@samitouri / QOSamiQemu
Code Commits Issues Pulls Sessions CI/CD
master
master
h 459 lines 17 KB
Raw
1 /*
2 * QEMU System Emulator block driver
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24 #ifndef BLOCK_IO_H
25 #define BLOCK_IO_H
26
27 #include "qemu/aiocb.h"
28 #include "qemu/aio-wait.h"
29 #include "block/block-common.h"
30 #include "block/graph-lock.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/iov.h"
33
34 /*
35 * I/O API functions. These functions are thread-safe, and therefore
36 * can run in any thread.
37 *
38 * These functions can only call functions from I/O and Common categories,
39 * but can be invoked by GS, "I/O or GS" and I/O APIs.
40 *
41 * All functions in this category must use the macro
42 * IO_CODE();
43 * to catch when they are accidentally called by the wrong API.
44 */
45
46 int co_wrapper_mixed_bdrv_rdlock
47 bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset, int64_t bytes,
48 BdrvRequestFlags flags);
49
50 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
51
52 int co_wrapper_mixed_bdrv_rdlock
53 bdrv_pread(BdrvChild *child, int64_t offset, int64_t bytes, void *buf,
54 BdrvRequestFlags flags);
55
56 int co_wrapper_mixed_bdrv_rdlock
57 bdrv_pwrite(BdrvChild *child, int64_t offset,int64_t bytes,
58 const void *buf, BdrvRequestFlags flags);
59
60 int co_wrapper_mixed_bdrv_rdlock
61 bdrv_pwrite_sync(BdrvChild *child, int64_t offset, int64_t bytes,
62 const void *buf, BdrvRequestFlags flags);
63
64 int coroutine_fn GRAPH_RDLOCK
65 bdrv_co_pwrite_sync(BdrvChild *child, int64_t offset, int64_t bytes,
66 const void *buf, BdrvRequestFlags flags);
67
68 /*
69 * Efficiently zero a region of the disk image. Note that this is a regular
70 * I/O request like read or write and should have a reasonable size. This
71 * function is not suitable for zeroing the entire image in a single request
72 * because it may allocate memory for the entire region.
73 */
74 int coroutine_fn GRAPH_RDLOCK
75 bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset, int64_t bytes,
76 BdrvRequestFlags flags);
77
78 int coroutine_fn GRAPH_RDLOCK
79 bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
80 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
81
82 int64_t coroutine_fn GRAPH_RDLOCK bdrv_co_nb_sectors(BlockDriverState *bs);
83 int64_t coroutine_mixed_fn bdrv_nb_sectors(BlockDriverState *bs);
84
85 int64_t coroutine_fn GRAPH_RDLOCK bdrv_co_getlength(BlockDriverState *bs);
86 int64_t co_wrapper_mixed_bdrv_rdlock bdrv_getlength(BlockDriverState *bs);
87
88 int64_t coroutine_fn GRAPH_RDLOCK
89 bdrv_co_get_allocated_file_size(BlockDriverState *bs);
90
91 int64_t co_wrapper_bdrv_rdlock
92 bdrv_get_allocated_file_size(BlockDriverState *bs);
93
94 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
95 BlockDriverState *in_bs, Error **errp);
96
97 int coroutine_fn GRAPH_RDLOCK
98 bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
99
100 void coroutine_fn GRAPH_RDLOCK
101 bdrv_co_delete_file_noerr(BlockDriverState *bs);
102
103
104 /* async block I/O */
105 void bdrv_aio_cancel_async(BlockAIOCB *acb);
106
107 /* sg packet commands */
108 int coroutine_fn GRAPH_RDLOCK
109 bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
110
111 /* Ensure contents are flushed to disk. */
112 int coroutine_fn GRAPH_RDLOCK bdrv_co_flush(BlockDriverState *bs);
113
114 int coroutine_fn GRAPH_RDLOCK bdrv_co_pdiscard(BdrvChild *child, int64_t offset,
115 int64_t bytes);
116
117 /* Report zone information of zone block device. */
118 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_report(BlockDriverState *bs,
119 int64_t offset,
120 unsigned int *nr_zones,
121 BlockZoneDescriptor *zones);
122 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_mgmt(BlockDriverState *bs,
123 BlockZoneOp op,
124 int64_t offset, int64_t len);
125 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_append(BlockDriverState *bs,
126 int64_t *offset,
127 QEMUIOVector *qiov,
128 BdrvRequestFlags flags);
129
130 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
131
132 int coroutine_fn GRAPH_RDLOCK
133 bdrv_co_block_status(BlockDriverState *bs, int64_t offset, int64_t bytes,
134 int64_t *pnum, int64_t *map, BlockDriverState **file);
135 int co_wrapper_mixed_bdrv_rdlock
136 bdrv_block_status(BlockDriverState *bs, int64_t offset, int64_t bytes,
137 int64_t *pnum, int64_t *map, BlockDriverState **file);
138
139 int coroutine_fn GRAPH_RDLOCK
140 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base,
141 int64_t offset, int64_t bytes, int64_t *pnum,
142 int64_t *map, BlockDriverState **file);
143 int co_wrapper_mixed_bdrv_rdlock
144 bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
145 int64_t offset, int64_t bytes, int64_t *pnum,
146 int64_t *map, BlockDriverState **file);
147
148 int coroutine_fn GRAPH_RDLOCK
149 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
150 int64_t *pnum);
151 int co_wrapper_mixed_bdrv_rdlock
152 bdrv_is_allocated(BlockDriverState *bs, int64_t offset,
153 int64_t bytes, int64_t *pnum);
154
155 int coroutine_fn GRAPH_RDLOCK
156 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
157 bool include_base, int64_t offset, int64_t bytes,
158 int64_t *pnum);
159 int co_wrapper_mixed_bdrv_rdlock
160 bdrv_is_allocated_above(BlockDriverState *bs, BlockDriverState *base,
161 bool include_base, int64_t offset,
162 int64_t bytes, int64_t *pnum);
163
164 int coroutine_fn GRAPH_RDLOCK
165 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes);
166 int coroutine_fn GRAPH_RDLOCK
167 bdrv_co_is_all_zeroes(BlockDriverState *bs);
168
169 int GRAPH_RDLOCK
170 bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
171 Error **errp);
172
173 bool bdrv_is_read_only(BlockDriverState *bs);
174 bool bdrv_is_writable(BlockDriverState *bs);
175 bool bdrv_is_sg(BlockDriverState *bs);
176 int bdrv_get_flags(BlockDriverState *bs);
177
178 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs);
179 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs);
180
181 void coroutine_fn GRAPH_RDLOCK
182 bdrv_co_lock_medium(BlockDriverState *bs, bool locked);
183
184 void coroutine_fn GRAPH_RDLOCK
185 bdrv_co_eject(BlockDriverState *bs, bool eject_flag);
186
187 const char *bdrv_get_format_name(BlockDriverState *bs);
188
189 bool GRAPH_RDLOCK bdrv_supports_compressed_writes(BlockDriverState *bs);
190 const char *bdrv_get_node_name(const BlockDriverState *bs);
191
192 const char * GRAPH_RDLOCK
193 bdrv_get_device_name(const BlockDriverState *bs);
194
195 const char * GRAPH_RDLOCK
196 bdrv_get_device_or_node_name(const BlockDriverState *bs);
197
198 int coroutine_fn GRAPH_RDLOCK
199 bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
200
201 int co_wrapper_mixed_bdrv_rdlock
202 bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
203
204 ImageInfoSpecific * GRAPH_RDLOCK
205 bdrv_get_specific_info(BlockDriverState *bs, Error **errp);
206
207 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
208 void bdrv_round_to_subclusters(BlockDriverState *bs,
209 int64_t offset, int64_t bytes,
210 int64_t *cluster_offset,
211 int64_t *cluster_bytes);
212
213 void bdrv_get_backing_filename(BlockDriverState *bs,
214 char *filename, int filename_size);
215
216 int coroutine_fn GRAPH_RDLOCK
217 bdrv_co_change_backing_file(BlockDriverState *bs, const char *backing_file,
218 const char *backing_fmt, bool warn);
219
220 int co_wrapper_bdrv_rdlock
221 bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
222 const char *backing_fmt, bool warn);
223
224 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
225 int64_t pos, int size);
226
227 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
228 int64_t pos, int size);
229
230 /*
231 * Returns the alignment in bytes that is required so that no bounce buffer
232 * is required throughout the stack
233 */
234 size_t bdrv_min_mem_align(BlockDriverState *bs);
235 /* Returns optimal alignment in bytes for bounce buffer */
236 size_t bdrv_opt_mem_align(BlockDriverState *bs);
237 void *qemu_blockalign(BlockDriverState *bs, size_t size);
238 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
239 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
240 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
241
242 void bdrv_enable_copy_on_read(BlockDriverState *bs);
243 void bdrv_disable_copy_on_read(BlockDriverState *bs);
244
245 void coroutine_fn GRAPH_RDLOCK
246 bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event);
247
248 void co_wrapper_mixed_bdrv_rdlock
249 bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
250
251 #define BLKDBG_CO_EVENT(child, evt) \
252 do { \
253 if (child) { \
254 bdrv_co_debug_event(child->bs, evt); \
255 } \
256 } while (0)
257
258 #define BLKDBG_EVENT(child, evt) \
259 do { \
260 if (child) { \
261 bdrv_debug_event(child->bs, evt); \
262 } \
263 } while (0)
264
265 /**
266 * bdrv_get_aio_context:
267 *
268 * Returns: the currently bound #AioContext
269 */
270 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
271
272 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
273
274 /**
275 * Move the current coroutine to the AioContext of @bs and return the old
276 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
277 * will wait for the operation to proceed until the corresponding
278 * bdrv_co_leave().
279 *
280 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
281 * this will deadlock.
282 */
283 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
284
285 /**
286 * Ends a section started by bdrv_co_enter(). Move the current coroutine back
287 * to old_ctx and decrease bs->in_flight again.
288 */
289 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
290
291 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
292
293 bool coroutine_fn GRAPH_RDLOCK
294 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
295 uint32_t granularity, Error **errp);
296 bool co_wrapper_bdrv_rdlock
297 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
298 uint32_t granularity, Error **errp);
299
300 /**
301 *
302 * bdrv_co_copy_range:
303 *
304 * Do offloaded copy between two children. If the operation is not implemented
305 * by the driver, or if the backend storage doesn't support it, a negative
306 * error code will be returned.
307 *
308 * Note: block layer doesn't emulate or fallback to a bounce buffer approach
309 * because usually the caller shouldn't attempt offloaded copy any more (e.g.
310 * calling copy_file_range(2)) after the first error, thus it should fall back
311 * to a read+write path in the caller level.
312 *
313 * @src: Source child to copy data from
314 * @src_offset: offset in @src image to read data
315 * @dst: Destination child to copy data to
316 * @dst_offset: offset in @dst image to write data
317 * @bytes: number of bytes to copy
318 * @flags: request flags. Supported flags:
319 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
320 * write on @dst as if bdrv_co_pwrite_zeroes is
321 * called. Used to simplify caller code, or
322 * during BlockDriver.bdrv_co_copy_range_from()
323 * recursion.
324 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
325 * requests currently in flight.
326 *
327 * Returns: 0 if succeeded; negative error code if failed.
328 **/
329 int coroutine_fn GRAPH_RDLOCK
330 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
331 BdrvChild *dst, int64_t dst_offset,
332 int64_t bytes, BdrvRequestFlags read_flags,
333 BdrvRequestFlags write_flags);
334
335 /*
336 * "I/O or GS" API functions. These functions can run without
337 * the BQL, but only in one specific iothread/main loop.
338 *
339 * More specifically, these functions use BDRV_POLL_WHILE(bs), which requires
340 * the caller to be either in the main thread or directly in the home thread
341 * that runs the bs AioContext. Calling them from another thread in another
342 * AioContext would cause deadlocks.
343 *
344 * Therefore, these functions are not proper I/O, because they
345 * can't run in *any* iothreads, but only in a specific one.
346 *
347 * These functions can call any function from I/O, Common and this
348 * categories, but must be invoked only by other "I/O or GS" and GS APIs.
349 *
350 * All functions in this category must use the macro
351 * IO_OR_GS_CODE();
352 * to catch when they are accidentally called by the wrong API.
353 */
354
355 #define BDRV_POLL_WHILE(bs, cond) ({ \
356 BlockDriverState *bs_ = (bs); \
357 IO_OR_GS_CODE(); \
358 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \
359 cond); })
360
361 void bdrv_drain(BlockDriverState *bs);
362
363 int co_wrapper_mixed_bdrv_rdlock
364 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
365 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
366
367 int co_wrapper_mixed_bdrv_rdlock
368 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
369
370 /* Invalidate any cached metadata used by image formats */
371 int co_wrapper_mixed_bdrv_rdlock
372 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
373
374 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs);
375
376 int co_wrapper_mixed_bdrv_rdlock
377 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
378
379 int co_wrapper_mixed_bdrv_rdlock
380 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
381
382 int co_wrapper_mixed_bdrv_rdlock
383 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
384
385 /**
386 * bdrv_parent_drained_begin_single:
387 *
388 * Begin a quiesced section for the parent of @c.
389 */
390 void GRAPH_RDLOCK bdrv_parent_drained_begin_single(BdrvChild *c);
391
392 /**
393 * bdrv_parent_drained_poll_single:
394 *
395 * Returns true if there is any pending activity to cease before @c can be
396 * called quiesced, false otherwise.
397 */
398 bool GRAPH_RDLOCK bdrv_parent_drained_poll_single(BdrvChild *c);
399
400 /**
401 * bdrv_parent_drained_end_single:
402 *
403 * End a quiesced section for the parent of @c.
404 */
405 void GRAPH_RDLOCK bdrv_parent_drained_end_single(BdrvChild *c);
406
407 /**
408 * bdrv_drain_poll:
409 *
410 * Poll for pending requests in @bs and its parents (except for @ignore_parent).
411 *
412 * If @ignore_bds_parents is true, parents that are BlockDriverStates must
413 * ignore the drain request because they will be drained separately (used for
414 * drain_all).
415 *
416 * This is part of bdrv_drained_begin.
417 */
418 bool GRAPH_RDLOCK
419 bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent,
420 bool ignore_bds_parents);
421
422 /**
423 * bdrv_drained_begin:
424 *
425 * Begin a quiesced section for exclusive access to the BDS, by disabling
426 * external request sources including NBD server, block jobs, and device model.
427 *
428 * This function can only be invoked by the main loop or a coroutine
429 * (regardless of the AioContext where it is running).
430 * If the coroutine is running in an Iothread AioContext, this function will
431 * just schedule a BH to run in the main loop.
432 * However, it cannot be directly called by an Iothread.
433 *
434 * This function can be recursive.
435 */
436 void GRAPH_UNLOCKED bdrv_drained_begin(BlockDriverState *bs);
437
438 /**
439 * bdrv_do_drained_begin_quiesce:
440 *
441 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
442 * running requests to complete.
443 */
444 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent);
445
446 /**
447 * bdrv_drained_end:
448 *
449 * End a quiescent section started by bdrv_drained_begin().
450 *
451 * This function can only be invoked by the main loop or a coroutine
452 * (regardless of the AioContext where it is running).
453 * If the coroutine is running in an Iothread AioContext, this function will
454 * just schedule a BH to run in the main loop.
455 * However, it cannot be directly called by an Iothread.
456 */
457 void bdrv_drained_end(BlockDriverState *bs);
458
459 #endif /* BLOCK_IO_H */