The benchmark runs a loop 5000 times. In the loop it reads the file name
from kprobe argument into stack by using bpf_probe_read_kernel_str(),
and compares the file name with a target character or string.
Three cases are compared: only compare one character, compare the whole
string by a home-made strncmp() and compare the whole string by
bpf_strcmp().
The following is the result:
x86-64 host:
one character: 2613499 ns
whole str by strncmp: 2920348 ns
whole str by helper: 2779332 ns
arm64 host:
one character: 3898867 ns
whole str by strncmp: 4396787 ns
whole str by helper: 3968113 ns
Compared with home-made strncmp, the performance of bpf_strncmp helper
improves 80% under x86-64 and 600% under arm64. The big performance win
on arm64 may comes from its arch-optimized strncmp().
Signed-off-by: Hou Tao <houtao1@xxxxxxxxxx>
---
.../bpf/prog_tests/test_strncmp_helper.c | 75 ++++++++++++
.../selftests/bpf/progs/strncmp_helper.c | 109 ++++++++++++++++++
2 files changed, 184 insertions(+)
create mode 100644 tools/testing/selftests/bpf/prog_tests/test_strncmp_helper.c
create mode 100644 tools/testing/selftests/bpf/progs/strncmp_helper.c
diff --git a/tools/testing/selftests/bpf/prog_tests/test_strncmp_helper.c b/tools/testing/selftests/bpf/prog_tests/test_strncmp_helper.c
new file mode 100644
index 000000000000..1b48ef3402ff
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/test_strncmp_helper.c
@@ -0,0 +1,75 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2021. Huawei Technologies Co., Ltd */
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <time.h>
+#include <test_progs.h>
+
+#include "strncmp_helper.skel.h"
+
+static void run_strncmp_bench(struct strncmp_helper *skel, const char *name,
+ int fd, int loop)
+{
+ struct timespec begin, end;
+ struct stat stat;
+ double nsecs;
+ int i;
+
+ skel->bss->equal = 0;
+ clock_gettime(CLOCK_MONOTONIC, &begin);
+ for (i = 0; i < loop; i++)
+ fstat(fd, &stat);
+ clock_gettime(CLOCK_MONOTONIC, &end);
+
+ nsecs = (end.tv_sec - begin.tv_sec) * 1e9 + (end.tv_nsec - begin.tv_nsec);
+ fprintf(stdout, "%s: loop %d nsecs %.0f\n", name, loop, nsecs);
+ fprintf(stdout, "equal nr %u\n", skel->bss->equal);
+}
+
+void test_test_strncmp_helper(void)
+{
+ const char *fpath = "/tmp/1234123412341234123412341234123412341234";
+ struct strncmp_helper *skel;
+ struct bpf_link *link;
+ int fd, loop;
+
+ skel = strncmp_helper__open_and_load();
+ if (!ASSERT_OK_PTR(skel, "helper load"))
+ return;
+
+ fd = open(fpath, O_CREAT | O_RDONLY, 0644);
+ if (!ASSERT_GE(fd, 0, "create file"))
+ goto close_prog;
+
+ loop = 5000;
+ skel->bss->pid = getpid();
+
+ link = bpf_program__attach(skel->progs.vfs_getattr_nocmp);
+ if (!ASSERT_EQ(libbpf_get_error(link), 0, "attach nocmp"))
+ goto clean_file;
+
+ run_strncmp_bench(skel, "nocmp", fd, loop);
+ bpf_link__destroy(link);
+
+ link = bpf_program__attach(skel->progs.vfs_getattr_cmp);
+ if (!ASSERT_EQ(libbpf_get_error(link), 0, "attach cmp"))
+ goto clean_file;
+
+ run_strncmp_bench(skel, "cmp", fd, loop);
+ bpf_link__destroy(link);
+
+ link = bpf_program__attach(skel->progs.vfs_getattr_cmp_v2);
+ if (!ASSERT_EQ(libbpf_get_error(link), 0, "attach cmp_v2"))
+ goto clean_file;
+
+ run_strncmp_bench(skel, "cmp_v2", fd, loop);
+ bpf_link__destroy(link);
+
+clean_file:
+ close(fd);
+ unlink(fpath);
+close_prog:
+ strncmp_helper__destroy(skel);
+}
diff --git a/tools/testing/selftests/bpf/progs/strncmp_helper.c b/tools/testing/selftests/bpf/progs/strncmp_helper.c
new file mode 100644
index 000000000000..f212c1f50099
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/strncmp_helper.c
@@ -0,0 +1,109 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2021. Huawei Technologies Co., Ltd */
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/bpf.h>
+#include <bpf/bpf_core_read.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+#define TARGET_FILE_NAME "1234123412341234123412341234123412341234"
+
+char _license[] SEC("license") = "GPL";
+
+__u32 pid = 0;
+__u32 equal = 0;
+
+struct qstr {
+ const char *name;
+} __attribute__((preserve_access_index));
+
+struct dentry {
+ struct qstr d_name;
+} __attribute__((preserve_access_index));
+
+struct path {
+ struct dentry *dentry;
+} __attribute__((preserve_access_index));
+
+static __always_inline int read_file_name(const struct path *path, char *buf,
+ unsigned int len)
+{
+ const char *name;
+ int err;
+
+ if ((bpf_get_current_pid_tgid() >> 32) != pid)
+ return -1;
+
+ name = BPF_CORE_READ(path, dentry, d_name.name);
+ err = bpf_probe_read_kernel_str(buf, len, name);
+ if (err <= 0)
+ return -1;
+
+ return err;
+}
+
+static __always_inline int local_strncmp(const char *s1, const char *s2,
+ unsigned int sz)
+{
+ int ret = 0;
+ unsigned int i;
+
+ for (i = 0; i < sz; i++) {
+ ret = s1[i] - s2[i];
+ if (ret || !s1[i])
+ break;
+ }
+
+ return ret;
+}
+
+SEC("kprobe/vfs_getattr")
+int BPF_KPROBE(vfs_getattr_nocmp, const struct path *path)
+{
+ char buf[64] = {0};
+ int err;
+
+ err = read_file_name(path, buf, sizeof(buf));
+ if (err < 0)
+ return 0;
+
+ if (buf[0] == '1')
+ equal++;
+
+ return 0;
+}
+
+SEC("kprobe/vfs_getattr")
+int BPF_KPROBE(vfs_getattr_cmp, const struct path *path)
+{
+ char buf[64] = {0};
+ int err;
+
+ err = read_file_name(path, buf, sizeof(buf));
+ if (err < 0)
+ return 0;
+
+ err = local_strncmp(TARGET_FILE_NAME, buf, sizeof(buf));
+ if (!err)
+ equal++;
+
+ return 0;
+}
+
+SEC("kprobe/vfs_getattr")
+int BPF_KPROBE(vfs_getattr_cmp_v2, const struct path *path)
+{
+ char buf[64] = {0};
+ int err;
+
+ err = read_file_name(path, buf, sizeof(buf));
+ if (err < 0)
+ return 0;
+
+ err = bpf_strncmp(TARGET_FILE_NAME, buf, sizeof(buf));
+ if (!err)
+ equal++;
+
+ return 0;
+}
--
2.29.2
