xref: /openbmc/qemu/include/block/block-io.h (revision 05e385d2)
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-common.h"
28 
29 /*
30  * I/O API functions. These functions are thread-safe, and therefore
31  * can run in any thread as long as the thread has called
32  * aio_context_acquire/release().
33  *
34  * These functions can only call functions from I/O and Common categories,
35  * but can be invoked by GS, "I/O or GS" and I/O APIs.
36  *
37  * All functions in this category must use the macro
38  * IO_CODE();
39  * to catch when they are accidentally called by the wrong API.
40  */
41 
42 int generated_co_wrapper bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
43                                             int64_t bytes,
44                                             BdrvRequestFlags flags);
45 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
46 int generated_co_wrapper bdrv_pread(BdrvChild *child, int64_t offset,
47                                     int64_t bytes, void *buf,
48                                     BdrvRequestFlags flags);
49 int generated_co_wrapper bdrv_pwrite(BdrvChild *child, int64_t offset,
50                                      int64_t bytes, const void *buf,
51                                      BdrvRequestFlags flags);
52 int generated_co_wrapper bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
53                                           int64_t bytes, const void *buf,
54                                           BdrvRequestFlags flags);
55 int coroutine_fn bdrv_co_pwrite_sync(BdrvChild *child, int64_t offset,
56                                      int64_t bytes, const void *buf,
57                                      BdrvRequestFlags flags);
58 /*
59  * Efficiently zero a region of the disk image.  Note that this is a regular
60  * I/O request like read or write and should have a reasonable size.  This
61  * function is not suitable for zeroing the entire image in a single request
62  * because it may allocate memory for the entire region.
63  */
64 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
65                                        int64_t bytes, BdrvRequestFlags flags);
66 
67 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
68                                   PreallocMode prealloc, BdrvRequestFlags flags,
69                                   Error **errp);
70 
71 int64_t bdrv_nb_sectors(BlockDriverState *bs);
72 int64_t bdrv_getlength(BlockDriverState *bs);
73 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
74 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
75                                BlockDriverState *in_bs, Error **errp);
76 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
77 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
78 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs);
79 
80 
81 /* async block I/O */
82 void bdrv_aio_cancel(BlockAIOCB *acb);
83 void bdrv_aio_cancel_async(BlockAIOCB *acb);
84 
85 /* sg packet commands */
86 int coroutine_fn bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
87 
88 /* Ensure contents are flushed to disk.  */
89 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
90 
91 int coroutine_fn bdrv_co_pdiscard(BdrvChild *child, int64_t offset,
92                                   int64_t bytes);
93 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
94 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
95                       int64_t bytes, int64_t *pnum, int64_t *map,
96                       BlockDriverState **file);
97 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
98                             int64_t offset, int64_t bytes, int64_t *pnum,
99                             int64_t *map, BlockDriverState **file);
100 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
101                       int64_t *pnum);
102 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
103                             bool include_base, int64_t offset, int64_t bytes,
104                             int64_t *pnum);
105 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
106                                       int64_t bytes);
107 
108 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
109                            bool ignore_allow_rdw, Error **errp);
110 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
111                               Error **errp);
112 bool bdrv_is_read_only(BlockDriverState *bs);
113 bool bdrv_is_writable(BlockDriverState *bs);
114 bool bdrv_is_sg(BlockDriverState *bs);
115 int bdrv_get_flags(BlockDriverState *bs);
116 bool bdrv_is_inserted(BlockDriverState *bs);
117 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
118 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
119 const char *bdrv_get_format_name(BlockDriverState *bs);
120 
121 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
122 const char *bdrv_get_node_name(const BlockDriverState *bs);
123 const char *bdrv_get_device_name(const BlockDriverState *bs);
124 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
125 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
126 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
127                                           Error **errp);
128 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
129 void bdrv_round_to_clusters(BlockDriverState *bs,
130                             int64_t offset, int64_t bytes,
131                             int64_t *cluster_offset,
132                             int64_t *cluster_bytes);
133 
134 void bdrv_get_backing_filename(BlockDriverState *bs,
135                                char *filename, int filename_size);
136 
137 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
138                       int64_t pos, int size);
139 
140 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
141                       int64_t pos, int size);
142 
143 /*
144  * Returns the alignment in bytes that is required so that no bounce buffer
145  * is required throughout the stack
146  */
147 size_t bdrv_min_mem_align(BlockDriverState *bs);
148 /* Returns optimal alignment in bytes for bounce buffer */
149 size_t bdrv_opt_mem_align(BlockDriverState *bs);
150 void *qemu_blockalign(BlockDriverState *bs, size_t size);
151 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
152 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
153 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
154 
155 void bdrv_enable_copy_on_read(BlockDriverState *bs);
156 void bdrv_disable_copy_on_read(BlockDriverState *bs);
157 
158 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
159 
160 #define BLKDBG_EVENT(child, evt) \
161     do { \
162         if (child) { \
163             bdrv_debug_event(child->bs, evt); \
164         } \
165     } while (0)
166 
167 /**
168  * bdrv_get_aio_context:
169  *
170  * Returns: the currently bound #AioContext
171  */
172 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
173 
174 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
175 
176 /**
177  * Move the current coroutine to the AioContext of @bs and return the old
178  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
179  * will wait for the operation to proceed until the corresponding
180  * bdrv_co_leave().
181  *
182  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
183  * this will deadlock.
184  */
185 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
186 
187 /**
188  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
189  * to old_ctx and decrease bs->in_flight again.
190  */
191 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
192 
193 /**
194  * Transfer control to @co in the aio context of @bs
195  */
196 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
197 
198 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
199 
200 void bdrv_io_plug(BlockDriverState *bs);
201 void bdrv_io_unplug(BlockDriverState *bs);
202 
203 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
204                                      uint32_t granularity, Error **errp);
205 
206 /**
207  *
208  * bdrv_co_copy_range:
209  *
210  * Do offloaded copy between two children. If the operation is not implemented
211  * by the driver, or if the backend storage doesn't support it, a negative
212  * error code will be returned.
213  *
214  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
215  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
216  * calling copy_file_range(2)) after the first error, thus it should fall back
217  * to a read+write path in the caller level.
218  *
219  * @src: Source child to copy data from
220  * @src_offset: offset in @src image to read data
221  * @dst: Destination child to copy data to
222  * @dst_offset: offset in @dst image to write data
223  * @bytes: number of bytes to copy
224  * @flags: request flags. Supported flags:
225  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
226  *                               write on @dst as if bdrv_co_pwrite_zeroes is
227  *                               called. Used to simplify caller code, or
228  *                               during BlockDriver.bdrv_co_copy_range_from()
229  *                               recursion.
230  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
231  *                                   requests currently in flight.
232  *
233  * Returns: 0 if succeeded; negative error code if failed.
234  **/
235 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
236                                     BdrvChild *dst, int64_t dst_offset,
237                                     int64_t bytes, BdrvRequestFlags read_flags,
238                                     BdrvRequestFlags write_flags);
239 
240 /**
241  * bdrv_drained_end_no_poll:
242  *
243  * Same as bdrv_drained_end(), but do not poll for the subgraph to
244  * actually become unquiesced.  Therefore, no graph changes will occur
245  * with this function.
246  *
247  * *drained_end_counter is incremented for every background operation
248  * that is scheduled, and will be decremented for every operation once
249  * it settles.  The caller must poll until it reaches 0.  The counter
250  * should be accessed using atomic operations only.
251  */
252 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
253 
254 
255 /*
256  * "I/O or GS" API functions. These functions can run without
257  * the BQL, but only in one specific iothread/main loop.
258  *
259  * More specifically, these functions use BDRV_POLL_WHILE(bs), which
260  * requires the caller to be either in the main thread and hold
261  * the BlockdriverState (bs) AioContext lock, or directly in the
262  * home thread that runs the bs AioContext. Calling them from
263  * another thread in another AioContext would cause deadlocks.
264  *
265  * Therefore, these functions are not proper I/O, because they
266  * can't run in *any* iothreads, but only in a specific one.
267  *
268  * These functions can call any function from I/O, Common and this
269  * categories, but must be invoked only by other "I/O or GS" and GS APIs.
270  *
271  * All functions in this category must use the macro
272  * IO_OR_GS_CODE();
273  * to catch when they are accidentally called by the wrong API.
274  */
275 
276 #define BDRV_POLL_WHILE(bs, cond) ({                       \
277     BlockDriverState *bs_ = (bs);                          \
278     IO_OR_GS_CODE();                                       \
279     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
280                    cond); })
281 
282 void bdrv_drain(BlockDriverState *bs);
283 
284 int generated_co_wrapper
285 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
286               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
287 
288 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
289                                     BdrvCheckMode fix);
290 
291 /* Invalidate any cached metadata used by image formats */
292 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
293                                                Error **errp);
294 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
295 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
296                                        int64_t bytes);
297 int generated_co_wrapper
298 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
299 int generated_co_wrapper
300 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
301 
302 /**
303  * bdrv_parent_drained_begin_single:
304  *
305  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
306  * any pending activity to cease.
307  */
308 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
309 
310 /**
311  * bdrv_parent_drained_end_single:
312  *
313  * End a quiesced section for the parent of @c.
314  *
315  * This polls @bs's AioContext until all scheduled sub-drained_ends
316  * have settled, which may result in graph changes.
317  */
318 void bdrv_parent_drained_end_single(BdrvChild *c);
319 
320 /**
321  * bdrv_drain_poll:
322  *
323  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
324  * and if @recursive is true its children as well (used for subtree drain).
325  *
326  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
327  * ignore the drain request because they will be drained separately (used for
328  * drain_all).
329  *
330  * This is part of bdrv_drained_begin.
331  */
332 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
333                      BdrvChild *ignore_parent, bool ignore_bds_parents);
334 
335 /**
336  * bdrv_drained_begin:
337  *
338  * Begin a quiesced section for exclusive access to the BDS, by disabling
339  * external request sources including NBD server, block jobs, and device model.
340  *
341  * This function can be recursive.
342  */
343 void bdrv_drained_begin(BlockDriverState *bs);
344 
345 /**
346  * bdrv_do_drained_begin_quiesce:
347  *
348  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
349  * running requests to complete.
350  */
351 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
352                                    BdrvChild *parent, bool ignore_bds_parents);
353 
354 /**
355  * Like bdrv_drained_begin, but recursively begins a quiesced section for
356  * exclusive access to all child nodes as well.
357  */
358 void bdrv_subtree_drained_begin(BlockDriverState *bs);
359 
360 /**
361  * bdrv_drained_end:
362  *
363  * End a quiescent section started by bdrv_drained_begin().
364  *
365  * This polls @bs's AioContext until all scheduled sub-drained_ends
366  * have settled.  On one hand, that may result in graph changes.  On
367  * the other, this requires that the caller either runs in the main
368  * loop; or that all involved nodes (@bs and all of its parents) are
369  * in the caller's AioContext.
370  */
371 void bdrv_drained_end(BlockDriverState *bs);
372 
373 /**
374  * End a quiescent section started by bdrv_subtree_drained_begin().
375  */
376 void bdrv_subtree_drained_end(BlockDriverState *bs);
377 
378 #endif /* BLOCK_IO_H */
379