On Wed, Mar 11, 2015 at 4:51 PM, Alex Deucher wrote: > On Wed, Mar 11, 2015 at 2:21 PM, Christian König > wrote: >> On 11.03.2015 16:44, Alex Deucher wrote: >>> >>> radeon_bo_create() calls radeon_ttm_placement_from_domain() >>> before ttm_bo_init() is called. radeon_ttm_placement_from_domain() >>> uses the ttm bo size to determine when to select top down >>> allocation but since the ttm bo is not initialized yet the >>> check is always false. >>> >>> Noticed-by: Oded Gabbay >>> Signed-off-by: Alex Deucher >>> Cc: stable@vger.kernel.org >> >> >> And I was already wondering why the heck the BOs always made this ping/pong >> in memory after creation. >> >> Patch is Reviewed-by: Christian König > > And fixing that promptly broke VCE due to vram location requirements. > Updated patch attached. Thoughts? And one more take to make things a bit more explicit for static kernel driver allocations. Alex > > Alex > >> >> Regards, >> Christian. >> >> >>> --- >>> drivers/gpu/drm/radeon/radeon.h | 3 ++- >>> drivers/gpu/drm/radeon/radeon_gem.c | 2 +- >>> drivers/gpu/drm/radeon/radeon_mn.c | 2 +- >>> drivers/gpu/drm/radeon/radeon_object.c | 17 ++++++++++------- >>> drivers/gpu/drm/radeon/radeon_ttm.c | 12 ++++++++---- >>> 5 files changed, 22 insertions(+), 14 deletions(-) >>> >>> diff --git a/drivers/gpu/drm/radeon/radeon.h >>> b/drivers/gpu/drm/radeon/radeon.h >>> index 5587603..726e89f 100644 >>> --- a/drivers/gpu/drm/radeon/radeon.h >>> +++ b/drivers/gpu/drm/radeon/radeon.h >>> @@ -2970,7 +2970,8 @@ extern void radeon_surface_init(struct radeon_device >>> *rdev); >>> extern int radeon_cs_parser_init(struct radeon_cs_parser *p, void >>> *data); >>> extern void radeon_legacy_set_clock_gating(struct radeon_device *rdev, >>> int enable); >>> extern void radeon_atom_set_clock_gating(struct radeon_device *rdev, int >>> enable); >>> -extern void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 >>> domain); >>> +extern void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 >>> domain, >>> + u64 size); >>> extern bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo); >>> extern int radeon_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, >>> uint32_t flags); >>> diff --git a/drivers/gpu/drm/radeon/radeon_gem.c >>> b/drivers/gpu/drm/radeon/radeon_gem.c >>> index ac3c131..d613d0c 100644 >>> --- a/drivers/gpu/drm/radeon/radeon_gem.c >>> +++ b/drivers/gpu/drm/radeon/radeon_gem.c >>> @@ -337,7 +337,7 @@ int radeon_gem_userptr_ioctl(struct drm_device *dev, >>> void *data, >>> goto release_object; >>> } >>> - radeon_ttm_placement_from_domain(bo, >>> RADEON_GEM_DOMAIN_GTT); >>> + radeon_ttm_placement_from_domain(bo, >>> RADEON_GEM_DOMAIN_GTT, bo->tbo.mem.size); >>> r = ttm_bo_validate(&bo->tbo, &bo->placement, true, >>> false); >>> radeon_bo_unreserve(bo); >>> up_read(¤t->mm->mmap_sem); >>> diff --git a/drivers/gpu/drm/radeon/radeon_mn.c >>> b/drivers/gpu/drm/radeon/radeon_mn.c >>> index a69bd44..e51f09b 100644 >>> --- a/drivers/gpu/drm/radeon/radeon_mn.c >>> +++ b/drivers/gpu/drm/radeon/radeon_mn.c >>> @@ -141,7 +141,7 @@ static void radeon_mn_invalidate_range_start(struct >>> mmu_notifier *mn, >>> DRM_ERROR("(%d) failed to wait for user >>> bo\n", r); >>> } >>> - radeon_ttm_placement_from_domain(bo, >>> RADEON_GEM_DOMAIN_CPU); >>> + radeon_ttm_placement_from_domain(bo, >>> RADEON_GEM_DOMAIN_CPU, bo->tbo.mem.size); >>> r = ttm_bo_validate(&bo->tbo, &bo->placement, false, >>> false); >>> if (r) >>> DRM_ERROR("(%d) failed to validate user bo\n", r); >>> diff --git a/drivers/gpu/drm/radeon/radeon_object.c >>> b/drivers/gpu/drm/radeon/radeon_object.c >>> index 43e0994..07f8fd5 100644 >>> --- a/drivers/gpu/drm/radeon/radeon_object.c >>> +++ b/drivers/gpu/drm/radeon/radeon_object.c >>> @@ -93,7 +93,8 @@ bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object >>> *bo) >>> return false; >>> } >>> -void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 >>> domain) >>> +void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain, >>> + u64 size) >>> { >>> u32 c = 0, i; >>> @@ -179,7 +180,7 @@ void radeon_ttm_placement_from_domain(struct >>> radeon_bo *rbo, u32 domain) >>> * improve fragmentation quality. >>> * 512kb was measured as the most optimal number. >>> */ >>> - if (rbo->tbo.mem.size > 512 * 1024) { >>> + if (size > 512 * 1024) { >>> for (i = 0; i < c; i++) { >>> rbo->placements[i].flags |= TTM_PL_FLAG_TOPDOWN; >>> } >>> @@ -252,7 +253,7 @@ int radeon_bo_create(struct radeon_device *rdev, >>> bo->flags &= ~RADEON_GEM_GTT_WC; >>> #endif >>> - radeon_ttm_placement_from_domain(bo, domain); >>> + radeon_ttm_placement_from_domain(bo, domain, size); >>> /* Kernel allocation are uninterruptible */ >>> down_read(&rdev->pm.mclk_lock); >>> r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type, >>> @@ -350,7 +351,7 @@ int radeon_bo_pin_restricted(struct radeon_bo *bo, u32 >>> domain, u64 max_offset, >>> return 0; >>> } >>> - radeon_ttm_placement_from_domain(bo, domain); >>> + radeon_ttm_placement_from_domain(bo, domain, bo->tbo.mem.size); >>> for (i = 0; i < bo->placement.num_placement; i++) { >>> /* force to pin into visible video ram */ >>> if ((bo->placements[i].flags & TTM_PL_FLAG_VRAM) && >>> @@ -557,7 +558,7 @@ int radeon_bo_list_validate(struct radeon_device >>> *rdev, >>> } >>> retry: >>> - radeon_ttm_placement_from_domain(bo, domain); >>> + radeon_ttm_placement_from_domain(bo, domain, >>> bo->tbo.mem.size); >>> if (ring == R600_RING_TYPE_UVD_INDEX) >>> radeon_uvd_force_into_uvd_segment(bo, >>> allowed); >>> @@ -800,7 +801,8 @@ int radeon_bo_fault_reserve_notify(struct >>> ttm_buffer_object *bo) >>> return 0; >>> /* hurrah the memory is not visible ! */ >>> - radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM); >>> + radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM, >>> + rbo->tbo.mem.size); >>> lpfn = rdev->mc.visible_vram_size >> PAGE_SHIFT; >>> for (i = 0; i < rbo->placement.num_placement; i++) { >>> /* Force into visible VRAM */ >>> @@ -810,7 +812,8 @@ int radeon_bo_fault_reserve_notify(struct >>> ttm_buffer_object *bo) >>> } >>> r = ttm_bo_validate(bo, &rbo->placement, false, false); >>> if (unlikely(r == -ENOMEM)) { >>> - radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_GTT); >>> + radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_GTT, >>> + rbo->tbo.mem.size); >>> return ttm_bo_validate(bo, &rbo->placement, false, false); >>> } else if (unlikely(r != 0)) { >>> return r; >>> diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c >>> b/drivers/gpu/drm/radeon/radeon_ttm.c >>> index d02aa1d..ce8ed2d 100644 >>> --- a/drivers/gpu/drm/radeon/radeon_ttm.c >>> +++ b/drivers/gpu/drm/radeon/radeon_ttm.c >>> @@ -197,7 +197,8 @@ static void radeon_evict_flags(struct >>> ttm_buffer_object *bo, >>> switch (bo->mem.mem_type) { >>> case TTM_PL_VRAM: >>> if >>> (rbo->rdev->ring[radeon_copy_ring_index(rbo->rdev)].ready == false) >>> - radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_CPU); >>> + radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_CPU, >>> + >>> rbo->tbo.mem.size); >>> else if (rbo->rdev->mc.visible_vram_size < >>> rbo->rdev->mc.real_vram_size && >>> bo->mem.start < (rbo->rdev->mc.visible_vram_size >>> >> PAGE_SHIFT)) { >>> unsigned fpfn = rbo->rdev->mc.visible_vram_size >> >>> PAGE_SHIFT; >>> @@ -209,7 +210,8 @@ static void radeon_evict_flags(struct >>> ttm_buffer_object *bo, >>> * BOs to be evicted from VRAM >>> */ >>> radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_VRAM | >>> - >>> RADEON_GEM_DOMAIN_GTT); >>> + >>> RADEON_GEM_DOMAIN_GTT, >>> + >>> rbo->tbo.mem.size); >>> rbo->placement.num_busy_placement = 0; >>> for (i = 0; i < rbo->placement.num_placement; i++) >>> { >>> if (rbo->placements[i].flags & >>> TTM_PL_FLAG_VRAM) { >>> @@ -222,11 +224,13 @@ static void radeon_evict_flags(struct >>> ttm_buffer_object *bo, >>> } >>> } >>> } else >>> - radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_GTT); >>> + radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_GTT, >>> + >>> rbo->tbo.mem.size); >>> break; >>> case TTM_PL_TT: >>> default: >>> - radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_CPU); >>> + radeon_ttm_placement_from_domain(rbo, >>> RADEON_GEM_DOMAIN_CPU, >>> + rbo->tbo.mem.size); >>> } >>> *placement = rbo->placement; >>> } >> >>