Loading...
1#ifndef _RAID1_H
2#define _RAID1_H
3
4struct raid1_info {
5 struct md_rdev *rdev;
6 sector_t head_position;
7
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;
12 sector_t seq_start;
13};
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.
21 * 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.
24 */
25
26struct pool_info {
27 struct mddev *mddev;
28 int raid_disks;
29};
30
31struct r1conf {
32 struct mddev *mddev;
33 struct raid1_info *mirrors; /* twice 'raid_disks' to
34 * allow for replacements.
35 */
36 int raid_disks;
37
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
44 /* 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
57 spinlock_t device_lock;
58
59 /* 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.
62 */
63 struct list_head retry_list;
64 /* A separate list of r1bio which just need raid_end_bio_io called.
65 * This mustn't happen for writes which had any errors if the superblock
66 * needs to be written.
67 */
68 struct list_head bio_end_io_list;
69
70 /* queue pending writes to be submitted on unplug */
71 struct bio_list pending_bio_list;
72 int pending_count;
73
74 /* for use when syncing mirrors:
75 * We don't allow both normal IO and resync/recovery IO at
76 * the same time - resync/recovery can only happen when there
77 * is no other IO. So when either is active, the other has to wait.
78 * See more details description in raid1.c near raise_barrier().
79 */
80 wait_queue_head_t wait_barrier;
81 spinlock_t resync_lock;
82 int nr_pending;
83 int nr_waiting;
84 int nr_queued;
85 int barrier;
86 int array_frozen;
87
88 /* Set to 1 if a full sync is needed, (fresh device added).
89 * Cleared when a sync completes.
90 */
91 int fullsync;
92
93 /* When the same as mddev->recovery_disabled we don't allow
94 * recovery to be attempted as we expect a read error.
95 */
96 int recovery_disabled;
97
98 /* poolinfo contains information about the content of the
99 * mempools - it changes when the array grows or shrinks
100 */
101 struct pool_info *poolinfo;
102 mempool_t *r1bio_pool;
103 mempool_t *r1buf_pool;
104
105 /* temporary buffer to synchronous IO when attempting to repair
106 * a read error.
107 */
108 struct page *tmppage;
109
110 /* When taking over an array from a different personality, we store
111 * the new thread here until we fully activate the array.
112 */
113 struct md_thread *thread;
114
115 /* Keep track of cluster resync window to send to other
116 * nodes.
117 */
118 sector_t cluster_sync_low;
119 sector_t cluster_sync_high;
120
121};
122
123/*
124 * this is our 'private' RAID1 bio.
125 *
126 * it contains information about what kind of IO operations were started
127 * for this RAID1 operation, and about their status:
128 */
129
130struct r1bio {
131 atomic_t remaining; /* 'have we finished' count,
132 * used from IRQ handlers
133 */
134 atomic_t behind_remaining; /* number of write-behind ios remaining
135 * in this BehindIO request
136 */
137 sector_t sector;
138 sector_t start_next_window;
139 int sectors;
140 unsigned long state;
141 struct mddev *mddev;
142 /*
143 * original bio going to /dev/mdx
144 */
145 struct bio *master_bio;
146 /*
147 * if the IO is in READ direction, then this is where we read
148 */
149 int read_disk;
150
151 struct list_head retry_list;
152 /* Next two are only valid when R1BIO_BehindIO is set */
153 struct bio_vec *behind_bvecs;
154 int behind_page_count;
155 /*
156 * if the IO is in WRITE direction, then multiple bios are used.
157 * We choose the number when they are allocated.
158 */
159 struct bio *bios[0];
160 /* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
161};
162
163/* bits for r1bio.state */
164enum r1bio_state {
165 R1BIO_Uptodate,
166 R1BIO_IsSync,
167 R1BIO_Degraded,
168 R1BIO_BehindIO,
169/* Set ReadError on bios that experience a readerror so that
170 * raid1d knows what to do with them.
171 */
172 R1BIO_ReadError,
173/* For write-behind requests, we call bi_end_io when
174 * the last non-write-behind device completes, providing
175 * any write was successful. Otherwise we call when
176 * any write-behind write succeeds, otherwise we call
177 * with failure when last write completes (and all failed).
178 * Record that bi_end_io was called with this flag...
179 */
180 R1BIO_Returned,
181/* If a write for this request means we can clear some
182 * known-bad-block records, we set this flag
183 */
184 R1BIO_MadeGood,
185 R1BIO_WriteError,
186 R1BIO_FailFast,
187};
188#endif
1#ifndef _RAID1_H
2#define _RAID1_H
3
4struct mirror_info {
5 struct md_rdev *rdev;
6 sector_t head_position;
7};
8
9/*
10 * memory pools need a pointer to the mddev, so they can force an unplug
11 * when memory is tight, and a count of the number of drives that the
12 * pool was allocated for, so they know how much to allocate and free.
13 * mddev->raid_disks cannot be used, as it can change while a pool is active
14 * These two datums are stored in a kmalloced struct.
15 * The 'raid_disks' here is twice the raid_disks in r1conf.
16 * This allows space for each 'real' device can have a replacement in the
17 * second half of the array.
18 */
19
20struct pool_info {
21 struct mddev *mddev;
22 int raid_disks;
23};
24
25struct r1conf {
26 struct mddev *mddev;
27 struct mirror_info *mirrors; /* twice 'raid_disks' to
28 * allow for replacements.
29 */
30 int raid_disks;
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 */
36 int last_used;
37 sector_t next_seq_sect;
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
44 spinlock_t device_lock;
45
46 /* list of 'struct r1bio' that need to be processed by raid1d,
47 * whether to retry a read, writeout a resync or recovery
48 * block, or anything else.
49 */
50 struct list_head retry_list;
51
52 /* queue pending writes to be submitted on unplug */
53 struct bio_list pending_bio_list;
54 int pending_count;
55
56 /* for use when syncing mirrors:
57 * We don't allow both normal IO and resync/recovery IO at
58 * the same time - resync/recovery can only happen when there
59 * is no other IO. So when either is active, the other has to wait.
60 * See more details description in raid1.c near raise_barrier().
61 */
62 wait_queue_head_t wait_barrier;
63 spinlock_t resync_lock;
64 int nr_pending;
65 int nr_waiting;
66 int nr_queued;
67 int barrier;
68
69 /* Set to 1 if a full sync is needed, (fresh device added).
70 * Cleared when a sync completes.
71 */
72 int fullsync;
73
74 /* When the same as mddev->recovery_disabled we don't allow
75 * recovery to be attempted as we expect a read error.
76 */
77 int recovery_disabled;
78
79
80 /* poolinfo contains information about the content of the
81 * mempools - it changes when the array grows or shrinks
82 */
83 struct pool_info *poolinfo;
84 mempool_t *r1bio_pool;
85 mempool_t *r1buf_pool;
86
87 /* temporary buffer to synchronous IO when attempting to repair
88 * a read error.
89 */
90 struct page *tmppage;
91
92
93 /* When taking over an array from a different personality, we store
94 * the new thread here until we fully activate the array.
95 */
96 struct md_thread *thread;
97};
98
99/*
100 * this is our 'private' RAID1 bio.
101 *
102 * it contains information about what kind of IO operations were started
103 * for this RAID1 operation, and about their status:
104 */
105
106struct r1bio {
107 atomic_t remaining; /* 'have we finished' count,
108 * used from IRQ handlers
109 */
110 atomic_t behind_remaining; /* number of write-behind ios remaining
111 * in this BehindIO request
112 */
113 sector_t sector;
114 int sectors;
115 unsigned long state;
116 struct mddev *mddev;
117 /*
118 * original bio going to /dev/mdx
119 */
120 struct bio *master_bio;
121 /*
122 * if the IO is in READ direction, then this is where we read
123 */
124 int read_disk;
125
126 struct list_head retry_list;
127 /* Next two are only valid when R1BIO_BehindIO is set */
128 struct bio_vec *behind_bvecs;
129 int behind_page_count;
130 /*
131 * if the IO is in WRITE direction, then multiple bios are used.
132 * We choose the number when they are allocated.
133 */
134 struct bio *bios[0];
135 /* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
136};
137
138/* when we get a read error on a read-only array, we redirect to another
139 * device without failing the first device, or trying to over-write to
140 * correct the read error. To keep track of bad blocks on a per-bio
141 * level, we store IO_BLOCKED in the appropriate 'bios' pointer
142 */
143#define IO_BLOCKED ((struct bio *)1)
144/* When we successfully write to a known bad-block, we need to remove the
145 * bad-block marking which must be done from process context. So we record
146 * the success by setting bios[n] to IO_MADE_GOOD
147 */
148#define IO_MADE_GOOD ((struct bio *)2)
149
150#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
151
152/* bits for r1bio.state */
153#define R1BIO_Uptodate 0
154#define R1BIO_IsSync 1
155#define R1BIO_Degraded 2
156#define R1BIO_BehindIO 3
157/* Set ReadError on bios that experience a readerror so that
158 * raid1d knows what to do with them.
159 */
160#define R1BIO_ReadError 4
161/* For write-behind requests, we call bi_end_io when
162 * the last non-write-behind device completes, providing
163 * any write was successful. Otherwise we call when
164 * any write-behind write succeeds, otherwise we call
165 * with failure when last write completes (and all failed).
166 * Record that bi_end_io was called with this flag...
167 */
168#define R1BIO_Returned 6
169/* If a write for this request means we can clear some
170 * known-bad-block records, we set this flag
171 */
172#define R1BIO_MadeGood 7
173#define R1BIO_WriteError 8
174
175extern int md_raid1_congested(struct mddev *mddev, int bits);
176
177#endif