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_async(BlockAIOCB *acb); 105 106 /* sg packet commands */ 107 int coroutine_fn GRAPH_RDLOCK 108 bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf); 109 110 /* Ensure contents are flushed to disk. */ 111 int coroutine_fn GRAPH_RDLOCK bdrv_co_flush(BlockDriverState *bs); 112 113 int coroutine_fn GRAPH_RDLOCK bdrv_co_pdiscard(BdrvChild *child, int64_t offset, 114 int64_t bytes); 115 116 /* Report zone information of zone block device. */ 117 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_report(BlockDriverState *bs, 118 int64_t offset, 119 unsigned int *nr_zones, 120 BlockZoneDescriptor *zones); 121 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_mgmt(BlockDriverState *bs, 122 BlockZoneOp op, 123 int64_t offset, int64_t len); 124 int coroutine_fn GRAPH_RDLOCK bdrv_co_zone_append(BlockDriverState *bs, 125 int64_t *offset, 126 QEMUIOVector *qiov, 127 BdrvRequestFlags flags); 128 129 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs); 130 131 int coroutine_fn GRAPH_RDLOCK 132 bdrv_co_block_status(BlockDriverState *bs, int64_t offset, int64_t bytes, 133 int64_t *pnum, int64_t *map, BlockDriverState **file); 134 int co_wrapper_mixed_bdrv_rdlock 135 bdrv_block_status(BlockDriverState *bs, int64_t offset, int64_t bytes, 136 int64_t *pnum, int64_t *map, BlockDriverState **file); 137 138 int coroutine_fn GRAPH_RDLOCK 139 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base, 140 int64_t offset, int64_t bytes, int64_t *pnum, 141 int64_t *map, BlockDriverState **file); 142 int co_wrapper_mixed_bdrv_rdlock 143 bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base, 144 int64_t offset, int64_t bytes, int64_t *pnum, 145 int64_t *map, BlockDriverState **file); 146 147 int coroutine_fn GRAPH_RDLOCK 148 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 149 int64_t *pnum); 150 int co_wrapper_mixed_bdrv_rdlock 151 bdrv_is_allocated(BlockDriverState *bs, int64_t offset, 152 int64_t bytes, int64_t *pnum); 153 154 int coroutine_fn GRAPH_RDLOCK 155 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 156 bool include_base, int64_t offset, int64_t bytes, 157 int64_t *pnum); 158 int co_wrapper_mixed_bdrv_rdlock 159 bdrv_is_allocated_above(BlockDriverState *bs, BlockDriverState *base, 160 bool include_base, int64_t offset, 161 int64_t bytes, int64_t *pnum); 162 163 int coroutine_fn GRAPH_RDLOCK 164 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes); 165 166 int GRAPH_RDLOCK 167 bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg, 168 Error **errp); 169 170 bool bdrv_is_read_only(BlockDriverState *bs); 171 bool bdrv_is_writable(BlockDriverState *bs); 172 bool bdrv_is_sg(BlockDriverState *bs); 173 int bdrv_get_flags(BlockDriverState *bs); 174 175 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs); 176 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs); 177 178 void coroutine_fn GRAPH_RDLOCK 179 bdrv_co_lock_medium(BlockDriverState *bs, bool locked); 180 181 void coroutine_fn GRAPH_RDLOCK 182 bdrv_co_eject(BlockDriverState *bs, bool eject_flag); 183 184 const char *bdrv_get_format_name(BlockDriverState *bs); 185 186 bool bdrv_supports_compressed_writes(BlockDriverState *bs); 187 const char *bdrv_get_node_name(const BlockDriverState *bs); 188 189 const char * GRAPH_RDLOCK 190 bdrv_get_device_name(const BlockDriverState *bs); 191 192 const char * GRAPH_RDLOCK 193 bdrv_get_device_or_node_name(const BlockDriverState *bs); 194 195 int coroutine_fn GRAPH_RDLOCK 196 bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 197 198 int co_wrapper_mixed_bdrv_rdlock 199 bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 200 201 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, 202 Error **errp); 203 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs); 204 void bdrv_round_to_subclusters(BlockDriverState *bs, 205 int64_t offset, int64_t bytes, 206 int64_t *cluster_offset, 207 int64_t *cluster_bytes); 208 209 void bdrv_get_backing_filename(BlockDriverState *bs, 210 char *filename, int filename_size); 211 212 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, 213 int64_t pos, int size); 214 215 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, 216 int64_t pos, int size); 217 218 /* 219 * Returns the alignment in bytes that is required so that no bounce buffer 220 * is required throughout the stack 221 */ 222 size_t bdrv_min_mem_align(BlockDriverState *bs); 223 /* Returns optimal alignment in bytes for bounce buffer */ 224 size_t bdrv_opt_mem_align(BlockDriverState *bs); 225 void *qemu_blockalign(BlockDriverState *bs, size_t size); 226 void *qemu_blockalign0(BlockDriverState *bs, size_t size); 227 void *qemu_try_blockalign(BlockDriverState *bs, size_t size); 228 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); 229 230 void bdrv_enable_copy_on_read(BlockDriverState *bs); 231 void bdrv_disable_copy_on_read(BlockDriverState *bs); 232 233 void coroutine_fn GRAPH_RDLOCK 234 bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event); 235 236 void co_wrapper_mixed_bdrv_rdlock 237 bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); 238 239 #define BLKDBG_CO_EVENT(child, evt) \ 240 do { \ 241 if (child) { \ 242 bdrv_co_debug_event(child->bs, evt); \ 243 } \ 244 } while (0) 245 246 #define BLKDBG_EVENT(child, evt) \ 247 do { \ 248 if (child) { \ 249 bdrv_debug_event(child->bs, evt); \ 250 } \ 251 } while (0) 252 253 /** 254 * bdrv_get_aio_context: 255 * 256 * Returns: the currently bound #AioContext 257 */ 258 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 259 260 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c); 261 262 /** 263 * Move the current coroutine to the AioContext of @bs and return the old 264 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 265 * will wait for the operation to proceed until the corresponding 266 * bdrv_co_leave(). 267 * 268 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 269 * this will deadlock. 270 */ 271 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 272 273 /** 274 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 275 * to old_ctx and decrease bs->in_flight again. 276 */ 277 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 278 279 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c); 280 281 bool coroutine_fn GRAPH_RDLOCK 282 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 283 uint32_t granularity, Error **errp); 284 bool co_wrapper_bdrv_rdlock 285 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 286 uint32_t granularity, Error **errp); 287 288 /** 289 * 290 * bdrv_co_copy_range: 291 * 292 * Do offloaded copy between two children. If the operation is not implemented 293 * by the driver, or if the backend storage doesn't support it, a negative 294 * error code will be returned. 295 * 296 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 297 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 298 * calling copy_file_range(2)) after the first error, thus it should fall back 299 * to a read+write path in the caller level. 300 * 301 * @src: Source child to copy data from 302 * @src_offset: offset in @src image to read data 303 * @dst: Destination child to copy data to 304 * @dst_offset: offset in @dst image to write data 305 * @bytes: number of bytes to copy 306 * @flags: request flags. Supported flags: 307 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 308 * write on @dst as if bdrv_co_pwrite_zeroes is 309 * called. Used to simplify caller code, or 310 * during BlockDriver.bdrv_co_copy_range_from() 311 * recursion. 312 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 313 * requests currently in flight. 314 * 315 * Returns: 0 if succeeded; negative error code if failed. 316 **/ 317 int coroutine_fn GRAPH_RDLOCK 318 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 319 BdrvChild *dst, int64_t dst_offset, 320 int64_t bytes, BdrvRequestFlags read_flags, 321 BdrvRequestFlags write_flags); 322 323 /* 324 * "I/O or GS" API functions. These functions can run without 325 * the BQL, but only in one specific iothread/main loop. 326 * 327 * More specifically, these functions use BDRV_POLL_WHILE(bs), which 328 * requires the caller to be either in the main thread and hold 329 * the BlockdriverState (bs) AioContext lock, or directly in the 330 * home thread that runs the bs AioContext. Calling them from 331 * another thread in another AioContext would cause deadlocks. 332 * 333 * Therefore, these functions are not proper I/O, because they 334 * can't run in *any* iothreads, but only in a specific one. 335 * 336 * These functions can call any function from I/O, Common and this 337 * categories, but must be invoked only by other "I/O or GS" and GS APIs. 338 * 339 * All functions in this category must use the macro 340 * IO_OR_GS_CODE(); 341 * to catch when they are accidentally called by the wrong API. 342 */ 343 344 #define BDRV_POLL_WHILE(bs, cond) ({ \ 345 BlockDriverState *bs_ = (bs); \ 346 IO_OR_GS_CODE(); \ 347 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 348 cond); }) 349 350 void bdrv_drain(BlockDriverState *bs); 351 352 int co_wrapper_mixed_bdrv_rdlock 353 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 354 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 355 356 int co_wrapper_mixed_bdrv_rdlock 357 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix); 358 359 /* Invalidate any cached metadata used by image formats */ 360 int co_wrapper_mixed_bdrv_rdlock 361 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp); 362 363 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs); 364 365 int co_wrapper_mixed_bdrv_rdlock 366 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes); 367 368 int co_wrapper_mixed_bdrv_rdlock 369 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 370 371 int co_wrapper_mixed_bdrv_rdlock 372 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 373 374 /** 375 * bdrv_parent_drained_begin_single: 376 * 377 * Begin a quiesced section for the parent of @c. 378 */ 379 void GRAPH_RDLOCK bdrv_parent_drained_begin_single(BdrvChild *c); 380 381 /** 382 * bdrv_parent_drained_poll_single: 383 * 384 * Returns true if there is any pending activity to cease before @c can be 385 * called quiesced, false otherwise. 386 */ 387 bool GRAPH_RDLOCK bdrv_parent_drained_poll_single(BdrvChild *c); 388 389 /** 390 * bdrv_parent_drained_end_single: 391 * 392 * End a quiesced section for the parent of @c. 393 */ 394 void GRAPH_RDLOCK bdrv_parent_drained_end_single(BdrvChild *c); 395 396 /** 397 * bdrv_drain_poll: 398 * 399 * Poll for pending requests in @bs and its parents (except for @ignore_parent). 400 * 401 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 402 * ignore the drain request because they will be drained separately (used for 403 * drain_all). 404 * 405 * This is part of bdrv_drained_begin. 406 */ 407 bool GRAPH_RDLOCK 408 bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent, 409 bool ignore_bds_parents); 410 411 /** 412 * bdrv_drained_begin: 413 * 414 * Begin a quiesced section for exclusive access to the BDS, by disabling 415 * external request sources including NBD server, block jobs, and device model. 416 * 417 * This function can only be invoked by the main loop or a coroutine 418 * (regardless of the AioContext where it is running). 419 * If the coroutine is running in an Iothread AioContext, this function will 420 * just schedule a BH to run in the main loop. 421 * However, it cannot be directly called by an Iothread. 422 * 423 * This function can be recursive. 424 */ 425 void bdrv_drained_begin(BlockDriverState *bs); 426 427 /** 428 * bdrv_do_drained_begin_quiesce: 429 * 430 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 431 * running requests to complete. 432 */ 433 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent); 434 435 /** 436 * bdrv_drained_end: 437 * 438 * End a quiescent section started by bdrv_drained_begin(). 439 * 440 * This function can only be invoked by the main loop or a coroutine 441 * (regardless of the AioContext where it is running). 442 * If the coroutine is running in an Iothread AioContext, this function will 443 * just schedule a BH to run in the main loop. 444 * However, it cannot be directly called by an Iothread. 445 */ 446 void bdrv_drained_end(BlockDriverState *bs); 447 448 #endif /* BLOCK_IO_H */ 449