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 bdrv_co_pwrite_sync(BdrvChild *child, int64_t offset, 64 int64_t bytes, const void *buf, 65 BdrvRequestFlags flags); 66 /* 67 * Efficiently zero a region of the disk image. Note that this is a regular 68 * I/O request like read or write and should have a reasonable size. This 69 * function is not suitable for zeroing the entire image in a single request 70 * because it may allocate memory for the entire region. 71 */ 72 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset, 73 int64_t bytes, BdrvRequestFlags flags); 74 75 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, 76 PreallocMode prealloc, BdrvRequestFlags flags, 77 Error **errp); 78 79 int64_t coroutine_fn bdrv_co_nb_sectors(BlockDriverState *bs); 80 int64_t co_wrapper_mixed bdrv_nb_sectors(BlockDriverState *bs); 81 82 int64_t coroutine_fn bdrv_co_getlength(BlockDriverState *bs); 83 int64_t co_wrapper_mixed bdrv_getlength(BlockDriverState *bs); 84 85 int64_t coroutine_fn bdrv_co_get_allocated_file_size(BlockDriverState *bs); 86 int64_t co_wrapper bdrv_get_allocated_file_size(BlockDriverState *bs); 87 88 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts, 89 BlockDriverState *in_bs, Error **errp); 90 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr); 91 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp); 92 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs); 93 94 95 /* async block I/O */ 96 void bdrv_aio_cancel(BlockAIOCB *acb); 97 void bdrv_aio_cancel_async(BlockAIOCB *acb); 98 99 /* sg packet commands */ 100 int coroutine_fn bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf); 101 102 /* Ensure contents are flushed to disk. */ 103 int coroutine_fn bdrv_co_flush(BlockDriverState *bs); 104 105 int coroutine_fn bdrv_co_pdiscard(BdrvChild *child, int64_t offset, 106 int64_t bytes); 107 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs); 108 int bdrv_block_status(BlockDriverState *bs, int64_t offset, 109 int64_t bytes, int64_t *pnum, int64_t *map, 110 BlockDriverState **file); 111 112 int coroutine_fn bdrv_co_block_status_above(BlockDriverState *bs, 113 BlockDriverState *base, 114 int64_t offset, int64_t bytes, 115 int64_t *pnum, int64_t *map, 116 BlockDriverState **file); 117 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base, 118 int64_t offset, int64_t bytes, int64_t *pnum, 119 int64_t *map, BlockDriverState **file); 120 121 int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, 122 int64_t bytes, int64_t *pnum); 123 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 124 int64_t *pnum); 125 126 int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, 127 BlockDriverState *base, 128 bool include_base, int64_t offset, 129 int64_t bytes, int64_t *pnum); 130 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 131 bool include_base, int64_t offset, int64_t bytes, 132 int64_t *pnum); 133 134 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, 135 int64_t bytes); 136 137 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only, 138 bool ignore_allow_rdw, Error **errp); 139 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg, 140 Error **errp); 141 bool bdrv_is_read_only(BlockDriverState *bs); 142 bool bdrv_is_writable(BlockDriverState *bs); 143 bool bdrv_is_sg(BlockDriverState *bs); 144 int bdrv_get_flags(BlockDriverState *bs); 145 146 bool coroutine_fn bdrv_co_is_inserted(BlockDriverState *bs); 147 bool co_wrapper bdrv_is_inserted(BlockDriverState *bs); 148 149 void bdrv_lock_medium(BlockDriverState *bs, bool locked); 150 void coroutine_fn bdrv_co_eject(BlockDriverState *bs, bool eject_flag); 151 152 const char *bdrv_get_format_name(BlockDriverState *bs); 153 154 bool bdrv_supports_compressed_writes(BlockDriverState *bs); 155 const char *bdrv_get_node_name(const BlockDriverState *bs); 156 const char *bdrv_get_device_name(const BlockDriverState *bs); 157 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs); 158 159 int coroutine_fn bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 160 int co_wrapper_mixed bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 161 162 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, 163 Error **errp); 164 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs); 165 void bdrv_round_to_clusters(BlockDriverState *bs, 166 int64_t offset, int64_t bytes, 167 int64_t *cluster_offset, 168 int64_t *cluster_bytes); 169 170 void bdrv_get_backing_filename(BlockDriverState *bs, 171 char *filename, int filename_size); 172 173 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, 174 int64_t pos, int size); 175 176 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, 177 int64_t pos, int size); 178 179 /* 180 * Returns the alignment in bytes that is required so that no bounce buffer 181 * is required throughout the stack 182 */ 183 size_t bdrv_min_mem_align(BlockDriverState *bs); 184 /* Returns optimal alignment in bytes for bounce buffer */ 185 size_t bdrv_opt_mem_align(BlockDriverState *bs); 186 void *qemu_blockalign(BlockDriverState *bs, size_t size); 187 void *qemu_blockalign0(BlockDriverState *bs, size_t size); 188 void *qemu_try_blockalign(BlockDriverState *bs, size_t size); 189 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); 190 191 void bdrv_enable_copy_on_read(BlockDriverState *bs); 192 void bdrv_disable_copy_on_read(BlockDriverState *bs); 193 194 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); 195 196 #define BLKDBG_EVENT(child, evt) \ 197 do { \ 198 if (child) { \ 199 bdrv_debug_event(child->bs, evt); \ 200 } \ 201 } while (0) 202 203 /** 204 * bdrv_get_aio_context: 205 * 206 * Returns: the currently bound #AioContext 207 */ 208 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 209 210 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c); 211 212 /** 213 * Move the current coroutine to the AioContext of @bs and return the old 214 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 215 * will wait for the operation to proceed until the corresponding 216 * bdrv_co_leave(). 217 * 218 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 219 * this will deadlock. 220 */ 221 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 222 223 /** 224 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 225 * to old_ctx and decrease bs->in_flight again. 226 */ 227 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 228 229 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c); 230 231 void coroutine_fn bdrv_co_io_plug(BlockDriverState *bs); 232 void coroutine_fn bdrv_co_io_unplug(BlockDriverState *bs); 233 234 bool coroutine_fn bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, 235 const char *name, 236 uint32_t granularity, 237 Error **errp); 238 bool co_wrapper bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, 239 const char *name, 240 uint32_t granularity, 241 Error **errp); 242 243 /** 244 * 245 * bdrv_co_copy_range: 246 * 247 * Do offloaded copy between two children. If the operation is not implemented 248 * by the driver, or if the backend storage doesn't support it, a negative 249 * error code will be returned. 250 * 251 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 252 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 253 * calling copy_file_range(2)) after the first error, thus it should fall back 254 * to a read+write path in the caller level. 255 * 256 * @src: Source child to copy data from 257 * @src_offset: offset in @src image to read data 258 * @dst: Destination child to copy data to 259 * @dst_offset: offset in @dst image to write data 260 * @bytes: number of bytes to copy 261 * @flags: request flags. Supported flags: 262 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 263 * write on @dst as if bdrv_co_pwrite_zeroes is 264 * called. Used to simplify caller code, or 265 * during BlockDriver.bdrv_co_copy_range_from() 266 * recursion. 267 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 268 * requests currently in flight. 269 * 270 * Returns: 0 if succeeded; negative error code if failed. 271 **/ 272 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 273 BdrvChild *dst, int64_t dst_offset, 274 int64_t bytes, BdrvRequestFlags read_flags, 275 BdrvRequestFlags write_flags); 276 277 /* 278 * "I/O or GS" API functions. These functions can run without 279 * the BQL, but only in one specific iothread/main loop. 280 * 281 * More specifically, these functions use BDRV_POLL_WHILE(bs), which 282 * requires the caller to be either in the main thread and hold 283 * the BlockdriverState (bs) AioContext lock, or directly in the 284 * home thread that runs the bs AioContext. Calling them from 285 * another thread in another AioContext would cause deadlocks. 286 * 287 * Therefore, these functions are not proper I/O, because they 288 * can't run in *any* iothreads, but only in a specific one. 289 * 290 * These functions can call any function from I/O, Common and this 291 * categories, but must be invoked only by other "I/O or GS" and GS APIs. 292 * 293 * All functions in this category must use the macro 294 * IO_OR_GS_CODE(); 295 * to catch when they are accidentally called by the wrong API. 296 */ 297 298 #define BDRV_POLL_WHILE(bs, cond) ({ \ 299 BlockDriverState *bs_ = (bs); \ 300 IO_OR_GS_CODE(); \ 301 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 302 cond); }) 303 304 void bdrv_drain(BlockDriverState *bs); 305 306 int co_wrapper_mixed_bdrv_rdlock 307 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 308 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 309 310 int co_wrapper_mixed_bdrv_rdlock 311 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix); 312 313 /* Invalidate any cached metadata used by image formats */ 314 int co_wrapper_mixed_bdrv_rdlock 315 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp); 316 317 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs); 318 319 int co_wrapper_mixed_bdrv_rdlock 320 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes); 321 322 int co_wrapper_mixed_bdrv_rdlock 323 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 324 325 int co_wrapper_mixed_bdrv_rdlock 326 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 327 328 /** 329 * bdrv_parent_drained_begin_single: 330 * 331 * Begin a quiesced section for the parent of @c. 332 */ 333 void bdrv_parent_drained_begin_single(BdrvChild *c); 334 335 /** 336 * bdrv_parent_drained_poll_single: 337 * 338 * Returns true if there is any pending activity to cease before @c can be 339 * called quiesced, false otherwise. 340 */ 341 bool bdrv_parent_drained_poll_single(BdrvChild *c); 342 343 /** 344 * bdrv_parent_drained_end_single: 345 * 346 * End a quiesced section for the parent of @c. 347 */ 348 void bdrv_parent_drained_end_single(BdrvChild *c); 349 350 /** 351 * bdrv_drain_poll: 352 * 353 * Poll for pending requests in @bs and its parents (except for @ignore_parent). 354 * 355 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 356 * ignore the drain request because they will be drained separately (used for 357 * drain_all). 358 * 359 * This is part of bdrv_drained_begin. 360 */ 361 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent, 362 bool ignore_bds_parents); 363 364 /** 365 * bdrv_drained_begin: 366 * 367 * Begin a quiesced section for exclusive access to the BDS, by disabling 368 * external request sources including NBD server, block jobs, and device model. 369 * 370 * This function can be recursive. 371 */ 372 void bdrv_drained_begin(BlockDriverState *bs); 373 374 /** 375 * bdrv_do_drained_begin_quiesce: 376 * 377 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 378 * running requests to complete. 379 */ 380 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent); 381 382 /** 383 * bdrv_drained_end: 384 * 385 * End a quiescent section started by bdrv_drained_begin(). 386 */ 387 void bdrv_drained_end(BlockDriverState *bs); 388 389 #endif /* BLOCK_IO_H */ 390