On Sun, Nov 24, 2013 at 03:27:58PM +0000, Al Viro wrote: > which seems to leave two candidates - follow_dotdot_rcu() and path_init(), > both setting nd->inode to nd->path.dentry->d_inode... > > Could you try to reproduce it with something like > if (read_seqretry(&mount_lock, nd->m_seq)) > goto failed; > slapped before the success exit in follow_dotdot_rcu(), just to see if > we are hitting some races with umount here? Still reproducable with the diff below. Fixed by reverting the RCU'd vfsmounts. diff --git a/fs/namei.c b/fs/namei.c index 8f77a8c..856e4d5 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -1158,6 +1158,10 @@ static int follow_dotdot_rcu(struct nameidata *nd) } follow_mount_rcu(nd); nd->inode = nd->path.dentry->d_inode; + + if (read_seqretry(&mount_lock, nd->m_seq)) + goto failed; + return 0; failed: -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
