On 03/12, Djalal Harouni wrote:
>
> On Sun, Mar 11, 2012 at 06:25:12PM +0100, Oleg Nesterov wrote:
> >
> > Well, I don't think it is right to add this counter into task_struct.
> >
> > It should be per-process, signal_struct makes more sense. Or may be
> > mm_struct.
> Some /proc/<pid>/{syscall,stack,...} do not operate on mm_struct so why we
> should add the: "acquire a reference to mm, get exec_id and mmput".
This could be simpler, just read the counter under task_lock(). And
unless I misread the next patches syscall/stack can use current->mm
lockless.
OK, nevermind.
> For the signal_struct currently I don't know, from a comment it seems that
> signal_struct can be shared!
Yes, it is shared, and that is why it makes sense for the per-process
data. All threads in the thread group (process) have the same ->signal.
And unlike ->mm, ->signal survives after exec.
Oleg.
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
