]>
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 | ||
14 | #include <inttypes.h> | |
967f97fa | 15 | |
64979a4d | 16 | #include "trace.h" |
fdfba1a2 | 17 | #include "exec/address-spaces.h" |
1de7afc9 | 18 | #include "qemu/error-report.h" |
0d09e41a | 19 | #include "hw/virtio/virtio.h" |
1de7afc9 | 20 | #include "qemu/atomic.h" |
0d09e41a | 21 | #include "hw/virtio/virtio-bus.h" |
967f97fa | 22 | |
6ce69d1c PM |
23 | /* |
24 | * The alignment to use between consumer and producer parts of vring. | |
25 | * x86 pagesize again. This is the default, used by transports like PCI | |
26 | * which don't provide a means for the guest to tell the host the alignment. | |
27 | */ | |
f46f15bc AL |
28 | #define VIRTIO_PCI_VRING_ALIGN 4096 |
29 | ||
967f97fa AL |
30 | typedef struct VRingDesc |
31 | { | |
32 | uint64_t addr; | |
33 | uint32_t len; | |
34 | uint16_t flags; | |
35 | uint16_t next; | |
36 | } VRingDesc; | |
37 | ||
38 | typedef struct VRingAvail | |
39 | { | |
40 | uint16_t flags; | |
41 | uint16_t idx; | |
42 | uint16_t ring[0]; | |
43 | } VRingAvail; | |
44 | ||
45 | typedef struct VRingUsedElem | |
46 | { | |
47 | uint32_t id; | |
48 | uint32_t len; | |
49 | } VRingUsedElem; | |
50 | ||
51 | typedef struct VRingUsed | |
52 | { | |
53 | uint16_t flags; | |
54 | uint16_t idx; | |
55 | VRingUsedElem ring[0]; | |
56 | } VRingUsed; | |
57 | ||
58 | typedef struct VRing | |
59 | { | |
60 | unsigned int num; | |
6ce69d1c | 61 | unsigned int align; |
a8170e5e AK |
62 | hwaddr desc; |
63 | hwaddr avail; | |
64 | hwaddr used; | |
967f97fa AL |
65 | } VRing; |
66 | ||
67 | struct VirtQueue | |
68 | { | |
69 | VRing vring; | |
a8170e5e | 70 | hwaddr pa; |
967f97fa | 71 | uint16_t last_avail_idx; |
bcbabae8 MT |
72 | /* Last used index value we have signalled on */ |
73 | uint16_t signalled_used; | |
74 | ||
75 | /* Last used index value we have signalled on */ | |
76 | bool signalled_used_valid; | |
77 | ||
78 | /* Notification enabled? */ | |
79 | bool notification; | |
80 | ||
e78a2b42 JW |
81 | uint16_t queue_index; |
82 | ||
967f97fa | 83 | int inuse; |
bcbabae8 | 84 | |
7055e687 | 85 | uint16_t vector; |
967f97fa | 86 | void (*handle_output)(VirtIODevice *vdev, VirtQueue *vq); |
1cbdabe2 MT |
87 | VirtIODevice *vdev; |
88 | EventNotifier guest_notifier; | |
89 | EventNotifier host_notifier; | |
967f97fa AL |
90 | }; |
91 | ||
967f97fa | 92 | /* virt queue functions */ |
53c25cea | 93 | static void virtqueue_init(VirtQueue *vq) |
967f97fa | 94 | { |
a8170e5e | 95 | hwaddr pa = vq->pa; |
53c25cea | 96 | |
967f97fa AL |
97 | vq->vring.desc = pa; |
98 | vq->vring.avail = pa + vq->vring.num * sizeof(VRingDesc); | |
f46f15bc AL |
99 | vq->vring.used = vring_align(vq->vring.avail + |
100 | offsetof(VRingAvail, ring[vq->vring.num]), | |
6ce69d1c | 101 | vq->vring.align); |
967f97fa AL |
102 | } |
103 | ||
a8170e5e | 104 | static inline uint64_t vring_desc_addr(hwaddr desc_pa, int i) |
967f97fa | 105 | { |
a8170e5e | 106 | hwaddr pa; |
5774cf98 | 107 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, addr); |
2c17449b | 108 | return ldq_phys(&address_space_memory, pa); |
967f97fa AL |
109 | } |
110 | ||
a8170e5e | 111 | static inline uint32_t vring_desc_len(hwaddr desc_pa, int i) |
967f97fa | 112 | { |
a8170e5e | 113 | hwaddr pa; |
5774cf98 | 114 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, len); |
fdfba1a2 | 115 | return ldl_phys(&address_space_memory, pa); |
967f97fa AL |
116 | } |
117 | ||
a8170e5e | 118 | static inline uint16_t vring_desc_flags(hwaddr desc_pa, int i) |
967f97fa | 119 | { |
a8170e5e | 120 | hwaddr pa; |
5774cf98 | 121 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, flags); |
41701aa4 | 122 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
123 | } |
124 | ||
a8170e5e | 125 | static inline uint16_t vring_desc_next(hwaddr desc_pa, int i) |
967f97fa | 126 | { |
a8170e5e | 127 | hwaddr pa; |
5774cf98 | 128 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, next); |
41701aa4 | 129 | return lduw_phys(&address_space_memory, pa); |
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); |
41701aa4 | 136 | return lduw_phys(&address_space_memory, 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); |
41701aa4 | 143 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
144 | } |
145 | ||
146 | static inline uint16_t vring_avail_ring(VirtQueue *vq, int i) | |
147 | { | |
a8170e5e | 148 | hwaddr pa; |
967f97fa | 149 | pa = vq->vring.avail + offsetof(VRingAvail, ring[i]); |
41701aa4 | 150 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
151 | } |
152 | ||
bcbabae8 MT |
153 | static inline uint16_t vring_used_event(VirtQueue *vq) |
154 | { | |
155 | return vring_avail_ring(vq, vq->vring.num); | |
156 | } | |
157 | ||
967f97fa AL |
158 | static inline void vring_used_ring_id(VirtQueue *vq, int i, uint32_t val) |
159 | { | |
a8170e5e | 160 | hwaddr pa; |
967f97fa | 161 | pa = vq->vring.used + offsetof(VRingUsed, ring[i].id); |
ab1da857 | 162 | stl_phys(&address_space_memory, pa, val); |
967f97fa AL |
163 | } |
164 | ||
165 | static inline void vring_used_ring_len(VirtQueue *vq, int i, uint32_t val) | |
166 | { | |
a8170e5e | 167 | hwaddr pa; |
967f97fa | 168 | pa = vq->vring.used + offsetof(VRingUsed, ring[i].len); |
ab1da857 | 169 | stl_phys(&address_space_memory, pa, val); |
967f97fa AL |
170 | } |
171 | ||
172 | static uint16_t vring_used_idx(VirtQueue *vq) | |
173 | { | |
a8170e5e | 174 | hwaddr pa; |
967f97fa | 175 | pa = vq->vring.used + offsetof(VRingUsed, idx); |
41701aa4 | 176 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
177 | } |
178 | ||
bcbabae8 | 179 | static inline void vring_used_idx_set(VirtQueue *vq, uint16_t val) |
967f97fa | 180 | { |
a8170e5e | 181 | hwaddr pa; |
967f97fa | 182 | pa = vq->vring.used + offsetof(VRingUsed, idx); |
5ce5944d | 183 | stw_phys(&address_space_memory, pa, val); |
967f97fa AL |
184 | } |
185 | ||
186 | static inline void vring_used_flags_set_bit(VirtQueue *vq, int mask) | |
187 | { | |
a8170e5e | 188 | hwaddr pa; |
967f97fa | 189 | pa = vq->vring.used + offsetof(VRingUsed, flags); |
5ce5944d EI |
190 | stw_phys(&address_space_memory, |
191 | pa, lduw_phys(&address_space_memory, pa) | mask); | |
967f97fa AL |
192 | } |
193 | ||
194 | static inline void vring_used_flags_unset_bit(VirtQueue *vq, int mask) | |
195 | { | |
a8170e5e | 196 | hwaddr pa; |
967f97fa | 197 | pa = vq->vring.used + offsetof(VRingUsed, flags); |
5ce5944d EI |
198 | stw_phys(&address_space_memory, |
199 | pa, lduw_phys(&address_space_memory, pa) & ~mask); | |
967f97fa AL |
200 | } |
201 | ||
bcbabae8 MT |
202 | static inline void vring_avail_event(VirtQueue *vq, uint16_t val) |
203 | { | |
a8170e5e | 204 | hwaddr pa; |
bcbabae8 MT |
205 | if (!vq->notification) { |
206 | return; | |
207 | } | |
208 | pa = vq->vring.used + offsetof(VRingUsed, ring[vq->vring.num]); | |
5ce5944d | 209 | stw_phys(&address_space_memory, pa, val); |
bcbabae8 MT |
210 | } |
211 | ||
967f97fa AL |
212 | void virtio_queue_set_notification(VirtQueue *vq, int enable) |
213 | { | |
bcbabae8 MT |
214 | vq->notification = enable; |
215 | if (vq->vdev->guest_features & (1 << VIRTIO_RING_F_EVENT_IDX)) { | |
216 | vring_avail_event(vq, vring_avail_idx(vq)); | |
217 | } else if (enable) { | |
967f97fa | 218 | vring_used_flags_unset_bit(vq, VRING_USED_F_NO_NOTIFY); |
bcbabae8 | 219 | } else { |
967f97fa | 220 | vring_used_flags_set_bit(vq, VRING_USED_F_NO_NOTIFY); |
bcbabae8 | 221 | } |
92045d80 MT |
222 | if (enable) { |
223 | /* Expose avail event/used flags before caller checks the avail idx. */ | |
224 | smp_mb(); | |
225 | } | |
967f97fa AL |
226 | } |
227 | ||
228 | int virtio_queue_ready(VirtQueue *vq) | |
229 | { | |
230 | return vq->vring.avail != 0; | |
231 | } | |
232 | ||
233 | int virtio_queue_empty(VirtQueue *vq) | |
234 | { | |
235 | return vring_avail_idx(vq) == vq->last_avail_idx; | |
236 | } | |
237 | ||
238 | void virtqueue_fill(VirtQueue *vq, const VirtQueueElement *elem, | |
239 | unsigned int len, unsigned int idx) | |
240 | { | |
241 | unsigned int offset; | |
242 | int i; | |
243 | ||
64979a4d SH |
244 | trace_virtqueue_fill(vq, elem, len, idx); |
245 | ||
967f97fa AL |
246 | offset = 0; |
247 | for (i = 0; i < elem->in_num; i++) { | |
248 | size_t size = MIN(len - offset, elem->in_sg[i].iov_len); | |
249 | ||
26b258e1 AL |
250 | cpu_physical_memory_unmap(elem->in_sg[i].iov_base, |
251 | elem->in_sg[i].iov_len, | |
252 | 1, size); | |
967f97fa | 253 | |
0cea71a2 | 254 | offset += size; |
967f97fa AL |
255 | } |
256 | ||
26b258e1 AL |
257 | for (i = 0; i < elem->out_num; i++) |
258 | cpu_physical_memory_unmap(elem->out_sg[i].iov_base, | |
259 | elem->out_sg[i].iov_len, | |
260 | 0, elem->out_sg[i].iov_len); | |
261 | ||
967f97fa AL |
262 | idx = (idx + vring_used_idx(vq)) % vq->vring.num; |
263 | ||
264 | /* Get a pointer to the next entry in the used ring. */ | |
265 | vring_used_ring_id(vq, idx, elem->index); | |
266 | vring_used_ring_len(vq, idx, len); | |
267 | } | |
268 | ||
269 | void virtqueue_flush(VirtQueue *vq, unsigned int count) | |
270 | { | |
bcbabae8 | 271 | uint16_t old, new; |
967f97fa | 272 | /* Make sure buffer is written before we update index. */ |
b90d2f35 | 273 | smp_wmb(); |
64979a4d | 274 | trace_virtqueue_flush(vq, count); |
bcbabae8 MT |
275 | old = vring_used_idx(vq); |
276 | new = old + count; | |
277 | vring_used_idx_set(vq, new); | |
967f97fa | 278 | vq->inuse -= count; |
bcbabae8 MT |
279 | if (unlikely((int16_t)(new - vq->signalled_used) < (uint16_t)(new - old))) |
280 | vq->signalled_used_valid = false; | |
967f97fa AL |
281 | } |
282 | ||
283 | void virtqueue_push(VirtQueue *vq, const VirtQueueElement *elem, | |
284 | unsigned int len) | |
285 | { | |
286 | virtqueue_fill(vq, elem, len, 0); | |
287 | virtqueue_flush(vq, 1); | |
288 | } | |
289 | ||
290 | static int virtqueue_num_heads(VirtQueue *vq, unsigned int idx) | |
291 | { | |
292 | uint16_t num_heads = vring_avail_idx(vq) - idx; | |
293 | ||
294 | /* Check it isn't doing very strange things with descriptor numbers. */ | |
bb6834cf | 295 | if (num_heads > vq->vring.num) { |
ce67ed65 SH |
296 | error_report("Guest moved used index from %u to %u", |
297 | idx, vring_avail_idx(vq)); | |
bb6834cf AL |
298 | exit(1); |
299 | } | |
a821ce59 MT |
300 | /* On success, callers read a descriptor at vq->last_avail_idx. |
301 | * Make sure descriptor read does not bypass avail index read. */ | |
302 | if (num_heads) { | |
303 | smp_rmb(); | |
304 | } | |
967f97fa AL |
305 | |
306 | return num_heads; | |
307 | } | |
308 | ||
309 | static unsigned int virtqueue_get_head(VirtQueue *vq, unsigned int idx) | |
310 | { | |
311 | unsigned int head; | |
312 | ||
313 | /* Grab the next descriptor number they're advertising, and increment | |
314 | * the index we've seen. */ | |
315 | head = vring_avail_ring(vq, idx % vq->vring.num); | |
316 | ||
317 | /* If their number is silly, that's a fatal mistake. */ | |
bb6834cf | 318 | if (head >= vq->vring.num) { |
ce67ed65 | 319 | error_report("Guest says index %u is available", head); |
bb6834cf AL |
320 | exit(1); |
321 | } | |
967f97fa AL |
322 | |
323 | return head; | |
324 | } | |
325 | ||
a8170e5e | 326 | static unsigned virtqueue_next_desc(hwaddr desc_pa, |
5774cf98 | 327 | unsigned int i, unsigned int max) |
967f97fa AL |
328 | { |
329 | unsigned int next; | |
330 | ||
331 | /* If this descriptor says it doesn't chain, we're done. */ | |
5774cf98 MM |
332 | if (!(vring_desc_flags(desc_pa, i) & VRING_DESC_F_NEXT)) |
333 | return max; | |
967f97fa AL |
334 | |
335 | /* Check they're not leading us off end of descriptors. */ | |
5774cf98 | 336 | next = vring_desc_next(desc_pa, i); |
967f97fa | 337 | /* Make sure compiler knows to grab that: we don't want it changing! */ |
b90d2f35 | 338 | smp_wmb(); |
967f97fa | 339 | |
5774cf98 | 340 | if (next >= max) { |
ce67ed65 | 341 | error_report("Desc next is %u", next); |
bb6834cf AL |
342 | exit(1); |
343 | } | |
967f97fa AL |
344 | |
345 | return next; | |
346 | } | |
347 | ||
0d8d7690 | 348 | void virtqueue_get_avail_bytes(VirtQueue *vq, unsigned int *in_bytes, |
e1f7b481 MT |
349 | unsigned int *out_bytes, |
350 | unsigned max_in_bytes, unsigned max_out_bytes) | |
967f97fa | 351 | { |
efeea6d0 | 352 | unsigned int idx; |
385ce95d | 353 | unsigned int total_bufs, in_total, out_total; |
967f97fa AL |
354 | |
355 | idx = vq->last_avail_idx; | |
356 | ||
efeea6d0 | 357 | total_bufs = in_total = out_total = 0; |
967f97fa | 358 | while (virtqueue_num_heads(vq, idx)) { |
efeea6d0 | 359 | unsigned int max, num_bufs, indirect = 0; |
a8170e5e | 360 | hwaddr desc_pa; |
967f97fa AL |
361 | int i; |
362 | ||
efeea6d0 MM |
363 | max = vq->vring.num; |
364 | num_bufs = total_bufs; | |
967f97fa | 365 | i = virtqueue_get_head(vq, idx++); |
efeea6d0 MM |
366 | desc_pa = vq->vring.desc; |
367 | ||
368 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_INDIRECT) { | |
369 | if (vring_desc_len(desc_pa, i) % sizeof(VRingDesc)) { | |
ce67ed65 | 370 | error_report("Invalid size for indirect buffer table"); |
efeea6d0 MM |
371 | exit(1); |
372 | } | |
373 | ||
374 | /* If we've got too many, that implies a descriptor loop. */ | |
375 | if (num_bufs >= max) { | |
ce67ed65 | 376 | error_report("Looped descriptor"); |
efeea6d0 MM |
377 | exit(1); |
378 | } | |
379 | ||
380 | /* loop over the indirect descriptor table */ | |
381 | indirect = 1; | |
382 | max = vring_desc_len(desc_pa, i) / sizeof(VRingDesc); | |
efeea6d0 | 383 | desc_pa = vring_desc_addr(desc_pa, i); |
1ae2757c | 384 | num_bufs = i = 0; |
efeea6d0 MM |
385 | } |
386 | ||
967f97fa AL |
387 | do { |
388 | /* If we've got too many, that implies a descriptor loop. */ | |
5774cf98 | 389 | if (++num_bufs > max) { |
ce67ed65 | 390 | error_report("Looped descriptor"); |
bb6834cf AL |
391 | exit(1); |
392 | } | |
967f97fa | 393 | |
5774cf98 | 394 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_WRITE) { |
0d8d7690 | 395 | in_total += vring_desc_len(desc_pa, i); |
967f97fa | 396 | } else { |
0d8d7690 | 397 | out_total += vring_desc_len(desc_pa, i); |
967f97fa | 398 | } |
e1f7b481 MT |
399 | if (in_total >= max_in_bytes && out_total >= max_out_bytes) { |
400 | goto done; | |
401 | } | |
5774cf98 | 402 | } while ((i = virtqueue_next_desc(desc_pa, i, max)) != max); |
efeea6d0 MM |
403 | |
404 | if (!indirect) | |
405 | total_bufs = num_bufs; | |
406 | else | |
407 | total_bufs++; | |
967f97fa | 408 | } |
e1f7b481 | 409 | done: |
0d8d7690 AS |
410 | if (in_bytes) { |
411 | *in_bytes = in_total; | |
412 | } | |
413 | if (out_bytes) { | |
414 | *out_bytes = out_total; | |
415 | } | |
416 | } | |
967f97fa | 417 | |
0d8d7690 AS |
418 | int virtqueue_avail_bytes(VirtQueue *vq, unsigned int in_bytes, |
419 | unsigned int out_bytes) | |
420 | { | |
421 | unsigned int in_total, out_total; | |
422 | ||
e1f7b481 MT |
423 | virtqueue_get_avail_bytes(vq, &in_total, &out_total, in_bytes, out_bytes); |
424 | return in_bytes <= in_total && out_bytes <= out_total; | |
967f97fa AL |
425 | } |
426 | ||
a8170e5e | 427 | void virtqueue_map_sg(struct iovec *sg, hwaddr *addr, |
42fb2e07 KW |
428 | size_t num_sg, int is_write) |
429 | { | |
430 | unsigned int i; | |
a8170e5e | 431 | hwaddr len; |
42fb2e07 | 432 | |
93725140 | 433 | if (num_sg > VIRTQUEUE_MAX_SIZE) { |
36cf2a37 MT |
434 | error_report("virtio: map attempt out of bounds: %zd > %d", |
435 | num_sg, VIRTQUEUE_MAX_SIZE); | |
436 | exit(1); | |
437 | } | |
438 | ||
42fb2e07 KW |
439 | for (i = 0; i < num_sg; i++) { |
440 | len = sg[i].iov_len; | |
441 | sg[i].iov_base = cpu_physical_memory_map(addr[i], &len, is_write); | |
442 | if (sg[i].iov_base == NULL || len != sg[i].iov_len) { | |
1a285899 | 443 | error_report("virtio: error trying to map MMIO memory"); |
42fb2e07 KW |
444 | exit(1); |
445 | } | |
446 | } | |
447 | } | |
448 | ||
967f97fa AL |
449 | int virtqueue_pop(VirtQueue *vq, VirtQueueElement *elem) |
450 | { | |
5774cf98 | 451 | unsigned int i, head, max; |
a8170e5e | 452 | hwaddr desc_pa = vq->vring.desc; |
967f97fa AL |
453 | |
454 | if (!virtqueue_num_heads(vq, vq->last_avail_idx)) | |
455 | return 0; | |
456 | ||
457 | /* When we start there are none of either input nor output. */ | |
458 | elem->out_num = elem->in_num = 0; | |
459 | ||
5774cf98 MM |
460 | max = vq->vring.num; |
461 | ||
967f97fa | 462 | i = head = virtqueue_get_head(vq, vq->last_avail_idx++); |
bcbabae8 MT |
463 | if (vq->vdev->guest_features & (1 << VIRTIO_RING_F_EVENT_IDX)) { |
464 | vring_avail_event(vq, vring_avail_idx(vq)); | |
465 | } | |
efeea6d0 MM |
466 | |
467 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_INDIRECT) { | |
468 | if (vring_desc_len(desc_pa, i) % sizeof(VRingDesc)) { | |
ce67ed65 | 469 | error_report("Invalid size for indirect buffer table"); |
efeea6d0 MM |
470 | exit(1); |
471 | } | |
472 | ||
473 | /* loop over the indirect descriptor table */ | |
474 | max = vring_desc_len(desc_pa, i) / sizeof(VRingDesc); | |
475 | desc_pa = vring_desc_addr(desc_pa, i); | |
476 | i = 0; | |
477 | } | |
478 | ||
42fb2e07 | 479 | /* Collect all the descriptors */ |
967f97fa AL |
480 | do { |
481 | struct iovec *sg; | |
482 | ||
5774cf98 | 483 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_WRITE) { |
c8eac1cf MT |
484 | if (elem->in_num >= ARRAY_SIZE(elem->in_sg)) { |
485 | error_report("Too many write descriptors in indirect table"); | |
486 | exit(1); | |
487 | } | |
5774cf98 | 488 | elem->in_addr[elem->in_num] = vring_desc_addr(desc_pa, i); |
967f97fa | 489 | sg = &elem->in_sg[elem->in_num++]; |
42fb2e07 | 490 | } else { |
c8eac1cf MT |
491 | if (elem->out_num >= ARRAY_SIZE(elem->out_sg)) { |
492 | error_report("Too many read descriptors in indirect table"); | |
493 | exit(1); | |
494 | } | |
42fb2e07 | 495 | elem->out_addr[elem->out_num] = vring_desc_addr(desc_pa, i); |
967f97fa | 496 | sg = &elem->out_sg[elem->out_num++]; |
42fb2e07 | 497 | } |
967f97fa | 498 | |
5774cf98 | 499 | sg->iov_len = vring_desc_len(desc_pa, i); |
967f97fa AL |
500 | |
501 | /* If we've got too many, that implies a descriptor loop. */ | |
5774cf98 | 502 | if ((elem->in_num + elem->out_num) > max) { |
ce67ed65 | 503 | error_report("Looped descriptor"); |
bb6834cf AL |
504 | exit(1); |
505 | } | |
5774cf98 | 506 | } while ((i = virtqueue_next_desc(desc_pa, i, max)) != max); |
967f97fa | 507 | |
42fb2e07 KW |
508 | /* Now map what we have collected */ |
509 | virtqueue_map_sg(elem->in_sg, elem->in_addr, elem->in_num, 1); | |
510 | virtqueue_map_sg(elem->out_sg, elem->out_addr, elem->out_num, 0); | |
511 | ||
967f97fa AL |
512 | elem->index = head; |
513 | ||
514 | vq->inuse++; | |
515 | ||
64979a4d | 516 | trace_virtqueue_pop(vq, elem, elem->in_num, elem->out_num); |
967f97fa AL |
517 | return elem->in_num + elem->out_num; |
518 | } | |
519 | ||
520 | /* virtio device */ | |
7055e687 MT |
521 | static void virtio_notify_vector(VirtIODevice *vdev, uint16_t vector) |
522 | { | |
1c819449 FK |
523 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
524 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
525 | ||
526 | if (k->notify) { | |
527 | k->notify(qbus->parent, vector); | |
7055e687 MT |
528 | } |
529 | } | |
967f97fa | 530 | |
53c25cea | 531 | void virtio_update_irq(VirtIODevice *vdev) |
967f97fa | 532 | { |
7055e687 | 533 | virtio_notify_vector(vdev, VIRTIO_NO_VECTOR); |
967f97fa AL |
534 | } |
535 | ||
4e1837f8 SH |
536 | void virtio_set_status(VirtIODevice *vdev, uint8_t val) |
537 | { | |
181103cd | 538 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
4e1837f8 SH |
539 | trace_virtio_set_status(vdev, val); |
540 | ||
181103cd FK |
541 | if (k->set_status) { |
542 | k->set_status(vdev, val); | |
4e1837f8 SH |
543 | } |
544 | vdev->status = val; | |
545 | } | |
546 | ||
53c25cea | 547 | void virtio_reset(void *opaque) |
967f97fa AL |
548 | { |
549 | VirtIODevice *vdev = opaque; | |
181103cd | 550 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
551 | int i; |
552 | ||
e0c472d8 MT |
553 | virtio_set_status(vdev, 0); |
554 | ||
181103cd FK |
555 | if (k->reset) { |
556 | k->reset(vdev); | |
557 | } | |
967f97fa | 558 | |
704a76fc | 559 | vdev->guest_features = 0; |
967f97fa AL |
560 | vdev->queue_sel = 0; |
561 | vdev->status = 0; | |
562 | vdev->isr = 0; | |
7055e687 MT |
563 | vdev->config_vector = VIRTIO_NO_VECTOR; |
564 | virtio_notify_vector(vdev, vdev->config_vector); | |
967f97fa AL |
565 | |
566 | for(i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
567 | vdev->vq[i].vring.desc = 0; | |
568 | vdev->vq[i].vring.avail = 0; | |
569 | vdev->vq[i].vring.used = 0; | |
570 | vdev->vq[i].last_avail_idx = 0; | |
53c25cea | 571 | vdev->vq[i].pa = 0; |
7055e687 | 572 | vdev->vq[i].vector = VIRTIO_NO_VECTOR; |
bcbabae8 MT |
573 | vdev->vq[i].signalled_used = 0; |
574 | vdev->vq[i].signalled_used_valid = false; | |
575 | vdev->vq[i].notification = true; | |
967f97fa AL |
576 | } |
577 | } | |
578 | ||
53c25cea | 579 | uint32_t virtio_config_readb(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 580 | { |
181103cd | 581 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
582 | uint8_t val; |
583 | ||
5f5a1318 | 584 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 585 | return (uint32_t)-1; |
5f5a1318 JW |
586 | } |
587 | ||
588 | k->get_config(vdev, vdev->config); | |
967f97fa | 589 | |
06dbfc6f | 590 | val = ldub_p(vdev->config + addr); |
967f97fa AL |
591 | return val; |
592 | } | |
593 | ||
53c25cea | 594 | uint32_t virtio_config_readw(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 595 | { |
181103cd | 596 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
597 | uint16_t val; |
598 | ||
5f5a1318 | 599 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 600 | return (uint32_t)-1; |
5f5a1318 JW |
601 | } |
602 | ||
603 | k->get_config(vdev, vdev->config); | |
967f97fa | 604 | |
06dbfc6f | 605 | val = lduw_p(vdev->config + addr); |
967f97fa AL |
606 | return val; |
607 | } | |
608 | ||
53c25cea | 609 | uint32_t virtio_config_readl(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 610 | { |
181103cd | 611 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
612 | uint32_t val; |
613 | ||
5f5a1318 | 614 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 615 | return (uint32_t)-1; |
5f5a1318 JW |
616 | } |
617 | ||
618 | k->get_config(vdev, vdev->config); | |
967f97fa | 619 | |
06dbfc6f | 620 | val = ldl_p(vdev->config + addr); |
967f97fa AL |
621 | return val; |
622 | } | |
623 | ||
53c25cea | 624 | void virtio_config_writeb(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 625 | { |
181103cd | 626 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
627 | uint8_t val = data; |
628 | ||
5f5a1318 | 629 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 630 | return; |
5f5a1318 | 631 | } |
967f97fa | 632 | |
06dbfc6f | 633 | stb_p(vdev->config + addr, val); |
967f97fa | 634 | |
181103cd FK |
635 | if (k->set_config) { |
636 | k->set_config(vdev, vdev->config); | |
637 | } | |
967f97fa AL |
638 | } |
639 | ||
53c25cea | 640 | void virtio_config_writew(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 641 | { |
181103cd | 642 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
643 | uint16_t val = data; |
644 | ||
5f5a1318 | 645 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 646 | return; |
5f5a1318 | 647 | } |
967f97fa | 648 | |
06dbfc6f | 649 | stw_p(vdev->config + addr, val); |
967f97fa | 650 | |
181103cd FK |
651 | if (k->set_config) { |
652 | k->set_config(vdev, vdev->config); | |
653 | } | |
967f97fa AL |
654 | } |
655 | ||
53c25cea | 656 | void virtio_config_writel(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 657 | { |
181103cd | 658 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
659 | uint32_t val = data; |
660 | ||
5f5a1318 | 661 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 662 | return; |
5f5a1318 | 663 | } |
967f97fa | 664 | |
06dbfc6f | 665 | stl_p(vdev->config + addr, val); |
967f97fa | 666 | |
181103cd FK |
667 | if (k->set_config) { |
668 | k->set_config(vdev, vdev->config); | |
669 | } | |
967f97fa AL |
670 | } |
671 | ||
a8170e5e | 672 | void virtio_queue_set_addr(VirtIODevice *vdev, int n, hwaddr addr) |
967f97fa | 673 | { |
7055e687 MT |
674 | vdev->vq[n].pa = addr; |
675 | virtqueue_init(&vdev->vq[n]); | |
53c25cea PB |
676 | } |
677 | ||
a8170e5e | 678 | hwaddr virtio_queue_get_addr(VirtIODevice *vdev, int n) |
53c25cea PB |
679 | { |
680 | return vdev->vq[n].pa; | |
681 | } | |
682 | ||
e63c0ba1 PM |
683 | void virtio_queue_set_num(VirtIODevice *vdev, int n, int num) |
684 | { | |
f6049f44 PM |
685 | /* Don't allow guest to flip queue between existent and |
686 | * nonexistent states, or to set it to an invalid size. | |
687 | */ | |
688 | if (!!num != !!vdev->vq[n].vring.num || | |
689 | num > VIRTQUEUE_MAX_SIZE || | |
690 | num < 0) { | |
691 | return; | |
e63c0ba1 | 692 | } |
f6049f44 PM |
693 | vdev->vq[n].vring.num = num; |
694 | virtqueue_init(&vdev->vq[n]); | |
e63c0ba1 PM |
695 | } |
696 | ||
53c25cea PB |
697 | int virtio_queue_get_num(VirtIODevice *vdev, int n) |
698 | { | |
699 | return vdev->vq[n].vring.num; | |
700 | } | |
967f97fa | 701 | |
c80decdb PB |
702 | int virtio_queue_get_id(VirtQueue *vq) |
703 | { | |
704 | VirtIODevice *vdev = vq->vdev; | |
705 | assert(vq >= &vdev->vq[0] && vq < &vdev->vq[VIRTIO_PCI_QUEUE_MAX]); | |
706 | return vq - &vdev->vq[0]; | |
707 | } | |
708 | ||
6ce69d1c PM |
709 | void virtio_queue_set_align(VirtIODevice *vdev, int n, int align) |
710 | { | |
711 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); | |
712 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
713 | ||
714 | /* Check that the transport told us it was going to do this | |
715 | * (so a buggy transport will immediately assert rather than | |
716 | * silently failing to migrate this state) | |
717 | */ | |
718 | assert(k->has_variable_vring_alignment); | |
719 | ||
720 | vdev->vq[n].vring.align = align; | |
721 | virtqueue_init(&vdev->vq[n]); | |
722 | } | |
723 | ||
25db9ebe SH |
724 | void virtio_queue_notify_vq(VirtQueue *vq) |
725 | { | |
726 | if (vq->vring.desc) { | |
727 | VirtIODevice *vdev = vq->vdev; | |
728 | trace_virtio_queue_notify(vdev, vq - vdev->vq, vq); | |
729 | vq->handle_output(vdev, vq); | |
730 | } | |
731 | } | |
732 | ||
53c25cea PB |
733 | void virtio_queue_notify(VirtIODevice *vdev, int n) |
734 | { | |
7157e2e2 | 735 | virtio_queue_notify_vq(&vdev->vq[n]); |
967f97fa AL |
736 | } |
737 | ||
7055e687 MT |
738 | uint16_t virtio_queue_vector(VirtIODevice *vdev, int n) |
739 | { | |
740 | return n < VIRTIO_PCI_QUEUE_MAX ? vdev->vq[n].vector : | |
741 | VIRTIO_NO_VECTOR; | |
742 | } | |
743 | ||
744 | void virtio_queue_set_vector(VirtIODevice *vdev, int n, uint16_t vector) | |
745 | { | |
746 | if (n < VIRTIO_PCI_QUEUE_MAX) | |
747 | vdev->vq[n].vector = vector; | |
748 | } | |
749 | ||
967f97fa AL |
750 | VirtQueue *virtio_add_queue(VirtIODevice *vdev, int queue_size, |
751 | void (*handle_output)(VirtIODevice *, VirtQueue *)) | |
752 | { | |
753 | int i; | |
754 | ||
755 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
756 | if (vdev->vq[i].vring.num == 0) | |
757 | break; | |
758 | } | |
759 | ||
760 | if (i == VIRTIO_PCI_QUEUE_MAX || queue_size > VIRTQUEUE_MAX_SIZE) | |
761 | abort(); | |
762 | ||
763 | vdev->vq[i].vring.num = queue_size; | |
6ce69d1c | 764 | vdev->vq[i].vring.align = VIRTIO_PCI_VRING_ALIGN; |
967f97fa AL |
765 | vdev->vq[i].handle_output = handle_output; |
766 | ||
767 | return &vdev->vq[i]; | |
768 | } | |
769 | ||
f23fd811 JW |
770 | void virtio_del_queue(VirtIODevice *vdev, int n) |
771 | { | |
772 | if (n < 0 || n >= VIRTIO_PCI_QUEUE_MAX) { | |
773 | abort(); | |
774 | } | |
775 | ||
776 | vdev->vq[n].vring.num = 0; | |
777 | } | |
778 | ||
1cbdabe2 MT |
779 | void virtio_irq(VirtQueue *vq) |
780 | { | |
64979a4d | 781 | trace_virtio_irq(vq); |
1cbdabe2 MT |
782 | vq->vdev->isr |= 0x01; |
783 | virtio_notify_vector(vq->vdev, vq->vector); | |
784 | } | |
785 | ||
bcbabae8 MT |
786 | /* Assuming a given event_idx value from the other size, if |
787 | * we have just incremented index from old to new_idx, | |
788 | * should we trigger an event? */ | |
789 | static inline int vring_need_event(uint16_t event, uint16_t new, uint16_t old) | |
967f97fa | 790 | { |
bcbabae8 MT |
791 | /* Note: Xen has similar logic for notification hold-off |
792 | * in include/xen/interface/io/ring.h with req_event and req_prod | |
793 | * corresponding to event_idx + 1 and new respectively. | |
794 | * Note also that req_event and req_prod in Xen start at 1, | |
795 | * event indexes in virtio start at 0. */ | |
796 | return (uint16_t)(new - event - 1) < (uint16_t)(new - old); | |
797 | } | |
798 | ||
799 | static bool vring_notify(VirtIODevice *vdev, VirtQueue *vq) | |
800 | { | |
801 | uint16_t old, new; | |
802 | bool v; | |
a281ebc1 MT |
803 | /* We need to expose used array entries before checking used event. */ |
804 | smp_mb(); | |
97b83deb | 805 | /* Always notify when queue is empty (when feature acknowledge) */ |
bcbabae8 MT |
806 | if (((vdev->guest_features & (1 << VIRTIO_F_NOTIFY_ON_EMPTY)) && |
807 | !vq->inuse && vring_avail_idx(vq) == vq->last_avail_idx)) { | |
808 | return true; | |
809 | } | |
810 | ||
811 | if (!(vdev->guest_features & (1 << VIRTIO_RING_F_EVENT_IDX))) { | |
812 | return !(vring_avail_flags(vq) & VRING_AVAIL_F_NO_INTERRUPT); | |
813 | } | |
814 | ||
815 | v = vq->signalled_used_valid; | |
816 | vq->signalled_used_valid = true; | |
817 | old = vq->signalled_used; | |
818 | new = vq->signalled_used = vring_used_idx(vq); | |
819 | return !v || vring_need_event(vring_used_event(vq), new, old); | |
820 | } | |
821 | ||
822 | void virtio_notify(VirtIODevice *vdev, VirtQueue *vq) | |
823 | { | |
824 | if (!vring_notify(vdev, vq)) { | |
967f97fa | 825 | return; |
bcbabae8 | 826 | } |
967f97fa | 827 | |
64979a4d | 828 | trace_virtio_notify(vdev, vq); |
967f97fa | 829 | vdev->isr |= 0x01; |
7055e687 | 830 | virtio_notify_vector(vdev, vq->vector); |
967f97fa AL |
831 | } |
832 | ||
833 | void virtio_notify_config(VirtIODevice *vdev) | |
834 | { | |
7625162c AL |
835 | if (!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) |
836 | return; | |
837 | ||
967f97fa | 838 | vdev->isr |= 0x03; |
7055e687 | 839 | virtio_notify_vector(vdev, vdev->config_vector); |
967f97fa AL |
840 | } |
841 | ||
842 | void virtio_save(VirtIODevice *vdev, QEMUFile *f) | |
843 | { | |
1c819449 FK |
844 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
845 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1b5fc0de | 846 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
847 | int i; |
848 | ||
1c819449 FK |
849 | if (k->save_config) { |
850 | k->save_config(qbus->parent, f); | |
851 | } | |
967f97fa | 852 | |
967f97fa AL |
853 | qemu_put_8s(f, &vdev->status); |
854 | qemu_put_8s(f, &vdev->isr); | |
855 | qemu_put_be16s(f, &vdev->queue_sel); | |
704a76fc | 856 | qemu_put_be32s(f, &vdev->guest_features); |
967f97fa AL |
857 | qemu_put_be32(f, vdev->config_len); |
858 | qemu_put_buffer(f, vdev->config, vdev->config_len); | |
859 | ||
860 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
861 | if (vdev->vq[i].vring.num == 0) | |
862 | break; | |
863 | } | |
864 | ||
865 | qemu_put_be32(f, i); | |
866 | ||
867 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
868 | if (vdev->vq[i].vring.num == 0) | |
869 | break; | |
870 | ||
871 | qemu_put_be32(f, vdev->vq[i].vring.num); | |
6ce69d1c PM |
872 | if (k->has_variable_vring_alignment) { |
873 | qemu_put_be32(f, vdev->vq[i].vring.align); | |
874 | } | |
53c25cea | 875 | qemu_put_be64(f, vdev->vq[i].pa); |
967f97fa | 876 | qemu_put_be16s(f, &vdev->vq[i].last_avail_idx); |
1c819449 FK |
877 | if (k->save_queue) { |
878 | k->save_queue(qbus->parent, i, f); | |
879 | } | |
967f97fa | 880 | } |
1b5fc0de GK |
881 | |
882 | if (vdc->save != NULL) { | |
883 | vdc->save(vdev, f); | |
884 | } | |
967f97fa AL |
885 | } |
886 | ||
ad0c9332 PB |
887 | int virtio_set_features(VirtIODevice *vdev, uint32_t val) |
888 | { | |
1c819449 FK |
889 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
890 | VirtioBusClass *vbusk = VIRTIO_BUS_GET_CLASS(qbus); | |
181103cd | 891 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
1c819449 | 892 | uint32_t supported_features = vbusk->get_features(qbus->parent); |
ad0c9332 PB |
893 | bool bad = (val & ~supported_features) != 0; |
894 | ||
895 | val &= supported_features; | |
181103cd FK |
896 | if (k->set_features) { |
897 | k->set_features(vdev, val); | |
ad0c9332 PB |
898 | } |
899 | vdev->guest_features = val; | |
900 | return bad ? -1 : 0; | |
901 | } | |
902 | ||
1b5fc0de | 903 | int virtio_load(VirtIODevice *vdev, QEMUFile *f, int version_id) |
967f97fa | 904 | { |
cc459952 | 905 | int i, ret; |
a890a2f9 | 906 | int32_t config_len; |
cc459952 | 907 | uint32_t num; |
6d74ca5a | 908 | uint32_t features; |
ad0c9332 | 909 | uint32_t supported_features; |
1c819449 FK |
910 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
911 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1b5fc0de | 912 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa | 913 | |
1c819449 FK |
914 | if (k->load_config) { |
915 | ret = k->load_config(qbus->parent, f); | |
ff24bd58 MT |
916 | if (ret) |
917 | return ret; | |
918 | } | |
967f97fa | 919 | |
967f97fa AL |
920 | qemu_get_8s(f, &vdev->status); |
921 | qemu_get_8s(f, &vdev->isr); | |
922 | qemu_get_be16s(f, &vdev->queue_sel); | |
4b53c2c7 MR |
923 | if (vdev->queue_sel >= VIRTIO_PCI_QUEUE_MAX) { |
924 | return -1; | |
925 | } | |
6d74ca5a | 926 | qemu_get_be32s(f, &features); |
ad0c9332 PB |
927 | |
928 | if (virtio_set_features(vdev, features) < 0) { | |
1c819449 | 929 | supported_features = k->get_features(qbus->parent); |
ce67ed65 SH |
930 | error_report("Features 0x%x unsupported. Allowed features: 0x%x", |
931 | features, supported_features); | |
6d74ca5a MT |
932 | return -1; |
933 | } | |
a890a2f9 | 934 | config_len = qemu_get_be32(f); |
2f5732e9 DDAG |
935 | |
936 | /* | |
937 | * There are cases where the incoming config can be bigger or smaller | |
938 | * than what we have; so load what we have space for, and skip | |
939 | * any excess that's in the stream. | |
940 | */ | |
941 | qemu_get_buffer(f, vdev->config, MIN(config_len, vdev->config_len)); | |
942 | ||
943 | while (config_len > vdev->config_len) { | |
944 | qemu_get_byte(f); | |
945 | config_len--; | |
a890a2f9 | 946 | } |
967f97fa AL |
947 | |
948 | num = qemu_get_be32(f); | |
949 | ||
cc459952 MT |
950 | if (num > VIRTIO_PCI_QUEUE_MAX) { |
951 | error_report("Invalid number of PCI queues: 0x%x", num); | |
952 | return -1; | |
953 | } | |
954 | ||
967f97fa AL |
955 | for (i = 0; i < num; i++) { |
956 | vdev->vq[i].vring.num = qemu_get_be32(f); | |
6ce69d1c PM |
957 | if (k->has_variable_vring_alignment) { |
958 | vdev->vq[i].vring.align = qemu_get_be32(f); | |
959 | } | |
53c25cea | 960 | vdev->vq[i].pa = qemu_get_be64(f); |
967f97fa | 961 | qemu_get_be16s(f, &vdev->vq[i].last_avail_idx); |
bcbabae8 MT |
962 | vdev->vq[i].signalled_used_valid = false; |
963 | vdev->vq[i].notification = true; | |
967f97fa | 964 | |
53c25cea | 965 | if (vdev->vq[i].pa) { |
1abeb5a6 | 966 | uint16_t nheads; |
53c25cea | 967 | virtqueue_init(&vdev->vq[i]); |
1abeb5a6 MT |
968 | nheads = vring_avail_idx(&vdev->vq[i]) - vdev->vq[i].last_avail_idx; |
969 | /* Check it isn't doing very strange things with descriptor numbers. */ | |
970 | if (nheads > vdev->vq[i].vring.num) { | |
971 | error_report("VQ %d size 0x%x Guest index 0x%x " | |
6daf194d | 972 | "inconsistent with Host index 0x%x: delta 0x%x", |
1abeb5a6 MT |
973 | i, vdev->vq[i].vring.num, |
974 | vring_avail_idx(&vdev->vq[i]), | |
975 | vdev->vq[i].last_avail_idx, nheads); | |
976 | return -1; | |
977 | } | |
978 | } else if (vdev->vq[i].last_avail_idx) { | |
979 | error_report("VQ %d address 0x0 " | |
6daf194d | 980 | "inconsistent with Host index 0x%x", |
1abeb5a6 MT |
981 | i, vdev->vq[i].last_avail_idx); |
982 | return -1; | |
258dc7c9 | 983 | } |
1c819449 FK |
984 | if (k->load_queue) { |
985 | ret = k->load_queue(qbus->parent, i, f); | |
ff24bd58 MT |
986 | if (ret) |
987 | return ret; | |
7055e687 | 988 | } |
967f97fa AL |
989 | } |
990 | ||
7055e687 | 991 | virtio_notify_vector(vdev, VIRTIO_NO_VECTOR); |
1b5fc0de GK |
992 | |
993 | if (vdc->load != NULL) { | |
994 | return vdc->load(vdev, f, version_id); | |
995 | } | |
996 | ||
ff24bd58 | 997 | return 0; |
967f97fa AL |
998 | } |
999 | ||
6a1a8cc7 | 1000 | void virtio_cleanup(VirtIODevice *vdev) |
b946a153 | 1001 | { |
85cf2a8d | 1002 | qemu_del_vm_change_state_handler(vdev->vmstate); |
6f79e06b | 1003 | g_free(vdev->config); |
7267c094 | 1004 | g_free(vdev->vq); |
8e05db92 FK |
1005 | } |
1006 | ||
1dfb4dd9 | 1007 | static void virtio_vmstate_change(void *opaque, int running, RunState state) |
85cf2a8d MT |
1008 | { |
1009 | VirtIODevice *vdev = opaque; | |
1c819449 FK |
1010 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
1011 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
85cf2a8d MT |
1012 | bool backend_run = running && (vdev->status & VIRTIO_CONFIG_S_DRIVER_OK); |
1013 | vdev->vm_running = running; | |
1014 | ||
1015 | if (backend_run) { | |
1016 | virtio_set_status(vdev, vdev->status); | |
1017 | } | |
1018 | ||
1c819449 FK |
1019 | if (k->vmstate_change) { |
1020 | k->vmstate_change(qbus->parent, backend_run); | |
85cf2a8d MT |
1021 | } |
1022 | ||
1023 | if (!backend_run) { | |
1024 | virtio_set_status(vdev, vdev->status); | |
1025 | } | |
1026 | } | |
1027 | ||
8e05db92 FK |
1028 | void virtio_init(VirtIODevice *vdev, const char *name, |
1029 | uint16_t device_id, size_t config_size) | |
967f97fa | 1030 | { |
b8193adb | 1031 | int i; |
53c25cea | 1032 | vdev->device_id = device_id; |
967f97fa AL |
1033 | vdev->status = 0; |
1034 | vdev->isr = 0; | |
1035 | vdev->queue_sel = 0; | |
7055e687 | 1036 | vdev->config_vector = VIRTIO_NO_VECTOR; |
7267c094 | 1037 | vdev->vq = g_malloc0(sizeof(VirtQueue) * VIRTIO_PCI_QUEUE_MAX); |
1354869c | 1038 | vdev->vm_running = runstate_is_running(); |
8e05db92 | 1039 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { |
b8193adb | 1040 | vdev->vq[i].vector = VIRTIO_NO_VECTOR; |
1cbdabe2 | 1041 | vdev->vq[i].vdev = vdev; |
e78a2b42 | 1042 | vdev->vq[i].queue_index = i; |
1cbdabe2 | 1043 | } |
967f97fa | 1044 | |
967f97fa AL |
1045 | vdev->name = name; |
1046 | vdev->config_len = config_size; | |
8e05db92 | 1047 | if (vdev->config_len) { |
7267c094 | 1048 | vdev->config = g_malloc0(config_size); |
8e05db92 | 1049 | } else { |
967f97fa | 1050 | vdev->config = NULL; |
8e05db92 FK |
1051 | } |
1052 | vdev->vmstate = qemu_add_vm_change_state_handler(virtio_vmstate_change, | |
1053 | vdev); | |
1054 | } | |
967f97fa | 1055 | |
a8170e5e | 1056 | hwaddr virtio_queue_get_desc_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1057 | { |
1058 | return vdev->vq[n].vring.desc; | |
1059 | } | |
1060 | ||
a8170e5e | 1061 | hwaddr virtio_queue_get_avail_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1062 | { |
1063 | return vdev->vq[n].vring.avail; | |
1064 | } | |
1065 | ||
a8170e5e | 1066 | hwaddr virtio_queue_get_used_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1067 | { |
1068 | return vdev->vq[n].vring.used; | |
1069 | } | |
1070 | ||
a8170e5e | 1071 | hwaddr virtio_queue_get_ring_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1072 | { |
1073 | return vdev->vq[n].vring.desc; | |
1074 | } | |
1075 | ||
a8170e5e | 1076 | hwaddr virtio_queue_get_desc_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1077 | { |
1078 | return sizeof(VRingDesc) * vdev->vq[n].vring.num; | |
1079 | } | |
1080 | ||
a8170e5e | 1081 | hwaddr virtio_queue_get_avail_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1082 | { |
1083 | return offsetof(VRingAvail, ring) + | |
2b3af999 | 1084 | sizeof(uint64_t) * vdev->vq[n].vring.num; |
1cbdabe2 MT |
1085 | } |
1086 | ||
a8170e5e | 1087 | hwaddr virtio_queue_get_used_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1088 | { |
1089 | return offsetof(VRingUsed, ring) + | |
1090 | sizeof(VRingUsedElem) * vdev->vq[n].vring.num; | |
1091 | } | |
1092 | ||
a8170e5e | 1093 | hwaddr virtio_queue_get_ring_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1094 | { |
1095 | return vdev->vq[n].vring.used - vdev->vq[n].vring.desc + | |
1096 | virtio_queue_get_used_size(vdev, n); | |
1097 | } | |
1098 | ||
1099 | uint16_t virtio_queue_get_last_avail_idx(VirtIODevice *vdev, int n) | |
1100 | { | |
1101 | return vdev->vq[n].last_avail_idx; | |
1102 | } | |
1103 | ||
1104 | void virtio_queue_set_last_avail_idx(VirtIODevice *vdev, int n, uint16_t idx) | |
1105 | { | |
1106 | vdev->vq[n].last_avail_idx = idx; | |
1107 | } | |
1108 | ||
6793dfd1 SH |
1109 | void virtio_queue_invalidate_signalled_used(VirtIODevice *vdev, int n) |
1110 | { | |
1111 | vdev->vq[n].signalled_used_valid = false; | |
1112 | } | |
1113 | ||
1cbdabe2 MT |
1114 | VirtQueue *virtio_get_queue(VirtIODevice *vdev, int n) |
1115 | { | |
1116 | return vdev->vq + n; | |
1117 | } | |
1118 | ||
e78a2b42 JW |
1119 | uint16_t virtio_get_queue_index(VirtQueue *vq) |
1120 | { | |
1121 | return vq->queue_index; | |
1122 | } | |
1123 | ||
15b2bd18 PB |
1124 | static void virtio_queue_guest_notifier_read(EventNotifier *n) |
1125 | { | |
1126 | VirtQueue *vq = container_of(n, VirtQueue, guest_notifier); | |
1127 | if (event_notifier_test_and_clear(n)) { | |
1128 | virtio_irq(vq); | |
1129 | } | |
1130 | } | |
1131 | ||
1132 | void virtio_queue_set_guest_notifier_fd_handler(VirtQueue *vq, bool assign, | |
1133 | bool with_irqfd) | |
1134 | { | |
1135 | if (assign && !with_irqfd) { | |
1136 | event_notifier_set_handler(&vq->guest_notifier, | |
1137 | virtio_queue_guest_notifier_read); | |
1138 | } else { | |
1139 | event_notifier_set_handler(&vq->guest_notifier, NULL); | |
1140 | } | |
1141 | if (!assign) { | |
1142 | /* Test and clear notifier before closing it, | |
1143 | * in case poll callback didn't have time to run. */ | |
1144 | virtio_queue_guest_notifier_read(&vq->guest_notifier); | |
1145 | } | |
1146 | } | |
1147 | ||
1cbdabe2 MT |
1148 | EventNotifier *virtio_queue_get_guest_notifier(VirtQueue *vq) |
1149 | { | |
1150 | return &vq->guest_notifier; | |
1151 | } | |
b1f416aa PB |
1152 | |
1153 | static void virtio_queue_host_notifier_read(EventNotifier *n) | |
1154 | { | |
1155 | VirtQueue *vq = container_of(n, VirtQueue, host_notifier); | |
1156 | if (event_notifier_test_and_clear(n)) { | |
1157 | virtio_queue_notify_vq(vq); | |
1158 | } | |
1159 | } | |
1160 | ||
26b9b5fe PB |
1161 | void virtio_queue_set_host_notifier_fd_handler(VirtQueue *vq, bool assign, |
1162 | bool set_handler) | |
b1f416aa | 1163 | { |
26b9b5fe | 1164 | if (assign && set_handler) { |
b1f416aa PB |
1165 | event_notifier_set_handler(&vq->host_notifier, |
1166 | virtio_queue_host_notifier_read); | |
1167 | } else { | |
1168 | event_notifier_set_handler(&vq->host_notifier, NULL); | |
26b9b5fe PB |
1169 | } |
1170 | if (!assign) { | |
b1f416aa PB |
1171 | /* Test and clear notifier before after disabling event, |
1172 | * in case poll callback didn't have time to run. */ | |
1173 | virtio_queue_host_notifier_read(&vq->host_notifier); | |
1174 | } | |
1175 | } | |
1176 | ||
1cbdabe2 MT |
1177 | EventNotifier *virtio_queue_get_host_notifier(VirtQueue *vq) |
1178 | { | |
1179 | return &vq->host_notifier; | |
1180 | } | |
8e05db92 | 1181 | |
1034e9cf FK |
1182 | void virtio_device_set_child_bus_name(VirtIODevice *vdev, char *bus_name) |
1183 | { | |
9e288406 | 1184 | g_free(vdev->bus_name); |
80e0090a | 1185 | vdev->bus_name = g_strdup(bus_name); |
1034e9cf FK |
1186 | } |
1187 | ||
1d244b42 AF |
1188 | static void virtio_device_realize(DeviceState *dev, Error **errp) |
1189 | { | |
1190 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); | |
1191 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); | |
1192 | Error *err = NULL; | |
1193 | ||
1d244b42 AF |
1194 | if (vdc->realize != NULL) { |
1195 | vdc->realize(dev, &err); | |
1196 | if (err != NULL) { | |
1197 | error_propagate(errp, err); | |
1198 | return; | |
1199 | } | |
8e05db92 | 1200 | } |
5e96f5d2 | 1201 | virtio_bus_device_plugged(vdev); |
8e05db92 FK |
1202 | } |
1203 | ||
1d244b42 | 1204 | static void virtio_device_unrealize(DeviceState *dev, Error **errp) |
1034e9cf | 1205 | { |
1d244b42 | 1206 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); |
306ec6c3 AF |
1207 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); |
1208 | Error *err = NULL; | |
1d244b42 | 1209 | |
83d07047 PB |
1210 | virtio_bus_device_unplugged(vdev); |
1211 | ||
306ec6c3 AF |
1212 | if (vdc->unrealize != NULL) { |
1213 | vdc->unrealize(dev, &err); | |
1214 | if (err != NULL) { | |
1215 | error_propagate(errp, err); | |
1216 | return; | |
1217 | } | |
5e96f5d2 | 1218 | } |
1d244b42 | 1219 | |
9e288406 MA |
1220 | g_free(vdev->bus_name); |
1221 | vdev->bus_name = NULL; | |
1034e9cf FK |
1222 | } |
1223 | ||
8e05db92 FK |
1224 | static void virtio_device_class_init(ObjectClass *klass, void *data) |
1225 | { | |
1226 | /* Set the default value here. */ | |
1227 | DeviceClass *dc = DEVICE_CLASS(klass); | |
1d244b42 AF |
1228 | |
1229 | dc->realize = virtio_device_realize; | |
1230 | dc->unrealize = virtio_device_unrealize; | |
8e05db92 FK |
1231 | dc->bus_type = TYPE_VIRTIO_BUS; |
1232 | } | |
1233 | ||
1234 | static const TypeInfo virtio_device_info = { | |
1235 | .name = TYPE_VIRTIO_DEVICE, | |
1236 | .parent = TYPE_DEVICE, | |
1237 | .instance_size = sizeof(VirtIODevice), | |
1238 | .class_init = virtio_device_class_init, | |
1239 | .abstract = true, | |
1240 | .class_size = sizeof(VirtioDeviceClass), | |
1241 | }; | |
1242 | ||
1243 | static void virtio_register_types(void) | |
1244 | { | |
1245 | type_register_static(&virtio_device_info); | |
1246 | } | |
1247 | ||
1248 | type_init(virtio_register_types) |