Em Wed, May 10, 2023 at 02:08:49PM +0100, Alan Maguire escreveu: > On 09/05/2023 22:21, Alexei Starovoitov wrote: > > On Tue, May 9, 2023 at 11:44 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote: > >> On 02/05/2023 04:40, Alexei Starovoitov wrote: > >> won't the pahole v1.25 changes help here; can you try applying > >> https://lore.kernel.org/bpf/1675949331-27935-1-git-send-email-alan.maguire@xxxxxxxxxx/ > >> ...and build using pahole; this should eliminate any functions > >> with inconsistent prototypes via > >> --skip_encoding_btf_inconsistent_proto > >> Do not encode functions with multiple inconsistent prototypes or > >> unexpected register use for their parameters, where the regis‐ > >> ters used do not match calling conventions. > >> I'll check this at my end too. > >> Alexei, if this works should we look at applying the above > >> again to bpf-next? If so I'll resend the patch. > > I've lost the track with pahole fixes. > > Did Arnaldo re-tag pahole 1.25 or released 1.26 with the fixes? > My understanding is the pahole repo is prepped for v1.25, > meaning that it will build with version 1.25 but it is > not officially released yet; Arnaldo will correct me if > I've got that wrong. Right, pahole 1.25 was released with those options/fixes. - Arnaldo > > Alan, > > please submit a fresh patch for bpf-next to enable > > --skip_encoding_btf_inconsistent_proto, so it can go through CI. > > I cannot test all combinations manually. > > > > Done; see [1]. If CI picks up the latest version from > the dwarves repo, it will see version 1.25. > > I've tested the above specific case along with general testing > using latest pahole. > > When running pahole with --verbose and > --skip_encoding_btf_inconsistent_proto > > we see: > > skipping addition of 'mm_init'(mm_init) due to multiple inconsistent > function prototypes > > > ...and > > bpftool btf dump file vmlinux |grep mm_init > > shows the function is not encoded in BTF due to these inconsistencies. > > Thanks! > > Alan > > [1] > https://lore.kernel.org/bpf/20230510130241.1696561-1-alan.maguire@xxxxxxxxxx/
