xref: /kvmtool/virtio/core.c (revision dc7a55d6bfe61b848832ad1e22273cee4db5d155)
1 #include <linux/virtio_ring.h>
2 #include <linux/types.h>
3 #include <sys/uio.h>
4 #include <stdlib.h>
5 
6 #include "kvm/guest_compat.h"
7 #include "kvm/barrier.h"
8 #include "kvm/virtio.h"
9 #include "kvm/virtio-pci.h"
10 #include "kvm/virtio-mmio.h"
11 #include "kvm/util.h"
12 #include "kvm/kvm.h"
13 
14 
15 const char* virtio_trans_name(enum virtio_trans trans)
16 {
17 	if (trans == VIRTIO_PCI)
18 		return "pci";
19 	else if (trans == VIRTIO_MMIO)
20 		return "mmio";
21 	return "unknown";
22 }
23 
24 struct vring_used_elem *virt_queue__set_used_elem(struct virt_queue *queue, u32 head, u32 len)
25 {
26 	struct vring_used_elem *used_elem;
27 	u16 idx = virtio_guest_to_host_u16(queue, queue->vring.used->idx);
28 
29 	used_elem	= &queue->vring.used->ring[idx % queue->vring.num];
30 	used_elem->id	= virtio_host_to_guest_u32(queue, head);
31 	used_elem->len	= virtio_host_to_guest_u32(queue, len);
32 
33 	/*
34 	 * Use wmb to assure that used elem was updated with head and len.
35 	 * We need a wmb here since we can't advance idx unless we're ready
36 	 * to pass the used element to the guest.
37 	 */
38 	wmb();
39 	idx++;
40 	queue->vring.used->idx = virtio_host_to_guest_u16(queue, idx);
41 
42 	/*
43 	 * Use wmb to assure used idx has been increased before we signal the guest.
44 	 * Without a wmb here the guest may ignore the queue since it won't see
45 	 * an updated idx.
46 	 */
47 	wmb();
48 
49 	return used_elem;
50 }
51 
52 static inline bool virt_desc__test_flag(struct virt_queue *vq,
53 					struct vring_desc *desc, u16 flag)
54 {
55 	return !!(virtio_guest_to_host_u16(vq, desc->flags) & flag);
56 }
57 
58 /*
59  * Each buffer in the virtqueues is actually a chain of descriptors.  This
60  * function returns the next descriptor in the chain, or vq->vring.num if we're
61  * at the end.
62  */
63 static unsigned next_desc(struct virt_queue *vq, struct vring_desc *desc,
64 			  unsigned int i, unsigned int max)
65 {
66 	unsigned int next;
67 
68 	/* If this descriptor says it doesn't chain, we're done. */
69 	if (!virt_desc__test_flag(vq, &desc[i], VRING_DESC_F_NEXT))
70 		return max;
71 
72 	/* Check they're not leading us off end of descriptors. */
73 	next = virtio_guest_to_host_u16(vq, desc[i].next);
74 	/* Make sure compiler knows to grab that: we don't want it changing! */
75 	wmb();
76 
77 	return next;
78 }
79 
80 u16 virt_queue__get_head_iov(struct virt_queue *vq, struct iovec iov[], u16 *out, u16 *in, u16 head, struct kvm *kvm)
81 {
82 	struct vring_desc *desc;
83 	u16 idx;
84 	u16 max;
85 
86 	idx = head;
87 	*out = *in = 0;
88 	max = vq->vring.num;
89 	desc = vq->vring.desc;
90 
91 	if (virt_desc__test_flag(vq, &desc[idx], VRING_DESC_F_INDIRECT)) {
92 		max = virtio_guest_to_host_u32(vq, desc[idx].len) / sizeof(struct vring_desc);
93 		desc = guest_flat_to_host(kvm, virtio_guest_to_host_u64(vq, desc[idx].addr));
94 		idx = 0;
95 	}
96 
97 	do {
98 		/* Grab the first descriptor, and check it's OK. */
99 		iov[*out + *in].iov_len = virtio_guest_to_host_u32(vq, desc[idx].len);
100 		iov[*out + *in].iov_base = guest_flat_to_host(kvm,
101 							      virtio_guest_to_host_u64(vq, desc[idx].addr));
102 		/* If this is an input descriptor, increment that count. */
103 		if (virt_desc__test_flag(vq, &desc[idx], VRING_DESC_F_WRITE))
104 			(*in)++;
105 		else
106 			(*out)++;
107 	} while ((idx = next_desc(vq, desc, idx, max)) != max);
108 
109 	return head;
110 }
111 
112 u16 virt_queue__get_iov(struct virt_queue *vq, struct iovec iov[], u16 *out, u16 *in, struct kvm *kvm)
113 {
114 	u16 head;
115 
116 	head = virt_queue__pop(vq);
117 
118 	return virt_queue__get_head_iov(vq, iov, out, in, head, kvm);
119 }
120 
121 /* in and out are relative to guest */
122 u16 virt_queue__get_inout_iov(struct kvm *kvm, struct virt_queue *queue,
123 			      struct iovec in_iov[], struct iovec out_iov[],
124 			      u16 *in, u16 *out)
125 {
126 	struct vring_desc *desc;
127 	u16 head, idx;
128 
129 	idx = head = virt_queue__pop(queue);
130 	*out = *in = 0;
131 	do {
132 		u64 addr;
133 		desc = virt_queue__get_desc(queue, idx);
134 		addr = virtio_guest_to_host_u64(queue, desc->addr);
135 		if (virt_desc__test_flag(queue, desc, VRING_DESC_F_WRITE)) {
136 			in_iov[*in].iov_base = guest_flat_to_host(kvm, addr);
137 			in_iov[*in].iov_len = virtio_guest_to_host_u32(queue, desc->len);
138 			(*in)++;
139 		} else {
140 			out_iov[*out].iov_base = guest_flat_to_host(kvm, addr);
141 			out_iov[*out].iov_len = virtio_guest_to_host_u32(queue, desc->len);
142 			(*out)++;
143 		}
144 		if (virt_desc__test_flag(queue, desc, VRING_DESC_F_NEXT))
145 			idx = virtio_guest_to_host_u16(queue, desc->next);
146 		else
147 			break;
148 	} while (1);
149 
150 	return head;
151 }
152 
153 int virtio__get_dev_specific_field(int offset, bool msix, u32 *config_off)
154 {
155 	if (msix) {
156 		if (offset < 4)
157 			return VIRTIO_PCI_O_MSIX;
158 		else
159 			offset -= 4;
160 	}
161 
162 	*config_off = offset;
163 
164 	return VIRTIO_PCI_O_CONFIG;
165 }
166 
167 bool virtio_queue__should_signal(struct virt_queue *vq)
168 {
169 	u16 old_idx, new_idx, event_idx;
170 
171 	old_idx		= vq->last_used_signalled;
172 	new_idx		= virtio_guest_to_host_u16(vq, vq->vring.used->idx);
173 	event_idx	= virtio_guest_to_host_u16(vq, vring_used_event(&vq->vring));
174 
175 	if (vring_need_event(event_idx, new_idx, old_idx)) {
176 		vq->last_used_signalled = new_idx;
177 		return true;
178 	}
179 
180 	return false;
181 }
182 
183 int virtio_init(struct kvm *kvm, void *dev, struct virtio_device *vdev,
184 		struct virtio_ops *ops, enum virtio_trans trans,
185 		int device_id, int subsys_id, int class)
186 {
187 	void *virtio;
188 
189 	switch (trans) {
190 	case VIRTIO_PCI:
191 		virtio = calloc(sizeof(struct virtio_pci), 1);
192 		if (!virtio)
193 			return -ENOMEM;
194 		vdev->virtio			= virtio;
195 		vdev->ops			= ops;
196 		vdev->ops->signal_vq		= virtio_pci__signal_vq;
197 		vdev->ops->signal_config	= virtio_pci__signal_config;
198 		vdev->ops->init			= virtio_pci__init;
199 		vdev->ops->exit			= virtio_pci__exit;
200 		vdev->ops->init(kvm, dev, vdev, device_id, subsys_id, class);
201 		break;
202 	case VIRTIO_MMIO:
203 		virtio = calloc(sizeof(struct virtio_mmio), 1);
204 		if (!virtio)
205 			return -ENOMEM;
206 		vdev->virtio			= virtio;
207 		vdev->ops			= ops;
208 		vdev->ops->signal_vq		= virtio_mmio_signal_vq;
209 		vdev->ops->signal_config	= virtio_mmio_signal_config;
210 		vdev->ops->init			= virtio_mmio_init;
211 		vdev->ops->exit			= virtio_mmio_exit;
212 		vdev->ops->init(kvm, dev, vdev, device_id, subsys_id, class);
213 		break;
214 	default:
215 		return -1;
216 	};
217 
218 	return 0;
219 }
220 
221 int virtio_compat_add_message(const char *device, const char *config)
222 {
223 	int len = 1024;
224 	int compat_id;
225 	char *title;
226 	char *desc;
227 
228 	title = malloc(len);
229 	if (!title)
230 		return -ENOMEM;
231 
232 	desc = malloc(len);
233 	if (!desc) {
234 		free(title);
235 		return -ENOMEM;
236 	}
237 
238 	snprintf(title, len, "%s device was not detected.", device);
239 	snprintf(desc,  len, "While you have requested a %s device, "
240 			     "the guest kernel did not initialize it.\n"
241 			     "\tPlease make sure that the guest kernel was "
242 			     "compiled with %s=y enabled in .config.",
243 			     device, config);
244 
245 	compat_id = compat__add_message(title, desc);
246 
247 	free(desc);
248 	free(title);
249 
250 	return compat_id;
251 }
252