blob: 14ebb288c1ef9446fcde8596ed5191ab50431cc3 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _RAID1_H
2#define _RAID1_H
3
Jonathan Brassow0eaf8222012-07-31 10:03:52 +10004struct raid1_info {
NeilBrown3cb03002011-10-11 16:45:26 +11005 struct md_rdev *rdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 sector_t head_position;
Shaohua Libe4d3282012-07-31 10:03:53 +10007
8 /* When choose the best device for a read (read_balance())
9 * we try to keep sequential reads one the same device
10 */
11 sector_t next_seq_sect;
Shaohua Li12cee5a2012-07-31 10:03:53 +100012 sector_t seq_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -070013};
14
15/*
16 * memory pools need a pointer to the mddev, so they can force an unplug
17 * when memory is tight, and a count of the number of drives that the
18 * pool was allocated for, so they know how much to allocate and free.
19 * mddev->raid_disks cannot be used, as it can change while a pool is active
20 * These two datums are stored in a kmalloced struct.
NeilBrown8f19ccb2011-12-23 10:17:56 +110021 * The 'raid_disks' here is twice the raid_disks in r1conf.
22 * This allows space for each 'real' device can have a replacement in the
23 * second half of the array.
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 */
25
26struct pool_info {
NeilBrownfd01b882011-10-11 16:47:53 +110027 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070028 int raid_disks;
29};
30
NeilBrowne8096362011-10-11 16:49:05 +110031struct r1conf {
NeilBrownfd01b882011-10-11 16:47:53 +110032 struct mddev *mddev;
Jonathan Brassow0eaf8222012-07-31 10:03:52 +100033 struct raid1_info *mirrors; /* twice 'raid_disks' to
NeilBrown8f19ccb2011-12-23 10:17:56 +110034 * allow for replacements.
35 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 int raid_disks;
NeilBrownce550c22011-10-07 14:22:33 +110037
NeilBrownce550c22011-10-07 14:22:33 +110038 /* During resync, read_balancing is only allowed on the part
39 * of the array that has been resynced. 'next_resync' tells us
40 * where that is.
41 */
42 sector_t next_resync;
43
majianpeng79ef3a82013-11-15 14:55:02 +080044 /* When raid1 starts resync, we divide array into four partitions
45 * |---------|--------------|---------------------|-------------|
46 * next_resync start_next_window end_window
47 * start_next_window = next_resync + NEXT_NORMALIO_DISTANCE
48 * end_window = start_next_window + NEXT_NORMALIO_DISTANCE
49 * current_window_requests means the count of normalIO between
50 * start_next_window and end_window.
51 * next_window_requests means the count of normalIO after end_window.
52 * */
53 sector_t start_next_window;
54 int current_window_requests;
55 int next_window_requests;
56
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 spinlock_t device_lock;
58
NeilBrown9f2c9d12011-10-11 16:48:43 +110059 /* list of 'struct r1bio' that need to be processed by raid1d,
60 * whether to retry a read, writeout a resync or recovery
61 * block, or anything else.
NeilBrownce550c22011-10-07 14:22:33 +110062 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 struct list_head retry_list;
NeilBrownce550c22011-10-07 14:22:33 +110064
65 /* queue pending writes to be submitted on unplug */
NeilBrown191ea9b2005-06-21 17:17:23 -070066 struct bio_list pending_bio_list;
NeilBrown34db0cd2011-10-11 16:50:01 +110067 int pending_count;
NeilBrown191ea9b2005-06-21 17:17:23 -070068
NeilBrownce550c22011-10-07 14:22:33 +110069 /* for use when syncing mirrors:
70 * We don't allow both normal IO and resync/recovery IO at
71 * the same time - resync/recovery can only happen when there
72 * is no other IO. So when either is active, the other has to wait.
73 * See more details description in raid1.c near raise_barrier().
74 */
75 wait_queue_head_t wait_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 spinlock_t resync_lock;
NeilBrown191ea9b2005-06-21 17:17:23 -070077 int nr_pending;
NeilBrown17999be2006-01-06 00:20:12 -080078 int nr_waiting;
NeilBrownddaf22a2006-01-06 00:20:19 -080079 int nr_queued;
NeilBrown191ea9b2005-06-21 17:17:23 -070080 int barrier;
majianpengb364e3d2013-11-14 15:16:18 +110081 int array_frozen;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
NeilBrownce550c22011-10-07 14:22:33 +110083 /* Set to 1 if a full sync is needed, (fresh device added).
84 * Cleared when a sync completes.
85 */
86 int fullsync;
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
NeilBrownce550c22011-10-07 14:22:33 +110088 /* When the same as mddev->recovery_disabled we don't allow
89 * recovery to be attempted as we expect a read error.
90 */
91 int recovery_disabled;
92
NeilBrownce550c22011-10-07 14:22:33 +110093 /* poolinfo contains information about the content of the
94 * mempools - it changes when the array grows or shrinks
95 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 struct pool_info *poolinfo;
NeilBrown9f2c9d12011-10-11 16:48:43 +110097 mempool_t *r1bio_pool;
98 mempool_t *r1buf_pool;
NeilBrown709ae482009-12-14 12:49:51 +110099
NeilBrownce550c22011-10-07 14:22:33 +1100100 /* temporary buffer to synchronous IO when attempting to repair
101 * a read error.
102 */
103 struct page *tmppage;
104
NeilBrown709ae482009-12-14 12:49:51 +1100105 /* When taking over an array from a different personality, we store
106 * the new thread here until we fully activate the array.
107 */
NeilBrown2b8bf342011-10-11 16:48:23 +1100108 struct md_thread *thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109};
110
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 * this is our 'private' RAID1 bio.
113 *
114 * it contains information about what kind of IO operations were started
115 * for this RAID1 operation, and about their status:
116 */
117
NeilBrown9f2c9d12011-10-11 16:48:43 +1100118struct r1bio {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 atomic_t remaining; /* 'have we finished' count,
120 * used from IRQ handlers
121 */
NeilBrown4b6d2872005-09-09 16:23:47 -0700122 atomic_t behind_remaining; /* number of write-behind ios remaining
123 * in this BehindIO request
124 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 sector_t sector;
majianpeng79ef3a82013-11-15 14:55:02 +0800126 sector_t start_next_window;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 int sectors;
128 unsigned long state;
NeilBrownfd01b882011-10-11 16:47:53 +1100129 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 /*
131 * original bio going to /dev/mdx
132 */
133 struct bio *master_bio;
134 /*
135 * if the IO is in READ direction, then this is where we read
136 */
137 int read_disk;
138
139 struct list_head retry_list;
NeilBrownaf6d7b72011-05-11 14:51:19 +1000140 /* Next two are only valid when R1BIO_BehindIO is set */
NeilBrown2ca68f52011-07-28 11:32:10 +1000141 struct bio_vec *behind_bvecs;
NeilBrownaf6d7b72011-05-11 14:51:19 +1000142 int behind_page_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 /*
144 * if the IO is in WRITE direction, then multiple bios are used.
145 * We choose the number when they are allocated.
146 */
147 struct bio *bios[0];
NeilBrown191ea9b2005-06-21 17:17:23 -0700148 /* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149};
150
151/* bits for r1bio.state */
152#define R1BIO_Uptodate 0
153#define R1BIO_IsSync 1
NeilBrown191ea9b2005-06-21 17:17:23 -0700154#define R1BIO_Degraded 2
NeilBrowna9701a32005-11-08 21:39:34 -0800155#define R1BIO_BehindIO 3
NeilBrownd2eb35a2011-07-28 11:31:48 +1000156/* Set ReadError on bios that experience a readerror so that
157 * raid1d knows what to do with them.
158 */
159#define R1BIO_ReadError 4
NeilBrown4b6d2872005-09-09 16:23:47 -0700160/* For write-behind requests, we call bi_end_io when
161 * the last non-write-behind device completes, providing
162 * any write was successful. Otherwise we call when
163 * any write-behind write succeeds, otherwise we call
164 * with failure when last write completes (and all failed).
165 * Record that bi_end_io was called with this flag...
166 */
NeilBrown9e71f9c2006-03-23 02:59:22 -0800167#define R1BIO_Returned 6
NeilBrown4367af52011-07-28 11:31:49 +1000168/* If a write for this request means we can clear some
169 * known-bad-block records, we set this flag
170 */
NeilBrowncd5ff9a12011-07-28 11:32:41 +1000171#define R1BIO_MadeGood 7
172#define R1BIO_WriteError 8
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173#endif