]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Monitor, Incremental: use device policies
[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 #ifdef __GLIBC__
29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
30 #elif !defined(lseek64)
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
33 # endif
34 #endif
35
36 #include <assert.h>
37 #include <sys/types.h>
38 #include <sys/stat.h>
39 #include <stdarg.h>
40 #include <stdint.h>
41 #include <stdlib.h>
42 #include <time.h>
43 #include <sys/time.h>
44 #include <getopt.h>
45 #include <fcntl.h>
46 #include <stdio.h>
47 #include <errno.h>
48 #include <string.h>
49 #include <syslog.h>
50 #include <stdbool.h>
51 #include <signal.h>
52 /* Newer glibc requires sys/sysmacros.h directly for makedev() */
53 #include <sys/sysmacros.h>
54 #ifdef __dietlibc__
55 #include <strings.h>
56 /* dietlibc has deprecated random and srandom!! */
57 #define random rand
58 #define srandom srand
59 #endif
60
61 #ifdef NO_COROSYNC
62 #define CS_OK 1
63 typedef uint64_t cmap_handle_t;
64 #else
65 #include <corosync/cmap.h>
66 #endif
67
68 #ifndef NO_DLM
69 #include <libdlm.h>
70 #include <errno.h>
71 #else
72 #define LKF_NOQUEUE 0x00000001
73 #define LKM_PWMODE 4
74 #define EUNLOCK 0x10002
75
76 typedef void *dlm_lshandle_t;
77
78 struct dlm_lksb {
79 int sb_status;
80 uint32_t sb_lkid;
81 char sb_flags;
82 char *sb_lvbptr;
83 };
84 #endif
85
86 #include <linux/kdev_t.h>
87 /*#include <linux/fs.h> */
88 #include <sys/mount.h>
89 #include <asm/types.h>
90 #include <sys/ioctl.h>
91 #define MD_MAJOR 9
92 #define MdpMinorShift 6
93
94 #ifndef BLKGETSIZE64
95 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
96 #endif
97
98 #define DEFAULT_CHUNK 512
99 #define DEFAULT_BITMAP_CHUNK 4096
100 #define DEFAULT_BITMAP_DELAY 5
101 #define DEFAULT_MAX_WRITE_BEHIND 256
102
103 /* DEV_NUM_PREF is a subpath to numbered MD devices, e.g. /dev/md1 or directory name.
104 * DEV_NUM_PREF_LEN is a length with Null byte excluded.
105 */
106 #ifndef DEV_NUM_PREF
107 #define DEV_NUM_PREF "/dev/md"
108 #define DEV_NUM_PREF_LEN (sizeof(DEV_NUM_PREF) - 1)
109 #endif /* DEV_NUM_PREF */
110
111 /* DEV_MD_DIR points to named MD devices directory.
112 * DEV_MD_DIR_LEN is a length with Null byte excluded.
113 */
114 #ifndef DEV_MD_DIR
115 #define DEV_MD_DIR "/dev/md/"
116 #define DEV_MD_DIR_LEN (sizeof(DEV_MD_DIR) - 1)
117 #endif /* DEV_MD_DIR */
118
119 /* MAP_DIR should be somewhere that persists across the pivotroot
120 * from early boot to late boot.
121 * /run seems to have emerged as the best standard.
122 */
123 #ifndef MAP_DIR
124 #define MAP_DIR "/run/mdadm"
125 #endif /* MAP_DIR */
126 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
127 * want something other than the default of "map"
128 */
129 #ifndef MAP_FILE
130 #define MAP_FILE "map"
131 #endif /* MAP_FILE */
132 /* MDMON_DIR is where pid and socket files used for communicating
133 * with mdmon normally live. Best is /var/run/mdadm as
134 * mdmon is needed at early boot then it needs to write there prior
135 * to /var/run being mounted read/write, and it also then needs to
136 * persist beyond when /var/run is mounter read-only. So, to be
137 * safe, the default is somewhere that is read/write early in the
138 * boot process and stays up as long as possible during shutdown.
139 */
140 #ifndef MDMON_DIR
141 #define MDMON_DIR "/run/mdadm"
142 #endif /* MDMON_DIR */
143
144 /* FAILED_SLOTS is where to save files storing recent removal of array
145 * member in order to allow future reuse of disk inserted in the same
146 * slot for array recovery
147 */
148 #ifndef FAILED_SLOTS_DIR
149 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
150 #endif /* FAILED_SLOTS */
151
152 #ifndef MDMON_SERVICE
153 #define MDMON_SERVICE "mdmon"
154 #endif /* MDMON_SERVICE */
155
156 #ifndef GROW_SERVICE
157 #define GROW_SERVICE "mdadm-grow-continue"
158 #endif /* GROW_SERVICE */
159
160 #include "md_u.h"
161 #include "md_p.h"
162 #include "bitmap.h"
163 #include "msg.h"
164
165 #include <endian.h>
166 /* Redhat don't like to #include <asm/byteorder.h>, and
167 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
168 * and there is no standard conversion function so... */
169 /* And dietlibc doesn't think byteswap is ok, so.. */
170 /* #include <byteswap.h> */
171 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
172 ((x) & 0xff00U) >> 8)
173 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
174 ((x) & 0xff000000U) >> 24 | \
175 ((x) & 0x0000ff00U) << 8 | \
176 ((x) & 0x00ff0000U) >> 8)
177 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
178 ((x) & 0xff00000000000000ULL) >> 56 | \
179 ((x) & 0x000000000000ff00ULL) << 40 | \
180 ((x) & 0x00ff000000000000ULL) >> 40 | \
181 ((x) & 0x0000000000ff0000ULL) << 24 | \
182 ((x) & 0x0000ff0000000000ULL) >> 24 | \
183 ((x) & 0x00000000ff000000ULL) << 8 | \
184 ((x) & 0x000000ff00000000ULL) >> 8)
185
186 #if !defined(__KLIBC__)
187 #if BYTE_ORDER == LITTLE_ENDIAN
188 #define __cpu_to_le16(_x) (unsigned int)(_x)
189 #define __cpu_to_le32(_x) (unsigned int)(_x)
190 #define __cpu_to_le64(_x) (unsigned long long)(_x)
191 #define __le16_to_cpu(_x) (unsigned int)(_x)
192 #define __le32_to_cpu(_x) (unsigned int)(_x)
193 #define __le64_to_cpu(_x) (unsigned long long)(_x)
194
195 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
196 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
197 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
198 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
199 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
200 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
201 #elif BYTE_ORDER == BIG_ENDIAN
202 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
203 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
204 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
205 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
206 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
207 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
208
209 #define __cpu_to_be16(_x) (unsigned int)(_x)
210 #define __cpu_to_be32(_x) (unsigned int)(_x)
211 #define __cpu_to_be64(_x) (unsigned long long)(_x)
212 #define __be16_to_cpu(_x) (unsigned int)(_x)
213 #define __be32_to_cpu(_x) (unsigned int)(_x)
214 #define __be64_to_cpu(_x) (unsigned long long)(_x)
215 #else
216 # error "unknown endianness."
217 #endif
218 #endif /* __KLIBC__ */
219
220 /*
221 * Partially stolen from include/linux/unaligned/packed_struct.h
222 */
223 struct __una_u16 { __u16 x; } __attribute__ ((packed));
224 struct __una_u32 { __u32 x; } __attribute__ ((packed));
225
226 static inline __u16 __get_unaligned16(const void *p)
227 {
228 const struct __una_u16 *ptr = (const struct __una_u16 *)p;
229 return ptr->x;
230 }
231
232 static inline __u32 __get_unaligned32(const void *p)
233 {
234 const struct __una_u32 *ptr = (const struct __una_u32 *)p;
235 return ptr->x;
236 }
237
238 static inline void __put_unaligned16(__u16 val, void *p)
239 {
240 struct __una_u16 *ptr = (struct __una_u16 *)p;
241 ptr->x = val;
242 }
243
244 static inline void __put_unaligned32(__u32 val, void *p)
245 {
246 struct __una_u32 *ptr = (struct __una_u32 *)p;
247 ptr->x = val;
248 }
249
250 /*
251 * Check at compile time that something is of a particular type.
252 * Always evaluates to 1 so you may use it easily in comparisons.
253 */
254
255 #define typecheck(type,x) \
256 ({ type __dummy; \
257 typeof(x) __dummy2; \
258 (void)(&__dummy == &__dummy2); \
259 1; \
260 })
261
262 /*
263 * These inlines deal with timer wrapping correctly.
264 *
265 * time_after(a,b) returns true if the time a is after time b.
266 */
267
268 #define time_after(a,b) \
269 (typecheck(unsigned int, a) && \
270 typecheck(unsigned int, b) && \
271 ((int)((b) - (a)) < 0))
272
273 #define time_before(a,b) time_after(b,a)
274
275 /*
276 * min()/max()/clamp() macros that also do
277 * strict type-checking.. See the
278 * "unnecessary" pointer comparison.
279 */
280 #define min(x, y) ({ \
281 typeof(x) _min1 = (x); \
282 typeof(y) _min2 = (y); \
283 (void) (&_min1 == &_min2); \
284 _min1 < _min2 ? _min1 : _min2; })
285
286 #define max(x, y) ({ \
287 typeof(x) _max1 = (x); \
288 typeof(y) _max2 = (y); \
289 (void) (&_max1 == &_max2); \
290 _max1 > _max2 ? _max1 : _max2; })
291
292 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
293
294 #define KIB_TO_BYTES(x) ((x) << 10)
295 #define SEC_TO_BYTES(x) ((x) << 9)
296
297 /**
298 * This is true for native and DDF, IMSM allows 16.
299 */
300 #define MD_NAME_MAX 32
301
302 extern const char Name[];
303
304 struct md_bb_entry {
305 unsigned long long sector;
306 int length;
307 };
308
309 struct md_bb {
310 int supported;
311 int count;
312 struct md_bb_entry *entries;
313 };
314
315 /* general information that might be extracted from a superblock */
316 struct mdinfo {
317 mdu_array_info_t array;
318 mdu_disk_info_t disk;
319 __u64 events;
320 int uuid[4];
321 char name[33];
322 unsigned long long data_offset;
323 unsigned long long new_data_offset;
324 unsigned long long component_size; /* same as array.size, except in
325 * sectors and up to 64bits.
326 */
327 unsigned long long custom_array_size; /* size for non-default sized
328 * arrays (in sectors)
329 */
330 #define NO_RESHAPE 0
331 #define VOLUME_RESHAPE 1
332 #define CONTAINER_RESHAPE 2
333 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
334 int reshape_active;
335 unsigned long long reshape_progress;
336 int recovery_blocked; /* for external metadata it
337 * indicates that there is
338 * reshape in progress in
339 * container,
340 * for native metadata it is
341 * reshape_active field mirror
342 */
343 int journal_device_required;
344 int journal_clean;
345
346 enum {
347 CONSISTENCY_POLICY_UNKNOWN,
348 CONSISTENCY_POLICY_NONE,
349 CONSISTENCY_POLICY_RESYNC,
350 CONSISTENCY_POLICY_BITMAP,
351 CONSISTENCY_POLICY_JOURNAL,
352 CONSISTENCY_POLICY_PPL,
353 } consistency_policy;
354
355 /* During reshape we can sometimes change the data_offset to avoid
356 * over-writing still-valid data. We need to know if there is space.
357 * So getinfo_super will fill in space_before and space_after in sectors.
358 * data_offset can be increased or decreased by this amount.
359 */
360 unsigned long long space_before, space_after;
361 union {
362 unsigned long long resync_start; /* per-array resync position */
363 unsigned long long recovery_start; /* per-device rebuild position */
364 #define MaxSector (~0ULL) /* resync/recovery complete position */
365 };
366 long bitmap_offset; /* 0 == none, 1 == a file */
367 unsigned int ppl_size;
368 int ppl_offset;
369 unsigned long long ppl_sector;
370 unsigned long safe_mode_delay; /* ms delay to mark clean */
371 int new_level, delta_disks, new_layout, new_chunk;
372 int errors;
373 unsigned long cache_size; /* size of raid456 stripe cache*/
374 int mismatch_cnt;
375 char text_version[50];
376
377 int container_member; /* for assembling external-metatdata arrays
378 * This is to be used internally by metadata
379 * handler only */
380 /**
381 * flag external handlers can set to indicate that subarrays have:
382 * - not enough disks to start (-1),
383 * - enough disks to start (0),
384 * - all expected disks (1).
385 */
386 int container_enough;
387 char sys_name[32];
388 struct mdinfo *devs;
389 struct mdinfo *next;
390
391 /* Device info for mdmon: */
392 int recovery_fd;
393 int state_fd;
394 int bb_fd;
395 int ubb_fd;
396 #define DS_FAULTY 1
397 #define DS_INSYNC 2
398 #define DS_WRITE_MOSTLY 4
399 #define DS_SPARE 8
400 #define DS_BLOCKED 16
401 #define DS_REMOVE 1024
402 #define DS_UNBLOCK 2048
403 int prev_state, curr_state, next_state;
404
405 /* info read from sysfs */
406 enum {
407 ARRAY_CLEAR,
408 ARRAY_INACTIVE,
409 ARRAY_SUSPENDED,
410 ARRAY_READONLY,
411 ARRAY_READ_AUTO,
412 ARRAY_CLEAN,
413 ARRAY_ACTIVE,
414 ARRAY_WRITE_PENDING,
415 ARRAY_ACTIVE_IDLE,
416 ARRAY_BROKEN,
417 ARRAY_UNKNOWN_STATE,
418 } array_state;
419 struct md_bb bb;
420 };
421
422 struct createinfo {
423 int uid;
424 int gid;
425 int autof;
426 int mode;
427 int names;
428 int bblist;
429 struct supertype *supertype;
430 };
431
432 struct spare_criteria {
433 bool criteria_set;
434 unsigned long long min_size;
435 unsigned int sector_size;
436 struct dev_policy *pols;
437 };
438
439 typedef enum mdadm_status {
440 MDADM_STATUS_SUCCESS = 0,
441 MDADM_STATUS_ERROR,
442 MDADM_STATUS_UNDEF,
443 } mdadm_status_t;
444
445 enum mode {
446 ASSEMBLE=1,
447 BUILD,
448 CREATE,
449 MANAGE,
450 MISC,
451 MONITOR,
452 GROW,
453 INCREMENTAL,
454 AUTODETECT,
455 mode_count
456 };
457
458 extern char short_options[];
459 extern char short_monitor_options[];
460 extern char short_bitmap_options[];
461 extern char short_bitmap_auto_options[];
462 extern struct option long_options[];
463 extern char Version[], Usage[], Help[], OptionHelp[],
464 *mode_help[],
465 Help_create[], Help_build[], Help_assemble[], Help_grow[],
466 Help_incr[],
467 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
468
469 /* for option that don't have short equivilents, we assign arbitrary
470 * numbers later than any 'short' character option.
471 */
472 enum special_options {
473 AssumeClean = 300,
474 WriteZeroes,
475 BitmapChunk,
476 WriteBehind,
477 ReAdd,
478 NoDegraded,
479 Sparc22,
480 BackupFile,
481 HomeHost,
482 AutoHomeHost,
483 AutoDetect,
484 Waitclean,
485 DetailPlatform,
486 KillSubarray,
487 UpdateSubarray,
488 IncrementalPath,
489 NoSharing,
490 HelpOptions,
491 Brief,
492 NoDevices,
493 ManageOpt,
494 Add,
495 AddSpare,
496 AddJournal,
497 Remove,
498 Fail,
499 Replace,
500 With,
501 MiscOpt,
502 WaitOpt,
503 ConfigFile,
504 ChunkSize,
505 WriteMostly,
506 FailFast,
507 NoFailFast,
508 Layout,
509 Auto,
510 Force,
511 SuperMinor,
512 EMail,
513 ProgramOpt,
514 Increment,
515 Fork,
516 Bitmap,
517 RebuildMapOpt,
518 InvalidBackup,
519 UdevRules,
520 FreezeReshape,
521 Continue,
522 OffRootOpt,
523 Prefer,
524 KillOpt,
525 DataOffset,
526 ExamineBB,
527 Dump,
528 Restore,
529 Action,
530 Nodes,
531 ClusterName,
532 ClusterConfirm,
533 WriteJournal,
534 ConsistencyPolicy,
535 };
536
537 enum update_opt {
538 UOPT_NAME = 1,
539 UOPT_PPL,
540 UOPT_NO_PPL,
541 UOPT_BITMAP,
542 UOPT_NO_BITMAP,
543 UOPT_SUBARRAY_ONLY,
544 UOPT_SPARC22,
545 UOPT_SUPER_MINOR,
546 UOPT_SUMMARIES,
547 UOPT_RESYNC,
548 UOPT_UUID,
549 UOPT_HOMEHOST,
550 UOPT_HOME_CLUSTER,
551 UOPT_NODES,
552 UOPT_DEVICESIZE,
553 UOPT_BBL,
554 UOPT_NO_BBL,
555 UOPT_FORCE_NO_BBL,
556 UOPT_METADATA,
557 UOPT_REVERT_RESHAPE,
558 UOPT_LAYOUT_ORIGINAL,
559 UOPT_LAYOUT_ALTERNATE,
560 UOPT_LAYOUT_UNSPECIFIED,
561 UOPT_BYTEORDER,
562 UOPT_HELP,
563 UOPT_USER_ONLY,
564 /*
565 * Code specific options, cannot be set by the user
566 */
567 UOPT_SPEC_FORCE_ONE,
568 UOPT_SPEC_FORCE_ARRAY,
569 UOPT_SPEC_ASSEMBLE,
570 UOPT_SPEC_LINEAR_GROW_NEW,
571 UOPT_SPEC_LINEAR_GROW_UPDATE,
572 UOPT_SPEC__RESHAPE_PROGRESS,
573 UOPT_SPEC_WRITEMOSTLY,
574 UOPT_SPEC_READWRITE,
575 UOPT_SPEC_FAILFAST,
576 UOPT_SPEC_NOFAILFAST,
577 UOPT_SPEC_REVERT_RESHAPE_NOBACKUP,
578 UOPT_UNDEFINED
579 };
580 extern void fprint_update_options(FILE *outf, enum update_opt update_mode);
581
582 enum prefix_standard {
583 JEDEC,
584 IEC
585 };
586
587 enum bitmap_update {
588 NoUpdate,
589 NameUpdate,
590 NodeNumUpdate,
591 };
592
593 enum flag_mode {
594 FlagDefault, FlagSet, FlagClear,
595 };
596
597 /* structures read from config file */
598 /* List of mddevice names and identifiers
599 * Identifiers can be:
600 * uuid=128-hex-uuid
601 * super-minor=decimal-minor-number-from-superblock
602 * devices=comma,separated,list,of,device,names,with,wildcards
603 *
604 * If multiple fields are present, the intersection of all matching
605 * devices is considered
606 */
607 #define UnSet (0xfffe)
608 struct mddev_ident {
609 char *devname;
610
611 int uuid_set;
612 int uuid[4];
613 char name[MD_NAME_MAX + 1];
614
615 int super_minor;
616
617 char *devices; /* comma separated list of device
618 * names with wild cards
619 */
620 int level;
621 int raid_disks;
622 int spare_disks;
623 struct supertype *st;
624 int autof; /* 1 for normal, 2 for partitioned */
625 char *spare_group;
626 char *bitmap_file;
627 int bitmap_fd;
628
629 char *container; /* /dev/whatever name of container, or
630 * uuid of container. You would expect
631 * this to be the 'devname' or UUID
632 * of some other entry.
633 */
634 char *member; /* subarray within a container */
635
636 struct mddev_ident *next;
637 union {
638 /* fields needed by different users of this structure */
639 int assembled; /* set when assembly succeeds */
640 };
641 };
642
643 struct context {
644 int readonly;
645 int runstop;
646 int verbose;
647 int brief;
648 int no_devices;
649 int force;
650 char *homehost;
651 int require_homehost;
652 char *prefer;
653 int export;
654 int test;
655 char *subarray;
656 enum update_opt update;
657 int scan;
658 int SparcAdjust;
659 int autof;
660 int delay;
661 int freeze_reshape;
662 char *backup_file;
663 int invalid_backup;
664 char *action;
665 int nodes;
666 char *homecluster;
667 };
668
669 struct shape {
670 int raiddisks;
671 int sparedisks;
672 int journaldisks;
673 int level;
674 int layout;
675 char *layout_str;
676 int chunk;
677 int bitmap_chunk;
678 char *bitmap_file;
679 int assume_clean;
680 bool write_zeroes;
681 int write_behind;
682 unsigned long long size;
683 unsigned long long data_offset;
684 int consistency_policy;
685 };
686
687 /* List of device names - wildcards expanded */
688 struct mddev_dev {
689 char *devname;
690 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
691 * 'A' for re_add.
692 * Not set for names read from .config
693 */
694 enum flag_mode writemostly;
695 enum flag_mode failfast;
696 int used; /* set when used */
697 long long data_offset;
698 struct mddev_dev *next;
699 };
700
701 typedef struct mapping {
702 char *name;
703 int num;
704 } mapping_t;
705
706 struct mdstat_ent {
707 char devnm[32];
708 int active;
709 char *level;
710 char *pattern; /* U for up, _ for down */
711 int percent; /* -1 if no resync */
712 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
713 int devcnt;
714 int raid_disks;
715 char * metadata_version;
716 struct dev_member {
717 char *name;
718 struct dev_member *next;
719 } *members;
720 struct mdstat_ent *next;
721 };
722
723 extern struct mdstat_ent *mdstat_read(int hold, int start);
724 extern void mdstat_close(void);
725 extern void free_mdstat(struct mdstat_ent *ms);
726 extern int mdstat_wait(int seconds);
727 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
728 extern int mddev_busy(char *devnm);
729 extern struct mdstat_ent *mdstat_by_component(char *name);
730 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
731
732 struct map_ent {
733 struct map_ent *next;
734 char devnm[32];
735 char metadata[20];
736 int uuid[4];
737 int bad;
738 char *path;
739 };
740 extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
741 int uuid[4], char *path);
742 extern void map_remove(struct map_ent **map, char *devnm);
743 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
744 extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
745 extern void map_free(struct map_ent *map);
746 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
747 extern void map_read(struct map_ent **melp);
748 extern int map_write(struct map_ent *mel);
749 extern void map_delete(struct map_ent **mapp, char *devnm);
750 extern void map_add(struct map_ent **melp,
751 char *devnm, char *metadata, int uuid[4], char *path);
752 extern int map_lock(struct map_ent **melp);
753 extern void map_unlock(struct map_ent **melp);
754 extern void map_fork(void);
755
756 /* various details can be requested */
757 enum sysfs_read_flags {
758 GET_LEVEL = (1 << 0),
759 GET_LAYOUT = (1 << 1),
760 GET_COMPONENT = (1 << 2),
761 GET_CHUNK = (1 << 3),
762 GET_CACHE = (1 << 4),
763 GET_MISMATCH = (1 << 5),
764 GET_VERSION = (1 << 6),
765 GET_DISKS = (1 << 7),
766 GET_SAFEMODE = (1 << 9),
767 GET_BITMAP_LOCATION = (1 << 10),
768
769 GET_DEVS = (1 << 20), /* gets role, major, minor */
770 GET_OFFSET = (1 << 21),
771 GET_SIZE = (1 << 22),
772 GET_STATE = (1 << 23),
773 GET_ERROR = (1 << 24),
774 GET_ARRAY_STATE = (1 << 25),
775 GET_CONSISTENCY_POLICY = (1 << 26),
776 GET_DEVS_ALL = (1 << 27),
777 };
778
779 #define SYSFS_MAX_BUF_SIZE 64
780
781 extern void sysfs_get_container_devnm(struct mdinfo *mdi, char *buf);
782
783 /* If fd >= 0, get the array it is open on,
784 * else use devnm.
785 */
786 extern int sysfs_open(char *devnm, char *devname, char *attr);
787 extern int sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
788 extern void sysfs_init_dev(struct mdinfo *mdi, dev_t devid);
789 extern void sysfs_free(struct mdinfo *sra);
790 extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
791 extern int sysfs_attr_match(const char *attr, const char *str);
792 extern int sysfs_match_word(const char *word, char **list);
793 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
794 char *name, char *val);
795 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
796 char *name, unsigned long long val);
797 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
798 char *name, long long val);
799 extern int sysfs_uevent(struct mdinfo *sra, char *event);
800 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
801 char *name);
802 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
803 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
804 char *name, unsigned long long *val);
805 extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
806 extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
807 char *name, unsigned long long *v1, unsigned long long *v2);
808 extern int sysfs_fd_get_str(int fd, char *val, int size);
809 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
810 char *name);
811 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
812 char *name, char *val, int size);
813 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
814 extern int sysfs_set_array(struct mdinfo *info, int vers);
815 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
816 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
817 extern int sysfs_unique_holder(char *devnm, long rdev);
818 extern int sysfs_freeze_array(struct mdinfo *sra);
819 extern int sysfs_wait(int fd, int *msec);
820 extern int load_sys(char *path, char *buf, int len);
821 extern int zero_disk_range(int fd, unsigned long long sector, size_t count);
822 extern int reshape_prepare_fdlist(char *devname,
823 struct mdinfo *sra,
824 int raid_disks,
825 int nrdisks,
826 unsigned long blocks,
827 char *backup_file,
828 int *fdlist,
829 unsigned long long *offsets);
830 extern void reshape_free_fdlist(int *fdlist,
831 unsigned long long *offsets,
832 int size);
833 extern int reshape_open_backup_file(char *backup,
834 int fd,
835 char *devname,
836 long blocks,
837 int *fdlist,
838 unsigned long long *offsets,
839 char *sysfs_name,
840 int restart);
841 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
842 unsigned int ndata, unsigned int odata);
843 extern char *locate_backup(char *name);
844 extern char *make_backup(char *name);
845
846 extern int save_stripes(int *source, unsigned long long *offsets,
847 int raid_disks, int chunk_size, int level, int layout,
848 int nwrites, int *dest,
849 unsigned long long start, unsigned long long length,
850 char *buf);
851 extern int restore_stripes(int *dest, unsigned long long *offsets,
852 int raid_disks, int chunk_size, int level, int layout,
853 int source, unsigned long long read_offset,
854 unsigned long long start, unsigned long long length,
855 char *src_buf);
856
857 #ifndef Sendmail
858 #define Sendmail "/usr/lib/sendmail -t"
859 #endif
860
861 #define SYSLOG_FACILITY LOG_DAEMON
862 extern char *map_num_s(mapping_t *map, int num);
863 extern char *map_num(mapping_t *map, int num);
864 extern int map_name(mapping_t *map, char *name);
865 extern mapping_t r0layout[], r5layout[], r6layout[],
866 pers[], modes[], faultylayout[];
867 extern mapping_t consistency_policies[], sysfs_array_states[], update_options[];
868
869 extern char *map_dev_preferred(int major, int minor, int create,
870 char *prefer);
871 static inline char *map_dev(int major, int minor, int create)
872 {
873 return map_dev_preferred(major, minor, create, NULL);
874 }
875
876 /**
877 * is_fd_valid() - check file descriptor.
878 * @fd: file descriptor.
879 *
880 * The function checks if @fd is nonnegative integer and shall be used only
881 * to verify open() result.
882 */
883 static inline int is_fd_valid(int fd)
884 {
885 return (fd > -1);
886 }
887
888 /**
889 * is_level456() - check whether given level is between inclusive 4 and 6.
890 * @level: level to check.
891 *
892 * Return: true if condition is met, false otherwise
893 */
894 static inline bool is_level456(int level)
895 {
896 return (level >= 4 && level <= 6);
897 }
898
899 /**
900 * close_fd() - verify, close and unset file descriptor.
901 * @fd: pointer to file descriptor.
902 *
903 * The function closes and invalidates file descriptor if appropriative. It
904 * ignores incorrect file descriptor quitely to simplify error handling.
905 */
906 static inline void close_fd(int *fd)
907 {
908 if (is_fd_valid(*fd) && close(*fd) == 0)
909 *fd = -1;
910 }
911
912 struct active_array;
913 struct metadata_update;
914
915 /* 'struct reshape' records the intermediate states of
916 * a general reshape.
917 * The starting geometry is converted to the 'before' geometry
918 * by at most an atomic level change. They could be the same.
919 * Similarly the 'after' geometry is converted to the final
920 * geometry by at most a level change.
921 * Note that 'before' and 'after' must have the same level.
922 * 'blocks' is the minimum number of sectors for a reshape unit.
923 * This will be a multiple of the stripe size in each of the
924 * 'before' and 'after' geometries.
925 * If 'blocks' is 0, no restriping is necessary.
926 * 'min_offset_change' is the minimum change to data_offset to
927 * allow the reshape to happen. It is at least the larger of
928 * the old and new chunk sizes, and typically the same as 'blocks'
929 * divided by number of data disks.
930 */
931 struct reshape {
932 int level;
933 int parity; /* number of parity blocks/devices */
934 struct {
935 int layout;
936 int data_disks;
937 } before, after;
938 unsigned long long backup_blocks;
939 unsigned long long min_offset_change;
940 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
941 unsigned long long new_size; /* New size of array in sectors */
942 };
943
944 /**
945 * struct dev_policy - Data structure for policy management.
946 * @next: pointer to next dev_policy.
947 * @name: policy name, category.
948 * @metadata: the metadata type it affects.
949 * @value: value of the policy.
950 *
951 * The functions to manipulate dev_policy lists do not free elements, so they must be statically
952 * allocated. @name and @metadata can be compared by address.
953 */
954 typedef struct dev_policy {
955 struct dev_policy *next;
956 char *name;
957 const char *metadata;
958 const char *value;
959 } dev_policy_t;
960
961 /* A superswitch provides entry point to a metadata handler.
962 *
963 * The superswitch primarily operates on some "metadata" that
964 * is accessed via the 'supertype'.
965 * This metadata has one of three possible sources.
966 * 1/ It is read from a single device. In this case it may not completely
967 * describe the array or arrays as some information might be on other
968 * devices.
969 * 2/ It is read from all devices in a container. In this case all
970 * information is present.
971 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
972 * be complete once enough ->add_to_super calls have completed.
973 *
974 * When creating an array inside a container, the metadata will be
975 * formed by a combination of 2 and 3. The metadata or the array is read,
976 * then new information is added.
977 *
978 * The metadata must sometimes have a concept of a 'current' array
979 * and a 'current' device.
980 * The 'current' array is set by init_super to be the newly created array,
981 * or is set by super_by_fd when it finds it is looking at an array inside
982 * a container.
983 *
984 * The 'current' device is either the device that the metadata was read from
985 * in case 1, or the last device added by add_to_super in case 3.
986 * Case 2 does not identify a 'current' device.
987 */
988 extern struct superswitch {
989
990 /* Used to report details of metadata read from a component
991 * device. ->load_super has been called.
992 */
993 void (*examine_super)(struct supertype *st, char *homehost);
994 void (*brief_examine_super)(struct supertype *st, int verbose);
995 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
996 void (*export_examine_super)(struct supertype *st);
997 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
998 int (*copy_metadata)(struct supertype *st, int from, int to);
999
1000 /* Used to report details of an active array.
1001 * ->load_super was possibly given a 'component' string.
1002 */
1003 void (*detail_super)(struct supertype *st, char *homehost,
1004 char *subarray);
1005 void (*brief_detail_super)(struct supertype *st, char *subarray);
1006 void (*export_detail_super)(struct supertype *st);
1007
1008 /* Optional: platform hardware / firmware details */
1009 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
1010 int (*export_detail_platform)(int verbose, char *controller_path);
1011
1012 /* Used:
1013 * to get uuid to storing in bitmap metadata
1014 * and 'reshape' backup-data metadata
1015 * To see if a device is being re-added to an array it was part of.
1016 */
1017 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
1018
1019 /* Extract generic details from metadata. This could be details about
1020 * the container, or about an individual array within the container.
1021 * The determination is made either by:
1022 * load_super being given a 'component' string.
1023 * validate_geometry determining what to create.
1024 * The info includes both array information and device information.
1025 * The particular device should be:
1026 * The last device added by add_to_super
1027 * The device the metadata was loaded from by load_super
1028 * If 'map' is present, then it is an array raid_disks long
1029 * (raid_disk must already be set and correct) and it is filled
1030 * with 1 for slots that are thought to be active and 0 for slots which
1031 * appear to be failed/missing.
1032 * *info is zeroed out before data is added.
1033 */
1034 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
1035 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
1036 /* Check if the given metadata is flagged as belonging to "this"
1037 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
1038 */
1039 int (*match_home)(struct supertype *st, char *homehost);
1040
1041 /* Make one of several generic modifications to metadata
1042 * prior to assembly (or other times).
1043 * sparc2.2 - first bug in early 0.90 metadata
1044 * super-minor - change name of 0.90 metadata
1045 * summaries - 'correct' any redundant data
1046 * resync - mark array as dirty to trigger a resync.
1047 * uuid - set new uuid - only 0.90 or 1.x
1048 * name - change the name of the array (where supported)
1049 * homehost - change which host this array is tied to.
1050 * devicesize - If metadata is at start of device, change recorded
1051 * device size to match actual device size
1052 * byteorder - swap bytes for 0.90 metadata
1053 *
1054 * force-one - mark that device as uptodate, not old or failed.
1055 * force-array - mark array as clean if it would not otherwise
1056 * assemble
1057 * assemble - not sure how this is different from force-one...
1058 * linear-grow-new - add a new device to a linear array, but don't
1059 * change the size: so superblock still matches
1060 * linear-grow-update - now change the size of the array.
1061 * writemostly - set the WriteMostly1 bit in the superblock devflags
1062 * readwrite - clear the WriteMostly1 bit in the superblock devflags
1063 * failfast - set the FailFast1 bit in the superblock
1064 * nofailfast - clear the FailFast1 bit
1065 * no-bitmap - clear any record that a bitmap is present.
1066 * bbl - add a bad-block-log if possible
1067 * no-bbl - remove any bad-block-log is it is empty.
1068 * force-no-bbl - remove any bad-block-log even if empty.
1069 * revert-reshape - If a reshape is in progress, modify metadata so
1070 * it will resume going in the opposite direction.
1071 */
1072 int (*update_super)(struct supertype *st, struct mdinfo *info,
1073 enum update_opt update,
1074 char *devname, int verbose,
1075 int uuid_set, char *homehost);
1076
1077 /* Create new metadata for new array as described. This could
1078 * be a new container, or an array in a pre-existing container.
1079 * Also used to zero metadata prior to writing it to invalidate old
1080 * metadata.
1081 */
1082 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
1083 struct shape *s, char *name,
1084 char *homehost, int *uuid,
1085 unsigned long long data_offset);
1086
1087 /* update the metadata to include new device, either at create or
1088 * when hot-adding a spare.
1089 */
1090 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
1091 int fd, char *devname,
1092 unsigned long long data_offset);
1093 /* update the metadata to delete a device,
1094 * when hot-removing.
1095 */
1096 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
1097
1098 /* Write metadata to one device when fixing problems or adding
1099 * a new device.
1100 */
1101 int (*store_super)(struct supertype *st, int fd);
1102
1103 /* Write all metadata for this array.
1104 */
1105 int (*write_init_super)(struct supertype *st);
1106 /* Check if metadata read from one device is compatible with an array,
1107 * used when assembling an array, or pseudo-assembling was with
1108 * "--examine --brief"
1109 * If "st" has not yet been loaded the superblock from, "tst" is
1110 * moved in, otherwise the superblock in 'st' is compared with
1111 * 'tst'.
1112 */
1113 int (*compare_super)(struct supertype *st, struct supertype *tst,
1114 int verbose);
1115 /* Load metadata from a single device. If 'devname' is not NULL
1116 * print error messages as appropriate */
1117 int (*load_super)(struct supertype *st, int fd, char *devname);
1118 /* 'fd' is a 'container' md array - load array metadata from the
1119 * whole container.
1120 */
1121 int (*load_container)(struct supertype *st, int fd, char *devname);
1122 /* If 'arg' is a valid name of this metadata type, allocate and
1123 * return a 'supertype' for the particular minor version */
1124 struct supertype * (*match_metadata_desc)(char *arg);
1125 /* If a device has the given size, and the data_offset has been
1126 * requested - work out how much space is available for data.
1127 * This involves adjusting for reserved space (e.g. bitmaps)
1128 * and for any rounding.
1129 * 'mdadm' only calls this for existing arrays where a possible
1130 * spare is being added. However some super-handlers call it
1131 * internally from validate_geometry when creating an array.
1132 */
1133 __u64 (*avail_size)(struct supertype *st, __u64 size,
1134 unsigned long long data_offset);
1135 /*
1136 * Return spare criteria for array:
1137 * - minimum disk size can be used in array;
1138 * - sector size can be used in array.
1139 */
1140 mdadm_status_t (*get_spare_criteria)(struct supertype *st, char *mddev_path,
1141 struct spare_criteria *sc);
1142 /* Find somewhere to put a bitmap - possibly auto-size it - and
1143 * update the metadata to record this. The array may be newly
1144 * created, in which case data_size may be updated, or it might
1145 * already exist. Metadata handler can know if init_super
1146 * has been called, but not write_init_super.
1147 * 0: Success
1148 * -Exxxx: On error
1149 */
1150 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
1151 int delay, int write_behind,
1152 unsigned long long size, int may_change, int major);
1153 /* Perform additional setup required to activate a bitmap.
1154 */
1155 int (*set_bitmap)(struct supertype *st, struct mdinfo *info);
1156 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1157 * md-native format bitmap
1158 */
1159 int (*locate_bitmap)(struct supertype *st, int fd, int node_num);
1160 /* if add_internal_bitmap succeeded for existing array, this
1161 * writes it out.
1162 */
1163 int (*write_bitmap)(struct supertype *st, int fd, enum bitmap_update update);
1164 /* Free the superblock and any other allocated data */
1165 void (*free_super)(struct supertype *st);
1166
1167 /* validate_geometry is called with an st returned by
1168 * match_metadata_desc.
1169 * It should check that the geometry described is compatible with
1170 * the metadata type. It will be called repeatedly as devices
1171 * added to validate changing size and new devices. If there are
1172 * inter-device dependencies, it should record sufficient details
1173 * so these can be validated.
1174 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1175 * Return value is:
1176 * 1: everything is OK
1177 * 0: not OK for some reason - if 'verbose', then error was reported.
1178 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1179 * type, but array is not acceptable for some reason
1180 * message was reported even if verbose is 0.
1181 */
1182 int (*validate_geometry)(struct supertype *st, int level, int layout,
1183 int raiddisks,
1184 int *chunk, unsigned long long size,
1185 unsigned long long data_offset,
1186 char *subdev, unsigned long long *freesize,
1187 int consistency_policy, int verbose);
1188
1189 /**
1190 * test_and_add_drive_policies() - test new and add custom policies from metadata handler.
1191 * @pols: list of currently recorded policies.
1192 * @disk_fd: file descriptor of the device to check.
1193 * @verbose: verbose flag.
1194 *
1195 * Used by IMSM to verify all drives in container/array, against requirements not recored
1196 * in superblock, like controller type for IMSM. It should check all drives even if
1197 * they are not actually used, because mdmon or kernel are free to use any drive assigned to
1198 * container automatically.
1199 *
1200 * Generating and comparison methods belong to metadata handler. It is not mandatory to be
1201 * implemented.
1202 *
1203 * Return: MDADM_STATUS_SUCCESS is expected on success.
1204 */
1205 mdadm_status_t (*test_and_add_drive_policies)(dev_policy_t **pols, int disk_fd,
1206 const int verbose);
1207
1208 /* Return a linked list of 'mdinfo' structures for all arrays
1209 * in the container. For non-containers, it is like
1210 * getinfo_super with an allocated mdinfo.*/
1211 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
1212 /* query the supertype for default geometry */
1213 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
1214 /* Permit subarray's to be deleted from inactive containers */
1215 int (*kill_subarray)(struct supertype *st,
1216 char *subarray_id); /* optional */
1217 /**
1218 * update_subarray() - Permit subarray to be modified.
1219 * @st: Supertype.
1220 * @subarray: Subarray name.
1221 * @update: Update option.
1222 * @ident: Optional identifiers.
1223 */
1224 int (*update_subarray)(struct supertype *st, char *subarray,
1225 enum update_opt update, struct mddev_ident *ident);
1226 /* Check if reshape is supported for this external format.
1227 * st is obtained from super_by_fd() where st->subarray[0] is
1228 * initialized to indicate if reshape is being performed at the
1229 * container or subarray level
1230 */
1231 #define APPLY_METADATA_CHANGES 1
1232 #define ROLLBACK_METADATA_CHANGES 0
1233
1234 int (*reshape_super)(struct supertype *st,
1235 unsigned long long size, int level,
1236 int layout, int chunksize, int raid_disks,
1237 int delta_disks, char *backup, char *dev,
1238 int direction,
1239 int verbose); /* optional */
1240 int (*manage_reshape)( /* optional */
1241 int afd, struct mdinfo *sra, struct reshape *reshape,
1242 struct supertype *st, unsigned long blocks,
1243 int *fds, unsigned long long *offsets,
1244 int dests, int *destfd, unsigned long long *destoffsets);
1245
1246 /* for mdmon */
1247 int (*open_new)(struct supertype *c, struct active_array *a,
1248 int inst);
1249
1250 /* Tell the metadata handler the current state of the array.
1251 * This covers whether it is known to be consistent (no pending writes)
1252 * and how far along a resync is known to have progressed
1253 * (in a->resync_start).
1254 * resync status is really irrelevant if the array is not consistent,
1255 * but some metadata (DDF!) have a place to record the distinction.
1256 * If 'consistent' is '2', then the array can mark it dirty if a
1257 * resync/recovery/whatever is required, or leave it clean if not.
1258 * Return value is 0 dirty (not consistent) and 1 if clean.
1259 * it is only really important if consistent is passed in as '2'.
1260 */
1261 int (*set_array_state)(struct active_array *a, int consistent);
1262
1263 /* When the state of a device might have changed, we call set_disk to
1264 * tell the metadata what the current state is.
1265 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1266 * transitions.
1267 * set_disk might be called when the state of the particular disk has
1268 * not in fact changed.
1269 */
1270 void (*set_disk)(struct active_array *a, int n, int state);
1271 void (*sync_metadata)(struct supertype *st);
1272 void (*process_update)(struct supertype *st,
1273 struct metadata_update *update);
1274 /* Prepare updates allocates extra memory that might be
1275 * needed. If the update cannot be understood, return 0.
1276 */
1277 int (*prepare_update)(struct supertype *st,
1278 struct metadata_update *update);
1279
1280 /* activate_spare will check if the array is degraded and, if it
1281 * is, try to find some spare space in the container.
1282 * On success, it add appropriate updates (For process_update) to
1283 * to the 'updates' list and returns a list of 'mdinfo' identifying
1284 * the device, or devices as there might be multiple missing
1285 * devices and multiple spares available.
1286 */
1287 struct mdinfo *(*activate_spare)(struct active_array *a,
1288 struct metadata_update **updates);
1289 /*
1290 * Return statically allocated string that represents metadata specific
1291 * controller domain of the disk. The domain is used in disk domain
1292 * matching functions. Disks belong to the same domain if the they have
1293 * the same domain from mdadm.conf and belong the same metadata domain.
1294 * Returning NULL or not providing this handler means that metadata
1295 * does not distinguish the differences between disks that belong to
1296 * different controllers. They are in the domain specified by
1297 * configuration file (mdadm.conf).
1298 * In case when the metadata has the notion of domains based on disk
1299 * it shall return NULL for disks that do not belong to the controller
1300 * the supported domains. Such disks will form another domain and won't
1301 * be mixed with supported ones.
1302 */
1303 const char *(*get_disk_controller_domain)(const char *path);
1304
1305 /* for external backup area */
1306 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
1307
1308 /* validate container after assemble */
1309 int (*validate_container)(struct mdinfo *info);
1310
1311 /* write initial empty PPL on device */
1312 int (*write_init_ppl)(struct supertype *st, struct mdinfo *info, int fd);
1313
1314 /* validate ppl before assemble */
1315 int (*validate_ppl)(struct supertype *st, struct mdinfo *info,
1316 struct mdinfo *disk);
1317
1318 /* records new bad block in metadata */
1319 int (*record_bad_block)(struct active_array *a, int n,
1320 unsigned long long sector, int length);
1321
1322 /* clears bad block from metadata */
1323 int (*clear_bad_block)(struct active_array *a, int n,
1324 unsigned long long sector, int length);
1325
1326 /* get list of bad blocks from metadata */
1327 struct md_bb *(*get_bad_blocks)(struct active_array *a, int n);
1328
1329 int swapuuid; /* true if uuid is bigending rather than hostendian */
1330 int external;
1331 const char *name; /* canonical metadata name */
1332 } *superlist[];
1333
1334 extern struct superswitch super0, super1;
1335 extern struct superswitch super_imsm, super_ddf;
1336 extern struct superswitch mbr, gpt;
1337
1338 void imsm_set_no_platform(int v);
1339
1340 struct metadata_update {
1341 int len;
1342 char *buf;
1343 void *space; /* allocated space that monitor will use */
1344 void **space_list; /* list of allocated spaces that monitor can
1345 * use or that it returned.
1346 */
1347 struct metadata_update *next;
1348 };
1349
1350 /* A supertype holds a particular collection of metadata.
1351 * It identifies the metadata type by the superswitch, and the particular
1352 * sub-version of that metadata type.
1353 * metadata read in or created is stored in 'sb' and 'info'.
1354 * There are also fields used by mdmon to track containers.
1355 *
1356 * A supertype may refer to:
1357 * Just an array, possibly in a container
1358 * A container, not identifying any particular array
1359 * Info read from just one device, not yet fully describing the array/container.
1360 *
1361 *
1362 * A supertype is created by:
1363 * super_by_fd
1364 * guess_super
1365 * dup_super
1366 */
1367 struct supertype {
1368 struct superswitch *ss;
1369 int minor_version;
1370 int max_devs;
1371 char container_devnm[32]; /* devnm of container */
1372 void *sb;
1373 void *info;
1374 void *other; /* Hack used to convert v0.90 to v1.0 */
1375 unsigned long long devsize;
1376 unsigned long long data_offset; /* used by v1.x only */
1377 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1378 HW/firmware related incompatability to load metadata.
1379 Used when examining metadata to display content of disk
1380 when user has no hw/firmare compatible system.
1381 */
1382 struct metadata_update *updates;
1383 struct metadata_update **update_tail;
1384
1385 /* extra stuff used by mdmon */
1386 struct active_array *arrays;
1387 int sock; /* listen to external programs */
1388 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1389 * external:/md0/12
1390 */
1391 int devcnt;
1392 int retry_soon;
1393 int nodes;
1394 char *cluster_name;
1395
1396 struct mdinfo *devs;
1397
1398 };
1399
1400 extern struct supertype *super_by_fd(int fd, char **subarray);
1401 enum guess_types { guess_any, guess_array, guess_partitions };
1402 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1403 static inline struct supertype *guess_super(int fd) {
1404 return guess_super_type(fd, guess_any);
1405 }
1406 extern struct supertype *dup_super(struct supertype *st);
1407 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1408 extern int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep);
1409 extern int must_be_container(int fd);
1410 void wait_for(char *dev, int fd);
1411
1412 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1413
1414 /* iterate over the sublist starting at list, having the same
1415 * 'name' as 'list', and matching the given metadata (Where
1416 * NULL matches anything
1417 */
1418 #define pol_for_each(item, list, _metadata) \
1419 for (item = list; \
1420 item && item->name == list->name; \
1421 item = item->next) \
1422 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1423 ; else
1424
1425 /*
1426 * policy records read from mdadm are largely just name-value pairs.
1427 * The names are constants, not strdupped
1428 */
1429 struct pol_rule {
1430 struct pol_rule *next;
1431 char *type; /* rule_policy or rule_part */
1432 struct rule {
1433 struct rule *next;
1434 char *name;
1435 char *value;
1436 char *dups; /* duplicates of 'value' with a partNN appended */
1437 } *rule;
1438 };
1439
1440 extern char rule_policy[], rule_part[];
1441 extern char rule_path[], rule_type[];
1442 extern char type_part[], type_disk[];
1443
1444 extern void policyline(char *line, char *type);
1445 extern void policy_add(char *type, ...);
1446 extern void policy_free(void);
1447
1448 extern struct dev_policy *path_policy(char **paths, char *type);
1449 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1450 extern struct dev_policy *devid_policy(int devid);
1451 extern void dev_policy_free(struct dev_policy *p);
1452
1453 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1454 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1455
1456 extern mdadm_status_t drive_test_and_add_policies(struct supertype *st, dev_policy_t **pols,
1457 int fd, const int verbose);
1458 extern mdadm_status_t sysfs_test_and_add_drive_policies(struct supertype *st, dev_policy_t **pols,
1459 struct mdinfo *mdi, const int verbose);
1460 extern mdadm_status_t mddev_test_and_add_drive_policies(struct supertype *st, dev_policy_t **pols,
1461 int array_fd, const int verbose);
1462
1463 enum policy_action {
1464 act_default,
1465 act_include,
1466 act_re_add,
1467 act_spare, /* This only applies to bare devices */
1468 act_spare_same_slot, /* this allows non-bare devices,
1469 * but only if recent removal */
1470 act_force_spare, /* this allow non-bare devices in any case */
1471 act_err
1472 };
1473
1474 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1475 enum policy_action want);
1476 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1477 enum policy_action want);
1478
1479 struct domainlist {
1480 struct domainlist *next;
1481 const char *dom;
1482 };
1483
1484 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1485 const char *metadata);
1486 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1487 const char *metadata);
1488 extern void domainlist_add_dev(struct domainlist **dom, int devid,
1489 const char *metadata);
1490 extern void domain_free(struct domainlist *dl);
1491 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1492 const char *metadata);
1493 void domain_add(struct domainlist **domp, char *domain);
1494
1495 extern void policy_save_path(char *id_path, struct map_ent *array);
1496 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1497
1498 extern void sysfs_rules_apply(char *devnm, struct mdinfo *dev);
1499 extern void sysfsline(char *line);
1500
1501 #if __GNUC__ < 3
1502 struct stat64;
1503 #endif
1504
1505 #define HAVE_NFTW we assume
1506 #define HAVE_FTW
1507
1508 #ifdef __UCLIBC__
1509 # include <features.h>
1510 # ifndef __UCLIBC_HAS_LFS__
1511 # define lseek64 lseek
1512 # endif
1513 # ifndef __UCLIBC_HAS_FTW__
1514 # undef HAVE_FTW
1515 # undef HAVE_NFTW
1516 # endif
1517 #endif
1518
1519 #ifdef __dietlibc__
1520 # undef HAVE_NFTW
1521 #endif
1522
1523 #if defined(__KLIBC__)
1524 # undef HAVE_NFTW
1525 # undef HAVE_FTW
1526 #endif
1527
1528 #ifndef HAVE_NFTW
1529 # define FTW_PHYS 1
1530 # ifndef HAVE_FTW
1531 struct FTW {};
1532 # endif
1533 #endif
1534
1535 #ifdef HAVE_FTW
1536 # include <ftw.h>
1537 #endif
1538
1539 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1540
1541 extern int Manage_ro(char *devname, int fd, int readonly);
1542 extern int Manage_run(char *devname, int fd, struct context *c);
1543 extern int Manage_stop(char *devname, int fd, int quiet,
1544 int will_retry);
1545 extern int Manage_subdevs(char *devname, int fd,
1546 struct mddev_dev *devlist, int verbose, int test,
1547 enum update_opt update, int force);
1548 extern int autodetect(void);
1549 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1550 extern int Grow_addbitmap(char *devname, int fd,
1551 struct context *c, struct shape *s);
1552 extern int Grow_reshape(char *devname, int fd,
1553 struct mddev_dev *devlist,
1554 struct context *c, struct shape *s);
1555 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1556 int *fdlist, int cnt, char *backup_file, int verbose);
1557 extern int Grow_continue(int mdfd, struct supertype *st,
1558 struct mdinfo *info, char *backup_file,
1559 int forked, int freeze_reshape);
1560 extern int Grow_consistency_policy(char *devname, int fd,
1561 struct context *c, struct shape *s);
1562
1563 extern int restore_backup(struct supertype *st,
1564 struct mdinfo *content,
1565 int working_disks,
1566 int spares,
1567 char **backup_filep,
1568 int verbose);
1569 extern int Grow_continue_command(char *devname, int fd,
1570 char *backup_file, int verbose);
1571
1572 extern int Assemble(struct supertype *st, char *mddev,
1573 struct mddev_ident *ident,
1574 struct mddev_dev *devlist,
1575 struct context *c);
1576
1577 extern int Build(struct mddev_ident *ident, struct mddev_dev *devlist, struct shape *s,
1578 struct context *c);
1579
1580 extern int Create(struct supertype *st, struct mddev_ident *ident, int subdevs,
1581 struct mddev_dev *devlist, struct shape *s, struct context *c);
1582
1583 extern int Detail(char *dev, struct context *c);
1584 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1585 extern int Query(char *dev);
1586 extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1587 extern int Examine(struct mddev_dev *devlist, struct context *c,
1588 struct supertype *forcest);
1589 extern int Monitor(struct mddev_dev *devlist,
1590 char *mailaddr, char *alert_cmd,
1591 struct context *c,
1592 int daemonise, int oneshot,
1593 int dosyslog, char *pidfile, int increments,
1594 int share);
1595
1596 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1597 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1598 extern int Update_subarray(char *dev, char *subarray, enum update_opt update, struct mddev_ident *ident, int quiet);
1599 extern int Wait(char *dev);
1600 extern int WaitClean(char *dev, int verbose);
1601 extern int SetAction(char *dev, char *action);
1602
1603 extern int Incremental(struct mddev_dev *devlist, struct context *c,
1604 struct supertype *st);
1605 extern void RebuildMap(void);
1606 extern int IncrementalScan(struct context *c, char *devnm);
1607 extern int IncrementalRemove(char *devname, char *path, int verbose);
1608 extern int CreateBitmap(char *filename, int force, char uuid[16],
1609 unsigned long chunksize, unsigned long daemon_sleep,
1610 unsigned long write_behind,
1611 unsigned long long array_size,
1612 int major);
1613 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1614 extern int IsBitmapDirty(char *filename);
1615 extern int Write_rules(char *rule_name);
1616 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1617
1618 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1619 static inline unsigned long long
1620 bitmap_bits(unsigned long long array_size, unsigned long chunksize)
1621 {
1622 return (array_size * 512 + chunksize - 1) / chunksize;
1623 }
1624
1625 extern int Dump_metadata(char *dev, char *dir, struct context *c,
1626 struct supertype *st);
1627 extern int Restore_metadata(char *dev, char *dir, struct context *c,
1628 struct supertype *st, int only);
1629
1630 int md_array_valid(int fd);
1631 int md_array_active(int fd);
1632 int md_array_is_active(struct mdinfo *info);
1633 int md_get_array_info(int fd, struct mdu_array_info_s *array);
1634 int md_set_array_info(int fd, struct mdu_array_info_s *array);
1635 int md_get_disk_info(int fd, struct mdu_disk_info_s *disk);
1636 extern int get_linux_version(void);
1637 extern int mdadm_version(char *version);
1638 extern unsigned long long parse_size(char *size);
1639 extern int parse_uuid(char *str, int uuid[4]);
1640 int default_layout(struct supertype *st, int level, int verbose);
1641 extern int is_near_layout_10(int layout);
1642 extern int parse_layout_10(char *layout);
1643 extern int parse_layout_faulty(char *layout);
1644 extern int parse_num(int *dest, const char *num);
1645 extern int parse_cluster_confirm_arg(char *inp, char **devname, int *slot);
1646 extern int check_ext2(int fd, char *name);
1647 extern int check_reiser(int fd, char *name);
1648 extern int check_raid(int fd, char *name);
1649 extern int check_partitions(int fd, char *dname,
1650 unsigned long long freesize,
1651 unsigned long long size);
1652 extern bool is_name_posix_compatible(const char *path);
1653 extern int fstat_is_blkdev(int fd, char *devname, dev_t *rdev);
1654 extern int stat_is_blkdev(char *devname, dev_t *rdev);
1655
1656 extern bool is_string_lq(const char * const str, size_t max_len);
1657 extern bool is_dev_alive(char *path);
1658 extern int get_mdp_major(void);
1659 extern int get_maj_min(char *dev, int *major, int *minor);
1660 extern bool is_bit_set(int *val, unsigned char index);
1661 extern int dev_open(char *dev, int flags);
1662 extern int open_dev(char *devnm);
1663 extern void reopen_mddev(int mdfd);
1664 extern int open_dev_flags(char *devnm, int flags);
1665 extern int open_dev_excl(char *devnm);
1666 extern int is_standard(char *dev, int *nump);
1667 extern int same_dev(char *one, char *two);
1668 extern int compare_paths (char* path1,char* path2);
1669 extern void enable_fds(int devices);
1670 extern void manage_fork_fds(int close_all);
1671 extern int continue_via_systemd(char *devnm, char *service_name, char *prefix);
1672
1673 extern void ident_init(struct mddev_ident *ident);
1674 extern mdadm_status_t ident_set_devname(struct mddev_ident *ident, const char *devname);
1675 extern mdadm_status_t ident_set_name(struct mddev_ident *ident, const char *name);
1676
1677 extern int parse_auto(char *str, char *msg, int config);
1678 extern struct mddev_ident *conf_get_ident(char *dev);
1679 extern struct mddev_dev *conf_get_devs(void);
1680 extern int conf_test_dev(char *devname);
1681 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1682 extern struct createinfo *conf_get_create_info(void);
1683 extern void set_conffile(char *file);
1684 extern char *conf_get_mailaddr(void);
1685 extern char *conf_get_mailfrom(void);
1686 extern char *conf_get_program(void);
1687 extern char *conf_get_homehost(int *require_homehostp);
1688 extern char *conf_get_homecluster(void);
1689 extern int conf_get_monitor_delay(void);
1690 extern char *conf_line(FILE *file);
1691 extern char *conf_word(FILE *file, int allow_key);
1692 extern void print_quoted(char *str);
1693 extern int use_udev(void);
1694 extern void print_escape(char *str);
1695 extern unsigned long GCD(unsigned long a, unsigned long b);
1696 extern int conf_name_is_free(char *name);
1697 extern bool is_devname_ignore(const char *devname);
1698 extern bool is_devname_md_numbered(const char *devname);
1699 extern bool is_devname_md_d_numbered(const char *devname);
1700 extern int conf_verify_devnames(struct mddev_ident *array_list);
1701 extern int devname_matches(char *name, char *match);
1702 extern struct mddev_ident *conf_match(struct supertype *st,
1703 struct mdinfo *info,
1704 char *devname,
1705 int verbose, int *rvp);
1706
1707 extern void free_line(char *line);
1708 extern int match_oneof(char *devices, char *devname);
1709 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1710 extern const int uuid_zero[4];
1711 extern int same_uuid(int a[4], int b[4], int swapuuid);
1712 extern void copy_uuid(void *a, int b[4], int swapuuid);
1713 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1714 extern char *fname_from_uuid(struct supertype *st,
1715 struct mdinfo *info, char *buf, char sep);
1716 extern unsigned long calc_csum(void *super, int bytes);
1717 extern int enough(int level, int raid_disks, int layout, int clean,
1718 char *avail);
1719 extern int ask(char *mesg);
1720 extern unsigned long long get_component_size(int fd);
1721 extern void remove_partitions(int fd);
1722 extern int test_partition(int fd);
1723 extern int test_partition_from_id(dev_t id);
1724 extern int get_data_disks(int level, int layout, int raid_disks);
1725 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1726 int chunksize, unsigned long long devsize);
1727 extern int flush_metadata_updates(struct supertype *st);
1728 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1729 extern int assemble_container_content(struct supertype *st, int mdfd,
1730 struct mdinfo *content,
1731 struct context *c,
1732 char *chosen_name, int *result);
1733 #define INCR_NO 1
1734 #define INCR_UNSAFE 2
1735 #define INCR_ALREADY 4
1736 #define INCR_YES 8
1737
1738 extern bool devid_matches_criteria(struct supertype *st, dev_t devid, struct spare_criteria *sc);
1739 extern bool disk_fd_matches_criteria(struct supertype *st, int disk_fd, struct spare_criteria *sc);
1740 extern struct mdinfo *container_choose_spares(struct supertype *st,
1741 struct spare_criteria *criteria,
1742 struct domainlist *domlist,
1743 char *spare_group,
1744 const char *metadata, int get_one);
1745 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1746 extern int add_disk(int mdfd, struct supertype *st,
1747 struct mdinfo *sra, struct mdinfo *info);
1748 extern int remove_disk(int mdfd, struct supertype *st,
1749 struct mdinfo *sra, struct mdinfo *info);
1750 extern int hot_remove_disk(int mdfd, unsigned long dev, int force);
1751 extern int sys_hot_remove_disk(int statefd, int force);
1752 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1753 unsigned long long min_recovery_start(struct mdinfo *array);
1754
1755 extern char *human_size(long long bytes);
1756 extern char *human_size_brief(long long bytes, int prefix);
1757 extern void print_r10_layout(int layout);
1758
1759 extern char *find_free_devnm(int use_partitions);
1760
1761 extern void put_md_name(char *name);
1762 extern char *devid2kname(dev_t devid);
1763 extern char *devid2devnm(dev_t devid);
1764 extern dev_t devnm2devid(char *devnm);
1765 extern char *get_md_name(char *devnm);
1766
1767 extern char DefaultConfFile[];
1768
1769 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1770 char *chosen, int block_udev);
1771 /* values for 'trustworthy' */
1772 #define LOCAL 1
1773 #define LOCAL_ANY 10
1774 #define FOREIGN 2
1775 #define METADATA 3
1776 extern int open_mddev(char *dev, int report_errors);
1777 extern int is_mddev(char *dev);
1778 extern int open_container(int fd);
1779 extern int metadata_container_matches(char *metadata, char *devnm);
1780 extern int metadata_subdev_matches(char *metadata, char *devnm);
1781 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1782 extern int is_subarray_active(char *subarray, char *devname);
1783 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1784 extern struct superswitch *version_to_superswitch(char *vers);
1785
1786 extern int mdmon_running(char *devnm);
1787 extern int mdmon_pid(char *devnm);
1788 extern int check_env(char *name);
1789 extern __u32 random32(void);
1790 extern void random_uuid(__u8 *buf);
1791 extern int start_mdmon(char *devnm);
1792
1793 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1794 struct supertype *st, unsigned long stripes,
1795 int *fds, unsigned long long *offsets,
1796 int dests, int *destfd, unsigned long long *destoffsets);
1797 void abort_reshape(struct mdinfo *sra);
1798
1799 void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1800
1801 extern char *stat2kname(struct stat *st);
1802 extern char *fd2kname(int fd);
1803 extern char *stat2devnm(struct stat *st);
1804 bool stat_is_md_dev(struct stat *st);
1805 extern char *fd2devnm(int fd);
1806
1807 extern int in_initrd(void);
1808
1809 struct cmap_hooks {
1810 void *cmap_handle; /* corosync lib related */
1811
1812 int (*initialize)(cmap_handle_t *handle);
1813 int (*get_string)(cmap_handle_t handle,
1814 const char *string,
1815 char **name);
1816 int (*finalize)(cmap_handle_t handle);
1817 };
1818
1819 extern void set_cmap_hooks(void);
1820 extern void set_hooks(void);
1821
1822 struct dlm_hooks {
1823 void *dlm_handle; /* dlm lib related */
1824
1825 dlm_lshandle_t (*create_lockspace)(const char *name,
1826 unsigned int mode);
1827 dlm_lshandle_t (*open_lockspace)(const char *name);
1828 int (*release_lockspace)(const char *name, dlm_lshandle_t ls,
1829 int force);
1830 int (*ls_lock)(dlm_lshandle_t lockspace, uint32_t mode,
1831 struct dlm_lksb *lksb, uint32_t flags,
1832 const void *name, unsigned int namelen,
1833 uint32_t parent, void (*astaddr) (void *astarg),
1834 void *astarg, void (*bastaddr) (void *astarg),
1835 void *range);
1836 int (*ls_unlock_wait)(dlm_lshandle_t lockspace, uint32_t lkid,
1837 uint32_t flags, struct dlm_lksb *lksb);
1838 int (*ls_get_fd)(dlm_lshandle_t ls);
1839 int (*dispatch)(int fd);
1840 };
1841
1842 extern int get_cluster_name(char **name);
1843 extern int dlm_funs_ready(void);
1844 extern int cluster_get_dlmlock(void);
1845 extern int cluster_release_dlmlock(void);
1846 extern void set_dlm_hooks(void);
1847
1848 #define MSEC_TO_NSEC(msec) ((msec) * 1000000)
1849 #define USEC_TO_NSEC(usec) ((usec) * 1000)
1850 extern void sleep_for(unsigned int sec, long nsec, bool wake_after_interrupt);
1851 extern bool is_directory(const char *path);
1852 extern bool is_file(const char *path);
1853 extern int s_gethostname(char *buf, int buf_len);
1854
1855 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1856 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1857 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1858 (ROUND_UP((unsigned long)(ptr), base)))
1859
1860 static inline int is_subarray(char *vers)
1861 {
1862 /* The version string for a 'subarray' (an array in a container)
1863 * is
1864 * /containername/componentname for normal read-write arrays
1865 * -containername/componentname for arrays which mdmon must not
1866 * reconfigure. They might be read-only
1867 * or might be undergoing reshape etc.
1868 * containername is e.g. md0, md_d1
1869 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1870 */
1871 return (*vers == '/' || *vers == '-');
1872 }
1873
1874 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1875 {
1876 return &ent->metadata_version[10+strlen(container)+1];
1877 }
1878
1879 /**
1880 * signal_s() - Wrapper for sigaction() with signal()-like interface.
1881 * @sig: The signal to set the signal handler to.
1882 * @handler: The signal handler.
1883 *
1884 * Return: previous handler or SIG_ERR on failure.
1885 */
1886 static inline sighandler_t signal_s(int sig, sighandler_t handler)
1887 {
1888 struct sigaction new_act = {0};
1889 struct sigaction old_act = {0};
1890
1891 new_act.sa_handler = handler;
1892
1893 if (sigaction(sig, &new_act, &old_act) == 0)
1894 return old_act.sa_handler;
1895
1896 return SIG_ERR;
1897 }
1898
1899 #ifdef DEBUG
1900 #define dprintf(fmt, arg...) \
1901 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1902 #define dprintf_cont(fmt, arg...) \
1903 fprintf(stderr, fmt, ##arg)
1904 #else
1905 #define dprintf(fmt, arg...) \
1906 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1907 #define dprintf_cont(fmt, arg...) \
1908 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1909 #endif
1910
1911 static inline int xasprintf(char **strp, const char *fmt, ...) {
1912 va_list ap;
1913 int ret;
1914 va_start(ap, fmt);
1915 ret = vasprintf(strp, fmt, ap);
1916 va_end(ap);
1917 assert(ret >= 0);
1918 return ret;
1919 }
1920
1921 #ifdef DEBUG
1922 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1923 #else
1924 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1925 #endif
1926 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1927
1928 #define pr_info(fmt, args...) printf("%s: "fmt, Name, ##args)
1929
1930 void *xmalloc(size_t len);
1931 void *xrealloc(void *ptr, size_t len);
1932 void *xcalloc(size_t num, size_t size);
1933 char *xstrdup(const char *str);
1934
1935 #define LEVEL_MULTIPATH (-4)
1936 #define LEVEL_LINEAR (-1)
1937 #define LEVEL_FAULTY (-5)
1938
1939 /* kernel module doesn't know about these */
1940 #define LEVEL_CONTAINER (-100)
1941 #define LEVEL_UNSUPPORTED (-200)
1942
1943 /* the kernel does know about this one ... */
1944 #define LEVEL_NONE (-1000000)
1945
1946 /* faulty stuff */
1947
1948 #define WriteTransient 0
1949 #define ReadTransient 1
1950 #define WritePersistent 2
1951 #define ReadPersistent 3
1952 #define WriteAll 4 /* doesn't go to device */
1953 #define ReadFixable 5
1954 #define Modes 6
1955
1956 #define ClearErrors 31
1957 #define ClearFaults 30
1958
1959 #define AllPersist 100 /* internal use only */
1960 #define NoPersist 101
1961
1962 #define ModeMask 0x1f
1963 #define ModeShift 5
1964
1965 #ifdef __TINYC__
1966 #undef minor
1967 #undef major
1968 #undef makedev
1969 #define minor(x) ((x)&0xff)
1970 #define major(x) (((x)>>8)&0xff)
1971 #define makedev(M,m) (((M)<<8) | (m))
1972 #endif
1973
1974 enum r0layout {
1975 RAID0_ORIG_LAYOUT = 1,
1976 RAID0_ALT_MULTIZONE_LAYOUT = 2,
1977 };
1978
1979 /* for raid4/5/6 */
1980 #define ALGORITHM_LEFT_ASYMMETRIC 0
1981 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1982 #define ALGORITHM_LEFT_SYMMETRIC 2
1983 #define ALGORITHM_RIGHT_SYMMETRIC 3
1984
1985 /* Define non-rotating (raid4) algorithms. These allow
1986 * conversion of raid4 to raid5.
1987 */
1988 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1989 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1990
1991 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1992 * Firstly, the exact positioning of the parity block is slightly
1993 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1994 * of DDF.
1995 * Secondly, or order of datablocks over which the Q syndrome is computed
1996 * is different.
1997 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1998 * These layouts are from the DDFv1.2 spec.
1999 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
2000 * leaves RLQ=3 as 'Vendor Specific'
2001 */
2002
2003 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
2004 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
2005 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
2006
2007 /* For every RAID5 algorithm we define a RAID6 algorithm
2008 * with exactly the same layout for data and parity, and
2009 * with the Q block always on the last device (N-1).
2010 * This allows trivial conversion from RAID5 to RAID6
2011 */
2012 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
2013 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
2014 #define ALGORITHM_LEFT_SYMMETRIC_6 18
2015 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
2016 #define ALGORITHM_PARITY_0_6 20
2017 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
2018
2019 /* Define PATH_MAX in case we don't use glibc or standard library does
2020 * not have PATH_MAX defined. Assume max path length is 4K characters.
2021 */
2022 #ifndef PATH_MAX
2023 #define PATH_MAX 4096
2024 #endif
2025
2026 #define RESYNC_NONE -1
2027 #define RESYNC_DELAYED -2
2028 #define RESYNC_PENDING -3
2029 #define RESYNC_REMOTE -4
2030 #define RESYNC_UNKNOWN -5
2031
2032 /* When using "GET_DISK_INFO" it isn't certain how high
2033 * we need to check. So we impose an absolute limit of
2034 * MAX_DISKS. This needs to be much more than the largest
2035 * number of devices any metadata can support. Currently
2036 * v1.x can support 1920
2037 */
2038 #define MAX_DISKS 4096
2039
2040 /* Sometimes the 'size' value passed needs to mean "Maximum".
2041 * In those cases with use MAX_SIZE
2042 */
2043 #define MAX_SIZE 1
2044
2045 /* We want to use unsigned numbers for sector counts, but need
2046 * a value for 'invalid'. Use '1'.
2047 */
2048 #define INVALID_SECTORS 1
2049 /* And another special number needed for --data_offset=variable */
2050 #define VARIABLE_OFFSET 3
2051
2052 /**
2053 * is_container() - check if @level is &LEVEL_CONTAINER
2054 * @level: level value
2055 *
2056 * return:
2057 * 1 if level is equal to &LEVEL_CONTAINER, 0 otherwise.
2058 */
2059 static inline int is_container(const int level)
2060 {
2061 if (level == LEVEL_CONTAINER)
2062 return 1;
2063 return 0;
2064 }
2065
2066 #define STR_COMMON_NONE "none"
2067
2068 /**
2069 * str_is_none() - check if @str starts with "none".
2070 * @str: string
2071 *
2072 * return:
2073 * true if string starts with "none", false otherwise.
2074 */
2075 static inline bool str_is_none(char *str)
2076 {
2077 if (strncmp(str, STR_COMMON_NONE, sizeof(STR_COMMON_NONE) - 1) == 0)
2078 return true;
2079 return false;
2080 }