blob: 37509d7134aaeb9409c8fbaf8f16257eef0b9cf2 [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;
NeilBrownf8c9e742012-05-21 09:28:33 +100037 } prev, geo;
NeilBrown5cf00fc2012-05-21 09:28:20 +100038 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 */
NeilBrownf8c9e742012-05-21 09:28:33 +100044 sector_t reshape_progress;
Trela, Maciejdab8b292010-03-08 16:02:45 +110045
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 struct list_head retry_list;
NeilBrown6cce3b22006-01-06 00:20:16 -080047 /* queue pending writes and submit them on unplug */
48 struct bio_list pending_bio_list;
NeilBrown34db0cd2011-10-11 16:50:01 +110049 int pending_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
51 spinlock_t resync_lock;
NeilBrown69335ef2011-12-23 10:17:54 +110052 int nr_pending;
53 int nr_waiting;
54 int nr_queued;
55 int barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070056 sector_t next_resync;
NeilBrown6cce3b22006-01-06 00:20:16 -080057 int fullsync; /* set to 1 if a full sync is needed,
58 * (fresh device added).
59 * Cleared when a sync completes.
60 */
NeilBrown69335ef2011-12-23 10:17:54 +110061 int have_replacement; /* There is at least one
62 * replacement device.
63 */
NeilBrown0a27ec92006-01-06 00:20:13 -080064 wait_queue_head_t wait_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
NeilBrown69335ef2011-12-23 10:17:54 +110066 mempool_t *r10bio_pool;
67 mempool_t *r10buf_pool;
NeilBrown4443ae12006-01-06 00:20:28 -080068 struct page *tmppage;
Trela, Maciejdab8b292010-03-08 16:02:45 +110069
70 /* When taking over an array from a different personality, we store
71 * the new thread here until we fully activate the array.
72 */
NeilBrown2b8bf342011-10-11 16:48:23 +110073 struct md_thread *thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -070074};
75
Linus Torvalds1da177e2005-04-16 15:20:36 -070076/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 * this is our 'private' RAID10 bio.
78 *
79 * it contains information about what kind of IO operations were started
80 * for this RAID10 operation, and about their status:
81 */
82
NeilBrown9f2c9d12011-10-11 16:48:43 +110083struct r10bio {
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 atomic_t remaining; /* 'have we finished' count,
85 * used from IRQ handlers
86 */
87 sector_t sector; /* virtual sector number */
88 int sectors;
89 unsigned long state;
NeilBrownfd01b882011-10-11 16:47:53 +110090 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 /*
92 * original bio going to /dev/mdx
93 */
94 struct bio *master_bio;
95 /*
96 * if the IO is in READ direction, then this is where we read
97 */
98 int read_slot;
99
100 struct list_head retry_list;
101 /*
102 * if the IO is in WRITE direction, then multiple bios are used,
103 * one for each copy.
104 * When resyncing we also use one for each copy.
105 * When reconstructing, we use 2 bios, one for read, one for write.
106 * We choose the number when they are allocated.
NeilBrown69335ef2011-12-23 10:17:54 +1100107 * We sometimes need an extra bio to write to the replacement.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 */
109 struct {
NeilBrown69335ef2011-12-23 10:17:54 +1100110 struct bio *bio;
111 union {
112 struct bio *repl_bio; /* used for resync and
113 * writes */
114 struct md_rdev *rdev; /* used for reads
115 * (read_slot >= 0) */
116 };
117 sector_t addr;
118 int devnum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 } devs[0];
120};
121
NeilBrown0eb3ff12006-01-06 00:20:29 -0800122/* when we get a read error on a read-only array, we redirect to another
123 * device without failing the first device, or trying to over-write to
124 * correct the read error. To keep track of bad blocks on a per-bio
125 * level, we store IO_BLOCKED in the appropriate 'bios' pointer
126 */
127#define IO_BLOCKED ((struct bio*)1)
NeilBrown749c55e2011-07-28 11:39:24 +1000128/* When we successfully write to a known bad-block, we need to remove the
129 * bad-block marking which must be done from process context. So we record
130 * the success by setting devs[n].bio to IO_MADE_GOOD
131 */
132#define IO_MADE_GOOD ((struct bio *)2)
133
134#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
NeilBrown0eb3ff12006-01-06 00:20:29 -0800135
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136/* bits for r10bio.state */
NeilBrown69335ef2011-12-23 10:17:54 +1100137enum r10bio_state {
138 R10BIO_Uptodate,
139 R10BIO_IsSync,
140 R10BIO_IsRecover,
141 R10BIO_Degraded,
NeilBrown856e08e2011-07-28 11:39:23 +1000142/* Set ReadError on bios that experience a read error
143 * so that raid10d knows what to do with them.
144 */
NeilBrown69335ef2011-12-23 10:17:54 +1100145 R10BIO_ReadError,
NeilBrown749c55e2011-07-28 11:39:24 +1000146/* If a write for this request means we can clear some
147 * known-bad-block records, we set this flag.
148 */
NeilBrown69335ef2011-12-23 10:17:54 +1100149 R10BIO_MadeGood,
150 R10BIO_WriteError,
NeilBrownf8c9e742012-05-21 09:28:33 +1000151/* During a reshape we might be performing IO on the
152 * 'previous' part of the array, in which case this
153 * flag is set
154 */
155 R10BIO_Previous,
NeilBrown69335ef2011-12-23 10:17:54 +1100156};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157#endif