[PATCH v3 bpf-next 0/4] libbpf: deprecate bpf_program__get_prog_info_linear

[PATCH v3 bpf-next 0/4] libbpf: deprecate bpf_program__get_prog_info_linear

[Dave Marchevsky]

bpf_program__get_prog_info_linear is a helper which wraps the
bpf_obj_get_info_by_fd BPF syscall with some niceties that put
all dynamic-length bpf_prog_info in one buffer contiguous with struct
bpf_prog_info, and simplify the selection of which dynamic data to grab.

The resultant combined struct, bpf_prog_info_linear, is persisted to
file by 'perf' to enable later annotation of BPF prog data. libbpf
includes some vaddr <-> offset conversion helpers for
struct bpf_prog_info_linear to simplify this.

This functionality is heavily tailored to perf's usecase, so its use as
a general prog info API should be deemphasized in favor of just calling
bpf_obj_get_info_by_fd, which can be more easily fit to purpose. Some
examples from caller migrations in this series:

  * Some callers weren't requesting or using dynamic-sized prog info and
    are well served by a simple get_info_by_fd call (e.g.
    dump_prog_id_as_func_ptr in bpftool)
  * Some callers were requesting all of a specific dynamic info type but
    only using the first record, so can avoid unnecessary malloc by
    only requesting 1 (e.g. profile_target_name in bpftool)
  * bpftool's do_dump saves some malloc/free by growing and reusing its
    dynamic prog_info buf as it loops over progs to grab info and dump.

Perf does need the full functionality of
bpf_program__get_prog_info_linear and its accompanying structs +
helpers, so copy the code to its codebase, migrate all other uses in the
tree, and deprecate the helper in libbpf.

Since the deprecated symbols continue to be included in perf some
renaming was necessary in perf's copy, otherwise functionality is
unchanged.

This work was previously discussed in libbpf's issue tracker [0].

[0]: https://github.com/libbpf/libbpf/issues/313

v2->v3:
  * Remove v2's patch 1 ("libbpf: Migrate internal use of
    bpf_program__get_prog_info_linear"), which was applied [Andrii]
  * Add new patch 1 migrating error checking of libbpf calls to
    new scheme [Andrii, Quentin]
  * In patch 2, fix != -1 error check of libbpf call, improper realloc
    handling, and get rid of confusing macros [Andrii]
  * In patch 4, deprecate starting from 0.6 instead of 0.7 [Andrii]

v1->v2: fix bpftool do_dump changes to clear bpf_prog_info after use and
correctly pass realloc'd ptr back (patch 2)

Dave Marchevsky (4):
  bpftool: Migrate -1 err checks of libbpf fn calls
  bpftool: use bpf_obj_get_info_by_fd directly
  perf: pull in bpf_program__get_prog_info_linear
  libbpf: deprecate bpf_program__get_prog_info_linear

 tools/bpf/bpftool/btf_dumper.c                |  42 +--
 tools/bpf/bpftool/prog.c                      | 158 ++++++++---
 tools/bpf/bpftool/struct_ops.c                |   2 +-
 tools/lib/bpf/libbpf.h                        |   3 +
 .../Documentation/perf.data-file-format.txt   |   2 +-
 tools/perf/util/Build                         |   1 +
 tools/perf/util/annotate.c                    |   3 +-
 tools/perf/util/bpf-event.c                   |  41 ++-
 tools/perf/util/bpf-event.h                   |   2 +-
 tools/perf/util/bpf-utils.c                   | 261 ++++++++++++++++++
 tools/perf/util/bpf-utils.h                   |  76 +++++
 tools/perf/util/bpf_counter.c                 |   6 +-
 tools/perf/util/dso.c                         |   1 +
 tools/perf/util/env.c                         |   1 +
 tools/perf/util/header.c                      |  13 +-
 15 files changed, 527 insertions(+), 85 deletions(-)
 create mode 100644 tools/perf/util/bpf-utils.c
 create mode 100644 tools/perf/util/bpf-utils.h

-- 
2.30.2

[PATCH v3 bpf-next 1/4] bpftool: Migrate -1 err checks of libbpf fn calls

[Dave Marchevsky]

Per [0], callers of libbpf functions with LIBBPF_STRICT_DIRECT_ERRS set
should handle negative error codes of various values (e.g. -EINVAL).
Migrate two callsites which were explicitly checking for -1 only to
handle the new scheme.

  [0]: https://github.com/libbpf/libbpf/wiki/Libbpf-1.0-migration-guide#direct-error-code-returning-libbpf_strict_direct_errs

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
---
 tools/bpf/bpftool/btf_dumper.c | 2 +-
 tools/bpf/bpftool/struct_ops.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/bpf/bpftool/btf_dumper.c b/tools/bpf/bpftool/btf_dumper.c
index 9c25286a5c73..6934e8634b94 100644
--- a/tools/bpf/bpftool/btf_dumper.c
+++ b/tools/bpf/bpftool/btf_dumper.c
@@ -52,7 +52,7 @@ static int dump_prog_id_as_func_ptr(const struct btf_dumper *d,
 
 	/* Get the bpf_prog's name.  Obtain from func_info. */
 	prog_fd = bpf_prog_get_fd_by_id(prog_id);
-	if (prog_fd == -1)
+	if (prog_fd < 0)
 		goto print;
 
 	prog_info = bpf_program__get_prog_info_linear(prog_fd,
diff --git a/tools/bpf/bpftool/struct_ops.c b/tools/bpf/bpftool/struct_ops.c
index ab2d2290569a..20f803dce2e4 100644
--- a/tools/bpf/bpftool/struct_ops.c
+++ b/tools/bpf/bpftool/struct_ops.c
@@ -252,7 +252,7 @@ static struct res do_one_id(const char *id_str, work_func func, void *data,
 	}
 
 	fd = bpf_map_get_fd_by_id(id);
-	if (fd == -1) {
+	if (fd < 0) {
 		p_err("can't get map by id (%lu): %s", id, strerror(errno));
 		res.nr_errs++;
 		return res;
-- 
2.30.2

[PATCH v3 bpf-next 2/4] bpftool: use bpf_obj_get_info_by_fd directly

[Dave Marchevsky]

To prepare for impending deprecation of libbpf's
bpf_program__get_prog_info_linear, migrate uses of this function to use
bpf_obj_get_info_by_fd.

Since the profile_target_name and dump_prog_id_as_func_ptr helpers were
only looking at the first func_info, avoid grabbing the rest to save a
malloc. For do_dump, add a more full-featured helper, but avoid
free/realloc of buffer when possible for multi-prog dumps.

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
---
 tools/bpf/bpftool/btf_dumper.c |  40 +++++----
 tools/bpf/bpftool/prog.c       | 158 ++++++++++++++++++++++++++-------
 2 files changed, 148 insertions(+), 50 deletions(-)

diff --git a/tools/bpf/bpftool/btf_dumper.c b/tools/bpf/bpftool/btf_dumper.c
index 6934e8634b94..f5dddf8ef404 100644
--- a/tools/bpf/bpftool/btf_dumper.c
+++ b/tools/bpf/bpftool/btf_dumper.c
@@ -32,14 +32,16 @@ static int dump_prog_id_as_func_ptr(const struct btf_dumper *d,
 				    const struct btf_type *func_proto,
 				    __u32 prog_id)
 {
-	struct bpf_prog_info_linear *prog_info = NULL;
 	const struct btf_type *func_type;
+	int prog_fd = -1, func_sig_len;
+	struct bpf_prog_info info = {};
+	__u32 info_len = sizeof(info);
 	const char *prog_name = NULL;
-	struct bpf_func_info *finfo;
 	struct btf *prog_btf = NULL;
-	struct bpf_prog_info *info;
-	int prog_fd, func_sig_len;
+	struct bpf_func_info finfo;
+	__u32 finfo_rec_size;
 	char prog_str[1024];
+	int err;
 
 	/* Get the ptr's func_proto */
 	func_sig_len = btf_dump_func(d->btf, prog_str, func_proto, NULL, 0,
@@ -55,22 +57,27 @@ static int dump_prog_id_as_func_ptr(const struct btf_dumper *d,
 	if (prog_fd < 0)
 		goto print;
 
-	prog_info = bpf_program__get_prog_info_linear(prog_fd,
-						1UL << BPF_PROG_INFO_FUNC_INFO);
-	close(prog_fd);
-	if (IS_ERR(prog_info)) {
-		prog_info = NULL;
+	err = bpf_obj_get_info_by_fd(prog_fd, &info, &info_len);
+	if (err)
 		goto print;
-	}
-	info = &prog_info->info;
 
-	if (!info->btf_id || !info->nr_func_info)
+	if (!info.btf_id || !info.nr_func_info)
+		goto print;
+
+	finfo_rec_size = info.func_info_rec_size;
+	memset(&info, 0, sizeof(info));
+	info.nr_func_info = 1;
+	info.func_info_rec_size = finfo_rec_size;
+	info.func_info = ptr_to_u64(&finfo);
+
+	err = bpf_obj_get_info_by_fd(prog_fd, &info, &info_len);
+	if (err)
 		goto print;
-	prog_btf = btf__load_from_kernel_by_id(info->btf_id);
+
+	prog_btf = btf__load_from_kernel_by_id(info.btf_id);
 	if (libbpf_get_error(prog_btf))
 		goto print;
-	finfo = u64_to_ptr(info->func_info);
-	func_type = btf__type_by_id(prog_btf, finfo->type_id);
+	func_type = btf__type_by_id(prog_btf, finfo.type_id);
 	if (!func_type || !btf_is_func(func_type))
 		goto print;
 
@@ -92,7 +99,8 @@ static int dump_prog_id_as_func_ptr(const struct btf_dumper *d,
 	prog_str[sizeof(prog_str) - 1] = '\0';
 	jsonw_string(d->jw, prog_str);
 	btf__free(prog_btf);
-	free(prog_info);
+	if (prog_fd >= 0)
+		close(prog_fd);
 	return 0;
 }
 
diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
index 515d22952602..0b6f1534d583 100644
--- a/tools/bpf/bpftool/prog.c
+++ b/tools/bpf/bpftool/prog.c
@@ -100,6 +100,76 @@ static enum bpf_attach_type parse_attach_type(const char *str)
 	return __MAX_BPF_ATTACH_TYPE;
 }
 
+static int prep_prog_info(struct bpf_prog_info *const info, enum dump_mode mode,
+			  void **info_data, size_t *const info_data_sz)
+{
+	struct bpf_prog_info holder = {};
+	size_t needed = 0;
+	void *ptr;
+
+	if (mode == DUMP_JITED) {
+		holder.jited_prog_len = info->jited_prog_len;
+		needed += info->jited_prog_len;
+	} else {
+		holder.xlated_prog_len = info->xlated_prog_len;
+		needed += info->xlated_prog_len;
+	}
+
+	holder.nr_jited_ksyms = info->nr_jited_ksyms;
+	needed += info->nr_jited_ksyms * sizeof(__u64);
+
+	holder.nr_jited_func_lens = info->nr_jited_func_lens;
+	needed += info->nr_jited_func_lens * sizeof(__u32);
+
+	holder.nr_func_info = info->nr_func_info;
+	holder.func_info_rec_size = info->func_info_rec_size;
+	needed += info->nr_func_info * info->func_info_rec_size;
+
+	holder.nr_line_info = info->nr_line_info;
+	holder.line_info_rec_size = info->line_info_rec_size;
+	needed += info->nr_line_info * info->line_info_rec_size;
+
+	holder.nr_jited_line_info = info->nr_jited_line_info;
+	holder.jited_line_info_rec_size = info->jited_line_info_rec_size;
+	needed += info->nr_jited_line_info * info->jited_line_info_rec_size;
+
+	if (needed > *info_data_sz) {
+		ptr = realloc(*info_data, needed);
+		if (!ptr)
+			return -1;
+
+		*info_data = ptr;
+		*info_data_sz = needed;
+	}
+	ptr = *info_data;
+
+	if (mode == DUMP_JITED) {
+		holder.jited_prog_insns = ptr_to_u64(ptr);
+		ptr += holder.jited_prog_len;
+	} else {
+		holder.xlated_prog_insns = ptr_to_u64(ptr);
+		ptr += holder.xlated_prog_len;
+	}
+
+	holder.jited_ksyms = ptr_to_u64(ptr);
+	ptr += holder.nr_jited_ksyms * sizeof(__u64);
+
+	holder.jited_func_lens = ptr_to_u64(ptr);
+	ptr += holder.nr_jited_func_lens * sizeof(__u32);
+
+	holder.func_info = ptr_to_u64(ptr);
+	ptr += holder.nr_func_info * holder.func_info_rec_size;
+
+	holder.line_info = ptr_to_u64(ptr);
+	ptr += holder.nr_line_info * holder.line_info_rec_size;
+
+	holder.jited_line_info = ptr_to_u64(ptr);
+	ptr += holder.nr_jited_line_info * holder.jited_line_info_rec_size;
+
+	*info = holder;
+	return 0;
+}
+
 static void print_boot_time(__u64 nsecs, char *buf, unsigned int size)
 {
 	struct timespec real_time_ts, boot_time_ts;
@@ -803,16 +873,18 @@ prog_dump(struct bpf_prog_info *info, enum dump_mode mode,
 
 static int do_dump(int argc, char **argv)
 {
-	struct bpf_prog_info_linear *info_linear;
+	struct bpf_prog_info info = {};
+	__u32 info_len = sizeof(info);
+	size_t info_data_sz = 0;
+	void *info_data = NULL;
 	char *filepath = NULL;
 	bool opcodes = false;
 	bool visual = false;
 	enum dump_mode mode;
 	bool linum = false;
-	int *fds = NULL;
 	int nb_fds, i = 0;
+	int *fds = NULL;
 	int err = -1;
-	__u64 arrays;
 
 	if (is_prefix(*argv, "jited")) {
 		if (disasm_init())
@@ -872,46 +944,46 @@ static int do_dump(int argc, char **argv)
 		goto exit_close;
 	}
 
-	if (mode == DUMP_JITED)
-		arrays = 1UL << BPF_PROG_INFO_JITED_INSNS;
-	else
-		arrays = 1UL << BPF_PROG_INFO_XLATED_INSNS;
-
-	arrays |= 1UL << BPF_PROG_INFO_JITED_KSYMS;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_FUNC_LENS;
-	arrays |= 1UL << BPF_PROG_INFO_FUNC_INFO;
-	arrays |= 1UL << BPF_PROG_INFO_LINE_INFO;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_LINE_INFO;
-
 	if (json_output && nb_fds > 1)
 		jsonw_start_array(json_wtr);	/* root array */
 	for (i = 0; i < nb_fds; i++) {
-		info_linear = bpf_program__get_prog_info_linear(fds[i], arrays);
-		if (IS_ERR_OR_NULL(info_linear)) {
+		err = bpf_obj_get_info_by_fd(fds[i], &info, &info_len);
+		if (err) {
+			p_err("can't get prog info: %s", strerror(errno));
+			break;
+		}
+
+		err = prep_prog_info(&info, mode, &info_data, &info_data_sz);
+		if (err) {
+			p_err("can't grow prog info_data");
+			break;
+		}
+
+		err = bpf_obj_get_info_by_fd(fds[i], &info, &info_len);
+		if (err) {
 			p_err("can't get prog info: %s", strerror(errno));
 			break;
 		}
 
 		if (json_output && nb_fds > 1) {
 			jsonw_start_object(json_wtr);	/* prog object */
-			print_prog_header_json(&info_linear->info);
+			print_prog_header_json(&info);
 			jsonw_name(json_wtr, "insns");
 		} else if (nb_fds > 1) {
-			print_prog_header_plain(&info_linear->info);
+			print_prog_header_plain(&info);
 		}
 
-		err = prog_dump(&info_linear->info, mode, filepath, opcodes,
-				visual, linum);
+		err = prog_dump(&info, mode, filepath, opcodes, visual, linum);
 
 		if (json_output && nb_fds > 1)
 			jsonw_end_object(json_wtr);	/* prog object */
 		else if (i != nb_fds - 1 && nb_fds > 1)
 			printf("\n");
 
-		free(info_linear);
 		if (err)
 			break;
 		close(fds[i]);
+		memset(&info, 0, sizeof(info));
 	}
 	if (json_output && nb_fds > 1)
 		jsonw_end_array(json_wtr);	/* root array */
@@ -920,6 +992,7 @@ static int do_dump(int argc, char **argv)
 	for (; i < nb_fds; i++)
 		close(fds[i]);
 exit_free:
+	free(info_data);
 	free(fds);
 	return err;
 }
@@ -2016,41 +2089,58 @@ static void profile_print_readings(void)
 
 static char *profile_target_name(int tgt_fd)
 {
-	struct bpf_prog_info_linear *info_linear;
-	struct bpf_func_info *func_info;
+	struct bpf_func_info func_info;
+	struct bpf_prog_info info = {};
+	__u32 info_len = sizeof(info);
 	const struct btf_type *t;
+	__u32 func_info_rec_size;
 	struct btf *btf = NULL;
 	char *name = NULL;
+	int err;
 
-	info_linear = bpf_program__get_prog_info_linear(
-		tgt_fd, 1UL << BPF_PROG_INFO_FUNC_INFO);
-	if (IS_ERR_OR_NULL(info_linear)) {
-		p_err("failed to get info_linear for prog FD %d", tgt_fd);
-		return NULL;
+	err = bpf_obj_get_info_by_fd(tgt_fd, &info, &info_len);
+	if (err) {
+		p_err("failed to bpf_obj_get_info_by_fd for prog FD %d", tgt_fd);
+		goto out;
 	}
 
-	if (info_linear->info.btf_id == 0) {
+	if (info.btf_id == 0) {
 		p_err("prog FD %d doesn't have valid btf", tgt_fd);
 		goto out;
 	}
 
-	btf = btf__load_from_kernel_by_id(info_linear->info.btf_id);
+	func_info_rec_size = info.func_info_rec_size;
+	if (info.nr_func_info == 0) {
+		p_err("bpf_obj_get_info_by_fd for prog FD %d found 0 func_info", tgt_fd);
+		goto out;
+	}
+
+	memset(&info, 0, sizeof(info));
+	info.nr_func_info = 1;
+	info.func_info_rec_size = func_info_rec_size;
+	info.func_info = ptr_to_u64(&func_info);
+
+	err = bpf_obj_get_info_by_fd(tgt_fd, &info, &info_len);
+	if (err) {
+		p_err("failed to get func_info for prog FD %d", tgt_fd);
+		goto out;
+	}
+
+	btf = btf__load_from_kernel_by_id(info.btf_id);
 	if (libbpf_get_error(btf)) {
 		p_err("failed to load btf for prog FD %d", tgt_fd);
 		goto out;
 	}
 
-	func_info = u64_to_ptr(info_linear->info.func_info);
-	t = btf__type_by_id(btf, func_info[0].type_id);
+	t = btf__type_by_id(btf, func_info.type_id);
 	if (!t) {
 		p_err("btf %d doesn't have type %d",
-		      info_linear->info.btf_id, func_info[0].type_id);
+		      info.btf_id, func_info.type_id);
 		goto out;
 	}
 	name = strdup(btf__name_by_offset(btf, t->name_off));
 out:
 	btf__free(btf);
-	free(info_linear);
 	return name;
 }
 
-- 
2.30.2

[PATCH v3 bpf-next 3/4] perf: pull in bpf_program__get_prog_info_linear

[Dave Marchevsky]

To prepare for impending deprecation of libbpf's
bpf_program__get_prog_info_linear, pull in the function and associated
helpers into the perf codebase and migrate existing uses to the perf
copy.

Since libbpf's deprecated definitions will still be visible to perf, it
is necessary to rename perf's definitions.

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
---
 .../Documentation/perf.data-file-format.txt   |   2 +-
 tools/perf/util/Build                         |   1 +
 tools/perf/util/annotate.c                    |   3 +-
 tools/perf/util/bpf-event.c                   |  41 ++-
 tools/perf/util/bpf-event.h                   |   2 +-
 tools/perf/util/bpf-utils.c                   | 261 ++++++++++++++++++
 tools/perf/util/bpf-utils.h                   |  76 +++++
 tools/perf/util/bpf_counter.c                 |   6 +-
 tools/perf/util/dso.c                         |   1 +
 tools/perf/util/env.c                         |   1 +
 tools/perf/util/header.c                      |  13 +-
 11 files changed, 374 insertions(+), 33 deletions(-)
 create mode 100644 tools/perf/util/bpf-utils.c
 create mode 100644 tools/perf/util/bpf-utils.h

diff --git a/tools/perf/Documentation/perf.data-file-format.txt b/tools/perf/Documentation/perf.data-file-format.txt
index e6ff8c898ada..f56d0e0fbff6 100644
--- a/tools/perf/Documentation/perf.data-file-format.txt
+++ b/tools/perf/Documentation/perf.data-file-format.txt
@@ -346,7 +346,7 @@ to special needs.
 
         HEADER_BPF_PROG_INFO = 25,
 
-struct bpf_prog_info_linear, which contains detailed information about
+struct perf_bpil, which contains detailed information about
 a BPF program, including type, id, tag, jited/xlated instructions, etc.
 
         HEADER_BPF_BTF = 26,
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index f2914d5bed6e..ee42da1d3639 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -201,6 +201,7 @@ endif
 perf-y += perf-hooks.o
 
 perf-$(CONFIG_LIBBPF) += bpf-event.o
+perf-$(CONFIG_LIBBPF) += bpf-utils.o
 
 perf-$(CONFIG_CXX) += c++/
 
diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c
index 0bae061b2d6d..f0e5a236b7e3 100644
--- a/tools/perf/util/annotate.c
+++ b/tools/perf/util/annotate.c
@@ -28,6 +28,7 @@
 #include "evsel.h"
 #include "evlist.h"
 #include "bpf-event.h"
+#include "bpf-utils.h"
 #include "block-range.h"
 #include "string2.h"
 #include "util/event.h"
@@ -1700,12 +1701,12 @@ static int symbol__disassemble_bpf(struct symbol *sym,
 {
 	struct annotation *notes = symbol__annotation(sym);
 	struct annotation_options *opts = args->options;
-	struct bpf_prog_info_linear *info_linear;
 	struct bpf_prog_linfo *prog_linfo = NULL;
 	struct bpf_prog_info_node *info_node;
 	int len = sym->end - sym->start;
 	disassembler_ftype disassemble;
 	struct map *map = args->ms.map;
+	struct perf_bpil *info_linear;
 	struct disassemble_info info;
 	struct dso *dso = map->dso;
 	int pc = 0, count, sub_id;
diff --git a/tools/perf/util/bpf-event.c b/tools/perf/util/bpf-event.c
index 388847bab6d9..a27badb0a53a 100644
--- a/tools/perf/util/bpf-event.c
+++ b/tools/perf/util/bpf-event.c
@@ -10,6 +10,7 @@
 #include <internal/lib.h>
 #include <symbol/kallsyms.h>
 #include "bpf-event.h"
+#include "bpf-utils.h"
 #include "debug.h"
 #include "dso.h"
 #include "symbol.h"
@@ -32,8 +33,6 @@ struct btf * __weak btf__load_from_kernel_by_id(__u32 id)
        return err ? ERR_PTR(err) : btf;
 }
 
-#define ptr_to_u64(ptr)    ((__u64)(unsigned long)(ptr))
-
 static int snprintf_hex(char *buf, size_t size, unsigned char *data, size_t len)
 {
 	int ret = 0;
@@ -48,9 +47,9 @@ static int machine__process_bpf_event_load(struct machine *machine,
 					   union perf_event *event,
 					   struct perf_sample *sample __maybe_unused)
 {
-	struct bpf_prog_info_linear *info_linear;
 	struct bpf_prog_info_node *info_node;
 	struct perf_env *env = machine->env;
+	struct perf_bpil *info_linear;
 	int id = event->bpf.id;
 	unsigned int i;
 
@@ -175,9 +174,9 @@ static int perf_event__synthesize_one_bpf_prog(struct perf_session *session,
 {
 	struct perf_record_ksymbol *ksymbol_event = &event->ksymbol;
 	struct perf_record_bpf_event *bpf_event = &event->bpf;
-	struct bpf_prog_info_linear *info_linear;
 	struct perf_tool *tool = session->tool;
 	struct bpf_prog_info_node *info_node;
+	struct perf_bpil *info_linear;
 	struct bpf_prog_info *info;
 	struct btf *btf = NULL;
 	struct perf_env *env;
@@ -191,15 +190,15 @@ static int perf_event__synthesize_one_bpf_prog(struct perf_session *session,
 	 */
 	env = session->data ? &session->header.env : &perf_env;
 
-	arrays = 1UL << BPF_PROG_INFO_JITED_KSYMS;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_FUNC_LENS;
-	arrays |= 1UL << BPF_PROG_INFO_FUNC_INFO;
-	arrays |= 1UL << BPF_PROG_INFO_PROG_TAGS;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_INSNS;
-	arrays |= 1UL << BPF_PROG_INFO_LINE_INFO;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_LINE_INFO;
+	arrays = 1UL << PERF_BPIL_JITED_KSYMS;
+	arrays |= 1UL << PERF_BPIL_JITED_FUNC_LENS;
+	arrays |= 1UL << PERF_BPIL_FUNC_INFO;
+	arrays |= 1UL << PERF_BPIL_PROG_TAGS;
+	arrays |= 1UL << PERF_BPIL_JITED_INSNS;
+	arrays |= 1UL << PERF_BPIL_LINE_INFO;
+	arrays |= 1UL << PERF_BPIL_JITED_LINE_INFO;
 
-	info_linear = bpf_program__get_prog_info_linear(fd, arrays);
+	info_linear = get_bpf_prog_info_linear(fd, arrays);
 	if (IS_ERR_OR_NULL(info_linear)) {
 		info_linear = NULL;
 		pr_debug("%s: failed to get BPF program info. aborting\n", __func__);
@@ -452,8 +451,8 @@ int perf_event__synthesize_bpf_events(struct perf_session *session,
 
 static void perf_env__add_bpf_info(struct perf_env *env, u32 id)
 {
-	struct bpf_prog_info_linear *info_linear;
 	struct bpf_prog_info_node *info_node;
+	struct perf_bpil *info_linear;
 	struct btf *btf = NULL;
 	u64 arrays;
 	u32 btf_id;
@@ -463,15 +462,15 @@ static void perf_env__add_bpf_info(struct perf_env *env, u32 id)
 	if (fd < 0)
 		return;
 
-	arrays = 1UL << BPF_PROG_INFO_JITED_KSYMS;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_FUNC_LENS;
-	arrays |= 1UL << BPF_PROG_INFO_FUNC_INFO;
-	arrays |= 1UL << BPF_PROG_INFO_PROG_TAGS;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_INSNS;
-	arrays |= 1UL << BPF_PROG_INFO_LINE_INFO;
-	arrays |= 1UL << BPF_PROG_INFO_JITED_LINE_INFO;
+	arrays = 1UL << PERF_BPIL_JITED_KSYMS;
+	arrays |= 1UL << PERF_BPIL_JITED_FUNC_LENS;
+	arrays |= 1UL << PERF_BPIL_FUNC_INFO;
+	arrays |= 1UL << PERF_BPIL_PROG_TAGS;
+	arrays |= 1UL << PERF_BPIL_JITED_INSNS;
+	arrays |= 1UL << PERF_BPIL_LINE_INFO;
+	arrays |= 1UL << PERF_BPIL_JITED_LINE_INFO;
 
-	info_linear = bpf_program__get_prog_info_linear(fd, arrays);
+	info_linear = get_bpf_prog_info_linear(fd, arrays);
 	if (IS_ERR_OR_NULL(info_linear)) {
 		pr_debug("%s: failed to get BPF program info. aborting\n", __func__);
 		goto out;
diff --git a/tools/perf/util/bpf-event.h b/tools/perf/util/bpf-event.h
index 68f315c3df5b..144a8a24cc69 100644
--- a/tools/perf/util/bpf-event.h
+++ b/tools/perf/util/bpf-event.h
@@ -19,7 +19,7 @@ struct evlist;
 struct target;
 
 struct bpf_prog_info_node {
-	struct bpf_prog_info_linear	*info_linear;
+	struct perf_bpil		*info_linear;
 	struct rb_node			rb_node;
 };
 
diff --git a/tools/perf/util/bpf-utils.c b/tools/perf/util/bpf-utils.c
new file mode 100644
index 000000000000..e271e05e51bc
--- /dev/null
+++ b/tools/perf/util/bpf-utils.c
@@ -0,0 +1,261 @@
+// SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+
+#include <errno.h>
+#include <stdlib.h>
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <bpf/bpf.h>
+#include "bpf-utils.h"
+#include "debug.h"
+
+struct bpil_array_desc {
+	int	array_offset;	/* e.g. offset of jited_prog_insns */
+	int	count_offset;	/* e.g. offset of jited_prog_len */
+	int	size_offset;	/* > 0: offset of rec size,
+				 * < 0: fix size of -size_offset
+				 */
+};
+
+static struct bpil_array_desc bpil_array_desc[] = {
+	[PERF_BPIL_JITED_INSNS] = {
+		offsetof(struct bpf_prog_info, jited_prog_insns),
+		offsetof(struct bpf_prog_info, jited_prog_len),
+		-1,
+	},
+	[PERF_BPIL_XLATED_INSNS] = {
+		offsetof(struct bpf_prog_info, xlated_prog_insns),
+		offsetof(struct bpf_prog_info, xlated_prog_len),
+		-1,
+	},
+	[PERF_BPIL_MAP_IDS] = {
+		offsetof(struct bpf_prog_info, map_ids),
+		offsetof(struct bpf_prog_info, nr_map_ids),
+		-(int)sizeof(__u32),
+	},
+	[PERF_BPIL_JITED_KSYMS] = {
+		offsetof(struct bpf_prog_info, jited_ksyms),
+		offsetof(struct bpf_prog_info, nr_jited_ksyms),
+		-(int)sizeof(__u64),
+	},
+	[PERF_BPIL_JITED_FUNC_LENS] = {
+		offsetof(struct bpf_prog_info, jited_func_lens),
+		offsetof(struct bpf_prog_info, nr_jited_func_lens),
+		-(int)sizeof(__u32),
+	},
+	[PERF_BPIL_FUNC_INFO] = {
+		offsetof(struct bpf_prog_info, func_info),
+		offsetof(struct bpf_prog_info, nr_func_info),
+		offsetof(struct bpf_prog_info, func_info_rec_size),
+	},
+	[PERF_BPIL_LINE_INFO] = {
+		offsetof(struct bpf_prog_info, line_info),
+		offsetof(struct bpf_prog_info, nr_line_info),
+		offsetof(struct bpf_prog_info, line_info_rec_size),
+	},
+	[PERF_BPIL_JITED_LINE_INFO] = {
+		offsetof(struct bpf_prog_info, jited_line_info),
+		offsetof(struct bpf_prog_info, nr_jited_line_info),
+		offsetof(struct bpf_prog_info, jited_line_info_rec_size),
+	},
+	[PERF_BPIL_PROG_TAGS] = {
+		offsetof(struct bpf_prog_info, prog_tags),
+		offsetof(struct bpf_prog_info, nr_prog_tags),
+		-(int)sizeof(__u8) * BPF_TAG_SIZE,
+	},
+
+};
+
+static __u32 bpf_prog_info_read_offset_u32(struct bpf_prog_info *info,
+					   int offset)
+{
+	__u32 *array = (__u32 *)info;
+
+	if (offset >= 0)
+		return array[offset / sizeof(__u32)];
+	return -(int)offset;
+}
+
+static __u64 bpf_prog_info_read_offset_u64(struct bpf_prog_info *info,
+					   int offset)
+{
+	__u64 *array = (__u64 *)info;
+
+	if (offset >= 0)
+		return array[offset / sizeof(__u64)];
+	return -(int)offset;
+}
+
+static void bpf_prog_info_set_offset_u32(struct bpf_prog_info *info, int offset,
+					 __u32 val)
+{
+	__u32 *array = (__u32 *)info;
+
+	if (offset >= 0)
+		array[offset / sizeof(__u32)] = val;
+}
+
+static void bpf_prog_info_set_offset_u64(struct bpf_prog_info *info, int offset,
+					 __u64 val)
+{
+	__u64 *array = (__u64 *)info;
+
+	if (offset >= 0)
+		array[offset / sizeof(__u64)] = val;
+}
+
+struct perf_bpil *
+get_bpf_prog_info_linear(int fd, __u64 arrays)
+{
+	struct bpf_prog_info info = {};
+	struct perf_bpil *info_linear;
+	__u32 info_len = sizeof(info);
+	__u32 data_len = 0;
+	int i, err;
+	void *ptr;
+
+	if (arrays >> PERF_BPIL_LAST_ARRAY)
+		return ERR_PTR(-EINVAL);
+
+	/* step 1: get array dimensions */
+	err = bpf_obj_get_info_by_fd(fd, &info, &info_len);
+	if (err) {
+		pr_debug("can't get prog info: %s", strerror(errno));
+		return ERR_PTR(-EFAULT);
+	}
+
+	/* step 2: calculate total size of all arrays */
+	for (i = PERF_BPIL_FIRST_ARRAY; i < PERF_BPIL_LAST_ARRAY; ++i) {
+		bool include_array = (arrays & (1UL << i)) > 0;
+		struct bpil_array_desc *desc;
+		__u32 count, size;
+
+		desc = bpil_array_desc + i;
+
+		/* kernel is too old to support this field */
+		if (info_len < desc->array_offset + sizeof(__u32) ||
+		    info_len < desc->count_offset + sizeof(__u32) ||
+		    (desc->size_offset > 0 && info_len < (__u32)desc->size_offset))
+			include_array = false;
+
+		if (!include_array) {
+			arrays &= ~(1UL << i);	/* clear the bit */
+			continue;
+		}
+
+		count = bpf_prog_info_read_offset_u32(&info, desc->count_offset);
+		size  = bpf_prog_info_read_offset_u32(&info, desc->size_offset);
+
+		data_len += count * size;
+	}
+
+	/* step 3: allocate continuous memory */
+	data_len = roundup(data_len, sizeof(__u64));
+	info_linear = malloc(sizeof(struct perf_bpil) + data_len);
+	if (!info_linear)
+		return ERR_PTR(-ENOMEM);
+
+	/* step 4: fill data to info_linear->info */
+	info_linear->arrays = arrays;
+	memset(&info_linear->info, 0, sizeof(info));
+	ptr = info_linear->data;
+
+	for (i = PERF_BPIL_FIRST_ARRAY; i < PERF_BPIL_LAST_ARRAY; ++i) {
+		struct bpil_array_desc *desc;
+		__u32 count, size;
+
+		if ((arrays & (1UL << i)) == 0)
+			continue;
+
+		desc  = bpil_array_desc + i;
+		count = bpf_prog_info_read_offset_u32(&info, desc->count_offset);
+		size  = bpf_prog_info_read_offset_u32(&info, desc->size_offset);
+		bpf_prog_info_set_offset_u32(&info_linear->info,
+					     desc->count_offset, count);
+		bpf_prog_info_set_offset_u32(&info_linear->info,
+					     desc->size_offset, size);
+		bpf_prog_info_set_offset_u64(&info_linear->info,
+					     desc->array_offset,
+					     ptr_to_u64(ptr));
+		ptr += count * size;
+	}
+
+	/* step 5: call syscall again to get required arrays */
+	err = bpf_obj_get_info_by_fd(fd, &info_linear->info, &info_len);
+	if (err) {
+		pr_debug("can't get prog info: %s", strerror(errno));
+		free(info_linear);
+		return ERR_PTR(-EFAULT);
+	}
+
+	/* step 6: verify the data */
+	for (i = PERF_BPIL_FIRST_ARRAY; i < PERF_BPIL_LAST_ARRAY; ++i) {
+		struct bpil_array_desc *desc;
+		__u32 v1, v2;
+
+		if ((arrays & (1UL << i)) == 0)
+			continue;
+
+		desc = bpil_array_desc + i;
+		v1 = bpf_prog_info_read_offset_u32(&info, desc->count_offset);
+		v2 = bpf_prog_info_read_offset_u32(&info_linear->info,
+						   desc->count_offset);
+		if (v1 != v2)
+			pr_warning("%s: mismatch in element count\n", __func__);
+
+		v1 = bpf_prog_info_read_offset_u32(&info, desc->size_offset);
+		v2 = bpf_prog_info_read_offset_u32(&info_linear->info,
+						   desc->size_offset);
+		if (v1 != v2)
+			pr_warning("%s: mismatch in rec size\n", __func__);
+	}
+
+	/* step 7: update info_len and data_len */
+	info_linear->info_len = sizeof(struct bpf_prog_info);
+	info_linear->data_len = data_len;
+
+	return info_linear;
+}
+
+void bpil_addr_to_offs(struct perf_bpil *info_linear)
+{
+	int i;
+
+	for (i = PERF_BPIL_FIRST_ARRAY; i < PERF_BPIL_LAST_ARRAY; ++i) {
+		struct bpil_array_desc *desc;
+		__u64 addr, offs;
+
+		if ((info_linear->arrays & (1UL << i)) == 0)
+			continue;
+
+		desc = bpil_array_desc + i;
+		addr = bpf_prog_info_read_offset_u64(&info_linear->info,
+						     desc->array_offset);
+		offs = addr - ptr_to_u64(info_linear->data);
+		bpf_prog_info_set_offset_u64(&info_linear->info,
+					     desc->array_offset, offs);
+	}
+}
+
+void bpil_offs_to_addr(struct perf_bpil *info_linear)
+{
+	int i;
+
+	for (i = PERF_BPIL_FIRST_ARRAY; i < PERF_BPIL_LAST_ARRAY; ++i) {
+		struct bpil_array_desc *desc;
+		__u64 addr, offs;
+
+		if ((info_linear->arrays & (1UL << i)) == 0)
+			continue;
+
+		desc = bpil_array_desc + i;
+		offs = bpf_prog_info_read_offset_u64(&info_linear->info,
+						     desc->array_offset);
+		addr = offs + ptr_to_u64(info_linear->data);
+		bpf_prog_info_set_offset_u64(&info_linear->info,
+					     desc->array_offset, addr);
+	}
+}
diff --git a/tools/perf/util/bpf-utils.h b/tools/perf/util/bpf-utils.h
new file mode 100644
index 000000000000..86a5055cdfad
--- /dev/null
+++ b/tools/perf/util/bpf-utils.h
@@ -0,0 +1,76 @@
+/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */
+
+#ifndef __PERF_BPF_UTILS_H
+#define __PERF_BPF_UTILS_H
+
+#define ptr_to_u64(ptr)    ((__u64)(unsigned long)(ptr))
+
+#ifdef HAVE_LIBBPF_SUPPORT
+
+#include <bpf/libbpf.h>
+
+/*
+ * Get bpf_prog_info in continuous memory
+ *
+ * struct bpf_prog_info has multiple arrays. The user has option to choose
+ * arrays to fetch from kernel. The following APIs provide an uniform way to
+ * fetch these data. All arrays in bpf_prog_info are stored in a single
+ * continuous memory region. This makes it easy to store the info in a
+ * file.
+ *
+ * Before writing perf_bpil to files, it is necessary to
+ * translate pointers in bpf_prog_info to offsets. Helper functions
+ * bpil_addr_to_offs() and bpil_offs_to_addr()
+ * are introduced to switch between pointers and offsets.
+ *
+ * Examples:
+ *   # To fetch map_ids and prog_tags:
+ *   __u64 arrays = (1UL << PERF_BPIL_MAP_IDS) |
+ *           (1UL << PERF_BPIL_PROG_TAGS);
+ *   struct perf_bpil *info_linear =
+ *           get_bpf_prog_info_linear(fd, arrays);
+ *
+ *   # To save data in file
+ *   bpil_addr_to_offs(info_linear);
+ *   write(f, info_linear, sizeof(*info_linear) + info_linear->data_len);
+ *
+ *   # To read data from file
+ *   read(f, info_linear, <proper_size>);
+ *   bpil_offs_to_addr(info_linear);
+ */
+enum perf_bpil_array_types {
+	PERF_BPIL_FIRST_ARRAY = 0,
+	PERF_BPIL_JITED_INSNS = 0,
+	PERF_BPIL_XLATED_INSNS,
+	PERF_BPIL_MAP_IDS,
+	PERF_BPIL_JITED_KSYMS,
+	PERF_BPIL_JITED_FUNC_LENS,
+	PERF_BPIL_FUNC_INFO,
+	PERF_BPIL_LINE_INFO,
+	PERF_BPIL_JITED_LINE_INFO,
+	PERF_BPIL_PROG_TAGS,
+	PERF_BPIL_LAST_ARRAY,
+};
+
+struct perf_bpil {
+	/* size of struct bpf_prog_info, when the tool is compiled */
+	__u32			info_len;
+	/* total bytes allocated for data, round up to 8 bytes */
+	__u32			data_len;
+	/* which arrays are included in data */
+	__u64			arrays;
+	struct bpf_prog_info	info;
+	__u8			data[];
+};
+
+struct perf_bpil *
+get_bpf_prog_info_linear(int fd, __u64 arrays);
+
+void
+bpil_addr_to_offs(struct perf_bpil *info_linear);
+
+void
+bpil_offs_to_addr(struct perf_bpil *info_linear);
+
+#endif /* HAVE_LIBBPF_SUPPORT */
+#endif /* __PERF_BPF_UTILS_H */
diff --git a/tools/perf/util/bpf_counter.c b/tools/perf/util/bpf_counter.c
index ba0f20853651..2b04df8c5f87 100644
--- a/tools/perf/util/bpf_counter.c
+++ b/tools/perf/util/bpf_counter.c
@@ -13,6 +13,7 @@
 #include <perf/bpf_perf.h>
 
 #include "bpf_counter.h"
+#include "bpf-utils.h"
 #include "counts.h"
 #include "debug.h"
 #include "evsel.h"
@@ -61,14 +62,13 @@ static int bpf_program_profiler__destroy(struct evsel *evsel)
 
 static char *bpf_target_prog_name(int tgt_fd)
 {
-	struct bpf_prog_info_linear *info_linear;
 	struct bpf_func_info *func_info;
+	struct perf_bpil *info_linear;
 	const struct btf_type *t;
 	struct btf *btf = NULL;
 	char *name = NULL;
 
-	info_linear = bpf_program__get_prog_info_linear(
-		tgt_fd, 1UL << BPF_PROG_INFO_FUNC_INFO);
+	info_linear = get_bpf_prog_info_linear(tgt_fd, 1UL << PERF_BPIL_FUNC_INFO);
 	if (IS_ERR_OR_NULL(info_linear)) {
 		pr_debug("failed to get info_linear for prog FD %d\n", tgt_fd);
 		return NULL;
diff --git a/tools/perf/util/dso.c b/tools/perf/util/dso.c
index 9ed9a5676d35..9cc8a1772b4b 100644
--- a/tools/perf/util/dso.c
+++ b/tools/perf/util/dso.c
@@ -14,6 +14,7 @@
 #ifdef HAVE_LIBBPF_SUPPORT
 #include <bpf/libbpf.h>
 #include "bpf-event.h"
+#include "bpf-utils.h"
 #endif
 #include "compress.h"
 #include "env.h"
diff --git a/tools/perf/util/env.c b/tools/perf/util/env.c
index cf773f0dec38..17f1dd0680b4 100644
--- a/tools/perf/util/env.c
+++ b/tools/perf/util/env.c
@@ -16,6 +16,7 @@ struct perf_env perf_env;
 
 #ifdef HAVE_LIBBPF_SUPPORT
 #include "bpf-event.h"
+#include "bpf-utils.h"
 #include <bpf/libbpf.h>
 
 void perf_env__insert_bpf_prog_info(struct perf_env *env,
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index 1c7414f66655..56511db8fa03 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -48,6 +48,7 @@
 #include "util/util.h" // perf_exe()
 #include "cputopo.h"
 #include "bpf-event.h"
+#include "bpf-utils.h"
 #include "clockid.h"
 #include "pmu-hybrid.h"
 
@@ -1006,17 +1007,17 @@ static int write_bpf_prog_info(struct feat_fd *ff,
 
 		node = rb_entry(next, struct bpf_prog_info_node, rb_node);
 		next = rb_next(&node->rb_node);
-		len = sizeof(struct bpf_prog_info_linear) +
+		len = sizeof(struct perf_bpil) +
 			node->info_linear->data_len;
 
 		/* before writing to file, translate address to offset */
-		bpf_program__bpil_addr_to_offs(node->info_linear);
+		bpil_addr_to_offs(node->info_linear);
 		ret = do_write(ff, node->info_linear, len);
 		/*
 		 * translate back to address even when do_write() fails,
 		 * so that this function never changes the data.
 		 */
-		bpf_program__bpil_offs_to_addr(node->info_linear);
+		bpil_offs_to_addr(node->info_linear);
 		if (ret < 0)
 			goto out;
 	}
@@ -3018,9 +3019,9 @@ static int process_dir_format(struct feat_fd *ff,
 #ifdef HAVE_LIBBPF_SUPPORT
 static int process_bpf_prog_info(struct feat_fd *ff, void *data __maybe_unused)
 {
-	struct bpf_prog_info_linear *info_linear;
 	struct bpf_prog_info_node *info_node;
 	struct perf_env *env = &ff->ph->env;
+	struct perf_bpil *info_linear;
 	u32 count, i;
 	int err = -1;
 
@@ -3049,7 +3050,7 @@ static int process_bpf_prog_info(struct feat_fd *ff, void *data __maybe_unused)
 			goto out;
 		}
 
-		info_linear = malloc(sizeof(struct bpf_prog_info_linear) +
+		info_linear = malloc(sizeof(struct perf_bpil) +
 				     data_len);
 		if (!info_linear)
 			goto out;
@@ -3071,7 +3072,7 @@ static int process_bpf_prog_info(struct feat_fd *ff, void *data __maybe_unused)
 			goto out;
 
 		/* after reading from file, translate offset to address */
-		bpf_program__bpil_offs_to_addr(info_linear);
+		bpil_offs_to_addr(info_linear);
 		info_node->info_linear = info_linear;
 		perf_env__insert_bpf_prog_info(env, info_node);
 	}
-- 
2.30.2

[PATCH v3 bpf-next 4/4] libbpf: deprecate bpf_program__get_prog_info_linear

[Dave Marchevsky]

As part of the road to libbpf 1.0, and discussed in libbpf issue tracker
[0], bpf_program__get_prog_info_linear and its associated structs and
helper functions should be deprecated. The functionality is too specific
to the needs of 'perf', and there's little/no out-of-tree usage to
preclude introduction of a more general helper in the future.

  [0] Closes: https://github.com/libbpf/libbpf/issues/313

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
---
 tools/lib/bpf/libbpf.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index 9de0f299706b..797f5f8a0e20 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -918,12 +918,15 @@ struct bpf_prog_info_linear {
 	__u8			data[];
 };
 
+LIBBPF_DEPRECATED_SINCE(0, 6, "use a custom linear prog_info wrapper")
 LIBBPF_API struct bpf_prog_info_linear *
 bpf_program__get_prog_info_linear(int fd, __u64 arrays);
 
+LIBBPF_DEPRECATED_SINCE(0, 6, "use a custom linear prog_info wrapper")
 LIBBPF_API void
 bpf_program__bpil_addr_to_offs(struct bpf_prog_info_linear *info_linear);
 
+LIBBPF_DEPRECATED_SINCE(0, 6, "use a custom linear prog_info wrapper")
 LIBBPF_API void
 bpf_program__bpil_offs_to_addr(struct bpf_prog_info_linear *info_linear);
 
-- 
2.30.2

Re: [PATCH v3 bpf-next 0/4] libbpf: deprecate bpf_program__get_prog_info_linear

[Andrii Nakryiko]

On Mon, Nov 1, 2021 at 3:46 PM Dave Marchevsky <davemarchevsky@fb.com> wrote:
>
> bpf_program__get_prog_info_linear is a helper which wraps the
> bpf_obj_get_info_by_fd BPF syscall with some niceties that put
> all dynamic-length bpf_prog_info in one buffer contiguous with struct
> bpf_prog_info, and simplify the selection of which dynamic data to grab.
>
> The resultant combined struct, bpf_prog_info_linear, is persisted to
> file by 'perf' to enable later annotation of BPF prog data. libbpf
> includes some vaddr <-> offset conversion helpers for
> struct bpf_prog_info_linear to simplify this.
>
> This functionality is heavily tailored to perf's usecase, so its use as
> a general prog info API should be deemphasized in favor of just calling
> bpf_obj_get_info_by_fd, which can be more easily fit to purpose. Some
> examples from caller migrations in this series:
>
>   * Some callers weren't requesting or using dynamic-sized prog info and
>     are well served by a simple get_info_by_fd call (e.g.
>     dump_prog_id_as_func_ptr in bpftool)
>   * Some callers were requesting all of a specific dynamic info type but
>     only using the first record, so can avoid unnecessary malloc by
>     only requesting 1 (e.g. profile_target_name in bpftool)
>   * bpftool's do_dump saves some malloc/free by growing and reusing its
>     dynamic prog_info buf as it loops over progs to grab info and dump.
>
> Perf does need the full functionality of
> bpf_program__get_prog_info_linear and its accompanying structs +
> helpers, so copy the code to its codebase, migrate all other uses in the
> tree, and deprecate the helper in libbpf.
>
> Since the deprecated symbols continue to be included in perf some
> renaming was necessary in perf's copy, otherwise functionality is
> unchanged.
>
> This work was previously discussed in libbpf's issue tracker [0].
>
> [0]: https://github.com/libbpf/libbpf/issues/313
>
> v2->v3:
>   * Remove v2's patch 1 ("libbpf: Migrate internal use of
>     bpf_program__get_prog_info_linear"), which was applied [Andrii]
>   * Add new patch 1 migrating error checking of libbpf calls to
>     new scheme [Andrii, Quentin]
>   * In patch 2, fix != -1 error check of libbpf call, improper realloc
>     handling, and get rid of confusing macros [Andrii]
>   * In patch 4, deprecate starting from 0.6 instead of 0.7 [Andrii]

LGTM. Quentin, can you please take a look and ack as well? Thanks!

>
> v1->v2: fix bpftool do_dump changes to clear bpf_prog_info after use and
> correctly pass realloc'd ptr back (patch 2)
>
> Dave Marchevsky (4):
>   bpftool: Migrate -1 err checks of libbpf fn calls
>   bpftool: use bpf_obj_get_info_by_fd directly
>   perf: pull in bpf_program__get_prog_info_linear
>   libbpf: deprecate bpf_program__get_prog_info_linear
>
>  tools/bpf/bpftool/btf_dumper.c                |  42 +--
>  tools/bpf/bpftool/prog.c                      | 158 ++++++++---
>  tools/bpf/bpftool/struct_ops.c                |   2 +-
>  tools/lib/bpf/libbpf.h                        |   3 +
>  .../Documentation/perf.data-file-format.txt   |   2 +-
>  tools/perf/util/Build                         |   1 +
>  tools/perf/util/annotate.c                    |   3 +-
>  tools/perf/util/bpf-event.c                   |  41 ++-
>  tools/perf/util/bpf-event.h                   |   2 +-
>  tools/perf/util/bpf-utils.c                   | 261 ++++++++++++++++++
>  tools/perf/util/bpf-utils.h                   |  76 +++++
>  tools/perf/util/bpf_counter.c                 |   6 +-
>  tools/perf/util/dso.c                         |   1 +
>  tools/perf/util/env.c                         |   1 +
>  tools/perf/util/header.c                      |  13 +-
>  15 files changed, 527 insertions(+), 85 deletions(-)
>  create mode 100644 tools/perf/util/bpf-utils.c
>  create mode 100644 tools/perf/util/bpf-utils.h
>
> --
> 2.30.2
>

Re: [PATCH v3 bpf-next 0/4] libbpf: deprecate bpf_program__get_prog_info_linear

[Quentin Monnet]

2021-11-02 16:06 UTC-0700 ~ Andrii Nakryiko <andrii.nakryiko@gmail.com>
> On Mon, Nov 1, 2021 at 3:46 PM Dave Marchevsky <davemarchevsky@fb.com> wrote:
>>
>> bpf_program__get_prog_info_linear is a helper which wraps the
>> bpf_obj_get_info_by_fd BPF syscall with some niceties that put
>> all dynamic-length bpf_prog_info in one buffer contiguous with struct
>> bpf_prog_info, and simplify the selection of which dynamic data to grab.
>>
>> The resultant combined struct, bpf_prog_info_linear, is persisted to
>> file by 'perf' to enable later annotation of BPF prog data. libbpf
>> includes some vaddr <-> offset conversion helpers for
>> struct bpf_prog_info_linear to simplify this.
>>
>> This functionality is heavily tailored to perf's usecase, so its use as
>> a general prog info API should be deemphasized in favor of just calling
>> bpf_obj_get_info_by_fd, which can be more easily fit to purpose. Some
>> examples from caller migrations in this series:
>>
>>   * Some callers weren't requesting or using dynamic-sized prog info and
>>     are well served by a simple get_info_by_fd call (e.g.
>>     dump_prog_id_as_func_ptr in bpftool)
>>   * Some callers were requesting all of a specific dynamic info type but
>>     only using the first record, so can avoid unnecessary malloc by
>>     only requesting 1 (e.g. profile_target_name in bpftool)
>>   * bpftool's do_dump saves some malloc/free by growing and reusing its
>>     dynamic prog_info buf as it loops over progs to grab info and dump.
>>
>> Perf does need the full functionality of
>> bpf_program__get_prog_info_linear and its accompanying structs +
>> helpers, so copy the code to its codebase, migrate all other uses in the
>> tree, and deprecate the helper in libbpf.
>>
>> Since the deprecated symbols continue to be included in perf some
>> renaming was necessary in perf's copy, otherwise functionality is
>> unchanged.
>>
>> This work was previously discussed in libbpf's issue tracker [0].
>>
>> [0]: https://github.com/libbpf/libbpf/issues/313
>>
>> v2->v3:
>>   * Remove v2's patch 1 ("libbpf: Migrate internal use of
>>     bpf_program__get_prog_info_linear"), which was applied [Andrii]
>>   * Add new patch 1 migrating error checking of libbpf calls to
>>     new scheme [Andrii, Quentin]
>>   * In patch 2, fix != -1 error check of libbpf call, improper realloc
>>     handling, and get rid of confusing macros [Andrii]
>>   * In patch 4, deprecate starting from 0.6 instead of 0.7 [Andrii]
> 
> LGTM. Quentin, can you please take a look and ack as well? Thanks!

Thanks Andrii for the Cc! I realised yesterday morning that I'd been hit
by the unsubscription incident and missed v3 of this set.

The changes look good to me, and you can add my tag to the first three
patches:

Acked-by: Quentin Monnet <quentin@isovalent.com>

Regarding patch 4, looking at the latest deprecations in libbpf, I would
have expected the functions to be deprecated starting in v0.7, and not v0.6?

Other than that, on patch 2 (apologies for not answering inline), it
would feel more natural, in do_dump()'s "for" loop in prog.c, to have
the memset() above the call to bpf_obj_get_info_by_fd() (and to skip the
zero-initialisation of "info") instead of at the end of the loop, which
means a useless memset() just before we exit the loop. But probably not
worth a respin just for that.

Thanks,
Quentin

Re: [PATCH v3 bpf-next 0/4] libbpf: deprecate bpf_program__get_prog_info_linear

[Andrii Nakryiko]

On Wed, Nov 3, 2021 at 4:26 AM Quentin Monnet <quentin@isovalent.com> wrote:
>
> 2021-11-02 16:06 UTC-0700 ~ Andrii Nakryiko <andrii.nakryiko@gmail.com>
> > On Mon, Nov 1, 2021 at 3:46 PM Dave Marchevsky <davemarchevsky@fb.com> wrote:
> >>
> >> bpf_program__get_prog_info_linear is a helper which wraps the
> >> bpf_obj_get_info_by_fd BPF syscall with some niceties that put
> >> all dynamic-length bpf_prog_info in one buffer contiguous with struct
> >> bpf_prog_info, and simplify the selection of which dynamic data to grab.
> >>
> >> The resultant combined struct, bpf_prog_info_linear, is persisted to
> >> file by 'perf' to enable later annotation of BPF prog data. libbpf
> >> includes some vaddr <-> offset conversion helpers for
> >> struct bpf_prog_info_linear to simplify this.
> >>
> >> This functionality is heavily tailored to perf's usecase, so its use as
> >> a general prog info API should be deemphasized in favor of just calling
> >> bpf_obj_get_info_by_fd, which can be more easily fit to purpose. Some
> >> examples from caller migrations in this series:
> >>
> >>   * Some callers weren't requesting or using dynamic-sized prog info and
> >>     are well served by a simple get_info_by_fd call (e.g.
> >>     dump_prog_id_as_func_ptr in bpftool)
> >>   * Some callers were requesting all of a specific dynamic info type but
> >>     only using the first record, so can avoid unnecessary malloc by
> >>     only requesting 1 (e.g. profile_target_name in bpftool)
> >>   * bpftool's do_dump saves some malloc/free by growing and reusing its
> >>     dynamic prog_info buf as it loops over progs to grab info and dump.
> >>
> >> Perf does need the full functionality of
> >> bpf_program__get_prog_info_linear and its accompanying structs +
> >> helpers, so copy the code to its codebase, migrate all other uses in the
> >> tree, and deprecate the helper in libbpf.
> >>
> >> Since the deprecated symbols continue to be included in perf some
> >> renaming was necessary in perf's copy, otherwise functionality is
> >> unchanged.
> >>
> >> This work was previously discussed in libbpf's issue tracker [0].
> >>
> >> [0]: https://github.com/libbpf/libbpf/issues/313
> >>
> >> v2->v3:
> >>   * Remove v2's patch 1 ("libbpf: Migrate internal use of
> >>     bpf_program__get_prog_info_linear"), which was applied [Andrii]
> >>   * Add new patch 1 migrating error checking of libbpf calls to
> >>     new scheme [Andrii, Quentin]
> >>   * In patch 2, fix != -1 error check of libbpf call, improper realloc
> >>     handling, and get rid of confusing macros [Andrii]
> >>   * In patch 4, deprecate starting from 0.6 instead of 0.7 [Andrii]
> >
> > LGTM. Quentin, can you please take a look and ack as well? Thanks!
>
> Thanks Andrii for the Cc! I realised yesterday morning that I'd been hit
> by the unsubscription incident and missed v3 of this set.

Yeah, super unfortunate this unsubscription. Had to go through that as well.

>
> The changes look good to me, and you can add my tag to the first three
> patches:
>
> Acked-by: Quentin Monnet <quentin@isovalent.com>

Thanks, I will.

>
> Regarding patch 4, looking at the latest deprecations in libbpf, I would
> have expected the functions to be deprecated starting in v0.7, and not v0.6?

The reason to do it in upcoming v0.6 is because there are no
replacement APIs that we are going to wait for. No point in delaying
the inevitable just for the sake of delaying it.

>
> Other than that, on patch 2 (apologies for not answering inline), it
> would feel more natural, in do_dump()'s "for" loop in prog.c, to have
> the memset() above the call to bpf_obj_get_info_by_fd() (and to skip the
> zero-initialisation of "info") instead of at the end of the loop, which
> means a useless memset() just before we exit the loop. But probably not
> worth a respin just for that.

Yeah, that bothered me a bit as well, but if you are also mentioning
that I'll just move it up as you suggested, while applying. Thanks.

>
> Thanks,
> Quentin

Re: [PATCH v3 bpf-next 0/4] libbpf: deprecate bpf_program__get_prog_info_linear

[Andrii Nakryiko]

On Wed, Nov 3, 2021 at 11:21 AM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Wed, Nov 3, 2021 at 4:26 AM Quentin Monnet <quentin@isovalent.com> wrote:
> >
> > 2021-11-02 16:06 UTC-0700 ~ Andrii Nakryiko <andrii.nakryiko@gmail.com>
> > > On Mon, Nov 1, 2021 at 3:46 PM Dave Marchevsky <davemarchevsky@fb.com> wrote:
> > >>
> > >> bpf_program__get_prog_info_linear is a helper which wraps the
> > >> bpf_obj_get_info_by_fd BPF syscall with some niceties that put
> > >> all dynamic-length bpf_prog_info in one buffer contiguous with struct
> > >> bpf_prog_info, and simplify the selection of which dynamic data to grab.
> > >>
> > >> The resultant combined struct, bpf_prog_info_linear, is persisted to
> > >> file by 'perf' to enable later annotation of BPF prog data. libbpf
> > >> includes some vaddr <-> offset conversion helpers for
> > >> struct bpf_prog_info_linear to simplify this.
> > >>
> > >> This functionality is heavily tailored to perf's usecase, so its use as
> > >> a general prog info API should be deemphasized in favor of just calling
> > >> bpf_obj_get_info_by_fd, which can be more easily fit to purpose. Some
> > >> examples from caller migrations in this series:
> > >>
> > >>   * Some callers weren't requesting or using dynamic-sized prog info and
> > >>     are well served by a simple get_info_by_fd call (e.g.
> > >>     dump_prog_id_as_func_ptr in bpftool)
> > >>   * Some callers were requesting all of a specific dynamic info type but
> > >>     only using the first record, so can avoid unnecessary malloc by
> > >>     only requesting 1 (e.g. profile_target_name in bpftool)
> > >>   * bpftool's do_dump saves some malloc/free by growing and reusing its
> > >>     dynamic prog_info buf as it loops over progs to grab info and dump.
> > >>
> > >> Perf does need the full functionality of
> > >> bpf_program__get_prog_info_linear and its accompanying structs +
> > >> helpers, so copy the code to its codebase, migrate all other uses in the
> > >> tree, and deprecate the helper in libbpf.
> > >>
> > >> Since the deprecated symbols continue to be included in perf some
> > >> renaming was necessary in perf's copy, otherwise functionality is
> > >> unchanged.
> > >>
> > >> This work was previously discussed in libbpf's issue tracker [0].
> > >>
> > >> [0]: https://github.com/libbpf/libbpf/issues/313
> > >>
> > >> v2->v3:
> > >>   * Remove v2's patch 1 ("libbpf: Migrate internal use of
> > >>     bpf_program__get_prog_info_linear"), which was applied [Andrii]
> > >>   * Add new patch 1 migrating error checking of libbpf calls to
> > >>     new scheme [Andrii, Quentin]
> > >>   * In patch 2, fix != -1 error check of libbpf call, improper realloc
> > >>     handling, and get rid of confusing macros [Andrii]
> > >>   * In patch 4, deprecate starting from 0.6 instead of 0.7 [Andrii]
> > >
> > > LGTM. Quentin, can you please take a look and ack as well? Thanks!
> >
> > Thanks Andrii for the Cc! I realised yesterday morning that I'd been hit
> > by the unsubscription incident and missed v3 of this set.
>
> Yeah, super unfortunate this unsubscription. Had to go through that as well.
>
> >
> > The changes look good to me, and you can add my tag to the first three
> > patches:
> >
> > Acked-by: Quentin Monnet <quentin@isovalent.com>
>
> Thanks, I will.
>
> >
> > Regarding patch 4, looking at the latest deprecations in libbpf, I would
> > have expected the functions to be deprecated starting in v0.7, and not v0.6?
>
> The reason to do it in upcoming v0.6 is because there are no
> replacement APIs that we are going to wait for. No point in delaying
> the inevitable just for the sake of delaying it.
>
> >
> > Other than that, on patch 2 (apologies for not answering inline), it
> > would feel more natural, in do_dump()'s "for" loop in prog.c, to have
> > the memset() above the call to bpf_obj_get_info_by_fd() (and to skip the
> > zero-initialisation of "info") instead of at the end of the loop, which
> > means a useless memset() just before we exit the loop. But probably not
> > worth a respin just for that.
>
> Yeah, that bothered me a bit as well, but if you are also mentioning
> that I'll just move it up as you suggested, while applying. Thanks.
>

Applied to bpf-next, thanks.

> >
> > Thanks,
> > Quentin