ftrace.c 12.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 * Code for replacing ftrace calls with jumps.
 *
 * Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
 *
 * Thanks goes to Ingo Molnar, for suggesting the idea.
 * Mathieu Desnoyers, for suggesting postponing the modifications.
 * Arjan van de Ven, for keeping me straight, and explaining to me
 * the dangers of modifying code on the run.
 */

#include <linux/spinlock.h>
#include <linux/hardirq.h>
14
#include <linux/uaccess.h>
15 16
#include <linux/ftrace.h>
#include <linux/percpu.h>
17
#include <linux/sched.h>
18 19 20
#include <linux/init.h>
#include <linux/list.h>

21 22
#include <trace/syscall.h>

23
#include <asm/cacheflush.h>
24
#include <asm/ftrace.h>
25
#include <asm/nops.h>
26
#include <asm/nmi.h>
27 28


29
#ifdef CONFIG_DYNAMIC_FTRACE
30

31 32 33 34 35 36 37 38 39 40 41 42
int ftrace_arch_code_modify_prepare(void)
{
	set_kernel_text_rw();
	return 0;
}

int ftrace_arch_code_modify_post_process(void)
{
	set_kernel_text_ro();
	return 0;
}

43
union ftrace_code_union {
44
	char code[MCOUNT_INSN_SIZE];
45 46 47 48 49 50
	struct {
		char e8;
		int offset;
	} __attribute__((packed));
};

51
static int ftrace_calc_offset(long ip, long addr)
52 53 54
{
	return (int)(addr - ip);
}
55

56
static unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
57 58
{
	static union ftrace_code_union calc;
59

60
	calc.e8		= 0xe8;
61
	calc.offset	= ftrace_calc_offset(ip + MCOUNT_INSN_SIZE, addr);
62 63 64 65 66 67

	/*
	 * No locking needed, this must be called via kstop_machine
	 * which in essence is like running on a uniprocessor machine.
	 */
	return calc.code;
68 69
}

70 71 72 73 74 75 76 77 78 79 80
/*
 * Modifying code must take extra care. On an SMP machine, if
 * the code being modified is also being executed on another CPU
 * that CPU will have undefined results and possibly take a GPF.
 * We use kstop_machine to stop other CPUS from exectuing code.
 * But this does not stop NMIs from happening. We still need
 * to protect against that. We separate out the modification of
 * the code to take care of this.
 *
 * Two buffers are added: An IP buffer and a "code" buffer.
 *
S
Steven Rostedt 已提交
81
 * 1) Put the instruction pointer into the IP buffer
82
 *    and the new code into the "code" buffer.
L
Lai Jiangshan 已提交
83 84 85 86 87
 * 2) Wait for any running NMIs to finish and set a flag that says
 *    we are modifying code, it is done in an atomic operation.
 * 3) Write the code
 * 4) clear the flag.
 * 5) Wait for any running NMIs to finish.
88 89 90 91 92 93 94 95 96 97 98
 *
 * If an NMI is executed, the first thing it does is to call
 * "ftrace_nmi_enter". This will check if the flag is set to write
 * and if it is, it will write what is in the IP and "code" buffers.
 *
 * The trick is, it does not matter if everyone is writing the same
 * content to the code location. Also, if a CPU is executing code
 * it is OK to write to that code location if the contents being written
 * are the same as what exists.
 */

L
Lai Jiangshan 已提交
99
#define MOD_CODE_WRITE_FLAG (1 << 31)	/* set when NMI should do the write */
100
static atomic_t nmi_running = ATOMIC_INIT(0);
S
Steven Rostedt 已提交
101 102 103
static int mod_code_status;		/* holds return value of text write */
static void *mod_code_ip;		/* holds the IP to write to */
static void *mod_code_newcode;		/* holds the text to write to the IP */
104

S
Steven Rostedt 已提交
105 106
static unsigned nmi_wait_count;
static atomic_t nmi_update_count = ATOMIC_INIT(0);
S
Steven Rostedt 已提交
107 108 109 110 111 112 113 114 115 116 117

int ftrace_arch_read_dyn_info(char *buf, int size)
{
	int r;

	r = snprintf(buf, size, "%u %u",
		     nmi_wait_count,
		     atomic_read(&nmi_update_count));
	return r;
}

L
Lai Jiangshan 已提交
118 119 120 121 122 123 124 125 126 127 128 129 130 131
static void clear_mod_flag(void)
{
	int old = atomic_read(&nmi_running);

	for (;;) {
		int new = old & ~MOD_CODE_WRITE_FLAG;

		if (old == new)
			break;

		old = atomic_cmpxchg(&nmi_running, old, new);
	}
}

132 133 134 135 136 137 138 139 140 141
static void ftrace_mod_code(void)
{
	/*
	 * Yes, more than one CPU process can be writing to mod_code_status.
	 *    (and the code itself)
	 * But if one were to fail, then they all should, and if one were
	 * to succeed, then they all should.
	 */
	mod_code_status = probe_kernel_write(mod_code_ip, mod_code_newcode,
					     MCOUNT_INSN_SIZE);
142 143 144

	/* if we fail, then kill any new writers */
	if (mod_code_status)
L
Lai Jiangshan 已提交
145
		clear_mod_flag();
146 147
}

148
void ftrace_nmi_enter(void)
149
{
L
Lai Jiangshan 已提交
150 151
	if (atomic_inc_return(&nmi_running) & MOD_CODE_WRITE_FLAG) {
		smp_rmb();
152
		ftrace_mod_code();
S
Steven Rostedt 已提交
153 154
		atomic_inc(&nmi_update_count);
	}
L
Lai Jiangshan 已提交
155 156
	/* Must have previous changes seen before executions */
	smp_mb();
157 158
}

159
void ftrace_nmi_exit(void)
160
{
161
	/* Finish all executions before clearing nmi_running */
L
Lai Jiangshan 已提交
162
	smp_mb();
163
	atomic_dec(&nmi_running);
164 165
}

L
Lai Jiangshan 已提交
166 167 168 169 170 171 172 173 174 175 176 177
static void wait_for_nmi_and_set_mod_flag(void)
{
	if (!atomic_cmpxchg(&nmi_running, 0, MOD_CODE_WRITE_FLAG))
		return;

	do {
		cpu_relax();
	} while (atomic_cmpxchg(&nmi_running, 0, MOD_CODE_WRITE_FLAG));

	nmi_wait_count++;
}

178 179
static void wait_for_nmi(void)
{
180
	if (!atomic_read(&nmi_running))
181
		return;
S
Steven Rostedt 已提交
182

183
	do {
184
		cpu_relax();
185
	} while (atomic_read(&nmi_running));
S
Steven Rostedt 已提交
186

187
	nmi_wait_count++;
188 189 190 191 192 193 194 195 196 197 198
}

static int
do_ftrace_mod_code(unsigned long ip, void *new_code)
{
	mod_code_ip = (void *)ip;
	mod_code_newcode = new_code;

	/* The buffers need to be visible before we let NMIs write them */
	smp_mb();

L
Lai Jiangshan 已提交
199
	wait_for_nmi_and_set_mod_flag();
200 201 202 203 204 205 206 207 208

	/* Make sure all running NMIs have finished before we write the code */
	smp_mb();

	ftrace_mod_code();

	/* Make sure the write happens before clearing the bit */
	smp_mb();

L
Lai Jiangshan 已提交
209
	clear_mod_flag();
210 211 212 213 214 215
	wait_for_nmi();

	return mod_code_status;
}


216 217 218 219


static unsigned char ftrace_nop[MCOUNT_INSN_SIZE];

220
static unsigned char *ftrace_nop_replace(void)
221 222 223 224
{
	return ftrace_nop;
}

225
static int
226 227 228
ftrace_modify_code(unsigned long ip, unsigned char *old_code,
		   unsigned char *new_code)
{
229
	unsigned char replaced[MCOUNT_INSN_SIZE];
230 231 232 233

	/*
	 * Note: Due to modules and __init, code can
	 *  disappear and change, we need to protect against faulting
S
Steven Rostedt 已提交
234
	 *  as well as code changing. We do this by using the
S
Steven Rostedt 已提交
235
	 *  probe_kernel_* functions.
236 237
	 *
	 * No real locking needed, this code is run through
238
	 * kstop_machine, or before SMP starts.
239
	 */
S
Steven Rostedt 已提交
240 241

	/* read the text we want to modify */
S
Steven Rostedt 已提交
242
	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
243
		return -EFAULT;
244

S
Steven Rostedt 已提交
245
	/* Make sure it is what we expect it to be */
246
	if (memcmp(replaced, old_code, MCOUNT_INSN_SIZE) != 0)
247
		return -EINVAL;
248

S
Steven Rostedt 已提交
249
	/* replace the text with the new text */
250
	if (do_ftrace_mod_code(ip, new_code))
251
		return -EPERM;
252 253

	sync_core();
254

255
	return 0;
256 257
}

258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280
int ftrace_make_nop(struct module *mod,
		    struct dyn_ftrace *rec, unsigned long addr)
{
	unsigned char *new, *old;
	unsigned long ip = rec->ip;

	old = ftrace_call_replace(ip, addr);
	new = ftrace_nop_replace();

	return ftrace_modify_code(rec->ip, old, new);
}

int ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
{
	unsigned char *new, *old;
	unsigned long ip = rec->ip;

	old = ftrace_nop_replace();
	new = ftrace_call_replace(ip, addr);

	return ftrace_modify_code(rec->ip, old, new);
}

281
int ftrace_update_ftrace_func(ftrace_func_t func)
282 283
{
	unsigned long ip = (unsigned long)(&ftrace_call);
284
	unsigned char old[MCOUNT_INSN_SIZE], *new;
285 286
	int ret;

287
	memcpy(old, &ftrace_call, MCOUNT_INSN_SIZE);
288 289 290 291 292 293 294
	new = ftrace_call_replace(ip, (unsigned long)func);
	ret = ftrace_modify_code(ip, old, new);

	return ret;
}

int __init ftrace_dyn_arch_init(void *data)
295
{
296 297 298 299
	extern const unsigned char ftrace_test_p6nop[];
	extern const unsigned char ftrace_test_nop5[];
	extern const unsigned char ftrace_test_jmp[];
	int faulted = 0;
300

301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
	/*
	 * There is no good nop for all x86 archs.
	 * We will default to using the P6_NOP5, but first we
	 * will test to make sure that the nop will actually
	 * work on this CPU. If it faults, we will then
	 * go to a lesser efficient 5 byte nop. If that fails
	 * we then just use a jmp as our nop. This isn't the most
	 * efficient nop, but we can not use a multi part nop
	 * since we would then risk being preempted in the middle
	 * of that nop, and if we enabled tracing then, it might
	 * cause a system crash.
	 *
	 * TODO: check the cpuid to determine the best nop.
	 */
	asm volatile (
		"ftrace_test_jmp:"
		"jmp ftrace_test_p6nop\n"
318 319 320
		"nop\n"
		"nop\n"
		"nop\n"  /* 2 byte jmp + 3 bytes */
321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
		"ftrace_test_p6nop:"
		P6_NOP5
		"jmp 1f\n"
		"ftrace_test_nop5:"
		".byte 0x66,0x66,0x66,0x66,0x90\n"
		"1:"
		".section .fixup, \"ax\"\n"
		"2:	movl $1, %0\n"
		"	jmp ftrace_test_nop5\n"
		"3:	movl $2, %0\n"
		"	jmp 1b\n"
		".previous\n"
		_ASM_EXTABLE(ftrace_test_p6nop, 2b)
		_ASM_EXTABLE(ftrace_test_nop5, 3b)
		: "=r"(faulted) : "0" (faulted));

	switch (faulted) {
	case 0:
		pr_info("ftrace: converting mcount calls to 0f 1f 44 00 00\n");
340
		memcpy(ftrace_nop, ftrace_test_p6nop, MCOUNT_INSN_SIZE);
341 342 343
		break;
	case 1:
		pr_info("ftrace: converting mcount calls to 66 66 66 66 90\n");
344
		memcpy(ftrace_nop, ftrace_test_nop5, MCOUNT_INSN_SIZE);
345 346
		break;
	case 2:
347
		pr_info("ftrace: converting mcount calls to jmp . + 5\n");
348
		memcpy(ftrace_nop, ftrace_test_jmp, MCOUNT_INSN_SIZE);
349 350 351 352 353
		break;
	}

	/* The return code is retured via data */
	*(unsigned long *)data = 0;
S
Steven Rostedt 已提交
354

355 356
	return 0;
}
357
#endif
358

359
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
360

361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404
#ifdef CONFIG_DYNAMIC_FTRACE
extern void ftrace_graph_call(void);

static int ftrace_mod_jmp(unsigned long ip,
			  int old_offset, int new_offset)
{
	unsigned char code[MCOUNT_INSN_SIZE];

	if (probe_kernel_read(code, (void *)ip, MCOUNT_INSN_SIZE))
		return -EFAULT;

	if (code[0] != 0xe9 || old_offset != *(int *)(&code[1]))
		return -EINVAL;

	*(int *)(&code[1]) = new_offset;

	if (do_ftrace_mod_code(ip, &code))
		return -EPERM;

	return 0;
}

int ftrace_enable_ftrace_graph_caller(void)
{
	unsigned long ip = (unsigned long)(&ftrace_graph_call);
	int old_offset, new_offset;

	old_offset = (unsigned long)(&ftrace_stub) - (ip + MCOUNT_INSN_SIZE);
	new_offset = (unsigned long)(&ftrace_graph_caller) - (ip + MCOUNT_INSN_SIZE);

	return ftrace_mod_jmp(ip, old_offset, new_offset);
}

int ftrace_disable_ftrace_graph_caller(void)
{
	unsigned long ip = (unsigned long)(&ftrace_graph_call);
	int old_offset, new_offset;

	old_offset = (unsigned long)(&ftrace_graph_caller) - (ip + MCOUNT_INSN_SIZE);
	new_offset = (unsigned long)(&ftrace_stub) - (ip + MCOUNT_INSN_SIZE);

	return ftrace_mod_jmp(ip, old_offset, new_offset);
}

405 406 407 408 409 410 411 412 413 414
#endif /* !CONFIG_DYNAMIC_FTRACE */

/*
 * Hook the return address and push it in the stack of return addrs
 * in current thread info.
 */
void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
{
	unsigned long old;
	int faulted;
415
	struct ftrace_graph_ent trace;
416 417 418 419
	unsigned long return_hooker = (unsigned long)
				&return_to_handler;

	/* Nmi's are currently unsupported */
420
	if (unlikely(in_nmi()))
421 422 423
		return;

	if (unlikely(atomic_read(&current->tracing_graph_pause)))
424 425 426 427 428 429 430 431
		return;

	/*
	 * Protect against fault, even if it shouldn't
	 * happen. This tool is too much intrusive to
	 * ignore such a protection.
	 */
	asm volatile(
432 433
		"1: " _ASM_MOV " (%[parent]), %[old]\n"
		"2: " _ASM_MOV " %[return_hooker], (%[parent])\n"
434
		"   movl $0, %[faulted]\n"
435
		"3:\n"
436 437

		".section .fixup, \"ax\"\n"
438 439
		"4: movl $1, %[faulted]\n"
		"   jmp 3b\n"
440 441
		".previous\n"

442 443
		_ASM_EXTABLE(1b, 4b)
		_ASM_EXTABLE(2b, 4b)
444

445
		: [old] "=&r" (old), [faulted] "=r" (faulted)
446
		: [parent] "r" (parent), [return_hooker] "r" (return_hooker)
447 448 449
		: "memory"
	);

S
Steven Rostedt 已提交
450 451 452
	if (unlikely(faulted)) {
		ftrace_graph_stop();
		WARN_ON(1);
453 454 455
		return;
	}

456
	if (ftrace_push_return_trace(old, self_addr, &trace.depth) == -EBUSY) {
457
		*parent = old;
458 459 460 461 462
		return;
	}

	trace.func = self_addr;

463 464 465 466 467
	/* Only trace if the calling function expects to */
	if (!ftrace_graph_entry(&trace)) {
		current->curr_ret_stack--;
		*parent = old;
	}
468
}
469
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532

#ifdef CONFIG_FTRACE_SYSCALLS

extern unsigned long __start_syscalls_metadata[];
extern unsigned long __stop_syscalls_metadata[];
extern unsigned long *sys_call_table;

static struct syscall_metadata **syscalls_metadata;

static struct syscall_metadata *find_syscall_meta(unsigned long *syscall)
{
	struct syscall_metadata *start;
	struct syscall_metadata *stop;
	char str[KSYM_SYMBOL_LEN];


	start = (struct syscall_metadata *)__start_syscalls_metadata;
	stop = (struct syscall_metadata *)__stop_syscalls_metadata;
	kallsyms_lookup((unsigned long) syscall, NULL, NULL, NULL, str);

	for ( ; start < stop; start++) {
		if (start->name && !strcmp(start->name, str))
			return start;
	}
	return NULL;
}

struct syscall_metadata *syscall_nr_to_meta(int nr)
{
	if (!syscalls_metadata || nr >= FTRACE_SYSCALL_MAX || nr < 0)
		return NULL;

	return syscalls_metadata[nr];
}

void arch_init_ftrace_syscalls(void)
{
	int i;
	struct syscall_metadata *meta;
	unsigned long **psys_syscall_table = &sys_call_table;
	static atomic_t refs;

	if (atomic_inc_return(&refs) != 1)
		goto end;

	syscalls_metadata = kzalloc(sizeof(*syscalls_metadata) *
					FTRACE_SYSCALL_MAX, GFP_KERNEL);
	if (!syscalls_metadata) {
		WARN_ON(1);
		return;
	}

	for (i = 0; i < FTRACE_SYSCALL_MAX; i++) {
		meta = find_syscall_meta(psys_syscall_table[i]);
		syscalls_metadata[i] = meta;
	}
	return;

	/* Paranoid: avoid overflow */
end:
	atomic_dec(&refs);
}
#endif