[linux.git] / virt / kvm / async_pf.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * kvm asynchronous fault support
 *
 * Copyright 2010 Red Hat, Inc.
 *
 * Author:
 *      Gleb Natapov <[email protected]>
 */

#include <linux/kvm_host.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/mmu_context.h>
#include <linux/sched/mm.h>

#include "async_pf.h"
#include <trace/events/kvm.h>

static struct kmem_cache *async_pf_cache;

int kvm_async_pf_init(void)
{
	async_pf_cache = KMEM_CACHE(kvm_async_pf, 0);

	if (!async_pf_cache)
		return -ENOMEM;

	return 0;
}

void kvm_async_pf_deinit(void)
{
	kmem_cache_destroy(async_pf_cache);
	async_pf_cache = NULL;
}

void kvm_async_pf_vcpu_init(struct kvm_vcpu *vcpu)
{
	INIT_LIST_HEAD(&vcpu->async_pf.done);
	INIT_LIST_HEAD(&vcpu->async_pf.queue);
	spin_lock_init(&vcpu->async_pf.lock);
}

static void async_pf_execute(struct work_struct *work)
{
	struct kvm_async_pf *apf =
		container_of(work, struct kvm_async_pf, work);
	struct kvm_vcpu *vcpu = apf->vcpu;
	struct mm_struct *mm = vcpu->kvm->mm;
	unsigned long addr = apf->addr;
	gpa_t cr2_or_gpa = apf->cr2_or_gpa;
	int locked = 1;
	bool first;

	might_sleep();

	/*
	 * Attempt to pin the VM's host address space, and simply skip gup() if
	 * acquiring a pin fail, i.e. if the process is exiting.  Note, KVM
	 * holds a reference to its associated mm_struct until the very end of
	 * kvm_destroy_vm(), i.e. the struct itself won't be freed before this
	 * work item is fully processed.
	 */
	if (mmget_not_zero(mm)) {
		mmap_read_lock(mm);
		get_user_pages_remote(mm, addr, 1, FOLL_WRITE, NULL, &locked);
		if (locked)
			mmap_read_unlock(mm);
		mmput(mm);
	}

	/*
	 * Notify and kick the vCPU even if faulting in the page failed, e.g.
	 * so that the vCPU can retry the fault synchronously.
	 */
	if (IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC))
		kvm_arch_async_page_present(vcpu, apf);

	spin_lock(&vcpu->async_pf.lock);
	first = list_empty(&vcpu->async_pf.done);
	list_add_tail(&apf->link, &vcpu->async_pf.done);
	apf->vcpu = NULL;
	spin_unlock(&vcpu->async_pf.lock);

	if (!IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC) && first)
		kvm_arch_async_page_present_queued(vcpu);

	/*
	 * apf may be freed by kvm_check_async_pf_completion() after
	 * this point
	 */

	trace_kvm_async_pf_completed(addr, cr2_or_gpa);

	__kvm_vcpu_wake_up(vcpu);
}

static void kvm_flush_and_free_async_pf_work(struct kvm_async_pf *work)
{
	/*
	 * The async #PF is "done", but KVM must wait for the work item itself,
	 * i.e. async_pf_execute(), to run to completion.  If KVM is a module,
	 * KVM must ensure *no* code owned by the KVM (the module) can be run
	 * after the last call to module_put().  Note, flushing the work item
	 * is always required when the item is taken off the completion queue.
	 * E.g. even if the vCPU handles the item in the "normal" path, the VM
	 * could be terminated before async_pf_execute() completes.
	 *
	 * Wake all events skip the queue and go straight done, i.e. don't
	 * need to be flushed (but sanity check that the work wasn't queued).
	 */
	if (work->wakeup_all)
		WARN_ON_ONCE(work->work.func);
	else
		flush_work(&work->work);
	kmem_cache_free(async_pf_cache, work);
}

void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu)
{
	spin_lock(&vcpu->async_pf.lock);

	/* cancel outstanding work queue item */
	while (!list_empty(&vcpu->async_pf.queue)) {
		struct kvm_async_pf *work =
			list_first_entry(&vcpu->async_pf.queue,
					 typeof(*work), queue);
		list_del(&work->queue);

		/*
		 * We know it's present in vcpu->async_pf.done, do
		 * nothing here.
		 */
		if (!work->vcpu)
			continue;

		spin_unlock(&vcpu->async_pf.lock);
#ifdef CONFIG_KVM_ASYNC_PF_SYNC
		flush_work(&work->work);
#else
		if (cancel_work_sync(&work->work))
			kmem_cache_free(async_pf_cache, work);
#endif
		spin_lock(&vcpu->async_pf.lock);
	}

	while (!list_empty(&vcpu->async_pf.done)) {
		struct kvm_async_pf *work =
			list_first_entry(&vcpu->async_pf.done,
					 typeof(*work), link);
		list_del(&work->link);

		spin_unlock(&vcpu->async_pf.lock);
		kvm_flush_and_free_async_pf_work(work);
		spin_lock(&vcpu->async_pf.lock);
	}
	spin_unlock(&vcpu->async_pf.lock);

	vcpu->async_pf.queued = 0;
}

void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu)
{
	struct kvm_async_pf *work;

	while (!list_empty_careful(&vcpu->async_pf.done) &&
	      kvm_arch_can_dequeue_async_page_present(vcpu)) {
		spin_lock(&vcpu->async_pf.lock);
		work = list_first_entry(&vcpu->async_pf.done, typeof(*work),
					      link);
		list_del(&work->link);
		spin_unlock(&vcpu->async_pf.lock);

		kvm_arch_async_page_ready(vcpu, work);
		if (!IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC))
			kvm_arch_async_page_present(vcpu, work);

		list_del(&work->queue);
		vcpu->async_pf.queued--;
		kvm_flush_and_free_async_pf_work(work);
	}
}

/*
 * Try to schedule a job to handle page fault asynchronously. Returns 'true' on
 * success, 'false' on failure (page fault has to be handled synchronously).
 */
bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
			unsigned long hva, struct kvm_arch_async_pf *arch)
{
	struct kvm_async_pf *work;

	if (vcpu->async_pf.queued >= ASYNC_PF_PER_VCPU)
		return false;

	/* Arch specific code should not do async PF in this case */
	if (unlikely(kvm_is_error_hva(hva)))
		return false;

	/*
	 * do alloc nowait since if we are going to sleep anyway we
	 * may as well sleep faulting in page
	 */
	work = kmem_cache_zalloc(async_pf_cache, GFP_NOWAIT | __GFP_NOWARN);
	if (!work)
		return false;

	work->wakeup_all = false;
	work->vcpu = vcpu;
	work->cr2_or_gpa = cr2_or_gpa;
	work->addr = hva;
	work->arch = *arch;

	INIT_WORK(&work->work, async_pf_execute);

	list_add_tail(&work->queue, &vcpu->async_pf.queue);
	vcpu->async_pf.queued++;
	work->notpresent_injected = kvm_arch_async_page_not_present(vcpu, work);

	schedule_work(&work->work);

	return true;
}

int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu)
{
	struct kvm_async_pf *work;
	bool first;

	if (!list_empty_careful(&vcpu->async_pf.done))
		return 0;

	work = kmem_cache_zalloc(async_pf_cache, GFP_ATOMIC);
	if (!work)
		return -ENOMEM;

	work->wakeup_all = true;
	INIT_LIST_HEAD(&work->queue); /* for list_del to work */

	spin_lock(&vcpu->async_pf.lock);
	first = list_empty(&vcpu->async_pf.done);
	list_add_tail(&work->link, &vcpu->async_pf.done);
	spin_unlock(&vcpu->async_pf.lock);

	if (!IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC) && first)
		kvm_arch_async_page_present_queued(vcpu);

	vcpu->async_pf.queued++;
	return 0;
}
Commit	Line	Data
775c8a3d	1	// SPDX-License-Identifier: GPL-2.0-only
af585b92 GN	2	/*
	3	* kvm asynchronous fault support
	4	*
	5	* Copyright 2010 Red Hat, Inc.
	6	*
	7	* Author:
	8	* Gleb Natapov <[email protected]>
af585b92 GN	9	*/
	10
	11	#include <linux/kvm_host.h>
	12	#include <linux/slab.h>
	13	#include <linux/module.h>
	14	#include <linux/mmu_context.h>
6e84f315	15	#include <linux/sched/mm.h>
af585b92 GN	16
	17	#include "async_pf.h"
	18	#include <trace/events/kvm.h>
	19
	20	static struct kmem_cache *async_pf_cache;
	21
	22	int kvm_async_pf_init(void)
	23	{
	24	async_pf_cache = KMEM_CACHE(kvm_async_pf, 0);
	25
	26	if (!async_pf_cache)
	27	return -ENOMEM;
	28
	29	return 0;
	30	}
	31
	32	void kvm_async_pf_deinit(void)
	33	{
4f52696a	34	kmem_cache_destroy(async_pf_cache);
af585b92 GN	35	async_pf_cache = NULL;
	36	}
	37
	38	void kvm_async_pf_vcpu_init(struct kvm_vcpu *vcpu)
	39	{
	40	INIT_LIST_HEAD(&vcpu->async_pf.done);
	41	INIT_LIST_HEAD(&vcpu->async_pf.queue);
	42	spin_lock_init(&vcpu->async_pf.lock);
	43	}
	44
	45	static void async_pf_execute(struct work_struct *work)
	46	{
af585b92 GN	47	struct kvm_async_pf *apf =
af585b92 GN	48	container_of(work, struct kvm_async_pf, work);
af585b92	49	struct kvm_vcpu *vcpu = apf->vcpu;
8284765f	50	struct mm_struct *mm = vcpu->kvm->mm;
af585b92	51	unsigned long addr = apf->addr;
736c291c	52	gpa_t cr2_or_gpa = apf->cr2_or_gpa;
8b7457ef	53	int locked = 1;
557a961a	54	bool first;
af585b92 GN	55
	56	might_sleep();
	57
1e987790	58	/*
8284765f SC	59	* Attempt to pin the VM's host address space, and simply skip gup() if
	60	* acquiring a pin fail, i.e. if the process is exiting. Note, KVM
	61	* holds a reference to its associated mm_struct until the very end of
	62	* kvm_destroy_vm(), i.e. the struct itself won't be freed before this
	63	* work item is fully processed.
1e987790	64	*/
8284765f SC	65	if (mmget_not_zero(mm)) {
	66	mmap_read_lock(mm);
	67	get_user_pages_remote(mm, addr, 1, FOLL_WRITE, NULL, &locked);
	68	if (locked)
	69	mmap_read_unlock(mm);
	70	mmput(mm);
	71	}
1e987790	72
8284765f SC	73	/*
	74	* Notify and kick the vCPU even if faulting in the page failed, e.g.
	75	* so that the vCPU can retry the fault synchronously.
	76	*/
4425f567 PB	77	if (IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC))
4425f567 PB	78	kvm_arch_async_page_present(vcpu, apf);
af585b92 GN	79
af585b92 GN	80	spin_lock(&vcpu->async_pf.lock);
557a961a	81	first = list_empty(&vcpu->async_pf.done);
af585b92	82	list_add_tail(&apf->link, &vcpu->async_pf.done);
22583f0d	83	apf->vcpu = NULL;
af585b92 GN	84	spin_unlock(&vcpu->async_pf.lock);
af585b92 GN	85
557a961a VK	86	if (!IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC) && first)
	87	kvm_arch_async_page_present_queued(vcpu);
	88
af585b92 GN	89	/*
	90	* apf may be freed by kvm_check_async_pf_completion() after
	91	* this point
	92	*/
	93
736c291c	94	trace_kvm_async_pf_completed(addr, cr2_or_gpa);
af585b92	95
d92a5d1c	96	__kvm_vcpu_wake_up(vcpu);
3d75b8aa SC	97	}
	98
	99	static void kvm_flush_and_free_async_pf_work(struct kvm_async_pf *work)
	100	{
	101	/*
	102	* The async #PF is "done", but KVM must wait for the work item itself,
	103	* i.e. async_pf_execute(), to run to completion. If KVM is a module,
	104	* KVM must ensure no code owned by the KVM (the module) can be run
	105	* after the last call to module_put(). Note, flushing the work item
	106	* is always required when the item is taken off the completion queue.
	107	* E.g. even if the vCPU handles the item in the "normal" path, the VM
	108	* could be terminated before async_pf_execute() completes.
	109	*
	110	* Wake all events skip the queue and go straight done, i.e. don't
	111	* need to be flushed (but sanity check that the work wasn't queued).
	112	*/
	113	if (work->wakeup_all)
	114	WARN_ON_ONCE(work->work.func);
	115	else
	116	flush_work(&work->work);
	117	kmem_cache_free(async_pf_cache, work);
af585b92 GN	118	}
	119
	120	void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu)
	121	{
22583f0d PB	122	spin_lock(&vcpu->async_pf.lock);
22583f0d PB	123
af585b92 GN	124	/* cancel outstanding work queue item */
	125	while (!list_empty(&vcpu->async_pf.queue)) {
	126	struct kvm_async_pf *work =
433da860 GT	127	list_first_entry(&vcpu->async_pf.queue,
433da860 GT	128	typeof(*work), queue);
af585b92	129	list_del(&work->queue);
9f2ceda4	130
22583f0d PB	131	/*
	132	* We know it's present in vcpu->async_pf.done, do
	133	* nothing here.
	134	*/
	135	if (!work->vcpu)
	136	continue;
	137
	138	spin_unlock(&vcpu->async_pf.lock);
9f2ceda4 DD	139	#ifdef CONFIG_KVM_ASYNC_PF_SYNC
	140	flush_work(&work->work);
	141	#else
8284765f	142	if (cancel_work_sync(&work->work))
af585b92	143	kmem_cache_free(async_pf_cache, work);
9f2ceda4	144	#endif
22583f0d	145	spin_lock(&vcpu->async_pf.lock);
af585b92 GN	146	}
af585b92 GN	147
af585b92 GN	148	while (!list_empty(&vcpu->async_pf.done)) {
af585b92 GN	149	struct kvm_async_pf *work =
433da860 GT	150	list_first_entry(&vcpu->async_pf.done,
433da860 GT	151	typeof(*work), link);
af585b92	152	list_del(&work->link);
3d75b8aa SC	153
	154	spin_unlock(&vcpu->async_pf.lock);
	155	kvm_flush_and_free_async_pf_work(work);
	156	spin_lock(&vcpu->async_pf.lock);
af585b92 GN	157	}
	158	spin_unlock(&vcpu->async_pf.lock);
	159
	160	vcpu->async_pf.queued = 0;
	161	}
	162
	163	void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu)
	164	{
	165	struct kvm_async_pf *work;
	166
15096ffc	167	while (!list_empty_careful(&vcpu->async_pf.done) &&
7c0ade6c	168	kvm_arch_can_dequeue_async_page_present(vcpu)) {
15096ffc XG	169	spin_lock(&vcpu->async_pf.lock);
	170	work = list_first_entry(&vcpu->async_pf.done, typeof(*work),
	171	link);
	172	list_del(&work->link);
	173	spin_unlock(&vcpu->async_pf.lock);
af585b92	174
f2e10669	175	kvm_arch_async_page_ready(vcpu, work);
4425f567 PB	176	if (!IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC))
4425f567 PB	177	kvm_arch_async_page_present(vcpu, work);
af585b92	178
15096ffc XG	179	list_del(&work->queue);
15096ffc XG	180	vcpu->async_pf.queued--;
3d75b8aa	181	kvm_flush_and_free_async_pf_work(work);
15096ffc	182	}
af585b92 GN	183	}
af585b92 GN	184
e8c22266 VK	185	/*
	186	* Try to schedule a job to handle page fault asynchronously. Returns 'true' on
	187	* success, 'false' on failure (page fault has to be handled synchronously).
	188	*/
	189	bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
	190	unsigned long hva, struct kvm_arch_async_pf *arch)
af585b92 GN	191	{
	192	struct kvm_async_pf *work;
	193
	194	if (vcpu->async_pf.queued >= ASYNC_PF_PER_VCPU)
e8c22266	195	return false;
af585b92	196
7863e346 VK	197	/* Arch specific code should not do async PF in this case */
7863e346 VK	198	if (unlikely(kvm_is_error_hva(hva)))
e8c22266	199	return false;
af585b92 GN	200
	201	/*
	202	* do alloc nowait since if we are going to sleep anyway we
	203	* may as well sleep faulting in page
	204	*/
d7444794	205	work = kmem_cache_zalloc(async_pf_cache, GFP_NOWAIT \| __GFP_NOWARN);
af585b92	206	if (!work)
e8c22266	207	return false;
af585b92	208
f2e10669	209	work->wakeup_all = false;
af585b92	210	work->vcpu = vcpu;
736c291c	211	work->cr2_or_gpa = cr2_or_gpa;
e0ead41a	212	work->addr = hva;
af585b92	213	work->arch = *arch;
af585b92	214
af585b92	215	INIT_WORK(&work->work, async_pf_execute);
af585b92 GN	216
	217	list_add_tail(&work->queue, &vcpu->async_pf.queue);
	218	vcpu->async_pf.queued++;
2a18b7e7	219	work->notpresent_injected = kvm_arch_async_page_not_present(vcpu, work);
7863e346 VK	220
	221	schedule_work(&work->work);
	222
e8c22266	223	return true;
af585b92	224	}
344d9588 GN	225
	226	int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu)
	227	{
	228	struct kvm_async_pf *work;
557a961a	229	bool first;
344d9588	230
64f638c7	231	if (!list_empty_careful(&vcpu->async_pf.done))
344d9588 GN	232	return 0;
	233
	234	work = kmem_cache_zalloc(async_pf_cache, GFP_ATOMIC);
	235	if (!work)
	236	return -ENOMEM;
	237
f2e10669	238	work->wakeup_all = true;
344d9588 GN	239	INIT_LIST_HEAD(&work->queue); /* for list_del to work */
344d9588 GN	240
64f638c7	241	spin_lock(&vcpu->async_pf.lock);
557a961a	242	first = list_empty(&vcpu->async_pf.done);
344d9588	243	list_add_tail(&work->link, &vcpu->async_pf.done);
64f638c7 XG	244	spin_unlock(&vcpu->async_pf.lock);
64f638c7 XG	245
557a961a VK	246	if (!IS_ENABLED(CONFIG_KVM_ASYNC_PF_SYNC) && first)
	247	kvm_arch_async_page_present_queued(vcpu);
	248
344d9588 GN	249	vcpu->async_pf.queued++;
	250	return 0;
	251	}