cgroup_disable=memory for 2.6.25?

cgroup_disable=memory for 2.6.25?

[Hugh Dickins]

Hi Balbir,

I'm rather surprised that nobody has pushed -mm's

cgroups-add-cgroup-support-for-enabling-controllers-at-boot-time.patch
cgroups-add-cgroup-support-for-enabling-controllers-at-boot-time-fix-boot-option-parsing.patch
memory-controller-make-memory-resource-control-aware-of-boot-options.patch

into 2.6.25: which was what I'd expected when I first suggested that
distros might want a way to build with the potential for mem cgroups,
but be able to switch off their significant overhead for everyone not
interested.

Ballpark figures, I'm trying to get this question out rather than
processing the exact numbers: CONFIG_CGROUP_MEM_RES_CTLR adds 15%
overhead to the affected paths, booting with cgroup_disable=memory
cuts that back to 1% overhead (due to slightly bigger struct page).

I'm no expert on distros, they may have no interest whatever in
CONFIG_CGROUP_MEM_RES_CTLR=y; and the rest of us can easily build
with or without it, or apply the cgroup_disable=memory patches.

But if those patches serve a purpose, shouldn't they be in 2.6.25?

Hugh

Re: cgroup_disable=memory for 2.6.25?

[Balbir Singh]

Hugh Dickins wrote:
> Hi Balbir,
> 
> I'm rather surprised that nobody has pushed -mm's
> 
> cgroups-add-cgroup-support-for-enabling-controllers-at-boot-time.patch
> cgroups-add-cgroup-support-for-enabling-controllers-at-boot-time-fix-boot-option-parsing.patch
> memory-controller-make-memory-resource-control-aware-of-boot-options.patch
> 
> into 2.6.25: which was what I'd expected when I first suggested that
> distros might want a way to build with the potential for mem cgroups,
> but be able to switch off their significant overhead for everyone not
> interested.
> 
> Ballpark figures, I'm trying to get this question out rather than
> processing the exact numbers: CONFIG_CGROUP_MEM_RES_CTLR adds 15%
> overhead to the affected paths, booting with cgroup_disable=memory
> cuts that back to 1% overhead (due to slightly bigger struct page).
> 
> I'm no expert on distros, they may have no interest whatever in
> CONFIG_CGROUP_MEM_RES_CTLR=y; and the rest of us can easily build
> with or without it, or apply the cgroup_disable=memory patches.
> 
> But if those patches serve a purpose, shouldn't they be in 2.6.25?

Hi, Hugh,

I expected those patches to make it into 2.6.25. But ever since 2.6.25-rc5-mm1,
the next -mm was for 2.6.25-rc8. I have been meaning to follow up with Andrew,
but lost with some other patches.

Andrew,

Could we please push these patches upstream before 2.6.25? Or is it too late?


-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

Re: cgroup_disable=memory for 2.6.25?

[Andrew Morton]

On Fri, 04 Apr 2008 01:57:14 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> I expected those patches to make it into 2.6.25.

My mind-ray receptors might need a tune-up.

> Could we please push these patches upstream before 2.6.25? Or is it too late?

Should be OK.  I borrowed a bit of Hugh's email and stuck it in the changelog
so that it at least has some attempt to describe why it's an urgent change.

Re: cgroup_disable=memory for 2.6.25?

[KAMEZAWA Hiroyuki]

On Thu, 3 Apr 2008 13:41:53 -0700
Andrew Morton <akpm@linux-foundation.org> wrote:

> On Fri, 04 Apr 2008 01:57:14 +0530
> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
> > I expected those patches to make it into 2.6.25.
> 
> My mind-ray receptors might need a tune-up.
> 
> > Could we please push these patches upstream before 2.6.25? Or is it too late?
> 
> Should be OK.  I borrowed a bit of Hugh's email and stuck it in the changelog
> so that it at least has some attempt to describe why it's an urgent change.

For example. Unix bench's execl test result on x86_64 was

== just after boot without mounting any cgroup fs.==
mem_cgorup=off : Execl Throughput       43.0     3150.1      732.6
mem_cgroup=on  : Execl Throughput       43.0     2932.6      682.0
==

I believe this is worth ot be urgent.

I'm now trying performance improvement patches but it's too late.

Thanks,
-Kame

Re: cgroup_disable=memory for 2.6.25?

[Balbir Singh]

Andrew Morton wrote:
> On Fri, 04 Apr 2008 01:57:14 +0530
> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
>> I expected those patches to make it into 2.6.25.
> 
> My mind-ray receptors might need a tune-up.
> 

I should have stressed a bit more on the priority.

>> Could we please push these patches upstream before 2.6.25? Or is it too late?
> 
> Should be OK.  I borrowed a bit of Hugh's email and stuck it in the changelog
> so that it at least has some attempt to describe why it's an urgent change.


Thanks, Andrew!

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL