Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _RAID1_H |
| 2 | #define _RAID1_H |
| 3 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | typedef struct mirror_info mirror_info_t; |
| 5 | |
| 6 | struct mirror_info { |
| 7 | mdk_rdev_t *rdev; |
| 8 | sector_t head_position; |
| 9 | }; |
| 10 | |
| 11 | /* |
| 12 | * memory pools need a pointer to the mddev, so they can force an unplug |
| 13 | * when memory is tight, and a count of the number of drives that the |
| 14 | * pool was allocated for, so they know how much to allocate and free. |
| 15 | * mddev->raid_disks cannot be used, as it can change while a pool is active |
| 16 | * These two datums are stored in a kmalloced struct. |
| 17 | */ |
| 18 | |
| 19 | struct pool_info { |
| 20 | mddev_t *mddev; |
| 21 | int raid_disks; |
| 22 | }; |
| 23 | |
| 24 | |
| 25 | typedef struct r1bio_s r1bio_t; |
| 26 | |
| 27 | struct r1_private_data_s { |
| 28 | mddev_t *mddev; |
| 29 | mirror_info_t *mirrors; |
| 30 | int raid_disks; |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 31 | |
| 32 | /* When choose the best device for a read (read_balance()) |
| 33 | * we try to keep sequential reads one the same device |
| 34 | * using 'last_used' and 'next_seq_sect' |
| 35 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 36 | int last_used; |
| 37 | sector_t next_seq_sect; |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 38 | /* 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 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | spinlock_t device_lock; |
| 45 | |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 46 | /* list of 'r1bio_t' that need to be processed by raid1d, whether |
| 47 | * to retry a read, writeout a resync or recovery block, or |
| 48 | * anything else. |
| 49 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 50 | struct list_head retry_list; |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 51 | |
| 52 | /* queue pending writes to be submitted on unplug */ |
NeilBrown | 191ea9b | 2005-06-21 17:17:23 -0700 | [diff] [blame] | 53 | struct bio_list pending_bio_list; |
NeilBrown | 191ea9b | 2005-06-21 17:17:23 -0700 | [diff] [blame] | 54 | |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 55 | /* for use when syncing mirrors: |
| 56 | * We don't allow both normal IO and resync/recovery IO at |
| 57 | * the same time - resync/recovery can only happen when there |
| 58 | * is no other IO. So when either is active, the other has to wait. |
| 59 | * See more details description in raid1.c near raise_barrier(). |
| 60 | */ |
| 61 | wait_queue_head_t wait_barrier; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 62 | spinlock_t resync_lock; |
NeilBrown | 191ea9b | 2005-06-21 17:17:23 -0700 | [diff] [blame] | 63 | int nr_pending; |
NeilBrown | 17999be | 2006-01-06 00:20:12 -0800 | [diff] [blame] | 64 | int nr_waiting; |
NeilBrown | ddaf22a | 2006-01-06 00:20:19 -0800 | [diff] [blame] | 65 | int nr_queued; |
NeilBrown | 191ea9b | 2005-06-21 17:17:23 -0700 | [diff] [blame] | 66 | int barrier; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 68 | /* Set to 1 if a full sync is needed, (fresh device added). |
| 69 | * Cleared when a sync completes. |
| 70 | */ |
| 71 | int fullsync; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 73 | /* When the same as mddev->recovery_disabled we don't allow |
| 74 | * recovery to be attempted as we expect a read error. |
| 75 | */ |
| 76 | int recovery_disabled; |
| 77 | |
| 78 | |
| 79 | /* poolinfo contains information about the content of the |
| 80 | * mempools - it changes when the array grows or shrinks |
| 81 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 82 | struct pool_info *poolinfo; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 83 | mempool_t *r1bio_pool; |
| 84 | mempool_t *r1buf_pool; |
NeilBrown | 709ae48 | 2009-12-14 12:49:51 +1100 | [diff] [blame] | 85 | |
NeilBrown | ce550c2 | 2011-10-07 14:22:33 +1100 | [diff] [blame] | 86 | /* temporary buffer to synchronous IO when attempting to repair |
| 87 | * a read error. |
| 88 | */ |
| 89 | struct page *tmppage; |
| 90 | |
| 91 | |
NeilBrown | 709ae48 | 2009-12-14 12:49:51 +1100 | [diff] [blame] | 92 | /* When taking over an array from a different personality, we store |
| 93 | * the new thread here until we fully activate the array. |
| 94 | */ |
| 95 | struct mdk_thread_s *thread; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 96 | }; |
| 97 | |
| 98 | typedef struct r1_private_data_s conf_t; |
| 99 | |
| 100 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 101 | * this is our 'private' RAID1 bio. |
| 102 | * |
| 103 | * it contains information about what kind of IO operations were started |
| 104 | * for this RAID1 operation, and about their status: |
| 105 | */ |
| 106 | |
| 107 | struct r1bio_s { |
| 108 | atomic_t remaining; /* 'have we finished' count, |
| 109 | * used from IRQ handlers |
| 110 | */ |
NeilBrown | 4b6d287 | 2005-09-09 16:23:47 -0700 | [diff] [blame] | 111 | atomic_t behind_remaining; /* number of write-behind ios remaining |
| 112 | * in this BehindIO request |
| 113 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 114 | sector_t sector; |
| 115 | int sectors; |
| 116 | unsigned long state; |
| 117 | mddev_t *mddev; |
| 118 | /* |
| 119 | * original bio going to /dev/mdx |
| 120 | */ |
| 121 | struct bio *master_bio; |
| 122 | /* |
| 123 | * if the IO is in READ direction, then this is where we read |
| 124 | */ |
| 125 | int read_disk; |
| 126 | |
| 127 | struct list_head retry_list; |
NeilBrown | af6d7b7 | 2011-05-11 14:51:19 +1000 | [diff] [blame] | 128 | /* Next two are only valid when R1BIO_BehindIO is set */ |
NeilBrown | 2ca68f5 | 2011-07-28 11:32:10 +1000 | [diff] [blame] | 129 | struct bio_vec *behind_bvecs; |
NeilBrown | af6d7b7 | 2011-05-11 14:51:19 +1000 | [diff] [blame] | 130 | int behind_page_count; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 131 | /* |
| 132 | * if the IO is in WRITE direction, then multiple bios are used. |
| 133 | * We choose the number when they are allocated. |
| 134 | */ |
| 135 | struct bio *bios[0]; |
NeilBrown | 191ea9b | 2005-06-21 17:17:23 -0700 | [diff] [blame] | 136 | /* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 137 | }; |
| 138 | |
NeilBrown | cf30a47 | 2006-01-06 00:20:23 -0800 | [diff] [blame] | 139 | /* when we get a read error on a read-only array, we redirect to another |
| 140 | * device without failing the first device, or trying to over-write to |
| 141 | * correct the read error. To keep track of bad blocks on a per-bio |
| 142 | * level, we store IO_BLOCKED in the appropriate 'bios' pointer |
| 143 | */ |
NeilBrown | 4367af5 | 2011-07-28 11:31:49 +1000 | [diff] [blame] | 144 | #define IO_BLOCKED ((struct bio *)1) |
| 145 | /* When we successfully write to a known bad-block, we need to remove the |
| 146 | * bad-block marking which must be done from process context. So we record |
| 147 | * the success by setting bios[n] to IO_MADE_GOOD |
| 148 | */ |
| 149 | #define IO_MADE_GOOD ((struct bio *)2) |
| 150 | |
| 151 | #define BIO_SPECIAL(bio) ((unsigned long)bio <= 2) |
NeilBrown | cf30a47 | 2006-01-06 00:20:23 -0800 | [diff] [blame] | 152 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 153 | /* bits for r1bio.state */ |
| 154 | #define R1BIO_Uptodate 0 |
| 155 | #define R1BIO_IsSync 1 |
NeilBrown | 191ea9b | 2005-06-21 17:17:23 -0700 | [diff] [blame] | 156 | #define R1BIO_Degraded 2 |
NeilBrown | a9701a3 | 2005-11-08 21:39:34 -0800 | [diff] [blame] | 157 | #define R1BIO_BehindIO 3 |
NeilBrown | d2eb35a | 2011-07-28 11:31:48 +1000 | [diff] [blame] | 158 | /* Set ReadError on bios that experience a readerror so that |
| 159 | * raid1d knows what to do with them. |
| 160 | */ |
| 161 | #define R1BIO_ReadError 4 |
NeilBrown | 4b6d287 | 2005-09-09 16:23:47 -0700 | [diff] [blame] | 162 | /* For write-behind requests, we call bi_end_io when |
| 163 | * the last non-write-behind device completes, providing |
| 164 | * any write was successful. Otherwise we call when |
| 165 | * any write-behind write succeeds, otherwise we call |
| 166 | * with failure when last write completes (and all failed). |
| 167 | * Record that bi_end_io was called with this flag... |
| 168 | */ |
NeilBrown | 9e71f9c | 2006-03-23 02:59:22 -0800 | [diff] [blame] | 169 | #define R1BIO_Returned 6 |
NeilBrown | 4367af5 | 2011-07-28 11:31:49 +1000 | [diff] [blame] | 170 | /* If a write for this request means we can clear some |
| 171 | * known-bad-block records, we set this flag |
| 172 | */ |
NeilBrown | cd5ff9a | 2011-07-28 11:32:41 +1000 | [diff] [blame] | 173 | #define R1BIO_MadeGood 7 |
| 174 | #define R1BIO_WriteError 8 |
NeilBrown | 4b6d287 | 2005-09-09 16:23:47 -0700 | [diff] [blame] | 175 | |
Jonathan Brassow | 1ed7242 | 2011-06-07 17:50:35 -0500 | [diff] [blame] | 176 | extern int md_raid1_congested(mddev_t *mddev, int bits); |
| 177 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 178 | #endif |