]> git.proxmox.com Git - qemu.git/blob - block_int.h
block: add I/O throttling algorithm
[qemu.git] / 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
54 typedef struct AIOPool {
55 void (*cancel)(BlockDriverAIOCB *acb);
56 int aiocb_size;
57 BlockDriverAIOCB *free_aiocb;
58 } AIOPool;
59
60 typedef struct BlockIOLimit {
61 int64_t bps[3];
62 int64_t iops[3];
63 } BlockIOLimit;
64
65 typedef struct BlockIOBaseValue {
66 uint64_t bytes[2];
67 uint64_t ios[2];
68 } BlockIOBaseValue;
69
70 struct BlockDriver {
71 const char *format_name;
72 int instance_size;
73 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
74 int (*bdrv_probe_device)(const char *filename);
75 int (*bdrv_open)(BlockDriverState *bs, int flags);
76 int (*bdrv_file_open)(BlockDriverState *bs, const char *filename, int flags);
77 int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
78 uint8_t *buf, int nb_sectors);
79 int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
80 const uint8_t *buf, int nb_sectors);
81 void (*bdrv_close)(BlockDriverState *bs);
82 int (*bdrv_create)(const char *filename, QEMUOptionParameter *options);
83 int (*bdrv_is_allocated)(BlockDriverState *bs, int64_t sector_num,
84 int nb_sectors, int *pnum);
85 int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
86 int (*bdrv_make_empty)(BlockDriverState *bs);
87 /* aio */
88 BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
89 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
90 BlockDriverCompletionFunc *cb, void *opaque);
91 BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
92 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
93 BlockDriverCompletionFunc *cb, void *opaque);
94 BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
95 BlockDriverCompletionFunc *cb, void *opaque);
96 BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
97 int64_t sector_num, int nb_sectors,
98 BlockDriverCompletionFunc *cb, void *opaque);
99
100 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
101 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
102 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
103 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
104 int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
105 int64_t sector_num, int nb_sectors);
106
107 /*
108 * Invalidate any cached meta-data.
109 */
110 void (*bdrv_invalidate_cache)(BlockDriverState *bs);
111
112 /*
113 * Flushes all data that was already written to the OS all the way down to
114 * the disk (for example raw-posix calls fsync()).
115 */
116 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
117
118 /*
119 * Flushes all internal caches to the OS. The data may still sit in a
120 * writeback cache of the host OS, but it will survive a crash of the qemu
121 * process.
122 */
123 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
124
125 int (*bdrv_aio_multiwrite)(BlockDriverState *bs, BlockRequest *reqs,
126 int num_reqs);
127 int (*bdrv_merge_requests)(BlockDriverState *bs, BlockRequest* a,
128 BlockRequest *b);
129
130
131 const char *protocol_name;
132 int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
133 int64_t (*bdrv_getlength)(BlockDriverState *bs);
134 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
135 int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
136 const uint8_t *buf, int nb_sectors);
137
138 int (*bdrv_snapshot_create)(BlockDriverState *bs,
139 QEMUSnapshotInfo *sn_info);
140 int (*bdrv_snapshot_goto)(BlockDriverState *bs,
141 const char *snapshot_id);
142 int (*bdrv_snapshot_delete)(BlockDriverState *bs, const char *snapshot_id);
143 int (*bdrv_snapshot_list)(BlockDriverState *bs,
144 QEMUSnapshotInfo **psn_info);
145 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
146 const char *snapshot_name);
147 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
148
149 int (*bdrv_save_vmstate)(BlockDriverState *bs, const uint8_t *buf,
150 int64_t pos, int size);
151 int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
152 int64_t pos, int size);
153
154 int (*bdrv_change_backing_file)(BlockDriverState *bs,
155 const char *backing_file, const char *backing_fmt);
156
157 /* removable device specific */
158 int (*bdrv_is_inserted)(BlockDriverState *bs);
159 int (*bdrv_media_changed)(BlockDriverState *bs);
160 void (*bdrv_eject)(BlockDriverState *bs, int eject_flag);
161 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
162
163 /* to control generic scsi devices */
164 int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
165 BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
166 unsigned long int req, void *buf,
167 BlockDriverCompletionFunc *cb, void *opaque);
168
169 /* List of options for creating images, terminated by name == NULL */
170 QEMUOptionParameter *create_options;
171
172
173 /*
174 * Returns 0 for completed check, -errno for internal errors.
175 * The check results are stored in result.
176 */
177 int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result);
178
179 void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
180
181 /*
182 * Returns 1 if newly created images are guaranteed to contain only
183 * zeros, 0 otherwise.
184 */
185 int (*bdrv_has_zero_init)(BlockDriverState *bs);
186
187 QLIST_ENTRY(BlockDriver) list;
188 };
189
190 struct BlockDriverState {
191 int64_t total_sectors; /* if we are reading a disk image, give its
192 size in sectors */
193 int read_only; /* if true, the media is read only */
194 int keep_read_only; /* if true, the media was requested to stay read only */
195 int open_flags; /* flags used to open the file, re-used for re-open */
196 int encrypted; /* if true, the media is encrypted */
197 int valid_key; /* if true, a valid encryption key has been set */
198 int sg; /* if true, the device is a /dev/sg* */
199
200 BlockDriver *drv; /* NULL means no media */
201 void *opaque;
202
203 void *dev; /* attached device model, if any */
204 /* TODO change to DeviceState when all users are qdevified */
205 const BlockDevOps *dev_ops;
206 void *dev_opaque;
207
208 char filename[1024];
209 char backing_file[1024]; /* if non zero, the image is a diff of
210 this file image */
211 char backing_format[16]; /* if non-zero and backing_file exists */
212 int is_temporary;
213
214 BlockDriverState *backing_hd;
215 BlockDriverState *file;
216
217 /* async read/write emulation */
218
219 void *sync_aiocb;
220
221 /* the time for latest disk I/O */
222 int64_t slice_time;
223 int64_t slice_start;
224 int64_t slice_end;
225 BlockIOLimit io_limits;
226 BlockIOBaseValue io_base;
227 CoQueue throttled_reqs;
228 QEMUTimer *block_timer;
229 bool io_limits_enabled;
230
231 /* I/O stats (display with "info blockstats"). */
232 uint64_t nr_bytes[BDRV_MAX_IOTYPE];
233 uint64_t nr_ops[BDRV_MAX_IOTYPE];
234 uint64_t total_time_ns[BDRV_MAX_IOTYPE];
235 uint64_t wr_highest_sector;
236
237 /* Whether the disk can expand beyond total_sectors */
238 int growable;
239
240 /* the memory alignment required for the buffers handled by this driver */
241 int buffer_alignment;
242
243 /* do we need to tell the quest if we have a volatile write cache? */
244 int enable_write_cache;
245
246 /* NOTE: the following infos are only hints for real hardware
247 drivers. They are not used by the block driver */
248 int cyls, heads, secs, translation;
249 BlockErrorAction on_read_error, on_write_error;
250 bool iostatus_enabled;
251 BlockDeviceIoStatus iostatus;
252 char device_name[32];
253 unsigned long *dirty_bitmap;
254 int64_t dirty_count;
255 int in_use; /* users other than guest access, eg. block migration */
256 QTAILQ_ENTRY(BlockDriverState) list;
257 void *private;
258 };
259
260 struct BlockDriverAIOCB {
261 AIOPool *pool;
262 BlockDriverState *bs;
263 BlockDriverCompletionFunc *cb;
264 void *opaque;
265 BlockDriverAIOCB *next;
266 };
267
268 void get_tmp_filename(char *filename, int size);
269
270 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
271 BlockDriverCompletionFunc *cb, void *opaque);
272 void qemu_aio_release(void *p);
273
274 void bdrv_set_io_limits(BlockDriverState *bs,
275 BlockIOLimit *io_limits);
276
277 #ifdef _WIN32
278 int is_windows_drive(const char *filename);
279 #endif
280
281 #endif /* BLOCK_INT_H */