]>
Commit | Line | Data |
---|---|---|
75818250 TS |
1 | /* |
2 | * QEMU Block driver for NBD | |
3 | * | |
f7651539 | 4 | * Copyright (c) 2019 Virtuozzo International GmbH. |
86f8cdf3 | 5 | * Copyright (C) 2016 Red Hat, Inc. |
75818250 | 6 | * Copyright (C) 2008 Bull S.A.S. |
bd5921b4 | 7 | * Author: Laurent Vivier <Laurent.Vivier@bull.net> |
75818250 TS |
8 | * |
9 | * Some parts: | |
10 | * Copyright (C) 2007 Anthony Liguori <anthony@codemonkey.ws> | |
11 | * | |
12 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
13 | * of this software and associated documentation files (the "Software"), to deal | |
14 | * in the Software without restriction, including without limitation the rights | |
15 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
16 | * copies of the Software, and to permit persons to whom the Software is | |
17 | * furnished to do so, subject to the following conditions: | |
18 | * | |
19 | * The above copyright notice and this permission notice shall be included in | |
20 | * all copies or substantial portions of the Software. | |
21 | * | |
22 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
23 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
24 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
25 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
26 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
27 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
28 | * THE SOFTWARE. | |
29 | */ | |
30 | ||
80c71a24 | 31 | #include "qemu/osdep.h" |
86f8cdf3 VSO |
32 | |
33 | #include "trace.h" | |
1de7afc9 | 34 | #include "qemu/uri.h" |
922a01a0 | 35 | #include "qemu/option.h" |
86f8cdf3 | 36 | #include "qemu/cutils.h" |
db725815 | 37 | #include "qemu/main-loop.h" |
fee091cd | 38 | #include "qemu/atomic.h" |
86f8cdf3 | 39 | |
9af23989 | 40 | #include "qapi/qapi-visit-sockets.h" |
2019d68b | 41 | #include "qapi/qmp/qstring.h" |
1dc4718d | 42 | #include "qapi/clone-visitor.h" |
86f8cdf3 VSO |
43 | |
44 | #include "block/qdict.h" | |
45 | #include "block/nbd.h" | |
46 | #include "block/block_int.h" | |
a71d597b | 47 | #include "block/coroutines.h" |
75818250 | 48 | |
fee091cd LS |
49 | #include "qemu/yank.h" |
50 | ||
1d45f8b5 | 51 | #define EN_OPTSTR ":exportname=" |
86f8cdf3 VSO |
52 | #define MAX_NBD_REQUESTS 16 |
53 | ||
54 | #define HANDLE_TO_INDEX(bs, handle) ((handle) ^ (uint64_t)(intptr_t)(bs)) | |
55 | #define INDEX_TO_HANDLE(bs, index) ((index) ^ (uint64_t)(intptr_t)(bs)) | |
56 | ||
57 | typedef struct { | |
58 | Coroutine *coroutine; | |
59 | uint64_t offset; /* original offset of the request */ | |
4ddb5d2f | 60 | bool receiving; /* sleeping in the yield in nbd_receive_replies */ |
1af77378 | 61 | bool reply_possible; /* reply header not yet received */ |
86f8cdf3 VSO |
62 | } NBDClientRequest; |
63 | ||
a34b1e5e | 64 | typedef enum NBDClientState { |
f7651539 VSO |
65 | NBD_CLIENT_CONNECTING_WAIT, |
66 | NBD_CLIENT_CONNECTING_NOWAIT, | |
a34b1e5e VSO |
67 | NBD_CLIENT_CONNECTED, |
68 | NBD_CLIENT_QUIT | |
69 | } NBDClientState; | |
70 | ||
611ae1d7 | 71 | typedef struct BDRVNBDState { |
95a078ea | 72 | QIOChannel *ioc; /* The current I/O channel */ |
86f8cdf3 VSO |
73 | NBDExportInfo info; |
74 | ||
75 | CoMutex send_mutex; | |
76 | CoQueue free_sema; | |
4ddb5d2f VSO |
77 | |
78 | CoMutex receive_mutex; | |
86f8cdf3 | 79 | int in_flight; |
a34b1e5e | 80 | NBDClientState state; |
86f8cdf3 | 81 | |
46f56631 | 82 | QEMUTimer *reconnect_delay_timer; |
be16b8bf | 83 | QEMUTimer *open_timer; |
46f56631 | 84 | |
86f8cdf3 VSO |
85 | NBDClientRequest requests[MAX_NBD_REQUESTS]; |
86 | NBDReply reply; | |
87 | BlockDriverState *bs; | |
03504d05 | 88 | |
8f071c9d VSO |
89 | /* Connection parameters */ |
90 | uint32_t reconnect_delay; | |
be16b8bf | 91 | uint32_t open_timeout; |
62cf396b | 92 | SocketAddress *saddr; |
491d6c7c | 93 | char *export, *tlscredsid; |
8f071c9d VSO |
94 | QCryptoTLSCreds *tlscreds; |
95 | const char *hostname; | |
96 | char *x_dirty_bitmap; | |
dbc7b014 | 97 | bool alloc_depth; |
1dc4718d | 98 | |
90ddc64f | 99 | NBDClientConnection *conn; |
75818250 TS |
100 | } BDRVNBDState; |
101 | ||
fee091cd | 102 | static void nbd_yank(void *opaque); |
f7651539 | 103 | |
bbba1c37 | 104 | static void nbd_clear_bdrvstate(BlockDriverState *bs) |
7f493662 | 105 | { |
bbba1c37 | 106 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
e8b35bf5 | 107 | |
248d4701 VSO |
108 | nbd_client_connection_release(s->conn); |
109 | s->conn = NULL; | |
bbba1c37 VSO |
110 | |
111 | yank_unregister_instance(BLOCKDEV_YANK_INSTANCE(bs->node_name)); | |
112 | ||
7f493662 PN |
113 | object_unref(OBJECT(s->tlscreds)); |
114 | qapi_free_SocketAddress(s->saddr); | |
115 | s->saddr = NULL; | |
116 | g_free(s->export); | |
117 | s->export = NULL; | |
118 | g_free(s->tlscredsid); | |
119 | s->tlscredsid = NULL; | |
120 | g_free(s->x_dirty_bitmap); | |
121 | s->x_dirty_bitmap = NULL; | |
122 | } | |
123 | ||
91e0998f VSO |
124 | static bool nbd_client_connected(BDRVNBDState *s) |
125 | { | |
126 | return qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTED; | |
127 | } | |
128 | ||
04a953b2 VSO |
129 | static bool nbd_recv_coroutine_wake_one(NBDClientRequest *req) |
130 | { | |
131 | if (req->receiving) { | |
132 | req->receiving = false; | |
133 | aio_co_wake(req->coroutine); | |
134 | return true; | |
135 | } | |
136 | ||
137 | return false; | |
138 | } | |
139 | ||
140 | static void nbd_recv_coroutines_wake(BDRVNBDState *s, bool all) | |
3bc0bd1f VSO |
141 | { |
142 | int i; | |
143 | ||
144 | for (i = 0; i < MAX_NBD_REQUESTS; i++) { | |
04a953b2 VSO |
145 | if (nbd_recv_coroutine_wake_one(&s->requests[i]) && !all) { |
146 | return; | |
3bc0bd1f VSO |
147 | } |
148 | } | |
149 | } | |
150 | ||
a34b1e5e VSO |
151 | static void nbd_channel_error(BDRVNBDState *s, int ret) |
152 | { | |
cb116da7 VSO |
153 | if (nbd_client_connected(s)) { |
154 | qio_channel_shutdown(s->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); | |
155 | } | |
156 | ||
f7651539 | 157 | if (ret == -EIO) { |
91e0998f | 158 | if (nbd_client_connected(s)) { |
f7651539 VSO |
159 | s->state = s->reconnect_delay ? NBD_CLIENT_CONNECTING_WAIT : |
160 | NBD_CLIENT_CONNECTING_NOWAIT; | |
161 | } | |
162 | } else { | |
f7651539 VSO |
163 | s->state = NBD_CLIENT_QUIT; |
164 | } | |
4ddb5d2f VSO |
165 | |
166 | nbd_recv_coroutines_wake(s, true); | |
a34b1e5e VSO |
167 | } |
168 | ||
46f56631 VSO |
169 | static void reconnect_delay_timer_del(BDRVNBDState *s) |
170 | { | |
171 | if (s->reconnect_delay_timer) { | |
46f56631 VSO |
172 | timer_free(s->reconnect_delay_timer); |
173 | s->reconnect_delay_timer = NULL; | |
174 | } | |
175 | } | |
176 | ||
177 | static void reconnect_delay_timer_cb(void *opaque) | |
178 | { | |
179 | BDRVNBDState *s = opaque; | |
180 | ||
fee091cd | 181 | if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) { |
46f56631 | 182 | s->state = NBD_CLIENT_CONNECTING_NOWAIT; |
4ddb5d2f | 183 | nbd_co_establish_connection_cancel(s->conn); |
46f56631 VSO |
184 | while (qemu_co_enter_next(&s->free_sema, NULL)) { |
185 | /* Resume all queued requests */ | |
186 | } | |
187 | } | |
188 | ||
189 | reconnect_delay_timer_del(s); | |
190 | } | |
191 | ||
192 | static void reconnect_delay_timer_init(BDRVNBDState *s, uint64_t expire_time_ns) | |
193 | { | |
fee091cd | 194 | if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTING_WAIT) { |
46f56631 VSO |
195 | return; |
196 | } | |
197 | ||
198 | assert(!s->reconnect_delay_timer); | |
199 | s->reconnect_delay_timer = aio_timer_new(bdrv_get_aio_context(s->bs), | |
200 | QEMU_CLOCK_REALTIME, | |
201 | SCALE_NS, | |
202 | reconnect_delay_timer_cb, s); | |
203 | timer_mod(s->reconnect_delay_timer, expire_time_ns); | |
204 | } | |
205 | ||
f7651539 VSO |
206 | static void nbd_teardown_connection(BlockDriverState *bs) |
207 | { | |
208 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; | |
209 | ||
4ddb5d2f VSO |
210 | assert(!s->in_flight); |
211 | ||
fbeb3e63 | 212 | if (s->ioc) { |
f7651539 | 213 | qio_channel_shutdown(s->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); |
4ddb5d2f VSO |
214 | yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), |
215 | nbd_yank, s->bs); | |
216 | object_unref(OBJECT(s->ioc)); | |
217 | s->ioc = NULL; | |
f7651539 | 218 | } |
fbeb3e63 | 219 | |
f7651539 | 220 | s->state = NBD_CLIENT_QUIT; |
f7651539 | 221 | } |
86f8cdf3 | 222 | |
be16b8bf VSO |
223 | static void open_timer_del(BDRVNBDState *s) |
224 | { | |
225 | if (s->open_timer) { | |
226 | timer_free(s->open_timer); | |
227 | s->open_timer = NULL; | |
228 | } | |
229 | } | |
230 | ||
231 | static void open_timer_cb(void *opaque) | |
232 | { | |
233 | BDRVNBDState *s = opaque; | |
234 | ||
235 | nbd_co_establish_connection_cancel(s->conn); | |
236 | open_timer_del(s); | |
237 | } | |
238 | ||
239 | static void open_timer_init(BDRVNBDState *s, uint64_t expire_time_ns) | |
240 | { | |
241 | assert(!s->open_timer); | |
242 | s->open_timer = aio_timer_new(bdrv_get_aio_context(s->bs), | |
243 | QEMU_CLOCK_REALTIME, | |
244 | SCALE_NS, | |
245 | open_timer_cb, s); | |
246 | timer_mod(s->open_timer, expire_time_ns); | |
247 | } | |
248 | ||
f7651539 VSO |
249 | static bool nbd_client_connecting(BDRVNBDState *s) |
250 | { | |
fee091cd LS |
251 | NBDClientState state = qatomic_load_acquire(&s->state); |
252 | return state == NBD_CLIENT_CONNECTING_WAIT || | |
253 | state == NBD_CLIENT_CONNECTING_NOWAIT; | |
f7651539 VSO |
254 | } |
255 | ||
256 | static bool nbd_client_connecting_wait(BDRVNBDState *s) | |
257 | { | |
fee091cd | 258 | return qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT; |
f7651539 VSO |
259 | } |
260 | ||
e9ba7788 VSO |
261 | /* |
262 | * Update @bs with information learned during a completed negotiation process. | |
263 | * Return failure if the server's advertised options are incompatible with the | |
264 | * client's needs. | |
265 | */ | |
266 | static int nbd_handle_updated_info(BlockDriverState *bs, Error **errp) | |
267 | { | |
268 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; | |
269 | int ret; | |
270 | ||
271 | if (s->x_dirty_bitmap) { | |
272 | if (!s->info.base_allocation) { | |
273 | error_setg(errp, "requested x-dirty-bitmap %s not found", | |
274 | s->x_dirty_bitmap); | |
275 | return -EINVAL; | |
276 | } | |
277 | if (strcmp(s->x_dirty_bitmap, "qemu:allocation-depth") == 0) { | |
278 | s->alloc_depth = true; | |
279 | } | |
280 | } | |
281 | ||
282 | if (s->info.flags & NBD_FLAG_READ_ONLY) { | |
283 | ret = bdrv_apply_auto_read_only(bs, "NBD export is read-only", errp); | |
284 | if (ret < 0) { | |
285 | return ret; | |
286 | } | |
287 | } | |
288 | ||
289 | if (s->info.flags & NBD_FLAG_SEND_FUA) { | |
290 | bs->supported_write_flags = BDRV_REQ_FUA; | |
291 | bs->supported_zero_flags |= BDRV_REQ_FUA; | |
292 | } | |
293 | ||
294 | if (s->info.flags & NBD_FLAG_SEND_WRITE_ZEROES) { | |
295 | bs->supported_zero_flags |= BDRV_REQ_MAY_UNMAP; | |
296 | if (s->info.flags & NBD_FLAG_SEND_FAST_ZERO) { | |
297 | bs->supported_zero_flags |= BDRV_REQ_NO_FALLBACK; | |
298 | } | |
299 | } | |
300 | ||
301 | trace_nbd_client_handshake_success(s->export); | |
302 | ||
303 | return 0; | |
304 | } | |
305 | ||
a71d597b VSO |
306 | int coroutine_fn nbd_co_do_establish_connection(BlockDriverState *bs, |
307 | Error **errp) | |
f7651539 | 308 | { |
51edbf53 | 309 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
fa35591b | 310 | int ret; |
4ddb5d2f | 311 | bool blocking = nbd_client_connecting_wait(s); |
f7651539 | 312 | |
51edbf53 VSO |
313 | assert(!s->ioc); |
314 | ||
4ddb5d2f | 315 | s->ioc = nbd_co_establish_connection(s->conn, &s->info, blocking, errp); |
51edbf53 VSO |
316 | if (!s->ioc) { |
317 | return -ECONNREFUSED; | |
318 | } | |
319 | ||
0b9cd6b9 LS |
320 | yank_register_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), nbd_yank, |
321 | bs); | |
322 | ||
51edbf53 VSO |
323 | ret = nbd_handle_updated_info(s->bs, NULL); |
324 | if (ret < 0) { | |
325 | /* | |
326 | * We have connected, but must fail for other reasons. | |
327 | * Send NBD_CMD_DISC as a courtesy to the server. | |
328 | */ | |
329 | NBDRequest request = { .type = NBD_CMD_DISC }; | |
330 | ||
331 | nbd_send_request(s->ioc, &request); | |
332 | ||
0b9cd6b9 LS |
333 | yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), |
334 | nbd_yank, bs); | |
51edbf53 VSO |
335 | object_unref(OBJECT(s->ioc)); |
336 | s->ioc = NULL; | |
337 | ||
338 | return ret; | |
339 | } | |
340 | ||
341 | qio_channel_set_blocking(s->ioc, false, NULL); | |
342 | qio_channel_attach_aio_context(s->ioc, bdrv_get_aio_context(bs)); | |
343 | ||
51edbf53 VSO |
344 | /* successfully connected */ |
345 | s->state = NBD_CLIENT_CONNECTED; | |
346 | qemu_co_queue_restart_all(&s->free_sema); | |
347 | ||
348 | return 0; | |
349 | } | |
350 | ||
4ddb5d2f | 351 | /* called under s->send_mutex */ |
51edbf53 VSO |
352 | static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s) |
353 | { | |
4ddb5d2f VSO |
354 | assert(nbd_client_connecting(s)); |
355 | assert(s->in_flight == 0); | |
f7651539 | 356 | |
4ddb5d2f VSO |
357 | if (nbd_client_connecting_wait(s) && s->reconnect_delay && |
358 | !s->reconnect_delay_timer) | |
359 | { | |
360 | /* | |
361 | * It's first reconnect attempt after switching to | |
362 | * NBD_CLIENT_CONNECTING_WAIT | |
363 | */ | |
364 | reconnect_delay_timer_init(s, | |
365 | qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + | |
366 | s->reconnect_delay * NANOSECONDS_PER_SECOND); | |
f7651539 VSO |
367 | } |
368 | ||
369 | /* | |
370 | * Now we are sure that nobody is accessing the channel, and no one will | |
371 | * try until we set the state to CONNECTED. | |
372 | */ | |
373 | ||
374 | /* Finalize previous connection if any */ | |
375 | if (s->ioc) { | |
8a509afd | 376 | qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc)); |
fee091cd LS |
377 | yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), |
378 | nbd_yank, s->bs); | |
f7651539 VSO |
379 | object_unref(OBJECT(s->ioc)); |
380 | s->ioc = NULL; | |
381 | } | |
382 | ||
51edbf53 | 383 | nbd_co_do_establish_connection(s->bs, NULL); |
3ce1fc16 HR |
384 | |
385 | /* | |
386 | * The reconnect attempt is done (maybe successfully, maybe not), so | |
387 | * we no longer need this timer. Delete it so it will not outlive | |
388 | * this I/O request (so draining removes all timers). | |
389 | */ | |
390 | reconnect_delay_timer_del(s); | |
f7651539 VSO |
391 | } |
392 | ||
4ddb5d2f | 393 | static coroutine_fn int nbd_receive_replies(BDRVNBDState *s, uint64_t handle) |
f7651539 | 394 | { |
4ddb5d2f VSO |
395 | int ret; |
396 | uint64_t ind = HANDLE_TO_INDEX(s, handle), ind2; | |
397 | QEMU_LOCK_GUARD(&s->receive_mutex); | |
f7651539 | 398 | |
4ddb5d2f VSO |
399 | while (true) { |
400 | if (s->reply.handle == handle) { | |
401 | /* We are done */ | |
402 | return 0; | |
f7651539 VSO |
403 | } |
404 | ||
4ddb5d2f VSO |
405 | if (!nbd_client_connected(s)) { |
406 | return -EIO; | |
407 | } | |
86f8cdf3 | 408 | |
4ddb5d2f VSO |
409 | if (s->reply.handle != 0) { |
410 | /* | |
411 | * Some other request is being handled now. It should already be | |
412 | * woken by whoever set s->reply.handle (or never wait in this | |
413 | * yield). So, we should not wake it here. | |
414 | */ | |
415 | ind2 = HANDLE_TO_INDEX(s, s->reply.handle); | |
416 | assert(!s->requests[ind2].receiving); | |
86f8cdf3 | 417 | |
4ddb5d2f VSO |
418 | s->requests[ind].receiving = true; |
419 | qemu_co_mutex_unlock(&s->receive_mutex); | |
f7651539 | 420 | |
4ddb5d2f VSO |
421 | qemu_coroutine_yield(); |
422 | /* | |
423 | * We may be woken for 3 reasons: | |
424 | * 1. From this function, executing in parallel coroutine, when our | |
425 | * handle is received. | |
426 | * 2. From nbd_channel_error(), when connection is lost. | |
427 | * 3. From nbd_co_receive_one_chunk(), when previous request is | |
428 | * finished and s->reply.handle set to 0. | |
429 | * Anyway, it's OK to lock the mutex and go to the next iteration. | |
430 | */ | |
f7651539 | 431 | |
4ddb5d2f VSO |
432 | qemu_co_mutex_lock(&s->receive_mutex); |
433 | assert(!s->requests[ind].receiving); | |
f7651539 VSO |
434 | continue; |
435 | } | |
436 | ||
4ddb5d2f | 437 | /* We are under mutex and handle is 0. We have to do the dirty work. */ |
86f8cdf3 | 438 | assert(s->reply.handle == 0); |
4ddb5d2f | 439 | ret = nbd_receive_reply(s->bs, s->ioc, &s->reply, NULL); |
86f8cdf3 | 440 | if (ret <= 0) { |
4ddb5d2f VSO |
441 | ret = ret ? ret : -EIO; |
442 | nbd_channel_error(s, ret); | |
443 | return ret; | |
86f8cdf3 | 444 | } |
4ddb5d2f | 445 | if (nbd_reply_is_structured(&s->reply) && !s->info.structured_reply) { |
a34b1e5e | 446 | nbd_channel_error(s, -EINVAL); |
4ddb5d2f | 447 | return -EINVAL; |
86f8cdf3 | 448 | } |
4ddb5d2f VSO |
449 | if (s->reply.handle == handle) { |
450 | /* We are done */ | |
451 | return 0; | |
452 | } | |
453 | ind2 = HANDLE_TO_INDEX(s, s->reply.handle); | |
1af77378 | 454 | if (ind2 >= MAX_NBD_REQUESTS || !s->requests[ind2].reply_possible) { |
4ddb5d2f VSO |
455 | nbd_channel_error(s, -EINVAL); |
456 | return -EINVAL; | |
457 | } | |
458 | nbd_recv_coroutine_wake_one(&s->requests[ind2]); | |
78c81a3f | 459 | } |
86f8cdf3 VSO |
460 | } |
461 | ||
462 | static int nbd_co_send_request(BlockDriverState *bs, | |
463 | NBDRequest *request, | |
464 | QEMUIOVector *qiov) | |
465 | { | |
611ae1d7 | 466 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
a34b1e5e | 467 | int rc, i = -1; |
86f8cdf3 VSO |
468 | |
469 | qemu_co_mutex_lock(&s->send_mutex); | |
4ddb5d2f VSO |
470 | |
471 | while (s->in_flight == MAX_NBD_REQUESTS || | |
472 | (!nbd_client_connected(s) && s->in_flight > 0)) | |
473 | { | |
86f8cdf3 VSO |
474 | qemu_co_queue_wait(&s->free_sema, &s->send_mutex); |
475 | } | |
a34b1e5e | 476 | |
4ddb5d2f VSO |
477 | if (nbd_client_connecting(s)) { |
478 | nbd_reconnect_attempt(s); | |
479 | } | |
480 | ||
91e0998f | 481 | if (!nbd_client_connected(s)) { |
a34b1e5e VSO |
482 | rc = -EIO; |
483 | goto err; | |
484 | } | |
485 | ||
86f8cdf3 VSO |
486 | s->in_flight++; |
487 | ||
488 | for (i = 0; i < MAX_NBD_REQUESTS; i++) { | |
489 | if (s->requests[i].coroutine == NULL) { | |
490 | break; | |
491 | } | |
492 | } | |
493 | ||
494 | g_assert(qemu_in_coroutine()); | |
495 | assert(i < MAX_NBD_REQUESTS); | |
496 | ||
497 | s->requests[i].coroutine = qemu_coroutine_self(); | |
498 | s->requests[i].offset = request->from; | |
499 | s->requests[i].receiving = false; | |
1af77378 | 500 | s->requests[i].reply_possible = true; |
86f8cdf3 VSO |
501 | |
502 | request->handle = INDEX_TO_HANDLE(s, i); | |
503 | ||
86f8cdf3 VSO |
504 | assert(s->ioc); |
505 | ||
506 | if (qiov) { | |
507 | qio_channel_set_cork(s->ioc, true); | |
508 | rc = nbd_send_request(s->ioc, request); | |
91e0998f | 509 | if (nbd_client_connected(s) && rc >= 0) { |
86f8cdf3 VSO |
510 | if (qio_channel_writev_all(s->ioc, qiov->iov, qiov->niov, |
511 | NULL) < 0) { | |
512 | rc = -EIO; | |
513 | } | |
514 | } else if (rc >= 0) { | |
515 | rc = -EIO; | |
516 | } | |
517 | qio_channel_set_cork(s->ioc, false); | |
518 | } else { | |
519 | rc = nbd_send_request(s->ioc, request); | |
520 | } | |
521 | ||
522 | err: | |
523 | if (rc < 0) { | |
a34b1e5e VSO |
524 | nbd_channel_error(s, rc); |
525 | if (i != -1) { | |
526 | s->requests[i].coroutine = NULL; | |
527 | s->in_flight--; | |
f7651539 VSO |
528 | qemu_co_queue_next(&s->free_sema); |
529 | } | |
86f8cdf3 VSO |
530 | } |
531 | qemu_co_mutex_unlock(&s->send_mutex); | |
532 | return rc; | |
533 | } | |
534 | ||
535 | static inline uint16_t payload_advance16(uint8_t **payload) | |
536 | { | |
537 | *payload += 2; | |
538 | return lduw_be_p(*payload - 2); | |
539 | } | |
540 | ||
541 | static inline uint32_t payload_advance32(uint8_t **payload) | |
542 | { | |
543 | *payload += 4; | |
544 | return ldl_be_p(*payload - 4); | |
545 | } | |
546 | ||
547 | static inline uint64_t payload_advance64(uint8_t **payload) | |
548 | { | |
549 | *payload += 8; | |
550 | return ldq_be_p(*payload - 8); | |
551 | } | |
552 | ||
611ae1d7 | 553 | static int nbd_parse_offset_hole_payload(BDRVNBDState *s, |
86f8cdf3 VSO |
554 | NBDStructuredReplyChunk *chunk, |
555 | uint8_t *payload, uint64_t orig_offset, | |
556 | QEMUIOVector *qiov, Error **errp) | |
557 | { | |
558 | uint64_t offset; | |
559 | uint32_t hole_size; | |
560 | ||
561 | if (chunk->length != sizeof(offset) + sizeof(hole_size)) { | |
562 | error_setg(errp, "Protocol error: invalid payload for " | |
563 | "NBD_REPLY_TYPE_OFFSET_HOLE"); | |
564 | return -EINVAL; | |
565 | } | |
566 | ||
567 | offset = payload_advance64(&payload); | |
568 | hole_size = payload_advance32(&payload); | |
569 | ||
570 | if (!hole_size || offset < orig_offset || hole_size > qiov->size || | |
571 | offset > orig_offset + qiov->size - hole_size) { | |
572 | error_setg(errp, "Protocol error: server sent chunk exceeding requested" | |
573 | " region"); | |
574 | return -EINVAL; | |
575 | } | |
611ae1d7 VSO |
576 | if (s->info.min_block && |
577 | !QEMU_IS_ALIGNED(hole_size, s->info.min_block)) { | |
86f8cdf3 VSO |
578 | trace_nbd_structured_read_compliance("hole"); |
579 | } | |
580 | ||
581 | qemu_iovec_memset(qiov, offset - orig_offset, 0, hole_size); | |
582 | ||
583 | return 0; | |
584 | } | |
585 | ||
586 | /* | |
587 | * nbd_parse_blockstatus_payload | |
588 | * Based on our request, we expect only one extent in reply, for the | |
589 | * base:allocation context. | |
590 | */ | |
611ae1d7 | 591 | static int nbd_parse_blockstatus_payload(BDRVNBDState *s, |
86f8cdf3 VSO |
592 | NBDStructuredReplyChunk *chunk, |
593 | uint8_t *payload, uint64_t orig_length, | |
594 | NBDExtent *extent, Error **errp) | |
595 | { | |
596 | uint32_t context_id; | |
597 | ||
598 | /* The server succeeded, so it must have sent [at least] one extent */ | |
599 | if (chunk->length < sizeof(context_id) + sizeof(*extent)) { | |
600 | error_setg(errp, "Protocol error: invalid payload for " | |
601 | "NBD_REPLY_TYPE_BLOCK_STATUS"); | |
602 | return -EINVAL; | |
603 | } | |
604 | ||
605 | context_id = payload_advance32(&payload); | |
611ae1d7 | 606 | if (s->info.context_id != context_id) { |
86f8cdf3 VSO |
607 | error_setg(errp, "Protocol error: unexpected context id %d for " |
608 | "NBD_REPLY_TYPE_BLOCK_STATUS, when negotiated context " | |
609 | "id is %d", context_id, | |
611ae1d7 | 610 | s->info.context_id); |
86f8cdf3 VSO |
611 | return -EINVAL; |
612 | } | |
613 | ||
614 | extent->length = payload_advance32(&payload); | |
615 | extent->flags = payload_advance32(&payload); | |
616 | ||
617 | if (extent->length == 0) { | |
618 | error_setg(errp, "Protocol error: server sent status chunk with " | |
619 | "zero length"); | |
620 | return -EINVAL; | |
621 | } | |
622 | ||
623 | /* | |
624 | * A server sending unaligned block status is in violation of the | |
625 | * protocol, but as qemu-nbd 3.1 is such a server (at least for | |
626 | * POSIX files that are not a multiple of 512 bytes, since qemu | |
627 | * rounds files up to 512-byte multiples but lseek(SEEK_HOLE) | |
628 | * still sees an implicit hole beyond the real EOF), it's nicer to | |
629 | * work around the misbehaving server. If the request included | |
630 | * more than the final unaligned block, truncate it back to an | |
631 | * aligned result; if the request was only the final block, round | |
632 | * up to the full block and change the status to fully-allocated | |
633 | * (always a safe status, even if it loses information). | |
634 | */ | |
611ae1d7 VSO |
635 | if (s->info.min_block && !QEMU_IS_ALIGNED(extent->length, |
636 | s->info.min_block)) { | |
86f8cdf3 | 637 | trace_nbd_parse_blockstatus_compliance("extent length is unaligned"); |
611ae1d7 | 638 | if (extent->length > s->info.min_block) { |
86f8cdf3 | 639 | extent->length = QEMU_ALIGN_DOWN(extent->length, |
611ae1d7 | 640 | s->info.min_block); |
86f8cdf3 | 641 | } else { |
611ae1d7 | 642 | extent->length = s->info.min_block; |
86f8cdf3 VSO |
643 | extent->flags = 0; |
644 | } | |
645 | } | |
646 | ||
647 | /* | |
648 | * We used NBD_CMD_FLAG_REQ_ONE, so the server should not have | |
649 | * sent us any more than one extent, nor should it have included | |
650 | * status beyond our request in that extent. However, it's easy | |
651 | * enough to ignore the server's noncompliance without killing the | |
652 | * connection; just ignore trailing extents, and clamp things to | |
653 | * the length of our request. | |
654 | */ | |
655 | if (chunk->length > sizeof(context_id) + sizeof(*extent)) { | |
656 | trace_nbd_parse_blockstatus_compliance("more than one extent"); | |
657 | } | |
658 | if (extent->length > orig_length) { | |
659 | extent->length = orig_length; | |
660 | trace_nbd_parse_blockstatus_compliance("extent length too large"); | |
661 | } | |
662 | ||
dbc7b014 EB |
663 | /* |
664 | * HACK: if we are using x-dirty-bitmaps to access | |
665 | * qemu:allocation-depth, treat all depths > 2 the same as 2, | |
666 | * since nbd_client_co_block_status is only expecting the low two | |
667 | * bits to be set. | |
668 | */ | |
669 | if (s->alloc_depth && extent->flags > 2) { | |
670 | extent->flags = 2; | |
671 | } | |
672 | ||
86f8cdf3 VSO |
673 | return 0; |
674 | } | |
675 | ||
676 | /* | |
677 | * nbd_parse_error_payload | |
678 | * on success @errp contains message describing nbd error reply | |
679 | */ | |
680 | static int nbd_parse_error_payload(NBDStructuredReplyChunk *chunk, | |
681 | uint8_t *payload, int *request_ret, | |
682 | Error **errp) | |
683 | { | |
684 | uint32_t error; | |
685 | uint16_t message_size; | |
686 | ||
687 | assert(chunk->type & (1 << 15)); | |
688 | ||
689 | if (chunk->length < sizeof(error) + sizeof(message_size)) { | |
690 | error_setg(errp, | |
691 | "Protocol error: invalid payload for structured error"); | |
692 | return -EINVAL; | |
693 | } | |
694 | ||
695 | error = nbd_errno_to_system_errno(payload_advance32(&payload)); | |
696 | if (error == 0) { | |
697 | error_setg(errp, "Protocol error: server sent structured error chunk " | |
698 | "with error = 0"); | |
699 | return -EINVAL; | |
700 | } | |
701 | ||
702 | *request_ret = -error; | |
703 | message_size = payload_advance16(&payload); | |
704 | ||
705 | if (message_size > chunk->length - sizeof(error) - sizeof(message_size)) { | |
706 | error_setg(errp, "Protocol error: server sent structured error chunk " | |
707 | "with incorrect message size"); | |
708 | return -EINVAL; | |
709 | } | |
710 | ||
711 | /* TODO: Add a trace point to mention the server complaint */ | |
712 | ||
713 | /* TODO handle ERROR_OFFSET */ | |
714 | ||
715 | return 0; | |
716 | } | |
717 | ||
611ae1d7 | 718 | static int nbd_co_receive_offset_data_payload(BDRVNBDState *s, |
86f8cdf3 VSO |
719 | uint64_t orig_offset, |
720 | QEMUIOVector *qiov, Error **errp) | |
721 | { | |
722 | QEMUIOVector sub_qiov; | |
723 | uint64_t offset; | |
724 | size_t data_size; | |
725 | int ret; | |
726 | NBDStructuredReplyChunk *chunk = &s->reply.structured; | |
727 | ||
728 | assert(nbd_reply_is_structured(&s->reply)); | |
729 | ||
730 | /* The NBD spec requires at least one byte of payload */ | |
731 | if (chunk->length <= sizeof(offset)) { | |
732 | error_setg(errp, "Protocol error: invalid payload for " | |
733 | "NBD_REPLY_TYPE_OFFSET_DATA"); | |
734 | return -EINVAL; | |
735 | } | |
736 | ||
737 | if (nbd_read64(s->ioc, &offset, "OFFSET_DATA offset", errp) < 0) { | |
738 | return -EIO; | |
739 | } | |
740 | ||
741 | data_size = chunk->length - sizeof(offset); | |
742 | assert(data_size); | |
743 | if (offset < orig_offset || data_size > qiov->size || | |
744 | offset > orig_offset + qiov->size - data_size) { | |
745 | error_setg(errp, "Protocol error: server sent chunk exceeding requested" | |
746 | " region"); | |
747 | return -EINVAL; | |
748 | } | |
749 | if (s->info.min_block && !QEMU_IS_ALIGNED(data_size, s->info.min_block)) { | |
750 | trace_nbd_structured_read_compliance("data"); | |
751 | } | |
752 | ||
753 | qemu_iovec_init(&sub_qiov, qiov->niov); | |
754 | qemu_iovec_concat(&sub_qiov, qiov, offset - orig_offset, data_size); | |
755 | ret = qio_channel_readv_all(s->ioc, sub_qiov.iov, sub_qiov.niov, errp); | |
756 | qemu_iovec_destroy(&sub_qiov); | |
757 | ||
758 | return ret < 0 ? -EIO : 0; | |
759 | } | |
760 | ||
761 | #define NBD_MAX_MALLOC_PAYLOAD 1000 | |
762 | static coroutine_fn int nbd_co_receive_structured_payload( | |
611ae1d7 | 763 | BDRVNBDState *s, void **payload, Error **errp) |
86f8cdf3 VSO |
764 | { |
765 | int ret; | |
766 | uint32_t len; | |
767 | ||
768 | assert(nbd_reply_is_structured(&s->reply)); | |
769 | ||
770 | len = s->reply.structured.length; | |
771 | ||
772 | if (len == 0) { | |
773 | return 0; | |
774 | } | |
775 | ||
776 | if (payload == NULL) { | |
777 | error_setg(errp, "Unexpected structured payload"); | |
778 | return -EINVAL; | |
779 | } | |
780 | ||
781 | if (len > NBD_MAX_MALLOC_PAYLOAD) { | |
782 | error_setg(errp, "Payload too large"); | |
783 | return -EINVAL; | |
784 | } | |
785 | ||
786 | *payload = g_new(char, len); | |
787 | ret = nbd_read(s->ioc, *payload, len, "structured payload", errp); | |
788 | if (ret < 0) { | |
789 | g_free(*payload); | |
790 | *payload = NULL; | |
791 | return ret; | |
792 | } | |
793 | ||
794 | return 0; | |
795 | } | |
796 | ||
797 | /* | |
798 | * nbd_co_do_receive_one_chunk | |
799 | * for simple reply: | |
800 | * set request_ret to received reply error | |
801 | * if qiov is not NULL: read payload to @qiov | |
802 | * for structured reply chunk: | |
803 | * if error chunk: read payload, set @request_ret, do not set @payload | |
804 | * else if offset_data chunk: read payload data to @qiov, do not set @payload | |
805 | * else: read payload to @payload | |
806 | * | |
807 | * If function fails, @errp contains corresponding error message, and the | |
808 | * connection with the server is suspect. If it returns 0, then the | |
809 | * transaction succeeded (although @request_ret may be a negative errno | |
810 | * corresponding to the server's error reply), and errp is unchanged. | |
811 | */ | |
812 | static coroutine_fn int nbd_co_do_receive_one_chunk( | |
611ae1d7 | 813 | BDRVNBDState *s, uint64_t handle, bool only_structured, |
86f8cdf3 VSO |
814 | int *request_ret, QEMUIOVector *qiov, void **payload, Error **errp) |
815 | { | |
816 | int ret; | |
817 | int i = HANDLE_TO_INDEX(s, handle); | |
818 | void *local_payload = NULL; | |
819 | NBDStructuredReplyChunk *chunk; | |
820 | ||
821 | if (payload) { | |
822 | *payload = NULL; | |
823 | } | |
824 | *request_ret = 0; | |
825 | ||
4ddb5d2f | 826 | nbd_receive_replies(s, handle); |
91e0998f | 827 | if (!nbd_client_connected(s)) { |
86f8cdf3 VSO |
828 | error_setg(errp, "Connection closed"); |
829 | return -EIO; | |
830 | } | |
831 | assert(s->ioc); | |
832 | ||
833 | assert(s->reply.handle == handle); | |
834 | ||
835 | if (nbd_reply_is_simple(&s->reply)) { | |
836 | if (only_structured) { | |
837 | error_setg(errp, "Protocol error: simple reply when structured " | |
838 | "reply chunk was expected"); | |
839 | return -EINVAL; | |
840 | } | |
841 | ||
842 | *request_ret = -nbd_errno_to_system_errno(s->reply.simple.error); | |
843 | if (*request_ret < 0 || !qiov) { | |
844 | return 0; | |
845 | } | |
846 | ||
847 | return qio_channel_readv_all(s->ioc, qiov->iov, qiov->niov, | |
848 | errp) < 0 ? -EIO : 0; | |
849 | } | |
850 | ||
851 | /* handle structured reply chunk */ | |
852 | assert(s->info.structured_reply); | |
853 | chunk = &s->reply.structured; | |
854 | ||
855 | if (chunk->type == NBD_REPLY_TYPE_NONE) { | |
856 | if (!(chunk->flags & NBD_REPLY_FLAG_DONE)) { | |
857 | error_setg(errp, "Protocol error: NBD_REPLY_TYPE_NONE chunk without" | |
858 | " NBD_REPLY_FLAG_DONE flag set"); | |
859 | return -EINVAL; | |
860 | } | |
861 | if (chunk->length) { | |
862 | error_setg(errp, "Protocol error: NBD_REPLY_TYPE_NONE chunk with" | |
863 | " nonzero length"); | |
864 | return -EINVAL; | |
865 | } | |
866 | return 0; | |
867 | } | |
868 | ||
869 | if (chunk->type == NBD_REPLY_TYPE_OFFSET_DATA) { | |
870 | if (!qiov) { | |
871 | error_setg(errp, "Unexpected NBD_REPLY_TYPE_OFFSET_DATA chunk"); | |
872 | return -EINVAL; | |
873 | } | |
874 | ||
875 | return nbd_co_receive_offset_data_payload(s, s->requests[i].offset, | |
876 | qiov, errp); | |
877 | } | |
878 | ||
879 | if (nbd_reply_type_is_error(chunk->type)) { | |
880 | payload = &local_payload; | |
881 | } | |
882 | ||
883 | ret = nbd_co_receive_structured_payload(s, payload, errp); | |
884 | if (ret < 0) { | |
885 | return ret; | |
886 | } | |
887 | ||
888 | if (nbd_reply_type_is_error(chunk->type)) { | |
889 | ret = nbd_parse_error_payload(chunk, local_payload, request_ret, errp); | |
890 | g_free(local_payload); | |
891 | return ret; | |
892 | } | |
893 | ||
894 | return 0; | |
895 | } | |
896 | ||
897 | /* | |
898 | * nbd_co_receive_one_chunk | |
899 | * Read reply, wake up connection_co and set s->quit if needed. | |
900 | * Return value is a fatal error code or normal nbd reply error code | |
901 | */ | |
902 | static coroutine_fn int nbd_co_receive_one_chunk( | |
611ae1d7 | 903 | BDRVNBDState *s, uint64_t handle, bool only_structured, |
86f8cdf3 VSO |
904 | int *request_ret, QEMUIOVector *qiov, NBDReply *reply, void **payload, |
905 | Error **errp) | |
906 | { | |
907 | int ret = nbd_co_do_receive_one_chunk(s, handle, only_structured, | |
908 | request_ret, qiov, payload, errp); | |
909 | ||
910 | if (ret < 0) { | |
5cf42b1c | 911 | memset(reply, 0, sizeof(*reply)); |
a34b1e5e | 912 | nbd_channel_error(s, ret); |
86f8cdf3 VSO |
913 | } else { |
914 | /* For assert at loop start in nbd_connection_entry */ | |
5cf42b1c | 915 | *reply = s->reply; |
86f8cdf3 | 916 | } |
f7651539 | 917 | s->reply.handle = 0; |
86f8cdf3 | 918 | |
4ddb5d2f | 919 | nbd_recv_coroutines_wake(s, false); |
86f8cdf3 VSO |
920 | |
921 | return ret; | |
922 | } | |
923 | ||
924 | typedef struct NBDReplyChunkIter { | |
925 | int ret; | |
926 | int request_ret; | |
927 | Error *err; | |
928 | bool done, only_structured; | |
929 | } NBDReplyChunkIter; | |
930 | ||
931 | static void nbd_iter_channel_error(NBDReplyChunkIter *iter, | |
932 | int ret, Error **local_err) | |
933 | { | |
d9366135 | 934 | assert(local_err && *local_err); |
86f8cdf3 VSO |
935 | assert(ret < 0); |
936 | ||
937 | if (!iter->ret) { | |
938 | iter->ret = ret; | |
939 | error_propagate(&iter->err, *local_err); | |
940 | } else { | |
941 | error_free(*local_err); | |
942 | } | |
943 | ||
944 | *local_err = NULL; | |
945 | } | |
946 | ||
947 | static void nbd_iter_request_error(NBDReplyChunkIter *iter, int ret) | |
948 | { | |
949 | assert(ret < 0); | |
950 | ||
951 | if (!iter->request_ret) { | |
952 | iter->request_ret = ret; | |
953 | } | |
954 | } | |
955 | ||
956 | /* | |
957 | * NBD_FOREACH_REPLY_CHUNK | |
958 | * The pointer stored in @payload requires g_free() to free it. | |
959 | */ | |
960 | #define NBD_FOREACH_REPLY_CHUNK(s, iter, handle, structured, \ | |
961 | qiov, reply, payload) \ | |
962 | for (iter = (NBDReplyChunkIter) { .only_structured = structured }; \ | |
963 | nbd_reply_chunk_iter_receive(s, &iter, handle, qiov, reply, payload);) | |
964 | ||
965 | /* | |
966 | * nbd_reply_chunk_iter_receive | |
967 | * The pointer stored in @payload requires g_free() to free it. | |
968 | */ | |
611ae1d7 | 969 | static bool nbd_reply_chunk_iter_receive(BDRVNBDState *s, |
86f8cdf3 VSO |
970 | NBDReplyChunkIter *iter, |
971 | uint64_t handle, | |
972 | QEMUIOVector *qiov, NBDReply *reply, | |
973 | void **payload) | |
974 | { | |
975 | int ret, request_ret; | |
976 | NBDReply local_reply; | |
977 | NBDStructuredReplyChunk *chunk; | |
978 | Error *local_err = NULL; | |
91e0998f | 979 | if (!nbd_client_connected(s)) { |
86f8cdf3 VSO |
980 | error_setg(&local_err, "Connection closed"); |
981 | nbd_iter_channel_error(iter, -EIO, &local_err); | |
982 | goto break_loop; | |
983 | } | |
984 | ||
985 | if (iter->done) { | |
986 | /* Previous iteration was last. */ | |
987 | goto break_loop; | |
988 | } | |
989 | ||
990 | if (reply == NULL) { | |
991 | reply = &local_reply; | |
992 | } | |
993 | ||
994 | ret = nbd_co_receive_one_chunk(s, handle, iter->only_structured, | |
995 | &request_ret, qiov, reply, payload, | |
996 | &local_err); | |
997 | if (ret < 0) { | |
998 | nbd_iter_channel_error(iter, ret, &local_err); | |
999 | } else if (request_ret < 0) { | |
1000 | nbd_iter_request_error(iter, request_ret); | |
1001 | } | |
1002 | ||
1003 | /* Do not execute the body of NBD_FOREACH_REPLY_CHUNK for simple reply. */ | |
91e0998f | 1004 | if (nbd_reply_is_simple(reply) || !nbd_client_connected(s)) { |
86f8cdf3 VSO |
1005 | goto break_loop; |
1006 | } | |
1007 | ||
1008 | chunk = &reply->structured; | |
1009 | iter->only_structured = true; | |
1010 | ||
1011 | if (chunk->type == NBD_REPLY_TYPE_NONE) { | |
1012 | /* NBD_REPLY_FLAG_DONE is already checked in nbd_co_receive_one_chunk */ | |
1013 | assert(chunk->flags & NBD_REPLY_FLAG_DONE); | |
1014 | goto break_loop; | |
1015 | } | |
1016 | ||
1017 | if (chunk->flags & NBD_REPLY_FLAG_DONE) { | |
1018 | /* This iteration is last. */ | |
1019 | iter->done = true; | |
1020 | } | |
1021 | ||
1022 | /* Execute the loop body */ | |
1023 | return true; | |
1024 | ||
1025 | break_loop: | |
1026 | s->requests[HANDLE_TO_INDEX(s, handle)].coroutine = NULL; | |
1027 | ||
1028 | qemu_co_mutex_lock(&s->send_mutex); | |
1029 | s->in_flight--; | |
4ddb5d2f | 1030 | qemu_co_queue_next(&s->free_sema); |
86f8cdf3 VSO |
1031 | qemu_co_mutex_unlock(&s->send_mutex); |
1032 | ||
1033 | return false; | |
1034 | } | |
1035 | ||
611ae1d7 | 1036 | static int nbd_co_receive_return_code(BDRVNBDState *s, uint64_t handle, |
86f8cdf3 VSO |
1037 | int *request_ret, Error **errp) |
1038 | { | |
1039 | NBDReplyChunkIter iter; | |
1040 | ||
1041 | NBD_FOREACH_REPLY_CHUNK(s, iter, handle, false, NULL, NULL, NULL) { | |
1042 | /* nbd_reply_chunk_iter_receive does all the work */ | |
1043 | } | |
1044 | ||
1045 | error_propagate(errp, iter.err); | |
1046 | *request_ret = iter.request_ret; | |
1047 | return iter.ret; | |
1048 | } | |
1049 | ||
611ae1d7 | 1050 | static int nbd_co_receive_cmdread_reply(BDRVNBDState *s, uint64_t handle, |
86f8cdf3 VSO |
1051 | uint64_t offset, QEMUIOVector *qiov, |
1052 | int *request_ret, Error **errp) | |
1053 | { | |
1054 | NBDReplyChunkIter iter; | |
1055 | NBDReply reply; | |
1056 | void *payload = NULL; | |
1057 | Error *local_err = NULL; | |
1058 | ||
1059 | NBD_FOREACH_REPLY_CHUNK(s, iter, handle, s->info.structured_reply, | |
1060 | qiov, &reply, &payload) | |
1061 | { | |
1062 | int ret; | |
1063 | NBDStructuredReplyChunk *chunk = &reply.structured; | |
1064 | ||
1065 | assert(nbd_reply_is_structured(&reply)); | |
1066 | ||
1067 | switch (chunk->type) { | |
1068 | case NBD_REPLY_TYPE_OFFSET_DATA: | |
1069 | /* | |
1070 | * special cased in nbd_co_receive_one_chunk, data is already | |
1071 | * in qiov | |
1072 | */ | |
1073 | break; | |
1074 | case NBD_REPLY_TYPE_OFFSET_HOLE: | |
1075 | ret = nbd_parse_offset_hole_payload(s, &reply.structured, payload, | |
1076 | offset, qiov, &local_err); | |
1077 | if (ret < 0) { | |
a34b1e5e | 1078 | nbd_channel_error(s, ret); |
86f8cdf3 VSO |
1079 | nbd_iter_channel_error(&iter, ret, &local_err); |
1080 | } | |
1081 | break; | |
1082 | default: | |
1083 | if (!nbd_reply_type_is_error(chunk->type)) { | |
1084 | /* not allowed reply type */ | |
a34b1e5e | 1085 | nbd_channel_error(s, -EINVAL); |
86f8cdf3 VSO |
1086 | error_setg(&local_err, |
1087 | "Unexpected reply type: %d (%s) for CMD_READ", | |
1088 | chunk->type, nbd_reply_type_lookup(chunk->type)); | |
1089 | nbd_iter_channel_error(&iter, -EINVAL, &local_err); | |
1090 | } | |
1091 | } | |
1092 | ||
1093 | g_free(payload); | |
1094 | payload = NULL; | |
1095 | } | |
1096 | ||
1097 | error_propagate(errp, iter.err); | |
1098 | *request_ret = iter.request_ret; | |
1099 | return iter.ret; | |
1100 | } | |
1101 | ||
611ae1d7 | 1102 | static int nbd_co_receive_blockstatus_reply(BDRVNBDState *s, |
86f8cdf3 VSO |
1103 | uint64_t handle, uint64_t length, |
1104 | NBDExtent *extent, | |
1105 | int *request_ret, Error **errp) | |
1106 | { | |
1107 | NBDReplyChunkIter iter; | |
1108 | NBDReply reply; | |
1109 | void *payload = NULL; | |
1110 | Error *local_err = NULL; | |
1111 | bool received = false; | |
1112 | ||
1113 | assert(!extent->length); | |
1114 | NBD_FOREACH_REPLY_CHUNK(s, iter, handle, false, NULL, &reply, &payload) { | |
1115 | int ret; | |
1116 | NBDStructuredReplyChunk *chunk = &reply.structured; | |
1117 | ||
1118 | assert(nbd_reply_is_structured(&reply)); | |
1119 | ||
1120 | switch (chunk->type) { | |
1121 | case NBD_REPLY_TYPE_BLOCK_STATUS: | |
1122 | if (received) { | |
a34b1e5e | 1123 | nbd_channel_error(s, -EINVAL); |
86f8cdf3 VSO |
1124 | error_setg(&local_err, "Several BLOCK_STATUS chunks in reply"); |
1125 | nbd_iter_channel_error(&iter, -EINVAL, &local_err); | |
1126 | } | |
1127 | received = true; | |
1128 | ||
1129 | ret = nbd_parse_blockstatus_payload(s, &reply.structured, | |
1130 | payload, length, extent, | |
1131 | &local_err); | |
1132 | if (ret < 0) { | |
a34b1e5e | 1133 | nbd_channel_error(s, ret); |
86f8cdf3 VSO |
1134 | nbd_iter_channel_error(&iter, ret, &local_err); |
1135 | } | |
1136 | break; | |
1137 | default: | |
1138 | if (!nbd_reply_type_is_error(chunk->type)) { | |
a34b1e5e | 1139 | nbd_channel_error(s, -EINVAL); |
86f8cdf3 VSO |
1140 | error_setg(&local_err, |
1141 | "Unexpected reply type: %d (%s) " | |
1142 | "for CMD_BLOCK_STATUS", | |
1143 | chunk->type, nbd_reply_type_lookup(chunk->type)); | |
1144 | nbd_iter_channel_error(&iter, -EINVAL, &local_err); | |
1145 | } | |
1146 | } | |
1147 | ||
1148 | g_free(payload); | |
1149 | payload = NULL; | |
1150 | } | |
1151 | ||
1152 | if (!extent->length && !iter.request_ret) { | |
1153 | error_setg(&local_err, "Server did not reply with any status extents"); | |
1154 | nbd_iter_channel_error(&iter, -EIO, &local_err); | |
1155 | } | |
1156 | ||
1157 | error_propagate(errp, iter.err); | |
1158 | *request_ret = iter.request_ret; | |
1159 | return iter.ret; | |
1160 | } | |
1161 | ||
1162 | static int nbd_co_request(BlockDriverState *bs, NBDRequest *request, | |
1163 | QEMUIOVector *write_qiov) | |
1164 | { | |
1165 | int ret, request_ret; | |
1166 | Error *local_err = NULL; | |
611ae1d7 | 1167 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1168 | |
1169 | assert(request->type != NBD_CMD_READ); | |
1170 | if (write_qiov) { | |
1171 | assert(request->type == NBD_CMD_WRITE); | |
1172 | assert(request->len == iov_size(write_qiov->iov, write_qiov->niov)); | |
1173 | } else { | |
1174 | assert(request->type != NBD_CMD_WRITE); | |
1175 | } | |
86f8cdf3 | 1176 | |
f7651539 VSO |
1177 | do { |
1178 | ret = nbd_co_send_request(bs, request, write_qiov); | |
1179 | if (ret < 0) { | |
1180 | continue; | |
1181 | } | |
1182 | ||
1183 | ret = nbd_co_receive_return_code(s, request->handle, | |
1184 | &request_ret, &local_err); | |
1185 | if (local_err) { | |
1186 | trace_nbd_co_request_fail(request->from, request->len, | |
1187 | request->handle, request->flags, | |
1188 | request->type, | |
1189 | nbd_cmd_lookup(request->type), | |
1190 | ret, error_get_pretty(local_err)); | |
1191 | error_free(local_err); | |
1192 | local_err = NULL; | |
1193 | } | |
1194 | } while (ret < 0 && nbd_client_connecting_wait(s)); | |
1195 | ||
86f8cdf3 VSO |
1196 | return ret ? ret : request_ret; |
1197 | } | |
1198 | ||
f7ef38dd VSO |
1199 | static int nbd_client_co_preadv(BlockDriverState *bs, int64_t offset, |
1200 | int64_t bytes, QEMUIOVector *qiov, | |
1201 | BdrvRequestFlags flags) | |
86f8cdf3 VSO |
1202 | { |
1203 | int ret, request_ret; | |
1204 | Error *local_err = NULL; | |
611ae1d7 | 1205 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1206 | NBDRequest request = { |
1207 | .type = NBD_CMD_READ, | |
1208 | .from = offset, | |
1209 | .len = bytes, | |
1210 | }; | |
1211 | ||
1212 | assert(bytes <= NBD_MAX_BUFFER_SIZE); | |
1213 | assert(!flags); | |
1214 | ||
1215 | if (!bytes) { | |
1216 | return 0; | |
1217 | } | |
1218 | /* | |
1219 | * Work around the fact that the block layer doesn't do | |
1220 | * byte-accurate sizing yet - if the read exceeds the server's | |
1221 | * advertised size because the block layer rounded size up, then | |
1222 | * truncate the request to the server and tail-pad with zero. | |
1223 | */ | |
611ae1d7 | 1224 | if (offset >= s->info.size) { |
86f8cdf3 VSO |
1225 | assert(bytes < BDRV_SECTOR_SIZE); |
1226 | qemu_iovec_memset(qiov, 0, 0, bytes); | |
1227 | return 0; | |
1228 | } | |
611ae1d7 VSO |
1229 | if (offset + bytes > s->info.size) { |
1230 | uint64_t slop = offset + bytes - s->info.size; | |
86f8cdf3 VSO |
1231 | |
1232 | assert(slop < BDRV_SECTOR_SIZE); | |
1233 | qemu_iovec_memset(qiov, bytes - slop, 0, slop); | |
1234 | request.len -= slop; | |
1235 | } | |
1236 | ||
f7651539 VSO |
1237 | do { |
1238 | ret = nbd_co_send_request(bs, &request, NULL); | |
1239 | if (ret < 0) { | |
1240 | continue; | |
1241 | } | |
1242 | ||
1243 | ret = nbd_co_receive_cmdread_reply(s, request.handle, offset, qiov, | |
1244 | &request_ret, &local_err); | |
1245 | if (local_err) { | |
1246 | trace_nbd_co_request_fail(request.from, request.len, request.handle, | |
1247 | request.flags, request.type, | |
1248 | nbd_cmd_lookup(request.type), | |
1249 | ret, error_get_pretty(local_err)); | |
1250 | error_free(local_err); | |
1251 | local_err = NULL; | |
1252 | } | |
1253 | } while (ret < 0 && nbd_client_connecting_wait(s)); | |
86f8cdf3 | 1254 | |
86f8cdf3 VSO |
1255 | return ret ? ret : request_ret; |
1256 | } | |
1257 | ||
e75abeda VSO |
1258 | static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, |
1259 | int64_t bytes, QEMUIOVector *qiov, | |
1260 | BdrvRequestFlags flags) | |
86f8cdf3 | 1261 | { |
611ae1d7 | 1262 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1263 | NBDRequest request = { |
1264 | .type = NBD_CMD_WRITE, | |
1265 | .from = offset, | |
1266 | .len = bytes, | |
1267 | }; | |
1268 | ||
611ae1d7 | 1269 | assert(!(s->info.flags & NBD_FLAG_READ_ONLY)); |
86f8cdf3 | 1270 | if (flags & BDRV_REQ_FUA) { |
611ae1d7 | 1271 | assert(s->info.flags & NBD_FLAG_SEND_FUA); |
86f8cdf3 VSO |
1272 | request.flags |= NBD_CMD_FLAG_FUA; |
1273 | } | |
1274 | ||
1275 | assert(bytes <= NBD_MAX_BUFFER_SIZE); | |
1276 | ||
1277 | if (!bytes) { | |
1278 | return 0; | |
1279 | } | |
1280 | return nbd_co_request(bs, &request, qiov); | |
1281 | } | |
1282 | ||
1283 | static int nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, | |
f34b2bcf | 1284 | int64_t bytes, BdrvRequestFlags flags) |
86f8cdf3 | 1285 | { |
611ae1d7 | 1286 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1287 | NBDRequest request = { |
1288 | .type = NBD_CMD_WRITE_ZEROES, | |
1289 | .from = offset, | |
f34b2bcf | 1290 | .len = bytes, /* .len is uint32_t actually */ |
86f8cdf3 VSO |
1291 | }; |
1292 | ||
f34b2bcf VSO |
1293 | assert(bytes <= UINT32_MAX); /* rely on max_pwrite_zeroes */ |
1294 | ||
611ae1d7 VSO |
1295 | assert(!(s->info.flags & NBD_FLAG_READ_ONLY)); |
1296 | if (!(s->info.flags & NBD_FLAG_SEND_WRITE_ZEROES)) { | |
86f8cdf3 VSO |
1297 | return -ENOTSUP; |
1298 | } | |
1299 | ||
1300 | if (flags & BDRV_REQ_FUA) { | |
611ae1d7 | 1301 | assert(s->info.flags & NBD_FLAG_SEND_FUA); |
86f8cdf3 VSO |
1302 | request.flags |= NBD_CMD_FLAG_FUA; |
1303 | } | |
1304 | if (!(flags & BDRV_REQ_MAY_UNMAP)) { | |
1305 | request.flags |= NBD_CMD_FLAG_NO_HOLE; | |
1306 | } | |
f061656c EB |
1307 | if (flags & BDRV_REQ_NO_FALLBACK) { |
1308 | assert(s->info.flags & NBD_FLAG_SEND_FAST_ZERO); | |
1309 | request.flags |= NBD_CMD_FLAG_FAST_ZERO; | |
1310 | } | |
86f8cdf3 VSO |
1311 | |
1312 | if (!bytes) { | |
1313 | return 0; | |
1314 | } | |
1315 | return nbd_co_request(bs, &request, NULL); | |
1316 | } | |
1317 | ||
1318 | static int nbd_client_co_flush(BlockDriverState *bs) | |
1319 | { | |
611ae1d7 | 1320 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1321 | NBDRequest request = { .type = NBD_CMD_FLUSH }; |
1322 | ||
611ae1d7 | 1323 | if (!(s->info.flags & NBD_FLAG_SEND_FLUSH)) { |
86f8cdf3 VSO |
1324 | return 0; |
1325 | } | |
1326 | ||
1327 | request.from = 0; | |
1328 | request.len = 0; | |
1329 | ||
1330 | return nbd_co_request(bs, &request, NULL); | |
1331 | } | |
1332 | ||
1333 | static int nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset, | |
0c802287 | 1334 | int64_t bytes) |
86f8cdf3 | 1335 | { |
611ae1d7 | 1336 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1337 | NBDRequest request = { |
1338 | .type = NBD_CMD_TRIM, | |
1339 | .from = offset, | |
0c802287 | 1340 | .len = bytes, /* len is uint32_t */ |
86f8cdf3 VSO |
1341 | }; |
1342 | ||
0c802287 VSO |
1343 | assert(bytes <= UINT32_MAX); /* rely on max_pdiscard */ |
1344 | ||
611ae1d7 VSO |
1345 | assert(!(s->info.flags & NBD_FLAG_READ_ONLY)); |
1346 | if (!(s->info.flags & NBD_FLAG_SEND_TRIM) || !bytes) { | |
86f8cdf3 VSO |
1347 | return 0; |
1348 | } | |
1349 | ||
1350 | return nbd_co_request(bs, &request, NULL); | |
1351 | } | |
1352 | ||
1353 | static int coroutine_fn nbd_client_co_block_status( | |
1354 | BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes, | |
1355 | int64_t *pnum, int64_t *map, BlockDriverState **file) | |
1356 | { | |
1357 | int ret, request_ret; | |
1358 | NBDExtent extent = { 0 }; | |
611ae1d7 | 1359 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1360 | Error *local_err = NULL; |
1361 | ||
1362 | NBDRequest request = { | |
1363 | .type = NBD_CMD_BLOCK_STATUS, | |
1364 | .from = offset, | |
6bf792b4 | 1365 | .len = MIN(QEMU_ALIGN_DOWN(INT_MAX, bs->bl.request_alignment), |
611ae1d7 | 1366 | MIN(bytes, s->info.size - offset)), |
86f8cdf3 VSO |
1367 | .flags = NBD_CMD_FLAG_REQ_ONE, |
1368 | }; | |
1369 | ||
611ae1d7 | 1370 | if (!s->info.base_allocation) { |
86f8cdf3 VSO |
1371 | *pnum = bytes; |
1372 | *map = offset; | |
1373 | *file = bs; | |
1374 | return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; | |
1375 | } | |
1376 | ||
1377 | /* | |
1378 | * Work around the fact that the block layer doesn't do | |
1379 | * byte-accurate sizing yet - if the status request exceeds the | |
1380 | * server's advertised size because the block layer rounded size | |
1381 | * up, we truncated the request to the server (above), or are | |
1382 | * called on just the hole. | |
1383 | */ | |
611ae1d7 | 1384 | if (offset >= s->info.size) { |
86f8cdf3 VSO |
1385 | *pnum = bytes; |
1386 | assert(bytes < BDRV_SECTOR_SIZE); | |
1387 | /* Intentionally don't report offset_valid for the hole */ | |
1388 | return BDRV_BLOCK_ZERO; | |
1389 | } | |
1390 | ||
611ae1d7 VSO |
1391 | if (s->info.min_block) { |
1392 | assert(QEMU_IS_ALIGNED(request.len, s->info.min_block)); | |
86f8cdf3 | 1393 | } |
f7651539 VSO |
1394 | do { |
1395 | ret = nbd_co_send_request(bs, &request, NULL); | |
1396 | if (ret < 0) { | |
1397 | continue; | |
1398 | } | |
1399 | ||
1400 | ret = nbd_co_receive_blockstatus_reply(s, request.handle, bytes, | |
1401 | &extent, &request_ret, | |
1402 | &local_err); | |
1403 | if (local_err) { | |
1404 | trace_nbd_co_request_fail(request.from, request.len, request.handle, | |
1405 | request.flags, request.type, | |
1406 | nbd_cmd_lookup(request.type), | |
1407 | ret, error_get_pretty(local_err)); | |
1408 | error_free(local_err); | |
1409 | local_err = NULL; | |
1410 | } | |
1411 | } while (ret < 0 && nbd_client_connecting_wait(s)); | |
86f8cdf3 | 1412 | |
86f8cdf3 VSO |
1413 | if (ret < 0 || request_ret < 0) { |
1414 | return ret ? ret : request_ret; | |
1415 | } | |
1416 | ||
1417 | assert(extent.length); | |
1418 | *pnum = extent.length; | |
1419 | *map = offset; | |
1420 | *file = bs; | |
1421 | return (extent.flags & NBD_STATE_HOLE ? 0 : BDRV_BLOCK_DATA) | | |
1422 | (extent.flags & NBD_STATE_ZERO ? BDRV_BLOCK_ZERO : 0) | | |
1423 | BDRV_BLOCK_OFFSET_VALID; | |
1424 | } | |
1425 | ||
e99754b4 ML |
1426 | static int nbd_client_reopen_prepare(BDRVReopenState *state, |
1427 | BlockReopenQueue *queue, Error **errp) | |
1428 | { | |
1429 | BDRVNBDState *s = (BDRVNBDState *)state->bs->opaque; | |
1430 | ||
1431 | if ((state->flags & BDRV_O_RDWR) && (s->info.flags & NBD_FLAG_READ_ONLY)) { | |
1432 | error_setg(errp, "Can't reopen read-only NBD mount as read/write"); | |
1433 | return -EACCES; | |
1434 | } | |
1435 | return 0; | |
1436 | } | |
1437 | ||
fee091cd LS |
1438 | static void nbd_yank(void *opaque) |
1439 | { | |
1440 | BlockDriverState *bs = opaque; | |
1441 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; | |
1442 | ||
1443 | qatomic_store_release(&s->state, NBD_CLIENT_QUIT); | |
95a078ea | 1444 | qio_channel_shutdown(QIO_CHANNEL(s->ioc), QIO_CHANNEL_SHUTDOWN_BOTH, NULL); |
fee091cd LS |
1445 | } |
1446 | ||
86f8cdf3 VSO |
1447 | static void nbd_client_close(BlockDriverState *bs) |
1448 | { | |
611ae1d7 | 1449 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
86f8cdf3 VSO |
1450 | NBDRequest request = { .type = NBD_CMD_DISC }; |
1451 | ||
f7651539 VSO |
1452 | if (s->ioc) { |
1453 | nbd_send_request(s->ioc, &request); | |
1454 | } | |
86f8cdf3 VSO |
1455 | |
1456 | nbd_teardown_connection(bs); | |
1457 | } | |
1458 | ||
86f8cdf3 | 1459 | |
8f071c9d VSO |
1460 | /* |
1461 | * Parse nbd_open options | |
1462 | */ | |
86f8cdf3 | 1463 | |
f53a1feb | 1464 | static int nbd_parse_uri(const char *filename, QDict *options) |
1d7d2a9d PB |
1465 | { |
1466 | URI *uri; | |
1467 | const char *p; | |
1468 | QueryParams *qp = NULL; | |
1469 | int ret = 0; | |
f53a1feb | 1470 | bool is_unix; |
1d7d2a9d PB |
1471 | |
1472 | uri = uri_parse(filename); | |
1473 | if (!uri) { | |
1474 | return -EINVAL; | |
1475 | } | |
1476 | ||
1477 | /* transport */ | |
f69165a8 | 1478 | if (!g_strcmp0(uri->scheme, "nbd")) { |
f53a1feb | 1479 | is_unix = false; |
f69165a8 | 1480 | } else if (!g_strcmp0(uri->scheme, "nbd+tcp")) { |
f53a1feb | 1481 | is_unix = false; |
f69165a8 | 1482 | } else if (!g_strcmp0(uri->scheme, "nbd+unix")) { |
f53a1feb | 1483 | is_unix = true; |
1d7d2a9d PB |
1484 | } else { |
1485 | ret = -EINVAL; | |
1486 | goto out; | |
1487 | } | |
1488 | ||
2485f22f EB |
1489 | p = uri->path ? uri->path : ""; |
1490 | if (p[0] == '/') { | |
1491 | p++; | |
1492 | } | |
1d7d2a9d | 1493 | if (p[0]) { |
46f5ac20 | 1494 | qdict_put_str(options, "export", p); |
1d7d2a9d PB |
1495 | } |
1496 | ||
1497 | qp = query_params_parse(uri->query); | |
f53a1feb | 1498 | if (qp->n > 1 || (is_unix && !qp->n) || (!is_unix && qp->n)) { |
1d7d2a9d PB |
1499 | ret = -EINVAL; |
1500 | goto out; | |
1501 | } | |
1502 | ||
f53a1feb | 1503 | if (is_unix) { |
1d7d2a9d PB |
1504 | /* nbd+unix:///export?socket=path */ |
1505 | if (uri->server || uri->port || strcmp(qp->p[0].name, "socket")) { | |
1506 | ret = -EINVAL; | |
1507 | goto out; | |
1508 | } | |
46f5ac20 EB |
1509 | qdict_put_str(options, "server.type", "unix"); |
1510 | qdict_put_str(options, "server.path", qp->p[0].value); | |
1d7d2a9d | 1511 | } else { |
23307908 | 1512 | QString *host; |
f84d431b HR |
1513 | char *port_str; |
1514 | ||
bebbf7fa | 1515 | /* nbd[+tcp]://host[:port]/export */ |
1d7d2a9d PB |
1516 | if (!uri->server) { |
1517 | ret = -EINVAL; | |
1518 | goto out; | |
1519 | } | |
f17c90be | 1520 | |
23307908 JT |
1521 | /* strip braces from literal IPv6 address */ |
1522 | if (uri->server[0] == '[') { | |
1523 | host = qstring_from_substr(uri->server, 1, | |
ba891d68 | 1524 | strlen(uri->server) - 1); |
23307908 JT |
1525 | } else { |
1526 | host = qstring_from_str(uri->server); | |
1527 | } | |
1528 | ||
46f5ac20 | 1529 | qdict_put_str(options, "server.type", "inet"); |
9445673e | 1530 | qdict_put(options, "server.host", host); |
f84d431b HR |
1531 | |
1532 | port_str = g_strdup_printf("%d", uri->port ?: NBD_DEFAULT_PORT); | |
46f5ac20 | 1533 | qdict_put_str(options, "server.port", port_str); |
f84d431b | 1534 | g_free(port_str); |
1d7d2a9d PB |
1535 | } |
1536 | ||
1537 | out: | |
1538 | if (qp) { | |
1539 | query_params_free(qp); | |
1540 | } | |
1541 | uri_free(uri); | |
1542 | return ret; | |
1543 | } | |
1544 | ||
48c38e0b HR |
1545 | static bool nbd_has_filename_options_conflict(QDict *options, Error **errp) |
1546 | { | |
1547 | const QDictEntry *e; | |
1548 | ||
1549 | for (e = qdict_first(options); e; e = qdict_next(options, e)) { | |
1550 | if (!strcmp(e->key, "host") || | |
1551 | !strcmp(e->key, "port") || | |
1552 | !strcmp(e->key, "path") || | |
491d6c7c HR |
1553 | !strcmp(e->key, "export") || |
1554 | strstart(e->key, "server.", NULL)) | |
48c38e0b HR |
1555 | { |
1556 | error_setg(errp, "Option '%s' cannot be used with a file name", | |
1557 | e->key); | |
1558 | return true; | |
1559 | } | |
1560 | } | |
1561 | ||
1562 | return false; | |
1563 | } | |
1564 | ||
6963a30d KW |
1565 | static void nbd_parse_filename(const char *filename, QDict *options, |
1566 | Error **errp) | |
75818250 | 1567 | { |
df18c04e | 1568 | g_autofree char *file = NULL; |
33897dc7 NT |
1569 | char *export_name; |
1570 | const char *host_spec; | |
75818250 | 1571 | const char *unixpath; |
75818250 | 1572 | |
48c38e0b | 1573 | if (nbd_has_filename_options_conflict(options, errp)) { |
681e7ad0 KW |
1574 | return; |
1575 | } | |
1576 | ||
1d7d2a9d | 1577 | if (strstr(filename, "://")) { |
6963a30d KW |
1578 | int ret = nbd_parse_uri(filename, options); |
1579 | if (ret < 0) { | |
1580 | error_setg(errp, "No valid URL specified"); | |
1581 | } | |
1582 | return; | |
1d7d2a9d PB |
1583 | } |
1584 | ||
7267c094 | 1585 | file = g_strdup(filename); |
1d45f8b5 | 1586 | |
33897dc7 NT |
1587 | export_name = strstr(file, EN_OPTSTR); |
1588 | if (export_name) { | |
1589 | if (export_name[strlen(EN_OPTSTR)] == 0) { | |
df18c04e | 1590 | return; |
1d45f8b5 | 1591 | } |
33897dc7 NT |
1592 | export_name[0] = 0; /* truncate 'file' */ |
1593 | export_name += strlen(EN_OPTSTR); | |
f53a1feb | 1594 | |
46f5ac20 | 1595 | qdict_put_str(options, "export", export_name); |
1d45f8b5 LV |
1596 | } |
1597 | ||
33897dc7 NT |
1598 | /* extract the host_spec - fail if it's not nbd:... */ |
1599 | if (!strstart(file, "nbd:", &host_spec)) { | |
6963a30d | 1600 | error_setg(errp, "File name string for NBD must start with 'nbd:'"); |
df18c04e | 1601 | return; |
1d45f8b5 | 1602 | } |
75818250 | 1603 | |
f53a1feb | 1604 | if (!*host_spec) { |
df18c04e | 1605 | return; |
f53a1feb KW |
1606 | } |
1607 | ||
33897dc7 NT |
1608 | /* are we a UNIX or TCP socket? */ |
1609 | if (strstart(host_spec, "unix:", &unixpath)) { | |
46f5ac20 EB |
1610 | qdict_put_str(options, "server.type", "unix"); |
1611 | qdict_put_str(options, "server.path", unixpath); | |
75818250 | 1612 | } else { |
0785bd7a | 1613 | InetSocketAddress *addr = g_new(InetSocketAddress, 1); |
f53a1feb | 1614 | |
0785bd7a MA |
1615 | if (inet_parse(addr, host_spec, errp)) { |
1616 | goto out_inet; | |
f17c90be | 1617 | } |
75818250 | 1618 | |
46f5ac20 EB |
1619 | qdict_put_str(options, "server.type", "inet"); |
1620 | qdict_put_str(options, "server.host", addr->host); | |
1621 | qdict_put_str(options, "server.port", addr->port); | |
0785bd7a | 1622 | out_inet: |
f53a1feb KW |
1623 | qapi_free_InetSocketAddress(addr); |
1624 | } | |
f53a1feb KW |
1625 | } |
1626 | ||
491d6c7c HR |
1627 | static bool nbd_process_legacy_socket_options(QDict *output_options, |
1628 | QemuOpts *legacy_opts, | |
1629 | Error **errp) | |
f53a1feb | 1630 | { |
491d6c7c HR |
1631 | const char *path = qemu_opt_get(legacy_opts, "path"); |
1632 | const char *host = qemu_opt_get(legacy_opts, "host"); | |
1633 | const char *port = qemu_opt_get(legacy_opts, "port"); | |
1634 | const QDictEntry *e; | |
f53a1feb | 1635 | |
491d6c7c HR |
1636 | if (!path && !host && !port) { |
1637 | return true; | |
1638 | } | |
03504d05 | 1639 | |
491d6c7c HR |
1640 | for (e = qdict_first(output_options); e; e = qdict_next(output_options, e)) |
1641 | { | |
1642 | if (strstart(e->key, "server.", NULL)) { | |
1643 | error_setg(errp, "Cannot use 'server' and path/host/port at the " | |
1644 | "same time"); | |
1645 | return false; | |
681e7ad0 | 1646 | } |
33897dc7 | 1647 | } |
491d6c7c HR |
1648 | |
1649 | if (path && host) { | |
1650 | error_setg(errp, "path and host may not be used at the same time"); | |
1651 | return false; | |
1652 | } else if (path) { | |
1653 | if (port) { | |
1654 | error_setg(errp, "port may not be used without host"); | |
1655 | return false; | |
1656 | } | |
1657 | ||
46f5ac20 EB |
1658 | qdict_put_str(output_options, "server.type", "unix"); |
1659 | qdict_put_str(output_options, "server.path", path); | |
491d6c7c | 1660 | } else if (host) { |
46f5ac20 EB |
1661 | qdict_put_str(output_options, "server.type", "inet"); |
1662 | qdict_put_str(output_options, "server.host", host); | |
1663 | qdict_put_str(output_options, "server.port", | |
1664 | port ?: stringify(NBD_DEFAULT_PORT)); | |
442045cb | 1665 | } |
f53a1feb | 1666 | |
491d6c7c HR |
1667 | return true; |
1668 | } | |
f53a1feb | 1669 | |
62cf396b MA |
1670 | static SocketAddress *nbd_config(BDRVNBDState *s, QDict *options, |
1671 | Error **errp) | |
491d6c7c | 1672 | { |
62cf396b | 1673 | SocketAddress *saddr = NULL; |
491d6c7c | 1674 | QDict *addr = NULL; |
491d6c7c | 1675 | Visitor *iv = NULL; |
491d6c7c HR |
1676 | |
1677 | qdict_extract_subqdict(options, &addr, "server."); | |
1678 | if (!qdict_size(addr)) { | |
1679 | error_setg(errp, "NBD server address missing"); | |
1680 | goto done; | |
f53a1feb KW |
1681 | } |
1682 | ||
af91062e MA |
1683 | iv = qobject_input_visitor_new_flat_confused(addr, errp); |
1684 | if (!iv) { | |
491d6c7c HR |
1685 | goto done; |
1686 | } | |
bebbf7fa | 1687 | |
af175e85 | 1688 | if (!visit_type_SocketAddress(iv, NULL, &saddr, errp)) { |
491d6c7c HR |
1689 | goto done; |
1690 | } | |
7a5ed437 | 1691 | |
6cc702be VSO |
1692 | if (socket_address_parse_named_fd(saddr, errp) < 0) { |
1693 | qapi_free_SocketAddress(saddr); | |
1694 | saddr = NULL; | |
1695 | goto done; | |
1696 | } | |
1697 | ||
491d6c7c | 1698 | done: |
cb3e7f08 | 1699 | qobject_unref(addr); |
491d6c7c | 1700 | visit_free(iv); |
7a5ed437 | 1701 | return saddr; |
33897dc7 | 1702 | } |
1d45f8b5 | 1703 | |
75822a12 DB |
1704 | static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp) |
1705 | { | |
1706 | Object *obj; | |
1707 | QCryptoTLSCreds *creds; | |
1708 | ||
1709 | obj = object_resolve_path_component( | |
1710 | object_get_objects_root(), id); | |
1711 | if (!obj) { | |
1712 | error_setg(errp, "No TLS credentials with id '%s'", | |
1713 | id); | |
1714 | return NULL; | |
1715 | } | |
1716 | creds = (QCryptoTLSCreds *) | |
1717 | object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS); | |
1718 | if (!creds) { | |
1719 | error_setg(errp, "Object with id '%s' is not TLS credentials", | |
1720 | id); | |
1721 | return NULL; | |
1722 | } | |
1723 | ||
7b3b6168 PMD |
1724 | if (!qcrypto_tls_creds_check_endpoint(creds, |
1725 | QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT, | |
1726 | errp)) { | |
75822a12 DB |
1727 | return NULL; |
1728 | } | |
1729 | object_ref(obj); | |
1730 | return creds; | |
1731 | } | |
1732 | ||
1733 | ||
7ccc44fd HR |
1734 | static QemuOptsList nbd_runtime_opts = { |
1735 | .name = "nbd", | |
1736 | .head = QTAILQ_HEAD_INITIALIZER(nbd_runtime_opts.head), | |
1737 | .desc = { | |
1738 | { | |
1739 | .name = "host", | |
1740 | .type = QEMU_OPT_STRING, | |
1741 | .help = "TCP host to connect to", | |
1742 | }, | |
1743 | { | |
1744 | .name = "port", | |
1745 | .type = QEMU_OPT_STRING, | |
1746 | .help = "TCP port to connect to", | |
1747 | }, | |
1748 | { | |
1749 | .name = "path", | |
1750 | .type = QEMU_OPT_STRING, | |
1751 | .help = "Unix socket path to connect to", | |
1752 | }, | |
1753 | { | |
1754 | .name = "export", | |
1755 | .type = QEMU_OPT_STRING, | |
1756 | .help = "Name of the NBD export to open", | |
1757 | }, | |
1758 | { | |
1759 | .name = "tls-creds", | |
1760 | .type = QEMU_OPT_STRING, | |
1761 | .help = "ID of the TLS credentials to use", | |
1762 | }, | |
216ee365 EB |
1763 | { |
1764 | .name = "x-dirty-bitmap", | |
1765 | .type = QEMU_OPT_STRING, | |
1766 | .help = "experimental: expose named dirty bitmap in place of " | |
1767 | "block status", | |
1768 | }, | |
b172ae2e VSO |
1769 | { |
1770 | .name = "reconnect-delay", | |
1771 | .type = QEMU_OPT_NUMBER, | |
1772 | .help = "On an unexpected disconnect, the nbd client tries to " | |
1773 | "connect again until succeeding or encountering a serious " | |
1774 | "error. During the first @reconnect-delay seconds, all " | |
1775 | "requests are paused and will be rerun on a successful " | |
1776 | "reconnect. After that time, any delayed requests and all " | |
1777 | "future requests before a successful reconnect will " | |
1778 | "immediately fail. Default 0", | |
1779 | }, | |
be16b8bf VSO |
1780 | { |
1781 | .name = "open-timeout", | |
1782 | .type = QEMU_OPT_NUMBER, | |
1783 | .help = "In seconds. If zero, the nbd driver tries the connection " | |
1784 | "only once, and fails to open if the connection fails. " | |
1785 | "If non-zero, the nbd driver will repeat connection " | |
1786 | "attempts until successful or until @open-timeout seconds " | |
1787 | "have elapsed. Default 0", | |
1788 | }, | |
c4365735 | 1789 | { /* end of list */ } |
7ccc44fd HR |
1790 | }, |
1791 | }; | |
1792 | ||
8f071c9d VSO |
1793 | static int nbd_process_options(BlockDriverState *bs, QDict *options, |
1794 | Error **errp) | |
33897dc7 NT |
1795 | { |
1796 | BDRVNBDState *s = bs->opaque; | |
8f071c9d | 1797 | QemuOpts *opts; |
75822a12 | 1798 | int ret = -EINVAL; |
ae255e52 | 1799 | |
7ccc44fd | 1800 | opts = qemu_opts_create(&nbd_runtime_opts, NULL, 0, &error_abort); |
af175e85 | 1801 | if (!qemu_opts_absorb_qdict(opts, options, errp)) { |
7ccc44fd HR |
1802 | goto error; |
1803 | } | |
1804 | ||
62cf396b | 1805 | /* Translate @host, @port, and @path to a SocketAddress */ |
491d6c7c HR |
1806 | if (!nbd_process_legacy_socket_options(options, opts, errp)) { |
1807 | goto error; | |
1808 | } | |
1809 | ||
33897dc7 | 1810 | /* Pop the config into our state object. Exit if invalid. */ |
491d6c7c HR |
1811 | s->saddr = nbd_config(s, options, errp); |
1812 | if (!s->saddr) { | |
75822a12 DB |
1813 | goto error; |
1814 | } | |
1815 | ||
491d6c7c | 1816 | s->export = g_strdup(qemu_opt_get(opts, "export")); |
93676c88 EB |
1817 | if (s->export && strlen(s->export) > NBD_MAX_STRING_SIZE) { |
1818 | error_setg(errp, "export name too long to send to server"); | |
1819 | goto error; | |
1820 | } | |
491d6c7c | 1821 | |
03504d05 HR |
1822 | s->tlscredsid = g_strdup(qemu_opt_get(opts, "tls-creds")); |
1823 | if (s->tlscredsid) { | |
8f071c9d VSO |
1824 | s->tlscreds = nbd_get_tls_creds(s->tlscredsid, errp); |
1825 | if (!s->tlscreds) { | |
75822a12 DB |
1826 | goto error; |
1827 | } | |
1828 | ||
ca0b64e5 | 1829 | /* TODO SOCKET_ADDRESS_KIND_FD where fd has AF_INET or AF_INET6 */ |
62cf396b | 1830 | if (s->saddr->type != SOCKET_ADDRESS_TYPE_INET) { |
75822a12 DB |
1831 | error_setg(errp, "TLS only supported over IP sockets"); |
1832 | goto error; | |
1833 | } | |
8f071c9d | 1834 | s->hostname = s->saddr->u.inet.host; |
33897dc7 NT |
1835 | } |
1836 | ||
8f071c9d | 1837 | s->x_dirty_bitmap = g_strdup(qemu_opt_get(opts, "x-dirty-bitmap")); |
93676c88 EB |
1838 | if (s->x_dirty_bitmap && strlen(s->x_dirty_bitmap) > NBD_MAX_STRING_SIZE) { |
1839 | error_setg(errp, "x-dirty-bitmap query too long to send to server"); | |
1840 | goto error; | |
1841 | } | |
1842 | ||
8f071c9d | 1843 | s->reconnect_delay = qemu_opt_get_number(opts, "reconnect-delay", 0); |
be16b8bf | 1844 | s->open_timeout = qemu_opt_get_number(opts, "open-timeout", 0); |
8f071c9d VSO |
1845 | |
1846 | ret = 0; | |
d42f78e9 | 1847 | |
75822a12 | 1848 | error: |
7ccc44fd | 1849 | qemu_opts_del(opts); |
75822a12 | 1850 | return ret; |
e183ef75 PB |
1851 | } |
1852 | ||
8f071c9d VSO |
1853 | static int nbd_open(BlockDriverState *bs, QDict *options, int flags, |
1854 | Error **errp) | |
1855 | { | |
1856 | int ret; | |
1857 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; | |
1858 | ||
8f071c9d VSO |
1859 | s->bs = bs; |
1860 | qemu_co_mutex_init(&s->send_mutex); | |
1861 | qemu_co_queue_init(&s->free_sema); | |
4ddb5d2f | 1862 | qemu_co_mutex_init(&s->receive_mutex); |
8f071c9d | 1863 | |
fee091cd LS |
1864 | if (!yank_register_instance(BLOCKDEV_YANK_INSTANCE(bs->node_name), errp)) { |
1865 | return -EEXIST; | |
1866 | } | |
1867 | ||
bbba1c37 VSO |
1868 | ret = nbd_process_options(bs, options, errp); |
1869 | if (ret < 0) { | |
1870 | goto fail; | |
1871 | } | |
1872 | ||
6d2b0332 VSO |
1873 | s->conn = nbd_client_connection_new(s->saddr, true, s->export, |
1874 | s->x_dirty_bitmap, s->tlscreds); | |
e8b35bf5 | 1875 | |
be16b8bf VSO |
1876 | if (s->open_timeout) { |
1877 | nbd_client_connection_enable_retry(s->conn); | |
1878 | open_timer_init(s, qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + | |
1879 | s->open_timeout * NANOSECONDS_PER_SECOND); | |
1880 | } | |
1881 | ||
4ddb5d2f | 1882 | s->state = NBD_CLIENT_CONNECTING_WAIT; |
a71d597b | 1883 | ret = nbd_do_establish_connection(bs, errp); |
8f071c9d | 1884 | if (ret < 0) { |
bbba1c37 | 1885 | goto fail; |
8f071c9d | 1886 | } |
8f071c9d | 1887 | |
4ddb5d2f | 1888 | nbd_client_connection_enable_retry(s->conn); |
8f071c9d VSO |
1889 | |
1890 | return 0; | |
bbba1c37 VSO |
1891 | |
1892 | fail: | |
1893 | nbd_clear_bdrvstate(bs); | |
1894 | return ret; | |
8f071c9d VSO |
1895 | } |
1896 | ||
1486d04a PB |
1897 | static int nbd_co_flush(BlockDriverState *bs) |
1898 | { | |
f53a829b | 1899 | return nbd_client_co_flush(bs); |
1486d04a PB |
1900 | } |
1901 | ||
fa21e6fa DL |
1902 | static void nbd_refresh_limits(BlockDriverState *bs, Error **errp) |
1903 | { | |
611ae1d7 | 1904 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; |
fd8d372d | 1905 | uint32_t min = s->info.min_block; |
081dd1fe EB |
1906 | uint32_t max = MIN_NON_ZERO(NBD_MAX_BUFFER_SIZE, s->info.max_block); |
1907 | ||
7da537f7 EB |
1908 | /* |
1909 | * If the server did not advertise an alignment: | |
1910 | * - a size that is not sector-aligned implies that an alignment | |
1911 | * of 1 can be used to access those tail bytes | |
1912 | * - advertisement of block status requires an alignment of 1, so | |
1913 | * that we don't violate block layer constraints that block | |
1914 | * status is always aligned (as we can't control whether the | |
1915 | * server will report sub-sector extents, such as a hole at EOF | |
1916 | * on an unaligned POSIX file) | |
1917 | * - otherwise, assume the server is so old that we are safer avoiding | |
1918 | * sub-sector requests | |
1919 | */ | |
1920 | if (!min) { | |
1921 | min = (!QEMU_IS_ALIGNED(s->info.size, BDRV_SECTOR_SIZE) || | |
1922 | s->info.base_allocation) ? 1 : BDRV_SECTOR_SIZE; | |
1923 | } | |
1924 | ||
1925 | bs->bl.request_alignment = min; | |
714eb0db | 1926 | bs->bl.max_pdiscard = QEMU_ALIGN_DOWN(INT_MAX, min); |
081dd1fe EB |
1927 | bs->bl.max_pwrite_zeroes = max; |
1928 | bs->bl.max_transfer = max; | |
1929 | ||
1930 | if (s->info.opt_block && | |
1931 | s->info.opt_block > bs->bl.opt_transfer) { | |
1932 | bs->bl.opt_transfer = s->info.opt_block; | |
1933 | } | |
fa21e6fa DL |
1934 | } |
1935 | ||
75818250 TS |
1936 | static void nbd_close(BlockDriverState *bs) |
1937 | { | |
f53a829b | 1938 | nbd_client_close(bs); |
bbba1c37 | 1939 | nbd_clear_bdrvstate(bs); |
75818250 TS |
1940 | } |
1941 | ||
a2b333c0 NS |
1942 | /* |
1943 | * NBD cannot truncate, but if the caller asks to truncate to the same size, or | |
1944 | * to a smaller size with exact=false, there is no reason to fail the | |
1945 | * operation. | |
1946 | * | |
1947 | * Preallocation mode is ignored since it does not seems useful to fail when | |
1948 | * we never change anything. | |
1949 | */ | |
1950 | static int coroutine_fn nbd_co_truncate(BlockDriverState *bs, int64_t offset, | |
1951 | bool exact, PreallocMode prealloc, | |
1952 | BdrvRequestFlags flags, Error **errp) | |
1953 | { | |
1954 | BDRVNBDState *s = bs->opaque; | |
1955 | ||
1956 | if (offset != s->info.size && exact) { | |
1957 | error_setg(errp, "Cannot resize NBD nodes"); | |
1958 | return -ENOTSUP; | |
1959 | } | |
1960 | ||
1961 | if (offset > s->info.size) { | |
1962 | error_setg(errp, "Cannot grow NBD nodes"); | |
1963 | return -EINVAL; | |
1964 | } | |
1965 | ||
1966 | return 0; | |
1967 | } | |
1968 | ||
75818250 TS |
1969 | static int64_t nbd_getlength(BlockDriverState *bs) |
1970 | { | |
1971 | BDRVNBDState *s = bs->opaque; | |
1972 | ||
611ae1d7 | 1973 | return s->info.size; |
75818250 TS |
1974 | } |
1975 | ||
998b3a1e | 1976 | static void nbd_refresh_filename(BlockDriverState *bs) |
2019d68b | 1977 | { |
03504d05 | 1978 | BDRVNBDState *s = bs->opaque; |
491d6c7c | 1979 | const char *host = NULL, *port = NULL, *path = NULL; |
5c86bdf1 | 1980 | size_t len = 0; |
491d6c7c | 1981 | |
62cf396b | 1982 | if (s->saddr->type == SOCKET_ADDRESS_TYPE_INET) { |
9445673e | 1983 | const InetSocketAddress *inet = &s->saddr->u.inet; |
491d6c7c HR |
1984 | if (!inet->has_ipv4 && !inet->has_ipv6 && !inet->has_to) { |
1985 | host = inet->host; | |
1986 | port = inet->port; | |
1987 | } | |
62cf396b | 1988 | } else if (s->saddr->type == SOCKET_ADDRESS_TYPE_UNIX) { |
9445673e MA |
1989 | path = s->saddr->u.q_unix.path; |
1990 | } /* else can't represent as pseudo-filename */ | |
2019d68b | 1991 | |
491d6c7c | 1992 | if (path && s->export) { |
5c86bdf1 EB |
1993 | len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), |
1994 | "nbd+unix:///%s?socket=%s", s->export, path); | |
491d6c7c | 1995 | } else if (path && !s->export) { |
5c86bdf1 EB |
1996 | len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), |
1997 | "nbd+unix://?socket=%s", path); | |
491d6c7c | 1998 | } else if (host && s->export) { |
5c86bdf1 EB |
1999 | len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), |
2000 | "nbd://%s:%s/%s", host, port, s->export); | |
491d6c7c | 2001 | } else if (host && !s->export) { |
5c86bdf1 EB |
2002 | len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), |
2003 | "nbd://%s:%s", host, port); | |
2004 | } | |
00d69986 | 2005 | if (len >= sizeof(bs->exact_filename)) { |
5c86bdf1 EB |
2006 | /* Name is too long to represent exactly, so leave it empty. */ |
2007 | bs->exact_filename[0] = '\0'; | |
ec0de768 | 2008 | } |
2019d68b HR |
2009 | } |
2010 | ||
8a6239c0 HR |
2011 | static char *nbd_dirname(BlockDriverState *bs, Error **errp) |
2012 | { | |
2013 | /* The generic bdrv_dirname() implementation is able to work out some | |
2014 | * directory name for NBD nodes, but that would be wrong. So far there is no | |
2015 | * specification for how "export paths" would work, so NBD does not have | |
2016 | * directory names. */ | |
2017 | error_setg(errp, "Cannot generate a base directory for NBD nodes"); | |
2018 | return NULL; | |
2019 | } | |
2020 | ||
2654267c HR |
2021 | static const char *const nbd_strong_runtime_opts[] = { |
2022 | "path", | |
2023 | "host", | |
2024 | "port", | |
2025 | "export", | |
2026 | "tls-creds", | |
2027 | "server.", | |
2028 | ||
2029 | NULL | |
2030 | }; | |
2031 | ||
c4f7f24e VSO |
2032 | static void nbd_cancel_in_flight(BlockDriverState *bs) |
2033 | { | |
2034 | BDRVNBDState *s = (BDRVNBDState *)bs->opaque; | |
2035 | ||
2036 | reconnect_delay_timer_del(s); | |
2037 | ||
2038 | if (s->state == NBD_CLIENT_CONNECTING_WAIT) { | |
2039 | s->state = NBD_CLIENT_CONNECTING_NOWAIT; | |
2040 | qemu_co_queue_restart_all(&s->free_sema); | |
2041 | } | |
4ddb5d2f VSO |
2042 | |
2043 | nbd_co_establish_connection_cancel(s->conn); | |
c4f7f24e VSO |
2044 | } |
2045 | ||
5efa9d5a | 2046 | static BlockDriver bdrv_nbd = { |
69447cd8 SH |
2047 | .format_name = "nbd", |
2048 | .protocol_name = "nbd", | |
2049 | .instance_size = sizeof(BDRVNBDState), | |
2050 | .bdrv_parse_filename = nbd_parse_filename, | |
5a5e7f8c ML |
2051 | .bdrv_co_create_opts = bdrv_co_create_opts_simple, |
2052 | .create_opts = &bdrv_create_opts_simple, | |
69447cd8 | 2053 | .bdrv_file_open = nbd_open, |
e99754b4 | 2054 | .bdrv_reopen_prepare = nbd_client_reopen_prepare, |
70c4fb26 EB |
2055 | .bdrv_co_preadv = nbd_client_co_preadv, |
2056 | .bdrv_co_pwritev = nbd_client_co_pwritev, | |
fa778fff | 2057 | .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, |
69447cd8 SH |
2058 | .bdrv_close = nbd_close, |
2059 | .bdrv_co_flush_to_os = nbd_co_flush, | |
447e57c3 | 2060 | .bdrv_co_pdiscard = nbd_client_co_pdiscard, |
fa21e6fa | 2061 | .bdrv_refresh_limits = nbd_refresh_limits, |
a2b333c0 | 2062 | .bdrv_co_truncate = nbd_co_truncate, |
69447cd8 | 2063 | .bdrv_getlength = nbd_getlength, |
2019d68b | 2064 | .bdrv_refresh_filename = nbd_refresh_filename, |
78a33ab5 | 2065 | .bdrv_co_block_status = nbd_client_co_block_status, |
8a6239c0 | 2066 | .bdrv_dirname = nbd_dirname, |
2654267c | 2067 | .strong_runtime_opts = nbd_strong_runtime_opts, |
c4f7f24e | 2068 | .bdrv_cancel_in_flight = nbd_cancel_in_flight, |
1d7d2a9d PB |
2069 | }; |
2070 | ||
2071 | static BlockDriver bdrv_nbd_tcp = { | |
69447cd8 SH |
2072 | .format_name = "nbd", |
2073 | .protocol_name = "nbd+tcp", | |
2074 | .instance_size = sizeof(BDRVNBDState), | |
2075 | .bdrv_parse_filename = nbd_parse_filename, | |
5a5e7f8c ML |
2076 | .bdrv_co_create_opts = bdrv_co_create_opts_simple, |
2077 | .create_opts = &bdrv_create_opts_simple, | |
69447cd8 | 2078 | .bdrv_file_open = nbd_open, |
e99754b4 | 2079 | .bdrv_reopen_prepare = nbd_client_reopen_prepare, |
70c4fb26 EB |
2080 | .bdrv_co_preadv = nbd_client_co_preadv, |
2081 | .bdrv_co_pwritev = nbd_client_co_pwritev, | |
fa778fff | 2082 | .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, |
69447cd8 SH |
2083 | .bdrv_close = nbd_close, |
2084 | .bdrv_co_flush_to_os = nbd_co_flush, | |
447e57c3 | 2085 | .bdrv_co_pdiscard = nbd_client_co_pdiscard, |
fa21e6fa | 2086 | .bdrv_refresh_limits = nbd_refresh_limits, |
a2b333c0 | 2087 | .bdrv_co_truncate = nbd_co_truncate, |
69447cd8 | 2088 | .bdrv_getlength = nbd_getlength, |
2019d68b | 2089 | .bdrv_refresh_filename = nbd_refresh_filename, |
78a33ab5 | 2090 | .bdrv_co_block_status = nbd_client_co_block_status, |
8a6239c0 | 2091 | .bdrv_dirname = nbd_dirname, |
2654267c | 2092 | .strong_runtime_opts = nbd_strong_runtime_opts, |
c4f7f24e | 2093 | .bdrv_cancel_in_flight = nbd_cancel_in_flight, |
1d7d2a9d PB |
2094 | }; |
2095 | ||
2096 | static BlockDriver bdrv_nbd_unix = { | |
69447cd8 SH |
2097 | .format_name = "nbd", |
2098 | .protocol_name = "nbd+unix", | |
2099 | .instance_size = sizeof(BDRVNBDState), | |
2100 | .bdrv_parse_filename = nbd_parse_filename, | |
5a5e7f8c ML |
2101 | .bdrv_co_create_opts = bdrv_co_create_opts_simple, |
2102 | .create_opts = &bdrv_create_opts_simple, | |
69447cd8 | 2103 | .bdrv_file_open = nbd_open, |
e99754b4 | 2104 | .bdrv_reopen_prepare = nbd_client_reopen_prepare, |
70c4fb26 EB |
2105 | .bdrv_co_preadv = nbd_client_co_preadv, |
2106 | .bdrv_co_pwritev = nbd_client_co_pwritev, | |
fa778fff | 2107 | .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, |
69447cd8 SH |
2108 | .bdrv_close = nbd_close, |
2109 | .bdrv_co_flush_to_os = nbd_co_flush, | |
447e57c3 | 2110 | .bdrv_co_pdiscard = nbd_client_co_pdiscard, |
fa21e6fa | 2111 | .bdrv_refresh_limits = nbd_refresh_limits, |
a2b333c0 | 2112 | .bdrv_co_truncate = nbd_co_truncate, |
69447cd8 | 2113 | .bdrv_getlength = nbd_getlength, |
2019d68b | 2114 | .bdrv_refresh_filename = nbd_refresh_filename, |
78a33ab5 | 2115 | .bdrv_co_block_status = nbd_client_co_block_status, |
8a6239c0 | 2116 | .bdrv_dirname = nbd_dirname, |
2654267c | 2117 | .strong_runtime_opts = nbd_strong_runtime_opts, |
c4f7f24e | 2118 | .bdrv_cancel_in_flight = nbd_cancel_in_flight, |
75818250 | 2119 | }; |
5efa9d5a AL |
2120 | |
2121 | static void bdrv_nbd_init(void) | |
2122 | { | |
2123 | bdrv_register(&bdrv_nbd); | |
1d7d2a9d PB |
2124 | bdrv_register(&bdrv_nbd_tcp); |
2125 | bdrv_register(&bdrv_nbd_unix); | |
5efa9d5a AL |
2126 | } |
2127 | ||
2128 | block_init(bdrv_nbd_init); |