[linux.git] / mm / mmu_notifier.c

/*
 *  linux/mm/mmu_notifier.c
 *
 *  Copyright (C) 2008  Qumranet, Inc.
 *  Copyright (C) 2008  SGI
 *             Christoph Lameter <[email protected]>
 *
 *  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);

void mmu_notifier_synchronize(void)
{
	/* Wait for any running method to finish. */
	srcu_barrier(&srcu);
}
EXPORT_SYMBOL_GPL(mmu_notifier_synchronize);

/*
 * 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);
}

void __mmu_notifier_invalidate_range_start(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_start)
			mn->ops->invalidate_range_start(mn, mm, start, end);
	}
	srcu_read_unlock(&srcu, id);
}
EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_start);

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

	id = srcu_read_lock(&srcu);
	hlist_for_each_entry_rcu(mn, &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, mm, start, end);
		if (mn->ops->invalidate_range_end)
			mn->ops->invalidate_range_end(mn, mm, start, end);
	}
	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);

/*
 * Must be called while holding mm->mmap_sem for either read or write.
 * The result is guaranteed to be valid until mm->mmap_sem is dropped.
 */
bool mm_has_blockable_invalidate_notifiers(struct mm_struct *mm)
{
	struct mmu_notifier *mn;
	int id;
	bool ret = false;

	WARN_ON_ONCE(!rwsem_is_locked(&mm->mmap_sem));

	if (!mm_has_notifiers(mm))
		return ret;

	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_start &&
		    !mn->ops->invalidate_range_end)
				continue;

		if (!(mn->ops->flags & MMU_INVALIDATE_DOES_NOT_BLOCK)) {
			ret = true;
			break;
		}
	}
	srcu_read_unlock(&srcu, id);
	return ret;
}

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 <[email protected]>
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
	38	void mmu_notifier_synchronize(void)
	39	{
	40	/* Wait for any running method to finish. */
	41	srcu_barrier(&srcu);
	42	}
	43	EXPORT_SYMBOL_GPL(mmu_notifier_synchronize);
	44
cddb8a5c AA	45	/*
	46	* This function can't run concurrently against mmu_notifier_register
	47	* because mm->mm_users > 0 during mmu_notifier_register and exit_mmap
	48	* runs with mm_users == 0. Other tasks may still invoke mmu notifiers
	49	* in parallel despite there being no task using this mm any more,
	50	* through the vmas outside of the exit_mmap context, such as with
	51	* vmtruncate. This serializes against mmu_notifier_unregister with
21a92735 SG	52	* the mmu_notifier_mm->lock in addition to SRCU and it serializes
21a92735 SG	53	* against the other mmu notifiers with SRCU. struct mmu_notifier_mm
cddb8a5c AA	54	* can't go away from under us as exit_mmap holds an mm_count pin
	55	* itself.
	56	*/
	57	void __mmu_notifier_release(struct mm_struct *mm)
	58	{
	59	struct mmu_notifier *mn;
21a92735	60	int id;
3ad3d901 XG	61
3ad3d901 XG	62	/*
d34883d4 XG	63	* SRCU here will block mmu_notifier_unregister until
d34883d4 XG	64	* ->release returns.
3ad3d901	65	*/
21a92735	66	id = srcu_read_lock(&srcu);
d34883d4 XG	67	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist)
	68	/*
	69	* If ->release runs before mmu_notifier_unregister it must be
	70	* handled, as it's the only way for the driver to flush all
	71	* existing sptes and stop the driver from establishing any more
	72	* sptes before all the pages in the mm are freed.
	73	*/
	74	if (mn->ops->release)
	75	mn->ops->release(mn, mm);
d34883d4	76
cddb8a5c AA	77	spin_lock(&mm->mmu_notifier_mm->lock);
	78	while (unlikely(!hlist_empty(&mm->mmu_notifier_mm->list))) {
	79	mn = hlist_entry(mm->mmu_notifier_mm->list.first,
	80	struct mmu_notifier,
	81	hlist);
	82	/*
d34883d4 XG	83	* We arrived before mmu_notifier_unregister so
	84	* mmu_notifier_unregister will do nothing other than to wait
	85	* for ->release to finish and for mmu_notifier_unregister to
	86	* return.
cddb8a5c AA	87	*/
cddb8a5c AA	88	hlist_del_init_rcu(&mn->hlist);
cddb8a5c AA	89	}
cddb8a5c AA	90	spin_unlock(&mm->mmu_notifier_mm->lock);
b972216e	91	srcu_read_unlock(&srcu, id);
cddb8a5c AA	92
cddb8a5c AA	93	/*
d34883d4 XG	94	* synchronize_srcu here prevents mmu_notifier_release from returning to
	95	* exit_mmap (which would proceed with freeing all pages in the mm)
	96	* until the ->release method returns, if it was invoked by
	97	* mmu_notifier_unregister.
	98	*
	99	* The mmu_notifier_mm can't go away from under us because one mm_count
	100	* is held by exit_mmap.
cddb8a5c	101	*/
21a92735	102	synchronize_srcu(&srcu);
cddb8a5c AA	103	}
	104
	105	/*
	106	* If no young bitflag is supported by the hardware, ->clear_flush_young can
	107	* unmap the address and return 1 or 0 depending if the mapping previously
	108	* existed or not.
	109	*/
	110	int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
57128468 ALC	111	unsigned long start,
57128468 ALC	112	unsigned long end)
cddb8a5c AA	113	{
cddb8a5c AA	114	struct mmu_notifier *mn;
21a92735	115	int young = 0, id;
cddb8a5c	116
21a92735	117	id = srcu_read_lock(&srcu);
b67bfe0d	118	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
cddb8a5c	119	if (mn->ops->clear_flush_young)
57128468	120	young \|= mn->ops->clear_flush_young(mn, mm, start, end);
cddb8a5c	121	}
21a92735	122	srcu_read_unlock(&srcu, id);
cddb8a5c AA	123
	124	return young;
	125	}
	126
1d7715c6 VD	127	int __mmu_notifier_clear_young(struct mm_struct *mm,
	128	unsigned long start,
	129	unsigned long end)
	130	{
	131	struct mmu_notifier *mn;
	132	int young = 0, id;
	133
	134	id = srcu_read_lock(&srcu);
	135	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
	136	if (mn->ops->clear_young)
	137	young \|= mn->ops->clear_young(mn, mm, start, end);
	138	}
	139	srcu_read_unlock(&srcu, id);
	140
	141	return young;
	142	}
	143
8ee53820 AA	144	int __mmu_notifier_test_young(struct mm_struct *mm,
	145	unsigned long address)
	146	{
	147	struct mmu_notifier *mn;
21a92735	148	int young = 0, id;
8ee53820	149
21a92735	150	id = srcu_read_lock(&srcu);
b67bfe0d	151	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
8ee53820 AA	152	if (mn->ops->test_young) {
	153	young = mn->ops->test_young(mn, mm, address);
	154	if (young)
	155	break;
	156	}
	157	}
21a92735	158	srcu_read_unlock(&srcu, id);
8ee53820 AA	159
	160	return young;
	161	}
	162
828502d3 IE	163	void __mmu_notifier_change_pte(struct mm_struct *mm, unsigned long address,
	164	pte_t pte)
	165	{
	166	struct mmu_notifier *mn;
21a92735	167	int id;
828502d3	168
21a92735	169	id = srcu_read_lock(&srcu);
b67bfe0d	170	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
828502d3 IE	171	if (mn->ops->change_pte)
828502d3 IE	172	mn->ops->change_pte(mn, mm, address, pte);
828502d3	173	}
21a92735	174	srcu_read_unlock(&srcu, id);
828502d3 IE	175	}
828502d3 IE	176
cddb8a5c AA	177	void __mmu_notifier_invalidate_range_start(struct mm_struct *mm,
	178	unsigned long start, unsigned long end)
	179	{
	180	struct mmu_notifier *mn;
21a92735	181	int id;
cddb8a5c	182
21a92735	183	id = srcu_read_lock(&srcu);
b67bfe0d	184	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
cddb8a5c AA	185	if (mn->ops->invalidate_range_start)
	186	mn->ops->invalidate_range_start(mn, mm, start, end);
	187	}
21a92735	188	srcu_read_unlock(&srcu, id);
cddb8a5c	189	}
fa794199	190	EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_start);
cddb8a5c AA	191
cddb8a5c AA	192	void __mmu_notifier_invalidate_range_end(struct mm_struct *mm,
4645b9fe JG	193	unsigned long start,
	194	unsigned long end,
	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);
b67bfe0d	201	hlist_for_each_entry_rcu(mn, &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)
0f0a327f	216	mn->ops->invalidate_range(mn, mm, start, end);
cddb8a5c AA	217	if (mn->ops->invalidate_range_end)
	218	mn->ops->invalidate_range_end(mn, mm, start, end);
	219	}
21a92735	220	srcu_read_unlock(&srcu, id);
cddb8a5c	221	}
fa794199	222	EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range_end);
cddb8a5c	223
0f0a327f JR	224	void __mmu_notifier_invalidate_range(struct mm_struct *mm,
	225	unsigned long start, unsigned long end)
	226	{
	227	struct mmu_notifier *mn;
	228	int id;
	229
	230	id = srcu_read_lock(&srcu);
	231	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
	232	if (mn->ops->invalidate_range)
	233	mn->ops->invalidate_range(mn, mm, start, end);
	234	}
	235	srcu_read_unlock(&srcu, id);
	236	}
	237	EXPORT_SYMBOL_GPL(__mmu_notifier_invalidate_range);
	238
5ff7091f DR	239	/*
	240	* Must be called while holding mm->mmap_sem for either read or write.
	241	* The result is guaranteed to be valid until mm->mmap_sem is dropped.
	242	*/
	243	bool mm_has_blockable_invalidate_notifiers(struct mm_struct *mm)
	244	{
	245	struct mmu_notifier *mn;
	246	int id;
	247	bool ret = false;
	248
	249	WARN_ON_ONCE(!rwsem_is_locked(&mm->mmap_sem));
	250
	251	if (!mm_has_notifiers(mm))
	252	return ret;
	253
	254	id = srcu_read_lock(&srcu);
	255	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
	256	if (!mn->ops->invalidate_range &&
	257	!mn->ops->invalidate_range_start &&
	258	!mn->ops->invalidate_range_end)
	259	continue;
	260
	261	if (!(mn->ops->flags & MMU_INVALIDATE_DOES_NOT_BLOCK)) {
	262	ret = true;
	263	break;
	264	}
	265	}
	266	srcu_read_unlock(&srcu, id);
	267	return ret;
	268	}
	269
cddb8a5c AA	270	static int do_mmu_notifier_register(struct mmu_notifier *mn,
	271	struct mm_struct *mm,
	272	int take_mmap_sem)
	273	{
	274	struct mmu_notifier_mm *mmu_notifier_mm;
	275	int ret;
	276
	277	BUG_ON(atomic_read(&mm->mm_users) <= 0);
	278
35cfa2b0 GS	279	ret = -ENOMEM;
	280	mmu_notifier_mm = kmalloc(sizeof(struct mmu_notifier_mm), GFP_KERNEL);
	281	if (unlikely(!mmu_notifier_mm))
	282	goto out;
	283
cddb8a5c AA	284	if (take_mmap_sem)
	285	down_write(&mm->mmap_sem);
	286	ret = mm_take_all_locks(mm);
	287	if (unlikely(ret))
35cfa2b0	288	goto out_clean;
cddb8a5c AA	289
	290	if (!mm_has_notifiers(mm)) {
	291	INIT_HLIST_HEAD(&mmu_notifier_mm->list);
	292	spin_lock_init(&mmu_notifier_mm->lock);
e0f3c3f7	293
cddb8a5c	294	mm->mmu_notifier_mm = mmu_notifier_mm;
35cfa2b0	295	mmu_notifier_mm = NULL;
cddb8a5c	296	}
f1f10076	297	mmgrab(mm);
cddb8a5c AA	298
	299	/*
	300	* Serialize the update against mmu_notifier_unregister. A
	301	* side note: mmu_notifier_release can't run concurrently with
	302	* us because we hold the mm_users pin (either implicitly as
	303	* current->mm or explicitly with get_task_mm() or similar).
	304	* We can't race against any other mmu notifier method either
	305	* thanks to mm_take_all_locks().
	306	*/
	307	spin_lock(&mm->mmu_notifier_mm->lock);
	308	hlist_add_head(&mn->hlist, &mm->mmu_notifier_mm->list);
	309	spin_unlock(&mm->mmu_notifier_mm->lock);
	310
	311	mm_drop_all_locks(mm);
35cfa2b0	312	out_clean:
cddb8a5c AA	313	if (take_mmap_sem)
cddb8a5c AA	314	up_write(&mm->mmap_sem);
35cfa2b0 GS	315	kfree(mmu_notifier_mm);
35cfa2b0 GS	316	out:
cddb8a5c AA	317	BUG_ON(atomic_read(&mm->mm_users) <= 0);
	318	return ret;
	319	}
	320
	321	/*
	322	* Must not hold mmap_sem nor any other VM related lock when calling
	323	* this registration function. Must also ensure mm_users can't go down
	324	* to zero while this runs to avoid races with mmu_notifier_release,
	325	* so mm has to be current->mm or the mm should be pinned safely such
	326	* as with get_task_mm(). If the mm is not current->mm, the mm_users
	327	* pin should be released by calling mmput after mmu_notifier_register
	328	* returns. mmu_notifier_unregister must be always called to
	329	* unregister the notifier. mm_count is automatically pinned to allow
	330	* mmu_notifier_unregister to safely run at any time later, before or
	331	* after exit_mmap. ->release will always be called before exit_mmap
	332	* frees the pages.
	333	*/
	334	int mmu_notifier_register(struct mmu_notifier mn, struct mm_struct mm)
	335	{
	336	return do_mmu_notifier_register(mn, mm, 1);
	337	}
	338	EXPORT_SYMBOL_GPL(mmu_notifier_register);
	339
	340	/*
	341	* Same as mmu_notifier_register but here the caller must hold the
	342	* mmap_sem in write mode.
	343	*/
	344	int __mmu_notifier_register(struct mmu_notifier mn, struct mm_struct mm)
	345	{
	346	return do_mmu_notifier_register(mn, mm, 0);
	347	}
	348	EXPORT_SYMBOL_GPL(__mmu_notifier_register);
	349
	350	/* this is called after the last mmu_notifier_unregister() returned */
	351	void __mmu_notifier_mm_destroy(struct mm_struct *mm)
	352	{
	353	BUG_ON(!hlist_empty(&mm->mmu_notifier_mm->list));
	354	kfree(mm->mmu_notifier_mm);
	355	mm->mmu_notifier_mm = LIST_POISON1; /* debug */
	356	}
	357
	358	/*
	359	* This releases the mm_count pin automatically and frees the mm
	360	* structure if it was the last user of it. It serializes against
21a92735 SG	361	* running mmu notifiers with SRCU and against mmu_notifier_unregister
21a92735 SG	362	* with the unregister lock + SRCU. All sptes must be dropped before
cddb8a5c AA	363	* calling mmu_notifier_unregister. ->release or any other notifier
	364	* method may be invoked concurrently with mmu_notifier_unregister,
	365	* and only after mmu_notifier_unregister returned we're guaranteed
	366	* that ->release or any other method can't run anymore.
	367	*/
	368	void mmu_notifier_unregister(struct mmu_notifier mn, struct mm_struct mm)
	369	{
	370	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	371
cddb8a5c	372	if (!hlist_unhashed(&mn->hlist)) {
d34883d4 XG	373	/*
	374	* SRCU here will force exit_mmap to wait for ->release to
	375	* finish before freeing the pages.
	376	*/
21a92735	377	int id;
3ad3d901	378
d34883d4	379	id = srcu_read_lock(&srcu);
cddb8a5c	380	/*
d34883d4 XG	381	* exit_mmap will block in mmu_notifier_release to guarantee
d34883d4 XG	382	* that ->release is called before freeing the pages.
cddb8a5c AA	383	*/
	384	if (mn->ops->release)
	385	mn->ops->release(mn, mm);
d34883d4	386	srcu_read_unlock(&srcu, id);
3ad3d901	387
d34883d4	388	spin_lock(&mm->mmu_notifier_mm->lock);
751efd86	389	/*
d34883d4 XG	390	* Can not use list_del_rcu() since __mmu_notifier_release
d34883d4 XG	391	* can delete it before we hold the lock.
751efd86	392	*/
d34883d4	393	hlist_del_init_rcu(&mn->hlist);
cddb8a5c	394	spin_unlock(&mm->mmu_notifier_mm->lock);
d34883d4	395	}
cddb8a5c AA	396
cddb8a5c AA	397	/*
d34883d4	398	* Wait for any running method to finish, of course including
83a35e36	399	* ->release if it was run by mmu_notifier_release instead of us.
cddb8a5c	400	*/
21a92735	401	synchronize_srcu(&srcu);
cddb8a5c AA	402
	403	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	404
	405	mmdrop(mm);
	406	}
	407	EXPORT_SYMBOL_GPL(mmu_notifier_unregister);
21a92735	408
b972216e PZ	409	/*
	410	* Same as mmu_notifier_unregister but no callback and no srcu synchronization.
	411	*/
	412	void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
	413	struct mm_struct *mm)
	414	{
	415	spin_lock(&mm->mmu_notifier_mm->lock);
	416	/*
	417	* Can not use list_del_rcu() since __mmu_notifier_release
	418	* can delete it before we hold the lock.
	419	*/
	420	hlist_del_init_rcu(&mn->hlist);
	421	spin_unlock(&mm->mmu_notifier_mm->lock);
	422
	423	BUG_ON(atomic_read(&mm->mm_count) <= 0);
	424	mmdrop(mm);
	425	}
	426	EXPORT_SYMBOL_GPL(mmu_notifier_unregister_no_release);