trace_syscalls.c 12 KB
Newer Older
1
#include <trace/syscall.h>
2
#include <trace/events/syscalls.h>
3
#include <linux/kernel.h>
4
#include <linux/ftrace.h>
5
#include <linux/perf_counter.h>
6 7 8 9 10
#include <asm/syscall.h>

#include "trace_output.h"
#include "trace.h"

11
static DEFINE_MUTEX(syscall_trace_lock);
12 13 14 15
static int sys_refcount_enter;
static int sys_refcount_exit;
static DECLARE_BITMAP(enabled_enter_syscalls, FTRACE_SYSCALL_MAX);
static DECLARE_BITMAP(enabled_exit_syscalls, FTRACE_SYSCALL_MAX);
16

17 18 19 20 21 22 23 24 25
enum print_line_t
print_syscall_enter(struct trace_iterator *iter, int flags)
{
	struct trace_seq *s = &iter->seq;
	struct trace_entry *ent = iter->ent;
	struct syscall_trace_enter *trace;
	struct syscall_metadata *entry;
	int i, ret, syscall;

26
	trace = (typeof(trace))ent;
27 28
	syscall = trace->nr;
	entry = syscall_nr_to_meta(syscall);
29

30 31 32
	if (!entry)
		goto end;

33 34 35 36 37
	if (entry->enter_id != ent->type) {
		WARN_ON_ONCE(1);
		goto end;
	}

38 39 40 41 42 43
	ret = trace_seq_printf(s, "%s(", entry->name);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	for (i = 0; i < entry->nb_args; i++) {
		/* parameter types */
44
		if (trace_flags & TRACE_ITER_VERBOSE) {
45 46 47 48 49
			ret = trace_seq_printf(s, "%s ", entry->types[i]);
			if (!ret)
				return TRACE_TYPE_PARTIAL_LINE;
		}
		/* parameter values */
50
		ret = trace_seq_printf(s, "%s: %lx%s", entry->args[i],
51
				       trace->args[i],
52
				       i == entry->nb_args - 1 ? "" : ", ");
53 54 55 56
		if (!ret)
			return TRACE_TYPE_PARTIAL_LINE;
	}

57 58 59 60
	ret = trace_seq_putc(s, ')');
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

61
end:
62 63 64 65
	ret =  trace_seq_putc(s, '\n');
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

66 67 68 69 70 71 72 73 74 75 76 77 78
	return TRACE_TYPE_HANDLED;
}

enum print_line_t
print_syscall_exit(struct trace_iterator *iter, int flags)
{
	struct trace_seq *s = &iter->seq;
	struct trace_entry *ent = iter->ent;
	struct syscall_trace_exit *trace;
	int syscall;
	struct syscall_metadata *entry;
	int ret;

79
	trace = (typeof(trace))ent;
80 81
	syscall = trace->nr;
	entry = syscall_nr_to_meta(syscall);
82

83 84 85 86 87
	if (!entry) {
		trace_seq_printf(s, "\n");
		return TRACE_TYPE_HANDLED;
	}

88 89 90 91 92
	if (entry->exit_id != ent->type) {
		WARN_ON_ONCE(1);
		return TRACE_TYPE_UNHANDLED;
	}

93 94 95 96 97 98 99 100
	ret = trace_seq_printf(s, "%s -> 0x%lx\n", entry->name,
				trace->ret);
	if (!ret)
		return TRACE_TYPE_PARTIAL_LINE;

	return TRACE_TYPE_HANDLED;
}

101 102 103 104 105 106 107
extern char *__bad_type_size(void);

#define SYSCALL_FIELD(type, name)					\
	sizeof(type) != sizeof(trace.name) ?				\
		__bad_type_size() :					\
		#type, #name, offsetof(typeof(trace), name), sizeof(trace.name)

108
int syscall_enter_format(struct ftrace_event_call *call, struct trace_seq *s)
109 110 111
{
	int i;
	int nr;
112
	int ret;
113
	struct syscall_metadata *entry;
114 115
	struct syscall_trace_enter trace;
	int offset = offsetof(struct syscall_trace_enter, args);
116

117
	nr = syscall_name_to_nr(call->data);
118 119 120
	entry = syscall_nr_to_meta(nr);

	if (!entry)
121 122 123 124 125 126
		return 0;

	ret = trace_seq_printf(s, "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n",
			       SYSCALL_FIELD(int, nr));
	if (!ret)
		return 0;
127 128 129 130 131 132

	for (i = 0; i < entry->nb_args; i++) {
		ret = trace_seq_printf(s, "\tfield:%s %s;", entry->types[i],
				        entry->args[i]);
		if (!ret)
			return 0;
133
		ret = trace_seq_printf(s, "\toffset:%d;\tsize:%zu;\n", offset,
134 135 136 137 138 139
				       sizeof(unsigned long));
		if (!ret)
			return 0;
		offset += sizeof(unsigned long);
	}

140
	trace_seq_puts(s, "\nprint fmt: \"");
141
	for (i = 0; i < entry->nb_args; i++) {
142
		ret = trace_seq_printf(s, "%s: 0x%%0%zulx%s", entry->args[i],
143
				        sizeof(unsigned long),
144
					i == entry->nb_args - 1 ? "" : ", ");
145 146 147
		if (!ret)
			return 0;
	}
148
	trace_seq_putc(s, '"');
149 150

	for (i = 0; i < entry->nb_args; i++) {
151 152
		ret = trace_seq_printf(s, ", ((unsigned long)(REC->%s))",
				       entry->args[i]);
153 154 155 156
		if (!ret)
			return 0;
	}

157
	return trace_seq_putc(s, '\n');
158 159
}

160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
int syscall_exit_format(struct ftrace_event_call *call, struct trace_seq *s)
{
	int ret;
	struct syscall_trace_exit trace;

	ret = trace_seq_printf(s,
			       "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n"
			       "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\n",
			       SYSCALL_FIELD(int, nr),
			       SYSCALL_FIELD(unsigned long, ret));
	if (!ret)
		return 0;

	return trace_seq_printf(s, "\nprint fmt: \"0x%%lx\", REC->ret\n");
}

176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195
int syscall_enter_define_fields(struct ftrace_event_call *call)
{
	struct syscall_trace_enter trace;
	struct syscall_metadata *meta;
	int ret;
	int nr;
	int i;
	int offset = offsetof(typeof(trace), args);

	nr = syscall_name_to_nr(call->data);
	meta = syscall_nr_to_meta(nr);

	if (!meta)
		return 0;

	ret = trace_define_common_fields(call);
	if (ret)
		return ret;

	for (i = 0; i < meta->nb_args; i++) {
196 197
		ret = trace_define_field(call, (char *)meta->types[i],
					 (char *)meta->args[i], offset,
198 199
					 sizeof(unsigned long), 0,
					 FILTER_OTHER);
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
		offset += sizeof(unsigned long);
	}

	return ret;
}

int syscall_exit_define_fields(struct ftrace_event_call *call)
{
	struct syscall_trace_exit trace;
	int ret;

	ret = trace_define_common_fields(call);
	if (ret)
		return ret;

215 216
	ret = trace_define_field(call, SYSCALL_FIELD(unsigned long, ret), 0,
				 FILTER_OTHER);
217 218 219 220

	return ret;
}

221
void ftrace_syscall_enter(struct pt_regs *regs, long id)
222
{
223 224 225 226
	struct syscall_trace_enter *entry;
	struct syscall_metadata *sys_data;
	struct ring_buffer_event *event;
	int size;
227 228 229
	int syscall_nr;

	syscall_nr = syscall_get_nr(current, regs);
230 231
	if (!test_bit(syscall_nr, enabled_enter_syscalls))
		return;
232

233 234 235 236 237 238
	sys_data = syscall_nr_to_meta(syscall_nr);
	if (!sys_data)
		return;

	size = sizeof(*entry) + sizeof(unsigned long) * sys_data->nb_args;

239
	event = trace_current_buffer_lock_reserve(sys_data->enter_id, size,
240 241 242 243 244 245 246 247
							0, 0);
	if (!event)
		return;

	entry = ring_buffer_event_data(event);
	entry->nr = syscall_nr;
	syscall_get_arguments(current, regs, 0, sys_data->nb_args, entry->args);

248 249
	if (!filter_current_check_discard(sys_data->enter_event, entry, event))
		trace_current_buffer_unlock_commit(event, 0, 0);
250 251
}

252
void ftrace_syscall_exit(struct pt_regs *regs, long ret)
253
{
254 255 256
	struct syscall_trace_exit *entry;
	struct syscall_metadata *sys_data;
	struct ring_buffer_event *event;
257 258 259
	int syscall_nr;

	syscall_nr = syscall_get_nr(current, regs);
260 261
	if (!test_bit(syscall_nr, enabled_exit_syscalls))
		return;
262

263 264 265 266
	sys_data = syscall_nr_to_meta(syscall_nr);
	if (!sys_data)
		return;

267
	event = trace_current_buffer_lock_reserve(sys_data->exit_id,
268 269 270 271 272 273 274 275
				sizeof(*entry), 0, 0);
	if (!event)
		return;

	entry = ring_buffer_event_data(event);
	entry->nr = syscall_nr;
	entry->ret = syscall_get_return_value(current, regs);

276 277
	if (!filter_current_check_discard(sys_data->exit_event, entry, event))
		trace_current_buffer_unlock_commit(event, 0, 0);
278 279
}

280
int reg_event_syscall_enter(struct ftrace_event_call *call)
281
{
282 283 284 285
	int ret = 0;
	int num;
	char *name;

286
	name = (char *)call->data;
287 288 289 290 291
	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return -ENOSYS;
	mutex_lock(&syscall_trace_lock);
	if (!sys_refcount_enter)
292
		ret = register_trace_sys_enter(ftrace_syscall_enter);
293 294 295 296 297 298 299 300 301
	if (ret) {
		pr_info("event trace: Could not activate"
				"syscall entry trace point");
	} else {
		set_bit(num, enabled_enter_syscalls);
		sys_refcount_enter++;
	}
	mutex_unlock(&syscall_trace_lock);
	return ret;
302 303
}

304
void unreg_event_syscall_enter(struct ftrace_event_call *call)
305
{
306 307
	int num;
	char *name;
308

309
	name = (char *)call->data;
310 311 312 313 314 315 316
	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return;
	mutex_lock(&syscall_trace_lock);
	sys_refcount_enter--;
	clear_bit(num, enabled_enter_syscalls);
	if (!sys_refcount_enter)
317
		unregister_trace_sys_enter(ftrace_syscall_enter);
318 319
	mutex_unlock(&syscall_trace_lock);
}
320

321
int reg_event_syscall_exit(struct ftrace_event_call *call)
322
{
323 324 325 326
	int ret = 0;
	int num;
	char *name;

327
	name = (char *)call->data;
328 329 330 331 332
	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return -ENOSYS;
	mutex_lock(&syscall_trace_lock);
	if (!sys_refcount_exit)
333
		ret = register_trace_sys_exit(ftrace_syscall_exit);
334 335 336 337 338 339
	if (ret) {
		pr_info("event trace: Could not activate"
				"syscall exit trace point");
	} else {
		set_bit(num, enabled_exit_syscalls);
		sys_refcount_exit++;
340
	}
341 342 343
	mutex_unlock(&syscall_trace_lock);
	return ret;
}
344

345
void unreg_event_syscall_exit(struct ftrace_event_call *call)
346 347 348
{
	int num;
	char *name;
349

350
	name = (char *)call->data;
351 352 353 354 355 356 357
	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return;
	mutex_lock(&syscall_trace_lock);
	sys_refcount_exit--;
	clear_bit(num, enabled_exit_syscalls);
	if (!sys_refcount_exit)
358
		unregister_trace_sys_exit(ftrace_syscall_exit);
359
	mutex_unlock(&syscall_trace_lock);
360
}
361 362 363 364 365 366 367 368

struct trace_event event_syscall_enter = {
	.trace			= print_syscall_enter,
};

struct trace_event event_syscall_exit = {
	.trace			= print_syscall_exit,
};
369 370

#ifdef CONFIG_EVENT_PROFILE
371

372 373 374 375 376 377 378
static DECLARE_BITMAP(enabled_prof_enter_syscalls, FTRACE_SYSCALL_MAX);
static DECLARE_BITMAP(enabled_prof_exit_syscalls, FTRACE_SYSCALL_MAX);
static int sys_prof_refcount_enter;
static int sys_prof_refcount_exit;

static void prof_syscall_enter(struct pt_regs *regs, long id)
{
379
	struct syscall_trace_enter *rec;
380 381
	struct syscall_metadata *sys_data;
	int syscall_nr;
382
	int size;
383 384 385 386 387 388 389 390 391

	syscall_nr = syscall_get_nr(current, regs);
	if (!test_bit(syscall_nr, enabled_prof_enter_syscalls))
		return;

	sys_data = syscall_nr_to_meta(syscall_nr);
	if (!sys_data)
		return;

392 393 394 395 396 397 398 399 400 401 402
	/* get the size after alignment with the u32 buffer size field */
	size = sizeof(unsigned long) * sys_data->nb_args + sizeof(*rec);
	size = ALIGN(size + sizeof(u32), sizeof(u64));
	size -= sizeof(u32);

	do {
		char raw_data[size];

		/* zero the dead bytes from align to not leak stack to user */
		*(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL;

403 404 405 406
		rec = (struct syscall_trace_enter *) raw_data;
		tracing_generic_entry_update(&rec->ent, 0, 0);
		rec->ent.type = sys_data->enter_id;
		rec->nr = syscall_nr;
407 408 409 410
		syscall_get_arguments(current, regs, 0, sys_data->nb_args,
				       (unsigned long *)&rec->args);
		perf_tpcounter_event(sys_data->enter_id, 0, 1, rec, size);
	} while(0);
411 412 413 414 415 416 417 418 419 420 421 422 423
}

int reg_prof_syscall_enter(char *name)
{
	int ret = 0;
	int num;

	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return -ENOSYS;

	mutex_lock(&syscall_trace_lock);
	if (!sys_prof_refcount_enter)
424
		ret = register_trace_sys_enter(prof_syscall_enter);
425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
	if (ret) {
		pr_info("event trace: Could not activate"
				"syscall entry trace point");
	} else {
		set_bit(num, enabled_prof_enter_syscalls);
		sys_prof_refcount_enter++;
	}
	mutex_unlock(&syscall_trace_lock);
	return ret;
}

void unreg_prof_syscall_enter(char *name)
{
	int num;

	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return;

	mutex_lock(&syscall_trace_lock);
	sys_prof_refcount_enter--;
	clear_bit(num, enabled_prof_enter_syscalls);
	if (!sys_prof_refcount_enter)
448
		unregister_trace_sys_enter(prof_syscall_enter);
449 450 451 452 453 454
	mutex_unlock(&syscall_trace_lock);
}

static void prof_syscall_exit(struct pt_regs *regs, long ret)
{
	struct syscall_metadata *sys_data;
455
	struct syscall_trace_exit rec;
456 457 458 459 460 461 462 463 464 465
	int syscall_nr;

	syscall_nr = syscall_get_nr(current, regs);
	if (!test_bit(syscall_nr, enabled_prof_exit_syscalls))
		return;

	sys_data = syscall_nr_to_meta(syscall_nr);
	if (!sys_data)
		return;

466 467 468
	tracing_generic_entry_update(&rec.ent, 0, 0);
	rec.ent.type = sys_data->exit_id;
	rec.nr = syscall_nr;
469 470 471
	rec.ret = syscall_get_return_value(current, regs);

	perf_tpcounter_event(sys_data->exit_id, 0, 1, &rec, sizeof(rec));
472 473 474 475 476 477 478 479 480 481 482 483 484
}

int reg_prof_syscall_exit(char *name)
{
	int ret = 0;
	int num;

	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return -ENOSYS;

	mutex_lock(&syscall_trace_lock);
	if (!sys_prof_refcount_exit)
485
		ret = register_trace_sys_exit(prof_syscall_exit);
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508
	if (ret) {
		pr_info("event trace: Could not activate"
				"syscall entry trace point");
	} else {
		set_bit(num, enabled_prof_exit_syscalls);
		sys_prof_refcount_exit++;
	}
	mutex_unlock(&syscall_trace_lock);
	return ret;
}

void unreg_prof_syscall_exit(char *name)
{
	int num;

	num = syscall_name_to_nr(name);
	if (num < 0 || num >= FTRACE_SYSCALL_MAX)
		return;

	mutex_lock(&syscall_trace_lock);
	sys_prof_refcount_exit--;
	clear_bit(num, enabled_prof_exit_syscalls);
	if (!sys_prof_refcount_exit)
509
		unregister_trace_sys_exit(prof_syscall_exit);
510 511 512 513 514 515
	mutex_unlock(&syscall_trace_lock);
}

#endif