xref: /linux/drivers/md/raid1.h (revision f2ee442115c9b6219083c019939a9cc0c9abb2f8)
1 #ifndef _RAID1_H
2 #define _RAID1_H
3 
4 struct 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  */
16 
17 struct pool_info {
18 	struct mddev *mddev;
19 	int	raid_disks;
20 };
21 
22 struct r1conf {
23 	struct mddev		*mddev;
24 	struct mirror_info		*mirrors;
25 	int			raid_disks;
26 
27 	/* When choose the best device for a read (read_balance())
28 	 * we try to keep sequential reads one the same device
29 	 * using 'last_used' and 'next_seq_sect'
30 	 */
31 	int			last_used;
32 	sector_t		next_seq_sect;
33 	/* During resync, read_balancing is only allowed on the part
34 	 * of the array that has been resynced.  'next_resync' tells us
35 	 * where that is.
36 	 */
37 	sector_t		next_resync;
38 
39 	spinlock_t		device_lock;
40 
41 	/* list of 'struct r1bio' that need to be processed by raid1d,
42 	 * whether to retry a read, writeout a resync or recovery
43 	 * block, or anything else.
44 	 */
45 	struct list_head	retry_list;
46 
47 	/* queue pending writes to be submitted on unplug */
48 	struct bio_list		pending_bio_list;
49 	int			pending_count;
50 
51 	/* for use when syncing mirrors:
52 	 * We don't allow both normal IO and resync/recovery IO at
53 	 * the same time - resync/recovery can only happen when there
54 	 * is no other IO.  So when either is active, the other has to wait.
55 	 * See more details description in raid1.c near raise_barrier().
56 	 */
57 	wait_queue_head_t	wait_barrier;
58 	spinlock_t		resync_lock;
59 	int			nr_pending;
60 	int			nr_waiting;
61 	int			nr_queued;
62 	int			barrier;
63 
64 	/* Set to 1 if a full sync is needed, (fresh device added).
65 	 * Cleared when a sync completes.
66 	 */
67 	int			fullsync;
68 
69 	/* When the same as mddev->recovery_disabled we don't allow
70 	 * recovery to be attempted as we expect a read error.
71 	 */
72 	int			recovery_disabled;
73 
74 
75 	/* poolinfo contains information about the content of the
76 	 * mempools - it changes when the array grows or shrinks
77 	 */
78 	struct pool_info	*poolinfo;
79 	mempool_t		*r1bio_pool;
80 	mempool_t		*r1buf_pool;
81 
82 	/* temporary buffer to synchronous IO when attempting to repair
83 	 * a read error.
84 	 */
85 	struct page		*tmppage;
86 
87 
88 	/* When taking over an array from a different personality, we store
89 	 * the new thread here until we fully activate the array.
90 	 */
91 	struct md_thread	*thread;
92 };
93 
94 /*
95  * this is our 'private' RAID1 bio.
96  *
97  * it contains information about what kind of IO operations were started
98  * for this RAID1 operation, and about their status:
99  */
100 
101 struct r1bio {
102 	atomic_t		remaining; /* 'have we finished' count,
103 					    * used from IRQ handlers
104 					    */
105 	atomic_t		behind_remaining; /* number of write-behind ios remaining
106 						 * in this BehindIO request
107 						 */
108 	sector_t		sector;
109 	int			sectors;
110 	unsigned long		state;
111 	struct mddev		*mddev;
112 	/*
113 	 * original bio going to /dev/mdx
114 	 */
115 	struct bio		*master_bio;
116 	/*
117 	 * if the IO is in READ direction, then this is where we read
118 	 */
119 	int			read_disk;
120 
121 	struct list_head	retry_list;
122 	/* Next two are only valid when R1BIO_BehindIO is set */
123 	struct bio_vec		*behind_bvecs;
124 	int			behind_page_count;
125 	/*
126 	 * if the IO is in WRITE direction, then multiple bios are used.
127 	 * We choose the number when they are allocated.
128 	 */
129 	struct bio		*bios[0];
130 	/* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
131 };
132 
133 /* when we get a read error on a read-only array, we redirect to another
134  * device without failing the first device, or trying to over-write to
135  * correct the read error.  To keep track of bad blocks on a per-bio
136  * level, we store IO_BLOCKED in the appropriate 'bios' pointer
137  */
138 #define IO_BLOCKED ((struct bio *)1)
139 /* When we successfully write to a known bad-block, we need to remove the
140  * bad-block marking which must be done from process context.  So we record
141  * the success by setting bios[n] to IO_MADE_GOOD
142  */
143 #define IO_MADE_GOOD ((struct bio *)2)
144 
145 #define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
146 
147 /* bits for r1bio.state */
148 #define	R1BIO_Uptodate	0
149 #define	R1BIO_IsSync	1
150 #define	R1BIO_Degraded	2
151 #define	R1BIO_BehindIO	3
152 /* Set ReadError on bios that experience a readerror so that
153  * raid1d knows what to do with them.
154  */
155 #define R1BIO_ReadError 4
156 /* For write-behind requests, we call bi_end_io when
157  * the last non-write-behind device completes, providing
158  * any write was successful.  Otherwise we call when
159  * any write-behind write succeeds, otherwise we call
160  * with failure when last write completes (and all failed).
161  * Record that bi_end_io was called with this flag...
162  */
163 #define	R1BIO_Returned 6
164 /* If a write for this request means we can clear some
165  * known-bad-block records, we set this flag
166  */
167 #define	R1BIO_MadeGood 7
168 #define	R1BIO_WriteError 8
169 
170 extern int md_raid1_congested(struct mddev *mddev, int bits);
171 
172 #endif
173