]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Create: cleanup after failed create in duplicated array member case
[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 #ifndef ALT_MAPFILE
91 #define ALT_MAPFILE "map"
92 #endif /* ALT_MAPFILE */
93
94 #include "md_u.h"
95 #include "md_p.h"
96 #include "bitmap.h"
97 #include "msg.h"
98
99 #include <endian.h>
100 /* Redhat don't like to #include <asm/byteorder.h>, and
101 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
102 * and there is no standard conversion function so... */
103 /* And dietlibc doesn't think byteswap is ok, so.. */
104 /* #include <byteswap.h> */
105 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
106 ((x) & 0xff00U) >> 8)
107 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
108 ((x) & 0xff000000U) >> 24 | \
109 ((x) & 0x0000ff00U) << 8 | \
110 ((x) & 0x00ff0000U) >> 8)
111 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
112 ((x) & 0xff00000000000000ULL) >> 56 | \
113 ((x) & 0x000000000000ff00ULL) << 40 | \
114 ((x) & 0x00ff000000000000ULL) >> 40 | \
115 ((x) & 0x0000000000ff0000ULL) << 24 | \
116 ((x) & 0x0000ff0000000000ULL) >> 24 | \
117 ((x) & 0x00000000ff000000ULL) << 8 | \
118 ((x) & 0x000000ff00000000ULL) >> 8)
119
120 #if !defined(__KLIBC__)
121 #if BYTE_ORDER == LITTLE_ENDIAN
122 #define __cpu_to_le16(_x) (_x)
123 #define __cpu_to_le32(_x) (_x)
124 #define __cpu_to_le64(_x) (_x)
125 #define __le16_to_cpu(_x) (_x)
126 #define __le32_to_cpu(_x) (_x)
127 #define __le64_to_cpu(_x) (_x)
128
129 #define __cpu_to_be16(_x) bswap_16(_x)
130 #define __cpu_to_be32(_x) bswap_32(_x)
131 #define __cpu_to_be64(_x) bswap_64(_x)
132 #define __be16_to_cpu(_x) bswap_16(_x)
133 #define __be32_to_cpu(_x) bswap_32(_x)
134 #define __be64_to_cpu(_x) bswap_64(_x)
135 #elif BYTE_ORDER == BIG_ENDIAN
136 #define __cpu_to_le16(_x) bswap_16(_x)
137 #define __cpu_to_le32(_x) bswap_32(_x)
138 #define __cpu_to_le64(_x) bswap_64(_x)
139 #define __le16_to_cpu(_x) bswap_16(_x)
140 #define __le32_to_cpu(_x) bswap_32(_x)
141 #define __le64_to_cpu(_x) bswap_64(_x)
142
143 #define __cpu_to_be16(_x) (_x)
144 #define __cpu_to_be32(_x) (_x)
145 #define __cpu_to_be64(_x) (_x)
146 #define __be16_to_cpu(_x) (_x)
147 #define __be32_to_cpu(_x) (_x)
148 #define __be64_to_cpu(_x) (_x)
149 #else
150 # error "unknown endianness."
151 #endif
152 #endif /* __KLIBC__ */
153
154
155 /*
156 * min()/max()/clamp() macros that also do
157 * strict type-checking.. See the
158 * "unnecessary" pointer comparison.
159 */
160 #define min(x, y) ({ \
161 typeof(x) _min1 = (x); \
162 typeof(y) _min2 = (y); \
163 (void) (&_min1 == &_min2); \
164 _min1 < _min2 ? _min1 : _min2; })
165
166 #define max(x, y) ({ \
167 typeof(x) _max1 = (x); \
168 typeof(y) _max2 = (y); \
169 (void) (&_max1 == &_max2); \
170 _max1 > _max2 ? _max1 : _max2; })
171
172 /* general information that might be extracted from a superblock */
173 struct mdinfo {
174 mdu_array_info_t array;
175 mdu_disk_info_t disk;
176 __u64 events;
177 int uuid[4];
178 char name[33];
179 unsigned long long data_offset;
180 unsigned long long component_size; /* same as array.size, except in
181 * sectors and up to 64bits.
182 */
183 unsigned long long custom_array_size; /* size for non-default sized
184 * arrays (in sectors)
185 */
186 int reshape_active;
187 unsigned long long reshape_progress;
188 union {
189 unsigned long long resync_start; /* per-array resync position */
190 unsigned long long recovery_start; /* per-device rebuild position */
191 #define MaxSector (~0ULL) /* resync/recovery complete position */
192 };
193 unsigned long safe_mode_delay; /* ms delay to mark clean */
194 int new_level, delta_disks, new_layout, new_chunk;
195 int errors;
196 int cache_size; /* size of raid456 stripe cache*/
197 int mismatch_cnt;
198 char text_version[50];
199 void *update_private; /* for passing metadata-format
200 * specific update data
201 * between successive calls to
202 * update_super()
203 */
204
205 int container_member; /* for assembling external-metatdata arrays
206 * This is to be used internally by metadata
207 * handler only */
208
209 char sys_name[20];
210 struct mdinfo *devs;
211 struct mdinfo *next;
212
213 /* Device info for mdmon: */
214 int recovery_fd;
215 int state_fd;
216 #define DS_FAULTY 1
217 #define DS_INSYNC 2
218 #define DS_WRITE_MOSTLY 4
219 #define DS_SPARE 8
220 #define DS_BLOCKED 16
221 #define DS_REMOVE 1024
222 #define DS_UNBLOCK 2048
223 int prev_state, curr_state, next_state;
224
225 };
226
227 struct createinfo {
228 int uid;
229 int gid;
230 int autof;
231 int mode;
232 int symlinks;
233 struct supertype *supertype;
234 };
235
236 #define Name "mdadm"
237
238 enum mode {
239 ASSEMBLE=1,
240 BUILD,
241 CREATE,
242 MANAGE,
243 MISC,
244 MONITOR,
245 GROW,
246 INCREMENTAL,
247 AUTODETECT,
248 };
249
250 extern char short_options[];
251 extern char short_bitmap_options[];
252 extern char short_bitmap_auto_options[];
253 extern struct option long_options[];
254 extern char Version[], Usage[], Help[], OptionHelp[],
255 Help_create[], Help_build[], Help_assemble[], Help_grow[],
256 Help_incr[],
257 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
258
259 /* for option that don't have short equivilents, we assign arbitrary
260 * small numbers. '1' means an undecorated option, so we start at '2'.
261 */
262 enum special_options {
263 AssumeClean = 2,
264 BitmapChunk,
265 WriteBehind,
266 ReAdd,
267 NoDegraded,
268 Sparc22,
269 BackupFile,
270 HomeHost,
271 AutoHomeHost,
272 Symlinks,
273 AutoDetect,
274 Waitclean,
275 DetailPlatform,
276 };
277
278 /* structures read from config file */
279 /* List of mddevice names and identifiers
280 * Identifiers can be:
281 * uuid=128-hex-uuid
282 * super-minor=decimal-minor-number-from-superblock
283 * devices=comma,separated,list,of,device,names,with,wildcards
284 *
285 * If multiple fields are present, the intersection of all matching
286 * devices is considered
287 */
288 #define UnSet (0xfffe)
289 typedef struct mddev_ident_s {
290 char *devname;
291
292 int uuid_set;
293 int uuid[4];
294 char name[33];
295
296 unsigned int super_minor;
297
298 char *devices; /* comma separated list of device
299 * names with wild cards
300 */
301 int level;
302 unsigned int raid_disks;
303 unsigned int spare_disks;
304 struct supertype *st;
305 int autof; /* 1 for normal, 2 for partitioned */
306 char *spare_group;
307 char *bitmap_file;
308 int bitmap_fd;
309
310 char *container; /* /dev/whatever name of container, or
311 * uuid of container. You would expect
312 * this to be the 'devname' or UUID
313 * of some other entry.
314 */
315 char *member; /* subarray within a container */
316
317 struct mddev_ident_s *next;
318 union {
319 /* fields needed by different users of this structure */
320 int assembled; /* set when assembly succeeds */
321 };
322 } *mddev_ident_t;
323
324 /* List of device names - wildcards expanded */
325 typedef struct mddev_dev_s {
326 char *devname;
327 char disposition; /* 'a' for add, 'r' for remove, 'f' for fail.
328 * Not set for names read from .config
329 */
330 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
331 char re_add;
332 char used; /* set when used */
333 struct mdinfo *content; /* If devname is a container, this might list
334 * the remaining member arrays. */
335 struct mddev_dev_s *next;
336 } *mddev_dev_t;
337
338 typedef struct mapping {
339 char *name;
340 int num;
341 } mapping_t;
342
343
344 struct mdstat_ent {
345 char *dev;
346 int devnum;
347 int active;
348 char *level;
349 char *pattern; /* U or up, _ for down */
350 int percent; /* -1 if no resync */
351 int resync; /* 1 if resync, 0 if recovery */
352 int devcnt;
353 int raid_disks;
354 int chunk_size;
355 char * metadata_version;
356 struct mdstat_ent *next;
357 };
358
359 extern struct mdstat_ent *mdstat_read(int hold, int start);
360 extern void free_mdstat(struct mdstat_ent *ms);
361 extern void mdstat_wait(int seconds);
362 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
363 extern int mddev_busy(int devnum);
364
365 struct map_ent {
366 struct map_ent *next;
367 int devnum;
368 char metadata[20];
369 int uuid[4];
370 int bad;
371 char *path;
372 };
373 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
374 int uuid[4], char *path);
375 extern void map_remove(struct map_ent **map, int devnum);
376 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
377 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
378 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
379 extern void map_read(struct map_ent **melp);
380 extern int map_write(struct map_ent *mel);
381 extern void map_delete(struct map_ent **mapp, int devnum);
382 extern void map_free(struct map_ent *map);
383 extern void map_add(struct map_ent **melp,
384 int devnum, char *metadata, int uuid[4], char *path);
385 extern int map_lock(struct map_ent **melp);
386 extern void map_unlock(struct map_ent **melp);
387
388 /* various details can be requested */
389 enum sysfs_read_flags {
390 GET_LEVEL = (1 << 0),
391 GET_LAYOUT = (1 << 1),
392 GET_COMPONENT = (1 << 2),
393 GET_CHUNK = (1 << 3),
394 GET_CACHE = (1 << 4),
395 GET_MISMATCH = (1 << 5),
396 GET_VERSION = (1 << 6),
397 GET_DISKS = (1 << 7),
398 GET_DEGRADED = (1 << 8),
399 GET_SAFEMODE = (1 << 9),
400 GET_DEVS = (1 << 10), /* gets role, major, minor */
401 GET_OFFSET = (1 << 11),
402 GET_SIZE = (1 << 12),
403 GET_STATE = (1 << 13),
404 GET_ERROR = (1 << 14),
405 SKIP_GONE_DEVS = (1 << 15),
406 };
407
408 /* If fd >= 0, get the array it is open on,
409 * else use devnum. >=0 -> major9. <0.....
410 */
411 extern int sysfs_open(int devnum, char *devname, char *attr);
412 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
413 extern void sysfs_free(struct mdinfo *sra);
414 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
415 extern int sysfs_attr_match(const char *attr, const char *str);
416 extern int sysfs_match_word(const char *word, char **list);
417 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
418 char *name, char *val);
419 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
420 char *name, unsigned long long val);
421 extern int sysfs_uevent(struct mdinfo *sra, char *event);
422 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
423 char *name);
424 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
425 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
426 char *name, unsigned long long *val);
427 extern int sysfs_fd_get_str(int fd, char *val, int size);
428 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
429 char *name, char *val, int size);
430 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
431 extern int sysfs_set_array(struct mdinfo *info, int vers);
432 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
433 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
434 extern int sysfs_unique_holder(int devnum, long rdev);
435 extern int load_sys(char *path, char *buf);
436
437
438 extern int save_stripes(int *source, unsigned long long *offsets,
439 int raid_disks, int chunk_size, int level, int layout,
440 int nwrites, int *dest,
441 unsigned long long start, unsigned long long length,
442 char *buf);
443 extern int restore_stripes(int *dest, unsigned long long *offsets,
444 int raid_disks, int chunk_size, int level, int layout,
445 int source, unsigned long long read_offset,
446 unsigned long long start, unsigned long long length);
447
448 #ifndef Sendmail
449 #define Sendmail "/usr/lib/sendmail -t"
450 #endif
451
452 #define SYSLOG_FACILITY LOG_DAEMON
453
454 extern char *map_num(mapping_t *map, int num);
455 extern int map_name(mapping_t *map, char *name);
456 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
457
458 extern char *map_dev(int major, int minor, int create);
459
460 struct active_array;
461 struct metadata_update;
462
463 /* A superswitch provides entry point the a metadata handler.
464 *
465 * The super_switch primarily operates on some "metadata" that
466 * is accessed via the 'supertype'.
467 * This metadata has one of three possible sources.
468 * 1/ It is read from a single device. In this case it may not completely
469 * describe the array or arrays as some information might be on other
470 * devices.
471 * 2/ It is read from all devices in a container. In this case all
472 * information is present.
473 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
474 * be complete once enough ->add_to_super calls have completed.
475 *
476 * When creating an array inside a container, the metadata will be
477 * formed by a combination of 2 and 3. The metadata or the array is read,
478 * then new information is added.
479 *
480 * The metadata must sometimes have a concept of a 'current' array
481 * and a 'current' device.
482 * The 'current' array is set by init_super to be the newly created array,
483 * or is set by super_by_fd when it finds it is looking at an array inside
484 * a container.
485 *
486 * The 'current' device is either the device that the metadata was read from
487 * in case 1, or the last device added by add_to_super in case 3.
488 * Case 2 does not identify a 'current' device.
489 */
490 extern struct superswitch {
491
492 /* Used to report details of metadata read from a component
493 * device. ->load_super has been called.
494 */
495 void (*examine_super)(struct supertype *st, char *homehost);
496 void (*brief_examine_super)(struct supertype *st, int verbose);
497 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
498 void (*export_examine_super)(struct supertype *st);
499
500 /* Used to report details of an active array.
501 * ->load_super was possibly given a 'component' string.
502 */
503 void (*detail_super)(struct supertype *st, char *homehost);
504 void (*brief_detail_super)(struct supertype *st);
505 void (*export_detail_super)(struct supertype *st);
506
507 /* Optional: platform hardware / firmware details */
508 int (*detail_platform)(int verbose, int enumerate_only);
509
510 /* Used:
511 * to get uuid to storing in bitmap metadata
512 * and 'reshape' backup-data metadata
513 * To see if a device is being re-added to an array it was part of.
514 */
515 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
516
517 /* Extract generic details from metadata. This could be details about
518 * the container, or about an individual array within the container.
519 * The determination is made either by:
520 * load_super being given a 'component' string.
521 * validate_geometry determining what to create.
522 * The info includes both array information and device information.
523 * The particular device should be:
524 * The last device added by add_to_super
525 * The device the metadata was loaded from by load_super
526 */
527 void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
528
529 /* Check if the given metadata is flagged as belonging to "this"
530 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
531 */
532 int (*match_home)(struct supertype *st, char *homehost);
533
534 /* Make one of several generic modifications to metadata
535 * prior to assembly (or other times).
536 * sparc2.2 - first bug in early 0.90 metadata
537 * super-minor - change name of 0.90 metadata
538 * summaries - 'correct' any redundant data
539 * resync - mark array as dirty to trigger a resync.
540 * uuid - set new uuid - only 0.90 or 1.x
541 * name - change the name of the array (where supported)
542 * homehost - change which host this array is tied to.
543 * devicesize - If metadata is at start of device, change recorded
544 * device size to match actual device size
545 * byteorder - swap bytes for 0.90 metadata
546 *
547 * force-one - mark that device as uptodate, not old or failed.
548 * force-array - mark array as clean if it would not otherwise
549 * assemble
550 * assemble - not sure how this is different from force-one...
551 * linear-grow-new - add a new device to a linear array, but don't
552 * change the size: so superblock still matches
553 * linear-grow-update - now change the size of the array.
554 */
555 int (*update_super)(struct supertype *st, struct mdinfo *info,
556 char *update,
557 char *devname, int verbose,
558 int uuid_set, char *homehost);
559
560 /* Create new metadata for new array as described. This could
561 * be a new container, or an array in a pre-existing container.
562 * Also used to zero metadata prior to writing it to invalidate old
563 * metadata.
564 */
565 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
566 unsigned long long size, char *name,
567 char *homehost, int *uuid);
568
569 /* update the metadata to include new device, either at create or
570 * when hot-adding a spare.
571 */
572 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
573 int fd, char *devname);
574
575 /* Write metadata to one device when fixing problems or adding
576 * a new device.
577 */
578 int (*store_super)(struct supertype *st, int fd);
579
580 /* Write all metadata for this array.
581 */
582 int (*write_init_super)(struct supertype *st);
583 int (*compare_super)(struct supertype *st, struct supertype *tst);
584 int (*load_super)(struct supertype *st, int fd, char *devname);
585 struct supertype * (*match_metadata_desc)(char *arg);
586 __u64 (*avail_size)(struct supertype *st, __u64 size);
587 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
588 int delay, int write_behind,
589 unsigned long long size, int may_change, int major);
590 void (*locate_bitmap)(struct supertype *st, int fd);
591 int (*write_bitmap)(struct supertype *st, int fd);
592 void (*free_super)(struct supertype *st);
593
594 /* validate_geometry is called with an st returned by
595 * match_metadata_desc.
596 * It should check that the geometry described in compatible with
597 * the metadata type. It will be called repeatedly as devices
598 * added to validate changing size and new devices. If there are
599 * inter-device dependencies, it should record sufficient details
600 * so these can be validated.
601 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
602 */
603 int (*validate_geometry)(struct supertype *st, int level, int layout,
604 int raiddisks,
605 int chunk, unsigned long long size,
606 char *subdev, unsigned long long *freesize,
607 int verbose);
608
609 struct mdinfo *(*container_content)(struct supertype *st);
610 /* Allow a metadata handler to override mdadm's default layouts */
611 int (*default_layout)(int level); /* optional */
612
613 /* for mdmon */
614 int (*open_new)(struct supertype *c, struct active_array *a,
615 char *inst);
616
617 /* Tell the metadata handler the current state of the array.
618 * This covers whether it is known to be consistent (no pending writes)
619 * and how far along a resync is known to have progressed
620 * (in a->resync_start).
621 * resync status is really irrelevant if the array is not consistent,
622 * but some metadata (DDF!) have a place to record the distinction.
623 * If 'consistent' is '2', then the array can mark it dirty if a
624 * resync/recovery/whatever is required, or leave it clean if not.
625 * Return value is 0 dirty (not consistent) and 1 if clean.
626 * it is only really important if consistent is passed in as '2'.
627 */
628 int (*set_array_state)(struct active_array *a, int consistent);
629
630 /* When the state of a device might have changed, we call set_disk to
631 * tell the metadata what the current state is.
632 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
633 * transitions.
634 * set_disk might be called when the state of the particular disk has
635 * not in fact changed.
636 */
637 void (*set_disk)(struct active_array *a, int n, int state);
638 void (*sync_metadata)(struct supertype *st);
639 void (*process_update)(struct supertype *st,
640 struct metadata_update *update);
641 void (*prepare_update)(struct supertype *st,
642 struct metadata_update *update);
643
644 /* activate_spare will check if the array is degraded and, if it
645 * is, try to find some spare space in the container.
646 * On success, it add appropriate updates (For process_update) to
647 * to the 'updates' list and returns a list of 'mdinfo' identifying
648 * the device, or devices as there might be multiple missing
649 * devices and multiple spares available.
650 */
651 struct mdinfo *(*activate_spare)(struct active_array *a,
652 struct metadata_update **updates);
653
654 int swapuuid; /* true if uuid is bigending rather than hostendian */
655 int external;
656 const char *name; /* canonical metadata name */
657 } super0, super1, super_ddf, *superlist[];
658
659 extern struct superswitch super_imsm;
660
661 struct metadata_update {
662 int len;
663 char *buf;
664 void *space; /* allocated space that monitor will use */
665 struct metadata_update *next;
666 };
667
668 /* A supertype holds a particular collection of metadata.
669 * It identifies the metadata type by the superswitch, and the particular
670 * sub-version of that metadata type.
671 * metadata read in or created is stored in 'sb' and 'info'.
672 * There are also fields used by mdmon to track containers.
673 *
674 * A supertype may refer to:
675 * Just an array, possibly in a container
676 * A container, not identifying any particular array
677 * Info read from just one device, not yet fully describing the array/container.
678 *
679 *
680 * A supertype is created by:
681 * super_by_fd
682 * guess_super
683 * dup_super
684 */
685 struct supertype {
686 struct superswitch *ss;
687 int minor_version;
688 int max_devs;
689 int container_dev; /* devnum of container */
690 char subarray[32]; /* name of array inside container */
691 void *sb;
692 void *info;
693 int loaded_container; /* Set if load_super found a container,
694 * not just one device */
695
696 struct metadata_update *updates;
697 struct metadata_update **update_tail;
698
699 /* extra stuff used by mdmon */
700 struct active_array *arrays;
701 int sock; /* listen to external programs */
702 int devnum;
703 char *devname; /* e.g. md0. This appears in metadata_verison:
704 * external:/md0/12
705 */
706 int devcnt;
707
708 struct mdinfo *devs;
709
710 };
711
712 extern struct supertype *super_by_fd(int fd);
713 extern struct supertype *guess_super(int fd);
714 extern struct supertype *dup_super(struct supertype *st);
715 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
716 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
717 mdu_disk_info_t *disk);
718 void wait_for(char *dev, int fd);
719
720 #if __GNUC__ < 3
721 struct stat64;
722 #endif
723
724 #define HAVE_NFTW we assume
725 #define HAVE_FTW
726
727 #ifdef __UCLIBC__
728 # include <features.h>
729 # ifndef __UCLIBC_HAS_LFS__
730 # define lseek64 lseek
731 # endif
732 # ifndef __UCLIBC_HAS_FTW__
733 # undef HAVE_FTW
734 # undef HAVE_NFTW
735 # endif
736 #endif
737
738 #ifdef __dietlibc__
739 # undef HAVE_NFTW
740 #endif
741
742 #if defined(__KLIBC__)
743 # undef HAVE_NFTW
744 # undef HAVE_FTW
745 #endif
746
747 #ifndef HAVE_NFTW
748 # define FTW_PHYS 1
749 # ifndef HAVE_FTW
750 struct FTW {};
751 # endif
752 #endif
753
754 #ifdef HAVE_FTW
755 # include <ftw.h>
756 #endif
757
758 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
759
760
761 extern int Manage_ro(char *devname, int fd, int readonly);
762 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
763 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
764 extern int Manage_subdevs(char *devname, int fd,
765 mddev_dev_t devlist, int verbose);
766 extern int autodetect(void);
767 extern int Grow_Add_device(char *devname, int fd, char *newdev);
768 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
769 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
770 long long size,
771 int level, char *layout_str, int chunksize, int raid_disks);
772 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
773 int *fdlist, int cnt, char *backup_file, int verbose);
774 extern int Grow_continue(int mdfd, struct supertype *st,
775 struct mdinfo *info, char *backup_file);
776
777 extern int Assemble(struct supertype *st, char *mddev,
778 mddev_ident_t ident,
779 mddev_dev_t devlist, char *backup_file,
780 int readonly, int runstop,
781 char *update, char *homehost, int require_homehost,
782 int verbose, int force);
783
784 extern int Build(char *mddev, int chunk, int level, int layout,
785 int raiddisks, mddev_dev_t devlist, int assume_clean,
786 char *bitmap_file, int bitmap_chunk, int write_behind,
787 int delay, int verbose, int autof, unsigned long long size);
788
789
790 extern int Create(struct supertype *st, char *mddev,
791 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
792 char *name, char *homehost, int *uuid,
793 int subdevs, mddev_dev_t devlist,
794 int runstop, int verbose, int force, int assume_clean,
795 char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
796
797 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
798 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
799 extern int Query(char *dev);
800 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
801 int SparcAdjust, struct supertype *forcest, char *homehost);
802 extern int Monitor(mddev_dev_t devlist,
803 char *mailaddr, char *alert_cmd,
804 int period, int daemonise, int scan, int oneshot,
805 int dosyslog, int test, char *pidfile, int increments);
806
807 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
808 extern int Wait(char *dev);
809 extern int WaitClean(char *dev, int sock, int verbose);
810
811 extern int Incremental(char *devname, int verbose, int runstop,
812 struct supertype *st, char *homehost, int require_homehost,
813 int autof);
814 extern int Incremental_container(struct supertype *st, char *devname,
815 int verbose, int runstop, int autof,
816 int trustworthy);
817 extern void RebuildMap(void);
818 extern int IncrementalScan(int verbose);
819
820 extern int CreateBitmap(char *filename, int force, char uuid[16],
821 unsigned long chunksize, unsigned long daemon_sleep,
822 unsigned long write_behind,
823 unsigned long long array_size,
824 int major);
825 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
826 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
827 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
828
829 extern int md_get_version(int fd);
830 extern int get_linux_version(void);
831 extern long long parse_size(char *size);
832 extern int parse_uuid(char *str, int uuid[4]);
833 extern int parse_layout_10(char *layout);
834 extern int parse_layout_faulty(char *layout);
835 extern int check_ext2(int fd, char *name);
836 extern int check_reiser(int fd, char *name);
837 extern int check_raid(int fd, char *name);
838 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
839
840 extern int get_mdp_major(void);
841 extern int dev_open(char *dev, int flags);
842 extern int open_dev(int devnum);
843 extern int open_dev_excl(int devnum);
844 extern int is_standard(char *dev, int *nump);
845 extern int same_dev(char *one, char *two);
846
847 extern int parse_auto(char *str, char *msg, int config);
848 extern mddev_ident_t conf_get_ident(char *dev);
849 extern mddev_dev_t conf_get_devs(void);
850 extern int conf_test_dev(char *devname);
851 extern int conf_test_metadata(const char *version, int is_homehost);
852 extern struct createinfo *conf_get_create_info(void);
853 extern void set_conffile(char *file);
854 extern char *conf_get_mailaddr(void);
855 extern char *conf_get_mailfrom(void);
856 extern char *conf_get_program(void);
857 extern char *conf_get_homehost(int *require_homehostp);
858 extern char *conf_line(FILE *file);
859 extern char *conf_word(FILE *file, int allow_key);
860 extern int conf_name_is_free(char *name);
861 extern int devname_matches(char *name, char *match);
862 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
863
864 extern void free_line(char *line);
865 extern int match_oneof(char *devices, char *devname);
866 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
867 extern const int uuid_match_any[4];
868 extern int same_uuid(int a[4], int b[4], int swapuuid);
869 extern void copy_uuid(void *a, int b[4], int swapuuid);
870 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
871 extern char *fname_from_uuid(struct supertype *st,
872 struct mdinfo *info, char *buf, char sep);
873 extern unsigned long calc_csum(void *super, int bytes);
874 extern int enough(int level, int raid_disks, int layout, int clean,
875 char *avail, int avail_disks);
876 extern int ask(char *mesg);
877 extern unsigned long long get_component_size(int fd);
878 extern void remove_partitions(int fd);
879 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
880 int chunksize, unsigned long long devsize);
881 extern int flush_metadata_updates(struct supertype *st);
882 extern void append_metadata_update(struct supertype *st, void *buf, int len);
883 extern int assemble_container_content(struct supertype *st, int mdfd,
884 struct mdinfo *content, int runstop,
885 char *chosen_name, int verbose);
886
887 extern int add_disk(int mdfd, struct supertype *st,
888 struct mdinfo *sra, struct mdinfo *info);
889 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
890 unsigned long long min_recovery_start(struct mdinfo *array);
891
892 extern char *human_size(long long bytes);
893 extern char *human_size_brief(long long bytes);
894 extern void print_r10_layout(int layout);
895
896 #define NoMdDev (1<<23)
897 extern int find_free_devnum(int use_partitions);
898
899 extern void put_md_name(char *name);
900 extern char *get_md_name(int dev);
901
902 extern char DefaultConfFile[];
903
904 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
905 char *chosen);
906 /* values for 'trustworthy' */
907 #define LOCAL 1
908 #define LOCAL_ANY 10
909 #define FOREIGN 2
910 #define METADATA 3
911 extern int open_mddev(char *dev, int report_errors);
912 extern int open_container(int fd);
913
914 extern char *pid_dir;
915 extern int mdmon_running(int devnum);
916 extern int mdmon_pid(int devnum);
917 extern int check_env(char *name);
918 extern __u32 random32(void);
919 extern int start_mdmon(int devnum);
920
921 extern char *devnum2devname(int num);
922 extern int devname2devnum(char *name);
923 extern int stat2devnum(struct stat *st);
924 extern int fd2devnum(int fd);
925
926 static inline int dev2major(int d)
927 {
928 if (d >= 0)
929 return MD_MAJOR;
930 else
931 return get_mdp_major();
932 }
933
934 static inline int dev2minor(int d)
935 {
936 if (d >= 0)
937 return d;
938 return (-1-d) << MdpMinorShift;
939 }
940
941 static inline int ROUND_UP(int a, int base)
942 {
943 return ((a+base-1)/base)*base;
944 }
945
946 static inline int is_subarray(char *vers)
947 {
948 /* The version string for a 'subarray' (an array in a container)
949 * is
950 * /containername/componentname for normal read-write arrays
951 * -containername/componentname for read-only arrays.
952 * containername is e.g. md0, md_d1
953 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
954 */
955 return (*vers == '/' || *vers == '-');
956 }
957
958 #ifdef DEBUG
959 #define dprintf(fmt, arg...) \
960 fprintf(stderr, fmt, ##arg)
961 #else
962 #define dprintf(fmt, arg...) \
963 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
964 #endif
965 #include <assert.h>
966 #include <stdarg.h>
967 static inline int xasprintf(char **strp, const char *fmt, ...) {
968 va_list ap;
969 int ret;
970 va_start(ap, fmt);
971 ret = vasprintf(strp, fmt, ap);
972 va_end(ap);
973 assert(ret >= 0);
974 return ret;
975 }
976
977 #define LEVEL_MULTIPATH (-4)
978 #define LEVEL_LINEAR (-1)
979 #define LEVEL_FAULTY (-5)
980
981 /* kernel module doesn't know about these */
982 #define LEVEL_CONTAINER (-100)
983 #define LEVEL_UNSUPPORTED (-200)
984
985
986 /* faulty stuff */
987
988 #define WriteTransient 0
989 #define ReadTransient 1
990 #define WritePersistent 2
991 #define ReadPersistent 3
992 #define WriteAll 4 /* doesn't go to device */
993 #define ReadFixable 5
994 #define Modes 6
995
996 #define ClearErrors 31
997 #define ClearFaults 30
998
999 #define AllPersist 100 /* internal use only */
1000 #define NoPersist 101
1001
1002 #define ModeMask 0x1f
1003 #define ModeShift 5
1004
1005
1006 #ifdef __TINYC__
1007 #undef minor
1008 #undef major
1009 #undef makedev
1010 #define minor(x) ((x)&0xff)
1011 #define major(x) (((x)>>8)&0xff)
1012 #define makedev(M,m) (((M)<<8) | (m))
1013 #endif
1014
1015 /* for raid4/5/6 */
1016 #define ALGORITHM_LEFT_ASYMMETRIC 0
1017 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1018 #define ALGORITHM_LEFT_SYMMETRIC 2
1019 #define ALGORITHM_RIGHT_SYMMETRIC 3
1020
1021 /* Define non-rotating (raid4) algorithms. These allow
1022 * conversion of raid4 to raid5.
1023 */
1024 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1025 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1026
1027 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1028 * Firstly, the exact positioning of the parity block is slightly
1029 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1030 * of DDF.
1031 * Secondly, or order of datablocks over which the Q syndrome is computed
1032 * is different.
1033 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1034 * These layouts are from the DDFv1.2 spec.
1035 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1036 * leaves RLQ=3 as 'Vendor Specific'
1037 */
1038
1039 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1040 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1041 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1042
1043
1044 /* For every RAID5 algorithm we define a RAID6 algorithm
1045 * with exactly the same layout for data and parity, and
1046 * with the Q block always on the last device (N-1).
1047 * This allows trivial conversion from RAID5 to RAID6
1048 */
1049 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1050 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1051 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1052 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1053 #define ALGORITHM_PARITY_0_6 20
1054 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1055
1056 /* Define PATH_MAX in case we don't use glibc or standard library does
1057 * not have PATH_MAX defined. Assume max path length is 4K characters.
1058 */
1059 #ifndef PATH_MAX
1060 #define PATH_MAX 4096
1061 #endif
1062