Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1/*
2 md_k.h : kernel internal structure of the Linux MD driver
3 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
4
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2, or (at your option)
8 any later version.
9
10 You should have received a copy of the GNU General Public License
11 (for example /usr/src/linux/COPYING); if not, write to the Free
12 Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
13*/
14
15#ifndef _MD_K_H
16#define _MD_K_H
17
18/* and dm-bio-list.h is not under include/linux because.... ??? */
19#include "../../../drivers/md/dm-bio-list.h"
20
21#define MD_RESERVED 0UL
22#define LINEAR 1UL
23#define RAID0 2UL
24#define RAID1 3UL
25#define RAID5 4UL
26#define TRANSLUCENT 5UL
27#define HSM 6UL
28#define MULTIPATH 7UL
29#define RAID6 8UL
30#define RAID10 9UL
31#define FAULTY 10UL
32#define MAX_PERSONALITY 11UL
33
34#define LEVEL_MULTIPATH (-4)
35#define LEVEL_LINEAR (-1)
36#define LEVEL_FAULTY (-5)
37
38#define MaxSector (~(sector_t)0)
39#define MD_THREAD_NAME_MAX 14
40
41static inline int pers_to_level (int pers)
42{
43 switch (pers) {
44 case FAULTY: return LEVEL_FAULTY;
45 case MULTIPATH: return LEVEL_MULTIPATH;
46 case HSM: return -3;
47 case TRANSLUCENT: return -2;
48 case LINEAR: return LEVEL_LINEAR;
49 case RAID0: return 0;
50 case RAID1: return 1;
51 case RAID5: return 5;
52 case RAID6: return 6;
53 case RAID10: return 10;
54 }
55 BUG();
56 return MD_RESERVED;
57}
58
59static inline int level_to_pers (int level)
60{
61 switch (level) {
62 case LEVEL_FAULTY: return FAULTY;
63 case LEVEL_MULTIPATH: return MULTIPATH;
64 case -3: return HSM;
65 case -2: return TRANSLUCENT;
66 case LEVEL_LINEAR: return LINEAR;
67 case 0: return RAID0;
68 case 1: return RAID1;
69 case 4:
70 case 5: return RAID5;
71 case 6: return RAID6;
72 case 10: return RAID10;
73 }
74 return MD_RESERVED;
75}
76
77typedef struct mddev_s mddev_t;
78typedef struct mdk_rdev_s mdk_rdev_t;
79
80#define MAX_MD_DEVS 256 /* Max number of md dev */
81
82/*
83 * options passed in raidrun:
84 */
85
86#define MAX_CHUNK_SIZE (4096*1024)
87
88/*
89 * MD's 'extended' device
90 */
91struct mdk_rdev_s
92{
93 struct list_head same_set; /* RAID devices within the same set */
94
95 sector_t size; /* Device size (in blocks) */
96 mddev_t *mddev; /* RAID array if running */
97 unsigned long last_events; /* IO event timestamp */
98
99 struct block_device *bdev; /* block device handle */
100
101 struct page *sb_page;
102 int sb_loaded;
103 sector_t data_offset; /* start of data in array */
104 sector_t sb_offset;
105 int sb_size; /* bytes in the superblock */
106 int preferred_minor; /* autorun support */
107
108 /* A device can be in one of three states based on two flags:
109 * Not working: faulty==1 in_sync==0
110 * Fully working: faulty==0 in_sync==1
111 * Working, but not
112 * in sync with array
113 * faulty==0 in_sync==0
114 *
115 * It can never have faulty==1, in_sync==1
116 * This reduces the burden of testing multiple flags in many cases
117 */
118 int faulty; /* if faulty do not issue IO requests */
119 int in_sync; /* device is a full member of the array */
120
121 unsigned long flags; /* Should include faulty and in_sync here. */
122#define WriteMostly 4 /* Avoid reading if at all possible */
123
124 int desc_nr; /* descriptor index in the superblock */
125 int raid_disk; /* role of device in array */
126 int saved_raid_disk; /* role that device used to have in the
127 * array and could again if we did a partial
128 * resync from the bitmap
129 */
130
131 atomic_t nr_pending; /* number of pending requests.
132 * only maintained for arrays that
133 * support hot removal
134 */
135};
136
137typedef struct mdk_personality_s mdk_personality_t;
138
139struct mddev_s
140{
141 void *private;
142 mdk_personality_t *pers;
143 dev_t unit;
144 int md_minor;
145 struct list_head disks;
146 int sb_dirty;
147 int ro;
148
149 struct gendisk *gendisk;
150
151 /* Superblock information */
152 int major_version,
153 minor_version,
154 patch_version;
155 int persistent;
156 int chunk_size;
157 time_t ctime, utime;
158 int level, layout;
159 int raid_disks;
160 int max_disks;
161 sector_t size; /* used size of component devices */
162 sector_t array_size; /* exported array size */
163 __u64 events;
164
165 char uuid[16];
166
167 struct mdk_thread_s *thread; /* management thread */
168 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
169 sector_t curr_resync; /* blocks scheduled */
170 unsigned long resync_mark; /* a recent timestamp */
171 sector_t resync_mark_cnt;/* blocks written at resync_mark */
172
173 sector_t resync_max_sectors; /* may be set by personality */
174 /* recovery/resync flags
175 * NEEDED: we might need to start a resync/recover
176 * RUNNING: a thread is running, or about to be started
177 * SYNC: actually doing a resync, not a recovery
178 * ERR: and IO error was detected - abort the resync/recovery
179 * INTR: someone requested a (clean) early abort.
180 * DONE: thread is done and is waiting to be reaped
181 */
182#define MD_RECOVERY_RUNNING 0
183#define MD_RECOVERY_SYNC 1
184#define MD_RECOVERY_ERR 2
185#define MD_RECOVERY_INTR 3
186#define MD_RECOVERY_DONE 4
187#define MD_RECOVERY_NEEDED 5
188 unsigned long recovery;
189
190 int in_sync; /* know to not need resync */
191 struct semaphore reconfig_sem;
192 atomic_t active;
193
194 int changed; /* true if we might need to reread partition info */
195 int degraded; /* whether md should consider
196 * adding a spare
197 */
198
199 atomic_t recovery_active; /* blocks scheduled, but not written */
200 wait_queue_head_t recovery_wait;
201 sector_t recovery_cp;
202
203 spinlock_t write_lock;
204 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
205 atomic_t pending_writes; /* number of active superblock writes */
206
207 unsigned int safemode; /* if set, update "clean" superblock
208 * when no writes pending.
209 */
210 unsigned int safemode_delay;
211 struct timer_list safemode_timer;
212 atomic_t writes_pending;
213 request_queue_t *queue; /* for plugging ... */
214
215 atomic_t write_behind; /* outstanding async IO */
216 unsigned int max_write_behind; /* 0 = sync */
217
218 struct bitmap *bitmap; /* the bitmap for the device */
219 struct file *bitmap_file; /* the bitmap file */
220 long bitmap_offset; /* offset from superblock of
221 * start of bitmap. May be
222 * negative, but not '0'
223 */
224 long default_bitmap_offset; /* this is the offset to use when
225 * hot-adding a bitmap. It should
226 * eventually be settable by sysfs.
227 */
228
229 struct list_head all_mddevs;
230};
231
232
233static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
234{
235 int faulty = rdev->faulty;
236 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
237 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
238}
239
240static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
241{
242 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
243}
244
245struct mdk_personality_s
246{
247 char *name;
248 struct module *owner;
249 int (*make_request)(request_queue_t *q, struct bio *bio);
250 int (*run)(mddev_t *mddev);
251 int (*stop)(mddev_t *mddev);
252 void (*status)(struct seq_file *seq, mddev_t *mddev);
253 /* error_handler must set ->faulty and clear ->in_sync
254 * if appropriate, and should abort recovery if needed
255 */
256 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
257 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
258 int (*hot_remove_disk) (mddev_t *mddev, int number);
259 int (*spare_active) (mddev_t *mddev);
260 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
261 int (*resize) (mddev_t *mddev, sector_t sectors);
262 int (*reshape) (mddev_t *mddev, int raid_disks);
263 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
264 /* quiesce moves between quiescence states
265 * 0 - fully active
266 * 1 - no new requests allowed
267 * others - reserved
268 */
269 void (*quiesce) (mddev_t *mddev, int state);
270};
271
272
273static inline char * mdname (mddev_t * mddev)
274{
275 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
276}
277
278extern mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr);
279
280/*
281 * iterates through some rdev ringlist. It's safe to remove the
282 * current 'rdev'. Dont touch 'tmp' though.
283 */
284#define ITERATE_RDEV_GENERIC(head,rdev,tmp) \
285 \
286 for ((tmp) = (head).next; \
287 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
288 (tmp) = (tmp)->next, (tmp)->prev != &(head) \
289 ; )
290/*
291 * iterates through the 'same array disks' ringlist
292 */
293#define ITERATE_RDEV(mddev,rdev,tmp) \
294 ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
295
296/*
297 * Iterates through 'pending RAID disks'
298 */
299#define ITERATE_RDEV_PENDING(rdev,tmp) \
300 ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
301
302typedef struct mdk_thread_s {
303 void (*run) (mddev_t *mddev);
304 mddev_t *mddev;
305 wait_queue_head_t wqueue;
306 unsigned long flags;
307 struct completion *event;
308 struct task_struct *tsk;
309 unsigned long timeout;
310 const char *name;
311} mdk_thread_t;
312
313#define THREAD_WAKEUP 0
314
315#define __wait_event_lock_irq(wq, condition, lock, cmd) \
316do { \
317 wait_queue_t __wait; \
318 init_waitqueue_entry(&__wait, current); \
319 \
320 add_wait_queue(&wq, &__wait); \
321 for (;;) { \
322 set_current_state(TASK_UNINTERRUPTIBLE); \
323 if (condition) \
324 break; \
325 spin_unlock_irq(&lock); \
326 cmd; \
327 schedule(); \
328 spin_lock_irq(&lock); \
329 } \
330 current->state = TASK_RUNNING; \
331 remove_wait_queue(&wq, &__wait); \
332} while (0)
333
334#define wait_event_lock_irq(wq, condition, lock, cmd) \
335do { \
336 if (condition) \
337 break; \
338 __wait_event_lock_irq(wq, condition, lock, cmd); \
339} while (0)
340
341#endif
342