jcs.org
qemu with hax to log dma reads & writes
jcs.org/2018/11/12/vfio
1#ifndef BLOCK_H
2#define BLOCK_H
3
4#include "block/aio.h"
5#include "block/aio-wait.h"
6#include "qemu/iov.h"
7#include "qemu/coroutine.h"
8#include "block/accounting.h"
9#include "block/dirty-bitmap.h"
10#include "block/blockjob.h"
11#include "qemu/hbitmap.h"
12
13/* block.c */
14typedef struct BlockDriver BlockDriver;
15typedef struct BdrvChild BdrvChild;
16typedef struct BdrvChildClass BdrvChildClass;
17
18typedef struct BlockDriverInfo {
19 /* in bytes, 0 if irrelevant */
20 int cluster_size;
21 /* offset at which the VM state can be saved (0 if not possible) */
22 int64_t vm_state_offset;
23 bool is_dirty;
24 /*
25 * True if this block driver only supports compressed writes
26 */
27 bool needs_compressed_writes;
28} BlockDriverInfo;
29
30typedef struct BlockFragInfo {
31 uint64_t allocated_clusters;
32 uint64_t total_clusters;
33 uint64_t fragmented_clusters;
34 uint64_t compressed_clusters;
35} BlockFragInfo;
36
37typedef enum {
38 BDRV_REQ_COPY_ON_READ = 0x1,
39 BDRV_REQ_ZERO_WRITE = 0x2,
40
41 /*
42 * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
43 * that the block driver should unmap (discard) blocks if it is guaranteed
44 * that the result will read back as zeroes. The flag is only passed to the
45 * driver if the block device is opened with BDRV_O_UNMAP.
46 */
47 BDRV_REQ_MAY_UNMAP = 0x4,
48
49 BDRV_REQ_FUA = 0x10,
50 BDRV_REQ_WRITE_COMPRESSED = 0x20,
51
52 /* Signifies that this write request will not change the visible disk
53 * content. */
54 BDRV_REQ_WRITE_UNCHANGED = 0x40,
55
56 /*
57 * BDRV_REQ_SERIALISING forces request serialisation for writes.
58 * It is used to ensure that writes to the backing file of a backup process
59 * target cannot race with a read of the backup target that defers to the
60 * backing file.
61 *
62 * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
63 * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
64 * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
65 */
66 BDRV_REQ_SERIALISING = 0x80,
67
68 /* Execute the request only if the operation can be offloaded or otherwise
69 * be executed efficiently, but return an error instead of using a slow
70 * fallback. */
71 BDRV_REQ_NO_FALLBACK = 0x100,
72
73 /*
74 * BDRV_REQ_PREFETCH may be used only together with BDRV_REQ_COPY_ON_READ
75 * on read request and means that caller doesn't really need data to be
76 * written to qiov parameter which may be NULL.
77 */
78 BDRV_REQ_PREFETCH = 0x200,
79 /* Mask of valid flags */
80 BDRV_REQ_MASK = 0x3ff,
81} BdrvRequestFlags;
82
83typedef struct BlockSizes {
84 uint32_t phys;
85 uint32_t log;
86} BlockSizes;
87
88typedef struct HDGeometry {
89 uint32_t heads;
90 uint32_t sectors;
91 uint32_t cylinders;
92} HDGeometry;
93
94#define BDRV_O_RDWR 0x0002
95#define BDRV_O_RESIZE 0x0004 /* request permission for resizing the node */
96#define BDRV_O_SNAPSHOT 0x0008 /* open the file read only and save writes in a snapshot */
97#define BDRV_O_TEMPORARY 0x0010 /* delete the file after use */
98#define BDRV_O_NOCACHE 0x0020 /* do not use the host page cache */
99#define BDRV_O_NATIVE_AIO 0x0080 /* use native AIO instead of the thread pool */
100#define BDRV_O_NO_BACKING 0x0100 /* don't open the backing file */
101#define BDRV_O_NO_FLUSH 0x0200 /* disable flushing on this disk */
102#define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
103#define BDRV_O_INACTIVE 0x0800 /* consistency hint for migration handoff */
104#define BDRV_O_CHECK 0x1000 /* open solely for consistency check */
105#define BDRV_O_ALLOW_RDWR 0x2000 /* allow reopen to change from r/o to r/w */
106#define BDRV_O_UNMAP 0x4000 /* execute guest UNMAP/TRIM operations */
107#define BDRV_O_PROTOCOL 0x8000 /* if no block driver is explicitly given:
108 select an appropriate protocol driver,
109 ignoring the format layer */
110#define BDRV_O_NO_IO 0x10000 /* don't initialize for I/O */
111#define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
112#define BDRV_O_IO_URING 0x40000 /* use io_uring instead of the thread pool */
113
114#define BDRV_O_CACHE_MASK (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
115
116
117/* Option names of options parsed by the block layer */
118
119#define BDRV_OPT_CACHE_WB "cache.writeback"
120#define BDRV_OPT_CACHE_DIRECT "cache.direct"
121#define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
122#define BDRV_OPT_READ_ONLY "read-only"
123#define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
124#define BDRV_OPT_DISCARD "discard"
125#define BDRV_OPT_FORCE_SHARE "force-share"
126
127
128#define BDRV_SECTOR_BITS 9
129#define BDRV_SECTOR_SIZE (1ULL << BDRV_SECTOR_BITS)
130
131#define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
132 INT_MAX >> BDRV_SECTOR_BITS)
133#define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
134
135/*
136 * Allocation status flags for bdrv_block_status() and friends.
137 *
138 * Public flags:
139 * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
140 * BDRV_BLOCK_ZERO: offset reads as zero
141 * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
142 * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
143 * layer rather than any backing, set by block layer
144 * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
145 * layer, set by block layer
146 *
147 * Internal flags:
148 * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
149 * that the block layer recompute the answer from the returned
150 * BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
151 * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
152 * zeroes in file child of current block node inside
153 * returned region. Only valid together with both
154 * BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
155 * appear with BDRV_BLOCK_ZERO.
156 *
157 * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
158 * host offset within the returned BDS that is allocated for the
159 * corresponding raw guest data. However, whether that offset
160 * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
161 *
162 * DATA ZERO OFFSET_VALID
163 * t t t sectors read as zero, returned file is zero at offset
164 * t f t sectors read as valid from file at offset
165 * f t t sectors preallocated, read as zero, returned file not
166 * necessarily zero at offset
167 * f f t sectors preallocated but read from backing_hd,
168 * returned file contains garbage at offset
169 * t t f sectors preallocated, read as zero, unknown offset
170 * t f f sectors read from unknown file or offset
171 * f t f not allocated or unknown offset, read as zero
172 * f f f not allocated or unknown offset, read from backing_hd
173 */
174#define BDRV_BLOCK_DATA 0x01
175#define BDRV_BLOCK_ZERO 0x02
176#define BDRV_BLOCK_OFFSET_VALID 0x04
177#define BDRV_BLOCK_RAW 0x08
178#define BDRV_BLOCK_ALLOCATED 0x10
179#define BDRV_BLOCK_EOF 0x20
180#define BDRV_BLOCK_RECURSE 0x40
181
182typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
183
184typedef struct BDRVReopenState {
185 BlockDriverState *bs;
186 int flags;
187 BlockdevDetectZeroesOptions detect_zeroes;
188 bool backing_missing;
189 bool replace_backing_bs; /* new_backing_bs is ignored if this is false */
190 BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
191 uint64_t perm, shared_perm;
192 QDict *options;
193 QDict *explicit_options;
194 void *opaque;
195} BDRVReopenState;
196
197/*
198 * Block operation types
199 */
200typedef enum BlockOpType {
201 BLOCK_OP_TYPE_BACKUP_SOURCE,
202 BLOCK_OP_TYPE_BACKUP_TARGET,
203 BLOCK_OP_TYPE_CHANGE,
204 BLOCK_OP_TYPE_COMMIT_SOURCE,
205 BLOCK_OP_TYPE_COMMIT_TARGET,
206 BLOCK_OP_TYPE_DATAPLANE,
207 BLOCK_OP_TYPE_DRIVE_DEL,
208 BLOCK_OP_TYPE_EJECT,
209 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
210 BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
211 BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
212 BLOCK_OP_TYPE_MIRROR_SOURCE,
213 BLOCK_OP_TYPE_MIRROR_TARGET,
214 BLOCK_OP_TYPE_RESIZE,
215 BLOCK_OP_TYPE_STREAM,
216 BLOCK_OP_TYPE_REPLACE,
217 BLOCK_OP_TYPE_MAX,
218} BlockOpType;
219
220/* Block node permission constants */
221enum {
222 /**
223 * A user that has the "permission" of consistent reads is guaranteed that
224 * their view of the contents of the block device is complete and
225 * self-consistent, representing the contents of a disk at a specific
226 * point.
227 *
228 * For most block devices (including their backing files) this is true, but
229 * the property cannot be maintained in a few situations like for
230 * intermediate nodes of a commit block job.
231 */
232 BLK_PERM_CONSISTENT_READ = 0x01,
233
234 /** This permission is required to change the visible disk contents. */
235 BLK_PERM_WRITE = 0x02,
236
237 /**
238 * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
239 * required for writes to the block node when the caller promises that
240 * the visible disk content doesn't change.
241 *
242 * As the BLK_PERM_WRITE permission is strictly stronger, either is
243 * sufficient to perform an unchanging write.
244 */
245 BLK_PERM_WRITE_UNCHANGED = 0x04,
246
247 /** This permission is required to change the size of a block node. */
248 BLK_PERM_RESIZE = 0x08,
249
250 /**
251 * This permission is required to change the node that this BdrvChild
252 * points to.
253 */
254 BLK_PERM_GRAPH_MOD = 0x10,
255
256 BLK_PERM_ALL = 0x1f,
257
258 DEFAULT_PERM_PASSTHROUGH = BLK_PERM_CONSISTENT_READ
259 | BLK_PERM_WRITE
260 | BLK_PERM_WRITE_UNCHANGED
261 | BLK_PERM_RESIZE,
262
263 DEFAULT_PERM_UNCHANGED = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
264};
265
266/*
267 * Flags that parent nodes assign to child nodes to specify what kind of
268 * role(s) they take.
269 *
270 * At least one of DATA, METADATA, FILTERED, or COW must be set for
271 * every child.
272 */
273enum BdrvChildRoleBits {
274 /*
275 * This child stores data.
276 * Any node may have an arbitrary number of such children.
277 */
278 BDRV_CHILD_DATA = (1 << 0),
279
280 /*
281 * This child stores metadata.
282 * Any node may have an arbitrary number of metadata-storing
283 * children.
284 */
285 BDRV_CHILD_METADATA = (1 << 1),
286
287 /*
288 * A child that always presents exactly the same visible data as
289 * the parent, e.g. by virtue of the parent forwarding all reads
290 * and writes.
291 * This flag is mutually exclusive with DATA, METADATA, and COW.
292 * Any node may have at most one filtered child at a time.
293 */
294 BDRV_CHILD_FILTERED = (1 << 2),
295
296 /*
297 * Child from which to read all data that isn’t allocated in the
298 * parent (i.e., the backing child); such data is copied to the
299 * parent through COW (and optionally COR).
300 * This field is mutually exclusive with DATA, METADATA, and
301 * FILTERED.
302 * Any node may have at most one such backing child at a time.
303 */
304 BDRV_CHILD_COW = (1 << 3),
305
306 /*
307 * The primary child. For most drivers, this is the child whose
308 * filename applies best to the parent node.
309 * Any node may have at most one primary child at a time.
310 */
311 BDRV_CHILD_PRIMARY = (1 << 4),
312
313 /* Useful combination of flags */
314 BDRV_CHILD_IMAGE = BDRV_CHILD_DATA
315 | BDRV_CHILD_METADATA
316 | BDRV_CHILD_PRIMARY,
317};
318
319/* Mask of BdrvChildRoleBits values */
320typedef unsigned int BdrvChildRole;
321
322char *bdrv_perm_names(uint64_t perm);
323uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
324
325/* disk I/O throttling */
326void bdrv_init(void);
327void bdrv_init_with_whitelist(void);
328bool bdrv_uses_whitelist(void);
329int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
330BlockDriver *bdrv_find_protocol(const char *filename,
331 bool allow_protocol_prefix,
332 Error **errp);
333BlockDriver *bdrv_find_format(const char *format_name);
334int bdrv_create(BlockDriver *drv, const char* filename,
335 QemuOpts *opts, Error **errp);
336int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
337
338BlockDriverState *bdrv_new(void);
339void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
340 Error **errp);
341void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
342 Error **errp);
343
344int bdrv_parse_aio(const char *mode, int *flags);
345int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
346int bdrv_parse_discard_flags(const char *mode, int *flags);
347BdrvChild *bdrv_open_child(const char *filename,
348 QDict *options, const char *bdref_key,
349 BlockDriverState* parent,
350 const BdrvChildClass *child_class,
351 BdrvChildRole child_role,
352 bool allow_none, Error **errp);
353BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
354void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
355 Error **errp);
356int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
357 const char *bdref_key, Error **errp);
358BlockDriverState *bdrv_open(const char *filename, const char *reference,
359 QDict *options, int flags, Error **errp);
360BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
361 int flags, Error **errp);
362BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
363 BlockDriverState *bs, QDict *options,
364 bool keep_old_opts);
365int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
366int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
367 Error **errp);
368int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
369 BlockReopenQueue *queue, Error **errp);
370void bdrv_reopen_commit(BDRVReopenState *reopen_state);
371void bdrv_reopen_abort(BDRVReopenState *reopen_state);
372int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
373 int bytes, BdrvRequestFlags flags);
374int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
375int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
376int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
377int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
378int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
379int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
380 const void *buf, int count);
381/*
382 * Efficiently zero a region of the disk image. Note that this is a regular
383 * I/O request like read or write and should have a reasonable size. This
384 * function is not suitable for zeroing the entire image in a single request
385 * because it may allocate memory for the entire region.
386 */
387int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
388 int bytes, BdrvRequestFlags flags);
389BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
390 const char *backing_file);
391void bdrv_refresh_filename(BlockDriverState *bs);
392
393int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
394 PreallocMode prealloc, BdrvRequestFlags flags,
395 Error **errp);
396int bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
397 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
398
399int64_t bdrv_nb_sectors(BlockDriverState *bs);
400int64_t bdrv_getlength(BlockDriverState *bs);
401int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
402BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
403 BlockDriverState *in_bs, Error **errp);
404void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
405void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
406int bdrv_commit(BlockDriverState *bs);
407int bdrv_make_empty(BdrvChild *c, Error **errp);
408int bdrv_change_backing_file(BlockDriverState *bs,
409 const char *backing_file, const char *backing_fmt);
410void bdrv_register(BlockDriver *bdrv);
411int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
412 const char *backing_file_str);
413BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
414 BlockDriverState *bs);
415BlockDriverState *bdrv_find_base(BlockDriverState *bs);
416bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
417 Error **errp);
418int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
419 Error **errp);
420void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
421int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
422
423
424typedef struct BdrvCheckResult {
425 int corruptions;
426 int leaks;
427 int check_errors;
428 int corruptions_fixed;
429 int leaks_fixed;
430 int64_t image_end_offset;
431 BlockFragInfo bfi;
432} BdrvCheckResult;
433
434typedef enum {
435 BDRV_FIX_LEAKS = 1,
436 BDRV_FIX_ERRORS = 2,
437} BdrvCheckMode;
438
439int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
440
441/* The units of offset and total_work_size may be chosen arbitrarily by the
442 * block driver; total_work_size may change during the course of the amendment
443 * operation */
444typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
445 int64_t total_work_size, void *opaque);
446int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
447 BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
448 bool force,
449 Error **errp);
450
451/* check if a named node can be replaced when doing drive-mirror */
452BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
453 const char *node_name, Error **errp);
454
455/* async block I/O */
456void bdrv_aio_cancel(BlockAIOCB *acb);
457void bdrv_aio_cancel_async(BlockAIOCB *acb);
458
459/* sg packet commands */
460int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
461
462/* Invalidate any cached metadata used by image formats */
463void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
464void bdrv_invalidate_cache_all(Error **errp);
465int bdrv_inactivate_all(void);
466
467/* Ensure contents are flushed to disk. */
468int bdrv_flush(BlockDriverState *bs);
469int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
470int bdrv_flush_all(void);
471void bdrv_close_all(void);
472void bdrv_drain(BlockDriverState *bs);
473void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
474void bdrv_drain_all_begin(void);
475void bdrv_drain_all_end(void);
476void bdrv_drain_all(void);
477
478#define BDRV_POLL_WHILE(bs, cond) ({ \
479 BlockDriverState *bs_ = (bs); \
480 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \
481 cond); })
482
483int bdrv_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
484int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
485int bdrv_has_zero_init_1(BlockDriverState *bs);
486int bdrv_has_zero_init(BlockDriverState *bs);
487bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
488int bdrv_block_status(BlockDriverState *bs, int64_t offset,
489 int64_t bytes, int64_t *pnum, int64_t *map,
490 BlockDriverState **file);
491int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
492 int64_t offset, int64_t bytes, int64_t *pnum,
493 int64_t *map, BlockDriverState **file);
494int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
495 int64_t *pnum);
496int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
497 bool include_base, int64_t offset, int64_t bytes,
498 int64_t *pnum);
499
500bool bdrv_is_read_only(BlockDriverState *bs);
501int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
502 bool ignore_allow_rdw, Error **errp);
503int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
504 Error **errp);
505bool bdrv_is_writable(BlockDriverState *bs);
506bool bdrv_is_sg(BlockDriverState *bs);
507bool bdrv_is_inserted(BlockDriverState *bs);
508void bdrv_lock_medium(BlockDriverState *bs, bool locked);
509void bdrv_eject(BlockDriverState *bs, bool eject_flag);
510const char *bdrv_get_format_name(BlockDriverState *bs);
511BlockDriverState *bdrv_find_node(const char *node_name);
512BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
513XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
514BlockDriverState *bdrv_lookup_bs(const char *device,
515 const char *node_name,
516 Error **errp);
517bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
518BlockDriverState *bdrv_next_node(BlockDriverState *bs);
519BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
520
521typedef struct BdrvNextIterator {
522 enum {
523 BDRV_NEXT_BACKEND_ROOTS,
524 BDRV_NEXT_MONITOR_OWNED,
525 } phase;
526 BlockBackend *blk;
527 BlockDriverState *bs;
528} BdrvNextIterator;
529
530BlockDriverState *bdrv_first(BdrvNextIterator *it);
531BlockDriverState *bdrv_next(BdrvNextIterator *it);
532void bdrv_next_cleanup(BdrvNextIterator *it);
533
534BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
535bool bdrv_is_encrypted(BlockDriverState *bs);
536void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
537 void *opaque, bool read_only);
538const char *bdrv_get_node_name(const BlockDriverState *bs);
539const char *bdrv_get_device_name(const BlockDriverState *bs);
540const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
541int bdrv_get_flags(BlockDriverState *bs);
542int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
543ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
544 Error **errp);
545BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
546void bdrv_round_to_clusters(BlockDriverState *bs,
547 int64_t offset, int64_t bytes,
548 int64_t *cluster_offset,
549 int64_t *cluster_bytes);
550
551void bdrv_get_backing_filename(BlockDriverState *bs,
552 char *filename, int filename_size);
553char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
554char *bdrv_get_full_backing_filename_from_filename(const char *backed,
555 const char *backing,
556 Error **errp);
557char *bdrv_dirname(BlockDriverState *bs, Error **errp);
558
559int path_has_protocol(const char *path);
560int path_is_absolute(const char *path);
561char *path_combine(const char *base_path, const char *filename);
562
563int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
564int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
565int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
566 int64_t pos, int size);
567
568int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
569 int64_t pos, int size);
570
571void bdrv_img_create(const char *filename, const char *fmt,
572 const char *base_filename, const char *base_fmt,
573 char *options, uint64_t img_size, int flags,
574 bool quiet, Error **errp);
575
576/* Returns the alignment in bytes that is required so that no bounce buffer
577 * is required throughout the stack */
578size_t bdrv_min_mem_align(BlockDriverState *bs);
579/* Returns optimal alignment in bytes for bounce buffer */
580size_t bdrv_opt_mem_align(BlockDriverState *bs);
581void *qemu_blockalign(BlockDriverState *bs, size_t size);
582void *qemu_blockalign0(BlockDriverState *bs, size_t size);
583void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
584void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
585bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
586
587void bdrv_enable_copy_on_read(BlockDriverState *bs);
588void bdrv_disable_copy_on_read(BlockDriverState *bs);
589
590void bdrv_ref(BlockDriverState *bs);
591void bdrv_unref(BlockDriverState *bs);
592void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
593BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
594 BlockDriverState *child_bs,
595 const char *child_name,
596 const BdrvChildClass *child_class,
597 BdrvChildRole child_role,
598 Error **errp);
599
600bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
601void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
602void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
603void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
604void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
605bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
606
607#define BLKDBG_EVENT(child, evt) \
608 do { \
609 if (child) { \
610 bdrv_debug_event(child->bs, evt); \
611 } \
612 } while (0)
613
614void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
615
616int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
617 const char *tag);
618int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
619int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
620bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
621
622/**
623 * bdrv_get_aio_context:
624 *
625 * Returns: the currently bound #AioContext
626 */
627AioContext *bdrv_get_aio_context(BlockDriverState *bs);
628
629/**
630 * Transfer control to @co in the aio context of @bs
631 */
632void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
633
634void bdrv_set_aio_context_ignore(BlockDriverState *bs,
635 AioContext *new_context, GSList **ignore);
636int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
637 Error **errp);
638int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
639 BdrvChild *ignore_child, Error **errp);
640bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
641 GSList **ignore, Error **errp);
642bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
643 GSList **ignore, Error **errp);
644int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
645int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
646
647void bdrv_io_plug(BlockDriverState *bs);
648void bdrv_io_unplug(BlockDriverState *bs);
649
650/**
651 * bdrv_parent_drained_begin_single:
652 *
653 * Begin a quiesced section for the parent of @c. If @poll is true, wait for
654 * any pending activity to cease.
655 */
656void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
657
658/**
659 * bdrv_parent_drained_end_single:
660 *
661 * End a quiesced section for the parent of @c.
662 *
663 * This polls @bs's AioContext until all scheduled sub-drained_ends
664 * have settled, which may result in graph changes.
665 */
666void bdrv_parent_drained_end_single(BdrvChild *c);
667
668/**
669 * bdrv_drain_poll:
670 *
671 * Poll for pending requests in @bs, its parents (except for @ignore_parent),
672 * and if @recursive is true its children as well (used for subtree drain).
673 *
674 * If @ignore_bds_parents is true, parents that are BlockDriverStates must
675 * ignore the drain request because they will be drained separately (used for
676 * drain_all).
677 *
678 * This is part of bdrv_drained_begin.
679 */
680bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
681 BdrvChild *ignore_parent, bool ignore_bds_parents);
682
683/**
684 * bdrv_drained_begin:
685 *
686 * Begin a quiesced section for exclusive access to the BDS, by disabling
687 * external request sources including NBD server and device model. Note that
688 * this doesn't block timers or coroutines from submitting more requests, which
689 * means block_job_pause is still necessary.
690 *
691 * This function can be recursive.
692 */
693void bdrv_drained_begin(BlockDriverState *bs);
694
695/**
696 * bdrv_do_drained_begin_quiesce:
697 *
698 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
699 * running requests to complete.
700 */
701void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
702 BdrvChild *parent, bool ignore_bds_parents);
703
704/**
705 * Like bdrv_drained_begin, but recursively begins a quiesced section for
706 * exclusive access to all child nodes as well.
707 */
708void bdrv_subtree_drained_begin(BlockDriverState *bs);
709
710/**
711 * bdrv_drained_end:
712 *
713 * End a quiescent section started by bdrv_drained_begin().
714 *
715 * This polls @bs's AioContext until all scheduled sub-drained_ends
716 * have settled. On one hand, that may result in graph changes. On
717 * the other, this requires that the caller either runs in the main
718 * loop; or that all involved nodes (@bs and all of its parents) are
719 * in the caller's AioContext.
720 */
721void bdrv_drained_end(BlockDriverState *bs);
722
723/**
724 * bdrv_drained_end_no_poll:
725 *
726 * Same as bdrv_drained_end(), but do not poll for the subgraph to
727 * actually become unquiesced. Therefore, no graph changes will occur
728 * with this function.
729 *
730 * *drained_end_counter is incremented for every background operation
731 * that is scheduled, and will be decremented for every operation once
732 * it settles. The caller must poll until it reaches 0. The counter
733 * should be accessed using atomic operations only.
734 */
735void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
736
737/**
738 * End a quiescent section started by bdrv_subtree_drained_begin().
739 */
740void bdrv_subtree_drained_end(BlockDriverState *bs);
741
742void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
743 Error **errp);
744void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
745
746bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
747 uint32_t granularity, Error **errp);
748/**
749 *
750 * bdrv_register_buf/bdrv_unregister_buf:
751 *
752 * Register/unregister a buffer for I/O. For example, VFIO drivers are
753 * interested to know the memory areas that would later be used for I/O, so
754 * that they can prepare IOMMU mapping etc., to get better performance.
755 */
756void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
757void bdrv_unregister_buf(BlockDriverState *bs, void *host);
758
759/**
760 *
761 * bdrv_co_copy_range:
762 *
763 * Do offloaded copy between two children. If the operation is not implemented
764 * by the driver, or if the backend storage doesn't support it, a negative
765 * error code will be returned.
766 *
767 * Note: block layer doesn't emulate or fallback to a bounce buffer approach
768 * because usually the caller shouldn't attempt offloaded copy any more (e.g.
769 * calling copy_file_range(2)) after the first error, thus it should fall back
770 * to a read+write path in the caller level.
771 *
772 * @src: Source child to copy data from
773 * @src_offset: offset in @src image to read data
774 * @dst: Destination child to copy data to
775 * @dst_offset: offset in @dst image to write data
776 * @bytes: number of bytes to copy
777 * @flags: request flags. Supported flags:
778 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
779 * write on @dst as if bdrv_co_pwrite_zeroes is
780 * called. Used to simplify caller code, or
781 * during BlockDriver.bdrv_co_copy_range_from()
782 * recursion.
783 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
784 * requests currently in flight.
785 *
786 * Returns: 0 if succeeded; negative error code if failed.
787 **/
788int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
789 BdrvChild *dst, uint64_t dst_offset,
790 uint64_t bytes, BdrvRequestFlags read_flags,
791 BdrvRequestFlags write_flags);
792#endif