]>
Commit | Line | Data |
---|---|---|
967f97fa AL |
1 | /* |
2 | * Virtio Support | |
3 | * | |
4 | * Copyright IBM, Corp. 2007 | |
5 | * | |
6 | * Authors: | |
7 | * Anthony Liguori <aliguori@us.ibm.com> | |
8 | * | |
9 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
10 | * the COPYING file in the top-level directory. | |
11 | * | |
12 | */ | |
13 | ||
9b8bfe21 | 14 | #include "qemu/osdep.h" |
da34e65c | 15 | #include "qapi/error.h" |
4771d756 PB |
16 | #include "qemu-common.h" |
17 | #include "cpu.h" | |
64979a4d | 18 | #include "trace.h" |
fdfba1a2 | 19 | #include "exec/address-spaces.h" |
1de7afc9 | 20 | #include "qemu/error-report.h" |
0d09e41a | 21 | #include "hw/virtio/virtio.h" |
1de7afc9 | 22 | #include "qemu/atomic.h" |
0d09e41a | 23 | #include "hw/virtio/virtio-bus.h" |
6b321a3d | 24 | #include "migration/migration.h" |
cee3ca00 | 25 | #include "hw/virtio/virtio-access.h" |
967f97fa | 26 | |
6ce69d1c PM |
27 | /* |
28 | * The alignment to use between consumer and producer parts of vring. | |
29 | * x86 pagesize again. This is the default, used by transports like PCI | |
30 | * which don't provide a means for the guest to tell the host the alignment. | |
31 | */ | |
f46f15bc AL |
32 | #define VIRTIO_PCI_VRING_ALIGN 4096 |
33 | ||
967f97fa AL |
34 | typedef struct VRingDesc |
35 | { | |
36 | uint64_t addr; | |
37 | uint32_t len; | |
38 | uint16_t flags; | |
39 | uint16_t next; | |
40 | } VRingDesc; | |
41 | ||
42 | typedef struct VRingAvail | |
43 | { | |
44 | uint16_t flags; | |
45 | uint16_t idx; | |
46 | uint16_t ring[0]; | |
47 | } VRingAvail; | |
48 | ||
49 | typedef struct VRingUsedElem | |
50 | { | |
51 | uint32_t id; | |
52 | uint32_t len; | |
53 | } VRingUsedElem; | |
54 | ||
55 | typedef struct VRingUsed | |
56 | { | |
57 | uint16_t flags; | |
58 | uint16_t idx; | |
59 | VRingUsedElem ring[0]; | |
60 | } VRingUsed; | |
61 | ||
62 | typedef struct VRing | |
63 | { | |
64 | unsigned int num; | |
46c5d082 | 65 | unsigned int num_default; |
6ce69d1c | 66 | unsigned int align; |
a8170e5e AK |
67 | hwaddr desc; |
68 | hwaddr avail; | |
69 | hwaddr used; | |
967f97fa AL |
70 | } VRing; |
71 | ||
72 | struct VirtQueue | |
73 | { | |
74 | VRing vring; | |
be1fea9b VM |
75 | |
76 | /* Next head to pop */ | |
967f97fa | 77 | uint16_t last_avail_idx; |
b796fcd1 | 78 | |
be1fea9b VM |
79 | /* Last avail_idx read from VQ. */ |
80 | uint16_t shadow_avail_idx; | |
81 | ||
b796fcd1 VM |
82 | uint16_t used_idx; |
83 | ||
bcbabae8 MT |
84 | /* Last used index value we have signalled on */ |
85 | uint16_t signalled_used; | |
86 | ||
87 | /* Last used index value we have signalled on */ | |
88 | bool signalled_used_valid; | |
89 | ||
90 | /* Notification enabled? */ | |
91 | bool notification; | |
92 | ||
e78a2b42 JW |
93 | uint16_t queue_index; |
94 | ||
967f97fa | 95 | int inuse; |
bcbabae8 | 96 | |
7055e687 | 97 | uint16_t vector; |
967f97fa | 98 | void (*handle_output)(VirtIODevice *vdev, VirtQueue *vq); |
344dc16f | 99 | void (*handle_aio_output)(VirtIODevice *vdev, VirtQueue *vq); |
1cbdabe2 MT |
100 | VirtIODevice *vdev; |
101 | EventNotifier guest_notifier; | |
102 | EventNotifier host_notifier; | |
e0d686bf | 103 | QLIST_ENTRY(VirtQueue) node; |
967f97fa AL |
104 | }; |
105 | ||
967f97fa | 106 | /* virt queue functions */ |
ab223c95 | 107 | void virtio_queue_update_rings(VirtIODevice *vdev, int n) |
967f97fa | 108 | { |
ab223c95 | 109 | VRing *vring = &vdev->vq[n].vring; |
53c25cea | 110 | |
ab223c95 CH |
111 | if (!vring->desc) { |
112 | /* not yet setup -> nothing to do */ | |
113 | return; | |
114 | } | |
115 | vring->avail = vring->desc + vring->num * sizeof(VRingDesc); | |
116 | vring->used = vring_align(vring->avail + | |
117 | offsetof(VRingAvail, ring[vring->num]), | |
118 | vring->align); | |
967f97fa AL |
119 | } |
120 | ||
aa570d6f PB |
121 | static void vring_desc_read(VirtIODevice *vdev, VRingDesc *desc, |
122 | hwaddr desc_pa, int i) | |
967f97fa | 123 | { |
aa570d6f PB |
124 | address_space_read(&address_space_memory, desc_pa + i * sizeof(VRingDesc), |
125 | MEMTXATTRS_UNSPECIFIED, (void *)desc, sizeof(VRingDesc)); | |
126 | virtio_tswap64s(vdev, &desc->addr); | |
127 | virtio_tswap32s(vdev, &desc->len); | |
128 | virtio_tswap16s(vdev, &desc->flags); | |
129 | virtio_tswap16s(vdev, &desc->next); | |
967f97fa AL |
130 | } |
131 | ||
132 | static inline uint16_t vring_avail_flags(VirtQueue *vq) | |
133 | { | |
a8170e5e | 134 | hwaddr pa; |
967f97fa | 135 | pa = vq->vring.avail + offsetof(VRingAvail, flags); |
cee3ca00 | 136 | return virtio_lduw_phys(vq->vdev, pa); |
967f97fa AL |
137 | } |
138 | ||
139 | static inline uint16_t vring_avail_idx(VirtQueue *vq) | |
140 | { | |
a8170e5e | 141 | hwaddr pa; |
967f97fa | 142 | pa = vq->vring.avail + offsetof(VRingAvail, idx); |
be1fea9b VM |
143 | vq->shadow_avail_idx = virtio_lduw_phys(vq->vdev, pa); |
144 | return vq->shadow_avail_idx; | |
967f97fa AL |
145 | } |
146 | ||
147 | static inline uint16_t vring_avail_ring(VirtQueue *vq, int i) | |
148 | { | |
a8170e5e | 149 | hwaddr pa; |
967f97fa | 150 | pa = vq->vring.avail + offsetof(VRingAvail, ring[i]); |
cee3ca00 | 151 | return virtio_lduw_phys(vq->vdev, pa); |
967f97fa AL |
152 | } |
153 | ||
e9600c6c | 154 | static inline uint16_t vring_get_used_event(VirtQueue *vq) |
bcbabae8 MT |
155 | { |
156 | return vring_avail_ring(vq, vq->vring.num); | |
157 | } | |
158 | ||
1cdd2ee5 VM |
159 | static inline void vring_used_write(VirtQueue *vq, VRingUsedElem *uelem, |
160 | int i) | |
967f97fa | 161 | { |
a8170e5e | 162 | hwaddr pa; |
1cdd2ee5 VM |
163 | virtio_tswap32s(vq->vdev, &uelem->id); |
164 | virtio_tswap32s(vq->vdev, &uelem->len); | |
165 | pa = vq->vring.used + offsetof(VRingUsed, ring[i]); | |
166 | address_space_write(&address_space_memory, pa, MEMTXATTRS_UNSPECIFIED, | |
167 | (void *)uelem, sizeof(VRingUsedElem)); | |
967f97fa AL |
168 | } |
169 | ||
170 | static uint16_t vring_used_idx(VirtQueue *vq) | |
171 | { | |
a8170e5e | 172 | hwaddr pa; |
967f97fa | 173 | pa = vq->vring.used + offsetof(VRingUsed, idx); |
cee3ca00 | 174 | return virtio_lduw_phys(vq->vdev, pa); |
967f97fa AL |
175 | } |
176 | ||
bcbabae8 | 177 | static inline void vring_used_idx_set(VirtQueue *vq, uint16_t val) |
967f97fa | 178 | { |
a8170e5e | 179 | hwaddr pa; |
967f97fa | 180 | pa = vq->vring.used + offsetof(VRingUsed, idx); |
cee3ca00 | 181 | virtio_stw_phys(vq->vdev, pa, val); |
b796fcd1 | 182 | vq->used_idx = val; |
967f97fa AL |
183 | } |
184 | ||
185 | static inline void vring_used_flags_set_bit(VirtQueue *vq, int mask) | |
186 | { | |
cee3ca00 | 187 | VirtIODevice *vdev = vq->vdev; |
a8170e5e | 188 | hwaddr pa; |
967f97fa | 189 | pa = vq->vring.used + offsetof(VRingUsed, flags); |
cee3ca00 | 190 | virtio_stw_phys(vdev, pa, virtio_lduw_phys(vdev, pa) | mask); |
967f97fa AL |
191 | } |
192 | ||
193 | static inline void vring_used_flags_unset_bit(VirtQueue *vq, int mask) | |
194 | { | |
cee3ca00 | 195 | VirtIODevice *vdev = vq->vdev; |
a8170e5e | 196 | hwaddr pa; |
967f97fa | 197 | pa = vq->vring.used + offsetof(VRingUsed, flags); |
cee3ca00 | 198 | virtio_stw_phys(vdev, pa, virtio_lduw_phys(vdev, pa) & ~mask); |
967f97fa AL |
199 | } |
200 | ||
e9600c6c | 201 | static inline void vring_set_avail_event(VirtQueue *vq, uint16_t val) |
bcbabae8 | 202 | { |
a8170e5e | 203 | hwaddr pa; |
bcbabae8 MT |
204 | if (!vq->notification) { |
205 | return; | |
206 | } | |
207 | pa = vq->vring.used + offsetof(VRingUsed, ring[vq->vring.num]); | |
cee3ca00 | 208 | virtio_stw_phys(vq->vdev, pa, val); |
bcbabae8 MT |
209 | } |
210 | ||
967f97fa AL |
211 | void virtio_queue_set_notification(VirtQueue *vq, int enable) |
212 | { | |
bcbabae8 | 213 | vq->notification = enable; |
95129d6f | 214 | if (virtio_vdev_has_feature(vq->vdev, VIRTIO_RING_F_EVENT_IDX)) { |
e9600c6c | 215 | vring_set_avail_event(vq, vring_avail_idx(vq)); |
bcbabae8 | 216 | } else if (enable) { |
967f97fa | 217 | vring_used_flags_unset_bit(vq, VRING_USED_F_NO_NOTIFY); |
bcbabae8 | 218 | } else { |
967f97fa | 219 | vring_used_flags_set_bit(vq, VRING_USED_F_NO_NOTIFY); |
bcbabae8 | 220 | } |
92045d80 MT |
221 | if (enable) { |
222 | /* Expose avail event/used flags before caller checks the avail idx. */ | |
223 | smp_mb(); | |
224 | } | |
967f97fa AL |
225 | } |
226 | ||
227 | int virtio_queue_ready(VirtQueue *vq) | |
228 | { | |
229 | return vq->vring.avail != 0; | |
230 | } | |
231 | ||
be1fea9b VM |
232 | /* Fetch avail_idx from VQ memory only when we really need to know if |
233 | * guest has added some buffers. */ | |
967f97fa AL |
234 | int virtio_queue_empty(VirtQueue *vq) |
235 | { | |
be1fea9b VM |
236 | if (vq->shadow_avail_idx != vq->last_avail_idx) { |
237 | return 0; | |
238 | } | |
239 | ||
967f97fa AL |
240 | return vring_avail_idx(vq) == vq->last_avail_idx; |
241 | } | |
242 | ||
ce317461 JW |
243 | static void virtqueue_unmap_sg(VirtQueue *vq, const VirtQueueElement *elem, |
244 | unsigned int len) | |
967f97fa AL |
245 | { |
246 | unsigned int offset; | |
247 | int i; | |
248 | ||
967f97fa AL |
249 | offset = 0; |
250 | for (i = 0; i < elem->in_num; i++) { | |
251 | size_t size = MIN(len - offset, elem->in_sg[i].iov_len); | |
252 | ||
26b258e1 AL |
253 | cpu_physical_memory_unmap(elem->in_sg[i].iov_base, |
254 | elem->in_sg[i].iov_len, | |
255 | 1, size); | |
967f97fa | 256 | |
0cea71a2 | 257 | offset += size; |
967f97fa AL |
258 | } |
259 | ||
26b258e1 AL |
260 | for (i = 0; i < elem->out_num; i++) |
261 | cpu_physical_memory_unmap(elem->out_sg[i].iov_base, | |
262 | elem->out_sg[i].iov_len, | |
263 | 0, elem->out_sg[i].iov_len); | |
ce317461 JW |
264 | } |
265 | ||
29b9f5ef JW |
266 | void virtqueue_discard(VirtQueue *vq, const VirtQueueElement *elem, |
267 | unsigned int len) | |
268 | { | |
269 | vq->last_avail_idx--; | |
270 | virtqueue_unmap_sg(vq, elem, len); | |
271 | } | |
272 | ||
ce317461 JW |
273 | void virtqueue_fill(VirtQueue *vq, const VirtQueueElement *elem, |
274 | unsigned int len, unsigned int idx) | |
275 | { | |
1cdd2ee5 VM |
276 | VRingUsedElem uelem; |
277 | ||
ce317461 JW |
278 | trace_virtqueue_fill(vq, elem, len, idx); |
279 | ||
280 | virtqueue_unmap_sg(vq, elem, len); | |
26b258e1 | 281 | |
b796fcd1 | 282 | idx = (idx + vq->used_idx) % vq->vring.num; |
967f97fa | 283 | |
1cdd2ee5 VM |
284 | uelem.id = elem->index; |
285 | uelem.len = len; | |
286 | vring_used_write(vq, &uelem, idx); | |
967f97fa AL |
287 | } |
288 | ||
289 | void virtqueue_flush(VirtQueue *vq, unsigned int count) | |
290 | { | |
bcbabae8 | 291 | uint16_t old, new; |
967f97fa | 292 | /* Make sure buffer is written before we update index. */ |
b90d2f35 | 293 | smp_wmb(); |
64979a4d | 294 | trace_virtqueue_flush(vq, count); |
b796fcd1 | 295 | old = vq->used_idx; |
bcbabae8 MT |
296 | new = old + count; |
297 | vring_used_idx_set(vq, new); | |
967f97fa | 298 | vq->inuse -= count; |
bcbabae8 MT |
299 | if (unlikely((int16_t)(new - vq->signalled_used) < (uint16_t)(new - old))) |
300 | vq->signalled_used_valid = false; | |
967f97fa AL |
301 | } |
302 | ||
303 | void virtqueue_push(VirtQueue *vq, const VirtQueueElement *elem, | |
304 | unsigned int len) | |
305 | { | |
306 | virtqueue_fill(vq, elem, len, 0); | |
307 | virtqueue_flush(vq, 1); | |
308 | } | |
309 | ||
310 | static int virtqueue_num_heads(VirtQueue *vq, unsigned int idx) | |
311 | { | |
312 | uint16_t num_heads = vring_avail_idx(vq) - idx; | |
313 | ||
314 | /* Check it isn't doing very strange things with descriptor numbers. */ | |
bb6834cf | 315 | if (num_heads > vq->vring.num) { |
ce67ed65 | 316 | error_report("Guest moved used index from %u to %u", |
be1fea9b | 317 | idx, vq->shadow_avail_idx); |
bb6834cf AL |
318 | exit(1); |
319 | } | |
a821ce59 MT |
320 | /* On success, callers read a descriptor at vq->last_avail_idx. |
321 | * Make sure descriptor read does not bypass avail index read. */ | |
322 | if (num_heads) { | |
323 | smp_rmb(); | |
324 | } | |
967f97fa AL |
325 | |
326 | return num_heads; | |
327 | } | |
328 | ||
329 | static unsigned int virtqueue_get_head(VirtQueue *vq, unsigned int idx) | |
330 | { | |
331 | unsigned int head; | |
332 | ||
333 | /* Grab the next descriptor number they're advertising, and increment | |
334 | * the index we've seen. */ | |
335 | head = vring_avail_ring(vq, idx % vq->vring.num); | |
336 | ||
337 | /* If their number is silly, that's a fatal mistake. */ | |
bb6834cf | 338 | if (head >= vq->vring.num) { |
ce67ed65 | 339 | error_report("Guest says index %u is available", head); |
bb6834cf AL |
340 | exit(1); |
341 | } | |
967f97fa AL |
342 | |
343 | return head; | |
344 | } | |
345 | ||
aa570d6f PB |
346 | static unsigned virtqueue_read_next_desc(VirtIODevice *vdev, VRingDesc *desc, |
347 | hwaddr desc_pa, unsigned int max) | |
967f97fa AL |
348 | { |
349 | unsigned int next; | |
350 | ||
351 | /* If this descriptor says it doesn't chain, we're done. */ | |
aa570d6f | 352 | if (!(desc->flags & VRING_DESC_F_NEXT)) { |
5774cf98 | 353 | return max; |
cee3ca00 | 354 | } |
967f97fa AL |
355 | |
356 | /* Check they're not leading us off end of descriptors. */ | |
aa570d6f | 357 | next = desc->next; |
967f97fa | 358 | /* Make sure compiler knows to grab that: we don't want it changing! */ |
b90d2f35 | 359 | smp_wmb(); |
967f97fa | 360 | |
5774cf98 | 361 | if (next >= max) { |
ce67ed65 | 362 | error_report("Desc next is %u", next); |
bb6834cf AL |
363 | exit(1); |
364 | } | |
967f97fa | 365 | |
aa570d6f | 366 | vring_desc_read(vdev, desc, desc_pa, next); |
967f97fa AL |
367 | return next; |
368 | } | |
369 | ||
0d8d7690 | 370 | void virtqueue_get_avail_bytes(VirtQueue *vq, unsigned int *in_bytes, |
e1f7b481 MT |
371 | unsigned int *out_bytes, |
372 | unsigned max_in_bytes, unsigned max_out_bytes) | |
967f97fa | 373 | { |
efeea6d0 | 374 | unsigned int idx; |
385ce95d | 375 | unsigned int total_bufs, in_total, out_total; |
967f97fa AL |
376 | |
377 | idx = vq->last_avail_idx; | |
378 | ||
efeea6d0 | 379 | total_bufs = in_total = out_total = 0; |
967f97fa | 380 | while (virtqueue_num_heads(vq, idx)) { |
cee3ca00 | 381 | VirtIODevice *vdev = vq->vdev; |
efeea6d0 | 382 | unsigned int max, num_bufs, indirect = 0; |
aa570d6f | 383 | VRingDesc desc; |
a8170e5e | 384 | hwaddr desc_pa; |
967f97fa AL |
385 | int i; |
386 | ||
efeea6d0 MM |
387 | max = vq->vring.num; |
388 | num_bufs = total_bufs; | |
967f97fa | 389 | i = virtqueue_get_head(vq, idx++); |
efeea6d0 | 390 | desc_pa = vq->vring.desc; |
aa570d6f | 391 | vring_desc_read(vdev, &desc, desc_pa, i); |
efeea6d0 | 392 | |
aa570d6f PB |
393 | if (desc.flags & VRING_DESC_F_INDIRECT) { |
394 | if (desc.len % sizeof(VRingDesc)) { | |
ce67ed65 | 395 | error_report("Invalid size for indirect buffer table"); |
efeea6d0 MM |
396 | exit(1); |
397 | } | |
398 | ||
399 | /* If we've got too many, that implies a descriptor loop. */ | |
400 | if (num_bufs >= max) { | |
ce67ed65 | 401 | error_report("Looped descriptor"); |
efeea6d0 MM |
402 | exit(1); |
403 | } | |
404 | ||
405 | /* loop over the indirect descriptor table */ | |
406 | indirect = 1; | |
aa570d6f PB |
407 | max = desc.len / sizeof(VRingDesc); |
408 | desc_pa = desc.addr; | |
1ae2757c | 409 | num_bufs = i = 0; |
aa570d6f | 410 | vring_desc_read(vdev, &desc, desc_pa, i); |
efeea6d0 MM |
411 | } |
412 | ||
967f97fa AL |
413 | do { |
414 | /* If we've got too many, that implies a descriptor loop. */ | |
5774cf98 | 415 | if (++num_bufs > max) { |
ce67ed65 | 416 | error_report("Looped descriptor"); |
bb6834cf AL |
417 | exit(1); |
418 | } | |
967f97fa | 419 | |
aa570d6f PB |
420 | if (desc.flags & VRING_DESC_F_WRITE) { |
421 | in_total += desc.len; | |
967f97fa | 422 | } else { |
aa570d6f | 423 | out_total += desc.len; |
967f97fa | 424 | } |
e1f7b481 MT |
425 | if (in_total >= max_in_bytes && out_total >= max_out_bytes) { |
426 | goto done; | |
427 | } | |
aa570d6f | 428 | } while ((i = virtqueue_read_next_desc(vdev, &desc, desc_pa, max)) != max); |
efeea6d0 MM |
429 | |
430 | if (!indirect) | |
431 | total_bufs = num_bufs; | |
432 | else | |
433 | total_bufs++; | |
967f97fa | 434 | } |
e1f7b481 | 435 | done: |
0d8d7690 AS |
436 | if (in_bytes) { |
437 | *in_bytes = in_total; | |
438 | } | |
439 | if (out_bytes) { | |
440 | *out_bytes = out_total; | |
441 | } | |
442 | } | |
967f97fa | 443 | |
0d8d7690 AS |
444 | int virtqueue_avail_bytes(VirtQueue *vq, unsigned int in_bytes, |
445 | unsigned int out_bytes) | |
446 | { | |
447 | unsigned int in_total, out_total; | |
448 | ||
e1f7b481 MT |
449 | virtqueue_get_avail_bytes(vq, &in_total, &out_total, in_bytes, out_bytes); |
450 | return in_bytes <= in_total && out_bytes <= out_total; | |
967f97fa AL |
451 | } |
452 | ||
3b3b0628 PB |
453 | static void virtqueue_map_desc(unsigned int *p_num_sg, hwaddr *addr, struct iovec *iov, |
454 | unsigned int max_num_sg, bool is_write, | |
455 | hwaddr pa, size_t sz) | |
456 | { | |
457 | unsigned num_sg = *p_num_sg; | |
458 | assert(num_sg <= max_num_sg); | |
459 | ||
460 | while (sz) { | |
461 | hwaddr len = sz; | |
462 | ||
463 | if (num_sg == max_num_sg) { | |
464 | error_report("virtio: too many write descriptors in indirect table"); | |
465 | exit(1); | |
466 | } | |
467 | ||
468 | iov[num_sg].iov_base = cpu_physical_memory_map(pa, &len, is_write); | |
469 | iov[num_sg].iov_len = len; | |
470 | addr[num_sg] = pa; | |
471 | ||
472 | sz -= len; | |
473 | pa += len; | |
474 | num_sg++; | |
475 | } | |
476 | *p_num_sg = num_sg; | |
477 | } | |
478 | ||
8059feee MT |
479 | static void virtqueue_map_iovec(struct iovec *sg, hwaddr *addr, |
480 | unsigned int *num_sg, unsigned int max_size, | |
481 | int is_write) | |
42fb2e07 KW |
482 | { |
483 | unsigned int i; | |
a8170e5e | 484 | hwaddr len; |
42fb2e07 | 485 | |
8059feee MT |
486 | /* Note: this function MUST validate input, some callers |
487 | * are passing in num_sg values received over the network. | |
488 | */ | |
489 | /* TODO: teach all callers that this can fail, and return failure instead | |
490 | * of asserting here. | |
491 | * When we do, we might be able to re-enable NDEBUG below. | |
492 | */ | |
493 | #ifdef NDEBUG | |
494 | #error building with NDEBUG is not supported | |
495 | #endif | |
496 | assert(*num_sg <= max_size); | |
497 | ||
498 | for (i = 0; i < *num_sg; i++) { | |
42fb2e07 KW |
499 | len = sg[i].iov_len; |
500 | sg[i].iov_base = cpu_physical_memory_map(addr[i], &len, is_write); | |
8059feee | 501 | if (!sg[i].iov_base) { |
1a285899 | 502 | error_report("virtio: error trying to map MMIO memory"); |
42fb2e07 KW |
503 | exit(1); |
504 | } | |
3b3b0628 PB |
505 | if (len != sg[i].iov_len) { |
506 | error_report("virtio: unexpected memory split"); | |
8059feee MT |
507 | exit(1); |
508 | } | |
42fb2e07 KW |
509 | } |
510 | } | |
511 | ||
8059feee MT |
512 | void virtqueue_map(VirtQueueElement *elem) |
513 | { | |
514 | virtqueue_map_iovec(elem->in_sg, elem->in_addr, &elem->in_num, | |
3724650d | 515 | VIRTQUEUE_MAX_SIZE, 1); |
8059feee | 516 | virtqueue_map_iovec(elem->out_sg, elem->out_addr, &elem->out_num, |
3724650d PB |
517 | VIRTQUEUE_MAX_SIZE, 0); |
518 | } | |
519 | ||
520 | void *virtqueue_alloc_element(size_t sz, unsigned out_num, unsigned in_num) | |
521 | { | |
522 | VirtQueueElement *elem; | |
523 | size_t in_addr_ofs = QEMU_ALIGN_UP(sz, __alignof__(elem->in_addr[0])); | |
524 | size_t out_addr_ofs = in_addr_ofs + in_num * sizeof(elem->in_addr[0]); | |
525 | size_t out_addr_end = out_addr_ofs + out_num * sizeof(elem->out_addr[0]); | |
526 | size_t in_sg_ofs = QEMU_ALIGN_UP(out_addr_end, __alignof__(elem->in_sg[0])); | |
527 | size_t out_sg_ofs = in_sg_ofs + in_num * sizeof(elem->in_sg[0]); | |
528 | size_t out_sg_end = out_sg_ofs + out_num * sizeof(elem->out_sg[0]); | |
529 | ||
530 | assert(sz >= sizeof(VirtQueueElement)); | |
531 | elem = g_malloc(out_sg_end); | |
532 | elem->out_num = out_num; | |
533 | elem->in_num = in_num; | |
534 | elem->in_addr = (void *)elem + in_addr_ofs; | |
535 | elem->out_addr = (void *)elem + out_addr_ofs; | |
536 | elem->in_sg = (void *)elem + in_sg_ofs; | |
537 | elem->out_sg = (void *)elem + out_sg_ofs; | |
538 | return elem; | |
8059feee MT |
539 | } |
540 | ||
51b19ebe | 541 | void *virtqueue_pop(VirtQueue *vq, size_t sz) |
967f97fa | 542 | { |
5774cf98 | 543 | unsigned int i, head, max; |
a8170e5e | 544 | hwaddr desc_pa = vq->vring.desc; |
cee3ca00 | 545 | VirtIODevice *vdev = vq->vdev; |
51b19ebe | 546 | VirtQueueElement *elem; |
3b3b0628 PB |
547 | unsigned out_num, in_num; |
548 | hwaddr addr[VIRTQUEUE_MAX_SIZE]; | |
549 | struct iovec iov[VIRTQUEUE_MAX_SIZE]; | |
aa570d6f | 550 | VRingDesc desc; |
967f97fa | 551 | |
be1fea9b | 552 | if (virtio_queue_empty(vq)) { |
51b19ebe PB |
553 | return NULL; |
554 | } | |
be1fea9b VM |
555 | /* Needed after virtio_queue_empty(), see comment in |
556 | * virtqueue_num_heads(). */ | |
557 | smp_rmb(); | |
967f97fa AL |
558 | |
559 | /* When we start there are none of either input nor output. */ | |
3b3b0628 | 560 | out_num = in_num = 0; |
967f97fa | 561 | |
5774cf98 MM |
562 | max = vq->vring.num; |
563 | ||
967f97fa | 564 | i = head = virtqueue_get_head(vq, vq->last_avail_idx++); |
95129d6f | 565 | if (virtio_vdev_has_feature(vdev, VIRTIO_RING_F_EVENT_IDX)) { |
e9600c6c | 566 | vring_set_avail_event(vq, vq->last_avail_idx); |
bcbabae8 | 567 | } |
efeea6d0 | 568 | |
aa570d6f PB |
569 | vring_desc_read(vdev, &desc, desc_pa, i); |
570 | if (desc.flags & VRING_DESC_F_INDIRECT) { | |
571 | if (desc.len % sizeof(VRingDesc)) { | |
ce67ed65 | 572 | error_report("Invalid size for indirect buffer table"); |
efeea6d0 MM |
573 | exit(1); |
574 | } | |
575 | ||
576 | /* loop over the indirect descriptor table */ | |
aa570d6f PB |
577 | max = desc.len / sizeof(VRingDesc); |
578 | desc_pa = desc.addr; | |
efeea6d0 | 579 | i = 0; |
aa570d6f | 580 | vring_desc_read(vdev, &desc, desc_pa, i); |
efeea6d0 MM |
581 | } |
582 | ||
42fb2e07 | 583 | /* Collect all the descriptors */ |
967f97fa | 584 | do { |
aa570d6f | 585 | if (desc.flags & VRING_DESC_F_WRITE) { |
3b3b0628 | 586 | virtqueue_map_desc(&in_num, addr + out_num, iov + out_num, |
aa570d6f | 587 | VIRTQUEUE_MAX_SIZE - out_num, true, desc.addr, desc.len); |
42fb2e07 | 588 | } else { |
3b3b0628 PB |
589 | if (in_num) { |
590 | error_report("Incorrect order for descriptors"); | |
c8eac1cf MT |
591 | exit(1); |
592 | } | |
3b3b0628 | 593 | virtqueue_map_desc(&out_num, addr, iov, |
aa570d6f | 594 | VIRTQUEUE_MAX_SIZE, false, desc.addr, desc.len); |
42fb2e07 | 595 | } |
967f97fa | 596 | |
967f97fa | 597 | /* If we've got too many, that implies a descriptor loop. */ |
3b3b0628 | 598 | if ((in_num + out_num) > max) { |
ce67ed65 | 599 | error_report("Looped descriptor"); |
bb6834cf AL |
600 | exit(1); |
601 | } | |
aa570d6f | 602 | } while ((i = virtqueue_read_next_desc(vdev, &desc, desc_pa, max)) != max); |
967f97fa | 603 | |
3b3b0628 PB |
604 | /* Now copy what we have collected and mapped */ |
605 | elem = virtqueue_alloc_element(sz, out_num, in_num); | |
967f97fa | 606 | elem->index = head; |
3b3b0628 PB |
607 | for (i = 0; i < out_num; i++) { |
608 | elem->out_addr[i] = addr[i]; | |
609 | elem->out_sg[i] = iov[i]; | |
610 | } | |
611 | for (i = 0; i < in_num; i++) { | |
612 | elem->in_addr[i] = addr[out_num + i]; | |
613 | elem->in_sg[i] = iov[out_num + i]; | |
614 | } | |
967f97fa AL |
615 | |
616 | vq->inuse++; | |
617 | ||
64979a4d | 618 | trace_virtqueue_pop(vq, elem, elem->in_num, elem->out_num); |
51b19ebe | 619 | return elem; |
967f97fa AL |
620 | } |
621 | ||
3724650d PB |
622 | /* Reading and writing a structure directly to QEMUFile is *awful*, but |
623 | * it is what QEMU has always done by mistake. We can change it sooner | |
624 | * or later by bumping the version number of the affected vm states. | |
625 | * In the meanwhile, since the in-memory layout of VirtQueueElement | |
626 | * has changed, we need to marshal to and from the layout that was | |
627 | * used before the change. | |
628 | */ | |
629 | typedef struct VirtQueueElementOld { | |
630 | unsigned int index; | |
631 | unsigned int out_num; | |
632 | unsigned int in_num; | |
633 | hwaddr in_addr[VIRTQUEUE_MAX_SIZE]; | |
634 | hwaddr out_addr[VIRTQUEUE_MAX_SIZE]; | |
635 | struct iovec in_sg[VIRTQUEUE_MAX_SIZE]; | |
636 | struct iovec out_sg[VIRTQUEUE_MAX_SIZE]; | |
637 | } VirtQueueElementOld; | |
638 | ||
ab281c17 PB |
639 | void *qemu_get_virtqueue_element(QEMUFile *f, size_t sz) |
640 | { | |
3724650d PB |
641 | VirtQueueElement *elem; |
642 | VirtQueueElementOld data; | |
643 | int i; | |
644 | ||
645 | qemu_get_buffer(f, (uint8_t *)&data, sizeof(VirtQueueElementOld)); | |
646 | ||
647 | elem = virtqueue_alloc_element(sz, data.out_num, data.in_num); | |
648 | elem->index = data.index; | |
649 | ||
650 | for (i = 0; i < elem->in_num; i++) { | |
651 | elem->in_addr[i] = data.in_addr[i]; | |
652 | } | |
653 | ||
654 | for (i = 0; i < elem->out_num; i++) { | |
655 | elem->out_addr[i] = data.out_addr[i]; | |
656 | } | |
657 | ||
658 | for (i = 0; i < elem->in_num; i++) { | |
659 | /* Base is overwritten by virtqueue_map. */ | |
660 | elem->in_sg[i].iov_base = 0; | |
661 | elem->in_sg[i].iov_len = data.in_sg[i].iov_len; | |
662 | } | |
663 | ||
664 | for (i = 0; i < elem->out_num; i++) { | |
665 | /* Base is overwritten by virtqueue_map. */ | |
666 | elem->out_sg[i].iov_base = 0; | |
667 | elem->out_sg[i].iov_len = data.out_sg[i].iov_len; | |
668 | } | |
669 | ||
ab281c17 PB |
670 | virtqueue_map(elem); |
671 | return elem; | |
672 | } | |
673 | ||
674 | void qemu_put_virtqueue_element(QEMUFile *f, VirtQueueElement *elem) | |
675 | { | |
3724650d PB |
676 | VirtQueueElementOld data; |
677 | int i; | |
678 | ||
679 | memset(&data, 0, sizeof(data)); | |
680 | data.index = elem->index; | |
681 | data.in_num = elem->in_num; | |
682 | data.out_num = elem->out_num; | |
683 | ||
684 | for (i = 0; i < elem->in_num; i++) { | |
685 | data.in_addr[i] = elem->in_addr[i]; | |
686 | } | |
687 | ||
688 | for (i = 0; i < elem->out_num; i++) { | |
689 | data.out_addr[i] = elem->out_addr[i]; | |
690 | } | |
691 | ||
692 | for (i = 0; i < elem->in_num; i++) { | |
693 | /* Base is overwritten by virtqueue_map when loading. Do not | |
694 | * save it, as it would leak the QEMU address space layout. */ | |
695 | data.in_sg[i].iov_len = elem->in_sg[i].iov_len; | |
696 | } | |
697 | ||
698 | for (i = 0; i < elem->out_num; i++) { | |
699 | /* Do not save iov_base as above. */ | |
700 | data.out_sg[i].iov_len = elem->out_sg[i].iov_len; | |
701 | } | |
702 | qemu_put_buffer(f, (uint8_t *)&data, sizeof(VirtQueueElementOld)); | |
ab281c17 PB |
703 | } |
704 | ||
967f97fa | 705 | /* virtio device */ |
7055e687 MT |
706 | static void virtio_notify_vector(VirtIODevice *vdev, uint16_t vector) |
707 | { | |
1c819449 FK |
708 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
709 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
710 | ||
711 | if (k->notify) { | |
712 | k->notify(qbus->parent, vector); | |
7055e687 MT |
713 | } |
714 | } | |
967f97fa | 715 | |
53c25cea | 716 | void virtio_update_irq(VirtIODevice *vdev) |
967f97fa | 717 | { |
7055e687 | 718 | virtio_notify_vector(vdev, VIRTIO_NO_VECTOR); |
967f97fa AL |
719 | } |
720 | ||
0b352fd6 CH |
721 | static int virtio_validate_features(VirtIODevice *vdev) |
722 | { | |
723 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
724 | ||
725 | if (k->validate_features) { | |
726 | return k->validate_features(vdev); | |
727 | } else { | |
728 | return 0; | |
729 | } | |
730 | } | |
731 | ||
732 | int virtio_set_status(VirtIODevice *vdev, uint8_t val) | |
4e1837f8 | 733 | { |
181103cd | 734 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
4e1837f8 SH |
735 | trace_virtio_set_status(vdev, val); |
736 | ||
95129d6f | 737 | if (virtio_vdev_has_feature(vdev, VIRTIO_F_VERSION_1)) { |
0b352fd6 CH |
738 | if (!(vdev->status & VIRTIO_CONFIG_S_FEATURES_OK) && |
739 | val & VIRTIO_CONFIG_S_FEATURES_OK) { | |
740 | int ret = virtio_validate_features(vdev); | |
741 | ||
742 | if (ret) { | |
743 | return ret; | |
744 | } | |
745 | } | |
746 | } | |
181103cd FK |
747 | if (k->set_status) { |
748 | k->set_status(vdev, val); | |
4e1837f8 SH |
749 | } |
750 | vdev->status = val; | |
0b352fd6 | 751 | return 0; |
4e1837f8 SH |
752 | } |
753 | ||
616a6552 GK |
754 | bool target_words_bigendian(void); |
755 | static enum virtio_device_endian virtio_default_endian(void) | |
756 | { | |
757 | if (target_words_bigendian()) { | |
758 | return VIRTIO_DEVICE_ENDIAN_BIG; | |
759 | } else { | |
760 | return VIRTIO_DEVICE_ENDIAN_LITTLE; | |
761 | } | |
762 | } | |
763 | ||
764 | static enum virtio_device_endian virtio_current_cpu_endian(void) | |
765 | { | |
766 | CPUClass *cc = CPU_GET_CLASS(current_cpu); | |
767 | ||
768 | if (cc->virtio_is_big_endian(current_cpu)) { | |
769 | return VIRTIO_DEVICE_ENDIAN_BIG; | |
770 | } else { | |
771 | return VIRTIO_DEVICE_ENDIAN_LITTLE; | |
772 | } | |
773 | } | |
774 | ||
53c25cea | 775 | void virtio_reset(void *opaque) |
967f97fa AL |
776 | { |
777 | VirtIODevice *vdev = opaque; | |
181103cd | 778 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
779 | int i; |
780 | ||
e0c472d8 | 781 | virtio_set_status(vdev, 0); |
616a6552 GK |
782 | if (current_cpu) { |
783 | /* Guest initiated reset */ | |
784 | vdev->device_endian = virtio_current_cpu_endian(); | |
785 | } else { | |
786 | /* System reset */ | |
787 | vdev->device_endian = virtio_default_endian(); | |
788 | } | |
e0c472d8 | 789 | |
181103cd FK |
790 | if (k->reset) { |
791 | k->reset(vdev); | |
792 | } | |
967f97fa | 793 | |
704a76fc | 794 | vdev->guest_features = 0; |
967f97fa AL |
795 | vdev->queue_sel = 0; |
796 | vdev->status = 0; | |
797 | vdev->isr = 0; | |
7055e687 MT |
798 | vdev->config_vector = VIRTIO_NO_VECTOR; |
799 | virtio_notify_vector(vdev, vdev->config_vector); | |
967f97fa | 800 | |
87b3bd1c | 801 | for(i = 0; i < VIRTIO_QUEUE_MAX; i++) { |
967f97fa AL |
802 | vdev->vq[i].vring.desc = 0; |
803 | vdev->vq[i].vring.avail = 0; | |
804 | vdev->vq[i].vring.used = 0; | |
805 | vdev->vq[i].last_avail_idx = 0; | |
be1fea9b | 806 | vdev->vq[i].shadow_avail_idx = 0; |
b796fcd1 | 807 | vdev->vq[i].used_idx = 0; |
e0d686bf | 808 | virtio_queue_set_vector(vdev, i, VIRTIO_NO_VECTOR); |
bcbabae8 MT |
809 | vdev->vq[i].signalled_used = 0; |
810 | vdev->vq[i].signalled_used_valid = false; | |
811 | vdev->vq[i].notification = true; | |
46c5d082 | 812 | vdev->vq[i].vring.num = vdev->vq[i].vring.num_default; |
967f97fa AL |
813 | } |
814 | } | |
815 | ||
53c25cea | 816 | uint32_t virtio_config_readb(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 817 | { |
181103cd | 818 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
819 | uint8_t val; |
820 | ||
5f5a1318 | 821 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 822 | return (uint32_t)-1; |
5f5a1318 JW |
823 | } |
824 | ||
825 | k->get_config(vdev, vdev->config); | |
967f97fa | 826 | |
06dbfc6f | 827 | val = ldub_p(vdev->config + addr); |
967f97fa AL |
828 | return val; |
829 | } | |
830 | ||
53c25cea | 831 | uint32_t virtio_config_readw(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 832 | { |
181103cd | 833 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
834 | uint16_t val; |
835 | ||
5f5a1318 | 836 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 837 | return (uint32_t)-1; |
5f5a1318 JW |
838 | } |
839 | ||
840 | k->get_config(vdev, vdev->config); | |
967f97fa | 841 | |
06dbfc6f | 842 | val = lduw_p(vdev->config + addr); |
967f97fa AL |
843 | return val; |
844 | } | |
845 | ||
53c25cea | 846 | uint32_t virtio_config_readl(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 847 | { |
181103cd | 848 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
849 | uint32_t val; |
850 | ||
5f5a1318 | 851 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 852 | return (uint32_t)-1; |
5f5a1318 JW |
853 | } |
854 | ||
855 | k->get_config(vdev, vdev->config); | |
967f97fa | 856 | |
06dbfc6f | 857 | val = ldl_p(vdev->config + addr); |
967f97fa AL |
858 | return val; |
859 | } | |
860 | ||
53c25cea | 861 | void virtio_config_writeb(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 862 | { |
181103cd | 863 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
864 | uint8_t val = data; |
865 | ||
5f5a1318 | 866 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 867 | return; |
5f5a1318 | 868 | } |
967f97fa | 869 | |
06dbfc6f | 870 | stb_p(vdev->config + addr, val); |
967f97fa | 871 | |
181103cd FK |
872 | if (k->set_config) { |
873 | k->set_config(vdev, vdev->config); | |
874 | } | |
967f97fa AL |
875 | } |
876 | ||
53c25cea | 877 | void virtio_config_writew(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 878 | { |
181103cd | 879 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
880 | uint16_t val = data; |
881 | ||
5f5a1318 | 882 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 883 | return; |
5f5a1318 | 884 | } |
967f97fa | 885 | |
06dbfc6f | 886 | stw_p(vdev->config + addr, val); |
967f97fa | 887 | |
181103cd FK |
888 | if (k->set_config) { |
889 | k->set_config(vdev, vdev->config); | |
890 | } | |
967f97fa AL |
891 | } |
892 | ||
53c25cea | 893 | void virtio_config_writel(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 894 | { |
181103cd | 895 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
896 | uint32_t val = data; |
897 | ||
5f5a1318 | 898 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 899 | return; |
5f5a1318 | 900 | } |
967f97fa | 901 | |
06dbfc6f | 902 | stl_p(vdev->config + addr, val); |
967f97fa | 903 | |
181103cd FK |
904 | if (k->set_config) { |
905 | k->set_config(vdev, vdev->config); | |
906 | } | |
967f97fa AL |
907 | } |
908 | ||
adfb743c MT |
909 | uint32_t virtio_config_modern_readb(VirtIODevice *vdev, uint32_t addr) |
910 | { | |
911 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
912 | uint8_t val; | |
913 | ||
914 | if (addr + sizeof(val) > vdev->config_len) { | |
915 | return (uint32_t)-1; | |
916 | } | |
917 | ||
918 | k->get_config(vdev, vdev->config); | |
919 | ||
920 | val = ldub_p(vdev->config + addr); | |
921 | return val; | |
922 | } | |
923 | ||
924 | uint32_t virtio_config_modern_readw(VirtIODevice *vdev, uint32_t addr) | |
925 | { | |
926 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
927 | uint16_t val; | |
928 | ||
929 | if (addr + sizeof(val) > vdev->config_len) { | |
930 | return (uint32_t)-1; | |
931 | } | |
932 | ||
933 | k->get_config(vdev, vdev->config); | |
934 | ||
935 | val = lduw_le_p(vdev->config + addr); | |
936 | return val; | |
937 | } | |
938 | ||
939 | uint32_t virtio_config_modern_readl(VirtIODevice *vdev, uint32_t addr) | |
940 | { | |
941 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
942 | uint32_t val; | |
943 | ||
944 | if (addr + sizeof(val) > vdev->config_len) { | |
945 | return (uint32_t)-1; | |
946 | } | |
947 | ||
948 | k->get_config(vdev, vdev->config); | |
949 | ||
950 | val = ldl_le_p(vdev->config + addr); | |
951 | return val; | |
952 | } | |
953 | ||
954 | void virtio_config_modern_writeb(VirtIODevice *vdev, | |
955 | uint32_t addr, uint32_t data) | |
956 | { | |
957 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
958 | uint8_t val = data; | |
959 | ||
960 | if (addr + sizeof(val) > vdev->config_len) { | |
961 | return; | |
962 | } | |
963 | ||
964 | stb_p(vdev->config + addr, val); | |
965 | ||
966 | if (k->set_config) { | |
967 | k->set_config(vdev, vdev->config); | |
968 | } | |
969 | } | |
970 | ||
971 | void virtio_config_modern_writew(VirtIODevice *vdev, | |
972 | uint32_t addr, uint32_t data) | |
973 | { | |
974 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
975 | uint16_t val = data; | |
976 | ||
977 | if (addr + sizeof(val) > vdev->config_len) { | |
978 | return; | |
979 | } | |
980 | ||
981 | stw_le_p(vdev->config + addr, val); | |
982 | ||
983 | if (k->set_config) { | |
984 | k->set_config(vdev, vdev->config); | |
985 | } | |
986 | } | |
987 | ||
988 | void virtio_config_modern_writel(VirtIODevice *vdev, | |
989 | uint32_t addr, uint32_t data) | |
990 | { | |
991 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); | |
992 | uint32_t val = data; | |
993 | ||
994 | if (addr + sizeof(val) > vdev->config_len) { | |
995 | return; | |
996 | } | |
997 | ||
998 | stl_le_p(vdev->config + addr, val); | |
999 | ||
1000 | if (k->set_config) { | |
1001 | k->set_config(vdev, vdev->config); | |
1002 | } | |
1003 | } | |
1004 | ||
a8170e5e | 1005 | void virtio_queue_set_addr(VirtIODevice *vdev, int n, hwaddr addr) |
967f97fa | 1006 | { |
ab223c95 CH |
1007 | vdev->vq[n].vring.desc = addr; |
1008 | virtio_queue_update_rings(vdev, n); | |
53c25cea PB |
1009 | } |
1010 | ||
a8170e5e | 1011 | hwaddr virtio_queue_get_addr(VirtIODevice *vdev, int n) |
53c25cea | 1012 | { |
ab223c95 CH |
1013 | return vdev->vq[n].vring.desc; |
1014 | } | |
1015 | ||
1016 | void virtio_queue_set_rings(VirtIODevice *vdev, int n, hwaddr desc, | |
1017 | hwaddr avail, hwaddr used) | |
1018 | { | |
1019 | vdev->vq[n].vring.desc = desc; | |
1020 | vdev->vq[n].vring.avail = avail; | |
1021 | vdev->vq[n].vring.used = used; | |
53c25cea PB |
1022 | } |
1023 | ||
e63c0ba1 PM |
1024 | void virtio_queue_set_num(VirtIODevice *vdev, int n, int num) |
1025 | { | |
f6049f44 PM |
1026 | /* Don't allow guest to flip queue between existent and |
1027 | * nonexistent states, or to set it to an invalid size. | |
1028 | */ | |
1029 | if (!!num != !!vdev->vq[n].vring.num || | |
1030 | num > VIRTQUEUE_MAX_SIZE || | |
1031 | num < 0) { | |
1032 | return; | |
e63c0ba1 | 1033 | } |
f6049f44 | 1034 | vdev->vq[n].vring.num = num; |
e63c0ba1 PM |
1035 | } |
1036 | ||
e0d686bf JW |
1037 | VirtQueue *virtio_vector_first_queue(VirtIODevice *vdev, uint16_t vector) |
1038 | { | |
1039 | return QLIST_FIRST(&vdev->vector_queues[vector]); | |
1040 | } | |
1041 | ||
1042 | VirtQueue *virtio_vector_next_queue(VirtQueue *vq) | |
1043 | { | |
1044 | return QLIST_NEXT(vq, node); | |
1045 | } | |
1046 | ||
53c25cea PB |
1047 | int virtio_queue_get_num(VirtIODevice *vdev, int n) |
1048 | { | |
1049 | return vdev->vq[n].vring.num; | |
1050 | } | |
967f97fa | 1051 | |
8ad176aa JW |
1052 | int virtio_get_num_queues(VirtIODevice *vdev) |
1053 | { | |
1054 | int i; | |
1055 | ||
87b3bd1c | 1056 | for (i = 0; i < VIRTIO_QUEUE_MAX; i++) { |
8ad176aa JW |
1057 | if (!virtio_queue_get_num(vdev, i)) { |
1058 | break; | |
1059 | } | |
1060 | } | |
1061 | ||
1062 | return i; | |
1063 | } | |
1064 | ||
c80decdb PB |
1065 | int virtio_queue_get_id(VirtQueue *vq) |
1066 | { | |
1067 | VirtIODevice *vdev = vq->vdev; | |
87b3bd1c | 1068 | assert(vq >= &vdev->vq[0] && vq < &vdev->vq[VIRTIO_QUEUE_MAX]); |
c80decdb PB |
1069 | return vq - &vdev->vq[0]; |
1070 | } | |
1071 | ||
6ce69d1c PM |
1072 | void virtio_queue_set_align(VirtIODevice *vdev, int n, int align) |
1073 | { | |
1074 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); | |
1075 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1076 | ||
ab223c95 | 1077 | /* virtio-1 compliant devices cannot change the alignment */ |
95129d6f | 1078 | if (virtio_vdev_has_feature(vdev, VIRTIO_F_VERSION_1)) { |
ab223c95 CH |
1079 | error_report("tried to modify queue alignment for virtio-1 device"); |
1080 | return; | |
1081 | } | |
6ce69d1c PM |
1082 | /* Check that the transport told us it was going to do this |
1083 | * (so a buggy transport will immediately assert rather than | |
1084 | * silently failing to migrate this state) | |
1085 | */ | |
1086 | assert(k->has_variable_vring_alignment); | |
1087 | ||
1088 | vdev->vq[n].vring.align = align; | |
ab223c95 | 1089 | virtio_queue_update_rings(vdev, n); |
6ce69d1c PM |
1090 | } |
1091 | ||
344dc16f MT |
1092 | static void virtio_queue_notify_aio_vq(VirtQueue *vq) |
1093 | { | |
1094 | if (vq->vring.desc && vq->handle_aio_output) { | |
1095 | VirtIODevice *vdev = vq->vdev; | |
1096 | ||
1097 | trace_virtio_queue_notify(vdev, vq - vdev->vq, vq); | |
1098 | vq->handle_aio_output(vdev, vq); | |
1099 | } | |
1100 | } | |
1101 | ||
2b2cbcad | 1102 | static void virtio_queue_notify_vq(VirtQueue *vq) |
25db9ebe | 1103 | { |
9e0f5b81 | 1104 | if (vq->vring.desc && vq->handle_output) { |
25db9ebe | 1105 | VirtIODevice *vdev = vq->vdev; |
9e0f5b81 | 1106 | |
25db9ebe SH |
1107 | trace_virtio_queue_notify(vdev, vq - vdev->vq, vq); |
1108 | vq->handle_output(vdev, vq); | |
1109 | } | |
1110 | } | |
1111 | ||
53c25cea PB |
1112 | void virtio_queue_notify(VirtIODevice *vdev, int n) |
1113 | { | |
7157e2e2 | 1114 | virtio_queue_notify_vq(&vdev->vq[n]); |
967f97fa AL |
1115 | } |
1116 | ||
7055e687 MT |
1117 | uint16_t virtio_queue_vector(VirtIODevice *vdev, int n) |
1118 | { | |
87b3bd1c | 1119 | return n < VIRTIO_QUEUE_MAX ? vdev->vq[n].vector : |
7055e687 MT |
1120 | VIRTIO_NO_VECTOR; |
1121 | } | |
1122 | ||
1123 | void virtio_queue_set_vector(VirtIODevice *vdev, int n, uint16_t vector) | |
1124 | { | |
e0d686bf JW |
1125 | VirtQueue *vq = &vdev->vq[n]; |
1126 | ||
87b3bd1c | 1127 | if (n < VIRTIO_QUEUE_MAX) { |
e0d686bf JW |
1128 | if (vdev->vector_queues && |
1129 | vdev->vq[n].vector != VIRTIO_NO_VECTOR) { | |
1130 | QLIST_REMOVE(vq, node); | |
1131 | } | |
7055e687 | 1132 | vdev->vq[n].vector = vector; |
e0d686bf JW |
1133 | if (vdev->vector_queues && |
1134 | vector != VIRTIO_NO_VECTOR) { | |
1135 | QLIST_INSERT_HEAD(&vdev->vector_queues[vector], vq, node); | |
1136 | } | |
1137 | } | |
7055e687 MT |
1138 | } |
1139 | ||
967f97fa AL |
1140 | VirtQueue *virtio_add_queue(VirtIODevice *vdev, int queue_size, |
1141 | void (*handle_output)(VirtIODevice *, VirtQueue *)) | |
1142 | { | |
1143 | int i; | |
1144 | ||
87b3bd1c | 1145 | for (i = 0; i < VIRTIO_QUEUE_MAX; i++) { |
967f97fa AL |
1146 | if (vdev->vq[i].vring.num == 0) |
1147 | break; | |
1148 | } | |
1149 | ||
87b3bd1c | 1150 | if (i == VIRTIO_QUEUE_MAX || queue_size > VIRTQUEUE_MAX_SIZE) |
967f97fa AL |
1151 | abort(); |
1152 | ||
1153 | vdev->vq[i].vring.num = queue_size; | |
46c5d082 | 1154 | vdev->vq[i].vring.num_default = queue_size; |
6ce69d1c | 1155 | vdev->vq[i].vring.align = VIRTIO_PCI_VRING_ALIGN; |
967f97fa | 1156 | vdev->vq[i].handle_output = handle_output; |
344dc16f | 1157 | vdev->vq[i].handle_aio_output = NULL; |
967f97fa AL |
1158 | |
1159 | return &vdev->vq[i]; | |
1160 | } | |
1161 | ||
f23fd811 JW |
1162 | void virtio_del_queue(VirtIODevice *vdev, int n) |
1163 | { | |
87b3bd1c | 1164 | if (n < 0 || n >= VIRTIO_QUEUE_MAX) { |
f23fd811 JW |
1165 | abort(); |
1166 | } | |
1167 | ||
1168 | vdev->vq[n].vring.num = 0; | |
46c5d082 | 1169 | vdev->vq[n].vring.num_default = 0; |
f23fd811 JW |
1170 | } |
1171 | ||
1cbdabe2 MT |
1172 | void virtio_irq(VirtQueue *vq) |
1173 | { | |
64979a4d | 1174 | trace_virtio_irq(vq); |
1cbdabe2 MT |
1175 | vq->vdev->isr |= 0x01; |
1176 | virtio_notify_vector(vq->vdev, vq->vector); | |
1177 | } | |
1178 | ||
adb3feda | 1179 | bool virtio_should_notify(VirtIODevice *vdev, VirtQueue *vq) |
bcbabae8 MT |
1180 | { |
1181 | uint16_t old, new; | |
1182 | bool v; | |
a281ebc1 MT |
1183 | /* We need to expose used array entries before checking used event. */ |
1184 | smp_mb(); | |
97b83deb | 1185 | /* Always notify when queue is empty (when feature acknowledge) */ |
95129d6f | 1186 | if (virtio_vdev_has_feature(vdev, VIRTIO_F_NOTIFY_ON_EMPTY) && |
be1fea9b | 1187 | !vq->inuse && virtio_queue_empty(vq)) { |
bcbabae8 MT |
1188 | return true; |
1189 | } | |
1190 | ||
95129d6f | 1191 | if (!virtio_vdev_has_feature(vdev, VIRTIO_RING_F_EVENT_IDX)) { |
bcbabae8 MT |
1192 | return !(vring_avail_flags(vq) & VRING_AVAIL_F_NO_INTERRUPT); |
1193 | } | |
1194 | ||
1195 | v = vq->signalled_used_valid; | |
1196 | vq->signalled_used_valid = true; | |
1197 | old = vq->signalled_used; | |
b796fcd1 | 1198 | new = vq->signalled_used = vq->used_idx; |
e9600c6c | 1199 | return !v || vring_need_event(vring_get_used_event(vq), new, old); |
bcbabae8 MT |
1200 | } |
1201 | ||
1202 | void virtio_notify(VirtIODevice *vdev, VirtQueue *vq) | |
1203 | { | |
adb3feda | 1204 | if (!virtio_should_notify(vdev, vq)) { |
967f97fa | 1205 | return; |
bcbabae8 | 1206 | } |
967f97fa | 1207 | |
64979a4d | 1208 | trace_virtio_notify(vdev, vq); |
967f97fa | 1209 | vdev->isr |= 0x01; |
7055e687 | 1210 | virtio_notify_vector(vdev, vq->vector); |
967f97fa AL |
1211 | } |
1212 | ||
1213 | void virtio_notify_config(VirtIODevice *vdev) | |
1214 | { | |
7625162c AL |
1215 | if (!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) |
1216 | return; | |
1217 | ||
967f97fa | 1218 | vdev->isr |= 0x03; |
b8f05908 | 1219 | vdev->generation++; |
7055e687 | 1220 | virtio_notify_vector(vdev, vdev->config_vector); |
967f97fa AL |
1221 | } |
1222 | ||
616a6552 GK |
1223 | static bool virtio_device_endian_needed(void *opaque) |
1224 | { | |
1225 | VirtIODevice *vdev = opaque; | |
1226 | ||
1227 | assert(vdev->device_endian != VIRTIO_DEVICE_ENDIAN_UNKNOWN); | |
95129d6f | 1228 | if (!virtio_vdev_has_feature(vdev, VIRTIO_F_VERSION_1)) { |
3c185597 CH |
1229 | return vdev->device_endian != virtio_default_endian(); |
1230 | } | |
1231 | /* Devices conforming to VIRTIO 1.0 or later are always LE. */ | |
1232 | return vdev->device_endian != VIRTIO_DEVICE_ENDIAN_LITTLE; | |
616a6552 GK |
1233 | } |
1234 | ||
019a3edb GH |
1235 | static bool virtio_64bit_features_needed(void *opaque) |
1236 | { | |
1237 | VirtIODevice *vdev = opaque; | |
1238 | ||
1239 | return (vdev->host_features >> 32) != 0; | |
1240 | } | |
1241 | ||
74aae7b2 JW |
1242 | static bool virtio_virtqueue_needed(void *opaque) |
1243 | { | |
1244 | VirtIODevice *vdev = opaque; | |
1245 | ||
1246 | return virtio_host_has_feature(vdev, VIRTIO_F_VERSION_1); | |
1247 | } | |
1248 | ||
46c5d082 CH |
1249 | static bool virtio_ringsize_needed(void *opaque) |
1250 | { | |
1251 | VirtIODevice *vdev = opaque; | |
1252 | int i; | |
1253 | ||
1254 | for (i = 0; i < VIRTIO_QUEUE_MAX; i++) { | |
1255 | if (vdev->vq[i].vring.num != vdev->vq[i].vring.num_default) { | |
1256 | return true; | |
1257 | } | |
1258 | } | |
1259 | return false; | |
1260 | } | |
1261 | ||
a6df8adf JW |
1262 | static bool virtio_extra_state_needed(void *opaque) |
1263 | { | |
1264 | VirtIODevice *vdev = opaque; | |
1265 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); | |
1266 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1267 | ||
1268 | return k->has_extra_state && | |
1269 | k->has_extra_state(qbus->parent); | |
1270 | } | |
1271 | ||
50e5ae4d | 1272 | static const VMStateDescription vmstate_virtqueue = { |
74aae7b2 | 1273 | .name = "virtqueue_state", |
50e5ae4d DDAG |
1274 | .version_id = 1, |
1275 | .minimum_version_id = 1, | |
1276 | .fields = (VMStateField[]) { | |
1277 | VMSTATE_UINT64(vring.avail, struct VirtQueue), | |
1278 | VMSTATE_UINT64(vring.used, struct VirtQueue), | |
1279 | VMSTATE_END_OF_LIST() | |
1280 | } | |
74aae7b2 JW |
1281 | }; |
1282 | ||
1283 | static const VMStateDescription vmstate_virtio_virtqueues = { | |
1284 | .name = "virtio/virtqueues", | |
1285 | .version_id = 1, | |
1286 | .minimum_version_id = 1, | |
1287 | .needed = &virtio_virtqueue_needed, | |
1288 | .fields = (VMStateField[]) { | |
3e996cc5 DDAG |
1289 | VMSTATE_STRUCT_VARRAY_POINTER_KNOWN(vq, struct VirtIODevice, |
1290 | VIRTIO_QUEUE_MAX, 0, vmstate_virtqueue, VirtQueue), | |
74aae7b2 JW |
1291 | VMSTATE_END_OF_LIST() |
1292 | } | |
1293 | }; | |
1294 | ||
50e5ae4d | 1295 | static const VMStateDescription vmstate_ringsize = { |
46c5d082 | 1296 | .name = "ringsize_state", |
50e5ae4d DDAG |
1297 | .version_id = 1, |
1298 | .minimum_version_id = 1, | |
1299 | .fields = (VMStateField[]) { | |
1300 | VMSTATE_UINT32(vring.num_default, struct VirtQueue), | |
1301 | VMSTATE_END_OF_LIST() | |
1302 | } | |
46c5d082 CH |
1303 | }; |
1304 | ||
1305 | static const VMStateDescription vmstate_virtio_ringsize = { | |
1306 | .name = "virtio/ringsize", | |
1307 | .version_id = 1, | |
1308 | .minimum_version_id = 1, | |
1309 | .needed = &virtio_ringsize_needed, | |
1310 | .fields = (VMStateField[]) { | |
3e996cc5 DDAG |
1311 | VMSTATE_STRUCT_VARRAY_POINTER_KNOWN(vq, struct VirtIODevice, |
1312 | VIRTIO_QUEUE_MAX, 0, vmstate_ringsize, VirtQueue), | |
46c5d082 CH |
1313 | VMSTATE_END_OF_LIST() |
1314 | } | |
1315 | }; | |
1316 | ||
a6df8adf JW |
1317 | static int get_extra_state(QEMUFile *f, void *pv, size_t size) |
1318 | { | |
1319 | VirtIODevice *vdev = pv; | |
1320 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); | |
1321 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1322 | ||
1323 | if (!k->load_extra_state) { | |
1324 | return -1; | |
1325 | } else { | |
1326 | return k->load_extra_state(qbus->parent, f); | |
1327 | } | |
1328 | } | |
1329 | ||
1330 | static void put_extra_state(QEMUFile *f, void *pv, size_t size) | |
1331 | { | |
1332 | VirtIODevice *vdev = pv; | |
1333 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); | |
1334 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1335 | ||
1336 | k->save_extra_state(qbus->parent, f); | |
1337 | } | |
1338 | ||
1339 | static const VMStateInfo vmstate_info_extra_state = { | |
1340 | .name = "virtqueue_extra_state", | |
1341 | .get = get_extra_state, | |
1342 | .put = put_extra_state, | |
1343 | }; | |
1344 | ||
1345 | static const VMStateDescription vmstate_virtio_extra_state = { | |
1346 | .name = "virtio/extra_state", | |
1347 | .version_id = 1, | |
1348 | .minimum_version_id = 1, | |
1349 | .needed = &virtio_extra_state_needed, | |
1350 | .fields = (VMStateField[]) { | |
1351 | { | |
1352 | .name = "extra_state", | |
1353 | .version_id = 0, | |
1354 | .field_exists = NULL, | |
1355 | .size = 0, | |
1356 | .info = &vmstate_info_extra_state, | |
1357 | .flags = VMS_SINGLE, | |
1358 | .offset = 0, | |
1359 | }, | |
1360 | VMSTATE_END_OF_LIST() | |
1361 | } | |
1362 | }; | |
1363 | ||
616a6552 GK |
1364 | static const VMStateDescription vmstate_virtio_device_endian = { |
1365 | .name = "virtio/device_endian", | |
1366 | .version_id = 1, | |
1367 | .minimum_version_id = 1, | |
5cd8cada | 1368 | .needed = &virtio_device_endian_needed, |
616a6552 GK |
1369 | .fields = (VMStateField[]) { |
1370 | VMSTATE_UINT8(device_endian, VirtIODevice), | |
1371 | VMSTATE_END_OF_LIST() | |
1372 | } | |
1373 | }; | |
1374 | ||
019a3edb GH |
1375 | static const VMStateDescription vmstate_virtio_64bit_features = { |
1376 | .name = "virtio/64bit_features", | |
1377 | .version_id = 1, | |
1378 | .minimum_version_id = 1, | |
5cd8cada | 1379 | .needed = &virtio_64bit_features_needed, |
019a3edb GH |
1380 | .fields = (VMStateField[]) { |
1381 | VMSTATE_UINT64(guest_features, VirtIODevice), | |
1382 | VMSTATE_END_OF_LIST() | |
1383 | } | |
1384 | }; | |
1385 | ||
6b321a3d GK |
1386 | static const VMStateDescription vmstate_virtio = { |
1387 | .name = "virtio", | |
1388 | .version_id = 1, | |
1389 | .minimum_version_id = 1, | |
1390 | .minimum_version_id_old = 1, | |
1391 | .fields = (VMStateField[]) { | |
1392 | VMSTATE_END_OF_LIST() | |
616a6552 | 1393 | }, |
5cd8cada JQ |
1394 | .subsections = (const VMStateDescription*[]) { |
1395 | &vmstate_virtio_device_endian, | |
1396 | &vmstate_virtio_64bit_features, | |
74aae7b2 | 1397 | &vmstate_virtio_virtqueues, |
46c5d082 | 1398 | &vmstate_virtio_ringsize, |
a6df8adf | 1399 | &vmstate_virtio_extra_state, |
5cd8cada | 1400 | NULL |
6b321a3d GK |
1401 | } |
1402 | }; | |
1403 | ||
967f97fa AL |
1404 | void virtio_save(VirtIODevice *vdev, QEMUFile *f) |
1405 | { | |
1c819449 FK |
1406 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
1407 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1b5fc0de | 1408 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(vdev); |
019a3edb | 1409 | uint32_t guest_features_lo = (vdev->guest_features & 0xffffffff); |
967f97fa AL |
1410 | int i; |
1411 | ||
1c819449 FK |
1412 | if (k->save_config) { |
1413 | k->save_config(qbus->parent, f); | |
1414 | } | |
967f97fa | 1415 | |
967f97fa AL |
1416 | qemu_put_8s(f, &vdev->status); |
1417 | qemu_put_8s(f, &vdev->isr); | |
1418 | qemu_put_be16s(f, &vdev->queue_sel); | |
019a3edb | 1419 | qemu_put_be32s(f, &guest_features_lo); |
967f97fa AL |
1420 | qemu_put_be32(f, vdev->config_len); |
1421 | qemu_put_buffer(f, vdev->config, vdev->config_len); | |
1422 | ||
87b3bd1c | 1423 | for (i = 0; i < VIRTIO_QUEUE_MAX; i++) { |
967f97fa AL |
1424 | if (vdev->vq[i].vring.num == 0) |
1425 | break; | |
1426 | } | |
1427 | ||
1428 | qemu_put_be32(f, i); | |
1429 | ||
87b3bd1c | 1430 | for (i = 0; i < VIRTIO_QUEUE_MAX; i++) { |
967f97fa AL |
1431 | if (vdev->vq[i].vring.num == 0) |
1432 | break; | |
1433 | ||
1434 | qemu_put_be32(f, vdev->vq[i].vring.num); | |
6ce69d1c PM |
1435 | if (k->has_variable_vring_alignment) { |
1436 | qemu_put_be32(f, vdev->vq[i].vring.align); | |
1437 | } | |
ab223c95 CH |
1438 | /* XXX virtio-1 devices */ |
1439 | qemu_put_be64(f, vdev->vq[i].vring.desc); | |
967f97fa | 1440 | qemu_put_be16s(f, &vdev->vq[i].last_avail_idx); |
1c819449 FK |
1441 | if (k->save_queue) { |
1442 | k->save_queue(qbus->parent, i, f); | |
1443 | } | |
967f97fa | 1444 | } |
1b5fc0de GK |
1445 | |
1446 | if (vdc->save != NULL) { | |
1447 | vdc->save(vdev, f); | |
1448 | } | |
6b321a3d GK |
1449 | |
1450 | /* Subsections */ | |
8118f095 | 1451 | vmstate_save_state(f, &vmstate_virtio, vdev, NULL); |
967f97fa AL |
1452 | } |
1453 | ||
6c0196d7 | 1454 | static int virtio_set_features_nocheck(VirtIODevice *vdev, uint64_t val) |
ad0c9332 | 1455 | { |
181103cd | 1456 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
6b8f1020 | 1457 | bool bad = (val & ~(vdev->host_features)) != 0; |
ad0c9332 | 1458 | |
6b8f1020 | 1459 | val &= vdev->host_features; |
181103cd FK |
1460 | if (k->set_features) { |
1461 | k->set_features(vdev, val); | |
ad0c9332 PB |
1462 | } |
1463 | vdev->guest_features = val; | |
1464 | return bad ? -1 : 0; | |
1465 | } | |
1466 | ||
6c0196d7 CH |
1467 | int virtio_set_features(VirtIODevice *vdev, uint64_t val) |
1468 | { | |
1469 | /* | |
1470 | * The driver must not attempt to set features after feature negotiation | |
1471 | * has finished. | |
1472 | */ | |
1473 | if (vdev->status & VIRTIO_CONFIG_S_FEATURES_OK) { | |
1474 | return -EINVAL; | |
1475 | } | |
1476 | return virtio_set_features_nocheck(vdev, val); | |
1477 | } | |
1478 | ||
1b5fc0de | 1479 | int virtio_load(VirtIODevice *vdev, QEMUFile *f, int version_id) |
967f97fa | 1480 | { |
cc459952 | 1481 | int i, ret; |
a890a2f9 | 1482 | int32_t config_len; |
cc459952 | 1483 | uint32_t num; |
6d74ca5a | 1484 | uint32_t features; |
1c819449 FK |
1485 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
1486 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1b5fc0de | 1487 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa | 1488 | |
616a6552 GK |
1489 | /* |
1490 | * We poison the endianness to ensure it does not get used before | |
1491 | * subsections have been loaded. | |
1492 | */ | |
1493 | vdev->device_endian = VIRTIO_DEVICE_ENDIAN_UNKNOWN; | |
1494 | ||
1c819449 FK |
1495 | if (k->load_config) { |
1496 | ret = k->load_config(qbus->parent, f); | |
ff24bd58 MT |
1497 | if (ret) |
1498 | return ret; | |
1499 | } | |
967f97fa | 1500 | |
967f97fa AL |
1501 | qemu_get_8s(f, &vdev->status); |
1502 | qemu_get_8s(f, &vdev->isr); | |
1503 | qemu_get_be16s(f, &vdev->queue_sel); | |
87b3bd1c | 1504 | if (vdev->queue_sel >= VIRTIO_QUEUE_MAX) { |
4b53c2c7 MR |
1505 | return -1; |
1506 | } | |
6d74ca5a | 1507 | qemu_get_be32s(f, &features); |
ad0c9332 | 1508 | |
a890a2f9 | 1509 | config_len = qemu_get_be32(f); |
2f5732e9 DDAG |
1510 | |
1511 | /* | |
1512 | * There are cases where the incoming config can be bigger or smaller | |
1513 | * than what we have; so load what we have space for, and skip | |
1514 | * any excess that's in the stream. | |
1515 | */ | |
1516 | qemu_get_buffer(f, vdev->config, MIN(config_len, vdev->config_len)); | |
1517 | ||
1518 | while (config_len > vdev->config_len) { | |
1519 | qemu_get_byte(f); | |
1520 | config_len--; | |
a890a2f9 | 1521 | } |
967f97fa AL |
1522 | |
1523 | num = qemu_get_be32(f); | |
1524 | ||
87b3bd1c | 1525 | if (num > VIRTIO_QUEUE_MAX) { |
8a1be662 | 1526 | error_report("Invalid number of virtqueues: 0x%x", num); |
cc459952 MT |
1527 | return -1; |
1528 | } | |
1529 | ||
967f97fa AL |
1530 | for (i = 0; i < num; i++) { |
1531 | vdev->vq[i].vring.num = qemu_get_be32(f); | |
6ce69d1c PM |
1532 | if (k->has_variable_vring_alignment) { |
1533 | vdev->vq[i].vring.align = qemu_get_be32(f); | |
1534 | } | |
ab223c95 | 1535 | vdev->vq[i].vring.desc = qemu_get_be64(f); |
967f97fa | 1536 | qemu_get_be16s(f, &vdev->vq[i].last_avail_idx); |
bcbabae8 MT |
1537 | vdev->vq[i].signalled_used_valid = false; |
1538 | vdev->vq[i].notification = true; | |
967f97fa | 1539 | |
ab223c95 CH |
1540 | if (vdev->vq[i].vring.desc) { |
1541 | /* XXX virtio-1 devices */ | |
1542 | virtio_queue_update_rings(vdev, i); | |
1abeb5a6 MT |
1543 | } else if (vdev->vq[i].last_avail_idx) { |
1544 | error_report("VQ %d address 0x0 " | |
6daf194d | 1545 | "inconsistent with Host index 0x%x", |
1abeb5a6 MT |
1546 | i, vdev->vq[i].last_avail_idx); |
1547 | return -1; | |
258dc7c9 | 1548 | } |
1c819449 FK |
1549 | if (k->load_queue) { |
1550 | ret = k->load_queue(qbus->parent, i, f); | |
ff24bd58 MT |
1551 | if (ret) |
1552 | return ret; | |
7055e687 | 1553 | } |
967f97fa AL |
1554 | } |
1555 | ||
7055e687 | 1556 | virtio_notify_vector(vdev, VIRTIO_NO_VECTOR); |
1b5fc0de GK |
1557 | |
1558 | if (vdc->load != NULL) { | |
6b321a3d GK |
1559 | ret = vdc->load(vdev, f, version_id); |
1560 | if (ret) { | |
1561 | return ret; | |
1562 | } | |
1b5fc0de GK |
1563 | } |
1564 | ||
616a6552 GK |
1565 | /* Subsections */ |
1566 | ret = vmstate_load_state(f, &vmstate_virtio, vdev, 1); | |
1567 | if (ret) { | |
1568 | return ret; | |
1569 | } | |
1570 | ||
1571 | if (vdev->device_endian == VIRTIO_DEVICE_ENDIAN_UNKNOWN) { | |
1572 | vdev->device_endian = virtio_default_endian(); | |
1573 | } | |
1574 | ||
019a3edb GH |
1575 | if (virtio_64bit_features_needed(vdev)) { |
1576 | /* | |
1577 | * Subsection load filled vdev->guest_features. Run them | |
1578 | * through virtio_set_features to sanity-check them against | |
1579 | * host_features. | |
1580 | */ | |
1581 | uint64_t features64 = vdev->guest_features; | |
6c0196d7 | 1582 | if (virtio_set_features_nocheck(vdev, features64) < 0) { |
019a3edb GH |
1583 | error_report("Features 0x%" PRIx64 " unsupported. " |
1584 | "Allowed features: 0x%" PRIx64, | |
1585 | features64, vdev->host_features); | |
1586 | return -1; | |
1587 | } | |
1588 | } else { | |
6c0196d7 | 1589 | if (virtio_set_features_nocheck(vdev, features) < 0) { |
019a3edb GH |
1590 | error_report("Features 0x%x unsupported. " |
1591 | "Allowed features: 0x%" PRIx64, | |
1592 | features, vdev->host_features); | |
1593 | return -1; | |
1594 | } | |
1595 | } | |
1596 | ||
616a6552 | 1597 | for (i = 0; i < num; i++) { |
ab223c95 | 1598 | if (vdev->vq[i].vring.desc) { |
616a6552 GK |
1599 | uint16_t nheads; |
1600 | nheads = vring_avail_idx(&vdev->vq[i]) - vdev->vq[i].last_avail_idx; | |
1601 | /* Check it isn't doing strange things with descriptor numbers. */ | |
1602 | if (nheads > vdev->vq[i].vring.num) { | |
1603 | error_report("VQ %d size 0x%x Guest index 0x%x " | |
1604 | "inconsistent with Host index 0x%x: delta 0x%x", | |
1605 | i, vdev->vq[i].vring.num, | |
1606 | vring_avail_idx(&vdev->vq[i]), | |
1607 | vdev->vq[i].last_avail_idx, nheads); | |
1608 | return -1; | |
1609 | } | |
b796fcd1 | 1610 | vdev->vq[i].used_idx = vring_used_idx(&vdev->vq[i]); |
be1fea9b | 1611 | vdev->vq[i].shadow_avail_idx = vring_avail_idx(&vdev->vq[i]); |
616a6552 GK |
1612 | } |
1613 | } | |
1614 | ||
1615 | return 0; | |
967f97fa AL |
1616 | } |
1617 | ||
6a1a8cc7 | 1618 | void virtio_cleanup(VirtIODevice *vdev) |
b946a153 | 1619 | { |
85cf2a8d | 1620 | qemu_del_vm_change_state_handler(vdev->vmstate); |
6f79e06b | 1621 | g_free(vdev->config); |
7267c094 | 1622 | g_free(vdev->vq); |
e0d686bf | 1623 | g_free(vdev->vector_queues); |
8e05db92 FK |
1624 | } |
1625 | ||
1dfb4dd9 | 1626 | static void virtio_vmstate_change(void *opaque, int running, RunState state) |
85cf2a8d MT |
1627 | { |
1628 | VirtIODevice *vdev = opaque; | |
1c819449 FK |
1629 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
1630 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
85cf2a8d | 1631 | bool backend_run = running && (vdev->status & VIRTIO_CONFIG_S_DRIVER_OK); |
9e8e8c48 | 1632 | vdev->vm_running = running; |
85cf2a8d MT |
1633 | |
1634 | if (backend_run) { | |
1635 | virtio_set_status(vdev, vdev->status); | |
1636 | } | |
1637 | ||
1c819449 FK |
1638 | if (k->vmstate_change) { |
1639 | k->vmstate_change(qbus->parent, backend_run); | |
85cf2a8d MT |
1640 | } |
1641 | ||
1642 | if (!backend_run) { | |
1643 | virtio_set_status(vdev, vdev->status); | |
1644 | } | |
1645 | } | |
1646 | ||
c8075caf GA |
1647 | void virtio_instance_init_common(Object *proxy_obj, void *data, |
1648 | size_t vdev_size, const char *vdev_name) | |
1649 | { | |
1650 | DeviceState *vdev = data; | |
1651 | ||
1652 | object_initialize(vdev, vdev_size, vdev_name); | |
1653 | object_property_add_child(proxy_obj, "virtio-backend", OBJECT(vdev), NULL); | |
1654 | object_unref(OBJECT(vdev)); | |
1655 | qdev_alias_all_properties(vdev, proxy_obj); | |
1656 | } | |
1657 | ||
8e05db92 FK |
1658 | void virtio_init(VirtIODevice *vdev, const char *name, |
1659 | uint16_t device_id, size_t config_size) | |
967f97fa | 1660 | { |
e0d686bf JW |
1661 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
1662 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
b8193adb | 1663 | int i; |
e0d686bf JW |
1664 | int nvectors = k->query_nvectors ? k->query_nvectors(qbus->parent) : 0; |
1665 | ||
1666 | if (nvectors) { | |
1667 | vdev->vector_queues = | |
1668 | g_malloc0(sizeof(*vdev->vector_queues) * nvectors); | |
1669 | } | |
1670 | ||
53c25cea | 1671 | vdev->device_id = device_id; |
967f97fa AL |
1672 | vdev->status = 0; |
1673 | vdev->isr = 0; | |
1674 | vdev->queue_sel = 0; | |
7055e687 | 1675 | vdev->config_vector = VIRTIO_NO_VECTOR; |
87b3bd1c | 1676 | vdev->vq = g_malloc0(sizeof(VirtQueue) * VIRTIO_QUEUE_MAX); |
1354869c | 1677 | vdev->vm_running = runstate_is_running(); |
87b3bd1c | 1678 | for (i = 0; i < VIRTIO_QUEUE_MAX; i++) { |
b8193adb | 1679 | vdev->vq[i].vector = VIRTIO_NO_VECTOR; |
1cbdabe2 | 1680 | vdev->vq[i].vdev = vdev; |
e78a2b42 | 1681 | vdev->vq[i].queue_index = i; |
1cbdabe2 | 1682 | } |
967f97fa | 1683 | |
967f97fa AL |
1684 | vdev->name = name; |
1685 | vdev->config_len = config_size; | |
8e05db92 | 1686 | if (vdev->config_len) { |
7267c094 | 1687 | vdev->config = g_malloc0(config_size); |
8e05db92 | 1688 | } else { |
967f97fa | 1689 | vdev->config = NULL; |
8e05db92 FK |
1690 | } |
1691 | vdev->vmstate = qemu_add_vm_change_state_handler(virtio_vmstate_change, | |
1692 | vdev); | |
616a6552 | 1693 | vdev->device_endian = virtio_default_endian(); |
5669655a | 1694 | vdev->use_guest_notifier_mask = true; |
8e05db92 | 1695 | } |
967f97fa | 1696 | |
a8170e5e | 1697 | hwaddr virtio_queue_get_desc_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1698 | { |
1699 | return vdev->vq[n].vring.desc; | |
1700 | } | |
1701 | ||
a8170e5e | 1702 | hwaddr virtio_queue_get_avail_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1703 | { |
1704 | return vdev->vq[n].vring.avail; | |
1705 | } | |
1706 | ||
a8170e5e | 1707 | hwaddr virtio_queue_get_used_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1708 | { |
1709 | return vdev->vq[n].vring.used; | |
1710 | } | |
1711 | ||
a8170e5e | 1712 | hwaddr virtio_queue_get_ring_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1713 | { |
1714 | return vdev->vq[n].vring.desc; | |
1715 | } | |
1716 | ||
a8170e5e | 1717 | hwaddr virtio_queue_get_desc_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1718 | { |
1719 | return sizeof(VRingDesc) * vdev->vq[n].vring.num; | |
1720 | } | |
1721 | ||
a8170e5e | 1722 | hwaddr virtio_queue_get_avail_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1723 | { |
1724 | return offsetof(VRingAvail, ring) + | |
50764fc8 | 1725 | sizeof(uint16_t) * vdev->vq[n].vring.num; |
1cbdabe2 MT |
1726 | } |
1727 | ||
a8170e5e | 1728 | hwaddr virtio_queue_get_used_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1729 | { |
1730 | return offsetof(VRingUsed, ring) + | |
1731 | sizeof(VRingUsedElem) * vdev->vq[n].vring.num; | |
1732 | } | |
1733 | ||
a8170e5e | 1734 | hwaddr virtio_queue_get_ring_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1735 | { |
1736 | return vdev->vq[n].vring.used - vdev->vq[n].vring.desc + | |
1737 | virtio_queue_get_used_size(vdev, n); | |
1738 | } | |
1739 | ||
1740 | uint16_t virtio_queue_get_last_avail_idx(VirtIODevice *vdev, int n) | |
1741 | { | |
1742 | return vdev->vq[n].last_avail_idx; | |
1743 | } | |
1744 | ||
1745 | void virtio_queue_set_last_avail_idx(VirtIODevice *vdev, int n, uint16_t idx) | |
1746 | { | |
1747 | vdev->vq[n].last_avail_idx = idx; | |
be1fea9b | 1748 | vdev->vq[n].shadow_avail_idx = idx; |
1cbdabe2 MT |
1749 | } |
1750 | ||
6793dfd1 SH |
1751 | void virtio_queue_invalidate_signalled_used(VirtIODevice *vdev, int n) |
1752 | { | |
1753 | vdev->vq[n].signalled_used_valid = false; | |
1754 | } | |
1755 | ||
1cbdabe2 MT |
1756 | VirtQueue *virtio_get_queue(VirtIODevice *vdev, int n) |
1757 | { | |
1758 | return vdev->vq + n; | |
1759 | } | |
1760 | ||
e78a2b42 JW |
1761 | uint16_t virtio_get_queue_index(VirtQueue *vq) |
1762 | { | |
1763 | return vq->queue_index; | |
1764 | } | |
1765 | ||
15b2bd18 PB |
1766 | static void virtio_queue_guest_notifier_read(EventNotifier *n) |
1767 | { | |
1768 | VirtQueue *vq = container_of(n, VirtQueue, guest_notifier); | |
1769 | if (event_notifier_test_and_clear(n)) { | |
1770 | virtio_irq(vq); | |
1771 | } | |
1772 | } | |
1773 | ||
1774 | void virtio_queue_set_guest_notifier_fd_handler(VirtQueue *vq, bool assign, | |
1775 | bool with_irqfd) | |
1776 | { | |
1777 | if (assign && !with_irqfd) { | |
1778 | event_notifier_set_handler(&vq->guest_notifier, | |
1779 | virtio_queue_guest_notifier_read); | |
1780 | } else { | |
1781 | event_notifier_set_handler(&vq->guest_notifier, NULL); | |
1782 | } | |
1783 | if (!assign) { | |
1784 | /* Test and clear notifier before closing it, | |
1785 | * in case poll callback didn't have time to run. */ | |
1786 | virtio_queue_guest_notifier_read(&vq->guest_notifier); | |
1787 | } | |
1788 | } | |
1789 | ||
1cbdabe2 MT |
1790 | EventNotifier *virtio_queue_get_guest_notifier(VirtQueue *vq) |
1791 | { | |
1792 | return &vq->guest_notifier; | |
1793 | } | |
b1f416aa | 1794 | |
344dc16f | 1795 | static void virtio_queue_host_notifier_aio_read(EventNotifier *n) |
b1f416aa PB |
1796 | { |
1797 | VirtQueue *vq = container_of(n, VirtQueue, host_notifier); | |
1798 | if (event_notifier_test_and_clear(n)) { | |
344dc16f | 1799 | virtio_queue_notify_aio_vq(vq); |
b1f416aa PB |
1800 | } |
1801 | } | |
1802 | ||
a1afb606 | 1803 | void virtio_queue_aio_set_host_notifier_handler(VirtQueue *vq, AioContext *ctx, |
a378b49a PB |
1804 | void (*handle_output)(VirtIODevice *, |
1805 | VirtQueue *)) | |
a1afb606 | 1806 | { |
a378b49a PB |
1807 | if (handle_output) { |
1808 | vq->handle_aio_output = handle_output; | |
a1afb606 | 1809 | aio_set_event_notifier(ctx, &vq->host_notifier, true, |
344dc16f | 1810 | virtio_queue_host_notifier_aio_read); |
a1afb606 PB |
1811 | } else { |
1812 | aio_set_event_notifier(ctx, &vq->host_notifier, true, NULL); | |
a1afb606 PB |
1813 | /* Test and clear notifier before after disabling event, |
1814 | * in case poll callback didn't have time to run. */ | |
344dc16f | 1815 | virtio_queue_host_notifier_aio_read(&vq->host_notifier); |
a378b49a | 1816 | vq->handle_aio_output = NULL; |
344dc16f MT |
1817 | } |
1818 | } | |
1819 | ||
1820 | static void virtio_queue_host_notifier_read(EventNotifier *n) | |
1821 | { | |
1822 | VirtQueue *vq = container_of(n, VirtQueue, host_notifier); | |
1823 | if (event_notifier_test_and_clear(n)) { | |
1824 | virtio_queue_notify_vq(vq); | |
a1afb606 PB |
1825 | } |
1826 | } | |
1827 | ||
26b9b5fe PB |
1828 | void virtio_queue_set_host_notifier_fd_handler(VirtQueue *vq, bool assign, |
1829 | bool set_handler) | |
b1f416aa | 1830 | { |
26b9b5fe | 1831 | if (assign && set_handler) { |
b1f416aa PB |
1832 | event_notifier_set_handler(&vq->host_notifier, |
1833 | virtio_queue_host_notifier_read); | |
1834 | } else { | |
1835 | event_notifier_set_handler(&vq->host_notifier, NULL); | |
26b9b5fe PB |
1836 | } |
1837 | if (!assign) { | |
b1f416aa PB |
1838 | /* Test and clear notifier before after disabling event, |
1839 | * in case poll callback didn't have time to run. */ | |
1840 | virtio_queue_host_notifier_read(&vq->host_notifier); | |
1841 | } | |
1842 | } | |
1843 | ||
1cbdabe2 MT |
1844 | EventNotifier *virtio_queue_get_host_notifier(VirtQueue *vq) |
1845 | { | |
1846 | return &vq->host_notifier; | |
1847 | } | |
8e05db92 | 1848 | |
1034e9cf FK |
1849 | void virtio_device_set_child_bus_name(VirtIODevice *vdev, char *bus_name) |
1850 | { | |
9e288406 | 1851 | g_free(vdev->bus_name); |
80e0090a | 1852 | vdev->bus_name = g_strdup(bus_name); |
1034e9cf FK |
1853 | } |
1854 | ||
1d244b42 AF |
1855 | static void virtio_device_realize(DeviceState *dev, Error **errp) |
1856 | { | |
1857 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); | |
1858 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); | |
1859 | Error *err = NULL; | |
1860 | ||
1d244b42 AF |
1861 | if (vdc->realize != NULL) { |
1862 | vdc->realize(dev, &err); | |
1863 | if (err != NULL) { | |
1864 | error_propagate(errp, err); | |
1865 | return; | |
1866 | } | |
8e05db92 | 1867 | } |
e8398045 JW |
1868 | |
1869 | virtio_bus_device_plugged(vdev, &err); | |
1870 | if (err != NULL) { | |
1871 | error_propagate(errp, err); | |
1872 | return; | |
1873 | } | |
8e05db92 FK |
1874 | } |
1875 | ||
1d244b42 | 1876 | static void virtio_device_unrealize(DeviceState *dev, Error **errp) |
1034e9cf | 1877 | { |
1d244b42 | 1878 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); |
306ec6c3 AF |
1879 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); |
1880 | Error *err = NULL; | |
1d244b42 | 1881 | |
83d07047 PB |
1882 | virtio_bus_device_unplugged(vdev); |
1883 | ||
306ec6c3 AF |
1884 | if (vdc->unrealize != NULL) { |
1885 | vdc->unrealize(dev, &err); | |
1886 | if (err != NULL) { | |
1887 | error_propagate(errp, err); | |
1888 | return; | |
1889 | } | |
5e96f5d2 | 1890 | } |
1d244b42 | 1891 | |
9e288406 MA |
1892 | g_free(vdev->bus_name); |
1893 | vdev->bus_name = NULL; | |
1034e9cf FK |
1894 | } |
1895 | ||
6b8f1020 CH |
1896 | static Property virtio_properties[] = { |
1897 | DEFINE_VIRTIO_COMMON_FEATURES(VirtIODevice, host_features), | |
1898 | DEFINE_PROP_END_OF_LIST(), | |
1899 | }; | |
1900 | ||
8e05db92 FK |
1901 | static void virtio_device_class_init(ObjectClass *klass, void *data) |
1902 | { | |
1903 | /* Set the default value here. */ | |
1904 | DeviceClass *dc = DEVICE_CLASS(klass); | |
1d244b42 AF |
1905 | |
1906 | dc->realize = virtio_device_realize; | |
1907 | dc->unrealize = virtio_device_unrealize; | |
8e05db92 | 1908 | dc->bus_type = TYPE_VIRTIO_BUS; |
6b8f1020 | 1909 | dc->props = virtio_properties; |
8e05db92 FK |
1910 | } |
1911 | ||
1912 | static const TypeInfo virtio_device_info = { | |
1913 | .name = TYPE_VIRTIO_DEVICE, | |
1914 | .parent = TYPE_DEVICE, | |
1915 | .instance_size = sizeof(VirtIODevice), | |
1916 | .class_init = virtio_device_class_init, | |
1917 | .abstract = true, | |
1918 | .class_size = sizeof(VirtioDeviceClass), | |
1919 | }; | |
1920 | ||
1921 | static void virtio_register_types(void) | |
1922 | { | |
1923 | type_register_static(&virtio_device_info); | |
1924 | } | |
1925 | ||
1926 | type_init(virtio_register_types) |