xe_vm *migrate_vm;
-
- migrate_vm = xe_migrate_get_vm(tile->migrate);
- if (xe->info.has_usm) {
- struct xe_hw_engine *hwe = xe_gt_hw_engine(gt,
- XE_ENGINE_CLASS_COPY,
- gt->usm.reserved_bcs_instance,
- false);
-
- if (!hwe)
- return ERR_PTR(-EINVAL);
-
- q = xe_exec_queue_create(xe, migrate_vm,
- BIT(hwe->logical_instance), 1, hwe,
- flags, extensions);
- } else {
- q = xe_exec_queue_create_class(xe, gt, migrate_vm,
- XE_ENGINE_CLASS_COPY, flags,
- extensions);
- }
- xe_vm_put(migrate_vm);
-
- return q;
-}
-
-void xe_exec_queue_destroy(struct kref *ref)
-{
- struct xe_exec_queue *q = container_of(ref, struct xe_exec_queue, refcount);
- struct xe_exec_queue *eq, *next;
-
- xe_exec_queue_last_fence_put_unlocked(q);
- if (!(q->flags & EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD)) {
- list_for_each_entry_safe(eq, next, &q->multi_gt_list,
- multi_gt_link)
- xe_exec_queue_put(eq);
- }
-
- q->ops->fini(q);
-}
-
-void xe_exec_queue_fini(struct xe_exec_queue *q)
-{
- int i;
-
- for (i = 0; i < q->width; ++i)
- xe_lrc_put(q->lrc[i]);
- __xe_exec_queue_free(q);
-}
-
-void xe_exec_queue_assign_name(struct xe_exec_queue *q, u32 instance)
-{
- switch (q->class) {
- case XE_ENGINE_CLASS_RENDER:
- snprintf(q->name, sizeof(q->name), "rcs%d", instance);
- break;
- case XE_ENGINE_CLASS_VIDEO_DECODE:
- snprintf(q->name, sizeof(q->name), "vcs%d", instance);
- break;
- case XE_ENGINE_CLASS_VIDEO_ENHANCE:
- snprintf(q->name, sizeof(q->name), "vecs%d", instance);
- break;
- case XE_ENGINE_CLASS_COPY:
- snprintf(q->name, sizeof(q->name), "bcs%d", instance);
- break;
- case XE_ENGINE_CLASS_COMPUTE:
- snprintf(q->name, sizeof(q->name), "ccs%d", instance);
- break;
- case XE_ENGINE_CLASS_OTHER:
- snprintf(q->name, sizeof(q->name), "gsccs%d", instance);
- break;
- default:
- XE_WARN_ON(q->class);
- }
-}
-
-struct xe_exec_queue *xe_exec_queue_lookup(struct xe_file *xef, u32 id)
-{
- struct xe_exec_queue *q;
-
- mutex_lock(&xef->exec_queue.lock);
- q = xa_load(&xef->exec_queue.xa, id);
- if (q)
- xe_exec_queue_get(q);
- mutex_unlock(&xef->exec_queue.lock);
-
- return q;
-}
-
-enum xe_exec_queue_priority
-xe_exec_queue_device_get_max_priority(struct xe_device *xe)
-{
- return capable(CAP_SYS_NICE) ? XE_EXEC_QUEUE_PRIORITY_HIGH :
- XE_EXEC_QUEUE_PRIORITY_NORMAL;
-}
-
-static int exec_queue_set_priority(struct xe_device *xe, struct xe_exec_queue *q,
- u64 value)
-{
- if (XE_IOCTL_DBG(xe, value > XE_EXEC_QUEUE_PRIORITY_HIGH))
- return -EINVAL;
-
- if (XE_IOCTL_DBG(xe, value > xe_exec_queue_device_get_max_priority(xe)))
- return -EPERM;
-
- q->sched_props.priority = value;
- return 0;
-}
-
-static bool xe_exec_queue_enforce_schedule_limit(void)
-{
-#if IS_ENABLED(CONFIG_DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT)
- return true;
-#else
- return !capable(CAP_SYS_NICE);
-#endif
-}
-
-static void
-xe_exec_queue_get_prop_minmax(struct xe_hw_engine_class_intf *eclass,
- enum xe_exec_queue_sched_prop prop,
- u32 *min, u32 *max)
-{
- switch (prop) {
- case XE_EXEC_QUEUE_JOB_TIMEOUT:
- *min = eclass->sched_props.job_timeout_min;
- *max = eclass->sched_props.job_timeout_max;
- break;
- case XE_EXEC_QUEUE_TIMESLICE:
- *min = eclass->sched_props.timeslice_min;
- *max = eclass->sched_props.timeslice_max;
- break;
- case XE_EXEC_QUEUE_PREEMPT_TIMEOUT:
- *min = eclass->sched_props.preempt_timeout_min;
- *max = eclass->sched_props.preempt_timeout_max;
- break;
- default:
- break;
- }
-#if IS_ENABLED(CONFIG_DRM_XE_ENABLE_SCHEDTIMEOUT_LIMIT)
- if (capable(CAP_SYS_NICE)) {
- switch (prop) {
- case XE_EXEC_QUEUE_JOB_TIMEOUT:
- *min = XE_HW_ENGINE_JOB_TIMEOUT_MIN;
- *max = XE_HW_ENGINE_JOB_TIMEOUT_MAX;
- break;
- case XE_EXEC_QUEUE_TIMESLICE:
- *min = XE_HW_ENGINE_TIMESLICE_MIN;
- *max = XE_HW_ENGINE_TIMESLICE_MAX;
- break;
- case XE_EXEC_QUEUE_PREEMPT_TIMEOUT:
- *min = XE_HW_ENGINE_PREEMPT_TIMEOUT_MIN;
- *max = XE_HW_ENGINE_PREEMPT_TIMEOUT_MAX;
- break;
- default:
- break;
- }
- }
-#endif
-}
-
-static int exec_queue_set_timeslice(struct xe_device *xe, struct xe_exec_queue *q,
- u64 value)
-{
- u32 min = 0, max = 0;
-
- xe_exec_queue_get_prop_minmax(q->hwe->eclass,
- XE_EXEC_QUEUE_TIMESLICE, &min, &max);
-
- if (xe_exec_queue_enforce_schedule_limit() &&
- !xe_hw_engine_timeout_in_range(value, min, max))
- return -EINVAL;
-
- q->sched_props.timeslice_us = value;
- return 0;
-}
-
-typedef int (*xe_exec_queue_set_property_fn)(struct xe_device *xe,
- struct xe_exec_queue *q,
- u64 value);
-
-static const xe_exec_queue_set_property_fn exec_queue_set_property_funcs[] = {
- [DRM_XE_EXEC_QUEUE_SET_PROPERTY_PRIORITY] = exec_queue_set_priority,
- [DRM_XE_EXEC_QUEUE_SET_PROPERTY_TIMESLICE] = exec_queue_set_timeslice,
-};
-
-static int exec_queue_user_ext_set_property(struct xe_device *xe,
- struct xe_exec_queue *q,
- u64 extension)
-{
- u64 __user *address = u64_to_user_ptr(extension);
- struct drm_xe_ext_set_property ext;
- int err;
- u32 idx;
-
- err = __copy_from_user(&ext, address, sizeof(ext));
- if (XE_IOCTL_DBG(xe, err))
- return -EFAULT;
-
- if (XE_IOCTL_DBG(xe, ext.property >=
- ARRAY_SIZE(exec_queue_set_property_funcs)) ||
- XE_IOCTL_DBG(xe, ext.pad) ||
- XE_IOCTL_DBG(xe, ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_PRIORITY &&
- ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_TIMESLICE))
- return -EINVAL;
-
- idx = array_index_nospec(ext.property, ARRAY_SIZE(exec_queue_set_property_funcs));
- if (!exec_queue_set_property_funcs[idx])
- return -EINVAL;
-
- return exec_queue_set_property_funcs[idx](xe, q, ext.value);
-}
-
-typedef int (*xe_exec_queue_user_extension_fn)(struct xe_device *xe,
- struct xe_exec_queue *q,
- u64 extension);
-
-static const xe_exec_queue_user_extension_fn exec_queue_user_extension_funcs[] = {
- [DRM_XE_EXEC_QUEUE_EXTENSION_SET_PROPERTY] = exec_queue_user_ext_set_property,
-};
-
-#define MAX_USER_EXTENSIONS 16
-static int exec_queue_user_extensions(struct xe_device *xe, struct xe_exec_queue *q,
- u64 extensions, int ext_number)
-{
- u64 __user *address = u64_to_user_ptr(extensions);
- struct drm_xe_user_extension ext;
- int err;
- u32 idx;
-
- if (XE_IOCTL_DBG(xe, ext_number >= MAX_USER_EXTENSIONS))
- return -E2BIG;
-
- err = __copy_from_user(&ext, address, sizeof(ext));
- if (XE_IOCTL_DBG(xe, err))
- return -EFAULT;
-
- if (XE_IOCTL_DBG(xe, ext.pad) ||
- XE_IOCTL_DBG(xe, ext.name >=
- ARRAY_SIZE(exec_queue_user_extension_funcs)))
- return -EINVAL;
-
- idx = array_index_nospec(ext.name,
- ARRAY_SIZE(exec_queue_user_extension_funcs));
- err = exec_queue_user_extension_funcs[idx](xe, q, extensions);
- if (XE_IOCTL_DBG(xe, err))
- return err;
-
- if (ext.next_extension)
- return exec_queue_user_extensions(xe, q, ext.next_extension,
- ++ext_number);
-
- return 0;
-}
-
-<<<<<<<
-=======
-static const enum xe_engine_class user_to_xe_engine_class[] = {
- [DRM_XE_ENGINE_CLASS_RENDER] = XE_ENGINE_CLASS_RENDER,
- [DRM_XE_ENGINE_CLASS_COPY] = XE_ENGINE_CLASS_COPY,
- [DRM_XE_ENGINE_CLASS_VIDEO_DECODE] = XE_ENGINE_CLASS_VIDEO_DECODE,
- [DRM_XE_ENGINE_CLASS_VIDEO_ENHANCE] = XE_ENGINE_CLASS_VIDEO_ENHANCE,
- [DRM_XE_ENGINE_CLASS_COMPUTE] = XE_ENGINE_CLASS_COMPUTE,
-};
-
-static struct xe_hw_engine *
-find_hw_engine(struct xe_device *xe,
- struct drm_xe_engine_class_instance eci)
-{
- u32 idx;
-
- if (eci.engine_class >= ARRAY_SIZE(user_to_xe_engine_class))
- return NULL;
-
- if (eci.gt_id >= xe->info.gt_count)
- return NULL;
-
- idx = array_index_nospec(eci.engine_class,
- ARRAY_SIZE(user_to_xe_engine_class));
-
- return xe_gt_hw_engine(xe_device_get_gt(xe, eci.gt_id),
- user_to_xe_engine_class[idx],
- eci.engine_instance, true);
-}
-
-static u32 bind_exec_queue_logical_mask(struct xe_device *xe, struct xe_gt *gt,
- struct drm_xe_engine_class_instance *eci,
- u16 width, u16 num_placements)
-{
- struct xe_hw_engine *hwe;
- enum xe_hw_engine_id id;
- u32 logical_mask = 0;
-
- if (XE_IOCTL_DBG(xe, width != 1))
- return 0;
- if (XE_IOCTL_DBG(xe, num_placements != 1))
- return 0;
- if (XE_IOCTL_DBG(xe, eci[0].engine_instance != 0))
- return 0;
-
- eci[0].engine_class = DRM_XE_ENGINE_CLASS_COPY;
-
- for_each_hw_engine(hwe, gt, id) {
- if (xe_hw_engine_is_reserved(hwe))
- continue;
-
- if (hwe->class ==
- user_to_xe_engine_class[DRM_XE_ENGINE_CLASS_COPY])
- logical_mask |= BIT(hwe->logical_instance);
- }
-
- return logical_mask;
-}
-
->>>>>>>
-static u32 calc_validate_logical_mask(struct xe_device *xe, struct xe_gt *gt,
- struct drm_xe_engine_class_instance *eci,
- u16 width, u16 num_placements)
-{
- int len = width * num_placements;
- int i, j, n;
- u16 class;
- u16 gt_id;
- u32 return_mask = 0, prev_mask;
-
- if (XE_IOCTL_DBG(xe, !xe_device_uc_enabled(xe) &&
- len > 1))
- return 0;
-
- for (i = 0; i < width; ++i) {
- u32 current_mask = 0;
-
- for (j = 0; j < num_placements; ++j) {
- struct xe_hw_engine *hwe;
-
- n = j * width + i;
-
- hwe = find_hw_engine(xe, eci[n]);
- if (XE_IOCTL_DBG(xe, !hwe))
- return 0;
-
- if (XE_IOCTL_DBG(xe, xe_hw_engine_is_reserved(hwe)))
- return 0;
-
- if (XE_IOCTL_DBG(xe, n && eci[n].gt_id != gt_id) ||
- XE_IOCTL_DBG(xe, n && eci[n].engine_class != class))
- return 0;
-
- class = eci[n].engine_class;
- gt_id = eci[n].gt_id;
-
- if (width == 1 || !i)
- return_mask |= BIT(eci[n].engine_instance);
- current_mask |= BIT(eci[n].engine_instance);
- }
-
- /* Parallel submissions must be logically contiguous */
- if (i && XE_IOCTL_DBG(xe, current_mask != prev_mask << 1))
- return 0;
-
- prev_mask = current_mask;
- }
-
- return return_mask;
-}
-
-int xe_exec_queue_create_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct xe_device *xe = to_xe_device(dev);
- struct xe_file *xef = to_xe_file(file);
- struct drm_xe_exec_queue_create *args = data;
- struct drm_xe_engine_class_instance eci[XE_HW_ENGINE_MAX_INSTANCE];
- struct drm_xe_engine_class_instance __user *user_eci =
- u64_to_user_ptr(args->instances);
- struct xe_hw_engine *hwe;
- struct xe_vm *vm;
- struct xe_gt *gt;
- struct xe_tile *tile;
- struct xe_exec_queue *q = NULL;
- u32 logical_mask;
- u32 id;
- u32 len;
- int err;
-
- if (XE_IOCTL_DBG(xe, args->flags) ||
- XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
- return -EINVAL;
-
- len = args->width * args->num_placements;
- if (XE_IOCTL_DBG(xe, !len || len > XE_HW_ENGINE_MAX_INSTANCE))
- return -EINVAL;
-
- err = __copy_from_user(eci, user_eci,
- sizeof(struct drm_xe_engine_class_instance) *
- len);
- if (XE_IOCTL_DBG(xe, err))
- return -EFAULT;
-
- if (XE_IOCTL_DBG(xe, eci[0].gt_id >= xe->info.gt_count))
- return -EINVAL;
-
- if (eci[0].engine_class == DRM_XE_ENGINE_CLASS_VM_BIND) {
- if (XE_IOCTL_DBG(xe, args->width != 1) ||
- XE_IOCTL_DBG(xe, args->num_placements != 1) ||
- XE_IOCTL_DBG(xe, eci[0].engine_instance != 0))
- return -EINVAL;
-
- for_each_tile(tile, xe, id) {
- struct xe_exec_queue *new;
- u32 flags = EXEC_QUEUE_FLAG_VM;
-
- if (id)
- flags |= EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD;
-
-<<<<<<<
- eci[0].gt_id = gt->info.id;
- logical_mask = bind_exec_queue_logical_mask(xe, gt, eci,
- args->width,
- args->num_placements);
- if (XE_IOCTL_DBG(xe, !logical_mask))
- return -EINVAL;
-
- hwe = find_hw_engine(xe, eci[0]);
- if (XE_IOCTL_DBG(xe, !hwe))
- return -EINVAL;
-
- /* The migration vm doesn't hold rpm ref */
- xe_pm_runtime_get_noresume(xe);
-
- flags = EXEC_QUEUE_FLAG_VM | (id ? EXEC_QUEUE_FLAG_BIND_ENGINE_CHILD : 0);
-
- migrate_vm = xe_migrate_get_vm(gt_to_tile(gt)->migrate);
- new = xe_exec_queue_create(xe, migrate_vm, logical_mask,
- args->width, hwe, flags,
- args->extensions);
-
- xe_pm_runtime_put(xe); /* now held by engine */
-
- xe_vm_put(migrate_vm);
-=======
- new = xe_exec_queue_create_bind(xe, tile, flags,
- args->extensions);
->>>>>>>
- if (IS_ERR(new)) {
- err = PTR_ERR(new);
- if (q)
- goto put_exec_queue;
- return err;
- }
- if (id == 0)
- q = new;
- else
- list_add_tail(&new->multi_gt_list,
- &q->multi_gt_link);
- }
- } else {
- gt = xe_device_get_gt(xe, eci[0].gt_id);
- logical_mask = calc_validate_logical_mask(xe, gt, eci,
- args->width,
- args->num_placements);
- if (XE_IOCTL_DBG(xe, !logical_mask))
- return -EINVAL;
-
- hwe = find_hw_engine(xe, eci[0]);
- if (XE_IOCTL_DBG(xe, !hwe))
- return -EINVAL;
-
- vm = xe_vm_lookup(xef, args->vm_id);
- if (XE_IOCTL_DBG(xe, !vm))
- return -ENOENT;
-
- err = down_read_interruptible(&vm->lock);
- if (err) {
- xe_vm_put(vm);
- return err;
- }
-
- if (XE_IOCTL_DBG(xe, xe_vm_is_closed_or_banned(vm))) {
- up_read(&vm->lock);
- xe_vm_put(vm);
- return -ENOENT;
- }
-
- q = xe_exec_queue_create(xe, vm, logical_mask,
- args->width, hwe, 0,
- args->extensions);
- up_read(&vm->lock);
- xe_vm_put(vm);
- if (IS_ERR(q))
- return PTR_ERR(q);
-
- if (xe_vm_in_preempt_fence_mode(vm)) {
- q->lr.context = dma_fence_context_alloc(1);
-
- err = xe_vm_add_compute_exec_queue(vm, q);
- if (XE_IOCTL_DBG(xe, err))
- goto put_exec_queue;
- }
- }
-
- mutex_lock(&xef->exec_queue.lock);
- err = xa_alloc(&xef->exec_queue.xa, &id, q, xa_limit_32b, GFP_KERNEL);
- mutex_unlock(&xef->exec_queue.lock);
- if (err)
- goto kill_exec_queue;
-
- args->exec_queue_id = id;
- q->xef = xe_file_get(xef);
-
- return 0;
-
-kill_exec_queue:
- xe_exec_queue_kill(q);
-put_exec_queue:
- xe_exec_queue_put(q);
- return err;
-}
-
-int xe_exec_queue_get_property_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct xe_device *xe = to_xe_device(dev);
- struct xe_file *xef = to_xe_file(file);
- struct drm_xe_exec_queue_get_property *args = data;
- struct xe_exec_queue *q;
- int ret;
-
- if (XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
- return -EINVAL;
-
- q = xe_exec_queue_lookup(xef, args->exec_queue_id);
- if (XE_IOCTL_DBG(xe, !q))
- return -ENOENT;
-
- switch (args->property) {
- case DRM_XE_EXEC_QUEUE_GET_PROPERTY_BAN:
- args->value = q->ops->reset_status(q);
- ret = 0;
- break;
- default:
- ret = -EINVAL;
- }
-
- xe_exec_queue_put(q);
-
- return ret;
-}
-
-/**
- * xe_exec_queue_is_lr() - Whether an exec_queue is long-running
- * @q: The exec_queue
- *
- * Return: True if the exec_queue is long-running, false otherwise.
- */
-bool xe_exec_queue_is_lr(struct xe_exec_queue *q)
-{
- return q->vm && xe_vm_in_lr_mode(q->vm) &&
- !(q->flags & EXEC_QUEUE_FLAG_VM);
-}
-
-static s32 xe_exec_queue_num_job_inflight(struct xe_exec_queue *q)
-{
- return q->lrc[0]->fence_ctx.next_seqno - xe_lrc_seqno(q->lrc[0]) - 1;
-}
-
-/**
- * xe_exec_queue_ring_full() - Whether an exec_queue's ring is full
- * @q: The exec_queue
- *
- * Return: True if the exec_queue's ring is full, false otherwise.
- */
-bool xe_exec_queue_ring_full(struct xe_exec_queue *q)
-{
- struct xe_lrc *lrc = q->lrc[0];
- s32 max_job = lrc->ring.size / MAX_JOB_SIZE_BYTES;
-
- return xe_exec_queue_num_job_inflight(q) >= max_job;
-}
-
-/**
- * xe_exec_queue_is_idle() - Whether an exec_queue is idle.
- * @q: The exec_queue
- *
- * FIXME: Need to determine what to use as the short-lived
- * timeline lock for the exec_queues, so that the return value
- * of this function becomes more than just an advisory
- * snapshot in time. The timeline lock must protect the
- * seqno from racing submissions on the same exec_queue.
- * Typically vm->resv, but user-created timeline locks use the migrate vm
- * and never grabs the migrate vm->resv so we have a race there.
- *
- * Return: True if the exec_queue is idle, false otherwise.
- */
-bool xe_exec_queue_is_idle(struct xe_exec_queue *q)
-{
- if (xe_exec_queue_is_parallel(q)) {
- int i;
-
- for (i = 0; i < q->width; ++i) {
- if (xe_lrc_seqno(q->lrc[i]) !=
- q->lrc[i]->fence_ctx.next_seqno - 1)
- return false;
- }
-
- return true;
- }
-
- return xe_lrc_seqno(q->lrc[0]) ==
- q->lrc[0]->fence_ctx.next_seqno - 1;
-}
-
-/**
- * xe_exec_queue_update_run_ticks() - Update run time in ticks for this exec queue
- * from hw
- * @q: The exec queue
- *
- * Update the timestamp saved by HW for this exec queue and save run ticks
- * calculated by using the delta from last update.
- */
-void xe_exec_queue_update_run_ticks(struct xe_exec_queue *q)
-{
- struct xe_file *xef;
- struct xe_lrc *lrc;
- u32 old_ts, new_ts;
-
- /*
- * Jobs that are run during driver load may use an exec_queue, but are
- * not associated with a user xe file, so avoid accumulating busyness
- * for kernel specific work.
- */
- if (!q->vm || !q->vm->xef)
- return;
-
- xef = q->vm->xef;
-
- /*
- * Only sample the first LRC. For parallel submission, all of them are
- * scheduled together and we compensate that below by multiplying by
- * width - this may introduce errors if that premise is not true and
- * they don't exit 100% aligned. On the other hand, looping through
- * the LRCs and reading them in different time could also introduce
- * errors.
- */
- lrc = q->lrc[0];
- new_ts = xe_lrc_update_timestamp(lrc, &old_ts);
- xef->run_ticks[q->class] += (new_ts - old_ts) * q->width;
-}
-
-void xe_exec_queue_kill(struct xe_exec_queue *q)
-{
- struct xe_exec_queue *eq = q, *next;
-
- list_for_each_entry_safe(eq, next, &eq->multi_gt_list,
- multi_gt_link) {
- q->ops->kill(eq);
- xe_vm_remove_compute_exec_queue(q->vm, eq);
- }
-
- q->ops->kill(q);
- xe_vm_remove_compute_exec_queue(q->vm, q);
-}
-
-int xe_exec_queue_destroy_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct xe_device *xe = to_xe_device(dev);
- struct xe_file *xef = to_xe_file(file);
- struct drm_xe_exec_queue_destroy *args = data;
- struct xe_exec_queue *q;
-
- if (XE_IOCTL_DBG(xe, args->pad) ||
- XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
- return -EINVAL;
-
- mutex_lock(&xef->exec_queue.lock);
- q = xa_erase(&xef->exec_queue.xa, args->exec_queue_id);
- mutex_unlock(&xef->exec_queue.lock);
- if (XE_IOCTL_DBG(xe, !q))
- return -ENOENT;
-
- xe_exec_queue_kill(q);
-
- trace_xe_exec_queue_close(q);
- xe_exec_queue_put(q);
-
- return 0;
-}
-
-static void xe_exec_queue_last_fence_lockdep_assert(struct xe_exec_queue *q,
- struct xe_vm *vm)
-{
- if (q->flags & EXEC_QUEUE_FLAG_VM)
- lockdep_assert_held(&vm->lock);
- else
- xe_vm_assert_held(vm);
-}
-
-/**
- * xe_exec_queue_last_fence_put() - Drop ref to last fence
- * @q: The exec queue
- * @vm: The VM the engine does a bind or exec for
- */
-void xe_exec_queue_last_fence_put(struct xe_exec_queue *q, struct xe_vm *vm)
-{
- xe_exec_queue_last_fence_lockdep_assert(q, vm);
-
- if (q->last_fence) {
- dma_fence_put(q->last_fence);
- q->last_fence = NULL;
- }
-}
-
-/**
- * xe_exec_queue_last_fence_put_unlocked() - Drop ref to last fence unlocked
- * @q: The exec queue
- *
- * Only safe to be called from xe_exec_queue_destroy().
- */
-void xe_exec_queue_last_fence_put_unlocked(struct xe_exec_queue *q)
-{
- if (q->last_fence) {
- dma_fence_put(q->last_fence);
- q->last_fence = NULL;
- }
-}
-
-/**
- * xe_exec_queue_last_fence_get() - Get last fence
- * @q: The exec queue
- * @vm: The VM the engine does a bind or exec for
- *
- * Get last fence, takes a ref
- *
- * Returns: last fence if not signaled, dma fence stub if signaled
- */
-struct dma_fence *xe_exec_queue_last_fence_get(struct xe_exec_queue *q,
- struct xe_vm *vm)
-{
- struct dma_fence *fence;
-
- xe_exec_queue_last_fence_lockdep_assert(q, vm);
-
- if (q->last_fence &&
- test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &q->last_fence->flags))
- xe_exec_queue_last_fence_put(q, vm);
-
- fence = q->last_fence ? q->last_fence : dma_fence_get_stub();
- dma_fence_get(fence);
- return fence;
-}
-
-/**
- * xe_exec_queue_last_fence_set() - Set last fence
- * @q: The exec queue
- * @vm: The VM the engine does a bind or exec for
- * @fence: The fence
- *
- * Set the last fence for the engine. Increases reference count for fence, when
- * closing engine xe_exec_queue_last_fence_put should be called.
- */
-void xe_exec_queue_last_fence_set(struct xe_exec_queue *q, struct xe_vm *vm,
- struct dma_fence *fence)
-{
- xe_exec_queue_last_fence_lockdep_assert(q, vm);
-
- xe_exec_queue_last_fence_put(q, vm);
- q->last_fence = dma_fence_get(fence);
-}
diff --git a/rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/postimage b/rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/postimage
deleted file mode 100644
index 710b1e2170c1..000000000000
--- a/rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/postimage
+++ /dev/null
@@ -1,451 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.probe_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.probe_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.probe_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
- struct intel_display *display = &xe->display;
-
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(display);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- struct intel_display *display = &xe->display;
- int err;
-
- if (!xe->info.probe_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(display);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(display);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.probe_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.probe_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- struct intel_display *display = &xe->display;
-
- if (!xe->info.probe_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(display);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.probe_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-/* TODO: System and runtime suspend/resume sequences will be sanitized as a follow-up. */
-void xe_display_pm_runtime_suspend(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- if (xe->d3cold.allowed)
- xe_display_pm_suspend(xe, true);
-
- intel_hpd_poll_enable(xe);
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- struct intel_display *display = &xe->display;
- bool s2idle = suspend_to_idle();
- if (!xe->info.probe_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
- if (!runtime && has_display(xe)) {
- drm_kms_helper_poll_disable(&xe->drm);
- intel_display_driver_disable_user_access(xe);
- intel_display_driver_suspend(xe);
- }
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
- if (!runtime && has_display(xe))
- intel_display_driver_suspend_access(xe);
-
- intel_encoder_suspend_all(&xe->display);
-
- intel_opregion_suspend(display, s2idle ? PCI_D1 : PCI_D3cold);
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.probe_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_runtime_resume(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_hpd_poll_disable(xe);
-
- if (xe->d3cold.allowed)
- xe_display_pm_resume(xe, true);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- struct intel_display *display = &xe->display;
-
- if (!xe->info.probe_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
- if (!runtime && has_display(xe))
- intel_display_driver_resume_access(xe);
-
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime && has_display(xe)) {
- intel_display_driver_resume(xe);
- drm_kms_helper_poll_enable(&xe->drm);
- intel_display_driver_enable_user_access(xe);
- intel_hpd_poll_disable(xe);
- }
-
- intel_opregion_resume(display);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.probe_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.probe_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/preimage b/rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/preimage
deleted file mode 100644
index 161960d76996..000000000000
--- a/rr-cache/c0d8d1174cf12b663da53bdc6388629683a9b906/preimage
+++ /dev/null
@@ -1,480 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.probe_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.probe_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.probe_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
- struct intel_display *display = &xe->display;
-
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(display);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- struct intel_display *display = &xe->display;
- int err;
-
- if (!xe->info.probe_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(display);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(display);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.probe_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.probe_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- struct intel_display *display = &xe->display;
-
- if (!xe->info.probe_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(display);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.probe_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-/* TODO: System and runtime suspend/resume sequences will be sanitized as a follow-up. */
-void xe_display_pm_runtime_suspend(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- if (xe->d3cold.allowed)
- xe_display_pm_suspend(xe, true);
-
- intel_hpd_poll_enable(xe);
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- struct intel_display *display = &xe->display;
- bool s2idle = suspend_to_idle();
- if (!xe->info.probe_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
-<<<<<<<
- if (!runtime && has_display(xe)) {
- drm_kms_helper_poll_disable(&xe->drm);
- intel_display_driver_disable_user_access(xe);
-=======
- if (has_display(xe)) {
- drm_kms_helper_poll_disable(&xe->drm);
- if (!runtime)
- intel_display_driver_disable_user_access(xe);
- }
-
- if (!runtime)
->>>>>>>
- intel_display_driver_suspend(xe);
- }
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
-<<<<<<<
- if (!runtime && has_display(xe))
- intel_display_driver_suspend_access(xe);
-
- intel_encoder_suspend_all(&xe->display);
-
- intel_opregion_suspend(display, s2idle ? PCI_D1 : PCI_D3cold);
-=======
- if (!runtime && has_display(xe)) {
- intel_display_driver_suspend_access(xe);
- intel_encoder_suspend_all(&xe->display);
- }
-
- intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold);
->>>>>>>
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.probe_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_runtime_resume(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_hpd_poll_disable(xe);
-
- if (xe->d3cold.allowed)
- xe_display_pm_resume(xe, true);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.probe_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- struct intel_display *display = &xe->display;
-
- if (!xe->info.probe_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
- if (!runtime && has_display(xe))
- intel_display_driver_resume_access(xe);
-
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime && has_display(xe)) {
- intel_display_driver_resume(xe);
-<<<<<<<
- drm_kms_helper_poll_enable(&xe->drm);
- intel_display_driver_enable_user_access(xe);
- intel_hpd_poll_disable(xe);
- }
-=======
-
- if (has_display(xe)) {
- drm_kms_helper_poll_enable(&xe->drm);
- if (!runtime)
- intel_display_driver_enable_user_access(xe);
- }
- intel_hpd_poll_disable(xe);
->>>>>>>
-
- intel_opregion_resume(display);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.probe_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.probe_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage b/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage
deleted file mode 100644
index 1e34dee9d205..000000000000
--- a/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage
+++ /dev/null
@@ -1,423 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.enable_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.enable_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(xe);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(xe);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(xe);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.enable_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(xe);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_disable(&xe->drm);
-
- if (!runtime)
- intel_display_driver_suspend(xe);
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
-<<<<<<<
- if (!runtime && has_display(xe)) {
- intel_display_driver_suspend_access(xe);
- intel_encoder_suspend_all(&xe->display);
- }
-=======
- intel_encoder_suspend_all(&xe->display);
->>>>>>>
-
- intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime)
- intel_display_driver_resume(xe);
-
- intel_hpd_poll_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_enable(&xe->drm);
-
- intel_opregion_resume(xe);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.enable_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.1 b/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.1
deleted file mode 100644
index 1e34dee9d205..000000000000
--- a/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.1
+++ /dev/null
@@ -1,423 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.enable_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.enable_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(xe);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(xe);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(xe);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.enable_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(xe);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_disable(&xe->drm);
-
- if (!runtime)
- intel_display_driver_suspend(xe);
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
-<<<<<<<
- if (!runtime && has_display(xe)) {
- intel_display_driver_suspend_access(xe);
- intel_encoder_suspend_all(&xe->display);
- }
-=======
- intel_encoder_suspend_all(&xe->display);
->>>>>>>
-
- intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime)
- intel_display_driver_resume(xe);
-
- intel_hpd_poll_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_enable(&xe->drm);
-
- intel_opregion_resume(xe);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.enable_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.2 b/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.2
deleted file mode 100644
index 1e34dee9d205..000000000000
--- a/rr-cache/d038209860ab485ed4c5dbc38ab24be8e4da6180/preimage.2
+++ /dev/null
@@ -1,423 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2023 Intel Corporation
- */
-
-#include "xe_display.h"
-#include "regs/xe_regs.h"
-
-#include <linux/fb.h>
-
-#include <drm/drm_drv.h>
-#include <drm/drm_managed.h>
-#include <drm/xe_drm.h>
-
-#include "soc/intel_dram.h"
-#include "i915_drv.h" /* FIXME: HAS_DISPLAY() depends on this */
-#include "intel_acpi.h"
-#include "intel_audio.h"
-#include "intel_bw.h"
-#include "intel_display.h"
-#include "intel_display_driver.h"
-#include "intel_display_irq.h"
-#include "intel_display_types.h"
-#include "intel_dmc.h"
-#include "intel_dp.h"
-#include "intel_encoder.h"
-#include "intel_fbdev.h"
-#include "intel_hdcp.h"
-#include "intel_hotplug.h"
-#include "intel_opregion.h"
-#include "xe_module.h"
-
-/* Xe device functions */
-
-static bool has_display(struct xe_device *xe)
-{
- return HAS_DISPLAY(xe);
-}
-
-/**
- * xe_display_driver_probe_defer - Detect if we need to wait for other drivers
- * early on
- * @pdev: PCI device
- *
- * Returns: true if probe needs to be deferred, false otherwise
- */
-bool xe_display_driver_probe_defer(struct pci_dev *pdev)
-{
- if (!xe_modparam.enable_display)
- return 0;
-
- return intel_display_driver_probe_defer(pdev);
-}
-
-/**
- * xe_display_driver_set_hooks - Add driver flags and hooks for display
- * @driver: DRM device driver
- *
- * Set features and function hooks in @driver that are needed for driving the
- * display IP. This sets the driver's capability of driving display, regardless
- * if the device has it enabled
- */
-void xe_display_driver_set_hooks(struct drm_driver *driver)
-{
- if (!xe_modparam.enable_display)
- return;
-
- driver->driver_features |= DRIVER_MODESET | DRIVER_ATOMIC;
-}
-
-static void unset_display_features(struct xe_device *xe)
-{
- xe->drm.driver_features &= ~(DRIVER_MODESET | DRIVER_ATOMIC);
-}
-
-static void display_destroy(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- destroy_workqueue(xe->display.hotplug.dp_wq);
-}
-
-/**
- * xe_display_create - create display struct
- * @xe: XE device instance
- *
- * Initialize all fields used by the display part.
- *
- * TODO: once everything can be inside a single struct, make the struct opaque
- * to the rest of xe and return it to be xe->display.
- *
- * Returns: 0 on success
- */
-int xe_display_create(struct xe_device *xe)
-{
- spin_lock_init(&xe->display.fb_tracking.lock);
-
- xe->display.hotplug.dp_wq = alloc_ordered_workqueue("xe-dp", 0);
-
- return drmm_add_action_or_reset(&xe->drm, display_destroy, NULL);
-}
-
-static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
-{
- struct xe_device *xe = to_xe_device(dev);
-
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_cleanup(xe);
-}
-
-int xe_display_init_nommio(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- /* Fake uncore lock */
- spin_lock_init(&xe->uncore.lock);
-
- /* This must be called before any calls to HAS_PCH_* */
- intel_detect_pch(xe);
-
- return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
-}
-
-static void xe_display_fini_noirq(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_noirq(xe);
- intel_opregion_cleanup(xe);
-}
-
-int xe_display_init_noirq(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- intel_display_driver_early_probe(xe);
-
- /* Early display init.. */
- intel_opregion_setup(xe);
-
- /*
- * Fill the dram structure to get the system dram info. This will be
- * used for memory latency calculation.
- */
- intel_dram_detect(xe);
-
- intel_bw_init_hw(xe);
-
- intel_display_device_info_runtime_init(xe);
-
- err = intel_display_driver_probe_noirq(xe);
- if (err) {
- intel_opregion_cleanup(xe);
- return err;
- }
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noirq, xe);
-}
-
-static void xe_display_fini_noaccel(void *arg)
-{
- struct xe_device *xe = arg;
-
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove_nogem(xe);
-}
-
-int xe_display_init_noaccel(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- return 0;
-
- err = intel_display_driver_probe_nogem(xe);
- if (err)
- return err;
-
- return devm_add_action_or_reset(xe->drm.dev, xe_display_fini_noaccel, xe);
-}
-
-int xe_display_init(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return 0;
-
- return intel_display_driver_probe(xe);
-}
-
-void xe_display_fini(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_hpd_poll_fini(xe);
-
- intel_hdcp_component_fini(xe);
- intel_audio_deinit(xe);
-}
-
-void xe_display_register(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_register(xe);
- intel_register_dsm_handler();
- intel_power_domains_enable(xe);
-}
-
-void xe_display_unregister(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_unregister_dsm_handler();
- intel_power_domains_disable(xe);
- intel_display_driver_unregister(xe);
-}
-
-void xe_display_driver_remove(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_driver_remove(xe);
-}
-
-/* IRQ-related functions */
-
-void xe_display_irq_handler(struct xe_device *xe, u32 master_ctl)
-{
- if (!xe->info.enable_display)
- return;
-
- if (master_ctl & DISPLAY_IRQ)
- gen11_display_irq_handler(xe);
-}
-
-void xe_display_irq_enable(struct xe_device *xe, u32 gu_misc_iir)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gu_misc_iir & GU_MISC_GSE)
- intel_opregion_asle_intr(xe);
-}
-
-void xe_display_irq_reset(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- gen11_display_irq_reset(xe);
-}
-
-void xe_display_irq_postinstall(struct xe_device *xe, struct xe_gt *gt)
-{
- if (!xe->info.enable_display)
- return;
-
- if (gt->info.id == XE_GT0)
- gen11_de_irq_postinstall(xe);
-}
-
-static bool suspend_to_idle(void)
-{
-#if IS_ENABLED(CONFIG_ACPI_SLEEP)
- if (acpi_target_system_state() < ACPI_STATE_S3)
- return true;
-#endif
- return false;
-}
-
-static void xe_display_flush_cleanup_work(struct xe_device *xe)
-{
- struct intel_crtc *crtc;
-
- for_each_intel_crtc(&xe->drm, crtc) {
- struct drm_crtc_commit *commit;
-
- spin_lock(&crtc->base.commit_lock);
- commit = list_first_entry_or_null(&crtc->base.commit_list,
- struct drm_crtc_commit, commit_entry);
- if (commit)
- drm_crtc_commit_get(commit);
- spin_unlock(&crtc->base.commit_lock);
-
- if (commit) {
- wait_for_completion(&commit->cleanup_done);
- drm_crtc_commit_put(commit);
- }
- }
-}
-
-void xe_display_pm_suspend(struct xe_device *xe, bool runtime)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- /*
- * We do a lot of poking in a lot of registers, make sure they work
- * properly.
- */
- intel_power_domains_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_disable(&xe->drm);
-
- if (!runtime)
- intel_display_driver_suspend(xe);
-
- xe_display_flush_cleanup_work(xe);
-
- intel_dp_mst_suspend(xe);
-
- intel_hpd_cancel_work(xe);
-
-<<<<<<<
- if (!runtime && has_display(xe)) {
- intel_display_driver_suspend_access(xe);
- intel_encoder_suspend_all(&xe->display);
- }
-=======
- intel_encoder_suspend_all(&xe->display);
->>>>>>>
-
- intel_opregion_suspend(xe, s2idle ? PCI_D1 : PCI_D3cold);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_SUSPENDED, true);
-
- intel_dmc_suspend(xe);
-}
-
-void xe_display_pm_suspend_late(struct xe_device *xe)
-{
- bool s2idle = suspend_to_idle();
- if (!xe->info.enable_display)
- return;
-
- intel_power_domains_suspend(xe, s2idle);
-
- intel_display_power_suspend_late(xe);
-}
-
-void xe_display_pm_resume_early(struct xe_device *xe)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_display_power_resume_early(xe);
-
- intel_power_domains_resume(xe);
-}
-
-void xe_display_pm_resume(struct xe_device *xe, bool runtime)
-{
- if (!xe->info.enable_display)
- return;
-
- intel_dmc_resume(xe);
-
- if (has_display(xe))
- drm_mode_config_reset(&xe->drm);
-
- intel_display_driver_init_hw(xe);
- intel_hpd_init(xe);
-
- /* MST sideband requires HPD interrupts enabled */
- intel_dp_mst_resume(xe);
- if (!runtime)
- intel_display_driver_resume(xe);
-
- intel_hpd_poll_disable(xe);
- if (has_display(xe))
- drm_kms_helper_poll_enable(&xe->drm);
-
- intel_opregion_resume(xe);
-
- intel_fbdev_set_suspend(&xe->drm, FBINFO_STATE_RUNNING, false);
-
- intel_power_domains_enable(xe);
-}
-
-static void display_device_remove(struct drm_device *dev, void *arg)
-{
- struct xe_device *xe = arg;
-
- intel_display_device_remove(xe);
-}
-
-int xe_display_probe(struct xe_device *xe)
-{
- int err;
-
- if (!xe->info.enable_display)
- goto no_display;
-
- intel_display_device_probe(xe);
-
- err = drmm_add_action_or_reset(&xe->drm, display_device_remove, xe);
- if (err)
- return err;
-
- if (has_display(xe))
- return 0;
-
-no_display:
- xe->info.enable_display = false;
- unset_display_features(xe);
- return 0;
-}
diff --git a/rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.10 b/rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.10
deleted file mode 100644
index 110e70f7ee7b..000000000000
--- a/rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.10
+++ /dev/null
@@ -1,2244 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2022 Intel Corporation
- */
-
-#include "xe_guc_submit.h"
-
-#include <linux/bitfield.h>
-#include <linux/bitmap.h>
-#include <linux/circ_buf.h>
-#include <linux/delay.h>
-#include <linux/dma-fence-array.h>
-#include <linux/math64.h>
-
-#include <drm/drm_managed.h>
-
-#include "abi/guc_actions_abi.h"
-#include "abi/guc_klvs_abi.h"
-#include "regs/xe_lrc_layout.h"
-#include "xe_assert.h"
-#include "xe_devcoredump.h"
-#include "xe_device.h"
-#include "xe_exec_queue.h"
-#include "xe_force_wake.h"
-#include "xe_gpu_scheduler.h"
-#include "xe_gt.h"
-#include "xe_gt_clock.h"
-#include "xe_gt_printk.h"
-#include "xe_guc.h"
-#include "xe_guc_ct.h"
-#include "xe_guc_exec_queue_types.h"
-#include "xe_guc_id_mgr.h"
-#include "xe_guc_submit_types.h"
-#include "xe_hw_engine.h"
-#include "xe_hw_fence.h"
-#include "xe_lrc.h"
-#include "xe_macros.h"
-#include "xe_map.h"
-#include "xe_mocs.h"
-#include "xe_pm.h"
-#include "xe_ring_ops_types.h"
-#include "xe_sched_job.h"
-#include "xe_trace.h"
-#include "xe_vm.h"
-
-static struct xe_guc *
-exec_queue_to_guc(struct xe_exec_queue *q)
-{
- return &q->gt->uc.guc;
-}
-
-/*
- * Helpers for engine state, using an atomic as some of the bits can transition
- * as the same time (e.g. a suspend can be happning at the same time as schedule
- * engine done being processed).
- */
-#define EXEC_QUEUE_STATE_REGISTERED (1 << 0)
-#define EXEC_QUEUE_STATE_ENABLED (1 << 1)
-#define EXEC_QUEUE_STATE_PENDING_ENABLE (1 << 2)
-#define EXEC_QUEUE_STATE_PENDING_DISABLE (1 << 3)
-#define EXEC_QUEUE_STATE_DESTROYED (1 << 4)
-#define EXEC_QUEUE_STATE_SUSPENDED (1 << 5)
-#define EXEC_QUEUE_STATE_RESET (1 << 6)
-#define EXEC_QUEUE_STATE_KILLED (1 << 7)
-#define EXEC_QUEUE_STATE_WEDGED (1 << 8)
-#define EXEC_QUEUE_STATE_BANNED (1 << 9)
-#define EXEC_QUEUE_STATE_CHECK_TIMEOUT (1 << 10)
-#define EXEC_QUEUE_STATE_EXTRA_REF (1 << 11)
-
-static bool exec_queue_registered(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_REGISTERED;
-}
-
-static void set_exec_queue_registered(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_REGISTERED, &q->guc->state);
-}
-
-static void clear_exec_queue_registered(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_REGISTERED, &q->guc->state);
-}
-
-static bool exec_queue_enabled(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_ENABLED;
-}
-
-static void set_exec_queue_enabled(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_ENABLED, &q->guc->state);
-}
-
-static void clear_exec_queue_enabled(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_ENABLED, &q->guc->state);
-}
-
-static bool exec_queue_pending_enable(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_PENDING_ENABLE;
-}
-
-static void set_exec_queue_pending_enable(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_PENDING_ENABLE, &q->guc->state);
-}
-
-static void clear_exec_queue_pending_enable(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_PENDING_ENABLE, &q->guc->state);
-}
-
-static bool exec_queue_pending_disable(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_PENDING_DISABLE;
-}
-
-static void set_exec_queue_pending_disable(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_PENDING_DISABLE, &q->guc->state);
-}
-
-static void clear_exec_queue_pending_disable(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_PENDING_DISABLE, &q->guc->state);
-}
-
-static bool exec_queue_destroyed(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_DESTROYED;
-}
-
-static void set_exec_queue_destroyed(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_DESTROYED, &q->guc->state);
-}
-
-static bool exec_queue_banned(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_BANNED;
-}
-
-static void set_exec_queue_banned(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_BANNED, &q->guc->state);
-}
-
-static bool exec_queue_suspended(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_SUSPENDED;
-}
-
-static void set_exec_queue_suspended(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_SUSPENDED, &q->guc->state);
-}
-
-static void clear_exec_queue_suspended(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_SUSPENDED, &q->guc->state);
-}
-
-static bool exec_queue_reset(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_RESET;
-}
-
-static void set_exec_queue_reset(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_RESET, &q->guc->state);
-}
-
-static bool exec_queue_killed(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_KILLED;
-}
-
-static void set_exec_queue_killed(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_KILLED, &q->guc->state);
-}
-
-static bool exec_queue_wedged(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_WEDGED;
-}
-
-static void set_exec_queue_wedged(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_WEDGED, &q->guc->state);
-}
-
-static bool exec_queue_check_timeout(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_CHECK_TIMEOUT;
-}
-
-static void set_exec_queue_check_timeout(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_CHECK_TIMEOUT, &q->guc->state);
-}
-
-static void clear_exec_queue_check_timeout(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_CHECK_TIMEOUT, &q->guc->state);
-}
-
-static bool exec_queue_extra_ref(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_EXTRA_REF;
-}
-
-static void set_exec_queue_extra_ref(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_EXTRA_REF, &q->guc->state);
-}
-
-static bool exec_queue_killed_or_banned_or_wedged(struct xe_exec_queue *q)
-{
- return (atomic_read(&q->guc->state) &
- (EXEC_QUEUE_STATE_WEDGED | EXEC_QUEUE_STATE_KILLED |
- EXEC_QUEUE_STATE_BANNED));
-}
-
-#ifdef CONFIG_PROVE_LOCKING
-static int alloc_submit_wq(struct xe_guc *guc)
-{
- int i;
-
- for (i = 0; i < NUM_SUBMIT_WQ; ++i) {
- guc->submission_state.submit_wq_pool[i] =
- alloc_ordered_workqueue("submit_wq", 0);
- if (!guc->submission_state.submit_wq_pool[i])
- goto err_free;
- }
-
- return 0;
-
-err_free:
- while (i)
- destroy_workqueue(guc->submission_state.submit_wq_pool[--i]);
-
- return -ENOMEM;
-}
-
-static void free_submit_wq(struct xe_guc *guc)
-{
- int i;
-
- for (i = 0; i < NUM_SUBMIT_WQ; ++i)
- destroy_workqueue(guc->submission_state.submit_wq_pool[i]);
-}
-
-static struct workqueue_struct *get_submit_wq(struct xe_guc *guc)
-{
- int idx = guc->submission_state.submit_wq_idx++ % NUM_SUBMIT_WQ;
-
- return guc->submission_state.submit_wq_pool[idx];
-}
-#else
-static int alloc_submit_wq(struct xe_guc *guc)
-{
- return 0;
-}
-
-static void free_submit_wq(struct xe_guc *guc)
-{
-
-}
-
-static struct workqueue_struct *get_submit_wq(struct xe_guc *guc)
-{
- return NULL;
-}
-#endif
-
-static void guc_submit_fini(struct drm_device *drm, void *arg)
-{
- struct xe_guc *guc = arg;
-
- xa_destroy(&guc->submission_state.exec_queue_lookup);
- free_submit_wq(guc);
-}
-
-static void guc_submit_wedged_fini(void *arg)
-{
- struct xe_guc *guc = arg;
- struct xe_exec_queue *q;
- unsigned long index;
-
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- if (exec_queue_wedged(q))
- xe_exec_queue_put(q);
-}
-
-static const struct xe_exec_queue_ops guc_exec_queue_ops;
-
-static void primelockdep(struct xe_guc *guc)
-{
- if (!IS_ENABLED(CONFIG_LOCKDEP))
- return;
-
- fs_reclaim_acquire(GFP_KERNEL);
-
- mutex_lock(&guc->submission_state.lock);
- mutex_unlock(&guc->submission_state.lock);
-
- fs_reclaim_release(GFP_KERNEL);
-}
-
-/**
- * xe_guc_submit_init() - Initialize GuC submission.
- * @guc: the &xe_guc to initialize
- * @num_ids: number of GuC context IDs to use
- *
- * The bare-metal or PF driver can pass ~0 as &num_ids to indicate that all
- * GuC context IDs supported by the GuC firmware should be used for submission.
- *
- * Only VF drivers will have to provide explicit number of GuC context IDs
- * that they can use for submission.
- *
- * Return: 0 on success or a negative error code on failure.
- */
-int xe_guc_submit_init(struct xe_guc *guc, unsigned int num_ids)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_gt *gt = guc_to_gt(guc);
- int err;
-
- err = drmm_mutex_init(&xe->drm, &guc->submission_state.lock);
- if (err)
- return err;
-
- err = xe_guc_id_mgr_init(&guc->submission_state.idm, num_ids);
- if (err)
- return err;
-
- err = alloc_submit_wq(guc);
- if (err)
- return err;
-
- gt->exec_queue_ops = &guc_exec_queue_ops;
-
- xa_init(&guc->submission_state.exec_queue_lookup);
-
- primelockdep(guc);
-
- return drmm_add_action_or_reset(&xe->drm, guc_submit_fini, guc);
-}
-
-static void __release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q, u32 xa_count)
-{
- int i;
-
- lockdep_assert_held(&guc->submission_state.lock);
-
- for (i = 0; i < xa_count; ++i)
- xa_erase(&guc->submission_state.exec_queue_lookup, q->guc->id + i);
-
- xe_guc_id_mgr_release_locked(&guc->submission_state.idm,
- q->guc->id, q->width);
-}
-
-static int alloc_guc_id(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- int ret;
- void *ptr;
- int i;
-
- /*
- * Must use GFP_NOWAIT as this lock is in the dma fence signalling path,
- * worse case user gets -ENOMEM on engine create and has to try again.
- *
- * FIXME: Have caller pre-alloc or post-alloc /w GFP_KERNEL to prevent
- * failure.
- */
- lockdep_assert_held(&guc->submission_state.lock);
-
- ret = xe_guc_id_mgr_reserve_locked(&guc->submission_state.idm,
- q->width);
- if (ret < 0)
- return ret;
-
- q->guc->id = ret;
-
- for (i = 0; i < q->width; ++i) {
- ptr = xa_store(&guc->submission_state.exec_queue_lookup,
- q->guc->id + i, q, GFP_NOWAIT);
- if (IS_ERR(ptr)) {
- ret = PTR_ERR(ptr);
- goto err_release;
- }
- }
-
- return 0;
-
-err_release:
- __release_guc_id(guc, q, i);
-
- return ret;
-}
-
-static void release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- mutex_lock(&guc->submission_state.lock);
- __release_guc_id(guc, q, q->width);
- mutex_unlock(&guc->submission_state.lock);
-}
-
-struct exec_queue_policy {
- u32 count;
- struct guc_update_exec_queue_policy h2g;
-};
-
-static u32 __guc_exec_queue_policy_action_size(struct exec_queue_policy *policy)
-{
- size_t bytes = sizeof(policy->h2g.header) +
- (sizeof(policy->h2g.klv[0]) * policy->count);
-
- return bytes / sizeof(u32);
-}
-
-static void __guc_exec_queue_policy_start_klv(struct exec_queue_policy *policy,
- u16 guc_id)
-{
- policy->h2g.header.action =
- XE_GUC_ACTION_HOST2GUC_UPDATE_CONTEXT_POLICIES;
- policy->h2g.header.guc_id = guc_id;
- policy->count = 0;
-}
-
-#define MAKE_EXEC_QUEUE_POLICY_ADD(func, id) \
-static void __guc_exec_queue_policy_add_##func(struct exec_queue_policy *policy, \
- u32 data) \
-{ \
- XE_WARN_ON(policy->count >= GUC_CONTEXT_POLICIES_KLV_NUM_IDS); \
-\
- policy->h2g.klv[policy->count].kl = \
- FIELD_PREP(GUC_KLV_0_KEY, \
- GUC_CONTEXT_POLICIES_KLV_ID_##id) | \
- FIELD_PREP(GUC_KLV_0_LEN, 1); \
- policy->h2g.klv[policy->count].value = data; \
- policy->count++; \
-}
-
-MAKE_EXEC_QUEUE_POLICY_ADD(execution_quantum, EXECUTION_QUANTUM)
-MAKE_EXEC_QUEUE_POLICY_ADD(preemption_timeout, PREEMPTION_TIMEOUT)
-MAKE_EXEC_QUEUE_POLICY_ADD(priority, SCHEDULING_PRIORITY)
-#undef MAKE_EXEC_QUEUE_POLICY_ADD
-
-static const int xe_exec_queue_prio_to_guc[] = {
- [XE_EXEC_QUEUE_PRIORITY_LOW] = GUC_CLIENT_PRIORITY_NORMAL,
- [XE_EXEC_QUEUE_PRIORITY_NORMAL] = GUC_CLIENT_PRIORITY_KMD_NORMAL,
- [XE_EXEC_QUEUE_PRIORITY_HIGH] = GUC_CLIENT_PRIORITY_HIGH,
- [XE_EXEC_QUEUE_PRIORITY_KERNEL] = GUC_CLIENT_PRIORITY_KMD_HIGH,
-};
-
-static void init_policies(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- struct exec_queue_policy policy;
- struct xe_device *xe = guc_to_xe(guc);
- enum xe_exec_queue_priority prio = q->sched_props.priority;
- u32 timeslice_us = q->sched_props.timeslice_us;
- u32 preempt_timeout_us = q->sched_props.preempt_timeout_us;
-
- xe_assert(xe, exec_queue_registered(q));
-
- __guc_exec_queue_policy_start_klv(&policy, q->guc->id);
- __guc_exec_queue_policy_add_priority(&policy, xe_exec_queue_prio_to_guc[prio]);
- __guc_exec_queue_policy_add_execution_quantum(&policy, timeslice_us);
- __guc_exec_queue_policy_add_preemption_timeout(&policy, preempt_timeout_us);
-
- xe_guc_ct_send(&guc->ct, (u32 *)&policy.h2g,
- __guc_exec_queue_policy_action_size(&policy), 0, 0);
-}
-
-static void set_min_preemption_timeout(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- struct exec_queue_policy policy;
-
- __guc_exec_queue_policy_start_klv(&policy, q->guc->id);
- __guc_exec_queue_policy_add_preemption_timeout(&policy, 1);
-
- xe_guc_ct_send(&guc->ct, (u32 *)&policy.h2g,
- __guc_exec_queue_policy_action_size(&policy), 0, 0);
-}
-
-#define parallel_read(xe_, map_, field_) \
- xe_map_rd_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
- field_)
-#define parallel_write(xe_, map_, field_, val_) \
- xe_map_wr_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
- field_, val_)
-
-static void __register_mlrc_exec_queue(struct xe_guc *guc,
- struct xe_exec_queue *q,
- struct guc_ctxt_registration_info *info)
-{
-#define MAX_MLRC_REG_SIZE (13 + XE_HW_ENGINE_MAX_INSTANCE * 2)
- struct xe_device *xe = guc_to_xe(guc);
- u32 action[MAX_MLRC_REG_SIZE];
- int len = 0;
- int i;
-
- xe_assert(xe, xe_exec_queue_is_parallel(q));
-
- action[len++] = XE_GUC_ACTION_REGISTER_CONTEXT_MULTI_LRC;
- action[len++] = info->flags;
- action[len++] = info->context_idx;
- action[len++] = info->engine_class;
- action[len++] = info->engine_submit_mask;
- action[len++] = info->wq_desc_lo;
- action[len++] = info->wq_desc_hi;
- action[len++] = info->wq_base_lo;
- action[len++] = info->wq_base_hi;
- action[len++] = info->wq_size;
- action[len++] = q->width;
- action[len++] = info->hwlrca_lo;
- action[len++] = info->hwlrca_hi;
-
- for (i = 1; i < q->width; ++i) {
- struct xe_lrc *lrc = q->lrc[i];
-
- action[len++] = lower_32_bits(xe_lrc_descriptor(lrc));
- action[len++] = upper_32_bits(xe_lrc_descriptor(lrc));
- }
-
- xe_assert(xe, len <= MAX_MLRC_REG_SIZE);
-#undef MAX_MLRC_REG_SIZE
-
- xe_guc_ct_send(&guc->ct, action, len, 0, 0);
-}
-
-static void __register_exec_queue(struct xe_guc *guc,
- struct guc_ctxt_registration_info *info)
-{
- u32 action[] = {
- XE_GUC_ACTION_REGISTER_CONTEXT,
- info->flags,
- info->context_idx,
- info->engine_class,
- info->engine_submit_mask,
- info->wq_desc_lo,
- info->wq_desc_hi,
- info->wq_base_lo,
- info->wq_base_hi,
- info->wq_size,
- info->hwlrca_lo,
- info->hwlrca_hi,
- };
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action), 0, 0);
-}
-
-static void register_exec_queue(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_lrc *lrc = q->lrc[0];
- struct guc_ctxt_registration_info info;
-
- xe_assert(xe, !exec_queue_registered(q));
-
- memset(&info, 0, sizeof(info));
- info.context_idx = q->guc->id;
- info.engine_class = xe_engine_class_to_guc_class(q->class);
- info.engine_submit_mask = q->logical_mask;
- info.hwlrca_lo = lower_32_bits(xe_lrc_descriptor(lrc));
- info.hwlrca_hi = upper_32_bits(xe_lrc_descriptor(lrc));
- info.flags = CONTEXT_REGISTRATION_FLAG_KMD;
-
- if (xe_exec_queue_is_parallel(q)) {
- u64 ggtt_addr = xe_lrc_parallel_ggtt_addr(lrc);
- struct iosys_map map = xe_lrc_parallel_map(lrc);
-
- info.wq_desc_lo = lower_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq_desc));
- info.wq_desc_hi = upper_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq_desc));
- info.wq_base_lo = lower_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq[0]));
- info.wq_base_hi = upper_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq[0]));
- info.wq_size = WQ_SIZE;
-
- q->guc->wqi_head = 0;
- q->guc->wqi_tail = 0;
- xe_map_memset(xe, &map, 0, 0, PARALLEL_SCRATCH_SIZE - WQ_SIZE);
- parallel_write(xe, map, wq_desc.wq_status, WQ_STATUS_ACTIVE);
- }
-
- /*
- * We must keep a reference for LR engines if engine is registered with
- * the GuC as jobs signal immediately and can't destroy an engine if the
- * GuC has a reference to it.
- */
- if (xe_exec_queue_is_lr(q))
- xe_exec_queue_get(q);
-
- set_exec_queue_registered(q);
- trace_xe_exec_queue_register(q);
- if (xe_exec_queue_is_parallel(q))
- __register_mlrc_exec_queue(guc, q, &info);
- else
- __register_exec_queue(guc, &info);
- init_policies(guc, q);
-}
-
-static u32 wq_space_until_wrap(struct xe_exec_queue *q)
-{
- return (WQ_SIZE - q->guc->wqi_tail);
-}
-
-static int wq_wait_for_space(struct xe_exec_queue *q, u32 wqi_size)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
- unsigned int sleep_period_ms = 1;
-
-#define AVAILABLE_SPACE \
- CIRC_SPACE(q->guc->wqi_tail, q->guc->wqi_head, WQ_SIZE)
- if (wqi_size > AVAILABLE_SPACE) {
-try_again:
- q->guc->wqi_head = parallel_read(xe, map, wq_desc.head);
- if (wqi_size > AVAILABLE_SPACE) {
- if (sleep_period_ms == 1024) {
- xe_gt_reset_async(q->gt);
- return -ENODEV;
- }
-
- msleep(sleep_period_ms);
- sleep_period_ms <<= 1;
- goto try_again;
- }
- }
-#undef AVAILABLE_SPACE
-
- return 0;
-}
-
-static int wq_noop_append(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
- u32 len_dw = wq_space_until_wrap(q) / sizeof(u32) - 1;
-
- if (wq_wait_for_space(q, wq_space_until_wrap(q)))
- return -ENODEV;
-
- xe_assert(xe, FIELD_FIT(WQ_LEN_MASK, len_dw));
-
- parallel_write(xe, map, wq[q->guc->wqi_tail / sizeof(u32)],
- FIELD_PREP(WQ_TYPE_MASK, WQ_TYPE_NOOP) |
- FIELD_PREP(WQ_LEN_MASK, len_dw));
- q->guc->wqi_tail = 0;
-
- return 0;
-}
-
-static void wq_item_append(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
-#define WQ_HEADER_SIZE 4 /* Includes 1 LRC address too */
- u32 wqi[XE_HW_ENGINE_MAX_INSTANCE + (WQ_HEADER_SIZE - 1)];
- u32 wqi_size = (q->width + (WQ_HEADER_SIZE - 1)) * sizeof(u32);
- u32 len_dw = (wqi_size / sizeof(u32)) - 1;
- int i = 0, j;
-
- if (wqi_size > wq_space_until_wrap(q)) {
- if (wq_noop_append(q))
- return;
- }
- if (wq_wait_for_space(q, wqi_size))
- return;
-
- wqi[i++] = FIELD_PREP(WQ_TYPE_MASK, WQ_TYPE_MULTI_LRC) |
- FIELD_PREP(WQ_LEN_MASK, len_dw);
- wqi[i++] = xe_lrc_descriptor(q->lrc[0]);
- wqi[i++] = FIELD_PREP(WQ_GUC_ID_MASK, q->guc->id) |
- FIELD_PREP(WQ_RING_TAIL_MASK, q->lrc[0]->ring.tail / sizeof(u64));
- wqi[i++] = 0;
- for (j = 1; j < q->width; ++j) {
- struct xe_lrc *lrc = q->lrc[j];
-
- wqi[i++] = lrc->ring.tail / sizeof(u64);
- }
-
- xe_assert(xe, i == wqi_size / sizeof(u32));
-
- iosys_map_incr(&map, offsetof(struct guc_submit_parallel_scratch,
- wq[q->guc->wqi_tail / sizeof(u32)]));
- xe_map_memcpy_to(xe, &map, 0, wqi, wqi_size);
- q->guc->wqi_tail += wqi_size;
- xe_assert(xe, q->guc->wqi_tail <= WQ_SIZE);
-
- xe_device_wmb(xe);
-
- map = xe_lrc_parallel_map(q->lrc[0]);
- parallel_write(xe, map, wq_desc.tail, q->guc->wqi_tail);
-}
-
-#define RESUME_PENDING ~0x0ull
-static void submit_exec_queue(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_lrc *lrc = q->lrc[0];
- u32 action[3];
- u32 g2h_len = 0;
- u32 num_g2h = 0;
- int len = 0;
- bool extra_submit = false;
-
- xe_assert(xe, exec_queue_registered(q));
-
- if (xe_exec_queue_is_parallel(q))
- wq_item_append(q);
- else
- xe_lrc_set_ring_tail(lrc, lrc->ring.tail);
-
- if (exec_queue_suspended(q) && !xe_exec_queue_is_parallel(q))
- return;
-
- if (!exec_queue_enabled(q) && !exec_queue_suspended(q)) {
- action[len++] = XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET;
- action[len++] = q->guc->id;
- action[len++] = GUC_CONTEXT_ENABLE;
- g2h_len = G2H_LEN_DW_SCHED_CONTEXT_MODE_SET;
- num_g2h = 1;
- if (xe_exec_queue_is_parallel(q))
- extra_submit = true;
-
- q->guc->resume_time = RESUME_PENDING;
- set_exec_queue_pending_enable(q);
- set_exec_queue_enabled(q);
- trace_xe_exec_queue_scheduling_enable(q);
- } else {
- action[len++] = XE_GUC_ACTION_SCHED_CONTEXT;
- action[len++] = q->guc->id;
- trace_xe_exec_queue_submit(q);
- }
-
- xe_guc_ct_send(&guc->ct, action, len, g2h_len, num_g2h);
-
- if (extra_submit) {
- len = 0;
- action[len++] = XE_GUC_ACTION_SCHED_CONTEXT;
- action[len++] = q->guc->id;
- trace_xe_exec_queue_submit(q);
-
- xe_guc_ct_send(&guc->ct, action, len, 0, 0);
- }
-}
-
-static struct dma_fence *
-guc_exec_queue_run_job(struct drm_sched_job *drm_job)
-{
- struct xe_sched_job *job = to_xe_sched_job(drm_job);
- struct xe_exec_queue *q = job->q;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- bool lr = xe_exec_queue_is_lr(q);
-
- xe_assert(xe, !(exec_queue_destroyed(q) || exec_queue_pending_disable(q)) ||
- exec_queue_banned(q) || exec_queue_suspended(q));
-
- trace_xe_sched_job_run(job);
-
- if (!exec_queue_killed_or_banned_or_wedged(q) && !xe_sched_job_is_error(job)) {
- if (!exec_queue_registered(q))
- register_exec_queue(q);
- if (!lr) /* LR jobs are emitted in the exec IOCTL */
- q->ring_ops->emit_job(job);
- submit_exec_queue(q);
- }
-
- if (lr) {
- xe_sched_job_set_error(job, -EOPNOTSUPP);
- return NULL;
- } else if (test_and_set_bit(JOB_FLAG_SUBMIT, &job->fence->flags)) {
- return job->fence;
- } else {
- return dma_fence_get(job->fence);
- }
-}
-
-static void guc_exec_queue_free_job(struct drm_sched_job *drm_job)
-{
- struct xe_sched_job *job = to_xe_sched_job(drm_job);
-
- xe_exec_queue_update_run_ticks(job->q);
-
- trace_xe_sched_job_free(job);
- xe_sched_job_put(job);
-}
-
-static int guc_read_stopped(struct xe_guc *guc)
-{
- return atomic_read(&guc->submission_state.stopped);
-}
-
-#define MAKE_SCHED_CONTEXT_ACTION(q, enable_disable) \
- u32 action[] = { \
- XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET, \
- q->guc->id, \
- GUC_CONTEXT_##enable_disable, \
- }
-
-static void disable_scheduling_deregister(struct xe_guc *guc,
- struct xe_exec_queue *q)
-{
- MAKE_SCHED_CONTEXT_ACTION(q, DISABLE);
- struct xe_device *xe = guc_to_xe(guc);
- int ret;
-
- set_min_preemption_timeout(guc, q);
- smp_rmb();
- ret = wait_event_timeout(guc->ct.wq, !exec_queue_pending_enable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret) {
- struct xe_gpu_scheduler *sched = &q->guc->sched;
-
- drm_warn(&xe->drm, "Pending enable failed to respond");
- xe_sched_submission_start(sched);
- xe_gt_reset_async(q->gt);
- xe_sched_tdr_queue_imm(sched);
- return;
- }
-
- clear_exec_queue_enabled(q);
- set_exec_queue_pending_disable(q);
- set_exec_queue_destroyed(q);
- trace_xe_exec_queue_scheduling_disable(q);
-
- /*
- * Reserve space for both G2H here as the 2nd G2H is sent from a G2H
- * handler and we are not allowed to reserved G2H space in handlers.
- */
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_SCHED_CONTEXT_MODE_SET +
- G2H_LEN_DW_DEREGISTER_CONTEXT, 2);
-}
-
-static void xe_guc_exec_queue_trigger_cleanup(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- /** to wakeup xe_wait_user_fence ioctl if exec queue is reset */
- wake_up_all(&xe->ufence_wq);
-
- if (xe_exec_queue_is_lr(q))
- queue_work(guc_to_gt(guc)->ordered_wq, &q->guc->lr_tdr);
- else
- xe_sched_tdr_queue_imm(&q->guc->sched);
-}
-
-/**
- * xe_guc_submit_wedge() - Wedge GuC submission
- * @guc: the GuC object
- *
- * Save exec queue's registered with GuC state by taking a ref to each queue.
- * Register a DRMM handler to drop refs upon driver unload.
- */
-void xe_guc_submit_wedge(struct xe_guc *guc)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- unsigned long index;
- int err;
-
- xe_gt_assert(guc_to_gt(guc), guc_to_xe(guc)->wedged.mode);
-
- err = devm_add_action_or_reset(guc_to_xe(guc)->drm.dev,
- guc_submit_wedged_fini, guc);
- if (err) {
- drm_err(&xe->drm, "Failed to register xe_guc_submit clean-up on wedged.mode=2. Although device is wedged.\n");
- return;
- }
-
- mutex_lock(&guc->submission_state.lock);
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- if (xe_exec_queue_get_unless_zero(q))
- set_exec_queue_wedged(q);
- mutex_unlock(&guc->submission_state.lock);
-}
-
-static bool guc_submit_hint_wedged(struct xe_guc *guc)
-{
- struct xe_device *xe = guc_to_xe(guc);
-
- if (xe->wedged.mode != 2)
- return false;
-
- if (xe_device_wedged(xe))
- return true;
-
- xe_device_declare_wedged(xe);
-
- return true;
-}
-
-static void xe_guc_exec_queue_lr_cleanup(struct work_struct *w)
-{
- struct xe_guc_exec_queue *ge =
- container_of(w, struct xe_guc_exec_queue, lr_tdr);
- struct xe_exec_queue *q = ge->q;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_gpu_scheduler *sched = &ge->sched;
- bool wedged;
-
- xe_assert(xe, xe_exec_queue_is_lr(q));
- trace_xe_exec_queue_lr_cleanup(q);
-
- wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
-
- /* Kill the run_job / process_msg entry points */
- xe_sched_submission_stop(sched);
-
- /*
- * Engine state now mostly stable, disable scheduling / deregister if
- * needed. This cleanup routine might be called multiple times, where
- * the actual async engine deregister drops the final engine ref.
- * Calling disable_scheduling_deregister will mark the engine as
- * destroyed and fire off the CT requests to disable scheduling /
- * deregister, which we only want to do once. We also don't want to mark
- * the engine as pending_disable again as this may race with the
- * xe_guc_deregister_done_handler() which treats it as an unexpected
- * state.
- */
- if (!wedged && exec_queue_registered(q) && !exec_queue_destroyed(q)) {
- struct xe_guc *guc = exec_queue_to_guc(q);
- int ret;
-
- set_exec_queue_banned(q);
- disable_scheduling_deregister(guc, q);
-
- /*
- * Must wait for scheduling to be disabled before signalling
- * any fences, if GT broken the GT reset code should signal us.
- */
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_disable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret) {
- drm_warn(&xe->drm, "Schedule disable failed to respond");
- xe_sched_submission_start(sched);
- xe_gt_reset_async(q->gt);
- return;
- }
- }
-
- xe_sched_submission_start(sched);
-}
-
-#define ADJUST_FIVE_PERCENT(__t) mul_u64_u32_div(__t, 105, 100)
-
-static bool check_timeout(struct xe_exec_queue *q, struct xe_sched_job *job)
-{
- struct xe_gt *gt = guc_to_gt(exec_queue_to_guc(q));
- u32 ctx_timestamp = xe_lrc_ctx_timestamp(q->lrc[0]);
- u32 ctx_job_timestamp = xe_lrc_ctx_job_timestamp(q->lrc[0]);
- u32 timeout_ms = q->sched_props.job_timeout_ms;
- u32 diff;
- u64 running_time_ms;
-
- /*
- * Counter wraps at ~223s at the usual 19.2MHz, be paranoid catch
- * possible overflows with a high timeout.
- */
- xe_gt_assert(gt, timeout_ms < 100 * MSEC_PER_SEC);
-
- if (ctx_timestamp < ctx_job_timestamp)
- diff = ctx_timestamp + U32_MAX - ctx_job_timestamp;
- else
- diff = ctx_timestamp - ctx_job_timestamp;
-
- /*
- * Ensure timeout is within 5% to account for an GuC scheduling latency
- */
- running_time_ms =
- ADJUST_FIVE_PERCENT(xe_gt_clock_interval_to_ms(gt, diff));
-
- xe_gt_dbg(gt,
- "Check job timeout: seqno=%u, lrc_seqno=%u, guc_id=%d, running_time_ms=%llu, timeout_ms=%u, diff=0x%08x",
- xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
- q->guc->id, running_time_ms, timeout_ms, diff);
-
- return running_time_ms >= timeout_ms;
-}
-
-static void enable_scheduling(struct xe_exec_queue *q)
-{
- MAKE_SCHED_CONTEXT_ACTION(q, ENABLE);
- struct xe_guc *guc = exec_queue_to_guc(q);
- int ret;
-
- xe_gt_assert(guc_to_gt(guc), !exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_enable(q));
-
- set_exec_queue_pending_enable(q);
- set_exec_queue_enabled(q);
- trace_xe_exec_queue_scheduling_enable(q);
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_SCHED_CONTEXT_MODE_SET, 1);
-
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_enable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret || guc_read_stopped(guc)) {
- xe_gt_warn(guc_to_gt(guc), "Schedule enable failed to respond");
- set_exec_queue_banned(q);
- xe_gt_reset_async(q->gt);
- xe_sched_tdr_queue_imm(&q->guc->sched);
- }
-}
-
-static void disable_scheduling(struct xe_exec_queue *q, bool immediate)
-{
- MAKE_SCHED_CONTEXT_ACTION(q, DISABLE);
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- xe_gt_assert(guc_to_gt(guc), !exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
-
- if (immediate)
- set_min_preemption_timeout(guc, q);
- clear_exec_queue_enabled(q);
- set_exec_queue_pending_disable(q);
- trace_xe_exec_queue_scheduling_disable(q);
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_SCHED_CONTEXT_MODE_SET, 1);
-}
-
-static void __deregister_exec_queue(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- u32 action[] = {
- XE_GUC_ACTION_DEREGISTER_CONTEXT,
- q->guc->id,
- };
-
- xe_gt_assert(guc_to_gt(guc), !exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_enable(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
-
- set_exec_queue_destroyed(q);
- trace_xe_exec_queue_deregister(q);
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_DEREGISTER_CONTEXT, 1);
-}
-
-static enum drm_gpu_sched_stat
-guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
-{
- struct xe_sched_job *job = to_xe_sched_job(drm_job);
- struct xe_sched_job *tmp_job;
- struct xe_exec_queue *q = job->q;
- struct xe_gpu_scheduler *sched = &q->guc->sched;
- struct xe_guc *guc = exec_queue_to_guc(q);
-<<<<<<<
-=======
- const char *process_name = "no process";
->>>>>>>
- int err = -ETIME;
- int i = 0;
- bool wedged, skip_timeout_check;
-
- /*
- * TDR has fired before free job worker. Common if exec queue
- * immediately closed after last fence signaled.
- */
- if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &job->fence->flags)) {
- guc_exec_queue_free_job(drm_job);
-
- return DRM_GPU_SCHED_STAT_NOMINAL;
- }
-
- /* Kill the run_job entry point */
- xe_sched_submission_stop(sched);
-
- /* Must check all state after stopping scheduler */
- skip_timeout_check = exec_queue_reset(q) ||
- exec_queue_killed_or_banned_or_wedged(q) ||
- exec_queue_destroyed(q);
-
- /* Job hasn't started, can't be timed out */
- if (!skip_timeout_check && !xe_sched_job_started(job))
- goto rearm;
-
- /*
- * XXX: Sampling timeout doesn't work in wedged mode as we have to
- * modify scheduling state to read timestamp. We could read the
- * timestamp from a register to accumulate current running time but this
- * doesn't work for SRIOV. For now assuming timeouts in wedged mode are
- * genuine timeouts.
- */
- wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
-
- /* Engine state now stable, disable scheduling to check timestamp */
- if (!wedged && exec_queue_registered(q)) {
- int ret;
-
- if (exec_queue_reset(q))
- err = -EIO;
-
- if (!exec_queue_destroyed(q)) {
- /*
- * Wait for any pending G2H to flush out before
- * modifying state
- */
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_enable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret || guc_read_stopped(guc))
- goto trigger_reset;
-
- /*
- * Flag communicates to G2H handler that schedule
- * disable originated from a timeout check. The G2H then
- * avoid triggering cleanup or deregistering the exec
- * queue.
- */
- set_exec_queue_check_timeout(q);
- disable_scheduling(q, skip_timeout_check);
- }
-
- /*
- * Must wait for scheduling to be disabled before signalling
- * any fences, if GT broken the GT reset code should signal us.
- *
- * FIXME: Tests can generate a ton of 0x6000 (IOMMU CAT fault
- * error) messages which can cause the schedule disable to get
- * lost. If this occurs, trigger a GT reset to recover.
- */
- smp_rmb();
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_disable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret || guc_read_stopped(guc)) {
-trigger_reset:
- if (!ret)
- xe_gt_warn(guc_to_gt(guc), "Schedule disable failed to respond");
- set_exec_queue_extra_ref(q);
- xe_exec_queue_get(q); /* GT reset owns this */
- set_exec_queue_banned(q);
- xe_gt_reset_async(q->gt);
- xe_sched_tdr_queue_imm(sched);
- goto rearm;
- }
- }
-
- /*
- * Check if job is actually timed out, if so restart job execution and TDR
- */
- if (!wedged && !skip_timeout_check && !check_timeout(q, job) &&
- !exec_queue_reset(q) && exec_queue_registered(q)) {
- clear_exec_queue_check_timeout(q);
- goto sched_enable;
- }
-
-<<<<<<<
- if (q->vm && q->vm->xef) {
- process_name = q->vm->xef->process_name;
- pid = q->vm->xef->pid;
- }
- xe_gt_notice(guc_to_gt(guc), "Timedout job: seqno=%u, lrc_seqno=%u, guc_id=%d, flags=0x%lx in %s [%d]",
- xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
- q->guc->id, q->flags, process_name, pid);
-
-=======
- xe_gt_notice(guc_to_gt(guc), "Timedout job: seqno=%u, lrc_seqno=%u, guc_id=%d, flags=0x%lx",
- xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
- q->guc->id, q->flags);
->>>>>>>
- trace_xe_sched_job_timedout(job);
-
- if (!exec_queue_killed(q))
- xe_devcoredump(job);
-
- /*
- * Kernel jobs should never fail, nor should VM jobs if they do
- * somethings has gone wrong and the GT needs a reset
- */
- xe_gt_WARN(q->gt, q->flags & EXEC_QUEUE_FLAG_KERNEL,
- "Kernel-submitted job timed out\n");
- xe_gt_WARN(q->gt, q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q),
- "VM job timed out on non-killed execqueue\n");
- if (!wedged && (q->flags & EXEC_QUEUE_FLAG_KERNEL ||
- (q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q)))) {
- if (!xe_sched_invalidate_job(job, 2)) {
- clear_exec_queue_check_timeout(q);
- xe_gt_reset_async(q->gt);
- goto rearm;
- }
- }
-
- /* Finish cleaning up exec queue via deregister */
- set_exec_queue_banned(q);
- if (!wedged && exec_queue_registered(q) && !exec_queue_destroyed(q)) {
- set_exec_queue_extra_ref(q);
- xe_exec_queue_get(q);
- __deregister_exec_queue(guc, q);
- }
-
- /* Stop fence signaling */
- xe_hw_fence_irq_stop(q->fence_irq);
-
- /*
- * Fence state now stable, stop / start scheduler which cleans up any
- * fences that are complete
- */
- xe_sched_add_pending_job(sched, job);
- xe_sched_submission_start(sched);
-
- xe_guc_exec_queue_trigger_cleanup(q);
-
- /* Mark all outstanding jobs as bad, thus completing them */
- spin_lock(&sched->base.job_list_lock);
- list_for_each_entry(tmp_job, &sched->base.pending_list, drm.list)
- xe_sched_job_set_error(tmp_job, !i++ ? err : -ECANCELED);
- spin_unlock(&sched->base.job_list_lock);
-
- /* Start fence signaling */
- xe_hw_fence_irq_start(q->fence_irq);
-
- return DRM_GPU_SCHED_STAT_NOMINAL;
-
-sched_enable:
- enable_scheduling(q);
-rearm:
- /*
- * XXX: Ideally want to adjust timeout based on current exection time
- * but there is not currently an easy way to do in DRM scheduler. With
- * some thought, do this in a follow up.
- */
- xe_sched_add_pending_job(sched, job);
- xe_sched_submission_start(sched);
-
- return DRM_GPU_SCHED_STAT_NOMINAL;
-}
-
-static void __guc_exec_queue_fini_async(struct work_struct *w)
-{
- struct xe_guc_exec_queue *ge =
- container_of(w, struct xe_guc_exec_queue, fini_async);
- struct xe_exec_queue *q = ge->q;
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- xe_pm_runtime_get(guc_to_xe(guc));
- trace_xe_exec_queue_destroy(q);
-
- if (xe_exec_queue_is_lr(q))
- cancel_work_sync(&ge->lr_tdr);
- release_guc_id(guc, q);
- xe_sched_entity_fini(&ge->entity);
- xe_sched_fini(&ge->sched);
-
- kfree(ge);
- xe_exec_queue_fini(q);
- xe_pm_runtime_put(guc_to_xe(guc));
-}
-
-static void guc_exec_queue_fini_async(struct xe_exec_queue *q)
-{
- INIT_WORK(&q->guc->fini_async, __guc_exec_queue_fini_async);
-
- /* We must block on kernel engines so slabs are empty on driver unload */
- if (q->flags & EXEC_QUEUE_FLAG_PERMANENT || exec_queue_wedged(q))
- __guc_exec_queue_fini_async(&q->guc->fini_async);
- else
- queue_work(system_wq, &q->guc->fini_async);
-}
-
-static void __guc_exec_queue_fini(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- /*
- * Might be done from within the GPU scheduler, need to do async as we
- * fini the scheduler when the engine is fini'd, the scheduler can't
- * complete fini within itself (circular dependency). Async resolves
- * this we and don't really care when everything is fini'd, just that it
- * is.
- */
- guc_exec_queue_fini_async(q);
-}
-
-static void __guc_exec_queue_process_msg_cleanup(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, !(q->flags & EXEC_QUEUE_FLAG_PERMANENT));
- trace_xe_exec_queue_cleanup_entity(q);
-
- if (exec_queue_registered(q))
- disable_scheduling_deregister(guc, q);
- else
- __guc_exec_queue_fini(guc, q);
-}
-
-static bool guc_exec_queue_allowed_to_change_state(struct xe_exec_queue *q)
-{
- return !exec_queue_killed_or_banned_or_wedged(q) && exec_queue_registered(q);
-}
-
-static void __guc_exec_queue_process_msg_set_sched_props(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- if (guc_exec_queue_allowed_to_change_state(q))
- init_policies(guc, q);
- kfree(msg);
-}
-
-static void suspend_fence_signal(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, exec_queue_suspended(q) || exec_queue_killed(q) ||
- guc_read_stopped(guc));
- xe_assert(xe, q->guc->suspend_pending);
-
- q->guc->suspend_pending = false;
- smp_wmb();
- wake_up(&q->guc->suspend_wait);
-}
-
-static void __guc_exec_queue_process_msg_suspend(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- if (guc_exec_queue_allowed_to_change_state(q) && !exec_queue_suspended(q) &&
- exec_queue_enabled(q)) {
- wait_event(guc->ct.wq, q->guc->resume_time != RESUME_PENDING ||
- guc_read_stopped(guc));
-
- if (!guc_read_stopped(guc)) {
- s64 since_resume_ms =
- ktime_ms_delta(ktime_get(),
- q->guc->resume_time);
- s64 wait_ms = q->vm->preempt.min_run_period_ms -
- since_resume_ms;
-
- if (wait_ms > 0 && q->guc->resume_time)
- msleep(wait_ms);
-
- set_exec_queue_suspended(q);
- disable_scheduling(q, false);
- }
- } else if (q->guc->suspend_pending) {
- set_exec_queue_suspended(q);
- suspend_fence_signal(q);
- }
-}
-
-static void __guc_exec_queue_process_msg_resume(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
-
- if (guc_exec_queue_allowed_to_change_state(q)) {
- q->guc->resume_time = RESUME_PENDING;
- clear_exec_queue_suspended(q);
- enable_scheduling(q);
- } else {
- clear_exec_queue_suspended(q);
- }
-}
-
-#define CLEANUP 1 /* Non-zero values to catch uninitialized msg */
-#define SET_SCHED_PROPS 2
-#define SUSPEND 3
-#define RESUME 4
-
-static void guc_exec_queue_process_msg(struct xe_sched_msg *msg)
-{
- trace_xe_sched_msg_recv(msg);
-
- switch (msg->opcode) {
- case CLEANUP:
- __guc_exec_queue_process_msg_cleanup(msg);
- break;
- case SET_SCHED_PROPS:
- __guc_exec_queue_process_msg_set_sched_props(msg);
- break;
- case SUSPEND:
- __guc_exec_queue_process_msg_suspend(msg);
- break;
- case RESUME:
- __guc_exec_queue_process_msg_resume(msg);
- break;
- default:
- XE_WARN_ON("Unknown message type");
- }
-
- xe_pm_runtime_put(guc_to_xe(exec_queue_to_guc(msg->private_data)));
-}
-
-static const struct drm_sched_backend_ops drm_sched_ops = {
- .run_job = guc_exec_queue_run_job,
- .free_job = guc_exec_queue_free_job,
- .timedout_job = guc_exec_queue_timedout_job,
-};
-
-static const struct xe_sched_backend_ops xe_sched_ops = {
- .process_msg = guc_exec_queue_process_msg,
-};
-
-static int guc_exec_queue_init(struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_guc_exec_queue *ge;
- long timeout;
- int err;
-
- xe_assert(xe, xe_device_uc_enabled(guc_to_xe(guc)));
-
- ge = kzalloc(sizeof(*ge), GFP_KERNEL);
- if (!ge)
- return -ENOMEM;
-
- q->guc = ge;
- ge->q = q;
- init_waitqueue_head(&ge->suspend_wait);
-
- timeout = (q->vm && xe_vm_in_lr_mode(q->vm)) ? MAX_SCHEDULE_TIMEOUT :
- msecs_to_jiffies(q->sched_props.job_timeout_ms);
- err = xe_sched_init(&ge->sched, &drm_sched_ops, &xe_sched_ops,
- get_submit_wq(guc),
- q->lrc[0]->ring.size / MAX_JOB_SIZE_BYTES, 64,
- timeout, guc_to_gt(guc)->ordered_wq, NULL,
- q->name, gt_to_xe(q->gt)->drm.dev);
- if (err)
- goto err_free;
-
- sched = &ge->sched;
- err = xe_sched_entity_init(&ge->entity, sched);
- if (err)
- goto err_sched;
-
- if (xe_exec_queue_is_lr(q))
- INIT_WORK(&q->guc->lr_tdr, xe_guc_exec_queue_lr_cleanup);
-
- mutex_lock(&guc->submission_state.lock);
-
- err = alloc_guc_id(guc, q);
- if (err)
- goto err_entity;
-
- q->entity = &ge->entity;
-
- if (guc_read_stopped(guc))
- xe_sched_stop(sched);
-
- mutex_unlock(&guc->submission_state.lock);
-
- xe_exec_queue_assign_name(q, q->guc->id);
-
- trace_xe_exec_queue_create(q);
-
- return 0;
-
-err_entity:
- mutex_unlock(&guc->submission_state.lock);
- xe_sched_entity_fini(&ge->entity);
-err_sched:
- xe_sched_fini(&ge->sched);
-err_free:
- kfree(ge);
-
- return err;
-}
-
-static void guc_exec_queue_kill(struct xe_exec_queue *q)
-{
- trace_xe_exec_queue_kill(q);
- set_exec_queue_killed(q);
- xe_guc_exec_queue_trigger_cleanup(q);
-}
-
-static void guc_exec_queue_add_msg(struct xe_exec_queue *q, struct xe_sched_msg *msg,
- u32 opcode)
-{
- xe_pm_runtime_get_noresume(guc_to_xe(exec_queue_to_guc(q)));
-
- INIT_LIST_HEAD(&msg->link);
- msg->opcode = opcode;
- msg->private_data = q;
-
- trace_xe_sched_msg_add(msg);
- xe_sched_add_msg(&q->guc->sched, msg);
-}
-
-#define STATIC_MSG_CLEANUP 0
-#define STATIC_MSG_SUSPEND 1
-#define STATIC_MSG_RESUME 2
-static void guc_exec_queue_fini(struct xe_exec_queue *q)
-{
- struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_CLEANUP;
-
- if (!(q->flags & EXEC_QUEUE_FLAG_PERMANENT) && !exec_queue_wedged(q))
- guc_exec_queue_add_msg(q, msg, CLEANUP);
- else
- __guc_exec_queue_fini(exec_queue_to_guc(q), q);
-}
-
-static int guc_exec_queue_set_priority(struct xe_exec_queue *q,
- enum xe_exec_queue_priority priority)
-{
- struct xe_sched_msg *msg;
-
- if (q->sched_props.priority == priority ||
- exec_queue_killed_or_banned_or_wedged(q))
- return 0;
-
- msg = kmalloc(sizeof(*msg), GFP_KERNEL);
- if (!msg)
- return -ENOMEM;
-
- q->sched_props.priority = priority;
- guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
-
- return 0;
-}
-
-static int guc_exec_queue_set_timeslice(struct xe_exec_queue *q, u32 timeslice_us)
-{
- struct xe_sched_msg *msg;
-
- if (q->sched_props.timeslice_us == timeslice_us ||
- exec_queue_killed_or_banned_or_wedged(q))
- return 0;
-
- msg = kmalloc(sizeof(*msg), GFP_KERNEL);
- if (!msg)
- return -ENOMEM;
-
- q->sched_props.timeslice_us = timeslice_us;
- guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
-
- return 0;
-}
-
-static int guc_exec_queue_set_preempt_timeout(struct xe_exec_queue *q,
- u32 preempt_timeout_us)
-{
- struct xe_sched_msg *msg;
-
- if (q->sched_props.preempt_timeout_us == preempt_timeout_us ||
- exec_queue_killed_or_banned_or_wedged(q))
- return 0;
-
- msg = kmalloc(sizeof(*msg), GFP_KERNEL);
- if (!msg)
- return -ENOMEM;
-
- q->sched_props.preempt_timeout_us = preempt_timeout_us;
- guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
-
- return 0;
-}
-
-static int guc_exec_queue_suspend(struct xe_exec_queue *q)
-{
- struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_SUSPEND;
-
- if (exec_queue_killed_or_banned_or_wedged(q) || q->guc->suspend_pending)
- return -EINVAL;
-
- q->guc->suspend_pending = true;
- guc_exec_queue_add_msg(q, msg, SUSPEND);
-
- return 0;
-}
-
-static void guc_exec_queue_suspend_wait(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- wait_event(q->guc->suspend_wait, !q->guc->suspend_pending ||
- guc_read_stopped(guc));
-}
-
-static void guc_exec_queue_resume(struct xe_exec_queue *q)
-{
- struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_RESUME;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, !q->guc->suspend_pending);
-
- guc_exec_queue_add_msg(q, msg, RESUME);
-}
-
-static bool guc_exec_queue_reset_status(struct xe_exec_queue *q)
-{
- return exec_queue_reset(q) || exec_queue_killed_or_banned_or_wedged(q);
-}
-
-/*
- * All of these functions are an abstraction layer which other parts of XE can
- * use to trap into the GuC backend. All of these functions, aside from init,
- * really shouldn't do much other than trap into the DRM scheduler which
- * synchronizes these operations.
- */
-static const struct xe_exec_queue_ops guc_exec_queue_ops = {
- .init = guc_exec_queue_init,
- .kill = guc_exec_queue_kill,
- .fini = guc_exec_queue_fini,
- .set_priority = guc_exec_queue_set_priority,
- .set_timeslice = guc_exec_queue_set_timeslice,
- .set_preempt_timeout = guc_exec_queue_set_preempt_timeout,
- .suspend = guc_exec_queue_suspend,
- .suspend_wait = guc_exec_queue_suspend_wait,
- .resume = guc_exec_queue_resume,
- .reset_status = guc_exec_queue_reset_status,
-};
-
-static void guc_exec_queue_stop(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched = &q->guc->sched;
-
- /* Stop scheduling + flush any DRM scheduler operations */
- xe_sched_submission_stop(sched);
-
- /* Clean up lost G2H + reset engine state */
- if (exec_queue_registered(q)) {
- if (exec_queue_extra_ref(q) || xe_exec_queue_is_lr(q))
- xe_exec_queue_put(q);
- else if (exec_queue_destroyed(q))
- __guc_exec_queue_fini(guc, q);
- }
- if (q->guc->suspend_pending) {
- set_exec_queue_suspended(q);
- suspend_fence_signal(q);
- }
- atomic_and(EXEC_QUEUE_STATE_WEDGED | EXEC_QUEUE_STATE_BANNED |
- EXEC_QUEUE_STATE_KILLED | EXEC_QUEUE_STATE_DESTROYED |
- EXEC_QUEUE_STATE_SUSPENDED,
- &q->guc->state);
- q->guc->resume_time = 0;
- trace_xe_exec_queue_stop(q);
-
- /*
- * Ban any engine (aside from kernel and engines used for VM ops) with a
- * started but not complete job or if a job has gone through a GT reset
- * more than twice.
- */
- if (!(q->flags & (EXEC_QUEUE_FLAG_KERNEL | EXEC_QUEUE_FLAG_VM))) {
- struct xe_sched_job *job = xe_sched_first_pending_job(sched);
- bool ban = false;
-
- if (job) {
- if ((xe_sched_job_started(job) &&
- !xe_sched_job_completed(job)) ||
- xe_sched_invalidate_job(job, 2)) {
- trace_xe_sched_job_ban(job);
- ban = true;
- }
- } else if (xe_exec_queue_is_lr(q) &&
- (xe_lrc_ring_head(q->lrc[0]) != xe_lrc_ring_tail(q->lrc[0]))) {
- ban = true;
- }
-
- if (ban) {
- set_exec_queue_banned(q);
- xe_guc_exec_queue_trigger_cleanup(q);
- }
- }
-}
-
-int xe_guc_submit_reset_prepare(struct xe_guc *guc)
-{
- int ret;
-
- /*
- * Using an atomic here rather than submission_state.lock as this
- * function can be called while holding the CT lock (engine reset
- * failure). submission_state.lock needs the CT lock to resubmit jobs.
- * Atomic is not ideal, but it works to prevent against concurrent reset
- * and releasing any TDRs waiting on guc->submission_state.stopped.
- */
- ret = atomic_fetch_or(1, &guc->submission_state.stopped);
- smp_wmb();
- wake_up_all(&guc->ct.wq);
-
- return ret;
-}
-
-void xe_guc_submit_reset_wait(struct xe_guc *guc)
-{
- wait_event(guc->ct.wq, xe_device_wedged(guc_to_xe(guc)) ||
- !guc_read_stopped(guc));
-}
-
-void xe_guc_submit_stop(struct xe_guc *guc)
-{
- struct xe_exec_queue *q;
- unsigned long index;
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, guc_read_stopped(guc) == 1);
-
- mutex_lock(&guc->submission_state.lock);
-
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- guc_exec_queue_stop(guc, q);
-
- mutex_unlock(&guc->submission_state.lock);
-
- /*
- * No one can enter the backend at this point, aside from new engine
- * creation which is protected by guc->submission_state.lock.
- */
-
-}
-
-static void guc_exec_queue_start(struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched = &q->guc->sched;
-
- if (!exec_queue_killed_or_banned_or_wedged(q)) {
- int i;
-
- trace_xe_exec_queue_resubmit(q);
- for (i = 0; i < q->width; ++i)
- xe_lrc_set_ring_head(q->lrc[i], q->lrc[i]->ring.tail);
- xe_sched_resubmit_jobs(sched);
- }
-
- xe_sched_submission_start(sched);
-}
-
-int xe_guc_submit_start(struct xe_guc *guc)
-{
- struct xe_exec_queue *q;
- unsigned long index;
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, guc_read_stopped(guc) == 1);
-
- mutex_lock(&guc->submission_state.lock);
- atomic_dec(&guc->submission_state.stopped);
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- guc_exec_queue_start(q);
- mutex_unlock(&guc->submission_state.lock);
-
- wake_up_all(&guc->ct.wq);
-
- return 0;
-}
-
-static struct xe_exec_queue *
-g2h_exec_queue_lookup(struct xe_guc *guc, u32 guc_id)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
-
- if (unlikely(guc_id >= GUC_ID_MAX)) {
- drm_err(&xe->drm, "Invalid guc_id %u", guc_id);
- return NULL;
- }
-
- q = xa_load(&guc->submission_state.exec_queue_lookup, guc_id);
- if (unlikely(!q)) {
- drm_err(&xe->drm, "Not engine present for guc_id %u", guc_id);
- return NULL;
- }
-
- xe_assert(xe, guc_id >= q->guc->id);
- xe_assert(xe, guc_id < (q->guc->id + q->width));
-
- return q;
-}
-
-static void deregister_exec_queue(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- u32 action[] = {
- XE_GUC_ACTION_DEREGISTER_CONTEXT,
- q->guc->id,
- };
-
- xe_gt_assert(guc_to_gt(guc), exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_enable(q));
-
- trace_xe_exec_queue_deregister(q);
-
- xe_guc_ct_send_g2h_handler(&guc->ct, action, ARRAY_SIZE(action));
-}
-
-static void handle_sched_done(struct xe_guc *guc, struct xe_exec_queue *q,
- u32 runnable_state)
-{
- trace_xe_exec_queue_scheduling_done(q);
-
- if (runnable_state == 1) {
- xe_gt_assert(guc_to_gt(guc), exec_queue_pending_enable(q));
-
- q->guc->resume_time = ktime_get();
- clear_exec_queue_pending_enable(q);
- smp_wmb();
- wake_up_all(&guc->ct.wq);
- } else {
- bool check_timeout = exec_queue_check_timeout(q);
-
- xe_gt_assert(guc_to_gt(guc), runnable_state == 0);
- xe_gt_assert(guc_to_gt(guc), exec_queue_pending_disable(q));
-
- clear_exec_queue_pending_disable(q);
- if (q->guc->suspend_pending) {
- suspend_fence_signal(q);
- } else {
- if (exec_queue_banned(q) || check_timeout) {
- smp_wmb();
- wake_up_all(&guc->ct.wq);
- }
- if (!check_timeout)
- deregister_exec_queue(guc, q);
- }
- }
-}
-
-int xe_guc_sched_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
- u32 runnable_state = msg[1];
-
- if (unlikely(len < 2)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- if (unlikely(!exec_queue_pending_enable(q) &&
- !exec_queue_pending_disable(q))) {
- xe_gt_err(guc_to_gt(guc),
- "SCHED_DONE: Unexpected engine state 0x%04x, guc_id=%d, runnable_state=%u",
- atomic_read(&q->guc->state), q->guc->id,
- runnable_state);
- return -EPROTO;
- }
-
- handle_sched_done(guc, q, runnable_state);
-
- return 0;
-}
-
-static void handle_deregister_done(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- trace_xe_exec_queue_deregister_done(q);
-
- clear_exec_queue_registered(q);
-
- if (exec_queue_extra_ref(q) || xe_exec_queue_is_lr(q))
- xe_exec_queue_put(q);
- else
- __guc_exec_queue_fini(guc, q);
-}
-
-int xe_guc_deregister_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
-
- if (unlikely(len < 1)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- if (!exec_queue_destroyed(q) || exec_queue_pending_disable(q) ||
- exec_queue_pending_enable(q) || exec_queue_enabled(q)) {
- xe_gt_err(guc_to_gt(guc),
- "DEREGISTER_DONE: Unexpected engine state 0x%04x, guc_id=%d",
- atomic_read(&q->guc->state), q->guc->id);
- return -EPROTO;
- }
-
- handle_deregister_done(guc, q);
-
- return 0;
-}
-
-int xe_guc_exec_queue_reset_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_gt *gt = guc_to_gt(guc);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
-
- if (unlikely(len < 1)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- xe_gt_info(gt, "Engine reset: engine_class=%s, logical_mask: 0x%x, guc_id=%d",
- xe_hw_engine_class_to_str(q->class), q->logical_mask, guc_id);
-
- /* FIXME: Do error capture, most likely async */
-
- trace_xe_exec_queue_reset(q);
-
- /*
- * A banned engine is a NOP at this point (came from
- * guc_exec_queue_timedout_job). Otherwise, kick drm scheduler to cancel
- * jobs by setting timeout of the job to the minimum value kicking
- * guc_exec_queue_timedout_job.
- */
- set_exec_queue_reset(q);
- if (!exec_queue_banned(q) && !exec_queue_check_timeout(q))
- xe_guc_exec_queue_trigger_cleanup(q);
-
- return 0;
-}
-
-int xe_guc_exec_queue_memory_cat_error_handler(struct xe_guc *guc, u32 *msg,
- u32 len)
-{
- struct xe_gt *gt = guc_to_gt(guc);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
-
- if (unlikely(len < 1)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- xe_gt_dbg(gt, "Engine memory cat error: engine_class=%s, logical_mask: 0x%x, guc_id=%d",
- xe_hw_engine_class_to_str(q->class), q->logical_mask, guc_id);
-
- trace_xe_exec_queue_memory_cat_error(q);
-
- /* Treat the same as engine reset */
- set_exec_queue_reset(q);
- if (!exec_queue_banned(q) && !exec_queue_check_timeout(q))
- xe_guc_exec_queue_trigger_cleanup(q);
-
- return 0;
-}
-
-int xe_guc_exec_queue_reset_failure_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_device *xe = guc_to_xe(guc);
- u8 guc_class, instance;
- u32 reason;
-
- if (unlikely(len != 3)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- guc_class = msg[0];
- instance = msg[1];
- reason = msg[2];
-
- /* Unexpected failure of a hardware feature, log an actual error */
- drm_err(&xe->drm, "GuC engine reset request failed on %d:%d because 0x%08X",
- guc_class, instance, reason);
-
- xe_gt_reset_async(guc_to_gt(guc));
-
- return 0;
-}
-
-static void
-guc_exec_queue_wq_snapshot_capture(struct xe_exec_queue *q,
- struct xe_guc_submit_exec_queue_snapshot *snapshot)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
- int i;
-
- snapshot->guc.wqi_head = q->guc->wqi_head;
- snapshot->guc.wqi_tail = q->guc->wqi_tail;
- snapshot->parallel.wq_desc.head = parallel_read(xe, map, wq_desc.head);
- snapshot->parallel.wq_desc.tail = parallel_read(xe, map, wq_desc.tail);
- snapshot->parallel.wq_desc.status = parallel_read(xe, map,
- wq_desc.wq_status);
-
- if (snapshot->parallel.wq_desc.head !=
- snapshot->parallel.wq_desc.tail) {
- for (i = snapshot->parallel.wq_desc.head;
- i != snapshot->parallel.wq_desc.tail;
- i = (i + sizeof(u32)) % WQ_SIZE)
- snapshot->parallel.wq[i / sizeof(u32)] =
- parallel_read(xe, map, wq[i / sizeof(u32)]);
- }
-}
-
-static void
-guc_exec_queue_wq_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snapshot,
- struct drm_printer *p)
-{
- int i;
-
- drm_printf(p, "\tWQ head: %u (internal), %d (memory)\n",
- snapshot->guc.wqi_head, snapshot->parallel.wq_desc.head);
- drm_printf(p, "\tWQ tail: %u (internal), %d (memory)\n",
- snapshot->guc.wqi_tail, snapshot->parallel.wq_desc.tail);
- drm_printf(p, "\tWQ status: %u\n", snapshot->parallel.wq_desc.status);
-
- if (snapshot->parallel.wq_desc.head !=
- snapshot->parallel.wq_desc.tail) {
- for (i = snapshot->parallel.wq_desc.head;
- i != snapshot->parallel.wq_desc.tail;
- i = (i + sizeof(u32)) % WQ_SIZE)
- drm_printf(p, "\tWQ[%zu]: 0x%08x\n", i / sizeof(u32),
- snapshot->parallel.wq[i / sizeof(u32)]);
- }
-}
-
-/**
- * xe_guc_exec_queue_snapshot_capture - Take a quick snapshot of the GuC Engine.
- * @q: faulty exec queue
- *
- * This can be printed out in a later stage like during dev_coredump
- * analysis.
- *
- * Returns: a GuC Submit Engine snapshot object that must be freed by the
- * caller, using `xe_guc_exec_queue_snapshot_free`.
- */
-struct xe_guc_submit_exec_queue_snapshot *
-xe_guc_exec_queue_snapshot_capture(struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched = &q->guc->sched;
- struct xe_guc_submit_exec_queue_snapshot *snapshot;
- int i;
-
- snapshot = kzalloc(sizeof(*snapshot), GFP_ATOMIC);
-
- if (!snapshot)
- return NULL;
-
- snapshot->guc.id = q->guc->id;
- memcpy(&snapshot->name, &q->name, sizeof(snapshot->name));
- snapshot->class = q->class;
- snapshot->logical_mask = q->logical_mask;
- snapshot->width = q->width;
- snapshot->refcount = kref_read(&q->refcount);
- snapshot->sched_timeout = sched->base.timeout;
- snapshot->sched_props.timeslice_us = q->sched_props.timeslice_us;
- snapshot->sched_props.preempt_timeout_us =
- q->sched_props.preempt_timeout_us;
-
- snapshot->lrc = kmalloc_array(q->width, sizeof(struct xe_lrc_snapshot *),
- GFP_ATOMIC);
-
- if (snapshot->lrc) {
- for (i = 0; i < q->width; ++i) {
- struct xe_lrc *lrc = q->lrc[i];
-
- snapshot->lrc[i] = xe_lrc_snapshot_capture(lrc);
- }
- }
-
- snapshot->schedule_state = atomic_read(&q->guc->state);
- snapshot->exec_queue_flags = q->flags;
-
- snapshot->parallel_execution = xe_exec_queue_is_parallel(q);
- if (snapshot->parallel_execution)
- guc_exec_queue_wq_snapshot_capture(q, snapshot);
-
- spin_lock(&sched->base.job_list_lock);
- snapshot->pending_list_size = list_count_nodes(&sched->base.pending_list);
- snapshot->pending_list = kmalloc_array(snapshot->pending_list_size,
- sizeof(struct pending_list_snapshot),
- GFP_ATOMIC);
-
- if (snapshot->pending_list) {
- struct xe_sched_job *job_iter;
-
- i = 0;
- list_for_each_entry(job_iter, &sched->base.pending_list, drm.list) {
- snapshot->pending_list[i].seqno =
- xe_sched_job_seqno(job_iter);
- snapshot->pending_list[i].fence =
- dma_fence_is_signaled(job_iter->fence) ? 1 : 0;
- snapshot->pending_list[i].finished =
- dma_fence_is_signaled(&job_iter->drm.s_fence->finished)
- ? 1 : 0;
- i++;
- }
- }
-
- spin_unlock(&sched->base.job_list_lock);
-
- return snapshot;
-}
-
-/**
- * xe_guc_exec_queue_snapshot_capture_delayed - Take delayed part of snapshot of the GuC Engine.
- * @snapshot: Previously captured snapshot of job.
- *
- * This captures some data that requires taking some locks, so it cannot be done in signaling path.
- */
-void
-xe_guc_exec_queue_snapshot_capture_delayed(struct xe_guc_submit_exec_queue_snapshot *snapshot)
-{
- int i;
-
- if (!snapshot || !snapshot->lrc)
- return;
-
- for (i = 0; i < snapshot->width; ++i)
- xe_lrc_snapshot_capture_delayed(snapshot->lrc[i]);
-}
-
-/**
- * xe_guc_exec_queue_snapshot_print - Print out a given GuC Engine snapshot.
- * @snapshot: GuC Submit Engine snapshot object.
- * @p: drm_printer where it will be printed out.
- *
- * This function prints out a given GuC Submit Engine snapshot object.
- */
-void
-xe_guc_exec_queue_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snapshot,
- struct drm_printer *p)
-{
- int i;
-
- if (!snapshot)
- return;
-
- drm_printf(p, "\nGuC ID: %d\n", snapshot->guc.id);
- drm_printf(p, "\tName: %s\n", snapshot->name);
- drm_printf(p, "\tClass: %d\n", snapshot->class);
- drm_printf(p, "\tLogical mask: 0x%x\n", snapshot->logical_mask);
- drm_printf(p, "\tWidth: %d\n", snapshot->width);
- drm_printf(p, "\tRef: %d\n", snapshot->refcount);
- drm_printf(p, "\tTimeout: %ld (ms)\n", snapshot->sched_timeout);
- drm_printf(p, "\tTimeslice: %u (us)\n",
- snapshot->sched_props.timeslice_us);
- drm_printf(p, "\tPreempt timeout: %u (us)\n",
- snapshot->sched_props.preempt_timeout_us);
-
- for (i = 0; snapshot->lrc && i < snapshot->width; ++i)
- xe_lrc_snapshot_print(snapshot->lrc[i], p);
-
- drm_printf(p, "\tSchedule State: 0x%x\n", snapshot->schedule_state);
- drm_printf(p, "\tFlags: 0x%lx\n", snapshot->exec_queue_flags);
-
- if (snapshot->parallel_execution)
- guc_exec_queue_wq_snapshot_print(snapshot, p);
-
- for (i = 0; snapshot->pending_list && i < snapshot->pending_list_size;
- i++)
- drm_printf(p, "\tJob: seqno=%d, fence=%d, finished=%d\n",
- snapshot->pending_list[i].seqno,
- snapshot->pending_list[i].fence,
- snapshot->pending_list[i].finished);
-}
-
-/**
- * xe_guc_exec_queue_snapshot_free - Free all allocated objects for a given
- * snapshot.
- * @snapshot: GuC Submit Engine snapshot object.
- *
- * This function free all the memory that needed to be allocated at capture
- * time.
- */
-void xe_guc_exec_queue_snapshot_free(struct xe_guc_submit_exec_queue_snapshot *snapshot)
-{
- int i;
-
- if (!snapshot)
- return;
-
- if (snapshot->lrc) {
- for (i = 0; i < snapshot->width; i++)
- xe_lrc_snapshot_free(snapshot->lrc[i]);
- kfree(snapshot->lrc);
- }
- kfree(snapshot->pending_list);
- kfree(snapshot);
-}
-
-static void guc_exec_queue_print(struct xe_exec_queue *q, struct drm_printer *p)
-{
- struct xe_guc_submit_exec_queue_snapshot *snapshot;
-
- snapshot = xe_guc_exec_queue_snapshot_capture(q);
- xe_guc_exec_queue_snapshot_print(snapshot, p);
- xe_guc_exec_queue_snapshot_free(snapshot);
-}
-
-/**
- * xe_guc_submit_print - GuC Submit Print.
- * @guc: GuC.
- * @p: drm_printer where it will be printed out.
- *
- * This function capture and prints snapshots of **all** GuC Engines.
- */
-void xe_guc_submit_print(struct xe_guc *guc, struct drm_printer *p)
-{
- struct xe_exec_queue *q;
- unsigned long index;
-
- if (!xe_device_uc_enabled(guc_to_xe(guc)))
- return;
-
- mutex_lock(&guc->submission_state.lock);
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- guc_exec_queue_print(q, p);
- mutex_unlock(&guc->submission_state.lock);
-}
diff --git a/rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.9 b/rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.9
deleted file mode 100644
index 110e70f7ee7b..000000000000
--- a/rr-cache/e7c9aafc2297a37f89715cfeed48ccbfb82f76bb/preimage.9
+++ /dev/null
@@ -1,2244 +0,0 @@
-// SPDX-License-Identifier: MIT
-/*
- * Copyright © 2022 Intel Corporation
- */
-
-#include "xe_guc_submit.h"
-
-#include <linux/bitfield.h>
-#include <linux/bitmap.h>
-#include <linux/circ_buf.h>
-#include <linux/delay.h>
-#include <linux/dma-fence-array.h>
-#include <linux/math64.h>
-
-#include <drm/drm_managed.h>
-
-#include "abi/guc_actions_abi.h"
-#include "abi/guc_klvs_abi.h"
-#include "regs/xe_lrc_layout.h"
-#include "xe_assert.h"
-#include "xe_devcoredump.h"
-#include "xe_device.h"
-#include "xe_exec_queue.h"
-#include "xe_force_wake.h"
-#include "xe_gpu_scheduler.h"
-#include "xe_gt.h"
-#include "xe_gt_clock.h"
-#include "xe_gt_printk.h"
-#include "xe_guc.h"
-#include "xe_guc_ct.h"
-#include "xe_guc_exec_queue_types.h"
-#include "xe_guc_id_mgr.h"
-#include "xe_guc_submit_types.h"
-#include "xe_hw_engine.h"
-#include "xe_hw_fence.h"
-#include "xe_lrc.h"
-#include "xe_macros.h"
-#include "xe_map.h"
-#include "xe_mocs.h"
-#include "xe_pm.h"
-#include "xe_ring_ops_types.h"
-#include "xe_sched_job.h"
-#include "xe_trace.h"
-#include "xe_vm.h"
-
-static struct xe_guc *
-exec_queue_to_guc(struct xe_exec_queue *q)
-{
- return &q->gt->uc.guc;
-}
-
-/*
- * Helpers for engine state, using an atomic as some of the bits can transition
- * as the same time (e.g. a suspend can be happning at the same time as schedule
- * engine done being processed).
- */
-#define EXEC_QUEUE_STATE_REGISTERED (1 << 0)
-#define EXEC_QUEUE_STATE_ENABLED (1 << 1)
-#define EXEC_QUEUE_STATE_PENDING_ENABLE (1 << 2)
-#define EXEC_QUEUE_STATE_PENDING_DISABLE (1 << 3)
-#define EXEC_QUEUE_STATE_DESTROYED (1 << 4)
-#define EXEC_QUEUE_STATE_SUSPENDED (1 << 5)
-#define EXEC_QUEUE_STATE_RESET (1 << 6)
-#define EXEC_QUEUE_STATE_KILLED (1 << 7)
-#define EXEC_QUEUE_STATE_WEDGED (1 << 8)
-#define EXEC_QUEUE_STATE_BANNED (1 << 9)
-#define EXEC_QUEUE_STATE_CHECK_TIMEOUT (1 << 10)
-#define EXEC_QUEUE_STATE_EXTRA_REF (1 << 11)
-
-static bool exec_queue_registered(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_REGISTERED;
-}
-
-static void set_exec_queue_registered(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_REGISTERED, &q->guc->state);
-}
-
-static void clear_exec_queue_registered(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_REGISTERED, &q->guc->state);
-}
-
-static bool exec_queue_enabled(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_ENABLED;
-}
-
-static void set_exec_queue_enabled(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_ENABLED, &q->guc->state);
-}
-
-static void clear_exec_queue_enabled(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_ENABLED, &q->guc->state);
-}
-
-static bool exec_queue_pending_enable(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_PENDING_ENABLE;
-}
-
-static void set_exec_queue_pending_enable(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_PENDING_ENABLE, &q->guc->state);
-}
-
-static void clear_exec_queue_pending_enable(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_PENDING_ENABLE, &q->guc->state);
-}
-
-static bool exec_queue_pending_disable(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_PENDING_DISABLE;
-}
-
-static void set_exec_queue_pending_disable(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_PENDING_DISABLE, &q->guc->state);
-}
-
-static void clear_exec_queue_pending_disable(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_PENDING_DISABLE, &q->guc->state);
-}
-
-static bool exec_queue_destroyed(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_DESTROYED;
-}
-
-static void set_exec_queue_destroyed(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_DESTROYED, &q->guc->state);
-}
-
-static bool exec_queue_banned(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_BANNED;
-}
-
-static void set_exec_queue_banned(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_BANNED, &q->guc->state);
-}
-
-static bool exec_queue_suspended(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_SUSPENDED;
-}
-
-static void set_exec_queue_suspended(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_SUSPENDED, &q->guc->state);
-}
-
-static void clear_exec_queue_suspended(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_SUSPENDED, &q->guc->state);
-}
-
-static bool exec_queue_reset(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_RESET;
-}
-
-static void set_exec_queue_reset(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_RESET, &q->guc->state);
-}
-
-static bool exec_queue_killed(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_KILLED;
-}
-
-static void set_exec_queue_killed(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_KILLED, &q->guc->state);
-}
-
-static bool exec_queue_wedged(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_WEDGED;
-}
-
-static void set_exec_queue_wedged(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_WEDGED, &q->guc->state);
-}
-
-static bool exec_queue_check_timeout(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_CHECK_TIMEOUT;
-}
-
-static void set_exec_queue_check_timeout(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_CHECK_TIMEOUT, &q->guc->state);
-}
-
-static void clear_exec_queue_check_timeout(struct xe_exec_queue *q)
-{
- atomic_and(~EXEC_QUEUE_STATE_CHECK_TIMEOUT, &q->guc->state);
-}
-
-static bool exec_queue_extra_ref(struct xe_exec_queue *q)
-{
- return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_EXTRA_REF;
-}
-
-static void set_exec_queue_extra_ref(struct xe_exec_queue *q)
-{
- atomic_or(EXEC_QUEUE_STATE_EXTRA_REF, &q->guc->state);
-}
-
-static bool exec_queue_killed_or_banned_or_wedged(struct xe_exec_queue *q)
-{
- return (atomic_read(&q->guc->state) &
- (EXEC_QUEUE_STATE_WEDGED | EXEC_QUEUE_STATE_KILLED |
- EXEC_QUEUE_STATE_BANNED));
-}
-
-#ifdef CONFIG_PROVE_LOCKING
-static int alloc_submit_wq(struct xe_guc *guc)
-{
- int i;
-
- for (i = 0; i < NUM_SUBMIT_WQ; ++i) {
- guc->submission_state.submit_wq_pool[i] =
- alloc_ordered_workqueue("submit_wq", 0);
- if (!guc->submission_state.submit_wq_pool[i])
- goto err_free;
- }
-
- return 0;
-
-err_free:
- while (i)
- destroy_workqueue(guc->submission_state.submit_wq_pool[--i]);
-
- return -ENOMEM;
-}
-
-static void free_submit_wq(struct xe_guc *guc)
-{
- int i;
-
- for (i = 0; i < NUM_SUBMIT_WQ; ++i)
- destroy_workqueue(guc->submission_state.submit_wq_pool[i]);
-}
-
-static struct workqueue_struct *get_submit_wq(struct xe_guc *guc)
-{
- int idx = guc->submission_state.submit_wq_idx++ % NUM_SUBMIT_WQ;
-
- return guc->submission_state.submit_wq_pool[idx];
-}
-#else
-static int alloc_submit_wq(struct xe_guc *guc)
-{
- return 0;
-}
-
-static void free_submit_wq(struct xe_guc *guc)
-{
-
-}
-
-static struct workqueue_struct *get_submit_wq(struct xe_guc *guc)
-{
- return NULL;
-}
-#endif
-
-static void guc_submit_fini(struct drm_device *drm, void *arg)
-{
- struct xe_guc *guc = arg;
-
- xa_destroy(&guc->submission_state.exec_queue_lookup);
- free_submit_wq(guc);
-}
-
-static void guc_submit_wedged_fini(void *arg)
-{
- struct xe_guc *guc = arg;
- struct xe_exec_queue *q;
- unsigned long index;
-
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- if (exec_queue_wedged(q))
- xe_exec_queue_put(q);
-}
-
-static const struct xe_exec_queue_ops guc_exec_queue_ops;
-
-static void primelockdep(struct xe_guc *guc)
-{
- if (!IS_ENABLED(CONFIG_LOCKDEP))
- return;
-
- fs_reclaim_acquire(GFP_KERNEL);
-
- mutex_lock(&guc->submission_state.lock);
- mutex_unlock(&guc->submission_state.lock);
-
- fs_reclaim_release(GFP_KERNEL);
-}
-
-/**
- * xe_guc_submit_init() - Initialize GuC submission.
- * @guc: the &xe_guc to initialize
- * @num_ids: number of GuC context IDs to use
- *
- * The bare-metal or PF driver can pass ~0 as &num_ids to indicate that all
- * GuC context IDs supported by the GuC firmware should be used for submission.
- *
- * Only VF drivers will have to provide explicit number of GuC context IDs
- * that they can use for submission.
- *
- * Return: 0 on success or a negative error code on failure.
- */
-int xe_guc_submit_init(struct xe_guc *guc, unsigned int num_ids)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_gt *gt = guc_to_gt(guc);
- int err;
-
- err = drmm_mutex_init(&xe->drm, &guc->submission_state.lock);
- if (err)
- return err;
-
- err = xe_guc_id_mgr_init(&guc->submission_state.idm, num_ids);
- if (err)
- return err;
-
- err = alloc_submit_wq(guc);
- if (err)
- return err;
-
- gt->exec_queue_ops = &guc_exec_queue_ops;
-
- xa_init(&guc->submission_state.exec_queue_lookup);
-
- primelockdep(guc);
-
- return drmm_add_action_or_reset(&xe->drm, guc_submit_fini, guc);
-}
-
-static void __release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q, u32 xa_count)
-{
- int i;
-
- lockdep_assert_held(&guc->submission_state.lock);
-
- for (i = 0; i < xa_count; ++i)
- xa_erase(&guc->submission_state.exec_queue_lookup, q->guc->id + i);
-
- xe_guc_id_mgr_release_locked(&guc->submission_state.idm,
- q->guc->id, q->width);
-}
-
-static int alloc_guc_id(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- int ret;
- void *ptr;
- int i;
-
- /*
- * Must use GFP_NOWAIT as this lock is in the dma fence signalling path,
- * worse case user gets -ENOMEM on engine create and has to try again.
- *
- * FIXME: Have caller pre-alloc or post-alloc /w GFP_KERNEL to prevent
- * failure.
- */
- lockdep_assert_held(&guc->submission_state.lock);
-
- ret = xe_guc_id_mgr_reserve_locked(&guc->submission_state.idm,
- q->width);
- if (ret < 0)
- return ret;
-
- q->guc->id = ret;
-
- for (i = 0; i < q->width; ++i) {
- ptr = xa_store(&guc->submission_state.exec_queue_lookup,
- q->guc->id + i, q, GFP_NOWAIT);
- if (IS_ERR(ptr)) {
- ret = PTR_ERR(ptr);
- goto err_release;
- }
- }
-
- return 0;
-
-err_release:
- __release_guc_id(guc, q, i);
-
- return ret;
-}
-
-static void release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- mutex_lock(&guc->submission_state.lock);
- __release_guc_id(guc, q, q->width);
- mutex_unlock(&guc->submission_state.lock);
-}
-
-struct exec_queue_policy {
- u32 count;
- struct guc_update_exec_queue_policy h2g;
-};
-
-static u32 __guc_exec_queue_policy_action_size(struct exec_queue_policy *policy)
-{
- size_t bytes = sizeof(policy->h2g.header) +
- (sizeof(policy->h2g.klv[0]) * policy->count);
-
- return bytes / sizeof(u32);
-}
-
-static void __guc_exec_queue_policy_start_klv(struct exec_queue_policy *policy,
- u16 guc_id)
-{
- policy->h2g.header.action =
- XE_GUC_ACTION_HOST2GUC_UPDATE_CONTEXT_POLICIES;
- policy->h2g.header.guc_id = guc_id;
- policy->count = 0;
-}
-
-#define MAKE_EXEC_QUEUE_POLICY_ADD(func, id) \
-static void __guc_exec_queue_policy_add_##func(struct exec_queue_policy *policy, \
- u32 data) \
-{ \
- XE_WARN_ON(policy->count >= GUC_CONTEXT_POLICIES_KLV_NUM_IDS); \
-\
- policy->h2g.klv[policy->count].kl = \
- FIELD_PREP(GUC_KLV_0_KEY, \
- GUC_CONTEXT_POLICIES_KLV_ID_##id) | \
- FIELD_PREP(GUC_KLV_0_LEN, 1); \
- policy->h2g.klv[policy->count].value = data; \
- policy->count++; \
-}
-
-MAKE_EXEC_QUEUE_POLICY_ADD(execution_quantum, EXECUTION_QUANTUM)
-MAKE_EXEC_QUEUE_POLICY_ADD(preemption_timeout, PREEMPTION_TIMEOUT)
-MAKE_EXEC_QUEUE_POLICY_ADD(priority, SCHEDULING_PRIORITY)
-#undef MAKE_EXEC_QUEUE_POLICY_ADD
-
-static const int xe_exec_queue_prio_to_guc[] = {
- [XE_EXEC_QUEUE_PRIORITY_LOW] = GUC_CLIENT_PRIORITY_NORMAL,
- [XE_EXEC_QUEUE_PRIORITY_NORMAL] = GUC_CLIENT_PRIORITY_KMD_NORMAL,
- [XE_EXEC_QUEUE_PRIORITY_HIGH] = GUC_CLIENT_PRIORITY_HIGH,
- [XE_EXEC_QUEUE_PRIORITY_KERNEL] = GUC_CLIENT_PRIORITY_KMD_HIGH,
-};
-
-static void init_policies(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- struct exec_queue_policy policy;
- struct xe_device *xe = guc_to_xe(guc);
- enum xe_exec_queue_priority prio = q->sched_props.priority;
- u32 timeslice_us = q->sched_props.timeslice_us;
- u32 preempt_timeout_us = q->sched_props.preempt_timeout_us;
-
- xe_assert(xe, exec_queue_registered(q));
-
- __guc_exec_queue_policy_start_klv(&policy, q->guc->id);
- __guc_exec_queue_policy_add_priority(&policy, xe_exec_queue_prio_to_guc[prio]);
- __guc_exec_queue_policy_add_execution_quantum(&policy, timeslice_us);
- __guc_exec_queue_policy_add_preemption_timeout(&policy, preempt_timeout_us);
-
- xe_guc_ct_send(&guc->ct, (u32 *)&policy.h2g,
- __guc_exec_queue_policy_action_size(&policy), 0, 0);
-}
-
-static void set_min_preemption_timeout(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- struct exec_queue_policy policy;
-
- __guc_exec_queue_policy_start_klv(&policy, q->guc->id);
- __guc_exec_queue_policy_add_preemption_timeout(&policy, 1);
-
- xe_guc_ct_send(&guc->ct, (u32 *)&policy.h2g,
- __guc_exec_queue_policy_action_size(&policy), 0, 0);
-}
-
-#define parallel_read(xe_, map_, field_) \
- xe_map_rd_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
- field_)
-#define parallel_write(xe_, map_, field_, val_) \
- xe_map_wr_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
- field_, val_)
-
-static void __register_mlrc_exec_queue(struct xe_guc *guc,
- struct xe_exec_queue *q,
- struct guc_ctxt_registration_info *info)
-{
-#define MAX_MLRC_REG_SIZE (13 + XE_HW_ENGINE_MAX_INSTANCE * 2)
- struct xe_device *xe = guc_to_xe(guc);
- u32 action[MAX_MLRC_REG_SIZE];
- int len = 0;
- int i;
-
- xe_assert(xe, xe_exec_queue_is_parallel(q));
-
- action[len++] = XE_GUC_ACTION_REGISTER_CONTEXT_MULTI_LRC;
- action[len++] = info->flags;
- action[len++] = info->context_idx;
- action[len++] = info->engine_class;
- action[len++] = info->engine_submit_mask;
- action[len++] = info->wq_desc_lo;
- action[len++] = info->wq_desc_hi;
- action[len++] = info->wq_base_lo;
- action[len++] = info->wq_base_hi;
- action[len++] = info->wq_size;
- action[len++] = q->width;
- action[len++] = info->hwlrca_lo;
- action[len++] = info->hwlrca_hi;
-
- for (i = 1; i < q->width; ++i) {
- struct xe_lrc *lrc = q->lrc[i];
-
- action[len++] = lower_32_bits(xe_lrc_descriptor(lrc));
- action[len++] = upper_32_bits(xe_lrc_descriptor(lrc));
- }
-
- xe_assert(xe, len <= MAX_MLRC_REG_SIZE);
-#undef MAX_MLRC_REG_SIZE
-
- xe_guc_ct_send(&guc->ct, action, len, 0, 0);
-}
-
-static void __register_exec_queue(struct xe_guc *guc,
- struct guc_ctxt_registration_info *info)
-{
- u32 action[] = {
- XE_GUC_ACTION_REGISTER_CONTEXT,
- info->flags,
- info->context_idx,
- info->engine_class,
- info->engine_submit_mask,
- info->wq_desc_lo,
- info->wq_desc_hi,
- info->wq_base_lo,
- info->wq_base_hi,
- info->wq_size,
- info->hwlrca_lo,
- info->hwlrca_hi,
- };
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action), 0, 0);
-}
-
-static void register_exec_queue(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_lrc *lrc = q->lrc[0];
- struct guc_ctxt_registration_info info;
-
- xe_assert(xe, !exec_queue_registered(q));
-
- memset(&info, 0, sizeof(info));
- info.context_idx = q->guc->id;
- info.engine_class = xe_engine_class_to_guc_class(q->class);
- info.engine_submit_mask = q->logical_mask;
- info.hwlrca_lo = lower_32_bits(xe_lrc_descriptor(lrc));
- info.hwlrca_hi = upper_32_bits(xe_lrc_descriptor(lrc));
- info.flags = CONTEXT_REGISTRATION_FLAG_KMD;
-
- if (xe_exec_queue_is_parallel(q)) {
- u64 ggtt_addr = xe_lrc_parallel_ggtt_addr(lrc);
- struct iosys_map map = xe_lrc_parallel_map(lrc);
-
- info.wq_desc_lo = lower_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq_desc));
- info.wq_desc_hi = upper_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq_desc));
- info.wq_base_lo = lower_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq[0]));
- info.wq_base_hi = upper_32_bits(ggtt_addr +
- offsetof(struct guc_submit_parallel_scratch, wq[0]));
- info.wq_size = WQ_SIZE;
-
- q->guc->wqi_head = 0;
- q->guc->wqi_tail = 0;
- xe_map_memset(xe, &map, 0, 0, PARALLEL_SCRATCH_SIZE - WQ_SIZE);
- parallel_write(xe, map, wq_desc.wq_status, WQ_STATUS_ACTIVE);
- }
-
- /*
- * We must keep a reference for LR engines if engine is registered with
- * the GuC as jobs signal immediately and can't destroy an engine if the
- * GuC has a reference to it.
- */
- if (xe_exec_queue_is_lr(q))
- xe_exec_queue_get(q);
-
- set_exec_queue_registered(q);
- trace_xe_exec_queue_register(q);
- if (xe_exec_queue_is_parallel(q))
- __register_mlrc_exec_queue(guc, q, &info);
- else
- __register_exec_queue(guc, &info);
- init_policies(guc, q);
-}
-
-static u32 wq_space_until_wrap(struct xe_exec_queue *q)
-{
- return (WQ_SIZE - q->guc->wqi_tail);
-}
-
-static int wq_wait_for_space(struct xe_exec_queue *q, u32 wqi_size)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
- unsigned int sleep_period_ms = 1;
-
-#define AVAILABLE_SPACE \
- CIRC_SPACE(q->guc->wqi_tail, q->guc->wqi_head, WQ_SIZE)
- if (wqi_size > AVAILABLE_SPACE) {
-try_again:
- q->guc->wqi_head = parallel_read(xe, map, wq_desc.head);
- if (wqi_size > AVAILABLE_SPACE) {
- if (sleep_period_ms == 1024) {
- xe_gt_reset_async(q->gt);
- return -ENODEV;
- }
-
- msleep(sleep_period_ms);
- sleep_period_ms <<= 1;
- goto try_again;
- }
- }
-#undef AVAILABLE_SPACE
-
- return 0;
-}
-
-static int wq_noop_append(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
- u32 len_dw = wq_space_until_wrap(q) / sizeof(u32) - 1;
-
- if (wq_wait_for_space(q, wq_space_until_wrap(q)))
- return -ENODEV;
-
- xe_assert(xe, FIELD_FIT(WQ_LEN_MASK, len_dw));
-
- parallel_write(xe, map, wq[q->guc->wqi_tail / sizeof(u32)],
- FIELD_PREP(WQ_TYPE_MASK, WQ_TYPE_NOOP) |
- FIELD_PREP(WQ_LEN_MASK, len_dw));
- q->guc->wqi_tail = 0;
-
- return 0;
-}
-
-static void wq_item_append(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
-#define WQ_HEADER_SIZE 4 /* Includes 1 LRC address too */
- u32 wqi[XE_HW_ENGINE_MAX_INSTANCE + (WQ_HEADER_SIZE - 1)];
- u32 wqi_size = (q->width + (WQ_HEADER_SIZE - 1)) * sizeof(u32);
- u32 len_dw = (wqi_size / sizeof(u32)) - 1;
- int i = 0, j;
-
- if (wqi_size > wq_space_until_wrap(q)) {
- if (wq_noop_append(q))
- return;
- }
- if (wq_wait_for_space(q, wqi_size))
- return;
-
- wqi[i++] = FIELD_PREP(WQ_TYPE_MASK, WQ_TYPE_MULTI_LRC) |
- FIELD_PREP(WQ_LEN_MASK, len_dw);
- wqi[i++] = xe_lrc_descriptor(q->lrc[0]);
- wqi[i++] = FIELD_PREP(WQ_GUC_ID_MASK, q->guc->id) |
- FIELD_PREP(WQ_RING_TAIL_MASK, q->lrc[0]->ring.tail / sizeof(u64));
- wqi[i++] = 0;
- for (j = 1; j < q->width; ++j) {
- struct xe_lrc *lrc = q->lrc[j];
-
- wqi[i++] = lrc->ring.tail / sizeof(u64);
- }
-
- xe_assert(xe, i == wqi_size / sizeof(u32));
-
- iosys_map_incr(&map, offsetof(struct guc_submit_parallel_scratch,
- wq[q->guc->wqi_tail / sizeof(u32)]));
- xe_map_memcpy_to(xe, &map, 0, wqi, wqi_size);
- q->guc->wqi_tail += wqi_size;
- xe_assert(xe, q->guc->wqi_tail <= WQ_SIZE);
-
- xe_device_wmb(xe);
-
- map = xe_lrc_parallel_map(q->lrc[0]);
- parallel_write(xe, map, wq_desc.tail, q->guc->wqi_tail);
-}
-
-#define RESUME_PENDING ~0x0ull
-static void submit_exec_queue(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_lrc *lrc = q->lrc[0];
- u32 action[3];
- u32 g2h_len = 0;
- u32 num_g2h = 0;
- int len = 0;
- bool extra_submit = false;
-
- xe_assert(xe, exec_queue_registered(q));
-
- if (xe_exec_queue_is_parallel(q))
- wq_item_append(q);
- else
- xe_lrc_set_ring_tail(lrc, lrc->ring.tail);
-
- if (exec_queue_suspended(q) && !xe_exec_queue_is_parallel(q))
- return;
-
- if (!exec_queue_enabled(q) && !exec_queue_suspended(q)) {
- action[len++] = XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET;
- action[len++] = q->guc->id;
- action[len++] = GUC_CONTEXT_ENABLE;
- g2h_len = G2H_LEN_DW_SCHED_CONTEXT_MODE_SET;
- num_g2h = 1;
- if (xe_exec_queue_is_parallel(q))
- extra_submit = true;
-
- q->guc->resume_time = RESUME_PENDING;
- set_exec_queue_pending_enable(q);
- set_exec_queue_enabled(q);
- trace_xe_exec_queue_scheduling_enable(q);
- } else {
- action[len++] = XE_GUC_ACTION_SCHED_CONTEXT;
- action[len++] = q->guc->id;
- trace_xe_exec_queue_submit(q);
- }
-
- xe_guc_ct_send(&guc->ct, action, len, g2h_len, num_g2h);
-
- if (extra_submit) {
- len = 0;
- action[len++] = XE_GUC_ACTION_SCHED_CONTEXT;
- action[len++] = q->guc->id;
- trace_xe_exec_queue_submit(q);
-
- xe_guc_ct_send(&guc->ct, action, len, 0, 0);
- }
-}
-
-static struct dma_fence *
-guc_exec_queue_run_job(struct drm_sched_job *drm_job)
-{
- struct xe_sched_job *job = to_xe_sched_job(drm_job);
- struct xe_exec_queue *q = job->q;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- bool lr = xe_exec_queue_is_lr(q);
-
- xe_assert(xe, !(exec_queue_destroyed(q) || exec_queue_pending_disable(q)) ||
- exec_queue_banned(q) || exec_queue_suspended(q));
-
- trace_xe_sched_job_run(job);
-
- if (!exec_queue_killed_or_banned_or_wedged(q) && !xe_sched_job_is_error(job)) {
- if (!exec_queue_registered(q))
- register_exec_queue(q);
- if (!lr) /* LR jobs are emitted in the exec IOCTL */
- q->ring_ops->emit_job(job);
- submit_exec_queue(q);
- }
-
- if (lr) {
- xe_sched_job_set_error(job, -EOPNOTSUPP);
- return NULL;
- } else if (test_and_set_bit(JOB_FLAG_SUBMIT, &job->fence->flags)) {
- return job->fence;
- } else {
- return dma_fence_get(job->fence);
- }
-}
-
-static void guc_exec_queue_free_job(struct drm_sched_job *drm_job)
-{
- struct xe_sched_job *job = to_xe_sched_job(drm_job);
-
- xe_exec_queue_update_run_ticks(job->q);
-
- trace_xe_sched_job_free(job);
- xe_sched_job_put(job);
-}
-
-static int guc_read_stopped(struct xe_guc *guc)
-{
- return atomic_read(&guc->submission_state.stopped);
-}
-
-#define MAKE_SCHED_CONTEXT_ACTION(q, enable_disable) \
- u32 action[] = { \
- XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET, \
- q->guc->id, \
- GUC_CONTEXT_##enable_disable, \
- }
-
-static void disable_scheduling_deregister(struct xe_guc *guc,
- struct xe_exec_queue *q)
-{
- MAKE_SCHED_CONTEXT_ACTION(q, DISABLE);
- struct xe_device *xe = guc_to_xe(guc);
- int ret;
-
- set_min_preemption_timeout(guc, q);
- smp_rmb();
- ret = wait_event_timeout(guc->ct.wq, !exec_queue_pending_enable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret) {
- struct xe_gpu_scheduler *sched = &q->guc->sched;
-
- drm_warn(&xe->drm, "Pending enable failed to respond");
- xe_sched_submission_start(sched);
- xe_gt_reset_async(q->gt);
- xe_sched_tdr_queue_imm(sched);
- return;
- }
-
- clear_exec_queue_enabled(q);
- set_exec_queue_pending_disable(q);
- set_exec_queue_destroyed(q);
- trace_xe_exec_queue_scheduling_disable(q);
-
- /*
- * Reserve space for both G2H here as the 2nd G2H is sent from a G2H
- * handler and we are not allowed to reserved G2H space in handlers.
- */
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_SCHED_CONTEXT_MODE_SET +
- G2H_LEN_DW_DEREGISTER_CONTEXT, 2);
-}
-
-static void xe_guc_exec_queue_trigger_cleanup(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- /** to wakeup xe_wait_user_fence ioctl if exec queue is reset */
- wake_up_all(&xe->ufence_wq);
-
- if (xe_exec_queue_is_lr(q))
- queue_work(guc_to_gt(guc)->ordered_wq, &q->guc->lr_tdr);
- else
- xe_sched_tdr_queue_imm(&q->guc->sched);
-}
-
-/**
- * xe_guc_submit_wedge() - Wedge GuC submission
- * @guc: the GuC object
- *
- * Save exec queue's registered with GuC state by taking a ref to each queue.
- * Register a DRMM handler to drop refs upon driver unload.
- */
-void xe_guc_submit_wedge(struct xe_guc *guc)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- unsigned long index;
- int err;
-
- xe_gt_assert(guc_to_gt(guc), guc_to_xe(guc)->wedged.mode);
-
- err = devm_add_action_or_reset(guc_to_xe(guc)->drm.dev,
- guc_submit_wedged_fini, guc);
- if (err) {
- drm_err(&xe->drm, "Failed to register xe_guc_submit clean-up on wedged.mode=2. Although device is wedged.\n");
- return;
- }
-
- mutex_lock(&guc->submission_state.lock);
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- if (xe_exec_queue_get_unless_zero(q))
- set_exec_queue_wedged(q);
- mutex_unlock(&guc->submission_state.lock);
-}
-
-static bool guc_submit_hint_wedged(struct xe_guc *guc)
-{
- struct xe_device *xe = guc_to_xe(guc);
-
- if (xe->wedged.mode != 2)
- return false;
-
- if (xe_device_wedged(xe))
- return true;
-
- xe_device_declare_wedged(xe);
-
- return true;
-}
-
-static void xe_guc_exec_queue_lr_cleanup(struct work_struct *w)
-{
- struct xe_guc_exec_queue *ge =
- container_of(w, struct xe_guc_exec_queue, lr_tdr);
- struct xe_exec_queue *q = ge->q;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_gpu_scheduler *sched = &ge->sched;
- bool wedged;
-
- xe_assert(xe, xe_exec_queue_is_lr(q));
- trace_xe_exec_queue_lr_cleanup(q);
-
- wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
-
- /* Kill the run_job / process_msg entry points */
- xe_sched_submission_stop(sched);
-
- /*
- * Engine state now mostly stable, disable scheduling / deregister if
- * needed. This cleanup routine might be called multiple times, where
- * the actual async engine deregister drops the final engine ref.
- * Calling disable_scheduling_deregister will mark the engine as
- * destroyed and fire off the CT requests to disable scheduling /
- * deregister, which we only want to do once. We also don't want to mark
- * the engine as pending_disable again as this may race with the
- * xe_guc_deregister_done_handler() which treats it as an unexpected
- * state.
- */
- if (!wedged && exec_queue_registered(q) && !exec_queue_destroyed(q)) {
- struct xe_guc *guc = exec_queue_to_guc(q);
- int ret;
-
- set_exec_queue_banned(q);
- disable_scheduling_deregister(guc, q);
-
- /*
- * Must wait for scheduling to be disabled before signalling
- * any fences, if GT broken the GT reset code should signal us.
- */
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_disable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret) {
- drm_warn(&xe->drm, "Schedule disable failed to respond");
- xe_sched_submission_start(sched);
- xe_gt_reset_async(q->gt);
- return;
- }
- }
-
- xe_sched_submission_start(sched);
-}
-
-#define ADJUST_FIVE_PERCENT(__t) mul_u64_u32_div(__t, 105, 100)
-
-static bool check_timeout(struct xe_exec_queue *q, struct xe_sched_job *job)
-{
- struct xe_gt *gt = guc_to_gt(exec_queue_to_guc(q));
- u32 ctx_timestamp = xe_lrc_ctx_timestamp(q->lrc[0]);
- u32 ctx_job_timestamp = xe_lrc_ctx_job_timestamp(q->lrc[0]);
- u32 timeout_ms = q->sched_props.job_timeout_ms;
- u32 diff;
- u64 running_time_ms;
-
- /*
- * Counter wraps at ~223s at the usual 19.2MHz, be paranoid catch
- * possible overflows with a high timeout.
- */
- xe_gt_assert(gt, timeout_ms < 100 * MSEC_PER_SEC);
-
- if (ctx_timestamp < ctx_job_timestamp)
- diff = ctx_timestamp + U32_MAX - ctx_job_timestamp;
- else
- diff = ctx_timestamp - ctx_job_timestamp;
-
- /*
- * Ensure timeout is within 5% to account for an GuC scheduling latency
- */
- running_time_ms =
- ADJUST_FIVE_PERCENT(xe_gt_clock_interval_to_ms(gt, diff));
-
- xe_gt_dbg(gt,
- "Check job timeout: seqno=%u, lrc_seqno=%u, guc_id=%d, running_time_ms=%llu, timeout_ms=%u, diff=0x%08x",
- xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
- q->guc->id, running_time_ms, timeout_ms, diff);
-
- return running_time_ms >= timeout_ms;
-}
-
-static void enable_scheduling(struct xe_exec_queue *q)
-{
- MAKE_SCHED_CONTEXT_ACTION(q, ENABLE);
- struct xe_guc *guc = exec_queue_to_guc(q);
- int ret;
-
- xe_gt_assert(guc_to_gt(guc), !exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_enable(q));
-
- set_exec_queue_pending_enable(q);
- set_exec_queue_enabled(q);
- trace_xe_exec_queue_scheduling_enable(q);
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_SCHED_CONTEXT_MODE_SET, 1);
-
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_enable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret || guc_read_stopped(guc)) {
- xe_gt_warn(guc_to_gt(guc), "Schedule enable failed to respond");
- set_exec_queue_banned(q);
- xe_gt_reset_async(q->gt);
- xe_sched_tdr_queue_imm(&q->guc->sched);
- }
-}
-
-static void disable_scheduling(struct xe_exec_queue *q, bool immediate)
-{
- MAKE_SCHED_CONTEXT_ACTION(q, DISABLE);
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- xe_gt_assert(guc_to_gt(guc), !exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
-
- if (immediate)
- set_min_preemption_timeout(guc, q);
- clear_exec_queue_enabled(q);
- set_exec_queue_pending_disable(q);
- trace_xe_exec_queue_scheduling_disable(q);
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_SCHED_CONTEXT_MODE_SET, 1);
-}
-
-static void __deregister_exec_queue(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- u32 action[] = {
- XE_GUC_ACTION_DEREGISTER_CONTEXT,
- q->guc->id,
- };
-
- xe_gt_assert(guc_to_gt(guc), !exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_enable(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
-
- set_exec_queue_destroyed(q);
- trace_xe_exec_queue_deregister(q);
-
- xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
- G2H_LEN_DW_DEREGISTER_CONTEXT, 1);
-}
-
-static enum drm_gpu_sched_stat
-guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
-{
- struct xe_sched_job *job = to_xe_sched_job(drm_job);
- struct xe_sched_job *tmp_job;
- struct xe_exec_queue *q = job->q;
- struct xe_gpu_scheduler *sched = &q->guc->sched;
- struct xe_guc *guc = exec_queue_to_guc(q);
-<<<<<<<
-=======
- const char *process_name = "no process";
->>>>>>>
- int err = -ETIME;
- int i = 0;
- bool wedged, skip_timeout_check;
-
- /*
- * TDR has fired before free job worker. Common if exec queue
- * immediately closed after last fence signaled.
- */
- if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &job->fence->flags)) {
- guc_exec_queue_free_job(drm_job);
-
- return DRM_GPU_SCHED_STAT_NOMINAL;
- }
-
- /* Kill the run_job entry point */
- xe_sched_submission_stop(sched);
-
- /* Must check all state after stopping scheduler */
- skip_timeout_check = exec_queue_reset(q) ||
- exec_queue_killed_or_banned_or_wedged(q) ||
- exec_queue_destroyed(q);
-
- /* Job hasn't started, can't be timed out */
- if (!skip_timeout_check && !xe_sched_job_started(job))
- goto rearm;
-
- /*
- * XXX: Sampling timeout doesn't work in wedged mode as we have to
- * modify scheduling state to read timestamp. We could read the
- * timestamp from a register to accumulate current running time but this
- * doesn't work for SRIOV. For now assuming timeouts in wedged mode are
- * genuine timeouts.
- */
- wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
-
- /* Engine state now stable, disable scheduling to check timestamp */
- if (!wedged && exec_queue_registered(q)) {
- int ret;
-
- if (exec_queue_reset(q))
- err = -EIO;
-
- if (!exec_queue_destroyed(q)) {
- /*
- * Wait for any pending G2H to flush out before
- * modifying state
- */
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_enable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret || guc_read_stopped(guc))
- goto trigger_reset;
-
- /*
- * Flag communicates to G2H handler that schedule
- * disable originated from a timeout check. The G2H then
- * avoid triggering cleanup or deregistering the exec
- * queue.
- */
- set_exec_queue_check_timeout(q);
- disable_scheduling(q, skip_timeout_check);
- }
-
- /*
- * Must wait for scheduling to be disabled before signalling
- * any fences, if GT broken the GT reset code should signal us.
- *
- * FIXME: Tests can generate a ton of 0x6000 (IOMMU CAT fault
- * error) messages which can cause the schedule disable to get
- * lost. If this occurs, trigger a GT reset to recover.
- */
- smp_rmb();
- ret = wait_event_timeout(guc->ct.wq,
- !exec_queue_pending_disable(q) ||
- guc_read_stopped(guc), HZ * 5);
- if (!ret || guc_read_stopped(guc)) {
-trigger_reset:
- if (!ret)
- xe_gt_warn(guc_to_gt(guc), "Schedule disable failed to respond");
- set_exec_queue_extra_ref(q);
- xe_exec_queue_get(q); /* GT reset owns this */
- set_exec_queue_banned(q);
- xe_gt_reset_async(q->gt);
- xe_sched_tdr_queue_imm(sched);
- goto rearm;
- }
- }
-
- /*
- * Check if job is actually timed out, if so restart job execution and TDR
- */
- if (!wedged && !skip_timeout_check && !check_timeout(q, job) &&
- !exec_queue_reset(q) && exec_queue_registered(q)) {
- clear_exec_queue_check_timeout(q);
- goto sched_enable;
- }
-
-<<<<<<<
- if (q->vm && q->vm->xef) {
- process_name = q->vm->xef->process_name;
- pid = q->vm->xef->pid;
- }
- xe_gt_notice(guc_to_gt(guc), "Timedout job: seqno=%u, lrc_seqno=%u, guc_id=%d, flags=0x%lx in %s [%d]",
- xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
- q->guc->id, q->flags, process_name, pid);
-
-=======
- xe_gt_notice(guc_to_gt(guc), "Timedout job: seqno=%u, lrc_seqno=%u, guc_id=%d, flags=0x%lx",
- xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
- q->guc->id, q->flags);
->>>>>>>
- trace_xe_sched_job_timedout(job);
-
- if (!exec_queue_killed(q))
- xe_devcoredump(job);
-
- /*
- * Kernel jobs should never fail, nor should VM jobs if they do
- * somethings has gone wrong and the GT needs a reset
- */
- xe_gt_WARN(q->gt, q->flags & EXEC_QUEUE_FLAG_KERNEL,
- "Kernel-submitted job timed out\n");
- xe_gt_WARN(q->gt, q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q),
- "VM job timed out on non-killed execqueue\n");
- if (!wedged && (q->flags & EXEC_QUEUE_FLAG_KERNEL ||
- (q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q)))) {
- if (!xe_sched_invalidate_job(job, 2)) {
- clear_exec_queue_check_timeout(q);
- xe_gt_reset_async(q->gt);
- goto rearm;
- }
- }
-
- /* Finish cleaning up exec queue via deregister */
- set_exec_queue_banned(q);
- if (!wedged && exec_queue_registered(q) && !exec_queue_destroyed(q)) {
- set_exec_queue_extra_ref(q);
- xe_exec_queue_get(q);
- __deregister_exec_queue(guc, q);
- }
-
- /* Stop fence signaling */
- xe_hw_fence_irq_stop(q->fence_irq);
-
- /*
- * Fence state now stable, stop / start scheduler which cleans up any
- * fences that are complete
- */
- xe_sched_add_pending_job(sched, job);
- xe_sched_submission_start(sched);
-
- xe_guc_exec_queue_trigger_cleanup(q);
-
- /* Mark all outstanding jobs as bad, thus completing them */
- spin_lock(&sched->base.job_list_lock);
- list_for_each_entry(tmp_job, &sched->base.pending_list, drm.list)
- xe_sched_job_set_error(tmp_job, !i++ ? err : -ECANCELED);
- spin_unlock(&sched->base.job_list_lock);
-
- /* Start fence signaling */
- xe_hw_fence_irq_start(q->fence_irq);
-
- return DRM_GPU_SCHED_STAT_NOMINAL;
-
-sched_enable:
- enable_scheduling(q);
-rearm:
- /*
- * XXX: Ideally want to adjust timeout based on current exection time
- * but there is not currently an easy way to do in DRM scheduler. With
- * some thought, do this in a follow up.
- */
- xe_sched_add_pending_job(sched, job);
- xe_sched_submission_start(sched);
-
- return DRM_GPU_SCHED_STAT_NOMINAL;
-}
-
-static void __guc_exec_queue_fini_async(struct work_struct *w)
-{
- struct xe_guc_exec_queue *ge =
- container_of(w, struct xe_guc_exec_queue, fini_async);
- struct xe_exec_queue *q = ge->q;
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- xe_pm_runtime_get(guc_to_xe(guc));
- trace_xe_exec_queue_destroy(q);
-
- if (xe_exec_queue_is_lr(q))
- cancel_work_sync(&ge->lr_tdr);
- release_guc_id(guc, q);
- xe_sched_entity_fini(&ge->entity);
- xe_sched_fini(&ge->sched);
-
- kfree(ge);
- xe_exec_queue_fini(q);
- xe_pm_runtime_put(guc_to_xe(guc));
-}
-
-static void guc_exec_queue_fini_async(struct xe_exec_queue *q)
-{
- INIT_WORK(&q->guc->fini_async, __guc_exec_queue_fini_async);
-
- /* We must block on kernel engines so slabs are empty on driver unload */
- if (q->flags & EXEC_QUEUE_FLAG_PERMANENT || exec_queue_wedged(q))
- __guc_exec_queue_fini_async(&q->guc->fini_async);
- else
- queue_work(system_wq, &q->guc->fini_async);
-}
-
-static void __guc_exec_queue_fini(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- /*
- * Might be done from within the GPU scheduler, need to do async as we
- * fini the scheduler when the engine is fini'd, the scheduler can't
- * complete fini within itself (circular dependency). Async resolves
- * this we and don't really care when everything is fini'd, just that it
- * is.
- */
- guc_exec_queue_fini_async(q);
-}
-
-static void __guc_exec_queue_process_msg_cleanup(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, !(q->flags & EXEC_QUEUE_FLAG_PERMANENT));
- trace_xe_exec_queue_cleanup_entity(q);
-
- if (exec_queue_registered(q))
- disable_scheduling_deregister(guc, q);
- else
- __guc_exec_queue_fini(guc, q);
-}
-
-static bool guc_exec_queue_allowed_to_change_state(struct xe_exec_queue *q)
-{
- return !exec_queue_killed_or_banned_or_wedged(q) && exec_queue_registered(q);
-}
-
-static void __guc_exec_queue_process_msg_set_sched_props(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- if (guc_exec_queue_allowed_to_change_state(q))
- init_policies(guc, q);
- kfree(msg);
-}
-
-static void suspend_fence_signal(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, exec_queue_suspended(q) || exec_queue_killed(q) ||
- guc_read_stopped(guc));
- xe_assert(xe, q->guc->suspend_pending);
-
- q->guc->suspend_pending = false;
- smp_wmb();
- wake_up(&q->guc->suspend_wait);
-}
-
-static void __guc_exec_queue_process_msg_suspend(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- if (guc_exec_queue_allowed_to_change_state(q) && !exec_queue_suspended(q) &&
- exec_queue_enabled(q)) {
- wait_event(guc->ct.wq, q->guc->resume_time != RESUME_PENDING ||
- guc_read_stopped(guc));
-
- if (!guc_read_stopped(guc)) {
- s64 since_resume_ms =
- ktime_ms_delta(ktime_get(),
- q->guc->resume_time);
- s64 wait_ms = q->vm->preempt.min_run_period_ms -
- since_resume_ms;
-
- if (wait_ms > 0 && q->guc->resume_time)
- msleep(wait_ms);
-
- set_exec_queue_suspended(q);
- disable_scheduling(q, false);
- }
- } else if (q->guc->suspend_pending) {
- set_exec_queue_suspended(q);
- suspend_fence_signal(q);
- }
-}
-
-static void __guc_exec_queue_process_msg_resume(struct xe_sched_msg *msg)
-{
- struct xe_exec_queue *q = msg->private_data;
-
- if (guc_exec_queue_allowed_to_change_state(q)) {
- q->guc->resume_time = RESUME_PENDING;
- clear_exec_queue_suspended(q);
- enable_scheduling(q);
- } else {
- clear_exec_queue_suspended(q);
- }
-}
-
-#define CLEANUP 1 /* Non-zero values to catch uninitialized msg */
-#define SET_SCHED_PROPS 2
-#define SUSPEND 3
-#define RESUME 4
-
-static void guc_exec_queue_process_msg(struct xe_sched_msg *msg)
-{
- trace_xe_sched_msg_recv(msg);
-
- switch (msg->opcode) {
- case CLEANUP:
- __guc_exec_queue_process_msg_cleanup(msg);
- break;
- case SET_SCHED_PROPS:
- __guc_exec_queue_process_msg_set_sched_props(msg);
- break;
- case SUSPEND:
- __guc_exec_queue_process_msg_suspend(msg);
- break;
- case RESUME:
- __guc_exec_queue_process_msg_resume(msg);
- break;
- default:
- XE_WARN_ON("Unknown message type");
- }
-
- xe_pm_runtime_put(guc_to_xe(exec_queue_to_guc(msg->private_data)));
-}
-
-static const struct drm_sched_backend_ops drm_sched_ops = {
- .run_job = guc_exec_queue_run_job,
- .free_job = guc_exec_queue_free_job,
- .timedout_job = guc_exec_queue_timedout_job,
-};
-
-static const struct xe_sched_backend_ops xe_sched_ops = {
- .process_msg = guc_exec_queue_process_msg,
-};
-
-static int guc_exec_queue_init(struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_guc_exec_queue *ge;
- long timeout;
- int err;
-
- xe_assert(xe, xe_device_uc_enabled(guc_to_xe(guc)));
-
- ge = kzalloc(sizeof(*ge), GFP_KERNEL);
- if (!ge)
- return -ENOMEM;
-
- q->guc = ge;
- ge->q = q;
- init_waitqueue_head(&ge->suspend_wait);
-
- timeout = (q->vm && xe_vm_in_lr_mode(q->vm)) ? MAX_SCHEDULE_TIMEOUT :
- msecs_to_jiffies(q->sched_props.job_timeout_ms);
- err = xe_sched_init(&ge->sched, &drm_sched_ops, &xe_sched_ops,
- get_submit_wq(guc),
- q->lrc[0]->ring.size / MAX_JOB_SIZE_BYTES, 64,
- timeout, guc_to_gt(guc)->ordered_wq, NULL,
- q->name, gt_to_xe(q->gt)->drm.dev);
- if (err)
- goto err_free;
-
- sched = &ge->sched;
- err = xe_sched_entity_init(&ge->entity, sched);
- if (err)
- goto err_sched;
-
- if (xe_exec_queue_is_lr(q))
- INIT_WORK(&q->guc->lr_tdr, xe_guc_exec_queue_lr_cleanup);
-
- mutex_lock(&guc->submission_state.lock);
-
- err = alloc_guc_id(guc, q);
- if (err)
- goto err_entity;
-
- q->entity = &ge->entity;
-
- if (guc_read_stopped(guc))
- xe_sched_stop(sched);
-
- mutex_unlock(&guc->submission_state.lock);
-
- xe_exec_queue_assign_name(q, q->guc->id);
-
- trace_xe_exec_queue_create(q);
-
- return 0;
-
-err_entity:
- mutex_unlock(&guc->submission_state.lock);
- xe_sched_entity_fini(&ge->entity);
-err_sched:
- xe_sched_fini(&ge->sched);
-err_free:
- kfree(ge);
-
- return err;
-}
-
-static void guc_exec_queue_kill(struct xe_exec_queue *q)
-{
- trace_xe_exec_queue_kill(q);
- set_exec_queue_killed(q);
- xe_guc_exec_queue_trigger_cleanup(q);
-}
-
-static void guc_exec_queue_add_msg(struct xe_exec_queue *q, struct xe_sched_msg *msg,
- u32 opcode)
-{
- xe_pm_runtime_get_noresume(guc_to_xe(exec_queue_to_guc(q)));
-
- INIT_LIST_HEAD(&msg->link);
- msg->opcode = opcode;
- msg->private_data = q;
-
- trace_xe_sched_msg_add(msg);
- xe_sched_add_msg(&q->guc->sched, msg);
-}
-
-#define STATIC_MSG_CLEANUP 0
-#define STATIC_MSG_SUSPEND 1
-#define STATIC_MSG_RESUME 2
-static void guc_exec_queue_fini(struct xe_exec_queue *q)
-{
- struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_CLEANUP;
-
- if (!(q->flags & EXEC_QUEUE_FLAG_PERMANENT) && !exec_queue_wedged(q))
- guc_exec_queue_add_msg(q, msg, CLEANUP);
- else
- __guc_exec_queue_fini(exec_queue_to_guc(q), q);
-}
-
-static int guc_exec_queue_set_priority(struct xe_exec_queue *q,
- enum xe_exec_queue_priority priority)
-{
- struct xe_sched_msg *msg;
-
- if (q->sched_props.priority == priority ||
- exec_queue_killed_or_banned_or_wedged(q))
- return 0;
-
- msg = kmalloc(sizeof(*msg), GFP_KERNEL);
- if (!msg)
- return -ENOMEM;
-
- q->sched_props.priority = priority;
- guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
-
- return 0;
-}
-
-static int guc_exec_queue_set_timeslice(struct xe_exec_queue *q, u32 timeslice_us)
-{
- struct xe_sched_msg *msg;
-
- if (q->sched_props.timeslice_us == timeslice_us ||
- exec_queue_killed_or_banned_or_wedged(q))
- return 0;
-
- msg = kmalloc(sizeof(*msg), GFP_KERNEL);
- if (!msg)
- return -ENOMEM;
-
- q->sched_props.timeslice_us = timeslice_us;
- guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
-
- return 0;
-}
-
-static int guc_exec_queue_set_preempt_timeout(struct xe_exec_queue *q,
- u32 preempt_timeout_us)
-{
- struct xe_sched_msg *msg;
-
- if (q->sched_props.preempt_timeout_us == preempt_timeout_us ||
- exec_queue_killed_or_banned_or_wedged(q))
- return 0;
-
- msg = kmalloc(sizeof(*msg), GFP_KERNEL);
- if (!msg)
- return -ENOMEM;
-
- q->sched_props.preempt_timeout_us = preempt_timeout_us;
- guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
-
- return 0;
-}
-
-static int guc_exec_queue_suspend(struct xe_exec_queue *q)
-{
- struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_SUSPEND;
-
- if (exec_queue_killed_or_banned_or_wedged(q) || q->guc->suspend_pending)
- return -EINVAL;
-
- q->guc->suspend_pending = true;
- guc_exec_queue_add_msg(q, msg, SUSPEND);
-
- return 0;
-}
-
-static void guc_exec_queue_suspend_wait(struct xe_exec_queue *q)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
-
- wait_event(q->guc->suspend_wait, !q->guc->suspend_pending ||
- guc_read_stopped(guc));
-}
-
-static void guc_exec_queue_resume(struct xe_exec_queue *q)
-{
- struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_RESUME;
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, !q->guc->suspend_pending);
-
- guc_exec_queue_add_msg(q, msg, RESUME);
-}
-
-static bool guc_exec_queue_reset_status(struct xe_exec_queue *q)
-{
- return exec_queue_reset(q) || exec_queue_killed_or_banned_or_wedged(q);
-}
-
-/*
- * All of these functions are an abstraction layer which other parts of XE can
- * use to trap into the GuC backend. All of these functions, aside from init,
- * really shouldn't do much other than trap into the DRM scheduler which
- * synchronizes these operations.
- */
-static const struct xe_exec_queue_ops guc_exec_queue_ops = {
- .init = guc_exec_queue_init,
- .kill = guc_exec_queue_kill,
- .fini = guc_exec_queue_fini,
- .set_priority = guc_exec_queue_set_priority,
- .set_timeslice = guc_exec_queue_set_timeslice,
- .set_preempt_timeout = guc_exec_queue_set_preempt_timeout,
- .suspend = guc_exec_queue_suspend,
- .suspend_wait = guc_exec_queue_suspend_wait,
- .resume = guc_exec_queue_resume,
- .reset_status = guc_exec_queue_reset_status,
-};
-
-static void guc_exec_queue_stop(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched = &q->guc->sched;
-
- /* Stop scheduling + flush any DRM scheduler operations */
- xe_sched_submission_stop(sched);
-
- /* Clean up lost G2H + reset engine state */
- if (exec_queue_registered(q)) {
- if (exec_queue_extra_ref(q) || xe_exec_queue_is_lr(q))
- xe_exec_queue_put(q);
- else if (exec_queue_destroyed(q))
- __guc_exec_queue_fini(guc, q);
- }
- if (q->guc->suspend_pending) {
- set_exec_queue_suspended(q);
- suspend_fence_signal(q);
- }
- atomic_and(EXEC_QUEUE_STATE_WEDGED | EXEC_QUEUE_STATE_BANNED |
- EXEC_QUEUE_STATE_KILLED | EXEC_QUEUE_STATE_DESTROYED |
- EXEC_QUEUE_STATE_SUSPENDED,
- &q->guc->state);
- q->guc->resume_time = 0;
- trace_xe_exec_queue_stop(q);
-
- /*
- * Ban any engine (aside from kernel and engines used for VM ops) with a
- * started but not complete job or if a job has gone through a GT reset
- * more than twice.
- */
- if (!(q->flags & (EXEC_QUEUE_FLAG_KERNEL | EXEC_QUEUE_FLAG_VM))) {
- struct xe_sched_job *job = xe_sched_first_pending_job(sched);
- bool ban = false;
-
- if (job) {
- if ((xe_sched_job_started(job) &&
- !xe_sched_job_completed(job)) ||
- xe_sched_invalidate_job(job, 2)) {
- trace_xe_sched_job_ban(job);
- ban = true;
- }
- } else if (xe_exec_queue_is_lr(q) &&
- (xe_lrc_ring_head(q->lrc[0]) != xe_lrc_ring_tail(q->lrc[0]))) {
- ban = true;
- }
-
- if (ban) {
- set_exec_queue_banned(q);
- xe_guc_exec_queue_trigger_cleanup(q);
- }
- }
-}
-
-int xe_guc_submit_reset_prepare(struct xe_guc *guc)
-{
- int ret;
-
- /*
- * Using an atomic here rather than submission_state.lock as this
- * function can be called while holding the CT lock (engine reset
- * failure). submission_state.lock needs the CT lock to resubmit jobs.
- * Atomic is not ideal, but it works to prevent against concurrent reset
- * and releasing any TDRs waiting on guc->submission_state.stopped.
- */
- ret = atomic_fetch_or(1, &guc->submission_state.stopped);
- smp_wmb();
- wake_up_all(&guc->ct.wq);
-
- return ret;
-}
-
-void xe_guc_submit_reset_wait(struct xe_guc *guc)
-{
- wait_event(guc->ct.wq, xe_device_wedged(guc_to_xe(guc)) ||
- !guc_read_stopped(guc));
-}
-
-void xe_guc_submit_stop(struct xe_guc *guc)
-{
- struct xe_exec_queue *q;
- unsigned long index;
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, guc_read_stopped(guc) == 1);
-
- mutex_lock(&guc->submission_state.lock);
-
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- guc_exec_queue_stop(guc, q);
-
- mutex_unlock(&guc->submission_state.lock);
-
- /*
- * No one can enter the backend at this point, aside from new engine
- * creation which is protected by guc->submission_state.lock.
- */
-
-}
-
-static void guc_exec_queue_start(struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched = &q->guc->sched;
-
- if (!exec_queue_killed_or_banned_or_wedged(q)) {
- int i;
-
- trace_xe_exec_queue_resubmit(q);
- for (i = 0; i < q->width; ++i)
- xe_lrc_set_ring_head(q->lrc[i], q->lrc[i]->ring.tail);
- xe_sched_resubmit_jobs(sched);
- }
-
- xe_sched_submission_start(sched);
-}
-
-int xe_guc_submit_start(struct xe_guc *guc)
-{
- struct xe_exec_queue *q;
- unsigned long index;
- struct xe_device *xe = guc_to_xe(guc);
-
- xe_assert(xe, guc_read_stopped(guc) == 1);
-
- mutex_lock(&guc->submission_state.lock);
- atomic_dec(&guc->submission_state.stopped);
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- guc_exec_queue_start(q);
- mutex_unlock(&guc->submission_state.lock);
-
- wake_up_all(&guc->ct.wq);
-
- return 0;
-}
-
-static struct xe_exec_queue *
-g2h_exec_queue_lookup(struct xe_guc *guc, u32 guc_id)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
-
- if (unlikely(guc_id >= GUC_ID_MAX)) {
- drm_err(&xe->drm, "Invalid guc_id %u", guc_id);
- return NULL;
- }
-
- q = xa_load(&guc->submission_state.exec_queue_lookup, guc_id);
- if (unlikely(!q)) {
- drm_err(&xe->drm, "Not engine present for guc_id %u", guc_id);
- return NULL;
- }
-
- xe_assert(xe, guc_id >= q->guc->id);
- xe_assert(xe, guc_id < (q->guc->id + q->width));
-
- return q;
-}
-
-static void deregister_exec_queue(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- u32 action[] = {
- XE_GUC_ACTION_DEREGISTER_CONTEXT,
- q->guc->id,
- };
-
- xe_gt_assert(guc_to_gt(guc), exec_queue_destroyed(q));
- xe_gt_assert(guc_to_gt(guc), exec_queue_registered(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_disable(q));
- xe_gt_assert(guc_to_gt(guc), !exec_queue_pending_enable(q));
-
- trace_xe_exec_queue_deregister(q);
-
- xe_guc_ct_send_g2h_handler(&guc->ct, action, ARRAY_SIZE(action));
-}
-
-static void handle_sched_done(struct xe_guc *guc, struct xe_exec_queue *q,
- u32 runnable_state)
-{
- trace_xe_exec_queue_scheduling_done(q);
-
- if (runnable_state == 1) {
- xe_gt_assert(guc_to_gt(guc), exec_queue_pending_enable(q));
-
- q->guc->resume_time = ktime_get();
- clear_exec_queue_pending_enable(q);
- smp_wmb();
- wake_up_all(&guc->ct.wq);
- } else {
- bool check_timeout = exec_queue_check_timeout(q);
-
- xe_gt_assert(guc_to_gt(guc), runnable_state == 0);
- xe_gt_assert(guc_to_gt(guc), exec_queue_pending_disable(q));
-
- clear_exec_queue_pending_disable(q);
- if (q->guc->suspend_pending) {
- suspend_fence_signal(q);
- } else {
- if (exec_queue_banned(q) || check_timeout) {
- smp_wmb();
- wake_up_all(&guc->ct.wq);
- }
- if (!check_timeout)
- deregister_exec_queue(guc, q);
- }
- }
-}
-
-int xe_guc_sched_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
- u32 runnable_state = msg[1];
-
- if (unlikely(len < 2)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- if (unlikely(!exec_queue_pending_enable(q) &&
- !exec_queue_pending_disable(q))) {
- xe_gt_err(guc_to_gt(guc),
- "SCHED_DONE: Unexpected engine state 0x%04x, guc_id=%d, runnable_state=%u",
- atomic_read(&q->guc->state), q->guc->id,
- runnable_state);
- return -EPROTO;
- }
-
- handle_sched_done(guc, q, runnable_state);
-
- return 0;
-}
-
-static void handle_deregister_done(struct xe_guc *guc, struct xe_exec_queue *q)
-{
- trace_xe_exec_queue_deregister_done(q);
-
- clear_exec_queue_registered(q);
-
- if (exec_queue_extra_ref(q) || xe_exec_queue_is_lr(q))
- xe_exec_queue_put(q);
- else
- __guc_exec_queue_fini(guc, q);
-}
-
-int xe_guc_deregister_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
-
- if (unlikely(len < 1)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- if (!exec_queue_destroyed(q) || exec_queue_pending_disable(q) ||
- exec_queue_pending_enable(q) || exec_queue_enabled(q)) {
- xe_gt_err(guc_to_gt(guc),
- "DEREGISTER_DONE: Unexpected engine state 0x%04x, guc_id=%d",
- atomic_read(&q->guc->state), q->guc->id);
- return -EPROTO;
- }
-
- handle_deregister_done(guc, q);
-
- return 0;
-}
-
-int xe_guc_exec_queue_reset_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_gt *gt = guc_to_gt(guc);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
-
- if (unlikely(len < 1)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- xe_gt_info(gt, "Engine reset: engine_class=%s, logical_mask: 0x%x, guc_id=%d",
- xe_hw_engine_class_to_str(q->class), q->logical_mask, guc_id);
-
- /* FIXME: Do error capture, most likely async */
-
- trace_xe_exec_queue_reset(q);
-
- /*
- * A banned engine is a NOP at this point (came from
- * guc_exec_queue_timedout_job). Otherwise, kick drm scheduler to cancel
- * jobs by setting timeout of the job to the minimum value kicking
- * guc_exec_queue_timedout_job.
- */
- set_exec_queue_reset(q);
- if (!exec_queue_banned(q) && !exec_queue_check_timeout(q))
- xe_guc_exec_queue_trigger_cleanup(q);
-
- return 0;
-}
-
-int xe_guc_exec_queue_memory_cat_error_handler(struct xe_guc *guc, u32 *msg,
- u32 len)
-{
- struct xe_gt *gt = guc_to_gt(guc);
- struct xe_device *xe = guc_to_xe(guc);
- struct xe_exec_queue *q;
- u32 guc_id = msg[0];
-
- if (unlikely(len < 1)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- q = g2h_exec_queue_lookup(guc, guc_id);
- if (unlikely(!q))
- return -EPROTO;
-
- xe_gt_dbg(gt, "Engine memory cat error: engine_class=%s, logical_mask: 0x%x, guc_id=%d",
- xe_hw_engine_class_to_str(q->class), q->logical_mask, guc_id);
-
- trace_xe_exec_queue_memory_cat_error(q);
-
- /* Treat the same as engine reset */
- set_exec_queue_reset(q);
- if (!exec_queue_banned(q) && !exec_queue_check_timeout(q))
- xe_guc_exec_queue_trigger_cleanup(q);
-
- return 0;
-}
-
-int xe_guc_exec_queue_reset_failure_handler(struct xe_guc *guc, u32 *msg, u32 len)
-{
- struct xe_device *xe = guc_to_xe(guc);
- u8 guc_class, instance;
- u32 reason;
-
- if (unlikely(len != 3)) {
- drm_err(&xe->drm, "Invalid length %u", len);
- return -EPROTO;
- }
-
- guc_class = msg[0];
- instance = msg[1];
- reason = msg[2];
-
- /* Unexpected failure of a hardware feature, log an actual error */
- drm_err(&xe->drm, "GuC engine reset request failed on %d:%d because 0x%08X",
- guc_class, instance, reason);
-
- xe_gt_reset_async(guc_to_gt(guc));
-
- return 0;
-}
-
-static void
-guc_exec_queue_wq_snapshot_capture(struct xe_exec_queue *q,
- struct xe_guc_submit_exec_queue_snapshot *snapshot)
-{
- struct xe_guc *guc = exec_queue_to_guc(q);
- struct xe_device *xe = guc_to_xe(guc);
- struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
- int i;
-
- snapshot->guc.wqi_head = q->guc->wqi_head;
- snapshot->guc.wqi_tail = q->guc->wqi_tail;
- snapshot->parallel.wq_desc.head = parallel_read(xe, map, wq_desc.head);
- snapshot->parallel.wq_desc.tail = parallel_read(xe, map, wq_desc.tail);
- snapshot->parallel.wq_desc.status = parallel_read(xe, map,
- wq_desc.wq_status);
-
- if (snapshot->parallel.wq_desc.head !=
- snapshot->parallel.wq_desc.tail) {
- for (i = snapshot->parallel.wq_desc.head;
- i != snapshot->parallel.wq_desc.tail;
- i = (i + sizeof(u32)) % WQ_SIZE)
- snapshot->parallel.wq[i / sizeof(u32)] =
- parallel_read(xe, map, wq[i / sizeof(u32)]);
- }
-}
-
-static void
-guc_exec_queue_wq_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snapshot,
- struct drm_printer *p)
-{
- int i;
-
- drm_printf(p, "\tWQ head: %u (internal), %d (memory)\n",
- snapshot->guc.wqi_head, snapshot->parallel.wq_desc.head);
- drm_printf(p, "\tWQ tail: %u (internal), %d (memory)\n",
- snapshot->guc.wqi_tail, snapshot->parallel.wq_desc.tail);
- drm_printf(p, "\tWQ status: %u\n", snapshot->parallel.wq_desc.status);
-
- if (snapshot->parallel.wq_desc.head !=
- snapshot->parallel.wq_desc.tail) {
- for (i = snapshot->parallel.wq_desc.head;
- i != snapshot->parallel.wq_desc.tail;
- i = (i + sizeof(u32)) % WQ_SIZE)
- drm_printf(p, "\tWQ[%zu]: 0x%08x\n", i / sizeof(u32),
- snapshot->parallel.wq[i / sizeof(u32)]);
- }
-}
-
-/**
- * xe_guc_exec_queue_snapshot_capture - Take a quick snapshot of the GuC Engine.
- * @q: faulty exec queue
- *
- * This can be printed out in a later stage like during dev_coredump
- * analysis.
- *
- * Returns: a GuC Submit Engine snapshot object that must be freed by the
- * caller, using `xe_guc_exec_queue_snapshot_free`.
- */
-struct xe_guc_submit_exec_queue_snapshot *
-xe_guc_exec_queue_snapshot_capture(struct xe_exec_queue *q)
-{
- struct xe_gpu_scheduler *sched = &q->guc->sched;
- struct xe_guc_submit_exec_queue_snapshot *snapshot;
- int i;
-
- snapshot = kzalloc(sizeof(*snapshot), GFP_ATOMIC);
-
- if (!snapshot)
- return NULL;
-
- snapshot->guc.id = q->guc->id;
- memcpy(&snapshot->name, &q->name, sizeof(snapshot->name));
- snapshot->class = q->class;
- snapshot->logical_mask = q->logical_mask;
- snapshot->width = q->width;
- snapshot->refcount = kref_read(&q->refcount);
- snapshot->sched_timeout = sched->base.timeout;
- snapshot->sched_props.timeslice_us = q->sched_props.timeslice_us;
- snapshot->sched_props.preempt_timeout_us =
- q->sched_props.preempt_timeout_us;
-
- snapshot->lrc = kmalloc_array(q->width, sizeof(struct xe_lrc_snapshot *),
- GFP_ATOMIC);
-
- if (snapshot->lrc) {
- for (i = 0; i < q->width; ++i) {
- struct xe_lrc *lrc = q->lrc[i];
-
- snapshot->lrc[i] = xe_lrc_snapshot_capture(lrc);
- }
- }
-
- snapshot->schedule_state = atomic_read(&q->guc->state);
- snapshot->exec_queue_flags = q->flags;
-
- snapshot->parallel_execution = xe_exec_queue_is_parallel(q);
- if (snapshot->parallel_execution)
- guc_exec_queue_wq_snapshot_capture(q, snapshot);
-
- spin_lock(&sched->base.job_list_lock);
- snapshot->pending_list_size = list_count_nodes(&sched->base.pending_list);
- snapshot->pending_list = kmalloc_array(snapshot->pending_list_size,
- sizeof(struct pending_list_snapshot),
- GFP_ATOMIC);
-
- if (snapshot->pending_list) {
- struct xe_sched_job *job_iter;
-
- i = 0;
- list_for_each_entry(job_iter, &sched->base.pending_list, drm.list) {
- snapshot->pending_list[i].seqno =
- xe_sched_job_seqno(job_iter);
- snapshot->pending_list[i].fence =
- dma_fence_is_signaled(job_iter->fence) ? 1 : 0;
- snapshot->pending_list[i].finished =
- dma_fence_is_signaled(&job_iter->drm.s_fence->finished)
- ? 1 : 0;
- i++;
- }
- }
-
- spin_unlock(&sched->base.job_list_lock);
-
- return snapshot;
-}
-
-/**
- * xe_guc_exec_queue_snapshot_capture_delayed - Take delayed part of snapshot of the GuC Engine.
- * @snapshot: Previously captured snapshot of job.
- *
- * This captures some data that requires taking some locks, so it cannot be done in signaling path.
- */
-void
-xe_guc_exec_queue_snapshot_capture_delayed(struct xe_guc_submit_exec_queue_snapshot *snapshot)
-{
- int i;
-
- if (!snapshot || !snapshot->lrc)
- return;
-
- for (i = 0; i < snapshot->width; ++i)
- xe_lrc_snapshot_capture_delayed(snapshot->lrc[i]);
-}
-
-/**
- * xe_guc_exec_queue_snapshot_print - Print out a given GuC Engine snapshot.
- * @snapshot: GuC Submit Engine snapshot object.
- * @p: drm_printer where it will be printed out.
- *
- * This function prints out a given GuC Submit Engine snapshot object.
- */
-void
-xe_guc_exec_queue_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snapshot,
- struct drm_printer *p)
-{
- int i;
-
- if (!snapshot)
- return;
-
- drm_printf(p, "\nGuC ID: %d\n", snapshot->guc.id);
- drm_printf(p, "\tName: %s\n", snapshot->name);
- drm_printf(p, "\tClass: %d\n", snapshot->class);
- drm_printf(p, "\tLogical mask: 0x%x\n", snapshot->logical_mask);
- drm_printf(p, "\tWidth: %d\n", snapshot->width);
- drm_printf(p, "\tRef: %d\n", snapshot->refcount);
- drm_printf(p, "\tTimeout: %ld (ms)\n", snapshot->sched_timeout);
- drm_printf(p, "\tTimeslice: %u (us)\n",
- snapshot->sched_props.timeslice_us);
- drm_printf(p, "\tPreempt timeout: %u (us)\n",
- snapshot->sched_props.preempt_timeout_us);
-
- for (i = 0; snapshot->lrc && i < snapshot->width; ++i)
- xe_lrc_snapshot_print(snapshot->lrc[i], p);
-
- drm_printf(p, "\tSchedule State: 0x%x\n", snapshot->schedule_state);
- drm_printf(p, "\tFlags: 0x%lx\n", snapshot->exec_queue_flags);
-
- if (snapshot->parallel_execution)
- guc_exec_queue_wq_snapshot_print(snapshot, p);
-
- for (i = 0; snapshot->pending_list && i < snapshot->pending_list_size;
- i++)
- drm_printf(p, "\tJob: seqno=%d, fence=%d, finished=%d\n",
- snapshot->pending_list[i].seqno,
- snapshot->pending_list[i].fence,
- snapshot->pending_list[i].finished);
-}
-
-/**
- * xe_guc_exec_queue_snapshot_free - Free all allocated objects for a given
- * snapshot.
- * @snapshot: GuC Submit Engine snapshot object.
- *
- * This function free all the memory that needed to be allocated at capture
- * time.
- */
-void xe_guc_exec_queue_snapshot_free(struct xe_guc_submit_exec_queue_snapshot *snapshot)
-{
- int i;
-
- if (!snapshot)
- return;
-
- if (snapshot->lrc) {
- for (i = 0; i < snapshot->width; i++)
- xe_lrc_snapshot_free(snapshot->lrc[i]);
- kfree(snapshot->lrc);
- }
- kfree(snapshot->pending_list);
- kfree(snapshot);
-}
-
-static void guc_exec_queue_print(struct xe_exec_queue *q, struct drm_printer *p)
-{
- struct xe_guc_submit_exec_queue_snapshot *snapshot;
-
- snapshot = xe_guc_exec_queue_snapshot_capture(q);
- xe_guc_exec_queue_snapshot_print(snapshot, p);
- xe_guc_exec_queue_snapshot_free(snapshot);
-}
-
-/**
- * xe_guc_submit_print - GuC Submit Print.
- * @guc: GuC.
- * @p: drm_printer where it will be printed out.
- *
- * This function capture and prints snapshots of **all** GuC Engines.
- */
-void xe_guc_submit_print(struct xe_guc *guc, struct drm_printer *p)
-{
- struct xe_exec_queue *q;
- unsigned long index;
-
- if (!xe_device_uc_enabled(guc_to_xe(guc)))
- return;
-
- mutex_lock(&guc->submission_state.lock);
- xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
- guc_exec_queue_print(q, p);
- mutex_unlock(&guc->submission_state.lock);
-}