2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2006-2007 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 takes from Common RAID DDF Specification Revision 1.2
25 * (July 28 2006). Reused by permission of SNIA.
28 #define HAVE_STDINT_H 1
34 static inline int ROUND_UP(int a
, int base
)
36 return ((a
+base
-1)/base
)*base
;
39 /* a non-official T10 name for creation GUIDs */
40 static char T10
[] = "Linux-MD";
42 /* DDF timestamps are 1980 based, so we need to add
43 * second-in-decade-of-seventies to convert to linux timestamps.
44 * 10 years with 2 leap years.
46 #define DECADE (3600*24*(365*10+2))
49 const unsigned char *buf
,
52 /* The DDF metadata handling.
53 * DDF metadata lives at the end of the device.
54 * The last 512 byte block provides an 'anchor' which is used to locate
55 * the rest of the metadata which usually lives immediately behind the anchor.
58 * - all multibyte numeric fields are bigendian.
59 * - all strings are space padded.
63 /* Primary Raid Level (PRL) */
64 #define DDF_RAID0 0x00
65 #define DDF_RAID1 0x01
66 #define DDF_RAID3 0x03
67 #define DDF_RAID4 0x04
68 #define DDF_RAID5 0x05
69 #define DDF_RAID1E 0x11
71 #define DDF_CONCAT 0x1f
72 #define DDF_RAID5E 0x15
73 #define DDF_RAID5EE 0x25
74 #define DDF_RAID6 0x16 /* Vendor unique layout */
76 /* Raid Level Qualifier (RLQ) */
77 #define DDF_RAID0_SIMPLE 0x00
78 #define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
79 #define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
80 #define DDF_RAID3_0 0x00 /* parity in first extent */
81 #define DDF_RAID3_N 0x01 /* parity in last extent */
82 #define DDF_RAID4_0 0x00 /* parity in first extent */
83 #define DDF_RAID4_N 0x01 /* parity in last extent */
84 /* these apply to raid5e and raid5ee as well */
85 #define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
86 #define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
87 #define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
89 #define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
90 #define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
92 /* Secondary RAID Level (SRL) */
93 #define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
94 #define DDF_2MIRRORED 0x01
95 #define DDF_2CONCAT 0x02
96 #define DDF_2SPANNED 0x03 /* This is also weird - be careful */
99 #define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
100 #define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
101 #define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
102 #define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
103 #define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
104 #define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
105 #define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
106 #define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
107 #define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
108 #define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
110 #define DDF_GUID_LEN 24
111 #define DDF_REVISION "01.00.00"
116 char guid
[DDF_GUID_LEN
];
117 char revision
[8]; /* 01.00.00 */
118 __u32 seq
; /* starts at '1' */
123 __u8 pad0
; /* 0xff */
124 __u8 pad1
[12]; /* 12 * 0xff */
125 /* 64 bytes so far */
126 __u8 header_ext
[32]; /* reserved: fill with 0xff */
130 __u8 pad2
[3]; /* 0xff */
131 __u32 workspace_len
; /* sectors for vendor space -
132 * at least 32768(sectors) */
134 __u16 max_pd_entries
; /* one of 15, 63, 255, 1023, 4095 */
135 __u16 max_vd_entries
; /* 2^(4,6,8,10,12)-1 : i.e. as above */
136 __u16 max_partitions
; /* i.e. max num of configuration
137 record entries per disk */
138 __u16 config_record_len
; /* 1 +ROUNDUP(max_primary_element_entries
140 __u16 max_primary_element_entries
; /* 16, 64, 256, 1024, or 4096 */
141 __u8 pad3
[54]; /* 0xff */
142 /* 192 bytes so far */
143 __u32 controller_section_offset
;
144 __u32 controller_section_length
;
145 __u32 phys_section_offset
;
146 __u32 phys_section_length
;
147 __u32 virt_section_offset
;
148 __u32 virt_section_length
;
149 __u32 config_section_offset
;
150 __u32 config_section_length
;
151 __u32 data_section_offset
;
152 __u32 data_section_length
;
153 __u32 bbm_section_offset
;
154 __u32 bbm_section_length
;
155 __u32 diag_space_offset
;
156 __u32 diag_space_length
;
159 /* 256 bytes so far */
160 __u8 pad4
[256]; /* 0xff */
164 #define DDF_HEADER_ANCHOR 0x00
165 #define DDF_HEADER_PRIMARY 0x01
166 #define DDF_HEADER_SECONDARY 0x02
168 /* The content of the 'controller section' - global scope */
169 struct ddf_controller_data
{
172 char guid
[DDF_GUID_LEN
];
173 struct controller_type
{
180 __u8 pad
[8]; /* 0xff */
181 __u8 vendor_data
[448];
184 /* The content of phys_section - global scope */
191 struct phys_disk_entry
{
192 char guid
[DDF_GUID_LEN
];
196 __u64 config_size
; /* DDF structures must be after here */
197 char path
[18]; /* another horrible structure really */
202 /* phys_disk_entry.type is a bitmap - bigendian remember */
203 #define DDF_Forced_PD_GUID 1
204 #define DDF_Active_in_VD 2
205 #define DDF_Global_Spare 4
206 #define DDF_Spare 8 /* overrides Global_spare */
207 #define DDF_Foreign 16
208 #define DDF_Legacy 32 /* no DDF on this device */
210 #define DDF_Interface_mask 0xf00
211 #define DDF_Interface_SCSI 0x100
212 #define DDF_Interface_SAS 0x200
213 #define DDF_Interface_SATA 0x300
214 #define DDF_Interface_FC 0x400
216 /* phys_disk_entry.state is a bigendian bitmap */
218 #define DDF_Failed 2 /* overrides 1,4,8 */
219 #define DDF_Rebuilding 4
220 #define DDF_Transition 8
222 #define DDF_ReadErrors 32
223 #define DDF_Missing 64
225 /* The content of the virt_section global scope */
226 struct virtual_disk
{
229 __u16 populated_vdes
;
232 struct virtual_entry
{
233 char guid
[DDF_GUID_LEN
];
235 __u16 pad0
; /* 0xffff */
245 /* virtual_entry.type is a bitmap - bigendian */
247 #define DDF_Enforce_Groups 2
248 #define DDF_Unicode 4
249 #define DDF_Owner_Valid 8
251 /* virtual_entry.state is a bigendian bitmap */
252 #define DDF_state_mask 0x7
253 #define DDF_state_optimal 0x0
254 #define DDF_state_degraded 0x1
255 #define DDF_state_deleted 0x2
256 #define DDF_state_missing 0x3
257 #define DDF_state_failed 0x4
259 #define DDF_state_morphing 0x8
260 #define DDF_state_inconsistent 0x10
262 /* virtual_entry.init_state is a bigendian bitmap */
263 #define DDF_initstate_mask 0x03
264 #define DDF_init_not 0x00
265 #define DDF_init_quick 0x01
266 #define DDF_init_full 0x02
268 #define DDF_access_mask 0xc0
269 #define DDF_access_rw 0x00
270 #define DDF_access_ro 0x80
271 #define DDF_access_blocked 0xc0
273 /* The content of the config_section - local scope
274 * It has multiple records each config_record_len sectors
275 * They can be vd_config or spare_assign
281 char guid
[DDF_GUID_LEN
];
285 __u16 prim_elmnt_count
;
286 __u8 chunk_shift
; /* 0 == 512, 1==1024 etc */
289 __u8 sec_elmnt_count
;
292 __u64 blocks
; /* blocks per component could be different
293 * on different component devices...(only
294 * for concat I hope) */
295 __u64 array_blocks
; /* blocks in array */
303 __u8 v0
[32]; /* reserved- 0xff */
304 __u8 v1
[32]; /* reserved- 0xff */
305 __u8 v2
[16]; /* reserved- 0xff */
306 __u8 v3
[16]; /* reserved- 0xff */
308 __u32 phys_refnum
[0]; /* refnum of each disk in sequence */
309 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
310 bvd are always the same size */
313 /* vd_config.cache_pol[7] is a bitmap */
314 #define DDF_cache_writeback 1 /* else writethrough */
315 #define DDF_cache_wadaptive 2 /* only applies if writeback */
316 #define DDF_cache_readahead 4
317 #define DDF_cache_radaptive 8 /* only if doing read-ahead */
318 #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
319 #define DDF_cache_wallowed 32 /* enable write caching */
320 #define DDF_cache_rallowed 64 /* enable read caching */
322 struct spare_assign
{
328 __u16 populated
; /* SAEs used */
329 __u16 max
; /* max SAEs */
331 struct spare_assign_entry
{
332 char guid
[DDF_GUID_LEN
];
333 __u16 secondary_element
;
337 /* spare_assign.type is a bitmap */
338 #define DDF_spare_dedicated 0x1 /* else global */
339 #define DDF_spare_revertible 0x2 /* else committable */
340 #define DDF_spare_active 0x4 /* else not active */
341 #define DDF_spare_affinity 0x8 /* enclosure affinity */
343 /* The data_section contents - local scope */
347 char guid
[DDF_GUID_LEN
];
348 __u32 refnum
; /* crc of some magic drive data ... */
349 __u8 forced_ref
; /* set when above was not result of magic */
350 __u8 forced_guid
; /* set if guid was forced rather than magic */
355 /* bbm_section content */
356 struct bad_block_log
{
363 struct mapped_block
{
364 __u64 defective_start
;
365 __u32 replacement_start
;
371 /* Struct for internally holding ddf structures */
372 /* The DDF structure stored on each device is potentially
373 * quite different, as some data is global and some is local.
374 * The global data is:
377 * - Physical disk records
378 * - Virtual disk records
380 * - Configuration records
381 * - Physical Disk data section
382 * ( and Bad block and vendor which I don't care about yet).
384 * The local data is parsed into separate lists as it is read
385 * and reconstructed for writing. This means that we only need
386 * to make config changes once and they are automatically
387 * propagated to all devices.
388 * Note that the ddf_super has space of the conf and disk data
389 * for this disk and also for a list of all such data.
390 * The list is only used for the superblock that is being
391 * built in Create or Assemble to describe the whole array.
394 struct ddf_header anchor
, primary
, secondary
, *active
;
395 struct ddf_controller_data controller
;
396 struct phys_disk
*phys
;
397 struct virtual_disk
*virt
;
402 __u64
*lba_offset
; /* location in 'conf' of
404 struct vd_config conf
;
405 } *conflist
, *newconf
;
408 struct disk_data disk
;
412 struct vcl
*vlist
[0]; /* max_part+1 in size */
417 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
420 extern struct superswitch super_ddf_container
, super_ddf_bvd
, super_ddf
;
422 static int calc_crc(void *buf
, int len
)
424 /* crcs are always at the same place as in the ddf_header */
425 struct ddf_header
*ddf
= buf
;
426 __u32 oldcrc
= ddf
->crc
;
428 ddf
->crc
= 0xffffffff;
430 newcrc
= crc32(0, buf
, len
);
435 static int load_ddf_header(int fd
, unsigned long long lba
,
436 unsigned long long size
,
438 struct ddf_header
*hdr
, struct ddf_header
*anchor
)
440 /* read a ddf header (primary or secondary) from fd/lba
441 * and check that it is consistent with anchor
443 * magic, crc, guid, rev, and LBA's header_type, and
444 * everything after header_type must be the same
449 if (lseek64(fd
, lba
<<9, 0) < 0)
452 if (read(fd
, hdr
, 512) != 512)
455 if (hdr
->magic
!= DDF_HEADER_MAGIC
)
457 if (calc_crc(hdr
, 512) != hdr
->crc
)
459 if (memcmp(anchor
->guid
, hdr
->guid
, DDF_GUID_LEN
) != 0 ||
460 memcmp(anchor
->revision
, hdr
->revision
, 8) != 0 ||
461 anchor
->primary_lba
!= hdr
->primary_lba
||
462 anchor
->secondary_lba
!= hdr
->secondary_lba
||
464 memcmp(anchor
->pad2
, hdr
->pad2
, 512 -
465 offsetof(struct ddf_header
, pad2
)) != 0)
468 /* Looks good enough to me... */
472 static void *load_section(int fd
, struct ddf_super
*super
, void *buf
,
473 __u32 offset_be
, __u32 len_be
, int check
)
475 unsigned long long offset
= __be32_to_cpu(offset_be
);
476 unsigned long long len
= __be32_to_cpu(len_be
);
477 int dofree
= (buf
== NULL
);
480 if (len
!= 2 && len
!= 8 && len
!= 32
481 && len
!= 128 && len
!= 512)
487 /* All pre-allocated sections are a single block */
491 buf
= malloc(len
<<9);
495 if (super
->active
->type
== 1)
496 offset
+= __be64_to_cpu(super
->active
->primary_lba
);
498 offset
+= __be64_to_cpu(super
->active
->secondary_lba
);
500 if (lseek64(fd
, offset
<<9, 0) != (offset
<<9)) {
505 if (read(fd
, buf
, len
<<9) != (len
<<9)) {
513 static int load_ddf_headers(int fd
, struct ddf_super
*super
, char *devname
)
515 unsigned long long dsize
;
517 get_dev_size(fd
, NULL
, &dsize
);
519 if (lseek64(fd
, dsize
-512, 0) < 0) {
522 Name
": Cannot seek to anchor block on %s: %s\n",
523 devname
, strerror(errno
));
526 if (read(fd
, &super
->anchor
, 512) != 512) {
529 Name
": Cannot read anchor block on %s: %s\n",
530 devname
, strerror(errno
));
533 if (super
->anchor
.magic
!= DDF_HEADER_MAGIC
) {
535 fprintf(stderr
, Name
": no DDF anchor found on %s\n",
539 if (calc_crc(&super
->anchor
, 512) != super
->anchor
.crc
) {
541 fprintf(stderr
, Name
": bad CRC on anchor on %s\n",
545 if (memcmp(super
->anchor
.revision
, DDF_REVISION
, 8) != 0) {
547 fprintf(stderr
, Name
": can only support super revision"
548 " %.8s, not %.8s on %s\n",
549 DDF_REVISION
, super
->anchor
.revision
, devname
);
552 if (load_ddf_header(fd
, __be64_to_cpu(super
->anchor
.primary_lba
),
554 &super
->primary
, &super
->anchor
) == 0) {
557 Name
": Failed to load primary DDF header "
561 super
->active
= &super
->primary
;
562 if (load_ddf_header(fd
, __be64_to_cpu(super
->anchor
.secondary_lba
),
564 &super
->secondary
, &super
->anchor
)) {
565 if ((__be32_to_cpu(super
->primary
.seq
)
566 < __be32_to_cpu(super
->secondary
.seq
) &&
567 !super
->secondary
.openflag
)
568 || (__be32_to_cpu(super
->primary
.seq
)
569 == __be32_to_cpu(super
->secondary
.seq
) &&
570 super
->primary
.openflag
&& !super
->secondary
.openflag
)
572 super
->active
= &super
->secondary
;
577 static int load_ddf_global(int fd
, struct ddf_super
*super
, char *devname
)
580 ok
= load_section(fd
, super
, &super
->controller
,
581 super
->active
->controller_section_offset
,
582 super
->active
->controller_section_length
,
584 super
->phys
= load_section(fd
, super
, NULL
,
585 super
->active
->phys_section_offset
,
586 super
->active
->phys_section_length
,
588 super
->pdsize
= __be32_to_cpu(super
->active
->phys_section_length
) * 512;
590 super
->virt
= load_section(fd
, super
, NULL
,
591 super
->active
->virt_section_offset
,
592 super
->active
->virt_section_length
,
594 super
->vdsize
= __be32_to_cpu(super
->active
->virt_section_length
) * 512;
604 super
->conflist
= NULL
;
609 static int load_ddf_local(int fd
, struct ddf_super
*super
,
610 char *devname
, int keep
)
619 /* First the local disk info */
620 super
->max_part
= __be16_to_cpu(super
->active
->max_partitions
);
621 dl
= malloc(sizeof(*dl
) +
622 (super
->max_part
+1) * sizeof(dl
->vlist
[0]));
624 load_section(fd
, super
, &dl
->disk
,
625 super
->active
->data_section_offset
,
626 super
->active
->data_section_length
,
628 dl
->devname
= devname
? strdup(devname
) : NULL
;
631 dl
->major
= major(stb
.st_rdev
);
632 dl
->minor
= minor(stb
.st_rdev
);
633 dl
->next
= super
->dlist
;
634 dl
->fd
= keep
? fd
: -1;
635 for (i
=0 ; i
< super
->max_part
+ 1 ; i
++)
639 /* Now the config list. */
640 /* 'conf' is an array of config entries, some of which are
641 * probably invalid. Those which are good need to be copied into
644 conflen
= __be16_to_cpu(super
->active
->config_record_len
);
646 conf
= load_section(fd
, super
, NULL
,
647 super
->active
->config_section_offset
,
648 super
->active
->config_section_length
,
652 i
< __be32_to_cpu(super
->active
->config_section_length
);
654 struct vd_config
*vd
=
655 (struct vd_config
*)((char*)conf
+ i
*512);
658 if (vd
->magic
!= DDF_VD_CONF_MAGIC
)
660 for (vcl
= super
->conflist
; vcl
; vcl
= vcl
->next
) {
661 if (memcmp(vcl
->conf
.guid
,
662 vd
->guid
, DDF_GUID_LEN
) == 0)
667 dl
->vlist
[i
/conflen
] = vcl
;
668 if (__be32_to_cpu(vd
->seqnum
) <=
669 __be32_to_cpu(vcl
->conf
.seqnum
))
672 vcl
= malloc(conflen
*512 + offsetof(struct vcl
, conf
));
673 vcl
->next
= super
->conflist
;
674 super
->conflist
= vcl
;
676 memcpy(&vcl
->conf
, vd
, conflen
*512);
677 mppe
= __be16_to_cpu(super
->anchor
.max_primary_element_entries
);
678 vcl
->lba_offset
= (__u64
*)
679 &vcl
->conf
.phys_refnum
[mppe
];
680 dl
->vlist
[i
/conflen
] = vcl
;
688 static int load_super_ddf_all(struct supertype
*st
, int fd
,
689 void **sbp
, char *devname
, int keep_fd
);
691 static int load_super_ddf(struct supertype
*st
, int fd
,
694 unsigned long long dsize
;
695 struct ddf_super
*super
;
699 if (load_super_ddf_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
703 if (get_dev_size(fd
, devname
, &dsize
) == 0)
706 /* 32M is a lower bound */
707 if (dsize
<= 32*1024*1024) {
710 Name
": %s is too small for ddf: "
711 "size is %llu sectors.\n",
719 Name
": %s is an odd size for ddf: "
720 "size is %llu bytes.\n",
726 super
= malloc(sizeof(*super
));
728 fprintf(stderr
, Name
": malloc of %zu failed.\n",
732 memset(super
, 0, sizeof(*super
));
734 rv
= load_ddf_headers(fd
, super
, devname
);
740 /* Have valid headers and have chosen the best. Let's read in the rest*/
742 rv
= load_ddf_global(fd
, super
, devname
);
747 Name
": Failed to load all information "
748 "sections on %s\n", devname
);
753 load_ddf_local(fd
, super
, devname
, 0);
755 /* Should possibly check the sections .... */
758 if (st
->ss
== NULL
) {
760 st
->minor_version
= 0;
767 static void free_super_ddf(struct supertype
*st
)
769 struct ddf_super
*ddf
= st
->sb
;
774 while (ddf
->conflist
) {
775 struct vcl
*v
= ddf
->conflist
;
776 ddf
->conflist
= v
->next
;
780 struct dl
*d
= ddf
->dlist
;
781 ddf
->dlist
= d
->next
;
790 static struct supertype
*match_metadata_desc_ddf(char *arg
)
792 /* 'ddf' only support containers */
793 struct supertype
*st
;
794 if (strcmp(arg
, "ddf") != 0 &&
795 strcmp(arg
, "default") != 0
799 st
= malloc(sizeof(*st
));
802 st
->minor_version
= 0;
807 static struct supertype
*match_metadata_desc_ddf_bvd(char *arg
)
809 struct supertype
*st
;
810 if (strcmp(arg
, "ddf/bvd") != 0 &&
811 strcmp(arg
, "bvd") != 0 &&
812 strcmp(arg
, "default") != 0
816 st
= malloc(sizeof(*st
));
817 st
->ss
= &super_ddf_bvd
;
819 st
->minor_version
= 0;
823 static struct supertype
*match_metadata_desc_ddf_svd(char *arg
)
825 struct supertype
*st
;
826 if (strcmp(arg
, "ddf/svd") != 0 &&
827 strcmp(arg
, "svd") != 0 &&
828 strcmp(arg
, "default") != 0
832 st
= malloc(sizeof(*st
));
833 st
->ss
= &super_ddf_svd
;
835 st
->minor_version
= 0;
842 static mapping_t ddf_state
[] = {
848 { "Partially Optimal", 5},
854 static mapping_t ddf_init_state
[] = {
855 { "Not Initialised", 0},
856 { "QuickInit in Progress", 1},
857 { "Fully Initialised", 2},
861 static mapping_t ddf_access
[] = {
865 { "Blocked (no access)", 3},
869 static mapping_t ddf_level
[] = {
870 { "RAID0", DDF_RAID0
},
871 { "RAID1", DDF_RAID1
},
872 { "RAID3", DDF_RAID3
},
873 { "RAID4", DDF_RAID4
},
874 { "RAID5", DDF_RAID5
},
875 { "RAID1E",DDF_RAID1E
},
877 { "CONCAT",DDF_CONCAT
},
878 { "RAID5E",DDF_RAID5E
},
879 { "RAID5EE",DDF_RAID5EE
},
880 { "RAID6", DDF_RAID6
},
883 static mapping_t ddf_sec_level
[] = {
884 { "Striped", DDF_2STRIPED
},
885 { "Mirrored", DDF_2MIRRORED
},
886 { "Concat", DDF_2CONCAT
},
887 { "Spanned", DDF_2SPANNED
},
895 static struct num_mapping ddf_level_num
[] = {
898 { DDF_RAID3
, LEVEL_UNSUPPORTED
},
900 { DDF_RAID1E
, LEVEL_UNSUPPORTED
},
901 { DDF_JBOD
, LEVEL_UNSUPPORTED
},
902 { DDF_CONCAT
, LEVEL_LINEAR
},
903 { DDF_RAID5E
, LEVEL_UNSUPPORTED
},
904 { DDF_RAID5EE
, LEVEL_UNSUPPORTED
},
909 static int map_num1(struct num_mapping
*map
, int num
)
912 for (i
=0 ; map
[i
].num1
!= MAXINT
; i
++)
913 if (map
[i
].num1
== num
)
919 static void print_guid(char *guid
, int tstamp
)
921 /* A GUIDs are part (or all) ASCII and part binary.
922 * They tend to be space padded.
923 * We ignore trailing spaces and print numbers
924 * <0x20 and >=0x7f as \xXX
925 * Some GUIDs have a time stamp in bytes 16-19.
926 * We print that if appropriate
928 int l
= DDF_GUID_LEN
;
930 while (l
&& guid
[l
-1] == ' ')
932 for (i
=0 ; i
<l
; i
++) {
933 if (guid
[i
] >= 0x20 && guid
[i
] < 0x7f)
934 fputc(guid
[i
], stdout
);
936 fprintf(stdout
, "\\x%02x", guid
[i
]&255);
939 time_t then
= __be32_to_cpu(*(__u32
*)(guid
+16)) + DECADE
;
942 tm
= localtime(&then
);
943 strftime(tbuf
, 100, " (%D %T)",tm
);
948 static void examine_vd(int n
, struct ddf_super
*sb
, char *guid
)
950 int crl
= __be16_to_cpu(sb
->anchor
.config_record_len
);
953 for (vcl
= sb
->conflist
; vcl
; vcl
= vcl
->next
) {
954 struct vd_config
*vc
= &vcl
->conf
;
956 if (calc_crc(vc
, crl
*512) != vc
->crc
)
958 if (memcmp(vc
->guid
, guid
, DDF_GUID_LEN
) != 0)
961 /* Ok, we know about this VD, let's give more details */
962 printf(" Raid Devices[%d] : %d\n", n
,
963 __be16_to_cpu(vc
->prim_elmnt_count
));
964 printf(" Chunk Size[%d] : %d sectors\n", n
,
965 1 << vc
->chunk_shift
);
966 printf(" Raid Level[%d] : %s\n", n
,
967 map_num(ddf_level
, vc
->prl
)?:"-unknown-");
968 if (vc
->sec_elmnt_count
!= 1) {
969 printf(" Secondary Position[%d] : %d of %d\n", n
,
970 vc
->sec_elmnt_seq
, vc
->sec_elmnt_count
);
971 printf(" Secondary Level[%d] : %s\n", n
,
972 map_num(ddf_sec_level
, vc
->srl
) ?: "-unknown-");
974 printf(" Device Size[%d] : %llu\n", n
,
975 __be64_to_cpu(vc
->blocks
)/2);
976 printf(" Array Size[%d] : %llu\n", n
,
977 __be64_to_cpu(vc
->array_blocks
)/2);
981 static void examine_vds(struct ddf_super
*sb
)
983 int cnt
= __be16_to_cpu(sb
->virt
->populated_vdes
);
985 printf(" Virtual Disks : %d\n", cnt
);
987 for (i
=0; i
<cnt
; i
++) {
988 struct virtual_entry
*ve
= &sb
->virt
->entries
[i
];
989 printf(" VD GUID[%d] : ", i
); print_guid(ve
->guid
, 1);
991 printf(" unit[%d] : %d\n", i
, __be16_to_cpu(ve
->unit
));
992 printf(" state[%d] : %s, %s%s\n", i
,
993 map_num(ddf_state
, ve
->state
& 7),
994 (ve
->state
& 8) ? "Morphing, ": "",
995 (ve
->state
& 16)? "Not Consistent" : "Consistent");
996 printf(" init state[%d] : %s\n", i
,
997 map_num(ddf_init_state
, ve
->init_state
&3));
998 printf(" access[%d] : %s\n", i
,
999 map_num(ddf_access
, (ve
->init_state
>>6) & 3));
1000 printf(" Name[%d] : %.16s\n", i
, ve
->name
);
1001 examine_vd(i
, sb
, ve
->guid
);
1003 if (cnt
) printf("\n");
1006 static void examine_pds(struct ddf_super
*sb
)
1008 int cnt
= __be16_to_cpu(sb
->phys
->used_pdes
);
1011 printf(" Physical Disks : %d\n", cnt
);
1013 for (i
=0 ; i
<cnt
; i
++) {
1014 struct phys_disk_entry
*pd
= &sb
->phys
->entries
[i
];
1015 int type
= __be16_to_cpu(pd
->type
);
1016 int state
= __be16_to_cpu(pd
->state
);
1018 printf(" PD GUID[%d] : ", i
); print_guid(pd
->guid
, 0);
1020 printf(" ref[%d] : %08x\n", i
,
1021 __be32_to_cpu(pd
->refnum
));
1022 printf(" mode[%d] : %s%s%s%s%s\n", i
,
1023 (type
&2) ? "active":"",
1024 (type
&4) ? "Global Spare":"",
1025 (type
&8) ? "spare" : "",
1026 (type
&16)? ", foreign" : "",
1027 (type
&32)? "pass-through" : "");
1028 printf(" state[%d] : %s%s%s%s%s%s%s\n", i
,
1029 (state
&1)? "Online": "Offline",
1030 (state
&2)? ", Failed": "",
1031 (state
&4)? ", Rebuilding": "",
1032 (state
&8)? ", in-transition": "",
1033 (state
&16)? ", SMART errors": "",
1034 (state
&32)? ", Unrecovered Read Errors": "",
1035 (state
&64)? ", Missing" : "");
1036 printf(" Avail Size[%d] : %llu K\n", i
,
1037 __be64_to_cpu(pd
->config_size
)>>1);
1038 for (dl
= sb
->dlist
; dl
; dl
= dl
->next
) {
1039 if (dl
->disk
.refnum
== pd
->refnum
) {
1040 char *dv
= map_dev(dl
->major
, dl
->minor
, 0);
1042 printf(" Device[%d] : %s\n",
1050 static void examine_super_ddf(struct supertype
*st
, char *homehost
)
1052 struct ddf_super
*sb
= st
->sb
;
1054 printf(" Magic : %08x\n", __be32_to_cpu(sb
->anchor
.magic
));
1055 printf(" Version : %.8s\n", sb
->anchor
.revision
);
1056 printf("Controller GUID : "); print_guid(sb
->controller
.guid
, 0);
1058 printf(" Container GUID : "); print_guid(sb
->anchor
.guid
, 1);
1060 printf(" Seq : %08x\n", __be32_to_cpu(sb
->active
->seq
));
1061 printf(" Redundant hdr : %s\n", sb
->secondary
.magic
== DDF_HEADER_MAGIC
1067 static void brief_examine_super_ddf(struct supertype
*st
)
1069 /* We just write a generic DDF ARRAY entry
1070 * The uuid is all hex, 6 groups of 4 bytes
1072 struct ddf_super
*ddf
= st
->sb
;
1074 printf("ARRAY /dev/ddf UUID=");
1075 for (i
= 0; i
< DDF_GUID_LEN
; i
++) {
1076 printf("%02x", ddf
->anchor
.guid
[i
]);
1077 if ((i
&3) == 0 && i
!= 0)
1083 static void detail_super_ddf(struct supertype
*st
, char *homehost
)
1086 * Could print DDF GUID
1087 * Need to find which array
1088 * If whole, briefly list all arrays
1093 static void brief_detail_super_ddf(struct supertype
*st
)
1095 /* FIXME I really need to know which array we are detailing.
1096 * Can that be stored in ddf_super??
1098 // struct ddf_super *ddf = st->sb;
1104 static int match_home_ddf(struct supertype
*st
, char *homehost
)
1106 /* It matches 'this' host if the controller is a
1107 * Linux-MD controller with vendor_data matching
1110 struct ddf_super
*ddf
= st
->sb
;
1111 int len
= strlen(homehost
);
1113 return (memcmp(ddf
->controller
.guid
, T10
, 8) == 0 &&
1114 len
< sizeof(ddf
->controller
.vendor_data
) &&
1115 memcmp(ddf
->controller
.vendor_data
, homehost
,len
) == 0 &&
1116 ddf
->controller
.vendor_data
[len
] == 0);
1119 static struct vd_config
*find_vdcr(struct ddf_super
*ddf
)
1121 /* FIXME this just picks off the first one */
1122 return &ddf
->conflist
->conf
;
1125 static void uuid_from_super_ddf(struct supertype
*st
, int uuid
[4])
1127 /* The uuid returned here is used for:
1128 * uuid to put into bitmap file (Create, Grow)
1129 * uuid for backup header when saving critical section (Grow)
1130 * comparing uuids when re-adding a device into an array
1131 * For each of these we can make do with a truncated
1132 * or hashed uuid rather than the original, as long as
1134 * In each case the uuid required is that of the data-array,
1135 * not the device-set.
1136 * In the case of SVD we assume the BVD is of interest,
1137 * though that might be the case if a bitmap were made for
1138 * a mirrored SVD - worry about that later.
1139 * So we need to find the VD configuration record for the
1140 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1141 * The first 16 bytes of the sha1 of these is used.
1143 struct ddf_super
*ddf
= st
->sb
;
1144 struct vd_config
*vd
= find_vdcr(ddf
);
1147 memset(uuid
, 0, sizeof (uuid
));
1150 struct sha1_ctx ctx
;
1151 sha1_init_ctx(&ctx
);
1152 sha1_process_bytes(&vd
->guid
, DDF_GUID_LEN
, &ctx
);
1153 if (vd
->sec_elmnt_count
> 1)
1154 sha1_process_bytes(&vd
->sec_elmnt_seq
, 1, &ctx
);
1155 sha1_finish_ctx(&ctx
, buf
);
1156 memcpy(uuid
, buf
, sizeof(uuid
));
1160 static void getinfo_super_ddf(struct supertype
*st
, struct mdinfo
*info
)
1162 struct ddf_super
*ddf
= st
->sb
;
1165 info
->array
.major_version
= 1000;
1166 info
->array
.minor_version
= 0; /* FIXME use ddf->revision somehow */
1167 info
->array
.patch_version
= 0;
1168 info
->array
.raid_disks
= __be16_to_cpu(ddf
->phys
->used_pdes
);
1169 info
->array
.level
= LEVEL_CONTAINER
;
1170 info
->array
.layout
= 0;
1171 info
->array
.md_minor
= -1;
1172 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*(__u32
*)
1173 (ddf
->anchor
.guid
+16));
1174 info
->array
.utime
= 0;
1175 info
->array
.chunk_size
= 0;
1177 // info->data_offset = ???;
1178 // info->component_size = ???;
1180 info
->disk
.major
= 0;
1181 info
->disk
.minor
= 0;
1182 info
->disk
.number
= __be32_to_cpu(ddf
->dlist
->disk
.refnum
);
1183 // info->disk.raid_disk = find refnum in the table and use index;
1184 info
->disk
.raid_disk
= -1;
1185 for (i
= 0; i
< __be16_to_cpu(ddf
->phys
->max_pdes
) ; i
++)
1186 if (ddf
->phys
->entries
[i
].refnum
== ddf
->dlist
->disk
.refnum
) {
1187 info
->disk
.raid_disk
= i
;
1190 info
->disk
.state
= (1 << MD_DISK_SYNC
);
1192 info
->reshape_active
= 0;
1194 // uuid_from_super_ddf(info->uuid, sbv);
1196 // info->name[] ?? ;
1199 static void getinfo_super_n_container(struct supertype
*st
, struct mdinfo
*info
)
1201 /* just need offset and size */
1202 struct ddf_super
*ddf
= st
->sb
;
1203 int n
= info
->disk
.number
;
1205 info
->data_offset
= __be64_to_cpu(ddf
->phys
->entries
[n
].config_size
);
1206 info
->component_size
= 32*1024*1024 / 512;
1209 static int rlq_to_layout(int rlq
, int prl
, int raiddisks
);
1211 static void getinfo_super_ddf_bvd(struct supertype
*st
, struct mdinfo
*info
)
1213 struct ddf_super
*ddf
= st
->sb
;
1214 struct vd_config
*vd
= find_vdcr(ddf
);
1216 /* FIXME this returns BVD info - what if we want SVD ?? */
1218 info
->array
.major_version
= 1000;
1219 info
->array
.minor_version
= 0; /* FIXME use ddf->revision somehow */
1220 info
->array
.patch_version
= 0;
1221 info
->array
.raid_disks
= __be16_to_cpu(vd
->prim_elmnt_count
);
1222 info
->array
.level
= map_num1(ddf_level_num
, vd
->prl
);
1223 info
->array
.layout
= rlq_to_layout(vd
->rlq
, vd
->prl
,
1224 info
->array
.raid_disks
);
1225 info
->array
.md_minor
= -1;
1226 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*(__u32
*)(vd
->guid
+16));
1227 info
->array
.utime
= DECADE
+ __be32_to_cpu(vd
->timestamp
);
1228 info
->array
.chunk_size
= 512 << vd
->chunk_shift
;
1230 // info->data_offset = ???;
1231 // info->component_size = ???;
1233 info
->disk
.major
= 0;
1234 info
->disk
.minor
= 0;
1235 // info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1236 // info->disk.raid_disk = find refnum in the table and use index;
1237 // info->disk.state = ???;
1239 uuid_from_super_ddf(st
, info
->uuid
);
1241 // info->name[] ?? ;
1244 static void getinfo_super_n_bvd(struct supertype
*st
, struct mdinfo
*info
)
1246 /* Find the particular details for info->disk.raid_disk.
1247 * This includes data_offset, component_size,
1249 struct ddf_super
*ddf
= st
->sb
;
1250 __u64
*lba_offset
= ddf
->newconf
->lba_offset
;
1251 struct vd_config
*conf
= &ddf
->newconf
->conf
;
1252 info
->data_offset
= __be64_to_cpu(lba_offset
[info
->disk
.raid_disk
]);
1253 info
->component_size
= __be64_to_cpu(conf
->blocks
);
1256 static int update_super_ddf(struct supertype
*st
, struct mdinfo
*info
,
1258 char *devname
, int verbose
,
1259 int uuid_set
, char *homehost
)
1261 /* For 'assemble' and 'force' we need to return non-zero if any
1262 * change was made. For others, the return value is ignored.
1263 * Update options are:
1264 * force-one : This device looks a bit old but needs to be included,
1265 * update age info appropriately.
1266 * assemble: clear any 'faulty' flag to allow this device to
1268 * force-array: Array is degraded but being forced, mark it clean
1269 * if that will be needed to assemble it.
1271 * newdev: not used ????
1272 * grow: Array has gained a new device - this is currently for
1274 * resync: mark as dirty so a resync will happen.
1275 * uuid: Change the uuid of the array to match watch is given
1276 * homehost: update the recorded homehost
1277 * name: update the name - preserving the homehost
1278 * _reshape_progress: record new reshape_progress position.
1280 * Following are not relevant for this version:
1281 * sparc2.2 : update from old dodgey metadata
1282 * super-minor: change the preferred_minor number
1283 * summaries: update redundant counters.
1286 // struct ddf_super *ddf = st->sb;
1287 // struct vd_config *vd = find_vdcr(ddf);
1288 // struct virtual_entry *ve = find_ve(ddf);
1291 /* we don't need to handle "force-*" or "assemble" as
1292 * there is no need to 'trick' the kernel. We the metadata is
1293 * first updated to activate the array, all the implied modifications
1297 if (strcmp(update
, "grow") == 0) {
1300 if (strcmp(update
, "resync") == 0) {
1301 // info->resync_checkpoint = 0;
1303 /* We ignore UUID updates as they make even less sense
1306 if (strcmp(update
, "homehost") == 0) {
1307 /* homehost is stored in controller->vendor_data,
1308 * or it is when we are the vendor
1310 // if (info->vendor_is_local)
1311 // strcpy(ddf->controller.vendor_data, homehost);
1313 if (strcmp(update
, "name") == 0) {
1314 /* name is stored in virtual_entry->name */
1315 // memset(ve->name, ' ', 16);
1316 // strncpy(ve->name, info->name, 16);
1318 if (strcmp(update
, "_reshape_progress") == 0) {
1319 /* We don't support reshape yet */
1322 // update_all_csum(ddf);
1327 static void make_header_guid(char *guid
)
1331 /* Create a DDF Header of Virtual Disk GUID */
1333 /* 24 bytes of fiction required.
1334 * first 8 are a 'vendor-id' - "Linux-MD"
1335 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1336 * Remaining 8 random number plus timestamp
1338 memcpy(guid
, T10
, sizeof(T10
));
1339 stamp
= __cpu_to_be32(0xdeadbeef);
1340 memcpy(guid
+8, &stamp
, 4);
1341 stamp
= __cpu_to_be32(0);
1342 memcpy(guid
+12, &stamp
, 4);
1343 stamp
= __cpu_to_be32(time(0) - DECADE
);
1344 memcpy(guid
+16, &stamp
, 4);
1345 rfd
= open("/dev/urandom", O_RDONLY
);
1346 if (rfd
< 0 || read(rfd
, &stamp
, 4) != 4)
1348 memcpy(guid
+20, &stamp
, 4);
1349 if (rfd
>= 0) close(rfd
);
1351 static int init_super_ddf(struct supertype
*st
,
1352 mdu_array_info_t
*info
,
1353 unsigned long long size
, char *name
, char *homehost
,
1356 /* This is primarily called by Create when creating a new array.
1357 * We will then get add_to_super called for each component, and then
1358 * write_init_super called to write it out to each device.
1359 * For DDF, Create can create on fresh devices or on a pre-existing
1361 * To create on a pre-existing array a different method will be called.
1362 * This one is just for fresh drives.
1364 * We need to create the entire 'ddf' structure which includes:
1365 * DDF headers - these are easy.
1366 * Controller data - a Sector describing this controller .. not that
1367 * this is a controller exactly.
1368 * Physical Disk Record - one entry per device, so
1369 * leave plenty of space.
1370 * Virtual Disk Records - again, just leave plenty of space.
1371 * This just lists VDs, doesn't give details
1372 * Config records - describes the VDs that use this disk
1373 * DiskData - describes 'this' device.
1374 * BadBlockManagement - empty
1375 * Diag Space - empty
1376 * Vendor Logs - Could we put bitmaps here?
1379 struct ddf_super
*ddf
;
1382 int max_phys_disks
, max_virt_disks
;
1383 unsigned long long sector
;
1387 struct phys_disk
*pd
;
1388 struct virtual_disk
*vd
;
1390 ddf
= malloc(sizeof(*ddf
));
1391 ddf
->dlist
= NULL
; /* no physical disks yet */
1392 ddf
->conflist
= NULL
; /* No virtual disks yet */
1394 /* At least 32MB *must* be reserved for the ddf. So let's just
1395 * start 32MB from the end, and put the primary header there.
1396 * Don't do secondary for now.
1397 * We don't know exactly where that will be yet as it could be
1398 * different on each device. To just set up the lengths.
1402 ddf
->anchor
.magic
= DDF_HEADER_MAGIC
;
1403 make_header_guid(ddf
->anchor
.guid
);
1405 memcpy(ddf
->anchor
.revision
, DDF_REVISION
, 8);
1406 ddf
->anchor
.seq
= __cpu_to_be32(1);
1407 ddf
->anchor
.timestamp
= __cpu_to_be32(time(0) - DECADE
);
1408 ddf
->anchor
.openflag
= 0xFF;
1409 ddf
->anchor
.foreignflag
= 0;
1410 ddf
->anchor
.enforcegroups
= 0; /* Is this best?? */
1411 ddf
->anchor
.pad0
= 0xff;
1412 memset(ddf
->anchor
.pad1
, 0xff, 12);
1413 memset(ddf
->anchor
.header_ext
, 0xff, 32);
1414 ddf
->anchor
.primary_lba
= ~(__u64
)0;
1415 ddf
->anchor
.secondary_lba
= ~(__u64
)0;
1416 ddf
->anchor
.type
= DDF_HEADER_ANCHOR
;
1417 memset(ddf
->anchor
.pad2
, 0xff, 3);
1418 ddf
->anchor
.workspace_len
= __cpu_to_be32(32768); /* Must be reserved */
1419 ddf
->anchor
.workspace_lba
= ~(__u64
)0; /* Put this at bottom
1420 of 32M reserved.. */
1421 max_phys_disks
= 1023; /* Should be enough */
1422 ddf
->anchor
.max_pd_entries
= __cpu_to_be16(max_phys_disks
);
1423 max_virt_disks
= 255;
1424 ddf
->anchor
.max_vd_entries
= __cpu_to_be16(max_virt_disks
); /* ?? */
1425 ddf
->anchor
.max_partitions
= __cpu_to_be16(64); /* ?? */
1427 ddf
->anchor
.config_record_len
= __cpu_to_be16(1 + 256*12/512);
1428 ddf
->anchor
.max_primary_element_entries
= __cpu_to_be16(256);
1429 memset(ddf
->anchor
.pad3
, 0xff, 54);
1431 /* controller sections is one sector long immediately
1432 * after the ddf header */
1434 ddf
->anchor
.controller_section_offset
= __cpu_to_be32(sector
);
1435 ddf
->anchor
.controller_section_length
= __cpu_to_be32(1);
1438 /* phys is 8 sectors after that */
1439 pdsize
= ROUND_UP(sizeof(struct phys_disk
) +
1440 sizeof(struct phys_disk_entry
)*max_phys_disks
,
1442 switch(pdsize
/512) {
1443 case 2: case 8: case 32: case 128: case 512: break;
1446 ddf
->anchor
.phys_section_offset
= __cpu_to_be32(sector
);
1447 ddf
->anchor
.phys_section_length
=
1448 __cpu_to_be32(pdsize
/512); /* max_primary_element_entries/8 */
1449 sector
+= pdsize
/512;
1451 /* virt is another 32 sectors */
1452 vdsize
= ROUND_UP(sizeof(struct virtual_disk
) +
1453 sizeof(struct virtual_entry
) * max_virt_disks
,
1455 switch(vdsize
/512) {
1456 case 2: case 8: case 32: case 128: case 512: break;
1459 ddf
->anchor
.virt_section_offset
= __cpu_to_be32(sector
);
1460 ddf
->anchor
.virt_section_length
=
1461 __cpu_to_be32(vdsize
/512); /* max_vd_entries/8 */
1462 sector
+= vdsize
/512;
1464 clen
= (1 + 256*12/512) * (64+1);
1465 ddf
->anchor
.config_section_offset
= __cpu_to_be32(sector
);
1466 ddf
->anchor
.config_section_length
= __cpu_to_be32(clen
);
1469 ddf
->anchor
.data_section_offset
= __cpu_to_be32(sector
);
1470 ddf
->anchor
.data_section_length
= __cpu_to_be32(1);
1473 ddf
->anchor
.bbm_section_length
= __cpu_to_be32(0);
1474 ddf
->anchor
.bbm_section_offset
= __cpu_to_be32(0xFFFFFFFF);
1475 ddf
->anchor
.diag_space_length
= __cpu_to_be32(0);
1476 ddf
->anchor
.diag_space_offset
= __cpu_to_be32(0xFFFFFFFF);
1477 ddf
->anchor
.vendor_length
= __cpu_to_be32(0);
1478 ddf
->anchor
.vendor_offset
= __cpu_to_be32(0xFFFFFFFF);
1480 memset(ddf
->anchor
.pad4
, 0xff, 256);
1482 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
1483 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
1485 ddf
->primary
.openflag
= 1; /* I guess.. */
1486 ddf
->primary
.type
= DDF_HEADER_PRIMARY
;
1488 ddf
->secondary
.openflag
= 1; /* I guess.. */
1489 ddf
->secondary
.type
= DDF_HEADER_SECONDARY
;
1491 ddf
->active
= &ddf
->primary
;
1493 ddf
->controller
.magic
= DDF_CONTROLLER_MAGIC
;
1495 /* 24 more bytes of fiction required.
1496 * first 8 are a 'vendor-id' - "Linux-MD"
1497 * Remaining 16 are serial number.... maybe a hostname would do?
1499 memcpy(ddf
->controller
.guid
, T10
, sizeof(T10
));
1500 gethostname(hostname
, 17);
1502 hostlen
= strlen(hostname
);
1503 memcpy(ddf
->controller
.guid
+ 24 - hostlen
, hostname
, hostlen
);
1504 for (i
= strlen(T10
) ; i
+hostlen
< 24; i
++)
1505 ddf
->controller
.guid
[i
] = ' ';
1507 ddf
->controller
.type
.vendor_id
= __cpu_to_be16(0xDEAD);
1508 ddf
->controller
.type
.device_id
= __cpu_to_be16(0xBEEF);
1509 ddf
->controller
.type
.sub_vendor_id
= 0;
1510 ddf
->controller
.type
.sub_device_id
= 0;
1511 memcpy(ddf
->controller
.product_id
, "What Is My PID??", 16);
1512 memset(ddf
->controller
.pad
, 0xff, 8);
1513 memset(ddf
->controller
.vendor_data
, 0xff, 448);
1515 pd
= ddf
->phys
= malloc(pdsize
);
1516 ddf
->pdsize
= pdsize
;
1518 memset(pd
, 0xff, pdsize
);
1519 memset(pd
, 0, sizeof(*pd
));
1520 pd
->magic
= DDF_PHYS_DATA_MAGIC
;
1521 pd
->used_pdes
= __cpu_to_be16(0);
1522 pd
->max_pdes
= __cpu_to_be16(max_phys_disks
);
1523 memset(pd
->pad
, 0xff, 52);
1525 vd
= ddf
->virt
= malloc(vdsize
);
1526 ddf
->vdsize
= vdsize
;
1527 memset(vd
, 0, vdsize
);
1528 vd
->magic
= DDF_VIRT_RECORDS_MAGIC
;
1529 vd
->populated_vdes
= __cpu_to_be16(0);
1530 vd
->max_vdes
= __cpu_to_be16(max_virt_disks
);
1531 memset(vd
->pad
, 0xff, 52);
1533 for (i
=0; i
<max_virt_disks
; i
++)
1534 memset(&vd
->entries
[i
], 0xff, sizeof(struct virtual_entry
));
1540 static int all_ff(char *guid
)
1543 for (i
= 0; i
< DDF_GUID_LEN
; i
++)
1544 if (guid
[i
] != (char)0xff)
1548 static int chunk_to_shift(int chunksize
)
1550 return ffs(chunksize
/512)-1;
1553 static int level_to_prl(int level
)
1556 case LEVEL_LINEAR
: return DDF_CONCAT
;
1557 case 0: return DDF_RAID0
;
1558 case 1: return DDF_RAID1
;
1559 case 4: return DDF_RAID4
;
1560 case 5: return DDF_RAID5
;
1561 case 6: return DDF_RAID6
;
1565 static int layout_to_rlq(int level
, int layout
, int raiddisks
)
1569 return DDF_RAID0_SIMPLE
;
1572 case 2: return DDF_RAID1_SIMPLE
;
1573 case 3: return DDF_RAID1_MULTI
;
1578 case 0: return DDF_RAID4_N
;
1584 case ALGORITHM_LEFT_ASYMMETRIC
:
1585 return DDF_RAID5_N_RESTART
;
1586 case ALGORITHM_RIGHT_ASYMMETRIC
:
1587 return DDF_RAID5_0_RESTART
;
1588 case ALGORITHM_LEFT_SYMMETRIC
:
1589 return DDF_RAID5_N_CONTINUE
;
1590 case ALGORITHM_RIGHT_SYMMETRIC
:
1591 return -1; /* not mentioned in standard */
1597 static int rlq_to_layout(int rlq
, int prl
, int raiddisks
)
1601 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1603 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1611 return -1; /* FIXME this isn't checked */
1616 case DDF_RAID5_N_RESTART
:
1617 return ALGORITHM_LEFT_ASYMMETRIC
;
1618 case DDF_RAID5_0_RESTART
:
1619 return ALGORITHM_RIGHT_ASYMMETRIC
;
1620 case DDF_RAID5_N_CONTINUE
:
1621 return ALGORITHM_LEFT_SYMMETRIC
;
1629 static int init_super_ddf_bvd(struct supertype
*st
,
1630 mdu_array_info_t
*info
,
1631 unsigned long long size
,
1632 char *name
, char *homehost
,
1635 /* We are creating a BVD inside a pre-existing container.
1636 * so st->sb is already set.
1637 * We need to create a new vd_config and a new virtual_entry
1639 struct ddf_super
*ddf
= st
->sb
;
1641 struct virtual_entry
*ve
;
1643 struct vd_config
*vc
;
1647 if (__be16_to_cpu(ddf
->virt
->populated_vdes
)
1648 >= __be16_to_cpu(ddf
->virt
->max_vdes
)) {
1649 fprintf(stderr
, Name
": This ddf already has the "
1650 "maximum of %d virtual devices\n",
1651 __be16_to_cpu(ddf
->virt
->max_vdes
));
1655 for (venum
= 0; venum
< __be16_to_cpu(ddf
->virt
->max_vdes
); venum
++)
1656 if (all_ff(ddf
->virt
->entries
[venum
].guid
))
1658 if (venum
== __be16_to_cpu(ddf
->virt
->max_vdes
)) {
1659 fprintf(stderr
, Name
": Cannot find spare slot for "
1660 "virtual disk - DDF is corrupt\n");
1663 ve
= &ddf
->virt
->entries
[venum
];
1664 st
->container_member
= venum
;
1666 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1667 * timestamp, random number
1669 make_header_guid(ve
->guid
);
1670 ve
->unit
= __cpu_to_be16(info
->md_minor
);
1672 ve
->guid_crc
= crc32(0, (unsigned char*)ddf
->anchor
.guid
, DDF_GUID_LEN
);
1676 if (!(info
->state
& 1))
1677 ve
->init_state
= DDF_state_inconsistent
;
1678 memset(ve
->pad1
, 0xff, 14);
1679 memset(ve
->name
, ' ', 16);
1681 strncpy(ve
->name
, name
, 16);
1682 ddf
->virt
->populated_vdes
=
1683 __cpu_to_be16(__be16_to_cpu(ddf
->virt
->populated_vdes
)+1);
1685 /* Now create a new vd_config */
1686 conflen
= __be16_to_cpu(ddf
->active
->config_record_len
);
1687 vcl
= malloc(offsetof(struct vcl
, conf
) + conflen
* 512);
1688 mppe
= __be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
1689 vcl
->lba_offset
= (__u64
*) &vcl
->conf
.phys_refnum
[mppe
];
1693 vc
->magic
= DDF_VD_CONF_MAGIC
;
1694 memcpy(vc
->guid
, ve
->guid
, DDF_GUID_LEN
);
1695 vc
->timestamp
= __cpu_to_be32(time(0)-DECADE
);
1696 vc
->seqnum
= __cpu_to_be32(1);
1697 memset(vc
->pad0
, 0xff, 24);
1698 vc
->prim_elmnt_count
= __cpu_to_be16(info
->raid_disks
);
1699 vc
->chunk_shift
= chunk_to_shift(info
->chunk_size
);
1700 vc
->prl
= level_to_prl(info
->level
);
1701 vc
->rlq
= layout_to_rlq(info
->level
, info
->layout
, info
->raid_disks
);
1702 vc
->sec_elmnt_count
= 1;
1703 vc
->sec_elmnt_seq
= 0;
1705 vc
->blocks
= __cpu_to_be64(info
->size
* 2);
1706 vc
->array_blocks
= __cpu_to_be64(
1707 calc_array_size(info
->level
, info
->raid_disks
, info
->layout
,
1708 info
->chunk_size
, info
->size
*2));
1709 memset(vc
->pad1
, 0xff, 8);
1710 vc
->spare_refs
[0] = 0xffffffff;
1711 vc
->spare_refs
[1] = 0xffffffff;
1712 vc
->spare_refs
[2] = 0xffffffff;
1713 vc
->spare_refs
[3] = 0xffffffff;
1714 vc
->spare_refs
[4] = 0xffffffff;
1715 vc
->spare_refs
[5] = 0xffffffff;
1716 vc
->spare_refs
[6] = 0xffffffff;
1717 vc
->spare_refs
[7] = 0xffffffff;
1718 memset(vc
->cache_pol
, 0, 8);
1720 memset(vc
->pad2
, 0xff, 3);
1721 memset(vc
->pad3
, 0xff, 52);
1722 memset(vc
->pad4
, 0xff, 192);
1723 memset(vc
->v0
, 0xff, 32);
1724 memset(vc
->v1
, 0xff, 32);
1725 memset(vc
->v2
, 0xff, 16);
1726 memset(vc
->v3
, 0xff, 16);
1727 memset(vc
->vendor
, 0xff, 32);
1729 memset(vc
->phys_refnum
, 0xff, 4*mppe
);
1730 memset(vc
->phys_refnum
+mppe
, 0x00, 8*mppe
);
1732 vcl
->next
= ddf
->conflist
;
1733 ddf
->conflist
= vcl
;
1738 static void add_to_super_ddf_bvd(struct supertype
*st
,
1739 mdu_disk_info_t
*dk
, int fd
, char *devname
)
1741 /* fd and devname identify a device with-in the ddf container (st).
1742 * dk identifies a location in the new BVD.
1743 * We need to find suitable free space in that device and update
1744 * the phys_refnum and lba_offset for the newly created vd_config.
1745 * We might also want to update the type in the phys_disk
1749 struct ddf_super
*ddf
= st
->sb
;
1750 struct vd_config
*vc
;
1754 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
1755 if (dl
->major
== dk
->major
&&
1756 dl
->minor
== dk
->minor
)
1758 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1761 vc
= &ddf
->newconf
->conf
;
1762 vc
->phys_refnum
[dk
->raid_disk
] = dl
->disk
.refnum
;
1763 mppe
= __be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
1764 lba_offset
= (__u64
*)(vc
->phys_refnum
+ mppe
);
1765 lba_offset
[dk
->raid_disk
] = 0; /* FIXME */
1767 dl
->vlist
[0] =ddf
->newconf
; /* FIXME */
1770 dl
->devname
= devname
;
1773 /* add a device to a container, either while creating it or while
1774 * expanding a pre-existing container
1776 static void add_to_super_ddf(struct supertype
*st
,
1777 mdu_disk_info_t
*dk
, int fd
, char *devname
)
1779 struct ddf_super
*ddf
= st
->sb
;
1783 unsigned long long size
;
1784 struct phys_disk_entry
*pde
;
1788 /* This is device numbered dk->number. We need to create
1789 * a phys_disk entry and a more detailed disk_data entry.
1792 dd
= malloc(sizeof(*dd
) + sizeof(dd
->vlist
[0]) * (ddf
->max_part
+1));
1793 dd
->major
= major(stb
.st_rdev
);
1794 dd
->minor
= minor(stb
.st_rdev
);
1795 dd
->devname
= devname
;
1796 dd
->next
= ddf
->dlist
;
1799 dd
->disk
.magic
= DDF_PHYS_DATA_MAGIC
;
1801 tm
= localtime(&now
);
1802 sprintf(dd
->disk
.guid
, "%8s%04d%02d%02d",
1803 T10
, tm
->tm_year
+1900, tm
->tm_mon
+1, tm
->tm_mday
);
1804 *(__u32
*)(dd
->disk
.guid
+ 16) = random();
1805 *(__u32
*)(dd
->disk
.guid
+ 20) = random();
1807 dd
->disk
.refnum
= random(); /* and hope for the best FIXME check this is unique!!*/
1808 dd
->disk
.forced_ref
= 1;
1809 dd
->disk
.forced_guid
= 1;
1810 memset(dd
->disk
.vendor
, ' ', 32);
1811 memcpy(dd
->disk
.vendor
, "Linux", 5);
1812 memset(dd
->disk
.pad
, 0xff, 442);
1813 for (i
= 0; i
< ddf
->max_part
+1 ; i
++)
1814 dd
->vlist
[i
] = NULL
;
1816 n
= __be16_to_cpu(ddf
->phys
->used_pdes
);
1817 pde
= &ddf
->phys
->entries
[n
];
1819 ddf
->phys
->used_pdes
= __cpu_to_be16(n
);
1821 memcpy(pde
->guid
, dd
->disk
.guid
, DDF_GUID_LEN
);
1822 pde
->refnum
= dd
->disk
.refnum
;
1823 pde
->type
= __cpu_to_be16(DDF_Forced_PD_GUID
|DDF_Global_Spare
);
1824 pde
->state
= __cpu_to_be16(DDF_Online
);
1825 get_dev_size(fd
, NULL
, &size
);
1826 /* We are required to reserve 32Meg, and record the size in sectors */
1827 pde
->config_size
= __cpu_to_be64( (size
- 32*1024*1024) / 512);
1828 sprintf(pde
->path
, "%17.17s","Information: nil") ;
1829 memset(pde
->pad
, 0xff, 6);
1835 * This is the write_init_super method for a ddf container. It is
1836 * called when creating a container or adding another device to a
1841 static int write_init_super_ddf(struct supertype
*st
)
1844 struct ddf_super
*ddf
= st
->sb
;
1850 unsigned long long size
, sector
;
1852 for (d
= ddf
->dlist
; d
; d
=d
->next
) {
1858 /* We need to fill in the primary, (secondary) and workspace
1859 * lba's in the headers, set their checksums,
1860 * Also checksum phys, virt....
1862 * Then write everything out, finally the anchor is written.
1864 get_dev_size(fd
, NULL
, &size
);
1866 ddf
->anchor
.workspace_lba
= __cpu_to_be64(size
- 32*1024*2);
1867 ddf
->anchor
.primary_lba
= __cpu_to_be64(size
- 16*1024*2);
1868 ddf
->anchor
.seq
= __cpu_to_be32(1);
1869 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
1870 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
1872 ddf
->anchor
.openflag
= 0xFF; /* 'open' means nothing */
1873 ddf
->anchor
.seq
= 0xFFFFFFFF; /* no sequencing in anchor */
1874 ddf
->anchor
.crc
= calc_crc(&ddf
->anchor
, 512);
1876 ddf
->primary
.openflag
= 0;
1877 ddf
->primary
.type
= DDF_HEADER_PRIMARY
;
1879 ddf
->secondary
.openflag
= 0;
1880 ddf
->secondary
.type
= DDF_HEADER_SECONDARY
;
1882 ddf
->primary
.crc
= calc_crc(&ddf
->primary
, 512);
1883 ddf
->secondary
.crc
= calc_crc(&ddf
->secondary
, 512);
1885 sector
= size
- 16*1024*2;
1886 lseek64(fd
, sector
<<9, 0);
1887 write(fd
, &ddf
->primary
, 512);
1889 ddf
->controller
.crc
= calc_crc(&ddf
->controller
, 512);
1890 write(fd
, &ddf
->controller
, 512);
1892 ddf
->phys
->crc
= calc_crc(ddf
->phys
, ddf
->pdsize
);
1894 write(fd
, ddf
->phys
, ddf
->pdsize
);
1896 ddf
->virt
->crc
= calc_crc(ddf
->virt
, ddf
->vdsize
);
1897 write(fd
, ddf
->virt
, ddf
->vdsize
);
1899 /* Now write lots of config records. */
1900 n_config
= __be16_to_cpu(ddf
->active
->max_partitions
);
1901 conf_size
= __be16_to_cpu(ddf
->active
->config_record_len
) * 512;
1902 for (i
= 0 ; i
<= n_config
; i
++) {
1903 struct vcl
*c
= d
->vlist
[i
];
1906 c
->conf
.crc
= calc_crc(&c
->conf
, conf_size
);
1907 write(fd
, &c
->conf
, conf_size
);
1909 __u32 sig
= 0xffffffff;
1911 lseek64(fd
, conf_size
-4, SEEK_CUR
);
1914 d
->disk
.crc
= calc_crc(&d
->disk
, 512);
1915 write(fd
, &d
->disk
, 512);
1917 /* Maybe do the same for secondary */
1919 lseek64(fd
, (size
-1)*512, SEEK_SET
);
1920 write(fd
, &ddf
->anchor
, 512);
1927 static __u64
avail_size_ddf(struct supertype
*st
, __u64 devsize
)
1929 /* We must reserve the last 32Meg */
1930 if (devsize
<= 32*1024*2)
1932 return devsize
- 32*1024*2;
1936 int validate_geometry_ddf(struct supertype
*st
,
1937 int level
, int layout
, int raiddisks
,
1938 int chunk
, unsigned long long size
,
1939 char *dev
, unsigned long long *freesize
)
1945 /* ddf potentially supports lots of things, but it depends on
1946 * what devices are offered (and maybe kernel version?)
1947 * If given unused devices, we will make a container.
1948 * If given devices in a container, we will make a BVD.
1949 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
1952 if (level
== LEVEL_CONTAINER
) {
1953 st
->ss
= &super_ddf_container
;
1955 int rv
=st
->ss
->validate_geometry(st
, level
, layout
,
1962 return st
->ss
->validate_geometry(st
, level
, layout
, raiddisks
,
1963 chunk
, size
, dev
, freesize
);
1967 /* creating in a given container */
1968 st
->ss
= &super_ddf_bvd
;
1970 int rv
=st
->ss
->validate_geometry(st
, level
, layout
,
1977 return st
->ss
->validate_geometry(st
, level
, layout
, raiddisks
,
1978 chunk
, size
, dev
, freesize
);
1980 /* FIXME should exclude MULTIPATH, or more appropriately, allow
1981 * only known levels.
1986 /* This device needs to be either a device in a 'ddf' container,
1987 * or it needs to be a 'ddf-bvd' array.
1990 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1992 sra
= sysfs_read(fd
, 0, GET_VERSION
);
1994 if (sra
&& sra
->array
.major_version
== -1 &&
1995 strcmp(sra
->text_version
, "ddf-bvd") == 0) {
1996 st
->ss
= &super_ddf_svd
;
1997 return st
->ss
->validate_geometry(st
, level
, layout
,
1998 raiddisks
, chunk
, size
,
2003 Name
": Cannot create this array on device %s\n",
2007 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
2008 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2009 dev
, strerror(errno
));
2012 /* Well, it is in use by someone, maybe a 'ddf' container. */
2013 cfd
= open_container(fd
);
2016 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
2020 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
2022 if (sra
&& sra
->array
.major_version
== -1 &&
2023 strcmp(sra
->text_version
, "ddf") == 0) {
2024 /* This is a member of a ddf container. Load the container
2025 * and try to create a bvd
2027 struct ddf_super
*ddf
;
2028 st
->ss
= &super_ddf_bvd
;
2029 if (load_super_ddf_all(st
, cfd
, (void **)&ddf
, NULL
, 1) == 0) {
2031 st
->container_dev
= fd2devnum(cfd
);
2032 st
->container_member
= 27; // FIXME
2034 return st
->ss
->validate_geometry(st
, level
, layout
,
2035 raiddisks
, chunk
, size
,
2040 fprintf(stderr
, Name
": Cannot use %s: Already in use\n",
2045 int validate_geometry_ddf_container(struct supertype
*st
,
2046 int level
, int layout
, int raiddisks
,
2047 int chunk
, unsigned long long size
,
2048 char *dev
, unsigned long long *freesize
)
2051 unsigned long long ldsize
;
2053 if (level
!= LEVEL_CONTAINER
)
2058 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2060 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2061 dev
, strerror(errno
));
2064 if (!get_dev_size(fd
, dev
, &ldsize
)) {
2070 *freesize
= avail_size_ddf(st
, ldsize
);
2076 unsigned long long start
, size
;
2078 int cmp_extent(const void *av
, const void *bv
)
2080 const struct extent
*a
= av
;
2081 const struct extent
*b
= bv
;
2082 if (a
->start
< b
->start
)
2084 if (a
->start
> b
->start
)
2089 struct extent
*get_extents(struct ddf_super
*ddf
, struct dl
*dl
)
2091 /* find a list of used extents on the give physical device
2092 * (dnum) or the given ddf.
2093 * Return a malloced array of 'struct extent'
2095 FIXME ignore DDF_Legacy devices?
2103 for (dnum
= 0; dnum
< ddf
->phys
->used_pdes
; dnum
++)
2104 if (memcmp(dl
->disk
.guid
,
2105 ddf
->phys
->entries
[dnum
].guid
,
2109 if (dnum
== ddf
->phys
->used_pdes
)
2112 rv
= malloc(sizeof(struct extent
) * (ddf
->max_part
+ 2));
2116 for (i
= 0; i
< ddf
->max_part
+1; i
++) {
2117 struct vcl
*v
= dl
->vlist
[i
];
2120 for (j
=0; j
< v
->conf
.prim_elmnt_count
; j
++)
2121 if (v
->conf
.phys_refnum
[j
] == dl
->disk
.refnum
) {
2122 /* This device plays role 'j' in 'v'. */
2123 rv
[n
].start
= __be64_to_cpu(v
->lba_offset
[j
]);
2124 rv
[n
].size
= __be64_to_cpu(v
->conf
.blocks
);
2129 qsort(rv
, n
, sizeof(*rv
), cmp_extent
);
2131 rv
[n
].start
= __be64_to_cpu(ddf
->phys
->entries
[dnum
].config_size
);
2136 int validate_geometry_ddf_bvd(struct supertype
*st
,
2137 int level
, int layout
, int raiddisks
,
2138 int chunk
, unsigned long long size
,
2139 char *dev
, unsigned long long *freesize
)
2142 struct ddf_super
*ddf
= st
->sb
;
2144 unsigned long long pos
= 0;
2145 unsigned long long maxsize
;
2148 /* ddf/bvd supports lots of things, but not containers */
2149 if (level
== LEVEL_CONTAINER
)
2151 /* We must have the container info already read in. */
2156 /* General test: make sure there is space for
2157 * 'raiddisks' device extents of size 'size'.
2159 unsigned long long minsize
= size
;
2163 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2168 e
= get_extents(ddf
, dl
);
2171 unsigned long long esize
;
2172 esize
= e
[i
].start
- pos
;
2173 if (esize
>= minsize
)
2175 pos
= e
[i
].start
+ e
[i
].size
;
2177 } while (e
[i
-1].size
);
2182 if (dcnt
< raiddisks
) {
2183 fprintf(stderr
, Name
": Not enough devices with space "
2184 "for this array (%d < %d)\n",
2190 /* This device must be a member of the set */
2191 if (stat(dev
, &stb
) < 0)
2193 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
2195 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
2196 if (dl
->major
== major(stb
.st_rdev
) &&
2197 dl
->minor
== minor(stb
.st_rdev
))
2201 fprintf(stderr
, Name
": %s is not in the same DDF set\n",
2205 e
= get_extents(ddf
, dl
);
2209 unsigned long long esize
;
2210 esize
= e
[i
].start
- pos
;
2211 if (esize
>= maxsize
)
2213 pos
= e
[i
].start
+ e
[i
].size
;
2215 } while (e
[i
-1].size
);
2216 *freesize
= maxsize
;
2221 int validate_geometry_ddf_svd(struct supertype
*st
,
2222 int level
, int layout
, int raiddisks
,
2223 int chunk
, unsigned long long size
,
2224 char *dev
, unsigned long long *freesize
)
2226 /* dd/svd only supports striped, mirrored, concat, spanned... */
2227 if (level
!= LEVEL_LINEAR
&&
2235 static int load_super_ddf_all(struct supertype
*st
, int fd
,
2236 void **sbp
, char *devname
, int keep_fd
)
2239 struct ddf_super
*super
;
2240 struct mdinfo
*sd
, *best
= NULL
;
2246 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
2249 if (sra
->array
.major_version
!= -1 ||
2250 sra
->array
.minor_version
!= -2 ||
2251 strcmp(sra
->text_version
, "ddf") != 0)
2254 super
= malloc(sizeof(*super
));
2257 memset(super
, 0, sizeof(*super
));
2259 /* first, try each device, and choose the best ddf */
2260 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
2262 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2263 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
2266 rv
= load_ddf_headers(dfd
, super
, NULL
);
2267 if (!keep_fd
) close(dfd
);
2269 seq
= __be32_to_cpu(super
->active
->seq
);
2270 if (super
->active
->openflag
)
2272 if (!best
|| seq
> bestseq
) {
2280 /* OK, load this ddf */
2281 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
2282 dfd
= dev_open(nm
, O_RDONLY
);
2285 load_ddf_headers(dfd
, super
, NULL
);
2286 load_ddf_global(dfd
, super
, NULL
);
2288 /* Now we need the device-local bits */
2289 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
2290 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2291 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
2294 seq
= load_ddf_local(dfd
, super
, NULL
, keep_fd
);
2295 if (!keep_fd
) close(dfd
);
2298 if (st
->ss
== NULL
) {
2299 st
->ss
= &super_ddf_container
;
2300 st
->minor_version
= 0;
2309 static struct mdinfo
*container_content_ddf(struct supertype
*st
)
2311 /* Given a container loaded by load_super_ddf_all,
2312 * extract information about all the arrays into
2315 * For each vcl in conflist: create an mdinfo, fill it in,
2316 * then look for matching devices (phys_refnum) in dlist
2317 * and create appropriate device mdinfo.
2319 struct ddf_super
*ddf
= st
->sb
;
2320 struct mdinfo
*rest
= NULL
;
2323 for (vc
= ddf
->conflist
; vc
; vc
=vc
->next
)
2327 struct mdinfo
*this;
2328 this = malloc(sizeof(*this));
2329 memset(this, 0, sizeof(*this));
2333 this->array
.major_version
= 1000;
2334 this->array
.minor_version
= 0;
2335 this->array
.patch_version
= 0;
2336 this->array
.level
= map_num1(ddf_level_num
, vc
->conf
.prl
);
2337 this->array
.raid_disks
=
2338 __be16_to_cpu(vc
->conf
.prim_elmnt_count
);
2339 /* FIXME this should be mapped */
2340 this->array
.layout
= vc
->conf
.rlq
;
2341 this->array
.md_minor
= -1;
2342 this->array
.ctime
= DECADE
+
2343 __be32_to_cpu(*(__u32
*)(vc
->conf
.guid
+16));
2344 this->array
.utime
= DECADE
+
2345 __be32_to_cpu(vc
->conf
.timestamp
);
2346 this->array
.chunk_size
= 512 << vc
->conf
.chunk_shift
;
2348 for (i
=0; i
< __be16_to_cpu(ddf
->virt
->populated_vdes
); i
++)
2349 if (memcmp(ddf
->virt
->entries
[i
].guid
,
2350 vc
->conf
.guid
, DDF_GUID_LEN
) == 0)
2352 if (ddf
->virt
->entries
[i
].state
& DDF_state_inconsistent
)
2353 this->array
.state
= 0;
2355 this->array
.state
= 1;
2356 memcpy(this->name
, ddf
->virt
->entries
[i
].name
, 32);
2359 memset(this->uuid
, 0, sizeof(this->uuid
));
2360 this->component_size
= __be64_to_cpu(vc
->conf
.blocks
);
2361 this->array
.size
= this->component_size
/ 2;
2362 this->container_member
= i
;
2364 mppe
= __be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
2365 for (i
=0 ; i
< mppe
; i
++) {
2369 if (vc
->conf
.phys_refnum
[i
] == 0xFFFFFFFF)
2372 this->array
.working_disks
++;
2374 for (d
= ddf
->dlist
; d
; d
=d
->next
)
2375 if (d
->disk
.refnum
== vc
->conf
.phys_refnum
[i
])
2380 dev
= malloc(sizeof(*dev
));
2381 memset(dev
, 0, sizeof(*dev
));
2382 dev
->next
= this->devs
;
2385 dev
->disk
.number
= __be32_to_cpu(d
->disk
.refnum
);
2386 dev
->disk
.major
= d
->major
;
2387 dev
->disk
.minor
= d
->minor
;
2388 dev
->disk
.raid_disk
= i
;
2389 dev
->disk
.state
= (1<<MD_DISK_SYNC
)|(1<<MD_DISK_ACTIVE
);
2391 dev
->events
= __le32_to_cpu(ddf
->primary
.seq
);
2392 dev
->data_offset
= vc
->lba_offset
[i
];
2393 dev
->component_size
= __be64_to_cpu(vc
->conf
.blocks
);
2395 strcpy(dev
->name
, d
->devname
);
2401 static int init_zero_ddf(struct supertype
*st
,
2402 mdu_array_info_t
*info
,
2403 unsigned long long size
, char *name
,
2404 char *homehost
, int *uuid
)
2410 static int store_zero_ddf(struct supertype
*st
, int fd
)
2412 unsigned long long dsize
;
2414 memset(buf
, 0, 512);
2417 if (!get_dev_size(fd
, NULL
, &dsize
))
2420 lseek64(fd
, dsize
-512, 0);
2421 write(fd
, buf
, 512);
2425 static int compare_super_ddf(struct supertype
*st
, struct supertype
*tst
)
2429 * 0 same, or first was empty, and second was copied
2430 * 1 second had wrong number
2432 * 3 wrong other info
2434 struct ddf_super
*first
= st
->sb
;
2435 struct ddf_super
*second
= tst
->sb
;
2443 if (memcmp(first
->anchor
.guid
, second
->anchor
.guid
, DDF_GUID_LEN
) != 0)
2446 /* FIXME should I look at anything else? */
2450 static int ddf_open_new(struct supertype
*c
, struct active_array
*a
, int inst
)
2452 fprintf(stderr
, "ddf: open_new %d\n", inst
);
2456 static void ddf_mark_clean(struct active_array
*a
, unsigned long long sync_pos
)
2458 fprintf(stderr
, "ddf: mark clean %llu\n", sync_pos
);
2461 static void ddf_mark_dirty(struct active_array
*a
)
2463 fprintf(stderr
, "ddf: mark dirty\n");
2466 static void ddf_mark_sync(struct active_array
*a
, unsigned long long resync
)
2468 fprintf(stderr
, "ddf: mark sync\n");
2471 static void ddf_set_disk(struct active_array
*a
, int n
, int state
)
2473 fprintf(stderr
, "ddf: set_disk %d\n", n
);
2476 static void ddf_sync_metadata(struct active_array
*a
)
2478 fprintf(stderr
, "ddf: sync_metadata\n");
2481 struct superswitch super_ddf
= {
2483 .examine_super
= examine_super_ddf
,
2484 .brief_examine_super
= brief_examine_super_ddf
,
2485 .detail_super
= detail_super_ddf
,
2486 .brief_detail_super
= brief_detail_super_ddf
,
2487 .validate_geometry
= validate_geometry_ddf
,
2489 .match_home
= match_home_ddf
,
2490 .uuid_from_super
= uuid_from_super_ddf
,
2491 .getinfo_super
= getinfo_super_ddf
,
2492 .update_super
= update_super_ddf
,
2494 .avail_size
= avail_size_ddf
,
2496 .compare_super
= compare_super_ddf
,
2498 .load_super
= load_super_ddf
,
2499 .init_super
= init_zero_ddf
,
2500 .store_super
= store_zero_ddf
,
2501 .free_super
= free_super_ddf
,
2502 .match_metadata_desc
= match_metadata_desc_ddf
,
2503 .getinfo_super_n
= getinfo_super_n_container
,
2509 .text_version
= "ddf",
2512 .open_new
= ddf_open_new
,
2513 .load_super
= load_super_ddf
,
2514 .mark_clean
= ddf_mark_clean
,
2515 .mark_dirty
= ddf_mark_dirty
,
2516 .mark_sync
= ddf_mark_sync
,
2517 .set_disk
= ddf_set_disk
,
2518 .sync_metadata
= ddf_sync_metadata
,
2523 /* Super_ddf_container is set by validate_geometry_ddf when given a
2524 * device that is not part of any array
2526 struct superswitch super_ddf_container
= {
2528 .validate_geometry
= validate_geometry_ddf_container
,
2529 .write_init_super
= write_init_super_ddf
,
2532 .init_super
= init_super_ddf
,
2533 .add_to_super
= add_to_super_ddf
,
2535 .free_super
= free_super_ddf
,
2537 .container_content
= container_content_ddf
,
2542 .text_version
= "ddf",
2545 struct superswitch super_ddf_bvd
= {
2547 // .detail_super = detail_super_ddf_bvd,
2548 // .brief_detail_super = brief_detail_super_ddf_bvd,
2549 .validate_geometry
= validate_geometry_ddf_bvd
,
2550 .write_init_super
= write_init_super_ddf
,
2552 .update_super
= update_super_ddf
,
2553 .init_super
= init_super_ddf_bvd
,
2554 .add_to_super
= add_to_super_ddf_bvd
,
2555 .getinfo_super
= getinfo_super_ddf_bvd
,
2556 .getinfo_super_n
= getinfo_super_n_bvd
,
2558 .load_super
= load_super_ddf
,
2559 .free_super
= free_super_ddf
,
2560 .match_metadata_desc
= match_metadata_desc_ddf_bvd
,
2566 .text_version
= "ddf",
2569 struct superswitch super_ddf_svd
= {
2571 // .detail_super = detail_super_ddf_svd,
2572 // .brief_detail_super = brief_detail_super_ddf_svd,
2573 .validate_geometry
= validate_geometry_ddf_svd
,
2575 .update_super
= update_super_ddf
,
2576 .init_super
= init_super_ddf
,
2578 .load_super
= load_super_ddf
,
2579 .free_super
= free_super_ddf
,
2580 .match_metadata_desc
= match_metadata_desc_ddf_svd
,
2585 .text_version
= "ddf",