Skip to contentSkip to conversation

libbpf: ensure libbpf always opens files with O_CLOEXEC

PATCH Applied Andrii Nakryiko 4 participants 5 messages May 25, 2023 – May 26, 2023

Subsystems (2)
lore
Share & tools
lore

Andrii Nakryiko

author maintainer
#0 lore raw

PATCH 1/2 libbpf: ensure libbpf always opens files with O_CLOEXEC

Make sure that libbpf code always gets FD with O_CLOEXEC flag set, regardless if file is open through open() or fopen(). For the latter this means to add "e" to mode string, which is supported since pretty ancient glibc v2.7.

I also dropped outdated TODO comment in usdt.c, which was already completed.

---
 tools/lib/bpf/btf.c           | 2 +-
 tools/lib/bpf/libbpf.c        | 6 +++---
 tools/lib/bpf/libbpf_probes.c | 2 +-
 tools/lib/bpf/usdt.c          | 5 ++---
 4 files changed, 7 insertions(+), 8 deletions(-)
tools/lib/bpf/btf.c +1 -1
index 0a2c079244b6..8484b563b53d 100644
@@ -1064,7 +1064,7 @@ static struct btf *btf_parse_raw(const char *path, struct btf *base_btf)
 	int err = 0;
 	long sz;
 
-	f = fopen(path, "rb");
+	f = fopen(path, "rbe");
 	if (!f) {
 		err = -errno;
 		goto err_out;
tools/lib/bpf/libbpf.c +3 -3
index 1ceb3a97dadc..60ef4c5e3bee 100644
@@ -4351,7 +4351,7 @@ static int bpf_get_map_info_from_fdinfo(int fd, struct bpf_map_info *info)
 	snprintf(file, sizeof(file), "/proc/%d/fdinfo/%d", getpid(), fd);
 	memset(info, 0, sizeof(*info));
 
-	fp = fopen(file, "r");
+	fp = fopen(file, "re");
 	if (!fp) {
 		err = -errno;
 		pr_warn("failed to open %s: %d. No procfs support?\n", file,
@@ -7455,7 +7455,7 @@ int libbpf_kallsyms_parse(kallsyms_cb_t cb, void *ctx)
 	int ret, err = 0;
 	FILE *f;
 
-	f = fopen("/proc/kallsyms", "r");
+	f = fopen("/proc/kallsyms", "re");
 	if (!f) {
 		err = -errno;
 		pr_warn("failed to open /proc/kallsyms: %d\n", err);
@@ -10075,7 +10075,7 @@ static int parse_uint_from_file(const char *file, const char *fmt)
 	int err, ret;
 	FILE *f;
 
-	f = fopen(file, "r");
+	f = fopen(file, "re");
 	if (!f) {
 		err = -errno;
 		pr_debug("failed to open '%s': %s\n", file,
tools/lib/bpf/libbpf_probes.c +1 -1
index 6065f408a59c..5f78a023b474 100644
@@ -38,7 +38,7 @@ static __u32 get_ubuntu_kernel_version(void)
 	if (faccessat(AT_FDCWD, ubuntu_kver_file, R_OK, AT_EACCESS) != 0)
 		return 0;
 
-	f = fopen(ubuntu_kver_file, "r");
+	f = fopen(ubuntu_kver_file, "re");
 	if (!f)
 		return 0;
 
tools/lib/bpf/usdt.c +2 -3
index 086eef355ab3..f1a141555f08 100644
@@ -466,7 +466,7 @@ static int parse_vma_segs(int pid, const char *lib_path, struct elf_seg **segs,
 
 proceed:
 	sprintf(line, "/proc/%d/maps", pid);
-	f = fopen(line, "r");
+	f = fopen(line, "re");
 	if (!f) {
 		err = -errno;
 		pr_warn("usdt: failed to open '%s' to get base addr of '%s': %d\n",
@@ -954,8 +954,7 @@ struct bpf_link *usdt_manager_attach_usdt(struct usdt_manager *man, const struct
 	spec_map_fd = bpf_map__fd(man->specs_map);
 	ip_map_fd = bpf_map__fd(man->ip_to_spec_id_map);
 
-	/* TODO: perform path resolution similar to uprobe's */
-	fd = open(path, O_RDONLY);
+	fd = open(path, O_RDONLY | O_CLOEXEC);
 	if (fd < 0) {
 		err = -errno;
 		pr_warn("usdt: failed to open ELF binary '%s': %d\n", path, err);
Signature · 1 line
2.34.1
author maintainer
#1 lore raw

PATCH 2/2 libbpf: ensure FD >= 3 during bpf_map__reuse_fd()

Improve bpf_map__reuse_fd() logic and ensure that dup'ed map FD is "good" (>= 3) and has O_CLOEXEC flags. Use fcntl(F_DUPFD_CLOEXEC) for that, similarly to ensure_good_fd() helper we already use in low-level APIs that work with bpf() syscall.

---
 tools/lib/bpf/libbpf.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)
tools/lib/bpf/libbpf.c +6 -7
index 60ef4c5e3bee..47632606b06d 100644
@@ -4414,18 +4414,17 @@ int bpf_map__reuse_fd(struct bpf_map *map, int fd)
 	if (!new_name)
 		return libbpf_err(-errno);
 
-	new_fd = open("/", O_RDONLY | O_CLOEXEC);
+	/*
+	 * Like dup(), but make sure new FD is >= 3 and has O_CLOEXEC set.
+	 * This is similar to what we do in ensure_good_fd(), but without
+	 * closing original FD.
+	 */
+	new_fd = fcntl(fd, F_DUPFD_CLOEXEC, 3);
 	if (new_fd < 0) {
 		err = -errno;
 		goto err_free_new_name;
 	}
 
-	new_fd = dup3(fd, new_fd, O_CLOEXEC);
-	if (new_fd < 0) {
-		err = -errno;
-		goto err_close_new_fd;
-	}
-
 	err = zclose(map->fd);
 	if (err) {
 		err = -errno;
Signature · 1 line
2.34.1
#2 lore raw
In reply to Andrii Nakryiko
Quoted 21 lines · Andrii Nakryiko

Make sure that libbpf code always gets FD with O_CLOEXEC flag set, regardless if file is open through open() or fopen(). For the latter this means to add "e" to mode string, which is supported since pretty ancient glibc v2.7.

I also dropped outdated TODO comment in usdt.c, which was already completed.

---
 tools/lib/bpf/btf.c           | 2 +-
 tools/lib/bpf/libbpf.c        | 6 +++---
 tools/lib/bpf/libbpf_probes.c | 2 +-
 tools/lib/bpf/usdt.c          | 5 ++---
 4 files changed, 7 insertions(+), 8 deletions(-)
tools/lib/bpf/btf.c +0 -0
index 0a2c079244b6..8484b563b53d 100644

tools/lib/bpf/btf.c static struct btf *btf_parse_raw(const char *path, struct btf *base_btf) · line 1067

Quoted 5 lines
tools/lib/bpf/btf.c +1 -1
@@ -1064,7 +1064,7 @@ static struct btf *btf_parse_raw(const char *path, struct btf *base_btf)
 	int err = 0;
 	long sz;
 
-	f = fopen(path, "rb");
+	f = fopen(path, "rbe");

You might as well drop the "b". That's a thing only on non-POSIX systems. So unless you want to support windows with this, you can drop it with zero effect.

Thanks for doing this, much appreciated!

Lennart

--
Lennart Poettering, Berlin

maintainer
#3 lore raw
Quoted 26 lines · Lennart Poettering
Quoted 24 lines · Andrii Nakryiko

Make sure that libbpf code always gets FD with O_CLOEXEC flag set, regardless if file is open through open() or fopen(). For the latter this means to add "e" to mode string, which is supported since pretty ancient glibc v2.7.

I also dropped outdated TODO comment in usdt.c, which was already completed.

---
  tools/lib/bpf/btf.c           | 2 +-
  tools/lib/bpf/libbpf.c        | 6 +++---
  tools/lib/bpf/libbpf_probes.c | 2 +-
  tools/lib/bpf/usdt.c          | 5 ++---
  4 files changed, 7 insertions(+), 8 deletions(-)
tools/lib/bpf/btf.c +0 -0
index 0a2c079244b6..8484b563b53d 100644
@@ -1064,7 +1064,7 @@ static struct btf *btf_parse_raw(const char *path, struct btf *base_btf)
  	int err = 0;
  	long sz;
 
Quoted 6 lines

tools/lib/bpf/btf.c static struct btf *btf_parse_raw(const char *path, struct btf *base_btf) · line 1067

Quoted 2 lines
tools/lib/bpf/btf.c +1 -1
-	f = fopen(path, "rb");
+	f = fopen(path, "rbe");

You might as well drop the "b". That's a thing only on non-POSIX systems. So unless you want to support windows with this, you can drop it with zero effect.

Iiuc, the library is also imported by the 'ebpf for windows' project [0], so we might need to keep the 'b' intact in that case.

Thanks,
Daniel

[0] https://github.com/microsoft/ebpf-for-windows/#architectural-overview

#4 lore raw
In reply to Andrii Nakryiko
This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:

Hello:

This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:

Quoted 8 lines · On Thu, 25 May 2023 15:13:10 -0700 you wrote:

Make sure that libbpf code always gets FD with O_CLOEXEC flag set, regardless if file is open through open() or fopen(). For the latter this means to add "e" to mode string, which is supported since pretty ancient glibc v2.7.

I also dropped outdated TODO comment in usdt.c, which was already completed.

[...]

Here is the summary with links:
- [bpf-next,1/2] libbpf: ensure libbpf always opens files with O_CLOEXEC

    https://git.kernel.org/bpf/bpf-next/c/59842c5451fe
  - [bpf-next,2/2] libbpf: ensure FD >= 3 during bpf_map__reuse_fd()
    https://git.kernel.org/bpf/bpf-next/c/4aadd2920b81

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html