[PATCH] fs: ioctl_fsthaw should use freeze_bdev if possible

[PATCH] fs: ioctl_fsthaw should use freeze_bdev if possible

[Dmitry Monakhov]

Commit: 18e9e5104fcd  breaks connection between freeze_bdev()/thaw_bdev() and
freeze_super()/thaw_super()
In other workds:
freeze_bdev()/thaw_bdev() may be called multiple times, but
freeze_super()/thaw_super() only once (will resunt -EBUSY on second call)

BUT user is still allowed to mix both interfaces:

freeze_bdev () call it first time
 ->bdev->bd_fsfreeze_count++
 ->freeze_super()
   return 0
freeze_super() second time
 ->bdev->bd_fsfreeze_count++
   return 0

One can call ioctl_fsthaw( #xfs_freeze -u $MNT)
ioctl_fsthaw
 ->thaw_super() -> fs is unfrozen now
   return 0

This leave bdev in broken state because:
bd_fsfreeze_count > 0 but superblock is not frozen.
Visiable effect:
0) Any freeze_bdev() will not call freeze_super() because
   (bdev->bd_fsfreeze_count > 0)
1) thaw_bdev() will fail because thaw_super() will return EINVAL (filesystem
   is not frozen) and bdev->bd_fsfreeze_count will not being decremented.
2) filesystem may not being mounted because mount_bdev will fail
   because (bdev->bd_fsfreeze_count > 0)

#TEST_CASE:
# Use xfstests 068'th test
(while true ;do dmsetup suspend $SCRATCH_DEV ;dmsetup resume $SCRATCH_DEV;done)&
pid=$!
./check 068 068 068 068 068
kill $pid
wait $pid

In order to fix that let's always use {freeze,thaw}_bdev() functions if
super block has bdev, and directly call {freeze,thaw)_super() otherwise.

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
---
 fs/ioctl.c |   16 +++++++++++++---
 1 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/fs/ioctl.c b/fs/ioctl.c
index 3bdad6d..ea47ccc 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -513,6 +513,7 @@ static int ioctl_fioasync(unsigned int fd, struct file *filp,
 static int ioctl_fsfreeze(struct file *filp)
 {
 	struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
+	int ret = 0;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
@@ -521,8 +522,15 @@ static int ioctl_fsfreeze(struct file *filp)
 	if (sb->s_op->freeze_fs == NULL)
 		return -EOPNOTSUPP;
 
-	/* Freeze */
-	return freeze_super(sb);
+	if (sb->s_bdev) {
+		struct super_block *fsb;
+		fsb = freeze_bdev(sb->s_bdev);
+		if (IS_ERR(fsb))
+			ret = PTR_ERR(fsb);
+	} else {
+		ret = freeze_super(sb);
+	}
+	return ret;
 }
 
 static int ioctl_fsthaw(struct file *filp)
@@ -532,7 +540,9 @@ static int ioctl_fsthaw(struct file *filp)
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	/* Thaw */
+	if (sb->s_bdev)
+		return thaw_bdev(sb->s_bdev, sb);
+
 	return thaw_super(sb);
 }
 
-- 
1.7.1

Re: [PATCH] fs: ioctl_fsthaw should use freeze_bdev if PING.

[Dmitry Monakhov]

On Thu, 21 Mar 2013 16:26:56 +0400, Dmitry Monakhov <dmonakhov@openvz.org> wrote:
Ping... Please take a look at this fix.
> Commit: 18e9e5104fcd  breaks connection between freeze_bdev()/thaw_bdev() and
> freeze_super()/thaw_super()
> In other workds:
> freeze_bdev()/thaw_bdev() may be called multiple times, but
> freeze_super()/thaw_super() only once (will resunt -EBUSY on second call)
> 
> BUT user is still allowed to mix both interfaces:
> 
> freeze_bdev () call it first time
>  ->bdev->bd_fsfreeze_count++
>  ->freeze_super()
>    return 0
> freeze_super() second time
>  ->bdev->bd_fsfreeze_count++
>    return 0
> 
> One can call ioctl_fsthaw( #xfs_freeze -u $MNT)
> ioctl_fsthaw
>  ->thaw_super() -> fs is unfrozen now
>    return 0
> 
> This leave bdev in broken state because:
> bd_fsfreeze_count > 0 but superblock is not frozen.
> Visiable effect:
> 0) Any freeze_bdev() will not call freeze_super() because
>    (bdev->bd_fsfreeze_count > 0)
> 1) thaw_bdev() will fail because thaw_super() will return EINVAL (filesystem
>    is not frozen) and bdev->bd_fsfreeze_count will not being decremented.
> 2) filesystem may not being mounted because mount_bdev will fail
>    because (bdev->bd_fsfreeze_count > 0)
> 
> #TEST_CASE:
> # Use xfstests 068'th test
> (while true ;do dmsetup suspend $SCRATCH_DEV ;dmsetup resume $SCRATCH_DEV;done)&
> pid=$!
> ./check 068 068 068 068 068
> kill $pid
> wait $pid
> 
> In order to fix that let's always use {freeze,thaw}_bdev() functions if
> super block has bdev, and directly call {freeze,thaw)_super() otherwise.
> 
> Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
> ---
>  fs/ioctl.c |   16 +++++++++++++---
>  1 files changed, 13 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/ioctl.c b/fs/ioctl.c
> index 3bdad6d..ea47ccc 100644
> --- a/fs/ioctl.c
> +++ b/fs/ioctl.c
> @@ -513,6 +513,7 @@ static int ioctl_fioasync(unsigned int fd, struct file *filp,
>  static int ioctl_fsfreeze(struct file *filp)
>  {
>  	struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
> +	int ret = 0;
>  
>  	if (!capable(CAP_SYS_ADMIN))
>  		return -EPERM;
> @@ -521,8 +522,15 @@ static int ioctl_fsfreeze(struct file *filp)
>  	if (sb->s_op->freeze_fs == NULL)
>  		return -EOPNOTSUPP;
>  
> -	/* Freeze */
> -	return freeze_super(sb);
> +	if (sb->s_bdev) {
> +		struct super_block *fsb;
> +		fsb = freeze_bdev(sb->s_bdev);
> +		if (IS_ERR(fsb))
> +			ret = PTR_ERR(fsb);
> +	} else {
> +		ret = freeze_super(sb);
> +	}
> +	return ret;
>  }
>  
>  static int ioctl_fsthaw(struct file *filp)
> @@ -532,7 +540,9 @@ static int ioctl_fsthaw(struct file *filp)
>  	if (!capable(CAP_SYS_ADMIN))
>  		return -EPERM;
>  
> -	/* Thaw */
> +	if (sb->s_bdev)
> +		return thaw_bdev(sb->s_bdev, sb);
> +
>  	return thaw_super(sb);
>  }
>  
> -- 
> 1.7.1
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] fs: ioctl_fsthaw should use freeze_bdev if PING.

[Jan Kara]

On Wed 10-04-13 21:57:37, Dmitry Monakhov wrote:
> On Thu, 21 Mar 2013 16:26:56 +0400, Dmitry Monakhov <dmonakhov@openvz.org> wrote:
> Ping... Please take a look at this fix.
  Actually, there are much more issues in this area. Fernando fixed most of
them in his patch set (last posting is here
  http://www.spinics.net/lists/linux-fsdevel/msg61266.html
I think) but it never got merged...

								Honza

> > Commit: 18e9e5104fcd  breaks connection between freeze_bdev()/thaw_bdev() and
> > freeze_super()/thaw_super()
> > In other workds:
> > freeze_bdev()/thaw_bdev() may be called multiple times, but
> > freeze_super()/thaw_super() only once (will resunt -EBUSY on second call)
> > 
> > BUT user is still allowed to mix both interfaces:
> > 
> > freeze_bdev () call it first time
> >  ->bdev->bd_fsfreeze_count++
> >  ->freeze_super()
> >    return 0
> > freeze_super() second time
> >  ->bdev->bd_fsfreeze_count++
> >    return 0
> > 
> > One can call ioctl_fsthaw( #xfs_freeze -u $MNT)
> > ioctl_fsthaw
> >  ->thaw_super() -> fs is unfrozen now
> >    return 0
> > 
> > This leave bdev in broken state because:
> > bd_fsfreeze_count > 0 but superblock is not frozen.
> > Visiable effect:
> > 0) Any freeze_bdev() will not call freeze_super() because
> >    (bdev->bd_fsfreeze_count > 0)
> > 1) thaw_bdev() will fail because thaw_super() will return EINVAL (filesystem
> >    is not frozen) and bdev->bd_fsfreeze_count will not being decremented.
> > 2) filesystem may not being mounted because mount_bdev will fail
> >    because (bdev->bd_fsfreeze_count > 0)
> > 
> > #TEST_CASE:
> > # Use xfstests 068'th test
> > (while true ;do dmsetup suspend $SCRATCH_DEV ;dmsetup resume $SCRATCH_DEV;done)&
> > pid=$!
> > ./check 068 068 068 068 068
> > kill $pid
> > wait $pid
> > 
> > In order to fix that let's always use {freeze,thaw}_bdev() functions if
> > super block has bdev, and directly call {freeze,thaw)_super() otherwise.
> > 
> > Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
> > ---
> >  fs/ioctl.c |   16 +++++++++++++---
> >  1 files changed, 13 insertions(+), 3 deletions(-)
> > 
> > diff --git a/fs/ioctl.c b/fs/ioctl.c
> > index 3bdad6d..ea47ccc 100644
> > --- a/fs/ioctl.c
> > +++ b/fs/ioctl.c
> > @@ -513,6 +513,7 @@ static int ioctl_fioasync(unsigned int fd, struct file *filp,
> >  static int ioctl_fsfreeze(struct file *filp)
> >  {
> >  	struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
> > +	int ret = 0;
> >  
> >  	if (!capable(CAP_SYS_ADMIN))
> >  		return -EPERM;
> > @@ -521,8 +522,15 @@ static int ioctl_fsfreeze(struct file *filp)
> >  	if (sb->s_op->freeze_fs == NULL)
> >  		return -EOPNOTSUPP;
> >  
> > -	/* Freeze */
> > -	return freeze_super(sb);
> > +	if (sb->s_bdev) {
> > +		struct super_block *fsb;
> > +		fsb = freeze_bdev(sb->s_bdev);
> > +		if (IS_ERR(fsb))
> > +			ret = PTR_ERR(fsb);
> > +	} else {
> > +		ret = freeze_super(sb);
> > +	}
> > +	return ret;
> >  }
> >  
> >  static int ioctl_fsthaw(struct file *filp)
> > @@ -532,7 +540,9 @@ static int ioctl_fsthaw(struct file *filp)
> >  	if (!capable(CAP_SYS_ADMIN))
> >  		return -EPERM;
> >  
> > -	/* Thaw */
> > +	if (sb->s_bdev)
> > +		return thaw_bdev(sb->s_bdev, sb);
> > +
> >  	return thaw_super(sb);
> >  }
> >  
> > -- 
> > 1.7.1
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR