xref: /openbmc/qemu/include/block/block-io.h (revision 5a3d2c35)
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_async(BlockAIOCB *acb);
105 
106 /* sg packet commands */
107 int coroutine_fn GRAPH_RDLOCK
108 bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
109 
110 /* Ensure contents are flushed to disk.  */
111 int coroutine_fn GRAPH_RDLOCK bdrv_co_flush(BlockDriverState *bs);
112 
113 int coroutine_fn GRAPH_RDLOCK bdrv_co_pdiscard(BdrvChild *child, int64_t offset,
114                                                int64_t bytes);
115 
116 /* Report zone information of zone block device. */
117 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_report(BlockDriverState *bs,
118                                                   int64_t offset,
119                                                   unsigned int *nr_zones,
120                                                   BlockZoneDescriptor *zones);
121 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_mgmt(BlockDriverState *bs,
122                                                 BlockZoneOp op,
123                                                 int64_t offset, int64_t len);
124 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_append(BlockDriverState *bs,
125                                                   int64_t *offset,
126                                                   QEMUIOVector *qiov,
127                                                   BdrvRequestFlags flags);
128 
129 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
130 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
131                       int64_t bytes, int64_t *pnum, int64_t *map,
132                       BlockDriverState **file);
133 
134 int coroutine_fn GRAPH_RDLOCK
135 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base,
136                            int64_t offset, int64_t bytes, int64_t *pnum,
137                            int64_t *map, BlockDriverState **file);
138 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
139                             int64_t offset, int64_t bytes, int64_t *pnum,
140                             int64_t *map, BlockDriverState **file);
141 
142 int coroutine_fn GRAPH_RDLOCK
143 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
144                      int64_t *pnum);
145 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
146                       int64_t *pnum);
147 
148 int coroutine_fn GRAPH_RDLOCK
149 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
150                            bool include_base, int64_t offset, int64_t bytes,
151                            int64_t *pnum);
152 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
153                             bool include_base, int64_t offset, int64_t bytes,
154                             int64_t *pnum);
155 
156 int coroutine_fn GRAPH_RDLOCK
157 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes);
158 
159 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
160                               Error **errp);
161 bool bdrv_is_read_only(BlockDriverState *bs);
162 bool bdrv_is_writable(BlockDriverState *bs);
163 bool bdrv_is_sg(BlockDriverState *bs);
164 int bdrv_get_flags(BlockDriverState *bs);
165 
166 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs);
167 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs);
168 
169 void coroutine_fn GRAPH_RDLOCK
170 bdrv_co_lock_medium(BlockDriverState *bs, bool locked);
171 
172 void coroutine_fn GRAPH_RDLOCK
173 bdrv_co_eject(BlockDriverState *bs, bool eject_flag);
174 
175 const char *bdrv_get_format_name(BlockDriverState *bs);
176 
177 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
178 const char *bdrv_get_node_name(const BlockDriverState *bs);
179 const char *bdrv_get_device_name(const BlockDriverState *bs);
180 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
181 
182 int coroutine_fn GRAPH_RDLOCK
183 bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
184 
185 int co_wrapper_mixed_bdrv_rdlock
186 bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
187 
188 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
189                                           Error **errp);
190 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
191 void bdrv_round_to_subclusters(BlockDriverState *bs,
192                                int64_t offset, int64_t bytes,
193                                int64_t *cluster_offset,
194                                int64_t *cluster_bytes);
195 
196 void bdrv_get_backing_filename(BlockDriverState *bs,
197                                char *filename, int filename_size);
198 
199 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
200                       int64_t pos, int size);
201 
202 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
203                       int64_t pos, int size);
204 
205 /*
206  * Returns the alignment in bytes that is required so that no bounce buffer
207  * is required throughout the stack
208  */
209 size_t bdrv_min_mem_align(BlockDriverState *bs);
210 /* Returns optimal alignment in bytes for bounce buffer */
211 size_t bdrv_opt_mem_align(BlockDriverState *bs);
212 void *qemu_blockalign(BlockDriverState *bs, size_t size);
213 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
214 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
215 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
216 
217 void bdrv_enable_copy_on_read(BlockDriverState *bs);
218 void bdrv_disable_copy_on_read(BlockDriverState *bs);
219 
220 void coroutine_fn GRAPH_RDLOCK
221 bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event);
222 
223 void co_wrapper_mixed_bdrv_rdlock
224 bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
225 
226 #define BLKDBG_CO_EVENT(child, evt) \
227     do { \
228         if (child) { \
229             bdrv_co_debug_event(child->bs, evt); \
230         } \
231     } while (0)
232 
233 #define BLKDBG_EVENT(child, evt) \
234     do { \
235         if (child) { \
236             bdrv_debug_event(child->bs, evt); \
237         } \
238     } while (0)
239 
240 /**
241  * bdrv_get_aio_context:
242  *
243  * Returns: the currently bound #AioContext
244  */
245 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
246 
247 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
248 
249 /**
250  * Move the current coroutine to the AioContext of @bs and return the old
251  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
252  * will wait for the operation to proceed until the corresponding
253  * bdrv_co_leave().
254  *
255  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
256  * this will deadlock.
257  */
258 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
259 
260 /**
261  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
262  * to old_ctx and decrease bs->in_flight again.
263  */
264 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
265 
266 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
267 
268 bool coroutine_fn GRAPH_RDLOCK
269 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
270                                    uint32_t granularity, Error **errp);
271 bool co_wrapper_bdrv_rdlock
272 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
273                                 uint32_t granularity, Error **errp);
274 
275 /**
276  *
277  * bdrv_co_copy_range:
278  *
279  * Do offloaded copy between two children. If the operation is not implemented
280  * by the driver, or if the backend storage doesn't support it, a negative
281  * error code will be returned.
282  *
283  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
284  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
285  * calling copy_file_range(2)) after the first error, thus it should fall back
286  * to a read+write path in the caller level.
287  *
288  * @src: Source child to copy data from
289  * @src_offset: offset in @src image to read data
290  * @dst: Destination child to copy data to
291  * @dst_offset: offset in @dst image to write data
292  * @bytes: number of bytes to copy
293  * @flags: request flags. Supported flags:
294  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
295  *                               write on @dst as if bdrv_co_pwrite_zeroes is
296  *                               called. Used to simplify caller code, or
297  *                               during BlockDriver.bdrv_co_copy_range_from()
298  *                               recursion.
299  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
300  *                                   requests currently in flight.
301  *
302  * Returns: 0 if succeeded; negative error code if failed.
303  **/
304 int coroutine_fn GRAPH_RDLOCK
305 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
306                    BdrvChild *dst, int64_t dst_offset,
307                    int64_t bytes, BdrvRequestFlags read_flags,
308                    BdrvRequestFlags write_flags);
309 
310 /*
311  * "I/O or GS" API functions. These functions can run without
312  * the BQL, but only in one specific iothread/main loop.
313  *
314  * More specifically, these functions use BDRV_POLL_WHILE(bs), which
315  * requires the caller to be either in the main thread and hold
316  * the BlockdriverState (bs) AioContext lock, or directly in the
317  * home thread that runs the bs AioContext. Calling them from
318  * another thread in another AioContext would cause deadlocks.
319  *
320  * Therefore, these functions are not proper I/O, because they
321  * can't run in *any* iothreads, but only in a specific one.
322  *
323  * These functions can call any function from I/O, Common and this
324  * categories, but must be invoked only by other "I/O or GS" and GS APIs.
325  *
326  * All functions in this category must use the macro
327  * IO_OR_GS_CODE();
328  * to catch when they are accidentally called by the wrong API.
329  */
330 
331 #define BDRV_POLL_WHILE(bs, cond) ({                       \
332     BlockDriverState *bs_ = (bs);                          \
333     IO_OR_GS_CODE();                                       \
334     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
335                    cond); })
336 
337 void bdrv_drain(BlockDriverState *bs);
338 
339 int co_wrapper_mixed_bdrv_rdlock
340 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
341               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
342 
343 int co_wrapper_mixed_bdrv_rdlock
344 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
345 
346 /* Invalidate any cached metadata used by image formats */
347 int co_wrapper_mixed_bdrv_rdlock
348 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
349 
350 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs);
351 
352 int co_wrapper_mixed_bdrv_rdlock
353 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
354 
355 int co_wrapper_mixed_bdrv_rdlock
356 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
357 
358 int co_wrapper_mixed_bdrv_rdlock
359 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
360 
361 /**
362  * bdrv_parent_drained_begin_single:
363  *
364  * Begin a quiesced section for the parent of @c.
365  */
366 void bdrv_parent_drained_begin_single(BdrvChild *c);
367 
368 /**
369  * bdrv_parent_drained_poll_single:
370  *
371  * Returns true if there is any pending activity to cease before @c can be
372  * called quiesced, false otherwise.
373  */
374 bool bdrv_parent_drained_poll_single(BdrvChild *c);
375 
376 /**
377  * bdrv_parent_drained_end_single:
378  *
379  * End a quiesced section for the parent of @c.
380  */
381 void bdrv_parent_drained_end_single(BdrvChild *c);
382 
383 /**
384  * bdrv_drain_poll:
385  *
386  * Poll for pending requests in @bs and its parents (except for @ignore_parent).
387  *
388  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
389  * ignore the drain request because they will be drained separately (used for
390  * drain_all).
391  *
392  * This is part of bdrv_drained_begin.
393  */
394 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent,
395                      bool ignore_bds_parents);
396 
397 /**
398  * bdrv_drained_begin:
399  *
400  * Begin a quiesced section for exclusive access to the BDS, by disabling
401  * external request sources including NBD server, block jobs, and device model.
402  *
403  * This function can be recursive.
404  */
405 void bdrv_drained_begin(BlockDriverState *bs);
406 
407 /**
408  * bdrv_do_drained_begin_quiesce:
409  *
410  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
411  * running requests to complete.
412  */
413 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent);
414 
415 /**
416  * bdrv_drained_end:
417  *
418  * End a quiescent section started by bdrv_drained_begin().
419  */
420 void bdrv_drained_end(BlockDriverState *bs);
421 
422 #endif /* BLOCK_IO_H */
423