[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAP4=nvQ6mcka9GAKOYY=M3LWo7qGoXy7GuMB8c2M92si+yMFcw@mail.gmail.com>
Date: Mon, 4 Aug 2025 13:04:27 +0200
From: Tomas Glozar <tglozar@...hat.com>
To: Costa Shulyupin <costa.shul@...hat.com>
Cc: Steven Rostedt <rostedt@...dmis.org>, linux-trace-kernel@...r.kernel.org,
linux-kernel@...r.kernel.org, John Kacur <jkacur@...hat.com>,
Luis Goncalves <lgoncalv@...hat.com>, Arnaldo Carvalho de Melo <acme@...nel.org>, Chang Yin <cyin@...hat.com>
Subject: Re: [PATCH 2/8] rtla/timerlat: Add action on threshold feature
ne 3. 8. 2025 v 13:08 odesÃlatel Costa Shulyupin <costa.shul@...hat.com> napsal:
>
> The term "threshold" is ambiguous. The use of the term is inconsistent
> across the tools. In osnoise top and hist, it means "the minimum delta
> to be considered noise," which conflicts with the semantics of the
> `--on-threshold <action>` option. To avoid confusion, I propose
> introducing "low" and "high" thresholds and updating the sources and
> documentation accordingly.
>
There is already existing naming for that: "minimum delta to be
considered noise" is called "sample threshold" while the timerlat one
is called "stop tracing threshold" (well, the source simply calls the
latter "stop tracing" or "stop"); timerlat also has another threshold,
the stack threshold (specified with -s).
The patchset (well, the v2 of it) has been already merged upstream, so
I'd prefer keeping the naming for rtla-timerlat, and if this is
implemented in rtla-osnoise in the future, a different name can be
used (--on-stop/--on-stop-threshold maybe?). The documentation clearly
says the threshold is the one specified by either -i or -T [1]. What
do you think?
Tomas
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/tools/rtla/common_timerlat_options.rst#n62
Powered by blists - more mailing lists