MIME-Version: 1.0
References: <20230201135737.800527-1-jolsa@kernel.org> <Y9vSZhBBCbshI3eM@casper.infradead.org>
 <Y9vX49CtDzyg3B/8@krava>
In-Reply-To: <Y9vX49CtDzyg3B/8@krava>
From:   Hao Luo <haoluo@google.com>
Date:   Wed, 8 Feb 2023 23:12:34 -0800
Message-ID: <CA+khW7juLEcrTOd7iKG3C_WY8L265XKNo0iLzV1fE=o-cyeHcQ@mail.gmail.com>
Subject: Re: [RFC 0/5] mm/bpf/perf: Store build id in file object
To:     Jiri Olsa <olsajiri@gmail.com>
Cc:     Matthew Wilcox <willy@infradead.org>,
        Alexei Starovoitov <ast@kernel.org>,
        Andrii Nakryiko <andrii@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        bpf@vger.kernel.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-perf-users@vger.kernel.org, Martin KaFai Lau <kafai@fb.com>,
        Song Liu <songliubraving@fb.com>, Yonghong Song <yhs@fb.com>,
        John Fastabend <john.fastabend@gmail.com>,
        KP Singh <kpsingh@chromium.org>,
        Stanislav Fomichev <sdf@google.com>,
        Daniel Borkmann <daniel@iogearbox.net>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Feb 2, 2023 at 7:33 AM Jiri Olsa <olsajiri@gmail.com> wrote:
>
> On Thu, Feb 02, 2023 at 03:10:30PM +0000, Matthew Wilcox wrote:
> > On Wed, Feb 01, 2023 at 02:57:32PM +0100, Jiri Olsa wrote:
> > > hi,
> > > we have a use cases for bpf programs to use binary file's build id.
> >
> > What is your use case?  Is it some hobbyist thing or is it something
> > that distro kernels are all going to enable?
> >
>
> our use case is for hubble/tetragon [1] and we are asked to report
> buildid of executed binary.. but the monitoring process is running
> in its own pod and can't access the the binaries outside of it, so
> we need to be able to read it in kernel
>
> I understand Hao Luo has also use case for that [2]
>

Sorry for the late reply.

We use BPF to profile stacktraces and build id is more useful than
instruction addresses. However, sometimes we need to record
stacktraces from an atomic context. In that case, if the page that
contains the build id string is not in the page cache, we would fail
to get build id. Storing the build id in file object solves this
problem and helps us get build id more reliably.

> jirka
>
>
> [1] https://github.com/cilium/tetragon/
> [2] https://lore.kernel.org/bpf/CA+khW7gAYHmoUkq0UqTiZjdOqARLG256USj3uFwi6z_FyZf31w@mail.gmail.com/