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 */
232 be16 used_pdes
; /* This is a counter, not a max - the list
233 * of used entries may not be dense */
236 struct phys_disk_entry
{
237 char guid
[DDF_GUID_LEN
];
241 be64 config_size
; /* DDF structures must be after here */
242 char path
[18]; /* Another horrible structure really
243 * but is "used for information
249 /* phys_disk_entry.type is a bitmap - bigendian remember */
250 #define DDF_Forced_PD_GUID 1
251 #define DDF_Active_in_VD 2
252 #define DDF_Global_Spare 4 /* VD_CONF records are ignored */
253 #define DDF_Spare 8 /* overrides Global_spare */
254 #define DDF_Foreign 16
255 #define DDF_Legacy 32 /* no DDF on this device */
257 #define DDF_Interface_mask 0xf00
258 #define DDF_Interface_SCSI 0x100
259 #define DDF_Interface_SAS 0x200
260 #define DDF_Interface_SATA 0x300
261 #define DDF_Interface_FC 0x400
263 /* phys_disk_entry.state is a bigendian bitmap */
265 #define DDF_Failed 2 /* overrides 1,4,8 */
266 #define DDF_Rebuilding 4
267 #define DDF_Transition 8
269 #define DDF_ReadErrors 32
270 #define DDF_Missing 64
272 /* The content of the virt_section global scope */
273 struct virtual_disk
{
274 be32 magic
; /* DDF_VIRT_RECORDS_MAGIC */
279 struct virtual_entry
{
280 char guid
[DDF_GUID_LEN
];
282 __u16 pad0
; /* 0xffff */
292 /* virtual_entry.type is a bitmap - bigendian */
294 #define DDF_Enforce_Groups 2
295 #define DDF_Unicode 4
296 #define DDF_Owner_Valid 8
298 /* virtual_entry.state is a bigendian bitmap */
299 #define DDF_state_mask 0x7
300 #define DDF_state_optimal 0x0
301 #define DDF_state_degraded 0x1
302 #define DDF_state_deleted 0x2
303 #define DDF_state_missing 0x3
304 #define DDF_state_failed 0x4
305 #define DDF_state_part_optimal 0x5
307 #define DDF_state_morphing 0x8
308 #define DDF_state_inconsistent 0x10
310 /* virtual_entry.init_state is a bigendian bitmap */
311 #define DDF_initstate_mask 0x03
312 #define DDF_init_not 0x00
313 #define DDF_init_quick 0x01 /* initialisation is progress.
314 * i.e. 'state_inconsistent' */
315 #define DDF_init_full 0x02
317 #define DDF_access_mask 0xc0
318 #define DDF_access_rw 0x00
319 #define DDF_access_ro 0x80
320 #define DDF_access_blocked 0xc0
322 /* The content of the config_section - local scope
323 * It has multiple records each config_record_len sectors
324 * They can be vd_config or spare_assign
328 be32 magic
; /* DDF_VD_CONF_MAGIC */
330 char guid
[DDF_GUID_LEN
];
334 be16 prim_elmnt_count
;
335 __u8 chunk_shift
; /* 0 == 512, 1==1024 etc */
338 __u8 sec_elmnt_count
;
341 be64 blocks
; /* blocks per component could be different
342 * on different component devices...(only
343 * for concat I hope) */
344 be64 array_blocks
; /* blocks in array */
352 __u8 v0
[32]; /* reserved- 0xff */
353 __u8 v1
[32]; /* reserved- 0xff */
354 __u8 v2
[16]; /* reserved- 0xff */
355 __u8 v3
[16]; /* reserved- 0xff */
357 be32 phys_refnum
[0]; /* refnum of each disk in sequence */
358 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
359 bvd are always the same size */
361 #define LBA_OFFSET(ddf, vd) ((be64 *) &(vd)->phys_refnum[(ddf)->mppe])
363 /* vd_config.cache_pol[7] is a bitmap */
364 #define DDF_cache_writeback 1 /* else writethrough */
365 #define DDF_cache_wadaptive 2 /* only applies if writeback */
366 #define DDF_cache_readahead 4
367 #define DDF_cache_radaptive 8 /* only if doing read-ahead */
368 #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
369 #define DDF_cache_wallowed 32 /* enable write caching */
370 #define DDF_cache_rallowed 64 /* enable read caching */
372 struct spare_assign
{
373 be32 magic
; /* DDF_SPARE_ASSIGN_MAGIC */
378 be16 populated
; /* SAEs used */
379 be16 max
; /* max SAEs */
381 struct spare_assign_entry
{
382 char guid
[DDF_GUID_LEN
];
383 be16 secondary_element
;
387 /* spare_assign.type is a bitmap */
388 #define DDF_spare_dedicated 0x1 /* else global */
389 #define DDF_spare_revertible 0x2 /* else committable */
390 #define DDF_spare_active 0x4 /* else not active */
391 #define DDF_spare_affinity 0x8 /* enclosure affinity */
393 /* The data_section contents - local scope */
395 be32 magic
; /* DDF_PHYS_DATA_MAGIC */
397 char guid
[DDF_GUID_LEN
];
398 be32 refnum
; /* crc of some magic drive data ... */
399 __u8 forced_ref
; /* set when above was not result of magic */
400 __u8 forced_guid
; /* set if guid was forced rather than magic */
405 /* bbm_section content */
406 struct bad_block_log
{
413 struct mapped_block
{
414 be64 defective_start
;
415 be32 replacement_start
;
421 /* Struct for internally holding ddf structures */
422 /* The DDF structure stored on each device is potentially
423 * quite different, as some data is global and some is local.
424 * The global data is:
427 * - Physical disk records
428 * - Virtual disk records
430 * - Configuration records
431 * - Physical Disk data section
432 * ( and Bad block and vendor which I don't care about yet).
434 * The local data is parsed into separate lists as it is read
435 * and reconstructed for writing. This means that we only need
436 * to make config changes once and they are automatically
437 * propagated to all devices.
438 * The global (config and disk data) records are each in a list
439 * of separate data structures. When writing we find the entry
440 * or entries applicable to the particular device.
443 struct ddf_header anchor
, primary
, secondary
;
444 struct ddf_controller_data controller
;
445 struct ddf_header
*active
;
446 struct phys_disk
*phys
;
447 struct virtual_disk
*virt
;
450 unsigned int max_part
, mppe
, conf_rec_len
;
458 unsigned int vcnum
; /* index into ->virt */
459 /* For an array with a secondary level there are
460 * multiple vd_config structures, all with the same
461 * guid but with different sec_elmnt_seq.
462 * One of these structures is in 'conf' below.
463 * The others are in other_bvds, not in any
466 struct vd_config
**other_bvds
;
467 __u64
*block_sizes
; /* NULL if all the same */
470 struct vd_config conf
;
471 } *conflist
, *currentconf
;
480 unsigned long long size
; /* sectors */
481 be64 primary_lba
; /* sectors */
482 be64 secondary_lba
; /* sectors */
483 be64 workspace_lba
; /* sectors */
484 int pdnum
; /* index in ->phys */
485 struct spare_assign
*spare
;
486 void *mdupdate
; /* hold metadata update */
488 /* These fields used by auto-layout */
489 int raiddisk
; /* slot to fill in autolayout */
493 struct disk_data disk
;
494 struct vcl
*vlist
[0]; /* max_part in size */
499 static int load_super_ddf_all(struct supertype
*st
, int fd
,
500 void **sbp
, char *devname
);
501 static int get_svd_state(const struct ddf_super
*, const struct vcl
*);
503 validate_geometry_ddf_container(struct supertype
*st
,
504 int level
, int layout
, int raiddisks
,
505 int chunk
, unsigned long long size
,
506 unsigned long long data_offset
,
507 char *dev
, unsigned long long *freesize
,
510 static int validate_geometry_ddf_bvd(struct supertype
*st
,
511 int level
, int layout
, int raiddisks
,
512 int *chunk
, unsigned long long size
,
513 unsigned long long data_offset
,
514 char *dev
, unsigned long long *freesize
,
518 static void free_super_ddf(struct supertype
*st
);
519 static int all_ff(const char *guid
);
520 static unsigned int get_pd_index_from_refnum(const struct vcl
*vc
,
521 be32 refnum
, unsigned int nmax
,
522 const struct vd_config
**bvd
,
524 static void getinfo_super_ddf(struct supertype
*st
, struct mdinfo
*info
, char *map
);
525 static void uuid_from_ddf_guid(const char *guid
, int uuid
[4]);
526 static void uuid_from_super_ddf(struct supertype
*st
, int uuid
[4]);
527 static void _ddf_array_name(char *name
, const struct ddf_super
*ddf
, int i
);
528 static void getinfo_super_ddf_bvd(struct supertype
*st
, struct mdinfo
*info
, char *map
);
529 static int init_super_ddf_bvd(struct supertype
*st
,
530 mdu_array_info_t
*info
,
531 unsigned long long size
,
532 char *name
, char *homehost
,
533 int *uuid
, unsigned long long data_offset
);
536 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
540 static void pr_state(struct ddf_super
*ddf
, const char *msg
)
543 dprintf("%s/%s: ", __func__
, msg
);
544 for (i
= 0; i
< be16_to_cpu(ddf
->active
->max_vd_entries
); i
++) {
545 if (all_ff(ddf
->virt
->entries
[i
].guid
))
547 dprintf("%u(s=%02x i=%02x) ", i
,
548 ddf
->virt
->entries
[i
].state
,
549 ddf
->virt
->entries
[i
].init_state
);
554 static void pr_state(const struct ddf_super
*ddf
, const char *msg
) {}
557 static void _ddf_set_updates_pending(struct ddf_super
*ddf
, const char *func
)
559 if (ddf
->updates_pending
)
561 ddf
->updates_pending
= 1;
562 ddf
->active
->seq
= cpu_to_be32((be32_to_cpu(ddf
->active
->seq
)+1));
566 #define ddf_set_updates_pending(x) _ddf_set_updates_pending((x), __func__)
568 static be32
calc_crc(void *buf
, int len
)
570 /* crcs are always at the same place as in the ddf_header */
571 struct ddf_header
*ddf
= buf
;
572 be32 oldcrc
= ddf
->crc
;
574 ddf
->crc
= cpu_to_be32(0xffffffff);
576 newcrc
= crc32(0, buf
, len
);
578 /* The crc is stored (like everything) bigendian, so convert
579 * here for simplicity
581 return cpu_to_be32(newcrc
);
584 #define DDF_INVALID_LEVEL 0xff
585 #define DDF_NO_SECONDARY 0xff
586 static int err_bad_md_layout(const mdu_array_info_t
*array
)
588 pr_err("RAID%d layout %x with %d disks is unsupported for DDF\n",
589 array
->level
, array
->layout
, array
->raid_disks
);
593 static int layout_md2ddf(const mdu_array_info_t
*array
,
594 struct vd_config
*conf
)
596 be16 prim_elmnt_count
= cpu_to_be16(array
->raid_disks
);
597 __u8 prl
= DDF_INVALID_LEVEL
, rlq
= 0;
598 __u8 sec_elmnt_count
= 1;
599 __u8 srl
= DDF_NO_SECONDARY
;
601 switch (array
->level
) {
606 rlq
= DDF_RAID0_SIMPLE
;
610 switch (array
->raid_disks
) {
612 rlq
= DDF_RAID1_SIMPLE
;
615 rlq
= DDF_RAID1_MULTI
;
618 return err_bad_md_layout(array
);
623 if (array
->layout
!= 0)
624 return err_bad_md_layout(array
);
629 switch (array
->layout
) {
630 case ALGORITHM_LEFT_ASYMMETRIC
:
631 rlq
= DDF_RAID5_N_RESTART
;
633 case ALGORITHM_RIGHT_ASYMMETRIC
:
634 rlq
= DDF_RAID5_0_RESTART
;
636 case ALGORITHM_LEFT_SYMMETRIC
:
637 rlq
= DDF_RAID5_N_CONTINUE
;
639 case ALGORITHM_RIGHT_SYMMETRIC
:
640 /* not mentioned in standard */
642 return err_bad_md_layout(array
);
647 switch (array
->layout
) {
648 case ALGORITHM_ROTATING_N_RESTART
:
649 rlq
= DDF_RAID5_N_RESTART
;
651 case ALGORITHM_ROTATING_ZERO_RESTART
:
652 rlq
= DDF_RAID6_0_RESTART
;
654 case ALGORITHM_ROTATING_N_CONTINUE
:
655 rlq
= DDF_RAID5_N_CONTINUE
;
658 return err_bad_md_layout(array
);
663 if (array
->raid_disks
% 2 == 0 && array
->layout
== 0x102) {
664 rlq
= DDF_RAID1_SIMPLE
;
665 prim_elmnt_count
= cpu_to_be16(2);
666 sec_elmnt_count
= array
->raid_disks
/ 2;
667 } else if (array
->raid_disks
% 3 == 0
668 && array
->layout
== 0x103) {
669 rlq
= DDF_RAID1_MULTI
;
670 prim_elmnt_count
= cpu_to_be16(3);
671 sec_elmnt_count
= array
->raid_disks
/ 3;
673 return err_bad_md_layout(array
);
678 return err_bad_md_layout(array
);
681 conf
->prim_elmnt_count
= prim_elmnt_count
;
684 conf
->sec_elmnt_count
= sec_elmnt_count
;
688 static int err_bad_ddf_layout(const struct vd_config
*conf
)
690 pr_err("DDF RAID %u qualifier %u with %u disks is unsupported\n",
691 conf
->prl
, conf
->rlq
, be16_to_cpu(conf
->prim_elmnt_count
));
695 static int layout_ddf2md(const struct vd_config
*conf
,
696 mdu_array_info_t
*array
)
698 int level
= LEVEL_UNSUPPORTED
;
700 int raiddisks
= be16_to_cpu(conf
->prim_elmnt_count
);
702 if (conf
->sec_elmnt_count
> 1) {
703 /* see also check_secondary() */
704 if (conf
->prl
!= DDF_RAID1
||
705 (conf
->srl
!= DDF_2STRIPED
&& conf
->srl
!= DDF_2SPANNED
)) {
706 pr_err("Unsupported secondary RAID level %u/%u\n",
707 conf
->prl
, conf
->srl
);
710 if (raiddisks
== 2 && conf
->rlq
== DDF_RAID1_SIMPLE
)
712 else if (raiddisks
== 3 && conf
->rlq
== DDF_RAID1_MULTI
)
715 return err_bad_ddf_layout(conf
);
716 raiddisks
*= conf
->sec_elmnt_count
;
723 level
= LEVEL_LINEAR
;
726 if (conf
->rlq
!= DDF_RAID0_SIMPLE
)
727 return err_bad_ddf_layout(conf
);
731 if (!((conf
->rlq
== DDF_RAID1_SIMPLE
&& raiddisks
== 2) ||
732 (conf
->rlq
== DDF_RAID1_MULTI
&& raiddisks
== 3)))
733 return err_bad_ddf_layout(conf
);
737 if (conf
->rlq
!= DDF_RAID4_N
)
738 return err_bad_ddf_layout(conf
);
743 case DDF_RAID5_N_RESTART
:
744 layout
= ALGORITHM_LEFT_ASYMMETRIC
;
746 case DDF_RAID5_0_RESTART
:
747 layout
= ALGORITHM_RIGHT_ASYMMETRIC
;
749 case DDF_RAID5_N_CONTINUE
:
750 layout
= ALGORITHM_LEFT_SYMMETRIC
;
753 return err_bad_ddf_layout(conf
);
759 case DDF_RAID5_N_RESTART
:
760 layout
= ALGORITHM_ROTATING_N_RESTART
;
762 case DDF_RAID6_0_RESTART
:
763 layout
= ALGORITHM_ROTATING_ZERO_RESTART
;
765 case DDF_RAID5_N_CONTINUE
:
766 layout
= ALGORITHM_ROTATING_N_CONTINUE
;
769 return err_bad_ddf_layout(conf
);
774 return err_bad_ddf_layout(conf
);
778 array
->level
= level
;
779 array
->layout
= layout
;
780 array
->raid_disks
= raiddisks
;
784 static int load_ddf_header(int fd
, unsigned long long lba
,
785 unsigned long long size
,
787 struct ddf_header
*hdr
, struct ddf_header
*anchor
)
789 /* read a ddf header (primary or secondary) from fd/lba
790 * and check that it is consistent with anchor
792 * magic, crc, guid, rev, and LBA's header_type, and
793 * everything after header_type must be the same
798 if (lseek64(fd
, lba
<<9, 0) < 0)
801 if (read(fd
, hdr
, 512) != 512)
804 if (!be32_eq(hdr
->magic
, DDF_HEADER_MAGIC
)) {
805 pr_err("%s: bad header magic\n", __func__
);
808 if (!be32_eq(calc_crc(hdr
, 512), hdr
->crc
)) {
809 pr_err("%s: bad CRC\n", __func__
);
812 if (memcmp(anchor
->guid
, hdr
->guid
, DDF_GUID_LEN
) != 0 ||
813 memcmp(anchor
->revision
, hdr
->revision
, 8) != 0 ||
814 !be64_eq(anchor
->primary_lba
, hdr
->primary_lba
) ||
815 !be64_eq(anchor
->secondary_lba
, hdr
->secondary_lba
) ||
817 memcmp(anchor
->pad2
, hdr
->pad2
, 512 -
818 offsetof(struct ddf_header
, pad2
)) != 0) {
819 pr_err("%s: header mismatch\n", __func__
);
823 /* Looks good enough to me... */
827 static void *load_section(int fd
, struct ddf_super
*super
, void *buf
,
828 be32 offset_be
, be32 len_be
, int check
)
830 unsigned long long offset
= be32_to_cpu(offset_be
);
831 unsigned long long len
= be32_to_cpu(len_be
);
832 int dofree
= (buf
== NULL
);
835 if (len
!= 2 && len
!= 8 && len
!= 32
836 && len
!= 128 && len
!= 512)
841 if (!buf
&& posix_memalign(&buf
, 512, len
<<9) != 0)
847 if (super
->active
->type
== 1)
848 offset
+= be64_to_cpu(super
->active
->primary_lba
);
850 offset
+= be64_to_cpu(super
->active
->secondary_lba
);
852 if ((unsigned long long)lseek64(fd
, offset
<<9, 0) != (offset
<<9)) {
857 if ((unsigned long long)read(fd
, buf
, len
<<9) != (len
<<9)) {
865 static int load_ddf_headers(int fd
, struct ddf_super
*super
, char *devname
)
867 unsigned long long dsize
;
869 get_dev_size(fd
, NULL
, &dsize
);
871 if (lseek64(fd
, dsize
-512, 0) < 0) {
873 pr_err("Cannot seek to anchor block on %s: %s\n",
874 devname
, strerror(errno
));
877 if (read(fd
, &super
->anchor
, 512) != 512) {
879 pr_err("Cannot read anchor block on %s: %s\n",
880 devname
, strerror(errno
));
883 if (!be32_eq(super
->anchor
.magic
, DDF_HEADER_MAGIC
)) {
885 pr_err("no DDF anchor found on %s\n",
889 if (!be32_eq(calc_crc(&super
->anchor
, 512), super
->anchor
.crc
)) {
891 pr_err("bad CRC on anchor on %s\n",
895 if (memcmp(super
->anchor
.revision
, DDF_REVISION_0
, 8) != 0 &&
896 memcmp(super
->anchor
.revision
, DDF_REVISION_2
, 8) != 0) {
898 pr_err("can only support super revision"
899 " %.8s and earlier, not %.8s on %s\n",
900 DDF_REVISION_2
, super
->anchor
.revision
,devname
);
903 super
->active
= NULL
;
904 if (load_ddf_header(fd
, be64_to_cpu(super
->anchor
.primary_lba
),
906 &super
->primary
, &super
->anchor
) == 0) {
908 pr_err("Failed to load primary DDF header "
911 super
->active
= &super
->primary
;
913 if (load_ddf_header(fd
, be64_to_cpu(super
->anchor
.secondary_lba
),
915 &super
->secondary
, &super
->anchor
)) {
916 if (super
->active
== NULL
917 || (be32_to_cpu(super
->primary
.seq
)
918 < be32_to_cpu(super
->secondary
.seq
) &&
919 !super
->secondary
.openflag
)
920 || (be32_to_cpu(super
->primary
.seq
)
921 == be32_to_cpu(super
->secondary
.seq
) &&
922 super
->primary
.openflag
&& !super
->secondary
.openflag
)
924 super
->active
= &super
->secondary
;
925 } else if (devname
&&
926 be64_to_cpu(super
->anchor
.secondary_lba
) != ~(__u64
)0)
927 pr_err("Failed to load secondary DDF header on %s\n",
929 if (super
->active
== NULL
)
934 static int load_ddf_global(int fd
, struct ddf_super
*super
, char *devname
)
937 ok
= load_section(fd
, super
, &super
->controller
,
938 super
->active
->controller_section_offset
,
939 super
->active
->controller_section_length
,
941 super
->phys
= load_section(fd
, super
, NULL
,
942 super
->active
->phys_section_offset
,
943 super
->active
->phys_section_length
,
945 super
->pdsize
= be32_to_cpu(super
->active
->phys_section_length
) * 512;
947 super
->virt
= load_section(fd
, super
, NULL
,
948 super
->active
->virt_section_offset
,
949 super
->active
->virt_section_length
,
951 super
->vdsize
= be32_to_cpu(super
->active
->virt_section_length
) * 512;
961 super
->conflist
= NULL
;
964 super
->max_part
= be16_to_cpu(super
->active
->max_partitions
);
965 super
->mppe
= be16_to_cpu(super
->active
->max_primary_element_entries
);
966 super
->conf_rec_len
= be16_to_cpu(super
->active
->config_record_len
);
970 #define DDF_UNUSED_BVD 0xff
971 static int alloc_other_bvds(const struct ddf_super
*ddf
, struct vcl
*vcl
)
973 unsigned int n_vds
= vcl
->conf
.sec_elmnt_count
- 1;
974 unsigned int i
, vdsize
;
977 vcl
->other_bvds
= NULL
;
980 vdsize
= ddf
->conf_rec_len
* 512;
981 if (posix_memalign(&p
, 512, n_vds
*
982 (vdsize
+ sizeof(struct vd_config
*))) != 0)
984 vcl
->other_bvds
= (struct vd_config
**) (p
+ n_vds
* vdsize
);
985 for (i
= 0; i
< n_vds
; i
++) {
986 vcl
->other_bvds
[i
] = p
+ i
* vdsize
;
987 memset(vcl
->other_bvds
[i
], 0, vdsize
);
988 vcl
->other_bvds
[i
]->sec_elmnt_seq
= DDF_UNUSED_BVD
;
993 static void add_other_bvd(struct vcl
*vcl
, struct vd_config
*vd
,
997 for (i
= 0; i
< vcl
->conf
.sec_elmnt_count
-1; i
++)
998 if (vcl
->other_bvds
[i
]->sec_elmnt_seq
== vd
->sec_elmnt_seq
)
1001 if (i
< vcl
->conf
.sec_elmnt_count
-1) {
1002 if (be32_to_cpu(vd
->seqnum
) <=
1003 be32_to_cpu(vcl
->other_bvds
[i
]->seqnum
))
1006 for (i
= 0; i
< vcl
->conf
.sec_elmnt_count
-1; i
++)
1007 if (vcl
->other_bvds
[i
]->sec_elmnt_seq
== DDF_UNUSED_BVD
)
1009 if (i
== vcl
->conf
.sec_elmnt_count
-1) {
1010 pr_err("no space for sec level config %u, count is %u\n",
1011 vd
->sec_elmnt_seq
, vcl
->conf
.sec_elmnt_count
);
1015 memcpy(vcl
->other_bvds
[i
], vd
, len
);
1018 static int load_ddf_local(int fd
, struct ddf_super
*super
,
1019 char *devname
, int keep
)
1025 unsigned int confsec
;
1027 unsigned int max_virt_disks
=
1028 be16_to_cpu(super
->active
->max_vd_entries
);
1029 unsigned long long dsize
;
1031 /* First the local disk info */
1032 if (posix_memalign((void**)&dl
, 512,
1034 (super
->max_part
) * sizeof(dl
->vlist
[0])) != 0) {
1035 pr_err("%s could not allocate disk info buffer\n",
1040 load_section(fd
, super
, &dl
->disk
,
1041 super
->active
->data_section_offset
,
1042 super
->active
->data_section_length
,
1044 dl
->devname
= devname
? xstrdup(devname
) : NULL
;
1047 dl
->major
= major(stb
.st_rdev
);
1048 dl
->minor
= minor(stb
.st_rdev
);
1049 dl
->next
= super
->dlist
;
1050 dl
->fd
= keep
? fd
: -1;
1053 if (get_dev_size(fd
, devname
, &dsize
))
1054 dl
->size
= dsize
>> 9;
1055 /* If the disks have different sizes, the LBAs will differ
1056 * between phys disks.
1057 * At this point here, the values in super->active must be valid
1058 * for this phys disk. */
1059 dl
->primary_lba
= super
->active
->primary_lba
;
1060 dl
->secondary_lba
= super
->active
->secondary_lba
;
1061 dl
->workspace_lba
= super
->active
->workspace_lba
;
1063 for (i
= 0 ; i
< super
->max_part
; i
++)
1064 dl
->vlist
[i
] = NULL
;
1067 for (i
= 0; i
< be16_to_cpu(super
->active
->max_pd_entries
); i
++)
1068 if (memcmp(super
->phys
->entries
[i
].guid
,
1069 dl
->disk
.guid
, DDF_GUID_LEN
) == 0)
1072 /* Now the config list. */
1073 /* 'conf' is an array of config entries, some of which are
1074 * probably invalid. Those which are good need to be copied into
1078 conf
= load_section(fd
, super
, super
->conf
,
1079 super
->active
->config_section_offset
,
1080 super
->active
->config_section_length
,
1085 confsec
< be32_to_cpu(super
->active
->config_section_length
);
1086 confsec
+= super
->conf_rec_len
) {
1087 struct vd_config
*vd
=
1088 (struct vd_config
*)((char*)conf
+ confsec
*512);
1091 if (be32_eq(vd
->magic
, DDF_SPARE_ASSIGN_MAGIC
)) {
1094 if (posix_memalign((void**)&dl
->spare
, 512,
1095 super
->conf_rec_len
*512) != 0) {
1096 pr_err("%s could not allocate spare info buf\n",
1101 memcpy(dl
->spare
, vd
, super
->conf_rec_len
*512);
1104 if (!be32_eq(vd
->magic
, DDF_VD_CONF_MAGIC
))
1105 /* Must be vendor-unique - I cannot handle those */
1108 for (vcl
= super
->conflist
; vcl
; vcl
= vcl
->next
) {
1109 if (memcmp(vcl
->conf
.guid
,
1110 vd
->guid
, DDF_GUID_LEN
) == 0)
1115 dl
->vlist
[vnum
++] = vcl
;
1116 if (vcl
->other_bvds
!= NULL
&&
1117 vcl
->conf
.sec_elmnt_seq
!= vd
->sec_elmnt_seq
) {
1118 add_other_bvd(vcl
, vd
, super
->conf_rec_len
*512);
1121 if (be32_to_cpu(vd
->seqnum
) <=
1122 be32_to_cpu(vcl
->conf
.seqnum
))
1125 if (posix_memalign((void**)&vcl
, 512,
1126 (super
->conf_rec_len
*512 +
1127 offsetof(struct vcl
, conf
))) != 0) {
1128 pr_err("%s could not allocate vcl buf\n",
1132 vcl
->next
= super
->conflist
;
1133 vcl
->block_sizes
= NULL
; /* FIXME not for CONCAT */
1134 vcl
->conf
.sec_elmnt_count
= vd
->sec_elmnt_count
;
1135 if (alloc_other_bvds(super
, vcl
) != 0) {
1136 pr_err("%s could not allocate other bvds\n",
1141 super
->conflist
= vcl
;
1142 dl
->vlist
[vnum
++] = vcl
;
1144 memcpy(&vcl
->conf
, vd
, super
->conf_rec_len
*512);
1145 for (i
=0; i
< max_virt_disks
; i
++)
1146 if (memcmp(super
->virt
->entries
[i
].guid
,
1147 vcl
->conf
.guid
, DDF_GUID_LEN
)==0)
1149 if (i
< max_virt_disks
)
1156 static int load_super_ddf(struct supertype
*st
, int fd
,
1159 unsigned long long dsize
;
1160 struct ddf_super
*super
;
1163 if (get_dev_size(fd
, devname
, &dsize
) == 0)
1166 if (test_partition(fd
))
1167 /* DDF is not allowed on partitions */
1170 /* 32M is a lower bound */
1171 if (dsize
<= 32*1024*1024) {
1173 pr_err("%s is too small for ddf: "
1174 "size is %llu sectors.\n",
1180 pr_err("%s is an odd size for ddf: "
1181 "size is %llu bytes.\n",
1188 if (posix_memalign((void**)&super
, 512, sizeof(*super
))!= 0) {
1189 pr_err("malloc of %zu failed.\n",
1193 memset(super
, 0, sizeof(*super
));
1195 rv
= load_ddf_headers(fd
, super
, devname
);
1201 /* Have valid headers and have chosen the best. Let's read in the rest*/
1203 rv
= load_ddf_global(fd
, super
, devname
);
1207 pr_err("Failed to load all information "
1208 "sections on %s\n", devname
);
1213 rv
= load_ddf_local(fd
, super
, devname
, 0);
1217 pr_err("Failed to load all information "
1218 "sections on %s\n", devname
);
1223 /* Should possibly check the sections .... */
1226 if (st
->ss
== NULL
) {
1227 st
->ss
= &super_ddf
;
1228 st
->minor_version
= 0;
1235 static void free_super_ddf(struct supertype
*st
)
1237 struct ddf_super
*ddf
= st
->sb
;
1243 while (ddf
->conflist
) {
1244 struct vcl
*v
= ddf
->conflist
;
1245 ddf
->conflist
= v
->next
;
1247 free(v
->block_sizes
);
1250 v->other_bvds[0] points to beginning of buffer,
1251 see alloc_other_bvds()
1253 free(v
->other_bvds
[0]);
1256 while (ddf
->dlist
) {
1257 struct dl
*d
= ddf
->dlist
;
1258 ddf
->dlist
= d
->next
;
1265 while (ddf
->add_list
) {
1266 struct dl
*d
= ddf
->add_list
;
1267 ddf
->add_list
= d
->next
;
1278 static struct supertype
*match_metadata_desc_ddf(char *arg
)
1280 /* 'ddf' only supports containers */
1281 struct supertype
*st
;
1282 if (strcmp(arg
, "ddf") != 0 &&
1283 strcmp(arg
, "default") != 0
1287 st
= xcalloc(1, sizeof(*st
));
1288 st
->ss
= &super_ddf
;
1290 st
->minor_version
= 0;
1297 static mapping_t ddf_state
[] = {
1303 { "Partially Optimal", 5},
1309 static mapping_t ddf_init_state
[] = {
1310 { "Not Initialised", 0},
1311 { "QuickInit in Progress", 1},
1312 { "Fully Initialised", 2},
1316 static mapping_t ddf_access
[] = {
1320 { "Blocked (no access)", 3},
1324 static mapping_t ddf_level
[] = {
1325 { "RAID0", DDF_RAID0
},
1326 { "RAID1", DDF_RAID1
},
1327 { "RAID3", DDF_RAID3
},
1328 { "RAID4", DDF_RAID4
},
1329 { "RAID5", DDF_RAID5
},
1330 { "RAID1E",DDF_RAID1E
},
1331 { "JBOD", DDF_JBOD
},
1332 { "CONCAT",DDF_CONCAT
},
1333 { "RAID5E",DDF_RAID5E
},
1334 { "RAID5EE",DDF_RAID5EE
},
1335 { "RAID6", DDF_RAID6
},
1338 static mapping_t ddf_sec_level
[] = {
1339 { "Striped", DDF_2STRIPED
},
1340 { "Mirrored", DDF_2MIRRORED
},
1341 { "Concat", DDF_2CONCAT
},
1342 { "Spanned", DDF_2SPANNED
},
1347 static int all_ff(const char *guid
)
1350 for (i
= 0; i
< DDF_GUID_LEN
; i
++)
1351 if (guid
[i
] != (char)0xff)
1356 static const char *guid_str(const char *guid
)
1358 static char buf
[DDF_GUID_LEN
*2+1];
1361 for (i
= 0; i
< DDF_GUID_LEN
; i
++) {
1362 unsigned char c
= guid
[i
];
1363 if (c
>= 32 && c
< 127)
1364 p
+= sprintf(p
, "%c", c
);
1366 p
+= sprintf(p
, "%02x", c
);
1369 return (const char *) buf
;
1373 static void print_guid(char *guid
, int tstamp
)
1375 /* A GUIDs are part (or all) ASCII and part binary.
1376 * They tend to be space padded.
1377 * We print the GUID in HEX, then in parentheses add
1378 * any initial ASCII sequence, and a possible
1379 * time stamp from bytes 16-19
1381 int l
= DDF_GUID_LEN
;
1384 for (i
=0 ; i
<DDF_GUID_LEN
; i
++) {
1385 if ((i
&3)==0 && i
!= 0) printf(":");
1386 printf("%02X", guid
[i
]&255);
1390 while (l
&& guid
[l
-1] == ' ')
1392 for (i
=0 ; i
<l
; i
++) {
1393 if (guid
[i
] >= 0x20 && guid
[i
] < 0x7f)
1394 fputc(guid
[i
], stdout
);
1399 time_t then
= __be32_to_cpu(*(__u32
*)(guid
+16)) + DECADE
;
1402 tm
= localtime(&then
);
1403 strftime(tbuf
, 100, " %D %T",tm
);
1404 fputs(tbuf
, stdout
);
1409 static void examine_vd(int n
, struct ddf_super
*sb
, char *guid
)
1411 int crl
= sb
->conf_rec_len
;
1414 for (vcl
= sb
->conflist
; vcl
; vcl
= vcl
->next
) {
1416 struct vd_config
*vc
= &vcl
->conf
;
1418 if (!be32_eq(calc_crc(vc
, crl
*512), vc
->crc
))
1420 if (memcmp(vc
->guid
, guid
, DDF_GUID_LEN
) != 0)
1423 /* Ok, we know about this VD, let's give more details */
1424 printf(" Raid Devices[%d] : %d (", n
,
1425 be16_to_cpu(vc
->prim_elmnt_count
));
1426 for (i
= 0; i
< be16_to_cpu(vc
->prim_elmnt_count
); i
++) {
1428 int cnt
= be16_to_cpu(sb
->phys
->max_pdes
);
1429 for (j
=0; j
<cnt
; j
++)
1430 if (be32_eq(vc
->phys_refnum
[i
],
1431 sb
->phys
->entries
[j
].refnum
))
1440 if (vc
->chunk_shift
!= 255)
1441 printf(" Chunk Size[%d] : %d sectors\n", n
,
1442 1 << vc
->chunk_shift
);
1443 printf(" Raid Level[%d] : %s\n", n
,
1444 map_num(ddf_level
, vc
->prl
)?:"-unknown-");
1445 if (vc
->sec_elmnt_count
!= 1) {
1446 printf(" Secondary Position[%d] : %d of %d\n", n
,
1447 vc
->sec_elmnt_seq
, vc
->sec_elmnt_count
);
1448 printf(" Secondary Level[%d] : %s\n", n
,
1449 map_num(ddf_sec_level
, vc
->srl
) ?: "-unknown-");
1451 printf(" Device Size[%d] : %llu\n", n
,
1452 be64_to_cpu(vc
->blocks
)/2);
1453 printf(" Array Size[%d] : %llu\n", n
,
1454 be64_to_cpu(vc
->array_blocks
)/2);
1458 static void examine_vds(struct ddf_super
*sb
)
1460 int cnt
= be16_to_cpu(sb
->virt
->populated_vdes
);
1462 printf(" Virtual Disks : %d\n", cnt
);
1464 for (i
= 0; i
< be16_to_cpu(sb
->virt
->max_vdes
); i
++) {
1465 struct virtual_entry
*ve
= &sb
->virt
->entries
[i
];
1466 if (all_ff(ve
->guid
))
1469 printf(" VD GUID[%d] : ", i
); print_guid(ve
->guid
, 1);
1471 printf(" unit[%d] : %d\n", i
, be16_to_cpu(ve
->unit
));
1472 printf(" state[%d] : %s, %s%s\n", i
,
1473 map_num(ddf_state
, ve
->state
& 7),
1474 (ve
->state
& DDF_state_morphing
) ? "Morphing, ": "",
1475 (ve
->state
& DDF_state_inconsistent
)? "Not Consistent" : "Consistent");
1476 printf(" init state[%d] : %s\n", i
,
1477 map_num(ddf_init_state
, ve
->init_state
&DDF_initstate_mask
));
1478 printf(" access[%d] : %s\n", i
,
1479 map_num(ddf_access
, (ve
->init_state
& DDF_access_mask
) >> 6));
1480 printf(" Name[%d] : %.16s\n", i
, ve
->name
);
1481 examine_vd(i
, sb
, ve
->guid
);
1483 if (cnt
) printf("\n");
1486 static void examine_pds(struct ddf_super
*sb
)
1488 int cnt
= be16_to_cpu(sb
->phys
->max_pdes
);
1491 printf(" Physical Disks : %d\n", cnt
);
1492 printf(" Number RefNo Size Device Type/State\n");
1494 for (i
=0 ; i
<cnt
; i
++) {
1495 struct phys_disk_entry
*pd
= &sb
->phys
->entries
[i
];
1496 int type
= be16_to_cpu(pd
->type
);
1497 int state
= be16_to_cpu(pd
->state
);
1499 if (be32_to_cpu(pd
->refnum
) == 0xffffffff)
1502 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1504 printf(" %3d %08x ", i
,
1505 be32_to_cpu(pd
->refnum
));
1507 be64_to_cpu(pd
->config_size
)>>1);
1508 for (dl
= sb
->dlist
; dl
; dl
= dl
->next
) {
1509 if (be32_eq(dl
->disk
.refnum
, pd
->refnum
)) {
1510 char *dv
= map_dev(dl
->major
, dl
->minor
, 0);
1512 printf("%-15s", dv
);
1519 printf(" %s%s%s%s%s",
1520 (type
&2) ? "active":"",
1521 (type
&4) ? "Global-Spare":"",
1522 (type
&8) ? "spare" : "",
1523 (type
&16)? ", foreign" : "",
1524 (type
&32)? "pass-through" : "");
1525 if (state
& DDF_Failed
)
1526 /* This over-rides these three */
1527 state
&= ~(DDF_Online
|DDF_Rebuilding
|DDF_Transition
);
1528 printf("/%s%s%s%s%s%s%s",
1529 (state
&1)? "Online": "Offline",
1530 (state
&2)? ", Failed": "",
1531 (state
&4)? ", Rebuilding": "",
1532 (state
&8)? ", in-transition": "",
1533 (state
&16)? ", SMART-errors": "",
1534 (state
&32)? ", Unrecovered-Read-Errors": "",
1535 (state
&64)? ", Missing" : "");
1540 static void examine_super_ddf(struct supertype
*st
, char *homehost
)
1542 struct ddf_super
*sb
= st
->sb
;
1544 printf(" Magic : %08x\n", be32_to_cpu(sb
->anchor
.magic
));
1545 printf(" Version : %.8s\n", sb
->anchor
.revision
);
1546 printf("Controller GUID : "); print_guid(sb
->controller
.guid
, 0);
1548 printf(" Container GUID : "); print_guid(sb
->anchor
.guid
, 1);
1550 printf(" Seq : %08x\n", be32_to_cpu(sb
->active
->seq
));
1551 printf(" Redundant hdr : %s\n", (be32_eq(sb
->secondary
.magic
,
1558 static unsigned int get_vd_num_of_subarray(struct supertype
*st
)
1561 * Figure out the VD number for this supertype.
1562 * Returns DDF_CONTAINER for the container itself,
1563 * and DDF_NOTFOUND on error.
1565 struct ddf_super
*ddf
= st
->sb
;
1570 if (*st
->container_devnm
== '\0')
1571 return DDF_CONTAINER
;
1573 sra
= sysfs_read(-1, st
->devnm
, GET_VERSION
);
1574 if (!sra
|| sra
->array
.major_version
!= -1 ||
1575 sra
->array
.minor_version
!= -2 ||
1576 !is_subarray(sra
->text_version
))
1577 return DDF_NOTFOUND
;
1579 sub
= strchr(sra
->text_version
+ 1, '/');
1581 vcnum
= strtoul(sub
+ 1, &end
, 10);
1582 if (sub
== NULL
|| *sub
== '\0' || *end
!= '\0' ||
1583 vcnum
>= be16_to_cpu(ddf
->active
->max_vd_entries
))
1584 return DDF_NOTFOUND
;
1589 static void brief_examine_super_ddf(struct supertype
*st
, int verbose
)
1591 /* We just write a generic DDF ARRAY entry
1595 getinfo_super_ddf(st
, &info
, NULL
);
1596 fname_from_uuid(st
, &info
, nbuf
, ':');
1598 printf("ARRAY metadata=ddf UUID=%s\n", nbuf
+ 5);
1601 static void brief_examine_subarrays_ddf(struct supertype
*st
, int verbose
)
1603 /* We write a DDF ARRAY member entry for each vd, identifying container
1604 * by uuid and member by unit number and uuid.
1606 struct ddf_super
*ddf
= st
->sb
;
1610 getinfo_super_ddf(st
, &info
, NULL
);
1611 fname_from_uuid(st
, &info
, nbuf
, ':');
1613 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++) {
1614 struct virtual_entry
*ve
= &ddf
->virt
->entries
[i
];
1618 if (all_ff(ve
->guid
))
1620 memcpy(vcl
.conf
.guid
, ve
->guid
, DDF_GUID_LEN
);
1621 ddf
->currentconf
=&vcl
;
1623 uuid_from_super_ddf(st
, info
.uuid
);
1624 fname_from_uuid(st
, &info
, nbuf1
, ':');
1625 _ddf_array_name(namebuf
, ddf
, i
);
1626 printf("ARRAY%s%s container=%s member=%d UUID=%s\n",
1627 namebuf
[0] == '\0' ? "" : " /dev/md/", namebuf
,
1628 nbuf
+5, i
, nbuf1
+5);
1632 static void export_examine_super_ddf(struct supertype
*st
)
1636 getinfo_super_ddf(st
, &info
, NULL
);
1637 fname_from_uuid(st
, &info
, nbuf
, ':');
1638 printf("MD_METADATA=ddf\n");
1639 printf("MD_LEVEL=container\n");
1640 printf("MD_UUID=%s\n", nbuf
+5);
1641 printf("MD_DEVICES=%u\n",
1642 be16_to_cpu(((struct ddf_super
*)st
->sb
)->phys
->used_pdes
));
1645 static int copy_metadata_ddf(struct supertype
*st
, int from
, int to
)
1648 unsigned long long dsize
, offset
;
1650 struct ddf_header
*ddf
;
1653 /* The meta consists of an anchor, a primary, and a secondary.
1654 * This all lives at the end of the device.
1655 * So it is easiest to find the earliest of primary and
1656 * secondary, and copy everything from there.
1658 * Anchor is 512 from end. It contains primary_lba and secondary_lba
1659 * we choose one of those
1662 if (posix_memalign(&buf
, 4096, 4096) != 0)
1665 if (!get_dev_size(from
, NULL
, &dsize
))
1668 if (lseek64(from
, dsize
-512, 0) < 0)
1670 if (read(from
, buf
, 512) != 512)
1673 if (!be32_eq(ddf
->magic
, DDF_HEADER_MAGIC
) ||
1674 !be32_eq(calc_crc(ddf
, 512), ddf
->crc
) ||
1675 (memcmp(ddf
->revision
, DDF_REVISION_0
, 8) != 0 &&
1676 memcmp(ddf
->revision
, DDF_REVISION_2
, 8) != 0))
1679 offset
= dsize
- 512;
1680 if ((be64_to_cpu(ddf
->primary_lba
) << 9) < offset
)
1681 offset
= be64_to_cpu(ddf
->primary_lba
) << 9;
1682 if ((be64_to_cpu(ddf
->secondary_lba
) << 9) < offset
)
1683 offset
= be64_to_cpu(ddf
->secondary_lba
) << 9;
1685 bytes
= dsize
- offset
;
1687 if (lseek64(from
, offset
, 0) < 0 ||
1688 lseek64(to
, offset
, 0) < 0)
1690 while (written
< bytes
) {
1691 int n
= bytes
- written
;
1694 if (read(from
, buf
, n
) != n
)
1696 if (write(to
, buf
, n
) != n
)
1707 static void detail_super_ddf(struct supertype
*st
, char *homehost
)
1710 * Could print DDF GUID
1711 * Need to find which array
1712 * If whole, briefly list all arrays
1717 static const char *vendors_with_variable_volume_UUID
[] = {
1721 static int volume_id_is_reliable(const struct ddf_super
*ddf
)
1723 int n
= ARRAY_SIZE(vendors_with_variable_volume_UUID
);
1725 for (i
= 0; i
< n
; i
++)
1726 if (!memcmp(ddf
->controller
.guid
,
1727 vendors_with_variable_volume_UUID
[i
], 8))
1732 static void uuid_of_ddf_subarray(const struct ddf_super
*ddf
,
1733 unsigned int vcnum
, int uuid
[4])
1735 char buf
[DDF_GUID_LEN
+18], sha
[20], *p
;
1736 struct sha1_ctx ctx
;
1737 if (volume_id_is_reliable(ddf
)) {
1738 uuid_from_ddf_guid(ddf
->virt
->entries
[vcnum
].guid
, uuid
);
1742 * Some fake RAID BIOSes (in particular, LSI ones) change the
1743 * VD GUID at every boot. These GUIDs are not suitable for
1744 * identifying an array. Luckily the header GUID appears to
1746 * We construct a pseudo-UUID from the header GUID and those
1747 * properties of the subarray that we expect to remain constant.
1749 memset(buf
, 0, sizeof(buf
));
1751 memcpy(p
, ddf
->anchor
.guid
, DDF_GUID_LEN
);
1753 memcpy(p
, ddf
->virt
->entries
[vcnum
].name
, 16);
1755 *((__u16
*) p
) = vcnum
;
1756 sha1_init_ctx(&ctx
);
1757 sha1_process_bytes(buf
, sizeof(buf
), &ctx
);
1758 sha1_finish_ctx(&ctx
, sha
);
1759 memcpy(uuid
, sha
, 4*4);
1762 static void brief_detail_super_ddf(struct supertype
*st
)
1766 struct ddf_super
*ddf
= st
->sb
;
1767 unsigned int vcnum
= get_vd_num_of_subarray(st
);
1768 if (vcnum
== DDF_CONTAINER
)
1769 uuid_from_super_ddf(st
, info
.uuid
);
1770 else if (vcnum
== DDF_NOTFOUND
)
1773 uuid_of_ddf_subarray(ddf
, vcnum
, info
.uuid
);
1774 fname_from_uuid(st
, &info
, nbuf
,':');
1775 printf(" UUID=%s", nbuf
+ 5);
1779 static int match_home_ddf(struct supertype
*st
, char *homehost
)
1781 /* It matches 'this' host if the controller is a
1782 * Linux-MD controller with vendor_data matching
1783 * the hostname. It would be nice if we could
1784 * test against controller found in /sys or somewhere...
1786 struct ddf_super
*ddf
= st
->sb
;
1791 len
= strlen(homehost
);
1793 return (memcmp(ddf
->controller
.guid
, T10
, 8) == 0 &&
1794 len
< sizeof(ddf
->controller
.vendor_data
) &&
1795 memcmp(ddf
->controller
.vendor_data
, homehost
,len
) == 0 &&
1796 ddf
->controller
.vendor_data
[len
] == 0);
1800 static int find_index_in_bvd(const struct ddf_super
*ddf
,
1801 const struct vd_config
*conf
, unsigned int n
,
1802 unsigned int *n_bvd
)
1805 * Find the index of the n-th valid physical disk in this BVD.
1806 * Unused entries can be sprinkled in with the used entries,
1811 i
< ddf
->mppe
&& j
< be16_to_cpu(conf
->prim_elmnt_count
);
1813 if (be32_to_cpu(conf
->phys_refnum
[i
]) != 0xffffffff) {
1821 dprintf("%s: couldn't find BVD member %u (total %u)\n",
1822 __func__
, n
, be16_to_cpu(conf
->prim_elmnt_count
));
1826 /* Given a member array instance number, and a raid disk within that instance,
1827 * find the vd_config structure. The offset of the given disk in the phys_refnum
1828 * table is returned in n_bvd.
1829 * For two-level members with a secondary raid level the vd_config for
1830 * the appropriate BVD is returned.
1831 * The return value is always &vlc->conf, where vlc is returned in last pointer.
1833 static struct vd_config
*find_vdcr(struct ddf_super
*ddf
, unsigned int inst
,
1835 unsigned int *n_bvd
, struct vcl
**vcl
)
1839 for (v
= ddf
->conflist
; v
; v
= v
->next
) {
1840 unsigned int nsec
, ibvd
= 0;
1841 struct vd_config
*conf
;
1842 if (inst
!= v
->vcnum
)
1845 if (conf
->sec_elmnt_count
== 1) {
1846 if (find_index_in_bvd(ddf
, conf
, n
, n_bvd
)) {
1852 if (v
->other_bvds
== NULL
) {
1853 pr_err("%s: BUG: other_bvds is NULL, nsec=%u\n",
1854 __func__
, conf
->sec_elmnt_count
);
1857 nsec
= n
/ be16_to_cpu(conf
->prim_elmnt_count
);
1858 if (conf
->sec_elmnt_seq
!= nsec
) {
1859 for (ibvd
= 1; ibvd
< conf
->sec_elmnt_count
; ibvd
++) {
1860 if (v
->other_bvds
[ibvd
-1]->sec_elmnt_seq
1864 if (ibvd
== conf
->sec_elmnt_count
)
1866 conf
= v
->other_bvds
[ibvd
-1];
1868 if (!find_index_in_bvd(ddf
, conf
,
1869 n
- nsec
*conf
->sec_elmnt_count
, n_bvd
))
1871 dprintf("%s: found disk %u as member %u in bvd %d of array %u\n"
1872 , __func__
, n
, *n_bvd
, ibvd
, inst
);
1877 pr_err("%s: Could't find disk %d in array %u\n", __func__
, n
, inst
);
1882 static int find_phys(const struct ddf_super
*ddf
, be32 phys_refnum
)
1884 /* Find the entry in phys_disk which has the given refnum
1885 * and return it's index
1888 for (i
= 0; i
< be16_to_cpu(ddf
->phys
->max_pdes
); i
++)
1889 if (be32_eq(ddf
->phys
->entries
[i
].refnum
, phys_refnum
))
1894 static void uuid_from_ddf_guid(const char *guid
, int uuid
[4])
1897 struct sha1_ctx ctx
;
1898 sha1_init_ctx(&ctx
);
1899 sha1_process_bytes(guid
, DDF_GUID_LEN
, &ctx
);
1900 sha1_finish_ctx(&ctx
, buf
);
1901 memcpy(uuid
, buf
, 4*4);
1904 static void uuid_from_super_ddf(struct supertype
*st
, int uuid
[4])
1906 /* The uuid returned here is used for:
1907 * uuid to put into bitmap file (Create, Grow)
1908 * uuid for backup header when saving critical section (Grow)
1909 * comparing uuids when re-adding a device into an array
1910 * In these cases the uuid required is that of the data-array,
1911 * not the device-set.
1912 * uuid to recognise same set when adding a missing device back
1913 * to an array. This is a uuid for the device-set.
1915 * For each of these we can make do with a truncated
1916 * or hashed uuid rather than the original, as long as
1918 * In the case of SVD we assume the BVD is of interest,
1919 * though that might be the case if a bitmap were made for
1920 * a mirrored SVD - worry about that later.
1921 * So we need to find the VD configuration record for the
1922 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1923 * The first 16 bytes of the sha1 of these is used.
1925 struct ddf_super
*ddf
= st
->sb
;
1926 struct vcl
*vcl
= ddf
->currentconf
;
1929 uuid_of_ddf_subarray(ddf
, vcl
->vcnum
, uuid
);
1931 uuid_from_ddf_guid(ddf
->anchor
.guid
, uuid
);
1934 static void getinfo_super_ddf(struct supertype
*st
, struct mdinfo
*info
, char *map
)
1936 struct ddf_super
*ddf
= st
->sb
;
1937 int map_disks
= info
->array
.raid_disks
;
1940 if (ddf
->currentconf
) {
1941 getinfo_super_ddf_bvd(st
, info
, map
);
1944 memset(info
, 0, sizeof(*info
));
1946 info
->array
.raid_disks
= be16_to_cpu(ddf
->phys
->used_pdes
);
1947 info
->array
.level
= LEVEL_CONTAINER
;
1948 info
->array
.layout
= 0;
1949 info
->array
.md_minor
= -1;
1950 cptr
= (__u32
*)(ddf
->anchor
.guid
+ 16);
1951 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*cptr
);
1953 info
->array
.utime
= 0;
1954 info
->array
.chunk_size
= 0;
1955 info
->container_enough
= 1;
1957 info
->disk
.major
= 0;
1958 info
->disk
.minor
= 0;
1960 struct phys_disk_entry
*pde
= NULL
;
1961 info
->disk
.number
= be32_to_cpu(ddf
->dlist
->disk
.refnum
);
1962 info
->disk
.raid_disk
= find_phys(ddf
, ddf
->dlist
->disk
.refnum
);
1964 info
->data_offset
= be64_to_cpu(ddf
->phys
->
1965 entries
[info
->disk
.raid_disk
].
1967 info
->component_size
= ddf
->dlist
->size
- info
->data_offset
;
1968 if (info
->disk
.raid_disk
>= 0)
1969 pde
= ddf
->phys
->entries
+ info
->disk
.raid_disk
;
1971 !(be16_to_cpu(pde
->state
) & DDF_Failed
))
1972 info
->disk
.state
= (1 << MD_DISK_SYNC
) | (1 << MD_DISK_ACTIVE
);
1974 info
->disk
.state
= 1 << MD_DISK_FAULTY
;
1976 info
->events
= be32_to_cpu(ddf
->active
->seq
);
1978 info
->disk
.number
= -1;
1979 info
->disk
.raid_disk
= -1;
1980 // info->disk.raid_disk = find refnum in the table and use index;
1981 info
->disk
.state
= (1 << MD_DISK_SYNC
) | (1 << MD_DISK_ACTIVE
);
1984 info
->recovery_start
= MaxSector
;
1985 info
->reshape_active
= 0;
1986 info
->recovery_blocked
= 0;
1989 info
->array
.major_version
= -1;
1990 info
->array
.minor_version
= -2;
1991 strcpy(info
->text_version
, "ddf");
1992 info
->safe_mode_delay
= 0;
1994 uuid_from_super_ddf(st
, info
->uuid
);
1998 for (i
= 0 ; i
< map_disks
; i
++) {
1999 if (i
< info
->array
.raid_disks
&&
2000 !(be16_to_cpu(ddf
->phys
->entries
[i
].state
)
2009 /* size of name must be at least 17 bytes! */
2010 static void _ddf_array_name(char *name
, const struct ddf_super
*ddf
, int i
)
2013 memcpy(name
, ddf
->virt
->entries
[i
].name
, 16);
2015 for(j
= 0; j
< 16; j
++)
2020 static void getinfo_super_ddf_bvd(struct supertype
*st
, struct mdinfo
*info
, char *map
)
2022 struct ddf_super
*ddf
= st
->sb
;
2023 struct vcl
*vc
= ddf
->currentconf
;
2024 int cd
= ddf
->currentdev
;
2028 int map_disks
= info
->array
.raid_disks
;
2030 struct vd_config
*conf
;
2032 memset(info
, 0, sizeof(*info
));
2033 if (layout_ddf2md(&vc
->conf
, &info
->array
) == -1)
2035 info
->array
.md_minor
= -1;
2036 cptr
= (__u32
*)(vc
->conf
.guid
+ 16);
2037 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*cptr
);
2038 info
->array
.utime
= DECADE
+ be32_to_cpu(vc
->conf
.timestamp
);
2039 info
->array
.chunk_size
= 512 << vc
->conf
.chunk_shift
;
2040 info
->custom_array_size
= 0;
2043 n_prim
= be16_to_cpu(conf
->prim_elmnt_count
);
2044 if (conf
->sec_elmnt_count
> 1 && cd
>= n_prim
) {
2045 int ibvd
= cd
/ n_prim
- 1;
2047 conf
= vc
->other_bvds
[ibvd
];
2050 if (cd
>= 0 && (unsigned)cd
< ddf
->mppe
) {
2052 be64_to_cpu(LBA_OFFSET(ddf
, conf
)[cd
]);
2053 if (vc
->block_sizes
)
2054 info
->component_size
= vc
->block_sizes
[cd
];
2056 info
->component_size
= be64_to_cpu(conf
->blocks
);
2059 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2060 if (be32_eq(dl
->disk
.refnum
, conf
->phys_refnum
[cd
]))
2063 info
->disk
.major
= 0;
2064 info
->disk
.minor
= 0;
2065 info
->disk
.state
= 0;
2066 if (dl
&& dl
->pdnum
>= 0) {
2067 info
->disk
.major
= dl
->major
;
2068 info
->disk
.minor
= dl
->minor
;
2069 info
->disk
.raid_disk
= cd
+ conf
->sec_elmnt_seq
2070 * be16_to_cpu(conf
->prim_elmnt_count
);
2071 info
->disk
.number
= dl
->pdnum
;
2072 info
->disk
.state
= 0;
2073 if (info
->disk
.number
>= 0 &&
2074 (be16_to_cpu(ddf
->phys
->entries
[info
->disk
.number
].state
) & DDF_Online
) &&
2075 !(be16_to_cpu(ddf
->phys
->entries
[info
->disk
.number
].state
) & DDF_Failed
))
2076 info
->disk
.state
= (1<<MD_DISK_SYNC
)|(1<<MD_DISK_ACTIVE
);
2077 info
->events
= be32_to_cpu(ddf
->active
->seq
);
2080 info
->container_member
= ddf
->currentconf
->vcnum
;
2082 info
->recovery_start
= MaxSector
;
2083 info
->resync_start
= 0;
2084 info
->reshape_active
= 0;
2085 info
->recovery_blocked
= 0;
2086 if (!(ddf
->virt
->entries
[info
->container_member
].state
2087 & DDF_state_inconsistent
) &&
2088 (ddf
->virt
->entries
[info
->container_member
].init_state
2089 & DDF_initstate_mask
)
2091 info
->resync_start
= MaxSector
;
2093 uuid_from_super_ddf(st
, info
->uuid
);
2095 info
->array
.major_version
= -1;
2096 info
->array
.minor_version
= -2;
2097 sprintf(info
->text_version
, "/%s/%d",
2098 st
->container_devnm
,
2099 info
->container_member
);
2100 info
->safe_mode_delay
= DDF_SAFE_MODE_DELAY
;
2102 _ddf_array_name(info
->name
, ddf
, info
->container_member
);
2105 for (j
= 0; j
< map_disks
; j
++) {
2107 if (j
< info
->array
.raid_disks
) {
2108 int i
= find_phys(ddf
, vc
->conf
.phys_refnum
[j
]);
2110 (be16_to_cpu(ddf
->phys
->entries
[i
].state
)
2112 !(be16_to_cpu(ddf
->phys
->entries
[i
].state
)
2119 static int update_super_ddf(struct supertype
*st
, struct mdinfo
*info
,
2121 char *devname
, int verbose
,
2122 int uuid_set
, char *homehost
)
2124 /* For 'assemble' and 'force' we need to return non-zero if any
2125 * change was made. For others, the return value is ignored.
2126 * Update options are:
2127 * force-one : This device looks a bit old but needs to be included,
2128 * update age info appropriately.
2129 * assemble: clear any 'faulty' flag to allow this device to
2131 * force-array: Array is degraded but being forced, mark it clean
2132 * if that will be needed to assemble it.
2134 * newdev: not used ????
2135 * grow: Array has gained a new device - this is currently for
2137 * resync: mark as dirty so a resync will happen.
2138 * uuid: Change the uuid of the array to match what is given
2139 * homehost: update the recorded homehost
2140 * name: update the name - preserving the homehost
2141 * _reshape_progress: record new reshape_progress position.
2143 * Following are not relevant for this version:
2144 * sparc2.2 : update from old dodgey metadata
2145 * super-minor: change the preferred_minor number
2146 * summaries: update redundant counters.
2149 // struct ddf_super *ddf = st->sb;
2150 // struct vd_config *vd = find_vdcr(ddf, info->container_member);
2151 // struct virtual_entry *ve = find_ve(ddf);
2153 /* we don't need to handle "force-*" or "assemble" as
2154 * there is no need to 'trick' the kernel. When the metadata is
2155 * first updated to activate the array, all the implied modifications
2159 if (strcmp(update
, "grow") == 0) {
2161 } else if (strcmp(update
, "resync") == 0) {
2162 // info->resync_checkpoint = 0;
2163 } else if (strcmp(update
, "homehost") == 0) {
2164 /* homehost is stored in controller->vendor_data,
2165 * or it is when we are the vendor
2167 // if (info->vendor_is_local)
2168 // strcpy(ddf->controller.vendor_data, homehost);
2170 } else if (strcmp(update
, "name") == 0) {
2171 /* name is stored in virtual_entry->name */
2172 // memset(ve->name, ' ', 16);
2173 // strncpy(ve->name, info->name, 16);
2175 } else if (strcmp(update
, "_reshape_progress") == 0) {
2176 /* We don't support reshape yet */
2177 } else if (strcmp(update
, "assemble") == 0 ) {
2178 /* Do nothing, just succeed */
2183 // update_all_csum(ddf);
2188 static void make_header_guid(char *guid
)
2191 /* Create a DDF Header of Virtual Disk GUID */
2193 /* 24 bytes of fiction required.
2194 * first 8 are a 'vendor-id' - "Linux-MD"
2195 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
2196 * Remaining 8 random number plus timestamp
2198 memcpy(guid
, T10
, sizeof(T10
));
2199 stamp
= cpu_to_be32(0xdeadbeef);
2200 memcpy(guid
+8, &stamp
, 4);
2201 stamp
= cpu_to_be32(0);
2202 memcpy(guid
+12, &stamp
, 4);
2203 stamp
= cpu_to_be32(time(0) - DECADE
);
2204 memcpy(guid
+16, &stamp
, 4);
2205 stamp
._v32
= random32();
2206 memcpy(guid
+20, &stamp
, 4);
2209 static unsigned int find_unused_vde(const struct ddf_super
*ddf
)
2212 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++) {
2213 if (all_ff(ddf
->virt
->entries
[i
].guid
))
2216 return DDF_NOTFOUND
;
2219 static unsigned int find_vde_by_name(const struct ddf_super
*ddf
,
2224 return DDF_NOTFOUND
;
2225 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++) {
2226 if (all_ff(ddf
->virt
->entries
[i
].guid
))
2228 if (!strncmp(name
, ddf
->virt
->entries
[i
].name
,
2229 sizeof(ddf
->virt
->entries
[i
].name
)))
2232 return DDF_NOTFOUND
;
2236 static unsigned int find_vde_by_guid(const struct ddf_super
*ddf
,
2240 if (guid
== NULL
|| all_ff(guid
))
2241 return DDF_NOTFOUND
;
2242 for (i
= 0; i
< be16_to_cpu(ddf
->virt
->max_vdes
); i
++)
2243 if (!memcmp(ddf
->virt
->entries
[i
].guid
, guid
, DDF_GUID_LEN
))
2245 return DDF_NOTFOUND
;
2249 static int init_super_ddf(struct supertype
*st
,
2250 mdu_array_info_t
*info
,
2251 unsigned long long size
, char *name
, char *homehost
,
2252 int *uuid
, unsigned long long data_offset
)
2254 /* This is primarily called by Create when creating a new array.
2255 * We will then get add_to_super called for each component, and then
2256 * write_init_super called to write it out to each device.
2257 * For DDF, Create can create on fresh devices or on a pre-existing
2259 * To create on a pre-existing array a different method will be called.
2260 * This one is just for fresh drives.
2262 * We need to create the entire 'ddf' structure which includes:
2263 * DDF headers - these are easy.
2264 * Controller data - a Sector describing this controller .. not that
2265 * this is a controller exactly.
2266 * Physical Disk Record - one entry per device, so
2267 * leave plenty of space.
2268 * Virtual Disk Records - again, just leave plenty of space.
2269 * This just lists VDs, doesn't give details.
2270 * Config records - describe the VDs that use this disk
2271 * DiskData - describes 'this' device.
2272 * BadBlockManagement - empty
2273 * Diag Space - empty
2274 * Vendor Logs - Could we put bitmaps here?
2277 struct ddf_super
*ddf
;
2280 int max_phys_disks
, max_virt_disks
;
2281 unsigned long long sector
;
2285 struct phys_disk
*pd
;
2286 struct virtual_disk
*vd
;
2288 if (data_offset
!= INVALID_SECTORS
) {
2289 pr_err("data-offset not supported by DDF\n");
2294 return init_super_ddf_bvd(st
, info
, size
, name
, homehost
, uuid
,
2297 if (posix_memalign((void**)&ddf
, 512, sizeof(*ddf
)) != 0) {
2298 pr_err("%s could not allocate superblock\n", __func__
);
2301 memset(ddf
, 0, sizeof(*ddf
));
2305 /* zeroing superblock */
2309 /* At least 32MB *must* be reserved for the ddf. So let's just
2310 * start 32MB from the end, and put the primary header there.
2311 * Don't do secondary for now.
2312 * We don't know exactly where that will be yet as it could be
2313 * different on each device. So just set up the lengths.
2316 ddf
->anchor
.magic
= DDF_HEADER_MAGIC
;
2317 make_header_guid(ddf
->anchor
.guid
);
2319 memcpy(ddf
->anchor
.revision
, DDF_REVISION_2
, 8);
2320 ddf
->anchor
.seq
= cpu_to_be32(1);
2321 ddf
->anchor
.timestamp
= cpu_to_be32(time(0) - DECADE
);
2322 ddf
->anchor
.openflag
= 0xFF;
2323 ddf
->anchor
.foreignflag
= 0;
2324 ddf
->anchor
.enforcegroups
= 0; /* Is this best?? */
2325 ddf
->anchor
.pad0
= 0xff;
2326 memset(ddf
->anchor
.pad1
, 0xff, 12);
2327 memset(ddf
->anchor
.header_ext
, 0xff, 32);
2328 ddf
->anchor
.primary_lba
= cpu_to_be64(~(__u64
)0);
2329 ddf
->anchor
.secondary_lba
= cpu_to_be64(~(__u64
)0);
2330 ddf
->anchor
.type
= DDF_HEADER_ANCHOR
;
2331 memset(ddf
->anchor
.pad2
, 0xff, 3);
2332 ddf
->anchor
.workspace_len
= cpu_to_be32(32768); /* Must be reserved */
2333 /* Put this at bottom of 32M reserved.. */
2334 ddf
->anchor
.workspace_lba
= cpu_to_be64(~(__u64
)0);
2335 max_phys_disks
= 1023; /* Should be enough, 4095 is also allowed */
2336 ddf
->anchor
.max_pd_entries
= cpu_to_be16(max_phys_disks
);
2337 max_virt_disks
= 255; /* 15, 63, 255, 1024, 4095 are all allowed */
2338 ddf
->anchor
.max_vd_entries
= cpu_to_be16(max_virt_disks
);
2340 ddf
->anchor
.max_partitions
= cpu_to_be16(ddf
->max_part
);
2341 ddf
->mppe
= 256; /* 16, 64, 256, 1024, 4096 are all allowed */
2342 ddf
->conf_rec_len
= 1 + ROUND_UP(ddf
->mppe
* (4+8), 512)/512;
2343 ddf
->anchor
.config_record_len
= cpu_to_be16(ddf
->conf_rec_len
);
2344 ddf
->anchor
.max_primary_element_entries
= cpu_to_be16(ddf
->mppe
);
2345 memset(ddf
->anchor
.pad3
, 0xff, 54);
2346 /* Controller section is one sector long immediately
2347 * after the ddf header */
2349 ddf
->anchor
.controller_section_offset
= cpu_to_be32(sector
);
2350 ddf
->anchor
.controller_section_length
= cpu_to_be32(1);
2353 /* phys is 8 sectors after that */
2354 pdsize
= ROUND_UP(sizeof(struct phys_disk
) +
2355 sizeof(struct phys_disk_entry
)*max_phys_disks
,
2357 switch(pdsize
/512) {
2358 case 2: case 8: case 32: case 128: case 512: break;
2361 ddf
->anchor
.phys_section_offset
= cpu_to_be32(sector
);
2362 ddf
->anchor
.phys_section_length
=
2363 cpu_to_be32(pdsize
/512); /* max_primary_element_entries/8 */
2364 sector
+= pdsize
/512;
2366 /* virt is another 32 sectors */
2367 vdsize
= ROUND_UP(sizeof(struct virtual_disk
) +
2368 sizeof(struct virtual_entry
) * max_virt_disks
,
2370 switch(vdsize
/512) {
2371 case 2: case 8: case 32: case 128: case 512: break;
2374 ddf
->anchor
.virt_section_offset
= cpu_to_be32(sector
);
2375 ddf
->anchor
.virt_section_length
=
2376 cpu_to_be32(vdsize
/512); /* max_vd_entries/8 */
2377 sector
+= vdsize
/512;
2379 clen
= ddf
->conf_rec_len
* (ddf
->max_part
+1);
2380 ddf
->anchor
.config_section_offset
= cpu_to_be32(sector
);
2381 ddf
->anchor
.config_section_length
= cpu_to_be32(clen
);
2384 ddf
->anchor
.data_section_offset
= cpu_to_be32(sector
);
2385 ddf
->anchor
.data_section_length
= cpu_to_be32(1);
2388 ddf
->anchor
.bbm_section_length
= cpu_to_be32(0);
2389 ddf
->anchor
.bbm_section_offset
= cpu_to_be32(0xFFFFFFFF);
2390 ddf
->anchor
.diag_space_length
= cpu_to_be32(0);
2391 ddf
->anchor
.diag_space_offset
= cpu_to_be32(0xFFFFFFFF);
2392 ddf
->anchor
.vendor_length
= cpu_to_be32(0);
2393 ddf
->anchor
.vendor_offset
= cpu_to_be32(0xFFFFFFFF);
2395 memset(ddf
->anchor
.pad4
, 0xff, 256);
2397 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
2398 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
2400 ddf
->primary
.openflag
= 1; /* I guess.. */
2401 ddf
->primary
.type
= DDF_HEADER_PRIMARY
;
2403 ddf
->secondary
.openflag
= 1; /* I guess.. */
2404 ddf
->secondary
.type
= DDF_HEADER_SECONDARY
;
2406 ddf
->active
= &ddf
->primary
;
2408 ddf
->controller
.magic
= DDF_CONTROLLER_MAGIC
;
2410 /* 24 more bytes of fiction required.
2411 * first 8 are a 'vendor-id' - "Linux-MD"
2412 * Remaining 16 are serial number.... maybe a hostname would do?
2414 memcpy(ddf
->controller
.guid
, T10
, sizeof(T10
));
2415 gethostname(hostname
, sizeof(hostname
));
2416 hostname
[sizeof(hostname
) - 1] = 0;
2417 hostlen
= strlen(hostname
);
2418 memcpy(ddf
->controller
.guid
+ 24 - hostlen
, hostname
, hostlen
);
2419 for (i
= strlen(T10
) ; i
+hostlen
< 24; i
++)
2420 ddf
->controller
.guid
[i
] = ' ';
2422 ddf
->controller
.type
.vendor_id
= cpu_to_be16(0xDEAD);
2423 ddf
->controller
.type
.device_id
= cpu_to_be16(0xBEEF);
2424 ddf
->controller
.type
.sub_vendor_id
= cpu_to_be16(0);
2425 ddf
->controller
.type
.sub_device_id
= cpu_to_be16(0);
2426 memcpy(ddf
->controller
.product_id
, "What Is My PID??", 16);
2427 memset(ddf
->controller
.pad
, 0xff, 8);
2428 memset(ddf
->controller
.vendor_data
, 0xff, 448);
2429 if (homehost
&& strlen(homehost
) < 440)
2430 strcpy((char*)ddf
->controller
.vendor_data
, homehost
);
2432 if (posix_memalign((void**)&pd
, 512, pdsize
) != 0) {
2433 pr_err("%s could not allocate pd\n", __func__
);
2437 ddf
->pdsize
= pdsize
;
2439 memset(pd
, 0xff, pdsize
);
2440 memset(pd
, 0, sizeof(*pd
));
2441 pd
->magic
= DDF_PHYS_RECORDS_MAGIC
;
2442 pd
->used_pdes
= cpu_to_be16(0);
2443 pd
->max_pdes
= cpu_to_be16(max_phys_disks
);
2444 memset(pd
->pad
, 0xff, 52);
2445 for (i
= 0; i
< max_phys_disks
; i
++)
2446 memset(pd
->entries
[i
].guid
, 0xff, DDF_GUID_LEN
);
2448 if (posix_memalign((void**)&vd
, 512, vdsize
) != 0) {
2449 pr_err("%s could not allocate vd\n", __func__
);
2453 ddf
->vdsize
= vdsize
;
2454 memset(vd
, 0, vdsize
);
2455 vd
->magic
= DDF_VIRT_RECORDS_MAGIC
;
2456 vd
->populated_vdes
= cpu_to_be16(0);
2457 vd
->max_vdes
= cpu_to_be16(max_virt_disks
);
2458 memset(vd
->pad
, 0xff, 52);
2460 for (i
=0; i
<max_virt_disks
; i
++)
2461 memset(&vd
->entries
[i
], 0xff, sizeof(struct virtual_entry
));
2464 ddf_set_updates_pending(ddf
);
2468 static int chunk_to_shift(int chunksize
)
2470 return ffs(chunksize
/512)-1;
2475 unsigned long long start
, size
;
2477 static int cmp_extent(const void *av
, const void *bv
)
2479 const struct extent
*a
= av
;
2480 const struct extent
*b
= bv
;
2481 if (a
->start
< b
->start
)
2483 if (a
->start
> b
->start
)
2488 static struct extent
*get_extents(struct ddf_super
*ddf
, struct dl
*dl
)
2490 /* Find a list of used extents on the give physical device
2491 * (dnum) of the given ddf.
2492 * Return a malloced array of 'struct extent'
2501 state
= be16_to_cpu(ddf
->phys
->entries
[dl
->pdnum
].state
);
2503 if ((state
& (DDF_Online
|DDF_Failed
|DDF_Missing
)) != DDF_Online
)
2506 rv
= xmalloc(sizeof(struct extent
) * (ddf
->max_part
+ 2));
2508 for (i
= 0; i
< ddf
->max_part
; i
++) {
2509 const struct vd_config
*bvd
;
2511 struct vcl
*v
= dl
->vlist
[i
];
2513 get_pd_index_from_refnum(v
, dl
->disk
.refnum
, ddf
->mppe
,
2514 &bvd
, &ibvd
) == DDF_NOTFOUND
)
2516 rv
[n
].start
= be64_to_cpu(LBA_OFFSET(ddf
, bvd
)[ibvd
]);
2517 rv
[n
].size
= be64_to_cpu(bvd
->blocks
);
2520 qsort(rv
, n
, sizeof(*rv
), cmp_extent
);
2522 rv
[n
].start
= be64_to_cpu(ddf
->phys
->entries
[dl
->pdnum
].config_size
);
2528 static int init_super_ddf_bvd(struct supertype
*st
,
2529 mdu_array_info_t
*info
,
2530 unsigned long long size
,
2531 char *name
, char *homehost
,
2532 int *uuid
, unsigned long long data_offset
)
2534 /* We are creating a BVD inside a pre-existing container.
2535 * so st->sb is already set.
2536 * We need to create a new vd_config and a new virtual_entry
2538 struct ddf_super
*ddf
= st
->sb
;
2539 unsigned int venum
, i
;
2540 struct virtual_entry
*ve
;
2542 struct vd_config
*vc
;
2544 if (find_vde_by_name(ddf
, name
) != DDF_NOTFOUND
) {
2545 pr_err("This ddf already has an array called %s\n", name
);
2548 venum
= find_unused_vde(ddf
);
2549 if (venum
== DDF_NOTFOUND
) {
2550 pr_err("Cannot find spare slot for virtual disk\n");
2553 ve
= &ddf
->virt
->entries
[venum
];
2555 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
2556 * timestamp, random number
2558 make_header_guid(ve
->guid
);
2559 ve
->unit
= cpu_to_be16(info
->md_minor
);
2561 ve
->guid_crc
._v16
= crc32(0, (unsigned char *)ddf
->anchor
.guid
,
2563 ve
->type
= cpu_to_be16(0);
2564 ve
->state
= DDF_state_degraded
; /* Will be modified as devices are added */
2565 if (info
->state
& 1) /* clean */
2566 ve
->init_state
= DDF_init_full
;
2568 ve
->init_state
= DDF_init_not
;
2570 memset(ve
->pad1
, 0xff, 14);
2571 memset(ve
->name
, ' ', 16);
2573 strncpy(ve
->name
, name
, 16);
2574 ddf
->virt
->populated_vdes
=
2575 cpu_to_be16(be16_to_cpu(ddf
->virt
->populated_vdes
)+1);
2577 /* Now create a new vd_config */
2578 if (posix_memalign((void**)&vcl
, 512,
2579 (offsetof(struct vcl
, conf
) + ddf
->conf_rec_len
* 512)) != 0) {
2580 pr_err("%s could not allocate vd_config\n", __func__
);
2584 vcl
->block_sizes
= NULL
; /* FIXME not for CONCAT */
2587 vc
->magic
= DDF_VD_CONF_MAGIC
;
2588 memcpy(vc
->guid
, ve
->guid
, DDF_GUID_LEN
);
2589 vc
->timestamp
= cpu_to_be32(time(0)-DECADE
);
2590 vc
->seqnum
= cpu_to_be32(1);
2591 memset(vc
->pad0
, 0xff, 24);
2592 vc
->chunk_shift
= chunk_to_shift(info
->chunk_size
);
2593 if (layout_md2ddf(info
, vc
) == -1 ||
2594 be16_to_cpu(vc
->prim_elmnt_count
) > ddf
->mppe
) {
2595 pr_err("%s: unsupported RAID level/layout %d/%d with %d disks\n",
2596 __func__
, info
->level
, info
->layout
, info
->raid_disks
);
2600 vc
->sec_elmnt_seq
= 0;
2601 if (alloc_other_bvds(ddf
, vcl
) != 0) {
2602 pr_err("%s could not allocate other bvds\n",
2607 vc
->blocks
= cpu_to_be64(info
->size
* 2);
2608 vc
->array_blocks
= cpu_to_be64(
2609 calc_array_size(info
->level
, info
->raid_disks
, info
->layout
,
2610 info
->chunk_size
, info
->size
*2));
2611 memset(vc
->pad1
, 0xff, 8);
2612 vc
->spare_refs
[0] = cpu_to_be32(0xffffffff);
2613 vc
->spare_refs
[1] = cpu_to_be32(0xffffffff);
2614 vc
->spare_refs
[2] = cpu_to_be32(0xffffffff);
2615 vc
->spare_refs
[3] = cpu_to_be32(0xffffffff);
2616 vc
->spare_refs
[4] = cpu_to_be32(0xffffffff);
2617 vc
->spare_refs
[5] = cpu_to_be32(0xffffffff);
2618 vc
->spare_refs
[6] = cpu_to_be32(0xffffffff);
2619 vc
->spare_refs
[7] = cpu_to_be32(0xffffffff);
2620 memset(vc
->cache_pol
, 0, 8);
2622 memset(vc
->pad2
, 0xff, 3);
2623 memset(vc
->pad3
, 0xff, 52);
2624 memset(vc
->pad4
, 0xff, 192);
2625 memset(vc
->v0
, 0xff, 32);
2626 memset(vc
->v1
, 0xff, 32);
2627 memset(vc
->v2
, 0xff, 16);
2628 memset(vc
->v3
, 0xff, 16);
2629 memset(vc
->vendor
, 0xff, 32);
2631 memset(vc
->phys_refnum
, 0xff, 4*ddf
->mppe
);
2632 memset(vc
->phys_refnum
+ddf
->mppe
, 0x00, 8*ddf
->mppe
);
2634 for (i
= 1; i
< vc
->sec_elmnt_count
; i
++) {
2635 memcpy(vcl
->other_bvds
[i
-1], vc
, ddf
->conf_rec_len
* 512);
2636 vcl
->other_bvds
[i
-1]->sec_elmnt_seq
= i
;
2639 vcl
->next
= ddf
->conflist
;
2640 ddf
->conflist
= vcl
;
2641 ddf
->currentconf
= vcl
;
2642 ddf_set_updates_pending(ddf
);
2647 static void add_to_super_ddf_bvd(struct supertype
*st
,
2648 mdu_disk_info_t
*dk
, int fd
, char *devname
)
2650 /* fd and devname identify a device within the ddf container (st).
2651 * dk identifies a location in the new BVD.
2652 * We need to find suitable free space in that device and update
2653 * the phys_refnum and lba_offset for the newly created vd_config.
2654 * We might also want to update the type in the phys_disk
2657 * Alternately: fd == -1 and we have already chosen which device to
2658 * use and recorded in dlist->raid_disk;
2661 struct ddf_super
*ddf
= st
->sb
;
2662 struct vd_config
*vc
;
2664 unsigned long long blocks
, pos
, esize
;
2666 unsigned int raid_disk
= dk
->raid_disk
;
2669 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2670 if (dl
->raiddisk
== dk
->raid_disk
)
2673 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2674 if (dl
->major
== dk
->major
&&
2675 dl
->minor
== dk
->minor
)
2678 if (!dl
|| dl
->pdnum
< 0 || ! (dk
->state
& (1<<MD_DISK_SYNC
)))
2681 vc
= &ddf
->currentconf
->conf
;
2682 if (vc
->sec_elmnt_count
> 1) {
2683 unsigned int n
= be16_to_cpu(vc
->prim_elmnt_count
);
2685 vc
= ddf
->currentconf
->other_bvds
[raid_disk
/ n
- 1];
2689 ex
= get_extents(ddf
, dl
);
2694 blocks
= be64_to_cpu(vc
->blocks
);
2695 if (ddf
->currentconf
->block_sizes
)
2696 blocks
= ddf
->currentconf
->block_sizes
[dk
->raid_disk
];
2700 esize
= ex
[i
].start
- pos
;
2701 if (esize
>= blocks
)
2703 pos
= ex
[i
].start
+ ex
[i
].size
;
2705 } while (ex
[i
-1].size
);
2711 ddf
->currentdev
= dk
->raid_disk
;
2712 vc
->phys_refnum
[raid_disk
] = dl
->disk
.refnum
;
2713 LBA_OFFSET(ddf
, vc
)[raid_disk
] = cpu_to_be64(pos
);
2715 for (i
= 0; i
< ddf
->max_part
; i
++)
2716 if (dl
->vlist
[i
] == NULL
)
2718 if (i
== ddf
->max_part
)
2720 dl
->vlist
[i
] = ddf
->currentconf
;
2725 dl
->devname
= devname
;
2727 /* Check if we can mark array as optimal yet */
2728 i
= ddf
->currentconf
->vcnum
;
2729 ddf
->virt
->entries
[i
].state
=
2730 (ddf
->virt
->entries
[i
].state
& ~DDF_state_mask
)
2731 | get_svd_state(ddf
, ddf
->currentconf
);
2732 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
2733 cpu_to_be16(DDF_Global_Spare
));
2734 be16_set(ddf
->phys
->entries
[dl
->pdnum
].type
,
2735 cpu_to_be16(DDF_Active_in_VD
));
2736 dprintf("%s: added disk %d/%08x to VD %d/%s as disk %d\n",
2737 __func__
, dl
->pdnum
, be32_to_cpu(dl
->disk
.refnum
),
2738 ddf
->currentconf
->vcnum
, guid_str(vc
->guid
),
2740 ddf_set_updates_pending(ddf
);
2743 static unsigned int find_unused_pde(const struct ddf_super
*ddf
)
2746 for (i
= 0; i
< be16_to_cpu(ddf
->phys
->max_pdes
); i
++) {
2747 if (all_ff(ddf
->phys
->entries
[i
].guid
))
2750 return DDF_NOTFOUND
;
2753 static void _set_config_size(struct phys_disk_entry
*pde
, const struct dl
*dl
)
2756 cfs
= min(dl
->size
- 32*1024*2ULL, be64_to_cpu(dl
->primary_lba
));
2757 t
= be64_to_cpu(dl
->secondary_lba
);
2761 * Some vendor DDF structures interpret workspace_lba
2762 * very differently than we do: Make a sanity check on the value.
2764 t
= be64_to_cpu(dl
->workspace_lba
);
2766 __u64 wsp
= cfs
- t
;
2767 if (wsp
> 1024*1024*2ULL && wsp
> dl
->size
/ 16) {
2768 pr_err("%s: %x:%x: workspace size 0x%llx too big, ignoring\n",
2769 __func__
, dl
->major
, dl
->minor
, wsp
);
2773 pde
->config_size
= cpu_to_be64(cfs
);
2774 dprintf("%s: %x:%x config_size %llx, DDF structure is %llx blocks\n",
2775 __func__
, dl
->major
, dl
->minor
, cfs
, dl
->size
-cfs
);
2778 /* Add a device to a container, either while creating it or while
2779 * expanding a pre-existing container
2781 static int add_to_super_ddf(struct supertype
*st
,
2782 mdu_disk_info_t
*dk
, int fd
, char *devname
,
2783 unsigned long long data_offset
)
2785 struct ddf_super
*ddf
= st
->sb
;
2789 unsigned long long size
;
2790 struct phys_disk_entry
*pde
;
2795 if (ddf
->currentconf
) {
2796 add_to_super_ddf_bvd(st
, dk
, fd
, devname
);
2800 /* This is device numbered dk->number. We need to create
2801 * a phys_disk entry and a more detailed disk_data entry.
2804 n
= find_unused_pde(ddf
);
2805 if (n
== DDF_NOTFOUND
) {
2806 pr_err("%s: No free slot in array, cannot add disk\n",
2810 pde
= &ddf
->phys
->entries
[n
];
2811 get_dev_size(fd
, NULL
, &size
);
2812 if (size
<= 32*1024*1024) {
2813 pr_err("%s: device size must be at least 32MB\n",
2819 if (posix_memalign((void**)&dd
, 512,
2820 sizeof(*dd
) + sizeof(dd
->vlist
[0]) * ddf
->max_part
) != 0) {
2821 pr_err("%s could allocate buffer for new disk, aborting\n",
2825 dd
->major
= major(stb
.st_rdev
);
2826 dd
->minor
= minor(stb
.st_rdev
);
2827 dd
->devname
= devname
;
2831 dd
->disk
.magic
= DDF_PHYS_DATA_MAGIC
;
2833 tm
= localtime(&now
);
2834 sprintf(dd
->disk
.guid
, "%8s%04d%02d%02d",
2835 T10
, tm
->tm_year
+1900, tm
->tm_mon
+1, tm
->tm_mday
);
2836 tptr
= (__u32
*)(dd
->disk
.guid
+ 16);
2837 *tptr
++ = random32();
2841 /* Cannot be bothered finding a CRC of some irrelevant details*/
2842 dd
->disk
.refnum
._v32
= random32();
2843 for (i
= be16_to_cpu(ddf
->active
->max_pd_entries
);
2845 if (be32_eq(ddf
->phys
->entries
[i
-1].refnum
,
2850 dd
->disk
.forced_ref
= 1;
2851 dd
->disk
.forced_guid
= 1;
2852 memset(dd
->disk
.vendor
, ' ', 32);
2853 memcpy(dd
->disk
.vendor
, "Linux", 5);
2854 memset(dd
->disk
.pad
, 0xff, 442);
2855 for (i
= 0; i
< ddf
->max_part
; i
++)
2856 dd
->vlist
[i
] = NULL
;
2860 if (st
->update_tail
) {
2861 int len
= (sizeof(struct phys_disk
) +
2862 sizeof(struct phys_disk_entry
));
2863 struct phys_disk
*pd
;
2866 pd
->magic
= DDF_PHYS_RECORDS_MAGIC
;
2867 pd
->used_pdes
= cpu_to_be16(n
);
2868 pde
= &pd
->entries
[0];
2871 ddf
->phys
->used_pdes
= cpu_to_be16(
2872 1 + be16_to_cpu(ddf
->phys
->used_pdes
));
2874 memcpy(pde
->guid
, dd
->disk
.guid
, DDF_GUID_LEN
);
2875 pde
->refnum
= dd
->disk
.refnum
;
2876 pde
->type
= cpu_to_be16(DDF_Forced_PD_GUID
| DDF_Global_Spare
);
2877 pde
->state
= cpu_to_be16(DDF_Online
);
2880 * If there is already a device in dlist, try to reserve the same
2881 * amount of workspace. Otherwise, use 32MB.
2882 * We checked disk size above already.
2884 #define __calc_lba(new, old, lba, mb) do { \
2885 unsigned long long dif; \
2886 if ((old) != NULL) \
2887 dif = (old)->size - be64_to_cpu((old)->lba); \
2889 dif = (new)->size; \
2890 if ((new)->size > dif) \
2891 (new)->lba = cpu_to_be64((new)->size - dif); \
2893 (new)->lba = cpu_to_be64((new)->size - (mb*1024*2)); \
2895 __calc_lba(dd
, ddf
->dlist
, workspace_lba
, 32);
2896 __calc_lba(dd
, ddf
->dlist
, primary_lba
, 16);
2897 if (ddf
->dlist
== NULL
||
2898 be64_to_cpu(ddf
->dlist
->secondary_lba
) != ~(__u64
)0)
2899 __calc_lba(dd
, ddf
->dlist
, secondary_lba
, 32);
2900 _set_config_size(pde
, dd
);
2902 sprintf(pde
->path
, "%17.17s","Information: nil") ;
2903 memset(pde
->pad
, 0xff, 6);
2905 if (st
->update_tail
) {
2906 dd
->next
= ddf
->add_list
;
2909 dd
->next
= ddf
->dlist
;
2911 ddf_set_updates_pending(ddf
);
2917 static int remove_from_super_ddf(struct supertype
*st
, mdu_disk_info_t
*dk
)
2919 struct ddf_super
*ddf
= st
->sb
;
2922 /* mdmon has noticed that this disk (dk->major/dk->minor) has
2923 * disappeared from the container.
2924 * We need to arrange that it disappears from the metadata and
2925 * internal data structures too.
2926 * Most of the work is done by ddf_process_update which edits
2927 * the metadata and closes the file handle and attaches the memory
2928 * where free_updates will free it.
2930 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2931 if (dl
->major
== dk
->major
&&
2932 dl
->minor
== dk
->minor
)
2934 if (!dl
|| dl
->pdnum
< 0)
2937 if (st
->update_tail
) {
2938 int len
= (sizeof(struct phys_disk
) +
2939 sizeof(struct phys_disk_entry
));
2940 struct phys_disk
*pd
;
2943 pd
->magic
= DDF_PHYS_RECORDS_MAGIC
;
2944 pd
->used_pdes
= cpu_to_be16(dl
->pdnum
);
2945 pd
->entries
[0].state
= cpu_to_be16(DDF_Missing
);
2946 append_metadata_update(st
, pd
, len
);
2953 * This is the write_init_super method for a ddf container. It is
2954 * called when creating a container or adding another device to a
2958 static int __write_ddf_structure(struct dl
*d
, struct ddf_super
*ddf
, __u8 type
)
2960 unsigned long long sector
;
2961 struct ddf_header
*header
;
2962 int fd
, i
, n_config
, conf_size
, buf_size
;
2969 case DDF_HEADER_PRIMARY
:
2970 header
= &ddf
->primary
;
2971 sector
= be64_to_cpu(header
->primary_lba
);
2973 case DDF_HEADER_SECONDARY
:
2974 header
= &ddf
->secondary
;
2975 sector
= be64_to_cpu(header
->secondary_lba
);
2980 if (sector
== ~(__u64
)0)
2983 header
->type
= type
;
2984 header
->openflag
= 1;
2985 header
->crc
= calc_crc(header
, 512);
2987 lseek64(fd
, sector
<<9, 0);
2988 if (write(fd
, header
, 512) < 0)
2991 ddf
->controller
.crc
= calc_crc(&ddf
->controller
, 512);
2992 if (write(fd
, &ddf
->controller
, 512) < 0)
2995 ddf
->phys
->crc
= calc_crc(ddf
->phys
, ddf
->pdsize
);
2996 if (write(fd
, ddf
->phys
, ddf
->pdsize
) < 0)
2998 ddf
->virt
->crc
= calc_crc(ddf
->virt
, ddf
->vdsize
);
2999 if (write(fd
, ddf
->virt
, ddf
->vdsize
) < 0)
3002 /* Now write lots of config records. */
3003 n_config
= ddf
->max_part
;
3004 conf_size
= ddf
->conf_rec_len
* 512;
3006 buf_size
= conf_size
* (n_config
+ 1);
3008 if (posix_memalign((void**)&conf
, 512, buf_size
) != 0)
3012 for (i
= 0 ; i
<= n_config
; i
++) {
3014 struct vd_config
*vdc
= NULL
;
3015 if (i
== n_config
) {
3016 c
= (struct vcl
*)d
->spare
;
3023 get_pd_index_from_refnum(
3026 (const struct vd_config
**)&vdc
,
3030 dprintf("writing conf record %i on disk %08x for %s/%u\n",
3031 i
, be32_to_cpu(d
->disk
.refnum
),
3032 guid_str(vdc
->guid
),
3033 vdc
->sec_elmnt_seq
);
3034 vdc
->seqnum
= header
->seq
;
3035 vdc
->crc
= calc_crc(vdc
, conf_size
);
3036 memcpy(conf
+ i
*conf_size
, vdc
, conf_size
);
3038 memset(conf
+ i
*conf_size
, 0xff, conf_size
);
3040 if (write(fd
, conf
, buf_size
) != buf_size
)
3043 d
->disk
.crc
= calc_crc(&d
->disk
, 512);
3044 if (write(fd
, &d
->disk
, 512) < 0)
3049 header
->openflag
= 0;
3050 header
->crc
= calc_crc(header
, 512);
3052 lseek64(fd
, sector
<<9, 0);
3053 if (write(fd
, header
, 512) < 0)
3059 static int _write_super_to_disk(struct ddf_super
*ddf
, struct dl
*d
)
3061 unsigned long long size
;
3066 /* We need to fill in the primary, (secondary) and workspace
3067 * lba's in the headers, set their checksums,
3068 * Also checksum phys, virt....
3070 * Then write everything out, finally the anchor is written.
3072 get_dev_size(fd
, NULL
, &size
);
3074 if (be64_to_cpu(d
->workspace_lba
) != 0ULL)
3075 ddf
->anchor
.workspace_lba
= d
->workspace_lba
;
3077 ddf
->anchor
.workspace_lba
=
3078 cpu_to_be64(size
- 32*1024*2);
3079 if (be64_to_cpu(d
->primary_lba
) != 0ULL)
3080 ddf
->anchor
.primary_lba
= d
->primary_lba
;
3082 ddf
->anchor
.primary_lba
=
3083 cpu_to_be64(size
- 16*1024*2);
3084 if (be64_to_cpu(d
->secondary_lba
) != 0ULL)
3085 ddf
->anchor
.secondary_lba
= d
->secondary_lba
;
3087 ddf
->anchor
.secondary_lba
=
3088 cpu_to_be64(size
- 32*1024*2);
3089 ddf
->anchor
.seq
= ddf
->active
->seq
;
3090 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
3091 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
3093 ddf
->anchor
.openflag
= 0xFF; /* 'open' means nothing */
3094 ddf
->anchor
.seq
= cpu_to_be32(0xFFFFFFFF); /* no sequencing in anchor */
3095 ddf
->anchor
.crc
= calc_crc(&ddf
->anchor
, 512);
3097 if (!__write_ddf_structure(d
, ddf
, DDF_HEADER_PRIMARY
))
3100 if (!__write_ddf_structure(d
, ddf
, DDF_HEADER_SECONDARY
))
3103 lseek64(fd
, (size
-1)*512, SEEK_SET
);
3104 if (write(fd
, &ddf
->anchor
, 512) < 0)
3111 static int __write_init_super_ddf(struct supertype
*st
)
3113 struct ddf_super
*ddf
= st
->sb
;
3118 pr_state(ddf
, __func__
);
3120 /* try to write updated metadata,
3121 * if we catch a failure move on to the next disk
3123 for (d
= ddf
->dlist
; d
; d
=d
->next
) {
3125 successes
+= _write_super_to_disk(ddf
, d
);
3128 return attempts
!= successes
;
3131 static int write_init_super_ddf(struct supertype
*st
)
3133 struct ddf_super
*ddf
= st
->sb
;
3134 struct vcl
*currentconf
= ddf
->currentconf
;
3136 /* We are done with currentconf - reset it so st refers to the container */
3137 ddf
->currentconf
= NULL
;
3139 if (st
->update_tail
) {
3140 /* queue the virtual_disk and vd_config as metadata updates */
3141 struct virtual_disk
*vd
;
3142 struct vd_config
*vc
;
3147 /* Must be adding a physical disk to the container */
3148 int len
= (sizeof(struct phys_disk
) +
3149 sizeof(struct phys_disk_entry
));
3151 /* adding a disk to the container. */
3155 append_metadata_update(st
, ddf
->add_list
->mdupdate
, len
);
3156 ddf
->add_list
->mdupdate
= NULL
;
3160 /* Newly created VD */
3162 /* First the virtual disk. We have a slightly fake header */
3163 len
= sizeof(struct virtual_disk
) + sizeof(struct virtual_entry
);
3166 vd
->entries
[0] = ddf
->virt
->entries
[currentconf
->vcnum
];
3167 vd
->populated_vdes
= cpu_to_be16(currentconf
->vcnum
);
3168 append_metadata_update(st
, vd
, len
);
3170 /* Then the vd_config */
3171 len
= ddf
->conf_rec_len
* 512;
3172 tlen
= len
* currentconf
->conf
.sec_elmnt_count
;
3174 memcpy(vc
, ¤tconf
->conf
, len
);
3175 for (i
= 1; i
< currentconf
->conf
.sec_elmnt_count
; i
++)
3176 memcpy((char *)vc
+ i
*len
, currentconf
->other_bvds
[i
-1],
3178 append_metadata_update(st
, vc
, tlen
);
3180 /* FIXME I need to close the fds! */
3185 for (d
= ddf
->dlist
; d
; d
=d
->next
)
3186 while (Kill(d
->devname
, NULL
, 0, -1, 1) == 0);
3187 return __write_init_super_ddf(st
);
3193 static __u64
avail_size_ddf(struct supertype
*st
, __u64 devsize
,
3194 unsigned long long data_offset
)
3196 /* We must reserve the last 32Meg */
3197 if (devsize
<= 32*1024*2)
3199 return devsize
- 32*1024*2;
3204 static int reserve_space(struct supertype
*st
, int raiddisks
,
3205 unsigned long long size
, int chunk
,
3206 unsigned long long *freesize
)
3208 /* Find 'raiddisks' spare extents at least 'size' big (but
3209 * only caring about multiples of 'chunk') and remember
3210 * them. If size==0, find the largest size possible.
3211 * Report available size in *freesize
3212 * If space cannot be found, fail.
3215 struct ddf_super
*ddf
= st
->sb
;
3218 for (dl
= ddf
->dlist
; dl
; dl
=dl
->next
) {
3222 /* Now find largest extent on each device */
3223 for (dl
= ddf
->dlist
; dl
; dl
=dl
->next
) {
3224 struct extent
*e
= get_extents(ddf
, dl
);
3225 unsigned long long pos
= 0;
3228 unsigned long long minsize
= size
;
3236 unsigned long long esize
;
3237 esize
= e
[i
].start
- pos
;
3238 if (esize
>= minsize
) {
3242 pos
= e
[i
].start
+ e
[i
].size
;
3244 } while (e
[i
-1].size
);
3247 dl
->esize
= minsize
;
3251 if (cnt
< raiddisks
) {
3252 pr_err("not enough devices with space to create array.\n");
3253 return 0; /* No enough free spaces large enough */
3256 /* choose the largest size of which there are at least 'raiddisk' */
3257 for (dl
= ddf
->dlist
; dl
; dl
=dl
->next
) {
3259 if (dl
->esize
<= size
)
3261 /* This is bigger than 'size', see if there are enough */
3263 for (dl2
= ddf
->dlist
; dl2
; dl2
=dl2
->next
)
3264 if (dl2
->esize
>= dl
->esize
)
3266 if (cnt
>= raiddisks
)
3270 size
= size
/ chunk
;
3275 pr_err("not enough spare devices to create array.\n");
3279 /* We have a 'size' of which there are enough spaces.
3280 * We simply do a first-fit */
3282 for (dl
= ddf
->dlist
; dl
&& cnt
< raiddisks
; dl
=dl
->next
) {
3283 if (dl
->esize
< size
)
3292 static int validate_geometry_ddf(struct supertype
*st
,
3293 int level
, int layout
, int raiddisks
,
3294 int *chunk
, unsigned long long size
,
3295 unsigned long long data_offset
,
3296 char *dev
, unsigned long long *freesize
,
3303 /* ddf potentially supports lots of things, but it depends on
3304 * what devices are offered (and maybe kernel version?)
3305 * If given unused devices, we will make a container.
3306 * If given devices in a container, we will make a BVD.
3307 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
3310 if (*chunk
== UnSet
)
3311 *chunk
= DEFAULT_CHUNK
;
3313 if (level
== LEVEL_NONE
)
3314 level
= LEVEL_CONTAINER
;
3315 if (level
== LEVEL_CONTAINER
) {
3316 /* Must be a fresh device to add to a container */
3317 return validate_geometry_ddf_container(st
, level
, layout
,
3319 size
, data_offset
, dev
,
3325 mdu_array_info_t array
= {
3328 .raid_disks
= raiddisks
3330 struct vd_config conf
;
3331 if (layout_md2ddf(&array
, &conf
) == -1) {
3333 pr_err("DDF does not support level %d /layout %d arrays with %d disks\n",
3334 level
, layout
, raiddisks
);
3337 /* Should check layout? etc */
3339 if (st
->sb
&& freesize
) {
3340 /* --create was given a container to create in.
3341 * So we need to check that there are enough
3342 * free spaces and return the amount of space.
3343 * We may as well remember which drives were
3344 * chosen so that add_to_super/getinfo_super
3347 return reserve_space(st
, raiddisks
, size
, *chunk
, freesize
);
3353 /* A container has already been opened, so we are
3354 * creating in there. Maybe a BVD, maybe an SVD.
3355 * Should make a distinction one day.
3357 return validate_geometry_ddf_bvd(st
, level
, layout
, raiddisks
,
3358 chunk
, size
, data_offset
, dev
,
3362 /* This is the first device for the array.
3363 * If it is a container, we read it in and do automagic allocations,
3364 * no other devices should be given.
3365 * Otherwise it must be a member device of a container, and we
3366 * do manual allocation.
3367 * Later we should check for a BVD and make an SVD.
3369 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
3371 sra
= sysfs_read(fd
, NULL
, GET_VERSION
);
3373 if (sra
&& sra
->array
.major_version
== -1 &&
3374 strcmp(sra
->text_version
, "ddf") == 0) {
3376 /* find space for 'n' devices. */
3377 /* remember the devices */
3378 /* Somehow return the fact that we have enough */
3382 pr_err("ddf: Cannot create this array "
3383 "on device %s - a container is required.\n",
3387 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
3389 pr_err("ddf: Cannot open %s: %s\n",
3390 dev
, strerror(errno
));
3393 /* Well, it is in use by someone, maybe a 'ddf' container. */
3394 cfd
= open_container(fd
);
3398 pr_err("ddf: Cannot use %s: %s\n",
3399 dev
, strerror(EBUSY
));
3402 sra
= sysfs_read(cfd
, NULL
, GET_VERSION
);
3404 if (sra
&& sra
->array
.major_version
== -1 &&
3405 strcmp(sra
->text_version
, "ddf") == 0) {
3406 /* This is a member of a ddf container. Load the container
3407 * and try to create a bvd
3409 struct ddf_super
*ddf
;
3410 if (load_super_ddf_all(st
, cfd
, (void **)&ddf
, NULL
) == 0) {
3412 strcpy(st
->container_devnm
, fd2devnm(cfd
));
3414 return validate_geometry_ddf_bvd(st
, level
, layout
,
3415 raiddisks
, chunk
, size
,
3421 } else /* device may belong to a different container */
3428 validate_geometry_ddf_container(struct supertype
*st
,
3429 int level
, int layout
, int raiddisks
,
3430 int chunk
, unsigned long long size
,
3431 unsigned long long data_offset
,
3432 char *dev
, unsigned long long *freesize
,
3436 unsigned long long ldsize
;
3438 if (level
!= LEVEL_CONTAINER
)
3443 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
3446 pr_err("ddf: Cannot open %s: %s\n",
3447 dev
, strerror(errno
));
3450 if (!get_dev_size(fd
, dev
, &ldsize
)) {
3456 *freesize
= avail_size_ddf(st
, ldsize
>> 9, INVALID_SECTORS
);
3463 static int validate_geometry_ddf_bvd(struct supertype
*st
,
3464 int level
, int layout
, int raiddisks
,
3465 int *chunk
, unsigned long long size
,
3466 unsigned long long data_offset
,
3467 char *dev
, unsigned long long *freesize
,
3471 struct ddf_super
*ddf
= st
->sb
;
3473 unsigned long long pos
= 0;
3474 unsigned long long maxsize
;
3477 /* ddf/bvd supports lots of things, but not containers */
3478 if (level
== LEVEL_CONTAINER
) {
3480 pr_err("DDF cannot create a container within an container\n");
3483 /* We must have the container info already read in. */
3488 /* General test: make sure there is space for
3489 * 'raiddisks' device extents of size 'size'.
3491 unsigned long long minsize
= size
;
3495 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
3500 e
= get_extents(ddf
, dl
);
3503 unsigned long long esize
;
3504 esize
= e
[i
].start
- pos
;
3505 if (esize
>= minsize
)
3507 pos
= e
[i
].start
+ e
[i
].size
;
3509 } while (e
[i
-1].size
);
3514 if (dcnt
< raiddisks
) {
3516 pr_err("ddf: Not enough devices with "
3517 "space for this array (%d < %d)\n",
3523 /* This device must be a member of the set */
3524 if (stat(dev
, &stb
) < 0)
3526 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
3528 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
3529 if (dl
->major
== (int)major(stb
.st_rdev
) &&
3530 dl
->minor
== (int)minor(stb
.st_rdev
))
3535 pr_err("ddf: %s is not in the "
3540 e
= get_extents(ddf
, dl
);
3545 unsigned long long esize
;
3546 esize
= e
[i
].start
- pos
;
3547 if (esize
>= maxsize
)
3549 pos
= e
[i
].start
+ e
[i
].size
;
3551 } while (e
[i
-1].size
);
3552 *freesize
= maxsize
;
3558 static int load_super_ddf_all(struct supertype
*st
, int fd
,
3559 void **sbp
, char *devname
)
3562 struct ddf_super
*super
;
3563 struct mdinfo
*sd
, *best
= NULL
;
3569 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
3572 if (sra
->array
.major_version
!= -1 ||
3573 sra
->array
.minor_version
!= -2 ||
3574 strcmp(sra
->text_version
, "ddf") != 0)
3577 if (posix_memalign((void**)&super
, 512, sizeof(*super
)) != 0)
3579 memset(super
, 0, sizeof(*super
));
3581 /* first, try each device, and choose the best ddf */
3582 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
3584 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3585 dfd
= dev_open(nm
, O_RDONLY
);
3588 rv
= load_ddf_headers(dfd
, super
, NULL
);
3591 seq
= be32_to_cpu(super
->active
->seq
);
3592 if (super
->active
->openflag
)
3594 if (!best
|| seq
> bestseq
) {
3602 /* OK, load this ddf */
3603 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
3604 dfd
= dev_open(nm
, O_RDONLY
);
3607 load_ddf_headers(dfd
, super
, NULL
);
3608 load_ddf_global(dfd
, super
, NULL
);
3610 /* Now we need the device-local bits */
3611 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
3614 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3615 dfd
= dev_open(nm
, O_RDWR
);
3618 rv
= load_ddf_headers(dfd
, super
, NULL
);
3620 rv
= load_ddf_local(dfd
, super
, NULL
, 1);
3626 if (st
->ss
== NULL
) {
3627 st
->ss
= &super_ddf
;
3628 st
->minor_version
= 0;
3631 strcpy(st
->container_devnm
, fd2devnm(fd
));
3635 static int load_container_ddf(struct supertype
*st
, int fd
,
3638 return load_super_ddf_all(st
, fd
, &st
->sb
, devname
);
3641 #endif /* MDASSEMBLE */
3643 static int check_secondary(const struct vcl
*vc
)
3645 const struct vd_config
*conf
= &vc
->conf
;
3648 /* The only DDF secondary RAID level md can support is
3649 * RAID 10, if the stripe sizes and Basic volume sizes
3651 * Other configurations could in theory be supported by exposing
3652 * the BVDs to user space and using device mapper for the secondary
3653 * mapping. So far we don't support that.
3656 __u64 sec_elements
[4] = {0, 0, 0, 0};
3657 #define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63)))
3658 #define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0)
3660 if (vc
->other_bvds
== NULL
) {
3661 pr_err("No BVDs for secondary RAID found\n");
3664 if (conf
->prl
!= DDF_RAID1
) {
3665 pr_err("Secondary RAID level only supported for mirrored BVD\n");
3668 if (conf
->srl
!= DDF_2STRIPED
&& conf
->srl
!= DDF_2SPANNED
) {
3669 pr_err("Secondary RAID level %d is unsupported\n",
3673 __set_sec_seen(conf
->sec_elmnt_seq
);
3674 for (i
= 0; i
< conf
->sec_elmnt_count
-1; i
++) {
3675 const struct vd_config
*bvd
= vc
->other_bvds
[i
];
3676 if (bvd
->sec_elmnt_seq
== DDF_UNUSED_BVD
)
3678 if (bvd
->srl
!= conf
->srl
) {
3679 pr_err("Inconsistent secondary RAID level across BVDs\n");
3682 if (bvd
->prl
!= conf
->prl
) {
3683 pr_err("Different RAID levels for BVDs are unsupported\n");
3686 if (!be16_eq(bvd
->prim_elmnt_count
, conf
->prim_elmnt_count
)) {
3687 pr_err("All BVDs must have the same number of primary elements\n");
3690 if (bvd
->chunk_shift
!= conf
->chunk_shift
) {
3691 pr_err("Different strip sizes for BVDs are unsupported\n");
3694 if (!be64_eq(bvd
->array_blocks
, conf
->array_blocks
)) {
3695 pr_err("Different BVD sizes are unsupported\n");
3698 __set_sec_seen(bvd
->sec_elmnt_seq
);
3700 for (i
= 0; i
< conf
->sec_elmnt_count
; i
++) {
3701 if (!__was_sec_seen(i
)) {
3702 pr_err("BVD %d is missing\n", i
);
3709 static unsigned int get_pd_index_from_refnum(const struct vcl
*vc
,
3710 be32 refnum
, unsigned int nmax
,
3711 const struct vd_config
**bvd
,
3714 unsigned int i
, j
, n
, sec
, cnt
;
3716 cnt
= be16_to_cpu(vc
->conf
.prim_elmnt_count
);
3717 sec
= (vc
->conf
.sec_elmnt_count
== 1 ? 0 : vc
->conf
.sec_elmnt_seq
);
3719 for (i
= 0, j
= 0 ; i
< nmax
; i
++) {
3720 /* j counts valid entries for this BVD */
3721 if (be32_eq(vc
->conf
.phys_refnum
[i
], refnum
)) {
3724 return sec
* cnt
+ j
;
3726 if (be32_to_cpu(vc
->conf
.phys_refnum
[i
]) != 0xffffffff)
3729 if (vc
->other_bvds
== NULL
)
3732 for (n
= 1; n
< vc
->conf
.sec_elmnt_count
; n
++) {
3733 struct vd_config
*vd
= vc
->other_bvds
[n
-1];
3734 sec
= vd
->sec_elmnt_seq
;
3735 if (sec
== DDF_UNUSED_BVD
)
3737 for (i
= 0, j
= 0 ; i
< nmax
; i
++) {
3738 if (be32_eq(vd
->phys_refnum
[i
], refnum
)) {
3741 return sec
* cnt
+ j
;
3743 if (be32_to_cpu(vd
->phys_refnum
[i
]) != 0xffffffff)
3749 return DDF_NOTFOUND
;
3752 static struct mdinfo
*container_content_ddf(struct supertype
*st
, char *subarray
)
3754 /* Given a container loaded by load_super_ddf_all,
3755 * extract information about all the arrays into
3758 * For each vcl in conflist: create an mdinfo, fill it in,
3759 * then look for matching devices (phys_refnum) in dlist
3760 * and create appropriate device mdinfo.
3762 struct ddf_super
*ddf
= st
->sb
;
3763 struct mdinfo
*rest
= NULL
;
3766 for (vc
= ddf
->conflist
; vc
; vc
=vc
->next
) {
3768 struct mdinfo
*this;
3774 (strtoul(subarray
, &ep
, 10) != vc
->vcnum
||
3778 if (vc
->conf
.sec_elmnt_count
> 1) {
3779 if (check_secondary(vc
) != 0)
3783 this = xcalloc(1, sizeof(*this));
3787 if (layout_ddf2md(&vc
->conf
, &this->array
))
3789 this->array
.md_minor
= -1;
3790 this->array
.major_version
= -1;
3791 this->array
.minor_version
= -2;
3792 this->safe_mode_delay
= DDF_SAFE_MODE_DELAY
;
3793 cptr
= (__u32
*)(vc
->conf
.guid
+ 16);
3794 this->array
.ctime
= DECADE
+ __be32_to_cpu(*cptr
);
3795 this->array
.utime
= DECADE
+
3796 be32_to_cpu(vc
->conf
.timestamp
);
3797 this->array
.chunk_size
= 512 << vc
->conf
.chunk_shift
;
3800 if ((ddf
->virt
->entries
[i
].state
& DDF_state_inconsistent
) ||
3801 (ddf
->virt
->entries
[i
].init_state
& DDF_initstate_mask
) !=
3803 this->array
.state
= 0;
3804 this->resync_start
= 0;
3806 this->array
.state
= 1;
3807 this->resync_start
= MaxSector
;
3809 _ddf_array_name(this->name
, ddf
, i
);
3810 memset(this->uuid
, 0, sizeof(this->uuid
));
3811 this->component_size
= be64_to_cpu(vc
->conf
.blocks
);
3812 this->array
.size
= this->component_size
/ 2;
3813 this->container_member
= i
;
3815 ddf
->currentconf
= vc
;
3816 uuid_from_super_ddf(st
, this->uuid
);
3818 ddf
->currentconf
= NULL
;
3820 sprintf(this->text_version
, "/%s/%d",
3821 st
->container_devnm
, this->container_member
);
3823 for (pd
= 0; pd
< be16_to_cpu(ddf
->phys
->max_pdes
); pd
++) {
3826 const struct vd_config
*bvd
;
3830 if (be32_to_cpu(ddf
->phys
->entries
[pd
].refnum
)
3834 stt
= be16_to_cpu(ddf
->phys
->entries
[pd
].state
);
3835 if ((stt
& (DDF_Online
|DDF_Failed
|DDF_Rebuilding
))
3839 i
= get_pd_index_from_refnum(
3840 vc
, ddf
->phys
->entries
[pd
].refnum
,
3841 ddf
->mppe
, &bvd
, &iphys
);
3842 if (i
== DDF_NOTFOUND
)
3845 this->array
.working_disks
++;
3847 for (d
= ddf
->dlist
; d
; d
=d
->next
)
3848 if (be32_eq(d
->disk
.refnum
,
3849 ddf
->phys
->entries
[pd
].refnum
))
3852 /* Haven't found that one yet, maybe there are others */
3855 dev
= xcalloc(1, sizeof(*dev
));
3856 dev
->next
= this->devs
;
3859 dev
->disk
.number
= be32_to_cpu(d
->disk
.refnum
);
3860 dev
->disk
.major
= d
->major
;
3861 dev
->disk
.minor
= d
->minor
;
3862 dev
->disk
.raid_disk
= i
;
3863 dev
->disk
.state
= (1<<MD_DISK_SYNC
)|(1<<MD_DISK_ACTIVE
);
3864 dev
->recovery_start
= MaxSector
;
3866 dev
->events
= be32_to_cpu(ddf
->active
->seq
);
3868 be64_to_cpu(LBA_OFFSET(ddf
, bvd
)[iphys
]);
3869 dev
->component_size
= be64_to_cpu(bvd
->blocks
);
3871 strcpy(dev
->name
, d
->devname
);
3877 static int store_super_ddf(struct supertype
*st
, int fd
)
3879 struct ddf_super
*ddf
= st
->sb
;
3880 unsigned long long dsize
;
3887 if (!get_dev_size(fd
, NULL
, &dsize
))
3890 if (ddf
->dlist
|| ddf
->conflist
) {
3895 if (fstat(fd
, &sta
) == -1 || !S_ISBLK(sta
.st_mode
)) {
3896 pr_err("%s: file descriptor for invalid device\n",
3900 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
3901 if (dl
->major
== (int)major(sta
.st_rdev
) &&
3902 dl
->minor
== (int)minor(sta
.st_rdev
))
3905 pr_err("%s: couldn't find disk %d/%d\n", __func__
,
3906 (int)major(sta
.st_rdev
),
3907 (int)minor(sta
.st_rdev
));
3912 ret
= (_write_super_to_disk(ddf
, dl
) != 1);
3917 if (posix_memalign(&buf
, 512, 512) != 0)
3919 memset(buf
, 0, 512);
3921 lseek64(fd
, dsize
-512, 0);
3922 rc
= write(fd
, buf
, 512);
3929 static int compare_super_ddf(struct supertype
*st
, struct supertype
*tst
)
3933 * 0 same, or first was empty, and second was copied
3934 * 1 second had wrong magic number - but that isn't possible
3936 * 3 wrong other info
3938 struct ddf_super
*first
= st
->sb
;
3939 struct ddf_super
*second
= tst
->sb
;
3940 struct dl
*dl1
, *dl2
;
3941 struct vcl
*vl1
, *vl2
;
3942 unsigned int max_vds
, max_pds
, pd
, vd
;
3950 if (memcmp(first
->anchor
.guid
, second
->anchor
.guid
, DDF_GUID_LEN
) != 0)
3953 /* It is only OK to compare info in the anchor. Anything else
3954 * could be changing due to a reconfig so must be ignored.
3955 * guid really should be enough anyway.
3958 if (!be32_eq(first
->active
->seq
, second
->active
->seq
)) {
3959 dprintf("%s: sequence number mismatch %u<->%u\n", __func__
,
3960 be32_to_cpu(first
->active
->seq
),
3961 be32_to_cpu(second
->active
->seq
));
3966 * At this point we are fairly sure that the meta data matches.
3967 * But the new disk may contain additional local data.
3968 * Add it to the super block.
3970 max_vds
= be16_to_cpu(first
->active
->max_vd_entries
);
3971 max_pds
= be16_to_cpu(first
->phys
->max_pdes
);
3972 for (vl2
= second
->conflist
; vl2
; vl2
= vl2
->next
) {
3973 for (vl1
= first
->conflist
; vl1
; vl1
= vl1
->next
)
3974 if (!memcmp(vl1
->conf
.guid
, vl2
->conf
.guid
,
3978 if (vl1
->other_bvds
!= NULL
&&
3979 vl1
->conf
.sec_elmnt_seq
!=
3980 vl2
->conf
.sec_elmnt_seq
) {
3981 dprintf("%s: adding BVD %u\n", __func__
,
3982 vl2
->conf
.sec_elmnt_seq
);
3983 add_other_bvd(vl1
, &vl2
->conf
,
3984 first
->conf_rec_len
*512);
3989 if (posix_memalign((void **)&vl1
, 512,
3990 (first
->conf_rec_len
*512 +
3991 offsetof(struct vcl
, conf
))) != 0) {
3992 pr_err("%s could not allocate vcl buf\n",
3997 vl1
->next
= first
->conflist
;
3998 vl1
->block_sizes
= NULL
;
3999 memcpy(&vl1
->conf
, &vl2
->conf
, first
->conf_rec_len
*512);
4000 if (alloc_other_bvds(first
, vl1
) != 0) {
4001 pr_err("%s could not allocate other bvds\n",
4006 for (vd
= 0; vd
< max_vds
; vd
++)
4007 if (!memcmp(first
->virt
->entries
[vd
].guid
,
4008 vl1
->conf
.guid
, DDF_GUID_LEN
))
4011 dprintf("%s: added config for VD %u\n", __func__
, vl1
->vcnum
);
4012 first
->conflist
= vl1
;
4015 for (dl2
= second
->dlist
; dl2
; dl2
= dl2
->next
) {
4016 for (dl1
= first
->dlist
; dl1
; dl1
= dl1
->next
)
4017 if (be32_eq(dl1
->disk
.refnum
, dl2
->disk
.refnum
))
4022 if (posix_memalign((void **)&dl1
, 512,
4023 sizeof(*dl1
) + (first
->max_part
) * sizeof(dl1
->vlist
[0]))
4025 pr_err("%s could not allocate disk info buffer\n",
4029 memcpy(dl1
, dl2
, sizeof(*dl1
));
4030 dl1
->mdupdate
= NULL
;
4031 dl1
->next
= first
->dlist
;
4033 for (pd
= 0; pd
< max_pds
; pd
++)
4034 if (be32_eq(first
->phys
->entries
[pd
].refnum
,
4037 dl1
->pdnum
= pd
< max_pds
? (int)pd
: -1;
4039 if (posix_memalign((void **)&dl1
->spare
, 512,
4040 first
->conf_rec_len
*512) != 0) {
4041 pr_err("%s could not allocate spare info buf\n",
4045 memcpy(dl1
->spare
, dl2
->spare
, first
->conf_rec_len
*512);
4047 for (vd
= 0 ; vd
< first
->max_part
; vd
++) {
4048 if (!dl2
->vlist
[vd
]) {
4049 dl1
->vlist
[vd
] = NULL
;
4052 for (vl1
= first
->conflist
; vl1
; vl1
= vl1
->next
) {
4053 if (!memcmp(vl1
->conf
.guid
,
4054 dl2
->vlist
[vd
]->conf
.guid
,
4057 dl1
->vlist
[vd
] = vl1
;
4061 dprintf("%s: added disk %d: %08x\n", __func__
, dl1
->pdnum
,
4062 be32_to_cpu(dl1
->disk
.refnum
));
4070 * A new array 'a' has been started which claims to be instance 'inst'
4071 * within container 'c'.
4072 * We need to confirm that the array matches the metadata in 'c' so
4073 * that we don't corrupt any metadata.
4075 static int ddf_open_new(struct supertype
*c
, struct active_array
*a
, char *inst
)
4077 struct ddf_super
*ddf
= c
->sb
;
4081 static const char faulty
[] = "faulty";
4083 if (all_ff(ddf
->virt
->entries
[n
].guid
)) {
4084 pr_err("%s: subarray %d doesn't exist\n", __func__
, n
);
4087 dprintf("%s: new subarray %d, GUID: %s\n", __func__
, n
,
4088 guid_str(ddf
->virt
->entries
[n
].guid
));
4089 for (dev
= a
->info
.devs
; dev
; dev
= dev
->next
) {
4090 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4091 if (dl
->major
== dev
->disk
.major
&&
4092 dl
->minor
== dev
->disk
.minor
)
4094 if (!dl
|| dl
->pdnum
< 0) {
4095 pr_err("%s: device %d/%d of subarray %d not found in meta data\n",
4096 __func__
, dev
->disk
.major
, dev
->disk
.minor
, n
);
4099 if ((be16_to_cpu(ddf
->phys
->entries
[dl
->pdnum
].state
) &
4100 (DDF_Online
|DDF_Missing
|DDF_Failed
)) != DDF_Online
) {
4101 pr_err("%s: new subarray %d contains broken device %d/%d (%02x)\n",
4102 __func__
, n
, dl
->major
, dl
->minor
,
4104 ddf
->phys
->entries
[dl
->pdnum
].state
));
4105 if (write(dev
->state_fd
, faulty
, sizeof(faulty
)-1) !=
4107 pr_err("Write to state_fd failed\n");
4108 dev
->curr_state
= DS_FAULTY
;
4111 a
->info
.container_member
= n
;
4115 static void handle_missing(struct ddf_super
*ddf
, struct active_array
*a
, int inst
)
4117 /* This member array is being activated. If any devices
4118 * are missing they must now be marked as failed.
4120 struct vd_config
*vc
;
4128 for (n
= 0; ; n
++) {
4129 vc
= find_vdcr(ddf
, inst
, n
, &n_bvd
, &vcl
);
4132 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4133 if (be32_eq(dl
->disk
.refnum
, vc
->phys_refnum
[n_bvd
]))
4136 /* Found this disk, so not missing */
4139 /* Mark the device as failed/missing. */
4140 pd
= find_phys(ddf
, vc
->phys_refnum
[n_bvd
]);
4141 if (pd
>= 0 && be16_and(ddf
->phys
->entries
[pd
].state
,
4142 cpu_to_be16(DDF_Online
))) {
4143 be16_clear(ddf
->phys
->entries
[pd
].state
,
4144 cpu_to_be16(DDF_Online
));
4145 be16_set(ddf
->phys
->entries
[pd
].state
,
4146 cpu_to_be16(DDF_Failed
|DDF_Missing
));
4147 vc
->phys_refnum
[n_bvd
] = cpu_to_be32(0);
4148 ddf_set_updates_pending(ddf
);
4151 /* Mark the array as Degraded */
4152 state
= get_svd_state(ddf
, vcl
);
4153 if (ddf
->virt
->entries
[inst
].state
!=
4154 ((ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4156 ddf
->virt
->entries
[inst
].state
=
4157 (ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4159 a
->check_degraded
= 1;
4160 ddf_set_updates_pending(ddf
);
4166 * The array 'a' is to be marked clean in the metadata.
4167 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4168 * clean up to the point (in sectors). If that cannot be recorded in the
4169 * metadata, then leave it as dirty.
4171 * For DDF, we need to clear the DDF_state_inconsistent bit in the
4172 * !global! virtual_disk.virtual_entry structure.
4174 static int ddf_set_array_state(struct active_array
*a
, int consistent
)
4176 struct ddf_super
*ddf
= a
->container
->sb
;
4177 int inst
= a
->info
.container_member
;
4178 int old
= ddf
->virt
->entries
[inst
].state
;
4179 if (consistent
== 2) {
4180 handle_missing(ddf
, a
, inst
);
4181 /* Should check if a recovery should be started FIXME */
4183 if (!is_resync_complete(&a
->info
))
4187 ddf
->virt
->entries
[inst
].state
&= ~DDF_state_inconsistent
;
4189 ddf
->virt
->entries
[inst
].state
|= DDF_state_inconsistent
;
4190 if (old
!= ddf
->virt
->entries
[inst
].state
)
4191 ddf_set_updates_pending(ddf
);
4193 old
= ddf
->virt
->entries
[inst
].init_state
;
4194 ddf
->virt
->entries
[inst
].init_state
&= ~DDF_initstate_mask
;
4195 if (is_resync_complete(&a
->info
))
4196 ddf
->virt
->entries
[inst
].init_state
|= DDF_init_full
;
4197 else if (a
->info
.resync_start
== 0)
4198 ddf
->virt
->entries
[inst
].init_state
|= DDF_init_not
;
4200 ddf
->virt
->entries
[inst
].init_state
|= DDF_init_quick
;
4201 if (old
!= ddf
->virt
->entries
[inst
].init_state
)
4202 ddf_set_updates_pending(ddf
);
4204 dprintf("ddf mark %d/%s (%d) %s %llu\n", inst
,
4205 guid_str(ddf
->virt
->entries
[inst
].guid
), a
->curr_state
,
4206 consistent
?"clean":"dirty",
4207 a
->info
.resync_start
);
4211 static int get_bvd_state(const struct ddf_super
*ddf
,
4212 const struct vd_config
*vc
)
4214 unsigned int i
, n_bvd
, working
= 0;
4215 unsigned int n_prim
= be16_to_cpu(vc
->prim_elmnt_count
);
4217 for (i
= 0; i
< n_prim
; i
++) {
4218 if (!find_index_in_bvd(ddf
, vc
, i
, &n_bvd
))
4220 pd
= find_phys(ddf
, vc
->phys_refnum
[n_bvd
]);
4223 st
= be16_to_cpu(ddf
->phys
->entries
[pd
].state
);
4224 if ((st
& (DDF_Online
|DDF_Failed
|DDF_Rebuilding
))
4229 state
= DDF_state_degraded
;
4230 if (working
== n_prim
)
4231 state
= DDF_state_optimal
;
4237 state
= DDF_state_failed
;
4241 state
= DDF_state_failed
;
4242 else if (working
>= 2)
4243 state
= DDF_state_part_optimal
;
4247 if (working
< n_prim
- 1)
4248 state
= DDF_state_failed
;
4251 if (working
< n_prim
- 2)
4252 state
= DDF_state_failed
;
4253 else if (working
== n_prim
- 1)
4254 state
= DDF_state_part_optimal
;
4260 static int secondary_state(int state
, int other
, int seclevel
)
4262 if (state
== DDF_state_optimal
&& other
== DDF_state_optimal
)
4263 return DDF_state_optimal
;
4264 if (seclevel
== DDF_2MIRRORED
) {
4265 if (state
== DDF_state_optimal
|| other
== DDF_state_optimal
)
4266 return DDF_state_part_optimal
;
4267 if (state
== DDF_state_failed
&& other
== DDF_state_failed
)
4268 return DDF_state_failed
;
4269 return DDF_state_degraded
;
4271 if (state
== DDF_state_failed
|| other
== DDF_state_failed
)
4272 return DDF_state_failed
;
4273 if (state
== DDF_state_degraded
|| other
== DDF_state_degraded
)
4274 return DDF_state_degraded
;
4275 return DDF_state_part_optimal
;
4279 static int get_svd_state(const struct ddf_super
*ddf
, const struct vcl
*vcl
)
4281 int state
= get_bvd_state(ddf
, &vcl
->conf
);
4283 for (i
= 1; i
< vcl
->conf
.sec_elmnt_count
; i
++) {
4284 state
= secondary_state(
4286 get_bvd_state(ddf
, vcl
->other_bvds
[i
-1]),
4293 * The state of each disk is stored in the global phys_disk structure
4294 * in phys_disk.entries[n].state.
4295 * This makes various combinations awkward.
4296 * - When a device fails in any array, it must be failed in all arrays
4297 * that include a part of this device.
4298 * - When a component is rebuilding, we cannot include it officially in the
4299 * array unless this is the only array that uses the device.
4301 * So: when transitioning:
4302 * Online -> failed, just set failed flag. monitor will propagate
4303 * spare -> online, the device might need to be added to the array.
4304 * spare -> failed, just set failed. Don't worry if in array or not.
4306 static void ddf_set_disk(struct active_array
*a
, int n
, int state
)
4308 struct ddf_super
*ddf
= a
->container
->sb
;
4309 unsigned int inst
= a
->info
.container_member
, n_bvd
;
4311 struct vd_config
*vc
= find_vdcr(ddf
, inst
, (unsigned int)n
,
4317 dprintf("%s: %d to %x\n", __func__
, n
, state
);
4319 dprintf("ddf: cannot find instance %d!!\n", inst
);
4322 /* Find the matching slot in 'info'. */
4323 for (mdi
= a
->info
.devs
; mdi
; mdi
= mdi
->next
)
4324 if (mdi
->disk
.raid_disk
== n
)
4327 pr_err("%s: cannot find raid disk %d\n",
4332 /* and find the 'dl' entry corresponding to that. */
4333 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4334 if (mdi
->state_fd
>= 0 &&
4335 mdi
->disk
.major
== dl
->major
&&
4336 mdi
->disk
.minor
== dl
->minor
)
4339 pr_err("%s: cannot find raid disk %d (%d/%d)\n",
4341 mdi
->disk
.major
, mdi
->disk
.minor
);
4345 pd
= find_phys(ddf
, vc
->phys_refnum
[n_bvd
]);
4346 if (pd
< 0 || pd
!= dl
->pdnum
) {
4347 /* disk doesn't currently exist or has changed.
4348 * If it is now in_sync, insert it. */
4349 dprintf("%s: phys disk not found for %d: %d/%d ref %08x\n",
4350 __func__
, dl
->pdnum
, dl
->major
, dl
->minor
,
4351 be32_to_cpu(dl
->disk
.refnum
));
4352 dprintf("%s: array %u disk %u ref %08x pd %d\n",
4353 __func__
, inst
, n_bvd
,
4354 be32_to_cpu(vc
->phys_refnum
[n_bvd
]), pd
);
4355 if ((state
& DS_INSYNC
) && ! (state
& DS_FAULTY
)) {
4356 pd
= dl
->pdnum
; /* FIXME: is this really correct ? */
4357 vc
->phys_refnum
[n_bvd
] = dl
->disk
.refnum
;
4358 LBA_OFFSET(ddf
, vc
)[n_bvd
] =
4359 cpu_to_be64(mdi
->data_offset
);
4360 be16_clear(ddf
->phys
->entries
[pd
].type
,
4361 cpu_to_be16(DDF_Global_Spare
));
4362 be16_set(ddf
->phys
->entries
[pd
].type
,
4363 cpu_to_be16(DDF_Active_in_VD
));
4364 ddf_set_updates_pending(ddf
);
4367 be16 old
= ddf
->phys
->entries
[pd
].state
;
4368 if (state
& DS_FAULTY
)
4369 be16_set(ddf
->phys
->entries
[pd
].state
,
4370 cpu_to_be16(DDF_Failed
));
4371 if (state
& DS_INSYNC
) {
4372 be16_set(ddf
->phys
->entries
[pd
].state
,
4373 cpu_to_be16(DDF_Online
));
4374 be16_clear(ddf
->phys
->entries
[pd
].state
,
4375 cpu_to_be16(DDF_Rebuilding
));
4377 if (!be16_eq(old
, ddf
->phys
->entries
[pd
].state
))
4378 ddf_set_updates_pending(ddf
);
4381 dprintf("ddf: set_disk %d (%08x) to %x->%02x\n", n
,
4382 be32_to_cpu(dl
->disk
.refnum
), state
,
4383 be16_to_cpu(ddf
->phys
->entries
[pd
].state
));
4385 /* Now we need to check the state of the array and update
4386 * virtual_disk.entries[n].state.
4387 * It needs to be one of "optimal", "degraded", "failed".
4388 * I don't understand 'deleted' or 'missing'.
4390 state
= get_svd_state(ddf
, vcl
);
4392 if (ddf
->virt
->entries
[inst
].state
!=
4393 ((ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4395 ddf
->virt
->entries
[inst
].state
=
4396 (ddf
->virt
->entries
[inst
].state
& ~DDF_state_mask
)
4398 ddf_set_updates_pending(ddf
);
4403 static void ddf_sync_metadata(struct supertype
*st
)
4406 * Write all data to all devices.
4407 * Later, we might be able to track whether only local changes
4408 * have been made, or whether any global data has been changed,
4409 * but ddf is sufficiently weird that it probably always
4410 * changes global data ....
4412 struct ddf_super
*ddf
= st
->sb
;
4413 if (!ddf
->updates_pending
)
4415 ddf
->updates_pending
= 0;
4416 __write_init_super_ddf(st
);
4417 dprintf("ddf: sync_metadata\n");
4420 static int del_from_conflist(struct vcl
**list
, const char *guid
)
4424 for (p
= list
; p
&& *p
; p
= &((*p
)->next
))
4425 if (!memcmp((*p
)->conf
.guid
, guid
, DDF_GUID_LEN
)) {
4432 static int _kill_subarray_ddf(struct ddf_super
*ddf
, const char *guid
)
4435 unsigned int vdnum
, i
;
4436 vdnum
= find_vde_by_guid(ddf
, guid
);
4437 if (vdnum
== DDF_NOTFOUND
) {
4438 pr_err("%s: could not find VD %s\n", __func__
,
4442 if (del_from_conflist(&ddf
->conflist
, guid
) == 0) {
4443 pr_err("%s: could not find conf %s\n", __func__
,
4447 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4448 for (i
= 0; i
< ddf
->max_part
; i
++)
4449 if (dl
->vlist
[i
] != NULL
&&
4450 !memcmp(dl
->vlist
[i
]->conf
.guid
, guid
,
4452 dl
->vlist
[i
] = NULL
;
4453 memset(ddf
->virt
->entries
[vdnum
].guid
, 0xff, DDF_GUID_LEN
);
4454 dprintf("%s: deleted %s\n", __func__
, guid_str(guid
));
4458 static int kill_subarray_ddf(struct supertype
*st
)
4460 struct ddf_super
*ddf
= st
->sb
;
4462 * currentconf is set in container_content_ddf,
4463 * called with subarray arg
4465 struct vcl
*victim
= ddf
->currentconf
;
4466 struct vd_config
*conf
;
4469 ddf
->currentconf
= NULL
;
4471 pr_err("%s: nothing to kill\n", __func__
);
4474 conf
= &victim
->conf
;
4475 vdnum
= find_vde_by_guid(ddf
, conf
->guid
);
4476 if (vdnum
== DDF_NOTFOUND
) {
4477 pr_err("%s: could not find VD %s\n", __func__
,
4478 guid_str(conf
->guid
));
4481 if (st
->update_tail
) {
4482 struct virtual_disk
*vd
;
4483 int len
= sizeof(struct virtual_disk
)
4484 + sizeof(struct virtual_entry
);
4487 pr_err("%s: failed to allocate %d bytes\n", __func__
,
4491 memset(vd
, 0 , len
);
4492 vd
->magic
= DDF_VIRT_RECORDS_MAGIC
;
4493 vd
->populated_vdes
= cpu_to_be16(0);
4494 memcpy(vd
->entries
[0].guid
, conf
->guid
, DDF_GUID_LEN
);
4495 /* we use DDF_state_deleted as marker */
4496 vd
->entries
[0].state
= DDF_state_deleted
;
4497 append_metadata_update(st
, vd
, len
);
4499 _kill_subarray_ddf(ddf
, conf
->guid
);
4500 ddf_set_updates_pending(ddf
);
4501 ddf_sync_metadata(st
);
4506 static void copy_matching_bvd(struct ddf_super
*ddf
,
4507 struct vd_config
*conf
,
4508 const struct metadata_update
*update
)
4511 be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
4512 unsigned int len
= ddf
->conf_rec_len
* 512;
4514 struct vd_config
*vc
;
4515 for (p
= update
->buf
; p
< update
->buf
+ update
->len
; p
+= len
) {
4516 vc
= (struct vd_config
*) p
;
4517 if (vc
->sec_elmnt_seq
== conf
->sec_elmnt_seq
) {
4518 memcpy(conf
->phys_refnum
, vc
->phys_refnum
,
4519 mppe
* (sizeof(__u32
) + sizeof(__u64
)));
4523 pr_err("%s: no match for BVD %d of %s in update\n", __func__
,
4524 conf
->sec_elmnt_seq
, guid_str(conf
->guid
));
4527 static void ddf_process_update(struct supertype
*st
,
4528 struct metadata_update
*update
)
4530 /* Apply this update to the metadata.
4531 * The first 4 bytes are a DDF_*_MAGIC which guides
4533 * Possible update are:
4534 * DDF_PHYS_RECORDS_MAGIC
4535 * Add a new physical device or remove an old one.
4536 * Changes to this record only happen implicitly.
4537 * used_pdes is the device number.
4538 * DDF_VIRT_RECORDS_MAGIC
4539 * Add a new VD. Possibly also change the 'access' bits.
4540 * populated_vdes is the entry number.
4542 * New or updated VD. the VIRT_RECORD must already
4543 * exist. For an update, phys_refnum and lba_offset
4544 * (at least) are updated, and the VD_CONF must
4545 * be written to precisely those devices listed with
4547 * DDF_SPARE_ASSIGN_MAGIC
4548 * replacement Spare Assignment Record... but for which device?
4551 * - to create a new array, we send a VIRT_RECORD and
4552 * a VD_CONF. Then assemble and start the array.
4553 * - to activate a spare we send a VD_CONF to add the phys_refnum
4554 * and offset. This will also mark the spare as active with
4555 * a spare-assignment record.
4557 struct ddf_super
*ddf
= st
->sb
;
4558 be32
*magic
= (be32
*)update
->buf
;
4559 struct phys_disk
*pd
;
4560 struct virtual_disk
*vd
;
4561 struct vd_config
*vc
;
4565 unsigned int pdnum
, pd2
, len
;
4567 dprintf("Process update %x\n", be32_to_cpu(*magic
));
4569 if (be32_eq(*magic
, DDF_PHYS_RECORDS_MAGIC
)) {
4570 if (update
->len
!= (sizeof(struct phys_disk
) +
4571 sizeof(struct phys_disk_entry
)))
4573 pd
= (struct phys_disk
*)update
->buf
;
4575 ent
= be16_to_cpu(pd
->used_pdes
);
4576 if (ent
>= be16_to_cpu(ddf
->phys
->max_pdes
))
4578 if (be16_and(pd
->entries
[0].state
, cpu_to_be16(DDF_Missing
))) {
4580 /* removing this disk. */
4581 be16_set(ddf
->phys
->entries
[ent
].state
,
4582 cpu_to_be16(DDF_Missing
));
4583 for (dlp
= &ddf
->dlist
; *dlp
; dlp
= &(*dlp
)->next
) {
4584 struct dl
*dl
= *dlp
;
4585 if (dl
->pdnum
== (signed)ent
) {
4588 /* FIXME this doesn't free
4595 ddf_set_updates_pending(ddf
);
4598 if (!all_ff(ddf
->phys
->entries
[ent
].guid
))
4600 ddf
->phys
->entries
[ent
] = pd
->entries
[0];
4601 ddf
->phys
->used_pdes
= cpu_to_be16
4602 (1 + be16_to_cpu(ddf
->phys
->used_pdes
));
4603 ddf_set_updates_pending(ddf
);
4604 if (ddf
->add_list
) {
4605 struct active_array
*a
;
4606 struct dl
*al
= ddf
->add_list
;
4607 ddf
->add_list
= al
->next
;
4609 al
->next
= ddf
->dlist
;
4612 /* As a device has been added, we should check
4613 * for any degraded devices that might make
4614 * use of this spare */
4615 for (a
= st
->arrays
; a
; a
=a
->next
)
4616 a
->check_degraded
= 1;
4618 } else if (be32_eq(*magic
, DDF_VIRT_RECORDS_MAGIC
)) {
4619 if (update
->len
!= (sizeof(struct virtual_disk
) +
4620 sizeof(struct virtual_entry
)))
4622 vd
= (struct virtual_disk
*)update
->buf
;
4624 if (vd
->entries
[0].state
== DDF_state_deleted
) {
4625 if (_kill_subarray_ddf(ddf
, vd
->entries
[0].guid
))
4628 ent
= find_vde_by_guid(ddf
, vd
->entries
[0].guid
);
4629 if (ent
!= DDF_NOTFOUND
) {
4630 dprintf("%s: VD %s exists already in slot %d\n",
4631 __func__
, guid_str(vd
->entries
[0].guid
),
4635 ent
= find_unused_vde(ddf
);
4636 if (ent
== DDF_NOTFOUND
)
4638 ddf
->virt
->entries
[ent
] = vd
->entries
[0];
4639 ddf
->virt
->populated_vdes
=
4642 ddf
->virt
->populated_vdes
));
4643 dprintf("%s: added VD %s in slot %d(s=%02x i=%02x)\n",
4644 __func__
, guid_str(vd
->entries
[0].guid
), ent
,
4645 ddf
->virt
->entries
[ent
].state
,
4646 ddf
->virt
->entries
[ent
].init_state
);
4648 ddf_set_updates_pending(ddf
);
4651 else if (be32_eq(*magic
, DDF_VD_CONF_MAGIC
)) {
4652 vc
= (struct vd_config
*)update
->buf
;
4653 len
= ddf
->conf_rec_len
* 512;
4654 if ((unsigned int)update
->len
!= len
* vc
->sec_elmnt_count
) {
4655 pr_err("%s: %s: insufficient data (%d) for %u BVDs\n",
4656 __func__
, guid_str(vc
->guid
), update
->len
,
4657 vc
->sec_elmnt_count
);
4660 for (vcl
= ddf
->conflist
; vcl
; vcl
= vcl
->next
)
4661 if (memcmp(vcl
->conf
.guid
, vc
->guid
, DDF_GUID_LEN
) == 0)
4663 dprintf("%s: conf update for %s (%s)\n", __func__
,
4664 guid_str(vc
->guid
), (vcl
? "old" : "new"));
4666 /* An update, just copy the phys_refnum and lba_offset
4671 copy_matching_bvd(ddf
, &vcl
->conf
, update
);
4672 for (k
= 0; k
< be16_to_cpu(vc
->prim_elmnt_count
); k
++)
4673 dprintf("BVD %u has %08x at %llu\n", 0,
4674 be32_to_cpu(vcl
->conf
.phys_refnum
[k
]),
4675 be64_to_cpu(LBA_OFFSET(ddf
,
4677 for (i
= 1; i
< vc
->sec_elmnt_count
; i
++) {
4678 copy_matching_bvd(ddf
, vcl
->other_bvds
[i
-1],
4680 for (k
= 0; k
< be16_to_cpu(
4681 vc
->prim_elmnt_count
); k
++)
4682 dprintf("BVD %u has %08x at %llu\n", i
,
4684 (vcl
->other_bvds
[i
-1]->
4689 vcl
->other_bvds
[i
-1])[k
]));
4696 vcl
= update
->space
;
4697 update
->space
= NULL
;
4698 vcl
->next
= ddf
->conflist
;
4699 memcpy(&vcl
->conf
, vc
, len
);
4700 ent
= find_vde_by_guid(ddf
, vc
->guid
);
4701 if (ent
== DDF_NOTFOUND
)
4704 ddf
->conflist
= vcl
;
4705 for (i
= 1; i
< vc
->sec_elmnt_count
; i
++)
4706 memcpy(vcl
->other_bvds
[i
-1],
4707 update
->buf
+ len
* i
, len
);
4709 /* Set DDF_Transition on all Failed devices - to help
4710 * us detect those that are no longer in use
4712 for (pdnum
= 0; pdnum
< be16_to_cpu(ddf
->phys
->max_pdes
);
4714 if (be16_and(ddf
->phys
->entries
[pdnum
].state
,
4715 cpu_to_be16(DDF_Failed
)))
4716 be16_set(ddf
->phys
->entries
[pdnum
].state
,
4717 cpu_to_be16(DDF_Transition
));
4718 /* Now make sure vlist is correct for each dl. */
4719 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
4720 unsigned int vn
= 0;
4721 int in_degraded
= 0;
4725 for (vcl
= ddf
->conflist
; vcl
; vcl
= vcl
->next
) {
4726 unsigned int dn
, ibvd
;
4727 const struct vd_config
*conf
;
4729 dn
= get_pd_index_from_refnum(vcl
,
4733 if (dn
== DDF_NOTFOUND
)
4735 dprintf("dev %d/%08x has %s (sec=%u) at %d\n",
4737 be32_to_cpu(dl
->disk
.refnum
),
4738 guid_str(conf
->guid
),
4739 conf
->sec_elmnt_seq
, vn
);
4740 /* Clear the Transition flag */
4742 (ddf
->phys
->entries
[dl
->pdnum
].state
,
4743 cpu_to_be16(DDF_Failed
)))
4744 be16_clear(ddf
->phys
4745 ->entries
[dl
->pdnum
].state
,
4746 cpu_to_be16(DDF_Transition
));
4747 dl
->vlist
[vn
++] = vcl
;
4748 vstate
= ddf
->virt
->entries
[vcl
->vcnum
].state
4750 if (vstate
== DDF_state_degraded
||
4751 vstate
== DDF_state_part_optimal
)
4754 while (vn
< ddf
->max_part
)
4755 dl
->vlist
[vn
++] = NULL
;
4757 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4758 cpu_to_be16(DDF_Global_Spare
));
4759 if (!be16_and(ddf
->phys
4760 ->entries
[dl
->pdnum
].type
,
4761 cpu_to_be16(DDF_Active_in_VD
))) {
4763 ->entries
[dl
->pdnum
].type
,
4764 cpu_to_be16(DDF_Active_in_VD
));
4767 ->entries
[dl
->pdnum
]
4774 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4775 cpu_to_be16(DDF_Global_Spare
));
4776 be16_set(ddf
->phys
->entries
[dl
->pdnum
].type
,
4777 cpu_to_be16(DDF_Spare
));
4779 if (!dl
->vlist
[0] && !dl
->spare
) {
4780 be16_set(ddf
->phys
->entries
[dl
->pdnum
].type
,
4781 cpu_to_be16(DDF_Global_Spare
));
4782 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4783 cpu_to_be16(DDF_Spare
));
4784 be16_clear(ddf
->phys
->entries
[dl
->pdnum
].type
,
4785 cpu_to_be16(DDF_Active_in_VD
));
4789 /* Now remove any 'Failed' devices that are not part
4790 * of any VD. They will have the Transition flag set.
4791 * Once done, we need to update all dl->pdnum numbers.
4794 for (pdnum
= 0; pdnum
< be16_to_cpu(ddf
->phys
->max_pdes
);
4796 if (be32_to_cpu(ddf
->phys
->entries
[pdnum
].refnum
) ==
4799 if (be16_and(ddf
->phys
->entries
[pdnum
].state
,
4800 cpu_to_be16(DDF_Failed
))
4801 && be16_and(ddf
->phys
->entries
[pdnum
].state
,
4802 cpu_to_be16(DDF_Transition
))) {
4803 /* skip this one unless in dlist*/
4804 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4805 if (dl
->pdnum
== (int)pdnum
)
4813 ddf
->phys
->entries
[pd2
] =
4814 ddf
->phys
->entries
[pdnum
];
4815 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
4816 if (dl
->pdnum
== (int)pdnum
)
4821 ddf
->phys
->used_pdes
= cpu_to_be16(pd2
);
4822 while (pd2
< pdnum
) {
4823 memset(ddf
->phys
->entries
[pd2
].guid
, 0xff,
4828 ddf_set_updates_pending(ddf
);
4830 /* case DDF_SPARE_ASSIGN_MAGIC */
4833 static void ddf_prepare_update(struct supertype
*st
,
4834 struct metadata_update
*update
)
4836 /* This update arrived at managemon.
4837 * We are about to pass it to monitor.
4838 * If a malloc is needed, do it here.
4840 struct ddf_super
*ddf
= st
->sb
;
4841 be32
*magic
= (be32
*)update
->buf
;
4842 if (be32_eq(*magic
, DDF_VD_CONF_MAGIC
)) {
4844 struct vd_config
*conf
= (struct vd_config
*) update
->buf
;
4845 if (posix_memalign(&update
->space
, 512,
4846 offsetof(struct vcl
, conf
)
4847 + ddf
->conf_rec_len
* 512) != 0) {
4848 update
->space
= NULL
;
4851 vcl
= update
->space
;
4852 vcl
->conf
.sec_elmnt_count
= conf
->sec_elmnt_count
;
4853 if (alloc_other_bvds(ddf
, vcl
) != 0) {
4854 free(update
->space
);
4855 update
->space
= NULL
;
4861 * Check degraded state of a RAID10.
4862 * returns 2 for good, 1 for degraded, 0 for failed, and -1 for error
4864 static int raid10_degraded(struct mdinfo
*info
)
4872 n_prim
= info
->array
.layout
& ~0x100;
4873 n_bvds
= info
->array
.raid_disks
/ n_prim
;
4874 found
= xmalloc(n_bvds
);
4877 memset(found
, 0, n_bvds
);
4878 for (d
= info
->devs
; d
; d
= d
->next
) {
4879 i
= d
->disk
.raid_disk
/ n_prim
;
4881 pr_err("%s: BUG: invalid raid disk\n", __func__
);
4884 if (d
->state_fd
> 0)
4888 for (i
= 0; i
< n_bvds
; i
++)
4890 dprintf("%s: BVD %d/%d failed\n", __func__
, i
, n_bvds
);
4893 } else if (found
[i
] < n_prim
) {
4894 dprintf("%s: BVD %d/%d degraded\n", __func__
, i
,
4904 * Check if the array 'a' is degraded but not failed.
4905 * If it is, find as many spares as are available and needed and
4906 * arrange for their inclusion.
4907 * We only choose devices which are not already in the array,
4908 * and prefer those with a spare-assignment to this array.
4909 * Otherwise we choose global spares - assuming always that
4910 * there is enough room.
4911 * For each spare that we assign, we return an 'mdinfo' which
4912 * describes the position for the device in the array.
4913 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
4914 * the new phys_refnum and lba_offset values.
4916 * Only worry about BVDs at the moment.
4918 static struct mdinfo
*ddf_activate_spare(struct active_array
*a
,
4919 struct metadata_update
**updates
)
4923 struct ddf_super
*ddf
= a
->container
->sb
;
4925 struct mdinfo
*rv
= NULL
;
4927 struct metadata_update
*mu
;
4932 struct vd_config
*vc
;
4935 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
4936 if ((d
->curr_state
& DS_FAULTY
) &&
4938 /* wait for Removal to happen */
4940 if (d
->state_fd
>= 0)
4944 dprintf("%s: working=%d (%d) level=%d\n", __func__
, working
,
4945 a
->info
.array
.raid_disks
,
4946 a
->info
.array
.level
);
4947 if (working
== a
->info
.array
.raid_disks
)
4948 return NULL
; /* array not degraded */
4949 switch (a
->info
.array
.level
) {
4952 return NULL
; /* failed */
4956 if (working
< a
->info
.array
.raid_disks
- 1)
4957 return NULL
; /* failed */
4960 if (working
< a
->info
.array
.raid_disks
- 2)
4961 return NULL
; /* failed */
4964 if (raid10_degraded(&a
->info
) < 1)
4967 default: /* concat or stripe */
4968 return NULL
; /* failed */
4971 /* For each slot, if it is not working, find a spare */
4973 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
4974 for (d
= a
->info
.devs
; d
; d
= d
->next
)
4975 if (d
->disk
.raid_disk
== i
)
4977 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
4978 if (d
&& (d
->state_fd
>= 0))
4981 /* OK, this device needs recovery. Find a spare */
4983 for ( ; dl
; dl
= dl
->next
) {
4984 unsigned long long esize
;
4985 unsigned long long pos
;
4988 int is_dedicated
= 0;
4995 state
= ddf
->phys
->entries
[dl
->pdnum
].state
;
4997 cpu_to_be16(DDF_Failed
|DDF_Missing
)) ||
4999 cpu_to_be16(DDF_Online
)))
5002 /* If in this array, skip */
5003 for (d2
= a
->info
.devs
; d2
; d2
= d2
->next
)
5004 if (d2
->state_fd
>= 0 &&
5005 d2
->disk
.major
== dl
->major
&&
5006 d2
->disk
.minor
== dl
->minor
) {
5007 dprintf("%x:%x (%08x) already in array\n",
5008 dl
->major
, dl
->minor
,
5009 be32_to_cpu(dl
->disk
.refnum
));
5014 if (be16_and(ddf
->phys
->entries
[dl
->pdnum
].type
,
5015 cpu_to_be16(DDF_Spare
))) {
5016 /* Check spare assign record */
5018 if (dl
->spare
->type
& DDF_spare_dedicated
) {
5019 /* check spare_ents for guid */
5025 if (memcmp(dl
->spare
->spare_ents
[j
].guid
,
5026 ddf
->virt
->entries
[a
->info
.container_member
].guid
,
5033 } else if (be16_and(ddf
->phys
->entries
[dl
->pdnum
].type
,
5034 cpu_to_be16(DDF_Global_Spare
))) {
5036 } else if (!be16_and(ddf
->phys
5037 ->entries
[dl
->pdnum
].state
,
5038 cpu_to_be16(DDF_Failed
))) {
5039 /* we can possibly use some of this */
5042 if ( ! (is_dedicated
||
5043 (is_global
&& global_ok
))) {
5044 dprintf("%x:%x not suitable: %d %d\n", dl
->major
, dl
->minor
,
5045 is_dedicated
, is_global
);
5049 /* We are allowed to use this device - is there space?
5050 * We need a->info.component_size sectors */
5051 ex
= get_extents(ddf
, dl
);
5053 dprintf("cannot get extents\n");
5060 esize
= ex
[j
].start
- pos
;
5061 if (esize
>= a
->info
.component_size
)
5063 pos
= ex
[j
].start
+ ex
[j
].size
;
5065 } while (ex
[j
-1].size
);
5068 if (esize
< a
->info
.component_size
) {
5069 dprintf("%x:%x has no room: %llu %llu\n",
5070 dl
->major
, dl
->minor
,
5071 esize
, a
->info
.component_size
);
5076 /* Cool, we have a device with some space at pos */
5077 di
= xcalloc(1, sizeof(*di
));
5078 di
->disk
.number
= i
;
5079 di
->disk
.raid_disk
= i
;
5080 di
->disk
.major
= dl
->major
;
5081 di
->disk
.minor
= dl
->minor
;
5083 di
->recovery_start
= 0;
5084 di
->data_offset
= pos
;
5085 di
->component_size
= a
->info
.component_size
;
5088 dprintf("%x:%x (%08x) to be %d at %llu\n",
5089 dl
->major
, dl
->minor
,
5090 be32_to_cpu(dl
->disk
.refnum
), i
, pos
);
5094 if (!dl
&& ! global_ok
) {
5095 /* not enough dedicated spares, try global */
5103 /* No spares found */
5105 /* Now 'rv' has a list of devices to return.
5106 * Create a metadata_update record to update the
5107 * phys_refnum and lba_offset values
5109 vc
= find_vdcr(ddf
, a
->info
.container_member
, rv
->disk
.raid_disk
,
5114 mu
= xmalloc(sizeof(*mu
));
5115 if (posix_memalign(&mu
->space
, 512, sizeof(struct vcl
)) != 0) {
5120 mu
->len
= ddf
->conf_rec_len
* 512 * vcl
->conf
.sec_elmnt_count
;
5121 mu
->buf
= xmalloc(mu
->len
);
5123 mu
->space_list
= NULL
;
5124 mu
->next
= *updates
;
5125 memcpy(mu
->buf
, &vcl
->conf
, ddf
->conf_rec_len
* 512);
5126 for (j
= 1; j
< vcl
->conf
.sec_elmnt_count
; j
++)
5127 memcpy(mu
->buf
+ j
* ddf
->conf_rec_len
* 512,
5128 vcl
->other_bvds
[j
-1], ddf
->conf_rec_len
* 512);
5130 vc
= (struct vd_config
*)mu
->buf
;
5131 for (di
= rv
; di
; di
= di
->next
) {
5132 unsigned int i_sec
, i_prim
;
5133 i_sec
= di
->disk
.raid_disk
5134 / be16_to_cpu(vcl
->conf
.prim_elmnt_count
);
5135 i_prim
= di
->disk
.raid_disk
5136 % be16_to_cpu(vcl
->conf
.prim_elmnt_count
);
5137 vc
= (struct vd_config
*)(mu
->buf
5138 + i_sec
* ddf
->conf_rec_len
* 512);
5139 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
5140 if (dl
->major
== di
->disk
.major
5141 && dl
->minor
== di
->disk
.minor
)
5143 if (!dl
|| dl
->pdnum
< 0) {
5144 pr_err("%s: BUG: can't find disk %d (%d/%d)\n",
5145 __func__
, di
->disk
.raid_disk
,
5146 di
->disk
.major
, di
->disk
.minor
);
5149 vc
->phys_refnum
[i_prim
] = ddf
->phys
->entries
[dl
->pdnum
].refnum
;
5150 LBA_OFFSET(ddf
, vc
)[i_prim
] = cpu_to_be64(di
->data_offset
);
5151 dprintf("BVD %u gets %u: %08x at %llu\n", i_sec
, i_prim
,
5152 be32_to_cpu(vc
->phys_refnum
[i_prim
]),
5153 be64_to_cpu(LBA_OFFSET(ddf
, vc
)[i_prim
]));
5158 #endif /* MDASSEMBLE */
5160 static int ddf_level_to_layout(int level
)
5167 return ALGORITHM_LEFT_SYMMETRIC
;
5169 return ALGORITHM_ROTATING_N_CONTINUE
;
5177 static void default_geometry_ddf(struct supertype
*st
, int *level
, int *layout
, int *chunk
)
5179 if (level
&& *level
== UnSet
)
5180 *level
= LEVEL_CONTAINER
;
5182 if (level
&& layout
&& *layout
== UnSet
)
5183 *layout
= ddf_level_to_layout(*level
);
5186 struct superswitch super_ddf
= {
5188 .examine_super
= examine_super_ddf
,
5189 .brief_examine_super
= brief_examine_super_ddf
,
5190 .brief_examine_subarrays
= brief_examine_subarrays_ddf
,
5191 .export_examine_super
= export_examine_super_ddf
,
5192 .detail_super
= detail_super_ddf
,
5193 .brief_detail_super
= brief_detail_super_ddf
,
5194 .validate_geometry
= validate_geometry_ddf
,
5195 .write_init_super
= write_init_super_ddf
,
5196 .add_to_super
= add_to_super_ddf
,
5197 .remove_from_super
= remove_from_super_ddf
,
5198 .load_container
= load_container_ddf
,
5199 .copy_metadata
= copy_metadata_ddf
,
5200 .kill_subarray
= kill_subarray_ddf
,
5202 .match_home
= match_home_ddf
,
5203 .uuid_from_super
= uuid_from_super_ddf
,
5204 .getinfo_super
= getinfo_super_ddf
,
5205 .update_super
= update_super_ddf
,
5207 .avail_size
= avail_size_ddf
,
5209 .compare_super
= compare_super_ddf
,
5211 .load_super
= load_super_ddf
,
5212 .init_super
= init_super_ddf
,
5213 .store_super
= store_super_ddf
,
5214 .free_super
= free_super_ddf
,
5215 .match_metadata_desc
= match_metadata_desc_ddf
,
5216 .container_content
= container_content_ddf
,
5217 .default_geometry
= default_geometry_ddf
,
5223 .open_new
= ddf_open_new
,
5224 .set_array_state
= ddf_set_array_state
,
5225 .set_disk
= ddf_set_disk
,
5226 .sync_metadata
= ddf_sync_metadata
,
5227 .process_update
= ddf_process_update
,
5228 .prepare_update
= ddf_prepare_update
,
5229 .activate_spare
= ddf_activate_spare
,