2008-06-15 14:53:22

by Aneesh Kumar K.V

[permalink] [raw]
Subject: [PATCH] ext4: Fix file system hang due to corrupt jinode values.

While commiting the transaction we set the i_flags
of jbd2_inode to JI_COMMIT_RUNNING. This is needed to ensure
that when we are commiting the transaction we don't release
the inode via iput. In ext4_clear_inode we wait for the
flag to be cleared before we release the inode.

In ext4_new_inode we call iput for different failure cases
before initializing the jbd2_inode. This cause the failure
case to hang in iput as show below. Fix the same by initializing
jbd2_inode early.

[<c022d20e>] ? prepare_to_wait+0x49/0x4e
[<c02cf430>] jbd2_journal_release_jbd_inode+0x86/0xcc
[<c022d0e4>] ? wake_bit_function+0x0/0x43
[<c02b329b>] ext4_clear_inode+0x45/0x4a
[<c02748f7>] clear_inode+0x5f/0xae
[<c0274e8f>] generic_drop_inode+0x119/0x13a
[<c0274071>] iput+0x4b/0x4e
[<c02ab200>] ext4_new_inode+0xb29/0xb7a
[<c0272761>] ? dput+0x19/0xdc
[<c02cca5f>] ? jbd2_journal_start+0x8c/0xb9
[<c02b1cb4>] ext4_create+0x81/0xe7
[<c02b1c33>] ? ext4_create+0x0/0xe7
[<c026aa3c>] vfs_create+0x6c/0xdb
[<c026cfd6>] do_filp_open+0x1a6/0x6b5
[<c026284e>] do_sys_open+0x40/0xb6
[<c0262906>] sys_open+0x1e/0x26
[<c0202be6>] syscall_call+0x7/0xb

Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
fs/ext4/ialloc.c | 1 -
fs/ext4/super.c | 1 +
2 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 7823405..dc8bfc4 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -820,7 +820,6 @@ struct inode *ext4_new_inode(handle_t *handle, struct inode * dir, int mode)
ei->i_state = EXT4_STATE_NEW;

ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize;
- jbd2_journal_init_jbd_inode(&ei->jinode, inode);

ret = inode;
if(DQUOT_ALLOC_INODE(inode)) {
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index e207555..fde1ae9 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -578,6 +578,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
ei->i_allocated_meta_blocks = 0;
ei->i_delalloc_reserved_flag = 0;
spin_lock_init(&(ei->i_block_reservation_lock));
+ jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
return &ei->vfs_inode;
}

--
1.5.6.rc2.15.g457bb.dirty



2008-06-16 13:45:04

by Jan Kara

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix file system hang due to corrupt jinode values.

> While commiting the transaction we set the i_flags
> of jbd2_inode to JI_COMMIT_RUNNING. This is needed to ensure
> that when we are commiting the transaction we don't release
> the inode via iput. In ext4_clear_inode we wait for the
> flag to be cleared before we release the inode.
>
> In ext4_new_inode we call iput for different failure cases
> before initializing the jbd2_inode. This cause the failure
> case to hang in iput as show below. Fix the same by initializing
> jbd2_inode early.
>
> [<c022d20e>] ? prepare_to_wait+0x49/0x4e
> [<c02cf430>] jbd2_journal_release_jbd_inode+0x86/0xcc
> [<c022d0e4>] ? wake_bit_function+0x0/0x43
> [<c02b329b>] ext4_clear_inode+0x45/0x4a
> [<c02748f7>] clear_inode+0x5f/0xae
> [<c0274e8f>] generic_drop_inode+0x119/0x13a
> [<c0274071>] iput+0x4b/0x4e
> [<c02ab200>] ext4_new_inode+0xb29/0xb7a
> [<c0272761>] ? dput+0x19/0xdc
> [<c02cca5f>] ? jbd2_journal_start+0x8c/0xb9
> [<c02b1cb4>] ext4_create+0x81/0xe7
> [<c02b1c33>] ? ext4_create+0x0/0xe7
> [<c026aa3c>] vfs_create+0x6c/0xdb
> [<c026cfd6>] do_filp_open+0x1a6/0x6b5
> [<c026284e>] do_sys_open+0x40/0xb6
> [<c0262906>] sys_open+0x1e/0x26
> [<c0202be6>] syscall_call+0x7/0xb
>
> Signed-off-by: Aneesh Kumar K.V <[email protected]>
Oh yes, thanks for fixing this. I think there is place in
super.c which also calls jbd2_journal_init_jbd_inode() and which doesn't
need to call it when you do so already in alloc_inode().

Honza
> ---
> fs/ext4/ialloc.c | 1 -
> fs/ext4/super.c | 1 +
> 2 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
> index 7823405..dc8bfc4 100644
> --- a/fs/ext4/ialloc.c
> +++ b/fs/ext4/ialloc.c
> @@ -820,7 +820,6 @@ struct inode *ext4_new_inode(handle_t *handle, struct inode * dir, int mode)
> ei->i_state = EXT4_STATE_NEW;
>
> ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize;
> - jbd2_journal_init_jbd_inode(&ei->jinode, inode);
>
> ret = inode;
> if(DQUOT_ALLOC_INODE(inode)) {
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index e207555..fde1ae9 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -578,6 +578,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
> ei->i_allocated_meta_blocks = 0;
> ei->i_delalloc_reserved_flag = 0;
> spin_lock_init(&(ei->i_block_reservation_lock));
> + jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
> return &ei->vfs_inode;
> }
>
> --
> 1.5.6.rc2.15.g457bb.dirty
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
Jan Kara <[email protected]>
SuSE CR Labs

2008-06-16 16:30:43

by Aneesh Kumar K.V

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix file system hang due to corrupt jinode values.

On Mon, Jun 16, 2008 at 03:45:03PM +0200, Jan Kara wrote:
> > While commiting the transaction we set the i_flags
> > of jbd2_inode to JI_COMMIT_RUNNING. This is needed to ensure
> > that when we are commiting the transaction we don't release
> > the inode via iput. In ext4_clear_inode we wait for the
> > flag to be cleared before we release the inode.
> >
> > In ext4_new_inode we call iput for different failure cases
> > before initializing the jbd2_inode. This cause the failure
> > case to hang in iput as show below. Fix the same by initializing
> > jbd2_inode early.
> >
> > [<c022d20e>] ? prepare_to_wait+0x49/0x4e
> > [<c02cf430>] jbd2_journal_release_jbd_inode+0x86/0xcc
> > [<c022d0e4>] ? wake_bit_function+0x0/0x43
> > [<c02b329b>] ext4_clear_inode+0x45/0x4a
> > [<c02748f7>] clear_inode+0x5f/0xae
> > [<c0274e8f>] generic_drop_inode+0x119/0x13a
> > [<c0274071>] iput+0x4b/0x4e
> > [<c02ab200>] ext4_new_inode+0xb29/0xb7a
> > [<c0272761>] ? dput+0x19/0xdc
> > [<c02cca5f>] ? jbd2_journal_start+0x8c/0xb9
> > [<c02b1cb4>] ext4_create+0x81/0xe7
> > [<c02b1c33>] ? ext4_create+0x0/0xe7
> > [<c026aa3c>] vfs_create+0x6c/0xdb
> > [<c026cfd6>] do_filp_open+0x1a6/0x6b5
> > [<c026284e>] do_sys_open+0x40/0xb6
> > [<c0262906>] sys_open+0x1e/0x26
> > [<c0202be6>] syscall_call+0x7/0xb
> >
> > Signed-off-by: Aneesh Kumar K.V <[email protected]>
> Oh yes, thanks for fixing this. I think there is place in
> super.c which also calls jbd2_journal_init_jbd_inode() and which doesn't
> need to call it when you do so already in alloc_inode().
>

Like this ?

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index fef2574..43d67e0 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3125,7 +3125,6 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
ei->i_default_acl = EXT4_ACL_NOT_CACHED;
#endif
ei->i_block_alloc_info = NULL;
- jbd2_journal_init_jbd_inode(&ei->jinode, inode);

ret = __ext4_get_inode_loc(inode, &iloc, 0);
if (ret < 0)
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 034f00f..e511b95 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -2259,8 +2259,6 @@ static int ext4_mb_init_backend(struct super_block *sb)
printk(KERN_ERR "EXT4-fs: can't get new inode\n");
goto err_freesgi;
}
- jbd2_journal_init_jbd_inode(&EXT4_I(sbi->s_buddy_cache)->jinode,
- sbi->s_buddy_cache);
EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;

metalen = sizeof(*meta_group_info) << EXT4_DESC_PER_BLOCK_BITS(sb);

2008-06-16 17:27:06

by Jan Kara

[permalink] [raw]
Subject: Re: [PATCH] ext4: Fix file system hang due to corrupt jinode values.

> On Mon, Jun 16, 2008 at 03:45:03PM +0200, Jan Kara wrote:
> > > While commiting the transaction we set the i_flags
> > > of jbd2_inode to JI_COMMIT_RUNNING. This is needed to ensure
> > > that when we are commiting the transaction we don't release
> > > the inode via iput. In ext4_clear_inode we wait for the
> > > flag to be cleared before we release the inode.
> > >
> > > In ext4_new_inode we call iput for different failure cases
> > > before initializing the jbd2_inode. This cause the failure
> > > case to hang in iput as show below. Fix the same by initializing
> > > jbd2_inode early.
> > >
> > > [<c022d20e>] ? prepare_to_wait+0x49/0x4e
> > > [<c02cf430>] jbd2_journal_release_jbd_inode+0x86/0xcc
> > > [<c022d0e4>] ? wake_bit_function+0x0/0x43
> > > [<c02b329b>] ext4_clear_inode+0x45/0x4a
> > > [<c02748f7>] clear_inode+0x5f/0xae
> > > [<c0274e8f>] generic_drop_inode+0x119/0x13a
> > > [<c0274071>] iput+0x4b/0x4e
> > > [<c02ab200>] ext4_new_inode+0xb29/0xb7a
> > > [<c0272761>] ? dput+0x19/0xdc
> > > [<c02cca5f>] ? jbd2_journal_start+0x8c/0xb9
> > > [<c02b1cb4>] ext4_create+0x81/0xe7
> > > [<c02b1c33>] ? ext4_create+0x0/0xe7
> > > [<c026aa3c>] vfs_create+0x6c/0xdb
> > > [<c026cfd6>] do_filp_open+0x1a6/0x6b5
> > > [<c026284e>] do_sys_open+0x40/0xb6
> > > [<c0262906>] sys_open+0x1e/0x26
> > > [<c0202be6>] syscall_call+0x7/0xb
> > >
> > > Signed-off-by: Aneesh Kumar K.V <[email protected]>
> > Oh yes, thanks for fixing this. I think there is place in
> > super.c which also calls jbd2_journal_init_jbd_inode() and which doesn't
> > need to call it when you do so already in alloc_inode().
> >
>
> Like this ?
Yes, so it was mballoc.c and inode.c :). Thanks. You can add
Acked-by: Jan Kara <[email protected]>
to the merged patch (or maybe even fold your fix into the original
patch).

> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index fef2574..43d67e0 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -3125,7 +3125,6 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
> ei->i_default_acl = EXT4_ACL_NOT_CACHED;
> #endif
> ei->i_block_alloc_info = NULL;
> - jbd2_journal_init_jbd_inode(&ei->jinode, inode);
>
> ret = __ext4_get_inode_loc(inode, &iloc, 0);
> if (ret < 0)
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index 034f00f..e511b95 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -2259,8 +2259,6 @@ static int ext4_mb_init_backend(struct super_block *sb)
> printk(KERN_ERR "EXT4-fs: can't get new inode\n");
> goto err_freesgi;
> }
> - jbd2_journal_init_jbd_inode(&EXT4_I(sbi->s_buddy_cache)->jinode,
> - sbi->s_buddy_cache);
> EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;
>
> metalen = sizeof(*meta_group_info) << EXT4_DESC_PER_BLOCK_BITS(sb);
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

Honza
--
Jan Kara <[email protected]>
SuSE CR Labs