]>
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" |
18e46a03 | 18 | #include "sysemu/blockdev.h" |
373340b2 | 19 | #include "sysemu/sysemu.h" |
e688df6b | 20 | #include "qapi/error.h" |
9af23989 | 21 | #include "qapi/qapi-events-block.h" |
f348b6d1 | 22 | #include "qemu/id.h" |
922a01a0 | 23 | #include "qemu/option.h" |
1e98fefd | 24 | #include "trace.h" |
5f7772c4 | 25 | #include "migration/misc.h" |
a7f53e26 MA |
26 | |
27 | /* Number of coroutines to reserve per attached device model */ | |
28 | #define COROUTINE_POOL_RESERVATION 64 | |
26f54e9a | 29 | |
1bf1cbc9 KW |
30 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
31 | ||
4981bdec HR |
32 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb); |
33 | ||
d03654ea SH |
34 | typedef struct BlockBackendAioNotifier { |
35 | void (*attached_aio_context)(AioContext *new_context, void *opaque); | |
36 | void (*detach_aio_context)(void *opaque); | |
37 | void *opaque; | |
38 | QLIST_ENTRY(BlockBackendAioNotifier) list; | |
39 | } BlockBackendAioNotifier; | |
40 | ||
26f54e9a MA |
41 | struct BlockBackend { |
42 | char *name; | |
43 | int refcnt; | |
f21d96d0 | 44 | BdrvChild *root; |
26f8b3a8 | 45 | DriveInfo *legacy_dinfo; /* null unless created by drive_new() */ |
2cf22d6a | 46 | QTAILQ_ENTRY(BlockBackend) link; /* for block_backends */ |
9492b0b9 | 47 | QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */ |
f2cd875d | 48 | BlockBackendPublic public; |
a7f53e26 | 49 | |
d09ea2d2 | 50 | DeviceState *dev; /* attached device model, if any */ |
a7f53e26 MA |
51 | const BlockDevOps *dev_ops; |
52 | void *dev_opaque; | |
68e9ec01 HR |
53 | |
54 | /* the block size for which the guest device expects atomicity */ | |
55 | int guest_block_size; | |
7f0e9da6 | 56 | |
281d22d8 HR |
57 | /* If the BDS tree is removed, some of its options are stored here (which |
58 | * can be used to restore those options in the new BDS on insert) */ | |
59 | BlockBackendRootState root_state; | |
60 | ||
bfd18d1e KW |
61 | bool enable_write_cache; |
62 | ||
7f0e9da6 HR |
63 | /* I/O stats (display with "info blockstats"). */ |
64 | BlockAcctStats stats; | |
373340b2 HR |
65 | |
66 | BlockdevOnError on_read_error, on_write_error; | |
67 | bool iostatus_enabled; | |
68 | BlockDeviceIoStatus iostatus; | |
3301f6c6 | 69 | |
981776b3 KW |
70 | uint64_t perm; |
71 | uint64_t shared_perm; | |
d35ff5e6 | 72 | bool disable_perm; |
981776b3 | 73 | |
980b0f94 | 74 | bool allow_aio_context_change; |
c10c9d96 KW |
75 | bool allow_write_beyond_eof; |
76 | ||
3301f6c6 | 77 | NotifierList remove_bs_notifiers, insert_bs_notifiers; |
d03654ea | 78 | QLIST_HEAD(, BlockBackendAioNotifier) aio_notifiers; |
f4d9cc88 JS |
79 | |
80 | int quiesce_counter; | |
5f7772c4 | 81 | VMChangeStateEntry *vmsh; |
ca2e2144 | 82 | bool force_allow_inactivate; |
33f2a757 SH |
83 | |
84 | /* Number of in-flight aio requests. BlockDriverState also counts | |
85 | * in-flight requests but aio requests can exist even when blk->root is | |
86 | * NULL, so we cannot rely on its counter for that case. | |
87 | * Accessed with atomic ops. | |
88 | */ | |
89 | unsigned int in_flight; | |
26f54e9a MA |
90 | }; |
91 | ||
e7f7d676 HR |
92 | typedef struct BlockBackendAIOCB { |
93 | BlockAIOCB common; | |
4981bdec | 94 | BlockBackend *blk; |
e7f7d676 HR |
95 | int ret; |
96 | } BlockBackendAIOCB; | |
97 | ||
98 | static const AIOCBInfo block_backend_aiocb_info = { | |
4981bdec | 99 | .get_aio_context = blk_aiocb_get_aio_context, |
e7f7d676 HR |
100 | .aiocb_size = sizeof(BlockBackendAIOCB), |
101 | }; | |
102 | ||
8fb3c76c | 103 | static void drive_info_del(DriveInfo *dinfo); |
7c8eece4 | 104 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs); |
8fb3c76c | 105 | |
2cf22d6a HR |
106 | /* All BlockBackends */ |
107 | static QTAILQ_HEAD(, BlockBackend) block_backends = | |
108 | QTAILQ_HEAD_INITIALIZER(block_backends); | |
109 | ||
9492b0b9 HR |
110 | /* All BlockBackends referenced by the monitor and which are iterated through by |
111 | * blk_next() */ | |
112 | static QTAILQ_HEAD(, BlockBackend) monitor_block_backends = | |
113 | QTAILQ_HEAD_INITIALIZER(monitor_block_backends); | |
26f54e9a | 114 | |
f21d96d0 KW |
115 | static void blk_root_inherit_options(int *child_flags, QDict *child_options, |
116 | int parent_flags, QDict *parent_options) | |
117 | { | |
118 | /* We're not supposed to call this function for root nodes */ | |
119 | abort(); | |
120 | } | |
c2066af0 | 121 | static void blk_root_drained_begin(BdrvChild *child); |
fe5258a5 | 122 | static bool blk_root_drained_poll(BdrvChild *child); |
c2066af0 | 123 | static void blk_root_drained_end(BdrvChild *child); |
f21d96d0 | 124 | |
5c8cab48 KW |
125 | static void blk_root_change_media(BdrvChild *child, bool load); |
126 | static void blk_root_resize(BdrvChild *child); | |
127 | ||
38475269 KW |
128 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, |
129 | GSList **ignore, Error **errp); | |
130 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
131 | GSList **ignore); | |
132 | ||
b5411555 KW |
133 | static char *blk_root_get_parent_desc(BdrvChild *child) |
134 | { | |
135 | BlockBackend *blk = child->opaque; | |
136 | char *dev_id; | |
137 | ||
138 | if (blk->name) { | |
139 | return g_strdup(blk->name); | |
140 | } | |
141 | ||
142 | dev_id = blk_get_attached_dev_id(blk); | |
143 | if (*dev_id) { | |
144 | return dev_id; | |
145 | } else { | |
146 | /* TODO Callback into the BB owner for something more detailed */ | |
147 | g_free(dev_id); | |
148 | return g_strdup("a block device"); | |
149 | } | |
150 | } | |
151 | ||
4c265bf9 KW |
152 | static const char *blk_root_get_name(BdrvChild *child) |
153 | { | |
154 | return blk_name(child->opaque); | |
155 | } | |
156 | ||
5f7772c4 FZ |
157 | static void blk_vm_state_changed(void *opaque, int running, RunState state) |
158 | { | |
159 | Error *local_err = NULL; | |
160 | BlockBackend *blk = opaque; | |
161 | ||
162 | if (state == RUN_STATE_INMIGRATE) { | |
163 | return; | |
164 | } | |
165 | ||
166 | qemu_del_vm_change_state_handler(blk->vmsh); | |
167 | blk->vmsh = NULL; | |
168 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); | |
169 | if (local_err) { | |
170 | error_report_err(local_err); | |
171 | } | |
172 | } | |
173 | ||
4417ab7a KW |
174 | /* |
175 | * Notifies the user of the BlockBackend that migration has completed. qdev | |
176 | * devices can tighten their permissions in response (specifically revoke | |
177 | * shared write permissions that we needed for storage migration). | |
178 | * | |
179 | * If an error is returned, the VM cannot be allowed to be resumed. | |
180 | */ | |
181 | static void blk_root_activate(BdrvChild *child, Error **errp) | |
182 | { | |
183 | BlockBackend *blk = child->opaque; | |
184 | Error *local_err = NULL; | |
185 | ||
186 | if (!blk->disable_perm) { | |
187 | return; | |
188 | } | |
189 | ||
190 | blk->disable_perm = false; | |
191 | ||
5f7772c4 FZ |
192 | blk_set_perm(blk, blk->perm, BLK_PERM_ALL, &local_err); |
193 | if (local_err) { | |
194 | error_propagate(errp, local_err); | |
195 | blk->disable_perm = true; | |
196 | return; | |
197 | } | |
198 | ||
199 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
200 | /* Activation can happen when migration process is still active, for | |
201 | * example when nbd_server_add is called during non-shared storage | |
202 | * migration. Defer the shared_perm update to migration completion. */ | |
203 | if (!blk->vmsh) { | |
204 | blk->vmsh = qemu_add_vm_change_state_handler(blk_vm_state_changed, | |
205 | blk); | |
206 | } | |
207 | return; | |
208 | } | |
209 | ||
4417ab7a KW |
210 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); |
211 | if (local_err) { | |
212 | error_propagate(errp, local_err); | |
213 | blk->disable_perm = true; | |
214 | return; | |
215 | } | |
216 | } | |
217 | ||
ca2e2144 FZ |
218 | void blk_set_force_allow_inactivate(BlockBackend *blk) |
219 | { | |
220 | blk->force_allow_inactivate = true; | |
221 | } | |
222 | ||
c16de8f5 FZ |
223 | static bool blk_can_inactivate(BlockBackend *blk) |
224 | { | |
ca2e2144 | 225 | /* If it is a guest device, inactivate is ok. */ |
c16de8f5 FZ |
226 | if (blk->dev || blk_name(blk)[0]) { |
227 | return true; | |
228 | } | |
229 | ||
ca2e2144 FZ |
230 | /* Inactivating means no more writes to the image can be done, |
231 | * even if those writes would be changes invisible to the | |
232 | * guest. For block job BBs that satisfy this, we can just allow | |
233 | * it. This is the case for mirror job source, which is required | |
234 | * by libvirt non-shared block migration. */ | |
235 | if (!(blk->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED))) { | |
236 | return true; | |
237 | } | |
238 | ||
239 | return blk->force_allow_inactivate; | |
c16de8f5 FZ |
240 | } |
241 | ||
cfa1a572 KW |
242 | static int blk_root_inactivate(BdrvChild *child) |
243 | { | |
244 | BlockBackend *blk = child->opaque; | |
245 | ||
246 | if (blk->disable_perm) { | |
247 | return 0; | |
248 | } | |
249 | ||
c16de8f5 | 250 | if (!blk_can_inactivate(blk)) { |
cfa1a572 KW |
251 | return -EPERM; |
252 | } | |
253 | ||
254 | blk->disable_perm = true; | |
255 | if (blk->root) { | |
256 | bdrv_child_try_set_perm(blk->root, 0, BLK_PERM_ALL, &error_abort); | |
257 | } | |
258 | ||
259 | return 0; | |
260 | } | |
261 | ||
d03654ea SH |
262 | static void blk_root_attach(BdrvChild *child) |
263 | { | |
264 | BlockBackend *blk = child->opaque; | |
265 | BlockBackendAioNotifier *notifier; | |
266 | ||
267 | trace_blk_root_attach(child, blk, child->bs); | |
268 | ||
269 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
270 | bdrv_add_aio_context_notifier(child->bs, | |
271 | notifier->attached_aio_context, | |
272 | notifier->detach_aio_context, | |
273 | notifier->opaque); | |
274 | } | |
275 | } | |
276 | ||
277 | static void blk_root_detach(BdrvChild *child) | |
278 | { | |
279 | BlockBackend *blk = child->opaque; | |
280 | BlockBackendAioNotifier *notifier; | |
281 | ||
282 | trace_blk_root_detach(child, blk, child->bs); | |
283 | ||
284 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
285 | bdrv_remove_aio_context_notifier(child->bs, | |
286 | notifier->attached_aio_context, | |
287 | notifier->detach_aio_context, | |
288 | notifier->opaque); | |
289 | } | |
290 | } | |
291 | ||
f21d96d0 | 292 | static const BdrvChildRole child_root = { |
c2066af0 KW |
293 | .inherit_options = blk_root_inherit_options, |
294 | ||
5c8cab48 KW |
295 | .change_media = blk_root_change_media, |
296 | .resize = blk_root_resize, | |
4c265bf9 | 297 | .get_name = blk_root_get_name, |
b5411555 | 298 | .get_parent_desc = blk_root_get_parent_desc, |
5c8cab48 | 299 | |
c2066af0 | 300 | .drained_begin = blk_root_drained_begin, |
fe5258a5 | 301 | .drained_poll = blk_root_drained_poll, |
c2066af0 | 302 | .drained_end = blk_root_drained_end, |
4417ab7a KW |
303 | |
304 | .activate = blk_root_activate, | |
cfa1a572 | 305 | .inactivate = blk_root_inactivate, |
d03654ea SH |
306 | |
307 | .attach = blk_root_attach, | |
308 | .detach = blk_root_detach, | |
38475269 KW |
309 | |
310 | .can_set_aio_ctx = blk_root_can_set_aio_ctx, | |
311 | .set_aio_ctx = blk_root_set_aio_ctx, | |
f21d96d0 KW |
312 | }; |
313 | ||
26f54e9a | 314 | /* |
efaa7c4e | 315 | * Create a new BlockBackend with a reference count of one. |
6d0eb64d KW |
316 | * |
317 | * @perm is a bitmasks of BLK_PERM_* constants which describes the permissions | |
318 | * to request for a block driver node that is attached to this BlockBackend. | |
319 | * @shared_perm is a bitmask which describes which permissions may be granted | |
320 | * to other users of the attached node. | |
321 | * Both sets of permissions can be changed later using blk_set_perm(). | |
322 | * | |
26f54e9a MA |
323 | * Return the new BlockBackend on success, null on failure. |
324 | */ | |
6d0eb64d | 325 | BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm) |
26f54e9a MA |
326 | { |
327 | BlockBackend *blk; | |
328 | ||
26f54e9a | 329 | blk = g_new0(BlockBackend, 1); |
26f54e9a | 330 | blk->refcnt = 1; |
6d0eb64d KW |
331 | blk->perm = perm; |
332 | blk->shared_perm = shared_perm; | |
0c3169df KW |
333 | blk_set_enable_write_cache(blk, true); |
334 | ||
cb53460b KW |
335 | blk->on_read_error = BLOCKDEV_ON_ERROR_REPORT; |
336 | blk->on_write_error = BLOCKDEV_ON_ERROR_ENOSPC; | |
337 | ||
9caa6f3d | 338 | block_acct_init(&blk->stats); |
27ccdd52 | 339 | |
3301f6c6 HR |
340 | notifier_list_init(&blk->remove_bs_notifiers); |
341 | notifier_list_init(&blk->insert_bs_notifiers); | |
d03654ea | 342 | QLIST_INIT(&blk->aio_notifiers); |
27ccdd52 | 343 | |
2cf22d6a | 344 | QTAILQ_INSERT_TAIL(&block_backends, blk, link); |
26f54e9a MA |
345 | return blk; |
346 | } | |
347 | ||
7e7d56d9 | 348 | /* |
28eb9b12 | 349 | * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. |
ca49a4fd HR |
350 | * |
351 | * Just as with bdrv_open(), after having called this function the reference to | |
352 | * @options belongs to the block layer (even on failure). | |
353 | * | |
354 | * TODO: Remove @filename and @flags; it should be possible to specify a whole | |
355 | * BDS tree just by specifying the @options QDict (or @reference, | |
356 | * alternatively). At the time of adding this function, this is not possible, | |
357 | * though, so callers of this function have to be able to specify @filename and | |
358 | * @flags. | |
359 | */ | |
efaa7c4e HR |
360 | BlockBackend *blk_new_open(const char *filename, const char *reference, |
361 | QDict *options, int flags, Error **errp) | |
ca49a4fd HR |
362 | { |
363 | BlockBackend *blk; | |
28eb9b12 | 364 | BlockDriverState *bs; |
1f4ad7d3 | 365 | uint64_t perm = 0; |
c62d32f5 KW |
366 | |
367 | /* blk_new_open() is mainly used in .bdrv_create implementations and the | |
368 | * tools where sharing isn't a concern because the BDS stays private, so we | |
369 | * just request permission according to the flags. | |
370 | * | |
371 | * The exceptions are xen_disk and blockdev_init(); in these cases, the | |
372 | * caller of blk_new_open() doesn't make use of the permissions, but they | |
373 | * shouldn't hurt either. We can still share everything here because the | |
374 | * guest devices will add their own blockers if they can't share. */ | |
1f4ad7d3 KW |
375 | if ((flags & BDRV_O_NO_IO) == 0) { |
376 | perm |= BLK_PERM_CONSISTENT_READ; | |
377 | if (flags & BDRV_O_RDWR) { | |
378 | perm |= BLK_PERM_WRITE; | |
379 | } | |
c62d32f5 KW |
380 | } |
381 | if (flags & BDRV_O_RESIZE) { | |
382 | perm |= BLK_PERM_RESIZE; | |
383 | } | |
ca49a4fd | 384 | |
c62d32f5 | 385 | blk = blk_new(perm, BLK_PERM_ALL); |
5b363937 HR |
386 | bs = bdrv_open(filename, reference, options, flags, errp); |
387 | if (!bs) { | |
ca49a4fd HR |
388 | blk_unref(blk); |
389 | return NULL; | |
390 | } | |
391 | ||
d5e6f437 | 392 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
50bfbe93 FZ |
393 | perm, BLK_PERM_ALL, blk, errp); |
394 | if (!blk->root) { | |
395 | bdrv_unref(bs); | |
396 | blk_unref(blk); | |
397 | return NULL; | |
398 | } | |
72e775c7 | 399 | |
ca49a4fd HR |
400 | return blk; |
401 | } | |
402 | ||
26f54e9a MA |
403 | static void blk_delete(BlockBackend *blk) |
404 | { | |
405 | assert(!blk->refcnt); | |
e5e78550 | 406 | assert(!blk->name); |
a7f53e26 | 407 | assert(!blk->dev); |
022cdc9f | 408 | if (blk->public.throttle_group_member.throttle_state) { |
1606e4cf EB |
409 | blk_io_limits_disable(blk); |
410 | } | |
f21d96d0 | 411 | if (blk->root) { |
13855c6b | 412 | blk_remove_bs(blk); |
7e7d56d9 | 413 | } |
5f7772c4 FZ |
414 | if (blk->vmsh) { |
415 | qemu_del_vm_change_state_handler(blk->vmsh); | |
416 | blk->vmsh = NULL; | |
417 | } | |
3301f6c6 HR |
418 | assert(QLIST_EMPTY(&blk->remove_bs_notifiers.notifiers)); |
419 | assert(QLIST_EMPTY(&blk->insert_bs_notifiers.notifiers)); | |
d03654ea | 420 | assert(QLIST_EMPTY(&blk->aio_notifiers)); |
2cf22d6a | 421 | QTAILQ_REMOVE(&block_backends, blk, link); |
18e46a03 | 422 | drive_info_del(blk->legacy_dinfo); |
979e9b03 | 423 | block_acct_cleanup(&blk->stats); |
26f54e9a MA |
424 | g_free(blk); |
425 | } | |
426 | ||
8fb3c76c MA |
427 | static void drive_info_del(DriveInfo *dinfo) |
428 | { | |
429 | if (!dinfo) { | |
430 | return; | |
431 | } | |
432 | qemu_opts_del(dinfo->opts); | |
8fb3c76c MA |
433 | g_free(dinfo); |
434 | } | |
435 | ||
f636ae85 AG |
436 | int blk_get_refcnt(BlockBackend *blk) |
437 | { | |
438 | return blk ? blk->refcnt : 0; | |
439 | } | |
440 | ||
26f54e9a MA |
441 | /* |
442 | * Increment @blk's reference count. | |
443 | * @blk must not be null. | |
444 | */ | |
445 | void blk_ref(BlockBackend *blk) | |
446 | { | |
5ca9d21b | 447 | assert(blk->refcnt > 0); |
26f54e9a MA |
448 | blk->refcnt++; |
449 | } | |
450 | ||
451 | /* | |
452 | * Decrement @blk's reference count. | |
453 | * If this drops it to zero, destroy @blk. | |
454 | * For convenience, do nothing if @blk is null. | |
455 | */ | |
456 | void blk_unref(BlockBackend *blk) | |
457 | { | |
458 | if (blk) { | |
459 | assert(blk->refcnt > 0); | |
5ca9d21b KW |
460 | if (blk->refcnt > 1) { |
461 | blk->refcnt--; | |
462 | } else { | |
463 | blk_drain(blk); | |
464 | /* blk_drain() cannot resurrect blk, nobody held a reference */ | |
465 | assert(blk->refcnt == 1); | |
466 | blk->refcnt = 0; | |
26f54e9a MA |
467 | blk_delete(blk); |
468 | } | |
469 | } | |
470 | } | |
471 | ||
2cf22d6a HR |
472 | /* |
473 | * Behaves similarly to blk_next() but iterates over all BlockBackends, even the | |
474 | * ones which are hidden (i.e. are not referenced by the monitor). | |
475 | */ | |
a429b9b5 | 476 | BlockBackend *blk_all_next(BlockBackend *blk) |
2cf22d6a HR |
477 | { |
478 | return blk ? QTAILQ_NEXT(blk, link) | |
479 | : QTAILQ_FIRST(&block_backends); | |
480 | } | |
481 | ||
d8da3cef HR |
482 | void blk_remove_all_bs(void) |
483 | { | |
74d1b8fc | 484 | BlockBackend *blk = NULL; |
d8da3cef | 485 | |
2cf22d6a | 486 | while ((blk = blk_all_next(blk)) != NULL) { |
d8da3cef HR |
487 | AioContext *ctx = blk_get_aio_context(blk); |
488 | ||
489 | aio_context_acquire(ctx); | |
f21d96d0 | 490 | if (blk->root) { |
d8da3cef HR |
491 | blk_remove_bs(blk); |
492 | } | |
493 | aio_context_release(ctx); | |
494 | } | |
495 | } | |
496 | ||
26f54e9a | 497 | /* |
9492b0b9 | 498 | * Return the monitor-owned BlockBackend after @blk. |
26f54e9a MA |
499 | * If @blk is null, return the first one. |
500 | * Else, return @blk's next sibling, which may be null. | |
501 | * | |
502 | * To iterate over all BlockBackends, do | |
503 | * for (blk = blk_next(NULL); blk; blk = blk_next(blk)) { | |
504 | * ... | |
505 | * } | |
506 | */ | |
507 | BlockBackend *blk_next(BlockBackend *blk) | |
508 | { | |
9492b0b9 HR |
509 | return blk ? QTAILQ_NEXT(blk, monitor_link) |
510 | : QTAILQ_FIRST(&monitor_block_backends); | |
26f54e9a MA |
511 | } |
512 | ||
7c8eece4 KW |
513 | /* Iterates over all top-level BlockDriverStates, i.e. BDSs that are owned by |
514 | * the monitor or attached to a BlockBackend */ | |
88be7b4b | 515 | BlockDriverState *bdrv_next(BdrvNextIterator *it) |
7c8eece4 | 516 | { |
5e003f17 HR |
517 | BlockDriverState *bs, *old_bs; |
518 | ||
519 | /* Must be called from the main loop */ | |
520 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
981f4f57 | 521 | |
7c8eece4 KW |
522 | /* First, return all root nodes of BlockBackends. In order to avoid |
523 | * returning a BDS twice when multiple BBs refer to it, we only return it | |
524 | * if the BB is the first one in the parent list of the BDS. */ | |
525 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
5e003f17 HR |
526 | BlockBackend *old_blk = it->blk; |
527 | ||
528 | old_bs = old_blk ? blk_bs(old_blk) : NULL; | |
529 | ||
7c8eece4 KW |
530 | do { |
531 | it->blk = blk_all_next(it->blk); | |
88be7b4b KW |
532 | bs = it->blk ? blk_bs(it->blk) : NULL; |
533 | } while (it->blk && (bs == NULL || bdrv_first_blk(bs) != it->blk)); | |
7c8eece4 | 534 | |
5e003f17 HR |
535 | if (it->blk) { |
536 | blk_ref(it->blk); | |
537 | } | |
538 | blk_unref(old_blk); | |
539 | ||
88be7b4b | 540 | if (bs) { |
5e003f17 HR |
541 | bdrv_ref(bs); |
542 | bdrv_unref(old_bs); | |
88be7b4b | 543 | return bs; |
7c8eece4 KW |
544 | } |
545 | it->phase = BDRV_NEXT_MONITOR_OWNED; | |
5e003f17 HR |
546 | } else { |
547 | old_bs = it->bs; | |
7c8eece4 KW |
548 | } |
549 | ||
550 | /* Then return the monitor-owned BDSes without a BB attached. Ignore all | |
551 | * BDSes that are attached to a BlockBackend here; they have been handled | |
552 | * by the above block already */ | |
981f4f57 | 553 | do { |
7c8eece4 | 554 | it->bs = bdrv_next_monitor_owned(it->bs); |
88be7b4b KW |
555 | bs = it->bs; |
556 | } while (bs && bdrv_has_blk(bs)); | |
557 | ||
5e003f17 HR |
558 | if (bs) { |
559 | bdrv_ref(bs); | |
560 | } | |
561 | bdrv_unref(old_bs); | |
562 | ||
88be7b4b KW |
563 | return bs; |
564 | } | |
565 | ||
5e003f17 | 566 | static void bdrv_next_reset(BdrvNextIterator *it) |
88be7b4b KW |
567 | { |
568 | *it = (BdrvNextIterator) { | |
569 | .phase = BDRV_NEXT_BACKEND_ROOTS, | |
570 | }; | |
5e003f17 | 571 | } |
981f4f57 | 572 | |
5e003f17 HR |
573 | BlockDriverState *bdrv_first(BdrvNextIterator *it) |
574 | { | |
575 | bdrv_next_reset(it); | |
88be7b4b | 576 | return bdrv_next(it); |
981f4f57 HR |
577 | } |
578 | ||
5e003f17 HR |
579 | /* Must be called when aborting a bdrv_next() iteration before |
580 | * bdrv_next() returns NULL */ | |
581 | void bdrv_next_cleanup(BdrvNextIterator *it) | |
582 | { | |
583 | /* Must be called from the main loop */ | |
584 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
585 | ||
586 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
587 | if (it->blk) { | |
588 | bdrv_unref(blk_bs(it->blk)); | |
589 | blk_unref(it->blk); | |
590 | } | |
591 | } else { | |
592 | bdrv_unref(it->bs); | |
593 | } | |
594 | ||
595 | bdrv_next_reset(it); | |
596 | } | |
597 | ||
e5e78550 HR |
598 | /* |
599 | * Add a BlockBackend into the list of backends referenced by the monitor, with | |
600 | * the given @name acting as the handle for the monitor. | |
601 | * Strictly for use by blockdev.c. | |
602 | * | |
603 | * @name must not be null or empty. | |
604 | * | |
605 | * Returns true on success and false on failure. In the latter case, an Error | |
606 | * object is returned through @errp. | |
607 | */ | |
608 | bool monitor_add_blk(BlockBackend *blk, const char *name, Error **errp) | |
609 | { | |
610 | assert(!blk->name); | |
611 | assert(name && name[0]); | |
612 | ||
613 | if (!id_wellformed(name)) { | |
614 | error_setg(errp, "Invalid device name"); | |
615 | return false; | |
616 | } | |
617 | if (blk_by_name(name)) { | |
618 | error_setg(errp, "Device with id '%s' already exists", name); | |
619 | return false; | |
620 | } | |
621 | if (bdrv_find_node(name)) { | |
622 | error_setg(errp, | |
623 | "Device name '%s' conflicts with an existing node name", | |
624 | name); | |
625 | return false; | |
626 | } | |
627 | ||
628 | blk->name = g_strdup(name); | |
629 | QTAILQ_INSERT_TAIL(&monitor_block_backends, blk, monitor_link); | |
630 | return true; | |
631 | } | |
632 | ||
633 | /* | |
634 | * Remove a BlockBackend from the list of backends referenced by the monitor. | |
635 | * Strictly for use by blockdev.c. | |
636 | */ | |
637 | void monitor_remove_blk(BlockBackend *blk) | |
638 | { | |
639 | if (!blk->name) { | |
640 | return; | |
641 | } | |
642 | ||
643 | QTAILQ_REMOVE(&monitor_block_backends, blk, monitor_link); | |
644 | g_free(blk->name); | |
645 | blk->name = NULL; | |
646 | } | |
647 | ||
26f54e9a | 648 | /* |
7e7d56d9 | 649 | * Return @blk's name, a non-null string. |
e5e78550 | 650 | * Returns an empty string iff @blk is not referenced by the monitor. |
26f54e9a | 651 | */ |
0731a50f | 652 | const char *blk_name(const BlockBackend *blk) |
26f54e9a | 653 | { |
e5e78550 | 654 | return blk->name ?: ""; |
26f54e9a MA |
655 | } |
656 | ||
657 | /* | |
658 | * Return the BlockBackend with name @name if it exists, else null. | |
659 | * @name must not be null. | |
660 | */ | |
661 | BlockBackend *blk_by_name(const char *name) | |
662 | { | |
74d1b8fc | 663 | BlockBackend *blk = NULL; |
26f54e9a MA |
664 | |
665 | assert(name); | |
74d1b8fc | 666 | while ((blk = blk_next(blk)) != NULL) { |
26f54e9a MA |
667 | if (!strcmp(name, blk->name)) { |
668 | return blk; | |
669 | } | |
670 | } | |
671 | return NULL; | |
672 | } | |
7e7d56d9 MA |
673 | |
674 | /* | |
675 | * Return the BlockDriverState attached to @blk if any, else null. | |
676 | */ | |
677 | BlockDriverState *blk_bs(BlockBackend *blk) | |
678 | { | |
f21d96d0 | 679 | return blk->root ? blk->root->bs : NULL; |
7e7d56d9 MA |
680 | } |
681 | ||
7c8eece4 | 682 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs) |
dde33812 KW |
683 | { |
684 | BdrvChild *child; | |
685 | QLIST_FOREACH(child, &bs->parents, next_parent) { | |
686 | if (child->role == &child_root) { | |
7c8eece4 | 687 | return child->opaque; |
dde33812 KW |
688 | } |
689 | } | |
690 | ||
7c8eece4 KW |
691 | return NULL; |
692 | } | |
693 | ||
694 | /* | |
695 | * Returns true if @bs has an associated BlockBackend. | |
696 | */ | |
697 | bool bdrv_has_blk(BlockDriverState *bs) | |
698 | { | |
699 | return bdrv_first_blk(bs) != NULL; | |
dde33812 KW |
700 | } |
701 | ||
b6c1bae5 KW |
702 | /* |
703 | * Returns true if @bs has only BlockBackends as parents. | |
704 | */ | |
705 | bool bdrv_is_root_node(BlockDriverState *bs) | |
706 | { | |
707 | BdrvChild *c; | |
708 | ||
709 | QLIST_FOREACH(c, &bs->parents, next_parent) { | |
710 | if (c->role != &child_root) { | |
711 | return false; | |
712 | } | |
713 | } | |
714 | ||
715 | return true; | |
716 | } | |
717 | ||
18e46a03 MA |
718 | /* |
719 | * Return @blk's DriveInfo if any, else null. | |
720 | */ | |
721 | DriveInfo *blk_legacy_dinfo(BlockBackend *blk) | |
722 | { | |
723 | return blk->legacy_dinfo; | |
724 | } | |
725 | ||
726 | /* | |
727 | * Set @blk's DriveInfo to @dinfo, and return it. | |
728 | * @blk must not have a DriveInfo set already. | |
729 | * No other BlockBackend may have the same DriveInfo set. | |
730 | */ | |
731 | DriveInfo *blk_set_legacy_dinfo(BlockBackend *blk, DriveInfo *dinfo) | |
732 | { | |
733 | assert(!blk->legacy_dinfo); | |
734 | return blk->legacy_dinfo = dinfo; | |
735 | } | |
736 | ||
737 | /* | |
738 | * Return the BlockBackend with DriveInfo @dinfo. | |
739 | * It must exist. | |
740 | */ | |
741 | BlockBackend *blk_by_legacy_dinfo(DriveInfo *dinfo) | |
742 | { | |
74d1b8fc | 743 | BlockBackend *blk = NULL; |
18e46a03 | 744 | |
74d1b8fc | 745 | while ((blk = blk_next(blk)) != NULL) { |
18e46a03 MA |
746 | if (blk->legacy_dinfo == dinfo) { |
747 | return blk; | |
748 | } | |
749 | } | |
750 | abort(); | |
751 | } | |
752 | ||
f2cd875d KW |
753 | /* |
754 | * Returns a pointer to the publicly accessible fields of @blk. | |
755 | */ | |
756 | BlockBackendPublic *blk_get_public(BlockBackend *blk) | |
757 | { | |
758 | return &blk->public; | |
759 | } | |
760 | ||
761 | /* | |
762 | * Returns a BlockBackend given the associated @public fields. | |
763 | */ | |
764 | BlockBackend *blk_by_public(BlockBackendPublic *public) | |
765 | { | |
766 | return container_of(public, BlockBackend, public); | |
767 | } | |
768 | ||
1c95f7e1 HR |
769 | /* |
770 | * Disassociates the currently associated BlockDriverState from @blk. | |
771 | */ | |
772 | void blk_remove_bs(BlockBackend *blk) | |
773 | { | |
c89bcf3a | 774 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
632a7735 | 775 | BlockDriverState *bs; |
022cdc9f | 776 | |
3301f6c6 | 777 | notifier_list_notify(&blk->remove_bs_notifiers, blk); |
c89bcf3a | 778 | if (tgm->throttle_state) { |
632a7735 ZL |
779 | bs = blk_bs(blk); |
780 | bdrv_drained_begin(bs); | |
c89bcf3a AG |
781 | throttle_group_detach_aio_context(tgm); |
782 | throttle_group_attach_aio_context(tgm, qemu_get_aio_context()); | |
632a7735 | 783 | bdrv_drained_end(bs); |
a5614993 | 784 | } |
1c95f7e1 | 785 | |
7ca7f0f6 KW |
786 | blk_update_root_state(blk); |
787 | ||
f45280cb GK |
788 | /* bdrv_root_unref_child() will cause blk->root to become stale and may |
789 | * switch to a completion coroutine later on. Let's drain all I/O here | |
790 | * to avoid that and a potential QEMU crash. | |
791 | */ | |
792 | blk_drain(blk); | |
f21d96d0 KW |
793 | bdrv_root_unref_child(blk->root); |
794 | blk->root = NULL; | |
1c95f7e1 HR |
795 | } |
796 | ||
0c3c36d6 HR |
797 | /* |
798 | * Associates a new BlockDriverState with @blk. | |
799 | */ | |
d7086422 | 800 | int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) |
0c3c36d6 | 801 | { |
c89bcf3a | 802 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
d5e6f437 | 803 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
d7086422 KW |
804 | blk->perm, blk->shared_perm, blk, errp); |
805 | if (blk->root == NULL) { | |
806 | return -EPERM; | |
807 | } | |
808 | bdrv_ref(bs); | |
3301f6c6 HR |
809 | |
810 | notifier_list_notify(&blk->insert_bs_notifiers, blk); | |
c89bcf3a AG |
811 | if (tgm->throttle_state) { |
812 | throttle_group_detach_aio_context(tgm); | |
813 | throttle_group_attach_aio_context(tgm, bdrv_get_aio_context(bs)); | |
7ca7f0f6 | 814 | } |
d7086422 KW |
815 | |
816 | return 0; | |
0c3c36d6 HR |
817 | } |
818 | ||
981776b3 KW |
819 | /* |
820 | * Sets the permission bitmasks that the user of the BlockBackend needs. | |
821 | */ | |
822 | int blk_set_perm(BlockBackend *blk, uint64_t perm, uint64_t shared_perm, | |
823 | Error **errp) | |
824 | { | |
825 | int ret; | |
826 | ||
d35ff5e6 | 827 | if (blk->root && !blk->disable_perm) { |
981776b3 KW |
828 | ret = bdrv_child_try_set_perm(blk->root, perm, shared_perm, errp); |
829 | if (ret < 0) { | |
830 | return ret; | |
831 | } | |
832 | } | |
833 | ||
834 | blk->perm = perm; | |
835 | blk->shared_perm = shared_perm; | |
836 | ||
837 | return 0; | |
838 | } | |
839 | ||
887354bd KW |
840 | void blk_get_perm(BlockBackend *blk, uint64_t *perm, uint64_t *shared_perm) |
841 | { | |
842 | *perm = blk->perm; | |
843 | *shared_perm = blk->shared_perm; | |
844 | } | |
845 | ||
d09ea2d2 TH |
846 | /* |
847 | * Attach device model @dev to @blk. | |
848 | * Return 0 on success, -EBUSY when a device model is attached already. | |
849 | */ | |
850 | int blk_attach_dev(BlockBackend *blk, DeviceState *dev) | |
4be74634 | 851 | { |
a7f53e26 MA |
852 | if (blk->dev) { |
853 | return -EBUSY; | |
854 | } | |
d35ff5e6 KW |
855 | |
856 | /* While migration is still incoming, we don't need to apply the | |
857 | * permissions of guest device BlockBackends. We might still have a block | |
858 | * job or NBD server writing to the image for storage migration. */ | |
859 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
860 | blk->disable_perm = true; | |
861 | } | |
862 | ||
84ebe375 | 863 | blk_ref(blk); |
a7f53e26 | 864 | blk->dev = dev; |
373340b2 | 865 | blk_iostatus_reset(blk); |
d35ff5e6 | 866 | |
a7f53e26 | 867 | return 0; |
4be74634 MA |
868 | } |
869 | ||
a7f53e26 MA |
870 | /* |
871 | * Detach device model @dev from @blk. | |
872 | * @dev must be currently attached to @blk. | |
873 | */ | |
d09ea2d2 | 874 | void blk_detach_dev(BlockBackend *blk, DeviceState *dev) |
4be74634 | 875 | { |
a7f53e26 MA |
876 | assert(blk->dev == dev); |
877 | blk->dev = NULL; | |
878 | blk->dev_ops = NULL; | |
879 | blk->dev_opaque = NULL; | |
68e9ec01 | 880 | blk->guest_block_size = 512; |
981776b3 | 881 | blk_set_perm(blk, 0, BLK_PERM_ALL, &error_abort); |
84ebe375 | 882 | blk_unref(blk); |
4be74634 MA |
883 | } |
884 | ||
a7f53e26 MA |
885 | /* |
886 | * Return the device model attached to @blk if any, else null. | |
887 | */ | |
d09ea2d2 | 888 | DeviceState *blk_get_attached_dev(BlockBackend *blk) |
a7f53e26 MA |
889 | { |
890 | return blk->dev; | |
891 | } | |
892 | ||
2d76e724 KW |
893 | /* Return the qdev ID, or if no ID is assigned the QOM path, of the block |
894 | * device attached to the BlockBackend. */ | |
77beef83 | 895 | char *blk_get_attached_dev_id(BlockBackend *blk) |
2d76e724 | 896 | { |
d09ea2d2 | 897 | DeviceState *dev = blk->dev; |
2d76e724 KW |
898 | |
899 | if (!dev) { | |
900 | return g_strdup(""); | |
901 | } else if (dev->id) { | |
902 | return g_strdup(dev->id); | |
903 | } | |
602414d1 LM |
904 | |
905 | return object_get_canonical_path(OBJECT(dev)) ?: g_strdup(""); | |
2d76e724 KW |
906 | } |
907 | ||
1c89e1fa KW |
908 | /* |
909 | * Return the BlockBackend which has the device model @dev attached if it | |
910 | * exists, else null. | |
911 | * | |
912 | * @dev must not be null. | |
913 | */ | |
914 | BlockBackend *blk_by_dev(void *dev) | |
915 | { | |
916 | BlockBackend *blk = NULL; | |
917 | ||
918 | assert(dev != NULL); | |
919 | while ((blk = blk_all_next(blk)) != NULL) { | |
920 | if (blk->dev == dev) { | |
921 | return blk; | |
922 | } | |
923 | } | |
924 | return NULL; | |
925 | } | |
926 | ||
a7f53e26 MA |
927 | /* |
928 | * Set @blk's device model callbacks to @ops. | |
929 | * @opaque is the opaque argument to pass to the callbacks. | |
930 | * This is for use by device models. | |
931 | */ | |
932 | void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, | |
933 | void *opaque) | |
934 | { | |
935 | blk->dev_ops = ops; | |
936 | blk->dev_opaque = opaque; | |
f4d9cc88 JS |
937 | |
938 | /* Are we currently quiesced? Should we enforce this right now? */ | |
939 | if (blk->quiesce_counter && ops->drained_begin) { | |
940 | ops->drained_begin(opaque); | |
941 | } | |
a7f53e26 MA |
942 | } |
943 | ||
944 | /* | |
945 | * Notify @blk's attached device model of media change. | |
39829a01 KW |
946 | * |
947 | * If @load is true, notify of media load. This action can fail, meaning that | |
948 | * the medium cannot be loaded. @errp is set then. | |
949 | * | |
950 | * If @load is false, notify of media eject. This can never fail. | |
951 | * | |
a7f53e26 MA |
952 | * Also send DEVICE_TRAY_MOVED events as appropriate. |
953 | */ | |
39829a01 | 954 | void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp) |
a7f53e26 MA |
955 | { |
956 | if (blk->dev_ops && blk->dev_ops->change_media_cb) { | |
f1f57066 | 957 | bool tray_was_open, tray_is_open; |
39829a01 | 958 | Error *local_err = NULL; |
a7f53e26 | 959 | |
f1f57066 | 960 | tray_was_open = blk_dev_is_tray_open(blk); |
39829a01 KW |
961 | blk->dev_ops->change_media_cb(blk->dev_opaque, load, &local_err); |
962 | if (local_err) { | |
963 | assert(load == true); | |
964 | error_propagate(errp, local_err); | |
965 | return; | |
966 | } | |
f1f57066 HR |
967 | tray_is_open = blk_dev_is_tray_open(blk); |
968 | ||
969 | if (tray_was_open != tray_is_open) { | |
2d76e724 | 970 | char *id = blk_get_attached_dev_id(blk); |
3ab72385 | 971 | qapi_event_send_device_tray_moved(blk_name(blk), id, tray_is_open); |
2d76e724 | 972 | g_free(id); |
a7f53e26 MA |
973 | } |
974 | } | |
975 | } | |
976 | ||
5c8cab48 KW |
977 | static void blk_root_change_media(BdrvChild *child, bool load) |
978 | { | |
39829a01 | 979 | blk_dev_change_media_cb(child->opaque, load, NULL); |
5c8cab48 KW |
980 | } |
981 | ||
a7f53e26 MA |
982 | /* |
983 | * Does @blk's attached device model have removable media? | |
984 | * %true if no device model is attached. | |
985 | */ | |
986 | bool blk_dev_has_removable_media(BlockBackend *blk) | |
987 | { | |
988 | return !blk->dev || (blk->dev_ops && blk->dev_ops->change_media_cb); | |
989 | } | |
990 | ||
8f3a73bc HR |
991 | /* |
992 | * Does @blk's attached device model have a tray? | |
993 | */ | |
994 | bool blk_dev_has_tray(BlockBackend *blk) | |
995 | { | |
996 | return blk->dev_ops && blk->dev_ops->is_tray_open; | |
997 | } | |
998 | ||
a7f53e26 MA |
999 | /* |
1000 | * Notify @blk's attached device model of a media eject request. | |
1001 | * If @force is true, the medium is about to be yanked out forcefully. | |
1002 | */ | |
1003 | void blk_dev_eject_request(BlockBackend *blk, bool force) | |
4be74634 | 1004 | { |
a7f53e26 MA |
1005 | if (blk->dev_ops && blk->dev_ops->eject_request_cb) { |
1006 | blk->dev_ops->eject_request_cb(blk->dev_opaque, force); | |
1007 | } | |
4be74634 MA |
1008 | } |
1009 | ||
a7f53e26 MA |
1010 | /* |
1011 | * Does @blk's attached device model have a tray, and is it open? | |
1012 | */ | |
1013 | bool blk_dev_is_tray_open(BlockBackend *blk) | |
4be74634 | 1014 | { |
8f3a73bc | 1015 | if (blk_dev_has_tray(blk)) { |
a7f53e26 MA |
1016 | return blk->dev_ops->is_tray_open(blk->dev_opaque); |
1017 | } | |
1018 | return false; | |
1019 | } | |
1020 | ||
1021 | /* | |
1022 | * Does @blk's attached device model have the medium locked? | |
1023 | * %false if the device model has no such lock. | |
1024 | */ | |
1025 | bool blk_dev_is_medium_locked(BlockBackend *blk) | |
1026 | { | |
1027 | if (blk->dev_ops && blk->dev_ops->is_medium_locked) { | |
1028 | return blk->dev_ops->is_medium_locked(blk->dev_opaque); | |
1029 | } | |
1030 | return false; | |
1031 | } | |
1032 | ||
1033 | /* | |
1034 | * Notify @blk's attached device model of a backend size change. | |
1035 | */ | |
5c8cab48 | 1036 | static void blk_root_resize(BdrvChild *child) |
a7f53e26 | 1037 | { |
5c8cab48 KW |
1038 | BlockBackend *blk = child->opaque; |
1039 | ||
a7f53e26 MA |
1040 | if (blk->dev_ops && blk->dev_ops->resize_cb) { |
1041 | blk->dev_ops->resize_cb(blk->dev_opaque); | |
1042 | } | |
1043 | } | |
1044 | ||
1045 | void blk_iostatus_enable(BlockBackend *blk) | |
1046 | { | |
373340b2 HR |
1047 | blk->iostatus_enabled = true; |
1048 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
1049 | } | |
1050 | ||
1051 | /* The I/O status is only enabled if the drive explicitly | |
1052 | * enables it _and_ the VM is configured to stop on errors */ | |
1053 | bool blk_iostatus_is_enabled(const BlockBackend *blk) | |
1054 | { | |
1055 | return (blk->iostatus_enabled && | |
1056 | (blk->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || | |
1057 | blk->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
1058 | blk->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
1059 | } | |
1060 | ||
1061 | BlockDeviceIoStatus blk_iostatus(const BlockBackend *blk) | |
1062 | { | |
1063 | return blk->iostatus; | |
1064 | } | |
1065 | ||
1066 | void blk_iostatus_disable(BlockBackend *blk) | |
1067 | { | |
1068 | blk->iostatus_enabled = false; | |
1069 | } | |
1070 | ||
1071 | void blk_iostatus_reset(BlockBackend *blk) | |
1072 | { | |
1073 | if (blk_iostatus_is_enabled(blk)) { | |
f21d96d0 | 1074 | BlockDriverState *bs = blk_bs(blk); |
373340b2 | 1075 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
f21d96d0 KW |
1076 | if (bs && bs->job) { |
1077 | block_job_iostatus_reset(bs->job); | |
373340b2 HR |
1078 | } |
1079 | } | |
1080 | } | |
1081 | ||
1082 | void blk_iostatus_set_err(BlockBackend *blk, int error) | |
1083 | { | |
1084 | assert(blk_iostatus_is_enabled(blk)); | |
1085 | if (blk->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
1086 | blk->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : | |
1087 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
1088 | } | |
4be74634 MA |
1089 | } |
1090 | ||
c10c9d96 KW |
1091 | void blk_set_allow_write_beyond_eof(BlockBackend *blk, bool allow) |
1092 | { | |
1093 | blk->allow_write_beyond_eof = allow; | |
1094 | } | |
1095 | ||
980b0f94 KW |
1096 | void blk_set_allow_aio_context_change(BlockBackend *blk, bool allow) |
1097 | { | |
1098 | blk->allow_aio_context_change = allow; | |
1099 | } | |
1100 | ||
e7f7d676 HR |
1101 | static int blk_check_byte_request(BlockBackend *blk, int64_t offset, |
1102 | size_t size) | |
1103 | { | |
1104 | int64_t len; | |
1105 | ||
1106 | if (size > INT_MAX) { | |
1107 | return -EIO; | |
1108 | } | |
1109 | ||
c09ba36c | 1110 | if (!blk_is_available(blk)) { |
e7f7d676 HR |
1111 | return -ENOMEDIUM; |
1112 | } | |
1113 | ||
e7f7d676 HR |
1114 | if (offset < 0) { |
1115 | return -EIO; | |
1116 | } | |
1117 | ||
c10c9d96 KW |
1118 | if (!blk->allow_write_beyond_eof) { |
1119 | len = blk_getlength(blk); | |
1120 | if (len < 0) { | |
1121 | return len; | |
1122 | } | |
1123 | ||
1124 | if (offset > len || len - offset < size) { | |
1125 | return -EIO; | |
1126 | } | |
e7f7d676 HR |
1127 | } |
1128 | ||
1129 | return 0; | |
1130 | } | |
1131 | ||
1e98fefd KW |
1132 | int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset, |
1133 | unsigned int bytes, QEMUIOVector *qiov, | |
1134 | BdrvRequestFlags flags) | |
4be74634 | 1135 | { |
1e98fefd | 1136 | int ret; |
99723548 | 1137 | BlockDriverState *bs = blk_bs(blk); |
1e98fefd | 1138 | |
99723548 | 1139 | trace_blk_co_preadv(blk, bs, offset, bytes, flags); |
1e98fefd KW |
1140 | |
1141 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1142 | if (ret < 0) { |
1143 | return ret; | |
1144 | } | |
1145 | ||
99723548 PB |
1146 | bdrv_inc_in_flight(bs); |
1147 | ||
441565b2 | 1148 | /* throttling disk I/O */ |
022cdc9f MP |
1149 | if (blk->public.throttle_group_member.throttle_state) { |
1150 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1151 | bytes, false); | |
441565b2 KW |
1152 | } |
1153 | ||
99723548 PB |
1154 | ret = bdrv_co_preadv(blk->root, offset, bytes, qiov, flags); |
1155 | bdrv_dec_in_flight(bs); | |
1156 | return ret; | |
1bf1cbc9 KW |
1157 | } |
1158 | ||
1e98fefd KW |
1159 | int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset, |
1160 | unsigned int bytes, QEMUIOVector *qiov, | |
1161 | BdrvRequestFlags flags) | |
a8823a3b | 1162 | { |
bfd18d1e | 1163 | int ret; |
99723548 | 1164 | BlockDriverState *bs = blk_bs(blk); |
bfd18d1e | 1165 | |
99723548 | 1166 | trace_blk_co_pwritev(blk, bs, offset, bytes, flags); |
1e98fefd | 1167 | |
bfd18d1e | 1168 | ret = blk_check_byte_request(blk, offset, bytes); |
a8823a3b KW |
1169 | if (ret < 0) { |
1170 | return ret; | |
1171 | } | |
1172 | ||
99723548 | 1173 | bdrv_inc_in_flight(bs); |
441565b2 | 1174 | /* throttling disk I/O */ |
022cdc9f MP |
1175 | if (blk->public.throttle_group_member.throttle_state) { |
1176 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1177 | bytes, true); | |
441565b2 KW |
1178 | } |
1179 | ||
bfd18d1e KW |
1180 | if (!blk->enable_write_cache) { |
1181 | flags |= BDRV_REQ_FUA; | |
1182 | } | |
1183 | ||
99723548 PB |
1184 | ret = bdrv_co_pwritev(blk->root, offset, bytes, qiov, flags); |
1185 | bdrv_dec_in_flight(bs); | |
1186 | return ret; | |
a8823a3b KW |
1187 | } |
1188 | ||
1bf1cbc9 KW |
1189 | typedef struct BlkRwCo { |
1190 | BlockBackend *blk; | |
1191 | int64_t offset; | |
c060332c | 1192 | void *iobuf; |
1bf1cbc9 KW |
1193 | int ret; |
1194 | BdrvRequestFlags flags; | |
1195 | } BlkRwCo; | |
1196 | ||
1197 | static void blk_read_entry(void *opaque) | |
1198 | { | |
1199 | BlkRwCo *rwco = opaque; | |
c060332c | 1200 | QEMUIOVector *qiov = rwco->iobuf; |
1bf1cbc9 | 1201 | |
c060332c DS |
1202 | rwco->ret = blk_co_preadv(rwco->blk, rwco->offset, qiov->size, |
1203 | qiov, rwco->flags); | |
4720cbee | 1204 | aio_wait_kick(); |
1bf1cbc9 KW |
1205 | } |
1206 | ||
a8823a3b KW |
1207 | static void blk_write_entry(void *opaque) |
1208 | { | |
1209 | BlkRwCo *rwco = opaque; | |
c060332c | 1210 | QEMUIOVector *qiov = rwco->iobuf; |
a8823a3b | 1211 | |
c060332c DS |
1212 | rwco->ret = blk_co_pwritev(rwco->blk, rwco->offset, qiov->size, |
1213 | qiov, rwco->flags); | |
4720cbee | 1214 | aio_wait_kick(); |
a8823a3b KW |
1215 | } |
1216 | ||
a55d3fba KW |
1217 | static int blk_prw(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1218 | int64_t bytes, CoroutineEntry co_entry, | |
1219 | BdrvRequestFlags flags) | |
1bf1cbc9 | 1220 | { |
ae5a9592 VSO |
1221 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); |
1222 | BlkRwCo rwco = { | |
1bf1cbc9 | 1223 | .blk = blk, |
a55d3fba | 1224 | .offset = offset, |
c060332c | 1225 | .iobuf = &qiov, |
fc1453cd | 1226 | .flags = flags, |
1bf1cbc9 KW |
1227 | .ret = NOT_DONE, |
1228 | }; | |
1229 | ||
35f106e6 PB |
1230 | if (qemu_in_coroutine()) { |
1231 | /* Fast-path if already in coroutine context */ | |
1232 | co_entry(&rwco); | |
1233 | } else { | |
1234 | Coroutine *co = qemu_coroutine_create(co_entry, &rwco); | |
e92f0e19 | 1235 | bdrv_coroutine_enter(blk_bs(blk), co); |
35f106e6 PB |
1236 | BDRV_POLL_WHILE(blk_bs(blk), rwco.ret == NOT_DONE); |
1237 | } | |
1bf1cbc9 KW |
1238 | |
1239 | return rwco.ret; | |
4be74634 MA |
1240 | } |
1241 | ||
b7d17f9f EB |
1242 | int blk_pread_unthrottled(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1243 | int count) | |
4be74634 | 1244 | { |
5bd51196 KW |
1245 | int ret; |
1246 | ||
b7d17f9f | 1247 | ret = blk_check_byte_request(blk, offset, count); |
e7f7d676 HR |
1248 | if (ret < 0) { |
1249 | return ret; | |
1250 | } | |
1251 | ||
c2066af0 | 1252 | blk_root_drained_begin(blk->root); |
b7d17f9f | 1253 | ret = blk_pread(blk, offset, buf, count); |
c2066af0 | 1254 | blk_root_drained_end(blk->root); |
5bd51196 | 1255 | return ret; |
4be74634 MA |
1256 | } |
1257 | ||
d004bd52 | 1258 | int blk_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 1259 | int bytes, BdrvRequestFlags flags) |
0df89e8e | 1260 | { |
f5a5ca79 | 1261 | return blk_prw(blk, offset, NULL, bytes, blk_write_entry, |
983a1600 | 1262 | flags | BDRV_REQ_ZERO_WRITE); |
0df89e8e KW |
1263 | } |
1264 | ||
720ff280 KW |
1265 | int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags) |
1266 | { | |
1267 | return bdrv_make_zero(blk->root, flags); | |
1268 | } | |
1269 | ||
c90e2a9c | 1270 | void blk_inc_in_flight(BlockBackend *blk) |
33f2a757 SH |
1271 | { |
1272 | atomic_inc(&blk->in_flight); | |
1273 | } | |
1274 | ||
c90e2a9c | 1275 | void blk_dec_in_flight(BlockBackend *blk) |
33f2a757 SH |
1276 | { |
1277 | atomic_dec(&blk->in_flight); | |
cfe29d82 | 1278 | aio_wait_kick(); |
33f2a757 SH |
1279 | } |
1280 | ||
e7f7d676 HR |
1281 | static void error_callback_bh(void *opaque) |
1282 | { | |
1283 | struct BlockBackendAIOCB *acb = opaque; | |
99723548 | 1284 | |
33f2a757 | 1285 | blk_dec_in_flight(acb->blk); |
e7f7d676 HR |
1286 | acb->common.cb(acb->common.opaque, acb->ret); |
1287 | qemu_aio_unref(acb); | |
1288 | } | |
1289 | ||
ca78ecfa PL |
1290 | BlockAIOCB *blk_abort_aio_request(BlockBackend *blk, |
1291 | BlockCompletionFunc *cb, | |
1292 | void *opaque, int ret) | |
e7f7d676 HR |
1293 | { |
1294 | struct BlockBackendAIOCB *acb; | |
e7f7d676 | 1295 | |
33f2a757 | 1296 | blk_inc_in_flight(blk); |
e7f7d676 | 1297 | acb = blk_aio_get(&block_backend_aiocb_info, blk, cb, opaque); |
4981bdec | 1298 | acb->blk = blk; |
e7f7d676 HR |
1299 | acb->ret = ret; |
1300 | ||
fffb6e12 | 1301 | aio_bh_schedule_oneshot(blk_get_aio_context(blk), error_callback_bh, acb); |
e7f7d676 HR |
1302 | return &acb->common; |
1303 | } | |
1304 | ||
57d6a428 KW |
1305 | typedef struct BlkAioEmAIOCB { |
1306 | BlockAIOCB common; | |
1307 | BlkRwCo rwco; | |
7fa84cd8 | 1308 | int bytes; |
57d6a428 | 1309 | bool has_returned; |
57d6a428 KW |
1310 | } BlkAioEmAIOCB; |
1311 | ||
1312 | static const AIOCBInfo blk_aio_em_aiocb_info = { | |
1313 | .aiocb_size = sizeof(BlkAioEmAIOCB), | |
1314 | }; | |
1315 | ||
1316 | static void blk_aio_complete(BlkAioEmAIOCB *acb) | |
1317 | { | |
57d6a428 KW |
1318 | if (acb->has_returned) { |
1319 | acb->common.cb(acb->common.opaque, acb->rwco.ret); | |
46aaf2a5 | 1320 | blk_dec_in_flight(acb->rwco.blk); |
57d6a428 KW |
1321 | qemu_aio_unref(acb); |
1322 | } | |
1323 | } | |
1324 | ||
1325 | static void blk_aio_complete_bh(void *opaque) | |
1326 | { | |
fffb6e12 | 1327 | BlkAioEmAIOCB *acb = opaque; |
fffb6e12 PB |
1328 | assert(acb->has_returned); |
1329 | blk_aio_complete(acb); | |
57d6a428 KW |
1330 | } |
1331 | ||
7fa84cd8 | 1332 | static BlockAIOCB *blk_aio_prwv(BlockBackend *blk, int64_t offset, int bytes, |
c060332c | 1333 | void *iobuf, CoroutineEntry co_entry, |
57d6a428 KW |
1334 | BdrvRequestFlags flags, |
1335 | BlockCompletionFunc *cb, void *opaque) | |
1336 | { | |
1337 | BlkAioEmAIOCB *acb; | |
1338 | Coroutine *co; | |
1339 | ||
33f2a757 | 1340 | blk_inc_in_flight(blk); |
57d6a428 KW |
1341 | acb = blk_aio_get(&blk_aio_em_aiocb_info, blk, cb, opaque); |
1342 | acb->rwco = (BlkRwCo) { | |
1343 | .blk = blk, | |
1344 | .offset = offset, | |
c060332c | 1345 | .iobuf = iobuf, |
57d6a428 KW |
1346 | .flags = flags, |
1347 | .ret = NOT_DONE, | |
1348 | }; | |
7fa84cd8 | 1349 | acb->bytes = bytes; |
57d6a428 KW |
1350 | acb->has_returned = false; |
1351 | ||
0b8b8753 | 1352 | co = qemu_coroutine_create(co_entry, acb); |
e92f0e19 | 1353 | bdrv_coroutine_enter(blk_bs(blk), co); |
57d6a428 KW |
1354 | |
1355 | acb->has_returned = true; | |
1356 | if (acb->rwco.ret != NOT_DONE) { | |
fffb6e12 PB |
1357 | aio_bh_schedule_oneshot(blk_get_aio_context(blk), |
1358 | blk_aio_complete_bh, acb); | |
57d6a428 KW |
1359 | } |
1360 | ||
1361 | return &acb->common; | |
1362 | } | |
1363 | ||
1364 | static void blk_aio_read_entry(void *opaque) | |
1365 | { | |
1366 | BlkAioEmAIOCB *acb = opaque; | |
1367 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1368 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1369 | |
c060332c | 1370 | assert(qiov->size == acb->bytes); |
7fa84cd8 | 1371 | rwco->ret = blk_co_preadv(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1372 | qiov, rwco->flags); |
57d6a428 KW |
1373 | blk_aio_complete(acb); |
1374 | } | |
1375 | ||
1376 | static void blk_aio_write_entry(void *opaque) | |
1377 | { | |
1378 | BlkAioEmAIOCB *acb = opaque; | |
1379 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1380 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1381 | |
c060332c | 1382 | assert(!qiov || qiov->size == acb->bytes); |
7fa84cd8 | 1383 | rwco->ret = blk_co_pwritev(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1384 | qiov, rwco->flags); |
57d6a428 KW |
1385 | blk_aio_complete(acb); |
1386 | } | |
1387 | ||
d004bd52 EB |
1388 | BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
1389 | int count, BdrvRequestFlags flags, | |
1390 | BlockCompletionFunc *cb, void *opaque) | |
4be74634 | 1391 | { |
983a1600 EB |
1392 | return blk_aio_prwv(blk, offset, count, NULL, blk_aio_write_entry, |
1393 | flags | BDRV_REQ_ZERO_WRITE, cb, opaque); | |
4be74634 MA |
1394 | } |
1395 | ||
1396 | int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count) | |
1397 | { | |
a55d3fba | 1398 | int ret = blk_prw(blk, offset, buf, count, blk_read_entry, 0); |
e7f7d676 HR |
1399 | if (ret < 0) { |
1400 | return ret; | |
1401 | } | |
a55d3fba | 1402 | return count; |
4be74634 MA |
1403 | } |
1404 | ||
8341f00d EB |
1405 | int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count, |
1406 | BdrvRequestFlags flags) | |
4be74634 | 1407 | { |
8341f00d EB |
1408 | int ret = blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
1409 | flags); | |
e7f7d676 HR |
1410 | if (ret < 0) { |
1411 | return ret; | |
1412 | } | |
a55d3fba | 1413 | return count; |
4be74634 MA |
1414 | } |
1415 | ||
1416 | int64_t blk_getlength(BlockBackend *blk) | |
1417 | { | |
c09ba36c HR |
1418 | if (!blk_is_available(blk)) { |
1419 | return -ENOMEDIUM; | |
1420 | } | |
1421 | ||
f21d96d0 | 1422 | return bdrv_getlength(blk_bs(blk)); |
4be74634 MA |
1423 | } |
1424 | ||
1425 | void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr) | |
1426 | { | |
f21d96d0 | 1427 | if (!blk_bs(blk)) { |
a46fc9c9 HR |
1428 | *nb_sectors_ptr = 0; |
1429 | } else { | |
f21d96d0 | 1430 | bdrv_get_geometry(blk_bs(blk), nb_sectors_ptr); |
a46fc9c9 | 1431 | } |
4be74634 MA |
1432 | } |
1433 | ||
1ef01253 HR |
1434 | int64_t blk_nb_sectors(BlockBackend *blk) |
1435 | { | |
c09ba36c HR |
1436 | if (!blk_is_available(blk)) { |
1437 | return -ENOMEDIUM; | |
1438 | } | |
1439 | ||
f21d96d0 | 1440 | return bdrv_nb_sectors(blk_bs(blk)); |
1ef01253 HR |
1441 | } |
1442 | ||
60cb2fa7 EB |
1443 | BlockAIOCB *blk_aio_preadv(BlockBackend *blk, int64_t offset, |
1444 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1445 | BlockCompletionFunc *cb, void *opaque) | |
1446 | { | |
1447 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1448 | blk_aio_read_entry, flags, cb, opaque); | |
1449 | } | |
1450 | ||
60cb2fa7 EB |
1451 | BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
1452 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1453 | BlockCompletionFunc *cb, void *opaque) | |
1454 | { | |
1455 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1456 | blk_aio_write_entry, flags, cb, opaque); | |
1457 | } | |
1458 | ||
be07a889 KW |
1459 | static void blk_aio_flush_entry(void *opaque) |
1460 | { | |
1461 | BlkAioEmAIOCB *acb = opaque; | |
1462 | BlkRwCo *rwco = &acb->rwco; | |
1463 | ||
1464 | rwco->ret = blk_co_flush(rwco->blk); | |
1465 | blk_aio_complete(acb); | |
1466 | } | |
1467 | ||
4be74634 MA |
1468 | BlockAIOCB *blk_aio_flush(BlockBackend *blk, |
1469 | BlockCompletionFunc *cb, void *opaque) | |
1470 | { | |
be07a889 | 1471 | return blk_aio_prwv(blk, 0, 0, NULL, blk_aio_flush_entry, 0, cb, opaque); |
4be74634 MA |
1472 | } |
1473 | ||
8c2e3dd5 KW |
1474 | static void blk_aio_pdiscard_entry(void *opaque) |
1475 | { | |
1476 | BlkAioEmAIOCB *acb = opaque; | |
1477 | BlkRwCo *rwco = &acb->rwco; | |
1478 | ||
1479 | rwco->ret = blk_co_pdiscard(rwco->blk, rwco->offset, acb->bytes); | |
1480 | blk_aio_complete(acb); | |
1481 | } | |
1482 | ||
1c6c4bb7 | 1483 | BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, |
f5a5ca79 | 1484 | int64_t offset, int bytes, |
1c6c4bb7 | 1485 | BlockCompletionFunc *cb, void *opaque) |
4be74634 | 1486 | { |
f5a5ca79 | 1487 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_pdiscard_entry, 0, |
8c2e3dd5 | 1488 | cb, opaque); |
4be74634 MA |
1489 | } |
1490 | ||
1491 | void blk_aio_cancel(BlockAIOCB *acb) | |
1492 | { | |
1493 | bdrv_aio_cancel(acb); | |
1494 | } | |
1495 | ||
1496 | void blk_aio_cancel_async(BlockAIOCB *acb) | |
1497 | { | |
1498 | bdrv_aio_cancel_async(acb); | |
1499 | } | |
1500 | ||
48af776a | 1501 | int blk_co_ioctl(BlockBackend *blk, unsigned long int req, void *buf) |
4be74634 | 1502 | { |
c09ba36c HR |
1503 | if (!blk_is_available(blk)) { |
1504 | return -ENOMEDIUM; | |
1505 | } | |
1506 | ||
48af776a KW |
1507 | return bdrv_co_ioctl(blk_bs(blk), req, buf); |
1508 | } | |
1509 | ||
1510 | static void blk_ioctl_entry(void *opaque) | |
1511 | { | |
1512 | BlkRwCo *rwco = opaque; | |
c060332c DS |
1513 | QEMUIOVector *qiov = rwco->iobuf; |
1514 | ||
48af776a | 1515 | rwco->ret = blk_co_ioctl(rwco->blk, rwco->offset, |
c060332c | 1516 | qiov->iov[0].iov_base); |
4720cbee | 1517 | aio_wait_kick(); |
48af776a KW |
1518 | } |
1519 | ||
1520 | int blk_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
1521 | { | |
1522 | return blk_prw(blk, req, buf, 0, blk_ioctl_entry, 0); | |
1523 | } | |
1524 | ||
1525 | static void blk_aio_ioctl_entry(void *opaque) | |
1526 | { | |
1527 | BlkAioEmAIOCB *acb = opaque; | |
1528 | BlkRwCo *rwco = &acb->rwco; | |
1529 | ||
c060332c DS |
1530 | rwco->ret = blk_co_ioctl(rwco->blk, rwco->offset, rwco->iobuf); |
1531 | ||
48af776a | 1532 | blk_aio_complete(acb); |
4be74634 MA |
1533 | } |
1534 | ||
1535 | BlockAIOCB *blk_aio_ioctl(BlockBackend *blk, unsigned long int req, void *buf, | |
1536 | BlockCompletionFunc *cb, void *opaque) | |
1537 | { | |
c060332c | 1538 | return blk_aio_prwv(blk, req, 0, buf, blk_aio_ioctl_entry, 0, cb, opaque); |
4be74634 MA |
1539 | } |
1540 | ||
f5a5ca79 | 1541 | int blk_co_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
2bb0dce7 | 1542 | { |
f5a5ca79 | 1543 | int ret = blk_check_byte_request(blk, offset, bytes); |
e7f7d676 HR |
1544 | if (ret < 0) { |
1545 | return ret; | |
1546 | } | |
1547 | ||
0b9fd3f4 | 1548 | return bdrv_co_pdiscard(blk->root, offset, bytes); |
2bb0dce7 HR |
1549 | } |
1550 | ||
1551 | int blk_co_flush(BlockBackend *blk) | |
1552 | { | |
c09ba36c HR |
1553 | if (!blk_is_available(blk)) { |
1554 | return -ENOMEDIUM; | |
1555 | } | |
1556 | ||
f21d96d0 | 1557 | return bdrv_co_flush(blk_bs(blk)); |
2bb0dce7 HR |
1558 | } |
1559 | ||
be07a889 | 1560 | static void blk_flush_entry(void *opaque) |
4be74634 | 1561 | { |
be07a889 KW |
1562 | BlkRwCo *rwco = opaque; |
1563 | rwco->ret = blk_co_flush(rwco->blk); | |
4720cbee | 1564 | aio_wait_kick(); |
be07a889 | 1565 | } |
c09ba36c | 1566 | |
be07a889 KW |
1567 | int blk_flush(BlockBackend *blk) |
1568 | { | |
1569 | return blk_prw(blk, 0, NULL, 0, blk_flush_entry, 0); | |
4be74634 MA |
1570 | } |
1571 | ||
97b0385a AY |
1572 | void blk_drain(BlockBackend *blk) |
1573 | { | |
33f2a757 SH |
1574 | BlockDriverState *bs = blk_bs(blk); |
1575 | ||
1576 | if (bs) { | |
1577 | bdrv_drained_begin(bs); | |
1578 | } | |
1579 | ||
1580 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 KW |
1581 | AIO_WAIT_WHILE(blk_get_aio_context(blk), |
1582 | atomic_mb_read(&blk->in_flight) > 0); | |
33f2a757 SH |
1583 | |
1584 | if (bs) { | |
1585 | bdrv_drained_end(bs); | |
a46fc9c9 | 1586 | } |
97b0385a AY |
1587 | } |
1588 | ||
4be74634 MA |
1589 | void blk_drain_all(void) |
1590 | { | |
33f2a757 SH |
1591 | BlockBackend *blk = NULL; |
1592 | ||
1593 | bdrv_drain_all_begin(); | |
1594 | ||
1595 | while ((blk = blk_all_next(blk)) != NULL) { | |
1596 | AioContext *ctx = blk_get_aio_context(blk); | |
1597 | ||
1598 | aio_context_acquire(ctx); | |
1599 | ||
1600 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 | 1601 | AIO_WAIT_WHILE(ctx, atomic_mb_read(&blk->in_flight) > 0); |
33f2a757 SH |
1602 | |
1603 | aio_context_release(ctx); | |
1604 | } | |
1605 | ||
1606 | bdrv_drain_all_end(); | |
4be74634 MA |
1607 | } |
1608 | ||
373340b2 HR |
1609 | void blk_set_on_error(BlockBackend *blk, BlockdevOnError on_read_error, |
1610 | BlockdevOnError on_write_error) | |
1611 | { | |
1612 | blk->on_read_error = on_read_error; | |
1613 | blk->on_write_error = on_write_error; | |
1614 | } | |
1615 | ||
4be74634 MA |
1616 | BlockdevOnError blk_get_on_error(BlockBackend *blk, bool is_read) |
1617 | { | |
373340b2 | 1618 | return is_read ? blk->on_read_error : blk->on_write_error; |
4be74634 MA |
1619 | } |
1620 | ||
1621 | BlockErrorAction blk_get_error_action(BlockBackend *blk, bool is_read, | |
1622 | int error) | |
1623 | { | |
373340b2 HR |
1624 | BlockdevOnError on_err = blk_get_on_error(blk, is_read); |
1625 | ||
1626 | switch (on_err) { | |
1627 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
1628 | return (error == ENOSPC) ? | |
1629 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
1630 | case BLOCKDEV_ON_ERROR_STOP: | |
1631 | return BLOCK_ERROR_ACTION_STOP; | |
1632 | case BLOCKDEV_ON_ERROR_REPORT: | |
1633 | return BLOCK_ERROR_ACTION_REPORT; | |
1634 | case BLOCKDEV_ON_ERROR_IGNORE: | |
1635 | return BLOCK_ERROR_ACTION_IGNORE; | |
8c398252 | 1636 | case BLOCKDEV_ON_ERROR_AUTO: |
373340b2 HR |
1637 | default: |
1638 | abort(); | |
1639 | } | |
4be74634 MA |
1640 | } |
1641 | ||
373340b2 HR |
1642 | static void send_qmp_error_event(BlockBackend *blk, |
1643 | BlockErrorAction action, | |
1644 | bool is_read, int error) | |
1645 | { | |
1646 | IoOperationType optype; | |
bfe1a14c | 1647 | BlockDriverState *bs = blk_bs(blk); |
373340b2 HR |
1648 | |
1649 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; | |
bfe1a14c KW |
1650 | qapi_event_send_block_io_error(blk_name(blk), !!bs, |
1651 | bs ? bdrv_get_node_name(bs) : NULL, optype, | |
2bf7e10f | 1652 | action, blk_iostatus_is_enabled(blk), |
3ab72385 | 1653 | error == ENOSPC, strerror(error)); |
373340b2 HR |
1654 | } |
1655 | ||
1656 | /* This is done by device models because, while the block layer knows | |
1657 | * about the error, it does not know whether an operation comes from | |
1658 | * the device or the block layer (from a job, for example). | |
1659 | */ | |
4be74634 MA |
1660 | void blk_error_action(BlockBackend *blk, BlockErrorAction action, |
1661 | bool is_read, int error) | |
1662 | { | |
373340b2 HR |
1663 | assert(error >= 0); |
1664 | ||
1665 | if (action == BLOCK_ERROR_ACTION_STOP) { | |
1666 | /* First set the iostatus, so that "info block" returns an iostatus | |
1667 | * that matches the events raised so far (an additional error iostatus | |
1668 | * is fine, but not a lost one). | |
1669 | */ | |
1670 | blk_iostatus_set_err(blk, error); | |
1671 | ||
1672 | /* Then raise the request to stop the VM and the event. | |
1673 | * qemu_system_vmstop_request_prepare has two effects. First, | |
1674 | * it ensures that the STOP event always comes after the | |
1675 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
1676 | * can observe the STOP event and do a "cont" before the STOP | |
1677 | * event is issued, the VM will not stop. In this case, vm_start() | |
1678 | * also ensures that the STOP/RESUME pair of events is emitted. | |
1679 | */ | |
1680 | qemu_system_vmstop_request_prepare(); | |
1681 | send_qmp_error_event(blk, action, is_read, error); | |
1682 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); | |
1683 | } else { | |
1684 | send_qmp_error_event(blk, action, is_read, error); | |
1685 | } | |
4be74634 MA |
1686 | } |
1687 | ||
96710565 | 1688 | bool blk_is_read_only(BlockBackend *blk) |
4be74634 | 1689 | { |
f21d96d0 KW |
1690 | BlockDriverState *bs = blk_bs(blk); |
1691 | ||
1692 | if (bs) { | |
1693 | return bdrv_is_read_only(bs); | |
061959e8 HR |
1694 | } else { |
1695 | return blk->root_state.read_only; | |
1696 | } | |
4be74634 MA |
1697 | } |
1698 | ||
96710565 | 1699 | bool blk_is_sg(BlockBackend *blk) |
4be74634 | 1700 | { |
f21d96d0 KW |
1701 | BlockDriverState *bs = blk_bs(blk); |
1702 | ||
1703 | if (!bs) { | |
96710565 | 1704 | return false; |
a46fc9c9 HR |
1705 | } |
1706 | ||
f21d96d0 | 1707 | return bdrv_is_sg(bs); |
4be74634 MA |
1708 | } |
1709 | ||
96710565 | 1710 | bool blk_enable_write_cache(BlockBackend *blk) |
4be74634 | 1711 | { |
bfd18d1e | 1712 | return blk->enable_write_cache; |
4be74634 MA |
1713 | } |
1714 | ||
1715 | void blk_set_enable_write_cache(BlockBackend *blk, bool wce) | |
1716 | { | |
bfd18d1e | 1717 | blk->enable_write_cache = wce; |
4be74634 MA |
1718 | } |
1719 | ||
2bb0dce7 HR |
1720 | void blk_invalidate_cache(BlockBackend *blk, Error **errp) |
1721 | { | |
f21d96d0 KW |
1722 | BlockDriverState *bs = blk_bs(blk); |
1723 | ||
1724 | if (!bs) { | |
c09ba36c HR |
1725 | error_setg(errp, "Device '%s' has no medium", blk->name); |
1726 | return; | |
1727 | } | |
1728 | ||
f21d96d0 | 1729 | bdrv_invalidate_cache(bs, errp); |
2bb0dce7 HR |
1730 | } |
1731 | ||
e031f750 | 1732 | bool blk_is_inserted(BlockBackend *blk) |
4be74634 | 1733 | { |
f21d96d0 KW |
1734 | BlockDriverState *bs = blk_bs(blk); |
1735 | ||
1736 | return bs && bdrv_is_inserted(bs); | |
db0284f8 HR |
1737 | } |
1738 | ||
1739 | bool blk_is_available(BlockBackend *blk) | |
1740 | { | |
1741 | return blk_is_inserted(blk) && !blk_dev_is_tray_open(blk); | |
4be74634 MA |
1742 | } |
1743 | ||
1744 | void blk_lock_medium(BlockBackend *blk, bool locked) | |
1745 | { | |
f21d96d0 KW |
1746 | BlockDriverState *bs = blk_bs(blk); |
1747 | ||
1748 | if (bs) { | |
1749 | bdrv_lock_medium(bs, locked); | |
a46fc9c9 | 1750 | } |
4be74634 MA |
1751 | } |
1752 | ||
1753 | void blk_eject(BlockBackend *blk, bool eject_flag) | |
1754 | { | |
f21d96d0 | 1755 | BlockDriverState *bs = blk_bs(blk); |
2d76e724 KW |
1756 | char *id; |
1757 | ||
f21d96d0 KW |
1758 | if (bs) { |
1759 | bdrv_eject(bs, eject_flag); | |
a46fc9c9 | 1760 | } |
c47ee043 JS |
1761 | |
1762 | /* Whether or not we ejected on the backend, | |
1763 | * the frontend experienced a tray event. */ | |
1764 | id = blk_get_attached_dev_id(blk); | |
1765 | qapi_event_send_device_tray_moved(blk_name(blk), id, | |
3ab72385 | 1766 | eject_flag); |
c47ee043 | 1767 | g_free(id); |
4be74634 MA |
1768 | } |
1769 | ||
1770 | int blk_get_flags(BlockBackend *blk) | |
1771 | { | |
f21d96d0 KW |
1772 | BlockDriverState *bs = blk_bs(blk); |
1773 | ||
1774 | if (bs) { | |
1775 | return bdrv_get_flags(bs); | |
061959e8 HR |
1776 | } else { |
1777 | return blk->root_state.open_flags; | |
1778 | } | |
4be74634 MA |
1779 | } |
1780 | ||
4841211e EB |
1781 | /* Returns the minimum request alignment, in bytes; guaranteed nonzero */ |
1782 | uint32_t blk_get_request_alignment(BlockBackend *blk) | |
1783 | { | |
1784 | BlockDriverState *bs = blk_bs(blk); | |
1785 | return bs ? bs->bl.request_alignment : BDRV_SECTOR_SIZE; | |
1786 | } | |
1787 | ||
5def6b80 EB |
1788 | /* Returns the maximum transfer length, in bytes; guaranteed nonzero */ |
1789 | uint32_t blk_get_max_transfer(BlockBackend *blk) | |
454057b7 | 1790 | { |
f21d96d0 | 1791 | BlockDriverState *bs = blk_bs(blk); |
5def6b80 | 1792 | uint32_t max = 0; |
f21d96d0 KW |
1793 | |
1794 | if (bs) { | |
5def6b80 | 1795 | max = bs->bl.max_transfer; |
a46fc9c9 | 1796 | } |
5def6b80 | 1797 | return MIN_NON_ZERO(max, INT_MAX); |
454057b7 PL |
1798 | } |
1799 | ||
648296e0 SH |
1800 | int blk_get_max_iov(BlockBackend *blk) |
1801 | { | |
f21d96d0 | 1802 | return blk->root->bs->bl.max_iov; |
648296e0 SH |
1803 | } |
1804 | ||
4be74634 MA |
1805 | void blk_set_guest_block_size(BlockBackend *blk, int align) |
1806 | { | |
68e9ec01 | 1807 | blk->guest_block_size = align; |
4be74634 MA |
1808 | } |
1809 | ||
f1c17521 PB |
1810 | void *blk_try_blockalign(BlockBackend *blk, size_t size) |
1811 | { | |
f21d96d0 | 1812 | return qemu_try_blockalign(blk ? blk_bs(blk) : NULL, size); |
f1c17521 PB |
1813 | } |
1814 | ||
4be74634 MA |
1815 | void *blk_blockalign(BlockBackend *blk, size_t size) |
1816 | { | |
f21d96d0 | 1817 | return qemu_blockalign(blk ? blk_bs(blk) : NULL, size); |
4be74634 MA |
1818 | } |
1819 | ||
1820 | bool blk_op_is_blocked(BlockBackend *blk, BlockOpType op, Error **errp) | |
1821 | { | |
f21d96d0 KW |
1822 | BlockDriverState *bs = blk_bs(blk); |
1823 | ||
1824 | if (!bs) { | |
a46fc9c9 HR |
1825 | return false; |
1826 | } | |
1827 | ||
f21d96d0 | 1828 | return bdrv_op_is_blocked(bs, op, errp); |
4be74634 MA |
1829 | } |
1830 | ||
1831 | void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason) | |
1832 | { | |
f21d96d0 KW |
1833 | BlockDriverState *bs = blk_bs(blk); |
1834 | ||
1835 | if (bs) { | |
1836 | bdrv_op_unblock(bs, op, reason); | |
a46fc9c9 | 1837 | } |
4be74634 MA |
1838 | } |
1839 | ||
1840 | void blk_op_block_all(BlockBackend *blk, Error *reason) | |
1841 | { | |
f21d96d0 KW |
1842 | BlockDriverState *bs = blk_bs(blk); |
1843 | ||
1844 | if (bs) { | |
1845 | bdrv_op_block_all(bs, reason); | |
a46fc9c9 | 1846 | } |
4be74634 MA |
1847 | } |
1848 | ||
1849 | void blk_op_unblock_all(BlockBackend *blk, Error *reason) | |
1850 | { | |
f21d96d0 KW |
1851 | BlockDriverState *bs = blk_bs(blk); |
1852 | ||
1853 | if (bs) { | |
1854 | bdrv_op_unblock_all(bs, reason); | |
a46fc9c9 | 1855 | } |
4be74634 MA |
1856 | } |
1857 | ||
1858 | AioContext *blk_get_aio_context(BlockBackend *blk) | |
1859 | { | |
7803696d | 1860 | return bdrv_get_aio_context(blk_bs(blk)); |
4981bdec HR |
1861 | } |
1862 | ||
1863 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) | |
1864 | { | |
1865 | BlockBackendAIOCB *blk_acb = DO_UPCAST(BlockBackendAIOCB, common, acb); | |
1866 | return blk_get_aio_context(blk_acb->blk); | |
4be74634 MA |
1867 | } |
1868 | ||
38475269 KW |
1869 | static void blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context, |
1870 | bool update_root_node) | |
4be74634 | 1871 | { |
f21d96d0 | 1872 | BlockDriverState *bs = blk_bs(blk); |
c61791fc | 1873 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
f21d96d0 KW |
1874 | |
1875 | if (bs) { | |
c61791fc | 1876 | if (tgm->throttle_state) { |
dc868fb0 | 1877 | bdrv_drained_begin(bs); |
c61791fc MP |
1878 | throttle_group_detach_aio_context(tgm); |
1879 | throttle_group_attach_aio_context(tgm, new_context); | |
dc868fb0 | 1880 | bdrv_drained_end(bs); |
7ca7f0f6 | 1881 | } |
38475269 KW |
1882 | if (update_root_node) { |
1883 | GSList *ignore = g_slist_prepend(NULL, blk->root); | |
1884 | bdrv_set_aio_context_ignore(bs, new_context, &ignore); | |
1885 | g_slist_free(ignore); | |
1886 | } | |
a46fc9c9 | 1887 | } |
4be74634 MA |
1888 | } |
1889 | ||
38475269 KW |
1890 | void blk_set_aio_context(BlockBackend *blk, AioContext *new_context) |
1891 | { | |
1892 | blk_do_set_aio_context(blk, new_context, true); | |
1893 | } | |
1894 | ||
1895 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
1896 | GSList **ignore, Error **errp) | |
1897 | { | |
1898 | BlockBackend *blk = child->opaque; | |
1899 | ||
980b0f94 KW |
1900 | if (blk->allow_aio_context_change) { |
1901 | return true; | |
1902 | } | |
1903 | ||
38475269 KW |
1904 | /* Only manually created BlockBackends that are not attached to anything |
1905 | * can change their AioContext without updating their user. */ | |
1906 | if (!blk->name || blk->dev) { | |
1907 | /* TODO Add BB name/QOM path */ | |
1908 | error_setg(errp, "Cannot change iothread of active block backend"); | |
1909 | return false; | |
1910 | } | |
1911 | ||
1912 | return true; | |
1913 | } | |
1914 | ||
1915 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
1916 | GSList **ignore) | |
1917 | { | |
1918 | BlockBackend *blk = child->opaque; | |
1919 | blk_do_set_aio_context(blk, ctx, false); | |
1920 | } | |
1921 | ||
2019ba0a HR |
1922 | void blk_add_aio_context_notifier(BlockBackend *blk, |
1923 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
1924 | void (*detach_aio_context)(void *opaque), void *opaque) | |
1925 | { | |
d03654ea | 1926 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
1927 | BlockDriverState *bs = blk_bs(blk); |
1928 | ||
d03654ea SH |
1929 | notifier = g_new(BlockBackendAioNotifier, 1); |
1930 | notifier->attached_aio_context = attached_aio_context; | |
1931 | notifier->detach_aio_context = detach_aio_context; | |
1932 | notifier->opaque = opaque; | |
1933 | QLIST_INSERT_HEAD(&blk->aio_notifiers, notifier, list); | |
1934 | ||
f21d96d0 KW |
1935 | if (bs) { |
1936 | bdrv_add_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
1937 | detach_aio_context, opaque); |
1938 | } | |
2019ba0a HR |
1939 | } |
1940 | ||
1941 | void blk_remove_aio_context_notifier(BlockBackend *blk, | |
1942 | void (*attached_aio_context)(AioContext *, | |
1943 | void *), | |
1944 | void (*detach_aio_context)(void *), | |
1945 | void *opaque) | |
1946 | { | |
d03654ea | 1947 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
1948 | BlockDriverState *bs = blk_bs(blk); |
1949 | ||
1950 | if (bs) { | |
1951 | bdrv_remove_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
1952 | detach_aio_context, opaque); |
1953 | } | |
d03654ea SH |
1954 | |
1955 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
1956 | if (notifier->attached_aio_context == attached_aio_context && | |
1957 | notifier->detach_aio_context == detach_aio_context && | |
1958 | notifier->opaque == opaque) { | |
1959 | QLIST_REMOVE(notifier, list); | |
1960 | g_free(notifier); | |
1961 | return; | |
1962 | } | |
1963 | } | |
1964 | ||
1965 | abort(); | |
2019ba0a HR |
1966 | } |
1967 | ||
3301f6c6 HR |
1968 | void blk_add_remove_bs_notifier(BlockBackend *blk, Notifier *notify) |
1969 | { | |
1970 | notifier_list_add(&blk->remove_bs_notifiers, notify); | |
1971 | } | |
1972 | ||
1973 | void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify) | |
1974 | { | |
1975 | notifier_list_add(&blk->insert_bs_notifiers, notify); | |
1976 | } | |
1977 | ||
4be74634 MA |
1978 | void blk_io_plug(BlockBackend *blk) |
1979 | { | |
f21d96d0 KW |
1980 | BlockDriverState *bs = blk_bs(blk); |
1981 | ||
1982 | if (bs) { | |
1983 | bdrv_io_plug(bs); | |
a46fc9c9 | 1984 | } |
4be74634 MA |
1985 | } |
1986 | ||
1987 | void blk_io_unplug(BlockBackend *blk) | |
1988 | { | |
f21d96d0 KW |
1989 | BlockDriverState *bs = blk_bs(blk); |
1990 | ||
1991 | if (bs) { | |
1992 | bdrv_io_unplug(bs); | |
a46fc9c9 | 1993 | } |
4be74634 MA |
1994 | } |
1995 | ||
1996 | BlockAcctStats *blk_get_stats(BlockBackend *blk) | |
1997 | { | |
7f0e9da6 | 1998 | return &blk->stats; |
4be74634 MA |
1999 | } |
2000 | ||
2001 | void *blk_aio_get(const AIOCBInfo *aiocb_info, BlockBackend *blk, | |
2002 | BlockCompletionFunc *cb, void *opaque) | |
2003 | { | |
2004 | return qemu_aio_get(aiocb_info, blk_bs(blk), cb, opaque); | |
2005 | } | |
1ef01253 | 2006 | |
d004bd52 | 2007 | int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 2008 | int bytes, BdrvRequestFlags flags) |
1ef01253 | 2009 | { |
f5a5ca79 | 2010 | return blk_co_pwritev(blk, offset, bytes, NULL, |
16aaf975 | 2011 | flags | BDRV_REQ_ZERO_WRITE); |
1ef01253 HR |
2012 | } |
2013 | ||
fe5c1355 PB |
2014 | int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, |
2015 | int count) | |
1ef01253 | 2016 | { |
35fadca8 PB |
2017 | return blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
2018 | BDRV_REQ_WRITE_COMPRESSED); | |
1ef01253 HR |
2019 | } |
2020 | ||
3a691c50 HR |
2021 | int blk_truncate(BlockBackend *blk, int64_t offset, PreallocMode prealloc, |
2022 | Error **errp) | |
1ef01253 | 2023 | { |
c09ba36c | 2024 | if (!blk_is_available(blk)) { |
ed3d2ec9 | 2025 | error_setg(errp, "No medium inserted"); |
c09ba36c HR |
2026 | return -ENOMEDIUM; |
2027 | } | |
2028 | ||
3a691c50 | 2029 | return bdrv_truncate(blk->root, offset, prealloc, errp); |
1ef01253 HR |
2030 | } |
2031 | ||
8c2e3dd5 | 2032 | static void blk_pdiscard_entry(void *opaque) |
1ef01253 | 2033 | { |
8c2e3dd5 | 2034 | BlkRwCo *rwco = opaque; |
c060332c DS |
2035 | QEMUIOVector *qiov = rwco->iobuf; |
2036 | ||
2037 | rwco->ret = blk_co_pdiscard(rwco->blk, rwco->offset, qiov->size); | |
4720cbee | 2038 | aio_wait_kick(); |
8c2e3dd5 | 2039 | } |
e7f7d676 | 2040 | |
f5a5ca79 | 2041 | int blk_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
8c2e3dd5 | 2042 | { |
f5a5ca79 | 2043 | return blk_prw(blk, offset, NULL, bytes, blk_pdiscard_entry, 0); |
1ef01253 HR |
2044 | } |
2045 | ||
2046 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, | |
2047 | int64_t pos, int size) | |
2048 | { | |
bfd18d1e KW |
2049 | int ret; |
2050 | ||
c09ba36c HR |
2051 | if (!blk_is_available(blk)) { |
2052 | return -ENOMEDIUM; | |
2053 | } | |
2054 | ||
bfd18d1e KW |
2055 | ret = bdrv_save_vmstate(blk_bs(blk), buf, pos, size); |
2056 | if (ret < 0) { | |
2057 | return ret; | |
2058 | } | |
2059 | ||
2060 | if (ret == size && !blk->enable_write_cache) { | |
2061 | ret = bdrv_flush(blk_bs(blk)); | |
2062 | } | |
2063 | ||
2064 | return ret < 0 ? ret : size; | |
1ef01253 HR |
2065 | } |
2066 | ||
2067 | int blk_load_vmstate(BlockBackend *blk, uint8_t *buf, int64_t pos, int size) | |
2068 | { | |
c09ba36c HR |
2069 | if (!blk_is_available(blk)) { |
2070 | return -ENOMEDIUM; | |
2071 | } | |
2072 | ||
f21d96d0 | 2073 | return bdrv_load_vmstate(blk_bs(blk), buf, pos, size); |
1ef01253 | 2074 | } |
f0272c4d ET |
2075 | |
2076 | int blk_probe_blocksizes(BlockBackend *blk, BlockSizes *bsz) | |
2077 | { | |
c09ba36c HR |
2078 | if (!blk_is_available(blk)) { |
2079 | return -ENOMEDIUM; | |
2080 | } | |
2081 | ||
f21d96d0 | 2082 | return bdrv_probe_blocksizes(blk_bs(blk), bsz); |
f0272c4d ET |
2083 | } |
2084 | ||
2085 | int blk_probe_geometry(BlockBackend *blk, HDGeometry *geo) | |
2086 | { | |
c09ba36c HR |
2087 | if (!blk_is_available(blk)) { |
2088 | return -ENOMEDIUM; | |
2089 | } | |
2090 | ||
f21d96d0 | 2091 | return bdrv_probe_geometry(blk_bs(blk), geo); |
f0272c4d | 2092 | } |
281d22d8 HR |
2093 | |
2094 | /* | |
2095 | * Updates the BlockBackendRootState object with data from the currently | |
2096 | * attached BlockDriverState. | |
2097 | */ | |
2098 | void blk_update_root_state(BlockBackend *blk) | |
2099 | { | |
f21d96d0 | 2100 | assert(blk->root); |
281d22d8 | 2101 | |
f21d96d0 KW |
2102 | blk->root_state.open_flags = blk->root->bs->open_flags; |
2103 | blk->root_state.read_only = blk->root->bs->read_only; | |
2104 | blk->root_state.detect_zeroes = blk->root->bs->detect_zeroes; | |
281d22d8 HR |
2105 | } |
2106 | ||
38cb18f5 | 2107 | /* |
b85114f8 KW |
2108 | * Returns the detect-zeroes setting to be used for bdrv_open() of a |
2109 | * BlockDriverState which is supposed to inherit the root state. | |
38cb18f5 | 2110 | */ |
b85114f8 | 2111 | bool blk_get_detect_zeroes_from_root_state(BlockBackend *blk) |
38cb18f5 | 2112 | { |
b85114f8 | 2113 | return blk->root_state.detect_zeroes; |
38cb18f5 HR |
2114 | } |
2115 | ||
2116 | /* | |
2117 | * Returns the flags to be used for bdrv_open() of a BlockDriverState which is | |
2118 | * supposed to inherit the root state. | |
2119 | */ | |
2120 | int blk_get_open_flags_from_root_state(BlockBackend *blk) | |
2121 | { | |
2122 | int bs_flags; | |
2123 | ||
2124 | bs_flags = blk->root_state.read_only ? 0 : BDRV_O_RDWR; | |
2125 | bs_flags |= blk->root_state.open_flags & ~BDRV_O_RDWR; | |
2126 | ||
2127 | return bs_flags; | |
2128 | } | |
2129 | ||
281d22d8 HR |
2130 | BlockBackendRootState *blk_get_root_state(BlockBackend *blk) |
2131 | { | |
2132 | return &blk->root_state; | |
2133 | } | |
1393f212 HR |
2134 | |
2135 | int blk_commit_all(void) | |
2136 | { | |
fe1a9cbc HR |
2137 | BlockBackend *blk = NULL; |
2138 | ||
2139 | while ((blk = blk_all_next(blk)) != NULL) { | |
2140 | AioContext *aio_context = blk_get_aio_context(blk); | |
2141 | ||
2142 | aio_context_acquire(aio_context); | |
f21d96d0 KW |
2143 | if (blk_is_inserted(blk) && blk->root->bs->backing) { |
2144 | int ret = bdrv_commit(blk->root->bs); | |
fe1a9cbc HR |
2145 | if (ret < 0) { |
2146 | aio_context_release(aio_context); | |
2147 | return ret; | |
2148 | } | |
2149 | } | |
2150 | aio_context_release(aio_context); | |
2151 | } | |
2152 | return 0; | |
2153 | } | |
2154 | ||
97148076 KW |
2155 | |
2156 | /* throttling disk I/O limits */ | |
2157 | void blk_set_io_limits(BlockBackend *blk, ThrottleConfig *cfg) | |
2158 | { | |
022cdc9f | 2159 | throttle_group_config(&blk->public.throttle_group_member, cfg); |
97148076 KW |
2160 | } |
2161 | ||
2162 | void blk_io_limits_disable(BlockBackend *blk) | |
2163 | { | |
48bf7ea8 AG |
2164 | BlockDriverState *bs = blk_bs(blk); |
2165 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; | |
2166 | assert(tgm->throttle_state); | |
2167 | if (bs) { | |
2168 | bdrv_drained_begin(bs); | |
2169 | } | |
2170 | throttle_group_unregister_tgm(tgm); | |
2171 | if (bs) { | |
2172 | bdrv_drained_end(bs); | |
2173 | } | |
97148076 KW |
2174 | } |
2175 | ||
2176 | /* should be called before blk_set_io_limits if a limit is set */ | |
2177 | void blk_io_limits_enable(BlockBackend *blk, const char *group) | |
2178 | { | |
022cdc9f | 2179 | assert(!blk->public.throttle_group_member.throttle_state); |
c61791fc MP |
2180 | throttle_group_register_tgm(&blk->public.throttle_group_member, |
2181 | group, blk_get_aio_context(blk)); | |
97148076 KW |
2182 | } |
2183 | ||
2184 | void blk_io_limits_update_group(BlockBackend *blk, const char *group) | |
2185 | { | |
2186 | /* this BB is not part of any group */ | |
022cdc9f | 2187 | if (!blk->public.throttle_group_member.throttle_state) { |
97148076 KW |
2188 | return; |
2189 | } | |
2190 | ||
2191 | /* this BB is a part of the same group than the one we want */ | |
022cdc9f MP |
2192 | if (!g_strcmp0(throttle_group_get_name(&blk->public.throttle_group_member), |
2193 | group)) { | |
97148076 KW |
2194 | return; |
2195 | } | |
2196 | ||
2197 | /* need to change the group this bs belong to */ | |
2198 | blk_io_limits_disable(blk); | |
2199 | blk_io_limits_enable(blk, group); | |
2200 | } | |
c2066af0 KW |
2201 | |
2202 | static void blk_root_drained_begin(BdrvChild *child) | |
2203 | { | |
2204 | BlockBackend *blk = child->opaque; | |
2205 | ||
f4d9cc88 JS |
2206 | if (++blk->quiesce_counter == 1) { |
2207 | if (blk->dev_ops && blk->dev_ops->drained_begin) { | |
2208 | blk->dev_ops->drained_begin(blk->dev_opaque); | |
2209 | } | |
2210 | } | |
2211 | ||
36fe1331 KW |
2212 | /* Note that blk->root may not be accessible here yet if we are just |
2213 | * attaching to a BlockDriverState that is drained. Use child instead. */ | |
2214 | ||
022cdc9f MP |
2215 | if (atomic_fetch_inc(&blk->public.throttle_group_member.io_limits_disabled) == 0) { |
2216 | throttle_group_restart_tgm(&blk->public.throttle_group_member); | |
c2066af0 KW |
2217 | } |
2218 | } | |
2219 | ||
fe5258a5 KW |
2220 | static bool blk_root_drained_poll(BdrvChild *child) |
2221 | { | |
2222 | BlockBackend *blk = child->opaque; | |
2223 | assert(blk->quiesce_counter); | |
2224 | return !!blk->in_flight; | |
2225 | } | |
2226 | ||
c2066af0 KW |
2227 | static void blk_root_drained_end(BdrvChild *child) |
2228 | { | |
2229 | BlockBackend *blk = child->opaque; | |
f4d9cc88 | 2230 | assert(blk->quiesce_counter); |
c2066af0 | 2231 | |
022cdc9f MP |
2232 | assert(blk->public.throttle_group_member.io_limits_disabled); |
2233 | atomic_dec(&blk->public.throttle_group_member.io_limits_disabled); | |
f4d9cc88 JS |
2234 | |
2235 | if (--blk->quiesce_counter == 0) { | |
2236 | if (blk->dev_ops && blk->dev_ops->drained_end) { | |
2237 | blk->dev_ops->drained_end(blk->dev_opaque); | |
2238 | } | |
2239 | } | |
c2066af0 | 2240 | } |
23d0ba93 FZ |
2241 | |
2242 | void blk_register_buf(BlockBackend *blk, void *host, size_t size) | |
2243 | { | |
2244 | bdrv_register_buf(blk_bs(blk), host, size); | |
2245 | } | |
2246 | ||
2247 | void blk_unregister_buf(BlockBackend *blk, void *host) | |
2248 | { | |
2249 | bdrv_unregister_buf(blk_bs(blk), host); | |
2250 | } | |
b5679fa4 FZ |
2251 | |
2252 | int coroutine_fn blk_co_copy_range(BlockBackend *blk_in, int64_t off_in, | |
2253 | BlockBackend *blk_out, int64_t off_out, | |
67b51fb9 VSO |
2254 | int bytes, BdrvRequestFlags read_flags, |
2255 | BdrvRequestFlags write_flags) | |
b5679fa4 FZ |
2256 | { |
2257 | int r; | |
2258 | r = blk_check_byte_request(blk_in, off_in, bytes); | |
2259 | if (r) { | |
2260 | return r; | |
2261 | } | |
2262 | r = blk_check_byte_request(blk_out, off_out, bytes); | |
2263 | if (r) { | |
2264 | return r; | |
2265 | } | |
2266 | return bdrv_co_copy_range(blk_in->root, off_in, | |
2267 | blk_out->root, off_out, | |
67b51fb9 | 2268 | bytes, read_flags, write_flags); |
b5679fa4 | 2269 | } |
5d3b4e99 VSO |
2270 | |
2271 | const BdrvChild *blk_root(BlockBackend *blk) | |
2272 | { | |
2273 | return blk->root; | |
2274 | } |