2019-04-25 09:55 UTC-0700 ~ Andrii Nakryiko <andr...@fb.com>
Add new `btf dump` sub-command to bpftool. It allows to dump
human-readable low-level BTF types representation of BTF types. BTF can
be retrieved from few different sources:
- from BTF object by ID;
- from PROG, if it has associated BTF;
- from MAP, if it has associated BTF data; it's possible to narrow
down types to either key type, value type, both, or all BTF types;
- from ELF file (.BTF section).
Output format mostly follows BPF verifier log format with few notable
exceptions:
- all the type/field/param/etc names are enclosed in single quotes to
allow easier grepping and to stand out a little bit more;
- FUNC_PROTO output follows STRUCT/UNION/ENUM format of having one
line per each argument; this is more uniform and allows easy
grepping, as opposed to succinct, but inconvenient format that BPF
verifier log is using.
Cc: Daniel Borkmann <dan...@iogearbox.net>
Cc: Alexei Starovoitov <a...@fb.com>
Cc: Yonghong Song <y...@fb.com>
Cc: Martin KaFai Lau <ka...@fb.com>
Cc: Song Liu <songliubrav...@fb.com>
Cc: Arnaldo Carvalho de Melo <a...@redhat.com>
Acked-by: Yonghong Song <y...@fb.com>
Signed-off-by: Andrii Nakryiko <andr...@fb.com>
---
tools/bpf/bpftool/btf.c | 580 +++++++++++++++++++++++++++++++++++++++
tools/bpf/bpftool/main.c | 3 +-
tools/bpf/bpftool/main.h | 1 +
3 files changed, 583 insertions(+), 1 deletion(-)
create mode 100644 tools/bpf/bpftool/btf.c
diff --git a/tools/bpf/bpftool/btf.c b/tools/bpf/bpftool/btf.c
new file mode 100644
index 000000000000..cbf04850c798
--- /dev/null
+++ b/tools/bpf/bpftool/btf.c
@@ -0,0 +1,580 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+/* Copyright (C) 2019 Facebook */
+
+#include <errno.h>
+#include <fcntl.h>
+#include <linux/err.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+#include <gelf.h>
+#include <bpf.h>
+#include <linux/btf.h>
Can we have as in prog.c/map.c: standard includes sorted alphabetically,
then linux/ includes, then bpf includes?
+
+#include "btf.h"
+#include "json_writer.h"
+#include "main.h"
+
+static const char * const btf_kind_str[NR_BTF_KINDS] = {
+ [BTF_KIND_UNKN] = "UNKNOWN",
+ [BTF_KIND_INT] = "INT",
+ [BTF_KIND_PTR] = "PTR",
+ [BTF_KIND_ARRAY] = "ARRAY",
+ [BTF_KIND_STRUCT] = "STRUCT",
+ [BTF_KIND_UNION] = "UNION",
+ [BTF_KIND_ENUM] = "ENUM",
+ [BTF_KIND_FWD] = "FWD",
+ [BTF_KIND_TYPEDEF] = "TYPEDEF",
+ [BTF_KIND_VOLATILE] = "VOLATILE",
+ [BTF_KIND_CONST] = "CONST",
+ [BTF_KIND_RESTRICT] = "RESTRICT",
+ [BTF_KIND_FUNC] = "FUNC",
+ [BTF_KIND_FUNC_PROTO] = "FUNC_PROTO",
+ [BTF_KIND_VAR] = "VAR",
+ [BTF_KIND_DATASEC] = "DATASEC",
+};
+
+static const char *btf_int_enc_str(__u8 encoding)
+{
+ switch (encoding) {
+ case 0:
+ return "(none)";
+ case BTF_INT_SIGNED:
+ return "SIGNED";
+ case BTF_INT_CHAR:
+ return "CHAR";
+ case BTF_INT_BOOL:
+ return "BOOL";
+ default:
+ return "UNKN";
+ }
+}
+
+static const char *btf_var_linkage_str(__u32 linkage)
+{
+ switch (linkage) {
+ case BTF_VAR_STATIC:
+ return "static";
+ case BTF_VAR_GLOBAL_ALLOCATED:
+ return "global-alloc";
+ default:
+ return "(unknown)";
+ }
+}
+
+static const char *btf_str(const struct btf *btf, __u32 off)
+{
+ if (!off)
+ return "(anon)";
+ return btf__name_by_offset(btf, off) ? : "(invalid)";
+}
+
+static int dump_btf_type(const struct btf *btf, __u32 id,
+ const struct btf_type *t)
+{
+ int kind = BTF_INFO_KIND(t->info);
+ int safe_kind = kind <= BTF_KIND_MAX ? kind : BTF_KIND_UNKN;
+ json_writer_t *w = json_wtr;
Can we keep reverse-Christmas tree style for declarations? Assigning the
values can be done on its own after the declarations.
+
+ if (json_output) {
+ jsonw_start_object(w);
+ jsonw_uint_field(w, "id", id);
+ jsonw_string_field(w, "kind", btf_kind_str[safe_kind]);
+ jsonw_string_field(w, "name", btf_str(btf, t->name_off));
+ } else {
+ printf("[%u] %s '%s'", id, btf_kind_str[safe_kind],
+ btf_str(btf, t->name_off));
+ }
+
+ switch (BTF_INFO_KIND(t->info)) {
+ case BTF_KIND_INT: {
+ __u32 v = *(__u32 *)(t + 1);
+ const char *enc = btf_int_enc_str(BTF_INT_ENCODING(v));
Same thing here.
+
+ if (json_output) {
+ jsonw_uint_field(w, "size", t->size);
+ jsonw_uint_field(w, "bits_offset", BTF_INT_OFFSET(v));
+ jsonw_uint_field(w, "nr_bits", BTF_INT_BITS(v));
+ jsonw_string_field(w, "encoding", enc);
+ } else {
+ printf(" size=%u bits_offset=%u nr_bits=%u encoding=%s",
+ t->size, BTF_INT_OFFSET(v), BTF_INT_BITS(v),
+ enc);
+ }
+ break;
+ }
+ case BTF_KIND_PTR:
[...]
printf(" type_id=%u", t->type);
+ break;
+ case BTF_KIND_FUNC_PROTO: {
+ const struct btf_param *p = (const void *)(t + 1);
+ __u16 vlen = BTF_INFO_VLEN(t->info);
+ int i;
+
+ if (json_output) {
+ jsonw_uint_field(w, "ret_type_id", t->type);
+ jsonw_uint_field(w, "vlen", vlen);
+ jsonw_name(w, "params");
+ jsonw_start_array(w);
+ } else {
+ printf(" ret_type_id=%u vlen=%u", t->type, vlen);
+ }
+ for (i = 0; i < vlen; i++, p++) {
+ const char *name = btf_str(btf, p->name_off);
+
+ if (json_output) {
+ jsonw_start_object(w);
+ jsonw_string_field(w, "name", name);
+ jsonw_uint_field(w, "type_id", p->type);
+ jsonw_end_object(w);
+ } else {
+ printf("\n\t'%s' type_id=%u", name, p->type);
+ }
+ }
+ if (json_output)
+ jsonw_end_array(w);
+ break;
+ }
+ case BTF_KIND_VAR: {
+ const struct btf_var *v = (const void *)(t + 1);
+ const char *linkage = btf_var_linkage_str(v->linkage);
And here please.
+
+ if (json_output) {
+ jsonw_uint_field(w, "type_id", t->type);
+ jsonw_string_field(w, "linkage", linkage);
+ } else {
+ printf(" type_id=%u, linkage=%s", t->type, linkage);
+ }
+ break;
+ }
+ case BTF_KIND_DATASEC: {
[...]
+static int do_help(int argc, char **argv)
+{
+ if (json_output) {
+ jsonw_null(json_wtr);
+ return 0;
+ }
+
+ fprintf(stderr,
+ "Usage: %s btf dump BTF_SRC\n"
Why so much space between "dump" and "BTF_SRC"?
