As mentioned by Arnaldo in a past discussion (Link:
https://lore.kernel.org/linux-perf-users/Z2m0pWiQgZNXhvxK@x1/), the
current documentation on --force-btf is vague and inaccurate, this patch
aims to provide a clearer explanation of the option.

Signed-off-by: Howard Chu <howardchu95@gmail.com>
Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
---
 tools/perf/Documentation/perf-trace.txt | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt
index 887dc37773d0..cf2445d2ba3d 100644
--- a/tools/perf/Documentation/perf-trace.txt
+++ b/tools/perf/Documentation/perf-trace.txt
@@ -246,10 +246,10 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs.
 	arguments to strings (pid to comm, syscall id to syscall name, etc).
 
 --force-btf::
-	Use btf_dump to pretty print syscall argument data, instead of using hand-crafted pretty
-	printers. This option is intended for testing BTF integration in perf trace. btf_dump-based
-	pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can
-	better pretty-print integer flags and struct pointers.
+	Use libbpf's btf_dump to pretty print syscall arguments, utilizing only the debug
+	information from BTF. This option is intended for testing BTF integration in perf
+	trace. Using libbpf is a fallback to perf trace's own prettifiers generated from the
+	kernel source, as the latter can better pretty-print integer flags and struct members.
 
 PAGEFAULTS
 ----------
-- 
2.45.2

Hello,

On Wed, Mar 26, 2025 at 09:04:16AM -0700, Howard Chu wrote:
> As mentioned by Arnaldo in a past discussion (Link:
> https://lore.kernel.org/linux-perf-users/Z2m0pWiQgZNXhvxK@x1/), the
> current documentation on --force-btf is vague and inaccurate, this patch
> aims to provide a clearer explanation of the option.
> 
> Signed-off-by: Howard Chu <howardchu95@gmail.com>
> Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
> ---
>  tools/perf/Documentation/perf-trace.txt | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt
> index 887dc37773d0..cf2445d2ba3d 100644
> --- a/tools/perf/Documentation/perf-trace.txt
> +++ b/tools/perf/Documentation/perf-trace.txt
> @@ -246,10 +246,10 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs.
>  	arguments to strings (pid to comm, syscall id to syscall name, etc).
>  
>  --force-btf::
> -	Use btf_dump to pretty print syscall argument data, instead of using hand-crafted pretty
> -	printers. This option is intended for testing BTF integration in perf trace. btf_dump-based
> -	pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can
> -	better pretty-print integer flags and struct pointers.
> +	Use libbpf's btf_dump to pretty print syscall arguments, utilizing only the debug
> +	information from BTF. This option is intended for testing BTF integration in perf
> +	trace. Using libbpf is a fallback to perf trace's own prettifiers generated from the
> +	kernel source, as the latter can better pretty-print integer flags and struct members.

"the latter" means our own prettifier, right?  Then why do we need the
fallback as it's better and always works?

Thanks,
Namhyung

>  
>  PAGEFAULTS
>  ----------
> -- 
> 2.45.2
>

Hello Namhyung,

On Wed, Mar 26, 2025 at 11:14 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Hello,
>
> On Wed, Mar 26, 2025 at 09:04:16AM -0700, Howard Chu wrote:
> > As mentioned by Arnaldo in a past discussion (Link:
> > https://lore.kernel.org/linux-perf-users/Z2m0pWiQgZNXhvxK@x1/), the
> > current documentation on --force-btf is vague and inaccurate, this patch
> > aims to provide a clearer explanation of the option.
> >
> > Signed-off-by: Howard Chu <howardchu95@gmail.com>
> > Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
> > ---
> >  tools/perf/Documentation/perf-trace.txt | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt
> > index 887dc37773d0..cf2445d2ba3d 100644
> > --- a/tools/perf/Documentation/perf-trace.txt
> > +++ b/tools/perf/Documentation/perf-trace.txt
> > @@ -246,10 +246,10 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs.
> >       arguments to strings (pid to comm, syscall id to syscall name, etc).
> >
> >  --force-btf::
> > -     Use btf_dump to pretty print syscall argument data, instead of using hand-crafted pretty
> > -     printers. This option is intended for testing BTF integration in perf trace. btf_dump-based
> > -     pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can
> > -     better pretty-print integer flags and struct pointers.
> > +     Use libbpf's btf_dump to pretty print syscall arguments, utilizing only the debug
> > +     information from BTF. This option is intended for testing BTF integration in perf
> > +     trace. Using libbpf is a fallback to perf trace's own prettifiers generated from the
> > +     kernel source, as the latter can better pretty-print integer flags and struct members.
>
> "the latter" means our own prettifier, right?  Then why do we need the
> fallback as it's better and always works?

Honestly, for debugging—if BTF exists, it always works, yes. I don't
know if libbpf is 'better' because our prettifiers are prettier :),
yet there are so few of them.

Thanks,
Howard

>
> Thanks,
> Namhyung
>
> >
> >  PAGEFAULTS
> >  ----------
> > --
> > 2.45.2
> >