On Mon, Jan 11, 2021 at 5:34 PM Yonghong Song <yhs@xxxxxx> wrote: > > > > On 1/11/21 1:37 PM, Andrii Nakryiko wrote: > > On Sun, Jan 10, 2021 at 8:15 PM Yonghong Song <yhs@xxxxxx> wrote: > >> > >> > >> > >> On 1/8/21 2:09 PM, Andrii Nakryiko wrote: > >>> Add support for searching for ksym externs not just in vmlinux BTF, but across > >>> all module BTFs, similarly to how it's done for CO-RE relocations. Kernels > >>> that expose module BTFs through sysfs are assumed to support new ldimm64 > >>> instruction extension with BTF FD provided in insn[1].imm field, so no extra > >>> feature detection is performed. > >>> > >>> Signed-off-by: Andrii Nakryiko <andrii@xxxxxxxxxx> > >>> --- > >>> tools/lib/bpf/libbpf.c | 47 +++++++++++++++++++++++++++--------------- > >>> 1 file changed, 30 insertions(+), 17 deletions(-) > >>> > >>> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c > >>> index 6ae748f6ea11..57559a71e4de 100644 > >>> --- a/tools/lib/bpf/libbpf.c > >>> +++ b/tools/lib/bpf/libbpf.c > >>> @@ -395,7 +395,8 @@ struct extern_desc { > >>> unsigned long long addr; > >>> > >>> /* target btf_id of the corresponding kernel var. */ > >>> - int vmlinux_btf_id; > >>> + int kernel_btf_obj_fd; > >>> + int kernel_btf_id; > >>> > >>> /* local btf_id of the ksym extern's type. */ > >>> __u32 type_id; > >>> @@ -6162,7 +6163,8 @@ bpf_object__relocate_data(struct bpf_object *obj, struct bpf_program *prog) > >>> } else /* EXT_KSYM */ { > >>> if (ext->ksym.type_id) { /* typed ksyms */ > >>> insn[0].src_reg = BPF_PSEUDO_BTF_ID; > >>> - insn[0].imm = ext->ksym.vmlinux_btf_id; > >>> + insn[0].imm = ext->ksym.kernel_btf_id; > >>> + insn[1].imm = ext->ksym.kernel_btf_obj_fd; > >>> } else { /* typeless ksyms */ > >>> insn[0].imm = (__u32)ext->ksym.addr; > >>> insn[1].imm = ext->ksym.addr >> 32; > >>> @@ -7319,7 +7321,8 @@ static int bpf_object__read_kallsyms_file(struct bpf_object *obj) > >>> static int bpf_object__resolve_ksyms_btf_id(struct bpf_object *obj) > >>> { > >>> struct extern_desc *ext; > >>> - int i, id; > >>> + struct btf *btf; > >>> + int i, j, id, btf_fd, err; > >>> > >>> for (i = 0; i < obj->nr_extern; i++) { > >>> const struct btf_type *targ_var, *targ_type; > >>> @@ -7331,8 +7334,22 @@ static int bpf_object__resolve_ksyms_btf_id(struct bpf_object *obj) > >>> if (ext->type != EXT_KSYM || !ext->ksym.type_id) > >>> continue; > >>> > >>> - id = btf__find_by_name_kind(obj->btf_vmlinux, ext->name, > >>> - BTF_KIND_VAR); > >>> + btf = obj->btf_vmlinux; > >>> + btf_fd = 0; > >>> + id = btf__find_by_name_kind(btf, ext->name, BTF_KIND_VAR); > >>> + if (id == -ENOENT) { > >>> + err = load_module_btfs(obj); > >>> + if (err) > >>> + return err; > >>> + > >>> + for (j = 0; j < obj->btf_module_cnt; j++) { > >>> + btf = obj->btf_modules[j].btf; > >>> + btf_fd = obj->btf_modules[j].fd; > >> > >> Do we have possibility btf_fd == 0 here? > > > > Extremely unlikely. But if we are really worried about 0 fd, we should > > handle that in a centralized fashion in libbpf. I.e., for any > > operation that can return FD, check if that FD is 0, and if yes, dup() > > it. And then make everything call that helper. So in the context of > > this patch I'm just ignoring such possibility. > Maybe at least add some comments here to document such a possibility? sure, will add > > > > >> > >>> + id = btf__find_by_name_kind(btf, ext->name, BTF_KIND_VAR); > >>> + if (id != -ENOENT) > >>> + break; > >>> + } > >>> + } > >>> if (id <= 0) { > >>> pr_warn("extern (ksym) '%s': failed to find BTF ID in vmlinux BTF.\n", > >>> ext->name); > >>> @@ -7343,24 +7360,19 @@ static int bpf_object__resolve_ksyms_btf_id(struct bpf_object *obj) > >>> local_type_id = ext->ksym.type_id; > >>> > >>> /* find target type_id */ > >>> - targ_var = btf__type_by_id(obj->btf_vmlinux, id); > >>> - targ_var_name = btf__name_by_offset(obj->btf_vmlinux, > >>> - targ_var->name_off); > >>> - targ_type = skip_mods_and_typedefs(obj->btf_vmlinux, > >>> - targ_var->type, > >>> - &targ_type_id); > >>> + targ_var = btf__type_by_id(btf, id); > >>> + targ_var_name = btf__name_by_offset(btf, targ_var->name_off); > >>> + targ_type = skip_mods_and_typedefs(btf, targ_var->type, &targ_type_id); > >>> > >>> ret = bpf_core_types_are_compat(obj->btf, local_type_id, > >>> - obj->btf_vmlinux, targ_type_id); > >>> + btf, targ_type_id); > >>> if (ret <= 0) { > >>> const struct btf_type *local_type; > >>> const char *targ_name, *local_name; > >>> > >>> local_type = btf__type_by_id(obj->btf, local_type_id); > >>> - local_name = btf__name_by_offset(obj->btf, > >>> - local_type->name_off); > >>> - targ_name = btf__name_by_offset(obj->btf_vmlinux, > >>> - targ_type->name_off); > >>> + local_name = btf__name_by_offset(obj->btf, local_type->name_off); > >>> + targ_name = btf__name_by_offset(btf, targ_type->name_off); > >>> > >>> pr_warn("extern (ksym) '%s': incompatible types, expected [%d] %s %s, but kernel has [%d] %s %s\n", > >>> ext->name, local_type_id, > >>> @@ -7370,7 +7382,8 @@ static int bpf_object__resolve_ksyms_btf_id(struct bpf_object *obj) > >>> } > >>> > >>> ext->is_set = true; > >>> - ext->ksym.vmlinux_btf_id = id; > >>> + ext->ksym.kernel_btf_obj_fd = btf_fd; > >>> + ext->ksym.kernel_btf_id = id; > >>> pr_debug("extern (ksym) '%s': resolved to [%d] %s %s\n", > >>> ext->name, id, btf_kind_str(targ_var), targ_var_name); > >>> } > >>>