* [PATCH] autofs4 - panic after mount fail
@ 2006-11-12 9:48 Ian Kent
2006-11-13 11:07 ` David Howells
0 siblings, 1 reply; 5+ messages in thread
From: Ian Kent @ 2006-11-12 9:48 UTC (permalink / raw)
To: Andrew Morton; +Cc: bibo,mao, David Howells, Kernel Mailing List
Hi Andrew,
Here is a patch to resolve the panic on failed mount of an autofs
filesystem originally reported by Mao Bibo.
It addresses two issues that happen after the mount fail. The first a NULL
pointer reference to a field (pipe) in the autofs superblock info
structure and second the lack of super block cleanup by the autofs and
autofs4 modules.
The work has been done against 2.6.19-rc4 but I have verified that the
patch also applies against 2.6.19-rc5 and 2.6.19-rc5-mm1.
Signed-off-by: Ian Kent <raven@themaw.net>
---
--- linux-2.6.19-rc4/fs/autofs4/waitq.c.mount-fail-panic 2006-11-10 15:34:42.000000000 +0800
+++ linux-2.6.19-rc4/fs/autofs4/waitq.c 2006-11-10 19:49:57.000000000 +0800
@@ -41,10 +41,8 @@ void autofs4_catatonic_mode(struct autof
wake_up_interruptible(&wq->queue);
wq = nwq;
}
- if (sbi->pipe) {
- fput(sbi->pipe); /* Close the pipe */
- sbi->pipe = NULL;
- }
+ fput(sbi->pipe); /* Close the pipe */
+ sbi->pipe = NULL;
}
static int autofs4_write(struct file *file, const void *addr, int bytes)
--- linux-2.6.19-rc4/fs/autofs4/inode.c.mount-fail-panic 2006-11-10 15:34:42.000000000 +0800
+++ linux-2.6.19-rc4/fs/autofs4/inode.c 2006-11-12 17:07:26.000000000 +0800
@@ -99,6 +99,9 @@ static void autofs4_force_release(struct
struct dentry *this_parent = sbi->sb->s_root;
struct list_head *next;
+ if (!sbi->sb->s_root)
+ return;
+
spin_lock(&dcache_lock);
repeat:
next = this_parent->d_subdirs.next;
@@ -146,6 +149,14 @@ void autofs4_kill_sb(struct super_block
{
struct autofs_sb_info *sbi = autofs4_sbi(sb);
+ /*
+ * In the event of a failure in get_sb_nodev the superblock
+ * info is not present so nothing else has been setup, so
+ * just exit when we are called from deactivate_super.
+ */
+ if (!sbi)
+ return;
+
sb->s_fs_info = NULL;
if ( !sbi->catatonic )
@@ -310,7 +321,8 @@ int autofs4_fill_super(struct super_bloc
s->s_fs_info = sbi;
sbi->magic = AUTOFS_SBI_MAGIC;
sbi->pipefd = -1;
- sbi->catatonic = 0;
+ sbi->pipe = NULL;
+ sbi->catatonic = 1;
sbi->exp_timeout = 0;
sbi->oz_pgrp = process_group(current);
sbi->sb = s;
@@ -388,6 +400,7 @@ int autofs4_fill_super(struct super_bloc
goto fail_fput;
sbi->pipe = pipe;
sbi->pipefd = pipefd;
+ sbi->catatonic = 0;
/*
* Success! Install the root dentry now to indicate completion.
@@ -412,6 +425,8 @@ fail_ino:
kfree(ino);
fail_free:
kfree(sbi);
+ s->s_fs_info = NULL;
+ kill_anon_super(s);
fail_unlock:
return -EINVAL;
}
--- linux-2.6.19-rc4/fs/autofs/waitq.c.mount-fail-panic 2006-09-20 11:42:06.000000000 +0800
+++ linux-2.6.19-rc4/fs/autofs/waitq.c 2006-11-10 19:49:57.000000000 +0800
@@ -41,6 +41,7 @@ void autofs_catatonic_mode(struct autofs
wq = nwq;
}
fput(sbi->pipe); /* Close the pipe */
+ sbi->pipe = NULL;
autofs_hash_dputall(&sbi->dirhash); /* Remove all dentry pointers */
}
--- linux-2.6.19-rc4/fs/autofs/inode.c.mount-fail-panic 2006-11-10 15:34:42.000000000 +0800
+++ linux-2.6.19-rc4/fs/autofs/inode.c 2006-11-12 17:07:58.000000000 +0800
@@ -25,6 +25,14 @@ void autofs_kill_sb(struct super_block *
struct autofs_sb_info *sbi = autofs_sbi(sb);
unsigned int n;
+ /*
+ * In the event of a failure in get_sb_nodev the superblock
+ * info is not present so nothing else has been setup, so
+ * just exit when we are called from deactivate_super.
+ */
+ if (!sbi)
+ return;
+
if ( !sbi->catatonic )
autofs_catatonic_mode(sbi); /* Free wait queues, close pipe */
@@ -136,7 +144,8 @@ int autofs_fill_super(struct super_block
s->s_fs_info = sbi;
sbi->magic = AUTOFS_SBI_MAGIC;
- sbi->catatonic = 0;
+ sbi->pipe = NULL;
+ sbi->catatonic = 1;
sbi->exp_timeout = 0;
sbi->oz_pgrp = process_group(current);
autofs_initialize_hash(&sbi->dirhash);
@@ -180,6 +189,7 @@ int autofs_fill_super(struct super_block
if ( !pipe->f_op || !pipe->f_op->write )
goto fail_fput;
sbi->pipe = pipe;
+ sbi->catatonic = 0;
/*
* Success! Install the root dentry now to indicate completion.
@@ -198,6 +208,8 @@ fail_iput:
iput(root_inode);
fail_free:
kfree(sbi);
+ s->s_fs_info = NULL;
+ kill_anon_super(s);
fail_unlock:
return -EINVAL;
}
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] autofs4 - panic after mount fail
2006-11-12 9:48 [PATCH] autofs4 - panic after mount fail Ian Kent
@ 2006-11-13 11:07 ` David Howells
2006-11-13 15:51 ` Ian Kent
0 siblings, 1 reply; 5+ messages in thread
From: David Howells @ 2006-11-13 11:07 UTC (permalink / raw)
To: Ian Kent; +Cc: Andrew Morton, bibo,mao, David Howells, Kernel Mailing List
Ian Kent <raven@themaw.net> wrote:
> - if (sbi->pipe) {
> - fput(sbi->pipe); /* Close the pipe */
> - sbi->pipe = NULL;
> - }
> + fput(sbi->pipe); /* Close the pipe */
> + sbi->pipe = NULL;
Ummm... Is that right? fput() doesn't check its argument for a NULL pointer,
so the original code shouldn't hurt and should give you an extra bit of
defense.
Other than that, it looks reasonable.
David
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] autofs4 - panic after mount fail
2006-11-13 11:07 ` David Howells
@ 2006-11-13 15:51 ` Ian Kent
2006-11-13 15:55 ` Ian Kent
0 siblings, 1 reply; 5+ messages in thread
From: Ian Kent @ 2006-11-13 15:51 UTC (permalink / raw)
To: David Howells; +Cc: Andrew Morton, bibo,mao, Kernel Mailing List
On Mon, 2006-11-13 at 11:07 +0000, David Howells wrote:
> Ian Kent <raven@themaw.net> wrote:
>
> > - if (sbi->pipe) {
> > - fput(sbi->pipe); /* Close the pipe */
> > - sbi->pipe = NULL;
> > - }
> > + fput(sbi->pipe); /* Close the pipe */
> > + sbi->pipe = NULL;
>
> Ummm... Is that right? fput() doesn't check its argument for a NULL pointer,
> so the original code shouldn't hurt and should give you an extra bit of
> defense.
HaHa .. sometimes it's people saying "get id of that you don't need it"
and other times it's the opposite.
Setting sbi->catatonic = 1 upon entry to autofs[4]_fill_super ensures
that autofs[4]_catatonic_mode is not called until after sbi->pipe is non
NULL. So I decided, in this case, to remove the test.
Ian
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] autofs4 - panic after mount fail
2006-11-13 15:51 ` Ian Kent
@ 2006-11-13 15:55 ` Ian Kent
2006-11-14 9:08 ` David Howells
0 siblings, 1 reply; 5+ messages in thread
From: Ian Kent @ 2006-11-13 15:55 UTC (permalink / raw)
To: David Howells; +Cc: Andrew Morton, bibo,mao, Kernel Mailing List
On Mon, 2006-11-13 at 23:51 +0800, Ian Kent wrote:
> On Mon, 2006-11-13 at 11:07 +0000, David Howells wrote:
> > Ian Kent <raven@themaw.net> wrote:
> >
> > > - if (sbi->pipe) {
> > > - fput(sbi->pipe); /* Close the pipe */
> > > - sbi->pipe = NULL;
> > > - }
> > > + fput(sbi->pipe); /* Close the pipe */
> > > + sbi->pipe = NULL;
> >
> > Ummm... Is that right? fput() doesn't check its argument for a NULL pointer,
> > so the original code shouldn't hurt and should give you an extra bit of
> > defense.
>
> HaHa .. sometimes it's people saying "get id of that you don't need it"
> and other times it's the opposite.
>
> Setting sbi->catatonic = 1 upon entry to autofs[4]_fill_super ensures
> that autofs[4]_catatonic_mode is not called until after sbi->pipe is non
> NULL. So I decided, in this case, to remove the test.
However, if you feel strongly about it I'll change it.
Ian
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] autofs4 - panic after mount fail
2006-11-13 15:55 ` Ian Kent
@ 2006-11-14 9:08 ` David Howells
0 siblings, 0 replies; 5+ messages in thread
From: David Howells @ 2006-11-14 9:08 UTC (permalink / raw)
To: Ian Kent; +Cc: David Howells, Andrew Morton, bibo,mao, Kernel Mailing List
Ian Kent <raven@themaw.net> wrote:
> However, if you feel strongly about it I'll change it.
Not really. I was just checking.
David
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2006-11-14 9:10 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-11-12 9:48 [PATCH] autofs4 - panic after mount fail Ian Kent
2006-11-13 11:07 ` David Howells
2006-11-13 15:51 ` Ian Kent
2006-11-13 15:55 ` Ian Kent
2006-11-14 9:08 ` David Howells
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox