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