xref: /openbmc/linux/kernel/trace/trace_boot.c (revision fddca7db)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * trace_boot.c
4  * Tracing kernel boot-time
5  */
6 
7 #define pr_fmt(fmt)	"trace_boot: " fmt
8 
9 #include <linux/bootconfig.h>
10 #include <linux/cpumask.h>
11 #include <linux/ftrace.h>
12 #include <linux/init.h>
13 #include <linux/kernel.h>
14 #include <linux/mutex.h>
15 #include <linux/string.h>
16 #include <linux/slab.h>
17 #include <linux/trace.h>
18 #include <linux/trace_events.h>
19 
20 #include "trace.h"
21 
22 #define MAX_BUF_LEN 256
23 
24 static void __init
trace_boot_set_instance_options(struct trace_array * tr,struct xbc_node * node)25 trace_boot_set_instance_options(struct trace_array *tr, struct xbc_node *node)
26 {
27 	struct xbc_node *anode;
28 	const char *p;
29 	char buf[MAX_BUF_LEN];
30 	unsigned long v = 0;
31 
32 	/* Common ftrace options */
33 	xbc_node_for_each_array_value(node, "options", anode, p) {
34 		if (strscpy(buf, p, ARRAY_SIZE(buf)) < 0) {
35 			pr_err("String is too long: %s\n", p);
36 			continue;
37 		}
38 
39 		if (trace_set_options(tr, buf) < 0)
40 			pr_err("Failed to set option: %s\n", buf);
41 	}
42 
43 	p = xbc_node_find_value(node, "tracing_on", NULL);
44 	if (p && *p != '\0') {
45 		if (kstrtoul(p, 10, &v))
46 			pr_err("Failed to set tracing on: %s\n", p);
47 		if (v)
48 			tracer_tracing_on(tr);
49 		else
50 			tracer_tracing_off(tr);
51 	}
52 
53 	p = xbc_node_find_value(node, "trace_clock", NULL);
54 	if (p && *p != '\0') {
55 		if (tracing_set_clock(tr, p) < 0)
56 			pr_err("Failed to set trace clock: %s\n", p);
57 	}
58 
59 	p = xbc_node_find_value(node, "buffer_size", NULL);
60 	if (p && *p != '\0') {
61 		v = memparse(p, NULL);
62 		if (v < PAGE_SIZE)
63 			pr_err("Buffer size is too small: %s\n", p);
64 		if (tracing_resize_ring_buffer(tr, v, RING_BUFFER_ALL_CPUS) < 0)
65 			pr_err("Failed to resize trace buffer to %s\n", p);
66 	}
67 
68 	p = xbc_node_find_value(node, "cpumask", NULL);
69 	if (p && *p != '\0') {
70 		cpumask_var_t new_mask;
71 
72 		if (alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
73 			if (cpumask_parse(p, new_mask) < 0 ||
74 			    tracing_set_cpumask(tr, new_mask) < 0)
75 				pr_err("Failed to set new CPU mask %s\n", p);
76 			free_cpumask_var(new_mask);
77 		}
78 	}
79 }
80 
81 #ifdef CONFIG_EVENT_TRACING
82 static void __init
trace_boot_enable_events(struct trace_array * tr,struct xbc_node * node)83 trace_boot_enable_events(struct trace_array *tr, struct xbc_node *node)
84 {
85 	struct xbc_node *anode;
86 	char buf[MAX_BUF_LEN];
87 	const char *p;
88 
89 	xbc_node_for_each_array_value(node, "events", anode, p) {
90 		if (strscpy(buf, p, ARRAY_SIZE(buf)) < 0) {
91 			pr_err("String is too long: %s\n", p);
92 			continue;
93 		}
94 
95 		if (ftrace_set_clr_event(tr, buf, 1) < 0)
96 			pr_err("Failed to enable event: %s\n", p);
97 	}
98 }
99 
100 #ifdef CONFIG_KPROBE_EVENTS
101 static int __init
trace_boot_add_kprobe_event(struct xbc_node * node,const char * event)102 trace_boot_add_kprobe_event(struct xbc_node *node, const char *event)
103 {
104 	struct dynevent_cmd cmd;
105 	struct xbc_node *anode;
106 	char buf[MAX_BUF_LEN];
107 	const char *val;
108 	int ret = 0;
109 
110 	xbc_node_for_each_array_value(node, "probes", anode, val) {
111 		kprobe_event_cmd_init(&cmd, buf, MAX_BUF_LEN);
112 
113 		ret = kprobe_event_gen_cmd_start(&cmd, event, val);
114 		if (ret) {
115 			pr_err("Failed to generate probe: %s\n", buf);
116 			break;
117 		}
118 
119 		ret = kprobe_event_gen_cmd_end(&cmd);
120 		if (ret) {
121 			pr_err("Failed to add probe: %s\n", buf);
122 			break;
123 		}
124 	}
125 
126 	return ret;
127 }
128 #else
129 static inline int __init
trace_boot_add_kprobe_event(struct xbc_node * node,const char * event)130 trace_boot_add_kprobe_event(struct xbc_node *node, const char *event)
131 {
132 	pr_err("Kprobe event is not supported.\n");
133 	return -ENOTSUPP;
134 }
135 #endif
136 
137 #ifdef CONFIG_SYNTH_EVENTS
138 static int __init
trace_boot_add_synth_event(struct xbc_node * node,const char * event)139 trace_boot_add_synth_event(struct xbc_node *node, const char *event)
140 {
141 	struct dynevent_cmd cmd;
142 	struct xbc_node *anode;
143 	char buf[MAX_BUF_LEN];
144 	const char *p;
145 	int ret;
146 
147 	synth_event_cmd_init(&cmd, buf, MAX_BUF_LEN);
148 
149 	ret = synth_event_gen_cmd_start(&cmd, event, NULL);
150 	if (ret)
151 		return ret;
152 
153 	xbc_node_for_each_array_value(node, "fields", anode, p) {
154 		ret = synth_event_add_field_str(&cmd, p);
155 		if (ret)
156 			return ret;
157 	}
158 
159 	ret = synth_event_gen_cmd_end(&cmd);
160 	if (ret < 0)
161 		pr_err("Failed to add synthetic event: %s\n", buf);
162 
163 	return ret;
164 }
165 #else
166 static inline int __init
trace_boot_add_synth_event(struct xbc_node * node,const char * event)167 trace_boot_add_synth_event(struct xbc_node *node, const char *event)
168 {
169 	pr_err("Synthetic event is not supported.\n");
170 	return -ENOTSUPP;
171 }
172 #endif
173 
174 #ifdef CONFIG_HIST_TRIGGERS
175 static int __init __printf(3, 4)
append_printf(char ** bufp,char * end,const char * fmt,...)176 append_printf(char **bufp, char *end, const char *fmt, ...)
177 {
178 	va_list args;
179 	int ret;
180 
181 	if (*bufp == end)
182 		return -ENOSPC;
183 
184 	va_start(args, fmt);
185 	ret = vsnprintf(*bufp, end - *bufp, fmt, args);
186 	if (ret < end - *bufp) {
187 		*bufp += ret;
188 	} else {
189 		*bufp = end;
190 		ret = -ERANGE;
191 	}
192 	va_end(args);
193 
194 	return ret;
195 }
196 
197 static int __init
append_str_nospace(char ** bufp,char * end,const char * str)198 append_str_nospace(char **bufp, char *end, const char *str)
199 {
200 	char *p = *bufp;
201 	int len;
202 
203 	while (p < end - 1 && *str != '\0') {
204 		if (!isspace(*str))
205 			*(p++) = *str;
206 		str++;
207 	}
208 	*p = '\0';
209 	if (p == end - 1) {
210 		*bufp = end;
211 		return -ENOSPC;
212 	}
213 	len = p - *bufp;
214 	*bufp = p;
215 	return (int)len;
216 }
217 
218 static int __init
trace_boot_hist_add_array(struct xbc_node * hnode,char ** bufp,char * end,const char * key)219 trace_boot_hist_add_array(struct xbc_node *hnode, char **bufp,
220 			  char *end, const char *key)
221 {
222 	struct xbc_node *anode;
223 	const char *p;
224 	char sep;
225 
226 	p = xbc_node_find_value(hnode, key, &anode);
227 	if (p) {
228 		if (!anode) {
229 			pr_err("hist.%s requires value(s).\n", key);
230 			return -EINVAL;
231 		}
232 
233 		append_printf(bufp, end, ":%s", key);
234 		sep = '=';
235 		xbc_array_for_each_value(anode, p) {
236 			append_printf(bufp, end, "%c%s", sep, p);
237 			if (sep == '=')
238 				sep = ',';
239 		}
240 	} else
241 		return -ENOENT;
242 
243 	return 0;
244 }
245 
246 static int __init
trace_boot_hist_add_one_handler(struct xbc_node * hnode,char ** bufp,char * end,const char * handler,const char * param)247 trace_boot_hist_add_one_handler(struct xbc_node *hnode, char **bufp,
248 				char *end, const char *handler,
249 				const char *param)
250 {
251 	struct xbc_node *knode, *anode;
252 	const char *p;
253 	char sep;
254 
255 	/* Compose 'handler' parameter */
256 	p = xbc_node_find_value(hnode, param, NULL);
257 	if (!p) {
258 		pr_err("hist.%s requires '%s' option.\n",
259 		       xbc_node_get_data(hnode), param);
260 		return -EINVAL;
261 	}
262 	append_printf(bufp, end, ":%s(%s)", handler, p);
263 
264 	/* Compose 'action' parameter */
265 	knode = xbc_node_find_subkey(hnode, "trace");
266 	if (!knode)
267 		knode = xbc_node_find_subkey(hnode, "save");
268 
269 	if (knode) {
270 		anode = xbc_node_get_child(knode);
271 		if (!anode || !xbc_node_is_value(anode)) {
272 			pr_err("hist.%s.%s requires value(s).\n",
273 			       xbc_node_get_data(hnode),
274 			       xbc_node_get_data(knode));
275 			return -EINVAL;
276 		}
277 
278 		append_printf(bufp, end, ".%s", xbc_node_get_data(knode));
279 		sep = '(';
280 		xbc_array_for_each_value(anode, p) {
281 			append_printf(bufp, end, "%c%s", sep, p);
282 			if (sep == '(')
283 				sep = ',';
284 		}
285 		append_printf(bufp, end, ")");
286 	} else if (xbc_node_find_subkey(hnode, "snapshot")) {
287 		append_printf(bufp, end, ".snapshot()");
288 	} else {
289 		pr_err("hist.%s requires an action.\n",
290 		       xbc_node_get_data(hnode));
291 		return -EINVAL;
292 	}
293 
294 	return 0;
295 }
296 
297 static int __init
trace_boot_hist_add_handlers(struct xbc_node * hnode,char ** bufp,char * end,const char * param)298 trace_boot_hist_add_handlers(struct xbc_node *hnode, char **bufp,
299 			     char *end, const char *param)
300 {
301 	struct xbc_node *node;
302 	const char *p, *handler;
303 	int ret = 0;
304 
305 	handler = xbc_node_get_data(hnode);
306 
307 	xbc_node_for_each_subkey(hnode, node) {
308 		p = xbc_node_get_data(node);
309 		if (!isdigit(p[0]))
310 			continue;
311 		/* All digit started node should be instances. */
312 		ret = trace_boot_hist_add_one_handler(node, bufp, end, handler, param);
313 		if (ret < 0)
314 			break;
315 	}
316 
317 	if (xbc_node_find_subkey(hnode, param))
318 		ret = trace_boot_hist_add_one_handler(hnode, bufp, end, handler, param);
319 
320 	return ret;
321 }
322 
323 /*
324  * Histogram boottime tracing syntax.
325  *
326  * ftrace.[instance.INSTANCE.]event.GROUP.EVENT.hist[.N] {
327  *	keys = <KEY>[,...]
328  *	values = <VAL>[,...]
329  *	sort = <SORT-KEY>[,...]
330  *	size = <ENTRIES>
331  *	name = <HISTNAME>
332  *	var { <VAR> = <EXPR> ... }
333  *	pause|continue|clear
334  *	onmax|onchange[.N] { var = <VAR>; <ACTION> [= <PARAM>] }
335  *	onmatch[.N] { event = <EVENT>; <ACTION> [= <PARAM>] }
336  *	filter = <FILTER>
337  * }
338  *
339  * Where <ACTION> are;
340  *
341  *	trace = <EVENT>, <ARG1>[, ...]
342  *	save = <ARG1>[, ...]
343  *	snapshot
344  */
345 static int __init
trace_boot_compose_hist_cmd(struct xbc_node * hnode,char * buf,size_t size)346 trace_boot_compose_hist_cmd(struct xbc_node *hnode, char *buf, size_t size)
347 {
348 	struct xbc_node *node, *knode;
349 	char *end = buf + size;
350 	const char *p;
351 	int ret = 0;
352 
353 	append_printf(&buf, end, "hist");
354 
355 	ret = trace_boot_hist_add_array(hnode, &buf, end, "keys");
356 	if (ret < 0) {
357 		if (ret == -ENOENT)
358 			pr_err("hist requires keys.\n");
359 		return -EINVAL;
360 	}
361 
362 	ret = trace_boot_hist_add_array(hnode, &buf, end, "values");
363 	if (ret == -EINVAL)
364 		return ret;
365 	ret = trace_boot_hist_add_array(hnode, &buf, end, "sort");
366 	if (ret == -EINVAL)
367 		return ret;
368 
369 	p = xbc_node_find_value(hnode, "size", NULL);
370 	if (p)
371 		append_printf(&buf, end, ":size=%s", p);
372 
373 	p = xbc_node_find_value(hnode, "name", NULL);
374 	if (p)
375 		append_printf(&buf, end, ":name=%s", p);
376 
377 	node = xbc_node_find_subkey(hnode, "var");
378 	if (node) {
379 		xbc_node_for_each_key_value(node, knode, p) {
380 			/* Expression must not include spaces. */
381 			append_printf(&buf, end, ":%s=",
382 				      xbc_node_get_data(knode));
383 			append_str_nospace(&buf, end, p);
384 		}
385 	}
386 
387 	/* Histogram control attributes (mutual exclusive) */
388 	if (xbc_node_find_value(hnode, "pause", NULL))
389 		append_printf(&buf, end, ":pause");
390 	else if (xbc_node_find_value(hnode, "continue", NULL))
391 		append_printf(&buf, end, ":continue");
392 	else if (xbc_node_find_value(hnode, "clear", NULL))
393 		append_printf(&buf, end, ":clear");
394 
395 	/* Histogram handler and actions */
396 	node = xbc_node_find_subkey(hnode, "onmax");
397 	if (node && trace_boot_hist_add_handlers(node, &buf, end, "var") < 0)
398 		return -EINVAL;
399 	node = xbc_node_find_subkey(hnode, "onchange");
400 	if (node && trace_boot_hist_add_handlers(node, &buf, end, "var") < 0)
401 		return -EINVAL;
402 	node = xbc_node_find_subkey(hnode, "onmatch");
403 	if (node && trace_boot_hist_add_handlers(node, &buf, end, "event") < 0)
404 		return -EINVAL;
405 
406 	p = xbc_node_find_value(hnode, "filter", NULL);
407 	if (p)
408 		append_printf(&buf, end, " if %s", p);
409 
410 	if (buf == end) {
411 		pr_err("hist exceeds the max command length.\n");
412 		return -E2BIG;
413 	}
414 
415 	return 0;
416 }
417 
418 static void __init
trace_boot_init_histograms(struct trace_event_file * file,struct xbc_node * hnode,char * buf,size_t size)419 trace_boot_init_histograms(struct trace_event_file *file,
420 			   struct xbc_node *hnode, char *buf, size_t size)
421 {
422 	struct xbc_node *node;
423 	const char *p;
424 	char *tmp;
425 
426 	xbc_node_for_each_subkey(hnode, node) {
427 		p = xbc_node_get_data(node);
428 		if (!isdigit(p[0]))
429 			continue;
430 		/* All digit started node should be instances. */
431 		if (trace_boot_compose_hist_cmd(node, buf, size) == 0) {
432 			tmp = kstrdup(buf, GFP_KERNEL);
433 			if (!tmp)
434 				return;
435 			if (trigger_process_regex(file, buf) < 0)
436 				pr_err("Failed to apply hist trigger: %s\n", tmp);
437 			kfree(tmp);
438 		}
439 	}
440 
441 	if (xbc_node_find_subkey(hnode, "keys")) {
442 		if (trace_boot_compose_hist_cmd(hnode, buf, size) == 0) {
443 			tmp = kstrdup(buf, GFP_KERNEL);
444 			if (!tmp)
445 				return;
446 			if (trigger_process_regex(file, buf) < 0)
447 				pr_err("Failed to apply hist trigger: %s\n", tmp);
448 			kfree(tmp);
449 		}
450 	}
451 }
452 #else
453 static void __init
trace_boot_init_histograms(struct trace_event_file * file,struct xbc_node * hnode,char * buf,size_t size)454 trace_boot_init_histograms(struct trace_event_file *file,
455 			   struct xbc_node *hnode, char *buf, size_t size)
456 {
457 	/* do nothing */
458 }
459 #endif
460 
461 static void __init
trace_boot_init_one_event(struct trace_array * tr,struct xbc_node * gnode,struct xbc_node * enode)462 trace_boot_init_one_event(struct trace_array *tr, struct xbc_node *gnode,
463 			  struct xbc_node *enode)
464 {
465 	struct trace_event_file *file;
466 	struct xbc_node *anode;
467 	char buf[MAX_BUF_LEN];
468 	const char *p, *group, *event;
469 
470 	group = xbc_node_get_data(gnode);
471 	event = xbc_node_get_data(enode);
472 
473 	if (!strcmp(group, "kprobes"))
474 		if (trace_boot_add_kprobe_event(enode, event) < 0)
475 			return;
476 	if (!strcmp(group, "synthetic"))
477 		if (trace_boot_add_synth_event(enode, event) < 0)
478 			return;
479 
480 	mutex_lock(&event_mutex);
481 	file = find_event_file(tr, group, event);
482 	if (!file) {
483 		pr_err("Failed to find event: %s:%s\n", group, event);
484 		goto out;
485 	}
486 
487 	p = xbc_node_find_value(enode, "filter", NULL);
488 	if (p && *p != '\0') {
489 		if (strscpy(buf, p, ARRAY_SIZE(buf)) < 0)
490 			pr_err("filter string is too long: %s\n", p);
491 		else if (apply_event_filter(file, buf) < 0)
492 			pr_err("Failed to apply filter: %s\n", buf);
493 	}
494 
495 	if (IS_ENABLED(CONFIG_HIST_TRIGGERS)) {
496 		xbc_node_for_each_array_value(enode, "actions", anode, p) {
497 			if (strscpy(buf, p, ARRAY_SIZE(buf)) < 0)
498 				pr_err("action string is too long: %s\n", p);
499 			else if (trigger_process_regex(file, buf) < 0)
500 				pr_err("Failed to apply an action: %s\n", p);
501 		}
502 		anode = xbc_node_find_subkey(enode, "hist");
503 		if (anode)
504 			trace_boot_init_histograms(file, anode, buf, ARRAY_SIZE(buf));
505 	} else if (xbc_node_find_value(enode, "actions", NULL))
506 		pr_err("Failed to apply event actions because CONFIG_HIST_TRIGGERS is not set.\n");
507 
508 	if (xbc_node_find_value(enode, "enable", NULL)) {
509 		if (trace_event_enable_disable(file, 1, 0) < 0)
510 			pr_err("Failed to enable event node: %s:%s\n",
511 				group, event);
512 	}
513 out:
514 	mutex_unlock(&event_mutex);
515 }
516 
517 static void __init
trace_boot_init_events(struct trace_array * tr,struct xbc_node * node)518 trace_boot_init_events(struct trace_array *tr, struct xbc_node *node)
519 {
520 	struct xbc_node *gnode, *enode;
521 	bool enable, enable_all = false;
522 	const char *data;
523 
524 	node = xbc_node_find_subkey(node, "event");
525 	if (!node)
526 		return;
527 	/* per-event key starts with "event.GROUP.EVENT" */
528 	xbc_node_for_each_subkey(node, gnode) {
529 		data = xbc_node_get_data(gnode);
530 		if (!strcmp(data, "enable")) {
531 			enable_all = true;
532 			continue;
533 		}
534 		enable = false;
535 		xbc_node_for_each_subkey(gnode, enode) {
536 			data = xbc_node_get_data(enode);
537 			if (!strcmp(data, "enable")) {
538 				enable = true;
539 				continue;
540 			}
541 			trace_boot_init_one_event(tr, gnode, enode);
542 		}
543 		/* Event enablement must be done after event settings */
544 		if (enable) {
545 			data = xbc_node_get_data(gnode);
546 			trace_array_set_clr_event(tr, data, NULL, true);
547 		}
548 	}
549 	/* Ditto */
550 	if (enable_all)
551 		trace_array_set_clr_event(tr, NULL, NULL, true);
552 }
553 #else
554 #define trace_boot_enable_events(tr, node) do {} while (0)
555 #define trace_boot_init_events(tr, node) do {} while (0)
556 #endif
557 
558 #ifdef CONFIG_DYNAMIC_FTRACE
559 static void __init
trace_boot_set_ftrace_filter(struct trace_array * tr,struct xbc_node * node)560 trace_boot_set_ftrace_filter(struct trace_array *tr, struct xbc_node *node)
561 {
562 	struct xbc_node *anode;
563 	const char *p;
564 	char *q;
565 
566 	xbc_node_for_each_array_value(node, "ftrace.filters", anode, p) {
567 		q = kstrdup(p, GFP_KERNEL);
568 		if (!q)
569 			return;
570 		if (ftrace_set_filter(tr->ops, q, strlen(q), 0) < 0)
571 			pr_err("Failed to add %s to ftrace filter\n", p);
572 		else
573 			ftrace_filter_param = true;
574 		kfree(q);
575 	}
576 	xbc_node_for_each_array_value(node, "ftrace.notraces", anode, p) {
577 		q = kstrdup(p, GFP_KERNEL);
578 		if (!q)
579 			return;
580 		if (ftrace_set_notrace(tr->ops, q, strlen(q), 0) < 0)
581 			pr_err("Failed to add %s to ftrace filter\n", p);
582 		else
583 			ftrace_filter_param = true;
584 		kfree(q);
585 	}
586 }
587 #else
588 #define trace_boot_set_ftrace_filter(tr, node) do {} while (0)
589 #endif
590 
591 static void __init
trace_boot_enable_tracer(struct trace_array * tr,struct xbc_node * node)592 trace_boot_enable_tracer(struct trace_array *tr, struct xbc_node *node)
593 {
594 	const char *p;
595 
596 	trace_boot_set_ftrace_filter(tr, node);
597 
598 	p = xbc_node_find_value(node, "tracer", NULL);
599 	if (p && *p != '\0') {
600 		if (tracing_set_tracer(tr, p) < 0)
601 			pr_err("Failed to set given tracer: %s\n", p);
602 	}
603 
604 	/* Since tracer can free snapshot buffer, allocate snapshot here.*/
605 	if (xbc_node_find_value(node, "alloc_snapshot", NULL)) {
606 		if (tracing_alloc_snapshot_instance(tr) < 0)
607 			pr_err("Failed to allocate snapshot buffer\n");
608 	}
609 }
610 
611 static void __init
trace_boot_init_one_instance(struct trace_array * tr,struct xbc_node * node)612 trace_boot_init_one_instance(struct trace_array *tr, struct xbc_node *node)
613 {
614 	trace_boot_set_instance_options(tr, node);
615 	trace_boot_init_events(tr, node);
616 	trace_boot_enable_events(tr, node);
617 	trace_boot_enable_tracer(tr, node);
618 }
619 
620 static void __init
trace_boot_init_instances(struct xbc_node * node)621 trace_boot_init_instances(struct xbc_node *node)
622 {
623 	struct xbc_node *inode;
624 	struct trace_array *tr;
625 	const char *p;
626 
627 	node = xbc_node_find_subkey(node, "instance");
628 	if (!node)
629 		return;
630 
631 	xbc_node_for_each_subkey(node, inode) {
632 		p = xbc_node_get_data(inode);
633 		if (!p || *p == '\0')
634 			continue;
635 
636 		tr = trace_array_get_by_name(p);
637 		if (!tr) {
638 			pr_err("Failed to get trace instance %s\n", p);
639 			continue;
640 		}
641 		trace_boot_init_one_instance(tr, inode);
642 		trace_array_put(tr);
643 	}
644 }
645 
trace_boot_init(void)646 static int __init trace_boot_init(void)
647 {
648 	struct xbc_node *trace_node;
649 	struct trace_array *tr;
650 
651 	trace_node = xbc_find_node("ftrace");
652 	if (!trace_node)
653 		return 0;
654 
655 	tr = top_trace_array();
656 	if (!tr)
657 		return 0;
658 
659 	/* Global trace array is also one instance */
660 	trace_boot_init_one_instance(tr, trace_node);
661 	trace_boot_init_instances(trace_node);
662 
663 	disable_tracing_selftest("running boot-time tracing");
664 
665 	return 0;
666 }
667 /*
668  * Start tracing at the end of core-initcall, so that it starts tracing
669  * from the beginning of postcore_initcall.
670  */
671 core_initcall_sync(trace_boot_init);
672