Re: [PATCH bpf-next] docs/bpf: add llvm_reloc.rst to explain llvm bpf relocations

[Yonghong Song <yhs@fb.com>]

On 5/24/21 1:33 AM, Lorenz Bauer wrote:
> On Sat, 22 May 2021 at 17:44, Yonghong Song <yhs@fb.com> wrote:
>>
>> Daniel, John and Lorenz,
>>
>> Could you help check how the new relocation scheme
>> may impact you? libbpf has a similar issue and is fixed by
>>     https://lore.kernel.org/bpf/20210522162341.3687617-1-yhs@fb.com/
>> In most cases, you should just change relocation enum number,
>> no relocation resolution is changed.
>>
>> Please let me know. Thanks!
> 
> Thank you for the heads up :) cilium/ebpf currently doesn't look at
> relocation types at all for better or worse. We simply collect
> "well-known" sections like maps, programs, etc. and only process
> relocations for these. So your change won't break cilium/ebpf, but it

Thanks for confirmation. So yes, you should be fine though.

> makes me wonder whether we should check the relocation type.

If the library is used in a control environment, e.g., the object
file is generated by llvm, bpftool linker, etc. You should be fine.
But yes, checking relocation types will make the library more robust.

> 
> Best
> Lorenz
>