Hi Mark,
On 2015/5/9 6:37, Mark Fasheh wrote:
> Thanks Joseph,
>
> This seems like a convoluted way of fixing the problem. The reason
> downconvert_thread_do_work is saving the # of blocked locks is so it doesn't
> spin forever while another process continually adds them - basically we want
> to
Thanks Joseph,
This seems like a convoluted way of fixing the problem. The reason
downconvert_thread_do_work is saving the # of blocked locks is so it doesn't
spin forever while another process continually adds them - basically we want
to do a certain amount of work and then move on. Instead why n
On Wed, 6 May 2015 19:12:54 +0800 Joseph Qi wrote:
> BUG_ON(list_empty(&osb->blocked_lock_list)) in
> ocfs2_downconvert_thread_do_work can be triggered in the following case:
> ocfs2dc has firstly saved osb->blocked_lock_count to local varibale
> processed, and then processes the dentry lockres.
BUG_ON(list_empty(&osb->blocked_lock_list)) in
ocfs2_downconvert_thread_do_work can be triggered in the following case:
ocfs2dc has firstly saved osb->blocked_lock_count to local varibale
processed, and then processes the dentry lockres.
During the dentry put, it calls iput and then deletes rw, ino
