[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20240604080856.907194a1fc3106be0cc94fa0@kernel.org>
Date: Tue, 4 Jun 2024 08:08:56 +0900
From: Masami Hiramatsu (Google) <mhiramat@...nel.org>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org, Mark
Rutland <mark.rutland@....com>, Mathieu Desnoyers
<mathieu.desnoyers@...icios.com>, Andrew Morton
<akpm@...ux-foundation.org>, Alexei Starovoitov
<alexei.starovoitov@...il.com>, Florent Revest <revest@...omium.org>,
Martin KaFai Lau <martin.lau@...ux.dev>, bpf <bpf@...r.kernel.org>, Sven
Schnelle <svens@...ux.ibm.com>, Alexei Starovoitov <ast@...nel.org>, Jiri
Olsa <jolsa@...nel.org>, Arnaldo Carvalho de Melo <acme@...nel.org>, Daniel
Borkmann <daniel@...earbox.net>, Alan Maguire <alan.maguire@...cle.com>,
Peter Zijlstra <peterz@...radead.org>, Thomas Gleixner
<tglx@...utronix.de>, Guo Ren <guoren@...nel.org>
Subject: Re: [PATCH v2 24/27] function_graph: Use static_call and branch to
optimize entry function
On Mon, 3 Jun 2024 11:07:52 -0400
Steven Rostedt <rostedt@...dmis.org> wrote:
> On Mon, 3 Jun 2024 11:00:18 -0400
> Steven Rostedt <rostedt@...dmis.org> wrote:
>
> > Yes, but that gets a bit complex, and requires the changing of all archs.
> > If it starts to become a problem, I rather add that as a feature. That is,
> > we can always go back to it. But for now, lets keep the complexity down.
>
> And if we were to go the route of calling a single fgraph_ops caller, I
> would want it choreographed with ftrace, such that the direct caller calls
> a different fgraph function only if it has only one graph caller on it and
> the fgraph loop function if a function has more than one. Just like the
> ftrace code does.
>
> If we are going to go that route, let's do it right.
Yes, that is what I expect. ftrace_ops has a callback for loop and subops
has another direct callbacks, and the ftrace_ops has a flag to direct subops
assign. In this case ftrace will assign the subops's direct callback for
single subops entries.
But anyway, this optimization can be done afterwards. Especially this feature
is important for fprobes on fgraph, until that, the current implementation is
enough.
Thank you,
>
> -- Steve
--
Masami Hiramatsu (Google) <mhiramat@...nel.org>
Powered by blists - more mailing lists