From: Lorenz Bauer <lmb@cloudflare.com>
To: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: bpf <bpf@vger.kernel.org>, Andrii Nakryiko <andrii@kernel.org>,
Florent Revest <revest@chromium.org>
Subject: Re: Portability of bpf_tracing.h
Date: Fri, 28 May 2021 09:29:53 +0100 [thread overview]
Message-ID: <CACAyw9-mHGrvrWozqngJ8X4qzqxB8Yku+AaL_Rv8RZhLXPRwJQ@mail.gmail.com> (raw)
In-Reply-To: <CAEf4BzZftL2q9qAoeXsO87-Wx9AbF8A1mLnBAtBrGo=XSx996g@mail.gmail.com>
On Wed, 26 May 2021 at 19:34, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
>
> So I did a bit of investigation and gathered struct pt_regs
> definitions from all the "supported" architectures in bpf_tracing.h.
> I'll leave it here for further reference.
>
> static unsigned long bpf_pt_regs_parm1(const void *regs)
> {
> if (___arch_is_x86)
> return ((struct pt_regs___x86 *)regs)->di;
> else if (___arch_is_s390)
> return ((struct pt_regs___s390 *)regs)->gprs[2];
> else if (___arch_is_powerpc)
> return ((struct pt_regs___powerpc *)regs)->gpr[3];
> else
> while(1); /* need some better way to force BPF verification failure */
> }
>
> And so on for other architectures and other helpers, you should get
> the idea from the above.
The idea of basing this on unique fields in types is neat, the
downside I see is that we encode the logic in the BPF bitstream. If in
the future struct pt_regs is changed, code breaks and we can't do much
about it. What if instead we replace ___arch_is_x86, etc. with a
.kconfig style constant load? The platform detection logic can then
live in libbpf or cilium/ebpf and can be evolved if needed. Instead of
while(1) we could use an illegal function call, like we do for
poisoned CORE relocations.
>
> As a shameless plug, if you'd like to see some more examples of using
> CO-RE for detecting kernel features, see [0]
>
> [0] https://nakryiko.com/posts/bpf-tips-printk/
>
> > > Well, obviously I'm not a fan of even more magic #defines. But I think
> > > we can achieve a similar effect with a more "lazy" approach. I.e., if
> > > user tries to use PT_REGS_xxx macros but doesn't specify the platform
> > > -- only then it gets compilation errors. There is stuff in
> > > bpf_tracing.h that doesn't need pt_regs, so we can't just outright do
> > > #error unconditinally. But we can do something like this:
> > >
> > > #else /* !bpf_target_defined */
> > >
> > > #define PT_REGS_PARM1(x) _Pragma("GCC error \"blah blah something
> > > user-facing\"")
> > >
> > > ... and so on for all macros
> > >
> > > #endif
> > >
> > > Thoughts?
> >
> > That would work for me, but it would change the behaviour for current
> > users of the header, no? That's why I added the magic define in the
> > first place.
>
> How so? If someone is using PT_REGS_PARM1 without setting target arch
> they should get compilation error about undefined macro. Here it will
> be the same thing, only if someone tries to use PT_REGS_PARM1() will
> they reach that _Pragma.
>
> Or am I missing something?
Right! Doing this makes sense regardless of the outcome of our discussion above.
--
Lorenz Bauer | Systems Engineer
6th Floor, County Hall/The Riverside Building, SE1 7PB, UK
www.cloudflare.com
next prev parent reply other threads:[~2021-05-28 8:30 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-24 15:05 Portability of bpf_tracing.h Lorenz Bauer
2021-05-24 17:47 ` Andrii Nakryiko
2021-05-24 19:30 ` John Fastabend
2021-05-25 0:13 ` Andrii Nakryiko
2021-05-26 9:13 ` Lorenz Bauer
2021-05-26 18:34 ` Andrii Nakryiko
2021-05-28 8:29 ` Lorenz Bauer [this message]
2021-05-30 0:51 ` Andrii Nakryiko
2021-06-10 14:09 ` Lorenz Bauer
2021-06-10 18:14 ` Alexei Starovoitov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CACAyw9-mHGrvrWozqngJ8X4qzqxB8Yku+AaL_Rv8RZhLXPRwJQ@mail.gmail.com \
--to=lmb@cloudflare.com \
--cc=andrii.nakryiko@gmail.com \
--cc=andrii@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=revest@chromium.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).