diff options
Diffstat (limited to 'drivers/gpu/drm/nouveau/nvkm')
210 files changed, 5428 insertions, 3193 deletions
diff --git a/drivers/gpu/drm/nouveau/nvkm/Kbuild b/drivers/gpu/drm/nouveau/nvkm/Kbuild index 2832147b676c..e664378f6eda 100644 --- a/drivers/gpu/drm/nouveau/nvkm/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/Kbuild @@ -1,3 +1,4 @@ include $(src)/nvkm/core/Kbuild +include $(src)/nvkm/falcon/Kbuild include $(src)/nvkm/subdev/Kbuild include $(src)/nvkm/engine/Kbuild diff --git a/drivers/gpu/drm/nouveau/nvkm/core/client.c b/drivers/gpu/drm/nouveau/nvkm/core/client.c index e1943910858e..0d3a896892b4 100644 --- a/drivers/gpu/drm/nouveau/nvkm/core/client.c +++ b/drivers/gpu/drm/nouveau/nvkm/core/client.c @@ -31,6 +31,43 @@ #include <nvif/if0000.h> #include <nvif/unpack.h> +static int +nvkm_uclient_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + union { + struct nvif_client_v0 v0; + } *args = argv; + struct nvkm_client *client; + int ret = -ENOSYS; + + if (!(ret = nvif_unpack(ret, &argv, &argc, args->v0, 0, 0, false))){ + args->v0.name[sizeof(args->v0.name) - 1] = 0; + ret = nvkm_client_new(args->v0.name, args->v0.device, NULL, + NULL, oclass->client->ntfy, &client); + if (ret) + return ret; + } else + return ret; + + client->object.client = oclass->client; + client->object.handle = oclass->handle; + client->object.route = oclass->route; + client->object.token = oclass->token; + client->object.object = oclass->object; + client->debug = oclass->client->debug; + *pobject = &client->object; + return 0; +} + +const struct nvkm_sclass +nvkm_uclient_sclass = { + .oclass = NVIF_CLASS_CLIENT, + .minver = 0, + .maxver = 0, + .ctor = nvkm_uclient_new, +}; + struct nvkm_client_notify { struct nvkm_client *client; struct nvkm_notify n; @@ -138,17 +175,30 @@ nvkm_client_notify_new(struct nvkm_object *object, return ret; } +static const struct nvkm_object_func nvkm_client; +struct nvkm_client * +nvkm_client_search(struct nvkm_client *client, u64 handle) +{ + struct nvkm_object *object; + + object = nvkm_object_search(client, handle, &nvkm_client); + if (IS_ERR(object)) + return (void *)object; + + return nvkm_client(object); +} + static int -nvkm_client_mthd_devlist(struct nvkm_object *object, void *data, u32 size) +nvkm_client_mthd_devlist(struct nvkm_client *client, void *data, u32 size) { union { - struct nv_client_devlist_v0 v0; + struct nvif_client_devlist_v0 v0; } *args = data; int ret = -ENOSYS; - nvif_ioctl(object, "client devlist size %d\n", size); + nvif_ioctl(&client->object, "client devlist size %d\n", size); if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, true))) { - nvif_ioctl(object, "client devlist vers %d count %d\n", + nvif_ioctl(&client->object, "client devlist vers %d count %d\n", args->v0.version, args->v0.count); if (size == sizeof(args->v0.device[0]) * args->v0.count) { ret = nvkm_device_list(args->v0.device, args->v0.count); @@ -167,9 +217,10 @@ nvkm_client_mthd_devlist(struct nvkm_object *object, void *data, u32 size) static int nvkm_client_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) { + struct nvkm_client *client = nvkm_client(object); switch (mthd) { - case NV_CLIENT_DEVLIST: - return nvkm_client_mthd_devlist(object, data, size); + case NVIF_CLIENT_V0_DEVLIST: + return nvkm_client_mthd_devlist(client, data, size); default: break; } @@ -190,7 +241,8 @@ nvkm_client_child_get(struct nvkm_object *object, int index, const struct nvkm_sclass *sclass; switch (index) { - case 0: sclass = &nvkm_udevice_sclass; break; + case 0: sclass = &nvkm_uclient_sclass; break; + case 1: sclass = &nvkm_udevice_sclass; break; default: return -EINVAL; } @@ -200,110 +252,54 @@ nvkm_client_child_get(struct nvkm_object *object, int index, return 0; } -static const struct nvkm_object_func -nvkm_client_object_func = { - .mthd = nvkm_client_mthd, - .sclass = nvkm_client_child_get, -}; - -void -nvkm_client_remove(struct nvkm_client *client, struct nvkm_object *object) -{ - if (!RB_EMPTY_NODE(&object->node)) - rb_erase(&object->node, &client->objroot); -} - -bool -nvkm_client_insert(struct nvkm_client *client, struct nvkm_object *object) -{ - struct rb_node **ptr = &client->objroot.rb_node; - struct rb_node *parent = NULL; - - while (*ptr) { - struct nvkm_object *this = - container_of(*ptr, typeof(*this), node); - parent = *ptr; - if (object->object < this->object) - ptr = &parent->rb_left; - else - if (object->object > this->object) - ptr = &parent->rb_right; - else - return false; - } - - rb_link_node(&object->node, parent, ptr); - rb_insert_color(&object->node, &client->objroot); - return true; -} - -struct nvkm_object * -nvkm_client_search(struct nvkm_client *client, u64 handle) -{ - struct rb_node *node = client->objroot.rb_node; - while (node) { - struct nvkm_object *object = - container_of(node, typeof(*object), node); - if (handle < object->object) - node = node->rb_left; - else - if (handle > object->object) - node = node->rb_right; - else - return object; - } - return NULL; -} - -int -nvkm_client_fini(struct nvkm_client *client, bool suspend) +static int +nvkm_client_fini(struct nvkm_object *object, bool suspend) { - struct nvkm_object *object = &client->object; + struct nvkm_client *client = nvkm_client(object); const char *name[2] = { "fini", "suspend" }; int i; nvif_debug(object, "%s notify\n", name[suspend]); for (i = 0; i < ARRAY_SIZE(client->notify); i++) nvkm_client_notify_put(client, i); - return nvkm_object_fini(&client->object, suspend); -} - -int -nvkm_client_init(struct nvkm_client *client) -{ - return nvkm_object_init(&client->object); + return 0; } -void -nvkm_client_del(struct nvkm_client **pclient) +static void * +nvkm_client_dtor(struct nvkm_object *object) { - struct nvkm_client *client = *pclient; + struct nvkm_client *client = nvkm_client(object); int i; - if (client) { - nvkm_client_fini(client, false); - for (i = 0; i < ARRAY_SIZE(client->notify); i++) - nvkm_client_notify_del(client, i); - nvkm_object_dtor(&client->object); - kfree(*pclient); - *pclient = NULL; - } + for (i = 0; i < ARRAY_SIZE(client->notify); i++) + nvkm_client_notify_del(client, i); + return client; } +static const struct nvkm_object_func +nvkm_client = { + .dtor = nvkm_client_dtor, + .fini = nvkm_client_fini, + .mthd = nvkm_client_mthd, + .sclass = nvkm_client_child_get, +}; + int nvkm_client_new(const char *name, u64 device, const char *cfg, - const char *dbg, struct nvkm_client **pclient) + const char *dbg, + int (*ntfy)(const void *, u32, const void *, u32), + struct nvkm_client **pclient) { - struct nvkm_oclass oclass = {}; + struct nvkm_oclass oclass = { .base = nvkm_uclient_sclass }; struct nvkm_client *client; if (!(client = *pclient = kzalloc(sizeof(*client), GFP_KERNEL))) return -ENOMEM; oclass.client = client; - nvkm_object_ctor(&nvkm_client_object_func, &oclass, &client->object); + nvkm_object_ctor(&nvkm_client, &oclass, &client->object); snprintf(client->name, sizeof(client->name), "%s", name); client->device = device; client->debug = nvkm_dbgopt(dbg, "CLIENT"); client->objroot = RB_ROOT; - client->dmaroot = RB_ROOT; + client->ntfy = ntfy; return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/core/engine.c b/drivers/gpu/drm/nouveau/nvkm/core/engine.c index ee8e5831fe37..b6c916954a10 100644 --- a/drivers/gpu/drm/nouveau/nvkm/core/engine.c +++ b/drivers/gpu/drm/nouveau/nvkm/core/engine.c @@ -27,6 +27,14 @@ #include <subdev/fb.h> +bool +nvkm_engine_chsw_load(struct nvkm_engine *engine) +{ + if (engine->func->chsw_load) + return engine->func->chsw_load(engine); + return false; +} + void nvkm_engine_unref(struct nvkm_engine **pengine) { diff --git a/drivers/gpu/drm/nouveau/nvkm/core/firmware.c b/drivers/gpu/drm/nouveau/nvkm/core/firmware.c index 34ecd4a7e0c1..058ff46b5f16 100644 --- a/drivers/gpu/drm/nouveau/nvkm/core/firmware.c +++ b/drivers/gpu/drm/nouveau/nvkm/core/firmware.c @@ -20,6 +20,7 @@ * DEALINGS IN THE SOFTWARE. */ #include <core/device.h> +#include <core/firmware.h> /** * nvkm_firmware_get - load firmware from the official nvidia/chip/ directory diff --git a/drivers/gpu/drm/nouveau/nvkm/core/ioctl.c b/drivers/gpu/drm/nouveau/nvkm/core/ioctl.c index b0db51847c36..be19bbe56bba 100644 --- a/drivers/gpu/drm/nouveau/nvkm/core/ioctl.c +++ b/drivers/gpu/drm/nouveau/nvkm/core/ioctl.c @@ -29,7 +29,8 @@ #include <nvif/ioctl.h> static int -nvkm_ioctl_nop(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_nop(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_nop_v0 v0; @@ -46,7 +47,8 @@ nvkm_ioctl_nop(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_sclass(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_sclass(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_sclass_v0 v0; @@ -78,12 +80,12 @@ nvkm_ioctl_sclass(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_new(struct nvkm_object *parent, void *data, u32 size) +nvkm_ioctl_new(struct nvkm_client *client, + struct nvkm_object *parent, void *data, u32 size) { union { struct nvif_ioctl_new_v0 v0; } *args = data; - struct nvkm_client *client = parent->client; struct nvkm_object *object = NULL; struct nvkm_oclass oclass; int ret = -ENOSYS, i = 0; @@ -104,9 +106,11 @@ nvkm_ioctl_new(struct nvkm_object *parent, void *data, u32 size) do { memset(&oclass, 0x00, sizeof(oclass)); - oclass.client = client; oclass.handle = args->v0.handle; + oclass.route = args->v0.route; + oclass.token = args->v0.token; oclass.object = args->v0.object; + oclass.client = client; oclass.parent = parent; ret = parent->func->sclass(parent, i++, &oclass); if (ret) @@ -125,10 +129,7 @@ nvkm_ioctl_new(struct nvkm_object *parent, void *data, u32 size) ret = nvkm_object_init(object); if (ret == 0) { list_add(&object->head, &parent->tree); - object->route = args->v0.route; - object->token = args->v0.token; - object->object = args->v0.object; - if (nvkm_client_insert(client, object)) { + if (nvkm_object_insert(object)) { client->data = object; return 0; } @@ -142,7 +143,8 @@ nvkm_ioctl_new(struct nvkm_object *parent, void *data, u32 size) } static int -nvkm_ioctl_del(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_del(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_del none; @@ -156,11 +158,12 @@ nvkm_ioctl_del(struct nvkm_object *object, void *data, u32 size) nvkm_object_del(&object); } - return ret; + return ret ? ret : 1; } static int -nvkm_ioctl_mthd(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_mthd(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_mthd_v0 v0; @@ -179,7 +182,8 @@ nvkm_ioctl_mthd(struct nvkm_object *object, void *data, u32 size) static int -nvkm_ioctl_rd(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_rd(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_rd_v0 v0; @@ -218,7 +222,8 @@ nvkm_ioctl_rd(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_wr(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_wr(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_wr_v0 v0; @@ -246,7 +251,8 @@ nvkm_ioctl_wr(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_map(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_map(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_map_v0 v0; @@ -264,7 +270,8 @@ nvkm_ioctl_map(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_unmap(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_unmap(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_unmap none; @@ -280,7 +287,8 @@ nvkm_ioctl_unmap(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_ntfy_new(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_ntfy_new(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { union { struct nvif_ioctl_ntfy_new_v0 v0; @@ -306,9 +314,9 @@ nvkm_ioctl_ntfy_new(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_ntfy_del(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_ntfy_del(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { - struct nvkm_client *client = object->client; union { struct nvif_ioctl_ntfy_del_v0 v0; } *args = data; @@ -325,9 +333,9 @@ nvkm_ioctl_ntfy_del(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_ntfy_get(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_ntfy_get(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { - struct nvkm_client *client = object->client; union { struct nvif_ioctl_ntfy_get_v0 v0; } *args = data; @@ -344,9 +352,9 @@ nvkm_ioctl_ntfy_get(struct nvkm_object *object, void *data, u32 size) } static int -nvkm_ioctl_ntfy_put(struct nvkm_object *object, void *data, u32 size) +nvkm_ioctl_ntfy_put(struct nvkm_client *client, + struct nvkm_object *object, void *data, u32 size) { - struct nvkm_client *client = object->client; union { struct nvif_ioctl_ntfy_put_v0 v0; } *args = data; @@ -364,7 +372,7 @@ nvkm_ioctl_ntfy_put(struct nvkm_object *object, void *data, u32 size) static struct { int version; - int (*func)(struct nvkm_object *, void *, u32); + int (*func)(struct nvkm_client *, struct nvkm_object *, void *, u32); } nvkm_ioctl_v0[] = { { 0x00, nvkm_ioctl_nop }, @@ -389,13 +397,10 @@ nvkm_ioctl_path(struct nvkm_client *client, u64 handle, u32 type, struct nvkm_object *object; int ret; - if (handle) - object = nvkm_client_search(client, handle); - else - object = &client->object; - if (unlikely(!object)) { + object = nvkm_object_search(client, handle, NULL); + if (IS_ERR(object)) { nvif_ioctl(&client->object, "object not found\n"); - return -ENOENT; + return PTR_ERR(object); } if (owner != NVIF_IOCTL_V0_OWNER_ANY && owner != object->route) { @@ -407,7 +412,7 @@ nvkm_ioctl_path(struct nvkm_client *client, u64 handle, u32 type, if (ret = -EINVAL, type < ARRAY_SIZE(nvkm_ioctl_v0)) { if (nvkm_ioctl_v0[type].version == 0) - ret = nvkm_ioctl_v0[type].func(object, data, size); + ret = nvkm_ioctl_v0[type].func(client, object, data, size); } return ret; @@ -436,12 +441,13 @@ nvkm_ioctl(struct nvkm_client *client, bool supervisor, &args->v0.route, &args->v0.token); } - nvif_ioctl(object, "return %d\n", ret); - if (hack) { - *hack = client->data; - client->data = NULL; + if (ret != 1) { + nvif_ioctl(object, "return %d\n", ret); + if (hack) { + *hack = client->data; + client->data = NULL; + } } - client->super = false; return ret; } diff --git a/drivers/gpu/drm/nouveau/nvkm/core/mm.c b/drivers/gpu/drm/nouveau/nvkm/core/mm.c index 09a1eee8fd33..fd19d652a7ab 100644 --- a/drivers/gpu/drm/nouveau/nvkm/core/mm.c +++ b/drivers/gpu/drm/nouveau/nvkm/core/mm.c @@ -147,6 +147,7 @@ nvkm_mm_head(struct nvkm_mm *mm, u8 heap, u8 type, u32 size_max, u32 size_min, if (!this) return -ENOMEM; + this->next = NULL; this->type = type; list_del(&this->fl_entry); *pnode = this; @@ -225,6 +226,7 @@ nvkm_mm_tail(struct nvkm_mm *mm, u8 heap, u8 type, u32 size_max, u32 size_min, if (!this) return -ENOMEM; + this->next = NULL; this->type = type; list_del(&this->fl_entry); *pnode = this; diff --git a/drivers/gpu/drm/nouveau/nvkm/core/object.c b/drivers/gpu/drm/nouveau/nvkm/core/object.c index 67aa7223dcd7..89d2e9da11c7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/core/object.c +++ b/drivers/gpu/drm/nouveau/nvkm/core/object.c @@ -25,6 +25,65 @@ #include <core/client.h> #include <core/engine.h> +struct nvkm_object * +nvkm_object_search(struct nvkm_client *client, u64 handle, + const struct nvkm_object_func *func) +{ + struct nvkm_object *object; + + if (handle) { + struct rb_node *node = client->objroot.rb_node; + while (node) { + object = rb_entry(node, typeof(*object), node); + if (handle < object->object) + node = node->rb_left; + else + if (handle > object->object) + node = node->rb_right; + else + goto done; + } + return ERR_PTR(-ENOENT); + } else { + object = &client->object; + } + +done: + if (unlikely(func && object->func != func)) + return ERR_PTR(-EINVAL); + return object; +} + +void +nvkm_object_remove(struct nvkm_object *object) +{ + if (!RB_EMPTY_NODE(&object->node)) + rb_erase(&object->node, &object->client->objroot); +} + +bool +nvkm_object_insert(struct nvkm_object *object) +{ + struct rb_node **ptr = &object->client->objroot.rb_node; + struct rb_node *parent = NULL; + + while (*ptr) { + struct nvkm_object *this = rb_entry(*ptr, typeof(*this), node); + parent = *ptr; + if (object->object < this->object) + ptr = &parent->rb_left; + else + if (object->object > this->object) + ptr = &parent->rb_right; + else + return false; + } + + rb_link_node(&object->node, parent, ptr); + rb_insert_color(&object->node, &object->client->objroot); + return true; +} + int nvkm_object_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) { @@ -214,7 +273,7 @@ nvkm_object_del(struct nvkm_object **pobject) struct nvkm_object *object = *pobject; if (object && !WARN_ON(!object->func)) { *pobject = nvkm_object_dtor(object); - nvkm_client_remove(object->client, object); + nvkm_object_remove(object); list_del(&object->head); kfree(*pobject); *pobject = NULL; @@ -230,6 +289,9 @@ nvkm_object_ctor(const struct nvkm_object_func *func, object->engine = nvkm_engine_ref(oclass->engine); object->oclass = oclass->base.oclass; object->handle = oclass->handle; + object->route = oclass->route; + object->token = oclass->token; + object->object = oclass->object; INIT_LIST_HEAD(&object->head); INIT_LIST_HEAD(&object->tree); RB_CLEAR_NODE(&object->node); diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild index a4458a8eb30a..255d81ccf916 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild @@ -4,4 +4,4 @@ nvkm-y += nvkm/engine/ce/gk104.o nvkm-y += nvkm/engine/ce/gm107.o nvkm-y += nvkm/engine/ce/gm200.o nvkm-y += nvkm/engine/ce/gp100.o -nvkm-y += nvkm/engine/ce/gp104.o +nvkm-y += nvkm/engine/ce/gp102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h index 05bb65608dfe..d9ca9636a3e3 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gf100_ce_data[] = { +static uint32_t gf100_ce_data[] = { /* 0x0000: ctx_object */ 0x00000000, /* 0x0004: ctx_query_address_high */ @@ -171,7 +171,7 @@ uint32_t gf100_ce_data[] = { 0x00000800, }; -uint32_t gf100_ce_code[] = { +static uint32_t gf100_ce_code[] = { /* 0x0000: main */ 0x04fe04bd, 0x3517f000, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h index 972281d10f38..f0a1cf31c7ca 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gt215_ce_data[] = { +static uint32_t gt215_ce_data[] = { /* 0x0000: ctx_object */ 0x00000000, /* 0x0004: ctx_dma */ @@ -183,7 +183,7 @@ uint32_t gt215_ce_data[] = { 0x00000800, }; -uint32_t gt215_ce_code[] = { +static uint32_t gt215_ce_code[] = { /* 0x0000: main */ 0x04fe04bd, 0x3517f000, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c index 20e019788a53..985c8f653874 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c @@ -27,7 +27,7 @@ #include <nvif/class.h> static const struct nvkm_engine_func -gp104_ce = { +gp102_ce = { .intr = gp100_ce_intr, .sclass = { { -1, -1, PASCAL_DMA_COPY_B }, @@ -37,8 +37,8 @@ gp104_ce = { }; int -gp104_ce_new(struct nvkm_device *device, int index, +gp102_ce_new(struct nvkm_device *device, int index, struct nvkm_engine **pengine) { - return nvkm_engine_new_(&gp104_ce, device, index, true, pengine); + return nvkm_engine_new_(&gp102_ce, device, index, true, pengine); } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c index 7218a067a6c5..273562dd6bbd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c @@ -993,7 +993,7 @@ nv92_chipset = { .mc = g84_mc_new, .mmu = nv50_mmu_new, .mxm = nv50_mxm_new, - .pci = g84_pci_new, + .pci = g92_pci_new, .therm = g84_therm_new, .timer = nv41_timer_new, .volt = nv40_volt_new, @@ -1357,7 +1357,7 @@ nvc0_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .ce[1] = gf100_ce_new, .disp = gt215_disp_new, @@ -1394,7 +1394,7 @@ nvc1_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .disp = gt215_disp_new, .dma = gf100_dma_new, @@ -1430,7 +1430,7 @@ nvc3_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .disp = gt215_disp_new, .dma = gf100_dma_new, @@ -1466,7 +1466,7 @@ nvc4_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .ce[1] = gf100_ce_new, .disp = gt215_disp_new, @@ -1503,7 +1503,7 @@ nvc8_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .ce[1] = gf100_ce_new, .disp = gt215_disp_new, @@ -1540,7 +1540,7 @@ nvce_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .ce[1] = gf100_ce_new, .disp = gt215_disp_new, @@ -1577,7 +1577,7 @@ nvcf_chipset = { .pmu = gf100_pmu_new, .therm = gt215_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .disp = gt215_disp_new, .dma = gf100_dma_new, @@ -1612,6 +1612,7 @@ nvd7_chipset = { .pci = gf106_pci_new, .therm = gf119_therm_new, .timer = nv41_timer_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .disp = gf119_disp_new, .dma = gf119_dma_new, @@ -1647,7 +1648,7 @@ nvd9_chipset = { .pmu = gf119_pmu_new, .therm = gf119_therm_new, .timer = nv41_timer_new, - .volt = nv40_volt_new, + .volt = gf100_volt_new, .ce[0] = gf100_ce_new, .disp = gf119_disp_new, .dma = gf119_dma_new, @@ -1851,7 +1852,7 @@ nvf1_chipset = { .fb = gk104_fb_new, .fuse = gf100_fuse_new, .gpio = gk104_gpio_new, - .i2c = gf119_i2c_new, + .i2c = gk104_i2c_new, .ibus = gk104_ibus_new, .iccsense = gf100_iccsense_new, .imem = nv50_instmem_new, @@ -1965,7 +1966,7 @@ nv117_chipset = { .fb = gm107_fb_new, .fuse = gm107_fuse_new, .gpio = gk104_gpio_new, - .i2c = gf119_i2c_new, + .i2c = gk104_i2c_new, .ibus = gk104_ibus_new, .iccsense = gf100_iccsense_new, .imem = nv50_instmem_new, @@ -1999,7 +2000,7 @@ nv118_chipset = { .fb = gm107_fb_new, .fuse = gm107_fuse_new, .gpio = gk104_gpio_new, - .i2c = gf119_i2c_new, + .i2c = gk104_i2c_new, .ibus = gk104_ibus_new, .iccsense = gf100_iccsense_new, .imem = nv50_instmem_new, @@ -2130,13 +2131,14 @@ nv12b_chipset = { .bar = gk20a_bar_new, .bus = gf100_bus_new, .clk = gm20b_clk_new, - .fb = gk20a_fb_new, + .fb = gm20b_fb_new, .fuse = gm107_fuse_new, .ibus = gk20a_ibus_new, .imem = gk20a_instmem_new, .ltc = gm200_ltc_new, .mc = gk20a_mc_new, .mmu = gf100_mmu_new, + .pmu = gm20b_pmu_new, .secboot = gm20b_secboot_new, .timer = gk20a_timer_new, .top = gk104_top_new, @@ -2166,6 +2168,7 @@ nv130_chipset = { .mmu = gf100_mmu_new, .secboot = gm200_secboot_new, .pci = gp100_pci_new, + .pmu = gp100_pmu_new, .timer = gk20a_timer_new, .top = gk104_top_new, .ce[0] = gp100_ce_new, @@ -2182,13 +2185,71 @@ nv130_chipset = { }; static const struct nvkm_device_chip +nv132_chipset = { + .name = "GP102", + .bar = gf100_bar_new, + .bios = nvkm_bios_new, + .bus = gf100_bus_new, + .devinit = gm200_devinit_new, + .fb = gp102_fb_new, + .fuse = gm107_fuse_new, + .gpio = gk104_gpio_new, + .i2c = gm200_i2c_new, + .ibus = gm200_ibus_new, + .imem = nv50_instmem_new, + .ltc = gp100_ltc_new, + .mc = gp100_mc_new, + .mmu = gf100_mmu_new, + .pci = gp100_pci_new, + .pmu = gp102_pmu_new, + .timer = gk20a_timer_new, + .top = gk104_top_new, + .ce[0] = gp102_ce_new, + .ce[1] = gp102_ce_new, + .ce[2] = gp102_ce_new, + .ce[3] = gp102_ce_new, + .disp = gp102_disp_new, + .dma = gf119_dma_new, + .fifo = gp100_fifo_new, +}; + +static const struct nvkm_device_chip nv134_chipset = { .name = "GP104", .bar = gf100_bar_new, .bios = nvkm_bios_new, .bus = gf100_bus_new, .devinit = gm200_devinit_new, - .fb = gp104_fb_new, + .fb = gp102_fb_new, + .fuse = gm107_fuse_new, + .gpio = gk104_gpio_new, + .i2c = gm200_i2c_new, + .ibus = gm200_ibus_new, + .imem = nv50_instmem_new, + .ltc = gp100_ltc_new, + .mc = gp100_mc_new, + .mmu = gf100_mmu_new, + .pci = gp100_pci_new, + .pmu = gp102_pmu_new, + .timer = gk20a_timer_new, + .top = gk104_top_new, + .ce[0] = gp102_ce_new, + .ce[1] = gp102_ce_new, + .ce[2] = gp102_ce_new, + .ce[3] = gp102_ce_new, + .disp = gp102_disp_new, + .dma = gf119_dma_new, + .fifo = gp100_fifo_new, +}; + +static const struct nvkm_device_chip +nv136_chipset = { + .name = "GP106", + .bar = gf100_bar_new, + .bios = nvkm_bios_new, + .bus = gf100_bus_new, + .devinit = gm200_devinit_new, + .fb = gp102_fb_new, .fuse = gm107_fuse_new, .gpio = gk104_gpio_new, .i2c = gm200_i2c_new, @@ -2198,13 +2259,14 @@ nv134_chipset = { .mc = gp100_mc_new, .mmu = gf100_mmu_new, .pci = gp100_pci_new, + .pmu = gp102_pmu_new, .timer = gk20a_timer_new, .top = gk104_top_new, - .ce[0] = gp104_ce_new, - .ce[1] = gp104_ce_new, - .ce[2] = gp104_ce_new, - .ce[3] = gp104_ce_new, - .disp = gp104_disp_new, + .ce[0] = gp102_ce_new, + .ce[1] = gp102_ce_new, + .ce[2] = gp102_ce_new, + .ce[3] = gp102_ce_new, + .disp = gp102_disp_new, .dma = gf119_dma_new, .fifo = gp100_fifo_new, }; @@ -2643,7 +2705,9 @@ nvkm_device_ctor(const struct nvkm_device_func *func, case 0x126: device->chip = &nv126_chipset; break; case 0x12b: device->chip = &nv12b_chipset; break; case 0x130: device->chip = &nv130_chipset; break; + case 0x132: device->chip = &nv132_chipset; break; case 0x134: device->chip = &nv134_chipset; break; + case 0x136: device->chip = &nv136_chipset; break; default: nvdev_error(device, "unknown chipset (%08x)\n", boot0); goto done; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c index 62ad0300cfa5..74a1ffa425f7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c @@ -1665,14 +1665,31 @@ nvkm_device_pci_new(struct pci_dev *pci_dev, const char *cfg, const char *dbg, *pdevice = &pdev->device; pdev->pdev = pci_dev; - return nvkm_device_ctor(&nvkm_device_pci_func, quirk, &pci_dev->dev, - pci_is_pcie(pci_dev) ? NVKM_DEVICE_PCIE : - pci_find_capability(pci_dev, PCI_CAP_ID_AGP) ? - NVKM_DEVICE_AGP : NVKM_DEVICE_PCI, - (u64)pci_domain_nr(pci_dev->bus) << 32 | - pci_dev->bus->number << 16 | - PCI_SLOT(pci_dev->devfn) << 8 | - PCI_FUNC(pci_dev->devfn), name, - cfg, dbg, detect, mmio, subdev_mask, - &pdev->device); + ret = nvkm_device_ctor(&nvkm_device_pci_func, quirk, &pci_dev->dev, + pci_is_pcie(pci_dev) ? NVKM_DEVICE_PCIE : + pci_find_capability(pci_dev, PCI_CAP_ID_AGP) ? + NVKM_DEVICE_AGP : NVKM_DEVICE_PCI, + (u64)pci_domain_nr(pci_dev->bus) << 32 | + pci_dev->bus->number << 16 | + PCI_SLOT(pci_dev->devfn) << 8 | + PCI_FUNC(pci_dev->devfn), name, + cfg, dbg, detect, mmio, subdev_mask, + &pdev->device); + + if (ret) + return ret; + + /* + * Set a preliminary DMA mask based on the .dma_bits member of the + * MMU subdevice. This allows other subdevices to create DMA mappings + * in their init() or oneinit() methods, which may be called before the + * TTM layer sets the DMA mask definitively. + * This is necessary for platforms where the default DMA mask of 32 + * does not cover any system memory, i.e., when all RAM is > 4 GB. + */ + if (pdev->device.mmu) + dma_set_mask_and_coherent(&pci_dev->dev, + DMA_BIT_MASK(pdev->device.mmu->dma_bits)); + + return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c index 9b638bd905ff..f2bc0b7d9b93 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c @@ -102,7 +102,7 @@ nvkm_device_tegra_probe_iommu(struct nvkm_device_tegra *tdev) if (iommu_present(&platform_bus_type)) { tdev->iommu.domain = iommu_domain_alloc(&platform_bus_type); - if (IS_ERR(tdev->iommu.domain)) + if (!tdev->iommu.domain) goto error; /* diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c index 79a8f71cf788..513ee6b79553 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c @@ -326,7 +326,7 @@ nvkm_udevice = { .sclass = nvkm_udevice_child_get, }; -int +static int nvkm_udevice_new(const struct nvkm_oclass *oclass, void *data, u32 size, struct nvkm_object **pobject) { diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild index 77a52b54a31e..fa05d16ae948 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild @@ -11,7 +11,7 @@ nvkm-y += nvkm/engine/disp/gk110.o nvkm-y += nvkm/engine/disp/gm107.o nvkm-y += nvkm/engine/disp/gm200.o nvkm-y += nvkm/engine/disp/gp100.o -nvkm-y += nvkm/engine/disp/gp104.o +nvkm-y += nvkm/engine/disp/gp102.o nvkm-y += nvkm/engine/disp/outp.o nvkm-y += nvkm/engine/disp/outpdp.o @@ -48,14 +48,14 @@ nvkm-y += nvkm/engine/disp/rootgk110.o nvkm-y += nvkm/engine/disp/rootgm107.o nvkm-y += nvkm/engine/disp/rootgm200.o nvkm-y += nvkm/engine/disp/rootgp100.o -nvkm-y += nvkm/engine/disp/rootgp104.o +nvkm-y += nvkm/engine/disp/rootgp102.o nvkm-y += nvkm/engine/disp/channv50.o nvkm-y += nvkm/engine/disp/changf119.o nvkm-y += nvkm/engine/disp/dmacnv50.o nvkm-y += nvkm/engine/disp/dmacgf119.o -nvkm-y += nvkm/engine/disp/dmacgp104.o +nvkm-y += nvkm/engine/disp/dmacgp102.o nvkm-y += nvkm/engine/disp/basenv50.o nvkm-y += nvkm/engine/disp/baseg84.o @@ -64,7 +64,7 @@ nvkm-y += nvkm/engine/disp/basegt215.o nvkm-y += nvkm/engine/disp/basegf119.o nvkm-y += nvkm/engine/disp/basegk104.o nvkm-y += nvkm/engine/disp/basegk110.o -nvkm-y += nvkm/engine/disp/basegp104.o +nvkm-y += nvkm/engine/disp/basegp102.o nvkm-y += nvkm/engine/disp/corenv50.o nvkm-y += nvkm/engine/disp/coreg84.o @@ -77,7 +77,7 @@ nvkm-y += nvkm/engine/disp/coregk110.o nvkm-y += nvkm/engine/disp/coregm107.o nvkm-y += nvkm/engine/disp/coregm200.o nvkm-y += nvkm/engine/disp/coregp100.o -nvkm-y += nvkm/engine/disp/coregp104.o +nvkm-y += nvkm/engine/disp/coregp102.o nvkm-y += nvkm/engine/disp/ovlynv50.o nvkm-y += nvkm/engine/disp/ovlyg84.o @@ -85,7 +85,7 @@ nvkm-y += nvkm/engine/disp/ovlygt200.o nvkm-y += nvkm/engine/disp/ovlygt215.o nvkm-y += nvkm/engine/disp/ovlygf119.o nvkm-y += nvkm/engine/disp/ovlygk104.o -nvkm-y += nvkm/engine/disp/ovlygp104.o +nvkm-y += nvkm/engine/disp/ovlygp102.o nvkm-y += nvkm/engine/disp/piocnv50.o nvkm-y += nvkm/engine/disp/piocgf119.o @@ -95,9 +95,11 @@ nvkm-y += nvkm/engine/disp/cursg84.o nvkm-y += nvkm/engine/disp/cursgt215.o nvkm-y += nvkm/engine/disp/cursgf119.o nvkm-y += nvkm/engine/disp/cursgk104.o +nvkm-y += nvkm/engine/disp/cursgp102.o nvkm-y += nvkm/engine/disp/oimmnv50.o nvkm-y += nvkm/engine/disp/oimmg84.o nvkm-y += nvkm/engine/disp/oimmgt215.o nvkm-y += nvkm/engine/disp/oimmgf119.o nvkm-y += nvkm/engine/disp/oimmgk104.o +nvkm-y += nvkm/engine/disp/oimmgp102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/basegp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/basegp102.c index 51688e37c54e..8a3cdeef8d2c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/basegp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/basegp102.c @@ -27,12 +27,12 @@ #include <nvif/class.h> const struct nv50_disp_dmac_oclass -gp104_disp_base_oclass = { +gp102_disp_base_oclass = { .base.oclass = GK110_DISP_BASE_CHANNEL_DMA, .base.minver = 0, .base.maxver = 0, .ctor = nv50_disp_base_new, - .func = &gp104_disp_dmac_func, + .func = &gp102_disp_dmac_func, .mthd = &gf119_disp_base_chan_mthd, .chid = 1, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c index dd2953bc9264..524a24eae1a0 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c @@ -82,7 +82,7 @@ nv50_disp_chan_mthd(struct nv50_disp_chan *chan, int debug) if (mthd->addr) { snprintf(cname_, sizeof(cname_), "%s %d", - mthd->name, chan->chid); + mthd->name, chan->chid.user); cname = cname_; } @@ -139,7 +139,7 @@ nv50_disp_chan_uevent_ctor(struct nvkm_object *object, void *data, u32 size, if (!(ret = nvif_unvers(ret, &data, &size, args->none))) { notify->size = sizeof(struct nvif_notify_uevent_rep); notify->types = 1; - notify->index = chan->chid; + notify->index = chan->chid.user; return 0; } @@ -153,27 +153,27 @@ nv50_disp_chan_uevent = { .fini = nv50_disp_chan_uevent_fini, }; -int +static int nv50_disp_chan_rd32(struct nvkm_object *object, u64 addr, u32 *data) { struct nv50_disp_chan *chan = nv50_disp_chan(object); struct nv50_disp *disp = chan->root->disp; struct nvkm_device *device = disp->base.engine.subdev.device; - *data = nvkm_rd32(device, 0x640000 + (chan->chid * 0x1000) + addr); + *data = nvkm_rd32(device, 0x640000 + (chan->chid.user * 0x1000) + addr); return 0; } -int +static int nv50_disp_chan_wr32(struct nvkm_object *object, u64 addr, u32 data) { struct nv50_disp_chan *chan = nv50_disp_chan(object); struct nv50_disp *disp = chan->root->disp; struct nvkm_device *device = disp->base.engine.subdev.device; - nvkm_wr32(device, 0x640000 + (chan->chid * 0x1000) + addr, data); + nvkm_wr32(device, 0x640000 + (chan->chid.user * 0x1000) + addr, data); return 0; } -int +static int nv50_disp_chan_ntfy(struct nvkm_object *object, u32 type, struct nvkm_event **pevent) { @@ -189,14 +189,14 @@ nv50_disp_chan_ntfy(struct nvkm_object *object, u32 type, return -EINVAL; } -int +static int nv50_disp_chan_map(struct nvkm_object *object, u64 *addr, u32 *size) { struct nv50_disp_chan *chan = nv50_disp_chan(object); struct nv50_disp *disp = chan->root->disp; struct nvkm_device *device = disp->base.engine.subdev.device; *addr = device->func->resource_addr(device, 0) + - 0x640000 + (chan->chid * 0x1000); + 0x640000 + (chan->chid.user * 0x1000); *size = 0x001000; return 0; } @@ -243,8 +243,8 @@ nv50_disp_chan_dtor(struct nvkm_object *object) { struct nv50_disp_chan *chan = nv50_disp_chan(object); struct nv50_disp *disp = chan->root->disp; - if (chan->chid >= 0) - disp->chan[chan->chid] = NULL; + if (chan->chid.user >= 0) + disp->chan[chan->chid.user] = NULL; return chan->func->dtor ? chan->func->dtor(chan) : chan; } @@ -263,7 +263,7 @@ nv50_disp_chan = { int nv50_disp_chan_ctor(const struct nv50_disp_chan_func *func, const struct nv50_disp_chan_mthd *mthd, - struct nv50_disp_root *root, int chid, int head, + struct nv50_disp_root *root, int ctrl, int user, int head, const struct nvkm_oclass *oclass, struct nv50_disp_chan *chan) { @@ -273,21 +273,22 @@ nv50_disp_chan_ctor(const struct nv50_disp_chan_func *func, chan->func = func; chan->mthd = mthd; chan->root = root; - chan->chid = chid; + chan->chid.ctrl = ctrl; + chan->chid.user = user; chan->head = head; - if (disp->chan[chan->chid]) { - chan->chid = -1; + if (disp->chan[chan->chid.user]) { + chan->chid.user = -1; return -EBUSY; } - disp->chan[chan->chid] = chan; + disp->chan[chan->chid.user] = chan; return 0; } int nv50_disp_chan_new_(const struct nv50_disp_chan_func *func, const struct nv50_disp_chan_mthd *mthd, - struct nv50_disp_root *root, int chid, int head, + struct nv50_disp_root *root, int ctrl, int user, int head, const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { @@ -297,5 +298,6 @@ nv50_disp_chan_new_(const struct nv50_disp_chan_func *func, return -ENOMEM; *pobject = &chan->object; - return nv50_disp_chan_ctor(func, mthd, root, chid, head, oclass, chan); + return nv50_disp_chan_ctor(func, mthd, root, ctrl, user, + head, oclass, chan); } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.h index f5f683d9fd20..737b38f6fbd2 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.h @@ -7,7 +7,11 @@ struct nv50_disp_chan { const struct nv50_disp_chan_func *func; const struct nv50_disp_chan_mthd *mthd; struct nv50_disp_root *root; - int chid; + + struct { + int ctrl; + int user; + } chid; int head; struct nvkm_object object; @@ -25,11 +29,11 @@ struct nv50_disp_chan_func { int nv50_disp_chan_ctor(const struct nv50_disp_chan_func *, const struct nv50_disp_chan_mthd *, - struct nv50_disp_root *, int chid, int head, + struct nv50_disp_root *, int ctrl, int user, int head, const struct nvkm_oclass *, struct nv50_disp_chan *); int nv50_disp_chan_new_(const struct nv50_disp_chan_func *, const struct nv50_disp_chan_mthd *, - struct nv50_disp_root *, int chid, int head, + struct nv50_disp_root *, int ctrl, int user, int head, const struct nvkm_oclass *, struct nvkm_object **); extern const struct nv50_disp_chan_func nv50_disp_pioc_func; @@ -90,13 +94,16 @@ extern const struct nv50_disp_chan_mthd gk104_disp_ovly_chan_mthd; struct nv50_disp_pioc_oclass { int (*ctor)(const struct nv50_disp_chan_func *, const struct nv50_disp_chan_mthd *, - struct nv50_disp_root *, int chid, + struct nv50_disp_root *, int ctrl, int user, const struct nvkm_oclass *, void *data, u32 size, struct nvkm_object **); struct nvkm_sclass base; const struct nv50_disp_chan_func *func; const struct nv50_disp_chan_mthd *mthd; - int chid; + struct { + int ctrl; + int user; + } chid; }; extern const struct nv50_disp_pioc_oclass nv50_disp_oimm_oclass; @@ -114,15 +121,17 @@ extern const struct nv50_disp_pioc_oclass gf119_disp_curs_oclass; extern const struct nv50_disp_pioc_oclass gk104_disp_oimm_oclass; extern const struct nv50_disp_pioc_oclass gk104_disp_curs_oclass; +extern const struct nv50_disp_pioc_oclass gp102_disp_oimm_oclass; +extern const struct nv50_disp_pioc_oclass gp102_disp_curs_oclass; int nv50_disp_curs_new(const struct nv50_disp_chan_func *, const struct nv50_disp_chan_mthd *, - struct nv50_disp_root *, int chid, + struct nv50_disp_root *, int ctrl, int user, const struct nvkm_oclass *, void *data, u32 size, struct nvkm_object **); int nv50_disp_oimm_new(const struct nv50_disp_chan_func *, const struct nv50_disp_chan_mthd *, - struct nv50_disp_root *, int chid, + struct nv50_disp_root *, int ctrl, int user, const struct nvkm_oclass *, void *data, u32 size, struct nvkm_object **); #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/coreg94.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/coreg94.c index 019379a3a01c..c65c9f3ff69f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/coreg94.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/coreg94.c @@ -26,7 +26,7 @@ #include <nvif/class.h> -const struct nv50_disp_mthd_list +static const struct nv50_disp_mthd_list g94_disp_core_mthd_sor = { .mthd = 0x0040, .addr = 0x000008, @@ -43,8 +43,8 @@ g94_disp_core_chan_mthd = { .prev = 0x000004, .data = { { "Global", 1, &nv50_disp_core_mthd_base }, - { "DAC", 3, &g84_disp_core_mthd_dac }, - { "SOR", 4, &g94_disp_core_mthd_sor }, + { "DAC", 3, &g84_disp_core_mthd_dac }, + { "SOR", 4, &g94_disp_core_mthd_sor }, { "PIOR", 3, &nv50_disp_core_mthd_pior }, { "HEAD", 2, &g84_disp_core_mthd_head }, {} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/coregp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/coregp102.c index 6922f4007b61..b0df4b752b8c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/coregp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/coregp102.c @@ -29,7 +29,7 @@ #include <nvif/class.h> static int -gp104_disp_core_init(struct nv50_disp_dmac *chan) +gp102_disp_core_init(struct nv50_disp_dmac *chan) { struct nv50_disp *disp = chan->base.root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; @@ -59,20 +59,20 @@ gp104_disp_core_init(struct nv50_disp_dmac *chan) return 0; } -const struct nv50_disp_dmac_func -gp104_disp_core_func = { - .init = gp104_disp_core_init, +static const struct nv50_disp_dmac_func +gp102_disp_core_func = { + .init = gp102_disp_core_init, .fini = gf119_disp_core_fini, .bind = gf119_disp_dmac_bind, }; const struct nv50_disp_dmac_oclass -gp104_disp_core_oclass = { - .base.oclass = GP104_DISP_CORE_CHANNEL_DMA, +gp102_disp_core_oclass = { + .base.oclass = GP102_DISP_CORE_CHANNEL_DMA, .base.minver = 0, .base.maxver = 0, .ctor = nv50_disp_core_new, - .func = &gp104_disp_core_func, + .func = &gp102_disp_core_func, .mthd = &gk104_disp_core_chan_mthd, .chid = 0, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursg84.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursg84.c index dd99fc7060b1..fa781b5a7e07 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursg84.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursg84.c @@ -33,5 +33,5 @@ g84_disp_curs_oclass = { .base.maxver = 0, .ctor = nv50_disp_curs_new, .func = &nv50_disp_pioc_func, - .chid = 7, + .chid = { 7, 7 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgf119.c index 2a1574e06ad6..2be6fb052c65 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgf119.c @@ -33,5 +33,5 @@ gf119_disp_curs_oclass = { .base.maxver = 0, .ctor = nv50_disp_curs_new, .func = &gf119_disp_pioc_func, - .chid = 13, + .chid = { 13, 13 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgk104.c index 28e8f06c9472..2a99db4bf8f8 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgk104.c @@ -33,5 +33,5 @@ gk104_disp_curs_oclass = { .base.maxver = 0, .ctor = nv50_disp_curs_new, .func = &gf119_disp_pioc_func, - .chid = 13, + .chid = { 13, 13 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgp102.c new file mode 100644 index 000000000000..e958210d8105 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgp102.c @@ -0,0 +1,37 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs <bskeggs@redhat.com> + */ +#include "channv50.h" +#include "rootnv50.h" + +#include <nvif/class.h> + +const struct nv50_disp_pioc_oclass +gp102_disp_curs_oclass = { + .base.oclass = GK104_DISP_CURSOR, + .base.minver = 0, + .base.maxver = 0, + .ctor = nv50_disp_curs_new, + .func = &gf119_disp_pioc_func, + .chid = { 13, 17 }, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgt215.c index d8a4b9ca139c..00a7f3564450 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursgt215.c @@ -33,5 +33,5 @@ gt215_disp_curs_oclass = { .base.maxver = 0, .ctor = nv50_disp_curs_new, .func = &nv50_disp_pioc_func, - .chid = 7, + .chid = { 7, 7 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursnv50.c index 8b1320499a0f..82ff82d8c1ab 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/cursnv50.c @@ -33,7 +33,7 @@ int nv50_disp_curs_new(const struct nv50_disp_chan_func *func, const struct nv50_disp_chan_mthd *mthd, - struct nv50_disp_root *root, int chid, + struct nv50_disp_root *root, int ctrl, int user, const struct nvkm_oclass *oclass, void *data, u32 size, struct nvkm_object **pobject) { @@ -54,7 +54,7 @@ nv50_disp_curs_new(const struct nv50_disp_chan_func *func, } else return ret; - return nv50_disp_chan_new_(func, mthd, root, chid + head, + return nv50_disp_chan_new_(func, mthd, root, ctrl + head, user + head, head, oclass, pobject); } @@ -65,5 +65,5 @@ nv50_disp_curs_oclass = { .base.maxver = 0, .ctor = nv50_disp_curs_new, .func = &nv50_disp_pioc_func, - .chid = 7, + .chid = { 7, 7 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgf119.c index a57f7cef307a..ce7cd74fbd5d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgf119.c @@ -32,8 +32,8 @@ gf119_disp_dmac_bind(struct nv50_disp_dmac *chan, struct nvkm_object *object, u32 handle) { return nvkm_ramht_insert(chan->base.root->ramht, object, - chan->base.chid, -9, handle, - chan->base.chid << 27 | 0x00000001); + chan->base.chid.user, -9, handle, + chan->base.chid.user << 27 | 0x00000001); } void @@ -42,22 +42,23 @@ gf119_disp_dmac_fini(struct nv50_disp_dmac *chan) struct nv50_disp *disp = chan->base.root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->base.chid; + int ctrl = chan->base.chid.ctrl; + int user = chan->base.chid.user; /* deactivate channel */ - nvkm_mask(device, 0x610490 + (chid * 0x0010), 0x00001010, 0x00001000); - nvkm_mask(device, 0x610490 + (chid * 0x0010), 0x00000003, 0x00000000); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00001010, 0x00001000); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00000003, 0x00000000); if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610490 + (chid * 0x10)) & 0x001e0000)) + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x001e0000)) break; ) < 0) { - nvkm_error(subdev, "ch %d fini: %08x\n", chid, - nvkm_rd32(device, 0x610490 + (chid * 0x10))); + nvkm_error(subdev, "ch %d fini: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); } /* disable error reporting and completion notification */ - nvkm_mask(device, 0x610090, 0x00000001 << chid, 0x00000000); - nvkm_mask(device, 0x6100a0, 0x00000001 << chid, 0x00000000); + nvkm_mask(device, 0x610090, 0x00000001 << user, 0x00000000); + nvkm_mask(device, 0x6100a0, 0x00000001 << user, 0x00000000); } static int @@ -66,26 +67,27 @@ gf119_disp_dmac_init(struct nv50_disp_dmac *chan) struct nv50_disp *disp = chan->base.root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->base.chid; + int ctrl = chan->base.chid.ctrl; + int user = chan->base.chid.user; /* enable error reporting */ - nvkm_mask(device, 0x6100a0, 0x00000001 << chid, 0x00000001 << chid); + nvkm_mask(device, 0x6100a0, 0x00000001 << user, 0x00000001 << user); /* initialise channel for dma command submission */ - nvkm_wr32(device, 0x610494 + (chid * 0x0010), chan->push); - nvkm_wr32(device, 0x610498 + (chid * 0x0010), 0x00010000); - nvkm_wr32(device, 0x61049c + (chid * 0x0010), 0x00000001); - nvkm_mask(device, 0x610490 + (chid * 0x0010), 0x00000010, 0x00000010); - nvkm_wr32(device, 0x640000 + (chid * 0x1000), 0x00000000); - nvkm_wr32(device, 0x610490 + (chid * 0x0010), 0x00000013); + nvkm_wr32(device, 0x610494 + (ctrl * 0x0010), chan->push); + nvkm_wr32(device, 0x610498 + (ctrl * 0x0010), 0x00010000); + nvkm_wr32(device, 0x61049c + (ctrl * 0x0010), 0x00000001); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000 + (ctrl * 0x1000), 0x00000000); + nvkm_wr32(device, 0x610490 + (ctrl * 0x0010), 0x00000013); /* wait for it to go inactive */ if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610490 + (chid * 0x10)) & 0x80000000)) + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x80000000)) break; ) < 0) { - nvkm_error(subdev, "ch %d init: %08x\n", chid, - nvkm_rd32(device, 0x610490 + (chid * 0x10))); + nvkm_error(subdev, "ch %d init: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); return -EBUSY; } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgp102.c index ad24c2c57696..cdead9500343 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgp102.c @@ -27,31 +27,32 @@ #include <subdev/timer.h> static int -gp104_disp_dmac_init(struct nv50_disp_dmac *chan) +gp102_disp_dmac_init(struct nv50_disp_dmac *chan) { struct nv50_disp *disp = chan->base.root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->base.chid; + int ctrl = chan->base.chid.ctrl; + int user = chan->base.chid.user; /* enable error reporting */ - nvkm_mask(device, 0x6100a0, 0x00000001 << chid, 0x00000001 << chid); + nvkm_mask(device, 0x6100a0, 0x00000001 << user, 0x00000001 << user); /* initialise channel for dma command submission */ - nvkm_wr32(device, 0x611494 + (chid * 0x0010), chan->push); - nvkm_wr32(device, 0x611498 + (chid * 0x0010), 0x00010000); - nvkm_wr32(device, 0x61149c + (chid * 0x0010), 0x00000001); - nvkm_mask(device, 0x610490 + (chid * 0x0010), 0x00000010, 0x00000010); - nvkm_wr32(device, 0x640000 + (chid * 0x1000), 0x00000000); - nvkm_wr32(device, 0x610490 + (chid * 0x0010), 0x00000013); + nvkm_wr32(device, 0x611494 + (ctrl * 0x0010), chan->push); + nvkm_wr32(device, 0x611498 + (ctrl * 0x0010), 0x00010000); + nvkm_wr32(device, 0x61149c + (ctrl * 0x0010), 0x00000001); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000 + (ctrl * 0x1000), 0x00000000); + nvkm_wr32(device, 0x610490 + (ctrl * 0x0010), 0x00000013); /* wait for it to go inactive */ if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610490 + (chid * 0x10)) & 0x80000000)) + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x80000000)) break; ) < 0) { - nvkm_error(subdev, "ch %d init: %08x\n", chid, - nvkm_rd32(device, 0x610490 + (chid * 0x10))); + nvkm_error(subdev, "ch %d init: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); return -EBUSY; } @@ -59,8 +60,8 @@ gp104_disp_dmac_init(struct nv50_disp_dmac *chan) } const struct nv50_disp_dmac_func -gp104_disp_dmac_func = { - .init = gp104_disp_dmac_init, +gp102_disp_dmac_func = { + .init = gp102_disp_dmac_init, .fini = gf119_disp_dmac_fini, .bind = gf119_disp_dmac_bind, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.c index 9c6645a357b9..070ec5e18fdb 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.c @@ -137,7 +137,6 @@ nv50_disp_dmac_new_(const struct nv50_disp_dmac_func *func, const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { - struct nvkm_device *device = root->disp->base.engine.subdev.device; struct nvkm_client *client = oclass->client; struct nvkm_dmaobj *dmaobj; struct nv50_disp_dmac *chan; @@ -149,13 +148,13 @@ nv50_disp_dmac_new_(const struct nv50_disp_dmac_func *func, chan->func = func; ret = nv50_disp_chan_ctor(&nv50_disp_dmac_func_, mthd, root, - chid, head, oclass, &chan->base); + chid, chid, head, oclass, &chan->base); if (ret) return ret; - dmaobj = nvkm_dma_search(device->dma, client, push); - if (!dmaobj) - return -ENOENT; + dmaobj = nvkm_dmaobj_search(client, push); + if (IS_ERR(dmaobj)) + return PTR_ERR(dmaobj); if (dmaobj->limit - dmaobj->start != 0xfff) return -EINVAL; @@ -179,9 +178,9 @@ nv50_disp_dmac_bind(struct nv50_disp_dmac *chan, struct nvkm_object *object, u32 handle) { return nvkm_ramht_insert(chan->base.root->ramht, object, - chan->base.chid, -10, handle, - chan->base.chid << 28 | - chan->base.chid); + chan->base.chid.user, -10, handle, + chan->base.chid.user << 28 | + chan->base.chid.user); } static void @@ -190,21 +189,22 @@ nv50_disp_dmac_fini(struct nv50_disp_dmac *chan) struct nv50_disp *disp = chan->base.root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->base.chid; + int ctrl = chan->base.chid.ctrl; + int user = chan->base.chid.user; /* deactivate channel */ - nvkm_mask(device, 0x610200 + (chid * 0x0010), 0x00001010, 0x00001000); - nvkm_mask(device, 0x610200 + (chid * 0x0010), 0x00000003, 0x00000000); + nvkm_mask(device, 0x610200 + (ctrl * 0x0010), 0x00001010, 0x00001000); + nvkm_mask(device, 0x610200 + (ctrl * 0x0010), 0x00000003, 0x00000000); if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610200 + (chid * 0x10)) & 0x001e0000)) + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x001e0000)) break; ) < 0) { - nvkm_error(subdev, "ch %d fini timeout, %08x\n", chid, - nvkm_rd32(device, 0x610200 + (chid * 0x10))); + nvkm_error(subdev, "ch %d fini timeout, %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); } /* disable error reporting and completion notifications */ - nvkm_mask(device, 0x610028, 0x00010001 << chid, 0x00000000 << chid); + nvkm_mask(device, 0x610028, 0x00010001 << user, 0x00000000 << user); } static int @@ -213,26 +213,27 @@ nv50_disp_dmac_init(struct nv50_disp_dmac *chan) struct nv50_disp *disp = chan->base.root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->base.chid; + int ctrl = chan->base.chid.ctrl; + int user = chan->base.chid.user; /* enable error reporting */ - nvkm_mask(device, 0x610028, 0x00010000 << chid, 0x00010000 << chid); + nvkm_mask(device, 0x610028, 0x00010000 << user, 0x00010000 << user); /* initialise channel for dma command submission */ - nvkm_wr32(device, 0x610204 + (chid * 0x0010), chan->push); - nvkm_wr32(device, 0x610208 + (chid * 0x0010), 0x00010000); - nvkm_wr32(device, 0x61020c + (chid * 0x0010), chid); - nvkm_mask(device, 0x610200 + (chid * 0x0010), 0x00000010, 0x00000010); - nvkm_wr32(device, 0x640000 + (chid * 0x1000), 0x00000000); - nvkm_wr32(device, 0x610200 + (chid * 0x0010), 0x00000013); + nvkm_wr32(device, 0x610204 + (ctrl * 0x0010), chan->push); + nvkm_wr32(device, 0x610208 + (ctrl * 0x0010), 0x00010000); + nvkm_wr32(device, 0x61020c + (ctrl * 0x0010), ctrl); + nvkm_mask(device, 0x610200 + (ctrl * 0x0010), 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000 + (ctrl * 0x1000), 0x00000000); + nvkm_wr32(device, 0x610200 + (ctrl * 0x0010), 0x00000013); /* wait for it to go inactive */ if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610200 + (chid * 0x10)) & 0x80000000)) + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x80000000)) break; ) < 0) { - nvkm_error(subdev, "ch %d init timeout, %08x\n", chid, - nvkm_rd32(device, 0x610200 + (chid * 0x10))); + nvkm_error(subdev, "ch %d init timeout, %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); return -EBUSY; } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.h index 43ac05857853..ea4a0d062e31 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacnv50.h @@ -30,7 +30,7 @@ int gf119_disp_dmac_bind(struct nv50_disp_dmac *, struct nvkm_object *, u32); extern const struct nv50_disp_dmac_func gf119_disp_core_func; void gf119_disp_core_fini(struct nv50_disp_dmac *); -extern const struct nv50_disp_dmac_func gp104_disp_dmac_func; +extern const struct nv50_disp_dmac_func gp102_disp_dmac_func; struct nv50_disp_dmac_oclass { int (*ctor)(const struct nv50_disp_dmac_func *, @@ -95,7 +95,7 @@ extern const struct nv50_disp_dmac_oclass gm200_disp_core_oclass; extern const struct nv50_disp_dmac_oclass gp100_disp_core_oclass; -extern const struct nv50_disp_dmac_oclass gp104_disp_core_oclass; -extern const struct nv50_disp_dmac_oclass gp104_disp_base_oclass; -extern const struct nv50_disp_dmac_oclass gp104_disp_ovly_oclass; +extern const struct nv50_disp_dmac_oclass gp102_disp_core_oclass; +extern const struct nv50_disp_dmac_oclass gp102_disp_base_oclass; +extern const struct nv50_disp_dmac_oclass gp102_disp_ovly_oclass; #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.c index 9688970eca47..4a93ceb850ac 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.c @@ -319,9 +319,8 @@ static const struct dp_rates { }; void -nvkm_dp_train(struct work_struct *w) +nvkm_dp_train(struct nvkm_output_dp *outp) { - struct nvkm_output_dp *outp = container_of(w, typeof(*outp), lt.work); struct nv50_disp *disp = nv50_disp(outp->base.disp); const struct dp_rates *cfg = nvkm_dp_rates; struct dp_state _dp = { @@ -353,9 +352,6 @@ nvkm_dp_train(struct work_struct *w) } cfg--; - /* disable link interrupt handling during link training */ - nvkm_notify_put(&outp->irq); - /* ensure sink is not in a low-power state */ if (!nvkm_rdaux(outp->aux, DPCD_SC00, &pwr, 1)) { if ((pwr & DPCD_SC00_SET_POWER) != DPCD_SC00_SET_POWER_D0) { @@ -400,9 +396,6 @@ nvkm_dp_train(struct work_struct *w) dp_link_train_fini(dp); - /* signal completion and enable link interrupt handling */ OUTP_DBG(&outp->base, "training complete"); atomic_set(&outp->lt.done, 1); - wake_up(&outp->lt.wait); - nvkm_notify_get(&outp->irq); } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.h index 6e10c5e0ef11..baf1dd9ff975 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dport.h @@ -1,6 +1,6 @@ #ifndef __NVKM_DISP_DPORT_H__ #define __NVKM_DISP_DPORT_H__ -#include <core/os.h> +struct nvkm_output_dp; /* DPCD Receiver Capabilities */ #define DPCD_RC00_DPCD_REV 0x00000 @@ -77,5 +77,5 @@ #define DPCD_SC00_SET_POWER_D0 0x01 #define DPCD_SC00_SET_POWER_D3 0x03 -void nvkm_dp_train(struct work_struct *); +void nvkm_dp_train(struct nvkm_output_dp *); #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c index 29e84b241cca..7b346ccc38b7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c @@ -203,17 +203,20 @@ gf119_disp_intr_unk2_0(struct nv50_disp *disp, int head) /* see note in nv50_disp_intr_unk20_0() */ if (outp && outp->info.type == DCB_OUTPUT_DP) { struct nvkm_output_dp *outpdp = nvkm_output_dp(outp); - struct nvbios_init init = { - .subdev = subdev, - .bios = subdev->device->bios, - .outp = &outp->info, - .crtc = head, - .offset = outpdp->info.script[4], - .execute = 1, - }; + if (!outpdp->lt.mst) { + struct nvbios_init init = { + .subdev = subdev, + .bios = subdev->device->bios, + .outp = &outp->info, + .crtc = head, + .offset = outpdp->info.script[4], + .execute = 1, + }; - nvbios_exec(&init); - atomic_set(&outpdp->lt.done, 0); + nvkm_notify_put(&outpdp->irq); + nvbios_exec(&init); + atomic_set(&outpdp->lt.done, 0); + } } } @@ -314,7 +317,7 @@ gf119_disp_intr_unk2_2(struct nv50_disp *disp, int head) break; } - if (nvkm_output_dp_train(outp, pclk, true)) + if (nvkm_output_dp_train(outp, pclk)) OUTP_ERR(outp, "link not trained before attach"); } else { if (disp->func->sor.magic) diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c index 3bf3380336e4..f5d613f82709 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c @@ -25,7 +25,7 @@ #include "rootnv50.h" static void -gp104_disp_intr_error(struct nv50_disp *disp, int chid) +gp102_disp_intr_error(struct nv50_disp *disp, int chid) { struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; @@ -51,12 +51,12 @@ gp104_disp_intr_error(struct nv50_disp *disp, int chid) } static const struct nv50_disp_func -gp104_disp = { +gp102_disp = { .intr = gf119_disp_intr, - .intr_error = gp104_disp_intr_error, + .intr_error = gp102_disp_intr_error, .uevent = &gf119_disp_chan_uevent, .super = gf119_disp_intr_supervisor, - .root = &gp104_disp_root_oclass, + .root = &gp102_disp_root_oclass, .head.vblank_init = gf119_disp_vblank_init, .head.vblank_fini = gf119_disp_vblank_fini, .head.scanoutpos = gf119_disp_root_scanoutpos, @@ -75,7 +75,7 @@ gp104_disp = { }; int -gp104_disp_new(struct nvkm_device *device, int index, struct nvkm_disp **pdisp) +gp102_disp_new(struct nvkm_device *device, int index, struct nvkm_disp **pdisp) { - return gf119_disp_new_(&gp104_disp, device, index, pdisp); + return gf119_disp_new_(&gp102_disp, device, index, pdisp); } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdagt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdagt215.c index 6f0436df0219..f8f2f16c22a2 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdagt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdagt215.c @@ -59,7 +59,7 @@ gt215_hda_eld(NV50_DISP_MTHD_V1) ); } for (i = 0; i < size; i++) - nvkm_wr32(device, 0x61c440 + soff, (i << 8) | args->v0.data[0]); + nvkm_wr32(device, 0x61c440 + soff, (i << 8) | args->v0.data[i]); for (; i < 0x60; i++) nvkm_wr32(device, 0x61c440 + soff, (i << 8)); nvkm_mask(device, 0x61c448 + soff, 0x80000003, 0x80000003); diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c index fbb8c7dc18fd..0db8efbf1c2e 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c @@ -433,8 +433,6 @@ nv50_disp_dptmds_war(struct nvkm_device *device) case 0x94: case 0x96: case 0x98: - case 0xaa: - case 0xac: return true; default: break; @@ -590,6 +588,7 @@ nv50_disp_intr_unk20_0(struct nv50_disp *disp, int head) .execute = 1, }; + nvkm_notify_put(&outpdp->irq); nvbios_exec(&init); atomic_set(&outpdp->lt.done, 0); } @@ -779,7 +778,7 @@ nv50_disp_intr_unk20_2(struct nv50_disp *disp, int head) break; } - if (nvkm_output_dp_train(outp, datarate / soff, true)) + if (nvkm_output_dp_train(outp, datarate / soff)) OUTP_ERR(outp, "link not trained before attach"); } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmg84.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmg84.c index 54a4ae8d66c6..5ad5d0f5db05 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmg84.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmg84.c @@ -33,5 +33,5 @@ g84_disp_oimm_oclass = { .base.maxver = 0, .ctor = nv50_disp_oimm_new, .func = &nv50_disp_pioc_func, - .chid = 5, + .chid = { 5, 5 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgf119.c index c658db54afc5..1f9fd3403f07 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgf119.c @@ -33,5 +33,5 @@ gf119_disp_oimm_oclass = { .base.maxver = 0, .ctor = nv50_disp_oimm_new, .func = &gf119_disp_pioc_func, - .chid = 9, + .chid = { 9, 9 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgk104.c index b1fde8c125d6..0c09fe85e952 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgk104.c @@ -33,5 +33,5 @@ gk104_disp_oimm_oclass = { .base.maxver = 0, .ctor = nv50_disp_oimm_new, .func = &gf119_disp_pioc_func, - .chid = 9, + .chid = { 9, 9 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgp102.c new file mode 100644 index 000000000000..abf82365c671 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgp102.c @@ -0,0 +1,37 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs <bskeggs@redhat.com> + */ +#include "channv50.h" +#include "rootnv50.h" + +#include <nvif/class.h> + +const struct nv50_disp_pioc_oclass +gp102_disp_oimm_oclass = { + .base.oclass = GK104_DISP_OVERLAY, + .base.minver = 0, + .base.maxver = 0, + .ctor = nv50_disp_oimm_new, + .func = &gf119_disp_pioc_func, + .chid = { 9, 13 }, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgt215.c index f4e7eb3d1177..1281db28aebd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmgt215.c @@ -33,5 +33,5 @@ gt215_disp_oimm_oclass = { .base.maxver = 0, .ctor = nv50_disp_oimm_new, .func = &nv50_disp_pioc_func, - .chid = 5, + .chid = { 5, 5 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmnv50.c index 3940b9c966ec..07540f3d32dc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/oimmnv50.c @@ -33,7 +33,7 @@ int nv50_disp_oimm_new(const struct nv50_disp_chan_func *func, const struct nv50_disp_chan_mthd *mthd, - struct nv50_disp_root *root, int chid, + struct nv50_disp_root *root, int ctrl, int user, const struct nvkm_oclass *oclass, void *data, u32 size, struct nvkm_object **pobject) { @@ -54,7 +54,7 @@ nv50_disp_oimm_new(const struct nv50_disp_chan_func *func, } else return ret; - return nv50_disp_chan_new_(func, mthd, root, chid + head, + return nv50_disp_chan_new_(func, mthd, root, ctrl + head, user + head, head, oclass, pobject); } @@ -65,5 +65,5 @@ nv50_disp_oimm_oclass = { .base.maxver = 0, .ctor = nv50_disp_oimm_new, .func = &nv50_disp_pioc_func, - .chid = 5, + .chid = { 5, 5 }, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.c index 3b7a9e7a1ea8..de36f73b14dc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.c @@ -31,7 +31,7 @@ #include <nvif/event.h> int -nvkm_output_dp_train(struct nvkm_output *base, u32 datarate, bool wait) +nvkm_output_dp_train(struct nvkm_output *base, u32 datarate) { struct nvkm_output_dp *outp = nvkm_output_dp(base); bool retrain = true; @@ -39,6 +39,8 @@ nvkm_output_dp_train(struct nvkm_output *base, u32 datarate, bool wait) u32 linkrate; int ret, i; + mutex_lock(&outp->mutex); + /* check that the link is trained at a high enough rate */ ret = nvkm_rdaux(outp->aux, DPCD_LC00_LINK_BW_SET, link, 2); if (ret) { @@ -88,19 +90,10 @@ done: outp->dpcd[DPCD_RC02] = outp->base.info.dpconf.link_nr; } - atomic_set(&outp->lt.done, 0); - schedule_work(&outp->lt.work); - } else { - nvkm_notify_get(&outp->irq); - } - - if (wait) { - if (!wait_event_timeout(outp->lt.wait, - atomic_read(&outp->lt.done), - msecs_to_jiffies(2000))) - ret = -ETIMEDOUT; + nvkm_dp_train(outp); } + mutex_unlock(&outp->mutex); return ret; } @@ -118,7 +111,7 @@ nvkm_output_dp_enable(struct nvkm_output_dp *outp, bool enable) if (!nvkm_rdaux(aux, DPCD_RC00_DPCD_REV, outp->dpcd, sizeof(outp->dpcd))) { - nvkm_output_dp_train(&outp->base, 0, true); + nvkm_output_dp_train(&outp->base, 0); return; } } @@ -165,10 +158,10 @@ nvkm_output_dp_irq(struct nvkm_notify *notify) }; OUTP_DBG(&outp->base, "IRQ: %d", line->mask); - nvkm_output_dp_train(&outp->base, 0, true); + nvkm_output_dp_train(&outp->base, 0); nvkm_event_send(&disp->hpd, rep.mask, conn->index, &rep, sizeof(rep)); - return NVKM_NOTIFY_DROP; + return NVKM_NOTIFY_KEEP; } static void @@ -177,7 +170,6 @@ nvkm_output_dp_fini(struct nvkm_output *base) struct nvkm_output_dp *outp = nvkm_output_dp(base); nvkm_notify_put(&outp->hpd); nvkm_notify_put(&outp->irq); - flush_work(&outp->lt.work); nvkm_output_dp_enable(outp, false); } @@ -187,6 +179,7 @@ nvkm_output_dp_init(struct nvkm_output *base) struct nvkm_output_dp *outp = nvkm_output_dp(base); nvkm_notify_put(&outp->base.conn->hpd); nvkm_output_dp_enable(outp, true); + nvkm_notify_get(&outp->irq); nvkm_notify_get(&outp->hpd); } @@ -238,11 +231,6 @@ nvkm_output_dp_ctor(const struct nvkm_output_dp_func *func, OUTP_DBG(&outp->base, "bios dp %02x %02x %02x %02x", outp->version, hdr, cnt, len); - /* link training */ - INIT_WORK(&outp->lt.work, nvkm_dp_train); - init_waitqueue_head(&outp->lt.wait); - atomic_set(&outp->lt.done, 0); - /* link maintenance */ ret = nvkm_notify_init(NULL, &i2c->event, nvkm_output_dp_irq, true, &(struct nvkm_i2c_ntfy_req) { @@ -257,6 +245,9 @@ nvkm_output_dp_ctor(const struct nvkm_output_dp_func *func, return ret; } + mutex_init(&outp->mutex); + atomic_set(&outp->lt.done, 0); + /* hotplug detect, replaces gpio-based mechanism with aux events */ ret = nvkm_notify_init(NULL, &i2c->event, nvkm_output_dp_hpd, true, &(struct nvkm_i2c_ntfy_req) { diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.h index 4e983f6d7032..3c83a561cd88 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outpdp.h @@ -29,10 +29,10 @@ struct nvkm_output_dp { bool present; u8 dpcd[16]; + struct mutex mutex; struct { - struct work_struct work; - wait_queue_head_t wait; atomic_t done; + bool mst; } lt; }; @@ -41,9 +41,11 @@ struct nvkm_output_dp_func { int (*lnk_pwr)(struct nvkm_output_dp *, int nr); int (*lnk_ctl)(struct nvkm_output_dp *, int nr, int bw, bool ef); int (*drv_ctl)(struct nvkm_output_dp *, int ln, int vs, int pe, int pc); + void (*vcpi)(struct nvkm_output_dp *, int head, u8 start_slot, + u8 num_slots, u16 pbn, u16 aligned_pbn); }; -int nvkm_output_dp_train(struct nvkm_output *, u32 rate, bool wait); +int nvkm_output_dp_train(struct nvkm_output *, u32 rate); int nvkm_output_dp_ctor(const struct nvkm_output_dp_func *, struct nvkm_disp *, int index, struct dcb_output *, struct nvkm_i2c_aux *, @@ -63,6 +65,7 @@ int gf119_sor_dp_new(struct nvkm_disp *, int, struct dcb_output *, struct nvkm_output **); int gf119_sor_dp_lnk_ctl(struct nvkm_output_dp *, int, int, bool); int gf119_sor_dp_drv_ctl(struct nvkm_output_dp *, int, int, int, int); +void gf119_sor_dp_vcpi(struct nvkm_output_dp *, int, u8, u8, u16, u16); int gm107_sor_dp_new(struct nvkm_disp *, int, struct dcb_output *, struct nvkm_output **); diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/ovlygp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ovlygp102.c index 97e2dd2d908e..589bd2f12b41 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/ovlygp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ovlygp102.c @@ -27,12 +27,12 @@ #include <nvif/class.h> const struct nv50_disp_dmac_oclass -gp104_disp_ovly_oclass = { +gp102_disp_ovly_oclass = { .base.oclass = GK104_DISP_OVERLAY_CONTROL_DMA, .base.minver = 0, .base.maxver = 0, .ctor = nv50_disp_ovly_new, - .func = &gp104_disp_dmac_func, + .func = &gp102_disp_dmac_func, .mthd = &gk104_disp_ovly_chan_mthd, .chid = 5, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocgf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocgf119.c index a625a9876e34..0abaa6431943 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocgf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocgf119.c @@ -32,20 +32,21 @@ gf119_disp_pioc_fini(struct nv50_disp_chan *chan) struct nv50_disp *disp = chan->root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->chid; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; - nvkm_mask(device, 0x610490 + (chid * 0x10), 0x00000001, 0x00000000); + nvkm_mask(device, 0x610490 + (ctrl * 0x10), 0x00000001, 0x00000000); if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610490 + (chid * 0x10)) & 0x00030000)) + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x00030000)) break; ) < 0) { - nvkm_error(subdev, "ch %d fini: %08x\n", chid, - nvkm_rd32(device, 0x610490 + (chid * 0x10))); + nvkm_error(subdev, "ch %d fini: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); } /* disable error reporting and completion notification */ - nvkm_mask(device, 0x610090, 0x00000001 << chid, 0x00000000); - nvkm_mask(device, 0x6100a0, 0x00000001 << chid, 0x00000000); + nvkm_mask(device, 0x610090, 0x00000001 << user, 0x00000000); + nvkm_mask(device, 0x6100a0, 0x00000001 << user, 0x00000000); } static int @@ -54,20 +55,21 @@ gf119_disp_pioc_init(struct nv50_disp_chan *chan) struct nv50_disp *disp = chan->root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->chid; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; /* enable error reporting */ - nvkm_mask(device, 0x6100a0, 0x00000001 << chid, 0x00000001 << chid); + nvkm_mask(device, 0x6100a0, 0x00000001 << user, 0x00000001 << user); /* activate channel */ - nvkm_wr32(device, 0x610490 + (chid * 0x10), 0x00000001); + nvkm_wr32(device, 0x610490 + (ctrl * 0x10), 0x00000001); if (nvkm_msec(device, 2000, - u32 tmp = nvkm_rd32(device, 0x610490 + (chid * 0x10)); + u32 tmp = nvkm_rd32(device, 0x610490 + (ctrl * 0x10)); if ((tmp & 0x00030000) == 0x00010000) break; ) < 0) { - nvkm_error(subdev, "ch %d init: %08x\n", chid, - nvkm_rd32(device, 0x610490 + (chid * 0x10))); + nvkm_error(subdev, "ch %d init: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); return -EBUSY; } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocnv50.c index 9d2618dacf20..0211e0e8a35f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/piocnv50.c @@ -32,15 +32,16 @@ nv50_disp_pioc_fini(struct nv50_disp_chan *chan) struct nv50_disp *disp = chan->root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->chid; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; - nvkm_mask(device, 0x610200 + (chid * 0x10), 0x00000001, 0x00000000); + nvkm_mask(device, 0x610200 + (ctrl * 0x10), 0x00000001, 0x00000000); if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610200 + (chid * 0x10)) & 0x00030000)) + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x00030000)) break; ) < 0) { - nvkm_error(subdev, "ch %d timeout: %08x\n", chid, - nvkm_rd32(device, 0x610200 + (chid * 0x10))); + nvkm_error(subdev, "ch %d timeout: %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); } } @@ -50,26 +51,27 @@ nv50_disp_pioc_init(struct nv50_disp_chan *chan) struct nv50_disp *disp = chan->root->disp; struct nvkm_subdev *subdev = &disp->base.engine.subdev; struct nvkm_device *device = subdev->device; - int chid = chan->chid; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; - nvkm_wr32(device, 0x610200 + (chid * 0x10), 0x00002000); + nvkm_wr32(device, 0x610200 + (ctrl * 0x10), 0x00002000); if (nvkm_msec(device, 2000, - if (!(nvkm_rd32(device, 0x610200 + (chid * 0x10)) & 0x00030000)) + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x00030000)) break; ) < 0) { - nvkm_error(subdev, "ch %d timeout0: %08x\n", chid, - nvkm_rd32(device, 0x610200 + (chid * 0x10))); + nvkm_error(subdev, "ch %d timeout0: %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); return -EBUSY; } - nvkm_wr32(device, 0x610200 + (chid * 0x10), 0x00000001); + nvkm_wr32(device, 0x610200 + (ctrl * 0x10), 0x00000001); if (nvkm_msec(device, 2000, - u32 tmp = nvkm_rd32(device, 0x610200 + (chid * 0x10)); + u32 tmp = nvkm_rd32(device, 0x610200 + (ctrl * 0x10)); if ((tmp & 0x00030000) == 0x00010000) break; ) < 0) { - nvkm_error(subdev, "ch %d timeout1: %08x\n", chid, - nvkm_rd32(device, 0x610200 + (chid * 0x10))); + nvkm_error(subdev, "ch %d timeout1: %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); return -EBUSY; } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootgp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootgp102.c index 8443e04dc626..37122ca579ad 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootgp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootgp102.c @@ -27,32 +27,32 @@ #include <nvif/class.h> static const struct nv50_disp_root_func -gp104_disp_root = { +gp102_disp_root = { .init = gf119_disp_root_init, .fini = gf119_disp_root_fini, .dmac = { - &gp104_disp_core_oclass, - &gp104_disp_base_oclass, - &gp104_disp_ovly_oclass, + &gp102_disp_core_oclass, + &gp102_disp_base_oclass, + &gp102_disp_ovly_oclass, }, .pioc = { - &gk104_disp_oimm_oclass, - &gk104_disp_curs_oclass, + &gp102_disp_oimm_oclass, + &gp102_disp_curs_oclass, }, }; static int -gp104_disp_root_new(struct nvkm_disp *disp, const struct nvkm_oclass *oclass, +gp102_disp_root_new(struct nvkm_disp *disp, const struct nvkm_oclass *oclass, void *data, u32 size, struct nvkm_object **pobject) { - return nv50_disp_root_new_(&gp104_disp_root, disp, oclass, + return nv50_disp_root_new_(&gp102_disp_root, disp, oclass, data, size, pobject); } const struct nvkm_disp_oclass -gp104_disp_root_oclass = { - .base.oclass = GP104_DISP, +gp102_disp_root_oclass = { + .base.oclass = GP102_DISP, .base.minver = -1, .base.maxver = -1, - .ctor = gp104_disp_root_new, + .ctor = gp102_disp_root_new, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.c index 2f9cecd81d04..e70dc6a9ff7d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.c @@ -66,7 +66,7 @@ nv50_disp_root_scanoutpos(NV50_DISP_MTHD_V0) return 0; } -int +static int nv50_disp_root_mthd_(struct nvkm_object *object, u32 mthd, void *data, u32 size) { union { @@ -173,13 +173,56 @@ nv50_disp_root_mthd_(struct nvkm_object *object, u32 mthd, void *data, u32 size) return 0; } else if (args->v0.state != 0) { - nvkm_output_dp_train(&outpdp->base, 0, true); + nvkm_output_dp_train(&outpdp->base, 0); return 0; } } else return ret; } break; + case NV50_DISP_MTHD_V1_SOR_DP_MST_LINK: { + struct nvkm_output_dp *outpdp = nvkm_output_dp(outp); + union { + struct nv50_disp_sor_dp_mst_link_v0 v0; + } *args = data; + int ret = -ENOSYS; + nvif_ioctl(object, "disp sor dp mst link size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, "disp sor dp mst link vers %d state %d\n", + args->v0.version, args->v0.state); + if (outpdp->lt.mst != !!args->v0.state) { + outpdp->lt.mst = !!args->v0.state; + atomic_set(&outpdp->lt.done, 0); + nvkm_output_dp_train(&outpdp->base, 0); + } + return 0; + } else + return ret; + } + break; + case NV50_DISP_MTHD_V1_SOR_DP_MST_VCPI: { + struct nvkm_output_dp *outpdp = nvkm_output_dp(outp); + union { + struct nv50_disp_sor_dp_mst_vcpi_v0 v0; + } *args = data; + int ret = -ENOSYS; + nvif_ioctl(object, "disp sor dp mst vcpi size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, "disp sor dp mst vcpi vers %d " + "slot %02x/%02x pbn %04x/%04x\n", + args->v0.version, args->v0.start_slot, + args->v0.num_slots, args->v0.pbn, + args->v0.aligned_pbn); + if (!outpdp->func->vcpi) + return -ENODEV; + outpdp->func->vcpi(outpdp, head, args->v0.start_slot, + args->v0.num_slots, args->v0.pbn, + args->v0.aligned_pbn); + return 0; + } else + return ret; + } + break; case NV50_DISP_MTHD_V1_PIOR_PWR: if (!func->pior.power) return -ENODEV; @@ -207,8 +250,8 @@ nv50_disp_root_pioc_new_(const struct nvkm_oclass *oclass, { const struct nv50_disp_pioc_oclass *sclass = oclass->priv; struct nv50_disp_root *root = nv50_disp_root(oclass->parent); - return sclass->ctor(sclass->func, sclass->mthd, root, sclass->chid, - oclass, data, size, pobject); + return sclass->ctor(sclass->func, sclass->mthd, root, sclass->chid.ctrl, + sclass->chid.user, oclass, data, size, pobject); } static int diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.h index ad00f1724b72..b147cf5b3518 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.h @@ -41,5 +41,5 @@ extern const struct nvkm_disp_oclass gk110_disp_root_oclass; extern const struct nvkm_disp_oclass gm107_disp_root_oclass; extern const struct nvkm_disp_oclass gm200_disp_root_oclass; extern const struct nvkm_disp_oclass gp100_disp_root_oclass; -extern const struct nvkm_disp_oclass gp104_disp_root_oclass; +extern const struct nvkm_disp_oclass gp102_disp_root_oclass; #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c index 1bb9d661e9b3..627b9ee1ddd2 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c @@ -39,21 +39,6 @@ g94_sor_loff(struct nvkm_output_dp *outp) } /******************************************************************************* - * TMDS/LVDS - ******************************************************************************/ -static const struct nvkm_output_func -g94_sor_output_func = { -}; - -int -g94_sor_output_new(struct nvkm_disp *disp, int index, - struct dcb_output *dcbE, struct nvkm_output **poutp) -{ - return nvkm_output_new_(&g94_sor_output_func, disp, - index, dcbE, poutp); -} - -/******************************************************************************* * DisplayPort ******************************************************************************/ u32 diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c index 49bd5da194e1..6ffdaa65aa77 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c @@ -56,11 +56,13 @@ gf119_sor_dp_lnk_ctl(struct nvkm_output_dp *outp, int nr, int bw, bool ef) clksor |= bw << 18; dpctrl |= ((1 << nr) - 1) << 16; + if (outp->lt.mst) + dpctrl |= 0x40000000; if (ef) dpctrl |= 0x00004000; nvkm_mask(device, 0x612300 + soff, 0x007c0000, clksor); - nvkm_mask(device, 0x61c10c + loff, 0x001f4000, dpctrl); + nvkm_mask(device, 0x61c10c + loff, 0x401f4000, dpctrl); return 0; } @@ -101,12 +103,24 @@ gf119_sor_dp_drv_ctl(struct nvkm_output_dp *outp, return 0; } +void +gf119_sor_dp_vcpi(struct nvkm_output_dp *outp, int head, u8 slot, + u8 slot_nr, u16 pbn, u16 aligned) +{ + struct nvkm_device *device = outp->base.disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x616588 + hoff, 0x00003f3f, (slot_nr << 8) | slot); + nvkm_mask(device, 0x61658c + hoff, 0xffffffff, (aligned << 16) | pbn); +} + static const struct nvkm_output_dp_func gf119_sor_dp_func = { .pattern = gf119_sor_dp_pattern, .lnk_pwr = g94_sor_dp_lnk_pwr, .lnk_ctl = gf119_sor_dp_lnk_ctl, .drv_ctl = gf119_sor_dp_drv_ctl, + .vcpi = gf119_sor_dp_vcpi, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm107.c index 37790b2617c5..4cf8ad4d18ab 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm107.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm107.c @@ -43,6 +43,7 @@ gm107_sor_dp_func = { .lnk_pwr = g94_sor_dp_lnk_pwr, .lnk_ctl = gf119_sor_dp_lnk_ctl, .drv_ctl = gf119_sor_dp_drv_ctl, + .vcpi = gf119_sor_dp_vcpi, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm200.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm200.c index c44fa7ea672a..81b788fa61be 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm200.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgm200.c @@ -120,6 +120,7 @@ gm200_sor_dp_func = { .lnk_pwr = gm200_sor_dp_lnk_pwr, .lnk_ctl = gf119_sor_dp_lnk_ctl, .drv_ctl = gm200_sor_dp_drv_ctl, + .vcpi = gf119_sor_dp_vcpi, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c index f11ebdd16c77..11b7b8fd5dda 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c @@ -28,24 +28,6 @@ #include <nvif/class.h> -struct nvkm_dmaobj * -nvkm_dma_search(struct nvkm_dma *dma, struct nvkm_client *client, u64 object) -{ - struct rb_node *node = client->dmaroot.rb_node; - while (node) { - struct nvkm_dmaobj *dmaobj = - container_of(node, typeof(*dmaobj), rb); - if (object < dmaobj->handle) - node = node->rb_left; - else - if (object > dmaobj->handle) - node = node->rb_right; - else - return dmaobj; - } - return NULL; -} - static int nvkm_dma_oclass_new(struct nvkm_device *device, const struct nvkm_oclass *oclass, void *data, u32 size, @@ -53,34 +35,12 @@ nvkm_dma_oclass_new(struct nvkm_device *device, { struct nvkm_dma *dma = nvkm_dma(oclass->engine); struct nvkm_dmaobj *dmaobj = NULL; - struct nvkm_client *client = oclass->client; - struct rb_node **ptr = &client->dmaroot.rb_node; - struct rb_node *parent = NULL; int ret; ret = dma->func->class_new(dma, oclass, data, size, &dmaobj); if (dmaobj) *pobject = &dmaobj->object; - if (ret) - return ret; - - dmaobj->handle = oclass->object; - - while (*ptr) { - struct nvkm_dmaobj *obj = container_of(*ptr, typeof(*obj), rb); - parent = *ptr; - if (dmaobj->handle < obj->handle) - ptr = &parent->rb_left; - else - if (dmaobj->handle > obj->handle) - ptr = &parent->rb_right; - else - return -EEXIST; - } - - rb_link_node(&dmaobj->rb, parent, ptr); - rb_insert_color(&dmaobj->rb, &client->dmaroot); - return 0; + return ret; } static const struct nvkm_device_oclass diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c index 13c661b1ef14..d20cc0681a88 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c @@ -31,6 +31,19 @@ #include <nvif/cl0002.h> #include <nvif/unpack.h> +static const struct nvkm_object_func nvkm_dmaobj_func; +struct nvkm_dmaobj * +nvkm_dmaobj_search(struct nvkm_client *client, u64 handle) +{ + struct nvkm_object *object; + + object = nvkm_object_search(client, handle, &nvkm_dmaobj_func); + if (IS_ERR(object)) + return (void *)object; + + return nvkm_dmaobj(object); +} + static int nvkm_dmaobj_bind(struct nvkm_object *base, struct nvkm_gpuobj *gpuobj, int align, struct nvkm_gpuobj **pgpuobj) @@ -42,10 +55,7 @@ nvkm_dmaobj_bind(struct nvkm_object *base, struct nvkm_gpuobj *gpuobj, static void * nvkm_dmaobj_dtor(struct nvkm_object *base) { - struct nvkm_dmaobj *dmaobj = nvkm_dmaobj(base); - if (!RB_EMPTY_NODE(&dmaobj->rb)) - rb_erase(&dmaobj->rb, &dmaobj->object.client->dmaroot); - return dmaobj; + return nvkm_dmaobj(base); } static const struct nvkm_object_func @@ -74,7 +84,6 @@ nvkm_dmaobj_ctor(const struct nvkm_dmaobj_func *func, struct nvkm_dma *dma, nvkm_object_ctor(&nvkm_dmaobj_func, oclass, &dmaobj->object); dmaobj->func = func; dmaobj->dma = dma; - RB_CLEAR_NODE(&dmaobj->rb); nvif_ioctl(parent, "create dma size %d\n", *psize); if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, true))) { diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c index 1c9682ae3a6b..660ca7aa95ea 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c @@ -32,6 +32,17 @@ #include <nvif/unpack.h> void +nvkm_fifo_recover_chan(struct nvkm_fifo *fifo, int chid) +{ + unsigned long flags; + if (WARN_ON(!fifo->func->recover_chan)) + return; + spin_lock_irqsave(&fifo->lock, flags); + fifo->func->recover_chan(fifo, chid); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +void nvkm_fifo_pause(struct nvkm_fifo *fifo, unsigned long *flags) { return fifo->func->pause(fifo, flags); @@ -55,19 +66,29 @@ nvkm_fifo_chan_put(struct nvkm_fifo *fifo, unsigned long flags, } struct nvkm_fifo_chan * -nvkm_fifo_chan_inst(struct nvkm_fifo *fifo, u64 inst, unsigned long *rflags) +nvkm_fifo_chan_inst_locked(struct nvkm_fifo *fifo, u64 inst) { struct nvkm_fifo_chan *chan; - unsigned long flags; - spin_lock_irqsave(&fifo->lock, flags); list_for_each_entry(chan, &fifo->chan, head) { if (chan->inst->addr == inst) { list_del(&chan->head); list_add(&chan->head, &fifo->chan); - *rflags = flags; return chan; } } + return NULL; +} + +struct nvkm_fifo_chan * +nvkm_fifo_chan_inst(struct nvkm_fifo *fifo, u64 inst, unsigned long *rflags) +{ + struct nvkm_fifo_chan *chan; + unsigned long flags; + spin_lock_irqsave(&fifo->lock, flags); + if ((chan = nvkm_fifo_chan_inst_locked(fifo, inst))) { + *rflags = flags; + return chan; + } spin_unlock_irqrestore(&fifo->lock, flags); return NULL; } @@ -90,9 +111,34 @@ nvkm_fifo_chan_chid(struct nvkm_fifo *fifo, int chid, unsigned long *rflags) return NULL; } +void +nvkm_fifo_kevent(struct nvkm_fifo *fifo, int chid) +{ + nvkm_event_send(&fifo->kevent, 1, chid, NULL, 0); +} + static int -nvkm_fifo_event_ctor(struct nvkm_object *object, void *data, u32 size, - struct nvkm_notify *notify) +nvkm_fifo_kevent_ctor(struct nvkm_object *object, void *data, u32 size, + struct nvkm_notify *notify) +{ + struct nvkm_fifo_chan *chan = nvkm_fifo_chan(object); + if (size == 0) { + notify->size = 0; + notify->types = 1; + notify->index = chan->chid; + return 0; + } + return -ENOSYS; +} + +static const struct nvkm_event_func +nvkm_fifo_kevent_func = { + .ctor = nvkm_fifo_kevent_ctor, +}; + +static int +nvkm_fifo_cevent_ctor(struct nvkm_object *object, void *data, u32 size, + struct nvkm_notify *notify) { if (size == 0) { notify->size = 0; @@ -104,10 +150,16 @@ nvkm_fifo_event_ctor(struct nvkm_object *object, void *data, u32 size, } static const struct nvkm_event_func -nvkm_fifo_event_func = { - .ctor = nvkm_fifo_event_ctor, +nvkm_fifo_cevent_func = { + .ctor = nvkm_fifo_cevent_ctor, }; +void +nvkm_fifo_cevent(struct nvkm_fifo *fifo) +{ + nvkm_event_send(&fifo->cevent, 1, 0, NULL, 0); +} + static void nvkm_fifo_uevent_fini(struct nvkm_event *event, int type, int index) { @@ -241,6 +293,7 @@ nvkm_fifo_dtor(struct nvkm_engine *engine) void *data = fifo; if (fifo->func->dtor) data = fifo->func->dtor(fifo); + nvkm_event_fini(&fifo->kevent); nvkm_event_fini(&fifo->cevent); nvkm_event_fini(&fifo->uevent); return data; @@ -283,5 +336,9 @@ nvkm_fifo_ctor(const struct nvkm_fifo_func *func, struct nvkm_device *device, return ret; } - return nvkm_event_init(&nvkm_fifo_event_func, 1, 1, &fifo->cevent); + ret = nvkm_event_init(&nvkm_fifo_cevent_func, 1, 1, &fifo->cevent); + if (ret) + return ret; + + return nvkm_event_init(&nvkm_fifo_kevent_func, 1, nr, &fifo->kevent); } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c index dc6d4678f228..fab760ae922f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c @@ -371,9 +371,9 @@ nvkm_fifo_chan_ctor(const struct nvkm_fifo_chan_func *func, /* allocate push buffer ctxdma instance */ if (push) { - dmaobj = nvkm_dma_search(device->dma, oclass->client, push); - if (!dmaobj) - return -ENOENT; + dmaobj = nvkm_dmaobj_search(client, push); + if (IS_ERR(dmaobj)) + return PTR_ERR(dmaobj); ret = nvkm_object_bind(&dmaobj->object, chan->inst, -16, &chan->push); @@ -410,6 +410,6 @@ nvkm_fifo_chan_ctor(const struct nvkm_fifo_chan_func *func, base + user * chan->chid; chan->size = user; - nvkm_event_send(&fifo->cevent, 1, 0, NULL, 0); + nvkm_fifo_cevent(fifo); return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h index 55dc415c5c08..d8019bdacd61 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h @@ -29,5 +29,5 @@ struct nvkm_fifo_chan_oclass { struct nvkm_sclass base; }; -int g84_fifo_chan_ntfy(struct nvkm_fifo_chan *, u32, struct nvkm_event **); +int gf100_fifo_chan_ntfy(struct nvkm_fifo_chan *, u32, struct nvkm_event **); #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chang84.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chang84.c index aeb3387a3fb0..61797c4dd07a 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chang84.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chang84.c @@ -30,12 +30,12 @@ #include <nvif/cl826e.h> -int +static int g84_fifo_chan_ntfy(struct nvkm_fifo_chan *chan, u32 type, struct nvkm_event **pevent) { switch (type) { - case G82_CHANNEL_DMA_V0_NTFY_UEVENT: + case NV826E_V0_NTFY_NON_STALL_INTERRUPT: *pevent = &chan->fifo->uevent; return 0; default: @@ -129,7 +129,7 @@ g84_fifo_chan_engine_fini(struct nvkm_fifo_chan *base, } -int +static int g84_fifo_chan_engine_init(struct nvkm_fifo_chan *base, struct nvkm_engine *engine) { @@ -170,7 +170,7 @@ g84_fifo_chan_engine_ctor(struct nvkm_fifo_chan *base, return nvkm_object_bind(object, NULL, 0, &chan->engn[engn]); } -int +static int g84_fifo_chan_object_ctor(struct nvkm_fifo_chan *base, struct nvkm_object *object) { diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c index 352a0baec84d..cd468ab1db12 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c @@ -68,7 +68,14 @@ gf100_fifo_runlist_commit(struct gf100_fifo *fifo) } nvkm_done(cur); - target = (nvkm_memory_target(cur) == NVKM_MEM_TARGET_HOST) ? 0x3 : 0x0; + switch (nvkm_memory_target(cur)) { + case NVKM_MEM_TARGET_VRAM: target = 0; break; + case NVKM_MEM_TARGET_NCOH: target = 3; break; + default: + mutex_unlock(&subdev->mutex); + WARN_ON(1); + return; + } nvkm_wr32(device, 0x002270, (nvkm_memory_addr(cur) >> 12) | (target << 28)); @@ -180,8 +187,10 @@ gf100_fifo_recover(struct gf100_fifo *fifo, struct nvkm_engine *engine, list_del_init(&chan->head); chan->killed = true; - fifo->recover.mask |= 1ULL << engine->subdev.index; + if (engine != &fifo->base.engine) + fifo->recover.mask |= 1ULL << engine->subdev.index; schedule_work(&fifo->recover.work); + nvkm_fifo_kevent(&fifo->base, chid); } static const struct nvkm_enum diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c index 103c0afaaa6d..3a24788c3185 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c @@ -27,11 +27,71 @@ #include <core/client.h> #include <core/gpuobj.h> #include <subdev/bar.h> +#include <subdev/timer.h> #include <subdev/top.h> #include <engine/sw.h> #include <nvif/class.h> +struct gk104_fifo_engine_status { + bool busy; + bool faulted; + bool chsw; + bool save; + bool load; + struct { + bool tsg; + u32 id; + } prev, next, *chan; +}; + +static void +gk104_fifo_engine_status(struct gk104_fifo *fifo, int engn, + struct gk104_fifo_engine_status *status) +{ + struct nvkm_engine *engine = fifo->engine[engn].engine; + struct nvkm_subdev *subdev = &fifo->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x002640 + (engn * 0x08)); + + status->busy = !!(stat & 0x80000000); + status->faulted = !!(stat & 0x40000000); + status->next.tsg = !!(stat & 0x10000000); + status->next.id = (stat & 0x0fff0000) >> 16; + status->chsw = !!(stat & 0x00008000); + status->save = !!(stat & 0x00004000); + status->load = !!(stat & 0x00002000); + status->prev.tsg = !!(stat & 0x00001000); + status->prev.id = (stat & 0x00000fff); + status->chan = NULL; + + if (status->busy && status->chsw) { + if (status->load && status->save) { + if (engine && nvkm_engine_chsw_load(engine)) + status->chan = &status->next; + else + status->chan = &status->prev; + } else + if (status->load) { + status->chan = &status->next; + } else { + status->chan = &status->prev; + } + } else + if (status->load) { + status->chan = &status->prev; + } + + nvkm_debug(subdev, "engine %02d: busy %d faulted %d chsw %d " + "save %d load %d %sid %d%s-> %sid %d%s\n", + engn, status->busy, status->faulted, + status->chsw, status->save, status->load, + status->prev.tsg ? "tsg" : "ch", status->prev.id, + status->chan == &status->prev ? "*" : " ", + status->next.tsg ? "tsg" : "ch", status->next.id, + status->chan == &status->next ? "*" : " "); +} + static int gk104_fifo_class_get(struct nvkm_fifo *base, int index, const struct nvkm_fifo_chan_oclass **psclass) @@ -83,10 +143,13 @@ gk104_fifo_runlist_commit(struct gk104_fifo *fifo, int runl) } nvkm_done(mem); - if (nvkm_memory_target(mem) == NVKM_MEM_TARGET_VRAM) - target = 0; - else - target = 3; + switch (nvkm_memory_target(mem)) { + case NVKM_MEM_TARGET_VRAM: target = 0; break; + case NVKM_MEM_TARGET_NCOH: target = 3; break; + default: + WARN_ON(1); + return; + } nvkm_wr32(device, 0x002270, (nvkm_memory_addr(mem) >> 12) | (target << 28)); @@ -149,31 +212,137 @@ gk104_fifo_recover_work(struct work_struct *w) nvkm_mask(device, 0x002630, runm, 0x00000000); } +static void gk104_fifo_recover_engn(struct gk104_fifo *fifo, int engn); + static void -gk104_fifo_recover(struct gk104_fifo *fifo, struct nvkm_engine *engine, - struct gk104_fifo_chan *chan) +gk104_fifo_recover_runl(struct gk104_fifo *fifo, int runl) { struct nvkm_subdev *subdev = &fifo->base.engine.subdev; struct nvkm_device *device = subdev->device; - u32 chid = chan->base.chid; - int engn; + const u32 runm = BIT(runl); - nvkm_error(subdev, "%s engine fault on channel %d, recovering...\n", - nvkm_subdev_name[engine->subdev.index], chid); assert_spin_locked(&fifo->base.lock); + if (fifo->recover.runm & runm) + return; + fifo->recover.runm |= runm; - nvkm_mask(device, 0x800004 + (chid * 0x08), 0x00000800, 0x00000800); - list_del_init(&chan->head); - chan->killed = true; + /* Block runlist to prevent channel assignment(s) from changing. */ + nvkm_mask(device, 0x002630, runm, runm); - for (engn = 0; engn < fifo->engine_nr; engn++) { - if (fifo->engine[engn].engine == engine) { - fifo->recover.engm |= BIT(engn); + /* Schedule recovery. */ + nvkm_warn(subdev, "runlist %d: scheduled for recovery\n", runl); + schedule_work(&fifo->recover.work); +} + +static void +gk104_fifo_recover_chan(struct nvkm_fifo *base, int chid) +{ + struct gk104_fifo *fifo = gk104_fifo(base); + struct nvkm_subdev *subdev = &fifo->base.engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 stat = nvkm_rd32(device, 0x800004 + (chid * 0x08)); + const u32 runl = (stat & 0x000f0000) >> 16; + const bool used = (stat & 0x00000001); + unsigned long engn, engm = fifo->runlist[runl].engm; + struct gk104_fifo_chan *chan; + + assert_spin_locked(&fifo->base.lock); + if (!used) + return; + + /* Lookup SW state for channel, and mark it as dead. */ + list_for_each_entry(chan, &fifo->runlist[runl].chan, head) { + if (chan->base.chid == chid) { + list_del_init(&chan->head); + chan->killed = true; + nvkm_fifo_kevent(&fifo->base, chid); break; } } - fifo->recover.runm |= BIT(chan->runl); + /* Disable channel. */ + nvkm_wr32(device, 0x800004 + (chid * 0x08), stat | 0x00000800); + nvkm_warn(subdev, "channel %d: killed\n", chid); + + /* Block channel assignments from changing during recovery. */ + gk104_fifo_recover_runl(fifo, runl); + + /* Schedule recovery for any engines the channel is on. */ + for_each_set_bit(engn, &engm, fifo->engine_nr) { + struct gk104_fifo_engine_status status; + gk104_fifo_engine_status(fifo, engn, &status); + if (!status.chan || status.chan->id != chid) + continue; + gk104_fifo_recover_engn(fifo, engn); + } +} + +static void +gk104_fifo_recover_engn(struct gk104_fifo *fifo, int engn) +{ + struct nvkm_engine *engine = fifo->engine[engn].engine; + struct nvkm_subdev *subdev = &fifo->base.engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 runl = fifo->engine[engn].runl; + const u32 engm = BIT(engn); + struct gk104_fifo_engine_status status; + int mmui = -1; + + assert_spin_locked(&fifo->base.lock); + if (fifo->recover.engm & engm) + return; + fifo->recover.engm |= engm; + + /* Block channel assignments from changing during recovery. */ + gk104_fifo_recover_runl(fifo, runl); + + /* Determine which channel (if any) is currently on the engine. */ + gk104_fifo_engine_status(fifo, engn, &status); + if (status.chan) { + /* The channel is not longer viable, kill it. */ + gk104_fifo_recover_chan(&fifo->base, status.chan->id); + } + + /* Determine MMU fault ID for the engine, if we're not being + * called from the fault handler already. + */ + if (!status.faulted && engine) { + mmui = nvkm_top_fault_id(device, engine->subdev.index); + if (mmui < 0) { + const struct nvkm_enum *en = fifo->func->fault.engine; + for (; en && en->name; en++) { + if (en->data2 == engine->subdev.index) { + mmui = en->value; + break; + } + } + } + WARN_ON(mmui < 0); + } + + /* Trigger a MMU fault for the engine. + * + * No good idea why this is needed, but nvgpu does something similar, + * and it makes recovery from CTXSW_TIMEOUT a lot more reliable. + */ + if (mmui >= 0) { + nvkm_wr32(device, 0x002a30 + (engn * 0x04), 0x00000100 | mmui); + + /* Wait for fault to trigger. */ + nvkm_msec(device, 2000, + gk104_fifo_engine_status(fifo, engn, &status); + if (status.faulted) + break; + ); + + /* Release MMU fault trigger, and ACK the fault. */ + nvkm_wr32(device, 0x002a30 + (engn * 0x04), 0x00000000); + nvkm_wr32(device, 0x00259c, BIT(mmui)); + nvkm_wr32(device, 0x002100, 0x10000000); + } + + /* Schedule recovery. */ + nvkm_warn(subdev, "engine %d: scheduled for recovery\n", engn); schedule_work(&fifo->recover.work); } @@ -211,34 +380,30 @@ static void gk104_fifo_intr_sched_ctxsw(struct gk104_fifo *fifo) { struct nvkm_device *device = fifo->base.engine.subdev.device; - struct gk104_fifo_chan *chan; - unsigned long flags; + unsigned long flags, engm = 0; u32 engn; + /* We need to ACK the SCHED_ERROR here, and prevent it reasserting, + * as MMU_FAULT cannot be triggered while it's pending. + */ spin_lock_irqsave(&fifo->base.lock, flags); + nvkm_mask(device, 0x002140, 0x00000100, 0x00000000); + nvkm_wr32(device, 0x002100, 0x00000100); + for (engn = 0; engn < fifo->engine_nr; engn++) { - struct nvkm_engine *engine = fifo->engine[engn].engine; - int runl = fifo->engine[engn].runl; - u32 stat = nvkm_rd32(device, 0x002640 + (engn * 0x08)); - u32 busy = (stat & 0x80000000); - u32 next = (stat & 0x0fff0000) >> 16; - u32 chsw = (stat & 0x00008000); - u32 save = (stat & 0x00004000); - u32 load = (stat & 0x00002000); - u32 prev = (stat & 0x00000fff); - u32 chid = load ? next : prev; - (void)save; - - if (!busy || !chsw) + struct gk104_fifo_engine_status status; + + gk104_fifo_engine_status(fifo, engn, &status); + if (!status.busy || !status.chsw) continue; - list_for_each_entry(chan, &fifo->runlist[runl].chan, head) { - if (chan->base.chid == chid && engine) { - gk104_fifo_recover(fifo, engine, chan); - break; - } - } + engm |= BIT(engn); } + + for_each_set_bit(engn, &engm, fifo->engine_nr) + gk104_fifo_recover_engn(fifo, engn); + + nvkm_mask(device, 0x002140, 0x00000100, 0x00000100); spin_unlock_irqrestore(&fifo->base.lock, flags); } @@ -301,6 +466,7 @@ gk104_fifo_intr_fault(struct gk104_fifo *fifo, int unit) struct nvkm_fifo_chan *chan; unsigned long flags; char gpcid[8] = "", en[16] = ""; + int engn; er = nvkm_enum_find(fifo->func->fault.reason, reason); eu = nvkm_enum_find(fifo->func->fault.engine, unit); @@ -342,7 +508,8 @@ gk104_fifo_intr_fault(struct gk104_fifo *fifo, int unit) snprintf(en, sizeof(en), "%s", eu->name); } - chan = nvkm_fifo_chan_inst(&fifo->base, (u64)inst << 12, &flags); + spin_lock_irqsave(&fifo->base.lock, flags); + chan = nvkm_fifo_chan_inst_locked(&fifo->base, (u64)inst << 12); nvkm_error(subdev, "%s fault at %010llx engine %02x [%s] client %02x [%s%s] " @@ -353,9 +520,23 @@ gk104_fifo_intr_fault(struct gk104_fifo *fifo, int unit) (u64)inst << 12, chan ? chan->object.client->name : "unknown"); - if (engine && chan) - gk104_fifo_recover(fifo, engine, (void *)chan); - nvkm_fifo_chan_put(&fifo->base, flags, &chan); + + /* Kill the channel that caused the fault. */ + if (chan) + gk104_fifo_recover_chan(&fifo->base, chan->chid); + + /* Channel recovery will probably have already done this for the + * correct engine(s), but just in case we can't find the channel + * information... + */ + for (engn = 0; engn < fifo->engine_nr && engine; engn++) { + if (fifo->engine[engn].engine == engine) { + gk104_fifo_recover_engn(fifo, engn); + break; + } + } + + spin_unlock_irqrestore(&fifo->base.lock, flags); } static const struct nvkm_bitfield gk104_fifo_pbdma_intr_0[] = { @@ -716,6 +897,7 @@ gk104_fifo_ = { .intr = gk104_fifo_intr, .uevent_init = gk104_fifo_uevent_init, .uevent_fini = gk104_fifo_uevent_fini, + .recover_chan = gk104_fifo_recover_chan, .class_get = gk104_fifo_class_get, }; @@ -743,14 +925,14 @@ gk104_fifo_fault_engine[] = { { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, { 0x06, "SCHED" }, - { 0x07, "HOST0" }, - { 0x08, "HOST1" }, - { 0x09, "HOST2" }, - { 0x0a, "HOST3" }, - { 0x0b, "HOST4" }, - { 0x0c, "HOST5" }, - { 0x0d, "HOST6" }, - { 0x0e, "HOST7" }, + { 0x07, "HOST0", NULL, NVKM_ENGINE_FIFO }, + { 0x08, "HOST1", NULL, NVKM_ENGINE_FIFO }, + { 0x09, "HOST2", NULL, NVKM_ENGINE_FIFO }, + { 0x0a, "HOST3", NULL, NVKM_ENGINE_FIFO }, + { 0x0b, "HOST4", NULL, NVKM_ENGINE_FIFO }, + { 0x0c, "HOST5", NULL, NVKM_ENGINE_FIFO }, + { 0x0d, "HOST6", NULL, NVKM_ENGINE_FIFO }, + { 0x0e, "HOST7", NULL, NVKM_ENGINE_FIFO }, { 0x0f, "HOSTSR" }, { 0x10, "MSVLD", NULL, NVKM_ENGINE_MSVLD }, { 0x11, "MSPPP", NULL, NVKM_ENGINE_MSPPP }, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c index bd1ff877aa06..29c080683b32 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c @@ -32,14 +32,14 @@ gm107_fifo_fault_engine[] = { { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, { 0x06, "SCHED" }, - { 0x07, "HOST0" }, - { 0x08, "HOST1" }, - { 0x09, "HOST2" }, - { 0x0a, "HOST3" }, - { 0x0b, "HOST4" }, - { 0x0c, "HOST5" }, - { 0x0d, "HOST6" }, - { 0x0e, "HOST7" }, + { 0x07, "HOST0", NULL, NVKM_ENGINE_FIFO }, + { 0x08, "HOST1", NULL, NVKM_ENGINE_FIFO }, + { 0x09, "HOST2", NULL, NVKM_ENGINE_FIFO }, + { 0x0a, "HOST3", NULL, NVKM_ENGINE_FIFO }, + { 0x0b, "HOST4", NULL, NVKM_ENGINE_FIFO }, + { 0x0c, "HOST5", NULL, NVKM_ENGINE_FIFO }, + { 0x0d, "HOST6", NULL, NVKM_ENGINE_FIFO }, + { 0x0e, "HOST7", NULL, NVKM_ENGINE_FIFO }, { 0x0f, "HOSTSR" }, { 0x13, "PERF" }, { 0x17, "PMU" }, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c index eff83f7fb705..b2635aea9f6e 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c @@ -30,17 +30,17 @@ gp100_fifo_fault_engine[] = { { 0x03, "IFB", NULL, NVKM_ENGINE_IFB }, { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, - { 0x06, "HOST0" }, - { 0x07, "HOST1" }, - { 0x08, "HOST2" }, - { 0x09, "HOST3" }, - { 0x0a, "HOST4" }, - { 0x0b, "HOST5" }, - { 0x0c, "HOST6" }, - { 0x0d, "HOST7" }, - { 0x0e, "HOST8" }, - { 0x0f, "HOST9" }, - { 0x10, "HOST10" }, + { 0x06, "HOST0", NULL, NVKM_ENGINE_FIFO }, + { 0x07, "HOST1", NULL, NVKM_ENGINE_FIFO }, + { 0x08, "HOST2", NULL, NVKM_ENGINE_FIFO }, + { 0x09, "HOST3", NULL, NVKM_ENGINE_FIFO }, + { 0x0a, "HOST4", NULL, NVKM_ENGINE_FIFO }, + { 0x0b, "HOST5", NULL, NVKM_ENGINE_FIFO }, + { 0x0c, "HOST6", NULL, NVKM_ENGINE_FIFO }, + { 0x0d, "HOST7", NULL, NVKM_ENGINE_FIFO }, + { 0x0e, "HOST8", NULL, NVKM_ENGINE_FIFO }, + { 0x0f, "HOST9", NULL, NVKM_ENGINE_FIFO }, + { 0x10, "HOST10", NULL, NVKM_ENGINE_FIFO }, { 0x13, "PERF" }, { 0x17, "PMU" }, { 0x18, "PTP" }, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c index cbc67f262322..f9e0377d3d24 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogf100.c @@ -32,6 +32,23 @@ #include <nvif/cl906f.h> #include <nvif/unpack.h> +int +gf100_fifo_chan_ntfy(struct nvkm_fifo_chan *chan, u32 type, + struct nvkm_event **pevent) +{ + switch (type) { + case NV906F_V0_NTFY_NON_STALL_INTERRUPT: + *pevent = &chan->fifo->uevent; + return 0; + case NV906F_V0_NTFY_KILLED: + *pevent = &chan->fifo->kevent; + return 0; + default: + break; + } + return -EINVAL; +} + static u32 gf100_fifo_gpfifo_engine_addr(struct nvkm_engine *engine) { @@ -60,6 +77,7 @@ gf100_fifo_gpfifo_engine_fini(struct nvkm_fifo_chan *base, struct nvkm_gpuobj *inst = chan->base.inst; int ret = 0; + mutex_lock(&subdev->mutex); nvkm_wr32(device, 0x002634, chan->base.chid); if (nvkm_msec(device, 2000, if (nvkm_rd32(device, 0x002634) == chan->base.chid) @@ -67,10 +85,12 @@ gf100_fifo_gpfifo_engine_fini(struct nvkm_fifo_chan *base, ) < 0) { nvkm_error(subdev, "channel %d [%s] kick timeout\n", chan->base.chid, chan->base.object.client->name); - ret = -EBUSY; - if (suspend) - return ret; + ret = -ETIMEDOUT; } + mutex_unlock(&subdev->mutex); + + if (ret && suspend) + return ret; if (offset) { nvkm_kmap(inst); @@ -181,7 +201,7 @@ gf100_fifo_gpfifo_func = { .dtor = gf100_fifo_gpfifo_dtor, .init = gf100_fifo_gpfifo_init, .fini = gf100_fifo_gpfifo_fini, - .ntfy = g84_fifo_chan_ntfy, + .ntfy = gf100_fifo_chan_ntfy, .engine_ctor = gf100_fifo_gpfifo_engine_ctor, .engine_dtor = gf100_fifo_gpfifo_engine_dtor, .engine_init = gf100_fifo_gpfifo_engine_init, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c index ed4351032ed6..8abf6f8ef445 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gpfifogk104.c @@ -40,7 +40,9 @@ gk104_fifo_gpfifo_kick(struct gk104_fifo_chan *chan) struct nvkm_subdev *subdev = &fifo->base.engine.subdev; struct nvkm_device *device = subdev->device; struct nvkm_client *client = chan->base.object.client; + int ret = 0; + mutex_lock(&subdev->mutex); nvkm_wr32(device, 0x002634, chan->base.chid); if (nvkm_msec(device, 2000, if (!(nvkm_rd32(device, 0x002634) & 0x00100000)) @@ -48,10 +50,11 @@ gk104_fifo_gpfifo_kick(struct gk104_fifo_chan *chan) ) < 0) { nvkm_error(subdev, "channel %d [%s] kick timeout\n", chan->base.chid, client->name); - return -EBUSY; + nvkm_fifo_recover_chan(&fifo->base, chan->base.chid); + ret = -ETIMEDOUT; } - - return 0; + mutex_unlock(&subdev->mutex); + return ret; } static u32 @@ -211,7 +214,7 @@ gk104_fifo_gpfifo_func = { .dtor = gk104_fifo_gpfifo_dtor, .init = gk104_fifo_gpfifo_init, .fini = gk104_fifo_gpfifo_fini, - .ntfy = g84_fifo_chan_ntfy, + .ntfy = gf100_fifo_chan_ntfy, .engine_ctor = gk104_fifo_gpfifo_engine_ctor, .engine_dtor = gk104_fifo_gpfifo_engine_dtor, .engine_init = gk104_fifo_gpfifo_engine_init, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h index f6dfb37d9429..f889b13b5e41 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h @@ -6,6 +6,12 @@ int nvkm_fifo_ctor(const struct nvkm_fifo_func *, struct nvkm_device *, int index, int nr, struct nvkm_fifo *); void nvkm_fifo_uevent(struct nvkm_fifo *); +void nvkm_fifo_cevent(struct nvkm_fifo *); +void nvkm_fifo_kevent(struct nvkm_fifo *, int chid); +void nvkm_fifo_recover_chan(struct nvkm_fifo *, int chid); + +struct nvkm_fifo_chan * +nvkm_fifo_chan_inst_locked(struct nvkm_fifo *, u64 inst); struct nvkm_fifo_chan_oclass; struct nvkm_fifo_func { @@ -18,6 +24,7 @@ struct nvkm_fifo_func { void (*start)(struct nvkm_fifo *, unsigned long *); void (*uevent_init)(struct nvkm_fifo *); void (*uevent_fini)(struct nvkm_fifo *); + void (*recover_chan)(struct nvkm_fifo *, int chid); int (*class_get)(struct nvkm_fifo *, int index, const struct nvkm_fifo_chan_oclass **); const struct nvkm_fifo_chan_oclass *chan[]; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c index 467065d1b4e6..cd8cf6f7024c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c @@ -25,6 +25,15 @@ #include <engine/fifo.h> +static bool +nvkm_gr_chsw_load(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->chsw_load) + return gr->func->chsw_load(gr); + return false; +} + static void nvkm_gr_tile(struct nvkm_engine *engine, int region, struct nvkm_fb_tile *tile) { @@ -106,6 +115,15 @@ nvkm_gr_init(struct nvkm_engine *engine) return gr->func->init(gr); } +static int +nvkm_gr_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->fini) + return gr->func->fini(gr, suspend); + return 0; +} + static void * nvkm_gr_dtor(struct nvkm_engine *engine) { @@ -120,8 +138,10 @@ nvkm_gr = { .dtor = nvkm_gr_dtor, .oneinit = nvkm_gr_oneinit, .init = nvkm_gr_init, + .fini = nvkm_gr_fini, .intr = nvkm_gr_intr, .tile = nvkm_gr_tile, + .chsw_load = nvkm_gr_chsw_load, .fifo.cclass = nvkm_gr_cclass_new, .fifo.sclass = nvkm_gr_oclass_get, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c index c925ade5880e..74a64e3fd59a 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c @@ -218,7 +218,7 @@ gf117_grctx_generate_attrib(struct gf100_grctx *info) } } -void +static void gf117_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) { struct nvkm_device *device = gr->base.engine.subdev.device; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c index 6d3c5011e18c..4c4b5ab6e46d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c @@ -933,7 +933,7 @@ gm107_grctx_generate_attrib(struct gf100_grctx *info) } } -void +static void gm107_grctx_generate_tpcid(struct gf100_gr *gr) { struct nvkm_device *device = gr->base.engine.subdev.device; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c index 1e13278cf306..c8bb9191f9a2 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c @@ -106,6 +106,7 @@ #define CP_SEEK_2 0x00c800ff #include "ctxnv40.h" +#include "nv50.h" #include <subdev/fb.h> diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h index 8cb240b65ec2..12a703fe355d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gf100_grgpc_data[] = { +static uint32_t gf100_grgpc_data[] = { /* 0x0000: gpc_mmio_list_head */ 0x00000064, /* 0x0004: gpc_mmio_list_tail */ @@ -36,7 +36,7 @@ uint32_t gf100_grgpc_data[] = { 0x00000000, }; -uint32_t gf100_grgpc_code[] = { +static uint32_t gf100_grgpc_code[] = { 0x03a10ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h index 550d6ba0933b..ffbfc51200f1 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gf117_grgpc_data[] = { +static uint32_t gf117_grgpc_data[] = { /* 0x0000: gpc_mmio_list_head */ 0x0000006c, /* 0x0004: gpc_mmio_list_tail */ @@ -40,7 +40,7 @@ uint32_t gf117_grgpc_data[] = { 0x00000000, }; -uint32_t gf117_grgpc_code[] = { +static uint32_t gf117_grgpc_code[] = { 0x03a10ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h index 271b59d365e5..357f662de571 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gk104_grgpc_data[] = { +static uint32_t gk104_grgpc_data[] = { /* 0x0000: gpc_mmio_list_head */ 0x0000006c, /* 0x0004: gpc_mmio_list_tail */ @@ -40,7 +40,7 @@ uint32_t gk104_grgpc_data[] = { 0x00000000, }; -uint32_t gk104_grgpc_code[] = { +static uint32_t gk104_grgpc_code[] = { 0x03a10ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h index 73b4a32c5d29..4ffc8212a85c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gk110_grgpc_data[] = { +static uint32_t gk110_grgpc_data[] = { /* 0x0000: gpc_mmio_list_head */ 0x0000006c, /* 0x0004: gpc_mmio_list_tail */ @@ -40,7 +40,7 @@ uint32_t gk110_grgpc_data[] = { 0x00000000, }; -uint32_t gk110_grgpc_code[] = { +static uint32_t gk110_grgpc_code[] = { 0x03a10ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h index 018169818317..09196206c9bc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h @@ -1,4 +1,4 @@ -uint32_t gk208_grgpc_data[] = { +static uint32_t gk208_grgpc_data[] = { /* 0x0000: gpc_mmio_list_head */ 0x0000006c, /* 0x0004: gpc_mmio_list_tail */ @@ -40,7 +40,7 @@ uint32_t gk208_grgpc_data[] = { 0x00000000, }; -uint32_t gk208_grgpc_code[] = { +static uint32_t gk208_grgpc_code[] = { 0x03140ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h index eca007f03fa9..6d7d004363d9 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h @@ -1,4 +1,4 @@ -uint32_t gm107_grgpc_data[] = { +static uint32_t gm107_grgpc_data[] = { /* 0x0000: gpc_mmio_list_head */ 0x0000006c, /* 0x0004: gpc_mmio_list_tail */ @@ -40,7 +40,7 @@ uint32_t gm107_grgpc_data[] = { 0x00000000, }; -uint32_t gm107_grgpc_code[] = { +static uint32_t gm107_grgpc_code[] = { 0x03410ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h index 8015b40a61d6..7538404b8b13 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gf100_grhub_data[] = { +static uint32_t gf100_grhub_data[] = { /* 0x0000: hub_mmio_list_head */ 0x00000300, /* 0x0004: hub_mmio_list_tail */ @@ -205,7 +205,7 @@ uint32_t gf100_grhub_data[] = { 0x0417e91c, }; -uint32_t gf100_grhub_code[] = { +static uint32_t gf100_grhub_code[] = { 0x039b0ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h index 2af90ec6852a..ce000a47ec6d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gf117_grhub_data[] = { +static uint32_t gf117_grhub_data[] = { /* 0x0000: hub_mmio_list_head */ 0x00000300, /* 0x0004: hub_mmio_list_tail */ @@ -205,7 +205,7 @@ uint32_t gf117_grhub_data[] = { 0x0417e91c, }; -uint32_t gf117_grhub_code[] = { +static uint32_t gf117_grhub_code[] = { 0x039b0ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h index e8b8c1c94700..1f26cb6a233c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gk104_grhub_data[] = { +static uint32_t gk104_grhub_data[] = { /* 0x0000: hub_mmio_list_head */ 0x00000300, /* 0x0004: hub_mmio_list_tail */ @@ -205,7 +205,7 @@ uint32_t gk104_grhub_data[] = { 0x0417e91c, }; -uint32_t gk104_grhub_code[] = { +static uint32_t gk104_grhub_code[] = { 0x039b0ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h index f4ed2fb6f714..70436d93efe3 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gk110_grhub_data[] = { +static uint32_t gk110_grhub_data[] = { /* 0x0000: hub_mmio_list_head */ 0x00000300, /* 0x0004: hub_mmio_list_tail */ @@ -205,7 +205,7 @@ uint32_t gk110_grhub_data[] = { 0x0417e91c, }; -uint32_t gk110_grhub_code[] = { +static uint32_t gk110_grhub_code[] = { 0x039b0ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h index ed488973c117..e0933a07426a 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h @@ -1,4 +1,4 @@ -uint32_t gk208_grhub_data[] = { +static uint32_t gk208_grhub_data[] = { /* 0x0000: hub_mmio_list_head */ 0x00000300, /* 0x0004: hub_mmio_list_tail */ @@ -205,7 +205,7 @@ uint32_t gk208_grhub_data[] = { 0x0417e91c, }; -uint32_t gk208_grhub_code[] = { +static uint32_t gk208_grhub_code[] = { 0x030e0ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h index 5c9051839557..9b432823bcbe 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h @@ -1,4 +1,4 @@ -uint32_t gm107_grhub_data[] = { +static uint32_t gm107_grhub_data[] = { /* 0x0000: hub_mmio_list_head */ 0x00000300, /* 0x0004: hub_mmio_list_tail */ @@ -205,7 +205,7 @@ uint32_t gm107_grhub_data[] = { 0x0417e91c, }; -uint32_t gm107_grhub_code[] = { +static uint32_t gm107_grhub_code[] = { 0x030e0ef5, /* 0x0004: queue_put */ 0x9800d898, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c index ce913300539f..da1ba74682b4 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c @@ -25,6 +25,8 @@ #include <subdev/timer.h> +#include <nvif/class.h> + static const struct nvkm_bitfield nv50_gr_status[] = { { 0x00000001, "BUSY" }, /* set when any bit is set */ { 0x00000002, "DISPATCH" }, @@ -180,11 +182,11 @@ g84_gr = { .tlb_flush = g84_gr_tlb_flush, .units = nv50_gr_units, .sclass = { - { -1, -1, 0x0030, &nv50_gr_object }, - { -1, -1, 0x502d, &nv50_gr_object }, - { -1, -1, 0x5039, &nv50_gr_object }, - { -1, -1, 0x50c0, &nv50_gr_object }, - { -1, -1, 0x8297, &nv50_gr_object }, + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, G82_TESLA, &nv50_gr_object }, {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c index 157919c788e6..f9acb8a944d2 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c @@ -702,6 +702,22 @@ gf100_gr_pack_mmio[] = { * PGRAPH engine/subdev functions ******************************************************************************/ +static bool +gf100_gr_chsw_load(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + if (!gr->firmware) { + u32 trace = nvkm_rd32(gr->base.engine.subdev.device, 0x40981c); + if (trace & 0x00000040) + return true; + } else { + u32 mthd = nvkm_rd32(gr->base.engine.subdev.device, 0x409808); + if (mthd & 0x00080000) + return true; + } + return false; +} + int gf100_gr_rops(struct gf100_gr *gr) { @@ -1041,6 +1057,13 @@ gf100_gr_trap_tpc(struct gf100_gr *gr, int gpc, int tpc) stat &= ~0x00000008; } + if (stat & 0x00000010) { + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0430)); + nvkm_error(subdev, "GPC%d/TPC%d/MPC: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0430), 0xc0000000); + stat &= ~0x00000010; + } + if (stat) { nvkm_error(subdev, "GPC%d/TPC%d/%08x: unknown\n", gpc, tpc, stat); } @@ -1129,7 +1152,7 @@ gf100_gr_trap_intr(struct gf100_gr *gr) if (trap & 0x00000008) { u32 stat = nvkm_rd32(device, 0x408030); - nvkm_snprintbf(error, sizeof(error), gf100_m2mf_error, + nvkm_snprintbf(error, sizeof(error), gf100_ccache_error, stat & 0x3fffffff); nvkm_error(subdev, "CCACHE %08x [%s]\n", stat, error); nvkm_wr32(device, 0x408030, 0xc0000000); @@ -1258,7 +1281,7 @@ gf100_gr_ctxctl_isr(struct gf100_gr *gr) struct nvkm_device *device = subdev->device; u32 stat = nvkm_rd32(device, 0x409c18); - if (stat & 0x00000001) { + if (!gr->firmware && (stat & 0x00000001)) { u32 code = nvkm_rd32(device, 0x409814); if (code == E_BAD_FWMTHD) { u32 class = nvkm_rd32(device, 0x409808); @@ -1270,15 +1293,14 @@ gf100_gr_ctxctl_isr(struct gf100_gr *gr) nvkm_error(subdev, "FECS MTHD subc %d class %04x " "mthd %04x data %08x\n", subc, class, mthd, data); - - nvkm_wr32(device, 0x409c20, 0x00000001); - stat &= ~0x00000001; } else { nvkm_error(subdev, "FECS ucode error %d\n", code); } + nvkm_wr32(device, 0x409c20, 0x00000001); + stat &= ~0x00000001; } - if (stat & 0x00080000) { + if (!gr->firmware && (stat & 0x00080000)) { nvkm_error(subdev, "FECS watchdog timeout\n"); gf100_gr_ctxctl_debug(gr); nvkm_wr32(device, 0x409c20, 0x00080000); @@ -1384,27 +1406,12 @@ gf100_gr_intr(struct nvkm_gr *base) nvkm_fifo_chan_put(device->fifo, flags, &chan); } -void -gf100_gr_init_fw(struct gf100_gr *gr, u32 fuc_base, +static void +gf100_gr_init_fw(struct nvkm_falcon *falcon, struct gf100_gr_fuc *code, struct gf100_gr_fuc *data) { - struct nvkm_device *device = gr->base.engine.subdev.device; - int i; - - nvkm_wr32(device, fuc_base + 0x01c0, 0x01000000); - for (i = 0; i < data->size / 4; i++) - nvkm_wr32(device, fuc_base + 0x01c4, data->data[i]); - - nvkm_wr32(device, fuc_base + 0x0180, 0x01000000); - for (i = 0; i < code->size / 4; i++) { - if ((i & 0x3f) == 0) - nvkm_wr32(device, fuc_base + 0x0188, i >> 6); - nvkm_wr32(device, fuc_base + 0x0184, code->data[i]); - } - - /* code must be padded to 0x40 words */ - for (; i & 0x3f; i++) - nvkm_wr32(device, fuc_base + 0x0184, 0); + nvkm_falcon_load_dmem(falcon, data->data, 0x0, data->size, 0); + nvkm_falcon_load_imem(falcon, code->data, 0x0, code->size, 0, 0, false); } static void @@ -1449,162 +1456,149 @@ gf100_gr_init_csdata(struct gf100_gr *gr, nvkm_wr32(device, falcon + 0x01c4, star + 4); } -int -gf100_gr_init_ctxctl(struct gf100_gr *gr) +/* Initialize context from an external (secure or not) firmware */ +static int +gf100_gr_init_ctxctl_ext(struct gf100_gr *gr) { - const struct gf100_grctx_func *grctx = gr->func->grctx; struct nvkm_subdev *subdev = &gr->base.engine.subdev; struct nvkm_device *device = subdev->device; struct nvkm_secboot *sb = device->secboot; - int i; int ret = 0; - if (gr->firmware) { - /* load fuc microcode */ - nvkm_mc_unk260(device, 0); - - /* securely-managed falcons must be reset using secure boot */ - if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_FECS)) - ret = nvkm_secboot_reset(sb, NVKM_SECBOOT_FALCON_FECS); - else - gf100_gr_init_fw(gr, 0x409000, &gr->fuc409c, - &gr->fuc409d); - if (ret) - return ret; + /* load fuc microcode */ + nvkm_mc_unk260(device, 0); - if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_GPCCS)) - ret = nvkm_secboot_reset(sb, NVKM_SECBOOT_FALCON_GPCCS); - else - gf100_gr_init_fw(gr, 0x41a000, &gr->fuc41ac, - &gr->fuc41ad); - if (ret) - return ret; + /* securely-managed falcons must be reset using secure boot */ + if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_FECS)) + ret = nvkm_secboot_reset(sb, NVKM_SECBOOT_FALCON_FECS); + else + gf100_gr_init_fw(gr->fecs, &gr->fuc409c, &gr->fuc409d); + if (ret) + return ret; - nvkm_mc_unk260(device, 1); - - /* start both of them running */ - nvkm_wr32(device, 0x409840, 0xffffffff); - nvkm_wr32(device, 0x41a10c, 0x00000000); - nvkm_wr32(device, 0x40910c, 0x00000000); - - if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_GPCCS)) - nvkm_secboot_start(sb, NVKM_SECBOOT_FALCON_GPCCS); - else - nvkm_wr32(device, 0x41a100, 0x00000002); - if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_FECS)) - nvkm_secboot_start(sb, NVKM_SECBOOT_FALCON_FECS); - else - nvkm_wr32(device, 0x409100, 0x00000002); - if (nvkm_msec(device, 2000, - if (nvkm_rd32(device, 0x409800) & 0x00000001) - break; - ) < 0) - return -EBUSY; + if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_GPCCS)) + ret = nvkm_secboot_reset(sb, NVKM_SECBOOT_FALCON_GPCCS); + else + gf100_gr_init_fw(gr->gpccs, &gr->fuc41ac, &gr->fuc41ad); + if (ret) + return ret; + + nvkm_mc_unk260(device, 1); + + /* start both of them running */ + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x41a10c, 0x00000000); + nvkm_wr32(device, 0x40910c, 0x00000000); + + nvkm_falcon_start(gr->gpccs); + nvkm_falcon_start(gr->fecs); + + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x00000001) + break; + ) < 0) + return -EBUSY; - nvkm_wr32(device, 0x409840, 0xffffffff); - nvkm_wr32(device, 0x409500, 0x7fffffff); - nvkm_wr32(device, 0x409504, 0x00000021); + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x7fffffff); + nvkm_wr32(device, 0x409504, 0x00000021); - nvkm_wr32(device, 0x409840, 0xffffffff); - nvkm_wr32(device, 0x409500, 0x00000000); - nvkm_wr32(device, 0x409504, 0x00000010); + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000010); + if (nvkm_msec(device, 2000, + if ((gr->size = nvkm_rd32(device, 0x409800))) + break; + ) < 0) + return -EBUSY; + + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000016); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) + return -EBUSY; + + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000025); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) + return -EBUSY; + + if (device->chipset >= 0xe0) { + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000030); if (nvkm_msec(device, 2000, - if ((gr->size = nvkm_rd32(device, 0x409800))) + if (nvkm_rd32(device, 0x409800)) break; ) < 0) return -EBUSY; - nvkm_wr32(device, 0x409840, 0xffffffff); - nvkm_wr32(device, 0x409500, 0x00000000); - nvkm_wr32(device, 0x409504, 0x00000016); + nvkm_wr32(device, 0x409810, 0xb00095c8); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000031); if (nvkm_msec(device, 2000, if (nvkm_rd32(device, 0x409800)) break; ) < 0) return -EBUSY; - nvkm_wr32(device, 0x409840, 0xffffffff); - nvkm_wr32(device, 0x409500, 0x00000000); - nvkm_wr32(device, 0x409504, 0x00000025); + nvkm_wr32(device, 0x409810, 0x00080420); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000032); if (nvkm_msec(device, 2000, if (nvkm_rd32(device, 0x409800)) break; ) < 0) return -EBUSY; - if (device->chipset >= 0xe0) { - nvkm_wr32(device, 0x409800, 0x00000000); - nvkm_wr32(device, 0x409500, 0x00000001); - nvkm_wr32(device, 0x409504, 0x00000030); - if (nvkm_msec(device, 2000, - if (nvkm_rd32(device, 0x409800)) - break; - ) < 0) - return -EBUSY; - - nvkm_wr32(device, 0x409810, 0xb00095c8); - nvkm_wr32(device, 0x409800, 0x00000000); - nvkm_wr32(device, 0x409500, 0x00000001); - nvkm_wr32(device, 0x409504, 0x00000031); - if (nvkm_msec(device, 2000, - if (nvkm_rd32(device, 0x409800)) - break; - ) < 0) - return -EBUSY; - - nvkm_wr32(device, 0x409810, 0x00080420); - nvkm_wr32(device, 0x409800, 0x00000000); - nvkm_wr32(device, 0x409500, 0x00000001); - nvkm_wr32(device, 0x409504, 0x00000032); - if (nvkm_msec(device, 2000, - if (nvkm_rd32(device, 0x409800)) - break; - ) < 0) - return -EBUSY; + nvkm_wr32(device, 0x409614, 0x00000070); + nvkm_wr32(device, 0x409614, 0x00000770); + nvkm_wr32(device, 0x40802c, 0x00000001); + } - nvkm_wr32(device, 0x409614, 0x00000070); - nvkm_wr32(device, 0x409614, 0x00000770); - nvkm_wr32(device, 0x40802c, 0x00000001); + if (gr->data == NULL) { + int ret = gf100_grctx_generate(gr); + if (ret) { + nvkm_error(subdev, "failed to construct context\n"); + return ret; } + } - if (gr->data == NULL) { - int ret = gf100_grctx_generate(gr); - if (ret) { - nvkm_error(subdev, "failed to construct context\n"); - return ret; - } - } + return 0; +} + +static int +gf100_gr_init_ctxctl_int(struct gf100_gr *gr) +{ + const struct gf100_grctx_func *grctx = gr->func->grctx; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; - return 0; - } else if (!gr->func->fecs.ucode) { return -ENOSYS; } /* load HUB microcode */ nvkm_mc_unk260(device, 0); - nvkm_wr32(device, 0x4091c0, 0x01000000); - for (i = 0; i < gr->func->fecs.ucode->data.size / 4; i++) - nvkm_wr32(device, 0x4091c4, gr->func->fecs.ucode->data.data[i]); - - nvkm_wr32(device, 0x409180, 0x01000000); - for (i = 0; i < gr->func->fecs.ucode->code.size / 4; i++) { - if ((i & 0x3f) == 0) - nvkm_wr32(device, 0x409188, i >> 6); - nvkm_wr32(device, 0x409184, gr->func->fecs.ucode->code.data[i]); - } + nvkm_falcon_load_dmem(gr->fecs, gr->func->fecs.ucode->data.data, 0x0, + gr->func->fecs.ucode->data.size, 0); + nvkm_falcon_load_imem(gr->fecs, gr->func->fecs.ucode->code.data, 0x0, + gr->func->fecs.ucode->code.size, 0, 0, false); /* load GPC microcode */ - nvkm_wr32(device, 0x41a1c0, 0x01000000); - for (i = 0; i < gr->func->gpccs.ucode->data.size / 4; i++) - nvkm_wr32(device, 0x41a1c4, gr->func->gpccs.ucode->data.data[i]); - - nvkm_wr32(device, 0x41a180, 0x01000000); - for (i = 0; i < gr->func->gpccs.ucode->code.size / 4; i++) { - if ((i & 0x3f) == 0) - nvkm_wr32(device, 0x41a188, i >> 6); - nvkm_wr32(device, 0x41a184, gr->func->gpccs.ucode->code.data[i]); - } + nvkm_falcon_load_dmem(gr->gpccs, gr->func->gpccs.ucode->data.data, 0x0, + gr->func->gpccs.ucode->data.size, 0); + nvkm_falcon_load_imem(gr->gpccs, gr->func->gpccs.ucode->code.data, 0x0, + gr->func->gpccs.ucode->code.size, 0, 0, false); nvkm_mc_unk260(device, 1); /* load register lists */ @@ -1636,6 +1630,19 @@ gf100_gr_init_ctxctl(struct gf100_gr *gr) return 0; } +int +gf100_gr_init_ctxctl(struct gf100_gr *gr) +{ + int ret; + + if (gr->firmware) + ret = gf100_gr_init_ctxctl_ext(gr); + else + ret = gf100_gr_init_ctxctl_int(gr); + + return ret; +} + static int gf100_gr_oneinit(struct nvkm_gr *base) { @@ -1701,14 +1708,36 @@ gf100_gr_oneinit(struct nvkm_gr *base) return 0; } -int +static int gf100_gr_init_(struct nvkm_gr *base) { struct gf100_gr *gr = gf100_gr(base); + struct nvkm_subdev *subdev = &base->engine.subdev; + u32 ret; + nvkm_pmu_pgob(gr->base.engine.subdev.device->pmu, false); + + ret = nvkm_falcon_get(gr->fecs, subdev); + if (ret) + return ret; + + ret = nvkm_falcon_get(gr->gpccs, subdev); + if (ret) + return ret; + return gr->func->init(gr); } +static int +gf100_gr_fini_(struct nvkm_gr *base, bool suspend) +{ + struct gf100_gr *gr = gf100_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + nvkm_falcon_put(gr->gpccs, subdev); + nvkm_falcon_put(gr->fecs, subdev); + return 0; +} + void gf100_gr_dtor_fw(struct gf100_gr_fuc *fuc) { @@ -1731,6 +1760,9 @@ gf100_gr_dtor(struct nvkm_gr *base) gr->func->dtor(gr); kfree(gr->data); + nvkm_falcon_del(&gr->gpccs); + nvkm_falcon_del(&gr->fecs); + gf100_gr_dtor_fw(&gr->fuc409c); gf100_gr_dtor_fw(&gr->fuc409d); gf100_gr_dtor_fw(&gr->fuc41ac); @@ -1749,13 +1781,59 @@ gf100_gr_ = { .dtor = gf100_gr_dtor, .oneinit = gf100_gr_oneinit, .init = gf100_gr_init_, + .fini = gf100_gr_fini_, .intr = gf100_gr_intr, .units = gf100_gr_units, .chan_new = gf100_gr_chan_new, .object_get = gf100_gr_object_get, + .chsw_load = gf100_gr_chsw_load, }; int +gf100_gr_ctor_fw_legacy(struct gf100_gr *gr, const char *fwname, + struct gf100_gr_fuc *fuc, int ret) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + const struct firmware *fw; + char f[32]; + + /* see if this firmware has a legacy path */ + if (!strcmp(fwname, "fecs_inst")) + fwname = "fuc409c"; + else if (!strcmp(fwname, "fecs_data")) + fwname = "fuc409d"; + else if (!strcmp(fwname, "gpccs_inst")) + fwname = "fuc41ac"; + else if (!strcmp(fwname, "gpccs_data")) + fwname = "fuc41ad"; + else { + /* nope, let's just return the error we got */ + nvkm_error(subdev, "failed to load %s\n", fwname); + return ret; + } + + /* yes, try to load from the legacy path */ + nvkm_debug(subdev, "%s: falling back to legacy path\n", fwname); + + snprintf(f, sizeof(f), "nouveau/nv%02x_%s", device->chipset, fwname); + ret = request_firmware(&fw, f, device->dev); + if (ret) { + snprintf(f, sizeof(f), "nouveau/%s", fwname); + ret = request_firmware(&fw, f, device->dev); + if (ret) { + nvkm_error(subdev, "failed to load %s\n", fwname); + return ret; + } + } + + fuc->size = fw->size; + fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL); + release_firmware(fw); + return (fuc->data != NULL) ? 0 : -ENOMEM; +} + +int gf100_gr_ctor_fw(struct gf100_gr *gr, const char *fwname, struct gf100_gr_fuc *fuc) { @@ -1765,10 +1843,8 @@ gf100_gr_ctor_fw(struct gf100_gr *gr, const char *fwname, int ret; ret = nvkm_firmware_get(device, fwname, &fw); - if (ret) { - nvkm_error(subdev, "failed to load %s\n", fwname); - return ret; - } + if (ret) + return gf100_gr_ctor_fw_legacy(gr, fwname, fuc, ret); fuc->size = fw->size; fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL); @@ -1780,6 +1856,7 @@ int gf100_gr_ctor(const struct gf100_gr_func *func, struct nvkm_device *device, int index, struct gf100_gr *gr) { + struct nvkm_subdev *subdev = &gr->base.engine.subdev; int ret; gr->func = func; @@ -1792,7 +1869,11 @@ gf100_gr_ctor(const struct gf100_gr_func *func, struct nvkm_device *device, if (ret) return ret; - return 0; + ret = nvkm_falcon_v1_new(subdev, "FECS", 0x409000, &gr->fecs); + if (ret) + return ret; + + return nvkm_falcon_v1_new(subdev, "GPCCS", 0x41a000, &gr->gpccs); } int diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h index 268b8d60ff73..db6ee3b06841 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h @@ -29,6 +29,7 @@ #include <core/gpuobj.h> #include <subdev/ltc.h> #include <subdev/mmu.h> +#include <engine/falcon.h> #define GPC_MAX 32 #define TPC_MAX_PER_GPC 8 @@ -75,6 +76,8 @@ struct gf100_gr { const struct gf100_gr_func *func; struct nvkm_gr base; + struct nvkm_falcon *fecs; + struct nvkm_falcon *gpccs; struct gf100_gr_fuc fuc409c; struct gf100_gr_fuc fuc409d; struct gf100_gr_fuc fuc41ac; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c index 70335f65c51e..0124e468086e 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c @@ -102,7 +102,7 @@ gf117_gr_pack_mmio[] = { #include "fuc/hubgf117.fuc3.h" -struct gf100_gr_ucode +static struct gf100_gr_ucode gf117_gr_fecs_ucode = { .code.data = gf117_grhub_code, .code.size = sizeof(gf117_grhub_code), @@ -112,7 +112,7 @@ gf117_gr_fecs_ucode = { #include "fuc/gpcgf117.fuc3.h" -struct gf100_gr_ucode +static struct gf100_gr_ucode gf117_gr_gpccs_ucode = { .code.data = gf117_grgpc_code, .code.size = sizeof(gf117_grgpc_code), diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c index 45f965f608a7..2c67fac576d1 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c @@ -308,7 +308,7 @@ gm107_gr_init_bios(struct gf100_gr *gr) } } -int +static int gm107_gr_init(struct gf100_gr *gr) { struct nvkm_device *device = gr->base.engine.subdev.device; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c index 2e68919f00b2..c711a55ce392 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c @@ -23,6 +23,8 @@ */ #include "nv50.h" +#include <nvif/class.h> + static const struct nvkm_gr_func gt200_gr = { .init = nv50_gr_init, @@ -31,11 +33,11 @@ gt200_gr = { .tlb_flush = g84_gr_tlb_flush, .units = nv50_gr_units, .sclass = { - { -1, -1, 0x0030, &nv50_gr_object }, - { -1, -1, 0x502d, &nv50_gr_object }, - { -1, -1, 0x5039, &nv50_gr_object }, - { -1, -1, 0x50c0, &nv50_gr_object }, - { -1, -1, 0x8397, &nv50_gr_object }, + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT200_TESLA, &nv50_gr_object }, {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c index 2bf7aac360cc..fa103df32ec7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c @@ -23,6 +23,8 @@ */ #include "nv50.h" +#include <nvif/class.h> + static const struct nvkm_gr_func gt215_gr = { .init = nv50_gr_init, @@ -31,12 +33,12 @@ gt215_gr = { .tlb_flush = g84_gr_tlb_flush, .units = nv50_gr_units, .sclass = { - { -1, -1, 0x0030, &nv50_gr_object }, - { -1, -1, 0x502d, &nv50_gr_object }, - { -1, -1, 0x5039, &nv50_gr_object }, - { -1, -1, 0x50c0, &nv50_gr_object }, - { -1, -1, 0x8597, &nv50_gr_object }, - { -1, -1, 0x85c0, &nv50_gr_object }, + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT214_TESLA, &nv50_gr_object }, + { -1, -1, GT214_COMPUTE, &nv50_gr_object }, {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c index 95d5219faf93..eb1a90644752 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c @@ -23,6 +23,8 @@ */ #include "nv50.h" +#include <nvif/class.h> + static const struct nvkm_gr_func mcp79_gr = { .init = nv50_gr_init, @@ -30,11 +32,11 @@ mcp79_gr = { .chan_new = nv50_gr_chan_new, .units = nv50_gr_units, .sclass = { - { -1, -1, 0x0030, &nv50_gr_object }, - { -1, -1, 0x502d, &nv50_gr_object }, - { -1, -1, 0x5039, &nv50_gr_object }, - { -1, -1, 0x50c0, &nv50_gr_object }, - { -1, -1, 0x8397, &nv50_gr_object }, + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT200_TESLA, &nv50_gr_object }, {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c index 027b58e5976b..c91eb56e9327 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c @@ -23,6 +23,8 @@ */ #include "nv50.h" +#include <nvif/class.h> + static const struct nvkm_gr_func mcp89_gr = { .init = nv50_gr_init, @@ -31,12 +33,12 @@ mcp89_gr = { .tlb_flush = g84_gr_tlb_flush, .units = nv50_gr_units, .sclass = { - { -1, -1, 0x0030, &nv50_gr_object }, - { -1, -1, 0x502d, &nv50_gr_object }, - { -1, -1, 0x5039, &nv50_gr_object }, - { -1, -1, 0x50c0, &nv50_gr_object }, - { -1, -1, 0x85c0, &nv50_gr_object }, - { -1, -1, 0x8697, &nv50_gr_object }, + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT214_COMPUTE, &nv50_gr_object }, + { -1, -1, GT21A_TESLA, &nv50_gr_object }, {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c index f1e15a4d4f64..b4e3c50badc7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c @@ -187,6 +187,7 @@ nv30_gr = { { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ { -1, -1, 0x0397, &nv04_gr_object }, /* rankine */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c index 300f5ed5de0b..e7ed04b935cd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c @@ -123,6 +123,7 @@ nv34_gr = { { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ { -1, -1, 0x0697, &nv04_gr_object }, /* rankine */ {} } diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c index 740df0f52c38..5e8abacbacc6 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c @@ -124,6 +124,7 @@ nv35_gr = { { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ { -1, -1, 0x0497, &nv04_gr_object }, /* rankine */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c index fca67de43f2b..df16ffda1749 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c @@ -27,6 +27,8 @@ #include <core/gpuobj.h> #include <engine/fifo.h> +#include <nvif/class.h> + u64 nv50_gr_units(struct nvkm_gr *gr) { @@ -778,11 +780,11 @@ nv50_gr = { .chan_new = nv50_gr_chan_new, .units = nv50_gr_units, .sclass = { - { -1, -1, 0x0030, &nv50_gr_object }, - { -1, -1, 0x502d, &nv50_gr_object }, - { -1, -1, 0x5039, &nv50_gr_object }, - { -1, -1, 0x5097, &nv50_gr_object }, - { -1, -1, 0x50c0, &nv50_gr_object }, + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_TESLA, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, {} } }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h index d8adcdf6985a..2a52d9f026ec 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h @@ -15,6 +15,7 @@ struct nvkm_gr_func { void *(*dtor)(struct nvkm_gr *); int (*oneinit)(struct nvkm_gr *); int (*init)(struct nvkm_gr *); + int (*fini)(struct nvkm_gr *, bool); void (*intr)(struct nvkm_gr *); void (*tile)(struct nvkm_gr *, int region, struct nvkm_fb_tile *); int (*tlb_flush)(struct nvkm_gr *); @@ -24,6 +25,7 @@ struct nvkm_gr_func { /* Returns chipset-specific counts of units packed into an u64. */ u64 (*units)(struct nvkm_gr *); + bool (*chsw_load)(struct nvkm_gr *); struct nvkm_sclass sclass[]; }; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c index 8616636ad7b4..dde89a4a0f5b 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c @@ -71,7 +71,7 @@ nvkm_perfdom_find(struct nvkm_pm *pm, int di) return NULL; } -struct nvkm_perfsig * +static struct nvkm_perfsig * nvkm_perfsig_find(struct nvkm_pm *pm, u8 di, u8 si, struct nvkm_perfdom **pdom) { struct nvkm_perfdom *dom = *pdom; @@ -699,7 +699,7 @@ nvkm_pm_oclass_get(struct nvkm_oclass *oclass, int index, return 1; } -int +static int nvkm_perfsrc_new(struct nvkm_pm *pm, struct nvkm_perfsig *sig, const struct nvkm_specsrc *spec) { diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c index d2901e9a7808..fe2532ee4145 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c @@ -102,7 +102,7 @@ gf100_pm_gpc[] = { {} }; -const struct nvkm_specdom +static const struct nvkm_specdom gf100_pm_part[] = { { 0xe0, (const struct nvkm_specsig[]) { { 0x0f, "part00_pbfb_00", gf100_pbfb_sources }, diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h b/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h index eca62221f299..4b57f8814560 100644 --- a/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h @@ -1,4 +1,4 @@ -uint32_t g98_sec_data[] = { +static uint32_t g98_sec_data[] = { /* 0x0000: ctx_dma */ /* 0x0000: ctx_dma_query */ 0x00000000, @@ -150,7 +150,7 @@ uint32_t g98_sec_data[] = { 0x00000000, }; -uint32_t g98_sec_code[] = { +static uint32_t g98_sec_code[] = { 0x17f004bd, 0x0010fe35, 0xf10004fe, diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/Kbuild b/drivers/gpu/drm/nouveau/nvkm/falcon/Kbuild new file mode 100644 index 000000000000..584863db9bfc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/falcon/Kbuild @@ -0,0 +1,2 @@ +nvkm-y += nvkm/falcon/base.o +nvkm-y += nvkm/falcon/v1.o diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/base.c b/drivers/gpu/drm/nouveau/nvkm/falcon/base.c new file mode 100644 index 000000000000..4852f313762f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/falcon/base.c @@ -0,0 +1,191 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include <subdev/mc.h> + +void +nvkm_falcon_load_imem(struct nvkm_falcon *falcon, void *data, u32 start, + u32 size, u16 tag, u8 port, bool secure) +{ + if (secure && !falcon->secret) { + nvkm_warn(falcon->user, + "writing with secure tag on a non-secure falcon!\n"); + return; + } + + falcon->func->load_imem(falcon, data, start, size, tag, port, + secure); +} + +void +nvkm_falcon_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start, + u32 size, u8 port) +{ + falcon->func->load_dmem(falcon, data, start, size, port); +} + +void +nvkm_falcon_read_dmem(struct nvkm_falcon *falcon, u32 start, u32 size, u8 port, + void *data) +{ + falcon->func->read_dmem(falcon, start, size, port, data); +} + +void +nvkm_falcon_bind_context(struct nvkm_falcon *falcon, struct nvkm_gpuobj *inst) +{ + if (!falcon->func->bind_context) { + nvkm_error(falcon->user, + "Context binding not supported on this falcon!\n"); + return; + } + + falcon->func->bind_context(falcon, inst); +} + +void +nvkm_falcon_set_start_addr(struct nvkm_falcon *falcon, u32 start_addr) +{ + falcon->func->set_start_addr(falcon, start_addr); +} + +void +nvkm_falcon_start(struct nvkm_falcon *falcon) +{ + falcon->func->start(falcon); +} + +int +nvkm_falcon_enable(struct nvkm_falcon *falcon) +{ + struct nvkm_device *device = falcon->owner->device; + enum nvkm_devidx id = falcon->owner->index; + int ret; + + nvkm_mc_enable(device, id); + ret = falcon->func->enable(falcon); + if (ret) { + nvkm_mc_disable(device, id); + return ret; + } + + return 0; +} + +void +nvkm_falcon_disable(struct nvkm_falcon *falcon) +{ + struct nvkm_device *device = falcon->owner->device; + enum nvkm_devidx id = falcon->owner->index; + + /* already disabled, return or wait_idle will timeout */ + if (!nvkm_mc_enabled(device, id)) + return; + + falcon->func->disable(falcon); + + nvkm_mc_disable(device, id); +} + +int +nvkm_falcon_reset(struct nvkm_falcon *falcon) +{ + nvkm_falcon_disable(falcon); + return nvkm_falcon_enable(falcon); +} + +int +nvkm_falcon_wait_for_halt(struct nvkm_falcon *falcon, u32 ms) +{ + return falcon->func->wait_for_halt(falcon, ms); +} + +int +nvkm_falcon_clear_interrupt(struct nvkm_falcon *falcon, u32 mask) +{ + return falcon->func->clear_interrupt(falcon, mask); +} + +void +nvkm_falcon_put(struct nvkm_falcon *falcon, const struct nvkm_subdev *user) +{ + mutex_lock(&falcon->mutex); + if (falcon->user == user) { + nvkm_debug(falcon->user, "released %s falcon\n", falcon->name); + falcon->user = NULL; + } + mutex_unlock(&falcon->mutex); +} + +int +nvkm_falcon_get(struct nvkm_falcon *falcon, const struct nvkm_subdev *user) +{ + mutex_lock(&falcon->mutex); + if (falcon->user) { + nvkm_error(user, "%s falcon already acquired by %s!\n", + falcon->name, nvkm_subdev_name[falcon->user->index]); + mutex_unlock(&falcon->mutex); + return -EBUSY; + } + + nvkm_debug(user, "acquired %s falcon\n", falcon->name); + falcon->user = user; + mutex_unlock(&falcon->mutex); + return 0; +} + +void +nvkm_falcon_ctor(const struct nvkm_falcon_func *func, + struct nvkm_subdev *subdev, const char *name, u32 addr, + struct nvkm_falcon *falcon) +{ + u32 reg; + + falcon->func = func; + falcon->owner = subdev; + falcon->name = name; + falcon->addr = addr; + mutex_init(&falcon->mutex); + + reg = nvkm_falcon_rd32(falcon, 0x12c); + falcon->version = reg & 0xf; + falcon->secret = (reg >> 4) & 0x3; + falcon->code.ports = (reg >> 8) & 0xf; + falcon->data.ports = (reg >> 12) & 0xf; + + reg = nvkm_falcon_rd32(falcon, 0x108); + falcon->code.limit = (reg & 0x1ff) << 8; + falcon->data.limit = (reg & 0x3fe00) >> 1; + + reg = nvkm_falcon_rd32(falcon, 0xc08); + falcon->debug = (reg >> 20) & 0x1; +} + +void +nvkm_falcon_del(struct nvkm_falcon **pfalcon) +{ + if (*pfalcon) { + kfree(*pfalcon); + *pfalcon = NULL; + } +} diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/priv.h b/drivers/gpu/drm/nouveau/nvkm/falcon/priv.h new file mode 100644 index 000000000000..97b56f759d0b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/falcon/priv.h @@ -0,0 +1,8 @@ +#ifndef __NVKM_FALCON_PRIV_H__ +#define __NVKM_FALCON_PRIV_H__ +#include <engine/falcon.h> + +void +nvkm_falcon_ctor(const struct nvkm_falcon_func *, struct nvkm_subdev *, + const char *, u32, struct nvkm_falcon *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c b/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c new file mode 100644 index 000000000000..b537f111f39c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c @@ -0,0 +1,266 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include <core/gpuobj.h> +#include <core/memory.h> +#include <subdev/timer.h> + +static void +nvkm_falcon_v1_load_imem(struct nvkm_falcon *falcon, void *data, u32 start, + u32 size, u16 tag, u8 port, bool secure) +{ + u8 rem = size % 4; + u32 reg; + int i; + + size -= rem; + + reg = start | BIT(24) | (secure ? BIT(28) : 0); + nvkm_falcon_wr32(falcon, 0x180 + (port * 16), reg); + for (i = 0; i < size / 4; i++) { + /* write new tag every 256B */ + if ((i & 0x3f) == 0) + nvkm_falcon_wr32(falcon, 0x188, tag++); + nvkm_falcon_wr32(falcon, 0x184, ((u32 *)data)[i]); + } + + /* + * If size is not a multiple of 4, mask the last work to ensure garbage + * does not get written + */ + if (rem) { + u32 extra = ((u32 *)data)[i]; + + /* write new tag every 256B */ + if ((i & 0x3f) == 0) + nvkm_falcon_wr32(falcon, 0x188, tag++); + nvkm_falcon_wr32(falcon, 0x184, extra & (BIT(rem * 8) - 1)); + ++i; + } + + /* code must be padded to 0x40 words */ + for (; i & 0x3f; i++) + nvkm_falcon_wr32(falcon, 0x184, 0); +} + +static void +nvkm_falcon_v1_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start, + u32 size, u8 port) +{ + u8 rem = size % 4; + int i; + + size -= rem; + + nvkm_falcon_wr32(falcon, 0x1c0 + (port * 16), start | (0x1 << 24)); + for (i = 0; i < size / 4; i++) + nvkm_falcon_wr32(falcon, 0x1c4, ((u32 *)data)[i]); + + /* + * If size is not a multiple of 4, mask the last work to ensure garbage + * does not get read + */ + if (rem) { + u32 extra = ((u32 *)data)[i]; + + nvkm_falcon_wr32(falcon, 0x1c4, extra & (BIT(rem * 8) - 1)); + } +} + +static void +nvkm_falcon_v1_read_dmem(struct nvkm_falcon *falcon, u32 start, u32 size, + u8 port, void *data) +{ + u8 rem = size % 4; + int i; + + size -= rem; + + nvkm_falcon_wr32(falcon, 0x1c0 + (port * 16), start | (0x1 << 25)); + for (i = 0; i < size / 4; i++) + ((u32 *)data)[i] = nvkm_falcon_rd32(falcon, 0x1c4); + + /* + * If size is not a multiple of 4, mask the last work to ensure garbage + * does not get read + */ + if (rem) { + u32 extra = nvkm_falcon_rd32(falcon, 0x1c4); + + for (i = size; i < size + rem; i++) { + ((u8 *)data)[i] = (u8)(extra & 0xff); + extra >>= 8; + } + } +} + +static void +nvkm_falcon_v1_bind_context(struct nvkm_falcon *falcon, struct nvkm_gpuobj *ctx) +{ + u32 inst_loc; + + /* disable instance block binding */ + if (ctx == NULL) { + nvkm_falcon_wr32(falcon, 0x10c, 0x0); + return; + } + + nvkm_falcon_wr32(falcon, 0x10c, 0x1); + + /* setup apertures - virtual */ + nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_UCODE, 0x4); + nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_VIRT, 0x0); + /* setup apertures - physical */ + nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_PHYS_VID, 0x4); + nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_PHYS_SYS_COH, 0x5); + nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_PHYS_SYS_NCOH, 0x6); + + /* Set context */ + switch (nvkm_memory_target(ctx->memory)) { + case NVKM_MEM_TARGET_VRAM: inst_loc = 0; break; + case NVKM_MEM_TARGET_NCOH: inst_loc = 3; break; + default: + WARN_ON(1); + return; + } + + /* Enable context */ + nvkm_falcon_mask(falcon, 0x048, 0x1, 0x1); + nvkm_falcon_wr32(falcon, 0x480, + ((ctx->addr >> 12) & 0xfffffff) | + (inst_loc << 28) | (1 << 30)); +} + +static void +nvkm_falcon_v1_set_start_addr(struct nvkm_falcon *falcon, u32 start_addr) +{ + nvkm_falcon_wr32(falcon, 0x104, start_addr); +} + +static void +nvkm_falcon_v1_start(struct nvkm_falcon *falcon) +{ + u32 reg = nvkm_falcon_rd32(falcon, 0x100); + + if (reg & BIT(6)) + nvkm_falcon_wr32(falcon, 0x130, 0x2); + else + nvkm_falcon_wr32(falcon, 0x100, 0x2); +} + +static int +nvkm_falcon_v1_wait_for_halt(struct nvkm_falcon *falcon, u32 ms) +{ + struct nvkm_device *device = falcon->owner->device; + int ret; + + ret = nvkm_wait_msec(device, ms, falcon->addr + 0x100, 0x10, 0x10); + if (ret < 0) + return ret; + + return 0; +} + +static int +nvkm_falcon_v1_clear_interrupt(struct nvkm_falcon *falcon, u32 mask) +{ + struct nvkm_device *device = falcon->owner->device; + int ret; + + /* clear interrupt(s) */ + nvkm_falcon_mask(falcon, 0x004, mask, mask); + /* wait until interrupts are cleared */ + ret = nvkm_wait_msec(device, 10, falcon->addr + 0x008, mask, 0x0); + if (ret < 0) + return ret; + + return 0; +} + +static int +falcon_v1_wait_idle(struct nvkm_falcon *falcon) +{ + struct nvkm_device *device = falcon->owner->device; + int ret; + + ret = nvkm_wait_msec(device, 10, falcon->addr + 0x04c, 0xffff, 0x0); + if (ret < 0) + return ret; + + return 0; +} + +static int +nvkm_falcon_v1_enable(struct nvkm_falcon *falcon) +{ + struct nvkm_device *device = falcon->owner->device; + int ret; + + ret = nvkm_wait_msec(device, 10, falcon->addr + 0x10c, 0x6, 0x0); + if (ret < 0) { + nvkm_error(falcon->user, "Falcon mem scrubbing timeout\n"); + return ret; + } + + ret = falcon_v1_wait_idle(falcon); + if (ret) + return ret; + + /* enable IRQs */ + nvkm_falcon_wr32(falcon, 0x010, 0xff); + + return 0; +} + +static void +nvkm_falcon_v1_disable(struct nvkm_falcon *falcon) +{ + /* disable IRQs and wait for any previous code to complete */ + nvkm_falcon_wr32(falcon, 0x014, 0xff); + falcon_v1_wait_idle(falcon); +} + +static const struct nvkm_falcon_func +nvkm_falcon_v1 = { + .load_imem = nvkm_falcon_v1_load_imem, + .load_dmem = nvkm_falcon_v1_load_dmem, + .read_dmem = nvkm_falcon_v1_read_dmem, + .bind_context = nvkm_falcon_v1_bind_context, + .start = nvkm_falcon_v1_start, + .wait_for_halt = nvkm_falcon_v1_wait_for_halt, + .clear_interrupt = nvkm_falcon_v1_clear_interrupt, + .enable = nvkm_falcon_v1_enable, + .disable = nvkm_falcon_v1_disable, + .set_start_addr = nvkm_falcon_v1_set_start_addr, +}; + +int +nvkm_falcon_v1_new(struct nvkm_subdev *owner, const char *name, u32 addr, + struct nvkm_falcon **pfalcon) +{ + struct nvkm_falcon *falcon; + if (!(falcon = *pfalcon = kzalloc(sizeof(*falcon), GFP_KERNEL))) + return -ENOMEM; + nvkm_falcon_ctor(&nvkm_falcon_v1, owner, name, addr, falcon); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bar/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bar/nv50.c index 370dcd8ff7b5..6eff637ac301 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bar/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bar/nv50.c @@ -84,7 +84,7 @@ nv50_bar_oneinit(struct nvkm_bar *base) start = 0x0100000000ULL; limit = start + device->func->resource_size(device, 3); - ret = nvkm_vm_new(device, start, limit, start, &bar3_lock, &vm); + ret = nvkm_vm_new(device, start, limit - start, start, &bar3_lock, &vm); if (ret) return ret; @@ -117,7 +117,7 @@ nv50_bar_oneinit(struct nvkm_bar *base) start = 0x0000000000ULL; limit = start + device->func->resource_size(device, 1); - ret = nvkm_vm_new(device, start, limit--, start, &bar1_lock, &vm); + ret = nvkm_vm_new(device, start, limit-- - start, start, &bar1_lock, &vm); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/Kbuild index dbcb0ef21587..6b4f1e06a38f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/Kbuild @@ -19,6 +19,7 @@ nvkm-y += nvkm/subdev/bios/pcir.o nvkm-y += nvkm/subdev/bios/perf.o nvkm-y += nvkm/subdev/bios/pll.o nvkm-y += nvkm/subdev/bios/pmu.o +nvkm-y += nvkm/subdev/bios/power_budget.o nvkm-y += nvkm/subdev/bios/ramcfg.o nvkm-y += nvkm/subdev/bios/rammap.o nvkm-y += nvkm/subdev/bios/shadow.o @@ -31,6 +32,7 @@ nvkm-y += nvkm/subdev/bios/timing.o nvkm-y += nvkm/subdev/bios/therm.o nvkm-y += nvkm/subdev/bios/vmap.o nvkm-y += nvkm/subdev/bios/volt.o +nvkm-y += nvkm/subdev/bios/vpstate.o nvkm-y += nvkm/subdev/bios/xpio.o nvkm-y += nvkm/subdev/bios/M0203.o nvkm-y += nvkm/subdev/bios/M0205.o diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/boost.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/boost.c index 3756ec91a88d..eaf74eb72983 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/boost.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/boost.c @@ -25,16 +25,16 @@ #include <subdev/bios/bit.h> #include <subdev/bios/boost.h> -u16 +u32 nvbios_boostTe(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, u8 *snr, u8 *ssz) { struct bit_entry bit_P; - u16 boost = 0x0000; + u32 boost = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 2) - boost = nvbios_rd16(bios, bit_P.offset + 0x30); + boost = nvbios_rd32(bios, bit_P.offset + 0x30); if (boost) { *ver = nvbios_rd08(bios, boost + 0); @@ -52,15 +52,15 @@ nvbios_boostTe(struct nvkm_bios *bios, } } - return 0x0000; + return 0; } -u16 +u32 nvbios_boostEe(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { u8 snr, ssz; - u16 data = nvbios_boostTe(bios, ver, hdr, cnt, len, &snr, &ssz); + u32 data = nvbios_boostTe(bios, ver, hdr, cnt, len, &snr, &ssz); if (data && idx < *cnt) { data = data + *hdr + (idx * (*len + (snr * ssz))); *hdr = *len; @@ -68,14 +68,14 @@ nvbios_boostEe(struct nvkm_bios *bios, int idx, *len = ssz; return data; } - return 0x0000; + return 0; } -u16 +u32 nvbios_boostEp(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, struct nvbios_boostE *info) { - u16 data = nvbios_boostEe(bios, idx, ver, hdr, cnt, len); + u32 data = nvbios_boostEe(bios, idx, ver, hdr, cnt, len); memset(info, 0x00, sizeof(*info)); if (data) { info->pstate = (nvbios_rd16(bios, data + 0x00) & 0x01e0) >> 5; @@ -85,7 +85,7 @@ nvbios_boostEp(struct nvkm_bios *bios, int idx, return data; } -u16 +u32 nvbios_boostEm(struct nvkm_bios *bios, u8 pstate, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, struct nvbios_boostE *info) { @@ -97,21 +97,21 @@ nvbios_boostEm(struct nvkm_bios *bios, u8 pstate, return data; } -u16 +u32 nvbios_boostSe(struct nvkm_bios *bios, int idx, - u16 data, u8 *ver, u8 *hdr, u8 cnt, u8 len) + u32 data, u8 *ver, u8 *hdr, u8 cnt, u8 len) { if (data && idx < cnt) { data = data + *hdr + (idx * len); *hdr = len; return data; } - return 0x0000; + return 0; } -u16 +u32 nvbios_boostSp(struct nvkm_bios *bios, int idx, - u16 data, u8 *ver, u8 *hdr, u8 cnt, u8 len, + u32 data, u8 *ver, u8 *hdr, u8 cnt, u8 len, struct nvbios_boostS *info) { data = nvbios_boostSe(bios, idx, data, ver, hdr, cnt, len); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/cstep.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/cstep.c index 32e01624a162..5063382d8a6c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/cstep.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/cstep.c @@ -25,16 +25,16 @@ #include <subdev/bios/bit.h> #include <subdev/bios/cstep.h> -u16 +u32 nvbios_cstepTe(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, u8 *xnr, u8 *xsz) { struct bit_entry bit_P; - u16 cstep = 0x0000; + u32 cstep = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 2) - cstep = nvbios_rd16(bios, bit_P.offset + 0x34); + cstep = nvbios_rd32(bios, bit_P.offset + 0x34); if (cstep) { *ver = nvbios_rd08(bios, cstep + 0); @@ -52,27 +52,27 @@ nvbios_cstepTe(struct nvkm_bios *bios, } } - return 0x0000; + return 0; } -u16 +u32 nvbios_cstepEe(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr) { u8 cnt, len, xnr, xsz; - u16 data = nvbios_cstepTe(bios, ver, hdr, &cnt, &len, &xnr, &xsz); + u32 data = nvbios_cstepTe(bios, ver, hdr, &cnt, &len, &xnr, &xsz); if (data && idx < cnt) { data = data + *hdr + (idx * len); *hdr = len; return data; } - return 0x0000; + return 0; } -u16 +u32 nvbios_cstepEp(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, struct nvbios_cstepE *info) { - u16 data = nvbios_cstepEe(bios, idx, ver, hdr); + u32 data = nvbios_cstepEe(bios, idx, ver, hdr); memset(info, 0x00, sizeof(*info)); if (data) { info->pstate = (nvbios_rd16(bios, data + 0x00) & 0x01e0) >> 5; @@ -81,7 +81,7 @@ nvbios_cstepEp(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, return data; } -u16 +u32 nvbios_cstepEm(struct nvkm_bios *bios, u8 pstate, u8 *ver, u8 *hdr, struct nvbios_cstepE *info) { @@ -93,24 +93,24 @@ nvbios_cstepEm(struct nvkm_bios *bios, u8 pstate, u8 *ver, u8 *hdr, return data; } -u16 +u32 nvbios_cstepXe(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr) { u8 cnt, len, xnr, xsz; - u16 data = nvbios_cstepTe(bios, ver, hdr, &cnt, &len, &xnr, &xsz); + u32 data = nvbios_cstepTe(bios, ver, hdr, &cnt, &len, &xnr, &xsz); if (data && idx < xnr) { data = data + *hdr + (cnt * len) + (idx * xsz); *hdr = xsz; return data; } - return 0x0000; + return 0; } -u16 +u32 nvbios_cstepXp(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, struct nvbios_cstepX *info) { - u16 data = nvbios_cstepXe(bios, idx, ver, hdr); + u32 data = nvbios_cstepXe(bios, idx, ver, hdr); memset(info, 0x00, sizeof(*info)); if (data) { info->freq = nvbios_rd16(bios, data + 0x00) * 1000; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/dp.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/dp.c index d89e78c4e689..972370ed36f0 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/dp.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/dp.c @@ -207,8 +207,11 @@ nvbios_dpcfg_match(struct nvkm_bios *bios, u16 outp, u8 pc, u8 vs, u8 pe, if (*ver >= 0x30) { const u8 vsoff[] = { 0, 4, 7, 9 }; idx = (pc * 10) + vsoff[vs] + pe; - if (*ver >= 0x40 && *hdr >= 0x12) + if (*ver >= 0x40 && *ver <= 0x41 && *hdr >= 0x12) idx += nvbios_rd08(bios, outp + 0x11) * 40; + else + if (*ver >= 0x42) + idx += nvbios_rd08(bios, outp + 0x11) * 10; } else { while ((data = nvbios_dpcfg_entry(bios, outp, ++idx, ver, hdr, cnt, len))) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/fan.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/fan.c index 80fed7e78dcb..456f9ea920dc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/fan.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/fan.c @@ -25,15 +25,15 @@ #include <subdev/bios/bit.h> #include <subdev/bios/fan.h> -u16 +static u32 nvbios_fan_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { struct bit_entry bit_P; - u16 fan = 0x0000; + u32 fan = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 2 && bit_P.length >= 0x5a) - fan = nvbios_rd16(bios, bit_P.offset + 0x58); + fan = nvbios_rd32(bios, bit_P.offset + 0x58); if (fan) { *ver = nvbios_rd08(bios, fan + 0); @@ -49,25 +49,25 @@ nvbios_fan_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) } } - return 0x0000; + return 0; } -u16 +static u32 nvbios_fan_entry(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { - u16 data = nvbios_fan_table(bios, ver, hdr, cnt, len); + u32 data = nvbios_fan_table(bios, ver, hdr, cnt, len); if (data && idx < *cnt) return data + *hdr + (idx * (*len)); - return 0x0000; + return 0; } -u16 +u32 nvbios_fan_parse(struct nvkm_bios *bios, struct nvbios_therm_fan *fan) { u8 ver, hdr, cnt, len; - u16 data = nvbios_fan_entry(bios, 0, &ver, &hdr, &cnt, &len); + u32 data = nvbios_fan_entry(bios, 0, &ver, &hdr, &cnt, &len); if (data) { u8 type = nvbios_rd08(bios, data + 0x00); switch (type) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/iccsense.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/iccsense.c index 084328028af1..3953d11844ea 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/iccsense.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/iccsense.c @@ -23,20 +23,21 @@ */ #include <subdev/bios.h> #include <subdev/bios/bit.h> +#include <subdev/bios/extdev.h> #include <subdev/bios/iccsense.h> -static u16 +static u32 nvbios_iccsense_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { struct bit_entry bit_P; - u16 iccsense; + u32 iccsense; if (bit_entry(bios, 'P', &bit_P) || bit_P.version != 2 || bit_P.length < 0x2c) return 0; - iccsense = nvbios_rd16(bios, bit_P.offset + 0x28); + iccsense = nvbios_rd32(bios, bit_P.offset + 0x28); if (!iccsense) return 0; @@ -60,7 +61,7 @@ nvbios_iccsense_parse(struct nvkm_bios *bios, struct nvbios_iccsense *iccsense) { struct nvkm_subdev *subdev = &bios->subdev; u8 ver, hdr, cnt, len, i; - u16 table, entry; + u32 table, entry; table = nvbios_iccsense_table(bios, &ver, &hdr, &cnt, &len); if (!table || !cnt) @@ -77,23 +78,47 @@ nvbios_iccsense_parse(struct nvkm_bios *bios, struct nvbios_iccsense *iccsense) return -ENOMEM; for (i = 0; i < cnt; ++i) { + struct nvbios_extdev_func extdev; struct pwr_rail_t *rail = &iccsense->rail[i]; + u8 res_start = 0; + int r; + entry = table + hdr + i * len; switch(ver) { case 0x10: rail->mode = nvbios_rd08(bios, entry + 0x1); rail->extdev_id = nvbios_rd08(bios, entry + 0x2); - rail->resistor_mohm = nvbios_rd08(bios, entry + 0x3); - rail->rail = nvbios_rd08(bios, entry + 0x4); + res_start = 0x3; break; case 0x20: rail->mode = nvbios_rd08(bios, entry); rail->extdev_id = nvbios_rd08(bios, entry + 0x1); - rail->resistor_mohm = nvbios_rd08(bios, entry + 0x5); - rail->rail = nvbios_rd08(bios, entry + 0x6); + res_start = 0x5; + break; + }; + + if (nvbios_extdev_parse(bios, rail->extdev_id, &extdev)) + continue; + + switch (extdev.type) { + case NVBIOS_EXTDEV_INA209: + case NVBIOS_EXTDEV_INA219: + rail->resistor_count = 1; + break; + case NVBIOS_EXTDEV_INA3221: + rail->resistor_count = 3; + break; + default: + rail->resistor_count = 0; break; }; + + for (r = 0; r < rail->resistor_count; ++r) { + rail->resistors[r].mohm = nvbios_rd08(bios, entry + res_start + r * 2); + rail->resistors[r].enabled = !(nvbios_rd08(bios, entry + res_start + r * 2 + 1) & 0x40); + } + rail->config = nvbios_rd16(bios, entry + res_start + rail->resistor_count * 2); } return 0; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/mxm.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/mxm.c index 3ddf0939ded3..994cc2d7759b 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/mxm.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/mxm.c @@ -81,7 +81,7 @@ mxm_sor_map(struct nvkm_bios *bios, u8 conn) u16 map = nvbios_rd16(bios, mxm + 4); if (map) { ver = nvbios_rd08(bios, map); - if (ver == 0x10) { + if (ver == 0x10 || ver == 0x11) { if (conn < nvbios_rd08(bios, map + 3)) { map += nvbios_rd08(bios, map + 1); map += conn; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/perf.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/perf.c index 636bfb665bb9..c3068358f695 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/perf.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/perf.c @@ -26,16 +26,16 @@ #include <subdev/bios/perf.h> #include <subdev/pci.h> -u16 +u32 nvbios_perf_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, u8 *snr, u8 *ssz) { struct bit_entry bit_P; - u16 perf = 0x0000; + u32 perf = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version <= 2) { - perf = nvbios_rd16(bios, bit_P.offset + 0); + perf = nvbios_rd32(bios, bit_P.offset + 0); if (perf) { *ver = nvbios_rd08(bios, perf + 0); *hdr = nvbios_rd08(bios, perf + 1); @@ -72,15 +72,15 @@ nvbios_perf_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, } } - return 0x0000; + return 0; } -u16 +u32 nvbios_perf_entry(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { u8 snr, ssz; - u16 perf = nvbios_perf_table(bios, ver, hdr, cnt, len, &snr, &ssz); + u32 perf = nvbios_perf_table(bios, ver, hdr, cnt, len, &snr, &ssz); if (perf && idx < *cnt) { perf = perf + *hdr + (idx * (*len + (snr * ssz))); *hdr = *len; @@ -88,14 +88,14 @@ nvbios_perf_entry(struct nvkm_bios *bios, int idx, *len = ssz; return perf; } - return 0x0000; + return 0; } -u16 +u32 nvbios_perfEp(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, struct nvbios_perfE *info) { - u16 perf = nvbios_perf_entry(bios, idx, ver, hdr, cnt, len); + u32 perf = nvbios_perf_entry(bios, idx, ver, hdr, cnt, len); memset(info, 0x00, sizeof(*info)); info->pstate = nvbios_rd08(bios, perf + 0x00); switch (!!perf * *ver) { @@ -163,7 +163,7 @@ nvbios_perfEp(struct nvkm_bios *bios, int idx, info->pcie_width = 0xff; break; default: - return 0x0000; + return 0; } return perf; } @@ -202,7 +202,7 @@ nvbios_perf_fan_parse(struct nvkm_bios *bios, struct nvbios_perf_fan *fan) { u8 ver, hdr, cnt, len, snr, ssz; - u16 perf = nvbios_perf_table(bios, &ver, &hdr, &cnt, &len, &snr, &ssz); + u32 perf = nvbios_perf_table(bios, &ver, &hdr, &cnt, &len, &snr, &ssz); if (!perf) return -ENODEV; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/power_budget.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/power_budget.c new file mode 100644 index 000000000000..617bfffce4ad --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/power_budget.c @@ -0,0 +1,126 @@ +/* + * Copyright 2016 Karol Herbst + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Karol Herbst + */ +#include <subdev/bios.h> +#include <subdev/bios/bit.h> +#include <subdev/bios/power_budget.h> + +static u32 +nvbios_power_budget_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, + u8 *len) +{ + struct bit_entry bit_P; + u32 power_budget; + + if (bit_entry(bios, 'P', &bit_P) || bit_P.version != 2 || + bit_P.length < 0x2c) + return 0; + + power_budget = nvbios_rd32(bios, bit_P.offset + 0x2c); + if (!power_budget) + return 0; + + *ver = nvbios_rd08(bios, power_budget); + switch (*ver) { + case 0x20: + case 0x30: + *hdr = nvbios_rd08(bios, power_budget + 0x1); + *len = nvbios_rd08(bios, power_budget + 0x2); + *cnt = nvbios_rd08(bios, power_budget + 0x3); + return power_budget; + default: + break; + } + + return 0; +} + +int +nvbios_power_budget_header(struct nvkm_bios *bios, + struct nvbios_power_budget *budget) +{ + struct nvkm_subdev *subdev = &bios->subdev; + u8 ver, hdr, cnt, len, cap_entry; + u32 header; + + if (!bios || !budget) + return -EINVAL; + + header = nvbios_power_budget_table(bios, &ver, &hdr, &cnt, &len); + if (!header || !cnt) + return -ENODEV; + + switch (ver) { + case 0x20: + cap_entry = nvbios_rd08(bios, header + 0x9); + break; + case 0x30: + cap_entry = nvbios_rd08(bios, header + 0xa); + break; + default: + cap_entry = 0xff; + } + + if (cap_entry >= cnt && cap_entry != 0xff) { + nvkm_warn(subdev, + "invalid cap_entry in power budget table found\n"); + budget->cap_entry = 0xff; + return -EINVAL; + } + + budget->offset = header; + budget->ver = ver; + budget->hlen = hdr; + budget->elen = len; + budget->ecount = cnt; + + budget->cap_entry = cap_entry; + + return 0; +} + +int +nvbios_power_budget_entry(struct nvkm_bios *bios, + struct nvbios_power_budget *budget, + u8 idx, struct nvbios_power_budget_entry *entry) +{ + u32 entry_offset; + + if (!bios || !budget || !budget->offset || idx >= budget->ecount + || !entry) + return -EINVAL; + + entry_offset = budget->offset + budget->hlen + idx * budget->elen; + + if (budget->ver >= 0x20) { + entry->min_w = nvbios_rd32(bios, entry_offset + 0x2); + entry->avg_w = nvbios_rd32(bios, entry_offset + 0x6); + entry->max_w = nvbios_rd32(bios, entry_offset + 0xa); + } else { + entry->min_w = 0; + entry->max_w = nvbios_rd32(bios, entry_offset + 0x2); + entry->avg_w = entry->max_w; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/priv.h index 212800ecdce9..7d1d3c6b4b72 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/priv.h @@ -12,6 +12,7 @@ struct nvbios_source { bool rw; bool ignore_checksum; bool no_pcir; + bool require_checksum; }; int nvbios_extend(struct nvkm_bios *, u32 length); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadow.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadow.c index b2557e87afdd..7deb81b6dbac 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadow.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadow.c @@ -86,9 +86,12 @@ shadow_image(struct nvkm_bios *bios, int idx, u32 offset, struct shadow *mthd) nvbios_checksum(&bios->data[image.base], image.size)) { nvkm_debug(subdev, "%08x: checksum failed\n", image.base); - if (mthd->func->rw) + if (!mthd->func->require_checksum) { + if (mthd->func->rw) + score += 1; score += 1; - score += 1; + } else + return 0; } else { score += 3; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadowacpi.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadowacpi.c index 8fecb5ff22a0..06572f8ce914 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadowacpi.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/shadowacpi.c @@ -99,6 +99,7 @@ nvbios_acpi_fast = { .init = acpi_init, .read = acpi_read_fast, .rw = false, + .require_checksum = true, }; const struct nvbios_source diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/therm.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/therm.c index a54cfec0550d..5babc5a7c7d5 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/therm.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/therm.c @@ -25,17 +25,17 @@ #include <subdev/bios/bit.h> #include <subdev/bios/therm.h> -static u16 +static u32 therm_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *len, u8 *cnt) { struct bit_entry bit_P; - u16 therm = 0; + u32 therm = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 1) - therm = nvbios_rd16(bios, bit_P.offset + 12); + therm = nvbios_rd32(bios, bit_P.offset + 12); else if (bit_P.version == 2) - therm = nvbios_rd16(bios, bit_P.offset + 16); + therm = nvbios_rd32(bios, bit_P.offset + 16); else nvkm_error(&bios->subdev, "unknown offset for thermal in BIT P %d\n", @@ -44,7 +44,7 @@ therm_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *len, u8 *cnt) /* exit now if we haven't found the thermal table */ if (!therm) - return 0x0000; + return 0; *ver = nvbios_rd08(bios, therm + 0); *hdr = nvbios_rd08(bios, therm + 1); @@ -53,14 +53,14 @@ therm_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *len, u8 *cnt) return therm + nvbios_rd08(bios, therm + 1); } -static u16 +static u32 nvbios_therm_entry(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len) { u8 hdr, cnt; - u16 therm = therm_table(bios, ver, &hdr, len, &cnt); + u32 therm = therm_table(bios, ver, &hdr, len, &cnt); if (therm && idx < cnt) return therm + idx * *len; - return 0x0000; + return 0; } int @@ -70,7 +70,7 @@ nvbios_therm_sensor_parse(struct nvkm_bios *bios, { s8 thrs_section, sensor_section, offset; u8 ver, len, i; - u16 entry; + u32 entry; /* we only support the core domain for now */ if (domain != NVBIOS_THERM_DOMAIN_CORE) @@ -154,7 +154,7 @@ nvbios_therm_fan_parse(struct nvkm_bios *bios, struct nvbios_therm_fan *fan) { struct nvbios_therm_trip_point *cur_trip = NULL; u8 ver, len, i; - u16 entry; + u32 entry; uint8_t duty_lut[] = { 0, 0, 25, 0, 40, 0, 50, 0, 75, 0, 85, 0, 100, 0, 100, 0 }; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/timing.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/timing.c index 99f6432ac0af..7e83c3985020 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/timing.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/timing.c @@ -25,19 +25,19 @@ #include <subdev/bios/bit.h> #include <subdev/bios/timing.h> -u16 +u32 nvbios_timingTe(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, u8 *snr, u8 *ssz) { struct bit_entry bit_P; - u16 timing = 0x0000; + u32 timing = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 1) - timing = nvbios_rd16(bios, bit_P.offset + 4); + timing = nvbios_rd32(bios, bit_P.offset + 4); else if (bit_P.version == 2) - timing = nvbios_rd16(bios, bit_P.offset + 8); + timing = nvbios_rd32(bios, bit_P.offset + 8); if (timing) { *ver = nvbios_rd08(bios, timing + 0); @@ -62,15 +62,15 @@ nvbios_timingTe(struct nvkm_bios *bios, } } - return 0x0000; + return 0; } -u16 +u32 nvbios_timingEe(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { u8 snr, ssz; - u16 timing = nvbios_timingTe(bios, ver, hdr, cnt, len, &snr, &ssz); + u32 timing = nvbios_timingTe(bios, ver, hdr, cnt, len, &snr, &ssz); if (timing && idx < *cnt) { timing += *hdr + idx * (*len + (snr * ssz)); *hdr = *len; @@ -78,14 +78,14 @@ nvbios_timingEe(struct nvkm_bios *bios, int idx, *len = ssz; return timing; } - return 0x0000; + return 0; } -u16 +u32 nvbios_timingEp(struct nvkm_bios *bios, int idx, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, struct nvbios_ramcfg *p) { - u16 data = nvbios_timingEe(bios, idx, ver, hdr, cnt, len), temp; + u32 data = nvbios_timingEe(bios, idx, ver, hdr, cnt, len), temp; p->timing_ver = *ver; p->timing_hdr = *hdr; switch (!!data * *ver) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vmap.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vmap.c index 2f13db745948..c228ca15fa3b 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vmap.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vmap.c @@ -25,15 +25,15 @@ #include <subdev/bios/bit.h> #include <subdev/bios/vmap.h> -u16 +u32 nvbios_vmap_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { struct bit_entry bit_P; - u16 vmap = 0x0000; + u32 vmap = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 2) { - vmap = nvbios_rd16(bios, bit_P.offset + 0x20); + vmap = nvbios_rd32(bios, bit_P.offset + 0x20); if (vmap) { *ver = nvbios_rd08(bios, vmap + 0); switch (*ver) { @@ -50,40 +50,50 @@ nvbios_vmap_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) } } - return 0x0000; + return 0; } -u16 +u32 nvbios_vmap_parse(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, struct nvbios_vmap *info) { - u16 vmap = nvbios_vmap_table(bios, ver, hdr, cnt, len); + u32 vmap = nvbios_vmap_table(bios, ver, hdr, cnt, len); memset(info, 0x00, sizeof(*info)); switch (!!vmap * *ver) { case 0x10: + info->max0 = 0xff; + info->max1 = 0xff; + info->max2 = 0xff; + break; case 0x20: + info->max0 = nvbios_rd08(bios, vmap + 0x7); + info->max1 = nvbios_rd08(bios, vmap + 0x8); + if (*len >= 0xc) + info->max2 = nvbios_rd08(bios, vmap + 0xc); + else + info->max2 = 0xff; break; } return vmap; } -u16 +u32 nvbios_vmap_entry(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len) { u8 hdr, cnt; - u16 vmap = nvbios_vmap_table(bios, ver, &hdr, &cnt, len); + u32 vmap = nvbios_vmap_table(bios, ver, &hdr, &cnt, len); if (vmap && idx < cnt) { vmap = vmap + hdr + (idx * *len); return vmap; } - return 0x0000; + return 0; } -u16 +u32 nvbios_vmap_entry_parse(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len, struct nvbios_vmap_entry *info) { - u16 vmap = nvbios_vmap_entry(bios, idx, ver, len); + u32 vmap = nvbios_vmap_entry(bios, idx, ver, len); memset(info, 0x00, sizeof(*info)); switch (!!vmap * *ver) { case 0x10: @@ -95,7 +105,7 @@ nvbios_vmap_entry_parse(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len, info->arg[2] = nvbios_rd32(bios, vmap + 0x10); break; case 0x20: - info->unk0 = nvbios_rd08(bios, vmap + 0x00); + info->mode = nvbios_rd08(bios, vmap + 0x00); info->link = nvbios_rd08(bios, vmap + 0x01); info->min = nvbios_rd32(bios, vmap + 0x02); info->max = nvbios_rd32(bios, vmap + 0x06); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/volt.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/volt.c index 6e0a33648be9..a7797a9e9cbc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/volt.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/volt.c @@ -25,18 +25,18 @@ #include <subdev/bios/bit.h> #include <subdev/bios/volt.h> -u16 +u32 nvbios_volt_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) { struct bit_entry bit_P; - u16 volt = 0x0000; + u32 volt = 0; if (!bit_entry(bios, 'P', &bit_P)) { if (bit_P.version == 2) - volt = nvbios_rd16(bios, bit_P.offset + 0x0c); + volt = nvbios_rd32(bios, bit_P.offset + 0x0c); else if (bit_P.version == 1) - volt = nvbios_rd16(bios, bit_P.offset + 0x10); + volt = nvbios_rd32(bios, bit_P.offset + 0x10); if (volt) { *ver = nvbios_rd08(bios, volt + 0); @@ -62,33 +62,37 @@ nvbios_volt_table(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len) } } - return 0x0000; + return 0; } -u16 +u32 nvbios_volt_parse(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, struct nvbios_volt *info) { - u16 volt = nvbios_volt_table(bios, ver, hdr, cnt, len); + u32 volt = nvbios_volt_table(bios, ver, hdr, cnt, len); memset(info, 0x00, sizeof(*info)); switch (!!volt * *ver) { case 0x12: info->type = NVBIOS_VOLT_GPIO; info->vidmask = nvbios_rd08(bios, volt + 0x04); + info->ranged = false; break; case 0x20: info->type = NVBIOS_VOLT_GPIO; info->vidmask = nvbios_rd08(bios, volt + 0x05); + info->ranged = false; break; case 0x30: info->type = NVBIOS_VOLT_GPIO; info->vidmask = nvbios_rd08(bios, volt + 0x04); + info->ranged = false; break; case 0x40: info->type = NVBIOS_VOLT_GPIO; info->base = nvbios_rd32(bios, volt + 0x04); info->step = nvbios_rd16(bios, volt + 0x08); info->vidmask = nvbios_rd08(bios, volt + 0x0b); + info->ranged = true; /* XXX: find the flag byte */ /*XXX*/ info->min = 0; info->max = info->base; @@ -104,32 +108,34 @@ nvbios_volt_parse(struct nvkm_bios *bios, u8 *ver, u8 *hdr, u8 *cnt, u8 *len, info->pwm_freq = nvbios_rd32(bios, volt + 0x5) / 1000; info->pwm_range = nvbios_rd32(bios, volt + 0x16); } else { - info->type = NVBIOS_VOLT_GPIO; - info->vidmask = nvbios_rd08(bios, volt + 0x06); - info->step = nvbios_rd16(bios, volt + 0x16); + info->type = NVBIOS_VOLT_GPIO; + info->vidmask = nvbios_rd08(bios, volt + 0x06); + info->step = nvbios_rd16(bios, volt + 0x16); + info->ranged = + !!(nvbios_rd08(bios, volt + 0x4) & 0x2); } break; } return volt; } -u16 +u32 nvbios_volt_entry(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len) { u8 hdr, cnt; - u16 volt = nvbios_volt_table(bios, ver, &hdr, &cnt, len); + u32 volt = nvbios_volt_table(bios, ver, &hdr, &cnt, len); if (volt && idx < cnt) { volt = volt + hdr + (idx * *len); return volt; } - return 0x0000; + return 0; } -u16 +u32 nvbios_volt_entry_parse(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len, struct nvbios_volt_entry *info) { - u16 volt = nvbios_volt_entry(bios, idx, ver, len); + u32 volt = nvbios_volt_entry(bios, idx, ver, len); memset(info, 0x00, sizeof(*info)); switch (!!volt * *ver) { case 0x12: @@ -142,7 +148,10 @@ nvbios_volt_entry_parse(struct nvkm_bios *bios, int idx, u8 *ver, u8 *len, info->vid = nvbios_rd08(bios, volt + 0x01) >> 2; break; case 0x40: + break; case 0x50: + info->voltage = nvbios_rd32(bios, volt) & 0x001fffff; + info->vid = (nvbios_rd32(bios, volt) >> 23) & 0xff; break; } return volt; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vpstate.c b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vpstate.c new file mode 100644 index 000000000000..f199270163d2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/bios/vpstate.c @@ -0,0 +1,82 @@ +/* + * Copyright 2016 Karol Herbst + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Karol Herbst + */ +#include <subdev/bios.h> +#include <subdev/bios/bit.h> +#include <subdev/bios/vpstate.h> + +static u32 +nvbios_vpstate_offset(struct nvkm_bios *b) +{ + struct bit_entry bit_P; + + if (!bit_entry(b, 'P', &bit_P)) { + if (bit_P.version == 2) + return nvbios_rd32(b, bit_P.offset + 0x38); + } + + return 0x0000; +} + +int +nvbios_vpstate_parse(struct nvkm_bios *b, struct nvbios_vpstate_header *h) +{ + if (!h) + return -EINVAL; + + h->offset = nvbios_vpstate_offset(b); + if (!h->offset) + return -ENODEV; + + h->version = nvbios_rd08(b, h->offset); + switch (h->version) { + case 0x10: + h->hlen = nvbios_rd08(b, h->offset + 0x1); + h->elen = nvbios_rd08(b, h->offset + 0x2); + h->slen = nvbios_rd08(b, h->offset + 0x3); + h->scount = nvbios_rd08(b, h->offset + 0x4); + h->ecount = nvbios_rd08(b, h->offset + 0x5); + + h->base_id = nvbios_rd08(b, h->offset + 0x0f); + h->boost_id = nvbios_rd08(b, h->offset + 0x10); + h->tdp_id = nvbios_rd08(b, h->offset + 0x11); + return 0; + default: + return -EINVAL; + } +} + +int +nvbios_vpstate_entry(struct nvkm_bios *b, struct nvbios_vpstate_header *h, + u8 idx, struct nvbios_vpstate_entry *e) +{ + u32 offset; + + if (!e || !h || idx > h->ecount) + return -EINVAL; + + offset = h->offset + h->hlen + idx * (h->elen + (h->slen * h->scount)); + e->pstate = nvbios_rd08(b, offset); + e->clock_mhz = nvbios_rd16(b, offset + 0x5); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/base.c index 7102c25320fc..e4c8d310d870 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/base.c @@ -27,6 +27,7 @@ #include <subdev/bios/boost.h> #include <subdev/bios/cstep.h> #include <subdev/bios/perf.h> +#include <subdev/bios/vpstate.h> #include <subdev/fb.h> #include <subdev/therm.h> #include <subdev/volt.h> @@ -43,13 +44,13 @@ nvkm_clk_adjust(struct nvkm_clk *clk, bool adjust, struct nvkm_bios *bios = clk->subdev.device->bios; struct nvbios_boostE boostE; u8 ver, hdr, cnt, len; - u16 data; + u32 data; data = nvbios_boostEm(bios, pstate, &ver, &hdr, &cnt, &len, &boostE); if (data) { struct nvbios_boostS boostS; u8 idx = 0, sver, shdr; - u16 subd; + u32 subd; input = max(boostE.min, input); input = min(boostE.max, input); @@ -74,6 +75,88 @@ nvkm_clk_adjust(struct nvkm_clk *clk, bool adjust, /****************************************************************************** * C-States *****************************************************************************/ +static bool +nvkm_cstate_valid(struct nvkm_clk *clk, struct nvkm_cstate *cstate, + u32 max_volt, int temp) +{ + const struct nvkm_domain *domain = clk->domains; + struct nvkm_volt *volt = clk->subdev.device->volt; + int voltage; + + while (domain && domain->name != nv_clk_src_max) { + if (domain->flags & NVKM_CLK_DOM_FLAG_VPSTATE) { + u32 freq = cstate->domain[domain->name]; + switch (clk->boost_mode) { + case NVKM_CLK_BOOST_NONE: + if (clk->base_khz && freq > clk->base_khz) + return false; + case NVKM_CLK_BOOST_BIOS: + if (clk->boost_khz && freq > clk->boost_khz) + return false; + } + } + domain++; + } + + if (!volt) + return true; + + voltage = nvkm_volt_map(volt, cstate->voltage, temp); + if (voltage < 0) + return false; + return voltage <= min(max_volt, volt->max_uv); +} + +static struct nvkm_cstate * +nvkm_cstate_find_best(struct nvkm_clk *clk, struct nvkm_pstate *pstate, + struct nvkm_cstate *start) +{ + struct nvkm_device *device = clk->subdev.device; + struct nvkm_volt *volt = device->volt; + struct nvkm_cstate *cstate; + int max_volt; + + if (!pstate || !start) + return NULL; + + if (!volt) + return start; + + max_volt = volt->max_uv; + if (volt->max0_id != 0xff) + max_volt = min(max_volt, + nvkm_volt_map(volt, volt->max0_id, clk->temp)); + if (volt->max1_id != 0xff) + max_volt = min(max_volt, + nvkm_volt_map(volt, volt->max1_id, clk->temp)); + if (volt->max2_id != 0xff) + max_volt = min(max_volt, + nvkm_volt_map(volt, volt->max2_id, clk->temp)); + + for (cstate = start; &cstate->head != &pstate->list; + cstate = list_entry(cstate->head.prev, typeof(*cstate), head)) { + if (nvkm_cstate_valid(clk, cstate, max_volt, clk->temp)) + break; + } + + return cstate; +} + +static struct nvkm_cstate * +nvkm_cstate_get(struct nvkm_clk *clk, struct nvkm_pstate *pstate, int cstatei) +{ + struct nvkm_cstate *cstate; + if (cstatei == NVKM_CLK_CSTATE_HIGHEST) + return list_last_entry(&pstate->list, typeof(*cstate), head); + else { + list_for_each_entry(cstate, &pstate->list, head) { + if (cstate->id == cstatei) + return cstate; + } + } + return NULL; +} + static int nvkm_cstate_prog(struct nvkm_clk *clk, struct nvkm_pstate *pstate, int cstatei) { @@ -85,7 +168,8 @@ nvkm_cstate_prog(struct nvkm_clk *clk, struct nvkm_pstate *pstate, int cstatei) int ret; if (!list_empty(&pstate->list)) { - cstate = list_entry(pstate->list.prev, typeof(*cstate), head); + cstate = nvkm_cstate_get(clk, pstate, cstatei); + cstate = nvkm_cstate_find_best(clk, pstate, cstate); } else { cstate = &pstate->base; } @@ -99,7 +183,8 @@ nvkm_cstate_prog(struct nvkm_clk *clk, struct nvkm_pstate *pstate, int cstatei) } if (volt) { - ret = nvkm_volt_set_id(volt, cstate->voltage, +1); + ret = nvkm_volt_set_id(volt, cstate->voltage, + pstate->base.voltage, clk->temp, +1); if (ret && ret != -ENODEV) { nvkm_error(subdev, "failed to raise voltage: %d\n", ret); return ret; @@ -113,7 +198,8 @@ nvkm_cstate_prog(struct nvkm_clk *clk, struct nvkm_pstate *pstate, int cstatei) } if (volt) { - ret = nvkm_volt_set_id(volt, cstate->voltage, -1); + ret = nvkm_volt_set_id(volt, cstate->voltage, + pstate->base.voltage, clk->temp, -1); if (ret && ret != -ENODEV) nvkm_error(subdev, "failed to lower voltage: %d\n", ret); } @@ -138,22 +224,27 @@ static int nvkm_cstate_new(struct nvkm_clk *clk, int idx, struct nvkm_pstate *pstate) { struct nvkm_bios *bios = clk->subdev.device->bios; + struct nvkm_volt *volt = clk->subdev.device->volt; const struct nvkm_domain *domain = clk->domains; struct nvkm_cstate *cstate = NULL; struct nvbios_cstepX cstepX; u8 ver, hdr; - u16 data; + u32 data; data = nvbios_cstepXp(bios, idx, &ver, &hdr, &cstepX); if (!data) return -ENOENT; + if (volt && nvkm_volt_map_min(volt, cstepX.voltage) > volt->max_uv) + return -EINVAL; + cstate = kzalloc(sizeof(*cstate), GFP_KERNEL); if (!cstate) return -ENOMEM; *cstate = pstate->base; cstate->voltage = cstepX.voltage; + cstate->id = idx; while (domain && domain->name != nv_clk_src_max) { if (domain->flags & NVKM_CLK_DOM_FLAG_CORE) { @@ -175,7 +266,7 @@ static int nvkm_pstate_prog(struct nvkm_clk *clk, int pstatei) { struct nvkm_subdev *subdev = &clk->subdev; - struct nvkm_ram *ram = subdev->device->fb->ram; + struct nvkm_fb *fb = subdev->device->fb; struct nvkm_pci *pci = subdev->device->pci; struct nvkm_pstate *pstate; int ret, idx = 0; @@ -190,7 +281,8 @@ nvkm_pstate_prog(struct nvkm_clk *clk, int pstatei) nvkm_pcie_set_link(pci, pstate->pcie_speed, pstate->pcie_width); - if (ram && ram->func->calc) { + if (fb && fb->ram && fb->ram->func->calc) { + struct nvkm_ram *ram = fb->ram; int khz = pstate->base.domain[nv_clk_src_mem]; do { ret = ram->func->calc(ram, khz); @@ -200,7 +292,7 @@ nvkm_pstate_prog(struct nvkm_clk *clk, int pstatei) ram->func->tidy(ram); } - return nvkm_cstate_prog(clk, pstate, 0); + return nvkm_cstate_prog(clk, pstate, NVKM_CLK_CSTATE_HIGHEST); } static void @@ -214,14 +306,14 @@ nvkm_pstate_work(struct work_struct *work) return; clk->pwrsrc = power_supply_is_system_supplied(); - nvkm_trace(subdev, "P %d PWR %d U(AC) %d U(DC) %d A %d T %d D %d\n", + nvkm_trace(subdev, "P %d PWR %d U(AC) %d U(DC) %d A %d T %d°C D %d\n", clk->pstate, clk->pwrsrc, clk->ustate_ac, clk->ustate_dc, - clk->astate, clk->tstate, clk->dstate); + clk->astate, clk->temp, clk->dstate); pstate = clk->pwrsrc ? clk->ustate_ac : clk->ustate_dc; if (clk->state_nr && pstate != -1) { pstate = (pstate < 0) ? clk->astate : pstate; - pstate = min(pstate, clk->state_nr - 1 + clk->tstate); + pstate = min(pstate, clk->state_nr - 1); pstate = max(pstate, clk->dstate); } else { pstate = clk->pstate = -1; @@ -316,7 +408,7 @@ nvkm_pstate_new(struct nvkm_clk *clk, int idx) struct nvbios_cstepE cstepE; struct nvbios_perfE perfE; u8 ver, hdr, cnt, len; - u16 data; + u32 data; data = nvbios_perfEp(bios, idx, &ver, &hdr, &cnt, &len, &perfE); if (!data) @@ -448,13 +540,12 @@ nvkm_clk_astate(struct nvkm_clk *clk, int req, int rel, bool wait) } int -nvkm_clk_tstate(struct nvkm_clk *clk, int req, int rel) +nvkm_clk_tstate(struct nvkm_clk *clk, u8 temp) { - if (!rel) clk->tstate = req; - if ( rel) clk->tstate += rel; - clk->tstate = min(clk->tstate, 0); - clk->tstate = max(clk->tstate, -(clk->state_nr - 1)); - return nvkm_pstate_calc(clk, true); + if (clk->temp == temp) + return 0; + clk->temp = temp; + return nvkm_pstate_calc(clk, false); } int @@ -524,9 +615,9 @@ nvkm_clk_init(struct nvkm_subdev *subdev) return clk->func->init(clk); clk->astate = clk->state_nr - 1; - clk->tstate = 0; clk->dstate = 0; clk->pstate = -1; + clk->temp = 90; /* reasonable default value */ nvkm_pstate_calc(clk, true); return 0; } @@ -561,10 +652,22 @@ int nvkm_clk_ctor(const struct nvkm_clk_func *func, struct nvkm_device *device, int index, bool allow_reclock, struct nvkm_clk *clk) { + struct nvkm_subdev *subdev = &clk->subdev; + struct nvkm_bios *bios = device->bios; int ret, idx, arglen; const char *mode; + struct nvbios_vpstate_header h; + + nvkm_subdev_ctor(&nvkm_clk, device, index, subdev); + + if (bios && !nvbios_vpstate_parse(bios, &h)) { + struct nvbios_vpstate_entry base, boost; + if (!nvbios_vpstate_entry(bios, &h, h.boost_id, &boost)) + clk->boost_khz = boost.clock_mhz * 1000; + if (!nvbios_vpstate_entry(bios, &h, h.base_id, &base)) + clk->base_khz = base.clock_mhz * 1000; + } - nvkm_subdev_ctor(&nvkm_clk, device, index, &clk->subdev); clk->func = func; INIT_LIST_HEAD(&clk->states); clk->domains = func->domains; @@ -607,6 +710,8 @@ nvkm_clk_ctor(const struct nvkm_clk_func *func, struct nvkm_device *device, if (mode) clk->ustate_dc = nvkm_clk_nstate(clk, mode, arglen); + clk->boost_mode = nvkm_longopt(device->cfgopt, "NvBoost", + NVKM_CLK_BOOST_NONE); return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gf100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gf100.c index 89d5543118cf..7f67f9f5a550 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gf100.c @@ -457,7 +457,7 @@ gf100_clk = { { nv_clk_src_hubk06 , 0x00 }, { nv_clk_src_hubk01 , 0x01 }, { nv_clk_src_copy , 0x02 }, - { nv_clk_src_gpc , 0x03, 0, "core", 2000 }, + { nv_clk_src_gpc , 0x03, NVKM_CLK_DOM_FLAG_VPSTATE, "core", 2000 }, { nv_clk_src_rop , 0x04 }, { nv_clk_src_mem , 0x05, 0, "memory", 1000 }, { nv_clk_src_vdec , 0x06 }, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk104.c b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk104.c index 06bc0d2d6ae1..0b37e3da7feb 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk104.c @@ -491,7 +491,7 @@ gk104_clk = { .domains = { { nv_clk_src_crystal, 0xff }, { nv_clk_src_href , 0xff }, - { nv_clk_src_gpc , 0x00, NVKM_CLK_DOM_FLAG_CORE, "core", 2000 }, + { nv_clk_src_gpc , 0x00, NVKM_CLK_DOM_FLAG_CORE | NVKM_CLK_DOM_FLAG_VPSTATE, "core", 2000 }, { nv_clk_src_hubk07 , 0x01, NVKM_CLK_DOM_FLAG_CORE }, { nv_clk_src_rop , 0x02, NVKM_CLK_DOM_FLAG_CORE }, { nv_clk_src_mem , 0x03, 0, "memory", 500 }, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gt215.c b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gt215.c index 056702ef69aa..96e0941c8edd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gt215.c @@ -180,7 +180,7 @@ gt215_clk_read(struct nvkm_clk *base, enum nv_clk_src src) return 0; } -int +static int gt215_clk_info(struct nvkm_clk *base, int idx, u32 khz, struct gt215_clk_info *info) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/nv50.c index 5841f297973c..da1770e47490 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/nv50.c @@ -112,7 +112,7 @@ read_pll_src(struct nv50_clk *clk, u32 base) M = (coef & 0x000000ff) >> 0; break; default: - BUG_ON(1); + BUG(); } if (M) diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c index a410c0db8a08..1730371933df 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c @@ -26,6 +26,7 @@ #include <subdev/bios.h> #include <subdev/bios/bit.h> #include <subdev/bios/pmu.h> +#include <subdev/timer.h> static void pmu_code(struct nv50_devinit *init, u32 pmu, u32 img, u32 len, bool sec) @@ -123,21 +124,13 @@ gm200_devinit_post(struct nvkm_devinit *base, bool post) return -EINVAL; } - /* reset PMU and load init table parser ucode */ - if (post) { - nvkm_mask(device, 0x000200, 0x00002000, 0x00000000); - nvkm_mask(device, 0x000200, 0x00002000, 0x00002000); - nvkm_rd32(device, 0x000200); - while (nvkm_rd32(device, 0x10a10c) & 0x00000006) { - } - } - ret = pmu_load(init, 0x04, post, &exec, &args); if (ret) return ret; /* upload first chunk of init data */ if (post) { + // devinit tables u32 pmu = pmu_args(init, args + 0x08, 0x08); u32 img = nvbios_rd16(bios, bit_I.offset + 0x14); u32 len = nvbios_rd16(bios, bit_I.offset + 0x16); @@ -146,6 +139,7 @@ gm200_devinit_post(struct nvkm_devinit *base, bool post) /* upload second chunk of init data */ if (post) { + // devinit boot scripts u32 pmu = pmu_args(init, args + 0x08, 0x10); u32 img = nvbios_rd16(bios, bit_I.offset + 0x18); u32 len = nvbios_rd16(bios, bit_I.offset + 0x1a); @@ -156,8 +150,11 @@ gm200_devinit_post(struct nvkm_devinit *base, bool post) if (post) { nvkm_wr32(device, 0x10a040, 0x00005000); pmu_exec(init, exec); - while (!(nvkm_rd32(device, 0x10a040) & 0x00002000)) { - } + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x10a040) & 0x00002000) + break; + ) < 0) + return -ETIMEDOUT; } /* load and execute some other ucode image (bios therm?) */ diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/nv50.c index c714b097719c..59362f8dee22 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/nv50.c @@ -50,7 +50,7 @@ nv50_devinit_pll_set(struct nvkm_devinit *init, u32 type, u32 freq) ret = nv04_pll_calc(subdev, &info, freq, &N1, &M1, &N2, &M2, &P); if (!ret) { nvkm_error(subdev, "failed pll calculation\n"); - return ret; + return -EINVAL; } switch (info.type) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/Kbuild index edcc157e6ac8..63566ba12fbb 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/Kbuild @@ -24,8 +24,9 @@ nvkm-y += nvkm/subdev/fb/gk104.o nvkm-y += nvkm/subdev/fb/gk20a.o nvkm-y += nvkm/subdev/fb/gm107.o nvkm-y += nvkm/subdev/fb/gm200.o +nvkm-y += nvkm/subdev/fb/gm20b.o nvkm-y += nvkm/subdev/fb/gp100.o -nvkm-y += nvkm/subdev/fb/gp104.o +nvkm-y += nvkm/subdev/fb/gp102.o nvkm-y += nvkm/subdev/fb/ram.o nvkm-y += nvkm/subdev/fb/ramnv04.o diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.c index 76433cc66fff..3841ad6be99e 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.c @@ -50,24 +50,33 @@ gf100_fb_intr(struct nvkm_fb *base) } int -gf100_fb_oneinit(struct nvkm_fb *fb) +gf100_fb_oneinit(struct nvkm_fb *base) { - struct nvkm_device *device = fb->subdev.device; + struct gf100_fb *fb = gf100_fb(base); + struct nvkm_device *device = fb->base.subdev.device; int ret, size = 0x1000; size = nvkm_longopt(device->cfgopt, "MmuDebugBufferSize", size); size = min(size, 0x1000); ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, size, 0x1000, - false, &fb->mmu_rd); + false, &fb->base.mmu_rd); if (ret) return ret; ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, size, 0x1000, - false, &fb->mmu_wr); + false, &fb->base.mmu_wr); if (ret) return ret; + fb->r100c10_page = alloc_page(GFP_KERNEL | __GFP_ZERO); + if (fb->r100c10_page) { + fb->r100c10 = dma_map_page(device->dev, fb->r100c10_page, 0, + PAGE_SIZE, DMA_BIDIRECTIONAL); + if (dma_mapping_error(device->dev, fb->r100c10)) + return -EFAULT; + } + return 0; } @@ -123,14 +132,6 @@ gf100_fb_new_(const struct nvkm_fb_func *func, struct nvkm_device *device, nvkm_fb_ctor(func, device, index, &fb->base); *pfb = &fb->base; - fb->r100c10_page = alloc_page(GFP_KERNEL | __GFP_ZERO); - if (fb->r100c10_page) { - fb->r100c10 = dma_map_page(device->dev, fb->r100c10_page, 0, - PAGE_SIZE, DMA_BIDIRECTIONAL); - if (dma_mapping_error(device->dev, fb->r100c10)) - return -EFAULT; - } - return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.h b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.h index 449f431644b3..412eb89834e8 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.h @@ -16,4 +16,8 @@ void gf100_fb_init(struct nvkm_fb *); void gf100_fb_intr(struct nvkm_fb *); void gp100_fb_init(struct nvkm_fb *); + +void gm200_fb_init_page(struct nvkm_fb *fb); +void gm200_fb_init(struct nvkm_fb *base); + #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gk20a.c index f815fe2bbf08..5d34d6136616 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gk20a.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gk20a.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. + * Copyright (c) 2014-2016, NVIDIA CORPORATION. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -20,27 +20,21 @@ * DEALINGS IN THE SOFTWARE. */ #include "priv.h" +#include "gf100.h" -#include <core/memory.h> - -static void -gk20a_fb_init(struct nvkm_fb *fb) -{ - struct nvkm_device *device = fb->subdev.device; - nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(fb->mmu_wr) >> 8); - nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(fb->mmu_rd) >> 8); -} - +/* GK20A's FB is similar to GF100's, but without the ability to allocate VRAM */ static const struct nvkm_fb_func gk20a_fb = { + .dtor = gf100_fb_dtor, .oneinit = gf100_fb_oneinit, - .init = gk20a_fb_init, + .init = gf100_fb_init, .init_page = gf100_fb_init_page, + .intr = gf100_fb_intr, .memtype_valid = gf100_fb_memtype_valid, }; int gk20a_fb_new(struct nvkm_device *device, int index, struct nvkm_fb **pfb) { - return nvkm_fb_new_(&gk20a_fb, device, index, pfb); + return gf100_fb_new_(&gk20a_fb, device, index, pfb); } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm200.c index 62f653240be3..fe5886013ac0 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm200.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm200.c @@ -44,7 +44,7 @@ gm200_fb_init_page(struct nvkm_fb *fb) } } -static void +void gm200_fb_init(struct nvkm_fb *base) { struct gf100_fb *fb = gf100_fb(base); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm20b.c new file mode 100644 index 000000000000..b87c233bcd6d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm20b.c @@ -0,0 +1,40 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "gf100.h" + +/* GM20B's FB is similar to GM200, but without the ability to allocate VRAM */ +static const struct nvkm_fb_func +gm20b_fb = { + .dtor = gf100_fb_dtor, + .oneinit = gf100_fb_oneinit, + .init = gm200_fb_init, + .init_page = gm200_fb_init_page, + .intr = gf100_fb_intr, + .memtype_valid = gf100_fb_memtype_valid, +}; + +int +gm20b_fb_new(struct nvkm_device *device, int index, struct nvkm_fb **pfb) +{ + return gf100_fb_new_(&gm20b_fb, device, index, pfb); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp104.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp102.c index 92cb71861bec..73b4ae1c73dc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp104.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp102.c @@ -27,7 +27,7 @@ #include <core/memory.h> static const struct nvkm_fb_func -gp104_fb = { +gp102_fb = { .dtor = gf100_fb_dtor, .oneinit = gf100_fb_oneinit, .init = gp100_fb_init, @@ -37,7 +37,7 @@ gp104_fb = { }; int -gp104_fb_new(struct nvkm_device *device, int index, struct nvkm_fb **pfb) +gp102_fb_new(struct nvkm_device *device, int index, struct nvkm_fb **pfb) { - return gf100_fb_new_(&gp104_fb, device, index, pfb); + return gf100_fb_new_(&gp102_fb, device, index, pfb); } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/nv50.c index 1b5fb02eab2a..0595e0722bfc 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/nv50.c @@ -210,6 +210,23 @@ nv50_fb_intr(struct nvkm_fb *base) nvkm_fifo_chan_put(fifo, flags, &chan); } +static int +nv50_fb_oneinit(struct nvkm_fb *base) +{ + struct nv50_fb *fb = nv50_fb(base); + struct nvkm_device *device = fb->base.subdev.device; + + fb->r100c08_page = alloc_page(GFP_KERNEL | __GFP_ZERO); + if (fb->r100c08_page) { + fb->r100c08 = dma_map_page(device->dev, fb->r100c08_page, 0, + PAGE_SIZE, DMA_BIDIRECTIONAL); + if (dma_mapping_error(device->dev, fb->r100c08)) + return -EFAULT; + } + + return 0; +} + static void nv50_fb_init(struct nvkm_fb *base) { @@ -245,6 +262,7 @@ nv50_fb_dtor(struct nvkm_fb *base) static const struct nvkm_fb_func nv50_fb_ = { .dtor = nv50_fb_dtor, + .oneinit = nv50_fb_oneinit, .init = nv50_fb_init, .intr = nv50_fb_intr, .ram_new = nv50_fb_ram_new, @@ -263,16 +281,6 @@ nv50_fb_new_(const struct nv50_fb_func *func, struct nvkm_device *device, fb->func = func; *pfb = &fb->base; - fb->r100c08_page = alloc_page(GFP_KERNEL | __GFP_ZERO); - if (fb->r100c08_page) { - fb->r100c08 = dma_map_page(device->dev, fb->r100c08_page, 0, - PAGE_SIZE, DMA_BIDIRECTIONAL); - if (dma_mapping_error(device->dev, fb->r100c08)) - return -EFAULT; - } else { - nvkm_warn(&fb->base.subdev, "failed 100c08 page alloc\n"); - } - return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ram.h b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ram.h index b9ec0ae6723a..b60068b7d8f9 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ram.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ram.h @@ -24,6 +24,7 @@ int gf100_ram_ctor(const struct nvkm_ram_func *, struct nvkm_fb *, int gf100_ram_get(struct nvkm_ram *, u64, u32, u32, u32, struct nvkm_mem **); void gf100_ram_put(struct nvkm_ram *, struct nvkm_mem **); +int gk104_ram_ctor(struct nvkm_fb *, struct nvkm_ram **, u32); int gk104_ram_init(struct nvkm_ram *ram); /* RAM type-specific MR calculation routines */ diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgf100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgf100.c index 772425ca5a9e..6758da93a3a1 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgf100.c @@ -420,8 +420,6 @@ gf100_ram_tidy(struct nvkm_ram *base) ram_exec(&ram->fuc, false); } -extern const u8 gf100_pte_storage_type_map[256]; - void gf100_ram_put(struct nvkm_ram *ram, struct nvkm_mem **pmem) { @@ -447,7 +445,7 @@ gf100_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, { struct nvkm_ltc *ltc = ram->fb->subdev.device->ltc; struct nvkm_mm *mm = &ram->vram; - struct nvkm_mm_node *r; + struct nvkm_mm_node **node, *r; struct nvkm_mem *mem; int type = (memtype & 0x0ff); int back = (memtype & 0x800); @@ -464,7 +462,6 @@ gf100_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, if (!mem) return -ENOMEM; - INIT_LIST_HEAD(&mem->regions); mem->size = size; mutex_lock(&ram->fb->subdev.mutex); @@ -480,6 +477,7 @@ gf100_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, } mem->memtype = type; + node = &mem->mem; do { if (back) ret = nvkm_mm_tail(mm, 0, 1, size, ncmin, align, &r); @@ -491,13 +489,13 @@ gf100_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, return ret; } - list_add_tail(&r->rl_entry, &mem->regions); + *node = r; + node = &r->next; size -= r->length; } while (size); mutex_unlock(&ram->fb->subdev.mutex); - r = list_first_entry(&mem->regions, struct nvkm_mm_node, rl_entry); - mem->offset = (u64)r->offset << NVKM_RAM_MM_SHIFT; + mem->offset = (u64)mem->mem->offset << NVKM_RAM_MM_SHIFT; *pmem = mem; return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgk104.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgk104.c index 1fa3ade468ae..fb8a1239743d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgk104.c @@ -259,7 +259,9 @@ gk104_ram_calc_gddr5(struct gk104_ram *ram, u32 freq) ram_mask(fuc, 0x10f808, 0x40000000, 0x40000000); ram_block(fuc); - ram_wr32(fuc, 0x62c000, 0x0f0f0000); + + if (nvkm_device_engine(ram->base.fb->subdev.device, NVKM_ENGINE_DISP)) + ram_wr32(fuc, 0x62c000, 0x0f0f0000); /* MR1: turn termination on early, for some reason.. */ if ((ram->base.mr[1] & 0x03c) != 0x030) { @@ -658,7 +660,9 @@ gk104_ram_calc_gddr5(struct gk104_ram *ram, u32 freq) gk104_ram_train(fuc, 0x80020000, 0x01000000); ram_unblock(fuc); - ram_wr32(fuc, 0x62c000, 0x0f0f0f00); + + if (nvkm_device_engine(ram->base.fb->subdev.device, NVKM_ENGINE_DISP)) + ram_wr32(fuc, 0x62c000, 0x0f0f0f00); if (next->bios.rammap_11_08_01) data = 0x00000800; @@ -706,7 +710,9 @@ gk104_ram_calc_sddr3(struct gk104_ram *ram, u32 freq) ram_mask(fuc, 0x10f808, 0x40000000, 0x40000000); ram_block(fuc); - ram_wr32(fuc, 0x62c000, 0x0f0f0000); + + if (nvkm_device_engine(ram->base.fb->subdev.device, NVKM_ENGINE_DISP)) + ram_wr32(fuc, 0x62c000, 0x0f0f0000); if (vc == 1 && ram_have(fuc, gpio2E)) { u32 temp = ram_mask(fuc, gpio2E, 0x3000, fuc->r_func2E[1]); @@ -936,7 +942,9 @@ gk104_ram_calc_sddr3(struct gk104_ram *ram, u32 freq) ram_nsec(fuc, 1000); ram_unblock(fuc); - ram_wr32(fuc, 0x62c000, 0x0f0f0f00); + + if (nvkm_device_engine(ram->base.fb->subdev.device, NVKM_ENGINE_DISP)) + ram_wr32(fuc, 0x62c000, 0x0f0f0f00); if (next->bios.rammap_11_08_01) data = 0x00000800; @@ -981,7 +989,7 @@ gk104_pll_calc_hiclk(int target_khz, int crystal, int *N1, int *fN1, int *M1, int *P1, int *N2, int *M2, int *P2) { - int best_clk = 0, best_err = target_khz, p_ref, n_ref; + int best_err = target_khz, p_ref, n_ref; bool upper = false; *M1 = 1; @@ -1002,7 +1010,6 @@ gk104_pll_calc_hiclk(int target_khz, int crystal, /* we found a better combination */ if (cur_err < best_err) { best_err = cur_err; - best_clk = cur_clk; *N2 = cur_N; *N1 = n_ref; *P1 = p_ref; @@ -1014,7 +1021,6 @@ gk104_pll_calc_hiclk(int target_khz, int crystal, - target_khz; if (cur_err < best_err) { best_err = cur_err; - best_clk = cur_clk; *N2 = cur_N; *N1 = n_ref; *P1 = p_ref; @@ -1530,6 +1536,12 @@ gk104_ram_func = { int gk104_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram) { + return gk104_ram_ctor(fb, pram, 0x022554); +} + +int +gk104_ram_ctor(struct nvkm_fb *fb, struct nvkm_ram **pram, u32 maskaddr) +{ struct nvkm_subdev *subdev = &fb->subdev; struct nvkm_device *device = subdev->device; struct nvkm_bios *bios = device->bios; @@ -1544,7 +1556,7 @@ gk104_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram) return -ENOMEM; *pram = &ram->base; - ret = gf100_ram_ctor(&gk104_ram_func, fb, 0x022554, &ram->base); + ret = gf100_ram_ctor(&gk104_ram_func, fb, maskaddr, &ram->base); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgm107.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgm107.c index 43d807f6ca71..ac862d1d77bd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgm107.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgm107.c @@ -23,18 +23,8 @@ */ #include "ram.h" -static const struct nvkm_ram_func -gm107_ram_func = { - .init = gk104_ram_init, - .get = gf100_ram_get, - .put = gf100_ram_put, -}; - int gm107_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram) { - if (!(*pram = kzalloc(sizeof(**pram), GFP_KERNEL))) - return -ENOMEM; - - return gf100_ram_ctor(&gm107_ram_func, fb, 0x021c14, *pram); + return gk104_ram_ctor(fb, pram, 0x021c14); } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgp100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgp100.c index f3be408b5e5e..405faabe8dcd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgp100.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgp100.c @@ -92,13 +92,13 @@ gp100_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram) enum nvkm_ram_type type = nvkm_fb_bios_memtype(device->bios); const u32 rsvd_head = ( 256 * 1024); /* vga memory */ const u32 rsvd_tail = (1024 * 1024); /* vbios etc */ - u32 fbpa_num = nvkm_rd32(device, 0x022438), fbpa; + u32 fbpa_num = nvkm_rd32(device, 0x02243c), fbpa; u32 fbio_opt = nvkm_rd32(device, 0x021c14); u64 part, size = 0, comm = ~0ULL; bool mixed = false; int ret; - nvkm_debug(subdev, "022438: %08x\n", fbpa_num); + nvkm_debug(subdev, "02243c: %08x\n", fbpa_num); nvkm_debug(subdev, "021c14: %08x\n", fbio_opt); for (fbpa = 0; fbpa < fbpa_num; fbpa++) { if (!(fbio_opt & (1 << fbpa))) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgt215.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgt215.c index d15ea886df27..f10664372161 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgt215.c @@ -95,7 +95,7 @@ struct gt215_ram { struct gt215_ltrain ltrain; }; -void +static void gt215_link_train_calc(u32 *vals, struct gt215_ltrain *train) { int i, lo, hi; @@ -149,7 +149,7 @@ gt215_link_train_calc(u32 *vals, struct gt215_ltrain *train) /* * Link training for (at least) DDR3 */ -int +static int gt215_link_train(struct gt215_ram *ram) { struct gt215_ltrain *train = &ram->ltrain; @@ -267,7 +267,7 @@ out: return ret; } -int +static int gt215_link_train_init(struct gt215_ram *ram) { static const u32 pattern[16] = { @@ -333,7 +333,7 @@ gt215_link_train_init(struct gt215_ram *ram) return 0; } -void +static void gt215_link_train_fini(struct gt215_ram *ram) { if (ram->ltrain.mem) diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/rammcp77.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/rammcp77.c index 0a0e44b75577..017a91de74a0 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/rammcp77.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/rammcp77.c @@ -39,7 +39,7 @@ mcp77_ram_init(struct nvkm_ram *base) u32 flush = ((ram->base.size - (ram->poller_base + 0x40)) >> 5) - 1; /* Enable NISO poller for various clients and set their associated - * read address, only for MCP77/78 and MCP79/7A. (fd#25701) + * read address, only for MCP77/78 and MCP79/7A. (fd#27501) */ nvkm_wr32(device, 0x100c18, dniso); nvkm_mask(device, 0x100c14, 0x00000000, 0x00000001); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramnv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramnv50.c index 87bde8ff2d6b..6549b0588309 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramnv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramnv50.c @@ -496,15 +496,12 @@ nv50_ram_tidy(struct nvkm_ram *base) void __nv50_ram_put(struct nvkm_ram *ram, struct nvkm_mem *mem) { - struct nvkm_mm_node *this; - - while (!list_empty(&mem->regions)) { - this = list_first_entry(&mem->regions, typeof(*this), rl_entry); - - list_del(&this->rl_entry); - nvkm_mm_free(&ram->vram, &this); + struct nvkm_mm_node *next = mem->mem; + struct nvkm_mm_node *node; + while ((node = next)) { + next = node->next; + nvkm_mm_free(&ram->vram, &node); } - nvkm_mm_free(&ram->tags, &mem->tag); } @@ -530,7 +527,7 @@ nv50_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, { struct nvkm_mm *heap = &ram->vram; struct nvkm_mm *tags = &ram->tags; - struct nvkm_mm_node *r; + struct nvkm_mm_node **node, *r; struct nvkm_mem *mem; int comp = (memtype & 0x300) >> 8; int type = (memtype & 0x07f); @@ -559,11 +556,11 @@ nv50_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, comp = 0; } - INIT_LIST_HEAD(&mem->regions); mem->memtype = (comp << 7) | type; mem->size = max; type = nv50_fb_memtype[type]; + node = &mem->mem; do { if (back) ret = nvkm_mm_tail(heap, 0, type, max, min, align, &r); @@ -575,13 +572,13 @@ nv50_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin, return ret; } - list_add_tail(&r->rl_entry, &mem->regions); + *node = r; + node = &r->next; max -= r->length; } while (max); mutex_unlock(&ram->fb->subdev.mutex); - r = list_first_entry(&mem->regions, struct nvkm_mm_node, rl_entry); - mem->offset = (u64)r->offset << NVKM_RAM_MM_SHIFT; + mem->offset = (u64)mem->mem->offset << NVKM_RAM_MM_SHIFT; *pmem = mem; return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr2.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr2.c index b9f1ffdfc602..4dcd8742f2da 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr2.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr2.c @@ -23,6 +23,7 @@ * Ben Skeggs */ #include "priv.h" +#include "ram.h" struct ramxlat { int id; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr3.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr3.c index 26900333b1d6..eca8a445eab3 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr3.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fb/sddr3.c @@ -23,6 +23,7 @@ * Roy Spliet <rspliet@eclipso.eu> */ #include "priv.h" +#include "ram.h" struct ramxlat { int id; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gpio/gk104.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gpio/gk104.c index 3f45afd17d5a..2ead515b8530 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/gpio/gk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gpio/gk104.c @@ -37,7 +37,7 @@ gk104_gpio_intr_stat(struct nvkm_gpio *gpio, u32 *hi, u32 *lo) nvkm_wr32(device, 0x00dc80, intr1); } -void +static void gk104_gpio_intr_mask(struct nvkm_gpio *gpio, u32 type, u32 mask, u32 data) { struct nvkm_device *device = gpio->subdev.device; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/aux.c b/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/aux.c index f0851d57df2f..01d5c5a56e2e 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/aux.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/aux.c @@ -74,7 +74,7 @@ nvkm_i2c_aux_i2c_func(struct i2c_adapter *adap) return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL; } -const struct i2c_algorithm +static const struct i2c_algorithm nvkm_i2c_aux_i2c_algo = { .master_xfer = nvkm_i2c_aux_i2c_xfer, .functionality = nvkm_i2c_aux_i2c_func diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxg94.c b/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxg94.c index 954f5b76bfcf..b80236a4eeac 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxg94.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxg94.c @@ -79,7 +79,7 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry, struct g94_i2c_aux *aux = g94_i2c_aux(obj); struct nvkm_device *device = aux->base.pad->i2c->subdev.device; const u32 base = aux->ch * 0x50; - u32 ctrl, stat, timeout, retries; + u32 ctrl, stat, timeout, retries = 0; u32 xbuf[4] = {}; int ret, i; @@ -111,7 +111,7 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry, nvkm_wr32(device, 0x00e4e0 + base, addr); /* (maybe) retry transaction a number of times on failure... */ - for (retries = 0; !ret && retries < 32; retries++) { + do { /* reset, and delay a while if this is a retry */ nvkm_wr32(device, 0x00e4e4 + base, 0x80000000 | ctrl); nvkm_wr32(device, 0x00e4e4 + base, 0x00000000 | ctrl); @@ -131,20 +131,20 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry, goto out; } } while (ctrl & 0x00010000); - ret = 1; + ret = 0; /* read status, and check if transaction completed ok */ stat = nvkm_mask(device, 0x00e4e8 + base, 0, 0); if ((stat & 0x000f0000) == 0x00080000 || (stat & 0x000f0000) == 0x00020000) - ret = retry ? 0 : 1; + ret = 1; if ((stat & 0x00000100)) ret = -ETIMEDOUT; if ((stat & 0x00000e00)) ret = -EIO; AUX_TRACE(&aux->base, "%02d %08x %08x", retries, ctrl, stat); - } + } while (ret && retry && retries++ < 32); if (type & 1) { for (i = 0; i < 16; i += 4) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxgm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxgm200.c index 61d729b82c69..ed458c7f056b 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxgm200.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxgm200.c @@ -79,7 +79,7 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry, struct gm200_i2c_aux *aux = gm200_i2c_aux(obj); struct nvkm_device *device = aux->base.pad->i2c->subdev.device; const u32 base = aux->ch * 0x50; - u32 ctrl, stat, timeout, retries; + u32 ctrl, stat, timeout, retries = 0; u32 xbuf[4] = {}; int ret, i; @@ -111,7 +111,7 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry, nvkm_wr32(device, 0x00d950 + base, addr); /* (maybe) retry transaction a number of times on failure... */ - for (retries = 0; !ret && retries < 32; retries++) { + do { /* reset, and delay a while if this is a retry */ nvkm_wr32(device, 0x00d954 + base, 0x80000000 | ctrl); nvkm_wr32(device, 0x00d954 + base, 0x00000000 | ctrl); @@ -131,20 +131,20 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry, goto out; } } while (ctrl & 0x00010000); - ret = 1; + ret = 0; /* read status, and check if transaction completed ok */ stat = nvkm_mask(device, 0x00d958 + base, 0, 0); if ((stat & 0x000f0000) == 0x00080000 || (stat & 0x000f0000) == 0x00020000) - ret = retry ? 0 : 1; + ret = 1; if ((stat & 0x00000100)) ret = -ETIMEDOUT; if ((stat & 0x00000e00)) ret = -EIO; AUX_TRACE(&aux->base, "%02d %08x %08x", retries, ctrl, stat); - } + } while (ret && retry && retries++ < 32); if (type & 1) { for (i = 0; i < 16; i += 4) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/ibus/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/subdev/ibus/gk20a.c index b7159b338fac..1a4ab825852c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/ibus/gk20a.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/ibus/gk20a.c @@ -29,7 +29,7 @@ gk20a_ibus_init_ibus_ring(struct nvkm_subdev *ibus) nvkm_mask(device, 0x137250, 0x3f, 0); nvkm_mask(device, 0x000200, 0x20, 0); - usleep_range(20, 30); + udelay(20); nvkm_mask(device, 0x000200, 0x20, 0x20); nvkm_wr32(device, 0x12004c, 0x4); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/base.c index 41bd5d0f7692..fecfa6afcf54 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/base.c @@ -26,6 +26,7 @@ #include <subdev/bios.h> #include <subdev/bios/extdev.h> #include <subdev/bios/iccsense.h> +#include <subdev/bios/power_budget.h> #include <subdev/i2c.h> static bool @@ -96,60 +97,12 @@ nvkm_iccsense_ina3221_read(struct nvkm_iccsense *iccsense, } static void -nvkm_iccsense_ina209_config(struct nvkm_iccsense *iccsense, - struct nvkm_iccsense_sensor *sensor) -{ - struct nvkm_subdev *subdev = &iccsense->subdev; - /* configuration: - * 0x0007: 0x0007 shunt and bus continous - * 0x0078: 0x0078 128 samples shunt - * 0x0780: 0x0780 128 samples bus - * 0x1800: 0x0000 +-40 mV shunt range - * 0x2000: 0x0000 16V FSR - */ - u16 value = 0x07ff; - nvkm_debug(subdev, "config for sensor id %i: 0x%x\n", sensor->id, value); - nv_wr16i2cr(sensor->i2c, sensor->addr, 0x00, value); -} - -static void -nvkm_iccsense_ina3221_config(struct nvkm_iccsense *iccsense, - struct nvkm_iccsense_sensor *sensor) -{ - struct nvkm_subdev *subdev = &iccsense->subdev; - /* configuration: - * 0x0007: 0x0007 shunt and bus continous - * 0x0031: 0x0000 140 us conversion time shunt - * 0x01c0: 0x0000 140 us conversion time bus - * 0x0f00: 0x0f00 1024 samples - * 0x7000: 0x?000 channels - */ - u16 value = 0x0e07; - if (sensor->rail_mask & 0x1) - value |= 0x1 << 14; - if (sensor->rail_mask & 0x2) - value |= 0x1 << 13; - if (sensor->rail_mask & 0x4) - value |= 0x1 << 12; - nvkm_debug(subdev, "config for sensor id %i: 0x%x\n", sensor->id, value); - nv_wr16i2cr(sensor->i2c, sensor->addr, 0x00, value); -} - -static void nvkm_iccsense_sensor_config(struct nvkm_iccsense *iccsense, struct nvkm_iccsense_sensor *sensor) { - switch (sensor->type) { - case NVBIOS_EXTDEV_INA209: - case NVBIOS_EXTDEV_INA219: - nvkm_iccsense_ina209_config(iccsense, sensor); - break; - case NVBIOS_EXTDEV_INA3221: - nvkm_iccsense_ina3221_config(iccsense, sensor); - break; - default: - break; - } + struct nvkm_subdev *subdev = &iccsense->subdev; + nvkm_trace(subdev, "write config of extdev %i: 0x%04x\n", sensor->id, sensor->config); + nv_wr16i2cr(sensor->i2c, sensor->addr, 0x00, sensor->config); } int @@ -196,7 +149,6 @@ nvkm_iccsense_dtor(struct nvkm_subdev *subdev) static struct nvkm_iccsense_sensor* nvkm_iccsense_create_sensor(struct nvkm_iccsense *iccsense, u8 id) { - struct nvkm_subdev *subdev = &iccsense->subdev; struct nvkm_bios *bios = subdev->device->bios; struct nvkm_i2c *i2c = subdev->device->i2c; @@ -245,7 +197,7 @@ nvkm_iccsense_create_sensor(struct nvkm_iccsense *iccsense, u8 id) sensor->type = extdev.type; sensor->i2c = &i2c_bus->i2c; sensor->addr = addr; - sensor->rail_mask = 0x0; + sensor->config = 0x0; return sensor; } @@ -265,56 +217,79 @@ nvkm_iccsense_oneinit(struct nvkm_subdev *subdev) { struct nvkm_iccsense *iccsense = nvkm_iccsense(subdev); struct nvkm_bios *bios = subdev->device->bios; + struct nvbios_power_budget budget; struct nvbios_iccsense stbl; - int i; + int i, ret; + + if (!bios) + return 0; + + ret = nvbios_power_budget_header(bios, &budget); + if (!ret && budget.cap_entry != 0xff) { + struct nvbios_power_budget_entry entry; + ret = nvbios_power_budget_entry(bios, &budget, + budget.cap_entry, &entry); + if (!ret) { + iccsense->power_w_max = entry.avg_w; + iccsense->power_w_crit = entry.max_w; + } + } - if (!bios || nvbios_iccsense_parse(bios, &stbl) || !stbl.nr_entry) + if (nvbios_iccsense_parse(bios, &stbl) || !stbl.nr_entry) return 0; iccsense->data_valid = true; for (i = 0; i < stbl.nr_entry; ++i) { - struct pwr_rail_t *r = &stbl.rail[i]; - struct nvkm_iccsense_rail *rail; + struct pwr_rail_t *pwr_rail = &stbl.rail[i]; struct nvkm_iccsense_sensor *sensor; - int (*read)(struct nvkm_iccsense *, - struct nvkm_iccsense_rail *); + int r; - if (!r->mode || r->resistor_mohm == 0) + if (pwr_rail->mode != 1 || !pwr_rail->resistor_count) continue; - sensor = nvkm_iccsense_get_sensor(iccsense, r->extdev_id); + sensor = nvkm_iccsense_get_sensor(iccsense, pwr_rail->extdev_id); if (!sensor) continue; - switch (sensor->type) { - case NVBIOS_EXTDEV_INA209: - if (r->rail != 0) - continue; - read = nvkm_iccsense_ina209_read; - break; - case NVBIOS_EXTDEV_INA219: - if (r->rail != 0) + if (!sensor->config) + sensor->config = pwr_rail->config; + else if (sensor->config != pwr_rail->config) + nvkm_error(subdev, "config mismatch found for extdev %i\n", pwr_rail->extdev_id); + + for (r = 0; r < pwr_rail->resistor_count; ++r) { + struct nvkm_iccsense_rail *rail; + struct pwr_rail_resistor_t *res = &pwr_rail->resistors[r]; + int (*read)(struct nvkm_iccsense *, + struct nvkm_iccsense_rail *); + + if (!res->mohm || !res->enabled) continue; - read = nvkm_iccsense_ina219_read; - break; - case NVBIOS_EXTDEV_INA3221: - if (r->rail >= 3) + + switch (sensor->type) { + case NVBIOS_EXTDEV_INA209: + read = nvkm_iccsense_ina209_read; + break; + case NVBIOS_EXTDEV_INA219: + read = nvkm_iccsense_ina219_read; + break; + case NVBIOS_EXTDEV_INA3221: + read = nvkm_iccsense_ina3221_read; + break; + default: continue; - read = nvkm_iccsense_ina3221_read; - break; - default: - continue; + } + + rail = kmalloc(sizeof(*rail), GFP_KERNEL); + if (!rail) + return -ENOMEM; + + rail->read = read; + rail->sensor = sensor; + rail->idx = r; + rail->mohm = res->mohm; + nvkm_debug(subdev, "create rail for extdev %i: { idx: %i, mohm: %i }\n", pwr_rail->extdev_id, r, rail->mohm); + list_add_tail(&rail->head, &iccsense->rails); } - - rail = kmalloc(sizeof(*rail), GFP_KERNEL); - if (!rail) - return -ENOMEM; - sensor->rail_mask |= 1 << r->rail; - rail->read = read; - rail->sensor = sensor; - rail->idx = r->rail; - rail->mohm = r->resistor_mohm; - list_add_tail(&rail->head, &iccsense->rails); } return 0; } @@ -329,7 +304,8 @@ nvkm_iccsense_init(struct nvkm_subdev *subdev) return 0; } -struct nvkm_subdev_func iccsense_func = { +static const struct nvkm_subdev_func +iccsense_func = { .oneinit = nvkm_iccsense_oneinit, .init = nvkm_iccsense_init, .dtor = nvkm_iccsense_dtor, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/priv.h index b72c31d2f908..e90e0f6ed008 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/iccsense/priv.h @@ -10,7 +10,7 @@ struct nvkm_iccsense_sensor { enum nvbios_extdev_type type; struct i2c_adapter *i2c; u8 addr; - u8 rail_mask; + u16 config; }; struct nvkm_iccsense_rail { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/base.c index 8ed8f65ff664..10c987a654ec 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/base.c @@ -104,7 +104,7 @@ nvkm_instobj_dtor(struct nvkm_memory *memory) return iobj; } -const struct nvkm_memory_func +static const struct nvkm_memory_func nvkm_instobj_func = { .dtor = nvkm_instobj_dtor, .target = nvkm_instobj_target, @@ -156,7 +156,7 @@ nvkm_instobj_wr32_slow(struct nvkm_memory *memory, u64 offset, u32 data) return nvkm_wo32(iobj->parent, offset, data); } -const struct nvkm_memory_func +static const struct nvkm_memory_func nvkm_instobj_func_slow = { .dtor = nvkm_instobj_dtor, .target = nvkm_instobj_target, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c index a6a7fa0d7679..9dec58ec3d9f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/gk20a.c @@ -116,7 +116,7 @@ struct gk20a_instmem { static enum nvkm_memory_target gk20a_instobj_target(struct nvkm_memory *memory) { - return NVKM_MEM_TARGET_HOST; + return NVKM_MEM_TARGET_NCOH; } static u64 @@ -305,11 +305,11 @@ gk20a_instobj_dtor_iommu(struct nvkm_memory *memory) struct gk20a_instobj_iommu *node = gk20a_instobj_iommu(memory); struct gk20a_instmem *imem = node->base.imem; struct device *dev = imem->base.subdev.device->dev; - struct nvkm_mm_node *r; + struct nvkm_mm_node *r = node->base.mem.mem; unsigned long flags; int i; - if (unlikely(list_empty(&node->base.mem.regions))) + if (unlikely(!r)) goto out; spin_lock_irqsave(&imem->lock, flags); @@ -320,9 +320,6 @@ gk20a_instobj_dtor_iommu(struct nvkm_memory *memory) spin_unlock_irqrestore(&imem->lock, flags); - r = list_first_entry(&node->base.mem.regions, struct nvkm_mm_node, - rl_entry); - /* clear IOMMU bit to unmap pages */ r->offset &= ~BIT(imem->iommu_bit - imem->iommu_pgshift); @@ -404,10 +401,7 @@ gk20a_instobj_ctor_dma(struct gk20a_instmem *imem, u32 npages, u32 align, node->r.length = (npages << PAGE_SHIFT) >> 12; node->base.mem.offset = node->handle; - - INIT_LIST_HEAD(&node->base.mem.regions); - list_add_tail(&node->r.rl_entry, &node->base.mem.regions); - + node->base.mem.mem = &node->r; return 0; } @@ -484,10 +478,7 @@ gk20a_instobj_ctor_iommu(struct gk20a_instmem *imem, u32 npages, u32 align, r->offset |= BIT(imem->iommu_bit - imem->iommu_pgshift); node->base.mem.offset = ((u64)r->offset) << imem->iommu_pgshift; - - INIT_LIST_HEAD(&node->base.mem.regions); - list_add_tail(&r->rl_entry, &node->base.mem.regions); - + node->base.mem.mem = r; return 0; release_area: diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/ltc/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/ltc/base.c index 39c2a38e54f7..0c7ef250dcaf 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/ltc/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/ltc/base.c @@ -47,8 +47,10 @@ nvkm_ltc_tags_clear(struct nvkm_ltc *ltc, u32 first, u32 count) BUG_ON((first > limit) || (limit >= ltc->num_tags)); + mutex_lock(<c->subdev.mutex); ltc->func->cbc_clear(ltc, first, limit); ltc->func->cbc_wait(ltc); + mutex_unlock(<c->subdev.mutex); } int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mc/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mc/base.c index 6b25e25f9eba..09f669ac6630 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/mc/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mc/base.c @@ -161,6 +161,16 @@ nvkm_mc_enable(struct nvkm_device *device, enum nvkm_devidx devidx) } } +bool +nvkm_mc_enabled(struct nvkm_device *device, enum nvkm_devidx devidx) +{ + u64 pmc_enable = nvkm_mc_reset_mask(device, false, devidx); + + return (pmc_enable != 0) && + ((nvkm_rd32(device, 0x000200) & pmc_enable) == pmc_enable); +} + + static int nvkm_mc_fini(struct nvkm_subdev *subdev, bool suspend) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mc/g84.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mc/g84.c index c3d66ef5dc12..430a61c3df44 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/mc/g84.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mc/g84.c @@ -34,7 +34,7 @@ g84_mc_reset[] = { {} }; -const struct nvkm_mc_map +static const struct nvkm_mc_map g84_mc_intr[] = { { 0x04000000, NVKM_ENGINE_DISP }, { 0x00020000, NVKM_ENGINE_VP }, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/base.c index 5df9669ea39c..d06ad2c372bf 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/base.c @@ -31,7 +31,7 @@ nvkm_vm_map_at(struct nvkm_vma *vma, u64 delta, struct nvkm_mem *node) { struct nvkm_vm *vm = vma->vm; struct nvkm_mmu *mmu = vm->mmu; - struct nvkm_mm_node *r; + struct nvkm_mm_node *r = node->mem; int big = vma->node->type != mmu->func->spg_shift; u32 offset = vma->node->offset + (delta >> 12); u32 bits = vma->node->type - 12; @@ -41,7 +41,7 @@ nvkm_vm_map_at(struct nvkm_vma *vma, u64 delta, struct nvkm_mem *node) u32 end, len; delta = 0; - list_for_each_entry(r, &node->regions, rl_entry) { + while (r) { u64 phys = (u64)r->offset << 12; u32 num = r->length >> bits; @@ -65,7 +65,8 @@ nvkm_vm_map_at(struct nvkm_vma *vma, u64 delta, struct nvkm_mem *node) delta += (u64)len << vma->node->type; } - } + r = r->next; + }; mmu->func->flush(vm); } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/base.c index 21b65ee254e4..e3e2f5e83815 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/base.c @@ -250,6 +250,10 @@ nvkm_mxm_new_(struct nvkm_device *device, int index, struct nvkm_mxm **pmxm) } nvkm_info(&mxm->subdev, "BIOS version %d.%d\n", ver >> 4, ver & 0x0f); + nvkm_debug(&mxm->subdev, "module flags: %02x\n", + nvbios_rd08(bios, data + 0x01)); + nvkm_debug(&mxm->subdev, "config flags: %02x\n", + nvbios_rd08(bios, data + 0x02)); if (mxm_shadow(mxm, ver)) { nvkm_warn(&mxm->subdev, "failed to locate valid SIS\n"); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/mxms.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/mxms.c index 45a2f8e784f9..9abfa5e2fe9f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/mxms.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/mxms.c @@ -23,8 +23,8 @@ */ #include "mxms.h" -#define ROM16(x) le16_to_cpu(*(u16 *)&(x)) -#define ROM32(x) le32_to_cpu(*(u32 *)&(x)) +#define ROM16(x) get_unaligned_le16(&(x)) +#define ROM32(x) get_unaligned_le32(&(x)) static u8 * mxms_data(struct nvkm_mxm *mxm) diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/nv50.c index db14fad2ddfc..844971e5e874 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mxm/nv50.c @@ -190,8 +190,8 @@ mxm_dcb_sanitise(struct nvkm_mxm *mxm) struct nvkm_bios *bios = subdev->device->bios; u8 ver, hdr, cnt, len; u16 dcb = dcb_table(bios, &ver, &hdr, &cnt, &len); - if (dcb == 0x0000 || ver != 0x40) { - nvkm_debug(subdev, "unsupported DCB version\n"); + if (dcb == 0x0000 || (ver != 0x40 && ver != 0x41)) { + nvkm_warn(subdev, "unsupported DCB version\n"); return; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/Kbuild index 2a31b7d66a6d..87bf41cef0c6 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/Kbuild @@ -6,6 +6,7 @@ nvkm-y += nvkm/subdev/pci/nv40.o nvkm-y += nvkm/subdev/pci/nv46.o nvkm-y += nvkm/subdev/pci/nv4c.o nvkm-y += nvkm/subdev/pci/g84.o +nvkm-y += nvkm/subdev/pci/g92.o nvkm-y += nvkm/subdev/pci/g94.o nvkm-y += nvkm/subdev/pci/gf100.o nvkm-y += nvkm/subdev/pci/gf106.o diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g92.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g92.c new file mode 100644 index 000000000000..48874359d5f6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g92.c @@ -0,0 +1,57 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs <bskeggs@redhat.com> + */ +#include "priv.h" + +int +g92_pcie_version_supported(struct nvkm_pci *pci) +{ + if ((nvkm_pci_rd32(pci, 0x460) & 0x200) == 0x200) + return 2; + return 1; +} + +static const struct nvkm_pci_func +g92_pci_func = { + .init = g84_pci_init, + .rd32 = nv40_pci_rd32, + .wr08 = nv40_pci_wr08, + .wr32 = nv40_pci_wr32, + .msi_rearm = nv46_pci_msi_rearm, + + .pcie.init = g84_pcie_init, + .pcie.set_link = g84_pcie_set_link, + + .pcie.max_speed = g84_pcie_max_speed, + .pcie.cur_speed = g84_pcie_cur_speed, + + .pcie.set_version = g84_pcie_set_version, + .pcie.version = g84_pcie_version, + .pcie.version_supported = g92_pcie_version_supported, +}; + +int +g92_pci_new(struct nvkm_device *device, int index, struct nvkm_pci **ppci) +{ + return nvkm_pci_new_(&g92_pci_func, device, index, ppci); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g94.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g94.c index 43444123bc04..09adb37a5664 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g94.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/g94.c @@ -23,14 +23,6 @@ */ #include "priv.h" -int -g94_pcie_version_supported(struct nvkm_pci *pci) -{ - if ((nvkm_pci_rd32(pci, 0x460) & 0x200) == 0x200) - return 2; - return 1; -} - static const struct nvkm_pci_func g94_pci_func = { .init = g84_pci_init, @@ -47,7 +39,7 @@ g94_pci_func = { .pcie.set_version = g84_pcie_set_version, .pcie.version = g84_pcie_version, - .pcie.version_supported = g94_pcie_version_supported, + .pcie.version_supported = g92_pcie_version_supported, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf100.c index e30ea676baf6..00a5e7d3ee9d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf100.c @@ -92,7 +92,7 @@ gf100_pci_func = { .pcie.set_version = gf100_pcie_set_version, .pcie.version = gf100_pcie_version, - .pcie.version_supported = g94_pcie_version_supported, + .pcie.version_supported = g92_pcie_version_supported, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf106.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf106.c index c3b798c5c6dd..11bf419afe3f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf106.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/gf106.c @@ -39,7 +39,7 @@ gf106_pci_func = { .pcie.set_version = gf100_pcie_set_version, .pcie.version = gf100_pcie_version, - .pcie.version_supported = g94_pcie_version_supported, + .pcie.version_supported = g92_pcie_version_supported, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/priv.h index 23de3180aae5..86921ec962d6 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pci/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pci/priv.h @@ -44,7 +44,7 @@ enum nvkm_pcie_speed g84_pcie_max_speed(struct nvkm_pci *); int g84_pcie_init(struct nvkm_pci *); int g84_pcie_set_link(struct nvkm_pci *, enum nvkm_pcie_speed, u8); -int g94_pcie_version_supported(struct nvkm_pci *); +int g92_pcie_version_supported(struct nvkm_pci *); void gf100_pcie_set_version(struct nvkm_pci *, u8); int gf100_pcie_version(struct nvkm_pci *); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/Kbuild index 88b643b8664e..ca57c1e491b0 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/Kbuild @@ -8,3 +8,6 @@ nvkm-y += nvkm/subdev/pmu/gk110.o nvkm-y += nvkm/subdev/pmu/gk208.o nvkm-y += nvkm/subdev/pmu/gk20a.o nvkm-y += nvkm/subdev/pmu/gm107.o +nvkm-y += nvkm/subdev/pmu/gm20b.o +nvkm-y += nvkm/subdev/pmu/gp100.o +nvkm-y += nvkm/subdev/pmu/gp102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/base.c index 8dd164d13043..a73f690eb4b5 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/base.c @@ -32,251 +32,121 @@ nvkm_pmu_pgob(struct nvkm_pmu *pmu, bool enable) pmu->func->pgob(pmu, enable); } -int -nvkm_pmu_send(struct nvkm_pmu *pmu, u32 reply[2], - u32 process, u32 message, u32 data0, u32 data1) -{ - struct nvkm_subdev *subdev = &pmu->subdev; - struct nvkm_device *device = subdev->device; - u32 addr; - - mutex_lock(&subdev->mutex); - /* wait for a free slot in the fifo */ - addr = nvkm_rd32(device, 0x10a4a0); - if (nvkm_msec(device, 2000, - u32 tmp = nvkm_rd32(device, 0x10a4b0); - if (tmp != (addr ^ 8)) - break; - ) < 0) { - mutex_unlock(&subdev->mutex); - return -EBUSY; - } - - /* we currently only support a single process at a time waiting - * on a synchronous reply, take the PMU mutex and tell the - * receive handler what we're waiting for - */ - if (reply) { - pmu->recv.message = message; - pmu->recv.process = process; - } - - /* acquire data segment access */ - do { - nvkm_wr32(device, 0x10a580, 0x00000001); - } while (nvkm_rd32(device, 0x10a580) != 0x00000001); - - /* write the packet */ - nvkm_wr32(device, 0x10a1c0, 0x01000000 | (((addr & 0x07) << 4) + - pmu->send.base)); - nvkm_wr32(device, 0x10a1c4, process); - nvkm_wr32(device, 0x10a1c4, message); - nvkm_wr32(device, 0x10a1c4, data0); - nvkm_wr32(device, 0x10a1c4, data1); - nvkm_wr32(device, 0x10a4a0, (addr + 1) & 0x0f); - - /* release data segment access */ - nvkm_wr32(device, 0x10a580, 0x00000000); - - /* wait for reply, if requested */ - if (reply) { - wait_event(pmu->recv.wait, (pmu->recv.process == 0)); - reply[0] = pmu->recv.data[0]; - reply[1] = pmu->recv.data[1]; - } - - mutex_unlock(&subdev->mutex); - return 0; -} - static void nvkm_pmu_recv(struct work_struct *work) { - struct nvkm_pmu *pmu = container_of(work, struct nvkm_pmu, recv.work); - struct nvkm_subdev *subdev = &pmu->subdev; - struct nvkm_device *device = subdev->device; - u32 process, message, data0, data1; - - /* nothing to do if GET == PUT */ - u32 addr = nvkm_rd32(device, 0x10a4cc); - if (addr == nvkm_rd32(device, 0x10a4c8)) - return; - - /* acquire data segment access */ - do { - nvkm_wr32(device, 0x10a580, 0x00000002); - } while (nvkm_rd32(device, 0x10a580) != 0x00000002); - - /* read the packet */ - nvkm_wr32(device, 0x10a1c0, 0x02000000 | (((addr & 0x07) << 4) + - pmu->recv.base)); - process = nvkm_rd32(device, 0x10a1c4); - message = nvkm_rd32(device, 0x10a1c4); - data0 = nvkm_rd32(device, 0x10a1c4); - data1 = nvkm_rd32(device, 0x10a1c4); - nvkm_wr32(device, 0x10a4cc, (addr + 1) & 0x0f); - - /* release data segment access */ - nvkm_wr32(device, 0x10a580, 0x00000000); - - /* wake process if it's waiting on a synchronous reply */ - if (pmu->recv.process) { - if (process == pmu->recv.process && - message == pmu->recv.message) { - pmu->recv.data[0] = data0; - pmu->recv.data[1] = data1; - pmu->recv.process = 0; - wake_up(&pmu->recv.wait); - return; - } - } + struct nvkm_pmu *pmu = container_of(work, typeof(*pmu), recv.work); + return pmu->func->recv(pmu); +} - /* right now there's no other expected responses from the engine, - * so assume that any unexpected message is an error. - */ - nvkm_warn(subdev, "%c%c%c%c %08x %08x %08x %08x\n", - (char)((process & 0x000000ff) >> 0), - (char)((process & 0x0000ff00) >> 8), - (char)((process & 0x00ff0000) >> 16), - (char)((process & 0xff000000) >> 24), - process, message, data0, data1); +int +nvkm_pmu_send(struct nvkm_pmu *pmu, u32 reply[2], + u32 process, u32 message, u32 data0, u32 data1) +{ + if (!pmu || !pmu->func->send) + return -ENODEV; + return pmu->func->send(pmu, reply, process, message, data0, data1); } static void nvkm_pmu_intr(struct nvkm_subdev *subdev) { struct nvkm_pmu *pmu = nvkm_pmu(subdev); - struct nvkm_device *device = pmu->subdev.device; - u32 disp = nvkm_rd32(device, 0x10a01c); - u32 intr = nvkm_rd32(device, 0x10a008) & disp & ~(disp >> 16); - - if (intr & 0x00000020) { - u32 stat = nvkm_rd32(device, 0x10a16c); - if (stat & 0x80000000) { - nvkm_error(subdev, "UAS fault at %06x addr %08x\n", - stat & 0x00ffffff, - nvkm_rd32(device, 0x10a168)); - nvkm_wr32(device, 0x10a16c, 0x00000000); - intr &= ~0x00000020; - } - } - - if (intr & 0x00000040) { - schedule_work(&pmu->recv.work); - nvkm_wr32(device, 0x10a004, 0x00000040); - intr &= ~0x00000040; - } - - if (intr & 0x00000080) { - nvkm_info(subdev, "wr32 %06x %08x\n", - nvkm_rd32(device, 0x10a7a0), - nvkm_rd32(device, 0x10a7a4)); - nvkm_wr32(device, 0x10a004, 0x00000080); - intr &= ~0x00000080; - } - - if (intr) { - nvkm_error(subdev, "intr %08x\n", intr); - nvkm_wr32(device, 0x10a004, intr); - } + if (!pmu->func->intr) + return; + pmu->func->intr(pmu); } static int nvkm_pmu_fini(struct nvkm_subdev *subdev, bool suspend) { struct nvkm_pmu *pmu = nvkm_pmu(subdev); - struct nvkm_device *device = pmu->subdev.device; - nvkm_wr32(device, 0x10a014, 0x00000060); + if (pmu->func->fini) + pmu->func->fini(pmu); + flush_work(&pmu->recv.work); return 0; } static int -nvkm_pmu_init(struct nvkm_subdev *subdev) +nvkm_pmu_reset(struct nvkm_pmu *pmu) { - struct nvkm_pmu *pmu = nvkm_pmu(subdev); struct nvkm_device *device = pmu->subdev.device; - int i; - /* prevent previous ucode from running, wait for idle, reset */ - nvkm_wr32(device, 0x10a014, 0x0000ffff); /* INTR_EN_CLR = ALL */ + if (!(nvkm_rd32(device, 0x000200) & 0x00002000)) + return 0; + + /* Inhibit interrupts, and wait for idle. */ + nvkm_wr32(device, 0x10a014, 0x0000ffff); nvkm_msec(device, 2000, if (!nvkm_rd32(device, 0x10a04c)) break; ); - nvkm_mask(device, 0x000200, 0x00002000, 0x00000000); - nvkm_mask(device, 0x000200, 0x00002000, 0x00002000); - nvkm_rd32(device, 0x000200); + + /* Reset. */ + pmu->func->reset(pmu); + + /* Wait for IMEM/DMEM scrubbing to be complete. */ nvkm_msec(device, 2000, if (!(nvkm_rd32(device, 0x10a10c) & 0x00000006)) break; ); - /* upload data segment */ - nvkm_wr32(device, 0x10a1c0, 0x01000000); - for (i = 0; i < pmu->func->data.size / 4; i++) - nvkm_wr32(device, 0x10a1c4, pmu->func->data.data[i]); - - /* upload code segment */ - nvkm_wr32(device, 0x10a180, 0x01000000); - for (i = 0; i < pmu->func->code.size / 4; i++) { - if ((i & 0x3f) == 0) - nvkm_wr32(device, 0x10a188, i >> 6); - nvkm_wr32(device, 0x10a184, pmu->func->code.data[i]); - } - - /* start it running */ - nvkm_wr32(device, 0x10a10c, 0x00000000); - nvkm_wr32(device, 0x10a104, 0x00000000); - nvkm_wr32(device, 0x10a100, 0x00000002); - - /* wait for valid host->pmu ring configuration */ - if (nvkm_msec(device, 2000, - if (nvkm_rd32(device, 0x10a4d0)) - break; - ) < 0) - return -EBUSY; - pmu->send.base = nvkm_rd32(device, 0x10a4d0) & 0x0000ffff; - pmu->send.size = nvkm_rd32(device, 0x10a4d0) >> 16; + return 0; +} - /* wait for valid pmu->host ring configuration */ - if (nvkm_msec(device, 2000, - if (nvkm_rd32(device, 0x10a4dc)) - break; - ) < 0) - return -EBUSY; - pmu->recv.base = nvkm_rd32(device, 0x10a4dc) & 0x0000ffff; - pmu->recv.size = nvkm_rd32(device, 0x10a4dc) >> 16; +static int +nvkm_pmu_preinit(struct nvkm_subdev *subdev) +{ + struct nvkm_pmu *pmu = nvkm_pmu(subdev); + return nvkm_pmu_reset(pmu); +} - nvkm_wr32(device, 0x10a010, 0x000000e0); - return 0; +static int +nvkm_pmu_init(struct nvkm_subdev *subdev) +{ + struct nvkm_pmu *pmu = nvkm_pmu(subdev); + int ret = nvkm_pmu_reset(pmu); + if (ret == 0 && pmu->func->init) + ret = pmu->func->init(pmu); + return ret; } static void * nvkm_pmu_dtor(struct nvkm_subdev *subdev) { + struct nvkm_pmu *pmu = nvkm_pmu(subdev); + nvkm_falcon_del(&pmu->falcon); return nvkm_pmu(subdev); } static const struct nvkm_subdev_func nvkm_pmu = { .dtor = nvkm_pmu_dtor, + .preinit = nvkm_pmu_preinit, .init = nvkm_pmu_init, .fini = nvkm_pmu_fini, .intr = nvkm_pmu_intr, }; int +nvkm_pmu_ctor(const struct nvkm_pmu_func *func, struct nvkm_device *device, + int index, struct nvkm_pmu *pmu) +{ + nvkm_subdev_ctor(&nvkm_pmu, device, index, &pmu->subdev); + pmu->func = func; + INIT_WORK(&pmu->recv.work, nvkm_pmu_recv); + init_waitqueue_head(&pmu->recv.wait); + return nvkm_falcon_v1_new(&pmu->subdev, "PMU", 0x10a000, &pmu->falcon); +} + +int nvkm_pmu_new_(const struct nvkm_pmu_func *func, struct nvkm_device *device, int index, struct nvkm_pmu **ppmu) { struct nvkm_pmu *pmu; if (!(pmu = *ppmu = kzalloc(sizeof(*pmu), GFP_KERNEL))) return -ENOMEM; - nvkm_subdev_ctor(&nvkm_pmu, device, index, &pmu->subdev); - pmu->func = func; - INIT_WORK(&pmu->recv.work, nvkm_pmu_recv); - init_waitqueue_head(&pmu->recv.wait); - return 0; + return nvkm_pmu_ctor(func, device, index, *ppmu); } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf100.fuc3.h index e2faccffee6f..0bcf0b307a61 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf100.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf100.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gf100_pmu_data[] = { +static uint32_t gf100_pmu_data[] = { /* 0x0000: proc_kern */ 0x52544e49, 0x00000000, @@ -916,7 +916,7 @@ uint32_t gf100_pmu_data[] = { 0x00000000, }; -uint32_t gf100_pmu_code[] = { +static uint32_t gf100_pmu_code[] = { 0x03920ef5, /* 0x0004: rd32 */ 0x07a007f1, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf119.fuc4.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf119.fuc4.h index 2d5bdc539697..fe8905666c67 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf119.fuc4.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gf119.fuc4.h @@ -1,4 +1,4 @@ -uint32_t gf119_pmu_data[] = { +static uint32_t gf119_pmu_data[] = { /* 0x0000: proc_kern */ 0x52544e49, 0x00000000, @@ -915,7 +915,7 @@ uint32_t gf119_pmu_data[] = { 0x00000000, }; -uint32_t gf119_pmu_code[] = { +static uint32_t gf119_pmu_code[] = { 0x03410ef5, /* 0x0004: rd32 */ 0x07a007f1, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gk208.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gk208.fuc5.h index 3c731ff12871..9cf4e6fc724e 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gk208.fuc5.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gk208.fuc5.h @@ -1,4 +1,4 @@ -uint32_t gk208_pmu_data[] = { +static uint32_t gk208_pmu_data[] = { /* 0x0000: proc_kern */ 0x52544e49, 0x00000000, @@ -915,7 +915,7 @@ uint32_t gk208_pmu_data[] = { 0x00000000, }; -uint32_t gk208_pmu_code[] = { +static uint32_t gk208_pmu_code[] = { 0x02f90ef5, /* 0x0004: rd32 */ 0xf607a040, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gt215.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gt215.fuc3.h index e83341815ec6..5d692425b190 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gt215.fuc3.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/fuc/gt215.fuc3.h @@ -1,4 +1,4 @@ -uint32_t gt215_pmu_data[] = { +static uint32_t gt215_pmu_data[] = { /* 0x0000: proc_kern */ 0x52544e49, 0x00000000, @@ -916,7 +916,7 @@ uint32_t gt215_pmu_data[] = { 0x00000000, }; -uint32_t gt215_pmu_code[] = { +static uint32_t gt215_pmu_code[] = { 0x03920ef5, /* 0x0004: rd32 */ 0x07a007f1, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf100.c index aeb8ccd891fc..0e36d4cb7201 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf100.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf100.c @@ -30,6 +30,12 @@ gf100_pmu = { .code.size = sizeof(gf100_pmu_code), .data.data = gf100_pmu_data, .data.size = sizeof(gf100_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf119.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf119.c index fbc88d8ecd4d..0e4ba4248b15 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gf119.c @@ -30,6 +30,12 @@ gf119_pmu = { .code.size = sizeof(gf119_pmu_code), .data.data = gf119_pmu_data, .data.size = sizeof(gf119_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk104.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk104.c index 86f9f3b13f71..2ad858d825ac 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk104.c @@ -109,6 +109,12 @@ gk104_pmu = { .code.size = sizeof(gk104_pmu_code), .data.data = gk104_pmu_data, .data.size = sizeof(gk104_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, .pgob = gk104_pmu_pgob, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk110.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk110.c index ae255247c9d1..fc4b8ecfdaeb 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk110.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk110.c @@ -88,6 +88,12 @@ gk110_pmu = { .code.size = sizeof(gk110_pmu_code), .data.data = gk110_pmu_data, .data.size = sizeof(gk110_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, .pgob = gk110_pmu_pgob, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk208.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk208.c index 3b4917637902..e9a91277683a 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk208.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk208.c @@ -30,6 +30,12 @@ gk208_pmu = { .code.size = sizeof(gk208_pmu_code), .data.data = gk208_pmu_data, .data.size = sizeof(gk208_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, .pgob = gk110_pmu_pgob, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk20a.c index f996d90c9f0d..9ca0db796cbe 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk20a.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gk20a.c @@ -19,7 +19,7 @@ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. */ -#define gk20a_pmu(p) container_of((p), struct gk20a_pmu, base.subdev) +#define gk20a_pmu(p) container_of((p), struct gk20a_pmu, base) #include "priv.h" #include <subdev/clk.h> @@ -43,9 +43,8 @@ struct gk20a_pmu { }; struct gk20a_pmu_dvfs_dev_status { - unsigned long total; - unsigned long busy; - int cur_state; + u32 total; + u32 busy; }; static int @@ -56,13 +55,12 @@ gk20a_pmu_dvfs_target(struct gk20a_pmu *pmu, int *state) return nvkm_clk_astate(clk, *state, 0, false); } -static int +static void gk20a_pmu_dvfs_get_cur_state(struct gk20a_pmu *pmu, int *state) { struct nvkm_clk *clk = pmu->base.subdev.device->clk; *state = clk->pstate; - return 0; } static int @@ -90,28 +88,26 @@ gk20a_pmu_dvfs_get_target_state(struct gk20a_pmu *pmu, *state = level; - if (level == cur_level) - return 0; - else - return 1; + return (level != cur_level); } -static int +static void gk20a_pmu_dvfs_get_dev_status(struct gk20a_pmu *pmu, struct gk20a_pmu_dvfs_dev_status *status) { - struct nvkm_device *device = pmu->base.subdev.device; - status->busy = nvkm_rd32(device, 0x10a508 + (BUSY_SLOT * 0x10)); - status->total= nvkm_rd32(device, 0x10a508 + (CLK_SLOT * 0x10)); - return 0; + struct nvkm_falcon *falcon = pmu->base.falcon; + + status->busy = nvkm_falcon_rd32(falcon, 0x508 + (BUSY_SLOT * 0x10)); + status->total= nvkm_falcon_rd32(falcon, 0x508 + (CLK_SLOT * 0x10)); } static void gk20a_pmu_dvfs_reset_dev_status(struct gk20a_pmu *pmu) { - struct nvkm_device *device = pmu->base.subdev.device; - nvkm_wr32(device, 0x10a508 + (BUSY_SLOT * 0x10), 0x80000000); - nvkm_wr32(device, 0x10a508 + (CLK_SLOT * 0x10), 0x80000000); + struct nvkm_falcon *falcon = pmu->base.falcon; + + nvkm_falcon_wr32(falcon, 0x508 + (BUSY_SLOT * 0x10), 0x80000000); + nvkm_falcon_wr32(falcon, 0x508 + (CLK_SLOT * 0x10), 0x80000000); } static void @@ -127,7 +123,7 @@ gk20a_pmu_dvfs_work(struct nvkm_alarm *alarm) struct nvkm_timer *tmr = device->timer; struct nvkm_volt *volt = device->volt; u32 utilization = 0; - int state, ret; + int state; /* * The PMU is initialized before CLK and VOLT, so we have to make sure the @@ -136,11 +132,7 @@ gk20a_pmu_dvfs_work(struct nvkm_alarm *alarm) if (!clk || !volt) goto resched; - ret = gk20a_pmu_dvfs_get_dev_status(pmu, &status); - if (ret) { - nvkm_warn(subdev, "failed to get device status\n"); - goto resched; - } + gk20a_pmu_dvfs_get_dev_status(pmu, &status); if (status.total) utilization = div_u64((u64)status.busy * 100, status.total); @@ -150,11 +142,7 @@ gk20a_pmu_dvfs_work(struct nvkm_alarm *alarm) nvkm_trace(subdev, "utilization = %d %%, avg_load = %d %%\n", utilization, data->avg_load); - ret = gk20a_pmu_dvfs_get_cur_state(pmu, &state); - if (ret) { - nvkm_warn(subdev, "failed to get current state\n"); - goto resched; - } + gk20a_pmu_dvfs_get_cur_state(pmu, &state); if (gk20a_pmu_dvfs_get_target_state(pmu, &state, data->avg_load)) { nvkm_trace(subdev, "set new state to %d\n", state); @@ -166,32 +154,36 @@ resched: nvkm_timer_alarm(tmr, 100000000, alarm); } -static int -gk20a_pmu_fini(struct nvkm_subdev *subdev, bool suspend) +static void +gk20a_pmu_fini(struct nvkm_pmu *pmu) { - struct gk20a_pmu *pmu = gk20a_pmu(subdev); - nvkm_timer_alarm_cancel(subdev->device->timer, &pmu->alarm); - return 0; -} + struct gk20a_pmu *gpmu = gk20a_pmu(pmu); + nvkm_timer_alarm_cancel(pmu->subdev.device->timer, &gpmu->alarm); -static void * -gk20a_pmu_dtor(struct nvkm_subdev *subdev) -{ - return gk20a_pmu(subdev); + nvkm_falcon_put(pmu->falcon, &pmu->subdev); } static int -gk20a_pmu_init(struct nvkm_subdev *subdev) +gk20a_pmu_init(struct nvkm_pmu *pmu) { - struct gk20a_pmu *pmu = gk20a_pmu(subdev); - struct nvkm_device *device = pmu->base.subdev.device; + struct gk20a_pmu *gpmu = gk20a_pmu(pmu); + struct nvkm_subdev *subdev = &pmu->subdev; + struct nvkm_device *device = pmu->subdev.device; + struct nvkm_falcon *falcon = pmu->falcon; + int ret; + + ret = nvkm_falcon_get(falcon, subdev); + if (ret) { + nvkm_error(subdev, "cannot acquire %s falcon!\n", falcon->name); + return ret; + } /* init pwr perf counter */ - nvkm_wr32(device, 0x10a504 + (BUSY_SLOT * 0x10), 0x00200001); - nvkm_wr32(device, 0x10a50c + (BUSY_SLOT * 0x10), 0x00000002); - nvkm_wr32(device, 0x10a50c + (CLK_SLOT * 0x10), 0x00000003); + nvkm_falcon_wr32(falcon, 0x504 + (BUSY_SLOT * 0x10), 0x00200001); + nvkm_falcon_wr32(falcon, 0x50c + (BUSY_SLOT * 0x10), 0x00000002); + nvkm_falcon_wr32(falcon, 0x50c + (CLK_SLOT * 0x10), 0x00000003); - nvkm_timer_alarm(device->timer, 2000000000, &pmu->alarm); + nvkm_timer_alarm(device->timer, 2000000000, &gpmu->alarm); return 0; } @@ -202,26 +194,26 @@ gk20a_dvfs_data= { .p_smooth = 1, }; -static const struct nvkm_subdev_func +static const struct nvkm_pmu_func gk20a_pmu = { .init = gk20a_pmu_init, .fini = gk20a_pmu_fini, - .dtor = gk20a_pmu_dtor, + .reset = gt215_pmu_reset, }; int gk20a_pmu_new(struct nvkm_device *device, int index, struct nvkm_pmu **ppmu) { - static const struct nvkm_pmu_func func = {}; struct gk20a_pmu *pmu; if (!(pmu = kzalloc(sizeof(*pmu), GFP_KERNEL))) return -ENOMEM; - pmu->base.func = &func; *ppmu = &pmu->base; - nvkm_subdev_ctor(&gk20a_pmu, device, index, &pmu->base.subdev); + nvkm_pmu_ctor(&gk20a_pmu, device, index, &pmu->base); + pmu->data = &gk20a_dvfs_data; nvkm_alarm_init(&pmu->alarm, gk20a_pmu_dvfs_work); + return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm107.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm107.c index 31b8692b4641..9a248ed75f09 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm107.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm107.c @@ -32,6 +32,12 @@ gm107_pmu = { .code.size = sizeof(gm107_pmu_code), .data.data = gm107_pmu_data, .data.size = sizeof(gm107_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm20b.c new file mode 100644 index 000000000000..0b8a1cc4a0ee --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm20b.c @@ -0,0 +1,34 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include "priv.h" + +static const struct nvkm_pmu_func +gm20b_pmu = { + .reset = gt215_pmu_reset, +}; + +int +gm20b_pmu_new(struct nvkm_device *device, int index, struct nvkm_pmu **ppmu) +{ + return nvkm_pmu_new_(&gm20b_pmu, device, index, ppmu); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp100.c new file mode 100644 index 000000000000..6c41c20c85a7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp100.c @@ -0,0 +1,35 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs <bskeggs@redhat.com> + */ +#include "priv.h" + +static const struct nvkm_pmu_func +gp100_pmu = { + .reset = gt215_pmu_reset, +}; + +int +gp100_pmu_new(struct nvkm_device *device, int index, struct nvkm_pmu **ppmu) +{ + return nvkm_pmu_new_(&gp100_pmu, device, index, ppmu); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c new file mode 100644 index 000000000000..f017352206c9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c @@ -0,0 +1,43 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs <bskeggs@redhat.com> + */ +#include "priv.h" + +static void +gp102_pmu_reset(struct nvkm_pmu *pmu) +{ + struct nvkm_device *device = pmu->subdev.device; + nvkm_mask(device, 0x10a3c0, 0x00000001, 0x00000001); + nvkm_mask(device, 0x10a3c0, 0x00000001, 0x00000000); +} + +static const struct nvkm_pmu_func +gp102_pmu = { + .reset = gp102_pmu_reset, +}; + +int +gp102_pmu_new(struct nvkm_device *device, int index, struct nvkm_pmu **ppmu) +{ + return nvkm_pmu_new_(&gp102_pmu, device, index, ppmu); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gt215.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gt215.c index 8ba7fa4ca75b..90d428b3be97 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gt215.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gt215.c @@ -24,21 +24,229 @@ #include "priv.h" #include "fuc/gt215.fuc3.h" -static void +#include <subdev/timer.h> + +int +gt215_pmu_send(struct nvkm_pmu *pmu, u32 reply[2], + u32 process, u32 message, u32 data0, u32 data1) +{ + struct nvkm_subdev *subdev = &pmu->subdev; + struct nvkm_device *device = subdev->device; + u32 addr; + + mutex_lock(&subdev->mutex); + /* wait for a free slot in the fifo */ + addr = nvkm_rd32(device, 0x10a4a0); + if (nvkm_msec(device, 2000, + u32 tmp = nvkm_rd32(device, 0x10a4b0); + if (tmp != (addr ^ 8)) + break; + ) < 0) { + mutex_unlock(&subdev->mutex); + return -EBUSY; + } + + /* we currently only support a single process at a time waiting + * on a synchronous reply, take the PMU mutex and tell the + * receive handler what we're waiting for + */ + if (reply) { + pmu->recv.message = message; + pmu->recv.process = process; + } + + /* acquire data segment access */ + do { + nvkm_wr32(device, 0x10a580, 0x00000001); + } while (nvkm_rd32(device, 0x10a580) != 0x00000001); + + /* write the packet */ + nvkm_wr32(device, 0x10a1c0, 0x01000000 | (((addr & 0x07) << 4) + + pmu->send.base)); + nvkm_wr32(device, 0x10a1c4, process); + nvkm_wr32(device, 0x10a1c4, message); + nvkm_wr32(device, 0x10a1c4, data0); + nvkm_wr32(device, 0x10a1c4, data1); + nvkm_wr32(device, 0x10a4a0, (addr + 1) & 0x0f); + + /* release data segment access */ + nvkm_wr32(device, 0x10a580, 0x00000000); + + /* wait for reply, if requested */ + if (reply) { + wait_event(pmu->recv.wait, (pmu->recv.process == 0)); + reply[0] = pmu->recv.data[0]; + reply[1] = pmu->recv.data[1]; + } + + mutex_unlock(&subdev->mutex); + return 0; +} + +void +gt215_pmu_recv(struct nvkm_pmu *pmu) +{ + struct nvkm_subdev *subdev = &pmu->subdev; + struct nvkm_device *device = subdev->device; + u32 process, message, data0, data1; + + /* nothing to do if GET == PUT */ + u32 addr = nvkm_rd32(device, 0x10a4cc); + if (addr == nvkm_rd32(device, 0x10a4c8)) + return; + + /* acquire data segment access */ + do { + nvkm_wr32(device, 0x10a580, 0x00000002); + } while (nvkm_rd32(device, 0x10a580) != 0x00000002); + + /* read the packet */ + nvkm_wr32(device, 0x10a1c0, 0x02000000 | (((addr & 0x07) << 4) + + pmu->recv.base)); + process = nvkm_rd32(device, 0x10a1c4); + message = nvkm_rd32(device, 0x10a1c4); + data0 = nvkm_rd32(device, 0x10a1c4); + data1 = nvkm_rd32(device, 0x10a1c4); + nvkm_wr32(device, 0x10a4cc, (addr + 1) & 0x0f); + + /* release data segment access */ + nvkm_wr32(device, 0x10a580, 0x00000000); + + /* wake process if it's waiting on a synchronous reply */ + if (pmu->recv.process) { + if (process == pmu->recv.process && + message == pmu->recv.message) { + pmu->recv.data[0] = data0; + pmu->recv.data[1] = data1; + pmu->recv.process = 0; + wake_up(&pmu->recv.wait); + return; + } + } + + /* right now there's no other expected responses from the engine, + * so assume that any unexpected message is an error. + */ + nvkm_warn(subdev, "%c%c%c%c %08x %08x %08x %08x\n", + (char)((process & 0x000000ff) >> 0), + (char)((process & 0x0000ff00) >> 8), + (char)((process & 0x00ff0000) >> 16), + (char)((process & 0xff000000) >> 24), + process, message, data0, data1); +} + +void +gt215_pmu_intr(struct nvkm_pmu *pmu) +{ + struct nvkm_subdev *subdev = &pmu->subdev; + struct nvkm_device *device = subdev->device; + u32 disp = nvkm_rd32(device, 0x10a01c); + u32 intr = nvkm_rd32(device, 0x10a008) & disp & ~(disp >> 16); + + if (intr & 0x00000020) { + u32 stat = nvkm_rd32(device, 0x10a16c); + if (stat & 0x80000000) { + nvkm_error(subdev, "UAS fault at %06x addr %08x\n", + stat & 0x00ffffff, + nvkm_rd32(device, 0x10a168)); + nvkm_wr32(device, 0x10a16c, 0x00000000); + intr &= ~0x00000020; + } + } + + if (intr & 0x00000040) { + schedule_work(&pmu->recv.work); + nvkm_wr32(device, 0x10a004, 0x00000040); + intr &= ~0x00000040; + } + + if (intr & 0x00000080) { + nvkm_info(subdev, "wr32 %06x %08x\n", + nvkm_rd32(device, 0x10a7a0), + nvkm_rd32(device, 0x10a7a4)); + nvkm_wr32(device, 0x10a004, 0x00000080); + intr &= ~0x00000080; + } + + if (intr) { + nvkm_error(subdev, "intr %08x\n", intr); + nvkm_wr32(device, 0x10a004, intr); + } +} + +void +gt215_pmu_fini(struct nvkm_pmu *pmu) +{ + nvkm_wr32(pmu->subdev.device, 0x10a014, 0x00000060); +} + +void gt215_pmu_reset(struct nvkm_pmu *pmu) { struct nvkm_device *device = pmu->subdev.device; - nvkm_mask(device, 0x022210, 0x00000001, 0x00000000); - nvkm_mask(device, 0x022210, 0x00000001, 0x00000001); + nvkm_mask(device, 0x000200, 0x00002000, 0x00000000); + nvkm_mask(device, 0x000200, 0x00002000, 0x00002000); + nvkm_rd32(device, 0x000200); +} + +int +gt215_pmu_init(struct nvkm_pmu *pmu) +{ + struct nvkm_device *device = pmu->subdev.device; + int i; + + /* upload data segment */ + nvkm_wr32(device, 0x10a1c0, 0x01000000); + for (i = 0; i < pmu->func->data.size / 4; i++) + nvkm_wr32(device, 0x10a1c4, pmu->func->data.data[i]); + + /* upload code segment */ + nvkm_wr32(device, 0x10a180, 0x01000000); + for (i = 0; i < pmu->func->code.size / 4; i++) { + if ((i & 0x3f) == 0) + nvkm_wr32(device, 0x10a188, i >> 6); + nvkm_wr32(device, 0x10a184, pmu->func->code.data[i]); + } + + /* start it running */ + nvkm_wr32(device, 0x10a10c, 0x00000000); + nvkm_wr32(device, 0x10a104, 0x00000000); + nvkm_wr32(device, 0x10a100, 0x00000002); + + /* wait for valid host->pmu ring configuration */ + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x10a4d0)) + break; + ) < 0) + return -EBUSY; + pmu->send.base = nvkm_rd32(device, 0x10a4d0) & 0x0000ffff; + pmu->send.size = nvkm_rd32(device, 0x10a4d0) >> 16; + + /* wait for valid pmu->host ring configuration */ + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x10a4dc)) + break; + ) < 0) + return -EBUSY; + pmu->recv.base = nvkm_rd32(device, 0x10a4dc) & 0x0000ffff; + pmu->recv.size = nvkm_rd32(device, 0x10a4dc) >> 16; + + nvkm_wr32(device, 0x10a010, 0x000000e0); + return 0; } static const struct nvkm_pmu_func gt215_pmu = { - .reset = gt215_pmu_reset, .code.data = gt215_pmu_code, .code.size = sizeof(gt215_pmu_code), .data.data = gt215_pmu_data, .data.size = sizeof(gt215_pmu_data), + .reset = gt215_pmu_reset, + .init = gt215_pmu_init, + .fini = gt215_pmu_fini, + .intr = gt215_pmu_intr, + .send = gt215_pmu_send, + .recv = gt215_pmu_recv, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h index f38c88fae3d6..096cba069f72 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h @@ -4,12 +4,12 @@ #include <subdev/pmu.h> #include <subdev/pmu/fuc/os.h> +int nvkm_pmu_ctor(const struct nvkm_pmu_func *, struct nvkm_device *, + int index, struct nvkm_pmu *); int nvkm_pmu_new_(const struct nvkm_pmu_func *, struct nvkm_device *, int index, struct nvkm_pmu **); struct nvkm_pmu_func { - void (*reset)(struct nvkm_pmu *); - struct { u32 *data; u32 size; @@ -20,8 +20,22 @@ struct nvkm_pmu_func { u32 size; } data; + void (*reset)(struct nvkm_pmu *); + int (*init)(struct nvkm_pmu *); + void (*fini)(struct nvkm_pmu *); + void (*intr)(struct nvkm_pmu *); + int (*send)(struct nvkm_pmu *, u32 reply[2], u32 process, + u32 message, u32 data0, u32 data1); + void (*recv)(struct nvkm_pmu *); void (*pgob)(struct nvkm_pmu *, bool); }; +void gt215_pmu_reset(struct nvkm_pmu *); +int gt215_pmu_init(struct nvkm_pmu *); +void gt215_pmu_fini(struct nvkm_pmu *); +void gt215_pmu_intr(struct nvkm_pmu *); +void gt215_pmu_recv(struct nvkm_pmu *); +int gt215_pmu_send(struct nvkm_pmu *, u32[2], u32, u32, u32, u32); + void gk110_pmu_pgob(struct nvkm_pmu *, bool); #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/Kbuild index b02b868a6589..5076d1500f47 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/Kbuild @@ -1,3 +1,7 @@ nvkm-y += nvkm/subdev/secboot/base.o +nvkm-y += nvkm/subdev/secboot/ls_ucode_gr.o +nvkm-y += nvkm/subdev/secboot/acr.o +nvkm-y += nvkm/subdev/secboot/acr_r352.o +nvkm-y += nvkm/subdev/secboot/acr_r361.o nvkm-y += nvkm/subdev/secboot/gm200.o nvkm-y += nvkm/subdev/secboot/gm20b.o diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr.c new file mode 100644 index 000000000000..75dc06557877 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr.c @@ -0,0 +1,54 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include "acr.h" + +#include <core/firmware.h> + +/** + * Convenience function to duplicate a firmware file in memory and check that + * it has the required minimum size. + */ +void * +nvkm_acr_load_firmware(const struct nvkm_subdev *subdev, const char *name, + size_t min_size) +{ + const struct firmware *fw; + void *blob; + int ret; + + ret = nvkm_firmware_get(subdev->device, name, &fw); + if (ret) + return ERR_PTR(ret); + if (fw->size < min_size) { + nvkm_error(subdev, "%s is smaller than expected size %zu\n", + name, min_size); + nvkm_firmware_put(fw); + return ERR_PTR(-EINVAL); + } + blob = kmemdup(fw->data, fw->size, GFP_KERNEL); + nvkm_firmware_put(fw); + if (!blob) + return ERR_PTR(-ENOMEM); + + return blob; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr.h b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr.h new file mode 100644 index 000000000000..97795b342b6f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr.h @@ -0,0 +1,69 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#ifndef __NVKM_SECBOOT_ACR_H__ +#define __NVKM_SECBOOT_ACR_H__ + +#include "priv.h" + +struct nvkm_acr; + +/** + * struct nvkm_acr_func - properties and functions specific to an ACR + * + * @load: make the ACR ready to run on the given secboot device + * @reset: reset the specified falcon + * @start: start the specified falcon (assumed to have been reset) + */ +struct nvkm_acr_func { + void (*dtor)(struct nvkm_acr *); + int (*oneinit)(struct nvkm_acr *, struct nvkm_secboot *); + int (*fini)(struct nvkm_acr *, struct nvkm_secboot *, bool); + int (*load)(struct nvkm_acr *, struct nvkm_secboot *, + struct nvkm_gpuobj *, u64); + int (*reset)(struct nvkm_acr *, struct nvkm_secboot *, + enum nvkm_secboot_falcon); + int (*start)(struct nvkm_acr *, struct nvkm_secboot *, + enum nvkm_secboot_falcon); +}; + +/** + * struct nvkm_acr - instance of an ACR + * + * @boot_falcon: ID of the falcon that will perform secure boot + * @managed_falcons: bitfield of falcons managed by this ACR + * @start_address: virtual start address of the HS bootloader + */ +struct nvkm_acr { + const struct nvkm_acr_func *func; + const struct nvkm_subdev *subdev; + + enum nvkm_secboot_falcon boot_falcon; + unsigned long managed_falcons; + u32 start_address; +}; + +void *nvkm_acr_load_firmware(const struct nvkm_subdev *, const char *, size_t); + +struct nvkm_acr *acr_r352_new(unsigned long); +struct nvkm_acr *acr_r361_new(unsigned long); + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r352.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r352.c new file mode 100644 index 000000000000..1aa37ea18580 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r352.c @@ -0,0 +1,936 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include "acr_r352.h" + +#include <core/gpuobj.h> +#include <core/firmware.h> +#include <engine/falcon.h> + +/** + * struct hsf_fw_header - HS firmware descriptor + * @sig_dbg_offset: offset of the debug signature + * @sig_dbg_size: size of the debug signature + * @sig_prod_offset: offset of the production signature + * @sig_prod_size: size of the production signature + * @patch_loc: offset of the offset (sic) of where the signature is + * @patch_sig: offset of the offset (sic) to add to sig_*_offset + * @hdr_offset: offset of the load header (see struct hs_load_header) + * @hdr_size: size of above header + * + * This structure is embedded in the HS firmware image at + * hs_bin_hdr.header_offset. + */ +struct hsf_fw_header { + u32 sig_dbg_offset; + u32 sig_dbg_size; + u32 sig_prod_offset; + u32 sig_prod_size; + u32 patch_loc; + u32 patch_sig; + u32 hdr_offset; + u32 hdr_size; +}; + +/** + * struct acr_r352_flcn_bl_desc - DMEM bootloader descriptor + * @signature: 16B signature for secure code. 0s if no secure code + * @ctx_dma: DMA context to be used by BL while loading code/data + * @code_dma_base: 256B-aligned Physical FB Address where code is located + * (falcon's $xcbase register) + * @non_sec_code_off: offset from code_dma_base where the non-secure code is + * located. The offset must be multiple of 256 to help perf + * @non_sec_code_size: the size of the nonSecure code part. + * @sec_code_off: offset from code_dma_base where the secure code is + * located. The offset must be multiple of 256 to help perf + * @sec_code_size: offset from code_dma_base where the secure code is + * located. The offset must be multiple of 256 to help perf + * @code_entry_point: code entry point which will be invoked by BL after + * code is loaded. + * @data_dma_base: 256B aligned Physical FB Address where data is located. + * (falcon's $xdbase register) + * @data_size: size of data block. Should be multiple of 256B + * + * Structure used by the bootloader to load the rest of the code. This has + * to be filled by host and copied into DMEM at offset provided in the + * hsflcn_bl_desc.bl_desc_dmem_load_off. + */ +struct acr_r352_flcn_bl_desc { + u32 reserved[4]; + u32 signature[4]; + u32 ctx_dma; + u32 code_dma_base; + u32 non_sec_code_off; + u32 non_sec_code_size; + u32 sec_code_off; + u32 sec_code_size; + u32 code_entry_point; + u32 data_dma_base; + u32 data_size; + u32 code_dma_base1; + u32 data_dma_base1; +}; + +/** + * acr_r352_generate_flcn_bl_desc - generate generic BL descriptor for LS image + */ +static void +acr_r352_generate_flcn_bl_desc(const struct nvkm_acr *acr, + const struct ls_ucode_img *_img, u64 wpr_addr, + void *_desc) +{ + struct ls_ucode_img_r352 *img = ls_ucode_img_r352(_img); + struct acr_r352_flcn_bl_desc *desc = _desc; + const struct ls_ucode_img_desc *pdesc = &_img->ucode_desc; + u64 base, addr_code, addr_data; + + base = wpr_addr + img->lsb_header.ucode_off + pdesc->app_start_offset; + addr_code = (base + pdesc->app_resident_code_offset) >> 8; + addr_data = (base + pdesc->app_resident_data_offset) >> 8; + + desc->ctx_dma = FALCON_DMAIDX_UCODE; + desc->code_dma_base = lower_32_bits(addr_code); + desc->code_dma_base1 = upper_32_bits(addr_code); + desc->non_sec_code_off = pdesc->app_resident_code_offset; + desc->non_sec_code_size = pdesc->app_resident_code_size; + desc->code_entry_point = pdesc->app_imem_entry; + desc->data_dma_base = lower_32_bits(addr_data); + desc->data_dma_base1 = upper_32_bits(addr_data); + desc->data_size = pdesc->app_resident_data_size; +} + + +/** + * struct hsflcn_acr_desc - data section of the HS firmware + * + * This header is to be copied at the beginning of DMEM by the HS bootloader. + * + * @signature: signature of ACR ucode + * @wpr_region_id: region ID holding the WPR header and its details + * @wpr_offset: offset from the WPR region holding the wpr header + * @regions: region descriptors + * @nonwpr_ucode_blob_size: size of LS blob + * @nonwpr_ucode_blob_start: FB location of LS blob is + */ +struct hsflcn_acr_desc { + union { + u8 reserved_dmem[0x200]; + u32 signatures[4]; + } ucode_reserved_space; + u32 wpr_region_id; + u32 wpr_offset; + u32 mmu_mem_range; +#define FLCN_ACR_MAX_REGIONS 2 + struct { + u32 no_regions; + struct { + u32 start_addr; + u32 end_addr; + u32 region_id; + u32 read_mask; + u32 write_mask; + u32 client_mask; + } region_props[FLCN_ACR_MAX_REGIONS]; + } regions; + u32 ucode_blob_size; + u64 ucode_blob_base __aligned(8); + struct { + u32 vpr_enabled; + u32 vpr_start; + u32 vpr_end; + u32 hdcp_policies; + } vpr_desc; +}; + + +/* + * Low-secure blob creation + */ + +/** + * ls_ucode_img_load() - create a lsf_ucode_img and load it + */ +struct ls_ucode_img * +acr_r352_ls_ucode_img_load(const struct acr_r352 *acr, + enum nvkm_secboot_falcon falcon_id) +{ + const struct nvkm_subdev *subdev = acr->base.subdev; + struct ls_ucode_img_r352 *img; + int ret; + + img = kzalloc(sizeof(*img), GFP_KERNEL); + if (!img) + return ERR_PTR(-ENOMEM); + + img->base.falcon_id = falcon_id; + + ret = acr->func->ls_func[falcon_id]->load(subdev, &img->base); + + if (ret) { + kfree(img->base.ucode_data); + kfree(img->base.sig); + kfree(img); + return ERR_PTR(ret); + } + + /* Check that the signature size matches our expectations... */ + if (img->base.sig_size != sizeof(img->lsb_header.signature)) { + nvkm_error(subdev, "invalid signature size for %s falcon!\n", + nvkm_secboot_falcon_name[falcon_id]); + return ERR_PTR(-EINVAL); + } + + /* Copy signature to the right place */ + memcpy(&img->lsb_header.signature, img->base.sig, img->base.sig_size); + + /* not needed? the signature should already have the right value */ + img->lsb_header.signature.falcon_id = falcon_id; + + return &img->base; +} + +#define LSF_LSB_HEADER_ALIGN 256 +#define LSF_BL_DATA_ALIGN 256 +#define LSF_BL_DATA_SIZE_ALIGN 256 +#define LSF_BL_CODE_SIZE_ALIGN 256 +#define LSF_UCODE_DATA_ALIGN 4096 + +/** + * acr_r352_ls_img_fill_headers - fill the WPR and LSB headers of an image + * @acr: ACR to use + * @img: image to generate for + * @offset: offset in the WPR region where this image starts + * + * Allocate space in the WPR area from offset and write the WPR and LSB headers + * accordingly. + * + * Return: offset at the end of this image. + */ +static u32 +acr_r352_ls_img_fill_headers(struct acr_r352 *acr, + struct ls_ucode_img_r352 *img, u32 offset) +{ + struct ls_ucode_img *_img = &img->base; + struct acr_r352_lsf_wpr_header *whdr = &img->wpr_header; + struct acr_r352_lsf_lsb_header *lhdr = &img->lsb_header; + struct ls_ucode_img_desc *desc = &_img->ucode_desc; + const struct acr_r352_ls_func *func = + acr->func->ls_func[_img->falcon_id]; + + /* Fill WPR header */ + whdr->falcon_id = _img->falcon_id; + whdr->bootstrap_owner = acr->base.boot_falcon; + whdr->status = LSF_IMAGE_STATUS_COPY; + + /* Skip bootstrapping falcons started by someone else than ACR */ + if (acr->lazy_bootstrap & BIT(_img->falcon_id)) + whdr->lazy_bootstrap = 1; + + /* Align, save off, and include an LSB header size */ + offset = ALIGN(offset, LSF_LSB_HEADER_ALIGN); + whdr->lsb_offset = offset; + offset += sizeof(*lhdr); + + /* + * Align, save off, and include the original (static) ucode + * image size + */ + offset = ALIGN(offset, LSF_UCODE_DATA_ALIGN); + lhdr->ucode_off = offset; + offset += _img->ucode_size; + + /* + * For falcons that use a boot loader (BL), we append a loader + * desc structure on the end of the ucode image and consider + * this the boot loader data. The host will then copy the loader + * desc args to this space within the WPR region (before locking + * down) and the HS bin will then copy them to DMEM 0 for the + * loader. + */ + lhdr->bl_code_size = ALIGN(desc->bootloader_size, + LSF_BL_CODE_SIZE_ALIGN); + lhdr->ucode_size = ALIGN(desc->app_resident_data_offset, + LSF_BL_CODE_SIZE_ALIGN) + lhdr->bl_code_size; + lhdr->data_size = ALIGN(desc->app_size, LSF_BL_CODE_SIZE_ALIGN) + + lhdr->bl_code_size - lhdr->ucode_size; + /* + * Though the BL is located at 0th offset of the image, the VA + * is different to make sure that it doesn't collide the actual + * OS VA range + */ + lhdr->bl_imem_off = desc->bootloader_imem_offset; + lhdr->app_code_off = desc->app_start_offset + + desc->app_resident_code_offset; + lhdr->app_code_size = desc->app_resident_code_size; + lhdr->app_data_off = desc->app_start_offset + + desc->app_resident_data_offset; + lhdr->app_data_size = desc->app_resident_data_size; + + lhdr->flags = func->lhdr_flags; + if (_img->falcon_id == acr->base.boot_falcon) + lhdr->flags |= LSF_FLAG_DMACTL_REQ_CTX; + + /* Align and save off BL descriptor size */ + lhdr->bl_data_size = ALIGN(func->bl_desc_size, LSF_BL_DATA_SIZE_ALIGN); + + /* + * Align, save off, and include the additional BL data + */ + offset = ALIGN(offset, LSF_BL_DATA_ALIGN); + lhdr->bl_data_off = offset; + offset += lhdr->bl_data_size; + + return offset; +} + +/** + * acr_r352_ls_fill_headers - fill WPR and LSB headers of all managed images + */ +int +acr_r352_ls_fill_headers(struct acr_r352 *acr, struct list_head *imgs) +{ + struct ls_ucode_img_r352 *img; + struct list_head *l; + u32 count = 0; + u32 offset; + + /* Count the number of images to manage */ + list_for_each(l, imgs) + count++; + + /* + * Start with an array of WPR headers at the base of the WPR. + * The expectation here is that the secure falcon will do a single DMA + * read of this array and cache it internally so it's ok to pack these. + * Also, we add 1 to the falcon count to indicate the end of the array. + */ + offset = sizeof(img->wpr_header) * (count + 1); + + /* + * Walk the managed falcons, accounting for the LSB structs + * as well as the ucode images. + */ + list_for_each_entry(img, imgs, base.node) { + offset = acr_r352_ls_img_fill_headers(acr, img, offset); + } + + return offset; +} + +/** + * acr_r352_ls_write_wpr - write the WPR blob contents + */ +int +acr_r352_ls_write_wpr(struct acr_r352 *acr, struct list_head *imgs, + struct nvkm_gpuobj *wpr_blob, u32 wpr_addr) +{ + struct ls_ucode_img *_img; + u32 pos = 0; + + nvkm_kmap(wpr_blob); + + list_for_each_entry(_img, imgs, node) { + struct ls_ucode_img_r352 *img = ls_ucode_img_r352(_img); + const struct acr_r352_ls_func *ls_func = + acr->func->ls_func[_img->falcon_id]; + u8 gdesc[ls_func->bl_desc_size]; + + nvkm_gpuobj_memcpy_to(wpr_blob, pos, &img->wpr_header, + sizeof(img->wpr_header)); + + nvkm_gpuobj_memcpy_to(wpr_blob, img->wpr_header.lsb_offset, + &img->lsb_header, sizeof(img->lsb_header)); + + /* Generate and write BL descriptor */ + memset(gdesc, 0, ls_func->bl_desc_size); + ls_func->generate_bl_desc(&acr->base, _img, wpr_addr, gdesc); + + nvkm_gpuobj_memcpy_to(wpr_blob, img->lsb_header.bl_data_off, + gdesc, ls_func->bl_desc_size); + + /* Copy ucode */ + nvkm_gpuobj_memcpy_to(wpr_blob, img->lsb_header.ucode_off, + _img->ucode_data, _img->ucode_size); + + pos += sizeof(img->wpr_header); + } + + nvkm_wo32(wpr_blob, pos, NVKM_SECBOOT_FALCON_INVALID); + + nvkm_done(wpr_blob); + + return 0; +} + +/* Both size and address of WPR need to be 128K-aligned */ +#define WPR_ALIGNMENT 0x20000 +/** + * acr_r352_prepare_ls_blob() - prepare the LS blob + * + * For each securely managed falcon, load the FW, signatures and bootloaders and + * prepare a ucode blob. Then, compute the offsets in the WPR region for each + * blob, and finally write the headers and ucode blobs into a GPU object that + * will be copied into the WPR region by the HS firmware. + */ +static int +acr_r352_prepare_ls_blob(struct acr_r352 *acr, u64 wpr_addr, u32 wpr_size) +{ + const struct nvkm_subdev *subdev = acr->base.subdev; + struct list_head imgs; + struct ls_ucode_img *img, *t; + unsigned long managed_falcons = acr->base.managed_falcons; + int managed_count = 0; + u32 image_wpr_size; + int falcon_id; + int ret; + + INIT_LIST_HEAD(&imgs); + + /* Load all LS blobs */ + for_each_set_bit(falcon_id, &managed_falcons, NVKM_SECBOOT_FALCON_END) { + struct ls_ucode_img *img; + + img = acr->func->ls_ucode_img_load(acr, falcon_id); + if (IS_ERR(img)) { + ret = PTR_ERR(img); + goto cleanup; + } + + list_add_tail(&img->node, &imgs); + managed_count++; + } + + /* + * Fill the WPR and LSF headers with the right offsets and compute + * required WPR size + */ + image_wpr_size = acr->func->ls_fill_headers(acr, &imgs); + image_wpr_size = ALIGN(image_wpr_size, WPR_ALIGNMENT); + + /* Allocate GPU object that will contain the WPR region */ + ret = nvkm_gpuobj_new(subdev->device, image_wpr_size, WPR_ALIGNMENT, + false, NULL, &acr->ls_blob); + if (ret) + goto cleanup; + + nvkm_debug(subdev, "%d managed LS falcons, WPR size is %d bytes\n", + managed_count, image_wpr_size); + + /* If WPR address and size are not fixed, set them to fit the LS blob */ + if (wpr_size == 0) { + wpr_addr = acr->ls_blob->addr; + wpr_size = image_wpr_size; + /* + * But if the WPR region is set by the bootloader, it is illegal for + * the HS blob to be larger than this region. + */ + } else if (image_wpr_size > wpr_size) { + nvkm_error(subdev, "WPR region too small for FW blob!\n"); + nvkm_error(subdev, "required: %dB\n", image_wpr_size); + nvkm_error(subdev, "available: %dB\n", wpr_size); + ret = -ENOSPC; + goto cleanup; + } + + /* Write LS blob */ + ret = acr->func->ls_write_wpr(acr, &imgs, acr->ls_blob, wpr_addr); + if (ret) + nvkm_gpuobj_del(&acr->ls_blob); + +cleanup: + list_for_each_entry_safe(img, t, &imgs, node) { + kfree(img->ucode_data); + kfree(img->sig); + kfree(img); + } + + return ret; +} + + + + +/** + * acr_r352_hsf_patch_signature() - patch HS blob with correct signature + */ +static void +acr_r352_hsf_patch_signature(struct nvkm_secboot *sb, void *acr_image) +{ + struct fw_bin_header *hsbin_hdr = acr_image; + struct hsf_fw_header *fw_hdr = acr_image + hsbin_hdr->header_offset; + void *hs_data = acr_image + hsbin_hdr->data_offset; + void *sig; + u32 sig_size; + + /* Falcon in debug or production mode? */ + if (sb->boot_falcon->debug) { + sig = acr_image + fw_hdr->sig_dbg_offset; + sig_size = fw_hdr->sig_dbg_size; + } else { + sig = acr_image + fw_hdr->sig_prod_offset; + sig_size = fw_hdr->sig_prod_size; + } + + /* Patch signature */ + memcpy(hs_data + fw_hdr->patch_loc, sig + fw_hdr->patch_sig, sig_size); +} + +static void +acr_r352_fixup_hs_desc(struct acr_r352 *acr, struct nvkm_secboot *sb, + struct hsflcn_acr_desc *desc) +{ + struct nvkm_gpuobj *ls_blob = acr->ls_blob; + + /* WPR region information if WPR is not fixed */ + if (sb->wpr_size == 0) { + u32 wpr_start = ls_blob->addr; + u32 wpr_end = wpr_start + ls_blob->size; + + desc->wpr_region_id = 1; + desc->regions.no_regions = 2; + desc->regions.region_props[0].start_addr = wpr_start >> 8; + desc->regions.region_props[0].end_addr = wpr_end >> 8; + desc->regions.region_props[0].region_id = 1; + desc->regions.region_props[0].read_mask = 0xf; + desc->regions.region_props[0].write_mask = 0xc; + desc->regions.region_props[0].client_mask = 0x2; + } else { + desc->ucode_blob_base = ls_blob->addr; + desc->ucode_blob_size = ls_blob->size; + } +} + +static void +acr_r352_generate_hs_bl_desc(const struct hsf_load_header *hdr, void *_bl_desc, + u64 offset) +{ + struct acr_r352_flcn_bl_desc *bl_desc = _bl_desc; + u64 addr_code, addr_data; + + addr_code = offset >> 8; + addr_data = (offset + hdr->data_dma_base) >> 8; + + bl_desc->ctx_dma = FALCON_DMAIDX_VIRT; + bl_desc->code_dma_base = lower_32_bits(addr_code); + bl_desc->non_sec_code_off = hdr->non_sec_code_off; + bl_desc->non_sec_code_size = hdr->non_sec_code_size; + bl_desc->sec_code_off = hdr->app[0].sec_code_off; + bl_desc->sec_code_size = hdr->app[0].sec_code_size; + bl_desc->code_entry_point = 0; + bl_desc->data_dma_base = lower_32_bits(addr_data); + bl_desc->data_size = hdr->data_size; +} + +/** + * acr_r352_prepare_hs_blob - load and prepare a HS blob and BL descriptor + * + * @sb secure boot instance to prepare for + * @fw name of the HS firmware to load + * @blob pointer to gpuobj that will be allocated to receive the HS FW payload + * @bl_desc pointer to the BL descriptor to write for this firmware + * @patch whether we should patch the HS descriptor (only for HS loaders) + */ +static int +acr_r352_prepare_hs_blob(struct acr_r352 *acr, struct nvkm_secboot *sb, + const char *fw, struct nvkm_gpuobj **blob, + struct hsf_load_header *load_header, bool patch) +{ + struct nvkm_subdev *subdev = &sb->subdev; + void *acr_image; + struct fw_bin_header *hsbin_hdr; + struct hsf_fw_header *fw_hdr; + struct hsf_load_header *load_hdr; + void *acr_data; + int ret; + + acr_image = nvkm_acr_load_firmware(subdev, fw, 0); + if (IS_ERR(acr_image)) + return PTR_ERR(acr_image); + + hsbin_hdr = acr_image; + fw_hdr = acr_image + hsbin_hdr->header_offset; + load_hdr = acr_image + fw_hdr->hdr_offset; + acr_data = acr_image + hsbin_hdr->data_offset; + + /* Patch signature */ + acr_r352_hsf_patch_signature(sb, acr_image); + + /* Patch descriptor with WPR information? */ + if (patch) { + struct hsflcn_acr_desc *desc; + + desc = acr_data + load_hdr->data_dma_base; + acr_r352_fixup_hs_desc(acr, sb, desc); + } + + if (load_hdr->num_apps > ACR_R352_MAX_APPS) { + nvkm_error(subdev, "more apps (%d) than supported (%d)!", + load_hdr->num_apps, ACR_R352_MAX_APPS); + ret = -EINVAL; + goto cleanup; + } + memcpy(load_header, load_hdr, sizeof(*load_header) + + (sizeof(load_hdr->app[0]) * load_hdr->num_apps)); + + /* Create ACR blob and copy HS data to it */ + ret = nvkm_gpuobj_new(subdev->device, ALIGN(hsbin_hdr->data_size, 256), + 0x1000, false, NULL, blob); + if (ret) + goto cleanup; + + nvkm_kmap(*blob); + nvkm_gpuobj_memcpy_to(*blob, 0, acr_data, hsbin_hdr->data_size); + nvkm_done(*blob); + +cleanup: + kfree(acr_image); + + return ret; +} + +static int +acr_r352_prepare_hsbl_blob(struct acr_r352 *acr) +{ + const struct nvkm_subdev *subdev = acr->base.subdev; + struct fw_bin_header *hdr; + struct fw_bl_desc *hsbl_desc; + + acr->hsbl_blob = nvkm_acr_load_firmware(subdev, "acr/bl", 0); + if (IS_ERR(acr->hsbl_blob)) { + int ret = PTR_ERR(acr->hsbl_blob); + + acr->hsbl_blob = NULL; + return ret; + } + + hdr = acr->hsbl_blob; + hsbl_desc = acr->hsbl_blob + hdr->header_offset; + + /* virtual start address for boot vector */ + acr->base.start_address = hsbl_desc->start_tag << 8; + + return 0; +} + +/** + * acr_r352_load_blobs - load blobs common to all ACR V1 versions. + * + * This includes the LS blob, HS ucode loading blob, and HS bootloader. + * + * The HS ucode unload blob is only used on dGPU if the WPR region is variable. + */ +int +acr_r352_load_blobs(struct acr_r352 *acr, struct nvkm_secboot *sb) +{ + int ret; + + /* Firmware already loaded? */ + if (acr->firmware_ok) + return 0; + + /* Load and prepare the managed falcon's firmwares */ + ret = acr_r352_prepare_ls_blob(acr, sb->wpr_addr, sb->wpr_size); + if (ret) + return ret; + + /* Load the HS firmware that will load the LS firmwares */ + if (!acr->load_blob) { + ret = acr_r352_prepare_hs_blob(acr, sb, "acr/ucode_load", + &acr->load_blob, + &acr->load_bl_header, true); + if (ret) + return ret; + } + + /* If the ACR region is dynamically programmed, we need an unload FW */ + if (sb->wpr_size == 0) { + ret = acr_r352_prepare_hs_blob(acr, sb, "acr/ucode_unload", + &acr->unload_blob, + &acr->unload_bl_header, false); + if (ret) + return ret; + } + + /* Load the HS firmware bootloader */ + if (!acr->hsbl_blob) { + ret = acr_r352_prepare_hsbl_blob(acr); + if (ret) + return ret; + } + + acr->firmware_ok = true; + nvkm_debug(&sb->subdev, "LS blob successfully created\n"); + + return 0; +} + +/** + * acr_r352_load() - prepare HS falcon to run the specified blob, mapped + * at GPU address offset. + */ +static int +acr_r352_load(struct nvkm_acr *_acr, struct nvkm_secboot *sb, + struct nvkm_gpuobj *blob, u64 offset) +{ + struct acr_r352 *acr = acr_r352(_acr); + struct nvkm_falcon *falcon = sb->boot_falcon; + struct fw_bin_header *hdr = acr->hsbl_blob; + struct fw_bl_desc *hsbl_desc = acr->hsbl_blob + hdr->header_offset; + void *blob_data = acr->hsbl_blob + hdr->data_offset; + void *hsbl_code = blob_data + hsbl_desc->code_off; + void *hsbl_data = blob_data + hsbl_desc->data_off; + u32 code_size = ALIGN(hsbl_desc->code_size, 256); + const struct hsf_load_header *load_hdr; + const u32 bl_desc_size = acr->func->hs_bl_desc_size; + u8 bl_desc[bl_desc_size]; + + /* Find the bootloader descriptor for our blob and copy it */ + if (blob == acr->load_blob) { + load_hdr = &acr->load_bl_header; + } else if (blob == acr->unload_blob) { + load_hdr = &acr->unload_bl_header; + } else { + nvkm_error(_acr->subdev, "invalid secure boot blob!\n"); + return -EINVAL; + } + + /* + * Copy HS bootloader data + */ + nvkm_falcon_load_dmem(falcon, hsbl_data, 0x0, hsbl_desc->data_size, 0); + + /* Copy HS bootloader code to end of IMEM */ + nvkm_falcon_load_imem(falcon, hsbl_code, falcon->code.limit - code_size, + code_size, hsbl_desc->start_tag, 0, false); + + /* Generate the BL header */ + memset(bl_desc, 0, bl_desc_size); + acr->func->generate_hs_bl_desc(load_hdr, bl_desc, offset); + + /* + * Copy HS BL header where the HS descriptor expects it to be + */ + nvkm_falcon_load_dmem(falcon, bl_desc, hsbl_desc->dmem_load_off, + bl_desc_size, 0); + + return 0; +} + +static int +acr_r352_shutdown(struct acr_r352 *acr, struct nvkm_secboot *sb) +{ + int i; + + /* Run the unload blob to unprotect the WPR region */ + if (acr->unload_blob && sb->wpr_set) { + int ret; + + nvkm_debug(&sb->subdev, "running HS unload blob\n"); + ret = sb->func->run_blob(sb, acr->unload_blob); + if (ret) + return ret; + nvkm_debug(&sb->subdev, "HS unload blob completed\n"); + } + + for (i = 0; i < NVKM_SECBOOT_FALCON_END; i++) + acr->falcon_state[i] = NON_SECURE; + + sb->wpr_set = false; + + return 0; +} + +static int +acr_r352_bootstrap(struct acr_r352 *acr, struct nvkm_secboot *sb) +{ + int ret; + + if (sb->wpr_set) + return 0; + + /* Make sure all blobs are ready */ + ret = acr_r352_load_blobs(acr, sb); + if (ret) + return ret; + + nvkm_debug(&sb->subdev, "running HS load blob\n"); + ret = sb->func->run_blob(sb, acr->load_blob); + /* clear halt interrupt */ + nvkm_falcon_clear_interrupt(sb->boot_falcon, 0x10); + if (ret) + return ret; + nvkm_debug(&sb->subdev, "HS load blob completed\n"); + + sb->wpr_set = true; + + return 0; +} + +/* + * acr_r352_reset() - execute secure boot from the prepared state + * + * Load the HS bootloader and ask the falcon to run it. This will in turn + * load the HS firmware and run it, so once the falcon stops all the managed + * falcons should have their LS firmware loaded and be ready to run. + */ +static int +acr_r352_reset(struct nvkm_acr *_acr, struct nvkm_secboot *sb, + enum nvkm_secboot_falcon falcon) +{ + struct acr_r352 *acr = acr_r352(_acr); + int ret; + + /* + * Dummy GM200 implementation: perform secure boot each time we are + * called on FECS. Since only FECS and GPCCS are managed and started + * together, this ought to be safe. + * + * Once we have proper PMU firmware and support, this will be changed + * to a proper call to the PMU method. + */ + if (falcon != NVKM_SECBOOT_FALCON_FECS) + goto end; + + ret = acr_r352_shutdown(acr, sb); + if (ret) + return ret; + + acr_r352_bootstrap(acr, sb); + if (ret) + return ret; + +end: + acr->falcon_state[falcon] = RESET; + return 0; +} + +static int +acr_r352_start(struct nvkm_acr *_acr, struct nvkm_secboot *sb, + enum nvkm_secboot_falcon falcon) +{ + struct acr_r352 *acr = acr_r352(_acr); + const struct nvkm_subdev *subdev = &sb->subdev; + int base; + + switch (falcon) { + case NVKM_SECBOOT_FALCON_FECS: + base = 0x409000; + break; + case NVKM_SECBOOT_FALCON_GPCCS: + base = 0x41a000; + break; + default: + nvkm_error(subdev, "cannot start unhandled falcon!\n"); + return -EINVAL; + } + + nvkm_wr32(subdev->device, base + 0x130, 0x00000002); + acr->falcon_state[falcon] = RUNNING; + + return 0; +} + +static int +acr_r352_fini(struct nvkm_acr *_acr, struct nvkm_secboot *sb, bool suspend) +{ + struct acr_r352 *acr = acr_r352(_acr); + + return acr_r352_shutdown(acr, sb); +} + +static void +acr_r352_dtor(struct nvkm_acr *_acr) +{ + struct acr_r352 *acr = acr_r352(_acr); + + nvkm_gpuobj_del(&acr->unload_blob); + + kfree(acr->hsbl_blob); + nvkm_gpuobj_del(&acr->load_blob); + nvkm_gpuobj_del(&acr->ls_blob); + + kfree(acr); +} + +const struct acr_r352_ls_func +acr_r352_ls_fecs_func = { + .load = acr_ls_ucode_load_fecs, + .generate_bl_desc = acr_r352_generate_flcn_bl_desc, + .bl_desc_size = sizeof(struct acr_r352_flcn_bl_desc), +}; + +const struct acr_r352_ls_func +acr_r352_ls_gpccs_func = { + .load = acr_ls_ucode_load_gpccs, + .generate_bl_desc = acr_r352_generate_flcn_bl_desc, + .bl_desc_size = sizeof(struct acr_r352_flcn_bl_desc), + /* GPCCS will be loaded using PRI */ + .lhdr_flags = LSF_FLAG_FORCE_PRIV_LOAD, +}; + +const struct acr_r352_func +acr_r352_func = { + .generate_hs_bl_desc = acr_r352_generate_hs_bl_desc, + .hs_bl_desc_size = sizeof(struct acr_r352_flcn_bl_desc), + .ls_ucode_img_load = acr_r352_ls_ucode_img_load, + .ls_fill_headers = acr_r352_ls_fill_headers, + .ls_write_wpr = acr_r352_ls_write_wpr, + .ls_func = { + [NVKM_SECBOOT_FALCON_FECS] = &acr_r352_ls_fecs_func, + [NVKM_SECBOOT_FALCON_GPCCS] = &acr_r352_ls_gpccs_func, + }, +}; + +static const struct nvkm_acr_func +acr_r352_base_func = { + .dtor = acr_r352_dtor, + .fini = acr_r352_fini, + .load = acr_r352_load, + .reset = acr_r352_reset, + .start = acr_r352_start, +}; + +struct nvkm_acr * +acr_r352_new_(const struct acr_r352_func *func, + enum nvkm_secboot_falcon boot_falcon, + unsigned long managed_falcons) +{ + struct acr_r352 *acr; + + acr = kzalloc(sizeof(*acr), GFP_KERNEL); + if (!acr) + return ERR_PTR(-ENOMEM); + + acr->base.boot_falcon = boot_falcon; + acr->base.managed_falcons = managed_falcons; + acr->base.func = &acr_r352_base_func; + acr->func = func; + + return &acr->base; +} + +struct nvkm_acr * +acr_r352_new(unsigned long managed_falcons) +{ + return acr_r352_new_(&acr_r352_func, NVKM_SECBOOT_FALCON_PMU, + managed_falcons); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r352.h b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r352.h new file mode 100644 index 000000000000..ad5923b0fd3c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r352.h @@ -0,0 +1,250 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#ifndef __NVKM_SECBOOT_ACR_R352_H__ +#define __NVKM_SECBOOT_ACR_R352_H__ + +#include "acr.h" +#include "ls_ucode.h" + +struct ls_ucode_img; + +#define ACR_R352_MAX_APPS 8 + +/* + * + * LS blob structures + * + */ + +/** + * struct acr_r352_lsf_lsb_header - LS firmware header + * @signature: signature to verify the firmware against + * @ucode_off: offset of the ucode blob in the WPR region. The ucode + * blob contains the bootloader, code and data of the + * LS falcon + * @ucode_size: size of the ucode blob, including bootloader + * @data_size: size of the ucode blob data + * @bl_code_size: size of the bootloader code + * @bl_imem_off: offset in imem of the bootloader + * @bl_data_off: offset of the bootloader data in WPR region + * @bl_data_size: size of the bootloader data + * @app_code_off: offset of the app code relative to ucode_off + * @app_code_size: size of the app code + * @app_data_off: offset of the app data relative to ucode_off + * @app_data_size: size of the app data + * @flags: flags for the secure bootloader + * + * This structure is written into the WPR region for each managed falcon. Each + * instance is referenced by the lsb_offset member of the corresponding + * lsf_wpr_header. + */ +struct acr_r352_lsf_lsb_header { + /** + * LS falcon signatures + * @prd_keys: signature to use in production mode + * @dgb_keys: signature to use in debug mode + * @b_prd_present: whether the production key is present + * @b_dgb_present: whether the debug key is present + * @falcon_id: ID of the falcon the ucode applies to + */ + struct { + u8 prd_keys[2][16]; + u8 dbg_keys[2][16]; + u32 b_prd_present; + u32 b_dbg_present; + u32 falcon_id; + } signature; + u32 ucode_off; + u32 ucode_size; + u32 data_size; + u32 bl_code_size; + u32 bl_imem_off; + u32 bl_data_off; + u32 bl_data_size; + u32 app_code_off; + u32 app_code_size; + u32 app_data_off; + u32 app_data_size; + u32 flags; +#define LSF_FLAG_LOAD_CODE_AT_0 1 +#define LSF_FLAG_DMACTL_REQ_CTX 4 +#define LSF_FLAG_FORCE_PRIV_LOAD 8 +}; + +/** + * struct acr_r352_lsf_wpr_header - LS blob WPR Header + * @falcon_id: LS falcon ID + * @lsb_offset: offset of the lsb_lsf_header in the WPR region + * @bootstrap_owner: secure falcon reponsible for bootstrapping the LS falcon + * @lazy_bootstrap: skip bootstrapping by ACR + * @status: bootstrapping status + * + * An array of these is written at the beginning of the WPR region, one for + * each managed falcon. The array is terminated by an instance which falcon_id + * is LSF_FALCON_ID_INVALID. + */ +struct acr_r352_lsf_wpr_header { + u32 falcon_id; + u32 lsb_offset; + u32 bootstrap_owner; + u32 lazy_bootstrap; + u32 status; +#define LSF_IMAGE_STATUS_NONE 0 +#define LSF_IMAGE_STATUS_COPY 1 +#define LSF_IMAGE_STATUS_VALIDATION_CODE_FAILED 2 +#define LSF_IMAGE_STATUS_VALIDATION_DATA_FAILED 3 +#define LSF_IMAGE_STATUS_VALIDATION_DONE 4 +#define LSF_IMAGE_STATUS_VALIDATION_SKIPPED 5 +#define LSF_IMAGE_STATUS_BOOTSTRAP_READY 6 +}; + +/** + * struct ls_ucode_img_r352 - ucode image augmented with r352 headers + */ +struct ls_ucode_img_r352 { + struct ls_ucode_img base; + + struct acr_r352_lsf_wpr_header wpr_header; + struct acr_r352_lsf_lsb_header lsb_header; +}; +#define ls_ucode_img_r352(i) container_of(i, struct ls_ucode_img_r352, base) + + +/* + * HS blob structures + */ + +struct hsf_load_header_app { + u32 sec_code_off; + u32 sec_code_size; +}; + +/** + * struct hsf_load_header - HS firmware load header + */ +struct hsf_load_header { + u32 non_sec_code_off; + u32 non_sec_code_size; + u32 data_dma_base; + u32 data_size; + u32 num_apps; + struct hsf_load_header_app app[0]; +}; + +/** + * struct acr_r352_ls_func - manages a single LS firmware + * + * @load: load the external firmware into a ls_ucode_img + * @generate_bl_desc: function called on a block of bl_desc_size to generate the + * proper bootloader descriptor for this LS firmware + * @bl_desc_size: size of the bootloader descriptor + * @lhdr_flags: LS flags + */ +struct acr_r352_ls_func { + int (*load)(const struct nvkm_subdev *, struct ls_ucode_img *); + void (*generate_bl_desc)(const struct nvkm_acr *, + const struct ls_ucode_img *, u64, void *); + u32 bl_desc_size; + u32 lhdr_flags; +}; + +struct acr_r352; + +/** + * struct acr_r352_func - manages nuances between ACR versions + * + * @generate_hs_bl_desc: function called on a block of bl_desc_size to generate + * the proper HS bootloader descriptor + * @hs_bl_desc_size: size of the HS bootloader descriptor + */ +struct acr_r352_func { + void (*generate_hs_bl_desc)(const struct hsf_load_header *, void *, + u64); + u32 hs_bl_desc_size; + + struct ls_ucode_img *(*ls_ucode_img_load)(const struct acr_r352 *, + enum nvkm_secboot_falcon); + int (*ls_fill_headers)(struct acr_r352 *, struct list_head *); + int (*ls_write_wpr)(struct acr_r352 *, struct list_head *, + struct nvkm_gpuobj *, u32); + + const struct acr_r352_ls_func *ls_func[NVKM_SECBOOT_FALCON_END]; +}; + +/** + * struct acr_r352 - ACR data for driver release 352 (and beyond) + */ +struct acr_r352 { + struct nvkm_acr base; + const struct acr_r352_func *func; + + /* + * HS FW - lock WPR region (dGPU only) and load LS FWs + * on Tegra the HS FW copies the LS blob into the fixed WPR instead + */ + struct nvkm_gpuobj *load_blob; + struct { + struct hsf_load_header load_bl_header; + struct hsf_load_header_app __load_apps[ACR_R352_MAX_APPS]; + }; + + /* HS FW - unlock WPR region (dGPU only) */ + struct nvkm_gpuobj *unload_blob; + struct { + struct hsf_load_header unload_bl_header; + struct hsf_load_header_app __unload_apps[ACR_R352_MAX_APPS]; + }; + + /* HS bootloader */ + void *hsbl_blob; + + /* LS FWs, to be loaded by the HS ACR */ + struct nvkm_gpuobj *ls_blob; + + /* Firmware already loaded? */ + bool firmware_ok; + + /* Falcons to lazy-bootstrap */ + u32 lazy_bootstrap; + + /* To keep track of the state of all managed falcons */ + enum { + /* In non-secure state, no firmware loaded, no privileges*/ + NON_SECURE = 0, + /* In low-secure mode and ready to be started */ + RESET, + /* In low-secure mode and running */ + RUNNING, + } falcon_state[NVKM_SECBOOT_FALCON_END]; +}; +#define acr_r352(acr) container_of(acr, struct acr_r352, base) + +struct nvkm_acr *acr_r352_new_(const struct acr_r352_func *, + enum nvkm_secboot_falcon, unsigned long); + +struct ls_ucode_img *acr_r352_ls_ucode_img_load(const struct acr_r352 *, + enum nvkm_secboot_falcon); +int acr_r352_ls_fill_headers(struct acr_r352 *, struct list_head *); +int acr_r352_ls_write_wpr(struct acr_r352 *, struct list_head *, + struct nvkm_gpuobj *, u32); + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r361.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r361.c new file mode 100644 index 000000000000..f0aff1d98474 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/acr_r361.c @@ -0,0 +1,138 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include "acr_r352.h" + +#include <engine/falcon.h> + +/** + * struct acr_r361_flcn_bl_desc - DMEM bootloader descriptor + * @signature: 16B signature for secure code. 0s if no secure code + * @ctx_dma: DMA context to be used by BL while loading code/data + * @code_dma_base: 256B-aligned Physical FB Address where code is located + * (falcon's $xcbase register) + * @non_sec_code_off: offset from code_dma_base where the non-secure code is + * located. The offset must be multiple of 256 to help perf + * @non_sec_code_size: the size of the nonSecure code part. + * @sec_code_off: offset from code_dma_base where the secure code is + * located. The offset must be multiple of 256 to help perf + * @sec_code_size: offset from code_dma_base where the secure code is + * located. The offset must be multiple of 256 to help perf + * @code_entry_point: code entry point which will be invoked by BL after + * code is loaded. + * @data_dma_base: 256B aligned Physical FB Address where data is located. + * (falcon's $xdbase register) + * @data_size: size of data block. Should be multiple of 256B + * + * Structure used by the bootloader to load the rest of the code. This has + * to be filled by host and copied into DMEM at offset provided in the + * hsflcn_bl_desc.bl_desc_dmem_load_off. + */ +struct acr_r361_flcn_bl_desc { + u32 reserved[4]; + u32 signature[4]; + u32 ctx_dma; + struct flcn_u64 code_dma_base; + u32 non_sec_code_off; + u32 non_sec_code_size; + u32 sec_code_off; + u32 sec_code_size; + u32 code_entry_point; + struct flcn_u64 data_dma_base; + u32 data_size; +}; + +static void +acr_r361_generate_flcn_bl_desc(const struct nvkm_acr *acr, + const struct ls_ucode_img *_img, u64 wpr_addr, + void *_desc) +{ + struct ls_ucode_img_r352 *img = ls_ucode_img_r352(_img); + struct acr_r361_flcn_bl_desc *desc = _desc; + const struct ls_ucode_img_desc *pdesc = &img->base.ucode_desc; + u64 base, addr_code, addr_data; + + base = wpr_addr + img->lsb_header.ucode_off + pdesc->app_start_offset; + addr_code = base + pdesc->app_resident_code_offset; + addr_data = base + pdesc->app_resident_data_offset; + + desc->ctx_dma = FALCON_DMAIDX_UCODE; + desc->code_dma_base = u64_to_flcn64(addr_code); + desc->non_sec_code_off = pdesc->app_resident_code_offset; + desc->non_sec_code_size = pdesc->app_resident_code_size; + desc->code_entry_point = pdesc->app_imem_entry; + desc->data_dma_base = u64_to_flcn64(addr_data); + desc->data_size = pdesc->app_resident_data_size; +} + +static void +acr_r361_generate_hs_bl_desc(const struct hsf_load_header *hdr, void *_bl_desc, + u64 offset) +{ + struct acr_r361_flcn_bl_desc *bl_desc = _bl_desc; + + bl_desc->ctx_dma = FALCON_DMAIDX_VIRT; + bl_desc->code_dma_base = u64_to_flcn64(offset); + bl_desc->non_sec_code_off = hdr->non_sec_code_off; + bl_desc->non_sec_code_size = hdr->non_sec_code_size; + bl_desc->sec_code_off = hdr->app[0].sec_code_off; + bl_desc->sec_code_size = hdr->app[0].sec_code_size; + bl_desc->code_entry_point = 0; + bl_desc->data_dma_base = u64_to_flcn64(offset + hdr->data_dma_base); + bl_desc->data_size = hdr->data_size; +} + +const struct acr_r352_ls_func +acr_r361_ls_fecs_func = { + .load = acr_ls_ucode_load_fecs, + .generate_bl_desc = acr_r361_generate_flcn_bl_desc, + .bl_desc_size = sizeof(struct acr_r361_flcn_bl_desc), +}; + +const struct acr_r352_ls_func +acr_r361_ls_gpccs_func = { + .load = acr_ls_ucode_load_gpccs, + .generate_bl_desc = acr_r361_generate_flcn_bl_desc, + .bl_desc_size = sizeof(struct acr_r361_flcn_bl_desc), + /* GPCCS will be loaded using PRI */ + .lhdr_flags = LSF_FLAG_FORCE_PRIV_LOAD, +}; + +const struct acr_r352_func +acr_r361_func = { + .generate_hs_bl_desc = acr_r361_generate_hs_bl_desc, + .hs_bl_desc_size = sizeof(struct acr_r361_flcn_bl_desc), + .ls_ucode_img_load = acr_r352_ls_ucode_img_load, + .ls_fill_headers = acr_r352_ls_fill_headers, + .ls_write_wpr = acr_r352_ls_write_wpr, + .ls_func = { + [NVKM_SECBOOT_FALCON_FECS] = &acr_r361_ls_fecs_func, + [NVKM_SECBOOT_FALCON_GPCCS] = &acr_r361_ls_gpccs_func, + }, +}; + +struct nvkm_acr * +acr_r361_new(unsigned long managed_falcons) +{ + return acr_r352_new_(&acr_r361_func, NVKM_SECBOOT_FALCON_PMU, + managed_falcons); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/base.c index 314be2192b7d..27c9dfffb9a6 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/base.c @@ -19,184 +19,108 @@ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. */ + +/* + * Secure boot is the process by which NVIDIA-signed firmware is loaded into + * some of the falcons of a GPU. For production devices this is the only way + * for the firmware to access useful (but sensitive) registers. + * + * A Falcon microprocessor supporting advanced security modes can run in one of + * three modes: + * + * - Non-secure (NS). In this mode, functionality is similar to Falcon + * architectures before security modes were introduced (pre-Maxwell), but + * capability is restricted. In particular, certain registers may be + * inaccessible for reads and/or writes, and physical memory access may be + * disabled (on certain Falcon instances). This is the only possible mode that + * can be used if you don't have microcode cryptographically signed by NVIDIA. + * + * - Heavy Secure (HS). In this mode, the microprocessor is a black box - it's + * not possible to read or write any Falcon internal state or Falcon registers + * from outside the Falcon (for example, from the host system). The only way + * to enable this mode is by loading microcode that has been signed by NVIDIA. + * (The loading process involves tagging the IMEM block as secure, writing the + * signature into a Falcon register, and starting execution. The hardware will + * validate the signature, and if valid, grant HS privileges.) + * + * - Light Secure (LS). In this mode, the microprocessor has more privileges + * than NS but fewer than HS. Some of the microprocessor state is visible to + * host software to ease debugging. The only way to enable this mode is by HS + * microcode enabling LS mode. Some privileges available to HS mode are not + * available here. LS mode is introduced in GM20x. + * + * Secure boot consists in temporarily switching a HS-capable falcon (typically + * PMU) into HS mode in order to validate the LS firmwares of managed falcons, + * load them, and switch managed falcons into LS mode. Once secure boot + * completes, no falcon remains in HS mode. + * + * Secure boot requires a write-protected memory region (WPR) which can only be + * written by the secure falcon. On dGPU, the driver sets up the WPR region in + * video memory. On Tegra, it is set up by the bootloader and its location and + * size written into memory controller registers. + * + * The secure boot process takes place as follows: + * + * 1) A LS blob is constructed that contains all the LS firmwares we want to + * load, along with their signatures and bootloaders. + * + * 2) A HS blob (also called ACR) is created that contains the signed HS + * firmware in charge of loading the LS firmwares into their respective + * falcons. + * + * 3) The HS blob is loaded (via its own bootloader) and executed on the + * HS-capable falcon. It authenticates itself, switches the secure falcon to + * HS mode and setup the WPR region around the LS blob (dGPU) or copies the + * LS blob into the WPR region (Tegra). + * + * 4) The LS blob is now secure from all external tampering. The HS falcon + * checks the signatures of the LS firmwares and, if valid, switches the + * managed falcons to LS mode and makes them ready to run the LS firmware. + * + * 5) The managed falcons remain in LS mode and can be started. + * + */ + #include "priv.h" +#include "acr.h" #include <subdev/mc.h> #include <subdev/timer.h> +#include <subdev/pmu.h> -static const char * -managed_falcons_names[] = { +const char * +nvkm_secboot_falcon_name[] = { [NVKM_SECBOOT_FALCON_PMU] = "PMU", [NVKM_SECBOOT_FALCON_RESERVED] = "<reserved>", [NVKM_SECBOOT_FALCON_FECS] = "FECS", [NVKM_SECBOOT_FALCON_GPCCS] = "GPCCS", [NVKM_SECBOOT_FALCON_END] = "<invalid>", }; - -/* - * Helper falcon functions - */ - -static int -falcon_clear_halt_interrupt(struct nvkm_device *device, u32 base) -{ - int ret; - - /* clear halt interrupt */ - nvkm_mask(device, base + 0x004, 0x10, 0x10); - /* wait until halt interrupt is cleared */ - ret = nvkm_wait_msec(device, 10, base + 0x008, 0x10, 0x0); - if (ret < 0) - return ret; - - return 0; -} - -static int -falcon_wait_idle(struct nvkm_device *device, u32 base) -{ - int ret; - - ret = nvkm_wait_msec(device, 10, base + 0x04c, 0xffff, 0x0); - if (ret < 0) - return ret; - - return 0; -} - -static int -nvkm_secboot_falcon_enable(struct nvkm_secboot *sb) -{ - struct nvkm_device *device = sb->subdev.device; - int ret; - - /* enable engine */ - nvkm_mc_enable(device, sb->devidx); - ret = nvkm_wait_msec(device, 10, sb->base + 0x10c, 0x6, 0x0); - if (ret < 0) { - nvkm_error(&sb->subdev, "Falcon mem scrubbing timeout\n"); - nvkm_mc_disable(device, sb->devidx); - return ret; - } - - ret = falcon_wait_idle(device, sb->base); - if (ret) - return ret; - - /* enable IRQs */ - nvkm_wr32(device, sb->base + 0x010, 0xff); - nvkm_mc_intr_mask(device, sb->devidx, true); - - return 0; -} - -static int -nvkm_secboot_falcon_disable(struct nvkm_secboot *sb) -{ - struct nvkm_device *device = sb->subdev.device; - - /* disable IRQs and wait for any previous code to complete */ - nvkm_mc_intr_mask(device, sb->devidx, false); - nvkm_wr32(device, sb->base + 0x014, 0xff); - - falcon_wait_idle(device, sb->base); - - /* disable engine */ - nvkm_mc_disable(device, sb->devidx); - - return 0; -} - -int -nvkm_secboot_falcon_reset(struct nvkm_secboot *sb) -{ - int ret; - - ret = nvkm_secboot_falcon_disable(sb); - if (ret) - return ret; - - ret = nvkm_secboot_falcon_enable(sb); - if (ret) - return ret; - - return 0; -} - -/** - * nvkm_secboot_falcon_run - run the falcon that will perform secure boot - * - * This function is to be called after all chip-specific preparations have - * been completed. It will start the falcon to perform secure boot, wait for - * it to halt, and report if an error occurred. - */ -int -nvkm_secboot_falcon_run(struct nvkm_secboot *sb) -{ - struct nvkm_device *device = sb->subdev.device; - int ret; - - /* Start falcon */ - nvkm_wr32(device, sb->base + 0x100, 0x2); - - /* Wait for falcon halt */ - ret = nvkm_wait_msec(device, 100, sb->base + 0x100, 0x10, 0x10); - if (ret < 0) - return ret; - - /* If mailbox register contains an error code, then ACR has failed */ - ret = nvkm_rd32(device, sb->base + 0x040); - if (ret) { - nvkm_error(&sb->subdev, "ACR boot failed, ret 0x%08x", ret); - falcon_clear_halt_interrupt(device, sb->base); - return -EINVAL; - } - - return 0; -} - - /** * nvkm_secboot_reset() - reset specified falcon */ int -nvkm_secboot_reset(struct nvkm_secboot *sb, u32 falcon) +nvkm_secboot_reset(struct nvkm_secboot *sb, enum nvkm_secboot_falcon falcon) { /* Unmanaged falcon? */ - if (!(BIT(falcon) & sb->func->managed_falcons)) { + if (!(BIT(falcon) & sb->acr->managed_falcons)) { nvkm_error(&sb->subdev, "cannot reset unmanaged falcon!\n"); return -EINVAL; } - return sb->func->reset(sb, falcon); -} - -/** - * nvkm_secboot_start() - start specified falcon - */ -int -nvkm_secboot_start(struct nvkm_secboot *sb, u32 falcon) -{ - /* Unmanaged falcon? */ - if (!(BIT(falcon) & sb->func->managed_falcons)) { - nvkm_error(&sb->subdev, "cannot start unmanaged falcon!\n"); - return -EINVAL; - } - - return sb->func->start(sb, falcon); + return sb->acr->func->reset(sb->acr, sb, falcon); } /** * nvkm_secboot_is_managed() - check whether a given falcon is securely-managed */ bool -nvkm_secboot_is_managed(struct nvkm_secboot *secboot, - enum nvkm_secboot_falcon fid) +nvkm_secboot_is_managed(struct nvkm_secboot *sb, enum nvkm_secboot_falcon fid) { - if (!secboot) + if (!sb) return false; - return secboot->func->managed_falcons & BIT(fid); + return sb->acr->managed_falcons & BIT(fid); } static int @@ -205,9 +129,19 @@ nvkm_secboot_oneinit(struct nvkm_subdev *subdev) struct nvkm_secboot *sb = nvkm_secboot(subdev); int ret = 0; + switch (sb->acr->boot_falcon) { + case NVKM_SECBOOT_FALCON_PMU: + sb->boot_falcon = subdev->device->pmu->falcon; + break; + default: + nvkm_error(subdev, "Unmanaged boot falcon %s!\n", + nvkm_secboot_falcon_name[sb->acr->boot_falcon]); + return -EINVAL; + } + /* Call chip-specific init function */ - if (sb->func->init) - ret = sb->func->init(sb); + if (sb->func->oneinit) + ret = sb->func->oneinit(sb); if (ret) { nvkm_error(subdev, "Secure Boot initialization failed: %d\n", ret); @@ -249,7 +183,7 @@ nvkm_secboot = { }; int -nvkm_secboot_ctor(const struct nvkm_secboot_func *func, +nvkm_secboot_ctor(const struct nvkm_secboot_func *func, struct nvkm_acr *acr, struct nvkm_device *device, int index, struct nvkm_secboot *sb) { @@ -257,22 +191,14 @@ nvkm_secboot_ctor(const struct nvkm_secboot_func *func, nvkm_subdev_ctor(&nvkm_secboot, device, index, &sb->subdev); sb->func = func; - - /* setup the performing falcon's base address and masks */ - switch (func->boot_falcon) { - case NVKM_SECBOOT_FALCON_PMU: - sb->devidx = NVKM_SUBDEV_PMU; - sb->base = 0x10a000; - break; - default: - nvkm_error(&sb->subdev, "invalid secure boot falcon\n"); - return -EINVAL; - }; + sb->acr = acr; + acr->subdev = &sb->subdev; nvkm_debug(&sb->subdev, "securely managed falcons:\n"); - for_each_set_bit(fid, &sb->func->managed_falcons, + for_each_set_bit(fid, &sb->acr->managed_falcons, NVKM_SECBOOT_FALCON_END) - nvkm_debug(&sb->subdev, "- %s\n", managed_falcons_names[fid]); + nvkm_debug(&sb->subdev, "- %s\n", + nvkm_secboot_falcon_name[fid]); return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.c index f1e2dc914366..813c4eb0b25f 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.c @@ -20,1313 +20,84 @@ * DEALINGS IN THE SOFTWARE. */ -/* - * Secure boot is the process by which NVIDIA-signed firmware is loaded into - * some of the falcons of a GPU. For production devices this is the only way - * for the firmware to access useful (but sensitive) registers. - * - * A Falcon microprocessor supporting advanced security modes can run in one of - * three modes: - * - * - Non-secure (NS). In this mode, functionality is similar to Falcon - * architectures before security modes were introduced (pre-Maxwell), but - * capability is restricted. In particular, certain registers may be - * inaccessible for reads and/or writes, and physical memory access may be - * disabled (on certain Falcon instances). This is the only possible mode that - * can be used if you don't have microcode cryptographically signed by NVIDIA. - * - * - Heavy Secure (HS). In this mode, the microprocessor is a black box - it's - * not possible to read or write any Falcon internal state or Falcon registers - * from outside the Falcon (for example, from the host system). The only way - * to enable this mode is by loading microcode that has been signed by NVIDIA. - * (The loading process involves tagging the IMEM block as secure, writing the - * signature into a Falcon register, and starting execution. The hardware will - * validate the signature, and if valid, grant HS privileges.) - * - * - Light Secure (LS). In this mode, the microprocessor has more privileges - * than NS but fewer than HS. Some of the microprocessor state is visible to - * host software to ease debugging. The only way to enable this mode is by HS - * microcode enabling LS mode. Some privileges available to HS mode are not - * available here. LS mode is introduced in GM20x. - * - * Secure boot consists in temporarily switching a HS-capable falcon (typically - * PMU) into HS mode in order to validate the LS firmwares of managed falcons, - * load them, and switch managed falcons into LS mode. Once secure boot - * completes, no falcon remains in HS mode. - * - * Secure boot requires a write-protected memory region (WPR) which can only be - * written by the secure falcon. On dGPU, the driver sets up the WPR region in - * video memory. On Tegra, it is set up by the bootloader and its location and - * size written into memory controller registers. - * - * The secure boot process takes place as follows: - * - * 1) A LS blob is constructed that contains all the LS firmwares we want to - * load, along with their signatures and bootloaders. - * - * 2) A HS blob (also called ACR) is created that contains the signed HS - * firmware in charge of loading the LS firmwares into their respective - * falcons. - * - * 3) The HS blob is loaded (via its own bootloader) and executed on the - * HS-capable falcon. It authenticates itself, switches the secure falcon to - * HS mode and setup the WPR region around the LS blob (dGPU) or copies the - * LS blob into the WPR region (Tegra). - * - * 4) The LS blob is now secure from all external tampering. The HS falcon - * checks the signatures of the LS firmwares and, if valid, switches the - * managed falcons to LS mode and makes them ready to run the LS firmware. - * - * 5) The managed falcons remain in LS mode and can be started. - * - */ -#include "priv.h" +#include "acr.h" +#include "gm200.h" #include <core/gpuobj.h> -#include <core/firmware.h> #include <subdev/fb.h> - -enum { - FALCON_DMAIDX_UCODE = 0, - FALCON_DMAIDX_VIRT = 1, - FALCON_DMAIDX_PHYS_VID = 2, - FALCON_DMAIDX_PHYS_SYS_COH = 3, - FALCON_DMAIDX_PHYS_SYS_NCOH = 4, -}; - -/** - * struct fw_bin_header - header of firmware files - * @bin_magic: always 0x3b1d14f0 - * @bin_ver: version of the bin format - * @bin_size: entire image size including this header - * @header_offset: offset of the firmware/bootloader header in the file - * @data_offset: offset of the firmware/bootloader payload in the file - * @data_size: size of the payload - * - * This header is located at the beginning of the HS firmware and HS bootloader - * files, to describe where the headers and data can be found. - */ -struct fw_bin_header { - u32 bin_magic; - u32 bin_ver; - u32 bin_size; - u32 header_offset; - u32 data_offset; - u32 data_size; -}; - -/** - * struct fw_bl_desc - firmware bootloader descriptor - * @start_tag: starting tag of bootloader - * @desc_dmem_load_off: DMEM offset of flcn_bl_dmem_desc - * @code_off: offset of code section - * @code_size: size of code section - * @data_off: offset of data section - * @data_size: size of data section - * - * This structure is embedded in bootloader firmware files at to describe the - * IMEM and DMEM layout expected by the bootloader. - */ -struct fw_bl_desc { - u32 start_tag; - u32 dmem_load_off; - u32 code_off; - u32 code_size; - u32 data_off; - u32 data_size; -}; - - -/* - * - * LS blob structures - * - */ - -/** - * struct lsf_ucode_desc - LS falcon signatures - * @prd_keys: signature to use when the GPU is in production mode - * @dgb_keys: signature to use when the GPU is in debug mode - * @b_prd_present: whether the production key is present - * @b_dgb_present: whether the debug key is present - * @falcon_id: ID of the falcon the ucode applies to - * - * Directly loaded from a signature file. - */ -struct lsf_ucode_desc { - u8 prd_keys[2][16]; - u8 dbg_keys[2][16]; - u32 b_prd_present; - u32 b_dbg_present; - u32 falcon_id; -}; - -/** - * struct lsf_lsb_header - LS firmware header - * @signature: signature to verify the firmware against - * @ucode_off: offset of the ucode blob in the WPR region. The ucode - * blob contains the bootloader, code and data of the - * LS falcon - * @ucode_size: size of the ucode blob, including bootloader - * @data_size: size of the ucode blob data - * @bl_code_size: size of the bootloader code - * @bl_imem_off: offset in imem of the bootloader - * @bl_data_off: offset of the bootloader data in WPR region - * @bl_data_size: size of the bootloader data - * @app_code_off: offset of the app code relative to ucode_off - * @app_code_size: size of the app code - * @app_data_off: offset of the app data relative to ucode_off - * @app_data_size: size of the app data - * @flags: flags for the secure bootloader - * - * This structure is written into the WPR region for each managed falcon. Each - * instance is referenced by the lsb_offset member of the corresponding - * lsf_wpr_header. - */ -struct lsf_lsb_header { - struct lsf_ucode_desc signature; - u32 ucode_off; - u32 ucode_size; - u32 data_size; - u32 bl_code_size; - u32 bl_imem_off; - u32 bl_data_off; - u32 bl_data_size; - u32 app_code_off; - u32 app_code_size; - u32 app_data_off; - u32 app_data_size; - u32 flags; -#define LSF_FLAG_LOAD_CODE_AT_0 1 -#define LSF_FLAG_DMACTL_REQ_CTX 4 -#define LSF_FLAG_FORCE_PRIV_LOAD 8 -}; - -/** - * struct lsf_wpr_header - LS blob WPR Header - * @falcon_id: LS falcon ID - * @lsb_offset: offset of the lsb_lsf_header in the WPR region - * @bootstrap_owner: secure falcon reponsible for bootstrapping the LS falcon - * @lazy_bootstrap: skip bootstrapping by ACR - * @status: bootstrapping status - * - * An array of these is written at the beginning of the WPR region, one for - * each managed falcon. The array is terminated by an instance which falcon_id - * is LSF_FALCON_ID_INVALID. - */ -struct lsf_wpr_header { - u32 falcon_id; - u32 lsb_offset; - u32 bootstrap_owner; - u32 lazy_bootstrap; - u32 status; -#define LSF_IMAGE_STATUS_NONE 0 -#define LSF_IMAGE_STATUS_COPY 1 -#define LSF_IMAGE_STATUS_VALIDATION_CODE_FAILED 2 -#define LSF_IMAGE_STATUS_VALIDATION_DATA_FAILED 3 -#define LSF_IMAGE_STATUS_VALIDATION_DONE 4 -#define LSF_IMAGE_STATUS_VALIDATION_SKIPPED 5 -#define LSF_IMAGE_STATUS_BOOTSTRAP_READY 6 -}; - - -/** - * struct ls_ucode_img_desc - descriptor of firmware image - * @descriptor_size: size of this descriptor - * @image_size: size of the whole image - * @bootloader_start_offset: start offset of the bootloader in ucode image - * @bootloader_size: size of the bootloader - * @bootloader_imem_offset: start off set of the bootloader in IMEM - * @bootloader_entry_point: entry point of the bootloader in IMEM - * @app_start_offset: start offset of the LS firmware - * @app_size: size of the LS firmware's code and data - * @app_imem_offset: offset of the app in IMEM - * @app_imem_entry: entry point of the app in IMEM - * @app_dmem_offset: offset of the data in DMEM - * @app_resident_code_offset: offset of app code from app_start_offset - * @app_resident_code_size: size of the code - * @app_resident_data_offset: offset of data from app_start_offset - * @app_resident_data_size: size of data - * - * A firmware image contains the code, data, and bootloader of a given LS - * falcon in a single blob. This structure describes where everything is. - * - * This can be generated from a (bootloader, code, data) set if they have - * been loaded separately, or come directly from a file. - */ -struct ls_ucode_img_desc { - u32 descriptor_size; - u32 image_size; - u32 tools_version; - u32 app_version; - char date[64]; - u32 bootloader_start_offset; - u32 bootloader_size; - u32 bootloader_imem_offset; - u32 bootloader_entry_point; - u32 app_start_offset; - u32 app_size; - u32 app_imem_offset; - u32 app_imem_entry; - u32 app_dmem_offset; - u32 app_resident_code_offset; - u32 app_resident_code_size; - u32 app_resident_data_offset; - u32 app_resident_data_size; - u32 nb_overlays; - struct {u32 start; u32 size; } load_ovl[64]; - u32 compressed; -}; - -/** - * struct ls_ucode_img - temporary storage for loaded LS firmwares - * @node: to link within lsf_ucode_mgr - * @falcon_id: ID of the falcon this LS firmware is for - * @ucode_desc: loaded or generated map of ucode_data - * @ucode_header: header of the firmware - * @ucode_data: firmware payload (code and data) - * @ucode_size: size in bytes of data in ucode_data - * @wpr_header: WPR header to be written to the LS blob - * @lsb_header: LSB header to be written to the LS blob - * - * Preparing the WPR LS blob requires information about all the LS firmwares - * (size, etc) to be known. This structure contains all the data of one LS - * firmware. - */ -struct ls_ucode_img { - struct list_head node; - enum nvkm_secboot_falcon falcon_id; - - struct ls_ucode_img_desc ucode_desc; - u32 *ucode_header; - u8 *ucode_data; - u32 ucode_size; - - struct lsf_wpr_header wpr_header; - struct lsf_lsb_header lsb_header; -}; - -/** - * struct ls_ucode_mgr - manager for all LS falcon firmwares - * @count: number of managed LS falcons - * @wpr_size: size of the required WPR region in bytes - * @img_list: linked list of lsf_ucode_img - */ -struct ls_ucode_mgr { - u16 count; - u32 wpr_size; - struct list_head img_list; -}; - - -/* - * - * HS blob structures - * - */ - -/** - * struct hsf_fw_header - HS firmware descriptor - * @sig_dbg_offset: offset of the debug signature - * @sig_dbg_size: size of the debug signature - * @sig_prod_offset: offset of the production signature - * @sig_prod_size: size of the production signature - * @patch_loc: offset of the offset (sic) of where the signature is - * @patch_sig: offset of the offset (sic) to add to sig_*_offset - * @hdr_offset: offset of the load header (see struct hs_load_header) - * @hdr_size: size of above header - * - * This structure is embedded in the HS firmware image at - * hs_bin_hdr.header_offset. - */ -struct hsf_fw_header { - u32 sig_dbg_offset; - u32 sig_dbg_size; - u32 sig_prod_offset; - u32 sig_prod_size; - u32 patch_loc; - u32 patch_sig; - u32 hdr_offset; - u32 hdr_size; -}; - -/** - * struct hsf_load_header - HS firmware load header - */ -struct hsf_load_header { - u32 non_sec_code_off; - u32 non_sec_code_size; - u32 data_dma_base; - u32 data_size; - u32 num_apps; - struct { - u32 sec_code_off; - u32 sec_code_size; - } app[0]; -}; - -/** - * Convenience function to duplicate a firmware file in memory and check that - * it has the required minimum size. - */ -static void * -gm200_secboot_load_firmware(struct nvkm_subdev *subdev, const char *name, - size_t min_size) -{ - const struct firmware *fw; - void *blob; - int ret; - - ret = nvkm_firmware_get(subdev->device, name, &fw); - if (ret) - return ERR_PTR(ret); - if (fw->size < min_size) { - nvkm_error(subdev, "%s is smaller than expected size %zu\n", - name, min_size); - nvkm_firmware_put(fw); - return ERR_PTR(-EINVAL); - } - blob = kmemdup(fw->data, fw->size, GFP_KERNEL); - nvkm_firmware_put(fw); - if (!blob) - return ERR_PTR(-ENOMEM); - - return blob; -} - - -/* - * Low-secure blob creation - */ - -#define BL_DESC_BLK_SIZE 256 -/** - * Build a ucode image and descriptor from provided bootloader, code and data. - * - * @bl: bootloader image, including 16-bytes descriptor - * @code: LS firmware code segment - * @data: LS firmware data segment - * @desc: ucode descriptor to be written - * - * Return: allocated ucode image with corresponding descriptor information. desc - * is also updated to contain the right offsets within returned image. - */ -static void * -ls_ucode_img_build(const struct firmware *bl, const struct firmware *code, - const struct firmware *data, struct ls_ucode_img_desc *desc) -{ - struct fw_bin_header *bin_hdr = (void *)bl->data; - struct fw_bl_desc *bl_desc = (void *)bl->data + bin_hdr->header_offset; - void *bl_data = (void *)bl->data + bin_hdr->data_offset; - u32 pos = 0; - void *image; - - desc->bootloader_start_offset = pos; - desc->bootloader_size = ALIGN(bl_desc->code_size, sizeof(u32)); - desc->bootloader_imem_offset = bl_desc->start_tag * 256; - desc->bootloader_entry_point = bl_desc->start_tag * 256; - - pos = ALIGN(pos + desc->bootloader_size, BL_DESC_BLK_SIZE); - desc->app_start_offset = pos; - desc->app_size = ALIGN(code->size, BL_DESC_BLK_SIZE) + - ALIGN(data->size, BL_DESC_BLK_SIZE); - desc->app_imem_offset = 0; - desc->app_imem_entry = 0; - desc->app_dmem_offset = 0; - desc->app_resident_code_offset = 0; - desc->app_resident_code_size = ALIGN(code->size, BL_DESC_BLK_SIZE); - - pos = ALIGN(pos + desc->app_resident_code_size, BL_DESC_BLK_SIZE); - desc->app_resident_data_offset = pos - desc->app_start_offset; - desc->app_resident_data_size = ALIGN(data->size, BL_DESC_BLK_SIZE); - - desc->image_size = ALIGN(bl_desc->code_size, BL_DESC_BLK_SIZE) + - desc->app_size; - - image = kzalloc(desc->image_size, GFP_KERNEL); - if (!image) - return ERR_PTR(-ENOMEM); - - memcpy(image + desc->bootloader_start_offset, bl_data, - bl_desc->code_size); - memcpy(image + desc->app_start_offset, code->data, code->size); - memcpy(image + desc->app_start_offset + desc->app_resident_data_offset, - data->data, data->size); - - return image; -} - -/** - * ls_ucode_img_load_generic() - load and prepare a LS ucode image - * - * Load the LS microcode, bootloader and signature and pack them into a single - * blob. Also generate the corresponding ucode descriptor. - */ -static int -ls_ucode_img_load_generic(struct nvkm_subdev *subdev, - struct ls_ucode_img *img, const char *falcon_name, - const u32 falcon_id) -{ - const struct firmware *bl, *code, *data; - struct lsf_ucode_desc *lsf_desc; - char f[64]; - int ret; - - img->ucode_header = NULL; - - snprintf(f, sizeof(f), "gr/%s_bl", falcon_name); - ret = nvkm_firmware_get(subdev->device, f, &bl); - if (ret) - goto error; - - snprintf(f, sizeof(f), "gr/%s_inst", falcon_name); - ret = nvkm_firmware_get(subdev->device, f, &code); - if (ret) - goto free_bl; - - snprintf(f, sizeof(f), "gr/%s_data", falcon_name); - ret = nvkm_firmware_get(subdev->device, f, &data); - if (ret) - goto free_inst; - - img->ucode_data = ls_ucode_img_build(bl, code, data, - &img->ucode_desc); - if (IS_ERR(img->ucode_data)) { - ret = PTR_ERR(img->ucode_data); - goto free_data; - } - img->ucode_size = img->ucode_desc.image_size; - - snprintf(f, sizeof(f), "gr/%s_sig", falcon_name); - lsf_desc = gm200_secboot_load_firmware(subdev, f, sizeof(*lsf_desc)); - if (IS_ERR(lsf_desc)) { - ret = PTR_ERR(lsf_desc); - goto free_image; - } - /* not needed? the signature should already have the right value */ - lsf_desc->falcon_id = falcon_id; - memcpy(&img->lsb_header.signature, lsf_desc, sizeof(*lsf_desc)); - img->falcon_id = lsf_desc->falcon_id; - kfree(lsf_desc); - - /* success path - only free requested firmware files */ - goto free_data; - -free_image: - kfree(img->ucode_data); -free_data: - nvkm_firmware_put(data); -free_inst: - nvkm_firmware_put(code); -free_bl: - nvkm_firmware_put(bl); -error: - return ret; -} - -typedef int (*lsf_load_func)(struct nvkm_subdev *, struct ls_ucode_img *); - -static int -ls_ucode_img_load_fecs(struct nvkm_subdev *subdev, struct ls_ucode_img *img) -{ - return ls_ucode_img_load_generic(subdev, img, "fecs", - NVKM_SECBOOT_FALCON_FECS); -} - -static int -ls_ucode_img_load_gpccs(struct nvkm_subdev *subdev, struct ls_ucode_img *img) -{ - return ls_ucode_img_load_generic(subdev, img, "gpccs", - NVKM_SECBOOT_FALCON_GPCCS); -} - -/** - * ls_ucode_img_load() - create a lsf_ucode_img and load it - */ -static struct ls_ucode_img * -ls_ucode_img_load(struct nvkm_subdev *subdev, lsf_load_func load_func) -{ - struct ls_ucode_img *img; - int ret; - - img = kzalloc(sizeof(*img), GFP_KERNEL); - if (!img) - return ERR_PTR(-ENOMEM); - - ret = load_func(subdev, img); - if (ret) { - kfree(img); - return ERR_PTR(ret); - } - - return img; -} - -static const lsf_load_func lsf_load_funcs[] = { - [NVKM_SECBOOT_FALCON_END] = NULL, /* reserve enough space */ - [NVKM_SECBOOT_FALCON_FECS] = ls_ucode_img_load_fecs, - [NVKM_SECBOOT_FALCON_GPCCS] = ls_ucode_img_load_gpccs, -}; - -/** - * ls_ucode_img_populate_bl_desc() - populate a DMEM BL descriptor for LS image - * @img: ucode image to generate against - * @desc: descriptor to populate - * @sb: secure boot state to use for base addresses - * - * Populate the DMEM BL descriptor with the information contained in a - * ls_ucode_desc. - * - */ -static void -ls_ucode_img_populate_bl_desc(struct ls_ucode_img *img, u64 wpr_addr, - struct gm200_flcn_bl_desc *desc) -{ - struct ls_ucode_img_desc *pdesc = &img->ucode_desc; - u64 addr_base; - - addr_base = wpr_addr + img->lsb_header.ucode_off + - pdesc->app_start_offset; - - memset(desc, 0, sizeof(*desc)); - desc->ctx_dma = FALCON_DMAIDX_UCODE; - desc->code_dma_base.lo = lower_32_bits( - (addr_base + pdesc->app_resident_code_offset)); - desc->code_dma_base.hi = upper_32_bits( - (addr_base + pdesc->app_resident_code_offset)); - desc->non_sec_code_size = pdesc->app_resident_code_size; - desc->data_dma_base.lo = lower_32_bits( - (addr_base + pdesc->app_resident_data_offset)); - desc->data_dma_base.hi = upper_32_bits( - (addr_base + pdesc->app_resident_data_offset)); - desc->data_size = pdesc->app_resident_data_size; - desc->code_entry_point = pdesc->app_imem_entry; -} - -#define LSF_LSB_HEADER_ALIGN 256 -#define LSF_BL_DATA_ALIGN 256 -#define LSF_BL_DATA_SIZE_ALIGN 256 -#define LSF_BL_CODE_SIZE_ALIGN 256 -#define LSF_UCODE_DATA_ALIGN 4096 - -/** - * ls_ucode_img_fill_headers - fill the WPR and LSB headers of an image - * @gsb: secure boot device used - * @img: image to generate for - * @offset: offset in the WPR region where this image starts - * - * Allocate space in the WPR area from offset and write the WPR and LSB headers - * accordingly. - * - * Return: offset at the end of this image. - */ -static u32 -ls_ucode_img_fill_headers(struct gm200_secboot *gsb, struct ls_ucode_img *img, - u32 offset) -{ - struct lsf_wpr_header *whdr = &img->wpr_header; - struct lsf_lsb_header *lhdr = &img->lsb_header; - struct ls_ucode_img_desc *desc = &img->ucode_desc; - - if (img->ucode_header) { - nvkm_fatal(&gsb->base.subdev, - "images withough loader are not supported yet!\n"); - return offset; - } - - /* Fill WPR header */ - whdr->falcon_id = img->falcon_id; - whdr->bootstrap_owner = gsb->base.func->boot_falcon; - whdr->status = LSF_IMAGE_STATUS_COPY; - - /* Align, save off, and include an LSB header size */ - offset = ALIGN(offset, LSF_LSB_HEADER_ALIGN); - whdr->lsb_offset = offset; - offset += sizeof(struct lsf_lsb_header); - - /* - * Align, save off, and include the original (static) ucode - * image size - */ - offset = ALIGN(offset, LSF_UCODE_DATA_ALIGN); - lhdr->ucode_off = offset; - offset += img->ucode_size; - - /* - * For falcons that use a boot loader (BL), we append a loader - * desc structure on the end of the ucode image and consider - * this the boot loader data. The host will then copy the loader - * desc args to this space within the WPR region (before locking - * down) and the HS bin will then copy them to DMEM 0 for the - * loader. - */ - lhdr->bl_code_size = ALIGN(desc->bootloader_size, - LSF_BL_CODE_SIZE_ALIGN); - lhdr->ucode_size = ALIGN(desc->app_resident_data_offset, - LSF_BL_CODE_SIZE_ALIGN) + lhdr->bl_code_size; - lhdr->data_size = ALIGN(desc->app_size, LSF_BL_CODE_SIZE_ALIGN) + - lhdr->bl_code_size - lhdr->ucode_size; - /* - * Though the BL is located at 0th offset of the image, the VA - * is different to make sure that it doesn't collide the actual - * OS VA range - */ - lhdr->bl_imem_off = desc->bootloader_imem_offset; - lhdr->app_code_off = desc->app_start_offset + - desc->app_resident_code_offset; - lhdr->app_code_size = desc->app_resident_code_size; - lhdr->app_data_off = desc->app_start_offset + - desc->app_resident_data_offset; - lhdr->app_data_size = desc->app_resident_data_size; - - lhdr->flags = 0; - if (img->falcon_id == gsb->base.func->boot_falcon) - lhdr->flags = LSF_FLAG_DMACTL_REQ_CTX; - - /* GPCCS will be loaded using PRI */ - if (img->falcon_id == NVKM_SECBOOT_FALCON_GPCCS) - lhdr->flags |= LSF_FLAG_FORCE_PRIV_LOAD; - - /* Align (size bloat) and save off BL descriptor size */ - lhdr->bl_data_size = ALIGN(sizeof(struct gm200_flcn_bl_desc), - LSF_BL_DATA_SIZE_ALIGN); - /* - * Align, save off, and include the additional BL data - */ - offset = ALIGN(offset, LSF_BL_DATA_ALIGN); - lhdr->bl_data_off = offset; - offset += lhdr->bl_data_size; - - return offset; -} - -static void -ls_ucode_mgr_init(struct ls_ucode_mgr *mgr) -{ - memset(mgr, 0, sizeof(*mgr)); - INIT_LIST_HEAD(&mgr->img_list); -} - -static void -ls_ucode_mgr_cleanup(struct ls_ucode_mgr *mgr) -{ - struct ls_ucode_img *img, *t; - - list_for_each_entry_safe(img, t, &mgr->img_list, node) { - kfree(img->ucode_data); - kfree(img->ucode_header); - kfree(img); - } -} - -static void -ls_ucode_mgr_add_img(struct ls_ucode_mgr *mgr, struct ls_ucode_img *img) -{ - mgr->count++; - list_add_tail(&img->node, &mgr->img_list); -} - -/** - * ls_ucode_mgr_fill_headers - fill WPR and LSB headers of all managed images - */ -static void -ls_ucode_mgr_fill_headers(struct gm200_secboot *gsb, struct ls_ucode_mgr *mgr) -{ - struct ls_ucode_img *img; - u32 offset; - - /* - * Start with an array of WPR headers at the base of the WPR. - * The expectation here is that the secure falcon will do a single DMA - * read of this array and cache it internally so it's ok to pack these. - * Also, we add 1 to the falcon count to indicate the end of the array. - */ - offset = sizeof(struct lsf_wpr_header) * (mgr->count + 1); - - /* - * Walk the managed falcons, accounting for the LSB structs - * as well as the ucode images. - */ - list_for_each_entry(img, &mgr->img_list, node) { - offset = ls_ucode_img_fill_headers(gsb, img, offset); - } - - mgr->wpr_size = offset; -} +#include <engine/falcon.h> +#include <subdev/mc.h> /** - * ls_ucode_mgr_write_wpr - write the WPR blob contents - */ -static int -ls_ucode_mgr_write_wpr(struct gm200_secboot *gsb, struct ls_ucode_mgr *mgr, - struct nvkm_gpuobj *wpr_blob) -{ - struct ls_ucode_img *img; - u32 pos = 0; - - nvkm_kmap(wpr_blob); - - list_for_each_entry(img, &mgr->img_list, node) { - nvkm_gpuobj_memcpy_to(wpr_blob, pos, &img->wpr_header, - sizeof(img->wpr_header)); - - nvkm_gpuobj_memcpy_to(wpr_blob, img->wpr_header.lsb_offset, - &img->lsb_header, sizeof(img->lsb_header)); - - /* Generate and write BL descriptor */ - if (!img->ucode_header) { - u8 desc[gsb->func->bl_desc_size]; - struct gm200_flcn_bl_desc gdesc; - - ls_ucode_img_populate_bl_desc(img, gsb->wpr_addr, - &gdesc); - gsb->func->fixup_bl_desc(&gdesc, &desc); - nvkm_gpuobj_memcpy_to(wpr_blob, - img->lsb_header.bl_data_off, - &desc, gsb->func->bl_desc_size); - } - - /* Copy ucode */ - nvkm_gpuobj_memcpy_to(wpr_blob, img->lsb_header.ucode_off, - img->ucode_data, img->ucode_size); - - pos += sizeof(img->wpr_header); - } - - nvkm_wo32(wpr_blob, pos, NVKM_SECBOOT_FALCON_INVALID); - - nvkm_done(wpr_blob); - - return 0; -} - -/* Both size and address of WPR need to be 128K-aligned */ -#define WPR_ALIGNMENT 0x20000 -/** - * gm200_secboot_prepare_ls_blob() - prepare the LS blob + * gm200_secboot_run_blob() - run the given high-secure blob * - * For each securely managed falcon, load the FW, signatures and bootloaders and - * prepare a ucode blob. Then, compute the offsets in the WPR region for each - * blob, and finally write the headers and ucode blobs into a GPU object that - * will be copied into the WPR region by the HS firmware. - */ -static int -gm200_secboot_prepare_ls_blob(struct gm200_secboot *gsb) -{ - struct nvkm_secboot *sb = &gsb->base; - struct nvkm_device *device = sb->subdev.device; - struct ls_ucode_mgr mgr; - int falcon_id; - int ret; - - ls_ucode_mgr_init(&mgr); - - /* Load all LS blobs */ - for_each_set_bit(falcon_id, &gsb->base.func->managed_falcons, - NVKM_SECBOOT_FALCON_END) { - struct ls_ucode_img *img; - - img = ls_ucode_img_load(&sb->subdev, lsf_load_funcs[falcon_id]); - - if (IS_ERR(img)) { - ret = PTR_ERR(img); - goto cleanup; - } - ls_ucode_mgr_add_img(&mgr, img); - } - - /* - * Fill the WPR and LSF headers with the right offsets and compute - * required WPR size - */ - ls_ucode_mgr_fill_headers(gsb, &mgr); - mgr.wpr_size = ALIGN(mgr.wpr_size, WPR_ALIGNMENT); - - /* Allocate GPU object that will contain the WPR region */ - ret = nvkm_gpuobj_new(device, mgr.wpr_size, WPR_ALIGNMENT, false, NULL, - &gsb->ls_blob); - if (ret) - goto cleanup; - - nvkm_debug(&sb->subdev, "%d managed LS falcons, WPR size is %d bytes\n", - mgr.count, mgr.wpr_size); - - /* If WPR address and size are not fixed, set them to fit the LS blob */ - if (!gsb->wpr_size) { - gsb->wpr_addr = gsb->ls_blob->addr; - gsb->wpr_size = gsb->ls_blob->size; - } - - /* Write LS blob */ - ret = ls_ucode_mgr_write_wpr(gsb, &mgr, gsb->ls_blob); - if (ret) - nvkm_gpuobj_del(&gsb->ls_blob); - -cleanup: - ls_ucode_mgr_cleanup(&mgr); - - return ret; -} - -/* - * High-secure blob creation - */ - -/** - * gm200_secboot_hsf_patch_signature() - patch HS blob with correct signature - */ -static void -gm200_secboot_hsf_patch_signature(struct gm200_secboot *gsb, void *acr_image) -{ - struct nvkm_secboot *sb = &gsb->base; - struct fw_bin_header *hsbin_hdr = acr_image; - struct hsf_fw_header *fw_hdr = acr_image + hsbin_hdr->header_offset; - void *hs_data = acr_image + hsbin_hdr->data_offset; - void *sig; - u32 sig_size; - - /* Falcon in debug or production mode? */ - if ((nvkm_rd32(sb->subdev.device, sb->base + 0xc08) >> 20) & 0x1) { - sig = acr_image + fw_hdr->sig_dbg_offset; - sig_size = fw_hdr->sig_dbg_size; - } else { - sig = acr_image + fw_hdr->sig_prod_offset; - sig_size = fw_hdr->sig_prod_size; - } - - /* Patch signature */ - memcpy(hs_data + fw_hdr->patch_loc, sig + fw_hdr->patch_sig, sig_size); -} - -/** - * gm200_secboot_populate_hsf_bl_desc() - populate BL descriptor for HS image - */ -static void -gm200_secboot_populate_hsf_bl_desc(void *acr_image, - struct gm200_flcn_bl_desc *bl_desc) -{ - struct fw_bin_header *hsbin_hdr = acr_image; - struct hsf_fw_header *fw_hdr = acr_image + hsbin_hdr->header_offset; - struct hsf_load_header *load_hdr = acr_image + fw_hdr->hdr_offset; - - /* - * Descriptor for the bootloader that will load the ACR image into - * IMEM/DMEM memory. - */ - fw_hdr = acr_image + hsbin_hdr->header_offset; - load_hdr = acr_image + fw_hdr->hdr_offset; - memset(bl_desc, 0, sizeof(*bl_desc)); - bl_desc->ctx_dma = FALCON_DMAIDX_VIRT; - bl_desc->non_sec_code_off = load_hdr->non_sec_code_off; - bl_desc->non_sec_code_size = load_hdr->non_sec_code_size; - bl_desc->sec_code_off = load_hdr->app[0].sec_code_off; - bl_desc->sec_code_size = load_hdr->app[0].sec_code_size; - bl_desc->code_entry_point = 0; - /* - * We need to set code_dma_base to the virtual address of the acr_blob, - * and add this address to data_dma_base before writing it into DMEM - */ - bl_desc->code_dma_base.lo = 0; - bl_desc->data_dma_base.lo = load_hdr->data_dma_base; - bl_desc->data_size = load_hdr->data_size; -} - -/** - * gm200_secboot_prepare_hs_blob - load and prepare a HS blob and BL descriptor - * - * @gsb secure boot instance to prepare for - * @fw name of the HS firmware to load - * @blob pointer to gpuobj that will be allocated to receive the HS FW payload - * @bl_desc pointer to the BL descriptor to write for this firmware - * @patch whether we should patch the HS descriptor (only for HS loaders) - */ -static int -gm200_secboot_prepare_hs_blob(struct gm200_secboot *gsb, const char *fw, - struct nvkm_gpuobj **blob, - struct gm200_flcn_bl_desc *bl_desc, bool patch) -{ - struct nvkm_subdev *subdev = &gsb->base.subdev; - void *acr_image; - struct fw_bin_header *hsbin_hdr; - struct hsf_fw_header *fw_hdr; - void *acr_data; - struct hsf_load_header *load_hdr; - struct hsflcn_acr_desc *desc; - int ret; - - acr_image = gm200_secboot_load_firmware(subdev, fw, 0); - if (IS_ERR(acr_image)) - return PTR_ERR(acr_image); - hsbin_hdr = acr_image; - - /* Patch signature */ - gm200_secboot_hsf_patch_signature(gsb, acr_image); - - acr_data = acr_image + hsbin_hdr->data_offset; - - /* Patch descriptor? */ - if (patch) { - fw_hdr = acr_image + hsbin_hdr->header_offset; - load_hdr = acr_image + fw_hdr->hdr_offset; - desc = acr_data + load_hdr->data_dma_base; - gsb->func->fixup_hs_desc(gsb, desc); - } - - /* Generate HS BL descriptor */ - gm200_secboot_populate_hsf_bl_desc(acr_image, bl_desc); - - /* Create ACR blob and copy HS data to it */ - ret = nvkm_gpuobj_new(subdev->device, ALIGN(hsbin_hdr->data_size, 256), - 0x1000, false, NULL, blob); - if (ret) - goto cleanup; - - nvkm_kmap(*blob); - nvkm_gpuobj_memcpy_to(*blob, 0, acr_data, hsbin_hdr->data_size); - nvkm_done(*blob); - -cleanup: - kfree(acr_image); - - return ret; -} - -/* - * High-secure bootloader blob creation - */ - -static int -gm200_secboot_prepare_hsbl_blob(struct gm200_secboot *gsb) -{ - struct nvkm_subdev *subdev = &gsb->base.subdev; - - gsb->hsbl_blob = gm200_secboot_load_firmware(subdev, "acr/bl", 0); - if (IS_ERR(gsb->hsbl_blob)) { - int ret = PTR_ERR(gsb->hsbl_blob); - - gsb->hsbl_blob = NULL; - return ret; - } - - return 0; -} - -/** - * gm20x_secboot_prepare_blobs - load blobs common to all GM20X GPUs. - * - * This includes the LS blob, HS ucode loading blob, and HS bootloader. - * - * The HS ucode unload blob is only used on dGPU. */ int -gm20x_secboot_prepare_blobs(struct gm200_secboot *gsb) -{ - int ret; - - /* Load and prepare the managed falcon's firmwares */ - if (!gsb->ls_blob) { - ret = gm200_secboot_prepare_ls_blob(gsb); - if (ret) - return ret; - } - - /* Load the HS firmware that will load the LS firmwares */ - if (!gsb->acr_load_blob) { - ret = gm200_secboot_prepare_hs_blob(gsb, "acr/ucode_load", - &gsb->acr_load_blob, - &gsb->acr_load_bl_desc, true); - if (ret) - return ret; - } - - /* Load the HS firmware bootloader */ - if (!gsb->hsbl_blob) { - ret = gm200_secboot_prepare_hsbl_blob(gsb); - if (ret) - return ret; - } - - return 0; -} - -static int -gm200_secboot_prepare_blobs(struct gm200_secboot *gsb) -{ - int ret; - - ret = gm20x_secboot_prepare_blobs(gsb); - if (ret) - return ret; - - /* dGPU only: load the HS firmware that unprotects the WPR region */ - if (!gsb->acr_unload_blob) { - ret = gm200_secboot_prepare_hs_blob(gsb, "acr/ucode_unload", - &gsb->acr_unload_blob, - &gsb->acr_unload_bl_desc, false); - if (ret) - return ret; - } - - return 0; -} - -static int -gm200_secboot_blobs_ready(struct gm200_secboot *gsb) +gm200_secboot_run_blob(struct nvkm_secboot *sb, struct nvkm_gpuobj *blob) { + struct gm200_secboot *gsb = gm200_secboot(sb); struct nvkm_subdev *subdev = &gsb->base.subdev; + struct nvkm_falcon *falcon = gsb->base.boot_falcon; + struct nvkm_vma vma; int ret; - /* firmware already loaded, nothing to do... */ - if (gsb->firmware_ok) - return 0; - - ret = gsb->func->prepare_blobs(gsb); - if (ret) { - nvkm_error(subdev, "failed to load secure firmware\n"); - return ret; - } - - gsb->firmware_ok = true; - - return 0; -} - - -/* - * Secure Boot Execution - */ - -/** - * gm200_secboot_load_hs_bl() - load HS bootloader into DMEM and IMEM - */ -static void -gm200_secboot_load_hs_bl(struct gm200_secboot *gsb, void *data, u32 data_size) -{ - struct nvkm_device *device = gsb->base.subdev.device; - struct fw_bin_header *hdr = gsb->hsbl_blob; - struct fw_bl_desc *hsbl_desc = gsb->hsbl_blob + hdr->header_offset; - void *blob_data = gsb->hsbl_blob + hdr->data_offset; - void *hsbl_code = blob_data + hsbl_desc->code_off; - void *hsbl_data = blob_data + hsbl_desc->data_off; - u32 code_size = ALIGN(hsbl_desc->code_size, 256); - const u32 base = gsb->base.base; - u32 blk; - u32 tag; - int i; - - /* - * Copy HS bootloader data - */ - nvkm_wr32(device, base + 0x1c0, (0x00000000 | (0x1 << 24))); - for (i = 0; i < hsbl_desc->data_size / 4; i++) - nvkm_wr32(device, base + 0x1c4, ((u32 *)hsbl_data)[i]); - - /* - * Copy HS bootloader interface structure where the HS descriptor - * expects it to be - */ - nvkm_wr32(device, base + 0x1c0, - (hsbl_desc->dmem_load_off | (0x1 << 24))); - for (i = 0; i < data_size / 4; i++) - nvkm_wr32(device, base + 0x1c4, ((u32 *)data)[i]); - - /* Copy HS bootloader code to end of IMEM */ - blk = (nvkm_rd32(device, base + 0x108) & 0x1ff) - (code_size >> 8); - tag = hsbl_desc->start_tag; - nvkm_wr32(device, base + 0x180, ((blk & 0xff) << 8) | (0x1 << 24)); - for (i = 0; i < code_size / 4; i++) { - /* write new tag every 256B */ - if ((i & 0x3f) == 0) { - nvkm_wr32(device, base + 0x188, tag & 0xffff); - tag++; - } - nvkm_wr32(device, base + 0x184, ((u32 *)hsbl_code)[i]); - } - nvkm_wr32(device, base + 0x188, 0); -} - -/** - * gm200_secboot_setup_falcon() - set up the secure falcon for secure boot - */ -static int -gm200_secboot_setup_falcon(struct gm200_secboot *gsb) -{ - struct nvkm_device *device = gsb->base.subdev.device; - struct fw_bin_header *hdr = gsb->hsbl_blob; - struct fw_bl_desc *hsbl_desc = gsb->hsbl_blob + hdr->header_offset; - /* virtual start address for boot vector */ - u32 virt_addr = hsbl_desc->start_tag << 8; - const u32 base = gsb->base.base; - const u32 reg_base = base + 0xe00; - u32 inst_loc; - int ret; - - ret = nvkm_secboot_falcon_reset(&gsb->base); + ret = nvkm_falcon_get(falcon, subdev); if (ret) return ret; - /* setup apertures - virtual */ - nvkm_wr32(device, reg_base + 4 * (FALCON_DMAIDX_UCODE), 0x4); - nvkm_wr32(device, reg_base + 4 * (FALCON_DMAIDX_VIRT), 0x0); - /* setup apertures - physical */ - nvkm_wr32(device, reg_base + 4 * (FALCON_DMAIDX_PHYS_VID), 0x4); - nvkm_wr32(device, reg_base + 4 * (FALCON_DMAIDX_PHYS_SYS_COH), - 0x4 | 0x1); - nvkm_wr32(device, reg_base + 4 * (FALCON_DMAIDX_PHYS_SYS_NCOH), - 0x4 | 0x2); - - /* Set context */ - if (nvkm_memory_target(gsb->inst->memory) == NVKM_MEM_TARGET_VRAM) - inst_loc = 0x0; /* FB */ - else - inst_loc = 0x3; /* Non-coherent sysmem */ - - nvkm_mask(device, base + 0x048, 0x1, 0x1); - nvkm_wr32(device, base + 0x480, - ((gsb->inst->addr >> 12) & 0xfffffff) | - (inst_loc << 28) | (1 << 30)); - - /* Set boot vector to code's starting virtual address */ - nvkm_wr32(device, base + 0x104, virt_addr); - - return 0; -} - -/** - * gm200_secboot_run_hs_blob() - run the given high-secure blob - */ -static int -gm200_secboot_run_hs_blob(struct gm200_secboot *gsb, struct nvkm_gpuobj *blob, - struct gm200_flcn_bl_desc *desc) -{ - struct nvkm_vma vma; - u64 vma_addr; - const u32 bl_desc_size = gsb->func->bl_desc_size; - u8 bl_desc[bl_desc_size]; - int ret; - /* Map the HS firmware so the HS bootloader can see it */ ret = nvkm_gpuobj_map(blob, gsb->vm, NV_MEM_ACCESS_RW, &vma); - if (ret) + if (ret) { + nvkm_falcon_put(falcon, subdev); return ret; + } - /* Add the mapping address to the DMA bases */ - vma_addr = flcn64_to_u64(desc->code_dma_base) + vma.offset; - desc->code_dma_base.lo = lower_32_bits(vma_addr); - desc->code_dma_base.hi = upper_32_bits(vma_addr); - vma_addr = flcn64_to_u64(desc->data_dma_base) + vma.offset; - desc->data_dma_base.lo = lower_32_bits(vma_addr); - desc->data_dma_base.hi = upper_32_bits(vma_addr); - - /* Fixup the BL header */ - gsb->func->fixup_bl_desc(desc, &bl_desc); - - /* Reset the falcon and make it ready to run the HS bootloader */ - ret = gm200_secboot_setup_falcon(gsb); + /* Reset and set the falcon up */ + ret = nvkm_falcon_reset(falcon); if (ret) - goto done; + goto end; + nvkm_falcon_bind_context(falcon, gsb->inst); /* Load the HS bootloader into the falcon's IMEM/DMEM */ - gm200_secboot_load_hs_bl(gsb, &bl_desc, bl_desc_size); - - /* Start the HS bootloader */ - ret = nvkm_secboot_falcon_run(&gsb->base); + ret = sb->acr->func->load(sb->acr, &gsb->base, blob, vma.offset); if (ret) - goto done; + goto end; -done: - /* Restore the original DMA addresses */ - vma_addr = flcn64_to_u64(desc->code_dma_base) - vma.offset; - desc->code_dma_base.lo = lower_32_bits(vma_addr); - desc->code_dma_base.hi = upper_32_bits(vma_addr); - vma_addr = flcn64_to_u64(desc->data_dma_base) - vma.offset; - desc->data_dma_base.lo = lower_32_bits(vma_addr); - desc->data_dma_base.hi = upper_32_bits(vma_addr); + /* Disable interrupts as we will poll for the HALT bit */ + nvkm_mc_intr_mask(sb->subdev.device, falcon->owner->index, false); - /* We don't need the ACR firmware anymore */ - nvkm_gpuobj_unmap(&vma); - - return ret; -} + /* Set default error value in mailbox register */ + nvkm_falcon_wr32(falcon, 0x040, 0xdeada5a5); -/* - * gm200_secboot_reset() - execute secure boot from the prepared state - * - * Load the HS bootloader and ask the falcon to run it. This will in turn - * load the HS firmware and run it, so once the falcon stops all the managed - * falcons should have their LS firmware loaded and be ready to run. - */ -int -gm200_secboot_reset(struct nvkm_secboot *sb, enum nvkm_secboot_falcon falcon) -{ - struct gm200_secboot *gsb = gm200_secboot(sb); - int ret; - - /* Make sure all blobs are ready */ - ret = gm200_secboot_blobs_ready(gsb); + /* Start the HS bootloader */ + nvkm_falcon_set_start_addr(falcon, sb->acr->start_address); + nvkm_falcon_start(falcon); + ret = nvkm_falcon_wait_for_halt(falcon, 100); if (ret) - return ret; - - /* - * Dummy GM200 implementation: perform secure boot each time we are - * called on FECS. Since only FECS and GPCCS are managed and started - * together, this ought to be safe. - * - * Once we have proper PMU firmware and support, this will be changed - * to a proper call to the PMU method. - */ - if (falcon != NVKM_SECBOOT_FALCON_FECS) goto end; - /* If WPR is set and we have an unload blob, run it to unlock WPR */ - if (gsb->acr_unload_blob && - gsb->falcon_state[NVKM_SECBOOT_FALCON_FECS] != NON_SECURE) { - ret = gm200_secboot_run_hs_blob(gsb, gsb->acr_unload_blob, - &gsb->acr_unload_bl_desc); - if (ret) - return ret; + /* If mailbox register contains an error code, then ACR has failed */ + ret = nvkm_falcon_rd32(falcon, 0x040); + if (ret) { + nvkm_error(subdev, "ACR boot failed, ret 0x%08x", ret); + ret = -EINVAL; + goto end; } - /* Reload all managed falcons */ - ret = gm200_secboot_run_hs_blob(gsb, gsb->acr_load_blob, - &gsb->acr_load_bl_desc); - if (ret) - return ret; - end: - gsb->falcon_state[falcon] = RESET; - return 0; -} + /* Reenable interrupts */ + nvkm_mc_intr_mask(sb->subdev.device, falcon->owner->index, true); -int -gm200_secboot_start(struct nvkm_secboot *sb, enum nvkm_secboot_falcon falcon) -{ - struct gm200_secboot *gsb = gm200_secboot(sb); - int base; - - switch (falcon) { - case NVKM_SECBOOT_FALCON_FECS: - base = 0x409000; - break; - case NVKM_SECBOOT_FALCON_GPCCS: - base = 0x41a000; - break; - default: - nvkm_error(&sb->subdev, "cannot start unhandled falcon!\n"); - return -EINVAL; - } - - nvkm_wr32(sb->subdev.device, base + 0x130, 0x00000002); - gsb->falcon_state[falcon] = RUNNING; + /* We don't need the ACR firmware anymore */ + nvkm_gpuobj_unmap(&vma); + nvkm_falcon_put(falcon, subdev); - return 0; + return ret; } - - int -gm200_secboot_init(struct nvkm_secboot *sb) +gm200_secboot_oneinit(struct nvkm_secboot *sb) { struct gm200_secboot *gsb = gm200_secboot(sb); struct nvkm_device *device = sb->subdev.device; @@ -1361,24 +132,22 @@ gm200_secboot_init(struct nvkm_secboot *sb) nvkm_wo32(gsb->inst, 0x20c, upper_32_bits(vm_area_len - 1)); nvkm_done(gsb->inst); + if (sb->acr->func->oneinit) { + ret = sb->acr->func->oneinit(sb->acr, sb); + if (ret) + return ret; + } + return 0; } int gm200_secboot_fini(struct nvkm_secboot *sb, bool suspend) { - struct gm200_secboot *gsb = gm200_secboot(sb); int ret = 0; - int i; - - /* Run the unload blob to unprotect the WPR region */ - if (gsb->acr_unload_blob && - gsb->falcon_state[NVKM_SECBOOT_FALCON_FECS] != NON_SECURE) - ret = gm200_secboot_run_hs_blob(gsb, gsb->acr_unload_blob, - &gsb->acr_unload_bl_desc); - for (i = 0; i < NVKM_SECBOOT_FALCON_END; i++) - gsb->falcon_state[i] = NON_SECURE; + if (sb->acr->func->fini) + ret = sb->acr->func->fini(sb->acr, sb, suspend); return ret; } @@ -1388,11 +157,7 @@ gm200_secboot_dtor(struct nvkm_secboot *sb) { struct gm200_secboot *gsb = gm200_secboot(sb); - nvkm_gpuobj_del(&gsb->acr_unload_blob); - - kfree(gsb->hsbl_blob); - nvkm_gpuobj_del(&gsb->acr_load_blob); - nvkm_gpuobj_del(&gsb->ls_blob); + sb->acr->func->dtor(sb->acr); nvkm_vm_ref(NULL, &gsb->vm, gsb->pgd); nvkm_gpuobj_del(&gsb->pgd); @@ -1405,50 +170,9 @@ gm200_secboot_dtor(struct nvkm_secboot *sb) static const struct nvkm_secboot_func gm200_secboot = { .dtor = gm200_secboot_dtor, - .init = gm200_secboot_init, + .oneinit = gm200_secboot_oneinit, .fini = gm200_secboot_fini, - .reset = gm200_secboot_reset, - .start = gm200_secboot_start, - .managed_falcons = BIT(NVKM_SECBOOT_FALCON_FECS) | - BIT(NVKM_SECBOOT_FALCON_GPCCS), - .boot_falcon = NVKM_SECBOOT_FALCON_PMU, -}; - -/** - * gm200_fixup_bl_desc - just copy the BL descriptor - * - * Use the GM200 descriptor format by default. - */ -static void -gm200_secboot_fixup_bl_desc(const struct gm200_flcn_bl_desc *desc, void *ret) -{ - memcpy(ret, desc, sizeof(*desc)); -} - -static void -gm200_secboot_fixup_hs_desc(struct gm200_secboot *gsb, - struct hsflcn_acr_desc *desc) -{ - desc->ucode_blob_base = gsb->ls_blob->addr; - desc->ucode_blob_size = gsb->ls_blob->size; - - desc->wpr_offset = 0; - - /* WPR region information for the HS binary to set up */ - desc->wpr_region_id = 1; - desc->regions.no_regions = 1; - desc->regions.region_props[0].region_id = 1; - desc->regions.region_props[0].start_addr = gsb->wpr_addr >> 8; - desc->regions.region_props[0].end_addr = - (gsb->wpr_addr + gsb->wpr_size) >> 8; -} - -static const struct gm200_secboot_func -gm200_secboot_func = { - .bl_desc_size = sizeof(struct gm200_flcn_bl_desc), - .fixup_bl_desc = gm200_secboot_fixup_bl_desc, - .fixup_hs_desc = gm200_secboot_fixup_hs_desc, - .prepare_blobs = gm200_secboot_prepare_blobs, + .run_blob = gm200_secboot_run_blob, }; int @@ -1457,6 +181,12 @@ gm200_secboot_new(struct nvkm_device *device, int index, { int ret; struct gm200_secboot *gsb; + struct nvkm_acr *acr; + + acr = acr_r361_new(BIT(NVKM_SECBOOT_FALCON_FECS) | + BIT(NVKM_SECBOOT_FALCON_GPCCS)); + if (IS_ERR(acr)) + return PTR_ERR(acr); gsb = kzalloc(sizeof(*gsb), GFP_KERNEL); if (!gsb) { @@ -1465,15 +195,14 @@ gm200_secboot_new(struct nvkm_device *device, int index, } *psb = &gsb->base; - ret = nvkm_secboot_ctor(&gm200_secboot, device, index, &gsb->base); + ret = nvkm_secboot_ctor(&gm200_secboot, acr, device, index, &gsb->base); if (ret) return ret; - gsb->func = &gm200_secboot_func; - return 0; } + MODULE_FIRMWARE("nvidia/gm200/acr/bl.bin"); MODULE_FIRMWARE("nvidia/gm200/acr/ucode_load.bin"); MODULE_FIRMWARE("nvidia/gm200/acr/ucode_unload.bin"); diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.h b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.h new file mode 100644 index 000000000000..45adf1a3bc20 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm200.h @@ -0,0 +1,43 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#ifndef __NVKM_SECBOOT_GM200_H__ +#define __NVKM_SECBOOT_GM200_H__ + +#include "priv.h" + +struct gm200_secboot { + struct nvkm_secboot base; + + /* Instance block & address space used for HS FW execution */ + struct nvkm_gpuobj *inst; + struct nvkm_gpuobj *pgd; + struct nvkm_vm *vm; +}; +#define gm200_secboot(sb) container_of(sb, struct gm200_secboot, base) + +int gm200_secboot_oneinit(struct nvkm_secboot *); +int gm200_secboot_fini(struct nvkm_secboot *, bool); +void *gm200_secboot_dtor(struct nvkm_secboot *); +int gm200_secboot_run_blob(struct nvkm_secboot *, struct nvkm_gpuobj *); + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c index d5395ebfe8d3..6707b8edc086 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c @@ -20,103 +20,8 @@ * DEALINGS IN THE SOFTWARE. */ -#include "priv.h" - -#include <core/gpuobj.h> - -/* - * The BL header format used by GM20B's firmware is slightly different - * from the one of GM200. Fix the differences here. - */ -struct gm20b_flcn_bl_desc { - u32 reserved[4]; - u32 signature[4]; - u32 ctx_dma; - u32 code_dma_base; - u32 non_sec_code_off; - u32 non_sec_code_size; - u32 sec_code_off; - u32 sec_code_size; - u32 code_entry_point; - u32 data_dma_base; - u32 data_size; -}; - -static int -gm20b_secboot_prepare_blobs(struct gm200_secboot *gsb) -{ - struct nvkm_subdev *subdev = &gsb->base.subdev; - int acr_size; - int ret; - - ret = gm20x_secboot_prepare_blobs(gsb); - if (ret) - return ret; - - acr_size = gsb->acr_load_blob->size; - /* - * On Tegra the WPR region is set by the bootloader. It is illegal for - * the HS blob to be larger than this region. - */ - if (acr_size > gsb->wpr_size) { - nvkm_error(subdev, "WPR region too small for FW blob!\n"); - nvkm_error(subdev, "required: %dB\n", acr_size); - nvkm_error(subdev, "WPR size: %dB\n", gsb->wpr_size); - return -ENOSPC; - } - - return 0; -} - -/** - * gm20b_secboot_fixup_bl_desc - adapt BL descriptor to format used by GM20B FW - * - * There is only a slight format difference (DMA addresses being 32-bits and - * 256B-aligned) to address. - */ -static void -gm20b_secboot_fixup_bl_desc(const struct gm200_flcn_bl_desc *desc, void *ret) -{ - struct gm20b_flcn_bl_desc *gdesc = ret; - u64 addr; - - memcpy(gdesc->reserved, desc->reserved, sizeof(gdesc->reserved)); - memcpy(gdesc->signature, desc->signature, sizeof(gdesc->signature)); - gdesc->ctx_dma = desc->ctx_dma; - addr = desc->code_dma_base.hi; - addr <<= 32; - addr |= desc->code_dma_base.lo; - gdesc->code_dma_base = lower_32_bits(addr >> 8); - gdesc->non_sec_code_off = desc->non_sec_code_off; - gdesc->non_sec_code_size = desc->non_sec_code_size; - gdesc->sec_code_off = desc->sec_code_off; - gdesc->sec_code_size = desc->sec_code_size; - gdesc->code_entry_point = desc->code_entry_point; - addr = desc->data_dma_base.hi; - addr <<= 32; - addr |= desc->data_dma_base.lo; - gdesc->data_dma_base = lower_32_bits(addr >> 8); - gdesc->data_size = desc->data_size; -} - -static void -gm20b_secboot_fixup_hs_desc(struct gm200_secboot *gsb, - struct hsflcn_acr_desc *desc) -{ - desc->ucode_blob_base = gsb->ls_blob->addr; - desc->ucode_blob_size = gsb->ls_blob->size; - - desc->wpr_offset = 0; -} - -static const struct gm200_secboot_func -gm20b_secboot_func = { - .bl_desc_size = sizeof(struct gm20b_flcn_bl_desc), - .fixup_bl_desc = gm20b_secboot_fixup_bl_desc, - .fixup_hs_desc = gm20b_secboot_fixup_hs_desc, - .prepare_blobs = gm20b_secboot_prepare_blobs, -}; - +#include "acr.h" +#include "gm200.h" #ifdef CONFIG_ARCH_TEGRA #define TEGRA_MC_BASE 0x70019000 @@ -144,15 +49,15 @@ gm20b_tegra_read_wpr(struct gm200_secboot *gsb) nvkm_error(&sb->subdev, "Cannot map Tegra MC registers\n"); return PTR_ERR(mc); } - gsb->wpr_addr = ioread32_native(mc + MC_SECURITY_CARVEOUT2_BOM_0) | + sb->wpr_addr = ioread32_native(mc + MC_SECURITY_CARVEOUT2_BOM_0) | ((u64)ioread32_native(mc + MC_SECURITY_CARVEOUT2_BOM_HI_0) << 32); - gsb->wpr_size = ioread32_native(mc + MC_SECURITY_CARVEOUT2_SIZE_128K) + sb->wpr_size = ioread32_native(mc + MC_SECURITY_CARVEOUT2_SIZE_128K) << 17; cfg = ioread32_native(mc + MC_SECURITY_CARVEOUT2_CFG0); iounmap(mc); /* Check that WPR settings are valid */ - if (gsb->wpr_size == 0) { + if (sb->wpr_size == 0) { nvkm_error(&sb->subdev, "WPR region is empty\n"); return -EINVAL; } @@ -174,7 +79,7 @@ gm20b_tegra_read_wpr(struct gm200_secboot *gsb) #endif static int -gm20b_secboot_init(struct nvkm_secboot *sb) +gm20b_secboot_oneinit(struct nvkm_secboot *sb) { struct gm200_secboot *gsb = gm200_secboot(sb); int ret; @@ -183,17 +88,15 @@ gm20b_secboot_init(struct nvkm_secboot *sb) if (ret) return ret; - return gm200_secboot_init(sb); + return gm200_secboot_oneinit(sb); } static const struct nvkm_secboot_func gm20b_secboot = { .dtor = gm200_secboot_dtor, - .init = gm20b_secboot_init, - .reset = gm200_secboot_reset, - .start = gm200_secboot_start, - .managed_falcons = BIT(NVKM_SECBOOT_FALCON_FECS), - .boot_falcon = NVKM_SECBOOT_FALCON_PMU, + .oneinit = gm20b_secboot_oneinit, + .fini = gm200_secboot_fini, + .run_blob = gm200_secboot_run_blob, }; int @@ -202,6 +105,11 @@ gm20b_secboot_new(struct nvkm_device *device, int index, { int ret; struct gm200_secboot *gsb; + struct nvkm_acr *acr; + + acr = acr_r352_new(BIT(NVKM_SECBOOT_FALCON_FECS)); + if (IS_ERR(acr)) + return PTR_ERR(acr); gsb = kzalloc(sizeof(*gsb), GFP_KERNEL); if (!gsb) { @@ -210,12 +118,10 @@ gm20b_secboot_new(struct nvkm_device *device, int index, } *psb = &gsb->base; - ret = nvkm_secboot_ctor(&gm20b_secboot, device, index, &gsb->base); + ret = nvkm_secboot_ctor(&gm20b_secboot, acr, device, index, &gsb->base); if (ret) return ret; - gsb->func = &gm20b_secboot_func; - return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/ls_ucode.h b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/ls_ucode.h new file mode 100644 index 000000000000..00886cee57eb --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/ls_ucode.h @@ -0,0 +1,151 @@ +/* + * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#ifndef __NVKM_SECBOOT_LS_UCODE_H__ +#define __NVKM_SECBOOT_LS_UCODE_H__ + +#include <core/os.h> +#include <core/subdev.h> +#include <subdev/secboot.h> + + +/** + * struct ls_ucode_img_desc - descriptor of firmware image + * @descriptor_size: size of this descriptor + * @image_size: size of the whole image + * @bootloader_start_offset: start offset of the bootloader in ucode image + * @bootloader_size: size of the bootloader + * @bootloader_imem_offset: start off set of the bootloader in IMEM + * @bootloader_entry_point: entry point of the bootloader in IMEM + * @app_start_offset: start offset of the LS firmware + * @app_size: size of the LS firmware's code and data + * @app_imem_offset: offset of the app in IMEM + * @app_imem_entry: entry point of the app in IMEM + * @app_dmem_offset: offset of the data in DMEM + * @app_resident_code_offset: offset of app code from app_start_offset + * @app_resident_code_size: size of the code + * @app_resident_data_offset: offset of data from app_start_offset + * @app_resident_data_size: size of data + * + * A firmware image contains the code, data, and bootloader of a given LS + * falcon in a single blob. This structure describes where everything is. + * + * This can be generated from a (bootloader, code, data) set if they have + * been loaded separately, or come directly from a file. + */ +struct ls_ucode_img_desc { + u32 descriptor_size; + u32 image_size; + u32 tools_version; + u32 app_version; + char date[64]; + u32 bootloader_start_offset; + u32 bootloader_size; + u32 bootloader_imem_offset; + u32 bootloader_entry_point; + u32 app_start_offset; + u32 app_size; + u32 app_imem_offset; + u32 app_imem_entry; + u32 app_dmem_offset; + u32 app_resident_code_offset; + u32 app_resident_code_size; + u32 app_resident_data_offset; + u32 app_resident_data_size; + u32 nb_overlays; + struct {u32 start; u32 size; } load_ovl[64]; + u32 compressed; +}; + +/** + * struct ls_ucode_img - temporary storage for loaded LS firmwares + * @node: to link within lsf_ucode_mgr + * @falcon_id: ID of the falcon this LS firmware is for + * @ucode_desc: loaded or generated map of ucode_data + * @ucode_data: firmware payload (code and data) + * @ucode_size: size in bytes of data in ucode_data + * @sig: signature for this firmware + * @sig:size: size of the signature in bytes + * + * Preparing the WPR LS blob requires information about all the LS firmwares + * (size, etc) to be known. This structure contains all the data of one LS + * firmware. + */ +struct ls_ucode_img { + struct list_head node; + enum nvkm_secboot_falcon falcon_id; + + struct ls_ucode_img_desc ucode_desc; + u8 *ucode_data; + u32 ucode_size; + + u8 *sig; + u32 sig_size; +}; + +/** + * struct fw_bin_header - header of firmware files + * @bin_magic: always 0x3b1d14f0 + * @bin_ver: version of the bin format + * @bin_size: entire image size including this header + * @header_offset: offset of the firmware/bootloader header in the file + * @data_offset: offset of the firmware/bootloader payload in the file + * @data_size: size of the payload + * + * This header is located at the beginning of the HS firmware and HS bootloader + * files, to describe where the headers and data can be found. + */ +struct fw_bin_header { + u32 bin_magic; + u32 bin_ver; + u32 bin_size; + u32 header_offset; + u32 data_offset; + u32 data_size; +}; + +/** + * struct fw_bl_desc - firmware bootloader descriptor + * @start_tag: starting tag of bootloader + * @desc_dmem_load_off: DMEM offset of flcn_bl_dmem_desc + * @code_off: offset of code section + * @code_size: size of code section + * @data_off: offset of data section + * @data_size: size of data section + * + * This structure is embedded in bootloader firmware files at to describe the + * IMEM and DMEM layout expected by the bootloader. + */ +struct fw_bl_desc { + u32 start_tag; + u32 dmem_load_off; + u32 code_off; + u32 code_size; + u32 data_off; + u32 data_size; +}; + +int acr_ls_ucode_load_fecs(const struct nvkm_subdev *, struct ls_ucode_img *); +int acr_ls_ucode_load_gpccs(const struct nvkm_subdev *, struct ls_ucode_img *); + + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/ls_ucode_gr.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/ls_ucode_gr.c new file mode 100644 index 000000000000..40a6df77bb8a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/ls_ucode_gr.c @@ -0,0 +1,158 @@ +/* + * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + + +#include "ls_ucode.h" +#include "acr.h" + +#include <core/firmware.h> + +#define BL_DESC_BLK_SIZE 256 +/** + * Build a ucode image and descriptor from provided bootloader, code and data. + * + * @bl: bootloader image, including 16-bytes descriptor + * @code: LS firmware code segment + * @data: LS firmware data segment + * @desc: ucode descriptor to be written + * + * Return: allocated ucode image with corresponding descriptor information. desc + * is also updated to contain the right offsets within returned image. + */ +static void * +ls_ucode_img_build(const struct firmware *bl, const struct firmware *code, + const struct firmware *data, struct ls_ucode_img_desc *desc) +{ + struct fw_bin_header *bin_hdr = (void *)bl->data; + struct fw_bl_desc *bl_desc = (void *)bl->data + bin_hdr->header_offset; + void *bl_data = (void *)bl->data + bin_hdr->data_offset; + u32 pos = 0; + void *image; + + desc->bootloader_start_offset = pos; + desc->bootloader_size = ALIGN(bl_desc->code_size, sizeof(u32)); + desc->bootloader_imem_offset = bl_desc->start_tag * 256; + desc->bootloader_entry_point = bl_desc->start_tag * 256; + + pos = ALIGN(pos + desc->bootloader_size, BL_DESC_BLK_SIZE); + desc->app_start_offset = pos; + desc->app_size = ALIGN(code->size, BL_DESC_BLK_SIZE) + + ALIGN(data->size, BL_DESC_BLK_SIZE); + desc->app_imem_offset = 0; + desc->app_imem_entry = 0; + desc->app_dmem_offset = 0; + desc->app_resident_code_offset = 0; + desc->app_resident_code_size = ALIGN(code->size, BL_DESC_BLK_SIZE); + + pos = ALIGN(pos + desc->app_resident_code_size, BL_DESC_BLK_SIZE); + desc->app_resident_data_offset = pos - desc->app_start_offset; + desc->app_resident_data_size = ALIGN(data->size, BL_DESC_BLK_SIZE); + + desc->image_size = ALIGN(bl_desc->code_size, BL_DESC_BLK_SIZE) + + desc->app_size; + + image = kzalloc(desc->image_size, GFP_KERNEL); + if (!image) + return ERR_PTR(-ENOMEM); + + memcpy(image + desc->bootloader_start_offset, bl_data, + bl_desc->code_size); + memcpy(image + desc->app_start_offset, code->data, code->size); + memcpy(image + desc->app_start_offset + desc->app_resident_data_offset, + data->data, data->size); + + return image; +} + +/** + * ls_ucode_img_load_gr() - load and prepare a LS GR ucode image + * + * Load the LS microcode, bootloader and signature and pack them into a single + * blob. Also generate the corresponding ucode descriptor. + */ +static int +ls_ucode_img_load_gr(const struct nvkm_subdev *subdev, struct ls_ucode_img *img, + const char *falcon_name) +{ + const struct firmware *bl, *code, *data, *sig; + char f[64]; + int ret; + + snprintf(f, sizeof(f), "gr/%s_bl", falcon_name); + ret = nvkm_firmware_get(subdev->device, f, &bl); + if (ret) + goto error; + + snprintf(f, sizeof(f), "gr/%s_inst", falcon_name); + ret = nvkm_firmware_get(subdev->device, f, &code); + if (ret) + goto free_bl; + + snprintf(f, sizeof(f), "gr/%s_data", falcon_name); + ret = nvkm_firmware_get(subdev->device, f, &data); + if (ret) + goto free_inst; + + snprintf(f, sizeof(f), "gr/%s_sig", falcon_name); + ret = nvkm_firmware_get(subdev->device, f, &sig); + if (ret) + goto free_data; + img->sig = kmemdup(sig->data, sig->size, GFP_KERNEL); + if (!img->sig) { + ret = -ENOMEM; + goto free_sig; + } + img->sig_size = sig->size; + + img->ucode_data = ls_ucode_img_build(bl, code, data, + &img->ucode_desc); + if (IS_ERR(img->ucode_data)) { + ret = PTR_ERR(img->ucode_data); + goto free_data; + } + img->ucode_size = img->ucode_desc.image_size; + +free_sig: + nvkm_firmware_put(sig); +free_data: + nvkm_firmware_put(data); +free_inst: + nvkm_firmware_put(code); +free_bl: + nvkm_firmware_put(bl); +error: + return ret; +} + +int +acr_ls_ucode_load_fecs(const struct nvkm_subdev *subdev, + struct ls_ucode_img *img) +{ + return ls_ucode_img_load_gr(subdev, img, "fecs"); +} + +int +acr_ls_ucode_load_gpccs(const struct nvkm_subdev *subdev, + struct ls_ucode_img *img) +{ + return ls_ucode_img_load_gr(subdev, img, "gpccs"); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/priv.h index a9a8a0e1017e..936a65f5658c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/priv.h @@ -27,20 +27,16 @@ #include <subdev/mmu.h> struct nvkm_secboot_func { - int (*init)(struct nvkm_secboot *); + int (*oneinit)(struct nvkm_secboot *); int (*fini)(struct nvkm_secboot *, bool suspend); void *(*dtor)(struct nvkm_secboot *); - int (*reset)(struct nvkm_secboot *, enum nvkm_secboot_falcon); - int (*start)(struct nvkm_secboot *, enum nvkm_secboot_falcon); - - /* ID of the falcon that will perform secure boot */ - enum nvkm_secboot_falcon boot_falcon; - /* Bit-mask of IDs of managed falcons */ - unsigned long managed_falcons; + int (*run_blob)(struct nvkm_secboot *, struct nvkm_gpuobj *); }; -int nvkm_secboot_ctor(const struct nvkm_secboot_func *, struct nvkm_device *, - int index, struct nvkm_secboot *); +extern const char *nvkm_secboot_falcon_name[]; + +int nvkm_secboot_ctor(const struct nvkm_secboot_func *, struct nvkm_acr *, + struct nvkm_device *, int, struct nvkm_secboot *); int nvkm_secboot_falcon_reset(struct nvkm_secboot *); int nvkm_secboot_falcon_run(struct nvkm_secboot *); @@ -48,187 +44,20 @@ struct flcn_u64 { u32 lo; u32 hi; }; + static inline u64 flcn64_to_u64(const struct flcn_u64 f) { return ((u64)f.hi) << 32 | f.lo; } -/** - * struct gm200_flcn_bl_desc - DMEM bootloader descriptor - * @signature: 16B signature for secure code. 0s if no secure code - * @ctx_dma: DMA context to be used by BL while loading code/data - * @code_dma_base: 256B-aligned Physical FB Address where code is located - * (falcon's $xcbase register) - * @non_sec_code_off: offset from code_dma_base where the non-secure code is - * located. The offset must be multiple of 256 to help perf - * @non_sec_code_size: the size of the nonSecure code part. - * @sec_code_off: offset from code_dma_base where the secure code is - * located. The offset must be multiple of 256 to help perf - * @sec_code_size: offset from code_dma_base where the secure code is - * located. The offset must be multiple of 256 to help perf - * @code_entry_point: code entry point which will be invoked by BL after - * code is loaded. - * @data_dma_base: 256B aligned Physical FB Address where data is located. - * (falcon's $xdbase register) - * @data_size: size of data block. Should be multiple of 256B - * - * Structure used by the bootloader to load the rest of the code. This has - * to be filled by host and copied into DMEM at offset provided in the - * hsflcn_bl_desc.bl_desc_dmem_load_off. - */ -struct gm200_flcn_bl_desc { - u32 reserved[4]; - u32 signature[4]; - u32 ctx_dma; - struct flcn_u64 code_dma_base; - u32 non_sec_code_off; - u32 non_sec_code_size; - u32 sec_code_off; - u32 sec_code_size; - u32 code_entry_point; - struct flcn_u64 data_dma_base; - u32 data_size; -}; - -/** - * struct hsflcn_acr_desc - data section of the HS firmware - * - * This header is to be copied at the beginning of DMEM by the HS bootloader. - * - * @signature: signature of ACR ucode - * @wpr_region_id: region ID holding the WPR header and its details - * @wpr_offset: offset from the WPR region holding the wpr header - * @regions: region descriptors - * @nonwpr_ucode_blob_size: size of LS blob - * @nonwpr_ucode_blob_start: FB location of LS blob is - */ -struct hsflcn_acr_desc { - union { - u8 reserved_dmem[0x200]; - u32 signatures[4]; - } ucode_reserved_space; - u32 wpr_region_id; - u32 wpr_offset; - u32 mmu_mem_range; -#define FLCN_ACR_MAX_REGIONS 2 - struct { - u32 no_regions; - struct { - u32 start_addr; - u32 end_addr; - u32 region_id; - u32 read_mask; - u32 write_mask; - u32 client_mask; - } region_props[FLCN_ACR_MAX_REGIONS]; - } regions; - u32 ucode_blob_size; - u64 ucode_blob_base __aligned(8); - struct { - u32 vpr_enabled; - u32 vpr_start; - u32 vpr_end; - u32 hdcp_policies; - } vpr_desc; -}; - -/** - * Contains the whole secure boot state, allowing it to be performed as needed - * @wpr_addr: physical address of the WPR region - * @wpr_size: size in bytes of the WPR region - * @ls_blob: LS blob of all the LS firmwares, signatures, bootloaders - * @ls_blob_size: size of the LS blob - * @ls_blob_nb_regions: number of LS firmwares that will be loaded - * @acr_blob: HS blob - * @acr_blob_vma: mapping of the HS blob into the secure falcon's VM - * @acr_bl_desc: bootloader descriptor of the HS blob - * @hsbl_blob: HS blob bootloader - * @inst: instance block for HS falcon - * @pgd: page directory for the HS falcon - * @vm: address space used by the HS falcon - * @falcon_state: current state of the managed falcons - * @firmware_ok: whether the firmware blobs have been created - */ -struct gm200_secboot { - struct nvkm_secboot base; - const struct gm200_secboot_func *func; - - /* - * Address and size of the WPR region. On dGPU this will be the - * address of the LS blob. On Tegra this is a fixed region set by the - * bootloader - */ - u64 wpr_addr; - u32 wpr_size; - - /* - * HS FW - lock WPR region (dGPU only) and load LS FWs - * on Tegra the HS FW copies the LS blob into the fixed WPR instead - */ - struct nvkm_gpuobj *acr_load_blob; - struct gm200_flcn_bl_desc acr_load_bl_desc; - - /* HS FW - unlock WPR region (dGPU only) */ - struct nvkm_gpuobj *acr_unload_blob; - struct gm200_flcn_bl_desc acr_unload_bl_desc; - - /* HS bootloader */ - void *hsbl_blob; - - /* LS FWs, to be loaded by the HS ACR */ - struct nvkm_gpuobj *ls_blob; - - /* Instance block & address space used for HS FW execution */ - struct nvkm_gpuobj *inst; - struct nvkm_gpuobj *pgd; - struct nvkm_vm *vm; - - /* To keep track of the state of all managed falcons */ - enum { - /* In non-secure state, no firmware loaded, no privileges*/ - NON_SECURE = 0, - /* In low-secure mode and ready to be started */ - RESET, - /* In low-secure mode and running */ - RUNNING, - } falcon_state[NVKM_SECBOOT_FALCON_END]; - - bool firmware_ok; -}; -#define gm200_secboot(sb) container_of(sb, struct gm200_secboot, base) - -/** - * Contains functions we wish to abstract between GM200-like implementations - * @bl_desc_size: size of the BL descriptor used by this chip. - * @fixup_bl_desc: hook that generates the proper BL descriptor format from - * the generic GM200 format into a data array of size - * bl_desc_size - * @fixup_hs_desc: hook that twiddles the HS descriptor before it is used - * @prepare_blobs: prepares the various blobs needed for secure booting - */ -struct gm200_secboot_func { - /* - * Size of the bootloader descriptor for this chip. A block of this - * size is allocated before booting a falcon and the fixup_bl_desc - * callback is called on it - */ - u32 bl_desc_size; - void (*fixup_bl_desc)(const struct gm200_flcn_bl_desc *, void *); - - /* - * Chip-specific modifications of the HS descriptor can be done here. - * On dGPU this is used to fill the information about the WPR region - * we want the HS FW to set up. - */ - void (*fixup_hs_desc)(struct gm200_secboot *, struct hsflcn_acr_desc *); - int (*prepare_blobs)(struct gm200_secboot *); -}; +static inline struct flcn_u64 u64_to_flcn64(u64 u) +{ + struct flcn_u64 ret; -int gm200_secboot_init(struct nvkm_secboot *); -void *gm200_secboot_dtor(struct nvkm_secboot *); -int gm200_secboot_reset(struct nvkm_secboot *, u32); -int gm200_secboot_start(struct nvkm_secboot *, u32); + ret.hi = upper_32_bits(u); + ret.lo = lower_32_bits(u); -int gm20x_secboot_prepare_blobs(struct gm200_secboot *); + return ret; +} #endif diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/therm/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/therm/base.c index 8894fee30cbc..df949fa7d05d 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/therm/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/therm/base.c @@ -64,10 +64,9 @@ nvkm_therm_update_trip(struct nvkm_therm *therm) } static int -nvkm_therm_update_linear(struct nvkm_therm *therm) +nvkm_therm_compute_linear_duty(struct nvkm_therm *therm, u8 linear_min_temp, + u8 linear_max_temp) { - u8 linear_min_temp = therm->fan->bios.linear_min_temp; - u8 linear_max_temp = therm->fan->bios.linear_max_temp; u8 temp = therm->func->temp_get(therm); u16 duty; @@ -85,6 +84,21 @@ nvkm_therm_update_linear(struct nvkm_therm *therm) return duty; } +static int +nvkm_therm_update_linear(struct nvkm_therm *therm) +{ + u8 min = therm->fan->bios.linear_min_temp; + u8 max = therm->fan->bios.linear_max_temp; + return nvkm_therm_compute_linear_duty(therm, min, max); +} + +static int +nvkm_therm_update_linear_fallback(struct nvkm_therm *therm) +{ + u8 max = therm->bios_sensor.thrs_fan_boost.temp; + return nvkm_therm_compute_linear_duty(therm, 30, max); +} + static void nvkm_therm_update(struct nvkm_therm *therm, int mode) { @@ -119,6 +133,8 @@ nvkm_therm_update(struct nvkm_therm *therm, int mode) case NVBIOS_THERM_FAN_OTHER: if (therm->cstate) duty = therm->cstate; + else + duty = nvkm_therm_update_linear_fallback(therm); poll = false; break; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/top/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/top/base.c index fe063d5728e2..67ada1d9a28c 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/top/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/top/base.c @@ -95,6 +95,20 @@ nvkm_top_intr(struct nvkm_device *device, u32 intr, u64 *psubdevs) return intr & ~handled; } +int +nvkm_top_fault_id(struct nvkm_device *device, enum nvkm_devidx devidx) +{ + struct nvkm_top *top = device->top; + struct nvkm_top_device *info; + + list_for_each_entry(info, &top->device, head) { + if (info->index == devidx && info->fault >= 0) + return info->fault; + } + + return -ENOENT; +} + enum nvkm_devidx nvkm_top_fault(struct nvkm_device *device, int fault) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/Kbuild index c34076223b7b..bcd179ba11d0 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/Kbuild +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/Kbuild @@ -1,6 +1,7 @@ nvkm-y += nvkm/subdev/volt/base.o nvkm-y += nvkm/subdev/volt/gpio.o nvkm-y += nvkm/subdev/volt/nv40.o +nvkm-y += nvkm/subdev/volt/gf100.o nvkm-y += nvkm/subdev/volt/gk104.o nvkm-y += nvkm/subdev/volt/gk20a.o nvkm-y += nvkm/subdev/volt/gm20b.o diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/base.c index 1c3d23b0e84a..e344901cfdc7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/base.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/base.c @@ -26,6 +26,7 @@ #include <subdev/bios.h> #include <subdev/bios/vmap.h> #include <subdev/bios/volt.h> +#include <subdev/therm.h> int nvkm_volt_get(struct nvkm_volt *volt) @@ -50,33 +51,45 @@ static int nvkm_volt_set(struct nvkm_volt *volt, u32 uv) { struct nvkm_subdev *subdev = &volt->subdev; - int i, ret = -EINVAL; + int i, ret = -EINVAL, best_err = volt->max_uv, best = -1; if (volt->func->volt_set) return volt->func->volt_set(volt, uv); for (i = 0; i < volt->vid_nr; i++) { - if (volt->vid[i].uv == uv) { - ret = volt->func->vid_set(volt, volt->vid[i].vid); - nvkm_debug(subdev, "set %duv: %d\n", uv, ret); + int err = volt->vid[i].uv - uv; + if (err < 0 || err > best_err) + continue; + + best_err = err; + best = i; + if (best_err == 0) break; - } } + + if (best == -1) { + nvkm_error(subdev, "couldn't set %iuv\n", uv); + return ret; + } + + ret = volt->func->vid_set(volt, volt->vid[best].vid); + nvkm_debug(subdev, "set req %duv to %duv: %d\n", uv, + volt->vid[best].uv, ret); return ret; } -static int -nvkm_volt_map(struct nvkm_volt *volt, u8 id) +int +nvkm_volt_map_min(struct nvkm_volt *volt, u8 id) { struct nvkm_bios *bios = volt->subdev.device->bios; struct nvbios_vmap_entry info; u8 ver, len; - u16 vmap; + u32 vmap; vmap = nvbios_vmap_entry_parse(bios, id, &ver, &len, &info); if (vmap) { if (info.link != 0xff) { - int ret = nvkm_volt_map(volt, info.link); + int ret = nvkm_volt_map_min(volt, info.link); if (ret < 0) return ret; info.min += ret; @@ -88,19 +101,79 @@ nvkm_volt_map(struct nvkm_volt *volt, u8 id) } int -nvkm_volt_set_id(struct nvkm_volt *volt, u8 id, int condition) +nvkm_volt_map(struct nvkm_volt *volt, u8 id, u8 temp) +{ + struct nvkm_bios *bios = volt->subdev.device->bios; + struct nvbios_vmap_entry info; + u8 ver, len; + u32 vmap; + + vmap = nvbios_vmap_entry_parse(bios, id, &ver, &len, &info); + if (vmap) { + s64 result; + + if (volt->speedo < 0) + return volt->speedo; + + if (ver == 0x10 || (ver == 0x20 && info.mode == 0)) { + result = div64_s64((s64)info.arg[0], 10); + result += div64_s64((s64)info.arg[1] * volt->speedo, 10); + result += div64_s64((s64)info.arg[2] * volt->speedo * volt->speedo, 100000); + } else if (ver == 0x20) { + switch (info.mode) { + /* 0x0 handled above! */ + case 0x1: + result = ((s64)info.arg[0] * 15625) >> 18; + result += ((s64)info.arg[1] * volt->speedo * 15625) >> 18; + result += ((s64)info.arg[2] * temp * 15625) >> 10; + result += ((s64)info.arg[3] * volt->speedo * temp * 15625) >> 18; + result += ((s64)info.arg[4] * volt->speedo * volt->speedo * 15625) >> 30; + result += ((s64)info.arg[5] * temp * temp * 15625) >> 18; + break; + case 0x3: + result = (info.min + info.max) / 2; + break; + case 0x2: + default: + result = info.min; + break; + } + } else { + return -ENODEV; + } + + result = min(max(result, (s64)info.min), (s64)info.max); + + if (info.link != 0xff) { + int ret = nvkm_volt_map(volt, info.link, temp); + if (ret < 0) + return ret; + result += ret; + } + return result; + } + + return id ? id * 10000 : -ENODEV; +} + +int +nvkm_volt_set_id(struct nvkm_volt *volt, u8 id, u8 min_id, u8 temp, + int condition) { int ret; if (volt->func->set_id) return volt->func->set_id(volt, id, condition); - ret = nvkm_volt_map(volt, id); + ret = nvkm_volt_map(volt, id, temp); if (ret >= 0) { int prev = nvkm_volt_get(volt); if (!condition || prev < 0 || (condition < 0 && ret < prev) || (condition > 0 && ret > prev)) { + int min = nvkm_volt_map(volt, min_id, temp); + if (min >= 0) + ret = max(min, ret); ret = nvkm_volt_set(volt, ret); } else { ret = 0; @@ -112,14 +185,16 @@ nvkm_volt_set_id(struct nvkm_volt *volt, u8 id, int condition) static void nvkm_volt_parse_bios(struct nvkm_bios *bios, struct nvkm_volt *volt) { + struct nvkm_subdev *subdev = &bios->subdev; struct nvbios_volt_entry ivid; struct nvbios_volt info; u8 ver, hdr, cnt, len; - u16 data; + u32 data; int i; data = nvbios_volt_parse(bios, &ver, &hdr, &cnt, &len, &info); - if (data && info.vidmask && info.base && info.step) { + if (data && info.vidmask && info.base && info.step && info.ranged) { + nvkm_debug(subdev, "found ranged based VIDs\n"); volt->min_uv = info.min; volt->max_uv = info.max; for (i = 0; i < info.vidmask + 1; i++) { @@ -132,7 +207,8 @@ nvkm_volt_parse_bios(struct nvkm_bios *bios, struct nvkm_volt *volt) info.base += info.step; } volt->vid_mask = info.vidmask; - } else if (data && info.vidmask) { + } else if (data && info.vidmask && !info.ranged) { + nvkm_debug(subdev, "found entry based VIDs\n"); volt->min_uv = 0xffffffff; volt->max_uv = 0; for (i = 0; i < cnt; i++) { @@ -154,6 +230,14 @@ nvkm_volt_parse_bios(struct nvkm_bios *bios, struct nvkm_volt *volt) } static int +nvkm_volt_speedo_read(struct nvkm_volt *volt) +{ + if (volt->func->speedo_read) + return volt->func->speedo_read(volt); + return -EINVAL; +} + +static int nvkm_volt_init(struct nvkm_subdev *subdev) { struct nvkm_volt *volt = nvkm_volt(subdev); @@ -167,6 +251,21 @@ nvkm_volt_init(struct nvkm_subdev *subdev) return 0; } +static int +nvkm_volt_oneinit(struct nvkm_subdev *subdev) +{ + struct nvkm_volt *volt = nvkm_volt(subdev); + + volt->speedo = nvkm_volt_speedo_read(volt); + if (volt->speedo > 0) + nvkm_debug(&volt->subdev, "speedo %x\n", volt->speedo); + + if (volt->func->oneinit) + return volt->func->oneinit(volt); + + return 0; +} + static void * nvkm_volt_dtor(struct nvkm_subdev *subdev) { @@ -177,6 +276,7 @@ static const struct nvkm_subdev_func nvkm_volt = { .dtor = nvkm_volt_dtor, .init = nvkm_volt_init, + .oneinit = nvkm_volt_oneinit, }; void @@ -191,9 +291,22 @@ nvkm_volt_ctor(const struct nvkm_volt_func *func, struct nvkm_device *device, /* Assuming the non-bios device should build the voltage table later */ if (bios) { + u8 ver, hdr, cnt, len; + struct nvbios_vmap vmap; + nvkm_volt_parse_bios(bios, volt); nvkm_debug(&volt->subdev, "min: %iuv max: %iuv\n", volt->min_uv, volt->max_uv); + + if (nvbios_vmap_parse(bios, &ver, &hdr, &cnt, &len, &vmap)) { + volt->max0_id = vmap.max0; + volt->max1_id = vmap.max1; + volt->max2_id = vmap.max2; + } else { + volt->max0_id = 0xff; + volt->max1_id = 0xff; + volt->max2_id = 0xff; + } } if (volt->vid_nr) { diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gf100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gf100.c new file mode 100644 index 000000000000..d9ed6925ca64 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gf100.c @@ -0,0 +1,70 @@ +/* + * Copyright 2016 Karol Herbst + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Karol Herbst + */ +#include "priv.h" + +#include <subdev/fuse.h> + +static int +gf100_volt_speedo_read(struct nvkm_volt *volt) +{ + struct nvkm_device *device = volt->subdev.device; + struct nvkm_fuse *fuse = device->fuse; + + if (!fuse) + return -EINVAL; + + return nvkm_fuse_read(fuse, 0x1cc); +} + +int +gf100_volt_oneinit(struct nvkm_volt *volt) +{ + struct nvkm_subdev *subdev = &volt->subdev; + if (volt->speedo <= 0) + nvkm_error(subdev, "couldn't find speedo value, volting not " + "possible\n"); + return 0; +} + +static const struct nvkm_volt_func +gf100_volt = { + .oneinit = gf100_volt_oneinit, + .vid_get = nvkm_voltgpio_get, + .vid_set = nvkm_voltgpio_set, + .speedo_read = gf100_volt_speedo_read, +}; + +int +gf100_volt_new(struct nvkm_device *device, int index, struct nvkm_volt **pvolt) +{ + struct nvkm_volt *volt; + int ret; + + ret = nvkm_volt_new_(&gf100_volt, device, index, &volt); + *pvolt = volt; + if (ret) + return ret; + + return nvkm_voltgpio_init(volt); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gk104.c b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gk104.c index 420bd84d8483..1c744e029454 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gk104.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gk104.c @@ -27,6 +27,7 @@ #include <subdev/gpio.h> #include <subdev/bios.h> #include <subdev/bios/volt.h> +#include <subdev/fuse.h> #define gk104_volt(p) container_of((p), struct gk104_volt, base) struct gk104_volt { @@ -34,7 +35,7 @@ struct gk104_volt { struct nvbios_volt bios; }; -int +static int gk104_volt_get(struct nvkm_volt *base) { struct nvbios_volt *bios = &gk104_volt(base)->bios; @@ -47,7 +48,7 @@ gk104_volt_get(struct nvkm_volt *base) return bios->base + bios->pwm_range * duty / div; } -int +static int gk104_volt_set(struct nvkm_volt *base, u32 uv) { struct nvbios_volt *bios = &gk104_volt(base)->bios; @@ -64,13 +65,33 @@ gk104_volt_set(struct nvkm_volt *base, u32 uv) return 0; } +static int +gk104_volt_speedo_read(struct nvkm_volt *volt) +{ + struct nvkm_device *device = volt->subdev.device; + struct nvkm_fuse *fuse = device->fuse; + int ret; + + if (!fuse) + return -EINVAL; + + nvkm_wr32(device, 0x122634, 0x0); + ret = nvkm_fuse_read(fuse, 0x3a8); + nvkm_wr32(device, 0x122634, 0x41); + return ret; +} + static const struct nvkm_volt_func gk104_volt_pwm = { + .oneinit = gf100_volt_oneinit, .volt_get = gk104_volt_get, .volt_set = gk104_volt_set, + .speedo_read = gk104_volt_speedo_read, }, gk104_volt_gpio = { + .oneinit = gf100_volt_oneinit, .vid_get = nvkm_voltgpio_get, .vid_set = nvkm_voltgpio_set, + .speedo_read = gk104_volt_speedo_read, }; int diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gm20b.c index 74db4d28930f..2925b9cae681 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gm20b.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gm20b.c @@ -25,7 +25,7 @@ #include <core/tegra.h> -const struct cvb_coef gm20b_cvb_coef[] = { +static const struct cvb_coef gm20b_cvb_coef[] = { /* KHz, c0, c1, c2 */ /* 76800 */ { 1786666, -85625, 1632 }, /* 153600 */ { 1846729, -87525, 1632 }, @@ -58,7 +58,7 @@ static const struct cvb_coef gm20b_na_cvb_coef[] = { /* 998400 */ { 1316991, 8144, -940, 808, -21583, 226 }, }; -const u32 speedo_to_vmin[] = { +static const u32 speedo_to_vmin[] = { /* 0, 1, 2, 3, 4, */ 950000, 840000, 818750, 840000, 810000, }; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gpio.c b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gpio.c index d2bac1d77819..443c031b966b 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gpio.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/gpio.c @@ -25,6 +25,7 @@ #include <subdev/bios.h> #include <subdev/bios/gpio.h> #include <subdev/gpio.h> +#include "priv.h" static const u8 tags[] = { DCB_GPIO_VID0, DCB_GPIO_VID1, DCB_GPIO_VID2, DCB_GPIO_VID3, diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/priv.h index d5140d991161..354bafe4b4e2 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/volt/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/volt/priv.h @@ -9,11 +9,13 @@ int nvkm_volt_new_(const struct nvkm_volt_func *, struct nvkm_device *, int index, struct nvkm_volt **); struct nvkm_volt_func { + int (*oneinit)(struct nvkm_volt *); int (*volt_get)(struct nvkm_volt *); int (*volt_set)(struct nvkm_volt *, u32 uv); int (*vid_get)(struct nvkm_volt *); int (*vid_set)(struct nvkm_volt *, u8 vid); int (*set_id)(struct nvkm_volt *, u8 id, int condition); + int (*speedo_read)(struct nvkm_volt *); }; int nvkm_voltgpio_init(struct nvkm_volt *); @@ -23,4 +25,6 @@ int nvkm_voltgpio_set(struct nvkm_volt *, u8); int nvkm_voltpwm_init(struct nvkm_volt *volt); int nvkm_voltpwm_get(struct nvkm_volt *volt); int nvkm_voltpwm_set(struct nvkm_volt *volt, u32 uv); + +int gf100_volt_oneinit(struct nvkm_volt *); #endif |