blob: 157d69e83ff401972f395db9bb281faa8df75a65 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _RAID10_H
2#define _RAID10_H
3
Jonathan Brassowdc280d982012-07-31 10:03:52 +10004struct raid10_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;
Jonathan Brassowdc280d982012-07-31 10:03:52 +100016 struct raid10_info *mirrors;
17 struct raid10_info *mirrors_new, *mirrors_old;
Linus Torvalds1da177e2005-04-16 15:20:36 -070018 spinlock_t device_lock;
19
20 /* geometry */
NeilBrown5cf00fc2012-05-21 09:28:20 +100021 struct geom {
22 int raid_disks;
23 int near_copies; /* number of copies laid out
NeilBrown69335ef2011-12-23 10:17:54 +110024 * raid0 style */
NeilBrown5cf00fc2012-05-21 09:28:20 +100025 int far_copies; /* number of copies laid out
Linus Torvalds1da177e2005-04-16 15:20:36 -070026 * at large strides across drives
27 */
NeilBrown5cf00fc2012-05-21 09:28:20 +100028 int far_offset; /* far_copies are offset by 1
NeilBrown69335ef2011-12-23 10:17:54 +110029 * stripe instead of many
NeilBrownc93983b2006-06-26 00:27:41 -070030 */
NeilBrown5cf00fc2012-05-21 09:28:20 +100031 sector_t stride; /* distance between far copies.
NeilBrownc93983b2006-06-26 00:27:41 -070032 * This is size / far_copies unless
33 * far_offset, in which case it is
34 * 1 stripe.
Linus Torvalds1da177e2005-04-16 15:20:36 -070035 */
Jonathan Brassow475901a2013-02-21 13:28:10 +110036 int far_set_size; /* The number of devices in a set,
37 * where a 'set' are devices that
38 * contain far/offset copies of
39 * each other.
40 */
NeilBrown5cf00fc2012-05-21 09:28:20 +100041 int chunk_shift; /* shift from chunks to sectors */
42 sector_t chunk_mask;
NeilBrownf8c9e742012-05-21 09:28:33 +100043 } prev, geo;
NeilBrown5cf00fc2012-05-21 09:28:20 +100044 int copies; /* near_copies * far_copies.
45 * must be <= raid_disks
46 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
NeilBrown69335ef2011-12-23 10:17:54 +110048 sector_t dev_sectors; /* temp copy of
49 * mddev->dev_sectors */
NeilBrownf8c9e742012-05-21 09:28:33 +100050 sector_t reshape_progress;
NeilBrown3ea7daa2012-05-22 13:53:47 +100051 sector_t reshape_safe;
52 unsigned long reshape_checkpoint;
53 sector_t offset_diff;
Trela, Maciejdab8b292010-03-08 16:02:45 +110054
Linus Torvalds1da177e2005-04-16 15:20:36 -070055 struct list_head retry_list;
NeilBrown6cce3b232006-01-06 00:20:16 -080056 /* queue pending writes and submit them on unplug */
57 struct bio_list pending_bio_list;
NeilBrown34db0cd2011-10-11 16:50:01 +110058 int pending_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
60 spinlock_t resync_lock;
NeilBrown69335ef2011-12-23 10:17:54 +110061 int nr_pending;
62 int nr_waiting;
63 int nr_queued;
64 int barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 sector_t next_resync;
NeilBrown6cce3b232006-01-06 00:20:16 -080066 int fullsync; /* set to 1 if a full sync is needed,
67 * (fresh device added).
68 * Cleared when a sync completes.
69 */
NeilBrown69335ef2011-12-23 10:17:54 +110070 int have_replacement; /* There is at least one
71 * replacement device.
72 */
NeilBrown0a27ec92006-01-06 00:20:13 -080073 wait_queue_head_t wait_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
NeilBrown69335ef2011-12-23 10:17:54 +110075 mempool_t *r10bio_pool;
76 mempool_t *r10buf_pool;
NeilBrown4443ae12006-01-06 00:20:28 -080077 struct page *tmppage;
Trela, Maciejdab8b292010-03-08 16:02:45 +110078
79 /* When taking over an array from a different personality, we store
80 * the new thread here until we fully activate the array.
81 */
NeilBrown2b8bf342011-10-11 16:48:23 +110082 struct md_thread *thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -070083};
84
Linus Torvalds1da177e2005-04-16 15:20:36 -070085/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 * this is our 'private' RAID10 bio.
87 *
88 * it contains information about what kind of IO operations were started
89 * for this RAID10 operation, and about their status:
90 */
91
NeilBrown9f2c9d12011-10-11 16:48:43 +110092struct r10bio {
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 atomic_t remaining; /* 'have we finished' count,
94 * used from IRQ handlers
95 */
96 sector_t sector; /* virtual sector number */
97 int sectors;
98 unsigned long state;
NeilBrownfd01b882011-10-11 16:47:53 +110099 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 /*
101 * original bio going to /dev/mdx
102 */
103 struct bio *master_bio;
104 /*
105 * if the IO is in READ direction, then this is where we read
106 */
107 int read_slot;
108
109 struct list_head retry_list;
110 /*
111 * if the IO is in WRITE direction, then multiple bios are used,
112 * one for each copy.
113 * When resyncing we also use one for each copy.
114 * When reconstructing, we use 2 bios, one for read, one for write.
115 * We choose the number when they are allocated.
NeilBrown69335ef2011-12-23 10:17:54 +1100116 * We sometimes need an extra bio to write to the replacement.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 */
NeilBrowne0ee7782012-08-18 09:51:42 +1000118 struct r10dev {
NeilBrown69335ef2011-12-23 10:17:54 +1100119 struct bio *bio;
120 union {
121 struct bio *repl_bio; /* used for resync and
122 * writes */
123 struct md_rdev *rdev; /* used for reads
124 * (read_slot >= 0) */
125 };
126 sector_t addr;
127 int devnum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 } devs[0];
129};
130
131/* bits for r10bio.state */
NeilBrown69335ef2011-12-23 10:17:54 +1100132enum r10bio_state {
133 R10BIO_Uptodate,
134 R10BIO_IsSync,
135 R10BIO_IsRecover,
NeilBrown3ea7daa2012-05-22 13:53:47 +1000136 R10BIO_IsReshape,
NeilBrown69335ef2011-12-23 10:17:54 +1100137 R10BIO_Degraded,
NeilBrown856e08e2011-07-28 11:39:23 +1000138/* Set ReadError on bios that experience a read error
139 * so that raid10d knows what to do with them.
140 */
NeilBrown69335ef2011-12-23 10:17:54 +1100141 R10BIO_ReadError,
NeilBrown749c55e2011-07-28 11:39:24 +1000142/* If a write for this request means we can clear some
143 * known-bad-block records, we set this flag.
144 */
NeilBrown69335ef2011-12-23 10:17:54 +1100145 R10BIO_MadeGood,
146 R10BIO_WriteError,
NeilBrownf8c9e742012-05-21 09:28:33 +1000147/* During a reshape we might be performing IO on the
148 * 'previous' part of the array, in which case this
149 * flag is set
150 */
151 R10BIO_Previous,
NeilBrown69335ef2011-12-23 10:17:54 +1100152};
Jonathan Brassowcc4d1ef2012-07-31 10:03:53 +1000153
154extern int md_raid10_congested(struct mddev *mddev, int bits);
155
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156#endif