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

/*
 * VFIO-KVM bridge pseudo device
 *
 * Copyright (C) 2013 Red Hat, Inc.  All rights reserved.
 *     Author: Alex Williamson <[email protected]>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#include <linux/errno.h>
#include <linux/file.h>
#include <linux/kvm_host.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/slab.h>
#include <linux/uaccess.h>
#include <linux/vfio.h>
#include "vfio.h"

struct kvm_vfio_group {
	struct list_head node;
	struct vfio_group *vfio_group;
};

struct kvm_vfio {
	struct list_head group_list;
	struct mutex lock;
	bool noncoherent;
};

static struct vfio_group *kvm_vfio_group_get_external_user(struct file *filep)
{
	struct vfio_group *vfio_group;
	struct vfio_group *(*fn)(struct file *);

	fn = symbol_get(vfio_group_get_external_user);
	if (!fn)
		return ERR_PTR(-EINVAL);

	vfio_group = fn(filep);

	symbol_put(vfio_group_get_external_user);

	return vfio_group;
}

static void kvm_vfio_group_put_external_user(struct vfio_group *vfio_group)
{
	void (*fn)(struct vfio_group *);

	fn = symbol_get(vfio_group_put_external_user);
	if (!fn)
		return;

	fn(vfio_group);

	symbol_put(vfio_group_put_external_user);
}

static bool kvm_vfio_group_is_coherent(struct vfio_group *vfio_group)
{
	long (*fn)(struct vfio_group *, unsigned long);
	long ret;

	fn = symbol_get(vfio_external_check_extension);
	if (!fn)
		return false;

	ret = fn(vfio_group, VFIO_DMA_CC_IOMMU);

	symbol_put(vfio_external_check_extension);

	return ret > 0;
}

/*
 * Groups can use the same or different IOMMU domains.  If the same then
 * adding a new group may change the coherency of groups we've previously
 * been told about.  We don't want to care about any of that so we retest
 * each group and bail as soon as we find one that's noncoherent.  This
 * means we only ever [un]register_noncoherent_dma once for the whole device.
 */
static void kvm_vfio_update_coherency(struct kvm_device *dev)
{
	struct kvm_vfio *kv = dev->private;
	bool noncoherent = false;
	struct kvm_vfio_group *kvg;

	mutex_lock(&kv->lock);

	list_for_each_entry(kvg, &kv->group_list, node) {
		if (!kvm_vfio_group_is_coherent(kvg->vfio_group)) {
			noncoherent = true;
			break;
		}
	}

	if (noncoherent != kv->noncoherent) {
		kv->noncoherent = noncoherent;

		if (kv->noncoherent)
			kvm_arch_register_noncoherent_dma(dev->kvm);
		else
			kvm_arch_unregister_noncoherent_dma(dev->kvm);
	}

	mutex_unlock(&kv->lock);
}

static int kvm_vfio_set_group(struct kvm_device *dev, long attr, u64 arg)
{
	struct kvm_vfio *kv = dev->private;
	struct vfio_group *vfio_group;
	struct kvm_vfio_group *kvg;
	int32_t __user *argp = (int32_t __user *)(unsigned long)arg;
	struct fd f;
	int32_t fd;
	int ret;

	switch (attr) {
	case KVM_DEV_VFIO_GROUP_ADD:
		if (get_user(fd, argp))
			return -EFAULT;

		f = fdget(fd);
		if (!f.file)
			return -EBADF;

		vfio_group = kvm_vfio_group_get_external_user(f.file);
		fdput(f);

		if (IS_ERR(vfio_group))
			return PTR_ERR(vfio_group);

		mutex_lock(&kv->lock);

		list_for_each_entry(kvg, &kv->group_list, node) {
			if (kvg->vfio_group == vfio_group) {
				mutex_unlock(&kv->lock);
				kvm_vfio_group_put_external_user(vfio_group);
				return -EEXIST;
			}
		}

		kvg = kzalloc(sizeof(*kvg), GFP_KERNEL);
		if (!kvg) {
			mutex_unlock(&kv->lock);
			kvm_vfio_group_put_external_user(vfio_group);
			return -ENOMEM;
		}

		list_add_tail(&kvg->node, &kv->group_list);
		kvg->vfio_group = vfio_group;

		kvm_arch_start_assignment(dev->kvm);

		mutex_unlock(&kv->lock);

		kvm_vfio_update_coherency(dev);

		return 0;

	case KVM_DEV_VFIO_GROUP_DEL:
		if (get_user(fd, argp))
			return -EFAULT;

		f = fdget(fd);
		if (!f.file)
			return -EBADF;

		vfio_group = kvm_vfio_group_get_external_user(f.file);
		fdput(f);

		if (IS_ERR(vfio_group))
			return PTR_ERR(vfio_group);

		ret = -ENOENT;

		mutex_lock(&kv->lock);

		list_for_each_entry(kvg, &kv->group_list, node) {
			if (kvg->vfio_group != vfio_group)
				continue;

			list_del(&kvg->node);
			kvm_vfio_group_put_external_user(kvg->vfio_group);
			kfree(kvg);
			ret = 0;
			break;
		}

		kvm_arch_end_assignment(dev->kvm);

		mutex_unlock(&kv->lock);

		kvm_vfio_group_put_external_user(vfio_group);

		kvm_vfio_update_coherency(dev);

		return ret;
	}

	return -ENXIO;
}

static int kvm_vfio_set_attr(struct kvm_device *dev,
			     struct kvm_device_attr *attr)
{
	switch (attr->group) {
	case KVM_DEV_VFIO_GROUP:
		return kvm_vfio_set_group(dev, attr->attr, attr->addr);
	}

	return -ENXIO;
}

static int kvm_vfio_has_attr(struct kvm_device *dev,
			     struct kvm_device_attr *attr)
{
	switch (attr->group) {
	case KVM_DEV_VFIO_GROUP:
		switch (attr->attr) {
		case KVM_DEV_VFIO_GROUP_ADD:
		case KVM_DEV_VFIO_GROUP_DEL:
			return 0;
		}

		break;
	}

	return -ENXIO;
}

static void kvm_vfio_destroy(struct kvm_device *dev)
{
	struct kvm_vfio *kv = dev->private;
	struct kvm_vfio_group *kvg, *tmp;

	list_for_each_entry_safe(kvg, tmp, &kv->group_list, node) {
		kvm_vfio_group_put_external_user(kvg->vfio_group);
		list_del(&kvg->node);
		kfree(kvg);
		kvm_arch_end_assignment(dev->kvm);
	}

	kvm_vfio_update_coherency(dev);

	kfree(kv);
	kfree(dev); /* alloc by kvm_ioctl_create_device, free by .destroy */
}

static int kvm_vfio_create(struct kvm_device *dev, u32 type);

static struct kvm_device_ops kvm_vfio_ops = {
	.name = "kvm-vfio",
	.create = kvm_vfio_create,
	.destroy = kvm_vfio_destroy,
	.set_attr = kvm_vfio_set_attr,
	.has_attr = kvm_vfio_has_attr,
};

static int kvm_vfio_create(struct kvm_device *dev, u32 type)
{
	struct kvm_device *tmp;
	struct kvm_vfio *kv;

	/* Only one VFIO "device" per VM */
	list_for_each_entry(tmp, &dev->kvm->devices, vm_node)
		if (tmp->ops == &kvm_vfio_ops)
			return -EBUSY;

	kv = kzalloc(sizeof(*kv), GFP_KERNEL);
	if (!kv)
		return -ENOMEM;

	INIT_LIST_HEAD(&kv->group_list);
	mutex_init(&kv->lock);

	dev->private = kv;

	return 0;
}

int kvm_vfio_ops_init(void)
{
	return kvm_register_device_ops(&kvm_vfio_ops, KVM_DEV_TYPE_VFIO);
}

void kvm_vfio_ops_exit(void)
{
	kvm_unregister_device_ops(KVM_DEV_TYPE_VFIO);
}
Commit	Line	Data
ec53500f AW	1	/*
	2	* VFIO-KVM bridge pseudo device
	3	*
	4	* Copyright (C) 2013 Red Hat, Inc. All rights reserved.
	5	* Author: Alex Williamson <[email protected]>
	6	*
	7	* This program is free software; you can redistribute it and/or modify
	8	* it under the terms of the GNU General Public License version 2 as
	9	* published by the Free Software Foundation.
	10	*/
	11
	12	#include <linux/errno.h>
	13	#include <linux/file.h>
	14	#include <linux/kvm_host.h>
	15	#include <linux/list.h>
	16	#include <linux/module.h>
	17	#include <linux/mutex.h>
	18	#include <linux/slab.h>
	19	#include <linux/uaccess.h>
	20	#include <linux/vfio.h>
3c3c29fd	21	#include "vfio.h"
ec53500f AW	22
	23	struct kvm_vfio_group {
	24	struct list_head node;
	25	struct vfio_group *vfio_group;
	26	};
	27
	28	struct kvm_vfio {
	29	struct list_head group_list;
	30	struct mutex lock;
e0f0bbc5	31	bool noncoherent;
ec53500f AW	32	};
	33
	34	static struct vfio_group kvm_vfio_group_get_external_user(struct file filep)
	35	{
	36	struct vfio_group *vfio_group;
	37	struct vfio_group (fn)(struct file *);
	38
	39	fn = symbol_get(vfio_group_get_external_user);
	40	if (!fn)
	41	return ERR_PTR(-EINVAL);
	42
	43	vfio_group = fn(filep);
	44
	45	symbol_put(vfio_group_get_external_user);
	46
	47	return vfio_group;
	48	}
	49
	50	static void kvm_vfio_group_put_external_user(struct vfio_group *vfio_group)
	51	{
	52	void (fn)(struct vfio_group );
	53
	54	fn = symbol_get(vfio_group_put_external_user);
	55	if (!fn)
	56	return;
	57
	58	fn(vfio_group);
	59
	60	symbol_put(vfio_group_put_external_user);
	61	}
	62
9d830d47 AW	63	static bool kvm_vfio_group_is_coherent(struct vfio_group *vfio_group)
	64	{
	65	long (fn)(struct vfio_group , unsigned long);
	66	long ret;
	67
	68	fn = symbol_get(vfio_external_check_extension);
	69	if (!fn)
	70	return false;
	71
	72	ret = fn(vfio_group, VFIO_DMA_CC_IOMMU);
	73
	74	symbol_put(vfio_external_check_extension);
	75
	76	return ret > 0;
	77	}
	78
e0f0bbc5 AW	79	/*
	80	* Groups can use the same or different IOMMU domains. If the same then
	81	* adding a new group may change the coherency of groups we've previously
	82	* been told about. We don't want to care about any of that so we retest
	83	* each group and bail as soon as we find one that's noncoherent. This
	84	* means we only ever [un]register_noncoherent_dma once for the whole device.
	85	*/
	86	static void kvm_vfio_update_coherency(struct kvm_device *dev)
	87	{
	88	struct kvm_vfio *kv = dev->private;
	89	bool noncoherent = false;
	90	struct kvm_vfio_group *kvg;
	91
	92	mutex_lock(&kv->lock);
	93
	94	list_for_each_entry(kvg, &kv->group_list, node) {
9d830d47 AW	95	if (!kvm_vfio_group_is_coherent(kvg->vfio_group)) {
	96	noncoherent = true;
	97	break;
	98	}
e0f0bbc5 AW	99	}
	100
	101	if (noncoherent != kv->noncoherent) {
	102	kv->noncoherent = noncoherent;
	103
	104	if (kv->noncoherent)
	105	kvm_arch_register_noncoherent_dma(dev->kvm);
	106	else
	107	kvm_arch_unregister_noncoherent_dma(dev->kvm);
	108	}
	109
	110	mutex_unlock(&kv->lock);
	111	}
	112
ec53500f AW	113	static int kvm_vfio_set_group(struct kvm_device *dev, long attr, u64 arg)
	114	{
	115	struct kvm_vfio *kv = dev->private;
	116	struct vfio_group *vfio_group;
	117	struct kvm_vfio_group *kvg;
e81d1ad3	118	int32_t __user argp = (int32_t __user )(unsigned long)arg;
ec53500f AW	119	struct fd f;
	120	int32_t fd;
	121	int ret;
	122
	123	switch (attr) {
	124	case KVM_DEV_VFIO_GROUP_ADD:
e81d1ad3	125	if (get_user(fd, argp))
ec53500f AW	126	return -EFAULT;
	127
	128	f = fdget(fd);
	129	if (!f.file)
	130	return -EBADF;
	131
	132	vfio_group = kvm_vfio_group_get_external_user(f.file);
	133	fdput(f);
	134
	135	if (IS_ERR(vfio_group))
	136	return PTR_ERR(vfio_group);
	137
	138	mutex_lock(&kv->lock);
	139
	140	list_for_each_entry(kvg, &kv->group_list, node) {
	141	if (kvg->vfio_group == vfio_group) {
	142	mutex_unlock(&kv->lock);
	143	kvm_vfio_group_put_external_user(vfio_group);
	144	return -EEXIST;
	145	}
	146	}
	147
	148	kvg = kzalloc(sizeof(*kvg), GFP_KERNEL);
	149	if (!kvg) {
	150	mutex_unlock(&kv->lock);
	151	kvm_vfio_group_put_external_user(vfio_group);
	152	return -ENOMEM;
	153	}
	154
	155	list_add_tail(&kvg->node, &kv->group_list);
	156	kvg->vfio_group = vfio_group;
	157
5544eb9b PB	158	kvm_arch_start_assignment(dev->kvm);
5544eb9b PB	159
ec53500f AW	160	mutex_unlock(&kv->lock);
ec53500f AW	161
e0f0bbc5 AW	162	kvm_vfio_update_coherency(dev);
e0f0bbc5 AW	163
ec53500f AW	164	return 0;
	165
	166	case KVM_DEV_VFIO_GROUP_DEL:
e81d1ad3	167	if (get_user(fd, argp))
ec53500f AW	168	return -EFAULT;
	169
	170	f = fdget(fd);
	171	if (!f.file)
	172	return -EBADF;
	173
	174	vfio_group = kvm_vfio_group_get_external_user(f.file);
	175	fdput(f);
	176
	177	if (IS_ERR(vfio_group))
	178	return PTR_ERR(vfio_group);
	179
	180	ret = -ENOENT;
	181
	182	mutex_lock(&kv->lock);
	183
	184	list_for_each_entry(kvg, &kv->group_list, node) {
	185	if (kvg->vfio_group != vfio_group)
	186	continue;
	187
	188	list_del(&kvg->node);
	189	kvm_vfio_group_put_external_user(kvg->vfio_group);
	190	kfree(kvg);
	191	ret = 0;
	192	break;
	193	}
	194
5544eb9b PB	195	kvm_arch_end_assignment(dev->kvm);
5544eb9b PB	196
ec53500f AW	197	mutex_unlock(&kv->lock);
	198
	199	kvm_vfio_group_put_external_user(vfio_group);
	200
e0f0bbc5 AW	201	kvm_vfio_update_coherency(dev);
e0f0bbc5 AW	202
ec53500f AW	203	return ret;
	204	}
	205
	206	return -ENXIO;
	207	}
	208
	209	static int kvm_vfio_set_attr(struct kvm_device *dev,
	210	struct kvm_device_attr *attr)
	211	{
	212	switch (attr->group) {
	213	case KVM_DEV_VFIO_GROUP:
	214	return kvm_vfio_set_group(dev, attr->attr, attr->addr);
	215	}
	216
	217	return -ENXIO;
	218	}
	219
	220	static int kvm_vfio_has_attr(struct kvm_device *dev,
	221	struct kvm_device_attr *attr)
	222	{
	223	switch (attr->group) {
	224	case KVM_DEV_VFIO_GROUP:
	225	switch (attr->attr) {
	226	case KVM_DEV_VFIO_GROUP_ADD:
	227	case KVM_DEV_VFIO_GROUP_DEL:
	228	return 0;
	229	}
	230
	231	break;
	232	}
	233
	234	return -ENXIO;
	235	}
	236
	237	static void kvm_vfio_destroy(struct kvm_device *dev)
	238	{
	239	struct kvm_vfio *kv = dev->private;
	240	struct kvm_vfio_group kvg, tmp;
	241
	242	list_for_each_entry_safe(kvg, tmp, &kv->group_list, node) {
	243	kvm_vfio_group_put_external_user(kvg->vfio_group);
	244	list_del(&kvg->node);
	245	kfree(kvg);
5544eb9b	246	kvm_arch_end_assignment(dev->kvm);
ec53500f AW	247	}
ec53500f AW	248
e0f0bbc5 AW	249	kvm_vfio_update_coherency(dev);
e0f0bbc5 AW	250
ec53500f AW	251	kfree(kv);
	252	kfree(dev); /* alloc by kvm_ioctl_create_device, free by .destroy */
	253	}
	254
80ce1639 WD	255	static int kvm_vfio_create(struct kvm_device *dev, u32 type);
	256
	257	static struct kvm_device_ops kvm_vfio_ops = {
	258	.name = "kvm-vfio",
	259	.create = kvm_vfio_create,
	260	.destroy = kvm_vfio_destroy,
	261	.set_attr = kvm_vfio_set_attr,
	262	.has_attr = kvm_vfio_has_attr,
	263	};
	264
ec53500f AW	265	static int kvm_vfio_create(struct kvm_device *dev, u32 type)
	266	{
	267	struct kvm_device *tmp;
	268	struct kvm_vfio *kv;
	269
	270	/* Only one VFIO "device" per VM */
	271	list_for_each_entry(tmp, &dev->kvm->devices, vm_node)
	272	if (tmp->ops == &kvm_vfio_ops)
	273	return -EBUSY;
	274
	275	kv = kzalloc(sizeof(*kv), GFP_KERNEL);
	276	if (!kv)
	277	return -ENOMEM;
	278
	279	INIT_LIST_HEAD(&kv->group_list);
	280	mutex_init(&kv->lock);
	281
	282	dev->private = kv;
	283
	284	return 0;
	285	}
	286
3c3c29fd	287	int kvm_vfio_ops_init(void)
80ce1639 WD	288	{
	289	return kvm_register_device_ops(&kvm_vfio_ops, KVM_DEV_TYPE_VFIO);
	290	}
571ee1b6 WL	291
	292	void kvm_vfio_ops_exit(void)
	293	{
	294	kvm_unregister_device_ops(KVM_DEV_TYPE_VFIO);
	295	}