2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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
27 #include "block/aio-wait.h"
28 #include "block/block-common.h"
29 #include "qemu/coroutine.h"
33 * I/O API functions. These functions are thread-safe, and therefore
34 * can run in any thread as long as the thread has called
35 * aio_context_acquire/release().
37 * These functions can only call functions from I/O and Common categories,
38 * but can be invoked by GS, "I/O or GS" and I/O APIs.
40 * All functions in this category must use the macro
42 * to catch when they are accidentally called by the wrong API.
45 int co_wrapper_mixed_bdrv_rdlock
46 bdrv_pwrite_zeroes(BdrvChild
*child
, int64_t offset
, int64_t bytes
,
47 BdrvRequestFlags flags
);
49 int bdrv_make_zero(BdrvChild
*child
, BdrvRequestFlags flags
);
51 int co_wrapper_mixed_bdrv_rdlock
52 bdrv_pread(BdrvChild
*child
, int64_t offset
, int64_t bytes
, void *buf
,
53 BdrvRequestFlags flags
);
55 int co_wrapper_mixed_bdrv_rdlock
56 bdrv_pwrite(BdrvChild
*child
, int64_t offset
,int64_t bytes
,
57 const void *buf
, BdrvRequestFlags flags
);
59 int co_wrapper_mixed_bdrv_rdlock
60 bdrv_pwrite_sync(BdrvChild
*child
, int64_t offset
, int64_t bytes
,
61 const void *buf
, BdrvRequestFlags flags
);
63 int coroutine_fn GRAPH_RDLOCK
64 bdrv_co_pwrite_sync(BdrvChild
*child
, int64_t offset
, int64_t bytes
,
65 const void *buf
, BdrvRequestFlags flags
);
68 * Efficiently zero a region of the disk image. Note that this is a regular
69 * I/O request like read or write and should have a reasonable size. This
70 * function is not suitable for zeroing the entire image in a single request
71 * because it may allocate memory for the entire region.
73 int coroutine_fn GRAPH_RDLOCK
74 bdrv_co_pwrite_zeroes(BdrvChild
*child
, int64_t offset
, int64_t bytes
,
75 BdrvRequestFlags flags
);
77 int coroutine_fn GRAPH_RDLOCK
78 bdrv_co_truncate(BdrvChild
*child
, int64_t offset
, bool exact
,
79 PreallocMode prealloc
, BdrvRequestFlags flags
, Error
**errp
);
81 int64_t coroutine_fn GRAPH_RDLOCK
bdrv_co_nb_sectors(BlockDriverState
*bs
);
82 int64_t coroutine_mixed_fn
bdrv_nb_sectors(BlockDriverState
*bs
);
84 int64_t coroutine_fn GRAPH_RDLOCK
bdrv_co_getlength(BlockDriverState
*bs
);
85 int64_t co_wrapper_mixed_bdrv_rdlock
bdrv_getlength(BlockDriverState
*bs
);
87 int64_t coroutine_fn GRAPH_RDLOCK
88 bdrv_co_get_allocated_file_size(BlockDriverState
*bs
);
90 int64_t co_wrapper_bdrv_rdlock
91 bdrv_get_allocated_file_size(BlockDriverState
*bs
);
93 BlockMeasureInfo
*bdrv_measure(BlockDriver
*drv
, QemuOpts
*opts
,
94 BlockDriverState
*in_bs
, Error
**errp
);
96 int coroutine_fn GRAPH_RDLOCK
97 bdrv_co_delete_file(BlockDriverState
*bs
, Error
**errp
);
99 void coroutine_fn GRAPH_RDLOCK
100 bdrv_co_delete_file_noerr(BlockDriverState
*bs
);
103 /* async block I/O */
104 void bdrv_aio_cancel_async(BlockAIOCB
*acb
);
106 /* sg packet commands */
107 int coroutine_fn GRAPH_RDLOCK
108 bdrv_co_ioctl(BlockDriverState
*bs
, int req
, void *buf
);
110 /* Ensure contents are flushed to disk. */
111 int coroutine_fn GRAPH_RDLOCK
bdrv_co_flush(BlockDriverState
*bs
);
113 int coroutine_fn GRAPH_RDLOCK
bdrv_co_pdiscard(BdrvChild
*child
, int64_t offset
,
116 /* Report zone information of zone block device. */
117 int coroutine_fn GRAPH_RDLOCK
bdrv_co_zone_report(BlockDriverState
*bs
,
119 unsigned int *nr_zones
,
120 BlockZoneDescriptor
*zones
);
121 int coroutine_fn GRAPH_RDLOCK
bdrv_co_zone_mgmt(BlockDriverState
*bs
,
123 int64_t offset
, int64_t len
);
124 int coroutine_fn GRAPH_RDLOCK
bdrv_co_zone_append(BlockDriverState
*bs
,
127 BdrvRequestFlags flags
);
129 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState
*bs
);
131 int coroutine_fn GRAPH_RDLOCK
132 bdrv_co_block_status(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
,
133 int64_t *pnum
, int64_t *map
, BlockDriverState
**file
);
134 int co_wrapper_mixed_bdrv_rdlock
135 bdrv_block_status(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
,
136 int64_t *pnum
, int64_t *map
, BlockDriverState
**file
);
138 int coroutine_fn GRAPH_RDLOCK
139 bdrv_co_block_status_above(BlockDriverState
*bs
, BlockDriverState
*base
,
140 int64_t offset
, int64_t bytes
, int64_t *pnum
,
141 int64_t *map
, BlockDriverState
**file
);
142 int co_wrapper_mixed_bdrv_rdlock
143 bdrv_block_status_above(BlockDriverState
*bs
, BlockDriverState
*base
,
144 int64_t offset
, int64_t bytes
, int64_t *pnum
,
145 int64_t *map
, BlockDriverState
**file
);
147 int coroutine_fn GRAPH_RDLOCK
148 bdrv_co_is_allocated(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
,
150 int co_wrapper_mixed_bdrv_rdlock
151 bdrv_is_allocated(BlockDriverState
*bs
, int64_t offset
,
152 int64_t bytes
, int64_t *pnum
);
154 int coroutine_fn GRAPH_RDLOCK
155 bdrv_co_is_allocated_above(BlockDriverState
*top
, BlockDriverState
*base
,
156 bool include_base
, int64_t offset
, int64_t bytes
,
158 int co_wrapper_mixed_bdrv_rdlock
159 bdrv_is_allocated_above(BlockDriverState
*bs
, BlockDriverState
*base
,
160 bool include_base
, int64_t offset
,
161 int64_t bytes
, int64_t *pnum
);
163 int coroutine_fn GRAPH_RDLOCK
164 bdrv_co_is_zero_fast(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
);
167 bdrv_apply_auto_read_only(BlockDriverState
*bs
, const char *errmsg
,
170 bool bdrv_is_read_only(BlockDriverState
*bs
);
171 bool bdrv_is_writable(BlockDriverState
*bs
);
172 bool bdrv_is_sg(BlockDriverState
*bs
);
173 int bdrv_get_flags(BlockDriverState
*bs
);
175 bool coroutine_fn GRAPH_RDLOCK
bdrv_co_is_inserted(BlockDriverState
*bs
);
176 bool co_wrapper_bdrv_rdlock
bdrv_is_inserted(BlockDriverState
*bs
);
178 void coroutine_fn GRAPH_RDLOCK
179 bdrv_co_lock_medium(BlockDriverState
*bs
, bool locked
);
181 void coroutine_fn GRAPH_RDLOCK
182 bdrv_co_eject(BlockDriverState
*bs
, bool eject_flag
);
184 const char *bdrv_get_format_name(BlockDriverState
*bs
);
186 bool GRAPH_RDLOCK
bdrv_supports_compressed_writes(BlockDriverState
*bs
);
187 const char *bdrv_get_node_name(const BlockDriverState
*bs
);
189 const char * GRAPH_RDLOCK
190 bdrv_get_device_name(const BlockDriverState
*bs
);
192 const char * GRAPH_RDLOCK
193 bdrv_get_device_or_node_name(const BlockDriverState
*bs
);
195 int coroutine_fn GRAPH_RDLOCK
196 bdrv_co_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
198 int co_wrapper_mixed_bdrv_rdlock
199 bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
201 ImageInfoSpecific
* GRAPH_RDLOCK
202 bdrv_get_specific_info(BlockDriverState
*bs
, Error
**errp
);
204 BlockStatsSpecific
*bdrv_get_specific_stats(BlockDriverState
*bs
);
205 void bdrv_round_to_subclusters(BlockDriverState
*bs
,
206 int64_t offset
, int64_t bytes
,
207 int64_t *cluster_offset
,
208 int64_t *cluster_bytes
);
210 void bdrv_get_backing_filename(BlockDriverState
*bs
,
211 char *filename
, int filename_size
);
213 int coroutine_fn GRAPH_RDLOCK
214 bdrv_co_change_backing_file(BlockDriverState
*bs
, const char *backing_file
,
215 const char *backing_fmt
, bool warn
);
217 int co_wrapper_bdrv_rdlock
218 bdrv_change_backing_file(BlockDriverState
*bs
, const char *backing_file
,
219 const char *backing_fmt
, bool warn
);
221 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
222 int64_t pos
, int size
);
224 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
225 int64_t pos
, int size
);
228 * Returns the alignment in bytes that is required so that no bounce buffer
229 * is required throughout the stack
231 size_t bdrv_min_mem_align(BlockDriverState
*bs
);
232 /* Returns optimal alignment in bytes for bounce buffer */
233 size_t bdrv_opt_mem_align(BlockDriverState
*bs
);
234 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
);
235 void *qemu_blockalign0(BlockDriverState
*bs
, size_t size
);
236 void *qemu_try_blockalign(BlockDriverState
*bs
, size_t size
);
237 void *qemu_try_blockalign0(BlockDriverState
*bs
, size_t size
);
239 void bdrv_enable_copy_on_read(BlockDriverState
*bs
);
240 void bdrv_disable_copy_on_read(BlockDriverState
*bs
);
242 void coroutine_fn GRAPH_RDLOCK
243 bdrv_co_debug_event(BlockDriverState
*bs
, BlkdebugEvent event
);
245 void co_wrapper_mixed_bdrv_rdlock
246 bdrv_debug_event(BlockDriverState
*bs
, BlkdebugEvent event
);
248 #define BLKDBG_CO_EVENT(child, evt) \
251 bdrv_co_debug_event(child->bs, evt); \
255 #define BLKDBG_EVENT(child, evt) \
258 bdrv_debug_event(child->bs, evt); \
263 * bdrv_get_aio_context:
265 * Returns: the currently bound #AioContext
267 AioContext
*bdrv_get_aio_context(BlockDriverState
*bs
);
269 AioContext
*bdrv_child_get_parent_aio_context(BdrvChild
*c
);
272 * Move the current coroutine to the AioContext of @bs and return the old
273 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
274 * will wait for the operation to proceed until the corresponding
277 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
278 * this will deadlock.
280 AioContext
*coroutine_fn
bdrv_co_enter(BlockDriverState
*bs
);
283 * Ends a section started by bdrv_co_enter(). Move the current coroutine back
284 * to old_ctx and decrease bs->in_flight again.
286 void coroutine_fn
bdrv_co_leave(BlockDriverState
*bs
, AioContext
*old_ctx
);
288 AioContext
*child_of_bds_get_parent_aio_context(BdrvChild
*c
);
290 bool coroutine_fn GRAPH_RDLOCK
291 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState
*bs
, const char *name
,
292 uint32_t granularity
, Error
**errp
);
293 bool co_wrapper_bdrv_rdlock
294 bdrv_can_store_new_dirty_bitmap(BlockDriverState
*bs
, const char *name
,
295 uint32_t granularity
, Error
**errp
);
299 * bdrv_co_copy_range:
301 * Do offloaded copy between two children. If the operation is not implemented
302 * by the driver, or if the backend storage doesn't support it, a negative
303 * error code will be returned.
305 * Note: block layer doesn't emulate or fallback to a bounce buffer approach
306 * because usually the caller shouldn't attempt offloaded copy any more (e.g.
307 * calling copy_file_range(2)) after the first error, thus it should fall back
308 * to a read+write path in the caller level.
310 * @src: Source child to copy data from
311 * @src_offset: offset in @src image to read data
312 * @dst: Destination child to copy data to
313 * @dst_offset: offset in @dst image to write data
314 * @bytes: number of bytes to copy
315 * @flags: request flags. Supported flags:
316 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
317 * write on @dst as if bdrv_co_pwrite_zeroes is
318 * called. Used to simplify caller code, or
319 * during BlockDriver.bdrv_co_copy_range_from()
321 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
322 * requests currently in flight.
324 * Returns: 0 if succeeded; negative error code if failed.
326 int coroutine_fn GRAPH_RDLOCK
327 bdrv_co_copy_range(BdrvChild
*src
, int64_t src_offset
,
328 BdrvChild
*dst
, int64_t dst_offset
,
329 int64_t bytes
, BdrvRequestFlags read_flags
,
330 BdrvRequestFlags write_flags
);
333 * "I/O or GS" API functions. These functions can run without
334 * the BQL, but only in one specific iothread/main loop.
336 * More specifically, these functions use BDRV_POLL_WHILE(bs), which
337 * requires the caller to be either in the main thread and hold
338 * the BlockdriverState (bs) AioContext lock, or directly in the
339 * home thread that runs the bs AioContext. Calling them from
340 * another thread in another AioContext would cause deadlocks.
342 * Therefore, these functions are not proper I/O, because they
343 * can't run in *any* iothreads, but only in a specific one.
345 * These functions can call any function from I/O, Common and this
346 * categories, but must be invoked only by other "I/O or GS" and GS APIs.
348 * All functions in this category must use the macro
350 * to catch when they are accidentally called by the wrong API.
353 #define BDRV_POLL_WHILE(bs, cond) ({ \
354 BlockDriverState *bs_ = (bs); \
356 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \
359 void bdrv_drain(BlockDriverState
*bs
);
361 int co_wrapper_mixed_bdrv_rdlock
362 bdrv_truncate(BdrvChild
*child
, int64_t offset
, bool exact
,
363 PreallocMode prealloc
, BdrvRequestFlags flags
, Error
**errp
);
365 int co_wrapper_mixed_bdrv_rdlock
366 bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
, BdrvCheckMode fix
);
368 /* Invalidate any cached metadata used by image formats */
369 int co_wrapper_mixed_bdrv_rdlock
370 bdrv_invalidate_cache(BlockDriverState
*bs
, Error
**errp
);
372 int co_wrapper_mixed_bdrv_rdlock
bdrv_flush(BlockDriverState
*bs
);
374 int co_wrapper_mixed_bdrv_rdlock
375 bdrv_pdiscard(BdrvChild
*child
, int64_t offset
, int64_t bytes
);
377 int co_wrapper_mixed_bdrv_rdlock
378 bdrv_readv_vmstate(BlockDriverState
*bs
, QEMUIOVector
*qiov
, int64_t pos
);
380 int co_wrapper_mixed_bdrv_rdlock
381 bdrv_writev_vmstate(BlockDriverState
*bs
, QEMUIOVector
*qiov
, int64_t pos
);
384 * bdrv_parent_drained_begin_single:
386 * Begin a quiesced section for the parent of @c.
388 void GRAPH_RDLOCK
bdrv_parent_drained_begin_single(BdrvChild
*c
);
391 * bdrv_parent_drained_poll_single:
393 * Returns true if there is any pending activity to cease before @c can be
394 * called quiesced, false otherwise.
396 bool GRAPH_RDLOCK
bdrv_parent_drained_poll_single(BdrvChild
*c
);
399 * bdrv_parent_drained_end_single:
401 * End a quiesced section for the parent of @c.
403 void GRAPH_RDLOCK
bdrv_parent_drained_end_single(BdrvChild
*c
);
408 * Poll for pending requests in @bs and its parents (except for @ignore_parent).
410 * If @ignore_bds_parents is true, parents that are BlockDriverStates must
411 * ignore the drain request because they will be drained separately (used for
414 * This is part of bdrv_drained_begin.
417 bdrv_drain_poll(BlockDriverState
*bs
, BdrvChild
*ignore_parent
,
418 bool ignore_bds_parents
);
421 * bdrv_drained_begin:
423 * Begin a quiesced section for exclusive access to the BDS, by disabling
424 * external request sources including NBD server, block jobs, and device model.
426 * This function can only be invoked by the main loop or a coroutine
427 * (regardless of the AioContext where it is running).
428 * If the coroutine is running in an Iothread AioContext, this function will
429 * just schedule a BH to run in the main loop.
430 * However, it cannot be directly called by an Iothread.
432 * This function can be recursive.
434 void bdrv_drained_begin(BlockDriverState
*bs
);
437 * bdrv_do_drained_begin_quiesce:
439 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
440 * running requests to complete.
442 void bdrv_do_drained_begin_quiesce(BlockDriverState
*bs
, BdrvChild
*parent
);
447 * End a quiescent section started by bdrv_drained_begin().
449 * This function can only be invoked by the main loop or a coroutine
450 * (regardless of the AioContext where it is running).
451 * If the coroutine is running in an Iothread AioContext, this function will
452 * just schedule a BH to run in the main loop.
453 * However, it cannot be directly called by an Iothread.
455 void bdrv_drained_end(BlockDriverState
*bs
);
457 #endif /* BLOCK_IO_H */