On Wed, 2023-11-01 at 15:40 -0700, Andrii Nakryiko wrote: > On Wed, Nov 1, 2023 at 2:27 AM Lorenz Bauer <lorenz.bauer@xxxxxxxxxxxxx> wrote: > > > > On Tue, Oct 31, 2023 at 6:38 PM Andrii Nakryiko > > <andrii.nakryiko@xxxxxxxxx> wrote: > > > > > > I don't remember if this is intention or not, but the main part is > > > adjusting CO-RE relocation, the actual instruction value is less > > > important. But this is happening after static linking, because BTF is > > > deduplicated (there is a duplication in BTF generated by Clang). > > > > Ah I see! And the deduplication is done by libbpf during linking? So > > yes > > > far, we've been validating that the instruction immediate matches what > > is in ext_infos. Should I just stop doing that? > > probably, because I just checked libbpf's linker code, I don't think > we adjust instructions that have CO-RE relocations. We might probably > add that, but it's basically just BTF_TYPE_ID_LOCAL that would need > this special handling. If someone sends the patch I'll accept it :) > > > > > > There are at least two identical prototypes (which is strange and > > > might be worth looking into from Clang side). > > > > That would be good! > > Agreed, maybe Yonghong or Eduard can take a look when they get time? I'll take a look, probably on Sat/Sun.
