p71924506
/
linux_old1
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

samples, bpf: Refactor perf_event user program with libbpf bpf_link 

The bpf_program__attach of libbpf(using bpf_link) is much more intuitive
than the previous method using ioctl.

bpf_program__attach_perf_event manages the enable of perf_event and
attach of BPF programs to it, so there's no neeed to do this
directly with ioctl.

In addition, bpf_link provides consistency in the use of API because it
allows disable (detach, destroy) for multiple events to be treated as
one bpf_link__destroy. Also, bpf_link__destroy manages the close() of
perf_event fd.

This commit refactors samples that attach the bpf program to perf_event
by using libbbpf instead of ioctl. Also the bpf_load in the samples were
removed and migrated to use libbbpf API.

Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20200321100424.1593964-3-danieltimlee@gmail.com
This commit is contained in:
Daniel T. Lee 2020-03-21 19:04:24 +09:00 committed by Daniel Borkmann
parent 24a6034acc
commit aa5e2af660
3 changed files with 160 additions and 83 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
samples/bpf/Makefile
View File

@ -88,8 +88,8 @@ xdp2-objs := xdp1_user.o
xdp_router_ipv4-objs := xdp_router_ipv4_user.o xdp_router_ipv4-objs := xdp_router_ipv4_user.o
test_current_task_under_cgroup-objs := bpf_load.o $(CGROUP_HELPERS) \ test_current_task_under_cgroup-objs := bpf_load.o $(CGROUP_HELPERS) \
test_current_task_under_cgroup_user.o test_current_task_under_cgroup_user.o
trace_event-objs := bpf_load.o trace_event_user.o $(TRACE_HELPERS) trace_event-objs := trace_event_user.o $(TRACE_HELPERS)
sampleip-objs := bpf_load.o sampleip_user.o $(TRACE_HELPERS) sampleip-objs := sampleip_user.o $(TRACE_HELPERS)
tc_l2_redirect-objs := bpf_load.o tc_l2_redirect_user.o tc_l2_redirect-objs := bpf_load.o tc_l2_redirect_user.o
lwt_len_hist-objs := bpf_load.o lwt_len_hist_user.o lwt_len_hist-objs := bpf_load.o lwt_len_hist_user.o
xdp_tx_iptunnel-objs := xdp_tx_iptunnel_user.o xdp_tx_iptunnel-objs := xdp_tx_iptunnel_user.o

98
samples/bpf/sampleip_user.c
View File

@ -10,21 +10,23 @@
#include <errno.h> #include <errno.h>
#include <signal.h> #include <signal.h>
#include <string.h> #include <string.h>
#include <assert.h>
#include <linux/perf_event.h> #include <linux/perf_event.h>
#include <linux/ptrace.h> #include <linux/ptrace.h>
#include <linux/bpf.h> #include <linux/bpf.h>
#include <sys/ioctl.h> #include <bpf/bpf.h>
#include <bpf/libbpf.h> #include <bpf/libbpf.h>
#include "bpf_load.h"
#include "perf-sys.h" #include "perf-sys.h"
#include "trace_helpers.h" #include "trace_helpers.h"
#define __must_check
#include <linux/err.h>
#define DEFAULT_FREQ 99 #define DEFAULT_FREQ 99
#define DEFAULT_SECS 5 #define DEFAULT_SECS 5
#define MAX_IPS 8192 #define MAX_IPS 8192
#define PAGE_OFFSET 0xffff880000000000 #define PAGE_OFFSET 0xffff880000000000
static int map_fd;
static int nr_cpus; static int nr_cpus;
static void usage(void) static void usage(void)
@ -34,9 +36,10 @@ static void usage(void)
printf(" duration # sampling duration (seconds), default 5\n"); printf(" duration # sampling duration (seconds), default 5\n");
} }
static int sampling_start(int *pmu_fd, int freq) static int sampling_start(int freq, struct bpf_program *prog,
struct bpf_link *links[])
{ {
int i; int i, pmu_fd;
struct perf_event_attr pe_sample_attr = { struct perf_event_attr pe_sample_attr = {
.type = PERF_TYPE_SOFTWARE, .type = PERF_TYPE_SOFTWARE,
@ -47,26 +50,30 @@ static int sampling_start(int *pmu_fd, int freq)
}; };
for (i = 0; i < nr_cpus; i++) { for (i = 0; i < nr_cpus; i++) {
pmu_fd[i] = sys_perf_event_open(&pe_sample_attr, -1 /* pid */, i, pmu_fd = sys_perf_event_open(&pe_sample_attr, -1 /* pid */, i,
-1 /* group_fd */, 0 /* flags */); -1 /* group_fd */, 0 /* flags */);
if (pmu_fd[i] < 0) { if (pmu_fd < 0) {
fprintf(stderr, "ERROR: Initializing perf sampling\n"); fprintf(stderr, "ERROR: Initializing perf sampling\n");
return 1; return 1;
} }
assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_SET_BPF, links[i] = bpf_program__attach_perf_event(prog, pmu_fd);
prog_fd[0]) == 0); if (IS_ERR(links[i])) {
assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_ENABLE, 0) == 0); fprintf(stderr, "ERROR: Attach perf event\n");
links[i] = NULL;
close(pmu_fd);
return 1;
}
} }
return 0; return 0;
} }
static void sampling_end(int *pmu_fd) static void sampling_end(struct bpf_link *links[])
{ {
int i; int i;
for (i = 0; i < nr_cpus; i++) for (i = 0; i < nr_cpus; i++)
close(pmu_fd[i]); bpf_link__destroy(links[i]);
} }
struct ipcount { struct ipcount {
@ -128,14 +135,17 @@ static void print_ip_map(int fd)
static void int_exit(int sig) static void int_exit(int sig)
{ {
printf("\n"); printf("\n");
print_ip_map(map_fd[0]); print_ip_map(map_fd);
exit(0); exit(0);
} }
int main(int argc, char **argv) int main(int argc, char **argv)
{ {
int opt, freq = DEFAULT_FREQ, secs = DEFAULT_SECS, error = 1;
struct bpf_object *obj = NULL;
struct bpf_program *prog;
struct bpf_link **links;
char filename[256]; char filename[256];
int *pmu_fd, opt, freq = DEFAULT_FREQ, secs = DEFAULT_SECS;
/* process arguments */ /* process arguments */
while ((opt = getopt(argc, argv, "F:h")) != -1) { while ((opt = getopt(argc, argv, "F:h")) != -1) {
@ -163,38 +173,58 @@ int main(int argc, char **argv)
} }
/* create perf FDs for each CPU */ /* create perf FDs for each CPU */
nr_cpus = sysconf(_SC_NPROCESSORS_CONF); nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
pmu_fd = malloc(nr_cpus * sizeof(int)); links = calloc(nr_cpus, sizeof(struct bpf_link *));
if (pmu_fd == NULL) { if (!links) {
fprintf(stderr, "ERROR: malloc of pmu_fd\n"); fprintf(stderr, "ERROR: malloc of links\n");
return 1; goto cleanup;
}
snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
obj = bpf_object__open_file(filename, NULL);
if (IS_ERR(obj)) {
fprintf(stderr, "ERROR: opening BPF object file failed\n");
obj = NULL;
goto cleanup;
}
prog = bpf_object__find_program_by_name(obj, "do_sample");
if (!prog) {
fprintf(stderr, "ERROR: finding a prog in obj file failed\n");
goto cleanup;
} }
/* load BPF program */ /* load BPF program */
snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]); if (bpf_object__load(obj)) {
if (load_bpf_file(filename)) { fprintf(stderr, "ERROR: loading BPF object file failed\n");
fprintf(stderr, "ERROR: loading BPF program (errno %d):\n", goto cleanup;
errno);
if (strcmp(bpf_log_buf, "") == 0)
fprintf(stderr, "Try: ulimit -l unlimited\n");
else
fprintf(stderr, "%s", bpf_log_buf);
return 1;
} }
map_fd = bpf_object__find_map_fd_by_name(obj, "ip_map");
if (map_fd < 0) {
fprintf(stderr, "ERROR: finding a map in obj file failed\n");
goto cleanup;
}
signal(SIGINT, int_exit); signal(SIGINT, int_exit);
signal(SIGTERM, int_exit); signal(SIGTERM, int_exit);
/* do sampling */ /* do sampling */
printf("Sampling at %d Hertz for %d seconds. Ctrl-C also ends.\n", printf("Sampling at %d Hertz for %d seconds. Ctrl-C also ends.\n",
freq, secs); freq, secs);
if (sampling_start(pmu_fd, freq) != 0) if (sampling_start(freq, prog, links) != 0)
return 1; goto cleanup;
sleep(secs); sleep(secs);
sampling_end(pmu_fd); error = 0;
free(pmu_fd);
cleanup:
sampling_end(links);
/* output sample counts */ /* output sample counts */
print_ip_map(map_fd[0]); if (!error)
print_ip_map(map_fd);
return 0; free(links);
bpf_object__close(obj);
return error;
} }

141
samples/bpf/trace_event_user.c
View File

@ -6,22 +6,25 @@
#include <stdlib.h> #include <stdlib.h>
#include <stdbool.h> #include <stdbool.h>
#include <string.h> #include <string.h>
#include <fcntl.h>
#include <poll.h>
#include <sys/ioctl.h>
#include <linux/perf_event.h> #include <linux/perf_event.h>
#include <linux/bpf.h> #include <linux/bpf.h>
#include <signal.h> #include <signal.h>
#include <assert.h>
#include <errno.h> #include <errno.h>
#include <sys/resource.h> #include <sys/resource.h>
#include <bpf/bpf.h>
#include <bpf/libbpf.h> #include <bpf/libbpf.h>
#include "bpf_load.h"
#include "perf-sys.h" #include "perf-sys.h"
#include "trace_helpers.h" #include "trace_helpers.h"
#define __must_check
#include <linux/err.h>
#define SAMPLE_FREQ 50 #define SAMPLE_FREQ 50
static int pid;
/* counts, stackmap */
static int map_fd[2];
struct bpf_program *prog;
static bool sys_read_seen, sys_write_seen; static bool sys_read_seen, sys_write_seen;
static void print_ksym(__u64 addr) static void print_ksym(__u64 addr)
@ -91,10 +94,10 @@ static void print_stack(struct key_t *key, __u64 count)
} }
} }
static void int_exit(int sig) static void err_exit(int err)
{ {
kill(0, SIGKILL); kill(pid, SIGKILL);
exit(0); exit(err);
} }
static void print_stacks(void) static void print_stacks(void)
@ -102,7 +105,7 @@ static void print_stacks(void)
struct key_t key = {}, next_key; struct key_t key = {}, next_key;
__u64 value; __u64 value;
__u32 stackid = 0, next_id; __u32 stackid = 0, next_id;
int fd = map_fd[0], stack_map = map_fd[1]; int error = 1, fd = map_fd[0], stack_map = map_fd[1];
sys_read_seen = sys_write_seen = false; sys_read_seen = sys_write_seen = false;
while (bpf_map_get_next_key(fd, &key, &next_key) == 0) { while (bpf_map_get_next_key(fd, &key, &next_key) == 0) {
@ -114,7 +117,7 @@ static void print_stacks(void)
printf("\n"); printf("\n");
if (!sys_read_seen || !sys_write_seen) { if (!sys_read_seen || !sys_write_seen) {
printf("BUG kernel stack doesn't contain sys_read() and sys_write()\n"); printf("BUG kernel stack doesn't contain sys_read() and sys_write()\n");
int_exit(0); err_exit(error);
} }
/* clear stack map */ /* clear stack map */
@ -136,43 +139,52 @@ static inline int generate_load(void)
static void test_perf_event_all_cpu(struct perf_event_attr *attr) static void test_perf_event_all_cpu(struct perf_event_attr *attr)
{ {
int nr_cpus = sysconf(_SC_NPROCESSORS_CONF); int nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
int *pmu_fd = malloc(nr_cpus * sizeof(int)); struct bpf_link **links = calloc(nr_cpus, sizeof(struct bpf_link *));
int i, error = 0; int i, pmu_fd, error = 1;
if (!links) {
printf("malloc of links failed\n");
goto err;
}
/* system wide perf event, no need to inherit */ /* system wide perf event, no need to inherit */
attr->inherit = 0; attr->inherit = 0;
/* open perf_event on all cpus */ /* open perf_event on all cpus */
for (i = 0; i < nr_cpus; i++) { for (i = 0; i < nr_cpus; i++) {
pmu_fd[i] = sys_perf_event_open(attr, -1, i, -1, 0); pmu_fd = sys_perf_event_open(attr, -1, i, -1, 0);
if (pmu_fd[i] < 0) { if (pmu_fd < 0) {
printf("sys_perf_event_open failed\n"); printf("sys_perf_event_open failed\n");
error = 1;
goto all_cpu_err; goto all_cpu_err;
} }
assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_SET_BPF, prog_fd[0]) == 0); links[i] = bpf_program__attach_perf_event(prog, pmu_fd);
assert(ioctl(pmu_fd[i], PERF_EVENT_IOC_ENABLE) == 0); if (IS_ERR(links[i])) {
printf("bpf_program__attach_perf_event failed\n");
links[i] = NULL;
close(pmu_fd);
goto all_cpu_err;
}
} }
if (generate_load() < 0) { if (generate_load() < 0)
error = 1;
goto all_cpu_err; goto all_cpu_err;
}
print_stacks(); print_stacks();
error = 0;
all_cpu_err: all_cpu_err:
for (i--; i >= 0; i--) { for (i--; i >= 0; i--)
ioctl(pmu_fd[i], PERF_EVENT_IOC_DISABLE); bpf_link__destroy(links[i]);
close(pmu_fd[i]); err:
} free(links);
free(pmu_fd);
if (error) if (error)
int_exit(0); err_exit(error);
} }
static void test_perf_event_task(struct perf_event_attr *attr) static void test_perf_event_task(struct perf_event_attr *attr)
{ {
int pmu_fd, error = 0; struct bpf_link *link = NULL;
int pmu_fd, error = 1;
/* per task perf event, enable inherit so the "dd ..." command can be traced properly. /* per task perf event, enable inherit so the "dd ..." command can be traced properly.
* Enabling inherit will cause bpf_perf_prog_read_time helper failure. * Enabling inherit will cause bpf_perf_prog_read_time helper failure.
@ -183,21 +195,25 @@ static void test_perf_event_task(struct perf_event_attr *attr)
pmu_fd = sys_perf_event_open(attr, 0, -1, -1, 0); pmu_fd = sys_perf_event_open(attr, 0, -1, -1, 0);
if (pmu_fd < 0) { if (pmu_fd < 0) {
printf("sys_perf_event_open failed\n"); printf("sys_perf_event_open failed\n");
int_exit(0);
}
assert(ioctl(pmu_fd, PERF_EVENT_IOC_SET_BPF, prog_fd[0]) == 0);
assert(ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE) == 0);
if (generate_load() < 0) {
error = 1;
goto err; goto err;
} }
link = bpf_program__attach_perf_event(prog, pmu_fd);
if (IS_ERR(link)) {
printf("bpf_program__attach_perf_event failed\n");
link = NULL;
close(pmu_fd);
goto err;
}
if (generate_load() < 0)
goto err;
print_stacks(); print_stacks();
error = 0;
err: err:
ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE); bpf_link__destroy(link);
close(pmu_fd);
if (error) if (error)
int_exit(0); err_exit(error);
} }
static void test_bpf_perf_event(void) static void test_bpf_perf_event(void)
@ -282,29 +298,60 @@ static void test_bpf_perf_event(void)
int main(int argc, char **argv) int main(int argc, char **argv)
{ {
struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY}; struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY};
struct bpf_object *obj = NULL;
char filename[256]; char filename[256];
int error = 1;
snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]); snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
setrlimit(RLIMIT_MEMLOCK, &r); setrlimit(RLIMIT_MEMLOCK, &r);
signal(SIGINT, int_exit); signal(SIGINT, err_exit);
signal(SIGTERM, int_exit); signal(SIGTERM, err_exit);
if (load_kallsyms()) { if (load_kallsyms()) {
printf("failed to process /proc/kallsyms\n"); printf("failed to process /proc/kallsyms\n");
return 1; goto cleanup;
} }
if (load_bpf_file(filename)) { obj = bpf_object__open_file(filename, NULL);
printf("%s", bpf_log_buf); if (IS_ERR(obj)) {
return 2; printf("opening BPF object file failed\n");
obj = NULL;
goto cleanup;
} }
if (fork() == 0) { prog = bpf_object__find_program_by_name(obj, "bpf_prog1");
if (!prog) {
printf("finding a prog in obj file failed\n");
goto cleanup;
}
/* load BPF program */
if (bpf_object__load(obj)) {
printf("loading BPF object file failed\n");
goto cleanup;
}
map_fd[0] = bpf_object__find_map_fd_by_name(obj, "counts");
map_fd[1] = bpf_object__find_map_fd_by_name(obj, "stackmap");
if (map_fd[0] < 0 || map_fd[1] < 0) {
printf("finding a counts/stackmap map in obj file failed\n");
goto cleanup;
}
pid = fork();
if (pid == 0) {
read_trace_pipe(); read_trace_pipe();
return 0; return 0;
} else if (pid == -1) {
printf("couldn't spawn process\n");
goto cleanup;
} }
test_bpf_perf_event(); test_bpf_perf_event();
int_exit(0); error = 0;
return 0;
cleanup:
bpf_object__close(obj);
err_exit(error);
} }