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