xref: /openbmc/qemu/include/block/block_int.h (revision e38e943a)
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 #include "qemu/throttle.h"
39 
40 #define BLOCK_FLAG_ENCRYPT          1
41 #define BLOCK_FLAG_COMPAT6          4
42 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
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 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
55 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
56 #define BLOCK_OPT_REDUNDANCY        "redundancy"
57 
58 typedef struct BdrvTrackedRequest {
59     BlockDriverState *bs;
60     int64_t offset;
61     unsigned int bytes;
62     bool is_write;
63 
64     bool serialising;
65     int64_t overlap_offset;
66     unsigned int overlap_bytes;
67 
68     QLIST_ENTRY(BdrvTrackedRequest) list;
69     Coroutine *co; /* owner, used for deadlock detection */
70     CoQueue wait_queue; /* coroutines blocked on this request */
71 
72     struct BdrvTrackedRequest *waiting_for;
73 } BdrvTrackedRequest;
74 
75 struct BlockDriver {
76     const char *format_name;
77     int instance_size;
78 
79     /* set to true if the BlockDriver is a block filter */
80     bool is_filter;
81     /* for snapshots block filter like Quorum can implement the
82      * following recursive callback.
83      * It's purpose is to recurse on the filter children while calling
84      * bdrv_recurse_is_first_non_filter on them.
85      * For a sample implementation look in the future Quorum block filter.
86      */
87     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
88                                              BlockDriverState *candidate);
89 
90     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
91     int (*bdrv_probe_device)(const char *filename);
92 
93     /* Any driver implementing this callback is expected to be able to handle
94      * NULL file names in its .bdrv_open() implementation */
95     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
96     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
97      * this field set to true, except ones that are defined only by their
98      * child's bs.
99      * An example of the last type will be the quorum block driver.
100      */
101     bool bdrv_needs_filename;
102 
103     /* Set if a driver can support backing files */
104     bool supports_backing;
105 
106     /* For handling image reopen for split or non-split files */
107     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
108                                BlockReopenQueue *queue, Error **errp);
109     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
110     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
111 
112     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
113                      Error **errp);
114     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
115                           Error **errp);
116     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
117                      uint8_t *buf, int nb_sectors);
118     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
119                       const uint8_t *buf, int nb_sectors);
120     void (*bdrv_close)(BlockDriverState *bs);
121     void (*bdrv_rebind)(BlockDriverState *bs);
122     int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
123     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
124     int (*bdrv_make_empty)(BlockDriverState *bs);
125     /* aio */
126     BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
127         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
128         BlockDriverCompletionFunc *cb, void *opaque);
129     BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
130         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
131         BlockDriverCompletionFunc *cb, void *opaque);
132     BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
133         BlockDriverCompletionFunc *cb, void *opaque);
134     BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
135         int64_t sector_num, int nb_sectors,
136         BlockDriverCompletionFunc *cb, void *opaque);
137 
138     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
139         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
140     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
141         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
142     /*
143      * Efficiently zero a region of the disk image.  Typically an image format
144      * would use a compact metadata representation to implement this.  This
145      * function pointer may be NULL and .bdrv_co_writev() will be called
146      * instead.
147      */
148     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
149         int64_t sector_num, int nb_sectors, BdrvRequestFlags flags);
150     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
151         int64_t sector_num, int nb_sectors);
152     int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
153         int64_t sector_num, int nb_sectors, int *pnum);
154 
155     /*
156      * Invalidate any cached meta-data.
157      */
158     void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
159 
160     /*
161      * Flushes all data that was already written to the OS all the way down to
162      * the disk (for example raw-posix calls fsync()).
163      */
164     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
165 
166     /*
167      * Flushes all internal caches to the OS. The data may still sit in a
168      * writeback cache of the host OS, but it will survive a crash of the qemu
169      * process.
170      */
171     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
172 
173     const char *protocol_name;
174     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
175 
176     int64_t (*bdrv_getlength)(BlockDriverState *bs);
177     bool has_variable_length;
178     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
179 
180     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
181                                  const uint8_t *buf, int nb_sectors);
182 
183     int (*bdrv_snapshot_create)(BlockDriverState *bs,
184                                 QEMUSnapshotInfo *sn_info);
185     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
186                               const char *snapshot_id);
187     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
188                                 const char *snapshot_id,
189                                 const char *name,
190                                 Error **errp);
191     int (*bdrv_snapshot_list)(BlockDriverState *bs,
192                               QEMUSnapshotInfo **psn_info);
193     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
194                                   const char *snapshot_id,
195                                   const char *name,
196                                   Error **errp);
197     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
198     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
199 
200     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
201                              int64_t pos);
202     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
203                              int64_t pos, int size);
204 
205     int (*bdrv_change_backing_file)(BlockDriverState *bs,
206         const char *backing_file, const char *backing_fmt);
207 
208     /* removable device specific */
209     int (*bdrv_is_inserted)(BlockDriverState *bs);
210     int (*bdrv_media_changed)(BlockDriverState *bs);
211     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
212     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
213 
214     /* to control generic scsi devices */
215     int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
216     BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
217         unsigned long int req, void *buf,
218         BlockDriverCompletionFunc *cb, void *opaque);
219 
220     /* List of options for creating images, terminated by name == NULL */
221     QemuOptsList *create_opts;
222 
223     /*
224      * Returns 0 for completed check, -errno for internal errors.
225      * The check results are stored in result.
226      */
227     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
228         BdrvCheckMode fix);
229 
230     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts);
231 
232     void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
233 
234     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
235     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
236         const char *tag);
237     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
238         const char *tag);
239     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
240     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
241 
242     int (*bdrv_refresh_limits)(BlockDriverState *bs);
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     /* Remove fd handlers, timers, and other event loop callbacks so the event
251      * loop is no longer in use.  Called with no in-flight requests and in
252      * depth-first traversal order with parents before child nodes.
253      */
254     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
255 
256     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
257      * can be processed again.  Called with no in-flight requests and in
258      * depth-first traversal order with child nodes before parent nodes.
259      */
260     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
261                                     AioContext *new_context);
262 
263     QLIST_ENTRY(BlockDriver) list;
264 };
265 
266 typedef struct BlockLimits {
267     /* maximum number of sectors that can be discarded at once */
268     int max_discard;
269 
270     /* optimal alignment for discard requests in sectors */
271     int64_t discard_alignment;
272 
273     /* maximum number of sectors that can zeroized at once */
274     int max_write_zeroes;
275 
276     /* optimal alignment for write zeroes requests in sectors */
277     int64_t write_zeroes_alignment;
278 
279     /* optimal transfer length in sectors */
280     int opt_transfer_length;
281 
282     /* memory alignment so that no bounce buffer is needed */
283     size_t opt_mem_alignment;
284 } BlockLimits;
285 
286 typedef struct BdrvOpBlocker BdrvOpBlocker;
287 
288 /*
289  * Note: the function bdrv_append() copies and swaps contents of
290  * BlockDriverStates, so if you add new fields to this struct, please
291  * inspect bdrv_append() to determine if the new fields need to be
292  * copied as well.
293  */
294 struct BlockDriverState {
295     int64_t total_sectors; /* if we are reading a disk image, give its
296                               size in sectors */
297     int read_only; /* if true, the media is read only */
298     int open_flags; /* flags used to open the file, re-used for re-open */
299     int encrypted; /* if true, the media is encrypted */
300     int valid_key; /* if true, a valid encryption key has been set */
301     int sg;        /* if true, the device is a /dev/sg* */
302     int copy_on_read; /* if true, copy read backing sectors into image
303                          note this is a reference count */
304 
305     BlockDriver *drv; /* NULL means no media */
306     void *opaque;
307 
308     void *dev;                  /* attached device model, if any */
309     /* TODO change to DeviceState when all users are qdevified */
310     const BlockDevOps *dev_ops;
311     void *dev_opaque;
312 
313     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
314 
315     char filename[1024];
316     char backing_file[1024]; /* if non zero, the image is a diff of
317                                 this file image */
318     char backing_format[16]; /* if non-zero and backing_file exists */
319 
320     BlockDriverState *backing_hd;
321     BlockDriverState *file;
322 
323     NotifierList close_notifiers;
324 
325     /* Callback before write request is processed */
326     NotifierWithReturnList before_write_notifiers;
327 
328     /* number of in-flight serialising requests */
329     unsigned int serialising_in_flight;
330 
331     /* I/O throttling */
332     ThrottleState throttle_state;
333     CoQueue      throttled_reqs[2];
334     bool         io_limits_enabled;
335 
336     /* I/O stats (display with "info blockstats"). */
337     uint64_t nr_bytes[BDRV_MAX_IOTYPE];
338     uint64_t nr_ops[BDRV_MAX_IOTYPE];
339     uint64_t total_time_ns[BDRV_MAX_IOTYPE];
340     uint64_t wr_highest_sector;
341 
342     /* I/O Limits */
343     BlockLimits bl;
344 
345     /* Whether the disk can expand beyond total_sectors */
346     int growable;
347 
348     /* Whether produces zeros when read beyond eof */
349     bool zero_beyond_eof;
350 
351     /* Alignment requirement for offset/length of I/O requests */
352     unsigned int request_alignment;
353 
354     /* the block size for which the guest device expects atomicity */
355     int guest_block_size;
356 
357     /* do we need to tell the quest if we have a volatile write cache? */
358     int enable_write_cache;
359 
360     /* NOTE: the following infos are only hints for real hardware
361        drivers. They are not used by the block driver */
362     BlockdevOnError on_read_error, on_write_error;
363     bool iostatus_enabled;
364     BlockDeviceIoStatus iostatus;
365 
366     /* the following member gives a name to every node on the bs graph. */
367     char node_name[32];
368     /* element of the list of named nodes building the graph */
369     QTAILQ_ENTRY(BlockDriverState) node_list;
370     /* Device name is the name associated with the "drive" the guest sees */
371     char device_name[32];
372     /* element of the list of "drives" the guest sees */
373     QTAILQ_ENTRY(BlockDriverState) device_list;
374     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
375     int refcnt;
376 
377     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
378 
379     /* operation blockers */
380     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
381 
382     /* long-running background operation */
383     BlockJob *job;
384 
385     QDict *options;
386     BlockdevDetectZeroesOptions detect_zeroes;
387 
388     /* The error object in use for blocking operations on backing_hd */
389     Error *backing_blocker;
390 };
391 
392 int get_tmp_filename(char *filename, int size);
393 
394 void bdrv_set_io_limits(BlockDriverState *bs,
395                         ThrottleConfig *cfg);
396 
397 
398 /**
399  * bdrv_add_before_write_notifier:
400  *
401  * Register a callback that is invoked before write requests are processed but
402  * after any throttling or waiting for overlapping requests.
403  */
404 void bdrv_add_before_write_notifier(BlockDriverState *bs,
405                                     NotifierWithReturn *notifier);
406 
407 /**
408  * bdrv_detach_aio_context:
409  *
410  * May be called from .bdrv_detach_aio_context() to detach children from the
411  * current #AioContext.  This is only needed by block drivers that manage their
412  * own children.  Both ->file and ->backing_hd are automatically handled and
413  * block drivers should not call this function on them explicitly.
414  */
415 void bdrv_detach_aio_context(BlockDriverState *bs);
416 
417 /**
418  * bdrv_attach_aio_context:
419  *
420  * May be called from .bdrv_attach_aio_context() to attach children to the new
421  * #AioContext.  This is only needed by block drivers that manage their own
422  * children.  Both ->file and ->backing_hd are automatically handled and block
423  * drivers should not call this function on them explicitly.
424  */
425 void bdrv_attach_aio_context(BlockDriverState *bs,
426                              AioContext *new_context);
427 
428 #ifdef _WIN32
429 int is_windows_drive(const char *filename);
430 #endif
431 
432 /**
433  * stream_start:
434  * @bs: Block device to operate on.
435  * @base: Block device that will become the new base, or %NULL to
436  * flatten the whole backing file chain onto @bs.
437  * @base_id: The file name that will be written to @bs as the new
438  * backing file if the job completes.  Ignored if @base is %NULL.
439  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
440  * @on_error: The action to take upon error.
441  * @cb: Completion function for the job.
442  * @opaque: Opaque pointer value passed to @cb.
443  * @errp: Error object.
444  *
445  * Start a streaming operation on @bs.  Clusters that are unallocated
446  * in @bs, but allocated in any image between @base and @bs (both
447  * exclusive) will be written to @bs.  At the end of a successful
448  * streaming job, the backing file of @bs will be changed to
449  * @base_id in the written image and to @base in the live BlockDriverState.
450  */
451 void stream_start(BlockDriverState *bs, BlockDriverState *base,
452                   const char *base_id, int64_t speed, BlockdevOnError on_error,
453                   BlockDriverCompletionFunc *cb,
454                   void *opaque, Error **errp);
455 
456 /**
457  * commit_start:
458  * @bs: Active block device.
459  * @top: Top block device to be committed.
460  * @base: Block device that will be written into, and become the new top.
461  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
462  * @on_error: The action to take upon error.
463  * @cb: Completion function for the job.
464  * @opaque: Opaque pointer value passed to @cb.
465  * @errp: Error object.
466  *
467  */
468 void commit_start(BlockDriverState *bs, BlockDriverState *base,
469                  BlockDriverState *top, int64_t speed,
470                  BlockdevOnError on_error, BlockDriverCompletionFunc *cb,
471                  void *opaque, Error **errp);
472 /**
473  * commit_active_start:
474  * @bs: Active block device to be committed.
475  * @base: Block device that will be written into, and become the new top.
476  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
477  * @on_error: The action to take upon error.
478  * @cb: Completion function for the job.
479  * @opaque: Opaque pointer value passed to @cb.
480  * @errp: Error object.
481  *
482  */
483 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
484                          int64_t speed,
485                          BlockdevOnError on_error,
486                          BlockDriverCompletionFunc *cb,
487                          void *opaque, Error **errp);
488 /*
489  * mirror_start:
490  * @bs: Block device to operate on.
491  * @target: Block device to write to.
492  * @replaces: Block graph node name to replace once the mirror is done. Can
493  *            only be used when full mirroring is selected.
494  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
495  * @granularity: The chosen granularity for the dirty bitmap.
496  * @buf_size: The amount of data that can be in flight at one time.
497  * @mode: Whether to collapse all images in the chain to the target.
498  * @on_source_error: The action to take upon error reading from the source.
499  * @on_target_error: The action to take upon error writing to the target.
500  * @cb: Completion function for the job.
501  * @opaque: Opaque pointer value passed to @cb.
502  * @errp: Error object.
503  *
504  * Start a mirroring operation on @bs.  Clusters that are allocated
505  * in @bs will be written to @bs until the job is cancelled or
506  * manually completed.  At the end of a successful mirroring job,
507  * @bs will be switched to read from @target.
508  */
509 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
510                   const char *replaces,
511                   int64_t speed, int64_t granularity, int64_t buf_size,
512                   MirrorSyncMode mode, BlockdevOnError on_source_error,
513                   BlockdevOnError on_target_error,
514                   BlockDriverCompletionFunc *cb,
515                   void *opaque, Error **errp);
516 
517 /*
518  * backup_start:
519  * @bs: Block device to operate on.
520  * @target: Block device to write to.
521  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
522  * @sync_mode: What parts of the disk image should be copied to the destination.
523  * @on_source_error: The action to take upon error reading from the source.
524  * @on_target_error: The action to take upon error writing to the target.
525  * @cb: Completion function for the job.
526  * @opaque: Opaque pointer value passed to @cb.
527  *
528  * Start a backup operation on @bs.  Clusters in @bs are written to @target
529  * until the job is cancelled or manually completed.
530  */
531 void backup_start(BlockDriverState *bs, BlockDriverState *target,
532                   int64_t speed, MirrorSyncMode sync_mode,
533                   BlockdevOnError on_source_error,
534                   BlockdevOnError on_target_error,
535                   BlockDriverCompletionFunc *cb, void *opaque,
536                   Error **errp);
537 
538 #endif /* BLOCK_INT_H */
539