]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
imsm: return correct uuid for volume in detail
[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 #ifdef __GLIBC__
29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
30 #elif !defined(lseek64)
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 <stdint.h>
39 #include <stdlib.h>
40 #include <time.h>
41 #include <sys/time.h>
42 #include <getopt.h>
43 #include <fcntl.h>
44 #include <stdio.h>
45 #include <errno.h>
46 #include <string.h>
47 #include <syslog.h>
48 /* Newer glibc requires sys/sysmacros.h directly for makedev() */
49 #include <sys/sysmacros.h>
50 #ifdef __dietlibc__
51 #include <strings.h>
52 /* dietlibc has deprecated random and srandom!! */
53 #define random rand
54 #define srandom srand
55 #endif
56
57 #ifdef NO_COROSYNC
58 #define CS_OK 1
59 typedef uint64_t cmap_handle_t;
60 #else
61 #include <corosync/cmap.h>
62 #endif
63
64 #ifndef NO_DLM
65 #include <libdlm.h>
66 #include <errno.h>
67 #else
68 #define LKF_NOQUEUE 0x00000001
69 #define LKM_PWMODE 4
70 #define EUNLOCK 0x10002
71
72 typedef void *dlm_lshandle_t;
73
74 struct dlm_lksb {
75 int sb_status;
76 uint32_t sb_lkid;
77 char sb_flags;
78 char *sb_lvbptr;
79 };
80 #endif
81
82 #include <linux/kdev_t.h>
83 /*#include <linux/fs.h> */
84 #include <sys/mount.h>
85 #include <asm/types.h>
86 #include <sys/ioctl.h>
87 #define MD_MAJOR 9
88 #define MdpMinorShift 6
89
90 #ifndef BLKGETSIZE64
91 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
92 #endif
93
94 #define DEFAULT_CHUNK 512
95 #define DEFAULT_BITMAP_CHUNK 4096
96 #define DEFAULT_BITMAP_DELAY 5
97 #define DEFAULT_MAX_WRITE_BEHIND 256
98
99 /* MAP_DIR should be somewhere that persists across the pivotroot
100 * from early boot to late boot.
101 * /run seems to have emerged as the best standard.
102 */
103 #ifndef MAP_DIR
104 #define MAP_DIR "/run/mdadm"
105 #endif /* MAP_DIR */
106 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
107 * want something other than the default of "map"
108 */
109 #ifndef MAP_FILE
110 #define MAP_FILE "map"
111 #endif /* MAP_FILE */
112 /* MDMON_DIR is where pid and socket files used for communicating
113 * with mdmon normally live. Best is /var/run/mdadm as
114 * mdmon is needed at early boot then it needs to write there prior
115 * to /var/run being mounted read/write, and it also then needs to
116 * persist beyond when /var/run is mounter read-only. So, to be
117 * safe, the default is somewhere that is read/write early in the
118 * boot process and stays up as long as possible during shutdown.
119 */
120 #ifndef MDMON_DIR
121 #define MDMON_DIR "/run/mdadm"
122 #endif /* MDMON_DIR */
123
124 /* FAILED_SLOTS is where to save files storing recent removal of array
125 * member in order to allow future reuse of disk inserted in the same
126 * slot for array recovery
127 */
128 #ifndef FAILED_SLOTS_DIR
129 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
130 #endif /* FAILED_SLOTS */
131
132 #include "md_u.h"
133 #include "md_p.h"
134 #include "bitmap.h"
135 #include "msg.h"
136
137 #include <endian.h>
138 /* Redhat don't like to #include <asm/byteorder.h>, and
139 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
140 * and there is no standard conversion function so... */
141 /* And dietlibc doesn't think byteswap is ok, so.. */
142 /* #include <byteswap.h> */
143 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
144 ((x) & 0xff00U) >> 8)
145 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
146 ((x) & 0xff000000U) >> 24 | \
147 ((x) & 0x0000ff00U) << 8 | \
148 ((x) & 0x00ff0000U) >> 8)
149 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
150 ((x) & 0xff00000000000000ULL) >> 56 | \
151 ((x) & 0x000000000000ff00ULL) << 40 | \
152 ((x) & 0x00ff000000000000ULL) >> 40 | \
153 ((x) & 0x0000000000ff0000ULL) << 24 | \
154 ((x) & 0x0000ff0000000000ULL) >> 24 | \
155 ((x) & 0x00000000ff000000ULL) << 8 | \
156 ((x) & 0x000000ff00000000ULL) >> 8)
157
158 #if !defined(__KLIBC__)
159 #if BYTE_ORDER == LITTLE_ENDIAN
160 #define __cpu_to_le16(_x) (unsigned int)(_x)
161 #define __cpu_to_le32(_x) (unsigned int)(_x)
162 #define __cpu_to_le64(_x) (unsigned long long)(_x)
163 #define __le16_to_cpu(_x) (unsigned int)(_x)
164 #define __le32_to_cpu(_x) (unsigned int)(_x)
165 #define __le64_to_cpu(_x) (unsigned long long)(_x)
166
167 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
168 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
169 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
170 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
171 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
172 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
173 #elif BYTE_ORDER == BIG_ENDIAN
174 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
175 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
176 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
177 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
178 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
179 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
180
181 #define __cpu_to_be16(_x) (unsigned int)(_x)
182 #define __cpu_to_be32(_x) (unsigned int)(_x)
183 #define __cpu_to_be64(_x) (unsigned long long)(_x)
184 #define __be16_to_cpu(_x) (unsigned int)(_x)
185 #define __be32_to_cpu(_x) (unsigned int)(_x)
186 #define __be64_to_cpu(_x) (unsigned long long)(_x)
187 #else
188 # error "unknown endianness."
189 #endif
190 #endif /* __KLIBC__ */
191
192 /*
193 * Partially stolen from include/linux/unaligned/packed_struct.h
194 */
195 struct __una_u16 { __u16 x; } __attribute__ ((packed));
196 struct __una_u32 { __u32 x; } __attribute__ ((packed));
197
198 static inline __u16 __get_unaligned16(const void *p)
199 {
200 const struct __una_u16 *ptr = (const struct __una_u16 *)p;
201 return ptr->x;
202 }
203
204 static inline __u32 __get_unaligned32(const void *p)
205 {
206 const struct __una_u32 *ptr = (const struct __una_u32 *)p;
207 return ptr->x;
208 }
209
210 static inline void __put_unaligned16(__u16 val, void *p)
211 {
212 struct __una_u16 *ptr = (struct __una_u16 *)p;
213 ptr->x = val;
214 }
215
216 static inline void __put_unaligned32(__u32 val, void *p)
217 {
218 struct __una_u32 *ptr = (struct __una_u32 *)p;
219 ptr->x = val;
220 }
221
222 /*
223 * Check at compile time that something is of a particular type.
224 * Always evaluates to 1 so you may use it easily in comparisons.
225 */
226
227 #define typecheck(type,x) \
228 ({ type __dummy; \
229 typeof(x) __dummy2; \
230 (void)(&__dummy == &__dummy2); \
231 1; \
232 })
233
234 /*
235 * These inlines deal with timer wrapping correctly.
236 *
237 * time_after(a,b) returns true if the time a is after time b.
238 */
239
240 #define time_after(a,b) \
241 (typecheck(unsigned int, a) && \
242 typecheck(unsigned int, b) && \
243 ((int)((b) - (a)) < 0))
244
245 #define time_before(a,b) time_after(b,a)
246
247 /*
248 * min()/max()/clamp() macros that also do
249 * strict type-checking.. See the
250 * "unnecessary" pointer comparison.
251 */
252 #define min(x, y) ({ \
253 typeof(x) _min1 = (x); \
254 typeof(y) _min2 = (y); \
255 (void) (&_min1 == &_min2); \
256 _min1 < _min2 ? _min1 : _min2; })
257
258 #define max(x, y) ({ \
259 typeof(x) _max1 = (x); \
260 typeof(y) _max2 = (y); \
261 (void) (&_max1 == &_max2); \
262 _max1 > _max2 ? _max1 : _max2; })
263
264 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
265
266 extern const char Name[];
267
268 struct md_bb_entry {
269 unsigned long long sector;
270 int length;
271 };
272
273 struct md_bb {
274 int supported;
275 int count;
276 struct md_bb_entry *entries;
277 };
278
279 /* general information that might be extracted from a superblock */
280 struct mdinfo {
281 mdu_array_info_t array;
282 mdu_disk_info_t disk;
283 __u64 events;
284 int uuid[4];
285 char name[33];
286 unsigned long long data_offset;
287 unsigned long long new_data_offset;
288 unsigned long long component_size; /* same as array.size, except in
289 * sectors and up to 64bits.
290 */
291 unsigned long long custom_array_size; /* size for non-default sized
292 * arrays (in sectors)
293 */
294 #define NO_RESHAPE 0
295 #define VOLUME_RESHAPE 1
296 #define CONTAINER_RESHAPE 2
297 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
298 int reshape_active;
299 unsigned long long reshape_progress;
300 int recovery_blocked; /* for external metadata it
301 * indicates that there is
302 * reshape in progress in
303 * container,
304 * for native metadata it is
305 * reshape_active field mirror
306 */
307 int journal_device_required;
308 int journal_clean;
309
310 enum {
311 CONSISTENCY_POLICY_UNKNOWN,
312 CONSISTENCY_POLICY_NONE,
313 CONSISTENCY_POLICY_RESYNC,
314 CONSISTENCY_POLICY_BITMAP,
315 CONSISTENCY_POLICY_JOURNAL,
316 CONSISTENCY_POLICY_PPL,
317 } consistency_policy;
318
319 /* During reshape we can sometimes change the data_offset to avoid
320 * over-writing still-valid data. We need to know if there is space.
321 * So getinfo_super will fill in space_before and space_after in sectors.
322 * data_offset can be increased or decreased by this amount.
323 */
324 unsigned long long space_before, space_after;
325 union {
326 unsigned long long resync_start; /* per-array resync position */
327 unsigned long long recovery_start; /* per-device rebuild position */
328 #define MaxSector (~0ULL) /* resync/recovery complete position */
329 };
330 long bitmap_offset; /* 0 == none, 1 == a file */
331 unsigned int ppl_size;
332 int ppl_offset;
333 unsigned long long ppl_sector;
334 unsigned long safe_mode_delay; /* ms delay to mark clean */
335 int new_level, delta_disks, new_layout, new_chunk;
336 int errors;
337 unsigned long cache_size; /* size of raid456 stripe cache*/
338 int mismatch_cnt;
339 char text_version[50];
340
341 int container_member; /* for assembling external-metatdata arrays
342 * This is to be used internally by metadata
343 * handler only */
344 int container_enough; /* flag external handlers can set to
345 * indicate that subarrays have not enough (-1),
346 * enough to start (0), or all expected disks (1) */
347 char sys_name[32];
348 struct mdinfo *devs;
349 struct mdinfo *next;
350
351 /* Device info for mdmon: */
352 int recovery_fd;
353 int state_fd;
354 int bb_fd;
355 int ubb_fd;
356 #define DS_FAULTY 1
357 #define DS_INSYNC 2
358 #define DS_WRITE_MOSTLY 4
359 #define DS_SPARE 8
360 #define DS_BLOCKED 16
361 #define DS_REMOVE 1024
362 #define DS_UNBLOCK 2048
363 int prev_state, curr_state, next_state;
364
365 /* info read from sysfs */
366 enum {
367 ARRAY_CLEAR,
368 ARRAY_INACTIVE,
369 ARRAY_SUSPENDED,
370 ARRAY_READONLY,
371 ARRAY_READ_AUTO,
372 ARRAY_CLEAN,
373 ARRAY_ACTIVE,
374 ARRAY_WRITE_PENDING,
375 ARRAY_ACTIVE_IDLE,
376 ARRAY_BROKEN,
377 ARRAY_UNKNOWN_STATE,
378 } array_state;
379 struct md_bb bb;
380 };
381
382 struct createinfo {
383 int uid;
384 int gid;
385 int autof;
386 int mode;
387 int symlinks;
388 int names;
389 int bblist;
390 struct supertype *supertype;
391 };
392
393 struct spare_criteria {
394 unsigned long long min_size;
395 unsigned int sector_size;
396 };
397
398 enum mode {
399 ASSEMBLE=1,
400 BUILD,
401 CREATE,
402 MANAGE,
403 MISC,
404 MONITOR,
405 GROW,
406 INCREMENTAL,
407 AUTODETECT,
408 mode_count
409 };
410
411 extern char short_options[];
412 extern char short_bitmap_options[];
413 extern char short_bitmap_auto_options[];
414 extern struct option long_options[];
415 extern char Version[], Usage[], Help[], OptionHelp[],
416 *mode_help[],
417 Help_create[], Help_build[], Help_assemble[], Help_grow[],
418 Help_incr[],
419 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
420
421 /* for option that don't have short equivilents, we assign arbitrary
422 * numbers later than any 'short' character option.
423 */
424 enum special_options {
425 AssumeClean = 300,
426 BitmapChunk,
427 WriteBehind,
428 ReAdd,
429 NoDegraded,
430 Sparc22,
431 BackupFile,
432 HomeHost,
433 AutoHomeHost,
434 Symlinks,
435 AutoDetect,
436 Waitclean,
437 DetailPlatform,
438 KillSubarray,
439 UpdateSubarray,
440 IncrementalPath,
441 NoSharing,
442 HelpOptions,
443 Brief,
444 NoDevices,
445 ManageOpt,
446 Add,
447 AddSpare,
448 AddJournal,
449 Remove,
450 Fail,
451 Replace,
452 With,
453 MiscOpt,
454 WaitOpt,
455 ConfigFile,
456 ChunkSize,
457 WriteMostly,
458 FailFast,
459 NoFailFast,
460 Layout,
461 Auto,
462 Force,
463 SuperMinor,
464 EMail,
465 ProgramOpt,
466 Increment,
467 Fork,
468 Bitmap,
469 RebuildMapOpt,
470 InvalidBackup,
471 UdevRules,
472 FreezeReshape,
473 Continue,
474 OffRootOpt,
475 Prefer,
476 KillOpt,
477 DataOffset,
478 ExamineBB,
479 Dump,
480 Restore,
481 Action,
482 Nodes,
483 ClusterName,
484 ClusterConfirm,
485 WriteJournal,
486 ConsistencyPolicy,
487 };
488
489 enum prefix_standard {
490 JEDEC,
491 IEC
492 };
493
494 enum bitmap_update {
495 NoUpdate,
496 NameUpdate,
497 NodeNumUpdate,
498 };
499
500 enum flag_mode {
501 FlagDefault, FlagSet, FlagClear,
502 };
503
504 /* structures read from config file */
505 /* List of mddevice names and identifiers
506 * Identifiers can be:
507 * uuid=128-hex-uuid
508 * super-minor=decimal-minor-number-from-superblock
509 * devices=comma,separated,list,of,device,names,with,wildcards
510 *
511 * If multiple fields are present, the intersection of all matching
512 * devices is considered
513 */
514 #define UnSet (0xfffe)
515 struct mddev_ident {
516 char *devname;
517
518 int uuid_set;
519 int uuid[4];
520 char name[33];
521
522 int super_minor;
523
524 char *devices; /* comma separated list of device
525 * names with wild cards
526 */
527 int level;
528 int raid_disks;
529 int spare_disks;
530 struct supertype *st;
531 int autof; /* 1 for normal, 2 for partitioned */
532 char *spare_group;
533 char *bitmap_file;
534 int bitmap_fd;
535
536 char *container; /* /dev/whatever name of container, or
537 * uuid of container. You would expect
538 * this to be the 'devname' or UUID
539 * of some other entry.
540 */
541 char *member; /* subarray within a container */
542
543 struct mddev_ident *next;
544 union {
545 /* fields needed by different users of this structure */
546 int assembled; /* set when assembly succeeds */
547 };
548 };
549
550 struct context {
551 int readonly;
552 int runstop;
553 int verbose;
554 int brief;
555 int no_devices;
556 int force;
557 char *homehost;
558 int require_homehost;
559 char *prefer;
560 int export;
561 int test;
562 char *subarray;
563 char *update;
564 int scan;
565 int SparcAdjust;
566 int autof;
567 int delay;
568 int freeze_reshape;
569 char *backup_file;
570 int invalid_backup;
571 char *action;
572 int nodes;
573 char *homecluster;
574 };
575
576 struct shape {
577 int raiddisks;
578 int sparedisks;
579 int journaldisks;
580 int level;
581 int layout;
582 char *layout_str;
583 int chunk;
584 int bitmap_chunk;
585 char *bitmap_file;
586 int assume_clean;
587 int write_behind;
588 unsigned long long size;
589 int consistency_policy;
590 };
591
592 /* List of device names - wildcards expanded */
593 struct mddev_dev {
594 char *devname;
595 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
596 * 'A' for re_add.
597 * Not set for names read from .config
598 */
599 enum flag_mode writemostly;
600 enum flag_mode failfast;
601 int used; /* set when used */
602 long long data_offset;
603 struct mddev_dev *next;
604 };
605
606 typedef struct mapping {
607 char *name;
608 int num;
609 } mapping_t;
610
611 struct mdstat_ent {
612 char devnm[32];
613 int active;
614 char *level;
615 char *pattern; /* U for up, _ for down */
616 int percent; /* -1 if no resync */
617 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
618 int devcnt;
619 int raid_disks;
620 char * metadata_version;
621 struct dev_member {
622 char *name;
623 struct dev_member *next;
624 } *members;
625 struct mdstat_ent *next;
626 };
627
628 extern struct mdstat_ent *mdstat_read(int hold, int start);
629 extern void mdstat_close(void);
630 extern void free_mdstat(struct mdstat_ent *ms);
631 extern void mdstat_wait(int seconds);
632 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
633 extern int mddev_busy(char *devnm);
634 extern struct mdstat_ent *mdstat_by_component(char *name);
635 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
636
637 struct map_ent {
638 struct map_ent *next;
639 char devnm[32];
640 char metadata[20];
641 int uuid[4];
642 int bad;
643 char *path;
644 };
645 extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
646 int uuid[4], char *path);
647 extern void map_remove(struct map_ent **map, char *devnm);
648 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
649 extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
650 extern void map_free(struct map_ent *map);
651 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
652 extern void map_read(struct map_ent **melp);
653 extern int map_write(struct map_ent *mel);
654 extern void map_delete(struct map_ent **mapp, char *devnm);
655 extern void map_add(struct map_ent **melp,
656 char *devnm, char *metadata, int uuid[4], char *path);
657 extern int map_lock(struct map_ent **melp);
658 extern void map_unlock(struct map_ent **melp);
659 extern void map_fork(void);
660
661 /* various details can be requested */
662 enum sysfs_read_flags {
663 GET_LEVEL = (1 << 0),
664 GET_LAYOUT = (1 << 1),
665 GET_COMPONENT = (1 << 2),
666 GET_CHUNK = (1 << 3),
667 GET_CACHE = (1 << 4),
668 GET_MISMATCH = (1 << 5),
669 GET_VERSION = (1 << 6),
670 GET_DISKS = (1 << 7),
671 GET_SAFEMODE = (1 << 9),
672 GET_BITMAP_LOCATION = (1 << 10),
673
674 GET_DEVS = (1 << 20), /* gets role, major, minor */
675 GET_OFFSET = (1 << 21),
676 GET_SIZE = (1 << 22),
677 GET_STATE = (1 << 23),
678 GET_ERROR = (1 << 24),
679 GET_ARRAY_STATE = (1 << 25),
680 GET_CONSISTENCY_POLICY = (1 << 26),
681 GET_DEVS_ALL = (1 << 27),
682 };
683
684 /* If fd >= 0, get the array it is open on,
685 * else use devnm.
686 */
687 extern int sysfs_open(char *devnm, char *devname, char *attr);
688 extern int sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
689 extern void sysfs_init_dev(struct mdinfo *mdi, dev_t devid);
690 extern void sysfs_free(struct mdinfo *sra);
691 extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
692 extern int sysfs_attr_match(const char *attr, const char *str);
693 extern int sysfs_match_word(const char *word, char **list);
694 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
695 char *name, char *val);
696 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
697 char *name, unsigned long long val);
698 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
699 char *name, long long val);
700 extern int sysfs_uevent(struct mdinfo *sra, char *event);
701 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
702 char *name);
703 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
704 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
705 char *name, unsigned long long *val);
706 extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
707 extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
708 char *name, unsigned long long *v1, unsigned long long *v2);
709 extern int sysfs_fd_get_str(int fd, char *val, int size);
710 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
711 char *name);
712 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
713 char *name, char *val, int size);
714 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
715 extern int sysfs_set_array(struct mdinfo *info, int vers);
716 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
717 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
718 extern int sysfs_unique_holder(char *devnm, long rdev);
719 extern int sysfs_freeze_array(struct mdinfo *sra);
720 extern int sysfs_wait(int fd, int *msec);
721 extern int load_sys(char *path, char *buf, int len);
722 extern int zero_disk_range(int fd, unsigned long long sector, size_t count);
723 extern int reshape_prepare_fdlist(char *devname,
724 struct mdinfo *sra,
725 int raid_disks,
726 int nrdisks,
727 unsigned long blocks,
728 char *backup_file,
729 int *fdlist,
730 unsigned long long *offsets);
731 extern void reshape_free_fdlist(int *fdlist,
732 unsigned long long *offsets,
733 int size);
734 extern int reshape_open_backup_file(char *backup,
735 int fd,
736 char *devname,
737 long blocks,
738 int *fdlist,
739 unsigned long long *offsets,
740 char *sysfs_name,
741 int restart);
742 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
743 unsigned int ndata, unsigned int odata);
744 extern char *locate_backup(char *name);
745 extern char *make_backup(char *name);
746
747 extern int save_stripes(int *source, unsigned long long *offsets,
748 int raid_disks, int chunk_size, int level, int layout,
749 int nwrites, int *dest,
750 unsigned long long start, unsigned long long length,
751 char *buf);
752 extern int restore_stripes(int *dest, unsigned long long *offsets,
753 int raid_disks, int chunk_size, int level, int layout,
754 int source, unsigned long long read_offset,
755 unsigned long long start, unsigned long long length,
756 char *src_buf);
757
758 #ifndef Sendmail
759 #define Sendmail "/usr/lib/sendmail -t"
760 #endif
761
762 #define SYSLOG_FACILITY LOG_DAEMON
763
764 extern char *map_num(mapping_t *map, int num);
765 extern int map_name(mapping_t *map, char *name);
766 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
767 extern mapping_t consistency_policies[], sysfs_array_states[];
768
769 extern char *map_dev_preferred(int major, int minor, int create,
770 char *prefer);
771 static inline char *map_dev(int major, int minor, int create)
772 {
773 return map_dev_preferred(major, minor, create, NULL);
774 }
775
776 struct active_array;
777 struct metadata_update;
778
779 /* 'struct reshape' records the intermediate states of
780 * a general reshape.
781 * The starting geometry is converted to the 'before' geometry
782 * by at most an atomic level change. They could be the same.
783 * Similarly the 'after' geometry is converted to the final
784 * geometry by at most a level change.
785 * Note that 'before' and 'after' must have the same level.
786 * 'blocks' is the minimum number of sectors for a reshape unit.
787 * This will be a multiple of the stripe size in each of the
788 * 'before' and 'after' geometries.
789 * If 'blocks' is 0, no restriping is necessary.
790 * 'min_offset_change' is the minimum change to data_offset to
791 * allow the reshape to happen. It is at least the larger of
792 * the old and new chunk sizes, and typically the same as 'blocks'
793 * divided by number of data disks.
794 */
795 struct reshape {
796 int level;
797 int parity; /* number of parity blocks/devices */
798 struct {
799 int layout;
800 int data_disks;
801 } before, after;
802 unsigned long long backup_blocks;
803 unsigned long long min_offset_change;
804 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
805 unsigned long long new_size; /* New size of array in sectors */
806 };
807
808 /* A superswitch provides entry point to a metadata handler.
809 *
810 * The superswitch primarily operates on some "metadata" that
811 * is accessed via the 'supertype'.
812 * This metadata has one of three possible sources.
813 * 1/ It is read from a single device. In this case it may not completely
814 * describe the array or arrays as some information might be on other
815 * devices.
816 * 2/ It is read from all devices in a container. In this case all
817 * information is present.
818 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
819 * be complete once enough ->add_to_super calls have completed.
820 *
821 * When creating an array inside a container, the metadata will be
822 * formed by a combination of 2 and 3. The metadata or the array is read,
823 * then new information is added.
824 *
825 * The metadata must sometimes have a concept of a 'current' array
826 * and a 'current' device.
827 * The 'current' array is set by init_super to be the newly created array,
828 * or is set by super_by_fd when it finds it is looking at an array inside
829 * a container.
830 *
831 * The 'current' device is either the device that the metadata was read from
832 * in case 1, or the last device added by add_to_super in case 3.
833 * Case 2 does not identify a 'current' device.
834 */
835 extern struct superswitch {
836
837 /* Used to report details of metadata read from a component
838 * device. ->load_super has been called.
839 */
840 void (*examine_super)(struct supertype *st, char *homehost);
841 void (*brief_examine_super)(struct supertype *st, int verbose);
842 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
843 void (*export_examine_super)(struct supertype *st);
844 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
845 int (*copy_metadata)(struct supertype *st, int from, int to);
846
847 /* Used to report details of an active array.
848 * ->load_super was possibly given a 'component' string.
849 */
850 void (*detail_super)(struct supertype *st, char *homehost,
851 char *subarray);
852 void (*brief_detail_super)(struct supertype *st, char *subarray);
853 void (*export_detail_super)(struct supertype *st);
854
855 /* Optional: platform hardware / firmware details */
856 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
857 int (*export_detail_platform)(int verbose, char *controller_path);
858
859 /* Used:
860 * to get uuid to storing in bitmap metadata
861 * and 'reshape' backup-data metadata
862 * To see if a device is being re-added to an array it was part of.
863 */
864 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
865
866 /* Extract generic details from metadata. This could be details about
867 * the container, or about an individual array within the container.
868 * The determination is made either by:
869 * load_super being given a 'component' string.
870 * validate_geometry determining what to create.
871 * The info includes both array information and device information.
872 * The particular device should be:
873 * The last device added by add_to_super
874 * The device the metadata was loaded from by load_super
875 * If 'map' is present, then it is an array raid_disks long
876 * (raid_disk must already be set and correct) and it is filled
877 * with 1 for slots that are thought to be active and 0 for slots which
878 * appear to be failed/missing.
879 * *info is zeroed out before data is added.
880 */
881 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
882 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
883 /* Check if the given metadata is flagged as belonging to "this"
884 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
885 */
886 int (*match_home)(struct supertype *st, char *homehost);
887
888 /* Make one of several generic modifications to metadata
889 * prior to assembly (or other times).
890 * sparc2.2 - first bug in early 0.90 metadata
891 * super-minor - change name of 0.90 metadata
892 * summaries - 'correct' any redundant data
893 * resync - mark array as dirty to trigger a resync.
894 * uuid - set new uuid - only 0.90 or 1.x
895 * name - change the name of the array (where supported)
896 * homehost - change which host this array is tied to.
897 * devicesize - If metadata is at start of device, change recorded
898 * device size to match actual device size
899 * byteorder - swap bytes for 0.90 metadata
900 *
901 * force-one - mark that device as uptodate, not old or failed.
902 * force-array - mark array as clean if it would not otherwise
903 * assemble
904 * assemble - not sure how this is different from force-one...
905 * linear-grow-new - add a new device to a linear array, but don't
906 * change the size: so superblock still matches
907 * linear-grow-update - now change the size of the array.
908 * writemostly - set the WriteMostly1 bit in the superblock devflags
909 * readwrite - clear the WriteMostly1 bit in the superblock devflags
910 * failfast - set the FailFast1 bit in the superblock
911 * nofailfast - clear the FailFast1 bit
912 * no-bitmap - clear any record that a bitmap is present.
913 * bbl - add a bad-block-log if possible
914 * no-bbl - remove any bad-block-log is it is empty.
915 * force-no-bbl - remove any bad-block-log even if empty.
916 * revert-reshape - If a reshape is in progress, modify metadata so
917 * it will resume going in the opposite direction.
918 */
919 int (*update_super)(struct supertype *st, struct mdinfo *info,
920 char *update,
921 char *devname, int verbose,
922 int uuid_set, char *homehost);
923
924 /* Create new metadata for new array as described. This could
925 * be a new container, or an array in a pre-existing container.
926 * Also used to zero metadata prior to writing it to invalidate old
927 * metadata.
928 */
929 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
930 struct shape *s, char *name,
931 char *homehost, int *uuid,
932 unsigned long long data_offset);
933
934 /* update the metadata to include new device, either at create or
935 * when hot-adding a spare.
936 */
937 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
938 int fd, char *devname,
939 unsigned long long data_offset);
940 /* update the metadata to delete a device,
941 * when hot-removing.
942 */
943 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
944
945 /* Write metadata to one device when fixing problems or adding
946 * a new device.
947 */
948 int (*store_super)(struct supertype *st, int fd);
949
950 /* Write all metadata for this array.
951 */
952 int (*write_init_super)(struct supertype *st);
953 /* Check if metadata read from one device is compatible with an array,
954 * used when assembling an array, or pseudo-assembling was with
955 * "--examine --brief"
956 * If "st" has not yet been loaded the superblock from, "tst" is
957 * moved in, otherwise the superblock in 'st' is compared with
958 * 'tst'.
959 */
960 int (*compare_super)(struct supertype *st, struct supertype *tst);
961 /* Load metadata from a single device. If 'devname' is not NULL
962 * print error messages as appropriate */
963 int (*load_super)(struct supertype *st, int fd, char *devname);
964 /* 'fd' is a 'container' md array - load array metadata from the
965 * whole container.
966 */
967 int (*load_container)(struct supertype *st, int fd, char *devname);
968 /* If 'arg' is a valid name of this metadata type, allocate and
969 * return a 'supertype' for the particular minor version */
970 struct supertype * (*match_metadata_desc)(char *arg);
971 /* If a device has the given size, and the data_offset has been
972 * requested - work out how much space is available for data.
973 * This involves adjusting for reserved space (e.g. bitmaps)
974 * and for any rounding.
975 * 'mdadm' only calls this for existing arrays where a possible
976 * spare is being added. However some super-handlers call it
977 * internally from validate_geometry when creating an array.
978 */
979 __u64 (*avail_size)(struct supertype *st, __u64 size,
980 unsigned long long data_offset);
981 /*
982 * Return spare criteria for array:
983 * - minimum disk size can be used in array;
984 * - sector size can be used in array.
985 * Return values: 0 - for success and -EINVAL on error.
986 */
987 int (*get_spare_criteria)(struct supertype *st,
988 struct spare_criteria *sc);
989 /* Find somewhere to put a bitmap - possibly auto-size it - and
990 * update the metadata to record this. The array may be newly
991 * created, in which case data_size may be updated, or it might
992 * already exist. Metadata handler can know if init_super
993 * has been called, but not write_init_super.
994 * 0: Success
995 * -Exxxx: On error
996 */
997 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
998 int delay, int write_behind,
999 unsigned long long size, int may_change, int major);
1000 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1001 * md-native format bitmap
1002 */
1003 int (*locate_bitmap)(struct supertype *st, int fd, int node_num);
1004 /* if add_internal_bitmap succeeded for existing array, this
1005 * writes it out.
1006 */
1007 int (*write_bitmap)(struct supertype *st, int fd, enum bitmap_update update);
1008 /* Free the superblock and any other allocated data */
1009 void (*free_super)(struct supertype *st);
1010
1011 /* validate_geometry is called with an st returned by
1012 * match_metadata_desc.
1013 * It should check that the geometry described is compatible with
1014 * the metadata type. It will be called repeatedly as devices
1015 * added to validate changing size and new devices. If there are
1016 * inter-device dependencies, it should record sufficient details
1017 * so these can be validated.
1018 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1019 * Return value is:
1020 * 1: everything is OK
1021 * 0: not OK for some reason - if 'verbose', then error was reported.
1022 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1023 * type, but array is not acceptable for some reason
1024 * message was reported even if verbose is 0.
1025 */
1026 int (*validate_geometry)(struct supertype *st, int level, int layout,
1027 int raiddisks,
1028 int *chunk, unsigned long long size,
1029 unsigned long long data_offset,
1030 char *subdev, unsigned long long *freesize,
1031 int consistency_policy, int verbose);
1032
1033 /* Return a linked list of 'mdinfo' structures for all arrays
1034 * in the container. For non-containers, it is like
1035 * getinfo_super with an allocated mdinfo.*/
1036 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
1037 /* query the supertype for default geometry */
1038 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
1039 /* Permit subarray's to be deleted from inactive containers */
1040 int (*kill_subarray)(struct supertype *st); /* optional */
1041 /* Permit subarray's to be modified */
1042 int (*update_subarray)(struct supertype *st, char *subarray,
1043 char *update, struct mddev_ident *ident); /* optional */
1044 /* Check if reshape is supported for this external format.
1045 * st is obtained from super_by_fd() where st->subarray[0] is
1046 * initialized to indicate if reshape is being performed at the
1047 * container or subarray level
1048 */
1049 #define APPLY_METADATA_CHANGES 1
1050 #define ROLLBACK_METADATA_CHANGES 0
1051
1052 int (*reshape_super)(struct supertype *st,
1053 unsigned long long size, int level,
1054 int layout, int chunksize, int raid_disks,
1055 int delta_disks, char *backup, char *dev,
1056 int direction,
1057 int verbose); /* optional */
1058 int (*manage_reshape)( /* optional */
1059 int afd, struct mdinfo *sra, struct reshape *reshape,
1060 struct supertype *st, unsigned long blocks,
1061 int *fds, unsigned long long *offsets,
1062 int dests, int *destfd, unsigned long long *destoffsets);
1063
1064 /* for mdmon */
1065 int (*open_new)(struct supertype *c, struct active_array *a,
1066 char *inst);
1067
1068 /* Tell the metadata handler the current state of the array.
1069 * This covers whether it is known to be consistent (no pending writes)
1070 * and how far along a resync is known to have progressed
1071 * (in a->resync_start).
1072 * resync status is really irrelevant if the array is not consistent,
1073 * but some metadata (DDF!) have a place to record the distinction.
1074 * If 'consistent' is '2', then the array can mark it dirty if a
1075 * resync/recovery/whatever is required, or leave it clean if not.
1076 * Return value is 0 dirty (not consistent) and 1 if clean.
1077 * it is only really important if consistent is passed in as '2'.
1078 */
1079 int (*set_array_state)(struct active_array *a, int consistent);
1080
1081 /* When the state of a device might have changed, we call set_disk to
1082 * tell the metadata what the current state is.
1083 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1084 * transitions.
1085 * set_disk might be called when the state of the particular disk has
1086 * not in fact changed.
1087 */
1088 void (*set_disk)(struct active_array *a, int n, int state);
1089 void (*sync_metadata)(struct supertype *st);
1090 void (*process_update)(struct supertype *st,
1091 struct metadata_update *update);
1092 /* Prepare updates allocates extra memory that might be
1093 * needed. If the update cannot be understood, return 0.
1094 */
1095 int (*prepare_update)(struct supertype *st,
1096 struct metadata_update *update);
1097
1098 /* activate_spare will check if the array is degraded and, if it
1099 * is, try to find some spare space in the container.
1100 * On success, it add appropriate updates (For process_update) to
1101 * to the 'updates' list and returns a list of 'mdinfo' identifying
1102 * the device, or devices as there might be multiple missing
1103 * devices and multiple spares available.
1104 */
1105 struct mdinfo *(*activate_spare)(struct active_array *a,
1106 struct metadata_update **updates);
1107 /*
1108 * Return statically allocated string that represents metadata specific
1109 * controller domain of the disk. The domain is used in disk domain
1110 * matching functions. Disks belong to the same domain if the they have
1111 * the same domain from mdadm.conf and belong the same metadata domain.
1112 * Returning NULL or not providing this handler means that metadata
1113 * does not distinguish the differences between disks that belong to
1114 * different controllers. They are in the domain specified by
1115 * configuration file (mdadm.conf).
1116 * In case when the metadata has the notion of domains based on disk
1117 * it shall return NULL for disks that do not belong to the controller
1118 * the supported domains. Such disks will form another domain and won't
1119 * be mixed with supported ones.
1120 */
1121 const char *(*get_disk_controller_domain)(const char *path);
1122
1123 /* for external backup area */
1124 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
1125
1126 /* validate container after assemble */
1127 int (*validate_container)(struct mdinfo *info);
1128
1129 /* write initial empty PPL on device */
1130 int (*write_init_ppl)(struct supertype *st, struct mdinfo *info, int fd);
1131
1132 /* validate ppl before assemble */
1133 int (*validate_ppl)(struct supertype *st, struct mdinfo *info,
1134 struct mdinfo *disk);
1135
1136 /* records new bad block in metadata */
1137 int (*record_bad_block)(struct active_array *a, int n,
1138 unsigned long long sector, int length);
1139
1140 /* clears bad block from metadata */
1141 int (*clear_bad_block)(struct active_array *a, int n,
1142 unsigned long long sector, int length);
1143
1144 /* get list of bad blocks from metadata */
1145 struct md_bb *(*get_bad_blocks)(struct active_array *a, int n);
1146
1147 int swapuuid; /* true if uuid is bigending rather than hostendian */
1148 int external;
1149 const char *name; /* canonical metadata name */
1150 } *superlist[];
1151
1152 extern struct superswitch super0, super1;
1153 extern struct superswitch super_imsm, super_ddf;
1154 extern struct superswitch mbr, gpt;
1155
1156 struct metadata_update {
1157 int len;
1158 char *buf;
1159 void *space; /* allocated space that monitor will use */
1160 void **space_list; /* list of allocated spaces that monitor can
1161 * use or that it returned.
1162 */
1163 struct metadata_update *next;
1164 };
1165
1166 /* A supertype holds a particular collection of metadata.
1167 * It identifies the metadata type by the superswitch, and the particular
1168 * sub-version of that metadata type.
1169 * metadata read in or created is stored in 'sb' and 'info'.
1170 * There are also fields used by mdmon to track containers.
1171 *
1172 * A supertype may refer to:
1173 * Just an array, possibly in a container
1174 * A container, not identifying any particular array
1175 * Info read from just one device, not yet fully describing the array/container.
1176 *
1177 *
1178 * A supertype is created by:
1179 * super_by_fd
1180 * guess_super
1181 * dup_super
1182 */
1183 struct supertype {
1184 struct superswitch *ss;
1185 int minor_version;
1186 int max_devs;
1187 char container_devnm[32]; /* devnm of container */
1188 void *sb;
1189 void *info;
1190 void *other; /* Hack used to convert v0.90 to v1.0 */
1191 unsigned long long devsize;
1192 unsigned long long data_offset; /* used by v1.x only */
1193 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1194 HW/firmware related incompatability to load metadata.
1195 Used when examining metadata to display content of disk
1196 when user has no hw/firmare compatible system.
1197 */
1198 struct metadata_update *updates;
1199 struct metadata_update **update_tail;
1200
1201 /* extra stuff used by mdmon */
1202 struct active_array *arrays;
1203 int sock; /* listen to external programs */
1204 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1205 * external:/md0/12
1206 */
1207 int devcnt;
1208 int retry_soon;
1209 int nodes;
1210 char *cluster_name;
1211
1212 struct mdinfo *devs;
1213
1214 };
1215
1216 extern struct supertype *super_by_fd(int fd, char **subarray);
1217 enum guess_types { guess_any, guess_array, guess_partitions };
1218 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1219 static inline struct supertype *guess_super(int fd) {
1220 return guess_super_type(fd, guess_any);
1221 }
1222 extern struct supertype *dup_super(struct supertype *st);
1223 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1224 extern int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep);
1225 extern int must_be_container(int fd);
1226 extern int dev_size_from_id(dev_t id, unsigned long long *size);
1227 extern int dev_sector_size_from_id(dev_t id, unsigned int *size);
1228 void wait_for(char *dev, int fd);
1229
1230 /*
1231 * Data structures for policy management.
1232 * Each device can have a policy structure that lists
1233 * various name/value pairs each possibly with a metadata associated.
1234 * The policy list is sorted by name/value/metadata
1235 */
1236 struct dev_policy {
1237 struct dev_policy *next;
1238 char *name; /* None of these strings are allocated. They are
1239 * all just references to strings which are known
1240 * to exist elsewhere.
1241 * name and metadata can be compared by address equality.
1242 */
1243 const char *metadata;
1244 const char *value;
1245 };
1246
1247 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1248
1249 /* iterate over the sublist starting at list, having the same
1250 * 'name' as 'list', and matching the given metadata (Where
1251 * NULL matches anything
1252 */
1253 #define pol_for_each(item, list, _metadata) \
1254 for (item = list; \
1255 item && item->name == list->name; \
1256 item = item->next) \
1257 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1258 ; else
1259
1260 /*
1261 * policy records read from mdadm are largely just name-value pairs.
1262 * The names are constants, not strdupped
1263 */
1264 struct pol_rule {
1265 struct pol_rule *next;
1266 char *type; /* rule_policy or rule_part */
1267 struct rule {
1268 struct rule *next;
1269 char *name;
1270 char *value;
1271 char *dups; /* duplicates of 'value' with a partNN appended */
1272 } *rule;
1273 };
1274
1275 extern char rule_policy[], rule_part[];
1276 extern char rule_path[], rule_type[];
1277 extern char type_part[], type_disk[];
1278
1279 extern void policyline(char *line, char *type);
1280 extern void policy_add(char *type, ...);
1281 extern void policy_free(void);
1282
1283 extern struct dev_policy *path_policy(char **paths, char *type);
1284 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1285 extern struct dev_policy *devid_policy(int devid);
1286 extern void dev_policy_free(struct dev_policy *p);
1287
1288 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1289 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1290 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1291
1292 enum policy_action {
1293 act_default,
1294 act_include,
1295 act_re_add,
1296 act_spare, /* This only applies to bare devices */
1297 act_spare_same_slot, /* this allows non-bare devices,
1298 * but only if recent removal */
1299 act_force_spare, /* this allow non-bare devices in any case */
1300 act_err
1301 };
1302
1303 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1304 enum policy_action want);
1305 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1306 enum policy_action want);
1307
1308 struct domainlist {
1309 struct domainlist *next;
1310 const char *dom;
1311 };
1312
1313 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1314 const char *metadata);
1315 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1316 const char *metadata);
1317 extern void domainlist_add_dev(struct domainlist **dom, int devid,
1318 const char *metadata);
1319 extern void domain_free(struct domainlist *dl);
1320 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1321 const char *metadata);
1322 void domain_add(struct domainlist **domp, char *domain);
1323
1324 extern void policy_save_path(char *id_path, struct map_ent *array);
1325 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1326
1327 extern void sysfs_rules_apply(char *devnm, struct mdinfo *dev);
1328 extern void sysfsline(char *line);
1329
1330 #if __GNUC__ < 3
1331 struct stat64;
1332 #endif
1333
1334 #define HAVE_NFTW we assume
1335 #define HAVE_FTW
1336
1337 #ifdef __UCLIBC__
1338 # include <features.h>
1339 # ifndef __UCLIBC_HAS_LFS__
1340 # define lseek64 lseek
1341 # endif
1342 # ifndef __UCLIBC_HAS_FTW__
1343 # undef HAVE_FTW
1344 # undef HAVE_NFTW
1345 # endif
1346 #endif
1347
1348 #ifdef __dietlibc__
1349 # undef HAVE_NFTW
1350 #endif
1351
1352 #if defined(__KLIBC__)
1353 # undef HAVE_NFTW
1354 # undef HAVE_FTW
1355 #endif
1356
1357 #ifndef HAVE_NFTW
1358 # define FTW_PHYS 1
1359 # ifndef HAVE_FTW
1360 struct FTW {};
1361 # endif
1362 #endif
1363
1364 #ifdef HAVE_FTW
1365 # include <ftw.h>
1366 #endif
1367
1368 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1369
1370 extern int Manage_ro(char *devname, int fd, int readonly);
1371 extern int Manage_run(char *devname, int fd, struct context *c);
1372 extern int Manage_stop(char *devname, int fd, int quiet,
1373 int will_retry);
1374 extern int Manage_subdevs(char *devname, int fd,
1375 struct mddev_dev *devlist, int verbose, int test,
1376 char *update, int force);
1377 extern int autodetect(void);
1378 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1379 extern int Grow_addbitmap(char *devname, int fd,
1380 struct context *c, struct shape *s);
1381 extern int Grow_reshape(char *devname, int fd,
1382 struct mddev_dev *devlist,
1383 unsigned long long data_offset,
1384 struct context *c, struct shape *s);
1385 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1386 int *fdlist, int cnt, char *backup_file, int verbose);
1387 extern int Grow_continue(int mdfd, struct supertype *st,
1388 struct mdinfo *info, char *backup_file,
1389 int forked, int freeze_reshape);
1390 extern int Grow_consistency_policy(char *devname, int fd,
1391 struct context *c, struct shape *s);
1392
1393 extern int restore_backup(struct supertype *st,
1394 struct mdinfo *content,
1395 int working_disks,
1396 int spares,
1397 char **backup_filep,
1398 int verbose);
1399 extern int Grow_continue_command(char *devname, int fd,
1400 char *backup_file, int verbose);
1401
1402 extern int Assemble(struct supertype *st, char *mddev,
1403 struct mddev_ident *ident,
1404 struct mddev_dev *devlist,
1405 struct context *c);
1406
1407 extern int Build(char *mddev, struct mddev_dev *devlist,
1408 struct shape *s, struct context *c);
1409
1410 extern int Create(struct supertype *st, char *mddev,
1411 char *name, int *uuid,
1412 int subdevs, struct mddev_dev *devlist,
1413 struct shape *s,
1414 struct context *c,
1415 unsigned long long data_offset);
1416
1417 extern int Detail(char *dev, struct context *c);
1418 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1419 extern int Query(char *dev);
1420 extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1421 extern int Examine(struct mddev_dev *devlist, struct context *c,
1422 struct supertype *forcest);
1423 extern int Monitor(struct mddev_dev *devlist,
1424 char *mailaddr, char *alert_cmd,
1425 struct context *c,
1426 int daemonise, int oneshot,
1427 int dosyslog, char *pidfile, int increments,
1428 int share);
1429
1430 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1431 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1432 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1433 extern int Wait(char *dev);
1434 extern int WaitClean(char *dev, int verbose);
1435 extern int SetAction(char *dev, char *action);
1436
1437 extern int Incremental(struct mddev_dev *devlist, struct context *c,
1438 struct supertype *st);
1439 extern void RebuildMap(void);
1440 extern int IncrementalScan(struct context *c, char *devnm);
1441 extern int IncrementalRemove(char *devname, char *path, int verbose);
1442 extern int CreateBitmap(char *filename, int force, char uuid[16],
1443 unsigned long chunksize, unsigned long daemon_sleep,
1444 unsigned long write_behind,
1445 unsigned long long array_size,
1446 int major);
1447 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1448 extern int Write_rules(char *rule_name);
1449 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1450
1451 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1452 static inline unsigned long long
1453 bitmap_bits(unsigned long long array_size, unsigned long chunksize)
1454 {
1455 return (array_size * 512 + chunksize - 1) / chunksize;
1456 }
1457
1458 extern int Dump_metadata(char *dev, char *dir, struct context *c,
1459 struct supertype *st);
1460 extern int Restore_metadata(char *dev, char *dir, struct context *c,
1461 struct supertype *st, int only);
1462
1463 int md_array_valid(int fd);
1464 int md_array_active(int fd);
1465 int md_array_is_active(struct mdinfo *info);
1466 int md_get_array_info(int fd, struct mdu_array_info_s *array);
1467 int md_set_array_info(int fd, struct mdu_array_info_s *array);
1468 int md_get_disk_info(int fd, struct mdu_disk_info_s *disk);
1469 extern int get_linux_version(void);
1470 extern int mdadm_version(char *version);
1471 extern unsigned long long parse_size(char *size);
1472 extern int parse_uuid(char *str, int uuid[4]);
1473 extern int is_near_layout_10(int layout);
1474 extern int parse_layout_10(char *layout);
1475 extern int parse_layout_faulty(char *layout);
1476 extern long parse_num(char *num);
1477 extern int parse_cluster_confirm_arg(char *inp, char **devname, int *slot);
1478 extern int check_ext2(int fd, char *name);
1479 extern int check_reiser(int fd, char *name);
1480 extern int check_raid(int fd, char *name);
1481 extern int check_partitions(int fd, char *dname,
1482 unsigned long long freesize,
1483 unsigned long long size);
1484 extern int fstat_is_blkdev(int fd, char *devname, dev_t *rdev);
1485 extern int stat_is_blkdev(char *devname, dev_t *rdev);
1486
1487 extern int get_mdp_major(void);
1488 extern int get_maj_min(char *dev, int *major, int *minor);
1489 extern int dev_open(char *dev, int flags);
1490 extern int open_dev(char *devnm);
1491 extern void reopen_mddev(int mdfd);
1492 extern int open_dev_flags(char *devnm, int flags);
1493 extern int open_dev_excl(char *devnm);
1494 extern int is_standard(char *dev, int *nump);
1495 extern int same_dev(char *one, char *two);
1496 extern int compare_paths (char* path1,char* path2);
1497 extern void enable_fds(int devices);
1498
1499 extern int parse_auto(char *str, char *msg, int config);
1500 extern struct mddev_ident *conf_get_ident(char *dev);
1501 extern struct mddev_dev *conf_get_devs(void);
1502 extern int conf_test_dev(char *devname);
1503 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1504 extern struct createinfo *conf_get_create_info(void);
1505 extern void set_conffile(char *file);
1506 extern char *conf_get_mailaddr(void);
1507 extern char *conf_get_mailfrom(void);
1508 extern char *conf_get_program(void);
1509 extern char *conf_get_homehost(int *require_homehostp);
1510 extern char *conf_get_homecluster(void);
1511 extern char *conf_line(FILE *file);
1512 extern char *conf_word(FILE *file, int allow_key);
1513 extern void print_quoted(char *str);
1514 extern void print_escape(char *str);
1515 extern int use_udev(void);
1516 extern unsigned long GCD(unsigned long a, unsigned long b);
1517 extern int conf_name_is_free(char *name);
1518 extern int conf_verify_devnames(struct mddev_ident *array_list);
1519 extern int devname_matches(char *name, char *match);
1520 extern struct mddev_ident *conf_match(struct supertype *st,
1521 struct mdinfo *info,
1522 char *devname,
1523 int verbose, int *rvp);
1524
1525 extern void free_line(char *line);
1526 extern int match_oneof(char *devices, char *devname);
1527 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1528 extern const int uuid_zero[4];
1529 extern int same_uuid(int a[4], int b[4], int swapuuid);
1530 extern void copy_uuid(void *a, int b[4], int swapuuid);
1531 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1532 extern char *fname_from_uuid(struct supertype *st,
1533 struct mdinfo *info, char *buf, char sep);
1534 extern unsigned long calc_csum(void *super, int bytes);
1535 extern int enough(int level, int raid_disks, int layout, int clean,
1536 char *avail);
1537 extern int ask(char *mesg);
1538 extern unsigned long long get_component_size(int fd);
1539 extern void remove_partitions(int fd);
1540 extern int test_partition(int fd);
1541 extern int test_partition_from_id(dev_t id);
1542 extern int get_data_disks(int level, int layout, int raid_disks);
1543 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1544 int chunksize, unsigned long long devsize);
1545 extern int flush_metadata_updates(struct supertype *st);
1546 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1547 extern int assemble_container_content(struct supertype *st, int mdfd,
1548 struct mdinfo *content,
1549 struct context *c,
1550 char *chosen_name, int *result);
1551 #define INCR_NO 1
1552 #define INCR_UNSAFE 2
1553 #define INCR_ALREADY 4
1554 #define INCR_YES 8
1555 extern struct mdinfo *container_choose_spares(struct supertype *st,
1556 struct spare_criteria *criteria,
1557 struct domainlist *domlist,
1558 char *spare_group,
1559 const char *metadata, int get_one);
1560 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1561 extern int add_disk(int mdfd, struct supertype *st,
1562 struct mdinfo *sra, struct mdinfo *info);
1563 extern int remove_disk(int mdfd, struct supertype *st,
1564 struct mdinfo *sra, struct mdinfo *info);
1565 extern int hot_remove_disk(int mdfd, unsigned long dev, int force);
1566 extern int sys_hot_remove_disk(int statefd, int force);
1567 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1568 unsigned long long min_recovery_start(struct mdinfo *array);
1569
1570 extern char *human_size(long long bytes);
1571 extern char *human_size_brief(long long bytes, int prefix);
1572 extern void print_r10_layout(int layout);
1573
1574 extern char *find_free_devnm(int use_partitions);
1575
1576 extern void put_md_name(char *name);
1577 extern char *devid2kname(dev_t devid);
1578 extern char *devid2devnm(dev_t devid);
1579 extern dev_t devnm2devid(char *devnm);
1580 extern char *get_md_name(char *devnm);
1581
1582 extern char DefaultConfFile[];
1583
1584 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1585 char *chosen, int block_udev);
1586 /* values for 'trustworthy' */
1587 #define LOCAL 1
1588 #define LOCAL_ANY 10
1589 #define FOREIGN 2
1590 #define METADATA 3
1591 extern int open_mddev(char *dev, int report_errors);
1592 extern int open_container(int fd);
1593 extern int metadata_container_matches(char *metadata, char *devnm);
1594 extern int metadata_subdev_matches(char *metadata, char *devnm);
1595 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1596 extern int is_subarray_active(char *subarray, char *devname);
1597 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1598 extern struct superswitch *version_to_superswitch(char *vers);
1599
1600 extern int mdmon_running(char *devnm);
1601 extern int mdmon_pid(char *devnm);
1602 extern int check_env(char *name);
1603 extern __u32 random32(void);
1604 extern void random_uuid(__u8 *buf);
1605 extern int start_mdmon(char *devnm);
1606
1607 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1608 struct supertype *st, unsigned long stripes,
1609 int *fds, unsigned long long *offsets,
1610 int dests, int *destfd, unsigned long long *destoffsets);
1611 void abort_reshape(struct mdinfo *sra);
1612
1613 void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1614
1615 extern char *stat2kname(struct stat *st);
1616 extern char *fd2kname(int fd);
1617 extern char *stat2devnm(struct stat *st);
1618 extern char *fd2devnm(int fd);
1619 extern void udev_block(char *devnm);
1620 extern void udev_unblock(void);
1621
1622 extern int in_initrd(void);
1623
1624 struct cmap_hooks {
1625 void *cmap_handle; /* corosync lib related */
1626
1627 int (*initialize)(cmap_handle_t *handle);
1628 int (*get_string)(cmap_handle_t handle,
1629 const char *string,
1630 char **name);
1631 int (*finalize)(cmap_handle_t handle);
1632 };
1633
1634 extern void set_cmap_hooks(void);
1635 extern void set_hooks(void);
1636
1637 struct dlm_hooks {
1638 void *dlm_handle; /* dlm lib related */
1639
1640 dlm_lshandle_t (*create_lockspace)(const char *name,
1641 unsigned int mode);
1642 dlm_lshandle_t (*open_lockspace)(const char *name);
1643 int (*release_lockspace)(const char *name, dlm_lshandle_t ls,
1644 int force);
1645 int (*ls_lock)(dlm_lshandle_t lockspace, uint32_t mode,
1646 struct dlm_lksb *lksb, uint32_t flags,
1647 const void *name, unsigned int namelen,
1648 uint32_t parent, void (*astaddr) (void *astarg),
1649 void *astarg, void (*bastaddr) (void *astarg),
1650 void *range);
1651 int (*ls_unlock_wait)(dlm_lshandle_t lockspace, uint32_t lkid,
1652 uint32_t flags, struct dlm_lksb *lksb);
1653 int (*ls_get_fd)(dlm_lshandle_t ls);
1654 int (*dispatch)(int fd);
1655 };
1656
1657 extern int get_cluster_name(char **name);
1658 extern int dlm_funs_ready(void);
1659 extern int cluster_get_dlmlock(void);
1660 extern int cluster_release_dlmlock(void);
1661 extern void set_dlm_hooks(void);
1662
1663 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1664 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1665 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1666 (ROUND_UP((unsigned long)(ptr), base)))
1667
1668 static inline int is_subarray(char *vers)
1669 {
1670 /* The version string for a 'subarray' (an array in a container)
1671 * is
1672 * /containername/componentname for normal read-write arrays
1673 * -containername/componentname for arrays which mdmon must not
1674 * reconfigure. They might be read-only
1675 * or might be undergoing reshape etc.
1676 * containername is e.g. md0, md_d1
1677 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1678 */
1679 return (*vers == '/' || *vers == '-');
1680 }
1681
1682 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1683 {
1684 return &ent->metadata_version[10+strlen(container)+1];
1685 }
1686
1687 #ifdef DEBUG
1688 #define dprintf(fmt, arg...) \
1689 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1690 #define dprintf_cont(fmt, arg...) \
1691 fprintf(stderr, fmt, ##arg)
1692 #else
1693 #define dprintf(fmt, arg...) \
1694 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1695 #define dprintf_cont(fmt, arg...) \
1696 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1697 #endif
1698 #include <assert.h>
1699 #include <stdarg.h>
1700 static inline int xasprintf(char **strp, const char *fmt, ...) {
1701 va_list ap;
1702 int ret;
1703 va_start(ap, fmt);
1704 ret = vasprintf(strp, fmt, ap);
1705 va_end(ap);
1706 assert(ret >= 0);
1707 return ret;
1708 }
1709
1710 #ifdef DEBUG
1711 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1712 #else
1713 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1714 #endif
1715 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1716
1717 void *xmalloc(size_t len);
1718 void *xrealloc(void *ptr, size_t len);
1719 void *xcalloc(size_t num, size_t size);
1720 char *xstrdup(const char *str);
1721
1722 #define LEVEL_MULTIPATH (-4)
1723 #define LEVEL_LINEAR (-1)
1724 #define LEVEL_FAULTY (-5)
1725
1726 /* kernel module doesn't know about these */
1727 #define LEVEL_CONTAINER (-100)
1728 #define LEVEL_UNSUPPORTED (-200)
1729
1730 /* the kernel does know about this one ... */
1731 #define LEVEL_NONE (-1000000)
1732
1733 /* faulty stuff */
1734
1735 #define WriteTransient 0
1736 #define ReadTransient 1
1737 #define WritePersistent 2
1738 #define ReadPersistent 3
1739 #define WriteAll 4 /* doesn't go to device */
1740 #define ReadFixable 5
1741 #define Modes 6
1742
1743 #define ClearErrors 31
1744 #define ClearFaults 30
1745
1746 #define AllPersist 100 /* internal use only */
1747 #define NoPersist 101
1748
1749 #define ModeMask 0x1f
1750 #define ModeShift 5
1751
1752 #ifdef __TINYC__
1753 #undef minor
1754 #undef major
1755 #undef makedev
1756 #define minor(x) ((x)&0xff)
1757 #define major(x) (((x)>>8)&0xff)
1758 #define makedev(M,m) (((M)<<8) | (m))
1759 #endif
1760
1761 /* for raid4/5/6 */
1762 #define ALGORITHM_LEFT_ASYMMETRIC 0
1763 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1764 #define ALGORITHM_LEFT_SYMMETRIC 2
1765 #define ALGORITHM_RIGHT_SYMMETRIC 3
1766
1767 /* Define non-rotating (raid4) algorithms. These allow
1768 * conversion of raid4 to raid5.
1769 */
1770 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1771 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1772
1773 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1774 * Firstly, the exact positioning of the parity block is slightly
1775 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1776 * of DDF.
1777 * Secondly, or order of datablocks over which the Q syndrome is computed
1778 * is different.
1779 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1780 * These layouts are from the DDFv1.2 spec.
1781 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1782 * leaves RLQ=3 as 'Vendor Specific'
1783 */
1784
1785 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1786 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1787 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1788
1789 /* For every RAID5 algorithm we define a RAID6 algorithm
1790 * with exactly the same layout for data and parity, and
1791 * with the Q block always on the last device (N-1).
1792 * This allows trivial conversion from RAID5 to RAID6
1793 */
1794 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1795 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1796 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1797 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1798 #define ALGORITHM_PARITY_0_6 20
1799 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1800
1801 /* Define PATH_MAX in case we don't use glibc or standard library does
1802 * not have PATH_MAX defined. Assume max path length is 4K characters.
1803 */
1804 #ifndef PATH_MAX
1805 #define PATH_MAX 4096
1806 #endif
1807
1808 #define RESYNC_NONE -1
1809 #define RESYNC_DELAYED -2
1810 #define RESYNC_PENDING -3
1811 #define RESYNC_UNKNOWN -4
1812
1813 /* When using "GET_DISK_INFO" it isn't certain how high
1814 * we need to check. So we impose an absolute limit of
1815 * MAX_DISKS. This needs to be much more than the largest
1816 * number of devices any metadata can support. Currently
1817 * v1.x can support 1920
1818 */
1819 #define MAX_DISKS 4096
1820
1821 /* Sometimes the 'size' value passed needs to mean "Maximum".
1822 * In those cases with use MAX_SIZE
1823 */
1824 #define MAX_SIZE 1
1825
1826 /* We want to use unsigned numbers for sector counts, but need
1827 * a value for 'invalid'. Use '1'.
1828 */
1829 #define INVALID_SECTORS 1
1830 /* And another special number needed for --data_offset=variable */
1831 #define VARIABLE_OFFSET 3