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