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); 153 154 void bdrv_enable_copy_on_read(BlockDriverState *bs); 155 void bdrv_disable_copy_on_read(BlockDriverState *bs); 156 157 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); 158 159 #define BLKDBG_EVENT(child, evt) \ 160 do { \ 161 if (child) { \ 162 bdrv_debug_event(child->bs, evt); \ 163 } \ 164 } while (0) 165 166 /** 167 * bdrv_get_aio_context: 168 * 169 * Returns: the currently bound #AioContext 170 */ 171 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 172 173 /** 174 * Move the current coroutine to the AioContext of @bs and return the old 175 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 176 * will wait for the operation to proceed until the corresponding 177 * bdrv_co_leave(). 178 * 179 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 180 * this will deadlock. 181 */ 182 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 183 184 /** 185 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 186 * to old_ctx and decrease bs->in_flight again. 187 */ 188 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 189 190 /** 191 * Transfer control to @co in the aio context of @bs 192 */ 193 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co); 194 195 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c); 196 197 void bdrv_io_plug(BlockDriverState *bs); 198 void bdrv_io_unplug(BlockDriverState *bs); 199 200 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 201 uint32_t granularity, Error **errp); 202 203 /** 204 * 205 * bdrv_co_copy_range: 206 * 207 * Do offloaded copy between two children. If the operation is not implemented 208 * by the driver, or if the backend storage doesn't support it, a negative 209 * error code will be returned. 210 * 211 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 212 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 213 * calling copy_file_range(2)) after the first error, thus it should fall back 214 * to a read+write path in the caller level. 215 * 216 * @src: Source child to copy data from 217 * @src_offset: offset in @src image to read data 218 * @dst: Destination child to copy data to 219 * @dst_offset: offset in @dst image to write data 220 * @bytes: number of bytes to copy 221 * @flags: request flags. Supported flags: 222 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 223 * write on @dst as if bdrv_co_pwrite_zeroes is 224 * called. Used to simplify caller code, or 225 * during BlockDriver.bdrv_co_copy_range_from() 226 * recursion. 227 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 228 * requests currently in flight. 229 * 230 * Returns: 0 if succeeded; negative error code if failed. 231 **/ 232 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 233 BdrvChild *dst, int64_t dst_offset, 234 int64_t bytes, BdrvRequestFlags read_flags, 235 BdrvRequestFlags write_flags); 236 237 /** 238 * bdrv_drained_end_no_poll: 239 * 240 * Same as bdrv_drained_end(), but do not poll for the subgraph to 241 * actually become unquiesced. Therefore, no graph changes will occur 242 * with this function. 243 * 244 * *drained_end_counter is incremented for every background operation 245 * that is scheduled, and will be decremented for every operation once 246 * it settles. The caller must poll until it reaches 0. The counter 247 * should be accessed using atomic operations only. 248 */ 249 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter); 250 251 252 /* 253 * "I/O or GS" API functions. These functions can run without 254 * the BQL, but only in one specific iothread/main loop. 255 * 256 * More specifically, these functions use BDRV_POLL_WHILE(bs), which 257 * requires the caller to be either in the main thread and hold 258 * the BlockdriverState (bs) AioContext lock, or directly in the 259 * home thread that runs the bs AioContext. Calling them from 260 * another thread in another AioContext would cause deadlocks. 261 * 262 * Therefore, these functions are not proper I/O, because they 263 * can't run in *any* iothreads, but only in a specific one. 264 * 265 * These functions can call any function from I/O, Common and this 266 * categories, but must be invoked only by other "I/O or GS" and GS APIs. 267 * 268 * All functions in this category must use the macro 269 * IO_OR_GS_CODE(); 270 * to catch when they are accidentally called by the wrong API. 271 */ 272 273 #define BDRV_POLL_WHILE(bs, cond) ({ \ 274 BlockDriverState *bs_ = (bs); \ 275 IO_OR_GS_CODE(); \ 276 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 277 cond); }) 278 279 void bdrv_drain(BlockDriverState *bs); 280 281 int generated_co_wrapper 282 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 283 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 284 285 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, 286 BdrvCheckMode fix); 287 288 /* Invalidate any cached metadata used by image formats */ 289 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs, 290 Error **errp); 291 int generated_co_wrapper bdrv_flush(BlockDriverState *bs); 292 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset, 293 int64_t bytes); 294 int generated_co_wrapper 295 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 296 int generated_co_wrapper 297 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 298 299 /** 300 * bdrv_parent_drained_begin_single: 301 * 302 * Begin a quiesced section for the parent of @c. If @poll is true, wait for 303 * any pending activity to cease. 304 */ 305 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll); 306 307 /** 308 * bdrv_parent_drained_end_single: 309 * 310 * End a quiesced section for the parent of @c. 311 * 312 * This polls @bs's AioContext until all scheduled sub-drained_ends 313 * have settled, which may result in graph changes. 314 */ 315 void bdrv_parent_drained_end_single(BdrvChild *c); 316 317 /** 318 * bdrv_drain_poll: 319 * 320 * Poll for pending requests in @bs, its parents (except for @ignore_parent), 321 * and if @recursive is true its children as well (used for subtree drain). 322 * 323 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 324 * ignore the drain request because they will be drained separately (used for 325 * drain_all). 326 * 327 * This is part of bdrv_drained_begin. 328 */ 329 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive, 330 BdrvChild *ignore_parent, bool ignore_bds_parents); 331 332 /** 333 * bdrv_drained_begin: 334 * 335 * Begin a quiesced section for exclusive access to the BDS, by disabling 336 * external request sources including NBD server, block jobs, and device model. 337 * 338 * This function can be recursive. 339 */ 340 void bdrv_drained_begin(BlockDriverState *bs); 341 342 /** 343 * bdrv_do_drained_begin_quiesce: 344 * 345 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 346 * running requests to complete. 347 */ 348 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, 349 BdrvChild *parent, bool ignore_bds_parents); 350 351 /** 352 * Like bdrv_drained_begin, but recursively begins a quiesced section for 353 * exclusive access to all child nodes as well. 354 */ 355 void bdrv_subtree_drained_begin(BlockDriverState *bs); 356 357 /** 358 * bdrv_drained_end: 359 * 360 * End a quiescent section started by bdrv_drained_begin(). 361 * 362 * This polls @bs's AioContext until all scheduled sub-drained_ends 363 * have settled. On one hand, that may result in graph changes. On 364 * the other, this requires that the caller either runs in the main 365 * loop; or that all involved nodes (@bs and all of its parents) are 366 * in the caller's AioContext. 367 */ 368 void bdrv_drained_end(BlockDriverState *bs); 369 370 /** 371 * End a quiescent section started by bdrv_subtree_drained_begin(). 372 */ 373 void bdrv_subtree_drained_end(BlockDriverState *bs); 374 375 #endif /* BLOCK_IO_H */ 376