linux-kernel - Re: [PATCH] mm/zsmalloc: add trace events for zs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 13 Jun 2016 09:51:20 +0800
From:	Ganesh Mahendran <opensource.ganesh@...il.com>
To:	Sergey Senozhatsky <sergey.senozhatsky@...il.com>
Cc:	Minchan Kim <minchan@...nel.org>, Linux-MM <linux-mm@...ck.org>,
	linux-kernel <linux-kernel@...r.kernel.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Nitin Gupta <ngupta@...are.org>,
	Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>,
	rostedt@...dmis.org, mingo@...hat.com
Subject: Re: [PATCH] mm/zsmalloc: add trace events for zs_compact

Hi, Sergey:

2016-06-08 22:10 GMT+08:00 Sergey Senozhatsky <sergey.senozhatsky@...il.com>:
> Hello,
>
> On (06/08/16 14:39), Ganesh Mahendran wrote:
>> >> > On Tue, Jun 07, 2016 at 04:56:44PM +0800, Ganesh Mahendran wrote:
>> >> >> Currently zsmalloc is widely used in android device.
>> >> >> Sometimes, we want to see how frequently zs_compact is
>> >> >> triggered or how may pages freed by zs_compact(), or which
>> >> >> zsmalloc pool is compacted.
>> >> >>
>> >> >> Most of the time, user can get the brief information from
>> >> >> trace_mm_shrink_slab_[start | end], but in some senario,
>> >> >> they do not use zsmalloc shrinker, but trigger compaction manually.
>> >> >> So add some trace events in zs_compact is convenient. Also we
>> >> >> can add some zsmalloc specific information(pool name, total compact
>> >> >> pages, etc) in zsmalloc trace.
>> >> >
>> >> > Sorry, I cannot understand what's the problem now and what you want to
>> >> > solve. Could you elaborate it a bit?
>> >> >
>> >> > Thanks.
>> >>
>> >> We have backported the zs_compact() to our product(kernel 3.18).
>> >> It is usefull for a longtime running device.
>> >> But there is not a convenient way to get the detailed information
>> >> of zs_comapct() which is usefull for  performance optimization.
>> >> Information about how much time zs_compact used, which pool is
>> >> compacted, how many page freed, etc.
>
> sorry, couldn't check my email earlier.

Sorry for the late response. I'm off for a few days annual leave.

>
> zs_compact() is just one of the N sites that are getting called by
> the shrinker; optimization here will "solve" only 1/N of the problems.
> are there any trace events in any other shrinker callbacks?

Yes, there are trace events in ext4 shrinker:
-----
fs/ext4/extents_status.c: ext4_es_scan()
{
        trace_ext4_es_shrink_scan_enter(sbi->s_sb, nr_to_scan, ret);
...
        nr_shrunk = __es_shrink(sbi, nr_to_scan, NULL);

        trace_ext4_es_shrink_scan_exit(sbi->s_sb, nr_shrunk, ret);
}
-----

>
>
> why trace_mm_shrink_slab_start()/trace_mm_shrink_slab_end()/etc. don't work you?

I think trace_mm_shrink_slab_start/end is for general usage. If we
have some specific information(such as pool name,  shrink count
or others in the future), it will be convenient when we have
zs_compact trace event.

Also, if user do manually zs_compact instead of shrinker,
he can get zs_compact information from trace event.

Thanks.

>
>         -ss