]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Improve partition table code.
[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_BITMAP_CHUNK 4096
68 #define DEFAULT_BITMAP_DELAY 5
69 #define DEFAULT_MAX_WRITE_BEHIND 256
70
71 /* VAR_RUN is where pid and socket files used for communicating
72 * with mdmon normally live. It should be /var/run, but if
73 * it is too hard to remount /var/run as read-only rather than
74 * unmounting it at shutdown time, then it should be
75 * redefined to some place that comfortably persists until
76 * final shutdown, possibly in /dev if that is a tmpfs.
77 * Note: VAR_RUN does not need to be writable at shutdown,
78 * only during boot when "mdmon --takeover" is run.
79 */
80 #ifndef VAR_RUN
81 #define VAR_RUN "/var/run/mdadm"
82 #endif /* VAR_RUN */
83 /* ALT_RUN should be somewhere that persists across the pivotroot
84 * from early boot to late boot.
85 * If you don't have /lib/init/rw you might want to use /dev/.something
86 */
87 #ifndef ALT_RUN
88 #define ALT_RUN "/lib/init/rw/mdadm"
89 #endif /* ALT_RUN */
90
91 #include "md_u.h"
92 #include "md_p.h"
93 #include "bitmap.h"
94 #include "msg.h"
95
96 #include <endian.h>
97 /* Redhat don't like to #include <asm/byteorder.h>, and
98 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
99 * and there is no standard conversion function so... */
100 /* And dietlibc doesn't think byteswap is ok, so.. */
101 /* #include <byteswap.h> */
102 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
103 ((x) & 0xff00U) >> 8)
104 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
105 ((x) & 0xff000000U) >> 24 | \
106 ((x) & 0x0000ff00U) << 8 | \
107 ((x) & 0x00ff0000U) >> 8)
108 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
109 ((x) & 0xff00000000000000ULL) >> 56 | \
110 ((x) & 0x000000000000ff00ULL) << 40 | \
111 ((x) & 0x00ff000000000000ULL) >> 40 | \
112 ((x) & 0x0000000000ff0000ULL) << 24 | \
113 ((x) & 0x0000ff0000000000ULL) >> 24 | \
114 ((x) & 0x00000000ff000000ULL) << 8 | \
115 ((x) & 0x000000ff00000000ULL) >> 8)
116
117 #if !defined(__KLIBC__)
118 #if BYTE_ORDER == LITTLE_ENDIAN
119 #define __cpu_to_le16(_x) (_x)
120 #define __cpu_to_le32(_x) (_x)
121 #define __cpu_to_le64(_x) (_x)
122 #define __le16_to_cpu(_x) (_x)
123 #define __le32_to_cpu(_x) (_x)
124 #define __le64_to_cpu(_x) (_x)
125
126 #define __cpu_to_be16(_x) bswap_16(_x)
127 #define __cpu_to_be32(_x) bswap_32(_x)
128 #define __cpu_to_be64(_x) bswap_64(_x)
129 #define __be16_to_cpu(_x) bswap_16(_x)
130 #define __be32_to_cpu(_x) bswap_32(_x)
131 #define __be64_to_cpu(_x) bswap_64(_x)
132 #elif BYTE_ORDER == BIG_ENDIAN
133 #define __cpu_to_le16(_x) bswap_16(_x)
134 #define __cpu_to_le32(_x) bswap_32(_x)
135 #define __cpu_to_le64(_x) bswap_64(_x)
136 #define __le16_to_cpu(_x) bswap_16(_x)
137 #define __le32_to_cpu(_x) bswap_32(_x)
138 #define __le64_to_cpu(_x) bswap_64(_x)
139
140 #define __cpu_to_be16(_x) (_x)
141 #define __cpu_to_be32(_x) (_x)
142 #define __cpu_to_be64(_x) (_x)
143 #define __be16_to_cpu(_x) (_x)
144 #define __be32_to_cpu(_x) (_x)
145 #define __be64_to_cpu(_x) (_x)
146 #else
147 # error "unknown endianness."
148 #endif
149 #endif /* __KLIBC__ */
150
151
152 /*
153 * min()/max()/clamp() macros that also do
154 * strict type-checking.. See the
155 * "unnecessary" pointer comparison.
156 */
157 #define min(x, y) ({ \
158 typeof(x) _min1 = (x); \
159 typeof(y) _min2 = (y); \
160 (void) (&_min1 == &_min2); \
161 _min1 < _min2 ? _min1 : _min2; })
162
163 #define max(x, y) ({ \
164 typeof(x) _max1 = (x); \
165 typeof(y) _max2 = (y); \
166 (void) (&_max1 == &_max2); \
167 _max1 > _max2 ? _max1 : _max2; })
168
169 /* general information that might be extracted from a superblock */
170 struct mdinfo {
171 mdu_array_info_t array;
172 mdu_disk_info_t disk;
173 __u64 events;
174 int uuid[4];
175 char name[33];
176 unsigned long long data_offset;
177 unsigned long long component_size; /* same as array.size, except in
178 * sectors and up to 64bits.
179 */
180 unsigned long long custom_array_size; /* size for non-default sized
181 * arrays (in sectors)
182 */
183 int reshape_active;
184 unsigned long long reshape_progress;
185 union {
186 unsigned long long resync_start; /* per-array resync position */
187 unsigned long long recovery_start; /* per-device rebuild position */
188 #define MaxSector (~0ULL) /* resync/recovery complete position */
189 };
190 unsigned long safe_mode_delay; /* ms delay to mark clean */
191 int new_level, delta_disks, new_layout, new_chunk;
192 int errors;
193 int cache_size; /* size of raid456 stripe cache*/
194 int mismatch_cnt;
195 char text_version[50];
196 void *update_private; /* for passing metadata-format
197 * specific update data
198 * between successive calls to
199 * update_super()
200 */
201
202 int container_member; /* for assembling external-metatdata arrays
203 * This is to be used internally by metadata
204 * handler only */
205
206 char sys_name[20];
207 struct mdinfo *devs;
208 struct mdinfo *next;
209
210 /* Device info for mdmon: */
211 int recovery_fd;
212 int state_fd;
213 #define DS_FAULTY 1
214 #define DS_INSYNC 2
215 #define DS_WRITE_MOSTLY 4
216 #define DS_SPARE 8
217 #define DS_BLOCKED 16
218 #define DS_REMOVE 1024
219 #define DS_UNBLOCK 2048
220 int prev_state, curr_state, next_state;
221
222 };
223
224 struct createinfo {
225 int uid;
226 int gid;
227 int autof;
228 int mode;
229 int symlinks;
230 struct supertype *supertype;
231 };
232
233 #define Name "mdadm"
234
235 enum mode {
236 ASSEMBLE=1,
237 BUILD,
238 CREATE,
239 MANAGE,
240 MISC,
241 MONITOR,
242 GROW,
243 INCREMENTAL,
244 AUTODETECT,
245 };
246
247 extern char short_options[];
248 extern char short_bitmap_options[];
249 extern char short_bitmap_auto_options[];
250 extern struct option long_options[];
251 extern char Version[], Usage[], Help[], OptionHelp[],
252 Help_create[], Help_build[], Help_assemble[], Help_grow[],
253 Help_incr[],
254 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
255
256 /* for option that don't have short equivilents, we assign arbitrary
257 * small numbers. '1' means an undecorated option, so we start at '2'.
258 */
259 enum special_options {
260 AssumeClean = 2,
261 BitmapChunk,
262 WriteBehind,
263 ReAdd,
264 NoDegraded,
265 Sparc22,
266 BackupFile,
267 HomeHost,
268 AutoHomeHost,
269 Symlinks,
270 AutoDetect,
271 Waitclean,
272 DetailPlatform,
273 };
274
275 /* structures read from config file */
276 /* List of mddevice names and identifiers
277 * Identifiers can be:
278 * uuid=128-hex-uuid
279 * super-minor=decimal-minor-number-from-superblock
280 * devices=comma,separated,list,of,device,names,with,wildcards
281 *
282 * If multiple fields are present, the intersection of all matching
283 * devices is considered
284 */
285 #define UnSet (0xfffe)
286 typedef struct mddev_ident_s {
287 char *devname;
288
289 int uuid_set;
290 int uuid[4];
291 char name[33];
292
293 unsigned int super_minor;
294
295 char *devices; /* comma separated list of device
296 * names with wild cards
297 */
298 int level;
299 unsigned int raid_disks;
300 unsigned int spare_disks;
301 struct supertype *st;
302 int autof; /* 1 for normal, 2 for partitioned */
303 char *spare_group;
304 char *bitmap_file;
305 int bitmap_fd;
306
307 char *container; /* /dev/whatever name of container, or
308 * uuid of container. You would expect
309 * this to be the 'devname' or UUID
310 * of some other entry.
311 */
312 char *member; /* subarray within a container */
313
314 struct mddev_ident_s *next;
315 union {
316 /* fields needed by different users of this structure */
317 int assembled; /* set when assembly succeeds */
318 };
319 } *mddev_ident_t;
320
321 /* List of device names - wildcards expanded */
322 typedef struct mddev_dev_s {
323 char *devname;
324 char disposition; /* 'a' for add, 'r' for remove, 'f' for fail.
325 * Not set for names read from .config
326 */
327 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
328 char re_add;
329 char used; /* set when used */
330 struct mdinfo *content; /* If devname is a container, this might list
331 * the remaining member arrays. */
332 struct mddev_dev_s *next;
333 } *mddev_dev_t;
334
335 typedef struct mapping {
336 char *name;
337 int num;
338 } mapping_t;
339
340
341 struct mdstat_ent {
342 char *dev;
343 int devnum;
344 int active;
345 char *level;
346 char *pattern; /* U or up, _ for down */
347 int percent; /* -1 if no resync */
348 int resync; /* 1 if resync, 0 if recovery */
349 int devcnt;
350 int raid_disks;
351 int chunk_size;
352 char * metadata_version;
353 struct mdstat_ent *next;
354 };
355
356 extern struct mdstat_ent *mdstat_read(int hold, int start);
357 extern void free_mdstat(struct mdstat_ent *ms);
358 extern void mdstat_wait(int seconds);
359 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
360 extern int mddev_busy(int devnum);
361
362 struct map_ent {
363 struct map_ent *next;
364 int devnum;
365 char metadata[20];
366 int uuid[4];
367 int bad;
368 char *path;
369 };
370 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
371 int uuid[4], char *path);
372 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
373 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
374 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
375 extern void map_read(struct map_ent **melp);
376 extern int map_write(struct map_ent *mel);
377 extern void map_delete(struct map_ent **mapp, int devnum);
378 extern void map_free(struct map_ent *map);
379 extern void map_add(struct map_ent **melp,
380 int devnum, char *metadata, int uuid[4], char *path);
381 extern int map_lock(struct map_ent **melp);
382 extern void map_unlock(struct map_ent **melp);
383
384 /* various details can be requested */
385 enum sysfs_read_flags {
386 GET_LEVEL = (1 << 0),
387 GET_LAYOUT = (1 << 1),
388 GET_COMPONENT = (1 << 2),
389 GET_CHUNK = (1 << 3),
390 GET_CACHE = (1 << 4),
391 GET_MISMATCH = (1 << 5),
392 GET_VERSION = (1 << 6),
393 GET_DISKS = (1 << 7),
394 GET_DEGRADED = (1 << 8),
395 GET_SAFEMODE = (1 << 9),
396 GET_DEVS = (1 << 10), /* gets role, major, minor */
397 GET_OFFSET = (1 << 11),
398 GET_SIZE = (1 << 12),
399 GET_STATE = (1 << 13),
400 GET_ERROR = (1 << 14),
401 SKIP_GONE_DEVS = (1 << 15),
402 };
403
404 /* If fd >= 0, get the array it is open on,
405 * else use devnum. >=0 -> major9. <0.....
406 */
407 extern int sysfs_open(int devnum, char *devname, char *attr);
408 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
409 extern void sysfs_free(struct mdinfo *sra);
410 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
411 extern int sysfs_attr_match(const char *attr, const char *str);
412 extern int sysfs_match_word(const char *word, char **list);
413 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
414 char *name, char *val);
415 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
416 char *name, unsigned long long val);
417 extern int sysfs_uevent(struct mdinfo *sra, char *event);
418 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
419 char *name);
420 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
421 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
422 char *name, unsigned long long *val);
423 extern int sysfs_fd_get_str(int fd, char *val, int size);
424 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
425 char *name, char *val, int size);
426 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
427 extern int sysfs_set_array(struct mdinfo *info, int vers);
428 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
429 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
430 extern int sysfs_unique_holder(int devnum, long rdev);
431 extern int load_sys(char *path, char *buf);
432
433
434 extern int save_stripes(int *source, unsigned long long *offsets,
435 int raid_disks, int chunk_size, int level, int layout,
436 int nwrites, int *dest,
437 unsigned long long start, unsigned long long length,
438 char *buf);
439 extern int restore_stripes(int *dest, unsigned long long *offsets,
440 int raid_disks, int chunk_size, int level, int layout,
441 int source, unsigned long long read_offset,
442 unsigned long long start, unsigned long long length);
443
444 #ifndef Sendmail
445 #define Sendmail "/usr/lib/sendmail -t"
446 #endif
447
448 #define SYSLOG_FACILITY LOG_DAEMON
449
450 extern char *map_num(mapping_t *map, int num);
451 extern int map_name(mapping_t *map, char *name);
452 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
453
454 extern char *map_dev(int major, int minor, int create);
455
456 struct active_array;
457 struct metadata_update;
458
459 /* A superswitch provides entry point the a metadata handler.
460 *
461 * The super_switch primarily operates on some "metadata" that
462 * is accessed via the 'supertype'.
463 * This metadata has one of three possible sources.
464 * 1/ It is read from a single device. In this case it may not completely
465 * describe the array or arrays as some information might be on other
466 * devices.
467 * 2/ It is read from all devices in a container. In this case all
468 * information is present.
469 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
470 * be complete once enough ->add_to_super calls have completed.
471 *
472 * When creating an array inside a container, the metadata will be
473 * formed by a combination of 2 and 3. The metadata or the array is read,
474 * then new information is added.
475 *
476 * The metadata must sometimes have a concept of a 'current' array
477 * and a 'current' device.
478 * The 'current' array is set by init_super to be the newly created array,
479 * or is set by super_by_fd when it finds it is looking at an array inside
480 * a container.
481 *
482 * The 'current' device is either the device that the metadata was read from
483 * in case 1, or the last device added by add_to_super in case 3.
484 * Case 2 does not identify a 'current' device.
485 */
486 extern struct superswitch {
487
488 /* Used to report details of metadata read from a component
489 * device. ->load_super has been called.
490 */
491 void (*examine_super)(struct supertype *st, char *homehost);
492 void (*brief_examine_super)(struct supertype *st, int verbose);
493 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
494 void (*export_examine_super)(struct supertype *st);
495
496 /* Used to report details of an active array.
497 * ->load_super was possibly given a 'component' string.
498 */
499 void (*detail_super)(struct supertype *st, char *homehost);
500 void (*brief_detail_super)(struct supertype *st);
501 void (*export_detail_super)(struct supertype *st);
502
503 /* Optional: platform hardware / firmware details */
504 int (*detail_platform)(int verbose, int enumerate_only);
505
506 /* Used:
507 * to get uuid to storing in bitmap metadata
508 * and 'reshape' backup-data metadata
509 * To see if a device is being re-added to an array it was part of.
510 */
511 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
512
513 /* Extract generic details from metadata. This could be details about
514 * the container, or about an individual array within the container.
515 * The determination is made either by:
516 * load_super being given a 'component' string.
517 * validate_geometry determining what to create.
518 * The info includes both array information and device information.
519 * The particular device should be:
520 * The last device added by add_to_super
521 * The device the metadata was loaded from by load_super
522 */
523 void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
524
525 /* Check if the given metadata is flagged as belonging to "this"
526 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
527 */
528 int (*match_home)(struct supertype *st, char *homehost);
529
530 /* Make one of several generic modifications to metadata
531 * prior to assembly (or other times).
532 * sparc2.2 - first bug in early 0.90 metadata
533 * super-minor - change name of 0.90 metadata
534 * summaries - 'correct' any redundant data
535 * resync - mark array as dirty to trigger a resync.
536 * uuid - set new uuid - only 0.90 or 1.x
537 * name - change the name of the array (where supported)
538 * homehost - change which host this array is tied to.
539 * devicesize - If metadata is at start of device, change recorded
540 * device size to match actual device size
541 * byteorder - swap bytes for 0.90 metadata
542 *
543 * force-one - mark that device as uptodate, not old or failed.
544 * force-array - mark array as clean if it would not otherwise
545 * assemble
546 * assemble - not sure how this is different from force-one...
547 * linear-grow-new - add a new device to a linear array, but don't
548 * change the size: so superblock still matches
549 * linear-grow-update - now change the size of the array.
550 */
551 int (*update_super)(struct supertype *st, struct mdinfo *info,
552 char *update,
553 char *devname, int verbose,
554 int uuid_set, char *homehost);
555
556 /* Create new metadata for new array as described. This could
557 * be a new container, or an array in a pre-existing container.
558 * Also used to zero metadata prior to writing it to invalidate old
559 * metadata.
560 */
561 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
562 unsigned long long size, char *name,
563 char *homehost, int *uuid);
564
565 /* update the metadata to include new device, either at create or
566 * when hot-adding a spare.
567 */
568 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
569 int fd, char *devname);
570
571 /* Write metadata to one device when fixing problems or adding
572 * a new device.
573 */
574 int (*store_super)(struct supertype *st, int fd);
575
576 /* Write all metadata for this array.
577 */
578 int (*write_init_super)(struct supertype *st);
579 int (*compare_super)(struct supertype *st, struct supertype *tst);
580 int (*load_super)(struct supertype *st, int fd, char *devname);
581 struct supertype * (*match_metadata_desc)(char *arg);
582 __u64 (*avail_size)(struct supertype *st, __u64 size);
583 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
584 int delay, int write_behind,
585 unsigned long long size, int may_change, int major);
586 void (*locate_bitmap)(struct supertype *st, int fd);
587 int (*write_bitmap)(struct supertype *st, int fd);
588 void (*free_super)(struct supertype *st);
589
590 /* validate_geometry is called with an st returned by
591 * match_metadata_desc.
592 * It should check that the geometry described in compatible with
593 * the metadata type. It will be called repeatedly as devices
594 * added to validate changing size and new devices. If there are
595 * inter-device dependencies, it should record sufficient details
596 * so these can be validated.
597 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
598 */
599 int (*validate_geometry)(struct supertype *st, int level, int layout,
600 int raiddisks,
601 int chunk, unsigned long long size,
602 char *subdev, unsigned long long *freesize,
603 int verbose);
604
605 struct mdinfo *(*container_content)(struct supertype *st);
606 /* Allow a metadata handler to override mdadm's default layouts */
607 int (*default_layout)(int level); /* optional */
608
609 /* for mdmon */
610 int (*open_new)(struct supertype *c, struct active_array *a,
611 char *inst);
612
613 /* Tell the metadata handler the current state of the array.
614 * This covers whether it is known to be consistent (no pending writes)
615 * and how far along a resync is known to have progressed
616 * (in a->resync_start).
617 * resync status is really irrelevant if the array is not consistent,
618 * but some metadata (DDF!) have a place to record the distinction.
619 * If 'consistent' is '2', then the array can mark it dirty if a
620 * resync/recovery/whatever is required, or leave it clean if not.
621 * Return value is 0 dirty (not consistent) and 1 if clean.
622 * it is only really important if consistent is passed in as '2'.
623 */
624 int (*set_array_state)(struct active_array *a, int consistent);
625
626 /* When the state of a device might have changed, we call set_disk to
627 * tell the metadata what the current state is.
628 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
629 * transitions.
630 * set_disk might be called when the state of the particular disk has
631 * not in fact changed.
632 */
633 void (*set_disk)(struct active_array *a, int n, int state);
634 void (*sync_metadata)(struct supertype *st);
635 void (*process_update)(struct supertype *st,
636 struct metadata_update *update);
637 void (*prepare_update)(struct supertype *st,
638 struct metadata_update *update);
639
640 /* activate_spare will check if the array is degraded and, if it
641 * is, try to find some spare space in the container.
642 * On success, it add appropriate updates (For process_update) to
643 * to the 'updates' list and returns a list of 'mdinfo' identifying
644 * the device, or devices as there might be multiple missing
645 * devices and multiple spares available.
646 */
647 struct mdinfo *(*activate_spare)(struct active_array *a,
648 struct metadata_update **updates);
649
650 int swapuuid; /* true if uuid is bigending rather than hostendian */
651 int external;
652 const char *name; /* canonical metadata name */
653 } super0, super1, super_ddf, *superlist[];
654
655 extern struct superswitch super_imsm;
656
657 struct metadata_update {
658 int len;
659 char *buf;
660 void *space; /* allocated space that monitor will use */
661 struct metadata_update *next;
662 };
663
664 /* A supertype holds a particular collection of metadata.
665 * It identifies the metadata type by the superswitch, and the particular
666 * sub-version of that metadata type.
667 * metadata read in or created is stored in 'sb' and 'info'.
668 * There are also fields used by mdmon to track containers.
669 *
670 * A supertype may refer to:
671 * Just an array, possibly in a container
672 * A container, not identifying any particular array
673 * Info read from just one device, not yet fully describing the array/container.
674 *
675 *
676 * A supertype is created by:
677 * super_by_fd
678 * guess_super
679 * dup_super
680 */
681 struct supertype {
682 struct superswitch *ss;
683 int minor_version;
684 int max_devs;
685 int container_dev; /* devnum of container */
686 char subarray[32]; /* name of array inside container */
687 void *sb;
688 void *info;
689 int loaded_container; /* Set if load_super found a container,
690 * not just one device */
691
692 struct metadata_update *updates;
693 struct metadata_update **update_tail;
694
695 /* extra stuff used by mdmon */
696 struct active_array *arrays;
697 int sock; /* listen to external programs */
698 int devnum;
699 char *devname; /* e.g. md0. This appears in metadata_verison:
700 * external:/md0/12
701 */
702 int devcnt;
703
704 struct mdinfo *devs;
705
706 };
707
708 extern struct supertype *super_by_fd(int fd);
709 extern struct supertype *guess_super(int fd);
710 extern struct supertype *dup_super(struct supertype *st);
711 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
712 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
713 mdu_disk_info_t *disk);
714 void wait_for(char *dev, int fd);
715
716 #if __GNUC__ < 3
717 struct stat64;
718 #endif
719
720 #define HAVE_NFTW we assume
721 #define HAVE_FTW
722
723 #ifdef __UCLIBC__
724 # include <features.h>
725 # ifndef __UCLIBC_HAS_LFS__
726 # define lseek64 lseek
727 # endif
728 # ifndef __UCLIBC_HAS_FTW__
729 # undef HAVE_FTW
730 # undef HAVE_NFTW
731 # endif
732 #endif
733
734 #ifdef __dietlibc__
735 # undef HAVE_NFTW
736 #endif
737
738 #if defined(__KLIBC__)
739 # undef HAVE_NFTW
740 # undef HAVE_FTW
741 #endif
742
743 #ifndef HAVE_NFTW
744 # define FTW_PHYS 1
745 # ifndef HAVE_FTW
746 struct FTW {};
747 # endif
748 #endif
749
750 #ifdef HAVE_FTW
751 # include <ftw.h>
752 #endif
753
754 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
755
756
757 extern int Manage_ro(char *devname, int fd, int readonly);
758 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
759 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
760 extern int Manage_subdevs(char *devname, int fd,
761 mddev_dev_t devlist, int verbose);
762 extern int autodetect(void);
763 extern int Grow_Add_device(char *devname, int fd, char *newdev);
764 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
765 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
766 long long size,
767 int level, char *layout_str, int chunksize, int raid_disks);
768 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
769 int *fdlist, int cnt, char *backup_file, int verbose);
770 extern int Grow_continue(int mdfd, struct supertype *st,
771 struct mdinfo *info, char *backup_file);
772
773 extern int Assemble(struct supertype *st, char *mddev,
774 mddev_ident_t ident,
775 mddev_dev_t devlist, char *backup_file,
776 int readonly, int runstop,
777 char *update, char *homehost, int require_homehost,
778 int verbose, int force);
779
780 extern int Build(char *mddev, int chunk, int level, int layout,
781 int raiddisks, mddev_dev_t devlist, int assume_clean,
782 char *bitmap_file, int bitmap_chunk, int write_behind,
783 int delay, int verbose, int autof, unsigned long long size);
784
785
786 extern int Create(struct supertype *st, char *mddev,
787 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
788 char *name, char *homehost, int *uuid,
789 int subdevs, mddev_dev_t devlist,
790 int runstop, int verbose, int force, int assume_clean,
791 char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
792
793 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
794 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
795 extern int Query(char *dev);
796 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
797 int SparcAdjust, struct supertype *forcest, char *homehost);
798 extern int Monitor(mddev_dev_t devlist,
799 char *mailaddr, char *alert_cmd,
800 int period, int daemonise, int scan, int oneshot,
801 int dosyslog, int test, char *pidfile, int increments);
802
803 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
804 extern int Wait(char *dev);
805 extern int WaitClean(char *dev, int sock, int verbose);
806
807 extern int Incremental(char *devname, int verbose, int runstop,
808 struct supertype *st, char *homehost, int require_homehost,
809 int autof);
810 extern int Incremental_container(struct supertype *st, char *devname,
811 int verbose, int runstop, int autof,
812 int trustworthy);
813 extern void RebuildMap(void);
814 extern int IncrementalScan(int verbose);
815
816 extern int CreateBitmap(char *filename, int force, char uuid[16],
817 unsigned long chunksize, unsigned long daemon_sleep,
818 unsigned long write_behind,
819 unsigned long long array_size,
820 int major);
821 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
822 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
823 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
824
825 extern int md_get_version(int fd);
826 extern int get_linux_version(void);
827 extern long long parse_size(char *size);
828 extern int parse_uuid(char *str, int uuid[4]);
829 extern int parse_layout_10(char *layout);
830 extern int parse_layout_faulty(char *layout);
831 extern int check_ext2(int fd, char *name);
832 extern int check_reiser(int fd, char *name);
833 extern int check_raid(int fd, char *name);
834 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
835
836 extern int get_mdp_major(void);
837 extern int dev_open(char *dev, int flags);
838 extern int open_dev(int devnum);
839 extern int open_dev_excl(int devnum);
840 extern int is_standard(char *dev, int *nump);
841 extern int same_dev(char *one, char *two);
842
843 extern int parse_auto(char *str, char *msg, int config);
844 extern mddev_ident_t conf_get_ident(char *dev);
845 extern mddev_dev_t conf_get_devs(void);
846 extern int conf_test_dev(char *devname);
847 extern int conf_test_metadata(const char *version, int is_homehost);
848 extern struct createinfo *conf_get_create_info(void);
849 extern void set_conffile(char *file);
850 extern char *conf_get_mailaddr(void);
851 extern char *conf_get_mailfrom(void);
852 extern char *conf_get_program(void);
853 extern char *conf_get_homehost(int *require_homehostp);
854 extern char *conf_line(FILE *file);
855 extern char *conf_word(FILE *file, int allow_key);
856 extern int conf_name_is_free(char *name);
857 extern int devname_matches(char *name, char *match);
858 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
859
860 extern void free_line(char *line);
861 extern int match_oneof(char *devices, char *devname);
862 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
863 extern const int uuid_match_any[4];
864 extern int same_uuid(int a[4], int b[4], int swapuuid);
865 extern void copy_uuid(void *a, int b[4], int swapuuid);
866 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
867 extern char *fname_from_uuid(struct supertype *st,
868 struct mdinfo *info, char *buf, char sep);
869 extern unsigned long calc_csum(void *super, int bytes);
870 extern int enough(int level, int raid_disks, int layout, int clean,
871 char *avail, int avail_disks);
872 extern int ask(char *mesg);
873 extern unsigned long long get_component_size(int fd);
874 extern void remove_partitions(int fd);
875 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
876 int chunksize, unsigned long long devsize);
877 extern int flush_metadata_updates(struct supertype *st);
878 extern void append_metadata_update(struct supertype *st, void *buf, int len);
879 extern int assemble_container_content(struct supertype *st, int mdfd,
880 struct mdinfo *content, int runstop,
881 char *chosen_name, int verbose);
882
883 extern int add_disk(int mdfd, struct supertype *st,
884 struct mdinfo *sra, struct mdinfo *info);
885 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
886 unsigned long long min_recovery_start(struct mdinfo *array);
887
888 extern char *human_size(long long bytes);
889 extern char *human_size_brief(long long bytes);
890 extern void print_r10_layout(int layout);
891
892 #define NoMdDev (1<<23)
893 extern int find_free_devnum(int use_partitions);
894
895 extern void put_md_name(char *name);
896 extern char *get_md_name(int dev);
897
898 extern char DefaultConfFile[];
899
900 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
901 char *chosen);
902 /* values for 'trustworthy' */
903 #define LOCAL 1
904 #define LOCAL_ANY 10
905 #define FOREIGN 2
906 #define METADATA 3
907 extern int open_mddev(char *dev, int report_errors);
908 extern int open_container(int fd);
909
910 extern char *pid_dir;
911 extern int mdmon_running(int devnum);
912 extern int mdmon_pid(int devnum);
913 extern int check_env(char *name);
914 extern __u32 random32(void);
915 extern int start_mdmon(int devnum);
916
917 extern char *devnum2devname(int num);
918 extern int devname2devnum(char *name);
919 extern int stat2devnum(struct stat *st);
920 extern int fd2devnum(int fd);
921
922 static inline int dev2major(int d)
923 {
924 if (d >= 0)
925 return MD_MAJOR;
926 else
927 return get_mdp_major();
928 }
929
930 static inline int dev2minor(int d)
931 {
932 if (d >= 0)
933 return d;
934 return (-1-d) << MdpMinorShift;
935 }
936
937 static inline int ROUND_UP(int a, int base)
938 {
939 return ((a+base-1)/base)*base;
940 }
941
942 static inline int is_subarray(char *vers)
943 {
944 /* The version string for a 'subarray' (an array in a container)
945 * is
946 * /containername/componentname for normal read-write arrays
947 * -containername/componentname for read-only arrays.
948 * containername is e.g. md0, md_d1
949 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
950 */
951 return (*vers == '/' || *vers == '-');
952 }
953
954 #ifdef DEBUG
955 #define dprintf(fmt, arg...) \
956 fprintf(stderr, fmt, ##arg)
957 #else
958 #define dprintf(fmt, arg...) \
959 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
960 #endif
961 #include <assert.h>
962 #include <stdarg.h>
963 static inline int xasprintf(char **strp, const char *fmt, ...) {
964 va_list ap;
965 int ret;
966 va_start(ap, fmt);
967 ret = vasprintf(strp, fmt, ap);
968 va_end(ap);
969 assert(ret >= 0);
970 return ret;
971 }
972
973 #define LEVEL_MULTIPATH (-4)
974 #define LEVEL_LINEAR (-1)
975 #define LEVEL_FAULTY (-5)
976
977 /* kernel module doesn't know about these */
978 #define LEVEL_CONTAINER (-100)
979 #define LEVEL_UNSUPPORTED (-200)
980
981
982 /* faulty stuff */
983
984 #define WriteTransient 0
985 #define ReadTransient 1
986 #define WritePersistent 2
987 #define ReadPersistent 3
988 #define WriteAll 4 /* doesn't go to device */
989 #define ReadFixable 5
990 #define Modes 6
991
992 #define ClearErrors 31
993 #define ClearFaults 30
994
995 #define AllPersist 100 /* internal use only */
996 #define NoPersist 101
997
998 #define ModeMask 0x1f
999 #define ModeShift 5
1000
1001
1002 #ifdef __TINYC__
1003 #undef minor
1004 #undef major
1005 #undef makedev
1006 #define minor(x) ((x)&0xff)
1007 #define major(x) (((x)>>8)&0xff)
1008 #define makedev(M,m) (((M)<<8) | (m))
1009 #endif
1010
1011 /* for raid4/5/6 */
1012 #define ALGORITHM_LEFT_ASYMMETRIC 0
1013 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1014 #define ALGORITHM_LEFT_SYMMETRIC 2
1015 #define ALGORITHM_RIGHT_SYMMETRIC 3
1016
1017 /* Define non-rotating (raid4) algorithms. These allow
1018 * conversion of raid4 to raid5.
1019 */
1020 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1021 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1022
1023 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1024 * Firstly, the exact positioning of the parity block is slightly
1025 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1026 * of DDF.
1027 * Secondly, or order of datablocks over which the Q syndrome is computed
1028 * is different.
1029 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1030 * These layouts are from the DDFv1.2 spec.
1031 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1032 * leaves RLQ=3 as 'Vendor Specific'
1033 */
1034
1035 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1036 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1037 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1038
1039
1040 /* For every RAID5 algorithm we define a RAID6 algorithm
1041 * with exactly the same layout for data and parity, and
1042 * with the Q block always on the last device (N-1).
1043 * This allows trivial conversion from RAID5 to RAID6
1044 */
1045 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1046 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1047 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1048 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1049 #define ALGORITHM_PARITY_0_6 20
1050 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1051
1052 /* Define PATH_MAX in case we don't use glibc or standard library does
1053 * not have PATH_MAX defined. Assume max path length is 4K characters.
1054 */
1055 #ifndef PATH_MAX
1056 #define PATH_MAX 4096
1057 #endif
1058