linux-kernel - Re: [PATCH] mm: compaction: Trace compaction begin and end

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]

Message-ID: <52A19DDF.9050608@suse.cz>
Date:	Fri, 06 Dec 2013 10:50:23 +0100
From:	Vlastimil Babka <vbabka@...e.cz>
To:	Mel Gorman <mgorman@...e.de>
Cc:	Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org, Rik van Riel <riel@...hat.com>
Subject: Re: [PATCH] mm: compaction: Trace compaction begin and end

On 12/05/2013 10:05 AM, Mel Gorman wrote:
> On Wed, Dec 04, 2013 at 03:51:57PM +0100, Vlastimil Babka wrote:
>> On 12/04/2013 03:30 PM, Mel Gorman wrote:
>>> This patch adds two tracepoints for compaction begin and end of a zone. Using
>>> this it is possible to calculate how much time a workload is spending
>>> within compaction and potentially debug problems related to cached pfns
>>> for scanning.
>>
>> I guess for debugging pfns it would be also useful to print their
>> values also in mm_compaction_end.
>>
>
> What additional information would we get from it and what new
> conclusions could we draw? We could guess how much work the
> scanners did but the trace_mm_compaction_isolate_freepages and
> trace_mm_compaction_isolate_migratepages tracepoints already accurately
> tell us that. The scanner PFNs alone do not tell us if the cached pfns
> were updated and even if it did, the information can be changed by
> parallel resets so it would be hard to draw reasonable conclusions from
> the information. We could guess where compaction hotspots might be but
> without the skip information, we could not detect it accurately.  If we
> wanted to detect that accurately, the mm_compaction_isolate* tracepoints
> would be the one to update.

OK, I agree. I guess multiple compaction_begin events would hint at 
scanners being stuck anyway.

> I was primarily concerned about compaction time so I might be looking
> at this the wrong way but it feels like having the PFNs at the end of a
> compaction cycle would be of marginal benefit.
>
>>> In combination with the direct reclaim and slab trace points
>>> it should be possible to estimate most allocation-related overhead for
>>> a workload.
>>>
>>> Signed-off-by: Mel Gorman <mgorman@...e.de>
>>> ---
>>>   include/trace/events/compaction.h | 42 +++++++++++++++++++++++++++++++++++++++
>>>   mm/compaction.c                   |  4 ++++
>>>   2 files changed, 46 insertions(+)
>>>
>>> diff --git a/include/trace/events/compaction.h b/include/trace/events/compaction.h
>>> index fde1b3e..f4e115a 100644
>>> --- a/include/trace/events/compaction.h
>>> +++ b/include/trace/events/compaction.h
>>> @@ -67,6 +67,48 @@ TRACE_EVENT(mm_compaction_migratepages,
>>>   		__entry->nr_failed)
>>>   );
>>>
>>> +TRACE_EVENT(mm_compaction_begin,
>>> +	TP_PROTO(unsigned long zone_start, unsigned long migrate_start,
>>> +		unsigned long zone_end, unsigned long free_start),
>>> +
>>> +	TP_ARGS(zone_start, migrate_start, zone_end, free_start),
>>
>> IMHO a better order would be:
>>   zone_start, migrate_start, free_start, zone_end
>> (well especially in the TP_printk part anyway).
>>
>
> Ok, that would put them in PFN order which may be easier to visualise.
> I'll post a V2 with that change at least.
>

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