xref: /openbmc/qemu/include/block/block_int.h (revision 1d300b5f)
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 
38 #define BLOCK_FLAG_ENCRYPT          1
39 #define BLOCK_FLAG_COMPAT6          4
40 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
41 
42 #define BLOCK_IO_LIMIT_READ     0
43 #define BLOCK_IO_LIMIT_WRITE    1
44 #define BLOCK_IO_LIMIT_TOTAL    2
45 
46 #define BLOCK_IO_SLICE_TIME     100000000
47 #define NANOSECONDS_PER_SECOND  1000000000.0
48 
49 #define BLOCK_OPT_SIZE              "size"
50 #define BLOCK_OPT_ENCRYPT           "encryption"
51 #define BLOCK_OPT_COMPAT6           "compat6"
52 #define BLOCK_OPT_BACKING_FILE      "backing_file"
53 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
54 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
55 #define BLOCK_OPT_TABLE_SIZE        "table_size"
56 #define BLOCK_OPT_PREALLOC          "preallocation"
57 #define BLOCK_OPT_SUBFMT            "subformat"
58 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
59 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
60 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
61 
62 typedef struct BdrvTrackedRequest {
63     BlockDriverState *bs;
64     int64_t sector_num;
65     int nb_sectors;
66     bool is_write;
67     QLIST_ENTRY(BdrvTrackedRequest) list;
68     Coroutine *co; /* owner, used for deadlock detection */
69     CoQueue wait_queue; /* coroutines blocked on this request */
70 } BdrvTrackedRequest;
71 
72 
73 typedef struct BlockIOLimit {
74     int64_t bps[3];
75     int64_t iops[3];
76 } BlockIOLimit;
77 
78 typedef struct BlockIOBaseValue {
79     uint64_t bytes[2];
80     uint64_t ios[2];
81 } BlockIOBaseValue;
82 
83 struct BlockDriver {
84     const char *format_name;
85     int instance_size;
86     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
87     int (*bdrv_probe_device)(const char *filename);
88 
89     /* Any driver implementing this callback is expected to be able to handle
90      * NULL file names in its .bdrv_open() implementation */
91     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
92 
93     /* For handling image reopen for split or non-split files */
94     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
95                                BlockReopenQueue *queue, Error **errp);
96     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
97     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
98 
99     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags);
100     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags);
101     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
102                      uint8_t *buf, int nb_sectors);
103     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
104                       const uint8_t *buf, int nb_sectors);
105     void (*bdrv_close)(BlockDriverState *bs);
106     void (*bdrv_rebind)(BlockDriverState *bs);
107     int (*bdrv_create)(const char *filename, QEMUOptionParameter *options);
108     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
109     int (*bdrv_make_empty)(BlockDriverState *bs);
110     /* aio */
111     BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
112         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
113         BlockDriverCompletionFunc *cb, void *opaque);
114     BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
115         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
116         BlockDriverCompletionFunc *cb, void *opaque);
117     BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
118         BlockDriverCompletionFunc *cb, void *opaque);
119     BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
120         int64_t sector_num, int nb_sectors,
121         BlockDriverCompletionFunc *cb, void *opaque);
122 
123     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
124         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
125     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
126         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
127     /*
128      * Efficiently zero a region of the disk image.  Typically an image format
129      * would use a compact metadata representation to implement this.  This
130      * function pointer may be NULL and .bdrv_co_writev() will be called
131      * instead.
132      */
133     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
134         int64_t sector_num, int nb_sectors);
135     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
136         int64_t sector_num, int nb_sectors);
137     int coroutine_fn (*bdrv_co_is_allocated)(BlockDriverState *bs,
138         int64_t sector_num, int nb_sectors, int *pnum);
139 
140     /*
141      * Invalidate any cached meta-data.
142      */
143     void (*bdrv_invalidate_cache)(BlockDriverState *bs);
144 
145     /*
146      * Flushes all data that was already written to the OS all the way down to
147      * the disk (for example raw-posix calls fsync()).
148      */
149     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
150 
151     /*
152      * Flushes all internal caches to the OS. The data may still sit in a
153      * writeback cache of the host OS, but it will survive a crash of the qemu
154      * process.
155      */
156     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
157 
158     const char *protocol_name;
159     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
160     int64_t (*bdrv_getlength)(BlockDriverState *bs);
161     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
162     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
163                                  const uint8_t *buf, int nb_sectors);
164 
165     int (*bdrv_snapshot_create)(BlockDriverState *bs,
166                                 QEMUSnapshotInfo *sn_info);
167     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
168                               const char *snapshot_id);
169     int (*bdrv_snapshot_delete)(BlockDriverState *bs, const char *snapshot_id);
170     int (*bdrv_snapshot_list)(BlockDriverState *bs,
171                               QEMUSnapshotInfo **psn_info);
172     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
173                                   const char *snapshot_name);
174     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
175 
176     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
177                              int64_t pos);
178     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
179                              int64_t pos, int size);
180 
181     int (*bdrv_change_backing_file)(BlockDriverState *bs,
182         const char *backing_file, const char *backing_fmt);
183 
184     /* removable device specific */
185     int (*bdrv_is_inserted)(BlockDriverState *bs);
186     int (*bdrv_media_changed)(BlockDriverState *bs);
187     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
188     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
189 
190     /* to control generic scsi devices */
191     int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
192     BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
193         unsigned long int req, void *buf,
194         BlockDriverCompletionFunc *cb, void *opaque);
195 
196     /* List of options for creating images, terminated by name == NULL */
197     QEMUOptionParameter *create_options;
198 
199 
200     /*
201      * Returns 0 for completed check, -errno for internal errors.
202      * The check results are stored in result.
203      */
204     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
205         BdrvCheckMode fix);
206 
207     void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
208 
209     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
210     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
211         const char *tag);
212     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
213     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
214 
215     /*
216      * Returns 1 if newly created images are guaranteed to contain only
217      * zeros, 0 otherwise.
218      */
219     int (*bdrv_has_zero_init)(BlockDriverState *bs);
220 
221     QLIST_ENTRY(BlockDriver) list;
222 };
223 
224 /*
225  * Note: the function bdrv_append() copies and swaps contents of
226  * BlockDriverStates, so if you add new fields to this struct, please
227  * inspect bdrv_append() to determine if the new fields need to be
228  * copied as well.
229  */
230 struct BlockDriverState {
231     int64_t total_sectors; /* if we are reading a disk image, give its
232                               size in sectors */
233     int read_only; /* if true, the media is read only */
234     int open_flags; /* flags used to open the file, re-used for re-open */
235     int encrypted; /* if true, the media is encrypted */
236     int valid_key; /* if true, a valid encryption key has been set */
237     int sg;        /* if true, the device is a /dev/sg* */
238     int copy_on_read; /* if true, copy read backing sectors into image
239                          note this is a reference count */
240 
241     BlockDriver *drv; /* NULL means no media */
242     void *opaque;
243 
244     void *dev;                  /* attached device model, if any */
245     /* TODO change to DeviceState when all users are qdevified */
246     const BlockDevOps *dev_ops;
247     void *dev_opaque;
248 
249     char filename[1024];
250     char backing_file[1024]; /* if non zero, the image is a diff of
251                                 this file image */
252     char backing_format[16]; /* if non-zero and backing_file exists */
253     int is_temporary;
254 
255     BlockDriverState *backing_hd;
256     BlockDriverState *file;
257 
258     NotifierList close_notifiers;
259 
260     /* Callback before write request is processed */
261     NotifierWithReturnList before_write_notifiers;
262 
263     /* number of in-flight copy-on-read requests */
264     unsigned int copy_on_read_in_flight;
265 
266     /* the time for latest disk I/O */
267     int64_t slice_start;
268     int64_t slice_end;
269     BlockIOLimit io_limits;
270     BlockIOBaseValue slice_submitted;
271     CoQueue      throttled_reqs;
272     QEMUTimer    *block_timer;
273     bool         io_limits_enabled;
274 
275     /* I/O stats (display with "info blockstats"). */
276     uint64_t nr_bytes[BDRV_MAX_IOTYPE];
277     uint64_t nr_ops[BDRV_MAX_IOTYPE];
278     uint64_t total_time_ns[BDRV_MAX_IOTYPE];
279     uint64_t wr_highest_sector;
280 
281     /* Whether the disk can expand beyond total_sectors */
282     int growable;
283 
284     /* the memory alignment required for the buffers handled by this driver */
285     int buffer_alignment;
286 
287     /* do we need to tell the quest if we have a volatile write cache? */
288     int enable_write_cache;
289 
290     /* NOTE: the following infos are only hints for real hardware
291        drivers. They are not used by the block driver */
292     BlockdevOnError on_read_error, on_write_error;
293     bool iostatus_enabled;
294     BlockDeviceIoStatus iostatus;
295     char device_name[32];
296     HBitmap *dirty_bitmap;
297     int in_use; /* users other than guest access, eg. block migration */
298     QTAILQ_ENTRY(BlockDriverState) list;
299 
300     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
301 
302     /* long-running background operation */
303     BlockJob *job;
304 
305     QDict *options;
306 };
307 
308 int get_tmp_filename(char *filename, int size);
309 
310 void bdrv_set_io_limits(BlockDriverState *bs,
311                         BlockIOLimit *io_limits);
312 
313 /**
314  * bdrv_add_before_write_notifier:
315  *
316  * Register a callback that is invoked before write requests are processed but
317  * after any throttling or waiting for overlapping requests.
318  */
319 void bdrv_add_before_write_notifier(BlockDriverState *bs,
320                                     NotifierWithReturn *notifier);
321 
322 /**
323  * bdrv_get_aio_context:
324  *
325  * Returns: the currently bound #AioContext
326  */
327 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
328 
329 #ifdef _WIN32
330 int is_windows_drive(const char *filename);
331 #endif
332 void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv,
333                                enum MonitorEvent ev,
334                                BlockErrorAction action, bool is_read);
335 
336 /**
337  * stream_start:
338  * @bs: Block device to operate on.
339  * @base: Block device that will become the new base, or %NULL to
340  * flatten the whole backing file chain onto @bs.
341  * @base_id: The file name that will be written to @bs as the new
342  * backing file if the job completes.  Ignored if @base is %NULL.
343  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
344  * @on_error: The action to take upon error.
345  * @cb: Completion function for the job.
346  * @opaque: Opaque pointer value passed to @cb.
347  * @errp: Error object.
348  *
349  * Start a streaming operation on @bs.  Clusters that are unallocated
350  * in @bs, but allocated in any image between @base and @bs (both
351  * exclusive) will be written to @bs.  At the end of a successful
352  * streaming job, the backing file of @bs will be changed to
353  * @base_id in the written image and to @base in the live BlockDriverState.
354  */
355 void stream_start(BlockDriverState *bs, BlockDriverState *base,
356                   const char *base_id, int64_t speed, BlockdevOnError on_error,
357                   BlockDriverCompletionFunc *cb,
358                   void *opaque, Error **errp);
359 
360 /**
361  * commit_start:
362  * @bs: Top Block device
363  * @base: Block device that will be written into, and become the new top
364  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
365  * @on_error: The action to take upon error.
366  * @cb: Completion function for the job.
367  * @opaque: Opaque pointer value passed to @cb.
368  * @errp: Error object.
369  *
370  */
371 void commit_start(BlockDriverState *bs, BlockDriverState *base,
372                  BlockDriverState *top, int64_t speed,
373                  BlockdevOnError on_error, BlockDriverCompletionFunc *cb,
374                  void *opaque, Error **errp);
375 
376 /*
377  * mirror_start:
378  * @bs: Block device to operate on.
379  * @target: Block device to write to.
380  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
381  * @granularity: The chosen granularity for the dirty bitmap.
382  * @buf_size: The amount of data that can be in flight at one time.
383  * @mode: Whether to collapse all images in the chain to the target.
384  * @on_source_error: The action to take upon error reading from the source.
385  * @on_target_error: The action to take upon error writing to the target.
386  * @cb: Completion function for the job.
387  * @opaque: Opaque pointer value passed to @cb.
388  * @errp: Error object.
389  *
390  * Start a mirroring operation on @bs.  Clusters that are allocated
391  * in @bs will be written to @bs until the job is cancelled or
392  * manually completed.  At the end of a successful mirroring job,
393  * @bs will be switched to read from @target.
394  */
395 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
396                   int64_t speed, int64_t granularity, int64_t buf_size,
397                   MirrorSyncMode mode, BlockdevOnError on_source_error,
398                   BlockdevOnError on_target_error,
399                   BlockDriverCompletionFunc *cb,
400                   void *opaque, Error **errp);
401 
402 /*
403  * backup_start:
404  * @bs: Block device to operate on.
405  * @target: Block device to write to.
406  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
407  * @sync_mode: What parts of the disk image should be copied to the destination.
408  * @on_source_error: The action to take upon error reading from the source.
409  * @on_target_error: The action to take upon error writing to the target.
410  * @cb: Completion function for the job.
411  * @opaque: Opaque pointer value passed to @cb.
412  *
413  * Start a backup operation on @bs.  Clusters in @bs are written to @target
414  * until the job is cancelled or manually completed.
415  */
416 void backup_start(BlockDriverState *bs, BlockDriverState *target,
417                   int64_t speed, MirrorSyncMode sync_mode,
418                   BlockdevOnError on_source_error,
419                   BlockdevOnError on_target_error,
420                   BlockDriverCompletionFunc *cb, void *opaque,
421                   Error **errp);
422 
423 #endif /* BLOCK_INT_H */
424