include/block/block-common.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_COMMON_H
  25 #define BLOCK_COMMON_H
  26
  27 #include "block/aio.h"
  28 #include "block/aio-wait.h"
  29 #include "qemu/iov.h"
  30 #include "qemu/coroutine.h"
  31 #include "block/accounting.h"
  32 #include "block/dirty-bitmap.h"
  33 #include "block/blockjob.h"
  34 #include "qemu/hbitmap.h"
  35 #include "qemu/transactions.h"
  36
  37 /*
  38  * generated_co_wrapper
  39  *
  40  * Function specifier, which does nothing but mark functions to be
  41  * generated by scripts/block-coroutine-wrapper.py
  42  *
  43  * Read more in docs/devel/block-coroutine-wrapper.rst
  44  */
  45 #define generated_co_wrapper
  46
  47 /* block.c */
  48 typedef struct BlockDriver BlockDriver;
  49 typedef struct BdrvChild BdrvChild;
  50 typedef struct BdrvChildClass BdrvChildClass;
  51
  52 typedef struct BlockDriverInfo {
  53     /* in bytes, 0 if irrelevant */
  54     int cluster_size;
  55     /* offset at which the VM state can be saved (0 if not possible) */
  56     int64_t vm_state_offset;
  57     bool is_dirty;
  58     /*
  59      * True if this block driver only supports compressed writes
  60      */
  61     bool needs_compressed_writes;
  62 } BlockDriverInfo;
  63
  64 typedef struct BlockFragInfo {
  65     uint64_t allocated_clusters;
  66     uint64_t total_clusters;
  67     uint64_t fragmented_clusters;
  68     uint64_t compressed_clusters;
  69 } BlockFragInfo;
  70
  71 typedef enum {
  72     BDRV_REQ_COPY_ON_READ       = 0x1,
  73     BDRV_REQ_ZERO_WRITE         = 0x2,
  74
  75     /*
  76      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
  77      * that the block driver should unmap (discard) blocks if it is guaranteed
  78      * that the result will read back as zeroes. The flag is only passed to the
  79      * driver if the block device is opened with BDRV_O_UNMAP.
  80      */
  81     BDRV_REQ_MAY_UNMAP          = 0x4,
  82
  83     BDRV_REQ_FUA                = 0x10,
  84     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
  85
  86     /*
  87      * Signifies that this write request will not change the visible disk
  88      * content.
  89      */
  90     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
  91
  92     /*
  93      * Forces request serialisation. Use only with write requests.
  94      */
  95     BDRV_REQ_SERIALISING        = 0x80,
  96
  97     /*
  98      * Execute the request only if the operation can be offloaded or otherwise
  99      * be executed efficiently, but return an error instead of using a slow
 100      * fallback.
 101      */
 102     BDRV_REQ_NO_FALLBACK        = 0x100,
 103
 104     /*
 105      * BDRV_REQ_PREFETCH makes sense only in the context of copy-on-read
 106      * (i.e., together with the BDRV_REQ_COPY_ON_READ flag or when a COR
 107      * filter is involved), in which case it signals that the COR operation
 108      * need not read the data into memory (qiov) but only ensure they are
 109      * copied to the top layer (i.e., that COR operation is done).
 110      */
 111     BDRV_REQ_PREFETCH  = 0x200,
 112
 113     /*
 114      * If we need to wait for other requests, just fail immediately. Used
 115      * only together with BDRV_REQ_SERIALISING.
 116      */
 117     BDRV_REQ_NO_WAIT = 0x400,
 118
 119     /* Mask of valid flags */
 120     BDRV_REQ_MASK               = 0x7ff,
 121 } BdrvRequestFlags;
 122
 123 #define BDRV_O_NO_SHARE    0x0001 /* don't share permissions */
 124 #define BDRV_O_RDWR        0x0002
 125 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
 126 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save
 127                                      writes in a snapshot */
 128 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
 129 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
 130 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the
 131                                      thread pool */
 132 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
 133 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
 134 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
 135 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
 136 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
 137 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
 138 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
 139 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
 140                                       select an appropriate protocol driver,
 141                                       ignoring the format layer */
 142 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
 143 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening
 144                                       read-write fails */
 145 #define BDRV_O_IO_URING    0x40000 /* use io_uring instead of the thread pool */
 146
 147 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
 148
 149
 150 /* Option names of options parsed by the block layer */
 151
 152 #define BDRV_OPT_CACHE_WB       "cache.writeback"
 153 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
 154 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
 155 #define BDRV_OPT_READ_ONLY      "read-only"
 156 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
 157 #define BDRV_OPT_DISCARD        "discard"
 158 #define BDRV_OPT_FORCE_SHARE    "force-share"
 159
 160
 161 #define BDRV_SECTOR_BITS   9
 162 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
 163
 164 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
 165                                            INT_MAX >> BDRV_SECTOR_BITS)
 166 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
 167
 168 /*
 169  * We want allow aligning requests and disk length up to any 32bit alignment
 170  * and don't afraid of overflow.
 171  * To achieve it, and in the same time use some pretty number as maximum disk
 172  * size, let's define maximum "length" (a limit for any offset/bytes request and
 173  * for disk size) to be the greatest power of 2 less than INT64_MAX.
 174  */
 175 #define BDRV_MAX_ALIGNMENT (1L << 30)
 176 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT))
 177
 178 /*
 179  * Allocation status flags for bdrv_block_status() and friends.
 180  *
 181  * Public flags:
 182  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
 183  * BDRV_BLOCK_ZERO: offset reads as zero
 184  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
 185  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
 186  *                       layer rather than any backing, set by block layer
 187  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
 188  *                 layer, set by block layer
 189  *
 190  * Internal flags:
 191  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
 192  *                 that the block layer recompute the answer from the returned
 193  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
 194  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
 195  *                     zeroes in file child of current block node inside
 196  *                     returned region. Only valid together with both
 197  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
 198  *                     appear with BDRV_BLOCK_ZERO.
 199  *
 200  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
 201  * host offset within the returned BDS that is allocated for the
 202  * corresponding raw guest data.  However, whether that offset
 203  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
 204  *
 205  * DATA ZERO OFFSET_VALID
 206  *  t    t        t       sectors read as zero, returned file is zero at offset
 207  *  t    f        t       sectors read as valid from file at offset
 208  *  f    t        t       sectors preallocated, read as zero, returned file not
 209  *                        necessarily zero at offset
 210  *  f    f        t       sectors preallocated but read from backing_hd,
 211  *                        returned file contains garbage at offset
 212  *  t    t        f       sectors preallocated, read as zero, unknown offset
 213  *  t    f        f       sectors read from unknown file or offset
 214  *  f    t        f       not allocated or unknown offset, read as zero
 215  *  f    f        f       not allocated or unknown offset, read from backing_hd
 216  */
 217 #define BDRV_BLOCK_DATA         0x01
 218 #define BDRV_BLOCK_ZERO         0x02
 219 #define BDRV_BLOCK_OFFSET_VALID 0x04
 220 #define BDRV_BLOCK_RAW          0x08
 221 #define BDRV_BLOCK_ALLOCATED    0x10
 222 #define BDRV_BLOCK_EOF          0x20
 223 #define BDRV_BLOCK_RECURSE      0x40
 224
 225 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
 226
 227 typedef struct BDRVReopenState {
 228     BlockDriverState *bs;
 229     int flags;
 230     BlockdevDetectZeroesOptions detect_zeroes;
 231     bool backing_missing;
 232     BlockDriverState *old_backing_bs; /* keep pointer for permissions update */
 233     BlockDriverState *old_file_bs; /* keep pointer for permissions update */
 234     QDict *options;
 235     QDict *explicit_options;
 236     void *opaque;
 237 } BDRVReopenState;
 238
 239 /*
 240  * Block operation types
 241  */
 242 typedef enum BlockOpType {
 243     BLOCK_OP_TYPE_BACKUP_SOURCE,
 244     BLOCK_OP_TYPE_BACKUP_TARGET,
 245     BLOCK_OP_TYPE_CHANGE,
 246     BLOCK_OP_TYPE_COMMIT_SOURCE,
 247     BLOCK_OP_TYPE_COMMIT_TARGET,
 248     BLOCK_OP_TYPE_DATAPLANE,
 249     BLOCK_OP_TYPE_DRIVE_DEL,
 250     BLOCK_OP_TYPE_EJECT,
 251     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
 252     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
 253     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
 254     BLOCK_OP_TYPE_MIRROR_SOURCE,
 255     BLOCK_OP_TYPE_MIRROR_TARGET,
 256     BLOCK_OP_TYPE_RESIZE,
 257     BLOCK_OP_TYPE_STREAM,
 258     BLOCK_OP_TYPE_REPLACE,
 259     BLOCK_OP_TYPE_MAX,
 260 } BlockOpType;
 261
 262 /* Block node permission constants */
 263 enum {
 264     /**
 265      * A user that has the "permission" of consistent reads is guaranteed that
 266      * their view of the contents of the block device is complete and
 267      * self-consistent, representing the contents of a disk at a specific
 268      * point.
 269      *
 270      * For most block devices (including their backing files) this is true, but
 271      * the property cannot be maintained in a few situations like for
 272      * intermediate nodes of a commit block job.
 273      */
 274     BLK_PERM_CONSISTENT_READ    = 0x01,
 275
 276     /** This permission is required to change the visible disk contents. */
 277     BLK_PERM_WRITE              = 0x02,
 278
 279     /**
 280      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
 281      * required for writes to the block node when the caller promises that
 282      * the visible disk content doesn't change.
 283      *
 284      * As the BLK_PERM_WRITE permission is strictly stronger, either is
 285      * sufficient to perform an unchanging write.
 286      */
 287     BLK_PERM_WRITE_UNCHANGED    = 0x04,
 288
 289     /** This permission is required to change the size of a block node. */
 290     BLK_PERM_RESIZE             = 0x08,
 291
 292     /**
 293      * There was a now-removed bit BLK_PERM_GRAPH_MOD, with value of 0x10. QEMU
 294      * 6.1 and earlier may still lock the corresponding byte in block/file-posix
 295      * locking.  So, implementing some new permission should be very careful to
 296      * not interfere with this old unused thing.
 297      */
 298
 299     BLK_PERM_ALL                = 0x0f,
 300
 301     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
 302                                  | BLK_PERM_WRITE
 303                                  | BLK_PERM_WRITE_UNCHANGED
 304                                  | BLK_PERM_RESIZE,
 305
 306     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
 307 };
 308
 309 /*
 310  * Flags that parent nodes assign to child nodes to specify what kind of
 311  * role(s) they take.
 312  *
 313  * At least one of DATA, METADATA, FILTERED, or COW must be set for
 314  * every child.
 315  */
 316 enum BdrvChildRoleBits {
 317     /*
 318      * This child stores data.
 319      * Any node may have an arbitrary number of such children.
 320      */
 321     BDRV_CHILD_DATA         = (1 << 0),
 322
 323     /*
 324      * This child stores metadata.
 325      * Any node may have an arbitrary number of metadata-storing
 326      * children.
 327      */
 328     BDRV_CHILD_METADATA     = (1 << 1),
 329
 330     /*
 331      * A child that always presents exactly the same visible data as
 332      * the parent, e.g. by virtue of the parent forwarding all reads
 333      * and writes.
 334      * This flag is mutually exclusive with DATA, METADATA, and COW.
 335      * Any node may have at most one filtered child at a time.
 336      */
 337     BDRV_CHILD_FILTERED     = (1 << 2),
 338
 339     /*
 340      * Child from which to read all data that isn't allocated in the
 341      * parent (i.e., the backing child); such data is copied to the
 342      * parent through COW (and optionally COR).
 343      * This field is mutually exclusive with DATA, METADATA, and
 344      * FILTERED.
 345      * Any node may have at most one such backing child at a time.
 346      */
 347     BDRV_CHILD_COW          = (1 << 3),
 348
 349     /*
 350      * The primary child.  For most drivers, this is the child whose
 351      * filename applies best to the parent node.
 352      * Any node may have at most one primary child at a time.
 353      */
 354     BDRV_CHILD_PRIMARY      = (1 << 4),
 355
 356     /* Useful combination of flags */
 357     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
 358                               | BDRV_CHILD_METADATA
 359                               | BDRV_CHILD_PRIMARY,
 360 };
 361
 362 /* Mask of BdrvChildRoleBits values */
 363 typedef unsigned int BdrvChildRole;
 364
 365 typedef struct BdrvCheckResult {
 366     int corruptions;
 367     int leaks;
 368     int check_errors;
 369     int corruptions_fixed;
 370     int leaks_fixed;
 371     int64_t image_end_offset;
 372     BlockFragInfo bfi;
 373 } BdrvCheckResult;
 374
 375 typedef enum {
 376     BDRV_FIX_LEAKS    = 1,
 377     BDRV_FIX_ERRORS   = 2,
 378 } BdrvCheckMode;
 379
 380 typedef struct BlockSizes {
 381     uint32_t phys;
 382     uint32_t log;
 383 } BlockSizes;
 384
 385 typedef struct HDGeometry {
 386     uint32_t heads;
 387     uint32_t sectors;
 388     uint32_t cylinders;
 389 } HDGeometry;
 390
 391 /*
 392  * Common functions that are neither I/O nor Global State.
 393  *
 394  * These functions must never call any function from other categories
 395  * (I/O, "I/O or GS", Global State) except this one, but can be invoked by
 396  * all of them.
 397  */
 398
 399 char *bdrv_perm_names(uint64_t perm);
 400 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
 401
 402 void bdrv_init_with_whitelist(void);
 403 bool bdrv_uses_whitelist(void);
 404 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
 405
 406 int bdrv_parse_aio(const char *mode, int *flags);
 407 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
 408 int bdrv_parse_discard_flags(const char *mode, int *flags);
 409
 410 int path_has_protocol(const char *path);
 411 int path_is_absolute(const char *path);
 412 char *path_combine(const char *base_path, const char *filename);
 413
 414 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
 415                                                    const char *backing,
 416                                                    Error **errp);
 417
 418 #endif /* BLOCK_COMMON_H */