include/block/block_int.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_INT_H
  25 #define BLOCK_INT_H
  26
  27 #include "block/block.h"
  28 #include "qemu/option.h"
  29 #include "qemu/queue.h"
  30 #include "block/coroutine.h"
  31 #include "qemu/timer.h"
  32 #include "qapi-types.h"
  33 #include "qapi/qmp/qerror.h"
  34 #include "monitor/monitor.h"
  35 #include "qemu/hbitmap.h"
  36 #include "block/snapshot.h"
  37 #include "qemu/main-loop.h"
  38
  39 #define BLOCK_FLAG_ENCRYPT          1
  40 #define BLOCK_FLAG_COMPAT6          4
  41 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
  42
  43 #define BLOCK_IO_LIMIT_READ     0
  44 #define BLOCK_IO_LIMIT_WRITE    1
  45 #define BLOCK_IO_LIMIT_TOTAL    2
  46
  47 #define BLOCK_IO_SLICE_TIME     100000000
  48 #define NANOSECONDS_PER_SECOND  1000000000.0
  49
  50 #define BLOCK_OPT_SIZE              "size"
  51 #define BLOCK_OPT_ENCRYPT           "encryption"
  52 #define BLOCK_OPT_COMPAT6           "compat6"
  53 #define BLOCK_OPT_BACKING_FILE      "backing_file"
  54 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
  55 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
  56 #define BLOCK_OPT_TABLE_SIZE        "table_size"
  57 #define BLOCK_OPT_PREALLOC          "preallocation"
  58 #define BLOCK_OPT_SUBFMT            "subformat"
  59 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
  60 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
  61 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
  62
  63 typedef struct BdrvTrackedRequest {
  64     BlockDriverState *bs;
  65     int64_t sector_num;
  66     int nb_sectors;
  67     bool is_write;
  68     QLIST_ENTRY(BdrvTrackedRequest) list;
  69     Coroutine *co; /* owner, used for deadlock detection */
  70     CoQueue wait_queue; /* coroutines blocked on this request */
  71 } BdrvTrackedRequest;
  72
  73
  74 typedef struct BlockIOLimit {
  75     int64_t bps[3];
  76     int64_t iops[3];
  77 } BlockIOLimit;
  78
  79 typedef struct BlockIOBaseValue {
  80     uint64_t bytes[2];
  81     uint64_t ios[2];
  82 } BlockIOBaseValue;
  83
  84 struct BlockDriver {
  85     const char *format_name;
  86     int instance_size;
  87     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
  88     int (*bdrv_probe_device)(const char *filename);
  89
  90     /* Any driver implementing this callback is expected to be able to handle
  91      * NULL file names in its .bdrv_open() implementation */
  92     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
  93
  94     /* For handling image reopen for split or non-split files */
  95     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
  96                                BlockReopenQueue *queue, Error **errp);
  97     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
  98     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
  99
 100     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags);
 101     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags);
 102     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
 103                      uint8_t *buf, int nb_sectors);
 104     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
 105                       const uint8_t *buf, int nb_sectors);
 106     void (*bdrv_close)(BlockDriverState *bs);
 107     void (*bdrv_rebind)(BlockDriverState *bs);
 108     int (*bdrv_create)(const char *filename, QEMUOptionParameter *options);
 109     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
 110     int (*bdrv_make_empty)(BlockDriverState *bs);
 111     /* aio */
 112     BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
 113         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 114         BlockDriverCompletionFunc *cb, void *opaque);
 115     BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
 116         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 117         BlockDriverCompletionFunc *cb, void *opaque);
 118     BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
 119         BlockDriverCompletionFunc *cb, void *opaque);
 120     BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
 121         int64_t sector_num, int nb_sectors,
 122         BlockDriverCompletionFunc *cb, void *opaque);
 123
 124     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
 125         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 126     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
 127         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 128     /*
 129      * Efficiently zero a region of the disk image.  Typically an image format
 130      * would use a compact metadata representation to implement this.  This
 131      * function pointer may be NULL and .bdrv_co_writev() will be called
 132      * instead.
 133      */
 134     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
 135         int64_t sector_num, int nb_sectors);
 136     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
 137         int64_t sector_num, int nb_sectors);
 138     int coroutine_fn (*bdrv_co_is_allocated)(BlockDriverState *bs,
 139         int64_t sector_num, int nb_sectors, int *pnum);
 140
 141     /*
 142      * Invalidate any cached meta-data.
 143      */
 144     void (*bdrv_invalidate_cache)(BlockDriverState *bs);
 145
 146     /*
 147      * Flushes all data that was already written to the OS all the way down to
 148      * the disk (for example raw-posix calls fsync()).
 149      */
 150     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
 151
 152     /*
 153      * Flushes all internal caches to the OS. The data may still sit in a
 154      * writeback cache of the host OS, but it will survive a crash of the qemu
 155      * process.
 156      */
 157     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
 158
 159     const char *protocol_name;
 160     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
 161     int64_t (*bdrv_getlength)(BlockDriverState *bs);
 162     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
 163     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
 164                                  const uint8_t *buf, int nb_sectors);
 165
 166     int (*bdrv_snapshot_create)(BlockDriverState *bs,
 167                                 QEMUSnapshotInfo *sn_info);
 168     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
 169                               const char *snapshot_id);
 170     int (*bdrv_snapshot_delete)(BlockDriverState *bs, const char *snapshot_id);
 171     int (*bdrv_snapshot_list)(BlockDriverState *bs,
 172                               QEMUSnapshotInfo **psn_info);
 173     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
 174                                   const char *snapshot_name);
 175     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
 176
 177     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
 178                              int64_t pos);
 179     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
 180                              int64_t pos, int size);
 181
 182     int (*bdrv_change_backing_file)(BlockDriverState *bs,
 183         const char *backing_file, const char *backing_fmt);
 184
 185     /* removable device specific */
 186     int (*bdrv_is_inserted)(BlockDriverState *bs);
 187     int (*bdrv_media_changed)(BlockDriverState *bs);
 188     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
 189     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
 190
 191     /* to control generic scsi devices */
 192     int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
 193     BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
 194         unsigned long int req, void *buf,
 195         BlockDriverCompletionFunc *cb, void *opaque);
 196
 197     /* List of options for creating images, terminated by name == NULL */
 198     QEMUOptionParameter *create_options;
 199
 200
 201     /*
 202      * Returns 0 for completed check, -errno for internal errors.
 203      * The check results are stored in result.
 204      */
 205     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
 206         BdrvCheckMode fix);
 207
 208     void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
 209
 210     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
 211     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
 212         const char *tag);
 213     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
 214     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
 215
 216     /*
 217      * Returns 1 if newly created images are guaranteed to contain only
 218      * zeros, 0 otherwise.
 219      */
 220     int (*bdrv_has_zero_init)(BlockDriverState *bs);
 221
 222     QLIST_ENTRY(BlockDriver) list;
 223 };
 224
 225 /*
 226  * Note: the function bdrv_append() copies and swaps contents of
 227  * BlockDriverStates, so if you add new fields to this struct, please
 228  * inspect bdrv_append() to determine if the new fields need to be
 229  * copied as well.
 230  */
 231 struct BlockDriverState {
 232     int64_t total_sectors; /* if we are reading a disk image, give its
 233                               size in sectors */
 234     int read_only; /* if true, the media is read only */
 235     int open_flags; /* flags used to open the file, re-used for re-open */
 236     int encrypted; /* if true, the media is encrypted */
 237     int valid_key; /* if true, a valid encryption key has been set */
 238     int sg;        /* if true, the device is a /dev/sg* */
 239     int copy_on_read; /* if true, copy read backing sectors into image
 240                          note this is a reference count */
 241
 242     BlockDriver *drv; /* NULL means no media */
 243     void *opaque;
 244
 245     void *dev;                  /* attached device model, if any */
 246     /* TODO change to DeviceState when all users are qdevified */
 247     const BlockDevOps *dev_ops;
 248     void *dev_opaque;
 249
 250     char filename[1024];
 251     char backing_file[1024]; /* if non zero, the image is a diff of
 252                                 this file image */
 253     char backing_format[16]; /* if non-zero and backing_file exists */
 254     int is_temporary;
 255
 256     BlockDriverState *backing_hd;
 257     BlockDriverState *file;
 258
 259     NotifierList close_notifiers;
 260
 261     /* Callback before write request is processed */
 262     NotifierWithReturnList before_write_notifiers;
 263
 264     /* number of in-flight copy-on-read requests */
 265     unsigned int copy_on_read_in_flight;
 266
 267     /* the time for latest disk I/O */
 268     int64_t slice_start;
 269     int64_t slice_end;
 270     BlockIOLimit io_limits;
 271     BlockIOBaseValue slice_submitted;
 272     CoQueue      throttled_reqs;
 273     QEMUTimer    *block_timer;
 274     bool         io_limits_enabled;
 275
 276     /* I/O stats (display with "info blockstats"). */
 277     uint64_t nr_bytes[BDRV_MAX_IOTYPE];
 278     uint64_t nr_ops[BDRV_MAX_IOTYPE];
 279     uint64_t total_time_ns[BDRV_MAX_IOTYPE];
 280     uint64_t wr_highest_sector;
 281
 282     /* Whether the disk can expand beyond total_sectors */
 283     int growable;
 284
 285     /* Whether produces zeros when read beyond eof */
 286     bool zero_beyond_eof;
 287
 288     /* the memory alignment required for the buffers handled by this driver */
 289     int buffer_alignment;
 290
 291     /* do we need to tell the quest if we have a volatile write cache? */
 292     int enable_write_cache;
 293
 294     /* NOTE: the following infos are only hints for real hardware
 295        drivers. They are not used by the block driver */
 296     BlockdevOnError on_read_error, on_write_error;
 297     bool iostatus_enabled;
 298     BlockDeviceIoStatus iostatus;
 299     char device_name[32];
 300     HBitmap *dirty_bitmap;
 301     int in_use; /* users other than guest access, eg. block migration */
 302     QTAILQ_ENTRY(BlockDriverState) list;
 303
 304     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
 305
 306     /* long-running background operation */
 307     BlockJob *job;
 308
 309     QDict *options;
 310 };
 311
 312 int get_tmp_filename(char *filename, int size);
 313
 314 void bdrv_set_io_limits(BlockDriverState *bs,
 315                         BlockIOLimit *io_limits);
 316
 317 /**
 318  * bdrv_add_before_write_notifier:
 319  *
 320  * Register a callback that is invoked before write requests are processed but
 321  * after any throttling or waiting for overlapping requests.
 322  */
 323 void bdrv_add_before_write_notifier(BlockDriverState *bs,
 324                                     NotifierWithReturn *notifier);
 325
 326 /**
 327  * bdrv_get_aio_context:
 328  *
 329  * Returns: the currently bound #AioContext
 330  */
 331 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
 332
 333 #ifdef _WIN32
 334 int is_windows_drive(const char *filename);
 335 #endif
 336 void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv,
 337                                enum MonitorEvent ev,
 338                                BlockErrorAction action, bool is_read);
 339
 340 /**
 341  * stream_start:
 342  * @bs: Block device to operate on.
 343  * @base: Block device that will become the new base, or %NULL to
 344  * flatten the whole backing file chain onto @bs.
 345  * @base_id: The file name that will be written to @bs as the new
 346  * backing file if the job completes.  Ignored if @base is %NULL.
 347  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 348  * @on_error: The action to take upon error.
 349  * @cb: Completion function for the job.
 350  * @opaque: Opaque pointer value passed to @cb.
 351  * @errp: Error object.
 352  *
 353  * Start a streaming operation on @bs.  Clusters that are unallocated
 354  * in @bs, but allocated in any image between @base and @bs (both
 355  * exclusive) will be written to @bs.  At the end of a successful
 356  * streaming job, the backing file of @bs will be changed to
 357  * @base_id in the written image and to @base in the live BlockDriverState.
 358  */
 359 void stream_start(BlockDriverState *bs, BlockDriverState *base,
 360                   const char *base_id, int64_t speed, BlockdevOnError on_error,
 361                   BlockDriverCompletionFunc *cb,
 362                   void *opaque, Error **errp);
 363
 364 /**
 365  * commit_start:
 366  * @bs: Top Block device
 367  * @base: Block device that will be written into, and become the new top
 368  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 369  * @on_error: The action to take upon error.
 370  * @cb: Completion function for the job.
 371  * @opaque: Opaque pointer value passed to @cb.
 372  * @errp: Error object.
 373  *
 374  */
 375 void commit_start(BlockDriverState *bs, BlockDriverState *base,
 376                  BlockDriverState *top, int64_t speed,
 377                  BlockdevOnError on_error, BlockDriverCompletionFunc *cb,
 378                  void *opaque, Error **errp);
 379
 380 /*
 381  * mirror_start:
 382  * @bs: Block device to operate on.
 383  * @target: Block device to write to.
 384  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 385  * @granularity: The chosen granularity for the dirty bitmap.
 386  * @buf_size: The amount of data that can be in flight at one time.
 387  * @mode: Whether to collapse all images in the chain to the target.
 388  * @on_source_error: The action to take upon error reading from the source.
 389  * @on_target_error: The action to take upon error writing to the target.
 390  * @cb: Completion function for the job.
 391  * @opaque: Opaque pointer value passed to @cb.
 392  * @errp: Error object.
 393  *
 394  * Start a mirroring operation on @bs.  Clusters that are allocated
 395  * in @bs will be written to @bs until the job is cancelled or
 396  * manually completed.  At the end of a successful mirroring job,
 397  * @bs will be switched to read from @target.
 398  */
 399 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
 400                   int64_t speed, int64_t granularity, int64_t buf_size,
 401                   MirrorSyncMode mode, BlockdevOnError on_source_error,
 402                   BlockdevOnError on_target_error,
 403                   BlockDriverCompletionFunc *cb,
 404                   void *opaque, Error **errp);
 405
 406 /*
 407  * backup_start:
 408  * @bs: Block device to operate on.
 409  * @target: Block device to write to.
 410  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 411  * @sync_mode: What parts of the disk image should be copied to the destination.
 412  * @on_source_error: The action to take upon error reading from the source.
 413  * @on_target_error: The action to take upon error writing to the target.
 414  * @cb: Completion function for the job.
 415  * @opaque: Opaque pointer value passed to @cb.
 416  *
 417  * Start a backup operation on @bs.  Clusters in @bs are written to @target
 418  * until the job is cancelled or manually completed.
 419  */
 420 void backup_start(BlockDriverState *bs, BlockDriverState *target,
 421                   int64_t speed, MirrorSyncMode sync_mode,
 422                   BlockdevOnError on_source_error,
 423                   BlockdevOnError on_target_error,
 424                   BlockDriverCompletionFunc *cb, void *opaque,
 425                   Error **errp);
 426
 427 #endif /* BLOCK_INT_H */