On 2018年10月19日 19:26, zhoucm1 wrote:
On 2018年10月19日 18:50, Chris Wilson wrote:
Quoting Chunming Zhou (2018-10-19 11:26:41)
Signed-off-by: Chunming Zhou <david1.z...@amd.com>
Cc: Daniel Vetter <dan...@ffwll.ch>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: Christian König <christian.koe...@amd.com>
---
drivers/gpu/drm/drm_syncobj.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/drm_syncobj.c
b/drivers/gpu/drm/drm_syncobj.c
index 57bf6006394d..2f3c14cb5156 100644
--- a/drivers/gpu/drm/drm_syncobj.c
+++ b/drivers/gpu/drm/drm_syncobj.c
@@ -344,13 +344,16 @@ void drm_syncobj_replace_fence(struct
drm_syncobj *syncobj,
drm_syncobj_create_signal_pt(syncobj, fence, pt_value);
if (fence) {
struct drm_syncobj_cb *cur, *tmp;
+ struct list_head cb_list;
+ INIT_LIST_HEAD(&cb_list);
LIST_HEAD(cb_list); // does both in one
spin_lock(&syncobj->lock);
- list_for_each_entry_safe(cur, tmp,
&syncobj->cb_list, node) {
+ list_splice_init(&syncobj->cb_list, &cb_list);
Steal the snapshot of the list under the lock, ok.
+ spin_unlock(&syncobj->lock);
+ list_for_each_entry_safe(cur, tmp, &cb_list, node) {
list_del_init(&cur->node);
Races against external caller of drm_syncobj_remove_callback(). However,
it looks like that race is just fine, but we don't guard against the
struct drm_syncobj_cb itself being freed, leading to all sort of fun for
an interrupted drm_syncobj_array_wait_timeout.
Thanks quick review, I will use "while (!list_empty()) { e =
list_first_entry(); list_del(e)" to avoid deadlock.
this still cannot resolve freeing problem, do you mind I change
spinlock to mutex?
Thanks,
David Zhou
will send v2 in one minute.
Regards,
David Zhou
That kfree seems to undermine the validity of stealing the list.
-Chris
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
