[patch v2] fs: umount on symlinlk leaks mnt count

[patch v2] fs: umount on symlinlk leaks mnt count

[Vasily Averin]

[-- Attachment #1: Type: text/plain, Size: 563 bytes --]

v2: mntget(path->mnt) should not be called if mountpoint_last() returns 1.

Currently umount on symlink blocks following umount:

/vz is separate mount

# ls /vz/ -al | grep test
drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir

# umount -l /vz/testlink
umount: /vz/testlink: not mounted (expected)

# lsof /vz
# umount /vz
umount: /vz: device is busy. (unexpected)

In this case mountpoint_last() gets an extra refcount on path->mnt

Signed-off-by:	Vasily Averin <vvs@openvz.org>


[-- Attachment #2: diff-umount-symlink-ms-2 --]
[-- Type: text/plain, Size: 372 bytes --]

diff --git a/fs/namei.c b/fs/namei.c
index 985c6f3..9eb787e 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2256,9 +2256,10 @@ done:
 		goto out;
 	}
 	path->dentry = dentry;
-	path->mnt = mntget(nd->path.mnt);
+	path->mnt = nd->path.mnt;
 	if (should_follow_link(dentry, nd->flags & LOOKUP_FOLLOW))
 		return 1;
+	mntget(path->mnt);
 	follow_mount(path);
 	error = 0;
 out:

Re: [patch v2] fs: umount on symlinlk leaks mnt count

[Richard Weinberger]

On Sun, Jul 20, 2014 at 9:32 AM, Vasily Averin <vvs@parallels.com> wrote:
> v2: mntget(path->mnt) should not be called if mountpoint_last() returns 1.
>
> Currently umount on symlink blocks following umount:
>
> /vz is separate mount
>
> # ls /vz/ -al | grep test
> drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
> lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir
>
> # umount -l /vz/testlink
> umount: /vz/testlink: not mounted (expected)
>
> # lsof /vz
> # umount /vz
> umount: /vz: device is busy. (unexpected)
>
> In this case mountpoint_last() gets an extra refcount on path->mnt
>
> Signed-off-by:  Vasily Averin <vvs@openvz.org>

Sounds reasonable to me.
Can you please send the patch inline and not as attachment?

-- 
Thanks,
//richard

[PATCH v3] fs: umount on symlink leaks mnt count

[Vasily Averin]

v3: patch inline

Currently umount on symlink blocks following umount:

/vz is separate mount

drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir

umount: /vz/testlink: not mounted (expected)

umount: /vz: device is busy. (unexpected)

In this case mountpoint_last() gets an extra refcount on path->mnt

Signed-off-by: Vasily Averin <vvs@openvz.org>
---
 fs/namei.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/fs/namei.c b/fs/namei.c
index 985c6f3..9eb787e 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2256,9 +2256,10 @@ done:
 		goto out;
 	}
 	path->dentry = dentry;
-	path->mnt = mntget(nd->path.mnt);
+	path->mnt = nd->path.mnt;
 	if (should_follow_link(dentry, nd->flags & LOOKUP_FOLLOW))
 		return 1;
+	mntget(path->mnt);
 	follow_mount(path);
 	error = 0;
 out:
-- 
1.7.5.4

[PATCH v4] fs: umount on symlink leaks mnt count

[Vasily Averin]

v4: description corrected
v3: patch inline

Currently umount on symlink blocks following umount:

/vz is separate mount

# ls /vz/ -al | grep test
drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir
# umount -l /vz/testlink
umount: /vz/testlink: not mounted (expected)

# lsof /vz
# umount /vz
umount: /vz: device is busy. (unexpected)

In this case mountpoint_last() gets an extra refcount on path->mnt

Signed-off-by: Vasily Averin <vvs@openvz.org>
---
 fs/namei.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/fs/namei.c b/fs/namei.c
index 985c6f3..9eb787e 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2256,9 +2256,10 @@ done:
 		goto out;
 	}
 	path->dentry = dentry;
-	path->mnt = mntget(nd->path.mnt);
+	path->mnt = nd->path.mnt;
 	if (should_follow_link(dentry, nd->flags & LOOKUP_FOLLOW))
 		return 1;
+	mntget(path->mnt);
 	follow_mount(path);
 	error = 0;
 out:
-- 
1.7.5.4

Re: [PATCH v4] fs: umount on symlink leaks mnt count

[Jeff Layton]

On Mon, 21 Jul 2014 12:30:23 +0400
Vasily Averin <vvs@parallels.com> wrote:

> v4: description corrected
> v3: patch inline
> 
> Currently umount on symlink blocks following umount:
> 
> /vz is separate mount
> 
> # ls /vz/ -al | grep test
> drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
> lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir
> # umount -l /vz/testlink
> umount: /vz/testlink: not mounted (expected)
> 
> # lsof /vz
> # umount /vz
> umount: /vz: device is busy. (unexpected)
> 
> In this case mountpoint_last() gets an extra refcount on path->mnt
> 
> Signed-off-by: Vasily Averin <vvs@openvz.org>
> ---
>  fs/namei.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/namei.c b/fs/namei.c
> index 985c6f3..9eb787e 100644
> --- a/fs/namei.c
> +++ b/fs/namei.c
> @@ -2256,9 +2256,10 @@ done:
>  		goto out;
>  	}
>  	path->dentry = dentry;
> -	path->mnt = mntget(nd->path.mnt);
> +	path->mnt = nd->path.mnt;
>  	if (should_follow_link(dentry, nd->flags & LOOKUP_FOLLOW))
>  		return 1;
> +	mntget(path->mnt);
>  	follow_mount(path);
>  	error = 0;
>  out:

Looks correct, I think...

Acked-by:
-- 
Jeff Layton <jlayton@primarydata.com>

Re: [PATCH v4] fs: umount on symlink leaks mnt count

[Ian Kent]

On Mon, 2014-07-21 at 12:30 +0400, Vasily Averin wrote:
> v4: description corrected
> v3: patch inline
> 
> Currently umount on symlink blocks following umount:
> 
> /vz is separate mount
> 
> # ls /vz/ -al | grep test
> drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
> lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir
> # umount -l /vz/testlink
> umount: /vz/testlink: not mounted (expected)
> 
> # lsof /vz
> # umount /vz
> umount: /vz: device is busy. (unexpected)
> 
> In this case mountpoint_last() gets an extra refcount on path->mnt
> 
> Signed-off-by: Vasily Averin <vvs@openvz.org>

Acked-by: Ian Kent <raven@themaw.net>

I was working on this too.
I prefer this patch over my own.
Haven't tested it yet but it is very similar to the patch I was working
with which had seen some basic testing.

> ---
>  fs/namei.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/namei.c b/fs/namei.c
> index 985c6f3..9eb787e 100644
> --- a/fs/namei.c
> +++ b/fs/namei.c
> @@ -2256,9 +2256,10 @@ done:
>  		goto out;
>  	}
>  	path->dentry = dentry;
> -	path->mnt = mntget(nd->path.mnt);
> +	path->mnt = nd->path.mnt;
>  	if (should_follow_link(dentry, nd->flags & LOOKUP_FOLLOW))
>  		return 1;
> +	mntget(path->mnt);
>  	follow_mount(path);
>  	error = 0;
>  out:

Re: [PATCH v4] fs: umount on symlink leaks mnt count

[Vasily Averin]

CVE-2014-5045 was assigned for this issue

On 07/23/2014 10:06 AM, Ian Kent wrote:
> On Mon, 2014-07-21 at 12:30 +0400, Vasily Averin wrote:
>> Currently umount on symlink blocks following umount:
>>
>> /vz is separate mount
>>
>> # ls /vz/ -al | grep test
>> drwxr-xr-x.  2 root root       4096 Jul 19 01:14 testdir
>> lrwxrwxrwx.  1 root root         11 Jul 19 01:16 testlink -> /vz/testdir
>> # umount -l /vz/testlink
>> umount: /vz/testlink: not mounted (expected)
>>
>> # lsof /vz
>> # umount /vz
>> umount: /vz: device is busy. (unexpected)
>>
>> In this case mountpoint_last() gets an extra refcount on path->mnt
>>
>> Signed-off-by: Vasily Averin <vvs@openvz.org>
> 
> Acked-by: Ian Kent <raven@themaw.net>