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

#include <linux/module.h>
#include <linux/sched.h>
#include <linux/spinlock.h>
#include <linux/list.h>
#include <linux/kprobes.h>
#include <linux/mm.h>
#include <linux/vmalloc.h>
#include <asm/alternative.h>
#include <asm/sections.h>
#include <asm/pgtable.h>
#include <asm/mce.h>
#include <asm/nmi.h>
#include <asm/vsyscall.h>

#define MAX_PATCH_LEN (255-1)

#ifdef CONFIG_HOTPLUG_CPU
static int smp_alt_once;

static int __init bootonly(char *str)
{
	smp_alt_once = 1;
	return 1;
}
__setup("smp-alt-boot", bootonly);
#else
#define smp_alt_once 1
#endif

static int debug_alternative;

static int __init debug_alt(char *str)
{
	debug_alternative = 1;
	return 1;
}
__setup("debug-alternative", debug_alt);

static int noreplace_smp;

static int __init setup_noreplace_smp(char *str)
{
	noreplace_smp = 1;
	return 1;
}
__setup("noreplace-smp", setup_noreplace_smp);

#ifdef CONFIG_PARAVIRT
static int noreplace_paravirt = 0;

static int __init setup_noreplace_paravirt(char *str)
{
	noreplace_paravirt = 1;
	return 1;
}
__setup("noreplace-paravirt", setup_noreplace_paravirt);
#endif

#define DPRINTK(fmt, args...) if (debug_alternative) \
	printk(KERN_DEBUG fmt, args)

#ifdef GENERIC_NOP1
/* Use inline assembly to define this because the nops are defined
   as inline assembly strings in the include files and we cannot
   get them easily into strings. */
asm("\t.section .rodata, \"a\"\nintelnops: "
	GENERIC_NOP1 GENERIC_NOP2 GENERIC_NOP3 GENERIC_NOP4 GENERIC_NOP5 GENERIC_NOP6
	GENERIC_NOP7 GENERIC_NOP8);
extern const unsigned char intelnops[];
static const unsigned char *const intel_nops[ASM_NOP_MAX+1] = {
	NULL,
	intelnops,
	intelnops + 1,
	intelnops + 1 + 2,
	intelnops + 1 + 2 + 3,
	intelnops + 1 + 2 + 3 + 4,
	intelnops + 1 + 2 + 3 + 4 + 5,
	intelnops + 1 + 2 + 3 + 4 + 5 + 6,
	intelnops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
};
#endif

#ifdef K8_NOP1
asm("\t.section .rodata, \"a\"\nk8nops: "
	K8_NOP1 K8_NOP2 K8_NOP3 K8_NOP4 K8_NOP5 K8_NOP6
	K8_NOP7 K8_NOP8);
extern const unsigned char k8nops[];
static const unsigned char *const k8_nops[ASM_NOP_MAX+1] = {
	NULL,
	k8nops,
	k8nops + 1,
	k8nops + 1 + 2,
	k8nops + 1 + 2 + 3,
	k8nops + 1 + 2 + 3 + 4,
	k8nops + 1 + 2 + 3 + 4 + 5,
	k8nops + 1 + 2 + 3 + 4 + 5 + 6,
	k8nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
};
#endif

#ifdef K7_NOP1
asm("\t.section .rodata, \"a\"\nk7nops: "
	K7_NOP1 K7_NOP2 K7_NOP3 K7_NOP4 K7_NOP5 K7_NOP6
	K7_NOP7 K7_NOP8);
extern const unsigned char k7nops[];
static const unsigned char *const k7_nops[ASM_NOP_MAX+1] = {
	NULL,
	k7nops,
	k7nops + 1,
	k7nops + 1 + 2,
	k7nops + 1 + 2 + 3,
	k7nops + 1 + 2 + 3 + 4,
	k7nops + 1 + 2 + 3 + 4 + 5,
	k7nops + 1 + 2 + 3 + 4 + 5 + 6,
	k7nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
};
#endif

#ifdef P6_NOP1
asm("\t.section .rodata, \"a\"\np6nops: "
	P6_NOP1 P6_NOP2 P6_NOP3 P6_NOP4 P6_NOP5 P6_NOP6
	P6_NOP7 P6_NOP8);
extern const unsigned char p6nops[];
static const unsigned char *const p6_nops[ASM_NOP_MAX+1] = {
	NULL,
	p6nops,
	p6nops + 1,
	p6nops + 1 + 2,
	p6nops + 1 + 2 + 3,
	p6nops + 1 + 2 + 3 + 4,
	p6nops + 1 + 2 + 3 + 4 + 5,
	p6nops + 1 + 2 + 3 + 4 + 5 + 6,
	p6nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
};
#endif

#ifdef CONFIG_X86_64

extern char __vsyscall_0;
static inline const unsigned char*const * find_nop_table(void)
{
	return boot_cpu_data.x86_vendor != X86_VENDOR_INTEL ||
	       boot_cpu_data.x86 < 6 ? k8_nops : p6_nops;
}

#else /* CONFIG_X86_64 */

static const struct nop {
	int cpuid;
	const unsigned char *const *noptable;
} noptypes[] = {
	{ X86_FEATURE_K8, k8_nops },
	{ X86_FEATURE_K7, k7_nops },
	{ X86_FEATURE_P4, p6_nops },
	{ X86_FEATURE_P3, p6_nops },
	{ -1, NULL }
};

static const unsigned char*const * find_nop_table(void)
{
	const unsigned char *const *noptable = intel_nops;
	int i;

	for (i = 0; noptypes[i].cpuid >= 0; i++) {
		if (boot_cpu_has(noptypes[i].cpuid)) {
			noptable = noptypes[i].noptable;
			break;
		}
	}
	return noptable;
}

#endif /* CONFIG_X86_64 */

/* Use this to add nops to a buffer, then text_poke the whole buffer. */
static void add_nops(void *insns, unsigned int len)
{
	const unsigned char *const *noptable = find_nop_table();

	while (len > 0) {
		unsigned int noplen = len;
		if (noplen > ASM_NOP_MAX)
			noplen = ASM_NOP_MAX;
		memcpy(insns, noptable[noplen], noplen);
		insns += noplen;
		len -= noplen;
	}
}

extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
extern u8 *__smp_locks[], *__smp_locks_end[];

/* Replace instructions with better alternatives for this CPU type.
   This runs before SMP is initialized to avoid SMP problems with
   self modifying code. This implies that assymetric systems where
   APs have less capabilities than the boot processor are not handled.
   Tough. Make sure you disable such features by hand. */

void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
{
	struct alt_instr *a;
	char insnbuf[MAX_PATCH_LEN];

	DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end);
	for (a = start; a < end; a++) {
		u8 *instr = a->instr;
		BUG_ON(a->replacementlen > a->instrlen);
		BUG_ON(a->instrlen > sizeof(insnbuf));
		if (!boot_cpu_has(a->cpuid))
			continue;
#ifdef CONFIG_X86_64
		/* vsyscall code is not mapped yet. resolve it manually. */
		if (instr >= (u8 *)VSYSCALL_START && instr < (u8*)VSYSCALL_END) {
			instr = __va(instr - (u8*)VSYSCALL_START + (u8*)__pa_symbol(&__vsyscall_0));
			DPRINTK("%s: vsyscall fixup: %p => %p\n",
				__FUNCTION__, a->instr, instr);
		}
#endif
		memcpy(insnbuf, a->replacement, a->replacementlen);
		add_nops(insnbuf + a->replacementlen,
			 a->instrlen - a->replacementlen);
		text_poke(instr, insnbuf, a->instrlen);
	}
}

#ifdef CONFIG_SMP

static void alternatives_smp_lock(u8 **start, u8 **end, u8 *text, u8 *text_end)
{
	u8 **ptr;

	for (ptr = start; ptr < end; ptr++) {
		if (*ptr < text)
			continue;
		if (*ptr > text_end)
			continue;
		text_poke(*ptr, ((unsigned char []){0xf0}), 1); /* add lock prefix */
	};
}

static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end)
{
	u8 **ptr;
	char insn[1];

	if (noreplace_smp)
		return;

	add_nops(insn, 1);
	for (ptr = start; ptr < end; ptr++) {
		if (*ptr < text)
			continue;
		if (*ptr > text_end)
			continue;
		text_poke(*ptr, insn, 1);
	};
}

struct smp_alt_module {
	/* what is this ??? */
	struct module	*mod;
	char		*name;

	/* ptrs to lock prefixes */
	u8		**locks;
	u8		**locks_end;

	/* .text segment, needed to avoid patching init code ;) */
	u8		*text;
	u8		*text_end;

	struct list_head next;
};
static LIST_HEAD(smp_alt_modules);
static DEFINE_SPINLOCK(smp_alt);
static int smp_mode = 1;	/* protected by smp_alt */

void alternatives_smp_module_add(struct module *mod, char *name,
				 void *locks, void *locks_end,
				 void *text,  void *text_end)
{
	struct smp_alt_module *smp;
	unsigned long flags;

	if (noreplace_smp)
		return;

	if (smp_alt_once) {
		if (boot_cpu_has(X86_FEATURE_UP))
			alternatives_smp_unlock(locks, locks_end,
						text, text_end);
		return;
	}

	smp = kzalloc(sizeof(*smp), GFP_KERNEL);
	if (NULL == smp)
		return; /* we'll run the (safe but slow) SMP code then ... */

	smp->mod	= mod;
	smp->name	= name;
	smp->locks	= locks;
	smp->locks_end	= locks_end;
	smp->text	= text;
	smp->text_end	= text_end;
	DPRINTK("%s: locks %p -> %p, text %p -> %p, name %s\n",
		__FUNCTION__, smp->locks, smp->locks_end,
		smp->text, smp->text_end, smp->name);

	spin_lock_irqsave(&smp_alt, flags);
	list_add_tail(&smp->next, &smp_alt_modules);
	if (boot_cpu_has(X86_FEATURE_UP))
		alternatives_smp_unlock(smp->locks, smp->locks_end,
					smp->text, smp->text_end);
	spin_unlock_irqrestore(&smp_alt, flags);
}

void alternatives_smp_module_del(struct module *mod)
{
	struct smp_alt_module *item;
	unsigned long flags;

	if (smp_alt_once || noreplace_smp)
		return;

	spin_lock_irqsave(&smp_alt, flags);
	list_for_each_entry(item, &smp_alt_modules, next) {
		if (mod != item->mod)
			continue;
		list_del(&item->next);
		spin_unlock_irqrestore(&smp_alt, flags);
		DPRINTK("%s: %s\n", __FUNCTION__, item->name);
		kfree(item);
		return;
	}
	spin_unlock_irqrestore(&smp_alt, flags);
}

void alternatives_smp_switch(int smp)
{
	struct smp_alt_module *mod;
	unsigned long flags;

#ifdef CONFIG_LOCKDEP
	/*
	 * Older binutils section handling bug prevented
	 * alternatives-replacement from working reliably.
	 *
	 * If this still occurs then you should see a hang
	 * or crash shortly after this line:
	 */
	printk("lockdep: fixing up alternatives.\n");
#endif

	if (noreplace_smp || smp_alt_once)
		return;
	BUG_ON(!smp && (num_online_cpus() > 1));

	spin_lock_irqsave(&smp_alt, flags);

	/*
	 * Avoid unnecessary switches because it forces JIT based VMs to
	 * throw away all cached translations, which can be quite costly.
	 */
	if (smp == smp_mode) {
		/* nothing */
	} else if (smp) {
		printk(KERN_INFO "SMP alternatives: switching to SMP code\n");
		clear_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
		clear_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
		list_for_each_entry(mod, &smp_alt_modules, next)
			alternatives_smp_lock(mod->locks, mod->locks_end,
					      mod->text, mod->text_end);
	} else {
		printk(KERN_INFO "SMP alternatives: switching to UP code\n");
		set_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
		set_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
		list_for_each_entry(mod, &smp_alt_modules, next)
			alternatives_smp_unlock(mod->locks, mod->locks_end,
						mod->text, mod->text_end);
	}
	smp_mode = smp;
	spin_unlock_irqrestore(&smp_alt, flags);
}

#endif

#ifdef CONFIG_PARAVIRT
void apply_paravirt(struct paravirt_patch_site *start,
		    struct paravirt_patch_site *end)
{
	struct paravirt_patch_site *p;
	char insnbuf[MAX_PATCH_LEN];

	if (noreplace_paravirt)
		return;

	for (p = start; p < end; p++) {
		unsigned int used;

		BUG_ON(p->len > MAX_PATCH_LEN);
		/* prep the buffer with the original instructions */
		memcpy(insnbuf, p->instr, p->len);
		used = pv_init_ops.patch(p->instrtype, p->clobbers, insnbuf,
					 (unsigned long)p->instr, p->len);

		BUG_ON(used > p->len);

		/* Pad the rest with nops */
		add_nops(insnbuf + used, p->len - used);
		text_poke(p->instr, insnbuf, p->len);
	}
}
extern struct paravirt_patch_site __start_parainstructions[],
	__stop_parainstructions[];
#endif	/* CONFIG_PARAVIRT */

void __init alternative_instructions(void)
{
	unsigned long flags;

	/* The patching is not fully atomic, so try to avoid local interruptions
	   that might execute the to be patched code.
	   Other CPUs are not running. */
	stop_nmi();
#ifdef CONFIG_X86_MCE
	stop_mce();
#endif

	local_irq_save(flags);
	apply_alternatives(__alt_instructions, __alt_instructions_end);

	/* switch to patch-once-at-boottime-only mode and free the
	 * tables in case we know the number of CPUs will never ever
	 * change */
#ifdef CONFIG_HOTPLUG_CPU
	if (num_possible_cpus() < 2)
		smp_alt_once = 1;
#endif

#ifdef CONFIG_SMP
	if (smp_alt_once) {
		if (1 == num_possible_cpus()) {
			printk(KERN_INFO "SMP alternatives: switching to UP code\n");
			set_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
			set_cpu_cap(&cpu_data(0), X86_FEATURE_UP);

			alternatives_smp_unlock(__smp_locks, __smp_locks_end,
						_text, _etext);
		}
	} else {
		alternatives_smp_module_add(NULL, "core kernel",
					    __smp_locks, __smp_locks_end,
					    _text, _etext);

		/* Only switch to UP mode if we don't immediately boot others */
		if (num_possible_cpus() == 1 || setup_max_cpus <= 1)
			alternatives_smp_switch(0);
	}
#endif
 	apply_paravirt(__parainstructions, __parainstructions_end);
	local_irq_restore(flags);

	if (smp_alt_once)
		free_init_pages("SMP alternatives",
				(unsigned long)__smp_locks,
				(unsigned long)__smp_locks_end);

	restart_nmi();
#ifdef CONFIG_X86_MCE
	restart_mce();
#endif
}

/*
 * Warning:
 * When you use this code to patch more than one byte of an instruction
 * you need to make sure that other CPUs cannot execute this code in parallel.
 * Also no thread must be currently preempted in the middle of these instructions.
 * And on the local CPU you need to be protected again NMI or MCE handlers
 * seeing an inconsistent instruction while you patch.
 */
void __kprobes text_poke(void *addr, unsigned char *opcode, int len)
{
	memcpy(addr, opcode, len);
	sync_core();
	/* Could also do a CLFLUSH here to speed up CPU recovery; but
	   that causes hangs on some VIA CPUs. */
}
Commit	Line	Data
9a0b5817	1	#include <linux/module.h>
f6a57033	2	#include <linux/sched.h>
9a0b5817 GH	3	#include <linux/spinlock.h>
9a0b5817 GH	4	#include <linux/list.h>
19d36ccd AK	5	#include <linux/kprobes.h>
	6	#include <linux/mm.h>
	7	#include <linux/vmalloc.h>
9a0b5817 GH	8	#include <asm/alternative.h>
9a0b5817 GH	9	#include <asm/sections.h>
19d36ccd	10	#include <asm/pgtable.h>
8f4e956b AK	11	#include <asm/mce.h>
8f4e956b AK	12	#include <asm/nmi.h>
b097976e	13	#include <asm/vsyscall.h>
9a0b5817	14
ab144f5e AK	15	#define MAX_PATCH_LEN (255-1)
ab144f5e AK	16
09488165 JB	17	#ifdef CONFIG_HOTPLUG_CPU
09488165 JB	18	static int smp_alt_once;
9a0b5817	19
d167a518 GH	20	static int __init bootonly(char *str)
	21	{
	22	smp_alt_once = 1;
	23	return 1;
	24	}
b7fb4af0	25	__setup("smp-alt-boot", bootonly);
09488165 JB	26	#else
	27	#define smp_alt_once 1
	28	#endif
	29
	30	static int debug_alternative;
b7fb4af0	31
d167a518 GH	32	static int __init debug_alt(char *str)
	33	{
	34	debug_alternative = 1;
	35	return 1;
	36	}
d167a518 GH	37	__setup("debug-alternative", debug_alt);
d167a518 GH	38
09488165 JB	39	static int noreplace_smp;
09488165 JB	40
b7fb4af0 JF	41	static int __init setup_noreplace_smp(char *str)
	42	{
	43	noreplace_smp = 1;
	44	return 1;
	45	}
	46	__setup("noreplace-smp", setup_noreplace_smp);
	47
959b4fdf JF	48	#ifdef CONFIG_PARAVIRT
	49	static int noreplace_paravirt = 0;
	50
	51	static int __init setup_noreplace_paravirt(char *str)
	52	{
	53	noreplace_paravirt = 1;
	54	return 1;
	55	}
	56	__setup("noreplace-paravirt", setup_noreplace_paravirt);
	57	#endif
b7fb4af0	58
d167a518 GH	59	#define DPRINTK(fmt, args...) if (debug_alternative) \
	60	printk(KERN_DEBUG fmt, args)
	61
	62	#ifdef GENERIC_NOP1
9a0b5817 GH	63	/* Use inline assembly to define this because the nops are defined
	64	as inline assembly strings in the include files and we cannot
	65	get them easily into strings. */
121d7bf5	66	asm("\t.section .rodata, \"a\"\nintelnops: "
9a0b5817 GH	67	GENERIC_NOP1 GENERIC_NOP2 GENERIC_NOP3 GENERIC_NOP4 GENERIC_NOP5 GENERIC_NOP6
9a0b5817 GH	68	GENERIC_NOP7 GENERIC_NOP8);
121d7bf5 JB	69	extern const unsigned char intelnops[];
121d7bf5 JB	70	static const unsigned char *const intel_nops[ASM_NOP_MAX+1] = {
9a0b5817 GH	71	NULL,
	72	intelnops,
	73	intelnops + 1,
	74	intelnops + 1 + 2,
	75	intelnops + 1 + 2 + 3,
	76	intelnops + 1 + 2 + 3 + 4,
	77	intelnops + 1 + 2 + 3 + 4 + 5,
	78	intelnops + 1 + 2 + 3 + 4 + 5 + 6,
	79	intelnops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
	80	};
d167a518 GH	81	#endif
	82
	83	#ifdef K8_NOP1
121d7bf5	84	asm("\t.section .rodata, \"a\"\nk8nops: "
d167a518 GH	85	K8_NOP1 K8_NOP2 K8_NOP3 K8_NOP4 K8_NOP5 K8_NOP6
d167a518 GH	86	K8_NOP7 K8_NOP8);
121d7bf5 JB	87	extern const unsigned char k8nops[];
121d7bf5 JB	88	static const unsigned char *const k8_nops[ASM_NOP_MAX+1] = {
9a0b5817 GH	89	NULL,
	90	k8nops,
	91	k8nops + 1,
	92	k8nops + 1 + 2,
	93	k8nops + 1 + 2 + 3,
	94	k8nops + 1 + 2 + 3 + 4,
	95	k8nops + 1 + 2 + 3 + 4 + 5,
	96	k8nops + 1 + 2 + 3 + 4 + 5 + 6,
	97	k8nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
	98	};
d167a518 GH	99	#endif
	100
	101	#ifdef K7_NOP1
121d7bf5	102	asm("\t.section .rodata, \"a\"\nk7nops: "
d167a518 GH	103	K7_NOP1 K7_NOP2 K7_NOP3 K7_NOP4 K7_NOP5 K7_NOP6
d167a518 GH	104	K7_NOP7 K7_NOP8);
121d7bf5 JB	105	extern const unsigned char k7nops[];
121d7bf5 JB	106	static const unsigned char *const k7_nops[ASM_NOP_MAX+1] = {
9a0b5817 GH	107	NULL,
	108	k7nops,
	109	k7nops + 1,
	110	k7nops + 1 + 2,
	111	k7nops + 1 + 2 + 3,
	112	k7nops + 1 + 2 + 3 + 4,
	113	k7nops + 1 + 2 + 3 + 4 + 5,
	114	k7nops + 1 + 2 + 3 + 4 + 5 + 6,
	115	k7nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
	116	};
d167a518 GH	117	#endif
d167a518 GH	118
32c464f5 JB	119	#ifdef P6_NOP1
	120	asm("\t.section .rodata, \"a\"\np6nops: "
	121	P6_NOP1 P6_NOP2 P6_NOP3 P6_NOP4 P6_NOP5 P6_NOP6
	122	P6_NOP7 P6_NOP8);
	123	extern const unsigned char p6nops[];
	124	static const unsigned char *const p6_nops[ASM_NOP_MAX+1] = {
	125	NULL,
	126	p6nops,
	127	p6nops + 1,
	128	p6nops + 1 + 2,
	129	p6nops + 1 + 2 + 3,
	130	p6nops + 1 + 2 + 3 + 4,
	131	p6nops + 1 + 2 + 3 + 4 + 5,
	132	p6nops + 1 + 2 + 3 + 4 + 5 + 6,
	133	p6nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
	134	};
	135	#endif
	136
d167a518 GH	137	#ifdef CONFIG_X86_64
	138
	139	extern char __vsyscall_0;
121d7bf5	140	static inline const unsigned charconst find_nop_table(void)
d167a518	141	{
32c464f5 JB	142	return boot_cpu_data.x86_vendor != X86_VENDOR_INTEL \|\|
32c464f5 JB	143	boot_cpu_data.x86 < 6 ? k8_nops : p6_nops;
d167a518 GH	144	}
	145
	146	#else /* CONFIG_X86_64 */
	147
121d7bf5	148	static const struct nop {
9a0b5817	149	int cpuid;
121d7bf5	150	const unsigned char const noptable;
9a0b5817 GH	151	} noptypes[] = {
	152	{ X86_FEATURE_K8, k8_nops },
	153	{ X86_FEATURE_K7, k7_nops },
32c464f5 JB	154	{ X86_FEATURE_P4, p6_nops },
32c464f5 JB	155	{ X86_FEATURE_P3, p6_nops },
9a0b5817 GH	156	{ -1, NULL }
	157	};
	158
121d7bf5	159	static const unsigned charconst find_nop_table(void)
9a0b5817	160	{
121d7bf5	161	const unsigned char const noptable = intel_nops;
9a0b5817 GH	162	int i;
	163
	164	for (i = 0; noptypes[i].cpuid >= 0; i++) {
	165	if (boot_cpu_has(noptypes[i].cpuid)) {
	166	noptable = noptypes[i].noptable;
	167	break;
	168	}
	169	}
	170	return noptable;
	171	}
	172
d167a518 GH	173	#endif /* CONFIG_X86_64 */
d167a518 GH	174
ab144f5e AK	175	/* Use this to add nops to a buffer, then text_poke the whole buffer. */
ab144f5e AK	176	static void add_nops(void *insns, unsigned int len)
139ec7c4	177	{
121d7bf5	178	const unsigned char const noptable = find_nop_table();
139ec7c4 RR	179
	180	while (len > 0) {
	181	unsigned int noplen = len;
	182	if (noplen > ASM_NOP_MAX)
	183	noplen = ASM_NOP_MAX;
ab144f5e	184	memcpy(insns, noptable[noplen], noplen);
139ec7c4 RR	185	insns += noplen;
	186	len -= noplen;
	187	}
	188	}
	189
d167a518	190	extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
d167a518 GH	191	extern u8 __smp_locks[], __smp_locks_end[];
d167a518 GH	192
9a0b5817 GH	193	/* Replace instructions with better alternatives for this CPU type.
	194	This runs before SMP is initialized to avoid SMP problems with
	195	self modifying code. This implies that assymetric systems where
	196	APs have less capabilities than the boot processor are not handled.
	197	Tough. Make sure you disable such features by hand. */
	198
	199	void apply_alternatives(struct alt_instr start, struct alt_instr end)
	200	{
9a0b5817	201	struct alt_instr *a;
ab144f5e	202	char insnbuf[MAX_PATCH_LEN];
9a0b5817 GH	203
	204	DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end);
	205	for (a = start; a < end; a++) {
ab144f5e	206	u8 *instr = a->instr;
9a0b5817	207	BUG_ON(a->replacementlen > a->instrlen);
ab144f5e	208	BUG_ON(a->instrlen > sizeof(insnbuf));
9a0b5817 GH	209	if (!boot_cpu_has(a->cpuid))
9a0b5817 GH	210	continue;
d167a518 GH	211	#ifdef CONFIG_X86_64
	212	/* vsyscall code is not mapped yet. resolve it manually. */
	213	if (instr >= (u8 )VSYSCALL_START && instr < (u8)VSYSCALL_END) {
	214	instr = __va(instr - (u8)VSYSCALL_START + (u8)__pa_symbol(&__vsyscall_0));
	215	DPRINTK("%s: vsyscall fixup: %p => %p\n",
	216	__FUNCTION__, a->instr, instr);
	217	}
	218	#endif
ab144f5e AK	219	memcpy(insnbuf, a->replacement, a->replacementlen);
	220	add_nops(insnbuf + a->replacementlen,
	221	a->instrlen - a->replacementlen);
	222	text_poke(instr, insnbuf, a->instrlen);
9a0b5817 GH	223	}
	224	}
	225
8ec4d41f GH	226	#ifdef CONFIG_SMP
8ec4d41f GH	227
9a0b5817 GH	228	static void alternatives_smp_lock(u8 start, u8 end, u8 text, u8 text_end)
	229	{
	230	u8 **ptr;
	231
	232	for (ptr = start; ptr < end; ptr++) {
	233	if (*ptr < text)
	234	continue;
	235	if (*ptr > text_end)
	236	continue;
19d36ccd	237	text_poke(ptr, ((unsigned char []){0xf0}), 1); / add lock prefix */
9a0b5817 GH	238	};
	239	}
	240
	241	static void alternatives_smp_unlock(u8 start, u8 end, u8 text, u8 text_end)
	242	{
9a0b5817	243	u8 **ptr;
ab144f5e	244	char insn[1];
9a0b5817	245
b7fb4af0 JF	246	if (noreplace_smp)
	247	return;
	248
ab144f5e	249	add_nops(insn, 1);
9a0b5817 GH	250	for (ptr = start; ptr < end; ptr++) {
	251	if (*ptr < text)
	252	continue;
	253	if (*ptr > text_end)
	254	continue;
ab144f5e	255	text_poke(*ptr, insn, 1);
9a0b5817 GH	256	};
	257	}
	258
	259	struct smp_alt_module {
	260	/* what is this ??? */
	261	struct module *mod;
	262	char *name;
	263
	264	/* ptrs to lock prefixes */
	265	u8 **locks;
	266	u8 **locks_end;
	267
	268	/* .text segment, needed to avoid patching init code ;) */
	269	u8 *text;
	270	u8 *text_end;
	271
	272	struct list_head next;
	273	};
	274	static LIST_HEAD(smp_alt_modules);
	275	static DEFINE_SPINLOCK(smp_alt);
ca74a6f8	276	static int smp_mode = 1; /* protected by smp_alt */
9a0b5817	277
9a0b5817 GH	278	void alternatives_smp_module_add(struct module mod, char name,
	279	void locks, void locks_end,
	280	void text, void text_end)
	281	{
	282	struct smp_alt_module *smp;
	283	unsigned long flags;
	284
b7fb4af0 JF	285	if (noreplace_smp)
	286	return;
	287
9a0b5817 GH	288	if (smp_alt_once) {
	289	if (boot_cpu_has(X86_FEATURE_UP))
	290	alternatives_smp_unlock(locks, locks_end,
	291	text, text_end);
	292	return;
	293	}
	294
	295	smp = kzalloc(sizeof(*smp), GFP_KERNEL);
	296	if (NULL == smp)
	297	return; /* we'll run the (safe but slow) SMP code then ... */
	298
	299	smp->mod = mod;
	300	smp->name = name;
	301	smp->locks = locks;
	302	smp->locks_end = locks_end;
	303	smp->text = text;
	304	smp->text_end = text_end;
	305	DPRINTK("%s: locks %p -> %p, text %p -> %p, name %s\n",
	306	__FUNCTION__, smp->locks, smp->locks_end,
	307	smp->text, smp->text_end, smp->name);
	308
	309	spin_lock_irqsave(&smp_alt, flags);
	310	list_add_tail(&smp->next, &smp_alt_modules);
	311	if (boot_cpu_has(X86_FEATURE_UP))
	312	alternatives_smp_unlock(smp->locks, smp->locks_end,
	313	smp->text, smp->text_end);
	314	spin_unlock_irqrestore(&smp_alt, flags);
	315	}
	316
	317	void alternatives_smp_module_del(struct module *mod)
	318	{
	319	struct smp_alt_module *item;
	320	unsigned long flags;
	321
b7fb4af0	322	if (smp_alt_once \|\| noreplace_smp)
9a0b5817 GH	323	return;
	324
	325	spin_lock_irqsave(&smp_alt, flags);
	326	list_for_each_entry(item, &smp_alt_modules, next) {
	327	if (mod != item->mod)
	328	continue;
	329	list_del(&item->next);
	330	spin_unlock_irqrestore(&smp_alt, flags);
	331	DPRINTK("%s: %s\n", __FUNCTION__, item->name);
	332	kfree(item);
	333	return;
	334	}
	335	spin_unlock_irqrestore(&smp_alt, flags);
	336	}
	337
	338	void alternatives_smp_switch(int smp)
	339	{
	340	struct smp_alt_module *mod;
	341	unsigned long flags;
	342
3047e99e IM	343	#ifdef CONFIG_LOCKDEP
3047e99e IM	344	/*
17abecfe IM	345	* Older binutils section handling bug prevented
	346	* alternatives-replacement from working reliably.
	347	*
	348	* If this still occurs then you should see a hang
	349	* or crash shortly after this line:
3047e99e	350	*/
17abecfe	351	printk("lockdep: fixing up alternatives.\n");
3047e99e IM	352	#endif
3047e99e IM	353
b7fb4af0	354	if (noreplace_smp \|\| smp_alt_once)
9a0b5817 GH	355	return;
	356	BUG_ON(!smp && (num_online_cpus() > 1));
	357
	358	spin_lock_irqsave(&smp_alt, flags);
ca74a6f8 AK	359
	360	/*
	361	* Avoid unnecessary switches because it forces JIT based VMs to
	362	* throw away all cached translations, which can be quite costly.
	363	*/
	364	if (smp == smp_mode) {
	365	/* nothing */
	366	} else if (smp) {
9a0b5817	367	printk(KERN_INFO "SMP alternatives: switching to SMP code\n");
53756d37 JF	368	clear_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
53756d37 JF	369	clear_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
9a0b5817 GH	370	list_for_each_entry(mod, &smp_alt_modules, next)
	371	alternatives_smp_lock(mod->locks, mod->locks_end,
	372	mod->text, mod->text_end);
	373	} else {
	374	printk(KERN_INFO "SMP alternatives: switching to UP code\n");
53756d37 JF	375	set_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
53756d37 JF	376	set_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
9a0b5817 GH	377	list_for_each_entry(mod, &smp_alt_modules, next)
	378	alternatives_smp_unlock(mod->locks, mod->locks_end,
	379	mod->text, mod->text_end);
	380	}
ca74a6f8	381	smp_mode = smp;
9a0b5817 GH	382	spin_unlock_irqrestore(&smp_alt, flags);
	383	}
	384
8ec4d41f GH	385	#endif
8ec4d41f GH	386
139ec7c4	387	#ifdef CONFIG_PARAVIRT
98de032b JF	388	void apply_paravirt(struct paravirt_patch_site *start,
98de032b JF	389	struct paravirt_patch_site *end)
139ec7c4	390	{
98de032b	391	struct paravirt_patch_site *p;
ab144f5e	392	char insnbuf[MAX_PATCH_LEN];
139ec7c4	393
959b4fdf JF	394	if (noreplace_paravirt)
	395	return;
	396
139ec7c4 RR	397	for (p = start; p < end; p++) {
	398	unsigned int used;
	399
ab144f5e	400	BUG_ON(p->len > MAX_PATCH_LEN);
d34fda4a CW	401	/* prep the buffer with the original instructions */
d34fda4a CW	402	memcpy(insnbuf, p->instr, p->len);
93b1eab3 JF	403	used = pv_init_ops.patch(p->instrtype, p->clobbers, insnbuf,
93b1eab3 JF	404	(unsigned long)p->instr, p->len);
7f63c41c	405
63f70270 JF	406	BUG_ON(used > p->len);
63f70270 JF	407
139ec7c4	408	/* Pad the rest with nops */
ab144f5e AK	409	add_nops(insnbuf + used, p->len - used);
ab144f5e AK	410	text_poke(p->instr, insnbuf, p->len);
139ec7c4	411	}
139ec7c4	412	}
98de032b	413	extern struct paravirt_patch_site __start_parainstructions[],
139ec7c4 RR	414	__stop_parainstructions[];
	415	#endif /* CONFIG_PARAVIRT */
	416
9a0b5817 GH	417	void __init alternative_instructions(void)
9a0b5817 GH	418	{
e51959fa	419	unsigned long flags;
e51959fa	420
8f4e956b AK	421	/* The patching is not fully atomic, so try to avoid local interruptions
	422	that might execute the to be patched code.
	423	Other CPUs are not running. */
	424	stop_nmi();
d2d0251f	425	#ifdef CONFIG_X86_MCE
8f4e956b AK	426	stop_mce();
	427	#endif
	428
e51959fa	429	local_irq_save(flags);
9a0b5817 GH	430	apply_alternatives(__alt_instructions, __alt_instructions_end);
	431
	432	/* switch to patch-once-at-boottime-only mode and free the
	433	* tables in case we know the number of CPUs will never ever
	434	* change */
	435	#ifdef CONFIG_HOTPLUG_CPU
	436	if (num_possible_cpus() < 2)
	437	smp_alt_once = 1;
9a0b5817 GH	438	#endif
9a0b5817 GH	439
8ec4d41f	440	#ifdef CONFIG_SMP
9a0b5817 GH	441	if (smp_alt_once) {
	442	if (1 == num_possible_cpus()) {
	443	printk(KERN_INFO "SMP alternatives: switching to UP code\n");
53756d37 JF	444	set_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
	445	set_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
	446
9a0b5817 GH	447	alternatives_smp_unlock(__smp_locks, __smp_locks_end,
	448	_text, _etext);
	449	}
9a0b5817	450	} else {
9a0b5817 GH	451	alternatives_smp_module_add(NULL, "core kernel",
	452	__smp_locks, __smp_locks_end,
	453	_text, _etext);
ca74a6f8 AK	454
	455	/* Only switch to UP mode if we don't immediately boot others */
	456	if (num_possible_cpus() == 1 \|\| setup_max_cpus <= 1)
	457	alternatives_smp_switch(0);
9a0b5817	458	}
8ec4d41f	459	#endif
441d40dc	460	apply_paravirt(__parainstructions, __parainstructions_end);
e51959fa	461	local_irq_restore(flags);
8f4e956b	462
f68fd5f4 FW	463	if (smp_alt_once)
	464	free_init_pages("SMP alternatives",
	465	(unsigned long)__smp_locks,
	466	(unsigned long)__smp_locks_end);
	467
8f4e956b	468	restart_nmi();
d2d0251f	469	#ifdef CONFIG_X86_MCE
8f4e956b AK	470	restart_mce();
8f4e956b AK	471	#endif
9a0b5817	472	}
19d36ccd AK	473
	474	/*
	475	* Warning:
	476	* When you use this code to patch more than one byte of an instruction
	477	* you need to make sure that other CPUs cannot execute this code in parallel.
	478	* Also no thread must be currently preempted in the middle of these instructions.
	479	* And on the local CPU you need to be protected again NMI or MCE handlers
	480	* seeing an inconsistent instruction while you patch.
	481	*/
602033ed	482	void __kprobes text_poke(void addr, unsigned char opcode, int len)
19d36ccd	483	{
19d36ccd AK	484	memcpy(addr, opcode, len);
19d36ccd AK	485	sync_core();
a534b679 AK	486	/* Could also do a CLFLUSH here to speed up CPU recovery; but
a534b679 AK	487	that causes hangs on some VIA CPUs. */
19d36ccd	488	}