librbd: possible deadlock if cluster connection closed after image #11037
Conversation
|
looks good, hoping this in 10.2.3 |
|
retest this please |
|
@xiaoxichen If I created a 10.2.2 dev packages with this fix, would you be able to verify that it solves the problem? |
|
@dillaman of course, YES! |
|
@xiaoxichen Trusty v10.2.2 + this patch is available under the wip-17254-jewel branch. CentOS 7 builder failed due to environment issues. |
|
@dillaman ,thx, we are all in ubuntu trusty:) |
|
@dillaman ,deployed the patched library to the cluster. Lets bake and see. |
| Context *ctx = on_finish; | ||
| delete this; | ||
|
|
||
| ctx->complete(r); |
There was a problem hiding this comment.
@dillaman Shouldn't it complete with ret_val?
|
@dillaman , it goes well in our enviroment for more than 12 hours, looking good except it should complete with ret_val |
Fixes: http://tracker.ceph.com/issues/17254 Signed-off-by: Jason Dillaman <dillaman@redhat.com>
librbd: possible deadlock if cluster connection closed after image #11037
| Context *ctx = on_finish; | ||
| delete this; | ||
|
|
||
| ctx->complete(ret_val); |
There was a problem hiding this comment.
@dillaman Ah, missed this: we can't use ret_val here. Need to copy it before delete.
Fixes: http://tracker.ceph.com/issues/17254
Signed-off-by: Jason Dillaman dillaman@redhat.com