]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Make target to install binaries only
[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 #ifndef MDMON_SERVICE
133 #define MDMON_SERVICE "mdmon"
134 #endif /* MDMON_SERVICE */
135
136 #ifndef GROW_SERVICE
137 #define GROW_SERVICE "mdadm-grow-continue"
138 #endif /* GROW_SERVICE */
139
140 #include "md_u.h"
141 #include "md_p.h"
142 #include "bitmap.h"
143 #include "msg.h"
144
145 #include <endian.h>
146 /* Redhat don't like to #include <asm/byteorder.h>, and
147 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
148 * and there is no standard conversion function so... */
149 /* And dietlibc doesn't think byteswap is ok, so.. */
150 /* #include <byteswap.h> */
151 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
152 ((x) & 0xff00U) >> 8)
153 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
154 ((x) & 0xff000000U) >> 24 | \
155 ((x) & 0x0000ff00U) << 8 | \
156 ((x) & 0x00ff0000U) >> 8)
157 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
158 ((x) & 0xff00000000000000ULL) >> 56 | \
159 ((x) & 0x000000000000ff00ULL) << 40 | \
160 ((x) & 0x00ff000000000000ULL) >> 40 | \
161 ((x) & 0x0000000000ff0000ULL) << 24 | \
162 ((x) & 0x0000ff0000000000ULL) >> 24 | \
163 ((x) & 0x00000000ff000000ULL) << 8 | \
164 ((x) & 0x000000ff00000000ULL) >> 8)
165
166 #if !defined(__KLIBC__)
167 #if BYTE_ORDER == LITTLE_ENDIAN
168 #define __cpu_to_le16(_x) (unsigned int)(_x)
169 #define __cpu_to_le32(_x) (unsigned int)(_x)
170 #define __cpu_to_le64(_x) (unsigned long long)(_x)
171 #define __le16_to_cpu(_x) (unsigned int)(_x)
172 #define __le32_to_cpu(_x) (unsigned int)(_x)
173 #define __le64_to_cpu(_x) (unsigned long long)(_x)
174
175 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
176 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
177 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
178 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
179 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
180 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
181 #elif BYTE_ORDER == BIG_ENDIAN
182 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
183 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
184 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
185 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
186 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
187 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
188
189 #define __cpu_to_be16(_x) (unsigned int)(_x)
190 #define __cpu_to_be32(_x) (unsigned int)(_x)
191 #define __cpu_to_be64(_x) (unsigned long long)(_x)
192 #define __be16_to_cpu(_x) (unsigned int)(_x)
193 #define __be32_to_cpu(_x) (unsigned int)(_x)
194 #define __be64_to_cpu(_x) (unsigned long long)(_x)
195 #else
196 # error "unknown endianness."
197 #endif
198 #endif /* __KLIBC__ */
199
200 /*
201 * Partially stolen from include/linux/unaligned/packed_struct.h
202 */
203 struct __una_u16 { __u16 x; } __attribute__ ((packed));
204 struct __una_u32 { __u32 x; } __attribute__ ((packed));
205
206 static inline __u16 __get_unaligned16(const void *p)
207 {
208 const struct __una_u16 *ptr = (const struct __una_u16 *)p;
209 return ptr->x;
210 }
211
212 static inline __u32 __get_unaligned32(const void *p)
213 {
214 const struct __una_u32 *ptr = (const struct __una_u32 *)p;
215 return ptr->x;
216 }
217
218 static inline void __put_unaligned16(__u16 val, void *p)
219 {
220 struct __una_u16 *ptr = (struct __una_u16 *)p;
221 ptr->x = val;
222 }
223
224 static inline void __put_unaligned32(__u32 val, void *p)
225 {
226 struct __una_u32 *ptr = (struct __una_u32 *)p;
227 ptr->x = val;
228 }
229
230 /*
231 * Check at compile time that something is of a particular type.
232 * Always evaluates to 1 so you may use it easily in comparisons.
233 */
234
235 #define typecheck(type,x) \
236 ({ type __dummy; \
237 typeof(x) __dummy2; \
238 (void)(&__dummy == &__dummy2); \
239 1; \
240 })
241
242 /*
243 * These inlines deal with timer wrapping correctly.
244 *
245 * time_after(a,b) returns true if the time a is after time b.
246 */
247
248 #define time_after(a,b) \
249 (typecheck(unsigned int, a) && \
250 typecheck(unsigned int, b) && \
251 ((int)((b) - (a)) < 0))
252
253 #define time_before(a,b) time_after(b,a)
254
255 /*
256 * min()/max()/clamp() macros that also do
257 * strict type-checking.. See the
258 * "unnecessary" pointer comparison.
259 */
260 #define min(x, y) ({ \
261 typeof(x) _min1 = (x); \
262 typeof(y) _min2 = (y); \
263 (void) (&_min1 == &_min2); \
264 _min1 < _min2 ? _min1 : _min2; })
265
266 #define max(x, y) ({ \
267 typeof(x) _max1 = (x); \
268 typeof(y) _max2 = (y); \
269 (void) (&_max1 == &_max2); \
270 _max1 > _max2 ? _max1 : _max2; })
271
272 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
273
274 extern const char Name[];
275
276 struct md_bb_entry {
277 unsigned long long sector;
278 int length;
279 };
280
281 struct md_bb {
282 int supported;
283 int count;
284 struct md_bb_entry *entries;
285 };
286
287 /* general information that might be extracted from a superblock */
288 struct mdinfo {
289 mdu_array_info_t array;
290 mdu_disk_info_t disk;
291 __u64 events;
292 int uuid[4];
293 char name[33];
294 unsigned long long data_offset;
295 unsigned long long new_data_offset;
296 unsigned long long component_size; /* same as array.size, except in
297 * sectors and up to 64bits.
298 */
299 unsigned long long custom_array_size; /* size for non-default sized
300 * arrays (in sectors)
301 */
302 #define NO_RESHAPE 0
303 #define VOLUME_RESHAPE 1
304 #define CONTAINER_RESHAPE 2
305 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
306 int reshape_active;
307 unsigned long long reshape_progress;
308 int recovery_blocked; /* for external metadata it
309 * indicates that there is
310 * reshape in progress in
311 * container,
312 * for native metadata it is
313 * reshape_active field mirror
314 */
315 int journal_device_required;
316 int journal_clean;
317
318 enum {
319 CONSISTENCY_POLICY_UNKNOWN,
320 CONSISTENCY_POLICY_NONE,
321 CONSISTENCY_POLICY_RESYNC,
322 CONSISTENCY_POLICY_BITMAP,
323 CONSISTENCY_POLICY_JOURNAL,
324 CONSISTENCY_POLICY_PPL,
325 } consistency_policy;
326
327 /* During reshape we can sometimes change the data_offset to avoid
328 * over-writing still-valid data. We need to know if there is space.
329 * So getinfo_super will fill in space_before and space_after in sectors.
330 * data_offset can be increased or decreased by this amount.
331 */
332 unsigned long long space_before, space_after;
333 union {
334 unsigned long long resync_start; /* per-array resync position */
335 unsigned long long recovery_start; /* per-device rebuild position */
336 #define MaxSector (~0ULL) /* resync/recovery complete position */
337 };
338 long bitmap_offset; /* 0 == none, 1 == a file */
339 unsigned int ppl_size;
340 int ppl_offset;
341 unsigned long long ppl_sector;
342 unsigned long safe_mode_delay; /* ms delay to mark clean */
343 int new_level, delta_disks, new_layout, new_chunk;
344 int errors;
345 unsigned long cache_size; /* size of raid456 stripe cache*/
346 int mismatch_cnt;
347 char text_version[50];
348
349 int container_member; /* for assembling external-metatdata arrays
350 * This is to be used internally by metadata
351 * handler only */
352 int container_enough; /* flag external handlers can set to
353 * indicate that subarrays have not enough (-1),
354 * enough to start (0), or all expected disks (1) */
355 char sys_name[32];
356 struct mdinfo *devs;
357 struct mdinfo *next;
358
359 /* Device info for mdmon: */
360 int recovery_fd;
361 int state_fd;
362 int bb_fd;
363 int ubb_fd;
364 #define DS_FAULTY 1
365 #define DS_INSYNC 2
366 #define DS_WRITE_MOSTLY 4
367 #define DS_SPARE 8
368 #define DS_BLOCKED 16
369 #define DS_REMOVE 1024
370 #define DS_UNBLOCK 2048
371 int prev_state, curr_state, next_state;
372
373 /* info read from sysfs */
374 enum {
375 ARRAY_CLEAR,
376 ARRAY_INACTIVE,
377 ARRAY_SUSPENDED,
378 ARRAY_READONLY,
379 ARRAY_READ_AUTO,
380 ARRAY_CLEAN,
381 ARRAY_ACTIVE,
382 ARRAY_WRITE_PENDING,
383 ARRAY_ACTIVE_IDLE,
384 ARRAY_BROKEN,
385 ARRAY_UNKNOWN_STATE,
386 } array_state;
387 struct md_bb bb;
388 };
389
390 struct createinfo {
391 int uid;
392 int gid;
393 int autof;
394 int mode;
395 int symlinks;
396 int names;
397 int bblist;
398 struct supertype *supertype;
399 };
400
401 struct spare_criteria {
402 unsigned long long min_size;
403 unsigned int sector_size;
404 };
405
406 enum mode {
407 ASSEMBLE=1,
408 BUILD,
409 CREATE,
410 MANAGE,
411 MISC,
412 MONITOR,
413 GROW,
414 INCREMENTAL,
415 AUTODETECT,
416 mode_count
417 };
418
419 extern char short_options[];
420 extern char short_bitmap_options[];
421 extern char short_bitmap_auto_options[];
422 extern struct option long_options[];
423 extern char Version[], Usage[], Help[], OptionHelp[],
424 *mode_help[],
425 Help_create[], Help_build[], Help_assemble[], Help_grow[],
426 Help_incr[],
427 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
428
429 /* for option that don't have short equivilents, we assign arbitrary
430 * numbers later than any 'short' character option.
431 */
432 enum special_options {
433 AssumeClean = 300,
434 BitmapChunk,
435 WriteBehind,
436 ReAdd,
437 NoDegraded,
438 Sparc22,
439 BackupFile,
440 HomeHost,
441 AutoHomeHost,
442 Symlinks,
443 AutoDetect,
444 Waitclean,
445 DetailPlatform,
446 KillSubarray,
447 UpdateSubarray,
448 IncrementalPath,
449 NoSharing,
450 HelpOptions,
451 Brief,
452 NoDevices,
453 ManageOpt,
454 Add,
455 AddSpare,
456 AddJournal,
457 Remove,
458 Fail,
459 Replace,
460 With,
461 MiscOpt,
462 WaitOpt,
463 ConfigFile,
464 ChunkSize,
465 WriteMostly,
466 FailFast,
467 NoFailFast,
468 Layout,
469 Auto,
470 Force,
471 SuperMinor,
472 EMail,
473 ProgramOpt,
474 Increment,
475 Fork,
476 Bitmap,
477 RebuildMapOpt,
478 InvalidBackup,
479 UdevRules,
480 FreezeReshape,
481 Continue,
482 OffRootOpt,
483 Prefer,
484 KillOpt,
485 DataOffset,
486 ExamineBB,
487 Dump,
488 Restore,
489 Action,
490 Nodes,
491 ClusterName,
492 ClusterConfirm,
493 WriteJournal,
494 ConsistencyPolicy,
495 };
496
497 enum prefix_standard {
498 JEDEC,
499 IEC
500 };
501
502 enum bitmap_update {
503 NoUpdate,
504 NameUpdate,
505 NodeNumUpdate,
506 };
507
508 enum flag_mode {
509 FlagDefault, FlagSet, FlagClear,
510 };
511
512 /* structures read from config file */
513 /* List of mddevice names and identifiers
514 * Identifiers can be:
515 * uuid=128-hex-uuid
516 * super-minor=decimal-minor-number-from-superblock
517 * devices=comma,separated,list,of,device,names,with,wildcards
518 *
519 * If multiple fields are present, the intersection of all matching
520 * devices is considered
521 */
522 #define UnSet (0xfffe)
523 struct mddev_ident {
524 char *devname;
525
526 int uuid_set;
527 int uuid[4];
528 char name[33];
529
530 int super_minor;
531
532 char *devices; /* comma separated list of device
533 * names with wild cards
534 */
535 int level;
536 int raid_disks;
537 int spare_disks;
538 struct supertype *st;
539 int autof; /* 1 for normal, 2 for partitioned */
540 char *spare_group;
541 char *bitmap_file;
542 int bitmap_fd;
543
544 char *container; /* /dev/whatever name of container, or
545 * uuid of container. You would expect
546 * this to be the 'devname' or UUID
547 * of some other entry.
548 */
549 char *member; /* subarray within a container */
550
551 struct mddev_ident *next;
552 union {
553 /* fields needed by different users of this structure */
554 int assembled; /* set when assembly succeeds */
555 };
556 };
557
558 struct context {
559 int readonly;
560 int runstop;
561 int verbose;
562 int brief;
563 int no_devices;
564 int force;
565 char *homehost;
566 int require_homehost;
567 char *prefer;
568 int export;
569 int test;
570 char *subarray;
571 char *update;
572 int scan;
573 int SparcAdjust;
574 int autof;
575 int delay;
576 int freeze_reshape;
577 char *backup_file;
578 int invalid_backup;
579 char *action;
580 int nodes;
581 char *homecluster;
582 };
583
584 struct shape {
585 int raiddisks;
586 int sparedisks;
587 int journaldisks;
588 int level;
589 int layout;
590 char *layout_str;
591 int chunk;
592 int bitmap_chunk;
593 char *bitmap_file;
594 int assume_clean;
595 int write_behind;
596 unsigned long long size;
597 int consistency_policy;
598 };
599
600 /* List of device names - wildcards expanded */
601 struct mddev_dev {
602 char *devname;
603 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
604 * 'A' for re_add.
605 * Not set for names read from .config
606 */
607 enum flag_mode writemostly;
608 enum flag_mode failfast;
609 int used; /* set when used */
610 long long data_offset;
611 struct mddev_dev *next;
612 };
613
614 typedef struct mapping {
615 char *name;
616 int num;
617 } mapping_t;
618
619 struct mdstat_ent {
620 char devnm[32];
621 int active;
622 char *level;
623 char *pattern; /* U for up, _ for down */
624 int percent; /* -1 if no resync */
625 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
626 int devcnt;
627 int raid_disks;
628 char * metadata_version;
629 struct dev_member {
630 char *name;
631 struct dev_member *next;
632 } *members;
633 struct mdstat_ent *next;
634 };
635
636 extern struct mdstat_ent *mdstat_read(int hold, int start);
637 extern void mdstat_close(void);
638 extern void free_mdstat(struct mdstat_ent *ms);
639 extern int mdstat_wait(int seconds);
640 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
641 extern int mddev_busy(char *devnm);
642 extern struct mdstat_ent *mdstat_by_component(char *name);
643 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
644
645 struct map_ent {
646 struct map_ent *next;
647 char devnm[32];
648 char metadata[20];
649 int uuid[4];
650 int bad;
651 char *path;
652 };
653 extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
654 int uuid[4], char *path);
655 extern void map_remove(struct map_ent **map, char *devnm);
656 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
657 extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
658 extern void map_free(struct map_ent *map);
659 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
660 extern void map_read(struct map_ent **melp);
661 extern int map_write(struct map_ent *mel);
662 extern void map_delete(struct map_ent **mapp, char *devnm);
663 extern void map_add(struct map_ent **melp,
664 char *devnm, char *metadata, int uuid[4], char *path);
665 extern int map_lock(struct map_ent **melp);
666 extern void map_unlock(struct map_ent **melp);
667 extern void map_fork(void);
668
669 /* various details can be requested */
670 enum sysfs_read_flags {
671 GET_LEVEL = (1 << 0),
672 GET_LAYOUT = (1 << 1),
673 GET_COMPONENT = (1 << 2),
674 GET_CHUNK = (1 << 3),
675 GET_CACHE = (1 << 4),
676 GET_MISMATCH = (1 << 5),
677 GET_VERSION = (1 << 6),
678 GET_DISKS = (1 << 7),
679 GET_SAFEMODE = (1 << 9),
680 GET_BITMAP_LOCATION = (1 << 10),
681
682 GET_DEVS = (1 << 20), /* gets role, major, minor */
683 GET_OFFSET = (1 << 21),
684 GET_SIZE = (1 << 22),
685 GET_STATE = (1 << 23),
686 GET_ERROR = (1 << 24),
687 GET_ARRAY_STATE = (1 << 25),
688 GET_CONSISTENCY_POLICY = (1 << 26),
689 GET_DEVS_ALL = (1 << 27),
690 };
691
692 /* If fd >= 0, get the array it is open on,
693 * else use devnm.
694 */
695 extern int sysfs_open(char *devnm, char *devname, char *attr);
696 extern int sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
697 extern void sysfs_init_dev(struct mdinfo *mdi, dev_t devid);
698 extern void sysfs_free(struct mdinfo *sra);
699 extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
700 extern int sysfs_attr_match(const char *attr, const char *str);
701 extern int sysfs_match_word(const char *word, char **list);
702 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
703 char *name, char *val);
704 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
705 char *name, unsigned long long val);
706 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
707 char *name, long long val);
708 extern int sysfs_uevent(struct mdinfo *sra, char *event);
709 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
710 char *name);
711 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
712 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
713 char *name, unsigned long long *val);
714 extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
715 extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
716 char *name, unsigned long long *v1, unsigned long long *v2);
717 extern int sysfs_fd_get_str(int fd, char *val, int size);
718 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
719 char *name);
720 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
721 char *name, char *val, int size);
722 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
723 extern int sysfs_set_array(struct mdinfo *info, int vers);
724 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
725 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
726 extern int sysfs_unique_holder(char *devnm, long rdev);
727 extern int sysfs_freeze_array(struct mdinfo *sra);
728 extern int sysfs_wait(int fd, int *msec);
729 extern int load_sys(char *path, char *buf, int len);
730 extern int zero_disk_range(int fd, unsigned long long sector, size_t count);
731 extern int reshape_prepare_fdlist(char *devname,
732 struct mdinfo *sra,
733 int raid_disks,
734 int nrdisks,
735 unsigned long blocks,
736 char *backup_file,
737 int *fdlist,
738 unsigned long long *offsets);
739 extern void reshape_free_fdlist(int *fdlist,
740 unsigned long long *offsets,
741 int size);
742 extern int reshape_open_backup_file(char *backup,
743 int fd,
744 char *devname,
745 long blocks,
746 int *fdlist,
747 unsigned long long *offsets,
748 char *sysfs_name,
749 int restart);
750 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
751 unsigned int ndata, unsigned int odata);
752 extern char *locate_backup(char *name);
753 extern char *make_backup(char *name);
754
755 extern int save_stripes(int *source, unsigned long long *offsets,
756 int raid_disks, int chunk_size, int level, int layout,
757 int nwrites, int *dest,
758 unsigned long long start, unsigned long long length,
759 char *buf);
760 extern int restore_stripes(int *dest, unsigned long long *offsets,
761 int raid_disks, int chunk_size, int level, int layout,
762 int source, unsigned long long read_offset,
763 unsigned long long start, unsigned long long length,
764 char *src_buf);
765
766 #ifndef Sendmail
767 #define Sendmail "/usr/lib/sendmail -t"
768 #endif
769
770 #define SYSLOG_FACILITY LOG_DAEMON
771
772 extern char *map_num(mapping_t *map, int num);
773 extern int map_name(mapping_t *map, char *name);
774 extern mapping_t r0layout[], r5layout[], r6layout[],
775 pers[], modes[], faultylayout[];
776 extern mapping_t consistency_policies[], sysfs_array_states[];
777
778 extern char *map_dev_preferred(int major, int minor, int create,
779 char *prefer);
780 static inline char *map_dev(int major, int minor, int create)
781 {
782 return map_dev_preferred(major, minor, create, NULL);
783 }
784
785 struct active_array;
786 struct metadata_update;
787
788 /* 'struct reshape' records the intermediate states of
789 * a general reshape.
790 * The starting geometry is converted to the 'before' geometry
791 * by at most an atomic level change. They could be the same.
792 * Similarly the 'after' geometry is converted to the final
793 * geometry by at most a level change.
794 * Note that 'before' and 'after' must have the same level.
795 * 'blocks' is the minimum number of sectors for a reshape unit.
796 * This will be a multiple of the stripe size in each of the
797 * 'before' and 'after' geometries.
798 * If 'blocks' is 0, no restriping is necessary.
799 * 'min_offset_change' is the minimum change to data_offset to
800 * allow the reshape to happen. It is at least the larger of
801 * the old and new chunk sizes, and typically the same as 'blocks'
802 * divided by number of data disks.
803 */
804 struct reshape {
805 int level;
806 int parity; /* number of parity blocks/devices */
807 struct {
808 int layout;
809 int data_disks;
810 } before, after;
811 unsigned long long backup_blocks;
812 unsigned long long min_offset_change;
813 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
814 unsigned long long new_size; /* New size of array in sectors */
815 };
816
817 /* A superswitch provides entry point to a metadata handler.
818 *
819 * The superswitch primarily operates on some "metadata" that
820 * is accessed via the 'supertype'.
821 * This metadata has one of three possible sources.
822 * 1/ It is read from a single device. In this case it may not completely
823 * describe the array or arrays as some information might be on other
824 * devices.
825 * 2/ It is read from all devices in a container. In this case all
826 * information is present.
827 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
828 * be complete once enough ->add_to_super calls have completed.
829 *
830 * When creating an array inside a container, the metadata will be
831 * formed by a combination of 2 and 3. The metadata or the array is read,
832 * then new information is added.
833 *
834 * The metadata must sometimes have a concept of a 'current' array
835 * and a 'current' device.
836 * The 'current' array is set by init_super to be the newly created array,
837 * or is set by super_by_fd when it finds it is looking at an array inside
838 * a container.
839 *
840 * The 'current' device is either the device that the metadata was read from
841 * in case 1, or the last device added by add_to_super in case 3.
842 * Case 2 does not identify a 'current' device.
843 */
844 extern struct superswitch {
845
846 /* Used to report details of metadata read from a component
847 * device. ->load_super has been called.
848 */
849 void (*examine_super)(struct supertype *st, char *homehost);
850 void (*brief_examine_super)(struct supertype *st, int verbose);
851 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
852 void (*export_examine_super)(struct supertype *st);
853 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
854 int (*copy_metadata)(struct supertype *st, int from, int to);
855
856 /* Used to report details of an active array.
857 * ->load_super was possibly given a 'component' string.
858 */
859 void (*detail_super)(struct supertype *st, char *homehost,
860 char *subarray);
861 void (*brief_detail_super)(struct supertype *st, char *subarray);
862 void (*export_detail_super)(struct supertype *st);
863
864 /* Optional: platform hardware / firmware details */
865 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
866 int (*export_detail_platform)(int verbose, char *controller_path);
867
868 /* Used:
869 * to get uuid to storing in bitmap metadata
870 * and 'reshape' backup-data metadata
871 * To see if a device is being re-added to an array it was part of.
872 */
873 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
874
875 /* Extract generic details from metadata. This could be details about
876 * the container, or about an individual array within the container.
877 * The determination is made either by:
878 * load_super being given a 'component' string.
879 * validate_geometry determining what to create.
880 * The info includes both array information and device information.
881 * The particular device should be:
882 * The last device added by add_to_super
883 * The device the metadata was loaded from by load_super
884 * If 'map' is present, then it is an array raid_disks long
885 * (raid_disk must already be set and correct) and it is filled
886 * with 1 for slots that are thought to be active and 0 for slots which
887 * appear to be failed/missing.
888 * *info is zeroed out before data is added.
889 */
890 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
891 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
892 /* Check if the given metadata is flagged as belonging to "this"
893 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
894 */
895 int (*match_home)(struct supertype *st, char *homehost);
896
897 /* Make one of several generic modifications to metadata
898 * prior to assembly (or other times).
899 * sparc2.2 - first bug in early 0.90 metadata
900 * super-minor - change name of 0.90 metadata
901 * summaries - 'correct' any redundant data
902 * resync - mark array as dirty to trigger a resync.
903 * uuid - set new uuid - only 0.90 or 1.x
904 * name - change the name of the array (where supported)
905 * homehost - change which host this array is tied to.
906 * devicesize - If metadata is at start of device, change recorded
907 * device size to match actual device size
908 * byteorder - swap bytes for 0.90 metadata
909 *
910 * force-one - mark that device as uptodate, not old or failed.
911 * force-array - mark array as clean if it would not otherwise
912 * assemble
913 * assemble - not sure how this is different from force-one...
914 * linear-grow-new - add a new device to a linear array, but don't
915 * change the size: so superblock still matches
916 * linear-grow-update - now change the size of the array.
917 * writemostly - set the WriteMostly1 bit in the superblock devflags
918 * readwrite - clear the WriteMostly1 bit in the superblock devflags
919 * failfast - set the FailFast1 bit in the superblock
920 * nofailfast - clear the FailFast1 bit
921 * no-bitmap - clear any record that a bitmap is present.
922 * bbl - add a bad-block-log if possible
923 * no-bbl - remove any bad-block-log is it is empty.
924 * force-no-bbl - remove any bad-block-log even if empty.
925 * revert-reshape - If a reshape is in progress, modify metadata so
926 * it will resume going in the opposite direction.
927 */
928 int (*update_super)(struct supertype *st, struct mdinfo *info,
929 char *update,
930 char *devname, int verbose,
931 int uuid_set, char *homehost);
932
933 /* Create new metadata for new array as described. This could
934 * be a new container, or an array in a pre-existing container.
935 * Also used to zero metadata prior to writing it to invalidate old
936 * metadata.
937 */
938 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
939 struct shape *s, char *name,
940 char *homehost, int *uuid,
941 unsigned long long data_offset);
942
943 /* update the metadata to include new device, either at create or
944 * when hot-adding a spare.
945 */
946 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
947 int fd, char *devname,
948 unsigned long long data_offset);
949 /* update the metadata to delete a device,
950 * when hot-removing.
951 */
952 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
953
954 /* Write metadata to one device when fixing problems or adding
955 * a new device.
956 */
957 int (*store_super)(struct supertype *st, int fd);
958
959 /* Write all metadata for this array.
960 */
961 int (*write_init_super)(struct supertype *st);
962 /* Check if metadata read from one device is compatible with an array,
963 * used when assembling an array, or pseudo-assembling was with
964 * "--examine --brief"
965 * If "st" has not yet been loaded the superblock from, "tst" is
966 * moved in, otherwise the superblock in 'st' is compared with
967 * 'tst'.
968 */
969 int (*compare_super)(struct supertype *st, struct supertype *tst);
970 /* Load metadata from a single device. If 'devname' is not NULL
971 * print error messages as appropriate */
972 int (*load_super)(struct supertype *st, int fd, char *devname);
973 /* 'fd' is a 'container' md array - load array metadata from the
974 * whole container.
975 */
976 int (*load_container)(struct supertype *st, int fd, char *devname);
977 /* If 'arg' is a valid name of this metadata type, allocate and
978 * return a 'supertype' for the particular minor version */
979 struct supertype * (*match_metadata_desc)(char *arg);
980 /* If a device has the given size, and the data_offset has been
981 * requested - work out how much space is available for data.
982 * This involves adjusting for reserved space (e.g. bitmaps)
983 * and for any rounding.
984 * 'mdadm' only calls this for existing arrays where a possible
985 * spare is being added. However some super-handlers call it
986 * internally from validate_geometry when creating an array.
987 */
988 __u64 (*avail_size)(struct supertype *st, __u64 size,
989 unsigned long long data_offset);
990 /*
991 * Return spare criteria for array:
992 * - minimum disk size can be used in array;
993 * - sector size can be used in array.
994 * Return values: 0 - for success and -EINVAL on error.
995 */
996 int (*get_spare_criteria)(struct supertype *st,
997 struct spare_criteria *sc);
998 /* Find somewhere to put a bitmap - possibly auto-size it - and
999 * update the metadata to record this. The array may be newly
1000 * created, in which case data_size may be updated, or it might
1001 * already exist. Metadata handler can know if init_super
1002 * has been called, but not write_init_super.
1003 * 0: Success
1004 * -Exxxx: On error
1005 */
1006 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
1007 int delay, int write_behind,
1008 unsigned long long size, int may_change, int major);
1009 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1010 * md-native format bitmap
1011 */
1012 int (*locate_bitmap)(struct supertype *st, int fd, int node_num);
1013 /* if add_internal_bitmap succeeded for existing array, this
1014 * writes it out.
1015 */
1016 int (*write_bitmap)(struct supertype *st, int fd, enum bitmap_update update);
1017 /* Free the superblock and any other allocated data */
1018 void (*free_super)(struct supertype *st);
1019
1020 /* validate_geometry is called with an st returned by
1021 * match_metadata_desc.
1022 * It should check that the geometry described is compatible with
1023 * the metadata type. It will be called repeatedly as devices
1024 * added to validate changing size and new devices. If there are
1025 * inter-device dependencies, it should record sufficient details
1026 * so these can be validated.
1027 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1028 * Return value is:
1029 * 1: everything is OK
1030 * 0: not OK for some reason - if 'verbose', then error was reported.
1031 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1032 * type, but array is not acceptable for some reason
1033 * message was reported even if verbose is 0.
1034 */
1035 int (*validate_geometry)(struct supertype *st, int level, int layout,
1036 int raiddisks,
1037 int *chunk, unsigned long long size,
1038 unsigned long long data_offset,
1039 char *subdev, unsigned long long *freesize,
1040 int consistency_policy, int verbose);
1041
1042 /* Return a linked list of 'mdinfo' structures for all arrays
1043 * in the container. For non-containers, it is like
1044 * getinfo_super with an allocated mdinfo.*/
1045 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
1046 /* query the supertype for default geometry */
1047 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
1048 /* Permit subarray's to be deleted from inactive containers */
1049 int (*kill_subarray)(struct supertype *st,
1050 char *subarray_id); /* optional */
1051 /* Permit subarray's to be modified */
1052 int (*update_subarray)(struct supertype *st, char *subarray,
1053 char *update, struct mddev_ident *ident); /* optional */
1054 /* Check if reshape is supported for this external format.
1055 * st is obtained from super_by_fd() where st->subarray[0] is
1056 * initialized to indicate if reshape is being performed at the
1057 * container or subarray level
1058 */
1059 #define APPLY_METADATA_CHANGES 1
1060 #define ROLLBACK_METADATA_CHANGES 0
1061
1062 int (*reshape_super)(struct supertype *st,
1063 unsigned long long size, int level,
1064 int layout, int chunksize, int raid_disks,
1065 int delta_disks, char *backup, char *dev,
1066 int direction,
1067 int verbose); /* optional */
1068 int (*manage_reshape)( /* optional */
1069 int afd, struct mdinfo *sra, struct reshape *reshape,
1070 struct supertype *st, unsigned long blocks,
1071 int *fds, unsigned long long *offsets,
1072 int dests, int *destfd, unsigned long long *destoffsets);
1073
1074 /* for mdmon */
1075 int (*open_new)(struct supertype *c, struct active_array *a,
1076 char *inst);
1077
1078 /* Tell the metadata handler the current state of the array.
1079 * This covers whether it is known to be consistent (no pending writes)
1080 * and how far along a resync is known to have progressed
1081 * (in a->resync_start).
1082 * resync status is really irrelevant if the array is not consistent,
1083 * but some metadata (DDF!) have a place to record the distinction.
1084 * If 'consistent' is '2', then the array can mark it dirty if a
1085 * resync/recovery/whatever is required, or leave it clean if not.
1086 * Return value is 0 dirty (not consistent) and 1 if clean.
1087 * it is only really important if consistent is passed in as '2'.
1088 */
1089 int (*set_array_state)(struct active_array *a, int consistent);
1090
1091 /* When the state of a device might have changed, we call set_disk to
1092 * tell the metadata what the current state is.
1093 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1094 * transitions.
1095 * set_disk might be called when the state of the particular disk has
1096 * not in fact changed.
1097 */
1098 void (*set_disk)(struct active_array *a, int n, int state);
1099 void (*sync_metadata)(struct supertype *st);
1100 void (*process_update)(struct supertype *st,
1101 struct metadata_update *update);
1102 /* Prepare updates allocates extra memory that might be
1103 * needed. If the update cannot be understood, return 0.
1104 */
1105 int (*prepare_update)(struct supertype *st,
1106 struct metadata_update *update);
1107
1108 /* activate_spare will check if the array is degraded and, if it
1109 * is, try to find some spare space in the container.
1110 * On success, it add appropriate updates (For process_update) to
1111 * to the 'updates' list and returns a list of 'mdinfo' identifying
1112 * the device, or devices as there might be multiple missing
1113 * devices and multiple spares available.
1114 */
1115 struct mdinfo *(*activate_spare)(struct active_array *a,
1116 struct metadata_update **updates);
1117 /*
1118 * Return statically allocated string that represents metadata specific
1119 * controller domain of the disk. The domain is used in disk domain
1120 * matching functions. Disks belong to the same domain if the they have
1121 * the same domain from mdadm.conf and belong the same metadata domain.
1122 * Returning NULL or not providing this handler means that metadata
1123 * does not distinguish the differences between disks that belong to
1124 * different controllers. They are in the domain specified by
1125 * configuration file (mdadm.conf).
1126 * In case when the metadata has the notion of domains based on disk
1127 * it shall return NULL for disks that do not belong to the controller
1128 * the supported domains. Such disks will form another domain and won't
1129 * be mixed with supported ones.
1130 */
1131 const char *(*get_disk_controller_domain)(const char *path);
1132
1133 /* for external backup area */
1134 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
1135
1136 /* validate container after assemble */
1137 int (*validate_container)(struct mdinfo *info);
1138
1139 /* write initial empty PPL on device */
1140 int (*write_init_ppl)(struct supertype *st, struct mdinfo *info, int fd);
1141
1142 /* validate ppl before assemble */
1143 int (*validate_ppl)(struct supertype *st, struct mdinfo *info,
1144 struct mdinfo *disk);
1145
1146 /* records new bad block in metadata */
1147 int (*record_bad_block)(struct active_array *a, int n,
1148 unsigned long long sector, int length);
1149
1150 /* clears bad block from metadata */
1151 int (*clear_bad_block)(struct active_array *a, int n,
1152 unsigned long long sector, int length);
1153
1154 /* get list of bad blocks from metadata */
1155 struct md_bb *(*get_bad_blocks)(struct active_array *a, int n);
1156
1157 int swapuuid; /* true if uuid is bigending rather than hostendian */
1158 int external;
1159 const char *name; /* canonical metadata name */
1160 } *superlist[];
1161
1162 extern struct superswitch super0, super1;
1163 extern struct superswitch super_imsm, super_ddf;
1164 extern struct superswitch mbr, gpt;
1165
1166 struct metadata_update {
1167 int len;
1168 char *buf;
1169 void *space; /* allocated space that monitor will use */
1170 void **space_list; /* list of allocated spaces that monitor can
1171 * use or that it returned.
1172 */
1173 struct metadata_update *next;
1174 };
1175
1176 /* A supertype holds a particular collection of metadata.
1177 * It identifies the metadata type by the superswitch, and the particular
1178 * sub-version of that metadata type.
1179 * metadata read in or created is stored in 'sb' and 'info'.
1180 * There are also fields used by mdmon to track containers.
1181 *
1182 * A supertype may refer to:
1183 * Just an array, possibly in a container
1184 * A container, not identifying any particular array
1185 * Info read from just one device, not yet fully describing the array/container.
1186 *
1187 *
1188 * A supertype is created by:
1189 * super_by_fd
1190 * guess_super
1191 * dup_super
1192 */
1193 struct supertype {
1194 struct superswitch *ss;
1195 int minor_version;
1196 int max_devs;
1197 char container_devnm[32]; /* devnm of container */
1198 void *sb;
1199 void *info;
1200 void *other; /* Hack used to convert v0.90 to v1.0 */
1201 unsigned long long devsize;
1202 unsigned long long data_offset; /* used by v1.x only */
1203 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1204 HW/firmware related incompatability to load metadata.
1205 Used when examining metadata to display content of disk
1206 when user has no hw/firmare compatible system.
1207 */
1208 struct metadata_update *updates;
1209 struct metadata_update **update_tail;
1210
1211 /* extra stuff used by mdmon */
1212 struct active_array *arrays;
1213 int sock; /* listen to external programs */
1214 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1215 * external:/md0/12
1216 */
1217 int devcnt;
1218 int retry_soon;
1219 int nodes;
1220 char *cluster_name;
1221
1222 struct mdinfo *devs;
1223
1224 };
1225
1226 extern struct supertype *super_by_fd(int fd, char **subarray);
1227 enum guess_types { guess_any, guess_array, guess_partitions };
1228 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1229 static inline struct supertype *guess_super(int fd) {
1230 return guess_super_type(fd, guess_any);
1231 }
1232 extern struct supertype *dup_super(struct supertype *st);
1233 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1234 extern int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep);
1235 extern int must_be_container(int fd);
1236 extern int dev_size_from_id(dev_t id, unsigned long long *size);
1237 extern int dev_sector_size_from_id(dev_t id, unsigned int *size);
1238 void wait_for(char *dev, int fd);
1239
1240 /*
1241 * Data structures for policy management.
1242 * Each device can have a policy structure that lists
1243 * various name/value pairs each possibly with a metadata associated.
1244 * The policy list is sorted by name/value/metadata
1245 */
1246 struct dev_policy {
1247 struct dev_policy *next;
1248 char *name; /* None of these strings are allocated. They are
1249 * all just references to strings which are known
1250 * to exist elsewhere.
1251 * name and metadata can be compared by address equality.
1252 */
1253 const char *metadata;
1254 const char *value;
1255 };
1256
1257 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1258
1259 /* iterate over the sublist starting at list, having the same
1260 * 'name' as 'list', and matching the given metadata (Where
1261 * NULL matches anything
1262 */
1263 #define pol_for_each(item, list, _metadata) \
1264 for (item = list; \
1265 item && item->name == list->name; \
1266 item = item->next) \
1267 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1268 ; else
1269
1270 /*
1271 * policy records read from mdadm are largely just name-value pairs.
1272 * The names are constants, not strdupped
1273 */
1274 struct pol_rule {
1275 struct pol_rule *next;
1276 char *type; /* rule_policy or rule_part */
1277 struct rule {
1278 struct rule *next;
1279 char *name;
1280 char *value;
1281 char *dups; /* duplicates of 'value' with a partNN appended */
1282 } *rule;
1283 };
1284
1285 extern char rule_policy[], rule_part[];
1286 extern char rule_path[], rule_type[];
1287 extern char type_part[], type_disk[];
1288
1289 extern void policyline(char *line, char *type);
1290 extern void policy_add(char *type, ...);
1291 extern void policy_free(void);
1292
1293 extern struct dev_policy *path_policy(char **paths, char *type);
1294 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1295 extern struct dev_policy *devid_policy(int devid);
1296 extern void dev_policy_free(struct dev_policy *p);
1297
1298 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1299 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1300 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1301
1302 enum policy_action {
1303 act_default,
1304 act_include,
1305 act_re_add,
1306 act_spare, /* This only applies to bare devices */
1307 act_spare_same_slot, /* this allows non-bare devices,
1308 * but only if recent removal */
1309 act_force_spare, /* this allow non-bare devices in any case */
1310 act_err
1311 };
1312
1313 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1314 enum policy_action want);
1315 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1316 enum policy_action want);
1317
1318 struct domainlist {
1319 struct domainlist *next;
1320 const char *dom;
1321 };
1322
1323 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1324 const char *metadata);
1325 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1326 const char *metadata);
1327 extern void domainlist_add_dev(struct domainlist **dom, int devid,
1328 const char *metadata);
1329 extern void domain_free(struct domainlist *dl);
1330 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1331 const char *metadata);
1332 void domain_add(struct domainlist **domp, char *domain);
1333
1334 extern void policy_save_path(char *id_path, struct map_ent *array);
1335 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1336
1337 extern void sysfs_rules_apply(char *devnm, struct mdinfo *dev);
1338 extern void sysfsline(char *line);
1339
1340 #if __GNUC__ < 3
1341 struct stat64;
1342 #endif
1343
1344 #define HAVE_NFTW we assume
1345 #define HAVE_FTW
1346
1347 #ifdef __UCLIBC__
1348 # include <features.h>
1349 # ifndef __UCLIBC_HAS_LFS__
1350 # define lseek64 lseek
1351 # endif
1352 # ifndef __UCLIBC_HAS_FTW__
1353 # undef HAVE_FTW
1354 # undef HAVE_NFTW
1355 # endif
1356 #endif
1357
1358 #ifdef __dietlibc__
1359 # undef HAVE_NFTW
1360 #endif
1361
1362 #if defined(__KLIBC__)
1363 # undef HAVE_NFTW
1364 # undef HAVE_FTW
1365 #endif
1366
1367 #ifndef HAVE_NFTW
1368 # define FTW_PHYS 1
1369 # ifndef HAVE_FTW
1370 struct FTW {};
1371 # endif
1372 #endif
1373
1374 #ifdef HAVE_FTW
1375 # include <ftw.h>
1376 #endif
1377
1378 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1379
1380 extern int Manage_ro(char *devname, int fd, int readonly);
1381 extern int Manage_run(char *devname, int fd, struct context *c);
1382 extern int Manage_stop(char *devname, int fd, int quiet,
1383 int will_retry);
1384 extern int Manage_subdevs(char *devname, int fd,
1385 struct mddev_dev *devlist, int verbose, int test,
1386 char *update, int force);
1387 extern int autodetect(void);
1388 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1389 extern int Grow_addbitmap(char *devname, int fd,
1390 struct context *c, struct shape *s);
1391 extern int Grow_reshape(char *devname, int fd,
1392 struct mddev_dev *devlist,
1393 unsigned long long data_offset,
1394 struct context *c, struct shape *s);
1395 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1396 int *fdlist, int cnt, char *backup_file, int verbose);
1397 extern int Grow_continue(int mdfd, struct supertype *st,
1398 struct mdinfo *info, char *backup_file,
1399 int forked, int freeze_reshape);
1400 extern int Grow_consistency_policy(char *devname, int fd,
1401 struct context *c, struct shape *s);
1402
1403 extern int restore_backup(struct supertype *st,
1404 struct mdinfo *content,
1405 int working_disks,
1406 int spares,
1407 char **backup_filep,
1408 int verbose);
1409 extern int Grow_continue_command(char *devname, int fd,
1410 char *backup_file, int verbose);
1411
1412 extern int Assemble(struct supertype *st, char *mddev,
1413 struct mddev_ident *ident,
1414 struct mddev_dev *devlist,
1415 struct context *c);
1416
1417 extern int Build(char *mddev, struct mddev_dev *devlist,
1418 struct shape *s, struct context *c);
1419
1420 extern int Create(struct supertype *st, char *mddev,
1421 char *name, int *uuid,
1422 int subdevs, struct mddev_dev *devlist,
1423 struct shape *s,
1424 struct context *c,
1425 unsigned long long data_offset);
1426
1427 extern int Detail(char *dev, struct context *c);
1428 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1429 extern int Query(char *dev);
1430 extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1431 extern int Examine(struct mddev_dev *devlist, struct context *c,
1432 struct supertype *forcest);
1433 extern int Monitor(struct mddev_dev *devlist,
1434 char *mailaddr, char *alert_cmd,
1435 struct context *c,
1436 int daemonise, int oneshot,
1437 int dosyslog, char *pidfile, int increments,
1438 int share);
1439
1440 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1441 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1442 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1443 extern int Wait(char *dev);
1444 extern int WaitClean(char *dev, int verbose);
1445 extern int SetAction(char *dev, char *action);
1446
1447 extern int Incremental(struct mddev_dev *devlist, struct context *c,
1448 struct supertype *st);
1449 extern void RebuildMap(void);
1450 extern int IncrementalScan(struct context *c, char *devnm);
1451 extern int IncrementalRemove(char *devname, char *path, int verbose);
1452 extern int CreateBitmap(char *filename, int force, char uuid[16],
1453 unsigned long chunksize, unsigned long daemon_sleep,
1454 unsigned long write_behind,
1455 unsigned long long array_size,
1456 int major);
1457 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1458 extern int IsBitmapDirty(char *filename);
1459 extern int Write_rules(char *rule_name);
1460 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1461
1462 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1463 static inline unsigned long long
1464 bitmap_bits(unsigned long long array_size, unsigned long chunksize)
1465 {
1466 return (array_size * 512 + chunksize - 1) / chunksize;
1467 }
1468
1469 extern int Dump_metadata(char *dev, char *dir, struct context *c,
1470 struct supertype *st);
1471 extern int Restore_metadata(char *dev, char *dir, struct context *c,
1472 struct supertype *st, int only);
1473
1474 int md_array_valid(int fd);
1475 int md_array_active(int fd);
1476 int md_array_is_active(struct mdinfo *info);
1477 int md_get_array_info(int fd, struct mdu_array_info_s *array);
1478 int md_set_array_info(int fd, struct mdu_array_info_s *array);
1479 int md_get_disk_info(int fd, struct mdu_disk_info_s *disk);
1480 extern int get_linux_version(void);
1481 extern int mdadm_version(char *version);
1482 extern unsigned long long parse_size(char *size);
1483 extern int parse_uuid(char *str, int uuid[4]);
1484 extern int is_near_layout_10(int layout);
1485 extern int parse_layout_10(char *layout);
1486 extern int parse_layout_faulty(char *layout);
1487 extern long parse_num(char *num);
1488 extern int parse_cluster_confirm_arg(char *inp, char **devname, int *slot);
1489 extern int check_ext2(int fd, char *name);
1490 extern int check_reiser(int fd, char *name);
1491 extern int check_raid(int fd, char *name);
1492 extern int check_partitions(int fd, char *dname,
1493 unsigned long long freesize,
1494 unsigned long long size);
1495 extern int fstat_is_blkdev(int fd, char *devname, dev_t *rdev);
1496 extern int stat_is_blkdev(char *devname, dev_t *rdev);
1497
1498 extern int get_mdp_major(void);
1499 extern int get_maj_min(char *dev, int *major, int *minor);
1500 extern int dev_open(char *dev, int flags);
1501 extern int open_dev(char *devnm);
1502 extern void reopen_mddev(int mdfd);
1503 extern int open_dev_flags(char *devnm, int flags);
1504 extern int open_dev_excl(char *devnm);
1505 extern int is_standard(char *dev, int *nump);
1506 extern int same_dev(char *one, char *two);
1507 extern int compare_paths (char* path1,char* path2);
1508 extern void enable_fds(int devices);
1509 extern void manage_fork_fds(int close_all);
1510 extern int continue_via_systemd(char *devnm, char *service_name);
1511
1512 extern int parse_auto(char *str, char *msg, int config);
1513 extern struct mddev_ident *conf_get_ident(char *dev);
1514 extern struct mddev_dev *conf_get_devs(void);
1515 extern int conf_test_dev(char *devname);
1516 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1517 extern struct createinfo *conf_get_create_info(void);
1518 extern void set_conffile(char *file);
1519 extern char *conf_get_mailaddr(void);
1520 extern char *conf_get_mailfrom(void);
1521 extern char *conf_get_program(void);
1522 extern char *conf_get_homehost(int *require_homehostp);
1523 extern char *conf_get_homecluster(void);
1524 extern char *conf_line(FILE *file);
1525 extern char *conf_word(FILE *file, int allow_key);
1526 extern void print_quoted(char *str);
1527 extern void print_escape(char *str);
1528 extern int use_udev(void);
1529 extern unsigned long GCD(unsigned long a, unsigned long b);
1530 extern int conf_name_is_free(char *name);
1531 extern int conf_verify_devnames(struct mddev_ident *array_list);
1532 extern int devname_matches(char *name, char *match);
1533 extern struct mddev_ident *conf_match(struct supertype *st,
1534 struct mdinfo *info,
1535 char *devname,
1536 int verbose, int *rvp);
1537
1538 extern void free_line(char *line);
1539 extern int match_oneof(char *devices, char *devname);
1540 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1541 extern const int uuid_zero[4];
1542 extern int same_uuid(int a[4], int b[4], int swapuuid);
1543 extern void copy_uuid(void *a, int b[4], int swapuuid);
1544 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1545 extern char *fname_from_uuid(struct supertype *st,
1546 struct mdinfo *info, char *buf, char sep);
1547 extern unsigned long calc_csum(void *super, int bytes);
1548 extern int enough(int level, int raid_disks, int layout, int clean,
1549 char *avail);
1550 extern int ask(char *mesg);
1551 extern unsigned long long get_component_size(int fd);
1552 extern void remove_partitions(int fd);
1553 extern int test_partition(int fd);
1554 extern int test_partition_from_id(dev_t id);
1555 extern int get_data_disks(int level, int layout, int raid_disks);
1556 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1557 int chunksize, unsigned long long devsize);
1558 extern int flush_metadata_updates(struct supertype *st);
1559 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1560 extern int assemble_container_content(struct supertype *st, int mdfd,
1561 struct mdinfo *content,
1562 struct context *c,
1563 char *chosen_name, int *result);
1564 #define INCR_NO 1
1565 #define INCR_UNSAFE 2
1566 #define INCR_ALREADY 4
1567 #define INCR_YES 8
1568 extern struct mdinfo *container_choose_spares(struct supertype *st,
1569 struct spare_criteria *criteria,
1570 struct domainlist *domlist,
1571 char *spare_group,
1572 const char *metadata, int get_one);
1573 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1574 extern int add_disk(int mdfd, struct supertype *st,
1575 struct mdinfo *sra, struct mdinfo *info);
1576 extern int remove_disk(int mdfd, struct supertype *st,
1577 struct mdinfo *sra, struct mdinfo *info);
1578 extern int hot_remove_disk(int mdfd, unsigned long dev, int force);
1579 extern int sys_hot_remove_disk(int statefd, int force);
1580 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1581 unsigned long long min_recovery_start(struct mdinfo *array);
1582
1583 extern char *human_size(long long bytes);
1584 extern char *human_size_brief(long long bytes, int prefix);
1585 extern void print_r10_layout(int layout);
1586
1587 extern char *find_free_devnm(int use_partitions);
1588
1589 extern void put_md_name(char *name);
1590 extern char *devid2kname(dev_t devid);
1591 extern char *devid2devnm(dev_t devid);
1592 extern dev_t devnm2devid(char *devnm);
1593 extern char *get_md_name(char *devnm);
1594
1595 extern char DefaultConfFile[];
1596
1597 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1598 char *chosen, int block_udev);
1599 /* values for 'trustworthy' */
1600 #define LOCAL 1
1601 #define LOCAL_ANY 10
1602 #define FOREIGN 2
1603 #define METADATA 3
1604 extern int open_mddev(char *dev, int report_errors);
1605 extern int open_container(int fd);
1606 extern int metadata_container_matches(char *metadata, char *devnm);
1607 extern int metadata_subdev_matches(char *metadata, char *devnm);
1608 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1609 extern int is_subarray_active(char *subarray, char *devname);
1610 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1611 extern struct superswitch *version_to_superswitch(char *vers);
1612
1613 extern int mdmon_running(char *devnm);
1614 extern int mdmon_pid(char *devnm);
1615 extern int check_env(char *name);
1616 extern __u32 random32(void);
1617 extern void random_uuid(__u8 *buf);
1618 extern int start_mdmon(char *devnm);
1619
1620 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1621 struct supertype *st, unsigned long stripes,
1622 int *fds, unsigned long long *offsets,
1623 int dests, int *destfd, unsigned long long *destoffsets);
1624 void abort_reshape(struct mdinfo *sra);
1625
1626 void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1627
1628 extern char *stat2kname(struct stat *st);
1629 extern char *fd2kname(int fd);
1630 extern char *stat2devnm(struct stat *st);
1631 extern char *fd2devnm(int fd);
1632 extern void udev_block(char *devnm);
1633 extern void udev_unblock(void);
1634
1635 extern int in_initrd(void);
1636
1637 struct cmap_hooks {
1638 void *cmap_handle; /* corosync lib related */
1639
1640 int (*initialize)(cmap_handle_t *handle);
1641 int (*get_string)(cmap_handle_t handle,
1642 const char *string,
1643 char **name);
1644 int (*finalize)(cmap_handle_t handle);
1645 };
1646
1647 extern void set_cmap_hooks(void);
1648 extern void set_hooks(void);
1649
1650 struct dlm_hooks {
1651 void *dlm_handle; /* dlm lib related */
1652
1653 dlm_lshandle_t (*create_lockspace)(const char *name,
1654 unsigned int mode);
1655 dlm_lshandle_t (*open_lockspace)(const char *name);
1656 int (*release_lockspace)(const char *name, dlm_lshandle_t ls,
1657 int force);
1658 int (*ls_lock)(dlm_lshandle_t lockspace, uint32_t mode,
1659 struct dlm_lksb *lksb, uint32_t flags,
1660 const void *name, unsigned int namelen,
1661 uint32_t parent, void (*astaddr) (void *astarg),
1662 void *astarg, void (*bastaddr) (void *astarg),
1663 void *range);
1664 int (*ls_unlock_wait)(dlm_lshandle_t lockspace, uint32_t lkid,
1665 uint32_t flags, struct dlm_lksb *lksb);
1666 int (*ls_get_fd)(dlm_lshandle_t ls);
1667 int (*dispatch)(int fd);
1668 };
1669
1670 extern int get_cluster_name(char **name);
1671 extern int dlm_funs_ready(void);
1672 extern int cluster_get_dlmlock(void);
1673 extern int cluster_release_dlmlock(void);
1674 extern void set_dlm_hooks(void);
1675
1676 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1677 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1678 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1679 (ROUND_UP((unsigned long)(ptr), base)))
1680
1681 static inline int is_subarray(char *vers)
1682 {
1683 /* The version string for a 'subarray' (an array in a container)
1684 * is
1685 * /containername/componentname for normal read-write arrays
1686 * -containername/componentname for arrays which mdmon must not
1687 * reconfigure. They might be read-only
1688 * or might be undergoing reshape etc.
1689 * containername is e.g. md0, md_d1
1690 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1691 */
1692 return (*vers == '/' || *vers == '-');
1693 }
1694
1695 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1696 {
1697 return &ent->metadata_version[10+strlen(container)+1];
1698 }
1699
1700 #ifdef DEBUG
1701 #define dprintf(fmt, arg...) \
1702 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1703 #define dprintf_cont(fmt, arg...) \
1704 fprintf(stderr, fmt, ##arg)
1705 #else
1706 #define dprintf(fmt, arg...) \
1707 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1708 #define dprintf_cont(fmt, arg...) \
1709 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1710 #endif
1711 #include <assert.h>
1712 #include <stdarg.h>
1713 static inline int xasprintf(char **strp, const char *fmt, ...) {
1714 va_list ap;
1715 int ret;
1716 va_start(ap, fmt);
1717 ret = vasprintf(strp, fmt, ap);
1718 va_end(ap);
1719 assert(ret >= 0);
1720 return ret;
1721 }
1722
1723 #ifdef DEBUG
1724 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1725 #else
1726 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1727 #endif
1728 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1729
1730 void *xmalloc(size_t len);
1731 void *xrealloc(void *ptr, size_t len);
1732 void *xcalloc(size_t num, size_t size);
1733 char *xstrdup(const char *str);
1734
1735 #define LEVEL_MULTIPATH (-4)
1736 #define LEVEL_LINEAR (-1)
1737 #define LEVEL_FAULTY (-5)
1738
1739 /* kernel module doesn't know about these */
1740 #define LEVEL_CONTAINER (-100)
1741 #define LEVEL_UNSUPPORTED (-200)
1742
1743 /* the kernel does know about this one ... */
1744 #define LEVEL_NONE (-1000000)
1745
1746 /* faulty stuff */
1747
1748 #define WriteTransient 0
1749 #define ReadTransient 1
1750 #define WritePersistent 2
1751 #define ReadPersistent 3
1752 #define WriteAll 4 /* doesn't go to device */
1753 #define ReadFixable 5
1754 #define Modes 6
1755
1756 #define ClearErrors 31
1757 #define ClearFaults 30
1758
1759 #define AllPersist 100 /* internal use only */
1760 #define NoPersist 101
1761
1762 #define ModeMask 0x1f
1763 #define ModeShift 5
1764
1765 #ifdef __TINYC__
1766 #undef minor
1767 #undef major
1768 #undef makedev
1769 #define minor(x) ((x)&0xff)
1770 #define major(x) (((x)>>8)&0xff)
1771 #define makedev(M,m) (((M)<<8) | (m))
1772 #endif
1773
1774 enum r0layout {
1775 RAID0_ORIG_LAYOUT = 1,
1776 RAID0_ALT_MULTIZONE_LAYOUT = 2,
1777 };
1778
1779 /* for raid4/5/6 */
1780 #define ALGORITHM_LEFT_ASYMMETRIC 0
1781 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1782 #define ALGORITHM_LEFT_SYMMETRIC 2
1783 #define ALGORITHM_RIGHT_SYMMETRIC 3
1784
1785 /* Define non-rotating (raid4) algorithms. These allow
1786 * conversion of raid4 to raid5.
1787 */
1788 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1789 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1790
1791 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1792 * Firstly, the exact positioning of the parity block is slightly
1793 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1794 * of DDF.
1795 * Secondly, or order of datablocks over which the Q syndrome is computed
1796 * is different.
1797 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1798 * These layouts are from the DDFv1.2 spec.
1799 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1800 * leaves RLQ=3 as 'Vendor Specific'
1801 */
1802
1803 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1804 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1805 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1806
1807 /* For every RAID5 algorithm we define a RAID6 algorithm
1808 * with exactly the same layout for data and parity, and
1809 * with the Q block always on the last device (N-1).
1810 * This allows trivial conversion from RAID5 to RAID6
1811 */
1812 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1813 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1814 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1815 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1816 #define ALGORITHM_PARITY_0_6 20
1817 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1818
1819 /* Define PATH_MAX in case we don't use glibc or standard library does
1820 * not have PATH_MAX defined. Assume max path length is 4K characters.
1821 */
1822 #ifndef PATH_MAX
1823 #define PATH_MAX 4096
1824 #endif
1825
1826 #define RESYNC_NONE -1
1827 #define RESYNC_DELAYED -2
1828 #define RESYNC_PENDING -3
1829 #define RESYNC_REMOTE -4
1830 #define RESYNC_UNKNOWN -5
1831
1832 /* When using "GET_DISK_INFO" it isn't certain how high
1833 * we need to check. So we impose an absolute limit of
1834 * MAX_DISKS. This needs to be much more than the largest
1835 * number of devices any metadata can support. Currently
1836 * v1.x can support 1920
1837 */
1838 #define MAX_DISKS 4096
1839
1840 /* Sometimes the 'size' value passed needs to mean "Maximum".
1841 * In those cases with use MAX_SIZE
1842 */
1843 #define MAX_SIZE 1
1844
1845 /* We want to use unsigned numbers for sector counts, but need
1846 * a value for 'invalid'. Use '1'.
1847 */
1848 #define INVALID_SECTORS 1
1849 /* And another special number needed for --data_offset=variable */
1850 #define VARIABLE_OFFSET 3