lkml.org 
[lkml]   [2013]   [May]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH RESEND v3 00/11] mm: fixup changers of per cpu pageset's ->high and ->batch
On 05/13/2013 12:20 PM, Pekka Enberg wrote:
> Hi Cody,
>
> On Mon, May 13, 2013 at 10:08 PM, Cody P Schafer
> <cody@linux.vnet.ibm.com> wrote:
>> "Problems" with the current code:
>> 1. there is a lack of synchronization in setting ->high and ->batch in
>> percpu_pagelist_fraction_sysctl_handler()
>> 2. stop_machine() in zone_pcp_update() is unnecissary.
>> 3. zone_pcp_update() does not consider the case where percpu_pagelist_fraction is non-zero
>
> Maybe it's just me but I find the above problem description confusing.
> How does the problem manifest itself?

1. I've not reproduced this causing issues.
2. Calling zone_pcp_update() is slow.
3. Not reproduced either, but would cause percpu_pagelist_fraction (set
via sysctl) to be ignored after a call to zone_pcp_update() (for
example, after a memory hotplug).

> How did you find about it?

I'm writing some code that resizes zones and thus uses
zone_pcp_update(), and fixing broken things along the way.

> Why
> do we need to fix all three problems in the same patch set?

They all affect the same bit of code and fixing all of them means
restructuring the both of the sites where ->high and ->batch are set.

Additionally, splitting it out (if possible) would make it less clear
what the overall goal is, and would mean a few inter-patchset
dependencies, which are undesirable.



\
 
 \ /
  Last update: 2013-05-13 23:21    [W:0.149 / U:0.116 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site