]>
Commit | Line | Data |
---|---|---|
26f54e9a MA |
1 | /* |
2 | * QEMU Block backends | |
3 | * | |
60cb2fa7 | 4 | * Copyright (C) 2014-2016 Red Hat, Inc. |
26f54e9a MA |
5 | * |
6 | * Authors: | |
7 | * Markus Armbruster <armbru@redhat.com>, | |
8 | * | |
9 | * This work is licensed under the terms of the GNU LGPL, version 2.1 | |
10 | * or later. See the COPYING.LIB file in the top-level directory. | |
11 | */ | |
12 | ||
80c71a24 | 13 | #include "qemu/osdep.h" |
26f54e9a MA |
14 | #include "sysemu/block-backend.h" |
15 | #include "block/block_int.h" | |
373340b2 | 16 | #include "block/blockjob.h" |
16d36e29 | 17 | #include "block/coroutines.h" |
281d22d8 | 18 | #include "block/throttle-groups.h" |
13d4ff07 | 19 | #include "hw/qdev-core.h" |
18e46a03 | 20 | #include "sysemu/blockdev.h" |
54d31236 | 21 | #include "sysemu/runstate.h" |
e4ec5ad4 | 22 | #include "sysemu/replay.h" |
e688df6b | 23 | #include "qapi/error.h" |
9af23989 | 24 | #include "qapi/qapi-events-block.h" |
f348b6d1 | 25 | #include "qemu/id.h" |
db725815 | 26 | #include "qemu/main-loop.h" |
922a01a0 | 27 | #include "qemu/option.h" |
1e98fefd | 28 | #include "trace.h" |
5f7772c4 | 29 | #include "migration/misc.h" |
a7f53e26 MA |
30 | |
31 | /* Number of coroutines to reserve per attached device model */ | |
32 | #define COROUTINE_POOL_RESERVATION 64 | |
26f54e9a | 33 | |
1bf1cbc9 KW |
34 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
35 | ||
4981bdec HR |
36 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb); |
37 | ||
d03654ea SH |
38 | typedef struct BlockBackendAioNotifier { |
39 | void (*attached_aio_context)(AioContext *new_context, void *opaque); | |
40 | void (*detach_aio_context)(void *opaque); | |
41 | void *opaque; | |
42 | QLIST_ENTRY(BlockBackendAioNotifier) list; | |
43 | } BlockBackendAioNotifier; | |
44 | ||
26f54e9a MA |
45 | struct BlockBackend { |
46 | char *name; | |
47 | int refcnt; | |
f21d96d0 | 48 | BdrvChild *root; |
d861ab3a | 49 | AioContext *ctx; |
26f8b3a8 | 50 | DriveInfo *legacy_dinfo; /* null unless created by drive_new() */ |
2cf22d6a | 51 | QTAILQ_ENTRY(BlockBackend) link; /* for block_backends */ |
9492b0b9 | 52 | QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */ |
f2cd875d | 53 | BlockBackendPublic public; |
a7f53e26 | 54 | |
d09ea2d2 | 55 | DeviceState *dev; /* attached device model, if any */ |
a7f53e26 MA |
56 | const BlockDevOps *dev_ops; |
57 | void *dev_opaque; | |
68e9ec01 | 58 | |
281d22d8 HR |
59 | /* If the BDS tree is removed, some of its options are stored here (which |
60 | * can be used to restore those options in the new BDS on insert) */ | |
61 | BlockBackendRootState root_state; | |
62 | ||
bfd18d1e KW |
63 | bool enable_write_cache; |
64 | ||
7f0e9da6 HR |
65 | /* I/O stats (display with "info blockstats"). */ |
66 | BlockAcctStats stats; | |
373340b2 HR |
67 | |
68 | BlockdevOnError on_read_error, on_write_error; | |
69 | bool iostatus_enabled; | |
70 | BlockDeviceIoStatus iostatus; | |
3301f6c6 | 71 | |
981776b3 KW |
72 | uint64_t perm; |
73 | uint64_t shared_perm; | |
d35ff5e6 | 74 | bool disable_perm; |
981776b3 | 75 | |
980b0f94 | 76 | bool allow_aio_context_change; |
c10c9d96 KW |
77 | bool allow_write_beyond_eof; |
78 | ||
a2c4c3b1 | 79 | /* Protected by BQL */ |
3301f6c6 | 80 | NotifierList remove_bs_notifiers, insert_bs_notifiers; |
d03654ea | 81 | QLIST_HEAD(, BlockBackendAioNotifier) aio_notifiers; |
f4d9cc88 | 82 | |
c4d5bf99 | 83 | int quiesce_counter; /* atomic: written under BQL, read by other threads */ |
407ae2ae | 84 | QemuMutex queued_requests_lock; /* protects queued_requests */ |
cf312932 | 85 | CoQueue queued_requests; |
ef80ec50 | 86 | bool disable_request_queuing; /* atomic */ |
cf312932 | 87 | |
5f7772c4 | 88 | VMChangeStateEntry *vmsh; |
ca2e2144 | 89 | bool force_allow_inactivate; |
33f2a757 SH |
90 | |
91 | /* Number of in-flight aio requests. BlockDriverState also counts | |
92 | * in-flight requests but aio requests can exist even when blk->root is | |
93 | * NULL, so we cannot rely on its counter for that case. | |
94 | * Accessed with atomic ops. | |
95 | */ | |
96 | unsigned int in_flight; | |
26f54e9a MA |
97 | }; |
98 | ||
e7f7d676 HR |
99 | typedef struct BlockBackendAIOCB { |
100 | BlockAIOCB common; | |
4981bdec | 101 | BlockBackend *blk; |
e7f7d676 HR |
102 | int ret; |
103 | } BlockBackendAIOCB; | |
104 | ||
105 | static const AIOCBInfo block_backend_aiocb_info = { | |
4981bdec | 106 | .get_aio_context = blk_aiocb_get_aio_context, |
e7f7d676 HR |
107 | .aiocb_size = sizeof(BlockBackendAIOCB), |
108 | }; | |
109 | ||
8fb3c76c | 110 | static void drive_info_del(DriveInfo *dinfo); |
7c8eece4 | 111 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs); |
8fb3c76c | 112 | |
a2c4c3b1 | 113 | /* All BlockBackends. Protected by BQL. */ |
2cf22d6a HR |
114 | static QTAILQ_HEAD(, BlockBackend) block_backends = |
115 | QTAILQ_HEAD_INITIALIZER(block_backends); | |
116 | ||
a2c4c3b1 EGE |
117 | /* |
118 | * All BlockBackends referenced by the monitor and which are iterated through by | |
119 | * blk_next(). Protected by BQL. | |
120 | */ | |
9492b0b9 HR |
121 | static QTAILQ_HEAD(, BlockBackend) monitor_block_backends = |
122 | QTAILQ_HEAD_INITIALIZER(monitor_block_backends); | |
26f54e9a | 123 | |
3cdc69d3 | 124 | static void blk_root_inherit_options(BdrvChildRole role, bool parent_is_format, |
272c02ea | 125 | int *child_flags, QDict *child_options, |
f21d96d0 KW |
126 | int parent_flags, QDict *parent_options) |
127 | { | |
128 | /* We're not supposed to call this function for root nodes */ | |
129 | abort(); | |
130 | } | |
c2066af0 | 131 | static void blk_root_drained_begin(BdrvChild *child); |
fe5258a5 | 132 | static bool blk_root_drained_poll(BdrvChild *child); |
2f65df6e | 133 | static void blk_root_drained_end(BdrvChild *child); |
f21d96d0 | 134 | |
5c8cab48 KW |
135 | static void blk_root_change_media(BdrvChild *child, bool load); |
136 | static void blk_root_resize(BdrvChild *child); | |
137 | ||
33949396 EGE |
138 | static bool blk_root_change_aio_ctx(BdrvChild *child, AioContext *ctx, |
139 | GHashTable *visited, Transaction *tran, | |
140 | Error **errp); | |
38475269 | 141 | |
b5411555 KW |
142 | static char *blk_root_get_parent_desc(BdrvChild *child) |
143 | { | |
144 | BlockBackend *blk = child->opaque; | |
fd240a18 | 145 | g_autofree char *dev_id = NULL; |
b5411555 KW |
146 | |
147 | if (blk->name) { | |
fd240a18 | 148 | return g_strdup_printf("block device '%s'", blk->name); |
b5411555 KW |
149 | } |
150 | ||
151 | dev_id = blk_get_attached_dev_id(blk); | |
152 | if (*dev_id) { | |
fd240a18 | 153 | return g_strdup_printf("block device '%s'", dev_id); |
b5411555 KW |
154 | } else { |
155 | /* TODO Callback into the BB owner for something more detailed */ | |
fd240a18 | 156 | return g_strdup("an unnamed block device"); |
b5411555 KW |
157 | } |
158 | } | |
159 | ||
4c265bf9 KW |
160 | static const char *blk_root_get_name(BdrvChild *child) |
161 | { | |
162 | return blk_name(child->opaque); | |
163 | } | |
164 | ||
538f0497 | 165 | static void blk_vm_state_changed(void *opaque, bool running, RunState state) |
5f7772c4 FZ |
166 | { |
167 | Error *local_err = NULL; | |
168 | BlockBackend *blk = opaque; | |
169 | ||
170 | if (state == RUN_STATE_INMIGRATE) { | |
171 | return; | |
172 | } | |
173 | ||
174 | qemu_del_vm_change_state_handler(blk->vmsh); | |
175 | blk->vmsh = NULL; | |
176 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); | |
177 | if (local_err) { | |
178 | error_report_err(local_err); | |
179 | } | |
180 | } | |
181 | ||
4417ab7a KW |
182 | /* |
183 | * Notifies the user of the BlockBackend that migration has completed. qdev | |
184 | * devices can tighten their permissions in response (specifically revoke | |
185 | * shared write permissions that we needed for storage migration). | |
186 | * | |
187 | * If an error is returned, the VM cannot be allowed to be resumed. | |
188 | */ | |
189 | static void blk_root_activate(BdrvChild *child, Error **errp) | |
190 | { | |
191 | BlockBackend *blk = child->opaque; | |
192 | Error *local_err = NULL; | |
492a1196 | 193 | uint64_t saved_shared_perm; |
4417ab7a KW |
194 | |
195 | if (!blk->disable_perm) { | |
196 | return; | |
197 | } | |
198 | ||
199 | blk->disable_perm = false; | |
200 | ||
492a1196 HR |
201 | /* |
202 | * blk->shared_perm contains the permissions we want to share once | |
203 | * migration is really completely done. For now, we need to share | |
204 | * all; but we also need to retain blk->shared_perm, which is | |
205 | * overwritten by a successful blk_set_perm() call. Save it and | |
206 | * restore it below. | |
207 | */ | |
208 | saved_shared_perm = blk->shared_perm; | |
209 | ||
5f7772c4 FZ |
210 | blk_set_perm(blk, blk->perm, BLK_PERM_ALL, &local_err); |
211 | if (local_err) { | |
212 | error_propagate(errp, local_err); | |
213 | blk->disable_perm = true; | |
214 | return; | |
215 | } | |
492a1196 | 216 | blk->shared_perm = saved_shared_perm; |
5f7772c4 FZ |
217 | |
218 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
219 | /* Activation can happen when migration process is still active, for | |
220 | * example when nbd_server_add is called during non-shared storage | |
221 | * migration. Defer the shared_perm update to migration completion. */ | |
222 | if (!blk->vmsh) { | |
223 | blk->vmsh = qemu_add_vm_change_state_handler(blk_vm_state_changed, | |
224 | blk); | |
225 | } | |
226 | return; | |
227 | } | |
228 | ||
4417ab7a KW |
229 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); |
230 | if (local_err) { | |
231 | error_propagate(errp, local_err); | |
232 | blk->disable_perm = true; | |
233 | return; | |
234 | } | |
235 | } | |
236 | ||
ca2e2144 FZ |
237 | void blk_set_force_allow_inactivate(BlockBackend *blk) |
238 | { | |
0439c5a4 | 239 | GLOBAL_STATE_CODE(); |
ca2e2144 FZ |
240 | blk->force_allow_inactivate = true; |
241 | } | |
242 | ||
c16de8f5 FZ |
243 | static bool blk_can_inactivate(BlockBackend *blk) |
244 | { | |
ca2e2144 | 245 | /* If it is a guest device, inactivate is ok. */ |
c16de8f5 FZ |
246 | if (blk->dev || blk_name(blk)[0]) { |
247 | return true; | |
248 | } | |
249 | ||
ca2e2144 FZ |
250 | /* Inactivating means no more writes to the image can be done, |
251 | * even if those writes would be changes invisible to the | |
252 | * guest. For block job BBs that satisfy this, we can just allow | |
253 | * it. This is the case for mirror job source, which is required | |
254 | * by libvirt non-shared block migration. */ | |
255 | if (!(blk->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED))) { | |
256 | return true; | |
257 | } | |
258 | ||
259 | return blk->force_allow_inactivate; | |
c16de8f5 FZ |
260 | } |
261 | ||
cfa1a572 KW |
262 | static int blk_root_inactivate(BdrvChild *child) |
263 | { | |
264 | BlockBackend *blk = child->opaque; | |
265 | ||
266 | if (blk->disable_perm) { | |
267 | return 0; | |
268 | } | |
269 | ||
c16de8f5 | 270 | if (!blk_can_inactivate(blk)) { |
cfa1a572 KW |
271 | return -EPERM; |
272 | } | |
273 | ||
274 | blk->disable_perm = true; | |
275 | if (blk->root) { | |
276 | bdrv_child_try_set_perm(blk->root, 0, BLK_PERM_ALL, &error_abort); | |
277 | } | |
278 | ||
279 | return 0; | |
280 | } | |
281 | ||
d03654ea SH |
282 | static void blk_root_attach(BdrvChild *child) |
283 | { | |
284 | BlockBackend *blk = child->opaque; | |
285 | BlockBackendAioNotifier *notifier; | |
286 | ||
287 | trace_blk_root_attach(child, blk, child->bs); | |
288 | ||
289 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
290 | bdrv_add_aio_context_notifier(child->bs, | |
291 | notifier->attached_aio_context, | |
292 | notifier->detach_aio_context, | |
293 | notifier->opaque); | |
294 | } | |
295 | } | |
296 | ||
297 | static void blk_root_detach(BdrvChild *child) | |
298 | { | |
299 | BlockBackend *blk = child->opaque; | |
300 | BlockBackendAioNotifier *notifier; | |
301 | ||
302 | trace_blk_root_detach(child, blk, child->bs); | |
303 | ||
304 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
305 | bdrv_remove_aio_context_notifier(child->bs, | |
306 | notifier->attached_aio_context, | |
307 | notifier->detach_aio_context, | |
308 | notifier->opaque); | |
309 | } | |
310 | } | |
311 | ||
3ca1f322 VSO |
312 | static AioContext *blk_root_get_parent_aio_context(BdrvChild *c) |
313 | { | |
314 | BlockBackend *blk = c->opaque; | |
d5f8d79c | 315 | IO_CODE(); |
3ca1f322 VSO |
316 | |
317 | return blk_get_aio_context(blk); | |
318 | } | |
319 | ||
bd86fb99 | 320 | static const BdrvChildClass child_root = { |
c2066af0 KW |
321 | .inherit_options = blk_root_inherit_options, |
322 | ||
5c8cab48 KW |
323 | .change_media = blk_root_change_media, |
324 | .resize = blk_root_resize, | |
4c265bf9 | 325 | .get_name = blk_root_get_name, |
b5411555 | 326 | .get_parent_desc = blk_root_get_parent_desc, |
5c8cab48 | 327 | |
c2066af0 | 328 | .drained_begin = blk_root_drained_begin, |
fe5258a5 | 329 | .drained_poll = blk_root_drained_poll, |
c2066af0 | 330 | .drained_end = blk_root_drained_end, |
4417ab7a KW |
331 | |
332 | .activate = blk_root_activate, | |
cfa1a572 | 333 | .inactivate = blk_root_inactivate, |
d03654ea SH |
334 | |
335 | .attach = blk_root_attach, | |
336 | .detach = blk_root_detach, | |
38475269 | 337 | |
33949396 | 338 | .change_aio_ctx = blk_root_change_aio_ctx, |
3ca1f322 VSO |
339 | |
340 | .get_parent_aio_context = blk_root_get_parent_aio_context, | |
f21d96d0 KW |
341 | }; |
342 | ||
26f54e9a | 343 | /* |
efaa7c4e | 344 | * Create a new BlockBackend with a reference count of one. |
6d0eb64d KW |
345 | * |
346 | * @perm is a bitmasks of BLK_PERM_* constants which describes the permissions | |
347 | * to request for a block driver node that is attached to this BlockBackend. | |
348 | * @shared_perm is a bitmask which describes which permissions may be granted | |
349 | * to other users of the attached node. | |
350 | * Both sets of permissions can be changed later using blk_set_perm(). | |
351 | * | |
26f54e9a MA |
352 | * Return the new BlockBackend on success, null on failure. |
353 | */ | |
d861ab3a | 354 | BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm) |
26f54e9a MA |
355 | { |
356 | BlockBackend *blk; | |
357 | ||
0439c5a4 EGE |
358 | GLOBAL_STATE_CODE(); |
359 | ||
26f54e9a | 360 | blk = g_new0(BlockBackend, 1); |
26f54e9a | 361 | blk->refcnt = 1; |
d861ab3a | 362 | blk->ctx = ctx; |
6d0eb64d KW |
363 | blk->perm = perm; |
364 | blk->shared_perm = shared_perm; | |
0c3169df KW |
365 | blk_set_enable_write_cache(blk, true); |
366 | ||
cb53460b KW |
367 | blk->on_read_error = BLOCKDEV_ON_ERROR_REPORT; |
368 | blk->on_write_error = BLOCKDEV_ON_ERROR_ENOSPC; | |
369 | ||
9caa6f3d | 370 | block_acct_init(&blk->stats); |
27ccdd52 | 371 | |
407ae2ae | 372 | qemu_mutex_init(&blk->queued_requests_lock); |
cf312932 | 373 | qemu_co_queue_init(&blk->queued_requests); |
3301f6c6 HR |
374 | notifier_list_init(&blk->remove_bs_notifiers); |
375 | notifier_list_init(&blk->insert_bs_notifiers); | |
d03654ea | 376 | QLIST_INIT(&blk->aio_notifiers); |
27ccdd52 | 377 | |
2cf22d6a | 378 | QTAILQ_INSERT_TAIL(&block_backends, blk, link); |
26f54e9a MA |
379 | return blk; |
380 | } | |
381 | ||
a3aeeab5 EB |
382 | /* |
383 | * Create a new BlockBackend connected to an existing BlockDriverState. | |
384 | * | |
385 | * @perm is a bitmasks of BLK_PERM_* constants which describes the | |
386 | * permissions to request for @bs that is attached to this | |
387 | * BlockBackend. @shared_perm is a bitmask which describes which | |
388 | * permissions may be granted to other users of the attached node. | |
389 | * Both sets of permissions can be changed later using blk_set_perm(). | |
390 | * | |
391 | * Return the new BlockBackend on success, null on failure. | |
392 | */ | |
393 | BlockBackend *blk_new_with_bs(BlockDriverState *bs, uint64_t perm, | |
394 | uint64_t shared_perm, Error **errp) | |
395 | { | |
396 | BlockBackend *blk = blk_new(bdrv_get_aio_context(bs), perm, shared_perm); | |
397 | ||
0439c5a4 EGE |
398 | GLOBAL_STATE_CODE(); |
399 | ||
a3aeeab5 EB |
400 | if (blk_insert_bs(blk, bs, errp) < 0) { |
401 | blk_unref(blk); | |
402 | return NULL; | |
403 | } | |
404 | return blk; | |
405 | } | |
406 | ||
7e7d56d9 | 407 | /* |
28eb9b12 | 408 | * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. |
d861ab3a | 409 | * The new BlockBackend is in the main AioContext. |
ca49a4fd HR |
410 | * |
411 | * Just as with bdrv_open(), after having called this function the reference to | |
412 | * @options belongs to the block layer (even on failure). | |
413 | * | |
414 | * TODO: Remove @filename and @flags; it should be possible to specify a whole | |
415 | * BDS tree just by specifying the @options QDict (or @reference, | |
416 | * alternatively). At the time of adding this function, this is not possible, | |
417 | * though, so callers of this function have to be able to specify @filename and | |
418 | * @flags. | |
419 | */ | |
efaa7c4e HR |
420 | BlockBackend *blk_new_open(const char *filename, const char *reference, |
421 | QDict *options, int flags, Error **errp) | |
ca49a4fd HR |
422 | { |
423 | BlockBackend *blk; | |
28eb9b12 | 424 | BlockDriverState *bs; |
1f4ad7d3 | 425 | uint64_t perm = 0; |
35b7f4ab | 426 | uint64_t shared = BLK_PERM_ALL; |
c62d32f5 | 427 | |
0439c5a4 EGE |
428 | GLOBAL_STATE_CODE(); |
429 | ||
35b7f4ab KW |
430 | /* |
431 | * blk_new_open() is mainly used in .bdrv_create implementations and the | |
432 | * tools where sharing isn't a major concern because the BDS stays private | |
433 | * and the file is generally not supposed to be used by a second process, | |
434 | * so we just request permission according to the flags. | |
c62d32f5 KW |
435 | * |
436 | * The exceptions are xen_disk and blockdev_init(); in these cases, the | |
437 | * caller of blk_new_open() doesn't make use of the permissions, but they | |
438 | * shouldn't hurt either. We can still share everything here because the | |
35b7f4ab KW |
439 | * guest devices will add their own blockers if they can't share. |
440 | */ | |
1f4ad7d3 KW |
441 | if ((flags & BDRV_O_NO_IO) == 0) { |
442 | perm |= BLK_PERM_CONSISTENT_READ; | |
443 | if (flags & BDRV_O_RDWR) { | |
444 | perm |= BLK_PERM_WRITE; | |
445 | } | |
c62d32f5 KW |
446 | } |
447 | if (flags & BDRV_O_RESIZE) { | |
448 | perm |= BLK_PERM_RESIZE; | |
449 | } | |
35b7f4ab KW |
450 | if (flags & BDRV_O_NO_SHARE) { |
451 | shared = BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED; | |
452 | } | |
ca49a4fd | 453 | |
35b7f4ab | 454 | blk = blk_new(qemu_get_aio_context(), perm, shared); |
5b363937 HR |
455 | bs = bdrv_open(filename, reference, options, flags, errp); |
456 | if (!bs) { | |
ca49a4fd HR |
457 | blk_unref(blk); |
458 | return NULL; | |
459 | } | |
460 | ||
1f38f04e HR |
461 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
462 | BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY, | |
35b7f4ab | 463 | perm, shared, blk, errp); |
50bfbe93 | 464 | if (!blk->root) { |
50bfbe93 FZ |
465 | blk_unref(blk); |
466 | return NULL; | |
467 | } | |
72e775c7 | 468 | |
ca49a4fd HR |
469 | return blk; |
470 | } | |
471 | ||
26f54e9a MA |
472 | static void blk_delete(BlockBackend *blk) |
473 | { | |
474 | assert(!blk->refcnt); | |
e5e78550 | 475 | assert(!blk->name); |
a7f53e26 | 476 | assert(!blk->dev); |
022cdc9f | 477 | if (blk->public.throttle_group_member.throttle_state) { |
1606e4cf EB |
478 | blk_io_limits_disable(blk); |
479 | } | |
f21d96d0 | 480 | if (blk->root) { |
13855c6b | 481 | blk_remove_bs(blk); |
7e7d56d9 | 482 | } |
5f7772c4 FZ |
483 | if (blk->vmsh) { |
484 | qemu_del_vm_change_state_handler(blk->vmsh); | |
485 | blk->vmsh = NULL; | |
486 | } | |
3301f6c6 HR |
487 | assert(QLIST_EMPTY(&blk->remove_bs_notifiers.notifiers)); |
488 | assert(QLIST_EMPTY(&blk->insert_bs_notifiers.notifiers)); | |
d03654ea | 489 | assert(QLIST_EMPTY(&blk->aio_notifiers)); |
407ae2ae SH |
490 | assert(qemu_co_queue_empty(&blk->queued_requests)); |
491 | qemu_mutex_destroy(&blk->queued_requests_lock); | |
2cf22d6a | 492 | QTAILQ_REMOVE(&block_backends, blk, link); |
18e46a03 | 493 | drive_info_del(blk->legacy_dinfo); |
979e9b03 | 494 | block_acct_cleanup(&blk->stats); |
26f54e9a MA |
495 | g_free(blk); |
496 | } | |
497 | ||
8fb3c76c MA |
498 | static void drive_info_del(DriveInfo *dinfo) |
499 | { | |
500 | if (!dinfo) { | |
501 | return; | |
502 | } | |
503 | qemu_opts_del(dinfo->opts); | |
8fb3c76c MA |
504 | g_free(dinfo); |
505 | } | |
506 | ||
f636ae85 AG |
507 | int blk_get_refcnt(BlockBackend *blk) |
508 | { | |
0439c5a4 | 509 | GLOBAL_STATE_CODE(); |
f636ae85 AG |
510 | return blk ? blk->refcnt : 0; |
511 | } | |
512 | ||
26f54e9a MA |
513 | /* |
514 | * Increment @blk's reference count. | |
515 | * @blk must not be null. | |
516 | */ | |
517 | void blk_ref(BlockBackend *blk) | |
518 | { | |
5ca9d21b | 519 | assert(blk->refcnt > 0); |
0439c5a4 | 520 | GLOBAL_STATE_CODE(); |
26f54e9a MA |
521 | blk->refcnt++; |
522 | } | |
523 | ||
524 | /* | |
525 | * Decrement @blk's reference count. | |
526 | * If this drops it to zero, destroy @blk. | |
527 | * For convenience, do nothing if @blk is null. | |
528 | */ | |
529 | void blk_unref(BlockBackend *blk) | |
530 | { | |
0439c5a4 | 531 | GLOBAL_STATE_CODE(); |
26f54e9a MA |
532 | if (blk) { |
533 | assert(blk->refcnt > 0); | |
5ca9d21b KW |
534 | if (blk->refcnt > 1) { |
535 | blk->refcnt--; | |
536 | } else { | |
537 | blk_drain(blk); | |
538 | /* blk_drain() cannot resurrect blk, nobody held a reference */ | |
539 | assert(blk->refcnt == 1); | |
540 | blk->refcnt = 0; | |
26f54e9a MA |
541 | blk_delete(blk); |
542 | } | |
543 | } | |
544 | } | |
545 | ||
2cf22d6a HR |
546 | /* |
547 | * Behaves similarly to blk_next() but iterates over all BlockBackends, even the | |
548 | * ones which are hidden (i.e. are not referenced by the monitor). | |
549 | */ | |
a429b9b5 | 550 | BlockBackend *blk_all_next(BlockBackend *blk) |
2cf22d6a | 551 | { |
0439c5a4 | 552 | GLOBAL_STATE_CODE(); |
2cf22d6a HR |
553 | return blk ? QTAILQ_NEXT(blk, link) |
554 | : QTAILQ_FIRST(&block_backends); | |
555 | } | |
556 | ||
d8da3cef HR |
557 | void blk_remove_all_bs(void) |
558 | { | |
74d1b8fc | 559 | BlockBackend *blk = NULL; |
d8da3cef | 560 | |
0439c5a4 EGE |
561 | GLOBAL_STATE_CODE(); |
562 | ||
2cf22d6a | 563 | while ((blk = blk_all_next(blk)) != NULL) { |
d8da3cef HR |
564 | AioContext *ctx = blk_get_aio_context(blk); |
565 | ||
566 | aio_context_acquire(ctx); | |
f21d96d0 | 567 | if (blk->root) { |
d8da3cef HR |
568 | blk_remove_bs(blk); |
569 | } | |
570 | aio_context_release(ctx); | |
571 | } | |
572 | } | |
573 | ||
26f54e9a | 574 | /* |
9492b0b9 | 575 | * Return the monitor-owned BlockBackend after @blk. |
26f54e9a MA |
576 | * If @blk is null, return the first one. |
577 | * Else, return @blk's next sibling, which may be null. | |
578 | * | |
579 | * To iterate over all BlockBackends, do | |
580 | * for (blk = blk_next(NULL); blk; blk = blk_next(blk)) { | |
581 | * ... | |
582 | * } | |
583 | */ | |
584 | BlockBackend *blk_next(BlockBackend *blk) | |
585 | { | |
0439c5a4 | 586 | GLOBAL_STATE_CODE(); |
9492b0b9 HR |
587 | return blk ? QTAILQ_NEXT(blk, monitor_link) |
588 | : QTAILQ_FIRST(&monitor_block_backends); | |
26f54e9a MA |
589 | } |
590 | ||
7c8eece4 KW |
591 | /* Iterates over all top-level BlockDriverStates, i.e. BDSs that are owned by |
592 | * the monitor or attached to a BlockBackend */ | |
88be7b4b | 593 | BlockDriverState *bdrv_next(BdrvNextIterator *it) |
7c8eece4 | 594 | { |
5e003f17 HR |
595 | BlockDriverState *bs, *old_bs; |
596 | ||
597 | /* Must be called from the main loop */ | |
598 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
981f4f57 | 599 | |
7c8eece4 KW |
600 | /* First, return all root nodes of BlockBackends. In order to avoid |
601 | * returning a BDS twice when multiple BBs refer to it, we only return it | |
602 | * if the BB is the first one in the parent list of the BDS. */ | |
603 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
5e003f17 HR |
604 | BlockBackend *old_blk = it->blk; |
605 | ||
606 | old_bs = old_blk ? blk_bs(old_blk) : NULL; | |
607 | ||
7c8eece4 KW |
608 | do { |
609 | it->blk = blk_all_next(it->blk); | |
88be7b4b KW |
610 | bs = it->blk ? blk_bs(it->blk) : NULL; |
611 | } while (it->blk && (bs == NULL || bdrv_first_blk(bs) != it->blk)); | |
7c8eece4 | 612 | |
5e003f17 HR |
613 | if (it->blk) { |
614 | blk_ref(it->blk); | |
615 | } | |
616 | blk_unref(old_blk); | |
617 | ||
88be7b4b | 618 | if (bs) { |
5e003f17 HR |
619 | bdrv_ref(bs); |
620 | bdrv_unref(old_bs); | |
88be7b4b | 621 | return bs; |
7c8eece4 KW |
622 | } |
623 | it->phase = BDRV_NEXT_MONITOR_OWNED; | |
5e003f17 HR |
624 | } else { |
625 | old_bs = it->bs; | |
7c8eece4 KW |
626 | } |
627 | ||
628 | /* Then return the monitor-owned BDSes without a BB attached. Ignore all | |
629 | * BDSes that are attached to a BlockBackend here; they have been handled | |
630 | * by the above block already */ | |
981f4f57 | 631 | do { |
7c8eece4 | 632 | it->bs = bdrv_next_monitor_owned(it->bs); |
88be7b4b KW |
633 | bs = it->bs; |
634 | } while (bs && bdrv_has_blk(bs)); | |
635 | ||
5e003f17 HR |
636 | if (bs) { |
637 | bdrv_ref(bs); | |
638 | } | |
639 | bdrv_unref(old_bs); | |
640 | ||
88be7b4b KW |
641 | return bs; |
642 | } | |
643 | ||
5e003f17 | 644 | static void bdrv_next_reset(BdrvNextIterator *it) |
88be7b4b KW |
645 | { |
646 | *it = (BdrvNextIterator) { | |
647 | .phase = BDRV_NEXT_BACKEND_ROOTS, | |
648 | }; | |
5e003f17 | 649 | } |
981f4f57 | 650 | |
5e003f17 HR |
651 | BlockDriverState *bdrv_first(BdrvNextIterator *it) |
652 | { | |
0439c5a4 | 653 | GLOBAL_STATE_CODE(); |
5e003f17 | 654 | bdrv_next_reset(it); |
88be7b4b | 655 | return bdrv_next(it); |
981f4f57 HR |
656 | } |
657 | ||
5e003f17 HR |
658 | /* Must be called when aborting a bdrv_next() iteration before |
659 | * bdrv_next() returns NULL */ | |
660 | void bdrv_next_cleanup(BdrvNextIterator *it) | |
661 | { | |
662 | /* Must be called from the main loop */ | |
663 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
664 | ||
665 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
666 | if (it->blk) { | |
667 | bdrv_unref(blk_bs(it->blk)); | |
668 | blk_unref(it->blk); | |
669 | } | |
670 | } else { | |
671 | bdrv_unref(it->bs); | |
672 | } | |
673 | ||
674 | bdrv_next_reset(it); | |
675 | } | |
676 | ||
e5e78550 HR |
677 | /* |
678 | * Add a BlockBackend into the list of backends referenced by the monitor, with | |
679 | * the given @name acting as the handle for the monitor. | |
680 | * Strictly for use by blockdev.c. | |
681 | * | |
682 | * @name must not be null or empty. | |
683 | * | |
684 | * Returns true on success and false on failure. In the latter case, an Error | |
685 | * object is returned through @errp. | |
686 | */ | |
687 | bool monitor_add_blk(BlockBackend *blk, const char *name, Error **errp) | |
688 | { | |
689 | assert(!blk->name); | |
690 | assert(name && name[0]); | |
0439c5a4 | 691 | GLOBAL_STATE_CODE(); |
e5e78550 HR |
692 | |
693 | if (!id_wellformed(name)) { | |
694 | error_setg(errp, "Invalid device name"); | |
695 | return false; | |
696 | } | |
697 | if (blk_by_name(name)) { | |
698 | error_setg(errp, "Device with id '%s' already exists", name); | |
699 | return false; | |
700 | } | |
701 | if (bdrv_find_node(name)) { | |
702 | error_setg(errp, | |
703 | "Device name '%s' conflicts with an existing node name", | |
704 | name); | |
705 | return false; | |
706 | } | |
707 | ||
708 | blk->name = g_strdup(name); | |
709 | QTAILQ_INSERT_TAIL(&monitor_block_backends, blk, monitor_link); | |
710 | return true; | |
711 | } | |
712 | ||
713 | /* | |
714 | * Remove a BlockBackend from the list of backends referenced by the monitor. | |
715 | * Strictly for use by blockdev.c. | |
716 | */ | |
717 | void monitor_remove_blk(BlockBackend *blk) | |
718 | { | |
0439c5a4 EGE |
719 | GLOBAL_STATE_CODE(); |
720 | ||
e5e78550 HR |
721 | if (!blk->name) { |
722 | return; | |
723 | } | |
724 | ||
725 | QTAILQ_REMOVE(&monitor_block_backends, blk, monitor_link); | |
726 | g_free(blk->name); | |
727 | blk->name = NULL; | |
728 | } | |
729 | ||
26f54e9a | 730 | /* |
7e7d56d9 | 731 | * Return @blk's name, a non-null string. |
e5e78550 | 732 | * Returns an empty string iff @blk is not referenced by the monitor. |
26f54e9a | 733 | */ |
0731a50f | 734 | const char *blk_name(const BlockBackend *blk) |
26f54e9a | 735 | { |
37868b2a | 736 | IO_CODE(); |
e5e78550 | 737 | return blk->name ?: ""; |
26f54e9a MA |
738 | } |
739 | ||
740 | /* | |
741 | * Return the BlockBackend with name @name if it exists, else null. | |
742 | * @name must not be null. | |
743 | */ | |
744 | BlockBackend *blk_by_name(const char *name) | |
745 | { | |
74d1b8fc | 746 | BlockBackend *blk = NULL; |
26f54e9a | 747 | |
0439c5a4 | 748 | GLOBAL_STATE_CODE(); |
26f54e9a | 749 | assert(name); |
74d1b8fc | 750 | while ((blk = blk_next(blk)) != NULL) { |
26f54e9a MA |
751 | if (!strcmp(name, blk->name)) { |
752 | return blk; | |
753 | } | |
754 | } | |
755 | return NULL; | |
756 | } | |
7e7d56d9 MA |
757 | |
758 | /* | |
759 | * Return the BlockDriverState attached to @blk if any, else null. | |
760 | */ | |
761 | BlockDriverState *blk_bs(BlockBackend *blk) | |
762 | { | |
37868b2a | 763 | IO_CODE(); |
f21d96d0 | 764 | return blk->root ? blk->root->bs : NULL; |
7e7d56d9 MA |
765 | } |
766 | ||
7c8eece4 | 767 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs) |
dde33812 KW |
768 | { |
769 | BdrvChild *child; | |
bdb73476 EGE |
770 | |
771 | GLOBAL_STATE_CODE(); | |
772 | ||
dde33812 | 773 | QLIST_FOREACH(child, &bs->parents, next_parent) { |
bd86fb99 | 774 | if (child->klass == &child_root) { |
7c8eece4 | 775 | return child->opaque; |
dde33812 KW |
776 | } |
777 | } | |
778 | ||
7c8eece4 KW |
779 | return NULL; |
780 | } | |
781 | ||
782 | /* | |
783 | * Returns true if @bs has an associated BlockBackend. | |
784 | */ | |
785 | bool bdrv_has_blk(BlockDriverState *bs) | |
786 | { | |
0439c5a4 | 787 | GLOBAL_STATE_CODE(); |
7c8eece4 | 788 | return bdrv_first_blk(bs) != NULL; |
dde33812 KW |
789 | } |
790 | ||
b6c1bae5 KW |
791 | /* |
792 | * Returns true if @bs has only BlockBackends as parents. | |
793 | */ | |
794 | bool bdrv_is_root_node(BlockDriverState *bs) | |
795 | { | |
796 | BdrvChild *c; | |
797 | ||
0439c5a4 | 798 | GLOBAL_STATE_CODE(); |
b6c1bae5 | 799 | QLIST_FOREACH(c, &bs->parents, next_parent) { |
bd86fb99 | 800 | if (c->klass != &child_root) { |
b6c1bae5 KW |
801 | return false; |
802 | } | |
803 | } | |
804 | ||
805 | return true; | |
806 | } | |
807 | ||
18e46a03 MA |
808 | /* |
809 | * Return @blk's DriveInfo if any, else null. | |
810 | */ | |
811 | DriveInfo *blk_legacy_dinfo(BlockBackend *blk) | |
812 | { | |
c5be7445 | 813 | GLOBAL_STATE_CODE(); |
18e46a03 MA |
814 | return blk->legacy_dinfo; |
815 | } | |
816 | ||
817 | /* | |
818 | * Set @blk's DriveInfo to @dinfo, and return it. | |
819 | * @blk must not have a DriveInfo set already. | |
820 | * No other BlockBackend may have the same DriveInfo set. | |
821 | */ | |
822 | DriveInfo *blk_set_legacy_dinfo(BlockBackend *blk, DriveInfo *dinfo) | |
823 | { | |
824 | assert(!blk->legacy_dinfo); | |
c5be7445 | 825 | GLOBAL_STATE_CODE(); |
18e46a03 MA |
826 | return blk->legacy_dinfo = dinfo; |
827 | } | |
828 | ||
829 | /* | |
830 | * Return the BlockBackend with DriveInfo @dinfo. | |
831 | * It must exist. | |
832 | */ | |
833 | BlockBackend *blk_by_legacy_dinfo(DriveInfo *dinfo) | |
834 | { | |
74d1b8fc | 835 | BlockBackend *blk = NULL; |
c5be7445 | 836 | GLOBAL_STATE_CODE(); |
18e46a03 | 837 | |
74d1b8fc | 838 | while ((blk = blk_next(blk)) != NULL) { |
18e46a03 MA |
839 | if (blk->legacy_dinfo == dinfo) { |
840 | return blk; | |
841 | } | |
842 | } | |
843 | abort(); | |
844 | } | |
845 | ||
f2cd875d KW |
846 | /* |
847 | * Returns a pointer to the publicly accessible fields of @blk. | |
848 | */ | |
849 | BlockBackendPublic *blk_get_public(BlockBackend *blk) | |
850 | { | |
0439c5a4 | 851 | GLOBAL_STATE_CODE(); |
f2cd875d KW |
852 | return &blk->public; |
853 | } | |
854 | ||
855 | /* | |
856 | * Returns a BlockBackend given the associated @public fields. | |
857 | */ | |
858 | BlockBackend *blk_by_public(BlockBackendPublic *public) | |
859 | { | |
0439c5a4 | 860 | GLOBAL_STATE_CODE(); |
f2cd875d KW |
861 | return container_of(public, BlockBackend, public); |
862 | } | |
863 | ||
1c95f7e1 HR |
864 | /* |
865 | * Disassociates the currently associated BlockDriverState from @blk. | |
866 | */ | |
867 | void blk_remove_bs(BlockBackend *blk) | |
868 | { | |
c89bcf3a | 869 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
e6cada92 | 870 | BdrvChild *root; |
022cdc9f | 871 | |
0439c5a4 EGE |
872 | GLOBAL_STATE_CODE(); |
873 | ||
3301f6c6 | 874 | notifier_list_notify(&blk->remove_bs_notifiers, blk); |
c89bcf3a | 875 | if (tgm->throttle_state) { |
1e3552db SH |
876 | BlockDriverState *bs = blk_bs(blk); |
877 | ||
878 | /* | |
879 | * Take a ref in case blk_bs() changes across bdrv_drained_begin(), for | |
880 | * example, if a temporary filter node is removed by a blockjob. | |
881 | */ | |
882 | bdrv_ref(bs); | |
632a7735 | 883 | bdrv_drained_begin(bs); |
c89bcf3a AG |
884 | throttle_group_detach_aio_context(tgm); |
885 | throttle_group_attach_aio_context(tgm, qemu_get_aio_context()); | |
632a7735 | 886 | bdrv_drained_end(bs); |
1e3552db | 887 | bdrv_unref(bs); |
a5614993 | 888 | } |
1c95f7e1 | 889 | |
7ca7f0f6 KW |
890 | blk_update_root_state(blk); |
891 | ||
f45280cb GK |
892 | /* bdrv_root_unref_child() will cause blk->root to become stale and may |
893 | * switch to a completion coroutine later on. Let's drain all I/O here | |
894 | * to avoid that and a potential QEMU crash. | |
895 | */ | |
896 | blk_drain(blk); | |
e6cada92 | 897 | root = blk->root; |
f21d96d0 | 898 | blk->root = NULL; |
e6cada92 | 899 | bdrv_root_unref_child(root); |
1c95f7e1 HR |
900 | } |
901 | ||
0c3c36d6 HR |
902 | /* |
903 | * Associates a new BlockDriverState with @blk. | |
904 | */ | |
d7086422 | 905 | int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) |
0c3c36d6 | 906 | { |
c89bcf3a | 907 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
0439c5a4 | 908 | GLOBAL_STATE_CODE(); |
b441dc71 | 909 | bdrv_ref(bs); |
1f38f04e HR |
910 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
911 | BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY, | |
228ca37e | 912 | blk->perm, blk->shared_perm, |
1f38f04e | 913 | blk, errp); |
d7086422 KW |
914 | if (blk->root == NULL) { |
915 | return -EPERM; | |
916 | } | |
3301f6c6 HR |
917 | |
918 | notifier_list_notify(&blk->insert_bs_notifiers, blk); | |
c89bcf3a AG |
919 | if (tgm->throttle_state) { |
920 | throttle_group_detach_aio_context(tgm); | |
921 | throttle_group_attach_aio_context(tgm, bdrv_get_aio_context(bs)); | |
7ca7f0f6 | 922 | } |
d7086422 KW |
923 | |
924 | return 0; | |
0c3c36d6 HR |
925 | } |
926 | ||
ed089506 VSO |
927 | /* |
928 | * Change BlockDriverState associated with @blk. | |
929 | */ | |
930 | int blk_replace_bs(BlockBackend *blk, BlockDriverState *new_bs, Error **errp) | |
931 | { | |
0439c5a4 | 932 | GLOBAL_STATE_CODE(); |
ed089506 VSO |
933 | return bdrv_replace_child_bs(blk->root, new_bs, errp); |
934 | } | |
935 | ||
981776b3 KW |
936 | /* |
937 | * Sets the permission bitmasks that the user of the BlockBackend needs. | |
938 | */ | |
939 | int blk_set_perm(BlockBackend *blk, uint64_t perm, uint64_t shared_perm, | |
940 | Error **errp) | |
941 | { | |
942 | int ret; | |
0439c5a4 | 943 | GLOBAL_STATE_CODE(); |
981776b3 | 944 | |
d35ff5e6 | 945 | if (blk->root && !blk->disable_perm) { |
981776b3 KW |
946 | ret = bdrv_child_try_set_perm(blk->root, perm, shared_perm, errp); |
947 | if (ret < 0) { | |
948 | return ret; | |
949 | } | |
950 | } | |
951 | ||
952 | blk->perm = perm; | |
953 | blk->shared_perm = shared_perm; | |
954 | ||
955 | return 0; | |
956 | } | |
957 | ||
887354bd KW |
958 | void blk_get_perm(BlockBackend *blk, uint64_t *perm, uint64_t *shared_perm) |
959 | { | |
0439c5a4 | 960 | GLOBAL_STATE_CODE(); |
887354bd KW |
961 | *perm = blk->perm; |
962 | *shared_perm = blk->shared_perm; | |
963 | } | |
964 | ||
d09ea2d2 TH |
965 | /* |
966 | * Attach device model @dev to @blk. | |
967 | * Return 0 on success, -EBUSY when a device model is attached already. | |
968 | */ | |
969 | int blk_attach_dev(BlockBackend *blk, DeviceState *dev) | |
4be74634 | 970 | { |
0439c5a4 | 971 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
972 | if (blk->dev) { |
973 | return -EBUSY; | |
974 | } | |
d35ff5e6 KW |
975 | |
976 | /* While migration is still incoming, we don't need to apply the | |
977 | * permissions of guest device BlockBackends. We might still have a block | |
978 | * job or NBD server writing to the image for storage migration. */ | |
979 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
980 | blk->disable_perm = true; | |
981 | } | |
982 | ||
84ebe375 | 983 | blk_ref(blk); |
a7f53e26 | 984 | blk->dev = dev; |
373340b2 | 985 | blk_iostatus_reset(blk); |
d35ff5e6 | 986 | |
a7f53e26 | 987 | return 0; |
4be74634 MA |
988 | } |
989 | ||
a7f53e26 MA |
990 | /* |
991 | * Detach device model @dev from @blk. | |
992 | * @dev must be currently attached to @blk. | |
993 | */ | |
d09ea2d2 | 994 | void blk_detach_dev(BlockBackend *blk, DeviceState *dev) |
4be74634 | 995 | { |
a7f53e26 | 996 | assert(blk->dev == dev); |
0439c5a4 | 997 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
998 | blk->dev = NULL; |
999 | blk->dev_ops = NULL; | |
1000 | blk->dev_opaque = NULL; | |
981776b3 | 1001 | blk_set_perm(blk, 0, BLK_PERM_ALL, &error_abort); |
84ebe375 | 1002 | blk_unref(blk); |
4be74634 MA |
1003 | } |
1004 | ||
a7f53e26 MA |
1005 | /* |
1006 | * Return the device model attached to @blk if any, else null. | |
1007 | */ | |
d09ea2d2 | 1008 | DeviceState *blk_get_attached_dev(BlockBackend *blk) |
a7f53e26 | 1009 | { |
0439c5a4 | 1010 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
1011 | return blk->dev; |
1012 | } | |
1013 | ||
2d76e724 KW |
1014 | /* Return the qdev ID, or if no ID is assigned the QOM path, of the block |
1015 | * device attached to the BlockBackend. */ | |
77beef83 | 1016 | char *blk_get_attached_dev_id(BlockBackend *blk) |
2d76e724 | 1017 | { |
d09ea2d2 | 1018 | DeviceState *dev = blk->dev; |
37868b2a | 1019 | IO_CODE(); |
2d76e724 KW |
1020 | |
1021 | if (!dev) { | |
1022 | return g_strdup(""); | |
1023 | } else if (dev->id) { | |
1024 | return g_strdup(dev->id); | |
1025 | } | |
602414d1 LM |
1026 | |
1027 | return object_get_canonical_path(OBJECT(dev)) ?: g_strdup(""); | |
2d76e724 KW |
1028 | } |
1029 | ||
1c89e1fa KW |
1030 | /* |
1031 | * Return the BlockBackend which has the device model @dev attached if it | |
1032 | * exists, else null. | |
1033 | * | |
1034 | * @dev must not be null. | |
1035 | */ | |
1036 | BlockBackend *blk_by_dev(void *dev) | |
1037 | { | |
1038 | BlockBackend *blk = NULL; | |
1039 | ||
0439c5a4 EGE |
1040 | GLOBAL_STATE_CODE(); |
1041 | ||
1c89e1fa KW |
1042 | assert(dev != NULL); |
1043 | while ((blk = blk_all_next(blk)) != NULL) { | |
1044 | if (blk->dev == dev) { | |
1045 | return blk; | |
1046 | } | |
1047 | } | |
1048 | return NULL; | |
1049 | } | |
1050 | ||
a7f53e26 MA |
1051 | /* |
1052 | * Set @blk's device model callbacks to @ops. | |
1053 | * @opaque is the opaque argument to pass to the callbacks. | |
1054 | * This is for use by device models. | |
1055 | */ | |
1056 | void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, | |
1057 | void *opaque) | |
1058 | { | |
0439c5a4 | 1059 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
1060 | blk->dev_ops = ops; |
1061 | blk->dev_opaque = opaque; | |
f4d9cc88 JS |
1062 | |
1063 | /* Are we currently quiesced? Should we enforce this right now? */ | |
c4d5bf99 | 1064 | if (qatomic_read(&blk->quiesce_counter) && ops && ops->drained_begin) { |
f4d9cc88 JS |
1065 | ops->drained_begin(opaque); |
1066 | } | |
a7f53e26 MA |
1067 | } |
1068 | ||
1069 | /* | |
1070 | * Notify @blk's attached device model of media change. | |
39829a01 KW |
1071 | * |
1072 | * If @load is true, notify of media load. This action can fail, meaning that | |
1073 | * the medium cannot be loaded. @errp is set then. | |
1074 | * | |
1075 | * If @load is false, notify of media eject. This can never fail. | |
1076 | * | |
a7f53e26 MA |
1077 | * Also send DEVICE_TRAY_MOVED events as appropriate. |
1078 | */ | |
39829a01 | 1079 | void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp) |
a7f53e26 | 1080 | { |
0439c5a4 | 1081 | GLOBAL_STATE_CODE(); |
a7f53e26 | 1082 | if (blk->dev_ops && blk->dev_ops->change_media_cb) { |
f1f57066 | 1083 | bool tray_was_open, tray_is_open; |
39829a01 | 1084 | Error *local_err = NULL; |
a7f53e26 | 1085 | |
f1f57066 | 1086 | tray_was_open = blk_dev_is_tray_open(blk); |
39829a01 KW |
1087 | blk->dev_ops->change_media_cb(blk->dev_opaque, load, &local_err); |
1088 | if (local_err) { | |
1089 | assert(load == true); | |
1090 | error_propagate(errp, local_err); | |
1091 | return; | |
1092 | } | |
f1f57066 HR |
1093 | tray_is_open = blk_dev_is_tray_open(blk); |
1094 | ||
1095 | if (tray_was_open != tray_is_open) { | |
2d76e724 | 1096 | char *id = blk_get_attached_dev_id(blk); |
3ab72385 | 1097 | qapi_event_send_device_tray_moved(blk_name(blk), id, tray_is_open); |
2d76e724 | 1098 | g_free(id); |
a7f53e26 MA |
1099 | } |
1100 | } | |
1101 | } | |
1102 | ||
5c8cab48 KW |
1103 | static void blk_root_change_media(BdrvChild *child, bool load) |
1104 | { | |
39829a01 | 1105 | blk_dev_change_media_cb(child->opaque, load, NULL); |
5c8cab48 KW |
1106 | } |
1107 | ||
a7f53e26 MA |
1108 | /* |
1109 | * Does @blk's attached device model have removable media? | |
1110 | * %true if no device model is attached. | |
1111 | */ | |
1112 | bool blk_dev_has_removable_media(BlockBackend *blk) | |
1113 | { | |
b4ad82aa | 1114 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
1115 | return !blk->dev || (blk->dev_ops && blk->dev_ops->change_media_cb); |
1116 | } | |
1117 | ||
8f3a73bc HR |
1118 | /* |
1119 | * Does @blk's attached device model have a tray? | |
1120 | */ | |
1121 | bool blk_dev_has_tray(BlockBackend *blk) | |
1122 | { | |
967d7905 | 1123 | IO_CODE(); |
8f3a73bc HR |
1124 | return blk->dev_ops && blk->dev_ops->is_tray_open; |
1125 | } | |
1126 | ||
a7f53e26 MA |
1127 | /* |
1128 | * Notify @blk's attached device model of a media eject request. | |
1129 | * If @force is true, the medium is about to be yanked out forcefully. | |
1130 | */ | |
1131 | void blk_dev_eject_request(BlockBackend *blk, bool force) | |
4be74634 | 1132 | { |
b4ad82aa | 1133 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
1134 | if (blk->dev_ops && blk->dev_ops->eject_request_cb) { |
1135 | blk->dev_ops->eject_request_cb(blk->dev_opaque, force); | |
1136 | } | |
4be74634 MA |
1137 | } |
1138 | ||
a7f53e26 MA |
1139 | /* |
1140 | * Does @blk's attached device model have a tray, and is it open? | |
1141 | */ | |
1142 | bool blk_dev_is_tray_open(BlockBackend *blk) | |
4be74634 | 1143 | { |
967d7905 | 1144 | IO_CODE(); |
8f3a73bc | 1145 | if (blk_dev_has_tray(blk)) { |
a7f53e26 MA |
1146 | return blk->dev_ops->is_tray_open(blk->dev_opaque); |
1147 | } | |
1148 | return false; | |
1149 | } | |
1150 | ||
1151 | /* | |
1152 | * Does @blk's attached device model have the medium locked? | |
1153 | * %false if the device model has no such lock. | |
1154 | */ | |
1155 | bool blk_dev_is_medium_locked(BlockBackend *blk) | |
1156 | { | |
b4ad82aa | 1157 | GLOBAL_STATE_CODE(); |
a7f53e26 MA |
1158 | if (blk->dev_ops && blk->dev_ops->is_medium_locked) { |
1159 | return blk->dev_ops->is_medium_locked(blk->dev_opaque); | |
1160 | } | |
1161 | return false; | |
1162 | } | |
1163 | ||
1164 | /* | |
1165 | * Notify @blk's attached device model of a backend size change. | |
1166 | */ | |
5c8cab48 | 1167 | static void blk_root_resize(BdrvChild *child) |
a7f53e26 | 1168 | { |
5c8cab48 KW |
1169 | BlockBackend *blk = child->opaque; |
1170 | ||
a7f53e26 MA |
1171 | if (blk->dev_ops && blk->dev_ops->resize_cb) { |
1172 | blk->dev_ops->resize_cb(blk->dev_opaque); | |
1173 | } | |
1174 | } | |
1175 | ||
1176 | void blk_iostatus_enable(BlockBackend *blk) | |
1177 | { | |
0439c5a4 | 1178 | GLOBAL_STATE_CODE(); |
373340b2 HR |
1179 | blk->iostatus_enabled = true; |
1180 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
1181 | } | |
1182 | ||
1183 | /* The I/O status is only enabled if the drive explicitly | |
1184 | * enables it _and_ the VM is configured to stop on errors */ | |
1185 | bool blk_iostatus_is_enabled(const BlockBackend *blk) | |
1186 | { | |
37868b2a | 1187 | IO_CODE(); |
373340b2 HR |
1188 | return (blk->iostatus_enabled && |
1189 | (blk->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || | |
1190 | blk->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
1191 | blk->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
1192 | } | |
1193 | ||
1194 | BlockDeviceIoStatus blk_iostatus(const BlockBackend *blk) | |
1195 | { | |
0439c5a4 | 1196 | GLOBAL_STATE_CODE(); |
373340b2 HR |
1197 | return blk->iostatus; |
1198 | } | |
1199 | ||
1200 | void blk_iostatus_disable(BlockBackend *blk) | |
1201 | { | |
0439c5a4 | 1202 | GLOBAL_STATE_CODE(); |
373340b2 HR |
1203 | blk->iostatus_enabled = false; |
1204 | } | |
1205 | ||
1206 | void blk_iostatus_reset(BlockBackend *blk) | |
1207 | { | |
0439c5a4 | 1208 | GLOBAL_STATE_CODE(); |
373340b2 HR |
1209 | if (blk_iostatus_is_enabled(blk)) { |
1210 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
373340b2 HR |
1211 | } |
1212 | } | |
1213 | ||
1214 | void blk_iostatus_set_err(BlockBackend *blk, int error) | |
1215 | { | |
37868b2a | 1216 | IO_CODE(); |
373340b2 HR |
1217 | assert(blk_iostatus_is_enabled(blk)); |
1218 | if (blk->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
1219 | blk->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : | |
1220 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
1221 | } | |
4be74634 MA |
1222 | } |
1223 | ||
c10c9d96 KW |
1224 | void blk_set_allow_write_beyond_eof(BlockBackend *blk, bool allow) |
1225 | { | |
37868b2a | 1226 | IO_CODE(); |
c10c9d96 KW |
1227 | blk->allow_write_beyond_eof = allow; |
1228 | } | |
1229 | ||
980b0f94 KW |
1230 | void blk_set_allow_aio_context_change(BlockBackend *blk, bool allow) |
1231 | { | |
37868b2a | 1232 | IO_CODE(); |
980b0f94 KW |
1233 | blk->allow_aio_context_change = allow; |
1234 | } | |
1235 | ||
cf312932 KW |
1236 | void blk_set_disable_request_queuing(BlockBackend *blk, bool disable) |
1237 | { | |
37868b2a | 1238 | IO_CODE(); |
ef80ec50 | 1239 | qatomic_set(&blk->disable_request_queuing, disable); |
cf312932 KW |
1240 | } |
1241 | ||
c73ff92c EGE |
1242 | static int coroutine_fn GRAPH_RDLOCK |
1243 | blk_check_byte_request(BlockBackend *blk, int64_t offset, int64_t bytes) | |
e7f7d676 HR |
1244 | { |
1245 | int64_t len; | |
1246 | ||
aa78b825 | 1247 | if (bytes < 0) { |
e7f7d676 HR |
1248 | return -EIO; |
1249 | } | |
1250 | ||
c73ff92c | 1251 | if (!blk_co_is_available(blk)) { |
e7f7d676 HR |
1252 | return -ENOMEDIUM; |
1253 | } | |
1254 | ||
e7f7d676 HR |
1255 | if (offset < 0) { |
1256 | return -EIO; | |
1257 | } | |
1258 | ||
c10c9d96 | 1259 | if (!blk->allow_write_beyond_eof) { |
bd53086e | 1260 | len = bdrv_co_getlength(blk_bs(blk)); |
c10c9d96 KW |
1261 | if (len < 0) { |
1262 | return len; | |
1263 | } | |
1264 | ||
7242db63 | 1265 | if (offset > len || len - offset < bytes) { |
c10c9d96 KW |
1266 | return -EIO; |
1267 | } | |
e7f7d676 HR |
1268 | } |
1269 | ||
1270 | return 0; | |
1271 | } | |
1272 | ||
7f16476f | 1273 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
cf312932 KW |
1274 | static void coroutine_fn blk_wait_while_drained(BlockBackend *blk) |
1275 | { | |
7f16476f KW |
1276 | assert(blk->in_flight > 0); |
1277 | ||
ef80ec50 SH |
1278 | if (qatomic_read(&blk->quiesce_counter) && |
1279 | !qatomic_read(&blk->disable_request_queuing)) { | |
407ae2ae SH |
1280 | /* |
1281 | * Take lock before decrementing in flight counter so main loop thread | |
1282 | * waits for us to enqueue ourselves before it can leave the drained | |
1283 | * section. | |
1284 | */ | |
1285 | qemu_mutex_lock(&blk->queued_requests_lock); | |
7f16476f | 1286 | blk_dec_in_flight(blk); |
407ae2ae | 1287 | qemu_co_queue_wait(&blk->queued_requests, &blk->queued_requests_lock); |
7f16476f | 1288 | blk_inc_in_flight(blk); |
407ae2ae | 1289 | qemu_mutex_unlock(&blk->queued_requests_lock); |
cf312932 KW |
1290 | } |
1291 | } | |
1292 | ||
fbb92b67 | 1293 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
d1d3fc3d AF |
1294 | static int coroutine_fn |
1295 | blk_co_do_preadv_part(BlockBackend *blk, int64_t offset, int64_t bytes, | |
1296 | QEMUIOVector *qiov, size_t qiov_offset, | |
1297 | BdrvRequestFlags flags) | |
4be74634 | 1298 | { |
1e98fefd | 1299 | int ret; |
cf312932 | 1300 | BlockDriverState *bs; |
1581a70d | 1301 | IO_CODE(); |
1e98fefd | 1302 | |
cf312932 | 1303 | blk_wait_while_drained(blk); |
b9b10c35 | 1304 | GRAPH_RDLOCK_GUARD(); |
1e98fefd | 1305 | |
cf312932 KW |
1306 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1307 | bs = blk_bs(blk); | |
99723548 | 1308 | trace_blk_co_preadv(blk, bs, offset, bytes, flags); |
1e98fefd KW |
1309 | |
1310 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1311 | if (ret < 0) { |
1312 | return ret; | |
1313 | } | |
1314 | ||
99723548 PB |
1315 | bdrv_inc_in_flight(bs); |
1316 | ||
441565b2 | 1317 | /* throttling disk I/O */ |
022cdc9f MP |
1318 | if (blk->public.throttle_group_member.throttle_state) { |
1319 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1320 | bytes, false); | |
441565b2 KW |
1321 | } |
1322 | ||
d1d3fc3d AF |
1323 | ret = bdrv_co_preadv_part(blk->root, offset, bytes, qiov, qiov_offset, |
1324 | flags); | |
99723548 PB |
1325 | bdrv_dec_in_flight(bs); |
1326 | return ret; | |
1bf1cbc9 KW |
1327 | } |
1328 | ||
6f675c93 AF |
1329 | int coroutine_fn blk_co_pread(BlockBackend *blk, int64_t offset, int64_t bytes, |
1330 | void *buf, BdrvRequestFlags flags) | |
1331 | { | |
1332 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); | |
1333 | IO_OR_GS_CODE(); | |
1334 | ||
1335 | assert(bytes <= SIZE_MAX); | |
1336 | ||
1337 | return blk_co_preadv(blk, offset, bytes, &qiov, flags); | |
1338 | } | |
1339 | ||
fbb92b67 | 1340 | int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset, |
95479077 | 1341 | int64_t bytes, QEMUIOVector *qiov, |
fbb92b67 KW |
1342 | BdrvRequestFlags flags) |
1343 | { | |
1344 | int ret; | |
37868b2a | 1345 | IO_OR_GS_CODE(); |
fbb92b67 KW |
1346 | |
1347 | blk_inc_in_flight(blk); | |
d1d3fc3d AF |
1348 | ret = blk_co_do_preadv_part(blk, offset, bytes, qiov, 0, flags); |
1349 | blk_dec_in_flight(blk); | |
1350 | ||
1351 | return ret; | |
1352 | } | |
1353 | ||
1354 | int coroutine_fn blk_co_preadv_part(BlockBackend *blk, int64_t offset, | |
1355 | int64_t bytes, QEMUIOVector *qiov, | |
1356 | size_t qiov_offset, BdrvRequestFlags flags) | |
1357 | { | |
1358 | int ret; | |
1359 | IO_OR_GS_CODE(); | |
1360 | ||
1361 | blk_inc_in_flight(blk); | |
1362 | ret = blk_co_do_preadv_part(blk, offset, bytes, qiov, qiov_offset, flags); | |
fbb92b67 KW |
1363 | blk_dec_in_flight(blk); |
1364 | ||
1365 | return ret; | |
1366 | } | |
1367 | ||
1368 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ | |
07a64aa4 | 1369 | static int coroutine_fn |
70e8775e VSO |
1370 | blk_co_do_pwritev_part(BlockBackend *blk, int64_t offset, int64_t bytes, |
1371 | QEMUIOVector *qiov, size_t qiov_offset, | |
1372 | BdrvRequestFlags flags) | |
a8823a3b | 1373 | { |
bfd18d1e | 1374 | int ret; |
cf312932 | 1375 | BlockDriverState *bs; |
1581a70d | 1376 | IO_CODE(); |
bfd18d1e | 1377 | |
cf312932 | 1378 | blk_wait_while_drained(blk); |
b9b10c35 | 1379 | GRAPH_RDLOCK_GUARD(); |
bfd18d1e | 1380 | |
cf312932 KW |
1381 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1382 | bs = blk_bs(blk); | |
99723548 | 1383 | trace_blk_co_pwritev(blk, bs, offset, bytes, flags); |
1e98fefd | 1384 | |
bfd18d1e | 1385 | ret = blk_check_byte_request(blk, offset, bytes); |
a8823a3b KW |
1386 | if (ret < 0) { |
1387 | return ret; | |
1388 | } | |
1389 | ||
99723548 | 1390 | bdrv_inc_in_flight(bs); |
441565b2 | 1391 | /* throttling disk I/O */ |
022cdc9f MP |
1392 | if (blk->public.throttle_group_member.throttle_state) { |
1393 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1394 | bytes, true); | |
441565b2 KW |
1395 | } |
1396 | ||
bfd18d1e KW |
1397 | if (!blk->enable_write_cache) { |
1398 | flags |= BDRV_REQ_FUA; | |
1399 | } | |
1400 | ||
b3016864 VSO |
1401 | ret = bdrv_co_pwritev_part(blk->root, offset, bytes, qiov, qiov_offset, |
1402 | flags); | |
99723548 PB |
1403 | bdrv_dec_in_flight(bs); |
1404 | return ret; | |
a8823a3b KW |
1405 | } |
1406 | ||
fbb92b67 | 1407 | int coroutine_fn blk_co_pwritev_part(BlockBackend *blk, int64_t offset, |
34460feb | 1408 | int64_t bytes, |
fbb92b67 KW |
1409 | QEMUIOVector *qiov, size_t qiov_offset, |
1410 | BdrvRequestFlags flags) | |
1411 | { | |
1412 | int ret; | |
37868b2a | 1413 | IO_OR_GS_CODE(); |
fbb92b67 KW |
1414 | |
1415 | blk_inc_in_flight(blk); | |
70e8775e | 1416 | ret = blk_co_do_pwritev_part(blk, offset, bytes, qiov, qiov_offset, flags); |
fbb92b67 KW |
1417 | blk_dec_in_flight(blk); |
1418 | ||
1419 | return ret; | |
1420 | } | |
1421 | ||
6f675c93 AF |
1422 | int coroutine_fn blk_co_pwrite(BlockBackend *blk, int64_t offset, int64_t bytes, |
1423 | const void *buf, BdrvRequestFlags flags) | |
1424 | { | |
1425 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); | |
1426 | IO_OR_GS_CODE(); | |
1427 | ||
1428 | assert(bytes <= SIZE_MAX); | |
1429 | ||
1430 | return blk_co_pwritev(blk, offset, bytes, &qiov, flags); | |
1431 | } | |
1432 | ||
b3016864 | 1433 | int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset, |
34460feb | 1434 | int64_t bytes, QEMUIOVector *qiov, |
b3016864 VSO |
1435 | BdrvRequestFlags flags) |
1436 | { | |
37868b2a | 1437 | IO_OR_GS_CODE(); |
b3016864 VSO |
1438 | return blk_co_pwritev_part(blk, offset, bytes, qiov, 0, flags); |
1439 | } | |
1440 | ||
ff7e261b EGE |
1441 | int coroutine_fn blk_co_block_status_above(BlockBackend *blk, |
1442 | BlockDriverState *base, | |
1443 | int64_t offset, int64_t bytes, | |
1444 | int64_t *pnum, int64_t *map, | |
1445 | BlockDriverState **file) | |
1446 | { | |
1447 | IO_CODE(); | |
7ff9579e | 1448 | GRAPH_RDLOCK_GUARD(); |
ff7e261b EGE |
1449 | return bdrv_co_block_status_above(blk_bs(blk), base, offset, bytes, pnum, |
1450 | map, file); | |
1451 | } | |
1452 | ||
1453 | int coroutine_fn blk_co_is_allocated_above(BlockBackend *blk, | |
1454 | BlockDriverState *base, | |
1455 | bool include_base, int64_t offset, | |
1456 | int64_t bytes, int64_t *pnum) | |
1457 | { | |
1458 | IO_CODE(); | |
7ff9579e | 1459 | GRAPH_RDLOCK_GUARD(); |
ff7e261b EGE |
1460 | return bdrv_co_is_allocated_above(blk_bs(blk), base, include_base, offset, |
1461 | bytes, pnum); | |
1462 | } | |
1463 | ||
1bf1cbc9 KW |
1464 | typedef struct BlkRwCo { |
1465 | BlockBackend *blk; | |
1466 | int64_t offset; | |
c060332c | 1467 | void *iobuf; |
1bf1cbc9 KW |
1468 | int ret; |
1469 | BdrvRequestFlags flags; | |
1470 | } BlkRwCo; | |
1471 | ||
720ff280 KW |
1472 | int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags) |
1473 | { | |
0439c5a4 | 1474 | GLOBAL_STATE_CODE(); |
720ff280 KW |
1475 | return bdrv_make_zero(blk->root, flags); |
1476 | } | |
1477 | ||
c90e2a9c | 1478 | void blk_inc_in_flight(BlockBackend *blk) |
33f2a757 | 1479 | { |
37868b2a | 1480 | IO_CODE(); |
d73415a3 | 1481 | qatomic_inc(&blk->in_flight); |
33f2a757 SH |
1482 | } |
1483 | ||
c90e2a9c | 1484 | void blk_dec_in_flight(BlockBackend *blk) |
33f2a757 | 1485 | { |
37868b2a | 1486 | IO_CODE(); |
d73415a3 | 1487 | qatomic_dec(&blk->in_flight); |
cfe29d82 | 1488 | aio_wait_kick(); |
33f2a757 SH |
1489 | } |
1490 | ||
e7f7d676 HR |
1491 | static void error_callback_bh(void *opaque) |
1492 | { | |
1493 | struct BlockBackendAIOCB *acb = opaque; | |
99723548 | 1494 | |
33f2a757 | 1495 | blk_dec_in_flight(acb->blk); |
e7f7d676 HR |
1496 | acb->common.cb(acb->common.opaque, acb->ret); |
1497 | qemu_aio_unref(acb); | |
1498 | } | |
1499 | ||
ca78ecfa PL |
1500 | BlockAIOCB *blk_abort_aio_request(BlockBackend *blk, |
1501 | BlockCompletionFunc *cb, | |
1502 | void *opaque, int ret) | |
e7f7d676 HR |
1503 | { |
1504 | struct BlockBackendAIOCB *acb; | |
37868b2a | 1505 | IO_CODE(); |
e7f7d676 | 1506 | |
33f2a757 | 1507 | blk_inc_in_flight(blk); |
e7f7d676 | 1508 | acb = blk_aio_get(&block_backend_aiocb_info, blk, cb, opaque); |
4981bdec | 1509 | acb->blk = blk; |
e7f7d676 HR |
1510 | acb->ret = ret; |
1511 | ||
e4ec5ad4 PD |
1512 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1513 | error_callback_bh, acb); | |
e7f7d676 HR |
1514 | return &acb->common; |
1515 | } | |
1516 | ||
57d6a428 KW |
1517 | typedef struct BlkAioEmAIOCB { |
1518 | BlockAIOCB common; | |
1519 | BlkRwCo rwco; | |
a93d81c8 | 1520 | int64_t bytes; |
57d6a428 | 1521 | bool has_returned; |
57d6a428 KW |
1522 | } BlkAioEmAIOCB; |
1523 | ||
1d719ddc SH |
1524 | static AioContext *blk_aio_em_aiocb_get_aio_context(BlockAIOCB *acb_) |
1525 | { | |
1526 | BlkAioEmAIOCB *acb = container_of(acb_, BlkAioEmAIOCB, common); | |
1527 | ||
1528 | return blk_get_aio_context(acb->rwco.blk); | |
1529 | } | |
1530 | ||
57d6a428 KW |
1531 | static const AIOCBInfo blk_aio_em_aiocb_info = { |
1532 | .aiocb_size = sizeof(BlkAioEmAIOCB), | |
1d719ddc | 1533 | .get_aio_context = blk_aio_em_aiocb_get_aio_context, |
57d6a428 KW |
1534 | }; |
1535 | ||
1536 | static void blk_aio_complete(BlkAioEmAIOCB *acb) | |
1537 | { | |
57d6a428 KW |
1538 | if (acb->has_returned) { |
1539 | acb->common.cb(acb->common.opaque, acb->rwco.ret); | |
46aaf2a5 | 1540 | blk_dec_in_flight(acb->rwco.blk); |
57d6a428 KW |
1541 | qemu_aio_unref(acb); |
1542 | } | |
1543 | } | |
1544 | ||
1545 | static void blk_aio_complete_bh(void *opaque) | |
1546 | { | |
fffb6e12 | 1547 | BlkAioEmAIOCB *acb = opaque; |
fffb6e12 PB |
1548 | assert(acb->has_returned); |
1549 | blk_aio_complete(acb); | |
57d6a428 KW |
1550 | } |
1551 | ||
a93d81c8 VSO |
1552 | static BlockAIOCB *blk_aio_prwv(BlockBackend *blk, int64_t offset, |
1553 | int64_t bytes, | |
c060332c | 1554 | void *iobuf, CoroutineEntry co_entry, |
57d6a428 KW |
1555 | BdrvRequestFlags flags, |
1556 | BlockCompletionFunc *cb, void *opaque) | |
1557 | { | |
1558 | BlkAioEmAIOCB *acb; | |
1559 | Coroutine *co; | |
1560 | ||
33f2a757 | 1561 | blk_inc_in_flight(blk); |
57d6a428 KW |
1562 | acb = blk_aio_get(&blk_aio_em_aiocb_info, blk, cb, opaque); |
1563 | acb->rwco = (BlkRwCo) { | |
1564 | .blk = blk, | |
1565 | .offset = offset, | |
c060332c | 1566 | .iobuf = iobuf, |
57d6a428 KW |
1567 | .flags = flags, |
1568 | .ret = NOT_DONE, | |
1569 | }; | |
7fa84cd8 | 1570 | acb->bytes = bytes; |
57d6a428 KW |
1571 | acb->has_returned = false; |
1572 | ||
0b8b8753 | 1573 | co = qemu_coroutine_create(co_entry, acb); |
3d65110f | 1574 | aio_co_enter(blk_get_aio_context(blk), co); |
57d6a428 KW |
1575 | |
1576 | acb->has_returned = true; | |
1577 | if (acb->rwco.ret != NOT_DONE) { | |
e4ec5ad4 PD |
1578 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1579 | blk_aio_complete_bh, acb); | |
57d6a428 KW |
1580 | } |
1581 | ||
1582 | return &acb->common; | |
1583 | } | |
1584 | ||
881a4c55 | 1585 | static void coroutine_fn blk_aio_read_entry(void *opaque) |
57d6a428 KW |
1586 | { |
1587 | BlkAioEmAIOCB *acb = opaque; | |
1588 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1589 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1590 | |
c060332c | 1591 | assert(qiov->size == acb->bytes); |
d1d3fc3d AF |
1592 | rwco->ret = blk_co_do_preadv_part(rwco->blk, rwco->offset, acb->bytes, qiov, |
1593 | 0, rwco->flags); | |
57d6a428 KW |
1594 | blk_aio_complete(acb); |
1595 | } | |
1596 | ||
881a4c55 | 1597 | static void coroutine_fn blk_aio_write_entry(void *opaque) |
57d6a428 KW |
1598 | { |
1599 | BlkAioEmAIOCB *acb = opaque; | |
1600 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1601 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1602 | |
c060332c | 1603 | assert(!qiov || qiov->size == acb->bytes); |
70e8775e VSO |
1604 | rwco->ret = blk_co_do_pwritev_part(rwco->blk, rwco->offset, acb->bytes, |
1605 | qiov, 0, rwco->flags); | |
57d6a428 KW |
1606 | blk_aio_complete(acb); |
1607 | } | |
1608 | ||
d004bd52 | 1609 | BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
a93d81c8 | 1610 | int64_t bytes, BdrvRequestFlags flags, |
d004bd52 | 1611 | BlockCompletionFunc *cb, void *opaque) |
4be74634 | 1612 | { |
37868b2a | 1613 | IO_CODE(); |
a93d81c8 | 1614 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_write_entry, |
983a1600 | 1615 | flags | BDRV_REQ_ZERO_WRITE, cb, opaque); |
4be74634 MA |
1616 | } |
1617 | ||
c86422c5 | 1618 | int64_t coroutine_fn blk_co_getlength(BlockBackend *blk) |
4be74634 | 1619 | { |
37868b2a | 1620 | IO_CODE(); |
c73ff92c | 1621 | GRAPH_RDLOCK_GUARD(); |
c86422c5 | 1622 | |
c73ff92c | 1623 | if (!blk_co_is_available(blk)) { |
c09ba36c HR |
1624 | return -ENOMEDIUM; |
1625 | } | |
1626 | ||
c86422c5 | 1627 | return bdrv_co_getlength(blk_bs(blk)); |
4be74634 MA |
1628 | } |
1629 | ||
9ed98cae | 1630 | int64_t coroutine_fn blk_co_nb_sectors(BlockBackend *blk) |
4be74634 | 1631 | { |
e5203a3b PB |
1632 | BlockDriverState *bs = blk_bs(blk); |
1633 | ||
37868b2a | 1634 | IO_CODE(); |
d8fbf9aa KW |
1635 | GRAPH_RDLOCK_GUARD(); |
1636 | ||
e5203a3b | 1637 | if (!bs) { |
9ed98cae | 1638 | return -ENOMEDIUM; |
a46fc9c9 | 1639 | } else { |
9ed98cae | 1640 | return bdrv_co_nb_sectors(bs); |
a46fc9c9 | 1641 | } |
4be74634 MA |
1642 | } |
1643 | ||
81f730d4 PB |
1644 | /* |
1645 | * This wrapper is written by hand because this function is in the hot I/O path, | |
1646 | * via blk_get_geometry. | |
1647 | */ | |
1648 | int64_t coroutine_mixed_fn blk_nb_sectors(BlockBackend *blk) | |
1649 | { | |
1650 | BlockDriverState *bs = blk_bs(blk); | |
1651 | ||
1652 | IO_CODE(); | |
1653 | ||
1654 | if (!bs) { | |
1655 | return -ENOMEDIUM; | |
1656 | } else { | |
1657 | return bdrv_nb_sectors(bs); | |
1658 | } | |
1659 | } | |
1660 | ||
9ed98cae PB |
1661 | /* return 0 as number of sectors if no device present or error */ |
1662 | void coroutine_fn blk_co_get_geometry(BlockBackend *blk, | |
1663 | uint64_t *nb_sectors_ptr) | |
1ef01253 | 1664 | { |
9ed98cae PB |
1665 | int64_t ret = blk_co_nb_sectors(blk); |
1666 | *nb_sectors_ptr = ret < 0 ? 0 : ret; | |
1ef01253 HR |
1667 | } |
1668 | ||
81f730d4 PB |
1669 | /* |
1670 | * This wrapper is written by hand because this function is in the hot I/O path. | |
1671 | */ | |
1672 | void coroutine_mixed_fn blk_get_geometry(BlockBackend *blk, | |
1673 | uint64_t *nb_sectors_ptr) | |
1674 | { | |
1675 | int64_t ret = blk_nb_sectors(blk); | |
1676 | *nb_sectors_ptr = ret < 0 ? 0 : ret; | |
1677 | } | |
1678 | ||
60cb2fa7 EB |
1679 | BlockAIOCB *blk_aio_preadv(BlockBackend *blk, int64_t offset, |
1680 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1681 | BlockCompletionFunc *cb, void *opaque) | |
1682 | { | |
37868b2a | 1683 | IO_CODE(); |
a93d81c8 | 1684 | assert((uint64_t)qiov->size <= INT64_MAX); |
60cb2fa7 EB |
1685 | return blk_aio_prwv(blk, offset, qiov->size, qiov, |
1686 | blk_aio_read_entry, flags, cb, opaque); | |
1687 | } | |
1688 | ||
60cb2fa7 EB |
1689 | BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
1690 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1691 | BlockCompletionFunc *cb, void *opaque) | |
1692 | { | |
37868b2a | 1693 | IO_CODE(); |
73d4a113 | 1694 | assert((uint64_t)qiov->size <= INT64_MAX); |
60cb2fa7 EB |
1695 | return blk_aio_prwv(blk, offset, qiov->size, qiov, |
1696 | blk_aio_write_entry, flags, cb, opaque); | |
1697 | } | |
1698 | ||
4be74634 MA |
1699 | void blk_aio_cancel(BlockAIOCB *acb) |
1700 | { | |
0439c5a4 | 1701 | GLOBAL_STATE_CODE(); |
4be74634 MA |
1702 | bdrv_aio_cancel(acb); |
1703 | } | |
1704 | ||
1705 | void blk_aio_cancel_async(BlockAIOCB *acb) | |
1706 | { | |
37868b2a | 1707 | IO_CODE(); |
4be74634 MA |
1708 | bdrv_aio_cancel_async(acb); |
1709 | } | |
1710 | ||
fbb92b67 | 1711 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
df02da00 | 1712 | static int coroutine_fn |
70e8775e | 1713 | blk_co_do_ioctl(BlockBackend *blk, unsigned long int req, void *buf) |
4be74634 | 1714 | { |
1581a70d EGE |
1715 | IO_CODE(); |
1716 | ||
cf312932 | 1717 | blk_wait_while_drained(blk); |
26c518ab | 1718 | GRAPH_RDLOCK_GUARD(); |
cf312932 | 1719 | |
c73ff92c | 1720 | if (!blk_co_is_available(blk)) { |
c09ba36c HR |
1721 | return -ENOMEDIUM; |
1722 | } | |
1723 | ||
48af776a KW |
1724 | return bdrv_co_ioctl(blk_bs(blk), req, buf); |
1725 | } | |
1726 | ||
df02da00 AF |
1727 | int coroutine_fn blk_co_ioctl(BlockBackend *blk, unsigned long int req, |
1728 | void *buf) | |
48af776a | 1729 | { |
16d36e29 | 1730 | int ret; |
37868b2a | 1731 | IO_OR_GS_CODE(); |
c060332c | 1732 | |
16d36e29 | 1733 | blk_inc_in_flight(blk); |
df02da00 | 1734 | ret = blk_co_do_ioctl(blk, req, buf); |
16d36e29 | 1735 | blk_dec_in_flight(blk); |
48af776a | 1736 | |
16d36e29 | 1737 | return ret; |
48af776a KW |
1738 | } |
1739 | ||
881a4c55 | 1740 | static void coroutine_fn blk_aio_ioctl_entry(void *opaque) |
48af776a KW |
1741 | { |
1742 | BlkAioEmAIOCB *acb = opaque; | |
1743 | BlkRwCo *rwco = &acb->rwco; | |
1744 | ||
70e8775e | 1745 | rwco->ret = blk_co_do_ioctl(rwco->blk, rwco->offset, rwco->iobuf); |
c060332c | 1746 | |
48af776a | 1747 | blk_aio_complete(acb); |
4be74634 MA |
1748 | } |
1749 | ||
1750 | BlockAIOCB *blk_aio_ioctl(BlockBackend *blk, unsigned long int req, void *buf, | |
1751 | BlockCompletionFunc *cb, void *opaque) | |
1752 | { | |
37868b2a | 1753 | IO_CODE(); |
c060332c | 1754 | return blk_aio_prwv(blk, req, 0, buf, blk_aio_ioctl_entry, 0, cb, opaque); |
4be74634 MA |
1755 | } |
1756 | ||
fbb92b67 | 1757 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
07a64aa4 | 1758 | static int coroutine_fn |
70e8775e | 1759 | blk_co_do_pdiscard(BlockBackend *blk, int64_t offset, int64_t bytes) |
2bb0dce7 | 1760 | { |
cf312932 | 1761 | int ret; |
1581a70d | 1762 | IO_CODE(); |
cf312932 KW |
1763 | |
1764 | blk_wait_while_drained(blk); | |
9a5a1c62 | 1765 | GRAPH_RDLOCK_GUARD(); |
cf312932 KW |
1766 | |
1767 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1768 | if (ret < 0) { |
1769 | return ret; | |
1770 | } | |
1771 | ||
0b9fd3f4 | 1772 | return bdrv_co_pdiscard(blk->root, offset, bytes); |
2bb0dce7 HR |
1773 | } |
1774 | ||
881a4c55 | 1775 | static void coroutine_fn blk_aio_pdiscard_entry(void *opaque) |
564806c5 KW |
1776 | { |
1777 | BlkAioEmAIOCB *acb = opaque; | |
1778 | BlkRwCo *rwco = &acb->rwco; | |
1779 | ||
70e8775e | 1780 | rwco->ret = blk_co_do_pdiscard(rwco->blk, rwco->offset, acb->bytes); |
564806c5 KW |
1781 | blk_aio_complete(acb); |
1782 | } | |
1783 | ||
1784 | BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, | |
a93d81c8 | 1785 | int64_t offset, int64_t bytes, |
564806c5 KW |
1786 | BlockCompletionFunc *cb, void *opaque) |
1787 | { | |
37868b2a | 1788 | IO_CODE(); |
564806c5 KW |
1789 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_pdiscard_entry, 0, |
1790 | cb, opaque); | |
1791 | } | |
1792 | ||
2800637a VSO |
1793 | int coroutine_fn blk_co_pdiscard(BlockBackend *blk, int64_t offset, |
1794 | int64_t bytes) | |
fbb92b67 KW |
1795 | { |
1796 | int ret; | |
37868b2a | 1797 | IO_OR_GS_CODE(); |
fbb92b67 KW |
1798 | |
1799 | blk_inc_in_flight(blk); | |
70e8775e | 1800 | ret = blk_co_do_pdiscard(blk, offset, bytes); |
fbb92b67 KW |
1801 | blk_dec_in_flight(blk); |
1802 | ||
1803 | return ret; | |
1804 | } | |
1805 | ||
fbb92b67 | 1806 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
07a64aa4 | 1807 | static int coroutine_fn blk_co_do_flush(BlockBackend *blk) |
2bb0dce7 | 1808 | { |
1581a70d | 1809 | IO_CODE(); |
88095349 EGE |
1810 | blk_wait_while_drained(blk); |
1811 | GRAPH_RDLOCK_GUARD(); | |
cf312932 | 1812 | |
c73ff92c | 1813 | if (!blk_co_is_available(blk)) { |
c09ba36c HR |
1814 | return -ENOMEDIUM; |
1815 | } | |
1816 | ||
f21d96d0 | 1817 | return bdrv_co_flush(blk_bs(blk)); |
2bb0dce7 HR |
1818 | } |
1819 | ||
881a4c55 | 1820 | static void coroutine_fn blk_aio_flush_entry(void *opaque) |
564806c5 KW |
1821 | { |
1822 | BlkAioEmAIOCB *acb = opaque; | |
1823 | BlkRwCo *rwco = &acb->rwco; | |
1824 | ||
70e8775e | 1825 | rwco->ret = blk_co_do_flush(rwco->blk); |
564806c5 KW |
1826 | blk_aio_complete(acb); |
1827 | } | |
1828 | ||
1829 | BlockAIOCB *blk_aio_flush(BlockBackend *blk, | |
1830 | BlockCompletionFunc *cb, void *opaque) | |
1831 | { | |
37868b2a | 1832 | IO_CODE(); |
564806c5 KW |
1833 | return blk_aio_prwv(blk, 0, 0, NULL, blk_aio_flush_entry, 0, cb, opaque); |
1834 | } | |
1835 | ||
fbb92b67 KW |
1836 | int coroutine_fn blk_co_flush(BlockBackend *blk) |
1837 | { | |
1838 | int ret; | |
37868b2a | 1839 | IO_OR_GS_CODE(); |
fbb92b67 KW |
1840 | |
1841 | blk_inc_in_flight(blk); | |
70e8775e | 1842 | ret = blk_co_do_flush(blk); |
fbb92b67 KW |
1843 | blk_dec_in_flight(blk); |
1844 | ||
1845 | return ret; | |
1846 | } | |
1847 | ||
97b0385a AY |
1848 | void blk_drain(BlockBackend *blk) |
1849 | { | |
33f2a757 | 1850 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 1851 | GLOBAL_STATE_CODE(); |
33f2a757 SH |
1852 | |
1853 | if (bs) { | |
1e3552db | 1854 | bdrv_ref(bs); |
33f2a757 SH |
1855 | bdrv_drained_begin(bs); |
1856 | } | |
1857 | ||
1858 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 | 1859 | AIO_WAIT_WHILE(blk_get_aio_context(blk), |
6a8a98a0 | 1860 | qatomic_read(&blk->in_flight) > 0); |
33f2a757 SH |
1861 | |
1862 | if (bs) { | |
1863 | bdrv_drained_end(bs); | |
1e3552db | 1864 | bdrv_unref(bs); |
a46fc9c9 | 1865 | } |
97b0385a AY |
1866 | } |
1867 | ||
4be74634 MA |
1868 | void blk_drain_all(void) |
1869 | { | |
33f2a757 SH |
1870 | BlockBackend *blk = NULL; |
1871 | ||
0439c5a4 EGE |
1872 | GLOBAL_STATE_CODE(); |
1873 | ||
33f2a757 SH |
1874 | bdrv_drain_all_begin(); |
1875 | ||
1876 | while ((blk = blk_all_next(blk)) != NULL) { | |
33f2a757 | 1877 | /* We may have -ENOMEDIUM completions in flight */ |
d5eab432 | 1878 | AIO_WAIT_WHILE_UNLOCKED(NULL, qatomic_read(&blk->in_flight) > 0); |
33f2a757 SH |
1879 | } |
1880 | ||
1881 | bdrv_drain_all_end(); | |
4be74634 MA |
1882 | } |
1883 | ||
373340b2 HR |
1884 | void blk_set_on_error(BlockBackend *blk, BlockdevOnError on_read_error, |
1885 | BlockdevOnError on_write_error) | |
1886 | { | |
0439c5a4 | 1887 | GLOBAL_STATE_CODE(); |
373340b2 HR |
1888 | blk->on_read_error = on_read_error; |
1889 | blk->on_write_error = on_write_error; | |
1890 | } | |
1891 | ||
4be74634 MA |
1892 | BlockdevOnError blk_get_on_error(BlockBackend *blk, bool is_read) |
1893 | { | |
37868b2a | 1894 | IO_CODE(); |
373340b2 | 1895 | return is_read ? blk->on_read_error : blk->on_write_error; |
4be74634 MA |
1896 | } |
1897 | ||
1898 | BlockErrorAction blk_get_error_action(BlockBackend *blk, bool is_read, | |
1899 | int error) | |
1900 | { | |
373340b2 | 1901 | BlockdevOnError on_err = blk_get_on_error(blk, is_read); |
37868b2a | 1902 | IO_CODE(); |
373340b2 HR |
1903 | |
1904 | switch (on_err) { | |
1905 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
1906 | return (error == ENOSPC) ? | |
1907 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
1908 | case BLOCKDEV_ON_ERROR_STOP: | |
1909 | return BLOCK_ERROR_ACTION_STOP; | |
1910 | case BLOCKDEV_ON_ERROR_REPORT: | |
1911 | return BLOCK_ERROR_ACTION_REPORT; | |
1912 | case BLOCKDEV_ON_ERROR_IGNORE: | |
1913 | return BLOCK_ERROR_ACTION_IGNORE; | |
8c398252 | 1914 | case BLOCKDEV_ON_ERROR_AUTO: |
373340b2 HR |
1915 | default: |
1916 | abort(); | |
1917 | } | |
4be74634 MA |
1918 | } |
1919 | ||
373340b2 HR |
1920 | static void send_qmp_error_event(BlockBackend *blk, |
1921 | BlockErrorAction action, | |
1922 | bool is_read, int error) | |
1923 | { | |
1924 | IoOperationType optype; | |
bfe1a14c | 1925 | BlockDriverState *bs = blk_bs(blk); |
373340b2 HR |
1926 | |
1927 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; | |
54fde4ff | 1928 | qapi_event_send_block_io_error(blk_name(blk), |
bfe1a14c | 1929 | bs ? bdrv_get_node_name(bs) : NULL, optype, |
2bf7e10f | 1930 | action, blk_iostatus_is_enabled(blk), |
3ab72385 | 1931 | error == ENOSPC, strerror(error)); |
373340b2 HR |
1932 | } |
1933 | ||
1934 | /* This is done by device models because, while the block layer knows | |
1935 | * about the error, it does not know whether an operation comes from | |
1936 | * the device or the block layer (from a job, for example). | |
1937 | */ | |
4be74634 MA |
1938 | void blk_error_action(BlockBackend *blk, BlockErrorAction action, |
1939 | bool is_read, int error) | |
1940 | { | |
373340b2 | 1941 | assert(error >= 0); |
37868b2a | 1942 | IO_CODE(); |
373340b2 HR |
1943 | |
1944 | if (action == BLOCK_ERROR_ACTION_STOP) { | |
1945 | /* First set the iostatus, so that "info block" returns an iostatus | |
1946 | * that matches the events raised so far (an additional error iostatus | |
1947 | * is fine, but not a lost one). | |
1948 | */ | |
1949 | blk_iostatus_set_err(blk, error); | |
1950 | ||
1951 | /* Then raise the request to stop the VM and the event. | |
1952 | * qemu_system_vmstop_request_prepare has two effects. First, | |
1953 | * it ensures that the STOP event always comes after the | |
1954 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
1955 | * can observe the STOP event and do a "cont" before the STOP | |
1956 | * event is issued, the VM will not stop. In this case, vm_start() | |
1957 | * also ensures that the STOP/RESUME pair of events is emitted. | |
1958 | */ | |
1959 | qemu_system_vmstop_request_prepare(); | |
1960 | send_qmp_error_event(blk, action, is_read, error); | |
1961 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); | |
1962 | } else { | |
1963 | send_qmp_error_event(blk, action, is_read, error); | |
1964 | } | |
4be74634 MA |
1965 | } |
1966 | ||
86b1cf32 KW |
1967 | /* |
1968 | * Returns true if the BlockBackend can support taking write permissions | |
1969 | * (because its root node is not read-only). | |
1970 | */ | |
1971 | bool blk_supports_write_perm(BlockBackend *blk) | |
4be74634 | 1972 | { |
f21d96d0 | 1973 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 1974 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
1975 | |
1976 | if (bs) { | |
86b1cf32 | 1977 | return !bdrv_is_read_only(bs); |
061959e8 | 1978 | } else { |
260242a8 | 1979 | return blk->root_state.open_flags & BDRV_O_RDWR; |
061959e8 | 1980 | } |
4be74634 MA |
1981 | } |
1982 | ||
86b1cf32 KW |
1983 | /* |
1984 | * Returns true if the BlockBackend can be written to in its current | |
1985 | * configuration (i.e. if write permission have been requested) | |
1986 | */ | |
1987 | bool blk_is_writable(BlockBackend *blk) | |
1988 | { | |
37868b2a | 1989 | IO_CODE(); |
86b1cf32 KW |
1990 | return blk->perm & BLK_PERM_WRITE; |
1991 | } | |
1992 | ||
96710565 | 1993 | bool blk_is_sg(BlockBackend *blk) |
4be74634 | 1994 | { |
f21d96d0 | 1995 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 1996 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
1997 | |
1998 | if (!bs) { | |
96710565 | 1999 | return false; |
a46fc9c9 HR |
2000 | } |
2001 | ||
f21d96d0 | 2002 | return bdrv_is_sg(bs); |
4be74634 MA |
2003 | } |
2004 | ||
96710565 | 2005 | bool blk_enable_write_cache(BlockBackend *blk) |
4be74634 | 2006 | { |
37868b2a | 2007 | IO_CODE(); |
bfd18d1e | 2008 | return blk->enable_write_cache; |
4be74634 MA |
2009 | } |
2010 | ||
2011 | void blk_set_enable_write_cache(BlockBackend *blk, bool wce) | |
2012 | { | |
be8da05b | 2013 | IO_CODE(); |
bfd18d1e | 2014 | blk->enable_write_cache = wce; |
4be74634 MA |
2015 | } |
2016 | ||
3b717194 | 2017 | void blk_activate(BlockBackend *blk, Error **errp) |
2bb0dce7 | 2018 | { |
f21d96d0 | 2019 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2020 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
2021 | |
2022 | if (!bs) { | |
c09ba36c HR |
2023 | error_setg(errp, "Device '%s' has no medium", blk->name); |
2024 | return; | |
2025 | } | |
2026 | ||
a94750d9 | 2027 | bdrv_activate(bs, errp); |
2bb0dce7 HR |
2028 | } |
2029 | ||
1e97be91 | 2030 | bool coroutine_fn blk_co_is_inserted(BlockBackend *blk) |
4be74634 | 2031 | { |
f21d96d0 | 2032 | BlockDriverState *bs = blk_bs(blk); |
37868b2a | 2033 | IO_CODE(); |
c73ff92c | 2034 | assert_bdrv_graph_readable(); |
f21d96d0 | 2035 | |
1e97be91 | 2036 | return bs && bdrv_co_is_inserted(bs); |
db0284f8 HR |
2037 | } |
2038 | ||
c73ff92c | 2039 | bool coroutine_fn blk_co_is_available(BlockBackend *blk) |
db0284f8 | 2040 | { |
37868b2a | 2041 | IO_CODE(); |
c73ff92c | 2042 | return blk_co_is_inserted(blk) && !blk_dev_is_tray_open(blk); |
4be74634 MA |
2043 | } |
2044 | ||
2c75261c | 2045 | void coroutine_fn blk_co_lock_medium(BlockBackend *blk, bool locked) |
4be74634 | 2046 | { |
f21d96d0 | 2047 | BlockDriverState *bs = blk_bs(blk); |
37868b2a | 2048 | IO_CODE(); |
79a292e5 | 2049 | GRAPH_RDLOCK_GUARD(); |
f21d96d0 KW |
2050 | |
2051 | if (bs) { | |
2c75261c | 2052 | bdrv_co_lock_medium(bs, locked); |
a46fc9c9 | 2053 | } |
4be74634 MA |
2054 | } |
2055 | ||
2531b390 | 2056 | void coroutine_fn blk_co_eject(BlockBackend *blk, bool eject_flag) |
4be74634 | 2057 | { |
f21d96d0 | 2058 | BlockDriverState *bs = blk_bs(blk); |
2d76e724 | 2059 | char *id; |
37868b2a | 2060 | IO_CODE(); |
79a292e5 | 2061 | GRAPH_RDLOCK_GUARD(); |
2d76e724 | 2062 | |
f21d96d0 | 2063 | if (bs) { |
2531b390 | 2064 | bdrv_co_eject(bs, eject_flag); |
a46fc9c9 | 2065 | } |
c47ee043 JS |
2066 | |
2067 | /* Whether or not we ejected on the backend, | |
2068 | * the frontend experienced a tray event. */ | |
2069 | id = blk_get_attached_dev_id(blk); | |
2070 | qapi_event_send_device_tray_moved(blk_name(blk), id, | |
3ab72385 | 2071 | eject_flag); |
c47ee043 | 2072 | g_free(id); |
4be74634 MA |
2073 | } |
2074 | ||
2075 | int blk_get_flags(BlockBackend *blk) | |
2076 | { | |
f21d96d0 | 2077 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2078 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
2079 | |
2080 | if (bs) { | |
2081 | return bdrv_get_flags(bs); | |
061959e8 HR |
2082 | } else { |
2083 | return blk->root_state.open_flags; | |
2084 | } | |
4be74634 MA |
2085 | } |
2086 | ||
4841211e EB |
2087 | /* Returns the minimum request alignment, in bytes; guaranteed nonzero */ |
2088 | uint32_t blk_get_request_alignment(BlockBackend *blk) | |
2089 | { | |
2090 | BlockDriverState *bs = blk_bs(blk); | |
37868b2a | 2091 | IO_CODE(); |
4841211e EB |
2092 | return bs ? bs->bl.request_alignment : BDRV_SECTOR_SIZE; |
2093 | } | |
2094 | ||
24b36e98 PB |
2095 | /* Returns the maximum hardware transfer length, in bytes; guaranteed nonzero */ |
2096 | uint64_t blk_get_max_hw_transfer(BlockBackend *blk) | |
2097 | { | |
2098 | BlockDriverState *bs = blk_bs(blk); | |
2099 | uint64_t max = INT_MAX; | |
37868b2a | 2100 | IO_CODE(); |
24b36e98 PB |
2101 | |
2102 | if (bs) { | |
2103 | max = MIN_NON_ZERO(max, bs->bl.max_hw_transfer); | |
2104 | max = MIN_NON_ZERO(max, bs->bl.max_transfer); | |
2105 | } | |
2106 | return ROUND_DOWN(max, blk_get_request_alignment(blk)); | |
2107 | } | |
2108 | ||
5def6b80 EB |
2109 | /* Returns the maximum transfer length, in bytes; guaranteed nonzero */ |
2110 | uint32_t blk_get_max_transfer(BlockBackend *blk) | |
454057b7 | 2111 | { |
f21d96d0 | 2112 | BlockDriverState *bs = blk_bs(blk); |
b99f7fa0 | 2113 | uint32_t max = INT_MAX; |
37868b2a | 2114 | IO_CODE(); |
f21d96d0 KW |
2115 | |
2116 | if (bs) { | |
b99f7fa0 | 2117 | max = MIN_NON_ZERO(max, bs->bl.max_transfer); |
a46fc9c9 | 2118 | } |
b99f7fa0 | 2119 | return ROUND_DOWN(max, blk_get_request_alignment(blk)); |
454057b7 PL |
2120 | } |
2121 | ||
cc071629 PB |
2122 | int blk_get_max_hw_iov(BlockBackend *blk) |
2123 | { | |
37868b2a | 2124 | IO_CODE(); |
cc071629 PB |
2125 | return MIN_NON_ZERO(blk->root->bs->bl.max_hw_iov, |
2126 | blk->root->bs->bl.max_iov); | |
2127 | } | |
2128 | ||
648296e0 SH |
2129 | int blk_get_max_iov(BlockBackend *blk) |
2130 | { | |
37868b2a | 2131 | IO_CODE(); |
f21d96d0 | 2132 | return blk->root->bs->bl.max_iov; |
648296e0 SH |
2133 | } |
2134 | ||
f1c17521 PB |
2135 | void *blk_try_blockalign(BlockBackend *blk, size_t size) |
2136 | { | |
37868b2a | 2137 | IO_CODE(); |
f21d96d0 | 2138 | return qemu_try_blockalign(blk ? blk_bs(blk) : NULL, size); |
f1c17521 PB |
2139 | } |
2140 | ||
4be74634 MA |
2141 | void *blk_blockalign(BlockBackend *blk, size_t size) |
2142 | { | |
37868b2a | 2143 | IO_CODE(); |
f21d96d0 | 2144 | return qemu_blockalign(blk ? blk_bs(blk) : NULL, size); |
4be74634 MA |
2145 | } |
2146 | ||
2147 | bool blk_op_is_blocked(BlockBackend *blk, BlockOpType op, Error **errp) | |
2148 | { | |
f21d96d0 | 2149 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2150 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
2151 | |
2152 | if (!bs) { | |
a46fc9c9 HR |
2153 | return false; |
2154 | } | |
2155 | ||
f21d96d0 | 2156 | return bdrv_op_is_blocked(bs, op, errp); |
4be74634 MA |
2157 | } |
2158 | ||
2159 | void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason) | |
2160 | { | |
f21d96d0 | 2161 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2162 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
2163 | |
2164 | if (bs) { | |
2165 | bdrv_op_unblock(bs, op, reason); | |
a46fc9c9 | 2166 | } |
4be74634 MA |
2167 | } |
2168 | ||
2169 | void blk_op_block_all(BlockBackend *blk, Error *reason) | |
2170 | { | |
f21d96d0 | 2171 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2172 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
2173 | |
2174 | if (bs) { | |
2175 | bdrv_op_block_all(bs, reason); | |
a46fc9c9 | 2176 | } |
4be74634 MA |
2177 | } |
2178 | ||
2179 | void blk_op_unblock_all(BlockBackend *blk, Error *reason) | |
2180 | { | |
f21d96d0 | 2181 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2182 | GLOBAL_STATE_CODE(); |
f21d96d0 KW |
2183 | |
2184 | if (bs) { | |
2185 | bdrv_op_unblock_all(bs, reason); | |
a46fc9c9 | 2186 | } |
4be74634 MA |
2187 | } |
2188 | ||
2189 | AioContext *blk_get_aio_context(BlockBackend *blk) | |
2190 | { | |
d861ab3a | 2191 | BlockDriverState *bs = blk_bs(blk); |
37868b2a | 2192 | IO_CODE(); |
d861ab3a | 2193 | |
d861ab3a | 2194 | if (bs) { |
132ada80 KW |
2195 | AioContext *ctx = bdrv_get_aio_context(blk_bs(blk)); |
2196 | assert(ctx == blk->ctx); | |
d861ab3a KW |
2197 | } |
2198 | ||
2199 | return blk->ctx; | |
4981bdec HR |
2200 | } |
2201 | ||
2202 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) | |
2203 | { | |
2204 | BlockBackendAIOCB *blk_acb = DO_UPCAST(BlockBackendAIOCB, common, acb); | |
2205 | return blk_get_aio_context(blk_acb->blk); | |
4be74634 MA |
2206 | } |
2207 | ||
97896a48 KW |
2208 | static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2209 | bool update_root_node, Error **errp) | |
4be74634 | 2210 | { |
f21d96d0 | 2211 | BlockDriverState *bs = blk_bs(blk); |
c61791fc | 2212 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
97896a48 | 2213 | int ret; |
f21d96d0 KW |
2214 | |
2215 | if (bs) { | |
1e3552db SH |
2216 | bdrv_ref(bs); |
2217 | ||
97896a48 | 2218 | if (update_root_node) { |
f8be48ad EGE |
2219 | /* |
2220 | * update_root_node MUST be false for blk_root_set_aio_ctx_commit(), | |
2221 | * as we are already in the commit function of a transaction. | |
2222 | */ | |
a41cfda1 | 2223 | ret = bdrv_try_change_aio_context(bs, new_context, blk->root, errp); |
97896a48 | 2224 | if (ret < 0) { |
1e3552db | 2225 | bdrv_unref(bs); |
97896a48 KW |
2226 | return ret; |
2227 | } | |
2228 | } | |
af5b6ebe HR |
2229 | /* |
2230 | * Make blk->ctx consistent with the root node before we invoke any | |
2231 | * other operations like drain that might inquire blk->ctx | |
2232 | */ | |
2233 | blk->ctx = new_context; | |
c61791fc | 2234 | if (tgm->throttle_state) { |
dc868fb0 | 2235 | bdrv_drained_begin(bs); |
c61791fc MP |
2236 | throttle_group_detach_aio_context(tgm); |
2237 | throttle_group_attach_aio_context(tgm, new_context); | |
dc868fb0 | 2238 | bdrv_drained_end(bs); |
7ca7f0f6 | 2239 | } |
1e3552db SH |
2240 | |
2241 | bdrv_unref(bs); | |
af5b6ebe HR |
2242 | } else { |
2243 | blk->ctx = new_context; | |
a46fc9c9 | 2244 | } |
97896a48 KW |
2245 | |
2246 | return 0; | |
4be74634 MA |
2247 | } |
2248 | ||
97896a48 KW |
2249 | int blk_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2250 | Error **errp) | |
38475269 | 2251 | { |
0439c5a4 | 2252 | GLOBAL_STATE_CODE(); |
97896a48 | 2253 | return blk_do_set_aio_context(blk, new_context, true, errp); |
38475269 KW |
2254 | } |
2255 | ||
33949396 EGE |
2256 | typedef struct BdrvStateBlkRootContext { |
2257 | AioContext *new_ctx; | |
2258 | BlockBackend *blk; | |
2259 | } BdrvStateBlkRootContext; | |
2260 | ||
2261 | static void blk_root_set_aio_ctx_commit(void *opaque) | |
2262 | { | |
2263 | BdrvStateBlkRootContext *s = opaque; | |
2264 | BlockBackend *blk = s->blk; | |
2265 | ||
2266 | blk_do_set_aio_context(blk, s->new_ctx, false, &error_abort); | |
2267 | } | |
2268 | ||
2269 | static TransactionActionDrv set_blk_root_context = { | |
2270 | .commit = blk_root_set_aio_ctx_commit, | |
2271 | .clean = g_free, | |
2272 | }; | |
2273 | ||
2274 | static bool blk_root_change_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2275 | GHashTable *visited, Transaction *tran, | |
2276 | Error **errp) | |
38475269 KW |
2277 | { |
2278 | BlockBackend *blk = child->opaque; | |
33949396 | 2279 | BdrvStateBlkRootContext *s; |
38475269 | 2280 | |
33949396 EGE |
2281 | if (!blk->allow_aio_context_change) { |
2282 | /* | |
2283 | * Manually created BlockBackends (those with a name) that are not | |
2284 | * attached to anything can change their AioContext without updating | |
2285 | * their user; return an error for others. | |
2286 | */ | |
2287 | if (!blk->name || blk->dev) { | |
2288 | /* TODO Add BB name/QOM path */ | |
2289 | error_setg(errp, "Cannot change iothread of active block backend"); | |
2290 | return false; | |
2291 | } | |
980b0f94 KW |
2292 | } |
2293 | ||
33949396 EGE |
2294 | s = g_new(BdrvStateBlkRootContext, 1); |
2295 | *s = (BdrvStateBlkRootContext) { | |
2296 | .new_ctx = ctx, | |
2297 | .blk = blk, | |
2298 | }; | |
38475269 | 2299 | |
33949396 | 2300 | tran_add(tran, &set_blk_root_context, s); |
38475269 KW |
2301 | return true; |
2302 | } | |
2303 | ||
2019ba0a HR |
2304 | void blk_add_aio_context_notifier(BlockBackend *blk, |
2305 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
2306 | void (*detach_aio_context)(void *opaque), void *opaque) | |
2307 | { | |
d03654ea | 2308 | BlockBackendAioNotifier *notifier; |
f21d96d0 | 2309 | BlockDriverState *bs = blk_bs(blk); |
0439c5a4 | 2310 | GLOBAL_STATE_CODE(); |
f21d96d0 | 2311 | |
d03654ea SH |
2312 | notifier = g_new(BlockBackendAioNotifier, 1); |
2313 | notifier->attached_aio_context = attached_aio_context; | |
2314 | notifier->detach_aio_context = detach_aio_context; | |
2315 | notifier->opaque = opaque; | |
2316 | QLIST_INSERT_HEAD(&blk->aio_notifiers, notifier, list); | |
2317 | ||
f21d96d0 KW |
2318 | if (bs) { |
2319 | bdrv_add_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2320 | detach_aio_context, opaque); |
2321 | } | |
2019ba0a HR |
2322 | } |
2323 | ||
2324 | void blk_remove_aio_context_notifier(BlockBackend *blk, | |
2325 | void (*attached_aio_context)(AioContext *, | |
2326 | void *), | |
2327 | void (*detach_aio_context)(void *), | |
2328 | void *opaque) | |
2329 | { | |
d03654ea | 2330 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2331 | BlockDriverState *bs = blk_bs(blk); |
2332 | ||
0439c5a4 EGE |
2333 | GLOBAL_STATE_CODE(); |
2334 | ||
f21d96d0 KW |
2335 | if (bs) { |
2336 | bdrv_remove_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2337 | detach_aio_context, opaque); |
2338 | } | |
d03654ea SH |
2339 | |
2340 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
2341 | if (notifier->attached_aio_context == attached_aio_context && | |
2342 | notifier->detach_aio_context == detach_aio_context && | |
2343 | notifier->opaque == opaque) { | |
2344 | QLIST_REMOVE(notifier, list); | |
2345 | g_free(notifier); | |
2346 | return; | |
2347 | } | |
2348 | } | |
2349 | ||
2350 | abort(); | |
2019ba0a HR |
2351 | } |
2352 | ||
3301f6c6 HR |
2353 | void blk_add_remove_bs_notifier(BlockBackend *blk, Notifier *notify) |
2354 | { | |
0439c5a4 | 2355 | GLOBAL_STATE_CODE(); |
3301f6c6 HR |
2356 | notifier_list_add(&blk->remove_bs_notifiers, notify); |
2357 | } | |
2358 | ||
2359 | void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify) | |
2360 | { | |
0439c5a4 | 2361 | GLOBAL_STATE_CODE(); |
3301f6c6 HR |
2362 | notifier_list_add(&blk->insert_bs_notifiers, notify); |
2363 | } | |
2364 | ||
8f497454 | 2365 | void coroutine_fn blk_co_io_plug(BlockBackend *blk) |
4be74634 | 2366 | { |
f21d96d0 | 2367 | BlockDriverState *bs = blk_bs(blk); |
37868b2a | 2368 | IO_CODE(); |
c3827069 | 2369 | GRAPH_RDLOCK_GUARD(); |
f21d96d0 KW |
2370 | |
2371 | if (bs) { | |
8f497454 | 2372 | bdrv_co_io_plug(bs); |
a46fc9c9 | 2373 | } |
4be74634 MA |
2374 | } |
2375 | ||
09d9fc97 | 2376 | void coroutine_fn blk_co_io_unplug(BlockBackend *blk) |
4be74634 | 2377 | { |
f21d96d0 | 2378 | BlockDriverState *bs = blk_bs(blk); |
37868b2a | 2379 | IO_CODE(); |
c3827069 | 2380 | GRAPH_RDLOCK_GUARD(); |
f21d96d0 KW |
2381 | |
2382 | if (bs) { | |
09d9fc97 | 2383 | bdrv_co_io_unplug(bs); |
a46fc9c9 | 2384 | } |
4be74634 MA |
2385 | } |
2386 | ||
2387 | BlockAcctStats *blk_get_stats(BlockBackend *blk) | |
2388 | { | |
37868b2a | 2389 | IO_CODE(); |
7f0e9da6 | 2390 | return &blk->stats; |
4be74634 MA |
2391 | } |
2392 | ||
2393 | void *blk_aio_get(const AIOCBInfo *aiocb_info, BlockBackend *blk, | |
2394 | BlockCompletionFunc *cb, void *opaque) | |
2395 | { | |
37868b2a | 2396 | IO_CODE(); |
4be74634 MA |
2397 | return qemu_aio_get(aiocb_info, blk_bs(blk), cb, opaque); |
2398 | } | |
1ef01253 | 2399 | |
d004bd52 | 2400 | int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
34460feb | 2401 | int64_t bytes, BdrvRequestFlags flags) |
1ef01253 | 2402 | { |
37868b2a | 2403 | IO_OR_GS_CODE(); |
f5a5ca79 | 2404 | return blk_co_pwritev(blk, offset, bytes, NULL, |
16aaf975 | 2405 | flags | BDRV_REQ_ZERO_WRITE); |
1ef01253 HR |
2406 | } |
2407 | ||
2c9715fa AF |
2408 | int coroutine_fn blk_co_pwrite_compressed(BlockBackend *blk, int64_t offset, |
2409 | int64_t bytes, const void *buf) | |
1ef01253 | 2410 | { |
06f0325c | 2411 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); |
37868b2a | 2412 | IO_OR_GS_CODE(); |
2c9715fa AF |
2413 | return blk_co_pwritev_part(blk, offset, bytes, &qiov, 0, |
2414 | BDRV_REQ_WRITE_COMPRESSED); | |
1ef01253 HR |
2415 | } |
2416 | ||
015ed252 AF |
2417 | int coroutine_fn blk_co_truncate(BlockBackend *blk, int64_t offset, bool exact, |
2418 | PreallocMode prealloc, BdrvRequestFlags flags, | |
2419 | Error **errp) | |
1ef01253 | 2420 | { |
37868b2a | 2421 | IO_OR_GS_CODE(); |
c2b8e315 | 2422 | GRAPH_RDLOCK_GUARD(); |
c73ff92c | 2423 | if (!blk_co_is_available(blk)) { |
ed3d2ec9 | 2424 | error_setg(errp, "No medium inserted"); |
c09ba36c HR |
2425 | return -ENOMEDIUM; |
2426 | } | |
2427 | ||
015ed252 | 2428 | return bdrv_co_truncate(blk->root, offset, exact, prealloc, flags, errp); |
1ef01253 HR |
2429 | } |
2430 | ||
1ef01253 HR |
2431 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, |
2432 | int64_t pos, int size) | |
2433 | { | |
bfd18d1e | 2434 | int ret; |
0439c5a4 | 2435 | GLOBAL_STATE_CODE(); |
bfd18d1e | 2436 | |
c09ba36c HR |
2437 | if (!blk_is_available(blk)) { |
2438 | return -ENOMEDIUM; | |
2439 | } | |
2440 | ||
bfd18d1e KW |
2441 | ret = bdrv_save_vmstate(blk_bs(blk), buf, pos, size); |
2442 | if (ret < 0) { | |
2443 | return ret; | |
2444 | } | |
2445 | ||
2446 | if (ret == size && !blk->enable_write_cache) { | |
2447 | ret = bdrv_flush(blk_bs(blk)); | |
2448 | } | |
2449 | ||
2450 | return ret < 0 ? ret : size; | |
1ef01253 HR |
2451 | } |
2452 | ||
2453 | int blk_load_vmstate(BlockBackend *blk, uint8_t *buf, int64_t pos, int size) | |
2454 | { | |
0439c5a4 | 2455 | GLOBAL_STATE_CODE(); |
c09ba36c HR |
2456 | if (!blk_is_available(blk)) { |
2457 | return -ENOMEDIUM; | |
2458 | } | |
2459 | ||
f21d96d0 | 2460 | return bdrv_load_vmstate(blk_bs(blk), buf, pos, size); |
1ef01253 | 2461 | } |
f0272c4d ET |
2462 | |
2463 | int blk_probe_blocksizes(BlockBackend *blk, BlockSizes *bsz) | |
2464 | { | |
0439c5a4 | 2465 | GLOBAL_STATE_CODE(); |
c09ba36c HR |
2466 | if (!blk_is_available(blk)) { |
2467 | return -ENOMEDIUM; | |
2468 | } | |
2469 | ||
f21d96d0 | 2470 | return bdrv_probe_blocksizes(blk_bs(blk), bsz); |
f0272c4d ET |
2471 | } |
2472 | ||
2473 | int blk_probe_geometry(BlockBackend *blk, HDGeometry *geo) | |
2474 | { | |
0439c5a4 | 2475 | GLOBAL_STATE_CODE(); |
c09ba36c HR |
2476 | if (!blk_is_available(blk)) { |
2477 | return -ENOMEDIUM; | |
2478 | } | |
2479 | ||
f21d96d0 | 2480 | return bdrv_probe_geometry(blk_bs(blk), geo); |
f0272c4d | 2481 | } |
281d22d8 HR |
2482 | |
2483 | /* | |
2484 | * Updates the BlockBackendRootState object with data from the currently | |
2485 | * attached BlockDriverState. | |
2486 | */ | |
2487 | void blk_update_root_state(BlockBackend *blk) | |
2488 | { | |
0439c5a4 | 2489 | GLOBAL_STATE_CODE(); |
f21d96d0 | 2490 | assert(blk->root); |
281d22d8 | 2491 | |
f21d96d0 | 2492 | blk->root_state.open_flags = blk->root->bs->open_flags; |
f21d96d0 | 2493 | blk->root_state.detect_zeroes = blk->root->bs->detect_zeroes; |
281d22d8 HR |
2494 | } |
2495 | ||
38cb18f5 | 2496 | /* |
b85114f8 KW |
2497 | * Returns the detect-zeroes setting to be used for bdrv_open() of a |
2498 | * BlockDriverState which is supposed to inherit the root state. | |
38cb18f5 | 2499 | */ |
b85114f8 | 2500 | bool blk_get_detect_zeroes_from_root_state(BlockBackend *blk) |
38cb18f5 | 2501 | { |
0439c5a4 | 2502 | GLOBAL_STATE_CODE(); |
b85114f8 | 2503 | return blk->root_state.detect_zeroes; |
38cb18f5 HR |
2504 | } |
2505 | ||
2506 | /* | |
2507 | * Returns the flags to be used for bdrv_open() of a BlockDriverState which is | |
2508 | * supposed to inherit the root state. | |
2509 | */ | |
2510 | int blk_get_open_flags_from_root_state(BlockBackend *blk) | |
2511 | { | |
0439c5a4 | 2512 | GLOBAL_STATE_CODE(); |
260242a8 | 2513 | return blk->root_state.open_flags; |
38cb18f5 HR |
2514 | } |
2515 | ||
281d22d8 HR |
2516 | BlockBackendRootState *blk_get_root_state(BlockBackend *blk) |
2517 | { | |
0439c5a4 | 2518 | GLOBAL_STATE_CODE(); |
281d22d8 HR |
2519 | return &blk->root_state; |
2520 | } | |
1393f212 HR |
2521 | |
2522 | int blk_commit_all(void) | |
2523 | { | |
fe1a9cbc | 2524 | BlockBackend *blk = NULL; |
0439c5a4 | 2525 | GLOBAL_STATE_CODE(); |
fe1a9cbc HR |
2526 | |
2527 | while ((blk = blk_all_next(blk)) != NULL) { | |
2528 | AioContext *aio_context = blk_get_aio_context(blk); | |
9a71b9de | 2529 | BlockDriverState *unfiltered_bs = bdrv_skip_filters(blk_bs(blk)); |
fe1a9cbc HR |
2530 | |
2531 | aio_context_acquire(aio_context); | |
9a71b9de HR |
2532 | if (blk_is_inserted(blk) && bdrv_cow_child(unfiltered_bs)) { |
2533 | int ret; | |
2534 | ||
2535 | ret = bdrv_commit(unfiltered_bs); | |
fe1a9cbc HR |
2536 | if (ret < 0) { |
2537 | aio_context_release(aio_context); | |
2538 | return ret; | |
2539 | } | |
2540 | } | |
2541 | aio_context_release(aio_context); | |
2542 | } | |
2543 | return 0; | |
2544 | } | |
2545 | ||
97148076 KW |
2546 | |
2547 | /* throttling disk I/O limits */ | |
2548 | void blk_set_io_limits(BlockBackend *blk, ThrottleConfig *cfg) | |
2549 | { | |
0439c5a4 | 2550 | GLOBAL_STATE_CODE(); |
022cdc9f | 2551 | throttle_group_config(&blk->public.throttle_group_member, cfg); |
97148076 KW |
2552 | } |
2553 | ||
2554 | void blk_io_limits_disable(BlockBackend *blk) | |
2555 | { | |
48bf7ea8 AG |
2556 | BlockDriverState *bs = blk_bs(blk); |
2557 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; | |
2558 | assert(tgm->throttle_state); | |
0439c5a4 | 2559 | GLOBAL_STATE_CODE(); |
48bf7ea8 | 2560 | if (bs) { |
1e3552db | 2561 | bdrv_ref(bs); |
48bf7ea8 AG |
2562 | bdrv_drained_begin(bs); |
2563 | } | |
2564 | throttle_group_unregister_tgm(tgm); | |
2565 | if (bs) { | |
2566 | bdrv_drained_end(bs); | |
1e3552db | 2567 | bdrv_unref(bs); |
48bf7ea8 | 2568 | } |
97148076 KW |
2569 | } |
2570 | ||
2571 | /* should be called before blk_set_io_limits if a limit is set */ | |
2572 | void blk_io_limits_enable(BlockBackend *blk, const char *group) | |
2573 | { | |
022cdc9f | 2574 | assert(!blk->public.throttle_group_member.throttle_state); |
0439c5a4 | 2575 | GLOBAL_STATE_CODE(); |
c61791fc MP |
2576 | throttle_group_register_tgm(&blk->public.throttle_group_member, |
2577 | group, blk_get_aio_context(blk)); | |
97148076 KW |
2578 | } |
2579 | ||
2580 | void blk_io_limits_update_group(BlockBackend *blk, const char *group) | |
2581 | { | |
0439c5a4 | 2582 | GLOBAL_STATE_CODE(); |
97148076 | 2583 | /* this BB is not part of any group */ |
022cdc9f | 2584 | if (!blk->public.throttle_group_member.throttle_state) { |
97148076 KW |
2585 | return; |
2586 | } | |
2587 | ||
2588 | /* this BB is a part of the same group than the one we want */ | |
022cdc9f MP |
2589 | if (!g_strcmp0(throttle_group_get_name(&blk->public.throttle_group_member), |
2590 | group)) { | |
97148076 KW |
2591 | return; |
2592 | } | |
2593 | ||
2594 | /* need to change the group this bs belong to */ | |
2595 | blk_io_limits_disable(blk); | |
2596 | blk_io_limits_enable(blk, group); | |
2597 | } | |
c2066af0 KW |
2598 | |
2599 | static void blk_root_drained_begin(BdrvChild *child) | |
2600 | { | |
2601 | BlockBackend *blk = child->opaque; | |
d73415a3 | 2602 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
c2066af0 | 2603 | |
c4d5bf99 | 2604 | if (qatomic_fetch_inc(&blk->quiesce_counter) == 0) { |
f4d9cc88 JS |
2605 | if (blk->dev_ops && blk->dev_ops->drained_begin) { |
2606 | blk->dev_ops->drained_begin(blk->dev_opaque); | |
2607 | } | |
2608 | } | |
2609 | ||
36fe1331 KW |
2610 | /* Note that blk->root may not be accessible here yet if we are just |
2611 | * attaching to a BlockDriverState that is drained. Use child instead. */ | |
2612 | ||
d73415a3 SH |
2613 | if (qatomic_fetch_inc(&tgm->io_limits_disabled) == 0) { |
2614 | throttle_group_restart_tgm(tgm); | |
c2066af0 KW |
2615 | } |
2616 | } | |
2617 | ||
fe5258a5 KW |
2618 | static bool blk_root_drained_poll(BdrvChild *child) |
2619 | { | |
2620 | BlockBackend *blk = child->opaque; | |
095cc4d0 | 2621 | bool busy = false; |
c4d5bf99 | 2622 | assert(qatomic_read(&blk->quiesce_counter)); |
095cc4d0 SL |
2623 | |
2624 | if (blk->dev_ops && blk->dev_ops->drained_poll) { | |
2625 | busy = blk->dev_ops->drained_poll(blk->dev_opaque); | |
2626 | } | |
2627 | return busy || !!blk->in_flight; | |
fe5258a5 KW |
2628 | } |
2629 | ||
2f65df6e | 2630 | static void blk_root_drained_end(BdrvChild *child) |
c2066af0 KW |
2631 | { |
2632 | BlockBackend *blk = child->opaque; | |
c4d5bf99 | 2633 | assert(qatomic_read(&blk->quiesce_counter)); |
c2066af0 | 2634 | |
022cdc9f | 2635 | assert(blk->public.throttle_group_member.io_limits_disabled); |
d73415a3 | 2636 | qatomic_dec(&blk->public.throttle_group_member.io_limits_disabled); |
f4d9cc88 | 2637 | |
c4d5bf99 | 2638 | if (qatomic_fetch_dec(&blk->quiesce_counter) == 1) { |
f4d9cc88 JS |
2639 | if (blk->dev_ops && blk->dev_ops->drained_end) { |
2640 | blk->dev_ops->drained_end(blk->dev_opaque); | |
2641 | } | |
407ae2ae SH |
2642 | qemu_mutex_lock(&blk->queued_requests_lock); |
2643 | while (qemu_co_enter_next(&blk->queued_requests, | |
2644 | &blk->queued_requests_lock)) { | |
cf312932 KW |
2645 | /* Resume all queued requests */ |
2646 | } | |
407ae2ae | 2647 | qemu_mutex_unlock(&blk->queued_requests_lock); |
f4d9cc88 | 2648 | } |
c2066af0 | 2649 | } |
23d0ba93 | 2650 | |
f4ec04ba | 2651 | bool blk_register_buf(BlockBackend *blk, void *host, size_t size, Error **errp) |
23d0ba93 | 2652 | { |
71038951 SH |
2653 | BlockDriverState *bs = blk_bs(blk); |
2654 | ||
0439c5a4 | 2655 | GLOBAL_STATE_CODE(); |
71038951 SH |
2656 | |
2657 | if (bs) { | |
2658 | return bdrv_register_buf(bs, host, size, errp); | |
2659 | } | |
2660 | return true; | |
23d0ba93 FZ |
2661 | } |
2662 | ||
4f384011 | 2663 | void blk_unregister_buf(BlockBackend *blk, void *host, size_t size) |
23d0ba93 | 2664 | { |
71038951 SH |
2665 | BlockDriverState *bs = blk_bs(blk); |
2666 | ||
0439c5a4 | 2667 | GLOBAL_STATE_CODE(); |
71038951 SH |
2668 | |
2669 | if (bs) { | |
2670 | bdrv_unregister_buf(bs, host, size); | |
2671 | } | |
23d0ba93 | 2672 | } |
b5679fa4 FZ |
2673 | |
2674 | int coroutine_fn blk_co_copy_range(BlockBackend *blk_in, int64_t off_in, | |
2675 | BlockBackend *blk_out, int64_t off_out, | |
e192179b | 2676 | int64_t bytes, BdrvRequestFlags read_flags, |
67b51fb9 | 2677 | BdrvRequestFlags write_flags) |
b5679fa4 FZ |
2678 | { |
2679 | int r; | |
37868b2a | 2680 | IO_CODE(); |
c73ff92c | 2681 | GRAPH_RDLOCK_GUARD(); |
37868b2a | 2682 | |
b5679fa4 FZ |
2683 | r = blk_check_byte_request(blk_in, off_in, bytes); |
2684 | if (r) { | |
2685 | return r; | |
2686 | } | |
2687 | r = blk_check_byte_request(blk_out, off_out, bytes); | |
2688 | if (r) { | |
2689 | return r; | |
2690 | } | |
742bf09b | 2691 | |
b5679fa4 FZ |
2692 | return bdrv_co_copy_range(blk_in->root, off_in, |
2693 | blk_out->root, off_out, | |
67b51fb9 | 2694 | bytes, read_flags, write_flags); |
b5679fa4 | 2695 | } |
5d3b4e99 VSO |
2696 | |
2697 | const BdrvChild *blk_root(BlockBackend *blk) | |
2698 | { | |
0439c5a4 | 2699 | GLOBAL_STATE_CODE(); |
5d3b4e99 VSO |
2700 | return blk->root; |
2701 | } | |
2b7bbdbd HR |
2702 | |
2703 | int blk_make_empty(BlockBackend *blk, Error **errp) | |
2704 | { | |
0439c5a4 | 2705 | GLOBAL_STATE_CODE(); |
2b7bbdbd HR |
2706 | if (!blk_is_available(blk)) { |
2707 | error_setg(errp, "No medium inserted"); | |
2708 | return -ENOMEDIUM; | |
2709 | } | |
2710 | ||
2711 | return bdrv_make_empty(blk->root, errp); | |
2712 | } |