blob: 7c615613c3818c59264e34ebc2c6d91879c3d776 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _RAID10_H
2#define _RAID10_H
3
Linus Torvalds1da177e2005-04-16 15:20:36 -07004struct mirror_info {
NeilBrown69335ef2011-12-23 10:17:54 +11005 struct md_rdev *rdev, *replacement;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 sector_t head_position;
NeilBrown2bb77732011-07-27 11:00:36 +10007 int recovery_disabled; /* matches
8 * mddev->recovery_disabled
9 * when we shouldn't try
10 * recovering this device.
11 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070012};
13
NeilBrowne879a872011-10-11 16:49:02 +110014struct r10conf {
NeilBrownfd01b882011-10-11 16:47:53 +110015 struct mddev *mddev;
NeilBrown0f6d02d2011-10-11 16:48:46 +110016 struct mirror_info *mirrors;
Linus Torvalds1da177e2005-04-16 15:20:36 -070017 int raid_disks;
Linus Torvalds1da177e2005-04-16 15:20:36 -070018 spinlock_t device_lock;
19
20 /* geometry */
NeilBrown69335ef2011-12-23 10:17:54 +110021 int near_copies; /* number of copies laid out
22 * raid0 style */
Lucas De Marchi25985ed2011-03-30 22:57:33 -030023 int far_copies; /* number of copies laid out
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 * at large strides across drives
25 */
NeilBrown69335ef2011-12-23 10:17:54 +110026 int far_offset; /* far_copies are offset by 1
27 * stripe instead of many
NeilBrownc93983b2006-06-26 00:27:41 -070028 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070029 int copies; /* near_copies * far_copies.
30 * must be <= raid_disks
31 */
32 sector_t stride; /* distance between far copies.
NeilBrownc93983b2006-06-26 00:27:41 -070033 * This is size / far_copies unless
34 * far_offset, in which case it is
35 * 1 stripe.
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 */
37
NeilBrown69335ef2011-12-23 10:17:54 +110038 sector_t dev_sectors; /* temp copy of
39 * mddev->dev_sectors */
Trela, Maciejdab8b292010-03-08 16:02:45 +110040
NeilBrown69335ef2011-12-23 10:17:54 +110041 int chunk_shift; /* shift from chunks to sectors */
42 sector_t chunk_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44 struct list_head retry_list;
NeilBrown6cce3b22006-01-06 00:20:16 -080045 /* queue pending writes and submit them on unplug */
46 struct bio_list pending_bio_list;
NeilBrown34db0cd2011-10-11 16:50:01 +110047 int pending_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
49 spinlock_t resync_lock;
NeilBrown69335ef2011-12-23 10:17:54 +110050 int nr_pending;
51 int nr_waiting;
52 int nr_queued;
53 int barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070054 sector_t next_resync;
NeilBrown6cce3b22006-01-06 00:20:16 -080055 int fullsync; /* set to 1 if a full sync is needed,
56 * (fresh device added).
57 * Cleared when a sync completes.
58 */
NeilBrown69335ef2011-12-23 10:17:54 +110059 int have_replacement; /* There is at least one
60 * replacement device.
61 */
NeilBrown0a27ec92006-01-06 00:20:13 -080062 wait_queue_head_t wait_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
NeilBrown69335ef2011-12-23 10:17:54 +110064 mempool_t *r10bio_pool;
65 mempool_t *r10buf_pool;
NeilBrown4443ae12006-01-06 00:20:28 -080066 struct page *tmppage;
Trela, Maciejdab8b292010-03-08 16:02:45 +110067
68 /* When taking over an array from a different personality, we store
69 * the new thread here until we fully activate the array.
70 */
NeilBrown2b8bf342011-10-11 16:48:23 +110071 struct md_thread *thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072};
73
Linus Torvalds1da177e2005-04-16 15:20:36 -070074/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 * this is our 'private' RAID10 bio.
76 *
77 * it contains information about what kind of IO operations were started
78 * for this RAID10 operation, and about their status:
79 */
80
NeilBrown9f2c9d12011-10-11 16:48:43 +110081struct r10bio {
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 atomic_t remaining; /* 'have we finished' count,
83 * used from IRQ handlers
84 */
85 sector_t sector; /* virtual sector number */
86 int sectors;
87 unsigned long state;
NeilBrownfd01b882011-10-11 16:47:53 +110088 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 /*
90 * original bio going to /dev/mdx
91 */
92 struct bio *master_bio;
93 /*
94 * if the IO is in READ direction, then this is where we read
95 */
96 int read_slot;
97
98 struct list_head retry_list;
99 /*
100 * if the IO is in WRITE direction, then multiple bios are used,
101 * one for each copy.
102 * When resyncing we also use one for each copy.
103 * When reconstructing, we use 2 bios, one for read, one for write.
104 * We choose the number when they are allocated.
NeilBrown69335ef2011-12-23 10:17:54 +1100105 * We sometimes need an extra bio to write to the replacement.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 */
107 struct {
NeilBrown69335ef2011-12-23 10:17:54 +1100108 struct bio *bio;
109 union {
110 struct bio *repl_bio; /* used for resync and
111 * writes */
112 struct md_rdev *rdev; /* used for reads
113 * (read_slot >= 0) */
114 };
115 sector_t addr;
116 int devnum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 } devs[0];
118};
119
NeilBrown0eb3ff12006-01-06 00:20:29 -0800120/* when we get a read error on a read-only array, we redirect to another
121 * device without failing the first device, or trying to over-write to
122 * correct the read error. To keep track of bad blocks on a per-bio
123 * level, we store IO_BLOCKED in the appropriate 'bios' pointer
124 */
125#define IO_BLOCKED ((struct bio*)1)
NeilBrown749c55e2011-07-28 11:39:24 +1000126/* When we successfully write to a known bad-block, we need to remove the
127 * bad-block marking which must be done from process context. So we record
128 * the success by setting devs[n].bio to IO_MADE_GOOD
129 */
130#define IO_MADE_GOOD ((struct bio *)2)
131
132#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
NeilBrown0eb3ff12006-01-06 00:20:29 -0800133
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134/* bits for r10bio.state */
NeilBrown69335ef2011-12-23 10:17:54 +1100135enum r10bio_state {
136 R10BIO_Uptodate,
137 R10BIO_IsSync,
138 R10BIO_IsRecover,
139 R10BIO_Degraded,
NeilBrown856e08e2011-07-28 11:39:23 +1000140/* Set ReadError on bios that experience a read error
141 * so that raid10d knows what to do with them.
142 */
NeilBrown69335ef2011-12-23 10:17:54 +1100143 R10BIO_ReadError,
NeilBrown749c55e2011-07-28 11:39:24 +1000144/* If a write for this request means we can clear some
145 * known-bad-block records, we set this flag.
146 */
NeilBrown69335ef2011-12-23 10:17:54 +1100147 R10BIO_MadeGood,
148 R10BIO_WriteError,
149};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150#endif