Re: [PATCH v2] sched/debug: Add new tracepoint to track cpu_capacity

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Qais Yousef <qais.yousef@arm.com>
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>,
	bpf <bpf@vger.kernel.org>, Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>, Phil Auld <pauld@redhat.com>,
	"Peter Zijlstra (Intel)" <peterz@infradead.org>,
	Dietmar Eggemann <dietmar.eggemann@arm.com>,
	vincent.donnefort@arm.com, Ingo Molnar <mingo@redhat.com>,
	vincent.guittot@linaro.org, LKML <linux-kernel@vger.kernel.org>,
	Valentin Schneider <valentin.schneider@arm.com>
Subject: Re: [PATCH v2] sched/debug: Add new tracepoint to track cpu_capacity
Date: Wed, 6 Jan 2021 15:42:13 -0800	[thread overview]
Message-ID: <CAEf4BzaL8788pNdk4A9_EGTZF52MikCPJX1-fh3JO2uca6x9FQ@mail.gmail.com> (raw)
In-Reply-To: <20210106112712.6ec7yejhidauo432@e107158-lin>

On Wed, Jan 6, 2021 at 3:27 AM Qais Yousef <qais.yousef@arm.com> wrote:
>
> On 01/05/21 08:44, Alexei Starovoitov wrote:
> > > Any pointer to an example test I could base this on?
> >
> > selftests/bpf/
>
> I was hoping for something more elaborate. I thought there's something already
> there that do some verification for raw tracepoint that I could either extend
> or replicate. Otherwise this could end up being a time sink for me and I'm not
> keen on jumping down this rabbit hole.

One way would be to add either another custom tracepoint definition to
a test module or modify the existing one to be a bare tracepoint. See
links below.

If it's easy to trigger those tracepoints from user-space on demand,
writing a similar (to module_attach) selftest for in-kernel tracepoint
is trivial.

  [0] https://github.com/torvalds/linux/blob/master/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod-events.h
  [1] https://github.com/torvalds/linux/blob/master/tools/testing/selftests/bpf/progs/test_module_attach.c#L12-L18
  [2] https://github.com/torvalds/linux/blob/master/tools/testing/selftests/bpf/prog_tests/module_attach.c

>
> > > > - add a doc with contents from commit log.
> > >
> > > You're referring to the ABI part of the changelog, right?
> > >
> > > > The "Does bpf make things into an abi ?" question keeps coming back
> > > > over and over again.
> > > > Everytime we have the same answer that No, bpf cannot bake things into abi.
> > > > I think once it's spelled out somewhere in Documentation/ it would be easier to
> > > > repeat this message.
> > >
> > > How about a new Documentation/bpf/ABI.rst? I can write something up initially
> > > for us to discuss in detail when I post.
> >
> > There is Documentation/bpf/bpf_design_QA.rst
> > and we already have this text in there that was added back in 2017:
> >
> > Q: Does BPF have a stable ABI?
> > ------------------------------
> > A: YES. BPF instructions, arguments to BPF programs, set of helper
> > functions and their arguments, recognized return codes are all part
> > of ABI. However there is one specific exception to tracing programs
> > which are using helpers like bpf_probe_read() to walk kernel internal
> > data structures and compile with kernel internal headers. Both of these
> > kernel internals are subject to change and can break with newer kernels
> > such that the program needs to be adapted accordingly.
> >
> > I'm suggesting to add an additional section to this Q/A doc to include
> > more or less
> > the same text you had in the commit log.
>
> Works for me.
>
> Thanks
>
> --
> Qais Yousef