xref: /openbmc/qemu/include/block/block-io.h (revision 1141159c)
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 "block/aio-wait.h"
28 #include "block/block-common.h"
29 #include "qemu/coroutine.h"
30 #include "qemu/iov.h"
31 
32 /*
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().
36  *
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.
39  *
40  * All functions in this category must use the macro
41  * IO_CODE();
42  * to catch when they are accidentally called by the wrong API.
43  */
44 
45 int co_wrapper_mixed_bdrv_rdlock
46 bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset, int64_t bytes,
47                    BdrvRequestFlags flags);
48 
49 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
50 
51 int co_wrapper_mixed_bdrv_rdlock
52 bdrv_pread(BdrvChild *child, int64_t offset, int64_t bytes, void *buf,
53            BdrvRequestFlags flags);
54 
55 int co_wrapper_mixed_bdrv_rdlock
56 bdrv_pwrite(BdrvChild *child, int64_t offset,int64_t bytes,
57             const void *buf, BdrvRequestFlags flags);
58 
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);
62 
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);
66 
67 /*
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.
72  */
73 int coroutine_fn GRAPH_RDLOCK
74 bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset, int64_t bytes,
75                       BdrvRequestFlags flags);
76 
77 int coroutine_fn GRAPH_RDLOCK
78 bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
79                  PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
80 
81 int64_t coroutine_fn GRAPH_RDLOCK bdrv_co_nb_sectors(BlockDriverState *bs);
82 int64_t coroutine_mixed_fn bdrv_nb_sectors(BlockDriverState *bs);
83 
84 int64_t coroutine_fn GRAPH_RDLOCK bdrv_co_getlength(BlockDriverState *bs);
85 int64_t co_wrapper_mixed_bdrv_rdlock bdrv_getlength(BlockDriverState *bs);
86 
87 int64_t coroutine_fn GRAPH_RDLOCK
88 bdrv_co_get_allocated_file_size(BlockDriverState *bs);
89 
90 int64_t co_wrapper_bdrv_rdlock
91 bdrv_get_allocated_file_size(BlockDriverState *bs);
92 
93 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
94                                BlockDriverState *in_bs, Error **errp);
95 
96 int coroutine_fn GRAPH_RDLOCK
97 bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
98 
99 void coroutine_fn GRAPH_RDLOCK
100 bdrv_co_delete_file_noerr(BlockDriverState *bs);
101 
102 
103 /* async block I/O */
104 void bdrv_aio_cancel(BlockAIOCB *acb);
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 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
132                       int64_t bytes, int64_t *pnum, int64_t *map,
133                       BlockDriverState **file);
134 
135 int coroutine_fn GRAPH_RDLOCK
136 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base,
137                            int64_t offset, int64_t bytes, int64_t *pnum,
138                            int64_t *map, BlockDriverState **file);
139 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
140                             int64_t offset, int64_t bytes, int64_t *pnum,
141                             int64_t *map, BlockDriverState **file);
142 
143 int coroutine_fn GRAPH_RDLOCK
144 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
145                      int64_t *pnum);
146 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
147                       int64_t *pnum);
148 
149 int coroutine_fn GRAPH_RDLOCK
150 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
151                            bool include_base, int64_t offset, int64_t bytes,
152                            int64_t *pnum);
153 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
154                             bool include_base, int64_t offset, int64_t bytes,
155                             int64_t *pnum);
156 
157 int coroutine_fn GRAPH_RDLOCK
158 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes);
159 
160 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
161                               Error **errp);
162 bool bdrv_is_read_only(BlockDriverState *bs);
163 bool bdrv_is_writable(BlockDriverState *bs);
164 bool bdrv_is_sg(BlockDriverState *bs);
165 int bdrv_get_flags(BlockDriverState *bs);
166 
167 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs);
168 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs);
169 
170 void coroutine_fn GRAPH_RDLOCK
171 bdrv_co_lock_medium(BlockDriverState *bs, bool locked);
172 
173 void coroutine_fn GRAPH_RDLOCK
174 bdrv_co_eject(BlockDriverState *bs, bool eject_flag);
175 
176 const char *bdrv_get_format_name(BlockDriverState *bs);
177 
178 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
179 const char *bdrv_get_node_name(const BlockDriverState *bs);
180 const char *bdrv_get_device_name(const BlockDriverState *bs);
181 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
182 
183 int coroutine_fn GRAPH_RDLOCK
184 bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
185 
186 int co_wrapper_mixed_bdrv_rdlock
187 bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
188 
189 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
190                                           Error **errp);
191 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
192 void bdrv_round_to_clusters(BlockDriverState *bs,
193                             int64_t offset, int64_t bytes,
194                             int64_t *cluster_offset,
195                             int64_t *cluster_bytes);
196 
197 void bdrv_get_backing_filename(BlockDriverState *bs,
198                                char *filename, int filename_size);
199 
200 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
201                       int64_t pos, int size);
202 
203 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
204                       int64_t pos, int size);
205 
206 /*
207  * Returns the alignment in bytes that is required so that no bounce buffer
208  * is required throughout the stack
209  */
210 size_t bdrv_min_mem_align(BlockDriverState *bs);
211 /* Returns optimal alignment in bytes for bounce buffer */
212 size_t bdrv_opt_mem_align(BlockDriverState *bs);
213 void *qemu_blockalign(BlockDriverState *bs, size_t size);
214 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
215 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
216 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
217 
218 void bdrv_enable_copy_on_read(BlockDriverState *bs);
219 void bdrv_disable_copy_on_read(BlockDriverState *bs);
220 
221 void coroutine_fn GRAPH_RDLOCK
222 bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event);
223 
224 void co_wrapper_mixed_bdrv_rdlock
225 bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
226 
227 #define BLKDBG_EVENT(child, evt) \
228     do { \
229         if (child) { \
230             bdrv_debug_event(child->bs, evt); \
231         } \
232     } while (0)
233 
234 /**
235  * bdrv_get_aio_context:
236  *
237  * Returns: the currently bound #AioContext
238  */
239 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
240 
241 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
242 
243 /**
244  * Move the current coroutine to the AioContext of @bs and return the old
245  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
246  * will wait for the operation to proceed until the corresponding
247  * bdrv_co_leave().
248  *
249  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
250  * this will deadlock.
251  */
252 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
253 
254 /**
255  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
256  * to old_ctx and decrease bs->in_flight again.
257  */
258 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
259 
260 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
261 
262 void coroutine_fn GRAPH_RDLOCK bdrv_co_io_plug(BlockDriverState *bs);
263 void coroutine_fn GRAPH_RDLOCK bdrv_co_io_unplug(BlockDriverState *bs);
264 
265 bool coroutine_fn GRAPH_RDLOCK
266 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
267                                    uint32_t granularity, Error **errp);
268 bool co_wrapper_bdrv_rdlock
269 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
270                                 uint32_t granularity, Error **errp);
271 
272 /**
273  *
274  * bdrv_co_copy_range:
275  *
276  * Do offloaded copy between two children. If the operation is not implemented
277  * by the driver, or if the backend storage doesn't support it, a negative
278  * error code will be returned.
279  *
280  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
281  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
282  * calling copy_file_range(2)) after the first error, thus it should fall back
283  * to a read+write path in the caller level.
284  *
285  * @src: Source child to copy data from
286  * @src_offset: offset in @src image to read data
287  * @dst: Destination child to copy data to
288  * @dst_offset: offset in @dst image to write data
289  * @bytes: number of bytes to copy
290  * @flags: request flags. Supported flags:
291  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
292  *                               write on @dst as if bdrv_co_pwrite_zeroes is
293  *                               called. Used to simplify caller code, or
294  *                               during BlockDriver.bdrv_co_copy_range_from()
295  *                               recursion.
296  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
297  *                                   requests currently in flight.
298  *
299  * Returns: 0 if succeeded; negative error code if failed.
300  **/
301 int coroutine_fn GRAPH_RDLOCK
302 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
303                    BdrvChild *dst, int64_t dst_offset,
304                    int64_t bytes, BdrvRequestFlags read_flags,
305                    BdrvRequestFlags write_flags);
306 
307 /*
308  * "I/O or GS" API functions. These functions can run without
309  * the BQL, but only in one specific iothread/main loop.
310  *
311  * More specifically, these functions use BDRV_POLL_WHILE(bs), which
312  * requires the caller to be either in the main thread and hold
313  * the BlockdriverState (bs) AioContext lock, or directly in the
314  * home thread that runs the bs AioContext. Calling them from
315  * another thread in another AioContext would cause deadlocks.
316  *
317  * Therefore, these functions are not proper I/O, because they
318  * can't run in *any* iothreads, but only in a specific one.
319  *
320  * These functions can call any function from I/O, Common and this
321  * categories, but must be invoked only by other "I/O or GS" and GS APIs.
322  *
323  * All functions in this category must use the macro
324  * IO_OR_GS_CODE();
325  * to catch when they are accidentally called by the wrong API.
326  */
327 
328 #define BDRV_POLL_WHILE(bs, cond) ({                       \
329     BlockDriverState *bs_ = (bs);                          \
330     IO_OR_GS_CODE();                                       \
331     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
332                    cond); })
333 
334 void bdrv_drain(BlockDriverState *bs);
335 
336 int co_wrapper_mixed_bdrv_rdlock
337 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
338               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
339 
340 int co_wrapper_mixed_bdrv_rdlock
341 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
342 
343 /* Invalidate any cached metadata used by image formats */
344 int co_wrapper_mixed_bdrv_rdlock
345 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
346 
347 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs);
348 
349 int co_wrapper_mixed_bdrv_rdlock
350 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
351 
352 int co_wrapper_mixed_bdrv_rdlock
353 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
354 
355 int co_wrapper_mixed_bdrv_rdlock
356 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
357 
358 /**
359  * bdrv_parent_drained_begin_single:
360  *
361  * Begin a quiesced section for the parent of @c.
362  */
363 void bdrv_parent_drained_begin_single(BdrvChild *c);
364 
365 /**
366  * bdrv_parent_drained_poll_single:
367  *
368  * Returns true if there is any pending activity to cease before @c can be
369  * called quiesced, false otherwise.
370  */
371 bool bdrv_parent_drained_poll_single(BdrvChild *c);
372 
373 /**
374  * bdrv_parent_drained_end_single:
375  *
376  * End a quiesced section for the parent of @c.
377  */
378 void bdrv_parent_drained_end_single(BdrvChild *c);
379 
380 /**
381  * bdrv_drain_poll:
382  *
383  * Poll for pending requests in @bs and its parents (except for @ignore_parent).
384  *
385  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
386  * ignore the drain request because they will be drained separately (used for
387  * drain_all).
388  *
389  * This is part of bdrv_drained_begin.
390  */
391 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent,
392                      bool ignore_bds_parents);
393 
394 /**
395  * bdrv_drained_begin:
396  *
397  * Begin a quiesced section for exclusive access to the BDS, by disabling
398  * external request sources including NBD server, block jobs, and device model.
399  *
400  * This function can be recursive.
401  */
402 void bdrv_drained_begin(BlockDriverState *bs);
403 
404 /**
405  * bdrv_do_drained_begin_quiesce:
406  *
407  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
408  * running requests to complete.
409  */
410 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent);
411 
412 /**
413  * bdrv_drained_end:
414  *
415  * End a quiescent section started by bdrv_drained_begin().
416  */
417 void bdrv_drained_end(BlockDriverState *bs);
418 
419 #endif /* BLOCK_IO_H */
420