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