Crash in bm_evict_inode() during shutdown.

Crash in bm_evict_inode() during shutdown.

[Stanislaw Gruszka]

On 4.14-rc4 updated today, I have OOPS on bm_evict_inode(). It happens
every time during shutdown of the system.

Seems e = inode->i_private is NULL and e->flags crashes.

Picture with the OOPS call trace is here:
http://people.redhat.com/sgruszka/DSC_0733.JPG

Thanks
Stanislaw

Re: Crash in bm_evict_inode() during shutdown.

[Oleg Nesterov]

On 10/13, Stanislaw Gruszka wrote:
>
> On 4.14-rc4 updated today, I have OOPS on bm_evict_inode(). It happens
> every time during shutdown of the system.
>
> Seems e = inode->i_private is NULL and e->flags crashes.

Yes, sorry, my fault. Fixed by the patch from Eryu, already in -mm tree.

Oleg.

binfmt_misc: Node could be NULL when evicting inode

inode->i_private is assigned by a Node pointer only after
registering a new binary format, and it's possible to be NULL if
inode was created by bm_fill_super() (or iput() was called by the
error path in bm_register_write()), and this could result in NULL
pointer dereference when evicting such an inode. e.g. mount
binfmt_misc filesystem then umount it immediately:

  mount -t binfmt_misc binfmt_misc /proc/sys/fs/binfmt_misc
  umount /proc/sys/fs/binfmt_misc

[ 9379.678259] BUG: unable to handle kernel NULL pointer dereference at 0000000000000013
[ 9379.985952] IP: bm_evict_inode+0x16/0x40 [binfmt_misc]
...
[ 9380.964911] Call Trace:
[ 9380.977633]  evict+0xd3/0x1a0
[ 9380.994449]  iput+0x17d/0x1d0
[ 9381.010306]  dentry_unlink_inode+0xb9/0xf0
[ 9381.034046]  __dentry_kill+0xc7/0x170
[ 9381.055145]  shrink_dentry_list+0x122/0x280
[ 9381.078908]  shrink_dcache_parent+0x39/0x90
[ 9381.103082]  do_one_tree+0x12/0x40
[ 9381.122005]  shrink_dcache_for_umount+0x2d/0x90
[ 9381.146517]  generic_shutdown_super+0x1f/0x120
[ 9381.171644]  kill_litter_super+0x29/0x40
[ 9381.193513]  deactivate_locked_super+0x43/0x70
[ 9381.219177]  deactivate_super+0x45/0x60
[ 9381.240130]  cleanup_mnt+0x3f/0x70
[ 9381.259064]  __cleanup_mnt+0x12/0x20
[ 9381.279802]  task_work_run+0x86/0xa0
[ 9381.299612]  exit_to_usermode_loop+0x6d/0x99
[ 9381.323872]  syscall_return_slowpath+0xba/0xf0
[ 9381.350464]  entry_SYSCALL_64_fastpath+0xa3/0xa

Fix it by making sure Node (e) is not NULL.

Cc: Oleg Nesterov <oleg@redhat.com>
Fixes: 83f918274e4b ("exec: binfmt_misc: shift filp_close(interp_file) from kill_node() to bm_evict_inode()")
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
---
v2:
- update commit log as suggested by Oleg

Hi Andrew, could you please help replace the v1 patch in -mm tree with
this v2? The new commit log describes the problem more accurately.
Thanks!

 fs/binfmt_misc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/binfmt_misc.c b/fs/binfmt_misc.c
index 2a46762def31..a7c5a9861bef 100644
--- a/fs/binfmt_misc.c
+++ b/fs/binfmt_misc.c
@@ -596,7 +596,7 @@ static void bm_evict_inode(struct inode *inode)
 {
 	Node *e = inode->i_private;
 
-	if (e->flags & MISC_FMT_OPEN_FILE)
+	if (e && e->flags & MISC_FMT_OPEN_FILE)
 		filp_close(e->interp_file, NULL);
 
 	clear_inode(inode);
-- 
2.13.6

Re: Crash in bm_evict_inode() during shutdown.

[Stanislaw Gruszka]

On Fri, Oct 13, 2017 at 04:47:45PM +0200, Oleg Nesterov wrote:
> Yes, sorry, my fault. Fixed by the patch from Eryu, already in -mm tree.

No worries, I just used power button to "help" the system to shutdown :-)

Now is fixed in mainline, thanks.
Stanislaw