[PATCH AUTOSEL 6.1 04/25] drm/amdkfd: Fix a race condition of vram buffer unref in svm code

[Sasha Levin]

From: Xiaogang Chen <xiaogang.chen@xxxxxxx>

[ Upstream commit 709c348261618da7ed89d6c303e2ceb9e453ba74 ]

prange->svm_bo unref can happen in both mmu callback and a callback after
migrate to system ram. Both are async call in different tasks. Sync svm_bo
unref operation to avoid random "use-after-free".

Signed-off-by: Xiaogang Chen <xiaogang.chen@xxxxxxx>
Reviewed-by: Philip Yang <Philip.Yang@xxxxxxx>
Reviewed-by: Jesse Zhang <Jesse.Zhang@xxxxxxx>
Tested-by: Jesse Zhang <Jesse.Zhang@xxxxxxx>
Signed-off-by: Alex Deucher <alexander.deucher@xxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
 drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 63feea08904cb..86a6d6143f008 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -612,8 +612,15 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 
 void svm_range_vram_node_free(struct svm_range *prange)
 {
-	svm_range_bo_unref(prange->svm_bo);
-	prange->ttm_res = NULL;
+	/* serialize prange->svm_bo unref */
+	mutex_lock(&prange->lock);
+	/* prange->svm_bo has not been unref */
+	if (prange->ttm_res) {
+		prange->ttm_res = NULL;
+		mutex_unlock(&prange->lock);
+		svm_range_bo_unref(prange->svm_bo);
+	} else
+		mutex_unlock(&prange->lock);
 }
 
 struct amdgpu_device *
-- 
2.42.0