]>
Commit | Line | Data |
---|---|---|
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 | /* FAILED_SLOTS is where to save files storing recent removal of array | |
97 | * member in order to allow future reuse of disk inserted in the same | |
98 | * slot for array recovery | |
99 | */ | |
100 | #ifndef FAILED_SLOTS_DIR | |
101 | #define FAILED_SLOTS_DIR "/dev/.mdadm/failed-slots" | |
102 | #endif /* FAILED_SLOTS */ | |
103 | ||
104 | #include "md_u.h" | |
105 | #include "md_p.h" | |
106 | #include "bitmap.h" | |
107 | #include "msg.h" | |
108 | ||
109 | #include <endian.h> | |
110 | /* Redhat don't like to #include <asm/byteorder.h>, and | |
111 | * some time include <linux/byteorder/xxx_endian.h> isn't enough, | |
112 | * and there is no standard conversion function so... */ | |
113 | /* And dietlibc doesn't think byteswap is ok, so.. */ | |
114 | /* #include <byteswap.h> */ | |
115 | #define bswap_16(x) (((x) & 0x00ffU) << 8 | \ | |
116 | ((x) & 0xff00U) >> 8) | |
117 | #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \ | |
118 | ((x) & 0xff000000U) >> 24 | \ | |
119 | ((x) & 0x0000ff00U) << 8 | \ | |
120 | ((x) & 0x00ff0000U) >> 8) | |
121 | #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \ | |
122 | ((x) & 0xff00000000000000ULL) >> 56 | \ | |
123 | ((x) & 0x000000000000ff00ULL) << 40 | \ | |
124 | ((x) & 0x00ff000000000000ULL) >> 40 | \ | |
125 | ((x) & 0x0000000000ff0000ULL) << 24 | \ | |
126 | ((x) & 0x0000ff0000000000ULL) >> 24 | \ | |
127 | ((x) & 0x00000000ff000000ULL) << 8 | \ | |
128 | ((x) & 0x000000ff00000000ULL) >> 8) | |
129 | ||
130 | #if !defined(__KLIBC__) | |
131 | #if BYTE_ORDER == LITTLE_ENDIAN | |
132 | #define __cpu_to_le16(_x) (_x) | |
133 | #define __cpu_to_le32(_x) (_x) | |
134 | #define __cpu_to_le64(_x) (_x) | |
135 | #define __le16_to_cpu(_x) (_x) | |
136 | #define __le32_to_cpu(_x) (_x) | |
137 | #define __le64_to_cpu(_x) (_x) | |
138 | ||
139 | #define __cpu_to_be16(_x) bswap_16(_x) | |
140 | #define __cpu_to_be32(_x) bswap_32(_x) | |
141 | #define __cpu_to_be64(_x) bswap_64(_x) | |
142 | #define __be16_to_cpu(_x) bswap_16(_x) | |
143 | #define __be32_to_cpu(_x) bswap_32(_x) | |
144 | #define __be64_to_cpu(_x) bswap_64(_x) | |
145 | #elif BYTE_ORDER == BIG_ENDIAN | |
146 | #define __cpu_to_le16(_x) bswap_16(_x) | |
147 | #define __cpu_to_le32(_x) bswap_32(_x) | |
148 | #define __cpu_to_le64(_x) bswap_64(_x) | |
149 | #define __le16_to_cpu(_x) bswap_16(_x) | |
150 | #define __le32_to_cpu(_x) bswap_32(_x) | |
151 | #define __le64_to_cpu(_x) bswap_64(_x) | |
152 | ||
153 | #define __cpu_to_be16(_x) (_x) | |
154 | #define __cpu_to_be32(_x) (_x) | |
155 | #define __cpu_to_be64(_x) (_x) | |
156 | #define __be16_to_cpu(_x) (_x) | |
157 | #define __be32_to_cpu(_x) (_x) | |
158 | #define __be64_to_cpu(_x) (_x) | |
159 | #else | |
160 | # error "unknown endianness." | |
161 | #endif | |
162 | #endif /* __KLIBC__ */ | |
163 | ||
164 | ||
165 | /* | |
166 | * min()/max()/clamp() macros that also do | |
167 | * strict type-checking.. See the | |
168 | * "unnecessary" pointer comparison. | |
169 | */ | |
170 | #define min(x, y) ({ \ | |
171 | typeof(x) _min1 = (x); \ | |
172 | typeof(y) _min2 = (y); \ | |
173 | (void) (&_min1 == &_min2); \ | |
174 | _min1 < _min2 ? _min1 : _min2; }) | |
175 | ||
176 | #define max(x, y) ({ \ | |
177 | typeof(x) _max1 = (x); \ | |
178 | typeof(y) _max2 = (y); \ | |
179 | (void) (&_max1 == &_max2); \ | |
180 | _max1 > _max2 ? _max1 : _max2; }) | |
181 | ||
182 | /* general information that might be extracted from a superblock */ | |
183 | struct mdinfo { | |
184 | mdu_array_info_t array; | |
185 | mdu_disk_info_t disk; | |
186 | __u64 events; | |
187 | int uuid[4]; | |
188 | char name[33]; | |
189 | unsigned long long data_offset; | |
190 | unsigned long long component_size; /* same as array.size, except in | |
191 | * sectors and up to 64bits. | |
192 | */ | |
193 | unsigned long long custom_array_size; /* size for non-default sized | |
194 | * arrays (in sectors) | |
195 | */ | |
196 | int reshape_active; | |
197 | unsigned long long reshape_progress; | |
198 | union { | |
199 | unsigned long long resync_start; /* per-array resync position */ | |
200 | unsigned long long recovery_start; /* per-device rebuild position */ | |
201 | #define MaxSector (~0ULL) /* resync/recovery complete position */ | |
202 | }; | |
203 | unsigned long safe_mode_delay; /* ms delay to mark clean */ | |
204 | int new_level, delta_disks, new_layout, new_chunk; | |
205 | int errors; | |
206 | unsigned long cache_size; /* size of raid456 stripe cache*/ | |
207 | int mismatch_cnt; | |
208 | char text_version[50]; | |
209 | void *update_private; /* for passing metadata-format | |
210 | * specific update data | |
211 | * between successive calls to | |
212 | * update_super() | |
213 | */ | |
214 | ||
215 | int container_member; /* for assembling external-metatdata arrays | |
216 | * This is to be used internally by metadata | |
217 | * handler only */ | |
218 | int container_enough; /* flag external handlers can set to | |
219 | * indicate that subarrays have not enough (-1), | |
220 | * enough to start (0), or all expected disks (1) */ | |
221 | char sys_name[20]; | |
222 | struct mdinfo *devs; | |
223 | struct mdinfo *next; | |
224 | ||
225 | /* Device info for mdmon: */ | |
226 | int recovery_fd; | |
227 | int state_fd; | |
228 | #define DS_FAULTY 1 | |
229 | #define DS_INSYNC 2 | |
230 | #define DS_WRITE_MOSTLY 4 | |
231 | #define DS_SPARE 8 | |
232 | #define DS_BLOCKED 16 | |
233 | #define DS_REMOVE 1024 | |
234 | #define DS_UNBLOCK 2048 | |
235 | int prev_state, curr_state, next_state; | |
236 | ||
237 | }; | |
238 | ||
239 | struct createinfo { | |
240 | int uid; | |
241 | int gid; | |
242 | int autof; | |
243 | int mode; | |
244 | int symlinks; | |
245 | struct supertype *supertype; | |
246 | }; | |
247 | ||
248 | #define Name "mdadm" | |
249 | ||
250 | enum mode { | |
251 | ASSEMBLE=1, | |
252 | BUILD, | |
253 | CREATE, | |
254 | MANAGE, | |
255 | MISC, | |
256 | MONITOR, | |
257 | GROW, | |
258 | INCREMENTAL, | |
259 | AUTODETECT, | |
260 | }; | |
261 | ||
262 | extern char short_options[]; | |
263 | extern char short_bitmap_options[]; | |
264 | extern char short_bitmap_auto_options[]; | |
265 | extern struct option long_options[]; | |
266 | extern char Version[], Usage[], Help[], OptionHelp[], | |
267 | Help_create[], Help_build[], Help_assemble[], Help_grow[], | |
268 | Help_incr[], | |
269 | Help_manage[], Help_misc[], Help_monitor[], Help_config[]; | |
270 | ||
271 | /* for option that don't have short equivilents, we assign arbitrary | |
272 | * numbers later than any 'short' character option. | |
273 | */ | |
274 | enum special_options { | |
275 | AssumeClean = 300, | |
276 | BitmapChunk, | |
277 | WriteBehind, | |
278 | ReAdd, | |
279 | NoDegraded, | |
280 | Sparc22, | |
281 | BackupFile, | |
282 | HomeHost, | |
283 | AutoHomeHost, | |
284 | Symlinks, | |
285 | AutoDetect, | |
286 | Waitclean, | |
287 | DetailPlatform, | |
288 | KillSubarray, | |
289 | UpdateSubarray, | |
290 | IncrementalPath, | |
291 | NoSharing, | |
292 | HelpOptions, | |
293 | Brief, | |
294 | ManageOpt, | |
295 | Add, | |
296 | Remove, | |
297 | Fail, | |
298 | MiscOpt, | |
299 | WaitOpt, | |
300 | ConfigFile, | |
301 | ChunkSize, | |
302 | WriteMostly, | |
303 | Layout, | |
304 | Auto, | |
305 | Force, | |
306 | SuperMinor, | |
307 | EMail, | |
308 | ProgramOpt, | |
309 | Increment, | |
310 | Fork, | |
311 | Bitmap, | |
312 | RebuildMapOpt, | |
313 | InvalidBackup, | |
314 | }; | |
315 | ||
316 | /* structures read from config file */ | |
317 | /* List of mddevice names and identifiers | |
318 | * Identifiers can be: | |
319 | * uuid=128-hex-uuid | |
320 | * super-minor=decimal-minor-number-from-superblock | |
321 | * devices=comma,separated,list,of,device,names,with,wildcards | |
322 | * | |
323 | * If multiple fields are present, the intersection of all matching | |
324 | * devices is considered | |
325 | */ | |
326 | #define UnSet (0xfffe) | |
327 | struct mddev_ident { | |
328 | char *devname; | |
329 | ||
330 | int uuid_set; | |
331 | int uuid[4]; | |
332 | char name[33]; | |
333 | ||
334 | int super_minor; | |
335 | ||
336 | char *devices; /* comma separated list of device | |
337 | * names with wild cards | |
338 | */ | |
339 | int level; | |
340 | int raid_disks; | |
341 | int spare_disks; | |
342 | struct supertype *st; | |
343 | int autof; /* 1 for normal, 2 for partitioned */ | |
344 | char *spare_group; | |
345 | char *bitmap_file; | |
346 | int bitmap_fd; | |
347 | ||
348 | char *container; /* /dev/whatever name of container, or | |
349 | * uuid of container. You would expect | |
350 | * this to be the 'devname' or UUID | |
351 | * of some other entry. | |
352 | */ | |
353 | char *member; /* subarray within a container */ | |
354 | ||
355 | struct mddev_ident *next; | |
356 | union { | |
357 | /* fields needed by different users of this structure */ | |
358 | int assembled; /* set when assembly succeeds */ | |
359 | }; | |
360 | }; | |
361 | ||
362 | /* List of device names - wildcards expanded */ | |
363 | struct mddev_dev { | |
364 | char *devname; | |
365 | int disposition; /* 'a' for add, 'r' for remove, 'f' for fail. | |
366 | * Not set for names read from .config | |
367 | */ | |
368 | char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */ | |
369 | char re_add; | |
370 | char used; /* set when used */ | |
371 | struct mddev_dev *next; | |
372 | }; | |
373 | ||
374 | typedef struct mapping { | |
375 | char *name; | |
376 | int num; | |
377 | } mapping_t; | |
378 | ||
379 | ||
380 | struct mdstat_ent { | |
381 | char *dev; | |
382 | int devnum; | |
383 | int active; | |
384 | char *level; | |
385 | char *pattern; /* U or up, _ for down */ | |
386 | int percent; /* -1 if no resync */ | |
387 | int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */ | |
388 | int devcnt; | |
389 | int raid_disks; | |
390 | char * metadata_version; | |
391 | struct dev_member { | |
392 | char *name; | |
393 | struct dev_member *next; | |
394 | } *members; | |
395 | struct mdstat_ent *next; | |
396 | }; | |
397 | ||
398 | extern struct mdstat_ent *mdstat_read(int hold, int start); | |
399 | extern void free_mdstat(struct mdstat_ent *ms); | |
400 | extern void mdstat_wait(int seconds); | |
401 | extern void mdstat_wait_fd(int fd, const sigset_t *sigmask); | |
402 | extern int mddev_busy(int devnum); | |
403 | extern struct mdstat_ent *mdstat_by_component(char *name); | |
404 | extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container); | |
405 | ||
406 | struct map_ent { | |
407 | struct map_ent *next; | |
408 | int devnum; | |
409 | char metadata[20]; | |
410 | int uuid[4]; | |
411 | int bad; | |
412 | char *path; | |
413 | }; | |
414 | extern int map_update(struct map_ent **mpp, int devnum, char *metadata, | |
415 | int uuid[4], char *path); | |
416 | extern void map_remove(struct map_ent **map, int devnum); | |
417 | extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]); | |
418 | extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum); | |
419 | extern struct map_ent *map_by_name(struct map_ent **map, char *name); | |
420 | extern void map_read(struct map_ent **melp); | |
421 | extern int map_write(struct map_ent *mel); | |
422 | extern void map_delete(struct map_ent **mapp, int devnum); | |
423 | extern void map_free(struct map_ent *map); | |
424 | extern void map_add(struct map_ent **melp, | |
425 | int devnum, char *metadata, int uuid[4], char *path); | |
426 | extern int map_lock(struct map_ent **melp); | |
427 | extern void map_unlock(struct map_ent **melp); | |
428 | ||
429 | /* various details can be requested */ | |
430 | enum sysfs_read_flags { | |
431 | GET_LEVEL = (1 << 0), | |
432 | GET_LAYOUT = (1 << 1), | |
433 | GET_COMPONENT = (1 << 2), | |
434 | GET_CHUNK = (1 << 3), | |
435 | GET_CACHE = (1 << 4), | |
436 | GET_MISMATCH = (1 << 5), | |
437 | GET_VERSION = (1 << 6), | |
438 | GET_DISKS = (1 << 7), | |
439 | GET_DEGRADED = (1 << 8), | |
440 | GET_SAFEMODE = (1 << 9), | |
441 | GET_DEVS = (1 << 10), /* gets role, major, minor */ | |
442 | GET_OFFSET = (1 << 11), | |
443 | GET_SIZE = (1 << 12), | |
444 | GET_STATE = (1 << 13), | |
445 | GET_ERROR = (1 << 14), | |
446 | }; | |
447 | ||
448 | /* If fd >= 0, get the array it is open on, | |
449 | * else use devnum. >=0 -> major9. <0..... | |
450 | */ | |
451 | extern int sysfs_open(int devnum, char *devname, char *attr); | |
452 | extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum); | |
453 | extern void sysfs_free(struct mdinfo *sra); | |
454 | extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options); | |
455 | extern int sysfs_attr_match(const char *attr, const char *str); | |
456 | extern int sysfs_match_word(const char *word, char **list); | |
457 | extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev, | |
458 | char *name, char *val); | |
459 | extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev, | |
460 | char *name, unsigned long long val); | |
461 | extern int sysfs_uevent(struct mdinfo *sra, char *event); | |
462 | extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev, | |
463 | char *name); | |
464 | extern int sysfs_fd_get_ll(int fd, unsigned long long *val); | |
465 | extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev, | |
466 | char *name, unsigned long long *val); | |
467 | extern int sysfs_fd_get_str(int fd, char *val, int size); | |
468 | extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev, | |
469 | char *name); | |
470 | extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev, | |
471 | char *name, char *val, int size); | |
472 | extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms); | |
473 | extern int sysfs_set_array(struct mdinfo *info, int vers); | |
474 | extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume); | |
475 | extern int sysfs_disk_to_scsi_id(int fd, __u32 *id); | |
476 | extern int sysfs_unique_holder(int devnum, long rdev); | |
477 | extern int sysfs_freeze_array(struct mdinfo *sra); | |
478 | extern int load_sys(char *path, char *buf); | |
479 | extern int reshape_prepare_fdlist(char *devname, | |
480 | struct mdinfo *sra, | |
481 | int raid_disks, | |
482 | int nrdisks, | |
483 | unsigned long blocks, | |
484 | char *backup_file, | |
485 | int *fdlist, | |
486 | unsigned long long *offsets); | |
487 | extern void reshape_free_fdlist(int *fdlist, | |
488 | unsigned long long *offsets, | |
489 | int size); | |
490 | extern int reshape_open_backup_file(char *backup, | |
491 | int fd, | |
492 | char *devname, | |
493 | long blocks, | |
494 | int *fdlist, | |
495 | unsigned long long *offsets); | |
496 | extern unsigned long compute_backup_blocks(int nchunk, int ochunk, | |
497 | unsigned int ndata, unsigned int odata); | |
498 | ||
499 | extern int save_stripes(int *source, unsigned long long *offsets, | |
500 | int raid_disks, int chunk_size, int level, int layout, | |
501 | int nwrites, int *dest, | |
502 | unsigned long long start, unsigned long long length, | |
503 | char *buf); | |
504 | extern int restore_stripes(int *dest, unsigned long long *offsets, | |
505 | int raid_disks, int chunk_size, int level, int layout, | |
506 | int source, unsigned long long read_offset, | |
507 | unsigned long long start, unsigned long long length); | |
508 | ||
509 | #ifndef Sendmail | |
510 | #define Sendmail "/usr/lib/sendmail -t" | |
511 | #endif | |
512 | ||
513 | #define SYSLOG_FACILITY LOG_DAEMON | |
514 | ||
515 | extern char *map_num(mapping_t *map, int num); | |
516 | extern int map_name(mapping_t *map, char *name); | |
517 | extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[]; | |
518 | ||
519 | extern char *map_dev(int major, int minor, int create); | |
520 | ||
521 | struct active_array; | |
522 | struct metadata_update; | |
523 | ||
524 | /* A superswitch provides entry point the a metadata handler. | |
525 | * | |
526 | * The super_switch primarily operates on some "metadata" that | |
527 | * is accessed via the 'supertype'. | |
528 | * This metadata has one of three possible sources. | |
529 | * 1/ It is read from a single device. In this case it may not completely | |
530 | * describe the array or arrays as some information might be on other | |
531 | * devices. | |
532 | * 2/ It is read from all devices in a container. In this case all | |
533 | * information is present. | |
534 | * 3/ It is created by ->init_super / ->add_to_super. In this case it will | |
535 | * be complete once enough ->add_to_super calls have completed. | |
536 | * | |
537 | * When creating an array inside a container, the metadata will be | |
538 | * formed by a combination of 2 and 3. The metadata or the array is read, | |
539 | * then new information is added. | |
540 | * | |
541 | * The metadata must sometimes have a concept of a 'current' array | |
542 | * and a 'current' device. | |
543 | * The 'current' array is set by init_super to be the newly created array, | |
544 | * or is set by super_by_fd when it finds it is looking at an array inside | |
545 | * a container. | |
546 | * | |
547 | * The 'current' device is either the device that the metadata was read from | |
548 | * in case 1, or the last device added by add_to_super in case 3. | |
549 | * Case 2 does not identify a 'current' device. | |
550 | */ | |
551 | extern struct superswitch { | |
552 | ||
553 | /* Used to report details of metadata read from a component | |
554 | * device. ->load_super has been called. | |
555 | */ | |
556 | void (*examine_super)(struct supertype *st, char *homehost); | |
557 | void (*brief_examine_super)(struct supertype *st, int verbose); | |
558 | void (*brief_examine_subarrays)(struct supertype *st, int verbose); | |
559 | void (*export_examine_super)(struct supertype *st); | |
560 | ||
561 | /* Used to report details of an active array. | |
562 | * ->load_super was possibly given a 'component' string. | |
563 | */ | |
564 | void (*detail_super)(struct supertype *st, char *homehost); | |
565 | void (*brief_detail_super)(struct supertype *st); | |
566 | void (*export_detail_super)(struct supertype *st); | |
567 | ||
568 | /* Optional: platform hardware / firmware details */ | |
569 | int (*detail_platform)(int verbose, int enumerate_only); | |
570 | ||
571 | /* Used: | |
572 | * to get uuid to storing in bitmap metadata | |
573 | * and 'reshape' backup-data metadata | |
574 | * To see if a device is being re-added to an array it was part of. | |
575 | */ | |
576 | void (*uuid_from_super)(struct supertype *st, int uuid[4]); | |
577 | ||
578 | /* Extract generic details from metadata. This could be details about | |
579 | * the container, or about an individual array within the container. | |
580 | * The determination is made either by: | |
581 | * load_super being given a 'component' string. | |
582 | * validate_geometry determining what to create. | |
583 | * The info includes both array information and device information. | |
584 | * The particular device should be: | |
585 | * The last device added by add_to_super | |
586 | * The device the metadata was loaded from by load_super | |
587 | * If 'map' is present, then it is an array raid_disks long | |
588 | * (raid_disk must already be set and correct) and it is filled | |
589 | * with 1 for slots that are thought to be active and 0 for slots which | |
590 | * appear to be failed/missing. | |
591 | */ | |
592 | void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map); | |
593 | struct mdinfo *(*getinfo_super_disks)(struct supertype *st); | |
594 | /* Check if the given metadata is flagged as belonging to "this" | |
595 | * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost" | |
596 | */ | |
597 | int (*match_home)(struct supertype *st, char *homehost); | |
598 | ||
599 | /* Make one of several generic modifications to metadata | |
600 | * prior to assembly (or other times). | |
601 | * sparc2.2 - first bug in early 0.90 metadata | |
602 | * super-minor - change name of 0.90 metadata | |
603 | * summaries - 'correct' any redundant data | |
604 | * resync - mark array as dirty to trigger a resync. | |
605 | * uuid - set new uuid - only 0.90 or 1.x | |
606 | * name - change the name of the array (where supported) | |
607 | * homehost - change which host this array is tied to. | |
608 | * devicesize - If metadata is at start of device, change recorded | |
609 | * device size to match actual device size | |
610 | * byteorder - swap bytes for 0.90 metadata | |
611 | * | |
612 | * force-one - mark that device as uptodate, not old or failed. | |
613 | * force-array - mark array as clean if it would not otherwise | |
614 | * assemble | |
615 | * assemble - not sure how this is different from force-one... | |
616 | * linear-grow-new - add a new device to a linear array, but don't | |
617 | * change the size: so superblock still matches | |
618 | * linear-grow-update - now change the size of the array. | |
619 | */ | |
620 | int (*update_super)(struct supertype *st, struct mdinfo *info, | |
621 | char *update, | |
622 | char *devname, int verbose, | |
623 | int uuid_set, char *homehost); | |
624 | ||
625 | /* Create new metadata for new array as described. This could | |
626 | * be a new container, or an array in a pre-existing container. | |
627 | * Also used to zero metadata prior to writing it to invalidate old | |
628 | * metadata. | |
629 | */ | |
630 | int (*init_super)(struct supertype *st, mdu_array_info_t *info, | |
631 | unsigned long long size, char *name, | |
632 | char *homehost, int *uuid); | |
633 | ||
634 | /* update the metadata to include new device, either at create or | |
635 | * when hot-adding a spare. | |
636 | */ | |
637 | int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo, | |
638 | int fd, char *devname); | |
639 | /* update the metadata to delete a device, | |
640 | * when hot-removing. | |
641 | */ | |
642 | int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo); | |
643 | ||
644 | /* Write metadata to one device when fixing problems or adding | |
645 | * a new device. | |
646 | */ | |
647 | int (*store_super)(struct supertype *st, int fd); | |
648 | ||
649 | /* Write all metadata for this array. | |
650 | */ | |
651 | int (*write_init_super)(struct supertype *st); | |
652 | int (*compare_super)(struct supertype *st, struct supertype *tst); | |
653 | int (*load_super)(struct supertype *st, int fd, char *devname); | |
654 | int (*load_container)(struct supertype *st, int fd, char *devname); | |
655 | struct supertype * (*match_metadata_desc)(char *arg); | |
656 | __u64 (*avail_size)(struct supertype *st, __u64 size); | |
657 | unsigned long long (*min_acceptable_spare_size)(struct supertype *st); | |
658 | int (*add_internal_bitmap)(struct supertype *st, int *chunkp, | |
659 | int delay, int write_behind, | |
660 | unsigned long long size, int may_change, int major); | |
661 | void (*locate_bitmap)(struct supertype *st, int fd); | |
662 | int (*write_bitmap)(struct supertype *st, int fd); | |
663 | void (*free_super)(struct supertype *st); | |
664 | ||
665 | /* validate_geometry is called with an st returned by | |
666 | * match_metadata_desc. | |
667 | * It should check that the geometry described in compatible with | |
668 | * the metadata type. It will be called repeatedly as devices | |
669 | * added to validate changing size and new devices. If there are | |
670 | * inter-device dependencies, it should record sufficient details | |
671 | * so these can be validated. | |
672 | * Both 'size' and '*freesize' are in sectors. chunk is KiB. | |
673 | */ | |
674 | int (*validate_geometry)(struct supertype *st, int level, int layout, | |
675 | int raiddisks, | |
676 | int chunk, unsigned long long size, | |
677 | char *subdev, unsigned long long *freesize, | |
678 | int verbose); | |
679 | ||
680 | struct mdinfo *(*container_content)(struct supertype *st, char *subarray); | |
681 | /* query the supertype for default geometry */ | |
682 | void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */ | |
683 | /* Permit subarray's to be deleted from inactive containers */ | |
684 | int (*kill_subarray)(struct supertype *st); /* optional */ | |
685 | /* Permit subarray's to be modified */ | |
686 | int (*update_subarray)(struct supertype *st, char *subarray, | |
687 | char *update, struct mddev_ident *ident); /* optional */ | |
688 | /* Check if reshape is supported for this external format. | |
689 | * st is obtained from super_by_fd() where st->subarray[0] is | |
690 | * initialized to indicate if reshape is being performed at the | |
691 | * container or subarray level | |
692 | */ | |
693 | int (*reshape_super)(struct supertype *st, long long size, int level, | |
694 | int layout, int chunksize, int raid_disks, | |
695 | char *backup, char *dev, int verbose); /* optional */ | |
696 | int (*manage_reshape)(struct supertype *st, char *backup); /* optional */ | |
697 | ||
698 | /* for mdmon */ | |
699 | int (*open_new)(struct supertype *c, struct active_array *a, | |
700 | char *inst); | |
701 | ||
702 | /* Tell the metadata handler the current state of the array. | |
703 | * This covers whether it is known to be consistent (no pending writes) | |
704 | * and how far along a resync is known to have progressed | |
705 | * (in a->resync_start). | |
706 | * resync status is really irrelevant if the array is not consistent, | |
707 | * but some metadata (DDF!) have a place to record the distinction. | |
708 | * If 'consistent' is '2', then the array can mark it dirty if a | |
709 | * resync/recovery/whatever is required, or leave it clean if not. | |
710 | * Return value is 0 dirty (not consistent) and 1 if clean. | |
711 | * it is only really important if consistent is passed in as '2'. | |
712 | */ | |
713 | int (*set_array_state)(struct active_array *a, int consistent); | |
714 | ||
715 | /* When the state of a device might have changed, we call set_disk to | |
716 | * tell the metadata what the current state is. | |
717 | * Typically this happens on spare->in_sync and (spare|in_sync)->faulty | |
718 | * transitions. | |
719 | * set_disk might be called when the state of the particular disk has | |
720 | * not in fact changed. | |
721 | */ | |
722 | void (*set_disk)(struct active_array *a, int n, int state); | |
723 | void (*sync_metadata)(struct supertype *st); | |
724 | void (*process_update)(struct supertype *st, | |
725 | struct metadata_update *update); | |
726 | void (*prepare_update)(struct supertype *st, | |
727 | struct metadata_update *update); | |
728 | ||
729 | /* activate_spare will check if the array is degraded and, if it | |
730 | * is, try to find some spare space in the container. | |
731 | * On success, it add appropriate updates (For process_update) to | |
732 | * to the 'updates' list and returns a list of 'mdinfo' identifying | |
733 | * the device, or devices as there might be multiple missing | |
734 | * devices and multiple spares available. | |
735 | */ | |
736 | struct mdinfo *(*activate_spare)(struct active_array *a, | |
737 | struct metadata_update **updates); | |
738 | /* | |
739 | * Return statically allocated string that represents metadata specific | |
740 | * controller domain of the disk. The domain is used in disk domain | |
741 | * matching functions. Disks belong to the same domain if the they have | |
742 | * the same domain from mdadm.conf and belong the same metadata domain. | |
743 | * Returning NULL or not providing this handler means that metadata | |
744 | * does not distinguish the differences between disks that belong to | |
745 | * different controllers. They are in the domain specified by | |
746 | * configuration file (mdadm.conf). | |
747 | * In case when the metadata has the notion of domains based on disk | |
748 | * it shall return NULL for disks that do not belong to the controller | |
749 | * the supported domains. Such disks will form another domain and won't | |
750 | * be mixed with supported ones. | |
751 | */ | |
752 | const char *(*get_disk_controller_domain)(const char *path); | |
753 | ||
754 | int swapuuid; /* true if uuid is bigending rather than hostendian */ | |
755 | int external; | |
756 | const char *name; /* canonical metadata name */ | |
757 | } *superlist[]; | |
758 | ||
759 | extern struct superswitch super0, super1; | |
760 | extern struct superswitch super_imsm, super_ddf; | |
761 | extern struct superswitch mbr, gpt; | |
762 | ||
763 | struct metadata_update { | |
764 | int len; | |
765 | char *buf; | |
766 | void *space; /* allocated space that monitor will use */ | |
767 | void **space_list; /* list of allocated spaces that monitor can | |
768 | * use or that it returned. | |
769 | */ | |
770 | struct metadata_update *next; | |
771 | }; | |
772 | ||
773 | /* A supertype holds a particular collection of metadata. | |
774 | * It identifies the metadata type by the superswitch, and the particular | |
775 | * sub-version of that metadata type. | |
776 | * metadata read in or created is stored in 'sb' and 'info'. | |
777 | * There are also fields used by mdmon to track containers. | |
778 | * | |
779 | * A supertype may refer to: | |
780 | * Just an array, possibly in a container | |
781 | * A container, not identifying any particular array | |
782 | * Info read from just one device, not yet fully describing the array/container. | |
783 | * | |
784 | * | |
785 | * A supertype is created by: | |
786 | * super_by_fd | |
787 | * guess_super | |
788 | * dup_super | |
789 | */ | |
790 | struct supertype { | |
791 | struct superswitch *ss; | |
792 | int minor_version; | |
793 | int max_devs; | |
794 | int container_dev; /* devnum of container */ | |
795 | void *sb; | |
796 | void *info; | |
797 | ||
798 | struct metadata_update *updates; | |
799 | struct metadata_update **update_tail; | |
800 | ||
801 | /* extra stuff used by mdmon */ | |
802 | struct active_array *arrays; | |
803 | int sock; /* listen to external programs */ | |
804 | int devnum; | |
805 | char *devname; /* e.g. md0. This appears in metadata_verison: | |
806 | * external:/md0/12 | |
807 | */ | |
808 | int devcnt; | |
809 | ||
810 | struct mdinfo *devs; | |
811 | ||
812 | }; | |
813 | ||
814 | extern struct supertype *super_by_fd(int fd, char **subarray); | |
815 | enum guess_types { guess_any, guess_array, guess_partitions }; | |
816 | extern struct supertype *guess_super_type(int fd, enum guess_types guess_type); | |
817 | static inline struct supertype *guess_super(int fd) { | |
818 | return guess_super_type(fd, guess_any); | |
819 | } | |
820 | extern struct supertype *dup_super(struct supertype *st); | |
821 | extern int get_dev_size(int fd, char *dname, unsigned long long *sizep); | |
822 | extern int must_be_container(int fd); | |
823 | extern int dev_size_from_id(dev_t id, unsigned long long *size); | |
824 | extern void get_one_disk(int mdfd, mdu_array_info_t *ainf, | |
825 | mdu_disk_info_t *disk); | |
826 | void wait_for(char *dev, int fd); | |
827 | ||
828 | /* | |
829 | * Data structures for policy management. | |
830 | * Each device can have a policy structure that lists | |
831 | * various name/value pairs each possibly with a metadata associated. | |
832 | * The policy list is sorted by name/value/metadata | |
833 | */ | |
834 | struct dev_policy { | |
835 | struct dev_policy *next; | |
836 | char *name; /* None of these strings are allocated. They are | |
837 | * all just references to strings which are known | |
838 | * to exist elsewhere. | |
839 | * name and metadata can be compared by address equality. | |
840 | */ | |
841 | const char *metadata; | |
842 | const char *value; | |
843 | }; | |
844 | ||
845 | extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[]; | |
846 | ||
847 | /* iterate over the sublist starting at list, having the same | |
848 | * 'name' as 'list', and matching the given metadata (Where | |
849 | * NULL matches anything | |
850 | */ | |
851 | #define pol_for_each(item, list, _metadata) \ | |
852 | for (item = list; \ | |
853 | item && item->name == list->name; \ | |
854 | item = item->next) \ | |
855 | if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \ | |
856 | ; else | |
857 | ||
858 | /* | |
859 | * policy records read from mdadm are largely just name-value pairs. | |
860 | * The names are constants, not strdupped | |
861 | */ | |
862 | struct pol_rule { | |
863 | struct pol_rule *next; | |
864 | char *type; /* rule_policy or rule_part */ | |
865 | struct rule { | |
866 | struct rule *next; | |
867 | char *name; | |
868 | char *value; | |
869 | char *dups; /* duplicates of 'value' with a partNN appended */ | |
870 | } *rule; | |
871 | }; | |
872 | ||
873 | extern char rule_policy[], rule_part[]; | |
874 | extern char rule_path[], rule_type[]; | |
875 | extern char type_part[], type_disk[]; | |
876 | ||
877 | extern void policyline(char *line, char *type); | |
878 | extern void policy_add(char *type, ...); | |
879 | extern void policy_free(void); | |
880 | ||
881 | extern struct dev_policy *path_policy(char *path, char *type); | |
882 | extern struct dev_policy *disk_policy(struct mdinfo *disk); | |
883 | extern struct dev_policy *devnum_policy(int dev); | |
884 | extern void dev_policy_free(struct dev_policy *p); | |
885 | ||
886 | //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata); | |
887 | extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata); | |
888 | extern struct dev_policy *pol_find(struct dev_policy *pol, char *name); | |
889 | ||
890 | enum policy_action { | |
891 | act_default, | |
892 | act_include, | |
893 | act_re_add, | |
894 | act_spare, /* This only applies to bare devices */ | |
895 | act_spare_same_slot, /* this allows non-bare devices, | |
896 | * but only if recent removal */ | |
897 | act_force_spare, /* this allow non-bare devices in any case */ | |
898 | act_err | |
899 | }; | |
900 | ||
901 | extern int policy_action_allows(struct dev_policy *plist, const char *metadata, | |
902 | enum policy_action want); | |
903 | extern int disk_action_allows(struct mdinfo *disk, const char *metadata, | |
904 | enum policy_action want); | |
905 | ||
906 | struct domainlist { | |
907 | struct domainlist *next; | |
908 | const char *dom; | |
909 | }; | |
910 | ||
911 | extern int domain_test(struct domainlist *dom, struct dev_policy *pol, | |
912 | const char *metadata); | |
913 | extern struct domainlist *domain_from_array(struct mdinfo *mdi, | |
914 | const char *metadata); | |
915 | extern void domainlist_add_dev(struct domainlist **dom, int devnum, | |
916 | const char *metadata); | |
917 | extern void domain_free(struct domainlist *dl); | |
918 | extern void domain_merge(struct domainlist **domp, struct dev_policy *pol, | |
919 | const char *metadata); | |
920 | void domain_add(struct domainlist **domp, char *domain); | |
921 | ||
922 | extern void policy_save_path(char *id_path, struct map_ent *array); | |
923 | extern int policy_check_path(struct mdinfo *disk, struct map_ent *array); | |
924 | ||
925 | #if __GNUC__ < 3 | |
926 | struct stat64; | |
927 | #endif | |
928 | ||
929 | #define HAVE_NFTW we assume | |
930 | #define HAVE_FTW | |
931 | ||
932 | #ifdef __UCLIBC__ | |
933 | # include <features.h> | |
934 | # ifndef __UCLIBC_HAS_LFS__ | |
935 | # define lseek64 lseek | |
936 | # endif | |
937 | # ifndef __UCLIBC_HAS_FTW__ | |
938 | # undef HAVE_FTW | |
939 | # undef HAVE_NFTW | |
940 | # endif | |
941 | #endif | |
942 | ||
943 | #ifdef __dietlibc__ | |
944 | # undef HAVE_NFTW | |
945 | #endif | |
946 | ||
947 | #if defined(__KLIBC__) | |
948 | # undef HAVE_NFTW | |
949 | # undef HAVE_FTW | |
950 | #endif | |
951 | ||
952 | #ifndef HAVE_NFTW | |
953 | # define FTW_PHYS 1 | |
954 | # ifndef HAVE_FTW | |
955 | struct FTW {}; | |
956 | # endif | |
957 | #endif | |
958 | ||
959 | #ifdef HAVE_FTW | |
960 | # include <ftw.h> | |
961 | #endif | |
962 | ||
963 | extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s); | |
964 | ||
965 | ||
966 | extern int Manage_ro(char *devname, int fd, int readonly); | |
967 | extern int Manage_runstop(char *devname, int fd, int runstop, int quiet); | |
968 | extern int Manage_resize(char *devname, int fd, long long size, int raid_disks); | |
969 | extern int Manage_subdevs(char *devname, int fd, | |
970 | struct mddev_dev *devlist, int verbose, int test, | |
971 | char *update); | |
972 | extern int autodetect(void); | |
973 | extern int Grow_Add_device(char *devname, int fd, char *newdev); | |
974 | extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force); | |
975 | extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file, | |
976 | long long size, | |
977 | int level, char *layout_str, int chunksize, int raid_disks, | |
978 | int force); | |
979 | extern int Grow_restart(struct supertype *st, struct mdinfo *info, | |
980 | int *fdlist, int cnt, char *backup_file, int verbose); | |
981 | extern int Grow_continue(int mdfd, struct supertype *st, | |
982 | struct mdinfo *info, char *backup_file); | |
983 | ||
984 | extern int Assemble(struct supertype *st, char *mddev, | |
985 | struct mddev_ident *ident, | |
986 | struct mddev_dev *devlist, | |
987 | char *backup_file, int invalid_backup, | |
988 | int readonly, int runstop, | |
989 | char *update, char *homehost, int require_homehost, | |
990 | int verbose, int force); | |
991 | ||
992 | extern int Build(char *mddev, int chunk, int level, int layout, | |
993 | int raiddisks, struct mddev_dev *devlist, int assume_clean, | |
994 | char *bitmap_file, int bitmap_chunk, int write_behind, | |
995 | int delay, int verbose, int autof, unsigned long long size); | |
996 | ||
997 | ||
998 | extern int Create(struct supertype *st, char *mddev, | |
999 | int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks, | |
1000 | char *name, char *homehost, int *uuid, | |
1001 | int subdevs, struct mddev_dev *devlist, | |
1002 | int runstop, int verbose, int force, int assume_clean, | |
1003 | char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof); | |
1004 | ||
1005 | extern int Detail(char *dev, int brief, int export, int test, char *homehost); | |
1006 | extern int Detail_Platform(struct superswitch *ss, int scan, int verbose); | |
1007 | extern int Query(char *dev); | |
1008 | extern int Examine(struct mddev_dev *devlist, int brief, int export, int scan, | |
1009 | int SparcAdjust, struct supertype *forcest, char *homehost); | |
1010 | extern int Monitor(struct mddev_dev *devlist, | |
1011 | char *mailaddr, char *alert_cmd, | |
1012 | int period, int daemonise, int scan, int oneshot, | |
1013 | int dosyslog, int test, char *pidfile, int increments, | |
1014 | int share); | |
1015 | ||
1016 | extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl); | |
1017 | extern int Kill_subarray(char *dev, char *subarray, int quiet); | |
1018 | extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet); | |
1019 | extern int Wait(char *dev); | |
1020 | extern int WaitClean(char *dev, int sock, int verbose); | |
1021 | ||
1022 | extern int Incremental(char *devname, int verbose, int runstop, | |
1023 | struct supertype *st, char *homehost, int require_homehost, | |
1024 | int autof); | |
1025 | extern void RebuildMap(void); | |
1026 | extern int IncrementalScan(int verbose); | |
1027 | extern int IncrementalRemove(char *devname, char *path, int verbose); | |
1028 | extern int CreateBitmap(char *filename, int force, char uuid[16], | |
1029 | unsigned long chunksize, unsigned long daemon_sleep, | |
1030 | unsigned long write_behind, | |
1031 | unsigned long long array_size, | |
1032 | int major); | |
1033 | extern int ExamineBitmap(char *filename, int brief, struct supertype *st); | |
1034 | extern int bitmap_update_uuid(int fd, int *uuid, int swap); | |
1035 | extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb); | |
1036 | ||
1037 | extern int md_get_version(int fd); | |
1038 | extern int get_linux_version(void); | |
1039 | extern int mdadm_version(char *version); | |
1040 | extern long long parse_size(char *size); | |
1041 | extern int parse_uuid(char *str, int uuid[4]); | |
1042 | extern int parse_layout_10(char *layout); | |
1043 | extern int parse_layout_faulty(char *layout); | |
1044 | extern int check_ext2(int fd, char *name); | |
1045 | extern int check_reiser(int fd, char *name); | |
1046 | extern int check_raid(int fd, char *name); | |
1047 | extern int check_partitions(int fd, char *dname, unsigned long long freesize); | |
1048 | ||
1049 | extern int get_mdp_major(void); | |
1050 | extern int dev_open(char *dev, int flags); | |
1051 | extern int open_dev(int devnum); | |
1052 | extern int open_dev_excl(int devnum); | |
1053 | extern int is_standard(char *dev, int *nump); | |
1054 | extern int same_dev(char *one, char *two); | |
1055 | ||
1056 | extern int parse_auto(char *str, char *msg, int config); | |
1057 | extern struct mddev_ident *conf_get_ident(char *dev); | |
1058 | extern struct mddev_dev *conf_get_devs(void); | |
1059 | extern int conf_test_dev(char *devname); | |
1060 | extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost); | |
1061 | extern struct createinfo *conf_get_create_info(void); | |
1062 | extern void set_conffile(char *file); | |
1063 | extern char *conf_get_mailaddr(void); | |
1064 | extern char *conf_get_mailfrom(void); | |
1065 | extern char *conf_get_program(void); | |
1066 | extern char *conf_get_homehost(int *require_homehostp); | |
1067 | extern char *conf_line(FILE *file); | |
1068 | extern char *conf_word(FILE *file, int allow_key); | |
1069 | extern int conf_name_is_free(char *name); | |
1070 | extern int devname_matches(char *name, char *match); | |
1071 | extern struct mddev_ident *conf_match(struct mdinfo *info, struct supertype *st); | |
1072 | extern int experimental(void); | |
1073 | ||
1074 | extern void free_line(char *line); | |
1075 | extern int match_oneof(char *devices, char *devname); | |
1076 | extern void uuid_from_super(int uuid[4], mdp_super_t *super); | |
1077 | extern const int uuid_match_any[4]; | |
1078 | extern int same_uuid(int a[4], int b[4], int swapuuid); | |
1079 | extern void copy_uuid(void *a, int b[4], int swapuuid); | |
1080 | extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep); | |
1081 | extern char *fname_from_uuid(struct supertype *st, | |
1082 | struct mdinfo *info, char *buf, char sep); | |
1083 | extern unsigned long calc_csum(void *super, int bytes); | |
1084 | extern int enough(int level, int raid_disks, int layout, int clean, | |
1085 | char *avail, int avail_disks); | |
1086 | extern int enough_fd(int fd); | |
1087 | extern int ask(char *mesg); | |
1088 | extern unsigned long long get_component_size(int fd); | |
1089 | extern void remove_partitions(int fd); | |
1090 | extern int test_partition(int fd); | |
1091 | extern unsigned long long calc_array_size(int level, int raid_disks, int layout, | |
1092 | int chunksize, unsigned long long devsize); | |
1093 | extern int flush_metadata_updates(struct supertype *st); | |
1094 | extern void append_metadata_update(struct supertype *st, void *buf, int len); | |
1095 | extern int assemble_container_content(struct supertype *st, int mdfd, | |
1096 | struct mdinfo *content, int runstop, | |
1097 | char *chosen_name, int verbose); | |
1098 | ||
1099 | extern int add_disk(int mdfd, struct supertype *st, | |
1100 | struct mdinfo *sra, struct mdinfo *info); | |
1101 | extern int remove_disk(int mdfd, struct supertype *st, | |
1102 | struct mdinfo *sra, struct mdinfo *info); | |
1103 | extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info); | |
1104 | unsigned long long min_recovery_start(struct mdinfo *array); | |
1105 | ||
1106 | extern char *human_size(long long bytes); | |
1107 | extern char *human_size_brief(long long bytes); | |
1108 | extern void print_r10_layout(int layout); | |
1109 | ||
1110 | #define NoMdDev (1<<23) | |
1111 | extern int find_free_devnum(int use_partitions); | |
1112 | ||
1113 | extern void put_md_name(char *name); | |
1114 | extern char *get_md_name(int dev); | |
1115 | ||
1116 | extern char DefaultConfFile[]; | |
1117 | ||
1118 | extern int create_mddev(char *dev, char *name, int autof, int trustworthy, | |
1119 | char *chosen); | |
1120 | /* values for 'trustworthy' */ | |
1121 | #define LOCAL 1 | |
1122 | #define LOCAL_ANY 10 | |
1123 | #define FOREIGN 2 | |
1124 | #define METADATA 3 | |
1125 | extern int open_mddev(char *dev, int report_errors); | |
1126 | extern int open_container(int fd); | |
1127 | extern int is_container_member(struct mdstat_ent *ent, char *devname); | |
1128 | extern int is_subarray_active(char *subarray, char *devname); | |
1129 | int is_container_active(char *devname); | |
1130 | extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet); | |
1131 | extern struct superswitch *version_to_superswitch(char *vers); | |
1132 | ||
1133 | extern int mdmon_running(int devnum); | |
1134 | extern int mdmon_pid(int devnum); | |
1135 | extern int check_env(char *name); | |
1136 | extern __u32 random32(void); | |
1137 | extern int start_mdmon(int devnum); | |
1138 | ||
1139 | extern char *devnum2devname(int num); | |
1140 | extern void fmt_devname(char *name, int num); | |
1141 | extern int devname2devnum(char *name); | |
1142 | extern int stat2devnum(struct stat *st); | |
1143 | extern int fd2devnum(int fd); | |
1144 | ||
1145 | static inline int dev2major(int d) | |
1146 | { | |
1147 | if (d >= 0) | |
1148 | return MD_MAJOR; | |
1149 | else | |
1150 | return get_mdp_major(); | |
1151 | } | |
1152 | ||
1153 | static inline int dev2minor(int d) | |
1154 | { | |
1155 | if (d >= 0) | |
1156 | return d; | |
1157 | return (-1-d) << MdpMinorShift; | |
1158 | } | |
1159 | ||
1160 | static inline int ROUND_UP(int a, int base) | |
1161 | { | |
1162 | return ((a+base-1)/base)*base; | |
1163 | } | |
1164 | ||
1165 | static inline int is_subarray(char *vers) | |
1166 | { | |
1167 | /* The version string for a 'subarray' (an array in a container) | |
1168 | * is | |
1169 | * /containername/componentname for normal read-write arrays | |
1170 | * -containername/componentname for arrays which mdmon must not | |
1171 | * reconfigure. They might be read-only | |
1172 | * or might be undergoing reshape etc. | |
1173 | * containername is e.g. md0, md_d1 | |
1174 | * componentname is dependant on the metadata. e.g. '1' 'S1' ... | |
1175 | */ | |
1176 | return (*vers == '/' || *vers == '-'); | |
1177 | } | |
1178 | ||
1179 | static inline char *to_subarray(struct mdstat_ent *ent, char *container) | |
1180 | { | |
1181 | return &ent->metadata_version[10+strlen(container)+1]; | |
1182 | } | |
1183 | ||
1184 | #ifdef DEBUG | |
1185 | #define dprintf(fmt, arg...) \ | |
1186 | fprintf(stderr, fmt, ##arg) | |
1187 | #else | |
1188 | #define dprintf(fmt, arg...) \ | |
1189 | ({ if (0) fprintf(stderr, fmt, ##arg); 0; }) | |
1190 | #endif | |
1191 | #include <assert.h> | |
1192 | #include <stdarg.h> | |
1193 | static inline int xasprintf(char **strp, const char *fmt, ...) { | |
1194 | va_list ap; | |
1195 | int ret; | |
1196 | va_start(ap, fmt); | |
1197 | ret = vasprintf(strp, fmt, ap); | |
1198 | va_end(ap); | |
1199 | assert(ret >= 0); | |
1200 | return ret; | |
1201 | } | |
1202 | ||
1203 | #define LEVEL_MULTIPATH (-4) | |
1204 | #define LEVEL_LINEAR (-1) | |
1205 | #define LEVEL_FAULTY (-5) | |
1206 | ||
1207 | /* kernel module doesn't know about these */ | |
1208 | #define LEVEL_CONTAINER (-100) | |
1209 | #define LEVEL_UNSUPPORTED (-200) | |
1210 | ||
1211 | ||
1212 | /* faulty stuff */ | |
1213 | ||
1214 | #define WriteTransient 0 | |
1215 | #define ReadTransient 1 | |
1216 | #define WritePersistent 2 | |
1217 | #define ReadPersistent 3 | |
1218 | #define WriteAll 4 /* doesn't go to device */ | |
1219 | #define ReadFixable 5 | |
1220 | #define Modes 6 | |
1221 | ||
1222 | #define ClearErrors 31 | |
1223 | #define ClearFaults 30 | |
1224 | ||
1225 | #define AllPersist 100 /* internal use only */ | |
1226 | #define NoPersist 101 | |
1227 | ||
1228 | #define ModeMask 0x1f | |
1229 | #define ModeShift 5 | |
1230 | ||
1231 | ||
1232 | #ifdef __TINYC__ | |
1233 | #undef minor | |
1234 | #undef major | |
1235 | #undef makedev | |
1236 | #define minor(x) ((x)&0xff) | |
1237 | #define major(x) (((x)>>8)&0xff) | |
1238 | #define makedev(M,m) (((M)<<8) | (m)) | |
1239 | #endif | |
1240 | ||
1241 | /* for raid4/5/6 */ | |
1242 | #define ALGORITHM_LEFT_ASYMMETRIC 0 | |
1243 | #define ALGORITHM_RIGHT_ASYMMETRIC 1 | |
1244 | #define ALGORITHM_LEFT_SYMMETRIC 2 | |
1245 | #define ALGORITHM_RIGHT_SYMMETRIC 3 | |
1246 | ||
1247 | /* Define non-rotating (raid4) algorithms. These allow | |
1248 | * conversion of raid4 to raid5. | |
1249 | */ | |
1250 | #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */ | |
1251 | #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */ | |
1252 | ||
1253 | /* DDF RAID6 layouts differ from md/raid6 layouts in two ways. | |
1254 | * Firstly, the exact positioning of the parity block is slightly | |
1255 | * different between the 'LEFT_*' modes of md and the "_N_*" modes | |
1256 | * of DDF. | |
1257 | * Secondly, or order of datablocks over which the Q syndrome is computed | |
1258 | * is different. | |
1259 | * Consequently we have different layouts for DDF/raid6 than md/raid6. | |
1260 | * These layouts are from the DDFv1.2 spec. | |
1261 | * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but | |
1262 | * leaves RLQ=3 as 'Vendor Specific' | |
1263 | */ | |
1264 | ||
1265 | #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */ | |
1266 | #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */ | |
1267 | #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */ | |
1268 | ||
1269 | ||
1270 | /* For every RAID5 algorithm we define a RAID6 algorithm | |
1271 | * with exactly the same layout for data and parity, and | |
1272 | * with the Q block always on the last device (N-1). | |
1273 | * This allows trivial conversion from RAID5 to RAID6 | |
1274 | */ | |
1275 | #define ALGORITHM_LEFT_ASYMMETRIC_6 16 | |
1276 | #define ALGORITHM_RIGHT_ASYMMETRIC_6 17 | |
1277 | #define ALGORITHM_LEFT_SYMMETRIC_6 18 | |
1278 | #define ALGORITHM_RIGHT_SYMMETRIC_6 19 | |
1279 | #define ALGORITHM_PARITY_0_6 20 | |
1280 | #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N | |
1281 | ||
1282 | /* Define PATH_MAX in case we don't use glibc or standard library does | |
1283 | * not have PATH_MAX defined. Assume max path length is 4K characters. | |
1284 | */ | |
1285 | #ifndef PATH_MAX | |
1286 | #define PATH_MAX 4096 | |
1287 | #endif | |
1288 |