Re: [PATCH v2 00/20] objtool: vmlinux.o and CLANG LTO support

From: Sami Tolvanen
Date: Mon Jan 25 2021 - 17:46:47 EST


On Fri, Jan 22, 2021 at 6:26 PM Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:
>
> On Fri, Jan 22, 2021 at 05:32:43PM -0800, Nick Desaulniers wrote:
> > > In this specific case, find_func_by_offset returns NULL for
> > > .text..L.cfi.jumptable.43 at addend 0x8, because Clang doesn't emit
> > > jump table symbols for static functions:
> > >
> > > 0000000000000000 <__typeid__ZTSFjmiE_global_addr>:
> > > 0: e9 00 00 00 00 jmpq 5 <__typeid__ZTSFjmiE_global_addr+0x5>
> > > 1: R_X86_64_PLT32 io_serial_in-0x4
> > > 5: cc int3
> > > 6: cc int3
> > > 7: cc int3
> > > 8: e9 00 00 00 00 jmpq d <__typeid__ZTSFjmiE_global_addr+0xd>
> > > 9: R_X86_64_PLT32 mem32_serial_in-0x4
> > > d: cc int3
> > > e: cc int3
> > > f: cc int3
> > >
> > > Nick, do you remember if there were plans to change this?
> >
> > Do you have a link to any previous discussion to help jog my mind; I
> > don't really remember this one.
> >
> > Is it that `__typeid__ZTSFjmiE_global_addr` is the synthesized jump
> > table, and yet there is no `__typeid__ZTSFjmiE_global_addr` entry in
> > the symbol table?
>
> I think he means there's not a 'mem32_serial_in.cfi_jt' symbol at
> '__typeid__ZTSFjmiE_global_addr+8'. Probably more aggressive symbol
> pruning from the assembler.

Correct, the compiler is not emitting mem32_serial_in.cfi_jt here. I
seem to recall that the missing jump table symbols also made stack
traces harder to follow (__typeid__ZTSFjmiE_global_addr+8 is not very
readable), so ideally they shouldn't be pruned.

> It's fine though. I just need to rewrite the CFI support a bit.
>
> But that can come later. For now I'll just drop the two CFI-related
> patches from this set so I can merge the others next week.

Sure, sounds good.

Sami