linux-kernel - BUG? DOC: Really jiffies in /proc/schedstat?

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>] [day] [month] [year] [list]

Message-ID: <CAOTQ16LL_W+RqDWQLzsvnRaYGT_9E9EMN_9hdr-vie8hp83gAA@mail.gmail.com>
Date:   Fri, 22 Jun 2018 17:09:12 +0200
From:   Thommy Jakobsson <thommyj@...il.com>
To:     linux-kernel@...r.kernel.org
Subject: BUG? DOC: Really jiffies in /proc/schedstat?

Hi,

I am measuring cpu runtime for a couple of processes, and is using
/proc/schedstat and /proc/[pid]/schedstat to do so. According to the
schedstat doc
(https://www.kernel.org/doc/Documentation/scheduler/sched-stats.txt) the last
three values are in jiffies. Doing something like

root@...4-base-pd-b-rev1:~#  cat /proc/schedstat  | grep cpu3; sleep
2; cat /proc/schedstat  | grep cpu3;
cpu3 0 0 0 0 0 0 41155818663320 17173365985560 1011370906
cpu3 0 0 0 0 0 0 41157858979020 17176095339120 1011430425

on a heavy loaded system seems to suggest that at least the sum of running tasks
is in nanoseconds, not jiffies. Which I also think looks correct when
checking the
code.

For /proc/[pid]/schedstat the documentation is a bit vague on unit,
but I interpret it
to be the same as the corresponding global values. Which would mean jiffies.
Also here it looks to be in nanoseconds on my system though.

Am I just confused? is it configuration/system dependent? or is the
doc just plain
wrong?

Thanks,
Thommy Jakobsson