Skip to contentSkip to conversation

Fix leaks in libbpf and selftests

PATCH Applied Andrii Nakryiko 2 participants 11 messages Nov 7, 2021

Subsystems (4)
lore
Share & tools
lore

Andrii Nakryiko

author maintainer
#0 lore raw

v4 cover Fix leaks in libbpf and selftests

Fix all the memory leaks reported by ASAN. All but one are just improper resource clean up in selftests. But one memory leak was discovered in libbpf, leaving inner map's name leaked.

First patch fixes selftests' Makefile by passing through SAN_CFLAGS to linker. Without that compiling with SAN_CFLAGS=-fsanitize=address kept failing.

Running selftests under ASAN in BPF CI is the next step, we just need to make sure all the necessary libraries (libasan and liblsan) are installed on the host and inside the VM. Would be great to get some help with that, but for now make sure that test_progs run is clean from leak sanitizer errors.

v3->v4:
- rebase on latest bpf-next;
v2->v3:
- fix per-cpu array memory leaks in btf_iter.c selftests (Hengqi);
v1->v2:
- call bpf_map__destroy() conditionally if map->inner_map is present.

Andrii Nakryiko (9):
selftests/bpf: pass sanitizer flags to linker through LDFLAGS
libbpf: free up resources used by inner map definition selftests/bpf: fix memory leaks in btf_type_c_dump() helper selftests/bpf: free per-cpu values array in bpf_iter selftest selftests/bpf: free inner strings index in btf selftest selftests/bpf: clean up btf and btf_dump in dump_datasec test selftests/bpf: avoid duplicate btf__parse() call selftests/bpf: destroy XDP link correctly selftests/bpf: fix bpf_object leak in skb_ctx selftest

 tools/lib/bpf/libbpf.c                                   | 5 ++++-
 tools/testing/selftests/bpf/Makefile                     | 1 +
 tools/testing/selftests/bpf/btf_helpers.c                | 9 +++++++--
 tools/testing/selftests/bpf/prog_tests/bpf_iter.c        | 8 ++++----
 tools/testing/selftests/bpf/prog_tests/btf.c             | 6 ++----
 tools/testing/selftests/bpf/prog_tests/btf_dump.c        | 8 ++++++--
 tools/testing/selftests/bpf/prog_tests/core_reloc.c      | 2 +-
 .../testing/selftests/bpf/prog_tests/migrate_reuseport.c | 4 ++--
 tools/testing/selftests/bpf/prog_tests/skb_ctx.c         | 2 ++
 9 files changed, 29 insertions(+), 16 deletions(-)
Signature · 1 line
2.30.2
author maintainer
#1 lore raw

PATCH v4 1/9 selftests/bpf: pass sanitizer flags to linker through LDFLAGS

When adding -fsanitize=address to SAN_CFLAGS, it has to be passed both to compiler through CFLAGS as well as linker through LDFLAGS. Add SAN_CFLAGS into LDFLAGS to allow building selftests with ASAN.

---
 tools/testing/selftests/bpf/Makefile | 1 +
 1 file changed, 1 insertion(+)
tools/testing/selftests/bpf/Makefile +1 -0
index e19cc6473936..0468ea57650d 100644
@@ -24,6 +24,7 @@ SAN_CFLAGS	?=
 CFLAGS += -g -O0 -rdynamic -Wall $(GENFLAGS) $(SAN_CFLAGS)		\
 	  -I$(CURDIR) -I$(INCLUDE_DIR) -I$(GENDIR) -I$(LIBDIR)		\
 	  -I$(TOOLSINCDIR) -I$(APIDIR) -I$(OUTPUT)
+LDFLAGS += $(SAN_CFLAGS)
 LDLIBS += -lcap -lelf -lz -lrt -lpthread
 
 # Silence some warnings when compiled with clang
Signature · 1 line
2.30.2
author maintainer
#2 lore raw

PATCH v4 2/9 libbpf: free up resources used by inner map definition

In reply to Andrii Nakryiko

It's not enough to just free(map->inner_map), as inner_map itself can have extra memory allocated, like map name.

Fixes 646f02ffdd49 ("libbpf: Add BTF-defined map-in-map support")
---
 tools/lib/bpf/libbpf.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
tools/lib/bpf/libbpf.c +4 -1
index dfd15cc60ea7..d869ebee1e27 100644
@@ -9009,7 +9009,10 @@ int bpf_map__set_inner_map_fd(struct bpf_map *map, int fd)
 		pr_warn("error: inner_map_fd already specified\n");
 		return libbpf_err(-EINVAL);
 	}
-	zfree(&map->inner_map);
+	if (map->inner_map) {
+		bpf_map__destroy(map->inner_map);
+		zfree(&map->inner_map);
+	}
 	map->inner_map_fd = fd;
 	return 0;
 }
Signature · 1 line
2.30.2
author maintainer
#3 lore raw

PATCH v4 3/9 selftests/bpf: fix memory leaks in btf_type_c_dump() helper

In reply to Andrii Nakryiko

Free up memory and resources used by temporary allocated memstream and btf_dump instance.

---
 tools/testing/selftests/bpf/btf_helpers.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/btf_helpers.c +7 -2
index b5b6b013a245..3d1a748d09d8 100644
@@ -251,18 +251,23 @@ const char *btf_type_c_dump(const struct btf *btf)
 	d = btf_dump__new(btf, NULL, &opts, btf_dump_printf);
 	if (libbpf_get_error(d)) {
 		fprintf(stderr, "Failed to create btf_dump instance: %ld\n", libbpf_get_error(d));
-		return NULL;
+		goto err_out;
 	}
 
 	for (i = 1; i < btf__type_cnt(btf); i++) {
 		err = btf_dump__dump_type(d, i);
 		if (err) {
 			fprintf(stderr, "Failed to dump type [%d]: %d\n", i, err);
-			return NULL;
+			goto err_out;
 		}
 	}
 
+	btf_dump__free(d);
 	fflush(buf_file);
 	fclose(buf_file);
 	return buf;
+err_out:
+	btf_dump__free(d);
+	fclose(buf_file);
+	return NULL;
 }
Signature · 1 line
2.30.2
author maintainer
#4 lore raw

PATCH v4 4/9 selftests/bpf: free per-cpu values array in bpf_iter selftest

In reply to Andrii Nakryiko

Array holding per-cpu values wasn't freed. Fix that.

---
 tools/testing/selftests/bpf/prog_tests/bpf_iter.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +4 -4
index 9454331aaf85..3e10abce3e5a 100644
@@ -699,14 +699,13 @@ static void test_bpf_percpu_hash_map(void)
 	char buf[64];
 	void *val;
 
-	val = malloc(8 * bpf_num_possible_cpus());
-
 	skel = bpf_iter_bpf_percpu_hash_map__open();
 	if (CHECK(!skel, "bpf_iter_bpf_percpu_hash_map__open",
 		  "skeleton open failed\n"))
 		return;
 
 	skel->rodata->num_cpus = bpf_num_possible_cpus();
+	val = malloc(8 * bpf_num_possible_cpus());
 
 	err = bpf_iter_bpf_percpu_hash_map__load(skel);
 	if (CHECK(!skel, "bpf_iter_bpf_percpu_hash_map__load",
@@ -770,6 +769,7 @@ static void test_bpf_percpu_hash_map(void)
 	bpf_link__destroy(link);
 out:
 	bpf_iter_bpf_percpu_hash_map__destroy(skel);
+	free(val);
 }
 
 static void test_bpf_array_map(void)
@@ -870,14 +870,13 @@ static void test_bpf_percpu_array_map(void)
 	void *val;
 	int len;
 
-	val = malloc(8 * bpf_num_possible_cpus());
-
 	skel = bpf_iter_bpf_percpu_array_map__open();
 	if (CHECK(!skel, "bpf_iter_bpf_percpu_array_map__open",
 		  "skeleton open failed\n"))
 		return;
 
 	skel->rodata->num_cpus = bpf_num_possible_cpus();
+	val = malloc(8 * bpf_num_possible_cpus());
 
 	err = bpf_iter_bpf_percpu_array_map__load(skel);
 	if (CHECK(!skel, "bpf_iter_bpf_percpu_array_map__load",
@@ -933,6 +932,7 @@ static void test_bpf_percpu_array_map(void)
 	bpf_link__destroy(link);
 out:
 	bpf_iter_bpf_percpu_array_map__destroy(skel);
+	free(val);
 }
 
 /* An iterator program deletes all local storage in a map. */
Signature · 1 line
2.30.2
author maintainer
#5 lore raw

PATCH v4 5/9 selftests/bpf: free inner strings index in btf selftest

In reply to Andrii Nakryiko

Inner array of allocated strings wasn't freed on success. Now it's always freed.

---
 tools/testing/selftests/bpf/prog_tests/btf.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
tools/testing/selftests/bpf/prog_tests/btf.c +2 -4
index ac596cb06e40..ebd1aa4d09d6 100644
@@ -4046,11 +4046,9 @@ static void *btf_raw_create(const struct btf_header *hdr,
 			next_str_idx < strs_cnt ? strs_idx[next_str_idx] : NULL;
 
 done:
+	free(strs_idx);
 	if (err) {
-		if (raw_btf)
-			free(raw_btf);
-		if (strs_idx)
-			free(strs_idx);
+		free(raw_btf);
 		return NULL;
 	}
 	return raw_btf;
Signature · 1 line
2.30.2
author maintainer
#6 lore raw

PATCH v4 6/9 selftests/bpf: clean up btf and btf_dump in dump_datasec test

In reply to Andrii Nakryiko

Free up used resources at the end and on error. Also make it more obvious that there is btf__parse() call that creates struct btf instance.

---
 tools/testing/selftests/bpf/prog_tests/btf_dump.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/prog_tests/btf_dump.c +6 -2
index aa76360d8f49..a04961942dfa 100644
@@ -814,21 +814,25 @@ static void test_btf_datasec(struct btf *btf, struct btf_dump *d, char *str,
 
 static void test_btf_dump_datasec_data(char *str)
 {
-	struct btf *btf = btf__parse("xdping_kern.o", NULL);
+	struct btf *btf;
 	struct btf_dump_opts opts = { .ctx = str };
 	char license[4] = "GPL";
 	struct btf_dump *d;
 
+	btf = btf__parse("xdping_kern.o", NULL);
 	if (!ASSERT_OK_PTR(btf, "xdping_kern.o BTF not found"))
 		return;
 
 	d = btf_dump__new(btf, NULL, &opts, btf_dump_snprintf);
 	if (!ASSERT_OK_PTR(d, "could not create BTF dump"))
-		return;
+		goto out;
 
 	test_btf_datasec(btf, d, str, "license",
 			 "SEC(\"license\") char[4] _license = (char[4])['G','P','L',];",
 			 license, sizeof(license));
+out:
+	btf_dump__free(d);
+	btf__free(btf);
 }
 
 void test_btf_dump() {
Signature · 1 line
2.30.2
author maintainer
#7 lore raw

PATCH v4 7/9 selftests/bpf: avoid duplicate btf__parse() call

In reply to Andrii Nakryiko

btf__parse() is repeated after successful setup, leaving the first instance leaked. Remove redundant and premature call.

---
 tools/testing/selftests/bpf/prog_tests/core_reloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/bpf/prog_tests/core_reloc.c +1 -1
index 55ec85ba7375..1041d0c593f6 100644
@@ -433,7 +433,7 @@ static int setup_type_id_case_local(struct core_reloc_test_case *test)
 
 static int setup_type_id_case_success(struct core_reloc_test_case *test) {
 	struct core_reloc_type_id_output *exp = (void *)test->output;
-	struct btf *targ_btf = btf__parse(test->btf_src_file, NULL);
+	struct btf *targ_btf;
 	int err;
 
 	err = setup_type_id_case_local(test);
Signature · 1 line
2.30.2
author maintainer
#8 lore raw

PATCH v4 8/9 selftests/bpf: destroy XDP link correctly

In reply to Andrii Nakryiko

bpf_link__detach() was confused with bpf_link__destroy() and leaves leaked FD in the process. Fix the problem.

---
 tools/testing/selftests/bpf/prog_tests/migrate_reuseport.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/prog_tests/migrate_reuseport.c +2 -2
index 7589c03fd26b..eb2feaac81fe 100644
@@ -204,8 +204,8 @@ static int pass_ack(struct migrate_reuseport_test_case *test_case)
 {
 	int err;
 
-	err = bpf_link__detach(test_case->link);
-	if (!ASSERT_OK(err, "bpf_link__detach"))
+	err = bpf_link__destroy(test_case->link);
+	if (!ASSERT_OK(err, "bpf_link__destroy"))
 		return -1;
 
 	test_case->link = NULL;
Signature · 1 line
2.30.2
author maintainer
#9 lore raw

PATCH v4 9/9 selftests/bpf: fix bpf_object leak in skb_ctx selftest

In reply to Andrii Nakryiko

skb_ctx selftest didn't close bpf_object implicitly allocated by bpf_prog_test_load() helper. Fix the problem by explicitly calling bpf_object__close() at the end of the test.

---
 tools/testing/selftests/bpf/prog_tests/skb_ctx.c | 2 ++
 1 file changed, 2 insertions(+)
tools/testing/selftests/bpf/prog_tests/skb_ctx.c +2 -0
index d3106078838c..b5319ba2ee27 100644
@@ -111,4 +111,6 @@ void test_skb_ctx(void)
 		   "ctx_out_mark",
 		   "skb->mark == %u, expected %d\n",
 		   skb.mark, 10);
+
+	bpf_object__close(obj);
 }
Signature · 1 line
2.30.2
#10 lore raw
In reply to Andrii Nakryiko
Quoted 11 lines · Andrii Nakryiko

Fix all the memory leaks reported by ASAN. All but one are just improper resource clean up in selftests. But one memory leak was discovered in libbpf, leaving inner map's name leaked.

First patch fixes selftests' Makefile by passing through SAN_CFLAGS to linker. Without that compiling with SAN_CFLAGS=-fsanitize=address kept failing.

Running selftests under ASAN in BPF CI is the next step, we just need to make sure all the necessary libraries (libasan and liblsan) are installed on the host and inside the VM. Would be great to get some help with that, but for now make sure that test_progs run is clean from leak sanitizer errors.

Applied. Thanks