Re: [PATCH 7/9] bpf: Compile the BTF id whitelist data in vmlinux

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Jiri Olsa <jolsa@redhat.com>
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>,
	Jiri Olsa <jolsa@kernel.org>, Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Networking <netdev@vger.kernel.org>, bpf <bpf@vger.kernel.org>,
	Yonghong Song <yhs@fb.com>, Martin KaFai Lau <kafai@fb.com>,
	David Miller <davem@redhat.com>,
	John Fastabend <john.fastabend@gmail.com>,
	Jesper Dangaard Brouer <hawk@kernel.org>,
	Wenbo Zhang <ethercflow@gmail.com>,
	KP Singh <kpsingh@chromium.org>, Andrii Nakryiko <andriin@fb.com>,
	Brendan Gregg <bgregg@netflix.com>,
	Florent Revest <revest@chromium.org>,
	Al Viro <viro@zeniv.linux.org.uk>
Subject: Re: [PATCH 7/9] bpf: Compile the BTF id whitelist data in vmlinux
Date: Fri, 29 May 2020 13:48:58 -0700	[thread overview]
Message-ID: <CAEf4BzbM-5-_QzDhrJDFJefo-m0OWDhvjsK_F1vA-ja4URVE9Q@mail.gmail.com> (raw)
In-Reply-To: <20200528172349.GA506785@krava>

On Thu, May 28, 2020 at 10:24 AM Jiri Olsa <jolsa@redhat.com> wrote:
>
> On Thu, May 14, 2020 at 03:46:26PM -0700, Andrii Nakryiko wrote:
>
> SNIP
>
> > > I was thinking of putting the names in __init section and generate the BTF
> > > ids on kernel start, but the build time generation seemed more convenient..
> > > let's see the linking times with 'real size' whitelist and we can reconsider
> > >
> >
> > Being able to record such places where to put BTF ID in code would be
> > really nice, as Alexei mentioned. There are many potential use cases
> > where it would be good to have BTF IDs just put into arbitrary
> > variables/arrays. This would trigger compilation error, if someone
> > screws up the name, or function is renamed, or if function can be
> > compiled out under some configuration. E.g., assuming some reasonable
> > implementation of the macro
>
> hi,
> I'm struggling with this part.. to get some reasonable reference
> to function/name into 32 bits? any idea? ;-)
>

Well, you don't have to store actual pointer, right? E.g, emitting
something like this in assembly:

.global __BTF_ID___some_function
.type __BTF_ID___some_function, @object
.size __BTF_ID___some_function, 4
__BTF_ID___some_function:
.zero  4

Would reserve 4 bytes and emit __BTF_ID___some_function symbol. If we
can then post-process vmlinux image and for all symbols starting with
__BTF_ID___ find some_function BTF type id and put it into those 4
bytes, that should work, no?

Maybe generalize it to __BTF_ID__{func,struct,typedef}__some_function,
whatever, not sure. Just an idea.

> jirka
>
> >
> > static const u32 d_path_whitelist[] = {
> >     BTF_ID_FUNC(vfs_fallocate),
> > #ifdef CONFIG_WHATEVER
> >     BTF_ID_FUNC(do_truncate),
> > #endif
> > };
> >
> > Would be nice and very explicit. Given this is not going to be sorted,
> > you won't be able to use binary search, but if whitelists are
> > generally small, it should be fine as is. If not, hashmap could be
> > built in runtime and would be, probably, faster than binary search for
> > longer sets of BTF IDs.
> >
> > I wonder if we can do some assembly magic with generating extra
> > symbols and/or relocations to achieve this? What do you think? Is it
> > doable/desirable/better?
> >
> >
> > > thanks,
> > > jirka
> > >
> >
>