[PATCH -mm] remove node_ prefix_from ns subsystem

[PATCH -mm] remove node_ prefix_from ns subsystem

[Cedric Le Goater]

From: Cedric Le Goater <clg@fr.ibm.com>

This is a slight change in the namespace cgroup subsystem api.

Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>
---
 kernel/cgroup.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: 2.6.26-rc2-mm1/kernel/cgroup.c
===================================================================
--- 2.6.26-rc2-mm1.orig/kernel/cgroup.c
+++ 2.6.26-rc2-mm1/kernel/cgroup.c
@@ -2903,7 +2903,7 @@ int cgroup_clone(struct task_struct *tsk
 	cg = tsk->cgroups;
 	parent = task_cgroup(tsk, subsys->subsys_id);
 
-	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
+	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
 
 	/* Pin the hierarchy */
 	atomic_inc(&parent->root->sb->s_active);

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Serge E. Hallyn]

Quoting Cedric Le Goater (clg@fr.ibm.com):
> From: Cedric Le Goater <clg@fr.ibm.com>
> 
> This is a slight change in the namespace cgroup subsystem api.
> 
> Signed-off-by: Cedric Le Goater <clg@fr.ibm.com>

I agree that is preferable.

Acked-by: Serge Hallyn <serue@us.ibm.com>

> ---
>  kernel/cgroup.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> Index: 2.6.26-rc2-mm1/kernel/cgroup.c
> ===================================================================
> --- 2.6.26-rc2-mm1.orig/kernel/cgroup.c
> +++ 2.6.26-rc2-mm1/kernel/cgroup.c
> @@ -2903,7 +2903,7 @@ int cgroup_clone(struct task_struct *tsk
>  	cg = tsk->cgroups;
>  	parent = task_cgroup(tsk, subsys->subsys_id);
> 
> -	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
> +	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
> 
>  	/* Pin the hierarchy */
>  	atomic_inc(&parent->root->sb->s_active);

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Andrew Morton]

On Wed, 21 May 2008 17:47:21 +0200 Cedric Le Goater <clg-NmTC/0ZBporQT0dZR+AlfA@public.gmane.org> wrote:

> From: Cedric Le Goater <clg-NmTC/0ZBporQT0dZR+AlfA@public.gmane.org>
> 
> This is a slight change in the namespace cgroup subsystem api.
> 

What are the reasons for making this change?

> ---
>  kernel/cgroup.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> Index: 2.6.26-rc2-mm1/kernel/cgroup.c
> ===================================================================
> --- 2.6.26-rc2-mm1.orig/kernel/cgroup.c
> +++ 2.6.26-rc2-mm1/kernel/cgroup.c
> @@ -2903,7 +2903,7 @@ int cgroup_clone(struct task_struct *tsk
>  	cg = tsk->cgroups;
>  	parent = task_cgroup(tsk, subsys->subsys_id);
>  
> -	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
> +	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
>  
>  	/* Pin the hierarchy */
>  	atomic_inc(&parent->root->sb->s_active);

This is user-visible, isn't it?  Been there since October.  How can we
safely change it now?

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Paul Menage]

On Wed, May 21, 2008 at 10:06 PM, Andrew Morton
<akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org> wrote:
>> --- 2.6.26-rc2-mm1.orig/kernel/cgroup.c
>> +++ 2.6.26-rc2-mm1/kernel/cgroup.c
>> @@ -2903,7 +2903,7 @@ int cgroup_clone(struct task_struct *tsk
>>       cg = tsk->cgroups;
>>       parent = task_cgroup(tsk, subsys->subsys_id);
>>
>> -     snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
>> +     snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
>>
>>       /* Pin the hierarchy */
>>       atomic_inc(&parent->root->sb->s_active);
>
> This is user-visible, isn't it?  Been there since October.  How can we
> safely change it now?

It is user-visible, but I'd be extremely surprised if anyone was
depending on it or even using it at this point.

Paul

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Andrew Morton]

On Thu, 22 May 2008 00:11:56 -0700 "Paul Menage" <menage-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org> wrote:

> On Wed, May 21, 2008 at 10:06 PM, Andrew Morton
> <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org> wrote:
> >> --- 2.6.26-rc2-mm1.orig/kernel/cgroup.c
> >> +++ 2.6.26-rc2-mm1/kernel/cgroup.c
> >> @@ -2903,7 +2903,7 @@ int cgroup_clone(struct task_struct *tsk
> >>       cg = tsk->cgroups;
> >>       parent = task_cgroup(tsk, subsys->subsys_id);
> >>
> >> -     snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
> >> +     snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
> >>
> >>       /* Pin the hierarchy */
> >>       atomic_inc(&parent->root->sb->s_active);
> >
> > This is user-visible, isn't it?  Been there since October.  How can we
> > safely change it now?
> 
> It is user-visible, but I'd be extremely surprised if anyone was
> depending on it or even using it at this point.

After six months?

Oh well, I guess the number is small, and they'll know how to fix it.

What is the reason for making this change?

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Cedric Le Goater]

Andrew Morton wrote:
> On Thu, 22 May 2008 00:11:56 -0700 "Paul Menage" <menage@google.com> wrote:
> 
>> On Wed, May 21, 2008 at 10:06 PM, Andrew Morton
>> <akpm@linux-foundation.org> wrote:
>>>> --- 2.6.26-rc2-mm1.orig/kernel/cgroup.c
>>>> +++ 2.6.26-rc2-mm1/kernel/cgroup.c
>>>> @@ -2903,7 +2903,7 @@ int cgroup_clone(struct task_struct *tsk
>>>>       cg = tsk->cgroups;
>>>>       parent = task_cgroup(tsk, subsys->subsys_id);
>>>>
>>>> -     snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
>>>> +     snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
>>>>
>>>>       /* Pin the hierarchy */
>>>>       atomic_inc(&parent->root->sb->s_active);
>>> This is user-visible, isn't it?  Been there since October.  How can we
>>> safely change it now?
>> It is user-visible, but I'd be extremely surprised if anyone was
>> depending on it or even using it at this point.
> 
> After six months?

yes that's what we believe also.

> Oh well, I guess the number is small, and they'll know how to fix it.
> 
> What is the reason for making this change?

the 'node_' prefix really is superfluous and misleading. what's a node ?

C.

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Andrew Morton]

On Thu, 22 May 2008 09:36:23 +0200 Cedric Le Goater <clg-NmTC/0ZBporQT0dZR+AlfA@public.gmane.org> wrote:

> > What is the reason for making this change?
> 
> the 'node_' prefix really is superfluous and misleading. what's a node ?

For gawd's sake guys.  I've asked three times and now I'm getting
rhetorical questions for answers.  It's like drawing teeth.

Will someone please write a halfway decent description of this change? 
One which (at least) covers all the questions I've been asking?

What the change is, why it is being made, what the user-visible
presentation is, what the impact upon users is, why we think it won't
be a problem, etc?  The stuff which should have been right there from
day one, before the code change was even made?

Thanks.

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Paul Menage]

Hi Andrew,

On Thu, May 22, 2008 at 1:10 AM, Andrew Morton
<akpm@linux-foundation.org> wrote:
>
> What the change is, why it is being made, what the user-visible
> presentation is, what the impact upon users is, why we think it won't
> be a problem, etc?  The stuff which should have been right there from
> day one, before the code change was even made?

The change is that previously when cgroup_clone() was called
(currently only from the unshare path in ns_proxy cgroup, you'd get a
new group named "node_$pid" whereas now you'll get a group named after
just your pid.)

The only users who would notice it are those who are using the
ns_proxy cgroup subsystem to auto-create cgroups when namespaces are
unshared - something of an experimental feature, which I think really
needs more complete container/namespace support in order to be useful.
I suspect the only users are Cedric and Serge, or maybe a few others
on containers@lists.linux-foundation.org. And in fact it would only be
noticed by the users who make the assumption about how the name is
generated, rather than getting it from the /proc/<pid>/cgroups file
for the process in question.

Whether the change is actually needed or not I'm fairly agnostic on,
but I guess it is more elegant to just use the pid as the new group
name rather than adding a fairly arbitrary "node_" prefix on the
front.

Paul

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Andrew Morton]

On Thu, 22 May 2008 01:23:35 -0700 "Paul Menage" <menage-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org> wrote:

> Hi Andrew,
> 
> On Thu, May 22, 2008 at 1:10 AM, Andrew Morton
> <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org> wrote:
> >
> > What the change is, why it is being made, what the user-visible
> > presentation is, what the impact upon users is, why we think it won't
> > be a problem, etc?  The stuff which should have been right there from
> > day one, before the code change was even made?
> 
> The change is that previously when cgroup_clone() was called
> (currently only from the unshare path in ns_proxy cgroup, you'd get a
> new group named "node_$pid" whereas now you'll get a group named after
> just your pid.)
> 
> The only users who would notice it are those who are using the
> ns_proxy cgroup subsystem to auto-create cgroups when namespaces are
> unshared - something of an experimental feature, which I think really
> needs more complete container/namespace support in order to be useful.
> I suspect the only users are Cedric and Serge, or maybe a few others
> on containers-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cuFJUID1Z7QI@public.gmane.org And in fact it would only be
> noticed by the users who make the assumption about how the name is
> generated, rather than getting it from the /proc/<pid>/cgroups file
> for the process in question.
> 
> Whether the change is actually needed or not I'm fairly agnostic on,
> but I guess it is more elegant to just use the pid as the new group
> name rather than adding a fairly arbitrary "node_" prefix on the
> front.
> 

Well I suppose that as a non-back-compatible change we should feed it
into 2.6.25.x as well.  It's a bit unusual, but so doing will reduce the
number of hey-where-did-my-file-go discoveries.

Re: [PATCH -mm] remove node_ prefix_from ns subsystem

[Cedric Le Goater]

Andrew Morton wrote:
> On Thu, 22 May 2008 01:23:35 -0700 "Paul Menage" <menage@google.com> wrote:
> 
>> Hi Andrew,
>>
>> On Thu, May 22, 2008 at 1:10 AM, Andrew Morton
>> <akpm@linux-foundation.org> wrote:
>>> What the change is, why it is being made, what the user-visible
>>> presentation is, what the impact upon users is, why we think it won't
>>> be a problem, etc?  The stuff which should have been right there from
>>> day one, before the code change was even made?
>> The change is that previously when cgroup_clone() was called
>> (currently only from the unshare path in ns_proxy cgroup, you'd get a
>> new group named "node_$pid" whereas now you'll get a group named after
>> just your pid.)
>>
>> The only users who would notice it are those who are using the
>> ns_proxy cgroup subsystem to auto-create cgroups when namespaces are
>> unshared - something of an experimental feature, which I think really
>> needs more complete container/namespace support in order to be useful.
>> I suspect the only users are Cedric and Serge, or maybe a few others
>> on containers@lists.linux-foundation.org. And in fact it would only be
>> noticed by the users who make the assumption about how the name is
>> generated, rather than getting it from the /proc/<pid>/cgroups file
>> for the process in question.
>>
>> Whether the change is actually needed or not I'm fairly agnostic on,
>> but I guess it is more elegant to just use the pid as the new group
>> name rather than adding a fairly arbitrary "node_" prefix on the
>> front.

Thanks Paul.

> Well I suppose that as a non-back-compatible change we should feed it
> into 2.6.25.x as well.  It's a bit unusual, but so doing will reduce the
> number of hey-where-did-my-file-go discoveries.

Thanks andrew. I have being keeping that patch for so long that I didn't
realize that the ns cgroup subsystem had been already merged. sorry for 
the noise.

C.