Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/xe: Reject BO eviction if BO is bound to current VM
@ 2025-01-10 21:01 Oak Zeng
  2025-01-12 10:44 ` Thomas Hellström
                   ` (8 more replies)
  0 siblings, 9 replies; 12+ messages in thread
From: Oak Zeng @ 2025-01-10 21:01 UTC (permalink / raw)
  To: intel-xe; +Cc: Thomas.Hellstrom

This is a follow up fix for
https://patchwork.freedesktop.org/patch/msgid/20241203021929.1919730-1-oak.zeng@intel.com
The overall goal is to fail vm_bind when there is memory pressure. See more
details in the commit message of above patch. Abbove patch fixes the issue
when user pass in a vm_id parameter during gem_create. If user doesn't pass
in a vm_id during gem_create, above patch doesn't help.

This patch further reject BO eviction (which could be triggered by bo validation)
if BO is bound to the current VM. vm_bind could fail due to the eviction failure.
The BO to VM reverse mapping structure is used to determine whether BO is bound
to VM.

v2:
Move vm_bo definition from function scope to if(evict) clause (Thomas)
Further constraint the condition by adding ctx->resv (Thomas)
Add a short comment describe the change.

Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Oak Zeng <oak.zeng@intel.com>
---
 drivers/gpu/drm/xe/xe_bo.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
index 4a623a98d9f1f..b024c63d8d296 100644
--- a/drivers/gpu/drm/xe/xe_bo.c
+++ b/drivers/gpu/drm/xe/xe_bo.c
@@ -713,6 +713,21 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
 		goto out;
 	}
 
+	/* Reject BO eviction if BO is bound to current VM. */
+	if (evict && ctx->resv) {
+		struct drm_gpuvm_bo *vm_bo;
+
+		drm_gem_for_each_gpuvm_bo(vm_bo, &bo->ttm.base) {
+			struct xe_vm *vm = gpuvm_to_vm(vm_bo->vm);
+
+			if (xe_vm_resv(vm) == ctx->resv &&
+			    xe_vm_in_preempt_fence_mode(vm)) {
+				ret = -EBUSY;
+				goto out;
+			}
+		}
+	}
+
 	/*
 	 * Failed multi-hop where the old_mem is still marked as
 	 * TTM_PL_FLAG_TEMPORARY, should just be a dummy move.
-- 
2.26.3


^ permalink raw reply related	[flat|nested] 12+ messages in thread
* [PATCH] drm/xe: Reject BO eviction if BO is bound to current VM
@ 2024-12-17 23:13 Oak Zeng
  2025-01-10 14:45 ` Thomas Hellström
  0 siblings, 1 reply; 12+ messages in thread
From: Oak Zeng @ 2024-12-17 23:13 UTC (permalink / raw)
  To: intel-xe; +Cc: Thomas.Hellstrom, rodrigo.vivi

This is a follow up fix for https://patchwork.freedesktop.org/patch/msgid/20241203021929.1919730-1-oak.zeng@intel.com
The overall goal is to fail vm_bind when there is memory pressure. See more
details in the commit message of above patch. Abbove patch fixes the issue
when user pass in a vm_id parameter during gem_create. If user doesn't pass
in a vm_id during gem_create, above patch doesn't help.

This patch further reject BO eviction (which could be triggered by bo validation)
if BO is bound to the current VM. vm_bind could fail due to the eviction failure.
The BO to VM reverse mapping structure is used to determine whether BO is bound
to VM.

Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Oak Zeng <oak.zeng@intel.com>
---
 drivers/gpu/drm/xe/xe_bo.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
index 283cd02945708..abdeed1c325ea 100644
--- a/drivers/gpu/drm/xe/xe_bo.c
+++ b/drivers/gpu/drm/xe/xe_bo.c
@@ -664,6 +664,7 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
 	u32 old_mem_type = old_mem ? old_mem->mem_type : XE_PL_SYSTEM;
 	struct ttm_tt *ttm = ttm_bo->ttm;
 	struct xe_migrate *migrate = NULL;
+	struct drm_gpuvm_bo *vm_bo;
 	struct dma_fence *fence;
 	bool move_lacks_source;
 	bool tt_has_data;
@@ -713,6 +714,18 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
 		goto out;
 	}
 
+	if (evict) {
+		drm_gem_for_each_gpuvm_bo(vm_bo, &bo->ttm.base) {
+			struct xe_vm *vm = gpuvm_to_vm(vm_bo->vm);
+
+			if (xe_vm_resv(vm) == ctx->resv &&
+			    xe_vm_in_preempt_fence_mode(vm)) {
+				ret = -EBUSY;
+				goto out;
+			}
+		}
+	}
+
 	/*
 	 * Failed multi-hop where the old_mem is still marked as
 	 * TTM_PL_FLAG_TEMPORARY, should just be a dummy move.
-- 
2.26.3


^ permalink raw reply related	[flat|nested] 12+ messages in thread
end of thread, other threads:[~2025-01-13 20:48 UTC | newest]

Thread overview: 12+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-01-10 21:01 [PATCH] drm/xe: Reject BO eviction if BO is bound to current VM Oak Zeng
2025-01-12 10:44 ` Thomas Hellström
2025-01-13  5:43 ` ✓ CI.Patch_applied: success for drm/xe: Reject BO eviction if BO is bound to current VM (rev3) Patchwork
2025-01-13  5:43 ` ✗ CI.checkpatch: warning " Patchwork
2025-01-13  5:45 ` ✓ CI.KUnit: success " Patchwork
2025-01-13  6:12 ` ✓ CI.Build: " Patchwork
2025-01-13  6:16 ` ✓ CI.Hooks: " Patchwork
2025-01-13  6:18 ` ✓ CI.checksparse: " Patchwork
2025-01-13  6:46 ` ✓ Xe.CI.BAT: " Patchwork
2025-01-13 20:48 ` ✗ Xe.CI.Full: failure " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2024-12-17 23:13 [PATCH] drm/xe: Reject BO eviction if BO is bound to current VM Oak Zeng
2025-01-10 14:45 ` Thomas Hellström

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox