On Sat, Apr 1, 2023 at 5:27 AM Alyssa Ross <hi@xxxxxxxxx> wrote:
>
> On Sat, Apr 01, 2023 at 12:42:13AM +0900, Masahiro Yamada wrote:
> > On Fri, Mar 31, 2023 at 7:29 AM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> > >
> > > On Thu, Mar 30, 2023 at 06:22:24PM +0000, Alyssa Ross wrote:
> > > > Since 32ef9e5054ec, -Wa,-gdwarf-2 is no longer used in KBUILD_AFLAGS.
> > > > Instead, it includes -g, the appropriate -gdwarf-* flag, and also the
> > > > -Wa versions of both of those if building with Clang and GNU as. As a
> > > > result, debug info was being generated for the purgatory objects, even
> > > > though the intention was that it not be.
> > > >
> > > > Fixes: 32ef9e5054ec ("Makefile.debug: re-enable debug info for .S files")
> > > > Signed-off-by: Alyssa Ross <hi@xxxxxxxxx>
> > > > Cc: stable@xxxxxxxxxxxxxxx
> > > > Acked-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > >
> > > This is definitely more future proof.
> > >
> > > Reviewed-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> > > Tested-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> >
> >
> >
> > I prefer v3 since it is cleaner, but unfortunately
> > it does not work for Clang+GAS.
> >
> >
> > With v3 applied, I still see the debug info.
> >
> >
> >
> > $ make LLVM=1 LLVM_IAS=0 arch/x86/purgatory/setup-x86_64.o
> > UPD include/config/kernel.release
> > UPD include/generated/utsrelease.h
> > CALL scripts/checksyscalls.sh
> > DESCEND objtool
> > INSTALL libsubcmd_headers
> > AS arch/x86/purgatory/setup-x86_64.o
> > $ readelf -S arch/x86/purgatory/setup-x86_64.o
> > There are 18 section headers, starting at offset 0x14d8:
> >
> > Section Headers:
> > [Nr] Name Type Address Offset
> > Size EntSize Flags Link Info Align
> > [ 0] NULL 0000000000000000 00000000
> > 0000000000000000 0000000000000000 0 0 0
> > [ 1] .text PROGBITS 0000000000000000 00000040
> > 0000000000000027 0000000000000000 AX 0 0 16
> > [ 2] .rela.text RELA 0000000000000000 000012f8
> > 0000000000000060 0000000000000018 I 15 1 8
> > [ 3] .data PROGBITS 0000000000000000 00000067
> > 0000000000000000 0000000000000000 WA 0 0 1
> > [ 4] .bss NOBITS 0000000000000000 00001000
> > 0000000000001000 0000000000000000 WA 0 0 4096
> > [ 5] .rodata PROGBITS 0000000000000000 00001000
> > 0000000000000020 0000000000000000 A 0 0 16
> > [ 6] .rela.rodata RELA 0000000000000000 00001358
> > 0000000000000018 0000000000000018 I 15 5 8
> > [ 7] .debug_line PROGBITS 0000000000000000 00001020
> > 000000000000005f 0000000000000000 0 0 1
> > [ 8] .rela.debug_line RELA 0000000000000000 00001370
> > 0000000000000018 0000000000000018 I 15 7 8
> > [ 9] .debug_info PROGBITS 0000000000000000 0000107f
> > 0000000000000027 0000000000000000 0 0 1
> > [10] .rela.debug_info RELA 0000000000000000 00001388
> > 0000000000000090 0000000000000018 I 15 9 8
> > [11] .debug_abbrev PROGBITS 0000000000000000 000010a6
> > 0000000000000014 0000000000000000 0 0 1
> > [12] .debug_aranges PROGBITS 0000000000000000 000010c0
> > 0000000000000030 0000000000000000 0 0 16
> > [13] .rela.debug_[...] RELA 0000000000000000 00001418
> > 0000000000000030 0000000000000018 I 15 12 8
> > [14] .debug_str PROGBITS 0000000000000000 000010f0
> > 0000000000000054 0000000000000001 MS 0 0 1
> > [15] .symtab SYMTAB 0000000000000000 00001148
> > 0000000000000168 0000000000000018 16 12 8
> > [16] .strtab STRTAB 0000000000000000 000012b0
> > 0000000000000041 0000000000000000 0 0 1
> > [17] .shstrtab STRTAB 0000000000000000 00001448
> > 000000000000008d 0000000000000000 0 0 1
> > Key to Flags:
> > W (write), A (alloc), X (execute), M (merge), S (strings), I (info),
> > L (link order), O (extra OS processing required), G (group), T (TLS),
> > C (compressed), x (unknown), o (OS specific), E (exclude),
> > D (mbind), l (large), p (processor specific)
> >
> >
> >
> >
> >
> >
> > With -g0 given, GCC stops passing -g -gdwarf-4 down to GAS.
> >
> >
> > Clang does not do anything about -g0 for the external assembler.
>
> You're right. Thank you for your thoughtful testing — I forgot to check
> LLVM for v3. I thought maybe adding -Wa,-g0 would be enough, but it
> turns out that GAS doesn't support that. If -g has been specified,
> there doesn't seem to be any way to disable debug info again later in
> the command line.
>
> So we probably can't do better than v2 while LLVM_IAS=0 is supported.
> The only other option I see is (untested):
>
> asflags-y += -g0
> asflags-remove-y += -Wa,-g -Wa,-gdwarf-4 -Wa,-gdwarf-5
>
> But I don't like that option, because it means there are two completely
> different ways of doing it depending on the compiler setup, and it makes
> it even less likely anybody would remember to update asflags-remove-y
> when DWARF 6 comes around or whatever.
Agree.
This is uglier than v2.
If nobody comes up with a better idea,
I will pick up v2.
--
Best Regards
Masahiro Yamada
