]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Grow/raid10: support reducing the devices in a RAID10.
[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 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 struct supertype *supertype;
261 };
262
263 #define Name "mdadm"
264
265 enum mode {
266 ASSEMBLE=1,
267 BUILD,
268 CREATE,
269 MANAGE,
270 MISC,
271 MONITOR,
272 GROW,
273 INCREMENTAL,
274 AUTODETECT,
275 mode_count
276 };
277
278 extern char short_options[];
279 extern char short_bitmap_options[];
280 extern char short_bitmap_auto_options[];
281 extern struct option long_options[];
282 extern char Version[], Usage[], Help[], OptionHelp[],
283 *mode_help[],
284 Help_create[], Help_build[], Help_assemble[], Help_grow[],
285 Help_incr[],
286 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
287
288 /* for option that don't have short equivilents, we assign arbitrary
289 * numbers later than any 'short' character option.
290 */
291 enum special_options {
292 AssumeClean = 300,
293 BitmapChunk,
294 WriteBehind,
295 ReAdd,
296 NoDegraded,
297 Sparc22,
298 BackupFile,
299 HomeHost,
300 AutoHomeHost,
301 Symlinks,
302 AutoDetect,
303 Waitclean,
304 DetailPlatform,
305 KillSubarray,
306 UpdateSubarray,
307 IncrementalPath,
308 NoSharing,
309 HelpOptions,
310 Brief,
311 ManageOpt,
312 Add,
313 Remove,
314 Fail,
315 MiscOpt,
316 WaitOpt,
317 ConfigFile,
318 ChunkSize,
319 WriteMostly,
320 Layout,
321 Auto,
322 Force,
323 SuperMinor,
324 EMail,
325 ProgramOpt,
326 Increment,
327 Fork,
328 Bitmap,
329 RebuildMapOpt,
330 InvalidBackup,
331 UdevRules,
332 FreezeReshape,
333 Continue,
334 OffRootOpt,
335 Prefer,
336 KillOpt,
337 DataOffset,
338 };
339
340 enum prefix_standard {
341 JEDEC,
342 IEC
343 };
344
345 /* structures read from config file */
346 /* List of mddevice names and identifiers
347 * Identifiers can be:
348 * uuid=128-hex-uuid
349 * super-minor=decimal-minor-number-from-superblock
350 * devices=comma,separated,list,of,device,names,with,wildcards
351 *
352 * If multiple fields are present, the intersection of all matching
353 * devices is considered
354 */
355 #define UnSet (0xfffe)
356 struct mddev_ident {
357 char *devname;
358
359 int uuid_set;
360 int uuid[4];
361 char name[33];
362
363 int super_minor;
364
365 char *devices; /* comma separated list of device
366 * names with wild cards
367 */
368 int level;
369 int raid_disks;
370 int spare_disks;
371 struct supertype *st;
372 int autof; /* 1 for normal, 2 for partitioned */
373 char *spare_group;
374 char *bitmap_file;
375 int bitmap_fd;
376
377 char *container; /* /dev/whatever name of container, or
378 * uuid of container. You would expect
379 * this to be the 'devname' or UUID
380 * of some other entry.
381 */
382 char *member; /* subarray within a container */
383
384 struct mddev_ident *next;
385 union {
386 /* fields needed by different users of this structure */
387 int assembled; /* set when assembly succeeds */
388 };
389 };
390
391 struct context {
392 int readonly;
393 int runstop;
394 int verbose;
395 int brief;
396 int force;
397 char *homehost;
398 int require_homehost;
399 char *prefer;
400 int export;
401 int test;
402 char *subarray;
403 char *update;
404 int scan;
405 int SparcAdjust;
406 int autof;
407 int delay;
408 int freeze_reshape;
409 char *backup_file;
410 int invalid_backup;
411 };
412
413 struct shape {
414 int raiddisks;
415 int sparedisks;
416 int level;
417 int layout;
418 char *layout_str;
419 int chunk;
420 int bitmap_chunk;
421 char *bitmap_file;
422 int assume_clean;
423 int write_behind;
424 unsigned long long size;
425 };
426
427 /* List of device names - wildcards expanded */
428 struct mddev_dev {
429 char *devname;
430 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
431 * 'A' for re_add.
432 * Not set for names read from .config
433 */
434 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
435 char used; /* set when used */
436 struct mddev_dev *next;
437 };
438
439 typedef struct mapping {
440 char *name;
441 int num;
442 } mapping_t;
443
444
445 struct mdstat_ent {
446 char *dev;
447 int devnum;
448 int active;
449 char *level;
450 char *pattern; /* U or up, _ for down */
451 int percent; /* -1 if no resync */
452 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
453 int devcnt;
454 int raid_disks;
455 char * metadata_version;
456 struct dev_member {
457 char *name;
458 struct dev_member *next;
459 } *members;
460 struct mdstat_ent *next;
461 };
462
463 extern struct mdstat_ent *mdstat_read(int hold, int start);
464 extern void free_mdstat(struct mdstat_ent *ms);
465 extern void mdstat_wait(int seconds);
466 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
467 extern int mddev_busy(int devnum);
468 extern struct mdstat_ent *mdstat_by_component(char *name);
469 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container);
470
471 struct map_ent {
472 struct map_ent *next;
473 int devnum;
474 char metadata[20];
475 int uuid[4];
476 int bad;
477 char *path;
478 };
479 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
480 int uuid[4], char *path);
481 extern void map_remove(struct map_ent **map, int devnum);
482 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
483 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
484 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
485 extern void map_read(struct map_ent **melp);
486 extern int map_write(struct map_ent *mel);
487 extern void map_delete(struct map_ent **mapp, int devnum);
488 extern void map_free(struct map_ent *map);
489 extern void map_add(struct map_ent **melp,
490 int devnum, char *metadata, int uuid[4], char *path);
491 extern int map_lock(struct map_ent **melp);
492 extern void map_unlock(struct map_ent **melp);
493 extern void map_fork(void);
494
495 /* various details can be requested */
496 enum sysfs_read_flags {
497 GET_LEVEL = (1 << 0),
498 GET_LAYOUT = (1 << 1),
499 GET_COMPONENT = (1 << 2),
500 GET_CHUNK = (1 << 3),
501 GET_CACHE = (1 << 4),
502 GET_MISMATCH = (1 << 5),
503 GET_VERSION = (1 << 6),
504 GET_DISKS = (1 << 7),
505 GET_DEGRADED = (1 << 8),
506 GET_SAFEMODE = (1 << 9),
507 GET_BITMAP_LOCATION = (1 << 10),
508
509 GET_DEVS = (1 << 20), /* gets role, major, minor */
510 GET_OFFSET = (1 << 21),
511 GET_SIZE = (1 << 22),
512 GET_STATE = (1 << 23),
513 GET_ERROR = (1 << 24),
514 };
515
516 /* If fd >= 0, get the array it is open on,
517 * else use devnum. >=0 -> major9. <0.....
518 */
519 extern int sysfs_open(int devnum, char *devname, char *attr);
520 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
521 extern void sysfs_free(struct mdinfo *sra);
522 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
523 extern int sysfs_attr_match(const char *attr, const char *str);
524 extern int sysfs_match_word(const char *word, char **list);
525 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
526 char *name, char *val);
527 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
528 char *name, unsigned long long val);
529 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
530 char *name, long long val);
531 extern int sysfs_uevent(struct mdinfo *sra, char *event);
532 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
533 char *name);
534 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
535 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
536 char *name, unsigned long long *val);
537 extern int sysfs_fd_get_str(int fd, char *val, int size);
538 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
539 char *name);
540 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
541 char *name, char *val, int size);
542 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
543 extern int sysfs_set_array(struct mdinfo *info, int vers);
544 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
545 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
546 extern int sysfs_unique_holder(int devnum, long rdev);
547 extern int sysfs_freeze_array(struct mdinfo *sra);
548 extern int load_sys(char *path, char *buf);
549 extern int reshape_prepare_fdlist(char *devname,
550 struct mdinfo *sra,
551 int raid_disks,
552 int nrdisks,
553 unsigned long blocks,
554 char *backup_file,
555 int *fdlist,
556 unsigned long long *offsets);
557 extern void reshape_free_fdlist(int *fdlist,
558 unsigned long long *offsets,
559 int size);
560 extern int reshape_open_backup_file(char *backup,
561 int fd,
562 char *devname,
563 long blocks,
564 int *fdlist,
565 unsigned long long *offsets,
566 int restart);
567 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
568 unsigned int ndata, unsigned int odata);
569
570 extern int save_stripes(int *source, unsigned long long *offsets,
571 int raid_disks, int chunk_size, int level, int layout,
572 int nwrites, int *dest,
573 unsigned long long start, unsigned long long length,
574 char *buf);
575 extern int restore_stripes(int *dest, unsigned long long *offsets,
576 int raid_disks, int chunk_size, int level, int layout,
577 int source, unsigned long long read_offset,
578 unsigned long long start, unsigned long long length,
579 char *src_buf);
580
581 #ifndef Sendmail
582 #define Sendmail "/usr/lib/sendmail -t"
583 #endif
584
585 #define SYSLOG_FACILITY LOG_DAEMON
586
587 extern char *map_num(mapping_t *map, int num);
588 extern int map_name(mapping_t *map, char *name);
589 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
590
591 extern char *map_dev_preferred(int major, int minor, int create,
592 char *prefer);
593 static inline char *map_dev(int major, int minor, int create)
594 {
595 return map_dev_preferred(major, minor, create, NULL);
596 }
597
598 struct active_array;
599 struct metadata_update;
600
601
602 /* 'struct reshape' records the intermediate states of
603 * a general reshape.
604 * The starting geometry is converted to the 'before' geometry
605 * by at most an atomic level change. They could be the same.
606 * Similarly the 'after' geometry is converted to the final
607 * geometry by at most a level change.
608 * Note that 'before' and 'after' must have the same level.
609 * 'blocks' is the minimum number of sectors for a reshape unit.
610 * This will be a multiple of the stripe size in each of the
611 * 'before' and 'after' geometries.
612 * If 'blocks' is 0, no restriping is necessary.
613 */
614 struct reshape {
615 int level;
616 int parity; /* number of parity blocks/devices */
617 struct {
618 int layout;
619 int data_disks;
620 } before, after;
621 unsigned long long backup_blocks;
622 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
623 unsigned long long new_size; /* New size of array in sectors */
624 };
625
626 /* A superswitch provides entry point the a metadata handler.
627 *
628 * The super_switch primarily operates on some "metadata" that
629 * is accessed via the 'supertype'.
630 * This metadata has one of three possible sources.
631 * 1/ It is read from a single device. In this case it may not completely
632 * describe the array or arrays as some information might be on other
633 * devices.
634 * 2/ It is read from all devices in a container. In this case all
635 * information is present.
636 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
637 * be complete once enough ->add_to_super calls have completed.
638 *
639 * When creating an array inside a container, the metadata will be
640 * formed by a combination of 2 and 3. The metadata or the array is read,
641 * then new information is added.
642 *
643 * The metadata must sometimes have a concept of a 'current' array
644 * and a 'current' device.
645 * The 'current' array is set by init_super to be the newly created array,
646 * or is set by super_by_fd when it finds it is looking at an array inside
647 * a container.
648 *
649 * The 'current' device is either the device that the metadata was read from
650 * in case 1, or the last device added by add_to_super in case 3.
651 * Case 2 does not identify a 'current' device.
652 */
653 extern struct superswitch {
654
655 /* Used to report details of metadata read from a component
656 * device. ->load_super has been called.
657 */
658 void (*examine_super)(struct supertype *st, char *homehost);
659 void (*brief_examine_super)(struct supertype *st, int verbose);
660 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
661 void (*export_examine_super)(struct supertype *st);
662
663 /* Used to report details of an active array.
664 * ->load_super was possibly given a 'component' string.
665 */
666 void (*detail_super)(struct supertype *st, char *homehost);
667 void (*brief_detail_super)(struct supertype *st);
668 void (*export_detail_super)(struct supertype *st);
669
670 /* Optional: platform hardware / firmware details */
671 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
672 int (*export_detail_platform)(int verbose, char *controller_path);
673
674 /* Used:
675 * to get uuid to storing in bitmap metadata
676 * and 'reshape' backup-data metadata
677 * To see if a device is being re-added to an array it was part of.
678 */
679 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
680
681 /* Extract generic details from metadata. This could be details about
682 * the container, or about an individual array within the container.
683 * The determination is made either by:
684 * load_super being given a 'component' string.
685 * validate_geometry determining what to create.
686 * The info includes both array information and device information.
687 * The particular device should be:
688 * The last device added by add_to_super
689 * The device the metadata was loaded from by load_super
690 * If 'map' is present, then it is an array raid_disks long
691 * (raid_disk must already be set and correct) and it is filled
692 * with 1 for slots that are thought to be active and 0 for slots which
693 * appear to be failed/missing.
694 * *info is zeroed out before data is added.
695 */
696 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
697 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
698 /* Check if the given metadata is flagged as belonging to "this"
699 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
700 */
701 int (*match_home)(struct supertype *st, char *homehost);
702
703 /* Make one of several generic modifications to metadata
704 * prior to assembly (or other times).
705 * sparc2.2 - first bug in early 0.90 metadata
706 * super-minor - change name of 0.90 metadata
707 * summaries - 'correct' any redundant data
708 * resync - mark array as dirty to trigger a resync.
709 * uuid - set new uuid - only 0.90 or 1.x
710 * name - change the name of the array (where supported)
711 * homehost - change which host this array is tied to.
712 * devicesize - If metadata is at start of device, change recorded
713 * device size to match actual device size
714 * byteorder - swap bytes for 0.90 metadata
715 *
716 * force-one - mark that device as uptodate, not old or failed.
717 * force-array - mark array as clean if it would not otherwise
718 * assemble
719 * assemble - not sure how this is different from force-one...
720 * linear-grow-new - add a new device to a linear array, but don't
721 * change the size: so superblock still matches
722 * linear-grow-update - now change the size of the array.
723 * writemostly - set the WriteMostly1 bit in the superblock devflags
724 * readwrite - clear the WriteMostly1 bit in the superblock devflags
725 */
726 int (*update_super)(struct supertype *st, struct mdinfo *info,
727 char *update,
728 char *devname, int verbose,
729 int uuid_set, char *homehost);
730
731 /* Create new metadata for new array as described. This could
732 * be a new container, or an array in a pre-existing container.
733 * Also used to zero metadata prior to writing it to invalidate old
734 * metadata.
735 */
736 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
737 unsigned long long size, char *name,
738 char *homehost, int *uuid,
739 unsigned long long data_offset);
740
741 /* update the metadata to include new device, either at create or
742 * when hot-adding a spare.
743 */
744 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
745 int fd, char *devname);
746 /* update the metadata to delete a device,
747 * when hot-removing.
748 */
749 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
750
751 /* Write metadata to one device when fixing problems or adding
752 * a new device.
753 */
754 int (*store_super)(struct supertype *st, int fd);
755
756 /* Write all metadata for this array.
757 */
758 int (*write_init_super)(struct supertype *st);
759 int (*compare_super)(struct supertype *st, struct supertype *tst);
760 int (*load_super)(struct supertype *st, int fd, char *devname);
761 int (*load_container)(struct supertype *st, int fd, char *devname);
762 struct supertype * (*match_metadata_desc)(char *arg);
763 __u64 (*avail_size)(struct supertype *st, __u64 size,
764 unsigned long long data_offset);
765 unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
766 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
767 int delay, int write_behind,
768 unsigned long long size, int may_change, int major);
769 void (*locate_bitmap)(struct supertype *st, int fd);
770 int (*write_bitmap)(struct supertype *st, int fd);
771 void (*free_super)(struct supertype *st);
772
773 /* validate_geometry is called with an st returned by
774 * match_metadata_desc.
775 * It should check that the geometry described in compatible with
776 * the metadata type. It will be called repeatedly as devices
777 * added to validate changing size and new devices. If there are
778 * inter-device dependencies, it should record sufficient details
779 * so these can be validated.
780 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
781 * Return value is:
782 * 1: everything is OK
783 * 0: not OK for some reason - if 'verbose', then error was reported.
784 * -1: st->sb was NULL, 'subdev' is a member of a container of this
785 * types, but array is not acceptable for some reason
786 * message was reported even if verbose is 0.
787 */
788 int (*validate_geometry)(struct supertype *st, int level, int layout,
789 int raiddisks,
790 int *chunk, unsigned long long size,
791 unsigned long long data_offset,
792 char *subdev, unsigned long long *freesize,
793 int verbose);
794
795 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
796 /* query the supertype for default geometry */
797 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
798 /* Permit subarray's to be deleted from inactive containers */
799 int (*kill_subarray)(struct supertype *st); /* optional */
800 /* Permit subarray's to be modified */
801 int (*update_subarray)(struct supertype *st, char *subarray,
802 char *update, struct mddev_ident *ident); /* optional */
803 /* Check if reshape is supported for this external format.
804 * st is obtained from super_by_fd() where st->subarray[0] is
805 * initialized to indicate if reshape is being performed at the
806 * container or subarray level
807 */
808 #define APPLY_METADATA_CHANGES 1
809 #define ROLLBACK_METADATA_CHANGES 0
810
811 int (*reshape_super)(struct supertype *st,
812 unsigned long long size, int level,
813 int layout, int chunksize, int raid_disks,
814 int delta_disks, char *backup, char *dev,
815 int direction,
816 int verbose); /* optional */
817 int (*manage_reshape)( /* optional */
818 int afd, struct mdinfo *sra, struct reshape *reshape,
819 struct supertype *st, unsigned long blocks,
820 int *fds, unsigned long long *offsets,
821 int dests, int *destfd, unsigned long long *destoffsets);
822
823 /* for mdmon */
824 int (*open_new)(struct supertype *c, struct active_array *a,
825 char *inst);
826
827 /* Tell the metadata handler the current state of the array.
828 * This covers whether it is known to be consistent (no pending writes)
829 * and how far along a resync is known to have progressed
830 * (in a->resync_start).
831 * resync status is really irrelevant if the array is not consistent,
832 * but some metadata (DDF!) have a place to record the distinction.
833 * If 'consistent' is '2', then the array can mark it dirty if a
834 * resync/recovery/whatever is required, or leave it clean if not.
835 * Return value is 0 dirty (not consistent) and 1 if clean.
836 * it is only really important if consistent is passed in as '2'.
837 */
838 int (*set_array_state)(struct active_array *a, int consistent);
839
840 /* When the state of a device might have changed, we call set_disk to
841 * tell the metadata what the current state is.
842 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
843 * transitions.
844 * set_disk might be called when the state of the particular disk has
845 * not in fact changed.
846 */
847 void (*set_disk)(struct active_array *a, int n, int state);
848 void (*sync_metadata)(struct supertype *st);
849 void (*process_update)(struct supertype *st,
850 struct metadata_update *update);
851 void (*prepare_update)(struct supertype *st,
852 struct metadata_update *update);
853
854 /* activate_spare will check if the array is degraded and, if it
855 * is, try to find some spare space in the container.
856 * On success, it add appropriate updates (For process_update) to
857 * to the 'updates' list and returns a list of 'mdinfo' identifying
858 * the device, or devices as there might be multiple missing
859 * devices and multiple spares available.
860 */
861 struct mdinfo *(*activate_spare)(struct active_array *a,
862 struct metadata_update **updates);
863 /*
864 * Return statically allocated string that represents metadata specific
865 * controller domain of the disk. The domain is used in disk domain
866 * matching functions. Disks belong to the same domain if the they have
867 * the same domain from mdadm.conf and belong the same metadata domain.
868 * Returning NULL or not providing this handler means that metadata
869 * does not distinguish the differences between disks that belong to
870 * different controllers. They are in the domain specified by
871 * configuration file (mdadm.conf).
872 * In case when the metadata has the notion of domains based on disk
873 * it shall return NULL for disks that do not belong to the controller
874 * the supported domains. Such disks will form another domain and won't
875 * be mixed with supported ones.
876 */
877 const char *(*get_disk_controller_domain)(const char *path);
878
879 /* for external backup area */
880 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
881
882 int swapuuid; /* true if uuid is bigending rather than hostendian */
883 int external;
884 const char *name; /* canonical metadata name */
885 } *superlist[];
886
887 extern struct superswitch super0, super1;
888 extern struct superswitch super_imsm, super_ddf;
889 extern struct superswitch mbr, gpt;
890
891 struct metadata_update {
892 int len;
893 char *buf;
894 void *space; /* allocated space that monitor will use */
895 void **space_list; /* list of allocated spaces that monitor can
896 * use or that it returned.
897 */
898 struct metadata_update *next;
899 };
900
901 /* A supertype holds a particular collection of metadata.
902 * It identifies the metadata type by the superswitch, and the particular
903 * sub-version of that metadata type.
904 * metadata read in or created is stored in 'sb' and 'info'.
905 * There are also fields used by mdmon to track containers.
906 *
907 * A supertype may refer to:
908 * Just an array, possibly in a container
909 * A container, not identifying any particular array
910 * Info read from just one device, not yet fully describing the array/container.
911 *
912 *
913 * A supertype is created by:
914 * super_by_fd
915 * guess_super
916 * dup_super
917 */
918 struct supertype {
919 struct superswitch *ss;
920 int minor_version;
921 int max_devs;
922 int container_dev; /* devnum of container */
923 void *sb;
924 void *info;
925 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
926 HW/firmware related incompatability to load metadata.
927 Used when examining metadata to display content of disk
928 when user has no hw/firmare compatible system.
929 */
930 struct metadata_update *updates;
931 struct metadata_update **update_tail;
932
933 /* extra stuff used by mdmon */
934 struct active_array *arrays;
935 int sock; /* listen to external programs */
936 int devnum;
937 char *devname; /* e.g. md0. This appears in metadata_verison:
938 * external:/md0/12
939 */
940 int devcnt;
941 int retry_soon;
942
943 struct mdinfo *devs;
944
945 };
946
947 extern struct supertype *super_by_fd(int fd, char **subarray);
948 enum guess_types { guess_any, guess_array, guess_partitions };
949 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
950 static inline struct supertype *guess_super(int fd) {
951 return guess_super_type(fd, guess_any);
952 }
953 extern struct supertype *dup_super(struct supertype *st);
954 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
955 extern int must_be_container(int fd);
956 extern int dev_size_from_id(dev_t id, unsigned long long *size);
957 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
958 mdu_disk_info_t *disk);
959 void wait_for(char *dev, int fd);
960
961 /*
962 * Data structures for policy management.
963 * Each device can have a policy structure that lists
964 * various name/value pairs each possibly with a metadata associated.
965 * The policy list is sorted by name/value/metadata
966 */
967 struct dev_policy {
968 struct dev_policy *next;
969 char *name; /* None of these strings are allocated. They are
970 * all just references to strings which are known
971 * to exist elsewhere.
972 * name and metadata can be compared by address equality.
973 */
974 const char *metadata;
975 const char *value;
976 };
977
978 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
979
980 /* iterate over the sublist starting at list, having the same
981 * 'name' as 'list', and matching the given metadata (Where
982 * NULL matches anything
983 */
984 #define pol_for_each(item, list, _metadata) \
985 for (item = list; \
986 item && item->name == list->name; \
987 item = item->next) \
988 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
989 ; else
990
991 /*
992 * policy records read from mdadm are largely just name-value pairs.
993 * The names are constants, not strdupped
994 */
995 struct pol_rule {
996 struct pol_rule *next;
997 char *type; /* rule_policy or rule_part */
998 struct rule {
999 struct rule *next;
1000 char *name;
1001 char *value;
1002 char *dups; /* duplicates of 'value' with a partNN appended */
1003 } *rule;
1004 };
1005
1006 extern char rule_policy[], rule_part[];
1007 extern char rule_path[], rule_type[];
1008 extern char type_part[], type_disk[];
1009
1010 extern void policyline(char *line, char *type);
1011 extern void policy_add(char *type, ...);
1012 extern void policy_free(void);
1013
1014 extern struct dev_policy *path_policy(char *path, char *type);
1015 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1016 extern struct dev_policy *devnum_policy(int dev);
1017 extern void dev_policy_free(struct dev_policy *p);
1018
1019 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1020 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1021 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1022
1023 enum policy_action {
1024 act_default,
1025 act_include,
1026 act_re_add,
1027 act_spare, /* This only applies to bare devices */
1028 act_spare_same_slot, /* this allows non-bare devices,
1029 * but only if recent removal */
1030 act_force_spare, /* this allow non-bare devices in any case */
1031 act_err
1032 };
1033
1034 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1035 enum policy_action want);
1036 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1037 enum policy_action want);
1038
1039 struct domainlist {
1040 struct domainlist *next;
1041 const char *dom;
1042 };
1043
1044 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1045 const char *metadata);
1046 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1047 const char *metadata);
1048 extern void domainlist_add_dev(struct domainlist **dom, int devnum,
1049 const char *metadata);
1050 extern void domain_free(struct domainlist *dl);
1051 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1052 const char *metadata);
1053 void domain_add(struct domainlist **domp, char *domain);
1054
1055 extern void policy_save_path(char *id_path, struct map_ent *array);
1056 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1057
1058 #if __GNUC__ < 3
1059 struct stat64;
1060 #endif
1061
1062 #define HAVE_NFTW we assume
1063 #define HAVE_FTW
1064
1065 #ifdef __UCLIBC__
1066 # include <features.h>
1067 # ifndef __UCLIBC_HAS_LFS__
1068 # define lseek64 lseek
1069 # endif
1070 # ifndef __UCLIBC_HAS_FTW__
1071 # undef HAVE_FTW
1072 # undef HAVE_NFTW
1073 # endif
1074 #endif
1075
1076 #ifdef __dietlibc__
1077 # undef HAVE_NFTW
1078 #endif
1079
1080 #if defined(__KLIBC__)
1081 # undef HAVE_NFTW
1082 # undef HAVE_FTW
1083 #endif
1084
1085 #ifndef HAVE_NFTW
1086 # define FTW_PHYS 1
1087 # ifndef HAVE_FTW
1088 struct FTW {};
1089 # endif
1090 #endif
1091
1092 #ifdef HAVE_FTW
1093 # include <ftw.h>
1094 #endif
1095
1096 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1097
1098
1099 extern int Manage_ro(char *devname, int fd, int readonly);
1100 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet,
1101 int will_retry);
1102 extern int Manage_subdevs(char *devname, int fd,
1103 struct mddev_dev *devlist, int verbose, int test,
1104 char *update, int force);
1105 extern int autodetect(void);
1106 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1107 extern int Grow_addbitmap(char *devname, int fd,
1108 struct context *c, struct shape *s);
1109 extern int Grow_reshape(char *devname, int fd,
1110 struct mddev_dev *devlist,
1111 unsigned long long data_offset,
1112 struct context *c, struct shape *s);
1113 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1114 int *fdlist, int cnt, char *backup_file, int verbose);
1115 extern int Grow_continue(int mdfd, struct supertype *st,
1116 struct mdinfo *info, char *backup_file,
1117 int freeze_reshape);
1118
1119 extern int restore_backup(struct supertype *st,
1120 struct mdinfo *content,
1121 int working_disks,
1122 int spares,
1123 char *backup_file,
1124 int verbose);
1125 extern int Grow_continue_command(char *devname, int fd,
1126 char *backup_file, int verbose);
1127
1128 extern int Assemble(struct supertype *st, char *mddev,
1129 struct mddev_ident *ident,
1130 struct mddev_dev *devlist,
1131 struct context *c);
1132
1133 extern int Build(char *mddev, struct mddev_dev *devlist,
1134 struct shape *s, struct context *c);
1135
1136 extern int Create(struct supertype *st, char *mddev,
1137 char *name, int *uuid,
1138 int subdevs, struct mddev_dev *devlist,
1139 struct shape *s,
1140 struct context *c,
1141 unsigned long long data_offset);
1142
1143 extern int Detail(char *dev, struct context *c);
1144 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1145 extern int Query(char *dev);
1146 extern int Examine(struct mddev_dev *devlist, struct context *c,
1147 struct supertype *forcest);
1148 extern int Monitor(struct mddev_dev *devlist,
1149 char *mailaddr, char *alert_cmd,
1150 struct context *c,
1151 int daemonise, int oneshot,
1152 int dosyslog, char *pidfile, int increments,
1153 int share);
1154
1155 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1156 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1157 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1158 extern int Wait(char *dev);
1159 extern int WaitClean(char *dev, int sock, int verbose);
1160
1161 extern int Incremental(char *devname, struct context *c,
1162 struct supertype *st);
1163 extern void RebuildMap(void);
1164 extern int IncrementalScan(int verbose);
1165 extern int IncrementalRemove(char *devname, char *path, int verbose);
1166 extern int CreateBitmap(char *filename, int force, char uuid[16],
1167 unsigned long chunksize, unsigned long daemon_sleep,
1168 unsigned long write_behind,
1169 unsigned long long array_size,
1170 int major);
1171 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1172 extern int Write_rules(char *rule_name);
1173 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1174 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1175
1176 extern int md_get_version(int fd);
1177 extern int get_linux_version(void);
1178 extern int mdadm_version(char *version);
1179 extern unsigned long long parse_size(char *size);
1180 extern int parse_uuid(char *str, int uuid[4]);
1181 extern int parse_layout_10(char *layout);
1182 extern int parse_layout_faulty(char *layout);
1183 extern long parse_num(char *num);
1184 extern int check_ext2(int fd, char *name);
1185 extern int check_reiser(int fd, char *name);
1186 extern int check_raid(int fd, char *name);
1187 extern int check_partitions(int fd, char *dname,
1188 unsigned long long freesize,
1189 unsigned long long size);
1190
1191 extern int get_mdp_major(void);
1192 extern int dev_open(char *dev, int flags);
1193 extern int open_dev(int devnum);
1194 extern int open_dev_flags(int devnum, int flags);
1195 extern int open_dev_excl(int devnum);
1196 extern int is_standard(char *dev, int *nump);
1197 extern int same_dev(char *one, char *two);
1198 extern int compare_paths (char* path1,char* path2);
1199
1200 extern int parse_auto(char *str, char *msg, int config);
1201 extern struct mddev_ident *conf_get_ident(char *dev);
1202 extern struct mddev_dev *conf_get_devs(void);
1203 extern int conf_test_dev(char *devname);
1204 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1205 extern struct createinfo *conf_get_create_info(void);
1206 extern void set_conffile(char *file);
1207 extern char *conf_get_mailaddr(void);
1208 extern char *conf_get_mailfrom(void);
1209 extern char *conf_get_program(void);
1210 extern char *conf_get_homehost(int *require_homehostp);
1211 extern char *conf_line(FILE *file);
1212 extern char *conf_word(FILE *file, int allow_key);
1213 extern void print_quoted(char *str);
1214 extern void print_escape(char *str);
1215 extern int conf_name_is_free(char *name);
1216 extern int conf_verify_devnames(struct mddev_ident *array_list);
1217 extern int devname_matches(char *name, char *match);
1218 extern struct mddev_ident *conf_match(struct supertype *st,
1219 struct mdinfo *info,
1220 char *devname,
1221 int verbose, int *rvp);
1222 extern int experimental(void);
1223
1224 extern void free_line(char *line);
1225 extern int match_oneof(char *devices, char *devname);
1226 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1227 extern const int uuid_zero[4];
1228 extern int same_uuid(int a[4], int b[4], int swapuuid);
1229 extern void copy_uuid(void *a, int b[4], int swapuuid);
1230 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1231 extern char *fname_from_uuid(struct supertype *st,
1232 struct mdinfo *info, char *buf, char sep);
1233 extern unsigned long calc_csum(void *super, int bytes);
1234 extern int enough(int level, int raid_disks, int layout, int clean,
1235 char *avail);
1236 extern int enough_fd(int fd);
1237 extern int ask(char *mesg);
1238 extern unsigned long long get_component_size(int fd);
1239 extern void remove_partitions(int fd);
1240 extern int test_partition(int fd);
1241 extern int test_partition_from_id(dev_t id);
1242 extern int get_data_disks(int level, int layout, int raid_disks);
1243 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1244 int chunksize, unsigned long long devsize);
1245 extern int flush_metadata_updates(struct supertype *st);
1246 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1247 extern int assemble_container_content(struct supertype *st, int mdfd,
1248 struct mdinfo *content,
1249 struct context *c,
1250 char *chosen_name);
1251 extern struct mdinfo *container_choose_spares(struct supertype *st,
1252 unsigned long long min_size,
1253 struct domainlist *domlist,
1254 char *spare_group,
1255 const char *metadata, int get_one);
1256 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1257 extern int add_disk(int mdfd, struct supertype *st,
1258 struct mdinfo *sra, struct mdinfo *info);
1259 extern int remove_disk(int mdfd, struct supertype *st,
1260 struct mdinfo *sra, struct mdinfo *info);
1261 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1262 unsigned long long min_recovery_start(struct mdinfo *array);
1263
1264 extern char *human_size(long long bytes);
1265 extern char *human_size_brief(long long bytes, int prefix);
1266 extern void print_r10_layout(int layout);
1267
1268 #define NoMdDev (1<<23)
1269 extern int find_free_devnum(int use_partitions);
1270
1271 extern void put_md_name(char *name);
1272 extern char *get_md_name(int dev);
1273
1274 extern char DefaultConfFile[];
1275
1276 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1277 char *chosen);
1278 /* values for 'trustworthy' */
1279 #define LOCAL 1
1280 #define LOCAL_ANY 10
1281 #define FOREIGN 2
1282 #define METADATA 3
1283 extern int open_mddev(char *dev, int report_errors);
1284 extern int open_container(int fd);
1285 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1286 extern int is_subarray_active(char *subarray, char *devname);
1287 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1288 extern struct superswitch *version_to_superswitch(char *vers);
1289
1290 extern int mdmon_running(int devnum);
1291 extern int mdmon_pid(int devnum);
1292 extern int check_env(char *name);
1293 extern __u32 random32(void);
1294 extern int start_mdmon(int devnum);
1295
1296 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1297 struct supertype *st, unsigned long stripes,
1298 int *fds, unsigned long long *offsets,
1299 int dests, int *destfd, unsigned long long *destoffsets);
1300 void abort_reshape(struct mdinfo *sra);
1301
1302 extern char *devnum2devname(int num);
1303 extern void fmt_devname(char *name, int num);
1304 extern int devname2devnum(char *name);
1305 extern int stat2devnum(struct stat *st);
1306 extern int fd2devnum(int fd);
1307
1308 static inline int dev2major(int d)
1309 {
1310 if (d >= 0)
1311 return MD_MAJOR;
1312 else
1313 return get_mdp_major();
1314 }
1315
1316 static inline int dev2minor(int d)
1317 {
1318 if (d >= 0)
1319 return d;
1320 return (-1-d) << MdpMinorShift;
1321 }
1322
1323 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1324 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1325 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1326 (ROUND_UP((unsigned long)(ptr), base)))
1327
1328 static inline int is_subarray(char *vers)
1329 {
1330 /* The version string for a 'subarray' (an array in a container)
1331 * is
1332 * /containername/componentname for normal read-write arrays
1333 * -containername/componentname for arrays which mdmon must not
1334 * reconfigure. They might be read-only
1335 * or might be undergoing reshape etc.
1336 * containername is e.g. md0, md_d1
1337 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1338 */
1339 return (*vers == '/' || *vers == '-');
1340 }
1341
1342 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1343 {
1344 return &ent->metadata_version[10+strlen(container)+1];
1345 }
1346
1347 #ifdef DEBUG
1348 #define dprintf(fmt, arg...) \
1349 fprintf(stderr, fmt, ##arg)
1350 #else
1351 #define dprintf(fmt, arg...) \
1352 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1353 #endif
1354 #include <assert.h>
1355 #include <stdarg.h>
1356 static inline int xasprintf(char **strp, const char *fmt, ...) {
1357 va_list ap;
1358 int ret;
1359 va_start(ap, fmt);
1360 ret = vasprintf(strp, fmt, ap);
1361 va_end(ap);
1362 assert(ret >= 0);
1363 return ret;
1364 }
1365
1366 #define pr_err(fmt ...) fprintf(stderr, Name ": " fmt)
1367 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1368
1369 void *xmalloc(size_t len);
1370 void *xrealloc(void *ptr, size_t len);
1371 void *xcalloc(size_t num, size_t size);
1372 char *xstrdup(const char *str);
1373
1374 #define LEVEL_MULTIPATH (-4)
1375 #define LEVEL_LINEAR (-1)
1376 #define LEVEL_FAULTY (-5)
1377
1378 /* kernel module doesn't know about these */
1379 #define LEVEL_CONTAINER (-100)
1380 #define LEVEL_UNSUPPORTED (-200)
1381
1382
1383 /* faulty stuff */
1384
1385 #define WriteTransient 0
1386 #define ReadTransient 1
1387 #define WritePersistent 2
1388 #define ReadPersistent 3
1389 #define WriteAll 4 /* doesn't go to device */
1390 #define ReadFixable 5
1391 #define Modes 6
1392
1393 #define ClearErrors 31
1394 #define ClearFaults 30
1395
1396 #define AllPersist 100 /* internal use only */
1397 #define NoPersist 101
1398
1399 #define ModeMask 0x1f
1400 #define ModeShift 5
1401
1402
1403 #ifdef __TINYC__
1404 #undef minor
1405 #undef major
1406 #undef makedev
1407 #define minor(x) ((x)&0xff)
1408 #define major(x) (((x)>>8)&0xff)
1409 #define makedev(M,m) (((M)<<8) | (m))
1410 #endif
1411
1412 /* for raid4/5/6 */
1413 #define ALGORITHM_LEFT_ASYMMETRIC 0
1414 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1415 #define ALGORITHM_LEFT_SYMMETRIC 2
1416 #define ALGORITHM_RIGHT_SYMMETRIC 3
1417
1418 /* Define non-rotating (raid4) algorithms. These allow
1419 * conversion of raid4 to raid5.
1420 */
1421 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1422 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1423
1424 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1425 * Firstly, the exact positioning of the parity block is slightly
1426 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1427 * of DDF.
1428 * Secondly, or order of datablocks over which the Q syndrome is computed
1429 * is different.
1430 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1431 * These layouts are from the DDFv1.2 spec.
1432 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1433 * leaves RLQ=3 as 'Vendor Specific'
1434 */
1435
1436 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1437 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1438 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1439
1440
1441 /* For every RAID5 algorithm we define a RAID6 algorithm
1442 * with exactly the same layout for data and parity, and
1443 * with the Q block always on the last device (N-1).
1444 * This allows trivial conversion from RAID5 to RAID6
1445 */
1446 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1447 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1448 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1449 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1450 #define ALGORITHM_PARITY_0_6 20
1451 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1452
1453 /* Define PATH_MAX in case we don't use glibc or standard library does
1454 * not have PATH_MAX defined. Assume max path length is 4K characters.
1455 */
1456 #ifndef PATH_MAX
1457 #define PATH_MAX 4096
1458 #endif
1459
1460 #define RESYNC_NONE -1
1461 #define RESYNC_DELAYED -2
1462 #define RESYNC_PENDING -3
1463 #define RESYNC_UNKNOWN -4
1464
1465 /* When using "GET_DISK_INFO" it isn't certain how high
1466 * we need to check. So we impose an absolute limit of
1467 * MAX_DISKS. This needs to be much more than the largest
1468 * number of devices any metadata can support. Currently
1469 * v1.x can support 1920
1470 */
1471 #define MAX_DISKS 4096
1472
1473 /* Sometimes the 'size' value passed needs to mean "Maximum".
1474 * In those cases with use MAX_SIZE
1475 */
1476 #define MAX_SIZE 1
1477
1478 /* We want to use unsigned numbers for sector counts, but need
1479 * a value for 'invalid'. Use '1'.
1480 */
1481 #define INVALID_SECTORS 1
1482
1483 extern int __offroot;