[PATCH 5.15.y] drm/amdkfd: stop restore retries on device unplug

[Sasha Levin <sashal@kernel.org>]

From: Olivier Kaloudoff <olivier.kaloudoff@gmail.com>

[ Upstream commit 3254ed0b15f2fe656c9cb36ccb59b92422440bf8 ]

amdgpu_amdkfd_restore_userptr_worker() and svm_range_restore_work()
reschedule themselves when validation fails with evicted objects
pending. After a surprise removal the restore paths fail with -ENODEV
permanently, so rescheduling just spins the worker uselessly until
cancel_delayed_work_sync() in fini finally stops it - which can take
a while, since process teardown itself is delayed by the dead device.

Treat -ENODEV as terminal in both workers.

Cc: Xiaogang Chen <xiaogang.chen@amd.com>
Signed-off-by: Olivier Kaloudoff <olivier.kaloudoff@gmail.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 266fab83139600e58db7e900e930c722e619b78c)
Cc: stable@vger.kernel.org
[ sashal: Adapted notifier-locked retry handling to the existing atomic eviction counter checks. ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c    | 17 ++++++++++++-----
 drivers/gpu/drm/amd/amdkfd/kfd_svm.c            |  8 +++++---
 2 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 07aa9f8de2644..b89bcf84cfa9d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -2188,6 +2188,7 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
 	struct task_struct *usertask;
 	struct mm_struct *mm;
 	int evicted_bos;
+	int ret;
 
 	evicted_bos = atomic_read(&process_info->evicted_bos);
 	if (!evicted_bos)
@@ -2205,7 +2206,8 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
 
 	mutex_lock(&process_info->lock);
 
-	if (update_invalid_user_pages(process_info, mm))
+	ret = update_invalid_user_pages(process_info, mm);
+	if (ret)
 		goto unlock_out;
 	/* userptr_inval_list can be empty if all evicted userptr BOs
 	 * have been freed. In that case there is nothing to validate
@@ -2215,7 +2217,8 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
 		if (atomic_read(&process_info->evicted_bos) != evicted_bos)
 			goto unlock_out; /* Concurrent eviction, try again */
 
-		if (validate_invalid_user_pages(process_info))
+		ret = validate_invalid_user_pages(process_info);
+		if (ret)
 			goto unlock_out;
 	}
 	/* Final check for concurrent evicton and atomic update. If
@@ -2224,8 +2227,10 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
 	 * reference counting inside KFD will handle this case.
 	 */
 	if (atomic_cmpxchg(&process_info->evicted_bos, evicted_bos, 0) !=
-	    evicted_bos)
+	    evicted_bos) {
+		ret = -EAGAIN;
 		goto unlock_out;
+	}
 	evicted_bos = 0;
 	if (kgd2kfd_resume_mm(mm)) {
 		pr_err("%s: Failed to resume KFD\n", __func__);
@@ -2239,8 +2244,10 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
 	mmput(mm);
 	put_task_struct(usertask);
 
-	/* If validation failed, reschedule another attempt */
-	if (evicted_bos)
+	/* If validation failed, reschedule another attempt, unless the
+	 * device is gone for good
+	 */
+	if (evicted_bos && ret != -ENODEV)
 		schedule_delayed_work(&process_info->restore_userptr_work,
 			msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index ff776944793a8..0b20ef89c4ad5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -1575,7 +1575,7 @@ static void svm_range_restore_work(struct work_struct *work)
 	struct mm_struct *mm;
 	int evicted_ranges;
 	int invalid;
-	int r;
+	int r = 0;
 
 	svms = container_of(dwork, struct svm_range_list, restore_work);
 	evicted_ranges = atomic_read(&svms->evicted_ranges);
@@ -1645,8 +1645,10 @@ static void svm_range_restore_work(struct work_struct *work)
 	mutex_unlock(&svms->lock);
 	mmap_write_unlock(mm);
 
-	/* If validation failed, reschedule another attempt */
-	if (evicted_ranges) {
+	/* If validation failed, reschedule another attempt, unless the
+	 * device is gone for good
+	 */
+	if (evicted_ranges && r != -ENODEV) {
 		pr_debug("reschedule to restore svm range\n");
 		schedule_delayed_work(&svms->restore_work,
 			msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS));
-- 
2.53.0