[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240821160316.02c03c44@gandalf.local.home>
Date: Wed, 21 Aug 2024 16:03:16 -0400
From: Steven Rostedt <rostedt@...dmis.org>
To: tglozar@...hat.com
Cc: linux-trace-kernel@...r.kernel.org, linux-kernel@...r.kernel.org,
jkacur@...hat.com, "Luis Claudio R. Goncalves" <lgoncalv@...hat.com>
Subject: Re: [PATCH] tracing/timerlat: Check tlat_var for NULL in
timerlat_fd_release
On Tue, 20 Aug 2024 15:00:01 +0200
tglozar@...hat.com wrote:
> From: Tomas Glozar <tglozar@...hat.com>
>
> When running timerlat with a userspace workload (NO_OSNOISE_WORKLOAD),
> NULL pointer dereference can be triggered by sending consequent SIGINT
> and SIGTERM signals to the workload process. That then causes
> timerlat_fd_release to be called twice in a row, and the second time,
> hrtimer_cancel is called on a zeroed hrtimer struct, causing the NULL
> dereference.
>
> This can be reproduced using rtla:
> ```
> $ while true; do rtla timerlat top -u -q & PID=$!; sleep 5; \
> kill -INT $PID; sleep 0.001; kill -TERM $PID; wait $PID; done
> [1] 1675
> [1]+ Aborted (SIGTERM) rtla timerlat top -u -q
> [1] 1688
> client_loop: send disconnect: Broken pipe
> ```
> triggering the bug:
I'm able to reproduce this with the above. Unfortunately, I can still
reproduce it after applying this patch :-(
Looking at the code, the logic for handling the kthread seems off. I'll
spend a little time to see if I can figure it out.
-- Steve
Powered by blists - more mailing lists