2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2006-2014 Neil Brown <neilb@suse.de>
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.
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.
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
22 * Email: <neil@brown.name>
24 * Specifications for DDF taken from Common RAID DDF Specification Revision 1.2
25 * (July 28 2006). Reused by permission of SNIA.
28 #define HAVE_STDINT_H 1
34 /* a non-official T10 name for creation GUIDs */
35 static char T10
[] = "Linux-MD";
37 /* DDF timestamps are 1980 based, so we need to add
38 * second-in-decade-of-seventies to convert to linux timestamps.
39 * 10 years with 2 leap years.
41 #define DECADE (3600*24*(365*10+2))
44 const unsigned char *buf
,
47 #define DDF_NOTFOUND (~0U)
48 #define DDF_CONTAINER (DDF_NOTFOUND-1)
50 /* Default for safe_mode_delay. Same value as for IMSM.
52 static const int DDF_SAFE_MODE_DELAY
= 4000;
54 /* The DDF metadata handling.
55 * DDF metadata lives at the end of the device.
56 * The last 512 byte block provides an 'anchor' which is used to locate
57 * the rest of the metadata which usually lives immediately behind the anchor.
60 * - all multibyte numeric fields are bigendian.
61 * - all strings are space padded.
65 typedef struct __be16
{
68 #define be16_eq(x, y) ((x)._v16 == (y)._v16)
69 #define be16_and(x, y) ((x)._v16 & (y)._v16)
70 #define be16_or(x, y) ((x)._v16 | (y)._v16)
71 #define be16_clear(x, y) ((x)._v16 &= ~(y)._v16)
72 #define be16_set(x, y) ((x)._v16 |= (y)._v16)
74 typedef struct __be32
{
77 #define be32_eq(x, y) ((x)._v32 == (y)._v32)
79 typedef struct __be64
{
82 #define be64_eq(x, y) ((x)._v64 == (y)._v64)
84 #define be16_to_cpu(be) __be16_to_cpu((be)._v16)
85 static inline be16
cpu_to_be16(__u16 x
)
87 be16 be
= { ._v16
= __cpu_to_be16(x
) };
91 #define be32_to_cpu(be) __be32_to_cpu((be)._v32)
92 static inline be32
cpu_to_be32(__u32 x
)
94 be32 be
= { ._v32
= __cpu_to_be32(x
) };
98 #define be64_to_cpu(be) __be64_to_cpu((be)._v64)
99 static inline be64
cpu_to_be64(__u64 x
)
101 be64 be
= { ._v64
= __cpu_to_be64(x
) };
105 /* Primary Raid Level (PRL) */
106 #define DDF_RAID0 0x00
107 #define DDF_RAID1 0x01
108 #define DDF_RAID3 0x03
109 #define DDF_RAID4 0x04
110 #define DDF_RAID5 0x05
111 #define DDF_RAID1E 0x11
112 #define DDF_JBOD 0x0f
113 #define DDF_CONCAT 0x1f
114 #define DDF_RAID5E 0x15
115 #define DDF_RAID5EE 0x25
116 #define DDF_RAID6 0x06
118 /* Raid Level Qualifier (RLQ) */
119 #define DDF_RAID0_SIMPLE 0x00
120 #define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
121 #define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
122 #define DDF_RAID3_0 0x00 /* parity in first extent */
123 #define DDF_RAID3_N 0x01 /* parity in last extent */
124 #define DDF_RAID4_0 0x00 /* parity in first extent */
125 #define DDF_RAID4_N 0x01 /* parity in last extent */
126 /* these apply to raid5e and raid5ee as well */
127 #define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
128 #define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
129 #define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
130 #define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
132 #define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
133 #define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
135 /* Secondary RAID Level (SRL) */
136 #define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
137 #define DDF_2MIRRORED 0x01
138 #define DDF_2CONCAT 0x02
139 #define DDF_2SPANNED 0x03 /* This is also weird - be careful */
142 #define DDF_HEADER_MAGIC cpu_to_be32(0xDE11DE11)
143 #define DDF_CONTROLLER_MAGIC cpu_to_be32(0xAD111111)
144 #define DDF_PHYS_RECORDS_MAGIC cpu_to_be32(0x22222222)
145 #define DDF_PHYS_DATA_MAGIC cpu_to_be32(0x33333333)
146 #define DDF_VIRT_RECORDS_MAGIC cpu_to_be32(0xDDDDDDDD)
147 #define DDF_VD_CONF_MAGIC cpu_to_be32(0xEEEEEEEE)
148 #define DDF_SPARE_ASSIGN_MAGIC cpu_to_be32(0x55555555)
149 #define DDF_VU_CONF_MAGIC cpu_to_be32(0x88888888)
150 #define DDF_VENDOR_LOG_MAGIC cpu_to_be32(0x01dBEEF0)
151 #define DDF_BBM_LOG_MAGIC cpu_to_be32(0xABADB10C)
153 #define DDF_GUID_LEN 24
154 #define DDF_REVISION_0 "01.00.00"
155 #define DDF_REVISION_2 "01.02.00"
158 be32 magic
; /* DDF_HEADER_MAGIC */
160 char guid
[DDF_GUID_LEN
];
161 char revision
[8]; /* 01.02.00 */
162 be32 seq
; /* starts at '1' */
167 __u8 pad0
; /* 0xff */
168 __u8 pad1
[12]; /* 12 * 0xff */
169 /* 64 bytes so far */
170 __u8 header_ext
[32]; /* reserved: fill with 0xff */
174 __u8 pad2
[3]; /* 0xff */
175 be32 workspace_len
; /* sectors for vendor space -
176 * at least 32768(sectors) */
178 be16 max_pd_entries
; /* one of 15, 63, 255, 1023, 4095 */
179 be16 max_vd_entries
; /* 2^(4,6,8,10,12)-1 : i.e. as above */
180 be16 max_partitions
; /* i.e. max num of configuration
181 record entries per disk */
182 be16 config_record_len
; /* 1 +ROUNDUP(max_primary_element_entries
184 be16 max_primary_element_entries
; /* 16, 64, 256, 1024, or 4096 */
185 __u8 pad3
[54]; /* 0xff */
186 /* 192 bytes so far */
187 be32 controller_section_offset
;
188 be32 controller_section_length
;
189 be32 phys_section_offset
;
190 be32 phys_section_length
;
191 be32 virt_section_offset
;
192 be32 virt_section_length
;
193 be32 config_section_offset
;
194 be32 config_section_length
;
195 be32 data_section_offset
;
196 be32 data_section_length
;
197 be32 bbm_section_offset
;
198 be32 bbm_section_length
;
199 be32 diag_space_offset
;
200 be32 diag_space_length
;
203 /* 256 bytes so far */
204 __u8 pad4
[256]; /* 0xff */
208 #define DDF_HEADER_ANCHOR 0x00
209 #define DDF_HEADER_PRIMARY 0x01
210 #define DDF_HEADER_SECONDARY 0x02
212 /* The content of the 'controller section' - global scope */
213 struct ddf_controller_data
{
214 be32 magic
; /* DDF_CONTROLLER_MAGIC */
216 char guid
[DDF_GUID_LEN
];
217 struct controller_type
{
224 __u8 pad
[8]; /* 0xff */
225 __u8 vendor_data
[448];
228 /* The content of phys_section - global scope */
230 be32 magic
; /* DDF_PHYS_RECORDS_MAGIC */
235 struct phys_disk_entry
{
236 char guid
[DDF_GUID_LEN
];
240 be64 config_size
; /* DDF structures must be after here */
241 char path
[18]; /* Another horrible structure really
242 * but is "used for information
248 /* phys_disk_entry.type is a bitmap - bigendian remember */
249 #define DDF_Forced_PD_GUID 1
250 #define DDF_Active_in_VD 2
251 #define DDF_Global_Spare 4 /* VD_CONF records are ignored */
252 #define DDF_Spare 8 /* overrides Global_spare */
253 #define DDF_Foreign 16
254 #define DDF_Legacy 32 /* no DDF on this device */
256 #define DDF_Interface_mask 0xf00
257 #define DDF_Interface_SCSI 0x100
258 #define DDF_Interface_SAS 0x200
259 #define DDF_Interface_SATA 0x300
260 #define DDF_Interface_FC 0x400
262 /* phys_disk_entry.state is a bigendian bitmap */
264 #define DDF_Failed 2 /* overrides 1,4,8 */
265 #define DDF_Rebuilding 4
266 #define DDF_Transition 8
268 #define DDF_ReadErrors 32
269 #define DDF_Missing 64
271 /* The content of the virt_section global scope */
272 struct virtual_disk
{
273 be32 magic
; /* DDF_VIRT_RECORDS_MAGIC */
278 struct virtual_entry
{
279 char guid
[DDF_GUID_LEN
];
281 __u16 pad0
; /* 0xffff */
291 /* virtual_entry.type is a bitmap - bigendian */
293 #define DDF_Enforce_Groups 2
294 #define DDF_Unicode 4
295 #define DDF_Owner_Valid 8
297 /* virtual_entry.state is a bigendian bitmap */
298 #define DDF_state_mask 0x7
299 #define DDF_state_optimal 0x0
300 #define DDF_state_degraded 0x1
301 #define DDF_state_deleted 0x2
302 #define DDF_state_missing 0x3
303 #define DDF_state_failed 0x4
304 #define DDF_state_part_optimal 0x5
306 #define DDF_state_morphing 0x8
307 #define DDF_state_inconsistent 0x10
309 /* virtual_entry.init_state is a bigendian bitmap */
310 #define DDF_initstate_mask 0x03
311 #define DDF_init_not 0x00
312 #define DDF_init_quick 0x01 /* initialisation is progress.
313 * i.e. 'state_inconsistent' */
314 #define DDF_init_full 0x02
316 #define DDF_access_mask 0xc0
317 #define DDF_access_rw 0x00
318 #define DDF_access_ro 0x80
319 #define DDF_access_blocked 0xc0
321 /* The content of the config_section - local scope
322 * It has multiple records each config_record_len sectors
323 * They can be vd_config or spare_assign
327 be32 magic
; /* DDF_VD_CONF_MAGIC */
329 char guid
[DDF_GUID_LEN
];
333 be16 prim_elmnt_count
;
334 __u8 chunk_shift
; /* 0 == 512, 1==1024 etc */
337 __u8 sec_elmnt_count
;
340 be64 blocks
; /* blocks per component could be different
341 * on different component devices...(only
342 * for concat I hope) */
343 be64 array_blocks
; /* blocks in array */
351 __u8 v0
[32]; /* reserved- 0xff */
352 __u8 v1
[32]; /* reserved- 0xff */
353 __u8 v2
[16]; /* reserved- 0xff */
354 __u8 v3
[16]; /* reserved- 0xff */
356 be32 phys_refnum
[0]; /* refnum of each disk in sequence */
357 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
358 bvd are always the same size */
360 #define LBA_OFFSET(ddf, vd) ((be64 *) &(vd)->phys_refnum[(ddf)->mppe])
362 /* vd_config.cache_pol[7] is a bitmap */
363 #define DDF_cache_writeback 1 /* else writethrough */
364 #define DDF_cache_wadaptive 2 /* only applies if writeback */
365 #define DDF_cache_readahead 4
366 #define DDF_cache_radaptive 8 /* only if doing read-ahead */
367 #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
368 #define DDF_cache_wallowed 32 /* enable write caching */
369 #define DDF_cache_rallowed 64 /* enable read caching */
371 struct spare_assign
{
372 be32 magic
; /* DDF_SPARE_ASSIGN_MAGIC */
377 be16 populated
; /* SAEs used */
378 be16 max
; /* max SAEs */
380 struct spare_assign_entry
{
381 char guid
[DDF_GUID_LEN
];
382 be16 secondary_element
;
386 /* spare_assign.type is a bitmap */
387 #define DDF_spare_dedicated 0x1 /* else global */
388 #define DDF_spare_revertible 0x2 /* else committable */
389 #define DDF_spare_active 0x4 /* else not active */
390 #define DDF_spare_affinity 0x8 /* enclosure affinity */
392 /* The data_section contents - local scope */
394 be32 magic
; /* DDF_PHYS_DATA_MAGIC */
396 char guid
[DDF_GUID_LEN
];
397 be32 refnum
; /* crc of some magic drive data ... */
398 __u8 forced_ref
; /* set when above was not result of magic */
399 __u8 forced_guid
; /* set if guid was forced rather than magic */
404 /* bbm_section content */
405 struct bad_block_log
{
412 struct mapped_block
{
413 be64 defective_start
;
414 be32 replacement_start
;
420 /* Struct for internally holding ddf structures */
421 /* The DDF structure stored on each device is potentially
422 * quite different, as some data is global and some is local.
423 * The global data is:
426 * - Physical disk records
427 * - Virtual disk records
429 * - Configuration records
430 * - Physical Disk data section
431 * ( and Bad block and vendor which I don't care about yet).
433 * The local data is parsed into separate lists as it is read
434 * and reconstructed for writing. This means that we only need
435 * to make config changes once and they are automatically
436 * propagated to all devices.
437 * The global (config and disk data) records are each in a list
438 * of separate data structures. When writing we find the entry
439 * or entries applicable to the particular device.
442 struct ddf_header anchor
, primary
, secondary
;
443 struct ddf_controller_data controller
;
444 struct ddf_header
*active
;
445 struct phys_disk
*phys
;
446 struct virtual_disk
*virt
;
449 unsigned int max_part
, mppe
, conf_rec_len
;
457 unsigned int vcnum
; /* index into ->virt */
458 /* For an array with a secondary level there are
459 * multiple vd_config structures, all with the same
460 * guid but with different sec_elmnt_seq.
461 * One of these structures is in 'conf' below.
462 * The others are in other_bvds, not in any
465 struct vd_config
**other_bvds
;
466 __u64
*block_sizes
; /* NULL if all the same */
469 struct vd_config conf
;
470 } *conflist
, *currentconf
;
479 unsigned long long size
; /* sectors */
480 be64 primary_lba
; /* sectors */
481 be64 secondary_lba
; /* sectors */
482 be64 workspace_lba
; /* sectors */
483 int pdnum
; /* index in ->phys */
484 struct spare_assign
*spare
;
485 void *mdupdate
; /* hold metadata update */
487 /* These fields used by auto-layout */
488 int raiddisk
; /* slot to fill in autolayout */
492 struct disk_data disk
;
493 struct vcl
*vlist
[0]; /* max_part in size */
498 static int load_super_ddf_all(struct supertype
*st
, int fd
,
499 void **sbp
, char *devname
);
500 static int get_svd_state(const struct ddf_super
*, const struct vcl
*);
502 validate_geometry_ddf_container(struct supertype
*st
,
503 int level
, int layout
, int raiddisks
,
504 int chunk
, unsigned long long size
,
505 unsigned long long data_offset
,
506 char *dev
, unsigned long long *freesize
,
509 static int validate_geometry_ddf_bvd(struct supertype
*st
,
510 int level
, int layout
, int raiddisks
,
511 int *chunk
, unsigned long long size
,
512 unsigned long long data_offset
,
513 char *dev
, unsigned long long *freesize
,
517 static void free_super_ddf(struct supertype
*st
);
518 static int all_ff(const char *guid
);
519 static unsigned int get_pd_index_from_refnum(const struct vcl
*vc
,
520 be32 refnum
, unsigned int nmax
,
521 const struct vd_config
**bvd
,
523 static void getinfo_super_ddf(struct supertype
*st
, struct mdinfo
*info
, char *map
);
524 static void uuid_from_ddf_guid(const char *guid
, int uuid
[4]);
525 static void uuid_from_super_ddf(struct supertype
*st
, int uuid
[4]);
526 static void _ddf_array_name(char *name
, const struct ddf_super
*ddf
, int i
);
527 static void getinfo_super_ddf_bvd(struct supertype
*st
, struct mdinfo
*info
, char *map
);
528 static int init_super_ddf_bvd(struct supertype
*st
,
529 mdu_array_info_t
*info
,
530 unsigned long long size
,
531 char *name
, char *homehost
,
532 int *uuid
, unsigned long long data_offset
);
535 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
539 static void pr_state(struct ddf_super
*ddf
, const char *msg
)
542 dprintf("%s/%s: ", __func__
, msg
);
543 for (i
= 0; i
< be16_to_cpu(ddf
->active
->max_vd_entries
); i
++) {
544 if (all_ff(ddf
->virt
->entries
[i
].guid
))
546 dprintf("%u(s=%02x i=%02x) ", i
,
547 ddf
->virt
->entries
[i
].state
,
548 ddf
->virt
->entries
[i
].init_state
);
553 static void pr_state(const struct ddf_super
*ddf
, const char *msg
) {}
556 static void _ddf_set_updates_pending(struct ddf_super
*ddf
, const char *func
)
558 if (ddf
->updates_pending
)
560 ddf
->updates_pending
= 1;
561 ddf
->active
->seq
= cpu_to_be32((be32_to_cpu(ddf
->active
->seq
)+1));
565 #define ddf_set_updates_pending(x) _ddf_set_updates_pending((x), __func__)
567 static be32
calc_crc(void *buf
, int len
)
569 /* crcs are always at the same place as in the ddf_header */
570 struct ddf_header
*ddf
= buf
;
571 be32 oldcrc
= ddf
->crc
;
573 ddf
->crc
= cpu_to_be32(0xffffffff);
575 newcrc
= crc32(0, buf
, len
);
577 /* The crc is stored (like everything) bigendian, so convert
578 * here for simplicity
580 return cpu_to_be32(newcrc
);
583 #define DDF_INVALID_LEVEL 0xff
584 #define DDF_NO_SECONDARY 0xff
585 static int err_bad_md_layout(const mdu_array_info_t
*array
)
587 pr_err("RAID%d layout %x with %d disks is unsupported for DDF\n",
588 array
->level
, array
->layout
, array
->raid_disks
);
592 static int layout_md2ddf(const mdu_array_info_t
*array
,
593 struct vd_config
*conf
)
595 be16 prim_elmnt_count
= cpu_to_be16(array
->raid_disks
);
596 __u8 prl
= DDF_INVALID_LEVEL
, rlq
= 0;
597 __u8 sec_elmnt_count
= 1;
598 __u8 srl
= DDF_NO_SECONDARY
;
600 switch (array
->level
) {
605 rlq
= DDF_RAID0_SIMPLE
;
609 switch (array
->raid_disks
) {
611 rlq
= DDF_RAID1_SIMPLE
;
614 rlq
= DDF_RAID1_MULTI
;
617 return err_bad_md_layout(array
);
622 if (array
->layout
!= 0)
623 return err_bad_md_layout(array
);
628 switch (array
->layout
) {
629 case ALGORITHM_LEFT_ASYMMETRIC
:
630 rlq
= DDF_RAID5_N_RESTART
;
632 case ALGORITHM_RIGHT_ASYMMETRIC
:
633 rlq
= DDF_RAID5_0_RESTART
;
635 case ALGORITHM_LEFT_SYMMETRIC
:
636 rlq
= DDF_RAID5_N_CONTINUE
;
638 case ALGORITHM_RIGHT_SYMMETRIC
:
639 /* not mentioned in standard */
641 return err_bad_md_layout(array
);
646 switch (array
->layout
) {
647 case ALGORITHM_ROTATING_N_RESTART
:
648 rlq
= DDF_RAID5_N_RESTART
;
650 case ALGORITHM_ROTATING_ZERO_RESTART
:
651 rlq
= DDF_RAID6_0_RESTART
;
653 case ALGORITHM_ROTATING_N_CONTINUE
:
654 rlq
= DDF_RAID5_N_CONTINUE
;
657 return err_bad_md_layout(array
);
662 if (array
->raid_disks
% 2 == 0 && array
->layout
== 0x102) {
663 rlq
= DDF_RAID1_SIMPLE
;
664 prim_elmnt_count
= cpu_to_be16(2);
665 sec_elmnt_count
= array
->raid_disks
/ 2;
666 } else if (array
->raid_disks
% 3 == 0
667 && array
->layout
== 0x103) {
668 rlq
= DDF_RAID1_MULTI
;
669 prim_elmnt_count
= cpu_to_be16(3);
670 sec_elmnt_count
= array
->raid_disks
/ 3;
672 return err_bad_md_layout(array
);
677 return err_bad_md_layout(array
);
680 conf
->prim_elmnt_count
= prim_elmnt_count
;
683 conf
->sec_elmnt_count
= sec_elmnt_count
;
687 static int err_bad_ddf_layout(const struct vd_config
*conf
)
689 pr_err("DDF RAID %u qualifier %u with %u disks is unsupported\n",
690 conf
->prl
, conf
->rlq
, be16_to_cpu(conf
->prim_elmnt_count
));
694 static int layout_ddf2md(const struct vd_config
*conf
,
695 mdu_array_info_t
*array
)
697 int level
= LEVEL_UNSUPPORTED
;
699 int raiddisks
= be16_to_cpu(conf
->prim_elmnt_count
);
701 if (conf
->sec_elmnt_count
> 1) {
702 /* see also check_secondary() */
703 if (conf
->prl
!= DDF_RAID1
||
704 (conf
->srl
!= DDF_2STRIPED
&& conf
->srl
!= DDF_2SPANNED
)) {
705 pr_err("Unsupported secondary RAID level %u/%u\n",
706 conf
->prl
, conf
->srl
);
709 if (raiddisks
== 2 && conf
->rlq
== DDF_RAID1_SIMPLE
)
711 else if (raiddisks
== 3 && conf
->rlq
== DDF_RAID1_MULTI
)
714 return err_bad_ddf_layout(conf
);
715 raiddisks
*= conf
->sec_elmnt_count
;
722 level
= LEVEL_LINEAR
;
725 if (conf
->rlq
!= DDF_RAID0_SIMPLE
)
726 return err_bad_ddf_layout(conf
);
730 if (!((conf
->rlq
== DDF_RAID1_SIMPLE
&& raiddisks
== 2) ||
731 (conf
->rlq
== DDF_RAID1_MULTI
&& raiddisks
== 3)))
732 return err_bad_ddf_layout(conf
);
736 if (conf
->rlq
!= DDF_RAID4_N
)
737 return err_bad_ddf_layout(conf
);
742 case DDF_RAID5_N_RESTART
:
743 layout
= ALGORITHM_LEFT_ASYMMETRIC
;
745 case DDF_RAID5_0_RESTART
:
746 layout
= ALGORITHM_RIGHT_ASYMMETRIC
;
748 case DDF_RAID5_N_CONTINUE
:
749 layout
= ALGORITHM_LEFT_SYMMETRIC
;
752 return err_bad_ddf_layout(conf
);
758 case DDF_RAID5_N_RESTART
:
759 layout
= ALGORITHM_ROTATING_N_RESTART
;
761 case DDF_RAID6_0_RESTART
:
762 layout
= ALGORITHM_ROTATING_ZERO_RESTART
;
764 case DDF_RAID5_N_CONTINUE
:
765 layout
= ALGORITHM_ROTATING_N_CONTINUE
;
768 return err_bad_ddf_layout(conf
);
773 return err_bad_ddf_layout(conf
);
777 array
->level
= level
;
778 array
->layout
= layout
;
779 array
->raid_disks
= raiddisks
;
783 static int load_ddf_header(int fd
, unsigned long long lba
,
784 unsigned long long size
,
786 struct ddf_header
*hdr
, struct ddf_header
*anchor
)
788 /* read a ddf header (primary or secondary) from fd/lba
789 * and check that it is consistent with anchor
791 * magic, crc, guid, rev, and LBA's header_type, and
792 * everything after header_type must be the same
797 if (lseek64(fd
, lba
<<9, 0) < 0)
800 if (read(fd
, hdr
, 512) != 512)
803 if (!be32_eq(hdr
->magic
, DDF_HEADER_MAGIC
)) {
804 pr_err("%s: bad header magic\n", __func__
);
807 if (!be32_eq(calc_crc(hdr
, 512), hdr
->crc
)) {
808 pr_err("%s: bad CRC\n", __func__
);
811 if (memcmp(anchor
->guid
, hdr
->guid
, DDF_GUID_LEN
) != 0 ||
812 memcmp(anchor
->revision
, hdr
->revision
, 8) != 0 ||
813 !be64_eq(anchor
->primary_lba
, hdr
->primary_lba
) ||
814 !be64_eq(anchor
->secondary_lba
, hdr
->secondary_lba
) ||
816 memcmp(anchor
->pad2
, hdr
->pad2
, 512 -
817 offsetof(struct ddf_header
, pad2
)) != 0) {
818 pr_err("%s: header mismatch\n", __func__
);
822 /* Looks good enough to me... */
826 static void *load_section(int fd
, struct ddf_super
*super
, void *buf
,
827 be32 offset_be
, be32 len_be
, int check
)
829 unsigned long long offset
= be32_to_cpu(offset_be
);
830 unsigned long long len
= be32_to_cpu(len_be
);
831 int dofree
= (buf
== NULL
);
834 if (len
!= 2 && len
!= 8 && len
!= 32
835 && len
!= 128 && len
!= 512)
840 if (!buf
&& posix_memalign(&buf
, 512, len
<<9) != 0)
846 if (super
->active
->type
== 1)
847 offset
+= be64_to_cpu(super
->active
->primary_lba
);
849 offset
+= be64_to_cpu(super
->active
->secondary_lba
);
851 if ((unsigned long long)lseek64(fd
, offset
<<9, 0) != (offset
<<9)) {
856 if ((unsigned long long)read(fd
, buf
, len
<<9) != (len
<<9)) {
864 static int load_ddf_headers(int fd
, struct ddf_super
*super
, char *devname
)
866 unsigned long long dsize
;
868 get_dev_size(fd
, NULL
, &dsize
);
870 if (lseek64(fd
, dsize
-512, 0) < 0) {
872 pr_err("Cannot seek to anchor block on %s: %s\n",
873 devname
, strerror(errno
));
876 if (read(fd
, &super
->anchor
, 512) != 512) {
878 pr_err("Cannot read anchor block on %s: %s\n",
879 devname
, strerror(errno
));
882 if (!be32_eq(super
->anchor
.magic
, DDF_HEADER_MAGIC
)) {
884 pr_err("no DDF anchor found on %s\n",
888 if (!be32_eq(calc_crc(&super
->anchor
, 512), super
->anchor
.crc
)) {
890 pr_err("bad CRC on anchor on %s\n",
894 if (memcmp(super
->anchor
.revision
, DDF_REVISION_0
, 8) != 0 &&
895 memcmp(super
->anchor
.revision
, DDF_REVISION_2
, 8) != 0) {
897 pr_err("can only support super revision"
898 " %.8s and earlier, not %.8s on %s\n",
899 DDF_REVISION_2
, super
->anchor
.revision
,devname
);
902 super
->active
= NULL
;
903 if (load_ddf_header(fd
, be64_to_cpu(super
->anchor
.primary_lba
),
905 &super
->primary
, &super
->anchor
) == 0) {
907 pr_err("Failed to load primary DDF header "
910 super
->active
= &super
->primary
;
912 if (load_ddf_header(fd
, be64_to_cpu(super
->anchor
.secondary_lba
),
914 &super
->secondary
, &super
->anchor
)) {
915 if (super
->active
== NULL
916 || (be32_to_cpu(super
->primary
.seq
)
917 < be32_to_cpu(super
->secondary
.seq
) &&
918 !super
->secondary
.openflag
)
919 || (be32_to_cpu(super
->primary
.seq
)
920 == be32_to_cpu(super
->secondary
.seq
) &&
921 super
->primary
.openflag
&& !super
->secondary
.openflag
)
923 super
->active
= &super
->secondary
;
924 } else if (devname
&&
925 be64_to_cpu(super
->anchor
.secondary_lba
) != ~(__u64
)0)
926 pr_err("Failed to load secondary DDF header on %s\n",
928 if (super
->active
== NULL
)
933 static int load_ddf_global(int fd
, struct ddf_super
*super
, char *devname
)
936 ok
= load_section(fd
, super
, &super
->controller
,
937 super
->active
->controller_section_offset
,
938 super
->active
->controller_section_length
,
940 super
->phys
= load_section(fd
, super
, NULL
,
941 super
->active
->phys_section_offset
,
942 super
->active
->phys_section_length
,
944 super
->pdsize
= be32_to_cpu(super
->active
->phys_section_length
) * 512;
946 super
->virt
= load_section(fd
, super
, NULL
,
947 super
->active
->virt_section_offset
,
948 super
->active
->virt_section_length
,
950 super
->vdsize
= be32_to_cpu(super
->active
->virt_section_length
) * 512;
960 super
->conflist
= NULL
;
963 super
->max_part
= be16_to_cpu(super
->active
->max_partitions
);
964 super
->mppe
= be16_to_cpu(super
->active
->max_primary_element_entries
);
965 super
->conf_rec_len
= be16_to_cpu(super
->active
->config_record_len
);
969 #define DDF_UNUSED_BVD 0xff
970 static int alloc_other_bvds(const struct ddf_super
*ddf
, struct vcl
*vcl
)
972 unsigned int n_vds
= vcl
->conf
.sec_elmnt_count
- 1;
973 unsigned int i
, vdsize
;
976 vcl
->other_bvds
= NULL
;
979 vdsize
= ddf
->conf_rec_len
* 512;
980 if (posix_memalign(&p
, 512, n_vds
*
981 (vdsize
+ sizeof(struct vd_config
*))) != 0)
983 vcl
->other_bvds
= (struct vd_config
**) (p
+ n_vds
* vdsize
);
984 for (i
= 0; i
< n_vds
; i
++) {
985 vcl
->other_bvds
[i
] = p
+ i
* vdsize
;
986 memset(vcl
->other_bvds
[i
], 0, vdsize
);
987 vcl
->other_bvds
[i
]->sec_elmnt_seq
= DDF_UNUSED_BVD
;
992 static void add_other_bvd(struct vcl
*vcl
, struct vd_config
*vd
,
996 for (i
= 0; i
< vcl
->conf
.sec_elmnt_count
-1; i
++)
997 if (vcl
->other_bvds
[i
]->sec_elmnt_seq
== vd
->sec_elmnt_seq
)
1000 if (i
< vcl
->conf
.sec_elmnt_count
-1) {
1001 if (be32_to_cpu(vd
->seqnum
) <=
1002 be32_to_cpu(vcl
->other_bvds
[i
]->seqnum
))
1005 for (i
= 0; i
< vcl
->conf
.sec_elmnt_count
-1; i
++)
1006 if (vcl
->other_bvds
[i
]->sec_elmnt_seq
== DDF_UNUSED_BVD
)
1008 if (i
== vcl
->conf
.sec_elmnt_count
-1) {
1009 pr_err("no space for sec level config %u, count is %u\n",
1010 vd
->sec_elmnt_seq
, vcl
->conf
.sec_elmnt_count
);
1014 memcpy(vcl
->other_bvds
[i
], vd
, len
);
1017 static int load_ddf_local(int fd
, struct ddf_super
*super
,
1018 char *devname
, int keep
)
1024 unsigned int confsec
;
1026 unsigned int max_virt_disks
=
1027 be16_to_cpu(super
->active
->max_vd_entries
);
1028 unsigned long long dsize
;
1030 /* First the local disk info */
1031 if (posix_memalign((void**)&dl
, 512,
1033 (super
->max_part
) * sizeof(dl
->vlist
[0])) != 0) {
1034 pr_err("%s could not allocate disk info buffer\n",
1039 load_section(fd
, super
, &dl
->disk
,
1040 super
->active
->data_section_offset
,
1041 super
->active
->data_section_length
,
1043 dl
->devname
= devname
? xstrdup(devname
) : NULL
;
1046 dl
->major
= major(stb
.st_rdev
);
1047 dl
->minor
= minor(stb
.st_rdev
);
1048 dl
->next
= super
->dlist
;
1049 dl
->fd
= keep
? fd
: -1;
1052 if (get_dev_size(fd
, devname
, &dsize
))
1053 dl
->size
= dsize
>> 9;
1054 /* If the disks have different sizes, the LBAs will differ
1055 * between phys disks.
1056 * At this point here, the values in super->active must be valid
1057 * for this phys disk. */
1058 dl
->primary_lba
= super
->active
->primary_lba
;
1059 dl
->secondary_lba
= super
->active
->secondary_lba
;
1060 dl
->workspace_lba
= super
->active
->workspace_lba
;
1062 for (i
= 0 ; i
< super
->max_part
; i
++)
1063 dl
->vlist
[i
] = NULL
;
1066 for (i
= 0; i
< be16_to_cpu(super
->active
->max_pd_entries
); i
++)
1067 if (memcmp(super
->phys
->entries
[i
].guid
,
1068 dl
->disk
.guid
, DDF_GUID_LEN
) == 0)
1071 /* Now the config list. */
1072 /* 'conf' is an array of config entries, some of which are
1073 * probably invalid. Those which are good need to be copied into
1077 conf
= load_section(fd
, super
, super
->conf
,
1078 super
->active
->config_section_offset
,
1079 super
->active
->config_section_length
,
1084 confsec
< be32_to_cpu(super
->active
->config_section_length
);
1085 confsec
+= super
->conf_rec_len
) {
1086 struct vd_config
*vd
=
1087 (struct vd_config
*)((char*)conf
+ confsec
*512);
1090 if (be32_eq(vd
->magic
, DDF_SPARE_ASSIGN_MAGIC
)) {
1093 if (posix_memalign((void**)&dl
->spare
, 512,
1094 super
->conf_rec_len
*512) != 0) {
1095 pr_err("%s could not allocate spare info buf\n",
1100 memcpy(dl
->spare
, vd
, super
->conf_rec_len
*512);
1103 if (!be32_eq(vd
->magic
, DDF_VD_CONF_MAGIC
))
1104 /* Must be vendor-unique - I cannot handle those */
1107 for (vcl
= super
->conflist
; vcl
; vcl
= vcl
->next
) {
1108 if (memcmp(vcl
->conf
.guid
,
1109 vd
->guid
, DDF_GUID_LEN
) == 0)
1114 dl
->vlist
[vnum
++] = vcl
;
1115 if (vcl
->other_bvds
!= NULL
&&
1116 vcl
->conf
.sec_elmnt_seq
!= vd
->sec_elmnt_seq
) {
1117 add_other_bvd(vcl
, vd
, super
->conf_rec_len
*512);
1120 if (be32_to_cpu(vd
->seqnum
) <=
1121 be32_to_cpu(vcl
->conf
.seqnum
))
1124 if (posix_memalign((void**)&vcl
, 512,
1125 (super
->conf_rec_len
*512 +
1126 offsetof(struct vcl
, conf
))) != 0) {
1127 pr_err("%s could not allocate vcl buf\n",
1131 vcl
->next
= super
->conflist
;
1132 vcl
->block_sizes
= NULL
; /* FIXME not for CONCAT */
1133 vcl
->conf
.sec_elmnt_count
= vd
->sec_elmnt_count
;
1134 if (alloc_other_bvds(super
, vcl
) != 0) {
1135 pr_err("%s could not allocate other bvds\n",
1140 super
->conflist
= vcl
;
1141 dl
->vlist
[vnum
++] = vcl
;
1143 memcpy(&vcl
->conf
, vd
, super
->conf_rec_len
*512);
1144 for (i
=0; i
< max_virt_disks
; i
++)
1145 if (memcmp(super
->virt
->entries
[i
].guid
,
1146 vcl
->conf
.guid
, DDF_GUID_LEN
)==0)
1148 if (i
< max_virt_disks
)
1155 static int load_super_ddf(struct supertype
*st
, int fd
,
1158 unsigned long long dsize
;
1159 struct ddf_super
*super
;
1162 if (get_dev_size(fd
, devname
, &dsize
) == 0)
1165 if (test_partition(fd
))
1166 /* DDF is not allowed on partitions */
1169 /* 32M is a lower bound */
1170 if (dsize
<= 32*1024*1024) {
1172 pr_err("%s is too small for ddf: "
1173 "size is %llu sectors.\n",
1179 pr_err("%s is an odd size for ddf: "
1180 "size is %llu bytes.\n",
1187 if (posix_memalign((void**)&super
, 512, sizeof(*super
))!= 0) {
1188 pr_err("malloc of %zu failed.\n",
1192 memset(super
, 0, sizeof(*super
));
1194 rv
= load_ddf_headers(fd
, super
, devname
);
1200 /* Have valid headers and have chosen the best. Let's read in the rest*/
1202 rv
= load_ddf_global(fd
, super
, devname
);
1206 pr_err("Failed to load all information "
1207 "sections on %s\n", devname
);
1212 rv
= load_ddf_local(fd
, super
, devname
, 0);
1216 pr_err("Failed to load all information "
1217 "sections on %s\n", devname
);
1222 /* Should possibly check the sections .... */
1225 if (st
->ss
== NULL
) {
1226 st
->ss
= &super_ddf
;
1227 st
->minor_version
= 0;
1234 static void free_super_ddf(struct supertype
*st
)
1236 struct ddf_super
*ddf
= st
->sb
;
1242 while (ddf
->conflist
) {
1243 struct vcl
*v
= ddf
->conflist
;
1244 ddf
->conflist
= v
->next
;
1246 free(v
->block_sizes
);
1249 v->other_bvds[0] points to beginning of buffer,
1250 see alloc_other_bvds()
1252 free(v
->other_bvds
[0]);
1255 while (ddf
->dlist
) {
1256 struct dl
*d
= ddf
->dlist
;
1257 ddf
->dlist
= d
->next
;
1264 while (ddf
->add_list
) {
1265 struct dl
*d
= ddf
->add_list
;
1266 ddf
->add_list
= d
->next
;
1277 static struct supertype
*match_metadata_desc_ddf(char *arg
)
1279 /* 'ddf' only supports containers */
1280 struct supertype
*st
;
1281 if (strcmp(arg
, "ddf") != 0 &&
1282 strcmp(arg
, "default") != 0
1286 st
= xcalloc(1, sizeof(*st
));
1287 st
->ss
= &super_ddf
;
1289 st
->minor_version
= 0;
1296 static mapping_t ddf_state
[] = {
1302 { "Partially Optimal", 5},
1308 static mapping_t ddf_init_state
[] = {
1309 { "Not Initialised", 0},
1310 { "QuickInit in Progress", 1},
1311 { "Fully Initialised", 2},
1315 static mapping_t ddf_access
[] = {
1319 { "Blocked (no access)", 3},
1323 static mapping_t ddf_level
[] = {
1324 { "RAID0", DDF_RAID0
},
1325 { "RAID1", DDF_RAID1
},
1326 { "RAID3", DDF_RAID3
},
1327 { "RAID4", DDF_RAID4
},
1328 { "RAID5", DDF_RAID5
},
1329 { "RAID1E",DDF_RAID1E
},
1330 { "JBOD", DDF_JBOD
},
1331 { "CONCAT",DDF_CONCAT
},
1332 { "RAID5E",DDF_RAID5E
},
1333 { "RAID5EE",DDF_RAID5EE
},
1334 { "RAID6", DDF_RAID6
},
1337 static mapping_t ddf_sec_level
[] = {
1338 { "Striped", DDF_2STRIPED
},
1339 { "Mirrored", DDF_2MIRRORED
},
1340 { "Concat", DDF_2CONCAT
},
1341 { "Spanned", DDF_2SPANNED
},
1346 static int all_ff(const char *guid
)
1349 for (i
= 0; i
< DDF_GUID_LEN
; i
++)
1350 if (guid
[i
] != (char)0xff)
1355 static const char *guid_str(const char *guid
)
1357 static char buf
[DDF_GUID_LEN
*2+1];
1360 for (i
= 0; i
< DDF_GUID_LEN
; i
++) {
1361 unsigned char c
= guid
[i
];
1362 if (c
>= 32 && c
< 127)
1363 p
+= sprintf(p
, "%c", c
);
1365 p
+= sprintf(p
, "%02x", c
);
1368 return (const char *) buf
;
1372 static void print_guid(char *guid
, int tstamp
)
1374 /* A GUIDs are part (or all) ASCII and part binary.
1375 * They tend to be space padded.
1376 * We print the GUID in HEX, then in parentheses add
1377 * any initial ASCII sequence, and a possible
1378 * time stamp from bytes 16-19
1380 int l
= DDF_GUID_LEN
;
1383 for (i
=0 ; i
<DDF_GUID_LEN
; i
++) {
1384 if ((i
&3)==0 && i
!= 0) printf(":");
1385 printf("%02X", guid
[i
]&255);
1389 while (l
&& guid
[l
-1] == ' ')
1391 for (i
=0 ; i
<l
; i
++) {
1392 if (guid
[i
] >= 0x20 && guid
[i
] < 0x7f)
1393 fputc(guid
[i
], stdout
);
1398 time_t then
= __be32_to_cpu(*(__u32
*)(guid
+16)) + DECADE
;
1401 tm
= localtime(&then
);
1402 strftime(tbuf
, 100, " %D %T",tm
);
1403 fputs(tbuf
, stdout
);
1408 static void examine_vd(int n
, struct ddf_super
*sb
, char *guid
)
1410 int crl
= sb
->conf_rec_len
;
1413 for (vcl
= sb
->conflist
; vcl
; vcl
= vcl
->next
) {
1415 struct vd_config
*vc
= &vcl
->conf
;
1417 if (!be32_eq(calc_crc(vc
, crl
*512), vc
->crc
))
1419 if (memcmp(vc
->guid
, guid
, DDF_GUID_LEN
) != 0)
1422 /* Ok, we know about this VD, let's give more details */
1423 printf(" Raid Devices[%d] : %d (", n
,
1424 be16_to_cpu(vc
->prim_elmnt_count
));
1425 for (i
= 0; i
< be16_to_cpu(vc
->prim_elmnt_count
); i
++) {
1427 int cnt
= be16_to_cpu(sb
->phys
->used_pdes
);
1428 for (j
=0; j
<cnt
; j
++)
1429 if (be32_eq(vc
->phys_refnum
[i
],
1430 sb
->phys
->entries
[j
].refnum
))
1439 if (vc
->chunk_shift
!= 255)
1440 printf(" Chunk Size[%d] : %d sectors\n", n
,
1441 1 << vc
->chunk_shift
);
1442 printf(" Raid Level[%d] : %s\n", n
,
1443 map_num(ddf_level
, vc
->prl
)?:"-unknown-");
1444 if (vc
->sec_elmnt_count
!= 1) {
1445 printf(" Secondary Position[%d] : %d of %d\n", n
,
1446 vc
->sec_elmnt_seq
, vc
->sec_elmnt_count
);
1447 printf(" Secondary Level[%d] : %s\n", n
,
1448 map_num(ddf_sec_level
, vc
->srl
) ?: "-unknown-");
1450 printf(" Device Size[%d] : %llu\n", n
,
1451 be64_to_cpu(vc
->blocks
)/2);
1452 printf(" Array Size[%d] : %llu\n", n
,
1453 be64_to_cpu(vc
->array_blocks
)/2);
1457 static void examine_vds(struct ddf_super
*sb
)
1459 int cnt
= be16_to_cpu(sb
->virt
->populated_vdes
);
1461 printf(" Virtual Disks : %d\n", cnt
);
1463 for (i
= 0; i
< be16_to_cpu(sb
->virt
->max_vdes
); i
++) {
1464 struct virtual_entry
*ve
= &sb
->virt
->entries
[i
];
1465 if (all_ff(ve
->guid
))
1468 printf(" VD GUID[%d] : ", i
); print_guid(ve
->guid
, 1);
1470 printf(" unit[%d] : %d\n", i
, be16_to_cpu(ve
->unit
));
1471 printf(" state[%d] : %s, %s%s\n", i
,
1472 map_num(ddf_state
, ve
->state
& 7),
1473 (ve
->state
& DDF_state_morphing
) ? "Morphing, ": "",
1474 (ve
->state
& DDF_state_inconsistent
)? "Not Consistent" : "Consistent");
1475 printf(" init state[%d] : %s\n", i
,
1476 map_num(ddf_init_state
, ve
->init_state
&DDF_initstate_mask
));
1477 printf(" access[%d] : %s\n", i
,
1478 map_num(ddf_access
, (ve
->init_state
& DDF_access_mask
) >> 6));
1479 printf(" Name[%d] : %.16s\n", i
, ve
->name
);
1480 examine_vd(i
, sb
, ve
->guid
);
1482 if (cnt
) printf("\n");
1485 static void examine_pds(struct ddf_super
*sb
)
1487 int cnt
= be16_to_cpu(sb
->phys
->used_pdes
);
1490 printf(" Physical Disks : %d\n", cnt
);
1491 printf(" Number RefNo Size Device Type/State\n");
1493 for (i
=0 ; i
<cnt
; i
++) {
1494 struct phys_disk_entry
*pd
= &sb
->phys
->entries
[i
];
1495 int type
= be16_to_cpu(pd
->type
);
1496 int state
= be16_to_cpu(pd
->state
);
1498 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1500 printf(" %3d %08x ", i
,
1501 be32_to_cpu(pd
->refnum
));
1503 be64_to_cpu(pd
->config_size
)>>1);
1504 for (dl
= sb
->dlist
; dl
; dl
= dl
->next
) {
1505 if (be32_eq(dl
->disk
.refnum
, pd
->refnum
)) {
1506 char *dv
= map_dev(dl
->major
, dl
->minor
, 0);
1508 printf("%-15s", dv
);
1515 printf(" %s%s%s%s%s",
1516 (type
&2) ? "active":"",
1517 (type
&4) ? "Global-Spare":"",
1518 (type
&8) ? "spare" : "",
1519 (type
&16)? ", foreign" : "",
1520 (type
&32)? "pass-through" : "");
1521 if (state
& DDF_Failed
)
1522 /* This over-rides these three */
1523 state
&= ~(DDF_Online
|DDF_Rebuilding
|DDF_Transition
);
1524 printf("/%s%s%s%s%s%s%s",
1525 (state
&1)? "Online": "Offline",
1526 (state
&2)? ", Failed": "",
1527 (state
&4)? ", Rebuilding": "",
1528 (state
&8)? ", in-transition": "",
1529 (state
&16)? ", SMART-errors": "",
1530 (state
&32)? ", Unrecovered-Read-Errors": "",
1531 (state
&64)? ", Missing" : "");
1536 static void examine_super_ddf(struct supertype
*st
, char *homehost
)
1538 struct ddf_super
*sb
= st
->sb
;
1540 printf(" Magic : %08x\n", be32_to_cpu(sb
->anchor
.magic
));
1541 printf(" Version : %.8s\n", sb
->anchor
.revision
);
1542 printf("Controller GUID : "); print_guid(sb
->controller
.guid
, 0);
1544 printf(" Container GUID : "); print_guid(sb
->anchor
.guid
, 1);
1546 printf(" Seq : %08x\n", be32_to_cpu(sb
->active
->seq
));
1547 printf(" Redundant hdr : %s\n", (be32_eq(sb
->secondary
.magic
,
1554 static unsigned int get_vd_num_of_subarray(struct supertype
*st
)
1557 * Figure out the VD number for this supertype.
1558 * Returns DDF_CONTAINER for the container itself,
1559 * and DDF_NOTFOUND on error.
1561 struct ddf_super
*ddf
= st
->sb
;
1566 if (*st
->container_devnm
== '\0')
1567 return DDF_CONTAINER
;
1569 sra
= sysfs_read(-1, st
->devnm
, GET_VERSION
);
1570 if (!sra
|| sra
->array
.major_version
!= -1 ||
1571 sra
->array
.minor_version
!= -2 ||
1572 !is_subarray(sra
->text_version
))
1573 return DDF_NOTFOUND
;
1575 sub
= strchr(sra
->text_version
+ 1, '/');
1577 vcnum
= strtoul(sub
+ 1, &end
, 10);
1578 if (sub
== NULL
|| *sub
== '\0' || *end
!= '\0' ||
1579 vcnum
>= be16_to_cpu(ddf
->active
->max_vd_entries
))
1580 return DDF_NOTFOUND
;
1585 static void brief_examine_super_ddf(struct supertype
*st
, int verbose
)
1587 /* We just write a generic DDF ARRAY entry
1591 getinfo_super_ddf(st
, &info
, NULL
);
1592 fname_from_uuid(st
, &info
, nbuf
, ':');
1594 printf("ARRAY metadata=ddf UUID=%s\n", nbuf
+ 5);
1597 static void brief_examine_subarrays_ddf(struct supertype
*st
, int verbose
)
1599 /* We write a DDF ARRAY member entry for each vd, identifying container
1600 * by uuid and member by unit number and uuid.
1602 struct ddf_super
*ddf
= st
->sb
;
1606 getinfo_super_ddf(st
, &info
, NULL
);
1607 fname_from_uuid(st
, &info
, nbuf
, ':');
1609 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++) {
1610 struct virtual_entry
*ve
= &ddf
->virt
->entries
[i
];
1614 if (all_ff(ve
->guid
))
1616 memcpy(vcl
.conf
.guid
, ve
->guid
, DDF_GUID_LEN
);
1617 ddf
->currentconf
=&vcl
;
1619 uuid_from_super_ddf(st
, info
.uuid
);
1620 fname_from_uuid(st
, &info
, nbuf1
, ':');
1621 _ddf_array_name(namebuf
, ddf
, i
);
1622 printf("ARRAY%s%s container=%s member=%d UUID=%s\n",
1623 namebuf
[0] == '\0' ? "" : " /dev/md/", namebuf
,
1624 nbuf
+5, i
, nbuf1
+5);
1628 static void export_examine_super_ddf(struct supertype
*st
)
1632 getinfo_super_ddf(st
, &info
, NULL
);
1633 fname_from_uuid(st
, &info
, nbuf
, ':');
1634 printf("MD_METADATA=ddf\n");
1635 printf("MD_LEVEL=container\n");
1636 printf("MD_UUID=%s\n", nbuf
+5);
1637 printf("MD_DEVICES=%u\n",
1638 be16_to_cpu(((struct ddf_super
*)st
->sb
)->phys
->used_pdes
));
1641 static int copy_metadata_ddf(struct supertype
*st
, int from
, int to
)
1644 unsigned long long dsize
, offset
;
1646 struct ddf_header
*ddf
;
1649 /* The meta consists of an anchor, a primary, and a secondary.
1650 * This all lives at the end of the device.
1651 * So it is easiest to find the earliest of primary and
1652 * secondary, and copy everything from there.
1654 * Anchor is 512 from end. It contains primary_lba and secondary_lba
1655 * we choose one of those
1658 if (posix_memalign(&buf
, 4096, 4096) != 0)
1661 if (!get_dev_size(from
, NULL
, &dsize
))
1664 if (lseek64(from
, dsize
-512, 0) < 0)
1666 if (read(from
, buf
, 512) != 512)
1669 if (!be32_eq(ddf
->magic
, DDF_HEADER_MAGIC
) ||
1670 !be32_eq(calc_crc(ddf
, 512), ddf
->crc
) ||
1671 (memcmp(ddf
->revision
, DDF_REVISION_0
, 8) != 0 &&
1672 memcmp(ddf
->revision
, DDF_REVISION_2
, 8) != 0))
1675 offset
= dsize
- 512;
1676 if ((be64_to_cpu(ddf
->primary_lba
) << 9) < offset
)
1677 offset
= be64_to_cpu(ddf
->primary_lba
) << 9;
1678 if ((be64_to_cpu(ddf
->secondary_lba
) << 9) < offset
)
1679 offset
= be64_to_cpu(ddf
->secondary_lba
) << 9;
1681 bytes
= dsize
- offset
;
1683 if (lseek64(from
, offset
, 0) < 0 ||
1684 lseek64(to
, offset
, 0) < 0)
1686 while (written
< bytes
) {
1687 int n
= bytes
- written
;
1690 if (read(from
, buf
, n
) != n
)
1692 if (write(to
, buf
, n
) != n
)
1703 static void detail_super_ddf(struct supertype
*st
, char *homehost
)
1706 * Could print DDF GUID
1707 * Need to find which array
1708 * If whole, briefly list all arrays
1713 static const char *vendors_with_variable_volume_UUID
[] = {
1717 static int volume_id_is_reliable(const struct ddf_super
*ddf
)
1719 int n
= ARRAY_SIZE(vendors_with_variable_volume_UUID
);
1721 for (i
= 0; i
< n
; i
++)
1722 if (!memcmp(ddf
->controller
.guid
,
1723 vendors_with_variable_volume_UUID
[i
], 8))
1728 static void uuid_of_ddf_subarray(const struct ddf_super
*ddf
,
1729 unsigned int vcnum
, int uuid
[4])
1731 char buf
[DDF_GUID_LEN
+18], sha
[20], *p
;
1732 struct sha1_ctx ctx
;
1733 if (volume_id_is_reliable(ddf
)) {
1734 uuid_from_ddf_guid(ddf
->virt
->entries
[vcnum
].guid
, uuid
);
1738 * Some fake RAID BIOSes (in particular, LSI ones) change the
1739 * VD GUID at every boot. These GUIDs are not suitable for
1740 * identifying an array. Luckily the header GUID appears to
1742 * We construct a pseudo-UUID from the header GUID and those
1743 * properties of the subarray that we expect to remain constant.
1745 memset(buf
, 0, sizeof(buf
));
1747 memcpy(p
, ddf
->anchor
.guid
, DDF_GUID_LEN
);
1749 memcpy(p
, ddf
->virt
->entries
[vcnum
].name
, 16);
1751 *((__u16
*) p
) = vcnum
;
1752 sha1_init_ctx(&ctx
);
1753 sha1_process_bytes(buf
, sizeof(buf
), &ctx
);
1754 sha1_finish_ctx(&ctx
, sha
);
1755 memcpy(uuid
, sha
, 4*4);
1758 static void brief_detail_super_ddf(struct supertype
*st
)
1762 struct ddf_super
*ddf
= st
->sb
;
1763 unsigned int vcnum
= get_vd_num_of_subarray(st
);
1764 if (vcnum
== DDF_CONTAINER
)
1765 uuid_from_super_ddf(st
, info
.uuid
);
1766 else if (vcnum
== DDF_NOTFOUND
)
1769 uuid_of_ddf_subarray(ddf
, vcnum
, info
.uuid
);
1770 fname_from_uuid(st
, &info
, nbuf
,':');
1771 printf(" UUID=%s", nbuf
+ 5);
1775 static int match_home_ddf(struct supertype
*st
, char *homehost
)
1777 /* It matches 'this' host if the controller is a
1778 * Linux-MD controller with vendor_data matching
1779 * the hostname. It would be nice if we could
1780 * test against controller found in /sys or somewhere...
1782 struct ddf_super
*ddf
= st
->sb
;
1787 len
= strlen(homehost
);
1789 return (memcmp(ddf
->controller
.guid
, T10
, 8) == 0 &&
1790 len
< sizeof(ddf
->controller
.vendor_data
) &&
1791 memcmp(ddf
->controller
.vendor_data
, homehost
,len
) == 0 &&
1792 ddf
->controller
.vendor_data
[len
] == 0);
1796 static int find_index_in_bvd(const struct ddf_super
*ddf
,
1797 const struct vd_config
*conf
, unsigned int n
,
1798 unsigned int *n_bvd
)
1801 * Find the index of the n-th valid physical disk in this BVD.
1802 * Unused entries can be sprinkled in with the used entries,
1807 i
< ddf
->mppe
&& j
< be16_to_cpu(conf
->prim_elmnt_count
);
1809 if (be32_to_cpu(conf
->phys_refnum
[i
]) != 0xffffffff) {
1817 dprintf("%s: couldn't find BVD member %u (total %u)\n",
1818 __func__
, n
, be16_to_cpu(conf
->prim_elmnt_count
));
1822 /* Given a member array instance number, and a raid disk within that instance,
1823 * find the vd_config structure. The offset of the given disk in the phys_refnum
1824 * table is returned in n_bvd.
1825 * For two-level members with a secondary raid level the vd_config for
1826 * the appropriate BVD is returned.
1827 * The return value is always &vlc->conf, where vlc is returned in last pointer.
1829 static struct vd_config
*find_vdcr(struct ddf_super
*ddf
, unsigned int inst
,
1831 unsigned int *n_bvd
, struct vcl
**vcl
)
1835 for (v
= ddf
->conflist
; v
; v
= v
->next
) {
1836 unsigned int nsec
, ibvd
= 0;
1837 struct vd_config
*conf
;
1838 if (inst
!= v
->vcnum
)
1841 if (conf
->sec_elmnt_count
== 1) {
1842 if (find_index_in_bvd(ddf
, conf
, n
, n_bvd
)) {
1848 if (v
->other_bvds
== NULL
) {
1849 pr_err("%s: BUG: other_bvds is NULL, nsec=%u\n",
1850 __func__
, conf
->sec_elmnt_count
);
1853 nsec
= n
/ be16_to_cpu(conf
->prim_elmnt_count
);
1854 if (conf
->sec_elmnt_seq
!= nsec
) {
1855 for (ibvd
= 1; ibvd
< conf
->sec_elmnt_count
; ibvd
++) {
1856 if (v
->other_bvds
[ibvd
-1]->sec_elmnt_seq
1860 if (ibvd
== conf
->sec_elmnt_count
)
1862 conf
= v
->other_bvds
[ibvd
-1];
1864 if (!find_index_in_bvd(ddf
, conf
,
1865 n
- nsec
*conf
->sec_elmnt_count
, n_bvd
))
1867 dprintf("%s: found disk %u as member %u in bvd %d of array %u\n"
1868 , __func__
, n
, *n_bvd
, ibvd
, inst
);
1873 pr_err("%s: Could't find disk %d in array %u\n", __func__
, n
, inst
);
1878 static int find_phys(const struct ddf_super
*ddf
, be32 phys_refnum
)
1880 /* Find the entry in phys_disk which has the given refnum
1881 * and return it's index
1884 for (i
= 0; i
< be16_to_cpu(ddf
->phys
->max_pdes
); i
++)
1885 if (be32_eq(ddf
->phys
->entries
[i
].refnum
, phys_refnum
))
1890 static void uuid_from_ddf_guid(const char *guid
, int uuid
[4])
1893 struct sha1_ctx ctx
;
1894 sha1_init_ctx(&ctx
);
1895 sha1_process_bytes(guid
, DDF_GUID_LEN
, &ctx
);
1896 sha1_finish_ctx(&ctx
, buf
);
1897 memcpy(uuid
, buf
, 4*4);
1900 static void uuid_from_super_ddf(struct supertype
*st
, int uuid
[4])
1902 /* The uuid returned here is used for:
1903 * uuid to put into bitmap file (Create, Grow)
1904 * uuid for backup header when saving critical section (Grow)
1905 * comparing uuids when re-adding a device into an array
1906 * In these cases the uuid required is that of the data-array,
1907 * not the device-set.
1908 * uuid to recognise same set when adding a missing device back
1909 * to an array. This is a uuid for the device-set.
1911 * For each of these we can make do with a truncated
1912 * or hashed uuid rather than the original, as long as
1914 * In the case of SVD we assume the BVD is of interest,
1915 * though that might be the case if a bitmap were made for
1916 * a mirrored SVD - worry about that later.
1917 * So we need to find the VD configuration record for the
1918 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1919 * The first 16 bytes of the sha1 of these is used.
1921 struct ddf_super
*ddf
= st
->sb
;
1922 struct vcl
*vcl
= ddf
->currentconf
;
1925 uuid_of_ddf_subarray(ddf
, vcl
->vcnum
, uuid
);
1927 uuid_from_ddf_guid(ddf
->anchor
.guid
, uuid
);
1930 static void getinfo_super_ddf(struct supertype
*st
, struct mdinfo
*info
, char *map
)
1932 struct ddf_super
*ddf
= st
->sb
;
1933 int map_disks
= info
->array
.raid_disks
;
1936 if (ddf
->currentconf
) {
1937 getinfo_super_ddf_bvd(st
, info
, map
);
1940 memset(info
, 0, sizeof(*info
));
1942 info
->array
.raid_disks
= be16_to_cpu(ddf
->phys
->used_pdes
);
1943 info
->array
.level
= LEVEL_CONTAINER
;
1944 info
->array
.layout
= 0;
1945 info
->array
.md_minor
= -1;
1946 cptr
= (__u32
*)(ddf
->anchor
.guid
+ 16);
1947 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*cptr
);
1949 info
->array
.utime
= 0;
1950 info
->array
.chunk_size
= 0;
1951 info
->container_enough
= 1;
1953 info
->disk
.major
= 0;
1954 info
->disk
.minor
= 0;
1956 struct phys_disk_entry
*pde
= NULL
;
1957 info
->disk
.number
= be32_to_cpu(ddf
->dlist
->disk
.refnum
);
1958 info
->disk
.raid_disk
= find_phys(ddf
, ddf
->dlist
->disk
.refnum
);
1960 info
->data_offset
= be64_to_cpu(ddf
->phys
->
1961 entries
[info
->disk
.raid_disk
].
1963 info
->component_size
= ddf
->dlist
->size
- info
->data_offset
;
1964 if (info
->disk
.raid_disk
>= 0)
1965 pde
= ddf
->phys
->entries
+ info
->disk
.raid_disk
;
1967 !(be16_to_cpu(pde
->state
) & DDF_Failed
))
1968 info
->disk
.state
= (1 << MD_DISK_SYNC
) | (1 << MD_DISK_ACTIVE
);
1970 info
->disk
.state
= 1 << MD_DISK_FAULTY
;
1972 info
->events
= be32_to_cpu(ddf
->active
->seq
);
1974 info
->disk
.number
= -1;
1975 info
->disk
.raid_disk
= -1;
1976 // info->disk.raid_disk = find refnum in the table and use index;
1977 info
->disk
.state
= (1 << MD_DISK_SYNC
) | (1 << MD_DISK_ACTIVE
);
1980 info
->recovery_start
= MaxSector
;
1981 info
->reshape_active
= 0;
1982 info
->recovery_blocked
= 0;
1985 info
->array
.major_version
= -1;
1986 info
->array
.minor_version
= -2;
1987 strcpy(info
->text_version
, "ddf");
1988 info
->safe_mode_delay
= 0;
1990 uuid_from_super_ddf(st
, info
->uuid
);
1994 for (i
= 0 ; i
< map_disks
; i
++) {
1995 if (i
< info
->array
.raid_disks
&&
1996 !(be16_to_cpu(ddf
->phys
->entries
[i
].state
)
2005 /* size of name must be at least 17 bytes! */
2006 static void _ddf_array_name(char *name
, const struct ddf_super
*ddf
, int i
)
2009 memcpy(name
, ddf
->virt
->entries
[i
].name
, 16);
2011 for(j
= 0; j
< 16; j
++)
2016 static void getinfo_super_ddf_bvd(struct supertype
*st
, struct mdinfo
*info
, char *map
)
2018 struct ddf_super
*ddf
= st
->sb
;
2019 struct vcl
*vc
= ddf
->currentconf
;
2020 int cd
= ddf
->currentdev
;
2024 int map_disks
= info
->array
.raid_disks
;
2026 struct vd_config
*conf
;
2028 memset(info
, 0, sizeof(*info
));
2029 if (layout_ddf2md(&vc
->conf
, &info
->array
) == -1)
2031 info
->array
.md_minor
= -1;
2032 cptr
= (__u32
*)(vc
->conf
.guid
+ 16);
2033 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*cptr
);
2034 info
->array
.utime
= DECADE
+ be32_to_cpu(vc
->conf
.timestamp
);
2035 info
->array
.chunk_size
= 512 << vc
->conf
.chunk_shift
;
2036 info
->custom_array_size
= 0;
2039 n_prim
= be16_to_cpu(conf
->prim_elmnt_count
);
2040 if (conf
->sec_elmnt_count
> 1 && cd
>= n_prim
) {
2041 int ibvd
= cd
/ n_prim
- 1;
2043 conf
= vc
->other_bvds
[ibvd
];
2046 if (cd
>= 0 && (unsigned)cd
< ddf
->mppe
) {
2048 be64_to_cpu(LBA_OFFSET(ddf
, conf
)[cd
]);
2049 if (vc
->block_sizes
)
2050 info
->component_size
= vc
->block_sizes
[cd
];
2052 info
->component_size
= be64_to_cpu(conf
->blocks
);
2055 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2056 if (be32_eq(dl
->disk
.refnum
, conf
->phys_refnum
[cd
]))
2059 info
->disk
.major
= 0;
2060 info
->disk
.minor
= 0;
2061 info
->disk
.state
= 0;
2063 info
->disk
.major
= dl
->major
;
2064 info
->disk
.minor
= dl
->minor
;
2065 info
->disk
.raid_disk
= cd
+ conf
->sec_elmnt_seq
2066 * be16_to_cpu(conf
->prim_elmnt_count
);
2067 info
->disk
.number
= dl
->pdnum
;
2068 info
->disk
.state
= 0;
2069 if (info
->disk
.number
>= 0 &&
2070 (be16_to_cpu(ddf
->phys
->entries
[info
->disk
.number
].state
) & DDF_Online
) &&
2071 !(be16_to_cpu(ddf
->phys
->entries
[info
->disk
.number
].state
) & DDF_Failed
))
2072 info
->disk
.state
= (1<<MD_DISK_SYNC
)|(1<<MD_DISK_ACTIVE
);
2073 info
->events
= be32_to_cpu(ddf
->active
->seq
);
2076 info
->container_member
= ddf
->currentconf
->vcnum
;
2078 info
->recovery_start
= MaxSector
;
2079 info
->resync_start
= 0;
2080 info
->reshape_active
= 0;
2081 info
->recovery_blocked
= 0;
2082 if (!(ddf
->virt
->entries
[info
->container_member
].state
2083 & DDF_state_inconsistent
) &&
2084 (ddf
->virt
->entries
[info
->container_member
].init_state
2085 & DDF_initstate_mask
)
2087 info
->resync_start
= MaxSector
;
2089 uuid_from_super_ddf(st
, info
->uuid
);
2091 info
->array
.major_version
= -1;
2092 info
->array
.minor_version
= -2;
2093 sprintf(info
->text_version
, "/%s/%d",
2094 st
->container_devnm
,
2095 info
->container_member
);
2096 info
->safe_mode_delay
= DDF_SAFE_MODE_DELAY
;
2098 _ddf_array_name(info
->name
, ddf
, info
->container_member
);
2101 for (j
= 0; j
< map_disks
; j
++) {
2103 if (j
< info
->array
.raid_disks
) {
2104 int i
= find_phys(ddf
, vc
->conf
.phys_refnum
[j
]);
2106 (be16_to_cpu(ddf
->phys
->entries
[i
].state
)
2108 !(be16_to_cpu(ddf
->phys
->entries
[i
].state
)
2115 static int update_super_ddf(struct supertype
*st
, struct mdinfo
*info
,
2117 char *devname
, int verbose
,
2118 int uuid_set
, char *homehost
)
2120 /* For 'assemble' and 'force' we need to return non-zero if any
2121 * change was made. For others, the return value is ignored.
2122 * Update options are:
2123 * force-one : This device looks a bit old but needs to be included,
2124 * update age info appropriately.
2125 * assemble: clear any 'faulty' flag to allow this device to
2127 * force-array: Array is degraded but being forced, mark it clean
2128 * if that will be needed to assemble it.
2130 * newdev: not used ????
2131 * grow: Array has gained a new device - this is currently for
2133 * resync: mark as dirty so a resync will happen.
2134 * uuid: Change the uuid of the array to match what is given
2135 * homehost: update the recorded homehost
2136 * name: update the name - preserving the homehost
2137 * _reshape_progress: record new reshape_progress position.
2139 * Following are not relevant for this version:
2140 * sparc2.2 : update from old dodgey metadata
2141 * super-minor: change the preferred_minor number
2142 * summaries: update redundant counters.
2145 // struct ddf_super *ddf = st->sb;
2146 // struct vd_config *vd = find_vdcr(ddf, info->container_member);
2147 // struct virtual_entry *ve = find_ve(ddf);
2149 /* we don't need to handle "force-*" or "assemble" as
2150 * there is no need to 'trick' the kernel. When the metadata is
2151 * first updated to activate the array, all the implied modifications
2155 if (strcmp(update
, "grow") == 0) {
2157 } else if (strcmp(update
, "resync") == 0) {
2158 // info->resync_checkpoint = 0;
2159 } else if (strcmp(update
, "homehost") == 0) {
2160 /* homehost is stored in controller->vendor_data,
2161 * or it is when we are the vendor
2163 // if (info->vendor_is_local)
2164 // strcpy(ddf->controller.vendor_data, homehost);
2166 } else if (strcmp(update
, "name") == 0) {
2167 /* name is stored in virtual_entry->name */
2168 // memset(ve->name, ' ', 16);
2169 // strncpy(ve->name, info->name, 16);
2171 } else if (strcmp(update
, "_reshape_progress") == 0) {
2172 /* We don't support reshape yet */
2173 } else if (strcmp(update
, "assemble") == 0 ) {
2174 /* Do nothing, just succeed */
2179 // update_all_csum(ddf);
2184 static void make_header_guid(char *guid
)
2187 /* Create a DDF Header of Virtual Disk GUID */
2189 /* 24 bytes of fiction required.
2190 * first 8 are a 'vendor-id' - "Linux-MD"
2191 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
2192 * Remaining 8 random number plus timestamp
2194 memcpy(guid
, T10
, sizeof(T10
));
2195 stamp
= cpu_to_be32(0xdeadbeef);
2196 memcpy(guid
+8, &stamp
, 4);
2197 stamp
= cpu_to_be32(0);
2198 memcpy(guid
+12, &stamp
, 4);
2199 stamp
= cpu_to_be32(time(0) - DECADE
);
2200 memcpy(guid
+16, &stamp
, 4);
2201 stamp
._v32
= random32();
2202 memcpy(guid
+20, &stamp
, 4);
2205 static unsigned int find_unused_vde(const struct ddf_super
*ddf
)
2208 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++) {
2209 if (all_ff(ddf
->virt
->entries
[i
].guid
))
2212 return DDF_NOTFOUND
;
2215 static unsigned int find_vde_by_name(const struct ddf_super
*ddf
,
2220 return DDF_NOTFOUND
;
2221 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++) {
2222 if (all_ff(ddf
->virt
->entries
[i
].guid
))
2224 if (!strncmp(name
, ddf
->virt
->entries
[i
].name
,
2225 sizeof(ddf
->virt
->entries
[i
].name
)))
2228 return DDF_NOTFOUND
;
2232 static unsigned int find_vde_by_guid(const struct ddf_super
*ddf
,
2236 if (guid
== NULL
|| all_ff(guid
))
2237 return DDF_NOTFOUND
;
2238 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++)
2239 if (!memcmp(ddf
->virt
->entries
[i
].guid
, guid
, DDF_GUID_LEN
))
2241 return DDF_NOTFOUND
;
2245 static int init_super_ddf(struct supertype
*st
,
2246 mdu_array_info_t
*info
,
2247 unsigned long long size
, char *name
, char *homehost
,
2248 int *uuid
, unsigned long long data_offset
)
2250 /* This is primarily called by Create when creating a new array.
2251 * We will then get add_to_super called for each component, and then
2252 * write_init_super called to write it out to each device.
2253 * For DDF, Create can create on fresh devices or on a pre-existing
2255 * To create on a pre-existing array a different method will be called.
2256 * This one is just for fresh drives.
2258 * We need to create the entire 'ddf' structure which includes:
2259 * DDF headers - these are easy.
2260 * Controller data - a Sector describing this controller .. not that
2261 * this is a controller exactly.
2262 * Physical Disk Record - one entry per device, so
2263 * leave plenty of space.
2264 * Virtual Disk Records - again, just leave plenty of space.
2265 * This just lists VDs, doesn't give details.
2266 * Config records - describe the VDs that use this disk
2267 * DiskData - describes 'this' device.
2268 * BadBlockManagement - empty
2269 * Diag Space - empty
2270 * Vendor Logs - Could we put bitmaps here?
2273 struct ddf_super
*ddf
;
2276 int max_phys_disks
, max_virt_disks
;
2277 unsigned long long sector
;
2281 struct phys_disk
*pd
;
2282 struct virtual_disk
*vd
;
2284 if (data_offset
!= INVALID_SECTORS
) {
2285 pr_err("data-offset not supported by DDF\n");
2290 return init_super_ddf_bvd(st
, info
, size
, name
, homehost
, uuid
,
2293 if (posix_memalign((void**)&ddf
, 512, sizeof(*ddf
)) != 0) {
2294 pr_err("%s could not allocate superblock\n", __func__
);
2297 memset(ddf
, 0, sizeof(*ddf
));
2301 /* zeroing superblock */
2305 /* At least 32MB *must* be reserved for the ddf. So let's just
2306 * start 32MB from the end, and put the primary header there.
2307 * Don't do secondary for now.
2308 * We don't know exactly where that will be yet as it could be
2309 * different on each device. So just set up the lengths.
2312 ddf
->anchor
.magic
= DDF_HEADER_MAGIC
;
2313 make_header_guid(ddf
->anchor
.guid
);
2315 memcpy(ddf
->anchor
.revision
, DDF_REVISION_2
, 8);
2316 ddf
->anchor
.seq
= cpu_to_be32(1);
2317 ddf
->anchor
.timestamp
= cpu_to_be32(time(0) - DECADE
);
2318 ddf
->anchor
.openflag
= 0xFF;
2319 ddf
->anchor
.foreignflag
= 0;
2320 ddf
->anchor
.enforcegroups
= 0; /* Is this best?? */
2321 ddf
->anchor
.pad0
= 0xff;
2322 memset(ddf
->anchor
.pad1
, 0xff, 12);
2323 memset(ddf
->anchor
.header_ext
, 0xff, 32);
2324 ddf
->anchor
.primary_lba
= cpu_to_be64(~(__u64
)0);
2325 ddf
->anchor
.secondary_lba
= cpu_to_be64(~(__u64
)0);
2326 ddf
->anchor
.type
= DDF_HEADER_ANCHOR
;
2327 memset(ddf
->anchor
.pad2
, 0xff, 3);
2328 ddf
->anchor
.workspace_len
= cpu_to_be32(32768); /* Must be reserved */
2329 /* Put this at bottom of 32M reserved.. */
2330 ddf
->anchor
.workspace_lba
= cpu_to_be64(~(__u64
)0);
2331 max_phys_disks
= 1023; /* Should be enough, 4095 is also allowed */
2332 ddf
->anchor
.max_pd_entries
= cpu_to_be16(max_phys_disks
);
2333 max_virt_disks
= 255; /* 15, 63, 255, 1024, 4095 are all allowed */
2334 ddf
->anchor
.max_vd_entries
= cpu_to_be16(max_virt_disks
);
2336 ddf
->anchor
.max_partitions
= cpu_to_be16(ddf
->max_part
);
2337 ddf
->mppe
= 256; /* 16, 64, 256, 1024, 4096 are all allowed */
2338 ddf
->conf_rec_len
= 1 + ROUND_UP(ddf
->mppe
* (4+8), 512)/512;
2339 ddf
->anchor
.config_record_len
= cpu_to_be16(ddf
->conf_rec_len
);
2340 ddf
->anchor
.max_primary_element_entries
= cpu_to_be16(ddf
->mppe
);
2341 memset(ddf
->anchor
.pad3
, 0xff, 54);
2342 /* Controller section is one sector long immediately
2343 * after the ddf header */
2345 ddf
->anchor
.controller_section_offset
= cpu_to_be32(sector
);
2346 ddf
->anchor
.controller_section_length
= cpu_to_be32(1);
2349 /* phys is 8 sectors after that */
2350 pdsize
= ROUND_UP(sizeof(struct phys_disk
) +
2351 sizeof(struct phys_disk_entry
)*max_phys_disks
,
2353 switch(pdsize
/512) {
2354 case 2: case 8: case 32: case 128: case 512: break;
2357 ddf
->anchor
.phys_section_offset
= cpu_to_be32(sector
);
2358 ddf
->anchor
.phys_section_length
=
2359 cpu_to_be32(pdsize
/512); /* max_primary_element_entries/8 */
2360 sector
+= pdsize
/512;
2362 /* virt is another 32 sectors */
2363 vdsize
= ROUND_UP(sizeof(struct virtual_disk
) +
2364 sizeof(struct virtual_entry
) * max_virt_disks
,
2366 switch(vdsize
/512) {
2367 case 2: case 8: case 32: case 128: case 512: break;
2370 ddf
->anchor
.virt_section_offset
= cpu_to_be32(sector
);
2371 ddf
->anchor
.virt_section_length
=
2372 cpu_to_be32(vdsize
/512); /* max_vd_entries/8 */
2373 sector
+= vdsize
/512;
2375 clen
= ddf
->conf_rec_len
* (ddf
->max_part
+1);
2376 ddf
->anchor
.config_section_offset
= cpu_to_be32(sector
);
2377 ddf
->anchor
.config_section_length
= cpu_to_be32(clen
);
2380 ddf
->anchor
.data_section_offset
= cpu_to_be32(sector
);
2381 ddf
->anchor
.data_section_length
= cpu_to_be32(1);
2384 ddf
->anchor
.bbm_section_length
= cpu_to_be32(0);
2385 ddf
->anchor
.bbm_section_offset
= cpu_to_be32(0xFFFFFFFF);
2386 ddf
->anchor
.diag_space_length
= cpu_to_be32(0);
2387 ddf
->anchor
.diag_space_offset
= cpu_to_be32(0xFFFFFFFF);
2388 ddf
->anchor
.vendor_length
= cpu_to_be32(0);
2389 ddf
->anchor
.vendor_offset
= cpu_to_be32(0xFFFFFFFF);
2391 memset(ddf
->anchor
.pad4
, 0xff, 256);
2393 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
2394 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
2396 ddf
->primary
.openflag
= 1; /* I guess.. */
2397 ddf
->primary
.type
= DDF_HEADER_PRIMARY
;
2399 ddf
->secondary
.openflag
= 1; /* I guess.. */
2400 ddf
->secondary
.type
= DDF_HEADER_SECONDARY
;
2402 ddf
->active
= &ddf
->primary
;
2404 ddf
->controller
.magic
= DDF_CONTROLLER_MAGIC
;
2406 /* 24 more bytes of fiction required.
2407 * first 8 are a 'vendor-id' - "Linux-MD"
2408 * Remaining 16 are serial number.... maybe a hostname would do?
2410 memcpy(ddf
->controller
.guid
, T10
, sizeof(T10
));
2411 gethostname(hostname
, sizeof(hostname
));
2412 hostname
[sizeof(hostname
) - 1] = 0;
2413 hostlen
= strlen(hostname
);
2414 memcpy(ddf
->controller
.guid
+ 24 - hostlen
, hostname
, hostlen
);
2415 for (i
= strlen(T10
) ; i
+hostlen
< 24; i
++)
2416 ddf
->controller
.guid
[i
] = ' ';
2418 ddf
->controller
.type
.vendor_id
= cpu_to_be16(0xDEAD);
2419 ddf
->controller
.type
.device_id
= cpu_to_be16(0xBEEF);
2420 ddf
->controller
.type
.sub_vendor_id
= cpu_to_be16(0);
2421 ddf
->controller
.type
.sub_device_id
= cpu_to_be16(0);
2422 memcpy(ddf
->controller
.product_id
, "What Is My PID??", 16);
2423 memset(ddf
->controller
.pad
, 0xff, 8);
2424 memset(ddf
->controller
.vendor_data
, 0xff, 448);
2425 if (homehost
&& strlen(homehost
) < 440)
2426 strcpy((char*)ddf
->controller
.vendor_data
, homehost
);
2428 if (posix_memalign((void**)&pd
, 512, pdsize
) != 0) {
2429 pr_err("%s could not allocate pd\n", __func__
);
2433 ddf
->pdsize
= pdsize
;
2435 memset(pd
, 0xff, pdsize
);
2436 memset(pd
, 0, sizeof(*pd
));
2437 pd
->magic
= DDF_PHYS_RECORDS_MAGIC
;
2438 pd
->used_pdes
= cpu_to_be16(0);
2439 pd
->max_pdes
= cpu_to_be16(max_phys_disks
);
2440 memset(pd
->pad
, 0xff, 52);
2441 for (i
= 0; i
< max_phys_disks
; i
++)
2442 memset(pd
->entries
[i
].guid
, 0xff, DDF_GUID_LEN
);
2444 if (posix_memalign((void**)&vd
, 512, vdsize
) != 0) {
2445 pr_err("%s could not allocate vd\n", __func__
);
2449 ddf
->vdsize
= vdsize
;
2450 memset(vd
, 0, vdsize
);
2451 vd
->magic
= DDF_VIRT_RECORDS_MAGIC
;
2452 vd
->populated_vdes
= cpu_to_be16(0);
2453 vd
->max_vdes
= cpu_to_be16(max_virt_disks
);
2454 memset(vd
->pad
, 0xff, 52);
2456 for (i
=0; i
<max_virt_disks
; i
++)
2457 memset(&vd
->entries
[i
], 0xff, sizeof(struct virtual_entry
));
2460 ddf_set_updates_pending(ddf
);
2464 static int chunk_to_shift(int chunksize
)
2466 return ffs(chunksize
/512)-1;
2471 unsigned long long start
, size
;
2473 static int cmp_extent(const void *av
, const void *bv
)
2475 const struct extent
*a
= av
;
2476 const struct extent
*b
= bv
;
2477 if (a
->start
< b
->start
)
2479 if (a
->start
> b
->start
)
2484 static struct extent
*get_extents(struct ddf_super
*ddf
, struct dl
*dl
)
2486 /* Find a list of used extents on the give physical device
2487 * (dnum) of the given ddf.
2488 * Return a malloced array of 'struct extent'
2497 state
= be16_to_cpu(ddf
->phys
->entries
[dl
->pdnum
].state
);
2499 if ((state
& (DDF_Online
|DDF_Failed
|DDF_Missing
)) != DDF_Online
)
2502 rv
= xmalloc(sizeof(struct extent
) * (ddf
->max_part
+ 2));
2504 for (i
= 0; i
< ddf
->max_part
; i
++) {
2505 const struct vd_config
*bvd
;
2507 struct vcl
*v
= dl
->vlist
[i
];
2509 get_pd_index_from_refnum(v
, dl
->disk
.refnum
, ddf
->mppe
,
2510 &bvd
, &ibvd
) == DDF_NOTFOUND
)
2512 rv
[n
].start
= be64_to_cpu(LBA_OFFSET(ddf
, bvd
)[ibvd
]);
2513 rv
[n
].size
= be64_to_cpu(bvd
->blocks
);
2516 qsort(rv
, n
, sizeof(*rv
), cmp_extent
);
2518 rv
[n
].start
= be64_to_cpu(ddf
->phys
->entries
[dl
->pdnum
].config_size
);
2524 static int init_super_ddf_bvd(struct supertype
*st
,
2525 mdu_array_info_t
*info
,
2526 unsigned long long size
,
2527 char *name
, char *homehost
,
2528 int *uuid
, unsigned long long data_offset
)
2530 /* We are creating a BVD inside a pre-existing container.
2531 * so st->sb is already set.
2532 * We need to create a new vd_config and a new virtual_entry
2534 struct ddf_super
*ddf
= st
->sb
;
2535 unsigned int venum
, i
;
2536 struct virtual_entry
*ve
;
2538 struct vd_config
*vc
;
2540 if (find_vde_by_name(ddf
, name
) != DDF_NOTFOUND
) {
2541 pr_err("This ddf already has an array called %s\n", name
);
2544 venum
= find_unused_vde(ddf
);
2545 if (venum
== DDF_NOTFOUND
) {
2546 pr_err("Cannot find spare slot for virtual disk\n");
2549 ve
= &ddf
->virt
->entries
[venum
];
2551 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
2552 * timestamp, random number
2554 make_header_guid(ve
->guid
);
2555 ve
->unit
= cpu_to_be16(info
->md_minor
);
2557 ve
->guid_crc
._v16
= crc32(0, (unsigned char *)ddf
->anchor
.guid
,
2559 ve
->type
= cpu_to_be16(0);
2560 ve
->state
= DDF_state_degraded
; /* Will be modified as devices are added */
2561 if (info
->state
& 1) /* clean */
2562 ve
->init_state
= DDF_init_full
;
2564 ve
->init_state
= DDF_init_not
;
2566 memset(ve
->pad1
, 0xff, 14);
2567 memset(ve
->name
, ' ', 16);
2569 strncpy(ve
->name
, name
, 16);
2570 ddf
->virt
->populated_vdes
=
2571 cpu_to_be16(be16_to_cpu(ddf
->virt
->populated_vdes
)+1);
2573 /* Now create a new vd_config */
2574 if (posix_memalign((void**)&vcl
, 512,
2575 (offsetof(struct vcl
, conf
) + ddf
->conf_rec_len
* 512)) != 0) {
2576 pr_err("%s could not allocate vd_config\n", __func__
);
2580 vcl
->block_sizes
= NULL
; /* FIXME not for CONCAT */
2583 vc
->magic
= DDF_VD_CONF_MAGIC
;
2584 memcpy(vc
->guid
, ve
->guid
, DDF_GUID_LEN
);
2585 vc
->timestamp
= cpu_to_be32(time(0)-DECADE
);
2586 vc
->seqnum
= cpu_to_be32(1);
2587 memset(vc
->pad0
, 0xff, 24);
2588 vc
->chunk_shift
= chunk_to_shift(info
->chunk_size
);
2589 if (layout_md2ddf(info
, vc
) == -1 ||
2590 be16_to_cpu(vc
->prim_elmnt_count
) > ddf
->mppe
) {
2591 pr_err("%s: unsupported RAID level/layout %d/%d with %d disks\n",
2592 __func__
, info
->level
, info
->layout
, info
->raid_disks
);
2596 vc
->sec_elmnt_seq
= 0;
2597 if (alloc_other_bvds(ddf
, vcl
) != 0) {
2598 pr_err("%s could not allocate other bvds\n",
2603 vc
->blocks
= cpu_to_be64(info
->size
* 2);
2604 vc
->array_blocks
= cpu_to_be64(
2605 calc_array_size(info
->level
, info
->raid_disks
, info
->layout
,
2606 info
->chunk_size
, info
->size
*2));
2607 memset(vc
->pad1
, 0xff, 8);
2608 vc
->spare_refs
[0] = cpu_to_be32(0xffffffff);
2609 vc
->spare_refs
[1] = cpu_to_be32(0xffffffff);
2610 vc
->spare_refs
[2] = cpu_to_be32(0xffffffff);
2611 vc
->spare_refs
[3] = cpu_to_be32(0xffffffff);
2612 vc
->spare_refs
[4] = cpu_to_be32(0xffffffff);
2613 vc
->spare_refs
[5] = cpu_to_be32(0xffffffff);
2614 vc
->spare_refs
[6] = cpu_to_be32(0xffffffff);
2615 vc
->spare_refs
[7] = cpu_to_be32(0xffffffff);
2616 memset(vc
->cache_pol
, 0, 8);
2618 memset(vc
->pad2
, 0xff, 3);
2619 memset(vc
->pad3
, 0xff, 52);
2620 memset(vc
->pad4
, 0xff, 192);
2621 memset(vc
->v0
, 0xff, 32);
2622 memset(vc
->v1
, 0xff, 32);
2623 memset(vc
->v2
, 0xff, 16);
2624 memset(vc
->v3
, 0xff, 16);
2625 memset(vc
->vendor
, 0xff, 32);
2627 memset(vc
->phys_refnum
, 0xff, 4*ddf
->mppe
);
2628 memset(vc
->phys_refnum
+ddf
->mppe
, 0x00, 8*ddf
->mppe
);
2630 for (i
= 1; i
< vc
->sec_elmnt_count
; i
++) {
2631 memcpy(vcl
->other_bvds
[i
-1], vc
, ddf
->conf_rec_len
* 512);
2632 vcl
->other_bvds
[i
-1]->sec_elmnt_seq
= i
;
2635 vcl
->next
= ddf
->conflist
;
2636 ddf
->conflist
= vcl
;
2637 ddf
->currentconf
= vcl
;
2638 ddf_set_updates_pending(ddf
);
2643 static void add_to_super_ddf_bvd(struct supertype
*st
,
2644 mdu_disk_info_t
*dk
, int fd
, char *devname
)
2646 /* fd and devname identify a device within the ddf container (st).
2647 * dk identifies a location in the new BVD.
2648 * We need to find suitable free space in that device and update
2649 * the phys_refnum and lba_offset for the newly created vd_config.
2650 * We might also want to update the type in the phys_disk
2653 * Alternately: fd == -1 and we have already chosen which device to
2654 * use and recorded in dlist->raid_disk;
2657 struct ddf_super
*ddf
= st
->sb
;
2658 struct vd_config
*vc
;
2660 unsigned long long blocks
, pos
, esize
;
2662 unsigned int raid_disk
= dk
->raid_disk
;
2665 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2666 if (dl
->raiddisk
== dk
->raid_disk
)
2669 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2670 if (dl
->major
== dk
->major
&&
2671 dl
->minor
== dk
->minor
)
2674 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
2677 vc
= &ddf
->currentconf
->conf
;
2678 if (vc
->sec_elmnt_count
> 1) {
2679 unsigned int n
= be16_to_cpu(vc
->prim_elmnt_count
);
2681 vc
= ddf
->currentconf
->other_bvds
[raid_disk
/ n
- 1];
2685 ex
= get_extents(ddf
, dl
);
2690 blocks
= be64_to_cpu(vc
->blocks
);
2691 if (ddf
->currentconf
->block_sizes
)
2692 blocks
= ddf
->currentconf
->block_sizes
[dk
->raid_disk
];
2696 esize
= ex
[i
].start
- pos
;
2697 if (esize
>= blocks
)
2699 pos
= ex
[i
].start
+ ex
[i
].size
;
2701 } while (ex
[i
-1].size
);
2707 ddf
->currentdev
= dk
->raid_disk
;
2708 vc
->phys_refnum
[raid_disk
] = dl
->disk
.refnum
;
2709 LBA_OFFSET(ddf
, vc
)[raid_disk
] = cpu_to_be64(pos
);
2711 for (i
= 0; i
< ddf
->max_part
; i
++)
2712 if (dl
->vlist
[i
] == NULL
)
2714 if (i
== ddf
->max_part
)
2716 dl
->vlist
[i
] = ddf
->currentconf
;
2721 dl
->devname
= devname
;
2723 /* Check if we can mark array as optimal yet */
2724 i
= ddf
->currentconf
->vcnum
;
2725 ddf
->virt
->entries
[i
].state
=
2726 (ddf
->virt
->entries
[i
].state
& ~DDF_state_mask
)
2727 | get_svd_state(ddf
, ddf
->currentconf
);
2728 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
2729 cpu_to_be16(DDF_Global_Spare
));
2730 be16_set(ddf
->phys
->entries
[dl
->pdnum
].type
,
2731 cpu_to_be16(DDF_Active_in_VD
));
2732 dprintf("%s: added disk %d/%08x to VD %d/%s as disk %d\n",
2733 __func__
, dl
->pdnum
, be32_to_cpu(dl
->disk
.refnum
),
2734 ddf
->currentconf
->vcnum
, guid_str(vc
->guid
),
2736 ddf_set_updates_pending(ddf
);
2739 static unsigned int find_unused_pde(const struct ddf_super
*ddf
)
2742 for (i
= 0; i
< be16_to_cpu(ddf
->phys
->max_pdes
); i
++) {
2743 if (all_ff(ddf
->phys
->entries
[i
].guid
))
2746 return DDF_NOTFOUND
;
2749 static void _set_config_size(struct phys_disk_entry
*pde
, const struct dl
*dl
)
2752 cfs
= min(dl
->size
- 32*1024*2ULL, be64_to_cpu(dl
->primary_lba
));
2753 t
= be64_to_cpu(dl
->secondary_lba
);
2757 * Some vendor DDF structures interpret workspace_lba
2758 * very differently than we do: Make a sanity check on the value.
2760 t
= be64_to_cpu(dl
->workspace_lba
);
2762 __u64 wsp
= cfs
- t
;
2763 if (wsp
> 1024*1024*2ULL && wsp
> dl
->size
/ 16) {
2764 pr_err("%s: %x:%x: workspace size 0x%llx too big, ignoring\n",
2765 __func__
, dl
->major
, dl
->minor
, wsp
);
2769 pde
->config_size
= cpu_to_be64(cfs
);
2770 dprintf("%s: %x:%x config_size %llx, DDF structure is %llx blocks\n",
2771 __func__
, dl
->major
, dl
->minor
, cfs
, dl
->size
-cfs
);
2774 /* Add a device to a container, either while creating it or while
2775 * expanding a pre-existing container
2777 static int add_to_super_ddf(struct supertype
*st
,
2778 mdu_disk_info_t
*dk
, int fd
, char *devname
,
2779 unsigned long long data_offset
)
2781 struct ddf_super
*ddf
= st
->sb
;
2785 unsigned long long size
;
2786 struct phys_disk_entry
*pde
;
2791 if (ddf
->currentconf
) {
2792 add_to_super_ddf_bvd(st
, dk
, fd
, devname
);
2796 /* This is device numbered dk->number. We need to create
2797 * a phys_disk entry and a more detailed disk_data entry.
2800 n
= find_unused_pde(ddf
);
2801 if (n
== DDF_NOTFOUND
) {
2802 pr_err("%s: No free slot in array, cannot add disk\n",
2806 pde
= &ddf
->phys
->entries
[n
];
2807 get_dev_size(fd
, NULL
, &size
);
2808 if (size
<= 32*1024*1024) {
2809 pr_err("%s: device size must be at least 32MB\n",
2815 if (posix_memalign((void**)&dd
, 512,
2816 sizeof(*dd
) + sizeof(dd
->vlist
[0]) * ddf
->max_part
) != 0) {
2817 pr_err("%s could allocate buffer for new disk, aborting\n",
2821 dd
->major
= major(stb
.st_rdev
);
2822 dd
->minor
= minor(stb
.st_rdev
);
2823 dd
->devname
= devname
;
2827 dd
->disk
.magic
= DDF_PHYS_DATA_MAGIC
;
2829 tm
= localtime(&now
);
2830 sprintf(dd
->disk
.guid
, "%8s%04d%02d%02d",
2831 T10
, tm
->tm_year
+1900, tm
->tm_mon
+1, tm
->tm_mday
);
2832 tptr
= (__u32
*)(dd
->disk
.guid
+ 16);
2833 *tptr
++ = random32();
2837 /* Cannot be bothered finding a CRC of some irrelevant details*/
2838 dd
->disk
.refnum
._v32
= random32();
2839 for (i
= be16_to_cpu(ddf
->active
->max_pd_entries
);
2841 if (be32_eq(ddf
->phys
->entries
[i
-1].refnum
,
2846 dd
->disk
.forced_ref
= 1;
2847 dd
->disk
.forced_guid
= 1;
2848 memset(dd
->disk
.vendor
, ' ', 32);
2849 memcpy(dd
->disk
.vendor
, "Linux", 5);
2850 memset(dd
->disk
.pad
, 0xff, 442);
2851 for (i
= 0; i
< ddf
->max_part
; i
++)
2852 dd
->vlist
[i
] = NULL
;
2856 if (st
->update_tail
) {
2857 int len
= (sizeof(struct phys_disk
) +
2858 sizeof(struct phys_disk_entry
));
2859 struct phys_disk
*pd
;
2862 pd
->magic
= DDF_PHYS_RECORDS_MAGIC
;
2863 pd
->used_pdes
= cpu_to_be16(n
);
2864 pde
= &pd
->entries
[0];
2867 ddf
->phys
->used_pdes
= cpu_to_be16(
2868 1 + be16_to_cpu(ddf
->phys
->used_pdes
));
2870 memcpy(pde
->guid
, dd
->disk
.guid
, DDF_GUID_LEN
);
2871 pde
->refnum
= dd
->disk
.refnum
;
2872 pde
->type
= cpu_to_be16(DDF_Forced_PD_GUID
| DDF_Global_Spare
);
2873 pde
->state
= cpu_to_be16(DDF_Online
);
2876 * If there is already a device in dlist, try to reserve the same
2877 * amount of workspace. Otherwise, use 32MB.
2878 * We checked disk size above already.
2880 #define __calc_lba(new, old, lba, mb) do { \
2881 unsigned long long dif; \
2882 if ((old) != NULL) \
2883 dif = (old)->size - be64_to_cpu((old)->lba); \
2885 dif = (new)->size; \
2886 if ((new)->size > dif) \
2887 (new)->lba = cpu_to_be64((new)->size - dif); \
2889 (new)->lba = cpu_to_be64((new)->size - (mb*1024*2)); \
2891 __calc_lba(dd
, ddf
->dlist
, workspace_lba
, 32);
2892 __calc_lba(dd
, ddf
->dlist
, primary_lba
, 16);
2893 if (ddf
->dlist
== NULL
||
2894 be64_to_cpu(ddf
->dlist
->secondary_lba
) != ~(__u64
)0)
2895 __calc_lba(dd
, ddf
->dlist
, secondary_lba
, 32);
2896 _set_config_size(pde
, dd
);
2898 sprintf(pde
->path
, "%17.17s","Information: nil") ;
2899 memset(pde
->pad
, 0xff, 6);
2901 if (st
->update_tail
) {
2902 dd
->next
= ddf
->add_list
;
2905 dd
->next
= ddf
->dlist
;
2907 ddf_set_updates_pending(ddf
);
2913 static int remove_from_super_ddf(struct supertype
*st
, mdu_disk_info_t
*dk
)
2915 struct ddf_super
*ddf
= st
->sb
;
2918 /* mdmon has noticed that this disk (dk->major/dk->minor) has
2919 * disappeared from the container.
2920 * We need to arrange that it disappears from the metadata and
2921 * internal data structures too.
2922 * Most of the work is done by ddf_process_update which edits
2923 * the metadata and closes the file handle and attaches the memory
2924 * where free_updates will free it.
2926 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2927 if (dl
->major
== dk
->major
&&
2928 dl
->minor
== dk
->minor
)
2930 if (!dl
|| dl
->pdnum
< 0)
2933 if (st
->update_tail
) {
2934 int len
= (sizeof(struct phys_disk
) +
2935 sizeof(struct phys_disk_entry
));
2936 struct phys_disk
*pd
;
2939 pd
->magic
= DDF_PHYS_RECORDS_MAGIC
;
2940 pd
->used_pdes
= cpu_to_be16(dl
->pdnum
);
2941 pd
->entries
[0].state
= cpu_to_be16(DDF_Missing
);
2942 append_metadata_update(st
, pd
, len
);
2949 * This is the write_init_super method for a ddf container. It is
2950 * called when creating a container or adding another device to a
2954 static int __write_ddf_structure(struct dl
*d
, struct ddf_super
*ddf
, __u8 type
)
2956 unsigned long long sector
;
2957 struct ddf_header
*header
;
2958 int fd
, i
, n_config
, conf_size
, buf_size
;
2965 case DDF_HEADER_PRIMARY
:
2966 header
= &ddf
->primary
;
2967 sector
= be64_to_cpu(header
->primary_lba
);
2969 case DDF_HEADER_SECONDARY
:
2970 header
= &ddf
->secondary
;
2971 sector
= be64_to_cpu(header
->secondary_lba
);
2976 if (sector
== ~(__u64
)0)
2979 header
->type
= type
;
2980 header
->openflag
= 1;
2981 header
->crc
= calc_crc(header
, 512);
2983 lseek64(fd
, sector
<<9, 0);
2984 if (write(fd
, header
, 512) < 0)
2987 ddf
->controller
.crc
= calc_crc(&ddf
->controller
, 512);
2988 if (write(fd
, &ddf
->controller
, 512) < 0)
2991 ddf
->phys
->crc
= calc_crc(ddf
->phys
, ddf
->pdsize
);
2992 if (write(fd
, ddf
->phys
, ddf
->pdsize
) < 0)
2994 ddf
->virt
->crc
= calc_crc(ddf
->virt
, ddf
->vdsize
);
2995 if (write(fd
, ddf
->virt
, ddf
->vdsize
) < 0)
2998 /* Now write lots of config records. */
2999 n_config
= ddf
->max_part
;
3000 conf_size
= ddf
->conf_rec_len
* 512;
3002 buf_size
= conf_size
* (n_config
+ 1);
3004 if (posix_memalign((void**)&conf
, 512, buf_size
) != 0)
3008 for (i
= 0 ; i
<= n_config
; i
++) {
3010 struct vd_config
*vdc
= NULL
;
3011 if (i
== n_config
) {
3012 c
= (struct vcl
*)d
->spare
;
3019 get_pd_index_from_refnum(
3022 (const struct vd_config
**)&vdc
,
3026 dprintf("writing conf record %i on disk %08x for %s/%u\n",
3027 i
, be32_to_cpu(d
->disk
.refnum
),
3028 guid_str(vdc
->guid
),
3029 vdc
->sec_elmnt_seq
);
3030 vdc
->seqnum
= header
->seq
;
3031 vdc
->crc
= calc_crc(vdc
, conf_size
);
3032 memcpy(conf
+ i
*conf_size
, vdc
, conf_size
);
3034 memset(conf
+ i
*conf_size
, 0xff, conf_size
);
3036 if (write(fd
, conf
, buf_size
) != buf_size
)
3039 d
->disk
.crc
= calc_crc(&d
->disk
, 512);
3040 if (write(fd
, &d
->disk
, 512) < 0)
3045 header
->openflag
= 0;
3046 header
->crc
= calc_crc(header
, 512);
3048 lseek64(fd
, sector
<<9, 0);
3049 if (write(fd
, header
, 512) < 0)
3055 static int _write_super_to_disk(struct ddf_super
*ddf
, struct dl
*d
)
3057 unsigned long long size
;
3062 /* We need to fill in the primary, (secondary) and workspace
3063 * lba's in the headers, set their checksums,
3064 * Also checksum phys, virt....
3066 * Then write everything out, finally the anchor is written.
3068 get_dev_size(fd
, NULL
, &size
);
3070 if (be64_to_cpu(d
->workspace_lba
) != 0ULL)
3071 ddf
->anchor
.workspace_lba
= d
->workspace_lba
;
3073 ddf
->anchor
.workspace_lba
=
3074 cpu_to_be64(size
- 32*1024*2);
3075 if (be64_to_cpu(d
->primary_lba
) != 0ULL)
3076 ddf
->anchor
.primary_lba
= d
->primary_lba
;
3078 ddf
->anchor
.primary_lba
=
3079 cpu_to_be64(size
- 16*1024*2);
3080 if (be64_to_cpu(d
->secondary_lba
) != 0ULL)
3081 ddf
->anchor
.secondary_lba
= d
->secondary_lba
;
3083 ddf
->anchor
.secondary_lba
=
3084 cpu_to_be64(size
- 32*1024*2);
3085 ddf
->anchor
.seq
= ddf
->active
->seq
;
3086 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
3087 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
3089 ddf
->anchor
.openflag
= 0xFF; /* 'open' means nothing */
3090 ddf
->anchor
.seq
= cpu_to_be32(0xFFFFFFFF); /* no sequencing in anchor */
3091 ddf
->anchor
.crc
= calc_crc(&ddf
->anchor
, 512);
3093 if (!__write_ddf_structure(d
, ddf
, DDF_HEADER_PRIMARY
))
3096 if (!__write_ddf_structure(d
, ddf
, DDF_HEADER_SECONDARY
))
3099 lseek64(fd
, (size
-1)*512, SEEK_SET
);
3100 if (write(fd
, &ddf
->anchor
, 512) < 0)
3107 static int __write_init_super_ddf(struct supertype
*st
)
3109 struct ddf_super
*ddf
= st
->sb
;
3114 pr_state(ddf
, __func__
);
3116 /* try to write updated metadata,
3117 * if we catch a failure move on to the next disk
3119 for (d
= ddf
->dlist
; d
; d
=d
->next
) {
3121 successes
+= _write_super_to_disk(ddf
, d
);
3124 return attempts
!= successes
;
3127 static int write_init_super_ddf(struct supertype
*st
)
3129 struct ddf_super
*ddf
= st
->sb
;
3130 struct vcl
*currentconf
= ddf
->currentconf
;
3132 /* We are done with currentconf - reset it so st refers to the container */
3133 ddf
->currentconf
= NULL
;
3135 if (st
->update_tail
) {
3136 /* queue the virtual_disk and vd_config as metadata updates */
3137 struct virtual_disk
*vd
;
3138 struct vd_config
*vc
;
3143 /* Must be adding a physical disk to the container */
3144 int len
= (sizeof(struct phys_disk
) +
3145 sizeof(struct phys_disk_entry
));
3147 /* adding a disk to the container. */
3151 append_metadata_update(st
, ddf
->add_list
->mdupdate
, len
);
3152 ddf
->add_list
->mdupdate
= NULL
;
3156 /* Newly created VD */
3158 /* First the virtual disk. We have a slightly fake header */
3159 len
= sizeof(struct virtual_disk
) + sizeof(struct virtual_entry
);
3162 vd
->entries
[0] = ddf
->virt
->entries
[currentconf
->vcnum
];
3163 vd
->populated_vdes
= cpu_to_be16(currentconf
->vcnum
);
3164 append_metadata_update(st
, vd
, len
);
3166 /* Then the vd_config */
3167 len
= ddf
->conf_rec_len
* 512;
3168 tlen
= len
* currentconf
->conf
.sec_elmnt_count
;
3170 memcpy(vc
, ¤tconf
->conf
, len
);
3171 for (i
= 1; i
< currentconf
->conf
.sec_elmnt_count
; i
++)
3172 memcpy((char *)vc
+ i
*len
, currentconf
->other_bvds
[i
-1],
3174 append_metadata_update(st
, vc
, tlen
);
3176 /* FIXME I need to close the fds! */
3181 for (d
= ddf
->dlist
; d
; d
=d
->next
)
3182 while (Kill(d
->devname
, NULL
, 0, -1, 1) == 0);
3183 return __write_init_super_ddf(st
);
3189 static __u64
avail_size_ddf(struct supertype
*st
, __u64 devsize
,
3190 unsigned long long data_offset
)
3192 /* We must reserve the last 32Meg */
3193 if (devsize
<= 32*1024*2)
3195 return devsize
- 32*1024*2;
3200 static int reserve_space(struct supertype
*st
, int raiddisks
,
3201 unsigned long long size
, int chunk
,
3202 unsigned long long *freesize
)
3204 /* Find 'raiddisks' spare extents at least 'size' big (but
3205 * only caring about multiples of 'chunk') and remember
3206 * them. If size==0, find the largest size possible.
3207 * Report available size in *freesize
3208 * If space cannot be found, fail.
3211 struct ddf_super
*ddf
= st
->sb
;
3214 for (dl
= ddf
->dlist
; dl
; dl
=dl
->next
) {
3218 /* Now find largest extent on each device */
3219 for (dl
= ddf
->dlist
; dl
; dl
=dl
->next
) {
3220 struct extent
*e
= get_extents(ddf
, dl
);
3221 unsigned long long pos
= 0;
3224 unsigned long long minsize
= size
;
3232 unsigned long long esize
;
3233 esize
= e
[i
].start
- pos
;
3234 if (esize
>= minsize
) {
3238 pos
= e
[i
].start
+ e
[i
].size
;
3240 } while (e
[i
-1].size
);
3243 dl
->esize
= minsize
;
3247 if (cnt
< raiddisks
) {
3248 pr_err("not enough devices with space to create array.\n");
3249 return 0; /* No enough free spaces large enough */
3252 /* choose the largest size of which there are at least 'raiddisk' */
3253 for (dl
= ddf
->dlist
; dl
; dl
=dl
->next
) {
3255 if (dl
->esize
<= size
)
3257 /* This is bigger than 'size', see if there are enough */
3259 for (dl2
= ddf
->dlist
; dl2
; dl2
=dl2
->next
)
3260 if (dl2
->esize
>= dl
->esize
)
3262 if (cnt
>= raiddisks
)
3266 size
= size
/ chunk
;
3271 pr_err("not enough spare devices to create array.\n");
3275 /* We have a 'size' of which there are enough spaces.
3276 * We simply do a first-fit */
3278 for (dl
= ddf
->dlist
; dl
&& cnt
< raiddisks
; dl
=dl
->next
) {
3279 if (dl
->esize
< size
)
3288 static int validate_geometry_ddf(struct supertype
*st
,
3289 int level
, int layout
, int raiddisks
,
3290 int *chunk
, unsigned long long size
,
3291 unsigned long long data_offset
,
3292 char *dev
, unsigned long long *freesize
,
3299 /* ddf potentially supports lots of things, but it depends on
3300 * what devices are offered (and maybe kernel version?)
3301 * If given unused devices, we will make a container.
3302 * If given devices in a container, we will make a BVD.
3303 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
3306 if (*chunk
== UnSet
)
3307 *chunk
= DEFAULT_CHUNK
;
3309 if (level
== LEVEL_NONE
)
3310 level
= LEVEL_CONTAINER
;
3311 if (level
== LEVEL_CONTAINER
) {
3312 /* Must be a fresh device to add to a container */
3313 return validate_geometry_ddf_container(st
, level
, layout
,
3315 size
, data_offset
, dev
,
3321 mdu_array_info_t array
= {
3324 .raid_disks
= raiddisks
3326 struct vd_config conf
;
3327 if (layout_md2ddf(&array
, &conf
) == -1) {
3329 pr_err("DDF does not support level %d /layout %d arrays with %d disks\n",
3330 level
, layout
, raiddisks
);
3333 /* Should check layout? etc */
3335 if (st
->sb
&& freesize
) {
3336 /* --create was given a container to create in.
3337 * So we need to check that there are enough
3338 * free spaces and return the amount of space.
3339 * We may as well remember which drives were
3340 * chosen so that add_to_super/getinfo_super
3343 return reserve_space(st
, raiddisks
, size
, *chunk
, freesize
);
3349 /* A container has already been opened, so we are
3350 * creating in there. Maybe a BVD, maybe an SVD.
3351 * Should make a distinction one day.
3353 return validate_geometry_ddf_bvd(st
, level
, layout
, raiddisks
,
3354 chunk
, size
, data_offset
, dev
,
3358 /* This is the first device for the array.
3359 * If it is a container, we read it in and do automagic allocations,
3360 * no other devices should be given.
3361 * Otherwise it must be a member device of a container, and we
3362 * do manual allocation.
3363 * Later we should check for a BVD and make an SVD.
3365 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
3367 sra
= sysfs_read(fd
, NULL
, GET_VERSION
);
3369 if (sra
&& sra
->array
.major_version
== -1 &&
3370 strcmp(sra
->text_version
, "ddf") == 0) {
3372 /* find space for 'n' devices. */
3373 /* remember the devices */
3374 /* Somehow return the fact that we have enough */
3378 pr_err("ddf: Cannot create this array "
3379 "on device %s - a container is required.\n",
3383 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
3385 pr_err("ddf: Cannot open %s: %s\n",
3386 dev
, strerror(errno
));
3389 /* Well, it is in use by someone, maybe a 'ddf' container. */
3390 cfd
= open_container(fd
);
3394 pr_err("ddf: Cannot use %s: %s\n",
3395 dev
, strerror(EBUSY
));
3398 sra
= sysfs_read(cfd
, NULL
, GET_VERSION
);
3400 if (sra
&& sra
->array
.major_version
== -1 &&
3401 strcmp(sra
->text_version
, "ddf") == 0) {
3402 /* This is a member of a ddf container. Load the container
3403 * and try to create a bvd
3405 struct ddf_super
*ddf
;
3406 if (load_super_ddf_all(st
, cfd
, (void **)&ddf
, NULL
) == 0) {
3408 strcpy(st
->container_devnm
, fd2devnm(cfd
));
3410 return validate_geometry_ddf_bvd(st
, level
, layout
,
3411 raiddisks
, chunk
, size
,
3417 } else /* device may belong to a different container */
3424 validate_geometry_ddf_container(struct supertype
*st
,
3425 int level
, int layout
, int raiddisks
,
3426 int chunk
, unsigned long long size
,
3427 unsigned long long data_offset
,
3428 char *dev
, unsigned long long *freesize
,
3432 unsigned long long ldsize
;
3434 if (level
!= LEVEL_CONTAINER
)
3439 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
3442 pr_err("ddf: Cannot open %s: %s\n",
3443 dev
, strerror(errno
));
3446 if (!get_dev_size(fd
, dev
, &ldsize
)) {
3452 *freesize
= avail_size_ddf(st
, ldsize
>> 9, INVALID_SECTORS
);
3459 static int validate_geometry_ddf_bvd(struct supertype
*st
,
3460 int level
, int layout
, int raiddisks
,
3461 int *chunk
, unsigned long long size
,
3462 unsigned long long data_offset
,
3463 char *dev
, unsigned long long *freesize
,
3467 struct ddf_super
*ddf
= st
->sb
;
3469 unsigned long long pos
= 0;
3470 unsigned long long maxsize
;
3473 /* ddf/bvd supports lots of things, but not containers */
3474 if (level
== LEVEL_CONTAINER
) {
3476 pr_err("DDF cannot create a container within an container\n");
3479 /* We must have the container info already read in. */
3484 /* General test: make sure there is space for
3485 * 'raiddisks' device extents of size 'size'.
3487 unsigned long long minsize
= size
;
3491 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
3496 e
= get_extents(ddf
, dl
);
3499 unsigned long long esize
;
3500 esize
= e
[i
].start
- pos
;
3501 if (esize
>= minsize
)
3503 pos
= e
[i
].start
+ e
[i
].size
;
3505 } while (e
[i
-1].size
);
3510 if (dcnt
< raiddisks
) {
3512 pr_err("ddf: Not enough devices with "
3513 "space for this array (%d < %d)\n",
3519 /* This device must be a member of the set */
3520 if (stat(dev
, &stb
) < 0)
3522 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
3524 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
3525 if (dl
->major
== (int)major(stb
.st_rdev
) &&
3526 dl
->minor
== (int)minor(stb
.st_rdev
))
3531 pr_err("ddf: %s is not in the "
3536 e
= get_extents(ddf
, dl
);
3541 unsigned long long esize
;
3542 esize
= e
[i
].start
- pos
;
3543 if (esize
>= maxsize
)
3545 pos
= e
[i
].start
+ e
[i
].size
;
3547 } while (e
[i
-1].size
);
3548 *freesize
= maxsize
;
3554 static int load_super_ddf_all(struct supertype
*st
, int fd
,
3555 void **sbp
, char *devname
)
3558 struct ddf_super
*super
;
3559 struct mdinfo
*sd
, *best
= NULL
;
3565 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
3568 if (sra
->array
.major_version
!= -1 ||
3569 sra
->array
.minor_version
!= -2 ||
3570 strcmp(sra
->text_version
, "ddf") != 0)
3573 if (posix_memalign((void**)&super
, 512, sizeof(*super
)) != 0)
3575 memset(super
, 0, sizeof(*super
));
3577 /* first, try each device, and choose the best ddf */
3578 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
3580 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3581 dfd
= dev_open(nm
, O_RDONLY
);
3584 rv
= load_ddf_headers(dfd
, super
, NULL
);
3587 seq
= be32_to_cpu(super
->active
->seq
);
3588 if (super
->active
->openflag
)
3590 if (!best
|| seq
> bestseq
) {
3598 /* OK, load this ddf */
3599 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
3600 dfd
= dev_open(nm
, O_RDONLY
);
3603 load_ddf_headers(dfd
, super
, NULL
);
3604 load_ddf_global(dfd
, super
, NULL
);
3606 /* Now we need the device-local bits */
3607 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
3610 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3611 dfd
= dev_open(nm
, O_RDWR
);
3614 rv
= load_ddf_headers(dfd
, super
, NULL
);
3616 rv
= load_ddf_local(dfd
, super
, NULL
, 1);
3622 if (st
->ss
== NULL
) {
3623 st
->ss
= &super_ddf
;
3624 st
->minor_version
= 0;
3627 strcpy(st
->container_devnm
, fd2devnm(fd
));
3631 static int load_container_ddf(struct supertype
*st
, int fd
,
3634 return load_super_ddf_all(st
, fd
, &st
->sb
, devname
);
3637 #endif /* MDASSEMBLE */
3639 static int check_secondary(const struct vcl
*vc
)
3641 const struct vd_config
*conf
= &vc
->conf
;
3644 /* The only DDF secondary RAID level md can support is
3645 * RAID 10, if the stripe sizes and Basic volume sizes
3647 * Other configurations could in theory be supported by exposing
3648 * the BVDs to user space and using device mapper for the secondary
3649 * mapping. So far we don't support that.
3652 __u64 sec_elements
[4] = {0, 0, 0, 0};
3653 #define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63)))
3654 #define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0)
3656 if (vc
->other_bvds
== NULL
) {
3657 pr_err("No BVDs for secondary RAID found\n");
3660 if (conf
->prl
!= DDF_RAID1
) {
3661 pr_err("Secondary RAID level only supported for mirrored BVD\n");
3664 if (conf
->srl
!= DDF_2STRIPED
&& conf
->srl
!= DDF_2SPANNED
) {
3665 pr_err("Secondary RAID level %d is unsupported\n",
3669 __set_sec_seen(conf
->sec_elmnt_seq
);
3670 for (i
= 0; i
< conf
->sec_elmnt_count
-1; i
++) {
3671 const struct vd_config
*bvd
= vc
->other_bvds
[i
];
3672 if (bvd
->sec_elmnt_seq
== DDF_UNUSED_BVD
)
3674 if (bvd
->srl
!= conf
->srl
) {
3675 pr_err("Inconsistent secondary RAID level across BVDs\n");
3678 if (bvd
->prl
!= conf
->prl
) {
3679 pr_err("Different RAID levels for BVDs are unsupported\n");
3682 if (!be16_eq(bvd
->prim_elmnt_count
, conf
->prim_elmnt_count
)) {
3683 pr_err("All BVDs must have the same number of primary elements\n");
3686 if (bvd
->chunk_shift
!= conf
->chunk_shift
) {
3687 pr_err("Different strip sizes for BVDs are unsupported\n");
3690 if (!be64_eq(bvd
->array_blocks
, conf
->array_blocks
)) {
3691 pr_err("Different BVD sizes are unsupported\n");
3694 __set_sec_seen(bvd
->sec_elmnt_seq
);
3696 for (i
= 0; i
< conf
->sec_elmnt_count
; i
++) {
3697 if (!__was_sec_seen(i
)) {
3698 pr_err("BVD %d is missing\n", i
);
3705 static unsigned int get_pd_index_from_refnum(const struct vcl
*vc
,
3706 be32 refnum
, unsigned int nmax
,
3707 const struct vd_config
**bvd
,
3710 unsigned int i
, j
, n
, sec
, cnt
;
3712 cnt
= be16_to_cpu(vc
->conf
.prim_elmnt_count
);
3713 sec
= (vc
->conf
.sec_elmnt_count
== 1 ? 0 : vc
->conf
.sec_elmnt_seq
);
3715 for (i
= 0, j
= 0 ; i
< nmax
; i
++) {
3716 /* j counts valid entries for this BVD */
3717 if (be32_eq(vc
->conf
.phys_refnum
[i
], refnum
)) {
3720 return sec
* cnt
+ j
;
3722 if (be32_to_cpu(vc
->conf
.phys_refnum
[i
]) != 0xffffffff)
3725 if (vc
->other_bvds
== NULL
)
3728 for (n
= 1; n
< vc
->conf
.sec_elmnt_count
; n
++) {
3729 struct vd_config
*vd
= vc
->other_bvds
[n
-1];
3730 sec
= vd
->sec_elmnt_seq
;
3731 if (sec
== DDF_UNUSED_BVD
)
3733 for (i
= 0, j
= 0 ; i
< nmax
; i
++) {
3734 if (be32_eq(vd
->phys_refnum
[i
], refnum
)) {
3737 return sec
* cnt
+ j
;
3739 if (be32_to_cpu(vd
->phys_refnum
[i
]) != 0xffffffff)
3745 return DDF_NOTFOUND
;
3748 static struct mdinfo
*container_content_ddf(struct supertype
*st
, char *subarray
)
3750 /* Given a container loaded by load_super_ddf_all,
3751 * extract information about all the arrays into
3754 * For each vcl in conflist: create an mdinfo, fill it in,
3755 * then look for matching devices (phys_refnum) in dlist
3756 * and create appropriate device mdinfo.
3758 struct ddf_super
*ddf
= st
->sb
;
3759 struct mdinfo
*rest
= NULL
;
3762 for (vc
= ddf
->conflist
; vc
; vc
=vc
->next
) {
3764 struct mdinfo
*this;
3770 (strtoul(subarray
, &ep
, 10) != vc
->vcnum
||
3774 if (vc
->conf
.sec_elmnt_count
> 1) {
3775 if (check_secondary(vc
) != 0)
3779 this = xcalloc(1, sizeof(*this));
3783 if (layout_ddf2md(&vc
->conf
, &this->array
))
3785 this->array
.md_minor
= -1;
3786 this->array
.major_version
= -1;
3787 this->array
.minor_version
= -2;
3788 this->safe_mode_delay
= DDF_SAFE_MODE_DELAY
;
3789 cptr
= (__u32
*)(vc
->conf
.guid
+ 16);
3790 this->array
.ctime
= DECADE
+ __be32_to_cpu(*cptr
);
3791 this->array
.utime
= DECADE
+
3792 be32_to_cpu(vc
->conf
.timestamp
);
3793 this->array
.chunk_size
= 512 << vc
->conf
.chunk_shift
;
3796 if ((ddf
->virt
->entries
[i
].state
& DDF_state_inconsistent
) ||
3797 (ddf
->virt
->entries
[i
].init_state
& DDF_initstate_mask
) !=
3799 this->array
.state
= 0;
3800 this->resync_start
= 0;
3802 this->array
.state
= 1;
3803 this->resync_start
= MaxSector
;
3805 _ddf_array_name(this->name
, ddf
, i
);
3806 memset(this->uuid
, 0, sizeof(this->uuid
));
3807 this->component_size
= be64_to_cpu(vc
->conf
.blocks
);
3808 this->array
.size
= this->component_size
/ 2;
3809 this->container_member
= i
;
3811 ddf
->currentconf
= vc
;
3812 uuid_from_super_ddf(st
, this->uuid
);
3814 ddf
->currentconf
= NULL
;
3816 sprintf(this->text_version
, "/%s/%d",
3817 st
->container_devnm
, this->container_member
);
3819 for (pd
= 0; pd
< be16_to_cpu(ddf
->phys
->used_pdes
); pd
++) {
3822 const struct vd_config
*bvd
;
3826 if (be32_to_cpu(ddf
->phys
->entries
[pd
].refnum
)
3830 stt
= be16_to_cpu(ddf
->phys
->entries
[pd
].state
);
3831 if ((stt
& (DDF_Online
|DDF_Failed
|DDF_Rebuilding
))
3835 i
= get_pd_index_from_refnum(
3836 vc
, ddf
->phys
->entries
[pd
].refnum
,
3837 ddf
->mppe
, &bvd
, &iphys
);
3838 if (i
== DDF_NOTFOUND
)
3841 this->array
.working_disks
++;
3843 for (d
= ddf
->dlist
; d
; d
=d
->next
)
3844 if (be32_eq(d
->disk
.refnum
,
3845 ddf
->phys
->entries
[pd
].refnum
))
3848 /* Haven't found that one yet, maybe there are others */
3851 dev
= xcalloc(1, sizeof(*dev
));
3852 dev
->next
= this->devs
;
3855 dev
->disk
.number
= be32_to_cpu(d
->disk
.refnum
);
3856 dev
->disk
.major
= d
->major
;
3857 dev
->disk
.minor
= d
->minor
;
3858 dev
->disk
.raid_disk
= i
;
3859 dev
->disk
.state
= (1<<MD_DISK_SYNC
)|(1<<MD_DISK_ACTIVE
);
3860 dev
->recovery_start
= MaxSector
;
3862 dev
->events
= be32_to_cpu(ddf
->active
->seq
);
3864 be64_to_cpu(LBA_OFFSET(ddf
, bvd
)[iphys
]);
3865 dev
->component_size
= be64_to_cpu(bvd
->blocks
);
3867 strcpy(dev
->name
, d
->devname
);
3873 static int store_super_ddf(struct supertype
*st
, int fd
)
3875 struct ddf_super
*ddf
= st
->sb
;
3876 unsigned long long dsize
;
3883 if (!get_dev_size(fd
, NULL
, &dsize
))
3886 if (ddf
->dlist
|| ddf
->conflist
) {
3891 if (fstat(fd
, &sta
) == -1 || !S_ISBLK(sta
.st_mode
)) {
3892 pr_err("%s: file descriptor for invalid device\n",
3896 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
3897 if (dl
->major
== (int)major(sta
.st_rdev
) &&
3898 dl
->minor
== (int)minor(sta
.st_rdev
))
3901 pr_err("%s: couldn't find disk %d/%d\n", __func__
,
3902 (int)major(sta
.st_rdev
),
3903 (int)minor(sta
.st_rdev
));
3908 ret
= (_write_super_to_disk(ddf
, dl
) != 1);
3913 if (posix_memalign(&buf
, 512, 512) != 0)
3915 memset(buf
, 0, 512);
3917 lseek64(fd
, dsize
-512, 0);
3918 rc
= write(fd
, buf
, 512);
3925 static int compare_super_ddf(struct supertype
*st
, struct supertype
*tst
)
3929 * 0 same, or first was empty, and second was copied
3930 * 1 second had wrong magic number - but that isn't possible
3932 * 3 wrong other info
3934 struct ddf_super
*first
= st
->sb
;
3935 struct ddf_super
*second
= tst
->sb
;
3936 struct dl
*dl1
, *dl2
;
3937 struct vcl
*vl1
, *vl2
;
3938 unsigned int max_vds
, max_pds
, pd
, vd
;
3946 if (memcmp(first
->anchor
.guid
, second
->anchor
.guid
, DDF_GUID_LEN
) != 0)
3949 if (first
->max_part
!= second
->max_part
||
3950 !be16_eq(first
->phys
->used_pdes
, second
->phys
->used_pdes
) ||
3951 !be16_eq(first
->virt
->populated_vdes
,
3952 second
->virt
->populated_vdes
)) {
3953 dprintf("%s: PD/VD number mismatch\n", __func__
);
3957 max_pds
= be16_to_cpu(first
->phys
->used_pdes
);
3958 for (dl2
= second
->dlist
; dl2
; dl2
= dl2
->next
) {
3959 for (pd
= 0; pd
< max_pds
; pd
++)
3960 if (be32_eq(first
->phys
->entries
[pd
].refnum
,
3963 if (pd
== max_pds
) {
3964 dprintf("%s: no match for disk %08x\n", __func__
,
3965 be32_to_cpu(dl2
->disk
.refnum
));
3970 max_vds
= be16_to_cpu(first
->active
->max_vd_entries
);
3971 for (vl2
= second
->conflist
; vl2
; vl2
= vl2
->next
) {
3972 if (!be32_eq(vl2
->conf
.magic
, DDF_VD_CONF_MAGIC
))
3974 for (vd
= 0; vd
< max_vds
; vd
++)
3975 if (!memcmp(first
->virt
->entries
[vd
].guid
,
3976 vl2
->conf
.guid
, DDF_GUID_LEN
))
3978 if (vd
== max_vds
) {
3979 dprintf("%s: no match for VD config\n", __func__
);
3983 /* FIXME should I look at anything else? */
3986 * At this point we are fairly sure that the meta data matches.
3987 * But the new disk may contain additional local data.
3988 * Add it to the super block.
3990 for (vl2
= second
->conflist
; vl2
; vl2
= vl2
->next
) {
3991 for (vl1
= first
->conflist
; vl1
; vl1
= vl1
->next
)
3992 if (!memcmp(vl1
->conf
.guid
, vl2
->conf
.guid
,
3996 if (vl1
->other_bvds
!= NULL
&&
3997 vl1
->conf
.sec_elmnt_seq
!=
3998 vl2
->conf
.sec_elmnt_seq
) {
3999 dprintf("%s: adding BVD %u\n", __func__
,
4000 vl2
->conf
.sec_elmnt_seq
);
4001 add_other_bvd(vl1
, &vl2
->conf
,
4002 first
->conf_rec_len
*512);
4007 if (posix_memalign((void **)&vl1
, 512,
4008 (first
->conf_rec_len
*512 +
4009 offsetof(struct vcl
, conf
))) != 0) {
4010 pr_err("%s could not allocate vcl buf\n",
4015 vl1
->next
= first
->conflist
;
4016 vl1
->block_sizes
= NULL
;
4017 memcpy(&vl1
->conf
, &vl2
->conf
, first
->conf_rec_len
*512);
4018 if (alloc_other_bvds(first
, vl1
) != 0) {
4019 pr_err("%s could not allocate other bvds\n",
4024 for (vd
= 0; vd
< max_vds
; vd
++)
4025 if (!memcmp(first
->virt
->entries
[vd
].guid
,
4026 vl1
->conf
.guid
, DDF_GUID_LEN
))
4029 dprintf("%s: added config for VD %u\n", __func__
, vl1
->vcnum
);
4030 first
->conflist
= vl1
;
4033 for (dl2
= second
->dlist
; dl2
; dl2
= dl2
->next
) {
4034 for (dl1
= first
->dlist
; dl1
; dl1
= dl1
->next
)
4035 if (be32_eq(dl1
->disk
.refnum
, dl2
->disk
.refnum
))
4040 if (posix_memalign((void **)&dl1
, 512,
4041 sizeof(*dl1
) + (first
->max_part
) * sizeof(dl1
->vlist
[0]))
4043 pr_err("%s could not allocate disk info buffer\n",
4047 memcpy(dl1
, dl2
, sizeof(*dl1
));
4048 dl1
->mdupdate
= NULL
;
4049 dl1
->next
= first
->dlist
;
4051 for (pd
= 0; pd
< max_pds
; pd
++)
4052 if (be32_eq(first
->phys
->entries
[pd
].refnum
,
4057 if (posix_memalign((void **)&dl1
->spare
, 512,
4058 first
->conf_rec_len
*512) != 0) {
4059 pr_err("%s could not allocate spare info buf\n",
4063 memcpy(dl1
->spare
, dl2
->spare
, first
->conf_rec_len
*512);
4065 for (vd
= 0 ; vd
< first
->max_part
; vd
++) {
4066 if (!dl2
->vlist
[vd
]) {
4067 dl1
->vlist
[vd
] = NULL
;
4070 for (vl1
= first
->conflist
; vl1
; vl1
= vl1
->next
) {
4071 if (!memcmp(vl1
->conf
.guid
,
4072 dl2
->vlist
[vd
]->conf
.guid
,
4075 dl1
->vlist
[vd
] = vl1
;
4079 dprintf("%s: added disk %d: %08x\n", __func__
, dl1
->pdnum
,
4080 be32_to_cpu(dl1
->disk
.refnum
));
4088 * A new array 'a' has been started which claims to be instance 'inst'
4089 * within container 'c'.
4090 * We need to confirm that the array matches the metadata in 'c' so
4091 * that we don't corrupt any metadata.
4093 static int ddf_open_new(struct supertype
*c
, struct active_array
*a
, char *inst
)
4095 struct ddf_super
*ddf
= c
->sb
;
4099 static const char faulty
[] = "faulty";
4101 if (all_ff(ddf
->virt
->entries
[n
].guid
)) {
4102 pr_err("%s: subarray %d doesn't exist\n", __func__
, n
);
4105 dprintf("%s: new subarray %d, GUID: %s\n", __func__
, n
,
4106 guid_str(ddf
->virt
->entries
[n
].guid
));
4107 for (dev
= a
->info
.devs
; dev
; dev
= dev
->next
) {
4108 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4109 if (dl
->major
== dev
->disk
.major
&&
4110 dl
->minor
== dev
->disk
.minor
)
4112 if (!dl
|| dl
->pdnum
< 0) {
4113 pr_err("%s: device %d/%d of subarray %d not found in meta data\n",
4114 __func__
, dev
->disk
.major
, dev
->disk
.minor
, n
);
4117 if ((be16_to_cpu(ddf
->phys
->entries
[dl
->pdnum
].state
) &
4118 (DDF_Online
|DDF_Missing
|DDF_Failed
)) != DDF_Online
) {
4119 pr_err("%s: new subarray %d contains broken device %d/%d (%02x)\n",
4120 __func__
, n
, dl
->major
, dl
->minor
,
4122 ddf
->phys
->entries
[dl
->pdnum
].state
));
4123 if (write(dev
->state_fd
, faulty
, sizeof(faulty
)-1) !=
4125 pr_err("Write to state_fd failed\n");
4126 dev
->curr_state
= DS_FAULTY
;
4129 a
->info
.container_member
= n
;
4133 static void handle_missing(struct ddf_super
*ddf
, struct active_array
*a
, int inst
)
4135 /* This member array is being activated. If any devices
4136 * are missing they must now be marked as failed.
4138 struct vd_config
*vc
;
4146 for (n
= 0; ; n
++) {
4147 vc
= find_vdcr(ddf
, inst
, n
, &n_bvd
, &vcl
);
4150 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4151 if (be32_eq(dl
->disk
.refnum
, vc
->phys_refnum
[n_bvd
]))
4154 /* Found this disk, so not missing */
4157 /* Mark the device as failed/missing. */
4158 pd
= find_phys(ddf
, vc
->phys_refnum
[n_bvd
]);
4159 if (pd
>= 0 && be16_and(ddf
->phys
->entries
[pd
].state
,
4160 cpu_to_be16(DDF_Online
))) {
4161 be16_clear(ddf
->phys
->entries
[pd
].state
,
4162 cpu_to_be16(DDF_Online
));
4163 be16_set(ddf
->phys
->entries
[pd
].state
,
4164 cpu_to_be16(DDF_Failed
|DDF_Missing
));
4165 vc
->phys_refnum
[n_bvd
] = cpu_to_be32(0);
4166 ddf_set_updates_pending(ddf
);
4169 /* Mark the array as Degraded */
4170 state
= get_svd_state(ddf
, vcl
);
4171 if (ddf
->virt
->entries
[inst
].state
!=
4172 ((ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4174 ddf
->virt
->entries
[inst
].state
=
4175 (ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4177 a
->check_degraded
= 1;
4178 ddf_set_updates_pending(ddf
);
4184 * The array 'a' is to be marked clean in the metadata.
4185 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4186 * clean up to the point (in sectors). If that cannot be recorded in the
4187 * metadata, then leave it as dirty.
4189 * For DDF, we need to clear the DDF_state_inconsistent bit in the
4190 * !global! virtual_disk.virtual_entry structure.
4192 static int ddf_set_array_state(struct active_array
*a
, int consistent
)
4194 struct ddf_super
*ddf
= a
->container
->sb
;
4195 int inst
= a
->info
.container_member
;
4196 int old
= ddf
->virt
->entries
[inst
].state
;
4197 if (consistent
== 2) {
4198 handle_missing(ddf
, a
, inst
);
4199 /* Should check if a recovery should be started FIXME */
4201 if (!is_resync_complete(&a
->info
))
4205 ddf
->virt
->entries
[inst
].state
&= ~DDF_state_inconsistent
;
4207 ddf
->virt
->entries
[inst
].state
|= DDF_state_inconsistent
;
4208 if (old
!= ddf
->virt
->entries
[inst
].state
)
4209 ddf_set_updates_pending(ddf
);
4211 old
= ddf
->virt
->entries
[inst
].init_state
;
4212 ddf
->virt
->entries
[inst
].init_state
&= ~DDF_initstate_mask
;
4213 if (is_resync_complete(&a
->info
))
4214 ddf
->virt
->entries
[inst
].init_state
|= DDF_init_full
;
4215 else if (a
->info
.resync_start
== 0)
4216 ddf
->virt
->entries
[inst
].init_state
|= DDF_init_not
;
4218 ddf
->virt
->entries
[inst
].init_state
|= DDF_init_quick
;
4219 if (old
!= ddf
->virt
->entries
[inst
].init_state
)
4220 ddf_set_updates_pending(ddf
);
4222 dprintf("ddf mark %d/%s (%d) %s %llu\n", inst
,
4223 guid_str(ddf
->virt
->entries
[inst
].guid
), a
->curr_state
,
4224 consistent
?"clean":"dirty",
4225 a
->info
.resync_start
);
4229 static int get_bvd_state(const struct ddf_super
*ddf
,
4230 const struct vd_config
*vc
)
4232 unsigned int i
, n_bvd
, working
= 0;
4233 unsigned int n_prim
= be16_to_cpu(vc
->prim_elmnt_count
);
4235 for (i
= 0; i
< n_prim
; i
++) {
4236 if (!find_index_in_bvd(ddf
, vc
, i
, &n_bvd
))
4238 pd
= find_phys(ddf
, vc
->phys_refnum
[n_bvd
]);
4241 st
= be16_to_cpu(ddf
->phys
->entries
[pd
].state
);
4242 if ((st
& (DDF_Online
|DDF_Failed
|DDF_Rebuilding
))
4247 state
= DDF_state_degraded
;
4248 if (working
== n_prim
)
4249 state
= DDF_state_optimal
;
4255 state
= DDF_state_failed
;
4259 state
= DDF_state_failed
;
4260 else if (working
>= 2)
4261 state
= DDF_state_part_optimal
;
4265 if (working
< n_prim
- 1)
4266 state
= DDF_state_failed
;
4269 if (working
< n_prim
- 2)
4270 state
= DDF_state_failed
;
4271 else if (working
== n_prim
- 1)
4272 state
= DDF_state_part_optimal
;
4278 static int secondary_state(int state
, int other
, int seclevel
)
4280 if (state
== DDF_state_optimal
&& other
== DDF_state_optimal
)
4281 return DDF_state_optimal
;
4282 if (seclevel
== DDF_2MIRRORED
) {
4283 if (state
== DDF_state_optimal
|| other
== DDF_state_optimal
)
4284 return DDF_state_part_optimal
;
4285 if (state
== DDF_state_failed
&& other
== DDF_state_failed
)
4286 return DDF_state_failed
;
4287 return DDF_state_degraded
;
4289 if (state
== DDF_state_failed
|| other
== DDF_state_failed
)
4290 return DDF_state_failed
;
4291 if (state
== DDF_state_degraded
|| other
== DDF_state_degraded
)
4292 return DDF_state_degraded
;
4293 return DDF_state_part_optimal
;
4297 static int get_svd_state(const struct ddf_super
*ddf
, const struct vcl
*vcl
)
4299 int state
= get_bvd_state(ddf
, &vcl
->conf
);
4301 for (i
= 1; i
< vcl
->conf
.sec_elmnt_count
; i
++) {
4302 state
= secondary_state(
4304 get_bvd_state(ddf
, vcl
->other_bvds
[i
-1]),
4311 * The state of each disk is stored in the global phys_disk structure
4312 * in phys_disk.entries[n].state.
4313 * This makes various combinations awkward.
4314 * - When a device fails in any array, it must be failed in all arrays
4315 * that include a part of this device.
4316 * - When a component is rebuilding, we cannot include it officially in the
4317 * array unless this is the only array that uses the device.
4319 * So: when transitioning:
4320 * Online -> failed, just set failed flag. monitor will propagate
4321 * spare -> online, the device might need to be added to the array.
4322 * spare -> failed, just set failed. Don't worry if in array or not.
4324 static void ddf_set_disk(struct active_array
*a
, int n
, int state
)
4326 struct ddf_super
*ddf
= a
->container
->sb
;
4327 unsigned int inst
= a
->info
.container_member
, n_bvd
;
4329 struct vd_config
*vc
= find_vdcr(ddf
, inst
, (unsigned int)n
,
4335 dprintf("%s: %d to %x\n", __func__
, n
, state
);
4337 dprintf("ddf: cannot find instance %d!!\n", inst
);
4340 /* Find the matching slot in 'info'. */
4341 for (mdi
= a
->info
.devs
; mdi
; mdi
= mdi
->next
)
4342 if (mdi
->disk
.raid_disk
== n
)
4345 pr_err("%s: cannot find raid disk %d\n",
4350 /* and find the 'dl' entry corresponding to that. */
4351 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4352 if (mdi
->state_fd
>= 0 &&
4353 mdi
->disk
.major
== dl
->major
&&
4354 mdi
->disk
.minor
== dl
->minor
)
4357 pr_err("%s: cannot find raid disk %d (%d/%d)\n",
4359 mdi
->disk
.major
, mdi
->disk
.minor
);
4363 pd
= find_phys(ddf
, vc
->phys_refnum
[n_bvd
]);
4364 if (pd
< 0 || pd
!= dl
->pdnum
) {
4365 /* disk doesn't currently exist or has changed.
4366 * If it is now in_sync, insert it. */
4367 dprintf("%s: phys disk not found for %d: %d/%d ref %08x\n",
4368 __func__
, dl
->pdnum
, dl
->major
, dl
->minor
,
4369 be32_to_cpu(dl
->disk
.refnum
));
4370 dprintf("%s: array %u disk %u ref %08x pd %d\n",
4371 __func__
, inst
, n_bvd
,
4372 be32_to_cpu(vc
->phys_refnum
[n_bvd
]), pd
);
4373 if ((state
& DS_INSYNC
) && ! (state
& DS_FAULTY
)) {
4374 pd
= dl
->pdnum
; /* FIXME: is this really correct ? */
4375 vc
->phys_refnum
[n_bvd
] = dl
->disk
.refnum
;
4376 LBA_OFFSET(ddf
, vc
)[n_bvd
] =
4377 cpu_to_be64(mdi
->data_offset
);
4378 be16_clear(ddf
->phys
->entries
[pd
].type
,
4379 cpu_to_be16(DDF_Global_Spare
));
4380 be16_set(ddf
->phys
->entries
[pd
].type
,
4381 cpu_to_be16(DDF_Active_in_VD
));
4382 ddf_set_updates_pending(ddf
);
4385 be16 old
= ddf
->phys
->entries
[pd
].state
;
4386 if (state
& DS_FAULTY
)
4387 be16_set(ddf
->phys
->entries
[pd
].state
,
4388 cpu_to_be16(DDF_Failed
));
4389 if (state
& DS_INSYNC
) {
4390 be16_set(ddf
->phys
->entries
[pd
].state
,
4391 cpu_to_be16(DDF_Online
));
4392 be16_clear(ddf
->phys
->entries
[pd
].state
,
4393 cpu_to_be16(DDF_Rebuilding
));
4395 if (!be16_eq(old
, ddf
->phys
->entries
[pd
].state
))
4396 ddf_set_updates_pending(ddf
);
4399 dprintf("ddf: set_disk %d (%08x) to %x->%02x\n", n
,
4400 be32_to_cpu(dl
->disk
.refnum
), state
,
4401 be16_to_cpu(ddf
->phys
->entries
[pd
].state
));
4403 /* Now we need to check the state of the array and update
4404 * virtual_disk.entries[n].state.
4405 * It needs to be one of "optimal", "degraded", "failed".
4406 * I don't understand 'deleted' or 'missing'.
4408 state
= get_svd_state(ddf
, vcl
);
4410 if (ddf
->virt
->entries
[inst
].state
!=
4411 ((ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4413 ddf
->virt
->entries
[inst
].state
=
4414 (ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4416 ddf_set_updates_pending(ddf
);
4421 static void ddf_sync_metadata(struct supertype
*st
)
4424 * Write all data to all devices.
4425 * Later, we might be able to track whether only local changes
4426 * have been made, or whether any global data has been changed,
4427 * but ddf is sufficiently weird that it probably always
4428 * changes global data ....
4430 struct ddf_super
*ddf
= st
->sb
;
4431 if (!ddf
->updates_pending
)
4433 ddf
->updates_pending
= 0;
4434 __write_init_super_ddf(st
);
4435 dprintf("ddf: sync_metadata\n");
4438 static int del_from_conflist(struct vcl
**list
, const char *guid
)
4442 for (p
= list
; p
&& *p
; p
= &((*p
)->next
))
4443 if (!memcmp((*p
)->conf
.guid
, guid
, DDF_GUID_LEN
)) {
4450 static int _kill_subarray_ddf(struct ddf_super
*ddf
, const char *guid
)
4453 unsigned int vdnum
, i
;
4454 vdnum
= find_vde_by_guid(ddf
, guid
);
4455 if (vdnum
== DDF_NOTFOUND
) {
4456 pr_err("%s: could not find VD %s\n", __func__
,
4460 if (del_from_conflist(&ddf
->conflist
, guid
) == 0) {
4461 pr_err("%s: could not find conf %s\n", __func__
,
4465 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4466 for (i
= 0; i
< ddf
->max_part
; i
++)
4467 if (dl
->vlist
[i
] != NULL
&&
4468 !memcmp(dl
->vlist
[i
]->conf
.guid
, guid
,
4470 dl
->vlist
[i
] = NULL
;
4471 memset(ddf
->virt
->entries
[vdnum
].guid
, 0xff, DDF_GUID_LEN
);
4472 dprintf("%s: deleted %s\n", __func__
, guid_str(guid
));
4476 static int kill_subarray_ddf(struct supertype
*st
)
4478 struct ddf_super
*ddf
= st
->sb
;
4480 * currentconf is set in container_content_ddf,
4481 * called with subarray arg
4483 struct vcl
*victim
= ddf
->currentconf
;
4484 struct vd_config
*conf
;
4487 ddf
->currentconf
= NULL
;
4489 pr_err("%s: nothing to kill\n", __func__
);
4492 conf
= &victim
->conf
;
4493 vdnum
= find_vde_by_guid(ddf
, conf
->guid
);
4494 if (vdnum
== DDF_NOTFOUND
) {
4495 pr_err("%s: could not find VD %s\n", __func__
,
4496 guid_str(conf
->guid
));
4499 if (st
->update_tail
) {
4500 struct virtual_disk
*vd
;
4501 int len
= sizeof(struct virtual_disk
)
4502 + sizeof(struct virtual_entry
);
4505 pr_err("%s: failed to allocate %d bytes\n", __func__
,
4509 memset(vd
, 0 , len
);
4510 vd
->magic
= DDF_VIRT_RECORDS_MAGIC
;
4511 vd
->populated_vdes
= cpu_to_be16(0);
4512 memcpy(vd
->entries
[0].guid
, conf
->guid
, DDF_GUID_LEN
);
4513 /* we use DDF_state_deleted as marker */
4514 vd
->entries
[0].state
= DDF_state_deleted
;
4515 append_metadata_update(st
, vd
, len
);
4517 _kill_subarray_ddf(ddf
, conf
->guid
);
4518 ddf_set_updates_pending(ddf
);
4519 ddf_sync_metadata(st
);
4524 static void copy_matching_bvd(struct ddf_super
*ddf
,
4525 struct vd_config
*conf
,
4526 const struct metadata_update
*update
)
4529 be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
4530 unsigned int len
= ddf
->conf_rec_len
* 512;
4532 struct vd_config
*vc
;
4533 for (p
= update
->buf
; p
< update
->buf
+ update
->len
; p
+= len
) {
4534 vc
= (struct vd_config
*) p
;
4535 if (vc
->sec_elmnt_seq
== conf
->sec_elmnt_seq
) {
4536 memcpy(conf
->phys_refnum
, vc
->phys_refnum
,
4537 mppe
* (sizeof(__u32
) + sizeof(__u64
)));
4541 pr_err("%s: no match for BVD %d of %s in update\n", __func__
,
4542 conf
->sec_elmnt_seq
, guid_str(conf
->guid
));
4545 static void ddf_process_update(struct supertype
*st
,
4546 struct metadata_update
*update
)
4548 /* Apply this update to the metadata.
4549 * The first 4 bytes are a DDF_*_MAGIC which guides
4551 * Possible update are:
4552 * DDF_PHYS_RECORDS_MAGIC
4553 * Add a new physical device or remove an old one.
4554 * Changes to this record only happen implicitly.
4555 * used_pdes is the device number.
4556 * DDF_VIRT_RECORDS_MAGIC
4557 * Add a new VD. Possibly also change the 'access' bits.
4558 * populated_vdes is the entry number.
4560 * New or updated VD. the VIRT_RECORD must already
4561 * exist. For an update, phys_refnum and lba_offset
4562 * (at least) are updated, and the VD_CONF must
4563 * be written to precisely those devices listed with
4565 * DDF_SPARE_ASSIGN_MAGIC
4566 * replacement Spare Assignment Record... but for which device?
4569 * - to create a new array, we send a VIRT_RECORD and
4570 * a VD_CONF. Then assemble and start the array.
4571 * - to activate a spare we send a VD_CONF to add the phys_refnum
4572 * and offset. This will also mark the spare as active with
4573 * a spare-assignment record.
4575 struct ddf_super
*ddf
= st
->sb
;
4576 be32
*magic
= (be32
*)update
->buf
;
4577 struct phys_disk
*pd
;
4578 struct virtual_disk
*vd
;
4579 struct vd_config
*vc
;
4583 unsigned int pdnum
, pd2
, len
;
4585 dprintf("Process update %x\n", be32_to_cpu(*magic
));
4587 if (be32_eq(*magic
, DDF_PHYS_RECORDS_MAGIC
)) {
4588 if (update
->len
!= (sizeof(struct phys_disk
) +
4589 sizeof(struct phys_disk_entry
)))
4591 pd
= (struct phys_disk
*)update
->buf
;
4593 ent
= be16_to_cpu(pd
->used_pdes
);
4594 if (ent
>= be16_to_cpu(ddf
->phys
->max_pdes
))
4596 if (be16_and(pd
->entries
[0].state
, cpu_to_be16(DDF_Missing
))) {
4598 /* removing this disk. */
4599 be16_set(ddf
->phys
->entries
[ent
].state
,
4600 cpu_to_be16(DDF_Missing
));
4601 for (dlp
= &ddf
->dlist
; *dlp
; dlp
= &(*dlp
)->next
) {
4602 struct dl
*dl
= *dlp
;
4603 if (dl
->pdnum
== (signed)ent
) {
4606 /* FIXME this doesn't free
4613 ddf_set_updates_pending(ddf
);
4616 if (!all_ff(ddf
->phys
->entries
[ent
].guid
))
4618 ddf
->phys
->entries
[ent
] = pd
->entries
[0];
4619 ddf
->phys
->used_pdes
= cpu_to_be16
4620 (1 + be16_to_cpu(ddf
->phys
->used_pdes
));
4621 ddf_set_updates_pending(ddf
);
4622 if (ddf
->add_list
) {
4623 struct active_array
*a
;
4624 struct dl
*al
= ddf
->add_list
;
4625 ddf
->add_list
= al
->next
;
4627 al
->next
= ddf
->dlist
;
4630 /* As a device has been added, we should check
4631 * for any degraded devices that might make
4632 * use of this spare */
4633 for (a
= st
->arrays
; a
; a
=a
->next
)
4634 a
->check_degraded
= 1;
4636 } else if (be32_eq(*magic
, DDF_VIRT_RECORDS_MAGIC
)) {
4637 if (update
->len
!= (sizeof(struct virtual_disk
) +
4638 sizeof(struct virtual_entry
)))
4640 vd
= (struct virtual_disk
*)update
->buf
;
4642 if (vd
->entries
[0].state
== DDF_state_deleted
) {
4643 if (_kill_subarray_ddf(ddf
, vd
->entries
[0].guid
))
4646 ent
= find_vde_by_guid(ddf
, vd
->entries
[0].guid
);
4647 if (ent
!= DDF_NOTFOUND
) {
4648 dprintf("%s: VD %s exists already in slot %d\n",
4649 __func__
, guid_str(vd
->entries
[0].guid
),
4653 ent
= find_unused_vde(ddf
);
4654 if (ent
== DDF_NOTFOUND
)
4656 ddf
->virt
->entries
[ent
] = vd
->entries
[0];
4657 ddf
->virt
->populated_vdes
=
4660 ddf
->virt
->populated_vdes
));
4661 dprintf("%s: added VD %s in slot %d(s=%02x i=%02x)\n",
4662 __func__
, guid_str(vd
->entries
[0].guid
), ent
,
4663 ddf
->virt
->entries
[ent
].state
,
4664 ddf
->virt
->entries
[ent
].init_state
);
4666 ddf_set_updates_pending(ddf
);
4669 else if (be32_eq(*magic
, DDF_VD_CONF_MAGIC
)) {
4670 vc
= (struct vd_config
*)update
->buf
;
4671 len
= ddf
->conf_rec_len
* 512;
4672 if ((unsigned int)update
->len
!= len
* vc
->sec_elmnt_count
) {
4673 pr_err("%s: %s: insufficient data (%d) for %u BVDs\n",
4674 __func__
, guid_str(vc
->guid
), update
->len
,
4675 vc
->sec_elmnt_count
);
4678 for (vcl
= ddf
->conflist
; vcl
; vcl
= vcl
->next
)
4679 if (memcmp(vcl
->conf
.guid
, vc
->guid
, DDF_GUID_LEN
) == 0)
4681 dprintf("%s: conf update for %s (%s)\n", __func__
,
4682 guid_str(vc
->guid
), (vcl
? "old" : "new"));
4684 /* An update, just copy the phys_refnum and lba_offset
4689 copy_matching_bvd(ddf
, &vcl
->conf
, update
);
4690 for (k
= 0; k
< be16_to_cpu(vc
->prim_elmnt_count
); k
++)
4691 dprintf("BVD %u has %08x at %llu\n", 0,
4692 be32_to_cpu(vcl
->conf
.phys_refnum
[k
]),
4693 be64_to_cpu(LBA_OFFSET(ddf
,
4695 for (i
= 1; i
< vc
->sec_elmnt_count
; i
++) {
4696 copy_matching_bvd(ddf
, vcl
->other_bvds
[i
-1],
4698 for (k
= 0; k
< be16_to_cpu(
4699 vc
->prim_elmnt_count
); k
++)
4700 dprintf("BVD %u has %08x at %llu\n", i
,
4702 (vcl
->other_bvds
[i
-1]->
4707 vcl
->other_bvds
[i
-1])[k
]));
4714 vcl
= update
->space
;
4715 update
->space
= NULL
;
4716 vcl
->next
= ddf
->conflist
;
4717 memcpy(&vcl
->conf
, vc
, len
);
4718 ent
= find_vde_by_guid(ddf
, vc
->guid
);
4719 if (ent
== DDF_NOTFOUND
)
4722 ddf
->conflist
= vcl
;
4723 for (i
= 1; i
< vc
->sec_elmnt_count
; i
++)
4724 memcpy(vcl
->other_bvds
[i
-1],
4725 update
->buf
+ len
* i
, len
);
4727 /* Set DDF_Transition on all Failed devices - to help
4728 * us detect those that are no longer in use
4730 for (pdnum
= 0; pdnum
< be16_to_cpu(ddf
->phys
->used_pdes
);
4732 if (be16_and(ddf
->phys
->entries
[pdnum
].state
,
4733 cpu_to_be16(DDF_Failed
)))
4734 be16_set(ddf
->phys
->entries
[pdnum
].state
,
4735 cpu_to_be16(DDF_Transition
));
4736 /* Now make sure vlist is correct for each dl. */
4737 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
4738 unsigned int vn
= 0;
4739 int in_degraded
= 0;
4743 for (vcl
= ddf
->conflist
; vcl
; vcl
= vcl
->next
) {
4744 unsigned int dn
, ibvd
;
4745 const struct vd_config
*conf
;
4747 dn
= get_pd_index_from_refnum(vcl
,
4751 if (dn
== DDF_NOTFOUND
)
4753 dprintf("dev %d/%08x has %s (sec=%u) at %d\n",
4755 be32_to_cpu(dl
->disk
.refnum
),
4756 guid_str(conf
->guid
),
4757 conf
->sec_elmnt_seq
, vn
);
4758 /* Clear the Transition flag */
4760 (ddf
->phys
->entries
[dl
->pdnum
].state
,
4761 cpu_to_be16(DDF_Failed
)))
4762 be16_clear(ddf
->phys
4763 ->entries
[dl
->pdnum
].state
,
4764 cpu_to_be16(DDF_Transition
));
4765 dl
->vlist
[vn
++] = vcl
;
4766 vstate
= ddf
->virt
->entries
[vcl
->vcnum
].state
4768 if (vstate
== DDF_state_degraded
||
4769 vstate
== DDF_state_part_optimal
)
4772 while (vn
< ddf
->max_part
)
4773 dl
->vlist
[vn
++] = NULL
;
4775 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4776 cpu_to_be16(DDF_Global_Spare
));
4777 if (!be16_and(ddf
->phys
4778 ->entries
[dl
->pdnum
].type
,
4779 cpu_to_be16(DDF_Active_in_VD
))) {
4781 ->entries
[dl
->pdnum
].type
,
4782 cpu_to_be16(DDF_Active_in_VD
));
4785 ->entries
[dl
->pdnum
]
4792 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4793 cpu_to_be16(DDF_Global_Spare
));
4794 be16_set(ddf
->phys
->entries
[dl
->pdnum
].type
,
4795 cpu_to_be16(DDF_Spare
));
4797 if (!dl
->vlist
[0] && !dl
->spare
) {
4798 be16_set(ddf
->phys
->entries
[dl
->pdnum
].type
,
4799 cpu_to_be16(DDF_Global_Spare
));
4800 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4801 cpu_to_be16(DDF_Spare
));
4802 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4803 cpu_to_be16(DDF_Active_in_VD
));
4807 /* Now remove any 'Failed' devices that are not part
4808 * of any VD. They will have the Transition flag set.
4809 * Once done, we need to update all dl->pdnum numbers.
4812 for (pdnum
= 0; pdnum
< be16_to_cpu(ddf
->phys
->used_pdes
);
4814 if (be16_and(ddf
->phys
->entries
[pdnum
].state
,
4815 cpu_to_be16(DDF_Failed
))
4816 && be16_and(ddf
->phys
->entries
[pdnum
].state
,
4817 cpu_to_be16(DDF_Transition
))) {
4818 /* skip this one unless in dlist*/
4819 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4820 if (dl
->pdnum
== (int)pdnum
)
4828 ddf
->phys
->entries
[pd2
] =
4829 ddf
->phys
->entries
[pdnum
];
4830 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4831 if (dl
->pdnum
== (int)pdnum
)
4836 ddf
->phys
->used_pdes
= cpu_to_be16(pd2
);
4837 while (pd2
< pdnum
) {
4838 memset(ddf
->phys
->entries
[pd2
].guid
, 0xff,
4843 ddf_set_updates_pending(ddf
);
4845 /* case DDF_SPARE_ASSIGN_MAGIC */
4848 static void ddf_prepare_update(struct supertype
*st
,
4849 struct metadata_update
*update
)
4851 /* This update arrived at managemon.
4852 * We are about to pass it to monitor.
4853 * If a malloc is needed, do it here.
4855 struct ddf_super
*ddf
= st
->sb
;
4856 be32
*magic
= (be32
*)update
->buf
;
4857 if (be32_eq(*magic
, DDF_VD_CONF_MAGIC
)) {
4859 struct vd_config
*conf
= (struct vd_config
*) update
->buf
;
4860 if (posix_memalign(&update
->space
, 512,
4861 offsetof(struct vcl
, conf
)
4862 + ddf
->conf_rec_len
* 512) != 0) {
4863 update
->space
= NULL
;
4866 vcl
= update
->space
;
4867 vcl
->conf
.sec_elmnt_count
= conf
->sec_elmnt_count
;
4868 if (alloc_other_bvds(ddf
, vcl
) != 0) {
4869 free(update
->space
);
4870 update
->space
= NULL
;
4876 * Check degraded state of a RAID10.
4877 * returns 2 for good, 1 for degraded, 0 for failed, and -1 for error
4879 static int raid10_degraded(struct mdinfo
*info
)
4887 n_prim
= info
->array
.layout
& ~0x100;
4888 n_bvds
= info
->array
.raid_disks
/ n_prim
;
4889 found
= xmalloc(n_bvds
);
4892 memset(found
, 0, n_bvds
);
4893 for (d
= info
->devs
; d
; d
= d
->next
) {
4894 i
= d
->disk
.raid_disk
/ n_prim
;
4896 pr_err("%s: BUG: invalid raid disk\n", __func__
);
4899 if (d
->state_fd
> 0)
4903 for (i
= 0; i
< n_bvds
; i
++)
4905 dprintf("%s: BVD %d/%d failed\n", __func__
, i
, n_bvds
);
4908 } else if (found
[i
] < n_prim
) {
4909 dprintf("%s: BVD %d/%d degraded\n", __func__
, i
,
4919 * Check if the array 'a' is degraded but not failed.
4920 * If it is, find as many spares as are available and needed and
4921 * arrange for their inclusion.
4922 * We only choose devices which are not already in the array,
4923 * and prefer those with a spare-assignment to this array.
4924 * Otherwise we choose global spares - assuming always that
4925 * there is enough room.
4926 * For each spare that we assign, we return an 'mdinfo' which
4927 * describes the position for the device in the array.
4928 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
4929 * the new phys_refnum and lba_offset values.
4931 * Only worry about BVDs at the moment.
4933 static struct mdinfo
*ddf_activate_spare(struct active_array
*a
,
4934 struct metadata_update
**updates
)
4938 struct ddf_super
*ddf
= a
->container
->sb
;
4940 struct mdinfo
*rv
= NULL
;
4942 struct metadata_update
*mu
;
4947 struct vd_config
*vc
;
4950 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
4951 if ((d
->curr_state
& DS_FAULTY
) &&
4953 /* wait for Removal to happen */
4955 if (d
->state_fd
>= 0)
4959 dprintf("%s: working=%d (%d) level=%d\n", __func__
, working
,
4960 a
->info
.array
.raid_disks
,
4961 a
->info
.array
.level
);
4962 if (working
== a
->info
.array
.raid_disks
)
4963 return NULL
; /* array not degraded */
4964 switch (a
->info
.array
.level
) {
4967 return NULL
; /* failed */
4971 if (working
< a
->info
.array
.raid_disks
- 1)
4972 return NULL
; /* failed */
4975 if (working
< a
->info
.array
.raid_disks
- 2)
4976 return NULL
; /* failed */
4979 if (raid10_degraded(&a
->info
) < 1)
4982 default: /* concat or stripe */
4983 return NULL
; /* failed */
4986 /* For each slot, if it is not working, find a spare */
4988 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
4989 for (d
= a
->info
.devs
; d
; d
= d
->next
)
4990 if (d
->disk
.raid_disk
== i
)
4992 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
4993 if (d
&& (d
->state_fd
>= 0))
4996 /* OK, this device needs recovery. Find a spare */
4998 for ( ; dl
; dl
= dl
->next
) {
4999 unsigned long long esize
;
5000 unsigned long long pos
;
5003 int is_dedicated
= 0;
5010 state
= ddf
->phys
->entries
[dl
->pdnum
].state
;
5012 cpu_to_be16(DDF_Failed
|DDF_Missing
)) ||
5014 cpu_to_be16(DDF_Online
)))
5017 /* If in this array, skip */
5018 for (d2
= a
->info
.devs
; d2
; d2
= d2
->next
)
5019 if (d2
->state_fd
>= 0 &&
5020 d2
->disk
.major
== dl
->major
&&
5021 d2
->disk
.minor
== dl
->minor
) {
5022 dprintf("%x:%x (%08x) already in array\n",
5023 dl
->major
, dl
->minor
,
5024 be32_to_cpu(dl
->disk
.refnum
));
5029 if (be16_and(ddf
->phys
->entries
[dl
->pdnum
].type
,
5030 cpu_to_be16(DDF_Spare
))) {
5031 /* Check spare assign record */
5033 if (dl
->spare
->type
& DDF_spare_dedicated
) {
5034 /* check spare_ents for guid */
5040 if (memcmp(dl
->spare
->spare_ents
[j
].guid
,
5041 ddf
->virt
->entries
[a
->info
.container_member
].guid
,
5048 } else if (be16_and(ddf
->phys
->entries
[dl
->pdnum
].type
,
5049 cpu_to_be16(DDF_Global_Spare
))) {
5051 } else if (!be16_and(ddf
->phys
5052 ->entries
[dl
->pdnum
].state
,
5053 cpu_to_be16(DDF_Failed
))) {
5054 /* we can possibly use some of this */
5057 if ( ! (is_dedicated
||
5058 (is_global
&& global_ok
))) {
5059 dprintf("%x:%x not suitable: %d %d\n", dl
->major
, dl
->minor
,
5060 is_dedicated
, is_global
);
5064 /* We are allowed to use this device - is there space?
5065 * We need a->info.component_size sectors */
5066 ex
= get_extents(ddf
, dl
);
5068 dprintf("cannot get extents\n");
5075 esize
= ex
[j
].start
- pos
;
5076 if (esize
>= a
->info
.component_size
)
5078 pos
= ex
[j
].start
+ ex
[j
].size
;
5080 } while (ex
[j
-1].size
);
5083 if (esize
< a
->info
.component_size
) {
5084 dprintf("%x:%x has no room: %llu %llu\n",
5085 dl
->major
, dl
->minor
,
5086 esize
, a
->info
.component_size
);
5091 /* Cool, we have a device with some space at pos */
5092 di
= xcalloc(1, sizeof(*di
));
5093 di
->disk
.number
= i
;
5094 di
->disk
.raid_disk
= i
;
5095 di
->disk
.major
= dl
->major
;
5096 di
->disk
.minor
= dl
->minor
;
5098 di
->recovery_start
= 0;
5099 di
->data_offset
= pos
;
5100 di
->component_size
= a
->info
.component_size
;
5103 dprintf("%x:%x (%08x) to be %d at %llu\n",
5104 dl
->major
, dl
->minor
,
5105 be32_to_cpu(dl
->disk
.refnum
), i
, pos
);
5109 if (!dl
&& ! global_ok
) {
5110 /* not enough dedicated spares, try global */
5118 /* No spares found */
5120 /* Now 'rv' has a list of devices to return.
5121 * Create a metadata_update record to update the
5122 * phys_refnum and lba_offset values
5124 vc
= find_vdcr(ddf
, a
->info
.container_member
, rv
->disk
.raid_disk
,
5129 mu
= xmalloc(sizeof(*mu
));
5130 if (posix_memalign(&mu
->space
, 512, sizeof(struct vcl
)) != 0) {
5135 mu
->len
= ddf
->conf_rec_len
* 512 * vcl
->conf
.sec_elmnt_count
;
5136 mu
->buf
= xmalloc(mu
->len
);
5138 mu
->space_list
= NULL
;
5139 mu
->next
= *updates
;
5140 memcpy(mu
->buf
, &vcl
->conf
, ddf
->conf_rec_len
* 512);
5141 for (j
= 1; j
< vcl
->conf
.sec_elmnt_count
; j
++)
5142 memcpy(mu
->buf
+ j
* ddf
->conf_rec_len
* 512,
5143 vcl
->other_bvds
[j
-1], ddf
->conf_rec_len
* 512);
5145 vc
= (struct vd_config
*)mu
->buf
;
5146 for (di
= rv
; di
; di
= di
->next
) {
5147 unsigned int i_sec
, i_prim
;
5148 i_sec
= di
->disk
.raid_disk
5149 / be16_to_cpu(vcl
->conf
.prim_elmnt_count
);
5150 i_prim
= di
->disk
.raid_disk
5151 % be16_to_cpu(vcl
->conf
.prim_elmnt_count
);
5152 vc
= (struct vd_config
*)(mu
->buf
5153 + i_sec
* ddf
->conf_rec_len
* 512);
5154 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
5155 if (dl
->major
== di
->disk
.major
5156 && dl
->minor
== di
->disk
.minor
)
5158 if (!dl
|| dl
->pdnum
< 0) {
5159 pr_err("%s: BUG: can't find disk %d (%d/%d)\n",
5160 __func__
, di
->disk
.raid_disk
,
5161 di
->disk
.major
, di
->disk
.minor
);
5164 vc
->phys_refnum
[i_prim
] = ddf
->phys
->entries
[dl
->pdnum
].refnum
;
5165 LBA_OFFSET(ddf
, vc
)[i_prim
] = cpu_to_be64(di
->data_offset
);
5166 dprintf("BVD %u gets %u: %08x at %llu\n", i_sec
, i_prim
,
5167 be32_to_cpu(vc
->phys_refnum
[i_prim
]),
5168 be64_to_cpu(LBA_OFFSET(ddf
, vc
)[i_prim
]));
5173 #endif /* MDASSEMBLE */
5175 static int ddf_level_to_layout(int level
)
5182 return ALGORITHM_LEFT_SYMMETRIC
;
5184 return ALGORITHM_ROTATING_N_CONTINUE
;
5192 static void default_geometry_ddf(struct supertype
*st
, int *level
, int *layout
, int *chunk
)
5194 if (level
&& *level
== UnSet
)
5195 *level
= LEVEL_CONTAINER
;
5197 if (level
&& layout
&& *layout
== UnSet
)
5198 *layout
= ddf_level_to_layout(*level
);
5201 struct superswitch super_ddf
= {
5203 .examine_super
= examine_super_ddf
,
5204 .brief_examine_super
= brief_examine_super_ddf
,
5205 .brief_examine_subarrays
= brief_examine_subarrays_ddf
,
5206 .export_examine_super
= export_examine_super_ddf
,
5207 .detail_super
= detail_super_ddf
,
5208 .brief_detail_super
= brief_detail_super_ddf
,
5209 .validate_geometry
= validate_geometry_ddf
,
5210 .write_init_super
= write_init_super_ddf
,
5211 .add_to_super
= add_to_super_ddf
,
5212 .remove_from_super
= remove_from_super_ddf
,
5213 .load_container
= load_container_ddf
,
5214 .copy_metadata
= copy_metadata_ddf
,
5215 .kill_subarray
= kill_subarray_ddf
,
5217 .match_home
= match_home_ddf
,
5218 .uuid_from_super
= uuid_from_super_ddf
,
5219 .getinfo_super
= getinfo_super_ddf
,
5220 .update_super
= update_super_ddf
,
5222 .avail_size
= avail_size_ddf
,
5224 .compare_super
= compare_super_ddf
,
5226 .load_super
= load_super_ddf
,
5227 .init_super
= init_super_ddf
,
5228 .store_super
= store_super_ddf
,
5229 .free_super
= free_super_ddf
,
5230 .match_metadata_desc
= match_metadata_desc_ddf
,
5231 .container_content
= container_content_ddf
,
5232 .default_geometry
= default_geometry_ddf
,
5238 .open_new
= ddf_open_new
,
5239 .set_array_state
= ddf_set_array_state
,
5240 .set_disk
= ddf_set_disk
,
5241 .sync_metadata
= ddf_sync_metadata
,
5242 .process_update
= ddf_process_update
,
5243 .prepare_update
= ddf_prepare_update
,
5244 .activate_spare
= ddf_activate_spare
,