Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <169139090386.324433.6412259486776991296.stgit@devnote2>
 <169139093899.324433.3739544465888158332.stgit@devnote2> <CABRcYmK6X6okNKNu9ZjgLEO+JMGL42j7idE8QPZ_EpYA9S9UZQ@mail.gmail.com>
 <20230809234318.08784e46d0b7d88c1bccedbe@kernel.org>
In-Reply-To: <20230809234318.08784e46d0b7d88c1bccedbe@kernel.org>
From:   Florent Revest <revest@chromium.org>
Date:   Wed, 9 Aug 2023 17:45:29 +0200
Message-ID: <CABRcYmLe05UiK+-mCq5LA0d1Xomdpb+R_5A5HLBLbuBqfBCwUA@mail.gmail.com>
Subject: Re: [RFC PATCH v2 3/6] fprobe: rethook: Use fprobe_regs in fprobe
 exit handler and rethook
To:     Masami Hiramatsu <mhiramat@kernel.org>
Cc:     Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        linux-trace-kernel@vger.kernel.org,
        LKML <linux-kernel@vger.kernel.org>,
        Martin KaFai Lau <martin.lau@linux.dev>,
        bpf <bpf@vger.kernel.org>, Sven Schnelle <svens@linux.ibm.com>,
        Alexei Starovoitov <ast@kernel.org>,
        Jiri Olsa <jolsa@kernel.org>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Alan Maguire <alan.maguire@oracle.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Aug 9, 2023 at 4:43=E2=80=AFPM Masami Hiramatsu <mhiramat@kernel.or=
g> wrote:
>
> > I think there are two things that can be meant with "rethook uses ftrac=
e_regs":
> >
> > - rethook callbacks receive a ftrace_regs (that's what you do further d=
own)
> > - rethook can hook to a traced function using a ftrace_regs (that's
> > what you use in fprobe now)
> >
> > But I think the second proposition shouldn't imply that rethook_hook
> > can _only_ hook to ftrace_regs. For the kprobe use case, I think there
> > should also be a rethook_hook_pt_regs() that operates on a pt_regs. We
> > could have a default implementation of rethook_hook that calls into
> > the other (or vice versa) on HAVE_FTRACE_REGS_COMPATIBLE_WITH_PT_REGS
> > but I think it's good to separate these two APIs
>
> Yeah, so for simplying the 2nd case, I added this dependency.
>
> diff --git a/arch/Kconfig b/arch/Kconfig
> index aff2746c8af2..e321bdb8b22b 100644
> --- a/arch/Kconfig
> +++ b/arch/Kconfig
> @@ -201,6 +201,7 @@ config KRETPROBE_ON_RETHOOK
>         def_bool y
>         depends on HAVE_RETHOOK
>         depends on KRETPROBES
> +       depends on HAVE_PT_REGS_COMPAT_FTRACE_REGS || !HAVE_DYNAMIC_FTRAC=
E_WITH_ARGS
>         select RETHOOK
>
> This is the point why I said that "do not remove kretprobe trampoline".
> If there is arch dependent kretprobe trampoline, kretprobe does not use
> the rethook for hooking return. And eventually I would like to remove
> kretprobe itself (replace it with fprobe + rethook). If so, I don't want
> to pay more efforts on this part, and keep kretprobe on rethook as it is.

What are your thoughts on kprobe + rethook though ?

If that's something you think is worth having, then in this case, it
seems that having a rethook_hook_pt_regs() API would help users.

If that's a frankenstein use case you don't want to support then I
agree we can live without this API and get away with the cast
protected by the depends on HAVE_PT_REGS_COMPAT_FTRACE_REGS...