[net-next-2.6.git] / arch / x86 / mm / tlb.c

#include <linux/init.h>

#include <linux/mm.h>
#include <linux/spinlock.h>
#include <linux/smp.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/cpu.h>

#include <asm/tlbflush.h>
#include <asm/mmu_context.h>
#include <asm/cache.h>
#include <asm/apic.h>
#include <asm/uv/uv.h>

DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate)
			= { &init_mm, 0, };

/*
 *	Smarter SMP flushing macros.
 *		c/o Linus Torvalds.
 *
 *	These mean you can really definitely utterly forget about
 *	writing to user space from interrupts. (Its not allowed anyway).
 *
 *	Optimizations Manfred Spraul <manfred@colorfullife.com>
 *
 *	More scalable flush, from Andi Kleen
 *
 *	To avoid global state use 8 different call vectors.
 *	Each CPU uses a specific vector to trigger flushes on other
 *	CPUs. Depending on the received vector the target CPUs look into
 *	the right array slot for the flush data.
 *
 *	With more than 8 CPUs they are hashed to the 8 available
 *	vectors. The limited global vector space forces us to this right now.
 *	In future when interrupts are split into per CPU domains this could be
 *	fixed, at the cost of triggering multiple IPIs in some cases.
 */

union smp_flush_state {
	struct {
		struct mm_struct *flush_mm;
		unsigned long flush_va;
		raw_spinlock_t tlbstate_lock;
		DECLARE_BITMAP(flush_cpumask, NR_CPUS);
	};
	char pad[INTERNODE_CACHE_BYTES];
} ____cacheline_internodealigned_in_smp;

/* State is put into the per CPU data section, but padded
   to a full cache line because other CPUs can access it and we don't
   want false sharing in the per cpu data segment. */
static union smp_flush_state flush_state[NUM_INVALIDATE_TLB_VECTORS];

static DEFINE_PER_CPU_READ_MOSTLY(int, tlb_vector_offset);

/*
 * We cannot call mmdrop() because we are in interrupt context,
 * instead update mm->cpu_vm_mask.
 */
void leave_mm(int cpu)
{
	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK)
		BUG();
	cpumask_clear_cpu(cpu,
			  mm_cpumask(percpu_read(cpu_tlbstate.active_mm)));
	load_cr3(swapper_pg_dir);
}
EXPORT_SYMBOL_GPL(leave_mm);

/*
 *
 * The flush IPI assumes that a thread switch happens in this order:
 * [cpu0: the cpu that switches]
 * 1) switch_mm() either 1a) or 1b)
 * 1a) thread switch to a different mm
 * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask);
 *	Stop ipi delivery for the old mm. This is not synchronized with
 *	the other cpus, but smp_invalidate_interrupt ignore flush ipis
 *	for the wrong mm, and in the worst case we perform a superfluous
 *	tlb flush.
 * 1a2) set cpu mmu_state to TLBSTATE_OK
 *	Now the smp_invalidate_interrupt won't call leave_mm if cpu0
 *	was in lazy tlb mode.
 * 1a3) update cpu active_mm
 *	Now cpu0 accepts tlb flushes for the new mm.
 * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask);
 *	Now the other cpus will send tlb flush ipis.
 * 1a4) change cr3.
 * 1b) thread switch without mm change
 *	cpu active_mm is correct, cpu0 already handles
 *	flush ipis.
 * 1b1) set cpu mmu_state to TLBSTATE_OK
 * 1b2) test_and_set the cpu bit in cpu_vm_mask.
 *	Atomically set the bit [other cpus will start sending flush ipis],
 *	and test the bit.
 * 1b3) if the bit was 0: leave_mm was called, flush the tlb.
 * 2) switch %%esp, ie current
 *
 * The interrupt must handle 2 special cases:
 * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm.
 * - the cpu performs speculative tlb reads, i.e. even if the cpu only
 *   runs in kernel space, the cpu could load tlb entries for user space
 *   pages.
 *
 * The good news is that cpu mmu_state is local to each cpu, no
 * write/read ordering problems.
 */

/*
 * TLB flush IPI:
 *
 * 1) Flush the tlb entries if the cpu uses the mm that's being flushed.
 * 2) Leave the mm if we are in the lazy tlb mode.
 *
 * Interrupts are disabled.
 */

/*
 * FIXME: use of asmlinkage is not consistent.  On x86_64 it's noop
 * but still used for documentation purpose but the usage is slightly
 * inconsistent.  On x86_32, asmlinkage is regparm(0) but interrupt
 * entry calls in with the first parameter in %eax.  Maybe define
 * intrlinkage?
 */
#ifdef CONFIG_X86_64
asmlinkage
#endif
void smp_invalidate_interrupt(struct pt_regs *regs)
{
	unsigned int cpu;
	unsigned int sender;
	union smp_flush_state *f;

	cpu = smp_processor_id();
	/*
	 * orig_rax contains the negated interrupt vector.
	 * Use that to determine where the sender put the data.
	 */
	sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START;
	f = &flush_state[sender];

	if (!cpumask_test_cpu(cpu, to_cpumask(f->flush_cpumask)))
		goto out;
		/*
		 * This was a BUG() but until someone can quote me the
		 * line from the intel manual that guarantees an IPI to
		 * multiple CPUs is retried _only_ on the erroring CPUs
		 * its staying as a return
		 *
		 * BUG();
		 */

	if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) {
		if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) {
			if (f->flush_va == TLB_FLUSH_ALL)
				local_flush_tlb();
			else
				__flush_tlb_one(f->flush_va);
		} else
			leave_mm(cpu);
	}
out:
	ack_APIC_irq();
	smp_mb__before_clear_bit();
	cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask));
	smp_mb__after_clear_bit();
	inc_irq_stat(irq_tlb_count);
}

static void flush_tlb_others_ipi(const struct cpumask *cpumask,
				 struct mm_struct *mm, unsigned long va)
{
	unsigned int sender;
	union smp_flush_state *f;

	/* Caller has disabled preemption */
	sender = this_cpu_read(tlb_vector_offset);
	f = &flush_state[sender];

	/*
	 * Could avoid this lock when
	 * num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is
	 * probably not worth checking this for a cache-hot lock.
	 */
	raw_spin_lock(&f->tlbstate_lock);

	f->flush_mm = mm;
	f->flush_va = va;
	if (cpumask_andnot(to_cpumask(f->flush_cpumask), cpumask, cpumask_of(smp_processor_id()))) {
		/*
		 * We have to send the IPI only to
		 * CPUs affected.
		 */
		apic->send_IPI_mask(to_cpumask(f->flush_cpumask),
			      INVALIDATE_TLB_VECTOR_START + sender);

		while (!cpumask_empty(to_cpumask(f->flush_cpumask)))
			cpu_relax();
	}

	f->flush_mm = NULL;
	f->flush_va = 0;
	raw_spin_unlock(&f->tlbstate_lock);
}

void native_flush_tlb_others(const struct cpumask *cpumask,
			     struct mm_struct *mm, unsigned long va)
{
	if (is_uv_system()) {
		unsigned int cpu;

		cpu = get_cpu();
		cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu);
		if (cpumask)
			flush_tlb_others_ipi(cpumask, mm, va);
		put_cpu();
		return;
	}
	flush_tlb_others_ipi(cpumask, mm, va);
}

static void __cpuinit calculate_tlb_offset(void)
{
	int cpu, node, nr_node_vecs;
	/*
	 * we are changing tlb_vector_offset for each CPU in runtime, but this
	 * will not cause inconsistency, as the write is atomic under X86. we
	 * might see more lock contentions in a short time, but after all CPU's
	 * tlb_vector_offset are changed, everything should go normal
	 *
	 * Note: if NUM_INVALIDATE_TLB_VECTORS % nr_online_nodes !=0, we might
	 * waste some vectors.
	 **/
	if (nr_online_nodes > NUM_INVALIDATE_TLB_VECTORS)
		nr_node_vecs = 1;
	else
		nr_node_vecs = NUM_INVALIDATE_TLB_VECTORS/nr_online_nodes;

	for_each_online_node(node) {
		int node_offset = (node % NUM_INVALIDATE_TLB_VECTORS) *
			nr_node_vecs;
		int cpu_offset = 0;
		for_each_cpu(cpu, cpumask_of_node(node)) {
			per_cpu(tlb_vector_offset, cpu) = node_offset +
				cpu_offset;
			cpu_offset++;
			cpu_offset = cpu_offset % nr_node_vecs;
		}
	}
}

static int __cpuinit tlb_cpuhp_notify(struct notifier_block *n,
		unsigned long action, void *hcpu)
{
	switch (action & 0xf) {
	case CPU_ONLINE:
	case CPU_DEAD:
		calculate_tlb_offset();
	}
	return NOTIFY_OK;
}

static int __cpuinit init_smp_flush(void)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(flush_state); i++)
		raw_spin_lock_init(&flush_state[i].tlbstate_lock);

	calculate_tlb_offset();
	hotcpu_notifier(tlb_cpuhp_notify, 0);
	return 0;
}
core_initcall(init_smp_flush);

void flush_tlb_current_task(void)
{
	struct mm_struct *mm = current->mm;

	preempt_disable();

	local_flush_tlb();
	if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
		flush_tlb_others(mm_cpumask(mm), mm, TLB_FLUSH_ALL);
	preempt_enable();
}

void flush_tlb_mm(struct mm_struct *mm)
{
	preempt_disable();

	if (current->active_mm == mm) {
		if (current->mm)
			local_flush_tlb();
		else
			leave_mm(smp_processor_id());
	}
	if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
		flush_tlb_others(mm_cpumask(mm), mm, TLB_FLUSH_ALL);

	preempt_enable();
}

void flush_tlb_page(struct vm_area_struct *vma, unsigned long va)
{
	struct mm_struct *mm = vma->vm_mm;

	preempt_disable();

	if (current->active_mm == mm) {
		if (current->mm)
			__flush_tlb_one(va);
		else
			leave_mm(smp_processor_id());
	}

	if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
		flush_tlb_others(mm_cpumask(mm), mm, va);

	preempt_enable();
}

static void do_flush_tlb_all(void *info)
{
	__flush_tlb_all();
	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY)
		leave_mm(smp_processor_id());
}

void flush_tlb_all(void)
{
	on_each_cpu(do_flush_tlb_all, NULL, 1);
}
Commit	Line	Data
c048fdfe GC	1	#include <linux/init.h>
	2
	3	#include <linux/mm.h>
c048fdfe GC	4	#include <linux/spinlock.h>
c048fdfe GC	5	#include <linux/smp.h>
c048fdfe	6	#include <linux/interrupt.h>
6dd01bed	7	#include <linux/module.h>
93296720	8	#include <linux/cpu.h>
c048fdfe	9
c048fdfe	10	#include <asm/tlbflush.h>
c048fdfe	11	#include <asm/mmu_context.h>
350f8f56	12	#include <asm/cache.h>
6dd01bed	13	#include <asm/apic.h>
bdbcdd48	14	#include <asm/uv/uv.h>
5af5573e	15
9eb912d1 BG	16	DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate)
	17	= { &init_mm, 0, };
	18
c048fdfe GC	19	/*
	20	* Smarter SMP flushing macros.
	21	* c/o Linus Torvalds.
	22	*
	23	* These mean you can really definitely utterly forget about
	24	* writing to user space from interrupts. (Its not allowed anyway).
	25	*
	26	* Optimizations Manfred Spraul <manfred@colorfullife.com>
	27	*
	28	* More scalable flush, from Andi Kleen
	29	*
	30	* To avoid global state use 8 different call vectors.
	31	* Each CPU uses a specific vector to trigger flushes on other
	32	* CPUs. Depending on the received vector the target CPUs look into
09b3ec73	33	* the right array slot for the flush data.
c048fdfe GC	34	*
	35	* With more than 8 CPUs they are hashed to the 8 available
	36	* vectors. The limited global vector space forces us to this right now.
	37	* In future when interrupts are split into per CPU domains this could be
	38	* fixed, at the cost of triggering multiple IPIs in some cases.
	39	*/
	40
	41	union smp_flush_state {
	42	struct {
c048fdfe GC	43	struct mm_struct *flush_mm;
c048fdfe GC	44	unsigned long flush_va;
39c662f6	45	raw_spinlock_t tlbstate_lock;
4595f962	46	DECLARE_BITMAP(flush_cpumask, NR_CPUS);
c048fdfe	47	};
350f8f56	48	char pad[INTERNODE_CACHE_BYTES];
09b3ec73	49	} ____cacheline_internodealigned_in_smp;
c048fdfe GC	50
	51	/* State is put into the per CPU data section, but padded
	52	to a full cache line because other CPUs can access it and we don't
	53	want false sharing in the per cpu data segment. */
09b3ec73	54	static union smp_flush_state flush_state[NUM_INVALIDATE_TLB_VECTORS];
c048fdfe	55
93296720 SL	56	static DEFINE_PER_CPU_READ_MOSTLY(int, tlb_vector_offset);
93296720 SL	57
c048fdfe GC	58	/*
	59	* We cannot call mmdrop() because we are in interrupt context,
	60	* instead update mm->cpu_vm_mask.
	61	*/
	62	void leave_mm(int cpu)
	63	{
9eb912d1	64	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK)
c048fdfe	65	BUG();
78f1c4d6 RR	66	cpumask_clear_cpu(cpu,
78f1c4d6 RR	67	mm_cpumask(percpu_read(cpu_tlbstate.active_mm)));
c048fdfe GC	68	load_cr3(swapper_pg_dir);
	69	}
	70	EXPORT_SYMBOL_GPL(leave_mm);
	71
	72	/*
	73	*
	74	* The flush IPI assumes that a thread switch happens in this order:
	75	* [cpu0: the cpu that switches]
	76	* 1) switch_mm() either 1a) or 1b)
	77	* 1a) thread switch to a different mm
	78	* 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask);
	79	* Stop ipi delivery for the old mm. This is not synchronized with
	80	* the other cpus, but smp_invalidate_interrupt ignore flush ipis
	81	* for the wrong mm, and in the worst case we perform a superfluous
	82	* tlb flush.
	83	* 1a2) set cpu mmu_state to TLBSTATE_OK
	84	* Now the smp_invalidate_interrupt won't call leave_mm if cpu0
	85	* was in lazy tlb mode.
	86	* 1a3) update cpu active_mm
	87	* Now cpu0 accepts tlb flushes for the new mm.
	88	* 1a4) cpu_set(cpu, new_mm->cpu_vm_mask);
	89	* Now the other cpus will send tlb flush ipis.
	90	* 1a4) change cr3.
	91	* 1b) thread switch without mm change
	92	* cpu active_mm is correct, cpu0 already handles
	93	* flush ipis.
	94	* 1b1) set cpu mmu_state to TLBSTATE_OK
	95	* 1b2) test_and_set the cpu bit in cpu_vm_mask.
	96	* Atomically set the bit [other cpus will start sending flush ipis],
	97	* and test the bit.
	98	* 1b3) if the bit was 0: leave_mm was called, flush the tlb.
	99	* 2) switch %%esp, ie current
	100	*
	101	* The interrupt must handle 2 special cases:
	102	* - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm.
	103	* - the cpu performs speculative tlb reads, i.e. even if the cpu only
	104	* runs in kernel space, the cpu could load tlb entries for user space
	105	* pages.
	106	*
	107	* The good news is that cpu mmu_state is local to each cpu, no
	108	* write/read ordering problems.
	109	*/
	110
	111	/*
	112	* TLB flush IPI:
	113	*
	114	* 1) Flush the tlb entries if the cpu uses the mm that's being flushed.
	115	* 2) Leave the mm if we are in the lazy tlb mode.
	116	*
	117	* Interrupts are disabled.
	118	*/
	119
02cf94c3 TH	120	/*
	121	* FIXME: use of asmlinkage is not consistent. On x86_64 it's noop
	122	* but still used for documentation purpose but the usage is slightly
	123	* inconsistent. On x86_32, asmlinkage is regparm(0) but interrupt
	124	* entry calls in with the first parameter in %eax. Maybe define
	125	* intrlinkage?
	126	*/
	127	#ifdef CONFIG_X86_64
	128	asmlinkage
	129	#endif
	130	void smp_invalidate_interrupt(struct pt_regs *regs)
c048fdfe	131	{
6dd01bed TH	132	unsigned int cpu;
6dd01bed TH	133	unsigned int sender;
c048fdfe GC	134	union smp_flush_state *f;
	135
	136	cpu = smp_processor_id();
	137	/*
	138	* orig_rax contains the negated interrupt vector.
	139	* Use that to determine where the sender put the data.
	140	*/
	141	sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START;
09b3ec73	142	f = &flush_state[sender];
c048fdfe	143
4595f962	144	if (!cpumask_test_cpu(cpu, to_cpumask(f->flush_cpumask)))
c048fdfe GC	145	goto out;
	146	/*
	147	* This was a BUG() but until someone can quote me the
	148	* line from the intel manual that guarantees an IPI to
	149	* multiple CPUs is retried _only_ on the erroring CPUs
	150	* its staying as a return
	151	*
	152	* BUG();
	153	*/
	154
9eb912d1 BG	155	if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) {
9eb912d1 BG	156	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) {
c048fdfe GC	157	if (f->flush_va == TLB_FLUSH_ALL)
	158	local_flush_tlb();
	159	else
	160	__flush_tlb_one(f->flush_va);
	161	} else
	162	leave_mm(cpu);
	163	}
	164	out:
	165	ack_APIC_irq();
6dd01bed	166	smp_mb__before_clear_bit();
4595f962	167	cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask));
6dd01bed	168	smp_mb__after_clear_bit();
8ae93669	169	inc_irq_stat(irq_tlb_count);
c048fdfe GC	170	}
c048fdfe GC	171
4595f962 RR	172	static void flush_tlb_others_ipi(const struct cpumask *cpumask,
4595f962 RR	173	struct mm_struct *mm, unsigned long va)
c048fdfe	174	{
6dd01bed	175	unsigned int sender;
c048fdfe	176	union smp_flush_state *f;
1812924b	177
c048fdfe	178	/* Caller has disabled preemption */
93296720	179	sender = this_cpu_read(tlb_vector_offset);
09b3ec73	180	f = &flush_state[sender];
c048fdfe GC	181
	182	/*
	183	* Could avoid this lock when
	184	* num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is
	185	* probably not worth checking this for a cache-hot lock.
	186	*/
39c662f6	187	raw_spin_lock(&f->tlbstate_lock);
c048fdfe GC	188
	189	f->flush_mm = mm;
	190	f->flush_va = va;
b04e6373 LT	191	if (cpumask_andnot(to_cpumask(f->flush_cpumask), cpumask, cpumask_of(smp_processor_id()))) {
	192	/*
	193	* We have to send the IPI only to
	194	* CPUs affected.
	195	*/
	196	apic->send_IPI_mask(to_cpumask(f->flush_cpumask),
	197	INVALIDATE_TLB_VECTOR_START + sender);
c048fdfe	198
b04e6373 LT	199	while (!cpumask_empty(to_cpumask(f->flush_cpumask)))
	200	cpu_relax();
	201	}
c048fdfe GC	202
	203	f->flush_mm = NULL;
	204	f->flush_va = 0;
39c662f6	205	raw_spin_unlock(&f->tlbstate_lock);
c048fdfe GC	206	}
c048fdfe GC	207
4595f962 RR	208	void native_flush_tlb_others(const struct cpumask *cpumask,
	209	struct mm_struct *mm, unsigned long va)
	210	{
	211	if (is_uv_system()) {
bdbcdd48	212	unsigned int cpu;
0e21990a	213
bdbcdd48 TH	214	cpu = get_cpu();
	215	cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu);
	216	if (cpumask)
	217	flush_tlb_others_ipi(cpumask, mm, va);
	218	put_cpu();
0e21990a	219	return;
4595f962 RR	220	}
	221	flush_tlb_others_ipi(cpumask, mm, va);
	222	}
	223
93296720 SL	224	static void __cpuinit calculate_tlb_offset(void)
	225	{
	226	int cpu, node, nr_node_vecs;
	227	/*
	228	* we are changing tlb_vector_offset for each CPU in runtime, but this
	229	* will not cause inconsistency, as the write is atomic under X86. we
	230	* might see more lock contentions in a short time, but after all CPU's
	231	* tlb_vector_offset are changed, everything should go normal
	232	*
	233	* Note: if NUM_INVALIDATE_TLB_VECTORS % nr_online_nodes !=0, we might
	234	* waste some vectors.
	235	**/
	236	if (nr_online_nodes > NUM_INVALIDATE_TLB_VECTORS)
	237	nr_node_vecs = 1;
	238	else
	239	nr_node_vecs = NUM_INVALIDATE_TLB_VECTORS/nr_online_nodes;
	240
	241	for_each_online_node(node) {
	242	int node_offset = (node % NUM_INVALIDATE_TLB_VECTORS) *
	243	nr_node_vecs;
	244	int cpu_offset = 0;
	245	for_each_cpu(cpu, cpumask_of_node(node)) {
	246	per_cpu(tlb_vector_offset, cpu) = node_offset +
	247	cpu_offset;
	248	cpu_offset++;
	249	cpu_offset = cpu_offset % nr_node_vecs;
	250	}
	251	}
	252	}
	253
cf38d0ba	254	static int __cpuinit tlb_cpuhp_notify(struct notifier_block *n,
93296720 SL	255	unsigned long action, void *hcpu)
	256	{
	257	switch (action & 0xf) {
	258	case CPU_ONLINE:
	259	case CPU_DEAD:
	260	calculate_tlb_offset();
	261	}
	262	return NOTIFY_OK;
	263	}
	264
a4928cff	265	static int __cpuinit init_smp_flush(void)
c048fdfe GC	266	{
	267	int i;
	268
09b3ec73	269	for (i = 0; i < ARRAY_SIZE(flush_state); i++)
39c662f6	270	raw_spin_lock_init(&flush_state[i].tlbstate_lock);
7c04e64a	271
93296720 SL	272	calculate_tlb_offset();
93296720 SL	273	hotcpu_notifier(tlb_cpuhp_notify, 0);
c048fdfe GC	274	return 0;
	275	}
	276	core_initcall(init_smp_flush);
	277
	278	void flush_tlb_current_task(void)
	279	{
	280	struct mm_struct *mm = current->mm;
c048fdfe GC	281
c048fdfe GC	282	preempt_disable();
c048fdfe GC	283
c048fdfe GC	284	local_flush_tlb();
78f1c4d6 RR	285	if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
78f1c4d6 RR	286	flush_tlb_others(mm_cpumask(mm), mm, TLB_FLUSH_ALL);
c048fdfe GC	287	preempt_enable();
	288	}
	289
	290	void flush_tlb_mm(struct mm_struct *mm)
	291	{
c048fdfe	292	preempt_disable();
c048fdfe GC	293
	294	if (current->active_mm == mm) {
	295	if (current->mm)
	296	local_flush_tlb();
	297	else
	298	leave_mm(smp_processor_id());
	299	}
78f1c4d6 RR	300	if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
78f1c4d6 RR	301	flush_tlb_others(mm_cpumask(mm), mm, TLB_FLUSH_ALL);
c048fdfe GC	302
	303	preempt_enable();
	304	}
	305
	306	void flush_tlb_page(struct vm_area_struct *vma, unsigned long va)
	307	{
	308	struct mm_struct *mm = vma->vm_mm;
c048fdfe GC	309
c048fdfe GC	310	preempt_disable();
c048fdfe GC	311
	312	if (current->active_mm == mm) {
	313	if (current->mm)
	314	__flush_tlb_one(va);
	315	else
	316	leave_mm(smp_processor_id());
	317	}
	318
78f1c4d6 RR	319	if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
78f1c4d6 RR	320	flush_tlb_others(mm_cpumask(mm), mm, va);
c048fdfe GC	321
	322	preempt_enable();
	323	}
	324
	325	static void do_flush_tlb_all(void *info)
	326	{
c048fdfe	327	__flush_tlb_all();
9eb912d1	328	if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY)
3f8afb77	329	leave_mm(smp_processor_id());
c048fdfe GC	330	}
	331
	332	void flush_tlb_all(void)
	333	{
15c8b6c1	334	on_each_cpu(do_flush_tlb_all, NULL, 1);
c048fdfe	335	}