Re: [PATCH 0/3] extend get/setrlimit to support setting rlimits external to a process (v7)

[Oleg Nesterov <oleg@redhat.com>]

On 11/09, Jiri Slaby wrote:
>
> On 11/09/2009 05:40 PM, Oleg Nesterov wrote:
> > Also, I don't understand why wthis code get's both pid and task_struct().
>
> And what do you suggest?

Well, I didn't see this patch, except the part quoted in your email...
(btw, thanks for ccing me).

We can call find_task_by_vpid() under rcu and do get_task_struct().
Or, given that we need tasklist anyway we can do find_task_by_vpid()
under tasklist and do not get() at all.

> > And the "if (resource >= RLIM_NLIMITS)" check is racy afaics, see
> > http://marc.info/?l=linux-kernel&m=125200862124872
>
> Your patch (which I have in my series btw) is likely needed for the
> setprlimit syscall, having this on my mind again from now on. But the
> 'if' above is a different story. 'resource' is an index here.

Yes, thanks, I misread this check ;)

> And as a bonus, what I found out now is that /proc/*/limits
> (proc_info_read->proc_pid_limits) doesn't necessarily reflect current
> limits. Since task_lock(current->group_leader) is not held, values of
> one limit may be from the old as well as the currently updated one.
>
> Am I right and do we care at all

I don't know, but personally I think we don't care.

Oleg.