1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3    md.h : kernel internal structure of the Linux MD driver
4           Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
5 
6 */
7 
8 #ifndef _MD_MD_H
9 #define _MD_MD_H
10 
11 #include <linux/blkdev.h>
12 #include <linux/backing-dev.h>
13 #include <linux/badblocks.h>
14 #include <linux/kobject.h>
15 #include <linux/list.h>
16 #include <linux/mm.h>
17 #include <linux/mutex.h>
18 #include <linux/timer.h>
19 #include <linux/wait.h>
20 #include <linux/workqueue.h>
21 #include "md-cluster.h"
22 
23 #define MaxSector (~(sector_t)0)
24 
25 /*
26  * These flags should really be called "NO_RETRY" rather than
27  * "FAILFAST" because they don't make any promise about time lapse,
28  * only about the number of retries, which will be zero.
29  * REQ_FAILFAST_DRIVER is not included because
30  * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.")
31  * seems to suggest that the errors it avoids retrying should usually
32  * be retried.
33  */
34 #define	MD_FAILFAST	(REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
35 
36 /*
37  * The struct embedded in rdev is used to serialize IO.
38  */
39 struct serial_in_rdev {
40 	struct rb_root_cached serial_rb;
41 	spinlock_t serial_lock;
42 	wait_queue_head_t serial_io_wait;
43 };
44 
45 /*
46  * MD's 'extended' device
47  */
48 struct md_rdev {
49 	struct list_head same_set;	/* RAID devices within the same set */
50 
51 	sector_t sectors;		/* Device size (in 512bytes sectors) */
52 	struct mddev *mddev;		/* RAID array if running */
53 	int last_events;		/* IO event timestamp */
54 
55 	/*
56 	 * If meta_bdev is non-NULL, it means that a separate device is
57 	 * being used to store the metadata (superblock/bitmap) which
58 	 * would otherwise be contained on the same device as the data (bdev).
59 	 */
60 	struct block_device *meta_bdev;
61 	struct block_device *bdev;	/* block device handle */
62 
63 	struct page	*sb_page, *bb_page;
64 	int		sb_loaded;
65 	__u64		sb_events;
66 	sector_t	data_offset;	/* start of data in array */
67 	sector_t	new_data_offset;/* only relevant while reshaping */
68 	sector_t	sb_start;	/* offset of the super block (in 512byte sectors) */
69 	int		sb_size;	/* bytes in the superblock */
70 	int		preferred_minor;	/* autorun support */
71 
72 	struct kobject	kobj;
73 
74 	/* A device can be in one of three states based on two flags:
75 	 * Not working:   faulty==1 in_sync==0
76 	 * Fully working: faulty==0 in_sync==1
77 	 * Working, but not
78 	 * in sync with array
79 	 *                faulty==0 in_sync==0
80 	 *
81 	 * It can never have faulty==1, in_sync==1
82 	 * This reduces the burden of testing multiple flags in many cases
83 	 */
84 
85 	unsigned long	flags;	/* bit set of 'enum flag_bits' bits. */
86 	wait_queue_head_t blocked_wait;
87 
88 	int desc_nr;			/* descriptor index in the superblock */
89 	int raid_disk;			/* role of device in array */
90 	int new_raid_disk;		/* role that the device will have in
91 					 * the array after a level-change completes.
92 					 */
93 	int saved_raid_disk;		/* role that device used to have in the
94 					 * array and could again if we did a partial
95 					 * resync from the bitmap
96 					 */
97 	union {
98 		sector_t recovery_offset;/* If this device has been partially
99 					 * recovered, this is where we were
100 					 * up to.
101 					 */
102 		sector_t journal_tail;	/* If this device is a journal device,
103 					 * this is the journal tail (journal
104 					 * recovery start point)
105 					 */
106 	};
107 
108 	atomic_t	nr_pending;	/* number of pending requests.
109 					 * only maintained for arrays that
110 					 * support hot removal
111 					 */
112 	atomic_t	read_errors;	/* number of consecutive read errors that
113 					 * we have tried to ignore.
114 					 */
115 	time64_t	last_read_error;	/* monotonic time since our
116 						 * last read error
117 						 */
118 	atomic_t	corrected_errors; /* number of corrected read errors,
119 					   * for reporting to userspace and storing
120 					   * in superblock.
121 					   */
122 
123 	struct serial_in_rdev *serial;  /* used for raid1 io serialization */
124 
125 	struct kernfs_node *sysfs_state; /* handle for 'state'
126 					   * sysfs entry */
127 	/* handle for 'unacknowledged_bad_blocks' sysfs dentry */
128 	struct kernfs_node *sysfs_unack_badblocks;
129 	/* handle for 'bad_blocks' sysfs dentry */
130 	struct kernfs_node *sysfs_badblocks;
131 	struct badblocks badblocks;
132 
133 	struct {
134 		short offset;	/* Offset from superblock to start of PPL.
135 				 * Not used by external metadata. */
136 		unsigned int size;	/* Size in sectors of the PPL space */
137 		sector_t sector;	/* First sector of the PPL space */
138 	} ppl;
139 };
140 enum flag_bits {
141 	Faulty,			/* device is known to have a fault */
142 	In_sync,		/* device is in_sync with rest of array */
143 	Bitmap_sync,		/* ..actually, not quite In_sync.  Need a
144 				 * bitmap-based recovery to get fully in sync.
145 				 * The bit is only meaningful before device
146 				 * has been passed to pers->hot_add_disk.
147 				 */
148 	WriteMostly,		/* Avoid reading if at all possible */
149 	AutoDetected,		/* added by auto-detect */
150 	Blocked,		/* An error occurred but has not yet
151 				 * been acknowledged by the metadata
152 				 * handler, so don't allow writes
153 				 * until it is cleared */
154 	WriteErrorSeen,		/* A write error has been seen on this
155 				 * device
156 				 */
157 	FaultRecorded,		/* Intermediate state for clearing
158 				 * Blocked.  The Fault is/will-be
159 				 * recorded in the metadata, but that
160 				 * metadata hasn't been stored safely
161 				 * on disk yet.
162 				 */
163 	BlockedBadBlocks,	/* A writer is blocked because they
164 				 * found an unacknowledged bad-block.
165 				 * This can safely be cleared at any
166 				 * time, and the writer will re-check.
167 				 * It may be set at any time, and at
168 				 * worst the writer will timeout and
169 				 * re-check.  So setting it as
170 				 * accurately as possible is good, but
171 				 * not absolutely critical.
172 				 */
173 	WantReplacement,	/* This device is a candidate to be
174 				 * hot-replaced, either because it has
175 				 * reported some faults, or because
176 				 * of explicit request.
177 				 */
178 	Replacement,		/* This device is a replacement for
179 				 * a want_replacement device with same
180 				 * raid_disk number.
181 				 */
182 	Candidate,		/* For clustered environments only:
183 				 * This device is seen locally but not
184 				 * by the whole cluster
185 				 */
186 	Journal,		/* This device is used as journal for
187 				 * raid-5/6.
188 				 * Usually, this device should be faster
189 				 * than other devices in the array
190 				 */
191 	ClusterRemove,
192 	RemoveSynchronized,	/* synchronize_rcu() was called after
193 				 * this device was known to be faulty,
194 				 * so it is safe to remove without
195 				 * another synchronize_rcu() call.
196 				 */
197 	ExternalBbl,            /* External metadata provides bad
198 				 * block management for a disk
199 				 */
200 	FailFast,		/* Minimal retries should be attempted on
201 				 * this device, so use REQ_FAILFAST_DEV.
202 				 * Also don't try to repair failed reads.
203 				 * It is expects that no bad block log
204 				 * is present.
205 				 */
206 	LastDev,		/* Seems to be the last working dev as
207 				 * it didn't fail, so don't use FailFast
208 				 * any more for metadata
209 				 */
210 	CollisionCheck,		/*
211 				 * check if there is collision between raid1
212 				 * serial bios.
213 				 */
214 	Holder,			/* rdev is used as holder while opening
215 				 * underlying disk exclusively.
216 				 */
217 };
218 
is_badblock(struct md_rdev * rdev,sector_t s,int sectors,sector_t * first_bad,int * bad_sectors)219 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
220 			      sector_t *first_bad, int *bad_sectors)
221 {
222 	if (unlikely(rdev->badblocks.count)) {
223 		int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
224 					sectors,
225 					first_bad, bad_sectors);
226 		if (rv)
227 			*first_bad -= rdev->data_offset;
228 		return rv;
229 	}
230 	return 0;
231 }
232 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
233 			      int is_new);
234 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
235 				int is_new);
236 struct md_cluster_info;
237 
238 /**
239  * enum mddev_flags - md device flags.
240  * @MD_ARRAY_FIRST_USE: First use of array, needs initialization.
241  * @MD_CLOSING: If set, we are closing the array, do not open it then.
242  * @MD_JOURNAL_CLEAN: A raid with journal is already clean.
243  * @MD_HAS_JOURNAL: The raid array has journal feature set.
244  * @MD_CLUSTER_RESYNC_LOCKED: cluster raid only, which means node, already took
245  *			       resync lock, need to release the lock.
246  * @MD_FAILFAST_SUPPORTED: Using MD_FAILFAST on metadata writes is supported as
247  *			    calls to md_error() will never cause the array to
248  *			    become failed.
249  * @MD_HAS_PPL:  The raid array has PPL feature set.
250  * @MD_HAS_MULTIPLE_PPLS: The raid array has multiple PPLs feature set.
251  * @MD_ALLOW_SB_UPDATE: md_check_recovery is allowed to update the metadata
252  *			 without taking reconfig_mutex.
253  * @MD_UPDATING_SB: md_check_recovery is updating the metadata without
254  *		     explicitly holding reconfig_mutex.
255  * @MD_NOT_READY: do_md_run() is active, so 'array_state', ust not report that
256  *		   array is ready yet.
257  * @MD_BROKEN: This is used to stop writes and mark array as failed.
258  * @MD_DELETED: This device is being deleted
259  *
260  * change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added
261  */
262 enum mddev_flags {
263 	MD_ARRAY_FIRST_USE,
264 	MD_CLOSING,
265 	MD_JOURNAL_CLEAN,
266 	MD_HAS_JOURNAL,
267 	MD_CLUSTER_RESYNC_LOCKED,
268 	MD_FAILFAST_SUPPORTED,
269 	MD_HAS_PPL,
270 	MD_HAS_MULTIPLE_PPLS,
271 	MD_ALLOW_SB_UPDATE,
272 	MD_UPDATING_SB,
273 	MD_NOT_READY,
274 	MD_BROKEN,
275 	MD_DELETED,
276 };
277 
278 enum mddev_sb_flags {
279 	MD_SB_CHANGE_DEVS,		/* Some device status has changed */
280 	MD_SB_CHANGE_CLEAN,	/* transition to or from 'clean' */
281 	MD_SB_CHANGE_PENDING,	/* switch from 'clean' to 'active' in progress */
282 	MD_SB_NEED_REWRITE,	/* metadata write needs to be repeated */
283 };
284 
285 #define NR_SERIAL_INFOS		8
286 /* record current range of serialize IOs */
287 struct serial_info {
288 	struct rb_node node;
289 	sector_t start;		/* start sector of rb node */
290 	sector_t last;		/* end sector of rb node */
291 	sector_t _subtree_last; /* highest sector in subtree of rb node */
292 };
293 
294 /*
295  * mddev->curr_resync stores the current sector of the resync but
296  * also has some overloaded values.
297  */
298 enum {
299 	/* No resync in progress */
300 	MD_RESYNC_NONE = 0,
301 	/* Yielded to allow another conflicting resync to commence */
302 	MD_RESYNC_YIELDED = 1,
303 	/* Delayed to check that there is no conflict with another sync */
304 	MD_RESYNC_DELAYED = 2,
305 	/* Any value greater than or equal to this is in an active resync */
306 	MD_RESYNC_ACTIVE = 3,
307 };
308 
309 struct mddev {
310 	void				*private;
311 	struct md_personality		*pers;
312 	dev_t				unit;
313 	int				md_minor;
314 	struct list_head		disks;
315 	unsigned long			flags;
316 	unsigned long			sb_flags;
317 
318 	int				suspended;
319 	struct percpu_ref		active_io;
320 	int				ro;
321 	int				sysfs_active; /* set when sysfs deletes
322 						       * are happening, so run/
323 						       * takeover/stop are not safe
324 						       */
325 	struct gendisk			*gendisk;
326 
327 	struct kobject			kobj;
328 	int				hold_active;
329 #define	UNTIL_IOCTL	1
330 #define	UNTIL_STOP	2
331 
332 	/* Superblock information */
333 	int				major_version,
334 					minor_version,
335 					patch_version;
336 	int				persistent;
337 	int				external;	/* metadata is
338 							 * managed externally */
339 	char				metadata_type[17]; /* externally set*/
340 	int				chunk_sectors;
341 	time64_t			ctime, utime;
342 	int				level, layout;
343 	char				clevel[16];
344 	int				raid_disks;
345 	int				max_disks;
346 	sector_t			dev_sectors;	/* used size of
347 							 * component devices */
348 	sector_t			array_sectors; /* exported array size */
349 	int				external_size; /* size managed
350 							* externally */
351 	__u64				events;
352 	/* If the last 'event' was simply a clean->dirty transition, and
353 	 * we didn't write it to the spares, then it is safe and simple
354 	 * to just decrement the event count on a dirty->clean transition.
355 	 * So we record that possibility here.
356 	 */
357 	int				can_decrease_events;
358 
359 	char				uuid[16];
360 
361 	/* If the array is being reshaped, we need to record the
362 	 * new shape and an indication of where we are up to.
363 	 * This is written to the superblock.
364 	 * If reshape_position is MaxSector, then no reshape is happening (yet).
365 	 */
366 	sector_t			reshape_position;
367 	int				delta_disks, new_level, new_layout;
368 	int				new_chunk_sectors;
369 	int				reshape_backwards;
370 
371 	struct md_thread __rcu		*thread;	/* management thread */
372 	struct md_thread __rcu		*sync_thread;	/* doing resync or reconstruct */
373 
374 	/* 'last_sync_action' is initialized to "none".  It is set when a
375 	 * sync operation (i.e "data-check", "requested-resync", "resync",
376 	 * "recovery", or "reshape") is started.  It holds this value even
377 	 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
378 	 * or finished).  It is overwritten when a new sync operation is begun.
379 	 */
380 	char				*last_sync_action;
381 	sector_t			curr_resync;	/* last block scheduled */
382 	/* As resync requests can complete out of order, we cannot easily track
383 	 * how much resync has been completed.  So we occasionally pause until
384 	 * everything completes, then set curr_resync_completed to curr_resync.
385 	 * As such it may be well behind the real resync mark, but it is a value
386 	 * we are certain of.
387 	 */
388 	sector_t			curr_resync_completed;
389 	unsigned long			resync_mark;	/* a recent timestamp */
390 	sector_t			resync_mark_cnt;/* blocks written at resync_mark */
391 	sector_t			curr_mark_cnt; /* blocks scheduled now */
392 
393 	sector_t			resync_max_sectors; /* may be set by personality */
394 
395 	atomic64_t			resync_mismatches; /* count of sectors where
396 							    * parity/replica mismatch found
397 							    */
398 
399 	/* allow user-space to request suspension of IO to regions of the array */
400 	sector_t			suspend_lo;
401 	sector_t			suspend_hi;
402 	/* if zero, use the system-wide default */
403 	int				sync_speed_min;
404 	int				sync_speed_max;
405 
406 	/* resync even though the same disks are shared among md-devices */
407 	int				parallel_resync;
408 
409 	int				ok_start_degraded;
410 
411 	unsigned long			recovery;
412 	/* If a RAID personality determines that recovery (of a particular
413 	 * device) will fail due to a read error on the source device, it
414 	 * takes a copy of this number and does not attempt recovery again
415 	 * until this number changes.
416 	 */
417 	int				recovery_disabled;
418 
419 	int				in_sync;	/* know to not need resync */
420 	/* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
421 	 * that we are never stopping an array while it is open.
422 	 * 'reconfig_mutex' protects all other reconfiguration.
423 	 * These locks are separate due to conflicting interactions
424 	 * with disk->open_mutex.
425 	 * Lock ordering is:
426 	 *  reconfig_mutex -> disk->open_mutex
427 	 *  disk->open_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
428 	 */
429 	struct mutex			open_mutex;
430 	struct mutex			reconfig_mutex;
431 	atomic_t			active;		/* general refcount */
432 	atomic_t			openers;	/* number of active opens */
433 
434 	int				changed;	/* True if we might need to
435 							 * reread partition info */
436 	int				degraded;	/* whether md should consider
437 							 * adding a spare
438 							 */
439 
440 	atomic_t			recovery_active; /* blocks scheduled, but not written */
441 	wait_queue_head_t		recovery_wait;
442 	sector_t			recovery_cp;
443 	sector_t			resync_min;	/* user requested sync
444 							 * starts here */
445 	sector_t			resync_max;	/* resync should pause
446 							 * when it gets here */
447 
448 	struct kernfs_node		*sysfs_state;	/* handle for 'array_state'
449 							 * file in sysfs.
450 							 */
451 	struct kernfs_node		*sysfs_action;  /* handle for 'sync_action' */
452 	struct kernfs_node		*sysfs_completed;	/*handle for 'sync_completed' */
453 	struct kernfs_node		*sysfs_degraded;	/*handle for 'degraded' */
454 	struct kernfs_node		*sysfs_level;		/*handle for 'level' */
455 
456 	struct work_struct del_work;	/* used for delayed sysfs removal */
457 
458 	/* "lock" protects:
459 	 *   flush_bio transition from NULL to !NULL
460 	 *   rdev superblocks, events
461 	 *   clearing MD_CHANGE_*
462 	 *   in_sync - and related safemode and MD_CHANGE changes
463 	 *   pers (also protected by reconfig_mutex and pending IO).
464 	 *   clearing ->bitmap
465 	 *   clearing ->bitmap_info.file
466 	 *   changing ->resync_{min,max}
467 	 *   setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
468 	 */
469 	spinlock_t			lock;
470 	wait_queue_head_t		sb_wait;	/* for waiting on superblock updates */
471 	atomic_t			pending_writes;	/* number of active superblock writes */
472 
473 	unsigned int			safemode;	/* if set, update "clean" superblock
474 							 * when no writes pending.
475 							 */
476 	unsigned int			safemode_delay;
477 	struct timer_list		safemode_timer;
478 	struct percpu_ref		writes_pending;
479 	int				sync_checkers;	/* # of threads checking writes_pending */
480 	struct request_queue		*queue;	/* for plugging ... */
481 
482 	struct bitmap			*bitmap; /* the bitmap for the device */
483 	struct {
484 		struct file		*file; /* the bitmap file */
485 		loff_t			offset; /* offset from superblock of
486 						 * start of bitmap. May be
487 						 * negative, but not '0'
488 						 * For external metadata, offset
489 						 * from start of device.
490 						 */
491 		unsigned long		space; /* space available at this offset */
492 		loff_t			default_offset; /* this is the offset to use when
493 							 * hot-adding a bitmap.  It should
494 							 * eventually be settable by sysfs.
495 							 */
496 		unsigned long		default_space; /* space available at
497 							* default offset */
498 		struct mutex		mutex;
499 		unsigned long		chunksize;
500 		unsigned long		daemon_sleep; /* how many jiffies between updates? */
501 		unsigned long		max_write_behind; /* write-behind mode */
502 		int			external;
503 		int			nodes; /* Maximum number of nodes in the cluster */
504 		char                    cluster_name[64]; /* Name of the cluster */
505 	} bitmap_info;
506 
507 	atomic_t			max_corr_read_errors; /* max read retries */
508 	struct list_head		all_mddevs;
509 
510 	const struct attribute_group	*to_remove;
511 
512 	struct bio_set			bio_set;
513 	struct bio_set			sync_set; /* for sync operations like
514 						   * metadata and bitmap writes
515 						   */
516 	struct bio_set			io_clone_set;
517 
518 	/* Generic flush handling.
519 	 * The last to finish preflush schedules a worker to submit
520 	 * the rest of the request (without the REQ_PREFLUSH flag).
521 	 */
522 	struct bio *flush_bio;
523 	atomic_t flush_pending;
524 	ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed
525 						* flush was started.
526 						*/
527 	struct work_struct flush_work;
528 	struct work_struct event_work;	/* used by dm to report failure event */
529 	mempool_t *serial_info_pool;
530 	void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
531 	struct md_cluster_info		*cluster_info;
532 	unsigned int			good_device_nr;	/* good device num within cluster raid */
533 	unsigned int			noio_flag; /* for memalloc scope API */
534 
535 	/*
536 	 * Temporarily store rdev that will be finally removed when
537 	 * reconfig_mutex is unlocked, protected by reconfig_mutex.
538 	 */
539 	struct list_head		deleting;
540 
541 	/* Used to synchronize idle and frozen for action_store() */
542 	struct mutex			sync_mutex;
543 	/* The sequence number for sync thread */
544 	atomic_t sync_seq;
545 
546 	bool	has_superblocks:1;
547 	bool	fail_last_dev:1;
548 	bool	serialize_policy:1;
549 };
550 
551 enum recovery_flags {
552 	/*
553 	 * If neither SYNC or RESHAPE are set, then it is a recovery.
554 	 */
555 	MD_RECOVERY_RUNNING,	/* a thread is running, or about to be started */
556 	MD_RECOVERY_SYNC,	/* actually doing a resync, not a recovery */
557 	MD_RECOVERY_RECOVER,	/* doing recovery, or need to try it. */
558 	MD_RECOVERY_INTR,	/* resync needs to be aborted for some reason */
559 	MD_RECOVERY_DONE,	/* thread is done and is waiting to be reaped */
560 	MD_RECOVERY_NEEDED,	/* we might need to start a resync/recover */
561 	MD_RECOVERY_REQUESTED,	/* user-space has requested a sync (used with SYNC) */
562 	MD_RECOVERY_CHECK,	/* user-space request for check-only, no repair */
563 	MD_RECOVERY_RESHAPE,	/* A reshape is happening */
564 	MD_RECOVERY_FROZEN,	/* User request to abort, and not restart, any action */
565 	MD_RECOVERY_ERROR,	/* sync-action interrupted because io-error */
566 	MD_RECOVERY_WAIT,	/* waiting for pers->start() to finish */
567 	MD_RESYNCING_REMOTE,	/* remote node is running resync thread */
568 };
569 
570 enum md_ro_state {
571 	MD_RDWR,
572 	MD_RDONLY,
573 	MD_AUTO_READ,
574 	MD_MAX_STATE
575 };
576 
md_is_rdwr(struct mddev * mddev)577 static inline bool md_is_rdwr(struct mddev *mddev)
578 {
579 	return (mddev->ro == MD_RDWR);
580 }
581 
is_md_suspended(struct mddev * mddev)582 static inline bool is_md_suspended(struct mddev *mddev)
583 {
584 	return percpu_ref_is_dying(&mddev->active_io);
585 }
586 
mddev_lock(struct mddev * mddev)587 static inline int __must_check mddev_lock(struct mddev *mddev)
588 {
589 	return mutex_lock_interruptible(&mddev->reconfig_mutex);
590 }
591 
592 /* Sometimes we need to take the lock in a situation where
593  * failure due to interrupts is not acceptable.
594  */
mddev_lock_nointr(struct mddev * mddev)595 static inline void mddev_lock_nointr(struct mddev *mddev)
596 {
597 	mutex_lock(&mddev->reconfig_mutex);
598 }
599 
mddev_trylock(struct mddev * mddev)600 static inline int mddev_trylock(struct mddev *mddev)
601 {
602 	return mutex_trylock(&mddev->reconfig_mutex);
603 }
604 extern void mddev_unlock(struct mddev *mddev);
605 
md_sync_acct(struct block_device * bdev,unsigned long nr_sectors)606 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
607 {
608 	atomic_add(nr_sectors, &bdev->bd_disk->sync_io);
609 }
610 
md_sync_acct_bio(struct bio * bio,unsigned long nr_sectors)611 static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors)
612 {
613 	md_sync_acct(bio->bi_bdev, nr_sectors);
614 }
615 
616 struct md_personality
617 {
618 	char *name;
619 	int level;
620 	struct list_head list;
621 	struct module *owner;
622 	bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio);
623 	/*
624 	 * start up works that do NOT require md_thread. tasks that
625 	 * requires md_thread should go into start()
626 	 */
627 	int (*run)(struct mddev *mddev);
628 	/* start up works that require md threads */
629 	int (*start)(struct mddev *mddev);
630 	void (*free)(struct mddev *mddev, void *priv);
631 	void (*status)(struct seq_file *seq, struct mddev *mddev);
632 	/* error_handler must set ->faulty and clear ->in_sync
633 	 * if appropriate, and should abort recovery if needed
634 	 */
635 	void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
636 	int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
637 	int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
638 	int (*spare_active) (struct mddev *mddev);
639 	sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
640 	int (*resize) (struct mddev *mddev, sector_t sectors);
641 	sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
642 	int (*check_reshape) (struct mddev *mddev);
643 	int (*start_reshape) (struct mddev *mddev);
644 	void (*finish_reshape) (struct mddev *mddev);
645 	void (*update_reshape_pos) (struct mddev *mddev);
646 	void (*prepare_suspend) (struct mddev *mddev);
647 	/* quiesce suspends or resumes internal processing.
648 	 * 1 - stop new actions and wait for action io to complete
649 	 * 0 - return to normal behaviour
650 	 */
651 	void (*quiesce) (struct mddev *mddev, int quiesce);
652 	/* takeover is used to transition an array from one
653 	 * personality to another.  The new personality must be able
654 	 * to handle the data in the current layout.
655 	 * e.g. 2drive raid1 -> 2drive raid5
656 	 *      ndrive raid5 -> degraded n+1drive raid6 with special layout
657 	 * If the takeover succeeds, a new 'private' structure is returned.
658 	 * This needs to be installed and then ->run used to activate the
659 	 * array.
660 	 */
661 	void *(*takeover) (struct mddev *mddev);
662 	/* Changes the consistency policy of an active array. */
663 	int (*change_consistency_policy)(struct mddev *mddev, const char *buf);
664 };
665 
666 struct md_sysfs_entry {
667 	struct attribute attr;
668 	ssize_t (*show)(struct mddev *, char *);
669 	ssize_t (*store)(struct mddev *, const char *, size_t);
670 };
671 extern const struct attribute_group md_bitmap_group;
672 
sysfs_get_dirent_safe(struct kernfs_node * sd,char * name)673 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
674 {
675 	if (sd)
676 		return sysfs_get_dirent(sd, name);
677 	return sd;
678 }
sysfs_notify_dirent_safe(struct kernfs_node * sd)679 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
680 {
681 	if (sd)
682 		sysfs_notify_dirent(sd);
683 }
684 
mdname(struct mddev * mddev)685 static inline char * mdname (struct mddev * mddev)
686 {
687 	return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
688 }
689 
sysfs_link_rdev(struct mddev * mddev,struct md_rdev * rdev)690 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
691 {
692 	char nm[20];
693 	if (!test_bit(Replacement, &rdev->flags) &&
694 	    !test_bit(Journal, &rdev->flags) &&
695 	    mddev->kobj.sd) {
696 		sprintf(nm, "rd%d", rdev->raid_disk);
697 		return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
698 	} else
699 		return 0;
700 }
701 
sysfs_unlink_rdev(struct mddev * mddev,struct md_rdev * rdev)702 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
703 {
704 	char nm[20];
705 	if (!test_bit(Replacement, &rdev->flags) &&
706 	    !test_bit(Journal, &rdev->flags) &&
707 	    mddev->kobj.sd) {
708 		sprintf(nm, "rd%d", rdev->raid_disk);
709 		sysfs_remove_link(&mddev->kobj, nm);
710 	}
711 }
712 
713 /*
714  * iterates through some rdev ringlist. It's safe to remove the
715  * current 'rdev'. Dont touch 'tmp' though.
716  */
717 #define rdev_for_each_list(rdev, tmp, head)				\
718 	list_for_each_entry_safe(rdev, tmp, head, same_set)
719 
720 /*
721  * iterates through the 'same array disks' ringlist
722  */
723 #define rdev_for_each(rdev, mddev)				\
724 	list_for_each_entry(rdev, &((mddev)->disks), same_set)
725 
726 #define rdev_for_each_safe(rdev, tmp, mddev)				\
727 	list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
728 
729 #define rdev_for_each_rcu(rdev, mddev)				\
730 	list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
731 
732 struct md_thread {
733 	void			(*run) (struct md_thread *thread);
734 	struct mddev		*mddev;
735 	wait_queue_head_t	wqueue;
736 	unsigned long		flags;
737 	struct task_struct	*tsk;
738 	unsigned long		timeout;
739 	void			*private;
740 };
741 
742 struct md_io_clone {
743 	struct mddev	*mddev;
744 	struct bio	*orig_bio;
745 	unsigned long	start_time;
746 	struct bio	bio_clone;
747 };
748 
749 #define THREAD_WAKEUP  0
750 
safe_put_page(struct page * p)751 static inline void safe_put_page(struct page *p)
752 {
753 	if (p) put_page(p);
754 }
755 
756 extern int register_md_personality(struct md_personality *p);
757 extern int unregister_md_personality(struct md_personality *p);
758 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
759 		struct module *module);
760 extern int unregister_md_cluster_operations(void);
761 extern int md_setup_cluster(struct mddev *mddev, int nodes);
762 extern void md_cluster_stop(struct mddev *mddev);
763 extern struct md_thread *md_register_thread(
764 	void (*run)(struct md_thread *thread),
765 	struct mddev *mddev,
766 	const char *name);
767 extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **threadp);
768 extern void md_wakeup_thread(struct md_thread __rcu *thread);
769 extern void md_check_recovery(struct mddev *mddev);
770 extern void md_reap_sync_thread(struct mddev *mddev);
771 extern int mddev_init_writes_pending(struct mddev *mddev);
772 extern bool md_write_start(struct mddev *mddev, struct bio *bi);
773 extern void md_write_inc(struct mddev *mddev, struct bio *bi);
774 extern void md_write_end(struct mddev *mddev);
775 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
776 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
777 extern void md_finish_reshape(struct mddev *mddev);
778 void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev,
779 			struct bio *bio, sector_t start, sector_t size);
780 void md_account_bio(struct mddev *mddev, struct bio **bio);
781 
782 extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio);
783 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
784 			   sector_t sector, int size, struct page *page);
785 extern int md_super_wait(struct mddev *mddev);
786 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
787 		struct page *page, blk_opf_t opf, bool metadata_op);
788 extern void md_do_sync(struct md_thread *thread);
789 extern void md_new_event(void);
790 extern void md_allow_write(struct mddev *mddev);
791 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
792 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
793 extern int md_check_no_bitmap(struct mddev *mddev);
794 extern int md_integrity_register(struct mddev *mddev);
795 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
796 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
797 
798 extern void mddev_init(struct mddev *mddev);
799 struct mddev *md_alloc(dev_t dev, char *name);
800 void mddev_put(struct mddev *mddev);
801 extern int md_run(struct mddev *mddev);
802 extern int md_start(struct mddev *mddev);
803 extern void md_stop(struct mddev *mddev);
804 extern void md_stop_writes(struct mddev *mddev);
805 extern int md_rdev_init(struct md_rdev *rdev);
806 extern void md_rdev_clear(struct md_rdev *rdev);
807 
808 extern void md_handle_request(struct mddev *mddev, struct bio *bio);
809 extern void mddev_suspend(struct mddev *mddev);
810 extern void mddev_resume(struct mddev *mddev);
811 
812 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
813 extern void md_update_sb(struct mddev *mddev, int force);
814 extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
815 				     bool is_suspend);
816 extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
817 				      bool is_suspend);
818 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
819 struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev);
820 
is_rdev_broken(struct md_rdev * rdev)821 static inline bool is_rdev_broken(struct md_rdev *rdev)
822 {
823 	return !disk_live(rdev->bdev->bd_disk);
824 }
825 
rdev_dec_pending(struct md_rdev * rdev,struct mddev * mddev)826 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
827 {
828 	int faulty = test_bit(Faulty, &rdev->flags);
829 	if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
830 		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
831 		md_wakeup_thread(mddev->thread);
832 	}
833 }
834 
835 extern struct md_cluster_operations *md_cluster_ops;
mddev_is_clustered(struct mddev * mddev)836 static inline int mddev_is_clustered(struct mddev *mddev)
837 {
838 	return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
839 }
840 
841 /* clear unsupported mddev_flags */
mddev_clear_unsupported_flags(struct mddev * mddev,unsigned long unsupported_flags)842 static inline void mddev_clear_unsupported_flags(struct mddev *mddev,
843 	unsigned long unsupported_flags)
844 {
845 	mddev->flags &= ~unsupported_flags;
846 }
847 
mddev_check_write_zeroes(struct mddev * mddev,struct bio * bio)848 static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio)
849 {
850 	if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
851 	    !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors)
852 		mddev->queue->limits.max_write_zeroes_sectors = 0;
853 }
854 
855 struct mdu_array_info_s;
856 struct mdu_disk_info_s;
857 
858 extern int mdp_major;
859 extern struct workqueue_struct *md_bitmap_wq;
860 void md_autostart_arrays(int part);
861 int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info);
862 int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info);
863 int do_md_run(struct mddev *mddev);
864 
865 extern const struct block_device_operations md_fops;
866 
867 #endif /* _MD_MD_H */
868