705bd9b53b3ae28e9cd403640a81e4705d4d467e
[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 #ifdef __GLIBC__
49 /* Newer glibc requires sys/sysmacros.h directly for makedev() */
50 #include <sys/sysmacros.h>
51 #endif
52 #ifdef __dietlibc__
53 #include <strings.h>
54 /* dietlibc has deprecated random and srandom!! */
55 #define random rand
56 #define srandom srand
57 #endif
58
59 #ifdef NO_COROSYNC
60 #define CS_OK 1
61 typedef uint64_t cmap_handle_t;
62 #else
63 #include <corosync/cmap.h>
64 #endif
65
66 #ifndef NO_DLM
67 #include <libdlm.h>
68 #include <errno.h>
69 #else
70 #define LKF_NOQUEUE 0x00000001
71 #define LKM_PWMODE 4
72 #define EUNLOCK 0x10002
73
74 typedef void *dlm_lshandle_t;
75
76 struct dlm_lksb {
77 int sb_status;
78 uint32_t sb_lkid;
79 char sb_flags;
80 char *sb_lvbptr;
81 };
82 #endif
83
84 #include <linux/kdev_t.h>
85 /*#include <linux/fs.h> */
86 #include <sys/mount.h>
87 #include <asm/types.h>
88 #include <sys/ioctl.h>
89 #define MD_MAJOR 9
90 #define MdpMinorShift 6
91
92 #ifndef BLKGETSIZE64
93 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
94 #endif
95
96 #define DEFAULT_CHUNK 512
97 #define DEFAULT_BITMAP_CHUNK 4096
98 #define DEFAULT_BITMAP_DELAY 5
99 #define DEFAULT_MAX_WRITE_BEHIND 256
100
101 /* MAP_DIR should be somewhere that persists across the pivotroot
102 * from early boot to late boot.
103 * /run seems to have emerged as the best standard.
104 */
105 #ifndef MAP_DIR
106 #define MAP_DIR "/run/mdadm"
107 #endif /* MAP_DIR */
108 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
109 * want something other than the default of "map"
110 */
111 #ifndef MAP_FILE
112 #define MAP_FILE "map"
113 #endif /* MAP_FILE */
114 /* MDMON_DIR is where pid and socket files used for communicating
115 * with mdmon normally live. Best is /var/run/mdadm as
116 * mdmon is needed at early boot then it needs to write there prior
117 * to /var/run being mounted read/write, and it also then needs to
118 * persist beyond when /var/run is mounter read-only. So, to be
119 * safe, the default is somewhere that is read/write early in the
120 * boot process and stays up as long as possible during shutdown.
121 */
122 #ifndef MDMON_DIR
123 #define MDMON_DIR "/run/mdadm"
124 #endif /* MDMON_DIR */
125
126 /* FAILED_SLOTS is where to save files storing recent removal of array
127 * member in order to allow future reuse of disk inserted in the same
128 * slot for array recovery
129 */
130 #ifndef FAILED_SLOTS_DIR
131 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
132 #endif /* FAILED_SLOTS */
133
134 #include "md_u.h"
135 #include "md_p.h"
136 #include "bitmap.h"
137 #include "msg.h"
138
139 #include <endian.h>
140 /* Redhat don't like to #include <asm/byteorder.h>, and
141 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
142 * and there is no standard conversion function so... */
143 /* And dietlibc doesn't think byteswap is ok, so.. */
144 /* #include <byteswap.h> */
145 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
146 ((x) & 0xff00U) >> 8)
147 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
148 ((x) & 0xff000000U) >> 24 | \
149 ((x) & 0x0000ff00U) << 8 | \
150 ((x) & 0x00ff0000U) >> 8)
151 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
152 ((x) & 0xff00000000000000ULL) >> 56 | \
153 ((x) & 0x000000000000ff00ULL) << 40 | \
154 ((x) & 0x00ff000000000000ULL) >> 40 | \
155 ((x) & 0x0000000000ff0000ULL) << 24 | \
156 ((x) & 0x0000ff0000000000ULL) >> 24 | \
157 ((x) & 0x00000000ff000000ULL) << 8 | \
158 ((x) & 0x000000ff00000000ULL) >> 8)
159
160 #if !defined(__KLIBC__)
161 #if BYTE_ORDER == LITTLE_ENDIAN
162 #define __cpu_to_le16(_x) (unsigned int)(_x)
163 #define __cpu_to_le32(_x) (unsigned int)(_x)
164 #define __cpu_to_le64(_x) (unsigned long long)(_x)
165 #define __le16_to_cpu(_x) (unsigned int)(_x)
166 #define __le32_to_cpu(_x) (unsigned int)(_x)
167 #define __le64_to_cpu(_x) (unsigned long long)(_x)
168
169 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
170 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
171 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
172 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
173 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
174 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
175 #elif BYTE_ORDER == BIG_ENDIAN
176 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
177 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
178 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
179 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
180 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
181 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
182
183 #define __cpu_to_be16(_x) (unsigned int)(_x)
184 #define __cpu_to_be32(_x) (unsigned int)(_x)
185 #define __cpu_to_be64(_x) (unsigned long long)(_x)
186 #define __be16_to_cpu(_x) (unsigned int)(_x)
187 #define __be32_to_cpu(_x) (unsigned int)(_x)
188 #define __be64_to_cpu(_x) (unsigned long long)(_x)
189 #else
190 # error "unknown endianness."
191 #endif
192 #endif /* __KLIBC__ */
193
194 /*
195 * Check at compile time that something is of a particular type.
196 * Always evaluates to 1 so you may use it easily in comparisons.
197 */
198
199 #define typecheck(type,x) \
200 ({ type __dummy; \
201 typeof(x) __dummy2; \
202 (void)(&__dummy == &__dummy2); \
203 1; \
204 })
205
206 /*
207 * These inlines deal with timer wrapping correctly.
208 *
209 * time_after(a,b) returns true if the time a is after time b.
210 */
211
212 #define time_after(a,b) \
213 (typecheck(unsigned int, a) && \
214 typecheck(unsigned int, b) && \
215 ((int)((b) - (a)) < 0))
216
217 #define time_before(a,b) time_after(b,a)
218
219 /*
220 * min()/max()/clamp() macros that also do
221 * strict type-checking.. See the
222 * "unnecessary" pointer comparison.
223 */
224 #define min(x, y) ({ \
225 typeof(x) _min1 = (x); \
226 typeof(y) _min2 = (y); \
227 (void) (&_min1 == &_min2); \
228 _min1 < _min2 ? _min1 : _min2; })
229
230 #define max(x, y) ({ \
231 typeof(x) _max1 = (x); \
232 typeof(y) _max2 = (y); \
233 (void) (&_max1 == &_max2); \
234 _max1 > _max2 ? _max1 : _max2; })
235
236 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
237
238 extern const char Name[];
239
240 struct md_bb_entry {
241 unsigned long long sector;
242 int length;
243 };
244
245 struct md_bb {
246 int supported;
247 int count;
248 struct md_bb_entry *entries;
249 };
250
251 /* general information that might be extracted from a superblock */
252 struct mdinfo {
253 mdu_array_info_t array;
254 mdu_disk_info_t disk;
255 __u64 events;
256 int uuid[4];
257 char name[33];
258 unsigned long long data_offset;
259 unsigned long long new_data_offset;
260 unsigned long long component_size; /* same as array.size, except in
261 * sectors and up to 64bits.
262 */
263 unsigned long long custom_array_size; /* size for non-default sized
264 * arrays (in sectors)
265 */
266 #define NO_RESHAPE 0
267 #define VOLUME_RESHAPE 1
268 #define CONTAINER_RESHAPE 2
269 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
270 int reshape_active;
271 unsigned long long reshape_progress;
272 int recovery_blocked; /* for external metadata it
273 * indicates that there is
274 * reshape in progress in
275 * container,
276 * for native metadata it is
277 * reshape_active field mirror
278 */
279 int journal_device_required;
280 int journal_clean;
281
282 enum {
283 CONSISTENCY_POLICY_UNKNOWN,
284 CONSISTENCY_POLICY_NONE,
285 CONSISTENCY_POLICY_RESYNC,
286 CONSISTENCY_POLICY_BITMAP,
287 CONSISTENCY_POLICY_JOURNAL,
288 CONSISTENCY_POLICY_PPL,
289 } consistency_policy;
290
291 /* During reshape we can sometimes change the data_offset to avoid
292 * over-writing still-valid data. We need to know if there is space.
293 * So getinfo_super will fill in space_before and space_after in sectors.
294 * data_offset can be increased or decreased by this amount.
295 */
296 unsigned long long space_before, space_after;
297 union {
298 unsigned long long resync_start; /* per-array resync position */
299 unsigned long long recovery_start; /* per-device rebuild position */
300 #define MaxSector (~0ULL) /* resync/recovery complete position */
301 };
302 long bitmap_offset; /* 0 == none, 1 == a file */
303 unsigned int ppl_size;
304 int ppl_offset;
305 unsigned long long ppl_sector;
306 unsigned long safe_mode_delay; /* ms delay to mark clean */
307 int new_level, delta_disks, new_layout, new_chunk;
308 int errors;
309 unsigned long cache_size; /* size of raid456 stripe cache*/
310 int mismatch_cnt;
311 char text_version[50];
312
313 int container_member; /* for assembling external-metatdata arrays
314 * This is to be used internally by metadata
315 * handler only */
316 int container_enough; /* flag external handlers can set to
317 * indicate that subarrays have not enough (-1),
318 * enough to start (0), or all expected disks (1) */
319 char sys_name[32];
320 struct mdinfo *devs;
321 struct mdinfo *next;
322
323 /* Device info for mdmon: */
324 int recovery_fd;
325 int state_fd;
326 int bb_fd;
327 int ubb_fd;
328 #define DS_FAULTY 1
329 #define DS_INSYNC 2
330 #define DS_WRITE_MOSTLY 4
331 #define DS_SPARE 8
332 #define DS_BLOCKED 16
333 #define DS_REMOVE 1024
334 #define DS_UNBLOCK 2048
335 int prev_state, curr_state, next_state;
336
337 /* info read from sysfs */
338 enum {
339 ARRAY_CLEAR,
340 ARRAY_INACTIVE,
341 ARRAY_SUSPENDED,
342 ARRAY_READONLY,
343 ARRAY_READ_AUTO,
344 ARRAY_CLEAN,
345 ARRAY_ACTIVE,
346 ARRAY_WRITE_PENDING,
347 ARRAY_ACTIVE_IDLE,
348 ARRAY_UNKNOWN_STATE,
349 } array_state;
350 struct md_bb bb;
351 };
352
353 struct createinfo {
354 int uid;
355 int gid;
356 int autof;
357 int mode;
358 int symlinks;
359 int names;
360 int bblist;
361 struct supertype *supertype;
362 };
363
364 struct spare_criteria {
365 unsigned long long min_size;
366 unsigned int sector_size;
367 };
368
369 enum mode {
370 ASSEMBLE=1,
371 BUILD,
372 CREATE,
373 MANAGE,
374 MISC,
375 MONITOR,
376 GROW,
377 INCREMENTAL,
378 AUTODETECT,
379 mode_count
380 };
381
382 extern char short_options[];
383 extern char short_bitmap_options[];
384 extern char short_bitmap_auto_options[];
385 extern struct option long_options[];
386 extern char Version[], Usage[], Help[], OptionHelp[],
387 *mode_help[],
388 Help_create[], Help_build[], Help_assemble[], Help_grow[],
389 Help_incr[],
390 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
391
392 /* for option that don't have short equivilents, we assign arbitrary
393 * numbers later than any 'short' character option.
394 */
395 enum special_options {
396 AssumeClean = 300,
397 BitmapChunk,
398 WriteBehind,
399 ReAdd,
400 NoDegraded,
401 Sparc22,
402 BackupFile,
403 HomeHost,
404 AutoHomeHost,
405 Symlinks,
406 AutoDetect,
407 Waitclean,
408 DetailPlatform,
409 KillSubarray,
410 UpdateSubarray,
411 IncrementalPath,
412 NoSharing,
413 HelpOptions,
414 Brief,
415 ManageOpt,
416 Add,
417 AddSpare,
418 AddJournal,
419 Remove,
420 Fail,
421 Replace,
422 With,
423 MiscOpt,
424 WaitOpt,
425 ConfigFile,
426 ChunkSize,
427 WriteMostly,
428 FailFast,
429 NoFailFast,
430 Layout,
431 Auto,
432 Force,
433 SuperMinor,
434 EMail,
435 ProgramOpt,
436 Increment,
437 Fork,
438 Bitmap,
439 RebuildMapOpt,
440 InvalidBackup,
441 UdevRules,
442 FreezeReshape,
443 Continue,
444 OffRootOpt,
445 Prefer,
446 KillOpt,
447 DataOffset,
448 ExamineBB,
449 Dump,
450 Restore,
451 Action,
452 Nodes,
453 ClusterName,
454 ClusterConfirm,
455 WriteJournal,
456 ConsistencyPolicy,
457 };
458
459 enum prefix_standard {
460 JEDEC,
461 IEC
462 };
463
464 enum bitmap_update {
465 NoUpdate,
466 NameUpdate,
467 NodeNumUpdate,
468 };
469
470 enum flag_mode {
471 FlagDefault, FlagSet, FlagClear,
472 };
473
474 /* structures read from config file */
475 /* List of mddevice names and identifiers
476 * Identifiers can be:
477 * uuid=128-hex-uuid
478 * super-minor=decimal-minor-number-from-superblock
479 * devices=comma,separated,list,of,device,names,with,wildcards
480 *
481 * If multiple fields are present, the intersection of all matching
482 * devices is considered
483 */
484 #define UnSet (0xfffe)
485 struct mddev_ident {
486 char *devname;
487
488 int uuid_set;
489 int uuid[4];
490 char name[33];
491
492 int super_minor;
493
494 char *devices; /* comma separated list of device
495 * names with wild cards
496 */
497 int level;
498 int raid_disks;
499 int spare_disks;
500 struct supertype *st;
501 int autof; /* 1 for normal, 2 for partitioned */
502 char *spare_group;
503 char *bitmap_file;
504 int bitmap_fd;
505
506 char *container; /* /dev/whatever name of container, or
507 * uuid of container. You would expect
508 * this to be the 'devname' or UUID
509 * of some other entry.
510 */
511 char *member; /* subarray within a container */
512
513 struct mddev_ident *next;
514 union {
515 /* fields needed by different users of this structure */
516 int assembled; /* set when assembly succeeds */
517 };
518 };
519
520 struct context {
521 int readonly;
522 int runstop;
523 int verbose;
524 int brief;
525 int force;
526 char *homehost;
527 int require_homehost;
528 char *prefer;
529 int export;
530 int test;
531 char *subarray;
532 char *update;
533 int scan;
534 int SparcAdjust;
535 int autof;
536 int delay;
537 int freeze_reshape;
538 char *backup_file;
539 int invalid_backup;
540 char *action;
541 int nodes;
542 char *homecluster;
543 };
544
545 struct shape {
546 int raiddisks;
547 int sparedisks;
548 int journaldisks;
549 int level;
550 int layout;
551 char *layout_str;
552 int chunk;
553 int bitmap_chunk;
554 char *bitmap_file;
555 int assume_clean;
556 int write_behind;
557 unsigned long long size;
558 int consistency_policy;
559 };
560
561 /* List of device names - wildcards expanded */
562 struct mddev_dev {
563 char *devname;
564 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
565 * 'A' for re_add.
566 * Not set for names read from .config
567 */
568 enum flag_mode writemostly;
569 enum flag_mode failfast;
570 int used; /* set when used */
571 long long data_offset;
572 struct mddev_dev *next;
573 };
574
575 typedef struct mapping {
576 char *name;
577 int num;
578 } mapping_t;
579
580 struct mdstat_ent {
581 char devnm[32];
582 int active;
583 char *level;
584 char *pattern; /* U for up, _ for down */
585 int percent; /* -1 if no resync */
586 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
587 int devcnt;
588 int raid_disks;
589 char * metadata_version;
590 struct dev_member {
591 char *name;
592 struct dev_member *next;
593 } *members;
594 struct mdstat_ent *next;
595 };
596
597 extern struct mdstat_ent *mdstat_read(int hold, int start);
598 extern void mdstat_close(void);
599 extern void free_mdstat(struct mdstat_ent *ms);
600 extern void mdstat_wait(int seconds);
601 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
602 extern int mddev_busy(char *devnm);
603 extern struct mdstat_ent *mdstat_by_component(char *name);
604 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
605
606 struct map_ent {
607 struct map_ent *next;
608 char devnm[32];
609 char metadata[20];
610 int uuid[4];
611 int bad;
612 char *path;
613 };
614 extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
615 int uuid[4], char *path);
616 extern void map_remove(struct map_ent **map, char *devnm);
617 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
618 extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
619 extern void map_free(struct map_ent *map);
620 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
621 extern void map_read(struct map_ent **melp);
622 extern int map_write(struct map_ent *mel);
623 extern void map_delete(struct map_ent **mapp, char *devnm);
624 extern void map_add(struct map_ent **melp,
625 char *devnm, char *metadata, int uuid[4], char *path);
626 extern int map_lock(struct map_ent **melp);
627 extern void map_unlock(struct map_ent **melp);
628 extern void map_fork(void);
629
630 /* various details can be requested */
631 enum sysfs_read_flags {
632 GET_LEVEL = (1 << 0),
633 GET_LAYOUT = (1 << 1),
634 GET_COMPONENT = (1 << 2),
635 GET_CHUNK = (1 << 3),
636 GET_CACHE = (1 << 4),
637 GET_MISMATCH = (1 << 5),
638 GET_VERSION = (1 << 6),
639 GET_DISKS = (1 << 7),
640 GET_SAFEMODE = (1 << 9),
641 GET_BITMAP_LOCATION = (1 << 10),
642
643 GET_DEVS = (1 << 20), /* gets role, major, minor */
644 GET_OFFSET = (1 << 21),
645 GET_SIZE = (1 << 22),
646 GET_STATE = (1 << 23),
647 GET_ERROR = (1 << 24),
648 GET_ARRAY_STATE = (1 << 25),
649 GET_CONSISTENCY_POLICY = (1 << 26),
650 };
651
652 /* If fd >= 0, get the array it is open on,
653 * else use devnm.
654 */
655 extern int sysfs_open(char *devnm, char *devname, char *attr);
656 extern int sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
657 extern void sysfs_init_dev(struct mdinfo *mdi, dev_t devid);
658 extern void sysfs_free(struct mdinfo *sra);
659 extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
660 extern int sysfs_attr_match(const char *attr, const char *str);
661 extern int sysfs_match_word(const char *word, char **list);
662 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
663 char *name, char *val);
664 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
665 char *name, unsigned long long val);
666 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
667 char *name, long long val);
668 extern int sysfs_uevent(struct mdinfo *sra, char *event);
669 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
670 char *name);
671 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
672 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
673 char *name, unsigned long long *val);
674 extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
675 extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
676 char *name, unsigned long long *v1, unsigned long long *v2);
677 extern int sysfs_fd_get_str(int fd, char *val, int size);
678 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
679 char *name);
680 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
681 char *name, char *val, int size);
682 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
683 extern int sysfs_set_array(struct mdinfo *info, int vers);
684 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
685 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
686 extern int sysfs_unique_holder(char *devnm, long rdev);
687 extern int sysfs_freeze_array(struct mdinfo *sra);
688 extern int sysfs_wait(int fd, int *msec);
689 extern int load_sys(char *path, char *buf, int len);
690 extern int zero_disk_range(int fd, unsigned long long sector, size_t count);
691 extern int reshape_prepare_fdlist(char *devname,
692 struct mdinfo *sra,
693 int raid_disks,
694 int nrdisks,
695 unsigned long blocks,
696 char *backup_file,
697 int *fdlist,
698 unsigned long long *offsets);
699 extern void reshape_free_fdlist(int *fdlist,
700 unsigned long long *offsets,
701 int size);
702 extern int reshape_open_backup_file(char *backup,
703 int fd,
704 char *devname,
705 long blocks,
706 int *fdlist,
707 unsigned long long *offsets,
708 char *sysfs_name,
709 int restart);
710 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
711 unsigned int ndata, unsigned int odata);
712 extern char *locate_backup(char *name);
713 extern char *make_backup(char *name);
714
715 extern int save_stripes(int *source, unsigned long long *offsets,
716 int raid_disks, int chunk_size, int level, int layout,
717 int nwrites, int *dest,
718 unsigned long long start, unsigned long long length,
719 char *buf);
720 extern int restore_stripes(int *dest, unsigned long long *offsets,
721 int raid_disks, int chunk_size, int level, int layout,
722 int source, unsigned long long read_offset,
723 unsigned long long start, unsigned long long length,
724 char *src_buf);
725
726 #ifndef Sendmail
727 #define Sendmail "/usr/lib/sendmail -t"
728 #endif
729
730 #define SYSLOG_FACILITY LOG_DAEMON
731
732 extern char *map_num(mapping_t *map, int num);
733 extern int map_name(mapping_t *map, char *name);
734 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
735 extern mapping_t consistency_policies[], sysfs_array_states[];
736
737 extern char *map_dev_preferred(int major, int minor, int create,
738 char *prefer);
739 static inline char *map_dev(int major, int minor, int create)
740 {
741 return map_dev_preferred(major, minor, create, NULL);
742 }
743
744 struct active_array;
745 struct metadata_update;
746
747 /* 'struct reshape' records the intermediate states of
748 * a general reshape.
749 * The starting geometry is converted to the 'before' geometry
750 * by at most an atomic level change. They could be the same.
751 * Similarly the 'after' geometry is converted to the final
752 * geometry by at most a level change.
753 * Note that 'before' and 'after' must have the same level.
754 * 'blocks' is the minimum number of sectors for a reshape unit.
755 * This will be a multiple of the stripe size in each of the
756 * 'before' and 'after' geometries.
757 * If 'blocks' is 0, no restriping is necessary.
758 * 'min_offset_change' is the minimum change to data_offset to
759 * allow the reshape to happen. It is at least the larger of
760 * the old and new chunk sizes, and typically the same as 'blocks'
761 * divided by number of data disks.
762 */
763 struct reshape {
764 int level;
765 int parity; /* number of parity blocks/devices */
766 struct {
767 int layout;
768 int data_disks;
769 } before, after;
770 unsigned long long backup_blocks;
771 unsigned long long min_offset_change;
772 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
773 unsigned long long new_size; /* New size of array in sectors */
774 };
775
776 /* A superswitch provides entry point to a metadata handler.
777 *
778 * The superswitch primarily operates on some "metadata" that
779 * is accessed via the 'supertype'.
780 * This metadata has one of three possible sources.
781 * 1/ It is read from a single device. In this case it may not completely
782 * describe the array or arrays as some information might be on other
783 * devices.
784 * 2/ It is read from all devices in a container. In this case all
785 * information is present.
786 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
787 * be complete once enough ->add_to_super calls have completed.
788 *
789 * When creating an array inside a container, the metadata will be
790 * formed by a combination of 2 and 3. The metadata or the array is read,
791 * then new information is added.
792 *
793 * The metadata must sometimes have a concept of a 'current' array
794 * and a 'current' device.
795 * The 'current' array is set by init_super to be the newly created array,
796 * or is set by super_by_fd when it finds it is looking at an array inside
797 * a container.
798 *
799 * The 'current' device is either the device that the metadata was read from
800 * in case 1, or the last device added by add_to_super in case 3.
801 * Case 2 does not identify a 'current' device.
802 */
803 extern struct superswitch {
804
805 /* Used to report details of metadata read from a component
806 * device. ->load_super has been called.
807 */
808 void (*examine_super)(struct supertype *st, char *homehost);
809 void (*brief_examine_super)(struct supertype *st, int verbose);
810 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
811 void (*export_examine_super)(struct supertype *st);
812 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
813 int (*copy_metadata)(struct supertype *st, int from, int to);
814
815 /* Used to report details of an active array.
816 * ->load_super was possibly given a 'component' string.
817 */
818 void (*detail_super)(struct supertype *st, char *homehost);
819 void (*brief_detail_super)(struct supertype *st);
820 void (*export_detail_super)(struct supertype *st);
821
822 /* Optional: platform hardware / firmware details */
823 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
824 int (*export_detail_platform)(int verbose, char *controller_path);
825
826 /* Used:
827 * to get uuid to storing in bitmap metadata
828 * and 'reshape' backup-data metadata
829 * To see if a device is being re-added to an array it was part of.
830 */
831 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
832
833 /* Extract generic details from metadata. This could be details about
834 * the container, or about an individual array within the container.
835 * The determination is made either by:
836 * load_super being given a 'component' string.
837 * validate_geometry determining what to create.
838 * The info includes both array information and device information.
839 * The particular device should be:
840 * The last device added by add_to_super
841 * The device the metadata was loaded from by load_super
842 * If 'map' is present, then it is an array raid_disks long
843 * (raid_disk must already be set and correct) and it is filled
844 * with 1 for slots that are thought to be active and 0 for slots which
845 * appear to be failed/missing.
846 * *info is zeroed out before data is added.
847 */
848 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
849 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
850 /* Check if the given metadata is flagged as belonging to "this"
851 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
852 */
853 int (*match_home)(struct supertype *st, char *homehost);
854
855 /* Make one of several generic modifications to metadata
856 * prior to assembly (or other times).
857 * sparc2.2 - first bug in early 0.90 metadata
858 * super-minor - change name of 0.90 metadata
859 * summaries - 'correct' any redundant data
860 * resync - mark array as dirty to trigger a resync.
861 * uuid - set new uuid - only 0.90 or 1.x
862 * name - change the name of the array (where supported)
863 * homehost - change which host this array is tied to.
864 * devicesize - If metadata is at start of device, change recorded
865 * device size to match actual device size
866 * byteorder - swap bytes for 0.90 metadata
867 *
868 * force-one - mark that device as uptodate, not old or failed.
869 * force-array - mark array as clean if it would not otherwise
870 * assemble
871 * assemble - not sure how this is different from force-one...
872 * linear-grow-new - add a new device to a linear array, but don't
873 * change the size: so superblock still matches
874 * linear-grow-update - now change the size of the array.
875 * writemostly - set the WriteMostly1 bit in the superblock devflags
876 * readwrite - clear the WriteMostly1 bit in the superblock devflags
877 * failfast - set the FailFast1 bit in the superblock
878 * nofailfast - clear the FailFast1 bit
879 * no-bitmap - clear any record that a bitmap is present.
880 * bbl - add a bad-block-log if possible
881 * no-bbl - remove any bad-block-log is it is empty.
882 * force-no-bbl - remove any bad-block-log even if empty.
883 * revert-reshape - If a reshape is in progress, modify metadata so
884 * it will resume going in the opposite direction.
885 */
886 int (*update_super)(struct supertype *st, struct mdinfo *info,
887 char *update,
888 char *devname, int verbose,
889 int uuid_set, char *homehost);
890
891 /* Create new metadata for new array as described. This could
892 * be a new container, or an array in a pre-existing container.
893 * Also used to zero metadata prior to writing it to invalidate old
894 * metadata.
895 */
896 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
897 struct shape *s, char *name,
898 char *homehost, int *uuid,
899 unsigned long long data_offset);
900
901 /* update the metadata to include new device, either at create or
902 * when hot-adding a spare.
903 */
904 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
905 int fd, char *devname,
906 unsigned long long data_offset);
907 /* update the metadata to delete a device,
908 * when hot-removing.
909 */
910 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
911
912 /* Write metadata to one device when fixing problems or adding
913 * a new device.
914 */
915 int (*store_super)(struct supertype *st, int fd);
916
917 /* Write all metadata for this array.
918 */
919 int (*write_init_super)(struct supertype *st);
920 /* Check if metadata read from one device is compatible with an array,
921 * used when assembling an array, or pseudo-assembling was with
922 * "--examine --brief"
923 * If "st" has not yet been loaded the superblock from, "tst" is
924 * moved in, otherwise the superblock in 'st' is compared with
925 * 'tst'.
926 */
927 int (*compare_super)(struct supertype *st, struct supertype *tst);
928 /* Load metadata from a single device. If 'devname' is not NULL
929 * print error messages as appropriate */
930 int (*load_super)(struct supertype *st, int fd, char *devname);
931 /* 'fd' is a 'container' md array - load array metadata from the
932 * whole container.
933 */
934 int (*load_container)(struct supertype *st, int fd, char *devname);
935 /* If 'arg' is a valid name of this metadata type, allocate and
936 * return a 'supertype' for the particular minor version */
937 struct supertype * (*match_metadata_desc)(char *arg);
938 /* If a device has the given size, and the data_offset has been
939 * requested - work out how much space is available for data.
940 * This involves adjusting for reserved space (e.g. bitmaps)
941 * and for any rounding.
942 * 'mdadm' only calls this for existing arrays where a possible
943 * spare is being added. However some super-handlers call it
944 * internally from validate_geometry when creating an array.
945 */
946 __u64 (*avail_size)(struct supertype *st, __u64 size,
947 unsigned long long data_offset);
948 /*
949 * Return spare criteria for array:
950 * - minimum disk size can be used in array;
951 * - sector size can be used in array.
952 * Return values: 0 - for success and -EINVAL on error.
953 */
954 int (*get_spare_criteria)(struct supertype *st,
955 struct spare_criteria *sc);
956 /* Find somewhere to put a bitmap - possibly auto-size it - and
957 * update the metadata to record this. The array may be newly
958 * created, in which case data_size may be updated, or it might
959 * already exist. Metadata handler can know if init_super
960 * has been called, but not write_init_super.
961 * 0: Success
962 * -Exxxx: On error
963 */
964 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
965 int delay, int write_behind,
966 unsigned long long size, int may_change, int major);
967 /* Seek 'fd' to start of write-intent-bitmap. Must be an
968 * md-native format bitmap
969 */
970 int (*locate_bitmap)(struct supertype *st, int fd, int node_num);
971 /* if add_internal_bitmap succeeded for existing array, this
972 * writes it out.
973 */
974 int (*write_bitmap)(struct supertype *st, int fd, enum bitmap_update update);
975 /* Free the superblock and any other allocated data */
976 void (*free_super)(struct supertype *st);
977
978 /* validate_geometry is called with an st returned by
979 * match_metadata_desc.
980 * It should check that the geometry described is compatible with
981 * the metadata type. It will be called repeatedly as devices
982 * added to validate changing size and new devices. If there are
983 * inter-device dependencies, it should record sufficient details
984 * so these can be validated.
985 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
986 * Return value is:
987 * 1: everything is OK
988 * 0: not OK for some reason - if 'verbose', then error was reported.
989 * -1: st->sb was NULL, 'subdev' is a member of a container of this
990 * type, but array is not acceptable for some reason
991 * message was reported even if verbose is 0.
992 */
993 int (*validate_geometry)(struct supertype *st, int level, int layout,
994 int raiddisks,
995 int *chunk, unsigned long long size,
996 unsigned long long data_offset,
997 char *subdev, unsigned long long *freesize,
998 int consistency_policy, int verbose);
999
1000 /* Return a linked list of 'mdinfo' structures for all arrays
1001 * in the container. For non-containers, it is like
1002 * getinfo_super with an allocated mdinfo.*/
1003 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
1004 /* query the supertype for default geometry */
1005 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
1006 /* Permit subarray's to be deleted from inactive containers */
1007 int (*kill_subarray)(struct supertype *st); /* optional */
1008 /* Permit subarray's to be modified */
1009 int (*update_subarray)(struct supertype *st, char *subarray,
1010 char *update, struct mddev_ident *ident); /* optional */
1011 /* Check if reshape is supported for this external format.
1012 * st is obtained from super_by_fd() where st->subarray[0] is
1013 * initialized to indicate if reshape is being performed at the
1014 * container or subarray level
1015 */
1016 #define APPLY_METADATA_CHANGES 1
1017 #define ROLLBACK_METADATA_CHANGES 0
1018
1019 int (*reshape_super)(struct supertype *st,
1020 unsigned long long size, int level,
1021 int layout, int chunksize, int raid_disks,
1022 int delta_disks, char *backup, char *dev,
1023 int direction,
1024 int verbose); /* optional */
1025 int (*manage_reshape)( /* optional */
1026 int afd, struct mdinfo *sra, struct reshape *reshape,
1027 struct supertype *st, unsigned long blocks,
1028 int *fds, unsigned long long *offsets,
1029 int dests, int *destfd, unsigned long long *destoffsets);
1030
1031 /* for mdmon */
1032 int (*open_new)(struct supertype *c, struct active_array *a,
1033 char *inst);
1034
1035 /* Tell the metadata handler the current state of the array.
1036 * This covers whether it is known to be consistent (no pending writes)
1037 * and how far along a resync is known to have progressed
1038 * (in a->resync_start).
1039 * resync status is really irrelevant if the array is not consistent,
1040 * but some metadata (DDF!) have a place to record the distinction.
1041 * If 'consistent' is '2', then the array can mark it dirty if a
1042 * resync/recovery/whatever is required, or leave it clean if not.
1043 * Return value is 0 dirty (not consistent) and 1 if clean.
1044 * it is only really important if consistent is passed in as '2'.
1045 */
1046 int (*set_array_state)(struct active_array *a, int consistent);
1047
1048 /* When the state of a device might have changed, we call set_disk to
1049 * tell the metadata what the current state is.
1050 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1051 * transitions.
1052 * set_disk might be called when the state of the particular disk has
1053 * not in fact changed.
1054 */
1055 void (*set_disk)(struct active_array *a, int n, int state);
1056 void (*sync_metadata)(struct supertype *st);
1057 void (*process_update)(struct supertype *st,
1058 struct metadata_update *update);
1059 /* Prepare updates allocates extra memory that might be
1060 * needed. If the update cannot be understood, return 0.
1061 */
1062 int (*prepare_update)(struct supertype *st,
1063 struct metadata_update *update);
1064
1065 /* activate_spare will check if the array is degraded and, if it
1066 * is, try to find some spare space in the container.
1067 * On success, it add appropriate updates (For process_update) to
1068 * to the 'updates' list and returns a list of 'mdinfo' identifying
1069 * the device, or devices as there might be multiple missing
1070 * devices and multiple spares available.
1071 */
1072 struct mdinfo *(*activate_spare)(struct active_array *a,
1073 struct metadata_update **updates);
1074 /*
1075 * Return statically allocated string that represents metadata specific
1076 * controller domain of the disk. The domain is used in disk domain
1077 * matching functions. Disks belong to the same domain if the they have
1078 * the same domain from mdadm.conf and belong the same metadata domain.
1079 * Returning NULL or not providing this handler means that metadata
1080 * does not distinguish the differences between disks that belong to
1081 * different controllers. They are in the domain specified by
1082 * configuration file (mdadm.conf).
1083 * In case when the metadata has the notion of domains based on disk
1084 * it shall return NULL for disks that do not belong to the controller
1085 * the supported domains. Such disks will form another domain and won't
1086 * be mixed with supported ones.
1087 */
1088 const char *(*get_disk_controller_domain)(const char *path);
1089
1090 /* for external backup area */
1091 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
1092
1093 /* validate container after assemble */
1094 int (*validate_container)(struct mdinfo *info);
1095
1096 /* write initial empty PPL on device */
1097 int (*write_init_ppl)(struct supertype *st, struct mdinfo *info, int fd);
1098
1099 /* validate ppl before assemble */
1100 int (*validate_ppl)(struct supertype *st, struct mdinfo *info,
1101 struct mdinfo *disk);
1102
1103 /* records new bad block in metadata */
1104 int (*record_bad_block)(struct active_array *a, int n,
1105 unsigned long long sector, int length);
1106
1107 /* clears bad block from metadata */
1108 int (*clear_bad_block)(struct active_array *a, int n,
1109 unsigned long long sector, int length);
1110
1111 /* get list of bad blocks from metadata */
1112 struct md_bb *(*get_bad_blocks)(struct active_array *a, int n);
1113
1114 int swapuuid; /* true if uuid is bigending rather than hostendian */
1115 int external;
1116 const char *name; /* canonical metadata name */
1117 } *superlist[];
1118
1119 extern struct superswitch super0, super1;
1120 extern struct superswitch super_imsm, super_ddf;
1121 extern struct superswitch mbr, gpt;
1122
1123 struct metadata_update {
1124 int len;
1125 char *buf;
1126 void *space; /* allocated space that monitor will use */
1127 void **space_list; /* list of allocated spaces that monitor can
1128 * use or that it returned.
1129 */
1130 struct metadata_update *next;
1131 };
1132
1133 /* A supertype holds a particular collection of metadata.
1134 * It identifies the metadata type by the superswitch, and the particular
1135 * sub-version of that metadata type.
1136 * metadata read in or created is stored in 'sb' and 'info'.
1137 * There are also fields used by mdmon to track containers.
1138 *
1139 * A supertype may refer to:
1140 * Just an array, possibly in a container
1141 * A container, not identifying any particular array
1142 * Info read from just one device, not yet fully describing the array/container.
1143 *
1144 *
1145 * A supertype is created by:
1146 * super_by_fd
1147 * guess_super
1148 * dup_super
1149 */
1150 struct supertype {
1151 struct superswitch *ss;
1152 int minor_version;
1153 int max_devs;
1154 char container_devnm[32]; /* devnm of container */
1155 void *sb;
1156 void *info;
1157 void *other; /* Hack used to convert v0.90 to v1.0 */
1158 unsigned long long devsize;
1159 unsigned long long data_offset; /* used by v1.x only */
1160 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1161 HW/firmware related incompatability to load metadata.
1162 Used when examining metadata to display content of disk
1163 when user has no hw/firmare compatible system.
1164 */
1165 struct metadata_update *updates;
1166 struct metadata_update **update_tail;
1167
1168 /* extra stuff used by mdmon */
1169 struct active_array *arrays;
1170 int sock; /* listen to external programs */
1171 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1172 * external:/md0/12
1173 */
1174 int devcnt;
1175 int retry_soon;
1176 int nodes;
1177 char *cluster_name;
1178
1179 struct mdinfo *devs;
1180
1181 };
1182
1183 extern struct supertype *super_by_fd(int fd, char **subarray);
1184 enum guess_types { guess_any, guess_array, guess_partitions };
1185 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1186 static inline struct supertype *guess_super(int fd) {
1187 return guess_super_type(fd, guess_any);
1188 }
1189 extern struct supertype *dup_super(struct supertype *st);
1190 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1191 extern int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep);
1192 extern int must_be_container(int fd);
1193 extern int dev_size_from_id(dev_t id, unsigned long long *size);
1194 extern int dev_sector_size_from_id(dev_t id, unsigned int *size);
1195 void wait_for(char *dev, int fd);
1196
1197 /*
1198 * Data structures for policy management.
1199 * Each device can have a policy structure that lists
1200 * various name/value pairs each possibly with a metadata associated.
1201 * The policy list is sorted by name/value/metadata
1202 */
1203 struct dev_policy {
1204 struct dev_policy *next;
1205 char *name; /* None of these strings are allocated. They are
1206 * all just references to strings which are known
1207 * to exist elsewhere.
1208 * name and metadata can be compared by address equality.
1209 */
1210 const char *metadata;
1211 const char *value;
1212 };
1213
1214 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1215
1216 /* iterate over the sublist starting at list, having the same
1217 * 'name' as 'list', and matching the given metadata (Where
1218 * NULL matches anything
1219 */
1220 #define pol_for_each(item, list, _metadata) \
1221 for (item = list; \
1222 item && item->name == list->name; \
1223 item = item->next) \
1224 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1225 ; else
1226
1227 /*
1228 * policy records read from mdadm are largely just name-value pairs.
1229 * The names are constants, not strdupped
1230 */
1231 struct pol_rule {
1232 struct pol_rule *next;
1233 char *type; /* rule_policy or rule_part */
1234 struct rule {
1235 struct rule *next;
1236 char *name;
1237 char *value;
1238 char *dups; /* duplicates of 'value' with a partNN appended */
1239 } *rule;
1240 };
1241
1242 extern char rule_policy[], rule_part[];
1243 extern char rule_path[], rule_type[];
1244 extern char type_part[], type_disk[];
1245
1246 extern void policyline(char *line, char *type);
1247 extern void policy_add(char *type, ...);
1248 extern void policy_free(void);
1249
1250 extern struct dev_policy *path_policy(char **paths, char *type);
1251 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1252 extern struct dev_policy *devid_policy(int devid);
1253 extern void dev_policy_free(struct dev_policy *p);
1254
1255 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1256 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1257 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1258
1259 enum policy_action {
1260 act_default,
1261 act_include,
1262 act_re_add,
1263 act_spare, /* This only applies to bare devices */
1264 act_spare_same_slot, /* this allows non-bare devices,
1265 * but only if recent removal */
1266 act_force_spare, /* this allow non-bare devices in any case */
1267 act_err
1268 };
1269
1270 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1271 enum policy_action want);
1272 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1273 enum policy_action want);
1274
1275 struct domainlist {
1276 struct domainlist *next;
1277 const char *dom;
1278 };
1279
1280 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1281 const char *metadata);
1282 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1283 const char *metadata);
1284 extern void domainlist_add_dev(struct domainlist **dom, int devid,
1285 const char *metadata);
1286 extern void domain_free(struct domainlist *dl);
1287 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1288 const char *metadata);
1289 void domain_add(struct domainlist **domp, char *domain);
1290
1291 extern void policy_save_path(char *id_path, struct map_ent *array);
1292 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1293
1294 #if __GNUC__ < 3
1295 struct stat64;
1296 #endif
1297
1298 #define HAVE_NFTW we assume
1299 #define HAVE_FTW
1300
1301 #ifdef __UCLIBC__
1302 # include <features.h>
1303 # ifndef __UCLIBC_HAS_LFS__
1304 # define lseek64 lseek
1305 # endif
1306 # ifndef __UCLIBC_HAS_FTW__
1307 # undef HAVE_FTW
1308 # undef HAVE_NFTW
1309 # endif
1310 #endif
1311
1312 #ifdef __dietlibc__
1313 # undef HAVE_NFTW
1314 #endif
1315
1316 #if defined(__KLIBC__)
1317 # undef HAVE_NFTW
1318 # undef HAVE_FTW
1319 #endif
1320
1321 #ifndef HAVE_NFTW
1322 # define FTW_PHYS 1
1323 # ifndef HAVE_FTW
1324 struct FTW {};
1325 # endif
1326 #endif
1327
1328 #ifdef HAVE_FTW
1329 # include <ftw.h>
1330 #endif
1331
1332 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1333
1334 extern int Manage_ro(char *devname, int fd, int readonly);
1335 extern int Manage_run(char *devname, int fd, struct context *c);
1336 extern int Manage_stop(char *devname, int fd, int quiet,
1337 int will_retry);
1338 extern int Manage_subdevs(char *devname, int fd,
1339 struct mddev_dev *devlist, int verbose, int test,
1340 char *update, int force);
1341 extern int autodetect(void);
1342 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1343 extern int Grow_addbitmap(char *devname, int fd,
1344 struct context *c, struct shape *s);
1345 extern int Grow_reshape(char *devname, int fd,
1346 struct mddev_dev *devlist,
1347 unsigned long long data_offset,
1348 struct context *c, struct shape *s);
1349 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1350 int *fdlist, int cnt, char *backup_file, int verbose);
1351 extern int Grow_continue(int mdfd, struct supertype *st,
1352 struct mdinfo *info, char *backup_file,
1353 int forked, int freeze_reshape);
1354 extern int Grow_consistency_policy(char *devname, int fd,
1355 struct context *c, struct shape *s);
1356
1357 extern int restore_backup(struct supertype *st,
1358 struct mdinfo *content,
1359 int working_disks,
1360 int spares,
1361 char **backup_filep,
1362 int verbose);
1363 extern int Grow_continue_command(char *devname, int fd,
1364 char *backup_file, int verbose);
1365
1366 extern int Assemble(struct supertype *st, char *mddev,
1367 struct mddev_ident *ident,
1368 struct mddev_dev *devlist,
1369 struct context *c);
1370
1371 extern int Build(char *mddev, struct mddev_dev *devlist,
1372 struct shape *s, struct context *c);
1373
1374 extern int Create(struct supertype *st, char *mddev,
1375 char *name, int *uuid,
1376 int subdevs, struct mddev_dev *devlist,
1377 struct shape *s,
1378 struct context *c,
1379 unsigned long long data_offset);
1380
1381 extern int Detail(char *dev, struct context *c);
1382 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1383 extern int Query(char *dev);
1384 extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1385 extern int Examine(struct mddev_dev *devlist, struct context *c,
1386 struct supertype *forcest);
1387 extern int Monitor(struct mddev_dev *devlist,
1388 char *mailaddr, char *alert_cmd,
1389 struct context *c,
1390 int daemonise, int oneshot,
1391 int dosyslog, char *pidfile, int increments,
1392 int share);
1393
1394 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1395 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1396 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1397 extern int Wait(char *dev);
1398 extern int WaitClean(char *dev, int verbose);
1399 extern int SetAction(char *dev, char *action);
1400
1401 extern int Incremental(struct mddev_dev *devlist, struct context *c,
1402 struct supertype *st);
1403 extern void RebuildMap(void);
1404 extern int IncrementalScan(struct context *c, char *devnm);
1405 extern int IncrementalRemove(char *devname, char *path, int verbose);
1406 extern int CreateBitmap(char *filename, int force, char uuid[16],
1407 unsigned long chunksize, unsigned long daemon_sleep,
1408 unsigned long write_behind,
1409 unsigned long long array_size,
1410 int major);
1411 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1412 extern int Write_rules(char *rule_name);
1413 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1414
1415 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1416 static inline unsigned long long
1417 bitmap_bits(unsigned long long array_size, unsigned long chunksize)
1418 {
1419 return (array_size * 512 + chunksize - 1) / chunksize;
1420 }
1421
1422 extern int Dump_metadata(char *dev, char *dir, struct context *c,
1423 struct supertype *st);
1424 extern int Restore_metadata(char *dev, char *dir, struct context *c,
1425 struct supertype *st, int only);
1426
1427 int md_array_valid(int fd);
1428 int md_array_active(int fd);
1429 int md_array_is_active(struct mdinfo *info);
1430 int md_get_array_info(int fd, struct mdu_array_info_s *array);
1431 int md_set_array_info(int fd, struct mdu_array_info_s *array);
1432 int md_get_disk_info(int fd, struct mdu_disk_info_s *disk);
1433 extern int get_linux_version(void);
1434 extern int mdadm_version(char *version);
1435 extern unsigned long long parse_size(char *size);
1436 extern int parse_uuid(char *str, int uuid[4]);
1437 extern int is_near_layout_10(int layout);
1438 extern int parse_layout_10(char *layout);
1439 extern int parse_layout_faulty(char *layout);
1440 extern long parse_num(char *num);
1441 extern int parse_cluster_confirm_arg(char *inp, char **devname, int *slot);
1442 extern int check_ext2(int fd, char *name);
1443 extern int check_reiser(int fd, char *name);
1444 extern int check_raid(int fd, char *name);
1445 extern int check_partitions(int fd, char *dname,
1446 unsigned long long freesize,
1447 unsigned long long size);
1448 extern int fstat_is_blkdev(int fd, char *devname, dev_t *rdev);
1449 extern int stat_is_blkdev(char *devname, dev_t *rdev);
1450
1451 extern int get_mdp_major(void);
1452 extern int get_maj_min(char *dev, int *major, int *minor);
1453 extern int dev_open(char *dev, int flags);
1454 extern int open_dev(char *devnm);
1455 extern void reopen_mddev(int mdfd);
1456 extern int open_dev_flags(char *devnm, int flags);
1457 extern int open_dev_excl(char *devnm);
1458 extern int is_standard(char *dev, int *nump);
1459 extern int same_dev(char *one, char *two);
1460 extern int compare_paths (char* path1,char* path2);
1461 extern void enable_fds(int devices);
1462
1463 extern int parse_auto(char *str, char *msg, int config);
1464 extern struct mddev_ident *conf_get_ident(char *dev);
1465 extern struct mddev_dev *conf_get_devs(void);
1466 extern int conf_test_dev(char *devname);
1467 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1468 extern struct createinfo *conf_get_create_info(void);
1469 extern void set_conffile(char *file);
1470 extern char *conf_get_mailaddr(void);
1471 extern char *conf_get_mailfrom(void);
1472 extern char *conf_get_program(void);
1473 extern char *conf_get_homehost(int *require_homehostp);
1474 extern char *conf_get_homecluster(void);
1475 extern char *conf_line(FILE *file);
1476 extern char *conf_word(FILE *file, int allow_key);
1477 extern void print_quoted(char *str);
1478 extern void print_escape(char *str);
1479 extern int use_udev(void);
1480 extern unsigned long GCD(unsigned long a, unsigned long b);
1481 extern int conf_name_is_free(char *name);
1482 extern int conf_verify_devnames(struct mddev_ident *array_list);
1483 extern int devname_matches(char *name, char *match);
1484 extern struct mddev_ident *conf_match(struct supertype *st,
1485 struct mdinfo *info,
1486 char *devname,
1487 int verbose, int *rvp);
1488
1489 extern void free_line(char *line);
1490 extern int match_oneof(char *devices, char *devname);
1491 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1492 extern const int uuid_zero[4];
1493 extern int same_uuid(int a[4], int b[4], int swapuuid);
1494 extern void copy_uuid(void *a, int b[4], int swapuuid);
1495 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1496 extern char *fname_from_uuid(struct supertype *st,
1497 struct mdinfo *info, char *buf, char sep);
1498 extern unsigned long calc_csum(void *super, int bytes);
1499 extern int enough(int level, int raid_disks, int layout, int clean,
1500 char *avail);
1501 extern int ask(char *mesg);
1502 extern unsigned long long get_component_size(int fd);
1503 extern void remove_partitions(int fd);
1504 extern int test_partition(int fd);
1505 extern int test_partition_from_id(dev_t id);
1506 extern int get_data_disks(int level, int layout, int raid_disks);
1507 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1508 int chunksize, unsigned long long devsize);
1509 extern int flush_metadata_updates(struct supertype *st);
1510 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1511 extern int assemble_container_content(struct supertype *st, int mdfd,
1512 struct mdinfo *content,
1513 struct context *c,
1514 char *chosen_name, int *result);
1515 #define INCR_NO 1
1516 #define INCR_UNSAFE 2
1517 #define INCR_ALREADY 4
1518 #define INCR_YES 8
1519 extern struct mdinfo *container_choose_spares(struct supertype *st,
1520 struct spare_criteria *criteria,
1521 struct domainlist *domlist,
1522 char *spare_group,
1523 const char *metadata, int get_one);
1524 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1525 extern int add_disk(int mdfd, struct supertype *st,
1526 struct mdinfo *sra, struct mdinfo *info);
1527 extern int remove_disk(int mdfd, struct supertype *st,
1528 struct mdinfo *sra, struct mdinfo *info);
1529 extern int hot_remove_disk(int mdfd, unsigned long dev, int force);
1530 extern int sys_hot_remove_disk(int statefd, int force);
1531 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1532 unsigned long long min_recovery_start(struct mdinfo *array);
1533
1534 extern char *human_size(long long bytes);
1535 extern char *human_size_brief(long long bytes, int prefix);
1536 extern void print_r10_layout(int layout);
1537
1538 extern char *find_free_devnm(int use_partitions);
1539
1540 extern void put_md_name(char *name);
1541 extern char *devid2kname(dev_t devid);
1542 extern char *devid2devnm(dev_t devid);
1543 extern dev_t devnm2devid(char *devnm);
1544 extern char *get_md_name(char *devnm);
1545
1546 extern char DefaultConfFile[];
1547
1548 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1549 char *chosen, int block_udev);
1550 /* values for 'trustworthy' */
1551 #define LOCAL 1
1552 #define LOCAL_ANY 10
1553 #define FOREIGN 2
1554 #define METADATA 3
1555 extern int open_mddev(char *dev, int report_errors);
1556 extern int open_container(int fd);
1557 extern int metadata_container_matches(char *metadata, char *devnm);
1558 extern int metadata_subdev_matches(char *metadata, char *devnm);
1559 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1560 extern int is_subarray_active(char *subarray, char *devname);
1561 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1562 extern struct superswitch *version_to_superswitch(char *vers);
1563
1564 extern int mdmon_running(char *devnm);
1565 extern int mdmon_pid(char *devnm);
1566 extern int check_env(char *name);
1567 extern __u32 random32(void);
1568 extern void random_uuid(__u8 *buf);
1569 extern int start_mdmon(char *devnm);
1570
1571 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1572 struct supertype *st, unsigned long stripes,
1573 int *fds, unsigned long long *offsets,
1574 int dests, int *destfd, unsigned long long *destoffsets);
1575 void abort_reshape(struct mdinfo *sra);
1576
1577 void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1578
1579 extern char *stat2kname(struct stat *st);
1580 extern char *fd2kname(int fd);
1581 extern char *stat2devnm(struct stat *st);
1582 extern char *fd2devnm(int fd);
1583 extern void udev_block(char *devnm);
1584 extern void udev_unblock(void);
1585
1586 extern int in_initrd(void);
1587
1588 struct cmap_hooks {
1589 void *cmap_handle; /* corosync lib related */
1590
1591 int (*initialize)(cmap_handle_t *handle);
1592 int (*get_string)(cmap_handle_t handle,
1593 const char *string,
1594 char **name);
1595 int (*finalize)(cmap_handle_t handle);
1596 };
1597
1598 extern void set_cmap_hooks(void);
1599 extern void set_hooks(void);
1600
1601 struct dlm_hooks {
1602 void *dlm_handle; /* dlm lib related */
1603
1604 dlm_lshandle_t (*create_lockspace)(const char *name,
1605 unsigned int mode);
1606 dlm_lshandle_t (*open_lockspace)(const char *name);
1607 int (*release_lockspace)(const char *name, dlm_lshandle_t ls,
1608 int force);
1609 int (*ls_lock)(dlm_lshandle_t lockspace, uint32_t mode,
1610 struct dlm_lksb *lksb, uint32_t flags,
1611 const void *name, unsigned int namelen,
1612 uint32_t parent, void (*astaddr) (void *astarg),
1613 void *astarg, void (*bastaddr) (void *astarg),
1614 void *range);
1615 int (*ls_unlock_wait)(dlm_lshandle_t lockspace, uint32_t lkid,
1616 uint32_t flags, struct dlm_lksb *lksb);
1617 int (*ls_get_fd)(dlm_lshandle_t ls);
1618 int (*dispatch)(int fd);
1619 };
1620
1621 extern int get_cluster_name(char **name);
1622 extern int dlm_funs_ready(void);
1623 extern int cluster_get_dlmlock(void);
1624 extern int cluster_release_dlmlock(void);
1625 extern void set_dlm_hooks(void);
1626
1627 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1628 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1629 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1630 (ROUND_UP((unsigned long)(ptr), base)))
1631
1632 static inline int is_subarray(char *vers)
1633 {
1634 /* The version string for a 'subarray' (an array in a container)
1635 * is
1636 * /containername/componentname for normal read-write arrays
1637 * -containername/componentname for arrays which mdmon must not
1638 * reconfigure. They might be read-only
1639 * or might be undergoing reshape etc.
1640 * containername is e.g. md0, md_d1
1641 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1642 */
1643 return (*vers == '/' || *vers == '-');
1644 }
1645
1646 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1647 {
1648 return &ent->metadata_version[10+strlen(container)+1];
1649 }
1650
1651 #ifdef DEBUG
1652 #define dprintf(fmt, arg...) \
1653 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1654 #define dprintf_cont(fmt, arg...) \
1655 fprintf(stderr, fmt, ##arg)
1656 #else
1657 #define dprintf(fmt, arg...) \
1658 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1659 #define dprintf_cont(fmt, arg...) \
1660 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1661 #endif
1662 #include <assert.h>
1663 #include <stdarg.h>
1664 static inline int xasprintf(char **strp, const char *fmt, ...) {
1665 va_list ap;
1666 int ret;
1667 va_start(ap, fmt);
1668 ret = vasprintf(strp, fmt, ap);
1669 va_end(ap);
1670 assert(ret >= 0);
1671 return ret;
1672 }
1673
1674 #ifdef DEBUG
1675 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1676 #else
1677 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1678 #endif
1679 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1680
1681 void *xmalloc(size_t len);
1682 void *xrealloc(void *ptr, size_t len);
1683 void *xcalloc(size_t num, size_t size);
1684 char *xstrdup(const char *str);
1685
1686 #define LEVEL_MULTIPATH (-4)
1687 #define LEVEL_LINEAR (-1)
1688 #define LEVEL_FAULTY (-5)
1689
1690 /* kernel module doesn't know about these */
1691 #define LEVEL_CONTAINER (-100)
1692 #define LEVEL_UNSUPPORTED (-200)
1693
1694 /* the kernel does know about this one ... */
1695 #define LEVEL_NONE (-1000000)
1696
1697 /* faulty stuff */
1698
1699 #define WriteTransient 0
1700 #define ReadTransient 1
1701 #define WritePersistent 2
1702 #define ReadPersistent 3
1703 #define WriteAll 4 /* doesn't go to device */
1704 #define ReadFixable 5
1705 #define Modes 6
1706
1707 #define ClearErrors 31
1708 #define ClearFaults 30
1709
1710 #define AllPersist 100 /* internal use only */
1711 #define NoPersist 101
1712
1713 #define ModeMask 0x1f
1714 #define ModeShift 5
1715
1716 #ifdef __TINYC__
1717 #undef minor
1718 #undef major
1719 #undef makedev
1720 #define minor(x) ((x)&0xff)
1721 #define major(x) (((x)>>8)&0xff)
1722 #define makedev(M,m) (((M)<<8) | (m))
1723 #endif
1724
1725 /* for raid4/5/6 */
1726 #define ALGORITHM_LEFT_ASYMMETRIC 0
1727 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1728 #define ALGORITHM_LEFT_SYMMETRIC 2
1729 #define ALGORITHM_RIGHT_SYMMETRIC 3
1730
1731 /* Define non-rotating (raid4) algorithms. These allow
1732 * conversion of raid4 to raid5.
1733 */
1734 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1735 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1736
1737 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1738 * Firstly, the exact positioning of the parity block is slightly
1739 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1740 * of DDF.
1741 * Secondly, or order of datablocks over which the Q syndrome is computed
1742 * is different.
1743 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1744 * These layouts are from the DDFv1.2 spec.
1745 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1746 * leaves RLQ=3 as 'Vendor Specific'
1747 */
1748
1749 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1750 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1751 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1752
1753 /* For every RAID5 algorithm we define a RAID6 algorithm
1754 * with exactly the same layout for data and parity, and
1755 * with the Q block always on the last device (N-1).
1756 * This allows trivial conversion from RAID5 to RAID6
1757 */
1758 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1759 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1760 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1761 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1762 #define ALGORITHM_PARITY_0_6 20
1763 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1764
1765 /* Define PATH_MAX in case we don't use glibc or standard library does
1766 * not have PATH_MAX defined. Assume max path length is 4K characters.
1767 */
1768 #ifndef PATH_MAX
1769 #define PATH_MAX 4096
1770 #endif
1771
1772 #define RESYNC_NONE -1
1773 #define RESYNC_DELAYED -2
1774 #define RESYNC_PENDING -3
1775 #define RESYNC_UNKNOWN -4
1776
1777 /* When using "GET_DISK_INFO" it isn't certain how high
1778 * we need to check. So we impose an absolute limit of
1779 * MAX_DISKS. This needs to be much more than the largest
1780 * number of devices any metadata can support. Currently
1781 * v1.x can support 1920
1782 */
1783 #define MAX_DISKS 4096
1784
1785 /* Sometimes the 'size' value passed needs to mean "Maximum".
1786 * In those cases with use MAX_SIZE
1787 */
1788 #define MAX_SIZE 1
1789
1790 /* We want to use unsigned numbers for sector counts, but need
1791 * a value for 'invalid'. Use '1'.
1792 */
1793 #define INVALID_SECTORS 1
1794 /* And another special number needed for --data_offset=variable */
1795 #define VARIABLE_OFFSET 3