ocfs2: __ocfs2_mknod_locked should return error when ocfs2_create_new_inode_locks() failed
PATCH 1/1 ocfs2: __ocfs2_mknod_locked should return error when ocfs2_create_new_inode_locks() failed
When ocfs2_create_new_inode_locks() return error, inode open lock may not be obtainted for this inode. So other nodes can remove this file and free dinode when inode still remain in memory on this node, which is not correct and may trigger BUG. So __ocfs2_mknod_locked should return error when ocfs2_create_new_inode_locks() failed.
Node_1 Node_2
create fileA, call ocfs2_mknod()
-> ocfs2_get_init_inode(), allocate inodeA -> ocfs2_claim_new_inode(), claim dinode(dinodeA) -> call ocfs2_create_new_inode_locks(),
create open lock failed, return error
-> __ocfs2_mknod_locked return success
unlink fileA try open lock succeed, and free dinodeA
create another file, call ocfs2_mknod()
-> ocfs2_get_init_inode(), allocate inodeB
-> ocfs2_claim_new_inode(), as Node_2 had freed dinodeA,
so claim dinodeA and update generation for dinodeA
call __ocfs2_drop_dl_inodes()->ocfs2_delete_inode() to free inodeA, and finally triggers BUG on(inode->i_generation != le32_to_cpu(fe->i_generation)) in function ocfs2_inode_lock_update().
--- fs/ocfs2/namei.c | 3 --- 1 file changed, 3 deletions(-)
fs/ocfs2/namei.c +0 -3
@@ -576,9 +576,6 @@ static int __ocfs2_mknod_locked(struct inode *dir,mlog_errno (status); } - status =0 ;/* error in ocfs2_create_new_inode_locks is not -* critical */ - leave:if (status <0 ) {if (*new_fe_bh) {
Signature · 1 line
1.8.4.3
Joel Becker
maintainerPATCH 1/1 ocfs2: __ocfs2_mknod_locked should return error when ocfs2_create_new_inode_locks() failed
Quoted 23 lines · Xue jiufei
When
ocfs2_create_new_inode_locks()return error, inode open lock may not be obtainted for this inode. So other nodes can remove this file and free dinode when inode still remain in memory on this node, which is not correct and may trigger BUG. So__ocfs2_mknod_lockedshould return error whenocfs2_create_new_inode_locks()failed.
Node_1Node_2
create fileA, callocfs2_mknod()
->ocfs2_get_init_inode(), allocate inodeA ->ocfs2_claim_new_inode(), claim dinode(dinodeA) -> callocfs2_create_new_inode_locks(),
create open lock failed, return error
->__ocfs2_mknod_lockedreturn successunlink fileA try open lock succeed, and free dinodeA
create another file, call
ocfs2_mknod()
->ocfs2_get_init_inode(), allocate inodeB
->ocfs2_claim_new_inode(), asNode_2had freed dinodeA,
so claim dinodeA and update generation for dinodeA
Quoted 4 lines
call __ocfs2_drop_dl_inodes()->ocfs2_delete_inode() to free inodeA, and finally triggers BUG
on(inode->i_generation != le32_to_cpu(fe->i_generation))in functionocfs2_inode_lock_update().
Wow, that's a deep race, and it's some salty old code. I think I buy your description of the problem. I'm trying to figure out why we didn't hit this before.
What workload or tests triggered this? Do you know why ocfs2_create_new_inode_locks() failed in the first place? I suspect it almost never fails, which is why we haven't seen this issue. Also, which cluster stack was involved?
Finally, have you tested the result? I believe that the iput() machinery will do the right thing, but tests are better than my intuition.
Joel
Quoted 28 lines
Signed-off-by joyce.xue <[email protected]>--- fs/ocfs2/namei.c | 3 --- 1 file changed, 3 deletions(-)fs/ocfs2/namei.c +0
-3@@ -576,9 +576,6 @@ static int __ocfs2_mknod_locked(struct inode *dir,mlog_errno (status); } - status =0 ;/* error in ocfs2_create_new_inode_locks is not -* critical */ - leave:if (status <0 ) {if (*new_fe_bh) {Signature · 7 lines
1.8.4.3 _______________________________________________ Ocfs2-devel mailing list Ocfs2-devel at oss.oracle.com https://oss.oracle.com/mailman/listinfo/ocfs2-devel
Signature · 6 lines
"Under capitalism, man exploits man. Under Communism, it's just the opposite." - John Kenneth Galbraith http://www.jlbec.org/ jlbec at evilplan.org
PATCH 1/1 ocfs2: __ocfs2_mknod_locked should return error when ocfs2_create_new_inode_locks() failed
Quoted 38 lines · Joel Becker
Quoted 23 lines · Xue jiufei
When
ocfs2_create_new_inode_locks()return error, inode open lock may not be obtainted for this inode. So other nodes can remove this file and free dinode when inode still remain in memory on this node, which is not correct and may trigger BUG. So__ocfs2_mknod_lockedshould return error whenocfs2_create_new_inode_locks()failed.
Node_1Node_2
create fileA, callocfs2_mknod()
->ocfs2_get_init_inode(), allocate inodeA ->ocfs2_claim_new_inode(), claim dinode(dinodeA) -> callocfs2_create_new_inode_locks(),
create open lock failed, return error
->__ocfs2_mknod_lockedreturn successunlink fileA try open lock succeed, and free dinodeA
create another file, call
ocfs2_mknod()
->ocfs2_get_init_inode(), allocate inodeB
->ocfs2_claim_new_inode(), asNode_2had freed dinodeA,
so claim dinodeA and update generation for dinodeAQuoted 4 lines
call __ocfs2_drop_dl_inodes()->ocfs2_delete_inode() to free inodeA, and finally triggers BUG
on(inode->i_generation != le32_to_cpu(fe->i_generation))in functionocfs2_inode_lock_update().Wow, that's a deep race, and it's some salty old code. I think I buy your description of the problem. I'm trying to figure out why we didn't hit this before.
What workload or tests triggered this? Do you know why
ocfs2_create_new_inode_locks()failed in the first place? I suspect it almost never fails, which is why we haven't seen this issue. Also, which cluster stack was involved?
Quoted 6 lines
Finally, have you tested the result? I believe that the
iput()machinery will do the right thing, but tests are better than my intuition.Joel
We triggered this BUG while running racer.sh, but no helpful messages remained and it can not be reproduced. So we analyzed function ocfs2_mknod() and found this situation. To reproduce this bug, we force ocfs2_create_new_inode_locks() to return error without abtaining open lock, and insert some delay before __ocfs2_drop_dl_inodes(). So I am sorry that I don't known why ocfs2_create_new_inode_locks() failed.
Quoted 29 lines
Quoted 28 lines
Signed-off-by joyce.xue <[email protected]>--- fs/ocfs2/namei.c | 3 --- 1 file changed, 3 deletions(-)fs/ocfs2/namei.c +0
-3@@ -576,9 +576,6 @@ static int __ocfs2_mknod_locked(struct inode *dir,mlog_errno (status); } - status =0 ;/* error in ocfs2_create_new_inode_locks is not -* critical */ - leave:if (status <0 ) {if (*new_fe_bh) {Signature · 7 lines
1.8.4.3 _______________________________________________ Ocfs2-devel mailing list Ocfs2-devel at oss.oracle.com https://oss.oracle.com/mailman/listinfo/ocfs2-devel