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