Re: [PATCH -mmotm 2/5] memcg: dirty memory documentation

From: Balbir Singh
Date: Wed Mar 17 2010 - 15:03:03 EST


* Greg Thelen <gthelen@xxxxxxxxxx> [2010-03-17 09:48:18]:

> On Mon, Mar 15, 2010 at 11:41 PM, Daisuke Nishimura
> <nishimura@xxxxxxxxxxxxxxxxx> wrote:
> > On Mon, 15 Mar 2010 00:26:39 +0100, Andrea Righi <arighi@xxxxxxxxxxx> wrote:
> >> Document cgroup dirty memory interfaces and statistics.
> >>
> >> Signed-off-by: Andrea Righi <arighi@xxxxxxxxxxx>
> >> ---
> >>  Documentation/cgroups/memory.txt |   36 ++++++++++++++++++++++++++++++++++++
> >>  1 files changed, 36 insertions(+), 0 deletions(-)
> >>
> >> diff --git a/Documentation/cgroups/memory.txt b/Documentation/cgroups/memory.txt
> >> index 49f86f3..38ca499 100644
> >> --- a/Documentation/cgroups/memory.txt
> >> +++ b/Documentation/cgroups/memory.txt
> >> @@ -310,6 +310,11 @@ cache            - # of bytes of page cache memory.
> >>  rss          - # of bytes of anonymous and swap cache memory.
> >>  pgpgin               - # of pages paged in (equivalent to # of charging events).
> >>  pgpgout              - # of pages paged out (equivalent to # of uncharging events).
> >> +filedirty    - # of pages that are waiting to get written back to the disk.
> >> +writeback    - # of pages that are actively being written back to the disk.
> >> +writeback_tmp        - # of pages used by FUSE for temporary writeback buffers.
> >> +nfs          - # of NFS pages sent to the server, but not yet committed to
> >> +               the actual storage.
>
> Should these new memory.stat counters (filedirty, etc) report byte
> counts rather than page counts? I am thinking that byte counters
> would make reporting more obvious depending on how heterogeneous page
> sizes are used. Byte counters would also agree with /proc/meminfo.
> Within the kernel we could still maintain page counts. The only
> change would be to the reporting routine, mem_cgroup_get_local_stat(),
> which would scale the page counts by PAGE_SIZE as it does for for
> cache,rss,etc.
>

I agree, byte counts would be better than page counts. pgpin and
pgpout are special cases where the pages matter, the size does not due
to the nature of the operation.

--
Three Cheers,
Balbir
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/