On Wed, Nov 10, 2021 at 2:11 PM Song Liu <songliubraving@xxxxxx> wrote: > > > > > On Nov 10, 2021, at 2:02 PM, Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote: > > > > On Wed, Nov 10, 2021 at 9:47 AM Song Liu <songliubraving@xxxxxx> wrote: > >> > >> -#ifdef CONFIG_DEBUG_INFO_BTF > >> -BTF_ID_LIST_GLOBAL(btf_sock_ids) > >> +BTF_ID_LIST_GLOBAL(btf_sock_ids, MAX_BTF_SOCK_TYPE) > >> #define BTF_SOCK_TYPE(name, type) BTF_ID(struct, type) > >> BTF_SOCK_TYPE_xxx > >> #undef BTF_SOCK_TYPE > >> -#else > >> -u32 btf_sock_ids[MAX_BTF_SOCK_TYPE]; > >> -#endif > > > > If we're trying to future proof it I think it would be better > > to combine it with MAX_BTF_SOCK_TYPE and BTF_SOCK_TYPE_xxx macro. > > (or have another macro that is tracing specific). > > That will help avoid cryptic btf_task_struct_ids[0|1|2] > > references in the code. > > Yeah, this makes sense. > > I am taking time off for tomorrow and Friday, so I probably won't > have time to implement this before 5.16-rc1. How about we ship > this fix as-is, and improve it later? It's not rc1 material. It's in bpf-next only. There is no rush, I think.
