diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gem/i915_gem_ttm.c')
| -rw-r--r-- | drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 259 |
1 files changed, 182 insertions, 77 deletions
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c index a4aa9500fa17..f65fe86c02b5 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c @@ -5,11 +5,13 @@ #include <linux/shmem_fs.h> -#include <drm/ttm/ttm_bo_driver.h> -#include <drm/ttm/ttm_placement.h> #include <drm/drm_buddy.h> +#include <drm/drm_print.h> +#include <drm/ttm/ttm_placement.h> +#include <drm/ttm/ttm_tt.h> #include "i915_drv.h" +#include "i915_jiffies.h" #include "i915_ttm_buddy_manager.h" #include "intel_memory_region.h" #include "intel_region_ttm.h" @@ -65,8 +67,6 @@ static const struct ttm_place sys_placement_flags = { static struct ttm_placement i915_sys_placement = { .num_placement = 1, .placement = &sys_placement_flags, - .num_busy_placement = 1, - .busy_placement = &sys_placement_flags, }; /** @@ -140,46 +140,44 @@ i915_ttm_place_from_region(const struct intel_memory_region *mr, if (flags & I915_BO_ALLOC_CONTIGUOUS) place->flags |= TTM_PL_FLAG_CONTIGUOUS; if (offset != I915_BO_INVALID_OFFSET) { + WARN_ON(overflows_type(offset >> PAGE_SHIFT, place->fpfn)); place->fpfn = offset >> PAGE_SHIFT; + WARN_ON(overflows_type(place->fpfn + (size >> PAGE_SHIFT), place->lpfn)); place->lpfn = place->fpfn + (size >> PAGE_SHIFT); - } else if (mr->io_size && mr->io_size < mr->total) { + } else if (resource_size(&mr->io) && resource_size(&mr->io) < mr->total) { if (flags & I915_BO_ALLOC_GPU_ONLY) { place->flags |= TTM_PL_FLAG_TOPDOWN; } else { place->fpfn = 0; - place->lpfn = mr->io_size >> PAGE_SHIFT; + WARN_ON(overflows_type(resource_size(&mr->io) >> PAGE_SHIFT, place->lpfn)); + place->lpfn = resource_size(&mr->io) >> PAGE_SHIFT; } } } static void i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj, - struct ttm_place *requested, - struct ttm_place *busy, + struct ttm_place *places, struct ttm_placement *placement) { unsigned int num_allowed = obj->mm.n_placements; unsigned int flags = obj->flags; unsigned int i; - placement->num_placement = 1; i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] : - obj->mm.region, requested, obj->bo_offset, + obj->mm.region, &places[0], obj->bo_offset, obj->base.size, flags); /* Cache this on object? */ - placement->num_busy_placement = num_allowed; - for (i = 0; i < placement->num_busy_placement; ++i) - i915_ttm_place_from_region(obj->mm.placements[i], busy + i, - obj->bo_offset, obj->base.size, flags); - - if (num_allowed == 0) { - *busy = *requested; - placement->num_busy_placement = 1; + for (i = 0; i < num_allowed; ++i) { + i915_ttm_place_from_region(obj->mm.placements[i], + &places[i + 1], obj->bo_offset, + obj->base.size, flags); + places[i + 1].flags |= TTM_PL_FLAG_FALLBACK; } - placement->placement = requested; - placement->busy_placement = busy; + placement->num_placement = num_allowed + 1; + placement->placement = places; } static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev, @@ -271,23 +269,21 @@ static struct ttm_tt *i915_ttm_tt_create(struct ttm_buffer_object *bo, { struct drm_i915_private *i915 = container_of(bo->bdev, typeof(*i915), bdev); - struct ttm_resource_manager *man = - ttm_manager_type(bo->bdev, bo->resource->mem_type); struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); unsigned long ccs_pages = 0; enum ttm_caching caching; struct i915_ttm_tt *i915_tt; int ret; - if (!obj) + if (i915_ttm_is_ghost_object(bo)) return NULL; i915_tt = kzalloc(sizeof(*i915_tt), GFP_KERNEL); if (!i915_tt) return NULL; - if (obj->flags & I915_BO_ALLOC_CPU_CLEAR && - man->use_tt) + if (obj->flags & I915_BO_ALLOC_CPU_CLEAR && (!bo->resource || + ttm_manager_type(bo->bdev, bo->resource->mem_type)->use_tt)) page_flags |= TTM_TT_FLAG_ZERO_ALLOC; caching = i915_ttm_select_tt_caching(obj); @@ -362,7 +358,7 @@ static bool i915_ttm_eviction_valuable(struct ttm_buffer_object *bo, { struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); - if (!obj) + if (i915_ttm_is_ghost_object(bo)) return false; /* @@ -471,7 +467,7 @@ static int i915_ttm_shrink(struct drm_i915_gem_object *obj, unsigned int flags) struct ttm_placement place = {}; int ret; - if (!bo->ttm || bo->resource->mem_type != TTM_PL_SYSTEM) + if (!bo->ttm || i915_ttm_cpu_maps_iomem(bo->resource)) return 0; GEM_BUG_ON(!i915_tt->is_shmem); @@ -509,18 +505,15 @@ static int i915_ttm_shrink(struct drm_i915_gem_object *obj, unsigned int flags) static void i915_ttm_delete_mem_notify(struct ttm_buffer_object *bo) { struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); - intel_wakeref_t wakeref = 0; - - if (bo->resource && likely(obj)) { - /* ttm_bo_release() already has dma_resv_lock */ - if (i915_ttm_cpu_maps_iomem(bo->resource)) - wakeref = intel_runtime_pm_get(&to_i915(obj->base.dev)->runtime_pm); + /* + * This gets called twice by ttm, so long as we have a ttm resource or + * ttm_tt then we can still safely call this. Due to pipeline-gutting, + * we maybe have NULL bo->resource, but in that case we should always + * have a ttm alive (like if the pages are swapped out). + */ + if ((bo->resource || bo->ttm) && !i915_ttm_is_ghost_object(bo)) { __i915_gem_object_pages_fini(obj); - - if (wakeref) - intel_runtime_pm_put(&to_i915(obj->base.dev)->runtime_pm, wakeref); - i915_ttm_free_cached_io_rsgt(obj); } } @@ -608,10 +601,17 @@ i915_ttm_resource_get_st(struct drm_i915_gem_object *obj, static int i915_ttm_truncate(struct drm_i915_gem_object *obj) { struct ttm_buffer_object *bo = i915_gem_to_ttm(obj); - int err; + long err; WARN_ON_ONCE(obj->mm.madv == I915_MADV_WILLNEED); + err = dma_resv_wait_timeout(bo->base.resv, DMA_RESV_USAGE_BOOKKEEP, + true, 15 * HZ); + if (err < 0) + return err; + if (err == 0) + return -EBUSY; + err = i915_ttm_move_notify(bo); if (err) return err; @@ -624,7 +624,7 @@ static void i915_ttm_swap_notify(struct ttm_buffer_object *bo) struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); int ret; - if (!obj) + if (i915_ttm_is_ghost_object(bo)) return; ret = i915_ttm_move_notify(bo); @@ -649,7 +649,7 @@ bool i915_ttm_resource_mappable(struct ttm_resource *res) if (!i915_ttm_cpu_maps_iomem(res)) return true; - return bman_res->used_visible_size == bman_res->base.num_pages; + return bman_res->used_visible_size == PFN_UP(bman_res->base.size); } static int i915_ttm_io_mem_reserve(struct ttm_device *bdev, struct ttm_resource *mem) @@ -657,7 +657,7 @@ static int i915_ttm_io_mem_reserve(struct ttm_device *bdev, struct ttm_resource struct drm_i915_gem_object *obj = i915_ttm_to_gem(mem->bo); bool unknown_state; - if (!obj) + if (i915_ttm_is_ghost_object(mem->bo)) return -EINVAL; if (!kref_get_unless_zero(&obj->base.refcount)) @@ -690,15 +690,59 @@ static unsigned long i915_ttm_io_mem_pfn(struct ttm_buffer_object *bo, unsigned long base; unsigned int ofs; - GEM_BUG_ON(!obj); + GEM_BUG_ON(i915_ttm_is_ghost_object(bo)); GEM_WARN_ON(bo->ttm); base = obj->mm.region->iomap.base - obj->mm.region->region.start; - sg = __i915_gem_object_get_sg(obj, &obj->ttm.get_io_page, page_offset, &ofs, true); + sg = i915_gem_object_page_iter_get_sg(obj, &obj->ttm.get_io_page, page_offset, &ofs); return ((base + sg_dma_address(sg)) >> PAGE_SHIFT) + ofs; } +static int i915_ttm_access_memory(struct ttm_buffer_object *bo, + unsigned long offset, void *buf, + int len, int write) +{ + struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); + resource_size_t iomap = obj->mm.region->iomap.base - + obj->mm.region->region.start; + unsigned long page = offset >> PAGE_SHIFT; + unsigned long bytes_left = len; + + /* + * TODO: For now just let it fail if the resource is non-mappable, + * otherwise we need to perform the memcpy from the gpu here, without + * interfering with the object (like moving the entire thing). + */ + if (!i915_ttm_resource_mappable(bo->resource)) + return -EIO; + + offset -= page << PAGE_SHIFT; + do { + unsigned long bytes = min(bytes_left, PAGE_SIZE - offset); + void __iomem *ptr; + dma_addr_t daddr; + + daddr = i915_gem_object_get_dma_address(obj, page); + ptr = ioremap_wc(iomap + daddr + offset, bytes); + if (!ptr) + return -EIO; + + if (write) + memcpy_toio(ptr, buf, bytes); + else + memcpy_fromio(buf, ptr, bytes); + iounmap(ptr); + + page++; + buf += bytes; + bytes_left -= bytes; + offset = 0; + } while (bytes_left); + + return len; +} + /* * All callbacks need to take care not to downcast a struct ttm_buffer_object * without checking its subclass, since it might be a TTM ghost object. @@ -715,6 +759,7 @@ static struct ttm_device_funcs i915_ttm_bo_driver = { .delete_mem_notify = i915_ttm_delete_mem_notify, .io_mem_reserve = i915_ttm_io_mem_reserve, .io_mem_pfn = i915_ttm_io_mem_pfn, + .access_memory = i915_ttm_access_memory, }; /** @@ -735,12 +780,16 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, .interruptible = true, .no_wait_gpu = false, }; - int real_num_busy; + struct ttm_placement initial_placement; + struct ttm_place initial_place; int ret; /* First try only the requested placement. No eviction. */ - real_num_busy = fetch_and_zero(&placement->num_busy_placement); - ret = ttm_bo_validate(bo, placement, &ctx); + initial_placement.num_placement = 1; + memcpy(&initial_place, placement->placement, sizeof(struct ttm_place)); + initial_place.flags |= TTM_PL_FLAG_DESIRED; + initial_placement.placement = &initial_place; + ret = ttm_bo_validate(bo, &initial_placement, &ctx); if (ret) { ret = i915_ttm_err_to_gem(ret); /* @@ -755,14 +804,13 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, * If the initial attempt fails, allow all accepted placements, * evicting if necessary. */ - placement->num_busy_placement = real_num_busy; ret = ttm_bo_validate(bo, placement, &ctx); if (ret) return i915_ttm_err_to_gem(ret); } if (bo->ttm && !ttm_tt_is_populated(bo->ttm)) { - ret = ttm_tt_populate(bo->bdev, bo->ttm, &ctx); + ret = ttm_bo_populate(bo, &ctx); if (ret) return ret; @@ -779,8 +827,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, GEM_BUG_ON(obj->mm.rsgt); obj->mm.rsgt = rsgt; - __i915_gem_object_set_pages(obj, &rsgt->table, - i915_sg_dma_sizes(rsgt->table.sgl)); + __i915_gem_object_set_pages(obj, &rsgt->table); } GEM_BUG_ON(bo->ttm && ((obj->base.size >> PAGE_SHIFT) < bo->ttm->num_pages)); @@ -790,13 +837,17 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) { - struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS]; + struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1]; struct ttm_placement placement; + /* restricted by sg_alloc_table */ + if (overflows_type(obj->base.size >> PAGE_SHIFT, unsigned int)) + return -E2BIG; + GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS); /* Move to the requested placement. */ - i915_ttm_placement_from_obj(obj, &requested, busy, &placement); + i915_ttm_placement_from_obj(obj, places, &placement); return __i915_ttm_get_pages(obj, &placement); } @@ -826,9 +877,7 @@ static int __i915_ttm_migrate(struct drm_i915_gem_object *obj, i915_ttm_place_from_region(mr, &requested, obj->bo_offset, obj->base.size, flags); placement.num_placement = 1; - placement.num_busy_placement = 1; placement.placement = &requested; - placement.busy_placement = &requested; ret = __i915_ttm_get_pages(obj, &placement); if (ret) @@ -947,7 +996,7 @@ void i915_ttm_adjust_lru(struct drm_i915_gem_object *obj) * If we need to place an LMEM resource which doesn't need CPU * access then we should try not to victimize mappable objects * first, since we likely end up stealing more of the mappable - * portion. And likewise when we try to find space for a mappble + * portion. And likewise when we try to find space for a mappable * object, we know not to ever victimize objects that don't * occupy any mappable pages. */ @@ -982,7 +1031,7 @@ static void i915_ttm_delayed_free(struct drm_i915_gem_object *obj) { GEM_BUG_ON(!obj->ttm.created); - ttm_bo_put(i915_gem_to_ttm(obj)); + ttm_bo_fini(i915_gem_to_ttm(obj)); } static vm_fault_t vm_fault_ttm(struct vm_fault *vmf) @@ -990,15 +1039,11 @@ static vm_fault_t vm_fault_ttm(struct vm_fault *vmf) struct vm_area_struct *area = vmf->vma; struct ttm_buffer_object *bo = area->vm_private_data; struct drm_device *dev = bo->base.dev; - struct drm_i915_gem_object *obj; - intel_wakeref_t wakeref = 0; + struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); + intel_wakeref_t wakeref = NULL; vm_fault_t ret; int idx; - obj = i915_ttm_to_gem(bo); - if (!obj) - return VM_FAULT_SIGBUS; - /* Sanity check that we allow writing into this object */ if (unlikely(i915_gem_object_is_readonly(obj) && area->vm_flags & VM_WRITE)) @@ -1013,10 +1058,27 @@ static vm_fault_t vm_fault_ttm(struct vm_fault *vmf) return VM_FAULT_SIGBUS; } - if (i915_ttm_cpu_maps_iomem(bo->resource)) - wakeref = intel_runtime_pm_get(&to_i915(obj->base.dev)->runtime_pm); + /* + * This must be swapped out with shmem ttm_tt (pipeline-gutting). + * Calling ttm_bo_validate() here with TTM_PL_SYSTEM should only go as + * far as far doing a ttm_bo_move_null(), which should skip all the + * other junk. + */ + if (!bo->resource) { + struct ttm_operation_ctx ctx = { + .interruptible = true, + .no_wait_gpu = true, /* should be idle already */ + }; + int err; - if (!i915_ttm_resource_mappable(bo->resource)) { + GEM_BUG_ON(!bo->ttm || !(bo->ttm->page_flags & TTM_TT_FLAG_SWAPPED)); + + err = ttm_bo_validate(bo, i915_ttm_sys_placement(), &ctx); + if (err) { + dma_resv_unlock(bo->base.resv); + return VM_FAULT_SIGBUS; + } + } else if (!i915_ttm_resource_mappable(bo->resource)) { int err = -ENODEV; int i; @@ -1024,7 +1086,7 @@ static vm_fault_t vm_fault_ttm(struct vm_fault *vmf) struct intel_memory_region *mr = obj->mm.placements[i]; unsigned int flags; - if (!mr->io_size && mr->type != INTEL_MEMORY_SYSTEM) + if (!resource_size(&mr->io) && mr->type != INTEL_MEMORY_SYSTEM) continue; flags = obj->flags; @@ -1035,13 +1097,18 @@ static vm_fault_t vm_fault_ttm(struct vm_fault *vmf) } if (err) { - drm_dbg(dev, "Unable to make resource CPU accessible\n"); + drm_dbg_ratelimited(dev, + "Unable to make resource CPU accessible(err = %pe)\n", + ERR_PTR(err)); dma_resv_unlock(bo->base.resv); ret = VM_FAULT_SIGBUS; goto out_rpm; } } + if (i915_ttm_cpu_maps_iomem(bo->resource)) + wakeref = intel_runtime_pm_get(&to_i915(obj->base.dev)->runtime_pm); + if (drm_dev_enter(dev, &idx)) { ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot, TTM_BO_VM_NUM_PREFAULT); @@ -1053,16 +1120,21 @@ static vm_fault_t vm_fault_ttm(struct vm_fault *vmf) if (ret == VM_FAULT_RETRY && !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT)) goto out_rpm; - /* ttm_bo_vm_reserve() already has dma_resv_lock */ + /* + * ttm_bo_vm_reserve() already has dma_resv_lock. + * userfault_count is protected by dma_resv lock and rpm wakeref. + */ if (ret == VM_FAULT_NOPAGE && wakeref && !obj->userfault_count) { obj->userfault_count = 1; - mutex_lock(&to_gt(to_i915(obj->base.dev))->lmem_userfault_lock); - list_add(&obj->userfault_link, &to_gt(to_i915(obj->base.dev))->lmem_userfault_list); - mutex_unlock(&to_gt(to_i915(obj->base.dev))->lmem_userfault_lock); + spin_lock(&to_i915(obj->base.dev)->runtime_pm.lmem_userfault_lock); + list_add(&obj->userfault_link, &to_i915(obj->base.dev)->runtime_pm.lmem_userfault_list); + spin_unlock(&to_i915(obj->base.dev)->runtime_pm.lmem_userfault_lock); + + GEM_WARN_ON(!i915_ttm_cpu_maps_iomem(bo->resource)); } - if (wakeref & CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND) - intel_wakeref_auto(&to_gt(to_i915(obj->base.dev))->userfault_wakeref, + if (wakeref && CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND != 0) + intel_wakeref_auto(&to_i915(obj->base.dev)->runtime_pm.userfault_wakeref, msecs_to_jiffies_timeout(CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND)); i915_ttm_adjust_lru(obj); @@ -1094,7 +1166,7 @@ static void ttm_vm_open(struct vm_area_struct *vma) struct drm_i915_gem_object *obj = i915_ttm_to_gem(vma->vm_private_data); - GEM_BUG_ON(!obj); + GEM_BUG_ON(i915_ttm_is_ghost_object(vma->vm_private_data)); i915_gem_object_get(obj); } @@ -1103,7 +1175,7 @@ static void ttm_vm_close(struct vm_area_struct *vma) struct drm_i915_gem_object *obj = i915_ttm_to_gem(vma->vm_private_data); - GEM_BUG_ON(!obj); + GEM_BUG_ON(i915_ttm_is_ghost_object(vma->vm_private_data)); i915_gem_object_put(obj); } @@ -1124,7 +1196,29 @@ static u64 i915_ttm_mmap_offset(struct drm_i915_gem_object *obj) static void i915_ttm_unmap_virtual(struct drm_i915_gem_object *obj) { + struct ttm_buffer_object *bo = i915_gem_to_ttm(obj); + intel_wakeref_t wakeref = NULL; + + assert_object_held_shared(obj); + + if (i915_ttm_cpu_maps_iomem(bo->resource)) { + wakeref = intel_runtime_pm_get(&to_i915(obj->base.dev)->runtime_pm); + + /* userfault_count is protected by obj lock and rpm wakeref. */ + if (obj->userfault_count) { + spin_lock(&to_i915(obj->base.dev)->runtime_pm.lmem_userfault_lock); + list_del(&obj->userfault_link); + spin_unlock(&to_i915(obj->base.dev)->runtime_pm.lmem_userfault_lock); + obj->userfault_count = 0; + } + } + + GEM_WARN_ON(obj->userfault_count); + ttm_bo_unmap_virtual(i915_gem_to_ttm(obj)); + + if (wakeref) + intel_runtime_pm_put(&to_i915(obj->base.dev)->runtime_pm, wakeref); } static const struct drm_i915_gem_object_ops i915_gem_ttm_obj_ops = { @@ -1177,7 +1271,7 @@ void i915_ttm_bo_destroy(struct ttm_buffer_object *bo) } } -/** +/* * __i915_gem_ttm_object_init - Initialize a ttm-backed i915 gem object * @mem: The initial memory region for the object. * @obj: The gem object. @@ -1233,12 +1327,23 @@ int __i915_gem_ttm_object_init(struct intel_memory_region *mem, * If this function fails, it will call the destructor, but * our caller still owns the object. So no freeing in the * destructor until obj->ttm.created is true. - * Similarly, in delayed_destroy, we can't call ttm_bo_put() + * Similarly, in delayed_destroy, we can't call ttm_bo_fini() * until successful initialization. */ ret = ttm_bo_init_reserved(&i915->bdev, i915_gem_to_ttm(obj), bo_type, &i915_sys_placement, page_size >> PAGE_SHIFT, &ctx, NULL, NULL, i915_ttm_bo_destroy); + + /* + * XXX: The ttm_bo_init_reserved() functions returns -ENOSPC if the size + * is too big to add vma. The direct function that returns -ENOSPC is + * drm_mm_insert_node_in_range(). To handle the same error as other code + * that returns -E2BIG when the size is too large, it converts -ENOSPC to + * -E2BIG. + */ + if (size >> PAGE_SHIFT > INT_MAX && ret == -ENOSPC) + ret = -E2BIG; + if (ret) return i915_ttm_err_to_gem(ret); |