]>
Commit | Line | Data |
---|---|---|
26f54e9a MA |
1 | /* |
2 | * QEMU Block backends | |
3 | * | |
60cb2fa7 | 4 | * Copyright (C) 2014-2016 Red Hat, Inc. |
26f54e9a MA |
5 | * |
6 | * Authors: | |
7 | * Markus Armbruster <armbru@redhat.com>, | |
8 | * | |
9 | * This work is licensed under the terms of the GNU LGPL, version 2.1 | |
10 | * or later. See the COPYING.LIB file in the top-level directory. | |
11 | */ | |
12 | ||
80c71a24 | 13 | #include "qemu/osdep.h" |
26f54e9a MA |
14 | #include "sysemu/block-backend.h" |
15 | #include "block/block_int.h" | |
373340b2 | 16 | #include "block/blockjob.h" |
281d22d8 | 17 | #include "block/throttle-groups.h" |
18e46a03 | 18 | #include "sysemu/blockdev.h" |
373340b2 | 19 | #include "sysemu/sysemu.h" |
e688df6b | 20 | #include "qapi/error.h" |
9af23989 | 21 | #include "qapi/qapi-events-block.h" |
f348b6d1 | 22 | #include "qemu/id.h" |
922a01a0 | 23 | #include "qemu/option.h" |
1e98fefd | 24 | #include "trace.h" |
5f7772c4 | 25 | #include "migration/misc.h" |
a7f53e26 MA |
26 | |
27 | /* Number of coroutines to reserve per attached device model */ | |
28 | #define COROUTINE_POOL_RESERVATION 64 | |
26f54e9a | 29 | |
1bf1cbc9 KW |
30 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
31 | ||
4981bdec HR |
32 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb); |
33 | ||
d03654ea SH |
34 | typedef struct BlockBackendAioNotifier { |
35 | void (*attached_aio_context)(AioContext *new_context, void *opaque); | |
36 | void (*detach_aio_context)(void *opaque); | |
37 | void *opaque; | |
38 | QLIST_ENTRY(BlockBackendAioNotifier) list; | |
39 | } BlockBackendAioNotifier; | |
40 | ||
26f54e9a MA |
41 | struct BlockBackend { |
42 | char *name; | |
43 | int refcnt; | |
f21d96d0 | 44 | BdrvChild *root; |
26f8b3a8 | 45 | DriveInfo *legacy_dinfo; /* null unless created by drive_new() */ |
2cf22d6a | 46 | QTAILQ_ENTRY(BlockBackend) link; /* for block_backends */ |
9492b0b9 | 47 | QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */ |
f2cd875d | 48 | BlockBackendPublic public; |
a7f53e26 | 49 | |
d09ea2d2 | 50 | DeviceState *dev; /* attached device model, if any */ |
a7f53e26 MA |
51 | const BlockDevOps *dev_ops; |
52 | void *dev_opaque; | |
68e9ec01 HR |
53 | |
54 | /* the block size for which the guest device expects atomicity */ | |
55 | int guest_block_size; | |
7f0e9da6 | 56 | |
281d22d8 HR |
57 | /* If the BDS tree is removed, some of its options are stored here (which |
58 | * can be used to restore those options in the new BDS on insert) */ | |
59 | BlockBackendRootState root_state; | |
60 | ||
bfd18d1e KW |
61 | bool enable_write_cache; |
62 | ||
7f0e9da6 HR |
63 | /* I/O stats (display with "info blockstats"). */ |
64 | BlockAcctStats stats; | |
373340b2 HR |
65 | |
66 | BlockdevOnError on_read_error, on_write_error; | |
67 | bool iostatus_enabled; | |
68 | BlockDeviceIoStatus iostatus; | |
3301f6c6 | 69 | |
981776b3 KW |
70 | uint64_t perm; |
71 | uint64_t shared_perm; | |
d35ff5e6 | 72 | bool disable_perm; |
981776b3 | 73 | |
c10c9d96 KW |
74 | bool allow_write_beyond_eof; |
75 | ||
3301f6c6 | 76 | NotifierList remove_bs_notifiers, insert_bs_notifiers; |
d03654ea | 77 | QLIST_HEAD(, BlockBackendAioNotifier) aio_notifiers; |
f4d9cc88 JS |
78 | |
79 | int quiesce_counter; | |
5f7772c4 | 80 | VMChangeStateEntry *vmsh; |
ca2e2144 | 81 | bool force_allow_inactivate; |
33f2a757 SH |
82 | |
83 | /* Number of in-flight aio requests. BlockDriverState also counts | |
84 | * in-flight requests but aio requests can exist even when blk->root is | |
85 | * NULL, so we cannot rely on its counter for that case. | |
86 | * Accessed with atomic ops. | |
87 | */ | |
88 | unsigned int in_flight; | |
26f54e9a MA |
89 | }; |
90 | ||
e7f7d676 HR |
91 | typedef struct BlockBackendAIOCB { |
92 | BlockAIOCB common; | |
4981bdec | 93 | BlockBackend *blk; |
e7f7d676 HR |
94 | int ret; |
95 | } BlockBackendAIOCB; | |
96 | ||
97 | static const AIOCBInfo block_backend_aiocb_info = { | |
4981bdec | 98 | .get_aio_context = blk_aiocb_get_aio_context, |
e7f7d676 HR |
99 | .aiocb_size = sizeof(BlockBackendAIOCB), |
100 | }; | |
101 | ||
8fb3c76c | 102 | static void drive_info_del(DriveInfo *dinfo); |
7c8eece4 | 103 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs); |
8fb3c76c | 104 | |
2cf22d6a HR |
105 | /* All BlockBackends */ |
106 | static QTAILQ_HEAD(, BlockBackend) block_backends = | |
107 | QTAILQ_HEAD_INITIALIZER(block_backends); | |
108 | ||
9492b0b9 HR |
109 | /* All BlockBackends referenced by the monitor and which are iterated through by |
110 | * blk_next() */ | |
111 | static QTAILQ_HEAD(, BlockBackend) monitor_block_backends = | |
112 | QTAILQ_HEAD_INITIALIZER(monitor_block_backends); | |
26f54e9a | 113 | |
f21d96d0 KW |
114 | static void blk_root_inherit_options(int *child_flags, QDict *child_options, |
115 | int parent_flags, QDict *parent_options) | |
116 | { | |
117 | /* We're not supposed to call this function for root nodes */ | |
118 | abort(); | |
119 | } | |
c2066af0 | 120 | static void blk_root_drained_begin(BdrvChild *child); |
fe5258a5 | 121 | static bool blk_root_drained_poll(BdrvChild *child); |
c2066af0 | 122 | static void blk_root_drained_end(BdrvChild *child); |
f21d96d0 | 123 | |
5c8cab48 KW |
124 | static void blk_root_change_media(BdrvChild *child, bool load); |
125 | static void blk_root_resize(BdrvChild *child); | |
126 | ||
b5411555 KW |
127 | static char *blk_root_get_parent_desc(BdrvChild *child) |
128 | { | |
129 | BlockBackend *blk = child->opaque; | |
130 | char *dev_id; | |
131 | ||
132 | if (blk->name) { | |
133 | return g_strdup(blk->name); | |
134 | } | |
135 | ||
136 | dev_id = blk_get_attached_dev_id(blk); | |
137 | if (*dev_id) { | |
138 | return dev_id; | |
139 | } else { | |
140 | /* TODO Callback into the BB owner for something more detailed */ | |
141 | g_free(dev_id); | |
142 | return g_strdup("a block device"); | |
143 | } | |
144 | } | |
145 | ||
4c265bf9 KW |
146 | static const char *blk_root_get_name(BdrvChild *child) |
147 | { | |
148 | return blk_name(child->opaque); | |
149 | } | |
150 | ||
5f7772c4 FZ |
151 | static void blk_vm_state_changed(void *opaque, int running, RunState state) |
152 | { | |
153 | Error *local_err = NULL; | |
154 | BlockBackend *blk = opaque; | |
155 | ||
156 | if (state == RUN_STATE_INMIGRATE) { | |
157 | return; | |
158 | } | |
159 | ||
160 | qemu_del_vm_change_state_handler(blk->vmsh); | |
161 | blk->vmsh = NULL; | |
162 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); | |
163 | if (local_err) { | |
164 | error_report_err(local_err); | |
165 | } | |
166 | } | |
167 | ||
4417ab7a KW |
168 | /* |
169 | * Notifies the user of the BlockBackend that migration has completed. qdev | |
170 | * devices can tighten their permissions in response (specifically revoke | |
171 | * shared write permissions that we needed for storage migration). | |
172 | * | |
173 | * If an error is returned, the VM cannot be allowed to be resumed. | |
174 | */ | |
175 | static void blk_root_activate(BdrvChild *child, Error **errp) | |
176 | { | |
177 | BlockBackend *blk = child->opaque; | |
178 | Error *local_err = NULL; | |
179 | ||
180 | if (!blk->disable_perm) { | |
181 | return; | |
182 | } | |
183 | ||
184 | blk->disable_perm = false; | |
185 | ||
5f7772c4 FZ |
186 | blk_set_perm(blk, blk->perm, BLK_PERM_ALL, &local_err); |
187 | if (local_err) { | |
188 | error_propagate(errp, local_err); | |
189 | blk->disable_perm = true; | |
190 | return; | |
191 | } | |
192 | ||
193 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
194 | /* Activation can happen when migration process is still active, for | |
195 | * example when nbd_server_add is called during non-shared storage | |
196 | * migration. Defer the shared_perm update to migration completion. */ | |
197 | if (!blk->vmsh) { | |
198 | blk->vmsh = qemu_add_vm_change_state_handler(blk_vm_state_changed, | |
199 | blk); | |
200 | } | |
201 | return; | |
202 | } | |
203 | ||
4417ab7a KW |
204 | blk_set_perm(blk, blk->perm, blk->shared_perm, &local_err); |
205 | if (local_err) { | |
206 | error_propagate(errp, local_err); | |
207 | blk->disable_perm = true; | |
208 | return; | |
209 | } | |
210 | } | |
211 | ||
ca2e2144 FZ |
212 | void blk_set_force_allow_inactivate(BlockBackend *blk) |
213 | { | |
214 | blk->force_allow_inactivate = true; | |
215 | } | |
216 | ||
c16de8f5 FZ |
217 | static bool blk_can_inactivate(BlockBackend *blk) |
218 | { | |
ca2e2144 | 219 | /* If it is a guest device, inactivate is ok. */ |
c16de8f5 FZ |
220 | if (blk->dev || blk_name(blk)[0]) { |
221 | return true; | |
222 | } | |
223 | ||
ca2e2144 FZ |
224 | /* Inactivating means no more writes to the image can be done, |
225 | * even if those writes would be changes invisible to the | |
226 | * guest. For block job BBs that satisfy this, we can just allow | |
227 | * it. This is the case for mirror job source, which is required | |
228 | * by libvirt non-shared block migration. */ | |
229 | if (!(blk->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED))) { | |
230 | return true; | |
231 | } | |
232 | ||
233 | return blk->force_allow_inactivate; | |
c16de8f5 FZ |
234 | } |
235 | ||
cfa1a572 KW |
236 | static int blk_root_inactivate(BdrvChild *child) |
237 | { | |
238 | BlockBackend *blk = child->opaque; | |
239 | ||
240 | if (blk->disable_perm) { | |
241 | return 0; | |
242 | } | |
243 | ||
c16de8f5 | 244 | if (!blk_can_inactivate(blk)) { |
cfa1a572 KW |
245 | return -EPERM; |
246 | } | |
247 | ||
248 | blk->disable_perm = true; | |
249 | if (blk->root) { | |
250 | bdrv_child_try_set_perm(blk->root, 0, BLK_PERM_ALL, &error_abort); | |
251 | } | |
252 | ||
253 | return 0; | |
254 | } | |
255 | ||
d03654ea SH |
256 | static void blk_root_attach(BdrvChild *child) |
257 | { | |
258 | BlockBackend *blk = child->opaque; | |
259 | BlockBackendAioNotifier *notifier; | |
260 | ||
261 | trace_blk_root_attach(child, blk, child->bs); | |
262 | ||
263 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
264 | bdrv_add_aio_context_notifier(child->bs, | |
265 | notifier->attached_aio_context, | |
266 | notifier->detach_aio_context, | |
267 | notifier->opaque); | |
268 | } | |
269 | } | |
270 | ||
271 | static void blk_root_detach(BdrvChild *child) | |
272 | { | |
273 | BlockBackend *blk = child->opaque; | |
274 | BlockBackendAioNotifier *notifier; | |
275 | ||
276 | trace_blk_root_detach(child, blk, child->bs); | |
277 | ||
278 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
279 | bdrv_remove_aio_context_notifier(child->bs, | |
280 | notifier->attached_aio_context, | |
281 | notifier->detach_aio_context, | |
282 | notifier->opaque); | |
283 | } | |
284 | } | |
285 | ||
f21d96d0 | 286 | static const BdrvChildRole child_root = { |
c2066af0 KW |
287 | .inherit_options = blk_root_inherit_options, |
288 | ||
5c8cab48 KW |
289 | .change_media = blk_root_change_media, |
290 | .resize = blk_root_resize, | |
4c265bf9 | 291 | .get_name = blk_root_get_name, |
b5411555 | 292 | .get_parent_desc = blk_root_get_parent_desc, |
5c8cab48 | 293 | |
c2066af0 | 294 | .drained_begin = blk_root_drained_begin, |
fe5258a5 | 295 | .drained_poll = blk_root_drained_poll, |
c2066af0 | 296 | .drained_end = blk_root_drained_end, |
4417ab7a KW |
297 | |
298 | .activate = blk_root_activate, | |
cfa1a572 | 299 | .inactivate = blk_root_inactivate, |
d03654ea SH |
300 | |
301 | .attach = blk_root_attach, | |
302 | .detach = blk_root_detach, | |
f21d96d0 KW |
303 | }; |
304 | ||
26f54e9a | 305 | /* |
efaa7c4e | 306 | * Create a new BlockBackend with a reference count of one. |
6d0eb64d KW |
307 | * |
308 | * @perm is a bitmasks of BLK_PERM_* constants which describes the permissions | |
309 | * to request for a block driver node that is attached to this BlockBackend. | |
310 | * @shared_perm is a bitmask which describes which permissions may be granted | |
311 | * to other users of the attached node. | |
312 | * Both sets of permissions can be changed later using blk_set_perm(). | |
313 | * | |
26f54e9a MA |
314 | * Return the new BlockBackend on success, null on failure. |
315 | */ | |
6d0eb64d | 316 | BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm) |
26f54e9a MA |
317 | { |
318 | BlockBackend *blk; | |
319 | ||
26f54e9a | 320 | blk = g_new0(BlockBackend, 1); |
26f54e9a | 321 | blk->refcnt = 1; |
6d0eb64d KW |
322 | blk->perm = perm; |
323 | blk->shared_perm = shared_perm; | |
0c3169df KW |
324 | blk_set_enable_write_cache(blk, true); |
325 | ||
cb53460b KW |
326 | blk->on_read_error = BLOCKDEV_ON_ERROR_REPORT; |
327 | blk->on_write_error = BLOCKDEV_ON_ERROR_ENOSPC; | |
328 | ||
9caa6f3d | 329 | block_acct_init(&blk->stats); |
27ccdd52 | 330 | |
3301f6c6 HR |
331 | notifier_list_init(&blk->remove_bs_notifiers); |
332 | notifier_list_init(&blk->insert_bs_notifiers); | |
d03654ea | 333 | QLIST_INIT(&blk->aio_notifiers); |
27ccdd52 | 334 | |
2cf22d6a | 335 | QTAILQ_INSERT_TAIL(&block_backends, blk, link); |
26f54e9a MA |
336 | return blk; |
337 | } | |
338 | ||
7e7d56d9 | 339 | /* |
28eb9b12 | 340 | * Creates a new BlockBackend, opens a new BlockDriverState, and connects both. |
ca49a4fd HR |
341 | * |
342 | * Just as with bdrv_open(), after having called this function the reference to | |
343 | * @options belongs to the block layer (even on failure). | |
344 | * | |
345 | * TODO: Remove @filename and @flags; it should be possible to specify a whole | |
346 | * BDS tree just by specifying the @options QDict (or @reference, | |
347 | * alternatively). At the time of adding this function, this is not possible, | |
348 | * though, so callers of this function have to be able to specify @filename and | |
349 | * @flags. | |
350 | */ | |
efaa7c4e HR |
351 | BlockBackend *blk_new_open(const char *filename, const char *reference, |
352 | QDict *options, int flags, Error **errp) | |
ca49a4fd HR |
353 | { |
354 | BlockBackend *blk; | |
28eb9b12 | 355 | BlockDriverState *bs; |
1f4ad7d3 | 356 | uint64_t perm = 0; |
c62d32f5 KW |
357 | |
358 | /* blk_new_open() is mainly used in .bdrv_create implementations and the | |
359 | * tools where sharing isn't a concern because the BDS stays private, so we | |
360 | * just request permission according to the flags. | |
361 | * | |
362 | * The exceptions are xen_disk and blockdev_init(); in these cases, the | |
363 | * caller of blk_new_open() doesn't make use of the permissions, but they | |
364 | * shouldn't hurt either. We can still share everything here because the | |
365 | * guest devices will add their own blockers if they can't share. */ | |
1f4ad7d3 KW |
366 | if ((flags & BDRV_O_NO_IO) == 0) { |
367 | perm |= BLK_PERM_CONSISTENT_READ; | |
368 | if (flags & BDRV_O_RDWR) { | |
369 | perm |= BLK_PERM_WRITE; | |
370 | } | |
c62d32f5 KW |
371 | } |
372 | if (flags & BDRV_O_RESIZE) { | |
373 | perm |= BLK_PERM_RESIZE; | |
374 | } | |
ca49a4fd | 375 | |
c62d32f5 | 376 | blk = blk_new(perm, BLK_PERM_ALL); |
5b363937 HR |
377 | bs = bdrv_open(filename, reference, options, flags, errp); |
378 | if (!bs) { | |
ca49a4fd HR |
379 | blk_unref(blk); |
380 | return NULL; | |
381 | } | |
382 | ||
d5e6f437 | 383 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
50bfbe93 FZ |
384 | perm, BLK_PERM_ALL, blk, errp); |
385 | if (!blk->root) { | |
386 | bdrv_unref(bs); | |
387 | blk_unref(blk); | |
388 | return NULL; | |
389 | } | |
72e775c7 | 390 | |
ca49a4fd HR |
391 | return blk; |
392 | } | |
393 | ||
26f54e9a MA |
394 | static void blk_delete(BlockBackend *blk) |
395 | { | |
396 | assert(!blk->refcnt); | |
e5e78550 | 397 | assert(!blk->name); |
a7f53e26 | 398 | assert(!blk->dev); |
022cdc9f | 399 | if (blk->public.throttle_group_member.throttle_state) { |
1606e4cf EB |
400 | blk_io_limits_disable(blk); |
401 | } | |
f21d96d0 | 402 | if (blk->root) { |
13855c6b | 403 | blk_remove_bs(blk); |
7e7d56d9 | 404 | } |
5f7772c4 FZ |
405 | if (blk->vmsh) { |
406 | qemu_del_vm_change_state_handler(blk->vmsh); | |
407 | blk->vmsh = NULL; | |
408 | } | |
3301f6c6 HR |
409 | assert(QLIST_EMPTY(&blk->remove_bs_notifiers.notifiers)); |
410 | assert(QLIST_EMPTY(&blk->insert_bs_notifiers.notifiers)); | |
d03654ea | 411 | assert(QLIST_EMPTY(&blk->aio_notifiers)); |
2cf22d6a | 412 | QTAILQ_REMOVE(&block_backends, blk, link); |
18e46a03 | 413 | drive_info_del(blk->legacy_dinfo); |
979e9b03 | 414 | block_acct_cleanup(&blk->stats); |
26f54e9a MA |
415 | g_free(blk); |
416 | } | |
417 | ||
8fb3c76c MA |
418 | static void drive_info_del(DriveInfo *dinfo) |
419 | { | |
420 | if (!dinfo) { | |
421 | return; | |
422 | } | |
423 | qemu_opts_del(dinfo->opts); | |
8fb3c76c MA |
424 | g_free(dinfo); |
425 | } | |
426 | ||
f636ae85 AG |
427 | int blk_get_refcnt(BlockBackend *blk) |
428 | { | |
429 | return blk ? blk->refcnt : 0; | |
430 | } | |
431 | ||
26f54e9a MA |
432 | /* |
433 | * Increment @blk's reference count. | |
434 | * @blk must not be null. | |
435 | */ | |
436 | void blk_ref(BlockBackend *blk) | |
437 | { | |
5ca9d21b | 438 | assert(blk->refcnt > 0); |
26f54e9a MA |
439 | blk->refcnt++; |
440 | } | |
441 | ||
442 | /* | |
443 | * Decrement @blk's reference count. | |
444 | * If this drops it to zero, destroy @blk. | |
445 | * For convenience, do nothing if @blk is null. | |
446 | */ | |
447 | void blk_unref(BlockBackend *blk) | |
448 | { | |
449 | if (blk) { | |
450 | assert(blk->refcnt > 0); | |
5ca9d21b KW |
451 | if (blk->refcnt > 1) { |
452 | blk->refcnt--; | |
453 | } else { | |
454 | blk_drain(blk); | |
455 | /* blk_drain() cannot resurrect blk, nobody held a reference */ | |
456 | assert(blk->refcnt == 1); | |
457 | blk->refcnt = 0; | |
26f54e9a MA |
458 | blk_delete(blk); |
459 | } | |
460 | } | |
461 | } | |
462 | ||
2cf22d6a HR |
463 | /* |
464 | * Behaves similarly to blk_next() but iterates over all BlockBackends, even the | |
465 | * ones which are hidden (i.e. are not referenced by the monitor). | |
466 | */ | |
a429b9b5 | 467 | BlockBackend *blk_all_next(BlockBackend *blk) |
2cf22d6a HR |
468 | { |
469 | return blk ? QTAILQ_NEXT(blk, link) | |
470 | : QTAILQ_FIRST(&block_backends); | |
471 | } | |
472 | ||
d8da3cef HR |
473 | void blk_remove_all_bs(void) |
474 | { | |
74d1b8fc | 475 | BlockBackend *blk = NULL; |
d8da3cef | 476 | |
2cf22d6a | 477 | while ((blk = blk_all_next(blk)) != NULL) { |
d8da3cef HR |
478 | AioContext *ctx = blk_get_aio_context(blk); |
479 | ||
480 | aio_context_acquire(ctx); | |
f21d96d0 | 481 | if (blk->root) { |
d8da3cef HR |
482 | blk_remove_bs(blk); |
483 | } | |
484 | aio_context_release(ctx); | |
485 | } | |
486 | } | |
487 | ||
26f54e9a | 488 | /* |
9492b0b9 | 489 | * Return the monitor-owned BlockBackend after @blk. |
26f54e9a MA |
490 | * If @blk is null, return the first one. |
491 | * Else, return @blk's next sibling, which may be null. | |
492 | * | |
493 | * To iterate over all BlockBackends, do | |
494 | * for (blk = blk_next(NULL); blk; blk = blk_next(blk)) { | |
495 | * ... | |
496 | * } | |
497 | */ | |
498 | BlockBackend *blk_next(BlockBackend *blk) | |
499 | { | |
9492b0b9 HR |
500 | return blk ? QTAILQ_NEXT(blk, monitor_link) |
501 | : QTAILQ_FIRST(&monitor_block_backends); | |
26f54e9a MA |
502 | } |
503 | ||
7c8eece4 KW |
504 | /* Iterates over all top-level BlockDriverStates, i.e. BDSs that are owned by |
505 | * the monitor or attached to a BlockBackend */ | |
88be7b4b | 506 | BlockDriverState *bdrv_next(BdrvNextIterator *it) |
7c8eece4 | 507 | { |
5e003f17 HR |
508 | BlockDriverState *bs, *old_bs; |
509 | ||
510 | /* Must be called from the main loop */ | |
511 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
981f4f57 | 512 | |
7c8eece4 KW |
513 | /* First, return all root nodes of BlockBackends. In order to avoid |
514 | * returning a BDS twice when multiple BBs refer to it, we only return it | |
515 | * if the BB is the first one in the parent list of the BDS. */ | |
516 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
5e003f17 HR |
517 | BlockBackend *old_blk = it->blk; |
518 | ||
519 | old_bs = old_blk ? blk_bs(old_blk) : NULL; | |
520 | ||
7c8eece4 KW |
521 | do { |
522 | it->blk = blk_all_next(it->blk); | |
88be7b4b KW |
523 | bs = it->blk ? blk_bs(it->blk) : NULL; |
524 | } while (it->blk && (bs == NULL || bdrv_first_blk(bs) != it->blk)); | |
7c8eece4 | 525 | |
5e003f17 HR |
526 | if (it->blk) { |
527 | blk_ref(it->blk); | |
528 | } | |
529 | blk_unref(old_blk); | |
530 | ||
88be7b4b | 531 | if (bs) { |
5e003f17 HR |
532 | bdrv_ref(bs); |
533 | bdrv_unref(old_bs); | |
88be7b4b | 534 | return bs; |
7c8eece4 KW |
535 | } |
536 | it->phase = BDRV_NEXT_MONITOR_OWNED; | |
5e003f17 HR |
537 | } else { |
538 | old_bs = it->bs; | |
7c8eece4 KW |
539 | } |
540 | ||
541 | /* Then return the monitor-owned BDSes without a BB attached. Ignore all | |
542 | * BDSes that are attached to a BlockBackend here; they have been handled | |
543 | * by the above block already */ | |
981f4f57 | 544 | do { |
7c8eece4 | 545 | it->bs = bdrv_next_monitor_owned(it->bs); |
88be7b4b KW |
546 | bs = it->bs; |
547 | } while (bs && bdrv_has_blk(bs)); | |
548 | ||
5e003f17 HR |
549 | if (bs) { |
550 | bdrv_ref(bs); | |
551 | } | |
552 | bdrv_unref(old_bs); | |
553 | ||
88be7b4b KW |
554 | return bs; |
555 | } | |
556 | ||
5e003f17 | 557 | static void bdrv_next_reset(BdrvNextIterator *it) |
88be7b4b KW |
558 | { |
559 | *it = (BdrvNextIterator) { | |
560 | .phase = BDRV_NEXT_BACKEND_ROOTS, | |
561 | }; | |
5e003f17 | 562 | } |
981f4f57 | 563 | |
5e003f17 HR |
564 | BlockDriverState *bdrv_first(BdrvNextIterator *it) |
565 | { | |
566 | bdrv_next_reset(it); | |
88be7b4b | 567 | return bdrv_next(it); |
981f4f57 HR |
568 | } |
569 | ||
5e003f17 HR |
570 | /* Must be called when aborting a bdrv_next() iteration before |
571 | * bdrv_next() returns NULL */ | |
572 | void bdrv_next_cleanup(BdrvNextIterator *it) | |
573 | { | |
574 | /* Must be called from the main loop */ | |
575 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); | |
576 | ||
577 | if (it->phase == BDRV_NEXT_BACKEND_ROOTS) { | |
578 | if (it->blk) { | |
579 | bdrv_unref(blk_bs(it->blk)); | |
580 | blk_unref(it->blk); | |
581 | } | |
582 | } else { | |
583 | bdrv_unref(it->bs); | |
584 | } | |
585 | ||
586 | bdrv_next_reset(it); | |
587 | } | |
588 | ||
e5e78550 HR |
589 | /* |
590 | * Add a BlockBackend into the list of backends referenced by the monitor, with | |
591 | * the given @name acting as the handle for the monitor. | |
592 | * Strictly for use by blockdev.c. | |
593 | * | |
594 | * @name must not be null or empty. | |
595 | * | |
596 | * Returns true on success and false on failure. In the latter case, an Error | |
597 | * object is returned through @errp. | |
598 | */ | |
599 | bool monitor_add_blk(BlockBackend *blk, const char *name, Error **errp) | |
600 | { | |
601 | assert(!blk->name); | |
602 | assert(name && name[0]); | |
603 | ||
604 | if (!id_wellformed(name)) { | |
605 | error_setg(errp, "Invalid device name"); | |
606 | return false; | |
607 | } | |
608 | if (blk_by_name(name)) { | |
609 | error_setg(errp, "Device with id '%s' already exists", name); | |
610 | return false; | |
611 | } | |
612 | if (bdrv_find_node(name)) { | |
613 | error_setg(errp, | |
614 | "Device name '%s' conflicts with an existing node name", | |
615 | name); | |
616 | return false; | |
617 | } | |
618 | ||
619 | blk->name = g_strdup(name); | |
620 | QTAILQ_INSERT_TAIL(&monitor_block_backends, blk, monitor_link); | |
621 | return true; | |
622 | } | |
623 | ||
624 | /* | |
625 | * Remove a BlockBackend from the list of backends referenced by the monitor. | |
626 | * Strictly for use by blockdev.c. | |
627 | */ | |
628 | void monitor_remove_blk(BlockBackend *blk) | |
629 | { | |
630 | if (!blk->name) { | |
631 | return; | |
632 | } | |
633 | ||
634 | QTAILQ_REMOVE(&monitor_block_backends, blk, monitor_link); | |
635 | g_free(blk->name); | |
636 | blk->name = NULL; | |
637 | } | |
638 | ||
26f54e9a | 639 | /* |
7e7d56d9 | 640 | * Return @blk's name, a non-null string. |
e5e78550 | 641 | * Returns an empty string iff @blk is not referenced by the monitor. |
26f54e9a | 642 | */ |
0731a50f | 643 | const char *blk_name(const BlockBackend *blk) |
26f54e9a | 644 | { |
e5e78550 | 645 | return blk->name ?: ""; |
26f54e9a MA |
646 | } |
647 | ||
648 | /* | |
649 | * Return the BlockBackend with name @name if it exists, else null. | |
650 | * @name must not be null. | |
651 | */ | |
652 | BlockBackend *blk_by_name(const char *name) | |
653 | { | |
74d1b8fc | 654 | BlockBackend *blk = NULL; |
26f54e9a MA |
655 | |
656 | assert(name); | |
74d1b8fc | 657 | while ((blk = blk_next(blk)) != NULL) { |
26f54e9a MA |
658 | if (!strcmp(name, blk->name)) { |
659 | return blk; | |
660 | } | |
661 | } | |
662 | return NULL; | |
663 | } | |
7e7d56d9 MA |
664 | |
665 | /* | |
666 | * Return the BlockDriverState attached to @blk if any, else null. | |
667 | */ | |
668 | BlockDriverState *blk_bs(BlockBackend *blk) | |
669 | { | |
f21d96d0 | 670 | return blk->root ? blk->root->bs : NULL; |
7e7d56d9 MA |
671 | } |
672 | ||
7c8eece4 | 673 | static BlockBackend *bdrv_first_blk(BlockDriverState *bs) |
dde33812 KW |
674 | { |
675 | BdrvChild *child; | |
676 | QLIST_FOREACH(child, &bs->parents, next_parent) { | |
677 | if (child->role == &child_root) { | |
7c8eece4 | 678 | return child->opaque; |
dde33812 KW |
679 | } |
680 | } | |
681 | ||
7c8eece4 KW |
682 | return NULL; |
683 | } | |
684 | ||
685 | /* | |
686 | * Returns true if @bs has an associated BlockBackend. | |
687 | */ | |
688 | bool bdrv_has_blk(BlockDriverState *bs) | |
689 | { | |
690 | return bdrv_first_blk(bs) != NULL; | |
dde33812 KW |
691 | } |
692 | ||
b6c1bae5 KW |
693 | /* |
694 | * Returns true if @bs has only BlockBackends as parents. | |
695 | */ | |
696 | bool bdrv_is_root_node(BlockDriverState *bs) | |
697 | { | |
698 | BdrvChild *c; | |
699 | ||
700 | QLIST_FOREACH(c, &bs->parents, next_parent) { | |
701 | if (c->role != &child_root) { | |
702 | return false; | |
703 | } | |
704 | } | |
705 | ||
706 | return true; | |
707 | } | |
708 | ||
18e46a03 MA |
709 | /* |
710 | * Return @blk's DriveInfo if any, else null. | |
711 | */ | |
712 | DriveInfo *blk_legacy_dinfo(BlockBackend *blk) | |
713 | { | |
714 | return blk->legacy_dinfo; | |
715 | } | |
716 | ||
717 | /* | |
718 | * Set @blk's DriveInfo to @dinfo, and return it. | |
719 | * @blk must not have a DriveInfo set already. | |
720 | * No other BlockBackend may have the same DriveInfo set. | |
721 | */ | |
722 | DriveInfo *blk_set_legacy_dinfo(BlockBackend *blk, DriveInfo *dinfo) | |
723 | { | |
724 | assert(!blk->legacy_dinfo); | |
725 | return blk->legacy_dinfo = dinfo; | |
726 | } | |
727 | ||
728 | /* | |
729 | * Return the BlockBackend with DriveInfo @dinfo. | |
730 | * It must exist. | |
731 | */ | |
732 | BlockBackend *blk_by_legacy_dinfo(DriveInfo *dinfo) | |
733 | { | |
74d1b8fc | 734 | BlockBackend *blk = NULL; |
18e46a03 | 735 | |
74d1b8fc | 736 | while ((blk = blk_next(blk)) != NULL) { |
18e46a03 MA |
737 | if (blk->legacy_dinfo == dinfo) { |
738 | return blk; | |
739 | } | |
740 | } | |
741 | abort(); | |
742 | } | |
743 | ||
f2cd875d KW |
744 | /* |
745 | * Returns a pointer to the publicly accessible fields of @blk. | |
746 | */ | |
747 | BlockBackendPublic *blk_get_public(BlockBackend *blk) | |
748 | { | |
749 | return &blk->public; | |
750 | } | |
751 | ||
752 | /* | |
753 | * Returns a BlockBackend given the associated @public fields. | |
754 | */ | |
755 | BlockBackend *blk_by_public(BlockBackendPublic *public) | |
756 | { | |
757 | return container_of(public, BlockBackend, public); | |
758 | } | |
759 | ||
1c95f7e1 HR |
760 | /* |
761 | * Disassociates the currently associated BlockDriverState from @blk. | |
762 | */ | |
763 | void blk_remove_bs(BlockBackend *blk) | |
764 | { | |
c89bcf3a | 765 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
632a7735 | 766 | BlockDriverState *bs; |
022cdc9f | 767 | |
3301f6c6 | 768 | notifier_list_notify(&blk->remove_bs_notifiers, blk); |
c89bcf3a | 769 | if (tgm->throttle_state) { |
632a7735 ZL |
770 | bs = blk_bs(blk); |
771 | bdrv_drained_begin(bs); | |
c89bcf3a AG |
772 | throttle_group_detach_aio_context(tgm); |
773 | throttle_group_attach_aio_context(tgm, qemu_get_aio_context()); | |
632a7735 | 774 | bdrv_drained_end(bs); |
a5614993 | 775 | } |
1c95f7e1 | 776 | |
7ca7f0f6 KW |
777 | blk_update_root_state(blk); |
778 | ||
f45280cb GK |
779 | /* bdrv_root_unref_child() will cause blk->root to become stale and may |
780 | * switch to a completion coroutine later on. Let's drain all I/O here | |
781 | * to avoid that and a potential QEMU crash. | |
782 | */ | |
783 | blk_drain(blk); | |
f21d96d0 KW |
784 | bdrv_root_unref_child(blk->root); |
785 | blk->root = NULL; | |
1c95f7e1 HR |
786 | } |
787 | ||
0c3c36d6 HR |
788 | /* |
789 | * Associates a new BlockDriverState with @blk. | |
790 | */ | |
d7086422 | 791 | int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp) |
0c3c36d6 | 792 | { |
c89bcf3a | 793 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
d5e6f437 | 794 | blk->root = bdrv_root_attach_child(bs, "root", &child_root, |
d7086422 KW |
795 | blk->perm, blk->shared_perm, blk, errp); |
796 | if (blk->root == NULL) { | |
797 | return -EPERM; | |
798 | } | |
799 | bdrv_ref(bs); | |
3301f6c6 HR |
800 | |
801 | notifier_list_notify(&blk->insert_bs_notifiers, blk); | |
c89bcf3a AG |
802 | if (tgm->throttle_state) { |
803 | throttle_group_detach_aio_context(tgm); | |
804 | throttle_group_attach_aio_context(tgm, bdrv_get_aio_context(bs)); | |
7ca7f0f6 | 805 | } |
d7086422 KW |
806 | |
807 | return 0; | |
0c3c36d6 HR |
808 | } |
809 | ||
981776b3 KW |
810 | /* |
811 | * Sets the permission bitmasks that the user of the BlockBackend needs. | |
812 | */ | |
813 | int blk_set_perm(BlockBackend *blk, uint64_t perm, uint64_t shared_perm, | |
814 | Error **errp) | |
815 | { | |
816 | int ret; | |
817 | ||
d35ff5e6 | 818 | if (blk->root && !blk->disable_perm) { |
981776b3 KW |
819 | ret = bdrv_child_try_set_perm(blk->root, perm, shared_perm, errp); |
820 | if (ret < 0) { | |
821 | return ret; | |
822 | } | |
823 | } | |
824 | ||
825 | blk->perm = perm; | |
826 | blk->shared_perm = shared_perm; | |
827 | ||
828 | return 0; | |
829 | } | |
830 | ||
887354bd KW |
831 | void blk_get_perm(BlockBackend *blk, uint64_t *perm, uint64_t *shared_perm) |
832 | { | |
833 | *perm = blk->perm; | |
834 | *shared_perm = blk->shared_perm; | |
835 | } | |
836 | ||
d09ea2d2 TH |
837 | /* |
838 | * Attach device model @dev to @blk. | |
839 | * Return 0 on success, -EBUSY when a device model is attached already. | |
840 | */ | |
841 | int blk_attach_dev(BlockBackend *blk, DeviceState *dev) | |
4be74634 | 842 | { |
a7f53e26 MA |
843 | if (blk->dev) { |
844 | return -EBUSY; | |
845 | } | |
d35ff5e6 KW |
846 | |
847 | /* While migration is still incoming, we don't need to apply the | |
848 | * permissions of guest device BlockBackends. We might still have a block | |
849 | * job or NBD server writing to the image for storage migration. */ | |
850 | if (runstate_check(RUN_STATE_INMIGRATE)) { | |
851 | blk->disable_perm = true; | |
852 | } | |
853 | ||
84ebe375 | 854 | blk_ref(blk); |
a7f53e26 | 855 | blk->dev = dev; |
373340b2 | 856 | blk_iostatus_reset(blk); |
d35ff5e6 | 857 | |
a7f53e26 | 858 | return 0; |
4be74634 MA |
859 | } |
860 | ||
a7f53e26 MA |
861 | /* |
862 | * Detach device model @dev from @blk. | |
863 | * @dev must be currently attached to @blk. | |
864 | */ | |
d09ea2d2 | 865 | void blk_detach_dev(BlockBackend *blk, DeviceState *dev) |
4be74634 | 866 | { |
a7f53e26 MA |
867 | assert(blk->dev == dev); |
868 | blk->dev = NULL; | |
869 | blk->dev_ops = NULL; | |
870 | blk->dev_opaque = NULL; | |
68e9ec01 | 871 | blk->guest_block_size = 512; |
981776b3 | 872 | blk_set_perm(blk, 0, BLK_PERM_ALL, &error_abort); |
84ebe375 | 873 | blk_unref(blk); |
4be74634 MA |
874 | } |
875 | ||
a7f53e26 MA |
876 | /* |
877 | * Return the device model attached to @blk if any, else null. | |
878 | */ | |
d09ea2d2 | 879 | DeviceState *blk_get_attached_dev(BlockBackend *blk) |
a7f53e26 MA |
880 | { |
881 | return blk->dev; | |
882 | } | |
883 | ||
2d76e724 KW |
884 | /* Return the qdev ID, or if no ID is assigned the QOM path, of the block |
885 | * device attached to the BlockBackend. */ | |
77beef83 | 886 | char *blk_get_attached_dev_id(BlockBackend *blk) |
2d76e724 | 887 | { |
d09ea2d2 | 888 | DeviceState *dev = blk->dev; |
2d76e724 KW |
889 | |
890 | if (!dev) { | |
891 | return g_strdup(""); | |
892 | } else if (dev->id) { | |
893 | return g_strdup(dev->id); | |
894 | } | |
602414d1 LM |
895 | |
896 | return object_get_canonical_path(OBJECT(dev)) ?: g_strdup(""); | |
2d76e724 KW |
897 | } |
898 | ||
1c89e1fa KW |
899 | /* |
900 | * Return the BlockBackend which has the device model @dev attached if it | |
901 | * exists, else null. | |
902 | * | |
903 | * @dev must not be null. | |
904 | */ | |
905 | BlockBackend *blk_by_dev(void *dev) | |
906 | { | |
907 | BlockBackend *blk = NULL; | |
908 | ||
909 | assert(dev != NULL); | |
910 | while ((blk = blk_all_next(blk)) != NULL) { | |
911 | if (blk->dev == dev) { | |
912 | return blk; | |
913 | } | |
914 | } | |
915 | return NULL; | |
916 | } | |
917 | ||
a7f53e26 MA |
918 | /* |
919 | * Set @blk's device model callbacks to @ops. | |
920 | * @opaque is the opaque argument to pass to the callbacks. | |
921 | * This is for use by device models. | |
922 | */ | |
923 | void blk_set_dev_ops(BlockBackend *blk, const BlockDevOps *ops, | |
924 | void *opaque) | |
925 | { | |
926 | blk->dev_ops = ops; | |
927 | blk->dev_opaque = opaque; | |
f4d9cc88 JS |
928 | |
929 | /* Are we currently quiesced? Should we enforce this right now? */ | |
930 | if (blk->quiesce_counter && ops->drained_begin) { | |
931 | ops->drained_begin(opaque); | |
932 | } | |
a7f53e26 MA |
933 | } |
934 | ||
935 | /* | |
936 | * Notify @blk's attached device model of media change. | |
39829a01 KW |
937 | * |
938 | * If @load is true, notify of media load. This action can fail, meaning that | |
939 | * the medium cannot be loaded. @errp is set then. | |
940 | * | |
941 | * If @load is false, notify of media eject. This can never fail. | |
942 | * | |
a7f53e26 MA |
943 | * Also send DEVICE_TRAY_MOVED events as appropriate. |
944 | */ | |
39829a01 | 945 | void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp) |
a7f53e26 MA |
946 | { |
947 | if (blk->dev_ops && blk->dev_ops->change_media_cb) { | |
f1f57066 | 948 | bool tray_was_open, tray_is_open; |
39829a01 | 949 | Error *local_err = NULL; |
a7f53e26 | 950 | |
f1f57066 | 951 | tray_was_open = blk_dev_is_tray_open(blk); |
39829a01 KW |
952 | blk->dev_ops->change_media_cb(blk->dev_opaque, load, &local_err); |
953 | if (local_err) { | |
954 | assert(load == true); | |
955 | error_propagate(errp, local_err); | |
956 | return; | |
957 | } | |
f1f57066 HR |
958 | tray_is_open = blk_dev_is_tray_open(blk); |
959 | ||
960 | if (tray_was_open != tray_is_open) { | |
2d76e724 | 961 | char *id = blk_get_attached_dev_id(blk); |
3ab72385 | 962 | qapi_event_send_device_tray_moved(blk_name(blk), id, tray_is_open); |
2d76e724 | 963 | g_free(id); |
a7f53e26 MA |
964 | } |
965 | } | |
966 | } | |
967 | ||
5c8cab48 KW |
968 | static void blk_root_change_media(BdrvChild *child, bool load) |
969 | { | |
39829a01 | 970 | blk_dev_change_media_cb(child->opaque, load, NULL); |
5c8cab48 KW |
971 | } |
972 | ||
a7f53e26 MA |
973 | /* |
974 | * Does @blk's attached device model have removable media? | |
975 | * %true if no device model is attached. | |
976 | */ | |
977 | bool blk_dev_has_removable_media(BlockBackend *blk) | |
978 | { | |
979 | return !blk->dev || (blk->dev_ops && blk->dev_ops->change_media_cb); | |
980 | } | |
981 | ||
8f3a73bc HR |
982 | /* |
983 | * Does @blk's attached device model have a tray? | |
984 | */ | |
985 | bool blk_dev_has_tray(BlockBackend *blk) | |
986 | { | |
987 | return blk->dev_ops && blk->dev_ops->is_tray_open; | |
988 | } | |
989 | ||
a7f53e26 MA |
990 | /* |
991 | * Notify @blk's attached device model of a media eject request. | |
992 | * If @force is true, the medium is about to be yanked out forcefully. | |
993 | */ | |
994 | void blk_dev_eject_request(BlockBackend *blk, bool force) | |
4be74634 | 995 | { |
a7f53e26 MA |
996 | if (blk->dev_ops && blk->dev_ops->eject_request_cb) { |
997 | blk->dev_ops->eject_request_cb(blk->dev_opaque, force); | |
998 | } | |
4be74634 MA |
999 | } |
1000 | ||
a7f53e26 MA |
1001 | /* |
1002 | * Does @blk's attached device model have a tray, and is it open? | |
1003 | */ | |
1004 | bool blk_dev_is_tray_open(BlockBackend *blk) | |
4be74634 | 1005 | { |
8f3a73bc | 1006 | if (blk_dev_has_tray(blk)) { |
a7f53e26 MA |
1007 | return blk->dev_ops->is_tray_open(blk->dev_opaque); |
1008 | } | |
1009 | return false; | |
1010 | } | |
1011 | ||
1012 | /* | |
1013 | * Does @blk's attached device model have the medium locked? | |
1014 | * %false if the device model has no such lock. | |
1015 | */ | |
1016 | bool blk_dev_is_medium_locked(BlockBackend *blk) | |
1017 | { | |
1018 | if (blk->dev_ops && blk->dev_ops->is_medium_locked) { | |
1019 | return blk->dev_ops->is_medium_locked(blk->dev_opaque); | |
1020 | } | |
1021 | return false; | |
1022 | } | |
1023 | ||
1024 | /* | |
1025 | * Notify @blk's attached device model of a backend size change. | |
1026 | */ | |
5c8cab48 | 1027 | static void blk_root_resize(BdrvChild *child) |
a7f53e26 | 1028 | { |
5c8cab48 KW |
1029 | BlockBackend *blk = child->opaque; |
1030 | ||
a7f53e26 MA |
1031 | if (blk->dev_ops && blk->dev_ops->resize_cb) { |
1032 | blk->dev_ops->resize_cb(blk->dev_opaque); | |
1033 | } | |
1034 | } | |
1035 | ||
1036 | void blk_iostatus_enable(BlockBackend *blk) | |
1037 | { | |
373340b2 HR |
1038 | blk->iostatus_enabled = true; |
1039 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; | |
1040 | } | |
1041 | ||
1042 | /* The I/O status is only enabled if the drive explicitly | |
1043 | * enables it _and_ the VM is configured to stop on errors */ | |
1044 | bool blk_iostatus_is_enabled(const BlockBackend *blk) | |
1045 | { | |
1046 | return (blk->iostatus_enabled && | |
1047 | (blk->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || | |
1048 | blk->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
1049 | blk->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
1050 | } | |
1051 | ||
1052 | BlockDeviceIoStatus blk_iostatus(const BlockBackend *blk) | |
1053 | { | |
1054 | return blk->iostatus; | |
1055 | } | |
1056 | ||
1057 | void blk_iostatus_disable(BlockBackend *blk) | |
1058 | { | |
1059 | blk->iostatus_enabled = false; | |
1060 | } | |
1061 | ||
1062 | void blk_iostatus_reset(BlockBackend *blk) | |
1063 | { | |
1064 | if (blk_iostatus_is_enabled(blk)) { | |
f21d96d0 | 1065 | BlockDriverState *bs = blk_bs(blk); |
373340b2 | 1066 | blk->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
f21d96d0 KW |
1067 | if (bs && bs->job) { |
1068 | block_job_iostatus_reset(bs->job); | |
373340b2 HR |
1069 | } |
1070 | } | |
1071 | } | |
1072 | ||
1073 | void blk_iostatus_set_err(BlockBackend *blk, int error) | |
1074 | { | |
1075 | assert(blk_iostatus_is_enabled(blk)); | |
1076 | if (blk->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
1077 | blk->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : | |
1078 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
1079 | } | |
4be74634 MA |
1080 | } |
1081 | ||
c10c9d96 KW |
1082 | void blk_set_allow_write_beyond_eof(BlockBackend *blk, bool allow) |
1083 | { | |
1084 | blk->allow_write_beyond_eof = allow; | |
1085 | } | |
1086 | ||
e7f7d676 HR |
1087 | static int blk_check_byte_request(BlockBackend *blk, int64_t offset, |
1088 | size_t size) | |
1089 | { | |
1090 | int64_t len; | |
1091 | ||
1092 | if (size > INT_MAX) { | |
1093 | return -EIO; | |
1094 | } | |
1095 | ||
c09ba36c | 1096 | if (!blk_is_available(blk)) { |
e7f7d676 HR |
1097 | return -ENOMEDIUM; |
1098 | } | |
1099 | ||
e7f7d676 HR |
1100 | if (offset < 0) { |
1101 | return -EIO; | |
1102 | } | |
1103 | ||
c10c9d96 KW |
1104 | if (!blk->allow_write_beyond_eof) { |
1105 | len = blk_getlength(blk); | |
1106 | if (len < 0) { | |
1107 | return len; | |
1108 | } | |
1109 | ||
1110 | if (offset > len || len - offset < size) { | |
1111 | return -EIO; | |
1112 | } | |
e7f7d676 HR |
1113 | } |
1114 | ||
1115 | return 0; | |
1116 | } | |
1117 | ||
1e98fefd KW |
1118 | int coroutine_fn blk_co_preadv(BlockBackend *blk, int64_t offset, |
1119 | unsigned int bytes, QEMUIOVector *qiov, | |
1120 | BdrvRequestFlags flags) | |
4be74634 | 1121 | { |
1e98fefd | 1122 | int ret; |
99723548 | 1123 | BlockDriverState *bs = blk_bs(blk); |
1e98fefd | 1124 | |
99723548 | 1125 | trace_blk_co_preadv(blk, bs, offset, bytes, flags); |
1e98fefd KW |
1126 | |
1127 | ret = blk_check_byte_request(blk, offset, bytes); | |
e7f7d676 HR |
1128 | if (ret < 0) { |
1129 | return ret; | |
1130 | } | |
1131 | ||
99723548 PB |
1132 | bdrv_inc_in_flight(bs); |
1133 | ||
441565b2 | 1134 | /* throttling disk I/O */ |
022cdc9f MP |
1135 | if (blk->public.throttle_group_member.throttle_state) { |
1136 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1137 | bytes, false); | |
441565b2 KW |
1138 | } |
1139 | ||
99723548 PB |
1140 | ret = bdrv_co_preadv(blk->root, offset, bytes, qiov, flags); |
1141 | bdrv_dec_in_flight(bs); | |
1142 | return ret; | |
1bf1cbc9 KW |
1143 | } |
1144 | ||
1e98fefd KW |
1145 | int coroutine_fn blk_co_pwritev(BlockBackend *blk, int64_t offset, |
1146 | unsigned int bytes, QEMUIOVector *qiov, | |
1147 | BdrvRequestFlags flags) | |
a8823a3b | 1148 | { |
bfd18d1e | 1149 | int ret; |
99723548 | 1150 | BlockDriverState *bs = blk_bs(blk); |
bfd18d1e | 1151 | |
99723548 | 1152 | trace_blk_co_pwritev(blk, bs, offset, bytes, flags); |
1e98fefd | 1153 | |
bfd18d1e | 1154 | ret = blk_check_byte_request(blk, offset, bytes); |
a8823a3b KW |
1155 | if (ret < 0) { |
1156 | return ret; | |
1157 | } | |
1158 | ||
99723548 | 1159 | bdrv_inc_in_flight(bs); |
441565b2 | 1160 | /* throttling disk I/O */ |
022cdc9f MP |
1161 | if (blk->public.throttle_group_member.throttle_state) { |
1162 | throttle_group_co_io_limits_intercept(&blk->public.throttle_group_member, | |
1163 | bytes, true); | |
441565b2 KW |
1164 | } |
1165 | ||
bfd18d1e KW |
1166 | if (!blk->enable_write_cache) { |
1167 | flags |= BDRV_REQ_FUA; | |
1168 | } | |
1169 | ||
99723548 PB |
1170 | ret = bdrv_co_pwritev(blk->root, offset, bytes, qiov, flags); |
1171 | bdrv_dec_in_flight(bs); | |
1172 | return ret; | |
a8823a3b KW |
1173 | } |
1174 | ||
1bf1cbc9 KW |
1175 | typedef struct BlkRwCo { |
1176 | BlockBackend *blk; | |
1177 | int64_t offset; | |
c060332c | 1178 | void *iobuf; |
1bf1cbc9 KW |
1179 | int ret; |
1180 | BdrvRequestFlags flags; | |
1181 | } BlkRwCo; | |
1182 | ||
1183 | static void blk_read_entry(void *opaque) | |
1184 | { | |
1185 | BlkRwCo *rwco = opaque; | |
c060332c | 1186 | QEMUIOVector *qiov = rwco->iobuf; |
1bf1cbc9 | 1187 | |
c060332c DS |
1188 | rwco->ret = blk_co_preadv(rwco->blk, rwco->offset, qiov->size, |
1189 | qiov, rwco->flags); | |
4720cbee | 1190 | aio_wait_kick(); |
1bf1cbc9 KW |
1191 | } |
1192 | ||
a8823a3b KW |
1193 | static void blk_write_entry(void *opaque) |
1194 | { | |
1195 | BlkRwCo *rwco = opaque; | |
c060332c | 1196 | QEMUIOVector *qiov = rwco->iobuf; |
a8823a3b | 1197 | |
c060332c DS |
1198 | rwco->ret = blk_co_pwritev(rwco->blk, rwco->offset, qiov->size, |
1199 | qiov, rwco->flags); | |
4720cbee | 1200 | aio_wait_kick(); |
a8823a3b KW |
1201 | } |
1202 | ||
a55d3fba KW |
1203 | static int blk_prw(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1204 | int64_t bytes, CoroutineEntry co_entry, | |
1205 | BdrvRequestFlags flags) | |
1bf1cbc9 | 1206 | { |
ae5a9592 VSO |
1207 | QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); |
1208 | BlkRwCo rwco = { | |
1bf1cbc9 | 1209 | .blk = blk, |
a55d3fba | 1210 | .offset = offset, |
c060332c | 1211 | .iobuf = &qiov, |
fc1453cd | 1212 | .flags = flags, |
1bf1cbc9 KW |
1213 | .ret = NOT_DONE, |
1214 | }; | |
1215 | ||
35f106e6 PB |
1216 | if (qemu_in_coroutine()) { |
1217 | /* Fast-path if already in coroutine context */ | |
1218 | co_entry(&rwco); | |
1219 | } else { | |
1220 | Coroutine *co = qemu_coroutine_create(co_entry, &rwco); | |
e92f0e19 | 1221 | bdrv_coroutine_enter(blk_bs(blk), co); |
35f106e6 PB |
1222 | BDRV_POLL_WHILE(blk_bs(blk), rwco.ret == NOT_DONE); |
1223 | } | |
1bf1cbc9 KW |
1224 | |
1225 | return rwco.ret; | |
4be74634 MA |
1226 | } |
1227 | ||
b7d17f9f EB |
1228 | int blk_pread_unthrottled(BlockBackend *blk, int64_t offset, uint8_t *buf, |
1229 | int count) | |
4be74634 | 1230 | { |
5bd51196 KW |
1231 | int ret; |
1232 | ||
b7d17f9f | 1233 | ret = blk_check_byte_request(blk, offset, count); |
e7f7d676 HR |
1234 | if (ret < 0) { |
1235 | return ret; | |
1236 | } | |
1237 | ||
c2066af0 | 1238 | blk_root_drained_begin(blk->root); |
b7d17f9f | 1239 | ret = blk_pread(blk, offset, buf, count); |
c2066af0 | 1240 | blk_root_drained_end(blk->root); |
5bd51196 | 1241 | return ret; |
4be74634 MA |
1242 | } |
1243 | ||
d004bd52 | 1244 | int blk_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 1245 | int bytes, BdrvRequestFlags flags) |
0df89e8e | 1246 | { |
f5a5ca79 | 1247 | return blk_prw(blk, offset, NULL, bytes, blk_write_entry, |
983a1600 | 1248 | flags | BDRV_REQ_ZERO_WRITE); |
0df89e8e KW |
1249 | } |
1250 | ||
720ff280 KW |
1251 | int blk_make_zero(BlockBackend *blk, BdrvRequestFlags flags) |
1252 | { | |
1253 | return bdrv_make_zero(blk->root, flags); | |
1254 | } | |
1255 | ||
c90e2a9c | 1256 | void blk_inc_in_flight(BlockBackend *blk) |
33f2a757 SH |
1257 | { |
1258 | atomic_inc(&blk->in_flight); | |
1259 | } | |
1260 | ||
c90e2a9c | 1261 | void blk_dec_in_flight(BlockBackend *blk) |
33f2a757 SH |
1262 | { |
1263 | atomic_dec(&blk->in_flight); | |
cfe29d82 | 1264 | aio_wait_kick(); |
33f2a757 SH |
1265 | } |
1266 | ||
e7f7d676 HR |
1267 | static void error_callback_bh(void *opaque) |
1268 | { | |
1269 | struct BlockBackendAIOCB *acb = opaque; | |
99723548 | 1270 | |
33f2a757 | 1271 | blk_dec_in_flight(acb->blk); |
e7f7d676 HR |
1272 | acb->common.cb(acb->common.opaque, acb->ret); |
1273 | qemu_aio_unref(acb); | |
1274 | } | |
1275 | ||
ca78ecfa PL |
1276 | BlockAIOCB *blk_abort_aio_request(BlockBackend *blk, |
1277 | BlockCompletionFunc *cb, | |
1278 | void *opaque, int ret) | |
e7f7d676 HR |
1279 | { |
1280 | struct BlockBackendAIOCB *acb; | |
e7f7d676 | 1281 | |
33f2a757 | 1282 | blk_inc_in_flight(blk); |
e7f7d676 | 1283 | acb = blk_aio_get(&block_backend_aiocb_info, blk, cb, opaque); |
4981bdec | 1284 | acb->blk = blk; |
e7f7d676 HR |
1285 | acb->ret = ret; |
1286 | ||
fffb6e12 | 1287 | aio_bh_schedule_oneshot(blk_get_aio_context(blk), error_callback_bh, acb); |
e7f7d676 HR |
1288 | return &acb->common; |
1289 | } | |
1290 | ||
57d6a428 KW |
1291 | typedef struct BlkAioEmAIOCB { |
1292 | BlockAIOCB common; | |
1293 | BlkRwCo rwco; | |
7fa84cd8 | 1294 | int bytes; |
57d6a428 | 1295 | bool has_returned; |
57d6a428 KW |
1296 | } BlkAioEmAIOCB; |
1297 | ||
1298 | static const AIOCBInfo blk_aio_em_aiocb_info = { | |
1299 | .aiocb_size = sizeof(BlkAioEmAIOCB), | |
1300 | }; | |
1301 | ||
1302 | static void blk_aio_complete(BlkAioEmAIOCB *acb) | |
1303 | { | |
57d6a428 KW |
1304 | if (acb->has_returned) { |
1305 | acb->common.cb(acb->common.opaque, acb->rwco.ret); | |
46aaf2a5 | 1306 | blk_dec_in_flight(acb->rwco.blk); |
57d6a428 KW |
1307 | qemu_aio_unref(acb); |
1308 | } | |
1309 | } | |
1310 | ||
1311 | static void blk_aio_complete_bh(void *opaque) | |
1312 | { | |
fffb6e12 | 1313 | BlkAioEmAIOCB *acb = opaque; |
fffb6e12 PB |
1314 | assert(acb->has_returned); |
1315 | blk_aio_complete(acb); | |
57d6a428 KW |
1316 | } |
1317 | ||
7fa84cd8 | 1318 | static BlockAIOCB *blk_aio_prwv(BlockBackend *blk, int64_t offset, int bytes, |
c060332c | 1319 | void *iobuf, CoroutineEntry co_entry, |
57d6a428 KW |
1320 | BdrvRequestFlags flags, |
1321 | BlockCompletionFunc *cb, void *opaque) | |
1322 | { | |
1323 | BlkAioEmAIOCB *acb; | |
1324 | Coroutine *co; | |
1325 | ||
33f2a757 | 1326 | blk_inc_in_flight(blk); |
57d6a428 KW |
1327 | acb = blk_aio_get(&blk_aio_em_aiocb_info, blk, cb, opaque); |
1328 | acb->rwco = (BlkRwCo) { | |
1329 | .blk = blk, | |
1330 | .offset = offset, | |
c060332c | 1331 | .iobuf = iobuf, |
57d6a428 KW |
1332 | .flags = flags, |
1333 | .ret = NOT_DONE, | |
1334 | }; | |
7fa84cd8 | 1335 | acb->bytes = bytes; |
57d6a428 KW |
1336 | acb->has_returned = false; |
1337 | ||
0b8b8753 | 1338 | co = qemu_coroutine_create(co_entry, acb); |
e92f0e19 | 1339 | bdrv_coroutine_enter(blk_bs(blk), co); |
57d6a428 KW |
1340 | |
1341 | acb->has_returned = true; | |
1342 | if (acb->rwco.ret != NOT_DONE) { | |
fffb6e12 PB |
1343 | aio_bh_schedule_oneshot(blk_get_aio_context(blk), |
1344 | blk_aio_complete_bh, acb); | |
57d6a428 KW |
1345 | } |
1346 | ||
1347 | return &acb->common; | |
1348 | } | |
1349 | ||
1350 | static void blk_aio_read_entry(void *opaque) | |
1351 | { | |
1352 | BlkAioEmAIOCB *acb = opaque; | |
1353 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1354 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1355 | |
c060332c | 1356 | assert(qiov->size == acb->bytes); |
7fa84cd8 | 1357 | rwco->ret = blk_co_preadv(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1358 | qiov, rwco->flags); |
57d6a428 KW |
1359 | blk_aio_complete(acb); |
1360 | } | |
1361 | ||
1362 | static void blk_aio_write_entry(void *opaque) | |
1363 | { | |
1364 | BlkAioEmAIOCB *acb = opaque; | |
1365 | BlkRwCo *rwco = &acb->rwco; | |
c060332c | 1366 | QEMUIOVector *qiov = rwco->iobuf; |
57d6a428 | 1367 | |
c060332c | 1368 | assert(!qiov || qiov->size == acb->bytes); |
7fa84cd8 | 1369 | rwco->ret = blk_co_pwritev(rwco->blk, rwco->offset, acb->bytes, |
c060332c | 1370 | qiov, rwco->flags); |
57d6a428 KW |
1371 | blk_aio_complete(acb); |
1372 | } | |
1373 | ||
d004bd52 EB |
1374 | BlockAIOCB *blk_aio_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
1375 | int count, BdrvRequestFlags flags, | |
1376 | BlockCompletionFunc *cb, void *opaque) | |
4be74634 | 1377 | { |
983a1600 EB |
1378 | return blk_aio_prwv(blk, offset, count, NULL, blk_aio_write_entry, |
1379 | flags | BDRV_REQ_ZERO_WRITE, cb, opaque); | |
4be74634 MA |
1380 | } |
1381 | ||
1382 | int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count) | |
1383 | { | |
a55d3fba | 1384 | int ret = blk_prw(blk, offset, buf, count, blk_read_entry, 0); |
e7f7d676 HR |
1385 | if (ret < 0) { |
1386 | return ret; | |
1387 | } | |
a55d3fba | 1388 | return count; |
4be74634 MA |
1389 | } |
1390 | ||
8341f00d EB |
1391 | int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count, |
1392 | BdrvRequestFlags flags) | |
4be74634 | 1393 | { |
8341f00d EB |
1394 | int ret = blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
1395 | flags); | |
e7f7d676 HR |
1396 | if (ret < 0) { |
1397 | return ret; | |
1398 | } | |
a55d3fba | 1399 | return count; |
4be74634 MA |
1400 | } |
1401 | ||
1402 | int64_t blk_getlength(BlockBackend *blk) | |
1403 | { | |
c09ba36c HR |
1404 | if (!blk_is_available(blk)) { |
1405 | return -ENOMEDIUM; | |
1406 | } | |
1407 | ||
f21d96d0 | 1408 | return bdrv_getlength(blk_bs(blk)); |
4be74634 MA |
1409 | } |
1410 | ||
1411 | void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr) | |
1412 | { | |
f21d96d0 | 1413 | if (!blk_bs(blk)) { |
a46fc9c9 HR |
1414 | *nb_sectors_ptr = 0; |
1415 | } else { | |
f21d96d0 | 1416 | bdrv_get_geometry(blk_bs(blk), nb_sectors_ptr); |
a46fc9c9 | 1417 | } |
4be74634 MA |
1418 | } |
1419 | ||
1ef01253 HR |
1420 | int64_t blk_nb_sectors(BlockBackend *blk) |
1421 | { | |
c09ba36c HR |
1422 | if (!blk_is_available(blk)) { |
1423 | return -ENOMEDIUM; | |
1424 | } | |
1425 | ||
f21d96d0 | 1426 | return bdrv_nb_sectors(blk_bs(blk)); |
1ef01253 HR |
1427 | } |
1428 | ||
60cb2fa7 EB |
1429 | BlockAIOCB *blk_aio_preadv(BlockBackend *blk, int64_t offset, |
1430 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1431 | BlockCompletionFunc *cb, void *opaque) | |
1432 | { | |
1433 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1434 | blk_aio_read_entry, flags, cb, opaque); | |
1435 | } | |
1436 | ||
60cb2fa7 EB |
1437 | BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
1438 | QEMUIOVector *qiov, BdrvRequestFlags flags, | |
1439 | BlockCompletionFunc *cb, void *opaque) | |
1440 | { | |
1441 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | |
1442 | blk_aio_write_entry, flags, cb, opaque); | |
1443 | } | |
1444 | ||
be07a889 KW |
1445 | static void blk_aio_flush_entry(void *opaque) |
1446 | { | |
1447 | BlkAioEmAIOCB *acb = opaque; | |
1448 | BlkRwCo *rwco = &acb->rwco; | |
1449 | ||
1450 | rwco->ret = blk_co_flush(rwco->blk); | |
1451 | blk_aio_complete(acb); | |
1452 | } | |
1453 | ||
4be74634 MA |
1454 | BlockAIOCB *blk_aio_flush(BlockBackend *blk, |
1455 | BlockCompletionFunc *cb, void *opaque) | |
1456 | { | |
be07a889 | 1457 | return blk_aio_prwv(blk, 0, 0, NULL, blk_aio_flush_entry, 0, cb, opaque); |
4be74634 MA |
1458 | } |
1459 | ||
8c2e3dd5 KW |
1460 | static void blk_aio_pdiscard_entry(void *opaque) |
1461 | { | |
1462 | BlkAioEmAIOCB *acb = opaque; | |
1463 | BlkRwCo *rwco = &acb->rwco; | |
1464 | ||
1465 | rwco->ret = blk_co_pdiscard(rwco->blk, rwco->offset, acb->bytes); | |
1466 | blk_aio_complete(acb); | |
1467 | } | |
1468 | ||
1c6c4bb7 | 1469 | BlockAIOCB *blk_aio_pdiscard(BlockBackend *blk, |
f5a5ca79 | 1470 | int64_t offset, int bytes, |
1c6c4bb7 | 1471 | BlockCompletionFunc *cb, void *opaque) |
4be74634 | 1472 | { |
f5a5ca79 | 1473 | return blk_aio_prwv(blk, offset, bytes, NULL, blk_aio_pdiscard_entry, 0, |
8c2e3dd5 | 1474 | cb, opaque); |
4be74634 MA |
1475 | } |
1476 | ||
1477 | void blk_aio_cancel(BlockAIOCB *acb) | |
1478 | { | |
1479 | bdrv_aio_cancel(acb); | |
1480 | } | |
1481 | ||
1482 | void blk_aio_cancel_async(BlockAIOCB *acb) | |
1483 | { | |
1484 | bdrv_aio_cancel_async(acb); | |
1485 | } | |
1486 | ||
48af776a | 1487 | int blk_co_ioctl(BlockBackend *blk, unsigned long int req, void *buf) |
4be74634 | 1488 | { |
c09ba36c HR |
1489 | if (!blk_is_available(blk)) { |
1490 | return -ENOMEDIUM; | |
1491 | } | |
1492 | ||
48af776a KW |
1493 | return bdrv_co_ioctl(blk_bs(blk), req, buf); |
1494 | } | |
1495 | ||
1496 | static void blk_ioctl_entry(void *opaque) | |
1497 | { | |
1498 | BlkRwCo *rwco = opaque; | |
c060332c DS |
1499 | QEMUIOVector *qiov = rwco->iobuf; |
1500 | ||
48af776a | 1501 | rwco->ret = blk_co_ioctl(rwco->blk, rwco->offset, |
c060332c | 1502 | qiov->iov[0].iov_base); |
4720cbee | 1503 | aio_wait_kick(); |
48af776a KW |
1504 | } |
1505 | ||
1506 | int blk_ioctl(BlockBackend *blk, unsigned long int req, void *buf) | |
1507 | { | |
1508 | return blk_prw(blk, req, buf, 0, blk_ioctl_entry, 0); | |
1509 | } | |
1510 | ||
1511 | static void blk_aio_ioctl_entry(void *opaque) | |
1512 | { | |
1513 | BlkAioEmAIOCB *acb = opaque; | |
1514 | BlkRwCo *rwco = &acb->rwco; | |
1515 | ||
c060332c DS |
1516 | rwco->ret = blk_co_ioctl(rwco->blk, rwco->offset, rwco->iobuf); |
1517 | ||
48af776a | 1518 | blk_aio_complete(acb); |
4be74634 MA |
1519 | } |
1520 | ||
1521 | BlockAIOCB *blk_aio_ioctl(BlockBackend *blk, unsigned long int req, void *buf, | |
1522 | BlockCompletionFunc *cb, void *opaque) | |
1523 | { | |
c060332c | 1524 | return blk_aio_prwv(blk, req, 0, buf, blk_aio_ioctl_entry, 0, cb, opaque); |
4be74634 MA |
1525 | } |
1526 | ||
f5a5ca79 | 1527 | int blk_co_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
2bb0dce7 | 1528 | { |
f5a5ca79 | 1529 | int ret = blk_check_byte_request(blk, offset, bytes); |
e7f7d676 HR |
1530 | if (ret < 0) { |
1531 | return ret; | |
1532 | } | |
1533 | ||
0b9fd3f4 | 1534 | return bdrv_co_pdiscard(blk->root, offset, bytes); |
2bb0dce7 HR |
1535 | } |
1536 | ||
1537 | int blk_co_flush(BlockBackend *blk) | |
1538 | { | |
c09ba36c HR |
1539 | if (!blk_is_available(blk)) { |
1540 | return -ENOMEDIUM; | |
1541 | } | |
1542 | ||
f21d96d0 | 1543 | return bdrv_co_flush(blk_bs(blk)); |
2bb0dce7 HR |
1544 | } |
1545 | ||
be07a889 | 1546 | static void blk_flush_entry(void *opaque) |
4be74634 | 1547 | { |
be07a889 KW |
1548 | BlkRwCo *rwco = opaque; |
1549 | rwco->ret = blk_co_flush(rwco->blk); | |
4720cbee | 1550 | aio_wait_kick(); |
be07a889 | 1551 | } |
c09ba36c | 1552 | |
be07a889 KW |
1553 | int blk_flush(BlockBackend *blk) |
1554 | { | |
1555 | return blk_prw(blk, 0, NULL, 0, blk_flush_entry, 0); | |
4be74634 MA |
1556 | } |
1557 | ||
97b0385a AY |
1558 | void blk_drain(BlockBackend *blk) |
1559 | { | |
33f2a757 SH |
1560 | BlockDriverState *bs = blk_bs(blk); |
1561 | ||
1562 | if (bs) { | |
1563 | bdrv_drained_begin(bs); | |
1564 | } | |
1565 | ||
1566 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 KW |
1567 | AIO_WAIT_WHILE(blk_get_aio_context(blk), |
1568 | atomic_mb_read(&blk->in_flight) > 0); | |
33f2a757 SH |
1569 | |
1570 | if (bs) { | |
1571 | bdrv_drained_end(bs); | |
a46fc9c9 | 1572 | } |
97b0385a AY |
1573 | } |
1574 | ||
4be74634 MA |
1575 | void blk_drain_all(void) |
1576 | { | |
33f2a757 SH |
1577 | BlockBackend *blk = NULL; |
1578 | ||
1579 | bdrv_drain_all_begin(); | |
1580 | ||
1581 | while ((blk = blk_all_next(blk)) != NULL) { | |
1582 | AioContext *ctx = blk_get_aio_context(blk); | |
1583 | ||
1584 | aio_context_acquire(ctx); | |
1585 | ||
1586 | /* We may have -ENOMEDIUM completions in flight */ | |
cfe29d82 | 1587 | AIO_WAIT_WHILE(ctx, atomic_mb_read(&blk->in_flight) > 0); |
33f2a757 SH |
1588 | |
1589 | aio_context_release(ctx); | |
1590 | } | |
1591 | ||
1592 | bdrv_drain_all_end(); | |
4be74634 MA |
1593 | } |
1594 | ||
373340b2 HR |
1595 | void blk_set_on_error(BlockBackend *blk, BlockdevOnError on_read_error, |
1596 | BlockdevOnError on_write_error) | |
1597 | { | |
1598 | blk->on_read_error = on_read_error; | |
1599 | blk->on_write_error = on_write_error; | |
1600 | } | |
1601 | ||
4be74634 MA |
1602 | BlockdevOnError blk_get_on_error(BlockBackend *blk, bool is_read) |
1603 | { | |
373340b2 | 1604 | return is_read ? blk->on_read_error : blk->on_write_error; |
4be74634 MA |
1605 | } |
1606 | ||
1607 | BlockErrorAction blk_get_error_action(BlockBackend *blk, bool is_read, | |
1608 | int error) | |
1609 | { | |
373340b2 HR |
1610 | BlockdevOnError on_err = blk_get_on_error(blk, is_read); |
1611 | ||
1612 | switch (on_err) { | |
1613 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
1614 | return (error == ENOSPC) ? | |
1615 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
1616 | case BLOCKDEV_ON_ERROR_STOP: | |
1617 | return BLOCK_ERROR_ACTION_STOP; | |
1618 | case BLOCKDEV_ON_ERROR_REPORT: | |
1619 | return BLOCK_ERROR_ACTION_REPORT; | |
1620 | case BLOCKDEV_ON_ERROR_IGNORE: | |
1621 | return BLOCK_ERROR_ACTION_IGNORE; | |
8c398252 | 1622 | case BLOCKDEV_ON_ERROR_AUTO: |
373340b2 HR |
1623 | default: |
1624 | abort(); | |
1625 | } | |
4be74634 MA |
1626 | } |
1627 | ||
373340b2 HR |
1628 | static void send_qmp_error_event(BlockBackend *blk, |
1629 | BlockErrorAction action, | |
1630 | bool is_read, int error) | |
1631 | { | |
1632 | IoOperationType optype; | |
bfe1a14c | 1633 | BlockDriverState *bs = blk_bs(blk); |
373340b2 HR |
1634 | |
1635 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; | |
bfe1a14c KW |
1636 | qapi_event_send_block_io_error(blk_name(blk), !!bs, |
1637 | bs ? bdrv_get_node_name(bs) : NULL, optype, | |
2bf7e10f | 1638 | action, blk_iostatus_is_enabled(blk), |
3ab72385 | 1639 | error == ENOSPC, strerror(error)); |
373340b2 HR |
1640 | } |
1641 | ||
1642 | /* This is done by device models because, while the block layer knows | |
1643 | * about the error, it does not know whether an operation comes from | |
1644 | * the device or the block layer (from a job, for example). | |
1645 | */ | |
4be74634 MA |
1646 | void blk_error_action(BlockBackend *blk, BlockErrorAction action, |
1647 | bool is_read, int error) | |
1648 | { | |
373340b2 HR |
1649 | assert(error >= 0); |
1650 | ||
1651 | if (action == BLOCK_ERROR_ACTION_STOP) { | |
1652 | /* First set the iostatus, so that "info block" returns an iostatus | |
1653 | * that matches the events raised so far (an additional error iostatus | |
1654 | * is fine, but not a lost one). | |
1655 | */ | |
1656 | blk_iostatus_set_err(blk, error); | |
1657 | ||
1658 | /* Then raise the request to stop the VM and the event. | |
1659 | * qemu_system_vmstop_request_prepare has two effects. First, | |
1660 | * it ensures that the STOP event always comes after the | |
1661 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
1662 | * can observe the STOP event and do a "cont" before the STOP | |
1663 | * event is issued, the VM will not stop. In this case, vm_start() | |
1664 | * also ensures that the STOP/RESUME pair of events is emitted. | |
1665 | */ | |
1666 | qemu_system_vmstop_request_prepare(); | |
1667 | send_qmp_error_event(blk, action, is_read, error); | |
1668 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); | |
1669 | } else { | |
1670 | send_qmp_error_event(blk, action, is_read, error); | |
1671 | } | |
4be74634 MA |
1672 | } |
1673 | ||
96710565 | 1674 | bool blk_is_read_only(BlockBackend *blk) |
4be74634 | 1675 | { |
f21d96d0 KW |
1676 | BlockDriverState *bs = blk_bs(blk); |
1677 | ||
1678 | if (bs) { | |
1679 | return bdrv_is_read_only(bs); | |
061959e8 HR |
1680 | } else { |
1681 | return blk->root_state.read_only; | |
1682 | } | |
4be74634 MA |
1683 | } |
1684 | ||
96710565 | 1685 | bool blk_is_sg(BlockBackend *blk) |
4be74634 | 1686 | { |
f21d96d0 KW |
1687 | BlockDriverState *bs = blk_bs(blk); |
1688 | ||
1689 | if (!bs) { | |
96710565 | 1690 | return false; |
a46fc9c9 HR |
1691 | } |
1692 | ||
f21d96d0 | 1693 | return bdrv_is_sg(bs); |
4be74634 MA |
1694 | } |
1695 | ||
96710565 | 1696 | bool blk_enable_write_cache(BlockBackend *blk) |
4be74634 | 1697 | { |
bfd18d1e | 1698 | return blk->enable_write_cache; |
4be74634 MA |
1699 | } |
1700 | ||
1701 | void blk_set_enable_write_cache(BlockBackend *blk, bool wce) | |
1702 | { | |
bfd18d1e | 1703 | blk->enable_write_cache = wce; |
4be74634 MA |
1704 | } |
1705 | ||
2bb0dce7 HR |
1706 | void blk_invalidate_cache(BlockBackend *blk, Error **errp) |
1707 | { | |
f21d96d0 KW |
1708 | BlockDriverState *bs = blk_bs(blk); |
1709 | ||
1710 | if (!bs) { | |
c09ba36c HR |
1711 | error_setg(errp, "Device '%s' has no medium", blk->name); |
1712 | return; | |
1713 | } | |
1714 | ||
f21d96d0 | 1715 | bdrv_invalidate_cache(bs, errp); |
2bb0dce7 HR |
1716 | } |
1717 | ||
e031f750 | 1718 | bool blk_is_inserted(BlockBackend *blk) |
4be74634 | 1719 | { |
f21d96d0 KW |
1720 | BlockDriverState *bs = blk_bs(blk); |
1721 | ||
1722 | return bs && bdrv_is_inserted(bs); | |
db0284f8 HR |
1723 | } |
1724 | ||
1725 | bool blk_is_available(BlockBackend *blk) | |
1726 | { | |
1727 | return blk_is_inserted(blk) && !blk_dev_is_tray_open(blk); | |
4be74634 MA |
1728 | } |
1729 | ||
1730 | void blk_lock_medium(BlockBackend *blk, bool locked) | |
1731 | { | |
f21d96d0 KW |
1732 | BlockDriverState *bs = blk_bs(blk); |
1733 | ||
1734 | if (bs) { | |
1735 | bdrv_lock_medium(bs, locked); | |
a46fc9c9 | 1736 | } |
4be74634 MA |
1737 | } |
1738 | ||
1739 | void blk_eject(BlockBackend *blk, bool eject_flag) | |
1740 | { | |
f21d96d0 | 1741 | BlockDriverState *bs = blk_bs(blk); |
2d76e724 KW |
1742 | char *id; |
1743 | ||
f21d96d0 KW |
1744 | if (bs) { |
1745 | bdrv_eject(bs, eject_flag); | |
a46fc9c9 | 1746 | } |
c47ee043 JS |
1747 | |
1748 | /* Whether or not we ejected on the backend, | |
1749 | * the frontend experienced a tray event. */ | |
1750 | id = blk_get_attached_dev_id(blk); | |
1751 | qapi_event_send_device_tray_moved(blk_name(blk), id, | |
3ab72385 | 1752 | eject_flag); |
c47ee043 | 1753 | g_free(id); |
4be74634 MA |
1754 | } |
1755 | ||
1756 | int blk_get_flags(BlockBackend *blk) | |
1757 | { | |
f21d96d0 KW |
1758 | BlockDriverState *bs = blk_bs(blk); |
1759 | ||
1760 | if (bs) { | |
1761 | return bdrv_get_flags(bs); | |
061959e8 HR |
1762 | } else { |
1763 | return blk->root_state.open_flags; | |
1764 | } | |
4be74634 MA |
1765 | } |
1766 | ||
5def6b80 EB |
1767 | /* Returns the maximum transfer length, in bytes; guaranteed nonzero */ |
1768 | uint32_t blk_get_max_transfer(BlockBackend *blk) | |
454057b7 | 1769 | { |
f21d96d0 | 1770 | BlockDriverState *bs = blk_bs(blk); |
5def6b80 | 1771 | uint32_t max = 0; |
f21d96d0 KW |
1772 | |
1773 | if (bs) { | |
5def6b80 | 1774 | max = bs->bl.max_transfer; |
a46fc9c9 | 1775 | } |
5def6b80 | 1776 | return MIN_NON_ZERO(max, INT_MAX); |
454057b7 PL |
1777 | } |
1778 | ||
648296e0 SH |
1779 | int blk_get_max_iov(BlockBackend *blk) |
1780 | { | |
f21d96d0 | 1781 | return blk->root->bs->bl.max_iov; |
648296e0 SH |
1782 | } |
1783 | ||
4be74634 MA |
1784 | void blk_set_guest_block_size(BlockBackend *blk, int align) |
1785 | { | |
68e9ec01 | 1786 | blk->guest_block_size = align; |
4be74634 MA |
1787 | } |
1788 | ||
f1c17521 PB |
1789 | void *blk_try_blockalign(BlockBackend *blk, size_t size) |
1790 | { | |
f21d96d0 | 1791 | return qemu_try_blockalign(blk ? blk_bs(blk) : NULL, size); |
f1c17521 PB |
1792 | } |
1793 | ||
4be74634 MA |
1794 | void *blk_blockalign(BlockBackend *blk, size_t size) |
1795 | { | |
f21d96d0 | 1796 | return qemu_blockalign(blk ? blk_bs(blk) : NULL, size); |
4be74634 MA |
1797 | } |
1798 | ||
1799 | bool blk_op_is_blocked(BlockBackend *blk, BlockOpType op, Error **errp) | |
1800 | { | |
f21d96d0 KW |
1801 | BlockDriverState *bs = blk_bs(blk); |
1802 | ||
1803 | if (!bs) { | |
a46fc9c9 HR |
1804 | return false; |
1805 | } | |
1806 | ||
f21d96d0 | 1807 | return bdrv_op_is_blocked(bs, op, errp); |
4be74634 MA |
1808 | } |
1809 | ||
1810 | void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason) | |
1811 | { | |
f21d96d0 KW |
1812 | BlockDriverState *bs = blk_bs(blk); |
1813 | ||
1814 | if (bs) { | |
1815 | bdrv_op_unblock(bs, op, reason); | |
a46fc9c9 | 1816 | } |
4be74634 MA |
1817 | } |
1818 | ||
1819 | void blk_op_block_all(BlockBackend *blk, Error *reason) | |
1820 | { | |
f21d96d0 KW |
1821 | BlockDriverState *bs = blk_bs(blk); |
1822 | ||
1823 | if (bs) { | |
1824 | bdrv_op_block_all(bs, reason); | |
a46fc9c9 | 1825 | } |
4be74634 MA |
1826 | } |
1827 | ||
1828 | void blk_op_unblock_all(BlockBackend *blk, Error *reason) | |
1829 | { | |
f21d96d0 KW |
1830 | BlockDriverState *bs = blk_bs(blk); |
1831 | ||
1832 | if (bs) { | |
1833 | bdrv_op_unblock_all(bs, reason); | |
a46fc9c9 | 1834 | } |
4be74634 MA |
1835 | } |
1836 | ||
1837 | AioContext *blk_get_aio_context(BlockBackend *blk) | |
1838 | { | |
7803696d | 1839 | return bdrv_get_aio_context(blk_bs(blk)); |
4981bdec HR |
1840 | } |
1841 | ||
1842 | static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb) | |
1843 | { | |
1844 | BlockBackendAIOCB *blk_acb = DO_UPCAST(BlockBackendAIOCB, common, acb); | |
1845 | return blk_get_aio_context(blk_acb->blk); | |
4be74634 MA |
1846 | } |
1847 | ||
1848 | void blk_set_aio_context(BlockBackend *blk, AioContext *new_context) | |
1849 | { | |
f21d96d0 | 1850 | BlockDriverState *bs = blk_bs(blk); |
c61791fc | 1851 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; |
f21d96d0 KW |
1852 | |
1853 | if (bs) { | |
c61791fc | 1854 | if (tgm->throttle_state) { |
dc868fb0 | 1855 | bdrv_drained_begin(bs); |
c61791fc MP |
1856 | throttle_group_detach_aio_context(tgm); |
1857 | throttle_group_attach_aio_context(tgm, new_context); | |
dc868fb0 | 1858 | bdrv_drained_end(bs); |
7ca7f0f6 | 1859 | } |
f21d96d0 | 1860 | bdrv_set_aio_context(bs, new_context); |
a46fc9c9 | 1861 | } |
4be74634 MA |
1862 | } |
1863 | ||
2019ba0a HR |
1864 | void blk_add_aio_context_notifier(BlockBackend *blk, |
1865 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
1866 | void (*detach_aio_context)(void *opaque), void *opaque) | |
1867 | { | |
d03654ea | 1868 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
1869 | BlockDriverState *bs = blk_bs(blk); |
1870 | ||
d03654ea SH |
1871 | notifier = g_new(BlockBackendAioNotifier, 1); |
1872 | notifier->attached_aio_context = attached_aio_context; | |
1873 | notifier->detach_aio_context = detach_aio_context; | |
1874 | notifier->opaque = opaque; | |
1875 | QLIST_INSERT_HEAD(&blk->aio_notifiers, notifier, list); | |
1876 | ||
f21d96d0 KW |
1877 | if (bs) { |
1878 | bdrv_add_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
1879 | detach_aio_context, opaque); |
1880 | } | |
2019ba0a HR |
1881 | } |
1882 | ||
1883 | void blk_remove_aio_context_notifier(BlockBackend *blk, | |
1884 | void (*attached_aio_context)(AioContext *, | |
1885 | void *), | |
1886 | void (*detach_aio_context)(void *), | |
1887 | void *opaque) | |
1888 | { | |
d03654ea | 1889 | BlockBackendAioNotifier *notifier; |
f21d96d0 KW |
1890 | BlockDriverState *bs = blk_bs(blk); |
1891 | ||
1892 | if (bs) { | |
1893 | bdrv_remove_aio_context_notifier(bs, attached_aio_context, | |
a46fc9c9 HR |
1894 | detach_aio_context, opaque); |
1895 | } | |
d03654ea SH |
1896 | |
1897 | QLIST_FOREACH(notifier, &blk->aio_notifiers, list) { | |
1898 | if (notifier->attached_aio_context == attached_aio_context && | |
1899 | notifier->detach_aio_context == detach_aio_context && | |
1900 | notifier->opaque == opaque) { | |
1901 | QLIST_REMOVE(notifier, list); | |
1902 | g_free(notifier); | |
1903 | return; | |
1904 | } | |
1905 | } | |
1906 | ||
1907 | abort(); | |
2019ba0a HR |
1908 | } |
1909 | ||
3301f6c6 HR |
1910 | void blk_add_remove_bs_notifier(BlockBackend *blk, Notifier *notify) |
1911 | { | |
1912 | notifier_list_add(&blk->remove_bs_notifiers, notify); | |
1913 | } | |
1914 | ||
1915 | void blk_add_insert_bs_notifier(BlockBackend *blk, Notifier *notify) | |
1916 | { | |
1917 | notifier_list_add(&blk->insert_bs_notifiers, notify); | |
1918 | } | |
1919 | ||
4be74634 MA |
1920 | void blk_io_plug(BlockBackend *blk) |
1921 | { | |
f21d96d0 KW |
1922 | BlockDriverState *bs = blk_bs(blk); |
1923 | ||
1924 | if (bs) { | |
1925 | bdrv_io_plug(bs); | |
a46fc9c9 | 1926 | } |
4be74634 MA |
1927 | } |
1928 | ||
1929 | void blk_io_unplug(BlockBackend *blk) | |
1930 | { | |
f21d96d0 KW |
1931 | BlockDriverState *bs = blk_bs(blk); |
1932 | ||
1933 | if (bs) { | |
1934 | bdrv_io_unplug(bs); | |
a46fc9c9 | 1935 | } |
4be74634 MA |
1936 | } |
1937 | ||
1938 | BlockAcctStats *blk_get_stats(BlockBackend *blk) | |
1939 | { | |
7f0e9da6 | 1940 | return &blk->stats; |
4be74634 MA |
1941 | } |
1942 | ||
1943 | void *blk_aio_get(const AIOCBInfo *aiocb_info, BlockBackend *blk, | |
1944 | BlockCompletionFunc *cb, void *opaque) | |
1945 | { | |
1946 | return qemu_aio_get(aiocb_info, blk_bs(blk), cb, opaque); | |
1947 | } | |
1ef01253 | 1948 | |
d004bd52 | 1949 | int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, |
f5a5ca79 | 1950 | int bytes, BdrvRequestFlags flags) |
1ef01253 | 1951 | { |
f5a5ca79 | 1952 | return blk_co_pwritev(blk, offset, bytes, NULL, |
16aaf975 | 1953 | flags | BDRV_REQ_ZERO_WRITE); |
1ef01253 HR |
1954 | } |
1955 | ||
fe5c1355 PB |
1956 | int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, |
1957 | int count) | |
1ef01253 | 1958 | { |
35fadca8 PB |
1959 | return blk_prw(blk, offset, (void *) buf, count, blk_write_entry, |
1960 | BDRV_REQ_WRITE_COMPRESSED); | |
1ef01253 HR |
1961 | } |
1962 | ||
3a691c50 HR |
1963 | int blk_truncate(BlockBackend *blk, int64_t offset, PreallocMode prealloc, |
1964 | Error **errp) | |
1ef01253 | 1965 | { |
c09ba36c | 1966 | if (!blk_is_available(blk)) { |
ed3d2ec9 | 1967 | error_setg(errp, "No medium inserted"); |
c09ba36c HR |
1968 | return -ENOMEDIUM; |
1969 | } | |
1970 | ||
3a691c50 | 1971 | return bdrv_truncate(blk->root, offset, prealloc, errp); |
1ef01253 HR |
1972 | } |
1973 | ||
8c2e3dd5 | 1974 | static void blk_pdiscard_entry(void *opaque) |
1ef01253 | 1975 | { |
8c2e3dd5 | 1976 | BlkRwCo *rwco = opaque; |
c060332c DS |
1977 | QEMUIOVector *qiov = rwco->iobuf; |
1978 | ||
1979 | rwco->ret = blk_co_pdiscard(rwco->blk, rwco->offset, qiov->size); | |
4720cbee | 1980 | aio_wait_kick(); |
8c2e3dd5 | 1981 | } |
e7f7d676 | 1982 | |
f5a5ca79 | 1983 | int blk_pdiscard(BlockBackend *blk, int64_t offset, int bytes) |
8c2e3dd5 | 1984 | { |
f5a5ca79 | 1985 | return blk_prw(blk, offset, NULL, bytes, blk_pdiscard_entry, 0); |
1ef01253 HR |
1986 | } |
1987 | ||
1988 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, | |
1989 | int64_t pos, int size) | |
1990 | { | |
bfd18d1e KW |
1991 | int ret; |
1992 | ||
c09ba36c HR |
1993 | if (!blk_is_available(blk)) { |
1994 | return -ENOMEDIUM; | |
1995 | } | |
1996 | ||
bfd18d1e KW |
1997 | ret = bdrv_save_vmstate(blk_bs(blk), buf, pos, size); |
1998 | if (ret < 0) { | |
1999 | return ret; | |
2000 | } | |
2001 | ||
2002 | if (ret == size && !blk->enable_write_cache) { | |
2003 | ret = bdrv_flush(blk_bs(blk)); | |
2004 | } | |
2005 | ||
2006 | return ret < 0 ? ret : size; | |
1ef01253 HR |
2007 | } |
2008 | ||
2009 | int blk_load_vmstate(BlockBackend *blk, uint8_t *buf, int64_t pos, int size) | |
2010 | { | |
c09ba36c HR |
2011 | if (!blk_is_available(blk)) { |
2012 | return -ENOMEDIUM; | |
2013 | } | |
2014 | ||
f21d96d0 | 2015 | return bdrv_load_vmstate(blk_bs(blk), buf, pos, size); |
1ef01253 | 2016 | } |
f0272c4d ET |
2017 | |
2018 | int blk_probe_blocksizes(BlockBackend *blk, BlockSizes *bsz) | |
2019 | { | |
c09ba36c HR |
2020 | if (!blk_is_available(blk)) { |
2021 | return -ENOMEDIUM; | |
2022 | } | |
2023 | ||
f21d96d0 | 2024 | return bdrv_probe_blocksizes(blk_bs(blk), bsz); |
f0272c4d ET |
2025 | } |
2026 | ||
2027 | int blk_probe_geometry(BlockBackend *blk, HDGeometry *geo) | |
2028 | { | |
c09ba36c HR |
2029 | if (!blk_is_available(blk)) { |
2030 | return -ENOMEDIUM; | |
2031 | } | |
2032 | ||
f21d96d0 | 2033 | return bdrv_probe_geometry(blk_bs(blk), geo); |
f0272c4d | 2034 | } |
281d22d8 HR |
2035 | |
2036 | /* | |
2037 | * Updates the BlockBackendRootState object with data from the currently | |
2038 | * attached BlockDriverState. | |
2039 | */ | |
2040 | void blk_update_root_state(BlockBackend *blk) | |
2041 | { | |
f21d96d0 | 2042 | assert(blk->root); |
281d22d8 | 2043 | |
f21d96d0 KW |
2044 | blk->root_state.open_flags = blk->root->bs->open_flags; |
2045 | blk->root_state.read_only = blk->root->bs->read_only; | |
2046 | blk->root_state.detect_zeroes = blk->root->bs->detect_zeroes; | |
281d22d8 HR |
2047 | } |
2048 | ||
38cb18f5 | 2049 | /* |
b85114f8 KW |
2050 | * Returns the detect-zeroes setting to be used for bdrv_open() of a |
2051 | * BlockDriverState which is supposed to inherit the root state. | |
38cb18f5 | 2052 | */ |
b85114f8 | 2053 | bool blk_get_detect_zeroes_from_root_state(BlockBackend *blk) |
38cb18f5 | 2054 | { |
b85114f8 | 2055 | return blk->root_state.detect_zeroes; |
38cb18f5 HR |
2056 | } |
2057 | ||
2058 | /* | |
2059 | * Returns the flags to be used for bdrv_open() of a BlockDriverState which is | |
2060 | * supposed to inherit the root state. | |
2061 | */ | |
2062 | int blk_get_open_flags_from_root_state(BlockBackend *blk) | |
2063 | { | |
2064 | int bs_flags; | |
2065 | ||
2066 | bs_flags = blk->root_state.read_only ? 0 : BDRV_O_RDWR; | |
2067 | bs_flags |= blk->root_state.open_flags & ~BDRV_O_RDWR; | |
2068 | ||
2069 | return bs_flags; | |
2070 | } | |
2071 | ||
281d22d8 HR |
2072 | BlockBackendRootState *blk_get_root_state(BlockBackend *blk) |
2073 | { | |
2074 | return &blk->root_state; | |
2075 | } | |
1393f212 HR |
2076 | |
2077 | int blk_commit_all(void) | |
2078 | { | |
fe1a9cbc HR |
2079 | BlockBackend *blk = NULL; |
2080 | ||
2081 | while ((blk = blk_all_next(blk)) != NULL) { | |
2082 | AioContext *aio_context = blk_get_aio_context(blk); | |
2083 | ||
2084 | aio_context_acquire(aio_context); | |
f21d96d0 KW |
2085 | if (blk_is_inserted(blk) && blk->root->bs->backing) { |
2086 | int ret = bdrv_commit(blk->root->bs); | |
fe1a9cbc HR |
2087 | if (ret < 0) { |
2088 | aio_context_release(aio_context); | |
2089 | return ret; | |
2090 | } | |
2091 | } | |
2092 | aio_context_release(aio_context); | |
2093 | } | |
2094 | return 0; | |
2095 | } | |
2096 | ||
97148076 KW |
2097 | |
2098 | /* throttling disk I/O limits */ | |
2099 | void blk_set_io_limits(BlockBackend *blk, ThrottleConfig *cfg) | |
2100 | { | |
022cdc9f | 2101 | throttle_group_config(&blk->public.throttle_group_member, cfg); |
97148076 KW |
2102 | } |
2103 | ||
2104 | void blk_io_limits_disable(BlockBackend *blk) | |
2105 | { | |
48bf7ea8 AG |
2106 | BlockDriverState *bs = blk_bs(blk); |
2107 | ThrottleGroupMember *tgm = &blk->public.throttle_group_member; | |
2108 | assert(tgm->throttle_state); | |
2109 | if (bs) { | |
2110 | bdrv_drained_begin(bs); | |
2111 | } | |
2112 | throttle_group_unregister_tgm(tgm); | |
2113 | if (bs) { | |
2114 | bdrv_drained_end(bs); | |
2115 | } | |
97148076 KW |
2116 | } |
2117 | ||
2118 | /* should be called before blk_set_io_limits if a limit is set */ | |
2119 | void blk_io_limits_enable(BlockBackend *blk, const char *group) | |
2120 | { | |
022cdc9f | 2121 | assert(!blk->public.throttle_group_member.throttle_state); |
c61791fc MP |
2122 | throttle_group_register_tgm(&blk->public.throttle_group_member, |
2123 | group, blk_get_aio_context(blk)); | |
97148076 KW |
2124 | } |
2125 | ||
2126 | void blk_io_limits_update_group(BlockBackend *blk, const char *group) | |
2127 | { | |
2128 | /* this BB is not part of any group */ | |
022cdc9f | 2129 | if (!blk->public.throttle_group_member.throttle_state) { |
97148076 KW |
2130 | return; |
2131 | } | |
2132 | ||
2133 | /* this BB is a part of the same group than the one we want */ | |
022cdc9f MP |
2134 | if (!g_strcmp0(throttle_group_get_name(&blk->public.throttle_group_member), |
2135 | group)) { | |
97148076 KW |
2136 | return; |
2137 | } | |
2138 | ||
2139 | /* need to change the group this bs belong to */ | |
2140 | blk_io_limits_disable(blk); | |
2141 | blk_io_limits_enable(blk, group); | |
2142 | } | |
c2066af0 KW |
2143 | |
2144 | static void blk_root_drained_begin(BdrvChild *child) | |
2145 | { | |
2146 | BlockBackend *blk = child->opaque; | |
2147 | ||
f4d9cc88 JS |
2148 | if (++blk->quiesce_counter == 1) { |
2149 | if (blk->dev_ops && blk->dev_ops->drained_begin) { | |
2150 | blk->dev_ops->drained_begin(blk->dev_opaque); | |
2151 | } | |
2152 | } | |
2153 | ||
36fe1331 KW |
2154 | /* Note that blk->root may not be accessible here yet if we are just |
2155 | * attaching to a BlockDriverState that is drained. Use child instead. */ | |
2156 | ||
022cdc9f MP |
2157 | if (atomic_fetch_inc(&blk->public.throttle_group_member.io_limits_disabled) == 0) { |
2158 | throttle_group_restart_tgm(&blk->public.throttle_group_member); | |
c2066af0 KW |
2159 | } |
2160 | } | |
2161 | ||
fe5258a5 KW |
2162 | static bool blk_root_drained_poll(BdrvChild *child) |
2163 | { | |
2164 | BlockBackend *blk = child->opaque; | |
2165 | assert(blk->quiesce_counter); | |
2166 | return !!blk->in_flight; | |
2167 | } | |
2168 | ||
c2066af0 KW |
2169 | static void blk_root_drained_end(BdrvChild *child) |
2170 | { | |
2171 | BlockBackend *blk = child->opaque; | |
f4d9cc88 | 2172 | assert(blk->quiesce_counter); |
c2066af0 | 2173 | |
022cdc9f MP |
2174 | assert(blk->public.throttle_group_member.io_limits_disabled); |
2175 | atomic_dec(&blk->public.throttle_group_member.io_limits_disabled); | |
f4d9cc88 JS |
2176 | |
2177 | if (--blk->quiesce_counter == 0) { | |
2178 | if (blk->dev_ops && blk->dev_ops->drained_end) { | |
2179 | blk->dev_ops->drained_end(blk->dev_opaque); | |
2180 | } | |
2181 | } | |
c2066af0 | 2182 | } |
23d0ba93 FZ |
2183 | |
2184 | void blk_register_buf(BlockBackend *blk, void *host, size_t size) | |
2185 | { | |
2186 | bdrv_register_buf(blk_bs(blk), host, size); | |
2187 | } | |
2188 | ||
2189 | void blk_unregister_buf(BlockBackend *blk, void *host) | |
2190 | { | |
2191 | bdrv_unregister_buf(blk_bs(blk), host); | |
2192 | } | |
b5679fa4 FZ |
2193 | |
2194 | int coroutine_fn blk_co_copy_range(BlockBackend *blk_in, int64_t off_in, | |
2195 | BlockBackend *blk_out, int64_t off_out, | |
67b51fb9 VSO |
2196 | int bytes, BdrvRequestFlags read_flags, |
2197 | BdrvRequestFlags write_flags) | |
b5679fa4 FZ |
2198 | { |
2199 | int r; | |
2200 | r = blk_check_byte_request(blk_in, off_in, bytes); | |
2201 | if (r) { | |
2202 | return r; | |
2203 | } | |
2204 | r = blk_check_byte_request(blk_out, off_out, bytes); | |
2205 | if (r) { | |
2206 | return r; | |
2207 | } | |
2208 | return bdrv_co_copy_range(blk_in->root, off_in, | |
2209 | blk_out->root, off_out, | |
67b51fb9 | 2210 | bytes, read_flags, write_flags); |
b5679fa4 | 2211 | } |
5d3b4e99 VSO |
2212 | |
2213 | const BdrvChild *blk_root(BlockBackend *blk) | |
2214 | { | |
2215 | return blk->root; | |
2216 | } |