xref: /openbmc/qemu/include/block/block-io.h (revision 6eece7f5)
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 co_wrapper_mixed_bdrv_rdlock 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 bdrv_co_get_allocated_file_size(BlockDriverState *bs);
88 int64_t co_wrapper bdrv_get_allocated_file_size(BlockDriverState *bs);
89 
90 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
91                                BlockDriverState *in_bs, Error **errp);
92 
93 void coroutine_fn GRAPH_RDLOCK
94 bdrv_co_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
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 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
118 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
119                       int64_t bytes, int64_t *pnum, int64_t *map,
120                       BlockDriverState **file);
121 
122 int coroutine_fn GRAPH_RDLOCK
123 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base,
124                            int64_t offset, int64_t bytes, int64_t *pnum,
125                            int64_t *map, BlockDriverState **file);
126 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
127                             int64_t offset, int64_t bytes, int64_t *pnum,
128                             int64_t *map, BlockDriverState **file);
129 
130 int coroutine_fn GRAPH_RDLOCK
131 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
132                      int64_t *pnum);
133 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
134                       int64_t *pnum);
135 
136 int coroutine_fn GRAPH_RDLOCK
137 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
138                            bool include_base, int64_t offset, int64_t bytes,
139                            int64_t *pnum);
140 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
141                             bool include_base, int64_t offset, int64_t bytes,
142                             int64_t *pnum);
143 
144 int coroutine_fn GRAPH_RDLOCK
145 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes);
146 
147 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
148                               Error **errp);
149 bool bdrv_is_read_only(BlockDriverState *bs);
150 bool bdrv_is_writable(BlockDriverState *bs);
151 bool bdrv_is_sg(BlockDriverState *bs);
152 int bdrv_get_flags(BlockDriverState *bs);
153 
154 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs);
155 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs);
156 
157 void coroutine_fn GRAPH_RDLOCK
158 bdrv_co_lock_medium(BlockDriverState *bs, bool locked);
159 
160 void coroutine_fn GRAPH_RDLOCK
161 bdrv_co_eject(BlockDriverState *bs, bool eject_flag);
162 
163 const char *bdrv_get_format_name(BlockDriverState *bs);
164 
165 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
166 const char *bdrv_get_node_name(const BlockDriverState *bs);
167 const char *bdrv_get_device_name(const BlockDriverState *bs);
168 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
169 
170 int coroutine_fn bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
171 int co_wrapper_mixed bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
172 
173 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
174                                           Error **errp);
175 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
176 void bdrv_round_to_clusters(BlockDriverState *bs,
177                             int64_t offset, int64_t bytes,
178                             int64_t *cluster_offset,
179                             int64_t *cluster_bytes);
180 
181 void bdrv_get_backing_filename(BlockDriverState *bs,
182                                char *filename, int filename_size);
183 
184 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
185                       int64_t pos, int size);
186 
187 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
188                       int64_t pos, int size);
189 
190 /*
191  * Returns the alignment in bytes that is required so that no bounce buffer
192  * is required throughout the stack
193  */
194 size_t bdrv_min_mem_align(BlockDriverState *bs);
195 /* Returns optimal alignment in bytes for bounce buffer */
196 size_t bdrv_opt_mem_align(BlockDriverState *bs);
197 void *qemu_blockalign(BlockDriverState *bs, size_t size);
198 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
199 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
200 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
201 
202 void bdrv_enable_copy_on_read(BlockDriverState *bs);
203 void bdrv_disable_copy_on_read(BlockDriverState *bs);
204 
205 void coroutine_fn bdrv_co_debug_event(BlockDriverState *bs,
206                                       BlkdebugEvent event);
207 void co_wrapper_mixed bdrv_debug_event(BlockDriverState *bs,
208                                        BlkdebugEvent event);
209 
210 #define BLKDBG_EVENT(child, evt) \
211     do { \
212         if (child) { \
213             bdrv_debug_event(child->bs, evt); \
214         } \
215     } while (0)
216 
217 /**
218  * bdrv_get_aio_context:
219  *
220  * Returns: the currently bound #AioContext
221  */
222 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
223 
224 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
225 
226 /**
227  * Move the current coroutine to the AioContext of @bs and return the old
228  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
229  * will wait for the operation to proceed until the corresponding
230  * bdrv_co_leave().
231  *
232  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
233  * this will deadlock.
234  */
235 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
236 
237 /**
238  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
239  * to old_ctx and decrease bs->in_flight again.
240  */
241 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
242 
243 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
244 
245 void coroutine_fn GRAPH_RDLOCK bdrv_co_io_plug(BlockDriverState *bs);
246 void coroutine_fn GRAPH_RDLOCK bdrv_co_io_unplug(BlockDriverState *bs);
247 
248 bool coroutine_fn GRAPH_RDLOCK
249 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
250                                    uint32_t granularity, Error **errp);
251 bool co_wrapper_bdrv_rdlock
252 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
253                                 uint32_t granularity, Error **errp);
254 
255 /**
256  *
257  * bdrv_co_copy_range:
258  *
259  * Do offloaded copy between two children. If the operation is not implemented
260  * by the driver, or if the backend storage doesn't support it, a negative
261  * error code will be returned.
262  *
263  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
264  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
265  * calling copy_file_range(2)) after the first error, thus it should fall back
266  * to a read+write path in the caller level.
267  *
268  * @src: Source child to copy data from
269  * @src_offset: offset in @src image to read data
270  * @dst: Destination child to copy data to
271  * @dst_offset: offset in @dst image to write data
272  * @bytes: number of bytes to copy
273  * @flags: request flags. Supported flags:
274  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
275  *                               write on @dst as if bdrv_co_pwrite_zeroes is
276  *                               called. Used to simplify caller code, or
277  *                               during BlockDriver.bdrv_co_copy_range_from()
278  *                               recursion.
279  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
280  *                                   requests currently in flight.
281  *
282  * Returns: 0 if succeeded; negative error code if failed.
283  **/
284 int coroutine_fn GRAPH_RDLOCK
285 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
286                    BdrvChild *dst, int64_t dst_offset,
287                    int64_t bytes, BdrvRequestFlags read_flags,
288                    BdrvRequestFlags write_flags);
289 
290 /*
291  * "I/O or GS" API functions. These functions can run without
292  * the BQL, but only in one specific iothread/main loop.
293  *
294  * More specifically, these functions use BDRV_POLL_WHILE(bs), which
295  * requires the caller to be either in the main thread and hold
296  * the BlockdriverState (bs) AioContext lock, or directly in the
297  * home thread that runs the bs AioContext. Calling them from
298  * another thread in another AioContext would cause deadlocks.
299  *
300  * Therefore, these functions are not proper I/O, because they
301  * can't run in *any* iothreads, but only in a specific one.
302  *
303  * These functions can call any function from I/O, Common and this
304  * categories, but must be invoked only by other "I/O or GS" and GS APIs.
305  *
306  * All functions in this category must use the macro
307  * IO_OR_GS_CODE();
308  * to catch when they are accidentally called by the wrong API.
309  */
310 
311 #define BDRV_POLL_WHILE(bs, cond) ({                       \
312     BlockDriverState *bs_ = (bs);                          \
313     IO_OR_GS_CODE();                                       \
314     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
315                    cond); })
316 
317 void bdrv_drain(BlockDriverState *bs);
318 
319 int co_wrapper_mixed_bdrv_rdlock
320 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
321               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
322 
323 int co_wrapper_mixed_bdrv_rdlock
324 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
325 
326 /* Invalidate any cached metadata used by image formats */
327 int co_wrapper_mixed_bdrv_rdlock
328 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
329 
330 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs);
331 
332 int co_wrapper_mixed_bdrv_rdlock
333 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
334 
335 int co_wrapper_mixed_bdrv_rdlock
336 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
337 
338 int co_wrapper_mixed_bdrv_rdlock
339 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
340 
341 /**
342  * bdrv_parent_drained_begin_single:
343  *
344  * Begin a quiesced section for the parent of @c.
345  */
346 void bdrv_parent_drained_begin_single(BdrvChild *c);
347 
348 /**
349  * bdrv_parent_drained_poll_single:
350  *
351  * Returns true if there is any pending activity to cease before @c can be
352  * called quiesced, false otherwise.
353  */
354 bool bdrv_parent_drained_poll_single(BdrvChild *c);
355 
356 /**
357  * bdrv_parent_drained_end_single:
358  *
359  * End a quiesced section for the parent of @c.
360  */
361 void bdrv_parent_drained_end_single(BdrvChild *c);
362 
363 /**
364  * bdrv_drain_poll:
365  *
366  * Poll for pending requests in @bs and its parents (except for @ignore_parent).
367  *
368  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
369  * ignore the drain request because they will be drained separately (used for
370  * drain_all).
371  *
372  * This is part of bdrv_drained_begin.
373  */
374 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent,
375                      bool ignore_bds_parents);
376 
377 /**
378  * bdrv_drained_begin:
379  *
380  * Begin a quiesced section for exclusive access to the BDS, by disabling
381  * external request sources including NBD server, block jobs, and device model.
382  *
383  * This function can be recursive.
384  */
385 void bdrv_drained_begin(BlockDriverState *bs);
386 
387 /**
388  * bdrv_do_drained_begin_quiesce:
389  *
390  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
391  * running requests to complete.
392  */
393 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent);
394 
395 /**
396  * bdrv_drained_end:
397  *
398  * End a quiescent section started by bdrv_drained_begin().
399  */
400 void bdrv_drained_end(BlockDriverState *bs);
401 
402 #endif /* BLOCK_IO_H */
403