]>
Commit | Line | Data |
---|---|---|
26f54e9a MA |
1 | /* |
2 | * QEMU Block backends | |
3 | * | |
60cb2fa7 | 4 | * Copyright (C) 2014-2016 Red Hat, Inc. |
26f54e9a MA |
5 | * |
6 | * Authors: | |
7 | * Markus Armbruster <armbru@redhat.com>, | |
8 | * | |
9 | * This work is licensed under the terms of the GNU LGPL, version 2.1 | |
10 | * or later. See the COPYING.LIB file in the top-level directory. | |
11 | */ | |
12 | ||
80c71a24 | 13 | #include "qemu/osdep.h" |
26f54e9a MA |
14 | #include "sysemu/block-backend.h" |
15 | #include "block/block_int.h" | |
373340b2 | 16 | #include "block/blockjob.h" |
281d22d8 | 17 | #include "block/throttle-groups.h" |
13d4ff07 | 18 | #include "hw/qdev-core.h" |
18e46a03 | 19 | #include "sysemu/blockdev.h" |
54d31236 | 20 | #include "sysemu/runstate.h" |
e4ec5ad4 PD |
21 | #include "sysemu/sysemu.h" |
22 | #include "sysemu/replay.h" | |
e688df6b | 23 | #include "qapi/error.h" |
9af23989 | 24 | #include "qapi/qapi-events-block.h" |
f348b6d1 | 25 | #include "qemu/id.h" |
db725815 | 26 | #include "qemu/main-loop.h" |
922a01a0 | 27 | #include "qemu/option.h" |
1e98fefd | 28 | #include "trace.h" |
5f7772c4 | 29 | #include "migration/misc.h" |
a7f53e26 MA |
30 | |
31 | /* Number of coroutines to reserve per attached device model */ | |
32 | #define COROUTINE_POOL_RESERVATION 64 | |
26f54e9a | 33 | |
1bf1cbc9 KW |
34 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
35 | ||
4981bdec HR |
36 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb); |
37 | ||
d03654ea SH |
38 | typedef struct BlockBackendAioNotifier { |
39 | void (*attached_aio_context)(AioContext *new_context, void *opaque); | |
40 | void (*detach_aio_context)(void *opaque); | |
41 | void *opaque; | |
42 | QLIST_ENTRY(BlockBackendAioNotifier) list; | |
43 | } BlockBackendAioNotifier; | |
44 | ||
26f54e9a MA |
45 | struct BlockBackend { |
46 | char *name; | |
47 | int refcnt; | |
f21d96d0 | 48 | BdrvChild *root; |
d861ab3a | 49 | AioContext *ctx; |
26f8b3a8 | 50 | DriveInfo *legacy_dinfo; /* null unless created by drive_new() */ |
2cf22d6a | 51 | QTAILQ_ENTRY(BlockBackend) link; /* for block_backends */ |
9492b0b9 | 52 | QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */ |
f2cd875d | 53 | BlockBackendPublic public; |
a7f53e26 | 54 | |
d09ea2d2 | 55 | DeviceState *dev; /* attached device model, if any */ |
a7f53e26 MA |
56 | const BlockDevOps *dev_ops; |
57 | void *dev_opaque; | |
68e9ec01 HR |
58 | |
59 | /* the block size for which the guest device expects atomicity */ | |
60 | int guest_block_size; | |
7f0e9da6 | 61 | |
281d22d8 HR |
62 | /* If the BDS tree is removed, some of its options are stored here (which |
63 | * can be used to restore those options in the new BDS on insert) */ | |
64 | BlockBackendRootState root_state; | |
65 | ||
bfd18d1e KW |
66 | bool enable_write_cache; |
67 | ||
7f0e9da6 HR |
68 | /* I/O stats (display with "info blockstats"). */ |
69 | BlockAcctStats stats; | |
373340b2 HR |
70 | |
71 | BlockdevOnError on_read_error, on_write_error; | |
72 | bool iostatus_enabled; | |
73 | BlockDeviceIoStatus iostatus; | |
3301f6c6 | 74 | |
981776b3 KW |
75 | uint64_t perm; |
76 | uint64_t shared_perm; | |
d35ff5e6 | 77 | bool disable_perm; |
981776b3 | 78 | |
980b0f94 | 79 | bool allow_aio_context_change; |
c10c9d96 KW |
80 | bool allow_write_beyond_eof; |
81 | ||
3301f6c6 | 82 | NotifierList remove_bs_notifiers, insert_bs_notifiers; |
d03654ea | 83 | QLIST_HEAD(, BlockBackendAioNotifier) aio_notifiers; |
f4d9cc88 JS |
84 | |
85 | int quiesce_counter; | |
cf312932 KW |
86 | CoQueue queued_requests; |
87 | bool disable_request_queuing; | |
88 | ||
5f7772c4 | 89 | VMChangeStateEntry *vmsh; |
ca2e2144 | 90 | bool force_allow_inactivate; |
33f2a757 SH |
91 | |
92 | /* Number of in-flight aio requests. BlockDriverState also counts | |
93 | * in-flight requests but aio requests can exist even when blk->root is | |
94 | * NULL, so we cannot rely on its counter for that case. | |
95 | * Accessed with atomic ops. | |
96 | */ | |
97 | unsigned int in_flight; | |
26f54e9a MA |
98 | }; |
99 | ||
e7f7d676 HR |
100 | typedef struct BlockBackendAIOCB { |
101 | BlockAIOCB common; | |
4981bdec | 102 | BlockBackend *blk; |
e7f7d676 HR |
103 | int ret; |
104 | } BlockBackendAIOCB; | |
105 | ||
106 | static const AIOCBInfo block_backend_aiocb_info = { | |
4981bdec | 107 | .get_aio_context = blk_aiocb_get_aio_context, |
e7f7d676 HR |
108 | .aiocb_size = sizeof(BlockBackendAIOCB), |
109 | }; | |
110 | ||
8fb3c76c | 111 | static void drive_info_del(DriveInfo *dinfo); |
7c8eece4 | 112 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs); |
8fb3c76c | 113 | |
2cf22d6a HR |
114 | /* All BlockBackends */ |
115 | static QTAILQ_HEAD(, BlockBackend) block_backends = | |
116 | QTAILQ_HEAD_INITIALIZER(block_backends); | |
117 | ||
9492b0b9 HR |
118 | /* All BlockBackends referenced by the monitor and which are iterated through by |
119 | * blk_next() */ | |
120 | static QTAILQ_HEAD(, BlockBackend) monitor_block_backends = | |
121 | QTAILQ_HEAD_INITIALIZER(monitor_block_backends); | |
26f54e9a | 122 | |
f21d96d0 KW |
123 | static void blk_root_inherit_options(int *child_flags, QDict *child_options, |
124 | int parent_flags, QDict *parent_options) | |
125 | { | |
126 | /* We're not supposed to call this function for root nodes */ | |
127 | abort(); | |
128 | } | |
c2066af0 | 129 | static void blk_root_drained_begin(BdrvChild *child); |
fe5258a5 | 130 | static bool blk_root_drained_poll(BdrvChild *child); |
e037c09c | 131 | static void blk_root_drained_end(BdrvChild *child, int *drained_end_counter); |
f21d96d0 | 132 | |
5c8cab48 KW |
133 | static void blk_root_change_media(BdrvChild *child, bool load); |
134 | static void blk_root_resize(BdrvChild *child); | |
135 | ||
38475269 KW |
136 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, |
137 | GSList **ignore, Error **errp); | |
138 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
139 | GSList **ignore); | |
140 | ||
b5411555 KW |
141 | static char *blk_root_get_parent_desc(BdrvChild *child) |
142 | { | |
143 | BlockBackend *blk = child->opaque; | |
144 | char *dev_id; | |
145 | ||
146 | if (blk->name) { | |
147 | return g_strdup(blk->name); | |
148 | } | |
149 | ||
150 | dev_id = blk_get_attached_dev_id(blk); | |
151 | if (*dev_id) { | |
152 | return dev_id; | |
153 | } else { | |
154 | /* TODO Callback into the BB owner for something more detailed */ | |
155 | g_free(dev_id); | |
156 | return g_strdup("a block device"); | |
157 | } | |
158 | } | |
159 | ||
4c265bf9 KW |
160 | static const char *blk_root_get_name(BdrvChild *child) |
161 | { | |
162 | return blk_name(child->opaque); | |
163 | } | |
164 | ||
5f7772c4 FZ |
165 | static void blk_vm_state_changed(void *opaque, int running, RunState state) |
166 | { | |
167 | Error *local_err = NULL; | |
168 | BlockBackend *blk = opaque; | |
169 | ||
170 | if (state == RUN_STATE_INMIGRATE) { | |
171 | return; | |
172 | } | |
173 | ||
174 | qemu_del_vm_change_state_handler(blk->vmsh); | |
175 | blk->vmsh = NULL; | |
176 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); | |
177 | if (local_err) { | |
178 | error_report_err(local_err); | |
179 | } | |
180 | } | |
181 | ||
4417ab7a KW |
182 | /* |
183 | * Notifies the user of the BlockBackend that migration has completed. qdev | |
184 | * devices can tighten their permissions in response (specifically revoke | |
185 | * shared write permissions that we needed for storage migration). | |
186 | * | |
187 | * If an error is returned, the VM cannot be allowed to be resumed. | |
188 | */ | |
189 | static void blk_root_activate(BdrvChild *child, Error **errp) | |
190 | { | |
191 | BlockBackend *blk = child->opaque; | |
192 | Error *local_err = NULL; | |
193 | ||
194 | if (!blk->disable_perm) { | |
195 | return; | |
196 | } | |
197 | ||
198 | blk->disable_perm = false; | |
199 | ||
5f7772c4 FZ |
200 | blk_set_perm(blk, blk->perm, BLK_PERM_ALL, &local_err); |
201 | if (local_err) { | |
202 | error_propagate(errp, local_err); | |
203 | blk->disable_perm = true; | |
204 | return; | |
205 | } | |
206 | ||
207 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
208 | /* Activation can happen when migration process is still active, for | |
209 | * example when nbd_server_add is called during non-shared storage | |
210 | * migration. Defer the shared_perm update to migration completion. */ | |
211 | if (!blk->vmsh) { | |
212 | blk->vmsh = qemu_add_vm_change_state_handler(blk_vm_state_changed, | |
213 | blk); | |
214 | } | |
215 | return; | |
216 | } | |
217 | ||
4417ab7a KW |
218 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); |
219 | if (local_err) { | |
220 | error_propagate(errp, local_err); | |
221 | blk->disable_perm = true; | |
222 | return; | |
223 | } | |
224 | } | |
225 | ||
ca2e2144 FZ |
226 | void blk_set_force_allow_inactivate(BlockBackend *blk) |
227 | { | |
228 | blk->force_allow_inactivate = true; | |
229 | } | |
230 | ||
c16de8f5 FZ |
231 | static bool blk_can_inactivate(BlockBackend *blk) |
232 | { | |
ca2e2144 | 233 | /* If it is a guest device, inactivate is ok. */ |
c16de8f5 FZ |
234 | if (blk->dev || blk_name(blk)[0]) { |
235 | return true; | |
236 | } | |
237 | ||
ca2e2144 FZ |
238 | /* Inactivating means no more writes to the image can be done, |
239 | * even if those writes would be changes invisible to the | |
240 | * guest. For block job BBs that satisfy this, we can just allow | |
241 | * it. This is the case for mirror job source, which is required | |
242 | * by libvirt non-shared block migration. */ | |
243 | if (!(blk->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED))) { | |
244 | return true; | |
245 | } | |
246 | ||
247 | return blk->force_allow_inactivate; | |
c16de8f5 FZ |
248 | } |
249 | ||
cfa1a572 KW |
250 | static int blk_root_inactivate(BdrvChild *child) |
251 | { | |
252 | BlockBackend *blk = child->opaque; | |
253 | ||
254 | if (blk->disable_perm) { | |
255 | return 0; | |
256 | } | |
257 | ||
c16de8f5 | 258 | if (!blk_can_inactivate(blk)) { |
cfa1a572 KW |
259 | return -EPERM; |
260 | } | |
261 | ||
262 | blk->disable_perm = true; | |
263 | if (blk->root) { | |
264 | bdrv_child_try_set_perm(blk->root, 0, BLK_PERM_ALL, &error_abort); | |
265 | } | |
266 | ||
267 | return 0; | |
268 | } | |
269 | ||
d03654ea SH |
270 | static void blk_root_attach(BdrvChild *child) |
271 | { | |
272 | BlockBackend *blk = child->opaque; | |
273 | BlockBackendAioNotifier *notifier; | |
274 | ||
275 | trace_blk_root_attach(child, blk, child->bs); | |
276 | ||
277 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
278 | bdrv_add_aio_context_notifier(child->bs, | |
279 | notifier->attached_aio_context, | |
280 | notifier->detach_aio_context, | |
281 | notifier->opaque); | |
282 | } | |
283 | } | |
284 | ||
285 | static void blk_root_detach(BdrvChild *child) | |
286 | { | |
287 | BlockBackend *blk = child->opaque; | |
288 | BlockBackendAioNotifier *notifier; | |
289 | ||
290 | trace_blk_root_detach(child, blk, child->bs); | |
291 | ||
292 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
293 | bdrv_remove_aio_context_notifier(child->bs, | |
294 | notifier->attached_aio_context, | |
295 | notifier->detach_aio_context, | |
296 | notifier->opaque); | |
297 | } | |
298 | } | |
299 | ||
f21d96d0 | 300 | static const BdrvChildRole child_root = { |
c2066af0 KW |
301 | .inherit_options = blk_root_inherit_options, |
302 | ||
5c8cab48 KW |
303 | .change_media = blk_root_change_media, |
304 | .resize = blk_root_resize, | |
4c265bf9 | 305 | .get_name = blk_root_get_name, |
b5411555 | 306 | .get_parent_desc = blk_root_get_parent_desc, |
5c8cab48 | 307 | |
c2066af0 | 308 | .drained_begin = blk_root_drained_begin, |
fe5258a5 | 309 | .drained_poll = blk_root_drained_poll, |
c2066af0 | 310 | .drained_end = blk_root_drained_end, |
4417ab7a KW |
311 | |
312 | .activate = blk_root_activate, | |
cfa1a572 | 313 | .inactivate = blk_root_inactivate, |
d03654ea SH |
314 | |
315 | .attach = blk_root_attach, | |
316 | .detach = blk_root_detach, | |
38475269 KW |
317 | |
318 | .can_set_aio_ctx = blk_root_can_set_aio_ctx, | |
319 | .set_aio_ctx = blk_root_set_aio_ctx, | |
f21d96d0 KW |
320 | }; |
321 | ||
26f54e9a | 322 | /* |
efaa7c4e | 323 | * Create a new BlockBackend with a reference count of one. |
6d0eb64d KW |
324 | * |
325 | * @perm is a bitmasks of BLK_PERM_* constants which describes the permissions | |
326 | * to request for a block driver node that is attached to this BlockBackend. | |
327 | * @shared_perm is a bitmask which describes which permissions may be granted | |
328 | * to other users of the attached node. | |
329 | * Both sets of permissions can be changed later using blk_set_perm(). | |
330 | * | |
26f54e9a MA |
331 | * Return the new BlockBackend on success, null on failure. |
332 | */ | |
d861ab3a | 333 | BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm) |
26f54e9a MA |
334 | { |
335 | BlockBackend *blk; | |
336 | ||
26f54e9a | 337 | blk = g_new0(BlockBackend, 1); |
26f54e9a | 338 | blk->refcnt = 1; |
d861ab3a | 339 | blk->ctx = ctx; |
6d0eb64d KW |
340 | blk->perm = perm; |
341 | blk->shared_perm = shared_perm; | |
0c3169df KW |
342 | blk_set_enable_write_cache(blk, true); |
343 | ||
cb53460b KW |
344 | blk->on_read_error = BLOCKDEV_ON_ERROR_REPORT; |
345 | blk->on_write_error = BLOCKDEV_ON_ERROR_ENOSPC; | |
346 | ||
9caa6f3d | 347 | block_acct_init(&blk->stats); |
27ccdd52 | 348 | |
cf312932 | 349 | qemu_co_queue_init(&blk->queued_requests); |
3301f6c6 HR |
350 | notifier_list_init(&blk->remove_bs_notifiers); |
351 | notifier_list_init(&blk->insert_bs_notifiers); | |
d03654ea | 352 | QLIST_INIT(&blk->aio_notifiers); |
27ccdd52 | 353 | |
2cf22d6a | 354 | QTAILQ_INSERT_TAIL(&block_backends, blk, link); |
26f54e9a MA |
355 | return blk; |
356 | } | |
357 | ||
7e7d56d9 | 358 | /* |
28eb9b12 | 359 | * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. |
d861ab3a | 360 | * The new BlockBackend is in the main AioContext. |
ca49a4fd HR |
361 | * |
362 | * Just as with bdrv_open(), after having called this function the reference to | |
363 | * @options belongs to the block layer (even on failure). | |
364 | * | |
365 | * TODO: Remove @filename and @flags; it should be possible to specify a whole | |
366 | * BDS tree just by specifying the @options QDict (or @reference, | |
367 | * alternatively). At the time of adding this function, this is not possible, | |
368 | * though, so callers of this function have to be able to specify @filename and | |
369 | * @flags. | |
370 | */ | |
efaa7c4e HR |
371 | BlockBackend *blk_new_open(const char *filename, const char *reference, |
372 | QDict *options, int flags, Error **errp) | |
ca49a4fd HR |
373 | { |
374 | BlockBackend *blk; | |
28eb9b12 | 375 | BlockDriverState *bs; |
1f4ad7d3 | 376 | uint64_t perm = 0; |
c62d32f5 KW |
377 | |
378 | /* blk_new_open() is mainly used in .bdrv_create implementations and the | |
379 | * tools where sharing isn't a concern because the BDS stays private, so we | |
380 | * just request permission according to the flags. | |
381 | * | |
382 | * The exceptions are xen_disk and blockdev_init(); in these cases, the | |
383 | * caller of blk_new_open() doesn't make use of the permissions, but they | |
384 | * shouldn't hurt either. We can still share everything here because the | |
385 | * guest devices will add their own blockers if they can't share. */ | |
1f4ad7d3 KW |
386 | if ((flags & BDRV_O_NO_IO) == 0) { |
387 | perm |= BLK_PERM_CONSISTENT_READ; | |
388 | if (flags & BDRV_O_RDWR) { | |
389 | perm |= BLK_PERM_WRITE; | |
390 | } | |
c62d32f5 KW |
391 | } |
392 | if (flags & BDRV_O_RESIZE) { | |
393 | perm |= BLK_PERM_RESIZE; | |
394 | } | |
ca49a4fd | 395 | |
d861ab3a | 396 | blk = blk_new(qemu_get_aio_context(), perm, BLK_PERM_ALL); |
5b363937 HR |
397 | bs = bdrv_open(filename, reference, options, flags, errp); |
398 | if (!bs) { | |
ca49a4fd HR |
399 | blk_unref(blk); |
400 | return NULL; | |
401 | } | |
402 | ||
132ada80 | 403 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx, |
50bfbe93 FZ |
404 | perm, BLK_PERM_ALL, blk, errp); |
405 | if (!blk->root) { | |
50bfbe93 FZ |
406 | blk_unref(blk); |
407 | return NULL; | |
408 | } | |
72e775c7 | 409 | |
ca49a4fd HR |
410 | return blk; |
411 | } | |
412 | ||
26f54e9a MA |
413 | static void blk_delete(BlockBackend *blk) |
414 | { | |
415 | assert(!blk->refcnt); | |
e5e78550 | 416 | assert(!blk->name); |
a7f53e26 | 417 | assert(!blk->dev); |
022cdc9f | 418 | if (blk->public.throttle_group_member.throttle_state) { |
1606e4cf EB |
419 | blk_io_limits_disable(blk); |
420 | } | |
f21d96d0 | 421 | if (blk->root) { |
13855c6b | 422 | blk_remove_bs(blk); |
7e7d56d9 | 423 | } |
5f7772c4 FZ |
424 | if (blk->vmsh) { |
425 | qemu_del_vm_change_state_handler(blk->vmsh); | |
426 | blk->vmsh = NULL; | |
427 | } | |
3301f6c6 HR |
428 | assert(QLIST_EMPTY(&blk->remove_bs_notifiers.notifiers)); |
429 | assert(QLIST_EMPTY(&blk->insert_bs_notifiers.notifiers)); | |
d03654ea | 430 | assert(QLIST_EMPTY(&blk->aio_notifiers)); |
2cf22d6a | 431 | QTAILQ_REMOVE(&block_backends, blk, link); |
18e46a03 | 432 | drive_info_del(blk->legacy_dinfo); |
979e9b03 | 433 | block_acct_cleanup(&blk->stats); |
26f54e9a MA |
434 | g_free(blk); |
435 | } | |
436 | ||
8fb3c76c MA |
437 | static void drive_info_del(DriveInfo *dinfo) |
438 | { | |
439 | if (!dinfo) { | |
440 | return; | |
441 | } | |
442 | qemu_opts_del(dinfo->opts); | |
8fb3c76c MA |
443 | g_free(dinfo); |
444 | } | |
445 | ||
f636ae85 AG |
446 | int blk_get_refcnt(BlockBackend *blk) |
447 | { | |
448 | return blk ? blk->refcnt : 0; | |
449 | } | |
450 | ||
26f54e9a MA |
451 | /* |
452 | * Increment @blk's reference count. | |
453 | * @blk must not be null. | |
454 | */ | |
455 | void blk_ref(BlockBackend *blk) | |
456 | { | |
5ca9d21b | 457 | assert(blk->refcnt > 0); |
26f54e9a MA |
458 | blk->refcnt++; |
459 | } | |
460 | ||
461 | /* | |
462 | * Decrement @blk's reference count. | |
463 | * If this drops it to zero, destroy @blk. | |
464 | * For convenience, do nothing if @blk is null. | |
465 | */ | |
466 | void blk_unref(BlockBackend *blk) | |
467 | { | |
468 | if (blk) { | |
469 | assert(blk->refcnt > 0); | |
5ca9d21b KW |
470 | if (blk->refcnt > 1) { |
471 | blk->refcnt--; | |
472 | } else { | |
473 | blk_drain(blk); | |
474 | /* blk_drain() cannot resurrect blk, nobody held a reference */ | |
475 | assert(blk->refcnt == 1); | |
476 | blk->refcnt = 0; | |
26f54e9a MA |
477 | blk_delete(blk); |
478 | } | |
479 | } | |
480 | } | |
481 | ||
2cf22d6a HR |
482 | /* |
483 | * Behaves similarly to blk_next() but iterates over all BlockBackends, even the | |
484 | * ones which are hidden (i.e. are not referenced by the monitor). | |
485 | */ | |
a429b9b5 | 486 | BlockBackend *blk_all_next(BlockBackend *blk) |
2cf22d6a HR |
487 | { |
488 | return blk ? QTAILQ_NEXT(blk, link) | |
489 | : QTAILQ_FIRST(&block_backends); | |
490 | } | |
491 | ||
d8da3cef HR |
492 | void blk_remove_all_bs(void) |
493 | { | |
74d1b8fc | 494 | BlockBackend *blk = NULL; |
d8da3cef | 495 | |
2cf22d6a | 496 | while ((blk = blk_all_next(blk)) != NULL) { |
d8da3cef HR |
497 | AioContext *ctx = blk_get_aio_context(blk); |
498 | ||
499 | aio_context_acquire(ctx); | |
f21d96d0 | 500 | if (blk->root) { |
d8da3cef HR |
501 | blk_remove_bs(blk); |
502 | } | |
503 | aio_context_release(ctx); | |
504 | } | |
505 | } | |
506 | ||
26f54e9a | 507 | /* |
9492b0b9 | 508 | * Return the monitor-owned BlockBackend after @blk. |
26f54e9a MA |
509 | * If @blk is null, return the first one. |
510 | * Else, return @blk's next sibling, which may be null. | |
511 | * | |
512 | * To iterate over all BlockBackends, do | |
513 | * for (blk = blk_next(NULL); blk; blk = blk_next(blk)) { | |
514 | * ... | |
515 | * } | |
516 | */ | |
517 | BlockBackend *blk_next(BlockBackend *blk) | |
518 | { | |
9492b0b9 HR |
519 | return blk ? QTAILQ_NEXT(blk, monitor_link) |
520 | : QTAILQ_FIRST(&monitor_block_backends); | |
26f54e9a MA |
521 | } |
522 | ||
7c8eece4 KW |
523 | /* Iterates over all top-level BlockDriverStates, i.e. BDSs that are owned by |
524 | * the monitor or attached to a BlockBackend */ | |
88be7b4b | 525 | BlockDriverState *bdrv_next(BdrvNextIterator *it) |
7c8eece4 | 526 | { |
5e003f17 HR |
527 | BlockDriverState *bs, *old_bs; |
528 | ||
529 | /* Must be called from the main loop */ | |
530 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
981f4f57 | 531 | |
7c8eece4 KW |
532 | /* First, return all root nodes of BlockBackends. In order to avoid |
533 | * returning a BDS twice when multiple BBs refer to it, we only return it | |
534 | * if the BB is the first one in the parent list of the BDS. */ | |
535 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
5e003f17 HR |
536 | BlockBackend *old_blk = it->blk; |
537 | ||
538 | old_bs = old_blk ? blk_bs(old_blk) : NULL; | |
539 | ||
7c8eece4 KW |
540 | do { |
541 | it->blk = blk_all_next(it->blk); | |
88be7b4b KW |
542 | bs = it->blk ? blk_bs(it->blk) : NULL; |
543 | } while (it->blk && (bs == NULL || bdrv_first_blk(bs) != it->blk)); | |
7c8eece4 | 544 | |
5e003f17 HR |
545 | if (it->blk) { |
546 | blk_ref(it->blk); | |
547 | } | |
548 | blk_unref(old_blk); | |
549 | ||
88be7b4b | 550 | if (bs) { |
5e003f17 HR |
551 | bdrv_ref(bs); |
552 | bdrv_unref(old_bs); | |
88be7b4b | 553 | return bs; |
7c8eece4 KW |
554 | } |
555 | it->phase = BDRV_NEXT_MONITOR_OWNED; | |
5e003f17 HR |
556 | } else { |
557 | old_bs = it->bs; | |
7c8eece4 KW |
558 | } |
559 | ||
560 | /* Then return the monitor-owned BDSes without a BB attached. Ignore all | |
561 | * BDSes that are attached to a BlockBackend here; they have been handled | |
562 | * by the above block already */ | |
981f4f57 | 563 | do { |
7c8eece4 | 564 | it->bs = bdrv_next_monitor_owned(it->bs); |
88be7b4b KW |
565 | bs = it->bs; |
566 | } while (bs && bdrv_has_blk(bs)); | |
567 | ||
5e003f17 HR |
568 | if (bs) { |
569 | bdrv_ref(bs); | |
570 | } | |
571 | bdrv_unref(old_bs); | |
572 | ||
88be7b4b KW |
573 | return bs; |
574 | } | |
575 | ||
5e003f17 | 576 | static void bdrv_next_reset(BdrvNextIterator *it) |
88be7b4b KW |
577 | { |
578 | *it = (BdrvNextIterator) { | |
579 | .phase = BDRV_NEXT_BACKEND_ROOTS, | |
580 | }; | |
5e003f17 | 581 | } |
981f4f57 | 582 | |
5e003f17 HR |
583 | BlockDriverState *bdrv_first(BdrvNextIterator *it) |
584 | { | |
585 | bdrv_next_reset(it); | |
88be7b4b | 586 | return bdrv_next(it); |
981f4f57 HR |
587 | } |
588 | ||
5e003f17 HR |
589 | /* Must be called when aborting a bdrv_next() iteration before |
590 | * bdrv_next() returns NULL */ | |
591 | void bdrv_next_cleanup(BdrvNextIterator *it) | |
592 | { | |
593 | /* Must be called from the main loop */ | |
594 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
595 | ||
596 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
597 | if (it->blk) { | |
598 | bdrv_unref(blk_bs(it->blk)); | |
599 | blk_unref(it->blk); | |
600 | } | |
601 | } else { | |
602 | bdrv_unref(it->bs); | |
603 | } | |
604 | ||
605 | bdrv_next_reset(it); | |
606 | } | |
607 | ||
e5e78550 HR |
608 | /* |
609 | * Add a BlockBackend into the list of backends referenced by the monitor, with | |
610 | * the given @name acting as the handle for the monitor. | |
611 | * Strictly for use by blockdev.c. | |
612 | * | |
613 | * @name must not be null or empty. | |
614 | * | |
615 | * Returns true on success and false on failure. In the latter case, an Error | |
616 | * object is returned through @errp. | |
617 | */ | |
618 | bool monitor_add_blk(BlockBackend *blk, const char *name, Error **errp) | |
619 | { | |
620 | assert(!blk->name); | |
621 | assert(name && name[0]); | |
622 | ||
623 | if (!id_wellformed(name)) { | |
624 | error_setg(errp, "Invalid device name"); | |
625 | return false; | |
626 | } | |
627 | if (blk_by_name(name)) { | |
628 | error_setg(errp, "Device with id '%s' already exists", name); | |
629 | return false; | |
630 | } | |
631 | if (bdrv_find_node(name)) { | |
632 | error_setg(errp, | |
633 | "Device name '%s' conflicts with an existing node name", | |
634 | name); | |
635 | return false; | |
636 | } | |
637 | ||
638 | blk->name = g_strdup(name); | |
639 | QTAILQ_INSERT_TAIL(&monitor_block_backends, blk, monitor_link); | |
640 | return true; | |
641 | } | |
642 | ||
643 | /* | |
644 | * Remove a BlockBackend from the list of backends referenced by the monitor. | |
645 | * Strictly for use by blockdev.c. | |
646 | */ | |
647 | void monitor_remove_blk(BlockBackend *blk) | |
648 | { | |
649 | if (!blk->name) { | |
650 | return; | |
651 | } | |
652 | ||
653 | QTAILQ_REMOVE(&monitor_block_backends, blk, monitor_link); | |
654 | g_free(blk->name); | |
655 | blk->name = NULL; | |
656 | } | |
657 | ||
26f54e9a | 658 | /* |
7e7d56d9 | 659 | * Return @blk's name, a non-null string. |
e5e78550 | 660 | * Returns an empty string iff @blk is not referenced by the monitor. |
26f54e9a | 661 | */ |
0731a50f | 662 | const char *blk_name(const BlockBackend *blk) |
26f54e9a | 663 | { |
e5e78550 | 664 | return blk->name ?: ""; |
26f54e9a MA |
665 | } |
666 | ||
667 | /* | |
668 | * Return the BlockBackend with name @name if it exists, else null. | |
669 | * @name must not be null. | |
670 | */ | |
671 | BlockBackend *blk_by_name(const char *name) | |
672 | { | |
74d1b8fc | 673 | BlockBackend *blk = NULL; |
26f54e9a MA |
674 | |
675 | assert(name); | |
74d1b8fc | 676 | while ((blk = blk_next(blk)) != NULL) { |
26f54e9a MA |
677 | if (!strcmp(name, blk->name)) { |
678 | return blk; | |
679 | } | |
680 | } | |
681 | return NULL; | |
682 | } | |
7e7d56d9 MA |
683 | |
684 | /* | |
685 | * Return the BlockDriverState attached to @blk if any, else null. | |
686 | */ | |
687 | BlockDriverState *blk_bs(BlockBackend *blk) | |
688 | { | |
f21d96d0 | 689 | return blk->root ? blk->root->bs : NULL; |
7e7d56d9 MA |
690 | } |
691 | ||
7c8eece4 | 692 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs) |
dde33812 KW |
693 | { |
694 | BdrvChild *child; | |
695 | QLIST_FOREACH(child, &bs->parents, next_parent) { | |
696 | if (child->role == &child_root) { | |
7c8eece4 | 697 | return child->opaque; |
dde33812 KW |
698 | } |
699 | } | |
700 | ||
7c8eece4 KW |
701 | return NULL; |
702 | } | |
703 | ||
704 | /* | |
705 | * Returns true if @bs has an associated BlockBackend. | |
706 | */ | |
707 | bool bdrv_has_blk(BlockDriverState *bs) | |
708 | { | |
709 | return bdrv_first_blk(bs) != NULL; | |
dde33812 KW |
710 | } |
711 | ||
b6c1bae5 KW |
712 | /* |
713 | * Returns true if @bs has only BlockBackends as parents. | |
714 | */ | |
715 | bool bdrv_is_root_node(BlockDriverState *bs) | |
716 | { | |
717 | BdrvChild *c; | |
718 | ||
719 | QLIST_FOREACH(c, &bs->parents, next_parent) { | |
720 | if (c->role != &child_root) { | |
721 | return false; | |
722 | } | |
723 | } | |
724 | ||
725 | return true; | |
726 | } | |
727 | ||
18e46a03 MA |
728 | /* |
729 | * Return @blk's DriveInfo if any, else null. | |
730 | */ | |
731 | DriveInfo *blk_legacy_dinfo(BlockBackend *blk) | |
732 | { | |
733 | return blk->legacy_dinfo; | |
734 | } | |
735 | ||
736 | /* | |
737 | * Set @blk's DriveInfo to @dinfo, and return it. | |
738 | * @blk must not have a DriveInfo set already. | |
739 | * No other BlockBackend may have the same DriveInfo set. | |
740 | */ | |
741 | DriveInfo *blk_set_legacy_dinfo(BlockBackend *blk, DriveInfo *dinfo) | |
742 | { | |
743 | assert(!blk->legacy_dinfo); | |
744 | return blk->legacy_dinfo = dinfo; | |
745 | } | |
746 | ||
747 | /* | |
748 | * Return the BlockBackend with DriveInfo @dinfo. | |
749 | * It must exist. | |
750 | */ | |
751 | BlockBackend *blk_by_legacy_dinfo(DriveInfo *dinfo) | |
752 | { | |
74d1b8fc | 753 | BlockBackend *blk = NULL; |
18e46a03 | 754 | |
74d1b8fc | 755 | while ((blk = blk_next(blk)) != NULL) { |
18e46a03 MA |
756 | if (blk->legacy_dinfo == dinfo) { |
757 | return blk; | |
758 | } | |
759 | } | |
760 | abort(); | |
761 | } | |
762 | ||
f2cd875d KW |
763 | /* |
764 | * Returns a pointer to the publicly accessible fields of @blk. | |
765 | */ | |
766 | BlockBackendPublic *blk_get_public(BlockBackend *blk) | |
767 | { | |
768 | return &blk->public; | |
769 | } | |
770 | ||
771 | /* | |
772 | * Returns a BlockBackend given the associated @public fields. | |
773 | */ | |
774 | BlockBackend *blk_by_public(BlockBackendPublic *public) | |
775 | { | |
776 | return container_of(public, BlockBackend, public); | |
777 | } | |
778 | ||
1c95f7e1 HR |
779 | /* |
780 | * Disassociates the currently associated BlockDriverState from @blk. | |
781 | */ | |
782 | void blk_remove_bs(BlockBackend *blk) | |
783 | { | |
c89bcf3a | 784 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
632a7735 | 785 | BlockDriverState *bs; |
022cdc9f | 786 | |
3301f6c6 | 787 | notifier_list_notify(&blk->remove_bs_notifiers, blk); |
c89bcf3a | 788 | if (tgm->throttle_state) { |
632a7735 ZL |
789 | bs = blk_bs(blk); |
790 | bdrv_drained_begin(bs); | |
c89bcf3a AG |
791 | throttle_group_detach_aio_context(tgm); |
792 | throttle_group_attach_aio_context(tgm, qemu_get_aio_context()); | |
632a7735 | 793 | bdrv_drained_end(bs); |
a5614993 | 794 | } |
1c95f7e1 | 795 | |
7ca7f0f6 KW |
796 | blk_update_root_state(blk); |
797 | ||
f45280cb GK |
798 | /* bdrv_root_unref_child() will cause blk->root to become stale and may |
799 | * switch to a completion coroutine later on. Let's drain all I/O here | |
800 | * to avoid that and a potential QEMU crash. | |
801 | */ | |
802 | blk_drain(blk); | |
f21d96d0 KW |
803 | bdrv_root_unref_child(blk->root); |
804 | blk->root = NULL; | |
1c95f7e1 HR |
805 | } |
806 | ||
0c3c36d6 HR |
807 | /* |
808 | * Associates a new BlockDriverState with @blk. | |
809 | */ | |
d7086422 | 810 | int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) |
0c3c36d6 | 811 | { |
c89bcf3a | 812 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
b441dc71 | 813 | bdrv_ref(bs); |
132ada80 | 814 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx, |
d7086422 KW |
815 | blk->perm, blk->shared_perm, blk, errp); |
816 | if (blk->root == NULL) { | |
817 | return -EPERM; | |
818 | } | |
3301f6c6 HR |
819 | |
820 | notifier_list_notify(&blk->insert_bs_notifiers, blk); | |
c89bcf3a AG |
821 | if (tgm->throttle_state) { |
822 | throttle_group_detach_aio_context(tgm); | |
823 | throttle_group_attach_aio_context(tgm, bdrv_get_aio_context(bs)); | |
7ca7f0f6 | 824 | } |
d7086422 KW |
825 | |
826 | return 0; | |
0c3c36d6 HR |
827 | } |
828 | ||
981776b3 KW |
829 | /* |
830 | * Sets the permission bitmasks that the user of the BlockBackend needs. | |
831 | */ | |
832 | int blk_set_perm(BlockBackend *blk, uint64_t perm, uint64_t shared_perm, | |
833 | Error **errp) | |
834 | { | |
835 | int ret; | |
836 | ||
d35ff5e6 | 837 | if (blk->root && !blk->disable_perm) { |
981776b3 KW |
838 | ret = bdrv_child_try_set_perm(blk->root, perm, shared_perm, errp); |
839 | if (ret < 0) { | |
840 | return ret; | |
841 | } | |
842 | } | |
843 | ||
844 | blk->perm = perm; | |
845 | blk->shared_perm = shared_perm; | |
846 | ||
847 | return 0; | |
848 | } | |
849 | ||
887354bd KW |
850 | void blk_get_perm(BlockBackend *blk, uint64_t *perm, uint64_t *shared_perm) |
851 | { | |
852 | *perm = blk->perm; | |
853 | *shared_perm = blk->shared_perm; | |
854 | } | |
855 | ||
d09ea2d2 TH |
856 | /* |
857 | * Attach device model @dev to @blk. | |
858 | * Return 0 on success, -EBUSY when a device model is attached already. | |
859 | */ | |
860 | int blk_attach_dev(BlockBackend *blk, DeviceState *dev) | |
4be74634 | 861 | { |
a7f53e26 MA |
862 | if (blk->dev) { |
863 | return -EBUSY; | |
864 | } | |
d35ff5e6 KW |
865 | |
866 | /* While migration is still incoming, we don't need to apply the | |
867 | * permissions of guest device BlockBackends. We might still have a block | |
868 | * job or NBD server writing to the image for storage migration. */ | |
869 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
870 | blk->disable_perm = true; | |
871 | } | |
872 | ||
84ebe375 | 873 | blk_ref(blk); |
a7f53e26 | 874 | blk->dev = dev; |
373340b2 | 875 | blk_iostatus_reset(blk); |
d35ff5e6 | 876 | |
a7f53e26 | 877 | return 0; |
4be74634 MA |
878 | } |
879 | ||
a7f53e26 MA |
880 | /* |
881 | * Detach device model @dev from @blk. | |
882 | * @dev must be currently attached to @blk. | |
883 | */ | |
d09ea2d2 | 884 | void blk_detach_dev(BlockBackend *blk, DeviceState *dev) |
4be74634 | 885 | { |
a7f53e26 MA |
886 | assert(blk->dev == dev); |
887 | blk->dev = NULL; | |
888 | blk->dev_ops = NULL; | |
889 | blk->dev_opaque = NULL; | |
68e9ec01 | 890 | blk->guest_block_size = 512; |
981776b3 | 891 | blk_set_perm(blk, 0, BLK_PERM_ALL, &error_abort); |
84ebe375 | 892 | blk_unref(blk); |
4be74634 MA |
893 | } |
894 | ||
a7f53e26 MA |
895 | /* |
896 | * Return the device model attached to @blk if any, else null. | |
897 | */ | |
d09ea2d2 | 898 | DeviceState *blk_get_attached_dev(BlockBackend *blk) |
a7f53e26 MA |
899 | { |
900 | return blk->dev; | |
901 | } | |
902 | ||
2d76e724 KW |
903 | /* Return the qdev ID, or if no ID is assigned the QOM path, of the block |
904 | * device attached to the BlockBackend. */ | |
77beef83 | 905 | char *blk_get_attached_dev_id(BlockBackend *blk) |
2d76e724 | 906 | { |
d09ea2d2 | 907 | DeviceState *dev = blk->dev; |
2d76e724 KW |
908 | |
909 | if (!dev) { | |
910 | return g_strdup(""); | |
911 | } else if (dev->id) { | |
912 | return g_strdup(dev->id); | |
913 | } | |
602414d1 LM |
914 | |
915 | return object_get_canonical_path(OBJECT(dev)) ?: g_strdup(""); | |
2d76e724 KW |
916 | } |
917 | ||
1c89e1fa KW |
918 | /* |
919 | * Return the BlockBackend which has the device model @dev attached if it | |
920 | * exists, else null. | |
921 | * | |
922 | * @dev must not be null. | |
923 | */ | |
924 | BlockBackend *blk_by_dev(void *dev) | |
925 | { | |
926 | BlockBackend *blk = NULL; | |
927 | ||
928 | assert(dev != NULL); | |
929 | while ((blk = blk_all_next(blk)) != NULL) { | |
930 | if (blk->dev == dev) { | |
931 | return blk; | |
932 | } | |
933 | } | |
934 | return NULL; | |
935 | } | |
936 | ||
a7f53e26 MA |
937 | /* |
938 | * Set @blk's device model callbacks to @ops. | |
939 | * @opaque is the opaque argument to pass to the callbacks. | |
940 | * This is for use by device models. | |
941 | */ | |
942 | void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, | |
943 | void *opaque) | |
944 | { | |
945 | blk->dev_ops = ops; | |
946 | blk->dev_opaque = opaque; | |
f4d9cc88 JS |
947 | |
948 | /* Are we currently quiesced? Should we enforce this right now? */ | |
949 | if (blk->quiesce_counter && ops->drained_begin) { | |
950 | ops->drained_begin(opaque); | |
951 | } | |
a7f53e26 MA |
952 | } |
953 | ||
954 | /* | |
955 | * Notify @blk's attached device model of media change. | |
39829a01 KW |
956 | * |
957 | * If @load is true, notify of media load. This action can fail, meaning that | |
958 | * the medium cannot be loaded. @errp is set then. | |
959 | * | |
960 | * If @load is false, notify of media eject. This can never fail. | |
961 | * | |
a7f53e26 MA |
962 | * Also send DEVICE_TRAY_MOVED events as appropriate. |
963 | */ | |
39829a01 | 964 | void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp) |
a7f53e26 MA |
965 | { |
966 | if (blk->dev_ops && blk->dev_ops->change_media_cb) { | |
f1f57066 | 967 | bool tray_was_open, tray_is_open; |
39829a01 | 968 | Error *local_err = NULL; |
a7f53e26 | 969 | |
f1f57066 | 970 | tray_was_open = blk_dev_is_tray_open(blk); |
39829a01 KW |
971 | blk->dev_ops->change_media_cb(blk->dev_opaque, load, &local_err); |
972 | if (local_err) { | |
973 | assert(load == true); | |
974 | error_propagate(errp, local_err); | |
975 | return; | |
976 | } | |
f1f57066 HR |
977 | tray_is_open = blk_dev_is_tray_open(blk); |
978 | ||
979 | if (tray_was_open != tray_is_open) { | |
2d76e724 | 980 | char *id = blk_get_attached_dev_id(blk); |
3ab72385 | 981 | qapi_event_send_device_tray_moved(blk_name(blk), id, tray_is_open); |
2d76e724 | 982 | g_free(id); |
a7f53e26 MA |
983 | } |
984 | } | |
985 | } | |
986 | ||
5c8cab48 KW |
987 | static void blk_root_change_media(BdrvChild *child, bool load) |
988 | { | |
39829a01 | 989 | blk_dev_change_media_cb(child->opaque, load, NULL); |
5c8cab48 KW |
990 | } |
991 | ||
a7f53e26 MA |
992 | /* |
993 | * Does @blk's attached device model have removable media? | |
994 | * %true if no device model is attached. | |
995 | */ | |
996 | bool blk_dev_has_removable_media(BlockBackend *blk) | |
997 | { | |
998 | return !blk->dev || (blk->dev_ops && blk->dev_ops->change_media_cb); | |
999 | } | |
1000 | ||
8f3a73bc HR |
1001 | /* |
1002 | * Does @blk's attached device model have a tray? | |
1003 | */ | |
1004 | bool blk_dev_has_tray(BlockBackend *blk) | |
1005 | { | |
1006 | return blk->dev_ops && blk->dev_ops->is_tray_open; | |
1007 | } | |
1008 | ||
a7f53e26 MA |
1009 | /* |
1010 | * Notify @blk's attached device model of a media eject request. | |
1011 | * If @force is true, the medium is about to be yanked out forcefully. | |
1012 | */ | |
1013 | void blk_dev_eject_request(BlockBackend *blk, bool force) | |
4be74634 | 1014 | { |
a7f53e26 MA |
1015 | if (blk->dev_ops && blk->dev_ops->eject_request_cb) { |
1016 | blk->dev_ops->eject_request_cb(blk->dev_opaque, force); | |
1017 | } | |
4be74634 MA |
1018 | } |
1019 | ||
a7f53e26 MA |
1020 | /* |
1021 | * Does @blk's attached device model have a tray, and is it open? | |
1022 | */ | |
1023 | bool blk_dev_is_tray_open(BlockBackend *blk) | |
4be74634 | 1024 | { |
8f3a73bc | 1025 | if (blk_dev_has_tray(blk)) { |
a7f53e26 MA |
1026 | return blk->dev_ops->is_tray_open(blk->dev_opaque); |
1027 | } | |
1028 | return false; | |
1029 | } | |
1030 | ||
1031 | /* | |
1032 | * Does @blk's attached device model have the medium locked? | |
1033 | * %false if the device model has no such lock. | |
1034 | */ | |
1035 | bool blk_dev_is_medium_locked(BlockBackend *blk) | |
1036 | { | |
1037 | if (blk->dev_ops && blk->dev_ops->is_medium_locked) { | |
1038 | return blk->dev_ops->is_medium_locked(blk->dev_opaque); | |
1039 | } | |
1040 | return false; | |
1041 | } | |
1042 | ||
1043 | /* | |
1044 | * Notify @blk's attached device model of a backend size change. | |
1045 | */ | |
5c8cab48 | 1046 | static void blk_root_resize(BdrvChild *child) |
a7f53e26 | 1047 | { |
5c8cab48 KW |
1048 | BlockBackend *blk = child->opaque; |
1049 | ||
a7f53e26 MA |
1050 | if (blk->dev_ops && blk->dev_ops->resize_cb) { |
1051 | blk->dev_ops->resize_cb(blk->dev_opaque); | |
1052 | } | |
1053 | } | |
1054 | ||
1055 | void blk_iostatus_enable(BlockBackend *blk) | |
1056 | { | |
373340b2 HR |
1057 | blk->iostatus_enabled = true; |
1058 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
1059 | } | |
1060 | ||
1061 | /* The I/O status is only enabled if the drive explicitly | |
1062 | * enables it _and_ the VM is configured to stop on errors */ | |
1063 | bool blk_iostatus_is_enabled(const BlockBackend *blk) | |
1064 | { | |
1065 | return (blk->iostatus_enabled && | |
1066 | (blk->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || | |
1067 | blk->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
1068 | blk->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
1069 | } | |
1070 | ||
1071 | BlockDeviceIoStatus blk_iostatus(const BlockBackend *blk) | |
1072 | { | |
1073 | return blk->iostatus; | |
1074 | } | |
1075 | ||
1076 | void blk_iostatus_disable(BlockBackend *blk) | |
1077 | { | |
1078 | blk->iostatus_enabled = false; | |
1079 | } | |
1080 | ||
1081 | void blk_iostatus_reset(BlockBackend *blk) | |
1082 | { | |
1083 | if (blk_iostatus_is_enabled(blk)) { | |
1084 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
373340b2 HR |
1085 | } |
1086 | } | |
1087 | ||
1088 | void blk_iostatus_set_err(BlockBackend *blk, int error) | |
1089 | { | |
1090 | assert(blk_iostatus_is_enabled(blk)); | |
1091 | if (blk->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
1092 | blk->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : | |
1093 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
1094 | } | |
4be74634 MA |
1095 | } |
1096 | ||
c10c9d96 KW |
1097 | void blk_set_allow_write_beyond_eof(BlockBackend *blk, bool allow) |
1098 | { | |
1099 | blk->allow_write_beyond_eof = allow; | |
1100 | } | |
1101 | ||
980b0f94 KW |
1102 | void blk_set_allow_aio_context_change(BlockBackend *blk, bool allow) |
1103 | { | |
1104 | blk->allow_aio_context_change = allow; | |
1105 | } | |
1106 | ||
cf312932 KW |
1107 | void blk_set_disable_request_queuing(BlockBackend *blk, bool disable) |
1108 | { | |
1109 | blk->disable_request_queuing = disable; | |
1110 | } | |
1111 | ||
e7f7d676 HR |
1112 | static int blk_check_byte_request(BlockBackend *blk, int64_t offset, |
1113 | size_t size) | |
1114 | { | |
1115 | int64_t len; | |
1116 | ||
1117 | if (size > INT_MAX) { | |
1118 | return -EIO; | |
1119 | } | |
1120 | ||
c09ba36c | 1121 | if (!blk_is_available(blk)) { |
e7f7d676 HR |
1122 | return -ENOMEDIUM; |
1123 | } | |
1124 | ||
e7f7d676 HR |
1125 | if (offset < 0) { |
1126 | return -EIO; | |
1127 | } | |
1128 | ||
c10c9d96 KW |
1129 | if (!blk->allow_write_beyond_eof) { |
1130 | len = blk_getlength(blk); | |
1131 | if (len < 0) { | |
1132 | return len; | |
1133 | } | |
1134 | ||
1135 | if (offset > len || len - offset < size) { | |
1136 | return -EIO; | |
1137 | } | |
e7f7d676 HR |
1138 | } |
1139 | ||
1140 | return 0; | |
1141 | } | |
1142 | ||
7f16476f | 1143 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
cf312932 KW |
1144 | static void coroutine_fn blk_wait_while_drained(BlockBackend *blk) |
1145 | { | |
7f16476f KW |
1146 | assert(blk->in_flight > 0); |
1147 | ||
cf312932 | 1148 | if (blk->quiesce_counter && !blk->disable_request_queuing) { |
7f16476f | 1149 | blk_dec_in_flight(blk); |
cf312932 | 1150 | qemu_co_queue_wait(&blk->queued_requests, NULL); |
7f16476f | 1151 | blk_inc_in_flight(blk); |
cf312932 KW |
1152 | } |
1153 | } | |
1154 | ||
fbb92b67 KW |
1155 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1156 | static int coroutine_fn | |
1157 | blk_do_preadv(BlockBackend *blk, int64_t offset, unsigned int bytes, | |
1158 | QEMUIOVector *qiov, BdrvRequestFlags flags) | |
4be74634 | 1159 | { |
1e98fefd | 1160 | int ret; |
cf312932 | 1161 | BlockDriverState *bs; |
1e98fefd | 1162 | |
cf312932 | 1163 | blk_wait_while_drained(blk); |
1e98fefd | 1164 | |
cf312932 KW |
1165 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1166 | bs = blk_bs(blk); | |
99723548 | 1167 | trace_blk_co_preadv(blk, bs, offset, bytes, flags); |
1e98fefd KW |
1168 | |
1169 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1170 | if (ret < 0) { |
1171 | return ret; | |
1172 | } | |
1173 | ||
99723548 PB |
1174 | bdrv_inc_in_flight(bs); |
1175 | ||
441565b2 | 1176 | /* throttling disk I/O */ |
022cdc9f MP |
1177 | if (blk->public.throttle_group_member.throttle_state) { |
1178 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1179 | bytes, false); | |
441565b2 KW |
1180 | } |
1181 | ||
99723548 PB |
1182 | ret = bdrv_co_preadv(blk->root, offset, bytes, qiov, flags); |
1183 | bdrv_dec_in_flight(bs); | |
1184 | return ret; | |
1bf1cbc9 KW |
1185 | } |
1186 | ||
fbb92b67 KW |
1187 | int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset, |
1188 | unsigned int bytes, QEMUIOVector *qiov, | |
1189 | BdrvRequestFlags flags) | |
1190 | { | |
1191 | int ret; | |
1192 | ||
1193 | blk_inc_in_flight(blk); | |
1194 | ret = blk_do_preadv(blk, offset, bytes, qiov, flags); | |
1195 | blk_dec_in_flight(blk); | |
1196 | ||
1197 | return ret; | |
1198 | } | |
1199 | ||
1200 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ | |
1201 | static int coroutine_fn | |
1202 | blk_do_pwritev_part(BlockBackend *blk, int64_t offset, unsigned int bytes, | |
1203 | QEMUIOVector *qiov, size_t qiov_offset, | |
1204 | BdrvRequestFlags flags) | |
a8823a3b | 1205 | { |
bfd18d1e | 1206 | int ret; |
cf312932 | 1207 | BlockDriverState *bs; |
bfd18d1e | 1208 | |
cf312932 | 1209 | blk_wait_while_drained(blk); |
bfd18d1e | 1210 | |
cf312932 KW |
1211 | /* Call blk_bs() only after waiting, the graph may have changed */ |
1212 | bs = blk_bs(blk); | |
99723548 | 1213 | trace_blk_co_pwritev(blk, bs, offset, bytes, flags); |
1e98fefd | 1214 | |
bfd18d1e | 1215 | ret = blk_check_byte_request(blk, offset, bytes); |
a8823a3b KW |
1216 | if (ret < 0) { |
1217 | return ret; | |
1218 | } | |
1219 | ||
99723548 | 1220 | bdrv_inc_in_flight(bs); |
441565b2 | 1221 | /* throttling disk I/O */ |
022cdc9f MP |
1222 | if (blk->public.throttle_group_member.throttle_state) { |
1223 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1224 | bytes, true); | |
441565b2 KW |
1225 | } |
1226 | ||
bfd18d1e KW |
1227 | if (!blk->enable_write_cache) { |
1228 | flags |= BDRV_REQ_FUA; | |
1229 | } | |
1230 | ||
b3016864 VSO |
1231 | ret = bdrv_co_pwritev_part(blk->root, offset, bytes, qiov, qiov_offset, |
1232 | flags); | |
99723548 PB |
1233 | bdrv_dec_in_flight(bs); |
1234 | return ret; | |
a8823a3b KW |
1235 | } |
1236 | ||
fbb92b67 KW |
1237 | int coroutine_fn blk_co_pwritev_part(BlockBackend *blk, int64_t offset, |
1238 | unsigned int bytes, | |
1239 | QEMUIOVector *qiov, size_t qiov_offset, | |
1240 | BdrvRequestFlags flags) | |
1241 | { | |
1242 | int ret; | |
1243 | ||
1244 | blk_inc_in_flight(blk); | |
1245 | ret = blk_do_pwritev_part(blk, offset, bytes, qiov, qiov_offset, flags); | |
1246 | blk_dec_in_flight(blk); | |
1247 | ||
1248 | return ret; | |
1249 | } | |
1250 | ||
b3016864 VSO |
1251 | int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset, |
1252 | unsigned int bytes, QEMUIOVector *qiov, | |
1253 | BdrvRequestFlags flags) | |
1254 | { | |
1255 | return blk_co_pwritev_part(blk, offset, bytes, qiov, 0, flags); | |
1256 | } | |
1257 | ||
1bf1cbc9 KW |
1258 | typedef struct BlkRwCo { |
1259 | BlockBackend *blk; | |
1260 | int64_t offset; | |
c060332c | 1261 | void *iobuf; |
1bf1cbc9 KW |
1262 | int ret; |
1263 | BdrvRequestFlags flags; | |
1264 | } BlkRwCo; | |
1265 | ||
1266 | static void blk_read_entry(void *opaque) | |
1267 | { | |
1268 | BlkRwCo *rwco = opaque; | |
c060332c | 1269 | QEMUIOVector *qiov = rwco->iobuf; |
1bf1cbc9 | 1270 | |
fbb92b67 | 1271 | rwco->ret = blk_do_preadv(rwco->blk, rwco->offset, qiov->size, |
c060332c | 1272 | qiov, rwco->flags); |
4720cbee | 1273 | aio_wait_kick(); |
1bf1cbc9 KW |
1274 | } |
1275 | ||
a8823a3b KW |
1276 | static void blk_write_entry(void *opaque) |
1277 | { | |
1278 | BlkRwCo *rwco = opaque; | |
c060332c | 1279 | QEMUIOVector *qiov = rwco->iobuf; |
a8823a3b | 1280 | |
fbb92b67 KW |
1281 | rwco->ret = blk_do_pwritev_part(rwco->blk, rwco->offset, qiov->size, |
1282 | qiov, 0, rwco->flags); | |
4720cbee | 1283 | aio_wait_kick(); |
a8823a3b KW |
1284 | } |
1285 | ||
a55d3fba KW |
1286 | static int blk_prw(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1287 | int64_t bytes, CoroutineEntry co_entry, | |
1288 | BdrvRequestFlags flags) | |
1bf1cbc9 | 1289 | { |
ae5a9592 VSO |
1290 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); |
1291 | BlkRwCo rwco = { | |
1bf1cbc9 | 1292 | .blk = blk, |
a55d3fba | 1293 | .offset = offset, |
c060332c | 1294 | .iobuf = &qiov, |
fc1453cd | 1295 | .flags = flags, |
1bf1cbc9 KW |
1296 | .ret = NOT_DONE, |
1297 | }; | |
1298 | ||
fbb92b67 | 1299 | blk_inc_in_flight(blk); |
35f106e6 PB |
1300 | if (qemu_in_coroutine()) { |
1301 | /* Fast-path if already in coroutine context */ | |
1302 | co_entry(&rwco); | |
1303 | } else { | |
1304 | Coroutine *co = qemu_coroutine_create(co_entry, &rwco); | |
e92f0e19 | 1305 | bdrv_coroutine_enter(blk_bs(blk), co); |
35f106e6 PB |
1306 | BDRV_POLL_WHILE(blk_bs(blk), rwco.ret == NOT_DONE); |
1307 | } | |
fbb92b67 | 1308 | blk_dec_in_flight(blk); |
1bf1cbc9 KW |
1309 | |
1310 | return rwco.ret; | |
4be74634 MA |
1311 | } |
1312 | ||
d004bd52 | 1313 | int blk_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 1314 | int bytes, BdrvRequestFlags flags) |
0df89e8e | 1315 | { |
f5a5ca79 | 1316 | return blk_prw(blk, offset, NULL, bytes, blk_write_entry, |
983a1600 | 1317 | flags | BDRV_REQ_ZERO_WRITE); |
0df89e8e KW |
1318 | } |
1319 | ||
720ff280 KW |
1320 | int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags) |
1321 | { | |
1322 | return bdrv_make_zero(blk->root, flags); | |
1323 | } | |
1324 | ||
c90e2a9c | 1325 | void blk_inc_in_flight(BlockBackend *blk) |
33f2a757 SH |
1326 | { |
1327 | atomic_inc(&blk->in_flight); | |
1328 | } | |
1329 | ||
c90e2a9c | 1330 | void blk_dec_in_flight(BlockBackend *blk) |
33f2a757 SH |
1331 | { |
1332 | atomic_dec(&blk->in_flight); | |
cfe29d82 | 1333 | aio_wait_kick(); |
33f2a757 SH |
1334 | } |
1335 | ||
e7f7d676 HR |
1336 | static void error_callback_bh(void *opaque) |
1337 | { | |
1338 | struct BlockBackendAIOCB *acb = opaque; | |
99723548 | 1339 | |
33f2a757 | 1340 | blk_dec_in_flight(acb->blk); |
e7f7d676 HR |
1341 | acb->common.cb(acb->common.opaque, acb->ret); |
1342 | qemu_aio_unref(acb); | |
1343 | } | |
1344 | ||
ca78ecfa PL |
1345 | BlockAIOCB *blk_abort_aio_request(BlockBackend *blk, |
1346 | BlockCompletionFunc *cb, | |
1347 | void *opaque, int ret) | |
e7f7d676 HR |
1348 | { |
1349 | struct BlockBackendAIOCB *acb; | |
e7f7d676 | 1350 | |
33f2a757 | 1351 | blk_inc_in_flight(blk); |
e7f7d676 | 1352 | acb = blk_aio_get(&block_backend_aiocb_info, blk, cb, opaque); |
4981bdec | 1353 | acb->blk = blk; |
e7f7d676 HR |
1354 | acb->ret = ret; |
1355 | ||
e4ec5ad4 PD |
1356 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1357 | error_callback_bh, acb); | |
e7f7d676 HR |
1358 | return &acb->common; |
1359 | } | |
1360 | ||
57d6a428 KW |
1361 | typedef struct BlkAioEmAIOCB { |
1362 | BlockAIOCB common; | |
1363 | BlkRwCo rwco; | |
7fa84cd8 | 1364 | int bytes; |
57d6a428 | 1365 | bool has_returned; |
57d6a428 KW |
1366 | } BlkAioEmAIOCB; |
1367 | ||
1368 | static const AIOCBInfo blk_aio_em_aiocb_info = { | |
1369 | .aiocb_size = sizeof(BlkAioEmAIOCB), | |
1370 | }; | |
1371 | ||
1372 | static void blk_aio_complete(BlkAioEmAIOCB *acb) | |
1373 | { | |
57d6a428 KW |
1374 | if (acb->has_returned) { |
1375 | acb->common.cb(acb->common.opaque, acb->rwco.ret); | |
46aaf2a5 | 1376 | blk_dec_in_flight(acb->rwco.blk); |
57d6a428 KW |
1377 | qemu_aio_unref(acb); |
1378 | } | |
1379 | } | |
1380 | ||
1381 | static void blk_aio_complete_bh(void *opaque) | |
1382 | { | |
fffb6e12 | 1383 | BlkAioEmAIOCB *acb = opaque; |
fffb6e12 PB |
1384 | assert(acb->has_returned); |
1385 | blk_aio_complete(acb); | |
57d6a428 KW |
1386 | } |
1387 | ||
7fa84cd8 | 1388 | static BlockAIOCB *blk_aio_prwv(BlockBackend *blk, int64_t offset, int bytes, |
c060332c | 1389 | void *iobuf, CoroutineEntry co_entry, |
57d6a428 KW |
1390 | BdrvRequestFlags flags, |
1391 | BlockCompletionFunc *cb, void *opaque) | |
1392 | { | |
1393 | BlkAioEmAIOCB *acb; | |
1394 | Coroutine *co; | |
1395 | ||
33f2a757 | 1396 | blk_inc_in_flight(blk); |
57d6a428 KW |
1397 | acb = blk_aio_get(&blk_aio_em_aiocb_info, blk, cb, opaque); |
1398 | acb->rwco = (BlkRwCo) { | |
1399 | .blk = blk, | |
1400 | .offset = offset, | |
c060332c | 1401 | .iobuf = iobuf, |
57d6a428 KW |
1402 | .flags = flags, |
1403 | .ret = NOT_DONE, | |
1404 | }; | |
7fa84cd8 | 1405 | acb->bytes = bytes; |
57d6a428 KW |
1406 | acb->has_returned = false; |
1407 | ||
0b8b8753 | 1408 | co = qemu_coroutine_create(co_entry, acb); |
e92f0e19 | 1409 | bdrv_coroutine_enter(blk_bs(blk), co); |
57d6a428 KW |
1410 | |
1411 | acb->has_returned = true; | |
1412 | if (acb->rwco.ret != NOT_DONE) { | |
e4ec5ad4 PD |
1413 | replay_bh_schedule_oneshot_event(blk_get_aio_context(blk), |
1414 | blk_aio_complete_bh, acb); | |
57d6a428 KW |
1415 | } |
1416 | ||
1417 | return &acb->common; | |
1418 | } | |
1419 | ||
1420 | static void blk_aio_read_entry(void *opaque) | |
1421 | { | |
1422 | BlkAioEmAIOCB *acb = opaque; | |
1423 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1424 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1425 | |
c060332c | 1426 | assert(qiov->size == acb->bytes); |
fbb92b67 | 1427 | rwco->ret = blk_do_preadv(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1428 | qiov, rwco->flags); |
57d6a428 KW |
1429 | blk_aio_complete(acb); |
1430 | } | |
1431 | ||
1432 | static void blk_aio_write_entry(void *opaque) | |
1433 | { | |
1434 | BlkAioEmAIOCB *acb = opaque; | |
1435 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1436 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1437 | |
c060332c | 1438 | assert(!qiov || qiov->size == acb->bytes); |
fbb92b67 KW |
1439 | rwco->ret = blk_do_pwritev_part(rwco->blk, rwco->offset, acb->bytes, |
1440 | qiov, 0, rwco->flags); | |
57d6a428 KW |
1441 | blk_aio_complete(acb); |
1442 | } | |
1443 | ||
d004bd52 EB |
1444 | BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
1445 | int count, BdrvRequestFlags flags, | |
1446 | BlockCompletionFunc *cb, void *opaque) | |
4be74634 | 1447 | { |
983a1600 EB |
1448 | return blk_aio_prwv(blk, offset, count, NULL, blk_aio_write_entry, |
1449 | flags | BDRV_REQ_ZERO_WRITE, cb, opaque); | |
4be74634 MA |
1450 | } |
1451 | ||
1452 | int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count) | |
1453 | { | |
a55d3fba | 1454 | int ret = blk_prw(blk, offset, buf, count, blk_read_entry, 0); |
e7f7d676 HR |
1455 | if (ret < 0) { |
1456 | return ret; | |
1457 | } | |
a55d3fba | 1458 | return count; |
4be74634 MA |
1459 | } |
1460 | ||
8341f00d EB |
1461 | int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count, |
1462 | BdrvRequestFlags flags) | |
4be74634 | 1463 | { |
8341f00d EB |
1464 | int ret = blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
1465 | flags); | |
e7f7d676 HR |
1466 | if (ret < 0) { |
1467 | return ret; | |
1468 | } | |
a55d3fba | 1469 | return count; |
4be74634 MA |
1470 | } |
1471 | ||
1472 | int64_t blk_getlength(BlockBackend *blk) | |
1473 | { | |
c09ba36c HR |
1474 | if (!blk_is_available(blk)) { |
1475 | return -ENOMEDIUM; | |
1476 | } | |
1477 | ||
f21d96d0 | 1478 | return bdrv_getlength(blk_bs(blk)); |
4be74634 MA |
1479 | } |
1480 | ||
1481 | void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr) | |
1482 | { | |
f21d96d0 | 1483 | if (!blk_bs(blk)) { |
a46fc9c9 HR |
1484 | *nb_sectors_ptr = 0; |
1485 | } else { | |
f21d96d0 | 1486 | bdrv_get_geometry(blk_bs(blk), nb_sectors_ptr); |
a46fc9c9 | 1487 | } |
4be74634 MA |
1488 | } |
1489 | ||
1ef01253 HR |
1490 | int64_t blk_nb_sectors(BlockBackend *blk) |
1491 | { | |
c09ba36c HR |
1492 | if (!blk_is_available(blk)) { |
1493 | return -ENOMEDIUM; | |
1494 | } | |
1495 | ||
f21d96d0 | 1496 | return bdrv_nb_sectors(blk_bs(blk)); |
1ef01253 HR |
1497 | } |
1498 | ||
60cb2fa7 EB |
1499 | BlockAIOCB *blk_aio_preadv(BlockBackend *blk, int64_t offset, |
1500 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1501 | BlockCompletionFunc *cb, void *opaque) | |
1502 | { | |
1503 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1504 | blk_aio_read_entry, flags, cb, opaque); | |
1505 | } | |
1506 | ||
60cb2fa7 EB |
1507 | BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
1508 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1509 | BlockCompletionFunc *cb, void *opaque) | |
1510 | { | |
1511 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1512 | blk_aio_write_entry, flags, cb, opaque); | |
1513 | } | |
1514 | ||
4be74634 MA |
1515 | void blk_aio_cancel(BlockAIOCB *acb) |
1516 | { | |
1517 | bdrv_aio_cancel(acb); | |
1518 | } | |
1519 | ||
1520 | void blk_aio_cancel_async(BlockAIOCB *acb) | |
1521 | { | |
1522 | bdrv_aio_cancel_async(acb); | |
1523 | } | |
1524 | ||
fbb92b67 KW |
1525 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1526 | static int coroutine_fn | |
1527 | blk_do_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
4be74634 | 1528 | { |
cf312932 KW |
1529 | blk_wait_while_drained(blk); |
1530 | ||
c09ba36c HR |
1531 | if (!blk_is_available(blk)) { |
1532 | return -ENOMEDIUM; | |
1533 | } | |
1534 | ||
48af776a KW |
1535 | return bdrv_co_ioctl(blk_bs(blk), req, buf); |
1536 | } | |
1537 | ||
1538 | static void blk_ioctl_entry(void *opaque) | |
1539 | { | |
1540 | BlkRwCo *rwco = opaque; | |
c060332c DS |
1541 | QEMUIOVector *qiov = rwco->iobuf; |
1542 | ||
fbb92b67 | 1543 | rwco->ret = blk_do_ioctl(rwco->blk, rwco->offset, qiov->iov[0].iov_base); |
4720cbee | 1544 | aio_wait_kick(); |
48af776a KW |
1545 | } |
1546 | ||
1547 | int blk_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
1548 | { | |
1549 | return blk_prw(blk, req, buf, 0, blk_ioctl_entry, 0); | |
1550 | } | |
1551 | ||
1552 | static void blk_aio_ioctl_entry(void *opaque) | |
1553 | { | |
1554 | BlkAioEmAIOCB *acb = opaque; | |
1555 | BlkRwCo *rwco = &acb->rwco; | |
1556 | ||
fbb92b67 | 1557 | rwco->ret = blk_do_ioctl(rwco->blk, rwco->offset, rwco->iobuf); |
c060332c | 1558 | |
48af776a | 1559 | blk_aio_complete(acb); |
4be74634 MA |
1560 | } |
1561 | ||
1562 | BlockAIOCB *blk_aio_ioctl(BlockBackend *blk, unsigned long int req, void *buf, | |
1563 | BlockCompletionFunc *cb, void *opaque) | |
1564 | { | |
c060332c | 1565 | return blk_aio_prwv(blk, req, 0, buf, blk_aio_ioctl_entry, 0, cb, opaque); |
4be74634 MA |
1566 | } |
1567 | ||
fbb92b67 KW |
1568 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1569 | static int coroutine_fn | |
1570 | blk_do_pdiscard(BlockBackend *blk, int64_t offset, int bytes) | |
2bb0dce7 | 1571 | { |
cf312932 KW |
1572 | int ret; |
1573 | ||
1574 | blk_wait_while_drained(blk); | |
1575 | ||
1576 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1577 | if (ret < 0) { |
1578 | return ret; | |
1579 | } | |
1580 | ||
0b9fd3f4 | 1581 | return bdrv_co_pdiscard(blk->root, offset, bytes); |
2bb0dce7 HR |
1582 | } |
1583 | ||
564806c5 KW |
1584 | static void blk_aio_pdiscard_entry(void *opaque) |
1585 | { | |
1586 | BlkAioEmAIOCB *acb = opaque; | |
1587 | BlkRwCo *rwco = &acb->rwco; | |
1588 | ||
fbb92b67 | 1589 | rwco->ret = blk_do_pdiscard(rwco->blk, rwco->offset, acb->bytes); |
564806c5 KW |
1590 | blk_aio_complete(acb); |
1591 | } | |
1592 | ||
1593 | BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, | |
1594 | int64_t offset, int bytes, | |
1595 | BlockCompletionFunc *cb, void *opaque) | |
1596 | { | |
1597 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_pdiscard_entry, 0, | |
1598 | cb, opaque); | |
1599 | } | |
1600 | ||
fbb92b67 KW |
1601 | int coroutine_fn blk_co_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
1602 | { | |
1603 | int ret; | |
1604 | ||
1605 | blk_inc_in_flight(blk); | |
1606 | ret = blk_do_pdiscard(blk, offset, bytes); | |
1607 | blk_dec_in_flight(blk); | |
1608 | ||
1609 | return ret; | |
1610 | } | |
1611 | ||
564806c5 KW |
1612 | static void blk_pdiscard_entry(void *opaque) |
1613 | { | |
1614 | BlkRwCo *rwco = opaque; | |
1615 | QEMUIOVector *qiov = rwco->iobuf; | |
1616 | ||
fbb92b67 | 1617 | rwco->ret = blk_do_pdiscard(rwco->blk, rwco->offset, qiov->size); |
564806c5 KW |
1618 | aio_wait_kick(); |
1619 | } | |
1620 | ||
1621 | int blk_pdiscard(BlockBackend *blk, int64_t offset, int bytes) | |
1622 | { | |
1623 | return blk_prw(blk, offset, NULL, bytes, blk_pdiscard_entry, 0); | |
1624 | } | |
1625 | ||
fbb92b67 KW |
1626 | /* To be called between exactly one pair of blk_inc/dec_in_flight() */ |
1627 | static int coroutine_fn blk_do_flush(BlockBackend *blk) | |
2bb0dce7 | 1628 | { |
cf312932 KW |
1629 | blk_wait_while_drained(blk); |
1630 | ||
c09ba36c HR |
1631 | if (!blk_is_available(blk)) { |
1632 | return -ENOMEDIUM; | |
1633 | } | |
1634 | ||
f21d96d0 | 1635 | return bdrv_co_flush(blk_bs(blk)); |
2bb0dce7 HR |
1636 | } |
1637 | ||
564806c5 KW |
1638 | static void blk_aio_flush_entry(void *opaque) |
1639 | { | |
1640 | BlkAioEmAIOCB *acb = opaque; | |
1641 | BlkRwCo *rwco = &acb->rwco; | |
1642 | ||
fbb92b67 | 1643 | rwco->ret = blk_do_flush(rwco->blk); |
564806c5 KW |
1644 | blk_aio_complete(acb); |
1645 | } | |
1646 | ||
1647 | BlockAIOCB *blk_aio_flush(BlockBackend *blk, | |
1648 | BlockCompletionFunc *cb, void *opaque) | |
1649 | { | |
1650 | return blk_aio_prwv(blk, 0, 0, NULL, blk_aio_flush_entry, 0, cb, opaque); | |
1651 | } | |
1652 | ||
fbb92b67 KW |
1653 | int coroutine_fn blk_co_flush(BlockBackend *blk) |
1654 | { | |
1655 | int ret; | |
1656 | ||
1657 | blk_inc_in_flight(blk); | |
1658 | ret = blk_do_flush(blk); | |
1659 | blk_dec_in_flight(blk); | |
1660 | ||
1661 | return ret; | |
1662 | } | |
1663 | ||
be07a889 | 1664 | static void blk_flush_entry(void *opaque) |
4be74634 | 1665 | { |
be07a889 | 1666 | BlkRwCo *rwco = opaque; |
fbb92b67 | 1667 | rwco->ret = blk_do_flush(rwco->blk); |
4720cbee | 1668 | aio_wait_kick(); |
be07a889 | 1669 | } |
c09ba36c | 1670 | |
be07a889 KW |
1671 | int blk_flush(BlockBackend *blk) |
1672 | { | |
1673 | return blk_prw(blk, 0, NULL, 0, blk_flush_entry, 0); | |
4be74634 MA |
1674 | } |
1675 | ||
97b0385a AY |
1676 | void blk_drain(BlockBackend *blk) |
1677 | { | |
33f2a757 SH |
1678 | BlockDriverState *bs = blk_bs(blk); |
1679 | ||
1680 | if (bs) { | |
1681 | bdrv_drained_begin(bs); | |
1682 | } | |
1683 | ||
1684 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 KW |
1685 | AIO_WAIT_WHILE(blk_get_aio_context(blk), |
1686 | atomic_mb_read(&blk->in_flight) > 0); | |
33f2a757 SH |
1687 | |
1688 | if (bs) { | |
1689 | bdrv_drained_end(bs); | |
a46fc9c9 | 1690 | } |
97b0385a AY |
1691 | } |
1692 | ||
4be74634 MA |
1693 | void blk_drain_all(void) |
1694 | { | |
33f2a757 SH |
1695 | BlockBackend *blk = NULL; |
1696 | ||
1697 | bdrv_drain_all_begin(); | |
1698 | ||
1699 | while ((blk = blk_all_next(blk)) != NULL) { | |
1700 | AioContext *ctx = blk_get_aio_context(blk); | |
1701 | ||
1702 | aio_context_acquire(ctx); | |
1703 | ||
1704 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 | 1705 | AIO_WAIT_WHILE(ctx, atomic_mb_read(&blk->in_flight) > 0); |
33f2a757 SH |
1706 | |
1707 | aio_context_release(ctx); | |
1708 | } | |
1709 | ||
1710 | bdrv_drain_all_end(); | |
4be74634 MA |
1711 | } |
1712 | ||
373340b2 HR |
1713 | void blk_set_on_error(BlockBackend *blk, BlockdevOnError on_read_error, |
1714 | BlockdevOnError on_write_error) | |
1715 | { | |
1716 | blk->on_read_error = on_read_error; | |
1717 | blk->on_write_error = on_write_error; | |
1718 | } | |
1719 | ||
4be74634 MA |
1720 | BlockdevOnError blk_get_on_error(BlockBackend *blk, bool is_read) |
1721 | { | |
373340b2 | 1722 | return is_read ? blk->on_read_error : blk->on_write_error; |
4be74634 MA |
1723 | } |
1724 | ||
1725 | BlockErrorAction blk_get_error_action(BlockBackend *blk, bool is_read, | |
1726 | int error) | |
1727 | { | |
373340b2 HR |
1728 | BlockdevOnError on_err = blk_get_on_error(blk, is_read); |
1729 | ||
1730 | switch (on_err) { | |
1731 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
1732 | return (error == ENOSPC) ? | |
1733 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
1734 | case BLOCKDEV_ON_ERROR_STOP: | |
1735 | return BLOCK_ERROR_ACTION_STOP; | |
1736 | case BLOCKDEV_ON_ERROR_REPORT: | |
1737 | return BLOCK_ERROR_ACTION_REPORT; | |
1738 | case BLOCKDEV_ON_ERROR_IGNORE: | |
1739 | return BLOCK_ERROR_ACTION_IGNORE; | |
8c398252 | 1740 | case BLOCKDEV_ON_ERROR_AUTO: |
373340b2 HR |
1741 | default: |
1742 | abort(); | |
1743 | } | |
4be74634 MA |
1744 | } |
1745 | ||
373340b2 HR |
1746 | static void send_qmp_error_event(BlockBackend *blk, |
1747 | BlockErrorAction action, | |
1748 | bool is_read, int error) | |
1749 | { | |
1750 | IoOperationType optype; | |
bfe1a14c | 1751 | BlockDriverState *bs = blk_bs(blk); |
373340b2 HR |
1752 | |
1753 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; | |
bfe1a14c KW |
1754 | qapi_event_send_block_io_error(blk_name(blk), !!bs, |
1755 | bs ? bdrv_get_node_name(bs) : NULL, optype, | |
2bf7e10f | 1756 | action, blk_iostatus_is_enabled(blk), |
3ab72385 | 1757 | error == ENOSPC, strerror(error)); |
373340b2 HR |
1758 | } |
1759 | ||
1760 | /* This is done by device models because, while the block layer knows | |
1761 | * about the error, it does not know whether an operation comes from | |
1762 | * the device or the block layer (from a job, for example). | |
1763 | */ | |
4be74634 MA |
1764 | void blk_error_action(BlockBackend *blk, BlockErrorAction action, |
1765 | bool is_read, int error) | |
1766 | { | |
373340b2 HR |
1767 | assert(error >= 0); |
1768 | ||
1769 | if (action == BLOCK_ERROR_ACTION_STOP) { | |
1770 | /* First set the iostatus, so that "info block" returns an iostatus | |
1771 | * that matches the events raised so far (an additional error iostatus | |
1772 | * is fine, but not a lost one). | |
1773 | */ | |
1774 | blk_iostatus_set_err(blk, error); | |
1775 | ||
1776 | /* Then raise the request to stop the VM and the event. | |
1777 | * qemu_system_vmstop_request_prepare has two effects. First, | |
1778 | * it ensures that the STOP event always comes after the | |
1779 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
1780 | * can observe the STOP event and do a "cont" before the STOP | |
1781 | * event is issued, the VM will not stop. In this case, vm_start() | |
1782 | * also ensures that the STOP/RESUME pair of events is emitted. | |
1783 | */ | |
1784 | qemu_system_vmstop_request_prepare(); | |
1785 | send_qmp_error_event(blk, action, is_read, error); | |
1786 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); | |
1787 | } else { | |
1788 | send_qmp_error_event(blk, action, is_read, error); | |
1789 | } | |
4be74634 MA |
1790 | } |
1791 | ||
96710565 | 1792 | bool blk_is_read_only(BlockBackend *blk) |
4be74634 | 1793 | { |
f21d96d0 KW |
1794 | BlockDriverState *bs = blk_bs(blk); |
1795 | ||
1796 | if (bs) { | |
1797 | return bdrv_is_read_only(bs); | |
061959e8 HR |
1798 | } else { |
1799 | return blk->root_state.read_only; | |
1800 | } | |
4be74634 MA |
1801 | } |
1802 | ||
96710565 | 1803 | bool blk_is_sg(BlockBackend *blk) |
4be74634 | 1804 | { |
f21d96d0 KW |
1805 | BlockDriverState *bs = blk_bs(blk); |
1806 | ||
1807 | if (!bs) { | |
96710565 | 1808 | return false; |
a46fc9c9 HR |
1809 | } |
1810 | ||
f21d96d0 | 1811 | return bdrv_is_sg(bs); |
4be74634 MA |
1812 | } |
1813 | ||
96710565 | 1814 | bool blk_enable_write_cache(BlockBackend *blk) |
4be74634 | 1815 | { |
bfd18d1e | 1816 | return blk->enable_write_cache; |
4be74634 MA |
1817 | } |
1818 | ||
1819 | void blk_set_enable_write_cache(BlockBackend *blk, bool wce) | |
1820 | { | |
bfd18d1e | 1821 | blk->enable_write_cache = wce; |
4be74634 MA |
1822 | } |
1823 | ||
2bb0dce7 HR |
1824 | void blk_invalidate_cache(BlockBackend *blk, Error **errp) |
1825 | { | |
f21d96d0 KW |
1826 | BlockDriverState *bs = blk_bs(blk); |
1827 | ||
1828 | if (!bs) { | |
c09ba36c HR |
1829 | error_setg(errp, "Device '%s' has no medium", blk->name); |
1830 | return; | |
1831 | } | |
1832 | ||
f21d96d0 | 1833 | bdrv_invalidate_cache(bs, errp); |
2bb0dce7 HR |
1834 | } |
1835 | ||
e031f750 | 1836 | bool blk_is_inserted(BlockBackend *blk) |
4be74634 | 1837 | { |
f21d96d0 KW |
1838 | BlockDriverState *bs = blk_bs(blk); |
1839 | ||
1840 | return bs && bdrv_is_inserted(bs); | |
db0284f8 HR |
1841 | } |
1842 | ||
1843 | bool blk_is_available(BlockBackend *blk) | |
1844 | { | |
1845 | return blk_is_inserted(blk) && !blk_dev_is_tray_open(blk); | |
4be74634 MA |
1846 | } |
1847 | ||
1848 | void blk_lock_medium(BlockBackend *blk, bool locked) | |
1849 | { | |
f21d96d0 KW |
1850 | BlockDriverState *bs = blk_bs(blk); |
1851 | ||
1852 | if (bs) { | |
1853 | bdrv_lock_medium(bs, locked); | |
a46fc9c9 | 1854 | } |
4be74634 MA |
1855 | } |
1856 | ||
1857 | void blk_eject(BlockBackend *blk, bool eject_flag) | |
1858 | { | |
f21d96d0 | 1859 | BlockDriverState *bs = blk_bs(blk); |
2d76e724 KW |
1860 | char *id; |
1861 | ||
f21d96d0 KW |
1862 | if (bs) { |
1863 | bdrv_eject(bs, eject_flag); | |
a46fc9c9 | 1864 | } |
c47ee043 JS |
1865 | |
1866 | /* Whether or not we ejected on the backend, | |
1867 | * the frontend experienced a tray event. */ | |
1868 | id = blk_get_attached_dev_id(blk); | |
1869 | qapi_event_send_device_tray_moved(blk_name(blk), id, | |
3ab72385 | 1870 | eject_flag); |
c47ee043 | 1871 | g_free(id); |
4be74634 MA |
1872 | } |
1873 | ||
1874 | int blk_get_flags(BlockBackend *blk) | |
1875 | { | |
f21d96d0 KW |
1876 | BlockDriverState *bs = blk_bs(blk); |
1877 | ||
1878 | if (bs) { | |
1879 | return bdrv_get_flags(bs); | |
061959e8 HR |
1880 | } else { |
1881 | return blk->root_state.open_flags; | |
1882 | } | |
4be74634 MA |
1883 | } |
1884 | ||
4841211e EB |
1885 | /* Returns the minimum request alignment, in bytes; guaranteed nonzero */ |
1886 | uint32_t blk_get_request_alignment(BlockBackend *blk) | |
1887 | { | |
1888 | BlockDriverState *bs = blk_bs(blk); | |
1889 | return bs ? bs->bl.request_alignment : BDRV_SECTOR_SIZE; | |
1890 | } | |
1891 | ||
5def6b80 EB |
1892 | /* Returns the maximum transfer length, in bytes; guaranteed nonzero */ |
1893 | uint32_t blk_get_max_transfer(BlockBackend *blk) | |
454057b7 | 1894 | { |
f21d96d0 | 1895 | BlockDriverState *bs = blk_bs(blk); |
5def6b80 | 1896 | uint32_t max = 0; |
f21d96d0 KW |
1897 | |
1898 | if (bs) { | |
5def6b80 | 1899 | max = bs->bl.max_transfer; |
a46fc9c9 | 1900 | } |
5def6b80 | 1901 | return MIN_NON_ZERO(max, INT_MAX); |
454057b7 PL |
1902 | } |
1903 | ||
648296e0 SH |
1904 | int blk_get_max_iov(BlockBackend *blk) |
1905 | { | |
f21d96d0 | 1906 | return blk->root->bs->bl.max_iov; |
648296e0 SH |
1907 | } |
1908 | ||
4be74634 MA |
1909 | void blk_set_guest_block_size(BlockBackend *blk, int align) |
1910 | { | |
68e9ec01 | 1911 | blk->guest_block_size = align; |
4be74634 MA |
1912 | } |
1913 | ||
f1c17521 PB |
1914 | void *blk_try_blockalign(BlockBackend *blk, size_t size) |
1915 | { | |
f21d96d0 | 1916 | return qemu_try_blockalign(blk ? blk_bs(blk) : NULL, size); |
f1c17521 PB |
1917 | } |
1918 | ||
4be74634 MA |
1919 | void *blk_blockalign(BlockBackend *blk, size_t size) |
1920 | { | |
f21d96d0 | 1921 | return qemu_blockalign(blk ? blk_bs(blk) : NULL, size); |
4be74634 MA |
1922 | } |
1923 | ||
1924 | bool blk_op_is_blocked(BlockBackend *blk, BlockOpType op, Error **errp) | |
1925 | { | |
f21d96d0 KW |
1926 | BlockDriverState *bs = blk_bs(blk); |
1927 | ||
1928 | if (!bs) { | |
a46fc9c9 HR |
1929 | return false; |
1930 | } | |
1931 | ||
f21d96d0 | 1932 | return bdrv_op_is_blocked(bs, op, errp); |
4be74634 MA |
1933 | } |
1934 | ||
1935 | void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason) | |
1936 | { | |
f21d96d0 KW |
1937 | BlockDriverState *bs = blk_bs(blk); |
1938 | ||
1939 | if (bs) { | |
1940 | bdrv_op_unblock(bs, op, reason); | |
a46fc9c9 | 1941 | } |
4be74634 MA |
1942 | } |
1943 | ||
1944 | void blk_op_block_all(BlockBackend *blk, Error *reason) | |
1945 | { | |
f21d96d0 KW |
1946 | BlockDriverState *bs = blk_bs(blk); |
1947 | ||
1948 | if (bs) { | |
1949 | bdrv_op_block_all(bs, reason); | |
a46fc9c9 | 1950 | } |
4be74634 MA |
1951 | } |
1952 | ||
1953 | void blk_op_unblock_all(BlockBackend *blk, Error *reason) | |
1954 | { | |
f21d96d0 KW |
1955 | BlockDriverState *bs = blk_bs(blk); |
1956 | ||
1957 | if (bs) { | |
1958 | bdrv_op_unblock_all(bs, reason); | |
a46fc9c9 | 1959 | } |
4be74634 MA |
1960 | } |
1961 | ||
1962 | AioContext *blk_get_aio_context(BlockBackend *blk) | |
1963 | { | |
d861ab3a KW |
1964 | BlockDriverState *bs = blk_bs(blk); |
1965 | ||
d861ab3a | 1966 | if (bs) { |
132ada80 KW |
1967 | AioContext *ctx = bdrv_get_aio_context(blk_bs(blk)); |
1968 | assert(ctx == blk->ctx); | |
d861ab3a KW |
1969 | } |
1970 | ||
1971 | return blk->ctx; | |
4981bdec HR |
1972 | } |
1973 | ||
1974 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) | |
1975 | { | |
1976 | BlockBackendAIOCB *blk_acb = DO_UPCAST(BlockBackendAIOCB, common, acb); | |
1977 | return blk_get_aio_context(blk_acb->blk); | |
4be74634 MA |
1978 | } |
1979 | ||
97896a48 KW |
1980 | static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context, |
1981 | bool update_root_node, Error **errp) | |
4be74634 | 1982 | { |
f21d96d0 | 1983 | BlockDriverState *bs = blk_bs(blk); |
c61791fc | 1984 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
97896a48 | 1985 | int ret; |
f21d96d0 KW |
1986 | |
1987 | if (bs) { | |
97896a48 KW |
1988 | if (update_root_node) { |
1989 | ret = bdrv_child_try_set_aio_context(bs, new_context, blk->root, | |
1990 | errp); | |
1991 | if (ret < 0) { | |
1992 | return ret; | |
1993 | } | |
1994 | } | |
c61791fc | 1995 | if (tgm->throttle_state) { |
dc868fb0 | 1996 | bdrv_drained_begin(bs); |
c61791fc MP |
1997 | throttle_group_detach_aio_context(tgm); |
1998 | throttle_group_attach_aio_context(tgm, new_context); | |
dc868fb0 | 1999 | bdrv_drained_end(bs); |
7ca7f0f6 | 2000 | } |
a46fc9c9 | 2001 | } |
97896a48 | 2002 | |
d861ab3a | 2003 | blk->ctx = new_context; |
97896a48 | 2004 | return 0; |
4be74634 MA |
2005 | } |
2006 | ||
97896a48 KW |
2007 | int blk_set_aio_context(BlockBackend *blk, AioContext *new_context, |
2008 | Error **errp) | |
38475269 | 2009 | { |
97896a48 | 2010 | return blk_do_set_aio_context(blk, new_context, true, errp); |
38475269 KW |
2011 | } |
2012 | ||
2013 | static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2014 | GSList **ignore, Error **errp) | |
2015 | { | |
2016 | BlockBackend *blk = child->opaque; | |
2017 | ||
980b0f94 KW |
2018 | if (blk->allow_aio_context_change) { |
2019 | return true; | |
2020 | } | |
2021 | ||
38475269 KW |
2022 | /* Only manually created BlockBackends that are not attached to anything |
2023 | * can change their AioContext without updating their user. */ | |
2024 | if (!blk->name || blk->dev) { | |
2025 | /* TODO Add BB name/QOM path */ | |
2026 | error_setg(errp, "Cannot change iothread of active block backend"); | |
2027 | return false; | |
2028 | } | |
2029 | ||
2030 | return true; | |
2031 | } | |
2032 | ||
2033 | static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx, | |
2034 | GSList **ignore) | |
2035 | { | |
2036 | BlockBackend *blk = child->opaque; | |
97896a48 | 2037 | blk_do_set_aio_context(blk, ctx, false, &error_abort); |
38475269 KW |
2038 | } |
2039 | ||
2019ba0a HR |
2040 | void blk_add_aio_context_notifier(BlockBackend *blk, |
2041 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
2042 | void (*detach_aio_context)(void *opaque), void *opaque) | |
2043 | { | |
d03654ea | 2044 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2045 | BlockDriverState *bs = blk_bs(blk); |
2046 | ||
d03654ea SH |
2047 | notifier = g_new(BlockBackendAioNotifier, 1); |
2048 | notifier->attached_aio_context = attached_aio_context; | |
2049 | notifier->detach_aio_context = detach_aio_context; | |
2050 | notifier->opaque = opaque; | |
2051 | QLIST_INSERT_HEAD(&blk->aio_notifiers, notifier, list); | |
2052 | ||
f21d96d0 KW |
2053 | if (bs) { |
2054 | bdrv_add_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2055 | detach_aio_context, opaque); |
2056 | } | |
2019ba0a HR |
2057 | } |
2058 | ||
2059 | void blk_remove_aio_context_notifier(BlockBackend *blk, | |
2060 | void (*attached_aio_context)(AioContext *, | |
2061 | void *), | |
2062 | void (*detach_aio_context)(void *), | |
2063 | void *opaque) | |
2064 | { | |
d03654ea | 2065 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
2066 | BlockDriverState *bs = blk_bs(blk); |
2067 | ||
2068 | if (bs) { | |
2069 | bdrv_remove_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
2070 | detach_aio_context, opaque); |
2071 | } | |
d03654ea SH |
2072 | |
2073 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
2074 | if (notifier->attached_aio_context == attached_aio_context && | |
2075 | notifier->detach_aio_context == detach_aio_context && | |
2076 | notifier->opaque == opaque) { | |
2077 | QLIST_REMOVE(notifier, list); | |
2078 | g_free(notifier); | |
2079 | return; | |
2080 | } | |
2081 | } | |
2082 | ||
2083 | abort(); | |
2019ba0a HR |
2084 | } |
2085 | ||
3301f6c6 HR |
2086 | void blk_add_remove_bs_notifier(BlockBackend *blk, Notifier *notify) |
2087 | { | |
2088 | notifier_list_add(&blk->remove_bs_notifiers, notify); | |
2089 | } | |
2090 | ||
2091 | void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify) | |
2092 | { | |
2093 | notifier_list_add(&blk->insert_bs_notifiers, notify); | |
2094 | } | |
2095 | ||
4be74634 MA |
2096 | void blk_io_plug(BlockBackend *blk) |
2097 | { | |
f21d96d0 KW |
2098 | BlockDriverState *bs = blk_bs(blk); |
2099 | ||
2100 | if (bs) { | |
2101 | bdrv_io_plug(bs); | |
a46fc9c9 | 2102 | } |
4be74634 MA |
2103 | } |
2104 | ||
2105 | void blk_io_unplug(BlockBackend *blk) | |
2106 | { | |
f21d96d0 KW |
2107 | BlockDriverState *bs = blk_bs(blk); |
2108 | ||
2109 | if (bs) { | |
2110 | bdrv_io_unplug(bs); | |
a46fc9c9 | 2111 | } |
4be74634 MA |
2112 | } |
2113 | ||
2114 | BlockAcctStats *blk_get_stats(BlockBackend *blk) | |
2115 | { | |
7f0e9da6 | 2116 | return &blk->stats; |
4be74634 MA |
2117 | } |
2118 | ||
2119 | void *blk_aio_get(const AIOCBInfo *aiocb_info, BlockBackend *blk, | |
2120 | BlockCompletionFunc *cb, void *opaque) | |
2121 | { | |
2122 | return qemu_aio_get(aiocb_info, blk_bs(blk), cb, opaque); | |
2123 | } | |
1ef01253 | 2124 | |
d004bd52 | 2125 | int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 2126 | int bytes, BdrvRequestFlags flags) |
1ef01253 | 2127 | { |
f5a5ca79 | 2128 | return blk_co_pwritev(blk, offset, bytes, NULL, |
16aaf975 | 2129 | flags | BDRV_REQ_ZERO_WRITE); |
1ef01253 HR |
2130 | } |
2131 | ||
fe5c1355 PB |
2132 | int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, |
2133 | int count) | |
1ef01253 | 2134 | { |
35fadca8 PB |
2135 | return blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
2136 | BDRV_REQ_WRITE_COMPRESSED); | |
1ef01253 HR |
2137 | } |
2138 | ||
c80d8b06 | 2139 | int blk_truncate(BlockBackend *blk, int64_t offset, bool exact, |
8c6242b6 | 2140 | PreallocMode prealloc, BdrvRequestFlags flags, Error **errp) |
1ef01253 | 2141 | { |
c09ba36c | 2142 | if (!blk_is_available(blk)) { |
ed3d2ec9 | 2143 | error_setg(errp, "No medium inserted"); |
c09ba36c HR |
2144 | return -ENOMEDIUM; |
2145 | } | |
2146 | ||
8c6242b6 | 2147 | return bdrv_truncate(blk->root, offset, exact, prealloc, flags, errp); |
1ef01253 HR |
2148 | } |
2149 | ||
1ef01253 HR |
2150 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, |
2151 | int64_t pos, int size) | |
2152 | { | |
bfd18d1e KW |
2153 | int ret; |
2154 | ||
c09ba36c HR |
2155 | if (!blk_is_available(blk)) { |
2156 | return -ENOMEDIUM; | |
2157 | } | |
2158 | ||
bfd18d1e KW |
2159 | ret = bdrv_save_vmstate(blk_bs(blk), buf, pos, size); |
2160 | if (ret < 0) { | |
2161 | return ret; | |
2162 | } | |
2163 | ||
2164 | if (ret == size && !blk->enable_write_cache) { | |
2165 | ret = bdrv_flush(blk_bs(blk)); | |
2166 | } | |
2167 | ||
2168 | return ret < 0 ? ret : size; | |
1ef01253 HR |
2169 | } |
2170 | ||
2171 | int blk_load_vmstate(BlockBackend *blk, uint8_t *buf, int64_t pos, int size) | |
2172 | { | |
c09ba36c HR |
2173 | if (!blk_is_available(blk)) { |
2174 | return -ENOMEDIUM; | |
2175 | } | |
2176 | ||
f21d96d0 | 2177 | return bdrv_load_vmstate(blk_bs(blk), buf, pos, size); |
1ef01253 | 2178 | } |
f0272c4d ET |
2179 | |
2180 | int blk_probe_blocksizes(BlockBackend *blk, BlockSizes *bsz) | |
2181 | { | |
c09ba36c HR |
2182 | if (!blk_is_available(blk)) { |
2183 | return -ENOMEDIUM; | |
2184 | } | |
2185 | ||
f21d96d0 | 2186 | return bdrv_probe_blocksizes(blk_bs(blk), bsz); |
f0272c4d ET |
2187 | } |
2188 | ||
2189 | int blk_probe_geometry(BlockBackend *blk, HDGeometry *geo) | |
2190 | { | |
c09ba36c HR |
2191 | if (!blk_is_available(blk)) { |
2192 | return -ENOMEDIUM; | |
2193 | } | |
2194 | ||
f21d96d0 | 2195 | return bdrv_probe_geometry(blk_bs(blk), geo); |
f0272c4d | 2196 | } |
281d22d8 HR |
2197 | |
2198 | /* | |
2199 | * Updates the BlockBackendRootState object with data from the currently | |
2200 | * attached BlockDriverState. | |
2201 | */ | |
2202 | void blk_update_root_state(BlockBackend *blk) | |
2203 | { | |
f21d96d0 | 2204 | assert(blk->root); |
281d22d8 | 2205 | |
f21d96d0 KW |
2206 | blk->root_state.open_flags = blk->root->bs->open_flags; |
2207 | blk->root_state.read_only = blk->root->bs->read_only; | |
2208 | blk->root_state.detect_zeroes = blk->root->bs->detect_zeroes; | |
281d22d8 HR |
2209 | } |
2210 | ||
38cb18f5 | 2211 | /* |
b85114f8 KW |
2212 | * Returns the detect-zeroes setting to be used for bdrv_open() of a |
2213 | * BlockDriverState which is supposed to inherit the root state. | |
38cb18f5 | 2214 | */ |
b85114f8 | 2215 | bool blk_get_detect_zeroes_from_root_state(BlockBackend *blk) |
38cb18f5 | 2216 | { |
b85114f8 | 2217 | return blk->root_state.detect_zeroes; |
38cb18f5 HR |
2218 | } |
2219 | ||
2220 | /* | |
2221 | * Returns the flags to be used for bdrv_open() of a BlockDriverState which is | |
2222 | * supposed to inherit the root state. | |
2223 | */ | |
2224 | int blk_get_open_flags_from_root_state(BlockBackend *blk) | |
2225 | { | |
2226 | int bs_flags; | |
2227 | ||
2228 | bs_flags = blk->root_state.read_only ? 0 : BDRV_O_RDWR; | |
2229 | bs_flags |= blk->root_state.open_flags & ~BDRV_O_RDWR; | |
2230 | ||
2231 | return bs_flags; | |
2232 | } | |
2233 | ||
281d22d8 HR |
2234 | BlockBackendRootState *blk_get_root_state(BlockBackend *blk) |
2235 | { | |
2236 | return &blk->root_state; | |
2237 | } | |
1393f212 HR |
2238 | |
2239 | int blk_commit_all(void) | |
2240 | { | |
fe1a9cbc HR |
2241 | BlockBackend *blk = NULL; |
2242 | ||
2243 | while ((blk = blk_all_next(blk)) != NULL) { | |
2244 | AioContext *aio_context = blk_get_aio_context(blk); | |
2245 | ||
2246 | aio_context_acquire(aio_context); | |
f21d96d0 KW |
2247 | if (blk_is_inserted(blk) && blk->root->bs->backing) { |
2248 | int ret = bdrv_commit(blk->root->bs); | |
fe1a9cbc HR |
2249 | if (ret < 0) { |
2250 | aio_context_release(aio_context); | |
2251 | return ret; | |
2252 | } | |
2253 | } | |
2254 | aio_context_release(aio_context); | |
2255 | } | |
2256 | return 0; | |
2257 | } | |
2258 | ||
97148076 KW |
2259 | |
2260 | /* throttling disk I/O limits */ | |
2261 | void blk_set_io_limits(BlockBackend *blk, ThrottleConfig *cfg) | |
2262 | { | |
022cdc9f | 2263 | throttle_group_config(&blk->public.throttle_group_member, cfg); |
97148076 KW |
2264 | } |
2265 | ||
2266 | void blk_io_limits_disable(BlockBackend *blk) | |
2267 | { | |
48bf7ea8 AG |
2268 | BlockDriverState *bs = blk_bs(blk); |
2269 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; | |
2270 | assert(tgm->throttle_state); | |
2271 | if (bs) { | |
2272 | bdrv_drained_begin(bs); | |
2273 | } | |
2274 | throttle_group_unregister_tgm(tgm); | |
2275 | if (bs) { | |
2276 | bdrv_drained_end(bs); | |
2277 | } | |
97148076 KW |
2278 | } |
2279 | ||
2280 | /* should be called before blk_set_io_limits if a limit is set */ | |
2281 | void blk_io_limits_enable(BlockBackend *blk, const char *group) | |
2282 | { | |
022cdc9f | 2283 | assert(!blk->public.throttle_group_member.throttle_state); |
c61791fc MP |
2284 | throttle_group_register_tgm(&blk->public.throttle_group_member, |
2285 | group, blk_get_aio_context(blk)); | |
97148076 KW |
2286 | } |
2287 | ||
2288 | void blk_io_limits_update_group(BlockBackend *blk, const char *group) | |
2289 | { | |
2290 | /* this BB is not part of any group */ | |
022cdc9f | 2291 | if (!blk->public.throttle_group_member.throttle_state) { |
97148076 KW |
2292 | return; |
2293 | } | |
2294 | ||
2295 | /* this BB is a part of the same group than the one we want */ | |
022cdc9f MP |
2296 | if (!g_strcmp0(throttle_group_get_name(&blk->public.throttle_group_member), |
2297 | group)) { | |
97148076 KW |
2298 | return; |
2299 | } | |
2300 | ||
2301 | /* need to change the group this bs belong to */ | |
2302 | blk_io_limits_disable(blk); | |
2303 | blk_io_limits_enable(blk, group); | |
2304 | } | |
c2066af0 KW |
2305 | |
2306 | static void blk_root_drained_begin(BdrvChild *child) | |
2307 | { | |
2308 | BlockBackend *blk = child->opaque; | |
2309 | ||
f4d9cc88 JS |
2310 | if (++blk->quiesce_counter == 1) { |
2311 | if (blk->dev_ops && blk->dev_ops->drained_begin) { | |
2312 | blk->dev_ops->drained_begin(blk->dev_opaque); | |
2313 | } | |
2314 | } | |
2315 | ||
36fe1331 KW |
2316 | /* Note that blk->root may not be accessible here yet if we are just |
2317 | * attaching to a BlockDriverState that is drained. Use child instead. */ | |
2318 | ||
022cdc9f MP |
2319 | if (atomic_fetch_inc(&blk->public.throttle_group_member.io_limits_disabled) == 0) { |
2320 | throttle_group_restart_tgm(&blk->public.throttle_group_member); | |
c2066af0 KW |
2321 | } |
2322 | } | |
2323 | ||
fe5258a5 KW |
2324 | static bool blk_root_drained_poll(BdrvChild *child) |
2325 | { | |
2326 | BlockBackend *blk = child->opaque; | |
2327 | assert(blk->quiesce_counter); | |
2328 | return !!blk->in_flight; | |
2329 | } | |
2330 | ||
e037c09c | 2331 | static void blk_root_drained_end(BdrvChild *child, int *drained_end_counter) |
c2066af0 KW |
2332 | { |
2333 | BlockBackend *blk = child->opaque; | |
f4d9cc88 | 2334 | assert(blk->quiesce_counter); |
c2066af0 | 2335 | |
022cdc9f MP |
2336 | assert(blk->public.throttle_group_member.io_limits_disabled); |
2337 | atomic_dec(&blk->public.throttle_group_member.io_limits_disabled); | |
f4d9cc88 JS |
2338 | |
2339 | if (--blk->quiesce_counter == 0) { | |
2340 | if (blk->dev_ops && blk->dev_ops->drained_end) { | |
2341 | blk->dev_ops->drained_end(blk->dev_opaque); | |
2342 | } | |
cf312932 KW |
2343 | while (qemu_co_enter_next(&blk->queued_requests, NULL)) { |
2344 | /* Resume all queued requests */ | |
2345 | } | |
f4d9cc88 | 2346 | } |
c2066af0 | 2347 | } |
23d0ba93 FZ |
2348 | |
2349 | void blk_register_buf(BlockBackend *blk, void *host, size_t size) | |
2350 | { | |
2351 | bdrv_register_buf(blk_bs(blk), host, size); | |
2352 | } | |
2353 | ||
2354 | void blk_unregister_buf(BlockBackend *blk, void *host) | |
2355 | { | |
2356 | bdrv_unregister_buf(blk_bs(blk), host); | |
2357 | } | |
b5679fa4 FZ |
2358 | |
2359 | int coroutine_fn blk_co_copy_range(BlockBackend *blk_in, int64_t off_in, | |
2360 | BlockBackend *blk_out, int64_t off_out, | |
67b51fb9 VSO |
2361 | int bytes, BdrvRequestFlags read_flags, |
2362 | BdrvRequestFlags write_flags) | |
b5679fa4 FZ |
2363 | { |
2364 | int r; | |
2365 | r = blk_check_byte_request(blk_in, off_in, bytes); | |
2366 | if (r) { | |
2367 | return r; | |
2368 | } | |
2369 | r = blk_check_byte_request(blk_out, off_out, bytes); | |
2370 | if (r) { | |
2371 | return r; | |
2372 | } | |
2373 | return bdrv_co_copy_range(blk_in->root, off_in, | |
2374 | blk_out->root, off_out, | |
67b51fb9 | 2375 | bytes, read_flags, write_flags); |
b5679fa4 | 2376 | } |
5d3b4e99 VSO |
2377 | |
2378 | const BdrvChild *blk_root(BlockBackend *blk) | |
2379 | { | |
2380 | return blk->root; | |
2381 | } |