Re: [RFC bpf-next 0/5] bpf, x86: Add bpf_get_func_ip helper

From: Jiri Olsa <jolsa@redhat.com>
To: Alan Maguire <alan.maguire@oracle.com>
Cc: Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andriin@fb.com>,
	netdev@vger.kernel.org, bpf@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>,
	Masami Hiramatsu <mhiramat@kernel.org>
Subject: Re: [RFC bpf-next 0/5] bpf, x86: Add bpf_get_func_ip helper
Date: Fri, 2 Jul 2021 10:16:48 +0200	[thread overview]
Message-ID: <YN7LcJu73nCz3Ips@krava> (raw)
In-Reply-To: <alpine.LRH.2.23.451.2107011819160.27594@localhost>

On Thu, Jul 01, 2021 at 06:22:45PM +0100, Alan Maguire wrote:
> On Tue, 29 Jun 2021, Jiri Olsa wrote:
> 
> > hi,
> > adding bpf_get_func_ip helper that returns IP address of the
> > caller function for trampoline and krobe programs.
> > 
> > There're 2 specific implementation of the bpf_get_func_ip
> > helper, one for trampoline progs and one for kprobe/kretprobe
> > progs.
> > 
> > The trampoline helper call is replaced/inlined by verifier
> > with simple move instruction. The kprobe/kretprobe is actual
> > helper call that returns prepared caller address.
> > 
> > The trampoline extra 3 instructions for storing IP address
> > is now optional, which I'm not completely sure is necessary,
> > so I plan to do some benchmarks, if it's noticeable, hence
> > the RFC. I'm also not completely sure about the kprobe/kretprobe
> > implementation.
> > 
> > Also available at:
> >   https://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
> >   bpf/get_func_ip
> > 
> > thanks,
> > jirka
> > 
> >
> 
> This is great Jiri! Feel free to add for the series:
> 
> Tested-by: Alan Maguire <alan.maguire@oracle.com>

great, thanks for testing

> 
> BTW I also verified that if we extend bpf_program__attach_kprobe() to
> support the function+offset format in the func_name argument for kprobes, 
> the following test will pass too:
> 
> __u64 test5_result = 0;
> SEC("kprobe/bpf_fentry_test5+0x6")
> int test5(struct pt_regs *ctx)
> {
>         __u64 addr = bpf_get_func_ip(ctx);
> 
>         test5_result = (const void *) addr == (&bpf_fentry_test5 + 0x6);
>         return 0;
> }

right, I did not think of this test, I'll add it

thanks,
jirka