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