Re: [PATCH v3] ipc: Modify message queue accounting to not take kernel data structures into account

[Davidlohr Bueso <dave@stgolabs.net>]

On Wed, 2015-07-08 at 15:17 -0400, Doug Ledford wrote:
> I responded yesterday in the v2 patch thread I believe.  In any case, I
> think this patch is fine, and can go to stable.  This patch doesn't
> actually change the math related to the rlimit checks (which is the main
> thing I wanted to correct in my original patches), instead it corrects a
> mistake I made.  At the time, I mistakenly thought that the qsize
> included the current message data total + the struct msg_msg size total.
>  It didn't, it was just the current user data total.  I added the rbtree
> nodes in order to keep the total accurate but I shouldn't have added the
> rbtree nodes to this total because none of the other kernel usage was
> previously included.

Exactly, this is what I was referring to when I suggested staying clear
from rlimit modifications. It just makes 0 sense to account for each
rbtree.

Thanks for taking a look!

> 
> Acked-by: Doug Ledford <dledford@redhat.com>

Acked-by: Davidlohr Bueso <dbueso@suse.de>