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 co_wrapper_mixed_bdrv_rdlock 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 bdrv_co_get_allocated_file_size(BlockDriverState *bs); 88 int64_t co_wrapper bdrv_get_allocated_file_size(BlockDriverState *bs); 89 90 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts, 91 BlockDriverState *in_bs, Error **errp); 92 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr); 93 94 int coroutine_fn GRAPH_RDLOCK 95 bdrv_co_delete_file(BlockDriverState *bs, Error **errp); 96 97 void coroutine_fn GRAPH_RDLOCK 98 bdrv_co_delete_file_noerr(BlockDriverState *bs); 99 100 101 /* async block I/O */ 102 void bdrv_aio_cancel(BlockAIOCB *acb); 103 void bdrv_aio_cancel_async(BlockAIOCB *acb); 104 105 /* sg packet commands */ 106 int coroutine_fn GRAPH_RDLOCK 107 bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf); 108 109 /* Ensure contents are flushed to disk. */ 110 int coroutine_fn GRAPH_RDLOCK bdrv_co_flush(BlockDriverState *bs); 111 112 int coroutine_fn GRAPH_RDLOCK bdrv_co_pdiscard(BdrvChild *child, int64_t offset, 113 int64_t bytes); 114 115 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs); 116 int bdrv_block_status(BlockDriverState *bs, int64_t offset, 117 int64_t bytes, int64_t *pnum, int64_t *map, 118 BlockDriverState **file); 119 120 int coroutine_fn GRAPH_RDLOCK 121 bdrv_co_block_status_above(BlockDriverState *bs, BlockDriverState *base, 122 int64_t offset, int64_t bytes, int64_t *pnum, 123 int64_t *map, BlockDriverState **file); 124 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base, 125 int64_t offset, int64_t bytes, int64_t *pnum, 126 int64_t *map, BlockDriverState **file); 127 128 int coroutine_fn GRAPH_RDLOCK 129 bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 130 int64_t *pnum); 131 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 132 int64_t *pnum); 133 134 int coroutine_fn GRAPH_RDLOCK 135 bdrv_co_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 136 bool include_base, int64_t offset, int64_t bytes, 137 int64_t *pnum); 138 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 139 bool include_base, int64_t offset, int64_t bytes, 140 int64_t *pnum); 141 142 int coroutine_fn GRAPH_RDLOCK 143 bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, int64_t bytes); 144 145 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg, 146 Error **errp); 147 bool bdrv_is_read_only(BlockDriverState *bs); 148 bool bdrv_is_writable(BlockDriverState *bs); 149 bool bdrv_is_sg(BlockDriverState *bs); 150 int bdrv_get_flags(BlockDriverState *bs); 151 152 bool coroutine_fn GRAPH_RDLOCK bdrv_co_is_inserted(BlockDriverState *bs); 153 bool co_wrapper_bdrv_rdlock bdrv_is_inserted(BlockDriverState *bs); 154 155 void coroutine_fn GRAPH_RDLOCK 156 bdrv_co_lock_medium(BlockDriverState *bs, bool locked); 157 158 void coroutine_fn GRAPH_RDLOCK 159 bdrv_co_eject(BlockDriverState *bs, bool eject_flag); 160 161 const char *bdrv_get_format_name(BlockDriverState *bs); 162 163 bool bdrv_supports_compressed_writes(BlockDriverState *bs); 164 const char *bdrv_get_node_name(const BlockDriverState *bs); 165 const char *bdrv_get_device_name(const BlockDriverState *bs); 166 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs); 167 168 int coroutine_fn bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 169 int co_wrapper_mixed bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 170 171 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, 172 Error **errp); 173 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs); 174 void bdrv_round_to_clusters(BlockDriverState *bs, 175 int64_t offset, int64_t bytes, 176 int64_t *cluster_offset, 177 int64_t *cluster_bytes); 178 179 void bdrv_get_backing_filename(BlockDriverState *bs, 180 char *filename, int filename_size); 181 182 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, 183 int64_t pos, int size); 184 185 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, 186 int64_t pos, int size); 187 188 /* 189 * Returns the alignment in bytes that is required so that no bounce buffer 190 * is required throughout the stack 191 */ 192 size_t bdrv_min_mem_align(BlockDriverState *bs); 193 /* Returns optimal alignment in bytes for bounce buffer */ 194 size_t bdrv_opt_mem_align(BlockDriverState *bs); 195 void *qemu_blockalign(BlockDriverState *bs, size_t size); 196 void *qemu_blockalign0(BlockDriverState *bs, size_t size); 197 void *qemu_try_blockalign(BlockDriverState *bs, size_t size); 198 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); 199 200 void bdrv_enable_copy_on_read(BlockDriverState *bs); 201 void bdrv_disable_copy_on_read(BlockDriverState *bs); 202 203 void coroutine_fn bdrv_co_debug_event(BlockDriverState *bs, 204 BlkdebugEvent event); 205 void co_wrapper_mixed bdrv_debug_event(BlockDriverState *bs, 206 BlkdebugEvent event); 207 208 #define BLKDBG_EVENT(child, evt) \ 209 do { \ 210 if (child) { \ 211 bdrv_debug_event(child->bs, evt); \ 212 } \ 213 } while (0) 214 215 /** 216 * bdrv_get_aio_context: 217 * 218 * Returns: the currently bound #AioContext 219 */ 220 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 221 222 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c); 223 224 /** 225 * Move the current coroutine to the AioContext of @bs and return the old 226 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 227 * will wait for the operation to proceed until the corresponding 228 * bdrv_co_leave(). 229 * 230 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 231 * this will deadlock. 232 */ 233 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 234 235 /** 236 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 237 * to old_ctx and decrease bs->in_flight again. 238 */ 239 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 240 241 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c); 242 243 void coroutine_fn GRAPH_RDLOCK bdrv_co_io_plug(BlockDriverState *bs); 244 void coroutine_fn GRAPH_RDLOCK bdrv_co_io_unplug(BlockDriverState *bs); 245 246 bool coroutine_fn GRAPH_RDLOCK 247 bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 248 uint32_t granularity, Error **errp); 249 bool co_wrapper_bdrv_rdlock 250 bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 251 uint32_t granularity, Error **errp); 252 253 /** 254 * 255 * bdrv_co_copy_range: 256 * 257 * Do offloaded copy between two children. If the operation is not implemented 258 * by the driver, or if the backend storage doesn't support it, a negative 259 * error code will be returned. 260 * 261 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 262 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 263 * calling copy_file_range(2)) after the first error, thus it should fall back 264 * to a read+write path in the caller level. 265 * 266 * @src: Source child to copy data from 267 * @src_offset: offset in @src image to read data 268 * @dst: Destination child to copy data to 269 * @dst_offset: offset in @dst image to write data 270 * @bytes: number of bytes to copy 271 * @flags: request flags. Supported flags: 272 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 273 * write on @dst as if bdrv_co_pwrite_zeroes is 274 * called. Used to simplify caller code, or 275 * during BlockDriver.bdrv_co_copy_range_from() 276 * recursion. 277 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 278 * requests currently in flight. 279 * 280 * Returns: 0 if succeeded; negative error code if failed. 281 **/ 282 int coroutine_fn GRAPH_RDLOCK 283 bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 284 BdrvChild *dst, int64_t dst_offset, 285 int64_t bytes, BdrvRequestFlags read_flags, 286 BdrvRequestFlags write_flags); 287 288 /* 289 * "I/O or GS" API functions. These functions can run without 290 * the BQL, but only in one specific iothread/main loop. 291 * 292 * More specifically, these functions use BDRV_POLL_WHILE(bs), which 293 * requires the caller to be either in the main thread and hold 294 * the BlockdriverState (bs) AioContext lock, or directly in the 295 * home thread that runs the bs AioContext. Calling them from 296 * another thread in another AioContext would cause deadlocks. 297 * 298 * Therefore, these functions are not proper I/O, because they 299 * can't run in *any* iothreads, but only in a specific one. 300 * 301 * These functions can call any function from I/O, Common and this 302 * categories, but must be invoked only by other "I/O or GS" and GS APIs. 303 * 304 * All functions in this category must use the macro 305 * IO_OR_GS_CODE(); 306 * to catch when they are accidentally called by the wrong API. 307 */ 308 309 #define BDRV_POLL_WHILE(bs, cond) ({ \ 310 BlockDriverState *bs_ = (bs); \ 311 IO_OR_GS_CODE(); \ 312 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 313 cond); }) 314 315 void bdrv_drain(BlockDriverState *bs); 316 317 int co_wrapper_mixed_bdrv_rdlock 318 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 319 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 320 321 int co_wrapper_mixed_bdrv_rdlock 322 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix); 323 324 /* Invalidate any cached metadata used by image formats */ 325 int co_wrapper_mixed_bdrv_rdlock 326 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp); 327 328 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs); 329 330 int co_wrapper_mixed_bdrv_rdlock 331 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes); 332 333 int co_wrapper_mixed_bdrv_rdlock 334 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 335 336 int co_wrapper_mixed_bdrv_rdlock 337 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 338 339 /** 340 * bdrv_parent_drained_begin_single: 341 * 342 * Begin a quiesced section for the parent of @c. 343 */ 344 void bdrv_parent_drained_begin_single(BdrvChild *c); 345 346 /** 347 * bdrv_parent_drained_poll_single: 348 * 349 * Returns true if there is any pending activity to cease before @c can be 350 * called quiesced, false otherwise. 351 */ 352 bool bdrv_parent_drained_poll_single(BdrvChild *c); 353 354 /** 355 * bdrv_parent_drained_end_single: 356 * 357 * End a quiesced section for the parent of @c. 358 */ 359 void bdrv_parent_drained_end_single(BdrvChild *c); 360 361 /** 362 * bdrv_drain_poll: 363 * 364 * Poll for pending requests in @bs and its parents (except for @ignore_parent). 365 * 366 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 367 * ignore the drain request because they will be drained separately (used for 368 * drain_all). 369 * 370 * This is part of bdrv_drained_begin. 371 */ 372 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent, 373 bool ignore_bds_parents); 374 375 /** 376 * bdrv_drained_begin: 377 * 378 * Begin a quiesced section for exclusive access to the BDS, by disabling 379 * external request sources including NBD server, block jobs, and device model. 380 * 381 * This function can be recursive. 382 */ 383 void bdrv_drained_begin(BlockDriverState *bs); 384 385 /** 386 * bdrv_do_drained_begin_quiesce: 387 * 388 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 389 * running requests to complete. 390 */ 391 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent); 392 393 /** 394 * bdrv_drained_end: 395 * 396 * End a quiescent section started by bdrv_drained_begin(). 397 */ 398 void bdrv_drained_end(BlockDriverState *bs); 399 400 #endif /* BLOCK_IO_H */ 401