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