lkml.org 
[lkml]   [2015]   [Jul]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/5] memcg: export struct mem_cgroup
On Wed 15-07-15 13:57:11, Andrew Morton wrote:
> On Wed, 15 Jul 2015 13:14:41 +0200 Michal Hocko <mhocko@kernel.org> wrote:
>
> > mem_cgroup structure is defined in mm/memcontrol.c currently which
> > means that the code outside of this file has to use external API even
> > for trivial access stuff.
> >
> > This patch exports mm_struct with its dependencies and makes some of the
> > exported functions inlines. This even helps to reduce the code size a bit
> > (make defconfig + CONFIG_MEMCG=y)
> >
> > text data bss dec hex filename
> > 12355346 1823792 1089536 15268674 e8fb42 vmlinux.before
> > 12354970 1823792 1089536 15268298 e8f9ca vmlinux.after
> >
> > This is not much (370B) but better than nothing. We also save a function
> > call in some hot paths like callers of mem_cgroup_count_vm_event which is
> > used for accounting.
> >
> > The patch doesn't introduce any functional changes.
> >
> > ...
> >
> > include/linux/memcontrol.h | 369 +++++++++++++++++++++++++++++++++++++++++----
>
> Boy, that's a ton of new stuff into the header file. Do we actually
> *need* to expose all this?

I am exporting struct mem_cgroup with its dependencies + some small
functions which allow to inline some really trivial code and helps to
generate a better code.

> Is some other patch dependent on it?

Without mem_cgroup visible outside of memcontrol.c we couldn't inline
and now we can also use some fields from mem_cgroup directly and get rid
of some really trivial access functions.

> If
> not then perhaps we shouldn't do this - if the code was already this
> way, I'd be attracted to a patch which was the reverse of this one!

I agree with Johannes who originally suggested to expose mem_cgroup that
it will allow for a better code later.

> There's some risk of build breakage here - just from a quick scan,
> memcontrol.h is going to need eventfd.h for eventfd_ctx. But what else
> is needed?

I have tested this with all{mod,yes,no}config + my battery of configs
which I am using for mm git tree testing + some randconfig without
issues. Sure there might be some config combo I haven't tested but I
guess it should be quite unlikely.
--
Michal Hocko
SUSE Labs


\
 
 \ /
  Last update: 2015-07-16 09:41    [W:0.179 / U:2.492 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site