xref: /openbmc/qemu/include/block/block_int.h (revision 8917c3bd)
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     /* For handling image reopen for split or non-split files */
80     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
81                                BlockReopenQueue *queue, Error **errp);
82     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
83     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
84 
85     int (*bdrv_open)(BlockDriverState *bs, int flags);
86     int (*bdrv_file_open)(BlockDriverState *bs, const char *filename, int flags);
87     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
88                      uint8_t *buf, int nb_sectors);
89     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
90                       const uint8_t *buf, int nb_sectors);
91     void (*bdrv_close)(BlockDriverState *bs);
92     void (*bdrv_rebind)(BlockDriverState *bs);
93     int (*bdrv_create)(const char *filename, QEMUOptionParameter *options);
94     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
95     int (*bdrv_make_empty)(BlockDriverState *bs);
96     /* aio */
97     BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
98         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
99         BlockDriverCompletionFunc *cb, void *opaque);
100     BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
101         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
102         BlockDriverCompletionFunc *cb, void *opaque);
103     BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
104         BlockDriverCompletionFunc *cb, void *opaque);
105     BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
106         int64_t sector_num, int nb_sectors,
107         BlockDriverCompletionFunc *cb, void *opaque);
108 
109     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
110         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
111     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
112         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
113     /*
114      * Efficiently zero a region of the disk image.  Typically an image format
115      * would use a compact metadata representation to implement this.  This
116      * function pointer may be NULL and .bdrv_co_writev() will be called
117      * instead.
118      */
119     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
120         int64_t sector_num, int nb_sectors);
121     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
122         int64_t sector_num, int nb_sectors);
123     int coroutine_fn (*bdrv_co_is_allocated)(BlockDriverState *bs,
124         int64_t sector_num, int nb_sectors, int *pnum);
125 
126     /*
127      * Invalidate any cached meta-data.
128      */
129     void (*bdrv_invalidate_cache)(BlockDriverState *bs);
130 
131     /*
132      * Flushes all data that was already written to the OS all the way down to
133      * the disk (for example raw-posix calls fsync()).
134      */
135     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
136 
137     /*
138      * Flushes all internal caches to the OS. The data may still sit in a
139      * writeback cache of the host OS, but it will survive a crash of the qemu
140      * process.
141      */
142     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
143 
144     const char *protocol_name;
145     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
146     int64_t (*bdrv_getlength)(BlockDriverState *bs);
147     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
148     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
149                                  const uint8_t *buf, int nb_sectors);
150 
151     int (*bdrv_snapshot_create)(BlockDriverState *bs,
152                                 QEMUSnapshotInfo *sn_info);
153     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
154                               const char *snapshot_id);
155     int (*bdrv_snapshot_delete)(BlockDriverState *bs, const char *snapshot_id);
156     int (*bdrv_snapshot_list)(BlockDriverState *bs,
157                               QEMUSnapshotInfo **psn_info);
158     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
159                                   const char *snapshot_name);
160     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
161 
162     int (*bdrv_save_vmstate)(BlockDriverState *bs, const uint8_t *buf,
163                              int64_t pos, int size);
164     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
165                              int64_t pos, int size);
166 
167     int (*bdrv_change_backing_file)(BlockDriverState *bs,
168         const char *backing_file, const char *backing_fmt);
169 
170     /* removable device specific */
171     int (*bdrv_is_inserted)(BlockDriverState *bs);
172     int (*bdrv_media_changed)(BlockDriverState *bs);
173     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
174     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
175 
176     /* to control generic scsi devices */
177     int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
178     BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
179         unsigned long int req, void *buf,
180         BlockDriverCompletionFunc *cb, void *opaque);
181 
182     /* List of options for creating images, terminated by name == NULL */
183     QEMUOptionParameter *create_options;
184 
185 
186     /*
187      * Returns 0 for completed check, -errno for internal errors.
188      * The check results are stored in result.
189      */
190     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
191         BdrvCheckMode fix);
192 
193     void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
194 
195     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
196     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
197         const char *tag);
198     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
199     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
200 
201     /*
202      * Returns 1 if newly created images are guaranteed to contain only
203      * zeros, 0 otherwise.
204      */
205     int (*bdrv_has_zero_init)(BlockDriverState *bs);
206 
207     QLIST_ENTRY(BlockDriver) list;
208 };
209 
210 /*
211  * Note: the function bdrv_append() copies and swaps contents of
212  * BlockDriverStates, so if you add new fields to this struct, please
213  * inspect bdrv_append() to determine if the new fields need to be
214  * copied as well.
215  */
216 struct BlockDriverState {
217     int64_t total_sectors; /* if we are reading a disk image, give its
218                               size in sectors */
219     int read_only; /* if true, the media is read only */
220     int open_flags; /* flags used to open the file, re-used for re-open */
221     int encrypted; /* if true, the media is encrypted */
222     int valid_key; /* if true, a valid encryption key has been set */
223     int sg;        /* if true, the device is a /dev/sg* */
224     int copy_on_read; /* if true, copy read backing sectors into image
225                          note this is a reference count */
226 
227     BlockDriver *drv; /* NULL means no media */
228     void *opaque;
229 
230     void *dev;                  /* attached device model, if any */
231     /* TODO change to DeviceState when all users are qdevified */
232     const BlockDevOps *dev_ops;
233     void *dev_opaque;
234 
235     char filename[1024];
236     char backing_file[1024]; /* if non zero, the image is a diff of
237                                 this file image */
238     char backing_format[16]; /* if non-zero and backing_file exists */
239     int is_temporary;
240 
241     BlockDriverState *backing_hd;
242     BlockDriverState *file;
243 
244     NotifierList close_notifiers;
245 
246     /* number of in-flight copy-on-read requests */
247     unsigned int copy_on_read_in_flight;
248 
249     /* the time for latest disk I/O */
250     int64_t slice_time;
251     int64_t slice_start;
252     int64_t slice_end;
253     BlockIOLimit io_limits;
254     BlockIOBaseValue  io_base;
255     CoQueue      throttled_reqs;
256     QEMUTimer    *block_timer;
257     bool         io_limits_enabled;
258 
259     /* I/O stats (display with "info blockstats"). */
260     uint64_t nr_bytes[BDRV_MAX_IOTYPE];
261     uint64_t nr_ops[BDRV_MAX_IOTYPE];
262     uint64_t total_time_ns[BDRV_MAX_IOTYPE];
263     uint64_t wr_highest_sector;
264 
265     /* Whether the disk can expand beyond total_sectors */
266     int growable;
267 
268     /* the memory alignment required for the buffers handled by this driver */
269     int buffer_alignment;
270 
271     /* do we need to tell the quest if we have a volatile write cache? */
272     int enable_write_cache;
273 
274     /* NOTE: the following infos are only hints for real hardware
275        drivers. They are not used by the block driver */
276     BlockdevOnError on_read_error, on_write_error;
277     bool iostatus_enabled;
278     BlockDeviceIoStatus iostatus;
279     char device_name[32];
280     HBitmap *dirty_bitmap;
281     int in_use; /* users other than guest access, eg. block migration */
282     QTAILQ_ENTRY(BlockDriverState) list;
283 
284     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
285 
286     /* long-running background operation */
287     BlockJob *job;
288 
289 };
290 
291 int get_tmp_filename(char *filename, int size);
292 
293 void bdrv_set_io_limits(BlockDriverState *bs,
294                         BlockIOLimit *io_limits);
295 
296 #ifdef _WIN32
297 int is_windows_drive(const char *filename);
298 #endif
299 void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv,
300                                enum MonitorEvent ev,
301                                BlockErrorAction action, bool is_read);
302 
303 /**
304  * stream_start:
305  * @bs: Block device to operate on.
306  * @base: Block device that will become the new base, or %NULL to
307  * flatten the whole backing file chain onto @bs.
308  * @base_id: The file name that will be written to @bs as the new
309  * backing file if the job completes.  Ignored if @base is %NULL.
310  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
311  * @on_error: The action to take upon error.
312  * @cb: Completion function for the job.
313  * @opaque: Opaque pointer value passed to @cb.
314  * @errp: Error object.
315  *
316  * Start a streaming operation on @bs.  Clusters that are unallocated
317  * in @bs, but allocated in any image between @base and @bs (both
318  * exclusive) will be written to @bs.  At the end of a successful
319  * streaming job, the backing file of @bs will be changed to
320  * @base_id in the written image and to @base in the live BlockDriverState.
321  */
322 void stream_start(BlockDriverState *bs, BlockDriverState *base,
323                   const char *base_id, int64_t speed, BlockdevOnError on_error,
324                   BlockDriverCompletionFunc *cb,
325                   void *opaque, Error **errp);
326 
327 /**
328  * commit_start:
329  * @bs: Top Block device
330  * @base: Block device that will be written into, and become the new top
331  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
332  * @on_error: The action to take upon error.
333  * @cb: Completion function for the job.
334  * @opaque: Opaque pointer value passed to @cb.
335  * @errp: Error object.
336  *
337  */
338 void commit_start(BlockDriverState *bs, BlockDriverState *base,
339                  BlockDriverState *top, int64_t speed,
340                  BlockdevOnError on_error, BlockDriverCompletionFunc *cb,
341                  void *opaque, Error **errp);
342 
343 /*
344  * mirror_start:
345  * @bs: Block device to operate on.
346  * @target: Block device to write to.
347  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
348  * @granularity: The chosen granularity for the dirty bitmap.
349  * @buf_size: The amount of data that can be in flight at one time.
350  * @mode: Whether to collapse all images in the chain to the target.
351  * @on_source_error: The action to take upon error reading from the source.
352  * @on_target_error: The action to take upon error writing to the target.
353  * @cb: Completion function for the job.
354  * @opaque: Opaque pointer value passed to @cb.
355  * @errp: Error object.
356  *
357  * Start a mirroring operation on @bs.  Clusters that are allocated
358  * in @bs will be written to @bs until the job is cancelled or
359  * manually completed.  At the end of a successful mirroring job,
360  * @bs will be switched to read from @target.
361  */
362 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
363                   int64_t speed, int64_t granularity, int64_t buf_size,
364                   MirrorSyncMode mode, BlockdevOnError on_source_error,
365                   BlockdevOnError on_target_error,
366                   BlockDriverCompletionFunc *cb,
367                   void *opaque, Error **errp);
368 
369 #endif /* BLOCK_INT_H */
370