include/block/block-io.h

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