bpf: permit multiple bpf attachments for a single perf tracepoint event

index 31ee304..9f78a682 100644
@@ -8191,10 +8191,10 @@ static void perf_event_free_bpf_prog(struct perf_event *event)
 {
 	struct bpf_prog *prog;
 
-	perf_event_free_bpf_handler(event);
-
-	if (!event->tp_event)
+	if (event->attr.type != PERF_TYPE_TRACEPOINT) {
+		perf_event_free_bpf_handler(event);
 		return;
+	}
 
 	prog = event->tp_event->prog;
 	if (prog && event->tp_event->bpf_prog_owner == event) {

index 1e334b2..172be7f 100644
@@ -273,18 +273,38 @@ int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
 				__u32 __user *prog_ids, u32 cnt);
 
-#define BPF_PROG_RUN_ARRAY(array, ctx, func)		\
+void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
+				struct bpf_prog *old_prog);
+int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
+			struct bpf_prog *exclude_prog,
+			struct bpf_prog *include_prog,
+			struct bpf_prog_array **new_array);
+
+#define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null)	\
 	({						\
-		struct bpf_prog **_prog;		\
+		struct bpf_prog **_prog, *__prog;	\
+		struct bpf_prog_array *_array;		\
 		u32 _ret = 1;				\
 		rcu_read_lock();			\
-		_prog = rcu_dereference(array)->progs;	\
-		for (; *_prog; _prog++)			\
-			_ret &= func(*_prog, ctx);	\
+		_array = rcu_dereference(array);	\
+		if (unlikely(check_non_null && !_array))\
+			goto _out;			\
+		_prog = _array->progs;			\
+		while ((__prog = READ_ONCE(*_prog))) {	\
+			_ret &= func(__prog, ctx);	\
+			_prog++;			\
+		}					\
+_out:							\
 		rcu_read_unlock();			\
 		_ret;					\
 	 })
 
+#define BPF_PROG_RUN_ARRAY(array, ctx, func)		\
+	__BPF_PROG_RUN_ARRAY(array, ctx, func, false)
+
+#define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func)	\
+	__BPF_PROG_RUN_ARRAY(array, ctx, func, true)
+
 #ifdef CONFIG_BPF_SYSCALL
 DECLARE_PER_CPU(int, bpf_prog_active);
 

index 2e0f222..c811464 100644
@@ -271,14 +271,37 @@ struct trace_event_call {
 #ifdef CONFIG_PERF_EVENTS
 	int				perf_refcount;
 	struct hlist_head __percpu	*perf_events;
-	struct bpf_prog			*prog;
-	struct perf_event		*bpf_prog_owner;
+	struct bpf_prog_array __rcu	*prog_array;
 
 	int	(*perf_perm)(struct trace_event_call *,
 			     struct perf_event *);
 #endif
 };
 
+#ifdef CONFIG_PERF_EVENTS
+static inline bool bpf_prog_array_valid(struct trace_event_call *call)
+{
+	/*
+	 * This inline function checks whether call->prog_array
+	 * is valid or not. The function is called in various places,
+	 * outside rcu_read_lock/unlock, as a heuristic to speed up execution.
+	 *
+	 * If this function returns true, and later call->prog_array
+	 * becomes false inside rcu_read_lock/unlock region,
+	 * we bail out then. If this function return false,
+	 * there is a risk that we might miss a few events if the checking
+	 * were delayed until inside rcu_read_lock/unlock region and
+	 * call->prog_array happened to become non-NULL then.
+	 *
+	 * Here, READ_ONCE() is used instead of rcu_access_pointer().
+	 * rcu_access_pointer() requires the actual definition of
+	 * "struct bpf_prog_array" while READ_ONCE() only needs
+	 * a declaration of the same type.
+	 */
+	return !!READ_ONCE(call->prog_array);
+}
+#endif
+
 static inline const char *
 trace_event_name(struct trace_event_call *call)
 {

@@ -435,9 +458,9 @@ trace_trigger_soft_disabled(struct trace_event_file *file)
 }
 
 #ifdef CONFIG_BPF_EVENTS
-unsigned int trace_call_bpf(struct bpf_prog *prog, void *ctx);
+unsigned int trace_call_bpf(struct trace_event_call *call, void *ctx);
 #else
-static inline unsigned int trace_call_bpf(struct bpf_prog *prog, void *ctx)
+static inline unsigned int trace_call_bpf(struct trace_event_call *call, void *ctx)
 {
 	return 1;
 }

@@ -499,6 +522,8 @@ extern void ftrace_profile_free_filter(struct perf_event *event);
 void perf_trace_buf_update(void *record, u16 type);
 void *perf_trace_buf_alloc(int size, struct pt_regs **regs, int *rctxp);
 
+int perf_event_attach_bpf_prog(struct perf_event *event, struct bpf_prog *prog);
+void perf_event_detach_bpf_prog(struct perf_event *event);
 void perf_trace_run_bpf_submit(void *raw_data, int size, int rctx,
 			       struct trace_event_call *call, u64 count,
 			       struct pt_regs *regs, struct hlist_head *head,

index 04fe68bb..14f127b6 100644
@@ -34,7 +34,6 @@ perf_trace_##call(void *__data, proto)					\
 	struct trace_event_call *event_call = __data;			\
 	struct trace_event_data_offsets_##call __maybe_unused __data_offsets;\
 	struct trace_event_raw_##call *entry;				\
-	struct bpf_prog *prog = event_call->prog;			\
 	struct pt_regs *__regs;						\
 	u64 __count = 1;						\
 	struct task_struct *__task = NULL;				\

@@ -46,8 +45,9 @@ perf_trace_##call(void *__data, proto)					\
 	__data_size = trace_event_get_offsets_##call(&__data_offsets, args); \
 									\
 	head = this_cpu_ptr(event_call->perf_events);			\
-	if (!prog && __builtin_constant_p(!__task) && !__task &&	\
-				hlist_empty(head))			\
+	if (!bpf_prog_array_valid(event_call) &&			\
+	    __builtin_constant_p(!__task) && !__task &&			\
+	    hlist_empty(head))						\
 		return;							\
 									\
 	__entry_size = ALIGN(__data_size + sizeof(*entry) + sizeof(u32),\

93 lines not on this page · raw message

index 9f78a682..9660ee6 100644
@@ -7954,11 +7954,9 @@ void perf_trace_run_bpf_submit(void *raw_data, int size, int rctx,
 			       struct pt_regs *regs, struct hlist_head *head,
 			       struct task_struct *task)
 {
-	struct bpf_prog *prog = call->prog;
-
-	if (prog) {
+	if (bpf_prog_array_valid(call)) {
 		*(struct pt_regs **)raw_data = regs;
-		if (!trace_call_bpf(prog, raw_data) || hlist_empty(head)) {
+		if (!trace_call_bpf(call, raw_data) || hlist_empty(head)) {
 			perf_swevent_put_recursion_context(rctx);
 			return;
 		}

@@ -8147,13 +8145,11 @@ static int perf_event_set_bpf_prog(struct perf_event *event, u32 prog_fd)
 {
 	bool is_kprobe, is_tracepoint, is_syscall_tp;
 	struct bpf_prog *prog;
+	int ret;
 
 	if (event->attr.type != PERF_TYPE_TRACEPOINT)
 		return perf_event_set_bpf_handler(event, prog_fd);
 
-	if (event->tp_event->prog)
-		return -EEXIST;
-
 	is_kprobe = event->tp_event->flags & TRACE_EVENT_FL_UKPROBE;
 	is_tracepoint = event->tp_event->flags & TRACE_EVENT_FL_TRACEPOINT;
 	is_syscall_tp = is_syscall_trace_event(event->tp_event);

@@ -8181,26 +8177,20 @@ static int perf_event_set_bpf_prog(struct perf_event *event, u32 prog_fd)
 			return -EACCES;
 		}
 	}
-	event->tp_event->prog = prog;
-	event->tp_event->bpf_prog_owner = event;
 
-	return 0;
+	ret = perf_event_attach_bpf_prog(event, prog);
+	if (ret)
+		bpf_prog_put(prog);
+	return ret;
 }
 
 static void perf_event_free_bpf_prog(struct perf_event *event)
 {
-	struct bpf_prog *prog;
-
 	if (event->attr.type != PERF_TYPE_TRACEPOINT) {
 		perf_event_free_bpf_handler(event);
 		return;
 	}
-
-	prog = event->tp_event->prog;
-	if (prog && event->tp_event->bpf_prog_owner == event) {
-		event->tp_event->prog = NULL;
-		bpf_prog_put(prog);
-	}
+	perf_event_detach_bpf_prog(event);
 }
 
 #else

103 lines not on this page · raw message

index 8a907e1..abf92e4 100644
@@ -1174,13 +1174,12 @@ static void
 kprobe_perf_func(struct trace_kprobe *tk, struct pt_regs *regs)
 {
 	struct trace_event_call *call = &tk->tp.call;
-	struct bpf_prog *prog = call->prog;
 	struct kprobe_trace_entry_head *entry;
 	struct hlist_head *head;
 	int size, __size, dsize;
 	int rctx;
 
-	if (prog && !trace_call_bpf(prog, regs))
+	if (bpf_prog_array_valid(call) && !trace_call_bpf(call, regs))
 		return;
 
 	head = this_cpu_ptr(call->perf_events);

@@ -1210,13 +1209,12 @@ kretprobe_perf_func(struct trace_kprobe *tk, struct kretprobe_instance *ri,
 		    struct pt_regs *regs)
 {
 	struct trace_event_call *call = &tk->tp.call;
-	struct bpf_prog *prog = call->prog;
 	struct kretprobe_trace_entry_head *entry;
 	struct hlist_head *head;
 	int size, __size, dsize;
 	int rctx;
 
-	if (prog && !trace_call_bpf(prog, regs))
+	if (bpf_prog_array_valid(call) && !trace_call_bpf(call, regs))
 		return;
 
 	head = this_cpu_ptr(call->perf_events);

index 696afe7..71a6af3 100644
@@ -559,9 +559,10 @@ static DECLARE_BITMAP(enabled_perf_exit_syscalls, NR_syscalls);
 static int sys_perf_refcount_enter;
 static int sys_perf_refcount_exit;
 
-static int perf_call_bpf_enter(struct bpf_prog *prog, struct pt_regs *regs,
-			      struct syscall_metadata *sys_data,
-			      struct syscall_trace_enter *rec) {
+static int perf_call_bpf_enter(struct trace_event_call *call, struct pt_regs *regs,
+			       struct syscall_metadata *sys_data,
+			       struct syscall_trace_enter *rec)
+{
 	struct syscall_tp_t {
 		unsigned long long regs;
 		unsigned long syscall_nr;

@@ -573,7 +574,7 @@ static int perf_call_bpf_enter(struct bpf_prog *prog, struct pt_regs *regs,
 	param.syscall_nr = rec->nr;
 	for (i = 0; i < sys_data->nb_args; i++)
 		param.args[i] = rec->args[i];
-	return trace_call_bpf(prog, &param);
+	return trace_call_bpf(call, &param);
 }
 
 static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)

@@ -581,7 +582,7 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	struct syscall_metadata *sys_data;
 	struct syscall_trace_enter *rec;
 	struct hlist_head *head;
-	struct bpf_prog *prog;
+	bool valid_prog_array;
 	int syscall_nr;
 	int rctx;
 	int size;

@@ -596,9 +597,9 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	if (!sys_data)
 		return;
 
-	prog = READ_ONCE(sys_data->enter_event->prog);
 	head = this_cpu_ptr(sys_data->enter_event->perf_events);
-	if (!prog && hlist_empty(head))
+	valid_prog_array = bpf_prog_array_valid(sys_data->enter_event);
+	if (!valid_prog_array && hlist_empty(head))
 		return;
 
 	/* get the size after alignment with the u32 buffer size field */

@@ -614,7 +615,8 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	syscall_get_arguments(current, regs, 0, sys_data->nb_args,
 			       (unsigned long *)&rec->args);
 
-	if ((prog && !perf_call_bpf_enter(prog, regs, sys_data, rec)) ||
+	if ((valid_prog_array &&
+	     !perf_call_bpf_enter(sys_data->enter_event, regs, sys_data, rec)) ||
 	    hlist_empty(head)) {
 		perf_swevent_put_recursion_context(rctx);
 		return;

@@ -659,8 +661,9 @@ static void perf_sysenter_disable(struct trace_event_call *call)
 	mutex_unlock(&syscall_trace_lock);
 }
 
-static int perf_call_bpf_exit(struct bpf_prog *prog, struct pt_regs *regs,
-			      struct syscall_trace_exit *rec) {
+static int perf_call_bpf_exit(struct trace_event_call *call, struct pt_regs *regs,
+			      struct syscall_trace_exit *rec)
+{
 	struct syscall_tp_t {
 		unsigned long long regs;
 		unsigned long syscall_nr;

@@ -670,7 +673,7 @@ static int perf_call_bpf_exit(struct bpf_prog *prog, struct pt_regs *regs,
 	*(struct pt_regs **)&param = regs;
 	param.syscall_nr = rec->nr;
 	param.ret = rec->ret;
-	return trace_call_bpf(prog, &param);
+	return trace_call_bpf(call, &param);
 }
 
 static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)

@@ -678,7 +681,7 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 	struct syscall_metadata *sys_data;
 	struct syscall_trace_exit *rec;
 	struct hlist_head *head;
-	struct bpf_prog *prog;
+	bool valid_prog_array;
 	int syscall_nr;
 	int rctx;
 	int size;

@@ -693,9 +696,9 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 	if (!sys_data)
 		return;
 
-	prog = READ_ONCE(sys_data->exit_event->prog);
 	head = this_cpu_ptr(sys_data->exit_event->perf_events);
-	if (!prog && hlist_empty(head))
+	valid_prog_array = bpf_prog_array_valid(sys_data->exit_event);
+	if (!valid_prog_array && hlist_empty(head))
 		return;
 
 	/* We can probably do that at build time */

@@ -709,7 +712,8 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 	rec->nr = syscall_nr;
 	rec->ret = syscall_get_return_value(current, regs);
 
-	if ((prog && !perf_call_bpf_exit(prog, regs, rec)) ||
+	if ((valid_prog_array &&
+	     !perf_call_bpf_exit(sys_data->exit_event, regs, rec)) ||
 	    hlist_empty(head)) {
 		perf_swevent_put_recursion_context(rctx);
 		return;

index 4525e02..153c0e4 100644
@@ -1113,13 +1113,12 @@ static void __uprobe_perf_func(struct trace_uprobe *tu,
 {
 	struct trace_event_call *call = &tu->tp.call;
 	struct uprobe_trace_entry_head *entry;
-	struct bpf_prog *prog = call->prog;
 	struct hlist_head *head;
 	void *data;
 	int size, esize;
 	int rctx;
 
-	if (prog && !trace_call_bpf(prog, regs))
+	if (bpf_prog_array_valid(call) && !trace_call_bpf(call, regs))
 		return;
 
 	esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));

index a3cb91e..9169d32 100644
@@ -23,6 +23,13 @@
  * This requires kernel CONFIG_FTRACE_SYSCALLS to be set.
  */
 
+static void usage(const char *cmd)
+{
+	printf("USAGE: %s [-i num_progs] [-h]\n", cmd);
+	printf("       -i num_progs      # number of progs of the test\n");
+	printf("       -h                # help\n");
+}
+
 static void verify_map(int map_id)
 {
 	__u32 key = 0;

@@ -32,22 +39,29 @@ static void verify_map(int map_id)
 		fprintf(stderr, "map_lookup failed: %s\n", strerror(errno));
 		return;
 	}
-	if (val == 0)
+	if (val == 0) {
 		fprintf(stderr, "failed: map #%d returns value 0\n", map_id);
+		return;
+	}
+	val = 0;
+	if (bpf_map_update_elem(map_id, &key, &val, BPF_ANY) != 0) {
+		fprintf(stderr, "map_update failed: %s\n", strerror(errno));
+		return;
+	}
 }
 
-int main(int argc, char **argv)
+static int test(char *filename, int num_progs)
 {
-	struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY};
-	char filename[256];
-	int fd;
+	int i, fd, map0_fds[num_progs], map1_fds[num_progs];
 
-	setrlimit(RLIMIT_MEMLOCK, &r);
-	snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
-
-	if (load_bpf_file(filename)) {
-		fprintf(stderr, "%s", bpf_log_buf);
-		return 1;
+	for (i = 0; i < num_progs; i++) {
+		if (load_bpf_file(filename)) {
+			fprintf(stderr, "%s", bpf_log_buf);
+			return 1;
+		}
+		printf("prog #%d: map ids %d %d\n", i, map_fd[0], map_fd[1]);
+		map0_fds[i] = map_fd[0];
+		map1_fds[i] = map_fd[1];
 	}
 
 	/* current load_bpf_file has perf_event_open default pid = -1

@@ -64,8 +78,34 @@ int main(int argc, char **argv)
 	close(fd);
 
 	/* verify the map */
-	verify_map(map_fd[0]);
-	verify_map(map_fd[1]);
+	for (i = 0; i < num_progs; i++) {
+		verify_map(map0_fds[i]);
+		verify_map(map1_fds[i]);
+	}
 
 	return 0;
 }
+
+int main(int argc, char **argv)
+{
+	struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY};
+	int opt, num_progs = 1;
+	char filename[256];
+
+	while ((opt = getopt(argc, argv, "i:h")) != -1) {
+		switch (opt) {
+		case 'i':
+			num_progs = atoi(optarg);
+			break;
+		case 'h':
+		default:
+			usage(argv[0]);
+			return 0;
+		}
+	}
+
+	setrlimit(RLIMIT_MEMLOCK, &r);
+	snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
+
+	return test(filename, num_progs);
+}