Merge tag 'drm-for-v4.15' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "This is the main drm pull request for v4.15.

  Core:
   - Atomic object lifetime fixes
   - Atomic iterator improvements
   - Sparse/smatch fixes
   - Legacy kms ioctls to be interruptible
   - EDID override improvements
   - fb/gem helper cleanups
   - Simple outreachy patches
   - Documentation improvements
   - Fix dma-buf rcu races
   - DRM mode object leasing for improving VR use cases.
   - vgaarb improvements for non-x86 platforms.

  New driver:
   - tve200: Faraday Technology TVE200 block.

     This "TV Encoder" encodes a ITU-T BT.656 stream and can be found in
     the StorLink SL3516 (later Cortina Systems CS3516) as well as the
     Grain Media GM8180.

  New bridges:
   - SiI9234 support

  New panels:
   - S6E63J0X03, OTM8009A, Seiko 43WVF1G, 7" rpi touch panel, Toshiba
     LT089AC19000, Innolux AT043TN24

  i915:
   - Remove Coffeelake from alpha support
   - Cannonlake workarounds
   - Infoframe refactoring for DisplayPort
   - VBT updates
   - DisplayPort vswing/emph/buffer translation refactoring
   - CCS fixes
   - Restore GPU clock boost on missed vblanks
   - Scatter list updates for userptr allocations
   - Gen9+ transition watermarks
   - Display IPC (Isochronous Priority Control)
   - Private PAT management
   - GVT: improved error handling and pci config sanitizing
   - Execlist refactoring
   - Transparent Huge Page support
   - User defined priorities support
   - HuC/GuC firmware refactoring
   - DP MST fixes
   - eDP power sequencing fixes
   - Use RCU instead of stop_machine
   - PSR state tracking support
   - Eviction fixes
   - BDW DP aux channel timeout fixes
   - LSPCON fixes
   - Cannonlake PLL fixes

  amdgpu:
   - Per VM BO support
   - Powerplay cleanups
   - CI powerplay support
   - PASID mgr for kfd
   - SR-IOV fixes
   - initial GPU reset for vega10
   - Prime mmap support
   - TTM updates
   - Clock query interface for Raven
   - Fence to handle ioctl
   - UVD encode ring support on Polaris
   - Transparent huge page DMA support
   - Compute LRU pipe tweaks
   - BO flag to allow buffers to opt out of implicit sync
   - CTX priority setting API
   - VRAM lost infrastructure plumbing

  qxl:
   - fix flicker since atomic rework

  amdkfd:
   - Further improvements from internal AMD tree
   - Usermode events
   - Drop radeon support

  nouveau:
   - Pascal temperature sensor support
   - Improved BAR2 handling
   - MMU rework to support Pascal MMU

  exynos:
   - Improved HDMI/mixer support
   - HDMI audio interface support

  tegra:
   - Prep work for tegra186
   - Cleanup/fixes

  msm:
   - Preemption support for a5xx
   - Display fixes for 8x96 (snapdragon 820)
   - Async cursor plane fixes
   - FW loading rework
   - GPU debugging improvements

  vc4:
   - Prep for DSI panels
   - fix T-format tiling scanout
   - New madvise ioctl

  Rockchip:
   - LVDS support

  omapdrm:
   - omap4 HDMI CEC support

  etnaviv:
   - GPU performance counters groundwork

  sun4i:
   - refactor driver load + TCON backend
   - HDMI improvements
   - A31 support
   - Misc fixes

  udl:
   - Probe/EDID read fixes.

  tilcdc:
   - Misc fixes.

  pl111:
   - Support more variants

  adv7511:
   - Improve EDID handling.
   - HDMI CEC support

  sii8620:
   - Add remote control support"

* tag 'drm-for-v4.15' of git://people.freedesktop.org/~airlied/linux: (1480 commits)
  drm/rockchip: analogix_dp: Use mutex rather than spinlock
  drm/mode_object: fix documentation for object lookups.
  drm/i915: Reorder context-close to avoid calling i915_vma_close() under RCU
  drm/i915: Move init_clock_gating() back to where it was
  drm/i915: Prune the reservation shared fence array
  drm/i915: Idle the GPU before shinking everything
  drm/i915: Lock llist_del_first() vs llist_del_all()
  drm/i915: Calculate ironlake intermediate watermarks correctly, v2.
  drm/i915: Disable lazy PPGTT page table optimization for vGPU
  drm/i915/execlists: Remove the priority "optimisation"
  drm/i915: Filter out spurious execlists context-switch interrupts
  drm/amdgpu: use irq-safe lock for kiq->ring_lock
  drm/amdgpu: bypass lru touch for KIQ ring submission
  drm/amdgpu: Potential uninitialized variable in amdgpu_vm_update_directories()
  drm/amdgpu: potential uninitialized variable in amdgpu_vce_ring_parse_cs()
  drm/amd/powerplay: initialize a variable before using it
  drm/amd/powerplay: suppress KASAN out of bounds warning in vega10_populate_all_memory_levels
  drm/amd/amdgpu: fix evicted VRAM bo adjudgement condition
  drm/vblank: Tune drm_crtc_accurate_vblank_count() WARN down to a debug
  drm/rockchip: add CONFIG_OF dependency for lvds
  ...

1 files changed, 233 insertions, 96 deletions

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c
index 6d512c062ae3..1ba7289684aa 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c
@@ -31,147 +31,293 @@
 
 struct nv50_instmem {
 	struct nvkm_instmem base;
-	unsigned long lock_flags;
-	spinlock_t lock;
 	u64 addr;
+
+	/* Mappings that can be evicted when BAR2 space has been exhausted. */
+	struct list_head lru;
 };
 
 /******************************************************************************
  * instmem object implementation
  *****************************************************************************/
-#define nv50_instobj(p) container_of((p), struct nv50_instobj, memory)
+#define nv50_instobj(p) container_of((p), struct nv50_instobj, base.memory)
 
 struct nv50_instobj {
-	struct nvkm_memory memory;
+	struct nvkm_instobj base;
 	struct nv50_instmem *imem;
-	struct nvkm_mem *mem;
-	struct nvkm_vma bar;
+	struct nvkm_memory *ram;
+	struct nvkm_vma *bar;
+	refcount_t maps;
 	void *map;
+	struct list_head lru;
 };
 
-static enum nvkm_memory_target
-nv50_instobj_target(struct nvkm_memory *memory)
+static void
+nv50_instobj_wr32_slow(struct nvkm_memory *memory, u64 offset, u32 data)
 {
-	return NVKM_MEM_TARGET_VRAM;
+	struct nv50_instobj *iobj = nv50_instobj(memory);
+	struct nv50_instmem *imem = iobj->imem;
+	struct nvkm_device *device = imem->base.subdev.device;
+	u64 base = (nvkm_memory_addr(iobj->ram) + offset) & 0xffffff00000ULL;
+	u64 addr = (nvkm_memory_addr(iobj->ram) + offset) & 0x000000fffffULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&imem->base.lock, flags);
+	if (unlikely(imem->addr != base)) {
+		nvkm_wr32(device, 0x001700, base >> 16);
+		imem->addr = base;
+	}
+	nvkm_wr32(device, 0x700000 + addr, data);
+	spin_unlock_irqrestore(&imem->base.lock, flags);
 }
 
-static u64
-nv50_instobj_addr(struct nvkm_memory *memory)
+static u32
+nv50_instobj_rd32_slow(struct nvkm_memory *memory, u64 offset)
 {
-	return nv50_instobj(memory)->mem->offset;
+	struct nv50_instobj *iobj = nv50_instobj(memory);
+	struct nv50_instmem *imem = iobj->imem;
+	struct nvkm_device *device = imem->base.subdev.device;
+	u64 base = (nvkm_memory_addr(iobj->ram) + offset) & 0xffffff00000ULL;
+	u64 addr = (nvkm_memory_addr(iobj->ram) + offset) & 0x000000fffffULL;
+	u32 data;
+	unsigned long flags;
+
+	spin_lock_irqsave(&imem->base.lock, flags);
+	if (unlikely(imem->addr != base)) {
+		nvkm_wr32(device, 0x001700, base >> 16);
+		imem->addr = base;
+	}
+	data = nvkm_rd32(device, 0x700000 + addr);
+	spin_unlock_irqrestore(&imem->base.lock, flags);
+	return data;
 }
 
-static u64
-nv50_instobj_size(struct nvkm_memory *memory)
+static const struct nvkm_memory_ptrs
+nv50_instobj_slow = {
+	.rd32 = nv50_instobj_rd32_slow,
+	.wr32 = nv50_instobj_wr32_slow,
+};
+
+static void
+nv50_instobj_wr32(struct nvkm_memory *memory, u64 offset, u32 data)
 {
-	return (u64)nv50_instobj(memory)->mem->size << NVKM_RAM_MM_SHIFT;
+	iowrite32_native(data, nv50_instobj(memory)->map + offset);
 }
 
+static u32
+nv50_instobj_rd32(struct nvkm_memory *memory, u64 offset)
+{
+	return ioread32_native(nv50_instobj(memory)->map + offset);
+}
+
+static const struct nvkm_memory_ptrs
+nv50_instobj_fast = {
+	.rd32 = nv50_instobj_rd32,
+	.wr32 = nv50_instobj_wr32,
+};
+
 static void
-nv50_instobj_boot(struct nvkm_memory *memory, struct nvkm_vm *vm)
+nv50_instobj_kmap(struct nv50_instobj *iobj, struct nvkm_vmm *vmm)
 {
-	struct nv50_instobj *iobj = nv50_instobj(memory);
-	struct nvkm_subdev *subdev = &iobj->imem->base.subdev;
+	struct nv50_instmem *imem = iobj->imem;
+	struct nv50_instobj *eobj;
+	struct nvkm_memory *memory = &iobj->base.memory;
+	struct nvkm_subdev *subdev = &imem->base.subdev;
 	struct nvkm_device *device = subdev->device;
+	struct nvkm_vma *bar = NULL, *ebar;
 	u64 size = nvkm_memory_size(memory);
-	void __iomem *map;
+	void *emap;
 	int ret;
 
-	iobj->map = ERR_PTR(-ENOMEM);
-
-	ret = nvkm_vm_get(vm, size, 12, NV_MEM_ACCESS_RW, &iobj->bar);
-	if (ret == 0) {
-		map = ioremap(device->func->resource_addr(device, 3) +
-			      (u32)iobj->bar.offset, size);
-		if (map) {
-			nvkm_memory_map(memory, &iobj->bar, 0);
-			iobj->map = map;
-		} else {
-			nvkm_warn(subdev, "PRAMIN ioremap failed\n");
-			nvkm_vm_put(&iobj->bar);
+	/* Attempt to allocate BAR2 address-space and map the object
+	 * into it.  The lock has to be dropped while doing this due
+	 * to the possibility of recursion for page table allocation.
+	 */
+	mutex_unlock(&subdev->mutex);
+	while ((ret = nvkm_vmm_get(vmm, 12, size, &bar))) {
+		/* Evict unused mappings, and keep retrying until we either
+		 * succeed,or there's no more objects left on the LRU.
+		 */
+		mutex_lock(&subdev->mutex);
+		eobj = list_first_entry_or_null(&imem->lru, typeof(*eobj), lru);
+		if (eobj) {
+			nvkm_debug(subdev, "evict %016llx %016llx @ %016llx\n",
+				   nvkm_memory_addr(&eobj->base.memory),
+				   nvkm_memory_size(&eobj->base.memory),
+				   eobj->bar->addr);
+			list_del_init(&eobj->lru);
+			ebar = eobj->bar;
+			eobj->bar = NULL;
+			emap = eobj->map;
+			eobj->map = NULL;
 		}
-	} else {
-		nvkm_warn(subdev, "PRAMIN exhausted\n");
+		mutex_unlock(&subdev->mutex);
+		if (!eobj)
+			break;
+		iounmap(emap);
+		nvkm_vmm_put(vmm, &ebar);
 	}
+
+	if (ret == 0)
+		ret = nvkm_memory_map(memory, 0, vmm, bar, NULL, 0);
+	mutex_lock(&subdev->mutex);
+	if (ret || iobj->bar) {
+		/* We either failed, or another thread beat us. */
+		mutex_unlock(&subdev->mutex);
+		nvkm_vmm_put(vmm, &bar);
+		mutex_lock(&subdev->mutex);
+		return;
+	}
+
+	/* Make the mapping visible to the host. */
+	iobj->bar = bar;
+	iobj->map = ioremap_wc(device->func->resource_addr(device, 3) +
+			       (u32)iobj->bar->addr, size);
+	if (!iobj->map) {
+		nvkm_warn(subdev, "PRAMIN ioremap failed\n");
+		nvkm_vmm_put(vmm, &iobj->bar);
+	}
+}
+
+static int
+nv50_instobj_map(struct nvkm_memory *memory, u64 offset, struct nvkm_vmm *vmm,
+		 struct nvkm_vma *vma, void *argv, u32 argc)
+{
+	memory = nv50_instobj(memory)->ram;
+	return nvkm_memory_map(memory, offset, vmm, vma, argv, argc);
 }
 
 static void
 nv50_instobj_release(struct nvkm_memory *memory)
 {
-	struct nv50_instmem *imem = nv50_instobj(memory)->imem;
-	spin_unlock_irqrestore(&imem->lock, imem->lock_flags);
+	struct nv50_instobj *iobj = nv50_instobj(memory);
+	struct nv50_instmem *imem = iobj->imem;
+	struct nvkm_subdev *subdev = &imem->base.subdev;
+
+	wmb();
+	nvkm_bar_flush(subdev->device->bar);
+
+	if (refcount_dec_and_mutex_lock(&iobj->maps, &subdev->mutex)) {
+		/* Add the now-unused mapping to the LRU instead of directly
+		 * unmapping it here, in case we need to map it again later.
+		 */
+		if (likely(iobj->lru.next) && iobj->map) {
+			BUG_ON(!list_empty(&iobj->lru));
+			list_add_tail(&iobj->lru, &imem->lru);
+		}
+
+		/* Switch back to NULL accessors when last map is gone. */
+		iobj->base.memory.ptrs = NULL;
+		mutex_unlock(&subdev->mutex);
+	}
 }
 
 static void __iomem *
 nv50_instobj_acquire(struct nvkm_memory *memory)
 {
 	struct nv50_instobj *iobj = nv50_instobj(memory);
-	struct nv50_instmem *imem = iobj->imem;
-	struct nvkm_bar *bar = imem->base.subdev.device->bar;
-	struct nvkm_vm *vm;
-	unsigned long flags;
+	struct nvkm_instmem *imem = &iobj->imem->base;
+	struct nvkm_vmm *vmm;
+	void __iomem *map = NULL;
 
-	if (!iobj->map && (vm = nvkm_bar_kmap(bar)))
-		nvkm_memory_boot(memory, vm);
-	if (!IS_ERR_OR_NULL(iobj->map))
+	/* Already mapped? */
+	if (refcount_inc_not_zero(&iobj->maps))
 		return iobj->map;
 
-	spin_lock_irqsave(&imem->lock, flags);
-	imem->lock_flags = flags;
-	return NULL;
-}
+	/* Take the lock, and re-check that another thread hasn't
+	 * already mapped the object in the meantime.
+	 */
+	mutex_lock(&imem->subdev.mutex);
+	if (refcount_inc_not_zero(&iobj->maps)) {
+		mutex_unlock(&imem->subdev.mutex);
+		return iobj->map;
+	}
 
-static u32
-nv50_instobj_rd32(struct nvkm_memory *memory, u64 offset)
-{
-	struct nv50_instobj *iobj = nv50_instobj(memory);
-	struct nv50_instmem *imem = iobj->imem;
-	struct nvkm_device *device = imem->base.subdev.device;
-	u64 base = (iobj->mem->offset + offset) & 0xffffff00000ULL;
-	u64 addr = (iobj->mem->offset + offset) & 0x000000fffffULL;
-	u32 data;
+	/* Attempt to get a direct CPU mapping of the object. */
+	if ((vmm = nvkm_bar_bar2_vmm(imem->subdev.device))) {
+		if (!iobj->map)
+			nv50_instobj_kmap(iobj, vmm);
+		map = iobj->map;
+	}
 
-	if (unlikely(imem->addr != base)) {
-		nvkm_wr32(device, 0x001700, base >> 16);
-		imem->addr = base;
+	if (!refcount_inc_not_zero(&iobj->maps)) {
+		/* Exclude object from eviction while it's being accessed. */
+		if (likely(iobj->lru.next))
+			list_del_init(&iobj->lru);
+
+		if (map)
+			iobj->base.memory.ptrs = &nv50_instobj_fast;
+		else
+			iobj->base.memory.ptrs = &nv50_instobj_slow;
+		refcount_inc(&iobj->maps);
 	}
-	data = nvkm_rd32(device, 0x700000 + addr);
-	return data;
+
+	mutex_unlock(&imem->subdev.mutex);
+	return map;
 }
 
 static void
-nv50_instobj_wr32(struct nvkm_memory *memory, u64 offset, u32 data)
+nv50_instobj_boot(struct nvkm_memory *memory, struct nvkm_vmm *vmm)
 {
 	struct nv50_instobj *iobj = nv50_instobj(memory);
-	struct nv50_instmem *imem = iobj->imem;
-	struct nvkm_device *device = imem->base.subdev.device;
-	u64 base = (iobj->mem->offset + offset) & 0xffffff00000ULL;
-	u64 addr = (iobj->mem->offset + offset) & 0x000000fffffULL;
-
-	if (unlikely(imem->addr != base)) {
-		nvkm_wr32(device, 0x001700, base >> 16);
-		imem->addr = base;
+	struct nvkm_instmem *imem = &iobj->imem->base;
+
+	/* Exclude bootstrapped objects (ie. the page tables for the
+	 * instmem BAR itself) from eviction.
+	 */
+	mutex_lock(&imem->subdev.mutex);
+	if (likely(iobj->lru.next)) {
+		list_del_init(&iobj->lru);
+		iobj->lru.next = NULL;
 	}
-	nvkm_wr32(device, 0x700000 + addr, data);
+
+	nv50_instobj_kmap(iobj, vmm);
+	nvkm_instmem_boot(imem);
+	mutex_unlock(&imem->subdev.mutex);
 }
 
-static void
-nv50_instobj_map(struct nvkm_memory *memory, struct nvkm_vma *vma, u64 offset)
+static u64
+nv50_instobj_size(struct nvkm_memory *memory)
 {
-	struct nv50_instobj *iobj = nv50_instobj(memory);
-	nvkm_vm_map_at(vma, offset, iobj->mem);
+	return nvkm_memory_size(nv50_instobj(memory)->ram);
+}
+
+static u64
+nv50_instobj_addr(struct nvkm_memory *memory)
+{
+	return nvkm_memory_addr(nv50_instobj(memory)->ram);
+}
+
+static enum nvkm_memory_target
+nv50_instobj_target(struct nvkm_memory *memory)
+{
+	return nvkm_memory_target(nv50_instobj(memory)->ram);
 }
 
 static void *
 nv50_instobj_dtor(struct nvkm_memory *memory)
 {
 	struct nv50_instobj *iobj = nv50_instobj(memory);
-	struct nvkm_ram *ram = iobj->imem->base.subdev.device->fb->ram;
-	if (!IS_ERR_OR_NULL(iobj->map)) {
-		nvkm_vm_put(&iobj->bar);
-		iounmap(iobj->map);
+	struct nvkm_instmem *imem = &iobj->imem->base;
+	struct nvkm_vma *bar;
+	void *map = map;
+
+	mutex_lock(&imem->subdev.mutex);
+	if (likely(iobj->lru.next))
+		list_del(&iobj->lru);
+	map = iobj->map;
+	bar = iobj->bar;
+	mutex_unlock(&imem->subdev.mutex);
+
+	if (map) {
+		struct nvkm_vmm *vmm = nvkm_bar_bar2_vmm(imem->subdev.device);
+		iounmap(map);
+		if (likely(vmm)) /* Can be NULL during BAR destructor. */
+			nvkm_vmm_put(vmm, &bar);
 	}
-	ram->func->put(ram, &iobj->mem);
+
+	nvkm_memory_unref(&iobj->ram);
+	nvkm_instobj_dtor(imem, &iobj->base);
 	return iobj;
 }
 
@@ -184,8 +330,6 @@ nv50_instobj_func = {
 	.boot = nv50_instobj_boot,
 	.acquire = nv50_instobj_acquire,
 	.release = nv50_instobj_release,
-	.rd32 = nv50_instobj_rd32,
-	.wr32 = nv50_instobj_wr32,
 	.map = nv50_instobj_map,
 };
 
@@ -195,25 +339,19 @@ nv50_instobj_new(struct nvkm_instmem *base, u32 size, u32 align, bool zero,
 {
 	struct nv50_instmem *imem = nv50_instmem(base);
 	struct nv50_instobj *iobj;
-	struct nvkm_ram *ram = imem->base.subdev.device->fb->ram;
-	int ret;
+	struct nvkm_device *device = imem->base.subdev.device;
+	u8 page = max(order_base_2(align), 12);
 
 	if (!(iobj = kzalloc(sizeof(*iobj), GFP_KERNEL)))
 		return -ENOMEM;
-	*pmemory = &iobj->memory;
+	*pmemory = &iobj->base.memory;
 
-	nvkm_memory_ctor(&nv50_instobj_func, &iobj->memory);
+	nvkm_instobj_ctor(&nv50_instobj_func, &imem->base, &iobj->base);
 	iobj->imem = imem;
+	refcount_set(&iobj->maps, 0);
+	INIT_LIST_HEAD(&iobj->lru);
 
-	size  = max((size  + 4095) & ~4095, (u32)4096);
-	align = max((align + 4095) & ~4095, (u32)4096);
-
-	ret = ram->func->get(ram, size, align, 0, 0x800, &iobj->mem);
-	if (ret)
-		return ret;
-
-	iobj->mem->page_shift = 12;
-	return 0;
+	return nvkm_ram_get(device, 0, 1, page, size, true, true, &iobj->ram);
 }
 
 /******************************************************************************
@@ -230,7 +368,6 @@ static const struct nvkm_instmem_func
 nv50_instmem = {
 	.fini = nv50_instmem_fini,
 	.memory_new = nv50_instobj_new,
-	.persistent = false,
 	.zero = false,
 };
 
@@ -243,7 +380,7 @@ nv50_instmem_new(struct nvkm_device *device, int index,
 	if (!(imem = kzalloc(sizeof(*imem), GFP_KERNEL)))
 		return -ENOMEM;
 	nvkm_instmem_ctor(&nv50_instmem, device, index, &imem->base);
-	spin_lock_init(&imem->lock);
+	INIT_LIST_HEAD(&imem->lru);
 	*pimem = &imem->base;
 	return 0;
 }