[PATCH v1 2/4] perf trace: Migrate BPF augmentation to use a skeleton

From: Ian Rogers
Date: Thu Aug 10 2023 - 14:49:15 EST

Previously a BPF event of augmented_raw_syscalls.c could be used to
enable augmentation of syscalls by perf trace. As BPF events are no
longer supported, switch to using a BPF skeleton which when attached
explicitly opens the sysenter and sysexit tracepoints.

The dump map is removed as debugging wasn't supported by the
augmentation and bpf_printk can be used when necessary.

Remove tools/perf/examples/bpf/augmented_raw_syscalls.c so that the
rename/migration to a BPF skeleton captures that this was the source.

Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
---
tools/perf/Makefile.perf | 1 +
tools/perf/builtin-trace.c | 180 +++++++++++-------
.../bpf_skel/augmented_raw_syscalls.bpf.c} | 27 +--
3 files changed, 131 insertions(+), 77 deletions(-)
rename tools/perf/{examples/bpf/augmented_raw_syscalls.c => util/bpf_skel/augmented_raw_syscalls.bpf.c} (96%)

diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
index 6ec5079fd697..0e1597712b95 100644
--- a/tools/perf/Makefile.perf
+++ b/tools/perf/Makefile.perf
@@ -1042,6 +1042,7 @@ SKELETONS += $(SKEL_OUT)/bperf_cgroup.skel.h $(SKEL_OUT)/func_latency.skel.h
SKELETONS += $(SKEL_OUT)/off_cpu.skel.h $(SKEL_OUT)/lock_contention.skel.h
SKELETONS += $(SKEL_OUT)/kwork_trace.skel.h $(SKEL_OUT)/sample_filter.skel.h
SKELETONS += $(SKEL_OUT)/bench_uprobe.skel.h
+SKELETONS += $(SKEL_OUT)/augmented_raw_syscalls.skel.h

$(SKEL_TMP_OUT) $(LIBAPI_OUTPUT) $(LIBBPF_OUTPUT) $(LIBPERF_OUTPUT) $(LIBSUBCMD_OUTPUT) $(LIBSYMBOL_OUTPUT):
$(Q)$(MKDIR) -p $@
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
index 59862467e781..8625fca42cd8 100644
--- a/tools/perf/builtin-trace.c
+++ b/tools/perf/builtin-trace.c
@@ -19,6 +19,9 @@
#ifdef HAVE_LIBBPF_SUPPORT
#include <bpf/bpf.h>
#include <bpf/libbpf.h>
+#ifdef HAVE_BPF_SKEL
+#include "bpf_skel/augmented_raw_syscalls.skel.h"
+#endif
#endif
#include "util/bpf_map.h"
#include "util/rlimit.h"
@@ -127,25 +130,19 @@ struct trace {
struct syscalltbl *sctbl;
struct {
struct syscall *table;
- struct { // per syscall BPF_MAP_TYPE_PROG_ARRAY
- struct bpf_map *sys_enter,
- *sys_exit;
- } prog_array;
struct {
struct evsel *sys_enter,
- *sys_exit,
- *augmented;
+ *sys_exit,
+ *bpf_output;
} events;
- struct bpf_program *unaugmented_prog;
} syscalls;
- struct {
- struct bpf_map *map;
- } dump;
+#ifdef HAVE_BPF_SKEL
+ struct augmented_raw_syscalls_bpf *skel;
+#endif
struct record_opts opts;
struct evlist *evlist;
struct machine *host;
struct thread *current;
- struct bpf_object *bpf_obj;
struct cgroup *cgroup;
u64 base_time;
FILE *output;
@@ -415,6 +412,7 @@ static int evsel__init_syscall_tp(struct evsel *evsel)
if (evsel__init_tp_uint_field(evsel, &sc->id, "__syscall_nr") &&
evsel__init_tp_uint_field(evsel, &sc->id, "nr"))
return -ENOENT;
+
return 0;
}

@@ -2845,7 +2843,7 @@ static int trace__event_handler(struct trace *trace, struct evsel *evsel,
if (thread)
trace__fprintf_comm_tid(trace, thread, trace->output);

- if (evsel == trace->syscalls.events.augmented) {
+ if (evsel == trace->syscalls.events.bpf_output) {
int id = perf_evsel__sc_tp_uint(evsel, id, sample);
struct syscall *sc = trace__syscall_info(trace, evsel, id);

@@ -3278,24 +3276,16 @@ static int trace__set_ev_qualifier_tp_filter(struct trace *trace)
goto out;
}

-#ifdef HAVE_LIBBPF_SUPPORT
-static struct bpf_map *trace__find_bpf_map_by_name(struct trace *trace, const char *name)
-{
- if (trace->bpf_obj == NULL)
- return NULL;
-
- return bpf_object__find_map_by_name(trace->bpf_obj, name);
-}
-
+#ifdef HAVE_BPF_SKEL
static struct bpf_program *trace__find_bpf_program_by_title(struct trace *trace, const char *name)
{
struct bpf_program *pos, *prog = NULL;
const char *sec_name;

- if (trace->bpf_obj == NULL)
+ if (trace->skel->obj == NULL)
return NULL;

- bpf_object__for_each_program(pos, trace->bpf_obj) {
+ bpf_object__for_each_program(pos, trace->skel->obj) {
sec_name = bpf_program__section_name(pos);
if (sec_name && !strcmp(sec_name, name)) {
prog = pos;
@@ -3313,12 +3303,14 @@ static struct bpf_program *trace__find_syscall_bpf_prog(struct trace *trace, str

if (prog_name == NULL) {
char default_prog_name[256];
- scnprintf(default_prog_name, sizeof(default_prog_name), "!syscalls:sys_%s_%s", type, sc->name);
+ scnprintf(default_prog_name, sizeof(default_prog_name), "tp/syscalls/sys_%s_%s",
+ type, sc->name);
prog = trace__find_bpf_program_by_title(trace, default_prog_name);
if (prog != NULL)
goto out_found;
if (sc->fmt && sc->fmt->alias) {
- scnprintf(default_prog_name, sizeof(default_prog_name), "!syscalls:sys_%s_%s", type, sc->fmt->alias);
+ scnprintf(default_prog_name, sizeof(default_prog_name),
+ "tp/syscalls/sys_%s_%s", type, sc->fmt->alias);
prog = trace__find_bpf_program_by_title(trace, default_prog_name);
if (prog != NULL)
goto out_found;
@@ -3336,7 +3328,7 @@ static struct bpf_program *trace__find_syscall_bpf_prog(struct trace *trace, str
pr_debug("Couldn't find BPF prog \"%s\" to associate with syscalls:sys_%s_%s, not augmenting it\n",
prog_name, type, sc->name);
out_unaugmented:
- return trace->syscalls.unaugmented_prog;
+ return trace->skel->progs.syscall_unaugmented;
}

static void trace__init_syscall_bpf_progs(struct trace *trace, int id)
@@ -3353,13 +3345,21 @@ static void trace__init_syscall_bpf_progs(struct trace *trace, int id)
static int trace__bpf_prog_sys_enter_fd(struct trace *trace, int id)
{
struct syscall *sc = trace__syscall_info(trace, NULL, id);
- return sc ? bpf_program__fd(sc->bpf_prog.sys_enter) : bpf_program__fd(trace->syscalls.unaugmented_prog);
+
+ if (sc)
+ return bpf_program__fd(sc->bpf_prog.sys_enter);
+
+ return bpf_program__fd(trace->skel->progs.syscall_unaugmented);
}

static int trace__bpf_prog_sys_exit_fd(struct trace *trace, int id)
{
struct syscall *sc = trace__syscall_info(trace, NULL, id);
- return sc ? bpf_program__fd(sc->bpf_prog.sys_exit) : bpf_program__fd(trace->syscalls.unaugmented_prog);
+
+ if (sc)
+ return bpf_program__fd(sc->bpf_prog.sys_exit);
+
+ return bpf_program__fd(trace->skel->progs.syscall_unaugmented);
}

static struct bpf_program *trace__find_usable_bpf_prog_entry(struct trace *trace, struct syscall *sc)
@@ -3384,7 +3384,7 @@ static struct bpf_program *trace__find_usable_bpf_prog_entry(struct trace *trace
bool is_candidate = false;

if (pair == NULL || pair == sc ||
- pair->bpf_prog.sys_enter == trace->syscalls.unaugmented_prog)
+ pair->bpf_prog.sys_enter == trace->skel->progs.syscall_unaugmented)
continue;

for (field = sc->args, candidate_field = pair->args;
@@ -3437,7 +3437,7 @@ static struct bpf_program *trace__find_usable_bpf_prog_entry(struct trace *trace
*/
if (pair_prog == NULL) {
pair_prog = trace__find_syscall_bpf_prog(trace, pair, pair->fmt ? pair->fmt->bpf_prog_name.sys_enter : NULL, "enter");
- if (pair_prog == trace->syscalls.unaugmented_prog)
+ if (pair_prog == trace->skel->progs.syscall_unaugmented)
goto next_candidate;
}

@@ -3452,8 +3452,8 @@ static struct bpf_program *trace__find_usable_bpf_prog_entry(struct trace *trace

static int trace__init_syscalls_bpf_prog_array_maps(struct trace *trace)
{
- int map_enter_fd = bpf_map__fd(trace->syscalls.prog_array.sys_enter),
- map_exit_fd = bpf_map__fd(trace->syscalls.prog_array.sys_exit);
+ int map_enter_fd = bpf_map__fd(trace->skel->maps.syscalls_sys_enter);
+ int map_exit_fd = bpf_map__fd(trace->skel->maps.syscalls_sys_exit);
int err = 0, key;

for (key = 0; key < trace->sctbl->syscalls.nr_entries; ++key) {
@@ -3515,7 +3515,7 @@ static int trace__init_syscalls_bpf_prog_array_maps(struct trace *trace)
* For now we're just reusing the sys_enter prog, and if it
* already has an augmenter, we don't need to find one.
*/
- if (sc->bpf_prog.sys_enter != trace->syscalls.unaugmented_prog)
+ if (sc->bpf_prog.sys_enter != trace->skel->progs.syscall_unaugmented)
continue;

/*
@@ -3538,22 +3538,9 @@ static int trace__init_syscalls_bpf_prog_array_maps(struct trace *trace)
break;
}

-
return err;
}
-
-#else // HAVE_LIBBPF_SUPPORT
-static struct bpf_map *trace__find_bpf_map_by_name(struct trace *trace __maybe_unused,
- const char *name __maybe_unused)
-{
- return NULL;
-}
-
-static int trace__init_syscalls_bpf_prog_array_maps(struct trace *trace __maybe_unused)
-{
- return 0;
-}
-#endif // HAVE_LIBBPF_SUPPORT
+#endif // HAVE_BPF_SKEL

static int trace__set_ev_qualifier_filter(struct trace *trace)
{
@@ -3917,13 +3904,31 @@ static int trace__run(struct trace *trace, int argc, const char **argv)
err = evlist__open(evlist);
if (err < 0)
goto out_error_open;
+#ifdef HAVE_BPF_SKEL
+ {
+ struct perf_cpu cpu;

+ /*
+ * Set up the __augmented_syscalls__ BPF map to hold for each
+ * CPU the bpf-output event's file descriptor.
+ */
+ perf_cpu_map__for_each_cpu(cpu, i, trace->syscalls.events.bpf_output->core.cpus) {
+ bpf_map__update_elem(trace->skel->maps.__augmented_syscalls__,
+ &cpu.cpu, sizeof(int),
+ xyarray__entry(trace->syscalls.events.bpf_output->core.fd,
+ cpu.cpu, 0),
+ sizeof(__u32), BPF_ANY);
+ }
+ }
+#endif
err = trace__set_filter_pids(trace);
if (err < 0)
goto out_error_mem;

- if (trace->syscalls.prog_array.sys_enter)
+#ifdef HAVE_BPF_SKEL
+ if (trace->skel->progs.sys_enter)
trace__init_syscalls_bpf_prog_array_maps(trace);
+#endif

if (trace->ev_qualifier_ids.nr > 0) {
err = trace__set_ev_qualifier_filter(trace);
@@ -3956,9 +3961,6 @@ static int trace__run(struct trace *trace, int argc, const char **argv)
if (err < 0)
goto out_error_apply_filters;

- if (trace->dump.map)
- bpf_map__fprintf(trace->dump.map, trace->output);
-
err = evlist__mmap(evlist, trace->opts.mmap_pages);
if (err < 0)
goto out_error_mmap;
@@ -4655,6 +4657,18 @@ static void trace__exit(struct trace *trace)
zfree(&trace->perfconfig_events);
}

+#ifdef HAVE_BPF_SKEL
+static int bpf__setup_bpf_output(struct evlist *evlist)
+{
+ int err = parse_event(evlist, "bpf-output/no-inherit=1,name=__augmented_syscalls__/");
+
+ if (err)
+ pr_debug("ERROR: failed to create the \"__augmented_syscalls__\" bpf-output event\n");
+
+ return err;
+}
+#endif
+
int cmd_trace(int argc, const char **argv)
{
const char *trace_usage[] = {
@@ -4686,7 +4700,6 @@ int cmd_trace(int argc, const char **argv)
.max_stack = UINT_MAX,
.max_events = ULONG_MAX,
};
- const char *map_dump_str = NULL;
const char *output_name = NULL;
const struct option trace_options[] = {
OPT_CALLBACK('e', "event", &trace, "event",
@@ -4720,9 +4733,6 @@ int cmd_trace(int argc, const char **argv)
OPT_CALLBACK(0, "duration", &trace, "float",
"show only events with duration > N.M ms",
trace__set_duration),
-#ifdef HAVE_LIBBPF_SUPPORT
- OPT_STRING(0, "map-dump", &map_dump_str, "BPF map", "BPF map to periodically dump"),
-#endif
OPT_BOOLEAN(0, "sched", &trace.sched, "show blocking scheduler events"),
OPT_INCR('v', "verbose", &verbose, "be more verbose"),
OPT_BOOLEAN('T', "time", &trace.full_time,
@@ -4849,16 +4859,55 @@ int cmd_trace(int argc, const char **argv)
"cgroup monitoring only available in system-wide mode");
}

- err = -1;
+#ifdef HAVE_BPF_SKEL
+ trace.skel = augmented_raw_syscalls_bpf__open();
+ if (!trace.skel) {
+ pr_debug("Failed to open augmented syscalls BPF skeleton");
+ } else {
+ /*
+ * Disable attaching the BPF programs except for sys_enter and
+ * sys_exit that tail call into this as necessary.
+ */
+ bpf_program__set_autoattach(trace.skel->progs.syscall_unaugmented,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_connect,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_sendto,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_open,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_openat,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_rename,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_renameat,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_perf_event_open,
+ /*autoattach=*/false);
+ bpf_program__set_autoattach(trace.skel->progs.sys_enter_clock_nanosleep,
+ /*autoattach=*/false);
+
+ err = augmented_raw_syscalls_bpf__load(trace.skel);

- if (map_dump_str) {
- trace.dump.map = trace__find_bpf_map_by_name(&trace, map_dump_str);
- if (trace.dump.map == NULL) {
- pr_err("ERROR: BPF map \"%s\" not found\n", map_dump_str);
- goto out;
+ if (err < 0) {
+ pr_debug("Failed to load augmented syscalls BPF skeleton\n");
+ } else {
+ augmented_raw_syscalls_bpf__attach(trace.skel);
+ trace__add_syscall_newtp(&trace);
}
}

+ err = bpf__setup_bpf_output(trace.evlist);
+ if (err) {
+ libbpf_strerror(err, bf, sizeof(bf));
+ pr_err("ERROR: Setup BPF output event failed: %s\n", bf);
+ goto out;
+ }
+ trace.syscalls.events.bpf_output = evlist__last(trace.evlist);
+ assert(!strcmp(evsel__name(trace.syscalls.events.bpf_output), "__augmented_syscalls__"));
+#endif
+ err = -1;
+
if (trace.trace_pgfaults) {
trace.opts.sample_address = true;
trace.opts.sample_time = true;
@@ -4909,7 +4958,7 @@ int cmd_trace(int argc, const char **argv)
* buffers that are being copied from kernel to userspace, think 'read'
* syscall.
*/
- if (trace.syscalls.events.augmented) {
+ if (trace.syscalls.events.bpf_output) {
evlist__for_each_entry(trace.evlist, evsel) {
bool raw_syscalls_sys_exit = strcmp(evsel__name(evsel), "raw_syscalls:sys_exit") == 0;

@@ -4918,9 +4967,9 @@ int cmd_trace(int argc, const char **argv)
goto init_augmented_syscall_tp;
}

- if (trace.syscalls.events.augmented->priv == NULL &&
+ if (trace.syscalls.events.bpf_output->priv == NULL &&
strstr(evsel__name(evsel), "syscalls:sys_enter")) {
- struct evsel *augmented = trace.syscalls.events.augmented;
+ struct evsel *augmented = trace.syscalls.events.bpf_output;
if (evsel__init_augmented_syscall_tp(augmented, evsel) ||
evsel__init_augmented_syscall_tp_args(augmented))
goto out;
@@ -5025,5 +5074,8 @@ int cmd_trace(int argc, const char **argv)
fclose(trace.output);
out:
trace__exit(&trace);
+#ifdef HAVE_BPF_SKEL
+ augmented_raw_syscalls_bpf__destroy(trace.skel);
+#endif
return err;
}
diff --git a/tools/perf/examples/bpf/augmented_raw_syscalls.c b/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
similarity index 96%
rename from tools/perf/examples/bpf/augmented_raw_syscalls.c
rename to tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
index 9a03189d33d3..70478b9460ee 100644
--- a/tools/perf/examples/bpf/augmented_raw_syscalls.c
+++ b/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
@@ -18,6 +18,8 @@
#include <bpf/bpf_helpers.h>
#include <linux/limits.h>

+#define MAX_CPUS 4096
+
// FIXME: These should come from system headers
typedef char bool;
typedef int pid_t;
@@ -34,7 +36,7 @@ struct __augmented_syscalls__ {
__uint(type, BPF_MAP_TYPE_PERF_EVENT_ARRAY);
__type(key, int);
__type(value, __u32);
- __uint(max_entries, __NR_CPUS__);
+ __uint(max_entries, MAX_CPUS);
} __augmented_syscalls__ SEC(".maps");

/*
@@ -170,7 +172,7 @@ unsigned int augmented_arg__read_str(struct augmented_arg *augmented_arg, const
return augmented_len;
}

-SEC("!raw_syscalls:unaugmented")
+SEC("tp/raw_syscalls/sys_enter")
int syscall_unaugmented(struct syscall_enter_args *args)
{
return 1;
@@ -182,7 +184,7 @@ int syscall_unaugmented(struct syscall_enter_args *args)
* on from there, reading the first syscall arg as a string, i.e. open's
* filename.
*/
-SEC("!syscalls:sys_enter_connect")
+SEC("tp/syscalls/sys_enter_connect")
int sys_enter_connect(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -201,7 +203,7 @@ int sys_enter_connect(struct syscall_enter_args *args)
return augmented__output(args, augmented_args, len + socklen);
}

-SEC("!syscalls:sys_enter_sendto")
+SEC("tp/syscalls/sys_enter_sendto")
int sys_enter_sendto(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -220,7 +222,7 @@ int sys_enter_sendto(struct syscall_enter_args *args)
return augmented__output(args, augmented_args, len + socklen);
}

-SEC("!syscalls:sys_enter_open")
+SEC("tp/syscalls/sys_enter_open")
int sys_enter_open(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -235,7 +237,7 @@ int sys_enter_open(struct syscall_enter_args *args)
return augmented__output(args, augmented_args, len);
}

-SEC("!syscalls:sys_enter_openat")
+SEC("tp/syscalls/sys_enter_openat")
int sys_enter_openat(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -250,7 +252,7 @@ int sys_enter_openat(struct syscall_enter_args *args)
return augmented__output(args, augmented_args, len);
}

-SEC("!syscalls:sys_enter_rename")
+SEC("tp/syscalls/sys_enter_rename")
int sys_enter_rename(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -267,7 +269,7 @@ int sys_enter_rename(struct syscall_enter_args *args)
return augmented__output(args, augmented_args, len);
}

-SEC("!syscalls:sys_enter_renameat")
+SEC("tp/syscalls/sys_enter_renameat")
int sys_enter_renameat(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -295,7 +297,7 @@ struct perf_event_attr_size {
__u32 size;
};

-SEC("!syscalls:sys_enter_perf_event_open")
+SEC("tp/syscalls/sys_enter_perf_event_open")
int sys_enter_perf_event_open(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -327,7 +329,7 @@ int sys_enter_perf_event_open(struct syscall_enter_args *args)
return 1; /* Failure: don't filter */
}

-SEC("!syscalls:sys_enter_clock_nanosleep")
+SEC("tp/syscalls/sys_enter_clock_nanosleep")
int sys_enter_clock_nanosleep(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args = augmented_args_payload();
@@ -358,7 +360,7 @@ static bool pid_filter__has(struct pids_filtered *pids, pid_t pid)
return bpf_map_lookup_elem(pids, &pid) != NULL;
}

-SEC("raw_syscalls:sys_enter")
+SEC("tp/raw_syscalls/sys_enter")
int sys_enter(struct syscall_enter_args *args)
{
struct augmented_args_payload *augmented_args;
@@ -371,7 +373,6 @@ int sys_enter(struct syscall_enter_args *args)
* We'll add to this as we add augmented syscalls right after that
* initial, non-augmented raw_syscalls:sys_enter payload.
*/
- unsigned int len = sizeof(augmented_args->args);

if (pid_filter__has(&pids_filtered, getpid()))
return 0;
@@ -393,7 +394,7 @@ int sys_enter(struct syscall_enter_args *args)
return 0;
}

-SEC("raw_syscalls:sys_exit")
+SEC("tp/raw_syscalls/sys_exit")
int sys_exit(struct syscall_exit_args *args)
{
struct syscall_exit_args exit_args;
--
2.41.0.640.ga95def55d0-goog