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