]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Incremental: restore assembly for inactive containers, block active
[thirdparty/mdadm.git] / mdadm.h
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neilb@suse.de>
23 */
24
25 #define _GNU_SOURCE
26 #define _FILE_OFFSET_BITS 64
27 #include <unistd.h>
28 #if !defined(__dietlibc__) && !defined(__KLIBC__)
29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
30 #else
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
33 # endif
34 #endif
35
36 #include <sys/types.h>
37 #include <sys/stat.h>
38 #include <stdlib.h>
39 #include <time.h>
40 #include <sys/time.h>
41 #include <getopt.h>
42 #include <fcntl.h>
43 #include <stdio.h>
44 #include <errno.h>
45 #include <string.h>
46 #include <syslog.h>
47 #ifdef __dietlibc__
48 #include <strings.h>
49 /* dietlibc has deprecated random and srandom!! */
50 #define random rand
51 #define srandom srand
52 #endif
53
54
55 #include <linux/kdev_t.h>
56 /*#include <linux/fs.h> */
57 #include <sys/mount.h>
58 #include <asm/types.h>
59 #include <sys/ioctl.h>
60 #define MD_MAJOR 9
61 #define MdpMinorShift 6
62
63 #ifndef BLKGETSIZE64
64 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
65 #endif
66
67 #define DEFAULT_BITMAP_CHUNK 4096
68 #define DEFAULT_BITMAP_DELAY 5
69 #define DEFAULT_MAX_WRITE_BEHIND 256
70
71 /* VAR_RUN is where pid and socket files used for communicating
72 * with mdmon normally live. It should be /var/run, but if
73 * it is too hard to remount /var/run as read-only rather than
74 * unmounting it at shutdown time, then it should be
75 * redefined to some place that comfortably persists until
76 * final shutdown, possibly in /dev if that is a tmpfs.
77 * Note: VAR_RUN does not need to be writable at shutdown,
78 * only during boot when "mdmon --takeover" is run.
79 */
80 #ifndef VAR_RUN
81 #define VAR_RUN "/var/run/mdadm"
82 #endif /* VAR_RUN */
83 /* ALT_RUN should be somewhere that persists across the pivotroot
84 * from early boot to late boot.
85 * If you don't have /lib/init/rw you might want to use /dev/.something
86 */
87 #ifndef ALT_RUN
88 #define ALT_RUN "/lib/init/rw/mdadm"
89 #endif /* ALT_RUN */
90 #ifndef ALT_MAPFILE
91 #define ALT_MAPFILE "map"
92 #endif /* ALT_MAPFILE */
93
94 #include "md_u.h"
95 #include "md_p.h"
96 #include "bitmap.h"
97 #include "msg.h"
98
99 #include <endian.h>
100 /* Redhat don't like to #include <asm/byteorder.h>, and
101 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
102 * and there is no standard conversion function so... */
103 /* And dietlibc doesn't think byteswap is ok, so.. */
104 /* #include <byteswap.h> */
105 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
106 ((x) & 0xff00U) >> 8)
107 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
108 ((x) & 0xff000000U) >> 24 | \
109 ((x) & 0x0000ff00U) << 8 | \
110 ((x) & 0x00ff0000U) >> 8)
111 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
112 ((x) & 0xff00000000000000ULL) >> 56 | \
113 ((x) & 0x000000000000ff00ULL) << 40 | \
114 ((x) & 0x00ff000000000000ULL) >> 40 | \
115 ((x) & 0x0000000000ff0000ULL) << 24 | \
116 ((x) & 0x0000ff0000000000ULL) >> 24 | \
117 ((x) & 0x00000000ff000000ULL) << 8 | \
118 ((x) & 0x000000ff00000000ULL) >> 8)
119
120 #if !defined(__KLIBC__)
121 #if BYTE_ORDER == LITTLE_ENDIAN
122 #define __cpu_to_le16(_x) (_x)
123 #define __cpu_to_le32(_x) (_x)
124 #define __cpu_to_le64(_x) (_x)
125 #define __le16_to_cpu(_x) (_x)
126 #define __le32_to_cpu(_x) (_x)
127 #define __le64_to_cpu(_x) (_x)
128
129 #define __cpu_to_be16(_x) bswap_16(_x)
130 #define __cpu_to_be32(_x) bswap_32(_x)
131 #define __cpu_to_be64(_x) bswap_64(_x)
132 #define __be16_to_cpu(_x) bswap_16(_x)
133 #define __be32_to_cpu(_x) bswap_32(_x)
134 #define __be64_to_cpu(_x) bswap_64(_x)
135 #elif BYTE_ORDER == BIG_ENDIAN
136 #define __cpu_to_le16(_x) bswap_16(_x)
137 #define __cpu_to_le32(_x) bswap_32(_x)
138 #define __cpu_to_le64(_x) bswap_64(_x)
139 #define __le16_to_cpu(_x) bswap_16(_x)
140 #define __le32_to_cpu(_x) bswap_32(_x)
141 #define __le64_to_cpu(_x) bswap_64(_x)
142
143 #define __cpu_to_be16(_x) (_x)
144 #define __cpu_to_be32(_x) (_x)
145 #define __cpu_to_be64(_x) (_x)
146 #define __be16_to_cpu(_x) (_x)
147 #define __be32_to_cpu(_x) (_x)
148 #define __be64_to_cpu(_x) (_x)
149 #else
150 # error "unknown endianness."
151 #endif
152 #endif /* __KLIBC__ */
153
154
155 /*
156 * min()/max()/clamp() macros that also do
157 * strict type-checking.. See the
158 * "unnecessary" pointer comparison.
159 */
160 #define min(x, y) ({ \
161 typeof(x) _min1 = (x); \
162 typeof(y) _min2 = (y); \
163 (void) (&_min1 == &_min2); \
164 _min1 < _min2 ? _min1 : _min2; })
165
166 #define max(x, y) ({ \
167 typeof(x) _max1 = (x); \
168 typeof(y) _max2 = (y); \
169 (void) (&_max1 == &_max2); \
170 _max1 > _max2 ? _max1 : _max2; })
171
172 /* general information that might be extracted from a superblock */
173 struct mdinfo {
174 mdu_array_info_t array;
175 mdu_disk_info_t disk;
176 __u64 events;
177 int uuid[4];
178 char name[33];
179 unsigned long long data_offset;
180 unsigned long long component_size; /* same as array.size, except in
181 * sectors and up to 64bits.
182 */
183 unsigned long long custom_array_size; /* size for non-default sized
184 * arrays (in sectors)
185 */
186 int reshape_active;
187 unsigned long long reshape_progress;
188 union {
189 unsigned long long resync_start; /* per-array resync position */
190 unsigned long long recovery_start; /* per-device rebuild position */
191 #define MaxSector (~0ULL) /* resync/recovery complete position */
192 };
193 unsigned long safe_mode_delay; /* ms delay to mark clean */
194 int new_level, delta_disks, new_layout, new_chunk;
195 int errors;
196 int cache_size; /* size of raid456 stripe cache*/
197 int mismatch_cnt;
198 char text_version[50];
199 void *update_private; /* for passing metadata-format
200 * specific update data
201 * between successive calls to
202 * update_super()
203 */
204
205 int container_member; /* for assembling external-metatdata arrays
206 * This is to be used internally by metadata
207 * handler only */
208 int container_enough; /* flag external handlers can set to
209 * indicate that subarrays have not enough (-1),
210 * enough to start (0), or all expected disks (1) */
211 char sys_name[20];
212 struct mdinfo *devs;
213 struct mdinfo *next;
214
215 /* Device info for mdmon: */
216 int recovery_fd;
217 int state_fd;
218 #define DS_FAULTY 1
219 #define DS_INSYNC 2
220 #define DS_WRITE_MOSTLY 4
221 #define DS_SPARE 8
222 #define DS_BLOCKED 16
223 #define DS_REMOVE 1024
224 #define DS_UNBLOCK 2048
225 int prev_state, curr_state, next_state;
226
227 };
228
229 struct createinfo {
230 int uid;
231 int gid;
232 int autof;
233 int mode;
234 int symlinks;
235 struct supertype *supertype;
236 };
237
238 #define Name "mdadm"
239
240 enum mode {
241 ASSEMBLE=1,
242 BUILD,
243 CREATE,
244 MANAGE,
245 MISC,
246 MONITOR,
247 GROW,
248 INCREMENTAL,
249 AUTODETECT,
250 };
251
252 extern char short_options[];
253 extern char short_bitmap_options[];
254 extern char short_bitmap_auto_options[];
255 extern struct option long_options[];
256 extern char Version[], Usage[], Help[], OptionHelp[],
257 Help_create[], Help_build[], Help_assemble[], Help_grow[],
258 Help_incr[],
259 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
260
261 /* for option that don't have short equivilents, we assign arbitrary
262 * small numbers. '1' means an undecorated option, so we start at '2'.
263 * (note we must stop before we get to 65 i.e. 'A')
264 */
265 enum special_options {
266 AssumeClean = 2,
267 BitmapChunk,
268 WriteBehind,
269 ReAdd,
270 NoDegraded,
271 Sparc22,
272 BackupFile, /* 8 */
273 HomeHost,
274 AutoHomeHost,
275 Symlinks,
276 AutoDetect,
277 Waitclean,
278 DetailPlatform,
279 KillSubarray,
280 UpdateSubarray, /* 16 */
281 };
282
283 /* structures read from config file */
284 /* List of mddevice names and identifiers
285 * Identifiers can be:
286 * uuid=128-hex-uuid
287 * super-minor=decimal-minor-number-from-superblock
288 * devices=comma,separated,list,of,device,names,with,wildcards
289 *
290 * If multiple fields are present, the intersection of all matching
291 * devices is considered
292 */
293 #define UnSet (0xfffe)
294 typedef struct mddev_ident_s {
295 char *devname;
296
297 int uuid_set;
298 int uuid[4];
299 char name[33];
300
301 unsigned int super_minor;
302
303 char *devices; /* comma separated list of device
304 * names with wild cards
305 */
306 int level;
307 unsigned int raid_disks;
308 unsigned int spare_disks;
309 struct supertype *st;
310 int autof; /* 1 for normal, 2 for partitioned */
311 char *spare_group;
312 char *bitmap_file;
313 int bitmap_fd;
314
315 char *container; /* /dev/whatever name of container, or
316 * uuid of container. You would expect
317 * this to be the 'devname' or UUID
318 * of some other entry.
319 */
320 char *member; /* subarray within a container */
321
322 struct mddev_ident_s *next;
323 union {
324 /* fields needed by different users of this structure */
325 int assembled; /* set when assembly succeeds */
326 };
327 } *mddev_ident_t;
328
329 /* List of device names - wildcards expanded */
330 typedef struct mddev_dev_s {
331 char *devname;
332 char disposition; /* 'a' for add, 'r' for remove, 'f' for fail.
333 * Not set for names read from .config
334 */
335 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
336 char re_add;
337 char used; /* set when used */
338 struct mdinfo *content; /* If devname is a container, this might list
339 * the remaining member arrays. */
340 struct mddev_dev_s *next;
341 } *mddev_dev_t;
342
343 typedef struct mapping {
344 char *name;
345 int num;
346 } mapping_t;
347
348
349 struct mdstat_ent {
350 char *dev;
351 int devnum;
352 int active;
353 char *level;
354 char *pattern; /* U or up, _ for down */
355 int percent; /* -1 if no resync */
356 int resync; /* 1 if resync, 0 if recovery */
357 int devcnt;
358 int raid_disks;
359 int chunk_size;
360 char * metadata_version;
361 struct dev_member {
362 char *name;
363 struct dev_member *next;
364 } *members;
365 struct mdstat_ent *next;
366 };
367
368 extern struct mdstat_ent *mdstat_read(int hold, int start);
369 extern void free_mdstat(struct mdstat_ent *ms);
370 extern void mdstat_wait(int seconds);
371 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
372 extern int mddev_busy(int devnum);
373 extern struct mdstat_ent *mdstat_by_component(char *name);
374
375 struct map_ent {
376 struct map_ent *next;
377 int devnum;
378 char metadata[20];
379 int uuid[4];
380 int bad;
381 char *path;
382 };
383 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
384 int uuid[4], char *path);
385 extern void map_remove(struct map_ent **map, int devnum);
386 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
387 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
388 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
389 extern void map_read(struct map_ent **melp);
390 extern int map_write(struct map_ent *mel);
391 extern void map_delete(struct map_ent **mapp, int devnum);
392 extern void map_free(struct map_ent *map);
393 extern void map_add(struct map_ent **melp,
394 int devnum, char *metadata, int uuid[4], char *path);
395 extern int map_lock(struct map_ent **melp);
396 extern void map_unlock(struct map_ent **melp);
397
398 /* various details can be requested */
399 enum sysfs_read_flags {
400 GET_LEVEL = (1 << 0),
401 GET_LAYOUT = (1 << 1),
402 GET_COMPONENT = (1 << 2),
403 GET_CHUNK = (1 << 3),
404 GET_CACHE = (1 << 4),
405 GET_MISMATCH = (1 << 5),
406 GET_VERSION = (1 << 6),
407 GET_DISKS = (1 << 7),
408 GET_DEGRADED = (1 << 8),
409 GET_SAFEMODE = (1 << 9),
410 GET_DEVS = (1 << 10), /* gets role, major, minor */
411 GET_OFFSET = (1 << 11),
412 GET_SIZE = (1 << 12),
413 GET_STATE = (1 << 13),
414 GET_ERROR = (1 << 14),
415 };
416
417 /* If fd >= 0, get the array it is open on,
418 * else use devnum. >=0 -> major9. <0.....
419 */
420 extern int sysfs_open(int devnum, char *devname, char *attr);
421 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
422 extern void sysfs_free(struct mdinfo *sra);
423 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
424 extern int sysfs_attr_match(const char *attr, const char *str);
425 extern int sysfs_match_word(const char *word, char **list);
426 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
427 char *name, char *val);
428 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
429 char *name, unsigned long long val);
430 extern int sysfs_uevent(struct mdinfo *sra, char *event);
431 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
432 char *name);
433 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
434 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
435 char *name, unsigned long long *val);
436 extern int sysfs_fd_get_str(int fd, char *val, int size);
437 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
438 char *name, char *val, int size);
439 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
440 extern int sysfs_set_array(struct mdinfo *info, int vers);
441 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
442 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
443 extern int sysfs_unique_holder(int devnum, long rdev);
444 extern int load_sys(char *path, char *buf);
445
446
447 extern int save_stripes(int *source, unsigned long long *offsets,
448 int raid_disks, int chunk_size, int level, int layout,
449 int nwrites, int *dest,
450 unsigned long long start, unsigned long long length,
451 char *buf);
452 extern int restore_stripes(int *dest, unsigned long long *offsets,
453 int raid_disks, int chunk_size, int level, int layout,
454 int source, unsigned long long read_offset,
455 unsigned long long start, unsigned long long length);
456
457 #ifndef Sendmail
458 #define Sendmail "/usr/lib/sendmail -t"
459 #endif
460
461 #define SYSLOG_FACILITY LOG_DAEMON
462
463 extern char *map_num(mapping_t *map, int num);
464 extern int map_name(mapping_t *map, char *name);
465 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
466
467 extern char *map_dev(int major, int minor, int create);
468
469 struct active_array;
470 struct metadata_update;
471
472 /* A superswitch provides entry point the a metadata handler.
473 *
474 * The super_switch primarily operates on some "metadata" that
475 * is accessed via the 'supertype'.
476 * This metadata has one of three possible sources.
477 * 1/ It is read from a single device. In this case it may not completely
478 * describe the array or arrays as some information might be on other
479 * devices.
480 * 2/ It is read from all devices in a container. In this case all
481 * information is present.
482 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
483 * be complete once enough ->add_to_super calls have completed.
484 *
485 * When creating an array inside a container, the metadata will be
486 * formed by a combination of 2 and 3. The metadata or the array is read,
487 * then new information is added.
488 *
489 * The metadata must sometimes have a concept of a 'current' array
490 * and a 'current' device.
491 * The 'current' array is set by init_super to be the newly created array,
492 * or is set by super_by_fd when it finds it is looking at an array inside
493 * a container.
494 *
495 * The 'current' device is either the device that the metadata was read from
496 * in case 1, or the last device added by add_to_super in case 3.
497 * Case 2 does not identify a 'current' device.
498 */
499 extern struct superswitch {
500
501 /* Used to report details of metadata read from a component
502 * device. ->load_super has been called.
503 */
504 void (*examine_super)(struct supertype *st, char *homehost);
505 void (*brief_examine_super)(struct supertype *st, int verbose);
506 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
507 void (*export_examine_super)(struct supertype *st);
508
509 /* Used to report details of an active array.
510 * ->load_super was possibly given a 'component' string.
511 */
512 void (*detail_super)(struct supertype *st, char *homehost);
513 void (*brief_detail_super)(struct supertype *st);
514 void (*export_detail_super)(struct supertype *st);
515
516 /* Optional: platform hardware / firmware details */
517 int (*detail_platform)(int verbose, int enumerate_only);
518
519 /* Used:
520 * to get uuid to storing in bitmap metadata
521 * and 'reshape' backup-data metadata
522 * To see if a device is being re-added to an array it was part of.
523 */
524 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
525
526 /* Extract generic details from metadata. This could be details about
527 * the container, or about an individual array within the container.
528 * The determination is made either by:
529 * load_super being given a 'component' string.
530 * validate_geometry determining what to create.
531 * The info includes both array information and device information.
532 * The particular device should be:
533 * The last device added by add_to_super
534 * The device the metadata was loaded from by load_super
535 */
536 void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
537
538 /* Check if the given metadata is flagged as belonging to "this"
539 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
540 */
541 int (*match_home)(struct supertype *st, char *homehost);
542
543 /* Make one of several generic modifications to metadata
544 * prior to assembly (or other times).
545 * sparc2.2 - first bug in early 0.90 metadata
546 * super-minor - change name of 0.90 metadata
547 * summaries - 'correct' any redundant data
548 * resync - mark array as dirty to trigger a resync.
549 * uuid - set new uuid - only 0.90 or 1.x
550 * name - change the name of the array (where supported)
551 * homehost - change which host this array is tied to.
552 * devicesize - If metadata is at start of device, change recorded
553 * device size to match actual device size
554 * byteorder - swap bytes for 0.90 metadata
555 *
556 * force-one - mark that device as uptodate, not old or failed.
557 * force-array - mark array as clean if it would not otherwise
558 * assemble
559 * assemble - not sure how this is different from force-one...
560 * linear-grow-new - add a new device to a linear array, but don't
561 * change the size: so superblock still matches
562 * linear-grow-update - now change the size of the array.
563 */
564 int (*update_super)(struct supertype *st, struct mdinfo *info,
565 char *update,
566 char *devname, int verbose,
567 int uuid_set, char *homehost);
568
569 /* Create new metadata for new array as described. This could
570 * be a new container, or an array in a pre-existing container.
571 * Also used to zero metadata prior to writing it to invalidate old
572 * metadata.
573 */
574 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
575 unsigned long long size, char *name,
576 char *homehost, int *uuid);
577
578 /* update the metadata to include new device, either at create or
579 * when hot-adding a spare.
580 */
581 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
582 int fd, char *devname);
583
584 /* Write metadata to one device when fixing problems or adding
585 * a new device.
586 */
587 int (*store_super)(struct supertype *st, int fd);
588
589 /* Write all metadata for this array.
590 */
591 int (*write_init_super)(struct supertype *st);
592 int (*compare_super)(struct supertype *st, struct supertype *tst);
593 int (*load_super)(struct supertype *st, int fd, char *devname);
594 struct supertype * (*match_metadata_desc)(char *arg);
595 __u64 (*avail_size)(struct supertype *st, __u64 size);
596 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
597 int delay, int write_behind,
598 unsigned long long size, int may_change, int major);
599 void (*locate_bitmap)(struct supertype *st, int fd);
600 int (*write_bitmap)(struct supertype *st, int fd);
601 void (*free_super)(struct supertype *st);
602
603 /* validate_geometry is called with an st returned by
604 * match_metadata_desc.
605 * It should check that the geometry described in compatible with
606 * the metadata type. It will be called repeatedly as devices
607 * added to validate changing size and new devices. If there are
608 * inter-device dependencies, it should record sufficient details
609 * so these can be validated.
610 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
611 */
612 int (*validate_geometry)(struct supertype *st, int level, int layout,
613 int raiddisks,
614 int chunk, unsigned long long size,
615 char *subdev, unsigned long long *freesize,
616 int verbose);
617
618 struct mdinfo *(*container_content)(struct supertype *st);
619 /* Allow a metadata handler to override mdadm's default layouts */
620 int (*default_layout)(int level); /* optional */
621 /* query the supertype for default chunk size */
622 int (*default_chunk)(struct supertype *st); /* optional */
623 /* Permit subarray's to be deleted from inactive containers */
624 int (*kill_subarray)(struct supertype *st); /* optional */
625 /* Permit subarray's to be modified */
626 int (*update_subarray)(struct supertype *st, char *update, mddev_ident_t ident); /* optional */
627
628 /* for mdmon */
629 int (*open_new)(struct supertype *c, struct active_array *a,
630 char *inst);
631
632 /* Tell the metadata handler the current state of the array.
633 * This covers whether it is known to be consistent (no pending writes)
634 * and how far along a resync is known to have progressed
635 * (in a->resync_start).
636 * resync status is really irrelevant if the array is not consistent,
637 * but some metadata (DDF!) have a place to record the distinction.
638 * If 'consistent' is '2', then the array can mark it dirty if a
639 * resync/recovery/whatever is required, or leave it clean if not.
640 * Return value is 0 dirty (not consistent) and 1 if clean.
641 * it is only really important if consistent is passed in as '2'.
642 */
643 int (*set_array_state)(struct active_array *a, int consistent);
644
645 /* When the state of a device might have changed, we call set_disk to
646 * tell the metadata what the current state is.
647 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
648 * transitions.
649 * set_disk might be called when the state of the particular disk has
650 * not in fact changed.
651 */
652 void (*set_disk)(struct active_array *a, int n, int state);
653 void (*sync_metadata)(struct supertype *st);
654 void (*process_update)(struct supertype *st,
655 struct metadata_update *update);
656 void (*prepare_update)(struct supertype *st,
657 struct metadata_update *update);
658
659 /* activate_spare will check if the array is degraded and, if it
660 * is, try to find some spare space in the container.
661 * On success, it add appropriate updates (For process_update) to
662 * to the 'updates' list and returns a list of 'mdinfo' identifying
663 * the device, or devices as there might be multiple missing
664 * devices and multiple spares available.
665 */
666 struct mdinfo *(*activate_spare)(struct active_array *a,
667 struct metadata_update **updates);
668
669 int swapuuid; /* true if uuid is bigending rather than hostendian */
670 int external;
671 const char *name; /* canonical metadata name */
672 } super0, super1, super_ddf, *superlist[];
673
674 extern struct superswitch super_imsm;
675
676 struct metadata_update {
677 int len;
678 char *buf;
679 void *space; /* allocated space that monitor will use */
680 struct metadata_update *next;
681 };
682
683 /* A supertype holds a particular collection of metadata.
684 * It identifies the metadata type by the superswitch, and the particular
685 * sub-version of that metadata type.
686 * metadata read in or created is stored in 'sb' and 'info'.
687 * There are also fields used by mdmon to track containers.
688 *
689 * A supertype may refer to:
690 * Just an array, possibly in a container
691 * A container, not identifying any particular array
692 * Info read from just one device, not yet fully describing the array/container.
693 *
694 *
695 * A supertype is created by:
696 * super_by_fd
697 * guess_super
698 * dup_super
699 */
700 struct supertype {
701 struct superswitch *ss;
702 int minor_version;
703 int max_devs;
704 int container_dev; /* devnum of container */
705 char subarray[32]; /* name of array inside container */
706 void *sb;
707 void *info;
708 int loaded_container; /* Set if load_super found a container,
709 * not just one device */
710
711 struct metadata_update *updates;
712 struct metadata_update **update_tail;
713
714 /* extra stuff used by mdmon */
715 struct active_array *arrays;
716 int sock; /* listen to external programs */
717 int devnum;
718 char *devname; /* e.g. md0. This appears in metadata_verison:
719 * external:/md0/12
720 */
721 int devcnt;
722
723 struct mdinfo *devs;
724
725 };
726
727 extern struct supertype *super_by_fd(int fd);
728 extern struct supertype *guess_super(int fd);
729 extern struct supertype *dup_super(struct supertype *st);
730 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
731 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
732 mdu_disk_info_t *disk);
733 void wait_for(char *dev, int fd);
734
735 #if __GNUC__ < 3
736 struct stat64;
737 #endif
738
739 #define HAVE_NFTW we assume
740 #define HAVE_FTW
741
742 #ifdef __UCLIBC__
743 # include <features.h>
744 # ifndef __UCLIBC_HAS_LFS__
745 # define lseek64 lseek
746 # endif
747 # ifndef __UCLIBC_HAS_FTW__
748 # undef HAVE_FTW
749 # undef HAVE_NFTW
750 # endif
751 #endif
752
753 #ifdef __dietlibc__
754 # undef HAVE_NFTW
755 #endif
756
757 #if defined(__KLIBC__)
758 # undef HAVE_NFTW
759 # undef HAVE_FTW
760 #endif
761
762 #ifndef HAVE_NFTW
763 # define FTW_PHYS 1
764 # ifndef HAVE_FTW
765 struct FTW {};
766 # endif
767 #endif
768
769 #ifdef HAVE_FTW
770 # include <ftw.h>
771 #endif
772
773 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
774
775
776 extern int Manage_ro(char *devname, int fd, int readonly);
777 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
778 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
779 extern int Manage_subdevs(char *devname, int fd,
780 mddev_dev_t devlist, int verbose, int test);
781 extern int autodetect(void);
782 extern int Grow_Add_device(char *devname, int fd, char *newdev);
783 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
784 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
785 long long size,
786 int level, char *layout_str, int chunksize, int raid_disks);
787 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
788 int *fdlist, int cnt, char *backup_file, int verbose);
789 extern int Grow_continue(int mdfd, struct supertype *st,
790 struct mdinfo *info, char *backup_file);
791
792 extern int Assemble(struct supertype *st, char *mddev,
793 mddev_ident_t ident,
794 mddev_dev_t devlist, char *backup_file,
795 int readonly, int runstop,
796 char *update, char *homehost, int require_homehost,
797 int verbose, int force);
798
799 extern int Build(char *mddev, int chunk, int level, int layout,
800 int raiddisks, mddev_dev_t devlist, int assume_clean,
801 char *bitmap_file, int bitmap_chunk, int write_behind,
802 int delay, int verbose, int autof, unsigned long long size);
803
804
805 extern int Create(struct supertype *st, char *mddev,
806 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
807 char *name, char *homehost, int *uuid,
808 int subdevs, mddev_dev_t devlist,
809 int runstop, int verbose, int force, int assume_clean,
810 char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
811
812 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
813 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
814 extern int Query(char *dev);
815 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
816 int SparcAdjust, struct supertype *forcest, char *homehost);
817 extern int Monitor(mddev_dev_t devlist,
818 char *mailaddr, char *alert_cmd,
819 int period, int daemonise, int scan, int oneshot,
820 int dosyslog, int test, char *pidfile, int increments);
821
822 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
823 extern int Kill_subarray(char *dev, char *subarray, int quiet);
824 extern int Update_subarray(char *dev, char *subarray, char *update, mddev_ident_t ident, int quiet);
825 extern int Wait(char *dev);
826 extern int WaitClean(char *dev, int sock, int verbose);
827
828 extern int Incremental(char *devname, int verbose, int runstop,
829 struct supertype *st, char *homehost, int require_homehost,
830 int autof);
831 extern int Incremental_container(struct supertype *st, char *devname,
832 int verbose, int runstop, int autof,
833 int trustworthy);
834 extern void RebuildMap(void);
835 extern int IncrementalScan(int verbose);
836 extern int IncrementalRemove(char *devname, int verbose);
837 extern int CreateBitmap(char *filename, int force, char uuid[16],
838 unsigned long chunksize, unsigned long daemon_sleep,
839 unsigned long write_behind,
840 unsigned long long array_size,
841 int major);
842 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
843 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
844 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
845
846 extern int md_get_version(int fd);
847 extern int get_linux_version(void);
848 extern long long parse_size(char *size);
849 extern int parse_uuid(char *str, int uuid[4]);
850 extern int parse_layout_10(char *layout);
851 extern int parse_layout_faulty(char *layout);
852 extern int check_ext2(int fd, char *name);
853 extern int check_reiser(int fd, char *name);
854 extern int check_raid(int fd, char *name);
855 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
856
857 extern int get_mdp_major(void);
858 extern int dev_open(char *dev, int flags);
859 extern int open_dev(int devnum);
860 extern int open_dev_excl(int devnum);
861 extern int is_standard(char *dev, int *nump);
862 extern int same_dev(char *one, char *two);
863
864 extern int parse_auto(char *str, char *msg, int config);
865 extern mddev_ident_t conf_get_ident(char *dev);
866 extern mddev_dev_t conf_get_devs(void);
867 extern int conf_test_dev(char *devname);
868 extern int conf_test_metadata(const char *version, int is_homehost);
869 extern struct createinfo *conf_get_create_info(void);
870 extern void set_conffile(char *file);
871 extern char *conf_get_mailaddr(void);
872 extern char *conf_get_mailfrom(void);
873 extern char *conf_get_program(void);
874 extern char *conf_get_homehost(int *require_homehostp);
875 extern char *conf_line(FILE *file);
876 extern char *conf_word(FILE *file, int allow_key);
877 extern int conf_name_is_free(char *name);
878 extern int devname_matches(char *name, char *match);
879 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
880
881 extern void free_line(char *line);
882 extern int match_oneof(char *devices, char *devname);
883 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
884 extern const int uuid_match_any[4];
885 extern int same_uuid(int a[4], int b[4], int swapuuid);
886 extern void copy_uuid(void *a, int b[4], int swapuuid);
887 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
888 extern char *fname_from_uuid(struct supertype *st,
889 struct mdinfo *info, char *buf, char sep);
890 extern unsigned long calc_csum(void *super, int bytes);
891 extern int enough(int level, int raid_disks, int layout, int clean,
892 char *avail, int avail_disks);
893 extern int ask(char *mesg);
894 extern unsigned long long get_component_size(int fd);
895 extern void remove_partitions(int fd);
896 extern int test_partition(int fd);
897 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
898 int chunksize, unsigned long long devsize);
899 extern int flush_metadata_updates(struct supertype *st);
900 extern void append_metadata_update(struct supertype *st, void *buf, int len);
901 extern int assemble_container_content(struct supertype *st, int mdfd,
902 struct mdinfo *content, int runstop,
903 char *chosen_name, int verbose);
904
905 extern int add_disk(int mdfd, struct supertype *st,
906 struct mdinfo *sra, struct mdinfo *info);
907 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
908 unsigned long long min_recovery_start(struct mdinfo *array);
909
910 extern char *human_size(long long bytes);
911 extern char *human_size_brief(long long bytes);
912 extern void print_r10_layout(int layout);
913
914 #define NoMdDev (1<<23)
915 extern int find_free_devnum(int use_partitions);
916
917 extern void put_md_name(char *name);
918 extern char *get_md_name(int dev);
919
920 extern char DefaultConfFile[];
921
922 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
923 char *chosen);
924 /* values for 'trustworthy' */
925 #define LOCAL 1
926 #define LOCAL_ANY 10
927 #define FOREIGN 2
928 #define METADATA 3
929 extern int open_mddev(char *dev, int report_errors);
930 extern int open_container(int fd);
931 extern int is_container_member(struct mdstat_ent *ent, char *devname);
932 extern int is_subarray_active(char *subarray, char *devname);
933 int is_container_active(char *devname);
934 extern int open_subarray(char *dev, struct supertype *st, int quiet);
935 extern struct superswitch *version_to_superswitch(char *vers);
936
937 extern char *pid_dir;
938 extern int mdmon_running(int devnum);
939 extern int mdmon_pid(int devnum);
940 extern int check_env(char *name);
941 extern __u32 random32(void);
942 extern int start_mdmon(int devnum);
943
944 extern char *devnum2devname(int num);
945 extern int devname2devnum(char *name);
946 extern int stat2devnum(struct stat *st);
947 extern int fd2devnum(int fd);
948
949 static inline int dev2major(int d)
950 {
951 if (d >= 0)
952 return MD_MAJOR;
953 else
954 return get_mdp_major();
955 }
956
957 static inline int dev2minor(int d)
958 {
959 if (d >= 0)
960 return d;
961 return (-1-d) << MdpMinorShift;
962 }
963
964 static inline int ROUND_UP(int a, int base)
965 {
966 return ((a+base-1)/base)*base;
967 }
968
969 static inline int is_subarray(char *vers)
970 {
971 /* The version string for a 'subarray' (an array in a container)
972 * is
973 * /containername/componentname for normal read-write arrays
974 * -containername/componentname for read-only arrays.
975 * containername is e.g. md0, md_d1
976 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
977 */
978 return (*vers == '/' || *vers == '-');
979 }
980
981 #ifdef DEBUG
982 #define dprintf(fmt, arg...) \
983 fprintf(stderr, fmt, ##arg)
984 #else
985 #define dprintf(fmt, arg...) \
986 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
987 #endif
988 #include <assert.h>
989 #include <stdarg.h>
990 static inline int xasprintf(char **strp, const char *fmt, ...) {
991 va_list ap;
992 int ret;
993 va_start(ap, fmt);
994 ret = vasprintf(strp, fmt, ap);
995 va_end(ap);
996 assert(ret >= 0);
997 return ret;
998 }
999
1000 #define LEVEL_MULTIPATH (-4)
1001 #define LEVEL_LINEAR (-1)
1002 #define LEVEL_FAULTY (-5)
1003
1004 /* kernel module doesn't know about these */
1005 #define LEVEL_CONTAINER (-100)
1006 #define LEVEL_UNSUPPORTED (-200)
1007
1008
1009 /* faulty stuff */
1010
1011 #define WriteTransient 0
1012 #define ReadTransient 1
1013 #define WritePersistent 2
1014 #define ReadPersistent 3
1015 #define WriteAll 4 /* doesn't go to device */
1016 #define ReadFixable 5
1017 #define Modes 6
1018
1019 #define ClearErrors 31
1020 #define ClearFaults 30
1021
1022 #define AllPersist 100 /* internal use only */
1023 #define NoPersist 101
1024
1025 #define ModeMask 0x1f
1026 #define ModeShift 5
1027
1028
1029 #ifdef __TINYC__
1030 #undef minor
1031 #undef major
1032 #undef makedev
1033 #define minor(x) ((x)&0xff)
1034 #define major(x) (((x)>>8)&0xff)
1035 #define makedev(M,m) (((M)<<8) | (m))
1036 #endif
1037
1038 /* for raid4/5/6 */
1039 #define ALGORITHM_LEFT_ASYMMETRIC 0
1040 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1041 #define ALGORITHM_LEFT_SYMMETRIC 2
1042 #define ALGORITHM_RIGHT_SYMMETRIC 3
1043
1044 /* Define non-rotating (raid4) algorithms. These allow
1045 * conversion of raid4 to raid5.
1046 */
1047 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1048 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1049
1050 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1051 * Firstly, the exact positioning of the parity block is slightly
1052 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1053 * of DDF.
1054 * Secondly, or order of datablocks over which the Q syndrome is computed
1055 * is different.
1056 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1057 * These layouts are from the DDFv1.2 spec.
1058 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1059 * leaves RLQ=3 as 'Vendor Specific'
1060 */
1061
1062 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1063 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1064 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1065
1066
1067 /* For every RAID5 algorithm we define a RAID6 algorithm
1068 * with exactly the same layout for data and parity, and
1069 * with the Q block always on the last device (N-1).
1070 * This allows trivial conversion from RAID5 to RAID6
1071 */
1072 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1073 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1074 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1075 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1076 #define ALGORITHM_PARITY_0_6 20
1077 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1078
1079 /* Define PATH_MAX in case we don't use glibc or standard library does
1080 * not have PATH_MAX defined. Assume max path length is 4K characters.
1081 */
1082 #ifndef PATH_MAX
1083 #define PATH_MAX 4096
1084 #endif
1085