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.h"
  28 #include "qemu-option.h"
  29 #include "qemu-queue.h"
  30 #include "qemu-coroutine.h"
  31 #include "qemu-timer.h"
  32 #include "qapi-types.h"
  33 #include "qerror.h"
  34
  35 #define BLOCK_FLAG_ENCRYPT          1
  36 #define BLOCK_FLAG_COMPAT6          4
  37 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
  38
  39 #define BLOCK_IO_LIMIT_READ     0
  40 #define BLOCK_IO_LIMIT_WRITE    1
  41 #define BLOCK_IO_LIMIT_TOTAL    2
  42
  43 #define BLOCK_IO_SLICE_TIME     100000000
  44 #define NANOSECONDS_PER_SECOND  1000000000.0
  45
  46 #define BLOCK_OPT_SIZE              "size"
  47 #define BLOCK_OPT_ENCRYPT           "encryption"
  48 #define BLOCK_OPT_COMPAT6           "compat6"
  49 #define BLOCK_OPT_BACKING_FILE      "backing_file"
  50 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
  51 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
  52 #define BLOCK_OPT_TABLE_SIZE        "table_size"
  53 #define BLOCK_OPT_PREALLOC          "preallocation"
  54 #define BLOCK_OPT_SUBFMT            "subformat"
  55 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
  56 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
  57
  58 typedef struct BdrvTrackedRequest BdrvTrackedRequest;
  59
  60 typedef struct BlockIOLimit {
  61     int64_t bps[3];
  62     int64_t iops[3];
  63 } BlockIOLimit;
  64
  65 typedef struct BlockIOBaseValue {
  66     uint64_t bytes[2];
  67     uint64_t ios[2];
  68 } BlockIOBaseValue;
  69
  70 typedef struct BlockJob BlockJob;
  71
  72 /**
  73  * BlockJobType:
  74  *
  75  * A class type for block job objects.
  76  */
  77 typedef struct BlockJobType {
  78     /** Derived BlockJob struct size */
  79     size_t instance_size;
  80
  81     /** String describing the operation, part of query-block-jobs QMP API */
  82     const char *job_type;
  83
  84     /** Optional callback for job types that support setting a speed limit */
  85     void (*set_speed)(BlockJob *job, int64_t speed, Error **errp);
  86 } BlockJobType;
  87
  88 /**
  89  * BlockJob:
  90  *
  91  * Long-running operation on a BlockDriverState.
  92  */
  93 struct BlockJob {
  94     /** The job type, including the job vtable.  */
  95     const BlockJobType *job_type;
  96
  97     /** The block device on which the job is operating.  */
  98     BlockDriverState *bs;
  99
 100     /**
 101      * The coroutine that executes the job.  If not NULL, it is
 102      * reentered when busy is false and the job is cancelled.
 103      */
 104     Coroutine *co;
 105
 106     /**
 107      * Set to true if the job should cancel itself.  The flag must
 108      * always be tested just before toggling the busy flag from false
 109      * to true.  After a job has been cancelled, it should only yield
 110      * if #qemu_aio_wait will ("sooner or later") reenter the coroutine.
 111      */
 112     bool cancelled;
 113
 114     /**
 115      * Set to false by the job while it is in a quiescent state, where
 116      * no I/O is pending and the job has yielded on any condition
 117      * that is not detected by #qemu_aio_wait, such as a timer.
 118      */
 119     bool busy;
 120
 121     /** Offset that is published by the query-block-jobs QMP API */
 122     int64_t offset;
 123
 124     /** Length that is published by the query-block-jobs QMP API */
 125     int64_t len;
 126
 127     /** Speed that was set with @block_job_set_speed.  */
 128     int64_t speed;
 129
 130     /** The completion function that will be called when the job completes.  */
 131     BlockDriverCompletionFunc *cb;
 132
 133     /** The opaque value that is passed to the completion function.  */
 134     void *opaque;
 135 };
 136
 137 struct BlockDriver {
 138     const char *format_name;
 139     int instance_size;
 140     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
 141     int (*bdrv_probe_device)(const char *filename);
 142
 143     /* For handling image reopen for split or non-split files */
 144     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
 145                                BlockReopenQueue *queue, Error **errp);
 146     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
 147     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
 148
 149     int (*bdrv_open)(BlockDriverState *bs, int flags);
 150     int (*bdrv_file_open)(BlockDriverState *bs, const char *filename, int flags);
 151     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
 152                      uint8_t *buf, int nb_sectors);
 153     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
 154                       const uint8_t *buf, int nb_sectors);
 155     void (*bdrv_close)(BlockDriverState *bs);
 156     void (*bdrv_rebind)(BlockDriverState *bs);
 157     int (*bdrv_create)(const char *filename, QEMUOptionParameter *options);
 158     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
 159     int (*bdrv_make_empty)(BlockDriverState *bs);
 160     /* aio */
 161     BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
 162         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 163         BlockDriverCompletionFunc *cb, void *opaque);
 164     BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
 165         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 166         BlockDriverCompletionFunc *cb, void *opaque);
 167     BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
 168         BlockDriverCompletionFunc *cb, void *opaque);
 169     BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
 170         int64_t sector_num, int nb_sectors,
 171         BlockDriverCompletionFunc *cb, void *opaque);
 172
 173     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
 174         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 175     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
 176         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 177     /*
 178      * Efficiently zero a region of the disk image.  Typically an image format
 179      * would use a compact metadata representation to implement this.  This
 180      * function pointer may be NULL and .bdrv_co_writev() will be called
 181      * instead.
 182      */
 183     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
 184         int64_t sector_num, int nb_sectors);
 185     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
 186         int64_t sector_num, int nb_sectors);
 187     int coroutine_fn (*bdrv_co_is_allocated)(BlockDriverState *bs,
 188         int64_t sector_num, int nb_sectors, int *pnum);
 189
 190     /*
 191      * Invalidate any cached meta-data.
 192      */
 193     void (*bdrv_invalidate_cache)(BlockDriverState *bs);
 194
 195     /*
 196      * Flushes all data that was already written to the OS all the way down to
 197      * the disk (for example raw-posix calls fsync()).
 198      */
 199     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
 200
 201     /*
 202      * Flushes all internal caches to the OS. The data may still sit in a
 203      * writeback cache of the host OS, but it will survive a crash of the qemu
 204      * process.
 205      */
 206     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
 207
 208     const char *protocol_name;
 209     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
 210     int64_t (*bdrv_getlength)(BlockDriverState *bs);
 211     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
 212     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
 213                                  const uint8_t *buf, int nb_sectors);
 214
 215     int (*bdrv_snapshot_create)(BlockDriverState *bs,
 216                                 QEMUSnapshotInfo *sn_info);
 217     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
 218                               const char *snapshot_id);
 219     int (*bdrv_snapshot_delete)(BlockDriverState *bs, const char *snapshot_id);
 220     int (*bdrv_snapshot_list)(BlockDriverState *bs,
 221                               QEMUSnapshotInfo **psn_info);
 222     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
 223                                   const char *snapshot_name);
 224     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
 225
 226     int (*bdrv_save_vmstate)(BlockDriverState *bs, const uint8_t *buf,
 227                              int64_t pos, int size);
 228     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
 229                              int64_t pos, int size);
 230
 231     int (*bdrv_change_backing_file)(BlockDriverState *bs,
 232         const char *backing_file, const char *backing_fmt);
 233
 234     /* removable device specific */
 235     int (*bdrv_is_inserted)(BlockDriverState *bs);
 236     int (*bdrv_media_changed)(BlockDriverState *bs);
 237     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
 238     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
 239
 240     /* to control generic scsi devices */
 241     int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
 242     BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
 243         unsigned long int req, void *buf,
 244         BlockDriverCompletionFunc *cb, void *opaque);
 245
 246     /* List of options for creating images, terminated by name == NULL */
 247     QEMUOptionParameter *create_options;
 248
 249
 250     /*
 251      * Returns 0 for completed check, -errno for internal errors.
 252      * The check results are stored in result.
 253      */
 254     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
 255         BdrvCheckMode fix);
 256
 257     void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
 258
 259     /*
 260      * Returns 1 if newly created images are guaranteed to contain only
 261      * zeros, 0 otherwise.
 262      */
 263     int (*bdrv_has_zero_init)(BlockDriverState *bs);
 264
 265     QLIST_ENTRY(BlockDriver) list;
 266 };
 267
 268 /*
 269  * Note: the function bdrv_append() copies and swaps contents of
 270  * BlockDriverStates, so if you add new fields to this struct, please
 271  * inspect bdrv_append() to determine if the new fields need to be
 272  * copied as well.
 273  */
 274 struct BlockDriverState {
 275     int64_t total_sectors; /* if we are reading a disk image, give its
 276                               size in sectors */
 277     int read_only; /* if true, the media is read only */
 278     int open_flags; /* flags used to open the file, re-used for re-open */
 279     int encrypted; /* if true, the media is encrypted */
 280     int valid_key; /* if true, a valid encryption key has been set */
 281     int sg;        /* if true, the device is a /dev/sg* */
 282     int copy_on_read; /* if true, copy read backing sectors into image
 283                          note this is a reference count */
 284
 285     BlockDriver *drv; /* NULL means no media */
 286     void *opaque;
 287
 288     void *dev;                  /* attached device model, if any */
 289     /* TODO change to DeviceState when all users are qdevified */
 290     const BlockDevOps *dev_ops;
 291     void *dev_opaque;
 292
 293     char filename[1024];
 294     char backing_file[1024]; /* if non zero, the image is a diff of
 295                                 this file image */
 296     char backing_format[16]; /* if non-zero and backing_file exists */
 297     int is_temporary;
 298
 299     BlockDriverState *backing_hd;
 300     BlockDriverState *file;
 301
 302     /* number of in-flight copy-on-read requests */
 303     unsigned int copy_on_read_in_flight;
 304
 305     /* the time for latest disk I/O */
 306     int64_t slice_time;
 307     int64_t slice_start;
 308     int64_t slice_end;
 309     BlockIOLimit io_limits;
 310     BlockIOBaseValue  io_base;
 311     CoQueue      throttled_reqs;
 312     QEMUTimer    *block_timer;
 313     bool         io_limits_enabled;
 314
 315     /* I/O stats (display with "info blockstats"). */
 316     uint64_t nr_bytes[BDRV_MAX_IOTYPE];
 317     uint64_t nr_ops[BDRV_MAX_IOTYPE];
 318     uint64_t total_time_ns[BDRV_MAX_IOTYPE];
 319     uint64_t wr_highest_sector;
 320
 321     /* Whether the disk can expand beyond total_sectors */
 322     int growable;
 323
 324     /* the memory alignment required for the buffers handled by this driver */
 325     int buffer_alignment;
 326
 327     /* do we need to tell the quest if we have a volatile write cache? */
 328     int enable_write_cache;
 329
 330     /* NOTE: the following infos are only hints for real hardware
 331        drivers. They are not used by the block driver */
 332     BlockErrorAction on_read_error, on_write_error;
 333     bool iostatus_enabled;
 334     BlockDeviceIoStatus iostatus;
 335     char device_name[32];
 336     unsigned long *dirty_bitmap;
 337     int64_t dirty_count;
 338     int in_use; /* users other than guest access, eg. block migration */
 339     QTAILQ_ENTRY(BlockDriverState) list;
 340
 341     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
 342
 343     /* long-running background operation */
 344     BlockJob *job;
 345
 346 };
 347
 348 int get_tmp_filename(char *filename, int size);
 349
 350 void bdrv_set_io_limits(BlockDriverState *bs,
 351                         BlockIOLimit *io_limits);
 352
 353 #ifdef _WIN32
 354 int is_windows_drive(const char *filename);
 355 #endif
 356
 357 /**
 358  * block_job_create:
 359  * @job_type: The class object for the newly-created job.
 360  * @bs: The block
 361  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 362  * @cb: Completion function for the job.
 363  * @opaque: Opaque pointer value passed to @cb.
 364  * @errp: Error object.
 365  *
 366  * Create a new long-running block device job and return it.  The job
 367  * will call @cb asynchronously when the job completes.  Note that
 368  * @bs may have been closed at the time the @cb it is called.  If
 369  * this is the case, the job may be reported as either cancelled or
 370  * completed.
 371  *
 372  * This function is not part of the public job interface; it should be
 373  * called from a wrapper that is specific to the job type.
 374  */
 375 void *block_job_create(const BlockJobType *job_type, BlockDriverState *bs,
 376                        int64_t speed, BlockDriverCompletionFunc *cb,
 377                        void *opaque, Error **errp);
 378
 379 /**
 380  * block_job_sleep_ns:
 381  * @job: The job that calls the function.
 382  * @clock: The clock to sleep on.
 383  * @ns: How many nanoseconds to stop for.
 384  *
 385  * Put the job to sleep (assuming that it wasn't canceled) for @ns
 386  * nanoseconds.  Canceling the job will interrupt the wait immediately.
 387  */
 388 void block_job_sleep_ns(BlockJob *job, QEMUClock *clock, int64_t ns);
 389
 390 /**
 391  * block_job_complete:
 392  * @job: The job being completed.
 393  * @ret: The status code.
 394  *
 395  * Call the completion function that was registered at creation time, and
 396  * free @job.
 397  */
 398 void block_job_complete(BlockJob *job, int ret);
 399
 400 /**
 401  * block_job_set_speed:
 402  * @job: The job to set the speed for.
 403  * @speed: The new value
 404  * @errp: Error object.
 405  *
 406  * Set a rate-limiting parameter for the job; the actual meaning may
 407  * vary depending on the job type.
 408  */
 409 void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp);
 410
 411 /**
 412  * block_job_cancel:
 413  * @job: The job to be canceled.
 414  *
 415  * Asynchronously cancel the specified job.
 416  */
 417 void block_job_cancel(BlockJob *job);
 418
 419 /**
 420  * block_job_is_cancelled:
 421  * @job: The job being queried.
 422  *
 423  * Returns whether the job is scheduled for cancellation.
 424  */
 425 bool block_job_is_cancelled(BlockJob *job);
 426
 427 /**
 428  * block_job_cancel:
 429  * @job: The job to be canceled.
 430  *
 431  * Asynchronously cancel the job and wait for it to reach a quiescent
 432  * state.  Note that the completion callback will still be called
 433  * asynchronously, hence it is *not* valid to call #bdrv_delete
 434  * immediately after #block_job_cancel_sync.  Users of block jobs
 435  * will usually protect the BlockDriverState objects with a reference
 436  * count, should this be a concern.
 437  *
 438  * Returns the return value from the job if the job actually completed
 439  * during the call, or -ECANCELED if it was canceled.
 440  */
 441 int block_job_cancel_sync(BlockJob *job);
 442
 443 /**
 444  * stream_start:
 445  * @bs: Block device to operate on.
 446  * @base: Block device that will become the new base, or %NULL to
 447  * flatten the whole backing file chain onto @bs.
 448  * @base_id: The file name that will be written to @bs as the new
 449  * backing file if the job completes.  Ignored if @base is %NULL.
 450  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 451  * @cb: Completion function for the job.
 452  * @opaque: Opaque pointer value passed to @cb.
 453  * @errp: Error object.
 454  *
 455  * Start a streaming operation on @bs.  Clusters that are unallocated
 456  * in @bs, but allocated in any image between @base and @bs (both
 457  * exclusive) will be written to @bs.  At the end of a successful
 458  * streaming job, the backing file of @bs will be changed to
 459  * @base_id in the written image and to @base in the live BlockDriverState.
 460  */
 461 void stream_start(BlockDriverState *bs, BlockDriverState *base,
 462                   const char *base_id, int64_t speed,
 463                   BlockDriverCompletionFunc *cb,
 464                   void *opaque, Error **errp);
 465
 466 /**
 467  * commit_start:
 468  * @bs: Top Block device
 469  * @base: Block device that will be written into, and become the new top
 470  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 471  * @on_error: The action to take upon error.
 472  * @cb: Completion function for the job.
 473  * @opaque: Opaque pointer value passed to @cb.
 474  * @errp: Error object.
 475  *
 476  */
 477 void commit_start(BlockDriverState *bs, BlockDriverState *base,
 478                  BlockDriverState *top, int64_t speed,
 479                  BlockErrorAction on_error, BlockDriverCompletionFunc *cb,
 480                  void *opaque, Error **errp);
 481
 482 #endif /* BLOCK_INT_H */