blob: 4c4942ac46fc27e82643e4e752871f504572a8b2 [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 spinlock_t device_lock;
18
19 /* geometry */
NeilBrown5cf00fc2012-05-21 09:28:20 +100020 struct geom {
21 int raid_disks;
22 int near_copies; /* number of copies laid out
NeilBrown69335ef2011-12-23 10:17:54 +110023 * raid0 style */
NeilBrown5cf00fc2012-05-21 09:28:20 +100024 int far_copies; /* number of copies laid out
Linus Torvalds1da177e2005-04-16 15:20:36 -070025 * at large strides across drives
26 */
NeilBrown5cf00fc2012-05-21 09:28:20 +100027 int far_offset; /* far_copies are offset by 1
NeilBrown69335ef2011-12-23 10:17:54 +110028 * stripe instead of many
NeilBrownc93983b2006-06-26 00:27:41 -070029 */
NeilBrown5cf00fc2012-05-21 09:28:20 +100030 sector_t stride; /* distance between far copies.
NeilBrownc93983b2006-06-26 00:27:41 -070031 * This is size / far_copies unless
32 * far_offset, in which case it is
33 * 1 stripe.
Linus Torvalds1da177e2005-04-16 15:20:36 -070034 */
NeilBrown5cf00fc2012-05-21 09:28:20 +100035 int chunk_shift; /* shift from chunks to sectors */
36 sector_t chunk_mask;
37 } geo;
38 int copies; /* near_copies * far_copies.
39 * must be <= raid_disks
40 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
NeilBrown69335ef2011-12-23 10:17:54 +110042 sector_t dev_sectors; /* temp copy of
43 * mddev->dev_sectors */
Trela, Maciejdab8b292010-03-08 16:02:45 +110044
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 struct list_head retry_list;
NeilBrown6cce3b22006-01-06 00:20:16 -080046 /* queue pending writes and submit them on unplug */
47 struct bio_list pending_bio_list;
NeilBrown34db0cd2011-10-11 16:50:01 +110048 int pending_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070049
50 spinlock_t resync_lock;
NeilBrown69335ef2011-12-23 10:17:54 +110051 int nr_pending;
52 int nr_waiting;
53 int nr_queued;
54 int barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070055 sector_t next_resync;
NeilBrown6cce3b22006-01-06 00:20:16 -080056 int fullsync; /* set to 1 if a full sync is needed,
57 * (fresh device added).
58 * Cleared when a sync completes.
59 */
NeilBrown69335ef2011-12-23 10:17:54 +110060 int have_replacement; /* There is at least one
61 * replacement device.
62 */
NeilBrown0a27ec92006-01-06 00:20:13 -080063 wait_queue_head_t wait_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
NeilBrown69335ef2011-12-23 10:17:54 +110065 mempool_t *r10bio_pool;
66 mempool_t *r10buf_pool;
NeilBrown4443ae12006-01-06 00:20:28 -080067 struct page *tmppage;
Trela, Maciejdab8b292010-03-08 16:02:45 +110068
69 /* When taking over an array from a different personality, we store
70 * the new thread here until we fully activate the array.
71 */
NeilBrown2b8bf342011-10-11 16:48:23 +110072 struct md_thread *thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073};
74
Linus Torvalds1da177e2005-04-16 15:20:36 -070075/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 * this is our 'private' RAID10 bio.
77 *
78 * it contains information about what kind of IO operations were started
79 * for this RAID10 operation, and about their status:
80 */
81
NeilBrown9f2c9d12011-10-11 16:48:43 +110082struct r10bio {
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 atomic_t remaining; /* 'have we finished' count,
84 * used from IRQ handlers
85 */
86 sector_t sector; /* virtual sector number */
87 int sectors;
88 unsigned long state;
NeilBrownfd01b882011-10-11 16:47:53 +110089 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 /*
91 * original bio going to /dev/mdx
92 */
93 struct bio *master_bio;
94 /*
95 * if the IO is in READ direction, then this is where we read
96 */
97 int read_slot;
98
99 struct list_head retry_list;
100 /*
101 * if the IO is in WRITE direction, then multiple bios are used,
102 * one for each copy.
103 * When resyncing we also use one for each copy.
104 * When reconstructing, we use 2 bios, one for read, one for write.
105 * We choose the number when they are allocated.
NeilBrown69335ef2011-12-23 10:17:54 +1100106 * We sometimes need an extra bio to write to the replacement.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 */
108 struct {
NeilBrown69335ef2011-12-23 10:17:54 +1100109 struct bio *bio;
110 union {
111 struct bio *repl_bio; /* used for resync and
112 * writes */
113 struct md_rdev *rdev; /* used for reads
114 * (read_slot >= 0) */
115 };
116 sector_t addr;
117 int devnum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 } devs[0];
119};
120
NeilBrown0eb3ff12006-01-06 00:20:29 -0800121/* when we get a read error on a read-only array, we redirect to another
122 * device without failing the first device, or trying to over-write to
123 * correct the read error. To keep track of bad blocks on a per-bio
124 * level, we store IO_BLOCKED in the appropriate 'bios' pointer
125 */
126#define IO_BLOCKED ((struct bio*)1)
NeilBrown749c55e2011-07-28 11:39:24 +1000127/* When we successfully write to a known bad-block, we need to remove the
128 * bad-block marking which must be done from process context. So we record
129 * the success by setting devs[n].bio to IO_MADE_GOOD
130 */
131#define IO_MADE_GOOD ((struct bio *)2)
132
133#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
NeilBrown0eb3ff12006-01-06 00:20:29 -0800134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135/* bits for r10bio.state */
NeilBrown69335ef2011-12-23 10:17:54 +1100136enum r10bio_state {
137 R10BIO_Uptodate,
138 R10BIO_IsSync,
139 R10BIO_IsRecover,
140 R10BIO_Degraded,
NeilBrown856e08e2011-07-28 11:39:23 +1000141/* Set ReadError on bios that experience a read error
142 * so that raid10d knows what to do with them.
143 */
NeilBrown69335ef2011-12-23 10:17:54 +1100144 R10BIO_ReadError,
NeilBrown749c55e2011-07-28 11:39:24 +1000145/* If a write for this request means we can clear some
146 * known-bad-block records, we set this flag.
147 */
NeilBrown69335ef2011-12-23 10:17:54 +1100148 R10BIO_MadeGood,
149 R10BIO_WriteError,
150};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151#endif