Re: [PATCH] perf symbol: Add LoongArch case in get_plt_sizes()

[Huacai Chen]

On Mon, May 22, 2023 at 3:59 PM Tiezhu Yang <yangtiezhu@xxxxxxxxxxx> wrote:
>
>
>
> On 05/22/2023 11:50 AM, Huacai Chen wrote:
> > Hi, Arnaldo,
> >
> > On Thu, May 18, 2023 at 8:16 PM Arnaldo Carvalho de Melo
> > <acme@xxxxxxxxxx> wrote:
> >>
> >> Em Thu, May 18, 2023 at 12:05:53PM +0800, Leo Yan escreveu:
> >>> On Thu, May 18, 2023 at 11:57:29AM +0800, Huacai Chen wrote:
> >>>> On Thu, May 18, 2023 at 11:21 AM Leo Yan <leo.yan@xxxxxxxxxx> wrote:
> >>>>>
> >>>>> On Thu, May 18, 2023 at 11:12:26AM +0800, Huacai Chen wrote:
> >>>>>> On Thu, May 18, 2023 at 11:06 AM Leo Yan <leo.yan@xxxxxxxxxx> wrote:
> >>>>>>>
> >>>>>>> On Thu, May 18, 2023 at 10:11:27AM +0800, Huacai Chen wrote:
> >>>>>>>> Queued, thanks.
> >>>>>>>
> >>>>>>> The patch is fine for me.
> >>>>>>>
> >>>>>>> Should not perf patches are to be merged via Arnaldo's tree?
> >>>>>>
> >>>>>> I think both are OK, if Arnaldo takes this patch, I will drop it.
> >>>>>
> >>>>> A good practice is to firstly inquiry the maintainers.
> >>>>>
> >>>>> AFAIK, Arnaldo will test perf patches before sending out pull request;
> >>>>> if perf patches are scattered out, it might be out of the testing
> >>>>> radar.
> >>>> OK, I know, thank you very much.
> >>>
> >>> You are welcome!
> >>>
> >>> I found the code base for bfd:
> >>> https://github.com/bminor/binutils-gdb/blob/master/bfd/elfnn-loongarch.c
> >>>
> >>> And this patch is consistent with above link, FWIW:
> >>>
> >>> Reviewed-by: Leo Yan <leo.yan@xxxxxxxxxx>
> >>
> >> Thanks, applied.
> > I'm very sorry that this patch breaks cross-build. We need some
> > additional modification.
> >
> > diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c
> > index 5d409c26a22e..b3dbf6ca99a7 100644
> > --- a/tools/perf/util/symbol-elf.c
> > +++ b/tools/perf/util/symbol-elf.c
> > @@ -35,6 +35,10 @@
> >  #define EM_AARCH64     183  /* ARM 64 bit */
> >  #endif
> >
> > +#ifndef EM_LOONGARCH
> > +#define EM_LOONGARCH   258
> > +#endif
> > +
> >  #ifndef ELF32_ST_VISIBILITY
> >  #define ELF32_ST_VISIBILITY(o) ((o) & 0x03)
> >  #endif
> >
> > Then, drop this patch to get an updated version, or let me send an
> > incremental patch?
> >
>
> I tested this patch on native LoongArch and x86 system, I did not
> hit the build error about undeclared EM_LOONGARCH both on LoongArch
> and x86, because EM_LOONGARCH is defined in /usr/include/elf.h
>
> Here is the x86 system info:
>
> [root@fedora yangtiezhu]# cat /etc/fedora-release
> Fedora release 38 (Thirty Eight)
> [root@fedora yangtiezhu]# uname -m
> x86_64
> [root@fedora yangtiezhu]# grep -rn -w EM_LOONGARCH /usr/include/
> /usr/include/linux/elf-em.h:54:#define EM_LOONGARCH    258    /*
> LoongArch */
> /usr/include/linux/audit.h:442:#define AUDIT_ARCH_LOONGARCH32
> (EM_LOONGARCH|__AUDIT_ARCH_LE)
> /usr/include/linux/audit.h:443:#define AUDIT_ARCH_LOONGARCH64
> (EM_LOONGARCH|__AUDIT_ARCH_64BIT|__AUDIT_ARCH_LE)
> /usr/include/elf.h:361:#define EM_LOONGARCH    258    /* LoongArch */
> [root@fedora yangtiezhu]# rpm -qf /usr/include/elf.h
> glibc-headers-x86-2.37-1.fc38.noarch
>
> If I am missing something, please let me know.
Fedora 38 has a very new toolchain, older distribution will meet build errors.

Huacai
>
> Anyway, it is not a bad thing to add the EM_LOONGARCH definition
> to avoid the build error on some systems which have no EM_LOONGARCH
> in the glibc header.
>
> Thanks,
> Tiezhu
>
>