[RFC] is_global_init() called on global init sub-thread

[RFC] is_global_init() called on global init sub-thread

[Sergey Senozhatsky]

Hello,

re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html

Oleg Nesterov wrote:

:Because is_global_init() is only true for the main thread of /sbin/init.
:
:Just look at oom_unkillable_task(). It tries to not kill init. But, say,
:select_bad_process() can happily find a sub-thread of is_global_init()
:and still kill it.

this is still the case, isn't it? at least in some -stable kernels.
is there (or was there) any reason this change has never been committed?
(I'm particularly interested in is_global_init()).

 static inline int is_global_init(struct task_struct *tsk)
 {
-	return tsk->pid == 1;
+	return task_tgid_nr(tsk) == 1;
 }

	-ss

Re: [RFC] is_global_init() called on global init sub-thread

[Serge E. Hallyn]

On Wed, Dec 30, 2015 at 03:25:42PM +0900, Sergey Senozhatsky wrote:
> Hello,
> 
> re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html
> 
> Oleg Nesterov wrote:
> 
> :Because is_global_init() is only true for the main thread of /sbin/init.
> :
> :Just look at oom_unkillable_task(). It tries to not kill init. But, say,
> :select_bad_process() can happily find a sub-thread of is_global_init()
> :and still kill it.
> 
> this is still the case, isn't it? at least in some -stable kernels.
> is there (or was there) any reason this change has never been committed?
> (I'm particularly interested in is_global_init()).

...  seems like it makes sense.  Can you remind us which init you're having
to deal with?

>  static inline int is_global_init(struct task_struct *tsk)
>  {
> -	return tsk->pid == 1;
> +	return task_tgid_nr(tsk) == 1;
>  }
> 
> 	-ss

Re: [RFC] is_global_init() called on global init sub-thread

[Sergey Senozhatsky]

On (12/31/15 19:08), Serge E. Hallyn wrote:
> > re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html
> > 
> > Oleg Nesterov wrote:
> > 
> > :Because is_global_init() is only true for the main thread of /sbin/init.
> > :
> > :Just look at oom_unkillable_task(). It tries to not kill init. But, say,
> > :select_bad_process() can happily find a sub-thread of is_global_init()
> > :and still kill it.
> > 
> > this is still the case, isn't it? at least in some -stable kernels.
> > is there (or was there) any reason this change has never been committed?
> > (I'm particularly interested in is_global_init()).
> 
> ...  seems like it makes sense.  Can you remind us which init you're having
> to deal with?
> 

systemd

	-ss

> >  static inline int is_global_init(struct task_struct *tsk)
> >  {
> > -	return tsk->pid == 1;
> > +	return task_tgid_nr(tsk) == 1;
> >  }

Re: [RFC] is_global_init() called on global init sub-thread

[Serge E. Hallyn]

On Fri, Jan 01, 2016 at 10:10:35AM +0900, Sergey Senozhatsky wrote:
> On (12/31/15 19:08), Serge E. Hallyn wrote:
> > > re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html
> > > 
> > > Oleg Nesterov wrote:
> > > 
> > > :Because is_global_init() is only true for the main thread of /sbin/init.
> > > :
> > > :Just look at oom_unkillable_task(). It tries to not kill init. But, say,
> > > :select_bad_process() can happily find a sub-thread of is_global_init()
> > > :and still kill it.
> > > 
> > > this is still the case, isn't it? at least in some -stable kernels.
> > > is there (or was there) any reason this change has never been committed?
> > > (I'm particularly interested in is_global_init()).
> > 
> > ...  seems like it makes sense.  Can you remind us which init you're having
> > to deal with?
> > 
> 
> systemd
> 
> 	-ss

Well it makes sense to me.  The question is whether we are protecting the
thing running as init, or the 'physical' thread with pid 1.  I think it's
the former, so let's push on this.  Please resend the patch with a proper
signed-off-by, and feel free to add

Acked-by: Serge Hallyn <serge.hallyn@canonical.com>

Re: [RFC] is_global_init() called on global init sub-thread

[Sergey Senozhatsky]

On (12/31/15 19:18), Serge E. Hallyn wrote:
> On Fri, Jan 01, 2016 at 10:10:35AM +0900, Sergey Senozhatsky wrote:
> > On (12/31/15 19:08), Serge E. Hallyn wrote:
> > > > re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html
> > > > 
> > > > Oleg Nesterov wrote:
> > > > 
> > > > :Because is_global_init() is only true for the main thread of /sbin/init.
> > > > :
> > > > :Just look at oom_unkillable_task(). It tries to not kill init. But, say,
> > > > :select_bad_process() can happily find a sub-thread of is_global_init()
> > > > :and still kill it.
> > > > 
> > > > this is still the case, isn't it? at least in some -stable kernels.
> > > > is there (or was there) any reason this change has never been committed?
> > > > (I'm particularly interested in is_global_init()).
> > > 
> > > ...  seems like it makes sense.  Can you remind us which init you're having
> > > to deal with?
> > > 
> > 
> > systemd
> > 
> > 	-ss
> 
> Well it makes sense to me.  The question is whether we are protecting the
> thing running as init, or the 'physical' thread with pid 1.  I think it's
> the former, so let's push on this.  Please resend the patch with a proper
> signed-off-by, and feel free to add

thanks. a bit puzzled, would reported-by Oleg and suggested-by Richard
be appropriate? (no objections if Oleg or Richard will submit it).

> Acked-by: Serge Hallyn <serge.hallyn@canonical.com>

	-ss

Re: [RFC] is_global_init() called on global init sub-thread

[Serge E. Hallyn]

On Fri, Jan 01, 2016 at 10:33:53AM +0900, Sergey Senozhatsky wrote:
> On (12/31/15 19:18), Serge E. Hallyn wrote:
> > On Fri, Jan 01, 2016 at 10:10:35AM +0900, Sergey Senozhatsky wrote:
> > > On (12/31/15 19:08), Serge E. Hallyn wrote:
> > > > > re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html
> > > > > 
> > > > > Oleg Nesterov wrote:
> > > > > 
> > > > > :Because is_global_init() is only true for the main thread of /sbin/init.
> > > > > :
> > > > > :Just look at oom_unkillable_task(). It tries to not kill init. But, say,
> > > > > :select_bad_process() can happily find a sub-thread of is_global_init()
> > > > > :and still kill it.
> > > > > 
> > > > > this is still the case, isn't it? at least in some -stable kernels.
> > > > > is there (or was there) any reason this change has never been committed?
> > > > > (I'm particularly interested in is_global_init()).
> > > > 
> > > > ...  seems like it makes sense.  Can you remind us which init you're having
> > > > to deal with?
> > > > 
> > > 
> > > systemd
> > > 
> > > 	-ss
> > 
> > Well it makes sense to me.  The question is whether we are protecting the
> > thing running as init, or the 'physical' thread with pid 1.  I think it's
> > the former, so let's push on this.  Please resend the patch with a proper
> > signed-off-by, and feel free to add
> 
> thanks. a bit puzzled, would reported-by Oleg and suggested-by Richard
> be appropriate?

Sounds good.

>  (no objections if Oleg or Richard will submit it).

> > Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
> 
> 	-ss

Re: [RFC] is_global_init() called on global init sub-thread

[Richard Guy Briggs]

On 16/01/01, Sergey Senozhatsky wrote:
> On (12/31/15 19:18), Serge E. Hallyn wrote:
> > On Fri, Jan 01, 2016 at 10:10:35AM +0900, Sergey Senozhatsky wrote:
> > > On (12/31/15 19:08), Serge E. Hallyn wrote:
> > > > > re-upping https://www.redhat.com/archives/linux-audit/2013-December/msg00086.html
> > > > > 
> > > > > Oleg Nesterov wrote:
> > > > > 
> > > > > :Because is_global_init() is only true for the main thread of /sbin/init.
> > > > > :
> > > > > :Just look at oom_unkillable_task(). It tries to not kill init. But, say,
> > > > > :select_bad_process() can happily find a sub-thread of is_global_init()
> > > > > :and still kill it.
> > > > > 
> > > > > this is still the case, isn't it? at least in some -stable kernels.
> > > > > is there (or was there) any reason this change has never been committed?
> > > > > (I'm particularly interested in is_global_init()).
> > > > 
> > > > ...  seems like it makes sense.  Can you remind us which init you're having
> > > > to deal with?
> > > > 
> > > 
> > > systemd
> > > 
> > > 	-ss
> > 
> > Well it makes sense to me.  The question is whether we are protecting the
> > thing running as init, or the 'physical' thread with pid 1.  I think it's
> > the former, so let's push on this.  Please resend the patch with a proper
> > signed-off-by, and feel free to add
> 
> thanks. a bit puzzled, would reported-by Oleg and suggested-by Richard
> be appropriate? (no objections if Oleg or Richard will submit it).

This works for me.  I have more patches related to it, but I'll
re-submit them later.

> > Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
> 
> 	-ss

- RGB

--
Richard Guy Briggs <rbriggs@redhat.com>
Senior Software Engineer, Kernel Security, AMER ENG Base Operating Systems, Red Hat
Remote, Ottawa, Canada
Voice: +1.647.777.2635, Internal: (81) 32635, Alt: +1.613.693.0684x3545