4 #include <linux/raid/xor.h>
5 #include <linux/dmaengine.h>
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.
15 * The flag bits that are used to represent these states are:
16 * R5_UPTODATE and R5_LOCKED
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
27 * The possible state transitions are:
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)
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:
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.
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
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.
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.
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.
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.
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.
95 * The inactive_list, handle_list and hash bucket lists are all protected by the
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
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.
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) ..
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
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
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
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.
158 * Operations state - intermediate states that are visible outside of
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)
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
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,
185 * enum reconstruct_states - handles writing or expanding a stripe
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,
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
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
217 enum check_states check_state;
218 enum reconstruct_states reconstruct_state;
219 spinlock_t stripe_lock;
221 struct r5worker_group *group;
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*/
228 struct r5l_io_unit *log_io;
229 struct ppl_io_unit *ppl_io;
232 struct list_head log_list;
233 sector_t log_start; /* first meta block on the journal */
234 struct list_head r5c; /* for r5c_cache->stripe_in_journal */
236 struct page *ppl_page; /* partial parity of this stripe */
238 * struct stripe_operations
239 * @target - STRIPE_OP_COMPUTE_BLK target
240 * @target2 - 2nd compute target in the raid6 case
241 * @zero_sum_result - P and Q verification flags
242 * @request - async service request flags for raid_run_ops
244 struct stripe_operations {
246 enum sum_check_flags zero_sum_result;
249 /* rreq and rvec are used for the replacement device when
250 * writing data to both devices.
252 struct bio req, rreq;
253 struct bio_vec vec, rvec;
254 struct page *page, *orig_page;
255 struct bio *toread, *read, *towrite, *written;
256 sector_t sector; /* sector of this page */
259 } dev[1]; /* allocated with extra space depending of RAID geometry */
262 /* stripe_head_state - collects and tracks the dynamic state of a stripe_head
265 struct stripe_head_state {
266 /* 'syncing' means that we need to read all devices, either
267 * to check/correct parity, or to reconstruct a missing device.
268 * 'replacing' means we are replacing one or more drives and
269 * the source is valid at this point so we don't need to
270 * read all devices, just the replacement targets.
272 int syncing, expanding, expanded, replacing;
273 int locked, uptodate, to_read, to_write, failed, written;
274 int to_fill, compute, req_compute, non_overwrite;
275 int injournal, just_cached;
277 int p_failed, q_failed;
278 int dec_preread_active;
279 unsigned long ops_request;
281 struct md_rdev *blocked_rdev;
282 int handle_bad_blocks;
284 int waiting_extra_page;
287 /* Flags for struct r5dev.flags */
289 R5_UPTODATE, /* page contains current data */
290 R5_LOCKED, /* IO has been submitted on "req" */
291 R5_DOUBLE_LOCKED,/* Cannot clear R5_LOCKED until 2 writes complete */
292 R5_OVERWRITE, /* towrite covers whole page */
293 /* and some that are internal to handle_stripe */
294 R5_Insync, /* rdev && rdev->in_sync at start */
295 R5_Wantread, /* want to schedule a read */
297 R5_Overlap, /* There is a pending overlapping request
299 R5_ReadNoMerge, /* prevent bio from merging in block-layer */
300 R5_ReadError, /* seen a read error here recently */
301 R5_ReWrite, /* have tried to over-write the readerror */
303 R5_Expanded, /* This block now has post-expand data */
304 R5_Wantcompute, /* compute_block in progress treat as
307 R5_Wantfill, /* dev->toread contains a bio that needs
310 R5_Wantdrain, /* dev->towrite needs to be drained */
311 R5_WantFUA, /* Write should be FUA */
312 R5_SyncIO, /* The IO is sync */
313 R5_WriteError, /* got a write error - need to record it */
314 R5_MadeGood, /* A bad block has been fixed by writing to it */
315 R5_ReadRepl, /* Will/did read from replacement rather than orig */
316 R5_MadeGoodRepl,/* A bad block on the replacement device has been
317 * fixed by writing to it */
318 R5_NeedReplace, /* This device has a replacement which is not
319 * up-to-date at this stripe. */
320 R5_WantReplace, /* We need to update the replacement, we have read
321 * data in, and now is a good time to write it out.
323 R5_Discard, /* Discard the stripe */
324 R5_SkipCopy, /* Don't copy data from bio to stripe cache */
325 R5_InJournal, /* data being written is in the journal device.
326 * if R5_InJournal is set for parity pd_idx, all the
327 * data and parity being written are in the journal
330 R5_OrigPageUPTDODATE, /* with write back cache, we read old data into
331 * dev->orig_page for prexor. When this flag is
332 * set, orig_page contains latest data in the
343 STRIPE_SYNC_REQUESTED,
347 STRIPE_PREREAD_ACTIVE,
352 STRIPE_EXPAND_SOURCE,
354 STRIPE_IO_STARTED, /* do not count towards 'bypass_count' */
355 STRIPE_FULL_WRITE, /* all blocks are set to be overwritten */
358 STRIPE_OPS_REQ_PENDING,
359 STRIPE_ON_UNPLUG_LIST,
361 STRIPE_ON_RELEASE_LIST,
364 STRIPE_BITMAP_PENDING, /* Being added to bitmap, don't add
367 STRIPE_LOG_TRAPPED, /* trapped into log (see raid5-cache.c)
368 * this bit is used in two scenarios:
371 * set in first entry of r5l_write_stripe
372 * clear in second entry of r5l_write_stripe
373 * used to bypass logic in handle_stripe
376 * set in r5c_try_caching_write()
377 * clear when journal write is done
378 * used to initiate r5c_cache_data()
379 * also used to bypass logic in handle_stripe
381 STRIPE_R5C_CACHING, /* the stripe is in caching phase
382 * see more detail in the raid5-cache.c
384 STRIPE_R5C_PARTIAL_STRIPE, /* in r5c cache (to-be/being handled or
385 * in conf->r5c_partial_stripe_list)
387 STRIPE_R5C_FULL_STRIPE, /* in r5c cache (to-be/being handled or
388 * in conf->r5c_full_stripe_list)
390 STRIPE_R5C_PREFLUSH, /* need to flush journal device */
393 #define STRIPE_EXPAND_SYNC_FLAGS \
394 ((1 << STRIPE_EXPAND_SOURCE) |\
395 (1 << STRIPE_EXPAND_READY) |\
396 (1 << STRIPE_EXPANDING) |\
397 (1 << STRIPE_SYNC_REQUESTED))
399 * Operation request flags
403 STRIPE_OP_COMPUTE_BLK,
406 STRIPE_OP_RECONSTRUCT,
408 STRIPE_OP_PARTIAL_PARITY,
412 * RAID parity calculation preferences
415 PARITY_DISABLE_RMW = 0,
421 * Pages requested from set_syndrome_sources()
425 SYNDROME_SRC_WANT_DRAIN,
426 SYNDROME_SRC_WRITTEN,
431 * To improve write throughput, we need to delay the handling of some
432 * stripes until there has been a chance that several write requests
433 * for the one stripe have all been collected.
434 * In particular, any write request that would require pre-reading
435 * is put on a "delayed" queue until there are no stripes currently
436 * in a pre-read phase. Further, if the "delayed" queue is empty when
437 * a stripe is put on it then we "plug" the queue and do not process it
438 * until an unplug call is made. (the unplug_io_fn() is called).
440 * When preread is initiated on a stripe, we set PREREAD_ACTIVE and add
441 * it to the count of prereading stripes.
442 * When write is initiated, or the stripe refcnt == 0 (just in case) we
443 * clear the PREREAD_ACTIVE flag and decrement the count
444 * Whenever the 'handle' queue is empty and the device is not plugged, we
445 * move any strips from delayed to handle and clear the DELAYED flag and set
447 * In stripe_handle, if we find pre-reading is necessary, we do it if
448 * PREREAD_ACTIVE is set, else we set DELAYED which will send it to the delayed queue.
449 * HANDLE gets cleared if stripe_handle leaves nothing locked.
453 struct md_rdev *rdev, *replacement;
454 struct page *extra_page; /* extra page to use in prexor */
461 #define NR_STRIPES 256
462 #define STRIPE_SIZE PAGE_SIZE
463 #define STRIPE_SHIFT (PAGE_SHIFT - 9)
464 #define STRIPE_SECTORS (STRIPE_SIZE>>9)
465 #define IO_THRESHOLD 1
466 #define BYPASS_THRESHOLD 1
467 #define NR_HASH (PAGE_SIZE / sizeof(struct hlist_head))
468 #define HASH_MASK (NR_HASH - 1)
469 #define MAX_STRIPE_BATCH 8
471 /* bio's attached to a stripe+device for I/O are linked together in bi_sector
472 * order without overlap. There may be several bio's per stripe+device, and
473 * a bio could span several devices.
474 * When walking this list for a particular stripe+device, we must never proceed
475 * beyond a bio that extends past this device, as the next bio might no longer
477 * This function is used to determine the 'next' bio in the list, given the
478 * sector of the current stripe+device
480 static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
482 int sectors = bio_sectors(bio);
484 if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
490 /* NOTE NR_STRIPE_HASH_LOCKS must remain below 64.
491 * This is because we sometimes take all the spinlocks
492 * and creating that much locking depth can cause
495 #define NR_STRIPE_HASH_LOCKS 8
496 #define STRIPE_HASH_LOCKS_MASK (NR_STRIPE_HASH_LOCKS - 1)
499 struct work_struct work;
500 struct r5worker_group *group;
501 struct list_head temp_inactive_list[NR_STRIPE_HASH_LOCKS];
505 struct r5worker_group {
506 struct list_head handle_list;
507 struct list_head loprio_list;
509 struct r5worker *workers;
514 * r5c journal modes of the array: write-back or write-through.
515 * write-through mode has identical behavior as existing log only
518 enum r5c_journal_mode {
519 R5C_JOURNAL_MODE_WRITE_THROUGH = 0,
520 R5C_JOURNAL_MODE_WRITE_BACK = 1,
523 enum r5_cache_state {
524 R5_INACTIVE_BLOCKED, /* release of inactive stripes blocked,
525 * waiting for 25% to be free
527 R5_ALLOC_MORE, /* It might help to allocate another
530 R5_DID_ALLOC, /* A stripe was allocated, don't allocate
531 * more until at least one has been
532 * released. This avoids flooding
535 R5C_LOG_TIGHT, /* log device space tight, need to
536 * prioritize stripes at last_checkpoint
538 R5C_LOG_CRITICAL, /* log device is running out of space,
539 * only process stripes that are already
542 R5C_EXTRA_PAGE_IN_USE, /* a stripe is using disk_info.extra_page
547 #define PENDING_IO_MAX 512
548 #define PENDING_IO_ONE_FLUSH 128
549 struct r5pending_data {
550 struct list_head sibling;
551 sector_t sector; /* stripe sector */
552 struct bio_list bios;
556 struct hlist_head *stripe_hashtbl;
557 /* only protect corresponding hash list and inactive_list */
558 spinlock_t hash_locks[NR_STRIPE_HASH_LOCKS];
561 int level, algorithm, rmw_level;
567 /* reshape_progress is the leading edge of a 'reshape'
568 * It has value MaxSector when no reshape is happening
569 * If delta_disks < 0, it is the last sector we started work on,
570 * else is it the next sector to work on.
572 sector_t reshape_progress;
573 /* reshape_safe is the trailing edge of a reshape. We know that
574 * before (or after) this address, all reshape has completed.
576 sector_t reshape_safe;
577 int previous_raid_disks;
578 int prev_chunk_sectors;
580 short generation; /* increments with every reshape */
581 seqcount_t gen_lock; /* lock against generation changes */
582 unsigned long reshape_checkpoint; /* Time we last updated
584 long long min_offset_diff; /* minimum difference between
586 * new_data_offset across all
587 * devices. May be negative,
588 * but is closest to zero.
591 struct list_head handle_list; /* stripes needing handling */
592 struct list_head loprio_list; /* low priority stripes */
593 struct list_head hold_list; /* preread ready stripes */
594 struct list_head delayed_list; /* stripes that have plugged requests */
595 struct list_head bitmap_list; /* stripes delaying awaiting bitmap update */
596 struct bio *retry_read_aligned; /* currently retrying aligned bios */
597 unsigned int retry_read_offset; /* sector offset into retry_read_aligned */
598 struct bio *retry_read_aligned_list; /* aligned bios retry list */
599 atomic_t preread_active_stripes; /* stripes with scheduled io */
600 atomic_t active_aligned_reads;
601 atomic_t pending_full_writes; /* full write backlog */
602 int bypass_count; /* bypassed prereads */
603 int bypass_threshold; /* preread nice */
604 int skip_copy; /* Don't copy data from bio to stripe cache */
605 struct list_head *last_hold; /* detect hold_list promotions */
607 atomic_t reshape_stripes; /* stripes with pending writes for reshape */
608 /* unfortunately we need two cache names as we temporarily have
612 char cache_name[2][32];
613 struct kmem_cache *slab_cache; /* for allocating stripes */
614 struct mutex cache_size_mutex; /* Protect changes to cache size */
616 int seq_flush, seq_write;
619 int fullsync; /* set to 1 if a full sync is needed,
620 * (fresh device added).
621 * Cleared when a sync completes.
623 int recovery_disabled;
624 /* per cpu variables */
625 struct raid5_percpu {
626 struct page *spare_page; /* Used when checking P/Q in raid6 */
627 struct flex_array *scribble; /* space for constructing buffer
628 * lists and performing address
633 int scribble_sectors;
634 struct hlist_node node;
639 atomic_t active_stripes;
640 struct list_head inactive_list[NR_STRIPE_HASH_LOCKS];
642 atomic_t r5c_cached_full_stripes;
643 struct list_head r5c_full_stripe_list;
644 atomic_t r5c_cached_partial_stripes;
645 struct list_head r5c_partial_stripe_list;
646 atomic_t r5c_flushing_full_stripes;
647 atomic_t r5c_flushing_partial_stripes;
649 atomic_t empty_inactive_list_nr;
650 struct llist_head released_stripes;
651 wait_queue_head_t wait_for_quiescent;
652 wait_queue_head_t wait_for_stripe;
653 wait_queue_head_t wait_for_overlap;
654 unsigned long cache_state;
655 struct shrinker shrinker;
656 int pool_size; /* number of disks in stripeheads in pool */
657 spinlock_t device_lock;
658 struct disk_info *disks;
659 struct bio_set *bio_split;
661 /* When taking over an array from a different personality, we store
662 * the new thread here until we fully activate the array.
664 struct md_thread *thread;
665 struct list_head temp_inactive_list[NR_STRIPE_HASH_LOCKS];
666 struct r5worker_group *worker_groups;
668 int worker_cnt_per_group;
672 spinlock_t pending_bios_lock;
673 bool batch_bio_dispatch;
674 struct r5pending_data *pending_data;
675 struct list_head free_list;
676 struct list_head pending_list;
677 int pending_data_cnt;
678 struct r5pending_data *next_pending_data;
683 * Our supported algorithms
685 #define ALGORITHM_LEFT_ASYMMETRIC 0 /* Rotating Parity N with Data Restart */
686 #define ALGORITHM_RIGHT_ASYMMETRIC 1 /* Rotating Parity 0 with Data Restart */
687 #define ALGORITHM_LEFT_SYMMETRIC 2 /* Rotating Parity N with Data Continuation */
688 #define ALGORITHM_RIGHT_SYMMETRIC 3 /* Rotating Parity 0 with Data Continuation */
690 /* Define non-rotating (raid4) algorithms. These allow
691 * conversion of raid4 to raid5.
693 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
694 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
696 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
697 * Firstly, the exact positioning of the parity block is slightly
698 * different between the 'LEFT_*' modes of md and the "_N_*" modes
700 * Secondly, or order of datablocks over which the Q syndrome is computed
702 * Consequently we have different layouts for DDF/raid6 than md/raid6.
703 * These layouts are from the DDFv1.2 spec.
704 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
705 * leaves RLQ=3 as 'Vendor Specific'
708 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
709 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
710 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
712 /* For every RAID5 algorithm we define a RAID6 algorithm
713 * with exactly the same layout for data and parity, and
714 * with the Q block always on the last device (N-1).
715 * This allows trivial conversion from RAID5 to RAID6
717 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
718 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
719 #define ALGORITHM_LEFT_SYMMETRIC_6 18
720 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
721 #define ALGORITHM_PARITY_0_6 20
722 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
724 static inline int algorithm_valid_raid5(int layout)
726 return (layout >= 0) &&
729 static inline int algorithm_valid_raid6(int layout)
731 return (layout >= 0 && layout <= 5)
733 (layout >= 8 && layout <= 10)
735 (layout >= 16 && layout <= 20);
738 static inline int algorithm_is_DDF(int layout)
740 return layout >= 8 && layout <= 10;
743 extern void md_raid5_kick_device(struct r5conf *conf);
744 extern int raid5_set_cache_size(struct mddev *mddev, int size);
745 extern sector_t raid5_compute_blocknr(struct stripe_head *sh, int i, int previous);
746 extern void raid5_release_stripe(struct stripe_head *sh);
747 extern sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
748 int previous, int *dd_idx,
749 struct stripe_head *sh);
750 extern struct stripe_head *
751 raid5_get_active_stripe(struct r5conf *conf, sector_t sector,
752 int previous, int noblock, int noquiesce);
753 extern int raid5_calc_degraded(struct r5conf *conf);
754 extern int r5c_journal_mode_set(struct mddev *mddev, int journal_mode);