Re: [PATCH v4] tracing: add 'accounted' entry into output of allocation tracepoints

From: Vasily Averin <vvs@openvz.org>
To: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	kernel@openvz.org, linux-kernel@vger.kernel.org,
	Steven Rostedt <rostedt@goodmis.org>,
	Ingo Molnar <mingo@redhat.com>,
	linux-mm@kvack.org, Shakeel Butt <shakeelb@google.com>,
	Roman Gushchin <roman.gushchin@linux.dev>,
	Vlastimil Babka <vbabka@suse.cz>,
	Matthew Wilcox <willy@infradead.org>,
	Joonsoo Kim <iamjoonsoo.kim@lge.com>,
	David Rientjes <rientjes@google.com>,
	Pekka Enberg <penberg@kernel.org>,
	Christoph Lameter <cl@linux.com>, Michal Hocko <mhocko@suse.com>,
	Muchun Song <songmuchun@bytedance.com>
Subject: Re: [PATCH v4] tracing: add 'accounted' entry into output of allocation tracepoints
Date: Sun, 22 May 2022 21:53:41 +0300	[thread overview]
Message-ID: <be8d4fec-9719-10ec-1299-691e498c2f7c@openvz.org> (raw)
In-Reply-To: <YonH2AqfEFYKrUV2@n2.us-central1-a.c.spheric-algebra-350919.internal>

On 5/22/22 08:19, Hyeonggon Yoo wrote:
> On Sun, May 22, 2022 at 07:33:08AM +0300, Vasily Averin wrote:
>> On 5/22/22 06:51, Hyeonggon Yoo wrote:
>>> If we decide to do that, it would be better to print
>>> something like:
>>> slab_flags=SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT|SLAB_STORE_USER
>>> instead of just printing 'accounted=true/false'. This patch is too
>>> specific to SLAB_ACCOUNT.
>>
>> Any extra output degrades performance.
> 
> No strong opinion but just a concern that maybe later someone want add
> something similar like 'reclaimable=true/false', 'dma=true/false', ...
> and I would prefer more general solution. (especially if we'll not
> change tracepoints after release because of backward compatibility)

I would like to quote Steven Rostedt:
https://lore.kernel.org/all/20220515180640.0ae2ead5@gandalf.local.home/
"
> Trace events are not ABI, but if we don't have a strong reason to break it,
> I'd preserve the old order.

Ideally everyone should be using libtraceevent which will parse the format
file for the needed entries.

Nothing (important) should be parsing the raw ascii from the trace files.
It's slow and unreliable. The raw format (trace_pipe_raw) files, along with
libtraceevent will handle fining the fields you are looking for, even if
the fields move around (internally or externally).

Then there's trace-cruncher (a python script that uses libtracefs and
libtraceevent) that will work too.

  https://github.com/vmware/trace-cruncher
"

Thank you,
	Vasily Averin