Re: [PATCH 01/11] objtool: Generic annotation infrastructure

[Peter Zijlstra]

On Fri, Nov 08, 2024 at 03:16:00PM +0100, Peter Zijlstra wrote:

> From an LLVM=-19 build we can see that:
> 
> $ readelf -WS tmp-build/arch/x86/kvm/vmx/vmenter.o | grep annotate
>   [13] .discard.annotate PROGBITS        0000000000000000 00028c 000018 08   M  0   0  1
> 
> $ readelf -WS tmp-build/arch/x86/kvm/kvm-intel.o | grep annotate
>   [ 3] .discard.annotate PROGBITS        0000000000000000 069fe0 0089d0 00   M  0   0  1
> 
> Which tells us that the translation unit itself has a sh_entsize of 8,
> while the linked object has sh_entsize of 0.
> 
> This then completely messes up the indexing objtool does, which relies
> on it being a sane number.
> 
> GCC/binutils very much does not do this, it retains the 8.

Anyway, for now I've added:

+       if (sec->sh.sh_entsize != 8) {
+               static bool warn = false;
+               if (!warn) {
+                       WARN("%s: dodgy linker, sh_entsize != 8", sec->name);
+                       warn = true;
+               }
+               sec->sh.sh_entsize = 8;
+       }

To objtool, this allows it function correctly and prints this reminder
to for us to figure out the linker story.