linux-kernel - Re: [PATCH] profiling: remove prof_cpu

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

Message-ID: <CAHk-=whRa0rQrdJc8SQMkC=CzzonzZk5QnZq7BgqD3+vKP+qWA@mail.gmail.com>
Date: Mon, 29 Jul 2024 12:36:06 -0700
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>, 
	Anna-Maria Gleixner <anna-maria@...utronix.de>, Andrew Morton <akpm@...ux-foundation.org>, 
	Arnd Bergmann <arnd@...db.de>, Mel Gorman <mgorman@...hsingularity.net>, 
	Michal Hocko <mhocko@...e.com>, Peter Zijlstra <peterz@...radead.org>, Vlastimil Babka <vbabka@...e.cz>, 
	Ingo Molnar <mingo@...nel.org>, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] profiling: remove prof_cpu_mask

On Sat, 27 Jul 2024 at 14:20, Thomas Gleixner <tglx@...utronix.de> wrote:
>
> Only people who indulge in nostalgia will notice :)

So let's see. Any bets on whether anybody actually notices?

I took Tetsuo's fix for the syzbot issue, and then I did a "remove all
the flip buffer code" in commit 2accfdb7eff6 ("profiling: attempt to
remove per-cpu profile flip buffer").

If somebody ends up having a real use-case for this old horrid code,
we may just have to remove it.

But it might also be the case that somebody actually does want the
boot-time profiling, and then the runtime overhead is annoying just
because they are on a multi-socket machine and the profiling just
keeps going - even after better profilers are available.

So it might be that nobody wants to actually re-instate the flip
buffer thing, but instead just turn the thing off entirely.

Technically you can do that by writing to /proc/profile with a
"profiling multiplier" that effectively turns it off, but very few
architectures actually support that (see "setup_profiling_timer()").

End result: maybe we should add a way to just say "I'm done profiling
now" if somebody reports that it causes performance issues after boot.

But I hope (and think it's very possible) that nobody will ever notice
any other way than from following this LKML discussion.

              Linus