linux-kernel - Re: [PATCH v11 00/14] unwind_user: x86: Deferred unwinding infrastructure

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: <aGVoAT6asYLUx4He@google.com>
Date: Wed, 2 Jul 2025 10:10:25 -0700
From: Namhyung Kim <namhyung@...nel.org>
To: Sam James <sam@...too.org>
Cc: fweimer@...hat.com, akpm@...ux-foundation.org, andrii@...nel.org,
	axboe@...nel.dk, beaub@...ux.microsoft.com, bpf@...r.kernel.org,
	indu.bhagat@...cle.com, jemarch@....org, jolsa@...nel.org,
	jpoimboe@...nel.org, jremus@...ux.ibm.com,
	linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org,
	mathieu.desnoyers@...icios.com, mhiramat@...nel.org,
	mingo@...nel.org, peterz@...radead.org, rostedt@...dmis.org,
	tglx@...utronix.de, torvalds@...ux-foundation.org, x86@...nel.org
Subject: Re: [PATCH v11 00/14] unwind_user: x86: Deferred unwinding
 infrastructure

Hello,

On Wed, Jul 02, 2025 at 12:44:51PM +0100, Sam James wrote:
> I started to play around with this properly last night and it was
> straightforward, fortunately.
> 
> Did initially attempt to backport to 6.15 but it was a victim of some
> mm refactoring and didn't seem worth to carry on w/ that route.
> 
> Started a rough page with notes for myself (but corrections & such
> welcome) at https://wiki.gentoo.org/wiki/Project:Toolchain/SFrame but
> honestly, it's immediately obvious (and beautiful) when it's working
> correctly. I've used Namhyung Kim's example from this thread but you can
> see it easily with `perf top -g` too.

I've looked at the page but it doesn't seem to work well unfortunately.
The working case has the symbols correct but the overheads are same.
It should have more 'Children' overhead than 'Self' like in the broken
case because 'children = self + callchain'.

Thanks,
Namhyung

> 
> In one of the commit messages in the perf series, Steven also gave `perf
> record -g -vv true` which was convenient for making sure it's correctly
> discovered deferred unwinding support.
> 
> I plan on doing measurements next and doing some more playing once I've
> built more userland with it.