find: WARNING: Hard link count is wrong for /proc/1

find: WARNING: Hard link count is wrong for /proc/1

[Vegard Nossum]

Hi,

2.6.26-rc4 seems to exhibit the following (new since at least 2.6.24)
warning when running find /proc/1:

find: WARNING: Hard link count is wrong for /proc/1: this may be a bug
in your filesystem driver.  Automatically turning on find's -noleaf
option.  Earlier results may have failed to include directories that
should have been searched.

# ls -l /proc/1
total 0
-r-------- 1 root root 0 Jun  2 02:20 auxv
--w------- 1 root root 0 Jun  2 02:20 clear_refs
-r--r--r-- 1 root root 0 Jun  2 02:20 cmdline
-rw-r--r-- 1 root root 0 Jun  2 02:20 coredump_filter
lrwxrwxrwx 1 root root 0 Jun  2 02:20 cwd -> /
-r-------- 1 root root 0 Jun  2 02:20 environ
lrwxrwxrwx 1 root root 0 Jun  2 02:20 exe -> /bin/bash
dr-x------ 2 root root 0 Jun  2 02:20 fd
dr-x------ 2 root root 0 Jun  2 02:20 fdinfo
-r-------- 1 root root 0 Jun  2 02:20 limits
-r--r--r-- 1 root root 0 Jun  2 02:20 maps
-rw------- 1 root root 0 Jun  2 02:20 mem
-r--r--r-- 1 root root 0 Jun  2 02:20 mountinfo
-r--r--r-- 1 root root 0 Jun  2 02:20 mounts
-r-------- 1 root root 0 Jun  2 02:20 mountstats
dr-xr-xr-x 5 root root 0 Jun  2 02:20 net
-rw-r--r-- 1 root root 0 Jun  2 02:20 oom_adj
-r--r--r-- 1 root root 0 Jun  2 02:20 oom_score
-r-------- 1 root root 0 Jun  2 02:20 pagemap
lrwxrwxrwx 1 root root 0 Jun  2 02:20 root -> /
-r--r--r-- 1 root root 0 Jun  2 02:20 smaps
-r--r--r-- 1 root root 0 Jun  2 02:20 stat
-r--r--r-- 1 root root 0 Jun  2 02:20 statm
-r--r--r-- 1 root root 0 Jun  2 02:20 status
dr-xr-xr-x 3 root root 0 Jun  2 02:20 task
-r--r--r-- 1 root root 0 Jun  2 02:20 wchan

# stat /proc/1
  File: `/proc/1'
  Size: 0               Blocks: 0          IO Block: 1024   directory
Device: 3h/3d   Inode: 39          Links: 5
Access: (0555/dr-xr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2008-06-02 02:19:07.418398897 -0400
Modify: 2008-06-02 02:19:07.418398897 -0400
Change: 2008-06-02 02:19:07.418398897 -0400

The other /proc/pid entries seem to have the same problem.


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036

Re: find: WARNING: Hard link count is wrong for /proc/1

[Vegard Nossum]

On Mon, Jun 2, 2008 at 8:31 AM, Vegard Nossum <vegard.nossum@gmail.com> wrote:
> Hi,
>
> 2.6.26-rc4 seems to exhibit the following (new since at least 2.6.24)
> warning when running find /proc/1:
>
> find: WARNING: Hard link count is wrong for /proc/1: this may be a bug
> in your filesystem driver.  Automatically turning on find's -noleaf
> option.  Earlier results may have failed to include directories that
> should have been searched.

My prime suspect is

commit e9720acd728a46cb40daa52c99a979f7c4ff195c
Author: Pavel Emelyanov <xemul@openvz.org>
Date:   Fri Mar 7 11:08:40 2008 -0800

    [NET]: Make /proc/net a symlink on /proc/self/net (v3)


as this patch introduces a new directory entry without bumping the nlink count.

Will this do? (Untested)

Vegard


>From a43ef05a8c1f8d45ca88e00fdd31c00ddfdba1b8 Mon Sep 17 00:00:00 2001
From: Vegard Nossum <vegard.nossum@gmail.com>
Date: Mon, 2 Jun 2008 08:57:45 +0200
Subject: [PATCH] proc: fix /proc/<pid> link count

commit e9720acd728a46cb40daa52c99a979f7c4ff195c
Author: Pavel Emelyanov <xemul@openvz.org>
Date:   Fri Mar 7 11:08:40 2008 -0800

    [NET]: Make /proc/net a symlink on /proc/self/net (v3)

introduced /proc/net without bumping the link count on /proc/self.

Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
---
 fs/proc/base.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/fs/proc/base.c b/fs/proc/base.c
index c447e07..10ee4a2 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2589,6 +2589,9 @@ static struct dentry
*proc_pid_instantiate(struct inode *dir,
 #ifdef CONFIG_SECURITY
 	inode->i_nlink += 1;
 #endif
+#ifdef CONFIG_NET
+	inode->i_nlink += 1;
+#endif

 	dentry->d_op = &pid_dentry_operations;

-- 
1.5.4.1

Re: find: WARNING: Hard link count is wrong for /proc/1

[Vegard Nossum]

On Mon, Jun 2, 2008 at 9:00 AM, Vegard Nossum <vegard.nossum@gmail.com> wrote:
...
> Will this do? (Untested)
>
> Vegard
>
>
> From a43ef05a8c1f8d45ca88e00fdd31c00ddfdba1b8 Mon Sep 17 00:00:00 2001
> From: Vegard Nossum <vegard.nossum@gmail.com>
> Date: Mon, 2 Jun 2008 08:57:45 +0200
> Subject: [PATCH] proc: fix /proc/<pid> link count
>
> commit e9720acd728a46cb40daa52c99a979f7c4ff195c
> Author: Pavel Emelyanov <xemul@openvz.org>
> Date:   Fri Mar 7 11:08:40 2008 -0800
>
>    [NET]: Make /proc/net a symlink on /proc/self/net (v3)
>
> introduced /proc/net without bumping the link count on /proc/self.

Oops ---^

Should say /proc/self/net. Sorry for the spamming.


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036

Re: find: WARNING: Hard link count is wrong for /proc/1

[Jan Kara]

> On Mon, Jun 2, 2008 at 8:31 AM, Vegard Nossum <vegard.nossum@gmail.com> wrote:
> > Hi,
> >
> > 2.6.26-rc4 seems to exhibit the following (new since at least 2.6.24)
> > warning when running find /proc/1:
> >
> > find: WARNING: Hard link count is wrong for /proc/1: this may be a bug
> > in your filesystem driver.  Automatically turning on find's -noleaf
> > option.  Earlier results may have failed to include directories that
> > should have been searched.
> 
> My prime suspect is
> 
> commit e9720acd728a46cb40daa52c99a979f7c4ff195c
> Author: Pavel Emelyanov <xemul@openvz.org>
> Date:   Fri Mar 7 11:08:40 2008 -0800
> 
>     [NET]: Make /proc/net a symlink on /proc/self/net (v3)
> 
> 
> as this patch introduces a new directory entry without bumping the nlink count.
> 
> Will this do? (Untested)
> 
> Vegard
> 
> 
> >From a43ef05a8c1f8d45ca88e00fdd31c00ddfdba1b8 Mon Sep 17 00:00:00 2001
> From: Vegard Nossum <vegard.nossum@gmail.com>
> Date: Mon, 2 Jun 2008 08:57:45 +0200
> Subject: [PATCH] proc: fix /proc/<pid> link count
> 
> commit e9720acd728a46cb40daa52c99a979f7c4ff195c
> Author: Pavel Emelyanov <xemul@openvz.org>
> Date:   Fri Mar 7 11:08:40 2008 -0800
> 
>     [NET]: Make /proc/net a symlink on /proc/self/net (v3)
> 
> introduced /proc/net without bumping the link count on /proc/self.
  Looks fine to me :).

								Honza
> Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
> ---
>  fs/proc/base.c |    3 +++
>  1 files changed, 3 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index c447e07..10ee4a2 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -2589,6 +2589,9 @@ static struct dentry
> *proc_pid_instantiate(struct inode *dir,
>  #ifdef CONFIG_SECURITY
>  	inode->i_nlink += 1;
>  #endif
> +#ifdef CONFIG_NET
> +	inode->i_nlink += 1;
> +#endif
> 
>  	dentry->d_op = &pid_dentry_operations;
> 
-- 
Jan Kara <jack@suse.cz>
SuSE CR Labs

Re: find: WARNING: Hard link count is wrong for /proc/1

[Pekka Enberg]

On Mon, Jun 2, 2008 at 10:00 AM, Vegard Nossum <vegard.nossum@gmail.com> wrote:
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index c447e07..10ee4a2 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -2589,6 +2589,9 @@ static struct dentry
> *proc_pid_instantiate(struct inode *dir,
>  #ifdef CONFIG_SECURITY
>        inode->i_nlink += 1;
>  #endif
> +#ifdef CONFIG_NET

#if defined(CONFIG_SECURITY) || defined(CONFIG_NET) ?

> +       inode->i_nlink += 1;
> +#endif
>
>        dentry->d_op = &pid_dentry_operations;
>
> --
> 1.5.4.1
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: find: WARNING: Hard link count is wrong for /proc/1

[Pekka Enberg]

On Mon, Jun 2, 2008 at 7:11 PM, Pekka Enberg <penberg@cs.helsinki.fi> wrote:
> On Mon, Jun 2, 2008 at 10:00 AM, Vegard Nossum <vegard.nossum@gmail.com> wrote:
>> diff --git a/fs/proc/base.c b/fs/proc/base.c
>> index c447e07..10ee4a2 100644
>> --- a/fs/proc/base.c
>> +++ b/fs/proc/base.c
>> @@ -2589,6 +2589,9 @@ static struct dentry
>> *proc_pid_instantiate(struct inode *dir,
>>  #ifdef CONFIG_SECURITY
>>        inode->i_nlink += 1;
>>  #endif
>> +#ifdef CONFIG_NET
>
> #if defined(CONFIG_SECURITY) || defined(CONFIG_NET) ?

Oh, forget it. I need to start thinking before posting.

Re: find: WARNING: Hard link count is wrong for /proc/1

[Rafael C. de Almeida]

Vegard Nossum wrote:
> Hi,
> 
> 2.6.26-rc4 seems to exhibit the following (new since at least 2.6.24)
> warning when running find /proc/1:
> 
> find: WARNING: Hard link count is wrong for /proc/1: this may be a bug
> in your filesystem driver.  Automatically turning on find's -noleaf
> option.  Earlier results may have failed to include directories that
> should have been searched.
> 
> # ls -l /proc/1
> total 0
> -r-------- 1 root root 0 Jun  2 02:20 auxv
> --w------- 1 root root 0 Jun  2 02:20 clear_refs
> -r--r--r-- 1 root root 0 Jun  2 02:20 cmdline
> -rw-r--r-- 1 root root 0 Jun  2 02:20 coredump_filter
> lrwxrwxrwx 1 root root 0 Jun  2 02:20 cwd -> /
> -r-------- 1 root root 0 Jun  2 02:20 environ
> lrwxrwxrwx 1 root root 0 Jun  2 02:20 exe -> /bin/bash
> dr-x------ 2 root root 0 Jun  2 02:20 fd
> dr-x------ 2 root root 0 Jun  2 02:20 fdinfo
> -r-------- 1 root root 0 Jun  2 02:20 limits
> -r--r--r-- 1 root root 0 Jun  2 02:20 maps
> -rw------- 1 root root 0 Jun  2 02:20 mem
> -r--r--r-- 1 root root 0 Jun  2 02:20 mountinfo
> -r--r--r-- 1 root root 0 Jun  2 02:20 mounts
> -r-------- 1 root root 0 Jun  2 02:20 mountstats
> dr-xr-xr-x 5 root root 0 Jun  2 02:20 net
> -rw-r--r-- 1 root root 0 Jun  2 02:20 oom_adj
> -r--r--r-- 1 root root 0 Jun  2 02:20 oom_score
> -r-------- 1 root root 0 Jun  2 02:20 pagemap
> lrwxrwxrwx 1 root root 0 Jun  2 02:20 root -> /
> -r--r--r-- 1 root root 0 Jun  2 02:20 smaps
> -r--r--r-- 1 root root 0 Jun  2 02:20 stat
> -r--r--r-- 1 root root 0 Jun  2 02:20 statm
> -r--r--r-- 1 root root 0 Jun  2 02:20 status
> dr-xr-xr-x 3 root root 0 Jun  2 02:20 task
> -r--r--r-- 1 root root 0 Jun  2 02:20 wchan
> 
> # stat /proc/1
>   File: `/proc/1'
>   Size: 0               Blocks: 0          IO Block: 1024   directory
> Device: 3h/3d   Inode: 39          Links: 5
> Access: (0555/dr-xr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
> Access: 2008-06-02 02:19:07.418398897 -0400
> Modify: 2008-06-02 02:19:07.418398897 -0400
> Change: 2008-06-02 02:19:07.418398897 -0400
> 
> The other /proc/pid entries seem to have the same problem.
> 
> 
> Vegard
> 
The same warning shows up on linux 2.6.25.4.

$ sudo stat /proc/1
  File: `/proc/1'
  Size: 0         	Blocks: 0          IO Block: 1024   directory
Device: 3h/3d	Inode: 111976      Links: 5
Access: (0555/dr-xr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2008-06-02 19:23:41.728134046 -0300
Modify: 2008-06-02 19:23:41.728134046 -0300
Change: 2008-06-02 19:23:41.728134046 -0300

This is the output of find before the warning shows up (it always happen
right after /proc/1/fdinfo/10):
$ sudo find /proc/1
/proc/1
/proc/1/task
/proc/1/task/1
/proc/1/task/1/fd
/proc/1/task/1/fd/10
/proc/1/task/1/fdinfo
/proc/1/task/1/fdinfo/10
/proc/1/task/1/environ
/proc/1/task/1/auxv
/proc/1/task/1/status
/proc/1/task/1/limits
/proc/1/task/1/cmdline
/proc/1/task/1/stat
/proc/1/task/1/statm
/proc/1/task/1/maps
/proc/1/task/1/mem
/proc/1/task/1/cwd
/proc/1/task/1/root
/proc/1/task/1/exe
/proc/1/task/1/mounts
/proc/1/task/1/clear_refs
/proc/1/task/1/smaps
/proc/1/task/1/pagemap
/proc/1/task/1/wchan
/proc/1/task/1/oom_score
/proc/1/task/1/oom_adj
/proc/1/fd
/proc/1/fd/10
/proc/1/fdinfo
/proc/1/fdinfo/10
find: WARNING: Hard link count is wrong for /proc/1: this may be a bug
in your filesystem driver.  Automatically turning on find's -noleaf
option.  Earlier results may have failed to include directories that
should have been searched.

Re: find: WARNING: Hard link count is wrong for /proc/1

[Eric W. Biederman]

From: Vegard Nossum <vegard.nossum@gmail.com>
Date: Mon, 2 Jun 2008 08:57:45 +0200
Subject: [PATCH] proc: fix /proc/<pid> link count

commit e9720acd728a46cb40daa52c99a979f7c4ff195c
Author: Pavel Emelyanov <xemul@openvz.org>
Date:   Fri Mar 7 11:08:40 2008 -0800

    [NET]: Make /proc/net a symlink on /proc/self/net (v3)

introduced /proc/net without bumping the link count on /proc/self.

Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>

---
 fs/proc/base.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/fs/proc/base.c b/fs/proc/base.c
index c447e07..10ee4a2 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2589,6 +2589,9 @@ static struct dentry
*proc_pid_instantiate(struct inode *dir,
 #ifdef CONFIG_SECURITY
 	inode->i_nlink += 1;
 #endif
+#ifdef CONFIG_NET
+	inode->i_nlink += 1;
+#endif

 	dentry->d_op = &pid_dentry_operations;

-- 
1.5.4.1

Re: find: WARNING: Hard link count is wrong for /proc/1

[Andrew Morton]

On Mon, 02 Jun 2008 18:59:46 -0700 ebiederm@xmission.com (Eric W. Biederman) wrote:

> 
> From: Vegard Nossum <vegard.nossum@gmail.com>
> Date: Mon, 2 Jun 2008 08:57:45 +0200
> Subject: [PATCH] proc: fix /proc/<pid> link count
> 
> commit e9720acd728a46cb40daa52c99a979f7c4ff195c
> Author: Pavel Emelyanov <xemul@openvz.org>
> Date:   Fri Mar 7 11:08:40 2008 -0800
> 
>     [NET]: Make /proc/net a symlink on /proc/self/net (v3)
> 
> introduced /proc/net without bumping the link count on /proc/self.
> 
> Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
> Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
> 
> ---
>  fs/proc/base.c |    3 +++
>  1 files changed, 3 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index c447e07..10ee4a2 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -2589,6 +2589,9 @@ static struct dentry
> *proc_pid_instantiate(struct inode *dir,
>  #ifdef CONFIG_SECURITY
>  	inode->i_nlink += 1;
>  #endif
> +#ifdef CONFIG_NET
> +	inode->i_nlink += 1;
> +#endif
> 
>  	dentry->d_op = &pid_dentry_operations;

A bit more explanation would be nice.

I assume that here we're accounting for a reference which net applies
to this dentry.  But it would be clearer and more robust to do that at
the site where that reference is really taken, not over in
proc_pid_instantiate().  Possible?

Re: find: WARNING: Hard link count is wrong for /proc/1

[Vegard Nossum]

On Tue, Jun 3, 2008 at 5:46 AM, Eric W. Biederman <ebiederm@xmission.com> wrote:
> Andrew Morton <akpm@linux-foundation.org> writes:
>>>  #ifdef CONFIG_SECURITY
>>>      inode->i_nlink += 1;
>>>  #endif
>>> +#ifdef CONFIG_NET
>>> +    inode->i_nlink += 1;
>>> +#endif
>>>
>>>      dentry->d_op = &pid_dentry_operations;
>>
>> A bit more explanation would be nice.
>>
>> I assume that here we're accounting for a reference which net applies
>> to this dentry.  But it would be clearer and more robust to do that at
>> the site where that reference is really taken, not over in
>> proc_pid_instantiate().  Possible?
>
> Yes.  Potentially we can read through the tid_base_stuff and
> count the total number of directories at runtime whenever we
> open a /proc/<pid> directory.  It is a smallish table and walking the
> entire table is likely to be lost in the noise, and that would
> certainly improve the maintainability of the code.
>
> Currently what we do is:
>        inode->i_nlink = 5;
> #ifdef CONFIG_SECURITY
>        inode->i_nlink += 1;
> #endif
>
> To get the hard link count of /proc/<pid> correct.
> With /proc/<pid>/net we have added one more optional possibly absent
> directory.
>
> So the above patch is trivially correct in the current form of the
> code but does not improve maintenance.
>
> I am tired and all I have the energy to do today is to review the
> patch and say: "Yes.  The patch works for the correct reasons, let's
> fix this bug."

I can try for the counting fix if you give me a bit of time. I have an
exam today and tomorrow, so I'll likely be able to submit a fix after
that :-)

Maybe you can apply the fix for now and I can follow up with the nicer
counting patch afterwards? (Or if anybody else wants to write it
that's also fine :-))


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036