Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20201221070029.1451176-1-namhyung@kernel.org> <20201221070029.1451176-3-namhyung@kernel.org>
 <20201228115006.GB450923@krava>
In-Reply-To: <20201228115006.GB450923@krava>
From:   Namhyung Kim <namhyung@kernel.org>
Date:   Tue, 29 Dec 2020 14:35:07 +0900
Message-ID: <CAM9d7ci4YapWyRK-tujg0THmgK6Ys9jCE7NSPHX1pP3rN4rt+w@mail.gmail.com>
Subject: Re: [PATCH 2/3] perf tools: Skip MMAP record synthesis for kernel threads
To:     Jiri Olsa <jolsa@redhat.com>
Cc:     Arnaldo Carvalho de Melo <acme@kernel.org>,
        Ingo Molnar <mingo@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Alexander Shishkin <alexander.shishkin@linux.intel.com>,
        LKML <linux-kernel@vger.kernel.org>,
        Stephane Eranian <eranian@google.com>,
        Andi Kleen <ak@linux.intel.com>,
        Ian Rogers <irogers@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Dec 28, 2020 at 8:50 PM Jiri Olsa <jolsa@redhat.com> wrote:
>
> On Mon, Dec 21, 2020 at 04:00:28PM +0900, Namhyung Kim wrote:
> > To synthesize information to resolve sample IPs, it needs to scan task
> > and mmap info from the /proc filesystem.  For each process, it
> > opens (and reads) status and maps file respectively.  But as kernel
> > threads don't have memory maps so we can skip the maps file.
> >
> > To find kernel threads, check "VmPeak:" line in /proc/<PID>/status
> > file.  It's about the peak virtual memory usage so only user-level
> > tasks have that.  Also check "Threads:" line (which follows the VmPeak
> > line whether or not it exists) to be sure it's read enough data - just
> > in case of deeply nested pid namespaces or large number of
> > supplementary groups are involved.
>
> how much does this save? for kernel threads the maps file is empty
> and we just open the file and close
>
> also perhaps just stat(".....maps") and check the size would be easier?

The numbers are in the cover letter and it's around 5% on an idle
machine which has mostly kernel threads.  I think most of the win
came from this change.

It's just to avoid those syscalls, so I wanted to use the available
info in the status file.

Thanks,
Namhyung