[linux-2.6-block.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);

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
	/*
	 * A not yet fixed binutils section handling bug prevents
	 * alternatives-replacement from working reliably, so turn
	 * it off:
	 */
	printk("lockdep: not fixing up alternatives.\n");
	return;
#endif

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

	spin_lock_irqsave(&smp_alt, flags);
	if (smp) {
		printk(KERN_INFO "SMP alternatives: switching to SMP code\n");
		clear_bit(X86_FEATURE_UP, boot_cpu_data.x86_capability);
		clear_bit(X86_FEATURE_UP, cpu_data(0).x86_capability);
		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_bit(X86_FEATURE_UP, boot_cpu_data.x86_capability);
		set_bit(X86_FEATURE_UP, cpu_data(0).x86_capability);
		list_for_each_entry(mod, &smp_alt_modules, next)
			alternatives_smp_unlock(mod->locks, mod->locks_end,
						mod->text, mod->text_end);
	}
	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_bit(X86_FEATURE_UP, boot_cpu_data.x86_capability);
			set_bit(X86_FEATURE_UP, cpu_data(0).x86_capability);
			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);
		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);
	276
9a0b5817 GH	277	void alternatives_smp_module_add(struct module mod, char name,
	278	void locks, void locks_end,
	279	void text, void text_end)
	280	{
	281	struct smp_alt_module *smp;
	282	unsigned long flags;
	283
b7fb4af0 JF	284	if (noreplace_smp)
	285	return;
	286
9a0b5817 GH	287	if (smp_alt_once) {
	288	if (boot_cpu_has(X86_FEATURE_UP))
	289	alternatives_smp_unlock(locks, locks_end,
	290	text, text_end);
	291	return;
	292	}
	293
	294	smp = kzalloc(sizeof(*smp), GFP_KERNEL);
	295	if (NULL == smp)
	296	return; /* we'll run the (safe but slow) SMP code then ... */
	297
	298	smp->mod = mod;
	299	smp->name = name;
	300	smp->locks = locks;
	301	smp->locks_end = locks_end;
	302	smp->text = text;
	303	smp->text_end = text_end;
	304	DPRINTK("%s: locks %p -> %p, text %p -> %p, name %s\n",
	305	__FUNCTION__, smp->locks, smp->locks_end,
	306	smp->text, smp->text_end, smp->name);
	307
	308	spin_lock_irqsave(&smp_alt, flags);
	309	list_add_tail(&smp->next, &smp_alt_modules);
	310	if (boot_cpu_has(X86_FEATURE_UP))
	311	alternatives_smp_unlock(smp->locks, smp->locks_end,
	312	smp->text, smp->text_end);
	313	spin_unlock_irqrestore(&smp_alt, flags);
	314	}
	315
	316	void alternatives_smp_module_del(struct module *mod)
	317	{
	318	struct smp_alt_module *item;
	319	unsigned long flags;
	320
b7fb4af0	321	if (smp_alt_once \|\| noreplace_smp)
9a0b5817 GH	322	return;
	323
	324	spin_lock_irqsave(&smp_alt, flags);
	325	list_for_each_entry(item, &smp_alt_modules, next) {
	326	if (mod != item->mod)
	327	continue;
	328	list_del(&item->next);
	329	spin_unlock_irqrestore(&smp_alt, flags);
	330	DPRINTK("%s: %s\n", __FUNCTION__, item->name);
	331	kfree(item);
	332	return;
	333	}
	334	spin_unlock_irqrestore(&smp_alt, flags);
	335	}
	336
	337	void alternatives_smp_switch(int smp)
	338	{
	339	struct smp_alt_module *mod;
	340	unsigned long flags;
	341
3047e99e IM	342	#ifdef CONFIG_LOCKDEP
	343	/*
	344	* A not yet fixed binutils section handling bug prevents
	345	* alternatives-replacement from working reliably, so turn
	346	* it off:
	347	*/
	348	printk("lockdep: not fixing up alternatives.\n");
	349	return;
	350	#endif
	351
b7fb4af0	352	if (noreplace_smp \|\| smp_alt_once)
9a0b5817 GH	353	return;
	354	BUG_ON(!smp && (num_online_cpus() > 1));
	355
	356	spin_lock_irqsave(&smp_alt, flags);
	357	if (smp) {
	358	printk(KERN_INFO "SMP alternatives: switching to SMP code\n");
	359	clear_bit(X86_FEATURE_UP, boot_cpu_data.x86_capability);
92cb7612	360	clear_bit(X86_FEATURE_UP, cpu_data(0).x86_capability);
9a0b5817 GH	361	list_for_each_entry(mod, &smp_alt_modules, next)
	362	alternatives_smp_lock(mod->locks, mod->locks_end,
	363	mod->text, mod->text_end);
	364	} else {
	365	printk(KERN_INFO "SMP alternatives: switching to UP code\n");
	366	set_bit(X86_FEATURE_UP, boot_cpu_data.x86_capability);
92cb7612	367	set_bit(X86_FEATURE_UP, cpu_data(0).x86_capability);
9a0b5817 GH	368	list_for_each_entry(mod, &smp_alt_modules, next)
	369	alternatives_smp_unlock(mod->locks, mod->locks_end,
	370	mod->text, mod->text_end);
	371	}
	372	spin_unlock_irqrestore(&smp_alt, flags);
	373	}
	374
8ec4d41f GH	375	#endif
8ec4d41f GH	376
139ec7c4	377	#ifdef CONFIG_PARAVIRT
98de032b JF	378	void apply_paravirt(struct paravirt_patch_site *start,
98de032b JF	379	struct paravirt_patch_site *end)
139ec7c4	380	{
98de032b	381	struct paravirt_patch_site *p;
ab144f5e	382	char insnbuf[MAX_PATCH_LEN];
139ec7c4	383
959b4fdf JF	384	if (noreplace_paravirt)
	385	return;
	386
139ec7c4 RR	387	for (p = start; p < end; p++) {
	388	unsigned int used;
	389
ab144f5e	390	BUG_ON(p->len > MAX_PATCH_LEN);
d34fda4a CW	391	/* prep the buffer with the original instructions */
d34fda4a CW	392	memcpy(insnbuf, p->instr, p->len);
93b1eab3 JF	393	used = pv_init_ops.patch(p->instrtype, p->clobbers, insnbuf,
93b1eab3 JF	394	(unsigned long)p->instr, p->len);
7f63c41c	395
63f70270 JF	396	BUG_ON(used > p->len);
63f70270 JF	397
139ec7c4	398	/* Pad the rest with nops */
ab144f5e AK	399	add_nops(insnbuf + used, p->len - used);
ab144f5e AK	400	text_poke(p->instr, insnbuf, p->len);
139ec7c4	401	}
139ec7c4	402	}
98de032b	403	extern struct paravirt_patch_site __start_parainstructions[],
139ec7c4 RR	404	__stop_parainstructions[];
	405	#endif /* CONFIG_PARAVIRT */
	406
9a0b5817 GH	407	void __init alternative_instructions(void)
9a0b5817 GH	408	{
e51959fa	409	unsigned long flags;
e51959fa	410
8f4e956b AK	411	/* The patching is not fully atomic, so try to avoid local interruptions
	412	that might execute the to be patched code.
	413	Other CPUs are not running. */
	414	stop_nmi();
d2d0251f	415	#ifdef CONFIG_X86_MCE
8f4e956b AK	416	stop_mce();
	417	#endif
	418
e51959fa	419	local_irq_save(flags);
9a0b5817 GH	420	apply_alternatives(__alt_instructions, __alt_instructions_end);
	421
	422	/* switch to patch-once-at-boottime-only mode and free the
	423	* tables in case we know the number of CPUs will never ever
	424	* change */
	425	#ifdef CONFIG_HOTPLUG_CPU
	426	if (num_possible_cpus() < 2)
	427	smp_alt_once = 1;
9a0b5817 GH	428	#endif
9a0b5817 GH	429
8ec4d41f	430	#ifdef CONFIG_SMP
9a0b5817 GH	431	if (smp_alt_once) {
	432	if (1 == num_possible_cpus()) {
	433	printk(KERN_INFO "SMP alternatives: switching to UP code\n");
	434	set_bit(X86_FEATURE_UP, boot_cpu_data.x86_capability);
92cb7612	435	set_bit(X86_FEATURE_UP, cpu_data(0).x86_capability);
9a0b5817 GH	436	alternatives_smp_unlock(__smp_locks, __smp_locks_end,
	437	_text, _etext);
	438	}
9a0b5817	439	} else {
9a0b5817 GH	440	alternatives_smp_module_add(NULL, "core kernel",
	441	__smp_locks, __smp_locks_end,
	442	_text, _etext);
	443	alternatives_smp_switch(0);
	444	}
8ec4d41f	445	#endif
441d40dc	446	apply_paravirt(__parainstructions, __parainstructions_end);
e51959fa	447	local_irq_restore(flags);
8f4e956b	448
f68fd5f4 FW	449	if (smp_alt_once)
	450	free_init_pages("SMP alternatives",
	451	(unsigned long)__smp_locks,
	452	(unsigned long)__smp_locks_end);
	453
8f4e956b	454	restart_nmi();
d2d0251f	455	#ifdef CONFIG_X86_MCE
8f4e956b AK	456	restart_mce();
8f4e956b AK	457	#endif
9a0b5817	458	}
19d36ccd AK	459
	460	/*
	461	* Warning:
	462	* When you use this code to patch more than one byte of an instruction
	463	* you need to make sure that other CPUs cannot execute this code in parallel.
	464	* Also no thread must be currently preempted in the middle of these instructions.
	465	* And on the local CPU you need to be protected again NMI or MCE handlers
	466	* seeing an inconsistent instruction while you patch.
	467	*/
602033ed	468	void __kprobes text_poke(void addr, unsigned char opcode, int len)
19d36ccd	469	{
19d36ccd AK	470	memcpy(addr, opcode, len);
19d36ccd AK	471	sync_core();
a534b679 AK	472	/* Could also do a CLFLUSH here to speed up CPU recovery; but
a534b679 AK	473	that causes hangs on some VIA CPUs. */
19d36ccd	474	}