[PATCH v3 2/5] perf build: Generate arm64's sysreg-defs.h and add to include path

[Ian Rogers]

On Mon, Nov 6, 2023 at 10:10 PM Ian Rogers <irogers at google.com> wrote:
>
> On Wed, Oct 18, 2023 at 7:12 AM Arnaldo Carvalho de Melo
> <acme at kernel.org> wrote:
> >
> > Em Tue, Oct 17, 2023 at 03:23:40PM -0700, Namhyung Kim escreveu:
> > > Hello,
> > >
> > > On Wed, Oct 11, 2023 at 12:58 PM Oliver Upton <oliver.upton at linux.dev> wrote:
> > > >
> > > > Start generating sysreg-defs.h in anticipation of updating sysreg.h to a
> > > > version that needs the generated output.
> > > >
> > > > Signed-off-by: Oliver Upton <oliver.upton at linux.dev>
> > >
> > > It seems we also need this on non-ARM archs to process ARM SPE data.
> > >
> > > Acked-by: Namhyung Kim <namhyung at kernel.org>
> >
> > When building with CORESIGHT=1, yes.
> >
> > I have it in my tests and:
> >
> > ⬢[acme at toolbox perf-tools-next]$ ls -la /tmp/build/perf-tools-next/util/arm-spe.o
> > -rw-r--r--. 1 acme acme 135432 Oct 17 16:49 /tmp/build/perf-tools-next/util/arm-spe.o
> > ⬢[acme at toolbox perf-tools-next]$ ldd /tmp/build/perf-tools-next/perf | grep csd
> >         libopencsd_c_api.so.1 => /lib64/libopencsd_c_api.so.1 (0x00007f36bfca5000)
> >         libopencsd.so.1 => /lib64/libopencsd.so.1 (0x00007f36be2e0000)
> > ⬢[acme at toolbox perf-tools-next]$ rpm -qf /lib64/libopencsd.so.1
> > opencsd-1.3.3-1.fc38.x86_64
> > ⬢[acme at toolbox perf-tools-next]$ rpm -q --qf "%{summary}\n" opencsd
> > An open source CoreSight(tm) Trace Decode library
> > ⬢[acme at toolbox perf-tools-next]$
> >
> > Well, double checked and arm-spe.o is built by default, only way to
> > disable it is using NO_AUXTRACE=1 in the make command line, but then
> > IIRC one needs linking with opencsd to decode all those traces, right?
> >
> > Anyway:
> >
> > Acked-by: Arnaldo Carvalho de Melo <acme at redhat.com>
> >
> > - Arnaldo
> >
> > > Thanks,
> > > Namhyung
> > >
> > >
> > > > ---
> > > >  tools/perf/Makefile.perf | 15 +++++++++++++--
> > > >  tools/perf/util/Build    |  2 +-
> > > >  2 files changed, 14 insertions(+), 3 deletions(-)
> > > >
> > > > diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> > > > index 37af6df7b978..14dedd11a1f5 100644
> > > > --- a/tools/perf/Makefile.perf
> > > > +++ b/tools/perf/Makefile.perf
> > > > @@ -443,6 +443,15 @@ drm_ioctl_tbl := $(srctree)/tools/perf/trace/beauty/drm_ioctl.sh
> > > >  # Create output directory if not already present
> > > >  _dummy := $(shell [ -d '$(beauty_ioctl_outdir)' ] || mkdir -p '$(beauty_ioctl_outdir)')
> > > >
> > > > +arm64_gen_sysreg_dir := $(srctree)/tools/arch/arm64/tools
> > > > +
> > > > +arm64-sysreg-defs: FORCE
> > > > +       $(Q)$(MAKE) -C $(arm64_gen_sysreg_dir)
>
> Should this not build an install_headers target? The generated code is
> going into the source tree as is, ignoring O= options to make.

I think on top of this, tools/perf/MANIFEST needs to have:
arch/arm64/tools/gen-sysreg.awk
arch/arm64/tools/sysreg
This will add these files to the release tar balls, it already
contains things like scripts/bpf_doc.py.
tools/arch/arm64/tools/Makefile is picked up by tools/arch being in
the MANIFEST.

Thanks,
Ian