[PATCH] drm/amdgpu: fix sem list soft lockup issue

[Junwei Zhang]

sem should be delete from the previous sem list
then add to the new sem list.

If adding sem to the list which is same as previous one,
it will cause endless loop when traverses the sem list.

[  264.133878] NMI watchdog: BUG: soft lockup - CPU#5 stuck for 22s!
...
[  264.133928] task: ffffa216ffbfbd00 task.stack: ffffa216f4220000
[  264.133929] RIP: 0010:[<ffffffffc0584523>]  [<ffffffffc0584523>]
amdgpu_sem_put+0x13/0x30 [amdgpu]
[  264.133983] RSP: 0018:ffffa216f4223cb0  EFLAGS: 00200286

Signed-off-by: Junwei Zhang <Jerry.Zhang at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 4 +++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c | 1 +
 2 files changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index ff20e11..aac43c8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -84,8 +84,10 @@ static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx)
 
 		mutex_lock(&ctx->rings[i].sem_lock);
 		/* release all the reset inserted SEM here */
-		list_for_each_entry_safe(sem, tmp, &ctx->rings[i].sem_list, list)
+		list_for_each_entry_safe(sem, tmp, &ctx->rings[i].sem_list, list) {
+			list_del_init(&sem->list);
 			amdgpu_sem_put(sem);
+		}
 
 		mutex_unlock(&ctx->rings[i].sem_lock);
 		mutex_destroy(&ctx->rings[i].sem_lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c
index 0302ea6..a7149ac 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sem.c
@@ -371,6 +371,7 @@ static int amdgpu_sem_cring_add(struct amdgpu_fpriv *fpriv,
 	if (r)
 		goto err;
 	mutex_lock(&ctx->rings[out_ring->idx].sem_lock);
+	list_del_init(&sem->list);
 	list_add(&sem->list, &ctx->rings[out_ring->idx].sem_list);
 	mutex_unlock(&ctx->rings[out_ring->idx].sem_lock);
 	amdgpu_sem_get(sem);
-- 
1.9.1