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 coroutine_mixed_fn 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 GRAPH_RDLOCK 88 bdrv_co_get_allocated_file_size(BlockDriverState *bs); 89 90 int64_t co_wrapper_bdrv_rdlock 91 bdrv_get_allocated_file_size(BlockDriverState *bs); 92 93 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts, 94 BlockDriverState *in_bs, Error **errp); 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 /* Report zone information of zone block device. */ 118 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_report(BlockDriverState *bs, 119 int64_t offset, 120 unsigned int *nr_zones, 121 BlockZoneDescriptor *zones); 122 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_mgmt(BlockDriverState *bs, 123 BlockZoneOp op, 124 int64_t offset, int64_t len); 125 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_append(BlockDriverState *bs, 126 int64_t *offset, 127 QEMUIOVector *qiov, 128 BdrvRequestFlags flags); 129 130 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs); 131 int bdrv_block_status(BlockDriverState *bs, int64_t offset, 132 int64_t bytes, int64_t *pnum, int64_t *map, 133 BlockDriverState **file); 134 135 int coroutine_fn GRAPH_RDLOCK 136 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base, 137 int64_t offset, int64_t bytes, int64_t *pnum, 138 int64_t *map, BlockDriverState **file); 139 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base, 140 int64_t offset, int64_t bytes, int64_t *pnum, 141 int64_t *map, BlockDriverState **file); 142 143 int coroutine_fn GRAPH_RDLOCK 144 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 145 int64_t *pnum); 146 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 147 int64_t *pnum); 148 149 int coroutine_fn GRAPH_RDLOCK 150 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 151 bool include_base, int64_t offset, int64_t bytes, 152 int64_t *pnum); 153 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 154 bool include_base, int64_t offset, int64_t bytes, 155 int64_t *pnum); 156 157 int coroutine_fn GRAPH_RDLOCK 158 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes); 159 160 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg, 161 Error **errp); 162 bool bdrv_is_read_only(BlockDriverState *bs); 163 bool bdrv_is_writable(BlockDriverState *bs); 164 bool bdrv_is_sg(BlockDriverState *bs); 165 int bdrv_get_flags(BlockDriverState *bs); 166 167 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs); 168 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs); 169 170 void coroutine_fn GRAPH_RDLOCK 171 bdrv_co_lock_medium(BlockDriverState *bs, bool locked); 172 173 void coroutine_fn GRAPH_RDLOCK 174 bdrv_co_eject(BlockDriverState *bs, bool eject_flag); 175 176 const char *bdrv_get_format_name(BlockDriverState *bs); 177 178 bool bdrv_supports_compressed_writes(BlockDriverState *bs); 179 const char *bdrv_get_node_name(const BlockDriverState *bs); 180 const char *bdrv_get_device_name(const BlockDriverState *bs); 181 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs); 182 183 int coroutine_fn GRAPH_RDLOCK 184 bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 185 186 int co_wrapper_mixed_bdrv_rdlock 187 bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 188 189 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, 190 Error **errp); 191 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs); 192 void bdrv_round_to_clusters(BlockDriverState *bs, 193 int64_t offset, int64_t bytes, 194 int64_t *cluster_offset, 195 int64_t *cluster_bytes); 196 197 void bdrv_get_backing_filename(BlockDriverState *bs, 198 char *filename, int filename_size); 199 200 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, 201 int64_t pos, int size); 202 203 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, 204 int64_t pos, int size); 205 206 /* 207 * Returns the alignment in bytes that is required so that no bounce buffer 208 * is required throughout the stack 209 */ 210 size_t bdrv_min_mem_align(BlockDriverState *bs); 211 /* Returns optimal alignment in bytes for bounce buffer */ 212 size_t bdrv_opt_mem_align(BlockDriverState *bs); 213 void *qemu_blockalign(BlockDriverState *bs, size_t size); 214 void *qemu_blockalign0(BlockDriverState *bs, size_t size); 215 void *qemu_try_blockalign(BlockDriverState *bs, size_t size); 216 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); 217 218 void bdrv_enable_copy_on_read(BlockDriverState *bs); 219 void bdrv_disable_copy_on_read(BlockDriverState *bs); 220 221 void coroutine_fn GRAPH_RDLOCK 222 bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event); 223 224 void co_wrapper_mixed_bdrv_rdlock 225 bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); 226 227 #define BLKDBG_CO_EVENT(child, evt) \ 228 do { \ 229 if (child) { \ 230 bdrv_co_debug_event(child->bs, evt); \ 231 } \ 232 } while (0) 233 234 #define BLKDBG_EVENT(child, evt) \ 235 do { \ 236 if (child) { \ 237 bdrv_debug_event(child->bs, evt); \ 238 } \ 239 } while (0) 240 241 /** 242 * bdrv_get_aio_context: 243 * 244 * Returns: the currently bound #AioContext 245 */ 246 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 247 248 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c); 249 250 /** 251 * Move the current coroutine to the AioContext of @bs and return the old 252 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 253 * will wait for the operation to proceed until the corresponding 254 * bdrv_co_leave(). 255 * 256 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 257 * this will deadlock. 258 */ 259 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 260 261 /** 262 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 263 * to old_ctx and decrease bs->in_flight again. 264 */ 265 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 266 267 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c); 268 269 bool coroutine_fn GRAPH_RDLOCK 270 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 271 uint32_t granularity, Error **errp); 272 bool co_wrapper_bdrv_rdlock 273 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 274 uint32_t granularity, Error **errp); 275 276 /** 277 * 278 * bdrv_co_copy_range: 279 * 280 * Do offloaded copy between two children. If the operation is not implemented 281 * by the driver, or if the backend storage doesn't support it, a negative 282 * error code will be returned. 283 * 284 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 285 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 286 * calling copy_file_range(2)) after the first error, thus it should fall back 287 * to a read+write path in the caller level. 288 * 289 * @src: Source child to copy data from 290 * @src_offset: offset in @src image to read data 291 * @dst: Destination child to copy data to 292 * @dst_offset: offset in @dst image to write data 293 * @bytes: number of bytes to copy 294 * @flags: request flags. Supported flags: 295 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 296 * write on @dst as if bdrv_co_pwrite_zeroes is 297 * called. Used to simplify caller code, or 298 * during BlockDriver.bdrv_co_copy_range_from() 299 * recursion. 300 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 301 * requests currently in flight. 302 * 303 * Returns: 0 if succeeded; negative error code if failed. 304 **/ 305 int coroutine_fn GRAPH_RDLOCK 306 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 307 BdrvChild *dst, int64_t dst_offset, 308 int64_t bytes, BdrvRequestFlags read_flags, 309 BdrvRequestFlags write_flags); 310 311 /* 312 * "I/O or GS" API functions. These functions can run without 313 * the BQL, but only in one specific iothread/main loop. 314 * 315 * More specifically, these functions use BDRV_POLL_WHILE(bs), which 316 * requires the caller to be either in the main thread and hold 317 * the BlockdriverState (bs) AioContext lock, or directly in the 318 * home thread that runs the bs AioContext. Calling them from 319 * another thread in another AioContext would cause deadlocks. 320 * 321 * Therefore, these functions are not proper I/O, because they 322 * can't run in *any* iothreads, but only in a specific one. 323 * 324 * These functions can call any function from I/O, Common and this 325 * categories, but must be invoked only by other "I/O or GS" and GS APIs. 326 * 327 * All functions in this category must use the macro 328 * IO_OR_GS_CODE(); 329 * to catch when they are accidentally called by the wrong API. 330 */ 331 332 #define BDRV_POLL_WHILE(bs, cond) ({ \ 333 BlockDriverState *bs_ = (bs); \ 334 IO_OR_GS_CODE(); \ 335 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 336 cond); }) 337 338 void bdrv_drain(BlockDriverState *bs); 339 340 int co_wrapper_mixed_bdrv_rdlock 341 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 342 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 343 344 int co_wrapper_mixed_bdrv_rdlock 345 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix); 346 347 /* Invalidate any cached metadata used by image formats */ 348 int co_wrapper_mixed_bdrv_rdlock 349 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp); 350 351 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs); 352 353 int co_wrapper_mixed_bdrv_rdlock 354 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes); 355 356 int co_wrapper_mixed_bdrv_rdlock 357 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 358 359 int co_wrapper_mixed_bdrv_rdlock 360 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 361 362 /** 363 * bdrv_parent_drained_begin_single: 364 * 365 * Begin a quiesced section for the parent of @c. 366 */ 367 void bdrv_parent_drained_begin_single(BdrvChild *c); 368 369 /** 370 * bdrv_parent_drained_poll_single: 371 * 372 * Returns true if there is any pending activity to cease before @c can be 373 * called quiesced, false otherwise. 374 */ 375 bool bdrv_parent_drained_poll_single(BdrvChild *c); 376 377 /** 378 * bdrv_parent_drained_end_single: 379 * 380 * End a quiesced section for the parent of @c. 381 */ 382 void bdrv_parent_drained_end_single(BdrvChild *c); 383 384 /** 385 * bdrv_drain_poll: 386 * 387 * Poll for pending requests in @bs and its parents (except for @ignore_parent). 388 * 389 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 390 * ignore the drain request because they will be drained separately (used for 391 * drain_all). 392 * 393 * This is part of bdrv_drained_begin. 394 */ 395 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent, 396 bool ignore_bds_parents); 397 398 /** 399 * bdrv_drained_begin: 400 * 401 * Begin a quiesced section for exclusive access to the BDS, by disabling 402 * external request sources including NBD server, block jobs, and device model. 403 * 404 * This function can be recursive. 405 */ 406 void bdrv_drained_begin(BlockDriverState *bs); 407 408 /** 409 * bdrv_do_drained_begin_quiesce: 410 * 411 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 412 * running requests to complete. 413 */ 414 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent); 415 416 /** 417 * bdrv_drained_end: 418 * 419 * End a quiescent section started by bdrv_drained_begin(). 420 */ 421 void bdrv_drained_end(BlockDriverState *bs); 422 423 #endif /* BLOCK_IO_H */ 424