[deliverable/linux.git] / virt / kvm / iommu.c

/*
 * Copyright (c) 2006, Intel Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
 * Place - Suite 330, Boston, MA 02111-1307 USA.
 *
 * Copyright (C) 2006-2008 Intel Corporation
 * Copyright IBM Corporation, 2008
 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
 *
 * Author: Allen M. Kay <allen.m.kay@intel.com>
 * Author: Weidong Han <weidong.han@intel.com>
 * Author: Ben-Ami Yassour <benami@il.ibm.com>
 */

#include <linux/list.h>
#include <linux/kvm_host.h>
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/stat.h>
#include <linux/dmar.h>
#include <linux/iommu.h>
#include <linux/intel-iommu.h>

static bool allow_unsafe_assigned_interrupts;
module_param_named(allow_unsafe_assigned_interrupts,
		   allow_unsafe_assigned_interrupts, bool, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(allow_unsafe_assigned_interrupts,
 "Enable device assignment on platforms without interrupt remapping support.");

static int kvm_iommu_unmap_memslots(struct kvm *kvm);
static void kvm_iommu_put_pages(struct kvm *kvm,
				gfn_t base_gfn, unsigned long npages);

static pfn_t kvm_pin_pages(struct kvm_memory_slot *slot, gfn_t gfn,
			   unsigned long size)
{
	gfn_t end_gfn;
	pfn_t pfn;

	pfn     = gfn_to_pfn_memslot(slot, gfn);
	end_gfn = gfn + (size >> PAGE_SHIFT);
	gfn    += 1;

	if (is_error_noslot_pfn(pfn))
		return pfn;

	while (gfn < end_gfn)
		gfn_to_pfn_memslot(slot, gfn++);

	return pfn;
}

int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
{
	gfn_t gfn, end_gfn;
	pfn_t pfn;
	int r = 0;
	struct iommu_domain *domain = kvm->arch.iommu_domain;
	int flags;

	/* check if iommu exists and in use */
	if (!domain)
		return 0;

	gfn     = slot->base_gfn;
	end_gfn = gfn + slot->npages;

	flags = IOMMU_READ | IOMMU_WRITE;
	if (kvm->arch.iommu_flags & KVM_IOMMU_CACHE_COHERENCY)
		flags |= IOMMU_CACHE;


	while (gfn < end_gfn) {
		unsigned long page_size;

		/* Check if already mapped */
		if (iommu_iova_to_phys(domain, gfn_to_gpa(gfn))) {
			gfn += 1;
			continue;
		}

		/* Get the page size we could use to map */
		page_size = kvm_host_page_size(kvm, gfn);

		/* Make sure the page_size does not exceed the memslot */
		while ((gfn + (page_size >> PAGE_SHIFT)) > end_gfn)
			page_size >>= 1;

		/* Make sure gfn is aligned to the page size we want to map */
		while ((gfn << PAGE_SHIFT) & (page_size - 1))
			page_size >>= 1;

		/*
		 * Pin all pages we are about to map in memory. This is
		 * important because we unmap and unpin in 4kb steps later.
		 */
		pfn = kvm_pin_pages(slot, gfn, page_size);
		if (is_error_noslot_pfn(pfn)) {
			gfn += 1;
			continue;
		}

		/* Map into IO address space */
		r = iommu_map(domain, gfn_to_gpa(gfn), pfn_to_hpa(pfn),
			      page_size, flags);
		if (r) {
			printk(KERN_ERR "kvm_iommu_map_address:"
			       "iommu failed to map pfn=%llx\n", pfn);
			goto unmap_pages;
		}

		gfn += page_size >> PAGE_SHIFT;


	}

	return 0;

unmap_pages:
	kvm_iommu_put_pages(kvm, slot->base_gfn, gfn);
	return r;
}

static int kvm_iommu_map_memslots(struct kvm *kvm)
{
	int idx, r = 0;
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;

	idx = srcu_read_lock(&kvm->srcu);
	slots = kvm_memslots(kvm);

	kvm_for_each_memslot(memslot, slots) {
		r = kvm_iommu_map_pages(kvm, memslot);
		if (r)
			break;
	}
	srcu_read_unlock(&kvm->srcu, idx);

	return r;
}

int kvm_assign_device(struct kvm *kvm,
		      struct kvm_assigned_dev_kernel *assigned_dev)
{
	struct pci_dev *pdev = NULL;
	struct iommu_domain *domain = kvm->arch.iommu_domain;
	int r, last_flags;

	/* check if iommu exists and in use */
	if (!domain)
		return 0;

	pdev = assigned_dev->dev;
	if (pdev == NULL)
		return -ENODEV;

	r = iommu_attach_device(domain, &pdev->dev);
	if (r) {
		dev_err(&pdev->dev, "kvm assign device failed ret %d", r);
		return r;
	}

	last_flags = kvm->arch.iommu_flags;
	if (iommu_domain_has_cap(kvm->arch.iommu_domain,
				 IOMMU_CAP_CACHE_COHERENCY))
		kvm->arch.iommu_flags |= KVM_IOMMU_CACHE_COHERENCY;

	/* Check if need to update IOMMU page table for guest memory */
	if ((last_flags ^ kvm->arch.iommu_flags) ==
			KVM_IOMMU_CACHE_COHERENCY) {
		kvm_iommu_unmap_memslots(kvm);
		r = kvm_iommu_map_memslots(kvm);
		if (r)
			goto out_unmap;
	}

	pdev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;

	printk(KERN_DEBUG "assign device %x:%x:%x.%x\n",
		assigned_dev->host_segnr,
		assigned_dev->host_busnr,
		PCI_SLOT(assigned_dev->host_devfn),
		PCI_FUNC(assigned_dev->host_devfn));

	return 0;
out_unmap:
	kvm_iommu_unmap_memslots(kvm);
	return r;
}

int kvm_deassign_device(struct kvm *kvm,
			struct kvm_assigned_dev_kernel *assigned_dev)
{
	struct iommu_domain *domain = kvm->arch.iommu_domain;
	struct pci_dev *pdev = NULL;

	/* check if iommu exists and in use */
	if (!domain)
		return 0;

	pdev = assigned_dev->dev;
	if (pdev == NULL)
		return -ENODEV;

	iommu_detach_device(domain, &pdev->dev);

	pdev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;

	printk(KERN_DEBUG "deassign device %x:%x:%x.%x\n",
		assigned_dev->host_segnr,
		assigned_dev->host_busnr,
		PCI_SLOT(assigned_dev->host_devfn),
		PCI_FUNC(assigned_dev->host_devfn));

	return 0;
}

int kvm_iommu_map_guest(struct kvm *kvm)
{
	int r;

	if (!iommu_present(&pci_bus_type)) {
		printk(KERN_ERR "%s: iommu not found\n", __func__);
		return -ENODEV;
	}

	mutex_lock(&kvm->slots_lock);

	kvm->arch.iommu_domain = iommu_domain_alloc(&pci_bus_type);
	if (!kvm->arch.iommu_domain) {
		r = -ENOMEM;
		goto out_unlock;
	}

	if (!allow_unsafe_assigned_interrupts &&
	    !iommu_domain_has_cap(kvm->arch.iommu_domain,
				  IOMMU_CAP_INTR_REMAP)) {
		printk(KERN_WARNING "%s: No interrupt remapping support,"
		       " disallowing device assignment."
		       " Re-enble with \"allow_unsafe_assigned_interrupts=1\""
		       " module option.\n", __func__);
		iommu_domain_free(kvm->arch.iommu_domain);
		kvm->arch.iommu_domain = NULL;
		r = -EPERM;
		goto out_unlock;
	}

	r = kvm_iommu_map_memslots(kvm);
	if (r)
		kvm_iommu_unmap_memslots(kvm);

out_unlock:
	mutex_unlock(&kvm->slots_lock);
	return r;
}

static void kvm_unpin_pages(struct kvm *kvm, pfn_t pfn, unsigned long npages)
{
	unsigned long i;

	for (i = 0; i < npages; ++i)
		kvm_release_pfn_clean(pfn + i);
}

static void kvm_iommu_put_pages(struct kvm *kvm,
				gfn_t base_gfn, unsigned long npages)
{
	struct iommu_domain *domain;
	gfn_t end_gfn, gfn;
	pfn_t pfn;
	u64 phys;

	domain  = kvm->arch.iommu_domain;
	end_gfn = base_gfn + npages;
	gfn     = base_gfn;

	/* check if iommu exists and in use */
	if (!domain)
		return;

	while (gfn < end_gfn) {
		unsigned long unmap_pages;
		size_t size;

		/* Get physical address */
		phys = iommu_iova_to_phys(domain, gfn_to_gpa(gfn));

		if (!phys) {
			gfn++;
			continue;
		}

		pfn  = phys >> PAGE_SHIFT;

		/* Unmap address from IO address space */
		size       = iommu_unmap(domain, gfn_to_gpa(gfn), PAGE_SIZE);
		unmap_pages = 1ULL << get_order(size);

		/* Unpin all pages we just unmapped to not leak any memory */
		kvm_unpin_pages(kvm, pfn, unmap_pages);

		gfn += unmap_pages;
	}
}

void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
{
	kvm_iommu_put_pages(kvm, slot->base_gfn, slot->npages);
}

static int kvm_iommu_unmap_memslots(struct kvm *kvm)
{
	int idx;
	struct kvm_memslots *slots;
	struct kvm_memory_slot *memslot;

	idx = srcu_read_lock(&kvm->srcu);
	slots = kvm_memslots(kvm);

	kvm_for_each_memslot(memslot, slots)
		kvm_iommu_unmap_pages(kvm, memslot);

	srcu_read_unlock(&kvm->srcu, idx);

	return 0;
}

int kvm_iommu_unmap_guest(struct kvm *kvm)
{
	struct iommu_domain *domain = kvm->arch.iommu_domain;

	/* check if iommu exists and in use */
	if (!domain)
		return 0;

	mutex_lock(&kvm->slots_lock);
	kvm_iommu_unmap_memslots(kvm);
	kvm->arch.iommu_domain = NULL;
	mutex_unlock(&kvm->slots_lock);

	iommu_domain_free(domain);
	return 0;
}
Commit	Line	Data
	1	/*
	2	* Copyright (c) 2006, Intel Corporation.
	3	*
	4	* This program is free software; you can redistribute it and/or modify it
	5	* under the terms and conditions of the GNU General Public License,
	6	* version 2, as published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope it will be useful, but WITHOUT
	9	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	10	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	11	* more details.
	12	*
	13	* You should have received a copy of the GNU General Public License along with
	14	* this program; if not, write to the Free Software Foundation, Inc., 59 Temple
	15	* Place - Suite 330, Boston, MA 02111-1307 USA.
	16	*
	17	* Copyright (C) 2006-2008 Intel Corporation
	18	* Copyright IBM Corporation, 2008
	19	* Copyright 2010 Red Hat, Inc. and/or its affiliates.
	20	*
	21	* Author: Allen M. Kay <allen.m.kay@intel.com>
	22	* Author: Weidong Han <weidong.han@intel.com>
	23	* Author: Ben-Ami Yassour <benami@il.ibm.com>
	24	*/
	25
	26	#include <linux/list.h>
	27	#include <linux/kvm_host.h>
	28	#include <linux/module.h>
	29	#include <linux/pci.h>
	30	#include <linux/stat.h>
	31	#include <linux/dmar.h>
	32	#include <linux/iommu.h>
	33	#include <linux/intel-iommu.h>
	34
	35	static bool allow_unsafe_assigned_interrupts;
	36	module_param_named(allow_unsafe_assigned_interrupts,
	37	allow_unsafe_assigned_interrupts, bool, S_IRUGO \| S_IWUSR);
	38	MODULE_PARM_DESC(allow_unsafe_assigned_interrupts,
	39	"Enable device assignment on platforms without interrupt remapping support.");
	40
	41	static int kvm_iommu_unmap_memslots(struct kvm *kvm);
	42	static void kvm_iommu_put_pages(struct kvm *kvm,
	43	gfn_t base_gfn, unsigned long npages);
	44
	45	static pfn_t kvm_pin_pages(struct kvm_memory_slot *slot, gfn_t gfn,
	46	unsigned long size)
	47	{
	48	gfn_t end_gfn;
	49	pfn_t pfn;
	50
	51	pfn = gfn_to_pfn_memslot(slot, gfn);
	52	end_gfn = gfn + (size >> PAGE_SHIFT);
	53	gfn += 1;
	54
	55	if (is_error_noslot_pfn(pfn))
	56	return pfn;
	57
	58	while (gfn < end_gfn)
	59	gfn_to_pfn_memslot(slot, gfn++);
	60
	61	return pfn;
	62	}
	63
	64	int kvm_iommu_map_pages(struct kvm kvm, struct kvm_memory_slot slot)
	65	{
	66	gfn_t gfn, end_gfn;
	67	pfn_t pfn;
	68	int r = 0;
	69	struct iommu_domain *domain = kvm->arch.iommu_domain;
	70	int flags;
	71
	72	/* check if iommu exists and in use */
	73	if (!domain)
	74	return 0;
	75
	76	gfn = slot->base_gfn;
	77	end_gfn = gfn + slot->npages;
	78
	79	flags = IOMMU_READ \| IOMMU_WRITE;
	80	if (kvm->arch.iommu_flags & KVM_IOMMU_CACHE_COHERENCY)
	81	flags \|= IOMMU_CACHE;
	82
	83
	84	while (gfn < end_gfn) {
	85	unsigned long page_size;
	86
	87	/* Check if already mapped */
	88	if (iommu_iova_to_phys(domain, gfn_to_gpa(gfn))) {
	89	gfn += 1;
	90	continue;
	91	}
	92
	93	/* Get the page size we could use to map */
	94	page_size = kvm_host_page_size(kvm, gfn);
	95
	96	/* Make sure the page_size does not exceed the memslot */
	97	while ((gfn + (page_size >> PAGE_SHIFT)) > end_gfn)
	98	page_size >>= 1;
	99
	100	/* Make sure gfn is aligned to the page size we want to map */
	101	while ((gfn << PAGE_SHIFT) & (page_size - 1))
	102	page_size >>= 1;
	103
	104	/*
	105	* Pin all pages we are about to map in memory. This is
	106	* important because we unmap and unpin in 4kb steps later.
	107	*/
	108	pfn = kvm_pin_pages(slot, gfn, page_size);
	109	if (is_error_noslot_pfn(pfn)) {
	110	gfn += 1;
	111	continue;
	112	}
	113
	114	/* Map into IO address space */
	115	r = iommu_map(domain, gfn_to_gpa(gfn), pfn_to_hpa(pfn),
	116	page_size, flags);
	117	if (r) {
	118	printk(KERN_ERR "kvm_iommu_map_address:"
	119	"iommu failed to map pfn=%llx\n", pfn);
	120	goto unmap_pages;
	121	}
	122
	123	gfn += page_size >> PAGE_SHIFT;
	124
	125
	126	}
	127
	128	return 0;
	129
	130	unmap_pages:
	131	kvm_iommu_put_pages(kvm, slot->base_gfn, gfn);
	132	return r;
	133	}
	134
	135	static int kvm_iommu_map_memslots(struct kvm *kvm)
	136	{
	137	int idx, r = 0;
	138	struct kvm_memslots *slots;
	139	struct kvm_memory_slot *memslot;
	140
	141	idx = srcu_read_lock(&kvm->srcu);
	142	slots = kvm_memslots(kvm);
	143
	144	kvm_for_each_memslot(memslot, slots) {
	145	r = kvm_iommu_map_pages(kvm, memslot);
	146	if (r)
	147	break;
	148	}
	149	srcu_read_unlock(&kvm->srcu, idx);
	150
	151	return r;
	152	}
	153
	154	int kvm_assign_device(struct kvm *kvm,
	155	struct kvm_assigned_dev_kernel *assigned_dev)
	156	{
	157	struct pci_dev *pdev = NULL;
	158	struct iommu_domain *domain = kvm->arch.iommu_domain;
	159	int r, last_flags;
	160
	161	/* check if iommu exists and in use */
	162	if (!domain)
	163	return 0;
	164
	165	pdev = assigned_dev->dev;
	166	if (pdev == NULL)
	167	return -ENODEV;
	168
	169	r = iommu_attach_device(domain, &pdev->dev);
	170	if (r) {
	171	dev_err(&pdev->dev, "kvm assign device failed ret %d", r);
	172	return r;
	173	}
	174
	175	last_flags = kvm->arch.iommu_flags;
	176	if (iommu_domain_has_cap(kvm->arch.iommu_domain,
	177	IOMMU_CAP_CACHE_COHERENCY))
	178	kvm->arch.iommu_flags \|= KVM_IOMMU_CACHE_COHERENCY;
	179
	180	/* Check if need to update IOMMU page table for guest memory */
	181	if ((last_flags ^ kvm->arch.iommu_flags) ==
	182	KVM_IOMMU_CACHE_COHERENCY) {
	183	kvm_iommu_unmap_memslots(kvm);
	184	r = kvm_iommu_map_memslots(kvm);
	185	if (r)
	186	goto out_unmap;
	187	}
	188
	189	pdev->dev_flags \|= PCI_DEV_FLAGS_ASSIGNED;
	190
	191	printk(KERN_DEBUG "assign device %x:%x:%x.%x\n",
	192	assigned_dev->host_segnr,
	193	assigned_dev->host_busnr,
	194	PCI_SLOT(assigned_dev->host_devfn),
	195	PCI_FUNC(assigned_dev->host_devfn));
	196
	197	return 0;
	198	out_unmap:
	199	kvm_iommu_unmap_memslots(kvm);
	200	return r;
	201	}
	202
	203	int kvm_deassign_device(struct kvm *kvm,
	204	struct kvm_assigned_dev_kernel *assigned_dev)
	205	{
	206	struct iommu_domain *domain = kvm->arch.iommu_domain;
	207	struct pci_dev *pdev = NULL;
	208
	209	/* check if iommu exists and in use */
	210	if (!domain)
	211	return 0;
	212
	213	pdev = assigned_dev->dev;
	214	if (pdev == NULL)
	215	return -ENODEV;
	216
	217	iommu_detach_device(domain, &pdev->dev);
	218
	219	pdev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
	220
	221	printk(KERN_DEBUG "deassign device %x:%x:%x.%x\n",
	222	assigned_dev->host_segnr,
	223	assigned_dev->host_busnr,
	224	PCI_SLOT(assigned_dev->host_devfn),
	225	PCI_FUNC(assigned_dev->host_devfn));
	226
	227	return 0;
	228	}
	229
	230	int kvm_iommu_map_guest(struct kvm *kvm)
	231	{
	232	int r;
	233
	234	if (!iommu_present(&pci_bus_type)) {
	235	printk(KERN_ERR "%s: iommu not found\n", __func__);
	236	return -ENODEV;
	237	}
	238
	239	mutex_lock(&kvm->slots_lock);
	240
	241	kvm->arch.iommu_domain = iommu_domain_alloc(&pci_bus_type);
	242	if (!kvm->arch.iommu_domain) {
	243	r = -ENOMEM;
	244	goto out_unlock;
	245	}
	246
	247	if (!allow_unsafe_assigned_interrupts &&
	248	!iommu_domain_has_cap(kvm->arch.iommu_domain,
	249	IOMMU_CAP_INTR_REMAP)) {
	250	printk(KERN_WARNING "%s: No interrupt remapping support,"
	251	" disallowing device assignment."
	252	" Re-enble with \"allow_unsafe_assigned_interrupts=1\""
	253	" module option.\n", __func__);
	254	iommu_domain_free(kvm->arch.iommu_domain);
	255	kvm->arch.iommu_domain = NULL;
	256	r = -EPERM;
	257	goto out_unlock;
	258	}
	259
	260	r = kvm_iommu_map_memslots(kvm);
	261	if (r)
	262	kvm_iommu_unmap_memslots(kvm);
	263
	264	out_unlock:
	265	mutex_unlock(&kvm->slots_lock);
	266	return r;
	267	}
	268
	269	static void kvm_unpin_pages(struct kvm *kvm, pfn_t pfn, unsigned long npages)
	270	{
	271	unsigned long i;
	272
	273	for (i = 0; i < npages; ++i)
	274	kvm_release_pfn_clean(pfn + i);
	275	}
	276
	277	static void kvm_iommu_put_pages(struct kvm *kvm,
	278	gfn_t base_gfn, unsigned long npages)
	279	{
	280	struct iommu_domain *domain;
	281	gfn_t end_gfn, gfn;
	282	pfn_t pfn;
	283	u64 phys;
	284
	285	domain = kvm->arch.iommu_domain;
	286	end_gfn = base_gfn + npages;
	287	gfn = base_gfn;
	288
	289	/* check if iommu exists and in use */
	290	if (!domain)
	291	return;
	292
	293	while (gfn < end_gfn) {
	294	unsigned long unmap_pages;
	295	size_t size;
	296
	297	/* Get physical address */
	298	phys = iommu_iova_to_phys(domain, gfn_to_gpa(gfn));
	299
	300	if (!phys) {
	301	gfn++;
	302	continue;
	303	}
	304
	305	pfn = phys >> PAGE_SHIFT;
	306
	307	/* Unmap address from IO address space */
	308	size = iommu_unmap(domain, gfn_to_gpa(gfn), PAGE_SIZE);
	309	unmap_pages = 1ULL << get_order(size);
	310
	311	/* Unpin all pages we just unmapped to not leak any memory */
	312	kvm_unpin_pages(kvm, pfn, unmap_pages);
	313
	314	gfn += unmap_pages;
	315	}
	316	}
	317
	318	void kvm_iommu_unmap_pages(struct kvm kvm, struct kvm_memory_slot slot)
	319	{
	320	kvm_iommu_put_pages(kvm, slot->base_gfn, slot->npages);
	321	}
	322
	323	static int kvm_iommu_unmap_memslots(struct kvm *kvm)
	324	{
	325	int idx;
	326	struct kvm_memslots *slots;
	327	struct kvm_memory_slot *memslot;
	328
	329	idx = srcu_read_lock(&kvm->srcu);
	330	slots = kvm_memslots(kvm);
	331
	332	kvm_for_each_memslot(memslot, slots)
	333	kvm_iommu_unmap_pages(kvm, memslot);
	334
	335	srcu_read_unlock(&kvm->srcu, idx);
	336
	337	return 0;
	338	}
	339
	340	int kvm_iommu_unmap_guest(struct kvm *kvm)
	341	{
	342	struct iommu_domain *domain = kvm->arch.iommu_domain;
	343
	344	/* check if iommu exists and in use */
	345	if (!domain)
	346	return 0;
	347
	348	mutex_lock(&kvm->slots_lock);
	349	kvm_iommu_unmap_memslots(kvm);
	350	kvm->arch.iommu_domain = NULL;
	351	mutex_unlock(&kvm->slots_lock);
	352
	353	iommu_domain_free(domain);
	354	return 0;
	355	}