On Thu, Oct 13, 2022 at 08:05:17AM -0700, Jakub Kicinski wrote: > On Wed, 5 Oct 2022 22:07:57 +0200 Jiri Olsa wrote: > > > Yeah, it's there on linux-next, too. > > > > > > Let me grab a fresh VM and try there. Maybe it's my system. Somehow. > > > > ok, I will look around what's the way to install that centos 8 thing > > Any luck? yea, sorry for delay, I reproduced that.. first objtool warnings ;-) [jolsa@centos8 linux-next]$ make UPD include/generated/compile.h CALL scripts/checksyscalls.sh DESCEND objtool DESCEND bpf/resolve_btfids CC init/version.o AR init/built-in.a AR built-in.a AR vmlinux.a LD vmlinux.o vmlinux.o: warning: objtool: relocate_restore_code+0x3c: relocation to !ENDBR: next_arg+0x8 vmlinux.o: warning: objtool: ___ksymtab+bpf_dispatcher_xdp_func+0x0: data relocation to !ENDBR: bpf_dispatcher_xdp_func+0x0 vmlinux.o: warning: objtool: bpf_dispatcher_xdp+0x20: data relocation to !ENDBR: bpf_dispatcher_xdp_func+0x0 ... Peter, as for objtool warnings, looks like with gcc we'll get endbr64 instruction generated after nops with centos gcc 8.5: ffffffff818d2e20 <bpf_dispatcher_xdp_func>: ffffffff818d2e20: 90 nop ffffffff818d2e21: 90 nop ffffffff818d2e22: 90 nop ffffffff818d2e23: 90 nop ffffffff818d2e24: 90 nop ffffffff818d2e25: f3 0f 1e fa endbr64 ffffffff818d2e29: e9 92 eb 52 00 jmpq ffffffff81e019c0 <__x86_indirect_thunk_rdx> ffffffff818d2e2e: 66 90 xchg %ax,%ax while latest gcc 12 will put it after: ffffffff82736900 <bpf_dispatcher_xdp_func>: ffffffff82736900: f3 0f 1e fa endbr64 ffffffff82736904: 90 nop ffffffff82736905: 90 nop ffffffff82736906: 90 nop ffffffff82736907: 90 nop ffffffff82736908: 90 nop ffffffff82736909: 41 54 push %r12 ffffffff8273690b: 49 89 f4 mov %rsi,%r12 ffffffff8273690e: 55 push %rbp ffffffff8273690f: 48 89 fd mov %rdi,%rbp ffffffff82736912: 53 push %rbx ffffffff82736913: 48 89 d3 mov %rdx,%rbx ffffffff82736916: e8 65 f6 cf fe call ffffffff81435f80 <__sanitizer_cov_trace_pc> ffffffff8273691b: 4c 89 e6 mov %r12,%rsi ffffffff8273691e: 48 89 ef mov %rbp,%rdi ffffffff82736921: 48 89 d8 mov %rbx,%rax ffffffff82736924: 5b pop %rbx ffffffff82736925: 5d pop %rbp ffffffff82736926: 41 5c pop %r12 ffffffff82736928: e9 b3 b6 8c 00 jmp ffffffff83001fe0 <__x86_indirect_thunk_array> any idea where's the problem? bad backport? ;-) in any case (unrelated), I'll check the bpf dispatcher code, I'm not sure the nop update code is aware of the endbr64 instruction jirka
