]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Add --data-offset flag for Create and Grow
[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 #if !defined(__dietlibc__) && !defined(__KLIBC__)
29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
30 #else
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 component_size; /* same as array.size, except in
192 * sectors and up to 64bits.
193 */
194 unsigned long long custom_array_size; /* size for non-default sized
195 * arrays (in sectors)
196 */
197 #define NO_RESHAPE 0
198 #define VOLUME_RESHAPE 1
199 #define CONTAINER_RESHAPE 2
200 int reshape_active;
201 unsigned long long reshape_progress;
202 int recovery_blocked; /* for external metadata it
203 * indicates that there is
204 * reshape in progress in
205 * container,
206 * for native metadata it is
207 * reshape_active field mirror
208 */
209 union {
210 unsigned long long resync_start; /* per-array resync position */
211 unsigned long long recovery_start; /* per-device rebuild position */
212 #define MaxSector (~0ULL) /* resync/recovery complete position */
213 };
214 long bitmap_offset; /* 0 == none, 1 == a file */
215 unsigned long safe_mode_delay; /* ms delay to mark clean */
216 int new_level, delta_disks, new_layout, new_chunk;
217 int errors;
218 unsigned long cache_size; /* size of raid456 stripe cache*/
219 int mismatch_cnt;
220 char text_version[50];
221
222 int container_member; /* for assembling external-metatdata arrays
223 * This is to be used internally by metadata
224 * handler only */
225 int container_enough; /* flag external handlers can set to
226 * indicate that subarrays have not enough (-1),
227 * enough to start (0), or all expected disks (1) */
228 char sys_name[20];
229 struct mdinfo *devs;
230 struct mdinfo *next;
231
232 /* Device info for mdmon: */
233 int recovery_fd;
234 int state_fd;
235 #define DS_FAULTY 1
236 #define DS_INSYNC 2
237 #define DS_WRITE_MOSTLY 4
238 #define DS_SPARE 8
239 #define DS_BLOCKED 16
240 #define DS_REMOVE 1024
241 #define DS_UNBLOCK 2048
242 int prev_state, curr_state, next_state;
243
244 };
245
246 struct createinfo {
247 int uid;
248 int gid;
249 int autof;
250 int mode;
251 int symlinks;
252 struct supertype *supertype;
253 };
254
255 #define Name "mdadm"
256
257 enum mode {
258 ASSEMBLE=1,
259 BUILD,
260 CREATE,
261 MANAGE,
262 MISC,
263 MONITOR,
264 GROW,
265 INCREMENTAL,
266 AUTODETECT,
267 mode_count
268 };
269
270 extern char short_options[];
271 extern char short_bitmap_options[];
272 extern char short_bitmap_auto_options[];
273 extern struct option long_options[];
274 extern char Version[], Usage[], Help[], OptionHelp[],
275 *mode_help[],
276 Help_create[], Help_build[], Help_assemble[], Help_grow[],
277 Help_incr[],
278 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
279
280 /* for option that don't have short equivilents, we assign arbitrary
281 * numbers later than any 'short' character option.
282 */
283 enum special_options {
284 AssumeClean = 300,
285 BitmapChunk,
286 WriteBehind,
287 ReAdd,
288 NoDegraded,
289 Sparc22,
290 BackupFile,
291 HomeHost,
292 AutoHomeHost,
293 Symlinks,
294 AutoDetect,
295 Waitclean,
296 DetailPlatform,
297 KillSubarray,
298 UpdateSubarray,
299 IncrementalPath,
300 NoSharing,
301 HelpOptions,
302 Brief,
303 ManageOpt,
304 Add,
305 Remove,
306 Fail,
307 MiscOpt,
308 WaitOpt,
309 ConfigFile,
310 ChunkSize,
311 WriteMostly,
312 Layout,
313 Auto,
314 Force,
315 SuperMinor,
316 EMail,
317 ProgramOpt,
318 Increment,
319 Fork,
320 Bitmap,
321 RebuildMapOpt,
322 InvalidBackup,
323 UdevRules,
324 FreezeReshape,
325 Continue,
326 OffRootOpt,
327 Prefer,
328 KillOpt,
329 DataOffset,
330 };
331
332 enum prefix_standard {
333 JEDEC,
334 IEC
335 };
336
337 /* structures read from config file */
338 /* List of mddevice names and identifiers
339 * Identifiers can be:
340 * uuid=128-hex-uuid
341 * super-minor=decimal-minor-number-from-superblock
342 * devices=comma,separated,list,of,device,names,with,wildcards
343 *
344 * If multiple fields are present, the intersection of all matching
345 * devices is considered
346 */
347 #define UnSet (0xfffe)
348 struct mddev_ident {
349 char *devname;
350
351 int uuid_set;
352 int uuid[4];
353 char name[33];
354
355 int super_minor;
356
357 char *devices; /* comma separated list of device
358 * names with wild cards
359 */
360 int level;
361 int raid_disks;
362 int spare_disks;
363 struct supertype *st;
364 int autof; /* 1 for normal, 2 for partitioned */
365 char *spare_group;
366 char *bitmap_file;
367 int bitmap_fd;
368
369 char *container; /* /dev/whatever name of container, or
370 * uuid of container. You would expect
371 * this to be the 'devname' or UUID
372 * of some other entry.
373 */
374 char *member; /* subarray within a container */
375
376 struct mddev_ident *next;
377 union {
378 /* fields needed by different users of this structure */
379 int assembled; /* set when assembly succeeds */
380 };
381 };
382
383 struct context {
384 int readonly;
385 int runstop;
386 int verbose;
387 int brief;
388 int force;
389 char *homehost;
390 int require_homehost;
391 char *prefer;
392 int export;
393 int test;
394 char *subarray;
395 char *update;
396 int scan;
397 int SparcAdjust;
398 int autof;
399 int delay;
400 int freeze_reshape;
401 char *backup_file;
402 int invalid_backup;
403 };
404
405 struct shape {
406 int raiddisks;
407 int sparedisks;
408 int level;
409 int layout;
410 char *layout_str;
411 int chunk;
412 int bitmap_chunk;
413 char *bitmap_file;
414 int assume_clean;
415 int write_behind;
416 unsigned long long size;
417 };
418
419 /* List of device names - wildcards expanded */
420 struct mddev_dev {
421 char *devname;
422 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
423 * 'A' for re_add.
424 * Not set for names read from .config
425 */
426 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
427 char used; /* set when used */
428 struct mddev_dev *next;
429 };
430
431 typedef struct mapping {
432 char *name;
433 int num;
434 } mapping_t;
435
436
437 struct mdstat_ent {
438 char *dev;
439 int devnum;
440 int active;
441 char *level;
442 char *pattern; /* U or up, _ for down */
443 int percent; /* -1 if no resync */
444 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
445 int devcnt;
446 int raid_disks;
447 char * metadata_version;
448 struct dev_member {
449 char *name;
450 struct dev_member *next;
451 } *members;
452 struct mdstat_ent *next;
453 };
454
455 extern struct mdstat_ent *mdstat_read(int hold, int start);
456 extern void free_mdstat(struct mdstat_ent *ms);
457 extern void mdstat_wait(int seconds);
458 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
459 extern int mddev_busy(int devnum);
460 extern struct mdstat_ent *mdstat_by_component(char *name);
461 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container);
462
463 struct map_ent {
464 struct map_ent *next;
465 int devnum;
466 char metadata[20];
467 int uuid[4];
468 int bad;
469 char *path;
470 };
471 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
472 int uuid[4], char *path);
473 extern void map_remove(struct map_ent **map, int devnum);
474 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
475 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
476 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
477 extern void map_read(struct map_ent **melp);
478 extern int map_write(struct map_ent *mel);
479 extern void map_delete(struct map_ent **mapp, int devnum);
480 extern void map_free(struct map_ent *map);
481 extern void map_add(struct map_ent **melp,
482 int devnum, char *metadata, int uuid[4], char *path);
483 extern int map_lock(struct map_ent **melp);
484 extern void map_unlock(struct map_ent **melp);
485 extern void map_fork(void);
486
487 /* various details can be requested */
488 enum sysfs_read_flags {
489 GET_LEVEL = (1 << 0),
490 GET_LAYOUT = (1 << 1),
491 GET_COMPONENT = (1 << 2),
492 GET_CHUNK = (1 << 3),
493 GET_CACHE = (1 << 4),
494 GET_MISMATCH = (1 << 5),
495 GET_VERSION = (1 << 6),
496 GET_DISKS = (1 << 7),
497 GET_DEGRADED = (1 << 8),
498 GET_SAFEMODE = (1 << 9),
499 GET_BITMAP_LOCATION = (1 << 10),
500
501 GET_DEVS = (1 << 20), /* gets role, major, minor */
502 GET_OFFSET = (1 << 21),
503 GET_SIZE = (1 << 22),
504 GET_STATE = (1 << 23),
505 GET_ERROR = (1 << 24),
506 };
507
508 /* If fd >= 0, get the array it is open on,
509 * else use devnum. >=0 -> major9. <0.....
510 */
511 extern int sysfs_open(int devnum, char *devname, char *attr);
512 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
513 extern void sysfs_free(struct mdinfo *sra);
514 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
515 extern int sysfs_attr_match(const char *attr, const char *str);
516 extern int sysfs_match_word(const char *word, char **list);
517 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
518 char *name, char *val);
519 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
520 char *name, unsigned long long val);
521 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
522 char *name, long long val);
523 extern int sysfs_uevent(struct mdinfo *sra, char *event);
524 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
525 char *name);
526 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
527 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
528 char *name, unsigned long long *val);
529 extern int sysfs_fd_get_str(int fd, char *val, int size);
530 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
531 char *name);
532 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
533 char *name, char *val, int size);
534 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
535 extern int sysfs_set_array(struct mdinfo *info, int vers);
536 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
537 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
538 extern int sysfs_unique_holder(int devnum, long rdev);
539 extern int sysfs_freeze_array(struct mdinfo *sra);
540 extern int load_sys(char *path, char *buf);
541 extern int reshape_prepare_fdlist(char *devname,
542 struct mdinfo *sra,
543 int raid_disks,
544 int nrdisks,
545 unsigned long blocks,
546 char *backup_file,
547 int *fdlist,
548 unsigned long long *offsets);
549 extern void reshape_free_fdlist(int *fdlist,
550 unsigned long long *offsets,
551 int size);
552 extern int reshape_open_backup_file(char *backup,
553 int fd,
554 char *devname,
555 long blocks,
556 int *fdlist,
557 unsigned long long *offsets,
558 int restart);
559 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
560 unsigned int ndata, unsigned int odata);
561
562 extern int save_stripes(int *source, unsigned long long *offsets,
563 int raid_disks, int chunk_size, int level, int layout,
564 int nwrites, int *dest,
565 unsigned long long start, unsigned long long length,
566 char *buf);
567 extern int restore_stripes(int *dest, unsigned long long *offsets,
568 int raid_disks, int chunk_size, int level, int layout,
569 int source, unsigned long long read_offset,
570 unsigned long long start, unsigned long long length,
571 char *src_buf);
572
573 #ifndef Sendmail
574 #define Sendmail "/usr/lib/sendmail -t"
575 #endif
576
577 #define SYSLOG_FACILITY LOG_DAEMON
578
579 extern char *map_num(mapping_t *map, int num);
580 extern int map_name(mapping_t *map, char *name);
581 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
582
583 extern char *map_dev_preferred(int major, int minor, int create,
584 char *prefer);
585 static inline char *map_dev(int major, int minor, int create)
586 {
587 return map_dev_preferred(major, minor, create, NULL);
588 }
589
590 struct active_array;
591 struct metadata_update;
592
593
594 /* 'struct reshape' records the intermediate states
595 * a general reshape.
596 * The starting geometry is converted to the 'before' geometry
597 * by at most an atomic level change. They could be the same.
598 * Similarly the 'after' geometry is converted to the final
599 * geometry by at most a level change.
600 * Note that 'before' and 'after' must have the same level.
601 * 'blocks' is the minimum number of sectors for a reshape unit.
602 * This will be a multiple of the stripe size in each of the
603 * 'before' and 'after' geometries.
604 * If 'blocks' is 0, no restriping is necessary.
605 */
606 struct reshape {
607 int level;
608 int parity; /* number of parity blocks/devices */
609 struct {
610 int layout;
611 int data_disks;
612 } before, after;
613 unsigned long long backup_blocks;
614 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
615 unsigned long long new_size; /* New size of array in sectors */
616 };
617
618 /* A superswitch provides entry point the a metadata handler.
619 *
620 * The super_switch primarily operates on some "metadata" that
621 * is accessed via the 'supertype'.
622 * This metadata has one of three possible sources.
623 * 1/ It is read from a single device. In this case it may not completely
624 * describe the array or arrays as some information might be on other
625 * devices.
626 * 2/ It is read from all devices in a container. In this case all
627 * information is present.
628 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
629 * be complete once enough ->add_to_super calls have completed.
630 *
631 * When creating an array inside a container, the metadata will be
632 * formed by a combination of 2 and 3. The metadata or the array is read,
633 * then new information is added.
634 *
635 * The metadata must sometimes have a concept of a 'current' array
636 * and a 'current' device.
637 * The 'current' array is set by init_super to be the newly created array,
638 * or is set by super_by_fd when it finds it is looking at an array inside
639 * a container.
640 *
641 * The 'current' device is either the device that the metadata was read from
642 * in case 1, or the last device added by add_to_super in case 3.
643 * Case 2 does not identify a 'current' device.
644 */
645 extern struct superswitch {
646
647 /* Used to report details of metadata read from a component
648 * device. ->load_super has been called.
649 */
650 void (*examine_super)(struct supertype *st, char *homehost);
651 void (*brief_examine_super)(struct supertype *st, int verbose);
652 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
653 void (*export_examine_super)(struct supertype *st);
654
655 /* Used to report details of an active array.
656 * ->load_super was possibly given a 'component' string.
657 */
658 void (*detail_super)(struct supertype *st, char *homehost);
659 void (*brief_detail_super)(struct supertype *st);
660 void (*export_detail_super)(struct supertype *st);
661
662 /* Optional: platform hardware / firmware details */
663 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
664 int (*export_detail_platform)(int verbose, char *controller_path);
665
666 /* Used:
667 * to get uuid to storing in bitmap metadata
668 * and 'reshape' backup-data metadata
669 * To see if a device is being re-added to an array it was part of.
670 */
671 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
672
673 /* Extract generic details from metadata. This could be details about
674 * the container, or about an individual array within the container.
675 * The determination is made either by:
676 * load_super being given a 'component' string.
677 * validate_geometry determining what to create.
678 * The info includes both array information and device information.
679 * The particular device should be:
680 * The last device added by add_to_super
681 * The device the metadata was loaded from by load_super
682 * If 'map' is present, then it is an array raid_disks long
683 * (raid_disk must already be set and correct) and it is filled
684 * with 1 for slots that are thought to be active and 0 for slots which
685 * appear to be failed/missing.
686 * *info is zeroed out before data is added.
687 */
688 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
689 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
690 /* Check if the given metadata is flagged as belonging to "this"
691 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
692 */
693 int (*match_home)(struct supertype *st, char *homehost);
694
695 /* Make one of several generic modifications to metadata
696 * prior to assembly (or other times).
697 * sparc2.2 - first bug in early 0.90 metadata
698 * super-minor - change name of 0.90 metadata
699 * summaries - 'correct' any redundant data
700 * resync - mark array as dirty to trigger a resync.
701 * uuid - set new uuid - only 0.90 or 1.x
702 * name - change the name of the array (where supported)
703 * homehost - change which host this array is tied to.
704 * devicesize - If metadata is at start of device, change recorded
705 * device size to match actual device size
706 * byteorder - swap bytes for 0.90 metadata
707 *
708 * force-one - mark that device as uptodate, not old or failed.
709 * force-array - mark array as clean if it would not otherwise
710 * assemble
711 * assemble - not sure how this is different from force-one...
712 * linear-grow-new - add a new device to a linear array, but don't
713 * change the size: so superblock still matches
714 * linear-grow-update - now change the size of the array.
715 * writemostly - set the WriteMostly1 bit in the superblock devflags
716 * readwrite - clear the WriteMostly1 bit in the superblock devflags
717 */
718 int (*update_super)(struct supertype *st, struct mdinfo *info,
719 char *update,
720 char *devname, int verbose,
721 int uuid_set, char *homehost);
722
723 /* Create new metadata for new array as described. This could
724 * be a new container, or an array in a pre-existing container.
725 * Also used to zero metadata prior to writing it to invalidate old
726 * metadata.
727 */
728 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
729 unsigned long long size, char *name,
730 char *homehost, int *uuid,
731 unsigned long long data_offset);
732
733 /* update the metadata to include new device, either at create or
734 * when hot-adding a spare.
735 */
736 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
737 int fd, char *devname);
738 /* update the metadata to delete a device,
739 * when hot-removing.
740 */
741 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
742
743 /* Write metadata to one device when fixing problems or adding
744 * a new device.
745 */
746 int (*store_super)(struct supertype *st, int fd);
747
748 /* Write all metadata for this array.
749 */
750 int (*write_init_super)(struct supertype *st);
751 int (*compare_super)(struct supertype *st, struct supertype *tst);
752 int (*load_super)(struct supertype *st, int fd, char *devname);
753 int (*load_container)(struct supertype *st, int fd, char *devname);
754 struct supertype * (*match_metadata_desc)(char *arg);
755 __u64 (*avail_size)(struct supertype *st, __u64 size,
756 unsigned long long data_offset);
757 unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
758 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
759 int delay, int write_behind,
760 unsigned long long size, int may_change, int major);
761 void (*locate_bitmap)(struct supertype *st, int fd);
762 int (*write_bitmap)(struct supertype *st, int fd);
763 void (*free_super)(struct supertype *st);
764
765 /* validate_geometry is called with an st returned by
766 * match_metadata_desc.
767 * It should check that the geometry described in compatible with
768 * the metadata type. It will be called repeatedly as devices
769 * added to validate changing size and new devices. If there are
770 * inter-device dependencies, it should record sufficient details
771 * so these can be validated.
772 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
773 * Return value is:
774 * 1: everything is OK
775 * 0: not OK for some reason - if 'verbose', then error was reported.
776 * -1: st->sb was NULL, 'subdev' is a member of a container of this
777 * types, but array is not acceptable for some reason
778 * message was reported even if verbose is 0.
779 */
780 int (*validate_geometry)(struct supertype *st, int level, int layout,
781 int raiddisks,
782 int *chunk, unsigned long long size,
783 unsigned long long data_offset,
784 char *subdev, unsigned long long *freesize,
785 int verbose);
786
787 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
788 /* query the supertype for default geometry */
789 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
790 /* Permit subarray's to be deleted from inactive containers */
791 int (*kill_subarray)(struct supertype *st); /* optional */
792 /* Permit subarray's to be modified */
793 int (*update_subarray)(struct supertype *st, char *subarray,
794 char *update, struct mddev_ident *ident); /* optional */
795 /* Check if reshape is supported for this external format.
796 * st is obtained from super_by_fd() where st->subarray[0] is
797 * initialized to indicate if reshape is being performed at the
798 * container or subarray level
799 */
800 #define APPLY_METADATA_CHANGES 1
801 #define ROLLBACK_METADATA_CHANGES 0
802
803 int (*reshape_super)(struct supertype *st,
804 unsigned long long size, int level,
805 int layout, int chunksize, int raid_disks,
806 int delta_disks, char *backup, char *dev,
807 int direction,
808 int verbose); /* optional */
809 int (*manage_reshape)( /* optional */
810 int afd, struct mdinfo *sra, struct reshape *reshape,
811 struct supertype *st, unsigned long blocks,
812 int *fds, unsigned long long *offsets,
813 int dests, int *destfd, unsigned long long *destoffsets);
814
815 /* for mdmon */
816 int (*open_new)(struct supertype *c, struct active_array *a,
817 char *inst);
818
819 /* Tell the metadata handler the current state of the array.
820 * This covers whether it is known to be consistent (no pending writes)
821 * and how far along a resync is known to have progressed
822 * (in a->resync_start).
823 * resync status is really irrelevant if the array is not consistent,
824 * but some metadata (DDF!) have a place to record the distinction.
825 * If 'consistent' is '2', then the array can mark it dirty if a
826 * resync/recovery/whatever is required, or leave it clean if not.
827 * Return value is 0 dirty (not consistent) and 1 if clean.
828 * it is only really important if consistent is passed in as '2'.
829 */
830 int (*set_array_state)(struct active_array *a, int consistent);
831
832 /* When the state of a device might have changed, we call set_disk to
833 * tell the metadata what the current state is.
834 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
835 * transitions.
836 * set_disk might be called when the state of the particular disk has
837 * not in fact changed.
838 */
839 void (*set_disk)(struct active_array *a, int n, int state);
840 void (*sync_metadata)(struct supertype *st);
841 void (*process_update)(struct supertype *st,
842 struct metadata_update *update);
843 void (*prepare_update)(struct supertype *st,
844 struct metadata_update *update);
845
846 /* activate_spare will check if the array is degraded and, if it
847 * is, try to find some spare space in the container.
848 * On success, it add appropriate updates (For process_update) to
849 * to the 'updates' list and returns a list of 'mdinfo' identifying
850 * the device, or devices as there might be multiple missing
851 * devices and multiple spares available.
852 */
853 struct mdinfo *(*activate_spare)(struct active_array *a,
854 struct metadata_update **updates);
855 /*
856 * Return statically allocated string that represents metadata specific
857 * controller domain of the disk. The domain is used in disk domain
858 * matching functions. Disks belong to the same domain if the they have
859 * the same domain from mdadm.conf and belong the same metadata domain.
860 * Returning NULL or not providing this handler means that metadata
861 * does not distinguish the differences between disks that belong to
862 * different controllers. They are in the domain specified by
863 * configuration file (mdadm.conf).
864 * In case when the metadata has the notion of domains based on disk
865 * it shall return NULL for disks that do not belong to the controller
866 * the supported domains. Such disks will form another domain and won't
867 * be mixed with supported ones.
868 */
869 const char *(*get_disk_controller_domain)(const char *path);
870
871 /* for external backup area */
872 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
873
874 int swapuuid; /* true if uuid is bigending rather than hostendian */
875 int external;
876 const char *name; /* canonical metadata name */
877 } *superlist[];
878
879 extern struct superswitch super0, super1;
880 extern struct superswitch super_imsm, super_ddf;
881 extern struct superswitch mbr, gpt;
882
883 struct metadata_update {
884 int len;
885 char *buf;
886 void *space; /* allocated space that monitor will use */
887 void **space_list; /* list of allocated spaces that monitor can
888 * use or that it returned.
889 */
890 struct metadata_update *next;
891 };
892
893 /* A supertype holds a particular collection of metadata.
894 * It identifies the metadata type by the superswitch, and the particular
895 * sub-version of that metadata type.
896 * metadata read in or created is stored in 'sb' and 'info'.
897 * There are also fields used by mdmon to track containers.
898 *
899 * A supertype may refer to:
900 * Just an array, possibly in a container
901 * A container, not identifying any particular array
902 * Info read from just one device, not yet fully describing the array/container.
903 *
904 *
905 * A supertype is created by:
906 * super_by_fd
907 * guess_super
908 * dup_super
909 */
910 struct supertype {
911 struct superswitch *ss;
912 int minor_version;
913 int max_devs;
914 int container_dev; /* devnum of container */
915 void *sb;
916 void *info;
917 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
918 HW/firmware related incompatability to load metadata.
919 Used when examining metadata to display content of disk
920 when user has no hw/firmare compatible system.
921 */
922 struct metadata_update *updates;
923 struct metadata_update **update_tail;
924
925 /* extra stuff used by mdmon */
926 struct active_array *arrays;
927 int sock; /* listen to external programs */
928 int devnum;
929 char *devname; /* e.g. md0. This appears in metadata_verison:
930 * external:/md0/12
931 */
932 int devcnt;
933 int retry_soon;
934
935 struct mdinfo *devs;
936
937 };
938
939 extern struct supertype *super_by_fd(int fd, char **subarray);
940 enum guess_types { guess_any, guess_array, guess_partitions };
941 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
942 static inline struct supertype *guess_super(int fd) {
943 return guess_super_type(fd, guess_any);
944 }
945 extern struct supertype *dup_super(struct supertype *st);
946 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
947 extern int must_be_container(int fd);
948 extern int dev_size_from_id(dev_t id, unsigned long long *size);
949 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
950 mdu_disk_info_t *disk);
951 void wait_for(char *dev, int fd);
952
953 /*
954 * Data structures for policy management.
955 * Each device can have a policy structure that lists
956 * various name/value pairs each possibly with a metadata associated.
957 * The policy list is sorted by name/value/metadata
958 */
959 struct dev_policy {
960 struct dev_policy *next;
961 char *name; /* None of these strings are allocated. They are
962 * all just references to strings which are known
963 * to exist elsewhere.
964 * name and metadata can be compared by address equality.
965 */
966 const char *metadata;
967 const char *value;
968 };
969
970 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
971
972 /* iterate over the sublist starting at list, having the same
973 * 'name' as 'list', and matching the given metadata (Where
974 * NULL matches anything
975 */
976 #define pol_for_each(item, list, _metadata) \
977 for (item = list; \
978 item && item->name == list->name; \
979 item = item->next) \
980 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
981 ; else
982
983 /*
984 * policy records read from mdadm are largely just name-value pairs.
985 * The names are constants, not strdupped
986 */
987 struct pol_rule {
988 struct pol_rule *next;
989 char *type; /* rule_policy or rule_part */
990 struct rule {
991 struct rule *next;
992 char *name;
993 char *value;
994 char *dups; /* duplicates of 'value' with a partNN appended */
995 } *rule;
996 };
997
998 extern char rule_policy[], rule_part[];
999 extern char rule_path[], rule_type[];
1000 extern char type_part[], type_disk[];
1001
1002 extern void policyline(char *line, char *type);
1003 extern void policy_add(char *type, ...);
1004 extern void policy_free(void);
1005
1006 extern struct dev_policy *path_policy(char *path, char *type);
1007 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1008 extern struct dev_policy *devnum_policy(int dev);
1009 extern void dev_policy_free(struct dev_policy *p);
1010
1011 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1012 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1013 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1014
1015 enum policy_action {
1016 act_default,
1017 act_include,
1018 act_re_add,
1019 act_spare, /* This only applies to bare devices */
1020 act_spare_same_slot, /* this allows non-bare devices,
1021 * but only if recent removal */
1022 act_force_spare, /* this allow non-bare devices in any case */
1023 act_err
1024 };
1025
1026 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1027 enum policy_action want);
1028 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1029 enum policy_action want);
1030
1031 struct domainlist {
1032 struct domainlist *next;
1033 const char *dom;
1034 };
1035
1036 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1037 const char *metadata);
1038 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1039 const char *metadata);
1040 extern void domainlist_add_dev(struct domainlist **dom, int devnum,
1041 const char *metadata);
1042 extern void domain_free(struct domainlist *dl);
1043 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1044 const char *metadata);
1045 void domain_add(struct domainlist **domp, char *domain);
1046
1047 extern void policy_save_path(char *id_path, struct map_ent *array);
1048 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1049
1050 #if __GNUC__ < 3
1051 struct stat64;
1052 #endif
1053
1054 #define HAVE_NFTW we assume
1055 #define HAVE_FTW
1056
1057 #ifdef __UCLIBC__
1058 # include <features.h>
1059 # ifndef __UCLIBC_HAS_LFS__
1060 # define lseek64 lseek
1061 # endif
1062 # ifndef __UCLIBC_HAS_FTW__
1063 # undef HAVE_FTW
1064 # undef HAVE_NFTW
1065 # endif
1066 #endif
1067
1068 #ifdef __dietlibc__
1069 # undef HAVE_NFTW
1070 #endif
1071
1072 #if defined(__KLIBC__)
1073 # undef HAVE_NFTW
1074 # undef HAVE_FTW
1075 #endif
1076
1077 #ifndef HAVE_NFTW
1078 # define FTW_PHYS 1
1079 # ifndef HAVE_FTW
1080 struct FTW {};
1081 # endif
1082 #endif
1083
1084 #ifdef HAVE_FTW
1085 # include <ftw.h>
1086 #endif
1087
1088 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1089
1090
1091 extern int Manage_ro(char *devname, int fd, int readonly);
1092 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet,
1093 int will_retry);
1094 extern int Manage_subdevs(char *devname, int fd,
1095 struct mddev_dev *devlist, int verbose, int test,
1096 char *update, int force);
1097 extern int autodetect(void);
1098 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1099 extern int Grow_addbitmap(char *devname, int fd,
1100 struct context *c, struct shape *s);
1101 extern int Grow_reshape(char *devname, int fd,
1102 struct mddev_dev *devlist,
1103 unsigned long long data_offset,
1104 struct context *c, struct shape *s);
1105 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1106 int *fdlist, int cnt, char *backup_file, int verbose);
1107 extern int Grow_continue(int mdfd, struct supertype *st,
1108 struct mdinfo *info, char *backup_file,
1109 int freeze_reshape);
1110
1111 extern int restore_backup(struct supertype *st,
1112 struct mdinfo *content,
1113 int working_disks,
1114 int spares,
1115 char *backup_file,
1116 int verbose);
1117 extern int Grow_continue_command(char *devname, int fd,
1118 char *backup_file, int verbose);
1119
1120 extern int Assemble(struct supertype *st, char *mddev,
1121 struct mddev_ident *ident,
1122 struct mddev_dev *devlist,
1123 struct context *c);
1124
1125 extern int Build(char *mddev, struct mddev_dev *devlist,
1126 struct shape *s, struct context *c);
1127
1128 extern int Create(struct supertype *st, char *mddev,
1129 char *name, int *uuid,
1130 int subdevs, struct mddev_dev *devlist,
1131 struct shape *s,
1132 struct context *c,
1133 unsigned long long data_offset);
1134
1135 extern int Detail(char *dev, struct context *c);
1136 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1137 extern int Query(char *dev);
1138 extern int Examine(struct mddev_dev *devlist, struct context *c,
1139 struct supertype *forcest);
1140 extern int Monitor(struct mddev_dev *devlist,
1141 char *mailaddr, char *alert_cmd,
1142 struct context *c,
1143 int daemonise, int oneshot,
1144 int dosyslog, char *pidfile, int increments,
1145 int share);
1146
1147 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1148 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1149 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1150 extern int Wait(char *dev);
1151 extern int WaitClean(char *dev, int sock, int verbose);
1152
1153 extern int Incremental(char *devname, struct context *c,
1154 struct supertype *st);
1155 extern void RebuildMap(void);
1156 extern int IncrementalScan(int verbose);
1157 extern int IncrementalRemove(char *devname, char *path, int verbose);
1158 extern int CreateBitmap(char *filename, int force, char uuid[16],
1159 unsigned long chunksize, unsigned long daemon_sleep,
1160 unsigned long write_behind,
1161 unsigned long long array_size,
1162 int major);
1163 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1164 extern int Write_rules(char *rule_name);
1165 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1166 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1167
1168 extern int md_get_version(int fd);
1169 extern int get_linux_version(void);
1170 extern int mdadm_version(char *version);
1171 extern unsigned long long parse_size(char *size);
1172 extern int parse_uuid(char *str, int uuid[4]);
1173 extern int parse_layout_10(char *layout);
1174 extern int parse_layout_faulty(char *layout);
1175 extern long parse_num(char *num);
1176 extern int check_ext2(int fd, char *name);
1177 extern int check_reiser(int fd, char *name);
1178 extern int check_raid(int fd, char *name);
1179 extern int check_partitions(int fd, char *dname,
1180 unsigned long long freesize,
1181 unsigned long long size);
1182
1183 extern int get_mdp_major(void);
1184 extern int dev_open(char *dev, int flags);
1185 extern int open_dev(int devnum);
1186 extern int open_dev_flags(int devnum, int flags);
1187 extern int open_dev_excl(int devnum);
1188 extern int is_standard(char *dev, int *nump);
1189 extern int same_dev(char *one, char *two);
1190 extern int compare_paths (char* path1,char* path2);
1191
1192 extern int parse_auto(char *str, char *msg, int config);
1193 extern struct mddev_ident *conf_get_ident(char *dev);
1194 extern struct mddev_dev *conf_get_devs(void);
1195 extern int conf_test_dev(char *devname);
1196 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1197 extern struct createinfo *conf_get_create_info(void);
1198 extern void set_conffile(char *file);
1199 extern char *conf_get_mailaddr(void);
1200 extern char *conf_get_mailfrom(void);
1201 extern char *conf_get_program(void);
1202 extern char *conf_get_homehost(int *require_homehostp);
1203 extern char *conf_line(FILE *file);
1204 extern char *conf_word(FILE *file, int allow_key);
1205 extern void print_quoted(char *str);
1206 extern void print_escape(char *str);
1207 extern int conf_name_is_free(char *name);
1208 extern int conf_verify_devnames(struct mddev_ident *array_list);
1209 extern int devname_matches(char *name, char *match);
1210 extern struct mddev_ident *conf_match(struct supertype *st,
1211 struct mdinfo *info,
1212 char *devname,
1213 int verbose, int *rvp);
1214 extern int experimental(void);
1215
1216 extern void free_line(char *line);
1217 extern int match_oneof(char *devices, char *devname);
1218 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1219 extern const int uuid_zero[4];
1220 extern int same_uuid(int a[4], int b[4], int swapuuid);
1221 extern void copy_uuid(void *a, int b[4], int swapuuid);
1222 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1223 extern char *fname_from_uuid(struct supertype *st,
1224 struct mdinfo *info, char *buf, char sep);
1225 extern unsigned long calc_csum(void *super, int bytes);
1226 extern int enough(int level, int raid_disks, int layout, int clean,
1227 char *avail);
1228 extern int enough_fd(int fd);
1229 extern int ask(char *mesg);
1230 extern unsigned long long get_component_size(int fd);
1231 extern void remove_partitions(int fd);
1232 extern int test_partition(int fd);
1233 extern int test_partition_from_id(dev_t id);
1234 extern int get_data_disks(int level, int layout, int raid_disks);
1235 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1236 int chunksize, unsigned long long devsize);
1237 extern int flush_metadata_updates(struct supertype *st);
1238 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1239 extern int assemble_container_content(struct supertype *st, int mdfd,
1240 struct mdinfo *content,
1241 struct context *c,
1242 char *chosen_name);
1243 extern struct mdinfo *container_choose_spares(struct supertype *st,
1244 unsigned long long min_size,
1245 struct domainlist *domlist,
1246 char *spare_group,
1247 const char *metadata, int get_one);
1248 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1249 extern int add_disk(int mdfd, struct supertype *st,
1250 struct mdinfo *sra, struct mdinfo *info);
1251 extern int remove_disk(int mdfd, struct supertype *st,
1252 struct mdinfo *sra, struct mdinfo *info);
1253 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1254 unsigned long long min_recovery_start(struct mdinfo *array);
1255
1256 extern char *human_size(long long bytes);
1257 extern char *human_size_brief(long long bytes, int prefix);
1258 extern void print_r10_layout(int layout);
1259
1260 #define NoMdDev (1<<23)
1261 extern int find_free_devnum(int use_partitions);
1262
1263 extern void put_md_name(char *name);
1264 extern char *get_md_name(int dev);
1265
1266 extern char DefaultConfFile[];
1267
1268 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1269 char *chosen);
1270 /* values for 'trustworthy' */
1271 #define LOCAL 1
1272 #define LOCAL_ANY 10
1273 #define FOREIGN 2
1274 #define METADATA 3
1275 extern int open_mddev(char *dev, int report_errors);
1276 extern int open_container(int fd);
1277 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1278 extern int is_subarray_active(char *subarray, char *devname);
1279 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1280 extern struct superswitch *version_to_superswitch(char *vers);
1281
1282 extern int mdmon_running(int devnum);
1283 extern int mdmon_pid(int devnum);
1284 extern int check_env(char *name);
1285 extern __u32 random32(void);
1286 extern int start_mdmon(int devnum);
1287
1288 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1289 struct supertype *st, unsigned long stripes,
1290 int *fds, unsigned long long *offsets,
1291 int dests, int *destfd, unsigned long long *destoffsets);
1292 void abort_reshape(struct mdinfo *sra);
1293
1294 extern char *devnum2devname(int num);
1295 extern void fmt_devname(char *name, int num);
1296 extern int devname2devnum(char *name);
1297 extern int stat2devnum(struct stat *st);
1298 extern int fd2devnum(int fd);
1299
1300 static inline int dev2major(int d)
1301 {
1302 if (d >= 0)
1303 return MD_MAJOR;
1304 else
1305 return get_mdp_major();
1306 }
1307
1308 static inline int dev2minor(int d)
1309 {
1310 if (d >= 0)
1311 return d;
1312 return (-1-d) << MdpMinorShift;
1313 }
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
1475 extern int __offroot;