08e454404bccb3494bc8789d1f9d3f97c42c767a
[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 /* MAP_DIR should be somewhere that persists across the pivotroot
72 * from early boot to late boot.
73 * Currently /dev seems to be the only option on most distros.
74 */
75 #ifndef MAP_DIR
76 #define MAP_DIR "/dev/.mdadm"
77 #endif /* MAP_DIR */
78 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
79 * want something other than the default of "map"
80 */
81 #ifndef MAP_FILE
82 #define MAP_FILE "map"
83 #endif /* MAP_FILE */
84 /* MDMON_DIR is where pid and socket files used for communicating
85 * with mdmon normally live. It *should* be /var/run, but when
86 * mdmon is needed at early boot then it needs to write there prior
87 * to /var/run being mounted read/write, and it also then needs to
88 * persist beyond when /var/run is mounter read-only. So, to be
89 * safe, the default is somewhere that is read/write early in the
90 * boot process and stays up as long as possible during shutdown.
91 */
92 #ifndef MDMON_DIR
93 #define MDMON_DIR "/dev/.mdadm/"
94 #endif /* MDMON_DIR */
95
96 #include "md_u.h"
97 #include "md_p.h"
98 #include "bitmap.h"
99 #include "msg.h"
100
101 #include <endian.h>
102 /* Redhat don't like to #include <asm/byteorder.h>, and
103 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
104 * and there is no standard conversion function so... */
105 /* And dietlibc doesn't think byteswap is ok, so.. */
106 /* #include <byteswap.h> */
107 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
108 ((x) & 0xff00U) >> 8)
109 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
110 ((x) & 0xff000000U) >> 24 | \
111 ((x) & 0x0000ff00U) << 8 | \
112 ((x) & 0x00ff0000U) >> 8)
113 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
114 ((x) & 0xff00000000000000ULL) >> 56 | \
115 ((x) & 0x000000000000ff00ULL) << 40 | \
116 ((x) & 0x00ff000000000000ULL) >> 40 | \
117 ((x) & 0x0000000000ff0000ULL) << 24 | \
118 ((x) & 0x0000ff0000000000ULL) >> 24 | \
119 ((x) & 0x00000000ff000000ULL) << 8 | \
120 ((x) & 0x000000ff00000000ULL) >> 8)
121
122 #if !defined(__KLIBC__)
123 #if BYTE_ORDER == LITTLE_ENDIAN
124 #define __cpu_to_le16(_x) (_x)
125 #define __cpu_to_le32(_x) (_x)
126 #define __cpu_to_le64(_x) (_x)
127 #define __le16_to_cpu(_x) (_x)
128 #define __le32_to_cpu(_x) (_x)
129 #define __le64_to_cpu(_x) (_x)
130
131 #define __cpu_to_be16(_x) bswap_16(_x)
132 #define __cpu_to_be32(_x) bswap_32(_x)
133 #define __cpu_to_be64(_x) bswap_64(_x)
134 #define __be16_to_cpu(_x) bswap_16(_x)
135 #define __be32_to_cpu(_x) bswap_32(_x)
136 #define __be64_to_cpu(_x) bswap_64(_x)
137 #elif BYTE_ORDER == BIG_ENDIAN
138 #define __cpu_to_le16(_x) bswap_16(_x)
139 #define __cpu_to_le32(_x) bswap_32(_x)
140 #define __cpu_to_le64(_x) bswap_64(_x)
141 #define __le16_to_cpu(_x) bswap_16(_x)
142 #define __le32_to_cpu(_x) bswap_32(_x)
143 #define __le64_to_cpu(_x) bswap_64(_x)
144
145 #define __cpu_to_be16(_x) (_x)
146 #define __cpu_to_be32(_x) (_x)
147 #define __cpu_to_be64(_x) (_x)
148 #define __be16_to_cpu(_x) (_x)
149 #define __be32_to_cpu(_x) (_x)
150 #define __be64_to_cpu(_x) (_x)
151 #else
152 # error "unknown endianness."
153 #endif
154 #endif /* __KLIBC__ */
155
156
157 /*
158 * min()/max()/clamp() macros that also do
159 * strict type-checking.. See the
160 * "unnecessary" pointer comparison.
161 */
162 #define min(x, y) ({ \
163 typeof(x) _min1 = (x); \
164 typeof(y) _min2 = (y); \
165 (void) (&_min1 == &_min2); \
166 _min1 < _min2 ? _min1 : _min2; })
167
168 #define max(x, y) ({ \
169 typeof(x) _max1 = (x); \
170 typeof(y) _max2 = (y); \
171 (void) (&_max1 == &_max2); \
172 _max1 > _max2 ? _max1 : _max2; })
173
174 /* general information that might be extracted from a superblock */
175 struct mdinfo {
176 mdu_array_info_t array;
177 mdu_disk_info_t disk;
178 __u64 events;
179 int uuid[4];
180 char name[33];
181 unsigned long long data_offset;
182 unsigned long long component_size; /* same as array.size, except in
183 * sectors and up to 64bits.
184 */
185 unsigned long long custom_array_size; /* size for non-default sized
186 * arrays (in sectors)
187 */
188 int reshape_active;
189 unsigned long long reshape_progress;
190 union {
191 unsigned long long resync_start; /* per-array resync position */
192 unsigned long long recovery_start; /* per-device rebuild position */
193 #define MaxSector (~0ULL) /* resync/recovery complete position */
194 };
195 unsigned long safe_mode_delay; /* ms delay to mark clean */
196 int new_level, delta_disks, new_layout, new_chunk;
197 int errors;
198 unsigned long cache_size; /* size of raid456 stripe cache*/
199 int mismatch_cnt;
200 char text_version[50];
201 void *update_private; /* for passing metadata-format
202 * specific update data
203 * between successive calls to
204 * update_super()
205 */
206
207 int container_member; /* for assembling external-metatdata arrays
208 * This is to be used internally by metadata
209 * handler only */
210 int container_enough; /* flag external handlers can set to
211 * indicate that subarrays have not enough (-1),
212 * enough to start (0), or all expected disks (1) */
213 char sys_name[20];
214 struct mdinfo *devs;
215 struct mdinfo *next;
216
217 /* Device info for mdmon: */
218 int recovery_fd;
219 int state_fd;
220 #define DS_FAULTY 1
221 #define DS_INSYNC 2
222 #define DS_WRITE_MOSTLY 4
223 #define DS_SPARE 8
224 #define DS_BLOCKED 16
225 #define DS_REMOVE 1024
226 #define DS_UNBLOCK 2048
227 int prev_state, curr_state, next_state;
228
229 };
230
231 struct createinfo {
232 int uid;
233 int gid;
234 int autof;
235 int mode;
236 int symlinks;
237 struct supertype *supertype;
238 };
239
240 #define Name "mdadm"
241
242 enum mode {
243 ASSEMBLE=1,
244 BUILD,
245 CREATE,
246 MANAGE,
247 MISC,
248 MONITOR,
249 GROW,
250 INCREMENTAL,
251 AUTODETECT,
252 };
253
254 extern char short_options[];
255 extern char short_bitmap_options[];
256 extern char short_bitmap_auto_options[];
257 extern struct option long_options[];
258 extern char Version[], Usage[], Help[], OptionHelp[],
259 Help_create[], Help_build[], Help_assemble[], Help_grow[],
260 Help_incr[],
261 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
262
263 /* for option that don't have short equivilents, we assign arbitrary
264 * small numbers. '1' means an undecorated option, so we start at '2'.
265 * (note we must stop before we get to 65 i.e. 'A')
266 */
267 enum special_options {
268 AssumeClean = 2,
269 BitmapChunk,
270 WriteBehind,
271 ReAdd,
272 NoDegraded,
273 Sparc22,
274 BackupFile, /* 8 */
275 HomeHost,
276 AutoHomeHost,
277 Symlinks,
278 AutoDetect,
279 Waitclean,
280 DetailPlatform,
281 KillSubarray,
282 UpdateSubarray, /* 16 */
283 };
284
285 /* structures read from config file */
286 /* List of mddevice names and identifiers
287 * Identifiers can be:
288 * uuid=128-hex-uuid
289 * super-minor=decimal-minor-number-from-superblock
290 * devices=comma,separated,list,of,device,names,with,wildcards
291 *
292 * If multiple fields are present, the intersection of all matching
293 * devices is considered
294 */
295 #define UnSet (0xfffe)
296 typedef struct mddev_ident_s {
297 char *devname;
298
299 int uuid_set;
300 int uuid[4];
301 char name[33];
302
303 int super_minor;
304
305 char *devices; /* comma separated list of device
306 * names with wild cards
307 */
308 int level;
309 int raid_disks;
310 int spare_disks;
311 struct supertype *st;
312 int autof; /* 1 for normal, 2 for partitioned */
313 char *spare_group;
314 char *bitmap_file;
315 int bitmap_fd;
316
317 char *container; /* /dev/whatever name of container, or
318 * uuid of container. You would expect
319 * this to be the 'devname' or UUID
320 * of some other entry.
321 */
322 char *member; /* subarray within a container */
323
324 struct mddev_ident_s *next;
325 union {
326 /* fields needed by different users of this structure */
327 int assembled; /* set when assembly succeeds */
328 };
329 } *mddev_ident_t;
330
331 /* List of device names - wildcards expanded */
332 typedef struct mddev_dev_s {
333 char *devname;
334 char disposition; /* 'a' for add, 'r' for remove, 'f' for fail.
335 * Not set for names read from .config
336 */
337 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
338 char re_add;
339 char used; /* set when used */
340 struct mdinfo *content; /* If devname is a container, this might list
341 * the remaining member arrays. */
342 struct mddev_dev_s *next;
343 } *mddev_dev_t;
344
345 typedef struct mapping {
346 char *name;
347 int num;
348 } mapping_t;
349
350
351 struct mdstat_ent {
352 char *dev;
353 int devnum;
354 int active;
355 char *level;
356 char *pattern; /* U or up, _ for down */
357 int percent; /* -1 if no resync */
358 int resync; /* 1 if resync, 0 if recovery */
359 int devcnt;
360 int raid_disks;
361 int chunk_size;
362 char * metadata_version;
363 struct dev_member {
364 char *name;
365 struct dev_member *next;
366 } *members;
367 struct mdstat_ent *next;
368 };
369
370 extern struct mdstat_ent *mdstat_read(int hold, int start);
371 extern void free_mdstat(struct mdstat_ent *ms);
372 extern void mdstat_wait(int seconds);
373 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
374 extern int mddev_busy(int devnum);
375 extern struct mdstat_ent *mdstat_by_component(char *name);
376
377 struct map_ent {
378 struct map_ent *next;
379 int devnum;
380 char metadata[20];
381 int uuid[4];
382 int bad;
383 char *path;
384 };
385 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
386 int uuid[4], char *path);
387 extern void map_remove(struct map_ent **map, int devnum);
388 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
389 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
390 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
391 extern void map_read(struct map_ent **melp);
392 extern int map_write(struct map_ent *mel);
393 extern void map_delete(struct map_ent **mapp, int devnum);
394 extern void map_free(struct map_ent *map);
395 extern void map_add(struct map_ent **melp,
396 int devnum, char *metadata, int uuid[4], char *path);
397 extern int map_lock(struct map_ent **melp);
398 extern void map_unlock(struct map_ent **melp);
399
400 /* various details can be requested */
401 enum sysfs_read_flags {
402 GET_LEVEL = (1 << 0),
403 GET_LAYOUT = (1 << 1),
404 GET_COMPONENT = (1 << 2),
405 GET_CHUNK = (1 << 3),
406 GET_CACHE = (1 << 4),
407 GET_MISMATCH = (1 << 5),
408 GET_VERSION = (1 << 6),
409 GET_DISKS = (1 << 7),
410 GET_DEGRADED = (1 << 8),
411 GET_SAFEMODE = (1 << 9),
412 GET_DEVS = (1 << 10), /* gets role, major, minor */
413 GET_OFFSET = (1 << 11),
414 GET_SIZE = (1 << 12),
415 GET_STATE = (1 << 13),
416 GET_ERROR = (1 << 14),
417 };
418
419 /* If fd >= 0, get the array it is open on,
420 * else use devnum. >=0 -> major9. <0.....
421 */
422 extern int sysfs_open(int devnum, char *devname, char *attr);
423 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
424 extern void sysfs_free(struct mdinfo *sra);
425 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
426 extern int sysfs_attr_match(const char *attr, const char *str);
427 extern int sysfs_match_word(const char *word, char **list);
428 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
429 char *name, char *val);
430 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
431 char *name, unsigned long long val);
432 extern int sysfs_uevent(struct mdinfo *sra, char *event);
433 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
434 char *name);
435 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
436 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
437 char *name, unsigned long long *val);
438 extern int sysfs_fd_get_str(int fd, char *val, int size);
439 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
440 char *name, char *val, int size);
441 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
442 extern int sysfs_set_array(struct mdinfo *info, int vers);
443 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
444 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
445 extern int sysfs_unique_holder(int devnum, long rdev);
446 extern int load_sys(char *path, char *buf);
447
448
449 extern int save_stripes(int *source, unsigned long long *offsets,
450 int raid_disks, int chunk_size, int level, int layout,
451 int nwrites, int *dest,
452 unsigned long long start, unsigned long long length,
453 char *buf);
454 extern int restore_stripes(int *dest, unsigned long long *offsets,
455 int raid_disks, int chunk_size, int level, int layout,
456 int source, unsigned long long read_offset,
457 unsigned long long start, unsigned long long length);
458
459 #ifndef Sendmail
460 #define Sendmail "/usr/lib/sendmail -t"
461 #endif
462
463 #define SYSLOG_FACILITY LOG_DAEMON
464
465 extern char *map_num(mapping_t *map, int num);
466 extern int map_name(mapping_t *map, char *name);
467 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
468
469 extern char *map_dev(int major, int minor, int create);
470
471 struct active_array;
472 struct metadata_update;
473
474 /* A superswitch provides entry point the a metadata handler.
475 *
476 * The super_switch primarily operates on some "metadata" that
477 * is accessed via the 'supertype'.
478 * This metadata has one of three possible sources.
479 * 1/ It is read from a single device. In this case it may not completely
480 * describe the array or arrays as some information might be on other
481 * devices.
482 * 2/ It is read from all devices in a container. In this case all
483 * information is present.
484 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
485 * be complete once enough ->add_to_super calls have completed.
486 *
487 * When creating an array inside a container, the metadata will be
488 * formed by a combination of 2 and 3. The metadata or the array is read,
489 * then new information is added.
490 *
491 * The metadata must sometimes have a concept of a 'current' array
492 * and a 'current' device.
493 * The 'current' array is set by init_super to be the newly created array,
494 * or is set by super_by_fd when it finds it is looking at an array inside
495 * a container.
496 *
497 * The 'current' device is either the device that the metadata was read from
498 * in case 1, or the last device added by add_to_super in case 3.
499 * Case 2 does not identify a 'current' device.
500 */
501 extern struct superswitch {
502
503 /* Used to report details of metadata read from a component
504 * device. ->load_super has been called.
505 */
506 void (*examine_super)(struct supertype *st, char *homehost);
507 void (*brief_examine_super)(struct supertype *st, int verbose);
508 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
509 void (*export_examine_super)(struct supertype *st);
510
511 /* Used to report details of an active array.
512 * ->load_super was possibly given a 'component' string.
513 */
514 void (*detail_super)(struct supertype *st, char *homehost);
515 void (*brief_detail_super)(struct supertype *st);
516 void (*export_detail_super)(struct supertype *st);
517
518 /* Optional: platform hardware / firmware details */
519 int (*detail_platform)(int verbose, int enumerate_only);
520
521 /* Used:
522 * to get uuid to storing in bitmap metadata
523 * and 'reshape' backup-data metadata
524 * To see if a device is being re-added to an array it was part of.
525 */
526 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
527
528 /* Extract generic details from metadata. This could be details about
529 * the container, or about an individual array within the container.
530 * The determination is made either by:
531 * load_super being given a 'component' string.
532 * validate_geometry determining what to create.
533 * The info includes both array information and device information.
534 * The particular device should be:
535 * The last device added by add_to_super
536 * The device the metadata was loaded from by load_super
537 */
538 void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
539
540 /* Check if the given metadata is flagged as belonging to "this"
541 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
542 */
543 int (*match_home)(struct supertype *st, char *homehost);
544
545 /* Make one of several generic modifications to metadata
546 * prior to assembly (or other times).
547 * sparc2.2 - first bug in early 0.90 metadata
548 * super-minor - change name of 0.90 metadata
549 * summaries - 'correct' any redundant data
550 * resync - mark array as dirty to trigger a resync.
551 * uuid - set new uuid - only 0.90 or 1.x
552 * name - change the name of the array (where supported)
553 * homehost - change which host this array is tied to.
554 * devicesize - If metadata is at start of device, change recorded
555 * device size to match actual device size
556 * byteorder - swap bytes for 0.90 metadata
557 *
558 * force-one - mark that device as uptodate, not old or failed.
559 * force-array - mark array as clean if it would not otherwise
560 * assemble
561 * assemble - not sure how this is different from force-one...
562 * linear-grow-new - add a new device to a linear array, but don't
563 * change the size: so superblock still matches
564 * linear-grow-update - now change the size of the array.
565 */
566 int (*update_super)(struct supertype *st, struct mdinfo *info,
567 char *update,
568 char *devname, int verbose,
569 int uuid_set, char *homehost);
570
571 /* Create new metadata for new array as described. This could
572 * be a new container, or an array in a pre-existing container.
573 * Also used to zero metadata prior to writing it to invalidate old
574 * metadata.
575 */
576 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
577 unsigned long long size, char *name,
578 char *homehost, int *uuid);
579
580 /* update the metadata to include new device, either at create or
581 * when hot-adding a spare.
582 */
583 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
584 int fd, char *devname);
585
586 /* Write metadata to one device when fixing problems or adding
587 * a new device.
588 */
589 int (*store_super)(struct supertype *st, int fd);
590
591 /* Write all metadata for this array.
592 */
593 int (*write_init_super)(struct supertype *st);
594 int (*compare_super)(struct supertype *st, struct supertype *tst);
595 int (*load_super)(struct supertype *st, int fd, char *devname);
596 struct supertype * (*match_metadata_desc)(char *arg);
597 __u64 (*avail_size)(struct supertype *st, __u64 size);
598 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
599 int delay, int write_behind,
600 unsigned long long size, int may_change, int major);
601 void (*locate_bitmap)(struct supertype *st, int fd);
602 int (*write_bitmap)(struct supertype *st, int fd);
603 void (*free_super)(struct supertype *st);
604
605 /* validate_geometry is called with an st returned by
606 * match_metadata_desc.
607 * It should check that the geometry described in compatible with
608 * the metadata type. It will be called repeatedly as devices
609 * added to validate changing size and new devices. If there are
610 * inter-device dependencies, it should record sufficient details
611 * so these can be validated.
612 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
613 */
614 int (*validate_geometry)(struct supertype *st, int level, int layout,
615 int raiddisks,
616 int chunk, unsigned long long size,
617 char *subdev, unsigned long long *freesize,
618 int verbose);
619
620 struct mdinfo *(*container_content)(struct supertype *st);
621 /* Allow a metadata handler to override mdadm's default layouts */
622 int (*default_layout)(int level); /* optional */
623 /* query the supertype for default chunk size */
624 int (*default_chunk)(struct supertype *st); /* optional */
625 /* Permit subarray's to be deleted from inactive containers */
626 int (*kill_subarray)(struct supertype *st); /* optional */
627 /* Permit subarray's to be modified */
628 int (*update_subarray)(struct supertype *st, char *update, mddev_ident_t ident); /* optional */
629
630 /* for mdmon */
631 int (*open_new)(struct supertype *c, struct active_array *a,
632 char *inst);
633
634 /* Tell the metadata handler the current state of the array.
635 * This covers whether it is known to be consistent (no pending writes)
636 * and how far along a resync is known to have progressed
637 * (in a->resync_start).
638 * resync status is really irrelevant if the array is not consistent,
639 * but some metadata (DDF!) have a place to record the distinction.
640 * If 'consistent' is '2', then the array can mark it dirty if a
641 * resync/recovery/whatever is required, or leave it clean if not.
642 * Return value is 0 dirty (not consistent) and 1 if clean.
643 * it is only really important if consistent is passed in as '2'.
644 */
645 int (*set_array_state)(struct active_array *a, int consistent);
646
647 /* When the state of a device might have changed, we call set_disk to
648 * tell the metadata what the current state is.
649 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
650 * transitions.
651 * set_disk might be called when the state of the particular disk has
652 * not in fact changed.
653 */
654 void (*set_disk)(struct active_array *a, int n, int state);
655 void (*sync_metadata)(struct supertype *st);
656 void (*process_update)(struct supertype *st,
657 struct metadata_update *update);
658 void (*prepare_update)(struct supertype *st,
659 struct metadata_update *update);
660
661 /* activate_spare will check if the array is degraded and, if it
662 * is, try to find some spare space in the container.
663 * On success, it add appropriate updates (For process_update) to
664 * to the 'updates' list and returns a list of 'mdinfo' identifying
665 * the device, or devices as there might be multiple missing
666 * devices and multiple spares available.
667 */
668 struct mdinfo *(*activate_spare)(struct active_array *a,
669 struct metadata_update **updates);
670
671 int swapuuid; /* true if uuid is bigending rather than hostendian */
672 int external;
673 const char *name; /* canonical metadata name */
674 } *superlist[];
675
676 extern struct superswitch super0, super1;
677 extern struct superswitch super_imsm, super_ddf;
678 extern struct superswitch mbr, gpt;
679
680 struct metadata_update {
681 int len;
682 char *buf;
683 void *space; /* allocated space that monitor will use */
684 struct metadata_update *next;
685 };
686
687 /* A supertype holds a particular collection of metadata.
688 * It identifies the metadata type by the superswitch, and the particular
689 * sub-version of that metadata type.
690 * metadata read in or created is stored in 'sb' and 'info'.
691 * There are also fields used by mdmon to track containers.
692 *
693 * A supertype may refer to:
694 * Just an array, possibly in a container
695 * A container, not identifying any particular array
696 * Info read from just one device, not yet fully describing the array/container.
697 *
698 *
699 * A supertype is created by:
700 * super_by_fd
701 * guess_super
702 * dup_super
703 */
704 struct supertype {
705 struct superswitch *ss;
706 int minor_version;
707 int max_devs;
708 int container_dev; /* devnum of container */
709 char subarray[32]; /* name of array inside container */
710 void *sb;
711 void *info;
712 int loaded_container; /* Set if load_super found a container,
713 * not just one device */
714
715 struct metadata_update *updates;
716 struct metadata_update **update_tail;
717
718 /* extra stuff used by mdmon */
719 struct active_array *arrays;
720 int sock; /* listen to external programs */
721 int devnum;
722 char *devname; /* e.g. md0. This appears in metadata_verison:
723 * external:/md0/12
724 */
725 int devcnt;
726
727 struct mdinfo *devs;
728
729 };
730
731 extern struct supertype *super_by_fd(int fd);
732 extern struct supertype *guess_super(int fd);
733 extern struct supertype *dup_super(struct supertype *st);
734 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
735 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
736 mdu_disk_info_t *disk);
737 void wait_for(char *dev, int fd);
738
739 /*
740 * Data structures for policy management.
741 * Each device can have a policy structure that lists
742 * various name/value pairs each possibly with a metadata associated.
743 * The policy list is sorted by name/value/metadata
744 */
745 struct dev_policy {
746 struct dev_policy *next;
747 char *name; /* None of these strings are allocated. They are
748 * all just references to strings which are known
749 * to exist elsewhere.
750 * name and metadata can be compared by address equality.
751 */
752 const char *metadata;
753 char *value;
754 };
755
756 extern char pol_act[], pol_domain[], pol_metadata[];
757
758 /* iterate over the sublist starting at list, having the same
759 * 'name' as 'list', and matching the given metadata (Where
760 * NULL matches anything
761 */
762 #define pol_for_each(item, list, _metadata) \
763 for (item = list; \
764 item && item->name == list->name; \
765 item = item->next) \
766 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
767 ; else
768
769 /*
770 * policy records read from mdadm are largely just name-value pairs.
771 * The names are constants, not strdupped
772 */
773 struct pol_rule {
774 struct pol_rule *next;
775 char *type; /* rule_policy or rule_part */
776 struct rule {
777 struct rule *next;
778 char *name;
779 char *value;
780 char *dups; /* duplicates of 'value' with a partNN appended */
781 } *rule;
782 };
783
784 extern char rule_policy[], rule_part[];
785 extern char rule_path[], rule_type[];
786 extern char type_part[], type_disk[];
787
788 extern void policyline(char *line, char *type);
789 extern void policy_free(void);
790
791 extern struct dev_policy *disk_policy(struct mdinfo *disk);
792 extern void dev_policy_free(struct dev_policy *p);
793
794 extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
795 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
796
797 enum policy_action {
798 act_default,
799 act_include,
800 act_re_add,
801 act_spare,
802 act_force_spare,
803 act_err
804 };
805
806 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
807 enum policy_action want);
808 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
809 enum policy_action want);
810
811 struct domainlist {
812 struct domainlist *next;
813 char *dom;
814 };
815
816 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
817 const char *metadata);
818 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
819 const char *metadata);
820 extern void domain_free(struct domainlist *dl);
821 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
822 const char *metadata);
823
824 #if __GNUC__ < 3
825 struct stat64;
826 #endif
827
828 #define HAVE_NFTW we assume
829 #define HAVE_FTW
830
831 #ifdef __UCLIBC__
832 # include <features.h>
833 # ifndef __UCLIBC_HAS_LFS__
834 # define lseek64 lseek
835 # endif
836 # ifndef __UCLIBC_HAS_FTW__
837 # undef HAVE_FTW
838 # undef HAVE_NFTW
839 # endif
840 #endif
841
842 #ifdef __dietlibc__
843 # undef HAVE_NFTW
844 #endif
845
846 #if defined(__KLIBC__)
847 # undef HAVE_NFTW
848 # undef HAVE_FTW
849 #endif
850
851 #ifndef HAVE_NFTW
852 # define FTW_PHYS 1
853 # ifndef HAVE_FTW
854 struct FTW {};
855 # endif
856 #endif
857
858 #ifdef HAVE_FTW
859 # include <ftw.h>
860 #endif
861
862 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
863
864
865 extern int Manage_ro(char *devname, int fd, int readonly);
866 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
867 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
868 extern int Manage_subdevs(char *devname, int fd,
869 mddev_dev_t devlist, int verbose, int test);
870 extern int autodetect(void);
871 extern int Grow_Add_device(char *devname, int fd, char *newdev);
872 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
873 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
874 long long size,
875 int level, char *layout_str, int chunksize, int raid_disks);
876 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
877 int *fdlist, int cnt, char *backup_file, int verbose);
878 extern int Grow_continue(int mdfd, struct supertype *st,
879 struct mdinfo *info, char *backup_file);
880
881 extern int Assemble(struct supertype *st, char *mddev,
882 mddev_ident_t ident,
883 mddev_dev_t devlist, char *backup_file,
884 int readonly, int runstop,
885 char *update, char *homehost, int require_homehost,
886 int verbose, int force);
887
888 extern int Build(char *mddev, int chunk, int level, int layout,
889 int raiddisks, mddev_dev_t devlist, int assume_clean,
890 char *bitmap_file, int bitmap_chunk, int write_behind,
891 int delay, int verbose, int autof, unsigned long long size);
892
893
894 extern int Create(struct supertype *st, char *mddev,
895 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
896 char *name, char *homehost, int *uuid,
897 int subdevs, mddev_dev_t devlist,
898 int runstop, int verbose, int force, int assume_clean,
899 char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
900
901 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
902 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
903 extern int Query(char *dev);
904 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
905 int SparcAdjust, struct supertype *forcest, char *homehost);
906 extern int Monitor(mddev_dev_t devlist,
907 char *mailaddr, char *alert_cmd,
908 int period, int daemonise, int scan, int oneshot,
909 int dosyslog, int test, char *pidfile, int increments);
910
911 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
912 extern int Kill_subarray(char *dev, char *subarray, int quiet);
913 extern int Update_subarray(char *dev, char *subarray, char *update, mddev_ident_t ident, int quiet);
914 extern int Wait(char *dev);
915 extern int WaitClean(char *dev, int sock, int verbose);
916
917 extern int Incremental(char *devname, int verbose, int runstop,
918 struct supertype *st, char *homehost, int require_homehost,
919 int autof);
920 extern int Incremental_container(struct supertype *st, char *devname,
921 int verbose, int runstop, int autof,
922 int trustworthy);
923 extern void RebuildMap(void);
924 extern int IncrementalScan(int verbose);
925 extern int IncrementalRemove(char *devname, int verbose);
926 extern int CreateBitmap(char *filename, int force, char uuid[16],
927 unsigned long chunksize, unsigned long daemon_sleep,
928 unsigned long write_behind,
929 unsigned long long array_size,
930 int major);
931 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
932 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
933 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
934
935 extern int md_get_version(int fd);
936 extern int get_linux_version(void);
937 extern long long parse_size(char *size);
938 extern int parse_uuid(char *str, int uuid[4]);
939 extern int parse_layout_10(char *layout);
940 extern int parse_layout_faulty(char *layout);
941 extern int check_ext2(int fd, char *name);
942 extern int check_reiser(int fd, char *name);
943 extern int check_raid(int fd, char *name);
944 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
945
946 extern int get_mdp_major(void);
947 extern int dev_open(char *dev, int flags);
948 extern int open_dev(int devnum);
949 extern int open_dev_excl(int devnum);
950 extern int is_standard(char *dev, int *nump);
951 extern int same_dev(char *one, char *two);
952
953 extern int parse_auto(char *str, char *msg, int config);
954 extern mddev_ident_t conf_get_ident(char *dev);
955 extern mddev_dev_t conf_get_devs(void);
956 extern int conf_test_dev(char *devname);
957 extern int conf_test_metadata(const char *version, int is_homehost);
958 extern struct createinfo *conf_get_create_info(void);
959 extern void set_conffile(char *file);
960 extern char *conf_get_mailaddr(void);
961 extern char *conf_get_mailfrom(void);
962 extern char *conf_get_program(void);
963 extern char *conf_get_homehost(int *require_homehostp);
964 extern char *conf_line(FILE *file);
965 extern char *conf_word(FILE *file, int allow_key);
966 extern int conf_name_is_free(char *name);
967 extern int devname_matches(char *name, char *match);
968 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
969
970 extern void free_line(char *line);
971 extern int match_oneof(char *devices, char *devname);
972 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
973 extern const int uuid_match_any[4];
974 extern int same_uuid(int a[4], int b[4], int swapuuid);
975 extern void copy_uuid(void *a, int b[4], int swapuuid);
976 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
977 extern char *fname_from_uuid(struct supertype *st,
978 struct mdinfo *info, char *buf, char sep);
979 extern unsigned long calc_csum(void *super, int bytes);
980 extern int enough(int level, int raid_disks, int layout, int clean,
981 char *avail, int avail_disks);
982 extern int ask(char *mesg);
983 extern unsigned long long get_component_size(int fd);
984 extern void remove_partitions(int fd);
985 extern int test_partition(int fd);
986 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
987 int chunksize, unsigned long long devsize);
988 extern int flush_metadata_updates(struct supertype *st);
989 extern void append_metadata_update(struct supertype *st, void *buf, int len);
990 extern int assemble_container_content(struct supertype *st, int mdfd,
991 struct mdinfo *content, int runstop,
992 char *chosen_name, int verbose);
993
994 extern int add_disk(int mdfd, struct supertype *st,
995 struct mdinfo *sra, struct mdinfo *info);
996 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
997 unsigned long long min_recovery_start(struct mdinfo *array);
998
999 extern char *human_size(long long bytes);
1000 extern char *human_size_brief(long long bytes);
1001 extern void print_r10_layout(int layout);
1002
1003 #define NoMdDev (1<<23)
1004 extern int find_free_devnum(int use_partitions);
1005
1006 extern void put_md_name(char *name);
1007 extern char *get_md_name(int dev);
1008
1009 extern char DefaultConfFile[];
1010
1011 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1012 char *chosen);
1013 /* values for 'trustworthy' */
1014 #define LOCAL 1
1015 #define LOCAL_ANY 10
1016 #define FOREIGN 2
1017 #define METADATA 3
1018 extern int open_mddev(char *dev, int report_errors);
1019 extern int open_container(int fd);
1020 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1021 extern int is_subarray_active(char *subarray, char *devname);
1022 int is_container_active(char *devname);
1023 extern int open_subarray(char *dev, struct supertype *st, int quiet);
1024 extern struct superswitch *version_to_superswitch(char *vers);
1025
1026 extern int mdmon_running(int devnum);
1027 extern int mdmon_pid(int devnum);
1028 extern int check_env(char *name);
1029 extern __u32 random32(void);
1030 extern int start_mdmon(int devnum);
1031
1032 extern char *devnum2devname(int num);
1033 extern int devname2devnum(char *name);
1034 extern int stat2devnum(struct stat *st);
1035 extern int fd2devnum(int fd);
1036
1037 static inline int dev2major(int d)
1038 {
1039 if (d >= 0)
1040 return MD_MAJOR;
1041 else
1042 return get_mdp_major();
1043 }
1044
1045 static inline int dev2minor(int d)
1046 {
1047 if (d >= 0)
1048 return d;
1049 return (-1-d) << MdpMinorShift;
1050 }
1051
1052 static inline int ROUND_UP(int a, int base)
1053 {
1054 return ((a+base-1)/base)*base;
1055 }
1056
1057 static inline int is_subarray(char *vers)
1058 {
1059 /* The version string for a 'subarray' (an array in a container)
1060 * is
1061 * /containername/componentname for normal read-write arrays
1062 * -containername/componentname for read-only arrays.
1063 * containername is e.g. md0, md_d1
1064 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1065 */
1066 return (*vers == '/' || *vers == '-');
1067 }
1068
1069 #ifdef DEBUG
1070 #define dprintf(fmt, arg...) \
1071 fprintf(stderr, fmt, ##arg)
1072 #else
1073 #define dprintf(fmt, arg...) \
1074 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1075 #endif
1076 #include <assert.h>
1077 #include <stdarg.h>
1078 static inline int xasprintf(char **strp, const char *fmt, ...) {
1079 va_list ap;
1080 int ret;
1081 va_start(ap, fmt);
1082 ret = vasprintf(strp, fmt, ap);
1083 va_end(ap);
1084 assert(ret >= 0);
1085 return ret;
1086 }
1087
1088 #define LEVEL_MULTIPATH (-4)
1089 #define LEVEL_LINEAR (-1)
1090 #define LEVEL_FAULTY (-5)
1091
1092 /* kernel module doesn't know about these */
1093 #define LEVEL_CONTAINER (-100)
1094 #define LEVEL_UNSUPPORTED (-200)
1095
1096
1097 /* faulty stuff */
1098
1099 #define WriteTransient 0
1100 #define ReadTransient 1
1101 #define WritePersistent 2
1102 #define ReadPersistent 3
1103 #define WriteAll 4 /* doesn't go to device */
1104 #define ReadFixable 5
1105 #define Modes 6
1106
1107 #define ClearErrors 31
1108 #define ClearFaults 30
1109
1110 #define AllPersist 100 /* internal use only */
1111 #define NoPersist 101
1112
1113 #define ModeMask 0x1f
1114 #define ModeShift 5
1115
1116
1117 #ifdef __TINYC__
1118 #undef minor
1119 #undef major
1120 #undef makedev
1121 #define minor(x) ((x)&0xff)
1122 #define major(x) (((x)>>8)&0xff)
1123 #define makedev(M,m) (((M)<<8) | (m))
1124 #endif
1125
1126 /* for raid4/5/6 */
1127 #define ALGORITHM_LEFT_ASYMMETRIC 0
1128 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1129 #define ALGORITHM_LEFT_SYMMETRIC 2
1130 #define ALGORITHM_RIGHT_SYMMETRIC 3
1131
1132 /* Define non-rotating (raid4) algorithms. These allow
1133 * conversion of raid4 to raid5.
1134 */
1135 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1136 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1137
1138 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1139 * Firstly, the exact positioning of the parity block is slightly
1140 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1141 * of DDF.
1142 * Secondly, or order of datablocks over which the Q syndrome is computed
1143 * is different.
1144 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1145 * These layouts are from the DDFv1.2 spec.
1146 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1147 * leaves RLQ=3 as 'Vendor Specific'
1148 */
1149
1150 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1151 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1152 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1153
1154
1155 /* For every RAID5 algorithm we define a RAID6 algorithm
1156 * with exactly the same layout for data and parity, and
1157 * with the Q block always on the last device (N-1).
1158 * This allows trivial conversion from RAID5 to RAID6
1159 */
1160 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1161 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1162 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1163 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1164 #define ALGORITHM_PARITY_0_6 20
1165 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1166
1167 /* Define PATH_MAX in case we don't use glibc or standard library does
1168 * not have PATH_MAX defined. Assume max path length is 4K characters.
1169 */
1170 #ifndef PATH_MAX
1171 #define PATH_MAX 4096
1172 #endif
1173