Re: NILFS crash on 2.6.37

Re: NILFS crash on 2.6.37

[Ryusuke Konishi]

(Cc'ed to linux-nilfs)
On Thu, 20 Jan 2011 21:49:18 -0500, Wakko Warner wrote:
> At one time I had created a filesystem with nilfs on it.  I don't know if
> it's nilfs or nilfs2.  Attempting to mount it with nilfs2 on 2.6.37 (also
> happens on 2.6.36.2), the kernel oopses.  Once the kernel oopses, any
> further attempts to mount it hangs the process.  I have also noticed that
> attempting to umount other filesystems also hangs the process.
> 
> [ 1202.579525] NILFS version 2 loaded
> [ 1202.581501] NILFS warning: broken superblock. using spare superblock.
> [ 1202.582093] NILFS warning: broken superblock. using spare superblock.
> [ 1202.821896] segctord starting. Construction interval = 5 seconds, CP frequency < 30 seconds
> [ 1202.824419] BUG: unable to handle kernel NULL pointer dereference at   (null)
> [ 1202.824561] IP: [<c03196bc>] memcpy+0xc/0x1b
> [ 1202.824667] *pdpt = 0000000033b94001 *pde = 0000000000000000 
> [ 1202.824797] Oops: 0002 [#1] SMP 
<snip>
> [ 1202.825136] Call Trace:
> [ 1202.825136]  [<f953720e>] ? nilfs_setup_super+0x6c/0xa5 [nilfs2]
> [ 1202.825136]  [<f95369e9>] ? nilfs_get_root_dentry+0x81/0xcb [nilfs2]
> [ 1202.825136]  [<f9537a08>] ? nilfs_mount+0x4f9/0x62c [nilfs2]
> [ 1202.825136]  [<c02745cf>] ? kstrdup+0x36/0x3f
> [ 1202.825136]  [<f953750f>] ? nilfs_mount+0x0/0x62c [nilfs2]
> [ 1202.825136]  [<c0293940>] ? vfs_kern_mount+0x4d/0x12c
> [ 1202.825136]  [<c02a5100>] ? get_fs_type+0x76/0x8f
> [ 1202.825136]  [<c0293a68>] ? do_kern_mount+0x33/0xbf
> [ 1202.825136]  [<c02a784a>] ? do_mount+0x2ed/0x714
> [ 1202.825136]  [<c02a6171>] ? copy_mount_options+0x28/0xfc
> [ 1202.825136]  [<c02a7ce3>] ? sys_mount+0x72/0xaf
> [ 1202.825136]  [<c0473085>] ? syscall_call+0x7/0xb
> [ 1202.825136] Code: 43 4c 04 83 fe 01 19 c0 24 fc 04 08 88 43 4d 5b 5e 5f c3 90 90 90 90 90 90 90 90 90 90 90 90 57 56 53 89 cb c1 e9 02 89 c7 89 d6 <f3> a5 89 d9 83 e1 03 74 02 f3 a4 5b 5e 5f c3 55 57 56 53 89 c5 
> [ 1202.825136] EIP: [<c03196bc>] memcpy+0xc/0x1b SS:ESP 0068:f3b35e74
> [ 1202.825136] CR2: 0000000000000000
> [ 1202.831111] ---[ end trace e5a2d0918d598efa ]---

Thanks for the report.

This seems to be a regression on 2.6.36 which could happen when one of
two superblocks is broken.

I'll fix it.

Ryusuke Konishi
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: NILFS crash on 2.6.37

[Ryusuke Konishi]

On Fri, 21 Jan 2011 16:37:38 +0900 (JST), Ryusuke Konishi wrote:
> (Cc'ed to linux-nilfs)
> On Thu, 20 Jan 2011 21:49:18 -0500, Wakko Warner wrote:
> > At one time I had created a filesystem with nilfs on it.  I don't know if
> > it's nilfs or nilfs2.  Attempting to mount it with nilfs2 on 2.6.37 (also
> > happens on 2.6.36.2), the kernel oopses.  Once the kernel oopses, any
> > further attempts to mount it hangs the process.  I have also noticed that
> > attempting to umount other filesystems also hangs the process.
> > 
> > [ 1202.579525] NILFS version 2 loaded
> > [ 1202.581501] NILFS warning: broken superblock. using spare superblock.
> > [ 1202.582093] NILFS warning: broken superblock. using spare superblock.
> > [ 1202.821896] segctord starting. Construction interval = 5 seconds, CP frequency < 30 seconds
> > [ 1202.824419] BUG: unable to handle kernel NULL pointer dereference at   (null)
> > [ 1202.824561] IP: [<c03196bc>] memcpy+0xc/0x1b
> > [ 1202.824667] *pdpt = 0000000033b94001 *pde = 0000000000000000 
> > [ 1202.824797] Oops: 0002 [#1] SMP 
> <snip>
> > [ 1202.825136] Call Trace:
> > [ 1202.825136]  [<f953720e>] ? nilfs_setup_super+0x6c/0xa5 [nilfs2]
> > [ 1202.825136]  [<f95369e9>] ? nilfs_get_root_dentry+0x81/0xcb [nilfs2]
> > [ 1202.825136]  [<f9537a08>] ? nilfs_mount+0x4f9/0x62c [nilfs2]
> > [ 1202.825136]  [<c02745cf>] ? kstrdup+0x36/0x3f
> > [ 1202.825136]  [<f953750f>] ? nilfs_mount+0x0/0x62c [nilfs2]
> > [ 1202.825136]  [<c0293940>] ? vfs_kern_mount+0x4d/0x12c
> > [ 1202.825136]  [<c02a5100>] ? get_fs_type+0x76/0x8f
> > [ 1202.825136]  [<c0293a68>] ? do_kern_mount+0x33/0xbf
> > [ 1202.825136]  [<c02a784a>] ? do_mount+0x2ed/0x714
> > [ 1202.825136]  [<c02a6171>] ? copy_mount_options+0x28/0xfc
> > [ 1202.825136]  [<c02a7ce3>] ? sys_mount+0x72/0xaf
> > [ 1202.825136]  [<c0473085>] ? syscall_call+0x7/0xb
> > [ 1202.825136] Code: 43 4c 04 83 fe 01 19 c0 24 fc 04 08 88 43 4d 5b 5e 5f c3 90 90 90 90 90 90 90 90 90 90 90 90 57 56 53 89 cb c1 e9 02 89 c7 89 d6 <f3> a5 89 d9 83 e1 03 74 02 f3 a4 5b 5e 5f c3 55 57 56 53 89 c5 
> > [ 1202.825136] EIP: [<c03196bc>] memcpy+0xc/0x1b SS:ESP 0068:f3b35e74
> > [ 1202.825136] CR2: 0000000000000000
> > [ 1202.831111] ---[ end trace e5a2d0918d598efa ]---
> 
> Thanks for the report.
> 
> This seems to be a regression on 2.6.36 which could happen when one of
> two superblocks is broken.
> 
> I'll fix it.
> 
> Ryusuke Konishi

Does the following patch fix your problem?

Ryusuke Konishi
---
diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
index 0994f6a7..58fd707 100644
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@@ -704,7 +704,8 @@ skip_mount_setup:
 	sbp[0]->s_state =
 		cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS);
 	/* synchronize sbp[1] with sbp[0] */
-	memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
+	if (sbp[1])
+		memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
 	return nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL);
 }
 
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: NILFS crash on 2.6.37

[Wakko Warner]

Ryusuke Konishi wrote:
> On Fri, 21 Jan 2011 16:37:38 +0900 (JST), Ryusuke Konishi wrote:
> > (Cc'ed to linux-nilfs)
> > > [ 1202.581501] NILFS warning: broken superblock. using spare superblock.
> > > [ 1202.582093] NILFS warning: broken superblock. using spare superblock.

Why does it say this 2 times?  Just curious.

> > Thanks for the report.
> > 
> > This seems to be a regression on 2.6.36 which could happen when one of
> > two superblocks is broken.
> > 
> > I'll fix it.
> 
> Does the following patch fix your problem?

> diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
> index 0994f6a7..58fd707 100644
> --- a/fs/nilfs2/super.c
> +++ b/fs/nilfs2/super.c
> @@ -704,7 +704,8 @@ skip_mount_setup:
>  	sbp[0]->s_state =
>  		cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS);
>  	/* synchronize sbp[1] with sbp[0] */
> -	memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
> +	if (sbp[1])
> +		memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
>  	return nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL);
>  }

That patch fix it.  Just a note, the 2.6.37 vanilla source was about 20
lines up from 704

-- 
 Microsoft has beaten Volkswagen's world record.  Volkswagen only created 22
 million bugs.

Re: NILFS crash on 2.6.37

[Ryusuke Konishi]

On Fri, 21 Jan 2011 20:57:48 -0500, Wakko Warner wrote:
> Ryusuke Konishi wrote:
> > On Fri, 21 Jan 2011 16:37:38 +0900 (JST), Ryusuke Konishi wrote:
> > > (Cc'ed to linux-nilfs)
> > Does the following patch fix your problem?
> 
> > diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
> > index 0994f6a7..58fd707 100644
> > --- a/fs/nilfs2/super.c
> > +++ b/fs/nilfs2/super.c
> > @@ -704,7 +704,8 @@ skip_mount_setup:
> >  	sbp[0]->s_state =
> >  		cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS);
> >  	/* synchronize sbp[1] with sbp[0] */
> > -	memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
> > +	if (sbp[1])
> > +		memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
> >  	return nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL);
> >  }
> 
> That patch fix it.  Just a note, the 2.6.37 vanilla source was about 20
> lines up from 704

Thanks for testing.  Will send upstream and to stable.


> > > > [ 1202.581501] NILFS warning: broken superblock. using spare superblock.
> > > > [ 1202.582093] NILFS warning: broken superblock. using spare superblock.
> 
> Why does it say this 2 times?  Just curious.

Because each super block is read twice in different block sizes.

Well, indeed it looks confusing.  I'll try to do something.


Ryusuke Konishi
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: NILFS crash on 2.6.37

[Wakko Warner]

Ryusuke Konishi wrote:
> On Fri, 21 Jan 2011 20:57:48 -0500, Wakko Warner wrote:
> > That patch fix it.  Just a note, the 2.6.37 vanilla source was about 20
> > lines up from 704
> 
> Thanks for testing.  Will send upstream and to stable.

No problem.

> > > > > [ 1202.581501] NILFS warning: broken superblock. using spare superblock.
> > > > > [ 1202.582093] NILFS warning: broken superblock. using spare superblock.
> > 
> > Why does it say this 2 times?  Just curious.
> 
> Because each super block is read twice in different block sizes.
> 
> Well, indeed it looks confusing.  I'll try to do something.

Ah, I see.

-- 
 Microsoft has beaten Volkswagen's world record.  Volkswagen only created 22
 million bugs.
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html