xref: /openbmc/linux/drivers/md/raid5.h (revision f3539c12)
1 #ifndef _RAID5_H
2 #define _RAID5_H
3 
4 #include <linux/raid/xor.h>
5 #include <linux/dmaengine.h>
6 
7 /*
8  *
9  * Each stripe contains one buffer per device.  Each buffer can be in
10  * one of a number of states stored in "flags".  Changes between
11  * these states happen *almost* exclusively under the protection of the
12  * STRIPE_ACTIVE flag.  Some very specific changes can happen in bi_end_io, and
13  * these are not protected by STRIPE_ACTIVE.
14  *
15  * The flag bits that are used to represent these states are:
16  *   R5_UPTODATE and R5_LOCKED
17  *
18  * State Empty == !UPTODATE, !LOCK
19  *        We have no data, and there is no active request
20  * State Want == !UPTODATE, LOCK
21  *        A read request is being submitted for this block
22  * State Dirty == UPTODATE, LOCK
23  *        Some new data is in this buffer, and it is being written out
24  * State Clean == UPTODATE, !LOCK
25  *        We have valid data which is the same as on disc
26  *
27  * The possible state transitions are:
28  *
29  *  Empty -> Want   - on read or write to get old data for  parity calc
30  *  Empty -> Dirty  - on compute_parity to satisfy write/sync request.
31  *  Empty -> Clean  - on compute_block when computing a block for failed drive
32  *  Want  -> Empty  - on failed read
33  *  Want  -> Clean  - on successful completion of read request
34  *  Dirty -> Clean  - on successful completion of write request
35  *  Dirty -> Clean  - on failed write
36  *  Clean -> Dirty  - on compute_parity to satisfy write/sync (RECONSTRUCT or RMW)
37  *
38  * The Want->Empty, Want->Clean, Dirty->Clean, transitions
39  * all happen in b_end_io at interrupt time.
40  * Each sets the Uptodate bit before releasing the Lock bit.
41  * This leaves one multi-stage transition:
42  *    Want->Dirty->Clean
43  * This is safe because thinking that a Clean buffer is actually dirty
44  * will at worst delay some action, and the stripe will be scheduled
45  * for attention after the transition is complete.
46  *
47  * There is one possibility that is not covered by these states.  That
48  * is if one drive has failed and there is a spare being rebuilt.  We
49  * can't distinguish between a clean block that has been generated
50  * from parity calculations, and a clean block that has been
51  * successfully written to the spare ( or to parity when resyncing).
52  * To distinguish these states we have a stripe bit STRIPE_INSYNC that
53  * is set whenever a write is scheduled to the spare, or to the parity
54  * disc if there is no spare.  A sync request clears this bit, and
55  * when we find it set with no buffers locked, we know the sync is
56  * complete.
57  *
58  * Buffers for the md device that arrive via make_request are attached
59  * to the appropriate stripe in one of two lists linked on b_reqnext.
60  * One list (bh_read) for read requests, one (bh_write) for write.
61  * There should never be more than one buffer on the two lists
62  * together, but we are not guaranteed of that so we allow for more.
63  *
64  * If a buffer is on the read list when the associated cache buffer is
65  * Uptodate, the data is copied into the read buffer and it's b_end_io
66  * routine is called.  This may happen in the end_request routine only
67  * if the buffer has just successfully been read.  end_request should
68  * remove the buffers from the list and then set the Uptodate bit on
69  * the buffer.  Other threads may do this only if they first check
70  * that the Uptodate bit is set.  Once they have checked that they may
71  * take buffers off the read queue.
72  *
73  * When a buffer on the write list is committed for write it is copied
74  * into the cache buffer, which is then marked dirty, and moved onto a
75  * third list, the written list (bh_written).  Once both the parity
76  * block and the cached buffer are successfully written, any buffer on
77  * a written list can be returned with b_end_io.
78  *
79  * The write list and read list both act as fifos.  The read list,
80  * write list and written list are protected by the device_lock.
81  * The device_lock is only for list manipulations and will only be
82  * held for a very short time.  It can be claimed from interrupts.
83  *
84  *
85  * Stripes in the stripe cache can be on one of two lists (or on
86  * neither).  The "inactive_list" contains stripes which are not
87  * currently being used for any request.  They can freely be reused
88  * for another stripe.  The "handle_list" contains stripes that need
89  * to be handled in some way.  Both of these are fifo queues.  Each
90  * stripe is also (potentially) linked to a hash bucket in the hash
91  * table so that it can be found by sector number.  Stripes that are
92  * not hashed must be on the inactive_list, and will normally be at
93  * the front.  All stripes start life this way.
94  *
95  * The inactive_list, handle_list and hash bucket lists are all protected by the
96  * device_lock.
97  *  - stripes have a reference counter. If count==0, they are on a list.
98  *  - If a stripe might need handling, STRIPE_HANDLE is set.
99  *  - When refcount reaches zero, then if STRIPE_HANDLE it is put on
100  *    handle_list else inactive_list
101  *
102  * This, combined with the fact that STRIPE_HANDLE is only ever
103  * cleared while a stripe has a non-zero count means that if the
104  * refcount is 0 and STRIPE_HANDLE is set, then it is on the
105  * handle_list and if recount is 0 and STRIPE_HANDLE is not set, then
106  * the stripe is on inactive_list.
107  *
108  * The possible transitions are:
109  *  activate an unhashed/inactive stripe (get_active_stripe())
110  *     lockdev check-hash unlink-stripe cnt++ clean-stripe hash-stripe unlockdev
111  *  activate a hashed, possibly active stripe (get_active_stripe())
112  *     lockdev check-hash if(!cnt++)unlink-stripe unlockdev
113  *  attach a request to an active stripe (add_stripe_bh())
114  *     lockdev attach-buffer unlockdev
115  *  handle a stripe (handle_stripe())
116  *     setSTRIPE_ACTIVE,  clrSTRIPE_HANDLE ...
117  *		(lockdev check-buffers unlockdev) ..
118  *		change-state ..
119  *		record io/ops needed clearSTRIPE_ACTIVE schedule io/ops
120  *  release an active stripe (release_stripe())
121  *     lockdev if (!--cnt) { if  STRIPE_HANDLE, add to handle_list else add to inactive-list } unlockdev
122  *
123  * The refcount counts each thread that have activated the stripe,
124  * plus raid5d if it is handling it, plus one for each active request
125  * on a cached buffer, and plus one if the stripe is undergoing stripe
126  * operations.
127  *
128  * The stripe operations are:
129  * -copying data between the stripe cache and user application buffers
130  * -computing blocks to save a disk access, or to recover a missing block
131  * -updating the parity on a write operation (reconstruct write and
132  *  read-modify-write)
133  * -checking parity correctness
134  * -running i/o to disk
135  * These operations are carried out by raid5_run_ops which uses the async_tx
136  * api to (optionally) offload operations to dedicated hardware engines.
137  * When requesting an operation handle_stripe sets the pending bit for the
138  * operation and increments the count.  raid5_run_ops is then run whenever
139  * the count is non-zero.
140  * There are some critical dependencies between the operations that prevent some
141  * from being requested while another is in flight.
142  * 1/ Parity check operations destroy the in cache version of the parity block,
143  *    so we prevent parity dependent operations like writes and compute_blocks
144  *    from starting while a check is in progress.  Some dma engines can perform
145  *    the check without damaging the parity block, in these cases the parity
146  *    block is re-marked up to date (assuming the check was successful) and is
147  *    not re-read from disk.
148  * 2/ When a write operation is requested we immediately lock the affected
149  *    blocks, and mark them as not up to date.  This causes new read requests
150  *    to be held off, as well as parity checks and compute block operations.
151  * 3/ Once a compute block operation has been requested handle_stripe treats
152  *    that block as if it is up to date.  raid5_run_ops guaruntees that any
153  *    operation that is dependent on the compute block result is initiated after
154  *    the compute block completes.
155  */
156 
157 /*
158  * Operations state - intermediate states that are visible outside of
159  *   STRIPE_ACTIVE.
160  * In general _idle indicates nothing is running, _run indicates a data
161  * processing operation is active, and _result means the data processing result
162  * is stable and can be acted upon.  For simple operations like biofill and
163  * compute that only have an _idle and _run state they are indicated with
164  * sh->state flags (STRIPE_BIOFILL_RUN and STRIPE_COMPUTE_RUN)
165  */
166 /**
167  * enum check_states - handles syncing / repairing a stripe
168  * @check_state_idle - check operations are quiesced
169  * @check_state_run - check operation is running
170  * @check_state_result - set outside lock when check result is valid
171  * @check_state_compute_run - check failed and we are repairing
172  * @check_state_compute_result - set outside lock when compute result is valid
173  */
174 enum check_states {
175 	check_state_idle = 0,
176 	check_state_run, /* xor parity check */
177 	check_state_run_q, /* q-parity check */
178 	check_state_run_pq, /* pq dual parity check */
179 	check_state_check_result,
180 	check_state_compute_run, /* parity repair */
181 	check_state_compute_result,
182 };
183 
184 /**
185  * enum reconstruct_states - handles writing or expanding a stripe
186  */
187 enum reconstruct_states {
188 	reconstruct_state_idle = 0,
189 	reconstruct_state_prexor_drain_run,	/* prexor-write */
190 	reconstruct_state_drain_run,		/* write */
191 	reconstruct_state_run,			/* expand */
192 	reconstruct_state_prexor_drain_result,
193 	reconstruct_state_drain_result,
194 	reconstruct_state_result,
195 };
196 
197 struct stripe_head {
198 	struct hlist_node	hash;
199 	struct list_head	lru;	      /* inactive_list or handle_list */
200 	struct llist_node	release_list;
201 	struct r5conf		*raid_conf;
202 	short			generation;	/* increments with every
203 						 * reshape */
204 	sector_t		sector;		/* sector of this row */
205 	short			pd_idx;		/* parity disk index */
206 	short			qd_idx;		/* 'Q' disk index for raid6 */
207 	short			ddf_layout;/* use DDF ordering to calculate Q */
208 	short			hash_lock_index;
209 	unsigned long		state;		/* state flags */
210 	atomic_t		count;	      /* nr of active thread/requests */
211 	int			bm_seq;	/* sequence number for bitmap flushes */
212 	int			disks;		/* disks in stripe */
213 	int			overwrite_disks; /* total overwrite disks in stripe,
214 						  * this is only checked when stripe
215 						  * has STRIPE_BATCH_READY
216 						  */
217 	enum check_states	check_state;
218 	enum reconstruct_states reconstruct_state;
219 	spinlock_t		stripe_lock;
220 	int			cpu;
221 	struct r5worker_group	*group;
222 
223 	struct stripe_head	*batch_head; /* protected by stripe lock */
224 	spinlock_t		batch_lock; /* only header's lock is useful */
225 	struct list_head	batch_list; /* protected by head's batch lock*/
226 
227 	struct r5l_io_unit	*log_io;
228 	struct list_head	log_list;
229 	/**
230 	 * struct stripe_operations
231 	 * @target - STRIPE_OP_COMPUTE_BLK target
232 	 * @target2 - 2nd compute target in the raid6 case
233 	 * @zero_sum_result - P and Q verification flags
234 	 * @request - async service request flags for raid_run_ops
235 	 */
236 	struct stripe_operations {
237 		int 		     target, target2;
238 		enum sum_check_flags zero_sum_result;
239 	} ops;
240 	struct r5dev {
241 		/* rreq and rvec are used for the replacement device when
242 		 * writing data to both devices.
243 		 */
244 		struct bio	req, rreq;
245 		struct bio_vec	vec, rvec;
246 		struct page	*page, *orig_page;
247 		struct bio	*toread, *read, *towrite, *written;
248 		sector_t	sector;			/* sector of this page */
249 		unsigned long	flags;
250 		u32		log_checksum;
251 	} dev[1]; /* allocated with extra space depending of RAID geometry */
252 };
253 
254 /* stripe_head_state - collects and tracks the dynamic state of a stripe_head
255  *     for handle_stripe.
256  */
257 struct stripe_head_state {
258 	/* 'syncing' means that we need to read all devices, either
259 	 * to check/correct parity, or to reconstruct a missing device.
260 	 * 'replacing' means we are replacing one or more drives and
261 	 * the source is valid at this point so we don't need to
262 	 * read all devices, just the replacement targets.
263 	 */
264 	int syncing, expanding, expanded, replacing;
265 	int locked, uptodate, to_read, to_write, failed, written;
266 	int to_fill, compute, req_compute, non_overwrite;
267 	int failed_num[2];
268 	int p_failed, q_failed;
269 	int dec_preread_active;
270 	unsigned long ops_request;
271 
272 	struct bio_list return_bi;
273 	struct md_rdev *blocked_rdev;
274 	int handle_bad_blocks;
275 	int log_failed;
276 };
277 
278 /* Flags for struct r5dev.flags */
279 enum r5dev_flags {
280 	R5_UPTODATE,	/* page contains current data */
281 	R5_LOCKED,	/* IO has been submitted on "req" */
282 	R5_DOUBLE_LOCKED,/* Cannot clear R5_LOCKED until 2 writes complete */
283 	R5_OVERWRITE,	/* towrite covers whole page */
284 /* and some that are internal to handle_stripe */
285 	R5_Insync,	/* rdev && rdev->in_sync at start */
286 	R5_Wantread,	/* want to schedule a read */
287 	R5_Wantwrite,
288 	R5_Overlap,	/* There is a pending overlapping request
289 			 * on this block */
290 	R5_ReadNoMerge, /* prevent bio from merging in block-layer */
291 	R5_ReadError,	/* seen a read error here recently */
292 	R5_ReWrite,	/* have tried to over-write the readerror */
293 
294 	R5_Expanded,	/* This block now has post-expand data */
295 	R5_Wantcompute,	/* compute_block in progress treat as
296 			 * uptodate
297 			 */
298 	R5_Wantfill,	/* dev->toread contains a bio that needs
299 			 * filling
300 			 */
301 	R5_Wantdrain,	/* dev->towrite needs to be drained */
302 	R5_WantFUA,	/* Write should be FUA */
303 	R5_SyncIO,	/* The IO is sync */
304 	R5_WriteError,	/* got a write error - need to record it */
305 	R5_MadeGood,	/* A bad block has been fixed by writing to it */
306 	R5_ReadRepl,	/* Will/did read from replacement rather than orig */
307 	R5_MadeGoodRepl,/* A bad block on the replacement device has been
308 			 * fixed by writing to it */
309 	R5_NeedReplace,	/* This device has a replacement which is not
310 			 * up-to-date at this stripe. */
311 	R5_WantReplace, /* We need to update the replacement, we have read
312 			 * data in, and now is a good time to write it out.
313 			 */
314 	R5_Discard,	/* Discard the stripe */
315 	R5_SkipCopy,	/* Don't copy data from bio to stripe cache */
316 };
317 
318 /*
319  * Stripe state
320  */
321 enum {
322 	STRIPE_ACTIVE,
323 	STRIPE_HANDLE,
324 	STRIPE_SYNC_REQUESTED,
325 	STRIPE_SYNCING,
326 	STRIPE_INSYNC,
327 	STRIPE_REPLACED,
328 	STRIPE_PREREAD_ACTIVE,
329 	STRIPE_DELAYED,
330 	STRIPE_DEGRADED,
331 	STRIPE_BIT_DELAY,
332 	STRIPE_EXPANDING,
333 	STRIPE_EXPAND_SOURCE,
334 	STRIPE_EXPAND_READY,
335 	STRIPE_IO_STARTED,	/* do not count towards 'bypass_count' */
336 	STRIPE_FULL_WRITE,	/* all blocks are set to be overwritten */
337 	STRIPE_BIOFILL_RUN,
338 	STRIPE_COMPUTE_RUN,
339 	STRIPE_OPS_REQ_PENDING,
340 	STRIPE_ON_UNPLUG_LIST,
341 	STRIPE_DISCARD,
342 	STRIPE_ON_RELEASE_LIST,
343 	STRIPE_BATCH_READY,
344 	STRIPE_BATCH_ERR,
345 	STRIPE_BITMAP_PENDING,	/* Being added to bitmap, don't add
346 				 * to batch yet.
347 				 */
348 	STRIPE_LOG_TRAPPED, /* trapped into log */
349 };
350 
351 #define STRIPE_EXPAND_SYNC_FLAGS \
352 	((1 << STRIPE_EXPAND_SOURCE) |\
353 	(1 << STRIPE_EXPAND_READY) |\
354 	(1 << STRIPE_EXPANDING) |\
355 	(1 << STRIPE_SYNC_REQUESTED))
356 /*
357  * Operation request flags
358  */
359 enum {
360 	STRIPE_OP_BIOFILL,
361 	STRIPE_OP_COMPUTE_BLK,
362 	STRIPE_OP_PREXOR,
363 	STRIPE_OP_BIODRAIN,
364 	STRIPE_OP_RECONSTRUCT,
365 	STRIPE_OP_CHECK,
366 };
367 
368 /*
369  * RAID parity calculation preferences
370  */
371 enum {
372 	PARITY_DISABLE_RMW = 0,
373 	PARITY_ENABLE_RMW,
374 	PARITY_PREFER_RMW,
375 };
376 
377 /*
378  * Pages requested from set_syndrome_sources()
379  */
380 enum {
381 	SYNDROME_SRC_ALL,
382 	SYNDROME_SRC_WANT_DRAIN,
383 	SYNDROME_SRC_WRITTEN,
384 };
385 /*
386  * Plugging:
387  *
388  * To improve write throughput, we need to delay the handling of some
389  * stripes until there has been a chance that several write requests
390  * for the one stripe have all been collected.
391  * In particular, any write request that would require pre-reading
392  * is put on a "delayed" queue until there are no stripes currently
393  * in a pre-read phase.  Further, if the "delayed" queue is empty when
394  * a stripe is put on it then we "plug" the queue and do not process it
395  * until an unplug call is made. (the unplug_io_fn() is called).
396  *
397  * When preread is initiated on a stripe, we set PREREAD_ACTIVE and add
398  * it to the count of prereading stripes.
399  * When write is initiated, or the stripe refcnt == 0 (just in case) we
400  * clear the PREREAD_ACTIVE flag and decrement the count
401  * Whenever the 'handle' queue is empty and the device is not plugged, we
402  * move any strips from delayed to handle and clear the DELAYED flag and set
403  * PREREAD_ACTIVE.
404  * In stripe_handle, if we find pre-reading is necessary, we do it if
405  * PREREAD_ACTIVE is set, else we set DELAYED which will send it to the delayed queue.
406  * HANDLE gets cleared if stripe_handle leaves nothing locked.
407  */
408 
409 struct disk_info {
410 	struct md_rdev	*rdev, *replacement;
411 };
412 
413 /* NOTE NR_STRIPE_HASH_LOCKS must remain below 64.
414  * This is because we sometimes take all the spinlocks
415  * and creating that much locking depth can cause
416  * problems.
417  */
418 #define NR_STRIPE_HASH_LOCKS 8
419 #define STRIPE_HASH_LOCKS_MASK (NR_STRIPE_HASH_LOCKS - 1)
420 
421 struct r5worker {
422 	struct work_struct work;
423 	struct r5worker_group *group;
424 	struct list_head temp_inactive_list[NR_STRIPE_HASH_LOCKS];
425 	bool working;
426 };
427 
428 struct r5worker_group {
429 	struct list_head handle_list;
430 	struct r5conf *conf;
431 	struct r5worker *workers;
432 	int stripes_cnt;
433 };
434 
435 struct r5conf {
436 	struct hlist_head	*stripe_hashtbl;
437 	/* only protect corresponding hash list and inactive_list */
438 	spinlock_t		hash_locks[NR_STRIPE_HASH_LOCKS];
439 	struct mddev		*mddev;
440 	int			chunk_sectors;
441 	int			level, algorithm, rmw_level;
442 	int			max_degraded;
443 	int			raid_disks;
444 	int			max_nr_stripes;
445 	int			min_nr_stripes;
446 
447 	/* reshape_progress is the leading edge of a 'reshape'
448 	 * It has value MaxSector when no reshape is happening
449 	 * If delta_disks < 0, it is the last sector we started work on,
450 	 * else is it the next sector to work on.
451 	 */
452 	sector_t		reshape_progress;
453 	/* reshape_safe is the trailing edge of a reshape.  We know that
454 	 * before (or after) this address, all reshape has completed.
455 	 */
456 	sector_t		reshape_safe;
457 	int			previous_raid_disks;
458 	int			prev_chunk_sectors;
459 	int			prev_algo;
460 	short			generation; /* increments with every reshape */
461 	seqcount_t		gen_lock;	/* lock against generation changes */
462 	unsigned long		reshape_checkpoint; /* Time we last updated
463 						     * metadata */
464 	long long		min_offset_diff; /* minimum difference between
465 						  * data_offset and
466 						  * new_data_offset across all
467 						  * devices.  May be negative,
468 						  * but is closest to zero.
469 						  */
470 
471 	struct list_head	handle_list; /* stripes needing handling */
472 	struct list_head	hold_list; /* preread ready stripes */
473 	struct list_head	delayed_list; /* stripes that have plugged requests */
474 	struct list_head	bitmap_list; /* stripes delaying awaiting bitmap update */
475 	struct bio		*retry_read_aligned; /* currently retrying aligned bios   */
476 	struct bio		*retry_read_aligned_list; /* aligned bios retry list  */
477 	atomic_t		preread_active_stripes; /* stripes with scheduled io */
478 	atomic_t		active_aligned_reads;
479 	atomic_t		pending_full_writes; /* full write backlog */
480 	int			bypass_count; /* bypassed prereads */
481 	int			bypass_threshold; /* preread nice */
482 	int			skip_copy; /* Don't copy data from bio to stripe cache */
483 	struct list_head	*last_hold; /* detect hold_list promotions */
484 
485 	/* bios to have bi_end_io called after metadata is synced */
486 	struct bio_list		return_bi;
487 
488 	atomic_t		reshape_stripes; /* stripes with pending writes for reshape */
489 	/* unfortunately we need two cache names as we temporarily have
490 	 * two caches.
491 	 */
492 	int			active_name;
493 	char			cache_name[2][32];
494 	struct kmem_cache	*slab_cache; /* for allocating stripes */
495 	struct mutex		cache_size_mutex; /* Protect changes to cache size */
496 
497 	int			seq_flush, seq_write;
498 	int			quiesce;
499 
500 	int			fullsync;  /* set to 1 if a full sync is needed,
501 					    * (fresh device added).
502 					    * Cleared when a sync completes.
503 					    */
504 	int			recovery_disabled;
505 	/* per cpu variables */
506 	struct raid5_percpu {
507 		struct page	*spare_page; /* Used when checking P/Q in raid6 */
508 		struct flex_array *scribble;   /* space for constructing buffer
509 					      * lists and performing address
510 					      * conversions
511 					      */
512 	} __percpu *percpu;
513 	int scribble_disks;
514 	int scribble_sectors;
515 #ifdef CONFIG_HOTPLUG_CPU
516 	struct notifier_block	cpu_notify;
517 #endif
518 
519 	/*
520 	 * Free stripes pool
521 	 */
522 	atomic_t		active_stripes;
523 	struct list_head	inactive_list[NR_STRIPE_HASH_LOCKS];
524 	atomic_t		empty_inactive_list_nr;
525 	struct llist_head	released_stripes;
526 	wait_queue_head_t	wait_for_quiescent;
527 	wait_queue_head_t	wait_for_stripe;
528 	wait_queue_head_t	wait_for_overlap;
529 	unsigned long		cache_state;
530 #define R5_INACTIVE_BLOCKED	1	/* release of inactive stripes blocked,
531 					 * waiting for 25% to be free
532 					 */
533 #define R5_ALLOC_MORE		2	/* It might help to allocate another
534 					 * stripe.
535 					 */
536 #define R5_DID_ALLOC		4	/* A stripe was allocated, don't allocate
537 					 * more until at least one has been
538 					 * released.  This avoids flooding
539 					 * the cache.
540 					 */
541 	struct shrinker		shrinker;
542 	int			pool_size; /* number of disks in stripeheads in pool */
543 	spinlock_t		device_lock;
544 	struct disk_info	*disks;
545 
546 	/* When taking over an array from a different personality, we store
547 	 * the new thread here until we fully activate the array.
548 	 */
549 	struct md_thread	*thread;
550 	struct list_head	temp_inactive_list[NR_STRIPE_HASH_LOCKS];
551 	struct r5worker_group	*worker_groups;
552 	int			group_cnt;
553 	int			worker_cnt_per_group;
554 	struct r5l_log		*log;
555 };
556 
557 
558 /*
559  * Our supported algorithms
560  */
561 #define ALGORITHM_LEFT_ASYMMETRIC	0 /* Rotating Parity N with Data Restart */
562 #define ALGORITHM_RIGHT_ASYMMETRIC	1 /* Rotating Parity 0 with Data Restart */
563 #define ALGORITHM_LEFT_SYMMETRIC	2 /* Rotating Parity N with Data Continuation */
564 #define ALGORITHM_RIGHT_SYMMETRIC	3 /* Rotating Parity 0 with Data Continuation */
565 
566 /* Define non-rotating (raid4) algorithms.  These allow
567  * conversion of raid4 to raid5.
568  */
569 #define ALGORITHM_PARITY_0		4 /* P or P,Q are initial devices */
570 #define ALGORITHM_PARITY_N		5 /* P or P,Q are final devices. */
571 
572 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
573  * Firstly, the exact positioning of the parity block is slightly
574  * different between the 'LEFT_*' modes of md and the "_N_*" modes
575  * of DDF.
576  * Secondly, or order of datablocks over which the Q syndrome is computed
577  * is different.
578  * Consequently we have different layouts for DDF/raid6 than md/raid6.
579  * These layouts are from the DDFv1.2 spec.
580  * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
581  * leaves RLQ=3 as 'Vendor Specific'
582  */
583 
584 #define ALGORITHM_ROTATING_ZERO_RESTART	8 /* DDF PRL=6 RLQ=1 */
585 #define ALGORITHM_ROTATING_N_RESTART	9 /* DDF PRL=6 RLQ=2 */
586 #define ALGORITHM_ROTATING_N_CONTINUE	10 /*DDF PRL=6 RLQ=3 */
587 
588 /* For every RAID5 algorithm we define a RAID6 algorithm
589  * with exactly the same layout for data and parity, and
590  * with the Q block always on the last device (N-1).
591  * This allows trivial conversion from RAID5 to RAID6
592  */
593 #define ALGORITHM_LEFT_ASYMMETRIC_6	16
594 #define ALGORITHM_RIGHT_ASYMMETRIC_6	17
595 #define ALGORITHM_LEFT_SYMMETRIC_6	18
596 #define ALGORITHM_RIGHT_SYMMETRIC_6	19
597 #define ALGORITHM_PARITY_0_6		20
598 #define ALGORITHM_PARITY_N_6		ALGORITHM_PARITY_N
599 
600 static inline int algorithm_valid_raid5(int layout)
601 {
602 	return (layout >= 0) &&
603 		(layout <= 5);
604 }
605 static inline int algorithm_valid_raid6(int layout)
606 {
607 	return (layout >= 0 && layout <= 5)
608 		||
609 		(layout >= 8 && layout <= 10)
610 		||
611 		(layout >= 16 && layout <= 20);
612 }
613 
614 static inline int algorithm_is_DDF(int layout)
615 {
616 	return layout >= 8 && layout <= 10;
617 }
618 
619 extern void md_raid5_kick_device(struct r5conf *conf);
620 extern int raid5_set_cache_size(struct mddev *mddev, int size);
621 extern sector_t raid5_compute_blocknr(struct stripe_head *sh, int i, int previous);
622 extern void raid5_release_stripe(struct stripe_head *sh);
623 extern sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
624 				     int previous, int *dd_idx,
625 				     struct stripe_head *sh);
626 extern struct stripe_head *
627 raid5_get_active_stripe(struct r5conf *conf, sector_t sector,
628 			int previous, int noblock, int noquiesce);
629 extern int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev);
630 extern void r5l_exit_log(struct r5l_log *log);
631 extern int r5l_write_stripe(struct r5l_log *log, struct stripe_head *head_sh);
632 extern void r5l_write_stripe_run(struct r5l_log *log);
633 extern void r5l_flush_stripe_to_raid(struct r5l_log *log);
634 extern void r5l_stripe_write_finished(struct stripe_head *sh);
635 extern int r5l_handle_flush_request(struct r5l_log *log, struct bio *bio);
636 extern void r5l_quiesce(struct r5l_log *log, int state);
637 extern bool r5l_log_disk_error(struct r5conf *conf);
638 #endif
639