[PATCH bpf-next v4 0/3] libbpf: use func name when pinning programs with LIBBPF_STRICT_SEC_NAME

Commit 15669e1dcd75 ("selftests/bpf: Normalize all the rest SEC() uses")
broke flow dissector tests. With the strict section names, bpftool isn't
able to pin all programs of the objects (all section names are the
same now). To bring it back to life let's do the following:

- teach libbpf to pin by func name with LIBBPF_STRICT_SEC_NAME
- enable strict mode in bpftool (breaking cli change)
- fix custom flow_dissector loader to use strict mode
- fix flow_dissector tests to use new pin names (func vs sec)

v4:
- fix comment spelling (Quentin Monnet)
- retry progtype without / (Quentin Monnet)

v3:
- clarify program pinning in LIBBPF_STRICT_SEC_NAME,
  for real this time (Andrii Nakryiko)
- fix possible segfault in __bpf_program__pin_name (Andrii Nakryiko)

v2:
- add github issue (Andrii Nakryiko)
- remove sec_name from bpf_program.pin_name comment (Andrii Nakryiko)
- add cover letter (Andrii Nakryiko)

Stanislav Fomichev (3):
  libbpf: use func name when pinning programs with
    LIBBPF_STRICT_SEC_NAME
  bpftool: conditionally append / to the progtype
  selftests/bpf: fix flow dissector tests

 tools/bpf/bpftool/main.c                       |  4 ++++
 tools/bpf/bpftool/prog.c                       |  9 +++++++--
 tools/lib/bpf/libbpf.c                         | 13 +++++++++++--
 tools/lib/bpf/libbpf_legacy.h                  |  3 +++
 .../selftests/bpf/flow_dissector_load.c        | 18 +++++++++++-------
 .../selftests/bpf/flow_dissector_load.h        | 10 ++--------
 .../selftests/bpf/test_flow_dissector.sh       | 10 +++++-----
 7 files changed, 43 insertions(+), 24 deletions(-)

-- 
2.33.0.1079.g6e70778dc9-goog

We can't use section name anymore because they are not unique
and pinning objects with multiple programs with the same
progtype/secname will fail.

Closes: https://github.com/libbpf/libbpf/issues/273
Fixes: 33a2c75c55e2 ("libbpf: add internal pin_name")
Reviewed-by: Quentin Monnet <redacted>
Signed-off-by: Stanislav Fomichev <redacted>
---
 tools/lib/bpf/libbpf.c        | 13 +++++++++++--
 tools/lib/bpf/libbpf_legacy.h |  3 +++
 2 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 760c7e346603..7f48eeb3ca82 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c

@@ -285,7 +285,7 @@ struct bpf_program {
 	size_t sub_insn_off;
 
 	char *name;
-	/* sec_name with / replaced by _; makes recursive pinning
+	/* name with / replaced by _; makes recursive pinning
 	 * in bpf_object__pin_programs easier
 	 */
 	char *pin_name;

@@ -614,7 +614,16 @@ static char *__bpf_program__pin_name(struct bpf_program *prog)
 {
 	char *name, *p;
 
-	name = p = strdup(prog->sec_name);
+	if (libbpf_mode & LIBBPF_STRICT_SEC_NAME)
+		name = strdup(prog->name);
+	else
+		name = strdup(prog->sec_name);
+
+	if (!name)
+		return NULL;
+
+	p = name;
+
 	while ((p = strchr(p, '/')))
 		*p = '_';
 

diff --git a/tools/lib/bpf/libbpf_legacy.h b/tools/lib/bpf/libbpf_legacy.h
index 74e6f860f703..29ccafab11a8 100644
--- a/tools/lib/bpf/libbpf_legacy.h
+++ b/tools/lib/bpf/libbpf_legacy.h

@@ -52,6 +52,9 @@ enum libbpf_strict_mode {
 	 * allowed, with LIBBPF_STRICT_SEC_PREFIX this will become
 	 * unrecognized by libbpf and would have to be just SEC("xdp") and
 	 * SEC("xdp") and SEC("perf_event").
+	 *
+	 * Note, in this mode the program pin path will be based on the
+	 * function name instead of section name.
 	 */
 	LIBBPF_STRICT_SEC_NAME = 0x04,
 

-- 
2.33.0.1079.g6e70778dc9-goog

Otherwise, attaching with bpftool doesn't work with strict section names.

Also, switch to libbpf strict mode to use the latest conventions
(note, I don't think we have any cli api guarantees?).

Cc: Quentin Monnet <redacted>
Signed-off-by: Stanislav Fomichev <redacted>
---
 tools/bpf/bpftool/main.c | 4 ++++
 tools/bpf/bpftool/prog.c | 9 +++++++--
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/tools/bpf/bpftool/main.c b/tools/bpf/bpftool/main.c
index 02eaaf065f65..8223bac1e401 100644
--- a/tools/bpf/bpftool/main.c
+++ b/tools/bpf/bpftool/main.c

@@ -409,6 +409,10 @@ int main(int argc, char **argv)
 	block_mount = false;
 	bin_name = argv[0];
 
+	ret = libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+	if (ret)
+		p_err("failed to enable libbpf strict mode: %d", ret);
+
 	hash_init(prog_table.table);
 	hash_init(map_table.table);
 	hash_init(link_table.table);

diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
index 277d51c4c5d9..b04990588ccf 100644
--- a/tools/bpf/bpftool/prog.c
+++ b/tools/bpf/bpftool/prog.c

@@ -1420,8 +1420,13 @@ static int load_with_options(int argc, char **argv, bool first_prog_only)
 			err = get_prog_type_by_name(type, &common_prog_type,
 						    &expected_attach_type);
 			free(type);
-			if (err < 0)
-				goto err_free_reuse_maps;
+			if (err < 0) {
+				err = get_prog_type_by_name(*argv, &common_prog_type,
+							    &expected_attach_type);
+				if (err < 0)
+
+					goto err_free_reuse_maps;
+			}
 
 			NEXT_ARG();
 		} else if (is_prefix(*argv, "map")) {

-- 
2.33.0.1079.g6e70778dc9-goog

- update custom loader to search by name, not section name
- update bpftool commands to use proper pin path

Signed-off-by: Stanislav Fomichev <redacted>
---
 .../selftests/bpf/flow_dissector_load.c        | 18 +++++++++++-------
 .../selftests/bpf/flow_dissector_load.h        | 10 ++--------
 .../selftests/bpf/test_flow_dissector.sh       | 10 +++++-----
 3 files changed, 18 insertions(+), 20 deletions(-)

diff --git a/tools/testing/selftests/bpf/flow_dissector_load.c b/tools/testing/selftests/bpf/flow_dissector_load.c
index 3fd83b9dc1bf..87fd1aa323a9 100644
--- a/tools/testing/selftests/bpf/flow_dissector_load.c
+++ b/tools/testing/selftests/bpf/flow_dissector_load.c

@@ -17,7 +17,7 @@
 const char *cfg_pin_path = "/sys/fs/bpf/flow_dissector";
 const char *cfg_map_name = "jmp_table";
 bool cfg_attach = true;
-char *cfg_section_name;
+char *cfg_prog_name;
 char *cfg_path_name;
 
 static void load_and_attach_program(void)

@@ -25,7 +25,11 @@ static void load_and_attach_program(void)
 	int prog_fd, ret;
 	struct bpf_object *obj;
 
-	ret = bpf_flow_load(&obj, cfg_path_name, cfg_section_name,
+	ret = libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+	if (ret)
+		error(1, 0, "failed to enable libbpf strict mode: %d", ret);
+
+	ret = bpf_flow_load(&obj, cfg_path_name, cfg_prog_name,
 			    cfg_map_name, NULL, &prog_fd, NULL);
 	if (ret)
 		error(1, 0, "bpf_flow_load %s", cfg_path_name);

@@ -75,15 +79,15 @@ static void parse_opts(int argc, char **argv)
 			break;
 		case 'p':
 			if (cfg_path_name)
-				error(1, 0, "only one prog name can be given");
+				error(1, 0, "only one path can be given");
 
 			cfg_path_name = optarg;
 			break;
 		case 's':
-			if (cfg_section_name)
-				error(1, 0, "only one section can be given");
+			if (cfg_prog_name)
+				error(1, 0, "only one prog can be given");
 
-			cfg_section_name = optarg;
+			cfg_prog_name = optarg;
 			break;
 		}
 	}

@@ -94,7 +98,7 @@ static void parse_opts(int argc, char **argv)
 	if (cfg_attach && !cfg_path_name)
 		error(1, 0, "must provide a path to the BPF program");
 
-	if (cfg_attach && !cfg_section_name)
+	if (cfg_attach && !cfg_prog_name)
 		error(1, 0, "must provide a section name");
 }
 

diff --git a/tools/testing/selftests/bpf/flow_dissector_load.h b/tools/testing/selftests/bpf/flow_dissector_load.h
index 7290401ec172..9d0acc2fc6cc 100644
--- a/tools/testing/selftests/bpf/flow_dissector_load.h
+++ b/tools/testing/selftests/bpf/flow_dissector_load.h

@@ -7,7 +7,7 @@
 
 static inline int bpf_flow_load(struct bpf_object **obj,
 				const char *path,
-				const char *section_name,
+				const char *prog_name,
 				const char *map_name,
 				const char *keys_map_name,
 				int *prog_fd,

@@ -23,13 +23,7 @@ static inline int bpf_flow_load(struct bpf_object **obj,
 	if (ret)
 		return ret;
 
-	main_prog = NULL;
-	bpf_object__for_each_program(prog, *obj) {
-		if (strcmp(section_name, bpf_program__section_name(prog)) == 0) {
-			main_prog = prog;
-			break;
-		}
-	}
+	main_prog = bpf_object__find_program_by_name(*obj, prog_name);
 	if (!main_prog)
 		return -1;
 

diff --git a/tools/testing/selftests/bpf/test_flow_dissector.sh b/tools/testing/selftests/bpf/test_flow_dissector.sh
index 174b72a64a4c..dbd91221727d 100755
--- a/tools/testing/selftests/bpf/test_flow_dissector.sh
+++ b/tools/testing/selftests/bpf/test_flow_dissector.sh

@@ -26,22 +26,22 @@ if [[ -z $(ip netns identify $$) ]]; then
 			type flow_dissector
 
 		if ! unshare --net $bpftool prog attach pinned \
-			/sys/fs/bpf/flow/flow_dissector flow_dissector; then
+			/sys/fs/bpf/flow/_dissect flow_dissector; then
 			echo "Unexpected unsuccessful attach in namespace" >&2
 			err=1
 		fi
 
-		$bpftool prog attach pinned /sys/fs/bpf/flow/flow_dissector \
+		$bpftool prog attach pinned /sys/fs/bpf/flow/_dissect \
 			flow_dissector
 
 		if unshare --net $bpftool prog attach pinned \
-			/sys/fs/bpf/flow/flow_dissector flow_dissector; then
+			/sys/fs/bpf/flow/_dissect flow_dissector; then
 			echo "Unexpected successful attach in namespace" >&2
 			err=1
 		fi
 
 		if ! $bpftool prog detach pinned \
-			/sys/fs/bpf/flow/flow_dissector flow_dissector; then
+			/sys/fs/bpf/flow/_dissect flow_dissector; then
 			echo "Failed to detach flow dissector" >&2
 			err=1
 		fi

@@ -95,7 +95,7 @@ else
 fi
 
 # Attach BPF program
-./flow_dissector_load -p bpf_flow.o -s flow_dissector
+./flow_dissector_load -p bpf_flow.o -s _dissect
 
 # Setup
 tc qdisc add dev lo ingress

-- 
2.33.0.1079.g6e70778dc9-goog

2021-10-21 09:56 UTC-0700 ~ Stanislav Fomichev [off-list ref]

Thanks a lot for the change! Can you please test it for e.g. an XDP
program? You should see that "bpftool prog load prog.o <path> type xdp"
prints a debug message from libbpf about the first attempt (above)
failing, before the second attempt (below) succeeds.

We need to get rid of this message. I think it should be easy, because
we explicitly "ask" for that message in get_prog_type_by_name(), in the
same file, if it fails to load in the first place.

Could you please update get_prog_type_by_name() to take an additional
switch as an argument, to tell if the debug-info should be retrieved
(then first attempt here would skip it, second would keep it)?
An alternative could be to move all the '/' and retries handling to that
function, and I think it would end up in bpftool keeping support for the
legacy object files with the former convention - but that would somewhat
defeat the objectives of the strict mode, so maybe not the best option.

+			if (err < 0) {

We could run the second attempt only on libbpf returning -ESRCH, maybe?

+				err = get_prog_type_by_name(*argv, &common_prog_type,
+							    &expected_attach_type);
+				if (err < 0)
+
+					goto err_free_reuse_maps;
+			}
 
 			NEXT_ARG();
 		} else if (is_prefix(*argv, "map")) {

On Thu, Oct 21, 2021 at 12:55 PM Quentin Monnet [off-list ref] wrote:

2021-10-21 09:56 UTC-0700 ~ Stanislav Fomichev [off-list ref]

quoted

Otherwise, attaching with bpftool doesn't work with strict section names.

Also, switch to libbpf strict mode to use the latest conventions
(note, I don't think we have any cli api guarantees?).

Cc: Quentin Monnet <redacted>
Signed-off-by: Stanislav Fomichev <redacted>
---
 tools/bpf/bpftool/main.c | 4 ++++
 tools/bpf/bpftool/prog.c | 9 +++++++--
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/tools/bpf/bpftool/main.c b/tools/bpf/bpftool/main.c
index 02eaaf065f65..8223bac1e401 100644
--- a/tools/bpf/bpftool/main.c
+++ b/tools/bpf/bpftool/main.c

@@ -409,6 +409,10 @@ int main(int argc, char **argv)
      block_mount = false;
      bin_name = argv[0];

+     ret = libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+     if (ret)
+             p_err("failed to enable libbpf strict mode: %d", ret);
+
      hash_init(prog_table.table);
      hash_init(map_table.table);
      hash_init(link_table.table);

diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
index 277d51c4c5d9..b04990588ccf 100644
--- a/tools/bpf/bpftool/prog.c
+++ b/tools/bpf/bpftool/prog.c

@@ -1420,8 +1420,13 @@ static int load_with_options(int argc, char **argv, bool first_prog_only)
                      err = get_prog_type_by_name(type, &common_prog_type,
                                                  &expected_attach_type);
                      free(type);
-                     if (err < 0)
-                             goto err_free_reuse_maps;

Thanks a lot for the change! Can you please test it for e.g. an XDP
program? You should see that "bpftool prog load prog.o <path> type xdp"
prints a debug message from libbpf about the first attempt (above)
failing, before the second attempt (below) succeeds.

We need to get rid of this message. I think it should be easy, because
we explicitly "ask" for that message in get_prog_type_by_name(), in the
same file, if it fails to load in the first place.

Could you please update get_prog_type_by_name() to take an additional
switch as an argument, to tell if the debug-info should be retrieved
(then first attempt here would skip it, second would keep it)?
An alternative could be to move all the '/' and retries handling to that
function, and I think it would end up in bpftool keeping support for the
legacy object files with the former convention - but that would somewhat
defeat the objectives of the strict mode, so maybe not the best option.

How about we call libbpf_prog_type_by_name with the provided argv
first and then, if it doesn't work, we fallback to appending '\' and
using get_prog_type_by_name ?

quoted

+                     if (err < 0) {

We could run the second attempt only on libbpf returning -ESRCH, maybe?

Not sure it matters here, why not always retry on error?

quoted

+                             err = get_prog_type_by_name(*argv, &common_prog_type,
+                                                         &expected_attach_type);
+                             if (err < 0)
+
+                                     goto err_free_reuse_maps;
+                     }

                      NEXT_ARG();
              } else if (is_prefix(*argv, "map")) {

On Thu, 21 Oct 2021 at 21:40, Stanislav Fomichev [off-list ref] wrote:

On Thu, Oct 21, 2021 at 12:55 PM Quentin Monnet [off-list ref] wrote:

quoted

2021-10-21 09:56 UTC-0700 ~ Stanislav Fomichev [off-list ref]

quoted

Otherwise, attaching with bpftool doesn't work with strict section names.

Also, switch to libbpf strict mode to use the latest conventions
(note, I don't think we have any cli api guarantees?).

Cc: Quentin Monnet <redacted>
Signed-off-by: Stanislav Fomichev <redacted>
---
 tools/bpf/bpftool/main.c | 4 ++++
 tools/bpf/bpftool/prog.c | 9 +++++++--
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/tools/bpf/bpftool/main.c b/tools/bpf/bpftool/main.c
index 02eaaf065f65..8223bac1e401 100644
--- a/tools/bpf/bpftool/main.c
+++ b/tools/bpf/bpftool/main.c

@@ -409,6 +409,10 @@ int main(int argc, char **argv)
      block_mount = false;
      bin_name = argv[0];

+     ret = libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
+     if (ret)
+             p_err("failed to enable libbpf strict mode: %d", ret);
+
      hash_init(prog_table.table);
      hash_init(map_table.table);
      hash_init(link_table.table);

diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
index 277d51c4c5d9..b04990588ccf 100644
--- a/tools/bpf/bpftool/prog.c
+++ b/tools/bpf/bpftool/prog.c

@@ -1420,8 +1420,13 @@ static int load_with_options(int argc, char **argv, bool first_prog_only)
                      err = get_prog_type_by_name(type, &common_prog_type,
                                                  &expected_attach_type);
                      free(type);
-                     if (err < 0)
-                             goto err_free_reuse_maps;

Thanks a lot for the change! Can you please test it for e.g. an XDP
program? You should see that "bpftool prog load prog.o <path> type xdp"
prints a debug message from libbpf about the first attempt (above)
failing, before the second attempt (below) succeeds.

We need to get rid of this message. I think it should be easy, because
we explicitly "ask" for that message in get_prog_type_by_name(), in the
same file, if it fails to load in the first place.

Could you please update get_prog_type_by_name() to take an additional
switch as an argument, to tell if the debug-info should be retrieved
(then first attempt here would skip it, second would keep it)?
An alternative could be to move all the '/' and retries handling to that
function, and I think it would end up in bpftool keeping support for the
legacy object files with the former convention - but that would somewhat
defeat the objectives of the strict mode, so maybe not the best option.

How about we call libbpf_prog_type_by_name with the provided argv
first and then, if it doesn't work, we fallback to appending '\' and
using get_prog_type_by_name ?

Yes it should work, too. Not sure of the order, maybe best to run
_with_ the '/' first, so that the debug message from libbpf (if
neither attempt succeeds) doesn't have the added slash? But that
doesn't matter much anyway.

quoted

quoted

+                     if (err < 0) {

We could run the second attempt only on libbpf returning -ESRCH, maybe?

Not sure it matters here, why not always retry on error?

In case the function failed for some other reason and we knew that
retrying with a '/' wouldn't work any better. But in practice that's
only if the name is NULL, and this wouldn't happen because we would
not reach that point, so right, doesn't matter much.