stable.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] fuse: fix live lock in fuse_iget()
@ 2021-03-04  9:09 Amir Goldstein
  2021-03-04 12:37 ` Miklos Szeredi
  0 siblings, 1 reply; 2+ messages in thread
From: Amir Goldstein @ 2021-03-04  9:09 UTC (permalink / raw)
  To: Miklos Szeredi; +Cc: Jan Kara, linux-fsdevel, stable

Commit 5d069dbe8aaf ("fuse: fix bad inode") replaced make_bad_inode()
in fuse_iget() with a private implementation fuse_make_bad().

The private implementation fails to remove the bad inode from inode
cache, so the retry loop with iget5_locked() finds the same bad inode
and marks it bad forever.

kmsg snip:

[ ] rcu: INFO: rcu_sched self-detected stall on CPU
...
[ ]  ? bit_wait_io+0x50/0x50
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ? find_inode.isra.32+0x60/0xb0
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ilookup5_nowait+0x65/0x90
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ilookup5.part.36+0x2e/0x80
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ? fuse_inode_eq+0x20/0x20
[ ]  iget5_locked+0x21/0x80
[ ]  ? fuse_inode_eq+0x20/0x20
[ ]  fuse_iget+0x96/0x1b0

Fixes: 5d069dbe8aaf ("fuse: fix bad inode")
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
---

Miklos,

This live lock showed up on a stress test in my system after upgarde to
kernel 5.10.  It's 100% reproducible when trying to rsync from a davfs2
filesystem with severel 1000 files.

Thanks,
Amir.

 fs/fuse/fuse_i.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h
index 68cca8d4db6e..63d97a15ffde 100644
--- a/fs/fuse/fuse_i.h
+++ b/fs/fuse/fuse_i.h
@@ -863,6 +863,7 @@ static inline u64 fuse_get_attr_version(struct fuse_conn *fc)
 
 static inline void fuse_make_bad(struct inode *inode)
 {
+	remove_inode_hash(inode);
 	set_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state);
 }
 
-- 
2.30.0


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] fuse: fix live lock in fuse_iget()
  2021-03-04  9:09 [PATCH] fuse: fix live lock in fuse_iget() Amir Goldstein
@ 2021-03-04 12:37 ` Miklos Szeredi
  0 siblings, 0 replies; 2+ messages in thread
From: Miklos Szeredi @ 2021-03-04 12:37 UTC (permalink / raw)
  To: Amir Goldstein; +Cc: Jan Kara, linux-fsdevel, stable

On Thu, Mar 4, 2021 at 10:09 AM Amir Goldstein <amir73il@gmail.com> wrote:
>
> Commit 5d069dbe8aaf ("fuse: fix bad inode") replaced make_bad_inode()
> in fuse_iget() with a private implementation fuse_make_bad().
>
> The private implementation fails to remove the bad inode from inode
> cache, so the retry loop with iget5_locked() finds the same bad inode
> and marks it bad forever.

Thanks, applied.

Miklos

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-03-04 12:39 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-04  9:09 [PATCH] fuse: fix live lock in fuse_iget() Amir Goldstein
2021-03-04 12:37 ` Miklos Szeredi

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).