lkml.org 
[lkml]   [2011]   [Sep]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] per-cgroup tcp buffer limitation
On 09/13/2011 03:46 PM, Paul Menage wrote:
> On Tue, Sep 13, 2011 at 11:11 AM, Glauber Costa<glommer@parallels.com> wrote:
>>
>> What if they are all updated under the same lock ?
>
> Right, that would be the kind of optimization that would remove the
> need for worrying about whether or not to account it. It would
> probably mean creating some memcg-specific structures like
> res-counters that could handle multiple values, since you'd need to
> update both the kernel charge and the total charge, in this cgroup
> *and* its ancestors.
>
> Paul
If we do that, we may have to commit to an intermediary user interface -
with controls to to determine if kernel memory is billed to kernel or
total, a enable/disable file, just to later render it pointless by a new
optimization - that we seem to agree that seems possible.

I think it is preferred to always assume kernel memory is accounted to
the kernel, and when we optimize it, no changes are made to what's
exposed to userspace.


\
 
 \ /
  Last update: 2011-09-13 21:11    [from the cache]
©2003-2011 Jasper Spaans