[deliverable/linux.git] / arch / x86 / kernel / ftrace.c

/*
 * 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>
#include <linux/uaccess.h>
#include <linux/ftrace.h>
#include <linux/percpu.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/list.h>

#include <asm/ftrace.h>
#include <linux/ftrace.h>
#include <asm/nops.h>
#include <asm/nmi.h>


#ifdef CONFIG_DYNAMIC_FTRACE

union ftrace_code_union {
	char code[MCOUNT_INSN_SIZE];
	struct {
		char e8;
		int offset;
	} __attribute__((packed));
};

static int ftrace_calc_offset(long ip, long addr)
{
	return (int)(addr - ip);
}

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

	calc.e8		= 0xe8;
	calc.offset	= ftrace_calc_offset(ip + MCOUNT_INSN_SIZE, addr);

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

/*
 * 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.
 *
 * 1) Put the instruction pointer into the IP buffer
 *    and the new code into the "code" buffer.
 * 2) Set a flag that says we are modifying code
 * 3) Wait for any running NMIs to finish.
 * 4) Write the code
 * 5) clear the flag.
 * 6) Wait for any running NMIs to finish.
 *
 * 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.
 */

static atomic_t in_nmi = ATOMIC_INIT(0);
static int mod_code_status;		/* holds return value of text write */
static int mod_code_write;		/* set when NMI should do the 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 */

static unsigned nmi_wait_count;
static atomic_t nmi_update_count = ATOMIC_INIT(0);

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;
}

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);

}

void ftrace_nmi_enter(void)
{
	atomic_inc(&in_nmi);
	/* Must have in_nmi seen before reading write flag */
	smp_mb();
	if (mod_code_write) {
		ftrace_mod_code();
		atomic_inc(&nmi_update_count);
	}
}

void ftrace_nmi_exit(void)
{
	/* Finish all executions before clearing in_nmi */
	smp_wmb();
	atomic_dec(&in_nmi);
}

static void wait_for_nmi(void)
{
	int waited = 0;

	while (atomic_read(&in_nmi)) {
		waited = 1;
		cpu_relax();
	}

	if (waited)
		nmi_wait_count++;
}

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_wmb();

	mod_code_write = 1;

	/* Make sure write bit is visible before we wait on NMIs */
	smp_mb();

	wait_for_nmi();

	/* 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_wmb();

	mod_code_write = 0;

	/* make sure NMIs see the cleared bit */
	smp_mb();

	wait_for_nmi();

	return mod_code_status;
}


static unsigned char ftrace_nop[MCOUNT_INSN_SIZE];

static unsigned char *ftrace_nop_replace(void)
{
	return ftrace_nop;
}

static int
ftrace_modify_code(unsigned long ip, unsigned char *old_code,
		   unsigned char *new_code)
{
	unsigned char replaced[MCOUNT_INSN_SIZE];

	/*
	 * Note: Due to modules and __init, code can
	 *  disappear and change, we need to protect against faulting
	 *  as well as code changing. We do this by using the
	 *  probe_kernel_* functions.
	 *
	 * No real locking needed, this code is run through
	 * kstop_machine, or before SMP starts.
	 */

	/* read the text we want to modify */
	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
		return -EFAULT;

	/* Make sure it is what we expect it to be */
	if (memcmp(replaced, old_code, MCOUNT_INSN_SIZE) != 0)
		return -EINVAL;

	/* replace the text with the new text */
	if (do_ftrace_mod_code(ip, new_code))
		return -EPERM;

	sync_core();

	return 0;
}

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);
}

int ftrace_update_ftrace_func(ftrace_func_t func)
{
	unsigned long ip = (unsigned long)(&ftrace_call);
	unsigned char old[MCOUNT_INSN_SIZE], *new;
	int ret;

	memcpy(old, &ftrace_call, MCOUNT_INSN_SIZE);
	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)
{
	extern const unsigned char ftrace_test_p6nop[];
	extern const unsigned char ftrace_test_nop5[];
	extern const unsigned char ftrace_test_jmp[];
	int faulted = 0;

	/*
	 * 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"
		"nop\n"
		"nop\n"
		"nop\n"  /* 2 byte jmp + 3 bytes */
		"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");
		memcpy(ftrace_nop, ftrace_test_p6nop, MCOUNT_INSN_SIZE);
		break;
	case 1:
		pr_info("ftrace: converting mcount calls to 66 66 66 66 90\n");
		memcpy(ftrace_nop, ftrace_test_nop5, MCOUNT_INSN_SIZE);
		break;
	case 2:
		pr_info("ftrace: converting mcount calls to jmp . + 5\n");
		memcpy(ftrace_nop, ftrace_test_jmp, MCOUNT_INSN_SIZE);
		break;
	}

	/* The return code is retured via data */
	*(unsigned long *)data = 0;

	return 0;
}
#endif

#ifdef CONFIG_FUNCTION_RET_TRACER

#ifndef CONFIG_DYNAMIC_FTRACE

/*
 * These functions are picked from those used on
 * this page for dynamic ftrace. They have been
 * simplified to ignore all traces in NMI context.
 */
static atomic_t in_nmi;

void ftrace_nmi_enter(void)
{
	atomic_inc(&in_nmi);
}

void ftrace_nmi_exit(void)
{
	atomic_dec(&in_nmi);
}
#endif /* !CONFIG_DYNAMIC_FTRACE */

/* Add a function return address to the trace stack on thread info.*/
static int push_return_trace(unsigned long ret, unsigned long long time,
				unsigned long func)
{
	int index;

	if (!current->ret_stack)
		return -EBUSY;

	/* The return trace stack is full */
	if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
		atomic_inc(&current->trace_overrun);
		return -EBUSY;
	}

	index = ++current->curr_ret_stack;
	barrier();
	current->ret_stack[index].ret = ret;
	current->ret_stack[index].func = func;
	current->ret_stack[index].calltime = time;

	return 0;
}

/* Retrieve a function return address to the trace stack on thread info.*/
static void pop_return_trace(unsigned long *ret, unsigned long long *time,
				unsigned long *func, unsigned long *overrun)
{
	int index;

	index = current->curr_ret_stack;
	*ret = current->ret_stack[index].ret;
	*func = current->ret_stack[index].func;
	*time = current->ret_stack[index].calltime;
	*overrun = atomic_read(&current->trace_overrun);
	current->curr_ret_stack--;
}

/*
 * Send the trace to the ring-buffer.
 * @return the original return address.
 */
unsigned long ftrace_return_to_handler(void)
{
	struct ftrace_retfunc trace;
	pop_return_trace(&trace.ret, &trace.calltime, &trace.func,
			&trace.overrun);
	trace.rettime = cpu_clock(raw_smp_processor_id());
	ftrace_function_return(&trace);

	return trace.ret;
}

/*
 * 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;
	unsigned long long calltime;
	int faulted;
	unsigned long return_hooker = (unsigned long)
				&return_to_handler;

	/* Nmi's are currently unsupported */
	if (atomic_read(&in_nmi))
		return;

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

		".section .fixup, \"ax\"\n"
		"3: movl $1, %[faulted]\n"
		".previous\n"

		".section __ex_table, \"a\"\n"
		"   .long 1b, 3b\n"
		"   .long 2b, 3b\n"
		".previous\n"

		: [parent_replaced] "=r" (parent), [old] "=r" (old),
		  [faulted] "=r" (faulted)
		: [parent_old] "0" (parent), [return_hooker] "r" (return_hooker)
		: "memory"
	);

	if (WARN_ON(faulted)) {
		unregister_ftrace_return();
		return;
	}

	if (WARN_ON(!__kernel_text_address(old))) {
		unregister_ftrace_return();
		*parent = old;
		return;
	}

	calltime = cpu_clock(raw_smp_processor_id());

	if (push_return_trace(old, calltime, self_addr) == -EBUSY)
		*parent = old;
}

#endif /* CONFIG_FUNCTION_RET_TRACER */
Commit	Line	Data
3d083395 SR	1	/*
	2	* Code for replacing ftrace calls with jumps.
	3	*
	4	* Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
	5	*
	6	* Thanks goes to Ingo Molnar, for suggesting the idea.
	7	* Mathieu Desnoyers, for suggesting postponing the modifications.
	8	* Arjan van de Ven, for keeping me straight, and explaining to me
	9	* the dangers of modifying code on the run.
	10	*/
	11
	12	#include <linux/spinlock.h>
	13	#include <linux/hardirq.h>
6f93fc07	14	#include <linux/uaccess.h>
3d083395 SR	15	#include <linux/ftrace.h>
3d083395 SR	16	#include <linux/percpu.h>
19b3e967	17	#include <linux/sched.h>
3d083395 SR	18	#include <linux/init.h>
	19	#include <linux/list.h>
	20
395a59d0	21	#include <asm/ftrace.h>
caf4b323	22	#include <linux/ftrace.h>
732f3ca7	23	#include <asm/nops.h>
caf4b323	24	#include <asm/nmi.h>
3d083395	25
3d083395	26
caf4b323	27	#ifdef CONFIG_DYNAMIC_FTRACE
3d083395	28
3d083395	29	union ftrace_code_union {
395a59d0	30	char code[MCOUNT_INSN_SIZE];
3d083395 SR	31	struct {
	32	char e8;
	33	int offset;
	34	} __attribute__((packed));
	35	};
	36
15adc048	37	static int ftrace_calc_offset(long ip, long addr)
3c1720f0 SR	38	{
	39	return (int)(addr - ip);
	40	}
3d083395	41
31e88909	42	static unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
3c1720f0 SR	43	{
3c1720f0 SR	44	static union ftrace_code_union calc;
3d083395	45
3c1720f0	46	calc.e8 = 0xe8;
395a59d0	47	calc.offset = ftrace_calc_offset(ip + MCOUNT_INSN_SIZE, addr);
3c1720f0 SR	48
	49	/*
	50	* No locking needed, this must be called via kstop_machine
	51	* which in essence is like running on a uniprocessor machine.
	52	*/
	53	return calc.code;
3d083395 SR	54	}
3d083395 SR	55
17666f02 SR	56	/*
	57	* Modifying code must take extra care. On an SMP machine, if
	58	* the code being modified is also being executed on another CPU
	59	* that CPU will have undefined results and possibly take a GPF.
	60	* We use kstop_machine to stop other CPUS from exectuing code.
	61	* But this does not stop NMIs from happening. We still need
	62	* to protect against that. We separate out the modification of
	63	* the code to take care of this.
	64	*
	65	* Two buffers are added: An IP buffer and a "code" buffer.
	66	*
a26a2a27	67	* 1) Put the instruction pointer into the IP buffer
17666f02 SR	68	* and the new code into the "code" buffer.
	69	* 2) Set a flag that says we are modifying code
	70	* 3) Wait for any running NMIs to finish.
	71	* 4) Write the code
	72	* 5) clear the flag.
	73	* 6) Wait for any running NMIs to finish.
	74	*
	75	* If an NMI is executed, the first thing it does is to call
	76	* "ftrace_nmi_enter". This will check if the flag is set to write
	77	* and if it is, it will write what is in the IP and "code" buffers.
	78	*
	79	* The trick is, it does not matter if everyone is writing the same
	80	* content to the code location. Also, if a CPU is executing code
	81	* it is OK to write to that code location if the contents being written
	82	* are the same as what exists.
	83	*/
	84
a26a2a27 SR	85	static atomic_t in_nmi = ATOMIC_INIT(0);
	86	static int mod_code_status; /* holds return value of text write */
	87	static int mod_code_write; /* set when NMI should do the write */
	88	static void mod_code_ip; / holds the IP to write to */
	89	static void mod_code_newcode; / holds the text to write to the IP */
17666f02	90
a26a2a27 SR	91	static unsigned nmi_wait_count;
a26a2a27 SR	92	static atomic_t nmi_update_count = ATOMIC_INIT(0);
b807c3d0 SR	93
	94	int ftrace_arch_read_dyn_info(char *buf, int size)
	95	{
	96	int r;
	97
	98	r = snprintf(buf, size, "%u %u",
	99	nmi_wait_count,
	100	atomic_read(&nmi_update_count));
	101	return r;
	102	}
	103
17666f02 SR	104	static void ftrace_mod_code(void)
	105	{
	106	/*
	107	* Yes, more than one CPU process can be writing to mod_code_status.
	108	* (and the code itself)
	109	* But if one were to fail, then they all should, and if one were
	110	* to succeed, then they all should.
	111	*/
	112	mod_code_status = probe_kernel_write(mod_code_ip, mod_code_newcode,
	113	MCOUNT_INSN_SIZE);
	114
	115	}
	116
	117	void ftrace_nmi_enter(void)
	118	{
	119	atomic_inc(&in_nmi);
	120	/* Must have in_nmi seen before reading write flag */
	121	smp_mb();
b807c3d0	122	if (mod_code_write) {
17666f02	123	ftrace_mod_code();
b807c3d0 SR	124	atomic_inc(&nmi_update_count);
b807c3d0 SR	125	}
17666f02 SR	126	}
	127
	128	void ftrace_nmi_exit(void)
	129	{
	130	/* Finish all executions before clearing in_nmi */
	131	smp_wmb();
	132	atomic_dec(&in_nmi);
	133	}
	134
	135	static void wait_for_nmi(void)
	136	{
b807c3d0 SR	137	int waited = 0;
	138
	139	while (atomic_read(&in_nmi)) {
	140	waited = 1;
17666f02	141	cpu_relax();
b807c3d0 SR	142	}
	143
	144	if (waited)
	145	nmi_wait_count++;
17666f02 SR	146	}
	147
	148	static int
	149	do_ftrace_mod_code(unsigned long ip, void *new_code)
	150	{
	151	mod_code_ip = (void *)ip;
	152	mod_code_newcode = new_code;
	153
	154	/* The buffers need to be visible before we let NMIs write them */
	155	smp_wmb();
	156
	157	mod_code_write = 1;
	158
	159	/* Make sure write bit is visible before we wait on NMIs */
	160	smp_mb();
	161
	162	wait_for_nmi();
	163
	164	/* Make sure all running NMIs have finished before we write the code */
	165	smp_mb();
	166
	167	ftrace_mod_code();
	168
	169	/* Make sure the write happens before clearing the bit */
	170	smp_wmb();
	171
	172	mod_code_write = 0;
	173
	174	/* make sure NMIs see the cleared bit */
	175	smp_mb();
	176
	177	wait_for_nmi();
	178
	179	return mod_code_status;
	180	}
	181
	182
caf4b323 FW	183
	184
	185	static unsigned char ftrace_nop[MCOUNT_INSN_SIZE];
	186
31e88909	187	static unsigned char *ftrace_nop_replace(void)
caf4b323 FW	188	{
	189	return ftrace_nop;
	190	}
	191
31e88909	192	static int
3d083395 SR	193	ftrace_modify_code(unsigned long ip, unsigned char *old_code,
	194	unsigned char *new_code)
	195	{
6f93fc07	196	unsigned char replaced[MCOUNT_INSN_SIZE];
3d083395 SR	197
	198	/*
	199	* Note: Due to modules and __init, code can
	200	* disappear and change, we need to protect against faulting
76aefee5	201	* as well as code changing. We do this by using the
ab9a0918	202	* probe_kernel_* functions.
3d083395 SR	203	*
3d083395 SR	204	* No real locking needed, this code is run through
6f93fc07	205	* kstop_machine, or before SMP starts.
3d083395	206	*/
76aefee5 SR	207
76aefee5 SR	208	/* read the text we want to modify */
ab9a0918	209	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
593eb8a2	210	return -EFAULT;
6f93fc07	211
76aefee5	212	/* Make sure it is what we expect it to be */
6f93fc07	213	if (memcmp(replaced, old_code, MCOUNT_INSN_SIZE) != 0)
593eb8a2	214	return -EINVAL;
3d083395	215
76aefee5	216	/* replace the text with the new text */
17666f02	217	if (do_ftrace_mod_code(ip, new_code))
593eb8a2	218	return -EPERM;
6f93fc07 SR	219
6f93fc07 SR	220	sync_core();
3d083395	221
6f93fc07	222	return 0;
3d083395 SR	223	}
3d083395 SR	224
31e88909 SR	225	int ftrace_make_nop(struct module *mod,
	226	struct dyn_ftrace *rec, unsigned long addr)
	227	{
	228	unsigned char new, old;
	229	unsigned long ip = rec->ip;
	230
	231	old = ftrace_call_replace(ip, addr);
	232	new = ftrace_nop_replace();
	233
	234	return ftrace_modify_code(rec->ip, old, new);
	235	}
	236
	237	int ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
	238	{
	239	unsigned char new, old;
	240	unsigned long ip = rec->ip;
	241
	242	old = ftrace_nop_replace();
	243	new = ftrace_call_replace(ip, addr);
	244
	245	return ftrace_modify_code(rec->ip, old, new);
	246	}
	247
15adc048	248	int ftrace_update_ftrace_func(ftrace_func_t func)
d61f82d0 SR	249	{
d61f82d0 SR	250	unsigned long ip = (unsigned long)(&ftrace_call);
395a59d0	251	unsigned char old[MCOUNT_INSN_SIZE], *new;
d61f82d0 SR	252	int ret;
d61f82d0 SR	253
395a59d0	254	memcpy(old, &ftrace_call, MCOUNT_INSN_SIZE);
d61f82d0 SR	255	new = ftrace_call_replace(ip, (unsigned long)func);
	256	ret = ftrace_modify_code(ip, old, new);
	257
	258	return ret;
	259	}
	260
d61f82d0	261	int __init ftrace_dyn_arch_init(void *data)
3d083395	262	{
732f3ca7 SR	263	extern const unsigned char ftrace_test_p6nop[];
	264	extern const unsigned char ftrace_test_nop5[];
	265	extern const unsigned char ftrace_test_jmp[];
	266	int faulted = 0;
d61f82d0	267
732f3ca7 SR	268	/*
	269	* There is no good nop for all x86 archs.
	270	* We will default to using the P6_NOP5, but first we
	271	* will test to make sure that the nop will actually
	272	* work on this CPU. If it faults, we will then
	273	* go to a lesser efficient 5 byte nop. If that fails
	274	* we then just use a jmp as our nop. This isn't the most
	275	* efficient nop, but we can not use a multi part nop
	276	* since we would then risk being preempted in the middle
	277	* of that nop, and if we enabled tracing then, it might
	278	* cause a system crash.
	279	*
	280	* TODO: check the cpuid to determine the best nop.
	281	*/
	282	asm volatile (
732f3ca7 SR	283	"ftrace_test_jmp:"
732f3ca7 SR	284	"jmp ftrace_test_p6nop\n"
8b27386a AK	285	"nop\n"
	286	"nop\n"
	287	"nop\n" /* 2 byte jmp + 3 bytes */
732f3ca7 SR	288	"ftrace_test_p6nop:"
	289	P6_NOP5
	290	"jmp 1f\n"
	291	"ftrace_test_nop5:"
	292	".byte 0x66,0x66,0x66,0x66,0x90\n"
732f3ca7 SR	293	"1:"
	294	".section .fixup, \"ax\"\n"
	295	"2: movl $1, %0\n"
	296	" jmp ftrace_test_nop5\n"
	297	"3: movl $2, %0\n"
	298	" jmp 1b\n"
	299	".previous\n"
	300	_ASM_EXTABLE(ftrace_test_p6nop, 2b)
	301	_ASM_EXTABLE(ftrace_test_nop5, 3b)
	302	: "=r"(faulted) : "0" (faulted));
	303
	304	switch (faulted) {
	305	case 0:
	306	pr_info("ftrace: converting mcount calls to 0f 1f 44 00 00\n");
8115f3f0	307	memcpy(ftrace_nop, ftrace_test_p6nop, MCOUNT_INSN_SIZE);
732f3ca7 SR	308	break;
	309	case 1:
	310	pr_info("ftrace: converting mcount calls to 66 66 66 66 90\n");
8115f3f0	311	memcpy(ftrace_nop, ftrace_test_nop5, MCOUNT_INSN_SIZE);
732f3ca7 SR	312	break;
732f3ca7 SR	313	case 2:
8b27386a	314	pr_info("ftrace: converting mcount calls to jmp . + 5\n");
8115f3f0	315	memcpy(ftrace_nop, ftrace_test_jmp, MCOUNT_INSN_SIZE);
732f3ca7 SR	316	break;
	317	}
	318
	319	/* The return code is retured via data */
	320	(unsigned long )data = 0;
dfa60aba	321
3d083395 SR	322	return 0;
3d083395 SR	323	}
caf4b323	324	#endif
e7d3737e FW	325
	326	#ifdef CONFIG_FUNCTION_RET_TRACER
	327
	328	#ifndef CONFIG_DYNAMIC_FTRACE
	329
	330	/*
	331	* These functions are picked from those used on
	332	* this page for dynamic ftrace. They have been
	333	* simplified to ignore all traces in NMI context.
	334	*/
	335	static atomic_t in_nmi;
	336
	337	void ftrace_nmi_enter(void)
	338	{
	339	atomic_inc(&in_nmi);
	340	}
	341
	342	void ftrace_nmi_exit(void)
	343	{
	344	atomic_dec(&in_nmi);
	345	}
	346	#endif /* !CONFIG_DYNAMIC_FTRACE */
	347
	348	/* Add a function return address to the trace stack on thread info.*/
	349	static int push_return_trace(unsigned long ret, unsigned long long time,
	350	unsigned long func)
	351	{
	352	int index;
f201ae23 FW	353
	354	if (!current->ret_stack)
	355	return -EBUSY;
e7d3737e FW	356
e7d3737e FW	357	/* The return trace stack is full */
f201ae23 FW	358	if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
f201ae23 FW	359	atomic_inc(&current->trace_overrun);
e7d3737e	360	return -EBUSY;
0231022c	361	}
e7d3737e	362
f201ae23	363	index = ++current->curr_ret_stack;
e7d3737e	364	barrier();
f201ae23 FW	365	current->ret_stack[index].ret = ret;
	366	current->ret_stack[index].func = func;
	367	current->ret_stack[index].calltime = time;
e7d3737e FW	368
	369	return 0;
	370	}
	371
	372	/* Retrieve a function return address to the trace stack on thread info.*/
	373	static void pop_return_trace(unsigned long ret, unsigned long long time,
0231022c	374	unsigned long func, unsigned long overrun)
e7d3737e FW	375	{
	376	int index;
	377
f201ae23 FW	378	index = current->curr_ret_stack;
	379	*ret = current->ret_stack[index].ret;
	380	*func = current->ret_stack[index].func;
	381	*time = current->ret_stack[index].calltime;
	382	*overrun = atomic_read(&current->trace_overrun);
	383	current->curr_ret_stack--;
e7d3737e FW	384	}
	385
	386	/*
	387	* Send the trace to the ring-buffer.
	388	* @return the original return address.
	389	*/
	390	unsigned long ftrace_return_to_handler(void)
	391	{
	392	struct ftrace_retfunc trace;
0231022c FW	393	pop_return_trace(&trace.ret, &trace.calltime, &trace.func,
0231022c FW	394	&trace.overrun);
e7d3737e FW	395	trace.rettime = cpu_clock(raw_smp_processor_id());
	396	ftrace_function_return(&trace);
	397
	398	return trace.ret;
	399	}
	400
	401	/*
	402	* Hook the return address and push it in the stack of return addrs
	403	* in current thread info.
	404	*/
	405	void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
	406	{
	407	unsigned long old;
	408	unsigned long long calltime;
	409	int faulted;
	410	unsigned long return_hooker = (unsigned long)
	411	&return_to_handler;
	412
	413	/* Nmi's are currently unsupported */
	414	if (atomic_read(&in_nmi))
	415	return;
	416
	417	/*
	418	* Protect against fault, even if it shouldn't
	419	* happen. This tool is too much intrusive to
	420	* ignore such a protection.
	421	*/
	422	asm volatile(
	423	"1: movl (%[parent_old]), %[old]\n"
	424	"2: movl %[return_hooker], (%[parent_replaced])\n"
	425	" movl $0, %[faulted]\n"
	426
	427	".section .fixup, \"ax\"\n"
	428	"3: movl $1, %[faulted]\n"
	429	".previous\n"
	430
	431	".section __ex_table, \"a\"\n"
	432	" .long 1b, 3b\n"
	433	" .long 2b, 3b\n"
	434	".previous\n"
	435
	436	: [parent_replaced] "=r" (parent), [old] "=r" (old),
	437	[faulted] "=r" (faulted)
	438	: [parent_old] "0" (parent), [return_hooker] "r" (return_hooker)
	439	: "memory"
	440	);
	441
	442	if (WARN_ON(faulted)) {
	443	unregister_ftrace_return();
	444	return;
	445	}
	446
	447	if (WARN_ON(!__kernel_text_address(old))) {
	448	unregister_ftrace_return();
	449	*parent = old;
	450	return;
	451	}
	452
	453	calltime = cpu_clock(raw_smp_processor_id());
	454
	455	if (push_return_trace(old, calltime, self_addr) == -EBUSY)
	456	*parent = old;
	457	}
	458
459	#endif /* CONFIG_FUNCTION_RET_TRACER */