]>
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" |
6b321a3d | 22 | #include "migration/migration.h" |
967f97fa | 23 | |
6ce69d1c PM |
24 | /* |
25 | * The alignment to use between consumer and producer parts of vring. | |
26 | * x86 pagesize again. This is the default, used by transports like PCI | |
27 | * which don't provide a means for the guest to tell the host the alignment. | |
28 | */ | |
f46f15bc AL |
29 | #define VIRTIO_PCI_VRING_ALIGN 4096 |
30 | ||
967f97fa AL |
31 | typedef struct VRingDesc |
32 | { | |
33 | uint64_t addr; | |
34 | uint32_t len; | |
35 | uint16_t flags; | |
36 | uint16_t next; | |
37 | } VRingDesc; | |
38 | ||
39 | typedef struct VRingAvail | |
40 | { | |
41 | uint16_t flags; | |
42 | uint16_t idx; | |
43 | uint16_t ring[0]; | |
44 | } VRingAvail; | |
45 | ||
46 | typedef struct VRingUsedElem | |
47 | { | |
48 | uint32_t id; | |
49 | uint32_t len; | |
50 | } VRingUsedElem; | |
51 | ||
52 | typedef struct VRingUsed | |
53 | { | |
54 | uint16_t flags; | |
55 | uint16_t idx; | |
56 | VRingUsedElem ring[0]; | |
57 | } VRingUsed; | |
58 | ||
59 | typedef struct VRing | |
60 | { | |
61 | unsigned int num; | |
6ce69d1c | 62 | unsigned int align; |
a8170e5e AK |
63 | hwaddr desc; |
64 | hwaddr avail; | |
65 | hwaddr used; | |
967f97fa AL |
66 | } VRing; |
67 | ||
68 | struct VirtQueue | |
69 | { | |
70 | VRing vring; | |
a8170e5e | 71 | hwaddr pa; |
967f97fa | 72 | uint16_t last_avail_idx; |
bcbabae8 MT |
73 | /* Last used index value we have signalled on */ |
74 | uint16_t signalled_used; | |
75 | ||
76 | /* Last used index value we have signalled on */ | |
77 | bool signalled_used_valid; | |
78 | ||
79 | /* Notification enabled? */ | |
80 | bool notification; | |
81 | ||
e78a2b42 JW |
82 | uint16_t queue_index; |
83 | ||
967f97fa | 84 | int inuse; |
bcbabae8 | 85 | |
7055e687 | 86 | uint16_t vector; |
967f97fa | 87 | void (*handle_output)(VirtIODevice *vdev, VirtQueue *vq); |
1cbdabe2 MT |
88 | VirtIODevice *vdev; |
89 | EventNotifier guest_notifier; | |
90 | EventNotifier host_notifier; | |
967f97fa AL |
91 | }; |
92 | ||
967f97fa | 93 | /* virt queue functions */ |
53c25cea | 94 | static void virtqueue_init(VirtQueue *vq) |
967f97fa | 95 | { |
a8170e5e | 96 | hwaddr pa = vq->pa; |
53c25cea | 97 | |
967f97fa AL |
98 | vq->vring.desc = pa; |
99 | vq->vring.avail = pa + vq->vring.num * sizeof(VRingDesc); | |
f46f15bc AL |
100 | vq->vring.used = vring_align(vq->vring.avail + |
101 | offsetof(VRingAvail, ring[vq->vring.num]), | |
6ce69d1c | 102 | vq->vring.align); |
967f97fa AL |
103 | } |
104 | ||
a8170e5e | 105 | static inline uint64_t vring_desc_addr(hwaddr desc_pa, int i) |
967f97fa | 106 | { |
a8170e5e | 107 | hwaddr pa; |
5774cf98 | 108 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, addr); |
2c17449b | 109 | return ldq_phys(&address_space_memory, pa); |
967f97fa AL |
110 | } |
111 | ||
a8170e5e | 112 | static inline uint32_t vring_desc_len(hwaddr desc_pa, int i) |
967f97fa | 113 | { |
a8170e5e | 114 | hwaddr pa; |
5774cf98 | 115 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, len); |
fdfba1a2 | 116 | return ldl_phys(&address_space_memory, pa); |
967f97fa AL |
117 | } |
118 | ||
a8170e5e | 119 | static inline uint16_t vring_desc_flags(hwaddr desc_pa, int i) |
967f97fa | 120 | { |
a8170e5e | 121 | hwaddr pa; |
5774cf98 | 122 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, flags); |
41701aa4 | 123 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
124 | } |
125 | ||
a8170e5e | 126 | static inline uint16_t vring_desc_next(hwaddr desc_pa, int i) |
967f97fa | 127 | { |
a8170e5e | 128 | hwaddr pa; |
5774cf98 | 129 | pa = desc_pa + sizeof(VRingDesc) * i + offsetof(VRingDesc, next); |
41701aa4 | 130 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
131 | } |
132 | ||
133 | static inline uint16_t vring_avail_flags(VirtQueue *vq) | |
134 | { | |
a8170e5e | 135 | hwaddr pa; |
967f97fa | 136 | pa = vq->vring.avail + offsetof(VRingAvail, flags); |
41701aa4 | 137 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
138 | } |
139 | ||
140 | static inline uint16_t vring_avail_idx(VirtQueue *vq) | |
141 | { | |
a8170e5e | 142 | hwaddr pa; |
967f97fa | 143 | pa = vq->vring.avail + offsetof(VRingAvail, idx); |
41701aa4 | 144 | return lduw_phys(&address_space_memory, pa); |
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]); |
41701aa4 | 151 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
152 | } |
153 | ||
bcbabae8 MT |
154 | static inline uint16_t vring_used_event(VirtQueue *vq) |
155 | { | |
156 | return vring_avail_ring(vq, vq->vring.num); | |
157 | } | |
158 | ||
967f97fa AL |
159 | static inline void vring_used_ring_id(VirtQueue *vq, int i, uint32_t val) |
160 | { | |
a8170e5e | 161 | hwaddr pa; |
967f97fa | 162 | pa = vq->vring.used + offsetof(VRingUsed, ring[i].id); |
ab1da857 | 163 | stl_phys(&address_space_memory, pa, val); |
967f97fa AL |
164 | } |
165 | ||
166 | static inline void vring_used_ring_len(VirtQueue *vq, int i, uint32_t val) | |
167 | { | |
a8170e5e | 168 | hwaddr pa; |
967f97fa | 169 | pa = vq->vring.used + offsetof(VRingUsed, ring[i].len); |
ab1da857 | 170 | stl_phys(&address_space_memory, pa, val); |
967f97fa AL |
171 | } |
172 | ||
173 | static uint16_t vring_used_idx(VirtQueue *vq) | |
174 | { | |
a8170e5e | 175 | hwaddr pa; |
967f97fa | 176 | pa = vq->vring.used + offsetof(VRingUsed, idx); |
41701aa4 | 177 | return lduw_phys(&address_space_memory, pa); |
967f97fa AL |
178 | } |
179 | ||
bcbabae8 | 180 | static inline void vring_used_idx_set(VirtQueue *vq, uint16_t val) |
967f97fa | 181 | { |
a8170e5e | 182 | hwaddr pa; |
967f97fa | 183 | pa = vq->vring.used + offsetof(VRingUsed, idx); |
5ce5944d | 184 | stw_phys(&address_space_memory, pa, val); |
967f97fa AL |
185 | } |
186 | ||
187 | static inline void vring_used_flags_set_bit(VirtQueue *vq, int mask) | |
188 | { | |
a8170e5e | 189 | hwaddr pa; |
967f97fa | 190 | pa = vq->vring.used + offsetof(VRingUsed, flags); |
5ce5944d EI |
191 | stw_phys(&address_space_memory, |
192 | pa, lduw_phys(&address_space_memory, pa) | mask); | |
967f97fa AL |
193 | } |
194 | ||
195 | static inline void vring_used_flags_unset_bit(VirtQueue *vq, int mask) | |
196 | { | |
a8170e5e | 197 | hwaddr pa; |
967f97fa | 198 | pa = vq->vring.used + offsetof(VRingUsed, flags); |
5ce5944d EI |
199 | stw_phys(&address_space_memory, |
200 | pa, lduw_phys(&address_space_memory, pa) & ~mask); | |
967f97fa AL |
201 | } |
202 | ||
bcbabae8 MT |
203 | static inline void vring_avail_event(VirtQueue *vq, uint16_t val) |
204 | { | |
a8170e5e | 205 | hwaddr pa; |
bcbabae8 MT |
206 | if (!vq->notification) { |
207 | return; | |
208 | } | |
209 | pa = vq->vring.used + offsetof(VRingUsed, ring[vq->vring.num]); | |
5ce5944d | 210 | stw_phys(&address_space_memory, pa, val); |
bcbabae8 MT |
211 | } |
212 | ||
967f97fa AL |
213 | void virtio_queue_set_notification(VirtQueue *vq, int enable) |
214 | { | |
bcbabae8 MT |
215 | vq->notification = enable; |
216 | if (vq->vdev->guest_features & (1 << VIRTIO_RING_F_EVENT_IDX)) { | |
217 | vring_avail_event(vq, vring_avail_idx(vq)); | |
218 | } else if (enable) { | |
967f97fa | 219 | vring_used_flags_unset_bit(vq, VRING_USED_F_NO_NOTIFY); |
bcbabae8 | 220 | } else { |
967f97fa | 221 | vring_used_flags_set_bit(vq, VRING_USED_F_NO_NOTIFY); |
bcbabae8 | 222 | } |
92045d80 MT |
223 | if (enable) { |
224 | /* Expose avail event/used flags before caller checks the avail idx. */ | |
225 | smp_mb(); | |
226 | } | |
967f97fa AL |
227 | } |
228 | ||
229 | int virtio_queue_ready(VirtQueue *vq) | |
230 | { | |
231 | return vq->vring.avail != 0; | |
232 | } | |
233 | ||
234 | int virtio_queue_empty(VirtQueue *vq) | |
235 | { | |
236 | return vring_avail_idx(vq) == vq->last_avail_idx; | |
237 | } | |
238 | ||
239 | void virtqueue_fill(VirtQueue *vq, const VirtQueueElement *elem, | |
240 | unsigned int len, unsigned int idx) | |
241 | { | |
242 | unsigned int offset; | |
243 | int i; | |
244 | ||
64979a4d SH |
245 | trace_virtqueue_fill(vq, elem, len, idx); |
246 | ||
967f97fa AL |
247 | offset = 0; |
248 | for (i = 0; i < elem->in_num; i++) { | |
249 | size_t size = MIN(len - offset, elem->in_sg[i].iov_len); | |
250 | ||
26b258e1 AL |
251 | cpu_physical_memory_unmap(elem->in_sg[i].iov_base, |
252 | elem->in_sg[i].iov_len, | |
253 | 1, size); | |
967f97fa | 254 | |
0cea71a2 | 255 | offset += size; |
967f97fa AL |
256 | } |
257 | ||
26b258e1 AL |
258 | for (i = 0; i < elem->out_num; i++) |
259 | cpu_physical_memory_unmap(elem->out_sg[i].iov_base, | |
260 | elem->out_sg[i].iov_len, | |
261 | 0, elem->out_sg[i].iov_len); | |
262 | ||
967f97fa AL |
263 | idx = (idx + vring_used_idx(vq)) % vq->vring.num; |
264 | ||
265 | /* Get a pointer to the next entry in the used ring. */ | |
266 | vring_used_ring_id(vq, idx, elem->index); | |
267 | vring_used_ring_len(vq, idx, len); | |
268 | } | |
269 | ||
270 | void virtqueue_flush(VirtQueue *vq, unsigned int count) | |
271 | { | |
bcbabae8 | 272 | uint16_t old, new; |
967f97fa | 273 | /* Make sure buffer is written before we update index. */ |
b90d2f35 | 274 | smp_wmb(); |
64979a4d | 275 | trace_virtqueue_flush(vq, count); |
bcbabae8 MT |
276 | old = vring_used_idx(vq); |
277 | new = old + count; | |
278 | vring_used_idx_set(vq, new); | |
967f97fa | 279 | vq->inuse -= count; |
bcbabae8 MT |
280 | if (unlikely((int16_t)(new - vq->signalled_used) < (uint16_t)(new - old))) |
281 | vq->signalled_used_valid = false; | |
967f97fa AL |
282 | } |
283 | ||
284 | void virtqueue_push(VirtQueue *vq, const VirtQueueElement *elem, | |
285 | unsigned int len) | |
286 | { | |
287 | virtqueue_fill(vq, elem, len, 0); | |
288 | virtqueue_flush(vq, 1); | |
289 | } | |
290 | ||
291 | static int virtqueue_num_heads(VirtQueue *vq, unsigned int idx) | |
292 | { | |
293 | uint16_t num_heads = vring_avail_idx(vq) - idx; | |
294 | ||
295 | /* Check it isn't doing very strange things with descriptor numbers. */ | |
bb6834cf | 296 | if (num_heads > vq->vring.num) { |
ce67ed65 SH |
297 | error_report("Guest moved used index from %u to %u", |
298 | idx, vring_avail_idx(vq)); | |
bb6834cf AL |
299 | exit(1); |
300 | } | |
a821ce59 MT |
301 | /* On success, callers read a descriptor at vq->last_avail_idx. |
302 | * Make sure descriptor read does not bypass avail index read. */ | |
303 | if (num_heads) { | |
304 | smp_rmb(); | |
305 | } | |
967f97fa AL |
306 | |
307 | return num_heads; | |
308 | } | |
309 | ||
310 | static unsigned int virtqueue_get_head(VirtQueue *vq, unsigned int idx) | |
311 | { | |
312 | unsigned int head; | |
313 | ||
314 | /* Grab the next descriptor number they're advertising, and increment | |
315 | * the index we've seen. */ | |
316 | head = vring_avail_ring(vq, idx % vq->vring.num); | |
317 | ||
318 | /* If their number is silly, that's a fatal mistake. */ | |
bb6834cf | 319 | if (head >= vq->vring.num) { |
ce67ed65 | 320 | error_report("Guest says index %u is available", head); |
bb6834cf AL |
321 | exit(1); |
322 | } | |
967f97fa AL |
323 | |
324 | return head; | |
325 | } | |
326 | ||
a8170e5e | 327 | static unsigned virtqueue_next_desc(hwaddr desc_pa, |
5774cf98 | 328 | unsigned int i, unsigned int max) |
967f97fa AL |
329 | { |
330 | unsigned int next; | |
331 | ||
332 | /* If this descriptor says it doesn't chain, we're done. */ | |
5774cf98 MM |
333 | if (!(vring_desc_flags(desc_pa, i) & VRING_DESC_F_NEXT)) |
334 | return max; | |
967f97fa AL |
335 | |
336 | /* Check they're not leading us off end of descriptors. */ | |
5774cf98 | 337 | next = vring_desc_next(desc_pa, i); |
967f97fa | 338 | /* Make sure compiler knows to grab that: we don't want it changing! */ |
b90d2f35 | 339 | smp_wmb(); |
967f97fa | 340 | |
5774cf98 | 341 | if (next >= max) { |
ce67ed65 | 342 | error_report("Desc next is %u", next); |
bb6834cf AL |
343 | exit(1); |
344 | } | |
967f97fa AL |
345 | |
346 | return next; | |
347 | } | |
348 | ||
0d8d7690 | 349 | void virtqueue_get_avail_bytes(VirtQueue *vq, unsigned int *in_bytes, |
e1f7b481 MT |
350 | unsigned int *out_bytes, |
351 | unsigned max_in_bytes, unsigned max_out_bytes) | |
967f97fa | 352 | { |
efeea6d0 | 353 | unsigned int idx; |
385ce95d | 354 | unsigned int total_bufs, in_total, out_total; |
967f97fa AL |
355 | |
356 | idx = vq->last_avail_idx; | |
357 | ||
efeea6d0 | 358 | total_bufs = in_total = out_total = 0; |
967f97fa | 359 | while (virtqueue_num_heads(vq, idx)) { |
efeea6d0 | 360 | unsigned int max, num_bufs, indirect = 0; |
a8170e5e | 361 | hwaddr desc_pa; |
967f97fa AL |
362 | int i; |
363 | ||
efeea6d0 MM |
364 | max = vq->vring.num; |
365 | num_bufs = total_bufs; | |
967f97fa | 366 | i = virtqueue_get_head(vq, idx++); |
efeea6d0 MM |
367 | desc_pa = vq->vring.desc; |
368 | ||
369 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_INDIRECT) { | |
370 | if (vring_desc_len(desc_pa, i) % sizeof(VRingDesc)) { | |
ce67ed65 | 371 | error_report("Invalid size for indirect buffer table"); |
efeea6d0 MM |
372 | exit(1); |
373 | } | |
374 | ||
375 | /* If we've got too many, that implies a descriptor loop. */ | |
376 | if (num_bufs >= max) { | |
ce67ed65 | 377 | error_report("Looped descriptor"); |
efeea6d0 MM |
378 | exit(1); |
379 | } | |
380 | ||
381 | /* loop over the indirect descriptor table */ | |
382 | indirect = 1; | |
383 | max = vring_desc_len(desc_pa, i) / sizeof(VRingDesc); | |
efeea6d0 | 384 | desc_pa = vring_desc_addr(desc_pa, i); |
1ae2757c | 385 | num_bufs = i = 0; |
efeea6d0 MM |
386 | } |
387 | ||
967f97fa AL |
388 | do { |
389 | /* If we've got too many, that implies a descriptor loop. */ | |
5774cf98 | 390 | if (++num_bufs > max) { |
ce67ed65 | 391 | error_report("Looped descriptor"); |
bb6834cf AL |
392 | exit(1); |
393 | } | |
967f97fa | 394 | |
5774cf98 | 395 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_WRITE) { |
0d8d7690 | 396 | in_total += vring_desc_len(desc_pa, i); |
967f97fa | 397 | } else { |
0d8d7690 | 398 | out_total += vring_desc_len(desc_pa, i); |
967f97fa | 399 | } |
e1f7b481 MT |
400 | if (in_total >= max_in_bytes && out_total >= max_out_bytes) { |
401 | goto done; | |
402 | } | |
5774cf98 | 403 | } while ((i = virtqueue_next_desc(desc_pa, i, max)) != max); |
efeea6d0 MM |
404 | |
405 | if (!indirect) | |
406 | total_bufs = num_bufs; | |
407 | else | |
408 | total_bufs++; | |
967f97fa | 409 | } |
e1f7b481 | 410 | done: |
0d8d7690 AS |
411 | if (in_bytes) { |
412 | *in_bytes = in_total; | |
413 | } | |
414 | if (out_bytes) { | |
415 | *out_bytes = out_total; | |
416 | } | |
417 | } | |
967f97fa | 418 | |
0d8d7690 AS |
419 | int virtqueue_avail_bytes(VirtQueue *vq, unsigned int in_bytes, |
420 | unsigned int out_bytes) | |
421 | { | |
422 | unsigned int in_total, out_total; | |
423 | ||
e1f7b481 MT |
424 | virtqueue_get_avail_bytes(vq, &in_total, &out_total, in_bytes, out_bytes); |
425 | return in_bytes <= in_total && out_bytes <= out_total; | |
967f97fa AL |
426 | } |
427 | ||
a8170e5e | 428 | void virtqueue_map_sg(struct iovec *sg, hwaddr *addr, |
42fb2e07 KW |
429 | size_t num_sg, int is_write) |
430 | { | |
431 | unsigned int i; | |
a8170e5e | 432 | hwaddr len; |
42fb2e07 | 433 | |
93725140 | 434 | if (num_sg > VIRTQUEUE_MAX_SIZE) { |
36cf2a37 MT |
435 | error_report("virtio: map attempt out of bounds: %zd > %d", |
436 | num_sg, VIRTQUEUE_MAX_SIZE); | |
437 | exit(1); | |
438 | } | |
439 | ||
42fb2e07 KW |
440 | for (i = 0; i < num_sg; i++) { |
441 | len = sg[i].iov_len; | |
442 | sg[i].iov_base = cpu_physical_memory_map(addr[i], &len, is_write); | |
443 | if (sg[i].iov_base == NULL || len != sg[i].iov_len) { | |
1a285899 | 444 | error_report("virtio: error trying to map MMIO memory"); |
42fb2e07 KW |
445 | exit(1); |
446 | } | |
447 | } | |
448 | } | |
449 | ||
967f97fa AL |
450 | int virtqueue_pop(VirtQueue *vq, VirtQueueElement *elem) |
451 | { | |
5774cf98 | 452 | unsigned int i, head, max; |
a8170e5e | 453 | hwaddr desc_pa = vq->vring.desc; |
967f97fa AL |
454 | |
455 | if (!virtqueue_num_heads(vq, vq->last_avail_idx)) | |
456 | return 0; | |
457 | ||
458 | /* When we start there are none of either input nor output. */ | |
459 | elem->out_num = elem->in_num = 0; | |
460 | ||
5774cf98 MM |
461 | max = vq->vring.num; |
462 | ||
967f97fa | 463 | i = head = virtqueue_get_head(vq, vq->last_avail_idx++); |
bcbabae8 MT |
464 | if (vq->vdev->guest_features & (1 << VIRTIO_RING_F_EVENT_IDX)) { |
465 | vring_avail_event(vq, vring_avail_idx(vq)); | |
466 | } | |
efeea6d0 MM |
467 | |
468 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_INDIRECT) { | |
469 | if (vring_desc_len(desc_pa, i) % sizeof(VRingDesc)) { | |
ce67ed65 | 470 | error_report("Invalid size for indirect buffer table"); |
efeea6d0 MM |
471 | exit(1); |
472 | } | |
473 | ||
474 | /* loop over the indirect descriptor table */ | |
475 | max = vring_desc_len(desc_pa, i) / sizeof(VRingDesc); | |
476 | desc_pa = vring_desc_addr(desc_pa, i); | |
477 | i = 0; | |
478 | } | |
479 | ||
42fb2e07 | 480 | /* Collect all the descriptors */ |
967f97fa AL |
481 | do { |
482 | struct iovec *sg; | |
483 | ||
5774cf98 | 484 | if (vring_desc_flags(desc_pa, i) & VRING_DESC_F_WRITE) { |
c8eac1cf MT |
485 | if (elem->in_num >= ARRAY_SIZE(elem->in_sg)) { |
486 | error_report("Too many write descriptors in indirect table"); | |
487 | exit(1); | |
488 | } | |
5774cf98 | 489 | elem->in_addr[elem->in_num] = vring_desc_addr(desc_pa, i); |
967f97fa | 490 | sg = &elem->in_sg[elem->in_num++]; |
42fb2e07 | 491 | } else { |
c8eac1cf MT |
492 | if (elem->out_num >= ARRAY_SIZE(elem->out_sg)) { |
493 | error_report("Too many read descriptors in indirect table"); | |
494 | exit(1); | |
495 | } | |
42fb2e07 | 496 | elem->out_addr[elem->out_num] = vring_desc_addr(desc_pa, i); |
967f97fa | 497 | sg = &elem->out_sg[elem->out_num++]; |
42fb2e07 | 498 | } |
967f97fa | 499 | |
5774cf98 | 500 | sg->iov_len = vring_desc_len(desc_pa, i); |
967f97fa AL |
501 | |
502 | /* If we've got too many, that implies a descriptor loop. */ | |
5774cf98 | 503 | if ((elem->in_num + elem->out_num) > max) { |
ce67ed65 | 504 | error_report("Looped descriptor"); |
bb6834cf AL |
505 | exit(1); |
506 | } | |
5774cf98 | 507 | } while ((i = virtqueue_next_desc(desc_pa, i, max)) != max); |
967f97fa | 508 | |
42fb2e07 KW |
509 | /* Now map what we have collected */ |
510 | virtqueue_map_sg(elem->in_sg, elem->in_addr, elem->in_num, 1); | |
511 | virtqueue_map_sg(elem->out_sg, elem->out_addr, elem->out_num, 0); | |
512 | ||
967f97fa AL |
513 | elem->index = head; |
514 | ||
515 | vq->inuse++; | |
516 | ||
64979a4d | 517 | trace_virtqueue_pop(vq, elem, elem->in_num, elem->out_num); |
967f97fa AL |
518 | return elem->in_num + elem->out_num; |
519 | } | |
520 | ||
521 | /* virtio device */ | |
7055e687 MT |
522 | static void virtio_notify_vector(VirtIODevice *vdev, uint16_t vector) |
523 | { | |
1c819449 FK |
524 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
525 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
526 | ||
527 | if (k->notify) { | |
528 | k->notify(qbus->parent, vector); | |
7055e687 MT |
529 | } |
530 | } | |
967f97fa | 531 | |
53c25cea | 532 | void virtio_update_irq(VirtIODevice *vdev) |
967f97fa | 533 | { |
7055e687 | 534 | virtio_notify_vector(vdev, VIRTIO_NO_VECTOR); |
967f97fa AL |
535 | } |
536 | ||
4e1837f8 SH |
537 | void virtio_set_status(VirtIODevice *vdev, uint8_t val) |
538 | { | |
181103cd | 539 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
4e1837f8 SH |
540 | trace_virtio_set_status(vdev, val); |
541 | ||
181103cd FK |
542 | if (k->set_status) { |
543 | k->set_status(vdev, val); | |
4e1837f8 SH |
544 | } |
545 | vdev->status = val; | |
546 | } | |
547 | ||
616a6552 GK |
548 | bool target_words_bigendian(void); |
549 | static enum virtio_device_endian virtio_default_endian(void) | |
550 | { | |
551 | if (target_words_bigendian()) { | |
552 | return VIRTIO_DEVICE_ENDIAN_BIG; | |
553 | } else { | |
554 | return VIRTIO_DEVICE_ENDIAN_LITTLE; | |
555 | } | |
556 | } | |
557 | ||
558 | static enum virtio_device_endian virtio_current_cpu_endian(void) | |
559 | { | |
560 | CPUClass *cc = CPU_GET_CLASS(current_cpu); | |
561 | ||
562 | if (cc->virtio_is_big_endian(current_cpu)) { | |
563 | return VIRTIO_DEVICE_ENDIAN_BIG; | |
564 | } else { | |
565 | return VIRTIO_DEVICE_ENDIAN_LITTLE; | |
566 | } | |
567 | } | |
568 | ||
53c25cea | 569 | void virtio_reset(void *opaque) |
967f97fa AL |
570 | { |
571 | VirtIODevice *vdev = opaque; | |
181103cd | 572 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
573 | int i; |
574 | ||
e0c472d8 | 575 | virtio_set_status(vdev, 0); |
616a6552 GK |
576 | if (current_cpu) { |
577 | /* Guest initiated reset */ | |
578 | vdev->device_endian = virtio_current_cpu_endian(); | |
579 | } else { | |
580 | /* System reset */ | |
581 | vdev->device_endian = virtio_default_endian(); | |
582 | } | |
e0c472d8 | 583 | |
181103cd FK |
584 | if (k->reset) { |
585 | k->reset(vdev); | |
586 | } | |
967f97fa | 587 | |
704a76fc | 588 | vdev->guest_features = 0; |
967f97fa AL |
589 | vdev->queue_sel = 0; |
590 | vdev->status = 0; | |
591 | vdev->isr = 0; | |
7055e687 MT |
592 | vdev->config_vector = VIRTIO_NO_VECTOR; |
593 | virtio_notify_vector(vdev, vdev->config_vector); | |
967f97fa AL |
594 | |
595 | for(i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
596 | vdev->vq[i].vring.desc = 0; | |
597 | vdev->vq[i].vring.avail = 0; | |
598 | vdev->vq[i].vring.used = 0; | |
599 | vdev->vq[i].last_avail_idx = 0; | |
53c25cea | 600 | vdev->vq[i].pa = 0; |
7055e687 | 601 | vdev->vq[i].vector = VIRTIO_NO_VECTOR; |
bcbabae8 MT |
602 | vdev->vq[i].signalled_used = 0; |
603 | vdev->vq[i].signalled_used_valid = false; | |
604 | vdev->vq[i].notification = true; | |
967f97fa AL |
605 | } |
606 | } | |
607 | ||
53c25cea | 608 | uint32_t virtio_config_readb(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 609 | { |
181103cd | 610 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
611 | uint8_t val; |
612 | ||
5f5a1318 | 613 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 614 | return (uint32_t)-1; |
5f5a1318 JW |
615 | } |
616 | ||
617 | k->get_config(vdev, vdev->config); | |
967f97fa | 618 | |
06dbfc6f | 619 | val = ldub_p(vdev->config + addr); |
967f97fa AL |
620 | return val; |
621 | } | |
622 | ||
53c25cea | 623 | uint32_t virtio_config_readw(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 624 | { |
181103cd | 625 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
626 | uint16_t val; |
627 | ||
5f5a1318 | 628 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 629 | return (uint32_t)-1; |
5f5a1318 JW |
630 | } |
631 | ||
632 | k->get_config(vdev, vdev->config); | |
967f97fa | 633 | |
06dbfc6f | 634 | val = lduw_p(vdev->config + addr); |
967f97fa AL |
635 | return val; |
636 | } | |
637 | ||
53c25cea | 638 | uint32_t virtio_config_readl(VirtIODevice *vdev, uint32_t addr) |
967f97fa | 639 | { |
181103cd | 640 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
641 | uint32_t val; |
642 | ||
5f5a1318 | 643 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 644 | return (uint32_t)-1; |
5f5a1318 JW |
645 | } |
646 | ||
647 | k->get_config(vdev, vdev->config); | |
967f97fa | 648 | |
06dbfc6f | 649 | val = ldl_p(vdev->config + addr); |
967f97fa AL |
650 | return val; |
651 | } | |
652 | ||
53c25cea | 653 | void virtio_config_writeb(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 654 | { |
181103cd | 655 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
656 | uint8_t val = data; |
657 | ||
5f5a1318 | 658 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 659 | return; |
5f5a1318 | 660 | } |
967f97fa | 661 | |
06dbfc6f | 662 | stb_p(vdev->config + addr, val); |
967f97fa | 663 | |
181103cd FK |
664 | if (k->set_config) { |
665 | k->set_config(vdev, vdev->config); | |
666 | } | |
967f97fa AL |
667 | } |
668 | ||
53c25cea | 669 | void virtio_config_writew(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 670 | { |
181103cd | 671 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
672 | uint16_t val = data; |
673 | ||
5f5a1318 | 674 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 675 | return; |
5f5a1318 | 676 | } |
967f97fa | 677 | |
06dbfc6f | 678 | stw_p(vdev->config + addr, val); |
967f97fa | 679 | |
181103cd FK |
680 | if (k->set_config) { |
681 | k->set_config(vdev, vdev->config); | |
682 | } | |
967f97fa AL |
683 | } |
684 | ||
53c25cea | 685 | void virtio_config_writel(VirtIODevice *vdev, uint32_t addr, uint32_t data) |
967f97fa | 686 | { |
181103cd | 687 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
688 | uint32_t val = data; |
689 | ||
5f5a1318 | 690 | if (addr + sizeof(val) > vdev->config_len) { |
967f97fa | 691 | return; |
5f5a1318 | 692 | } |
967f97fa | 693 | |
06dbfc6f | 694 | stl_p(vdev->config + addr, val); |
967f97fa | 695 | |
181103cd FK |
696 | if (k->set_config) { |
697 | k->set_config(vdev, vdev->config); | |
698 | } | |
967f97fa AL |
699 | } |
700 | ||
a8170e5e | 701 | void virtio_queue_set_addr(VirtIODevice *vdev, int n, hwaddr addr) |
967f97fa | 702 | { |
7055e687 MT |
703 | vdev->vq[n].pa = addr; |
704 | virtqueue_init(&vdev->vq[n]); | |
53c25cea PB |
705 | } |
706 | ||
a8170e5e | 707 | hwaddr virtio_queue_get_addr(VirtIODevice *vdev, int n) |
53c25cea PB |
708 | { |
709 | return vdev->vq[n].pa; | |
710 | } | |
711 | ||
e63c0ba1 PM |
712 | void virtio_queue_set_num(VirtIODevice *vdev, int n, int num) |
713 | { | |
f6049f44 PM |
714 | /* Don't allow guest to flip queue between existent and |
715 | * nonexistent states, or to set it to an invalid size. | |
716 | */ | |
717 | if (!!num != !!vdev->vq[n].vring.num || | |
718 | num > VIRTQUEUE_MAX_SIZE || | |
719 | num < 0) { | |
720 | return; | |
e63c0ba1 | 721 | } |
f6049f44 PM |
722 | vdev->vq[n].vring.num = num; |
723 | virtqueue_init(&vdev->vq[n]); | |
e63c0ba1 PM |
724 | } |
725 | ||
53c25cea PB |
726 | int virtio_queue_get_num(VirtIODevice *vdev, int n) |
727 | { | |
728 | return vdev->vq[n].vring.num; | |
729 | } | |
967f97fa | 730 | |
c80decdb PB |
731 | int virtio_queue_get_id(VirtQueue *vq) |
732 | { | |
733 | VirtIODevice *vdev = vq->vdev; | |
734 | assert(vq >= &vdev->vq[0] && vq < &vdev->vq[VIRTIO_PCI_QUEUE_MAX]); | |
735 | return vq - &vdev->vq[0]; | |
736 | } | |
737 | ||
6ce69d1c PM |
738 | void virtio_queue_set_align(VirtIODevice *vdev, int n, int align) |
739 | { | |
740 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); | |
741 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
742 | ||
743 | /* Check that the transport told us it was going to do this | |
744 | * (so a buggy transport will immediately assert rather than | |
745 | * silently failing to migrate this state) | |
746 | */ | |
747 | assert(k->has_variable_vring_alignment); | |
748 | ||
749 | vdev->vq[n].vring.align = align; | |
750 | virtqueue_init(&vdev->vq[n]); | |
751 | } | |
752 | ||
25db9ebe SH |
753 | void virtio_queue_notify_vq(VirtQueue *vq) |
754 | { | |
755 | if (vq->vring.desc) { | |
756 | VirtIODevice *vdev = vq->vdev; | |
757 | trace_virtio_queue_notify(vdev, vq - vdev->vq, vq); | |
758 | vq->handle_output(vdev, vq); | |
759 | } | |
760 | } | |
761 | ||
53c25cea PB |
762 | void virtio_queue_notify(VirtIODevice *vdev, int n) |
763 | { | |
7157e2e2 | 764 | virtio_queue_notify_vq(&vdev->vq[n]); |
967f97fa AL |
765 | } |
766 | ||
7055e687 MT |
767 | uint16_t virtio_queue_vector(VirtIODevice *vdev, int n) |
768 | { | |
769 | return n < VIRTIO_PCI_QUEUE_MAX ? vdev->vq[n].vector : | |
770 | VIRTIO_NO_VECTOR; | |
771 | } | |
772 | ||
773 | void virtio_queue_set_vector(VirtIODevice *vdev, int n, uint16_t vector) | |
774 | { | |
775 | if (n < VIRTIO_PCI_QUEUE_MAX) | |
776 | vdev->vq[n].vector = vector; | |
777 | } | |
778 | ||
967f97fa AL |
779 | VirtQueue *virtio_add_queue(VirtIODevice *vdev, int queue_size, |
780 | void (*handle_output)(VirtIODevice *, VirtQueue *)) | |
781 | { | |
782 | int i; | |
783 | ||
784 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
785 | if (vdev->vq[i].vring.num == 0) | |
786 | break; | |
787 | } | |
788 | ||
789 | if (i == VIRTIO_PCI_QUEUE_MAX || queue_size > VIRTQUEUE_MAX_SIZE) | |
790 | abort(); | |
791 | ||
792 | vdev->vq[i].vring.num = queue_size; | |
6ce69d1c | 793 | vdev->vq[i].vring.align = VIRTIO_PCI_VRING_ALIGN; |
967f97fa AL |
794 | vdev->vq[i].handle_output = handle_output; |
795 | ||
796 | return &vdev->vq[i]; | |
797 | } | |
798 | ||
f23fd811 JW |
799 | void virtio_del_queue(VirtIODevice *vdev, int n) |
800 | { | |
801 | if (n < 0 || n >= VIRTIO_PCI_QUEUE_MAX) { | |
802 | abort(); | |
803 | } | |
804 | ||
805 | vdev->vq[n].vring.num = 0; | |
806 | } | |
807 | ||
1cbdabe2 MT |
808 | void virtio_irq(VirtQueue *vq) |
809 | { | |
64979a4d | 810 | trace_virtio_irq(vq); |
1cbdabe2 MT |
811 | vq->vdev->isr |= 0x01; |
812 | virtio_notify_vector(vq->vdev, vq->vector); | |
813 | } | |
814 | ||
bcbabae8 MT |
815 | /* Assuming a given event_idx value from the other size, if |
816 | * we have just incremented index from old to new_idx, | |
817 | * should we trigger an event? */ | |
818 | static inline int vring_need_event(uint16_t event, uint16_t new, uint16_t old) | |
967f97fa | 819 | { |
bcbabae8 MT |
820 | /* Note: Xen has similar logic for notification hold-off |
821 | * in include/xen/interface/io/ring.h with req_event and req_prod | |
822 | * corresponding to event_idx + 1 and new respectively. | |
823 | * Note also that req_event and req_prod in Xen start at 1, | |
824 | * event indexes in virtio start at 0. */ | |
825 | return (uint16_t)(new - event - 1) < (uint16_t)(new - old); | |
826 | } | |
827 | ||
828 | static bool vring_notify(VirtIODevice *vdev, VirtQueue *vq) | |
829 | { | |
830 | uint16_t old, new; | |
831 | bool v; | |
a281ebc1 MT |
832 | /* We need to expose used array entries before checking used event. */ |
833 | smp_mb(); | |
97b83deb | 834 | /* Always notify when queue is empty (when feature acknowledge) */ |
bcbabae8 MT |
835 | if (((vdev->guest_features & (1 << VIRTIO_F_NOTIFY_ON_EMPTY)) && |
836 | !vq->inuse && vring_avail_idx(vq) == vq->last_avail_idx)) { | |
837 | return true; | |
838 | } | |
839 | ||
840 | if (!(vdev->guest_features & (1 << VIRTIO_RING_F_EVENT_IDX))) { | |
841 | return !(vring_avail_flags(vq) & VRING_AVAIL_F_NO_INTERRUPT); | |
842 | } | |
843 | ||
844 | v = vq->signalled_used_valid; | |
845 | vq->signalled_used_valid = true; | |
846 | old = vq->signalled_used; | |
847 | new = vq->signalled_used = vring_used_idx(vq); | |
848 | return !v || vring_need_event(vring_used_event(vq), new, old); | |
849 | } | |
850 | ||
851 | void virtio_notify(VirtIODevice *vdev, VirtQueue *vq) | |
852 | { | |
853 | if (!vring_notify(vdev, vq)) { | |
967f97fa | 854 | return; |
bcbabae8 | 855 | } |
967f97fa | 856 | |
64979a4d | 857 | trace_virtio_notify(vdev, vq); |
967f97fa | 858 | vdev->isr |= 0x01; |
7055e687 | 859 | virtio_notify_vector(vdev, vq->vector); |
967f97fa AL |
860 | } |
861 | ||
862 | void virtio_notify_config(VirtIODevice *vdev) | |
863 | { | |
7625162c AL |
864 | if (!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) |
865 | return; | |
866 | ||
967f97fa | 867 | vdev->isr |= 0x03; |
7055e687 | 868 | virtio_notify_vector(vdev, vdev->config_vector); |
967f97fa AL |
869 | } |
870 | ||
616a6552 GK |
871 | static bool virtio_device_endian_needed(void *opaque) |
872 | { | |
873 | VirtIODevice *vdev = opaque; | |
874 | ||
875 | assert(vdev->device_endian != VIRTIO_DEVICE_ENDIAN_UNKNOWN); | |
876 | return vdev->device_endian != virtio_default_endian(); | |
877 | } | |
878 | ||
879 | static const VMStateDescription vmstate_virtio_device_endian = { | |
880 | .name = "virtio/device_endian", | |
881 | .version_id = 1, | |
882 | .minimum_version_id = 1, | |
883 | .fields = (VMStateField[]) { | |
884 | VMSTATE_UINT8(device_endian, VirtIODevice), | |
885 | VMSTATE_END_OF_LIST() | |
886 | } | |
887 | }; | |
888 | ||
6b321a3d GK |
889 | static const VMStateDescription vmstate_virtio = { |
890 | .name = "virtio", | |
891 | .version_id = 1, | |
892 | .minimum_version_id = 1, | |
893 | .minimum_version_id_old = 1, | |
894 | .fields = (VMStateField[]) { | |
895 | VMSTATE_END_OF_LIST() | |
616a6552 GK |
896 | }, |
897 | .subsections = (VMStateSubsection[]) { | |
898 | { | |
899 | .vmsd = &vmstate_virtio_device_endian, | |
900 | .needed = &virtio_device_endian_needed | |
901 | }, | |
902 | { 0 } | |
6b321a3d GK |
903 | } |
904 | }; | |
905 | ||
967f97fa AL |
906 | void virtio_save(VirtIODevice *vdev, QEMUFile *f) |
907 | { | |
1c819449 FK |
908 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
909 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1b5fc0de | 910 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa AL |
911 | int i; |
912 | ||
1c819449 FK |
913 | if (k->save_config) { |
914 | k->save_config(qbus->parent, f); | |
915 | } | |
967f97fa | 916 | |
967f97fa AL |
917 | qemu_put_8s(f, &vdev->status); |
918 | qemu_put_8s(f, &vdev->isr); | |
919 | qemu_put_be16s(f, &vdev->queue_sel); | |
704a76fc | 920 | qemu_put_be32s(f, &vdev->guest_features); |
967f97fa AL |
921 | qemu_put_be32(f, vdev->config_len); |
922 | qemu_put_buffer(f, vdev->config, vdev->config_len); | |
923 | ||
924 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
925 | if (vdev->vq[i].vring.num == 0) | |
926 | break; | |
927 | } | |
928 | ||
929 | qemu_put_be32(f, i); | |
930 | ||
931 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { | |
932 | if (vdev->vq[i].vring.num == 0) | |
933 | break; | |
934 | ||
935 | qemu_put_be32(f, vdev->vq[i].vring.num); | |
6ce69d1c PM |
936 | if (k->has_variable_vring_alignment) { |
937 | qemu_put_be32(f, vdev->vq[i].vring.align); | |
938 | } | |
53c25cea | 939 | qemu_put_be64(f, vdev->vq[i].pa); |
967f97fa | 940 | qemu_put_be16s(f, &vdev->vq[i].last_avail_idx); |
1c819449 FK |
941 | if (k->save_queue) { |
942 | k->save_queue(qbus->parent, i, f); | |
943 | } | |
967f97fa | 944 | } |
1b5fc0de GK |
945 | |
946 | if (vdc->save != NULL) { | |
947 | vdc->save(vdev, f); | |
948 | } | |
6b321a3d GK |
949 | |
950 | /* Subsections */ | |
951 | vmstate_save_state(f, &vmstate_virtio, vdev); | |
967f97fa AL |
952 | } |
953 | ||
ad0c9332 PB |
954 | int virtio_set_features(VirtIODevice *vdev, uint32_t val) |
955 | { | |
1c819449 FK |
956 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
957 | VirtioBusClass *vbusk = VIRTIO_BUS_GET_CLASS(qbus); | |
181103cd | 958 | VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(vdev); |
1c819449 | 959 | uint32_t supported_features = vbusk->get_features(qbus->parent); |
ad0c9332 PB |
960 | bool bad = (val & ~supported_features) != 0; |
961 | ||
962 | val &= supported_features; | |
181103cd FK |
963 | if (k->set_features) { |
964 | k->set_features(vdev, val); | |
ad0c9332 PB |
965 | } |
966 | vdev->guest_features = val; | |
967 | return bad ? -1 : 0; | |
968 | } | |
969 | ||
1b5fc0de | 970 | int virtio_load(VirtIODevice *vdev, QEMUFile *f, int version_id) |
967f97fa | 971 | { |
cc459952 | 972 | int i, ret; |
a890a2f9 | 973 | int32_t config_len; |
cc459952 | 974 | uint32_t num; |
6d74ca5a | 975 | uint32_t features; |
ad0c9332 | 976 | uint32_t supported_features; |
1c819449 FK |
977 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
978 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
1b5fc0de | 979 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(vdev); |
967f97fa | 980 | |
616a6552 GK |
981 | /* |
982 | * We poison the endianness to ensure it does not get used before | |
983 | * subsections have been loaded. | |
984 | */ | |
985 | vdev->device_endian = VIRTIO_DEVICE_ENDIAN_UNKNOWN; | |
986 | ||
1c819449 FK |
987 | if (k->load_config) { |
988 | ret = k->load_config(qbus->parent, f); | |
ff24bd58 MT |
989 | if (ret) |
990 | return ret; | |
991 | } | |
967f97fa | 992 | |
967f97fa AL |
993 | qemu_get_8s(f, &vdev->status); |
994 | qemu_get_8s(f, &vdev->isr); | |
995 | qemu_get_be16s(f, &vdev->queue_sel); | |
4b53c2c7 MR |
996 | if (vdev->queue_sel >= VIRTIO_PCI_QUEUE_MAX) { |
997 | return -1; | |
998 | } | |
6d74ca5a | 999 | qemu_get_be32s(f, &features); |
ad0c9332 PB |
1000 | |
1001 | if (virtio_set_features(vdev, features) < 0) { | |
1c819449 | 1002 | supported_features = k->get_features(qbus->parent); |
ce67ed65 SH |
1003 | error_report("Features 0x%x unsupported. Allowed features: 0x%x", |
1004 | features, supported_features); | |
6d74ca5a MT |
1005 | return -1; |
1006 | } | |
a890a2f9 | 1007 | config_len = qemu_get_be32(f); |
2f5732e9 DDAG |
1008 | |
1009 | /* | |
1010 | * There are cases where the incoming config can be bigger or smaller | |
1011 | * than what we have; so load what we have space for, and skip | |
1012 | * any excess that's in the stream. | |
1013 | */ | |
1014 | qemu_get_buffer(f, vdev->config, MIN(config_len, vdev->config_len)); | |
1015 | ||
1016 | while (config_len > vdev->config_len) { | |
1017 | qemu_get_byte(f); | |
1018 | config_len--; | |
a890a2f9 | 1019 | } |
967f97fa AL |
1020 | |
1021 | num = qemu_get_be32(f); | |
1022 | ||
cc459952 MT |
1023 | if (num > VIRTIO_PCI_QUEUE_MAX) { |
1024 | error_report("Invalid number of PCI queues: 0x%x", num); | |
1025 | return -1; | |
1026 | } | |
1027 | ||
967f97fa AL |
1028 | for (i = 0; i < num; i++) { |
1029 | vdev->vq[i].vring.num = qemu_get_be32(f); | |
6ce69d1c PM |
1030 | if (k->has_variable_vring_alignment) { |
1031 | vdev->vq[i].vring.align = qemu_get_be32(f); | |
1032 | } | |
53c25cea | 1033 | vdev->vq[i].pa = qemu_get_be64(f); |
967f97fa | 1034 | qemu_get_be16s(f, &vdev->vq[i].last_avail_idx); |
bcbabae8 MT |
1035 | vdev->vq[i].signalled_used_valid = false; |
1036 | vdev->vq[i].notification = true; | |
967f97fa | 1037 | |
53c25cea PB |
1038 | if (vdev->vq[i].pa) { |
1039 | virtqueue_init(&vdev->vq[i]); | |
1abeb5a6 MT |
1040 | } else if (vdev->vq[i].last_avail_idx) { |
1041 | error_report("VQ %d address 0x0 " | |
6daf194d | 1042 | "inconsistent with Host index 0x%x", |
1abeb5a6 MT |
1043 | i, vdev->vq[i].last_avail_idx); |
1044 | return -1; | |
258dc7c9 | 1045 | } |
1c819449 FK |
1046 | if (k->load_queue) { |
1047 | ret = k->load_queue(qbus->parent, i, f); | |
ff24bd58 MT |
1048 | if (ret) |
1049 | return ret; | |
7055e687 | 1050 | } |
967f97fa AL |
1051 | } |
1052 | ||
7055e687 | 1053 | virtio_notify_vector(vdev, VIRTIO_NO_VECTOR); |
1b5fc0de GK |
1054 | |
1055 | if (vdc->load != NULL) { | |
6b321a3d GK |
1056 | ret = vdc->load(vdev, f, version_id); |
1057 | if (ret) { | |
1058 | return ret; | |
1059 | } | |
1b5fc0de GK |
1060 | } |
1061 | ||
616a6552 GK |
1062 | /* Subsections */ |
1063 | ret = vmstate_load_state(f, &vmstate_virtio, vdev, 1); | |
1064 | if (ret) { | |
1065 | return ret; | |
1066 | } | |
1067 | ||
1068 | if (vdev->device_endian == VIRTIO_DEVICE_ENDIAN_UNKNOWN) { | |
1069 | vdev->device_endian = virtio_default_endian(); | |
1070 | } | |
1071 | ||
1072 | for (i = 0; i < num; i++) { | |
1073 | if (vdev->vq[i].pa) { | |
1074 | uint16_t nheads; | |
1075 | nheads = vring_avail_idx(&vdev->vq[i]) - vdev->vq[i].last_avail_idx; | |
1076 | /* Check it isn't doing strange things with descriptor numbers. */ | |
1077 | if (nheads > vdev->vq[i].vring.num) { | |
1078 | error_report("VQ %d size 0x%x Guest index 0x%x " | |
1079 | "inconsistent with Host index 0x%x: delta 0x%x", | |
1080 | i, vdev->vq[i].vring.num, | |
1081 | vring_avail_idx(&vdev->vq[i]), | |
1082 | vdev->vq[i].last_avail_idx, nheads); | |
1083 | return -1; | |
1084 | } | |
1085 | } | |
1086 | } | |
1087 | ||
1088 | return 0; | |
967f97fa AL |
1089 | } |
1090 | ||
6a1a8cc7 | 1091 | void virtio_cleanup(VirtIODevice *vdev) |
b946a153 | 1092 | { |
85cf2a8d | 1093 | qemu_del_vm_change_state_handler(vdev->vmstate); |
6f79e06b | 1094 | g_free(vdev->config); |
7267c094 | 1095 | g_free(vdev->vq); |
8e05db92 FK |
1096 | } |
1097 | ||
1dfb4dd9 | 1098 | static void virtio_vmstate_change(void *opaque, int running, RunState state) |
85cf2a8d MT |
1099 | { |
1100 | VirtIODevice *vdev = opaque; | |
1c819449 FK |
1101 | BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); |
1102 | VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus); | |
85cf2a8d MT |
1103 | bool backend_run = running && (vdev->status & VIRTIO_CONFIG_S_DRIVER_OK); |
1104 | vdev->vm_running = running; | |
1105 | ||
1106 | if (backend_run) { | |
1107 | virtio_set_status(vdev, vdev->status); | |
1108 | } | |
1109 | ||
1c819449 FK |
1110 | if (k->vmstate_change) { |
1111 | k->vmstate_change(qbus->parent, backend_run); | |
85cf2a8d MT |
1112 | } |
1113 | ||
1114 | if (!backend_run) { | |
1115 | virtio_set_status(vdev, vdev->status); | |
1116 | } | |
1117 | } | |
1118 | ||
8e05db92 FK |
1119 | void virtio_init(VirtIODevice *vdev, const char *name, |
1120 | uint16_t device_id, size_t config_size) | |
967f97fa | 1121 | { |
b8193adb | 1122 | int i; |
53c25cea | 1123 | vdev->device_id = device_id; |
967f97fa AL |
1124 | vdev->status = 0; |
1125 | vdev->isr = 0; | |
1126 | vdev->queue_sel = 0; | |
7055e687 | 1127 | vdev->config_vector = VIRTIO_NO_VECTOR; |
7267c094 | 1128 | vdev->vq = g_malloc0(sizeof(VirtQueue) * VIRTIO_PCI_QUEUE_MAX); |
1354869c | 1129 | vdev->vm_running = runstate_is_running(); |
8e05db92 | 1130 | for (i = 0; i < VIRTIO_PCI_QUEUE_MAX; i++) { |
b8193adb | 1131 | vdev->vq[i].vector = VIRTIO_NO_VECTOR; |
1cbdabe2 | 1132 | vdev->vq[i].vdev = vdev; |
e78a2b42 | 1133 | vdev->vq[i].queue_index = i; |
1cbdabe2 | 1134 | } |
967f97fa | 1135 | |
967f97fa AL |
1136 | vdev->name = name; |
1137 | vdev->config_len = config_size; | |
8e05db92 | 1138 | if (vdev->config_len) { |
7267c094 | 1139 | vdev->config = g_malloc0(config_size); |
8e05db92 | 1140 | } else { |
967f97fa | 1141 | vdev->config = NULL; |
8e05db92 FK |
1142 | } |
1143 | vdev->vmstate = qemu_add_vm_change_state_handler(virtio_vmstate_change, | |
1144 | vdev); | |
616a6552 | 1145 | vdev->device_endian = virtio_default_endian(); |
8e05db92 | 1146 | } |
967f97fa | 1147 | |
a8170e5e | 1148 | hwaddr virtio_queue_get_desc_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1149 | { |
1150 | return vdev->vq[n].vring.desc; | |
1151 | } | |
1152 | ||
a8170e5e | 1153 | hwaddr virtio_queue_get_avail_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1154 | { |
1155 | return vdev->vq[n].vring.avail; | |
1156 | } | |
1157 | ||
a8170e5e | 1158 | hwaddr virtio_queue_get_used_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1159 | { |
1160 | return vdev->vq[n].vring.used; | |
1161 | } | |
1162 | ||
a8170e5e | 1163 | hwaddr virtio_queue_get_ring_addr(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1164 | { |
1165 | return vdev->vq[n].vring.desc; | |
1166 | } | |
1167 | ||
a8170e5e | 1168 | hwaddr virtio_queue_get_desc_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1169 | { |
1170 | return sizeof(VRingDesc) * vdev->vq[n].vring.num; | |
1171 | } | |
1172 | ||
a8170e5e | 1173 | hwaddr virtio_queue_get_avail_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1174 | { |
1175 | return offsetof(VRingAvail, ring) + | |
2b3af999 | 1176 | sizeof(uint64_t) * vdev->vq[n].vring.num; |
1cbdabe2 MT |
1177 | } |
1178 | ||
a8170e5e | 1179 | hwaddr virtio_queue_get_used_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1180 | { |
1181 | return offsetof(VRingUsed, ring) + | |
1182 | sizeof(VRingUsedElem) * vdev->vq[n].vring.num; | |
1183 | } | |
1184 | ||
a8170e5e | 1185 | hwaddr virtio_queue_get_ring_size(VirtIODevice *vdev, int n) |
1cbdabe2 MT |
1186 | { |
1187 | return vdev->vq[n].vring.used - vdev->vq[n].vring.desc + | |
1188 | virtio_queue_get_used_size(vdev, n); | |
1189 | } | |
1190 | ||
1191 | uint16_t virtio_queue_get_last_avail_idx(VirtIODevice *vdev, int n) | |
1192 | { | |
1193 | return vdev->vq[n].last_avail_idx; | |
1194 | } | |
1195 | ||
1196 | void virtio_queue_set_last_avail_idx(VirtIODevice *vdev, int n, uint16_t idx) | |
1197 | { | |
1198 | vdev->vq[n].last_avail_idx = idx; | |
1199 | } | |
1200 | ||
6793dfd1 SH |
1201 | void virtio_queue_invalidate_signalled_used(VirtIODevice *vdev, int n) |
1202 | { | |
1203 | vdev->vq[n].signalled_used_valid = false; | |
1204 | } | |
1205 | ||
1cbdabe2 MT |
1206 | VirtQueue *virtio_get_queue(VirtIODevice *vdev, int n) |
1207 | { | |
1208 | return vdev->vq + n; | |
1209 | } | |
1210 | ||
e78a2b42 JW |
1211 | uint16_t virtio_get_queue_index(VirtQueue *vq) |
1212 | { | |
1213 | return vq->queue_index; | |
1214 | } | |
1215 | ||
15b2bd18 PB |
1216 | static void virtio_queue_guest_notifier_read(EventNotifier *n) |
1217 | { | |
1218 | VirtQueue *vq = container_of(n, VirtQueue, guest_notifier); | |
1219 | if (event_notifier_test_and_clear(n)) { | |
1220 | virtio_irq(vq); | |
1221 | } | |
1222 | } | |
1223 | ||
1224 | void virtio_queue_set_guest_notifier_fd_handler(VirtQueue *vq, bool assign, | |
1225 | bool with_irqfd) | |
1226 | { | |
1227 | if (assign && !with_irqfd) { | |
1228 | event_notifier_set_handler(&vq->guest_notifier, | |
1229 | virtio_queue_guest_notifier_read); | |
1230 | } else { | |
1231 | event_notifier_set_handler(&vq->guest_notifier, NULL); | |
1232 | } | |
1233 | if (!assign) { | |
1234 | /* Test and clear notifier before closing it, | |
1235 | * in case poll callback didn't have time to run. */ | |
1236 | virtio_queue_guest_notifier_read(&vq->guest_notifier); | |
1237 | } | |
1238 | } | |
1239 | ||
1cbdabe2 MT |
1240 | EventNotifier *virtio_queue_get_guest_notifier(VirtQueue *vq) |
1241 | { | |
1242 | return &vq->guest_notifier; | |
1243 | } | |
b1f416aa PB |
1244 | |
1245 | static void virtio_queue_host_notifier_read(EventNotifier *n) | |
1246 | { | |
1247 | VirtQueue *vq = container_of(n, VirtQueue, host_notifier); | |
1248 | if (event_notifier_test_and_clear(n)) { | |
1249 | virtio_queue_notify_vq(vq); | |
1250 | } | |
1251 | } | |
1252 | ||
26b9b5fe PB |
1253 | void virtio_queue_set_host_notifier_fd_handler(VirtQueue *vq, bool assign, |
1254 | bool set_handler) | |
b1f416aa | 1255 | { |
26b9b5fe | 1256 | if (assign && set_handler) { |
b1f416aa PB |
1257 | event_notifier_set_handler(&vq->host_notifier, |
1258 | virtio_queue_host_notifier_read); | |
1259 | } else { | |
1260 | event_notifier_set_handler(&vq->host_notifier, NULL); | |
26b9b5fe PB |
1261 | } |
1262 | if (!assign) { | |
b1f416aa PB |
1263 | /* Test and clear notifier before after disabling event, |
1264 | * in case poll callback didn't have time to run. */ | |
1265 | virtio_queue_host_notifier_read(&vq->host_notifier); | |
1266 | } | |
1267 | } | |
1268 | ||
1cbdabe2 MT |
1269 | EventNotifier *virtio_queue_get_host_notifier(VirtQueue *vq) |
1270 | { | |
1271 | return &vq->host_notifier; | |
1272 | } | |
8e05db92 | 1273 | |
1034e9cf FK |
1274 | void virtio_device_set_child_bus_name(VirtIODevice *vdev, char *bus_name) |
1275 | { | |
9e288406 | 1276 | g_free(vdev->bus_name); |
80e0090a | 1277 | vdev->bus_name = g_strdup(bus_name); |
1034e9cf FK |
1278 | } |
1279 | ||
1d244b42 AF |
1280 | static void virtio_device_realize(DeviceState *dev, Error **errp) |
1281 | { | |
1282 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); | |
1283 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); | |
1284 | Error *err = NULL; | |
1285 | ||
1d244b42 AF |
1286 | if (vdc->realize != NULL) { |
1287 | vdc->realize(dev, &err); | |
1288 | if (err != NULL) { | |
1289 | error_propagate(errp, err); | |
1290 | return; | |
1291 | } | |
8e05db92 | 1292 | } |
5e96f5d2 | 1293 | virtio_bus_device_plugged(vdev); |
8e05db92 FK |
1294 | } |
1295 | ||
1d244b42 | 1296 | static void virtio_device_unrealize(DeviceState *dev, Error **errp) |
1034e9cf | 1297 | { |
1d244b42 | 1298 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); |
306ec6c3 AF |
1299 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); |
1300 | Error *err = NULL; | |
1d244b42 | 1301 | |
83d07047 PB |
1302 | virtio_bus_device_unplugged(vdev); |
1303 | ||
306ec6c3 AF |
1304 | if (vdc->unrealize != NULL) { |
1305 | vdc->unrealize(dev, &err); | |
1306 | if (err != NULL) { | |
1307 | error_propagate(errp, err); | |
1308 | return; | |
1309 | } | |
5e96f5d2 | 1310 | } |
1d244b42 | 1311 | |
9e288406 MA |
1312 | g_free(vdev->bus_name); |
1313 | vdev->bus_name = NULL; | |
1034e9cf FK |
1314 | } |
1315 | ||
8e05db92 FK |
1316 | static void virtio_device_class_init(ObjectClass *klass, void *data) |
1317 | { | |
1318 | /* Set the default value here. */ | |
1319 | DeviceClass *dc = DEVICE_CLASS(klass); | |
1d244b42 AF |
1320 | |
1321 | dc->realize = virtio_device_realize; | |
1322 | dc->unrealize = virtio_device_unrealize; | |
8e05db92 FK |
1323 | dc->bus_type = TYPE_VIRTIO_BUS; |
1324 | } | |
1325 | ||
1326 | static const TypeInfo virtio_device_info = { | |
1327 | .name = TYPE_VIRTIO_DEVICE, | |
1328 | .parent = TYPE_DEVICE, | |
1329 | .instance_size = sizeof(VirtIODevice), | |
1330 | .class_init = virtio_device_class_init, | |
1331 | .abstract = true, | |
1332 | .class_size = sizeof(VirtioDeviceClass), | |
1333 | }; | |
1334 | ||
1335 | static void virtio_register_types(void) | |
1336 | { | |
1337 | type_register_static(&virtio_device_info); | |
1338 | } | |
1339 | ||
1340 | type_init(virtio_register_types) |