xref: /openbmc/linux/drivers/md/raid1.h (revision e7065e20)
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  * 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 
20 struct pool_info {
21 	struct mddev *mddev;
22 	int	raid_disks;
23 };
24 
25 struct 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 
106 struct 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 
175 extern int md_raid1_congested(struct mddev *mddev, int bits);
176 
177 #endif
178