On Tue, Apr 11, 2023 at 11:44 AM Fangrui Song <maskray@xxxxxxxxxx> wrote:
>
> On Fri, Apr 7, 2023 at 11:35 AM Nick Desaulniers
> <ndesaulniers@xxxxxxxxxx> wrote:
> >
> > On Fri, Mar 24, 2023 at 2:22 PM Fangrui Song <maskray@xxxxxxxxxx> wrote:
> > >
> > > Commit 27f2a4db76e8 ("Makefile: fix GDB warning with CONFIG_RELR")
> > > added --use-android-relr-tags to fix a GDB warning
> > >
> > > BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'
> > >
> > > The GDB warning has been fixed in version 11.2.
> > >
> > > The DT_ANDROID_RELR tag was deprecated since DT_RELR was standardized.
> > > Thus, --use-android-relr-tags should be removed. While making the
> > > change, try -z pack-relative-relocs, which is supported since LLD 15.
> > > Keep supporting --pack-dyn-relocs=relr as well for older LLD versions.
> > >
> > > As of today, GNU ld supports the latter option for x86 and powerpc64
> > > ports and has no intention to support --pack-dyn-relocs=relr. In the
> > > absence of the glibc symbol version GLIBC_ABI_DT_RELR,
> > > --pack-dyn-relocs=relr and -z pack-relative-relocs are identical in
> > > ld.lld.
> > >
> > > Link: https://github.com/ClangBuiltLinux/linux/issues/1057
> > > Link: https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=a619b58721f0a03fd91c27670d3e4c2fb0d88f1e
> > > Signed-off-by: Fangrui Song <maskray@xxxxxxxxxx>
> >
> > Thanks v2 looks better. IIUC, this will first try to test+use
> > `--pack-dyn-relocs=relr` in preference to `-z pack-relative-relocs`.
> > Do we want to reorder the preference, for both the test and actual
> > flag used?
>
> Thanks for the comment. The order is deliberate and there is no
> obsolescence indication of --pack-dyn-relocs=relr.
>
> GNU ld and newer lld report warnings (instead of errors) for unknown
> -z options, and only errors lead to non-zero exit codes.
That makes sense; consider adding a note about that intention to the
commit message in a v3 along with:
Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
>
> Therefore, $(call ld-option,--pack-dyn-relocs=relr,-z
> pack-relative-relocs) has to place --pack-dyn-relocs=relr first.
> scripts/tools-support-relr.sh has to test --pack-dyn-relocs=relr first as well.
>
> > > ---
> > > Makefile | 3 ++-
> > > scripts/tools-support-relr.sh | 8 ++++++--
> > > 2 files changed, 8 insertions(+), 3 deletions(-)
> > > ---
> > > Changes from v1:
> > > * Keep supporting --pack-dyn-relocs=relr for older ld.lld versions
> > >
> > > diff --git a/Makefile b/Makefile
> > > index a2c310df2145..e23a85476d5d 100644
> > > --- a/Makefile
> > > +++ b/Makefile
> > > @@ -1113,7 +1113,8 @@ LDFLAGS_vmlinux += -X
> > > endif
> > >
> > > ifeq ($(CONFIG_RELR),y)
> > > -LDFLAGS_vmlinux += --pack-dyn-relocs=relr --use-android-relr-tags
> > > +# ld.lld before 15 did not support -z pack-relative-relocs.
> > > +LDFLAGS_vmlinux += $(call ld-option,--pack-dyn-relocs=relr,-z pack-relative-relocs)
> > > endif
> > >
> > > # We never want expected sections to be placed heuristically by the
> > > diff --git a/scripts/tools-support-relr.sh b/scripts/tools-support-relr.sh
> > > index cb55878bd5b8..4c121946e517 100755
> > > --- a/scripts/tools-support-relr.sh
> > > +++ b/scripts/tools-support-relr.sh
> > > @@ -7,8 +7,12 @@ trap "rm -f $tmp_file.o $tmp_file $tmp_file.bin" EXIT
> > > cat << "END" | $CC -c -x c - -o $tmp_file.o >/dev/null 2>&1
> > > void *p = &p;
> > > END
> > > -$LD $tmp_file.o -shared -Bsymbolic --pack-dyn-relocs=relr \
> > > - --use-android-relr-tags -o $tmp_file
> > > +
> > > +# ld.lld before 15 did not support -z pack-relative-relocs.
> > > +if ! $LD $tmp_file.o -shared -Bsymbolic --pack-dyn-relocs=relr -o $tmp_file 2>/dev/null; then
> > > + $LD $tmp_file.o -shared -Bsymbolic -z pack-relative-relocs -o $tmp_file 2>&1 |
> > > + grep -q pack-relative-relocs && exit 1
> > > +fi
> > >
> > > # Despite printing an error message, GNU nm still exits with exit code 0 if it
> > > # sees a relr section. So we need to check that nothing is printed to stderr.
> > > --
> > > 2.40.0.348.gf938b09366-goog
> > >
> >
> >
> > --
> > Thanks,
> > ~Nick Desaulniers
>
>
>
> --
> 宋方睿
--
Thanks,
~Nick Desaulniers
