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-common.h" 28 29 /* 30 * I/O API functions. These functions are thread-safe, and therefore 31 * can run in any thread as long as the thread has called 32 * aio_context_acquire/release(). 33 * 34 * These functions can only call functions from I/O and Common categories, 35 * but can be invoked by GS, "I/O or GS" and I/O APIs. 36 * 37 * All functions in this category must use the macro 38 * IO_CODE(); 39 * to catch when they are accidentally called by the wrong API. 40 */ 41 42 int co_wrapper_mixed_bdrv_rdlock 43 bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset, int64_t bytes, 44 BdrvRequestFlags flags); 45 46 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags); 47 48 int co_wrapper_mixed_bdrv_rdlock 49 bdrv_pread(BdrvChild *child, int64_t offset, int64_t bytes, void *buf, 50 BdrvRequestFlags flags); 51 52 int co_wrapper_mixed_bdrv_rdlock 53 bdrv_pwrite(BdrvChild *child, int64_t offset,int64_t bytes, 54 const void *buf, BdrvRequestFlags flags); 55 56 int co_wrapper_mixed_bdrv_rdlock 57 bdrv_pwrite_sync(BdrvChild *child, int64_t offset, int64_t bytes, 58 const void *buf, BdrvRequestFlags flags); 59 60 int coroutine_fn bdrv_co_pwrite_sync(BdrvChild *child, int64_t offset, 61 int64_t bytes, const void *buf, 62 BdrvRequestFlags flags); 63 /* 64 * Efficiently zero a region of the disk image. Note that this is a regular 65 * I/O request like read or write and should have a reasonable size. This 66 * function is not suitable for zeroing the entire image in a single request 67 * because it may allocate memory for the entire region. 68 */ 69 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset, 70 int64_t bytes, BdrvRequestFlags flags); 71 72 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, 73 PreallocMode prealloc, BdrvRequestFlags flags, 74 Error **errp); 75 76 int64_t bdrv_nb_sectors(BlockDriverState *bs); 77 int64_t bdrv_getlength(BlockDriverState *bs); 78 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs); 79 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts, 80 BlockDriverState *in_bs, Error **errp); 81 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr); 82 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp); 83 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs); 84 85 86 /* async block I/O */ 87 void bdrv_aio_cancel(BlockAIOCB *acb); 88 void bdrv_aio_cancel_async(BlockAIOCB *acb); 89 90 /* sg packet commands */ 91 int coroutine_fn bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf); 92 93 /* Ensure contents are flushed to disk. */ 94 int coroutine_fn bdrv_co_flush(BlockDriverState *bs); 95 96 int coroutine_fn bdrv_co_pdiscard(BdrvChild *child, int64_t offset, 97 int64_t bytes); 98 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs); 99 int bdrv_block_status(BlockDriverState *bs, int64_t offset, 100 int64_t bytes, int64_t *pnum, int64_t *map, 101 BlockDriverState **file); 102 103 int coroutine_fn bdrv_co_block_status_above(BlockDriverState *bs, 104 BlockDriverState *base, 105 int64_t offset, int64_t bytes, 106 int64_t *pnum, int64_t *map, 107 BlockDriverState **file); 108 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base, 109 int64_t offset, int64_t bytes, int64_t *pnum, 110 int64_t *map, BlockDriverState **file); 111 112 int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t offset, 113 int64_t bytes, int64_t *pnum); 114 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 115 int64_t *pnum); 116 117 int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, 118 BlockDriverState *base, 119 bool include_base, int64_t offset, 120 int64_t bytes, int64_t *pnum); 121 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 122 bool include_base, int64_t offset, int64_t bytes, 123 int64_t *pnum); 124 125 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, 126 int64_t bytes); 127 128 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only, 129 bool ignore_allow_rdw, Error **errp); 130 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg, 131 Error **errp); 132 bool bdrv_is_read_only(BlockDriverState *bs); 133 bool bdrv_is_writable(BlockDriverState *bs); 134 bool bdrv_is_sg(BlockDriverState *bs); 135 int bdrv_get_flags(BlockDriverState *bs); 136 bool bdrv_is_inserted(BlockDriverState *bs); 137 void bdrv_lock_medium(BlockDriverState *bs, bool locked); 138 void bdrv_eject(BlockDriverState *bs, bool eject_flag); 139 const char *bdrv_get_format_name(BlockDriverState *bs); 140 141 bool bdrv_supports_compressed_writes(BlockDriverState *bs); 142 const char *bdrv_get_node_name(const BlockDriverState *bs); 143 const char *bdrv_get_device_name(const BlockDriverState *bs); 144 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs); 145 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 146 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, 147 Error **errp); 148 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs); 149 void bdrv_round_to_clusters(BlockDriverState *bs, 150 int64_t offset, int64_t bytes, 151 int64_t *cluster_offset, 152 int64_t *cluster_bytes); 153 154 void bdrv_get_backing_filename(BlockDriverState *bs, 155 char *filename, int filename_size); 156 157 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, 158 int64_t pos, int size); 159 160 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, 161 int64_t pos, int size); 162 163 /* 164 * Returns the alignment in bytes that is required so that no bounce buffer 165 * is required throughout the stack 166 */ 167 size_t bdrv_min_mem_align(BlockDriverState *bs); 168 /* Returns optimal alignment in bytes for bounce buffer */ 169 size_t bdrv_opt_mem_align(BlockDriverState *bs); 170 void *qemu_blockalign(BlockDriverState *bs, size_t size); 171 void *qemu_blockalign0(BlockDriverState *bs, size_t size); 172 void *qemu_try_blockalign(BlockDriverState *bs, size_t size); 173 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); 174 175 void bdrv_enable_copy_on_read(BlockDriverState *bs); 176 void bdrv_disable_copy_on_read(BlockDriverState *bs); 177 178 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); 179 180 #define BLKDBG_EVENT(child, evt) \ 181 do { \ 182 if (child) { \ 183 bdrv_debug_event(child->bs, evt); \ 184 } \ 185 } while (0) 186 187 /** 188 * bdrv_get_aio_context: 189 * 190 * Returns: the currently bound #AioContext 191 */ 192 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 193 194 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c); 195 196 /** 197 * Move the current coroutine to the AioContext of @bs and return the old 198 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 199 * will wait for the operation to proceed until the corresponding 200 * bdrv_co_leave(). 201 * 202 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 203 * this will deadlock. 204 */ 205 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 206 207 /** 208 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 209 * to old_ctx and decrease bs->in_flight again. 210 */ 211 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 212 213 /** 214 * Transfer control to @co in the aio context of @bs 215 */ 216 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co); 217 218 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c); 219 220 void bdrv_io_plug(BlockDriverState *bs); 221 void bdrv_io_unplug(BlockDriverState *bs); 222 223 bool coroutine_fn bdrv_co_can_store_new_dirty_bitmap(BlockDriverState *bs, 224 const char *name, 225 uint32_t granularity, 226 Error **errp); 227 bool co_wrapper bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, 228 const char *name, 229 uint32_t granularity, 230 Error **errp); 231 232 /** 233 * 234 * bdrv_co_copy_range: 235 * 236 * Do offloaded copy between two children. If the operation is not implemented 237 * by the driver, or if the backend storage doesn't support it, a negative 238 * error code will be returned. 239 * 240 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 241 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 242 * calling copy_file_range(2)) after the first error, thus it should fall back 243 * to a read+write path in the caller level. 244 * 245 * @src: Source child to copy data from 246 * @src_offset: offset in @src image to read data 247 * @dst: Destination child to copy data to 248 * @dst_offset: offset in @dst image to write data 249 * @bytes: number of bytes to copy 250 * @flags: request flags. Supported flags: 251 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 252 * write on @dst as if bdrv_co_pwrite_zeroes is 253 * called. Used to simplify caller code, or 254 * during BlockDriver.bdrv_co_copy_range_from() 255 * recursion. 256 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 257 * requests currently in flight. 258 * 259 * Returns: 0 if succeeded; negative error code if failed. 260 **/ 261 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 262 BdrvChild *dst, int64_t dst_offset, 263 int64_t bytes, BdrvRequestFlags read_flags, 264 BdrvRequestFlags write_flags); 265 266 /* 267 * "I/O or GS" API functions. These functions can run without 268 * the BQL, but only in one specific iothread/main loop. 269 * 270 * More specifically, these functions use BDRV_POLL_WHILE(bs), which 271 * requires the caller to be either in the main thread and hold 272 * the BlockdriverState (bs) AioContext lock, or directly in the 273 * home thread that runs the bs AioContext. Calling them from 274 * another thread in another AioContext would cause deadlocks. 275 * 276 * Therefore, these functions are not proper I/O, because they 277 * can't run in *any* iothreads, but only in a specific one. 278 * 279 * These functions can call any function from I/O, Common and this 280 * categories, but must be invoked only by other "I/O or GS" and GS APIs. 281 * 282 * All functions in this category must use the macro 283 * IO_OR_GS_CODE(); 284 * to catch when they are accidentally called by the wrong API. 285 */ 286 287 #define BDRV_POLL_WHILE(bs, cond) ({ \ 288 BlockDriverState *bs_ = (bs); \ 289 IO_OR_GS_CODE(); \ 290 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 291 cond); }) 292 293 void bdrv_drain(BlockDriverState *bs); 294 295 int co_wrapper_mixed_bdrv_rdlock 296 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 297 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 298 299 int co_wrapper_mixed_bdrv_rdlock 300 bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix); 301 302 /* Invalidate any cached metadata used by image formats */ 303 int co_wrapper_mixed_bdrv_rdlock 304 bdrv_invalidate_cache(BlockDriverState *bs, Error **errp); 305 306 int co_wrapper_mixed_bdrv_rdlock bdrv_flush(BlockDriverState *bs); 307 308 int co_wrapper_mixed_bdrv_rdlock 309 bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes); 310 311 int co_wrapper_mixed_bdrv_rdlock 312 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 313 314 int co_wrapper_mixed_bdrv_rdlock 315 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 316 317 /** 318 * bdrv_parent_drained_begin_single: 319 * 320 * Begin a quiesced section for the parent of @c. 321 */ 322 void bdrv_parent_drained_begin_single(BdrvChild *c); 323 324 /** 325 * bdrv_parent_drained_poll_single: 326 * 327 * Returns true if there is any pending activity to cease before @c can be 328 * called quiesced, false otherwise. 329 */ 330 bool bdrv_parent_drained_poll_single(BdrvChild *c); 331 332 /** 333 * bdrv_parent_drained_end_single: 334 * 335 * End a quiesced section for the parent of @c. 336 */ 337 void bdrv_parent_drained_end_single(BdrvChild *c); 338 339 /** 340 * bdrv_drain_poll: 341 * 342 * Poll for pending requests in @bs and its parents (except for @ignore_parent). 343 * 344 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 345 * ignore the drain request because they will be drained separately (used for 346 * drain_all). 347 * 348 * This is part of bdrv_drained_begin. 349 */ 350 bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent, 351 bool ignore_bds_parents); 352 353 /** 354 * bdrv_drained_begin: 355 * 356 * Begin a quiesced section for exclusive access to the BDS, by disabling 357 * external request sources including NBD server, block jobs, and device model. 358 * 359 * This function can be recursive. 360 */ 361 void bdrv_drained_begin(BlockDriverState *bs); 362 363 /** 364 * bdrv_do_drained_begin_quiesce: 365 * 366 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 367 * running requests to complete. 368 */ 369 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, BdrvChild *parent); 370 371 /** 372 * bdrv_drained_end: 373 * 374 * End a quiescent section started by bdrv_drained_begin(). 375 */ 376 void bdrv_drained_end(BlockDriverState *bs); 377 378 #endif /* BLOCK_IO_H */ 379