linux-kernel - Re: [PATCH] tools/rtla: Add basic test suite

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: <CAP4=nvR4an1nzBvPyYv3YBY0Cu=fNwTkEN0_q63JxNAi6799mg@mail.gmail.com>
Date: Mon, 20 Jan 2025 08:38:44 +0100
From: Tomas Glozar <tglozar@...hat.com>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: linux-trace-kernel@...r.kernel.org, linux-kernel@...r.kernel.org, 
	John Kacur <jkacur@...hat.com>, Luis Goncalves <lgoncalv@...hat.com>, Chang Yin <cyin@...hat.com>, 
	Qiao Zhao <qzhao@...hat.com>
Subject: Re: [PATCH] tools/rtla: Add basic test suite

pá 17. 1. 2025 v 22:02 odesílatel Steven Rostedt <rostedt@...dmis.org> napsal:
>
> Now, what I also noticed was. When it did crash and die. The osnoise tracer
> no longer creates any osnoise threads. Can you take a look.
>

It's likely that you have NO_OSNOISE_WORKLOAD set. The default value
is OSNOISE_WORKLOAD; rtla relies on it not being set, and rtla
timerlat -u resets it to the original value after it ends, so usually,
everything works. But it breaks either when you set
NO_OSNOISE_WORKLOAD manually or when rtla timerlat -u crashes (and
fails to set it back).

> So that failure puts the system into a state that can no longer run osnoise.
> It takes a reboot to fix it.
>
> That should be fixed where a reboot is not needed.

I sent patches to fix this for timerlat [1] and as I commented in the
thread [2], the same has to be done for osnoise, too. I just haven't
sent that yet since I was working on the rtla hang in sample
collection bug [3]. I also discovered that OSNOISE_WORKLOAD is not the
only option that rtla relies on being set properly: stop_tracing_us is
also affected and possibly other options. rtla should set all osnoise
options properly on every run, not relying on them being set by the
cleanup of the previous rtla run.

[1] https://lore.kernel.org/linux-trace-kernel/20250107144823.239782-1-tglozar@redhat.com/T/
[2[ https://lore.kernel.org/linux-trace-kernel/CAP4=nvRCw_2hKPUyd7obZv2Xho_mjdSC7EmozFsoeS0OZC9xaA@mail.gmail.com/
[3] https://lore.kernel.org/linux-trace-kernel/20250116144931.649593-1-tglozar@redhat.com/

Tomas