]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Incr: don't exclude 'active' devices from auto inclusion in a container.
[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 /* MAP_DIR should be somewhere that persists across the pivotroot
72 * from early boot to late boot.
73 * Currently /dev seems to be the only option on most distros.
74 */
75 #ifndef MAP_DIR
76 #define MAP_DIR "/dev/.mdadm"
77 #endif /* MAP_DIR */
78 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
79 * want something other than the default of "map"
80 */
81 #ifndef MAP_FILE
82 #define MAP_FILE "map"
83 #endif /* MAP_FILE */
84 /* MDMON_DIR is where pid and socket files used for communicating
85 * with mdmon normally live. It *should* be /var/run, but when
86 * mdmon is needed at early boot then it needs to write there prior
87 * to /var/run being mounted read/write, and it also then needs to
88 * persist beyond when /var/run is mounter read-only. So, to be
89 * safe, the default is somewhere that is read/write early in the
90 * boot process and stays up as long as possible during shutdown.
91 */
92 #ifndef MDMON_DIR
93 #define MDMON_DIR "/dev/.mdadm/"
94 #endif /* MDMON_DIR */
95
96 #include "md_u.h"
97 #include "md_p.h"
98 #include "bitmap.h"
99 #include "msg.h"
100
101 #include <endian.h>
102 /* Redhat don't like to #include <asm/byteorder.h>, and
103 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
104 * and there is no standard conversion function so... */
105 /* And dietlibc doesn't think byteswap is ok, so.. */
106 /* #include <byteswap.h> */
107 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
108 ((x) & 0xff00U) >> 8)
109 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
110 ((x) & 0xff000000U) >> 24 | \
111 ((x) & 0x0000ff00U) << 8 | \
112 ((x) & 0x00ff0000U) >> 8)
113 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
114 ((x) & 0xff00000000000000ULL) >> 56 | \
115 ((x) & 0x000000000000ff00ULL) << 40 | \
116 ((x) & 0x00ff000000000000ULL) >> 40 | \
117 ((x) & 0x0000000000ff0000ULL) << 24 | \
118 ((x) & 0x0000ff0000000000ULL) >> 24 | \
119 ((x) & 0x00000000ff000000ULL) << 8 | \
120 ((x) & 0x000000ff00000000ULL) >> 8)
121
122 #if !defined(__KLIBC__)
123 #if BYTE_ORDER == LITTLE_ENDIAN
124 #define __cpu_to_le16(_x) (_x)
125 #define __cpu_to_le32(_x) (_x)
126 #define __cpu_to_le64(_x) (_x)
127 #define __le16_to_cpu(_x) (_x)
128 #define __le32_to_cpu(_x) (_x)
129 #define __le64_to_cpu(_x) (_x)
130
131 #define __cpu_to_be16(_x) bswap_16(_x)
132 #define __cpu_to_be32(_x) bswap_32(_x)
133 #define __cpu_to_be64(_x) bswap_64(_x)
134 #define __be16_to_cpu(_x) bswap_16(_x)
135 #define __be32_to_cpu(_x) bswap_32(_x)
136 #define __be64_to_cpu(_x) bswap_64(_x)
137 #elif BYTE_ORDER == BIG_ENDIAN
138 #define __cpu_to_le16(_x) bswap_16(_x)
139 #define __cpu_to_le32(_x) bswap_32(_x)
140 #define __cpu_to_le64(_x) bswap_64(_x)
141 #define __le16_to_cpu(_x) bswap_16(_x)
142 #define __le32_to_cpu(_x) bswap_32(_x)
143 #define __le64_to_cpu(_x) bswap_64(_x)
144
145 #define __cpu_to_be16(_x) (_x)
146 #define __cpu_to_be32(_x) (_x)
147 #define __cpu_to_be64(_x) (_x)
148 #define __be16_to_cpu(_x) (_x)
149 #define __be32_to_cpu(_x) (_x)
150 #define __be64_to_cpu(_x) (_x)
151 #else
152 # error "unknown endianness."
153 #endif
154 #endif /* __KLIBC__ */
155
156
157 /*
158 * min()/max()/clamp() macros that also do
159 * strict type-checking.. See the
160 * "unnecessary" pointer comparison.
161 */
162 #define min(x, y) ({ \
163 typeof(x) _min1 = (x); \
164 typeof(y) _min2 = (y); \
165 (void) (&_min1 == &_min2); \
166 _min1 < _min2 ? _min1 : _min2; })
167
168 #define max(x, y) ({ \
169 typeof(x) _max1 = (x); \
170 typeof(y) _max2 = (y); \
171 (void) (&_max1 == &_max2); \
172 _max1 > _max2 ? _max1 : _max2; })
173
174 /* general information that might be extracted from a superblock */
175 struct mdinfo {
176 mdu_array_info_t array;
177 mdu_disk_info_t disk;
178 __u64 events;
179 int uuid[4];
180 char name[33];
181 unsigned long long data_offset;
182 unsigned long long component_size; /* same as array.size, except in
183 * sectors and up to 64bits.
184 */
185 unsigned long long custom_array_size; /* size for non-default sized
186 * arrays (in sectors)
187 */
188 int reshape_active;
189 unsigned long long reshape_progress;
190 union {
191 unsigned long long resync_start; /* per-array resync position */
192 unsigned long long recovery_start; /* per-device rebuild position */
193 #define MaxSector (~0ULL) /* resync/recovery complete position */
194 };
195 unsigned long safe_mode_delay; /* ms delay to mark clean */
196 int new_level, delta_disks, new_layout, new_chunk;
197 int errors;
198 unsigned long cache_size; /* size of raid456 stripe cache*/
199 int mismatch_cnt;
200 char text_version[50];
201 void *update_private; /* for passing metadata-format
202 * specific update data
203 * between successive calls to
204 * update_super()
205 */
206
207 int container_member; /* for assembling external-metatdata arrays
208 * This is to be used internally by metadata
209 * handler only */
210 int container_enough; /* flag external handlers can set to
211 * indicate that subarrays have not enough (-1),
212 * enough to start (0), or all expected disks (1) */
213 char sys_name[20];
214 struct mdinfo *devs;
215 struct mdinfo *next;
216
217 /* Device info for mdmon: */
218 int recovery_fd;
219 int state_fd;
220 #define DS_FAULTY 1
221 #define DS_INSYNC 2
222 #define DS_WRITE_MOSTLY 4
223 #define DS_SPARE 8
224 #define DS_BLOCKED 16
225 #define DS_REMOVE 1024
226 #define DS_UNBLOCK 2048
227 int prev_state, curr_state, next_state;
228
229 };
230
231 struct createinfo {
232 int uid;
233 int gid;
234 int autof;
235 int mode;
236 int symlinks;
237 struct supertype *supertype;
238 };
239
240 #define Name "mdadm"
241
242 enum mode {
243 ASSEMBLE=1,
244 BUILD,
245 CREATE,
246 MANAGE,
247 MISC,
248 MONITOR,
249 GROW,
250 INCREMENTAL,
251 AUTODETECT,
252 };
253
254 extern char short_options[];
255 extern char short_bitmap_options[];
256 extern char short_bitmap_auto_options[];
257 extern struct option long_options[];
258 extern char Version[], Usage[], Help[], OptionHelp[],
259 Help_create[], Help_build[], Help_assemble[], Help_grow[],
260 Help_incr[],
261 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
262
263 /* for option that don't have short equivilents, we assign arbitrary
264 * small numbers. '1' means an undecorated option, so we start at '2'.
265 * (note we must stop before we get to 65 i.e. 'A')
266 */
267 enum special_options {
268 AssumeClean = 2,
269 BitmapChunk,
270 WriteBehind,
271 ReAdd,
272 NoDegraded,
273 Sparc22,
274 BackupFile, /* 8 */
275 HomeHost,
276 AutoHomeHost,
277 Symlinks,
278 AutoDetect,
279 Waitclean,
280 DetailPlatform,
281 KillSubarray,
282 UpdateSubarray, /* 16 */
283 };
284
285 /* structures read from config file */
286 /* List of mddevice names and identifiers
287 * Identifiers can be:
288 * uuid=128-hex-uuid
289 * super-minor=decimal-minor-number-from-superblock
290 * devices=comma,separated,list,of,device,names,with,wildcards
291 *
292 * If multiple fields are present, the intersection of all matching
293 * devices is considered
294 */
295 #define UnSet (0xfffe)
296 typedef struct mddev_ident_s {
297 char *devname;
298
299 int uuid_set;
300 int uuid[4];
301 char name[33];
302
303 int super_minor;
304
305 char *devices; /* comma separated list of device
306 * names with wild cards
307 */
308 int level;
309 int raid_disks;
310 int spare_disks;
311 struct supertype *st;
312 int autof; /* 1 for normal, 2 for partitioned */
313 char *spare_group;
314 char *bitmap_file;
315 int bitmap_fd;
316
317 char *container; /* /dev/whatever name of container, or
318 * uuid of container. You would expect
319 * this to be the 'devname' or UUID
320 * of some other entry.
321 */
322 char *member; /* subarray within a container */
323
324 struct mddev_ident_s *next;
325 union {
326 /* fields needed by different users of this structure */
327 int assembled; /* set when assembly succeeds */
328 };
329 } *mddev_ident_t;
330
331 /* List of device names - wildcards expanded */
332 typedef struct mddev_dev_s {
333 char *devname;
334 char disposition; /* 'a' for add, 'r' for remove, 'f' for fail.
335 * Not set for names read from .config
336 */
337 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
338 char re_add;
339 char used; /* set when used */
340 struct mdinfo *content; /* If devname is a container, this might list
341 * the remaining member arrays. */
342 struct mddev_dev_s *next;
343 } *mddev_dev_t;
344
345 typedef struct mapping {
346 char *name;
347 int num;
348 } mapping_t;
349
350
351 struct mdstat_ent {
352 char *dev;
353 int devnum;
354 int active;
355 char *level;
356 char *pattern; /* U or up, _ for down */
357 int percent; /* -1 if no resync */
358 int resync; /* 1 if resync, 0 if recovery */
359 int devcnt;
360 int raid_disks;
361 char * metadata_version;
362 struct dev_member {
363 char *name;
364 struct dev_member *next;
365 } *members;
366 struct mdstat_ent *next;
367 };
368
369 extern struct mdstat_ent *mdstat_read(int hold, int start);
370 extern void free_mdstat(struct mdstat_ent *ms);
371 extern void mdstat_wait(int seconds);
372 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
373 extern int mddev_busy(int devnum);
374 extern struct mdstat_ent *mdstat_by_component(char *name);
375
376 struct map_ent {
377 struct map_ent *next;
378 int devnum;
379 char metadata[20];
380 int uuid[4];
381 int bad;
382 char *path;
383 };
384 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
385 int uuid[4], char *path);
386 extern void map_remove(struct map_ent **map, int devnum);
387 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
388 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
389 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
390 extern void map_read(struct map_ent **melp);
391 extern int map_write(struct map_ent *mel);
392 extern void map_delete(struct map_ent **mapp, int devnum);
393 extern void map_free(struct map_ent *map);
394 extern void map_add(struct map_ent **melp,
395 int devnum, char *metadata, int uuid[4], char *path);
396 extern int map_lock(struct map_ent **melp);
397 extern void map_unlock(struct map_ent **melp);
398
399 /* various details can be requested */
400 enum sysfs_read_flags {
401 GET_LEVEL = (1 << 0),
402 GET_LAYOUT = (1 << 1),
403 GET_COMPONENT = (1 << 2),
404 GET_CHUNK = (1 << 3),
405 GET_CACHE = (1 << 4),
406 GET_MISMATCH = (1 << 5),
407 GET_VERSION = (1 << 6),
408 GET_DISKS = (1 << 7),
409 GET_DEGRADED = (1 << 8),
410 GET_SAFEMODE = (1 << 9),
411 GET_DEVS = (1 << 10), /* gets role, major, minor */
412 GET_OFFSET = (1 << 11),
413 GET_SIZE = (1 << 12),
414 GET_STATE = (1 << 13),
415 GET_ERROR = (1 << 14),
416 };
417
418 /* If fd >= 0, get the array it is open on,
419 * else use devnum. >=0 -> major9. <0.....
420 */
421 extern int sysfs_open(int devnum, char *devname, char *attr);
422 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
423 extern void sysfs_free(struct mdinfo *sra);
424 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
425 extern int sysfs_attr_match(const char *attr, const char *str);
426 extern int sysfs_match_word(const char *word, char **list);
427 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
428 char *name, char *val);
429 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
430 char *name, unsigned long long val);
431 extern int sysfs_uevent(struct mdinfo *sra, char *event);
432 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
433 char *name);
434 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
435 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
436 char *name, unsigned long long *val);
437 extern int sysfs_fd_get_str(int fd, char *val, int size);
438 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
439 char *name, char *val, int size);
440 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
441 extern int sysfs_set_array(struct mdinfo *info, int vers);
442 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
443 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
444 extern int sysfs_unique_holder(int devnum, long rdev);
445 extern int load_sys(char *path, char *buf);
446
447
448 extern int save_stripes(int *source, unsigned long long *offsets,
449 int raid_disks, int chunk_size, int level, int layout,
450 int nwrites, int *dest,
451 unsigned long long start, unsigned long long length,
452 char *buf);
453 extern int restore_stripes(int *dest, unsigned long long *offsets,
454 int raid_disks, int chunk_size, int level, int layout,
455 int source, unsigned long long read_offset,
456 unsigned long long start, unsigned long long length);
457
458 #ifndef Sendmail
459 #define Sendmail "/usr/lib/sendmail -t"
460 #endif
461
462 #define SYSLOG_FACILITY LOG_DAEMON
463
464 extern char *map_num(mapping_t *map, int num);
465 extern int map_name(mapping_t *map, char *name);
466 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
467
468 extern char *map_dev(int major, int minor, int create);
469
470 struct active_array;
471 struct metadata_update;
472
473 /* A superswitch provides entry point the a metadata handler.
474 *
475 * The super_switch primarily operates on some "metadata" that
476 * is accessed via the 'supertype'.
477 * This metadata has one of three possible sources.
478 * 1/ It is read from a single device. In this case it may not completely
479 * describe the array or arrays as some information might be on other
480 * devices.
481 * 2/ It is read from all devices in a container. In this case all
482 * information is present.
483 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
484 * be complete once enough ->add_to_super calls have completed.
485 *
486 * When creating an array inside a container, the metadata will be
487 * formed by a combination of 2 and 3. The metadata or the array is read,
488 * then new information is added.
489 *
490 * The metadata must sometimes have a concept of a 'current' array
491 * and a 'current' device.
492 * The 'current' array is set by init_super to be the newly created array,
493 * or is set by super_by_fd when it finds it is looking at an array inside
494 * a container.
495 *
496 * The 'current' device is either the device that the metadata was read from
497 * in case 1, or the last device added by add_to_super in case 3.
498 * Case 2 does not identify a 'current' device.
499 */
500 extern struct superswitch {
501
502 /* Used to report details of metadata read from a component
503 * device. ->load_super has been called.
504 */
505 void (*examine_super)(struct supertype *st, char *homehost);
506 void (*brief_examine_super)(struct supertype *st, int verbose);
507 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
508 void (*export_examine_super)(struct supertype *st);
509
510 /* Used to report details of an active array.
511 * ->load_super was possibly given a 'component' string.
512 */
513 void (*detail_super)(struct supertype *st, char *homehost);
514 void (*brief_detail_super)(struct supertype *st);
515 void (*export_detail_super)(struct supertype *st);
516
517 /* Optional: platform hardware / firmware details */
518 int (*detail_platform)(int verbose, int enumerate_only);
519
520 /* Used:
521 * to get uuid to storing in bitmap metadata
522 * and 'reshape' backup-data metadata
523 * To see if a device is being re-added to an array it was part of.
524 */
525 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
526
527 /* Extract generic details from metadata. This could be details about
528 * the container, or about an individual array within the container.
529 * The determination is made either by:
530 * load_super being given a 'component' string.
531 * validate_geometry determining what to create.
532 * The info includes both array information and device information.
533 * The particular device should be:
534 * The last device added by add_to_super
535 * The device the metadata was loaded from by load_super
536 */
537 void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
538
539 /* Check if the given metadata is flagged as belonging to "this"
540 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
541 */
542 int (*match_home)(struct supertype *st, char *homehost);
543
544 /* Make one of several generic modifications to metadata
545 * prior to assembly (or other times).
546 * sparc2.2 - first bug in early 0.90 metadata
547 * super-minor - change name of 0.90 metadata
548 * summaries - 'correct' any redundant data
549 * resync - mark array as dirty to trigger a resync.
550 * uuid - set new uuid - only 0.90 or 1.x
551 * name - change the name of the array (where supported)
552 * homehost - change which host this array is tied to.
553 * devicesize - If metadata is at start of device, change recorded
554 * device size to match actual device size
555 * byteorder - swap bytes for 0.90 metadata
556 *
557 * force-one - mark that device as uptodate, not old or failed.
558 * force-array - mark array as clean if it would not otherwise
559 * assemble
560 * assemble - not sure how this is different from force-one...
561 * linear-grow-new - add a new device to a linear array, but don't
562 * change the size: so superblock still matches
563 * linear-grow-update - now change the size of the array.
564 */
565 int (*update_super)(struct supertype *st, struct mdinfo *info,
566 char *update,
567 char *devname, int verbose,
568 int uuid_set, char *homehost);
569
570 /* Create new metadata for new array as described. This could
571 * be a new container, or an array in a pre-existing container.
572 * Also used to zero metadata prior to writing it to invalidate old
573 * metadata.
574 */
575 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
576 unsigned long long size, char *name,
577 char *homehost, int *uuid);
578
579 /* update the metadata to include new device, either at create or
580 * when hot-adding a spare.
581 */
582 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
583 int fd, char *devname);
584 /* update the metadata to delete a device,
585 * when hot-removing.
586 */
587 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
588
589 /* Write metadata to one device when fixing problems or adding
590 * a new device.
591 */
592 int (*store_super)(struct supertype *st, int fd);
593
594 /* Write all metadata for this array.
595 */
596 int (*write_init_super)(struct supertype *st);
597 int (*compare_super)(struct supertype *st, struct supertype *tst);
598 int (*load_super)(struct supertype *st, int fd, char *devname);
599 struct supertype * (*match_metadata_desc)(char *arg);
600 __u64 (*avail_size)(struct supertype *st, __u64 size);
601 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
602 int delay, int write_behind,
603 unsigned long long size, int may_change, int major);
604 void (*locate_bitmap)(struct supertype *st, int fd);
605 int (*write_bitmap)(struct supertype *st, int fd);
606 void (*free_super)(struct supertype *st);
607
608 /* validate_geometry is called with an st returned by
609 * match_metadata_desc.
610 * It should check that the geometry described in compatible with
611 * the metadata type. It will be called repeatedly as devices
612 * added to validate changing size and new devices. If there are
613 * inter-device dependencies, it should record sufficient details
614 * so these can be validated.
615 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
616 */
617 int (*validate_geometry)(struct supertype *st, int level, int layout,
618 int raiddisks,
619 int chunk, unsigned long long size,
620 char *subdev, unsigned long long *freesize,
621 int verbose);
622
623 struct mdinfo *(*container_content)(struct supertype *st);
624 /* Allow a metadata handler to override mdadm's default layouts */
625 int (*default_layout)(int level); /* optional */
626 /* query the supertype for default chunk size */
627 int (*default_chunk)(struct supertype *st); /* optional */
628 /* Permit subarray's to be deleted from inactive containers */
629 int (*kill_subarray)(struct supertype *st); /* optional */
630 /* Permit subarray's to be modified */
631 int (*update_subarray)(struct supertype *st, char *update, mddev_ident_t ident); /* optional */
632
633 /* for mdmon */
634 int (*open_new)(struct supertype *c, struct active_array *a,
635 char *inst);
636
637 /* Tell the metadata handler the current state of the array.
638 * This covers whether it is known to be consistent (no pending writes)
639 * and how far along a resync is known to have progressed
640 * (in a->resync_start).
641 * resync status is really irrelevant if the array is not consistent,
642 * but some metadata (DDF!) have a place to record the distinction.
643 * If 'consistent' is '2', then the array can mark it dirty if a
644 * resync/recovery/whatever is required, or leave it clean if not.
645 * Return value is 0 dirty (not consistent) and 1 if clean.
646 * it is only really important if consistent is passed in as '2'.
647 */
648 int (*set_array_state)(struct active_array *a, int consistent);
649
650 /* When the state of a device might have changed, we call set_disk to
651 * tell the metadata what the current state is.
652 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
653 * transitions.
654 * set_disk might be called when the state of the particular disk has
655 * not in fact changed.
656 */
657 void (*set_disk)(struct active_array *a, int n, int state);
658 void (*sync_metadata)(struct supertype *st);
659 void (*process_update)(struct supertype *st,
660 struct metadata_update *update);
661 void (*prepare_update)(struct supertype *st,
662 struct metadata_update *update);
663
664 /* activate_spare will check if the array is degraded and, if it
665 * is, try to find some spare space in the container.
666 * On success, it add appropriate updates (For process_update) to
667 * to the 'updates' list and returns a list of 'mdinfo' identifying
668 * the device, or devices as there might be multiple missing
669 * devices and multiple spares available.
670 */
671 struct mdinfo *(*activate_spare)(struct active_array *a,
672 struct metadata_update **updates);
673
674 int swapuuid; /* true if uuid is bigending rather than hostendian */
675 int external;
676 const char *name; /* canonical metadata name */
677 } super0, super1, super_ddf, *superlist[];
678
679 extern struct superswitch super_imsm;
680
681 struct metadata_update {
682 int len;
683 char *buf;
684 void *space; /* allocated space that monitor will use */
685 struct metadata_update *next;
686 };
687
688 /* A supertype holds a particular collection of metadata.
689 * It identifies the metadata type by the superswitch, and the particular
690 * sub-version of that metadata type.
691 * metadata read in or created is stored in 'sb' and 'info'.
692 * There are also fields used by mdmon to track containers.
693 *
694 * A supertype may refer to:
695 * Just an array, possibly in a container
696 * A container, not identifying any particular array
697 * Info read from just one device, not yet fully describing the array/container.
698 *
699 *
700 * A supertype is created by:
701 * super_by_fd
702 * guess_super
703 * dup_super
704 */
705 struct supertype {
706 struct superswitch *ss;
707 int minor_version;
708 int max_devs;
709 int container_dev; /* devnum of container */
710 char subarray[32]; /* name of array inside container */
711 void *sb;
712 void *info;
713 int loaded_container; /* Set if load_super found a container,
714 * not just one device */
715
716 struct metadata_update *updates;
717 struct metadata_update **update_tail;
718
719 /* extra stuff used by mdmon */
720 struct active_array *arrays;
721 int sock; /* listen to external programs */
722 int devnum;
723 char *devname; /* e.g. md0. This appears in metadata_verison:
724 * external:/md0/12
725 */
726 int devcnt;
727
728 struct mdinfo *devs;
729
730 };
731
732 extern struct supertype *super_by_fd(int fd);
733 extern struct supertype *guess_super(int fd);
734 extern struct supertype *dup_super(struct supertype *st);
735 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
736 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
737 mdu_disk_info_t *disk);
738 void wait_for(char *dev, int fd);
739
740 #if __GNUC__ < 3
741 struct stat64;
742 #endif
743
744 #define HAVE_NFTW we assume
745 #define HAVE_FTW
746
747 #ifdef __UCLIBC__
748 # include <features.h>
749 # ifndef __UCLIBC_HAS_LFS__
750 # define lseek64 lseek
751 # endif
752 # ifndef __UCLIBC_HAS_FTW__
753 # undef HAVE_FTW
754 # undef HAVE_NFTW
755 # endif
756 #endif
757
758 #ifdef __dietlibc__
759 # undef HAVE_NFTW
760 #endif
761
762 #if defined(__KLIBC__)
763 # undef HAVE_NFTW
764 # undef HAVE_FTW
765 #endif
766
767 #ifndef HAVE_NFTW
768 # define FTW_PHYS 1
769 # ifndef HAVE_FTW
770 struct FTW {};
771 # endif
772 #endif
773
774 #ifdef HAVE_FTW
775 # include <ftw.h>
776 #endif
777
778 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
779
780
781 extern int Manage_ro(char *devname, int fd, int readonly);
782 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
783 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
784 extern int Manage_subdevs(char *devname, int fd,
785 mddev_dev_t devlist, int verbose, int test);
786 extern int autodetect(void);
787 extern int Grow_Add_device(char *devname, int fd, char *newdev);
788 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
789 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
790 long long size,
791 int level, char *layout_str, int chunksize, int raid_disks);
792 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
793 int *fdlist, int cnt, char *backup_file, int verbose);
794 extern int Grow_continue(int mdfd, struct supertype *st,
795 struct mdinfo *info, char *backup_file);
796
797 extern int Assemble(struct supertype *st, char *mddev,
798 mddev_ident_t ident,
799 mddev_dev_t devlist, char *backup_file,
800 int readonly, int runstop,
801 char *update, char *homehost, int require_homehost,
802 int verbose, int force);
803
804 extern int Build(char *mddev, int chunk, int level, int layout,
805 int raiddisks, mddev_dev_t devlist, int assume_clean,
806 char *bitmap_file, int bitmap_chunk, int write_behind,
807 int delay, int verbose, int autof, unsigned long long size);
808
809
810 extern int Create(struct supertype *st, char *mddev,
811 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
812 char *name, char *homehost, int *uuid,
813 int subdevs, mddev_dev_t devlist,
814 int runstop, int verbose, int force, int assume_clean,
815 char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
816
817 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
818 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
819 extern int Query(char *dev);
820 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
821 int SparcAdjust, struct supertype *forcest, char *homehost);
822 extern int Monitor(mddev_dev_t devlist,
823 char *mailaddr, char *alert_cmd,
824 int period, int daemonise, int scan, int oneshot,
825 int dosyslog, int test, char *pidfile, int increments);
826
827 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
828 extern int Kill_subarray(char *dev, char *subarray, int quiet);
829 extern int Update_subarray(char *dev, char *subarray, char *update, mddev_ident_t ident, int quiet);
830 extern int Wait(char *dev);
831 extern int WaitClean(char *dev, int sock, int verbose);
832
833 extern int Incremental(char *devname, int verbose, int runstop,
834 struct supertype *st, char *homehost, int require_homehost,
835 int autof);
836 extern int Incremental_container(struct supertype *st, char *devname,
837 int verbose, int runstop, int autof,
838 int trustworthy);
839 extern void RebuildMap(void);
840 extern int IncrementalScan(int verbose);
841 extern int IncrementalRemove(char *devname, int verbose);
842 extern int CreateBitmap(char *filename, int force, char uuid[16],
843 unsigned long chunksize, unsigned long daemon_sleep,
844 unsigned long write_behind,
845 unsigned long long array_size,
846 int major);
847 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
848 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
849 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
850
851 extern int md_get_version(int fd);
852 extern int get_linux_version(void);
853 extern long long parse_size(char *size);
854 extern int parse_uuid(char *str, int uuid[4]);
855 extern int parse_layout_10(char *layout);
856 extern int parse_layout_faulty(char *layout);
857 extern int check_ext2(int fd, char *name);
858 extern int check_reiser(int fd, char *name);
859 extern int check_raid(int fd, char *name);
860 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
861
862 extern int get_mdp_major(void);
863 extern int dev_open(char *dev, int flags);
864 extern int open_dev(int devnum);
865 extern int open_dev_excl(int devnum);
866 extern int is_standard(char *dev, int *nump);
867 extern int same_dev(char *one, char *two);
868
869 extern int parse_auto(char *str, char *msg, int config);
870 extern mddev_ident_t conf_get_ident(char *dev);
871 extern mddev_dev_t conf_get_devs(void);
872 extern int conf_test_dev(char *devname);
873 extern int conf_test_metadata(const char *version, int is_homehost);
874 extern struct createinfo *conf_get_create_info(void);
875 extern void set_conffile(char *file);
876 extern char *conf_get_mailaddr(void);
877 extern char *conf_get_mailfrom(void);
878 extern char *conf_get_program(void);
879 extern char *conf_get_homehost(int *require_homehostp);
880 extern char *conf_line(FILE *file);
881 extern char *conf_word(FILE *file, int allow_key);
882 extern int conf_name_is_free(char *name);
883 extern int devname_matches(char *name, char *match);
884 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
885
886 extern void free_line(char *line);
887 extern int match_oneof(char *devices, char *devname);
888 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
889 extern const int uuid_match_any[4];
890 extern int same_uuid(int a[4], int b[4], int swapuuid);
891 extern void copy_uuid(void *a, int b[4], int swapuuid);
892 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
893 extern char *fname_from_uuid(struct supertype *st,
894 struct mdinfo *info, char *buf, char sep);
895 extern unsigned long calc_csum(void *super, int bytes);
896 extern int enough(int level, int raid_disks, int layout, int clean,
897 char *avail, int avail_disks);
898 extern int enough_fd(int fd);
899 extern int ask(char *mesg);
900 extern unsigned long long get_component_size(int fd);
901 extern void remove_partitions(int fd);
902 extern int test_partition(int fd);
903 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
904 int chunksize, unsigned long long devsize);
905 extern int flush_metadata_updates(struct supertype *st);
906 extern void append_metadata_update(struct supertype *st, void *buf, int len);
907 extern int assemble_container_content(struct supertype *st, int mdfd,
908 struct mdinfo *content, int runstop,
909 char *chosen_name, int verbose);
910
911 extern int add_disk(int mdfd, struct supertype *st,
912 struct mdinfo *sra, struct mdinfo *info);
913 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
914 unsigned long long min_recovery_start(struct mdinfo *array);
915
916 extern char *human_size(long long bytes);
917 extern char *human_size_brief(long long bytes);
918 extern void print_r10_layout(int layout);
919
920 #define NoMdDev (1<<23)
921 extern int find_free_devnum(int use_partitions);
922
923 extern void put_md_name(char *name);
924 extern char *get_md_name(int dev);
925
926 extern char DefaultConfFile[];
927
928 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
929 char *chosen);
930 /* values for 'trustworthy' */
931 #define LOCAL 1
932 #define LOCAL_ANY 10
933 #define FOREIGN 2
934 #define METADATA 3
935 extern int open_mddev(char *dev, int report_errors);
936 extern int open_container(int fd);
937 extern int is_container_member(struct mdstat_ent *ent, char *devname);
938 extern int is_subarray_active(char *subarray, char *devname);
939 extern int open_subarray(char *dev, struct supertype *st, int quiet);
940 extern struct superswitch *version_to_superswitch(char *vers);
941
942 extern int mdmon_running(int devnum);
943 extern int mdmon_pid(int devnum);
944 extern int check_env(char *name);
945 extern __u32 random32(void);
946 extern int start_mdmon(int devnum);
947
948 extern char *devnum2devname(int num);
949 extern int devname2devnum(char *name);
950 extern int stat2devnum(struct stat *st);
951 extern int fd2devnum(int fd);
952
953 static inline int dev2major(int d)
954 {
955 if (d >= 0)
956 return MD_MAJOR;
957 else
958 return get_mdp_major();
959 }
960
961 static inline int dev2minor(int d)
962 {
963 if (d >= 0)
964 return d;
965 return (-1-d) << MdpMinorShift;
966 }
967
968 static inline int ROUND_UP(int a, int base)
969 {
970 return ((a+base-1)/base)*base;
971 }
972
973 static inline int is_subarray(char *vers)
974 {
975 /* The version string for a 'subarray' (an array in a container)
976 * is
977 * /containername/componentname for normal read-write arrays
978 * -containername/componentname for read-only arrays.
979 * containername is e.g. md0, md_d1
980 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
981 */
982 return (*vers == '/' || *vers == '-');
983 }
984
985 #ifdef DEBUG
986 #define dprintf(fmt, arg...) \
987 fprintf(stderr, fmt, ##arg)
988 #else
989 #define dprintf(fmt, arg...) \
990 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
991 #endif
992 #include <assert.h>
993 #include <stdarg.h>
994 static inline int xasprintf(char **strp, const char *fmt, ...) {
995 va_list ap;
996 int ret;
997 va_start(ap, fmt);
998 ret = vasprintf(strp, fmt, ap);
999 va_end(ap);
1000 assert(ret >= 0);
1001 return ret;
1002 }
1003
1004 #define LEVEL_MULTIPATH (-4)
1005 #define LEVEL_LINEAR (-1)
1006 #define LEVEL_FAULTY (-5)
1007
1008 /* kernel module doesn't know about these */
1009 #define LEVEL_CONTAINER (-100)
1010 #define LEVEL_UNSUPPORTED (-200)
1011
1012
1013 /* faulty stuff */
1014
1015 #define WriteTransient 0
1016 #define ReadTransient 1
1017 #define WritePersistent 2
1018 #define ReadPersistent 3
1019 #define WriteAll 4 /* doesn't go to device */
1020 #define ReadFixable 5
1021 #define Modes 6
1022
1023 #define ClearErrors 31
1024 #define ClearFaults 30
1025
1026 #define AllPersist 100 /* internal use only */
1027 #define NoPersist 101
1028
1029 #define ModeMask 0x1f
1030 #define ModeShift 5
1031
1032
1033 #ifdef __TINYC__
1034 #undef minor
1035 #undef major
1036 #undef makedev
1037 #define minor(x) ((x)&0xff)
1038 #define major(x) (((x)>>8)&0xff)
1039 #define makedev(M,m) (((M)<<8) | (m))
1040 #endif
1041
1042 /* for raid4/5/6 */
1043 #define ALGORITHM_LEFT_ASYMMETRIC 0
1044 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1045 #define ALGORITHM_LEFT_SYMMETRIC 2
1046 #define ALGORITHM_RIGHT_SYMMETRIC 3
1047
1048 /* Define non-rotating (raid4) algorithms. These allow
1049 * conversion of raid4 to raid5.
1050 */
1051 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1052 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1053
1054 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1055 * Firstly, the exact positioning of the parity block is slightly
1056 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1057 * of DDF.
1058 * Secondly, or order of datablocks over which the Q syndrome is computed
1059 * is different.
1060 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1061 * These layouts are from the DDFv1.2 spec.
1062 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1063 * leaves RLQ=3 as 'Vendor Specific'
1064 */
1065
1066 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1067 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1068 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1069
1070
1071 /* For every RAID5 algorithm we define a RAID6 algorithm
1072 * with exactly the same layout for data and parity, and
1073 * with the Q block always on the last device (N-1).
1074 * This allows trivial conversion from RAID5 to RAID6
1075 */
1076 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1077 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1078 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1079 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1080 #define ALGORITHM_PARITY_0_6 20
1081 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1082
1083 /* Define PATH_MAX in case we don't use glibc or standard library does
1084 * not have PATH_MAX defined. Assume max path length is 4K characters.
1085 */
1086 #ifndef PATH_MAX
1087 #define PATH_MAX 4096
1088 #endif
1089