]>
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 | 21 | #include "sysemu/replay.h" |
e688df6b | 22 | #include "qapi/error.h" |
9af23989 | 23 | #include "qapi/qapi-events-block.h" |
f348b6d1 | 24 | #include "qemu/id.h" |
db725815 | 25 | #include "qemu/main-loop.h" |
922a01a0 | 26 | #include "qemu/option.h" |
1e98fefd | 27 | #include "trace.h" |
5f7772c4 | 28 | #include "migration/misc.h" |
a7f53e26 MA |
29 | |
30 | /* Number of coroutines to reserve per attached device model */ | |
31 | #define COROUTINE_POOL_RESERVATION 64 | |
26f54e9a | 32 | |
1bf1cbc9 KW |
33 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
34 | ||
4981bdec HR |
35 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb); |
36 | ||
d03654ea SH |
37 | typedef struct BlockBackendAioNotifier { |
38 | void (*attached_aio_context)(AioContext *new_context, void *opaque); | |
39 | void (*detach_aio_context)(void *opaque); | |
40 | void *opaque; | |
41 | QLIST_ENTRY(BlockBackendAioNotifier) list; | |
42 | } BlockBackendAioNotifier; | |
43 | ||
26f54e9a MA |
44 | struct BlockBackend { |
45 | char *name; | |
46 | int refcnt; | |
f21d96d0 | 47 | BdrvChild *root; |
d861ab3a | 48 | AioContext *ctx; |
26f8b3a8 | 49 | DriveInfo *legacy_dinfo; /* null unless created by drive_new() */ |
2cf22d6a | 50 | QTAILQ_ENTRY(BlockBackend) link; /* for block_backends */ |
9492b0b9 | 51 | QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */ |
f2cd875d | 52 | BlockBackendPublic public; |
a7f53e26 | 53 | |
d09ea2d2 | 54 | DeviceState *dev; /* attached device model, if any */ |
a7f53e26 MA |
55 | const BlockDevOps *dev_ops; |
56 | void *dev_opaque; | |
68e9ec01 HR |
57 | |
58 | /* the block size for which the guest device expects atomicity */ | |
59 | int guest_block_size; | |
7f0e9da6 | 60 | |
281d22d8 HR |
61 | /* If the BDS tree is removed, some of its options are stored here (which |
62 | * can be used to restore those options in the new BDS on insert) */ | |
63 | BlockBackendRootState root_state; | |
64 | ||
bfd18d1e KW |
65 | bool enable_write_cache; |
66 | ||
7f0e9da6 HR |
67 | /* I/O stats (display with "info blockstats"). */ |
68 | BlockAcctStats stats; | |
373340b2 HR |
69 | |
70 | BlockdevOnError on_read_error, on_write_error; | |
71 | bool iostatus_enabled; | |
72 | BlockDeviceIoStatus iostatus; | |
3301f6c6 | 73 | |
981776b3 KW |
74 | uint64_t perm; |
75 | uint64_t shared_perm; | |
d35ff5e6 | 76 | bool disable_perm; |
981776b3 | 77 | |
980b0f94 | 78 | bool allow_aio_context_change; |
c10c9d96 KW |
79 | bool allow_write_beyond_eof; |
80 | ||
3301f6c6 | 81 | NotifierList remove_bs_notifiers, insert_bs_notifiers; |
d03654ea | 82 | QLIST_HEAD(, BlockBackendAioNotifier) aio_notifiers; |
f4d9cc88 JS |
83 | |
84 | int quiesce_counter; | |
cf312932 KW |
85 | CoQueue queued_requests; |
86 | bool disable_request_queuing; | |
87 | ||
5f7772c4 | 88 | VMChangeStateEntry *vmsh; |
ca2e2144 | 89 | bool force_allow_inactivate; |
33f2a757 SH |
90 | |
91 | /* Number of in-flight aio requests. BlockDriverState also counts | |
92 | * in-flight requests but aio requests can exist even when blk->root is | |
93 | * NULL, so we cannot rely on its counter for that case. | |
94 | * Accessed with atomic ops. | |
95 | */ | |
96 | unsigned int in_flight; | |
26f54e9a MA |
97 | }; |
98 | ||
e7f7d676 HR |
99 | typedef struct BlockBackendAIOCB { |
100 | BlockAIOCB common; | |
4981bdec | 101 | BlockBackend *blk; |
e7f7d676 HR |
102 | int ret; |
103 | } BlockBackendAIOCB; | |
104 | ||
105 | static const AIOCBInfo block_backend_aiocb_info = { | |
4981bdec | 106 | .get_aio_context = blk_aiocb_get_aio_context, |
e7f7d676 HR |
107 | .aiocb_size = sizeof(BlockBackendAIOCB), |
108 | }; | |
109 | ||
8fb3c76c | 110 | static void drive_info_del(DriveInfo *dinfo); |
7c8eece4 | 111 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs); |
8fb3c76c | 112 | |
2cf22d6a HR |
113 | /* All BlockBackends */ |
114 | static QTAILQ_HEAD(, BlockBackend) block_backends = | |
115 | QTAILQ_HEAD_INITIALIZER(block_backends); | |
116 | ||
9492b0b9 HR |
117 | /* All BlockBackends referenced by the monitor and which are iterated through by |
118 | * blk_next() */ | |
119 | static QTAILQ_HEAD(, BlockBackend) monitor_block_backends = | |
120 | QTAILQ_HEAD_INITIALIZER(monitor_block_backends); | |
26f54e9a | 121 | |
3cdc69d3 | 122 | static void blk_root_inherit_options(BdrvChildRole role, bool parent_is_format, |
272c02ea | 123 | int *child_flags, QDict *child_options, |
f21d96d0 KW |
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; | |
fd240a18 | 144 | g_autofree char *dev_id = NULL; |
b5411555 KW |
145 | |
146 | if (blk->name) { | |
fd240a18 | 147 | return g_strdup_printf("block device '%s'", blk->name); |
b5411555 KW |
148 | } |
149 | ||
150 | dev_id = blk_get_attached_dev_id(blk); | |
151 | if (*dev_id) { | |
fd240a18 | 152 | return g_strdup_printf("block device '%s'", dev_id); |
b5411555 KW |
153 | } else { |
154 | /* TODO Callback into the BB owner for something more detailed */ | |
fd240a18 | 155 | return g_strdup("an unnamed block device"); |
b5411555 KW |
156 | } |
157 | } | |
158 | ||
4c265bf9 KW |
159 | static const char *blk_root_get_name(BdrvChild *child) |
160 | { | |
161 | return blk_name(child->opaque); | |
162 | } | |
163 | ||
538f0497 | 164 | static void blk_vm_state_changed(void *opaque, bool running, RunState state) |
5f7772c4 FZ |
165 | { |
166 | Error *local_err = NULL; | |
167 | BlockBackend *blk = opaque; | |
168 | ||
169 | if (state == RUN_STATE_INMIGRATE) { | |
170 | return; | |
171 | } | |
172 | ||
173 | qemu_del_vm_change_state_handler(blk->vmsh); | |
174 | blk->vmsh = NULL; | |
175 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); | |
176 | if (local_err) { | |
177 | error_report_err(local_err); | |
178 | } | |
179 | } | |
180 | ||
4417ab7a KW |
181 | /* |
182 | * Notifies the user of the BlockBackend that migration has completed. qdev | |
183 | * devices can tighten their permissions in response (specifically revoke | |
184 | * shared write permissions that we needed for storage migration). | |
185 | * | |
186 | * If an error is returned, the VM cannot be allowed to be resumed. | |
187 | */ | |
188 | static void blk_root_activate(BdrvChild *child, Error **errp) | |
189 | { | |
190 | BlockBackend *blk = child->opaque; | |
191 | Error *local_err = NULL; | |
192 | ||
193 | if (!blk->disable_perm) { | |
194 | return; | |
195 | } | |
196 | ||
197 | blk->disable_perm = false; | |
198 | ||
5f7772c4 FZ |
199 | blk_set_perm(blk, blk->perm, BLK_PERM_ALL, &local_err); |
200 | if (local_err) { | |
201 | error_propagate(errp, local_err); | |
202 | blk->disable_perm = true; | |
203 | return; | |
204 | } | |
205 | ||
206 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
207 | /* Activation can happen when migration process is still active, for | |
208 | * example when nbd_server_add is called during non-shared storage | |
209 | * migration. Defer the shared_perm update to migration completion. */ | |
210 | if (!blk->vmsh) { | |
211 | blk->vmsh = qemu_add_vm_change_state_handler(blk_vm_state_changed, | |
212 | blk); | |
213 | } | |
214 | return; | |
215 | } | |
216 | ||
4417ab7a KW |
217 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); |
218 | if (local_err) { | |
219 | error_propagate(errp, local_err); | |
220 | blk->disable_perm = true; | |
221 | return; | |
222 | } | |
223 | } | |
224 | ||
ca2e2144 FZ |
225 | void blk_set_force_allow_inactivate(BlockBackend *blk) |
226 | { | |
227 | blk->force_allow_inactivate = true; | |
228 | } | |
229 | ||
c16de8f5 FZ |
230 | static bool blk_can_inactivate(BlockBackend *blk) |
231 | { | |
ca2e2144 | 232 | /* If it is a guest device, inactivate is ok. */ |
c16de8f5 FZ |
233 | if (blk->dev || blk_name(blk)[0]) { |
234 | return true; | |
235 | } | |
236 | ||
ca2e2144 FZ |
237 | /* Inactivating means no more writes to the image can be done, |
238 | * even if those writes would be changes invisible to the | |
239 | * guest. For block job BBs that satisfy this, we can just allow | |
240 | * it. This is the case for mirror job source, which is required | |
241 | * by libvirt non-shared block migration. */ | |
242 | if (!(blk->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED))) { | |
243 | return true; | |
244 | } | |
245 | ||
246 | return blk->force_allow_inactivate; | |
c16de8f5 FZ |
247 | } |
248 | ||
cfa1a572 KW |
249 | static int blk_root_inactivate(BdrvChild *child) |
250 | { | |
251 | BlockBackend *blk = child->opaque; | |
252 | ||
253 | if (blk->disable_perm) { | |
254 | return 0; | |
255 | } | |
256 | ||
c16de8f5 | 257 | if (!blk_can_inactivate(blk)) { |
cfa1a572 KW |
258 | return -EPERM; |
259 | } | |
260 | ||
261 | blk->disable_perm = true; | |
262 | if (blk->root) { | |
263 | bdrv_child_try_set_perm(blk->root, 0, BLK_PERM_ALL, &error_abort); | |
264 | } | |
265 | ||
266 | return 0; | |
267 | } | |
268 | ||
d03654ea SH |
269 | static void blk_root_attach(BdrvChild *child) |
270 | { | |
271 | BlockBackend *blk = child->opaque; | |
272 | BlockBackendAioNotifier *notifier; | |
273 | ||
274 | trace_blk_root_attach(child, blk, child->bs); | |
275 | ||
276 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
277 | bdrv_add_aio_context_notifier(child->bs, | |
278 | notifier->attached_aio_context, | |
279 | notifier->detach_aio_context, | |
280 | notifier->opaque); | |
281 | } | |
282 | } | |
283 | ||
284 | static void blk_root_detach(BdrvChild *child) | |
285 | { | |
286 | BlockBackend *blk = child->opaque; | |
287 | BlockBackendAioNotifier *notifier; | |
288 | ||
289 | trace_blk_root_detach(child, blk, child->bs); | |
290 | ||
291 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
292 | bdrv_remove_aio_context_notifier(child->bs, | |
293 | notifier->attached_aio_context, | |
294 | notifier->detach_aio_context, | |
295 | notifier->opaque); | |
296 | } | |
297 | } | |
298 | ||
3ca1f322 VSO |
299 | static AioContext *blk_root_get_parent_aio_context(BdrvChild *c) |
300 | { | |
301 | BlockBackend *blk = c->opaque; | |
302 | ||
303 | return blk_get_aio_context(blk); | |
304 | } | |
305 | ||
bd86fb99 | 306 | static const BdrvChildClass child_root = { |
c2066af0 KW |
307 | .inherit_options = blk_root_inherit_options, |
308 | ||
5c8cab48 KW |
309 | .change_media = blk_root_change_media, |
310 | .resize = blk_root_resize, | |
4c265bf9 | 311 | .get_name = blk_root_get_name, |
b5411555 | 312 | .get_parent_desc = blk_root_get_parent_desc, |
5c8cab48 | 313 | |
c2066af0 | 314 | .drained_begin = blk_root_drained_begin, |
fe5258a5 | 315 | .drained_poll = blk_root_drained_poll, |
c2066af0 | 316 | .drained_end = blk_root_drained_end, |
4417ab7a KW |
317 | |
318 | .activate = blk_root_activate, | |
cfa1a572 | 319 | .inactivate = blk_root_inactivate, |
d03654ea SH |
320 | |
321 | .attach = blk_root_attach, | |
322 | .detach = blk_root_detach, | |
38475269 KW |
323 | |
324 | .can_set_aio_ctx = blk_root_can_set_aio_ctx, | |
325 | .set_aio_ctx = blk_root_set_aio_ctx, | |
3ca1f322 VSO |
326 | |
327 | .get_parent_aio_context = blk_root_get_parent_aio_context, | |
f21d96d0 KW |
328 | }; |
329 | ||
26f54e9a | 330 | /* |
efaa7c4e | 331 | * Create a new BlockBackend with a reference count of one. |
6d0eb64d KW |
332 | * |
333 | * @perm is a bitmasks of BLK_PERM_* constants which describes the permissions | |
334 | * to request for a block driver node that is attached to this BlockBackend. | |
335 | * @shared_perm is a bitmask which describes which permissions may be granted | |
336 | * to other users of the attached node. | |
337 | * Both sets of permissions can be changed later using blk_set_perm(). | |
338 | * | |
26f54e9a MA |
339 | * Return the new BlockBackend on success, null on failure. |
340 | */ | |
d861ab3a | 341 | BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm) |
26f54e9a MA |
342 | { |
343 | BlockBackend *blk; | |
344 | ||
26f54e9a | 345 | blk = g_new0(BlockBackend, 1); |
26f54e9a | 346 | blk->refcnt = 1; |
d861ab3a | 347 | blk->ctx = ctx; |
6d0eb64d KW |
348 | blk->perm = perm; |
349 | blk->shared_perm = shared_perm; | |
0c3169df KW |
350 | blk_set_enable_write_cache(blk, true); |
351 | ||
cb53460b KW |
352 | blk->on_read_error = BLOCKDEV_ON_ERROR_REPORT; |
353 | blk->on_write_error = BLOCKDEV_ON_ERROR_ENOSPC; | |
354 | ||
9caa6f3d | 355 | block_acct_init(&blk->stats); |
27ccdd52 | 356 | |
cf312932 | 357 | qemu_co_queue_init(&blk->queued_requests); |
3301f6c6 HR |
358 | notifier_list_init(&blk->remove_bs_notifiers); |
359 | notifier_list_init(&blk->insert_bs_notifiers); | |
d03654ea | 360 | QLIST_INIT(&blk->aio_notifiers); |
27ccdd52 | 361 | |
2cf22d6a | 362 | QTAILQ_INSERT_TAIL(&block_backends, blk, link); |
26f54e9a MA |
363 | return blk; |
364 | } | |
365 | ||
a3aeeab5 EB |
366 | /* |
367 | * Create a new BlockBackend connected to an existing BlockDriverState. | |
368 | * | |
369 | * @perm is a bitmasks of BLK_PERM_* constants which describes the | |
370 | * permissions to request for @bs that is attached to this | |
371 | * BlockBackend. @shared_perm is a bitmask which describes which | |
372 | * permissions may be granted to other users of the attached node. | |
373 | * Both sets of permissions can be changed later using blk_set_perm(). | |
374 | * | |
375 | * Return the new BlockBackend on success, null on failure. | |
376 | */ | |
377 | BlockBackend *blk_new_with_bs(BlockDriverState *bs, uint64_t perm, | |
378 | uint64_t shared_perm, Error **errp) | |
379 | { | |
380 | BlockBackend *blk = blk_new(bdrv_get_aio_context(bs), perm, shared_perm); | |
381 | ||
382 | if (blk_insert_bs(blk, bs, errp) < 0) { | |
383 | blk_unref(blk); | |
384 | return NULL; | |
385 | } | |
386 | return blk; | |
387 | } | |
388 | ||
7e7d56d9 | 389 | /* |
28eb9b12 | 390 | * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. |
d861ab3a | 391 | * The new BlockBackend is in the main AioContext. |
ca49a4fd HR |
392 | * |
393 | * Just as with bdrv_open(), after having called this function the reference to | |
394 | * @options belongs to the block layer (even on failure). | |
395 | * | |
396 | * TODO: Remove @filename and @flags; it should be possible to specify a whole | |
397 | * BDS tree just by specifying the @options QDict (or @reference, | |
398 | * alternatively). At the time of adding this function, this is not possible, | |
399 | * though, so callers of this function have to be able to specify @filename and | |
400 | * @flags. | |
401 | */ | |
efaa7c4e HR |
402 | BlockBackend *blk_new_open(const char *filename, const char *reference, |
403 | QDict *options, int flags, Error **errp) | |
ca49a4fd HR |
404 | { |
405 | BlockBackend *blk; | |
28eb9b12 | 406 | BlockDriverState *bs; |
1f4ad7d3 | 407 | uint64_t perm = 0; |
35b7f4ab | 408 | uint64_t shared = BLK_PERM_ALL; |
c62d32f5 | 409 | |
35b7f4ab KW |
410 | /* |
411 | * blk_new_open() is mainly used in .bdrv_create implementations and the | |
412 | * tools where sharing isn't a major concern because the BDS stays private | |
413 | * and the file is generally not supposed to be used by a second process, | |
414 | * so we just request permission according to the flags. | |
c62d32f5 KW |
415 | * |
416 | * The exceptions are xen_disk and blockdev_init(); in these cases, the | |
417 | * caller of blk_new_open() doesn't make use of the permissions, but they | |
418 | * shouldn't hurt either. We can still share everything here because the | |
35b7f4ab KW |
419 | * guest devices will add their own blockers if they can't share. |
420 | */ | |
1f4ad7d3 KW |
421 | if ((flags & BDRV_O_NO_IO) == 0) { |
422 | perm |= BLK_PERM_CONSISTENT_READ; | |
423 | if (flags & BDRV_O_RDWR) { | |
424 | perm |= BLK_PERM_WRITE; | |
425 | } | |
c62d32f5 KW |
426 | } |
427 | if (flags & BDRV_O_RESIZE) { | |
428 | perm |= BLK_PERM_RESIZE; | |
429 | } | |
35b7f4ab KW |
430 | if (flags & BDRV_O_NO_SHARE) { |
431 | shared = BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED; | |
432 | } | |
ca49a4fd | 433 | |
35b7f4ab | 434 | blk = blk_new(qemu_get_aio_context(), perm, shared); |
5b363937 HR |
435 | bs = bdrv_open(filename, reference, options, flags, errp); |
436 | if (!bs) { | |
ca49a4fd HR |
437 | blk_unref(blk); |
438 | return NULL; | |
439 | } | |
440 | ||
1f38f04e HR |
441 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
442 | BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY, | |
35b7f4ab | 443 | perm, shared, blk, errp); |
50bfbe93 | 444 | if (!blk->root) { |
50bfbe93 FZ |
445 | blk_unref(blk); |
446 | return NULL; | |
447 | } | |
72e775c7 | 448 | |
ca49a4fd HR |
449 | return blk; |
450 | } | |
451 | ||
26f54e9a MA |
452 | static void blk_delete(BlockBackend *blk) |
453 | { | |
454 | assert(!blk->refcnt); | |
e5e78550 | 455 | assert(!blk->name); |
a7f53e26 | 456 | assert(!blk->dev); |
022cdc9f | 457 | if (blk->public.throttle_group_member.throttle_state) { |
1606e4cf EB |
458 | blk_io_limits_disable(blk); |
459 | } | |
f21d96d0 | 460 | if (blk->root) { |
13855c6b | 461 | blk_remove_bs(blk); |
7e7d56d9 | 462 | } |
5f7772c4 FZ |
463 | if (blk->vmsh) { |
464 | qemu_del_vm_change_state_handler(blk->vmsh); | |
465 | blk->vmsh = NULL; | |
466 | } | |
3301f6c6 HR |
467 | assert(QLIST_EMPTY(&blk->remove_bs_notifiers.notifiers)); |
468 | assert(QLIST_EMPTY(&blk->insert_bs_notifiers.notifiers)); | |
d03654ea | 469 | assert(QLIST_EMPTY(&blk->aio_notifiers)); |
2cf22d6a | 470 | QTAILQ_REMOVE(&block_backends, blk, link); |
18e46a03 | 471 | drive_info_del(blk->legacy_dinfo); |
979e9b03 | 472 | block_acct_cleanup(&blk->stats); |
26f54e9a MA |
473 | g_free(blk); |
474 | } | |
475 | ||
8fb3c76c MA |
476 | static void drive_info_del(DriveInfo *dinfo) |
477 | { | |
478 | if (!dinfo) { | |
479 | return; | |
480 | } | |
481 | qemu_opts_del(dinfo->opts); | |
8fb3c76c MA |
482 | g_free(dinfo); |
483 | } | |
484 | ||
f636ae85 AG |
485 | int blk_get_refcnt(BlockBackend *blk) |
486 | { | |
487 | return blk ? blk->refcnt : 0; | |
488 | } | |
489 | ||
26f54e9a MA |
490 | /* |
491 | * Increment @blk's reference count. | |
492 | * @blk must not be null. | |
493 | */ | |
494 | void blk_ref(BlockBackend *blk) | |
495 | { | |
5ca9d21b | 496 | assert(blk->refcnt > 0); |
26f54e9a MA |
497 | blk->refcnt++; |
498 | } | |
499 | ||
500 | /* | |
501 | * Decrement @blk's reference count. | |
502 | * If this drops it to zero, destroy @blk. | |
503 | * For convenience, do nothing if @blk is null. | |
504 | */ | |
505 | void blk_unref(BlockBackend *blk) | |
506 | { | |
507 | if (blk) { | |
508 | assert(blk->refcnt > 0); | |
5ca9d21b KW |
509 | if (blk->refcnt > 1) { |
510 | blk->refcnt--; | |
511 | } else { | |
512 | blk_drain(blk); | |
513 | /* blk_drain() cannot resurrect blk, nobody held a reference */ | |
514 | assert(blk->refcnt == 1); | |
515 | blk->refcnt = 0; | |
26f54e9a MA |
516 | blk_delete(blk); |
517 | } | |
518 | } | |
519 | } | |
520 | ||
2cf22d6a HR |
521 | /* |
522 | * Behaves similarly to blk_next() but iterates over all BlockBackends, even the | |
523 | * ones which are hidden (i.e. are not referenced by the monitor). | |
524 | */ | |
a429b9b5 | 525 | BlockBackend *blk_all_next(BlockBackend *blk) |
2cf22d6a HR |
526 | { |
527 | return blk ? QTAILQ_NEXT(blk, link) | |
528 | : QTAILQ_FIRST(&block_backends); | |
529 | } | |
530 | ||
d8da3cef HR |
531 | void blk_remove_all_bs(void) |
532 | { | |
74d1b8fc | 533 | BlockBackend *blk = NULL; |
d8da3cef | 534 | |
2cf22d6a | 535 | while ((blk = blk_all_next(blk)) != NULL) { |
d8da3cef HR |
536 | AioContext *ctx = blk_get_aio_context(blk); |
537 | ||
538 | aio_context_acquire(ctx); | |
f21d96d0 | 539 | if (blk->root) { |
d8da3cef HR |
540 | blk_remove_bs(blk); |
541 | } | |
542 | aio_context_release(ctx); | |
543 | } | |
544 | } | |
545 | ||
26f54e9a | 546 | /* |
9492b0b9 | 547 | * Return the monitor-owned BlockBackend after @blk. |
26f54e9a MA |
548 | * If @blk is null, return the first one. |
549 | * Else, return @blk's next sibling, which may be null. | |
550 | * | |
551 | * To iterate over all BlockBackends, do | |
552 | * for (blk = blk_next(NULL); blk; blk = blk_next(blk)) { | |
553 | * ... | |
554 | * } | |
555 | */ | |
556 | BlockBackend *blk_next(BlockBackend *blk) | |
557 | { | |
9492b0b9 HR |
558 | return blk ? QTAILQ_NEXT(blk, monitor_link) |
559 | : QTAILQ_FIRST(&monitor_block_backends); | |
26f54e9a MA |
560 | } |
561 | ||
7c8eece4 KW |
562 | /* Iterates over all top-level BlockDriverStates, i.e. BDSs that are owned by |
563 | * the monitor or attached to a BlockBackend */ | |
88be7b4b | 564 | BlockDriverState *bdrv_next(BdrvNextIterator *it) |
7c8eece4 | 565 | { |
5e003f17 HR |
566 | BlockDriverState *bs, *old_bs; |
567 | ||
568 | /* Must be called from the main loop */ | |
569 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
981f4f57 | 570 | |
7c8eece4 KW |
571 | /* First, return all root nodes of BlockBackends. In order to avoid |
572 | * returning a BDS twice when multiple BBs refer to it, we only return it | |
573 | * if the BB is the first one in the parent list of the BDS. */ | |
574 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
5e003f17 HR |
575 | BlockBackend *old_blk = it->blk; |
576 | ||
577 | old_bs = old_blk ? blk_bs(old_blk) : NULL; | |
578 | ||
7c8eece4 KW |
579 | do { |
580 | it->blk = blk_all_next(it->blk); | |
88be7b4b KW |
581 | bs = it->blk ? blk_bs(it->blk) : NULL; |
582 | } while (it->blk && (bs == NULL || bdrv_first_blk(bs) != it->blk)); | |
7c8eece4 | 583 | |
5e003f17 HR |
584 | if (it->blk) { |
585 | blk_ref(it->blk); | |
586 | } | |
587 | blk_unref(old_blk); | |
588 | ||
88be7b4b | 589 | if (bs) { |
5e003f17 HR |
590 | bdrv_ref(bs); |
591 | bdrv_unref(old_bs); | |
88be7b4b | 592 | return bs; |
7c8eece4 KW |
593 | } |
594 | it->phase = BDRV_NEXT_MONITOR_OWNED; | |
5e003f17 HR |
595 | } else { |
596 | old_bs = it->bs; | |
7c8eece4 KW |
597 | } |
598 | ||
599 | /* Then return the monitor-owned BDSes without a BB attached. Ignore all | |
600 | * BDSes that are attached to a BlockBackend here; they have been handled | |
601 | * by the above block already */ | |
981f4f57 | 602 | do { |
7c8eece4 | 603 | it->bs = bdrv_next_monitor_owned(it->bs); |
88be7b4b KW |
604 | bs = it->bs; |
605 | } while (bs && bdrv_has_blk(bs)); | |
606 | ||
5e003f17 HR |
607 | if (bs) { |
608 | bdrv_ref(bs); | |
609 | } | |
610 | bdrv_unref(old_bs); | |
611 | ||
88be7b4b KW |
612 | return bs; |
613 | } | |
614 | ||
5e003f17 | 615 | static void bdrv_next_reset(BdrvNextIterator *it) |
88be7b4b KW |
616 | { |
617 | *it = (BdrvNextIterator) { | |
618 | .phase = BDRV_NEXT_BACKEND_ROOTS, | |
619 | }; | |
5e003f17 | 620 | } |
981f4f57 | 621 | |
5e003f17 HR |
622 | BlockDriverState *bdrv_first(BdrvNextIterator *it) |
623 | { | |
624 | bdrv_next_reset(it); | |
88be7b4b | 625 | return bdrv_next(it); |
981f4f57 HR |
626 | } |
627 | ||
5e003f17 HR |
628 | /* Must be called when aborting a bdrv_next() iteration before |
629 | * bdrv_next() returns NULL */ | |
630 | void bdrv_next_cleanup(BdrvNextIterator *it) | |
631 | { | |
632 | /* Must be called from the main loop */ | |
633 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
634 | ||
635 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
636 | if (it->blk) { | |
637 | bdrv_unref(blk_bs(it->blk)); | |
638 | blk_unref(it->blk); | |
639 | } | |
640 | } else { | |
641 | bdrv_unref(it->bs); | |
642 | } | |
643 | ||
644 | bdrv_next_reset(it); | |
645 | } | |
646 | ||
e5e78550 HR |
647 | /* |
648 | * Add a BlockBackend into the list of backends referenced by the monitor, with | |
649 | * the given @name acting as the handle for the monitor. | |
650 | * Strictly for use by blockdev.c. | |
651 | * | |
652 | * @name must not be null or empty. | |
653 | * | |
654 | * Returns true on success and false on failure. In the latter case, an Error | |
655 | * object is returned through @errp. | |
656 | */ | |
657 | bool monitor_add_blk(BlockBackend *blk, const char *name, Error **errp) | |
658 | { | |
659 | assert(!blk->name); | |
660 | assert(name && name[0]); | |
661 | ||
662 | if (!id_wellformed(name)) { | |
663 | error_setg(errp, "Invalid device name"); | |
664 | return false; | |
665 | } | |
666 | if (blk_by_name(name)) { | |
667 | error_setg(errp, "Device with id '%s' already exists", name); | |
668 | return false; | |
669 | } | |
670 | if (bdrv_find_node(name)) { | |
671 | error_setg(errp, | |
672 | "Device name '%s' conflicts with an existing node name", | |
673 | name); | |
674 | return false; | |
675 | } | |
676 | ||
677 | blk->name = g_strdup(name); | |
678 | QTAILQ_INSERT_TAIL(&monitor_block_backends, blk, monitor_link); | |
679 | return true; | |
680 | } | |
681 | ||
682 | /* | |
683 | * Remove a BlockBackend from the list of backends referenced by the monitor. | |
684 | * Strictly for use by blockdev.c. | |
685 | */ | |
686 | void monitor_remove_blk(BlockBackend *blk) | |
687 | { | |
688 | if (!blk->name) { | |
689 | return; | |
690 | } | |
691 | ||
692 | QTAILQ_REMOVE(&monitor_block_backends, blk, monitor_link); | |
693 | g_free(blk->name); | |
694 | blk->name = NULL; | |
695 | } | |
696 | ||
26f54e9a | 697 | /* |
7e7d56d9 | 698 | * Return @blk's name, a non-null string. |
e5e78550 | 699 | * Returns an empty string iff @blk is not referenced by the monitor. |
26f54e9a | 700 | */ |
0731a50f | 701 | const char *blk_name(const BlockBackend *blk) |
26f54e9a | 702 | { |
e5e78550 | 703 | return blk->name ?: ""; |
26f54e9a MA |
704 | } |
705 | ||
706 | /* | |
707 | * Return the BlockBackend with name @name if it exists, else null. | |
708 | * @name must not be null. | |
709 | */ | |
710 | BlockBackend *blk_by_name(const char *name) | |
711 | { | |
74d1b8fc | 712 | BlockBackend *blk = NULL; |
26f54e9a MA |
713 | |
714 | assert(name); | |
74d1b8fc | 715 | while ((blk = blk_next(blk)) != NULL) { |
26f54e9a MA |
716 | if (!strcmp(name, blk->name)) { |
717 | return blk; | |
718 | } | |
719 | } | |
720 | return NULL; | |
721 | } | |
7e7d56d9 MA |
722 | |
723 | /* | |
724 | * Return the BlockDriverState attached to @blk if any, else null. | |
725 | */ | |
726 | BlockDriverState *blk_bs(BlockBackend *blk) | |
727 | { | |
f21d96d0 | 728 | return blk->root ? blk->root->bs : NULL; |
7e7d56d9 MA |
729 | } |
730 | ||
7c8eece4 | 731 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs) |
dde33812 KW |
732 | { |
733 | BdrvChild *child; | |
734 | QLIST_FOREACH(child, &bs->parents, next_parent) { | |
bd86fb99 | 735 | if (child->klass == &child_root) { |
7c8eece4 | 736 | return child->opaque; |
dde33812 KW |
737 | } |
738 | } | |
739 | ||
7c8eece4 KW |
740 | return NULL; |
741 | } | |
742 | ||
743 | /* | |
744 | * Returns true if @bs has an associated BlockBackend. | |
745 | */ | |
746 | bool bdrv_has_blk(BlockDriverState *bs) | |
747 | { | |
748 | return bdrv_first_blk(bs) != NULL; | |
dde33812 KW |
749 | } |
750 | ||
b6c1bae5 KW |
751 | /* |
752 | * Returns true if @bs has only BlockBackends as parents. | |
753 | */ | |
754 | bool bdrv_is_root_node(BlockDriverState *bs) | |
755 | { | |
756 | BdrvChild *c; | |
757 | ||
758 | QLIST_FOREACH(c, &bs->parents, next_parent) { | |
bd86fb99 | 759 | if (c->klass != &child_root) { |
b6c1bae5 KW |
760 | return false; |
761 | } | |
762 | } | |
763 | ||
764 | return true; | |
765 | } | |
766 | ||
18e46a03 MA |
767 | /* |
768 | * Return @blk's DriveInfo if any, else null. | |
769 | */ | |
770 | DriveInfo *blk_legacy_dinfo(BlockBackend *blk) | |
771 | { | |
772 | return blk->legacy_dinfo; | |
773 | } | |
774 | ||
775 | /* | |
776 | * Set @blk's DriveInfo to @dinfo, and return it. | |
777 | * @blk must not have a DriveInfo set already. | |
778 | * No other BlockBackend may have the same DriveInfo set. | |
779 | */ | |
780 | DriveInfo *blk_set_legacy_dinfo(BlockBackend *blk, DriveInfo *dinfo) | |
781 | { | |
782 | assert(!blk->legacy_dinfo); | |
783 | return blk->legacy_dinfo = dinfo; | |
784 | } | |
785 | ||
786 | /* | |
787 | * Return the BlockBackend with DriveInfo @dinfo. | |
788 | * It must exist. | |
789 | */ | |
790 | BlockBackend *blk_by_legacy_dinfo(DriveInfo *dinfo) | |
791 | { | |
74d1b8fc | 792 | BlockBackend *blk = NULL; |
18e46a03 | 793 | |
74d1b8fc | 794 | while ((blk = blk_next(blk)) != NULL) { |
18e46a03 MA |
795 | if (blk->legacy_dinfo == dinfo) { |
796 | return blk; | |
797 | } | |
798 | } | |
799 | abort(); | |
800 | } | |
801 | ||
f2cd875d KW |
802 | /* |
803 | * Returns a pointer to the publicly accessible fields of @blk. | |
804 | */ | |
805 | BlockBackendPublic *blk_get_public(BlockBackend *blk) | |
806 | { | |
807 | return &blk->public; | |
808 | } | |
809 | ||
810 | /* | |
811 | * Returns a BlockBackend given the associated @public fields. | |
812 | */ | |
813 | BlockBackend *blk_by_public(BlockBackendPublic *public) | |
814 | { | |
815 | return container_of(public, BlockBackend, public); | |
816 | } | |
817 | ||
1c95f7e1 HR |
818 | /* |
819 | * Disassociates the currently associated BlockDriverState from @blk. | |
820 | */ | |
821 | void blk_remove_bs(BlockBackend *blk) | |
822 | { | |
c89bcf3a | 823 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
632a7735 | 824 | BlockDriverState *bs; |
e6cada92 | 825 | BdrvChild *root; |
022cdc9f | 826 | |
3301f6c6 | 827 | notifier_list_notify(&blk->remove_bs_notifiers, blk); |
c89bcf3a | 828 | if (tgm->throttle_state) { |
632a7735 ZL |
829 | bs = blk_bs(blk); |
830 | bdrv_drained_begin(bs); | |
c89bcf3a AG |
831 | throttle_group_detach_aio_context(tgm); |
832 | throttle_group_attach_aio_context(tgm, qemu_get_aio_context()); | |
632a7735 | 833 | bdrv_drained_end(bs); |
a5614993 | 834 | } |
1c95f7e1 | 835 | |
7ca7f0f6 KW |
836 | blk_update_root_state(blk); |
837 | ||
f45280cb GK |
838 | /* bdrv_root_unref_child() will cause blk->root to become stale and may |
839 | * switch to a completion coroutine later on. Let's drain all I/O here | |
840 | * to avoid that and a potential QEMU crash. | |
841 | */ | |
842 | blk_drain(blk); | |
e6cada92 | 843 | root = blk->root; |
f21d96d0 | 844 | blk->root = NULL; |
e6cada92 | 845 | bdrv_root_unref_child(root); |
1c95f7e1 HR |
846 | } |
847 | ||
0c3c36d6 HR |
848 | /* |
849 | * Associates a new BlockDriverState with @blk. | |
850 | */ | |
d7086422 | 851 | int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) |
0c3c36d6 | 852 | { |
c89bcf3a | 853 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
b441dc71 | 854 | bdrv_ref(bs); |
1f38f04e HR |
855 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
856 | BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY, | |
228ca37e | 857 | blk->perm, blk->shared_perm, |
1f38f04e | 858 | blk, errp); |
d7086422 KW |
859 | if (blk->root == NULL) { |
860 | return -EPERM; | |
861 | } | |
3301f6c6 HR |
862 | |
863 | notifier_list_notify(&blk->insert_bs_notifiers, blk); | |
c89bcf3a AG |
864 | if (tgm->throttle_state) { |
865 | throttle_group_detach_aio_context(tgm); | |
866 | throttle_group_attach_aio_context(tgm, bdrv_get_aio_context(bs)); | |
7ca7f0f6 | 867 | } |
d7086422 KW |
868 | |
869 | return 0; | |
0c3c36d6 HR |
870 | } |
871 | ||
981776b3 KW |
872 | /* |
873 | * Sets the permission bitmasks that the user of the BlockBackend needs. | |
874 | */ | |
875 | int blk_set_perm(BlockBackend *blk, uint64_t perm, uint64_t shared_perm, | |
876 | Error **errp) | |
877 | { | |
878 | int ret; | |
879 | ||
d35ff5e6 | 880 | if (blk->root && !blk->disable_perm) { |
981776b3 KW |
881 | ret = bdrv_child_try_set_perm(blk->root, perm, shared_perm, errp); |
882 | if (ret < 0) { | |
883 | return ret; | |
884 | } | |
885 | } | |
886 | ||
887 | blk->perm = perm; | |
888 | blk->shared_perm = shared_perm; | |
889 | ||
890 | return 0; | |
891 | } | |
892 | ||
887354bd KW |
893 | void blk_get_perm(BlockBackend *blk, uint64_t *perm, uint64_t *shared_perm) |
894 | { | |
895 | *perm = blk->perm; | |
896 | *shared_perm = blk->shared_perm; | |
897 | } | |
898 | ||
d09ea2d2 TH |
899 | /* |
900 | * Attach device model @dev to @blk. | |
901 | * Return 0 on success, -EBUSY when a device model is attached already. | |
902 | */ | |
903 | int blk_attach_dev(BlockBackend *blk, DeviceState *dev) | |
4be74634 | 904 | { |
a7f53e26 MA |
905 | if (blk->dev) { |
906 | return -EBUSY; | |
907 | } | |
d35ff5e6 KW |
908 | |
909 | /* While migration is still incoming, we don't need to apply the | |
910 | * permissions of guest device BlockBackends. We might still have a block | |
911 | * job or NBD server writing to the image for storage migration. */ | |
912 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
913 | blk->disable_perm = true; | |
914 | } | |
915 | ||
84ebe375 | 916 | blk_ref(blk); |
a7f53e26 | 917 | blk->dev = dev; |
373340b2 | 918 | blk_iostatus_reset(blk); |
d35ff5e6 | 919 | |
a7f53e26 | 920 | return 0; |
4be74634 MA |
921 | } |
922 | ||
a7f53e26 MA |
923 | /* |
924 | * Detach device model @dev from @blk. | |
925 | * @dev must be currently attached to @blk. | |
926 | */ | |
d09ea2d2 | 927 | void blk_detach_dev(BlockBackend *blk, DeviceState *dev) |
4be74634 | 928 | { |
a7f53e26 MA |
929 | assert(blk->dev == dev); |
930 | blk->dev = NULL; | |
931 | blk->dev_ops = NULL; | |
932 | blk->dev_opaque = NULL; | |
68e9ec01 | 933 | blk->guest_block_size = 512; |
981776b3 | 934 | blk_set_perm(blk, 0, BLK_PERM_ALL, &error_abort); |
84ebe375 | 935 | blk_unref(blk); |
4be74634 MA |
936 | } |
937 | ||
a7f53e26 MA |
938 | /* |
939 | * Return the device model attached to @blk if any, else null. | |
940 | */ | |
d09ea2d2 | 941 | DeviceState *blk_get_attached_dev(BlockBackend *blk) |
a7f53e26 MA |
942 | { |
943 | return blk->dev; | |
944 | } | |
945 | ||
2d76e724 KW |
946 | /* Return the qdev ID, or if no ID is assigned the QOM path, of the block |
947 | * device attached to the BlockBackend. */ | |
77beef83 | 948 | char *blk_get_attached_dev_id(BlockBackend *blk) |
2d76e724 | 949 | { |
d09ea2d2 | 950 | DeviceState *dev = blk->dev; |
2d76e724 KW |
951 | |
952 | if (!dev) { | |
953 | return g_strdup(""); | |
954 | } else if (dev->id) { | |
955 | return g_strdup(dev->id); | |
956 | } | |
602414d1 LM |
957 | |
958 | return object_get_canonical_path(OBJECT(dev)) ?: g_strdup(""); | |
2d76e724 KW |
959 | } |
960 | ||
1c89e1fa KW |
961 | /* |
962 | * Return the BlockBackend which has the device model @dev attached if it | |
963 | * exists, else null. | |
964 | * | |
965 | * @dev must not be null. | |
966 | */ | |
967 | BlockBackend *blk_by_dev(void *dev) | |
968 | { | |
969 | BlockBackend *blk = NULL; | |
970 | ||
971 | assert(dev != NULL); | |
972 | while ((blk = blk_all_next(blk)) != NULL) { | |
973 | if (blk->dev == dev) { | |
974 | return blk; | |
975 | } | |
976 | } | |
977 | return NULL; | |
978 | } | |
979 | ||
a7f53e26 MA |
980 | /* |
981 | * Set @blk's device model callbacks to @ops. | |
982 | * @opaque is the opaque argument to pass to the callbacks. | |
983 | * This is for use by device models. | |
984 | */ | |
985 | void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, | |
986 | void *opaque) | |
987 | { | |
988 | blk->dev_ops = ops; | |
989 | blk->dev_opaque = opaque; | |
f4d9cc88 JS |
990 | |
991 | /* Are we currently quiesced? Should we enforce this right now? */ | |
992 | if (blk->quiesce_counter && ops->drained_begin) { | |
993 | ops->drained_begin(opaque); | |
994 | } | |
a7f53e26 MA |
995 | } |
996 | ||
997 | /* | |
998 | * Notify @blk's attached device model of media change. | |
39829a01 KW |
999 | * |
1000 | * If @load is true, notify of media load. This action can fail, meaning that | |
1001 | * the medium cannot be loaded. @errp is set then. | |
1002 | * | |
1003 | * If @load is false, notify of media eject. This can never fail. | |
1004 | * | |
a7f53e26 MA |
1005 | * Also send DEVICE_TRAY_MOVED events as appropriate. |
1006 | */ | |
39829a01 | 1007 | void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp) |
a7f53e26 MA |
1008 | { |
1009 | if (blk->dev_ops && blk->dev_ops->change_media_cb) { | |
f1f57066 | 1010 | bool tray_was_open, tray_is_open; |
39829a01 | 1011 | Error *local_err = NULL; |
a7f53e26 | 1012 | |
f1f57066 | 1013 | tray_was_open = blk_dev_is_tray_open(blk); |
39829a01 KW |
1014 | blk->dev_ops->change_media_cb(blk->dev_opaque, load, &local_err); |
1015 | if (local_err) { | |
1016 | assert(load == true); | |
1017 | error_propagate(errp, local_err); | |
1018 | return; | |
1019 | } | |
f1f57066 HR |
1020 | tray_is_open = blk_dev_is_tray_open(blk); |
1021 | ||
1022 | if (tray_was_open != tray_is_open) { | |
2d76e724 | 1023 | char *id = blk_get_attached_dev_id(blk); |
3ab72385 | 1024 | qapi_event_send_device_tray_moved(blk_name(blk), id, tray_is_open); |
2d76e724 | 1025 | g_free(id); |
a7f53e26 MA |
1026 | } |
1027 | } | |
1028 | } | |
1029 | ||
5c8cab48 KW |
1030 | static void blk_root_change_media(BdrvChild *child, bool load) |
1031 | { | |
39829a01 | 1032 | blk_dev_change_media_cb(child->opaque, load, NULL); |
5c8cab48 KW |
1033 | } |
1034 | ||
a7f53e26 MA |
1035 | /* |
1036 | * Does @blk's attached device model have removable media? | |
1037 | * %true if no device model is attached. | |
1038 | */ | |
1039 | bool blk_dev_has_removable_media(BlockBackend *blk) | |
1040 | { | |
1041 | return !blk->dev || (blk->dev_ops && blk->dev_ops->change_media_cb); | |
1042 | } | |
1043 | ||
8f3a73bc HR |
1044 | /* |
1045 | * Does @blk's attached device model have a tray? | |
1046 | */ | |
1047 | bool blk_dev_has_tray(BlockBackend *blk) | |
1048 | { | |
1049 | return blk->dev_ops && blk->dev_ops->is_tray_open; | |
1050 | } | |
1051 | ||
a7f53e26 MA |
1052 | /* |
1053 | * Notify @blk's attached device model of a media eject request. | |
1054 | * If @force is true, the medium is about to be yanked out forcefully. | |
1055 | */ | |
1056 | void blk_dev_eject_request(BlockBackend *blk, bool force) | |
4be74634 | 1057 | { |
a7f53e26 MA |
1058 | if (blk->dev_ops && blk->dev_ops->eject_request_cb) { |
1059 | blk->dev_ops->eject_request_cb(blk->dev_opaque, force); | |
1060 | } | |
4be74634 MA |
1061 | } |
1062 | ||
a7f53e26 MA |
1063 | /* |
1064 | * Does @blk's attached device model have a tray, and is it open? | |
1065 | */ | |
1066 | bool blk_dev_is_tray_open(BlockBackend *blk) | |
4be74634 | 1067 | { |
8f3a73bc | 1068 | if (blk_dev_has_tray(blk)) { |
a7f53e26 MA |
1069 | return blk->dev_ops->is_tray_open(blk->dev_opaque); |
1070 | } | |
1071 | return false; | |
1072 | } | |
1073 | ||
1074 | /* | |
1075 | * Does @blk's attached device model have the medium locked? | |
1076 | * %false if the device model has no such lock. | |
1077 | */ | |
1078 | bool blk_dev_is_medium_locked(BlockBackend *blk) | |
1079 | { | |
1080 | if (blk->dev_ops && blk->dev_ops->is_medium_locked) { | |
1081 | return blk->dev_ops->is_medium_locked(blk->dev_opaque); | |
1082 | } | |
1083 | return false; | |
1084 | } | |
1085 | ||
1086 | /* | |
1087 | * Notify @blk's attached device model of a backend size change. | |
1088 | */ | |
5c8cab48 | 1089 | static void blk_root_resize(BdrvChild *child) |
a7f53e26 | 1090 | { |
5c8cab48 KW |
1091 | BlockBackend *blk = child->opaque; |
1092 | ||
a7f53e26 MA |
1093 | if (blk->dev_ops && blk->dev_ops->resize_cb) { |
1094 | blk->dev_ops->resize_cb(blk->dev_opaque); | |
1095 | } | |
1096 | } | |
1097 | ||
1098 | void blk_iostatus_enable(BlockBackend *blk) | |
1099 | { | |
373340b2 HR |
1100 | blk->iostatus_enabled = true; |
1101 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
1102 | } | |
1103 | ||
1104 | /* The I/O status is only enabled if the drive explicitly | |
1105 | * enables it _and_ the VM is configured to stop on errors */ | |
1106 | bool blk_iostatus_is_enabled(const BlockBackend *blk) | |
1107 | { | |
1108 | return (blk->iostatus_enabled && | |
1109 | (blk->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || | |
1110 | blk->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
1111 | blk->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
1112 | } | |
1113 | ||
1114 | BlockDeviceIoStatus blk_iostatus(const BlockBackend *blk) | |
1115 | { | |
1116 | return blk->iostatus; | |
1117 | } | |
1118 | ||
1119 | void blk_iostatus_disable(BlockBackend *blk) | |
1120 | { | |
1121 | blk->iostatus_enabled = false; | |
1122 | } | |
1123 | ||
1124 | void blk_iostatus_reset(BlockBackend *blk) | |
1125 | { | |
1126 | if (blk_iostatus_is_enabled(blk)) { | |
1127 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
373340b2 HR |
1128 | } |
1129 | } | |
1130 | ||
1131 | void blk_iostatus_set_err(BlockBackend *blk, int error) | |
1132 | { | |
1133 | assert(blk_iostatus_is_enabled(blk)); | |
1134 | if (blk->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
1135 | blk->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : | |
1136 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
1137 | } | |
4be74634 MA |
1138 | } |
1139 | ||
c10c9d96 KW |
1140 | void blk_set_allow_write_beyond_eof(BlockBackend *blk, bool allow) |
1141 | { | |
1142 | blk->allow_write_beyond_eof = allow; | |
1143 | } | |
1144 | ||
980b0f94 KW |
1145 | void blk_set_allow_aio_context_change(BlockBackend *blk, bool allow) |
1146 | { | |
1147 | blk->allow_aio_context_change = allow; | |
1148 | } | |
1149 | ||
cf312932 KW |
1150 | void blk_set_disable_request_queuing(BlockBackend *blk, bool disable) |
1151 | { | |
1152 | blk->disable_request_queuing = disable; | |
1153 | } | |
1154 | ||
e7f7d676 HR |
1155 | static int blk_check_byte_request(BlockBackend *blk, int64_t offset, |
1156 | size_t size) | |
1157 | { | |
1158 | int64_t len; | |
1159 | ||
1160 | if (size > INT_MAX) { | |
1161 | return -EIO; | |
1162 | } | |
1163 | ||
c09ba36c | 1164 | if (!blk_is_available(blk)) { |
e7f7d676 HR |
1165 | return -ENOMEDIUM; |
1166 | } | |
1167 | ||
e7f7d676 HR |
1168 | if (offset < 0) { |
1169 | return -EIO; | |
1170 | } | |
1171 | ||
c10c9d96 KW |
1172 | if (!blk->allow_write_beyond_eof) { |
1173 | len = blk_getlength(blk); | |
1174 | if (len < 0) { | |
1175 | return len; | |
1176 | } | |
1177 | ||
1178 | if (offset > len || len - offset < size) { | |
1179 | return -EIO; | |
1180 | } | |
e7f7d676 HR |
1181 | } |
1182 | ||
1183 | return 0; | |
1184 | } | |
1185 | ||
7f16476f | 1186 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
cf312932 KW |
1187 | static void coroutine_fn blk_wait_while_drained(BlockBackend *blk) |
1188 | { | |
7f16476f KW |
1189 | assert(blk->in_flight > 0); |
1190 | ||
cf312932 | 1191 | if (blk->quiesce_counter && !blk->disable_request_queuing) { |
7f16476f | 1192 | blk_dec_in_flight(blk); |
cf312932 | 1193 | qemu_co_queue_wait(&blk->queued_requests, NULL); |
7f16476f | 1194 | blk_inc_in_flight(blk); |
cf312932 KW |
1195 | } |
1196 | } | |
1197 | ||
fbb92b67 KW |
1198 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1199 | static int coroutine_fn | |
1200 | blk_do_preadv(BlockBackend *blk, int64_t offset, unsigned int bytes, | |
1201 | QEMUIOVector *qiov, BdrvRequestFlags flags) | |
4be74634 | 1202 | { |
1e98fefd | 1203 | int ret; |
cf312932 | 1204 | BlockDriverState *bs; |
1e98fefd | 1205 | |
cf312932 | 1206 | blk_wait_while_drained(blk); |
1e98fefd | 1207 | |
cf312932 KW |
1208 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1209 | bs = blk_bs(blk); | |
99723548 | 1210 | trace_blk_co_preadv(blk, bs, offset, bytes, flags); |
1e98fefd KW |
1211 | |
1212 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1213 | if (ret < 0) { |
1214 | return ret; | |
1215 | } | |
1216 | ||
99723548 PB |
1217 | bdrv_inc_in_flight(bs); |
1218 | ||
441565b2 | 1219 | /* throttling disk I/O */ |
022cdc9f MP |
1220 | if (blk->public.throttle_group_member.throttle_state) { |
1221 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1222 | bytes, false); | |
441565b2 KW |
1223 | } |
1224 | ||
99723548 PB |
1225 | ret = bdrv_co_preadv(blk->root, offset, bytes, qiov, flags); |
1226 | bdrv_dec_in_flight(bs); | |
1227 | return ret; | |
1bf1cbc9 KW |
1228 | } |
1229 | ||
fbb92b67 KW |
1230 | int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset, |
1231 | unsigned int bytes, QEMUIOVector *qiov, | |
1232 | BdrvRequestFlags flags) | |
1233 | { | |
1234 | int ret; | |
1235 | ||
1236 | blk_inc_in_flight(blk); | |
1237 | ret = blk_do_preadv(blk, offset, bytes, qiov, flags); | |
1238 | blk_dec_in_flight(blk); | |
1239 | ||
1240 | return ret; | |
1241 | } | |
1242 | ||
1243 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ | |
1244 | static int coroutine_fn | |
1245 | blk_do_pwritev_part(BlockBackend *blk, int64_t offset, unsigned int bytes, | |
1246 | QEMUIOVector *qiov, size_t qiov_offset, | |
1247 | BdrvRequestFlags flags) | |
a8823a3b | 1248 | { |
bfd18d1e | 1249 | int ret; |
cf312932 | 1250 | BlockDriverState *bs; |
bfd18d1e | 1251 | |
cf312932 | 1252 | blk_wait_while_drained(blk); |
bfd18d1e | 1253 | |
cf312932 KW |
1254 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1255 | bs = blk_bs(blk); | |
99723548 | 1256 | trace_blk_co_pwritev(blk, bs, offset, bytes, flags); |
1e98fefd | 1257 | |
bfd18d1e | 1258 | ret = blk_check_byte_request(blk, offset, bytes); |
a8823a3b KW |
1259 | if (ret < 0) { |
1260 | return ret; | |
1261 | } | |
1262 | ||
99723548 | 1263 | bdrv_inc_in_flight(bs); |
441565b2 | 1264 | /* throttling disk I/O */ |
022cdc9f MP |
1265 | if (blk->public.throttle_group_member.throttle_state) { |
1266 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1267 | bytes, true); | |
441565b2 KW |
1268 | } |
1269 | ||
bfd18d1e KW |
1270 | if (!blk->enable_write_cache) { |
1271 | flags |= BDRV_REQ_FUA; | |
1272 | } | |
1273 | ||
b3016864 VSO |
1274 | ret = bdrv_co_pwritev_part(blk->root, offset, bytes, qiov, qiov_offset, |
1275 | flags); | |
99723548 PB |
1276 | bdrv_dec_in_flight(bs); |
1277 | return ret; | |
a8823a3b KW |
1278 | } |
1279 | ||
fbb92b67 KW |
1280 | int coroutine_fn blk_co_pwritev_part(BlockBackend *blk, int64_t offset, |
1281 | unsigned int bytes, | |
1282 | QEMUIOVector *qiov, size_t qiov_offset, | |
1283 | BdrvRequestFlags flags) | |
1284 | { | |
1285 | int ret; | |
1286 | ||
1287 | blk_inc_in_flight(blk); | |
1288 | ret = blk_do_pwritev_part(blk, offset, bytes, qiov, qiov_offset, flags); | |
1289 | blk_dec_in_flight(blk); | |
1290 | ||
1291 | return ret; | |
1292 | } | |
1293 | ||
b3016864 VSO |
1294 | int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset, |
1295 | unsigned int bytes, QEMUIOVector *qiov, | |
1296 | BdrvRequestFlags flags) | |
1297 | { | |
1298 | return blk_co_pwritev_part(blk, offset, bytes, qiov, 0, flags); | |
1299 | } | |
1300 | ||
1bf1cbc9 KW |
1301 | typedef struct BlkRwCo { |
1302 | BlockBackend *blk; | |
1303 | int64_t offset; | |
c060332c | 1304 | void *iobuf; |
1bf1cbc9 KW |
1305 | int ret; |
1306 | BdrvRequestFlags flags; | |
1307 | } BlkRwCo; | |
1308 | ||
1309 | static void blk_read_entry(void *opaque) | |
1310 | { | |
1311 | BlkRwCo *rwco = opaque; | |
c060332c | 1312 | QEMUIOVector *qiov = rwco->iobuf; |
1bf1cbc9 | 1313 | |
fbb92b67 | 1314 | rwco->ret = blk_do_preadv(rwco->blk, rwco->offset, qiov->size, |
c060332c | 1315 | qiov, rwco->flags); |
4720cbee | 1316 | aio_wait_kick(); |
1bf1cbc9 KW |
1317 | } |
1318 | ||
a8823a3b KW |
1319 | static void blk_write_entry(void *opaque) |
1320 | { | |
1321 | BlkRwCo *rwco = opaque; | |
c060332c | 1322 | QEMUIOVector *qiov = rwco->iobuf; |
a8823a3b | 1323 | |
fbb92b67 KW |
1324 | rwco->ret = blk_do_pwritev_part(rwco->blk, rwco->offset, qiov->size, |
1325 | qiov, 0, rwco->flags); | |
4720cbee | 1326 | aio_wait_kick(); |
a8823a3b KW |
1327 | } |
1328 | ||
a55d3fba KW |
1329 | static int blk_prw(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1330 | int64_t bytes, CoroutineEntry co_entry, | |
1331 | BdrvRequestFlags flags) | |
1bf1cbc9 | 1332 | { |
ae5a9592 VSO |
1333 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); |
1334 | BlkRwCo rwco = { | |
1bf1cbc9 | 1335 | .blk = blk, |
a55d3fba | 1336 | .offset = offset, |
c060332c | 1337 | .iobuf = &qiov, |
fc1453cd | 1338 | .flags = flags, |
1bf1cbc9 KW |
1339 | .ret = NOT_DONE, |
1340 | }; | |
1341 | ||
fbb92b67 | 1342 | blk_inc_in_flight(blk); |
35f106e6 PB |
1343 | if (qemu_in_coroutine()) { |
1344 | /* Fast-path if already in coroutine context */ | |
1345 | co_entry(&rwco); | |
1346 | } else { | |
1347 | Coroutine *co = qemu_coroutine_create(co_entry, &rwco); | |
e92f0e19 | 1348 | bdrv_coroutine_enter(blk_bs(blk), co); |
35f106e6 PB |
1349 | BDRV_POLL_WHILE(blk_bs(blk), rwco.ret == NOT_DONE); |
1350 | } | |
fbb92b67 | 1351 | blk_dec_in_flight(blk); |
1bf1cbc9 KW |
1352 | |
1353 | return rwco.ret; | |
4be74634 MA |
1354 | } |
1355 | ||
d004bd52 | 1356 | int blk_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 1357 | int bytes, BdrvRequestFlags flags) |
0df89e8e | 1358 | { |
f5a5ca79 | 1359 | return blk_prw(blk, offset, NULL, bytes, blk_write_entry, |
983a1600 | 1360 | flags | BDRV_REQ_ZERO_WRITE); |
0df89e8e KW |
1361 | } |
1362 | ||
720ff280 KW |
1363 | int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags) |
1364 | { | |
1365 | return bdrv_make_zero(blk->root, flags); | |
1366 | } | |
1367 | ||
c90e2a9c | 1368 | void blk_inc_in_flight(BlockBackend *blk) |
33f2a757 | 1369 | { |
d73415a3 | 1370 | qatomic_inc(&blk->in_flight); |
33f2a757 SH |
1371 | } |
1372 | ||
c90e2a9c | 1373 | void blk_dec_in_flight(BlockBackend *blk) |
33f2a757 | 1374 | { |
d73415a3 | 1375 | qatomic_dec(&blk->in_flight); |
cfe29d82 | 1376 | aio_wait_kick(); |
33f2a757 SH |
1377 | } |
1378 | ||
e7f7d676 HR |
1379 | static void error_callback_bh(void *opaque) |
1380 | { | |
1381 | struct BlockBackendAIOCB *acb = opaque; | |
99723548 | 1382 | |
33f2a757 | 1383 | blk_dec_in_flight(acb->blk); |
e7f7d676 HR |
1384 | acb->common.cb(acb->common.opaque, acb->ret); |
1385 | qemu_aio_unref(acb); | |
1386 | } | |
1387 | ||
ca78ecfa PL |
1388 | BlockAIOCB *blk_abort_aio_request(BlockBackend *blk, |
1389 | BlockCompletionFunc *cb, | |
1390 | void *opaque, int ret) | |
e7f7d676 HR |
1391 | { |
1392 | struct BlockBackendAIOCB *acb; | |
e7f7d676 | 1393 | |
33f2a757 | 1394 | blk_inc_in_flight(blk); |
e7f7d676 | 1395 | acb = blk_aio_get(&block_backend_aiocb_info, blk, cb, opaque); |
4981bdec | 1396 | acb->blk = blk; |
e7f7d676 HR |
1397 | acb->ret = ret; |
1398 | ||
e4ec5ad4 PD |
1399 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1400 | error_callback_bh, acb); | |
e7f7d676 HR |
1401 | return &acb->common; |
1402 | } | |
1403 | ||
57d6a428 KW |
1404 | typedef struct BlkAioEmAIOCB { |
1405 | BlockAIOCB common; | |
1406 | BlkRwCo rwco; | |
7fa84cd8 | 1407 | int bytes; |
57d6a428 | 1408 | bool has_returned; |
57d6a428 KW |
1409 | } BlkAioEmAIOCB; |
1410 | ||
1d719ddc SH |
1411 | static AioContext *blk_aio_em_aiocb_get_aio_context(BlockAIOCB *acb_) |
1412 | { | |
1413 | BlkAioEmAIOCB *acb = container_of(acb_, BlkAioEmAIOCB, common); | |
1414 | ||
1415 | return blk_get_aio_context(acb->rwco.blk); | |
1416 | } | |
1417 | ||
57d6a428 KW |
1418 | static const AIOCBInfo blk_aio_em_aiocb_info = { |
1419 | .aiocb_size = sizeof(BlkAioEmAIOCB), | |
1d719ddc | 1420 | .get_aio_context = blk_aio_em_aiocb_get_aio_context, |
57d6a428 KW |
1421 | }; |
1422 | ||
1423 | static void blk_aio_complete(BlkAioEmAIOCB *acb) | |
1424 | { | |
57d6a428 KW |
1425 | if (acb->has_returned) { |
1426 | acb->common.cb(acb->common.opaque, acb->rwco.ret); | |
46aaf2a5 | 1427 | blk_dec_in_flight(acb->rwco.blk); |
57d6a428 KW |
1428 | qemu_aio_unref(acb); |
1429 | } | |
1430 | } | |
1431 | ||
1432 | static void blk_aio_complete_bh(void *opaque) | |
1433 | { | |
fffb6e12 | 1434 | BlkAioEmAIOCB *acb = opaque; |
fffb6e12 PB |
1435 | assert(acb->has_returned); |
1436 | blk_aio_complete(acb); | |
57d6a428 KW |
1437 | } |
1438 | ||
7fa84cd8 | 1439 | static BlockAIOCB *blk_aio_prwv(BlockBackend *blk, int64_t offset, int bytes, |
c060332c | 1440 | void *iobuf, CoroutineEntry co_entry, |
57d6a428 KW |
1441 | BdrvRequestFlags flags, |
1442 | BlockCompletionFunc *cb, void *opaque) | |
1443 | { | |
1444 | BlkAioEmAIOCB *acb; | |
1445 | Coroutine *co; | |
1446 | ||
33f2a757 | 1447 | blk_inc_in_flight(blk); |
57d6a428 KW |
1448 | acb = blk_aio_get(&blk_aio_em_aiocb_info, blk, cb, opaque); |
1449 | acb->rwco = (BlkRwCo) { | |
1450 | .blk = blk, | |
1451 | .offset = offset, | |
c060332c | 1452 | .iobuf = iobuf, |
57d6a428 KW |
1453 | .flags = flags, |
1454 | .ret = NOT_DONE, | |
1455 | }; | |
7fa84cd8 | 1456 | acb->bytes = bytes; |
57d6a428 KW |
1457 | acb->has_returned = false; |
1458 | ||
0b8b8753 | 1459 | co = qemu_coroutine_create(co_entry, acb); |
e92f0e19 | 1460 | bdrv_coroutine_enter(blk_bs(blk), co); |
57d6a428 KW |
1461 | |
1462 | acb->has_returned = true; | |
1463 | if (acb->rwco.ret != NOT_DONE) { | |
e4ec5ad4 PD |
1464 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1465 | blk_aio_complete_bh, acb); | |
57d6a428 KW |
1466 | } |
1467 | ||
1468 | return &acb->common; | |
1469 | } | |
1470 | ||
1471 | static void blk_aio_read_entry(void *opaque) | |
1472 | { | |
1473 | BlkAioEmAIOCB *acb = opaque; | |
1474 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1475 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1476 | |
c060332c | 1477 | assert(qiov->size == acb->bytes); |
fbb92b67 | 1478 | rwco->ret = blk_do_preadv(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1479 | qiov, rwco->flags); |
57d6a428 KW |
1480 | blk_aio_complete(acb); |
1481 | } | |
1482 | ||
1483 | static void blk_aio_write_entry(void *opaque) | |
1484 | { | |
1485 | BlkAioEmAIOCB *acb = opaque; | |
1486 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1487 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1488 | |
c060332c | 1489 | assert(!qiov || qiov->size == acb->bytes); |
fbb92b67 KW |
1490 | rwco->ret = blk_do_pwritev_part(rwco->blk, rwco->offset, acb->bytes, |
1491 | qiov, 0, rwco->flags); | |
57d6a428 KW |
1492 | blk_aio_complete(acb); |
1493 | } | |
1494 | ||
d004bd52 EB |
1495 | BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
1496 | int count, BdrvRequestFlags flags, | |
1497 | BlockCompletionFunc *cb, void *opaque) | |
4be74634 | 1498 | { |
983a1600 EB |
1499 | return blk_aio_prwv(blk, offset, count, NULL, blk_aio_write_entry, |
1500 | flags | BDRV_REQ_ZERO_WRITE, cb, opaque); | |
4be74634 MA |
1501 | } |
1502 | ||
1503 | int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count) | |
1504 | { | |
a55d3fba | 1505 | int ret = blk_prw(blk, offset, buf, count, blk_read_entry, 0); |
e7f7d676 HR |
1506 | if (ret < 0) { |
1507 | return ret; | |
1508 | } | |
a55d3fba | 1509 | return count; |
4be74634 MA |
1510 | } |
1511 | ||
8341f00d EB |
1512 | int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count, |
1513 | BdrvRequestFlags flags) | |
4be74634 | 1514 | { |
8341f00d EB |
1515 | int ret = blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
1516 | flags); | |
e7f7d676 HR |
1517 | if (ret < 0) { |
1518 | return ret; | |
1519 | } | |
a55d3fba | 1520 | return count; |
4be74634 MA |
1521 | } |
1522 | ||
1523 | int64_t blk_getlength(BlockBackend *blk) | |
1524 | { | |
c09ba36c HR |
1525 | if (!blk_is_available(blk)) { |
1526 | return -ENOMEDIUM; | |
1527 | } | |
1528 | ||
f21d96d0 | 1529 | return bdrv_getlength(blk_bs(blk)); |
4be74634 MA |
1530 | } |
1531 | ||
1532 | void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr) | |
1533 | { | |
f21d96d0 | 1534 | if (!blk_bs(blk)) { |
a46fc9c9 HR |
1535 | *nb_sectors_ptr = 0; |
1536 | } else { | |
f21d96d0 | 1537 | bdrv_get_geometry(blk_bs(blk), nb_sectors_ptr); |
a46fc9c9 | 1538 | } |
4be74634 MA |
1539 | } |
1540 | ||
1ef01253 HR |
1541 | int64_t blk_nb_sectors(BlockBackend *blk) |
1542 | { | |
c09ba36c HR |
1543 | if (!blk_is_available(blk)) { |
1544 | return -ENOMEDIUM; | |
1545 | } | |
1546 | ||
f21d96d0 | 1547 | return bdrv_nb_sectors(blk_bs(blk)); |
1ef01253 HR |
1548 | } |
1549 | ||
60cb2fa7 EB |
1550 | BlockAIOCB *blk_aio_preadv(BlockBackend *blk, int64_t offset, |
1551 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1552 | BlockCompletionFunc *cb, void *opaque) | |
1553 | { | |
1554 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1555 | blk_aio_read_entry, flags, cb, opaque); | |
1556 | } | |
1557 | ||
60cb2fa7 EB |
1558 | BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
1559 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1560 | BlockCompletionFunc *cb, void *opaque) | |
1561 | { | |
1562 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1563 | blk_aio_write_entry, flags, cb, opaque); | |
1564 | } | |
1565 | ||
4be74634 MA |
1566 | void blk_aio_cancel(BlockAIOCB *acb) |
1567 | { | |
1568 | bdrv_aio_cancel(acb); | |
1569 | } | |
1570 | ||
1571 | void blk_aio_cancel_async(BlockAIOCB *acb) | |
1572 | { | |
1573 | bdrv_aio_cancel_async(acb); | |
1574 | } | |
1575 | ||
fbb92b67 KW |
1576 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1577 | static int coroutine_fn | |
1578 | blk_do_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
4be74634 | 1579 | { |
cf312932 KW |
1580 | blk_wait_while_drained(blk); |
1581 | ||
c09ba36c HR |
1582 | if (!blk_is_available(blk)) { |
1583 | return -ENOMEDIUM; | |
1584 | } | |
1585 | ||
48af776a KW |
1586 | return bdrv_co_ioctl(blk_bs(blk), req, buf); |
1587 | } | |
1588 | ||
1589 | static void blk_ioctl_entry(void *opaque) | |
1590 | { | |
1591 | BlkRwCo *rwco = opaque; | |
c060332c DS |
1592 | QEMUIOVector *qiov = rwco->iobuf; |
1593 | ||
fbb92b67 | 1594 | rwco->ret = blk_do_ioctl(rwco->blk, rwco->offset, qiov->iov[0].iov_base); |
4720cbee | 1595 | aio_wait_kick(); |
48af776a KW |
1596 | } |
1597 | ||
1598 | int blk_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
1599 | { | |
1600 | return blk_prw(blk, req, buf, 0, blk_ioctl_entry, 0); | |
1601 | } | |
1602 | ||
1603 | static void blk_aio_ioctl_entry(void *opaque) | |
1604 | { | |
1605 | BlkAioEmAIOCB *acb = opaque; | |
1606 | BlkRwCo *rwco = &acb->rwco; | |
1607 | ||
fbb92b67 | 1608 | rwco->ret = blk_do_ioctl(rwco->blk, rwco->offset, rwco->iobuf); |
c060332c | 1609 | |
48af776a | 1610 | blk_aio_complete(acb); |
4be74634 MA |
1611 | } |
1612 | ||
1613 | BlockAIOCB *blk_aio_ioctl(BlockBackend *blk, unsigned long int req, void *buf, | |
1614 | BlockCompletionFunc *cb, void *opaque) | |
1615 | { | |
c060332c | 1616 | return blk_aio_prwv(blk, req, 0, buf, blk_aio_ioctl_entry, 0, cb, opaque); |
4be74634 MA |
1617 | } |
1618 | ||
fbb92b67 KW |
1619 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1620 | static int coroutine_fn | |
1621 | blk_do_pdiscard(BlockBackend *blk, int64_t offset, int bytes) | |
2bb0dce7 | 1622 | { |
cf312932 KW |
1623 | int ret; |
1624 | ||
1625 | blk_wait_while_drained(blk); | |
1626 | ||
1627 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1628 | if (ret < 0) { |
1629 | return ret; | |
1630 | } | |
1631 | ||
0b9fd3f4 | 1632 | return bdrv_co_pdiscard(blk->root, offset, bytes); |
2bb0dce7 HR |
1633 | } |
1634 | ||
564806c5 KW |
1635 | static void blk_aio_pdiscard_entry(void *opaque) |
1636 | { | |
1637 | BlkAioEmAIOCB *acb = opaque; | |
1638 | BlkRwCo *rwco = &acb->rwco; | |
1639 | ||
fbb92b67 | 1640 | rwco->ret = blk_do_pdiscard(rwco->blk, rwco->offset, acb->bytes); |
564806c5 KW |
1641 | blk_aio_complete(acb); |
1642 | } | |
1643 | ||
1644 | BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, | |
1645 | int64_t offset, int bytes, | |
1646 | BlockCompletionFunc *cb, void *opaque) | |
1647 | { | |
1648 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_pdiscard_entry, 0, | |
1649 | cb, opaque); | |
1650 | } | |
1651 | ||
fbb92b67 KW |
1652 | int coroutine_fn blk_co_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
1653 | { | |
1654 | int ret; | |
1655 | ||
1656 | blk_inc_in_flight(blk); | |
1657 | ret = blk_do_pdiscard(blk, offset, bytes); | |
1658 | blk_dec_in_flight(blk); | |
1659 | ||
1660 | return ret; | |
1661 | } | |
1662 | ||
564806c5 KW |
1663 | static void blk_pdiscard_entry(void *opaque) |
1664 | { | |
1665 | BlkRwCo *rwco = opaque; | |
1666 | QEMUIOVector *qiov = rwco->iobuf; | |
1667 | ||
fbb92b67 | 1668 | rwco->ret = blk_do_pdiscard(rwco->blk, rwco->offset, qiov->size); |
564806c5 KW |
1669 | aio_wait_kick(); |
1670 | } | |
1671 | ||
1672 | int blk_pdiscard(BlockBackend *blk, int64_t offset, int bytes) | |
1673 | { | |
1674 | return blk_prw(blk, offset, NULL, bytes, blk_pdiscard_entry, 0); | |
1675 | } | |
1676 | ||
fbb92b67 KW |
1677 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1678 | static int coroutine_fn blk_do_flush(BlockBackend *blk) | |
2bb0dce7 | 1679 | { |
cf312932 KW |
1680 | blk_wait_while_drained(blk); |
1681 | ||
c09ba36c HR |
1682 | if (!blk_is_available(blk)) { |
1683 | return -ENOMEDIUM; | |
1684 | } | |
1685 | ||
f21d96d0 | 1686 | return bdrv_co_flush(blk_bs(blk)); |
2bb0dce7 HR |
1687 | } |
1688 | ||
564806c5 KW |
1689 | static void blk_aio_flush_entry(void *opaque) |
1690 | { | |
1691 | BlkAioEmAIOCB *acb = opaque; | |
1692 | BlkRwCo *rwco = &acb->rwco; | |
1693 | ||
fbb92b67 | 1694 | rwco->ret = blk_do_flush(rwco->blk); |
564806c5 KW |
1695 | blk_aio_complete(acb); |
1696 | } | |
1697 | ||
1698 | BlockAIOCB *blk_aio_flush(BlockBackend *blk, | |
1699 | BlockCompletionFunc *cb, void *opaque) | |
1700 | { | |
1701 | return blk_aio_prwv(blk, 0, 0, NULL, blk_aio_flush_entry, 0, cb, opaque); | |
1702 | } | |
1703 | ||
fbb92b67 KW |
1704 | int coroutine_fn blk_co_flush(BlockBackend *blk) |
1705 | { | |
1706 | int ret; | |
1707 | ||
1708 | blk_inc_in_flight(blk); | |
1709 | ret = blk_do_flush(blk); | |
1710 | blk_dec_in_flight(blk); | |
1711 | ||
1712 | return ret; | |
1713 | } | |
1714 | ||
be07a889 | 1715 | static void blk_flush_entry(void *opaque) |
4be74634 | 1716 | { |
be07a889 | 1717 | BlkRwCo *rwco = opaque; |
fbb92b67 | 1718 | rwco->ret = blk_do_flush(rwco->blk); |
4720cbee | 1719 | aio_wait_kick(); |
be07a889 | 1720 | } |
c09ba36c | 1721 | |
be07a889 KW |
1722 | int blk_flush(BlockBackend *blk) |
1723 | { | |
1724 | return blk_prw(blk, 0, NULL, 0, blk_flush_entry, 0); | |
4be74634 MA |
1725 | } |
1726 | ||
97b0385a AY |
1727 | void blk_drain(BlockBackend *blk) |
1728 | { | |
33f2a757 SH |
1729 | BlockDriverState *bs = blk_bs(blk); |
1730 | ||
1731 | if (bs) { | |
1732 | bdrv_drained_begin(bs); | |
1733 | } | |
1734 | ||
1735 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 | 1736 | AIO_WAIT_WHILE(blk_get_aio_context(blk), |
d73415a3 | 1737 | qatomic_mb_read(&blk->in_flight) > 0); |
33f2a757 SH |
1738 | |
1739 | if (bs) { | |
1740 | bdrv_drained_end(bs); | |
a46fc9c9 | 1741 | } |
97b0385a AY |
1742 | } |
1743 | ||
4be74634 MA |
1744 | void blk_drain_all(void) |
1745 | { | |
33f2a757 SH |
1746 | BlockBackend *blk = NULL; |
1747 | ||
1748 | bdrv_drain_all_begin(); | |
1749 | ||
1750 | while ((blk = blk_all_next(blk)) != NULL) { | |
1751 | AioContext *ctx = blk_get_aio_context(blk); | |
1752 | ||
1753 | aio_context_acquire(ctx); | |
1754 | ||
1755 | /* We may have -ENOMEDIUM completions in flight */ | |
d73415a3 | 1756 | AIO_WAIT_WHILE(ctx, qatomic_mb_read(&blk->in_flight) > 0); |
33f2a757 SH |
1757 | |
1758 | aio_context_release(ctx); | |
1759 | } | |
1760 | ||
1761 | bdrv_drain_all_end(); | |
4be74634 MA |
1762 | } |
1763 | ||
373340b2 HR |
1764 | void blk_set_on_error(BlockBackend *blk, BlockdevOnError on_read_error, |
1765 | BlockdevOnError on_write_error) | |
1766 | { | |
1767 | blk->on_read_error = on_read_error; | |
1768 | blk->on_write_error = on_write_error; | |
1769 | } | |
1770 | ||
4be74634 MA |
1771 | BlockdevOnError blk_get_on_error(BlockBackend *blk, bool is_read) |
1772 | { | |
373340b2 | 1773 | return is_read ? blk->on_read_error : blk->on_write_error; |
4be74634 MA |
1774 | } |
1775 | ||
1776 | BlockErrorAction blk_get_error_action(BlockBackend *blk, bool is_read, | |
1777 | int error) | |
1778 | { | |
373340b2 HR |
1779 | BlockdevOnError on_err = blk_get_on_error(blk, is_read); |
1780 | ||
1781 | switch (on_err) { | |
1782 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
1783 | return (error == ENOSPC) ? | |
1784 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
1785 | case BLOCKDEV_ON_ERROR_STOP: | |
1786 | return BLOCK_ERROR_ACTION_STOP; | |
1787 | case BLOCKDEV_ON_ERROR_REPORT: | |
1788 | return BLOCK_ERROR_ACTION_REPORT; | |
1789 | case BLOCKDEV_ON_ERROR_IGNORE: | |
1790 | return BLOCK_ERROR_ACTION_IGNORE; | |
8c398252 | 1791 | case BLOCKDEV_ON_ERROR_AUTO: |
373340b2 HR |
1792 | default: |
1793 | abort(); | |
1794 | } | |
4be74634 MA |
1795 | } |
1796 | ||
373340b2 HR |
1797 | static void send_qmp_error_event(BlockBackend *blk, |
1798 | BlockErrorAction action, | |
1799 | bool is_read, int error) | |
1800 | { | |
1801 | IoOperationType optype; | |
bfe1a14c | 1802 | BlockDriverState *bs = blk_bs(blk); |
373340b2 HR |
1803 | |
1804 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; | |
bfe1a14c KW |
1805 | qapi_event_send_block_io_error(blk_name(blk), !!bs, |
1806 | bs ? bdrv_get_node_name(bs) : NULL, optype, | |
2bf7e10f | 1807 | action, blk_iostatus_is_enabled(blk), |
3ab72385 | 1808 | error == ENOSPC, strerror(error)); |
373340b2 HR |
1809 | } |
1810 | ||
1811 | /* This is done by device models because, while the block layer knows | |
1812 | * about the error, it does not know whether an operation comes from | |
1813 | * the device or the block layer (from a job, for example). | |
1814 | */ | |
4be74634 MA |
1815 | void blk_error_action(BlockBackend *blk, BlockErrorAction action, |
1816 | bool is_read, int error) | |
1817 | { | |
373340b2 HR |
1818 | assert(error >= 0); |
1819 | ||
1820 | if (action == BLOCK_ERROR_ACTION_STOP) { | |
1821 | /* First set the iostatus, so that "info block" returns an iostatus | |
1822 | * that matches the events raised so far (an additional error iostatus | |
1823 | * is fine, but not a lost one). | |
1824 | */ | |
1825 | blk_iostatus_set_err(blk, error); | |
1826 | ||
1827 | /* Then raise the request to stop the VM and the event. | |
1828 | * qemu_system_vmstop_request_prepare has two effects. First, | |
1829 | * it ensures that the STOP event always comes after the | |
1830 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
1831 | * can observe the STOP event and do a "cont" before the STOP | |
1832 | * event is issued, the VM will not stop. In this case, vm_start() | |
1833 | * also ensures that the STOP/RESUME pair of events is emitted. | |
1834 | */ | |
1835 | qemu_system_vmstop_request_prepare(); | |
1836 | send_qmp_error_event(blk, action, is_read, error); | |
1837 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); | |
1838 | } else { | |
1839 | send_qmp_error_event(blk, action, is_read, error); | |
1840 | } | |
4be74634 MA |
1841 | } |
1842 | ||
86b1cf32 KW |
1843 | /* |
1844 | * Returns true if the BlockBackend can support taking write permissions | |
1845 | * (because its root node is not read-only). | |
1846 | */ | |
1847 | bool blk_supports_write_perm(BlockBackend *blk) | |
4be74634 | 1848 | { |
f21d96d0 KW |
1849 | BlockDriverState *bs = blk_bs(blk); |
1850 | ||
1851 | if (bs) { | |
86b1cf32 | 1852 | return !bdrv_is_read_only(bs); |
061959e8 | 1853 | } else { |
260242a8 | 1854 | return blk->root_state.open_flags & BDRV_O_RDWR; |
061959e8 | 1855 | } |
4be74634 MA |
1856 | } |
1857 | ||
86b1cf32 KW |
1858 | /* |
1859 | * Returns true if the BlockBackend can be written to in its current | |
1860 | * configuration (i.e. if write permission have been requested) | |
1861 | */ | |
1862 | bool blk_is_writable(BlockBackend *blk) | |
1863 | { | |
1864 | return blk->perm & BLK_PERM_WRITE; | |
1865 | } | |
1866 | ||
96710565 | 1867 | bool blk_is_sg(BlockBackend *blk) |
4be74634 | 1868 | { |
f21d96d0 KW |
1869 | BlockDriverState *bs = blk_bs(blk); |
1870 | ||
1871 | if (!bs) { | |
96710565 | 1872 | return false; |
a46fc9c9 HR |
1873 | } |
1874 | ||
f21d96d0 | 1875 | return bdrv_is_sg(bs); |
4be74634 MA |
1876 | } |
1877 | ||
96710565 | 1878 | bool blk_enable_write_cache(BlockBackend *blk) |
4be74634 | 1879 | { |
bfd18d1e | 1880 | return blk->enable_write_cache; |
4be74634 MA |
1881 | } |
1882 | ||
1883 | void blk_set_enable_write_cache(BlockBackend *blk, bool wce) | |
1884 | { | |
bfd18d1e | 1885 | blk->enable_write_cache = wce; |
4be74634 MA |
1886 | } |
1887 | ||
2bb0dce7 HR |
1888 | void blk_invalidate_cache(BlockBackend *blk, Error **errp) |
1889 | { | |
f21d96d0 KW |
1890 | BlockDriverState *bs = blk_bs(blk); |
1891 | ||
1892 | if (!bs) { | |
c09ba36c HR |
1893 | error_setg(errp, "Device '%s' has no medium", blk->name); |
1894 | return; | |
1895 | } | |
1896 | ||
f21d96d0 | 1897 | bdrv_invalidate_cache(bs, errp); |
2bb0dce7 HR |
1898 | } |
1899 | ||
e031f750 | 1900 | bool blk_is_inserted(BlockBackend *blk) |
4be74634 | 1901 | { |
f21d96d0 KW |
1902 | BlockDriverState *bs = blk_bs(blk); |
1903 | ||
1904 | return bs && bdrv_is_inserted(bs); | |
db0284f8 HR |
1905 | } |
1906 | ||
1907 | bool blk_is_available(BlockBackend *blk) | |
1908 | { | |
1909 | return blk_is_inserted(blk) && !blk_dev_is_tray_open(blk); | |
4be74634 MA |
1910 | } |
1911 | ||
1912 | void blk_lock_medium(BlockBackend *blk, bool locked) | |
1913 | { | |
f21d96d0 KW |
1914 | BlockDriverState *bs = blk_bs(blk); |
1915 | ||
1916 | if (bs) { | |
1917 | bdrv_lock_medium(bs, locked); | |
a46fc9c9 | 1918 | } |
4be74634 MA |
1919 | } |
1920 | ||
1921 | void blk_eject(BlockBackend *blk, bool eject_flag) | |
1922 | { | |
f21d96d0 | 1923 | BlockDriverState *bs = blk_bs(blk); |
2d76e724 KW |
1924 | char *id; |
1925 | ||
f21d96d0 KW |
1926 | if (bs) { |
1927 | bdrv_eject(bs, eject_flag); | |
a46fc9c9 | 1928 | } |
c47ee043 JS |
1929 | |
1930 | /* Whether or not we ejected on the backend, | |
1931 | * the frontend experienced a tray event. */ | |
1932 | id = blk_get_attached_dev_id(blk); | |
1933 | qapi_event_send_device_tray_moved(blk_name(blk), id, | |
3ab72385 | 1934 | eject_flag); |
c47ee043 | 1935 | g_free(id); |
4be74634 MA |
1936 | } |
1937 | ||
1938 | int blk_get_flags(BlockBackend *blk) | |
1939 | { | |
f21d96d0 KW |
1940 | BlockDriverState *bs = blk_bs(blk); |
1941 | ||
1942 | if (bs) { | |
1943 | return bdrv_get_flags(bs); | |
061959e8 HR |
1944 | } else { |
1945 | return blk->root_state.open_flags; | |
1946 | } | |
4be74634 MA |
1947 | } |
1948 | ||
4841211e EB |
1949 | /* Returns the minimum request alignment, in bytes; guaranteed nonzero */ |
1950 | uint32_t blk_get_request_alignment(BlockBackend *blk) | |
1951 | { | |
1952 | BlockDriverState *bs = blk_bs(blk); | |
1953 | return bs ? bs->bl.request_alignment : BDRV_SECTOR_SIZE; | |
1954 | } | |
1955 | ||
5def6b80 EB |
1956 | /* Returns the maximum transfer length, in bytes; guaranteed nonzero */ |
1957 | uint32_t blk_get_max_transfer(BlockBackend *blk) | |
454057b7 | 1958 | { |
f21d96d0 | 1959 | BlockDriverState *bs = blk_bs(blk); |
5def6b80 | 1960 | uint32_t max = 0; |
f21d96d0 KW |
1961 | |
1962 | if (bs) { | |
5def6b80 | 1963 | max = bs->bl.max_transfer; |
a46fc9c9 | 1964 | } |
5def6b80 | 1965 | return MIN_NON_ZERO(max, INT_MAX); |
454057b7 PL |
1966 | } |
1967 | ||
648296e0 SH |
1968 | int blk_get_max_iov(BlockBackend *blk) |
1969 | { | |
f21d96d0 | 1970 | return blk->root->bs->bl.max_iov; |
648296e0 SH |
1971 | } |
1972 | ||
4be74634 MA |
1973 | void blk_set_guest_block_size(BlockBackend *blk, int align) |
1974 | { | |
68e9ec01 | 1975 | blk->guest_block_size = align; |
4be74634 MA |
1976 | } |
1977 | ||
f1c17521 PB |
1978 | void *blk_try_blockalign(BlockBackend *blk, size_t size) |
1979 | { | |
f21d96d0 | 1980 | return qemu_try_blockalign(blk ? blk_bs(blk) : NULL, size); |
f1c17521 PB |
1981 | } |
1982 | ||
4be74634 MA |
1983 | void *blk_blockalign(BlockBackend *blk, size_t size) |
1984 | { | |
f21d96d0 | 1985 | return qemu_blockalign(blk ? blk_bs(blk) : NULL, size); |
4be74634 MA |
1986 | } |
1987 | ||
1988 | bool blk_op_is_blocked(BlockBackend *blk, BlockOpType op, Error **errp) | |
1989 | { | |
f21d96d0 KW |
1990 | BlockDriverState *bs = blk_bs(blk); |
1991 | ||
1992 | if (!bs) { | |
a46fc9c9 HR |
1993 | return false; |
1994 | } | |
1995 | ||
f21d96d0 | 1996 | return bdrv_op_is_blocked(bs, op, errp); |
4be74634 MA |
1997 | } |
1998 | ||
1999 | void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason) | |
2000 | { | |
f21d96d0 KW |
2001 | BlockDriverState *bs = blk_bs(blk); |
2002 | ||
2003 | if (bs) { | |
2004 | bdrv_op_unblock(bs, op, reason); | |
a46fc9c9 | 2005 | } |
4be74634 MA |
2006 | } |
2007 | ||
2008 | void blk_op_block_all(BlockBackend *blk, Error *reason) | |
2009 | { | |
f21d96d0 KW |
2010 | BlockDriverState *bs = blk_bs(blk); |
2011 | ||
2012 | if (bs) { | |
2013 | bdrv_op_block_all(bs, reason); | |
a46fc9c9 | 2014 | } |
4be74634 MA |
2015 | } |
2016 | ||
2017 | void blk_op_unblock_all(BlockBackend *blk, Error *reason) | |
2018 | { | |
f21d96d0 KW |
2019 | BlockDriverState *bs = blk_bs(blk); |
2020 | ||
2021 | if (bs) { | |
2022 | bdrv_op_unblock_all(bs, reason); | |
a46fc9c9 | 2023 | } |
4be74634 MA |
2024 | } |
2025 | ||
2026 | AioContext *blk_get_aio_context(BlockBackend *blk) | |
2027 | { | |
d861ab3a KW |
2028 | BlockDriverState *bs = blk_bs(blk); |
2029 | ||
d861ab3a | 2030 | if (bs) { |
132ada80 KW |
2031 | AioContext *ctx = bdrv_get_aio_context(blk_bs(blk)); |
2032 | assert(ctx == blk->ctx); | |
d861ab3a KW |
2033 | } |
2034 | ||
2035 | return blk->ctx; | |
4981bdec HR |
2036 | } |
2037 | ||
2038 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) | |
2039 | { | |
2040 | BlockBackendAIOCB *blk_acb = DO_UPCAST(BlockBackendAIOCB, common, acb); | |
2041 | return blk_get_aio_context(blk_acb->blk); | |
4be74634 MA |
2042 | } |
2043 | ||
97896a48 KW |
2044 | static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2045 | bool update_root_node, Error **errp) | |
4be74634 | 2046 | { |
f21d96d0 | 2047 | BlockDriverState *bs = blk_bs(blk); |
c61791fc | 2048 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
97896a48 | 2049 | int ret; |
f21d96d0 KW |
2050 | |
2051 | if (bs) { | |
97896a48 KW |
2052 | if (update_root_node) { |
2053 | ret = bdrv_child_try_set_aio_context(bs, new_context, blk->root, | |
2054 | errp); | |
2055 | if (ret < 0) { | |
2056 | return ret; | |
2057 | } | |
2058 | } | |
c61791fc | 2059 | if (tgm->throttle_state) { |
dc868fb0 | 2060 | bdrv_drained_begin(bs); |
c61791fc MP |
2061 | throttle_group_detach_aio_context(tgm); |
2062 | throttle_group_attach_aio_context(tgm, new_context); | |
dc868fb0 | 2063 | bdrv_drained_end(bs); |
7ca7f0f6 | 2064 | } |
a46fc9c9 | 2065 | } |
97896a48 | 2066 | |
d861ab3a | 2067 | blk->ctx = new_context; |
97896a48 | 2068 | return 0; |
4be74634 MA |
2069 | } |
2070 | ||
97896a48 KW |
2071 | int blk_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2072 | Error **errp) | |
38475269 | 2073 | { |
97896a48 | 2074 | return blk_do_set_aio_context(blk, new_context, true, errp); |
38475269 KW |
2075 | } |
2076 | ||
2077 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2078 | GSList **ignore, Error **errp) | |
2079 | { | |
2080 | BlockBackend *blk = child->opaque; | |
2081 | ||
980b0f94 KW |
2082 | if (blk->allow_aio_context_change) { |
2083 | return true; | |
2084 | } | |
2085 | ||
38475269 KW |
2086 | /* Only manually created BlockBackends that are not attached to anything |
2087 | * can change their AioContext without updating their user. */ | |
2088 | if (!blk->name || blk->dev) { | |
2089 | /* TODO Add BB name/QOM path */ | |
2090 | error_setg(errp, "Cannot change iothread of active block backend"); | |
2091 | return false; | |
2092 | } | |
2093 | ||
2094 | return true; | |
2095 | } | |
2096 | ||
2097 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2098 | GSList **ignore) | |
2099 | { | |
2100 | BlockBackend *blk = child->opaque; | |
97896a48 | 2101 | blk_do_set_aio_context(blk, ctx, false, &error_abort); |
38475269 KW |
2102 | } |
2103 | ||
2019ba0a HR |
2104 | void blk_add_aio_context_notifier(BlockBackend *blk, |
2105 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
2106 | void (*detach_aio_context)(void *opaque), void *opaque) | |
2107 | { | |
d03654ea | 2108 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2109 | BlockDriverState *bs = blk_bs(blk); |
2110 | ||
d03654ea SH |
2111 | notifier = g_new(BlockBackendAioNotifier, 1); |
2112 | notifier->attached_aio_context = attached_aio_context; | |
2113 | notifier->detach_aio_context = detach_aio_context; | |
2114 | notifier->opaque = opaque; | |
2115 | QLIST_INSERT_HEAD(&blk->aio_notifiers, notifier, list); | |
2116 | ||
f21d96d0 KW |
2117 | if (bs) { |
2118 | bdrv_add_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2119 | detach_aio_context, opaque); |
2120 | } | |
2019ba0a HR |
2121 | } |
2122 | ||
2123 | void blk_remove_aio_context_notifier(BlockBackend *blk, | |
2124 | void (*attached_aio_context)(AioContext *, | |
2125 | void *), | |
2126 | void (*detach_aio_context)(void *), | |
2127 | void *opaque) | |
2128 | { | |
d03654ea | 2129 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2130 | BlockDriverState *bs = blk_bs(blk); |
2131 | ||
2132 | if (bs) { | |
2133 | bdrv_remove_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2134 | detach_aio_context, opaque); |
2135 | } | |
d03654ea SH |
2136 | |
2137 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
2138 | if (notifier->attached_aio_context == attached_aio_context && | |
2139 | notifier->detach_aio_context == detach_aio_context && | |
2140 | notifier->opaque == opaque) { | |
2141 | QLIST_REMOVE(notifier, list); | |
2142 | g_free(notifier); | |
2143 | return; | |
2144 | } | |
2145 | } | |
2146 | ||
2147 | abort(); | |
2019ba0a HR |
2148 | } |
2149 | ||
3301f6c6 HR |
2150 | void blk_add_remove_bs_notifier(BlockBackend *blk, Notifier *notify) |
2151 | { | |
2152 | notifier_list_add(&blk->remove_bs_notifiers, notify); | |
2153 | } | |
2154 | ||
2155 | void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify) | |
2156 | { | |
2157 | notifier_list_add(&blk->insert_bs_notifiers, notify); | |
2158 | } | |
2159 | ||
4be74634 MA |
2160 | void blk_io_plug(BlockBackend *blk) |
2161 | { | |
f21d96d0 KW |
2162 | BlockDriverState *bs = blk_bs(blk); |
2163 | ||
2164 | if (bs) { | |
2165 | bdrv_io_plug(bs); | |
a46fc9c9 | 2166 | } |
4be74634 MA |
2167 | } |
2168 | ||
2169 | void blk_io_unplug(BlockBackend *blk) | |
2170 | { | |
f21d96d0 KW |
2171 | BlockDriverState *bs = blk_bs(blk); |
2172 | ||
2173 | if (bs) { | |
2174 | bdrv_io_unplug(bs); | |
a46fc9c9 | 2175 | } |
4be74634 MA |
2176 | } |
2177 | ||
2178 | BlockAcctStats *blk_get_stats(BlockBackend *blk) | |
2179 | { | |
7f0e9da6 | 2180 | return &blk->stats; |
4be74634 MA |
2181 | } |
2182 | ||
2183 | void *blk_aio_get(const AIOCBInfo *aiocb_info, BlockBackend *blk, | |
2184 | BlockCompletionFunc *cb, void *opaque) | |
2185 | { | |
2186 | return qemu_aio_get(aiocb_info, blk_bs(blk), cb, opaque); | |
2187 | } | |
1ef01253 | 2188 | |
d004bd52 | 2189 | int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 2190 | int bytes, BdrvRequestFlags flags) |
1ef01253 | 2191 | { |
f5a5ca79 | 2192 | return blk_co_pwritev(blk, offset, bytes, NULL, |
16aaf975 | 2193 | flags | BDRV_REQ_ZERO_WRITE); |
1ef01253 HR |
2194 | } |
2195 | ||
fe5c1355 PB |
2196 | int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, |
2197 | int count) | |
1ef01253 | 2198 | { |
35fadca8 PB |
2199 | return blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
2200 | BDRV_REQ_WRITE_COMPRESSED); | |
1ef01253 HR |
2201 | } |
2202 | ||
c80d8b06 | 2203 | int blk_truncate(BlockBackend *blk, int64_t offset, bool exact, |
8c6242b6 | 2204 | PreallocMode prealloc, BdrvRequestFlags flags, Error **errp) |
1ef01253 | 2205 | { |
c09ba36c | 2206 | if (!blk_is_available(blk)) { |
ed3d2ec9 | 2207 | error_setg(errp, "No medium inserted"); |
c09ba36c HR |
2208 | return -ENOMEDIUM; |
2209 | } | |
2210 | ||
8c6242b6 | 2211 | return bdrv_truncate(blk->root, offset, exact, prealloc, flags, errp); |
1ef01253 HR |
2212 | } |
2213 | ||
1ef01253 HR |
2214 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, |
2215 | int64_t pos, int size) | |
2216 | { | |
bfd18d1e KW |
2217 | int ret; |
2218 | ||
c09ba36c HR |
2219 | if (!blk_is_available(blk)) { |
2220 | return -ENOMEDIUM; | |
2221 | } | |
2222 | ||
bfd18d1e KW |
2223 | ret = bdrv_save_vmstate(blk_bs(blk), buf, pos, size); |
2224 | if (ret < 0) { | |
2225 | return ret; | |
2226 | } | |
2227 | ||
2228 | if (ret == size && !blk->enable_write_cache) { | |
2229 | ret = bdrv_flush(blk_bs(blk)); | |
2230 | } | |
2231 | ||
2232 | return ret < 0 ? ret : size; | |
1ef01253 HR |
2233 | } |
2234 | ||
2235 | int blk_load_vmstate(BlockBackend *blk, uint8_t *buf, int64_t pos, int size) | |
2236 | { | |
c09ba36c HR |
2237 | if (!blk_is_available(blk)) { |
2238 | return -ENOMEDIUM; | |
2239 | } | |
2240 | ||
f21d96d0 | 2241 | return bdrv_load_vmstate(blk_bs(blk), buf, pos, size); |
1ef01253 | 2242 | } |
f0272c4d ET |
2243 | |
2244 | int blk_probe_blocksizes(BlockBackend *blk, BlockSizes *bsz) | |
2245 | { | |
c09ba36c HR |
2246 | if (!blk_is_available(blk)) { |
2247 | return -ENOMEDIUM; | |
2248 | } | |
2249 | ||
f21d96d0 | 2250 | return bdrv_probe_blocksizes(blk_bs(blk), bsz); |
f0272c4d ET |
2251 | } |
2252 | ||
2253 | int blk_probe_geometry(BlockBackend *blk, HDGeometry *geo) | |
2254 | { | |
c09ba36c HR |
2255 | if (!blk_is_available(blk)) { |
2256 | return -ENOMEDIUM; | |
2257 | } | |
2258 | ||
f21d96d0 | 2259 | return bdrv_probe_geometry(blk_bs(blk), geo); |
f0272c4d | 2260 | } |
281d22d8 HR |
2261 | |
2262 | /* | |
2263 | * Updates the BlockBackendRootState object with data from the currently | |
2264 | * attached BlockDriverState. | |
2265 | */ | |
2266 | void blk_update_root_state(BlockBackend *blk) | |
2267 | { | |
f21d96d0 | 2268 | assert(blk->root); |
281d22d8 | 2269 | |
f21d96d0 | 2270 | blk->root_state.open_flags = blk->root->bs->open_flags; |
f21d96d0 | 2271 | blk->root_state.detect_zeroes = blk->root->bs->detect_zeroes; |
281d22d8 HR |
2272 | } |
2273 | ||
38cb18f5 | 2274 | /* |
b85114f8 KW |
2275 | * Returns the detect-zeroes setting to be used for bdrv_open() of a |
2276 | * BlockDriverState which is supposed to inherit the root state. | |
38cb18f5 | 2277 | */ |
b85114f8 | 2278 | bool blk_get_detect_zeroes_from_root_state(BlockBackend *blk) |
38cb18f5 | 2279 | { |
b85114f8 | 2280 | return blk->root_state.detect_zeroes; |
38cb18f5 HR |
2281 | } |
2282 | ||
2283 | /* | |
2284 | * Returns the flags to be used for bdrv_open() of a BlockDriverState which is | |
2285 | * supposed to inherit the root state. | |
2286 | */ | |
2287 | int blk_get_open_flags_from_root_state(BlockBackend *blk) | |
2288 | { | |
260242a8 | 2289 | return blk->root_state.open_flags; |
38cb18f5 HR |
2290 | } |
2291 | ||
281d22d8 HR |
2292 | BlockBackendRootState *blk_get_root_state(BlockBackend *blk) |
2293 | { | |
2294 | return &blk->root_state; | |
2295 | } | |
1393f212 HR |
2296 | |
2297 | int blk_commit_all(void) | |
2298 | { | |
fe1a9cbc HR |
2299 | BlockBackend *blk = NULL; |
2300 | ||
2301 | while ((blk = blk_all_next(blk)) != NULL) { | |
2302 | AioContext *aio_context = blk_get_aio_context(blk); | |
9a71b9de | 2303 | BlockDriverState *unfiltered_bs = bdrv_skip_filters(blk_bs(blk)); |
fe1a9cbc HR |
2304 | |
2305 | aio_context_acquire(aio_context); | |
9a71b9de HR |
2306 | if (blk_is_inserted(blk) && bdrv_cow_child(unfiltered_bs)) { |
2307 | int ret; | |
2308 | ||
2309 | ret = bdrv_commit(unfiltered_bs); | |
fe1a9cbc HR |
2310 | if (ret < 0) { |
2311 | aio_context_release(aio_context); | |
2312 | return ret; | |
2313 | } | |
2314 | } | |
2315 | aio_context_release(aio_context); | |
2316 | } | |
2317 | return 0; | |
2318 | } | |
2319 | ||
97148076 KW |
2320 | |
2321 | /* throttling disk I/O limits */ | |
2322 | void blk_set_io_limits(BlockBackend *blk, ThrottleConfig *cfg) | |
2323 | { | |
022cdc9f | 2324 | throttle_group_config(&blk->public.throttle_group_member, cfg); |
97148076 KW |
2325 | } |
2326 | ||
2327 | void blk_io_limits_disable(BlockBackend *blk) | |
2328 | { | |
48bf7ea8 AG |
2329 | BlockDriverState *bs = blk_bs(blk); |
2330 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; | |
2331 | assert(tgm->throttle_state); | |
2332 | if (bs) { | |
2333 | bdrv_drained_begin(bs); | |
2334 | } | |
2335 | throttle_group_unregister_tgm(tgm); | |
2336 | if (bs) { | |
2337 | bdrv_drained_end(bs); | |
2338 | } | |
97148076 KW |
2339 | } |
2340 | ||
2341 | /* should be called before blk_set_io_limits if a limit is set */ | |
2342 | void blk_io_limits_enable(BlockBackend *blk, const char *group) | |
2343 | { | |
022cdc9f | 2344 | assert(!blk->public.throttle_group_member.throttle_state); |
c61791fc MP |
2345 | throttle_group_register_tgm(&blk->public.throttle_group_member, |
2346 | group, blk_get_aio_context(blk)); | |
97148076 KW |
2347 | } |
2348 | ||
2349 | void blk_io_limits_update_group(BlockBackend *blk, const char *group) | |
2350 | { | |
2351 | /* this BB is not part of any group */ | |
022cdc9f | 2352 | if (!blk->public.throttle_group_member.throttle_state) { |
97148076 KW |
2353 | return; |
2354 | } | |
2355 | ||
2356 | /* this BB is a part of the same group than the one we want */ | |
022cdc9f MP |
2357 | if (!g_strcmp0(throttle_group_get_name(&blk->public.throttle_group_member), |
2358 | group)) { | |
97148076 KW |
2359 | return; |
2360 | } | |
2361 | ||
2362 | /* need to change the group this bs belong to */ | |
2363 | blk_io_limits_disable(blk); | |
2364 | blk_io_limits_enable(blk, group); | |
2365 | } | |
c2066af0 KW |
2366 | |
2367 | static void blk_root_drained_begin(BdrvChild *child) | |
2368 | { | |
2369 | BlockBackend *blk = child->opaque; | |
d73415a3 | 2370 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
c2066af0 | 2371 | |
f4d9cc88 JS |
2372 | if (++blk->quiesce_counter == 1) { |
2373 | if (blk->dev_ops && blk->dev_ops->drained_begin) { | |
2374 | blk->dev_ops->drained_begin(blk->dev_opaque); | |
2375 | } | |
2376 | } | |
2377 | ||
36fe1331 KW |
2378 | /* Note that blk->root may not be accessible here yet if we are just |
2379 | * attaching to a BlockDriverState that is drained. Use child instead. */ | |
2380 | ||
d73415a3 SH |
2381 | if (qatomic_fetch_inc(&tgm->io_limits_disabled) == 0) { |
2382 | throttle_group_restart_tgm(tgm); | |
c2066af0 KW |
2383 | } |
2384 | } | |
2385 | ||
fe5258a5 KW |
2386 | static bool blk_root_drained_poll(BdrvChild *child) |
2387 | { | |
2388 | BlockBackend *blk = child->opaque; | |
095cc4d0 | 2389 | bool busy = false; |
fe5258a5 | 2390 | assert(blk->quiesce_counter); |
095cc4d0 SL |
2391 | |
2392 | if (blk->dev_ops && blk->dev_ops->drained_poll) { | |
2393 | busy = blk->dev_ops->drained_poll(blk->dev_opaque); | |
2394 | } | |
2395 | return busy || !!blk->in_flight; | |
fe5258a5 KW |
2396 | } |
2397 | ||
e037c09c | 2398 | static void blk_root_drained_end(BdrvChild *child, int *drained_end_counter) |
c2066af0 KW |
2399 | { |
2400 | BlockBackend *blk = child->opaque; | |
f4d9cc88 | 2401 | assert(blk->quiesce_counter); |
c2066af0 | 2402 | |
022cdc9f | 2403 | assert(blk->public.throttle_group_member.io_limits_disabled); |
d73415a3 | 2404 | qatomic_dec(&blk->public.throttle_group_member.io_limits_disabled); |
f4d9cc88 JS |
2405 | |
2406 | if (--blk->quiesce_counter == 0) { | |
2407 | if (blk->dev_ops && blk->dev_ops->drained_end) { | |
2408 | blk->dev_ops->drained_end(blk->dev_opaque); | |
2409 | } | |
cf312932 KW |
2410 | while (qemu_co_enter_next(&blk->queued_requests, NULL)) { |
2411 | /* Resume all queued requests */ | |
2412 | } | |
f4d9cc88 | 2413 | } |
c2066af0 | 2414 | } |
23d0ba93 FZ |
2415 | |
2416 | void blk_register_buf(BlockBackend *blk, void *host, size_t size) | |
2417 | { | |
2418 | bdrv_register_buf(blk_bs(blk), host, size); | |
2419 | } | |
2420 | ||
2421 | void blk_unregister_buf(BlockBackend *blk, void *host) | |
2422 | { | |
2423 | bdrv_unregister_buf(blk_bs(blk), host); | |
2424 | } | |
b5679fa4 FZ |
2425 | |
2426 | int coroutine_fn blk_co_copy_range(BlockBackend *blk_in, int64_t off_in, | |
2427 | BlockBackend *blk_out, int64_t off_out, | |
67b51fb9 VSO |
2428 | int bytes, BdrvRequestFlags read_flags, |
2429 | BdrvRequestFlags write_flags) | |
b5679fa4 FZ |
2430 | { |
2431 | int r; | |
2432 | r = blk_check_byte_request(blk_in, off_in, bytes); | |
2433 | if (r) { | |
2434 | return r; | |
2435 | } | |
2436 | r = blk_check_byte_request(blk_out, off_out, bytes); | |
2437 | if (r) { | |
2438 | return r; | |
2439 | } | |
2440 | return bdrv_co_copy_range(blk_in->root, off_in, | |
2441 | blk_out->root, off_out, | |
67b51fb9 | 2442 | bytes, read_flags, write_flags); |
b5679fa4 | 2443 | } |
5d3b4e99 VSO |
2444 | |
2445 | const BdrvChild *blk_root(BlockBackend *blk) | |
2446 | { | |
2447 | return blk->root; | |
2448 | } | |
2b7bbdbd HR |
2449 | |
2450 | int blk_make_empty(BlockBackend *blk, Error **errp) | |
2451 | { | |
2452 | if (!blk_is_available(blk)) { | |
2453 | error_setg(errp, "No medium inserted"); | |
2454 | return -ENOMEDIUM; | |
2455 | } | |
2456 | ||
2457 | return bdrv_make_empty(blk->root, errp); | |
2458 | } |