]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-ddf.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Grow: fix recent breakage - lseek return status.
[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
47/* The DDF metadata handling.
48 * DDF metadata lives at the end of the device.
49 * The last 512 byte block provides an 'anchor' which is used to locate
50 * the rest of the metadata which usually lives immediately behind the anchor.
51 *
52 * Note:
53 * - all multibyte numeric fields are bigendian.
54 * - all strings are space padded.
55 *
56 */
57
58/* Primary Raid Level (PRL) */
59#define DDF_RAID0 0x00
60#define DDF_RAID1 0x01
61#define DDF_RAID3 0x03
62#define DDF_RAID4 0x04
63#define DDF_RAID5 0x05
64#define DDF_RAID1E 0x11
65#define DDF_JBOD 0x0f
66#define DDF_CONCAT 0x1f
67#define DDF_RAID5E 0x15
68#define DDF_RAID5EE 0x25
59e36268 69#define DDF_RAID6 0x06
a322f70c
DW		 70
71/* Raid Level Qualifier (RLQ) */
72#define DDF_RAID0_SIMPLE 0x00
73#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
74#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
75#define DDF_RAID3_0 0x00 /* parity in first extent */
76#define DDF_RAID3_N 0x01 /* parity in last extent */
77#define DDF_RAID4_0 0x00 /* parity in first extent */
78#define DDF_RAID4_N 0x01 /* parity in last extent */
79/* these apply to raid5e and raid5ee as well */
80#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
59e36268 81#define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
a322f70c
DW		 82#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
83#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
84
85#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
86#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
87
88/* Secondary RAID Level (SRL) */
89#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
90#define DDF_2MIRRORED 0x01
91#define DDF_2CONCAT 0x02
92#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
93
94/* Magic numbers */
95#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
96#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
97#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
98#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
99#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
100#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
101#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
102#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
103#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
104#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
105
106#define DDF_GUID_LEN 24
59e36268
NB		 107#define DDF_REVISION_0 "01.00.00"
108#define DDF_REVISION_2 "01.02.00"
a322f70c
DW		 109
110struct ddf_header {
88c164f4 111 __u32 magic; /* DDF_HEADER_MAGIC */
a322f70c
DW		 112 __u32 crc;
113 char guid[DDF_GUID_LEN];
59e36268 114 char revision[8]; /* 01.02.00 */
a322f70c
DW		 115 __u32 seq; /* starts at '1' */
116 __u32 timestamp;
117 __u8 openflag;
118 __u8 foreignflag;
119 __u8 enforcegroups;
120 __u8 pad0; /* 0xff */
121 __u8 pad1[12]; /* 12 * 0xff */
122 /* 64 bytes so far */
123 __u8 header_ext[32]; /* reserved: fill with 0xff */
124 __u64 primary_lba;
125 __u64 secondary_lba;
126 __u8 type;
127 __u8 pad2[3]; /* 0xff */
128 __u32 workspace_len; /* sectors for vendor space -
129 * at least 32768(sectors) */
130 __u64 workspace_lba;
131 __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
132 __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
133 __u16 max_partitions; /* i.e. max num of configuration
134 record entries per disk */
135 __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
136 *12/512) */
137 __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
138 __u8 pad3[54]; /* 0xff */
139 /* 192 bytes so far */
140 __u32 controller_section_offset;
141 __u32 controller_section_length;
142 __u32 phys_section_offset;
143 __u32 phys_section_length;
144 __u32 virt_section_offset;
145 __u32 virt_section_length;
146 __u32 config_section_offset;
147 __u32 config_section_length;
148 __u32 data_section_offset;
149 __u32 data_section_length;
150 __u32 bbm_section_offset;
151 __u32 bbm_section_length;
152 __u32 diag_space_offset;
153 __u32 diag_space_length;
154 __u32 vendor_offset;
155 __u32 vendor_length;
156 /* 256 bytes so far */
157 __u8 pad4[256]; /* 0xff */
158};
159
160/* type field */
161#define DDF_HEADER_ANCHOR 0x00
162#define DDF_HEADER_PRIMARY 0x01
163#define DDF_HEADER_SECONDARY 0x02
164
165/* The content of the 'controller section' - global scope */
166struct ddf_controller_data {
88c164f4 167 __u32 magic; /* DDF_CONTROLLER_MAGIC */
a322f70c
DW		 168 __u32 crc;
169 char guid[DDF_GUID_LEN];
170 struct controller_type {
171 __u16 vendor_id;
172 __u16 device_id;
173 __u16 sub_vendor_id;
174 __u16 sub_device_id;
175 } type;
176 char product_id[16];
177 __u8 pad[8]; /* 0xff */
178 __u8 vendor_data[448];
179};
180
181/* The content of phys_section - global scope */
182struct phys_disk {
88c164f4 183 __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */
a322f70c
DW		 184 __u32 crc;
185 __u16 used_pdes;
186 __u16 max_pdes;
187 __u8 pad[52];
188 struct phys_disk_entry {
189 char guid[DDF_GUID_LEN];
190 __u32 refnum;
191 __u16 type;
192 __u16 state;
193 __u64 config_size; /* DDF structures must be after here */
194 char path[18]; /* another horrible structure really */
195 __u8 pad[6];
196 } entries[0];
197};
198
199/* phys_disk_entry.type is a bitmap - bigendian remember */
200#define DDF_Forced_PD_GUID 1
201#define DDF_Active_in_VD 2
88c164f4 202#define DDF_Global_Spare 4 /* VD_CONF records are ignored */
a322f70c
DW		 203#define DDF_Spare 8 /* overrides Global_spare */
204#define DDF_Foreign 16
205#define DDF_Legacy 32 /* no DDF on this device */
206
207#define DDF_Interface_mask 0xf00
208#define DDF_Interface_SCSI 0x100
209#define DDF_Interface_SAS 0x200
210#define DDF_Interface_SATA 0x300
211#define DDF_Interface_FC 0x400
212
213/* phys_disk_entry.state is a bigendian bitmap */
214#define DDF_Online 1
215#define DDF_Failed 2 /* overrides 1,4,8 */
216#define DDF_Rebuilding 4
217#define DDF_Transition 8
218#define DDF_SMART 16
219#define DDF_ReadErrors 32
220#define DDF_Missing 64
221
222/* The content of the virt_section global scope */
223struct virtual_disk {
88c164f4 224 __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */
a322f70c
DW		 225 __u32 crc;
226 __u16 populated_vdes;
227 __u16 max_vdes;
228 __u8 pad[52];
229 struct virtual_entry {
230 char guid[DDF_GUID_LEN];
231 __u16 unit;
232 __u16 pad0; /* 0xffff */
233 __u16 guid_crc;
234 __u16 type;
235 __u8 state;
236 __u8 init_state;
237 __u8 pad1[14];
238 char name[16];
239 } entries[0];
240};
241
242/* virtual_entry.type is a bitmap - bigendian */
243#define DDF_Shared 1
244#define DDF_Enforce_Groups 2
245#define DDF_Unicode 4
246#define DDF_Owner_Valid 8
247
248/* virtual_entry.state is a bigendian bitmap */
249#define DDF_state_mask 0x7
250#define DDF_state_optimal 0x0
251#define DDF_state_degraded 0x1
252#define DDF_state_deleted 0x2
253#define DDF_state_missing 0x3
254#define DDF_state_failed 0x4
7a7cc504 255#define DDF_state_part_optimal 0x5
a322f70c
DW		 256
257#define DDF_state_morphing 0x8
258#define DDF_state_inconsistent 0x10
259
260/* virtual_entry.init_state is a bigendian bitmap */
261#define DDF_initstate_mask 0x03
262#define DDF_init_not 0x00
7a7cc504
NB		 263#define DDF_init_quick 0x01 /* initialisation is progress.
264 * i.e. 'state_inconsistent' */
a322f70c
DW		 265#define DDF_init_full 0x02
266
267#define DDF_access_mask 0xc0
268#define DDF_access_rw 0x00
269#define DDF_access_ro 0x80
270#define DDF_access_blocked 0xc0
271
272/* The content of the config_section - local scope
273 * It has multiple records each config_record_len sectors
274 * They can be vd_config or spare_assign
275 */
276
277struct vd_config {
88c164f4 278 __u32 magic; /* DDF_VD_CONF_MAGIC */
a322f70c
DW		 279 __u32 crc;
280 char guid[DDF_GUID_LEN];
281 __u32 timestamp;
282 __u32 seqnum;
283 __u8 pad0[24];
284 __u16 prim_elmnt_count;
285 __u8 chunk_shift; /* 0 == 512, 1==1024 etc */
286 __u8 prl;
287 __u8 rlq;
288 __u8 sec_elmnt_count;
289 __u8 sec_elmnt_seq;
290 __u8 srl;
598f0d58
NB		 291 __u64 blocks; /* blocks per component could be different
292 * on different component devices...(only
293 * for concat I hope) */
294 __u64 array_blocks; /* blocks in array */
a322f70c
DW		 295 __u8 pad1[8];
296 __u32 spare_refs[8];
297 __u8 cache_pol[8];
298 __u8 bg_rate;
299 __u8 pad2[3];
300 __u8 pad3[52];
301 __u8 pad4[192];
302 __u8 v0[32]; /* reserved- 0xff */
303 __u8 v1[32]; /* reserved- 0xff */
304 __u8 v2[16]; /* reserved- 0xff */
305 __u8 v3[16]; /* reserved- 0xff */
306 __u8 vendor[32];
307 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
308 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
309 bvd are always the same size */
310};
311
312/* vd_config.cache_pol[7] is a bitmap */
313#define DDF_cache_writeback 1 /* else writethrough */
314#define DDF_cache_wadaptive 2 /* only applies if writeback */
315#define DDF_cache_readahead 4
316#define DDF_cache_radaptive 8 /* only if doing read-ahead */
317#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
318#define DDF_cache_wallowed 32 /* enable write caching */
319#define DDF_cache_rallowed 64 /* enable read caching */
320
321struct spare_assign {
88c164f4 322 __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */
a322f70c
DW		 323 __u32 crc;
324 __u32 timestamp;
325 __u8 reserved[7];
326 __u8 type;
327 __u16 populated; /* SAEs used */
328 __u16 max; /* max SAEs */
329 __u8 pad[8];
330 struct spare_assign_entry {
331 char guid[DDF_GUID_LEN];
332 __u16 secondary_element;
333 __u8 pad[6];
334 } spare_ents[0];
335};
336/* spare_assign.type is a bitmap */
337#define DDF_spare_dedicated 0x1 /* else global */
338#define DDF_spare_revertible 0x2 /* else committable */
339#define DDF_spare_active 0x4 /* else not active */
340#define DDF_spare_affinity 0x8 /* enclosure affinity */
341
342/* The data_section contents - local scope */
343struct disk_data {
88c164f4 344 __u32 magic; /* DDF_PHYS_DATA_MAGIC */
a322f70c
DW		 345 __u32 crc;
346 char guid[DDF_GUID_LEN];
347 __u32 refnum; /* crc of some magic drive data ... */
348 __u8 forced_ref; /* set when above was not result of magic */
349 __u8 forced_guid; /* set if guid was forced rather than magic */
350 __u8 vendor[32];
351 __u8 pad[442];
352};
353
354/* bbm_section content */
355struct bad_block_log {
356 __u32 magic;
357 __u32 crc;
358 __u16 entry_count;
359 __u32 spare_count;
360 __u8 pad[10];
361 __u64 first_spare;
362 struct mapped_block {
363 __u64 defective_start;
364 __u32 replacement_start;
365 __u16 remap_count;
366 __u8 pad[2];
367 } entries[0];
368};
369
370/* Struct for internally holding ddf structures */
371/* The DDF structure stored on each device is potentially
372 * quite different, as some data is global and some is local.
373 * The global data is:
374 * - ddf header
375 * - controller_data
376 * - Physical disk records
377 * - Virtual disk records
378 * The local data is:
379 * - Configuration records
380 * - Physical Disk data section
381 * ( and Bad block and vendor which I don't care about yet).
382 *
383 * The local data is parsed into separate lists as it is read
384 * and reconstructed for writing. This means that we only need
385 * to make config changes once and they are automatically
386 * propagated to all devices.
387 * Note that the ddf_super has space of the conf and disk data
388 * for this disk and also for a list of all such data.
389 * The list is only used for the superblock that is being
390 * built in Create or Assemble to describe the whole array.
391 */
392struct ddf_super {
6416d527 393 struct ddf_header anchor, primary, secondary;
a322f70c 394 struct ddf_controller_data controller;
6416d527 395 struct ddf_header *active;
a322f70c
DW		 396 struct phys_disk *phys;
397 struct virtual_disk *virt;
398 int pdsize, vdsize;
8c3b8c2c 399 int max_part, mppe, conf_rec_len;
d2ca6449 400 int currentdev;
18a2f463 401 int updates_pending;
a322f70c 402 struct vcl {
6416d527
NB		 403 union {
404 char space[512];
405 struct {
406 struct vcl *next;
407 __u64 *lba_offset; /* location in 'conf' of
408 * the lba table */
409 int vcnum; /* index into ->virt */
410 __u64 *block_sizes; /* NULL if all the same */
411 };
412 };
a322f70c 413 struct vd_config conf;
d2ca6449 414 } *conflist, *currentconf;
a322f70c 415 struct dl {
6416d527
NB		 416 union {
417 char space[512];
418 struct {
419 struct dl *next;
420 int major, minor;
421 char *devname;
422 int fd;
423 unsigned long long size; /* sectors */
424 int pdnum; /* index in ->phys */
425 struct spare_assign *spare;
8592f29d
N		 426 void *mdupdate; /* hold metadata update */
427
428 /* These fields used by auto-layout */
429 int raiddisk; /* slot to fill in autolayout */
430 __u64 esize;
6416d527
NB		 431 };
432 };
a322f70c 433 struct disk_data disk;
b2280677 434 struct vcl *vlist[0]; /* max_part in size */
2cc2983d 435 } *dlist, *add_list;
a322f70c
DW		 436};
437
438#ifndef offsetof
439#define offsetof(t,f) ((size_t)&(((t*)0)->f))
440#endif
441
a322f70c
DW		 442
443static int calc_crc(void *buf, int len)
444{
445 /* crcs are always at the same place as in the ddf_header */
446 struct ddf_header *ddf = buf;
447 __u32 oldcrc = ddf->crc;
448 __u32 newcrc;
449 ddf->crc = 0xffffffff;
450
451 newcrc = crc32(0, buf, len);
452 ddf->crc = oldcrc;
4abe6b70
N		 453 /* The crc is store (like everything) bigendian, so convert
454 * here for simplicity
455 */
456 return __cpu_to_be32(newcrc);
a322f70c
DW		 457}
458
459static int load_ddf_header(int fd, unsigned long long lba,
460 unsigned long long size,
461 int type,
462 struct ddf_header *hdr, struct ddf_header *anchor)
463{
464 /* read a ddf header (primary or secondary) from fd/lba
465 * and check that it is consistent with anchor
466 * Need to check:
467 * magic, crc, guid, rev, and LBA's header_type, and
468 * everything after header_type must be the same
469 */
470 if (lba >= size-1)
471 return 0;
472
473 if (lseek64(fd, lba<<9, 0) < 0)
474 return 0;
475
476 if (read(fd, hdr, 512) != 512)
477 return 0;
478
479 if (hdr->magic != DDF_HEADER_MAGIC)
480 return 0;
481 if (calc_crc(hdr, 512) != hdr->crc)
482 return 0;
483 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
484 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
485 anchor->primary_lba != hdr->primary_lba ||
486 anchor->secondary_lba != hdr->secondary_lba ||
487 hdr->type != type ||
488 memcmp(anchor->pad2, hdr->pad2, 512 -
489 offsetof(struct ddf_header, pad2)) != 0)
490 return 0;
491
492 /* Looks good enough to me... */
493 return 1;
494}
495
496static void *load_section(int fd, struct ddf_super *super, void *buf,
497 __u32 offset_be, __u32 len_be, int check)
498{
499 unsigned long long offset = __be32_to_cpu(offset_be);
500 unsigned long long len = __be32_to_cpu(len_be);
501 int dofree = (buf == NULL);
502
503 if (check)
504 if (len != 2 && len != 8 && len != 32
505 && len != 128 && len != 512)
506 return NULL;
507
508 if (len > 1024)
509 return NULL;
510 if (buf) {
511 /* All pre-allocated sections are a single block */
512 if (len != 1)
513 return NULL;
3d2c4fc7
DW		 514 } else if (posix_memalign(&buf, 512, len<<9) != 0)
515 buf = NULL;
6416d527 516
a322f70c
DW		 517 if (!buf)
518 return NULL;
519
520 if (super->active->type == 1)
521 offset += __be64_to_cpu(super->active->primary_lba);
522 else
523 offset += __be64_to_cpu(super->active->secondary_lba);
524
525 if (lseek64(fd, offset<<9, 0) != (offset<<9)) {
526 if (dofree)
527 free(buf);
528 return NULL;
529 }
530 if (read(fd, buf, len<<9) != (len<<9)) {
531 if (dofree)
532 free(buf);
533 return NULL;
534 }
535 return buf;
536}
537
538static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
539{
540 unsigned long long dsize;
541
542 get_dev_size(fd, NULL, &dsize);
543
544 if (lseek64(fd, dsize-512, 0) < 0) {
545 if (devname)
546 fprintf(stderr,
547 Name": Cannot seek to anchor block on %s: %s\n",
548 devname, strerror(errno));
549 return 1;
550 }
551 if (read(fd, &super->anchor, 512) != 512) {
552 if (devname)
553 fprintf(stderr,
554 Name ": Cannot read anchor block on %s: %s\n",
555 devname, strerror(errno));
556 return 1;
557 }
558 if (super->anchor.magic != DDF_HEADER_MAGIC) {
559 if (devname)
560 fprintf(stderr, Name ": no DDF anchor found on %s\n",
561 devname);
562 return 2;
563 }
564 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
565 if (devname)
566 fprintf(stderr, Name ": bad CRC on anchor on %s\n",
567 devname);
568 return 2;
569 }
59e36268
NB		 570 if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
571 memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
a322f70c
DW		 572 if (devname)
573 fprintf(stderr, Name ": can only support super revision"
59e36268
NB		 574 " %.8s and earlier, not %.8s on %s\n",
575 DDF_REVISION_2, super->anchor.revision,devname);
a322f70c
DW		 576 return 2;
577 }
578 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
579 dsize >> 9, 1,
580 &super->primary, &super->anchor) == 0) {
581 if (devname)
582 fprintf(stderr,
583 Name ": Failed to load primary DDF header "
584 "on %s\n", devname);
585 return 2;
586 }
587 super->active = &super->primary;
588 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
589 dsize >> 9, 2,
590 &super->secondary, &super->anchor)) {
591 if ((__be32_to_cpu(super->primary.seq)
592 < __be32_to_cpu(super->secondary.seq) &&
593 !super->secondary.openflag)
594 || (__be32_to_cpu(super->primary.seq)
595 == __be32_to_cpu(super->secondary.seq) &&
596 super->primary.openflag && !super->secondary.openflag)
597 )
598 super->active = &super->secondary;
599 }
600 return 0;
601}
602
603static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
604{
605 void *ok;
606 ok = load_section(fd, super, &super->controller,
607 super->active->controller_section_offset,
608 super->active->controller_section_length,
609 0);
610 super->phys = load_section(fd, super, NULL,
611 super->active->phys_section_offset,
612 super->active->phys_section_length,
613 1);
614 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
615
616 super->virt = load_section(fd, super, NULL,
617 super->active->virt_section_offset,
618 super->active->virt_section_length,
619 1);
620 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
621 if (!ok ||
622 !super->phys ||
623 !super->virt) {
624 free(super->phys);
625 free(super->virt);
a2349791
NB		 626 super->phys = NULL;
627 super->virt = NULL;
a322f70c
DW		 628 return 2;
629 }
630 super->conflist = NULL;
631 super->dlist = NULL;
8c3b8c2c
NB		 632
633 super->max_part = __be16_to_cpu(super->active->max_partitions);
634 super->mppe = __be16_to_cpu(super->active->max_primary_element_entries);
635 super->conf_rec_len = __be16_to_cpu(super->active->config_record_len);
a322f70c
DW		 636 return 0;
637}
638
639static int load_ddf_local(int fd, struct ddf_super *super,
640 char *devname, int keep)
641{
642 struct dl *dl;
643 struct stat stb;
644 char *conf;
645 int i;
e223334f 646 int confsec;
b2280677 647 int vnum;
59e36268 648 int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
d2ca6449 649 unsigned long long dsize;
a322f70c
DW		 650
651 /* First the local disk info */
3d2c4fc7 652 if (posix_memalign((void**)&dl, 512,
6416d527 653 sizeof(*dl) +
3d2c4fc7
DW		 654 (super->max_part) * sizeof(dl->vlist[0])) != 0) {
655 fprintf(stderr, Name ": %s could not allocate disk info buffer\n",
656 __func__);
657 return 1;
658 }
a322f70c
DW		 659
660 load_section(fd, super, &dl->disk,
661 super->active->data_section_offset,
662 super->active->data_section_length,
663 0);
664 dl->devname = devname ? strdup(devname) : NULL;
598f0d58 665
a322f70c
DW		 666 fstat(fd, &stb);
667 dl->major = major(stb.st_rdev);
668 dl->minor = minor(stb.st_rdev);
669 dl->next = super->dlist;
670 dl->fd = keep ? fd : -1;
d2ca6449
NB		 671
672 dl->size = 0;
673 if (get_dev_size(fd, devname, &dsize))
674 dl->size = dsize >> 9;
b2280677
NB		 675 dl->spare = NULL;
676 for (i=0 ; i < super->max_part ; i++)
a322f70c
DW		 677 dl->vlist[i] = NULL;
678 super->dlist = dl;
59e36268 679 dl->pdnum = -1;
5575e7d9
NB		 680 for (i=0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
681 if (memcmp(super->phys->entries[i].guid,
682 dl->disk.guid, DDF_GUID_LEN) == 0)
683 dl->pdnum = i;
684
a322f70c
DW		 685 /* Now the config list. */
686 /* 'conf' is an array of config entries, some of which are
687 * probably invalid. Those which are good need to be copied into
688 * the conflist
689 */
a322f70c
DW		 690
691 conf = load_section(fd, super, NULL,
692 super->active->config_section_offset,
693 super->active->config_section_length,
694 0);
695
b2280677 696 vnum = 0;
e223334f
N		 697 for (confsec = 0;
698 confsec < __be32_to_cpu(super->active->config_section_length);
699 confsec += super->conf_rec_len) {
a322f70c 700 struct vd_config *vd =
e223334f 701 (struct vd_config *)((char*)conf + confsec*512);
a322f70c
DW		 702 struct vcl *vcl;
703
b2280677
NB		 704 if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
705 if (dl->spare)
706 continue;
3d2c4fc7
DW		 707 if (posix_memalign((void**)&dl->spare, 512,
708 super->conf_rec_len*512) != 0) {
709 fprintf(stderr, Name
710 ": %s could not allocate spare info buf\n",
711 __func__);
712 return 1;
713 }
714
b2280677
NB		 715 memcpy(dl->spare, vd, super->conf_rec_len*512);
716 continue;
717 }
a322f70c
DW		 718 if (vd->magic != DDF_VD_CONF_MAGIC)
719 continue;
720 for (vcl = super->conflist; vcl; vcl = vcl->next) {
721 if (memcmp(vcl->conf.guid,
722 vd->guid, DDF_GUID_LEN) == 0)
723 break;
724 }
725
726 if (vcl) {
b2280677 727 dl->vlist[vnum++] = vcl;
a322f70c
DW		 728 if (__be32_to_cpu(vd->seqnum) <=
729 __be32_to_cpu(vcl->conf.seqnum))
730 continue;
59e36268 731 } else {
3d2c4fc7 732 if (posix_memalign((void**)&vcl, 512,
6416d527 733 (super->conf_rec_len*512 +
3d2c4fc7
DW		 734 offsetof(struct vcl, conf))) != 0) {
735 fprintf(stderr, Name
736 ": %s could not allocate vcl buf\n",
737 __func__);
738 return 1;
739 }
a322f70c 740 vcl->next = super->conflist;
59e36268 741 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
a322f70c 742 super->conflist = vcl;
b2280677 743 dl->vlist[vnum++] = vcl;
a322f70c 744 }
8c3b8c2c 745 memcpy(&vcl->conf, vd, super->conf_rec_len*512);
a322f70c 746 vcl->lba_offset = (__u64*)
8c3b8c2c 747 &vcl->conf.phys_refnum[super->mppe];
59e36268
NB		 748
749 for (i=0; i < max_virt_disks ; i++)
750 if (memcmp(super->virt->entries[i].guid,
751 vcl->conf.guid, DDF_GUID_LEN)==0)
752 break;
753 if (i < max_virt_disks)
754 vcl->vcnum = i;
a322f70c
DW		 755 }
756 free(conf);
757
758 return 0;
759}
760
761#ifndef MDASSEMBLE
762static int load_super_ddf_all(struct supertype *st, int fd,
763 void **sbp, char *devname, int keep_fd);
764#endif
37424f13
DW		 765
766static void free_super_ddf(struct supertype *st);
767
a322f70c
DW		 768static int load_super_ddf(struct supertype *st, int fd,
769 char *devname)
770{
771 unsigned long long dsize;
772 struct ddf_super *super;
773 int rv;
774
775#ifndef MDASSEMBLE
59e36268 776 /* if 'fd' is a container, load metadata from all the devices */
3dbccbcf 777 if (load_super_ddf_all(st, fd, &st->sb, devname, 1) == 0)
a322f70c
DW		 778 return 0;
779#endif
f7e7067b
NB		 780 if (st->subarray[0])
781 return 1; /* FIXME Is this correct */
a322f70c
DW		 782
783 if (get_dev_size(fd, devname, &dsize) == 0)
784 return 1;
785
786 /* 32M is a lower bound */
787 if (dsize <= 32*1024*1024) {
97320d7c 788 if (devname)
a322f70c
DW		 789 fprintf(stderr,
790 Name ": %s is too small for ddf: "
791 "size is %llu sectors.\n",
792 devname, dsize>>9);
97320d7c 793 return 1;
a322f70c
DW		 794 }
795 if (dsize & 511) {
97320d7c 796 if (devname)
a322f70c
DW		 797 fprintf(stderr,
798 Name ": %s is an odd size for ddf: "
799 "size is %llu bytes.\n",
800 devname, dsize);
97320d7c 801 return 1;
a322f70c
DW		 802 }
803
37424f13
DW		 804 free_super_ddf(st);
805
6416d527 806 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW		 807 fprintf(stderr, Name ": malloc of %zu failed.\n",
808 sizeof(*super));
809 return 1;
810 }
a2349791 811 memset(super, 0, sizeof(*super));
a322f70c
DW		 812
813 rv = load_ddf_headers(fd, super, devname);
814 if (rv) {
815 free(super);
816 return rv;
817 }
818
819 /* Have valid headers and have chosen the best. Let's read in the rest*/
820
821 rv = load_ddf_global(fd, super, devname);
822
823 if (rv) {
824 if (devname)
825 fprintf(stderr,
826 Name ": Failed to load all information "
827 "sections on %s\n", devname);
828 free(super);
829 return rv;
830 }
831
3d2c4fc7
DW		 832 rv = load_ddf_local(fd, super, devname, 0);
833
834 if (rv) {
835 if (devname)
836 fprintf(stderr,
837 Name ": Failed to load all information "
838 "sections on %s\n", devname);
839 free(super);
840 return rv;
841 }
a322f70c 842
af99d9ca
DW		 843 if (st->subarray[0]) {
844 struct vcl *v;
845
846 for (v = super->conflist; v; v = v->next)
847 if (v->vcnum == atoi(st->subarray))
848 super->currentconf = v;
849 if (!super->currentconf) {
850 free(super);
851 return 1;
852 }
853 }
854
a322f70c
DW		 855 /* Should possibly check the sections .... */
856
857 st->sb = super;
858 if (st->ss == NULL) {
859 st->ss = &super_ddf;
860 st->minor_version = 0;
861 st->max_devs = 512;
862 }
352452c3 863 st->loaded_container = 0;
a322f70c
DW		 864 return 0;
865
866}
867
868static void free_super_ddf(struct supertype *st)
869{
870 struct ddf_super *ddf = st->sb;
871 if (ddf == NULL)
872 return;
873 free(ddf->phys);
874 free(ddf->virt);
875 while (ddf->conflist) {
876 struct vcl *v = ddf->conflist;
877 ddf->conflist = v->next;
59e36268
NB		 878 if (v->block_sizes)
879 free(v->block_sizes);
a322f70c
DW		 880 free(v);
881 }
882 while (ddf->dlist) {
883 struct dl *d = ddf->dlist;
884 ddf->dlist = d->next;
885 if (d->fd >= 0)
886 close(d->fd);
b2280677
NB		 887 if (d->spare)
888 free(d->spare);
a322f70c
DW		 889 free(d);
890 }
891 free(ddf);
892 st->sb = NULL;
893}
894
895static struct supertype *match_metadata_desc_ddf(char *arg)
896{
897 /* 'ddf' only support containers */
898 struct supertype *st;
899 if (strcmp(arg, "ddf") != 0 &&
900 strcmp(arg, "default") != 0
901 )
902 return NULL;
903
904 st = malloc(sizeof(*st));
ef609477 905 memset(st, 0, sizeof(*st));
a322f70c
DW		 906 st->ss = &super_ddf;
907 st->max_devs = 512;
908 st->minor_version = 0;
909 st->sb = NULL;
910 return st;
911}
912
a322f70c
DW		 913
914#ifndef MDASSEMBLE
915
916static mapping_t ddf_state[] = {
917 { "Optimal", 0},
918 { "Degraded", 1},
919 { "Deleted", 2},
920 { "Missing", 3},
921 { "Failed", 4},
922 { "Partially Optimal", 5},
923 { "-reserved-", 6},
924 { "-reserved-", 7},
925 { NULL, 0}
926};
927
928static mapping_t ddf_init_state[] = {
929 { "Not Initialised", 0},
930 { "QuickInit in Progress", 1},
931 { "Fully Initialised", 2},
932 { "*UNKNOWN*", 3},
933 { NULL, 0}
934};
935static mapping_t ddf_access[] = {
936 { "Read/Write", 0},
937 { "Reserved", 1},
938 { "Read Only", 2},
939 { "Blocked (no access)", 3},
940 { NULL ,0}
941};
942
943static mapping_t ddf_level[] = {
944 { "RAID0", DDF_RAID0},
945 { "RAID1", DDF_RAID1},
946 { "RAID3", DDF_RAID3},
947 { "RAID4", DDF_RAID4},
948 { "RAID5", DDF_RAID5},
949 { "RAID1E",DDF_RAID1E},
950 { "JBOD", DDF_JBOD},
951 { "CONCAT",DDF_CONCAT},
952 { "RAID5E",DDF_RAID5E},
953 { "RAID5EE",DDF_RAID5EE},
954 { "RAID6", DDF_RAID6},
955 { NULL, 0}
956};
957static mapping_t ddf_sec_level[] = {
958 { "Striped", DDF_2STRIPED},
959 { "Mirrored", DDF_2MIRRORED},
960 { "Concat", DDF_2CONCAT},
961 { "Spanned", DDF_2SPANNED},
962 { NULL, 0}
963};
964#endif
965
966struct num_mapping {
967 int num1, num2;
968};
969static struct num_mapping ddf_level_num[] = {
970 { DDF_RAID0, 0 },
971 { DDF_RAID1, 1 },
972 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB		 973 { DDF_RAID4, 4 },
974 { DDF_RAID5, 5 },
a322f70c
DW		 975 { DDF_RAID1E, LEVEL_UNSUPPORTED },
976 { DDF_JBOD, LEVEL_UNSUPPORTED },
977 { DDF_CONCAT, LEVEL_LINEAR },
978 { DDF_RAID5E, LEVEL_UNSUPPORTED },
979 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
980 { DDF_RAID6, 6},
981 { MAXINT, MAXINT }
982};
983
984static int map_num1(struct num_mapping *map, int num)
985{
986 int i;
987 for (i=0 ; map[i].num1 != MAXINT; i++)
988 if (map[i].num1 == num)
989 break;
990 return map[i].num2;
991}
992
42dc2744
N		 993static int all_ff(char *guid)
994{
995 int i;
996 for (i = 0; i < DDF_GUID_LEN; i++)
997 if (guid[i] != (char)0xff)
998 return 0;
999 return 1;
1000}
1001
a322f70c
DW		 1002#ifndef MDASSEMBLE
1003static void print_guid(char *guid, int tstamp)
1004{
1005 /* A GUIDs are part (or all) ASCII and part binary.
1006 * They tend to be space padded.
59e36268
NB		 1007 * We print the GUID in HEX, then in parentheses add
1008 * any initial ASCII sequence, and a possible
1009 * time stamp from bytes 16-19
a322f70c
DW		 1010 */
1011 int l = DDF_GUID_LEN;
1012 int i;
59e36268
NB		 1013
1014 for (i=0 ; i<DDF_GUID_LEN ; i++) {
1015 if ((i&3)==0 && i != 0) printf(":");
1016 printf("%02X", guid[i]&255);
1017 }
1018
cfccea8c 1019 printf("\n (");
a322f70c
DW		 1020 while (l && guid[l-1] == ' ')
1021 l--;
1022 for (i=0 ; i<l ; i++) {
1023 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1024 fputc(guid[i], stdout);
1025 else
59e36268 1026 break;
a322f70c
DW		 1027 }
1028 if (tstamp) {
1029 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1030 char tbuf[100];
1031 struct tm *tm;
1032 tm = localtime(&then);
59e36268 1033 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW		 1034 fputs(tbuf, stdout);
1035 }
59e36268 1036 printf(")");
a322f70c
DW		 1037}
1038
1039static void examine_vd(int n, struct ddf_super *sb, char *guid)
1040{
8c3b8c2c 1041 int crl = sb->conf_rec_len;
a322f70c
DW		 1042 struct vcl *vcl;
1043
1044 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
b06e3095 1045 int i;
a322f70c
DW		 1046 struct vd_config *vc = &vcl->conf;
1047
1048 if (calc_crc(vc, crl*512) != vc->crc)
1049 continue;
1050 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1051 continue;
1052
1053 /* Ok, we know about this VD, let's give more details */
b06e3095 1054 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1055 __be16_to_cpu(vc->prim_elmnt_count));
b06e3095
N		 1056 for (i=0; i<__be16_to_cpu(vc->prim_elmnt_count); i++) {
1057 int j;
1058 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1059 for (j=0; j<cnt; j++)
1060 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1061 break;
1062 if (i) printf(" ");
1063 if (j < cnt)
1064 printf("%d", j);
1065 else
1066 printf("--");
1067 }
1068 printf(")\n");
1069 if (vc->chunk_shift != 255)
a322f70c
DW		 1070 printf(" Chunk Size[%d] : %d sectors\n", n,
1071 1 << vc->chunk_shift);
1072 printf(" Raid Level[%d] : %s\n", n,
1073 map_num(ddf_level, vc->prl)?:"-unknown-");
1074 if (vc->sec_elmnt_count != 1) {
1075 printf(" Secondary Position[%d] : %d of %d\n", n,
1076 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1077 printf(" Secondary Level[%d] : %s\n", n,
1078 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1079 }
1080 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1081 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1082 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1083 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW		 1084 }
1085}
1086
1087static void examine_vds(struct ddf_super *sb)
1088{
1089 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1090 int i;
1091 printf(" Virtual Disks : %d\n", cnt);
1092
1093 for (i=0; i<cnt; i++) {
1094 struct virtual_entry *ve = &sb->virt->entries[i];
b06e3095 1095 printf("\n");
a322f70c
DW		 1096 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1097 printf("\n");
1098 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1099 printf(" state[%d] : %s, %s%s\n", i,
1100 map_num(ddf_state, ve->state & 7),
1101 (ve->state & 8) ? "Morphing, ": "",
1102 (ve->state & 16)? "Not Consistent" : "Consistent");
1103 printf(" init state[%d] : %s\n", i,
1104 map_num(ddf_init_state, ve->init_state&3));
1105 printf(" access[%d] : %s\n", i,
1106 map_num(ddf_access, (ve->init_state>>6) & 3));
1107 printf(" Name[%d] : %.16s\n", i, ve->name);
1108 examine_vd(i, sb, ve->guid);
1109 }
1110 if (cnt) printf("\n");
1111}
1112
1113static void examine_pds(struct ddf_super *sb)
1114{
1115 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1116 int i;
1117 struct dl *dl;
1118 printf(" Physical Disks : %d\n", cnt);
962371a5 1119 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW		 1120
1121 for (i=0 ; i<cnt ; i++) {
1122 struct phys_disk_entry *pd = &sb->phys->entries[i];
1123 int type = __be16_to_cpu(pd->type);
1124 int state = __be16_to_cpu(pd->state);
1125
b06e3095
N		 1126 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1127 //printf("\n");
1128 printf(" %3d %08x ", i,
a322f70c 1129 __be32_to_cpu(pd->refnum));
c9b6907b
DL		 1130 printf("%8lluK ",
1131 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N		 1132 for (dl = sb->dlist; dl ; dl = dl->next) {
1133 if (dl->disk.refnum == pd->refnum) {
1134 char *dv = map_dev(dl->major, dl->minor, 0);
1135 if (dv) {
962371a5 1136 printf("%-15s", dv);
b06e3095
N		 1137 break;
1138 }
1139 }
1140 }
1141 if (!dl)
962371a5 1142 printf("%15s","");
b06e3095 1143 printf(" %s%s%s%s%s",
a322f70c 1144 (type&2) ? "active":"",
b06e3095 1145 (type&4) ? "Global-Spare":"",
a322f70c
DW		 1146 (type&8) ? "spare" : "",
1147 (type&16)? ", foreign" : "",
1148 (type&32)? "pass-through" : "");
b06e3095 1149 printf("/%s%s%s%s%s%s%s",
a322f70c
DW		 1150 (state&1)? "Online": "Offline",
1151 (state&2)? ", Failed": "",
1152 (state&4)? ", Rebuilding": "",
1153 (state&8)? ", in-transition": "",
b06e3095
N		 1154 (state&16)? ", SMART-errors": "",
1155 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1156 (state&64)? ", Missing" : "");
a322f70c
DW		 1157 printf("\n");
1158 }
1159}
1160
1161static void examine_super_ddf(struct supertype *st, char *homehost)
1162{
1163 struct ddf_super *sb = st->sb;
1164
1165 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1166 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB		 1167 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1168 printf("\n");
1169 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW		 1170 printf("\n");
1171 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1172 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1173 ?"yes" : "no");
1174 examine_vds(sb);
1175 examine_pds(sb);
1176}
1177
ff54de6e
N		 1178static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info);
1179
42dc2744 1180static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1181
061f2c6a 1182static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N		 1183{
1184 /* We just write a generic DDF ARRAY entry
1185 */
1186 struct mdinfo info;
1187 char nbuf[64];
1188 getinfo_super_ddf(st, &info);
1189 fname_from_uuid(st, &info, nbuf, ':');
1190
1191 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1192}
1193
1194static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW		 1195{
1196 /* We just write a generic DDF ARRAY entry
a322f70c 1197 */
42dc2744 1198 struct ddf_super *ddf = st->sb;
ff54de6e 1199 struct mdinfo info;
42dc2744 1200 int i;
ff54de6e
N		 1201 char nbuf[64];
1202 getinfo_super_ddf(st, &info);
1203 fname_from_uuid(st, &info, nbuf, ':');
42dc2744
N		 1204
1205 for (i=0; i<__be16_to_cpu(ddf->virt->max_vdes); i++) {
1206 struct virtual_entry *ve = &ddf->virt->entries[i];
1207 struct vcl vcl;
1208 char nbuf1[64];
1209 if (all_ff(ve->guid))
1210 continue;
1211 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1212 ddf->currentconf =&vcl;
1213 uuid_from_super_ddf(st, info.uuid);
1214 fname_from_uuid(st, &info, nbuf1, ':');
1215 printf("ARRAY container=%s member=%d UUID=%s\n",
1216 nbuf+5, i, nbuf1+5);
1217 }
a322f70c
DW		 1218}
1219
bceedeec
N		 1220static void export_examine_super_ddf(struct supertype *st)
1221{
1222 struct mdinfo info;
1223 char nbuf[64];
1224 getinfo_super_ddf(st, &info);
1225 fname_from_uuid(st, &info, nbuf, ':');
1226 printf("MD_METADATA=ddf\n");
1227 printf("MD_LEVEL=container\n");
1228 printf("MD_UUID=%s\n", nbuf+5);
1229}
1230
1231
a322f70c
DW		 1232static void detail_super_ddf(struct supertype *st, char *homehost)
1233{
1234 /* FIXME later
1235 * Could print DDF GUID
1236 * Need to find which array
1237 * If whole, briefly list all arrays
1238 * If one, give name
1239 */
1240}
1241
1242static void brief_detail_super_ddf(struct supertype *st)
1243{
1244 /* FIXME I really need to know which array we are detailing.
1245 * Can that be stored in ddf_super??
1246 */
1247// struct ddf_super *ddf = st->sb;
ff54de6e
N		 1248 struct mdinfo info;
1249 char nbuf[64];
1250 getinfo_super_ddf(st, &info);
1251 fname_from_uuid(st, &info, nbuf,':');
1252 printf(" UUID=%s", nbuf + 5);
a322f70c 1253}
a322f70c
DW		 1254#endif
1255
1256static int match_home_ddf(struct supertype *st, char *homehost)
1257{
1258 /* It matches 'this' host if the controller is a
1259 * Linux-MD controller with vendor_data matching
1260 * the hostname
1261 */
1262 struct ddf_super *ddf = st->sb;
d1d3482b
N		 1263 int len;
1264
1265 if (!homehost)
1266 return 0;
1267 len = strlen(homehost);
a322f70c
DW		 1268
1269 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1270 len < sizeof(ddf->controller.vendor_data) &&
1271 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1272 ddf->controller.vendor_data[len] == 0);
1273}
1274
0e600426 1275#ifndef MDASSEMBLE
7a7cc504 1276static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
a322f70c 1277{
7a7cc504 1278 struct vcl *v;
59e36268 1279
7a7cc504 1280 for (v = ddf->conflist; v; v = v->next)
59e36268 1281 if (inst == v->vcnum)
7a7cc504
NB		 1282 return &v->conf;
1283 return NULL;
1284}
0e600426 1285#endif
7a7cc504
NB		 1286
1287static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1288{
1289 /* Find the entry in phys_disk which has the given refnum
1290 * and return it's index
1291 */
1292 int i;
1293 for (i=0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
1294 if (ddf->phys->entries[i].refnum == phys_refnum)
1295 return i;
1296 return -1;
a322f70c
DW		 1297}
1298
1299static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1300{
1301 /* The uuid returned here is used for:
1302 * uuid to put into bitmap file (Create, Grow)
1303 * uuid for backup header when saving critical section (Grow)
1304 * comparing uuids when re-adding a device into an array
51006d85
N		 1305 * In these cases the uuid required is that of the data-array,
1306 * not the device-set.
1307 * uuid to recognise same set when adding a missing device back
1308 * to an array. This is a uuid for the device-set.
1309 *
a322f70c
DW		 1310 * For each of these we can make do with a truncated
1311 * or hashed uuid rather than the original, as long as
1312 * everyone agrees.
a322f70c
DW		 1313 * In the case of SVD we assume the BVD is of interest,
1314 * though that might be the case if a bitmap were made for
1315 * a mirrored SVD - worry about that later.
1316 * So we need to find the VD configuration record for the
1317 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1318 * The first 16 bytes of the sha1 of these is used.
1319 */
1320 struct ddf_super *ddf = st->sb;
d2ca6449 1321 struct vcl *vcl = ddf->currentconf;
c5afc314
N		 1322 char *guid;
1323 char buf[20];
1324 struct sha1_ctx ctx;
a322f70c 1325
c5afc314
N		 1326 if (vcl)
1327 guid = vcl->conf.guid;
1328 else
1329 guid = ddf->anchor.guid;
1330
1331 sha1_init_ctx(&ctx);
1332 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N		 1333 sha1_finish_ctx(&ctx, buf);
1334 memcpy(uuid, buf, 4*4);
a322f70c
DW		 1335}
1336
78e44928
NB		 1337static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info);
1338
a322f70c
DW		 1339static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1340{
1341 struct ddf_super *ddf = st->sb;
1342
78e44928
NB		 1343 if (ddf->currentconf) {
1344 getinfo_super_ddf_bvd(st, info);
1345 return;
1346 }
1347
a322f70c
DW		 1348 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1349 info->array.level = LEVEL_CONTAINER;
1350 info->array.layout = 0;
1351 info->array.md_minor = -1;
1352 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1353 (ddf->anchor.guid+16));
1354 info->array.utime = 0;
1355 info->array.chunk_size = 0;
1356
a322f70c
DW		 1357
1358 info->disk.major = 0;
1359 info->disk.minor = 0;
cba0191b
NB		 1360 if (ddf->dlist) {
1361 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1362 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB		 1363
1364 info->data_offset = __be64_to_cpu(ddf->phys->
1365 entries[info->disk.raid_disk].
1366 config_size);
1367 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB		 1368 } else {
1369 info->disk.number = -1;
661dce36 1370 info->disk.raid_disk = -1;
cba0191b
NB		 1371// info->disk.raid_disk = find refnum in the table and use index;
1372 }
f22385f9 1373 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1374
d2ca6449 1375
921d9e16 1376 info->recovery_start = MaxSector;
a19c88b8 1377 info->reshape_active = 0;
c5afc314 1378 info->name[0] = 0;
a322f70c 1379
f35f2525
N		 1380 info->array.major_version = -1;
1381 info->array.minor_version = -2;
159c3a1a 1382 strcpy(info->text_version, "ddf");
a67dd8cc 1383 info->safe_mode_delay = 0;
159c3a1a 1384
c5afc314 1385 uuid_from_super_ddf(st, info->uuid);
a322f70c 1386
a322f70c
DW		 1387}
1388
598f0d58
NB		 1389static int rlq_to_layout(int rlq, int prl, int raiddisks);
1390
a322f70c
DW		 1391static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1392{
1393 struct ddf_super *ddf = st->sb;
d2ca6449
NB		 1394 struct vcl *vc = ddf->currentconf;
1395 int cd = ddf->currentdev;
db42fa9b 1396 int j;
8592f29d 1397 struct dl *dl;
a322f70c
DW		 1398
1399 /* FIXME this returns BVD info - what if we want SVD ?? */
1400
d2ca6449
NB		 1401 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1402 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1403 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1404 info->array.raid_disks);
a322f70c 1405 info->array.md_minor = -1;
d2ca6449
NB		 1406 info->array.ctime = DECADE +
1407 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1408 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1409 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1410 info->custom_array_size = 0;
d2ca6449
NB		 1411
1412 if (cd >= 0 && cd < ddf->mppe) {
1413 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1414 if (vc->block_sizes)
1415 info->component_size = vc->block_sizes[cd];
1416 else
1417 info->component_size = __be64_to_cpu(vc->conf.blocks);
1418 }
a322f70c 1419
8592f29d
N		 1420 for (dl = ddf->dlist; dl ; dl = dl->next)
1421 if (dl->raiddisk == info->disk.raid_disk)
1422 break;
a322f70c
DW		 1423 info->disk.major = 0;
1424 info->disk.minor = 0;
8592f29d
N		 1425 if (dl) {
1426 info->disk.major = dl->major;
1427 info->disk.minor = dl->minor;
1428 }
a322f70c
DW		 1429// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1430// info->disk.raid_disk = find refnum in the table and use index;
1431// info->disk.state = ???;
1432
103f2410
NB		 1433 info->container_member = ddf->currentconf->vcnum;
1434
921d9e16 1435 info->recovery_start = MaxSector;
80d26cb2 1436 info->resync_start = 0;
624c5ad4 1437 info->reshape_active = 0;
80d26cb2
NB		 1438 if (!(ddf->virt->entries[info->container_member].state
1439 & DDF_state_inconsistent) &&
1440 (ddf->virt->entries[info->container_member].init_state
1441 & DDF_initstate_mask)
1442 == DDF_init_full)
b7528a20 1443 info->resync_start = MaxSector;
80d26cb2 1444
a322f70c
DW		 1445 uuid_from_super_ddf(st, info->uuid);
1446
3576e302 1447 info->container_member = atoi(st->subarray);
f35f2525
N		 1448 info->array.major_version = -1;
1449 info->array.minor_version = -2;
3576e302 1450 sprintf(info->text_version, "/%s/%s",
159c3a1a 1451 devnum2devname(st->container_dev),
3576e302 1452 st->subarray);
a67dd8cc 1453 info->safe_mode_delay = 200;
159c3a1a 1454
db42fa9b
N		 1455 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1456 info->name[16]=0;
1457 for(j=0; j<16; j++)
1458 if (info->name[j] == ' ')
1459 info->name[j] = 0;
a322f70c
DW		 1460}
1461
598f0d58 1462
a322f70c
DW		 1463static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1464 char *update,
1465 char *devname, int verbose,
1466 int uuid_set, char *homehost)
1467{
1468 /* For 'assemble' and 'force' we need to return non-zero if any
1469 * change was made. For others, the return value is ignored.
1470 * Update options are:
1471 * force-one : This device looks a bit old but needs to be included,
1472 * update age info appropriately.
1473 * assemble: clear any 'faulty' flag to allow this device to
1474 * be assembled.
1475 * force-array: Array is degraded but being forced, mark it clean
1476 * if that will be needed to assemble it.
1477 *
1478 * newdev: not used ????
1479 * grow: Array has gained a new device - this is currently for
1480 * linear only
1481 * resync: mark as dirty so a resync will happen.
59e36268 1482 * uuid: Change the uuid of the array to match what is given
a322f70c
DW		 1483 * homehost: update the recorded homehost
1484 * name: update the name - preserving the homehost
1485 * _reshape_progress: record new reshape_progress position.
1486 *
1487 * Following are not relevant for this version:
1488 * sparc2.2 : update from old dodgey metadata
1489 * super-minor: change the preferred_minor number
1490 * summaries: update redundant counters.
1491 */
1492 int rv = 0;
1493// struct ddf_super *ddf = st->sb;
7a7cc504 1494// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW		 1495// struct virtual_entry *ve = find_ve(ddf);
1496
a322f70c
DW		 1497 /* we don't need to handle "force-*" or "assemble" as
1498 * there is no need to 'trick' the kernel. We the metadata is
1499 * first updated to activate the array, all the implied modifications
1500 * will just happen.
1501 */
1502
1503 if (strcmp(update, "grow") == 0) {
1504 /* FIXME */
1505 }
1506 if (strcmp(update, "resync") == 0) {
1507// info->resync_checkpoint = 0;
1508 }
1509 /* We ignore UUID updates as they make even less sense
1510 * with DDF
1511 */
1512 if (strcmp(update, "homehost") == 0) {
1513 /* homehost is stored in controller->vendor_data,
1514 * or it is when we are the vendor
1515 */
1516// if (info->vendor_is_local)
1517// strcpy(ddf->controller.vendor_data, homehost);
1518 }
1519 if (strcmp(update, "name") == 0) {
1520 /* name is stored in virtual_entry->name */
1521// memset(ve->name, ' ', 16);
1522// strncpy(ve->name, info->name, 16);
1523 }
1524 if (strcmp(update, "_reshape_progress") == 0) {
1525 /* We don't support reshape yet */
1526 }
1527
1528// update_all_csum(ddf);
1529
1530 return rv;
1531}
1532
5f8097be
NB		 1533static void make_header_guid(char *guid)
1534{
1535 __u32 stamp;
5f8097be
NB		 1536 /* Create a DDF Header of Virtual Disk GUID */
1537
1538 /* 24 bytes of fiction required.
1539 * first 8 are a 'vendor-id' - "Linux-MD"
1540 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1541 * Remaining 8 random number plus timestamp
1542 */
1543 memcpy(guid, T10, sizeof(T10));
1544 stamp = __cpu_to_be32(0xdeadbeef);
1545 memcpy(guid+8, &stamp, 4);
1546 stamp = __cpu_to_be32(0);
1547 memcpy(guid+12, &stamp, 4);
1548 stamp = __cpu_to_be32(time(0) - DECADE);
1549 memcpy(guid+16, &stamp, 4);
bfb7ea78 1550 stamp = random32();
5f8097be 1551 memcpy(guid+20, &stamp, 4);
5f8097be 1552}
59e36268 1553
78e44928
NB		 1554static int init_super_ddf_bvd(struct supertype *st,
1555 mdu_array_info_t *info,
1556 unsigned long long size,
1557 char *name, char *homehost,
1558 int *uuid);
1559
a322f70c
DW		 1560static int init_super_ddf(struct supertype *st,
1561 mdu_array_info_t *info,
1562 unsigned long long size, char *name, char *homehost,
1563 int *uuid)
1564{
1565 /* This is primarily called by Create when creating a new array.
1566 * We will then get add_to_super called for each component, and then
1567 * write_init_super called to write it out to each device.
1568 * For DDF, Create can create on fresh devices or on a pre-existing
1569 * array.
1570 * To create on a pre-existing array a different method will be called.
1571 * This one is just for fresh drives.
1572 *
1573 * We need to create the entire 'ddf' structure which includes:
1574 * DDF headers - these are easy.
1575 * Controller data - a Sector describing this controller .. not that
1576 * this is a controller exactly.
1577 * Physical Disk Record - one entry per device, so
1578 * leave plenty of space.
1579 * Virtual Disk Records - again, just leave plenty of space.
1580 * This just lists VDs, doesn't give details
1581 * Config records - describes the VDs that use this disk
1582 * DiskData - describes 'this' device.
1583 * BadBlockManagement - empty
1584 * Diag Space - empty
1585 * Vendor Logs - Could we put bitmaps here?
1586 *
1587 */
1588 struct ddf_super *ddf;
1589 char hostname[17];
1590 int hostlen;
a322f70c
DW		 1591 int max_phys_disks, max_virt_disks;
1592 unsigned long long sector;
1593 int clen;
1594 int i;
1595 int pdsize, vdsize;
1596 struct phys_disk *pd;
1597 struct virtual_disk *vd;
1598
78e44928 1599 if (st->sb)
955e9ea1 1600 return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
ba7eb04f 1601
3d2c4fc7
DW		 1602 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1603 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1604 return 0;
1605 }
6264b437 1606 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW		 1607 ddf->dlist = NULL; /* no physical disks yet */
1608 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW		 1609 st->sb = ddf;
1610
1611 if (info == NULL) {
1612 /* zeroing superblock */
1613 return 0;
1614 }
a322f70c
DW		 1615
1616 /* At least 32MB *must* be reserved for the ddf. So let's just
1617 * start 32MB from the end, and put the primary header there.
1618 * Don't do secondary for now.
1619 * We don't know exactly where that will be yet as it could be
1620 * different on each device. To just set up the lengths.
1621 *
1622 */
1623
1624 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1625 make_header_guid(ddf->anchor.guid);
a322f70c 1626
59e36268 1627 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW		 1628 ddf->anchor.seq = __cpu_to_be32(1);
1629 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1630 ddf->anchor.openflag = 0xFF;
1631 ddf->anchor.foreignflag = 0;
1632 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1633 ddf->anchor.pad0 = 0xff;
1634 memset(ddf->anchor.pad1, 0xff, 12);
1635 memset(ddf->anchor.header_ext, 0xff, 32);
1636 ddf->anchor.primary_lba = ~(__u64)0;
1637 ddf->anchor.secondary_lba = ~(__u64)0;
1638 ddf->anchor.type = DDF_HEADER_ANCHOR;
1639 memset(ddf->anchor.pad2, 0xff, 3);
1640 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1641 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1642 of 32M reserved.. */
1643 max_phys_disks = 1023; /* Should be enough */
1644 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1645 max_virt_disks = 255;
1646 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1647 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1648 ddf->max_part = 64;
8c3b8c2c 1649 ddf->mppe = 256;
59e36268
NB		 1650 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1651 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1652 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1653 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW		 1654 /* controller sections is one sector long immediately
1655 * after the ddf header */
1656 sector = 1;
1657 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1658 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1659 sector += 1;
1660
1661 /* phys is 8 sectors after that */
1662 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1663 sizeof(struct phys_disk_entry)*max_phys_disks,
1664 512);
1665 switch(pdsize/512) {
1666 case 2: case 8: case 32: case 128: case 512: break;
1667 default: abort();
1668 }
1669 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1670 ddf->anchor.phys_section_length =
1671 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1672 sector += pdsize/512;
1673
1674 /* virt is another 32 sectors */
1675 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1676 sizeof(struct virtual_entry) * max_virt_disks,
1677 512);
1678 switch(vdsize/512) {
1679 case 2: case 8: case 32: case 128: case 512: break;
1680 default: abort();
1681 }
1682 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1683 ddf->anchor.virt_section_length =
1684 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1685 sector += vdsize/512;
1686
59e36268 1687 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW		 1688 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1689 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1690 sector += clen;
1691
1692 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1693 ddf->anchor.data_section_length = __cpu_to_be32(1);
1694 sector += 1;
1695
1696 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1697 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1698 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1699 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1700 ddf->anchor.vendor_length = __cpu_to_be32(0);
1701 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1702
1703 memset(ddf->anchor.pad4, 0xff, 256);
1704
1705 memcpy(&ddf->primary, &ddf->anchor, 512);
1706 memcpy(&ddf->secondary, &ddf->anchor, 512);
1707
1708 ddf->primary.openflag = 1; /* I guess.. */
1709 ddf->primary.type = DDF_HEADER_PRIMARY;
1710
1711 ddf->secondary.openflag = 1; /* I guess.. */
1712 ddf->secondary.type = DDF_HEADER_SECONDARY;
1713
1714 ddf->active = &ddf->primary;
1715
1716 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1717
1718 /* 24 more bytes of fiction required.
1719 * first 8 are a 'vendor-id' - "Linux-MD"
1720 * Remaining 16 are serial number.... maybe a hostname would do?
1721 */
1722 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW		 1723 gethostname(hostname, sizeof(hostname));
1724 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW		 1725 hostlen = strlen(hostname);
1726 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1727 for (i = strlen(T10) ; i+hostlen < 24; i++)
1728 ddf->controller.guid[i] = ' ';
1729
1730 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1731 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1732 ddf->controller.type.sub_vendor_id = 0;
1733 ddf->controller.type.sub_device_id = 0;
1734 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1735 memset(ddf->controller.pad, 0xff, 8);
1736 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N		 1737 if (homehost && strlen(homehost) < 440)
1738 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1739
3d2c4fc7
DW		 1740 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1741 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1742 return 0;
1743 }
6416d527 1744 ddf->phys = pd;
a322f70c
DW		 1745 ddf->pdsize = pdsize;
1746
1747 memset(pd, 0xff, pdsize);
1748 memset(pd, 0, sizeof(*pd));
076515ba 1749 pd->magic = DDF_PHYS_RECORDS_MAGIC;
a322f70c
DW		 1750 pd->used_pdes = __cpu_to_be16(0);
1751 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1752 memset(pd->pad, 0xff, 52);
1753
3d2c4fc7
DW		 1754 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1755 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1756 return 0;
1757 }
6416d527 1758 ddf->virt = vd;
a322f70c
DW		 1759 ddf->vdsize = vdsize;
1760 memset(vd, 0, vdsize);
1761 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1762 vd->populated_vdes = __cpu_to_be16(0);
1763 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1764 memset(vd->pad, 0xff, 52);
1765
5f8097be
NB		 1766 for (i=0; i<max_virt_disks; i++)
1767 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1768
a322f70c 1769 st->sb = ddf;
18a2f463 1770 ddf->updates_pending = 1;
a322f70c
DW		 1771 return 1;
1772}
1773
5f8097be
NB		 1774static int chunk_to_shift(int chunksize)
1775{
1776 return ffs(chunksize/512)-1;
1777}
1778
1779static int level_to_prl(int level)
1780{
1781 switch (level) {
1782 case LEVEL_LINEAR: return DDF_CONCAT;
1783 case 0: return DDF_RAID0;
1784 case 1: return DDF_RAID1;
1785 case 4: return DDF_RAID4;
1786 case 5: return DDF_RAID5;
1787 case 6: return DDF_RAID6;
1788 default: return -1;
1789 }
1790}
1791static int layout_to_rlq(int level, int layout, int raiddisks)
1792{
1793 switch(level) {
1794 case 0:
1795 return DDF_RAID0_SIMPLE;
1796 case 1:
1797 switch(raiddisks) {
1798 case 2: return DDF_RAID1_SIMPLE;
1799 case 3: return DDF_RAID1_MULTI;
1800 default: return -1;
1801 }
1802 case 4:
1803 switch(layout) {
1804 case 0: return DDF_RAID4_N;
1805 }
1806 break;
1807 case 5:
5f8097be
NB		 1808 switch(layout) {
1809 case ALGORITHM_LEFT_ASYMMETRIC:
1810 return DDF_RAID5_N_RESTART;
1811 case ALGORITHM_RIGHT_ASYMMETRIC:
b640a252 1812 return DDF_RAID5_0_RESTART;
5f8097be
NB		 1813 case ALGORITHM_LEFT_SYMMETRIC:
1814 return DDF_RAID5_N_CONTINUE;
1815 case ALGORITHM_RIGHT_SYMMETRIC:
1816 return -1; /* not mentioned in standard */
1817 }
b640a252
N		 1818 case 6:
1819 switch(layout) {
1820 case ALGORITHM_ROTATING_N_RESTART:
1821 return DDF_RAID5_N_RESTART;
1822 case ALGORITHM_ROTATING_ZERO_RESTART:
1823 return DDF_RAID6_0_RESTART;
1824 case ALGORITHM_ROTATING_N_CONTINUE:
1825 return DDF_RAID5_N_CONTINUE;
1826 }
5f8097be
NB		 1827 }
1828 return -1;
1829}
1830
598f0d58
NB		 1831static int rlq_to_layout(int rlq, int prl, int raiddisks)
1832{
1833 switch(prl) {
1834 case DDF_RAID0:
1835 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1836 case DDF_RAID1:
1837 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1838 on raiddisks*/
1839 case DDF_RAID4:
1840 switch(rlq) {
1841 case DDF_RAID4_N:
1842 return 0;
1843 default:
1844 /* not supported */
1845 return -1; /* FIXME this isn't checked */
1846 }
1847 case DDF_RAID5:
598f0d58
NB		 1848 switch(rlq) {
1849 case DDF_RAID5_N_RESTART:
1850 return ALGORITHM_LEFT_ASYMMETRIC;
1851 case DDF_RAID5_0_RESTART:
1852 return ALGORITHM_RIGHT_ASYMMETRIC;
1853 case DDF_RAID5_N_CONTINUE:
1854 return ALGORITHM_LEFT_SYMMETRIC;
1855 default:
1856 return -1;
1857 }
59e36268
NB		 1858 case DDF_RAID6:
1859 switch(rlq) {
1860 case DDF_RAID5_N_RESTART:
b640a252 1861 return ALGORITHM_ROTATING_N_RESTART;
59e36268 1862 case DDF_RAID6_0_RESTART:
b640a252 1863 return ALGORITHM_ROTATING_ZERO_RESTART;
59e36268 1864 case DDF_RAID5_N_CONTINUE:
b640a252 1865 return ALGORITHM_ROTATING_N_CONTINUE;
59e36268
NB		 1866 default:
1867 return -1;
1868 }
598f0d58
NB		 1869 }
1870 return -1;
1871}
1872
0e600426 1873#ifndef MDASSEMBLE
59e36268
NB		 1874struct extent {
1875 unsigned long long start, size;
1876};
78e44928 1877static int cmp_extent(const void *av, const void *bv)
59e36268
NB		 1878{
1879 const struct extent *a = av;
1880 const struct extent *b = bv;
1881 if (a->start < b->start)
1882 return -1;
1883 if (a->start > b->start)
1884 return 1;
1885 return 0;
1886}
1887
78e44928 1888static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB		 1889{
1890 /* find a list of used extents on the give physical device
1891 * (dnum) of the given ddf.
1892 * Return a malloced array of 'struct extent'
1893
1894FIXME ignore DDF_Legacy devices?
1895
1896 */
1897 struct extent *rv;
1898 int n = 0;
1899 int i, j;
1900
1901 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1902 if (!rv)
1903 return NULL;
1904
1905 for (i = 0; i < ddf->max_part; i++) {
1906 struct vcl *v = dl->vlist[i];
1907 if (v == NULL)
1908 continue;
1909 for (j=0; j < v->conf.prim_elmnt_count; j++)
1910 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1911 /* This device plays role 'j' in 'v'. */
1912 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1913 rv[n].size = __be64_to_cpu(v->conf.blocks);
1914 n++;
1915 break;
1916 }
1917 }
1918 qsort(rv, n, sizeof(*rv), cmp_extent);
1919
1920 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1921 rv[n].size = 0;
1922 return rv;
1923}
0e600426 1924#endif
59e36268 1925
5f8097be
NB		 1926static int init_super_ddf_bvd(struct supertype *st,
1927 mdu_array_info_t *info,
1928 unsigned long long size,
1929 char *name, char *homehost,
1930 int *uuid)
1931{
1932 /* We are creating a BVD inside a pre-existing container.
1933 * so st->sb is already set.
1934 * We need to create a new vd_config and a new virtual_entry
1935 */
1936 struct ddf_super *ddf = st->sb;
1937 int venum;
1938 struct virtual_entry *ve;
1939 struct vcl *vcl;
1940 struct vd_config *vc;
5f8097be
NB		 1941
1942 if (__be16_to_cpu(ddf->virt->populated_vdes)
1943 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1944 fprintf(stderr, Name": This ddf already has the "
1945 "maximum of %d virtual devices\n",
1946 __be16_to_cpu(ddf->virt->max_vdes));
1947 return 0;
1948 }
1949
1950 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1951 if (all_ff(ddf->virt->entries[venum].guid))
1952 break;
1953 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1954 fprintf(stderr, Name ": Cannot find spare slot for "
1955 "virtual disk - DDF is corrupt\n");
1956 return 0;
1957 }
1958 ve = &ddf->virt->entries[venum];
1959
1960 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1961 * timestamp, random number
1962 */
1963 make_header_guid(ve->guid);
1964 ve->unit = __cpu_to_be16(info->md_minor);
1965 ve->pad0 = 0xFFFF;
1966 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1967 ve->type = 0;
7a7cc504
NB		 1968 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1969 if (info->state & 1) /* clean */
1970 ve->init_state = DDF_init_full;
1971 else
1972 ve->init_state = DDF_init_not;
1973
5f8097be
NB		 1974 memset(ve->pad1, 0xff, 14);
1975 memset(ve->name, ' ', 16);
1976 if (name)
1977 strncpy(ve->name, name, 16);
1978 ddf->virt->populated_vdes =
1979 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1980
1981 /* Now create a new vd_config */
3d2c4fc7
DW		 1982 if (posix_memalign((void**)&vcl, 512,
1983 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
1984 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
1985 return 0;
1986 }
8c3b8c2c 1987 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268 1988 vcl->vcnum = venum;
f7e7067b 1989 sprintf(st->subarray, "%d", venum);
59e36268 1990 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB		 1991
1992 vc = &vcl->conf;
1993
1994 vc->magic = DDF_VD_CONF_MAGIC;
1995 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
1996 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
1997 vc->seqnum = __cpu_to_be32(1);
1998 memset(vc->pad0, 0xff, 24);
1999 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
2000 vc->chunk_shift = chunk_to_shift(info->chunk_size);
2001 vc->prl = level_to_prl(info->level);
2002 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
2003 vc->sec_elmnt_count = 1;
2004 vc->sec_elmnt_seq = 0;
2005 vc->srl = 0;
2006 vc->blocks = __cpu_to_be64(info->size * 2);
2007 vc->array_blocks = __cpu_to_be64(
2008 calc_array_size(info->level, info->raid_disks, info->layout,
2009 info->chunk_size, info->size*2));
2010 memset(vc->pad1, 0xff, 8);
2011 vc->spare_refs[0] = 0xffffffff;
2012 vc->spare_refs[1] = 0xffffffff;
2013 vc->spare_refs[2] = 0xffffffff;
2014 vc->spare_refs[3] = 0xffffffff;
2015 vc->spare_refs[4] = 0xffffffff;
2016 vc->spare_refs[5] = 0xffffffff;
2017 vc->spare_refs[6] = 0xffffffff;
2018 vc->spare_refs[7] = 0xffffffff;
2019 memset(vc->cache_pol, 0, 8);
2020 vc->bg_rate = 0x80;
2021 memset(vc->pad2, 0xff, 3);
2022 memset(vc->pad3, 0xff, 52);
2023 memset(vc->pad4, 0xff, 192);
2024 memset(vc->v0, 0xff, 32);
2025 memset(vc->v1, 0xff, 32);
2026 memset(vc->v2, 0xff, 16);
2027 memset(vc->v3, 0xff, 16);
2028 memset(vc->vendor, 0xff, 32);
598f0d58 2029
8c3b8c2c 2030 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2031 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be
NB		 2032
2033 vcl->next = ddf->conflist;
2034 ddf->conflist = vcl;
d2ca6449 2035 ddf->currentconf = vcl;
18a2f463 2036 ddf->updates_pending = 1;
5f8097be
NB		 2037 return 1;
2038}
2039
0e600426 2040#ifndef MDASSEMBLE
5f8097be
NB		 2041static void add_to_super_ddf_bvd(struct supertype *st,
2042 mdu_disk_info_t *dk, int fd, char *devname)
2043{
2044 /* fd and devname identify a device with-in the ddf container (st).
2045 * dk identifies a location in the new BVD.
2046 * We need to find suitable free space in that device and update
2047 * the phys_refnum and lba_offset for the newly created vd_config.
2048 * We might also want to update the type in the phys_disk
5575e7d9 2049 * section.
8592f29d
N		 2050 *
2051 * Alternately: fd == -1 and we have already chosen which device to
2052 * use and recorded in dlist->raid_disk;
5f8097be
NB		 2053 */
2054 struct dl *dl;
2055 struct ddf_super *ddf = st->sb;
2056 struct vd_config *vc;
2057 __u64 *lba_offset;
7a7cc504 2058 int working;
5575e7d9 2059 int i;
59e36268
NB		 2060 unsigned long long blocks, pos, esize;
2061 struct extent *ex;
5f8097be 2062
8592f29d
N		 2063 if (fd == -1) {
2064 for (dl = ddf->dlist; dl ; dl = dl->next)
2065 if (dl->raiddisk == dk->raid_disk)
2066 break;
2067 } else {
2068 for (dl = ddf->dlist; dl ; dl = dl->next)
2069 if (dl->major == dk->major &&
2070 dl->minor == dk->minor)
2071 break;
2072 }
5f8097be
NB		 2073 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2074 return;
2075
d2ca6449
NB		 2076 vc = &ddf->currentconf->conf;
2077 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB		 2078
2079 ex = get_extents(ddf, dl);
2080 if (!ex)
2081 return;
2082
2083 i = 0; pos = 0;
2084 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB		 2085 if (ddf->currentconf->block_sizes)
2086 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB		 2087
2088 do {
2089 esize = ex[i].start - pos;
2090 if (esize >= blocks)
2091 break;
2092 pos = ex[i].start + ex[i].size;
2093 i++;
2094 } while (ex[i-1].size);
2095
2096 free(ex);
2097 if (esize < blocks)
2098 return;
2099
d2ca6449 2100 ddf->currentdev = dk->raid_disk;
5f8097be 2101 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2102 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2103
5575e7d9
NB		 2104 for (i=0; i < ddf->max_part ; i++)
2105 if (dl->vlist[i] == NULL)
2106 break;
2107 if (i == ddf->max_part)
2108 return;
d2ca6449 2109 dl->vlist[i] = ddf->currentconf;
5f8097be 2110
8592f29d
N		 2111 if (fd >= 0)
2112 dl->fd = fd;
2113 if (devname)
2114 dl->devname = devname;
7a7cc504
NB		 2115
2116 /* Check how many working raid_disks, and if we can mark
2117 * array as optimal yet
2118 */
2119 working = 0;
5575e7d9 2120
7a7cc504
NB		 2121 for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
2122 if (vc->phys_refnum[i] != 0xffffffff)
2123 working++;
59e36268 2124
5575e7d9 2125 /* Find which virtual_entry */
d2ca6449 2126 i = ddf->currentconf->vcnum;
7a7cc504 2127 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2128 ddf->virt->entries[i].state =
2129 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB		 2130 | DDF_state_optimal;
2131
2132 if (vc->prl == DDF_RAID6 &&
2133 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2134 ddf->virt->entries[i].state =
2135 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2136 | DDF_state_part_optimal;
5575e7d9
NB		 2137
2138 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2139 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2140 ddf->updates_pending = 1;
5f8097be
NB		 2141}
2142
a322f70c
DW		 2143/* add a device to a container, either while creating it or while
2144 * expanding a pre-existing container
2145 */
f20c3968 2146static int add_to_super_ddf(struct supertype *st,
a322f70c
DW		 2147 mdu_disk_info_t *dk, int fd, char *devname)
2148{
2149 struct ddf_super *ddf = st->sb;
2150 struct dl *dd;
2151 time_t now;
2152 struct tm *tm;
2153 unsigned long long size;
2154 struct phys_disk_entry *pde;
2155 int n, i;
2156 struct stat stb;
2157
78e44928
NB		 2158 if (ddf->currentconf) {
2159 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2160 return 0;
78e44928
NB		 2161 }
2162
a322f70c
DW		 2163 /* This is device numbered dk->number. We need to create
2164 * a phys_disk entry and a more detailed disk_data entry.
2165 */
2166 fstat(fd, &stb);
3d2c4fc7
DW		 2167 if (posix_memalign((void**)&dd, 512,
2168 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2169 fprintf(stderr, Name
2170 ": %s could allocate buffer for new disk, aborting\n",
2171 __func__);
f20c3968 2172 return 1;
3d2c4fc7 2173 }
a322f70c
DW		 2174 dd->major = major(stb.st_rdev);
2175 dd->minor = minor(stb.st_rdev);
2176 dd->devname = devname;
a322f70c 2177 dd->fd = fd;
b2280677 2178 dd->spare = NULL;
a322f70c
DW		 2179
2180 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2181 now = time(0);
2182 tm = localtime(&now);
2183 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2184 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
bfb7ea78
N		 2185 *(__u32*)(dd->disk.guid + 16) = random32();
2186 *(__u32*)(dd->disk.guid + 20) = random32();
a322f70c 2187
59e36268
NB		 2188 do {
2189 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2190 dd->disk.refnum = random32();
59e36268
NB		 2191 for (i = __be16_to_cpu(ddf->active->max_pd_entries) - 1;
2192 i >= 0; i--)
2193 if (ddf->phys->entries[i].refnum == dd->disk.refnum)
2194 break;
2195 } while (i >= 0);
2196
a322f70c
DW		 2197 dd->disk.forced_ref = 1;
2198 dd->disk.forced_guid = 1;
2199 memset(dd->disk.vendor, ' ', 32);
2200 memcpy(dd->disk.vendor, "Linux", 5);
2201 memset(dd->disk.pad, 0xff, 442);
b2280677 2202 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW		 2203 dd->vlist[i] = NULL;
2204
2205 n = __be16_to_cpu(ddf->phys->used_pdes);
2206 pde = &ddf->phys->entries[n];
5575e7d9
NB		 2207 dd->pdnum = n;
2208
2cc2983d
N		 2209 if (st->update_tail) {
2210 int len = (sizeof(struct phys_disk) +
2211 sizeof(struct phys_disk_entry));
2212 struct phys_disk *pd;
2213
2214 pd = malloc(len);
2215 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2216 pd->used_pdes = __cpu_to_be16(n);
2217 pde = &pd->entries[0];
2218 dd->mdupdate = pd;
2219 } else {
2220 n++;
2221 ddf->phys->used_pdes = __cpu_to_be16(n);
2222 }
a322f70c
DW		 2223
2224 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2225 pde->refnum = dd->disk.refnum;
5575e7d9 2226 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW		 2227 pde->state = __cpu_to_be16(DDF_Online);
2228 get_dev_size(fd, NULL, &size);
2229 /* We are required to reserve 32Meg, and record the size in sectors */
2230 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2231 sprintf(pde->path, "%17.17s","Information: nil") ;
2232 memset(pde->pad, 0xff, 6);
2233
d2ca6449 2234 dd->size = size >> 9;
2cc2983d
N		 2235 if (st->update_tail) {
2236 dd->next = ddf->add_list;
2237 ddf->add_list = dd;
2238 } else {
2239 dd->next = ddf->dlist;
2240 ddf->dlist = dd;
2241 ddf->updates_pending = 1;
2242 }
f20c3968
DW		 2243
2244 return 0;
a322f70c
DW		 2245}
2246
2247/*
2248 * This is the write_init_super method for a ddf container. It is
2249 * called when creating a container or adding another device to a
2250 * container.
2251 */
2252
6416d527 2253static unsigned char null_conf[4096+512];
18a2f463 2254
7a7cc504 2255static int __write_init_super_ddf(struct supertype *st, int do_close)
a322f70c
DW		 2256{
2257
2258 struct ddf_super *ddf = st->sb;
2259 int i;
2260 struct dl *d;
2261 int n_config;
2262 int conf_size;
175593bf
DW		 2263 int attempts = 0;
2264 int successes = 0;
a322f70c
DW		 2265 unsigned long long size, sector;
2266
175593bf
DW		 2267 /* try to write updated metadata,
2268 * if we catch a failure move on to the next disk
2269 */
a322f70c
DW		 2270 for (d = ddf->dlist; d; d=d->next) {
2271 int fd = d->fd;
2272
2273 if (fd < 0)
2274 continue;
2275
175593bf 2276 attempts++;
a322f70c
DW		 2277 /* We need to fill in the primary, (secondary) and workspace
2278 * lba's in the headers, set their checksums,
2279 * Also checksum phys, virt....
2280 *
2281 * Then write everything out, finally the anchor is written.
2282 */
2283 get_dev_size(fd, NULL, &size);
2284 size /= 512;
2285 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2286 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2287 ddf->anchor.seq = __cpu_to_be32(1);
2288 memcpy(&ddf->primary, &ddf->anchor, 512);
2289 memcpy(&ddf->secondary, &ddf->anchor, 512);
2290
2291 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2292 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2293 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2294
2295 ddf->primary.openflag = 0;
2296 ddf->primary.type = DDF_HEADER_PRIMARY;
2297
2298 ddf->secondary.openflag = 0;
2299 ddf->secondary.type = DDF_HEADER_SECONDARY;
2300
2301 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2302 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2303
2304 sector = size - 16*1024*2;
2305 lseek64(fd, sector<<9, 0);
175593bf
DW		 2306 if (write(fd, &ddf->primary, 512) < 0)
2307 continue;
a322f70c
DW		 2308
2309 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW		 2310 if (write(fd, &ddf->controller, 512) < 0)
2311 continue;
a322f70c
DW		 2312
2313 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2314
175593bf
DW		 2315 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2316 continue;
a322f70c
DW		 2317
2318 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW		 2319 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2320 continue;
a322f70c
DW		 2321
2322 /* Now write lots of config records. */
8c3b8c2c
NB		 2323 n_config = ddf->max_part;
2324 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW		 2325 for (i = 0 ; i <= n_config ; i++) {
2326 struct vcl *c = d->vlist[i];
b2280677
NB		 2327 if (i == n_config)
2328 c = (struct vcl*)d->spare;
a322f70c
DW		 2329
2330 if (c) {
2331 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW		 2332 if (write(fd, &c->conf, conf_size) < 0)
2333 break;
a322f70c 2334 } else {
6416d527 2335 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB		 2336 if (null_conf[0] != 0xff)
2337 memset(null_conf, 0xff, sizeof(null_conf));
2338 int togo = conf_size;
6416d527 2339 while (togo > sizeof(null_conf)-512) {
175593bf
DW		 2340 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2341 break;
6416d527 2342 togo -= sizeof(null_conf)-512;
18a2f463 2343 }
175593bf
DW		 2344 if (write(fd, null_aligned, togo) < 0)
2345 break;
a322f70c
DW		 2346 }
2347 }
175593bf
DW		 2348 if (i <= n_config)
2349 continue;
a322f70c 2350 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW		 2351 if (write(fd, &d->disk, 512) < 0)
2352 continue;
a322f70c
DW		 2353
2354 /* Maybe do the same for secondary */
2355
2356 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW		 2357 if (write(fd, &ddf->anchor, 512) < 0)
2358 continue;
2359 successes++;
2360 }
2361
2362 if (do_close)
2363 for (d = ddf->dlist; d; d=d->next) {
2364 close(d->fd);
7a7cc504
NB		 2365 d->fd = -1;
2366 }
175593bf
DW		 2367
2368 return attempts != successes;
a322f70c 2369}
7a7cc504
NB		 2370
2371static int write_init_super_ddf(struct supertype *st)
2372{
9b1fb677
DW		 2373 struct ddf_super *ddf = st->sb;
2374 struct vcl *currentconf = ddf->currentconf;
2375
2376 /* we are done with currentconf reset it to point st at the container */
2377 ddf->currentconf = NULL;
edd8d13c
NB		 2378
2379 if (st->update_tail) {
2380 /* queue the virtual_disk and vd_config as metadata updates */
2381 struct virtual_disk *vd;
2382 struct vd_config *vc;
edd8d13c
NB		 2383 int len;
2384
9b1fb677 2385 if (!currentconf) {
2cc2983d
N		 2386 int len = (sizeof(struct phys_disk) +
2387 sizeof(struct phys_disk_entry));
2388
2389 /* adding a disk to the container. */
2390 if (!ddf->add_list)
2391 return 0;
2392
2393 append_metadata_update(st, ddf->add_list->mdupdate, len);
2394 ddf->add_list->mdupdate = NULL;
2395 return 0;
2396 }
2397
2398 /* Newly created VD */
2399
edd8d13c
NB		 2400 /* First the virtual disk. We have a slightly fake header */
2401 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2402 vd = malloc(len);
2403 *vd = *ddf->virt;
9b1fb677
DW		 2404 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2405 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB		 2406 append_metadata_update(st, vd, len);
2407
2408 /* Then the vd_config */
2409 len = ddf->conf_rec_len * 512;
2410 vc = malloc(len);
9b1fb677 2411 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB		 2412 append_metadata_update(st, vc, len);
2413
2414 /* FIXME I need to close the fds! */
2415 return 0;
2416 } else
2417 return __write_init_super_ddf(st, 1);
7a7cc504
NB		 2418}
2419
a322f70c
DW		 2420#endif
2421
2422static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2423{
2424 /* We must reserve the last 32Meg */
2425 if (devsize <= 32*1024*2)
2426 return 0;
2427 return devsize - 32*1024*2;
2428}
2429
2430#ifndef MDASSEMBLE
8592f29d
N		 2431
2432static int reserve_space(struct supertype *st, int raiddisks,
2433 unsigned long long size, int chunk,
2434 unsigned long long *freesize)
2435{
2436 /* Find 'raiddisks' spare extents at least 'size' big (but
2437 * only caring about multiples of 'chunk') and remember
2438 * them.
2439 * If the cannot be found, fail.
2440 */
2441 struct dl *dl;
2442 struct ddf_super *ddf = st->sb;
2443 int cnt = 0;
2444
2445 for (dl = ddf->dlist; dl ; dl=dl->next) {
2446 dl->raiddisk = -1;
2447 dl->esize = 0;
2448 }
2449 /* Now find largest extent on each device */
2450 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2451 struct extent *e = get_extents(ddf, dl);
2452 unsigned long long pos = 0;
2453 int i = 0;
2454 int found = 0;
2455 unsigned long long minsize = size;
2456
2457 if (size == 0)
2458 minsize = chunk;
2459
2460 if (!e)
2461 continue;
2462 do {
2463 unsigned long long esize;
2464 esize = e[i].start - pos;
2465 if (esize >= minsize) {
2466 found = 1;
2467 minsize = esize;
2468 }
2469 pos = e[i].start + e[i].size;
2470 i++;
2471 } while (e[i-1].size);
2472 if (found) {
2473 cnt++;
2474 dl->esize = minsize;
2475 }
2476 free(e);
2477 }
2478 if (cnt < raiddisks) {
2479 fprintf(stderr, Name ": not enough devices with space to create array.\n");
2480 return 0; /* No enough free spaces large enough */
2481 }
2482 if (size == 0) {
2483 /* choose the largest size of which there are at least 'raiddisk' */
2484 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2485 struct dl *dl2;
2486 if (dl->esize <= size)
2487 continue;
2488 /* This is bigger than 'size', see if there are enough */
2489 cnt = 0;
2490 for (dl2 = dl; dl2 ; dl2=dl2->next)
2491 if (dl2->esize >= dl->esize)
2492 cnt++;
2493 if (cnt >= raiddisks)
2494 size = dl->esize;
2495 }
2496 if (chunk) {
2497 size = size / chunk;
2498 size *= chunk;
2499 }
2500 *freesize = size;
2501 if (size < 32) {
2502 fprintf(stderr, Name ": not enough spare devices to create array.\n");
2503 return 0;
2504 }
2505 }
2506 /* We have a 'size' of which there are enough spaces.
2507 * We simply do a first-fit */
2508 cnt = 0;
2509 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
2510 if (dl->esize < size)
2511 continue;
2512
2513 dl->raiddisk = cnt;
2514 cnt++;
2515 }
2516 return 1;
2517}
2518
2519
2520
2c514b71
NB		 2521static int
2522validate_geometry_ddf_container(struct supertype *st,
2523 int level, int layout, int raiddisks,
2524 int chunk, unsigned long long size,
2525 char *dev, unsigned long long *freesize,
2526 int verbose);
78e44928
NB		 2527
2528static int validate_geometry_ddf_bvd(struct supertype *st,
2529 int level, int layout, int raiddisks,
2530 int chunk, unsigned long long size,
2c514b71
NB		 2531 char *dev, unsigned long long *freesize,
2532 int verbose);
78e44928
NB		 2533
2534static int validate_geometry_ddf(struct supertype *st,
2c514b71
NB		 2535 int level, int layout, int raiddisks,
2536 int chunk, unsigned long long size,
2537 char *dev, unsigned long long *freesize,
2538 int verbose)
a322f70c
DW		 2539{
2540 int fd;
2541 struct mdinfo *sra;
2542 int cfd;
2543
2544 /* ddf potentially supports lots of things, but it depends on
2545 * what devices are offered (and maybe kernel version?)
2546 * If given unused devices, we will make a container.
2547 * If given devices in a container, we will make a BVD.
2548 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2549 */
2550
2551 if (level == LEVEL_CONTAINER) {
78e44928
NB		 2552 /* Must be a fresh device to add to a container */
2553 return validate_geometry_ddf_container(st, level, layout,
2c514b71
NB		 2554 raiddisks, chunk,
2555 size, dev, freesize,
2556 verbose);
5f8097be
NB		 2557 }
2558
78e44928
NB		 2559 if (!dev) {
2560 /* Initial sanity check. Exclude illegal levels. */
2561 int i;
2562 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2563 if (ddf_level_num[i].num2 == level)
2564 break;
b42f577a
N		 2565 if (ddf_level_num[i].num1 == MAXINT) {
2566 if (verbose)
2567 fprintf(stderr, Name ": DDF does not support level %d arrays\n",
2568 level);
78e44928 2569 return 0;
b42f577a 2570 }
78e44928 2571 /* Should check layout? etc */
8592f29d
N		 2572
2573 if (st->sb && freesize) {
2574 /* --create was given a container to create in.
2575 * So we need to check that there are enough
2576 * free spaces and return the amount of space.
2577 * We may as well remember which drives were
2578 * chosen so that add_to_super/getinfo_super
2579 * can return them.
2580 */
2581 return reserve_space(st, raiddisks, size, chunk, freesize);
2582 }
a322f70c 2583 return 1;
78e44928 2584 }
a322f70c 2585
8592f29d
N		 2586 if (st->sb) {
2587 /* A container has already been opened, so we are
2588 * creating in there. Maybe a BVD, maybe an SVD.
2589 * Should make a distinction one day.
2590 */
2591 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2592 chunk, size, dev, freesize,
2593 verbose);
2594 }
78e44928
NB		 2595 /* This is the first device for the array.
2596 * If it is a container, we read it in and do automagic allocations,
2597 * no other devices should be given.
2598 * Otherwise it must be a member device of a container, and we
2599 * do manual allocation.
2600 * Later we should check for a BVD and make an SVD.
a322f70c 2601 */
a322f70c
DW		 2602 fd = open(dev, O_RDONLY|O_EXCL, 0);
2603 if (fd >= 0) {
2604 sra = sysfs_read(fd, 0, GET_VERSION);
2605 close(fd);
2606 if (sra && sra->array.major_version == -1 &&
78e44928
NB		 2607 strcmp(sra->text_version, "ddf") == 0) {
2608
2609 /* load super */
2610 /* find space for 'n' devices. */
2611 /* remember the devices */
2612 /* Somehow return the fact that we have enough */
a322f70c
DW		 2613 }
2614
2c514b71
NB		 2615 if (verbose)
2616 fprintf(stderr,
2617 Name ": ddf: Cannot create this array "
b42f577a 2618 "on device %s - a container is required.\n",
2c514b71 2619 dev);
a322f70c
DW		 2620 return 0;
2621 }
2622 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB		 2623 if (verbose)
2624 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2625 dev, strerror(errno));
a322f70c
DW		 2626 return 0;
2627 }
2628 /* Well, it is in use by someone, maybe a 'ddf' container. */
2629 cfd = open_container(fd);
2630 if (cfd < 0) {
2631 close(fd);
2c514b71
NB		 2632 if (verbose)
2633 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2634 dev, strerror(EBUSY));
a322f70c
DW		 2635 return 0;
2636 }
2637 sra = sysfs_read(cfd, 0, GET_VERSION);
2638 close(fd);
2639 if (sra && sra->array.major_version == -1 &&
2640 strcmp(sra->text_version, "ddf") == 0) {
2641 /* This is a member of a ddf container. Load the container
2642 * and try to create a bvd
2643 */
2644 struct ddf_super *ddf;
a322f70c 2645 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 2646 st->sb = ddf;
2f6079dc 2647 st->container_dev = fd2devnum(cfd);
a322f70c 2648 close(cfd);
78e44928 2649 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2650 raiddisks, chunk, size,
2c514b71
NB		 2651 dev, freesize,
2652 verbose);
a322f70c
DW		 2653 }
2654 close(cfd);
c42ec1ed
DW		 2655 } else /* device may belong to a different container */
2656 return 0;
2657
a322f70c
DW		 2658 return 1;
2659}
2660
2c514b71
NB		 2661static int
2662validate_geometry_ddf_container(struct supertype *st,
2663 int level, int layout, int raiddisks,
2664 int chunk, unsigned long long size,
2665 char *dev, unsigned long long *freesize,
2666 int verbose)
a322f70c
DW		 2667{
2668 int fd;
2669 unsigned long long ldsize;
2670
2671 if (level != LEVEL_CONTAINER)
2672 return 0;
2673 if (!dev)
2674 return 1;
2675
2676 fd = open(dev, O_RDONLY|O_EXCL, 0);
2677 if (fd < 0) {
2c514b71
NB		 2678 if (verbose)
2679 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2680 dev, strerror(errno));
a322f70c
DW		 2681 return 0;
2682 }
2683 if (!get_dev_size(fd, dev, &ldsize)) {
2684 close(fd);
2685 return 0;
2686 }
2687 close(fd);
2688
56aca704 2689 *freesize = avail_size_ddf(st, ldsize >> 9);
ea17e7aa
N		 2690 if (*freesize == 0)
2691 return 0;
a322f70c
DW		 2692
2693 return 1;
2694}
2695
78e44928
NB		 2696static int validate_geometry_ddf_bvd(struct supertype *st,
2697 int level, int layout, int raiddisks,
2698 int chunk, unsigned long long size,
2c514b71
NB		 2699 char *dev, unsigned long long *freesize,
2700 int verbose)
a322f70c
DW		 2701{
2702 struct stat stb;
2703 struct ddf_super *ddf = st->sb;
2704 struct dl *dl;
5f8097be
NB		 2705 unsigned long long pos = 0;
2706 unsigned long long maxsize;
2707 struct extent *e;
2708 int i;
a322f70c 2709 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N		 2710 if (level == LEVEL_CONTAINER) {
2711 if (verbose)
2712 fprintf(stderr, Name ": DDF cannot create a container within an container\n");
a322f70c 2713 return 0;
b42f577a 2714 }
a322f70c
DW		 2715 /* We must have the container info already read in. */
2716 if (!ddf)
2717 return 0;
2718
5f8097be
NB		 2719 if (!dev) {
2720 /* General test: make sure there is space for
2721 * 'raiddisks' device extents of size 'size'.
2722 */
2723 unsigned long long minsize = size;
2724 int dcnt = 0;
2725 if (minsize == 0)
2726 minsize = 8;
2727 for (dl = ddf->dlist; dl ; dl = dl->next)
2728 {
2729 int found = 0;
7e1432fb 2730 pos = 0;
5f8097be
NB		 2731
2732 i = 0;
2733 e = get_extents(ddf, dl);
2734 if (!e) continue;
2735 do {
2736 unsigned long long esize;
2737 esize = e[i].start - pos;
2738 if (esize >= minsize)
2739 found = 1;
2740 pos = e[i].start + e[i].size;
2741 i++;
2742 } while (e[i-1].size);
2743 if (found)
2744 dcnt++;
2745 free(e);
2746 }
2747 if (dcnt < raiddisks) {
2c514b71
NB		 2748 if (verbose)
2749 fprintf(stderr,
2750 Name ": ddf: Not enough devices with "
2751 "space for this array (%d < %d)\n",
2752 dcnt, raiddisks);
5f8097be
NB		 2753 return 0;
2754 }
2755 return 1;
2756 }
a322f70c
DW		 2757 /* This device must be a member of the set */
2758 if (stat(dev, &stb) < 0)
2759 return 0;
2760 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2761 return 0;
2762 for (dl = ddf->dlist ; dl ; dl = dl->next) {
2763 if (dl->major == major(stb.st_rdev) &&
2764 dl->minor == minor(stb.st_rdev))
2765 break;
2766 }
5f8097be 2767 if (!dl) {
2c514b71
NB		 2768 if (verbose)
2769 fprintf(stderr, Name ": ddf: %s is not in the "
2770 "same DDF set\n",
2771 dev);
5f8097be
NB		 2772 return 0;
2773 }
2774 e = get_extents(ddf, dl);
2775 maxsize = 0;
2776 i = 0;
2777 if (e) do {
2778 unsigned long long esize;
2779 esize = e[i].start - pos;
2780 if (esize >= maxsize)
2781 maxsize = esize;
2782 pos = e[i].start + e[i].size;
2783 i++;
2784 } while (e[i-1].size);
2785 *freesize = maxsize;
a322f70c
DW		 2786 // FIXME here I am
2787
2788 return 1;
2789}
59e36268 2790
a322f70c
DW		 2791static int load_super_ddf_all(struct supertype *st, int fd,
2792 void **sbp, char *devname, int keep_fd)
2793{
2794 struct mdinfo *sra;
2795 struct ddf_super *super;
2796 struct mdinfo *sd, *best = NULL;
2797 int bestseq = 0;
2798 int seq;
2799 char nm[20];
2800 int dfd;
dab4a513
DW		 2801 int devnum = fd2devnum(fd);
2802 enum sysfs_read_flags flags;
a322f70c 2803
dab4a513
DW		 2804 flags = GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE;
2805 if (mdmon_running(devnum))
2806 flags |= SKIP_GONE_DEVS;
2807
2808 sra = sysfs_read(fd, 0, flags);
a322f70c
DW		 2809 if (!sra)
2810 return 1;
2811 if (sra->array.major_version != -1 ||
2812 sra->array.minor_version != -2 ||
2813 strcmp(sra->text_version, "ddf") != 0)
2814 return 1;
2815
6416d527 2816 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2817 return 1;
a2349791 2818 memset(super, 0, sizeof(*super));
a322f70c
DW		 2819
2820 /* first, try each device, and choose the best ddf */
2821 for (sd = sra->devs ; sd ; sd = sd->next) {
2822 int rv;
2823 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB		 2824 dfd = dev_open(nm, O_RDONLY);
2825 if (dfd < 0)
a322f70c
DW		 2826 return 2;
2827 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2828 close(dfd);
a322f70c
DW		 2829 if (rv == 0) {
2830 seq = __be32_to_cpu(super->active->seq);
2831 if (super->active->openflag)
2832 seq--;
2833 if (!best || seq > bestseq) {
2834 bestseq = seq;
2835 best = sd;
2836 }
2837 }
2838 }
2839 if (!best)
2840 return 1;
2841 /* OK, load this ddf */
2842 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2843 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2844 if (dfd < 0)
a322f70c
DW		 2845 return 1;
2846 load_ddf_headers(dfd, super, NULL);
2847 load_ddf_global(dfd, super, NULL);
2848 close(dfd);
2849 /* Now we need the device-local bits */
2850 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW		 2851 int rv;
2852
a322f70c 2853 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2854 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
7a7cc504 2855 if (dfd < 0)
a322f70c 2856 return 2;
3d2c4fc7
DW		 2857 rv = load_ddf_headers(dfd, super, NULL);
2858 if (rv == 0)
2859 rv = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2860 if (!keep_fd) close(dfd);
3d2c4fc7
DW		 2861 if (rv)
2862 return 1;
a322f70c 2863 }
f7e7067b
NB		 2864 if (st->subarray[0]) {
2865 struct vcl *v;
2866
2867 for (v = super->conflist; v; v = v->next)
2868 if (v->vcnum == atoi(st->subarray))
d2ca6449
NB		 2869 super->currentconf = v;
2870 if (!super->currentconf)
f7e7067b
NB		 2871 return 1;
2872 }
a322f70c
DW		 2873 *sbp = super;
2874 if (st->ss == NULL) {
78e44928 2875 st->ss = &super_ddf;
a322f70c
DW		 2876 st->minor_version = 0;
2877 st->max_devs = 512;
75aa18b5 2878 st->container_dev = fd2devnum(fd);
a322f70c 2879 }
352452c3 2880 st->loaded_container = 1;
a322f70c
DW		 2881 return 0;
2882}
0e600426 2883#endif /* MDASSEMBLE */
a322f70c 2884
598f0d58
NB		 2885static struct mdinfo *container_content_ddf(struct supertype *st)
2886{
2887 /* Given a container loaded by load_super_ddf_all,
2888 * extract information about all the arrays into
2889 * an mdinfo tree.
2890 *
2891 * For each vcl in conflist: create an mdinfo, fill it in,
2892 * then look for matching devices (phys_refnum) in dlist
2893 * and create appropriate device mdinfo.
2894 */
2895 struct ddf_super *ddf = st->sb;
2896 struct mdinfo *rest = NULL;
2897 struct vcl *vc;
2898
2899 for (vc = ddf->conflist ; vc ; vc=vc->next)
2900 {
598f0d58 2901 int i;
db42fa9b 2902 int j;
598f0d58
NB		 2903 struct mdinfo *this;
2904 this = malloc(sizeof(*this));
2905 memset(this, 0, sizeof(*this));
2906 this->next = rest;
2907 rest = this;
2908
598f0d58
NB		 2909 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2910 this->array.raid_disks =
2911 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB		 2912 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2913 this->array.raid_disks);
598f0d58 2914 this->array.md_minor = -1;
f35f2525
N		 2915 this->array.major_version = -1;
2916 this->array.minor_version = -2;
598f0d58
NB		 2917 this->array.ctime = DECADE +
2918 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2919 this->array.utime = DECADE +
2920 __be32_to_cpu(vc->conf.timestamp);
2921 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2922
59e36268 2923 i = vc->vcnum;
7a7cc504
NB		 2924 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2925 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2926 DDF_init_full) {
598f0d58 2927 this->array.state = 0;
ed9d66aa
NB		 2928 this->resync_start = 0;
2929 } else {
598f0d58 2930 this->array.state = 1;
b7528a20 2931 this->resync_start = MaxSector;
ed9d66aa 2932 }
db42fa9b
N		 2933 memcpy(this->name, ddf->virt->entries[i].name, 16);
2934 this->name[16]=0;
2935 for(j=0; j<16; j++)
2936 if (this->name[j] == ' ')
2937 this->name[j] = 0;
598f0d58
NB		 2938
2939 memset(this->uuid, 0, sizeof(this->uuid));
2940 this->component_size = __be64_to_cpu(vc->conf.blocks);
2941 this->array.size = this->component_size / 2;
5f2aace8 2942 this->container_member = i;
598f0d58 2943
c5afc314
N		 2944 ddf->currentconf = vc;
2945 uuid_from_super_ddf(st, this->uuid);
2946 ddf->currentconf = NULL;
2947
60f18132
NB		 2948 sprintf(this->text_version, "/%s/%d",
2949 devnum2devname(st->container_dev),
2950 this->container_member);
2951
8c3b8c2c 2952 for (i=0 ; i < ddf->mppe ; i++) {
598f0d58
NB		 2953 struct mdinfo *dev;
2954 struct dl *d;
2955
2956 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2957 continue;
2958
c0d50ca5
N		 2959 this->array.working_disks++;
2960
598f0d58
NB		 2961 for (d = ddf->dlist; d ; d=d->next)
2962 if (d->disk.refnum == vc->conf.phys_refnum[i])
2963 break;
2964 if (d == NULL)
bc17324f
N		 2965 /* Haven't found that one yet, maybe there are others */
2966 continue;
2967
598f0d58
NB		 2968 dev = malloc(sizeof(*dev));
2969 memset(dev, 0, sizeof(*dev));
2970 dev->next = this->devs;
2971 this->devs = dev;
2972
2973 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2974 dev->disk.major = d->major;
2975 dev->disk.minor = d->minor;
2976 dev->disk.raid_disk = i;
2977 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
d23534e4 2978 dev->recovery_start = MaxSector;
598f0d58 2979
120f7677
NB		 2980 dev->events = __be32_to_cpu(ddf->primary.seq);
2981 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB		 2982 dev->component_size = __be64_to_cpu(vc->conf.blocks);
2983 if (d->devname)
2984 strcpy(dev->name, d->devname);
2985 }
2986 }
2987 return rest;
2988}
2989
955e9ea1 2990static int store_super_ddf(struct supertype *st, int fd)
a322f70c 2991{
955e9ea1 2992 struct ddf_super *ddf = st->sb;
a322f70c 2993 unsigned long long dsize;
6416d527 2994 void *buf;
3d2c4fc7 2995 int rc;
a322f70c 2996
955e9ea1
DW		 2997 if (!ddf)
2998 return 1;
2999
3000 /* ->dlist and ->conflist will be set for updates, currently not
3001 * supported
3002 */
3003 if (ddf->dlist || ddf->conflist)
3004 return 1;
3005
a322f70c
DW		 3006 if (!get_dev_size(fd, NULL, &dsize))
3007 return 1;
3008
3d2c4fc7
DW		 3009 if (posix_memalign(&buf, 512, 512) != 0)
3010 return 1;
6416d527
NB		 3011 memset(buf, 0, 512);
3012
a322f70c 3013 lseek64(fd, dsize-512, 0);
3d2c4fc7 3014 rc = write(fd, buf, 512);
6416d527 3015 free(buf);
3d2c4fc7
DW		 3016 if (rc < 0)
3017 return 1;
a322f70c
DW		 3018 return 0;
3019}
3020
a19c88b8
NB		 3021static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3022{
3023 /*
3024 * return:
3025 * 0 same, or first was empty, and second was copied
3026 * 1 second had wrong number
3027 * 2 wrong uuid
3028 * 3 wrong other info
3029 */
3030 struct ddf_super *first = st->sb;
3031 struct ddf_super *second = tst->sb;
3032
3033 if (!first) {
3034 st->sb = tst->sb;
3035 tst->sb = NULL;
3036 return 0;
3037 }
3038
3039 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3040 return 2;
3041
3042 /* FIXME should I look at anything else? */
3043 return 0;
3044}
3045
0e600426 3046#ifndef MDASSEMBLE
4e5528c6
NB		 3047/*
3048 * A new array 'a' has been started which claims to be instance 'inst'
3049 * within container 'c'.
3050 * We need to confirm that the array matches the metadata in 'c' so
3051 * that we don't corrupt any metadata.
3052 */
cba0191b 3053static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3054{
2c514b71 3055 dprintf("ddf: open_new %s\n", inst);
cba0191b 3056 a->info.container_member = atoi(inst);
549e9569
NB		 3057 return 0;
3058}
3059
4e5528c6
NB		 3060/*
3061 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3062 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB		 3063 * clean up to the point (in sectors). If that cannot be recorded in the
3064 * metadata, then leave it as dirty.
3065 *
3066 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3067 * !global! virtual_disk.virtual_entry structure.
3068 */
01f157d7 3069static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3070{
4e5528c6
NB		 3071 struct ddf_super *ddf = a->container->sb;
3072 int inst = a->info.container_member;
18a2f463 3073 int old = ddf->virt->entries[inst].state;
01f157d7
N		 3074 if (consistent == 2) {
3075 /* Should check if a recovery should be started FIXME */
3076 consistent = 1;
b7941fd6 3077 if (!is_resync_complete(&a->info))
01f157d7
N		 3078 consistent = 0;
3079 }
ed9d66aa
NB		 3080 if (consistent)
3081 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3082 else
4e5528c6 3083 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB		 3084 if (old != ddf->virt->entries[inst].state)
3085 ddf->updates_pending = 1;
3086
3087 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3088 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
b7941fd6 3089 if (is_resync_complete(&a->info))
ed9d66aa 3090 ddf->virt->entries[inst].init_state |= DDF_init_full;
b7941fd6 3091 else if (a->info.resync_start == 0)
ed9d66aa 3092 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3093 else
ed9d66aa 3094 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB		 3095 if (old != ddf->virt->entries[inst].init_state)
3096 ddf->updates_pending = 1;
ed9d66aa 3097
2c514b71 3098 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
b7941fd6 3099 a->info.resync_start);
01f157d7 3100 return consistent;
fd7cde1b
DW		 3101}
3102
7a7cc504
NB		 3103/*
3104 * The state of each disk is stored in the global phys_disk structure
3105 * in phys_disk.entries[n].state.
3106 * This makes various combinations awkward.
3107 * - When a device fails in any array, it must be failed in all arrays
3108 * that include a part of this device.
3109 * - When a component is rebuilding, we cannot include it officially in the
3110 * array unless this is the only array that uses the device.
3111 *
3112 * So: when transitioning:
3113 * Online -> failed, just set failed flag. monitor will propagate
3114 * spare -> online, the device might need to be added to the array.
3115 * spare -> failed, just set failed. Don't worry if in array or not.
3116 */
8d45d196 3117static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 3118{
7a7cc504
NB		 3119 struct ddf_super *ddf = a->container->sb;
3120 int inst = a->info.container_member;
3121 struct vd_config *vc = find_vdcr(ddf, inst);
3122 int pd = find_phys(ddf, vc->phys_refnum[n]);
3123 int i, st, working;
3124
3125 if (vc == NULL) {
2c514b71 3126 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB		 3127 return;
3128 }
3129 if (pd < 0) {
3130 /* disk doesn't currently exist. If it is now in_sync,
3131 * insert it. */
3132 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
3133 /* Find dev 'n' in a->info->devs, determine the
3134 * ddf refnum, and set vc->phys_refnum and update
3135 * phys->entries[]
3136 */
3137 /* FIXME */
3138 }
3139 } else {
18a2f463 3140 int old = ddf->phys->entries[pd].state;
7a7cc504
NB		 3141 if (state & DS_FAULTY)
3142 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
3143 if (state & DS_INSYNC) {
3144 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
3145 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
3146 }
18a2f463
NB		 3147 if (old != ddf->phys->entries[pd].state)
3148 ddf->updates_pending = 1;
7a7cc504
NB		 3149 }
3150
2c514b71 3151 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 3152
7a7cc504
NB		 3153 /* Now we need to check the state of the array and update
3154 * virtual_disk.entries[n].state.
3155 * It needs to be one of "optimal", "degraded", "failed".
3156 * I don't understand 'deleted' or 'missing'.
3157 */
3158 working = 0;
3159 for (i=0; i < a->info.array.raid_disks; i++) {
3160 pd = find_phys(ddf, vc->phys_refnum[i]);
3161 if (pd < 0)
3162 continue;
57632f4a
NB		 3163 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3164 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB		 3165 == DDF_Online)
3166 working++;
3167 }
3168 state = DDF_state_degraded;
3169 if (working == a->info.array.raid_disks)
3170 state = DDF_state_optimal;
3171 else switch(vc->prl) {
3172 case DDF_RAID0:
3173 case DDF_CONCAT:
3174 case DDF_JBOD:
3175 state = DDF_state_failed;
3176 break;
3177 case DDF_RAID1:
3178 if (working == 0)
3179 state = DDF_state_failed;
3180 break;
3181 case DDF_RAID4:
3182 case DDF_RAID5:
3183 if (working < a->info.array.raid_disks-1)
3184 state = DDF_state_failed;
3185 break;
3186 case DDF_RAID6:
3187 if (working < a->info.array.raid_disks-2)
3188 state = DDF_state_failed;
3189 else if (working == a->info.array.raid_disks-1)
3190 state = DDF_state_part_optimal;
3191 break;
3192 }
3193
18a2f463
NB		 3194 if (ddf->virt->entries[inst].state !=
3195 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
3196 | state)) {
3197
3198 ddf->virt->entries[inst].state =
3199 (ddf->virt->entries[inst].state & ~DDF_state_mask)
3200 | state;
3201 ddf->updates_pending = 1;
3202 }
7a7cc504 3203
549e9569
NB		 3204}
3205
2e735d19 3206static void ddf_sync_metadata(struct supertype *st)
549e9569 3207{
7a7cc504
NB		 3208
3209 /*
3210 * Write all data to all devices.
3211 * Later, we might be able to track whether only local changes
3212 * have been made, or whether any global data has been changed,
3213 * but ddf is sufficiently weird that it probably always
3214 * changes global data ....
3215 */
18a2f463
NB		 3216 struct ddf_super *ddf = st->sb;
3217 if (!ddf->updates_pending)
3218 return;
3219 ddf->updates_pending = 0;
2e735d19 3220 __write_init_super_ddf(st, 0);
2c514b71 3221 dprintf("ddf: sync_metadata\n");
549e9569
NB		 3222}
3223
88c164f4
NB		 3224static void ddf_process_update(struct supertype *st,
3225 struct metadata_update *update)
3226{
3227 /* Apply this update to the metadata.
3228 * The first 4 bytes are a DDF_*_MAGIC which guides
3229 * our actions.
3230 * Possible update are:
3231 * DDF_PHYS_RECORDS_MAGIC
3232 * Add a new physical device. Changes to this record
3233 * only happen implicitly.
3234 * used_pdes is the device number.
3235 * DDF_VIRT_RECORDS_MAGIC
3236 * Add a new VD. Possibly also change the 'access' bits.
3237 * populated_vdes is the entry number.
3238 * DDF_VD_CONF_MAGIC
3239 * New or updated VD. the VIRT_RECORD must already
3240 * exist. For an update, phys_refnum and lba_offset
3241 * (at least) are updated, and the VD_CONF must
3242 * be written to precisely those devices listed with
3243 * a phys_refnum.
3244 * DDF_SPARE_ASSIGN_MAGIC
3245 * replacement Spare Assignment Record... but for which device?
3246 *
3247 * So, e.g.:
3248 * - to create a new array, we send a VIRT_RECORD and
3249 * a VD_CONF. Then assemble and start the array.
3250 * - to activate a spare we send a VD_CONF to add the phys_refnum
3251 * and offset. This will also mark the spare as active with
3252 * a spare-assignment record.
3253 */
3254 struct ddf_super *ddf = st->sb;
3255 __u32 *magic = (__u32*)update->buf;
3256 struct phys_disk *pd;
3257 struct virtual_disk *vd;
3258 struct vd_config *vc;
3259 struct vcl *vcl;
3260 struct dl *dl;
3261 int mppe;
3262 int ent;
3263
2c514b71 3264 dprintf("Process update %x\n", *magic);
7e1432fb 3265
88c164f4
NB		 3266 switch (*magic) {
3267 case DDF_PHYS_RECORDS_MAGIC:
3268
3269 if (update->len != (sizeof(struct phys_disk) +
3270 sizeof(struct phys_disk_entry)))
3271 return;
3272 pd = (struct phys_disk*)update->buf;
3273
3274 ent = __be16_to_cpu(pd->used_pdes);
3275 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3276 return;
3277 if (!all_ff(ddf->phys->entries[ent].guid))
3278 return;
3279 ddf->phys->entries[ent] = pd->entries[0];
3280 ddf->phys->used_pdes = __cpu_to_be16(1 +
3281 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3282 ddf->updates_pending = 1;
2cc2983d
N		 3283 if (ddf->add_list) {
3284 struct active_array *a;
3285 struct dl *al = ddf->add_list;
3286 ddf->add_list = al->next;
3287
3288 al->next = ddf->dlist;
3289 ddf->dlist = al;
3290
3291 /* As a device has been added, we should check
3292 * for any degraded devices that might make
3293 * use of this spare */
3294 for (a = st->arrays ; a; a=a->next)
3295 a->check_degraded = 1;
3296 }
88c164f4
NB		 3297 break;
3298
3299 case DDF_VIRT_RECORDS_MAGIC:
3300
3301 if (update->len != (sizeof(struct virtual_disk) +
3302 sizeof(struct virtual_entry)))
3303 return;
3304 vd = (struct virtual_disk*)update->buf;
3305
3306 ent = __be16_to_cpu(vd->populated_vdes);
3307 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3308 return;
3309 if (!all_ff(ddf->virt->entries[ent].guid))
3310 return;
3311 ddf->virt->entries[ent] = vd->entries[0];
3312 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3313 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3314 ddf->updates_pending = 1;
88c164f4
NB		 3315 break;
3316
3317 case DDF_VD_CONF_MAGIC:
2c514b71 3318 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB		 3319
3320 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
edd8d13c 3321 if (update->len != ddf->conf_rec_len * 512)
88c164f4
NB		 3322 return;
3323 vc = (struct vd_config*)update->buf;
3324 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3325 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3326 break;
2c514b71 3327 dprintf("vcl = %p\n", vcl);
88c164f4
NB		 3328 if (vcl) {
3329 /* An update, just copy the phys_refnum and lba_offset
3330 * fields
3331 */
3332 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3333 mppe * (sizeof(__u32) + sizeof(__u64)));
3334 } else {
3335 /* A new VD_CONF */
e6b9548d
DW		 3336 if (!update->space)
3337 return;
88c164f4
NB		 3338 vcl = update->space;
3339 update->space = NULL;
3340 vcl->next = ddf->conflist;
edd8d13c 3341 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB		 3342 vcl->lba_offset = (__u64*)
3343 &vcl->conf.phys_refnum[mppe];
3344 ddf->conflist = vcl;
3345 }
3346 /* Now make sure vlist is correct for each dl. */
3347 for (dl = ddf->dlist; dl; dl = dl->next) {
3348 int dn;
3349 int vn = 0;
3350 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3351 for (dn=0; dn < ddf->mppe ; dn++)
3352 if (vcl->conf.phys_refnum[dn] ==
3353 dl->disk.refnum) {
2c514b71
NB		 3354 dprintf("dev %d has %p at %d\n",
3355 dl->pdnum, vcl, vn);
88c164f4
NB		 3356 dl->vlist[vn++] = vcl;
3357 break;
3358 }
3359 while (vn < ddf->max_part)
3360 dl->vlist[vn++] = NULL;
7e1432fb
NB		 3361 if (dl->vlist[0]) {
3362 ddf->phys->entries[dl->pdnum].type &=
3363 ~__cpu_to_be16(DDF_Global_Spare);
3364 ddf->phys->entries[dl->pdnum].type |=
3365 __cpu_to_be16(DDF_Active_in_VD);
3366 }
3367 if (dl->spare) {
3368 ddf->phys->entries[dl->pdnum].type &=
3369 ~__cpu_to_be16(DDF_Global_Spare);
3370 ddf->phys->entries[dl->pdnum].type |=
3371 __cpu_to_be16(DDF_Spare);
3372 }
3373 if (!dl->vlist[0] && !dl->spare) {
3374 ddf->phys->entries[dl->pdnum].type |=
3375 __cpu_to_be16(DDF_Global_Spare);
3376 ddf->phys->entries[dl->pdnum].type &=
3377 ~__cpu_to_be16(DDF_Spare |
3378 DDF_Active_in_VD);
3379 }
88c164f4 3380 }
18a2f463 3381 ddf->updates_pending = 1;
88c164f4
NB		 3382 break;
3383 case DDF_SPARE_ASSIGN_MAGIC:
3384 default: break;
3385 }
3386}
3387
edd8d13c
NB		 3388static void ddf_prepare_update(struct supertype *st,
3389 struct metadata_update *update)
3390{
3391 /* This update arrived at managemon.
3392 * We are about to pass it to monitor.
3393 * If a malloc is needed, do it here.
3394 */
3395 struct ddf_super *ddf = st->sb;
3396 __u32 *magic = (__u32*)update->buf;
3397 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3398 if (posix_memalign(&update->space, 512,
6416d527 3399 offsetof(struct vcl, conf)
e6b9548d
DW		 3400 + ddf->conf_rec_len * 512) != 0)
3401 update->space = NULL;
edd8d13c
NB		 3402}
3403
7e1432fb
NB		 3404/*
3405 * Check if the array 'a' is degraded but not failed.
3406 * If it is, find as many spares as are available and needed and
3407 * arrange for their inclusion.
3408 * We only choose devices which are not already in the array,
3409 * and prefer those with a spare-assignment to this array.
3410 * otherwise we choose global spares - assuming always that
3411 * there is enough room.
3412 * For each spare that we assign, we return an 'mdinfo' which
3413 * describes the position for the device in the array.
3414 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3415 * the new phys_refnum and lba_offset values.
3416 *
3417 * Only worry about BVDs at the moment.
3418 */
3419static struct mdinfo *ddf_activate_spare(struct active_array *a,
3420 struct metadata_update **updates)
3421{
3422 int working = 0;
3423 struct mdinfo *d;
3424 struct ddf_super *ddf = a->container->sb;
3425 int global_ok = 0;
3426 struct mdinfo *rv = NULL;
3427 struct mdinfo *di;
3428 struct metadata_update *mu;
3429 struct dl *dl;
3430 int i;
3431 struct vd_config *vc;
3432 __u64 *lba;
3433
7e1432fb
NB		 3434 for (d = a->info.devs ; d ; d = d->next) {
3435 if ((d->curr_state & DS_FAULTY) &&
3436 d->state_fd >= 0)
3437 /* wait for Removal to happen */
3438 return NULL;
3439 if (d->state_fd >= 0)
3440 working ++;
3441 }
3442
2c514b71
NB		 3443 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3444 a->info.array.level);
7e1432fb
NB		 3445 if (working == a->info.array.raid_disks)
3446 return NULL; /* array not degraded */
3447 switch (a->info.array.level) {
3448 case 1:
3449 if (working == 0)
3450 return NULL; /* failed */
3451 break;
3452 case 4:
3453 case 5:
3454 if (working < a->info.array.raid_disks - 1)
3455 return NULL; /* failed */
3456 break;
3457 case 6:
3458 if (working < a->info.array.raid_disks - 2)
3459 return NULL; /* failed */
3460 break;
3461 default: /* concat or stripe */
3462 return NULL; /* failed */
3463 }
3464
3465 /* For each slot, if it is not working, find a spare */
3466 dl = ddf->dlist;
3467 for (i = 0; i < a->info.array.raid_disks; i++) {
3468 for (d = a->info.devs ; d ; d = d->next)
3469 if (d->disk.raid_disk == i)
3470 break;
2c514b71 3471 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB		 3472 if (d && (d->state_fd >= 0))
3473 continue;
3474
3475 /* OK, this device needs recovery. Find a spare */
3476 again:
3477 for ( ; dl ; dl = dl->next) {
3478 unsigned long long esize;
3479 unsigned long long pos;
3480 struct mdinfo *d2;
3481 int is_global = 0;
3482 int is_dedicated = 0;
3483 struct extent *ex;
3484 int j;
3485 /* If in this array, skip */
3486 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3487 if (d2->disk.major == dl->major &&
3488 d2->disk.minor == dl->minor) {
2c514b71 3489 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB		 3490 break;
3491 }
3492 if (d2)
3493 continue;
3494 if (ddf->phys->entries[dl->pdnum].type &
3495 __cpu_to_be16(DDF_Spare)) {
3496 /* Check spare assign record */
3497 if (dl->spare) {
3498 if (dl->spare->type & DDF_spare_dedicated) {
3499 /* check spare_ents for guid */
3500 for (j = 0 ;
3501 j < __be16_to_cpu(dl->spare->populated);
3502 j++) {
3503 if (memcmp(dl->spare->spare_ents[j].guid,
3504 ddf->virt->entries[a->info.container_member].guid,
3505 DDF_GUID_LEN) == 0)
3506 is_dedicated = 1;
3507 }
3508 } else
3509 is_global = 1;
3510 }
3511 } else if (ddf->phys->entries[dl->pdnum].type &
3512 __cpu_to_be16(DDF_Global_Spare)) {
3513 is_global = 1;
3514 }
3515 if ( ! (is_dedicated ||
3516 (is_global && global_ok))) {
2c514b71 3517 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB		 3518 is_dedicated, is_global);
3519 continue;
3520 }
3521
3522 /* We are allowed to use this device - is there space?
3523 * We need a->info.component_size sectors */
3524 ex = get_extents(ddf, dl);
3525 if (!ex) {
2c514b71 3526 dprintf("cannot get extents\n");
7e1432fb
NB		 3527 continue;
3528 }
3529 j = 0; pos = 0;
3530 esize = 0;
3531
3532 do {
3533 esize = ex[j].start - pos;
3534 if (esize >= a->info.component_size)
3535 break;
3536 pos = ex[i].start + ex[i].size;
3537 i++;
3538 } while (ex[i-1].size);
3539
3540 free(ex);
3541 if (esize < a->info.component_size) {
2c514b71
NB		 3542 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3543 esize, a->info.component_size);
7e1432fb
NB		 3544 /* No room */
3545 continue;
3546 }
3547
3548 /* Cool, we have a device with some space at pos */
3549 di = malloc(sizeof(*di));
79244939
DW		 3550 if (!di)
3551 continue;
7e1432fb
NB		 3552 memset(di, 0, sizeof(*di));
3553 di->disk.number = i;
3554 di->disk.raid_disk = i;
3555 di->disk.major = dl->major;
3556 di->disk.minor = dl->minor;
3557 di->disk.state = 0;
d23534e4 3558 di->recovery_start = 0;
7e1432fb
NB		 3559 di->data_offset = pos;
3560 di->component_size = a->info.component_size;
3561 di->container_member = dl->pdnum;
3562 di->next = rv;
3563 rv = di;
2c514b71
NB		 3564 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3565 i, pos);
7e1432fb
NB		 3566
3567 break;
3568 }
3569 if (!dl && ! global_ok) {
3570 /* not enough dedicated spares, try global */
3571 global_ok = 1;
3572 dl = ddf->dlist;
3573 goto again;
3574 }
3575 }
3576
3577 if (!rv)
3578 /* No spares found */
3579 return rv;
3580 /* Now 'rv' has a list of devices to return.
3581 * Create a metadata_update record to update the
3582 * phys_refnum and lba_offset values
3583 */
904c1ef7 3584 mu = malloc(sizeof(*mu));
79244939
DW		 3585 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3586 free(mu);
3587 mu = NULL;
3588 }
3589 if (!mu) {
3590 while (rv) {
3591 struct mdinfo *n = rv->next;
3592
3593 free(rv);
3594 rv = n;
3595 }
3596 return NULL;
3597 }
3598
904c1ef7 3599 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB		 3600 mu->len = ddf->conf_rec_len;
3601 mu->next = *updates;
3602 vc = find_vdcr(ddf, a->info.container_member);
3603 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3604
3605 vc = (struct vd_config*)mu->buf;
3606 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3607 for (di = rv ; di ; di = di->next) {
3608 vc->phys_refnum[di->disk.raid_disk] =
3609 ddf->phys->entries[dl->pdnum].refnum;
3610 lba[di->disk.raid_disk] = di->data_offset;
3611 }
3612 *updates = mu;
3613 return rv;
3614}
0e600426 3615#endif /* MDASSEMBLE */
7e1432fb 3616
b640a252
N		 3617static int ddf_level_to_layout(int level)
3618{
3619 switch(level) {
3620 case 0:
3621 case 1:
3622 return 0;
3623 case 5:
3624 return ALGORITHM_LEFT_SYMMETRIC;
3625 case 6:
3626 return ALGORITHM_ROTATING_N_CONTINUE;
3627 case 10:
3628 return 0x102;
3629 default:
3630 return UnSet;
3631 }
3632}
3633
a322f70c
DW		 3634struct superswitch super_ddf = {
3635#ifndef MDASSEMBLE
3636 .examine_super = examine_super_ddf,
3637 .brief_examine_super = brief_examine_super_ddf,
4737ae25 3638 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 3639 .export_examine_super = export_examine_super_ddf,
a322f70c
DW		 3640 .detail_super = detail_super_ddf,
3641 .brief_detail_super = brief_detail_super_ddf,
3642 .validate_geometry = validate_geometry_ddf,
78e44928 3643 .write_init_super = write_init_super_ddf,
0e600426 3644 .add_to_super = add_to_super_ddf,
a322f70c
DW		 3645#endif
3646 .match_home = match_home_ddf,
3647 .uuid_from_super= uuid_from_super_ddf,
3648 .getinfo_super = getinfo_super_ddf,
3649 .update_super = update_super_ddf,
3650
3651 .avail_size = avail_size_ddf,
3652
a19c88b8
NB		 3653 .compare_super = compare_super_ddf,
3654
a322f70c 3655 .load_super = load_super_ddf,
ba7eb04f 3656 .init_super = init_super_ddf,
955e9ea1 3657 .store_super = store_super_ddf,
a322f70c
DW		 3658 .free_super = free_super_ddf,
3659 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3660 .container_content = container_content_ddf,
b640a252 3661 .default_layout = ddf_level_to_layout,
a322f70c 3662
a322f70c 3663 .external = 1,
549e9569 3664
0e600426 3665#ifndef MDASSEMBLE
549e9569
NB		 3666/* for mdmon */
3667 .open_new = ddf_open_new,
ed9d66aa 3668 .set_array_state= ddf_set_array_state,
549e9569
NB		 3669 .set_disk = ddf_set_disk,
3670 .sync_metadata = ddf_sync_metadata,
88c164f4 3671 .process_update = ddf_process_update,
edd8d13c 3672 .prepare_update = ddf_prepare_update,
7e1432fb 3673 .activate_spare = ddf_activate_spare,
0e600426 3674#endif
4cce4069 3675 .name = "ddf",
a322f70c 3676};
Unnamed repository; edit this file 'description' to name the repository.