Re: [PATCH v5 04/12] fprobe: Use ftrace_regs in fprobe entry handler

[Jiri Olsa <olsajiri@gmail.com>]

On Mon, Sep 25, 2023 at 09:15:15PM +0900, Masami Hiramatsu wrote:
> Hi Jiri,
> 
> On Mon, 25 Sep 2023 12:41:59 +0200
> Jiri Olsa <olsajiri@gmail.com> wrote:
> 
> > On Sun, Sep 24, 2023 at 10:36:36PM +0900, Masami Hiramatsu (Google) wrote:
> > > From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> > > 
> > > This allows fprobes to be available with CONFIG_DYNAMIC_FTRACE_WITH_ARGS
> > > instead of CONFIG_DYNAMIC_FTRACE_WITH_REGS, then we can enable fprobe
> > > on arm64.
> > > 
> > > Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> > > Acked-by: Florent Revest <revest@chromium.org>
> > 
> > I was getting bpf selftests failures with this patchset and when
> > bisecting I'm getting crash when running on top of this change
> 
> Thanks for bisecting!
> 
> > 
> > looks like it's missing some of the regs NULL checks added later?
> 
> yeah, if the RIP (arch_rethook_prepare+0x0/0x30) is correct, 
> 
> void arch_rethook_prepare(struct rethook_node *rh, struct ftrace_regs *fregs, bool mcount)
> 
> RSI (the 2nd argument) is NULL. This means fregs == NULL and caused the crash.
> I think ftrace_get_regs(fregs) for the entry handler may return NULL.
> 
> Ah, 
> 
> @@ -182,7 +182,7 @@ static void fprobe_init(struct fprobe *fp)
>  		fp->ops.func = fprobe_kprobe_handler;
>  	else
>  		fp->ops.func = fprobe_handler;
> -	fp->ops.flags |= FTRACE_OPS_FL_SAVE_REGS;
> +	fp->ops.flags |= FTRACE_OPS_FL_SAVE_ARGS;
>  }
>  
>  static int fprobe_init_rethook(struct fprobe *fp, int num)
> 
> This may cause the issue, it should keep REGS at this point (this must be done in
> [9/12]). But after applying [9/12], it shouldn't be a problem... 
> 
> Let me check it again.

that helped with the crash, I'll continue bisecting to find out
where it breaks the tests

thanks,
jirka