Re: [PATCH resend ] tracepoint: Print the function symbol when tracepoint_debug is set

[Mathieu Desnoyers]

On 2025-03-06 10:13, Steven Rostedt wrote:
> On Thu, 6 Mar 2025 11:37:38 +0800
> Shijie Huang <shijie@xxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
>
> > diff --git a/kernel/tracepoint.c b/kernel/tracepoint.c
> > index abfd0ac1177f..5a5041f32cc8 100644
> > --- a/kernel/tracepoint.c
> > +++ b/kernel/tracepoint.c
> > @@ -127,7 +127,8 @@ static void debug_print_probes(struct
> > tracepoint_func *funcs)
> >                   return;
> >
> >           for (i = 0; funcs[i].func; i++)
> > -               printk(KERN_DEBUG "Probe %d : %p\n", i, funcs[i].func);
> > +               printk(KERN_DEBUG "Probe %d : [ %p ] %pS\n",
> > +                               i, funcs[i].func, funcs[i].func);
> >    }
> >
> >
> > The output will look like:
> >
> > [   63.818829] Probe 0 : [ 0000000032848d41 ]
> > perf_trace_sched_wakeup_template+0x0/0x20
> > [   63.819287] Probe 0 : [ 00000000fe8cca4d ]
> > perf_trace_sched_switch+0x0/0x20
> > [   65.325638] Probe 0 : [ 0000000032848d41 ]
> > perf_trace_sched_wakeup_template+0x0/0x20
> > [   65.695631] Probe 0 : [ 00000000fe8cca4d ]
> > perf_trace_sched_switch+0x0/0x20
>
> I'm fine either way. Mathieu, what's your preference?
>
> Although the above shows a hashed %p. Is that even useful?

The hashed %p is not really useful.

I think we could just print %Sb

e.g. from core-api/printk-formats.rst:

        %pSb    versatile_init+0x0/0x110 [module_name ed5019fdf5e53be37cb1ba7899292d7e143b259e]

AFAIU, when the kallsyms_lookup_buildid fails, it prints the raw address instead
(see __sprint_symbol).

Thanks,

Mathieu

--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com