]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
IMSM: Add warning message when assemble spanned container
[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 <stdlib.h>
39 #include <time.h>
40 #include <sys/time.h>
41 #include <getopt.h>
42 #include <fcntl.h>
43 #include <stdio.h>
44 #include <errno.h>
45 #include <string.h>
46 #include <syslog.h>
47 #ifdef __dietlibc__
48 #include <strings.h>
49 /* dietlibc has deprecated random and srandom!! */
50 #define random rand
51 #define srandom srand
52 #endif
53
54 #include <linux/kdev_t.h>
55 /*#include <linux/fs.h> */
56 #include <sys/mount.h>
57 #include <asm/types.h>
58 #include <sys/ioctl.h>
59 #define MD_MAJOR 9
60 #define MdpMinorShift 6
61
62 #ifndef BLKGETSIZE64
63 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
64 #endif
65
66 #define DEFAULT_CHUNK 512
67 #define DEFAULT_BITMAP_CHUNK 4096
68 #define DEFAULT_BITMAP_DELAY 5
69 #define DEFAULT_MAX_WRITE_BEHIND 256
70
71 /* MAP_DIR should be somewhere that persists across the pivotroot
72 * from early boot to late boot.
73 * /run seems to have emerged as the best standard.
74 */
75 #ifndef MAP_DIR
76 #define MAP_DIR "/run/mdadm"
77 #endif /* MAP_DIR */
78 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
79 * want something other than the default of "map"
80 */
81 #ifndef MAP_FILE
82 #define MAP_FILE "map"
83 #endif /* MAP_FILE */
84 /* MDMON_DIR is where pid and socket files used for communicating
85 * with mdmon normally live. Best is /var/run/mdadm as
86 * mdmon is needed at early boot then it needs to write there prior
87 * to /var/run being mounted read/write, and it also then needs to
88 * persist beyond when /var/run is mounter read-only. So, to be
89 * safe, the default is somewhere that is read/write early in the
90 * boot process and stays up as long as possible during shutdown.
91 */
92 #ifndef MDMON_DIR
93 #define MDMON_DIR "/run/mdadm"
94 #endif /* MDMON_DIR */
95
96 /* FAILED_SLOTS is where to save files storing recent removal of array
97 * member in order to allow future reuse of disk inserted in the same
98 * slot for array recovery
99 */
100 #ifndef FAILED_SLOTS_DIR
101 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
102 #endif /* FAILED_SLOTS */
103
104 #include "md_u.h"
105 #include "md_p.h"
106 #include "bitmap.h"
107 #include "msg.h"
108
109 #include <endian.h>
110 /* Redhat don't like to #include <asm/byteorder.h>, and
111 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
112 * and there is no standard conversion function so... */
113 /* And dietlibc doesn't think byteswap is ok, so.. */
114 /* #include <byteswap.h> */
115 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
116 ((x) & 0xff00U) >> 8)
117 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
118 ((x) & 0xff000000U) >> 24 | \
119 ((x) & 0x0000ff00U) << 8 | \
120 ((x) & 0x00ff0000U) >> 8)
121 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
122 ((x) & 0xff00000000000000ULL) >> 56 | \
123 ((x) & 0x000000000000ff00ULL) << 40 | \
124 ((x) & 0x00ff000000000000ULL) >> 40 | \
125 ((x) & 0x0000000000ff0000ULL) << 24 | \
126 ((x) & 0x0000ff0000000000ULL) >> 24 | \
127 ((x) & 0x00000000ff000000ULL) << 8 | \
128 ((x) & 0x000000ff00000000ULL) >> 8)
129
130 #if !defined(__KLIBC__)
131 #if BYTE_ORDER == LITTLE_ENDIAN
132 #define __cpu_to_le16(_x) (unsigned int)(_x)
133 #define __cpu_to_le32(_x) (unsigned int)(_x)
134 #define __cpu_to_le64(_x) (unsigned long long)(_x)
135 #define __le16_to_cpu(_x) (unsigned int)(_x)
136 #define __le32_to_cpu(_x) (unsigned int)(_x)
137 #define __le64_to_cpu(_x) (unsigned long long)(_x)
138
139 #define __cpu_to_be16(_x) bswap_16(_x)
140 #define __cpu_to_be32(_x) bswap_32(_x)
141 #define __cpu_to_be64(_x) bswap_64(_x)
142 #define __be16_to_cpu(_x) bswap_16(_x)
143 #define __be32_to_cpu(_x) bswap_32(_x)
144 #define __be64_to_cpu(_x) bswap_64(_x)
145 #elif BYTE_ORDER == BIG_ENDIAN
146 #define __cpu_to_le16(_x) bswap_16(_x)
147 #define __cpu_to_le32(_x) bswap_32(_x)
148 #define __cpu_to_le64(_x) bswap_64(_x)
149 #define __le16_to_cpu(_x) bswap_16(_x)
150 #define __le32_to_cpu(_x) bswap_32(_x)
151 #define __le64_to_cpu(_x) bswap_64(_x)
152
153 #define __cpu_to_be16(_x) (unsigned int)(_x)
154 #define __cpu_to_be32(_x) (unsigned int)(_x)
155 #define __cpu_to_be64(_x) (unsigned long long)(_x)
156 #define __be16_to_cpu(_x) (unsigned int)(_x)
157 #define __be32_to_cpu(_x) (unsigned int)(_x)
158 #define __be64_to_cpu(_x) (unsigned long long)(_x)
159 #else
160 # error "unknown endianness."
161 #endif
162 #endif /* __KLIBC__ */
163
164 /*
165 * min()/max()/clamp() macros that also do
166 * strict type-checking.. See the
167 * "unnecessary" pointer comparison.
168 */
169 #define min(x, y) ({ \
170 typeof(x) _min1 = (x); \
171 typeof(y) _min2 = (y); \
172 (void) (&_min1 == &_min2); \
173 _min1 < _min2 ? _min1 : _min2; })
174
175 #define max(x, y) ({ \
176 typeof(x) _max1 = (x); \
177 typeof(y) _max2 = (y); \
178 (void) (&_max1 == &_max2); \
179 _max1 > _max2 ? _max1 : _max2; })
180
181 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
182
183 /* general information that might be extracted from a superblock */
184 struct mdinfo {
185 mdu_array_info_t array;
186 mdu_disk_info_t disk;
187 __u64 events;
188 int uuid[4];
189 char name[33];
190 unsigned long long data_offset;
191 unsigned long long new_data_offset;
192 unsigned long long component_size; /* same as array.size, except in
193 * sectors and up to 64bits.
194 */
195 unsigned long long custom_array_size; /* size for non-default sized
196 * arrays (in sectors)
197 */
198 #define NO_RESHAPE 0
199 #define VOLUME_RESHAPE 1
200 #define CONTAINER_RESHAPE 2
201 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
202 int reshape_active;
203 unsigned long long reshape_progress;
204 int recovery_blocked; /* for external metadata it
205 * indicates that there is
206 * reshape in progress in
207 * container,
208 * for native metadata it is
209 * reshape_active field mirror
210 */
211 /* During reshape we can sometimes change the data_offset to avoid
212 * over-writing still-valid data. We need to know if there is space.
213 * So getinfo_super will fill in space_before and space_after in sectors.
214 * data_offset can be increased or decreased by this amount.
215 */
216 unsigned long long space_before, space_after;
217 union {
218 unsigned long long resync_start; /* per-array resync position */
219 unsigned long long recovery_start; /* per-device rebuild position */
220 #define MaxSector (~0ULL) /* resync/recovery complete position */
221 };
222 long bitmap_offset; /* 0 == none, 1 == a file */
223 unsigned long safe_mode_delay; /* ms delay to mark clean */
224 int new_level, delta_disks, new_layout, new_chunk;
225 int errors;
226 unsigned long cache_size; /* size of raid456 stripe cache*/
227 int mismatch_cnt;
228 char text_version[50];
229
230 int container_member; /* for assembling external-metatdata arrays
231 * This is to be used internally by metadata
232 * handler only */
233 int container_enough; /* flag external handlers can set to
234 * indicate that subarrays have not enough (-1),
235 * enough to start (0), or all expected disks (1) */
236 char sys_name[20];
237 struct mdinfo *devs;
238 struct mdinfo *next;
239
240 /* Device info for mdmon: */
241 int recovery_fd;
242 int state_fd;
243 #define DS_FAULTY 1
244 #define DS_INSYNC 2
245 #define DS_WRITE_MOSTLY 4
246 #define DS_SPARE 8
247 #define DS_BLOCKED 16
248 #define DS_REMOVE 1024
249 #define DS_UNBLOCK 2048
250 int prev_state, curr_state, next_state;
251
252 };
253
254 struct createinfo {
255 int uid;
256 int gid;
257 int autof;
258 int mode;
259 int symlinks;
260 int names;
261 struct supertype *supertype;
262 };
263
264 #define Name "mdadm"
265
266 enum mode {
267 ASSEMBLE=1,
268 BUILD,
269 CREATE,
270 MANAGE,
271 MISC,
272 MONITOR,
273 GROW,
274 INCREMENTAL,
275 AUTODETECT,
276 mode_count
277 };
278
279 extern char short_options[];
280 extern char short_bitmap_options[];
281 extern char short_bitmap_auto_options[];
282 extern struct option long_options[];
283 extern char Version[], Usage[], Help[], OptionHelp[],
284 *mode_help[],
285 Help_create[], Help_build[], Help_assemble[], Help_grow[],
286 Help_incr[],
287 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
288
289 /* for option that don't have short equivilents, we assign arbitrary
290 * numbers later than any 'short' character option.
291 */
292 enum special_options {
293 AssumeClean = 300,
294 BitmapChunk,
295 WriteBehind,
296 ReAdd,
297 NoDegraded,
298 Sparc22,
299 BackupFile,
300 HomeHost,
301 AutoHomeHost,
302 Symlinks,
303 AutoDetect,
304 Waitclean,
305 DetailPlatform,
306 KillSubarray,
307 UpdateSubarray,
308 IncrementalPath,
309 NoSharing,
310 HelpOptions,
311 Brief,
312 ManageOpt,
313 Add,
314 AddSpare,
315 Remove,
316 Fail,
317 Replace,
318 With,
319 MiscOpt,
320 WaitOpt,
321 ConfigFile,
322 ChunkSize,
323 WriteMostly,
324 Layout,
325 Auto,
326 Force,
327 SuperMinor,
328 EMail,
329 ProgramOpt,
330 Increment,
331 Fork,
332 Bitmap,
333 RebuildMapOpt,
334 InvalidBackup,
335 UdevRules,
336 FreezeReshape,
337 Continue,
338 OffRootOpt,
339 Prefer,
340 KillOpt,
341 DataOffset,
342 ExamineBB,
343 Dump,
344 Restore,
345 Action,
346 };
347
348 enum prefix_standard {
349 JEDEC,
350 IEC
351 };
352
353 /* structures read from config file */
354 /* List of mddevice names and identifiers
355 * Identifiers can be:
356 * uuid=128-hex-uuid
357 * super-minor=decimal-minor-number-from-superblock
358 * devices=comma,separated,list,of,device,names,with,wildcards
359 *
360 * If multiple fields are present, the intersection of all matching
361 * devices is considered
362 */
363 #define UnSet (0xfffe)
364 struct mddev_ident {
365 char *devname;
366
367 int uuid_set;
368 int uuid[4];
369 char name[33];
370
371 int super_minor;
372
373 char *devices; /* comma separated list of device
374 * names with wild cards
375 */
376 int level;
377 int raid_disks;
378 int spare_disks;
379 struct supertype *st;
380 int autof; /* 1 for normal, 2 for partitioned */
381 char *spare_group;
382 char *bitmap_file;
383 int bitmap_fd;
384
385 char *container; /* /dev/whatever name of container, or
386 * uuid of container. You would expect
387 * this to be the 'devname' or UUID
388 * of some other entry.
389 */
390 char *member; /* subarray within a container */
391
392 struct mddev_ident *next;
393 union {
394 /* fields needed by different users of this structure */
395 int assembled; /* set when assembly succeeds */
396 };
397 };
398
399 struct context {
400 int readonly;
401 int runstop;
402 int verbose;
403 int brief;
404 int force;
405 char *homehost;
406 int require_homehost;
407 char *prefer;
408 int export;
409 int test;
410 char *subarray;
411 char *update;
412 int scan;
413 int SparcAdjust;
414 int autof;
415 int delay;
416 int freeze_reshape;
417 char *backup_file;
418 int invalid_backup;
419 char *action;
420 };
421
422 struct shape {
423 int raiddisks;
424 int sparedisks;
425 int level;
426 int layout;
427 char *layout_str;
428 int chunk;
429 int bitmap_chunk;
430 char *bitmap_file;
431 int assume_clean;
432 int write_behind;
433 unsigned long long size;
434 };
435
436 /* List of device names - wildcards expanded */
437 struct mddev_dev {
438 char *devname;
439 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
440 * 'A' for re_add.
441 * Not set for names read from .config
442 */
443 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
444 int used; /* set when used */
445 long long data_offset;
446 struct mddev_dev *next;
447 };
448
449 typedef struct mapping {
450 char *name;
451 int num;
452 } mapping_t;
453
454 struct mdstat_ent {
455 char *dev;
456 char devnm[32];
457 int active;
458 char *level;
459 char *pattern; /* U or up, _ for down */
460 int percent; /* -1 if no resync */
461 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
462 int devcnt;
463 int raid_disks;
464 char * metadata_version;
465 struct dev_member {
466 char *name;
467 struct dev_member *next;
468 } *members;
469 struct mdstat_ent *next;
470 };
471
472 extern struct mdstat_ent *mdstat_read(int hold, int start);
473 extern void mdstat_close(void);
474 extern void free_mdstat(struct mdstat_ent *ms);
475 extern void mdstat_wait(int seconds);
476 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
477 extern int mddev_busy(char *devnm);
478 extern struct mdstat_ent *mdstat_by_component(char *name);
479 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
480
481 struct map_ent {
482 struct map_ent *next;
483 char devnm[32];
484 char metadata[20];
485 int uuid[4];
486 int bad;
487 char *path;
488 };
489 extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
490 int uuid[4], char *path);
491 extern void map_remove(struct map_ent **map, char *devnm);
492 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
493 #ifdef MDASSEMBLE
494 static inline struct map_ent *map_by_devnm(struct map_ent **map, char *name)
495 {
496 return NULL;
497 }
498 static inline void map_free(struct map_ent *map)
499 {
500 }
501 #else
502 extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
503 extern void map_free(struct map_ent *map);
504 #endif
505 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
506 extern void map_read(struct map_ent **melp);
507 extern int map_write(struct map_ent *mel);
508 extern void map_delete(struct map_ent **mapp, char *devnm);
509 extern void map_add(struct map_ent **melp,
510 char *devnm, char *metadata, int uuid[4], char *path);
511 extern int map_lock(struct map_ent **melp);
512 extern void map_unlock(struct map_ent **melp);
513 extern void map_fork(void);
514
515 /* various details can be requested */
516 enum sysfs_read_flags {
517 GET_LEVEL = (1 << 0),
518 GET_LAYOUT = (1 << 1),
519 GET_COMPONENT = (1 << 2),
520 GET_CHUNK = (1 << 3),
521 GET_CACHE = (1 << 4),
522 GET_MISMATCH = (1 << 5),
523 GET_VERSION = (1 << 6),
524 GET_DISKS = (1 << 7),
525 GET_DEGRADED = (1 << 8),
526 GET_SAFEMODE = (1 << 9),
527 GET_BITMAP_LOCATION = (1 << 10),
528
529 GET_DEVS = (1 << 20), /* gets role, major, minor */
530 GET_OFFSET = (1 << 21),
531 GET_SIZE = (1 << 22),
532 GET_STATE = (1 << 23),
533 GET_ERROR = (1 << 24),
534 };
535
536 /* If fd >= 0, get the array it is open on,
537 * else use devnm.
538 */
539 extern int sysfs_open(char *devnm, char *devname, char *attr);
540 extern void sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
541 extern void sysfs_free(struct mdinfo *sra);
542 extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
543 extern int sysfs_attr_match(const char *attr, const char *str);
544 extern int sysfs_match_word(const char *word, char **list);
545 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
546 char *name, char *val);
547 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
548 char *name, unsigned long long val);
549 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
550 char *name, long long val);
551 extern int sysfs_uevent(struct mdinfo *sra, char *event);
552 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
553 char *name);
554 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
555 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
556 char *name, unsigned long long *val);
557 extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
558 extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
559 char *name, unsigned long long *v1, unsigned long long *v2);
560 extern int sysfs_fd_get_str(int fd, char *val, int size);
561 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
562 char *name);
563 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
564 char *name, char *val, int size);
565 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
566 extern int sysfs_set_array(struct mdinfo *info, int vers);
567 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
568 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
569 extern int sysfs_unique_holder(char *devnm, long rdev);
570 extern int sysfs_freeze_array(struct mdinfo *sra);
571 extern int sysfs_wait(int fd, int *msec);
572 extern int load_sys(char *path, char *buf);
573 extern int reshape_prepare_fdlist(char *devname,
574 struct mdinfo *sra,
575 int raid_disks,
576 int nrdisks,
577 unsigned long blocks,
578 char *backup_file,
579 int *fdlist,
580 unsigned long long *offsets);
581 extern void reshape_free_fdlist(int *fdlist,
582 unsigned long long *offsets,
583 int size);
584 extern int reshape_open_backup_file(char *backup,
585 int fd,
586 char *devname,
587 long blocks,
588 int *fdlist,
589 unsigned long long *offsets,
590 char *sysfs_name,
591 int restart);
592 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
593 unsigned int ndata, unsigned int odata);
594 extern char *locate_backup(char *name);
595 extern char *make_backup(char *name);
596
597 extern int save_stripes(int *source, unsigned long long *offsets,
598 int raid_disks, int chunk_size, int level, int layout,
599 int nwrites, int *dest,
600 unsigned long long start, unsigned long long length,
601 char *buf);
602 extern int restore_stripes(int *dest, unsigned long long *offsets,
603 int raid_disks, int chunk_size, int level, int layout,
604 int source, unsigned long long read_offset,
605 unsigned long long start, unsigned long long length,
606 char *src_buf);
607
608 #ifndef Sendmail
609 #define Sendmail "/usr/lib/sendmail -t"
610 #endif
611
612 #define SYSLOG_FACILITY LOG_DAEMON
613
614 extern char *map_num(mapping_t *map, int num);
615 extern int map_name(mapping_t *map, char *name);
616 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
617
618 extern char *map_dev_preferred(int major, int minor, int create,
619 char *prefer);
620 static inline char *map_dev(int major, int minor, int create)
621 {
622 return map_dev_preferred(major, minor, create, NULL);
623 }
624
625 struct active_array;
626 struct metadata_update;
627
628 /* 'struct reshape' records the intermediate states of
629 * a general reshape.
630 * The starting geometry is converted to the 'before' geometry
631 * by at most an atomic level change. They could be the same.
632 * Similarly the 'after' geometry is converted to the final
633 * geometry by at most a level change.
634 * Note that 'before' and 'after' must have the same level.
635 * 'blocks' is the minimum number of sectors for a reshape unit.
636 * This will be a multiple of the stripe size in each of the
637 * 'before' and 'after' geometries.
638 * If 'blocks' is 0, no restriping is necessary.
639 * 'min_offset_change' is the minimum change to data_offset to
640 * allow the reshape to happen. It is at least the larger of
641 * the old and new chunk sizes, and typically the same as 'blocks'
642 * divided by number of data disks.
643 */
644 struct reshape {
645 int level;
646 int parity; /* number of parity blocks/devices */
647 struct {
648 int layout;
649 int data_disks;
650 } before, after;
651 unsigned long long backup_blocks;
652 unsigned long long min_offset_change;
653 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
654 unsigned long long new_size; /* New size of array in sectors */
655 };
656
657 /* A superswitch provides entry point the a metadata handler.
658 *
659 * The superswitch primarily operates on some "metadata" that
660 * is accessed via the 'supertype'.
661 * This metadata has one of three possible sources.
662 * 1/ It is read from a single device. In this case it may not completely
663 * describe the array or arrays as some information might be on other
664 * devices.
665 * 2/ It is read from all devices in a container. In this case all
666 * information is present.
667 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
668 * be complete once enough ->add_to_super calls have completed.
669 *
670 * When creating an array inside a container, the metadata will be
671 * formed by a combination of 2 and 3. The metadata or the array is read,
672 * then new information is added.
673 *
674 * The metadata must sometimes have a concept of a 'current' array
675 * and a 'current' device.
676 * The 'current' array is set by init_super to be the newly created array,
677 * or is set by super_by_fd when it finds it is looking at an array inside
678 * a container.
679 *
680 * The 'current' device is either the device that the metadata was read from
681 * in case 1, or the last device added by add_to_super in case 3.
682 * Case 2 does not identify a 'current' device.
683 */
684 extern struct superswitch {
685
686 /* Used to report details of metadata read from a component
687 * device. ->load_super has been called.
688 */
689 void (*examine_super)(struct supertype *st, char *homehost);
690 void (*brief_examine_super)(struct supertype *st, int verbose);
691 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
692 void (*export_examine_super)(struct supertype *st);
693 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
694 int (*copy_metadata)(struct supertype *st, int from, int to);
695
696 /* Used to report details of an active array.
697 * ->load_super was possibly given a 'component' string.
698 */
699 void (*detail_super)(struct supertype *st, char *homehost);
700 void (*brief_detail_super)(struct supertype *st);
701 void (*export_detail_super)(struct supertype *st);
702
703 /* Optional: platform hardware / firmware details */
704 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
705 int (*export_detail_platform)(int verbose, char *controller_path);
706
707 /* Used:
708 * to get uuid to storing in bitmap metadata
709 * and 'reshape' backup-data metadata
710 * To see if a device is being re-added to an array it was part of.
711 */
712 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
713
714 /* Extract generic details from metadata. This could be details about
715 * the container, or about an individual array within the container.
716 * The determination is made either by:
717 * load_super being given a 'component' string.
718 * validate_geometry determining what to create.
719 * The info includes both array information and device information.
720 * The particular device should be:
721 * The last device added by add_to_super
722 * The device the metadata was loaded from by load_super
723 * If 'map' is present, then it is an array raid_disks long
724 * (raid_disk must already be set and correct) and it is filled
725 * with 1 for slots that are thought to be active and 0 for slots which
726 * appear to be failed/missing.
727 * *info is zeroed out before data is added.
728 */
729 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
730 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
731 /* Check if the given metadata is flagged as belonging to "this"
732 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
733 */
734 int (*match_home)(struct supertype *st, char *homehost);
735
736 /* Make one of several generic modifications to metadata
737 * prior to assembly (or other times).
738 * sparc2.2 - first bug in early 0.90 metadata
739 * super-minor - change name of 0.90 metadata
740 * summaries - 'correct' any redundant data
741 * resync - mark array as dirty to trigger a resync.
742 * uuid - set new uuid - only 0.90 or 1.x
743 * name - change the name of the array (where supported)
744 * homehost - change which host this array is tied to.
745 * devicesize - If metadata is at start of device, change recorded
746 * device size to match actual device size
747 * byteorder - swap bytes for 0.90 metadata
748 *
749 * force-one - mark that device as uptodate, not old or failed.
750 * force-array - mark array as clean if it would not otherwise
751 * assemble
752 * assemble - not sure how this is different from force-one...
753 * linear-grow-new - add a new device to a linear array, but don't
754 * change the size: so superblock still matches
755 * linear-grow-update - now change the size of the array.
756 * writemostly - set the WriteMostly1 bit in the superblock devflags
757 * readwrite - clear the WriteMostly1 bit in the superblock devflags
758 * no-bitmap - clear any record that a bitmap is present.
759 * bbl - add a bad-block-log if possible
760 * no-bbl - remove and bad-block-log is it is empty.
761 * revert-reshape - If a reshape is in progress, modify metadata so
762 * it will resume going in the opposite direction.
763 */
764 int (*update_super)(struct supertype *st, struct mdinfo *info,
765 char *update,
766 char *devname, int verbose,
767 int uuid_set, char *homehost);
768
769 /* Create new metadata for new array as described. This could
770 * be a new container, or an array in a pre-existing container.
771 * Also used to zero metadata prior to writing it to invalidate old
772 * metadata.
773 */
774 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
775 unsigned long long size, char *name,
776 char *homehost, int *uuid,
777 unsigned long long data_offset);
778
779 /* update the metadata to include new device, either at create or
780 * when hot-adding a spare.
781 */
782 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
783 int fd, char *devname,
784 unsigned long long data_offset);
785 /* update the metadata to delete a device,
786 * when hot-removing.
787 */
788 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
789
790 /* Write metadata to one device when fixing problems or adding
791 * a new device.
792 */
793 int (*store_super)(struct supertype *st, int fd);
794
795 /* Write all metadata for this array.
796 */
797 int (*write_init_super)(struct supertype *st);
798 /* Check if metadata read from one device is compatible with an array,
799 * used when assembling an array, or pseudo-assembling was with
800 * "--examine --brief"
801 * If "st" has not yet been loaded the superblock from, "tst" is
802 * moved in, otherwise the superblock in 'st' is compared with
803 * 'tst'.
804 */
805 int (*compare_super)(struct supertype *st, struct supertype *tst);
806 /* Load metadata from a single device. If 'devname' is not NULL
807 * print error messages as appropriate */
808 int (*load_super)(struct supertype *st, int fd, char *devname);
809 /* 'fd' is a 'container' md array - load array metadata from the
810 * whole container.
811 */
812 int (*load_container)(struct supertype *st, int fd, char *devname);
813 /* If 'arg' is a valid name of this metadata type, allocate and
814 * return a 'supertype' for the particular minor version */
815 struct supertype * (*match_metadata_desc)(char *arg);
816 /* If a device has the given size, and the data_offset has been
817 * requested - work out how much space is available for data.
818 * This involves adjusting for reserved space (e.g. bitmaps)
819 * and for any rounding.
820 * 'mdadm' only calls this for existing arrays where a possible
821 * spare is being added. However some super-handlers call it
822 * internally from validate_geometry when creating an array.
823 */
824 __u64 (*avail_size)(struct supertype *st, __u64 size,
825 unsigned long long data_offset);
826 /* This is similar to 'avail_size' in purpose, but is used for
827 * containers for which there is no 'component size' to compare.
828 * This reports that whole-device size which is a minimum
829 */
830 unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
831 /* Find somewhere to put a bitmap - possibly auto-size it - and
832 * update the metadata to record this. The array may be newly
833 * created, in which case data_size may be updated, or it might
834 * already exist. Metadata handler can know if init_super
835 * has been called, but not write_init_super.
836 */
837 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
838 int delay, int write_behind,
839 unsigned long long size, int may_change, int major);
840 /* Seek 'fd' to start of write-intent-bitmap. Must be an
841 * md-native format bitmap
842 */
843 void (*locate_bitmap)(struct supertype *st, int fd);
844 /* if add_internal_bitmap succeeded for existing array, this
845 * writes it out.
846 */
847 int (*write_bitmap)(struct supertype *st, int fd);
848 /* Free the superblock and any other allocated data */
849 void (*free_super)(struct supertype *st);
850
851 /* validate_geometry is called with an st returned by
852 * match_metadata_desc.
853 * It should check that the geometry described is compatible with
854 * the metadata type. It will be called repeatedly as devices
855 * added to validate changing size and new devices. If there are
856 * inter-device dependencies, it should record sufficient details
857 * so these can be validated.
858 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
859 * Return value is:
860 * 1: everything is OK
861 * 0: not OK for some reason - if 'verbose', then error was reported.
862 * -1: st->sb was NULL, 'subdev' is a member of a container of this
863 * type, but array is not acceptable for some reason
864 * message was reported even if verbose is 0.
865 */
866 int (*validate_geometry)(struct supertype *st, int level, int layout,
867 int raiddisks,
868 int *chunk, unsigned long long size,
869 unsigned long long data_offset,
870 char *subdev, unsigned long long *freesize,
871 int verbose);
872
873 /* Return a linked list of 'mdinfo' structures for all arrays
874 * in the container. For non-containers, it is like
875 * getinfo_super with an allocated mdinfo.*/
876 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
877 /* query the supertype for default geometry */
878 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
879 /* Permit subarray's to be deleted from inactive containers */
880 int (*kill_subarray)(struct supertype *st); /* optional */
881 /* Permit subarray's to be modified */
882 int (*update_subarray)(struct supertype *st, char *subarray,
883 char *update, struct mddev_ident *ident); /* optional */
884 /* Check if reshape is supported for this external format.
885 * st is obtained from super_by_fd() where st->subarray[0] is
886 * initialized to indicate if reshape is being performed at the
887 * container or subarray level
888 */
889 #define APPLY_METADATA_CHANGES 1
890 #define ROLLBACK_METADATA_CHANGES 0
891
892 int (*reshape_super)(struct supertype *st,
893 unsigned long long size, int level,
894 int layout, int chunksize, int raid_disks,
895 int delta_disks, char *backup, char *dev,
896 int direction,
897 int verbose); /* optional */
898 int (*manage_reshape)( /* optional */
899 int afd, struct mdinfo *sra, struct reshape *reshape,
900 struct supertype *st, unsigned long blocks,
901 int *fds, unsigned long long *offsets,
902 int dests, int *destfd, unsigned long long *destoffsets);
903
904 /* for mdmon */
905 int (*open_new)(struct supertype *c, struct active_array *a,
906 char *inst);
907
908 /* Tell the metadata handler the current state of the array.
909 * This covers whether it is known to be consistent (no pending writes)
910 * and how far along a resync is known to have progressed
911 * (in a->resync_start).
912 * resync status is really irrelevant if the array is not consistent,
913 * but some metadata (DDF!) have a place to record the distinction.
914 * If 'consistent' is '2', then the array can mark it dirty if a
915 * resync/recovery/whatever is required, or leave it clean if not.
916 * Return value is 0 dirty (not consistent) and 1 if clean.
917 * it is only really important if consistent is passed in as '2'.
918 */
919 int (*set_array_state)(struct active_array *a, int consistent);
920
921 /* When the state of a device might have changed, we call set_disk to
922 * tell the metadata what the current state is.
923 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
924 * transitions.
925 * set_disk might be called when the state of the particular disk has
926 * not in fact changed.
927 */
928 void (*set_disk)(struct active_array *a, int n, int state);
929 void (*sync_metadata)(struct supertype *st);
930 void (*process_update)(struct supertype *st,
931 struct metadata_update *update);
932 void (*prepare_update)(struct supertype *st,
933 struct metadata_update *update);
934
935 /* activate_spare will check if the array is degraded and, if it
936 * is, try to find some spare space in the container.
937 * On success, it add appropriate updates (For process_update) to
938 * to the 'updates' list and returns a list of 'mdinfo' identifying
939 * the device, or devices as there might be multiple missing
940 * devices and multiple spares available.
941 */
942 struct mdinfo *(*activate_spare)(struct active_array *a,
943 struct metadata_update **updates);
944 /*
945 * Return statically allocated string that represents metadata specific
946 * controller domain of the disk. The domain is used in disk domain
947 * matching functions. Disks belong to the same domain if the they have
948 * the same domain from mdadm.conf and belong the same metadata domain.
949 * Returning NULL or not providing this handler means that metadata
950 * does not distinguish the differences between disks that belong to
951 * different controllers. They are in the domain specified by
952 * configuration file (mdadm.conf).
953 * In case when the metadata has the notion of domains based on disk
954 * it shall return NULL for disks that do not belong to the controller
955 * the supported domains. Such disks will form another domain and won't
956 * be mixed with supported ones.
957 */
958 const char *(*get_disk_controller_domain)(const char *path);
959
960 /* for external backup area */
961 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
962
963 /* validate container after assemble */
964 int (*validate_container)(struct mdinfo *info);
965
966 int swapuuid; /* true if uuid is bigending rather than hostendian */
967 int external;
968 const char *name; /* canonical metadata name */
969 } *superlist[];
970
971 extern struct superswitch super0, super1;
972 extern struct superswitch super_imsm, super_ddf;
973 extern struct superswitch mbr, gpt;
974
975 struct metadata_update {
976 int len;
977 char *buf;
978 void *space; /* allocated space that monitor will use */
979 void **space_list; /* list of allocated spaces that monitor can
980 * use or that it returned.
981 */
982 struct metadata_update *next;
983 };
984
985 /* A supertype holds a particular collection of metadata.
986 * It identifies the metadata type by the superswitch, and the particular
987 * sub-version of that metadata type.
988 * metadata read in or created is stored in 'sb' and 'info'.
989 * There are also fields used by mdmon to track containers.
990 *
991 * A supertype may refer to:
992 * Just an array, possibly in a container
993 * A container, not identifying any particular array
994 * Info read from just one device, not yet fully describing the array/container.
995 *
996 *
997 * A supertype is created by:
998 * super_by_fd
999 * guess_super
1000 * dup_super
1001 */
1002 struct supertype {
1003 struct superswitch *ss;
1004 int minor_version;
1005 int max_devs;
1006 char container_devnm[32]; /* devnm of container */
1007 void *sb;
1008 void *info;
1009 void *other; /* Hack used to convert v0.90 to v1.0 */
1010 unsigned long long devsize;
1011 unsigned long long data_offset; /* used by v1.x only */
1012 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1013 HW/firmware related incompatability to load metadata.
1014 Used when examining metadata to display content of disk
1015 when user has no hw/firmare compatible system.
1016 */
1017 struct metadata_update *updates;
1018 struct metadata_update **update_tail;
1019
1020 /* extra stuff used by mdmon */
1021 struct active_array *arrays;
1022 int sock; /* listen to external programs */
1023 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1024 * external:/md0/12
1025 */
1026 int devcnt;
1027 int retry_soon;
1028
1029 struct mdinfo *devs;
1030
1031 };
1032
1033 extern struct supertype *super_by_fd(int fd, char **subarray);
1034 enum guess_types { guess_any, guess_array, guess_partitions };
1035 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1036 static inline struct supertype *guess_super(int fd) {
1037 return guess_super_type(fd, guess_any);
1038 }
1039 extern struct supertype *dup_super(struct supertype *st);
1040 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1041 extern int must_be_container(int fd);
1042 extern int dev_size_from_id(dev_t id, unsigned long long *size);
1043 void wait_for(char *dev, int fd);
1044
1045 /*
1046 * Data structures for policy management.
1047 * Each device can have a policy structure that lists
1048 * various name/value pairs each possibly with a metadata associated.
1049 * The policy list is sorted by name/value/metadata
1050 */
1051 struct dev_policy {
1052 struct dev_policy *next;
1053 char *name; /* None of these strings are allocated. They are
1054 * all just references to strings which are known
1055 * to exist elsewhere.
1056 * name and metadata can be compared by address equality.
1057 */
1058 const char *metadata;
1059 const char *value;
1060 };
1061
1062 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1063
1064 /* iterate over the sublist starting at list, having the same
1065 * 'name' as 'list', and matching the given metadata (Where
1066 * NULL matches anything
1067 */
1068 #define pol_for_each(item, list, _metadata) \
1069 for (item = list; \
1070 item && item->name == list->name; \
1071 item = item->next) \
1072 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1073 ; else
1074
1075 /*
1076 * policy records read from mdadm are largely just name-value pairs.
1077 * The names are constants, not strdupped
1078 */
1079 struct pol_rule {
1080 struct pol_rule *next;
1081 char *type; /* rule_policy or rule_part */
1082 struct rule {
1083 struct rule *next;
1084 char *name;
1085 char *value;
1086 char *dups; /* duplicates of 'value' with a partNN appended */
1087 } *rule;
1088 };
1089
1090 extern char rule_policy[], rule_part[];
1091 extern char rule_path[], rule_type[];
1092 extern char type_part[], type_disk[];
1093
1094 extern void policyline(char *line, char *type);
1095 extern void policy_add(char *type, ...);
1096 extern void policy_free(void);
1097
1098 extern struct dev_policy *path_policy(char *path, char *type);
1099 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1100 extern struct dev_policy *devid_policy(int devid);
1101 extern void dev_policy_free(struct dev_policy *p);
1102
1103 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1104 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1105 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1106
1107 enum policy_action {
1108 act_default,
1109 act_include,
1110 act_re_add,
1111 act_spare, /* This only applies to bare devices */
1112 act_spare_same_slot, /* this allows non-bare devices,
1113 * but only if recent removal */
1114 act_force_spare, /* this allow non-bare devices in any case */
1115 act_err
1116 };
1117
1118 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1119 enum policy_action want);
1120 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1121 enum policy_action want);
1122
1123 struct domainlist {
1124 struct domainlist *next;
1125 const char *dom;
1126 };
1127
1128 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1129 const char *metadata);
1130 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1131 const char *metadata);
1132 extern void domainlist_add_dev(struct domainlist **dom, int devid,
1133 const char *metadata);
1134 extern void domain_free(struct domainlist *dl);
1135 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1136 const char *metadata);
1137 void domain_add(struct domainlist **domp, char *domain);
1138
1139 extern void policy_save_path(char *id_path, struct map_ent *array);
1140 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1141
1142 #if __GNUC__ < 3
1143 struct stat64;
1144 #endif
1145
1146 #define HAVE_NFTW we assume
1147 #define HAVE_FTW
1148
1149 #ifdef __UCLIBC__
1150 # include <features.h>
1151 # ifndef __UCLIBC_HAS_LFS__
1152 # define lseek64 lseek
1153 # endif
1154 # ifndef __UCLIBC_HAS_FTW__
1155 # undef HAVE_FTW
1156 # undef HAVE_NFTW
1157 # endif
1158 #endif
1159
1160 #ifdef __dietlibc__
1161 # undef HAVE_NFTW
1162 #endif
1163
1164 #if defined(__KLIBC__)
1165 # undef HAVE_NFTW
1166 # undef HAVE_FTW
1167 #endif
1168
1169 #ifndef HAVE_NFTW
1170 # define FTW_PHYS 1
1171 # ifndef HAVE_FTW
1172 struct FTW {};
1173 # endif
1174 #endif
1175
1176 #ifdef HAVE_FTW
1177 # include <ftw.h>
1178 #endif
1179
1180 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1181
1182 extern int Manage_ro(char *devname, int fd, int readonly);
1183 extern int Manage_run(char *devname, int fd, struct context *c);
1184 extern int Manage_stop(char *devname, int fd, int quiet,
1185 int will_retry);
1186 extern int Manage_subdevs(char *devname, int fd,
1187 struct mddev_dev *devlist, int verbose, int test,
1188 char *update, int force);
1189 extern int autodetect(void);
1190 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1191 extern int Grow_addbitmap(char *devname, int fd,
1192 struct context *c, struct shape *s);
1193 extern int Grow_reshape(char *devname, int fd,
1194 struct mddev_dev *devlist,
1195 unsigned long long data_offset,
1196 struct context *c, struct shape *s);
1197 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1198 int *fdlist, int cnt, char *backup_file, int verbose);
1199 extern int Grow_continue(int mdfd, struct supertype *st,
1200 struct mdinfo *info, char *backup_file,
1201 int forked, int freeze_reshape);
1202
1203 extern int restore_backup(struct supertype *st,
1204 struct mdinfo *content,
1205 int working_disks,
1206 int spares,
1207 char **backup_filep,
1208 int verbose);
1209 extern int Grow_continue_command(char *devname, int fd,
1210 char *backup_file, int verbose);
1211
1212 extern int Assemble(struct supertype *st, char *mddev,
1213 struct mddev_ident *ident,
1214 struct mddev_dev *devlist,
1215 struct context *c);
1216
1217 extern int Build(char *mddev, struct mddev_dev *devlist,
1218 struct shape *s, struct context *c);
1219
1220 extern int Create(struct supertype *st, char *mddev,
1221 char *name, int *uuid,
1222 int subdevs, struct mddev_dev *devlist,
1223 struct shape *s,
1224 struct context *c,
1225 unsigned long long data_offset);
1226
1227 extern int Detail(char *dev, struct context *c);
1228 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1229 extern int Query(char *dev);
1230 extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1231 extern int Examine(struct mddev_dev *devlist, struct context *c,
1232 struct supertype *forcest);
1233 extern int Monitor(struct mddev_dev *devlist,
1234 char *mailaddr, char *alert_cmd,
1235 struct context *c,
1236 int daemonise, int oneshot,
1237 int dosyslog, char *pidfile, int increments,
1238 int share);
1239
1240 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1241 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1242 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1243 extern int Wait(char *dev);
1244 extern int WaitClean(char *dev, int sock, int verbose);
1245 extern int SetAction(char *dev, char *action);
1246
1247 extern int Incremental(struct mddev_dev *devlist, struct context *c,
1248 struct supertype *st);
1249 extern void RebuildMap(void);
1250 extern int IncrementalScan(struct context *c, char *devnm);
1251 extern int IncrementalRemove(char *devname, char *path, int verbose);
1252 extern int CreateBitmap(char *filename, int force, char uuid[16],
1253 unsigned long chunksize, unsigned long daemon_sleep,
1254 unsigned long write_behind,
1255 unsigned long long array_size,
1256 int major);
1257 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1258 extern int Write_rules(char *rule_name);
1259 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1260 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1261 extern int Dump_metadata(char *dev, char *dir, struct context *c,
1262 struct supertype *st);
1263 extern int Restore_metadata(char *dev, char *dir, struct context *c,
1264 struct supertype *st, int only);
1265
1266 extern int md_get_version(int fd);
1267 extern int get_linux_version(void);
1268 extern int mdadm_version(char *version);
1269 extern unsigned long long parse_size(char *size);
1270 extern int parse_uuid(char *str, int uuid[4]);
1271 extern int parse_layout_10(char *layout);
1272 extern int parse_layout_faulty(char *layout);
1273 extern long parse_num(char *num);
1274 extern int check_ext2(int fd, char *name);
1275 extern int check_reiser(int fd, char *name);
1276 extern int check_raid(int fd, char *name);
1277 extern int check_partitions(int fd, char *dname,
1278 unsigned long long freesize,
1279 unsigned long long size);
1280
1281 extern int get_mdp_major(void);
1282 extern int dev_open(char *dev, int flags);
1283 extern int open_dev(char *devnm);
1284 extern void reopen_mddev(int mdfd);
1285 extern int open_dev_flags(char *devnm, int flags);
1286 extern int open_dev_excl(char *devnm);
1287 extern int is_standard(char *dev, int *nump);
1288 extern int same_dev(char *one, char *two);
1289 extern int compare_paths (char* path1,char* path2);
1290 extern void enable_fds(int devices);
1291
1292 extern int parse_auto(char *str, char *msg, int config);
1293 extern struct mddev_ident *conf_get_ident(char *dev);
1294 extern struct mddev_dev *conf_get_devs(void);
1295 extern int conf_test_dev(char *devname);
1296 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1297 extern struct createinfo *conf_get_create_info(void);
1298 extern void set_conffile(char *file);
1299 extern char *conf_get_mailaddr(void);
1300 extern char *conf_get_mailfrom(void);
1301 extern char *conf_get_program(void);
1302 extern char *conf_get_homehost(int *require_homehostp);
1303 extern char *conf_line(FILE *file);
1304 extern char *conf_word(FILE *file, int allow_key);
1305 extern void print_quoted(char *str);
1306 extern void print_escape(char *str);
1307 extern int use_udev(void);
1308 extern unsigned long GCD(unsigned long a, unsigned long b);
1309 extern int conf_name_is_free(char *name);
1310 extern int conf_verify_devnames(struct mddev_ident *array_list);
1311 extern int devname_matches(char *name, char *match);
1312 extern struct mddev_ident *conf_match(struct supertype *st,
1313 struct mdinfo *info,
1314 char *devname,
1315 int verbose, int *rvp);
1316 extern int experimental(void);
1317
1318 extern void free_line(char *line);
1319 extern int match_oneof(char *devices, char *devname);
1320 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1321 extern const int uuid_zero[4];
1322 extern int same_uuid(int a[4], int b[4], int swapuuid);
1323 extern void copy_uuid(void *a, int b[4], int swapuuid);
1324 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1325 extern char *fname_from_uuid(struct supertype *st,
1326 struct mdinfo *info, char *buf, char sep);
1327 extern unsigned long calc_csum(void *super, int bytes);
1328 extern int enough(int level, int raid_disks, int layout, int clean,
1329 char *avail);
1330 extern int enough_fd(int fd);
1331 extern int ask(char *mesg);
1332 extern unsigned long long get_component_size(int fd);
1333 extern void remove_partitions(int fd);
1334 extern int test_partition(int fd);
1335 extern int test_partition_from_id(dev_t id);
1336 extern int get_data_disks(int level, int layout, int raid_disks);
1337 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1338 int chunksize, unsigned long long devsize);
1339 extern int flush_metadata_updates(struct supertype *st);
1340 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1341 extern int assemble_container_content(struct supertype *st, int mdfd,
1342 struct mdinfo *content,
1343 struct context *c,
1344 char *chosen_name, int *result);
1345 #define INCR_NO 1
1346 #define INCR_UNSAFE 2
1347 #define INCR_ALREADY 4
1348 #define INCR_YES 8
1349 extern struct mdinfo *container_choose_spares(struct supertype *st,
1350 unsigned long long min_size,
1351 struct domainlist *domlist,
1352 char *spare_group,
1353 const char *metadata, int get_one);
1354 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1355 extern int add_disk(int mdfd, struct supertype *st,
1356 struct mdinfo *sra, struct mdinfo *info);
1357 extern int remove_disk(int mdfd, struct supertype *st,
1358 struct mdinfo *sra, struct mdinfo *info);
1359 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1360 unsigned long long min_recovery_start(struct mdinfo *array);
1361
1362 extern char *human_size(long long bytes);
1363 extern char *human_size_brief(long long bytes, int prefix);
1364 extern void print_r10_layout(int layout);
1365
1366 extern char *find_free_devnm(int use_partitions);
1367
1368 extern void put_md_name(char *name);
1369 extern char *devid2kname(int devid);
1370 extern char *devid2devnm(int devid);
1371 extern int devnm2devid(char *devnm);
1372 extern char *get_md_name(char *devnm);
1373
1374 extern char DefaultConfFile[];
1375
1376 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1377 char *chosen);
1378 /* values for 'trustworthy' */
1379 #define LOCAL 1
1380 #define LOCAL_ANY 10
1381 #define FOREIGN 2
1382 #define METADATA 3
1383 extern int open_mddev(char *dev, int report_errors);
1384 extern int open_container(int fd);
1385 extern int metadata_container_matches(char *metadata, char *devnm);
1386 extern int metadata_subdev_matches(char *metadata, char *devnm);
1387 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1388 extern int is_subarray_active(char *subarray, char *devname);
1389 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1390 extern struct superswitch *version_to_superswitch(char *vers);
1391
1392 extern int mdmon_running(char *devnm);
1393 extern int mdmon_pid(char *devnm);
1394 extern int check_env(char *name);
1395 extern __u32 random32(void);
1396 extern int start_mdmon(char *devnm);
1397
1398 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1399 struct supertype *st, unsigned long stripes,
1400 int *fds, unsigned long long *offsets,
1401 int dests, int *destfd, unsigned long long *destoffsets);
1402 void abort_reshape(struct mdinfo *sra);
1403
1404 void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1405
1406 extern void fmt_devname(char *name, int num);
1407 extern char *stat2devnm(struct stat *st);
1408 extern char *fd2devnm(int fd);
1409
1410 extern int in_initrd(void);
1411
1412 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1413 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1414 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1415 (ROUND_UP((unsigned long)(ptr), base)))
1416
1417 static inline int is_subarray(char *vers)
1418 {
1419 /* The version string for a 'subarray' (an array in a container)
1420 * is
1421 * /containername/componentname for normal read-write arrays
1422 * -containername/componentname for arrays which mdmon must not
1423 * reconfigure. They might be read-only
1424 * or might be undergoing reshape etc.
1425 * containername is e.g. md0, md_d1
1426 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1427 */
1428 return (*vers == '/' || *vers == '-');
1429 }
1430
1431 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1432 {
1433 return &ent->metadata_version[10+strlen(container)+1];
1434 }
1435
1436 #ifdef DEBUG
1437 #define dprintf(fmt, arg...) \
1438 fprintf(stderr, fmt, ##arg)
1439 #else
1440 #define dprintf(fmt, arg...) \
1441 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1442 #endif
1443 #include <assert.h>
1444 #include <stdarg.h>
1445 static inline int xasprintf(char **strp, const char *fmt, ...) {
1446 va_list ap;
1447 int ret;
1448 va_start(ap, fmt);
1449 ret = vasprintf(strp, fmt, ap);
1450 va_end(ap);
1451 assert(ret >= 0);
1452 return ret;
1453 }
1454
1455 #define pr_err(fmt ...) fprintf(stderr, Name ": " fmt)
1456 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1457
1458 void *xmalloc(size_t len);
1459 void *xrealloc(void *ptr, size_t len);
1460 void *xcalloc(size_t num, size_t size);
1461 char *xstrdup(const char *str);
1462
1463 #define LEVEL_MULTIPATH (-4)
1464 #define LEVEL_LINEAR (-1)
1465 #define LEVEL_FAULTY (-5)
1466
1467 /* kernel module doesn't know about these */
1468 #define LEVEL_CONTAINER (-100)
1469 #define LEVEL_UNSUPPORTED (-200)
1470
1471 /* the kernel does know about this one ... */
1472 #define LEVEL_NONE (-1000000)
1473
1474 /* faulty stuff */
1475
1476 #define WriteTransient 0
1477 #define ReadTransient 1
1478 #define WritePersistent 2
1479 #define ReadPersistent 3
1480 #define WriteAll 4 /* doesn't go to device */
1481 #define ReadFixable 5
1482 #define Modes 6
1483
1484 #define ClearErrors 31
1485 #define ClearFaults 30
1486
1487 #define AllPersist 100 /* internal use only */
1488 #define NoPersist 101
1489
1490 #define ModeMask 0x1f
1491 #define ModeShift 5
1492
1493 #ifdef __TINYC__
1494 #undef minor
1495 #undef major
1496 #undef makedev
1497 #define minor(x) ((x)&0xff)
1498 #define major(x) (((x)>>8)&0xff)
1499 #define makedev(M,m) (((M)<<8) | (m))
1500 #endif
1501
1502 /* for raid4/5/6 */
1503 #define ALGORITHM_LEFT_ASYMMETRIC 0
1504 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1505 #define ALGORITHM_LEFT_SYMMETRIC 2
1506 #define ALGORITHM_RIGHT_SYMMETRIC 3
1507
1508 /* Define non-rotating (raid4) algorithms. These allow
1509 * conversion of raid4 to raid5.
1510 */
1511 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1512 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1513
1514 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1515 * Firstly, the exact positioning of the parity block is slightly
1516 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1517 * of DDF.
1518 * Secondly, or order of datablocks over which the Q syndrome is computed
1519 * is different.
1520 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1521 * These layouts are from the DDFv1.2 spec.
1522 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1523 * leaves RLQ=3 as 'Vendor Specific'
1524 */
1525
1526 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1527 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1528 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1529
1530 /* For every RAID5 algorithm we define a RAID6 algorithm
1531 * with exactly the same layout for data and parity, and
1532 * with the Q block always on the last device (N-1).
1533 * This allows trivial conversion from RAID5 to RAID6
1534 */
1535 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1536 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1537 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1538 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1539 #define ALGORITHM_PARITY_0_6 20
1540 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1541
1542 /* Define PATH_MAX in case we don't use glibc or standard library does
1543 * not have PATH_MAX defined. Assume max path length is 4K characters.
1544 */
1545 #ifndef PATH_MAX
1546 #define PATH_MAX 4096
1547 #endif
1548
1549 #define RESYNC_NONE -1
1550 #define RESYNC_DELAYED -2
1551 #define RESYNC_PENDING -3
1552 #define RESYNC_UNKNOWN -4
1553
1554 /* When using "GET_DISK_INFO" it isn't certain how high
1555 * we need to check. So we impose an absolute limit of
1556 * MAX_DISKS. This needs to be much more than the largest
1557 * number of devices any metadata can support. Currently
1558 * v1.x can support 1920
1559 */
1560 #define MAX_DISKS 4096
1561
1562 /* Sometimes the 'size' value passed needs to mean "Maximum".
1563 * In those cases with use MAX_SIZE
1564 */
1565 #define MAX_SIZE 1
1566
1567 /* We want to use unsigned numbers for sector counts, but need
1568 * a value for 'invalid'. Use '1'.
1569 */
1570 #define INVALID_SECTORS 1
1571 /* And another special number needed for --data_offset=variable */
1572 #define VARIABLE_OFFSET 3