On Mon, Jul 11, 2022 at 11:16:21AM +0200, Jiri Olsa wrote: > On Wed, Jun 29, 2022 at 06:44:52PM -0300, Arnaldo Carvalho de Melo wrote: > > Em Wed, Jun 29, 2022 at 12:12:13AM -0700, Yonghong Song escreveu: > > > Add support for enum64. For 64-bit enumerator value, > > > previously, the value is truncated into 32bit, e.g., > > > for the following enum in linux uapi bpf.h, > > > enum { > > > BPF_F_INDEX_MASK = 0xffffffffULL, > > > BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, > > > /* BPF_FUNC_perf_event_output for sk_buff input context. */ > > > BPF_F_CTXLEN_MASK = (0xfffffULL << 32), > > > }; > > > > Applied, added the entry for skip generating enums to the man page, > > added support to the pahole BTF loader, used the new pahole to build > > bpf-next/master, all seems ok, pushing to next on git.kernel.org so that > > the libbpf github CI can give it a go. > > > > To build with torvalds/master one has to add --skip_encoding_btf_enum64, > > I think, haven't tested with it, without it isn't working, libbpf > > complains at that btfids tool. > > > > Please check/test what is in there now: > > > > git://git.kernel.org/pub/scm/devel/pahole/pahole.git next > > https://git.kernel.org/pub/scm/devel/pahole/pahole.git/log/?h=next > > > > Unless someone screams I plan pushing out a new release, update fedora > > packages, etc early next week its overdue by now. > > I used this new pahole in kernel build and it looks ok, > but I'm getting following warning: > > BTFIDS vmlinux > WARN: resolve_btfids: unresolved symbol mptcp_sock > > might be specific to my .config, I'll check and let you know ok I have only FWD declaration of mptcp_sock struct, that's why it can't get resolved.. when I enabled CONFIG_MPTCP it goes away so no problem with pahole ;-) Tested-by: Jiri Olsa <jolsa@xxxxxxxxxx> the problem is with the btf_sock_ids array that carries mptcp_sock BTF id even when the CONFIG_MPTCP option is not enabled I'll follow up on that in separate email thanks, jirka