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;
602 super
->conflist
= NULL
;
607 static int load_ddf_local(int fd
, struct ddf_super
*super
,
608 char *devname
, int keep
)
617 /* First the local disk info */
618 super
->max_part
= __be16_to_cpu(super
->active
->max_partitions
);
619 dl
= malloc(sizeof(*dl
) +
620 (super
->max_part
+1) * sizeof(dl
->vlist
[0]));
622 load_section(fd
, super
, &dl
->disk
,
623 super
->active
->data_section_offset
,
624 super
->active
->data_section_length
,
626 dl
->devname
= devname
? strdup(devname
) : NULL
;
629 dl
->major
= major(stb
.st_rdev
);
630 dl
->minor
= minor(stb
.st_rdev
);
631 dl
->next
= super
->dlist
;
632 dl
->fd
= keep
? fd
: -1;
633 for (i
=0 ; i
< super
->max_part
+ 1 ; i
++)
637 /* Now the config list. */
638 /* 'conf' is an array of config entries, some of which are
639 * probably invalid. Those which are good need to be copied into
642 conflen
= __be16_to_cpu(super
->active
->config_record_len
);
644 conf
= load_section(fd
, super
, NULL
,
645 super
->active
->config_section_offset
,
646 super
->active
->config_section_length
,
650 i
< __be32_to_cpu(super
->active
->config_section_length
);
652 struct vd_config
*vd
=
653 (struct vd_config
*)((char*)conf
+ i
*512);
656 if (vd
->magic
!= DDF_VD_CONF_MAGIC
)
658 for (vcl
= super
->conflist
; vcl
; vcl
= vcl
->next
) {
659 if (memcmp(vcl
->conf
.guid
,
660 vd
->guid
, DDF_GUID_LEN
) == 0)
665 dl
->vlist
[i
/conflen
] = vcl
;
666 if (__be32_to_cpu(vd
->seqnum
) <=
667 __be32_to_cpu(vcl
->conf
.seqnum
))
670 vcl
= malloc(conflen
*512 + offsetof(struct vcl
, conf
));
671 vcl
->next
= super
->conflist
;
672 super
->conflist
= vcl
;
674 memcpy(&vcl
->conf
, vd
, conflen
*512);
675 mppe
= __be16_to_cpu(super
->anchor
.max_primary_element_entries
);
676 vcl
->lba_offset
= (__u64
*)
677 &vcl
->conf
.phys_refnum
[mppe
];
678 dl
->vlist
[i
/conflen
] = vcl
;
686 static int load_super_ddf_all(struct supertype
*st
, int fd
,
687 void **sbp
, char *devname
, int keep_fd
);
689 static int load_super_ddf(struct supertype
*st
, int fd
,
692 unsigned long long dsize
;
693 struct ddf_super
*super
;
697 if (load_super_ddf_all(st
, fd
, &st
->sb
, devname
, 0) == 0)
701 if (get_dev_size(fd
, devname
, &dsize
) == 0)
704 /* 32M is a lower bound */
705 if (dsize
<= 32*1024*1024) {
708 Name
": %s is too small for ddf: "
709 "size is %llu sectors.\n",
717 Name
": %s is an odd size for ddf: "
718 "size is %llu bytes.\n",
724 super
= malloc(sizeof(*super
));
726 fprintf(stderr
, Name
": malloc of %zu failed.\n",
731 rv
= load_ddf_headers(fd
, super
, devname
);
737 /* Have valid headers and have chosen the best. Let's read in the rest*/
739 rv
= load_ddf_global(fd
, super
, devname
);
744 Name
": Failed to load all information "
745 "sections on %s\n", devname
);
750 load_ddf_local(fd
, super
, devname
, 0);
752 /* Should possibly check the sections .... */
755 if (st
->ss
== NULL
) {
757 st
->minor_version
= 0;
764 static void free_super_ddf(struct supertype
*st
)
766 struct ddf_super
*ddf
= st
->sb
;
771 while (ddf
->conflist
) {
772 struct vcl
*v
= ddf
->conflist
;
773 ddf
->conflist
= v
->next
;
777 struct dl
*d
= ddf
->dlist
;
778 ddf
->dlist
= d
->next
;
787 static struct supertype
*match_metadata_desc_ddf(char *arg
)
789 /* 'ddf' only support containers */
790 struct supertype
*st
;
791 if (strcmp(arg
, "ddf") != 0 &&
792 strcmp(arg
, "default") != 0
796 st
= malloc(sizeof(*st
));
799 st
->minor_version
= 0;
804 static struct supertype
*match_metadata_desc_ddf_bvd(char *arg
)
806 struct supertype
*st
;
807 if (strcmp(arg
, "ddf/bvd") != 0 &&
808 strcmp(arg
, "bvd") != 0 &&
809 strcmp(arg
, "default") != 0
813 st
= malloc(sizeof(*st
));
814 st
->ss
= &super_ddf_bvd
;
816 st
->minor_version
= 0;
820 static struct supertype
*match_metadata_desc_ddf_svd(char *arg
)
822 struct supertype
*st
;
823 if (strcmp(arg
, "ddf/svd") != 0 &&
824 strcmp(arg
, "svd") != 0 &&
825 strcmp(arg
, "default") != 0
829 st
= malloc(sizeof(*st
));
830 st
->ss
= &super_ddf_svd
;
832 st
->minor_version
= 0;
839 static mapping_t ddf_state
[] = {
845 { "Partially Optimal", 5},
851 static mapping_t ddf_init_state
[] = {
852 { "Not Initialised", 0},
853 { "QuickInit in Progress", 1},
854 { "Fully Initialised", 2},
858 static mapping_t ddf_access
[] = {
862 { "Blocked (no access)", 3},
866 static mapping_t ddf_level
[] = {
867 { "RAID0", DDF_RAID0
},
868 { "RAID1", DDF_RAID1
},
869 { "RAID3", DDF_RAID3
},
870 { "RAID4", DDF_RAID4
},
871 { "RAID5", DDF_RAID5
},
872 { "RAID1E",DDF_RAID1E
},
874 { "CONCAT",DDF_CONCAT
},
875 { "RAID5E",DDF_RAID5E
},
876 { "RAID5EE",DDF_RAID5EE
},
877 { "RAID6", DDF_RAID6
},
880 static mapping_t ddf_sec_level
[] = {
881 { "Striped", DDF_2STRIPED
},
882 { "Mirrored", DDF_2MIRRORED
},
883 { "Concat", DDF_2CONCAT
},
884 { "Spanned", DDF_2SPANNED
},
892 static struct num_mapping ddf_level_num
[] = {
895 { DDF_RAID3
, LEVEL_UNSUPPORTED
},
897 { DDF_RAID1E
, LEVEL_UNSUPPORTED
},
898 { DDF_JBOD
, LEVEL_UNSUPPORTED
},
899 { DDF_CONCAT
, LEVEL_LINEAR
},
900 { DDF_RAID5E
, LEVEL_UNSUPPORTED
},
901 { DDF_RAID5EE
, LEVEL_UNSUPPORTED
},
906 static int map_num1(struct num_mapping
*map
, int num
)
909 for (i
=0 ; map
[i
].num1
!= MAXINT
; i
++)
910 if (map
[i
].num1
== num
)
916 static void print_guid(char *guid
, int tstamp
)
918 /* A GUIDs are part (or all) ASCII and part binary.
919 * They tend to be space padded.
920 * We ignore trailing spaces and print numbers
921 * <0x20 and >=0x7f as \xXX
922 * Some GUIDs have a time stamp in bytes 16-19.
923 * We print that if appropriate
925 int l
= DDF_GUID_LEN
;
927 while (l
&& guid
[l
-1] == ' ')
929 for (i
=0 ; i
<l
; i
++) {
930 if (guid
[i
] >= 0x20 && guid
[i
] < 0x7f)
931 fputc(guid
[i
], stdout
);
933 fprintf(stdout
, "\\x%02x", guid
[i
]&255);
936 time_t then
= __be32_to_cpu(*(__u32
*)(guid
+16)) + DECADE
;
939 tm
= localtime(&then
);
940 strftime(tbuf
, 100, " (%D %T)",tm
);
945 static void examine_vd(int n
, struct ddf_super
*sb
, char *guid
)
947 int crl
= __be16_to_cpu(sb
->anchor
.config_record_len
);
950 for (vcl
= sb
->conflist
; vcl
; vcl
= vcl
->next
) {
951 struct vd_config
*vc
= &vcl
->conf
;
953 if (calc_crc(vc
, crl
*512) != vc
->crc
)
955 if (memcmp(vc
->guid
, guid
, DDF_GUID_LEN
) != 0)
958 /* Ok, we know about this VD, let's give more details */
959 printf(" Raid Devices[%d] : %d\n", n
,
960 __be16_to_cpu(vc
->prim_elmnt_count
));
961 printf(" Chunk Size[%d] : %d sectors\n", n
,
962 1 << vc
->chunk_shift
);
963 printf(" Raid Level[%d] : %s\n", n
,
964 map_num(ddf_level
, vc
->prl
)?:"-unknown-");
965 if (vc
->sec_elmnt_count
!= 1) {
966 printf(" Secondary Position[%d] : %d of %d\n", n
,
967 vc
->sec_elmnt_seq
, vc
->sec_elmnt_count
);
968 printf(" Secondary Level[%d] : %s\n", n
,
969 map_num(ddf_sec_level
, vc
->srl
) ?: "-unknown-");
971 printf(" Device Size[%d] : %llu\n", n
,
972 __be64_to_cpu(vc
->blocks
)/2);
973 printf(" Array Size[%d] : %llu\n", n
,
974 __be64_to_cpu(vc
->array_blocks
)/2);
978 static void examine_vds(struct ddf_super
*sb
)
980 int cnt
= __be16_to_cpu(sb
->virt
->populated_vdes
);
982 printf(" Virtual Disks : %d\n", cnt
);
984 for (i
=0; i
<cnt
; i
++) {
985 struct virtual_entry
*ve
= &sb
->virt
->entries
[i
];
986 printf(" VD GUID[%d] : ", i
); print_guid(ve
->guid
, 1);
988 printf(" unit[%d] : %d\n", i
, __be16_to_cpu(ve
->unit
));
989 printf(" state[%d] : %s, %s%s\n", i
,
990 map_num(ddf_state
, ve
->state
& 7),
991 (ve
->state
& 8) ? "Morphing, ": "",
992 (ve
->state
& 16)? "Not Consistent" : "Consistent");
993 printf(" init state[%d] : %s\n", i
,
994 map_num(ddf_init_state
, ve
->init_state
&3));
995 printf(" access[%d] : %s\n", i
,
996 map_num(ddf_access
, (ve
->init_state
>>6) & 3));
997 printf(" Name[%d] : %.16s\n", i
, ve
->name
);
998 examine_vd(i
, sb
, ve
->guid
);
1000 if (cnt
) printf("\n");
1003 static void examine_pds(struct ddf_super
*sb
)
1005 int cnt
= __be16_to_cpu(sb
->phys
->used_pdes
);
1008 printf(" Physical Disks : %d\n", cnt
);
1010 for (i
=0 ; i
<cnt
; i
++) {
1011 struct phys_disk_entry
*pd
= &sb
->phys
->entries
[i
];
1012 int type
= __be16_to_cpu(pd
->type
);
1013 int state
= __be16_to_cpu(pd
->state
);
1015 printf(" PD GUID[%d] : ", i
); print_guid(pd
->guid
, 0);
1017 printf(" ref[%d] : %08x\n", i
,
1018 __be32_to_cpu(pd
->refnum
));
1019 printf(" mode[%d] : %s%s%s%s%s\n", i
,
1020 (type
&2) ? "active":"",
1021 (type
&4) ? "Global Spare":"",
1022 (type
&8) ? "spare" : "",
1023 (type
&16)? ", foreign" : "",
1024 (type
&32)? "pass-through" : "");
1025 printf(" state[%d] : %s%s%s%s%s%s%s\n", i
,
1026 (state
&1)? "Online": "Offline",
1027 (state
&2)? ", Failed": "",
1028 (state
&4)? ", Rebuilding": "",
1029 (state
&8)? ", in-transition": "",
1030 (state
&16)? ", SMART errors": "",
1031 (state
&32)? ", Unrecovered Read Errors": "",
1032 (state
&64)? ", Missing" : "");
1033 printf(" Avail Size[%d] : %llu K\n", i
,
1034 __be64_to_cpu(pd
->config_size
)>>1);
1035 for (dl
= sb
->dlist
; dl
; dl
= dl
->next
) {
1036 if (dl
->disk
.refnum
== pd
->refnum
) {
1037 char *dv
= map_dev(dl
->major
, dl
->minor
, 0);
1039 printf(" Device[%d] : %s\n",
1047 static void examine_super_ddf(struct supertype
*st
, char *homehost
)
1049 struct ddf_super
*sb
= st
->sb
;
1051 printf(" Magic : %08x\n", __be32_to_cpu(sb
->anchor
.magic
));
1052 printf(" Version : %.8s\n", sb
->anchor
.revision
);
1053 printf("Controller GUID : "); print_guid(sb
->controller
.guid
, 0);
1055 printf(" Container GUID : "); print_guid(sb
->anchor
.guid
, 1);
1057 printf(" Seq : %08x\n", __be32_to_cpu(sb
->active
->seq
));
1058 printf(" Redundant hdr : %s\n", sb
->secondary
.magic
== DDF_HEADER_MAGIC
1064 static void brief_examine_super_ddf(struct supertype
*st
)
1066 /* We just write a generic DDF ARRAY entry
1067 * The uuid is all hex, 6 groups of 4 bytes
1069 struct ddf_super
*ddf
= st
->sb
;
1071 printf("ARRAY /dev/ddf UUID=");
1072 for (i
= 0; i
< DDF_GUID_LEN
; i
++) {
1073 printf("%02x", ddf
->anchor
.guid
[i
]);
1074 if ((i
&3) == 0 && i
!= 0)
1080 static void detail_super_ddf(struct supertype
*st
, char *homehost
)
1083 * Could print DDF GUID
1084 * Need to find which array
1085 * If whole, briefly list all arrays
1090 static void brief_detail_super_ddf(struct supertype
*st
)
1092 /* FIXME I really need to know which array we are detailing.
1093 * Can that be stored in ddf_super??
1095 // struct ddf_super *ddf = st->sb;
1101 static int match_home_ddf(struct supertype
*st
, char *homehost
)
1103 /* It matches 'this' host if the controller is a
1104 * Linux-MD controller with vendor_data matching
1107 struct ddf_super
*ddf
= st
->sb
;
1108 int len
= strlen(homehost
);
1110 return (memcmp(ddf
->controller
.guid
, T10
, 8) == 0 &&
1111 len
< sizeof(ddf
->controller
.vendor_data
) &&
1112 memcmp(ddf
->controller
.vendor_data
, homehost
,len
) == 0 &&
1113 ddf
->controller
.vendor_data
[len
] == 0);
1116 static struct vd_config
*find_vdcr(struct ddf_super
*ddf
)
1118 /* FIXME this just picks off the first one */
1119 return &ddf
->conflist
->conf
;
1122 static void uuid_from_super_ddf(struct supertype
*st
, int uuid
[4])
1124 /* The uuid returned here is used for:
1125 * uuid to put into bitmap file (Create, Grow)
1126 * uuid for backup header when saving critical section (Grow)
1127 * comparing uuids when re-adding a device into an array
1128 * For each of these we can make do with a truncated
1129 * or hashed uuid rather than the original, as long as
1131 * In each case the uuid required is that of the data-array,
1132 * not the device-set.
1133 * In the case of SVD we assume the BVD is of interest,
1134 * though that might be the case if a bitmap were made for
1135 * a mirrored SVD - worry about that later.
1136 * So we need to find the VD configuration record for the
1137 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1138 * The first 16 bytes of the sha1 of these is used.
1140 struct ddf_super
*ddf
= st
->sb
;
1141 struct vd_config
*vd
= find_vdcr(ddf
);
1144 memset(uuid
, 0, sizeof (uuid
));
1147 struct sha1_ctx ctx
;
1148 sha1_init_ctx(&ctx
);
1149 sha1_process_bytes(&vd
->guid
, DDF_GUID_LEN
, &ctx
);
1150 if (vd
->sec_elmnt_count
> 1)
1151 sha1_process_bytes(&vd
->sec_elmnt_seq
, 1, &ctx
);
1152 sha1_finish_ctx(&ctx
, buf
);
1153 memcpy(uuid
, buf
, sizeof(uuid
));
1157 static void getinfo_super_ddf(struct supertype
*st
, struct mdinfo
*info
)
1159 struct ddf_super
*ddf
= st
->sb
;
1162 info
->array
.major_version
= 1000;
1163 info
->array
.minor_version
= 0; /* FIXME use ddf->revision somehow */
1164 info
->array
.patch_version
= 0;
1165 info
->array
.raid_disks
= __be16_to_cpu(ddf
->phys
->used_pdes
);
1166 info
->array
.level
= LEVEL_CONTAINER
;
1167 info
->array
.layout
= 0;
1168 info
->array
.md_minor
= -1;
1169 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*(__u32
*)
1170 (ddf
->anchor
.guid
+16));
1171 info
->array
.utime
= 0;
1172 info
->array
.chunk_size
= 0;
1174 // info->data_offset = ???;
1175 // info->component_size = ???;
1177 info
->disk
.major
= 0;
1178 info
->disk
.minor
= 0;
1179 info
->disk
.number
= __be32_to_cpu(ddf
->dlist
->disk
.refnum
);
1180 // info->disk.raid_disk = find refnum in the table and use index;
1181 info
->disk
.raid_disk
= -1;
1182 for (i
= 0; i
< __be16_to_cpu(ddf
->phys
->max_pdes
) ; i
++)
1183 if (ddf
->phys
->entries
[i
].refnum
== ddf
->dlist
->disk
.refnum
) {
1184 info
->disk
.raid_disk
= i
;
1187 info
->disk
.state
= (1 << MD_DISK_SYNC
);
1189 info
->reshape_active
= 0;
1191 // uuid_from_super_ddf(info->uuid, sbv);
1193 // info->name[] ?? ;
1196 static void getinfo_super_n_container(struct supertype
*st
, struct mdinfo
*info
)
1198 /* just need offset and size */
1199 struct ddf_super
*ddf
= st
->sb
;
1200 int n
= info
->disk
.number
;
1202 info
->data_offset
= __be64_to_cpu(ddf
->phys
->entries
[n
].config_size
);
1203 info
->component_size
= 32*1024*1024 / 512;
1206 static int rlq_to_layout(int rlq
, int prl
, int raiddisks
);
1208 static void getinfo_super_ddf_bvd(struct supertype
*st
, struct mdinfo
*info
)
1210 struct ddf_super
*ddf
= st
->sb
;
1211 struct vd_config
*vd
= find_vdcr(ddf
);
1213 /* FIXME this returns BVD info - what if we want SVD ?? */
1215 info
->array
.major_version
= 1000;
1216 info
->array
.minor_version
= 0; /* FIXME use ddf->revision somehow */
1217 info
->array
.patch_version
= 0;
1218 info
->array
.raid_disks
= __be16_to_cpu(vd
->prim_elmnt_count
);
1219 info
->array
.level
= map_num1(ddf_level_num
, vd
->prl
);
1220 info
->array
.layout
= rlq_to_layout(vd
->rlq
, vd
->prl
,
1221 info
->array
.raid_disks
);
1222 info
->array
.md_minor
= -1;
1223 info
->array
.ctime
= DECADE
+ __be32_to_cpu(*(__u32
*)(vd
->guid
+16));
1224 info
->array
.utime
= DECADE
+ __be32_to_cpu(vd
->timestamp
);
1225 info
->array
.chunk_size
= 512 << vd
->chunk_shift
;
1227 // info->data_offset = ???;
1228 // info->component_size = ???;
1230 info
->disk
.major
= 0;
1231 info
->disk
.minor
= 0;
1232 // info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1233 // info->disk.raid_disk = find refnum in the table and use index;
1234 // info->disk.state = ???;
1236 uuid_from_super_ddf(st
, info
->uuid
);
1238 // info->name[] ?? ;
1241 static void getinfo_super_n_bvd(struct supertype
*st
, struct mdinfo
*info
)
1243 /* Find the particular details for info->disk.raid_disk.
1244 * This includes data_offset, component_size,
1246 struct ddf_super
*ddf
= st
->sb
;
1247 __u64
*lba_offset
= ddf
->newconf
->lba_offset
;
1248 struct vd_config
*conf
= &ddf
->newconf
->conf
;
1249 info
->data_offset
= __be64_to_cpu(lba_offset
[info
->disk
.raid_disk
]);
1250 info
->component_size
= __be64_to_cpu(conf
->blocks
);
1253 static int update_super_ddf(struct supertype
*st
, struct mdinfo
*info
,
1255 char *devname
, int verbose
,
1256 int uuid_set
, char *homehost
)
1258 /* For 'assemble' and 'force' we need to return non-zero if any
1259 * change was made. For others, the return value is ignored.
1260 * Update options are:
1261 * force-one : This device looks a bit old but needs to be included,
1262 * update age info appropriately.
1263 * assemble: clear any 'faulty' flag to allow this device to
1265 * force-array: Array is degraded but being forced, mark it clean
1266 * if that will be needed to assemble it.
1268 * newdev: not used ????
1269 * grow: Array has gained a new device - this is currently for
1271 * resync: mark as dirty so a resync will happen.
1272 * uuid: Change the uuid of the array to match watch is given
1273 * homehost: update the recorded homehost
1274 * name: update the name - preserving the homehost
1275 * _reshape_progress: record new reshape_progress position.
1277 * Following are not relevant for this version:
1278 * sparc2.2 : update from old dodgey metadata
1279 * super-minor: change the preferred_minor number
1280 * summaries: update redundant counters.
1283 // struct ddf_super *ddf = st->sb;
1284 // struct vd_config *vd = find_vdcr(ddf);
1285 // struct virtual_entry *ve = find_ve(ddf);
1288 /* we don't need to handle "force-*" or "assemble" as
1289 * there is no need to 'trick' the kernel. We the metadata is
1290 * first updated to activate the array, all the implied modifications
1294 if (strcmp(update
, "grow") == 0) {
1297 if (strcmp(update
, "resync") == 0) {
1298 // info->resync_checkpoint = 0;
1300 /* We ignore UUID updates as they make even less sense
1303 if (strcmp(update
, "homehost") == 0) {
1304 /* homehost is stored in controller->vendor_data,
1305 * or it is when we are the vendor
1307 // if (info->vendor_is_local)
1308 // strcpy(ddf->controller.vendor_data, homehost);
1310 if (strcmp(update
, "name") == 0) {
1311 /* name is stored in virtual_entry->name */
1312 // memset(ve->name, ' ', 16);
1313 // strncpy(ve->name, info->name, 16);
1315 if (strcmp(update
, "_reshape_progress") == 0) {
1316 /* We don't support reshape yet */
1319 // update_all_csum(ddf);
1324 static void make_header_guid(char *guid
)
1328 /* Create a DDF Header of Virtual Disk GUID */
1330 /* 24 bytes of fiction required.
1331 * first 8 are a 'vendor-id' - "Linux-MD"
1332 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1333 * Remaining 8 random number plus timestamp
1335 memcpy(guid
, T10
, sizeof(T10
));
1336 stamp
= __cpu_to_be32(0xdeadbeef);
1337 memcpy(guid
+8, &stamp
, 4);
1338 stamp
= __cpu_to_be32(0);
1339 memcpy(guid
+12, &stamp
, 4);
1340 stamp
= __cpu_to_be32(time(0) - DECADE
);
1341 memcpy(guid
+16, &stamp
, 4);
1342 rfd
= open("/dev/urandom", O_RDONLY
);
1343 if (rfd
< 0 || read(rfd
, &stamp
, 4) != 4)
1345 memcpy(guid
+20, &stamp
, 4);
1346 if (rfd
>= 0) close(rfd
);
1348 static int init_super_ddf(struct supertype
*st
,
1349 mdu_array_info_t
*info
,
1350 unsigned long long size
, char *name
, char *homehost
,
1353 /* This is primarily called by Create when creating a new array.
1354 * We will then get add_to_super called for each component, and then
1355 * write_init_super called to write it out to each device.
1356 * For DDF, Create can create on fresh devices or on a pre-existing
1358 * To create on a pre-existing array a different method will be called.
1359 * This one is just for fresh drives.
1361 * We need to create the entire 'ddf' structure which includes:
1362 * DDF headers - these are easy.
1363 * Controller data - a Sector describing this controller .. not that
1364 * this is a controller exactly.
1365 * Physical Disk Record - one entry per device, so
1366 * leave plenty of space.
1367 * Virtual Disk Records - again, just leave plenty of space.
1368 * This just lists VDs, doesn't give details
1369 * Config records - describes the VDs that use this disk
1370 * DiskData - describes 'this' device.
1371 * BadBlockManagement - empty
1372 * Diag Space - empty
1373 * Vendor Logs - Could we put bitmaps here?
1376 struct ddf_super
*ddf
;
1379 int max_phys_disks
, max_virt_disks
;
1380 unsigned long long sector
;
1384 struct phys_disk
*pd
;
1385 struct virtual_disk
*vd
;
1387 ddf
= malloc(sizeof(*ddf
));
1388 ddf
->dlist
= NULL
; /* no physical disks yet */
1389 ddf
->conflist
= NULL
; /* No virtual disks yet */
1391 /* At least 32MB *must* be reserved for the ddf. So let's just
1392 * start 32MB from the end, and put the primary header there.
1393 * Don't do secondary for now.
1394 * We don't know exactly where that will be yet as it could be
1395 * different on each device. To just set up the lengths.
1399 ddf
->anchor
.magic
= DDF_HEADER_MAGIC
;
1400 make_header_guid(ddf
->anchor
.guid
);
1402 memcpy(ddf
->anchor
.revision
, DDF_REVISION
, 8);
1403 ddf
->anchor
.seq
= __cpu_to_be32(1);
1404 ddf
->anchor
.timestamp
= __cpu_to_be32(time(0) - DECADE
);
1405 ddf
->anchor
.openflag
= 0xFF;
1406 ddf
->anchor
.foreignflag
= 0;
1407 ddf
->anchor
.enforcegroups
= 0; /* Is this best?? */
1408 ddf
->anchor
.pad0
= 0xff;
1409 memset(ddf
->anchor
.pad1
, 0xff, 12);
1410 memset(ddf
->anchor
.header_ext
, 0xff, 32);
1411 ddf
->anchor
.primary_lba
= ~(__u64
)0;
1412 ddf
->anchor
.secondary_lba
= ~(__u64
)0;
1413 ddf
->anchor
.type
= DDF_HEADER_ANCHOR
;
1414 memset(ddf
->anchor
.pad2
, 0xff, 3);
1415 ddf
->anchor
.workspace_len
= __cpu_to_be32(32768); /* Must be reserved */
1416 ddf
->anchor
.workspace_lba
= ~(__u64
)0; /* Put this at bottom
1417 of 32M reserved.. */
1418 max_phys_disks
= 1023; /* Should be enough */
1419 ddf
->anchor
.max_pd_entries
= __cpu_to_be16(max_phys_disks
);
1420 max_virt_disks
= 255;
1421 ddf
->anchor
.max_vd_entries
= __cpu_to_be16(max_virt_disks
); /* ?? */
1422 ddf
->anchor
.max_partitions
= __cpu_to_be16(64); /* ?? */
1424 ddf
->anchor
.config_record_len
= __cpu_to_be16(1 + 256*12/512);
1425 ddf
->anchor
.max_primary_element_entries
= __cpu_to_be16(256);
1426 memset(ddf
->anchor
.pad3
, 0xff, 54);
1428 /* controller sections is one sector long immediately
1429 * after the ddf header */
1431 ddf
->anchor
.controller_section_offset
= __cpu_to_be32(sector
);
1432 ddf
->anchor
.controller_section_length
= __cpu_to_be32(1);
1435 /* phys is 8 sectors after that */
1436 pdsize
= ROUND_UP(sizeof(struct phys_disk
) +
1437 sizeof(struct phys_disk_entry
)*max_phys_disks
,
1439 switch(pdsize
/512) {
1440 case 2: case 8: case 32: case 128: case 512: break;
1443 ddf
->anchor
.phys_section_offset
= __cpu_to_be32(sector
);
1444 ddf
->anchor
.phys_section_length
=
1445 __cpu_to_be32(pdsize
/512); /* max_primary_element_entries/8 */
1446 sector
+= pdsize
/512;
1448 /* virt is another 32 sectors */
1449 vdsize
= ROUND_UP(sizeof(struct virtual_disk
) +
1450 sizeof(struct virtual_entry
) * max_virt_disks
,
1452 switch(vdsize
/512) {
1453 case 2: case 8: case 32: case 128: case 512: break;
1456 ddf
->anchor
.virt_section_offset
= __cpu_to_be32(sector
);
1457 ddf
->anchor
.virt_section_length
=
1458 __cpu_to_be32(vdsize
/512); /* max_vd_entries/8 */
1459 sector
+= vdsize
/512;
1461 clen
= (1 + 256*12/512) * (64+1);
1462 ddf
->anchor
.config_section_offset
= __cpu_to_be32(sector
);
1463 ddf
->anchor
.config_section_length
= __cpu_to_be32(clen
);
1466 ddf
->anchor
.data_section_offset
= __cpu_to_be32(sector
);
1467 ddf
->anchor
.data_section_length
= __cpu_to_be32(1);
1470 ddf
->anchor
.bbm_section_length
= __cpu_to_be32(0);
1471 ddf
->anchor
.bbm_section_offset
= __cpu_to_be32(0xFFFFFFFF);
1472 ddf
->anchor
.diag_space_length
= __cpu_to_be32(0);
1473 ddf
->anchor
.diag_space_offset
= __cpu_to_be32(0xFFFFFFFF);
1474 ddf
->anchor
.vendor_length
= __cpu_to_be32(0);
1475 ddf
->anchor
.vendor_offset
= __cpu_to_be32(0xFFFFFFFF);
1477 memset(ddf
->anchor
.pad4
, 0xff, 256);
1479 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
1480 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
1482 ddf
->primary
.openflag
= 1; /* I guess.. */
1483 ddf
->primary
.type
= DDF_HEADER_PRIMARY
;
1485 ddf
->secondary
.openflag
= 1; /* I guess.. */
1486 ddf
->secondary
.type
= DDF_HEADER_SECONDARY
;
1488 ddf
->active
= &ddf
->primary
;
1490 ddf
->controller
.magic
= DDF_CONTROLLER_MAGIC
;
1492 /* 24 more bytes of fiction required.
1493 * first 8 are a 'vendor-id' - "Linux-MD"
1494 * Remaining 16 are serial number.... maybe a hostname would do?
1496 memcpy(ddf
->controller
.guid
, T10
, sizeof(T10
));
1497 gethostname(hostname
, 17);
1499 hostlen
= strlen(hostname
);
1500 memcpy(ddf
->controller
.guid
+ 24 - hostlen
, hostname
, hostlen
);
1501 for (i
= strlen(T10
) ; i
+hostlen
< 24; i
++)
1502 ddf
->controller
.guid
[i
] = ' ';
1504 ddf
->controller
.type
.vendor_id
= __cpu_to_be16(0xDEAD);
1505 ddf
->controller
.type
.device_id
= __cpu_to_be16(0xBEEF);
1506 ddf
->controller
.type
.sub_vendor_id
= 0;
1507 ddf
->controller
.type
.sub_device_id
= 0;
1508 memcpy(ddf
->controller
.product_id
, "What Is My PID??", 16);
1509 memset(ddf
->controller
.pad
, 0xff, 8);
1510 memset(ddf
->controller
.vendor_data
, 0xff, 448);
1512 pd
= ddf
->phys
= malloc(pdsize
);
1513 ddf
->pdsize
= pdsize
;
1515 memset(pd
, 0xff, pdsize
);
1516 memset(pd
, 0, sizeof(*pd
));
1517 pd
->magic
= DDF_PHYS_DATA_MAGIC
;
1518 pd
->used_pdes
= __cpu_to_be16(0);
1519 pd
->max_pdes
= __cpu_to_be16(max_phys_disks
);
1520 memset(pd
->pad
, 0xff, 52);
1522 vd
= ddf
->virt
= malloc(vdsize
);
1523 ddf
->vdsize
= vdsize
;
1524 memset(vd
, 0, vdsize
);
1525 vd
->magic
= DDF_VIRT_RECORDS_MAGIC
;
1526 vd
->populated_vdes
= __cpu_to_be16(0);
1527 vd
->max_vdes
= __cpu_to_be16(max_virt_disks
);
1528 memset(vd
->pad
, 0xff, 52);
1530 for (i
=0; i
<max_virt_disks
; i
++)
1531 memset(&vd
->entries
[i
], 0xff, sizeof(struct virtual_entry
));
1537 static int all_ff(char *guid
)
1540 for (i
= 0; i
< DDF_GUID_LEN
; i
++)
1541 if (guid
[i
] != (char)0xff)
1545 static int chunk_to_shift(int chunksize
)
1547 return ffs(chunksize
/512)-1;
1550 static int level_to_prl(int level
)
1553 case LEVEL_LINEAR
: return DDF_CONCAT
;
1554 case 0: return DDF_RAID0
;
1555 case 1: return DDF_RAID1
;
1556 case 4: return DDF_RAID4
;
1557 case 5: return DDF_RAID5
;
1558 case 6: return DDF_RAID6
;
1562 static int layout_to_rlq(int level
, int layout
, int raiddisks
)
1566 return DDF_RAID0_SIMPLE
;
1569 case 2: return DDF_RAID1_SIMPLE
;
1570 case 3: return DDF_RAID1_MULTI
;
1575 case 0: return DDF_RAID4_N
;
1581 case ALGORITHM_LEFT_ASYMMETRIC
:
1582 return DDF_RAID5_N_RESTART
;
1583 case ALGORITHM_RIGHT_ASYMMETRIC
:
1584 return DDF_RAID5_0_RESTART
;
1585 case ALGORITHM_LEFT_SYMMETRIC
:
1586 return DDF_RAID5_N_CONTINUE
;
1587 case ALGORITHM_RIGHT_SYMMETRIC
:
1588 return -1; /* not mentioned in standard */
1594 static int rlq_to_layout(int rlq
, int prl
, int raiddisks
)
1598 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1600 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1608 return -1; /* FIXME this isn't checked */
1613 case DDF_RAID5_N_RESTART
:
1614 return ALGORITHM_LEFT_ASYMMETRIC
;
1615 case DDF_RAID5_0_RESTART
:
1616 return ALGORITHM_RIGHT_ASYMMETRIC
;
1617 case DDF_RAID5_N_CONTINUE
:
1618 return ALGORITHM_LEFT_SYMMETRIC
;
1626 static int init_super_ddf_bvd(struct supertype
*st
,
1627 mdu_array_info_t
*info
,
1628 unsigned long long size
,
1629 char *name
, char *homehost
,
1632 /* We are creating a BVD inside a pre-existing container.
1633 * so st->sb is already set.
1634 * We need to create a new vd_config and a new virtual_entry
1636 struct ddf_super
*ddf
= st
->sb
;
1638 struct virtual_entry
*ve
;
1640 struct vd_config
*vc
;
1644 if (__be16_to_cpu(ddf
->virt
->populated_vdes
)
1645 >= __be16_to_cpu(ddf
->virt
->max_vdes
)) {
1646 fprintf(stderr
, Name
": This ddf already has the "
1647 "maximum of %d virtual devices\n",
1648 __be16_to_cpu(ddf
->virt
->max_vdes
));
1652 for (venum
= 0; venum
< __be16_to_cpu(ddf
->virt
->max_vdes
); venum
++)
1653 if (all_ff(ddf
->virt
->entries
[venum
].guid
))
1655 if (venum
== __be16_to_cpu(ddf
->virt
->max_vdes
)) {
1656 fprintf(stderr
, Name
": Cannot find spare slot for "
1657 "virtual disk - DDF is corrupt\n");
1660 ve
= &ddf
->virt
->entries
[venum
];
1661 st
->container_member
= venum
;
1663 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1664 * timestamp, random number
1666 make_header_guid(ve
->guid
);
1667 ve
->unit
= __cpu_to_be16(info
->md_minor
);
1669 ve
->guid_crc
= crc32(0, (unsigned char*)ddf
->anchor
.guid
, DDF_GUID_LEN
);
1673 if (!(info
->state
& 1))
1674 ve
->init_state
= DDF_state_inconsistent
;
1675 memset(ve
->pad1
, 0xff, 14);
1676 memset(ve
->name
, ' ', 16);
1678 strncpy(ve
->name
, name
, 16);
1679 ddf
->virt
->populated_vdes
=
1680 __cpu_to_be16(__be16_to_cpu(ddf
->virt
->populated_vdes
)+1);
1682 /* Now create a new vd_config */
1683 conflen
= __be16_to_cpu(ddf
->active
->config_record_len
);
1684 vcl
= malloc(offsetof(struct vcl
, conf
) + conflen
* 512);
1685 mppe
= __be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
1686 vcl
->lba_offset
= (__u64
*) &vcl
->conf
.phys_refnum
[mppe
];
1690 vc
->magic
= DDF_VD_CONF_MAGIC
;
1691 memcpy(vc
->guid
, ve
->guid
, DDF_GUID_LEN
);
1692 vc
->timestamp
= __cpu_to_be32(time(0)-DECADE
);
1693 vc
->seqnum
= __cpu_to_be32(1);
1694 memset(vc
->pad0
, 0xff, 24);
1695 vc
->prim_elmnt_count
= __cpu_to_be16(info
->raid_disks
);
1696 vc
->chunk_shift
= chunk_to_shift(info
->chunk_size
);
1697 vc
->prl
= level_to_prl(info
->level
);
1698 vc
->rlq
= layout_to_rlq(info
->level
, info
->layout
, info
->raid_disks
);
1699 vc
->sec_elmnt_count
= 1;
1700 vc
->sec_elmnt_seq
= 0;
1702 vc
->blocks
= __cpu_to_be64(info
->size
* 2);
1703 vc
->array_blocks
= __cpu_to_be64(
1704 calc_array_size(info
->level
, info
->raid_disks
, info
->layout
,
1705 info
->chunk_size
, info
->size
*2));
1706 memset(vc
->pad1
, 0xff, 8);
1707 vc
->spare_refs
[0] = 0xffffffff;
1708 vc
->spare_refs
[1] = 0xffffffff;
1709 vc
->spare_refs
[2] = 0xffffffff;
1710 vc
->spare_refs
[3] = 0xffffffff;
1711 vc
->spare_refs
[4] = 0xffffffff;
1712 vc
->spare_refs
[5] = 0xffffffff;
1713 vc
->spare_refs
[6] = 0xffffffff;
1714 vc
->spare_refs
[7] = 0xffffffff;
1715 memset(vc
->cache_pol
, 0, 8);
1717 memset(vc
->pad2
, 0xff, 3);
1718 memset(vc
->pad3
, 0xff, 52);
1719 memset(vc
->pad4
, 0xff, 192);
1720 memset(vc
->v0
, 0xff, 32);
1721 memset(vc
->v1
, 0xff, 32);
1722 memset(vc
->v2
, 0xff, 16);
1723 memset(vc
->v3
, 0xff, 16);
1724 memset(vc
->vendor
, 0xff, 32);
1726 memset(vc
->phys_refnum
, 0xff, 4*mppe
);
1727 memset(vc
->phys_refnum
+mppe
, 0x00, 8*mppe
);
1729 vcl
->next
= ddf
->conflist
;
1730 ddf
->conflist
= vcl
;
1735 static void add_to_super_ddf_bvd(struct supertype
*st
,
1736 mdu_disk_info_t
*dk
, int fd
, char *devname
)
1738 /* fd and devname identify a device with-in the ddf container (st).
1739 * dk identifies a location in the new BVD.
1740 * We need to find suitable free space in that device and update
1741 * the phys_refnum and lba_offset for the newly created vd_config.
1742 * We might also want to update the type in the phys_disk
1746 struct ddf_super
*ddf
= st
->sb
;
1747 struct vd_config
*vc
;
1751 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
1752 if (dl
->major
== dk
->major
&&
1753 dl
->minor
== dk
->minor
)
1755 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1758 vc
= &ddf
->newconf
->conf
;
1759 vc
->phys_refnum
[dk
->raid_disk
] = dl
->disk
.refnum
;
1760 mppe
= __be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
1761 lba_offset
= (__u64
*)(vc
->phys_refnum
+ mppe
);
1762 lba_offset
[dk
->raid_disk
] = 0; /* FIXME */
1764 dl
->vlist
[0] =ddf
->newconf
; /* FIXME */
1767 dl
->devname
= devname
;
1770 /* add a device to a container, either while creating it or while
1771 * expanding a pre-existing container
1773 static void add_to_super_ddf(struct supertype
*st
,
1774 mdu_disk_info_t
*dk
, int fd
, char *devname
)
1776 struct ddf_super
*ddf
= st
->sb
;
1780 unsigned long long size
;
1781 struct phys_disk_entry
*pde
;
1785 /* This is device numbered dk->number. We need to create
1786 * a phys_disk entry and a more detailed disk_data entry.
1789 dd
= malloc(sizeof(*dd
) + sizeof(dd
->vlist
[0]) * (ddf
->max_part
+1));
1790 dd
->major
= major(stb
.st_rdev
);
1791 dd
->minor
= minor(stb
.st_rdev
);
1792 dd
->devname
= devname
;
1793 dd
->next
= ddf
->dlist
;
1796 dd
->disk
.magic
= DDF_PHYS_DATA_MAGIC
;
1798 tm
= localtime(&now
);
1799 sprintf(dd
->disk
.guid
, "%8s%04d%02d%02d",
1800 T10
, tm
->tm_year
+1900, tm
->tm_mon
+1, tm
->tm_mday
);
1801 *(__u32
*)(dd
->disk
.guid
+ 16) = random();
1802 *(__u32
*)(dd
->disk
.guid
+ 20) = random();
1804 dd
->disk
.refnum
= random(); /* and hope for the best FIXME check this is unique!!*/
1805 dd
->disk
.forced_ref
= 1;
1806 dd
->disk
.forced_guid
= 1;
1807 memset(dd
->disk
.vendor
, ' ', 32);
1808 memcpy(dd
->disk
.vendor
, "Linux", 5);
1809 memset(dd
->disk
.pad
, 0xff, 442);
1810 for (i
= 0; i
< ddf
->max_part
+1 ; i
++)
1811 dd
->vlist
[i
] = NULL
;
1813 n
= __be16_to_cpu(ddf
->phys
->used_pdes
);
1814 pde
= &ddf
->phys
->entries
[n
];
1816 ddf
->phys
->used_pdes
= __cpu_to_be16(n
);
1818 memcpy(pde
->guid
, dd
->disk
.guid
, DDF_GUID_LEN
);
1819 pde
->refnum
= dd
->disk
.refnum
;
1820 pde
->type
= __cpu_to_be16(DDF_Forced_PD_GUID
|DDF_Global_Spare
);
1821 pde
->state
= __cpu_to_be16(DDF_Online
);
1822 get_dev_size(fd
, NULL
, &size
);
1823 /* We are required to reserve 32Meg, and record the size in sectors */
1824 pde
->config_size
= __cpu_to_be64( (size
- 32*1024*1024) / 512);
1825 sprintf(pde
->path
, "%17.17s","Information: nil") ;
1826 memset(pde
->pad
, 0xff, 6);
1832 * This is the write_init_super method for a ddf container. It is
1833 * called when creating a container or adding another device to a
1838 static int write_init_super_ddf(struct supertype
*st
)
1841 struct ddf_super
*ddf
= st
->sb
;
1847 unsigned long long size
, sector
;
1849 for (d
= ddf
->dlist
; d
; d
=d
->next
) {
1855 /* We need to fill in the primary, (secondary) and workspace
1856 * lba's in the headers, set their checksums,
1857 * Also checksum phys, virt....
1859 * Then write everything out, finally the anchor is written.
1861 get_dev_size(fd
, NULL
, &size
);
1863 ddf
->anchor
.workspace_lba
= __cpu_to_be64(size
- 32*1024*2);
1864 ddf
->anchor
.primary_lba
= __cpu_to_be64(size
- 16*1024*2);
1865 ddf
->anchor
.seq
= __cpu_to_be32(1);
1866 memcpy(&ddf
->primary
, &ddf
->anchor
, 512);
1867 memcpy(&ddf
->secondary
, &ddf
->anchor
, 512);
1869 ddf
->anchor
.openflag
= 0xFF; /* 'open' means nothing */
1870 ddf
->anchor
.seq
= 0xFFFFFFFF; /* no sequencing in anchor */
1871 ddf
->anchor
.crc
= calc_crc(&ddf
->anchor
, 512);
1873 ddf
->primary
.openflag
= 0;
1874 ddf
->primary
.type
= DDF_HEADER_PRIMARY
;
1876 ddf
->secondary
.openflag
= 0;
1877 ddf
->secondary
.type
= DDF_HEADER_SECONDARY
;
1879 ddf
->primary
.crc
= calc_crc(&ddf
->primary
, 512);
1880 ddf
->secondary
.crc
= calc_crc(&ddf
->secondary
, 512);
1882 sector
= size
- 16*1024*2;
1883 lseek64(fd
, sector
<<9, 0);
1884 write(fd
, &ddf
->primary
, 512);
1886 ddf
->controller
.crc
= calc_crc(&ddf
->controller
, 512);
1887 write(fd
, &ddf
->controller
, 512);
1889 ddf
->phys
->crc
= calc_crc(ddf
->phys
, ddf
->pdsize
);
1891 write(fd
, ddf
->phys
, ddf
->pdsize
);
1893 ddf
->virt
->crc
= calc_crc(ddf
->virt
, ddf
->vdsize
);
1894 write(fd
, ddf
->virt
, ddf
->vdsize
);
1896 /* Now write lots of config records. */
1897 n_config
= __be16_to_cpu(ddf
->active
->max_partitions
);
1898 conf_size
= __be16_to_cpu(ddf
->active
->config_record_len
) * 512;
1899 for (i
= 0 ; i
<= n_config
; i
++) {
1900 struct vcl
*c
= d
->vlist
[i
];
1903 c
->conf
.crc
= calc_crc(&c
->conf
, conf_size
);
1904 write(fd
, &c
->conf
, conf_size
);
1906 __u32 sig
= 0xffffffff;
1908 lseek64(fd
, conf_size
-4, SEEK_CUR
);
1911 d
->disk
.crc
= calc_crc(&d
->disk
, 512);
1912 write(fd
, &d
->disk
, 512);
1914 /* Maybe do the same for secondary */
1916 lseek64(fd
, (size
-1)*512, SEEK_SET
);
1917 write(fd
, &ddf
->anchor
, 512);
1924 static __u64
avail_size_ddf(struct supertype
*st
, __u64 devsize
)
1926 /* We must reserve the last 32Meg */
1927 if (devsize
<= 32*1024*2)
1929 return devsize
- 32*1024*2;
1933 int validate_geometry_ddf(struct supertype
*st
,
1934 int level
, int layout
, int raiddisks
,
1935 int chunk
, unsigned long long size
,
1936 char *dev
, unsigned long long *freesize
)
1942 /* ddf potentially supports lots of things, but it depends on
1943 * what devices are offered (and maybe kernel version?)
1944 * If given unused devices, we will make a container.
1945 * If given devices in a container, we will make a BVD.
1946 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
1949 if (level
== LEVEL_CONTAINER
) {
1950 st
->ss
= &super_ddf_container
;
1952 int rv
=st
->ss
->validate_geometry(st
, level
, layout
,
1959 return st
->ss
->validate_geometry(st
, level
, layout
, raiddisks
,
1960 chunk
, size
, dev
, freesize
);
1964 /* creating in a given container */
1965 st
->ss
= &super_ddf_bvd
;
1967 int rv
=st
->ss
->validate_geometry(st
, level
, layout
,
1974 return st
->ss
->validate_geometry(st
, level
, layout
, raiddisks
,
1975 chunk
, size
, dev
, freesize
);
1977 /* FIXME should exclude MULTIPATH, or more appropriately, allow
1978 * only known levels.
1983 /* This device needs to be either a device in a 'ddf' container,
1984 * or it needs to be a 'ddf-bvd' array.
1987 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1989 sra
= sysfs_read(fd
, 0, GET_VERSION
);
1991 if (sra
&& sra
->array
.major_version
== -1 &&
1992 strcmp(sra
->text_version
, "ddf-bvd") == 0) {
1993 st
->ss
= &super_ddf_svd
;
1994 return st
->ss
->validate_geometry(st
, level
, layout
,
1995 raiddisks
, chunk
, size
,
2000 Name
": Cannot create this array on device %s\n",
2004 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
2005 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2006 dev
, strerror(errno
));
2009 /* Well, it is in use by someone, maybe a 'ddf' container. */
2010 cfd
= open_container(fd
);
2013 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
2017 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
2019 if (sra
&& sra
->array
.major_version
== -1 &&
2020 strcmp(sra
->text_version
, "ddf") == 0) {
2021 /* This is a member of a ddf container. Load the container
2022 * and try to create a bvd
2024 struct ddf_super
*ddf
;
2025 st
->ss
= &super_ddf_bvd
;
2026 if (load_super_ddf_all(st
, cfd
, (void **)&ddf
, NULL
, 1) == 0) {
2028 st
->container_dev
= fd2devnum(cfd
);
2029 st
->container_member
= 27; // FIXME
2031 return st
->ss
->validate_geometry(st
, level
, layout
,
2032 raiddisks
, chunk
, size
,
2037 fprintf(stderr
, Name
": Cannot use %s: Already in use\n",
2042 int validate_geometry_ddf_container(struct supertype
*st
,
2043 int level
, int layout
, int raiddisks
,
2044 int chunk
, unsigned long long size
,
2045 char *dev
, unsigned long long *freesize
)
2048 unsigned long long ldsize
;
2050 if (level
!= LEVEL_CONTAINER
)
2055 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2057 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2058 dev
, strerror(errno
));
2061 if (!get_dev_size(fd
, dev
, &ldsize
)) {
2067 *freesize
= avail_size_ddf(st
, ldsize
);
2073 unsigned long long start
, size
;
2075 int cmp_extent(const void *av
, const void *bv
)
2077 const struct extent
*a
= av
;
2078 const struct extent
*b
= bv
;
2079 if (a
->start
< b
->start
)
2081 if (a
->start
> b
->start
)
2086 struct extent
*get_extents(struct ddf_super
*ddf
, struct dl
*dl
)
2088 /* find a list of used extents on the give physical device
2089 * (dnum) or the given ddf.
2090 * Return a malloced array of 'struct extent'
2092 FIXME ignore DDF_Legacy devices?
2100 for (dnum
= 0; dnum
< ddf
->phys
->used_pdes
; dnum
++)
2101 if (memcmp(dl
->disk
.guid
,
2102 ddf
->phys
->entries
[dnum
].guid
,
2106 if (dnum
== ddf
->phys
->used_pdes
)
2109 rv
= malloc(sizeof(struct extent
) * (ddf
->max_part
+ 2));
2113 for (i
= 0; i
< ddf
->max_part
+1; i
++) {
2114 struct vcl
*v
= dl
->vlist
[i
];
2117 for (j
=0; j
< v
->conf
.prim_elmnt_count
; j
++)
2118 if (v
->conf
.phys_refnum
[j
] == dl
->disk
.refnum
) {
2119 /* This device plays role 'j' in 'v'. */
2120 rv
[n
].start
= __be64_to_cpu(v
->lba_offset
[j
]);
2121 rv
[n
].size
= __be64_to_cpu(v
->conf
.blocks
);
2126 qsort(rv
, n
, sizeof(*rv
), cmp_extent
);
2128 rv
[n
].start
= __be64_to_cpu(ddf
->phys
->entries
[dnum
].config_size
);
2133 int validate_geometry_ddf_bvd(struct supertype
*st
,
2134 int level
, int layout
, int raiddisks
,
2135 int chunk
, unsigned long long size
,
2136 char *dev
, unsigned long long *freesize
)
2139 struct ddf_super
*ddf
= st
->sb
;
2141 unsigned long long pos
= 0;
2142 unsigned long long maxsize
;
2145 /* ddf/bvd supports lots of things, but not containers */
2146 if (level
== LEVEL_CONTAINER
)
2148 /* We must have the container info already read in. */
2153 /* General test: make sure there is space for
2154 * 'raiddisks' device extents of size 'size'.
2156 unsigned long long minsize
= size
;
2160 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
)
2165 e
= get_extents(ddf
, dl
);
2168 unsigned long long esize
;
2169 esize
= e
[i
].start
- pos
;
2170 if (esize
>= minsize
)
2172 pos
= e
[i
].start
+ e
[i
].size
;
2174 } while (e
[i
-1].size
);
2179 if (dcnt
< raiddisks
) {
2180 fprintf(stderr
, Name
": Not enough devices with space "
2181 "for this array (%d < %d)\n",
2187 /* This device must be a member of the set */
2188 if (stat(dev
, &stb
) < 0)
2190 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
2192 for (dl
= ddf
->dlist
; dl
; dl
= dl
->next
) {
2193 if (dl
->major
== major(stb
.st_rdev
) &&
2194 dl
->minor
== minor(stb
.st_rdev
))
2198 fprintf(stderr
, Name
": %s is not in the same DDF set\n",
2202 e
= get_extents(ddf
, dl
);
2206 unsigned long long esize
;
2207 esize
= e
[i
].start
- pos
;
2208 if (esize
>= maxsize
)
2210 pos
= e
[i
].start
+ e
[i
].size
;
2212 } while (e
[i
-1].size
);
2213 *freesize
= maxsize
;
2218 int validate_geometry_ddf_svd(struct supertype
*st
,
2219 int level
, int layout
, int raiddisks
,
2220 int chunk
, unsigned long long size
,
2221 char *dev
, unsigned long long *freesize
)
2223 /* dd/svd only supports striped, mirrored, concat, spanned... */
2224 if (level
!= LEVEL_LINEAR
&&
2232 static int load_super_ddf_all(struct supertype
*st
, int fd
,
2233 void **sbp
, char *devname
, int keep_fd
)
2236 struct ddf_super
*super
;
2237 struct mdinfo
*sd
, *best
= NULL
;
2243 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
2246 if (sra
->array
.major_version
!= -1 ||
2247 sra
->array
.minor_version
!= -2 ||
2248 strcmp(sra
->text_version
, "ddf") != 0)
2251 super
= malloc(sizeof(*super
));
2255 /* first, try each device, and choose the best ddf */
2256 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
2258 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2259 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
2262 rv
= load_ddf_headers(dfd
, super
, NULL
);
2263 if (!keep_fd
) close(dfd
);
2265 seq
= __be32_to_cpu(super
->active
->seq
);
2266 if (super
->active
->openflag
)
2268 if (!best
|| seq
> bestseq
) {
2276 /* OK, load this ddf */
2277 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
2278 dfd
= dev_open(nm
, O_RDONLY
);
2281 load_ddf_headers(dfd
, super
, NULL
);
2282 load_ddf_global(dfd
, super
, NULL
);
2284 /* Now we need the device-local bits */
2285 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
2286 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2287 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
2290 seq
= load_ddf_local(dfd
, super
, NULL
, keep_fd
);
2291 if (!keep_fd
) close(dfd
);
2294 if (st
->ss
== NULL
) {
2295 st
->ss
= &super_ddf_container
;
2296 st
->minor_version
= 0;
2305 static struct mdinfo
*container_content_ddf(struct supertype
*st
)
2307 /* Given a container loaded by load_super_ddf_all,
2308 * extract information about all the arrays into
2311 * For each vcl in conflist: create an mdinfo, fill it in,
2312 * then look for matching devices (phys_refnum) in dlist
2313 * and create appropriate device mdinfo.
2315 struct ddf_super
*ddf
= st
->sb
;
2316 struct mdinfo
*rest
= NULL
;
2319 for (vc
= ddf
->conflist
; vc
; vc
=vc
->next
)
2323 struct mdinfo
*this;
2324 this = malloc(sizeof(*this));
2325 memset(this, 0, sizeof(*this));
2329 this->array
.major_version
= 1000;
2330 this->array
.minor_version
= 0;
2331 this->array
.patch_version
= 0;
2332 this->array
.level
= map_num1(ddf_level_num
, vc
->conf
.prl
);
2333 this->array
.raid_disks
=
2334 __be16_to_cpu(vc
->conf
.prim_elmnt_count
);
2335 /* FIXME this should be mapped */
2336 this->array
.layout
= vc
->conf
.rlq
;
2337 this->array
.md_minor
= -1;
2338 this->array
.ctime
= DECADE
+
2339 __be32_to_cpu(*(__u32
*)(vc
->conf
.guid
+16));
2340 this->array
.utime
= DECADE
+
2341 __be32_to_cpu(vc
->conf
.timestamp
);
2342 this->array
.chunk_size
= 512 << vc
->conf
.chunk_shift
;
2344 for (i
=0; i
< __be16_to_cpu(ddf
->virt
->populated_vdes
); i
++)
2345 if (memcmp(ddf
->virt
->entries
[i
].guid
,
2346 vc
->conf
.guid
, DDF_GUID_LEN
) == 0)
2348 if (ddf
->virt
->entries
[i
].state
& DDF_state_inconsistent
)
2349 this->array
.state
= 0;
2351 this->array
.state
= 1;
2352 memcpy(this->name
, ddf
->virt
->entries
[i
].name
, 32);
2355 memset(this->uuid
, 0, sizeof(this->uuid
));
2356 this->component_size
= __be64_to_cpu(vc
->conf
.blocks
);
2357 this->array
.size
= this->component_size
/ 2;
2358 this->container_member
= i
;
2360 mppe
= __be16_to_cpu(ddf
->anchor
.max_primary_element_entries
);
2361 for (i
=0 ; i
< mppe
; i
++) {
2365 if (vc
->conf
.phys_refnum
[i
] == 0xFFFFFFFF)
2368 this->array
.working_disks
++;
2370 for (d
= ddf
->dlist
; d
; d
=d
->next
)
2371 if (d
->disk
.refnum
== vc
->conf
.phys_refnum
[i
])
2376 dev
= malloc(sizeof(*dev
));
2377 memset(dev
, 0, sizeof(*dev
));
2378 dev
->next
= this->devs
;
2381 dev
->disk
.number
= __be32_to_cpu(d
->disk
.refnum
);
2382 dev
->disk
.major
= d
->major
;
2383 dev
->disk
.minor
= d
->minor
;
2384 dev
->disk
.raid_disk
= i
;
2385 dev
->disk
.state
= (1<<MD_DISK_SYNC
)|(1<<MD_DISK_ACTIVE
);
2387 dev
->events
= __le32_to_cpu(ddf
->primary
.seq
);
2388 dev
->data_offset
= vc
->lba_offset
[i
];
2389 dev
->component_size
= __be64_to_cpu(vc
->conf
.blocks
);
2391 strcpy(dev
->name
, d
->devname
);
2397 static int init_zero_ddf(struct supertype
*st
,
2398 mdu_array_info_t
*info
,
2399 unsigned long long size
, char *name
,
2400 char *homehost
, int *uuid
)
2406 static int store_zero_ddf(struct supertype
*st
, int fd
)
2408 unsigned long long dsize
;
2410 memset(buf
, 0, 512);
2413 if (!get_dev_size(fd
, NULL
, &dsize
))
2416 lseek64(fd
, dsize
-512, 0);
2417 write(fd
, buf
, 512);
2421 static int compare_super_ddf(struct supertype
*st
, struct supertype
*tst
)
2425 * 0 same, or first was empty, and second was copied
2426 * 1 second had wrong number
2428 * 3 wrong other info
2430 struct ddf_super
*first
= st
->sb
;
2431 struct ddf_super
*second
= tst
->sb
;
2439 if (memcmp(first
->anchor
.guid
, second
->anchor
.guid
, DDF_GUID_LEN
) != 0)
2442 /* FIXME should I look at anything else? */
2446 static int ddf_open_new(struct supertype
*c
, struct active_array
*a
, int inst
)
2448 fprintf(stderr
, "ddf: open_new %d\n", inst
);
2452 static void ddf_mark_clean(struct active_array
*a
, unsigned long long sync_pos
)
2454 fprintf(stderr
, "ddf: mark clean %llu\n", sync_pos
);
2457 static void ddf_mark_dirty(struct active_array
*a
)
2459 fprintf(stderr
, "ddf: mark dirty\n");
2462 static void ddf_mark_sync(struct active_array
*a
, unsigned long long resync
)
2464 fprintf(stderr
, "ddf: mark sync\n");
2467 static void ddf_set_disk(struct active_array
*a
, int n
)
2469 fprintf(stderr
, "ddf: set_disk %d\n", n
);
2472 static void ddf_sync_metadata(struct active_array
*a
)
2474 fprintf(stderr
, "ddf: sync_metadata\n");
2477 struct superswitch super_ddf
= {
2479 .examine_super
= examine_super_ddf
,
2480 .brief_examine_super
= brief_examine_super_ddf
,
2481 .detail_super
= detail_super_ddf
,
2482 .brief_detail_super
= brief_detail_super_ddf
,
2483 .validate_geometry
= validate_geometry_ddf
,
2485 .match_home
= match_home_ddf
,
2486 .uuid_from_super
= uuid_from_super_ddf
,
2487 .getinfo_super
= getinfo_super_ddf
,
2488 .update_super
= update_super_ddf
,
2490 .avail_size
= avail_size_ddf
,
2492 .compare_super
= compare_super_ddf
,
2494 .load_super
= load_super_ddf
,
2495 .init_super
= init_zero_ddf
,
2496 .store_super
= store_zero_ddf
,
2497 .free_super
= free_super_ddf
,
2498 .match_metadata_desc
= match_metadata_desc_ddf
,
2499 .getinfo_super_n
= getinfo_super_n_container
,
2505 .text_version
= "ddf",
2508 .open_new
= ddf_open_new
,
2509 .load_super
= load_super_ddf
,
2510 .mark_clean
= ddf_mark_clean
,
2511 .mark_dirty
= ddf_mark_dirty
,
2512 .mark_sync
= ddf_mark_sync
,
2513 .set_disk
= ddf_set_disk
,
2514 .sync_metadata
= ddf_sync_metadata
,
2519 /* Super_ddf_container is set by validate_geometry_ddf when given a
2520 * device that is not part of any array
2522 struct superswitch super_ddf_container
= {
2524 .validate_geometry
= validate_geometry_ddf_container
,
2525 .write_init_super
= write_init_super_ddf
,
2528 .init_super
= init_super_ddf
,
2529 .add_to_super
= add_to_super_ddf
,
2531 .free_super
= free_super_ddf
,
2533 .container_content
= container_content_ddf
,
2538 .text_version
= "ddf",
2541 struct superswitch super_ddf_bvd
= {
2543 // .detail_super = detail_super_ddf_bvd,
2544 // .brief_detail_super = brief_detail_super_ddf_bvd,
2545 .validate_geometry
= validate_geometry_ddf_bvd
,
2546 .write_init_super
= write_init_super_ddf
,
2548 .update_super
= update_super_ddf
,
2549 .init_super
= init_super_ddf_bvd
,
2550 .add_to_super
= add_to_super_ddf_bvd
,
2551 .getinfo_super
= getinfo_super_ddf_bvd
,
2552 .getinfo_super_n
= getinfo_super_n_bvd
,
2554 .load_super
= load_super_ddf
,
2555 .free_super
= free_super_ddf
,
2556 .match_metadata_desc
= match_metadata_desc_ddf_bvd
,
2562 .text_version
= "ddf",
2565 struct superswitch super_ddf_svd
= {
2567 // .detail_super = detail_super_ddf_svd,
2568 // .brief_detail_super = brief_detail_super_ddf_svd,
2569 .validate_geometry
= validate_geometry_ddf_svd
,
2571 .update_super
= update_super_ddf
,
2572 .init_super
= init_super_ddf
,
2574 .load_super
= load_super_ddf
,
2575 .free_super
= free_super_ddf
,
2576 .match_metadata_desc
= match_metadata_desc_ddf_svd
,
2581 .text_version
= "ddf",