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