]>
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" |
281d22d8 | 17 | #include "block/throttle-groups.h" |
13d4ff07 | 18 | #include "hw/qdev-core.h" |
18e46a03 | 19 | #include "sysemu/blockdev.h" |
54d31236 | 20 | #include "sysemu/runstate.h" |
e4ec5ad4 PD |
21 | #include "sysemu/sysemu.h" |
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 HR |
58 | |
59 | /* the block size for which the guest device expects atomicity */ | |
60 | int guest_block_size; | |
7f0e9da6 | 61 | |
281d22d8 HR |
62 | /* If the BDS tree is removed, some of its options are stored here (which |
63 | * can be used to restore those options in the new BDS on insert) */ | |
64 | BlockBackendRootState root_state; | |
65 | ||
bfd18d1e KW |
66 | bool enable_write_cache; |
67 | ||
7f0e9da6 HR |
68 | /* I/O stats (display with "info blockstats"). */ |
69 | BlockAcctStats stats; | |
373340b2 HR |
70 | |
71 | BlockdevOnError on_read_error, on_write_error; | |
72 | bool iostatus_enabled; | |
73 | BlockDeviceIoStatus iostatus; | |
3301f6c6 | 74 | |
981776b3 KW |
75 | uint64_t perm; |
76 | uint64_t shared_perm; | |
d35ff5e6 | 77 | bool disable_perm; |
981776b3 | 78 | |
980b0f94 | 79 | bool allow_aio_context_change; |
c10c9d96 KW |
80 | bool allow_write_beyond_eof; |
81 | ||
3301f6c6 | 82 | NotifierList remove_bs_notifiers, insert_bs_notifiers; |
d03654ea | 83 | QLIST_HEAD(, BlockBackendAioNotifier) aio_notifiers; |
f4d9cc88 JS |
84 | |
85 | int quiesce_counter; | |
cf312932 KW |
86 | CoQueue queued_requests; |
87 | bool disable_request_queuing; | |
88 | ||
5f7772c4 | 89 | VMChangeStateEntry *vmsh; |
ca2e2144 | 90 | bool force_allow_inactivate; |
33f2a757 SH |
91 | |
92 | /* Number of in-flight aio requests. BlockDriverState also counts | |
93 | * in-flight requests but aio requests can exist even when blk->root is | |
94 | * NULL, so we cannot rely on its counter for that case. | |
95 | * Accessed with atomic ops. | |
96 | */ | |
97 | unsigned int in_flight; | |
26f54e9a MA |
98 | }; |
99 | ||
e7f7d676 HR |
100 | typedef struct BlockBackendAIOCB { |
101 | BlockAIOCB common; | |
4981bdec | 102 | BlockBackend *blk; |
e7f7d676 HR |
103 | int ret; |
104 | } BlockBackendAIOCB; | |
105 | ||
106 | static const AIOCBInfo block_backend_aiocb_info = { | |
4981bdec | 107 | .get_aio_context = blk_aiocb_get_aio_context, |
e7f7d676 HR |
108 | .aiocb_size = sizeof(BlockBackendAIOCB), |
109 | }; | |
110 | ||
8fb3c76c | 111 | static void drive_info_del(DriveInfo *dinfo); |
7c8eece4 | 112 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs); |
8fb3c76c | 113 | |
2cf22d6a HR |
114 | /* All BlockBackends */ |
115 | static QTAILQ_HEAD(, BlockBackend) block_backends = | |
116 | QTAILQ_HEAD_INITIALIZER(block_backends); | |
117 | ||
9492b0b9 HR |
118 | /* All BlockBackends referenced by the monitor and which are iterated through by |
119 | * blk_next() */ | |
120 | static QTAILQ_HEAD(, BlockBackend) monitor_block_backends = | |
121 | QTAILQ_HEAD_INITIALIZER(monitor_block_backends); | |
26f54e9a | 122 | |
f21d96d0 KW |
123 | static void blk_root_inherit_options(int *child_flags, QDict *child_options, |
124 | int parent_flags, QDict *parent_options) | |
125 | { | |
126 | /* We're not supposed to call this function for root nodes */ | |
127 | abort(); | |
128 | } | |
c2066af0 | 129 | static void blk_root_drained_begin(BdrvChild *child); |
fe5258a5 | 130 | static bool blk_root_drained_poll(BdrvChild *child); |
e037c09c | 131 | static void blk_root_drained_end(BdrvChild *child, int *drained_end_counter); |
f21d96d0 | 132 | |
5c8cab48 KW |
133 | static void blk_root_change_media(BdrvChild *child, bool load); |
134 | static void blk_root_resize(BdrvChild *child); | |
135 | ||
38475269 KW |
136 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, |
137 | GSList **ignore, Error **errp); | |
138 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
139 | GSList **ignore); | |
140 | ||
b5411555 KW |
141 | static char *blk_root_get_parent_desc(BdrvChild *child) |
142 | { | |
143 | BlockBackend *blk = child->opaque; | |
144 | char *dev_id; | |
145 | ||
146 | if (blk->name) { | |
147 | return g_strdup(blk->name); | |
148 | } | |
149 | ||
150 | dev_id = blk_get_attached_dev_id(blk); | |
151 | if (*dev_id) { | |
152 | return dev_id; | |
153 | } else { | |
154 | /* TODO Callback into the BB owner for something more detailed */ | |
155 | g_free(dev_id); | |
156 | return g_strdup("a block device"); | |
157 | } | |
158 | } | |
159 | ||
4c265bf9 KW |
160 | static const char *blk_root_get_name(BdrvChild *child) |
161 | { | |
162 | return blk_name(child->opaque); | |
163 | } | |
164 | ||
5f7772c4 FZ |
165 | static void blk_vm_state_changed(void *opaque, int running, RunState state) |
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; | |
193 | ||
194 | if (!blk->disable_perm) { | |
195 | return; | |
196 | } | |
197 | ||
198 | blk->disable_perm = false; | |
199 | ||
5f7772c4 FZ |
200 | blk_set_perm(blk, blk->perm, BLK_PERM_ALL, &local_err); |
201 | if (local_err) { | |
202 | error_propagate(errp, local_err); | |
203 | blk->disable_perm = true; | |
204 | return; | |
205 | } | |
206 | ||
207 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
208 | /* Activation can happen when migration process is still active, for | |
209 | * example when nbd_server_add is called during non-shared storage | |
210 | * migration. Defer the shared_perm update to migration completion. */ | |
211 | if (!blk->vmsh) { | |
212 | blk->vmsh = qemu_add_vm_change_state_handler(blk_vm_state_changed, | |
213 | blk); | |
214 | } | |
215 | return; | |
216 | } | |
217 | ||
4417ab7a KW |
218 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); |
219 | if (local_err) { | |
220 | error_propagate(errp, local_err); | |
221 | blk->disable_perm = true; | |
222 | return; | |
223 | } | |
224 | } | |
225 | ||
ca2e2144 FZ |
226 | void blk_set_force_allow_inactivate(BlockBackend *blk) |
227 | { | |
228 | blk->force_allow_inactivate = true; | |
229 | } | |
230 | ||
c16de8f5 FZ |
231 | static bool blk_can_inactivate(BlockBackend *blk) |
232 | { | |
ca2e2144 | 233 | /* If it is a guest device, inactivate is ok. */ |
c16de8f5 FZ |
234 | if (blk->dev || blk_name(blk)[0]) { |
235 | return true; | |
236 | } | |
237 | ||
ca2e2144 FZ |
238 | /* Inactivating means no more writes to the image can be done, |
239 | * even if those writes would be changes invisible to the | |
240 | * guest. For block job BBs that satisfy this, we can just allow | |
241 | * it. This is the case for mirror job source, which is required | |
242 | * by libvirt non-shared block migration. */ | |
243 | if (!(blk->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED))) { | |
244 | return true; | |
245 | } | |
246 | ||
247 | return blk->force_allow_inactivate; | |
c16de8f5 FZ |
248 | } |
249 | ||
cfa1a572 KW |
250 | static int blk_root_inactivate(BdrvChild *child) |
251 | { | |
252 | BlockBackend *blk = child->opaque; | |
253 | ||
254 | if (blk->disable_perm) { | |
255 | return 0; | |
256 | } | |
257 | ||
c16de8f5 | 258 | if (!blk_can_inactivate(blk)) { |
cfa1a572 KW |
259 | return -EPERM; |
260 | } | |
261 | ||
262 | blk->disable_perm = true; | |
263 | if (blk->root) { | |
264 | bdrv_child_try_set_perm(blk->root, 0, BLK_PERM_ALL, &error_abort); | |
265 | } | |
266 | ||
267 | return 0; | |
268 | } | |
269 | ||
d03654ea SH |
270 | static void blk_root_attach(BdrvChild *child) |
271 | { | |
272 | BlockBackend *blk = child->opaque; | |
273 | BlockBackendAioNotifier *notifier; | |
274 | ||
275 | trace_blk_root_attach(child, blk, child->bs); | |
276 | ||
277 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
278 | bdrv_add_aio_context_notifier(child->bs, | |
279 | notifier->attached_aio_context, | |
280 | notifier->detach_aio_context, | |
281 | notifier->opaque); | |
282 | } | |
283 | } | |
284 | ||
285 | static void blk_root_detach(BdrvChild *child) | |
286 | { | |
287 | BlockBackend *blk = child->opaque; | |
288 | BlockBackendAioNotifier *notifier; | |
289 | ||
290 | trace_blk_root_detach(child, blk, child->bs); | |
291 | ||
292 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
293 | bdrv_remove_aio_context_notifier(child->bs, | |
294 | notifier->attached_aio_context, | |
295 | notifier->detach_aio_context, | |
296 | notifier->opaque); | |
297 | } | |
298 | } | |
299 | ||
f21d96d0 | 300 | static const BdrvChildRole child_root = { |
c2066af0 KW |
301 | .inherit_options = blk_root_inherit_options, |
302 | ||
5c8cab48 KW |
303 | .change_media = blk_root_change_media, |
304 | .resize = blk_root_resize, | |
4c265bf9 | 305 | .get_name = blk_root_get_name, |
b5411555 | 306 | .get_parent_desc = blk_root_get_parent_desc, |
5c8cab48 | 307 | |
c2066af0 | 308 | .drained_begin = blk_root_drained_begin, |
fe5258a5 | 309 | .drained_poll = blk_root_drained_poll, |
c2066af0 | 310 | .drained_end = blk_root_drained_end, |
4417ab7a KW |
311 | |
312 | .activate = blk_root_activate, | |
cfa1a572 | 313 | .inactivate = blk_root_inactivate, |
d03654ea SH |
314 | |
315 | .attach = blk_root_attach, | |
316 | .detach = blk_root_detach, | |
38475269 KW |
317 | |
318 | .can_set_aio_ctx = blk_root_can_set_aio_ctx, | |
319 | .set_aio_ctx = blk_root_set_aio_ctx, | |
f21d96d0 KW |
320 | }; |
321 | ||
26f54e9a | 322 | /* |
efaa7c4e | 323 | * Create a new BlockBackend with a reference count of one. |
6d0eb64d KW |
324 | * |
325 | * @perm is a bitmasks of BLK_PERM_* constants which describes the permissions | |
326 | * to request for a block driver node that is attached to this BlockBackend. | |
327 | * @shared_perm is a bitmask which describes which permissions may be granted | |
328 | * to other users of the attached node. | |
329 | * Both sets of permissions can be changed later using blk_set_perm(). | |
330 | * | |
26f54e9a MA |
331 | * Return the new BlockBackend on success, null on failure. |
332 | */ | |
d861ab3a | 333 | BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm) |
26f54e9a MA |
334 | { |
335 | BlockBackend *blk; | |
336 | ||
26f54e9a | 337 | blk = g_new0(BlockBackend, 1); |
26f54e9a | 338 | blk->refcnt = 1; |
d861ab3a | 339 | blk->ctx = ctx; |
6d0eb64d KW |
340 | blk->perm = perm; |
341 | blk->shared_perm = shared_perm; | |
0c3169df KW |
342 | blk_set_enable_write_cache(blk, true); |
343 | ||
cb53460b KW |
344 | blk->on_read_error = BLOCKDEV_ON_ERROR_REPORT; |
345 | blk->on_write_error = BLOCKDEV_ON_ERROR_ENOSPC; | |
346 | ||
9caa6f3d | 347 | block_acct_init(&blk->stats); |
27ccdd52 | 348 | |
cf312932 | 349 | qemu_co_queue_init(&blk->queued_requests); |
3301f6c6 HR |
350 | notifier_list_init(&blk->remove_bs_notifiers); |
351 | notifier_list_init(&blk->insert_bs_notifiers); | |
d03654ea | 352 | QLIST_INIT(&blk->aio_notifiers); |
27ccdd52 | 353 | |
2cf22d6a | 354 | QTAILQ_INSERT_TAIL(&block_backends, blk, link); |
26f54e9a MA |
355 | return blk; |
356 | } | |
357 | ||
a3aeeab5 EB |
358 | /* |
359 | * Create a new BlockBackend connected to an existing BlockDriverState. | |
360 | * | |
361 | * @perm is a bitmasks of BLK_PERM_* constants which describes the | |
362 | * permissions to request for @bs that is attached to this | |
363 | * BlockBackend. @shared_perm is a bitmask which describes which | |
364 | * permissions may be granted to other users of the attached node. | |
365 | * Both sets of permissions can be changed later using blk_set_perm(). | |
366 | * | |
367 | * Return the new BlockBackend on success, null on failure. | |
368 | */ | |
369 | BlockBackend *blk_new_with_bs(BlockDriverState *bs, uint64_t perm, | |
370 | uint64_t shared_perm, Error **errp) | |
371 | { | |
372 | BlockBackend *blk = blk_new(bdrv_get_aio_context(bs), perm, shared_perm); | |
373 | ||
374 | if (blk_insert_bs(blk, bs, errp) < 0) { | |
375 | blk_unref(blk); | |
376 | return NULL; | |
377 | } | |
378 | return blk; | |
379 | } | |
380 | ||
7e7d56d9 | 381 | /* |
28eb9b12 | 382 | * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. |
d861ab3a | 383 | * The new BlockBackend is in the main AioContext. |
ca49a4fd HR |
384 | * |
385 | * Just as with bdrv_open(), after having called this function the reference to | |
386 | * @options belongs to the block layer (even on failure). | |
387 | * | |
388 | * TODO: Remove @filename and @flags; it should be possible to specify a whole | |
389 | * BDS tree just by specifying the @options QDict (or @reference, | |
390 | * alternatively). At the time of adding this function, this is not possible, | |
391 | * though, so callers of this function have to be able to specify @filename and | |
392 | * @flags. | |
393 | */ | |
efaa7c4e HR |
394 | BlockBackend *blk_new_open(const char *filename, const char *reference, |
395 | QDict *options, int flags, Error **errp) | |
ca49a4fd HR |
396 | { |
397 | BlockBackend *blk; | |
28eb9b12 | 398 | BlockDriverState *bs; |
1f4ad7d3 | 399 | uint64_t perm = 0; |
c62d32f5 KW |
400 | |
401 | /* blk_new_open() is mainly used in .bdrv_create implementations and the | |
402 | * tools where sharing isn't a concern because the BDS stays private, so we | |
403 | * just request permission according to the flags. | |
404 | * | |
405 | * The exceptions are xen_disk and blockdev_init(); in these cases, the | |
406 | * caller of blk_new_open() doesn't make use of the permissions, but they | |
407 | * shouldn't hurt either. We can still share everything here because the | |
408 | * guest devices will add their own blockers if they can't share. */ | |
1f4ad7d3 KW |
409 | if ((flags & BDRV_O_NO_IO) == 0) { |
410 | perm |= BLK_PERM_CONSISTENT_READ; | |
411 | if (flags & BDRV_O_RDWR) { | |
412 | perm |= BLK_PERM_WRITE; | |
413 | } | |
c62d32f5 KW |
414 | } |
415 | if (flags & BDRV_O_RESIZE) { | |
416 | perm |= BLK_PERM_RESIZE; | |
417 | } | |
ca49a4fd | 418 | |
d861ab3a | 419 | blk = blk_new(qemu_get_aio_context(), perm, BLK_PERM_ALL); |
5b363937 HR |
420 | bs = bdrv_open(filename, reference, options, flags, errp); |
421 | if (!bs) { | |
ca49a4fd HR |
422 | blk_unref(blk); |
423 | return NULL; | |
424 | } | |
425 | ||
132ada80 | 426 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx, |
50bfbe93 FZ |
427 | perm, BLK_PERM_ALL, blk, errp); |
428 | if (!blk->root) { | |
50bfbe93 FZ |
429 | blk_unref(blk); |
430 | return NULL; | |
431 | } | |
72e775c7 | 432 | |
ca49a4fd HR |
433 | return blk; |
434 | } | |
435 | ||
26f54e9a MA |
436 | static void blk_delete(BlockBackend *blk) |
437 | { | |
438 | assert(!blk->refcnt); | |
e5e78550 | 439 | assert(!blk->name); |
a7f53e26 | 440 | assert(!blk->dev); |
022cdc9f | 441 | if (blk->public.throttle_group_member.throttle_state) { |
1606e4cf EB |
442 | blk_io_limits_disable(blk); |
443 | } | |
f21d96d0 | 444 | if (blk->root) { |
13855c6b | 445 | blk_remove_bs(blk); |
7e7d56d9 | 446 | } |
5f7772c4 FZ |
447 | if (blk->vmsh) { |
448 | qemu_del_vm_change_state_handler(blk->vmsh); | |
449 | blk->vmsh = NULL; | |
450 | } | |
3301f6c6 HR |
451 | assert(QLIST_EMPTY(&blk->remove_bs_notifiers.notifiers)); |
452 | assert(QLIST_EMPTY(&blk->insert_bs_notifiers.notifiers)); | |
d03654ea | 453 | assert(QLIST_EMPTY(&blk->aio_notifiers)); |
2cf22d6a | 454 | QTAILQ_REMOVE(&block_backends, blk, link); |
18e46a03 | 455 | drive_info_del(blk->legacy_dinfo); |
979e9b03 | 456 | block_acct_cleanup(&blk->stats); |
26f54e9a MA |
457 | g_free(blk); |
458 | } | |
459 | ||
8fb3c76c MA |
460 | static void drive_info_del(DriveInfo *dinfo) |
461 | { | |
462 | if (!dinfo) { | |
463 | return; | |
464 | } | |
465 | qemu_opts_del(dinfo->opts); | |
8fb3c76c MA |
466 | g_free(dinfo); |
467 | } | |
468 | ||
f636ae85 AG |
469 | int blk_get_refcnt(BlockBackend *blk) |
470 | { | |
471 | return blk ? blk->refcnt : 0; | |
472 | } | |
473 | ||
26f54e9a MA |
474 | /* |
475 | * Increment @blk's reference count. | |
476 | * @blk must not be null. | |
477 | */ | |
478 | void blk_ref(BlockBackend *blk) | |
479 | { | |
5ca9d21b | 480 | assert(blk->refcnt > 0); |
26f54e9a MA |
481 | blk->refcnt++; |
482 | } | |
483 | ||
484 | /* | |
485 | * Decrement @blk's reference count. | |
486 | * If this drops it to zero, destroy @blk. | |
487 | * For convenience, do nothing if @blk is null. | |
488 | */ | |
489 | void blk_unref(BlockBackend *blk) | |
490 | { | |
491 | if (blk) { | |
492 | assert(blk->refcnt > 0); | |
5ca9d21b KW |
493 | if (blk->refcnt > 1) { |
494 | blk->refcnt--; | |
495 | } else { | |
496 | blk_drain(blk); | |
497 | /* blk_drain() cannot resurrect blk, nobody held a reference */ | |
498 | assert(blk->refcnt == 1); | |
499 | blk->refcnt = 0; | |
26f54e9a MA |
500 | blk_delete(blk); |
501 | } | |
502 | } | |
503 | } | |
504 | ||
2cf22d6a HR |
505 | /* |
506 | * Behaves similarly to blk_next() but iterates over all BlockBackends, even the | |
507 | * ones which are hidden (i.e. are not referenced by the monitor). | |
508 | */ | |
a429b9b5 | 509 | BlockBackend *blk_all_next(BlockBackend *blk) |
2cf22d6a HR |
510 | { |
511 | return blk ? QTAILQ_NEXT(blk, link) | |
512 | : QTAILQ_FIRST(&block_backends); | |
513 | } | |
514 | ||
d8da3cef HR |
515 | void blk_remove_all_bs(void) |
516 | { | |
74d1b8fc | 517 | BlockBackend *blk = NULL; |
d8da3cef | 518 | |
2cf22d6a | 519 | while ((blk = blk_all_next(blk)) != NULL) { |
d8da3cef HR |
520 | AioContext *ctx = blk_get_aio_context(blk); |
521 | ||
522 | aio_context_acquire(ctx); | |
f21d96d0 | 523 | if (blk->root) { |
d8da3cef HR |
524 | blk_remove_bs(blk); |
525 | } | |
526 | aio_context_release(ctx); | |
527 | } | |
528 | } | |
529 | ||
26f54e9a | 530 | /* |
9492b0b9 | 531 | * Return the monitor-owned BlockBackend after @blk. |
26f54e9a MA |
532 | * If @blk is null, return the first one. |
533 | * Else, return @blk's next sibling, which may be null. | |
534 | * | |
535 | * To iterate over all BlockBackends, do | |
536 | * for (blk = blk_next(NULL); blk; blk = blk_next(blk)) { | |
537 | * ... | |
538 | * } | |
539 | */ | |
540 | BlockBackend *blk_next(BlockBackend *blk) | |
541 | { | |
9492b0b9 HR |
542 | return blk ? QTAILQ_NEXT(blk, monitor_link) |
543 | : QTAILQ_FIRST(&monitor_block_backends); | |
26f54e9a MA |
544 | } |
545 | ||
7c8eece4 KW |
546 | /* Iterates over all top-level BlockDriverStates, i.e. BDSs that are owned by |
547 | * the monitor or attached to a BlockBackend */ | |
88be7b4b | 548 | BlockDriverState *bdrv_next(BdrvNextIterator *it) |
7c8eece4 | 549 | { |
5e003f17 HR |
550 | BlockDriverState *bs, *old_bs; |
551 | ||
552 | /* Must be called from the main loop */ | |
553 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
981f4f57 | 554 | |
7c8eece4 KW |
555 | /* First, return all root nodes of BlockBackends. In order to avoid |
556 | * returning a BDS twice when multiple BBs refer to it, we only return it | |
557 | * if the BB is the first one in the parent list of the BDS. */ | |
558 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
5e003f17 HR |
559 | BlockBackend *old_blk = it->blk; |
560 | ||
561 | old_bs = old_blk ? blk_bs(old_blk) : NULL; | |
562 | ||
7c8eece4 KW |
563 | do { |
564 | it->blk = blk_all_next(it->blk); | |
88be7b4b KW |
565 | bs = it->blk ? blk_bs(it->blk) : NULL; |
566 | } while (it->blk && (bs == NULL || bdrv_first_blk(bs) != it->blk)); | |
7c8eece4 | 567 | |
5e003f17 HR |
568 | if (it->blk) { |
569 | blk_ref(it->blk); | |
570 | } | |
571 | blk_unref(old_blk); | |
572 | ||
88be7b4b | 573 | if (bs) { |
5e003f17 HR |
574 | bdrv_ref(bs); |
575 | bdrv_unref(old_bs); | |
88be7b4b | 576 | return bs; |
7c8eece4 KW |
577 | } |
578 | it->phase = BDRV_NEXT_MONITOR_OWNED; | |
5e003f17 HR |
579 | } else { |
580 | old_bs = it->bs; | |
7c8eece4 KW |
581 | } |
582 | ||
583 | /* Then return the monitor-owned BDSes without a BB attached. Ignore all | |
584 | * BDSes that are attached to a BlockBackend here; they have been handled | |
585 | * by the above block already */ | |
981f4f57 | 586 | do { |
7c8eece4 | 587 | it->bs = bdrv_next_monitor_owned(it->bs); |
88be7b4b KW |
588 | bs = it->bs; |
589 | } while (bs && bdrv_has_blk(bs)); | |
590 | ||
5e003f17 HR |
591 | if (bs) { |
592 | bdrv_ref(bs); | |
593 | } | |
594 | bdrv_unref(old_bs); | |
595 | ||
88be7b4b KW |
596 | return bs; |
597 | } | |
598 | ||
5e003f17 | 599 | static void bdrv_next_reset(BdrvNextIterator *it) |
88be7b4b KW |
600 | { |
601 | *it = (BdrvNextIterator) { | |
602 | .phase = BDRV_NEXT_BACKEND_ROOTS, | |
603 | }; | |
5e003f17 | 604 | } |
981f4f57 | 605 | |
5e003f17 HR |
606 | BlockDriverState *bdrv_first(BdrvNextIterator *it) |
607 | { | |
608 | bdrv_next_reset(it); | |
88be7b4b | 609 | return bdrv_next(it); |
981f4f57 HR |
610 | } |
611 | ||
5e003f17 HR |
612 | /* Must be called when aborting a bdrv_next() iteration before |
613 | * bdrv_next() returns NULL */ | |
614 | void bdrv_next_cleanup(BdrvNextIterator *it) | |
615 | { | |
616 | /* Must be called from the main loop */ | |
617 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
618 | ||
619 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
620 | if (it->blk) { | |
621 | bdrv_unref(blk_bs(it->blk)); | |
622 | blk_unref(it->blk); | |
623 | } | |
624 | } else { | |
625 | bdrv_unref(it->bs); | |
626 | } | |
627 | ||
628 | bdrv_next_reset(it); | |
629 | } | |
630 | ||
e5e78550 HR |
631 | /* |
632 | * Add a BlockBackend into the list of backends referenced by the monitor, with | |
633 | * the given @name acting as the handle for the monitor. | |
634 | * Strictly for use by blockdev.c. | |
635 | * | |
636 | * @name must not be null or empty. | |
637 | * | |
638 | * Returns true on success and false on failure. In the latter case, an Error | |
639 | * object is returned through @errp. | |
640 | */ | |
641 | bool monitor_add_blk(BlockBackend *blk, const char *name, Error **errp) | |
642 | { | |
643 | assert(!blk->name); | |
644 | assert(name && name[0]); | |
645 | ||
646 | if (!id_wellformed(name)) { | |
647 | error_setg(errp, "Invalid device name"); | |
648 | return false; | |
649 | } | |
650 | if (blk_by_name(name)) { | |
651 | error_setg(errp, "Device with id '%s' already exists", name); | |
652 | return false; | |
653 | } | |
654 | if (bdrv_find_node(name)) { | |
655 | error_setg(errp, | |
656 | "Device name '%s' conflicts with an existing node name", | |
657 | name); | |
658 | return false; | |
659 | } | |
660 | ||
661 | blk->name = g_strdup(name); | |
662 | QTAILQ_INSERT_TAIL(&monitor_block_backends, blk, monitor_link); | |
663 | return true; | |
664 | } | |
665 | ||
666 | /* | |
667 | * Remove a BlockBackend from the list of backends referenced by the monitor. | |
668 | * Strictly for use by blockdev.c. | |
669 | */ | |
670 | void monitor_remove_blk(BlockBackend *blk) | |
671 | { | |
672 | if (!blk->name) { | |
673 | return; | |
674 | } | |
675 | ||
676 | QTAILQ_REMOVE(&monitor_block_backends, blk, monitor_link); | |
677 | g_free(blk->name); | |
678 | blk->name = NULL; | |
679 | } | |
680 | ||
26f54e9a | 681 | /* |
7e7d56d9 | 682 | * Return @blk's name, a non-null string. |
e5e78550 | 683 | * Returns an empty string iff @blk is not referenced by the monitor. |
26f54e9a | 684 | */ |
0731a50f | 685 | const char *blk_name(const BlockBackend *blk) |
26f54e9a | 686 | { |
e5e78550 | 687 | return blk->name ?: ""; |
26f54e9a MA |
688 | } |
689 | ||
690 | /* | |
691 | * Return the BlockBackend with name @name if it exists, else null. | |
692 | * @name must not be null. | |
693 | */ | |
694 | BlockBackend *blk_by_name(const char *name) | |
695 | { | |
74d1b8fc | 696 | BlockBackend *blk = NULL; |
26f54e9a MA |
697 | |
698 | assert(name); | |
74d1b8fc | 699 | while ((blk = blk_next(blk)) != NULL) { |
26f54e9a MA |
700 | if (!strcmp(name, blk->name)) { |
701 | return blk; | |
702 | } | |
703 | } | |
704 | return NULL; | |
705 | } | |
7e7d56d9 MA |
706 | |
707 | /* | |
708 | * Return the BlockDriverState attached to @blk if any, else null. | |
709 | */ | |
710 | BlockDriverState *blk_bs(BlockBackend *blk) | |
711 | { | |
f21d96d0 | 712 | return blk->root ? blk->root->bs : NULL; |
7e7d56d9 MA |
713 | } |
714 | ||
7c8eece4 | 715 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs) |
dde33812 KW |
716 | { |
717 | BdrvChild *child; | |
718 | QLIST_FOREACH(child, &bs->parents, next_parent) { | |
719 | if (child->role == &child_root) { | |
7c8eece4 | 720 | return child->opaque; |
dde33812 KW |
721 | } |
722 | } | |
723 | ||
7c8eece4 KW |
724 | return NULL; |
725 | } | |
726 | ||
727 | /* | |
728 | * Returns true if @bs has an associated BlockBackend. | |
729 | */ | |
730 | bool bdrv_has_blk(BlockDriverState *bs) | |
731 | { | |
732 | return bdrv_first_blk(bs) != NULL; | |
dde33812 KW |
733 | } |
734 | ||
b6c1bae5 KW |
735 | /* |
736 | * Returns true if @bs has only BlockBackends as parents. | |
737 | */ | |
738 | bool bdrv_is_root_node(BlockDriverState *bs) | |
739 | { | |
740 | BdrvChild *c; | |
741 | ||
742 | QLIST_FOREACH(c, &bs->parents, next_parent) { | |
743 | if (c->role != &child_root) { | |
744 | return false; | |
745 | } | |
746 | } | |
747 | ||
748 | return true; | |
749 | } | |
750 | ||
18e46a03 MA |
751 | /* |
752 | * Return @blk's DriveInfo if any, else null. | |
753 | */ | |
754 | DriveInfo *blk_legacy_dinfo(BlockBackend *blk) | |
755 | { | |
756 | return blk->legacy_dinfo; | |
757 | } | |
758 | ||
759 | /* | |
760 | * Set @blk's DriveInfo to @dinfo, and return it. | |
761 | * @blk must not have a DriveInfo set already. | |
762 | * No other BlockBackend may have the same DriveInfo set. | |
763 | */ | |
764 | DriveInfo *blk_set_legacy_dinfo(BlockBackend *blk, DriveInfo *dinfo) | |
765 | { | |
766 | assert(!blk->legacy_dinfo); | |
767 | return blk->legacy_dinfo = dinfo; | |
768 | } | |
769 | ||
770 | /* | |
771 | * Return the BlockBackend with DriveInfo @dinfo. | |
772 | * It must exist. | |
773 | */ | |
774 | BlockBackend *blk_by_legacy_dinfo(DriveInfo *dinfo) | |
775 | { | |
74d1b8fc | 776 | BlockBackend *blk = NULL; |
18e46a03 | 777 | |
74d1b8fc | 778 | while ((blk = blk_next(blk)) != NULL) { |
18e46a03 MA |
779 | if (blk->legacy_dinfo == dinfo) { |
780 | return blk; | |
781 | } | |
782 | } | |
783 | abort(); | |
784 | } | |
785 | ||
f2cd875d KW |
786 | /* |
787 | * Returns a pointer to the publicly accessible fields of @blk. | |
788 | */ | |
789 | BlockBackendPublic *blk_get_public(BlockBackend *blk) | |
790 | { | |
791 | return &blk->public; | |
792 | } | |
793 | ||
794 | /* | |
795 | * Returns a BlockBackend given the associated @public fields. | |
796 | */ | |
797 | BlockBackend *blk_by_public(BlockBackendPublic *public) | |
798 | { | |
799 | return container_of(public, BlockBackend, public); | |
800 | } | |
801 | ||
1c95f7e1 HR |
802 | /* |
803 | * Disassociates the currently associated BlockDriverState from @blk. | |
804 | */ | |
805 | void blk_remove_bs(BlockBackend *blk) | |
806 | { | |
c89bcf3a | 807 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
632a7735 | 808 | BlockDriverState *bs; |
022cdc9f | 809 | |
3301f6c6 | 810 | notifier_list_notify(&blk->remove_bs_notifiers, blk); |
c89bcf3a | 811 | if (tgm->throttle_state) { |
632a7735 ZL |
812 | bs = blk_bs(blk); |
813 | bdrv_drained_begin(bs); | |
c89bcf3a AG |
814 | throttle_group_detach_aio_context(tgm); |
815 | throttle_group_attach_aio_context(tgm, qemu_get_aio_context()); | |
632a7735 | 816 | bdrv_drained_end(bs); |
a5614993 | 817 | } |
1c95f7e1 | 818 | |
7ca7f0f6 KW |
819 | blk_update_root_state(blk); |
820 | ||
f45280cb GK |
821 | /* bdrv_root_unref_child() will cause blk->root to become stale and may |
822 | * switch to a completion coroutine later on. Let's drain all I/O here | |
823 | * to avoid that and a potential QEMU crash. | |
824 | */ | |
825 | blk_drain(blk); | |
f21d96d0 KW |
826 | bdrv_root_unref_child(blk->root); |
827 | blk->root = NULL; | |
1c95f7e1 HR |
828 | } |
829 | ||
0c3c36d6 HR |
830 | /* |
831 | * Associates a new BlockDriverState with @blk. | |
832 | */ | |
d7086422 | 833 | int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) |
0c3c36d6 | 834 | { |
c89bcf3a | 835 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
b441dc71 | 836 | bdrv_ref(bs); |
132ada80 | 837 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx, |
d7086422 KW |
838 | blk->perm, blk->shared_perm, blk, errp); |
839 | if (blk->root == NULL) { | |
840 | return -EPERM; | |
841 | } | |
3301f6c6 HR |
842 | |
843 | notifier_list_notify(&blk->insert_bs_notifiers, blk); | |
c89bcf3a AG |
844 | if (tgm->throttle_state) { |
845 | throttle_group_detach_aio_context(tgm); | |
846 | throttle_group_attach_aio_context(tgm, bdrv_get_aio_context(bs)); | |
7ca7f0f6 | 847 | } |
d7086422 KW |
848 | |
849 | return 0; | |
0c3c36d6 HR |
850 | } |
851 | ||
981776b3 KW |
852 | /* |
853 | * Sets the permission bitmasks that the user of the BlockBackend needs. | |
854 | */ | |
855 | int blk_set_perm(BlockBackend *blk, uint64_t perm, uint64_t shared_perm, | |
856 | Error **errp) | |
857 | { | |
858 | int ret; | |
859 | ||
d35ff5e6 | 860 | if (blk->root && !blk->disable_perm) { |
981776b3 KW |
861 | ret = bdrv_child_try_set_perm(blk->root, perm, shared_perm, errp); |
862 | if (ret < 0) { | |
863 | return ret; | |
864 | } | |
865 | } | |
866 | ||
867 | blk->perm = perm; | |
868 | blk->shared_perm = shared_perm; | |
869 | ||
870 | return 0; | |
871 | } | |
872 | ||
887354bd KW |
873 | void blk_get_perm(BlockBackend *blk, uint64_t *perm, uint64_t *shared_perm) |
874 | { | |
875 | *perm = blk->perm; | |
876 | *shared_perm = blk->shared_perm; | |
877 | } | |
878 | ||
d09ea2d2 TH |
879 | /* |
880 | * Attach device model @dev to @blk. | |
881 | * Return 0 on success, -EBUSY when a device model is attached already. | |
882 | */ | |
883 | int blk_attach_dev(BlockBackend *blk, DeviceState *dev) | |
4be74634 | 884 | { |
a7f53e26 MA |
885 | if (blk->dev) { |
886 | return -EBUSY; | |
887 | } | |
d35ff5e6 KW |
888 | |
889 | /* While migration is still incoming, we don't need to apply the | |
890 | * permissions of guest device BlockBackends. We might still have a block | |
891 | * job or NBD server writing to the image for storage migration. */ | |
892 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
893 | blk->disable_perm = true; | |
894 | } | |
895 | ||
84ebe375 | 896 | blk_ref(blk); |
a7f53e26 | 897 | blk->dev = dev; |
373340b2 | 898 | blk_iostatus_reset(blk); |
d35ff5e6 | 899 | |
a7f53e26 | 900 | return 0; |
4be74634 MA |
901 | } |
902 | ||
a7f53e26 MA |
903 | /* |
904 | * Detach device model @dev from @blk. | |
905 | * @dev must be currently attached to @blk. | |
906 | */ | |
d09ea2d2 | 907 | void blk_detach_dev(BlockBackend *blk, DeviceState *dev) |
4be74634 | 908 | { |
a7f53e26 MA |
909 | assert(blk->dev == dev); |
910 | blk->dev = NULL; | |
911 | blk->dev_ops = NULL; | |
912 | blk->dev_opaque = NULL; | |
68e9ec01 | 913 | blk->guest_block_size = 512; |
981776b3 | 914 | blk_set_perm(blk, 0, BLK_PERM_ALL, &error_abort); |
84ebe375 | 915 | blk_unref(blk); |
4be74634 MA |
916 | } |
917 | ||
a7f53e26 MA |
918 | /* |
919 | * Return the device model attached to @blk if any, else null. | |
920 | */ | |
d09ea2d2 | 921 | DeviceState *blk_get_attached_dev(BlockBackend *blk) |
a7f53e26 MA |
922 | { |
923 | return blk->dev; | |
924 | } | |
925 | ||
2d76e724 KW |
926 | /* Return the qdev ID, or if no ID is assigned the QOM path, of the block |
927 | * device attached to the BlockBackend. */ | |
77beef83 | 928 | char *blk_get_attached_dev_id(BlockBackend *blk) |
2d76e724 | 929 | { |
d09ea2d2 | 930 | DeviceState *dev = blk->dev; |
2d76e724 KW |
931 | |
932 | if (!dev) { | |
933 | return g_strdup(""); | |
934 | } else if (dev->id) { | |
935 | return g_strdup(dev->id); | |
936 | } | |
602414d1 LM |
937 | |
938 | return object_get_canonical_path(OBJECT(dev)) ?: g_strdup(""); | |
2d76e724 KW |
939 | } |
940 | ||
1c89e1fa KW |
941 | /* |
942 | * Return the BlockBackend which has the device model @dev attached if it | |
943 | * exists, else null. | |
944 | * | |
945 | * @dev must not be null. | |
946 | */ | |
947 | BlockBackend *blk_by_dev(void *dev) | |
948 | { | |
949 | BlockBackend *blk = NULL; | |
950 | ||
951 | assert(dev != NULL); | |
952 | while ((blk = blk_all_next(blk)) != NULL) { | |
953 | if (blk->dev == dev) { | |
954 | return blk; | |
955 | } | |
956 | } | |
957 | return NULL; | |
958 | } | |
959 | ||
a7f53e26 MA |
960 | /* |
961 | * Set @blk's device model callbacks to @ops. | |
962 | * @opaque is the opaque argument to pass to the callbacks. | |
963 | * This is for use by device models. | |
964 | */ | |
965 | void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, | |
966 | void *opaque) | |
967 | { | |
968 | blk->dev_ops = ops; | |
969 | blk->dev_opaque = opaque; | |
f4d9cc88 JS |
970 | |
971 | /* Are we currently quiesced? Should we enforce this right now? */ | |
972 | if (blk->quiesce_counter && ops->drained_begin) { | |
973 | ops->drained_begin(opaque); | |
974 | } | |
a7f53e26 MA |
975 | } |
976 | ||
977 | /* | |
978 | * Notify @blk's attached device model of media change. | |
39829a01 KW |
979 | * |
980 | * If @load is true, notify of media load. This action can fail, meaning that | |
981 | * the medium cannot be loaded. @errp is set then. | |
982 | * | |
983 | * If @load is false, notify of media eject. This can never fail. | |
984 | * | |
a7f53e26 MA |
985 | * Also send DEVICE_TRAY_MOVED events as appropriate. |
986 | */ | |
39829a01 | 987 | void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp) |
a7f53e26 MA |
988 | { |
989 | if (blk->dev_ops && blk->dev_ops->change_media_cb) { | |
f1f57066 | 990 | bool tray_was_open, tray_is_open; |
39829a01 | 991 | Error *local_err = NULL; |
a7f53e26 | 992 | |
f1f57066 | 993 | tray_was_open = blk_dev_is_tray_open(blk); |
39829a01 KW |
994 | blk->dev_ops->change_media_cb(blk->dev_opaque, load, &local_err); |
995 | if (local_err) { | |
996 | assert(load == true); | |
997 | error_propagate(errp, local_err); | |
998 | return; | |
999 | } | |
f1f57066 HR |
1000 | tray_is_open = blk_dev_is_tray_open(blk); |
1001 | ||
1002 | if (tray_was_open != tray_is_open) { | |
2d76e724 | 1003 | char *id = blk_get_attached_dev_id(blk); |
3ab72385 | 1004 | qapi_event_send_device_tray_moved(blk_name(blk), id, tray_is_open); |
2d76e724 | 1005 | g_free(id); |
a7f53e26 MA |
1006 | } |
1007 | } | |
1008 | } | |
1009 | ||
5c8cab48 KW |
1010 | static void blk_root_change_media(BdrvChild *child, bool load) |
1011 | { | |
39829a01 | 1012 | blk_dev_change_media_cb(child->opaque, load, NULL); |
5c8cab48 KW |
1013 | } |
1014 | ||
a7f53e26 MA |
1015 | /* |
1016 | * Does @blk's attached device model have removable media? | |
1017 | * %true if no device model is attached. | |
1018 | */ | |
1019 | bool blk_dev_has_removable_media(BlockBackend *blk) | |
1020 | { | |
1021 | return !blk->dev || (blk->dev_ops && blk->dev_ops->change_media_cb); | |
1022 | } | |
1023 | ||
8f3a73bc HR |
1024 | /* |
1025 | * Does @blk's attached device model have a tray? | |
1026 | */ | |
1027 | bool blk_dev_has_tray(BlockBackend *blk) | |
1028 | { | |
1029 | return blk->dev_ops && blk->dev_ops->is_tray_open; | |
1030 | } | |
1031 | ||
a7f53e26 MA |
1032 | /* |
1033 | * Notify @blk's attached device model of a media eject request. | |
1034 | * If @force is true, the medium is about to be yanked out forcefully. | |
1035 | */ | |
1036 | void blk_dev_eject_request(BlockBackend *blk, bool force) | |
4be74634 | 1037 | { |
a7f53e26 MA |
1038 | if (blk->dev_ops && blk->dev_ops->eject_request_cb) { |
1039 | blk->dev_ops->eject_request_cb(blk->dev_opaque, force); | |
1040 | } | |
4be74634 MA |
1041 | } |
1042 | ||
a7f53e26 MA |
1043 | /* |
1044 | * Does @blk's attached device model have a tray, and is it open? | |
1045 | */ | |
1046 | bool blk_dev_is_tray_open(BlockBackend *blk) | |
4be74634 | 1047 | { |
8f3a73bc | 1048 | if (blk_dev_has_tray(blk)) { |
a7f53e26 MA |
1049 | return blk->dev_ops->is_tray_open(blk->dev_opaque); |
1050 | } | |
1051 | return false; | |
1052 | } | |
1053 | ||
1054 | /* | |
1055 | * Does @blk's attached device model have the medium locked? | |
1056 | * %false if the device model has no such lock. | |
1057 | */ | |
1058 | bool blk_dev_is_medium_locked(BlockBackend *blk) | |
1059 | { | |
1060 | if (blk->dev_ops && blk->dev_ops->is_medium_locked) { | |
1061 | return blk->dev_ops->is_medium_locked(blk->dev_opaque); | |
1062 | } | |
1063 | return false; | |
1064 | } | |
1065 | ||
1066 | /* | |
1067 | * Notify @blk's attached device model of a backend size change. | |
1068 | */ | |
5c8cab48 | 1069 | static void blk_root_resize(BdrvChild *child) |
a7f53e26 | 1070 | { |
5c8cab48 KW |
1071 | BlockBackend *blk = child->opaque; |
1072 | ||
a7f53e26 MA |
1073 | if (blk->dev_ops && blk->dev_ops->resize_cb) { |
1074 | blk->dev_ops->resize_cb(blk->dev_opaque); | |
1075 | } | |
1076 | } | |
1077 | ||
1078 | void blk_iostatus_enable(BlockBackend *blk) | |
1079 | { | |
373340b2 HR |
1080 | blk->iostatus_enabled = true; |
1081 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
1082 | } | |
1083 | ||
1084 | /* The I/O status is only enabled if the drive explicitly | |
1085 | * enables it _and_ the VM is configured to stop on errors */ | |
1086 | bool blk_iostatus_is_enabled(const BlockBackend *blk) | |
1087 | { | |
1088 | return (blk->iostatus_enabled && | |
1089 | (blk->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || | |
1090 | blk->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
1091 | blk->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
1092 | } | |
1093 | ||
1094 | BlockDeviceIoStatus blk_iostatus(const BlockBackend *blk) | |
1095 | { | |
1096 | return blk->iostatus; | |
1097 | } | |
1098 | ||
1099 | void blk_iostatus_disable(BlockBackend *blk) | |
1100 | { | |
1101 | blk->iostatus_enabled = false; | |
1102 | } | |
1103 | ||
1104 | void blk_iostatus_reset(BlockBackend *blk) | |
1105 | { | |
1106 | if (blk_iostatus_is_enabled(blk)) { | |
1107 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
373340b2 HR |
1108 | } |
1109 | } | |
1110 | ||
1111 | void blk_iostatus_set_err(BlockBackend *blk, int error) | |
1112 | { | |
1113 | assert(blk_iostatus_is_enabled(blk)); | |
1114 | if (blk->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
1115 | blk->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : | |
1116 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
1117 | } | |
4be74634 MA |
1118 | } |
1119 | ||
c10c9d96 KW |
1120 | void blk_set_allow_write_beyond_eof(BlockBackend *blk, bool allow) |
1121 | { | |
1122 | blk->allow_write_beyond_eof = allow; | |
1123 | } | |
1124 | ||
980b0f94 KW |
1125 | void blk_set_allow_aio_context_change(BlockBackend *blk, bool allow) |
1126 | { | |
1127 | blk->allow_aio_context_change = allow; | |
1128 | } | |
1129 | ||
cf312932 KW |
1130 | void blk_set_disable_request_queuing(BlockBackend *blk, bool disable) |
1131 | { | |
1132 | blk->disable_request_queuing = disable; | |
1133 | } | |
1134 | ||
e7f7d676 HR |
1135 | static int blk_check_byte_request(BlockBackend *blk, int64_t offset, |
1136 | size_t size) | |
1137 | { | |
1138 | int64_t len; | |
1139 | ||
1140 | if (size > INT_MAX) { | |
1141 | return -EIO; | |
1142 | } | |
1143 | ||
c09ba36c | 1144 | if (!blk_is_available(blk)) { |
e7f7d676 HR |
1145 | return -ENOMEDIUM; |
1146 | } | |
1147 | ||
e7f7d676 HR |
1148 | if (offset < 0) { |
1149 | return -EIO; | |
1150 | } | |
1151 | ||
c10c9d96 KW |
1152 | if (!blk->allow_write_beyond_eof) { |
1153 | len = blk_getlength(blk); | |
1154 | if (len < 0) { | |
1155 | return len; | |
1156 | } | |
1157 | ||
1158 | if (offset > len || len - offset < size) { | |
1159 | return -EIO; | |
1160 | } | |
e7f7d676 HR |
1161 | } |
1162 | ||
1163 | return 0; | |
1164 | } | |
1165 | ||
7f16476f | 1166 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
cf312932 KW |
1167 | static void coroutine_fn blk_wait_while_drained(BlockBackend *blk) |
1168 | { | |
7f16476f KW |
1169 | assert(blk->in_flight > 0); |
1170 | ||
cf312932 | 1171 | if (blk->quiesce_counter && !blk->disable_request_queuing) { |
7f16476f | 1172 | blk_dec_in_flight(blk); |
cf312932 | 1173 | qemu_co_queue_wait(&blk->queued_requests, NULL); |
7f16476f | 1174 | blk_inc_in_flight(blk); |
cf312932 KW |
1175 | } |
1176 | } | |
1177 | ||
fbb92b67 KW |
1178 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1179 | static int coroutine_fn | |
1180 | blk_do_preadv(BlockBackend *blk, int64_t offset, unsigned int bytes, | |
1181 | QEMUIOVector *qiov, BdrvRequestFlags flags) | |
4be74634 | 1182 | { |
1e98fefd | 1183 | int ret; |
cf312932 | 1184 | BlockDriverState *bs; |
1e98fefd | 1185 | |
cf312932 | 1186 | blk_wait_while_drained(blk); |
1e98fefd | 1187 | |
cf312932 KW |
1188 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1189 | bs = blk_bs(blk); | |
99723548 | 1190 | trace_blk_co_preadv(blk, bs, offset, bytes, flags); |
1e98fefd KW |
1191 | |
1192 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1193 | if (ret < 0) { |
1194 | return ret; | |
1195 | } | |
1196 | ||
99723548 PB |
1197 | bdrv_inc_in_flight(bs); |
1198 | ||
441565b2 | 1199 | /* throttling disk I/O */ |
022cdc9f MP |
1200 | if (blk->public.throttle_group_member.throttle_state) { |
1201 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1202 | bytes, false); | |
441565b2 KW |
1203 | } |
1204 | ||
99723548 PB |
1205 | ret = bdrv_co_preadv(blk->root, offset, bytes, qiov, flags); |
1206 | bdrv_dec_in_flight(bs); | |
1207 | return ret; | |
1bf1cbc9 KW |
1208 | } |
1209 | ||
fbb92b67 KW |
1210 | int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset, |
1211 | unsigned int bytes, QEMUIOVector *qiov, | |
1212 | BdrvRequestFlags flags) | |
1213 | { | |
1214 | int ret; | |
1215 | ||
1216 | blk_inc_in_flight(blk); | |
1217 | ret = blk_do_preadv(blk, offset, bytes, qiov, flags); | |
1218 | blk_dec_in_flight(blk); | |
1219 | ||
1220 | return ret; | |
1221 | } | |
1222 | ||
1223 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ | |
1224 | static int coroutine_fn | |
1225 | blk_do_pwritev_part(BlockBackend *blk, int64_t offset, unsigned int bytes, | |
1226 | QEMUIOVector *qiov, size_t qiov_offset, | |
1227 | BdrvRequestFlags flags) | |
a8823a3b | 1228 | { |
bfd18d1e | 1229 | int ret; |
cf312932 | 1230 | BlockDriverState *bs; |
bfd18d1e | 1231 | |
cf312932 | 1232 | blk_wait_while_drained(blk); |
bfd18d1e | 1233 | |
cf312932 KW |
1234 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1235 | bs = blk_bs(blk); | |
99723548 | 1236 | trace_blk_co_pwritev(blk, bs, offset, bytes, flags); |
1e98fefd | 1237 | |
bfd18d1e | 1238 | ret = blk_check_byte_request(blk, offset, bytes); |
a8823a3b KW |
1239 | if (ret < 0) { |
1240 | return ret; | |
1241 | } | |
1242 | ||
99723548 | 1243 | bdrv_inc_in_flight(bs); |
441565b2 | 1244 | /* throttling disk I/O */ |
022cdc9f MP |
1245 | if (blk->public.throttle_group_member.throttle_state) { |
1246 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1247 | bytes, true); | |
441565b2 KW |
1248 | } |
1249 | ||
bfd18d1e KW |
1250 | if (!blk->enable_write_cache) { |
1251 | flags |= BDRV_REQ_FUA; | |
1252 | } | |
1253 | ||
b3016864 VSO |
1254 | ret = bdrv_co_pwritev_part(blk->root, offset, bytes, qiov, qiov_offset, |
1255 | flags); | |
99723548 PB |
1256 | bdrv_dec_in_flight(bs); |
1257 | return ret; | |
a8823a3b KW |
1258 | } |
1259 | ||
fbb92b67 KW |
1260 | int coroutine_fn blk_co_pwritev_part(BlockBackend *blk, int64_t offset, |
1261 | unsigned int bytes, | |
1262 | QEMUIOVector *qiov, size_t qiov_offset, | |
1263 | BdrvRequestFlags flags) | |
1264 | { | |
1265 | int ret; | |
1266 | ||
1267 | blk_inc_in_flight(blk); | |
1268 | ret = blk_do_pwritev_part(blk, offset, bytes, qiov, qiov_offset, flags); | |
1269 | blk_dec_in_flight(blk); | |
1270 | ||
1271 | return ret; | |
1272 | } | |
1273 | ||
b3016864 VSO |
1274 | int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset, |
1275 | unsigned int bytes, QEMUIOVector *qiov, | |
1276 | BdrvRequestFlags flags) | |
1277 | { | |
1278 | return blk_co_pwritev_part(blk, offset, bytes, qiov, 0, flags); | |
1279 | } | |
1280 | ||
1bf1cbc9 KW |
1281 | typedef struct BlkRwCo { |
1282 | BlockBackend *blk; | |
1283 | int64_t offset; | |
c060332c | 1284 | void *iobuf; |
1bf1cbc9 KW |
1285 | int ret; |
1286 | BdrvRequestFlags flags; | |
1287 | } BlkRwCo; | |
1288 | ||
1289 | static void blk_read_entry(void *opaque) | |
1290 | { | |
1291 | BlkRwCo *rwco = opaque; | |
c060332c | 1292 | QEMUIOVector *qiov = rwco->iobuf; |
1bf1cbc9 | 1293 | |
fbb92b67 | 1294 | rwco->ret = blk_do_preadv(rwco->blk, rwco->offset, qiov->size, |
c060332c | 1295 | qiov, rwco->flags); |
4720cbee | 1296 | aio_wait_kick(); |
1bf1cbc9 KW |
1297 | } |
1298 | ||
a8823a3b KW |
1299 | static void blk_write_entry(void *opaque) |
1300 | { | |
1301 | BlkRwCo *rwco = opaque; | |
c060332c | 1302 | QEMUIOVector *qiov = rwco->iobuf; |
a8823a3b | 1303 | |
fbb92b67 KW |
1304 | rwco->ret = blk_do_pwritev_part(rwco->blk, rwco->offset, qiov->size, |
1305 | qiov, 0, rwco->flags); | |
4720cbee | 1306 | aio_wait_kick(); |
a8823a3b KW |
1307 | } |
1308 | ||
a55d3fba KW |
1309 | static int blk_prw(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1310 | int64_t bytes, CoroutineEntry co_entry, | |
1311 | BdrvRequestFlags flags) | |
1bf1cbc9 | 1312 | { |
ae5a9592 VSO |
1313 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); |
1314 | BlkRwCo rwco = { | |
1bf1cbc9 | 1315 | .blk = blk, |
a55d3fba | 1316 | .offset = offset, |
c060332c | 1317 | .iobuf = &qiov, |
fc1453cd | 1318 | .flags = flags, |
1bf1cbc9 KW |
1319 | .ret = NOT_DONE, |
1320 | }; | |
1321 | ||
fbb92b67 | 1322 | blk_inc_in_flight(blk); |
35f106e6 PB |
1323 | if (qemu_in_coroutine()) { |
1324 | /* Fast-path if already in coroutine context */ | |
1325 | co_entry(&rwco); | |
1326 | } else { | |
1327 | Coroutine *co = qemu_coroutine_create(co_entry, &rwco); | |
e92f0e19 | 1328 | bdrv_coroutine_enter(blk_bs(blk), co); |
35f106e6 PB |
1329 | BDRV_POLL_WHILE(blk_bs(blk), rwco.ret == NOT_DONE); |
1330 | } | |
fbb92b67 | 1331 | blk_dec_in_flight(blk); |
1bf1cbc9 KW |
1332 | |
1333 | return rwco.ret; | |
4be74634 MA |
1334 | } |
1335 | ||
d004bd52 | 1336 | int blk_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 1337 | int bytes, BdrvRequestFlags flags) |
0df89e8e | 1338 | { |
f5a5ca79 | 1339 | return blk_prw(blk, offset, NULL, bytes, blk_write_entry, |
983a1600 | 1340 | flags | BDRV_REQ_ZERO_WRITE); |
0df89e8e KW |
1341 | } |
1342 | ||
720ff280 KW |
1343 | int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags) |
1344 | { | |
1345 | return bdrv_make_zero(blk->root, flags); | |
1346 | } | |
1347 | ||
c90e2a9c | 1348 | void blk_inc_in_flight(BlockBackend *blk) |
33f2a757 SH |
1349 | { |
1350 | atomic_inc(&blk->in_flight); | |
1351 | } | |
1352 | ||
c90e2a9c | 1353 | void blk_dec_in_flight(BlockBackend *blk) |
33f2a757 SH |
1354 | { |
1355 | atomic_dec(&blk->in_flight); | |
cfe29d82 | 1356 | aio_wait_kick(); |
33f2a757 SH |
1357 | } |
1358 | ||
e7f7d676 HR |
1359 | static void error_callback_bh(void *opaque) |
1360 | { | |
1361 | struct BlockBackendAIOCB *acb = opaque; | |
99723548 | 1362 | |
33f2a757 | 1363 | blk_dec_in_flight(acb->blk); |
e7f7d676 HR |
1364 | acb->common.cb(acb->common.opaque, acb->ret); |
1365 | qemu_aio_unref(acb); | |
1366 | } | |
1367 | ||
ca78ecfa PL |
1368 | BlockAIOCB *blk_abort_aio_request(BlockBackend *blk, |
1369 | BlockCompletionFunc *cb, | |
1370 | void *opaque, int ret) | |
e7f7d676 HR |
1371 | { |
1372 | struct BlockBackendAIOCB *acb; | |
e7f7d676 | 1373 | |
33f2a757 | 1374 | blk_inc_in_flight(blk); |
e7f7d676 | 1375 | acb = blk_aio_get(&block_backend_aiocb_info, blk, cb, opaque); |
4981bdec | 1376 | acb->blk = blk; |
e7f7d676 HR |
1377 | acb->ret = ret; |
1378 | ||
e4ec5ad4 PD |
1379 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1380 | error_callback_bh, acb); | |
e7f7d676 HR |
1381 | return &acb->common; |
1382 | } | |
1383 | ||
57d6a428 KW |
1384 | typedef struct BlkAioEmAIOCB { |
1385 | BlockAIOCB common; | |
1386 | BlkRwCo rwco; | |
7fa84cd8 | 1387 | int bytes; |
57d6a428 | 1388 | bool has_returned; |
57d6a428 KW |
1389 | } BlkAioEmAIOCB; |
1390 | ||
1391 | static const AIOCBInfo blk_aio_em_aiocb_info = { | |
1392 | .aiocb_size = sizeof(BlkAioEmAIOCB), | |
1393 | }; | |
1394 | ||
1395 | static void blk_aio_complete(BlkAioEmAIOCB *acb) | |
1396 | { | |
57d6a428 KW |
1397 | if (acb->has_returned) { |
1398 | acb->common.cb(acb->common.opaque, acb->rwco.ret); | |
46aaf2a5 | 1399 | blk_dec_in_flight(acb->rwco.blk); |
57d6a428 KW |
1400 | qemu_aio_unref(acb); |
1401 | } | |
1402 | } | |
1403 | ||
1404 | static void blk_aio_complete_bh(void *opaque) | |
1405 | { | |
fffb6e12 | 1406 | BlkAioEmAIOCB *acb = opaque; |
fffb6e12 PB |
1407 | assert(acb->has_returned); |
1408 | blk_aio_complete(acb); | |
57d6a428 KW |
1409 | } |
1410 | ||
7fa84cd8 | 1411 | static BlockAIOCB *blk_aio_prwv(BlockBackend *blk, int64_t offset, int bytes, |
c060332c | 1412 | void *iobuf, CoroutineEntry co_entry, |
57d6a428 KW |
1413 | BdrvRequestFlags flags, |
1414 | BlockCompletionFunc *cb, void *opaque) | |
1415 | { | |
1416 | BlkAioEmAIOCB *acb; | |
1417 | Coroutine *co; | |
1418 | ||
33f2a757 | 1419 | blk_inc_in_flight(blk); |
57d6a428 KW |
1420 | acb = blk_aio_get(&blk_aio_em_aiocb_info, blk, cb, opaque); |
1421 | acb->rwco = (BlkRwCo) { | |
1422 | .blk = blk, | |
1423 | .offset = offset, | |
c060332c | 1424 | .iobuf = iobuf, |
57d6a428 KW |
1425 | .flags = flags, |
1426 | .ret = NOT_DONE, | |
1427 | }; | |
7fa84cd8 | 1428 | acb->bytes = bytes; |
57d6a428 KW |
1429 | acb->has_returned = false; |
1430 | ||
0b8b8753 | 1431 | co = qemu_coroutine_create(co_entry, acb); |
e92f0e19 | 1432 | bdrv_coroutine_enter(blk_bs(blk), co); |
57d6a428 KW |
1433 | |
1434 | acb->has_returned = true; | |
1435 | if (acb->rwco.ret != NOT_DONE) { | |
e4ec5ad4 PD |
1436 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1437 | blk_aio_complete_bh, acb); | |
57d6a428 KW |
1438 | } |
1439 | ||
1440 | return &acb->common; | |
1441 | } | |
1442 | ||
1443 | static void blk_aio_read_entry(void *opaque) | |
1444 | { | |
1445 | BlkAioEmAIOCB *acb = opaque; | |
1446 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1447 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1448 | |
c060332c | 1449 | assert(qiov->size == acb->bytes); |
fbb92b67 | 1450 | rwco->ret = blk_do_preadv(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1451 | qiov, rwco->flags); |
57d6a428 KW |
1452 | blk_aio_complete(acb); |
1453 | } | |
1454 | ||
1455 | static void blk_aio_write_entry(void *opaque) | |
1456 | { | |
1457 | BlkAioEmAIOCB *acb = opaque; | |
1458 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1459 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1460 | |
c060332c | 1461 | assert(!qiov || qiov->size == acb->bytes); |
fbb92b67 KW |
1462 | rwco->ret = blk_do_pwritev_part(rwco->blk, rwco->offset, acb->bytes, |
1463 | qiov, 0, rwco->flags); | |
57d6a428 KW |
1464 | blk_aio_complete(acb); |
1465 | } | |
1466 | ||
d004bd52 EB |
1467 | BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
1468 | int count, BdrvRequestFlags flags, | |
1469 | BlockCompletionFunc *cb, void *opaque) | |
4be74634 | 1470 | { |
983a1600 EB |
1471 | return blk_aio_prwv(blk, offset, count, NULL, blk_aio_write_entry, |
1472 | flags | BDRV_REQ_ZERO_WRITE, cb, opaque); | |
4be74634 MA |
1473 | } |
1474 | ||
1475 | int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count) | |
1476 | { | |
a55d3fba | 1477 | int ret = blk_prw(blk, offset, buf, count, blk_read_entry, 0); |
e7f7d676 HR |
1478 | if (ret < 0) { |
1479 | return ret; | |
1480 | } | |
a55d3fba | 1481 | return count; |
4be74634 MA |
1482 | } |
1483 | ||
8341f00d EB |
1484 | int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count, |
1485 | BdrvRequestFlags flags) | |
4be74634 | 1486 | { |
8341f00d EB |
1487 | int ret = blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
1488 | flags); | |
e7f7d676 HR |
1489 | if (ret < 0) { |
1490 | return ret; | |
1491 | } | |
a55d3fba | 1492 | return count; |
4be74634 MA |
1493 | } |
1494 | ||
1495 | int64_t blk_getlength(BlockBackend *blk) | |
1496 | { | |
c09ba36c HR |
1497 | if (!blk_is_available(blk)) { |
1498 | return -ENOMEDIUM; | |
1499 | } | |
1500 | ||
f21d96d0 | 1501 | return bdrv_getlength(blk_bs(blk)); |
4be74634 MA |
1502 | } |
1503 | ||
1504 | void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr) | |
1505 | { | |
f21d96d0 | 1506 | if (!blk_bs(blk)) { |
a46fc9c9 HR |
1507 | *nb_sectors_ptr = 0; |
1508 | } else { | |
f21d96d0 | 1509 | bdrv_get_geometry(blk_bs(blk), nb_sectors_ptr); |
a46fc9c9 | 1510 | } |
4be74634 MA |
1511 | } |
1512 | ||
1ef01253 HR |
1513 | int64_t blk_nb_sectors(BlockBackend *blk) |
1514 | { | |
c09ba36c HR |
1515 | if (!blk_is_available(blk)) { |
1516 | return -ENOMEDIUM; | |
1517 | } | |
1518 | ||
f21d96d0 | 1519 | return bdrv_nb_sectors(blk_bs(blk)); |
1ef01253 HR |
1520 | } |
1521 | ||
60cb2fa7 EB |
1522 | BlockAIOCB *blk_aio_preadv(BlockBackend *blk, int64_t offset, |
1523 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1524 | BlockCompletionFunc *cb, void *opaque) | |
1525 | { | |
1526 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1527 | blk_aio_read_entry, flags, cb, opaque); | |
1528 | } | |
1529 | ||
60cb2fa7 EB |
1530 | BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
1531 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1532 | BlockCompletionFunc *cb, void *opaque) | |
1533 | { | |
1534 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1535 | blk_aio_write_entry, flags, cb, opaque); | |
1536 | } | |
1537 | ||
4be74634 MA |
1538 | void blk_aio_cancel(BlockAIOCB *acb) |
1539 | { | |
1540 | bdrv_aio_cancel(acb); | |
1541 | } | |
1542 | ||
1543 | void blk_aio_cancel_async(BlockAIOCB *acb) | |
1544 | { | |
1545 | bdrv_aio_cancel_async(acb); | |
1546 | } | |
1547 | ||
fbb92b67 KW |
1548 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1549 | static int coroutine_fn | |
1550 | blk_do_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
4be74634 | 1551 | { |
cf312932 KW |
1552 | blk_wait_while_drained(blk); |
1553 | ||
c09ba36c HR |
1554 | if (!blk_is_available(blk)) { |
1555 | return -ENOMEDIUM; | |
1556 | } | |
1557 | ||
48af776a KW |
1558 | return bdrv_co_ioctl(blk_bs(blk), req, buf); |
1559 | } | |
1560 | ||
1561 | static void blk_ioctl_entry(void *opaque) | |
1562 | { | |
1563 | BlkRwCo *rwco = opaque; | |
c060332c DS |
1564 | QEMUIOVector *qiov = rwco->iobuf; |
1565 | ||
fbb92b67 | 1566 | rwco->ret = blk_do_ioctl(rwco->blk, rwco->offset, qiov->iov[0].iov_base); |
4720cbee | 1567 | aio_wait_kick(); |
48af776a KW |
1568 | } |
1569 | ||
1570 | int blk_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
1571 | { | |
1572 | return blk_prw(blk, req, buf, 0, blk_ioctl_entry, 0); | |
1573 | } | |
1574 | ||
1575 | static void blk_aio_ioctl_entry(void *opaque) | |
1576 | { | |
1577 | BlkAioEmAIOCB *acb = opaque; | |
1578 | BlkRwCo *rwco = &acb->rwco; | |
1579 | ||
fbb92b67 | 1580 | rwco->ret = blk_do_ioctl(rwco->blk, rwco->offset, rwco->iobuf); |
c060332c | 1581 | |
48af776a | 1582 | blk_aio_complete(acb); |
4be74634 MA |
1583 | } |
1584 | ||
1585 | BlockAIOCB *blk_aio_ioctl(BlockBackend *blk, unsigned long int req, void *buf, | |
1586 | BlockCompletionFunc *cb, void *opaque) | |
1587 | { | |
c060332c | 1588 | return blk_aio_prwv(blk, req, 0, buf, blk_aio_ioctl_entry, 0, cb, opaque); |
4be74634 MA |
1589 | } |
1590 | ||
fbb92b67 KW |
1591 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1592 | static int coroutine_fn | |
1593 | blk_do_pdiscard(BlockBackend *blk, int64_t offset, int bytes) | |
2bb0dce7 | 1594 | { |
cf312932 KW |
1595 | int ret; |
1596 | ||
1597 | blk_wait_while_drained(blk); | |
1598 | ||
1599 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1600 | if (ret < 0) { |
1601 | return ret; | |
1602 | } | |
1603 | ||
0b9fd3f4 | 1604 | return bdrv_co_pdiscard(blk->root, offset, bytes); |
2bb0dce7 HR |
1605 | } |
1606 | ||
564806c5 KW |
1607 | static void blk_aio_pdiscard_entry(void *opaque) |
1608 | { | |
1609 | BlkAioEmAIOCB *acb = opaque; | |
1610 | BlkRwCo *rwco = &acb->rwco; | |
1611 | ||
fbb92b67 | 1612 | rwco->ret = blk_do_pdiscard(rwco->blk, rwco->offset, acb->bytes); |
564806c5 KW |
1613 | blk_aio_complete(acb); |
1614 | } | |
1615 | ||
1616 | BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, | |
1617 | int64_t offset, int bytes, | |
1618 | BlockCompletionFunc *cb, void *opaque) | |
1619 | { | |
1620 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_pdiscard_entry, 0, | |
1621 | cb, opaque); | |
1622 | } | |
1623 | ||
fbb92b67 KW |
1624 | int coroutine_fn blk_co_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
1625 | { | |
1626 | int ret; | |
1627 | ||
1628 | blk_inc_in_flight(blk); | |
1629 | ret = blk_do_pdiscard(blk, offset, bytes); | |
1630 | blk_dec_in_flight(blk); | |
1631 | ||
1632 | return ret; | |
1633 | } | |
1634 | ||
564806c5 KW |
1635 | static void blk_pdiscard_entry(void *opaque) |
1636 | { | |
1637 | BlkRwCo *rwco = opaque; | |
1638 | QEMUIOVector *qiov = rwco->iobuf; | |
1639 | ||
fbb92b67 | 1640 | rwco->ret = blk_do_pdiscard(rwco->blk, rwco->offset, qiov->size); |
564806c5 KW |
1641 | aio_wait_kick(); |
1642 | } | |
1643 | ||
1644 | int blk_pdiscard(BlockBackend *blk, int64_t offset, int bytes) | |
1645 | { | |
1646 | return blk_prw(blk, offset, NULL, bytes, blk_pdiscard_entry, 0); | |
1647 | } | |
1648 | ||
fbb92b67 KW |
1649 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1650 | static int coroutine_fn blk_do_flush(BlockBackend *blk) | |
2bb0dce7 | 1651 | { |
cf312932 KW |
1652 | blk_wait_while_drained(blk); |
1653 | ||
c09ba36c HR |
1654 | if (!blk_is_available(blk)) { |
1655 | return -ENOMEDIUM; | |
1656 | } | |
1657 | ||
f21d96d0 | 1658 | return bdrv_co_flush(blk_bs(blk)); |
2bb0dce7 HR |
1659 | } |
1660 | ||
564806c5 KW |
1661 | static void blk_aio_flush_entry(void *opaque) |
1662 | { | |
1663 | BlkAioEmAIOCB *acb = opaque; | |
1664 | BlkRwCo *rwco = &acb->rwco; | |
1665 | ||
fbb92b67 | 1666 | rwco->ret = blk_do_flush(rwco->blk); |
564806c5 KW |
1667 | blk_aio_complete(acb); |
1668 | } | |
1669 | ||
1670 | BlockAIOCB *blk_aio_flush(BlockBackend *blk, | |
1671 | BlockCompletionFunc *cb, void *opaque) | |
1672 | { | |
1673 | return blk_aio_prwv(blk, 0, 0, NULL, blk_aio_flush_entry, 0, cb, opaque); | |
1674 | } | |
1675 | ||
fbb92b67 KW |
1676 | int coroutine_fn blk_co_flush(BlockBackend *blk) |
1677 | { | |
1678 | int ret; | |
1679 | ||
1680 | blk_inc_in_flight(blk); | |
1681 | ret = blk_do_flush(blk); | |
1682 | blk_dec_in_flight(blk); | |
1683 | ||
1684 | return ret; | |
1685 | } | |
1686 | ||
be07a889 | 1687 | static void blk_flush_entry(void *opaque) |
4be74634 | 1688 | { |
be07a889 | 1689 | BlkRwCo *rwco = opaque; |
fbb92b67 | 1690 | rwco->ret = blk_do_flush(rwco->blk); |
4720cbee | 1691 | aio_wait_kick(); |
be07a889 | 1692 | } |
c09ba36c | 1693 | |
be07a889 KW |
1694 | int blk_flush(BlockBackend *blk) |
1695 | { | |
1696 | return blk_prw(blk, 0, NULL, 0, blk_flush_entry, 0); | |
4be74634 MA |
1697 | } |
1698 | ||
97b0385a AY |
1699 | void blk_drain(BlockBackend *blk) |
1700 | { | |
33f2a757 SH |
1701 | BlockDriverState *bs = blk_bs(blk); |
1702 | ||
1703 | if (bs) { | |
1704 | bdrv_drained_begin(bs); | |
1705 | } | |
1706 | ||
1707 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 KW |
1708 | AIO_WAIT_WHILE(blk_get_aio_context(blk), |
1709 | atomic_mb_read(&blk->in_flight) > 0); | |
33f2a757 SH |
1710 | |
1711 | if (bs) { | |
1712 | bdrv_drained_end(bs); | |
a46fc9c9 | 1713 | } |
97b0385a AY |
1714 | } |
1715 | ||
4be74634 MA |
1716 | void blk_drain_all(void) |
1717 | { | |
33f2a757 SH |
1718 | BlockBackend *blk = NULL; |
1719 | ||
1720 | bdrv_drain_all_begin(); | |
1721 | ||
1722 | while ((blk = blk_all_next(blk)) != NULL) { | |
1723 | AioContext *ctx = blk_get_aio_context(blk); | |
1724 | ||
1725 | aio_context_acquire(ctx); | |
1726 | ||
1727 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 | 1728 | AIO_WAIT_WHILE(ctx, atomic_mb_read(&blk->in_flight) > 0); |
33f2a757 SH |
1729 | |
1730 | aio_context_release(ctx); | |
1731 | } | |
1732 | ||
1733 | bdrv_drain_all_end(); | |
4be74634 MA |
1734 | } |
1735 | ||
373340b2 HR |
1736 | void blk_set_on_error(BlockBackend *blk, BlockdevOnError on_read_error, |
1737 | BlockdevOnError on_write_error) | |
1738 | { | |
1739 | blk->on_read_error = on_read_error; | |
1740 | blk->on_write_error = on_write_error; | |
1741 | } | |
1742 | ||
4be74634 MA |
1743 | BlockdevOnError blk_get_on_error(BlockBackend *blk, bool is_read) |
1744 | { | |
373340b2 | 1745 | return is_read ? blk->on_read_error : blk->on_write_error; |
4be74634 MA |
1746 | } |
1747 | ||
1748 | BlockErrorAction blk_get_error_action(BlockBackend *blk, bool is_read, | |
1749 | int error) | |
1750 | { | |
373340b2 HR |
1751 | BlockdevOnError on_err = blk_get_on_error(blk, is_read); |
1752 | ||
1753 | switch (on_err) { | |
1754 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
1755 | return (error == ENOSPC) ? | |
1756 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
1757 | case BLOCKDEV_ON_ERROR_STOP: | |
1758 | return BLOCK_ERROR_ACTION_STOP; | |
1759 | case BLOCKDEV_ON_ERROR_REPORT: | |
1760 | return BLOCK_ERROR_ACTION_REPORT; | |
1761 | case BLOCKDEV_ON_ERROR_IGNORE: | |
1762 | return BLOCK_ERROR_ACTION_IGNORE; | |
8c398252 | 1763 | case BLOCKDEV_ON_ERROR_AUTO: |
373340b2 HR |
1764 | default: |
1765 | abort(); | |
1766 | } | |
4be74634 MA |
1767 | } |
1768 | ||
373340b2 HR |
1769 | static void send_qmp_error_event(BlockBackend *blk, |
1770 | BlockErrorAction action, | |
1771 | bool is_read, int error) | |
1772 | { | |
1773 | IoOperationType optype; | |
bfe1a14c | 1774 | BlockDriverState *bs = blk_bs(blk); |
373340b2 HR |
1775 | |
1776 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; | |
bfe1a14c KW |
1777 | qapi_event_send_block_io_error(blk_name(blk), !!bs, |
1778 | bs ? bdrv_get_node_name(bs) : NULL, optype, | |
2bf7e10f | 1779 | action, blk_iostatus_is_enabled(blk), |
3ab72385 | 1780 | error == ENOSPC, strerror(error)); |
373340b2 HR |
1781 | } |
1782 | ||
1783 | /* This is done by device models because, while the block layer knows | |
1784 | * about the error, it does not know whether an operation comes from | |
1785 | * the device or the block layer (from a job, for example). | |
1786 | */ | |
4be74634 MA |
1787 | void blk_error_action(BlockBackend *blk, BlockErrorAction action, |
1788 | bool is_read, int error) | |
1789 | { | |
373340b2 HR |
1790 | assert(error >= 0); |
1791 | ||
1792 | if (action == BLOCK_ERROR_ACTION_STOP) { | |
1793 | /* First set the iostatus, so that "info block" returns an iostatus | |
1794 | * that matches the events raised so far (an additional error iostatus | |
1795 | * is fine, but not a lost one). | |
1796 | */ | |
1797 | blk_iostatus_set_err(blk, error); | |
1798 | ||
1799 | /* Then raise the request to stop the VM and the event. | |
1800 | * qemu_system_vmstop_request_prepare has two effects. First, | |
1801 | * it ensures that the STOP event always comes after the | |
1802 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
1803 | * can observe the STOP event and do a "cont" before the STOP | |
1804 | * event is issued, the VM will not stop. In this case, vm_start() | |
1805 | * also ensures that the STOP/RESUME pair of events is emitted. | |
1806 | */ | |
1807 | qemu_system_vmstop_request_prepare(); | |
1808 | send_qmp_error_event(blk, action, is_read, error); | |
1809 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); | |
1810 | } else { | |
1811 | send_qmp_error_event(blk, action, is_read, error); | |
1812 | } | |
4be74634 MA |
1813 | } |
1814 | ||
96710565 | 1815 | bool blk_is_read_only(BlockBackend *blk) |
4be74634 | 1816 | { |
f21d96d0 KW |
1817 | BlockDriverState *bs = blk_bs(blk); |
1818 | ||
1819 | if (bs) { | |
1820 | return bdrv_is_read_only(bs); | |
061959e8 HR |
1821 | } else { |
1822 | return blk->root_state.read_only; | |
1823 | } | |
4be74634 MA |
1824 | } |
1825 | ||
96710565 | 1826 | bool blk_is_sg(BlockBackend *blk) |
4be74634 | 1827 | { |
f21d96d0 KW |
1828 | BlockDriverState *bs = blk_bs(blk); |
1829 | ||
1830 | if (!bs) { | |
96710565 | 1831 | return false; |
a46fc9c9 HR |
1832 | } |
1833 | ||
f21d96d0 | 1834 | return bdrv_is_sg(bs); |
4be74634 MA |
1835 | } |
1836 | ||
96710565 | 1837 | bool blk_enable_write_cache(BlockBackend *blk) |
4be74634 | 1838 | { |
bfd18d1e | 1839 | return blk->enable_write_cache; |
4be74634 MA |
1840 | } |
1841 | ||
1842 | void blk_set_enable_write_cache(BlockBackend *blk, bool wce) | |
1843 | { | |
bfd18d1e | 1844 | blk->enable_write_cache = wce; |
4be74634 MA |
1845 | } |
1846 | ||
2bb0dce7 HR |
1847 | void blk_invalidate_cache(BlockBackend *blk, Error **errp) |
1848 | { | |
f21d96d0 KW |
1849 | BlockDriverState *bs = blk_bs(blk); |
1850 | ||
1851 | if (!bs) { | |
c09ba36c HR |
1852 | error_setg(errp, "Device '%s' has no medium", blk->name); |
1853 | return; | |
1854 | } | |
1855 | ||
f21d96d0 | 1856 | bdrv_invalidate_cache(bs, errp); |
2bb0dce7 HR |
1857 | } |
1858 | ||
e031f750 | 1859 | bool blk_is_inserted(BlockBackend *blk) |
4be74634 | 1860 | { |
f21d96d0 KW |
1861 | BlockDriverState *bs = blk_bs(blk); |
1862 | ||
1863 | return bs && bdrv_is_inserted(bs); | |
db0284f8 HR |
1864 | } |
1865 | ||
1866 | bool blk_is_available(BlockBackend *blk) | |
1867 | { | |
1868 | return blk_is_inserted(blk) && !blk_dev_is_tray_open(blk); | |
4be74634 MA |
1869 | } |
1870 | ||
1871 | void blk_lock_medium(BlockBackend *blk, bool locked) | |
1872 | { | |
f21d96d0 KW |
1873 | BlockDriverState *bs = blk_bs(blk); |
1874 | ||
1875 | if (bs) { | |
1876 | bdrv_lock_medium(bs, locked); | |
a46fc9c9 | 1877 | } |
4be74634 MA |
1878 | } |
1879 | ||
1880 | void blk_eject(BlockBackend *blk, bool eject_flag) | |
1881 | { | |
f21d96d0 | 1882 | BlockDriverState *bs = blk_bs(blk); |
2d76e724 KW |
1883 | char *id; |
1884 | ||
f21d96d0 KW |
1885 | if (bs) { |
1886 | bdrv_eject(bs, eject_flag); | |
a46fc9c9 | 1887 | } |
c47ee043 JS |
1888 | |
1889 | /* Whether or not we ejected on the backend, | |
1890 | * the frontend experienced a tray event. */ | |
1891 | id = blk_get_attached_dev_id(blk); | |
1892 | qapi_event_send_device_tray_moved(blk_name(blk), id, | |
3ab72385 | 1893 | eject_flag); |
c47ee043 | 1894 | g_free(id); |
4be74634 MA |
1895 | } |
1896 | ||
1897 | int blk_get_flags(BlockBackend *blk) | |
1898 | { | |
f21d96d0 KW |
1899 | BlockDriverState *bs = blk_bs(blk); |
1900 | ||
1901 | if (bs) { | |
1902 | return bdrv_get_flags(bs); | |
061959e8 HR |
1903 | } else { |
1904 | return blk->root_state.open_flags; | |
1905 | } | |
4be74634 MA |
1906 | } |
1907 | ||
4841211e EB |
1908 | /* Returns the minimum request alignment, in bytes; guaranteed nonzero */ |
1909 | uint32_t blk_get_request_alignment(BlockBackend *blk) | |
1910 | { | |
1911 | BlockDriverState *bs = blk_bs(blk); | |
1912 | return bs ? bs->bl.request_alignment : BDRV_SECTOR_SIZE; | |
1913 | } | |
1914 | ||
5def6b80 EB |
1915 | /* Returns the maximum transfer length, in bytes; guaranteed nonzero */ |
1916 | uint32_t blk_get_max_transfer(BlockBackend *blk) | |
454057b7 | 1917 | { |
f21d96d0 | 1918 | BlockDriverState *bs = blk_bs(blk); |
5def6b80 | 1919 | uint32_t max = 0; |
f21d96d0 KW |
1920 | |
1921 | if (bs) { | |
5def6b80 | 1922 | max = bs->bl.max_transfer; |
a46fc9c9 | 1923 | } |
5def6b80 | 1924 | return MIN_NON_ZERO(max, INT_MAX); |
454057b7 PL |
1925 | } |
1926 | ||
648296e0 SH |
1927 | int blk_get_max_iov(BlockBackend *blk) |
1928 | { | |
f21d96d0 | 1929 | return blk->root->bs->bl.max_iov; |
648296e0 SH |
1930 | } |
1931 | ||
4be74634 MA |
1932 | void blk_set_guest_block_size(BlockBackend *blk, int align) |
1933 | { | |
68e9ec01 | 1934 | blk->guest_block_size = align; |
4be74634 MA |
1935 | } |
1936 | ||
f1c17521 PB |
1937 | void *blk_try_blockalign(BlockBackend *blk, size_t size) |
1938 | { | |
f21d96d0 | 1939 | return qemu_try_blockalign(blk ? blk_bs(blk) : NULL, size); |
f1c17521 PB |
1940 | } |
1941 | ||
4be74634 MA |
1942 | void *blk_blockalign(BlockBackend *blk, size_t size) |
1943 | { | |
f21d96d0 | 1944 | return qemu_blockalign(blk ? blk_bs(blk) : NULL, size); |
4be74634 MA |
1945 | } |
1946 | ||
1947 | bool blk_op_is_blocked(BlockBackend *blk, BlockOpType op, Error **errp) | |
1948 | { | |
f21d96d0 KW |
1949 | BlockDriverState *bs = blk_bs(blk); |
1950 | ||
1951 | if (!bs) { | |
a46fc9c9 HR |
1952 | return false; |
1953 | } | |
1954 | ||
f21d96d0 | 1955 | return bdrv_op_is_blocked(bs, op, errp); |
4be74634 MA |
1956 | } |
1957 | ||
1958 | void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason) | |
1959 | { | |
f21d96d0 KW |
1960 | BlockDriverState *bs = blk_bs(blk); |
1961 | ||
1962 | if (bs) { | |
1963 | bdrv_op_unblock(bs, op, reason); | |
a46fc9c9 | 1964 | } |
4be74634 MA |
1965 | } |
1966 | ||
1967 | void blk_op_block_all(BlockBackend *blk, Error *reason) | |
1968 | { | |
f21d96d0 KW |
1969 | BlockDriverState *bs = blk_bs(blk); |
1970 | ||
1971 | if (bs) { | |
1972 | bdrv_op_block_all(bs, reason); | |
a46fc9c9 | 1973 | } |
4be74634 MA |
1974 | } |
1975 | ||
1976 | void blk_op_unblock_all(BlockBackend *blk, Error *reason) | |
1977 | { | |
f21d96d0 KW |
1978 | BlockDriverState *bs = blk_bs(blk); |
1979 | ||
1980 | if (bs) { | |
1981 | bdrv_op_unblock_all(bs, reason); | |
a46fc9c9 | 1982 | } |
4be74634 MA |
1983 | } |
1984 | ||
1985 | AioContext *blk_get_aio_context(BlockBackend *blk) | |
1986 | { | |
d861ab3a KW |
1987 | BlockDriverState *bs = blk_bs(blk); |
1988 | ||
d861ab3a | 1989 | if (bs) { |
132ada80 KW |
1990 | AioContext *ctx = bdrv_get_aio_context(blk_bs(blk)); |
1991 | assert(ctx == blk->ctx); | |
d861ab3a KW |
1992 | } |
1993 | ||
1994 | return blk->ctx; | |
4981bdec HR |
1995 | } |
1996 | ||
1997 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) | |
1998 | { | |
1999 | BlockBackendAIOCB *blk_acb = DO_UPCAST(BlockBackendAIOCB, common, acb); | |
2000 | return blk_get_aio_context(blk_acb->blk); | |
4be74634 MA |
2001 | } |
2002 | ||
97896a48 KW |
2003 | static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2004 | bool update_root_node, Error **errp) | |
4be74634 | 2005 | { |
f21d96d0 | 2006 | BlockDriverState *bs = blk_bs(blk); |
c61791fc | 2007 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
97896a48 | 2008 | int ret; |
f21d96d0 KW |
2009 | |
2010 | if (bs) { | |
97896a48 KW |
2011 | if (update_root_node) { |
2012 | ret = bdrv_child_try_set_aio_context(bs, new_context, blk->root, | |
2013 | errp); | |
2014 | if (ret < 0) { | |
2015 | return ret; | |
2016 | } | |
2017 | } | |
c61791fc | 2018 | if (tgm->throttle_state) { |
dc868fb0 | 2019 | bdrv_drained_begin(bs); |
c61791fc MP |
2020 | throttle_group_detach_aio_context(tgm); |
2021 | throttle_group_attach_aio_context(tgm, new_context); | |
dc868fb0 | 2022 | bdrv_drained_end(bs); |
7ca7f0f6 | 2023 | } |
a46fc9c9 | 2024 | } |
97896a48 | 2025 | |
d861ab3a | 2026 | blk->ctx = new_context; |
97896a48 | 2027 | return 0; |
4be74634 MA |
2028 | } |
2029 | ||
97896a48 KW |
2030 | int blk_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2031 | Error **errp) | |
38475269 | 2032 | { |
97896a48 | 2033 | return blk_do_set_aio_context(blk, new_context, true, errp); |
38475269 KW |
2034 | } |
2035 | ||
2036 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2037 | GSList **ignore, Error **errp) | |
2038 | { | |
2039 | BlockBackend *blk = child->opaque; | |
2040 | ||
980b0f94 KW |
2041 | if (blk->allow_aio_context_change) { |
2042 | return true; | |
2043 | } | |
2044 | ||
38475269 KW |
2045 | /* Only manually created BlockBackends that are not attached to anything |
2046 | * can change their AioContext without updating their user. */ | |
2047 | if (!blk->name || blk->dev) { | |
2048 | /* TODO Add BB name/QOM path */ | |
2049 | error_setg(errp, "Cannot change iothread of active block backend"); | |
2050 | return false; | |
2051 | } | |
2052 | ||
2053 | return true; | |
2054 | } | |
2055 | ||
2056 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2057 | GSList **ignore) | |
2058 | { | |
2059 | BlockBackend *blk = child->opaque; | |
97896a48 | 2060 | blk_do_set_aio_context(blk, ctx, false, &error_abort); |
38475269 KW |
2061 | } |
2062 | ||
2019ba0a HR |
2063 | void blk_add_aio_context_notifier(BlockBackend *blk, |
2064 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
2065 | void (*detach_aio_context)(void *opaque), void *opaque) | |
2066 | { | |
d03654ea | 2067 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2068 | BlockDriverState *bs = blk_bs(blk); |
2069 | ||
d03654ea SH |
2070 | notifier = g_new(BlockBackendAioNotifier, 1); |
2071 | notifier->attached_aio_context = attached_aio_context; | |
2072 | notifier->detach_aio_context = detach_aio_context; | |
2073 | notifier->opaque = opaque; | |
2074 | QLIST_INSERT_HEAD(&blk->aio_notifiers, notifier, list); | |
2075 | ||
f21d96d0 KW |
2076 | if (bs) { |
2077 | bdrv_add_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2078 | detach_aio_context, opaque); |
2079 | } | |
2019ba0a HR |
2080 | } |
2081 | ||
2082 | void blk_remove_aio_context_notifier(BlockBackend *blk, | |
2083 | void (*attached_aio_context)(AioContext *, | |
2084 | void *), | |
2085 | void (*detach_aio_context)(void *), | |
2086 | void *opaque) | |
2087 | { | |
d03654ea | 2088 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2089 | BlockDriverState *bs = blk_bs(blk); |
2090 | ||
2091 | if (bs) { | |
2092 | bdrv_remove_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2093 | detach_aio_context, opaque); |
2094 | } | |
d03654ea SH |
2095 | |
2096 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
2097 | if (notifier->attached_aio_context == attached_aio_context && | |
2098 | notifier->detach_aio_context == detach_aio_context && | |
2099 | notifier->opaque == opaque) { | |
2100 | QLIST_REMOVE(notifier, list); | |
2101 | g_free(notifier); | |
2102 | return; | |
2103 | } | |
2104 | } | |
2105 | ||
2106 | abort(); | |
2019ba0a HR |
2107 | } |
2108 | ||
3301f6c6 HR |
2109 | void blk_add_remove_bs_notifier(BlockBackend *blk, Notifier *notify) |
2110 | { | |
2111 | notifier_list_add(&blk->remove_bs_notifiers, notify); | |
2112 | } | |
2113 | ||
2114 | void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify) | |
2115 | { | |
2116 | notifier_list_add(&blk->insert_bs_notifiers, notify); | |
2117 | } | |
2118 | ||
4be74634 MA |
2119 | void blk_io_plug(BlockBackend *blk) |
2120 | { | |
f21d96d0 KW |
2121 | BlockDriverState *bs = blk_bs(blk); |
2122 | ||
2123 | if (bs) { | |
2124 | bdrv_io_plug(bs); | |
a46fc9c9 | 2125 | } |
4be74634 MA |
2126 | } |
2127 | ||
2128 | void blk_io_unplug(BlockBackend *blk) | |
2129 | { | |
f21d96d0 KW |
2130 | BlockDriverState *bs = blk_bs(blk); |
2131 | ||
2132 | if (bs) { | |
2133 | bdrv_io_unplug(bs); | |
a46fc9c9 | 2134 | } |
4be74634 MA |
2135 | } |
2136 | ||
2137 | BlockAcctStats *blk_get_stats(BlockBackend *blk) | |
2138 | { | |
7f0e9da6 | 2139 | return &blk->stats; |
4be74634 MA |
2140 | } |
2141 | ||
2142 | void *blk_aio_get(const AIOCBInfo *aiocb_info, BlockBackend *blk, | |
2143 | BlockCompletionFunc *cb, void *opaque) | |
2144 | { | |
2145 | return qemu_aio_get(aiocb_info, blk_bs(blk), cb, opaque); | |
2146 | } | |
1ef01253 | 2147 | |
d004bd52 | 2148 | int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 2149 | int bytes, BdrvRequestFlags flags) |
1ef01253 | 2150 | { |
f5a5ca79 | 2151 | return blk_co_pwritev(blk, offset, bytes, NULL, |
16aaf975 | 2152 | flags | BDRV_REQ_ZERO_WRITE); |
1ef01253 HR |
2153 | } |
2154 | ||
fe5c1355 PB |
2155 | int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, |
2156 | int count) | |
1ef01253 | 2157 | { |
35fadca8 PB |
2158 | return blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
2159 | BDRV_REQ_WRITE_COMPRESSED); | |
1ef01253 HR |
2160 | } |
2161 | ||
c80d8b06 | 2162 | int blk_truncate(BlockBackend *blk, int64_t offset, bool exact, |
8c6242b6 | 2163 | PreallocMode prealloc, BdrvRequestFlags flags, Error **errp) |
1ef01253 | 2164 | { |
c09ba36c | 2165 | if (!blk_is_available(blk)) { |
ed3d2ec9 | 2166 | error_setg(errp, "No medium inserted"); |
c09ba36c HR |
2167 | return -ENOMEDIUM; |
2168 | } | |
2169 | ||
8c6242b6 | 2170 | return bdrv_truncate(blk->root, offset, exact, prealloc, flags, errp); |
1ef01253 HR |
2171 | } |
2172 | ||
1ef01253 HR |
2173 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, |
2174 | int64_t pos, int size) | |
2175 | { | |
bfd18d1e KW |
2176 | int ret; |
2177 | ||
c09ba36c HR |
2178 | if (!blk_is_available(blk)) { |
2179 | return -ENOMEDIUM; | |
2180 | } | |
2181 | ||
bfd18d1e KW |
2182 | ret = bdrv_save_vmstate(blk_bs(blk), buf, pos, size); |
2183 | if (ret < 0) { | |
2184 | return ret; | |
2185 | } | |
2186 | ||
2187 | if (ret == size && !blk->enable_write_cache) { | |
2188 | ret = bdrv_flush(blk_bs(blk)); | |
2189 | } | |
2190 | ||
2191 | return ret < 0 ? ret : size; | |
1ef01253 HR |
2192 | } |
2193 | ||
2194 | int blk_load_vmstate(BlockBackend *blk, uint8_t *buf, int64_t pos, int size) | |
2195 | { | |
c09ba36c HR |
2196 | if (!blk_is_available(blk)) { |
2197 | return -ENOMEDIUM; | |
2198 | } | |
2199 | ||
f21d96d0 | 2200 | return bdrv_load_vmstate(blk_bs(blk), buf, pos, size); |
1ef01253 | 2201 | } |
f0272c4d ET |
2202 | |
2203 | int blk_probe_blocksizes(BlockBackend *blk, BlockSizes *bsz) | |
2204 | { | |
c09ba36c HR |
2205 | if (!blk_is_available(blk)) { |
2206 | return -ENOMEDIUM; | |
2207 | } | |
2208 | ||
f21d96d0 | 2209 | return bdrv_probe_blocksizes(blk_bs(blk), bsz); |
f0272c4d ET |
2210 | } |
2211 | ||
2212 | int blk_probe_geometry(BlockBackend *blk, HDGeometry *geo) | |
2213 | { | |
c09ba36c HR |
2214 | if (!blk_is_available(blk)) { |
2215 | return -ENOMEDIUM; | |
2216 | } | |
2217 | ||
f21d96d0 | 2218 | return bdrv_probe_geometry(blk_bs(blk), geo); |
f0272c4d | 2219 | } |
281d22d8 HR |
2220 | |
2221 | /* | |
2222 | * Updates the BlockBackendRootState object with data from the currently | |
2223 | * attached BlockDriverState. | |
2224 | */ | |
2225 | void blk_update_root_state(BlockBackend *blk) | |
2226 | { | |
f21d96d0 | 2227 | assert(blk->root); |
281d22d8 | 2228 | |
f21d96d0 KW |
2229 | blk->root_state.open_flags = blk->root->bs->open_flags; |
2230 | blk->root_state.read_only = blk->root->bs->read_only; | |
2231 | blk->root_state.detect_zeroes = blk->root->bs->detect_zeroes; | |
281d22d8 HR |
2232 | } |
2233 | ||
38cb18f5 | 2234 | /* |
b85114f8 KW |
2235 | * Returns the detect-zeroes setting to be used for bdrv_open() of a |
2236 | * BlockDriverState which is supposed to inherit the root state. | |
38cb18f5 | 2237 | */ |
b85114f8 | 2238 | bool blk_get_detect_zeroes_from_root_state(BlockBackend *blk) |
38cb18f5 | 2239 | { |
b85114f8 | 2240 | return blk->root_state.detect_zeroes; |
38cb18f5 HR |
2241 | } |
2242 | ||
2243 | /* | |
2244 | * Returns the flags to be used for bdrv_open() of a BlockDriverState which is | |
2245 | * supposed to inherit the root state. | |
2246 | */ | |
2247 | int blk_get_open_flags_from_root_state(BlockBackend *blk) | |
2248 | { | |
2249 | int bs_flags; | |
2250 | ||
2251 | bs_flags = blk->root_state.read_only ? 0 : BDRV_O_RDWR; | |
2252 | bs_flags |= blk->root_state.open_flags & ~BDRV_O_RDWR; | |
2253 | ||
2254 | return bs_flags; | |
2255 | } | |
2256 | ||
281d22d8 HR |
2257 | BlockBackendRootState *blk_get_root_state(BlockBackend *blk) |
2258 | { | |
2259 | return &blk->root_state; | |
2260 | } | |
1393f212 HR |
2261 | |
2262 | int blk_commit_all(void) | |
2263 | { | |
fe1a9cbc HR |
2264 | BlockBackend *blk = NULL; |
2265 | ||
2266 | while ((blk = blk_all_next(blk)) != NULL) { | |
2267 | AioContext *aio_context = blk_get_aio_context(blk); | |
2268 | ||
2269 | aio_context_acquire(aio_context); | |
f21d96d0 KW |
2270 | if (blk_is_inserted(blk) && blk->root->bs->backing) { |
2271 | int ret = bdrv_commit(blk->root->bs); | |
fe1a9cbc HR |
2272 | if (ret < 0) { |
2273 | aio_context_release(aio_context); | |
2274 | return ret; | |
2275 | } | |
2276 | } | |
2277 | aio_context_release(aio_context); | |
2278 | } | |
2279 | return 0; | |
2280 | } | |
2281 | ||
97148076 KW |
2282 | |
2283 | /* throttling disk I/O limits */ | |
2284 | void blk_set_io_limits(BlockBackend *blk, ThrottleConfig *cfg) | |
2285 | { | |
022cdc9f | 2286 | throttle_group_config(&blk->public.throttle_group_member, cfg); |
97148076 KW |
2287 | } |
2288 | ||
2289 | void blk_io_limits_disable(BlockBackend *blk) | |
2290 | { | |
48bf7ea8 AG |
2291 | BlockDriverState *bs = blk_bs(blk); |
2292 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; | |
2293 | assert(tgm->throttle_state); | |
2294 | if (bs) { | |
2295 | bdrv_drained_begin(bs); | |
2296 | } | |
2297 | throttle_group_unregister_tgm(tgm); | |
2298 | if (bs) { | |
2299 | bdrv_drained_end(bs); | |
2300 | } | |
97148076 KW |
2301 | } |
2302 | ||
2303 | /* should be called before blk_set_io_limits if a limit is set */ | |
2304 | void blk_io_limits_enable(BlockBackend *blk, const char *group) | |
2305 | { | |
022cdc9f | 2306 | assert(!blk->public.throttle_group_member.throttle_state); |
c61791fc MP |
2307 | throttle_group_register_tgm(&blk->public.throttle_group_member, |
2308 | group, blk_get_aio_context(blk)); | |
97148076 KW |
2309 | } |
2310 | ||
2311 | void blk_io_limits_update_group(BlockBackend *blk, const char *group) | |
2312 | { | |
2313 | /* this BB is not part of any group */ | |
022cdc9f | 2314 | if (!blk->public.throttle_group_member.throttle_state) { |
97148076 KW |
2315 | return; |
2316 | } | |
2317 | ||
2318 | /* this BB is a part of the same group than the one we want */ | |
022cdc9f MP |
2319 | if (!g_strcmp0(throttle_group_get_name(&blk->public.throttle_group_member), |
2320 | group)) { | |
97148076 KW |
2321 | return; |
2322 | } | |
2323 | ||
2324 | /* need to change the group this bs belong to */ | |
2325 | blk_io_limits_disable(blk); | |
2326 | blk_io_limits_enable(blk, group); | |
2327 | } | |
c2066af0 KW |
2328 | |
2329 | static void blk_root_drained_begin(BdrvChild *child) | |
2330 | { | |
2331 | BlockBackend *blk = child->opaque; | |
2332 | ||
f4d9cc88 JS |
2333 | if (++blk->quiesce_counter == 1) { |
2334 | if (blk->dev_ops && blk->dev_ops->drained_begin) { | |
2335 | blk->dev_ops->drained_begin(blk->dev_opaque); | |
2336 | } | |
2337 | } | |
2338 | ||
36fe1331 KW |
2339 | /* Note that blk->root may not be accessible here yet if we are just |
2340 | * attaching to a BlockDriverState that is drained. Use child instead. */ | |
2341 | ||
022cdc9f MP |
2342 | if (atomic_fetch_inc(&blk->public.throttle_group_member.io_limits_disabled) == 0) { |
2343 | throttle_group_restart_tgm(&blk->public.throttle_group_member); | |
c2066af0 KW |
2344 | } |
2345 | } | |
2346 | ||
fe5258a5 KW |
2347 | static bool blk_root_drained_poll(BdrvChild *child) |
2348 | { | |
2349 | BlockBackend *blk = child->opaque; | |
2350 | assert(blk->quiesce_counter); | |
2351 | return !!blk->in_flight; | |
2352 | } | |
2353 | ||
e037c09c | 2354 | static void blk_root_drained_end(BdrvChild *child, int *drained_end_counter) |
c2066af0 KW |
2355 | { |
2356 | BlockBackend *blk = child->opaque; | |
f4d9cc88 | 2357 | assert(blk->quiesce_counter); |
c2066af0 | 2358 | |
022cdc9f MP |
2359 | assert(blk->public.throttle_group_member.io_limits_disabled); |
2360 | atomic_dec(&blk->public.throttle_group_member.io_limits_disabled); | |
f4d9cc88 JS |
2361 | |
2362 | if (--blk->quiesce_counter == 0) { | |
2363 | if (blk->dev_ops && blk->dev_ops->drained_end) { | |
2364 | blk->dev_ops->drained_end(blk->dev_opaque); | |
2365 | } | |
cf312932 KW |
2366 | while (qemu_co_enter_next(&blk->queued_requests, NULL)) { |
2367 | /* Resume all queued requests */ | |
2368 | } | |
f4d9cc88 | 2369 | } |
c2066af0 | 2370 | } |
23d0ba93 FZ |
2371 | |
2372 | void blk_register_buf(BlockBackend *blk, void *host, size_t size) | |
2373 | { | |
2374 | bdrv_register_buf(blk_bs(blk), host, size); | |
2375 | } | |
2376 | ||
2377 | void blk_unregister_buf(BlockBackend *blk, void *host) | |
2378 | { | |
2379 | bdrv_unregister_buf(blk_bs(blk), host); | |
2380 | } | |
b5679fa4 FZ |
2381 | |
2382 | int coroutine_fn blk_co_copy_range(BlockBackend *blk_in, int64_t off_in, | |
2383 | BlockBackend *blk_out, int64_t off_out, | |
67b51fb9 VSO |
2384 | int bytes, BdrvRequestFlags read_flags, |
2385 | BdrvRequestFlags write_flags) | |
b5679fa4 FZ |
2386 | { |
2387 | int r; | |
2388 | r = blk_check_byte_request(blk_in, off_in, bytes); | |
2389 | if (r) { | |
2390 | return r; | |
2391 | } | |
2392 | r = blk_check_byte_request(blk_out, off_out, bytes); | |
2393 | if (r) { | |
2394 | return r; | |
2395 | } | |
2396 | return bdrv_co_copy_range(blk_in->root, off_in, | |
2397 | blk_out->root, off_out, | |
67b51fb9 | 2398 | bytes, read_flags, write_flags); |
b5679fa4 | 2399 | } |
5d3b4e99 VSO |
2400 | |
2401 | const BdrvChild *blk_root(BlockBackend *blk) | |
2402 | { | |
2403 | return blk->root; | |
2404 | } | |
2b7bbdbd HR |
2405 | |
2406 | int blk_make_empty(BlockBackend *blk, Error **errp) | |
2407 | { | |
2408 | if (!blk_is_available(blk)) { | |
2409 | error_setg(errp, "No medium inserted"); | |
2410 | return -ENOMEDIUM; | |
2411 | } | |
2412 | ||
2413 | return bdrv_make_empty(blk->root, errp); | |
2414 | } |