]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
mdmon: allow pid to be stored in different directory.
[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 #define VAR_RUN "/var/run/mdadm"
72 /* ALT_RUN should be somewhere that persists across the pivotroot
73 * from early boot to late boot.
74 * If you don't have /lib/init/rw you might want to use /dev/.something
75 */
76 #ifndef ALT_RUN
77 #define ALT_RUN "/lib/init/rw/mdadm"
78 #endif /* ALT_RUN */
79
80 #include "md_u.h"
81 #include "md_p.h"
82 #include "bitmap.h"
83 #include "msg.h"
84
85 #include <endian.h>
86 /* Redhat don't like to #include <asm/byteorder.h>, and
87 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
88 * and there is no standard conversion function so... */
89 /* And dietlibc doesn't think byteswap is ok, so.. */
90 /* #include <byteswap.h> */
91 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
92 ((x) & 0xff00U) >> 8)
93 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
94 ((x) & 0xff000000U) >> 24 | \
95 ((x) & 0x0000ff00U) << 8 | \
96 ((x) & 0x00ff0000U) >> 8)
97 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
98 ((x) & 0xff00000000000000ULL) >> 56 | \
99 ((x) & 0x000000000000ff00ULL) << 40 | \
100 ((x) & 0x00ff000000000000ULL) >> 40 | \
101 ((x) & 0x0000000000ff0000ULL) << 24 | \
102 ((x) & 0x0000ff0000000000ULL) >> 24 | \
103 ((x) & 0x00000000ff000000ULL) << 8 | \
104 ((x) & 0x000000ff00000000ULL) >> 8)
105
106 #if !defined(__KLIBC__)
107 #if BYTE_ORDER == LITTLE_ENDIAN
108 #define __cpu_to_le16(_x) (_x)
109 #define __cpu_to_le32(_x) (_x)
110 #define __cpu_to_le64(_x) (_x)
111 #define __le16_to_cpu(_x) (_x)
112 #define __le32_to_cpu(_x) (_x)
113 #define __le64_to_cpu(_x) (_x)
114
115 #define __cpu_to_be16(_x) bswap_16(_x)
116 #define __cpu_to_be32(_x) bswap_32(_x)
117 #define __cpu_to_be64(_x) bswap_64(_x)
118 #define __be16_to_cpu(_x) bswap_16(_x)
119 #define __be32_to_cpu(_x) bswap_32(_x)
120 #define __be64_to_cpu(_x) bswap_64(_x)
121 #elif BYTE_ORDER == BIG_ENDIAN
122 #define __cpu_to_le16(_x) bswap_16(_x)
123 #define __cpu_to_le32(_x) bswap_32(_x)
124 #define __cpu_to_le64(_x) bswap_64(_x)
125 #define __le16_to_cpu(_x) bswap_16(_x)
126 #define __le32_to_cpu(_x) bswap_32(_x)
127 #define __le64_to_cpu(_x) bswap_64(_x)
128
129 #define __cpu_to_be16(_x) (_x)
130 #define __cpu_to_be32(_x) (_x)
131 #define __cpu_to_be64(_x) (_x)
132 #define __be16_to_cpu(_x) (_x)
133 #define __be32_to_cpu(_x) (_x)
134 #define __be64_to_cpu(_x) (_x)
135 #else
136 # error "unknown endianness."
137 #endif
138 #endif /* __KLIBC__ */
139
140
141 /*
142 * min()/max()/clamp() macros that also do
143 * strict type-checking.. See the
144 * "unnecessary" pointer comparison.
145 */
146 #define min(x, y) ({ \
147 typeof(x) _min1 = (x); \
148 typeof(y) _min2 = (y); \
149 (void) (&_min1 == &_min2); \
150 _min1 < _min2 ? _min1 : _min2; })
151
152 #define max(x, y) ({ \
153 typeof(x) _max1 = (x); \
154 typeof(y) _max2 = (y); \
155 (void) (&_max1 == &_max2); \
156 _max1 > _max2 ? _max1 : _max2; })
157
158 /* general information that might be extracted from a superblock */
159 struct mdinfo {
160 mdu_array_info_t array;
161 mdu_disk_info_t disk;
162 __u64 events;
163 int uuid[4];
164 char name[33];
165 unsigned long long data_offset;
166 unsigned long long component_size; /* same as array.size, except in
167 * sectors and up to 64bits.
168 */
169 unsigned long long custom_array_size; /* size for non-default sized
170 * arrays (in sectors)
171 */
172 int reshape_active;
173 unsigned long long reshape_progress;
174 union {
175 unsigned long long resync_start; /* per-array resync position */
176 unsigned long long recovery_start; /* per-device rebuild position */
177 #define MaxSector (~0ULL) /* resync/recovery complete position */
178 };
179 unsigned long safe_mode_delay; /* ms delay to mark clean */
180 int new_level, delta_disks, new_layout, new_chunk;
181 int errors;
182 int cache_size; /* size of raid456 stripe cache*/
183 int mismatch_cnt;
184 char text_version[50];
185 void *update_private; /* for passing metadata-format
186 * specific update data
187 * between successive calls to
188 * update_super()
189 */
190
191 int container_member; /* for assembling external-metatdata arrays
192 * This is to be used internally by metadata
193 * handler only */
194
195 char sys_name[20];
196 struct mdinfo *devs;
197 struct mdinfo *next;
198
199 /* Device info for mdmon: */
200 int recovery_fd;
201 int state_fd;
202 #define DS_FAULTY 1
203 #define DS_INSYNC 2
204 #define DS_WRITE_MOSTLY 4
205 #define DS_SPARE 8
206 #define DS_BLOCKED 16
207 #define DS_REMOVE 1024
208 #define DS_UNBLOCK 2048
209 int prev_state, curr_state, next_state;
210
211 };
212
213 struct createinfo {
214 int uid;
215 int gid;
216 int autof;
217 int mode;
218 int symlinks;
219 struct supertype *supertype;
220 };
221
222 #define Name "mdadm"
223
224 enum mode {
225 ASSEMBLE=1,
226 BUILD,
227 CREATE,
228 MANAGE,
229 MISC,
230 MONITOR,
231 GROW,
232 INCREMENTAL,
233 AUTODETECT,
234 };
235
236 extern char short_options[];
237 extern char short_bitmap_options[];
238 extern char short_bitmap_auto_options[];
239 extern struct option long_options[];
240 extern char Version[], Usage[], Help[], OptionHelp[],
241 Help_create[], Help_build[], Help_assemble[], Help_grow[],
242 Help_incr[],
243 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
244
245 /* for option that don't have short equivilents, we assign arbitrary
246 * small numbers. '1' means an undecorated option, so we start at '2'.
247 */
248 enum special_options {
249 AssumeClean = 2,
250 BitmapChunk,
251 WriteBehind,
252 ReAdd,
253 NoDegraded,
254 Sparc22,
255 BackupFile,
256 HomeHost,
257 AutoHomeHost,
258 Symlinks,
259 AutoDetect,
260 Waitclean,
261 DetailPlatform,
262 };
263
264 /* structures read from config file */
265 /* List of mddevice names and identifiers
266 * Identifiers can be:
267 * uuid=128-hex-uuid
268 * super-minor=decimal-minor-number-from-superblock
269 * devices=comma,separated,list,of,device,names,with,wildcards
270 *
271 * If multiple fields are present, the intersection of all matching
272 * devices is considered
273 */
274 #define UnSet (0xfffe)
275 typedef struct mddev_ident_s {
276 char *devname;
277
278 int uuid_set;
279 int uuid[4];
280 char name[33];
281
282 unsigned int super_minor;
283
284 char *devices; /* comma separated list of device
285 * names with wild cards
286 */
287 int level;
288 unsigned int raid_disks;
289 unsigned int spare_disks;
290 struct supertype *st;
291 int autof; /* 1 for normal, 2 for partitioned */
292 char *spare_group;
293 char *bitmap_file;
294 int bitmap_fd;
295
296 char *container; /* /dev/whatever name of container, or
297 * uuid of container. You would expect
298 * this to be the 'devname' or UUID
299 * of some other entry.
300 */
301 char *member; /* subarray within a container */
302
303 struct mddev_ident_s *next;
304 } *mddev_ident_t;
305
306 /* List of device names - wildcards expanded */
307 typedef struct mddev_dev_s {
308 char *devname;
309 char disposition; /* 'a' for add, 'r' for remove, 'f' for fail.
310 * Not set for names read from .config
311 */
312 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
313 char re_add;
314 char used; /* set when used */
315 struct mdinfo *content; /* If devname is a container, this might list
316 * the remaining member arrays. */
317 struct mddev_dev_s *next;
318 } *mddev_dev_t;
319
320 typedef struct mapping {
321 char *name;
322 int num;
323 } mapping_t;
324
325
326 struct mdstat_ent {
327 char *dev;
328 int devnum;
329 int active;
330 char *level;
331 char *pattern; /* U or up, _ for down */
332 int percent; /* -1 if no resync */
333 int resync; /* 1 if resync, 0 if recovery */
334 int devcnt;
335 int raid_disks;
336 int chunk_size;
337 char * metadata_version;
338 struct mdstat_ent *next;
339 };
340
341 extern struct mdstat_ent *mdstat_read(int hold, int start);
342 extern void free_mdstat(struct mdstat_ent *ms);
343 extern void mdstat_wait(int seconds);
344 extern void mdstat_wait_fd(int fd, int fd2, const sigset_t *sigmask);
345 extern int mddev_busy(int devnum);
346
347 struct map_ent {
348 struct map_ent *next;
349 int devnum;
350 char metadata[20];
351 int uuid[4];
352 int bad;
353 char *path;
354 };
355 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
356 int uuid[4], char *path);
357 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
358 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
359 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
360 extern void map_read(struct map_ent **melp);
361 extern int map_write(struct map_ent *mel);
362 extern void map_delete(struct map_ent **mapp, int devnum);
363 extern void map_free(struct map_ent *map);
364 extern void map_add(struct map_ent **melp,
365 int devnum, char *metadata, int uuid[4], char *path);
366 extern int map_lock(struct map_ent **melp);
367 extern void map_unlock(struct map_ent **melp);
368
369 /* various details can be requested */
370 enum sysfs_read_flags {
371 GET_LEVEL = (1 << 0),
372 GET_LAYOUT = (1 << 1),
373 GET_COMPONENT = (1 << 2),
374 GET_CHUNK = (1 << 3),
375 GET_CACHE = (1 << 4),
376 GET_MISMATCH = (1 << 5),
377 GET_VERSION = (1 << 6),
378 GET_DISKS = (1 << 7),
379 GET_DEGRADED = (1 << 8),
380 GET_SAFEMODE = (1 << 9),
381 GET_DEVS = (1 << 10), /* gets role, major, minor */
382 GET_OFFSET = (1 << 11),
383 GET_SIZE = (1 << 12),
384 GET_STATE = (1 << 13),
385 GET_ERROR = (1 << 14),
386 SKIP_GONE_DEVS = (1 << 15),
387 };
388
389 /* If fd >= 0, get the array it is open on,
390 * else use devnum. >=0 -> major9. <0.....
391 */
392 extern int sysfs_open(int devnum, char *devname, char *attr);
393 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
394 extern void sysfs_free(struct mdinfo *sra);
395 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
396 extern int sysfs_attr_match(const char *attr, const char *str);
397 extern int sysfs_match_word(const char *word, char **list);
398 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
399 char *name, char *val);
400 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
401 char *name, unsigned long long val);
402 extern int sysfs_uevent(struct mdinfo *sra, char *event);
403 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
404 char *name);
405 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
406 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
407 char *name, unsigned long long *val);
408 extern int sysfs_fd_get_str(int fd, char *val, int size);
409 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
410 char *name, char *val, int size);
411 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
412 extern int sysfs_set_array(struct mdinfo *info, int vers);
413 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
414 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
415 extern int sysfs_unique_holder(int devnum, long rdev);
416 extern int load_sys(char *path, char *buf);
417
418
419 extern int save_stripes(int *source, unsigned long long *offsets,
420 int raid_disks, int chunk_size, int level, int layout,
421 int nwrites, int *dest,
422 unsigned long long start, unsigned long long length,
423 char *buf);
424 extern int restore_stripes(int *dest, unsigned long long *offsets,
425 int raid_disks, int chunk_size, int level, int layout,
426 int source, unsigned long long read_offset,
427 unsigned long long start, unsigned long long length);
428
429 #ifndef Sendmail
430 #define Sendmail "/usr/lib/sendmail -t"
431 #endif
432
433 #define SYSLOG_FACILITY LOG_DAEMON
434
435 extern char *map_num(mapping_t *map, int num);
436 extern int map_name(mapping_t *map, char *name);
437 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
438
439 extern char *map_dev(int major, int minor, int create);
440
441 struct active_array;
442 struct metadata_update;
443
444 /* A superswitch provides entry point the a metadata handler.
445 *
446 * The super_switch primarily operates on some "metadata" that
447 * is accessed via the 'supertype'.
448 * This metadata has one of three possible sources.
449 * 1/ It is read from a single device. In this case it may not completely
450 * describe the array or arrays as some information might be on other
451 * devices.
452 * 2/ It is read from all devices in a container. In this case all
453 * information is present.
454 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
455 * be complete once enough ->add_to_super calls have completed.
456 *
457 * When creating an array inside a container, the metadata will be
458 * formed by a combination of 2 and 3. The metadata or the array is read,
459 * then new information is added.
460 *
461 * The metadata must sometimes have a concept of a 'current' array
462 * and a 'current' device.
463 * The 'current' array is set by init_super to be the newly created array,
464 * or is set by super_by_fd when it finds it is looking at an array inside
465 * a container.
466 *
467 * The 'current' device is either the device that the metadata was read from
468 * in case 1, or the last device added by add_to_super in case 3.
469 * Case 2 does not identify a 'current' device.
470 */
471 extern struct superswitch {
472
473 /* Used to report details of metadata read from a component
474 * device. ->load_super has been called.
475 */
476 void (*examine_super)(struct supertype *st, char *homehost);
477 void (*brief_examine_super)(struct supertype *st, int verbose);
478 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
479 void (*export_examine_super)(struct supertype *st);
480
481 /* Used to report details of an active array.
482 * ->load_super was possibly given a 'component' string.
483 */
484 void (*detail_super)(struct supertype *st, char *homehost);
485 void (*brief_detail_super)(struct supertype *st);
486 void (*export_detail_super)(struct supertype *st);
487
488 /* Optional: platform hardware / firmware details */
489 int (*detail_platform)(int verbose, int enumerate_only);
490
491 /* Used:
492 * to get uuid to storing in bitmap metadata
493 * and 'reshape' backup-data metadata
494 * To see if a device is being re-added to an array it was part of.
495 */
496 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
497
498 /* Extract generic details from metadata. This could be details about
499 * the container, or about an individual array within the container.
500 * The determination is made either by:
501 * load_super being given a 'component' string.
502 * validate_geometry determining what to create.
503 * The info includes both array information and device information.
504 * The particular device should be:
505 * The last device added by add_to_super
506 * The device the metadata was loaded from by load_super
507 */
508 void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
509
510 /* Check if the given metadata is flagged as belonging to "this"
511 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
512 */
513 int (*match_home)(struct supertype *st, char *homehost);
514
515 /* Make one of several generic modifications to metadata
516 * prior to assembly (or other times).
517 * sparc2.2 - first bug in early 0.90 metadata
518 * super-minor - change name of 0.90 metadata
519 * summaries - 'correct' any redundant data
520 * resync - mark array as dirty to trigger a resync.
521 * uuid - set new uuid - only 0.90 or 1.x
522 * name - change the name of the array (where supported)
523 * homehost - change which host this array is tied to.
524 * devicesize - If metadata is at start of device, change recorded
525 * device size to match actual device size
526 * byteorder - swap bytes for 0.90 metadata
527 *
528 * force-one - mark that device as uptodate, not old or failed.
529 * force-array - mark array as clean if it would not otherwise
530 * assemble
531 * assemble - not sure how this is different from force-one...
532 * linear-grow-new - add a new device to a linear array, but don't
533 * change the size: so superblock still matches
534 * linear-grow-update - now change the size of the array.
535 */
536 int (*update_super)(struct supertype *st, struct mdinfo *info,
537 char *update,
538 char *devname, int verbose,
539 int uuid_set, char *homehost);
540
541 /* Create new metadata for new array as described. This could
542 * be a new container, or an array in a pre-existing container.
543 * Also used to zero metadata prior to writing it to invalidate old
544 * metadata.
545 */
546 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
547 unsigned long long size, char *name,
548 char *homehost, int *uuid);
549
550 /* update the metadata to include new device, either at create or
551 * when hot-adding a spare.
552 */
553 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
554 int fd, char *devname);
555
556 /* Write metadata to one device when fixing problems or adding
557 * a new device.
558 */
559 int (*store_super)(struct supertype *st, int fd);
560
561 /* Write all metadata for this array.
562 */
563 int (*write_init_super)(struct supertype *st);
564 int (*compare_super)(struct supertype *st, struct supertype *tst);
565 int (*load_super)(struct supertype *st, int fd, char *devname);
566 struct supertype * (*match_metadata_desc)(char *arg);
567 __u64 (*avail_size)(struct supertype *st, __u64 size);
568 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
569 int delay, int write_behind,
570 unsigned long long size, int may_change, int major);
571 void (*locate_bitmap)(struct supertype *st, int fd);
572 int (*write_bitmap)(struct supertype *st, int fd);
573 void (*free_super)(struct supertype *st);
574
575 /* validate_geometry is called with an st returned by
576 * match_metadata_desc.
577 * It should check that the geometry described in compatible with
578 * the metadata type. It will be called repeatedly as devices
579 * added to validate changing size and new devices. If there are
580 * inter-device dependencies, it should record sufficient details
581 * so these can be validated.
582 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
583 */
584 int (*validate_geometry)(struct supertype *st, int level, int layout,
585 int raiddisks,
586 int chunk, unsigned long long size,
587 char *subdev, unsigned long long *freesize,
588 int verbose);
589
590 struct mdinfo *(*container_content)(struct supertype *st);
591 /* Allow a metadata handler to override mdadm's default layouts */
592 int (*default_layout)(int level); /* optional */
593
594 /* for mdmon */
595 int (*open_new)(struct supertype *c, struct active_array *a,
596 char *inst);
597
598 /* Tell the metadata handler the current state of the array.
599 * This covers whether it is known to be consistent (no pending writes)
600 * and how far along a resync is known to have progressed
601 * (in a->resync_start).
602 * resync status is really irrelevant if the array is not consistent,
603 * but some metadata (DDF!) have a place to record the distinction.
604 * If 'consistent' is '2', then the array can mark it dirty if a
605 * resync/recovery/whatever is required, or leave it clean if not.
606 * Return value is 0 dirty (not consistent) and 1 if clean.
607 * it is only really important if consistent is passed in as '2'.
608 */
609 int (*set_array_state)(struct active_array *a, int consistent);
610
611 /* When the state of a device might have changed, we call set_disk to
612 * tell the metadata what the current state is.
613 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
614 * transitions.
615 * set_disk might be called when the state of the particular disk has
616 * not in fact changed.
617 */
618 void (*set_disk)(struct active_array *a, int n, int state);
619 void (*sync_metadata)(struct supertype *st);
620 void (*process_update)(struct supertype *st,
621 struct metadata_update *update);
622 void (*prepare_update)(struct supertype *st,
623 struct metadata_update *update);
624
625 /* activate_spare will check if the array is degraded and, if it
626 * is, try to find some spare space in the container.
627 * On success, it add appropriate updates (For process_update) to
628 * to the 'updates' list and returns a list of 'mdinfo' identifying
629 * the device, or devices as there might be multiple missing
630 * devices and multiple spares available.
631 */
632 struct mdinfo *(*activate_spare)(struct active_array *a,
633 struct metadata_update **updates);
634
635 int swapuuid; /* true if uuid is bigending rather than hostendian */
636 int external;
637 const char *name; /* canonical metadata name */
638 } super0, super1, super_ddf, *superlist[];
639
640 extern struct superswitch super_imsm;
641
642 struct metadata_update {
643 int len;
644 char *buf;
645 void *space; /* allocated space that monitor will use */
646 struct metadata_update *next;
647 };
648
649 /* A supertype holds a particular collection of metadata.
650 * It identifies the metadata type by the superswitch, and the particular
651 * sub-version of that metadata type.
652 * metadata read in or created is stored in 'sb' and 'info'.
653 * There are also fields used by mdmon to track containers.
654 *
655 * A supertype may refer to:
656 * Just an array, possibly in a container
657 * A container, not identifying any particular array
658 * Info read from just one device, not yet fully describing the array/container.
659 *
660 *
661 * A supertype is created by:
662 * super_by_fd
663 * guess_super
664 * dup_super
665 */
666 struct supertype {
667 struct superswitch *ss;
668 int minor_version;
669 int max_devs;
670 int container_dev; /* devnum of container */
671 char subarray[32]; /* name of array inside container */
672 void *sb;
673 void *info;
674 int loaded_container; /* Set if load_super found a container,
675 * not just one device */
676
677 struct metadata_update *updates;
678 struct metadata_update **update_tail;
679
680 /* extra stuff used by mdmon */
681 struct active_array *arrays;
682 int sock; /* listen to external programs */
683 int devnum;
684 char *devname; /* e.g. md0. This appears in metadata_verison:
685 * external:/md0/12
686 */
687 int devcnt;
688
689 struct mdinfo *devs;
690
691 };
692
693 extern struct supertype *super_by_fd(int fd);
694 extern struct supertype *guess_super(int fd);
695 extern struct supertype *dup_super(struct supertype *st);
696 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
697 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
698 mdu_disk_info_t *disk);
699 void wait_for(char *dev, int fd);
700
701 #if __GNUC__ < 3
702 struct stat64;
703 #endif
704
705 #define HAVE_NFTW we assume
706 #define HAVE_FTW
707
708 #ifdef __UCLIBC__
709 # include <features.h>
710 # ifndef __UCLIBC_HAS_LFS__
711 # define lseek64 lseek
712 # endif
713 # ifndef __UCLIBC_HAS_FTW__
714 # undef HAVE_FTW
715 # undef HAVE_NFTW
716 # endif
717 #endif
718
719 #ifdef __dietlibc__
720 # undef HAVE_NFTW
721 #endif
722
723 #if defined(__KLIBC__)
724 # undef HAVE_NFTW
725 # undef HAVE_FTW
726 #endif
727
728 #ifndef HAVE_NFTW
729 # define FTW_PHYS 1
730 # ifndef HAVE_FTW
731 struct FTW {};
732 # endif
733 #endif
734
735 #ifdef HAVE_FTW
736 # include <ftw.h>
737 #endif
738
739 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
740
741
742 extern int Manage_ro(char *devname, int fd, int readonly);
743 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
744 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
745 extern int Manage_subdevs(char *devname, int fd,
746 mddev_dev_t devlist, int verbose);
747 extern int autodetect(void);
748 extern int Grow_Add_device(char *devname, int fd, char *newdev);
749 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
750 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
751 long long size,
752 int level, char *layout_str, int chunksize, int raid_disks);
753 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
754 int *fdlist, int cnt, char *backup_file, int verbose);
755 extern int Grow_continue(int mdfd, struct supertype *st,
756 struct mdinfo *info, char *backup_file);
757
758 extern int Assemble(struct supertype *st, char *mddev,
759 mddev_ident_t ident,
760 mddev_dev_t devlist, char *backup_file,
761 int readonly, int runstop,
762 char *update, char *homehost, int require_homehost,
763 int verbose, int force);
764
765 extern int Build(char *mddev, int chunk, int level, int layout,
766 int raiddisks, mddev_dev_t devlist, int assume_clean,
767 char *bitmap_file, int bitmap_chunk, int write_behind,
768 int delay, int verbose, int autof, unsigned long long size);
769
770
771 extern int Create(struct supertype *st, char *mddev,
772 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
773 char *name, char *homehost, int *uuid,
774 int subdevs, mddev_dev_t devlist,
775 int runstop, int verbose, int force, int assume_clean,
776 char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
777
778 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
779 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
780 extern int Query(char *dev);
781 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
782 int SparcAdjust, struct supertype *forcest, char *homehost);
783 extern int Monitor(mddev_dev_t devlist,
784 char *mailaddr, char *alert_cmd,
785 int period, int daemonise, int scan, int oneshot,
786 int dosyslog, int test, char *pidfile, int increments);
787
788 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
789 extern int Wait(char *dev);
790 extern int WaitClean(char *dev, int sock, int verbose);
791
792 extern int Incremental(char *devname, int verbose, int runstop,
793 struct supertype *st, char *homehost, int require_homehost,
794 int autof);
795 extern int Incremental_container(struct supertype *st, char *devname,
796 int verbose, int runstop, int autof,
797 int trustworthy);
798 extern void RebuildMap(void);
799 extern int IncrementalScan(int verbose);
800
801 extern int CreateBitmap(char *filename, int force, char uuid[16],
802 unsigned long chunksize, unsigned long daemon_sleep,
803 unsigned long write_behind,
804 unsigned long long array_size,
805 int major);
806 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
807 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
808 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
809
810 extern int md_get_version(int fd);
811 extern int get_linux_version(void);
812 extern long long parse_size(char *size);
813 extern int parse_uuid(char *str, int uuid[4]);
814 extern int parse_layout_10(char *layout);
815 extern int parse_layout_faulty(char *layout);
816 extern int check_ext2(int fd, char *name);
817 extern int check_reiser(int fd, char *name);
818 extern int check_raid(int fd, char *name);
819 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
820
821 extern int get_mdp_major(void);
822 extern int dev_open(char *dev, int flags);
823 extern int open_dev(int devnum);
824 extern int open_dev_excl(int devnum);
825 extern int is_standard(char *dev, int *nump);
826 extern int same_dev(char *one, char *two);
827
828 extern int parse_auto(char *str, char *msg, int config);
829 extern mddev_ident_t conf_get_ident(char *dev);
830 extern mddev_dev_t conf_get_devs(void);
831 extern int conf_test_dev(char *devname);
832 extern int conf_test_metadata(const char *version);
833 extern struct createinfo *conf_get_create_info(void);
834 extern void set_conffile(char *file);
835 extern char *conf_get_mailaddr(void);
836 extern char *conf_get_mailfrom(void);
837 extern char *conf_get_program(void);
838 extern char *conf_get_homehost(int *require_homehostp);
839 extern char *conf_line(FILE *file);
840 extern char *conf_word(FILE *file, int allow_key);
841 extern int conf_name_is_free(char *name);
842 extern int devname_matches(char *name, char *match);
843 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
844
845 extern void free_line(char *line);
846 extern int match_oneof(char *devices, char *devname);
847 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
848 extern const int uuid_match_any[4];
849 extern int same_uuid(int a[4], int b[4], int swapuuid);
850 extern void copy_uuid(void *a, int b[4], int swapuuid);
851 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
852 extern char *fname_from_uuid(struct supertype *st,
853 struct mdinfo *info, char *buf, char sep);
854 extern unsigned long calc_csum(void *super, int bytes);
855 extern int enough(int level, int raid_disks, int layout, int clean,
856 char *avail, int avail_disks);
857 extern int ask(char *mesg);
858 extern unsigned long long get_component_size(int fd);
859 extern void remove_partitions(int fd);
860 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
861 int chunksize, unsigned long long devsize);
862 extern int flush_metadata_updates(struct supertype *st);
863 extern void append_metadata_update(struct supertype *st, void *buf, int len);
864 extern int assemble_container_content(struct supertype *st, int mdfd,
865 struct mdinfo *content, int runstop,
866 char *chosen_name, int verbose);
867
868 extern int add_disk(int mdfd, struct supertype *st,
869 struct mdinfo *sra, struct mdinfo *info);
870 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
871 unsigned long long min_recovery_start(struct mdinfo *array);
872
873 extern char *human_size(long long bytes);
874 extern char *human_size_brief(long long bytes);
875 extern void print_r10_layout(int layout);
876
877 #define NoMdDev (1<<23)
878 extern int find_free_devnum(int use_partitions);
879
880 extern void put_md_name(char *name);
881 extern char *get_md_name(int dev);
882
883 extern char DefaultConfFile[];
884
885 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
886 char *chosen);
887 /* values for 'trustworthy' */
888 #define LOCAL 1
889 #define FOREIGN 2
890 #define METADATA 3
891 extern int open_mddev(char *dev, int report_errors);
892 extern int open_container(int fd);
893
894 extern char *pid_dir;
895 extern int mdmon_running(int devnum);
896 extern int mdmon_pid(int devnum);
897 extern int check_env(char *name);
898 extern __u32 random32(void);
899 extern int start_mdmon(int devnum);
900
901 extern char *devnum2devname(int num);
902 extern int devname2devnum(char *name);
903 extern int stat2devnum(struct stat *st);
904 extern int fd2devnum(int fd);
905
906 static inline int dev2major(int d)
907 {
908 if (d >= 0)
909 return MD_MAJOR;
910 else
911 return get_mdp_major();
912 }
913
914 static inline int dev2minor(int d)
915 {
916 if (d >= 0)
917 return d;
918 return (-1-d) << MdpMinorShift;
919 }
920
921 static inline int ROUND_UP(int a, int base)
922 {
923 return ((a+base-1)/base)*base;
924 }
925
926 static inline int is_subarray(char *vers)
927 {
928 /* The version string for a 'subarray' (an array in a container)
929 * is
930 * /containername/componentname for normal read-write arrays
931 * -containername/componentname for read-only arrays.
932 * containername is e.g. md0, md_d1
933 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
934 */
935 return (*vers == '/' || *vers == '-');
936 }
937
938 #ifdef DEBUG
939 #define dprintf(fmt, arg...) \
940 fprintf(stderr, fmt, ##arg)
941 #else
942 #define dprintf(fmt, arg...) \
943 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
944 #endif
945 #include <assert.h>
946 #include <stdarg.h>
947 static inline int xasprintf(char **strp, const char *fmt, ...) {
948 va_list ap;
949 int ret;
950 va_start(ap, fmt);
951 ret = vasprintf(strp, fmt, ap);
952 va_end(ap);
953 assert(ret >= 0);
954 return ret;
955 }
956
957 #define LEVEL_MULTIPATH (-4)
958 #define LEVEL_LINEAR (-1)
959 #define LEVEL_FAULTY (-5)
960
961 /* kernel module doesn't know about these */
962 #define LEVEL_CONTAINER (-100)
963 #define LEVEL_UNSUPPORTED (-200)
964
965
966 /* faulty stuff */
967
968 #define WriteTransient 0
969 #define ReadTransient 1
970 #define WritePersistent 2
971 #define ReadPersistent 3
972 #define WriteAll 4 /* doesn't go to device */
973 #define ReadFixable 5
974 #define Modes 6
975
976 #define ClearErrors 31
977 #define ClearFaults 30
978
979 #define AllPersist 100 /* internal use only */
980 #define NoPersist 101
981
982 #define ModeMask 0x1f
983 #define ModeShift 5
984
985
986 #ifdef __TINYC__
987 #undef minor
988 #undef major
989 #undef makedev
990 #define minor(x) ((x)&0xff)
991 #define major(x) (((x)>>8)&0xff)
992 #define makedev(M,m) (((M)<<8) | (m))
993 #endif
994
995 /* for raid4/5/6 */
996 #define ALGORITHM_LEFT_ASYMMETRIC 0
997 #define ALGORITHM_RIGHT_ASYMMETRIC 1
998 #define ALGORITHM_LEFT_SYMMETRIC 2
999 #define ALGORITHM_RIGHT_SYMMETRIC 3
1000
1001 /* Define non-rotating (raid4) algorithms. These allow
1002 * conversion of raid4 to raid5.
1003 */
1004 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1005 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1006
1007 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1008 * Firstly, the exact positioning of the parity block is slightly
1009 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1010 * of DDF.
1011 * Secondly, or order of datablocks over which the Q syndrome is computed
1012 * is different.
1013 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1014 * These layouts are from the DDFv1.2 spec.
1015 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1016 * leaves RLQ=3 as 'Vendor Specific'
1017 */
1018
1019 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1020 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1021 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1022
1023
1024 /* For every RAID5 algorithm we define a RAID6 algorithm
1025 * with exactly the same layout for data and parity, and
1026 * with the Q block always on the last device (N-1).
1027 * This allows trivial conversion from RAID5 to RAID6
1028 */
1029 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1030 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1031 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1032 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1033 #define ALGORITHM_PARITY_0_6 20
1034 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1035
1036 /* Define PATH_MAX in case we don't use glibc or standard library does
1037 * not have PATH_MAX defined. Assume max path length is 4K characters.
1038 */
1039 #ifndef PATH_MAX
1040 #define PATH_MAX 4096
1041 #endif
1042