]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-ddf.c
DDF: add_to_super_ddf: allow empty slots in phys disk table
[thirdparty/mdadm.git] / super-ddf.c
CommitLineData
a322f70c
DW
1/*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
e736b623 4 * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
a322f70c
DW
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neil@brown.name>
23 *
24 * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2
25 * (July 28 2006). Reused by permission of SNIA.
26 */
27
28#define HAVE_STDINT_H 1
29#include "mdadm.h"
549e9569 30#include "mdmon.h"
a322f70c
DW
31#include "sha1.h"
32#include <values.h>
33
a322f70c
DW
34/* a non-official T10 name for creation GUIDs */
35static char T10[] = "Linux-MD";
36
37/* DDF timestamps are 1980 based, so we need to add
38 * second-in-decade-of-seventies to convert to linux timestamps.
39 * 10 years with 2 leap years.
40 */
41#define DECADE (3600*24*(365*10+2))
42unsigned long crc32(
43 unsigned long crc,
44 const unsigned char *buf,
45 unsigned len);
46
bedbf68a 47#define DDF_NOTFOUND (~0U)
48#define DDF_CONTAINER (DDF_NOTFOUND-1)
49
a322f70c
DW
50/* The DDF metadata handling.
51 * DDF metadata lives at the end of the device.
52 * The last 512 byte block provides an 'anchor' which is used to locate
53 * the rest of the metadata which usually lives immediately behind the anchor.
54 *
55 * Note:
56 * - all multibyte numeric fields are bigendian.
57 * - all strings are space padded.
58 *
59 */
60
61/* Primary Raid Level (PRL) */
62#define DDF_RAID0 0x00
63#define DDF_RAID1 0x01
64#define DDF_RAID3 0x03
65#define DDF_RAID4 0x04
66#define DDF_RAID5 0x05
67#define DDF_RAID1E 0x11
68#define DDF_JBOD 0x0f
69#define DDF_CONCAT 0x1f
70#define DDF_RAID5E 0x15
71#define DDF_RAID5EE 0x25
59e36268 72#define DDF_RAID6 0x06
a322f70c
DW
73
74/* Raid Level Qualifier (RLQ) */
75#define DDF_RAID0_SIMPLE 0x00
76#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
77#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
78#define DDF_RAID3_0 0x00 /* parity in first extent */
79#define DDF_RAID3_N 0x01 /* parity in last extent */
80#define DDF_RAID4_0 0x00 /* parity in first extent */
81#define DDF_RAID4_N 0x01 /* parity in last extent */
82/* these apply to raid5e and raid5ee as well */
83#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
59e36268 84#define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
a322f70c
DW
85#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
86#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
87
88#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
89#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
90
91/* Secondary RAID Level (SRL) */
92#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
93#define DDF_2MIRRORED 0x01
94#define DDF_2CONCAT 0x02
95#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
96
97/* Magic numbers */
98#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
99#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
100#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
101#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
102#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
103#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
104#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
105#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
106#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
107#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
108
109#define DDF_GUID_LEN 24
59e36268
NB
110#define DDF_REVISION_0 "01.00.00"
111#define DDF_REVISION_2 "01.02.00"
a322f70c
DW
112
113struct ddf_header {
88c164f4 114 __u32 magic; /* DDF_HEADER_MAGIC */
a322f70c
DW
115 __u32 crc;
116 char guid[DDF_GUID_LEN];
59e36268 117 char revision[8]; /* 01.02.00 */
a322f70c
DW
118 __u32 seq; /* starts at '1' */
119 __u32 timestamp;
120 __u8 openflag;
121 __u8 foreignflag;
122 __u8 enforcegroups;
123 __u8 pad0; /* 0xff */
124 __u8 pad1[12]; /* 12 * 0xff */
125 /* 64 bytes so far */
126 __u8 header_ext[32]; /* reserved: fill with 0xff */
127 __u64 primary_lba;
128 __u64 secondary_lba;
129 __u8 type;
130 __u8 pad2[3]; /* 0xff */
131 __u32 workspace_len; /* sectors for vendor space -
132 * at least 32768(sectors) */
133 __u64 workspace_lba;
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
139 *12/512) */
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;
157 __u32 vendor_offset;
158 __u32 vendor_length;
159 /* 256 bytes so far */
160 __u8 pad4[256]; /* 0xff */
161};
162
163/* type field */
164#define DDF_HEADER_ANCHOR 0x00
165#define DDF_HEADER_PRIMARY 0x01
166#define DDF_HEADER_SECONDARY 0x02
167
168/* The content of the 'controller section' - global scope */
169struct ddf_controller_data {
88c164f4 170 __u32 magic; /* DDF_CONTROLLER_MAGIC */
a322f70c
DW
171 __u32 crc;
172 char guid[DDF_GUID_LEN];
173 struct controller_type {
174 __u16 vendor_id;
175 __u16 device_id;
176 __u16 sub_vendor_id;
177 __u16 sub_device_id;
178 } type;
179 char product_id[16];
180 __u8 pad[8]; /* 0xff */
181 __u8 vendor_data[448];
182};
183
184/* The content of phys_section - global scope */
185struct phys_disk {
88c164f4 186 __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */
a322f70c
DW
187 __u32 crc;
188 __u16 used_pdes;
189 __u16 max_pdes;
190 __u8 pad[52];
191 struct phys_disk_entry {
192 char guid[DDF_GUID_LEN];
193 __u32 refnum;
194 __u16 type;
195 __u16 state;
196 __u64 config_size; /* DDF structures must be after here */
197 char path[18]; /* another horrible structure really */
198 __u8 pad[6];
199 } entries[0];
200};
201
202/* phys_disk_entry.type is a bitmap - bigendian remember */
203#define DDF_Forced_PD_GUID 1
204#define DDF_Active_in_VD 2
88c164f4 205#define DDF_Global_Spare 4 /* VD_CONF records are ignored */
a322f70c
DW
206#define DDF_Spare 8 /* overrides Global_spare */
207#define DDF_Foreign 16
208#define DDF_Legacy 32 /* no DDF on this device */
209
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
215
216/* phys_disk_entry.state is a bigendian bitmap */
217#define DDF_Online 1
218#define DDF_Failed 2 /* overrides 1,4,8 */
219#define DDF_Rebuilding 4
220#define DDF_Transition 8
221#define DDF_SMART 16
222#define DDF_ReadErrors 32
223#define DDF_Missing 64
224
225/* The content of the virt_section global scope */
226struct virtual_disk {
88c164f4 227 __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */
a322f70c
DW
228 __u32 crc;
229 __u16 populated_vdes;
230 __u16 max_vdes;
231 __u8 pad[52];
232 struct virtual_entry {
233 char guid[DDF_GUID_LEN];
234 __u16 unit;
235 __u16 pad0; /* 0xffff */
236 __u16 guid_crc;
237 __u16 type;
238 __u8 state;
239 __u8 init_state;
240 __u8 pad1[14];
241 char name[16];
242 } entries[0];
243};
244
245/* virtual_entry.type is a bitmap - bigendian */
246#define DDF_Shared 1
247#define DDF_Enforce_Groups 2
248#define DDF_Unicode 4
249#define DDF_Owner_Valid 8
250
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
7a7cc504 258#define DDF_state_part_optimal 0x5
a322f70c
DW
259
260#define DDF_state_morphing 0x8
261#define DDF_state_inconsistent 0x10
262
263/* virtual_entry.init_state is a bigendian bitmap */
264#define DDF_initstate_mask 0x03
265#define DDF_init_not 0x00
7a7cc504
NB
266#define DDF_init_quick 0x01 /* initialisation is progress.
267 * i.e. 'state_inconsistent' */
a322f70c
DW
268#define DDF_init_full 0x02
269
270#define DDF_access_mask 0xc0
271#define DDF_access_rw 0x00
272#define DDF_access_ro 0x80
273#define DDF_access_blocked 0xc0
274
275/* The content of the config_section - local scope
276 * It has multiple records each config_record_len sectors
277 * They can be vd_config or spare_assign
278 */
279
280struct vd_config {
88c164f4 281 __u32 magic; /* DDF_VD_CONF_MAGIC */
a322f70c
DW
282 __u32 crc;
283 char guid[DDF_GUID_LEN];
284 __u32 timestamp;
285 __u32 seqnum;
286 __u8 pad0[24];
287 __u16 prim_elmnt_count;
288 __u8 chunk_shift; /* 0 == 512, 1==1024 etc */
289 __u8 prl;
290 __u8 rlq;
291 __u8 sec_elmnt_count;
292 __u8 sec_elmnt_seq;
293 __u8 srl;
598f0d58
NB
294 __u64 blocks; /* blocks per component could be different
295 * on different component devices...(only
296 * for concat I hope) */
297 __u64 array_blocks; /* blocks in array */
a322f70c
DW
298 __u8 pad1[8];
299 __u32 spare_refs[8];
300 __u8 cache_pol[8];
301 __u8 bg_rate;
302 __u8 pad2[3];
303 __u8 pad3[52];
304 __u8 pad4[192];
305 __u8 v0[32]; /* reserved- 0xff */
306 __u8 v1[32]; /* reserved- 0xff */
307 __u8 v2[16]; /* reserved- 0xff */
308 __u8 v3[16]; /* reserved- 0xff */
309 __u8 vendor[32];
310 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
311 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
312 bvd are always the same size */
313};
57a66662 314#define LBA_OFFSET(ddf, vd) ((__u64 *) &(vd)->phys_refnum[(ddf)->mppe])
a322f70c
DW
315
316/* vd_config.cache_pol[7] is a bitmap */
317#define DDF_cache_writeback 1 /* else writethrough */
318#define DDF_cache_wadaptive 2 /* only applies if writeback */
319#define DDF_cache_readahead 4
320#define DDF_cache_radaptive 8 /* only if doing read-ahead */
321#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
322#define DDF_cache_wallowed 32 /* enable write caching */
323#define DDF_cache_rallowed 64 /* enable read caching */
324
325struct spare_assign {
88c164f4 326 __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */
a322f70c
DW
327 __u32 crc;
328 __u32 timestamp;
329 __u8 reserved[7];
330 __u8 type;
331 __u16 populated; /* SAEs used */
332 __u16 max; /* max SAEs */
333 __u8 pad[8];
334 struct spare_assign_entry {
335 char guid[DDF_GUID_LEN];
336 __u16 secondary_element;
337 __u8 pad[6];
338 } spare_ents[0];
339};
340/* spare_assign.type is a bitmap */
341#define DDF_spare_dedicated 0x1 /* else global */
342#define DDF_spare_revertible 0x2 /* else committable */
343#define DDF_spare_active 0x4 /* else not active */
344#define DDF_spare_affinity 0x8 /* enclosure affinity */
345
346/* The data_section contents - local scope */
347struct disk_data {
88c164f4 348 __u32 magic; /* DDF_PHYS_DATA_MAGIC */
a322f70c
DW
349 __u32 crc;
350 char guid[DDF_GUID_LEN];
351 __u32 refnum; /* crc of some magic drive data ... */
352 __u8 forced_ref; /* set when above was not result of magic */
353 __u8 forced_guid; /* set if guid was forced rather than magic */
354 __u8 vendor[32];
355 __u8 pad[442];
356};
357
358/* bbm_section content */
359struct bad_block_log {
360 __u32 magic;
361 __u32 crc;
362 __u16 entry_count;
363 __u32 spare_count;
364 __u8 pad[10];
365 __u64 first_spare;
366 struct mapped_block {
367 __u64 defective_start;
368 __u32 replacement_start;
369 __u16 remap_count;
370 __u8 pad[2];
371 } entries[0];
372};
373
374/* Struct for internally holding ddf structures */
375/* The DDF structure stored on each device is potentially
376 * quite different, as some data is global and some is local.
377 * The global data is:
378 * - ddf header
379 * - controller_data
380 * - Physical disk records
381 * - Virtual disk records
382 * The local data is:
383 * - Configuration records
384 * - Physical Disk data section
385 * ( and Bad block and vendor which I don't care about yet).
386 *
387 * The local data is parsed into separate lists as it is read
388 * and reconstructed for writing. This means that we only need
389 * to make config changes once and they are automatically
390 * propagated to all devices.
391 * Note that the ddf_super has space of the conf and disk data
392 * for this disk and also for a list of all such data.
393 * The list is only used for the superblock that is being
394 * built in Create or Assemble to describe the whole array.
395 */
396struct ddf_super {
6416d527 397 struct ddf_header anchor, primary, secondary;
a322f70c 398 struct ddf_controller_data controller;
6416d527 399 struct ddf_header *active;
a322f70c
DW
400 struct phys_disk *phys;
401 struct virtual_disk *virt;
402 int pdsize, vdsize;
f21e18ca 403 unsigned int max_part, mppe, conf_rec_len;
d2ca6449 404 int currentdev;
18a2f463 405 int updates_pending;
a322f70c 406 struct vcl {
6416d527
NB
407 union {
408 char space[512];
409 struct {
410 struct vcl *next;
f21e18ca 411 unsigned int vcnum; /* index into ->virt */
8ec5d685 412 struct vd_config **other_bvds;
6416d527
NB
413 __u64 *block_sizes; /* NULL if all the same */
414 };
415 };
a322f70c 416 struct vd_config conf;
d2ca6449 417 } *conflist, *currentconf;
a322f70c 418 struct dl {
6416d527
NB
419 union {
420 char space[512];
421 struct {
422 struct dl *next;
423 int major, minor;
424 char *devname;
425 int fd;
426 unsigned long long size; /* sectors */
097bcf00 427 unsigned long long primary_lba; /* sectors */
428 unsigned long long secondary_lba; /* sectors */
429 unsigned long long workspace_lba; /* sectors */
6416d527
NB
430 int pdnum; /* index in ->phys */
431 struct spare_assign *spare;
8592f29d
N
432 void *mdupdate; /* hold metadata update */
433
434 /* These fields used by auto-layout */
435 int raiddisk; /* slot to fill in autolayout */
436 __u64 esize;
6416d527
NB
437 };
438 };
a322f70c 439 struct disk_data disk;
b2280677 440 struct vcl *vlist[0]; /* max_part in size */
2cc2983d 441 } *dlist, *add_list;
a322f70c
DW
442};
443
444#ifndef offsetof
445#define offsetof(t,f) ((size_t)&(((t*)0)->f))
446#endif
447
7d5a7ff3 448#if DEBUG
fb9d0acb 449static int all_ff(const char *guid);
7d5a7ff3 450static void pr_state(struct ddf_super *ddf, const char *msg)
451{
452 unsigned int i;
453 dprintf("%s/%s: ", __func__, msg);
454 for (i = 0; i < __be16_to_cpu(ddf->active->max_vd_entries); i++) {
455 if (all_ff(ddf->virt->entries[i].guid))
456 continue;
457 dprintf("%u(s=%02x i=%02x) ", i,
458 ddf->virt->entries[i].state,
459 ddf->virt->entries[i].init_state);
460 }
461 dprintf("\n");
462}
463#else
464static void pr_state(const struct ddf_super *ddf, const char *msg) {}
465#endif
466
467#define ddf_set_updates_pending(x) \
468 do { (x)->updates_pending = 1; pr_state(x, __func__); } while (0)
469
fcc22180 470static unsigned int get_pd_index_from_refnum(const struct vcl *vc,
471 __u32 refnum, unsigned int nmax,
472 const struct vd_config **bvd,
473 unsigned int *idx);
474
f21e18ca 475static unsigned int calc_crc(void *buf, int len)
a322f70c
DW
476{
477 /* crcs are always at the same place as in the ddf_header */
478 struct ddf_header *ddf = buf;
479 __u32 oldcrc = ddf->crc;
480 __u32 newcrc;
481 ddf->crc = 0xffffffff;
482
483 newcrc = crc32(0, buf, len);
484 ddf->crc = oldcrc;
4abe6b70
N
485 /* The crc is store (like everything) bigendian, so convert
486 * here for simplicity
487 */
488 return __cpu_to_be32(newcrc);
a322f70c
DW
489}
490
a3163bf0 491#define DDF_INVALID_LEVEL 0xff
492#define DDF_NO_SECONDARY 0xff
493static int err_bad_md_layout(const mdu_array_info_t *array)
494{
495 pr_err("RAID%d layout %x with %d disks is unsupported for DDF\n",
496 array->level, array->layout, array->raid_disks);
497 return DDF_INVALID_LEVEL;
498}
499
500static int layout_md2ddf(const mdu_array_info_t *array,
501 struct vd_config *conf)
502{
503 __u16 prim_elmnt_count = __cpu_to_be16(array->raid_disks);
504 __u8 prl = DDF_INVALID_LEVEL, rlq = 0;
505 __u8 sec_elmnt_count = 1;
506 __u8 srl = DDF_NO_SECONDARY;
507
508 switch (array->level) {
509 case LEVEL_LINEAR:
510 prl = DDF_CONCAT;
511 break;
512 case 0:
513 rlq = DDF_RAID0_SIMPLE;
514 prl = DDF_RAID0;
515 break;
516 case 1:
517 switch (array->raid_disks) {
518 case 2:
519 rlq = DDF_RAID1_SIMPLE;
520 break;
521 case 3:
522 rlq = DDF_RAID1_MULTI;
523 break;
524 default:
525 return err_bad_md_layout(array);
526 }
527 prl = DDF_RAID1;
528 break;
529 case 4:
530 if (array->layout != 0)
531 return err_bad_md_layout(array);
532 rlq = DDF_RAID4_N;
533 prl = DDF_RAID4;
534 break;
535 case 5:
536 switch (array->layout) {
537 case ALGORITHM_LEFT_ASYMMETRIC:
538 rlq = DDF_RAID5_N_RESTART;
539 break;
540 case ALGORITHM_RIGHT_ASYMMETRIC:
541 rlq = DDF_RAID5_0_RESTART;
542 break;
543 case ALGORITHM_LEFT_SYMMETRIC:
544 rlq = DDF_RAID5_N_CONTINUE;
545 break;
546 case ALGORITHM_RIGHT_SYMMETRIC:
547 /* not mentioned in standard */
548 default:
549 return err_bad_md_layout(array);
550 }
551 prl = DDF_RAID5;
552 break;
553 case 6:
554 switch (array->layout) {
555 case ALGORITHM_ROTATING_N_RESTART:
556 rlq = DDF_RAID5_N_RESTART;
557 break;
558 case ALGORITHM_ROTATING_ZERO_RESTART:
559 rlq = DDF_RAID6_0_RESTART;
560 break;
561 case ALGORITHM_ROTATING_N_CONTINUE:
562 rlq = DDF_RAID5_N_CONTINUE;
563 break;
564 default:
565 return err_bad_md_layout(array);
566 }
567 prl = DDF_RAID6;
568 break;
569 case 10:
570 if (array->raid_disks % 2 == 0 && array->layout == 0x102) {
571 rlq = DDF_RAID1_SIMPLE;
572 prim_elmnt_count = __cpu_to_be16(2);
573 sec_elmnt_count = array->raid_disks / 2;
574 } else if (array->raid_disks % 3 == 0
575 && array->layout == 0x103) {
576 rlq = DDF_RAID1_MULTI;
577 prim_elmnt_count = __cpu_to_be16(3);
578 sec_elmnt_count = array->raid_disks / 3;
579 } else
580 return err_bad_md_layout(array);
581 srl = DDF_2SPANNED;
582 prl = DDF_RAID1;
583 break;
584 default:
585 return err_bad_md_layout(array);
586 }
587 conf->prl = prl;
588 conf->prim_elmnt_count = prim_elmnt_count;
589 conf->rlq = rlq;
590 conf->srl = srl;
591 conf->sec_elmnt_count = sec_elmnt_count;
592 return 0;
593}
594
8a2848a7 595static int err_bad_ddf_layout(const struct vd_config *conf)
596{
597 pr_err("DDF RAID %u qualifier %u with %u disks is unsupported\n",
598 conf->prl, conf->rlq, __be16_to_cpu(conf->prim_elmnt_count));
599 return -1;
600}
601
602static int layout_ddf2md(const struct vd_config *conf,
603 mdu_array_info_t *array)
604{
605 int level = LEVEL_UNSUPPORTED;
606 int layout = 0;
607 int raiddisks = __be16_to_cpu(conf->prim_elmnt_count);
608
609 if (conf->sec_elmnt_count > 1) {
610 /* see also check_secondary() */
611 if (conf->prl != DDF_RAID1 ||
612 (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED)) {
613 pr_err("Unsupported secondary RAID level %u/%u\n",
614 conf->prl, conf->srl);
615 return -1;
616 }
617 if (raiddisks == 2 && conf->rlq == DDF_RAID1_SIMPLE)
618 layout = 0x102;
619 else if (raiddisks == 3 && conf->rlq == DDF_RAID1_MULTI)
620 layout = 0x103;
621 else
622 return err_bad_ddf_layout(conf);
623 raiddisks *= conf->sec_elmnt_count;
624 level = 10;
625 goto good;
626 }
627
628 switch (conf->prl) {
629 case DDF_CONCAT:
630 level = LEVEL_LINEAR;
631 break;
632 case DDF_RAID0:
633 if (conf->rlq != DDF_RAID0_SIMPLE)
634 return err_bad_ddf_layout(conf);
635 level = 0;
636 break;
637 case DDF_RAID1:
638 if (!((conf->rlq == DDF_RAID1_SIMPLE && raiddisks == 2) ||
639 (conf->rlq == DDF_RAID1_MULTI && raiddisks == 3)))
640 return err_bad_ddf_layout(conf);
641 level = 1;
642 break;
643 case DDF_RAID4:
644 if (conf->rlq != DDF_RAID4_N)
645 return err_bad_ddf_layout(conf);
646 level = 4;
647 break;
648 case DDF_RAID5:
649 switch (conf->rlq) {
650 case DDF_RAID5_N_RESTART:
651 layout = ALGORITHM_LEFT_ASYMMETRIC;
652 break;
653 case DDF_RAID5_0_RESTART:
654 layout = ALGORITHM_RIGHT_ASYMMETRIC;
655 break;
656 case DDF_RAID5_N_CONTINUE:
657 layout = ALGORITHM_LEFT_SYMMETRIC;
658 break;
659 default:
660 return err_bad_ddf_layout(conf);
661 }
662 level = 5;
663 break;
664 case DDF_RAID6:
665 switch (conf->rlq) {
666 case DDF_RAID5_N_RESTART:
667 layout = ALGORITHM_ROTATING_N_RESTART;
668 break;
669 case DDF_RAID6_0_RESTART:
670 layout = ALGORITHM_ROTATING_ZERO_RESTART;
671 break;
672 case DDF_RAID5_N_CONTINUE:
673 layout = ALGORITHM_ROTATING_N_CONTINUE;
674 break;
675 default:
676 return err_bad_ddf_layout(conf);
677 }
678 level = 6;
679 break;
680 default:
681 return err_bad_ddf_layout(conf);
682 };
683
684good:
685 array->level = level;
686 array->layout = layout;
687 array->raid_disks = raiddisks;
688 return 0;
689}
690
a322f70c
DW
691static int load_ddf_header(int fd, unsigned long long lba,
692 unsigned long long size,
693 int type,
694 struct ddf_header *hdr, struct ddf_header *anchor)
695{
696 /* read a ddf header (primary or secondary) from fd/lba
697 * and check that it is consistent with anchor
698 * Need to check:
699 * magic, crc, guid, rev, and LBA's header_type, and
700 * everything after header_type must be the same
701 */
702 if (lba >= size-1)
703 return 0;
704
705 if (lseek64(fd, lba<<9, 0) < 0)
706 return 0;
707
708 if (read(fd, hdr, 512) != 512)
709 return 0;
710
711 if (hdr->magic != DDF_HEADER_MAGIC)
712 return 0;
713 if (calc_crc(hdr, 512) != hdr->crc)
714 return 0;
715 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
716 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
717 anchor->primary_lba != hdr->primary_lba ||
718 anchor->secondary_lba != hdr->secondary_lba ||
719 hdr->type != type ||
720 memcmp(anchor->pad2, hdr->pad2, 512 -
721 offsetof(struct ddf_header, pad2)) != 0)
722 return 0;
723
724 /* Looks good enough to me... */
725 return 1;
726}
727
728static void *load_section(int fd, struct ddf_super *super, void *buf,
729 __u32 offset_be, __u32 len_be, int check)
730{
731 unsigned long long offset = __be32_to_cpu(offset_be);
732 unsigned long long len = __be32_to_cpu(len_be);
733 int dofree = (buf == NULL);
734
735 if (check)
736 if (len != 2 && len != 8 && len != 32
737 && len != 128 && len != 512)
738 return NULL;
739
740 if (len > 1024)
741 return NULL;
742 if (buf) {
743 /* All pre-allocated sections are a single block */
744 if (len != 1)
745 return NULL;
3d2c4fc7
DW
746 } else if (posix_memalign(&buf, 512, len<<9) != 0)
747 buf = NULL;
6416d527 748
a322f70c
DW
749 if (!buf)
750 return NULL;
751
752 if (super->active->type == 1)
753 offset += __be64_to_cpu(super->active->primary_lba);
754 else
755 offset += __be64_to_cpu(super->active->secondary_lba);
756
f21e18ca 757 if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) {
a322f70c
DW
758 if (dofree)
759 free(buf);
760 return NULL;
761 }
f21e18ca 762 if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) {
a322f70c
DW
763 if (dofree)
764 free(buf);
765 return NULL;
766 }
767 return buf;
768}
769
770static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
771{
772 unsigned long long dsize;
773
774 get_dev_size(fd, NULL, &dsize);
775
776 if (lseek64(fd, dsize-512, 0) < 0) {
777 if (devname)
e7b84f9d
N
778 pr_err("Cannot seek to anchor block on %s: %s\n",
779 devname, strerror(errno));
a322f70c
DW
780 return 1;
781 }
782 if (read(fd, &super->anchor, 512) != 512) {
783 if (devname)
e7b84f9d
N
784 pr_err("Cannot read anchor block on %s: %s\n",
785 devname, strerror(errno));
a322f70c
DW
786 return 1;
787 }
788 if (super->anchor.magic != DDF_HEADER_MAGIC) {
789 if (devname)
e7b84f9d 790 pr_err("no DDF anchor found on %s\n",
a322f70c
DW
791 devname);
792 return 2;
793 }
794 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
795 if (devname)
e7b84f9d 796 pr_err("bad CRC on anchor on %s\n",
a322f70c
DW
797 devname);
798 return 2;
799 }
59e36268
NB
800 if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
801 memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
a322f70c 802 if (devname)
e7b84f9d 803 pr_err("can only support super revision"
59e36268
NB
804 " %.8s and earlier, not %.8s on %s\n",
805 DDF_REVISION_2, super->anchor.revision,devname);
a322f70c
DW
806 return 2;
807 }
dbeb699a 808 super->active = NULL;
a322f70c
DW
809 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
810 dsize >> 9, 1,
811 &super->primary, &super->anchor) == 0) {
812 if (devname)
e7b84f9d
N
813 pr_err("Failed to load primary DDF header "
814 "on %s\n", devname);
dbeb699a 815 } else
816 super->active = &super->primary;
a322f70c
DW
817 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
818 dsize >> 9, 2,
819 &super->secondary, &super->anchor)) {
820 if ((__be32_to_cpu(super->primary.seq)
821 < __be32_to_cpu(super->secondary.seq) &&
822 !super->secondary.openflag)
823 || (__be32_to_cpu(super->primary.seq)
824 == __be32_to_cpu(super->secondary.seq) &&
825 super->primary.openflag && !super->secondary.openflag)
dbeb699a 826 || super->active == NULL
a322f70c
DW
827 )
828 super->active = &super->secondary;
dbeb699a 829 } else if (devname)
830 pr_err("Failed to load secondary DDF header on %s\n",
831 devname);
832 if (super->active == NULL)
833 return 2;
a322f70c
DW
834 return 0;
835}
836
837static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
838{
839 void *ok;
840 ok = load_section(fd, super, &super->controller,
841 super->active->controller_section_offset,
842 super->active->controller_section_length,
843 0);
844 super->phys = load_section(fd, super, NULL,
845 super->active->phys_section_offset,
846 super->active->phys_section_length,
847 1);
848 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
849
850 super->virt = load_section(fd, super, NULL,
851 super->active->virt_section_offset,
852 super->active->virt_section_length,
853 1);
854 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
855 if (!ok ||
856 !super->phys ||
857 !super->virt) {
858 free(super->phys);
859 free(super->virt);
a2349791
NB
860 super->phys = NULL;
861 super->virt = NULL;
a322f70c
DW
862 return 2;
863 }
864 super->conflist = NULL;
865 super->dlist = NULL;
8c3b8c2c
NB
866
867 super->max_part = __be16_to_cpu(super->active->max_partitions);
868 super->mppe = __be16_to_cpu(super->active->max_primary_element_entries);
869 super->conf_rec_len = __be16_to_cpu(super->active->config_record_len);
a322f70c
DW
870 return 0;
871}
872
3c48f7be 873#define DDF_UNUSED_BVD 0xff
874static int alloc_other_bvds(const struct ddf_super *ddf, struct vcl *vcl)
875{
876 unsigned int n_vds = vcl->conf.sec_elmnt_count - 1;
877 unsigned int i, vdsize;
878 void *p;
879 if (n_vds == 0) {
880 vcl->other_bvds = NULL;
881 return 0;
882 }
883 vdsize = ddf->conf_rec_len * 512;
884 if (posix_memalign(&p, 512, n_vds *
885 (vdsize + sizeof(struct vd_config *))) != 0)
886 return -1;
887 vcl->other_bvds = (struct vd_config **) (p + n_vds * vdsize);
888 for (i = 0; i < n_vds; i++) {
889 vcl->other_bvds[i] = p + i * vdsize;
890 memset(vcl->other_bvds[i], 0, vdsize);
891 vcl->other_bvds[i]->sec_elmnt_seq = DDF_UNUSED_BVD;
892 }
893 return 0;
894}
895
3dc821b0 896static void add_other_bvd(struct vcl *vcl, struct vd_config *vd,
897 unsigned int len)
898{
899 int i;
900 for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++)
3c48f7be 901 if (vcl->other_bvds[i]->sec_elmnt_seq == vd->sec_elmnt_seq)
3dc821b0 902 break;
903
904 if (i < vcl->conf.sec_elmnt_count-1) {
905 if (vd->seqnum <= vcl->other_bvds[i]->seqnum)
906 return;
907 } else {
908 for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++)
3c48f7be 909 if (vcl->other_bvds[i]->sec_elmnt_seq == DDF_UNUSED_BVD)
3dc821b0 910 break;
911 if (i == vcl->conf.sec_elmnt_count-1) {
912 pr_err("no space for sec level config %u, count is %u\n",
913 vd->sec_elmnt_seq, vcl->conf.sec_elmnt_count);
914 return;
915 }
3dc821b0 916 }
917 memcpy(vcl->other_bvds[i], vd, len);
918}
919
a322f70c
DW
920static int load_ddf_local(int fd, struct ddf_super *super,
921 char *devname, int keep)
922{
923 struct dl *dl;
924 struct stat stb;
925 char *conf;
f21e18ca
N
926 unsigned int i;
927 unsigned int confsec;
b2280677 928 int vnum;
f21e18ca 929 unsigned int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
d2ca6449 930 unsigned long long dsize;
a322f70c
DW
931
932 /* First the local disk info */
3d2c4fc7 933 if (posix_memalign((void**)&dl, 512,
6416d527 934 sizeof(*dl) +
3d2c4fc7 935 (super->max_part) * sizeof(dl->vlist[0])) != 0) {
e7b84f9d 936 pr_err("%s could not allocate disk info buffer\n",
3d2c4fc7
DW
937 __func__);
938 return 1;
939 }
a322f70c
DW
940
941 load_section(fd, super, &dl->disk,
942 super->active->data_section_offset,
943 super->active->data_section_length,
944 0);
503975b9 945 dl->devname = devname ? xstrdup(devname) : NULL;
598f0d58 946
a322f70c
DW
947 fstat(fd, &stb);
948 dl->major = major(stb.st_rdev);
949 dl->minor = minor(stb.st_rdev);
950 dl->next = super->dlist;
951 dl->fd = keep ? fd : -1;
d2ca6449
NB
952
953 dl->size = 0;
954 if (get_dev_size(fd, devname, &dsize))
955 dl->size = dsize >> 9;
097bcf00 956 /* If the disks have different sizes, the LBAs will differ
957 * between phys disks.
958 * At this point here, the values in super->active must be valid
959 * for this phys disk. */
960 dl->primary_lba = super->active->primary_lba;
961 dl->secondary_lba = super->active->secondary_lba;
962 dl->workspace_lba = super->active->workspace_lba;
b2280677 963 dl->spare = NULL;
f21e18ca 964 for (i = 0 ; i < super->max_part ; i++)
a322f70c
DW
965 dl->vlist[i] = NULL;
966 super->dlist = dl;
59e36268 967 dl->pdnum = -1;
f21e18ca 968 for (i = 0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
5575e7d9
NB
969 if (memcmp(super->phys->entries[i].guid,
970 dl->disk.guid, DDF_GUID_LEN) == 0)
971 dl->pdnum = i;
972
a322f70c
DW
973 /* Now the config list. */
974 /* 'conf' is an array of config entries, some of which are
975 * probably invalid. Those which are good need to be copied into
976 * the conflist
977 */
a322f70c
DW
978
979 conf = load_section(fd, super, NULL,
980 super->active->config_section_offset,
981 super->active->config_section_length,
982 0);
983
b2280677 984 vnum = 0;
e223334f
N
985 for (confsec = 0;
986 confsec < __be32_to_cpu(super->active->config_section_length);
987 confsec += super->conf_rec_len) {
a322f70c 988 struct vd_config *vd =
e223334f 989 (struct vd_config *)((char*)conf + confsec*512);
a322f70c
DW
990 struct vcl *vcl;
991
b2280677
NB
992 if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
993 if (dl->spare)
994 continue;
3d2c4fc7
DW
995 if (posix_memalign((void**)&dl->spare, 512,
996 super->conf_rec_len*512) != 0) {
e7b84f9d
N
997 pr_err("%s could not allocate spare info buf\n",
998 __func__);
3d2c4fc7
DW
999 return 1;
1000 }
613b0d17 1001
b2280677
NB
1002 memcpy(dl->spare, vd, super->conf_rec_len*512);
1003 continue;
1004 }
a322f70c
DW
1005 if (vd->magic != DDF_VD_CONF_MAGIC)
1006 continue;
1007 for (vcl = super->conflist; vcl; vcl = vcl->next) {
1008 if (memcmp(vcl->conf.guid,
1009 vd->guid, DDF_GUID_LEN) == 0)
1010 break;
1011 }
1012
1013 if (vcl) {
b2280677 1014 dl->vlist[vnum++] = vcl;
3dc821b0 1015 if (vcl->other_bvds != NULL &&
1016 vcl->conf.sec_elmnt_seq != vd->sec_elmnt_seq) {
1017 add_other_bvd(vcl, vd, super->conf_rec_len*512);
1018 continue;
1019 }
a322f70c
DW
1020 if (__be32_to_cpu(vd->seqnum) <=
1021 __be32_to_cpu(vcl->conf.seqnum))
1022 continue;
59e36268 1023 } else {
3d2c4fc7 1024 if (posix_memalign((void**)&vcl, 512,
6416d527 1025 (super->conf_rec_len*512 +
3d2c4fc7 1026 offsetof(struct vcl, conf))) != 0) {
e7b84f9d
N
1027 pr_err("%s could not allocate vcl buf\n",
1028 __func__);
3d2c4fc7
DW
1029 return 1;
1030 }
a322f70c 1031 vcl->next = super->conflist;
59e36268 1032 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
3c48f7be 1033 vcl->conf.sec_elmnt_count = vd->sec_elmnt_count;
1034 if (alloc_other_bvds(super, vcl) != 0) {
1035 pr_err("%s could not allocate other bvds\n",
1036 __func__);
1037 free(vcl);
1038 return 1;
1039 };
a322f70c 1040 super->conflist = vcl;
b2280677 1041 dl->vlist[vnum++] = vcl;
a322f70c 1042 }
8c3b8c2c 1043 memcpy(&vcl->conf, vd, super->conf_rec_len*512);
59e36268
NB
1044 for (i=0; i < max_virt_disks ; i++)
1045 if (memcmp(super->virt->entries[i].guid,
1046 vcl->conf.guid, DDF_GUID_LEN)==0)
1047 break;
1048 if (i < max_virt_disks)
1049 vcl->vcnum = i;
a322f70c
DW
1050 }
1051 free(conf);
1052
1053 return 0;
1054}
1055
1056#ifndef MDASSEMBLE
1057static int load_super_ddf_all(struct supertype *st, int fd,
e1902a7b 1058 void **sbp, char *devname);
a322f70c 1059#endif
37424f13
DW
1060
1061static void free_super_ddf(struct supertype *st);
1062
a322f70c
DW
1063static int load_super_ddf(struct supertype *st, int fd,
1064 char *devname)
1065{
1066 unsigned long long dsize;
1067 struct ddf_super *super;
1068 int rv;
1069
a322f70c
DW
1070 if (get_dev_size(fd, devname, &dsize) == 0)
1071 return 1;
1072
b31df436 1073 if (!st->ignore_hw_compat && test_partition(fd))
691c6ee1
N
1074 /* DDF is not allowed on partitions */
1075 return 1;
1076
a322f70c
DW
1077 /* 32M is a lower bound */
1078 if (dsize <= 32*1024*1024) {
97320d7c 1079 if (devname)
e7b84f9d
N
1080 pr_err("%s is too small for ddf: "
1081 "size is %llu sectors.\n",
1082 devname, dsize>>9);
97320d7c 1083 return 1;
a322f70c
DW
1084 }
1085 if (dsize & 511) {
97320d7c 1086 if (devname)
e7b84f9d
N
1087 pr_err("%s is an odd size for ddf: "
1088 "size is %llu bytes.\n",
1089 devname, dsize);
97320d7c 1090 return 1;
a322f70c
DW
1091 }
1092
37424f13
DW
1093 free_super_ddf(st);
1094
6416d527 1095 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
e7b84f9d 1096 pr_err("malloc of %zu failed.\n",
a322f70c
DW
1097 sizeof(*super));
1098 return 1;
1099 }
a2349791 1100 memset(super, 0, sizeof(*super));
a322f70c
DW
1101
1102 rv = load_ddf_headers(fd, super, devname);
1103 if (rv) {
1104 free(super);
1105 return rv;
1106 }
1107
1108 /* Have valid headers and have chosen the best. Let's read in the rest*/
1109
1110 rv = load_ddf_global(fd, super, devname);
1111
1112 if (rv) {
1113 if (devname)
e7b84f9d
N
1114 pr_err("Failed to load all information "
1115 "sections on %s\n", devname);
a322f70c
DW
1116 free(super);
1117 return rv;
1118 }
1119
3d2c4fc7
DW
1120 rv = load_ddf_local(fd, super, devname, 0);
1121
1122 if (rv) {
1123 if (devname)
e7b84f9d
N
1124 pr_err("Failed to load all information "
1125 "sections on %s\n", devname);
3d2c4fc7
DW
1126 free(super);
1127 return rv;
1128 }
a322f70c
DW
1129
1130 /* Should possibly check the sections .... */
1131
1132 st->sb = super;
1133 if (st->ss == NULL) {
1134 st->ss = &super_ddf;
1135 st->minor_version = 0;
1136 st->max_devs = 512;
1137 }
1138 return 0;
1139
1140}
1141
1142static void free_super_ddf(struct supertype *st)
1143{
1144 struct ddf_super *ddf = st->sb;
1145 if (ddf == NULL)
1146 return;
1147 free(ddf->phys);
1148 free(ddf->virt);
1149 while (ddf->conflist) {
1150 struct vcl *v = ddf->conflist;
1151 ddf->conflist = v->next;
59e36268
NB
1152 if (v->block_sizes)
1153 free(v->block_sizes);
3c48f7be 1154 if (v->other_bvds)
1155 /*
1156 v->other_bvds[0] points to beginning of buffer,
1157 see alloc_other_bvds()
1158 */
1159 free(v->other_bvds[0]);
a322f70c
DW
1160 free(v);
1161 }
1162 while (ddf->dlist) {
1163 struct dl *d = ddf->dlist;
1164 ddf->dlist = d->next;
1165 if (d->fd >= 0)
1166 close(d->fd);
b2280677
NB
1167 if (d->spare)
1168 free(d->spare);
a322f70c
DW
1169 free(d);
1170 }
8a38cb04
N
1171 while (ddf->add_list) {
1172 struct dl *d = ddf->add_list;
1173 ddf->add_list = d->next;
1174 if (d->fd >= 0)
1175 close(d->fd);
1176 if (d->spare)
1177 free(d->spare);
1178 free(d);
1179 }
a322f70c
DW
1180 free(ddf);
1181 st->sb = NULL;
1182}
1183
1184static struct supertype *match_metadata_desc_ddf(char *arg)
1185{
1186 /* 'ddf' only support containers */
1187 struct supertype *st;
1188 if (strcmp(arg, "ddf") != 0 &&
1189 strcmp(arg, "default") != 0
1190 )
1191 return NULL;
1192
503975b9 1193 st = xcalloc(1, sizeof(*st));
a322f70c
DW
1194 st->ss = &super_ddf;
1195 st->max_devs = 512;
1196 st->minor_version = 0;
1197 st->sb = NULL;
1198 return st;
1199}
1200
a322f70c
DW
1201#ifndef MDASSEMBLE
1202
1203static mapping_t ddf_state[] = {
1204 { "Optimal", 0},
1205 { "Degraded", 1},
1206 { "Deleted", 2},
1207 { "Missing", 3},
1208 { "Failed", 4},
1209 { "Partially Optimal", 5},
1210 { "-reserved-", 6},
1211 { "-reserved-", 7},
1212 { NULL, 0}
1213};
1214
1215static mapping_t ddf_init_state[] = {
1216 { "Not Initialised", 0},
1217 { "QuickInit in Progress", 1},
1218 { "Fully Initialised", 2},
1219 { "*UNKNOWN*", 3},
1220 { NULL, 0}
1221};
1222static mapping_t ddf_access[] = {
1223 { "Read/Write", 0},
1224 { "Reserved", 1},
1225 { "Read Only", 2},
1226 { "Blocked (no access)", 3},
1227 { NULL ,0}
1228};
1229
1230static mapping_t ddf_level[] = {
1231 { "RAID0", DDF_RAID0},
1232 { "RAID1", DDF_RAID1},
1233 { "RAID3", DDF_RAID3},
1234 { "RAID4", DDF_RAID4},
1235 { "RAID5", DDF_RAID5},
1236 { "RAID1E",DDF_RAID1E},
1237 { "JBOD", DDF_JBOD},
1238 { "CONCAT",DDF_CONCAT},
1239 { "RAID5E",DDF_RAID5E},
1240 { "RAID5EE",DDF_RAID5EE},
1241 { "RAID6", DDF_RAID6},
1242 { NULL, 0}
1243};
1244static mapping_t ddf_sec_level[] = {
1245 { "Striped", DDF_2STRIPED},
1246 { "Mirrored", DDF_2MIRRORED},
1247 { "Concat", DDF_2CONCAT},
1248 { "Spanned", DDF_2SPANNED},
1249 { NULL, 0}
1250};
1251#endif
1252
fb9d0acb 1253static int all_ff(const char *guid)
42dc2744
N
1254{
1255 int i;
1256 for (i = 0; i < DDF_GUID_LEN; i++)
1257 if (guid[i] != (char)0xff)
1258 return 0;
1259 return 1;
1260}
1261
a322f70c
DW
1262#ifndef MDASSEMBLE
1263static void print_guid(char *guid, int tstamp)
1264{
1265 /* A GUIDs are part (or all) ASCII and part binary.
1266 * They tend to be space padded.
59e36268
NB
1267 * We print the GUID in HEX, then in parentheses add
1268 * any initial ASCII sequence, and a possible
1269 * time stamp from bytes 16-19
a322f70c
DW
1270 */
1271 int l = DDF_GUID_LEN;
1272 int i;
59e36268
NB
1273
1274 for (i=0 ; i<DDF_GUID_LEN ; i++) {
1275 if ((i&3)==0 && i != 0) printf(":");
1276 printf("%02X", guid[i]&255);
1277 }
1278
cfccea8c 1279 printf("\n (");
a322f70c
DW
1280 while (l && guid[l-1] == ' ')
1281 l--;
1282 for (i=0 ; i<l ; i++) {
1283 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1284 fputc(guid[i], stdout);
1285 else
59e36268 1286 break;
a322f70c
DW
1287 }
1288 if (tstamp) {
1289 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1290 char tbuf[100];
1291 struct tm *tm;
1292 tm = localtime(&then);
59e36268 1293 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW
1294 fputs(tbuf, stdout);
1295 }
59e36268 1296 printf(")");
a322f70c
DW
1297}
1298
1299static void examine_vd(int n, struct ddf_super *sb, char *guid)
1300{
8c3b8c2c 1301 int crl = sb->conf_rec_len;
a322f70c
DW
1302 struct vcl *vcl;
1303
1304 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
f21e18ca 1305 unsigned int i;
a322f70c
DW
1306 struct vd_config *vc = &vcl->conf;
1307
1308 if (calc_crc(vc, crl*512) != vc->crc)
1309 continue;
1310 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1311 continue;
1312
1313 /* Ok, we know about this VD, let's give more details */
b06e3095 1314 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1315 __be16_to_cpu(vc->prim_elmnt_count));
f21e18ca 1316 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) {
b06e3095
N
1317 int j;
1318 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1319 for (j=0; j<cnt; j++)
1320 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1321 break;
1322 if (i) printf(" ");
1323 if (j < cnt)
1324 printf("%d", j);
1325 else
1326 printf("--");
1327 }
1328 printf(")\n");
1329 if (vc->chunk_shift != 255)
613b0d17
N
1330 printf(" Chunk Size[%d] : %d sectors\n", n,
1331 1 << vc->chunk_shift);
a322f70c
DW
1332 printf(" Raid Level[%d] : %s\n", n,
1333 map_num(ddf_level, vc->prl)?:"-unknown-");
1334 if (vc->sec_elmnt_count != 1) {
1335 printf(" Secondary Position[%d] : %d of %d\n", n,
1336 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1337 printf(" Secondary Level[%d] : %s\n", n,
1338 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1339 }
1340 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1341 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1342 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1343 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW
1344 }
1345}
1346
1347static void examine_vds(struct ddf_super *sb)
1348{
1349 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
fb9d0acb 1350 unsigned int i;
a322f70c
DW
1351 printf(" Virtual Disks : %d\n", cnt);
1352
fb9d0acb 1353 for (i = 0; i < __be16_to_cpu(sb->virt->max_vdes); i++) {
a322f70c 1354 struct virtual_entry *ve = &sb->virt->entries[i];
fb9d0acb 1355 if (all_ff(ve->guid))
1356 continue;
b06e3095 1357 printf("\n");
a322f70c
DW
1358 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1359 printf("\n");
1360 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1361 printf(" state[%d] : %s, %s%s\n", i,
1362 map_num(ddf_state, ve->state & 7),
1363 (ve->state & 8) ? "Morphing, ": "",
1364 (ve->state & 16)? "Not Consistent" : "Consistent");
1365 printf(" init state[%d] : %s\n", i,
1366 map_num(ddf_init_state, ve->init_state&3));
1367 printf(" access[%d] : %s\n", i,
1368 map_num(ddf_access, (ve->init_state>>6) & 3));
1369 printf(" Name[%d] : %.16s\n", i, ve->name);
1370 examine_vd(i, sb, ve->guid);
1371 }
1372 if (cnt) printf("\n");
1373}
1374
1375static void examine_pds(struct ddf_super *sb)
1376{
1377 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1378 int i;
1379 struct dl *dl;
1380 printf(" Physical Disks : %d\n", cnt);
962371a5 1381 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW
1382
1383 for (i=0 ; i<cnt ; i++) {
1384 struct phys_disk_entry *pd = &sb->phys->entries[i];
1385 int type = __be16_to_cpu(pd->type);
1386 int state = __be16_to_cpu(pd->state);
1387
b06e3095
N
1388 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1389 //printf("\n");
1390 printf(" %3d %08x ", i,
a322f70c 1391 __be32_to_cpu(pd->refnum));
613b0d17 1392 printf("%8lluK ",
c9b6907b 1393 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N
1394 for (dl = sb->dlist; dl ; dl = dl->next) {
1395 if (dl->disk.refnum == pd->refnum) {
1396 char *dv = map_dev(dl->major, dl->minor, 0);
1397 if (dv) {
962371a5 1398 printf("%-15s", dv);
b06e3095
N
1399 break;
1400 }
1401 }
1402 }
1403 if (!dl)
962371a5 1404 printf("%15s","");
b06e3095 1405 printf(" %s%s%s%s%s",
a322f70c 1406 (type&2) ? "active":"",
b06e3095 1407 (type&4) ? "Global-Spare":"",
a322f70c
DW
1408 (type&8) ? "spare" : "",
1409 (type&16)? ", foreign" : "",
1410 (type&32)? "pass-through" : "");
18cb4496
N
1411 if (state & DDF_Failed)
1412 /* This over-rides these three */
1413 state &= ~(DDF_Online|DDF_Rebuilding|DDF_Transition);
b06e3095 1414 printf("/%s%s%s%s%s%s%s",
a322f70c
DW
1415 (state&1)? "Online": "Offline",
1416 (state&2)? ", Failed": "",
1417 (state&4)? ", Rebuilding": "",
1418 (state&8)? ", in-transition": "",
b06e3095
N
1419 (state&16)? ", SMART-errors": "",
1420 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1421 (state&64)? ", Missing" : "");
a322f70c
DW
1422 printf("\n");
1423 }
1424}
1425
1426static void examine_super_ddf(struct supertype *st, char *homehost)
1427{
1428 struct ddf_super *sb = st->sb;
1429
1430 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1431 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB
1432 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1433 printf("\n");
1434 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW
1435 printf("\n");
1436 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1437 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1438 ?"yes" : "no");
1439 examine_vds(sb);
1440 examine_pds(sb);
1441}
1442
a5d85af7 1443static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map);
ff54de6e 1444
bedbf68a 1445static void uuid_from_ddf_guid(const char *guid, int uuid[4]);
42dc2744 1446static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1447
bedbf68a 1448static unsigned int get_vd_num_of_subarray(struct supertype *st)
1449{
1450 /*
1451 * Figure out the VD number for this supertype.
1452 * Returns DDF_CONTAINER for the container itself,
1453 * and DDF_NOTFOUND on error.
1454 */
1455 struct ddf_super *ddf = st->sb;
1456 struct mdinfo *sra;
1457 char *sub, *end;
1458 unsigned int vcnum;
1459
1460 if (*st->container_devnm == '\0')
1461 return DDF_CONTAINER;
1462
1463 sra = sysfs_read(-1, st->devnm, GET_VERSION);
1464 if (!sra || sra->array.major_version != -1 ||
1465 sra->array.minor_version != -2 ||
1466 !is_subarray(sra->text_version))
1467 return DDF_NOTFOUND;
1468
1469 sub = strchr(sra->text_version + 1, '/');
1470 if (sub != NULL)
1471 vcnum = strtoul(sub + 1, &end, 10);
1472 if (sub == NULL || *sub == '\0' || *end != '\0' ||
1473 vcnum >= __be16_to_cpu(ddf->active->max_vd_entries))
1474 return DDF_NOTFOUND;
1475
1476 return vcnum;
1477}
1478
061f2c6a 1479static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N
1480{
1481 /* We just write a generic DDF ARRAY entry
1482 */
1483 struct mdinfo info;
1484 char nbuf[64];
a5d85af7 1485 getinfo_super_ddf(st, &info, NULL);
4737ae25
N
1486 fname_from_uuid(st, &info, nbuf, ':');
1487
1488 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1489}
1490
1491static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW
1492{
1493 /* We just write a generic DDF ARRAY entry
a322f70c 1494 */
42dc2744 1495 struct ddf_super *ddf = st->sb;
ff54de6e 1496 struct mdinfo info;
f21e18ca 1497 unsigned int i;
ff54de6e 1498 char nbuf[64];
a5d85af7 1499 getinfo_super_ddf(st, &info, NULL);
ff54de6e 1500 fname_from_uuid(st, &info, nbuf, ':');
42dc2744 1501
f21e18ca 1502 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
42dc2744
N
1503 struct virtual_entry *ve = &ddf->virt->entries[i];
1504 struct vcl vcl;
1505 char nbuf1[64];
1506 if (all_ff(ve->guid))
1507 continue;
1508 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1509 ddf->currentconf =&vcl;
1510 uuid_from_super_ddf(st, info.uuid);
1511 fname_from_uuid(st, &info, nbuf1, ':');
1512 printf("ARRAY container=%s member=%d UUID=%s\n",
1513 nbuf+5, i, nbuf1+5);
1514 }
a322f70c
DW
1515}
1516
bceedeec
N
1517static void export_examine_super_ddf(struct supertype *st)
1518{
1519 struct mdinfo info;
1520 char nbuf[64];
a5d85af7 1521 getinfo_super_ddf(st, &info, NULL);
bceedeec
N
1522 fname_from_uuid(st, &info, nbuf, ':');
1523 printf("MD_METADATA=ddf\n");
1524 printf("MD_LEVEL=container\n");
1525 printf("MD_UUID=%s\n", nbuf+5);
1526}
bceedeec 1527
74db60b0
N
1528static int copy_metadata_ddf(struct supertype *st, int from, int to)
1529{
1530 void *buf;
1531 unsigned long long dsize, offset;
1532 int bytes;
1533 struct ddf_header *ddf;
1534 int written = 0;
1535
1536 /* The meta consists of an anchor, a primary, and a secondary.
1537 * This all lives at the end of the device.
1538 * So it is easiest to find the earliest of primary and
1539 * secondary, and copy everything from there.
1540 *
1541 * Anchor is 512 from end It contains primary_lba and secondary_lba
1542 * we choose one of those
1543 */
1544
1545 if (posix_memalign(&buf, 4096, 4096) != 0)
1546 return 1;
1547
1548 if (!get_dev_size(from, NULL, &dsize))
1549 goto err;
1550
1551 if (lseek64(from, dsize-512, 0) < 0)
1552 goto err;
1553 if (read(from, buf, 512) != 512)
1554 goto err;
1555 ddf = buf;
1556 if (ddf->magic != DDF_HEADER_MAGIC ||
1557 calc_crc(ddf, 512) != ddf->crc ||
1558 (memcmp(ddf->revision, DDF_REVISION_0, 8) != 0 &&
1559 memcmp(ddf->revision, DDF_REVISION_2, 8) != 0))
1560 goto err;
1561
1562 offset = dsize - 512;
1563 if ((__be64_to_cpu(ddf->primary_lba) << 9) < offset)
1564 offset = __be64_to_cpu(ddf->primary_lba) << 9;
1565 if ((__be64_to_cpu(ddf->secondary_lba) << 9) < offset)
1566 offset = __be64_to_cpu(ddf->secondary_lba) << 9;
1567
1568 bytes = dsize - offset;
1569
1570 if (lseek64(from, offset, 0) < 0 ||
1571 lseek64(to, offset, 0) < 0)
1572 goto err;
1573 while (written < bytes) {
1574 int n = bytes - written;
1575 if (n > 4096)
1576 n = 4096;
1577 if (read(from, buf, n) != n)
1578 goto err;
1579 if (write(to, buf, n) != n)
1580 goto err;
1581 written += n;
1582 }
1583 free(buf);
1584 return 0;
1585err:
1586 free(buf);
1587 return 1;
1588}
1589
a322f70c
DW
1590static void detail_super_ddf(struct supertype *st, char *homehost)
1591{
1592 /* FIXME later
1593 * Could print DDF GUID
1594 * Need to find which array
1595 * If whole, briefly list all arrays
1596 * If one, give name
1597 */
1598}
1599
1600static void brief_detail_super_ddf(struct supertype *st)
1601{
ff54de6e
N
1602 struct mdinfo info;
1603 char nbuf[64];
bedbf68a 1604 struct ddf_super *ddf = st->sb;
1605 unsigned int vcnum = get_vd_num_of_subarray(st);
1606 if (vcnum == DDF_CONTAINER)
1607 uuid_from_super_ddf(st, info.uuid);
1608 else if (vcnum == DDF_NOTFOUND)
1609 return;
1610 else
1611 uuid_from_ddf_guid(ddf->virt->entries[vcnum].guid, info.uuid);
ff54de6e
N
1612 fname_from_uuid(st, &info, nbuf,':');
1613 printf(" UUID=%s", nbuf + 5);
a322f70c 1614}
a322f70c
DW
1615#endif
1616
1617static int match_home_ddf(struct supertype *st, char *homehost)
1618{
1619 /* It matches 'this' host if the controller is a
1620 * Linux-MD controller with vendor_data matching
1621 * the hostname
1622 */
1623 struct ddf_super *ddf = st->sb;
f21e18ca 1624 unsigned int len;
d1d3482b
N
1625
1626 if (!homehost)
1627 return 0;
1628 len = strlen(homehost);
a322f70c
DW
1629
1630 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1631 len < sizeof(ddf->controller.vendor_data) &&
1632 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1633 ddf->controller.vendor_data[len] == 0);
1634}
1635
0e600426 1636#ifndef MDASSEMBLE
baba3f4e 1637static int find_index_in_bvd(const struct ddf_super *ddf,
1638 const struct vd_config *conf, unsigned int n,
1639 unsigned int *n_bvd)
1640{
1641 /*
1642 * Find the index of the n-th valid physical disk in this BVD
1643 */
1644 unsigned int i, j;
1645 for (i = 0, j = 0; i < ddf->mppe &&
1646 j < __be16_to_cpu(conf->prim_elmnt_count); i++) {
1647 if (conf->phys_refnum[i] != 0xffffffff) {
1648 if (n == j) {
1649 *n_bvd = i;
1650 return 1;
1651 }
1652 j++;
1653 }
1654 }
1655 dprintf("%s: couldn't find BVD member %u (total %u)\n",
1656 __func__, n, __be16_to_cpu(conf->prim_elmnt_count));
1657 return 0;
1658}
1659
1660static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst,
1661 unsigned int n,
1662 unsigned int *n_bvd, struct vcl **vcl)
a322f70c 1663{
7a7cc504 1664 struct vcl *v;
59e36268 1665
baba3f4e 1666 for (v = ddf->conflist; v; v = v->next) {
1667 unsigned int nsec, ibvd;
1668 struct vd_config *conf;
1669 if (inst != v->vcnum)
1670 continue;
1671 conf = &v->conf;
1672 if (conf->sec_elmnt_count == 1) {
1673 if (find_index_in_bvd(ddf, conf, n, n_bvd)) {
1674 *vcl = v;
1675 return conf;
1676 } else
1677 goto bad;
1678 }
1679 if (v->other_bvds == NULL) {
1680 pr_err("%s: BUG: other_bvds is NULL, nsec=%u\n",
1681 __func__, conf->sec_elmnt_count);
1682 goto bad;
1683 }
1684 nsec = n / __be16_to_cpu(conf->prim_elmnt_count);
1685 if (conf->sec_elmnt_seq != nsec) {
1686 for (ibvd = 1; ibvd < conf->sec_elmnt_count; ibvd++) {
baba3f4e 1687 if (v->other_bvds[ibvd-1]->sec_elmnt_seq
1688 == nsec)
1689 break;
1690 }
1691 if (ibvd == conf->sec_elmnt_count)
1692 goto bad;
1693 conf = v->other_bvds[ibvd-1];
1694 }
1695 if (!find_index_in_bvd(ddf, conf,
1696 n - nsec*conf->sec_elmnt_count, n_bvd))
1697 goto bad;
1698 dprintf("%s: found disk %u as member %u in bvd %d of array %u\n"
1699 , __func__, n, *n_bvd, ibvd-1, inst);
1700 *vcl = v;
1701 return conf;
1702 }
1703bad:
1704 pr_err("%s: Could't find disk %d in array %u\n", __func__, n, inst);
7a7cc504
NB
1705 return NULL;
1706}
0e600426 1707#endif
7a7cc504 1708
5ec636b7 1709static int find_phys(const struct ddf_super *ddf, __u32 phys_refnum)
7a7cc504
NB
1710{
1711 /* Find the entry in phys_disk which has the given refnum
1712 * and return it's index
1713 */
f21e18ca
N
1714 unsigned int i;
1715 for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
7a7cc504
NB
1716 if (ddf->phys->entries[i].refnum == phys_refnum)
1717 return i;
1718 return -1;
a322f70c
DW
1719}
1720
bedbf68a 1721static void uuid_from_ddf_guid(const char *guid, int uuid[4])
1722{
1723 char buf[20];
1724 struct sha1_ctx ctx;
1725 sha1_init_ctx(&ctx);
1726 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
1727 sha1_finish_ctx(&ctx, buf);
1728 memcpy(uuid, buf, 4*4);
1729}
1730
a322f70c
DW
1731static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1732{
1733 /* The uuid returned here is used for:
1734 * uuid to put into bitmap file (Create, Grow)
1735 * uuid for backup header when saving critical section (Grow)
1736 * comparing uuids when re-adding a device into an array
51006d85
N
1737 * In these cases the uuid required is that of the data-array,
1738 * not the device-set.
1739 * uuid to recognise same set when adding a missing device back
1740 * to an array. This is a uuid for the device-set.
613b0d17 1741 *
a322f70c
DW
1742 * For each of these we can make do with a truncated
1743 * or hashed uuid rather than the original, as long as
1744 * everyone agrees.
a322f70c
DW
1745 * In the case of SVD we assume the BVD is of interest,
1746 * though that might be the case if a bitmap were made for
1747 * a mirrored SVD - worry about that later.
1748 * So we need to find the VD configuration record for the
1749 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1750 * The first 16 bytes of the sha1 of these is used.
1751 */
1752 struct ddf_super *ddf = st->sb;
d2ca6449 1753 struct vcl *vcl = ddf->currentconf;
c5afc314 1754 char *guid;
a322f70c 1755
c5afc314
N
1756 if (vcl)
1757 guid = vcl->conf.guid;
1758 else
1759 guid = ddf->anchor.guid;
bedbf68a 1760 uuid_from_ddf_guid(guid, uuid);
a322f70c
DW
1761}
1762
a5d85af7 1763static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map);
78e44928 1764
a5d85af7 1765static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map)
a322f70c
DW
1766{
1767 struct ddf_super *ddf = st->sb;
a5d85af7 1768 int map_disks = info->array.raid_disks;
90fa1a29 1769 __u32 *cptr;
a322f70c 1770
78e44928 1771 if (ddf->currentconf) {
a5d85af7 1772 getinfo_super_ddf_bvd(st, info, map);
78e44928
NB
1773 return;
1774 }
95eeceeb 1775 memset(info, 0, sizeof(*info));
78e44928 1776
a322f70c
DW
1777 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1778 info->array.level = LEVEL_CONTAINER;
1779 info->array.layout = 0;
1780 info->array.md_minor = -1;
90fa1a29
JS
1781 cptr = (__u32 *)(ddf->anchor.guid + 16);
1782 info->array.ctime = DECADE + __be32_to_cpu(*cptr);
1783
a322f70c
DW
1784 info->array.utime = 0;
1785 info->array.chunk_size = 0;
510242aa 1786 info->container_enough = 1;
a322f70c 1787
a322f70c
DW
1788 info->disk.major = 0;
1789 info->disk.minor = 0;
cba0191b
NB
1790 if (ddf->dlist) {
1791 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1792 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB
1793
1794 info->data_offset = __be64_to_cpu(ddf->phys->
613b0d17
N
1795 entries[info->disk.raid_disk].
1796 config_size);
d2ca6449 1797 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB
1798 } else {
1799 info->disk.number = -1;
661dce36 1800 info->disk.raid_disk = -1;
cba0191b
NB
1801// info->disk.raid_disk = find refnum in the table and use index;
1802 }
f22385f9 1803 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1804
921d9e16 1805 info->recovery_start = MaxSector;
a19c88b8 1806 info->reshape_active = 0;
6e75048b 1807 info->recovery_blocked = 0;
c5afc314 1808 info->name[0] = 0;
a322f70c 1809
f35f2525
N
1810 info->array.major_version = -1;
1811 info->array.minor_version = -2;
159c3a1a 1812 strcpy(info->text_version, "ddf");
a67dd8cc 1813 info->safe_mode_delay = 0;
159c3a1a 1814
c5afc314 1815 uuid_from_super_ddf(st, info->uuid);
a322f70c 1816
a5d85af7
N
1817 if (map) {
1818 int i;
1819 for (i = 0 ; i < map_disks; i++) {
1820 if (i < info->array.raid_disks &&
1821 (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) &&
1822 !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed))
1823 map[i] = 1;
1824 else
1825 map[i] = 0;
1826 }
1827 }
a322f70c
DW
1828}
1829
a5d85af7 1830static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map)
a322f70c
DW
1831{
1832 struct ddf_super *ddf = st->sb;
d2ca6449
NB
1833 struct vcl *vc = ddf->currentconf;
1834 int cd = ddf->currentdev;
db42fa9b 1835 int j;
8592f29d 1836 struct dl *dl;
a5d85af7 1837 int map_disks = info->array.raid_disks;
90fa1a29 1838 __u32 *cptr;
a322f70c 1839
95eeceeb 1840 memset(info, 0, sizeof(*info));
8a2848a7 1841 if (layout_ddf2md(&vc->conf, &info->array) == -1)
1842 return;
a322f70c 1843 info->array.md_minor = -1;
90fa1a29
JS
1844 cptr = (__u32 *)(vc->conf.guid + 16);
1845 info->array.ctime = DECADE + __be32_to_cpu(*cptr);
d2ca6449
NB
1846 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1847 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1848 info->custom_array_size = 0;
d2ca6449 1849
f21e18ca 1850 if (cd >= 0 && (unsigned)cd < ddf->mppe) {
57a66662 1851 info->data_offset =
1852 __be64_to_cpu(LBA_OFFSET(ddf, &vc->conf)[cd]);
d2ca6449
NB
1853 if (vc->block_sizes)
1854 info->component_size = vc->block_sizes[cd];
1855 else
1856 info->component_size = __be64_to_cpu(vc->conf.blocks);
1857 }
a322f70c 1858
fb204fb2
N
1859 for (dl = ddf->dlist; dl ; dl = dl->next)
1860 if (dl->raiddisk == ddf->currentdev)
1861 break;
1862
a322f70c
DW
1863 info->disk.major = 0;
1864 info->disk.minor = 0;
fb204fb2 1865 info->disk.state = 0;
8592f29d
N
1866 if (dl) {
1867 info->disk.major = dl->major;
1868 info->disk.minor = dl->minor;
fb204fb2
N
1869 info->disk.raid_disk = dl->raiddisk;
1870 info->disk.number = dl->pdnum;
1871 info->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
8592f29d 1872 }
a322f70c 1873
103f2410
NB
1874 info->container_member = ddf->currentconf->vcnum;
1875
921d9e16 1876 info->recovery_start = MaxSector;
80d26cb2 1877 info->resync_start = 0;
624c5ad4 1878 info->reshape_active = 0;
6e75048b 1879 info->recovery_blocked = 0;
80d26cb2
NB
1880 if (!(ddf->virt->entries[info->container_member].state
1881 & DDF_state_inconsistent) &&
1882 (ddf->virt->entries[info->container_member].init_state
1883 & DDF_initstate_mask)
1884 == DDF_init_full)
b7528a20 1885 info->resync_start = MaxSector;
80d26cb2 1886
a322f70c
DW
1887 uuid_from_super_ddf(st, info->uuid);
1888
f35f2525
N
1889 info->array.major_version = -1;
1890 info->array.minor_version = -2;
9b63e648 1891 sprintf(info->text_version, "/%s/%d",
4dd2df09 1892 st->container_devnm,
9b63e648 1893 info->container_member);
a67dd8cc 1894 info->safe_mode_delay = 200;
159c3a1a 1895
db42fa9b
N
1896 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1897 info->name[16]=0;
1898 for(j=0; j<16; j++)
1899 if (info->name[j] == ' ')
1900 info->name[j] = 0;
a5d85af7
N
1901
1902 if (map)
1903 for (j = 0; j < map_disks; j++) {
1904 map[j] = 0;
1905 if (j < info->array.raid_disks) {
1906 int i = find_phys(ddf, vc->conf.phys_refnum[j]);
613b0d17 1907 if (i >= 0 &&
a5d85af7
N
1908 (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) &&
1909 !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed))
1910 map[i] = 1;
1911 }
1912 }
a322f70c
DW
1913}
1914
1915static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1916 char *update,
1917 char *devname, int verbose,
1918 int uuid_set, char *homehost)
1919{
1920 /* For 'assemble' and 'force' we need to return non-zero if any
1921 * change was made. For others, the return value is ignored.
1922 * Update options are:
1923 * force-one : This device looks a bit old but needs to be included,
1924 * update age info appropriately.
1925 * assemble: clear any 'faulty' flag to allow this device to
1926 * be assembled.
1927 * force-array: Array is degraded but being forced, mark it clean
1928 * if that will be needed to assemble it.
1929 *
1930 * newdev: not used ????
1931 * grow: Array has gained a new device - this is currently for
1932 * linear only
1933 * resync: mark as dirty so a resync will happen.
59e36268 1934 * uuid: Change the uuid of the array to match what is given
a322f70c
DW
1935 * homehost: update the recorded homehost
1936 * name: update the name - preserving the homehost
1937 * _reshape_progress: record new reshape_progress position.
1938 *
1939 * Following are not relevant for this version:
1940 * sparc2.2 : update from old dodgey metadata
1941 * super-minor: change the preferred_minor number
1942 * summaries: update redundant counters.
1943 */
1944 int rv = 0;
1945// struct ddf_super *ddf = st->sb;
7a7cc504 1946// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW
1947// struct virtual_entry *ve = find_ve(ddf);
1948
a322f70c
DW
1949 /* we don't need to handle "force-*" or "assemble" as
1950 * there is no need to 'trick' the kernel. We the metadata is
1951 * first updated to activate the array, all the implied modifications
1952 * will just happen.
1953 */
1954
1955 if (strcmp(update, "grow") == 0) {
1956 /* FIXME */
1e2b2765 1957 } else if (strcmp(update, "resync") == 0) {
a322f70c 1958// info->resync_checkpoint = 0;
1e2b2765 1959 } else if (strcmp(update, "homehost") == 0) {
a322f70c
DW
1960 /* homehost is stored in controller->vendor_data,
1961 * or it is when we are the vendor
1962 */
1963// if (info->vendor_is_local)
1964// strcpy(ddf->controller.vendor_data, homehost);
1e2b2765 1965 rv = -1;
f49208ec 1966 } else if (strcmp(update, "name") == 0) {
a322f70c
DW
1967 /* name is stored in virtual_entry->name */
1968// memset(ve->name, ' ', 16);
1969// strncpy(ve->name, info->name, 16);
1e2b2765 1970 rv = -1;
f49208ec 1971 } else if (strcmp(update, "_reshape_progress") == 0) {
a322f70c 1972 /* We don't support reshape yet */
f49208ec
N
1973 } else if (strcmp(update, "assemble") == 0 ) {
1974 /* Do nothing, just succeed */
1975 rv = 0;
1e2b2765
N
1976 } else
1977 rv = -1;
a322f70c
DW
1978
1979// update_all_csum(ddf);
1980
1981 return rv;
1982}
1983
5f8097be
NB
1984static void make_header_guid(char *guid)
1985{
1986 __u32 stamp;
5f8097be
NB
1987 /* Create a DDF Header of Virtual Disk GUID */
1988
1989 /* 24 bytes of fiction required.
1990 * first 8 are a 'vendor-id' - "Linux-MD"
1991 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1992 * Remaining 8 random number plus timestamp
1993 */
1994 memcpy(guid, T10, sizeof(T10));
1995 stamp = __cpu_to_be32(0xdeadbeef);
1996 memcpy(guid+8, &stamp, 4);
1997 stamp = __cpu_to_be32(0);
1998 memcpy(guid+12, &stamp, 4);
1999 stamp = __cpu_to_be32(time(0) - DECADE);
2000 memcpy(guid+16, &stamp, 4);
bfb7ea78 2001 stamp = random32();
5f8097be 2002 memcpy(guid+20, &stamp, 4);
5f8097be 2003}
59e36268 2004
fb9d0acb 2005static unsigned int find_unused_vde(const struct ddf_super *ddf)
2006{
2007 unsigned int i;
2008 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
2009 if (all_ff(ddf->virt->entries[i].guid))
2010 return i;
2011 }
2012 return DDF_NOTFOUND;
2013}
2014
2015static unsigned int find_vde_by_name(const struct ddf_super *ddf,
2016 const char *name)
2017{
2018 unsigned int i;
2019 if (name == NULL)
2020 return DDF_NOTFOUND;
2021 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
2022 if (all_ff(ddf->virt->entries[i].guid))
2023 continue;
2024 if (!strncmp(name, ddf->virt->entries[i].name,
2025 sizeof(ddf->virt->entries[i].name)))
2026 return i;
2027 }
2028 return DDF_NOTFOUND;
2029}
2030
2031static unsigned int find_vde_by_guid(const struct ddf_super *ddf,
2032 const char *guid)
2033{
2034 unsigned int i;
2035 if (guid == NULL || all_ff(guid))
2036 return DDF_NOTFOUND;
2037 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++)
2038 if (!memcmp(ddf->virt->entries[i].guid, guid, DDF_GUID_LEN))
2039 return i;
2040 return DDF_NOTFOUND;
2041}
2042
78e44928
NB
2043static int init_super_ddf_bvd(struct supertype *st,
2044 mdu_array_info_t *info,
2045 unsigned long long size,
2046 char *name, char *homehost,
83cd1e97 2047 int *uuid, unsigned long long data_offset);
78e44928 2048
a322f70c
DW
2049static int init_super_ddf(struct supertype *st,
2050 mdu_array_info_t *info,
2051 unsigned long long size, char *name, char *homehost,
83cd1e97 2052 int *uuid, unsigned long long data_offset)
a322f70c
DW
2053{
2054 /* This is primarily called by Create when creating a new array.
2055 * We will then get add_to_super called for each component, and then
2056 * write_init_super called to write it out to each device.
2057 * For DDF, Create can create on fresh devices or on a pre-existing
2058 * array.
2059 * To create on a pre-existing array a different method will be called.
2060 * This one is just for fresh drives.
2061 *
2062 * We need to create the entire 'ddf' structure which includes:
2063 * DDF headers - these are easy.
2064 * Controller data - a Sector describing this controller .. not that
2065 * this is a controller exactly.
2066 * Physical Disk Record - one entry per device, so
2067 * leave plenty of space.
2068 * Virtual Disk Records - again, just leave plenty of space.
2069 * This just lists VDs, doesn't give details
2070 * Config records - describes the VDs that use this disk
2071 * DiskData - describes 'this' device.
2072 * BadBlockManagement - empty
2073 * Diag Space - empty
2074 * Vendor Logs - Could we put bitmaps here?
2075 *
2076 */
2077 struct ddf_super *ddf;
2078 char hostname[17];
2079 int hostlen;
a322f70c
DW
2080 int max_phys_disks, max_virt_disks;
2081 unsigned long long sector;
2082 int clen;
2083 int i;
2084 int pdsize, vdsize;
2085 struct phys_disk *pd;
2086 struct virtual_disk *vd;
2087
83cd1e97 2088 if (data_offset != INVALID_SECTORS) {
ed503f89 2089 pr_err("data-offset not supported by DDF\n");
83cd1e97
N
2090 return 0;
2091 }
2092
78e44928 2093 if (st->sb)
83cd1e97
N
2094 return init_super_ddf_bvd(st, info, size, name, homehost, uuid,
2095 data_offset);
ba7eb04f 2096
3d2c4fc7 2097 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
e7b84f9d 2098 pr_err("%s could not allocate superblock\n", __func__);
3d2c4fc7
DW
2099 return 0;
2100 }
6264b437 2101 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW
2102 ddf->dlist = NULL; /* no physical disks yet */
2103 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW
2104 st->sb = ddf;
2105
2106 if (info == NULL) {
2107 /* zeroing superblock */
2108 return 0;
2109 }
a322f70c
DW
2110
2111 /* At least 32MB *must* be reserved for the ddf. So let's just
2112 * start 32MB from the end, and put the primary header there.
2113 * Don't do secondary for now.
2114 * We don't know exactly where that will be yet as it could be
2115 * different on each device. To just set up the lengths.
2116 *
2117 */
2118
2119 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 2120 make_header_guid(ddf->anchor.guid);
a322f70c 2121
59e36268 2122 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW
2123 ddf->anchor.seq = __cpu_to_be32(1);
2124 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
2125 ddf->anchor.openflag = 0xFF;
2126 ddf->anchor.foreignflag = 0;
2127 ddf->anchor.enforcegroups = 0; /* Is this best?? */
2128 ddf->anchor.pad0 = 0xff;
2129 memset(ddf->anchor.pad1, 0xff, 12);
2130 memset(ddf->anchor.header_ext, 0xff, 32);
2131 ddf->anchor.primary_lba = ~(__u64)0;
2132 ddf->anchor.secondary_lba = ~(__u64)0;
2133 ddf->anchor.type = DDF_HEADER_ANCHOR;
2134 memset(ddf->anchor.pad2, 0xff, 3);
2135 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
2136 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
2137 of 32M reserved.. */
2138 max_phys_disks = 1023; /* Should be enough */
2139 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
2140 max_virt_disks = 255;
2141 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
2142 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
2143 ddf->max_part = 64;
8c3b8c2c 2144 ddf->mppe = 256;
59e36268
NB
2145 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
2146 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
2147 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 2148 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW
2149 /* controller sections is one sector long immediately
2150 * after the ddf header */
2151 sector = 1;
2152 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
2153 ddf->anchor.controller_section_length = __cpu_to_be32(1);
2154 sector += 1;
2155
2156 /* phys is 8 sectors after that */
2157 pdsize = ROUND_UP(sizeof(struct phys_disk) +
2158 sizeof(struct phys_disk_entry)*max_phys_disks,
2159 512);
2160 switch(pdsize/512) {
2161 case 2: case 8: case 32: case 128: case 512: break;
2162 default: abort();
2163 }
2164 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
2165 ddf->anchor.phys_section_length =
2166 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
2167 sector += pdsize/512;
2168
2169 /* virt is another 32 sectors */
2170 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
2171 sizeof(struct virtual_entry) * max_virt_disks,
2172 512);
2173 switch(vdsize/512) {
2174 case 2: case 8: case 32: case 128: case 512: break;
2175 default: abort();
2176 }
2177 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
2178 ddf->anchor.virt_section_length =
2179 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
2180 sector += vdsize/512;
2181
59e36268 2182 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW
2183 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
2184 ddf->anchor.config_section_length = __cpu_to_be32(clen);
2185 sector += clen;
2186
2187 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
2188 ddf->anchor.data_section_length = __cpu_to_be32(1);
2189 sector += 1;
2190
2191 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
2192 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
2193 ddf->anchor.diag_space_length = __cpu_to_be32(0);
2194 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
2195 ddf->anchor.vendor_length = __cpu_to_be32(0);
2196 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
2197
2198 memset(ddf->anchor.pad4, 0xff, 256);
2199
2200 memcpy(&ddf->primary, &ddf->anchor, 512);
2201 memcpy(&ddf->secondary, &ddf->anchor, 512);
2202
2203 ddf->primary.openflag = 1; /* I guess.. */
2204 ddf->primary.type = DDF_HEADER_PRIMARY;
2205
2206 ddf->secondary.openflag = 1; /* I guess.. */
2207 ddf->secondary.type = DDF_HEADER_SECONDARY;
2208
2209 ddf->active = &ddf->primary;
2210
2211 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
2212
2213 /* 24 more bytes of fiction required.
2214 * first 8 are a 'vendor-id' - "Linux-MD"
2215 * Remaining 16 are serial number.... maybe a hostname would do?
2216 */
2217 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW
2218 gethostname(hostname, sizeof(hostname));
2219 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW
2220 hostlen = strlen(hostname);
2221 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
2222 for (i = strlen(T10) ; i+hostlen < 24; i++)
2223 ddf->controller.guid[i] = ' ';
2224
2225 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
2226 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
2227 ddf->controller.type.sub_vendor_id = 0;
2228 ddf->controller.type.sub_device_id = 0;
2229 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
2230 memset(ddf->controller.pad, 0xff, 8);
2231 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N
2232 if (homehost && strlen(homehost) < 440)
2233 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 2234
3d2c4fc7 2235 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
e7b84f9d 2236 pr_err("%s could not allocate pd\n", __func__);
3d2c4fc7
DW
2237 return 0;
2238 }
6416d527 2239 ddf->phys = pd;
a322f70c
DW
2240 ddf->pdsize = pdsize;
2241
2242 memset(pd, 0xff, pdsize);
2243 memset(pd, 0, sizeof(*pd));
076515ba 2244 pd->magic = DDF_PHYS_RECORDS_MAGIC;
a322f70c
DW
2245 pd->used_pdes = __cpu_to_be16(0);
2246 pd->max_pdes = __cpu_to_be16(max_phys_disks);
2247 memset(pd->pad, 0xff, 52);
4a3ca8ac 2248 for (i = 0; i < max_phys_disks; i++)
2249 memset(pd->entries[i].guid, 0xff, DDF_GUID_LEN);
a322f70c 2250
3d2c4fc7 2251 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
e7b84f9d 2252 pr_err("%s could not allocate vd\n", __func__);
3d2c4fc7
DW
2253 return 0;
2254 }
6416d527 2255 ddf->virt = vd;
a322f70c
DW
2256 ddf->vdsize = vdsize;
2257 memset(vd, 0, vdsize);
2258 vd->magic = DDF_VIRT_RECORDS_MAGIC;
2259 vd->populated_vdes = __cpu_to_be16(0);
2260 vd->max_vdes = __cpu_to_be16(max_virt_disks);
2261 memset(vd->pad, 0xff, 52);
2262
5f8097be
NB
2263 for (i=0; i<max_virt_disks; i++)
2264 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
2265
a322f70c 2266 st->sb = ddf;
7d5a7ff3 2267 ddf_set_updates_pending(ddf);
a322f70c
DW
2268 return 1;
2269}
2270
5f8097be
NB
2271static int chunk_to_shift(int chunksize)
2272{
2273 return ffs(chunksize/512)-1;
2274}
2275
0e600426 2276#ifndef MDASSEMBLE
59e36268
NB
2277struct extent {
2278 unsigned long long start, size;
2279};
78e44928 2280static int cmp_extent(const void *av, const void *bv)
59e36268
NB
2281{
2282 const struct extent *a = av;
2283 const struct extent *b = bv;
2284 if (a->start < b->start)
2285 return -1;
2286 if (a->start > b->start)
2287 return 1;
2288 return 0;
2289}
2290
78e44928 2291static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB
2292{
2293 /* find a list of used extents on the give physical device
2294 * (dnum) of the given ddf.
2295 * Return a malloced array of 'struct extent'
2296
613b0d17 2297 * FIXME ignore DDF_Legacy devices?
59e36268
NB
2298
2299 */
2300 struct extent *rv;
2301 int n = 0;
fcc22180 2302 unsigned int i;
59e36268 2303
503975b9 2304 rv = xmalloc(sizeof(struct extent) * (ddf->max_part + 2));
59e36268
NB
2305
2306 for (i = 0; i < ddf->max_part; i++) {
fcc22180 2307 const struct vd_config *bvd;
2308 unsigned int ibvd;
59e36268 2309 struct vcl *v = dl->vlist[i];
fcc22180 2310 if (v == NULL ||
2311 get_pd_index_from_refnum(v, dl->disk.refnum, ddf->mppe,
2312 &bvd, &ibvd) == DDF_NOTFOUND)
59e36268 2313 continue;
fcc22180 2314 rv[n].start = __be64_to_cpu(LBA_OFFSET(ddf, bvd)[ibvd]);
2315 rv[n].size = __be64_to_cpu(bvd->blocks);
2316 n++;
59e36268
NB
2317 }
2318 qsort(rv, n, sizeof(*rv), cmp_extent);
2319
2320 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
2321 rv[n].size = 0;
2322 return rv;
2323}
0e600426 2324#endif
59e36268 2325
5f8097be
NB
2326static int init_super_ddf_bvd(struct supertype *st,
2327 mdu_array_info_t *info,
2328 unsigned long long size,
2329 char *name, char *homehost,
83cd1e97 2330 int *uuid, unsigned long long data_offset)
5f8097be
NB
2331{
2332 /* We are creating a BVD inside a pre-existing container.
2333 * so st->sb is already set.
2334 * We need to create a new vd_config and a new virtual_entry
2335 */
2336 struct ddf_super *ddf = st->sb;
5aaf6c7b 2337 unsigned int venum, i;
5f8097be
NB
2338 struct virtual_entry *ve;
2339 struct vcl *vcl;
2340 struct vd_config *vc;
5f8097be 2341
fb9d0acb 2342 if (find_vde_by_name(ddf, name) != DDF_NOTFOUND) {
2343 pr_err("This ddf already has an array called %s\n", name);
5f8097be
NB
2344 return 0;
2345 }
fb9d0acb 2346 venum = find_unused_vde(ddf);
2347 if (venum == DDF_NOTFOUND) {
2348 pr_err("Cannot find spare slot for virtual disk\n");
5f8097be
NB
2349 return 0;
2350 }
2351 ve = &ddf->virt->entries[venum];
2352
2353 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
2354 * timestamp, random number
2355 */
2356 make_header_guid(ve->guid);
2357 ve->unit = __cpu_to_be16(info->md_minor);
2358 ve->pad0 = 0xFFFF;
2359 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
2360 ve->type = 0;
7a7cc504
NB
2361 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
2362 if (info->state & 1) /* clean */
2363 ve->init_state = DDF_init_full;
2364 else
2365 ve->init_state = DDF_init_not;
2366
5f8097be
NB
2367 memset(ve->pad1, 0xff, 14);
2368 memset(ve->name, ' ', 16);
2369 if (name)
2370 strncpy(ve->name, name, 16);
2371 ddf->virt->populated_vdes =
2372 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
2373
2374 /* Now create a new vd_config */
3d2c4fc7
DW
2375 if (posix_memalign((void**)&vcl, 512,
2376 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
e7b84f9d 2377 pr_err("%s could not allocate vd_config\n", __func__);
3d2c4fc7
DW
2378 return 0;
2379 }
59e36268
NB
2380 vcl->vcnum = venum;
2381 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB
2382 vc = &vcl->conf;
2383
2384 vc->magic = DDF_VD_CONF_MAGIC;
2385 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
2386 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
2387 vc->seqnum = __cpu_to_be32(1);
2388 memset(vc->pad0, 0xff, 24);
5f8097be 2389 vc->chunk_shift = chunk_to_shift(info->chunk_size);
a3163bf0 2390 if (layout_md2ddf(info, vc) == -1 ||
2391 __be16_to_cpu(vc->prim_elmnt_count) > ddf->mppe) {
2392 pr_err("%s: unsupported RAID level/layout %d/%d with %d disks\n",
2393 __func__, info->level, info->layout, info->raid_disks);
2394 free(vcl);
2395 return 0;
2396 }
5f8097be 2397 vc->sec_elmnt_seq = 0;
3c48f7be 2398 if (alloc_other_bvds(ddf, vcl) != 0) {
2399 pr_err("%s could not allocate other bvds\n",
2400 __func__);
2401 free(vcl);
2402 return 0;
2403 }
5f8097be
NB
2404 vc->blocks = __cpu_to_be64(info->size * 2);
2405 vc->array_blocks = __cpu_to_be64(
2406 calc_array_size(info->level, info->raid_disks, info->layout,
2407 info->chunk_size, info->size*2));
2408 memset(vc->pad1, 0xff, 8);
2409 vc->spare_refs[0] = 0xffffffff;
2410 vc->spare_refs[1] = 0xffffffff;
2411 vc->spare_refs[2] = 0xffffffff;
2412 vc->spare_refs[3] = 0xffffffff;
2413 vc->spare_refs[4] = 0xffffffff;
2414 vc->spare_refs[5] = 0xffffffff;
2415 vc->spare_refs[6] = 0xffffffff;
2416 vc->spare_refs[7] = 0xffffffff;
2417 memset(vc->cache_pol, 0, 8);
2418 vc->bg_rate = 0x80;
2419 memset(vc->pad2, 0xff, 3);
2420 memset(vc->pad3, 0xff, 52);
2421 memset(vc->pad4, 0xff, 192);
2422 memset(vc->v0, 0xff, 32);
2423 memset(vc->v1, 0xff, 32);
2424 memset(vc->v2, 0xff, 16);
2425 memset(vc->v3, 0xff, 16);
2426 memset(vc->vendor, 0xff, 32);
598f0d58 2427
8c3b8c2c 2428 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2429 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be 2430
5aaf6c7b 2431 for (i = 1; i < vc->sec_elmnt_count; i++) {
2432 memcpy(vcl->other_bvds[i-1], vc, ddf->conf_rec_len * 512);
2433 vcl->other_bvds[i-1]->sec_elmnt_seq = i;
2434 }
2435
5f8097be
NB
2436 vcl->next = ddf->conflist;
2437 ddf->conflist = vcl;
d2ca6449 2438 ddf->currentconf = vcl;
7d5a7ff3 2439 ddf_set_updates_pending(ddf);
5f8097be
NB
2440 return 1;
2441}
2442
0e600426 2443#ifndef MDASSEMBLE
5f8097be
NB
2444static void add_to_super_ddf_bvd(struct supertype *st,
2445 mdu_disk_info_t *dk, int fd, char *devname)
2446{
2447 /* fd and devname identify a device with-in the ddf container (st).
2448 * dk identifies a location in the new BVD.
2449 * We need to find suitable free space in that device and update
2450 * the phys_refnum and lba_offset for the newly created vd_config.
2451 * We might also want to update the type in the phys_disk
5575e7d9 2452 * section.
8592f29d
N
2453 *
2454 * Alternately: fd == -1 and we have already chosen which device to
2455 * use and recorded in dlist->raid_disk;
5f8097be
NB
2456 */
2457 struct dl *dl;
2458 struct ddf_super *ddf = st->sb;
2459 struct vd_config *vc;
2460 __u64 *lba_offset;
f21e18ca
N
2461 unsigned int working;
2462 unsigned int i;
59e36268
NB
2463 unsigned long long blocks, pos, esize;
2464 struct extent *ex;
5f8097be 2465
8592f29d
N
2466 if (fd == -1) {
2467 for (dl = ddf->dlist; dl ; dl = dl->next)
2468 if (dl->raiddisk == dk->raid_disk)
2469 break;
2470 } else {
2471 for (dl = ddf->dlist; dl ; dl = dl->next)
2472 if (dl->major == dk->major &&
2473 dl->minor == dk->minor)
2474 break;
2475 }
5f8097be
NB
2476 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2477 return;
2478
d2ca6449 2479 vc = &ddf->currentconf->conf;
57a66662 2480 lba_offset = LBA_OFFSET(ddf, &ddf->currentconf->conf);
59e36268
NB
2481
2482 ex = get_extents(ddf, dl);
2483 if (!ex)
2484 return;
2485
2486 i = 0; pos = 0;
2487 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB
2488 if (ddf->currentconf->block_sizes)
2489 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB
2490
2491 do {
2492 esize = ex[i].start - pos;
2493 if (esize >= blocks)
2494 break;
2495 pos = ex[i].start + ex[i].size;
2496 i++;
2497 } while (ex[i-1].size);
2498
2499 free(ex);
2500 if (esize < blocks)
2501 return;
2502
d2ca6449 2503 ddf->currentdev = dk->raid_disk;
5f8097be 2504 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2505 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2506
f21e18ca 2507 for (i = 0; i < ddf->max_part ; i++)
5575e7d9
NB
2508 if (dl->vlist[i] == NULL)
2509 break;
2510 if (i == ddf->max_part)
2511 return;
d2ca6449 2512 dl->vlist[i] = ddf->currentconf;
5f8097be 2513
8592f29d
N
2514 if (fd >= 0)
2515 dl->fd = fd;
2516 if (devname)
2517 dl->devname = devname;
7a7cc504
NB
2518
2519 /* Check how many working raid_disks, and if we can mark
2520 * array as optimal yet
2521 */
2522 working = 0;
5575e7d9 2523
f21e18ca 2524 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
7a7cc504
NB
2525 if (vc->phys_refnum[i] != 0xffffffff)
2526 working++;
59e36268 2527
5575e7d9 2528 /* Find which virtual_entry */
d2ca6449 2529 i = ddf->currentconf->vcnum;
7a7cc504 2530 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2531 ddf->virt->entries[i].state =
2532 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB
2533 | DDF_state_optimal;
2534
2535 if (vc->prl == DDF_RAID6 &&
2536 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2537 ddf->virt->entries[i].state =
2538 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2539 | DDF_state_part_optimal;
5575e7d9
NB
2540
2541 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2542 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
7d5a7ff3 2543 ddf_set_updates_pending(ddf);
5f8097be
NB
2544}
2545
4a3ca8ac 2546static unsigned int find_unused_pde(const struct ddf_super *ddf)
2547{
2548 unsigned int i;
2549 for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++) {
2550 if (all_ff(ddf->phys->entries[i].guid))
2551 return i;
2552 }
2553 return DDF_NOTFOUND;
2554}
2555
a322f70c
DW
2556/* add a device to a container, either while creating it or while
2557 * expanding a pre-existing container
2558 */
f20c3968 2559static int add_to_super_ddf(struct supertype *st,
72ca9bcf
N
2560 mdu_disk_info_t *dk, int fd, char *devname,
2561 unsigned long long data_offset)
a322f70c
DW
2562{
2563 struct ddf_super *ddf = st->sb;
2564 struct dl *dd;
2565 time_t now;
2566 struct tm *tm;
2567 unsigned long long size;
2568 struct phys_disk_entry *pde;
f21e18ca 2569 unsigned int n, i;
a322f70c 2570 struct stat stb;
90fa1a29 2571 __u32 *tptr;
a322f70c 2572
78e44928
NB
2573 if (ddf->currentconf) {
2574 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2575 return 0;
78e44928
NB
2576 }
2577
a322f70c
DW
2578 /* This is device numbered dk->number. We need to create
2579 * a phys_disk entry and a more detailed disk_data entry.
2580 */
2581 fstat(fd, &stb);
4a3ca8ac 2582 n = find_unused_pde(ddf);
2583 if (n == DDF_NOTFOUND) {
2584 pr_err("%s: No free slot in array, cannot add disk\n",
2585 __func__);
2586 return 1;
2587 }
2588 pde = &ddf->phys->entries[n];
2589
3d2c4fc7
DW
2590 if (posix_memalign((void**)&dd, 512,
2591 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
e7b84f9d
N
2592 pr_err("%s could allocate buffer for new disk, aborting\n",
2593 __func__);
f20c3968 2594 return 1;
3d2c4fc7 2595 }
a322f70c
DW
2596 dd->major = major(stb.st_rdev);
2597 dd->minor = minor(stb.st_rdev);
2598 dd->devname = devname;
a322f70c 2599 dd->fd = fd;
b2280677 2600 dd->spare = NULL;
a322f70c
DW
2601
2602 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2603 now = time(0);
2604 tm = localtime(&now);
2605 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2606 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
90fa1a29
JS
2607 tptr = (__u32 *)(dd->disk.guid + 16);
2608 *tptr++ = random32();
2609 *tptr = random32();
a322f70c 2610
59e36268
NB
2611 do {
2612 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2613 dd->disk.refnum = random32();
f21e18ca
N
2614 for (i = __be16_to_cpu(ddf->active->max_pd_entries);
2615 i > 0; i--)
2616 if (ddf->phys->entries[i-1].refnum == dd->disk.refnum)
59e36268 2617 break;
f21e18ca 2618 } while (i > 0);
59e36268 2619
a322f70c
DW
2620 dd->disk.forced_ref = 1;
2621 dd->disk.forced_guid = 1;
2622 memset(dd->disk.vendor, ' ', 32);
2623 memcpy(dd->disk.vendor, "Linux", 5);
2624 memset(dd->disk.pad, 0xff, 442);
b2280677 2625 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW
2626 dd->vlist[i] = NULL;
2627
5575e7d9
NB
2628 dd->pdnum = n;
2629
2cc2983d
N
2630 if (st->update_tail) {
2631 int len = (sizeof(struct phys_disk) +
2632 sizeof(struct phys_disk_entry));
2633 struct phys_disk *pd;
2634
503975b9 2635 pd = xmalloc(len);
2cc2983d
N
2636 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2637 pd->used_pdes = __cpu_to_be16(n);
2638 pde = &pd->entries[0];
2639 dd->mdupdate = pd;
4a3ca8ac 2640 } else
2641 ddf->phys->used_pdes = __cpu_to_be16(
2642 1 + __be16_to_cpu(ddf->phys->used_pdes));
a322f70c
DW
2643
2644 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2645 pde->refnum = dd->disk.refnum;
5575e7d9 2646 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW
2647 pde->state = __cpu_to_be16(DDF_Online);
2648 get_dev_size(fd, NULL, &size);
2649 /* We are required to reserve 32Meg, and record the size in sectors */
2650 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2651 sprintf(pde->path, "%17.17s","Information: nil") ;
2652 memset(pde->pad, 0xff, 6);
2653
d2ca6449 2654 dd->size = size >> 9;
2cc2983d
N
2655 if (st->update_tail) {
2656 dd->next = ddf->add_list;
2657 ddf->add_list = dd;
2658 } else {
2659 dd->next = ddf->dlist;
2660 ddf->dlist = dd;
7d5a7ff3 2661 ddf_set_updates_pending(ddf);
2cc2983d 2662 }
f20c3968
DW
2663
2664 return 0;
a322f70c
DW
2665}
2666
4dd968cc
N
2667static int remove_from_super_ddf(struct supertype *st, mdu_disk_info_t *dk)
2668{
2669 struct ddf_super *ddf = st->sb;
2670 struct dl *dl;
2671
2672 /* mdmon has noticed that this disk (dk->major/dk->minor) has
2673 * disappeared from the container.
2674 * We need to arrange that it disappears from the metadata and
2675 * internal data structures too.
2676 * Most of the work is done by ddf_process_update which edits
2677 * the metadata and closes the file handle and attaches the memory
2678 * where free_updates will free it.
2679 */
2680 for (dl = ddf->dlist; dl ; dl = dl->next)
2681 if (dl->major == dk->major &&
2682 dl->minor == dk->minor)
2683 break;
2684 if (!dl)
2685 return -1;
2686
2687 if (st->update_tail) {
2688 int len = (sizeof(struct phys_disk) +
2689 sizeof(struct phys_disk_entry));
2690 struct phys_disk *pd;
2691
503975b9 2692 pd = xmalloc(len);
4dd968cc
N
2693 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2694 pd->used_pdes = __cpu_to_be16(dl->pdnum);
2695 pd->entries[0].state = __cpu_to_be16(DDF_Missing);
2696 append_metadata_update(st, pd, len);
2697 }
2698 return 0;
2699}
2700
a322f70c
DW
2701/*
2702 * This is the write_init_super method for a ddf container. It is
2703 * called when creating a container or adding another device to a
2704 * container.
2705 */
42d5dfd9 2706#define NULL_CONF_SZ 4096
18a2f463 2707
7f798aca 2708static int __write_ddf_structure(struct dl *d, struct ddf_super *ddf, __u8 type,
2709 char *null_aligned)
a322f70c 2710{
7f798aca 2711 unsigned long long sector;
2712 struct ddf_header *header;
2713 int fd, i, n_config, conf_size;
a4057a88 2714 int ret = 0;
7f798aca 2715
2716 fd = d->fd;
2717
2718 switch (type) {
2719 case DDF_HEADER_PRIMARY:
2720 header = &ddf->primary;
2721 sector = __be64_to_cpu(header->primary_lba);
2722 break;
2723 case DDF_HEADER_SECONDARY:
2724 header = &ddf->secondary;
2725 sector = __be64_to_cpu(header->secondary_lba);
2726 break;
2727 default:
2728 return 0;
2729 }
2730
2731 header->type = type;
a4057a88 2732 header->openflag = 1;
7f798aca 2733 header->crc = calc_crc(header, 512);
2734
2735 lseek64(fd, sector<<9, 0);
2736 if (write(fd, header, 512) < 0)
a4057a88 2737 goto out;
7f798aca 2738
2739 ddf->controller.crc = calc_crc(&ddf->controller, 512);
2740 if (write(fd, &ddf->controller, 512) < 0)
a4057a88 2741 goto out;
a322f70c 2742
7f798aca 2743 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2744 if (write(fd, ddf->phys, ddf->pdsize) < 0)
a4057a88 2745 goto out;
7f798aca 2746 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
2747 if (write(fd, ddf->virt, ddf->vdsize) < 0)
a4057a88 2748 goto out;
7f798aca 2749
2750 /* Now write lots of config records. */
2751 n_config = ddf->max_part;
2752 conf_size = ddf->conf_rec_len * 512;
2753 for (i = 0 ; i <= n_config ; i++) {
e3c2a365 2754 struct vcl *c;
2755 struct vd_config *vdc = NULL;
2756 if (i == n_config) {
7f798aca 2757 c = (struct vcl *)d->spare;
e3c2a365 2758 if (c)
2759 vdc = &c->conf;
2760 } else {
2761 unsigned int dummy;
2762 c = d->vlist[i];
2763 if (c)
2764 get_pd_index_from_refnum(
2765 c, d->disk.refnum,
2766 ddf->mppe,
2767 (const struct vd_config **)&vdc,
2768 &dummy);
2769 }
7f798aca 2770 if (c) {
dacf3dc5 2771 vdc->seqnum = header->seq;
e3c2a365 2772 vdc->crc = calc_crc(vdc, conf_size);
2773 if (write(fd, vdc, conf_size) < 0)
7f798aca 2774 break;
2775 } else {
2776 unsigned int togo = conf_size;
2777 while (togo > NULL_CONF_SZ) {
2778 if (write(fd, null_aligned, NULL_CONF_SZ) < 0)
2779 break;
2780 togo -= NULL_CONF_SZ;
2781 }
2782 if (write(fd, null_aligned, togo) < 0)
2783 break;
2784 }
2785 }
2786 if (i <= n_config)
a4057a88 2787 goto out;
7f798aca 2788
2789 d->disk.crc = calc_crc(&d->disk, 512);
2790 if (write(fd, &d->disk, 512) < 0)
a4057a88 2791 goto out;
7f798aca 2792
a4057a88 2793 ret = 1;
2794out:
2795 header->openflag = 0;
2796 header->crc = calc_crc(header, 512);
2797
2798 lseek64(fd, sector<<9, 0);
2799 if (write(fd, header, 512) < 0)
2800 ret = 0;
2801
2802 return ret;
7f798aca 2803}
2804
2805static int __write_init_super_ddf(struct supertype *st)
2806{
a322f70c 2807 struct ddf_super *ddf = st->sb;
a322f70c 2808 struct dl *d;
175593bf
DW
2809 int attempts = 0;
2810 int successes = 0;
7f798aca 2811 unsigned long long size;
42d5dfd9 2812 char *null_aligned;
0175cbf6 2813 __u32 seq;
42d5dfd9 2814
7d5a7ff3 2815 pr_state(ddf, __func__);
42d5dfd9
JS
2816 if (posix_memalign((void**)&null_aligned, 4096, NULL_CONF_SZ) != 0) {
2817 return -ENOMEM;
2818 }
2819 memset(null_aligned, 0xff, NULL_CONF_SZ);
a322f70c 2820
dc9e279c 2821 seq = ddf->active->seq + 1;
0175cbf6 2822
175593bf
DW
2823 /* try to write updated metadata,
2824 * if we catch a failure move on to the next disk
2825 */
a322f70c
DW
2826 for (d = ddf->dlist; d; d=d->next) {
2827 int fd = d->fd;
2828
2829 if (fd < 0)
2830 continue;
2831
175593bf 2832 attempts++;
a322f70c
DW
2833 /* We need to fill in the primary, (secondary) and workspace
2834 * lba's in the headers, set their checksums,
2835 * Also checksum phys, virt....
2836 *
2837 * Then write everything out, finally the anchor is written.
2838 */
2839 get_dev_size(fd, NULL, &size);
2840 size /= 512;
097bcf00 2841 if (d->workspace_lba != 0)
2842 ddf->anchor.workspace_lba = d->workspace_lba;
2843 else
2844 ddf->anchor.workspace_lba =
2845 __cpu_to_be64(size - 32*1024*2);
2846 if (d->primary_lba != 0)
2847 ddf->anchor.primary_lba = d->primary_lba;
2848 else
2849 ddf->anchor.primary_lba =
2850 __cpu_to_be64(size - 16*1024*2);
2851 if (d->secondary_lba != 0)
2852 ddf->anchor.secondary_lba = d->secondary_lba;
2853 else
2854 ddf->anchor.secondary_lba =
2855 __cpu_to_be64(size - 32*1024*2);
0175cbf6 2856 ddf->anchor.seq = seq;
a322f70c
DW
2857 memcpy(&ddf->primary, &ddf->anchor, 512);
2858 memcpy(&ddf->secondary, &ddf->anchor, 512);
2859
2860 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2861 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2862 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2863
7f798aca 2864 if (!__write_ddf_structure(d, ddf, DDF_HEADER_PRIMARY,
2865 null_aligned))
175593bf 2866 continue;
a322f70c 2867
7f798aca 2868 if (!__write_ddf_structure(d, ddf, DDF_HEADER_SECONDARY,
2869 null_aligned))
175593bf 2870 continue;
a322f70c 2871
a322f70c 2872 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW
2873 if (write(fd, &ddf->anchor, 512) < 0)
2874 continue;
2875 successes++;
2876 }
42d5dfd9 2877 free(null_aligned);
175593bf 2878
175593bf 2879 return attempts != successes;
a322f70c 2880}
7a7cc504
NB
2881
2882static int write_init_super_ddf(struct supertype *st)
2883{
9b1fb677
DW
2884 struct ddf_super *ddf = st->sb;
2885 struct vcl *currentconf = ddf->currentconf;
2886
2887 /* we are done with currentconf reset it to point st at the container */
2888 ddf->currentconf = NULL;
edd8d13c
NB
2889
2890 if (st->update_tail) {
2891 /* queue the virtual_disk and vd_config as metadata updates */
2892 struct virtual_disk *vd;
2893 struct vd_config *vc;
edd8d13c
NB
2894 int len;
2895
9b1fb677 2896 if (!currentconf) {
2cc2983d
N
2897 int len = (sizeof(struct phys_disk) +
2898 sizeof(struct phys_disk_entry));
2899
2900 /* adding a disk to the container. */
2901 if (!ddf->add_list)
2902 return 0;
2903
2904 append_metadata_update(st, ddf->add_list->mdupdate, len);
2905 ddf->add_list->mdupdate = NULL;
2906 return 0;
2907 }
2908
2909 /* Newly created VD */
2910
edd8d13c
NB
2911 /* First the virtual disk. We have a slightly fake header */
2912 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
503975b9 2913 vd = xmalloc(len);
edd8d13c 2914 *vd = *ddf->virt;
9b1fb677
DW
2915 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2916 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB
2917 append_metadata_update(st, vd, len);
2918
2919 /* Then the vd_config */
2920 len = ddf->conf_rec_len * 512;
503975b9 2921 vc = xmalloc(len);
9b1fb677 2922 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB
2923 append_metadata_update(st, vc, len);
2924
2925 /* FIXME I need to close the fds! */
2926 return 0;
613b0d17 2927 } else {
d682f344
N
2928 struct dl *d;
2929 for (d = ddf->dlist; d; d=d->next)
ba728be7 2930 while (Kill(d->devname, NULL, 0, -1, 1) == 0);
1cc7f4fe 2931 return __write_init_super_ddf(st);
d682f344 2932 }
7a7cc504
NB
2933}
2934
a322f70c
DW
2935#endif
2936
387fcd59
N
2937static __u64 avail_size_ddf(struct supertype *st, __u64 devsize,
2938 unsigned long long data_offset)
a322f70c
DW
2939{
2940 /* We must reserve the last 32Meg */
2941 if (devsize <= 32*1024*2)
2942 return 0;
2943 return devsize - 32*1024*2;
2944}
2945
2946#ifndef MDASSEMBLE
8592f29d
N
2947
2948static int reserve_space(struct supertype *st, int raiddisks,
2949 unsigned long long size, int chunk,
2950 unsigned long long *freesize)
2951{
2952 /* Find 'raiddisks' spare extents at least 'size' big (but
2953 * only caring about multiples of 'chunk') and remember
2954 * them.
2955 * If the cannot be found, fail.
2956 */
2957 struct dl *dl;
2958 struct ddf_super *ddf = st->sb;
2959 int cnt = 0;
2960
2961 for (dl = ddf->dlist; dl ; dl=dl->next) {
613b0d17 2962 dl->raiddisk = -1;
8592f29d
N
2963 dl->esize = 0;
2964 }
2965 /* Now find largest extent on each device */
2966 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2967 struct extent *e = get_extents(ddf, dl);
2968 unsigned long long pos = 0;
2969 int i = 0;
2970 int found = 0;
2971 unsigned long long minsize = size;
2972
2973 if (size == 0)
2974 minsize = chunk;
2975
2976 if (!e)
2977 continue;
2978 do {
2979 unsigned long long esize;
2980 esize = e[i].start - pos;
2981 if (esize >= minsize) {
2982 found = 1;
2983 minsize = esize;
2984 }
2985 pos = e[i].start + e[i].size;
2986 i++;
2987 } while (e[i-1].size);
2988 if (found) {
2989 cnt++;
2990 dl->esize = minsize;
2991 }
2992 free(e);
2993 }
2994 if (cnt < raiddisks) {
e7b84f9d 2995 pr_err("not enough devices with space to create array.\n");
8592f29d
N
2996 return 0; /* No enough free spaces large enough */
2997 }
2998 if (size == 0) {
2999 /* choose the largest size of which there are at least 'raiddisk' */
3000 for (dl = ddf->dlist ; dl ; dl=dl->next) {
3001 struct dl *dl2;
3002 if (dl->esize <= size)
3003 continue;
3004 /* This is bigger than 'size', see if there are enough */
3005 cnt = 0;
7b80ad6a 3006 for (dl2 = ddf->dlist; dl2 ; dl2=dl2->next)
8592f29d
N
3007 if (dl2->esize >= dl->esize)
3008 cnt++;
3009 if (cnt >= raiddisks)
3010 size = dl->esize;
3011 }
3012 if (chunk) {
3013 size = size / chunk;
3014 size *= chunk;
3015 }
3016 *freesize = size;
3017 if (size < 32) {
e7b84f9d 3018 pr_err("not enough spare devices to create array.\n");
8592f29d
N
3019 return 0;
3020 }
3021 }
3022 /* We have a 'size' of which there are enough spaces.
3023 * We simply do a first-fit */
3024 cnt = 0;
3025 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
3026 if (dl->esize < size)
3027 continue;
613b0d17 3028
8592f29d
N
3029 dl->raiddisk = cnt;
3030 cnt++;
3031 }
3032 return 1;
3033}
3034
2c514b71
NB
3035static int
3036validate_geometry_ddf_container(struct supertype *st,
3037 int level, int layout, int raiddisks,
3038 int chunk, unsigned long long size,
af4348dd 3039 unsigned long long data_offset,
2c514b71
NB
3040 char *dev, unsigned long long *freesize,
3041 int verbose);
78e44928
NB
3042
3043static int validate_geometry_ddf_bvd(struct supertype *st,
3044 int level, int layout, int raiddisks,
c21e737b 3045 int *chunk, unsigned long long size,
af4348dd 3046 unsigned long long data_offset,
2c514b71
NB
3047 char *dev, unsigned long long *freesize,
3048 int verbose);
78e44928
NB
3049
3050static int validate_geometry_ddf(struct supertype *st,
2c514b71 3051 int level, int layout, int raiddisks,
c21e737b 3052 int *chunk, unsigned long long size,
af4348dd 3053 unsigned long long data_offset,
2c514b71
NB
3054 char *dev, unsigned long long *freesize,
3055 int verbose)
a322f70c
DW
3056{
3057 int fd;
3058 struct mdinfo *sra;
3059 int cfd;
3060
3061 /* ddf potentially supports lots of things, but it depends on
3062 * what devices are offered (and maybe kernel version?)
3063 * If given unused devices, we will make a container.
3064 * If given devices in a container, we will make a BVD.
3065 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
3066 */
3067
bb7295f1
N
3068 if (chunk && *chunk == UnSet)
3069 *chunk = DEFAULT_CHUNK;
3070
542ef4ec 3071 if (level == -1000000) level = LEVEL_CONTAINER;
a322f70c 3072 if (level == LEVEL_CONTAINER) {
78e44928
NB
3073 /* Must be a fresh device to add to a container */
3074 return validate_geometry_ddf_container(st, level, layout,
c21e737b 3075 raiddisks, chunk?*chunk:0,
af4348dd
N
3076 size, data_offset, dev,
3077 freesize,
2c514b71 3078 verbose);
5f8097be
NB
3079 }
3080
78e44928 3081 if (!dev) {
a3163bf0 3082 mdu_array_info_t array = {
3083 .level = level, .layout = layout,
3084 .raid_disks = raiddisks
3085 };
3086 struct vd_config conf;
3087 if (layout_md2ddf(&array, &conf) == -1) {
b42f577a 3088 if (verbose)
94b08b7c 3089 pr_err("DDF does not support level %d /layout %d arrays with %d disks\n",
3090 level, layout, raiddisks);
78e44928 3091 return 0;
b42f577a 3092 }
78e44928 3093 /* Should check layout? etc */
8592f29d
N
3094
3095 if (st->sb && freesize) {
3096 /* --create was given a container to create in.
3097 * So we need to check that there are enough
3098 * free spaces and return the amount of space.
3099 * We may as well remember which drives were
3100 * chosen so that add_to_super/getinfo_super
3101 * can return them.
3102 */
c21e737b 3103 return reserve_space(st, raiddisks, size, chunk?*chunk:0, freesize);
8592f29d 3104 }
a322f70c 3105 return 1;
78e44928 3106 }
a322f70c 3107
8592f29d
N
3108 if (st->sb) {
3109 /* A container has already been opened, so we are
3110 * creating in there. Maybe a BVD, maybe an SVD.
3111 * Should make a distinction one day.
3112 */
3113 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
af4348dd
N
3114 chunk, size, data_offset, dev,
3115 freesize,
8592f29d
N
3116 verbose);
3117 }
78e44928
NB
3118 /* This is the first device for the array.
3119 * If it is a container, we read it in and do automagic allocations,
3120 * no other devices should be given.
3121 * Otherwise it must be a member device of a container, and we
3122 * do manual allocation.
3123 * Later we should check for a BVD and make an SVD.
a322f70c 3124 */
a322f70c
DW
3125 fd = open(dev, O_RDONLY|O_EXCL, 0);
3126 if (fd >= 0) {
4dd2df09 3127 sra = sysfs_read(fd, NULL, GET_VERSION);
a322f70c
DW
3128 close(fd);
3129 if (sra && sra->array.major_version == -1 &&
78e44928
NB
3130 strcmp(sra->text_version, "ddf") == 0) {
3131
3132 /* load super */
3133 /* find space for 'n' devices. */
3134 /* remember the devices */
3135 /* Somehow return the fact that we have enough */
a322f70c
DW
3136 }
3137
2c514b71 3138 if (verbose)
e7b84f9d
N
3139 pr_err("ddf: Cannot create this array "
3140 "on device %s - a container is required.\n",
3141 dev);
a322f70c
DW
3142 return 0;
3143 }
3144 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71 3145 if (verbose)
e7b84f9d 3146 pr_err("ddf: Cannot open %s: %s\n",
613b0d17 3147 dev, strerror(errno));
a322f70c
DW
3148 return 0;
3149 }
3150 /* Well, it is in use by someone, maybe a 'ddf' container. */
3151 cfd = open_container(fd);
3152 if (cfd < 0) {
3153 close(fd);
2c514b71 3154 if (verbose)
e7b84f9d 3155 pr_err("ddf: Cannot use %s: %s\n",
613b0d17 3156 dev, strerror(EBUSY));
a322f70c
DW
3157 return 0;
3158 }
4dd2df09 3159 sra = sysfs_read(cfd, NULL, GET_VERSION);
a322f70c
DW
3160 close(fd);
3161 if (sra && sra->array.major_version == -1 &&
3162 strcmp(sra->text_version, "ddf") == 0) {
3163 /* This is a member of a ddf container. Load the container
3164 * and try to create a bvd
3165 */
3166 struct ddf_super *ddf;
e1902a7b 3167 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL) == 0) {
5f8097be 3168 st->sb = ddf;
4dd2df09 3169 strcpy(st->container_devnm, fd2devnm(cfd));
a322f70c 3170 close(cfd);
78e44928 3171 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 3172 raiddisks, chunk, size,
af4348dd 3173 data_offset,
2c514b71
NB
3174 dev, freesize,
3175 verbose);
a322f70c
DW
3176 }
3177 close(cfd);
c42ec1ed
DW
3178 } else /* device may belong to a different container */
3179 return 0;
3180
a322f70c
DW
3181 return 1;
3182}
3183
2c514b71
NB
3184static int
3185validate_geometry_ddf_container(struct supertype *st,
3186 int level, int layout, int raiddisks,
3187 int chunk, unsigned long long size,
af4348dd 3188 unsigned long long data_offset,
2c514b71
NB
3189 char *dev, unsigned long long *freesize,
3190 int verbose)
a322f70c
DW
3191{
3192 int fd;
3193 unsigned long long ldsize;
3194
3195 if (level != LEVEL_CONTAINER)
3196 return 0;
3197 if (!dev)
3198 return 1;
3199
3200 fd = open(dev, O_RDONLY|O_EXCL, 0);
3201 if (fd < 0) {
2c514b71 3202 if (verbose)
e7b84f9d 3203 pr_err("ddf: Cannot open %s: %s\n",
613b0d17 3204 dev, strerror(errno));
a322f70c
DW
3205 return 0;
3206 }
3207 if (!get_dev_size(fd, dev, &ldsize)) {
3208 close(fd);
3209 return 0;
3210 }
3211 close(fd);
3212
387fcd59 3213 *freesize = avail_size_ddf(st, ldsize >> 9, INVALID_SECTORS);
ea17e7aa
N
3214 if (*freesize == 0)
3215 return 0;
a322f70c
DW
3216
3217 return 1;
3218}
3219
78e44928
NB
3220static int validate_geometry_ddf_bvd(struct supertype *st,
3221 int level, int layout, int raiddisks,
c21e737b 3222 int *chunk, unsigned long long size,
af4348dd 3223 unsigned long long data_offset,
2c514b71
NB
3224 char *dev, unsigned long long *freesize,
3225 int verbose)
a322f70c
DW
3226{
3227 struct stat stb;
3228 struct ddf_super *ddf = st->sb;
3229 struct dl *dl;
5f8097be
NB
3230 unsigned long long pos = 0;
3231 unsigned long long maxsize;
3232 struct extent *e;
3233 int i;
a322f70c 3234 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N
3235 if (level == LEVEL_CONTAINER) {
3236 if (verbose)
e7b84f9d 3237 pr_err("DDF cannot create a container within an container\n");
a322f70c 3238 return 0;
b42f577a 3239 }
a322f70c
DW
3240 /* We must have the container info already read in. */
3241 if (!ddf)
3242 return 0;
3243
5f8097be
NB
3244 if (!dev) {
3245 /* General test: make sure there is space for
3246 * 'raiddisks' device extents of size 'size'.
3247 */
3248 unsigned long long minsize = size;
3249 int dcnt = 0;
3250 if (minsize == 0)
3251 minsize = 8;
3252 for (dl = ddf->dlist; dl ; dl = dl->next)
3253 {
3254 int found = 0;
7e1432fb 3255 pos = 0;
5f8097be
NB
3256
3257 i = 0;
3258 e = get_extents(ddf, dl);
3259 if (!e) continue;
3260 do {
3261 unsigned long long esize;
3262 esize = e[i].start - pos;
3263 if (esize >= minsize)
3264 found = 1;
3265 pos = e[i].start + e[i].size;
3266 i++;
3267 } while (e[i-1].size);
3268 if (found)
3269 dcnt++;
3270 free(e);
3271 }
3272 if (dcnt < raiddisks) {
2c514b71 3273 if (verbose)
e7b84f9d
N
3274 pr_err("ddf: Not enough devices with "
3275 "space for this array (%d < %d)\n",
3276 dcnt, raiddisks);
5f8097be
NB
3277 return 0;
3278 }
3279 return 1;
3280 }
a322f70c
DW
3281 /* This device must be a member of the set */
3282 if (stat(dev, &stb) < 0)
3283 return 0;
3284 if ((S_IFMT & stb.st_mode) != S_IFBLK)
3285 return 0;
3286 for (dl = ddf->dlist ; dl ; dl = dl->next) {
f21e18ca
N
3287 if (dl->major == (int)major(stb.st_rdev) &&
3288 dl->minor == (int)minor(stb.st_rdev))
a322f70c
DW
3289 break;
3290 }
5f8097be 3291 if (!dl) {
2c514b71 3292 if (verbose)
e7b84f9d 3293 pr_err("ddf: %s is not in the "
613b0d17
N
3294 "same DDF set\n",
3295 dev);
5f8097be
NB
3296 return 0;
3297 }
3298 e = get_extents(ddf, dl);
3299 maxsize = 0;
3300 i = 0;
3301 if (e) do {
613b0d17
N
3302 unsigned long long esize;
3303 esize = e[i].start - pos;
3304 if (esize >= maxsize)
3305 maxsize = esize;
3306 pos = e[i].start + e[i].size;
3307 i++;
3308 } while (e[i-1].size);
5f8097be 3309 *freesize = maxsize;
a322f70c
DW
3310 // FIXME here I am
3311
3312 return 1;
3313}
59e36268 3314
a322f70c 3315static int load_super_ddf_all(struct supertype *st, int fd,
e1902a7b 3316 void **sbp, char *devname)
a322f70c
DW
3317{
3318 struct mdinfo *sra;
3319 struct ddf_super *super;
3320 struct mdinfo *sd, *best = NULL;
3321 int bestseq = 0;
3322 int seq;
3323 char nm[20];
3324 int dfd;
3325
b526e52d 3326 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
a322f70c
DW
3327 if (!sra)
3328 return 1;
3329 if (sra->array.major_version != -1 ||
3330 sra->array.minor_version != -2 ||
3331 strcmp(sra->text_version, "ddf") != 0)
3332 return 1;
3333
6416d527 3334 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 3335 return 1;
a2349791 3336 memset(super, 0, sizeof(*super));
a322f70c
DW
3337
3338 /* first, try each device, and choose the best ddf */
3339 for (sd = sra->devs ; sd ; sd = sd->next) {
3340 int rv;
3341 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB
3342 dfd = dev_open(nm, O_RDONLY);
3343 if (dfd < 0)
a322f70c
DW
3344 return 2;
3345 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 3346 close(dfd);
a322f70c
DW
3347 if (rv == 0) {
3348 seq = __be32_to_cpu(super->active->seq);
3349 if (super->active->openflag)
3350 seq--;
3351 if (!best || seq > bestseq) {
3352 bestseq = seq;
3353 best = sd;
3354 }
3355 }
3356 }
3357 if (!best)
3358 return 1;
3359 /* OK, load this ddf */
3360 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
3361 dfd = dev_open(nm, O_RDONLY);
7a7cc504 3362 if (dfd < 0)
a322f70c
DW
3363 return 1;
3364 load_ddf_headers(dfd, super, NULL);
3365 load_ddf_global(dfd, super, NULL);
3366 close(dfd);
3367 /* Now we need the device-local bits */
3368 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW
3369 int rv;
3370
a322f70c 3371 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
e1902a7b 3372 dfd = dev_open(nm, O_RDWR);
7a7cc504 3373 if (dfd < 0)
a322f70c 3374 return 2;
3d2c4fc7
DW
3375 rv = load_ddf_headers(dfd, super, NULL);
3376 if (rv == 0)
e1902a7b 3377 rv = load_ddf_local(dfd, super, NULL, 1);
3d2c4fc7
DW
3378 if (rv)
3379 return 1;
a322f70c 3380 }
33414a01 3381
a322f70c
DW
3382 *sbp = super;
3383 if (st->ss == NULL) {
78e44928 3384 st->ss = &super_ddf;
a322f70c
DW
3385 st->minor_version = 0;
3386 st->max_devs = 512;
3387 }
4dd2df09 3388 strcpy(st->container_devnm, fd2devnm(fd));
a322f70c
DW
3389 return 0;
3390}
2b959fbf
N
3391
3392static int load_container_ddf(struct supertype *st, int fd,
3393 char *devname)
3394{
3395 return load_super_ddf_all(st, fd, &st->sb, devname);
3396}
3397
0e600426 3398#endif /* MDASSEMBLE */
a322f70c 3399
a5c7adb3 3400static int check_secondary(const struct vcl *vc)
3401{
3402 const struct vd_config *conf = &vc->conf;
3403 int i;
3404
3405 /* The only DDF secondary RAID level md can support is
3406 * RAID 10, if the stripe sizes and Basic volume sizes
3407 * are all equal.
3408 * Other configurations could in theory be supported by exposing
3409 * the BVDs to user space and using device mapper for the secondary
3410 * mapping. So far we don't support that.
3411 */
3412
3413 __u64 sec_elements[4] = {0, 0, 0, 0};
3414#define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63)))
3415#define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0)
3416
3417 if (vc->other_bvds == NULL) {
3418 pr_err("No BVDs for secondary RAID found\n");
3419 return -1;
3420 }
3421 if (conf->prl != DDF_RAID1) {
3422 pr_err("Secondary RAID level only supported for mirrored BVD\n");
3423 return -1;
3424 }
3425 if (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED) {
3426 pr_err("Secondary RAID level %d is unsupported\n",
3427 conf->srl);
3428 return -1;
3429 }
3430 __set_sec_seen(conf->sec_elmnt_seq);
3431 for (i = 0; i < conf->sec_elmnt_count-1; i++) {
3432 const struct vd_config *bvd = vc->other_bvds[i];
3c48f7be 3433 if (bvd->sec_elmnt_seq == DDF_UNUSED_BVD)
c98567ba 3434 continue;
a5c7adb3 3435 if (bvd->srl != conf->srl) {
3436 pr_err("Inconsistent secondary RAID level across BVDs\n");
3437 return -1;
3438 }
3439 if (bvd->prl != conf->prl) {
3440 pr_err("Different RAID levels for BVDs are unsupported\n");
3441 return -1;
3442 }
3443 if (bvd->prim_elmnt_count != conf->prim_elmnt_count) {
3444 pr_err("All BVDs must have the same number of primary elements\n");
3445 return -1;
3446 }
3447 if (bvd->chunk_shift != conf->chunk_shift) {
3448 pr_err("Different strip sizes for BVDs are unsupported\n");
3449 return -1;
3450 }
3451 if (bvd->array_blocks != conf->array_blocks) {
3452 pr_err("Different BVD sizes are unsupported\n");
3453 return -1;
3454 }
3455 __set_sec_seen(bvd->sec_elmnt_seq);
3456 }
3457 for (i = 0; i < conf->sec_elmnt_count; i++) {
3458 if (!__was_sec_seen(i)) {
3459 pr_err("BVD %d is missing\n", i);
3460 return -1;
3461 }
3462 }
3463 return 0;
3464}
3465
8a38db86 3466static unsigned int get_pd_index_from_refnum(const struct vcl *vc,
4e587018 3467 __u32 refnum, unsigned int nmax,
3468 const struct vd_config **bvd,
3469 unsigned int *idx)
8a38db86 3470{
4e587018 3471 unsigned int i, j, n, sec, cnt;
3472
3473 cnt = __be16_to_cpu(vc->conf.prim_elmnt_count);
3474 sec = (vc->conf.sec_elmnt_count == 1 ? 0 : vc->conf.sec_elmnt_seq);
3475
3476 for (i = 0, j = 0 ; i < nmax ; i++) {
3477 /* j counts valid entries for this BVD */
3478 if (vc->conf.phys_refnum[i] != 0xffffffff)
3479 j++;
3480 if (vc->conf.phys_refnum[i] == refnum) {
3481 *bvd = &vc->conf;
3482 *idx = i;
3483 return sec * cnt + j - 1;
3484 }
3485 }
3486 if (vc->other_bvds == NULL)
3487 goto bad;
3488
3489 for (n = 1; n < vc->conf.sec_elmnt_count; n++) {
3490 struct vd_config *vd = vc->other_bvds[n-1];
4e587018 3491 sec = vd->sec_elmnt_seq;
3c48f7be 3492 if (sec == DDF_UNUSED_BVD)
3493 continue;
4e587018 3494 for (i = 0, j = 0 ; i < nmax ; i++) {
3495 if (vd->phys_refnum[i] != 0xffffffff)
3496 j++;
3497 if (vd->phys_refnum[i] == refnum) {
3498 *bvd = vd;
3499 *idx = i;
3500 return sec * cnt + j - 1;
3501 }
3502 }
3503 }
3504bad:
3505 *bvd = NULL;
d6e7b083 3506 return DDF_NOTFOUND;
8a38db86 3507}
3508
00bbdbda 3509static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray)
598f0d58
NB
3510{
3511 /* Given a container loaded by load_super_ddf_all,
3512 * extract information about all the arrays into
3513 * an mdinfo tree.
3514 *
3515 * For each vcl in conflist: create an mdinfo, fill it in,
3516 * then look for matching devices (phys_refnum) in dlist
3517 * and create appropriate device mdinfo.
3518 */
3519 struct ddf_super *ddf = st->sb;
3520 struct mdinfo *rest = NULL;
3521 struct vcl *vc;
3522
3523 for (vc = ddf->conflist ; vc ; vc=vc->next)
3524 {
f21e18ca
N
3525 unsigned int i;
3526 unsigned int j;
598f0d58 3527 struct mdinfo *this;
00bbdbda 3528 char *ep;
90fa1a29 3529 __u32 *cptr;
8a38db86 3530 unsigned int pd;
00bbdbda
N
3531
3532 if (subarray &&
3533 (strtoul(subarray, &ep, 10) != vc->vcnum ||
3534 *ep != '\0'))
3535 continue;
3536
a5c7adb3 3537 if (vc->conf.sec_elmnt_count > 1) {
3538 if (check_secondary(vc) != 0)
3539 continue;
3540 }
3541
503975b9 3542 this = xcalloc(1, sizeof(*this));
598f0d58
NB
3543 this->next = rest;
3544 rest = this;
3545
8a2848a7 3546 if (layout_ddf2md(&vc->conf, &this->array))
3547 continue;
598f0d58 3548 this->array.md_minor = -1;
f35f2525
N
3549 this->array.major_version = -1;
3550 this->array.minor_version = -2;
90fa1a29
JS
3551 cptr = (__u32 *)(vc->conf.guid + 16);
3552 this->array.ctime = DECADE + __be32_to_cpu(*cptr);
598f0d58
NB
3553 this->array.utime = DECADE +
3554 __be32_to_cpu(vc->conf.timestamp);
3555 this->array.chunk_size = 512 << vc->conf.chunk_shift;
3556
59e36268 3557 i = vc->vcnum;
7a7cc504
NB
3558 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
3559 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 3560 DDF_init_full) {
598f0d58 3561 this->array.state = 0;
ed9d66aa
NB
3562 this->resync_start = 0;
3563 } else {
598f0d58 3564 this->array.state = 1;
b7528a20 3565 this->resync_start = MaxSector;
ed9d66aa 3566 }
db42fa9b
N
3567 memcpy(this->name, ddf->virt->entries[i].name, 16);
3568 this->name[16]=0;
3569 for(j=0; j<16; j++)
3570 if (this->name[j] == ' ')
3571 this->name[j] = 0;
598f0d58
NB
3572
3573 memset(this->uuid, 0, sizeof(this->uuid));
3574 this->component_size = __be64_to_cpu(vc->conf.blocks);
3575 this->array.size = this->component_size / 2;
5f2aace8 3576 this->container_member = i;
598f0d58 3577
c5afc314
N
3578 ddf->currentconf = vc;
3579 uuid_from_super_ddf(st, this->uuid);
3580 ddf->currentconf = NULL;
3581
60f18132 3582 sprintf(this->text_version, "/%s/%d",
4dd2df09 3583 st->container_devnm, this->container_member);
60f18132 3584
8a38db86 3585 for (pd = 0; pd < __be16_to_cpu(ddf->phys->used_pdes); pd++) {
598f0d58
NB
3586 struct mdinfo *dev;
3587 struct dl *d;
4e587018 3588 const struct vd_config *bvd;
3589 unsigned int iphys;
fa033bec 3590 int stt;
598f0d58 3591
8a38db86 3592 if (ddf->phys->entries[pd].refnum == 0xFFFFFFFF)
bc17324f 3593 continue;
0cf5ef67
N
3594
3595 stt = __be16_to_cpu(ddf->phys->entries[pd].state);
fa033bec
N
3596 if ((stt & (DDF_Online|DDF_Failed|DDF_Rebuilding))
3597 != DDF_Online)
3598 continue;
3599
8a38db86 3600 i = get_pd_index_from_refnum(
4e587018 3601 vc, ddf->phys->entries[pd].refnum,
3602 ddf->mppe, &bvd, &iphys);
d6e7b083 3603 if (i == DDF_NOTFOUND)
8a38db86 3604 continue;
3605
fa033bec 3606 this->array.working_disks++;
bc17324f 3607
0cf5ef67 3608 for (d = ddf->dlist; d ; d=d->next)
8a38db86 3609 if (d->disk.refnum ==
3610 ddf->phys->entries[pd].refnum)
0cf5ef67
N
3611 break;
3612 if (d == NULL)
3613 /* Haven't found that one yet, maybe there are others */
3614 continue;
3615
503975b9 3616 dev = xcalloc(1, sizeof(*dev));
598f0d58
NB
3617 dev->next = this->devs;
3618 this->devs = dev;
3619
3620 dev->disk.number = __be32_to_cpu(d->disk.refnum);
3621 dev->disk.major = d->major;
3622 dev->disk.minor = d->minor;
3623 dev->disk.raid_disk = i;
3624 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
d23534e4 3625 dev->recovery_start = MaxSector;
598f0d58 3626
120f7677 3627 dev->events = __be32_to_cpu(ddf->primary.seq);
57a66662 3628 dev->data_offset =
3629 __be64_to_cpu(LBA_OFFSET(ddf, bvd)[iphys]);
4e587018 3630 dev->component_size = __be64_to_cpu(bvd->blocks);
598f0d58
NB
3631 if (d->devname)
3632 strcpy(dev->name, d->devname);
3633 }
3634 }
3635 return rest;
3636}
3637
955e9ea1 3638static int store_super_ddf(struct supertype *st, int fd)
a322f70c 3639{
955e9ea1 3640 struct ddf_super *ddf = st->sb;
a322f70c 3641 unsigned long long dsize;
6416d527 3642 void *buf;
3d2c4fc7 3643 int rc;
a322f70c 3644
955e9ea1
DW
3645 if (!ddf)
3646 return 1;
3647
a322f70c
DW
3648 if (!get_dev_size(fd, NULL, &dsize))
3649 return 1;
3650
dbf98368 3651 if (ddf->dlist || ddf->conflist) {
3652 struct stat sta;
3653 struct dl *dl;
3654 int ofd, ret;
3655
3656 if (fstat(fd, &sta) == -1 || !S_ISBLK(sta.st_mode)) {
3657 pr_err("%s: file descriptor for invalid device\n",
3658 __func__);
3659 return 1;
3660 }
3661 for (dl = ddf->dlist; dl; dl = dl->next)
3662 if (dl->major == (int)major(sta.st_rdev) &&
3663 dl->minor == (int)minor(sta.st_rdev))
3664 break;
3665 if (!dl) {
3666 pr_err("%s: couldn't find disk %d/%d\n", __func__,
3667 (int)major(sta.st_rdev),
3668 (int)minor(sta.st_rdev));
3669 return 1;
3670 }
3671 /*
3672 For DDF, writing to just one disk makes no sense.
3673 We would run the risk of writing inconsistent meta data
3674 to the devices. So just call __write_init_super_ddf and
3675 write to all devices, including this one.
3676 Use the fd passed to this function, just in case dl->fd
3677 is invalid.
3678 */
3679 ofd = dl->fd;
3680 dl->fd = fd;
3681 ret = __write_init_super_ddf(st);
3682 dl->fd = ofd;
3683 return ret;
3684 }
3685
3d2c4fc7
DW
3686 if (posix_memalign(&buf, 512, 512) != 0)
3687 return 1;
6416d527
NB
3688 memset(buf, 0, 512);
3689
a322f70c 3690 lseek64(fd, dsize-512, 0);
3d2c4fc7 3691 rc = write(fd, buf, 512);
6416d527 3692 free(buf);
3d2c4fc7
DW
3693 if (rc < 0)
3694 return 1;
a322f70c
DW
3695 return 0;
3696}
3697
a19c88b8
NB
3698static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3699{
3700 /*
3701 * return:
3702 * 0 same, or first was empty, and second was copied
3703 * 1 second had wrong number
3704 * 2 wrong uuid
3705 * 3 wrong other info
3706 */
3707 struct ddf_super *first = st->sb;
3708 struct ddf_super *second = tst->sb;
4eefd651 3709 struct dl *dl1, *dl2;
3710 struct vcl *vl1, *vl2;
2d210697 3711 unsigned int max_vds, max_pds, pd, vd;
a19c88b8
NB
3712
3713 if (!first) {
3714 st->sb = tst->sb;
3715 tst->sb = NULL;
3716 return 0;
3717 }
3718
3719 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3720 return 2;
3721
2d210697 3722 if (first->anchor.seq != second->anchor.seq) {
3723 dprintf("%s: sequence number mismatch %u/%u\n", __func__,
3724 __be32_to_cpu(first->anchor.seq),
3725 __be32_to_cpu(second->anchor.seq));
3726 return 3;
3727 }
3728 if (first->max_part != second->max_part ||
3729 first->phys->used_pdes != second->phys->used_pdes ||
3730 first->virt->populated_vdes != second->virt->populated_vdes) {
3731 dprintf("%s: PD/VD number mismatch\n", __func__);
3732 return 3;
3733 }
3734
3735 max_pds = __be16_to_cpu(first->phys->used_pdes);
3736 for (dl2 = second->dlist; dl2; dl2 = dl2->next) {
3737 for (pd = 0; pd < max_pds; pd++)
3738 if (first->phys->entries[pd].refnum == dl2->disk.refnum)
3739 break;
3740 if (pd == max_pds) {
3741 dprintf("%s: no match for disk %08x\n", __func__,
3742 __be32_to_cpu(dl2->disk.refnum));
3743 return 3;
3744 }
3745 }
3746
3747 max_vds = __be16_to_cpu(first->active->max_vd_entries);
3748 for (vl2 = second->conflist; vl2; vl2 = vl2->next) {
3749 if (vl2->conf.magic != DDF_VD_CONF_MAGIC)
3750 continue;
3751 for (vd = 0; vd < max_vds; vd++)
3752 if (!memcmp(first->virt->entries[vd].guid,
3753 vl2->conf.guid, DDF_GUID_LEN))
3754 break;
3755 if (vd == max_vds) {
3756 dprintf("%s: no match for VD config\n", __func__);
3757 return 3;
3758 }
3759 }
a19c88b8 3760 /* FIXME should I look at anything else? */
2d210697 3761
4eefd651 3762 /*
3763 At this point we are fairly sure that the meta data matches.
3764 But the new disk may contain additional local data.
3765 Add it to the super block.
3766 */
3767 for (vl2 = second->conflist; vl2; vl2 = vl2->next) {
3768 for (vl1 = first->conflist; vl1; vl1 = vl1->next)
3769 if (!memcmp(vl1->conf.guid, vl2->conf.guid,
3770 DDF_GUID_LEN))
3771 break;
3772 if (vl1) {
3773 if (vl1->other_bvds != NULL &&
3774 vl1->conf.sec_elmnt_seq !=
3775 vl2->conf.sec_elmnt_seq) {
3776 dprintf("%s: adding BVD %u\n", __func__,
3777 vl2->conf.sec_elmnt_seq);
3778 add_other_bvd(vl1, &vl2->conf,
3779 first->conf_rec_len*512);
3780 }
3781 continue;
3782 }
3783
3784 if (posix_memalign((void **)&vl1, 512,
3785 (first->conf_rec_len*512 +
3786 offsetof(struct vcl, conf))) != 0) {
3787 pr_err("%s could not allocate vcl buf\n",
3788 __func__);
3789 return 3;
3790 }
3791
3792 vl1->next = first->conflist;
3793 vl1->block_sizes = NULL;
4eefd651 3794 memcpy(&vl1->conf, &vl2->conf, first->conf_rec_len*512);
3c48f7be 3795 if (alloc_other_bvds(first, vl1) != 0) {
3796 pr_err("%s could not allocate other bvds\n",
3797 __func__);
3798 free(vl1);
3799 return 3;
3800 }
4eefd651 3801 for (vd = 0; vd < max_vds; vd++)
3802 if (!memcmp(first->virt->entries[vd].guid,
3803 vl1->conf.guid, DDF_GUID_LEN))
3804 break;
3805 vl1->vcnum = vd;
3806 dprintf("%s: added config for VD %u\n", __func__, vl1->vcnum);
3807 first->conflist = vl1;
3808 }
3809
3810 for (dl2 = second->dlist; dl2; dl2 = dl2->next) {
3811 for (dl1 = first->dlist; dl1; dl1 = dl1->next)
3812 if (dl1->disk.refnum == dl2->disk.refnum)
3813 break;
3814 if (dl1)
3815 continue;
3816
3817 if (posix_memalign((void **)&dl1, 512,
3818 sizeof(*dl1) + (first->max_part) * sizeof(dl1->vlist[0]))
3819 != 0) {
3820 pr_err("%s could not allocate disk info buffer\n",
3821 __func__);
3822 return 3;
3823 }
3824 memcpy(dl1, dl2, sizeof(*dl1));
3825 dl1->mdupdate = NULL;
3826 dl1->next = first->dlist;
3827 dl1->fd = -1;
3828 for (pd = 0; pd < max_pds; pd++)
3829 if (first->phys->entries[pd].refnum == dl1->disk.refnum)
3830 break;
3831 dl1->pdnum = pd;
3832 if (dl2->spare) {
3833 if (posix_memalign((void **)&dl1->spare, 512,
3834 first->conf_rec_len*512) != 0) {
3835 pr_err("%s could not allocate spare info buf\n",
3836 __func__);
3837 return 3;
3838 }
3839 memcpy(dl1->spare, dl2->spare, first->conf_rec_len*512);
3840 }
3841 for (vd = 0 ; vd < first->max_part ; vd++) {
3842 if (!dl2->vlist[vd]) {
3843 dl1->vlist[vd] = NULL;
3844 continue;
3845 }
3846 for (vl1 = first->conflist; vl1; vl1 = vl1->next) {
3847 if (!memcmp(vl1->conf.guid,
3848 dl2->vlist[vd]->conf.guid,
3849 DDF_GUID_LEN))
3850 break;
3851 dl1->vlist[vd] = vl1;
3852 }
3853 }
3854 first->dlist = dl1;
3855 dprintf("%s: added disk %d: %08x\n", __func__, dl1->pdnum,
3856 dl1->disk.refnum);
3857 }
3858
a19c88b8
NB
3859 return 0;
3860}
3861
0e600426 3862#ifndef MDASSEMBLE
4e5528c6
NB
3863/*
3864 * A new array 'a' has been started which claims to be instance 'inst'
3865 * within container 'c'.
3866 * We need to confirm that the array matches the metadata in 'c' so
3867 * that we don't corrupt any metadata.
3868 */
cba0191b 3869static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3870{
a2aa439e 3871 struct ddf_super *ddf = c->sb;
3872 int n = atoi(inst);
fb9d0acb 3873 if (all_ff(ddf->virt->entries[n].guid)) {
3874 pr_err("%s: subarray %d doesn't exist\n", __func__, n);
a2aa439e 3875 return -ENODEV;
3876 }
3877 dprintf("ddf: open_new %d\n", n);
3878 a->info.container_member = n;
549e9569
NB
3879 return 0;
3880}
3881
4e5528c6
NB
3882/*
3883 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3884 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB
3885 * clean up to the point (in sectors). If that cannot be recorded in the
3886 * metadata, then leave it as dirty.
3887 *
3888 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3889 * !global! virtual_disk.virtual_entry structure.
3890 */
01f157d7 3891static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3892{
4e5528c6
NB
3893 struct ddf_super *ddf = a->container->sb;
3894 int inst = a->info.container_member;
18a2f463 3895 int old = ddf->virt->entries[inst].state;
01f157d7
N
3896 if (consistent == 2) {
3897 /* Should check if a recovery should be started FIXME */
3898 consistent = 1;
b7941fd6 3899 if (!is_resync_complete(&a->info))
01f157d7
N
3900 consistent = 0;
3901 }
ed9d66aa
NB
3902 if (consistent)
3903 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3904 else
4e5528c6 3905 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463 3906 if (old != ddf->virt->entries[inst].state)
7d5a7ff3 3907 ddf_set_updates_pending(ddf);
18a2f463
NB
3908
3909 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3910 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
b7941fd6 3911 if (is_resync_complete(&a->info))
ed9d66aa 3912 ddf->virt->entries[inst].init_state |= DDF_init_full;
b7941fd6 3913 else if (a->info.resync_start == 0)
ed9d66aa 3914 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3915 else
ed9d66aa 3916 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463 3917 if (old != ddf->virt->entries[inst].init_state)
7d5a7ff3 3918 ddf_set_updates_pending(ddf);
ed9d66aa 3919
2c514b71 3920 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
b7941fd6 3921 a->info.resync_start);
01f157d7 3922 return consistent;
fd7cde1b
DW
3923}
3924
5ec636b7 3925static int get_bvd_state(const struct ddf_super *ddf,
3926 const struct vd_config *vc)
3927{
3928 unsigned int i, n_bvd, working = 0;
3929 unsigned int n_prim = __be16_to_cpu(vc->prim_elmnt_count);
3930 int pd, st, state;
3931 for (i = 0; i < n_prim; i++) {
3932 if (!find_index_in_bvd(ddf, vc, i, &n_bvd))
3933 continue;
3934 pd = find_phys(ddf, vc->phys_refnum[n_bvd]);
3935 if (pd < 0)
3936 continue;
3937 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3938 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
3939 == DDF_Online)
3940 working++;
3941 }
3942
3943 state = DDF_state_degraded;
3944 if (working == n_prim)
3945 state = DDF_state_optimal;
3946 else
3947 switch (vc->prl) {
3948 case DDF_RAID0:
3949 case DDF_CONCAT:
3950 case DDF_JBOD:
3951 state = DDF_state_failed;
3952 break;
3953 case DDF_RAID1:
3954 if (working == 0)
3955 state = DDF_state_failed;
3956 else if (working >= 2)
3957 state = DDF_state_part_optimal;
3958 break;
3959 case DDF_RAID4:
3960 case DDF_RAID5:
3961 if (working < n_prim - 1)
3962 state = DDF_state_failed;
3963 break;
3964 case DDF_RAID6:
3965 if (working < n_prim - 2)
3966 state = DDF_state_failed;
3967 else if (working == n_prim - 1)
3968 state = DDF_state_part_optimal;
3969 break;
3970 }
3971 return state;
3972}
3973
0777d17d 3974static int secondary_state(int state, int other, int seclevel)
3975{
3976 if (state == DDF_state_optimal && other == DDF_state_optimal)
3977 return DDF_state_optimal;
3978 if (seclevel == DDF_2MIRRORED) {
3979 if (state == DDF_state_optimal || other == DDF_state_optimal)
3980 return DDF_state_part_optimal;
3981 if (state == DDF_state_failed && other == DDF_state_failed)
3982 return DDF_state_failed;
3983 return DDF_state_degraded;
3984 } else {
3985 if (state == DDF_state_failed || other == DDF_state_failed)
3986 return DDF_state_failed;
3987 if (state == DDF_state_degraded || other == DDF_state_degraded)
3988 return DDF_state_degraded;
3989 return DDF_state_part_optimal;
3990 }
3991}
3992
3993static int get_svd_state(const struct ddf_super *ddf, const struct vcl *vcl)
3994{
3995 int state = get_bvd_state(ddf, &vcl->conf);
3996 unsigned int i;
3997 for (i = 1; i < vcl->conf.sec_elmnt_count; i++) {
3998 state = secondary_state(
3999 state,
4000 get_bvd_state(ddf, vcl->other_bvds[i-1]),
4001 vcl->conf.srl);
4002 }
4003 return state;
4004}
4005
7a7cc504
NB
4006/*
4007 * The state of each disk is stored in the global phys_disk structure
4008 * in phys_disk.entries[n].state.
4009 * This makes various combinations awkward.
4010 * - When a device fails in any array, it must be failed in all arrays
4011 * that include a part of this device.
4012 * - When a component is rebuilding, we cannot include it officially in the
4013 * array unless this is the only array that uses the device.
4014 *
4015 * So: when transitioning:
4016 * Online -> failed, just set failed flag. monitor will propagate
4017 * spare -> online, the device might need to be added to the array.
4018 * spare -> failed, just set failed. Don't worry if in array or not.
4019 */
8d45d196 4020static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 4021{
7a7cc504 4022 struct ddf_super *ddf = a->container->sb;
baba3f4e 4023 unsigned int inst = a->info.container_member, n_bvd;
4024 struct vcl *vcl;
4025 struct vd_config *vc = find_vdcr(ddf, inst, (unsigned int)n,
4026 &n_bvd, &vcl);
4027 int pd;
e1316fab
N
4028 struct mdinfo *mdi;
4029 struct dl *dl;
7a7cc504
NB
4030
4031 if (vc == NULL) {
2c514b71 4032 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB
4033 return;
4034 }
e1316fab
N
4035 /* Find the matching slot in 'info'. */
4036 for (mdi = a->info.devs; mdi; mdi = mdi->next)
4037 if (mdi->disk.raid_disk == n)
4038 break;
4039 if (!mdi)
4040 return;
4041
4042 /* and find the 'dl' entry corresponding to that. */
4043 for (dl = ddf->dlist; dl; dl = dl->next)
77632af9
N
4044 if (mdi->state_fd >= 0 &&
4045 mdi->disk.major == dl->major &&
e1316fab
N
4046 mdi->disk.minor == dl->minor)
4047 break;
4048 if (!dl)
4049 return;
4050
baba3f4e 4051 pd = find_phys(ddf, vc->phys_refnum[n_bvd]);
e1316fab
N
4052 if (pd < 0 || pd != dl->pdnum) {
4053 /* disk doesn't currently exist or has changed.
4054 * If it is now in_sync, insert it. */
baba3f4e 4055 dprintf("%s: phys disk not found for %d: %d/%d ref %08x\n",
4056 __func__, dl->pdnum, dl->major, dl->minor,
4057 dl->disk.refnum);
4058 dprintf("%s: array %u disk %u ref %08x pd %d\n",
4059 __func__, inst, n_bvd, vc->phys_refnum[n_bvd], pd);
7a7cc504 4060 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
baba3f4e 4061 pd = dl->pdnum; /* FIXME: is this really correct ? */
4062 vc->phys_refnum[n_bvd] = dl->disk.refnum;
57a66662 4063 LBA_OFFSET(ddf, vc)[n_bvd] =
4064 __cpu_to_be64(mdi->data_offset);
e1316fab
N
4065 ddf->phys->entries[pd].type &=
4066 ~__cpu_to_be16(DDF_Global_Spare);
4067 ddf->phys->entries[pd].type |=
4068 __cpu_to_be16(DDF_Active_in_VD);
7d5a7ff3 4069 ddf_set_updates_pending(ddf);
7a7cc504
NB
4070 }
4071 } else {
18a2f463 4072 int old = ddf->phys->entries[pd].state;
7a7cc504
NB
4073 if (state & DS_FAULTY)
4074 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
4075 if (state & DS_INSYNC) {
4076 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
4077 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
4078 }
18a2f463 4079 if (old != ddf->phys->entries[pd].state)
7d5a7ff3 4080 ddf_set_updates_pending(ddf);
7a7cc504
NB
4081 }
4082
2c514b71 4083 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 4084
7a7cc504
NB
4085 /* Now we need to check the state of the array and update
4086 * virtual_disk.entries[n].state.
4087 * It needs to be one of "optimal", "degraded", "failed".
4088 * I don't understand 'deleted' or 'missing'.
4089 */
0777d17d 4090 state = get_svd_state(ddf, vcl);
7a7cc504 4091
18a2f463
NB
4092 if (ddf->virt->entries[inst].state !=
4093 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
4094 | state)) {
4095
4096 ddf->virt->entries[inst].state =
4097 (ddf->virt->entries[inst].state & ~DDF_state_mask)
4098 | state;
7d5a7ff3 4099 ddf_set_updates_pending(ddf);
18a2f463 4100 }
7a7cc504 4101
549e9569
NB
4102}
4103
2e735d19 4104static void ddf_sync_metadata(struct supertype *st)
549e9569 4105{
7a7cc504
NB
4106
4107 /*
4108 * Write all data to all devices.
4109 * Later, we might be able to track whether only local changes
4110 * have been made, or whether any global data has been changed,
4111 * but ddf is sufficiently weird that it probably always
4112 * changes global data ....
4113 */
18a2f463
NB
4114 struct ddf_super *ddf = st->sb;
4115 if (!ddf->updates_pending)
4116 return;
4117 ddf->updates_pending = 0;
1cc7f4fe 4118 __write_init_super_ddf(st);
2c514b71 4119 dprintf("ddf: sync_metadata\n");
549e9569
NB
4120}
4121
88c164f4
NB
4122static void ddf_process_update(struct supertype *st,
4123 struct metadata_update *update)
4124{
4125 /* Apply this update to the metadata.
4126 * The first 4 bytes are a DDF_*_MAGIC which guides
4127 * our actions.
4128 * Possible update are:
4129 * DDF_PHYS_RECORDS_MAGIC
4dd968cc
N
4130 * Add a new physical device or remove an old one.
4131 * Changes to this record only happen implicitly.
88c164f4
NB
4132 * used_pdes is the device number.
4133 * DDF_VIRT_RECORDS_MAGIC
4134 * Add a new VD. Possibly also change the 'access' bits.
4135 * populated_vdes is the entry number.
4136 * DDF_VD_CONF_MAGIC
4137 * New or updated VD. the VIRT_RECORD must already
4138 * exist. For an update, phys_refnum and lba_offset
4139 * (at least) are updated, and the VD_CONF must
4140 * be written to precisely those devices listed with
4141 * a phys_refnum.
4142 * DDF_SPARE_ASSIGN_MAGIC
4143 * replacement Spare Assignment Record... but for which device?
4144 *
4145 * So, e.g.:
4146 * - to create a new array, we send a VIRT_RECORD and
4147 * a VD_CONF. Then assemble and start the array.
4148 * - to activate a spare we send a VD_CONF to add the phys_refnum
4149 * and offset. This will also mark the spare as active with
4150 * a spare-assignment record.
4151 */
4152 struct ddf_super *ddf = st->sb;
4153 __u32 *magic = (__u32*)update->buf;
4154 struct phys_disk *pd;
4155 struct virtual_disk *vd;
4156 struct vd_config *vc;
4157 struct vcl *vcl;
4158 struct dl *dl;
f21e18ca
N
4159 unsigned int mppe;
4160 unsigned int ent;
c7079c84 4161 unsigned int pdnum, pd2;
88c164f4 4162
2c514b71 4163 dprintf("Process update %x\n", *magic);
7e1432fb 4164
88c164f4
NB
4165 switch (*magic) {
4166 case DDF_PHYS_RECORDS_MAGIC:
4167
4168 if (update->len != (sizeof(struct phys_disk) +
4169 sizeof(struct phys_disk_entry)))
4170 return;
4171 pd = (struct phys_disk*)update->buf;
4172
4173 ent = __be16_to_cpu(pd->used_pdes);
4174 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
4175 return;
4dd968cc
N
4176 if (pd->entries[0].state & __cpu_to_be16(DDF_Missing)) {
4177 struct dl **dlp;
4178 /* removing this disk. */
4179 ddf->phys->entries[ent].state |= __cpu_to_be16(DDF_Missing);
4180 for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) {
4181 struct dl *dl = *dlp;
4182 if (dl->pdnum == (signed)ent) {
4183 close(dl->fd);
4184 dl->fd = -1;
4185 /* FIXME this doesn't free
4186 * dl->devname */
4187 update->space = dl;
4188 *dlp = dl->next;
4189 break;
4190 }
4191 }
7d5a7ff3 4192 ddf_set_updates_pending(ddf);
4dd968cc
N
4193 return;
4194 }
88c164f4
NB
4195 if (!all_ff(ddf->phys->entries[ent].guid))
4196 return;
4197 ddf->phys->entries[ent] = pd->entries[0];
4198 ddf->phys->used_pdes = __cpu_to_be16(1 +
613b0d17 4199 __be16_to_cpu(ddf->phys->used_pdes));
7d5a7ff3 4200 ddf_set_updates_pending(ddf);
2cc2983d
N
4201 if (ddf->add_list) {
4202 struct active_array *a;
4203 struct dl *al = ddf->add_list;
4204 ddf->add_list = al->next;
4205
4206 al->next = ddf->dlist;
4207 ddf->dlist = al;
4208
4209 /* As a device has been added, we should check
4210 * for any degraded devices that might make
4211 * use of this spare */
4212 for (a = st->arrays ; a; a=a->next)
4213 a->check_degraded = 1;
4214 }
88c164f4
NB
4215 break;
4216
4217 case DDF_VIRT_RECORDS_MAGIC:
4218
4219 if (update->len != (sizeof(struct virtual_disk) +
4220 sizeof(struct virtual_entry)))
4221 return;
4222 vd = (struct virtual_disk*)update->buf;
4223
fb9d0acb 4224 ent = find_unused_vde(ddf);
4225 if (ent == DDF_NOTFOUND)
88c164f4
NB
4226 return;
4227 ddf->virt->entries[ent] = vd->entries[0];
4228 ddf->virt->populated_vdes = __cpu_to_be16(1 +
613b0d17 4229 __be16_to_cpu(ddf->virt->populated_vdes));
7d5a7ff3 4230 ddf_set_updates_pending(ddf);
88c164f4
NB
4231 break;
4232
4233 case DDF_VD_CONF_MAGIC:
2c514b71 4234 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB
4235
4236 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
f21e18ca 4237 if ((unsigned)update->len != ddf->conf_rec_len * 512)
88c164f4
NB
4238 return;
4239 vc = (struct vd_config*)update->buf;
4240 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
4241 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
4242 break;
2c514b71 4243 dprintf("vcl = %p\n", vcl);
88c164f4
NB
4244 if (vcl) {
4245 /* An update, just copy the phys_refnum and lba_offset
4246 * fields
4247 */
4248 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
4249 mppe * (sizeof(__u32) + sizeof(__u64)));
4250 } else {
4251 /* A new VD_CONF */
e6b9548d
DW
4252 if (!update->space)
4253 return;
88c164f4
NB
4254 vcl = update->space;
4255 update->space = NULL;
4256 vcl->next = ddf->conflist;
edd8d13c 4257 memcpy(&vcl->conf, vc, update->len);
fb9d0acb 4258 ent = find_vde_by_guid(ddf, vc->guid);
4259 if (ent == DDF_NOTFOUND)
4260 return;
4261 vcl->vcnum = ent;
88c164f4
NB
4262 ddf->conflist = vcl;
4263 }
c7079c84
N
4264 /* Set DDF_Transition on all Failed devices - to help
4265 * us detect those that are no longer in use
4266 */
4267 for (pdnum = 0; pdnum < __be16_to_cpu(ddf->phys->used_pdes); pdnum++)
4268 if (ddf->phys->entries[pdnum].state
4269 & __be16_to_cpu(DDF_Failed))
4270 ddf->phys->entries[pdnum].state
4271 |= __be16_to_cpu(DDF_Transition);
88c164f4
NB
4272 /* Now make sure vlist is correct for each dl. */
4273 for (dl = ddf->dlist; dl; dl = dl->next) {
f21e18ca
N
4274 unsigned int dn;
4275 unsigned int vn = 0;
8401644c 4276 int in_degraded = 0;
88c164f4
NB
4277 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
4278 for (dn=0; dn < ddf->mppe ; dn++)
4279 if (vcl->conf.phys_refnum[dn] ==
4280 dl->disk.refnum) {
8401644c 4281 int vstate;
2c514b71
NB
4282 dprintf("dev %d has %p at %d\n",
4283 dl->pdnum, vcl, vn);
c7079c84
N
4284 /* Clear the Transition flag */
4285 if (ddf->phys->entries[dl->pdnum].state
4286 & __be16_to_cpu(DDF_Failed))
4287 ddf->phys->entries[dl->pdnum].state &=
4288 ~__be16_to_cpu(DDF_Transition);
4289
88c164f4 4290 dl->vlist[vn++] = vcl;
8401644c
N
4291 vstate = ddf->virt->entries[vcl->vcnum].state
4292 & DDF_state_mask;
4293 if (vstate == DDF_state_degraded ||
4294 vstate == DDF_state_part_optimal)
4295 in_degraded = 1;
88c164f4
NB
4296 break;
4297 }
4298 while (vn < ddf->max_part)
4299 dl->vlist[vn++] = NULL;
7e1432fb
NB
4300 if (dl->vlist[0]) {
4301 ddf->phys->entries[dl->pdnum].type &=
4302 ~__cpu_to_be16(DDF_Global_Spare);
8401644c
N
4303 if (!(ddf->phys->entries[dl->pdnum].type &
4304 __cpu_to_be16(DDF_Active_in_VD))) {
613b0d17
N
4305 ddf->phys->entries[dl->pdnum].type |=
4306 __cpu_to_be16(DDF_Active_in_VD);
4307 if (in_degraded)
4308 ddf->phys->entries[dl->pdnum].state |=
4309 __cpu_to_be16(DDF_Rebuilding);
4310 }
7e1432fb
NB
4311 }
4312 if (dl->spare) {
4313 ddf->phys->entries[dl->pdnum].type &=
4314 ~__cpu_to_be16(DDF_Global_Spare);
4315 ddf->phys->entries[dl->pdnum].type |=
4316 __cpu_to_be16(DDF_Spare);
4317 }
4318 if (!dl->vlist[0] && !dl->spare) {
4319 ddf->phys->entries[dl->pdnum].type |=
4320 __cpu_to_be16(DDF_Global_Spare);
4321 ddf->phys->entries[dl->pdnum].type &=
4322 ~__cpu_to_be16(DDF_Spare |
4323 DDF_Active_in_VD);
4324 }
88c164f4 4325 }
c7079c84
N
4326
4327 /* Now remove any 'Failed' devices that are not part
4328 * of any VD. They will have the Transition flag set.
4329 * Once done, we need to update all dl->pdnum numbers.
4330 */
4331 pd2 = 0;
4332 for (pdnum = 0; pdnum < __be16_to_cpu(ddf->phys->used_pdes); pdnum++)
4333 if ((ddf->phys->entries[pdnum].state
4334 & __be16_to_cpu(DDF_Failed))
4335 && (ddf->phys->entries[pdnum].state
4336 & __be16_to_cpu(DDF_Transition)))
4337 /* skip this one */;
4338 else if (pdnum == pd2)
4339 pd2++;
4340 else {
4341 ddf->phys->entries[pd2] = ddf->phys->entries[pdnum];
4342 for (dl = ddf->dlist; dl; dl = dl->next)
4343 if (dl->pdnum == (int)pdnum)
4344 dl->pdnum = pd2;
4345 pd2++;
4346 }
4347 ddf->phys->used_pdes = __cpu_to_be16(pd2);
4348 while (pd2 < pdnum) {
4349 memset(ddf->phys->entries[pd2].guid, 0xff, DDF_GUID_LEN);
4350 pd2++;
4351 }
4352
7d5a7ff3 4353 ddf_set_updates_pending(ddf);
88c164f4
NB
4354 break;
4355 case DDF_SPARE_ASSIGN_MAGIC:
4356 default: break;
4357 }
4358}
4359
edd8d13c
NB
4360static void ddf_prepare_update(struct supertype *st,
4361 struct metadata_update *update)
4362{
4363 /* This update arrived at managemon.
4364 * We are about to pass it to monitor.
4365 * If a malloc is needed, do it here.
4366 */
4367 struct ddf_super *ddf = st->sb;
4368 __u32 *magic = (__u32*)update->buf;
4369 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 4370 if (posix_memalign(&update->space, 512,
613b0d17
N
4371 offsetof(struct vcl, conf)
4372 + ddf->conf_rec_len * 512) != 0)
e6b9548d 4373 update->space = NULL;
edd8d13c
NB
4374}
4375
7e1432fb
NB
4376/*
4377 * Check if the array 'a' is degraded but not failed.
4378 * If it is, find as many spares as are available and needed and
4379 * arrange for their inclusion.
4380 * We only choose devices which are not already in the array,
4381 * and prefer those with a spare-assignment to this array.
4382 * otherwise we choose global spares - assuming always that
4383 * there is enough room.
4384 * For each spare that we assign, we return an 'mdinfo' which
4385 * describes the position for the device in the array.
4386 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
4387 * the new phys_refnum and lba_offset values.
4388 *
4389 * Only worry about BVDs at the moment.
4390 */
4391static struct mdinfo *ddf_activate_spare(struct active_array *a,
4392 struct metadata_update **updates)
4393{
4394 int working = 0;
4395 struct mdinfo *d;
4396 struct ddf_super *ddf = a->container->sb;
4397 int global_ok = 0;
4398 struct mdinfo *rv = NULL;
4399 struct mdinfo *di;
4400 struct metadata_update *mu;
4401 struct dl *dl;
4402 int i;
baba3f4e 4403 struct vcl *vcl;
7e1432fb 4404 struct vd_config *vc;
baba3f4e 4405 unsigned int n_bvd;
7e1432fb 4406
7e1432fb
NB
4407 for (d = a->info.devs ; d ; d = d->next) {
4408 if ((d->curr_state & DS_FAULTY) &&
613b0d17 4409 d->state_fd >= 0)
7e1432fb
NB
4410 /* wait for Removal to happen */
4411 return NULL;
4412 if (d->state_fd >= 0)
4413 working ++;
4414 }
4415
2c514b71
NB
4416 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
4417 a->info.array.level);
7e1432fb
NB
4418 if (working == a->info.array.raid_disks)
4419 return NULL; /* array not degraded */
4420 switch (a->info.array.level) {
4421 case 1:
4422 if (working == 0)
4423 return NULL; /* failed */
4424 break;
4425 case 4:
4426 case 5:
4427 if (working < a->info.array.raid_disks - 1)
4428 return NULL; /* failed */
4429 break;
4430 case 6:
4431 if (working < a->info.array.raid_disks - 2)
4432 return NULL; /* failed */
4433 break;
4434 default: /* concat or stripe */
4435 return NULL; /* failed */
4436 }
4437
4438 /* For each slot, if it is not working, find a spare */
4439 dl = ddf->dlist;
4440 for (i = 0; i < a->info.array.raid_disks; i++) {
4441 for (d = a->info.devs ; d ; d = d->next)
4442 if (d->disk.raid_disk == i)
4443 break;
2c514b71 4444 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB
4445 if (d && (d->state_fd >= 0))
4446 continue;
4447
4448 /* OK, this device needs recovery. Find a spare */
4449 again:
4450 for ( ; dl ; dl = dl->next) {
4451 unsigned long long esize;
4452 unsigned long long pos;
4453 struct mdinfo *d2;
4454 int is_global = 0;
4455 int is_dedicated = 0;
4456 struct extent *ex;
f21e18ca 4457 unsigned int j;
7e1432fb
NB
4458 /* If in this array, skip */
4459 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
7590d562
N
4460 if (d2->state_fd >= 0 &&
4461 d2->disk.major == dl->major &&
7e1432fb 4462 d2->disk.minor == dl->minor) {
2c514b71 4463 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB
4464 break;
4465 }
4466 if (d2)
4467 continue;
4468 if (ddf->phys->entries[dl->pdnum].type &
4469 __cpu_to_be16(DDF_Spare)) {
4470 /* Check spare assign record */
4471 if (dl->spare) {
4472 if (dl->spare->type & DDF_spare_dedicated) {
4473 /* check spare_ents for guid */
4474 for (j = 0 ;
4475 j < __be16_to_cpu(dl->spare->populated);
4476 j++) {
4477 if (memcmp(dl->spare->spare_ents[j].guid,
4478 ddf->virt->entries[a->info.container_member].guid,
4479 DDF_GUID_LEN) == 0)
4480 is_dedicated = 1;
4481 }
4482 } else
4483 is_global = 1;
4484 }
4485 } else if (ddf->phys->entries[dl->pdnum].type &
4486 __cpu_to_be16(DDF_Global_Spare)) {
4487 is_global = 1;
e0e7aeaa
N
4488 } else if (!(ddf->phys->entries[dl->pdnum].state &
4489 __cpu_to_be16(DDF_Failed))) {
4490 /* we can possibly use some of this */
4491 is_global = 1;
7e1432fb
NB
4492 }
4493 if ( ! (is_dedicated ||
4494 (is_global && global_ok))) {
2c514b71 4495 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
613b0d17 4496 is_dedicated, is_global);
7e1432fb
NB
4497 continue;
4498 }
4499
4500 /* We are allowed to use this device - is there space?
4501 * We need a->info.component_size sectors */
4502 ex = get_extents(ddf, dl);
4503 if (!ex) {
2c514b71 4504 dprintf("cannot get extents\n");
7e1432fb
NB
4505 continue;
4506 }
4507 j = 0; pos = 0;
4508 esize = 0;
4509
4510 do {
4511 esize = ex[j].start - pos;
4512 if (esize >= a->info.component_size)
4513 break;
e5cc7d46
N
4514 pos = ex[j].start + ex[j].size;
4515 j++;
4516 } while (ex[j-1].size);
7e1432fb
NB
4517
4518 free(ex);
4519 if (esize < a->info.component_size) {
e5cc7d46
N
4520 dprintf("%x:%x has no room: %llu %llu\n",
4521 dl->major, dl->minor,
2c514b71 4522 esize, a->info.component_size);
7e1432fb
NB
4523 /* No room */
4524 continue;
4525 }
4526
4527 /* Cool, we have a device with some space at pos */
503975b9 4528 di = xcalloc(1, sizeof(*di));
7e1432fb
NB
4529 di->disk.number = i;
4530 di->disk.raid_disk = i;
4531 di->disk.major = dl->major;
4532 di->disk.minor = dl->minor;
4533 di->disk.state = 0;
d23534e4 4534 di->recovery_start = 0;
7e1432fb
NB
4535 di->data_offset = pos;
4536 di->component_size = a->info.component_size;
4537 di->container_member = dl->pdnum;
4538 di->next = rv;
4539 rv = di;
2c514b71
NB
4540 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
4541 i, pos);
7e1432fb
NB
4542
4543 break;
4544 }
4545 if (!dl && ! global_ok) {
4546 /* not enough dedicated spares, try global */
4547 global_ok = 1;
4548 dl = ddf->dlist;
4549 goto again;
4550 }
4551 }
4552
4553 if (!rv)
4554 /* No spares found */
4555 return rv;
4556 /* Now 'rv' has a list of devices to return.
4557 * Create a metadata_update record to update the
4558 * phys_refnum and lba_offset values
4559 */
503975b9
N
4560 mu = xmalloc(sizeof(*mu));
4561 if (posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
79244939
DW
4562 free(mu);
4563 mu = NULL;
4564 }
503975b9 4565 mu->buf = xmalloc(ddf->conf_rec_len * 512);
7590d562
N
4566 mu->len = ddf->conf_rec_len * 512;
4567 mu->space = NULL;
f50ae22e 4568 mu->space_list = NULL;
7e1432fb 4569 mu->next = *updates;
baba3f4e 4570 vc = find_vdcr(ddf, a->info.container_member, di->disk.raid_disk,
4571 &n_bvd, &vcl);
7e1432fb
NB
4572 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
4573
4574 vc = (struct vd_config*)mu->buf;
7e1432fb
NB
4575 for (di = rv ; di ; di = di->next) {
4576 vc->phys_refnum[di->disk.raid_disk] =
4577 ddf->phys->entries[dl->pdnum].refnum;
57a66662 4578 LBA_OFFSET(ddf, vc)[di->disk.raid_disk]
4579 = __cpu_to_be64(di->data_offset);
7e1432fb
NB
4580 }
4581 *updates = mu;
4582 return rv;
4583}
0e600426 4584#endif /* MDASSEMBLE */
7e1432fb 4585
b640a252
N
4586static int ddf_level_to_layout(int level)
4587{
4588 switch(level) {
4589 case 0:
4590 case 1:
4591 return 0;
4592 case 5:
4593 return ALGORITHM_LEFT_SYMMETRIC;
4594 case 6:
4595 return ALGORITHM_ROTATING_N_CONTINUE;
4596 case 10:
4597 return 0x102;
4598 default:
4599 return UnSet;
4600 }
4601}
4602
30f58b22
DW
4603static void default_geometry_ddf(struct supertype *st, int *level, int *layout, int *chunk)
4604{
4605 if (level && *level == UnSet)
4606 *level = LEVEL_CONTAINER;
4607
4608 if (level && layout && *layout == UnSet)
4609 *layout = ddf_level_to_layout(*level);
4610}
4611
a322f70c
DW
4612struct superswitch super_ddf = {
4613#ifndef MDASSEMBLE
4614 .examine_super = examine_super_ddf,
4615 .brief_examine_super = brief_examine_super_ddf,
4737ae25 4616 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 4617 .export_examine_super = export_examine_super_ddf,
a322f70c
DW
4618 .detail_super = detail_super_ddf,
4619 .brief_detail_super = brief_detail_super_ddf,
4620 .validate_geometry = validate_geometry_ddf,
78e44928 4621 .write_init_super = write_init_super_ddf,
0e600426 4622 .add_to_super = add_to_super_ddf,
4dd968cc 4623 .remove_from_super = remove_from_super_ddf,
2b959fbf 4624 .load_container = load_container_ddf,
74db60b0 4625 .copy_metadata = copy_metadata_ddf,
a322f70c
DW
4626#endif
4627 .match_home = match_home_ddf,
4628 .uuid_from_super= uuid_from_super_ddf,
4629 .getinfo_super = getinfo_super_ddf,
4630 .update_super = update_super_ddf,
4631
4632 .avail_size = avail_size_ddf,
4633
a19c88b8
NB
4634 .compare_super = compare_super_ddf,
4635
a322f70c 4636 .load_super = load_super_ddf,
ba7eb04f 4637 .init_super = init_super_ddf,
955e9ea1 4638 .store_super = store_super_ddf,
a322f70c
DW
4639 .free_super = free_super_ddf,
4640 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 4641 .container_content = container_content_ddf,
30f58b22 4642 .default_geometry = default_geometry_ddf,
a322f70c 4643
a322f70c 4644 .external = 1,
549e9569 4645
0e600426 4646#ifndef MDASSEMBLE
549e9569
NB
4647/* for mdmon */
4648 .open_new = ddf_open_new,
ed9d66aa 4649 .set_array_state= ddf_set_array_state,
549e9569
NB
4650 .set_disk = ddf_set_disk,
4651 .sync_metadata = ddf_sync_metadata,
88c164f4 4652 .process_update = ddf_process_update,
edd8d13c 4653 .prepare_update = ddf_prepare_update,
7e1432fb 4654 .activate_spare = ddf_activate_spare,
0e600426 4655#endif
4cce4069 4656 .name = "ddf",
a322f70c 4657};