xref: /openbmc/qemu/include/block/block-io.h (revision ffe98631)
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 bdrv_co_pwrite_sync(BdrvChild *child, int64_t offset,
64                                      int64_t bytes, const void *buf,
65                                      BdrvRequestFlags flags);
66 /*
67  * Efficiently zero a region of the disk image.  Note that this is a regular
68  * I/O request like read or write and should have a reasonable size.  This
69  * function is not suitable for zeroing the entire image in a single request
70  * because it may allocate memory for the entire region.
71  */
72 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
73                                        int64_t bytes, BdrvRequestFlags flags);
74 
75 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
76                                   PreallocMode prealloc, BdrvRequestFlags flags,
77                                   Error **errp);
78 
79 int64_t bdrv_nb_sectors(BlockDriverState *bs);
80 int64_t bdrv_getlength(BlockDriverState *bs);
81 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
82 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
83                                BlockDriverState *in_bs, Error **errp);
84 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
85 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
86 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs);
87 
88 
89 /* async block I/O */
90 void bdrv_aio_cancel(BlockAIOCB *acb);
91 void bdrv_aio_cancel_async(BlockAIOCB *acb);
92 
93 /* sg packet commands */
94 int coroutine_fn bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
95 
96 /* Ensure contents are flushed to disk.  */
97 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
98 
99 int coroutine_fn bdrv_co_pdiscard(BdrvChild *child, int64_t offset,
100                                   int64_t bytes);
101 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
102 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
103                       int64_t bytes, int64_t *pnum, int64_t *map,
104                       BlockDriverState **file);
105 
106 int coroutine_fn bdrv_co_block_status_above(BlockDriverState *bs,
107                                             BlockDriverState *base,
108                                             int64_t offset, int64_t bytes,
109                                             int64_t *pnum, int64_t *map,
110                                             BlockDriverState **file);
111 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
112                             int64_t offset, int64_t bytes, int64_t *pnum,
113                             int64_t *map, BlockDriverState **file);
114 
115 int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset,
116                                       int64_t bytes, int64_t *pnum);
117 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
118                       int64_t *pnum);
119 
120 int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top,
121                                             BlockDriverState *base,
122                                             bool include_base, int64_t offset,
123                                             int64_t bytes, int64_t *pnum);
124 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
125                             bool include_base, int64_t offset, int64_t bytes,
126                             int64_t *pnum);
127 
128 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
129                                       int64_t bytes);
130 
131 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
132                            bool ignore_allow_rdw, Error **errp);
133 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
134                               Error **errp);
135 bool bdrv_is_read_only(BlockDriverState *bs);
136 bool bdrv_is_writable(BlockDriverState *bs);
137 bool bdrv_is_sg(BlockDriverState *bs);
138 int bdrv_get_flags(BlockDriverState *bs);
139 bool bdrv_is_inserted(BlockDriverState *bs);
140 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
141 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
142 const char *bdrv_get_format_name(BlockDriverState *bs);
143 
144 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
145 const char *bdrv_get_node_name(const BlockDriverState *bs);
146 const char *bdrv_get_device_name(const BlockDriverState *bs);
147 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
148 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
149 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
150                                           Error **errp);
151 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
152 void bdrv_round_to_clusters(BlockDriverState *bs,
153                             int64_t offset, int64_t bytes,
154                             int64_t *cluster_offset,
155                             int64_t *cluster_bytes);
156 
157 void bdrv_get_backing_filename(BlockDriverState *bs,
158                                char *filename, int filename_size);
159 
160 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
161                       int64_t pos, int size);
162 
163 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
164                       int64_t pos, int size);
165 
166 /*
167  * Returns the alignment in bytes that is required so that no bounce buffer
168  * is required throughout the stack
169  */
170 size_t bdrv_min_mem_align(BlockDriverState *bs);
171 /* Returns optimal alignment in bytes for bounce buffer */
172 size_t bdrv_opt_mem_align(BlockDriverState *bs);
173 void *qemu_blockalign(BlockDriverState *bs, size_t size);
174 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
175 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
176 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
177 
178 void bdrv_enable_copy_on_read(BlockDriverState *bs);
179 void bdrv_disable_copy_on_read(BlockDriverState *bs);
180 
181 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
182 
183 #define BLKDBG_EVENT(child, evt) \
184     do { \
185         if (child) { \
186             bdrv_debug_event(child->bs, evt); \
187         } \
188     } while (0)
189 
190 /**
191  * bdrv_get_aio_context:
192  *
193  * Returns: the currently bound #AioContext
194  */
195 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
196 
197 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
198 
199 /**
200  * Move the current coroutine to the AioContext of @bs and return the old
201  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
202  * will wait for the operation to proceed until the corresponding
203  * bdrv_co_leave().
204  *
205  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
206  * this will deadlock.
207  */
208 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
209 
210 /**
211  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
212  * to old_ctx and decrease bs->in_flight again.
213  */
214 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
215 
216 /**
217  * Transfer control to @co in the aio context of @bs
218  */
219 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
220 
221 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
222 
223 void bdrv_io_plug(BlockDriverState *bs);
224 void bdrv_io_unplug(BlockDriverState *bs);
225 
226 bool coroutine_fn bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs,
227                                                      const char *name,
228                                                      uint32_t granularity,
229                                                      Error **errp);
230 bool co_wrapper bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs,
231                                                 const char *name,
232                                                 uint32_t granularity,
233                                                 Error **errp);
234 
235 /**
236  *
237  * bdrv_co_copy_range:
238  *
239  * Do offloaded copy between two children. If the operation is not implemented
240  * by the driver, or if the backend storage doesn't support it, a negative
241  * error code will be returned.
242  *
243  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
244  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
245  * calling copy_file_range(2)) after the first error, thus it should fall back
246  * to a read+write path in the caller level.
247  *
248  * @src: Source child to copy data from
249  * @src_offset: offset in @src image to read data
250  * @dst: Destination child to copy data to
251  * @dst_offset: offset in @dst image to write data
252  * @bytes: number of bytes to copy
253  * @flags: request flags. Supported flags:
254  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
255  *                               write on @dst as if bdrv_co_pwrite_zeroes is
256  *                               called. Used to simplify caller code, or
257  *                               during BlockDriver.bdrv_co_copy_range_from()
258  *                               recursion.
259  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
260  *                                   requests currently in flight.
261  *
262  * Returns: 0 if succeeded; negative error code if failed.
263  **/
264 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
265                                     BdrvChild *dst, int64_t dst_offset,
266                                     int64_t bytes, BdrvRequestFlags read_flags,
267                                     BdrvRequestFlags write_flags);
268 
269 /*
270  * "I/O or GS" API functions. These functions can run without
271  * the BQL, but only in one specific iothread/main loop.
272  *
273  * More specifically, these functions use BDRV_POLL_WHILE(bs), which
274  * requires the caller to be either in the main thread and hold
275  * the BlockdriverState (bs) AioContext lock, or directly in the
276  * home thread that runs the bs AioContext. Calling them from
277  * another thread in another AioContext would cause deadlocks.
278  *
279  * Therefore, these functions are not proper I/O, because they
280  * can't run in *any* iothreads, but only in a specific one.
281  *
282  * These functions can call any function from I/O, Common and this
283  * categories, but must be invoked only by other "I/O or GS" and GS APIs.
284  *
285  * All functions in this category must use the macro
286  * IO_OR_GS_CODE();
287  * to catch when they are accidentally called by the wrong API.
288  */
289 
290 #define BDRV_POLL_WHILE(bs, cond) ({                       \
291     BlockDriverState *bs_ = (bs);                          \
292     IO_OR_GS_CODE();                                       \
293     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
294                    cond); })
295 
296 void bdrv_drain(BlockDriverState *bs);
297 
298 int co_wrapper_mixed_bdrv_rdlock
299 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
300               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
301 
302 int co_wrapper_mixed_bdrv_rdlock
303 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
304 
305 /* Invalidate any cached metadata used by image formats */
306 int co_wrapper_mixed_bdrv_rdlock
307 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
308 
309 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs);
310 
311 int co_wrapper_mixed_bdrv_rdlock
312 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
313 
314 int co_wrapper_mixed_bdrv_rdlock
315 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
316 
317 int co_wrapper_mixed_bdrv_rdlock
318 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
319 
320 /**
321  * bdrv_parent_drained_begin_single:
322  *
323  * Begin a quiesced section for the parent of @c.
324  */
325 void bdrv_parent_drained_begin_single(BdrvChild *c);
326 
327 /**
328  * bdrv_parent_drained_poll_single:
329  *
330  * Returns true if there is any pending activity to cease before @c can be
331  * called quiesced, false otherwise.
332  */
333 bool bdrv_parent_drained_poll_single(BdrvChild *c);
334 
335 /**
336  * bdrv_parent_drained_end_single:
337  *
338  * End a quiesced section for the parent of @c.
339  */
340 void bdrv_parent_drained_end_single(BdrvChild *c);
341 
342 /**
343  * bdrv_drain_poll:
344  *
345  * Poll for pending requests in @bs and its parents (except for @ignore_parent).
346  *
347  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
348  * ignore the drain request because they will be drained separately (used for
349  * drain_all).
350  *
351  * This is part of bdrv_drained_begin.
352  */
353 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent,
354                      bool ignore_bds_parents);
355 
356 /**
357  * bdrv_drained_begin:
358  *
359  * Begin a quiesced section for exclusive access to the BDS, by disabling
360  * external request sources including NBD server, block jobs, and device model.
361  *
362  * This function can be recursive.
363  */
364 void bdrv_drained_begin(BlockDriverState *bs);
365 
366 /**
367  * bdrv_do_drained_begin_quiesce:
368  *
369  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
370  * running requests to complete.
371  */
372 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent);
373 
374 /**
375  * bdrv_drained_end:
376  *
377  * End a quiescent section started by bdrv_drained_begin().
378  */
379 void bdrv_drained_end(BlockDriverState *bs);
380 
381 #endif /* BLOCK_IO_H */
382