Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
2 | #ifndef _RAID10_H |
3 | #define _RAID10_H | |
4 | ||
f2785b52 N |
5 | /* Note: raid10_info.rdev can be set to NULL asynchronously by |
6 | * raid10_remove_disk. | |
7 | * There are three safe ways to access raid10_info.rdev. | |
8 | * 1/ when holding mddev->reconfig_mutex | |
9 | * 2/ when resync/recovery/reshape is known to be happening - i.e. in code | |
10 | * that is called as part of performing resync/recovery/reshape. | |
11 | * 3/ while holding rcu_read_lock(), use rcu_dereference to get the pointer | |
12 | * and if it is non-NULL, increment rdev->nr_pending before dropping the | |
13 | * RCU lock. | |
14 | * When .rdev is set to NULL, the nr_pending count checked again and if it has | |
15 | * been incremented, the pointer is put back in .rdev. | |
16 | */ | |
17 | ||
dc280d98 | 18 | struct raid10_info { |
69335ef3 | 19 | struct md_rdev *rdev, *replacement; |
1da177e4 | 20 | sector_t head_position; |
2bb77736 N |
21 | int recovery_disabled; /* matches |
22 | * mddev->recovery_disabled | |
23 | * when we shouldn't try | |
24 | * recovering this device. | |
25 | */ | |
1da177e4 LT |
26 | }; |
27 | ||
e879a879 | 28 | struct r10conf { |
fd01b88c | 29 | struct mddev *mddev; |
dc280d98 JB |
30 | struct raid10_info *mirrors; |
31 | struct raid10_info *mirrors_new, *mirrors_old; | |
1da177e4 LT |
32 | spinlock_t device_lock; |
33 | ||
34 | /* geometry */ | |
5cf00fcd N |
35 | struct geom { |
36 | int raid_disks; | |
37 | int near_copies; /* number of copies laid out | |
69335ef3 | 38 | * raid0 style */ |
5cf00fcd | 39 | int far_copies; /* number of copies laid out |
1da177e4 LT |
40 | * at large strides across drives |
41 | */ | |
5cf00fcd | 42 | int far_offset; /* far_copies are offset by 1 |
69335ef3 | 43 | * stripe instead of many |
c93983bf | 44 | */ |
5cf00fcd | 45 | sector_t stride; /* distance between far copies. |
c93983bf N |
46 | * This is size / far_copies unless |
47 | * far_offset, in which case it is | |
48 | * 1 stripe. | |
1da177e4 | 49 | */ |
475901af JB |
50 | int far_set_size; /* The number of devices in a set, |
51 | * where a 'set' are devices that | |
52 | * contain far/offset copies of | |
53 | * each other. | |
54 | */ | |
5cf00fcd N |
55 | int chunk_shift; /* shift from chunks to sectors */ |
56 | sector_t chunk_mask; | |
f8c9e74f | 57 | } prev, geo; |
5cf00fcd N |
58 | int copies; /* near_copies * far_copies. |
59 | * must be <= raid_disks | |
60 | */ | |
1da177e4 | 61 | |
69335ef3 N |
62 | sector_t dev_sectors; /* temp copy of |
63 | * mddev->dev_sectors */ | |
f8c9e74f | 64 | sector_t reshape_progress; |
3ea7daa5 N |
65 | sector_t reshape_safe; |
66 | unsigned long reshape_checkpoint; | |
67 | sector_t offset_diff; | |
dab8b292 | 68 | |
1da177e4 | 69 | struct list_head retry_list; |
95af587e N |
70 | /* A separate list of r1bio which just need raid_end_bio_io called. |
71 | * This mustn't happen for writes which had any errors if the superblock | |
72 | * needs to be written. | |
73 | */ | |
74 | struct list_head bio_end_io_list; | |
75 | ||
6cce3b23 N |
76 | /* queue pending writes and submit them on unplug */ |
77 | struct bio_list pending_bio_list; | |
1da177e4 | 78 | |
b9b083f9 | 79 | seqlock_t resync_lock; |
0e5313e2 | 80 | atomic_t nr_pending; |
69335ef3 N |
81 | int nr_waiting; |
82 | int nr_queued; | |
83 | int barrier; | |
0e5313e2 | 84 | int array_freeze_pending; |
1da177e4 | 85 | sector_t next_resync; |
6cce3b23 N |
86 | int fullsync; /* set to 1 if a full sync is needed, |
87 | * (fresh device added). | |
88 | * Cleared when a sync completes. | |
89 | */ | |
69335ef3 N |
90 | int have_replacement; /* There is at least one |
91 | * replacement device. | |
92 | */ | |
0a27ec96 | 93 | wait_queue_head_t wait_barrier; |
1da177e4 | 94 | |
afeee514 KO |
95 | mempool_t r10bio_pool; |
96 | mempool_t r10buf_pool; | |
4443ae10 | 97 | struct page *tmppage; |
afeee514 | 98 | struct bio_set bio_split; |
dab8b292 TM |
99 | |
100 | /* When taking over an array from a different personality, we store | |
101 | * the new thread here until we fully activate the array. | |
102 | */ | |
2b8bf345 | 103 | struct md_thread *thread; |
8db87912 GJ |
104 | |
105 | /* | |
106 | * Keep track of cluster resync window to send to other nodes. | |
107 | */ | |
108 | sector_t cluster_sync_low; | |
109 | sector_t cluster_sync_high; | |
1da177e4 LT |
110 | }; |
111 | ||
1da177e4 LT |
112 | /* |
113 | * this is our 'private' RAID10 bio. | |
114 | * | |
115 | * it contains information about what kind of IO operations were started | |
116 | * for this RAID10 operation, and about their status: | |
117 | */ | |
118 | ||
9f2c9d12 | 119 | struct r10bio { |
1da177e4 LT |
120 | atomic_t remaining; /* 'have we finished' count, |
121 | * used from IRQ handlers | |
122 | */ | |
123 | sector_t sector; /* virtual sector number */ | |
124 | int sectors; | |
125 | unsigned long state; | |
528bc2cf | 126 | unsigned long start_time; |
fd01b88c | 127 | struct mddev *mddev; |
1da177e4 LT |
128 | /* |
129 | * original bio going to /dev/mdx | |
130 | */ | |
131 | struct bio *master_bio; | |
132 | /* | |
133 | * if the IO is in READ direction, then this is where we read | |
134 | */ | |
135 | int read_slot; | |
136 | ||
137 | struct list_head retry_list; | |
138 | /* | |
139 | * if the IO is in WRITE direction, then multiple bios are used, | |
140 | * one for each copy. | |
141 | * When resyncing we also use one for each copy. | |
142 | * When reconstructing, we use 2 bios, one for read, one for write. | |
143 | * We choose the number when they are allocated. | |
69335ef3 | 144 | * We sometimes need an extra bio to write to the replacement. |
1da177e4 | 145 | */ |
e0ee7785 | 146 | struct r10dev { |
69335ef3 N |
147 | struct bio *bio; |
148 | union { | |
149 | struct bio *repl_bio; /* used for resync and | |
150 | * writes */ | |
151 | struct md_rdev *rdev; /* used for reads | |
152 | * (read_slot >= 0) */ | |
153 | }; | |
154 | sector_t addr; | |
155 | int devnum; | |
358369f0 | 156 | } devs[]; |
1da177e4 LT |
157 | }; |
158 | ||
159 | /* bits for r10bio.state */ | |
69335ef3 N |
160 | enum r10bio_state { |
161 | R10BIO_Uptodate, | |
162 | R10BIO_IsSync, | |
163 | R10BIO_IsRecover, | |
3ea7daa5 | 164 | R10BIO_IsReshape, |
69335ef3 | 165 | R10BIO_Degraded, |
856e08e2 N |
166 | /* Set ReadError on bios that experience a read error |
167 | * so that raid10d knows what to do with them. | |
168 | */ | |
69335ef3 | 169 | R10BIO_ReadError, |
749c55e9 N |
170 | /* If a write for this request means we can clear some |
171 | * known-bad-block records, we set this flag. | |
172 | */ | |
69335ef3 N |
173 | R10BIO_MadeGood, |
174 | R10BIO_WriteError, | |
f8c9e74f N |
175 | /* During a reshape we might be performing IO on the |
176 | * 'previous' part of the array, in which case this | |
177 | * flag is set | |
178 | */ | |
179 | R10BIO_Previous, | |
8d3ca83d N |
180 | /* failfast devices did receive failfast requests. */ |
181 | R10BIO_FailFast, | |
254c271d | 182 | R10BIO_Discard, |
69335ef3 | 183 | }; |
1da177e4 | 184 | #endif |