[thirdparty/kernel/linux.git] / mm / mmu_notifier.c

/*
 *  linux/mm/mmu_notifier.c
 *
 *  Copyright (C) 2008  Qumranet, Inc.
 *  Copyright (C) 2008  SGI
 *             Christoph Lameter <cl@linux.com>
 *
 *  This work is licensed under the terms of the GNU GPL, version 2. See
 *  the COPYING file in the top-level directory.
 */

#include <linux/rculist.h>
#include <linux/mmu_notifier.h>
#include <linux/export.h>
#include <linux/mm.h>
#include <linux/err.h>
#include <linux/srcu.h>
#include <linux/rcupdate.h>
#include <linux/sched.h>
#include <linux/sched/mm.h>
#include <linux/slab.h>

/* global SRCU for all MMs */
DEFINE_STATIC_SRCU(srcu);

/*
 * This function allows mmu_notifier::release callback to delay a call to
 * a function that will free appropriate resources. The function must be
 * quick and must not block.
 */
void mmu_notifier_call_srcu(struct rcu_head *rcu,
			    void (*func)(struct rcu_head *rcu))
{
	call_srcu(&srcu, rcu, func);
}
EXPORT_SYMBOL_GPL(mmu_notifier_call_srcu);

/*
 * This function can't run concurrently against mmu_notifier_register
 * because mm->mm_users > 0 during mmu_notifier_register and exit_mmap
 * runs with mm_users == 0. Other tasks may still invoke mmu notifiers
 * in parallel despite there being no task using this mm any more,
 * through the vmas outside of the exit_mmap context, such as with
 * vmtruncate. This serializes against mmu_notifier_unregister with
 * the mmu_notifier_mm->lock in addition to SRCU and it serializes
 * against the other mmu notifiers with SRCU. struct mmu_notifier_mm
 * can't go away from under us as exit_mmap holds an mm_count pin
 * itself.
 */
void __mmu_notifier_release(struct mm_struct *mm)
{
	struct mmu_notifier *mn;
	int id;

	/*
	 * SRCU here will block mmu_notifier_unregister until
	 * ->release returns.
	 */
	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist)
		/*
		 * If ->release runs before mmu_notifier_unregister it must be
		 * handled, as it's the only way for the driver to flush all
		 * existing sptes and stop the driver from establishing any more
		 * sptes before all the pages in the mm are freed.
		 */
		if (mn->ops->release)
			mn->ops->release(mn, mm);

	spin_lock(&mm->mmu_notifier_mm->lock);
	while (unlikely(!hlist_empty(&mm->mmu_notifier_mm->list))) {
		mn = hlist_entry(mm->mmu_notifier_mm->list.first,
				 struct mmu_notifier,
				 hlist);
		/*
		 * We arrived before mmu_notifier_unregister so
		 * mmu_notifier_unregister will do nothing other than to wait
		 * for ->release to finish and for mmu_notifier_unregister to
		 * return.
		 */
		hlist_del_init_rcu(&mn->hlist);
	}
	spin_unlock(&mm->mmu_notifier_mm->lock);
	srcu_read_unlock(&srcu, id);

	/*
	 * synchronize_srcu here prevents mmu_notifier_release from returning to
	 * exit_mmap (which would proceed with freeing all pages in the mm)
	 * until the ->release method returns, if it was invoked by
	 * mmu_notifier_unregister.
	 *
	 * The mmu_notifier_mm can't go away from under us because one mm_count
	 * is held by exit_mmap.
	 */
	synchronize_srcu(&srcu);
}

/*
 * If no young bitflag is supported by the hardware, ->clear_flush_young can
 * unmap the address and return 1 or 0 depending if the mapping previously
 * existed or not.
 */
int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
					unsigned long start,
					unsigned long end)
{
	struct mmu_notifier *mn;
	int young = 0, id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
		if (mn->ops->clear_flush_young)
			young |= mn->ops->clear_flush_young(mn, mm, start, end);
	}
	srcu_read_unlock(&srcu, id);

	return young;
}

int __mmu_notifier_clear_young(struct mm_struct *mm,
			       unsigned long start,
			       unsigned long end)
{
	struct mmu_notifier *mn;
	int young = 0, id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
		if (mn->ops->clear_young)
			young |= mn->ops->clear_young(mn, mm, start, end);
	}
	srcu_read_unlock(&srcu, id);

	return young;
}

int __mmu_notifier_test_young(struct mm_struct *mm,
			      unsigned long address)
{
	struct mmu_notifier *mn;
	int young = 0, id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
		if (mn->ops->test_young) {
			young = mn->ops->test_young(mn, mm, address);
			if (young)
				break;
		}
	}
	srcu_read_unlock(&srcu, id);

	return young;
}

void __mmu_notifier_change_pte(struct mm_struct *mm, unsigned long address,
			       pte_t pte)
{
	struct mmu_notifier *mn;
	int id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
		if (mn->ops->change_pte)
			mn->ops->change_pte(mn, mm, address, pte);
	}
	srcu_read_unlock(&srcu, id);
}

int __mmu_notifier_invalidate_range_start(struct mmu_notifier_range *range)
{
	struct mmu_notifier *mn;
	int ret = 0;
	int id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &range->mm->mmu_notifier_mm->list, hlist) {
		if (mn->ops->invalidate_range_start) {
			int _ret = mn->ops->invalidate_range_start(mn, range);
			if (_ret) {
				pr_info("%pS callback failed with %d in %sblockable context.\n",
					mn->ops->invalidate_range_start, _ret,
					!range->blockable ? "non-" : "");
				ret = _ret;
			}
		}
	}
	srcu_read_unlock(&srcu, id);

	return ret;
}
EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_start);

void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range,
					 bool only_end)
{
	struct mmu_notifier *mn;
	int id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &range->mm->mmu_notifier_mm->list, hlist) {
		/*
		 * Call invalidate_range here too to avoid the need for the
		 * subsystem of having to register an invalidate_range_end
		 * call-back when there is invalidate_range already. Usually a
		 * subsystem registers either invalidate_range_start()/end() or
		 * invalidate_range(), so this will be no additional overhead
		 * (besides the pointer check).
		 *
		 * We skip call to invalidate_range() if we know it is safe ie
		 * call site use mmu_notifier_invalidate_range_only_end() which
		 * is safe to do when we know that a call to invalidate_range()
		 * already happen under page table lock.
		 */
		if (!only_end && mn->ops->invalidate_range)
			mn->ops->invalidate_range(mn, range->mm,
						  range->start,
						  range->end);
		if (mn->ops->invalidate_range_end)
			mn->ops->invalidate_range_end(mn, range);
	}
	srcu_read_unlock(&srcu, id);
}
EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_end);

void __mmu_notifier_invalidate_range(struct mm_struct *mm,
				  unsigned long start, unsigned long end)
{
	struct mmu_notifier *mn;
	int id;

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
		if (mn->ops->invalidate_range)
			mn->ops->invalidate_range(mn, mm, start, end);
	}
	srcu_read_unlock(&srcu, id);
}
EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range);

static int do_mmu_notifier_register(struct mmu_notifier *mn,
				    struct mm_struct *mm,
				    int take_mmap_sem)
{
	struct mmu_notifier_mm *mmu_notifier_mm;
	int ret;

	BUG_ON(atomic_read(&mm->mm_users) <= 0);

	ret = -ENOMEM;
	mmu_notifier_mm = kmalloc(sizeof(struct mmu_notifier_mm), GFP_KERNEL);
	if (unlikely(!mmu_notifier_mm))
		goto out;

	if (take_mmap_sem)
		down_write(&mm->mmap_sem);
	ret = mm_take_all_locks(mm);
	if (unlikely(ret))
		goto out_clean;

	if (!mm_has_notifiers(mm)) {
		INIT_HLIST_HEAD(&mmu_notifier_mm->list);
		spin_lock_init(&mmu_notifier_mm->lock);

		mm->mmu_notifier_mm = mmu_notifier_mm;
		mmu_notifier_mm = NULL;
	}
	mmgrab(mm);

	/*
	 * Serialize the update against mmu_notifier_unregister. A
	 * side note: mmu_notifier_release can't run concurrently with
	 * us because we hold the mm_users pin (either implicitly as
	 * current->mm or explicitly with get_task_mm() or similar).
	 * We can't race against any other mmu notifier method either
	 * thanks to mm_take_all_locks().
	 */
	spin_lock(&mm->mmu_notifier_mm->lock);
	hlist_add_head(&mn->hlist, &mm->mmu_notifier_mm->list);
	spin_unlock(&mm->mmu_notifier_mm->lock);

	mm_drop_all_locks(mm);
out_clean:
	if (take_mmap_sem)
		up_write(&mm->mmap_sem);
	kfree(mmu_notifier_mm);
out:
	BUG_ON(atomic_read(&mm->mm_users) <= 0);
	return ret;
}

/*
 * Must not hold mmap_sem nor any other VM related lock when calling
 * this registration function. Must also ensure mm_users can't go down
 * to zero while this runs to avoid races with mmu_notifier_release,
 * so mm has to be current->mm or the mm should be pinned safely such
 * as with get_task_mm(). If the mm is not current->mm, the mm_users
 * pin should be released by calling mmput after mmu_notifier_register
 * returns. mmu_notifier_unregister must be always called to
 * unregister the notifier. mm_count is automatically pinned to allow
 * mmu_notifier_unregister to safely run at any time later, before or
 * after exit_mmap. ->release will always be called before exit_mmap
 * frees the pages.
 */
int mmu_notifier_register(struct mmu_notifier *mn, struct mm_struct *mm)
{
	return do_mmu_notifier_register(mn, mm, 1);
}
EXPORT_SYMBOL_GPL(mmu_notifier_register);

/*
 * Same as mmu_notifier_register but here the caller must hold the
 * mmap_sem in write mode.
 */
int __mmu_notifier_register(struct mmu_notifier *mn, struct mm_struct *mm)
{
	return do_mmu_notifier_register(mn, mm, 0);
}
EXPORT_SYMBOL_GPL(__mmu_notifier_register);

/* this is called after the last mmu_notifier_unregister() returned */
void __mmu_notifier_mm_destroy(struct mm_struct *mm)
{
	BUG_ON(!hlist_empty(&mm->mmu_notifier_mm->list));
	kfree(mm->mmu_notifier_mm);
	mm->mmu_notifier_mm = LIST_POISON1; /* debug */
}

/*
 * This releases the mm_count pin automatically and frees the mm
 * structure if it was the last user of it. It serializes against
 * running mmu notifiers with SRCU and against mmu_notifier_unregister
 * with the unregister lock + SRCU. All sptes must be dropped before
 * calling mmu_notifier_unregister. ->release or any other notifier
 * method may be invoked concurrently with mmu_notifier_unregister,
 * and only after mmu_notifier_unregister returned we're guaranteed
 * that ->release or any other method can't run anymore.
 */
void mmu_notifier_unregister(struct mmu_notifier *mn, struct mm_struct *mm)
{
	BUG_ON(atomic_read(&mm->mm_count) <= 0);

	if (!hlist_unhashed(&mn->hlist)) {
		/*
		 * SRCU here will force exit_mmap to wait for ->release to
		 * finish before freeing the pages.
		 */
		int id;

		id = srcu_read_lock(&srcu);
		/*
		 * exit_mmap will block in mmu_notifier_release to guarantee
		 * that ->release is called before freeing the pages.
		 */
		if (mn->ops->release)
			mn->ops->release(mn, mm);
		srcu_read_unlock(&srcu, id);

		spin_lock(&mm->mmu_notifier_mm->lock);
		/*
		 * Can not use list_del_rcu() since __mmu_notifier_release
		 * can delete it before we hold the lock.
		 */
		hlist_del_init_rcu(&mn->hlist);
		spin_unlock(&mm->mmu_notifier_mm->lock);
	}

	/*
	 * Wait for any running method to finish, of course including
	 * ->release if it was run by mmu_notifier_release instead of us.
	 */
	synchronize_srcu(&srcu);

	BUG_ON(atomic_read(&mm->mm_count) <= 0);

	mmdrop(mm);
}
EXPORT_SYMBOL_GPL(mmu_notifier_unregister);

/*
 * Same as mmu_notifier_unregister but no callback and no srcu synchronization.
 */
void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
					struct mm_struct *mm)
{
	spin_lock(&mm->mmu_notifier_mm->lock);
	/*
	 * Can not use list_del_rcu() since __mmu_notifier_release
	 * can delete it before we hold the lock.
	 */
	hlist_del_init_rcu(&mn->hlist);
	spin_unlock(&mm->mmu_notifier_mm->lock);

	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	mmdrop(mm);
}
EXPORT_SYMBOL_GPL(mmu_notifier_unregister_no_release);
Commit	Line	Data
cddb8a5c AA	1	/*
	2	* linux/mm/mmu_notifier.c
	3	*
	4	* Copyright (C) 2008 Qumranet, Inc.
	5	* Copyright (C) 2008 SGI
93e205a7	6	* Christoph Lameter <cl@linux.com>
cddb8a5c AA	7	*
	8	* This work is licensed under the terms of the GNU GPL, version 2. See
	9	* the COPYING file in the top-level directory.
	10	*/
	11
	12	#include <linux/rculist.h>
	13	#include <linux/mmu_notifier.h>
b95f1b31	14	#include <linux/export.h>
cddb8a5c AA	15	#include <linux/mm.h>
cddb8a5c AA	16	#include <linux/err.h>
21a92735	17	#include <linux/srcu.h>
cddb8a5c AA	18	#include <linux/rcupdate.h>
cddb8a5c AA	19	#include <linux/sched.h>
6e84f315	20	#include <linux/sched/mm.h>
5a0e3ad6	21	#include <linux/slab.h>
cddb8a5c	22
21a92735	23	/* global SRCU for all MMs */
dde8da6c	24	DEFINE_STATIC_SRCU(srcu);
21a92735	25
b972216e PZ	26	/*
	27	* This function allows mmu_notifier::release callback to delay a call to
	28	* a function that will free appropriate resources. The function must be
	29	* quick and must not block.
	30	*/
	31	void mmu_notifier_call_srcu(struct rcu_head *rcu,
	32	void (func)(struct rcu_head rcu))
	33	{
	34	call_srcu(&srcu, rcu, func);
	35	}
	36	EXPORT_SYMBOL_GPL(mmu_notifier_call_srcu);
	37
cddb8a5c AA	38	/*
	39	* This function can't run concurrently against mmu_notifier_register
	40	* because mm->mm_users > 0 during mmu_notifier_register and exit_mmap
	41	* runs with mm_users == 0. Other tasks may still invoke mmu notifiers
	42	* in parallel despite there being no task using this mm any more,
	43	* through the vmas outside of the exit_mmap context, such as with
	44	* vmtruncate. This serializes against mmu_notifier_unregister with
21a92735 SG	45	* the mmu_notifier_mm->lock in addition to SRCU and it serializes
21a92735 SG	46	* against the other mmu notifiers with SRCU. struct mmu_notifier_mm
cddb8a5c AA	47	* can't go away from under us as exit_mmap holds an mm_count pin
	48	* itself.
	49	*/
	50	void __mmu_notifier_release(struct mm_struct *mm)
	51	{
	52	struct mmu_notifier *mn;
21a92735	53	int id;
3ad3d901 XG	54
3ad3d901 XG	55	/*
d34883d4 XG	56	* SRCU here will block mmu_notifier_unregister until
d34883d4 XG	57	* ->release returns.
3ad3d901	58	*/
21a92735	59	id = srcu_read_lock(&srcu);
d34883d4 XG	60	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist)
	61	/*
	62	* If ->release runs before mmu_notifier_unregister it must be
	63	* handled, as it's the only way for the driver to flush all
	64	* existing sptes and stop the driver from establishing any more
	65	* sptes before all the pages in the mm are freed.
	66	*/
	67	if (mn->ops->release)
	68	mn->ops->release(mn, mm);
d34883d4	69
cddb8a5c AA	70	spin_lock(&mm->mmu_notifier_mm->lock);
	71	while (unlikely(!hlist_empty(&mm->mmu_notifier_mm->list))) {
	72	mn = hlist_entry(mm->mmu_notifier_mm->list.first,
	73	struct mmu_notifier,
	74	hlist);
	75	/*
d34883d4 XG	76	* We arrived before mmu_notifier_unregister so
	77	* mmu_notifier_unregister will do nothing other than to wait
	78	* for ->release to finish and for mmu_notifier_unregister to
	79	* return.
cddb8a5c AA	80	*/
cddb8a5c AA	81	hlist_del_init_rcu(&mn->hlist);
cddb8a5c AA	82	}
cddb8a5c AA	83	spin_unlock(&mm->mmu_notifier_mm->lock);
b972216e	84	srcu_read_unlock(&srcu, id);
cddb8a5c AA	85
cddb8a5c AA	86	/*
d34883d4 XG	87	* synchronize_srcu here prevents mmu_notifier_release from returning to
	88	* exit_mmap (which would proceed with freeing all pages in the mm)
	89	* until the ->release method returns, if it was invoked by
	90	* mmu_notifier_unregister.
	91	*
	92	* The mmu_notifier_mm can't go away from under us because one mm_count
	93	* is held by exit_mmap.
cddb8a5c	94	*/
21a92735	95	synchronize_srcu(&srcu);
cddb8a5c AA	96	}
	97
	98	/*
	99	* If no young bitflag is supported by the hardware, ->clear_flush_young can
	100	* unmap the address and return 1 or 0 depending if the mapping previously
	101	* existed or not.
	102	*/
	103	int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
57128468 ALC	104	unsigned long start,
57128468 ALC	105	unsigned long end)
cddb8a5c AA	106	{
cddb8a5c AA	107	struct mmu_notifier *mn;
21a92735	108	int young = 0, id;
cddb8a5c	109
21a92735	110	id = srcu_read_lock(&srcu);
b67bfe0d	111	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
cddb8a5c	112	if (mn->ops->clear_flush_young)
57128468	113	young \|= mn->ops->clear_flush_young(mn, mm, start, end);
cddb8a5c	114	}
21a92735	115	srcu_read_unlock(&srcu, id);
cddb8a5c AA	116
	117	return young;
	118	}
	119
1d7715c6 VD	120	int __mmu_notifier_clear_young(struct mm_struct *mm,
	121	unsigned long start,
	122	unsigned long end)
	123	{
	124	struct mmu_notifier *mn;
	125	int young = 0, id;
	126
	127	id = srcu_read_lock(&srcu);
	128	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
	129	if (mn->ops->clear_young)
	130	young \|= mn->ops->clear_young(mn, mm, start, end);
	131	}
	132	srcu_read_unlock(&srcu, id);
	133
	134	return young;
	135	}
	136
8ee53820 AA	137	int __mmu_notifier_test_young(struct mm_struct *mm,
	138	unsigned long address)
	139	{
	140	struct mmu_notifier *mn;
21a92735	141	int young = 0, id;
8ee53820	142
21a92735	143	id = srcu_read_lock(&srcu);
b67bfe0d	144	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
8ee53820 AA	145	if (mn->ops->test_young) {
	146	young = mn->ops->test_young(mn, mm, address);
	147	if (young)
	148	break;
	149	}
	150	}
21a92735	151	srcu_read_unlock(&srcu, id);
8ee53820 AA	152
	153	return young;
	154	}
	155
828502d3 IE	156	void __mmu_notifier_change_pte(struct mm_struct *mm, unsigned long address,
	157	pte_t pte)
	158	{
	159	struct mmu_notifier *mn;
21a92735	160	int id;
828502d3	161
21a92735	162	id = srcu_read_lock(&srcu);
b67bfe0d	163	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
828502d3 IE	164	if (mn->ops->change_pte)
828502d3 IE	165	mn->ops->change_pte(mn, mm, address, pte);
828502d3	166	}
21a92735	167	srcu_read_unlock(&srcu, id);
828502d3 IE	168	}
828502d3 IE	169
ac46d4f3	170	int __mmu_notifier_invalidate_range_start(struct mmu_notifier_range *range)
cddb8a5c AA	171	{
cddb8a5c AA	172	struct mmu_notifier *mn;
93065ac7	173	int ret = 0;
21a92735	174	int id;
cddb8a5c	175
21a92735	176	id = srcu_read_lock(&srcu);
ac46d4f3	177	hlist_for_each_entry_rcu(mn, &range->mm->mmu_notifier_mm->list, hlist) {
93065ac7	178	if (mn->ops->invalidate_range_start) {
5d6527a7	179	int _ret = mn->ops->invalidate_range_start(mn, range);
93065ac7 MH	180	if (_ret) {
93065ac7 MH	181	pr_info("%pS callback failed with %d in %sblockable context.\n",
ac46d4f3 JG	182	mn->ops->invalidate_range_start, _ret,
ac46d4f3 JG	183	!range->blockable ? "non-" : "");
93065ac7 MH	184	ret = _ret;
	185	}
	186	}
cddb8a5c	187	}
21a92735	188	srcu_read_unlock(&srcu, id);
93065ac7 MH	189
93065ac7 MH	190	return ret;
cddb8a5c	191	}
fa794199	192	EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_start);
cddb8a5c	193
ac46d4f3	194	void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range,
4645b9fe	195	bool only_end)
cddb8a5c AA	196	{
cddb8a5c AA	197	struct mmu_notifier *mn;
21a92735	198	int id;
cddb8a5c	199
21a92735	200	id = srcu_read_lock(&srcu);
ac46d4f3	201	hlist_for_each_entry_rcu(mn, &range->mm->mmu_notifier_mm->list, hlist) {
0f0a327f JR	202	/*
	203	* Call invalidate_range here too to avoid the need for the
	204	* subsystem of having to register an invalidate_range_end
	205	* call-back when there is invalidate_range already. Usually a
	206	* subsystem registers either invalidate_range_start()/end() or
	207	* invalidate_range(), so this will be no additional overhead
	208	* (besides the pointer check).
4645b9fe JG	209	*
	210	* We skip call to invalidate_range() if we know it is safe ie
	211	* call site use mmu_notifier_invalidate_range_only_end() which
	212	* is safe to do when we know that a call to invalidate_range()
	213	* already happen under page table lock.
0f0a327f	214	*/
4645b9fe	215	if (!only_end && mn->ops->invalidate_range)
ac46d4f3 JG	216	mn->ops->invalidate_range(mn, range->mm,
	217	range->start,
	218	range->end);
cddb8a5c	219	if (mn->ops->invalidate_range_end)
5d6527a7	220	mn->ops->invalidate_range_end(mn, range);
cddb8a5c	221	}
21a92735	222	srcu_read_unlock(&srcu, id);
cddb8a5c	223	}
fa794199	224	EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_end);
cddb8a5c	225
0f0a327f JR	226	void __mmu_notifier_invalidate_range(struct mm_struct *mm,
	227	unsigned long start, unsigned long end)
	228	{
	229	struct mmu_notifier *mn;
	230	int id;
	231
	232	id = srcu_read_lock(&srcu);
	233	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
	234	if (mn->ops->invalidate_range)
	235	mn->ops->invalidate_range(mn, mm, start, end);
	236	}
	237	srcu_read_unlock(&srcu, id);
	238	}
	239	EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range);
	240
cddb8a5c AA	241	static int do_mmu_notifier_register(struct mmu_notifier *mn,
	242	struct mm_struct *mm,
	243	int take_mmap_sem)
	244	{
	245	struct mmu_notifier_mm *mmu_notifier_mm;
	246	int ret;
	247
	248	BUG_ON(atomic_read(&mm->mm_users) <= 0);
	249
35cfa2b0 GS	250	ret = -ENOMEM;
	251	mmu_notifier_mm = kmalloc(sizeof(struct mmu_notifier_mm), GFP_KERNEL);
	252	if (unlikely(!mmu_notifier_mm))
	253	goto out;
	254
cddb8a5c AA	255	if (take_mmap_sem)
	256	down_write(&mm->mmap_sem);
	257	ret = mm_take_all_locks(mm);
	258	if (unlikely(ret))
35cfa2b0	259	goto out_clean;
cddb8a5c AA	260
	261	if (!mm_has_notifiers(mm)) {
	262	INIT_HLIST_HEAD(&mmu_notifier_mm->list);
	263	spin_lock_init(&mmu_notifier_mm->lock);
e0f3c3f7	264
cddb8a5c	265	mm->mmu_notifier_mm = mmu_notifier_mm;
35cfa2b0	266	mmu_notifier_mm = NULL;
cddb8a5c	267	}
f1f10076	268	mmgrab(mm);
cddb8a5c AA	269
	270	/*
	271	* Serialize the update against mmu_notifier_unregister. A
	272	* side note: mmu_notifier_release can't run concurrently with
	273	* us because we hold the mm_users pin (either implicitly as
	274	* current->mm or explicitly with get_task_mm() or similar).
	275	* We can't race against any other mmu notifier method either
	276	* thanks to mm_take_all_locks().
	277	*/
	278	spin_lock(&mm->mmu_notifier_mm->lock);
	279	hlist_add_head(&mn->hlist, &mm->mmu_notifier_mm->list);
	280	spin_unlock(&mm->mmu_notifier_mm->lock);
	281
	282	mm_drop_all_locks(mm);
35cfa2b0	283	out_clean:
cddb8a5c AA	284	if (take_mmap_sem)
cddb8a5c AA	285	up_write(&mm->mmap_sem);
35cfa2b0 GS	286	kfree(mmu_notifier_mm);
35cfa2b0 GS	287	out:
cddb8a5c AA	288	BUG_ON(atomic_read(&mm->mm_users) <= 0);
	289	return ret;
	290	}
	291
	292	/*
	293	* Must not hold mmap_sem nor any other VM related lock when calling
	294	* this registration function. Must also ensure mm_users can't go down
	295	* to zero while this runs to avoid races with mmu_notifier_release,
	296	* so mm has to be current->mm or the mm should be pinned safely such
	297	* as with get_task_mm(). If the mm is not current->mm, the mm_users
	298	* pin should be released by calling mmput after mmu_notifier_register
	299	* returns. mmu_notifier_unregister must be always called to
	300	* unregister the notifier. mm_count is automatically pinned to allow
	301	* mmu_notifier_unregister to safely run at any time later, before or
	302	* after exit_mmap. ->release will always be called before exit_mmap
	303	* frees the pages.
	304	*/
	305	int mmu_notifier_register(struct mmu_notifier mn, struct mm_struct mm)
	306	{
	307	return do_mmu_notifier_register(mn, mm, 1);
	308	}
	309	EXPORT_SYMBOL_GPL(mmu_notifier_register);
	310
	311	/*
	312	* Same as mmu_notifier_register but here the caller must hold the
	313	* mmap_sem in write mode.
	314	*/
	315	int __mmu_notifier_register(struct mmu_notifier mn, struct mm_struct mm)
	316	{
	317	return do_mmu_notifier_register(mn, mm, 0);
	318	}
	319	EXPORT_SYMBOL_GPL(__mmu_notifier_register);
	320
	321	/* this is called after the last mmu_notifier_unregister() returned */
	322	void __mmu_notifier_mm_destroy(struct mm_struct *mm)
	323	{
	324	BUG_ON(!hlist_empty(&mm->mmu_notifier_mm->list));
	325	kfree(mm->mmu_notifier_mm);
	326	mm->mmu_notifier_mm = LIST_POISON1; /* debug */
	327	}
	328
	329	/*
	330	* This releases the mm_count pin automatically and frees the mm
	331	* structure if it was the last user of it. It serializes against
21a92735 SG	332	* running mmu notifiers with SRCU and against mmu_notifier_unregister
21a92735 SG	333	* with the unregister lock + SRCU. All sptes must be dropped before
cddb8a5c AA	334	* calling mmu_notifier_unregister. ->release or any other notifier
	335	* method may be invoked concurrently with mmu_notifier_unregister,
	336	* and only after mmu_notifier_unregister returned we're guaranteed
	337	* that ->release or any other method can't run anymore.
	338	*/
	339	void mmu_notifier_unregister(struct mmu_notifier mn, struct mm_struct mm)
	340	{
	341	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	342
cddb8a5c	343	if (!hlist_unhashed(&mn->hlist)) {
d34883d4 XG	344	/*
	345	* SRCU here will force exit_mmap to wait for ->release to
	346	* finish before freeing the pages.
	347	*/
21a92735	348	int id;
3ad3d901	349
d34883d4	350	id = srcu_read_lock(&srcu);
cddb8a5c	351	/*
d34883d4 XG	352	* exit_mmap will block in mmu_notifier_release to guarantee
d34883d4 XG	353	* that ->release is called before freeing the pages.
cddb8a5c AA	354	*/
	355	if (mn->ops->release)
	356	mn->ops->release(mn, mm);
d34883d4	357	srcu_read_unlock(&srcu, id);
3ad3d901	358
d34883d4	359	spin_lock(&mm->mmu_notifier_mm->lock);
751efd86	360	/*
d34883d4 XG	361	* Can not use list_del_rcu() since __mmu_notifier_release
d34883d4 XG	362	* can delete it before we hold the lock.
751efd86	363	*/
d34883d4	364	hlist_del_init_rcu(&mn->hlist);
cddb8a5c	365	spin_unlock(&mm->mmu_notifier_mm->lock);
d34883d4	366	}
cddb8a5c AA	367
cddb8a5c AA	368	/*
d34883d4	369	* Wait for any running method to finish, of course including
83a35e36	370	* ->release if it was run by mmu_notifier_release instead of us.
cddb8a5c	371	*/
21a92735	372	synchronize_srcu(&srcu);
cddb8a5c AA	373
	374	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	375
	376	mmdrop(mm);
	377	}
	378	EXPORT_SYMBOL_GPL(mmu_notifier_unregister);
21a92735	379
b972216e PZ	380	/*
	381	* Same as mmu_notifier_unregister but no callback and no srcu synchronization.
	382	*/
	383	void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
	384	struct mm_struct *mm)
	385	{
	386	spin_lock(&mm->mmu_notifier_mm->lock);
	387	/*
	388	* Can not use list_del_rcu() since __mmu_notifier_release
	389	* can delete it before we hold the lock.
	390	*/
	391	hlist_del_init_rcu(&mn->hlist);
	392	spin_unlock(&mm->mmu_notifier_mm->lock);
	393
	394	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	395	mmdrop(mm);
	396	}
	397	EXPORT_SYMBOL_GPL(mmu_notifier_unregister_no_release);