xref: /openbmc/qemu/include/block/block-io.h (revision 2344f98e)
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 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 bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
92 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
93 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
94                       int64_t bytes, int64_t *pnum, int64_t *map,
95                       BlockDriverState **file);
96 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
97                             int64_t offset, int64_t bytes, int64_t *pnum,
98                             int64_t *map, BlockDriverState **file);
99 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
100                       int64_t *pnum);
101 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
102                             bool include_base, int64_t offset, int64_t bytes,
103                             int64_t *pnum);
104 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
105                                       int64_t bytes);
106 
107 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
108                            bool ignore_allow_rdw, Error **errp);
109 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
110                               Error **errp);
111 bool bdrv_is_read_only(BlockDriverState *bs);
112 bool bdrv_is_writable(BlockDriverState *bs);
113 bool bdrv_is_sg(BlockDriverState *bs);
114 int bdrv_get_flags(BlockDriverState *bs);
115 bool bdrv_is_inserted(BlockDriverState *bs);
116 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
117 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
118 const char *bdrv_get_format_name(BlockDriverState *bs);
119 
120 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
121 const char *bdrv_get_node_name(const BlockDriverState *bs);
122 const char *bdrv_get_device_name(const BlockDriverState *bs);
123 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
124 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
125 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
126                                           Error **errp);
127 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
128 void bdrv_round_to_clusters(BlockDriverState *bs,
129                             int64_t offset, int64_t bytes,
130                             int64_t *cluster_offset,
131                             int64_t *cluster_bytes);
132 
133 void bdrv_get_backing_filename(BlockDriverState *bs,
134                                char *filename, int filename_size);
135 
136 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
137                       int64_t pos, int size);
138 
139 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
140                       int64_t pos, int size);
141 
142 /*
143  * Returns the alignment in bytes that is required so that no bounce buffer
144  * is required throughout the stack
145  */
146 size_t bdrv_min_mem_align(BlockDriverState *bs);
147 /* Returns optimal alignment in bytes for bounce buffer */
148 size_t bdrv_opt_mem_align(BlockDriverState *bs);
149 void *qemu_blockalign(BlockDriverState *bs, size_t size);
150 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
151 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
152 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
153 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
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 /**
175  * Move the current coroutine to the AioContext of @bs and return the old
176  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
177  * will wait for the operation to proceed until the corresponding
178  * bdrv_co_leave().
179  *
180  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
181  * this will deadlock.
182  */
183 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
184 
185 /**
186  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
187  * to old_ctx and decrease bs->in_flight again.
188  */
189 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
190 
191 /**
192  * Transfer control to @co in the aio context of @bs
193  */
194 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
195 
196 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
197 
198 void bdrv_io_plug(BlockDriverState *bs);
199 void bdrv_io_unplug(BlockDriverState *bs);
200 
201 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
202                                      uint32_t granularity, Error **errp);
203 
204 /**
205  *
206  * bdrv_co_copy_range:
207  *
208  * Do offloaded copy between two children. If the operation is not implemented
209  * by the driver, or if the backend storage doesn't support it, a negative
210  * error code will be returned.
211  *
212  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
213  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
214  * calling copy_file_range(2)) after the first error, thus it should fall back
215  * to a read+write path in the caller level.
216  *
217  * @src: Source child to copy data from
218  * @src_offset: offset in @src image to read data
219  * @dst: Destination child to copy data to
220  * @dst_offset: offset in @dst image to write data
221  * @bytes: number of bytes to copy
222  * @flags: request flags. Supported flags:
223  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
224  *                               write on @dst as if bdrv_co_pwrite_zeroes is
225  *                               called. Used to simplify caller code, or
226  *                               during BlockDriver.bdrv_co_copy_range_from()
227  *                               recursion.
228  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
229  *                                   requests currently in flight.
230  *
231  * Returns: 0 if succeeded; negative error code if failed.
232  **/
233 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
234                                     BdrvChild *dst, int64_t dst_offset,
235                                     int64_t bytes, BdrvRequestFlags read_flags,
236                                     BdrvRequestFlags write_flags);
237 
238 /**
239  * bdrv_drained_end_no_poll:
240  *
241  * Same as bdrv_drained_end(), but do not poll for the subgraph to
242  * actually become unquiesced.  Therefore, no graph changes will occur
243  * with this function.
244  *
245  * *drained_end_counter is incremented for every background operation
246  * that is scheduled, and will be decremented for every operation once
247  * it settles.  The caller must poll until it reaches 0.  The counter
248  * should be accessed using atomic operations only.
249  */
250 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
251 
252 
253 /*
254  * "I/O or GS" API functions. These functions can run without
255  * the BQL, but only in one specific iothread/main loop.
256  *
257  * More specifically, these functions use BDRV_POLL_WHILE(bs), which
258  * requires the caller to be either in the main thread and hold
259  * the BlockdriverState (bs) AioContext lock, or directly in the
260  * home thread that runs the bs AioContext. Calling them from
261  * another thread in another AioContext would cause deadlocks.
262  *
263  * Therefore, these functions are not proper I/O, because they
264  * can't run in *any* iothreads, but only in a specific one.
265  *
266  * These functions can call any function from I/O, Common and this
267  * categories, but must be invoked only by other "I/O or GS" and GS APIs.
268  *
269  * All functions in this category must use the macro
270  * IO_OR_GS_CODE();
271  * to catch when they are accidentally called by the wrong API.
272  */
273 
274 #define BDRV_POLL_WHILE(bs, cond) ({                       \
275     BlockDriverState *bs_ = (bs);                          \
276     IO_OR_GS_CODE();                                       \
277     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
278                    cond); })
279 
280 void bdrv_drain(BlockDriverState *bs);
281 
282 int generated_co_wrapper
283 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
284               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
285 
286 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
287                                     BdrvCheckMode fix);
288 
289 /* Invalidate any cached metadata used by image formats */
290 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
291                                                Error **errp);
292 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
293 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
294                                        int64_t bytes);
295 int generated_co_wrapper
296 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
297 int generated_co_wrapper
298 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
299 
300 /**
301  * bdrv_parent_drained_begin_single:
302  *
303  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
304  * any pending activity to cease.
305  */
306 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
307 
308 /**
309  * bdrv_parent_drained_end_single:
310  *
311  * End a quiesced section for the parent of @c.
312  *
313  * This polls @bs's AioContext until all scheduled sub-drained_ends
314  * have settled, which may result in graph changes.
315  */
316 void bdrv_parent_drained_end_single(BdrvChild *c);
317 
318 /**
319  * bdrv_drain_poll:
320  *
321  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
322  * and if @recursive is true its children as well (used for subtree drain).
323  *
324  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
325  * ignore the drain request because they will be drained separately (used for
326  * drain_all).
327  *
328  * This is part of bdrv_drained_begin.
329  */
330 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
331                      BdrvChild *ignore_parent, bool ignore_bds_parents);
332 
333 /**
334  * bdrv_drained_begin:
335  *
336  * Begin a quiesced section for exclusive access to the BDS, by disabling
337  * external request sources including NBD server, block jobs, and device model.
338  *
339  * This function can be recursive.
340  */
341 void bdrv_drained_begin(BlockDriverState *bs);
342 
343 /**
344  * bdrv_do_drained_begin_quiesce:
345  *
346  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
347  * running requests to complete.
348  */
349 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
350                                    BdrvChild *parent, bool ignore_bds_parents);
351 
352 /**
353  * Like bdrv_drained_begin, but recursively begins a quiesced section for
354  * exclusive access to all child nodes as well.
355  */
356 void bdrv_subtree_drained_begin(BlockDriverState *bs);
357 
358 /**
359  * bdrv_drained_end:
360  *
361  * End a quiescent section started by bdrv_drained_begin().
362  *
363  * This polls @bs's AioContext until all scheduled sub-drained_ends
364  * have settled.  On one hand, that may result in graph changes.  On
365  * the other, this requires that the caller either runs in the main
366  * loop; or that all involved nodes (@bs and all of its parents) are
367  * in the caller's AioContext.
368  */
369 void bdrv_drained_end(BlockDriverState *bs);
370 
371 /**
372  * End a quiescent section started by bdrv_subtree_drained_begin().
373  */
374 void bdrv_subtree_drained_end(BlockDriverState *bs);
375 
376 #endif /* BLOCK_IO_H */
377