ftrace.c 10.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
/*
 * 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.
 */

12 13
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

14 15
#include <linux/spinlock.h>
#include <linux/hardirq.h>
16
#include <linux/uaccess.h>
17 18
#include <linux/ftrace.h>
#include <linux/percpu.h>
19
#include <linux/sched.h>
20 21
#include <linux/init.h>
#include <linux/list.h>
22
#include <linux/module.h>
23

24 25
#include <trace/syscall.h>

26
#include <asm/cacheflush.h>
27
#include <asm/ftrace.h>
28
#include <asm/nops.h>
29
#include <asm/nmi.h>
30 31


32
#ifdef CONFIG_DYNAMIC_FTRACE
33

34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
/*
 * modifying_code is set to notify NMIs that they need to use
 * memory barriers when entering or exiting. But we don't want
 * to burden NMIs with unnecessary memory barriers when code
 * modification is not being done (which is most of the time).
 *
 * A mutex is already held when ftrace_arch_code_modify_prepare
 * and post_process are called. No locks need to be taken here.
 *
 * Stop machine will make sure currently running NMIs are done
 * and new NMIs will see the updated variable before we need
 * to worry about NMIs doing memory barriers.
 */
static int modifying_code __read_mostly;
static DEFINE_PER_CPU(int, save_modifying_code);

50 51 52
int ftrace_arch_code_modify_prepare(void)
{
	set_kernel_text_rw();
53
	set_all_modules_text_rw();
54
	modifying_code = 1;
55 56 57 58 59
	return 0;
}

int ftrace_arch_code_modify_post_process(void)
{
60
	modifying_code = 0;
61
	set_all_modules_text_ro();
62 63 64 65
	set_kernel_text_ro();
	return 0;
}

66
union ftrace_code_union {
67
	char code[MCOUNT_INSN_SIZE];
68 69 70 71 72 73
	struct {
		char e8;
		int offset;
	} __attribute__((packed));
};

74
static int ftrace_calc_offset(long ip, long addr)
75 76 77
{
	return (int)(addr - ip);
}
78

79
static unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
80 81
{
	static union ftrace_code_union calc;
82

83
	calc.e8		= 0xe8;
84
	calc.offset	= ftrace_calc_offset(ip + MCOUNT_INSN_SIZE, addr);
85 86 87 88 89 90

	/*
	 * No locking needed, this must be called via kstop_machine
	 * which in essence is like running on a uniprocessor machine.
	 */
	return calc.code;
91 92
}

93 94 95 96 97 98 99 100 101 102 103
/*
 * 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
104
 * 1) Put the instruction pointer into the IP buffer
105
 *    and the new code into the "code" buffer.
L
Lai Jiangshan 已提交
106 107 108 109 110
 * 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.
111 112 113 114 115 116 117 118 119 120 121
 *
 * 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
122
#define MOD_CODE_WRITE_FLAG (1 << 31)	/* set when NMI should do the write */
123
static atomic_t nmi_running = ATOMIC_INIT(0);
S
Steven Rostedt 已提交
124 125 126
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 */
127

S
Steven Rostedt 已提交
128 129
static unsigned nmi_wait_count;
static atomic_t nmi_update_count = ATOMIC_INIT(0);
S
Steven Rostedt 已提交
130 131 132 133 134 135 136 137 138 139 140

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
141 142 143 144 145 146 147 148 149 150 151 152 153 154
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);
	}
}

155 156 157 158 159 160 161 162 163 164
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);
165 166 167

	/* if we fail, then kill any new writers */
	if (mod_code_status)
L
Lai Jiangshan 已提交
168
		clear_mod_flag();
169 170
}

171
void ftrace_nmi_enter(void)
172
{
T
Tejun Heo 已提交
173
	__this_cpu_write(save_modifying_code, modifying_code);
174

T
Tejun Heo 已提交
175
	if (!__this_cpu_read(save_modifying_code))
176 177
		return;

L
Lai Jiangshan 已提交
178 179
	if (atomic_inc_return(&nmi_running) & MOD_CODE_WRITE_FLAG) {
		smp_rmb();
180
		ftrace_mod_code();
S
Steven Rostedt 已提交
181 182
		atomic_inc(&nmi_update_count);
	}
L
Lai Jiangshan 已提交
183 184
	/* Must have previous changes seen before executions */
	smp_mb();
185 186
}

187
void ftrace_nmi_exit(void)
188
{
T
Tejun Heo 已提交
189
	if (!__this_cpu_read(save_modifying_code))
190 191
		return;

192
	/* Finish all executions before clearing nmi_running */
L
Lai Jiangshan 已提交
193
	smp_mb();
194
	atomic_dec(&nmi_running);
195 196
}

L
Lai Jiangshan 已提交
197 198 199 200 201 202 203 204 205 206 207 208
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++;
}

209 210
static void wait_for_nmi(void)
{
211
	if (!atomic_read(&nmi_running))
212
		return;
S
Steven Rostedt 已提交
213

214
	do {
215
		cpu_relax();
216
	} while (atomic_read(&nmi_running));
S
Steven Rostedt 已提交
217

218
	nmi_wait_count++;
219 220
}

221 222 223 224 225 226
static inline int
within(unsigned long addr, unsigned long start, unsigned long end)
{
	return addr >= start && addr < end;
}

227 228 229
static int
do_ftrace_mod_code(unsigned long ip, void *new_code)
{
230 231 232 233 234 235 236 237 238 239 240
	/*
	 * On x86_64, kernel text mappings are mapped read-only with
	 * CONFIG_DEBUG_RODATA. So we use the kernel identity mapping instead
	 * of the kernel text mapping to modify the kernel text.
	 *
	 * For 32bit kernels, these mappings are same and we can use
	 * kernel identity mapping to modify code.
	 */
	if (within(ip, (unsigned long)_text, (unsigned long)_etext))
		ip = (unsigned long)__va(__pa(ip));

241 242 243 244 245 246
	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
247
	wait_for_nmi_and_set_mod_flag();
248 249 250 251 252 253 254 255 256

	/* 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
257
	clear_mod_flag();
258 259 260 261 262
	wait_for_nmi();

	return mod_code_status;
}

263
static unsigned char *ftrace_nop_replace(void)
264
{
265
	return ideal_nop5;
266 267
}

268
static int
269 270 271
ftrace_modify_code(unsigned long ip, unsigned char *old_code,
		   unsigned char *new_code)
{
272
	unsigned char replaced[MCOUNT_INSN_SIZE];
273 274 275 276

	/*
	 * Note: Due to modules and __init, code can
	 *  disappear and change, we need to protect against faulting
S
Steven Rostedt 已提交
277
	 *  as well as code changing. We do this by using the
S
Steven Rostedt 已提交
278
	 *  probe_kernel_* functions.
279 280
	 *
	 * No real locking needed, this code is run through
281
	 * kstop_machine, or before SMP starts.
282
	 */
S
Steven Rostedt 已提交
283 284

	/* read the text we want to modify */
S
Steven Rostedt 已提交
285
	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
286
		return -EFAULT;
287

S
Steven Rostedt 已提交
288
	/* Make sure it is what we expect it to be */
289
	if (memcmp(replaced, old_code, MCOUNT_INSN_SIZE) != 0)
290
		return -EINVAL;
291

S
Steven Rostedt 已提交
292
	/* replace the text with the new text */
293
	if (do_ftrace_mod_code(ip, new_code))
294
		return -EPERM;
295 296

	sync_core();
297

298
	return 0;
299 300
}

301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323
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);
}

324
int ftrace_update_ftrace_func(ftrace_func_t func)
325 326
{
	unsigned long ip = (unsigned long)(&ftrace_call);
327
	unsigned char old[MCOUNT_INSN_SIZE], *new;
328 329
	int ret;

330
	memcpy(old, &ftrace_call, MCOUNT_INSN_SIZE);
331 332 333 334 335 336 337
	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)
338
{
339 340
	/* The return code is retured via data */
	*(unsigned long *)data = 0;
S
Steven Rostedt 已提交
341

342 343
	return 0;
}
344
#endif
345

346
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
347

348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391
#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);
}

392 393 394 395 396 397
#endif /* !CONFIG_DYNAMIC_FTRACE */

/*
 * Hook the return address and push it in the stack of return addrs
 * in current thread info.
 */
398 399
void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr,
			   unsigned long frame_pointer)
400 401 402
{
	unsigned long old;
	int faulted;
403
	struct ftrace_graph_ent trace;
404 405 406
	unsigned long return_hooker = (unsigned long)
				&return_to_handler;

407
	if (unlikely(atomic_read(&current->tracing_graph_pause)))
408 409 410 411 412 413 414 415
		return;

	/*
	 * Protect against fault, even if it shouldn't
	 * happen. This tool is too much intrusive to
	 * ignore such a protection.
	 */
	asm volatile(
416 417
		"1: " _ASM_MOV " (%[parent]), %[old]\n"
		"2: " _ASM_MOV " %[return_hooker], (%[parent])\n"
418
		"   movl $0, %[faulted]\n"
419
		"3:\n"
420 421

		".section .fixup, \"ax\"\n"
422 423
		"4: movl $1, %[faulted]\n"
		"   jmp 3b\n"
424 425
		".previous\n"

426 427
		_ASM_EXTABLE(1b, 4b)
		_ASM_EXTABLE(2b, 4b)
428

429
		: [old] "=&r" (old), [faulted] "=r" (faulted)
430
		: [parent] "r" (parent), [return_hooker] "r" (return_hooker)
431 432 433
		: "memory"
	);

S
Steven Rostedt 已提交
434 435 436
	if (unlikely(faulted)) {
		ftrace_graph_stop();
		WARN_ON(1);
437 438 439
		return;
	}

440 441
	if (ftrace_push_return_trace(old, self_addr, &trace.depth,
		    frame_pointer) == -EBUSY) {
442
		*parent = old;
443 444 445 446 447
		return;
	}

	trace.func = self_addr;

448 449 450 451 452
	/* Only trace if the calling function expects to */
	if (!ftrace_graph_entry(&trace)) {
		current->curr_ret_stack--;
		*parent = old;
	}
453
}
454
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */