Skip to contentSkip to conversation

libbpf: always NULL out pobj in bpf_prog_load_xattr

PATCH In review Lorenz Bauer 3 participants 6 messages May 2, 2019 – May 9, 2019

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 libbpf: always NULL out pobj in bpf_prog_load_xattr

Currently, code like the following segfaults if bpf_prog_load_xattr returns an error:

    struct bpf_object *obj;

    err = bpf_prog_load_xattr(&attr, &obj, &prog_fd);
    bpf_object__close(obj);
    if (err)
        ...

Unconditionally reset pobj to NULL at the start of the function to fix this.

---
 tools/lib/bpf/libbpf.c | 2 ++
 1 file changed, 2 insertions(+)
tools/lib/bpf/libbpf.c +2 -0
index 11a65db4b93f..2ddf3212b8f7 100644
@@ -3363,6 +3363,8 @@ int bpf_prog_load_xattr(const struct bpf_prog_load_attr *attr,
 	struct bpf_map *map;
 	int err;
 
+	*pobj = NULL;
+
 	if (!attr)
 		return -EINVAL;
 	if (!attr->file)
Signature · 1 line
2.19.1
#1 lore raw
Quoted 7 lines · Lorenz Bauer

Currently, code like the following segfaults if bpf_prog_load_xattr returns an error:

    struct bpf_object *obj;

    err = bpf_prog_load_xattr(&attr, &obj, &prog_fd);
    bpf_object__close(obj);

This is a bug. err should always be checked first before using obj or prog_fd. If there is err, calling bpf_object__close(NULL) is another redundancy.

Quoted 5 lines
    if (err)
        ...

Unconditionally reset pobj to NULL at the start of the function to fix this.

Hence, this change is unnecessary.

Quoted 22 lines
---
 tools/lib/bpf/libbpf.c | 2 ++
 1 file changed, 2 insertions(+)
tools/lib/bpf/libbpf.c +2 -0
index 11a65db4b93f..2ddf3212b8f7 100644
@@ -3363,6 +3363,8 @@ int bpf_prog_load_xattr(const struct bpf_prog_load_attr *attr,
 	struct bpf_map *map;
 	int err;
 
+	*pobj = NULL;
+
 	if (!attr)
 		return -EINVAL;
 	if (!attr->file)
Signature · 1 line
2.19.1
#2 lore raw
Quoted 5 lines · Martin Lau
Quoted 3 lines · Lorenz Bauer

Currently, code like the following segfaults if bpf_prog_load_xattr returns an error:

Quoted 7 lines
Quoted 4 lines
    struct bpf_object *obj;

    err = bpf_prog_load_xattr(&attr, &obj, &prog_fd);
    bpf_object__close(obj);

This is a bug. err should always be checked first before using obj or prog_fd. If there is err, calling bpf_object__close(NULL) is another redundancy.

Ack, I'll send another patch. The bug is in a couple of places, so I figured that this is something people do naturally.

Quoted 30 lines
Quoted 6 lines
    if (err)
        ...

Unconditionally reset pobj to NULL at the start of the function to fix this.

Hence, this change is unnecessary.

Quoted 22 lines
---
 tools/lib/bpf/libbpf.c | 2 ++
 1 file changed, 2 insertions(+)
tools/lib/bpf/libbpf.c +2 -1
index 11a65db4b93f..2ddf3212b8f7 100644
@@ -3363,6 +3363,8 @@ int bpf_prog_load_xattr(const struct bpf_prog_load_attr *attr,
      struct bpf_map *map;
      int err;
 
+     *pobj = NULL;
+
      if (!attr)
              return -EINVAL;
      if (!attr->file)
--

2.19.1

Signature · 4 lines
Lorenz Bauer  |  Systems Engineer
25 Lavington St., London SE1 0NZ

www.cloudflare.com
#3 lore raw

PATCH v2 1/1 selftests: bpf: initialize bpf_object pointers where needed

In reply to Lorenz Bauer

There are a few tests which call bpf_object__close on uninitialized bpf_object*, which may segfault. Explicitly zero-initialise these pointers to avoid this.

---
 tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c  | 2 +-
 tools/testing/selftests/bpf/prog_tests/task_fd_query_tp.c | 2 +-
 tools/testing/selftests/bpf/prog_tests/tp_attach_query.c  | 3 +++
 3 files changed, 5 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c +1 -1
index 23b159d95c3f..b74e2f6e96d0 100644
@@ -15,7 +15,7 @@ static int libbpf_debug_print(enum libbpf_print_level level,
 static int check_load(const char *file)
 {
 	struct bpf_prog_load_attr attr;
-	struct bpf_object *obj;
+	struct bpf_object *obj = NULL;
 	int err, prog_fd;
 
 	memset(&attr, 0, sizeof(struct bpf_prog_load_attr));
tools/testing/selftests/bpf/prog_tests/task_fd_query_tp.c +1 -1
index d636a4f39476..f9b70e81682b 100644
@@ -9,7 +9,7 @@ static void test_task_fd_query_tp_core(const char *probe_name,
 	struct perf_event_attr attr = {};
 	__u64 probe_offset, probe_addr;
 	__u32 len, prog_id, fd_type;
-	struct bpf_object *obj;
+	struct bpf_object *obj = NULL;
 	__u32 duration = 0;
 	char buf[256];
 
tools/testing/selftests/bpf/prog_tests/tp_attach_query.c +3 -0
index a2f476f91637..fb095e5cd9af 100644
@@ -13,6 +13,9 @@ void test_tp_attach_query(void)
 	struct bpf_prog_info prog_info;
 	char buf[256];
 
+	for (i = 0; i < num_progs; i++)
+		obj[i] = NULL;
+
 	snprintf(buf, sizeof(buf),
 		 "/sys/kernel/debug/tracing/events/sched/sched_switch/id");
 	efd = open(buf, O_RDONLY, 0);
Signature · 1 line
2.19.1
Acked-by on v2 1/1
#4 lore raw

Re: [PATCH bpf v2] selftests: bpf: initialize bpf_object pointers where needed

Quoted 3 lines · Lorenz Bauer

There are a few tests which call bpf_object__close on uninitialized bpf_object*, which may segfault. Explicitly zero-initialise these pointers to avoid this.

#5 lore raw

Re: [PATCH bpf v2] selftests: bpf: initialize bpf_object pointers where needed

Quoted 6 lines · Martin Lau
Quoted 3 lines · Lorenz Bauer

There are a few tests which call bpf_object__close on uninitialized bpf_object*, which may segfault. Explicitly zero-initialise these pointers to avoid this.

Applied. Thanks