[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20230720064028.1aeb3c18@gandalf.local.home>
Date: Thu, 20 Jul 2023 06:40:28 -0400
From: Steven Rostedt <rostedt@...dmis.org>
To: Tony Luck <tony.luck@...el.com>
Cc: Aristeu Rozanski <aris@...vo.org>, linux-kernel@...r.kernel.org
Subject: Re: rasdaemon broke between v6.0 and v6.3?
On Wed, 19 Jul 2023 16:35:54 -0700
Tony Luck <tony.luck@...el.com> wrote:
> [resend as plain text - sorry for the earlier HTML]
>
> An internal team is seeing tests that worked on v6.0 fail on v6.3. The problem is that
> rasdaemon isn’t waking up to process the “mce_record” trace events.
>
> Manually checking for them works:
>
> root@...251:/sys/kernel/debug/tracing>systemctl stop rasdaemon
> root@...251:/sys/kernel/debug/tracing>
> root@...251:/sys/kernel/debug/tracing>
> root@...251:/sys/kernel/debug/tracing>echo 1 > events/mce/mce_record/enable
> root@...251:/sys/kernel/debug/tracing>
> root@...251:/sys/kernel/debug/tracing>cat trace_pipe
> <...>-235 [000] ..... 596.892583: mce_record: CPU: 0, MCGc/s: f000c15/0, MC13: 8c00004200800090, IPID: 0000000000000000, ADDR/MISC/SYND: 0000000123450000/08000a80c2982086/0000000000000000, RIP: 00:<0000000000000000>, TSC: 14120b051a1, PROCESSOR: 0:c06f1, TIME: 1689802780, SOCKET: 0, APIC: 0
> kworker/0:2-235 [000] ..... 597.204343: mce_record: CPU: 0, MCGc/s: f000c15/0, MC255: 9c0000000000009f, IPID: 0000000000000000, ADDR/MISC/SYND: 0000000123450000/000000000000008c/0000000000000000, RIP: 00:<0000000000000000>, TSC: 0, PROCESSOR: 0:c06f1, TIME: 1689802781, SOCKET: 0, APIC: 0
>
> So their tests are injecting errors, and the trace event is firing.
>
> Is there some updated version of rasdaemon needed?
>
> Some kernel CONFIG option problem?
>
A bug was fixed that I think affected rasdaemon.
commit 3e46d910d8acf94e5360126593b68bf4fee4c4a1
Author: Shiju Jose <shiju.jose@...wei.com>
Date: Thu Feb 2 18:23:09 2023 +0000
tracing: Fix poll() and select() do not work on per_cpu trace_pipe and trace_pipe_raw
Make sure /sys/kernel/tracing/buffer_percent = 0
-- Steve
Powered by blists - more mailing lists