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