more ddf stuff
[thirdparty/mdadm.git] / super-ddf.c
CommitLineData
a322f70c
DW
1/*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2006-2007 Neil Brown <neilb@suse.de>
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"
30#include "sha1.h"
31#include <values.h>
32
33static inline int ROUND_UP(int a, int base)
34{
35 return ((a+base-1)/base)*base;
36}
37
38/* a non-official T10 name for creation GUIDs */
39static char T10[] = "Linux-MD";
40
41/* DDF timestamps are 1980 based, so we need to add
42 * second-in-decade-of-seventies to convert to linux timestamps.
43 * 10 years with 2 leap years.
44 */
45#define DECADE (3600*24*(365*10+2))
46unsigned long crc32(
47 unsigned long crc,
48 const unsigned char *buf,
49 unsigned len);
50
51/* The DDF metadata handling.
52 * DDF metadata lives at the end of the device.
53 * The last 512 byte block provides an 'anchor' which is used to locate
54 * the rest of the metadata which usually lives immediately behind the anchor.
55 *
56 * Note:
57 * - all multibyte numeric fields are bigendian.
58 * - all strings are space padded.
59 *
60 */
61
62/* Primary Raid Level (PRL) */
63#define DDF_RAID0 0x00
64#define DDF_RAID1 0x01
65#define DDF_RAID3 0x03
66#define DDF_RAID4 0x04
67#define DDF_RAID5 0x05
68#define DDF_RAID1E 0x11
69#define DDF_JBOD 0x0f
70#define DDF_CONCAT 0x1f
71#define DDF_RAID5E 0x15
72#define DDF_RAID5EE 0x25
73#define DDF_RAID6 0x16 /* Vendor unique layout */
74
75/* Raid Level Qualifier (RLQ) */
76#define DDF_RAID0_SIMPLE 0x00
77#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
78#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
79#define DDF_RAID3_0 0x00 /* parity in first extent */
80#define DDF_RAID3_N 0x01 /* parity in last extent */
81#define DDF_RAID4_0 0x00 /* parity in first extent */
82#define DDF_RAID4_N 0x01 /* parity in last extent */
83/* these apply to raid5e and raid5ee as well */
84#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
85#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
86#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
87
88#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
89#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
90
91/* Secondary RAID Level (SRL) */
92#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
93#define DDF_2MIRRORED 0x01
94#define DDF_2CONCAT 0x02
95#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
96
97/* Magic numbers */
98#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
99#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
100#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
101#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
102#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
103#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
104#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
105#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
106#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
107#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
108
109#define DDF_GUID_LEN 24
110#define DDF_REVISION "01.00.00"
111
112struct ddf_header {
113 __u32 magic;
114 __u32 crc;
115 char guid[DDF_GUID_LEN];
116 char revision[8]; /* 01.00.00 */
117 __u32 seq; /* starts at '1' */
118 __u32 timestamp;
119 __u8 openflag;
120 __u8 foreignflag;
121 __u8 enforcegroups;
122 __u8 pad0; /* 0xff */
123 __u8 pad1[12]; /* 12 * 0xff */
124 /* 64 bytes so far */
125 __u8 header_ext[32]; /* reserved: fill with 0xff */
126 __u64 primary_lba;
127 __u64 secondary_lba;
128 __u8 type;
129 __u8 pad2[3]; /* 0xff */
130 __u32 workspace_len; /* sectors for vendor space -
131 * at least 32768(sectors) */
132 __u64 workspace_lba;
133 __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
134 __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
135 __u16 max_partitions; /* i.e. max num of configuration
136 record entries per disk */
137 __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
138 *12/512) */
139 __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
140 __u8 pad3[54]; /* 0xff */
141 /* 192 bytes so far */
142 __u32 controller_section_offset;
143 __u32 controller_section_length;
144 __u32 phys_section_offset;
145 __u32 phys_section_length;
146 __u32 virt_section_offset;
147 __u32 virt_section_length;
148 __u32 config_section_offset;
149 __u32 config_section_length;
150 __u32 data_section_offset;
151 __u32 data_section_length;
152 __u32 bbm_section_offset;
153 __u32 bbm_section_length;
154 __u32 diag_space_offset;
155 __u32 diag_space_length;
156 __u32 vendor_offset;
157 __u32 vendor_length;
158 /* 256 bytes so far */
159 __u8 pad4[256]; /* 0xff */
160};
161
162/* type field */
163#define DDF_HEADER_ANCHOR 0x00
164#define DDF_HEADER_PRIMARY 0x01
165#define DDF_HEADER_SECONDARY 0x02
166
167/* The content of the 'controller section' - global scope */
168struct ddf_controller_data {
169 __u32 magic;
170 __u32 crc;
171 char guid[DDF_GUID_LEN];
172 struct controller_type {
173 __u16 vendor_id;
174 __u16 device_id;
175 __u16 sub_vendor_id;
176 __u16 sub_device_id;
177 } type;
178 char product_id[16];
179 __u8 pad[8]; /* 0xff */
180 __u8 vendor_data[448];
181};
182
183/* The content of phys_section - global scope */
184struct phys_disk {
185 __u32 magic;
186 __u32 crc;
187 __u16 used_pdes;
188 __u16 max_pdes;
189 __u8 pad[52];
190 struct phys_disk_entry {
191 char guid[DDF_GUID_LEN];
192 __u32 refnum;
193 __u16 type;
194 __u16 state;
195 __u64 config_size; /* DDF structures must be after here */
196 char path[18]; /* another horrible structure really */
197 __u8 pad[6];
198 } entries[0];
199};
200
201/* phys_disk_entry.type is a bitmap - bigendian remember */
202#define DDF_Forced_PD_GUID 1
203#define DDF_Active_in_VD 2
204#define DDF_Global_Spare 4
205#define DDF_Spare 8 /* overrides Global_spare */
206#define DDF_Foreign 16
207#define DDF_Legacy 32 /* no DDF on this device */
208
209#define DDF_Interface_mask 0xf00
210#define DDF_Interface_SCSI 0x100
211#define DDF_Interface_SAS 0x200
212#define DDF_Interface_SATA 0x300
213#define DDF_Interface_FC 0x400
214
215/* phys_disk_entry.state is a bigendian bitmap */
216#define DDF_Online 1
217#define DDF_Failed 2 /* overrides 1,4,8 */
218#define DDF_Rebuilding 4
219#define DDF_Transition 8
220#define DDF_SMART 16
221#define DDF_ReadErrors 32
222#define DDF_Missing 64
223
224/* The content of the virt_section global scope */
225struct virtual_disk {
226 __u32 magic;
227 __u32 crc;
228 __u16 populated_vdes;
229 __u16 max_vdes;
230 __u8 pad[52];
231 struct virtual_entry {
232 char guid[DDF_GUID_LEN];
233 __u16 unit;
234 __u16 pad0; /* 0xffff */
235 __u16 guid_crc;
236 __u16 type;
237 __u8 state;
238 __u8 init_state;
239 __u8 pad1[14];
240 char name[16];
241 } entries[0];
242};
243
244/* virtual_entry.type is a bitmap - bigendian */
245#define DDF_Shared 1
246#define DDF_Enforce_Groups 2
247#define DDF_Unicode 4
248#define DDF_Owner_Valid 8
249
250/* virtual_entry.state is a bigendian bitmap */
251#define DDF_state_mask 0x7
252#define DDF_state_optimal 0x0
253#define DDF_state_degraded 0x1
254#define DDF_state_deleted 0x2
255#define DDF_state_missing 0x3
256#define DDF_state_failed 0x4
257
258#define DDF_state_morphing 0x8
259#define DDF_state_inconsistent 0x10
260
261/* virtual_entry.init_state is a bigendian bitmap */
262#define DDF_initstate_mask 0x03
263#define DDF_init_not 0x00
264#define DDF_init_quick 0x01
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 {
278 __u32 magic;
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;
291 __u64 blocks;
292 __u64 array_blocks;
293 __u8 pad1[8];
294 __u32 spare_refs[8];
295 __u8 cache_pol[8];
296 __u8 bg_rate;
297 __u8 pad2[3];
298 __u8 pad3[52];
299 __u8 pad4[192];
300 __u8 v0[32]; /* reserved- 0xff */
301 __u8 v1[32]; /* reserved- 0xff */
302 __u8 v2[16]; /* reserved- 0xff */
303 __u8 v3[16]; /* reserved- 0xff */
304 __u8 vendor[32];
305 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
306 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
307 bvd are always the same size */
308};
309
310/* vd_config.cache_pol[7] is a bitmap */
311#define DDF_cache_writeback 1 /* else writethrough */
312#define DDF_cache_wadaptive 2 /* only applies if writeback */
313#define DDF_cache_readahead 4
314#define DDF_cache_radaptive 8 /* only if doing read-ahead */
315#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
316#define DDF_cache_wallowed 32 /* enable write caching */
317#define DDF_cache_rallowed 64 /* enable read caching */
318
319struct spare_assign {
320 __u32 magic;
321 __u32 crc;
322 __u32 timestamp;
323 __u8 reserved[7];
324 __u8 type;
325 __u16 populated; /* SAEs used */
326 __u16 max; /* max SAEs */
327 __u8 pad[8];
328 struct spare_assign_entry {
329 char guid[DDF_GUID_LEN];
330 __u16 secondary_element;
331 __u8 pad[6];
332 } spare_ents[0];
333};
334/* spare_assign.type is a bitmap */
335#define DDF_spare_dedicated 0x1 /* else global */
336#define DDF_spare_revertible 0x2 /* else committable */
337#define DDF_spare_active 0x4 /* else not active */
338#define DDF_spare_affinity 0x8 /* enclosure affinity */
339
340/* The data_section contents - local scope */
341struct disk_data {
342 __u32 magic;
343 __u32 crc;
344 char guid[DDF_GUID_LEN];
345 __u32 refnum; /* crc of some magic drive data ... */
346 __u8 forced_ref; /* set when above was not result of magic */
347 __u8 forced_guid; /* set if guid was forced rather than magic */
348 __u8 vendor[32];
349 __u8 pad[442];
350};
351
352/* bbm_section content */
353struct bad_block_log {
354 __u32 magic;
355 __u32 crc;
356 __u16 entry_count;
357 __u32 spare_count;
358 __u8 pad[10];
359 __u64 first_spare;
360 struct mapped_block {
361 __u64 defective_start;
362 __u32 replacement_start;
363 __u16 remap_count;
364 __u8 pad[2];
365 } entries[0];
366};
367
368/* Struct for internally holding ddf structures */
369/* The DDF structure stored on each device is potentially
370 * quite different, as some data is global and some is local.
371 * The global data is:
372 * - ddf header
373 * - controller_data
374 * - Physical disk records
375 * - Virtual disk records
376 * The local data is:
377 * - Configuration records
378 * - Physical Disk data section
379 * ( and Bad block and vendor which I don't care about yet).
380 *
381 * The local data is parsed into separate lists as it is read
382 * and reconstructed for writing. This means that we only need
383 * to make config changes once and they are automatically
384 * propagated to all devices.
385 * Note that the ddf_super has space of the conf and disk data
386 * for this disk and also for a list of all such data.
387 * The list is only used for the superblock that is being
388 * built in Create or Assemble to describe the whole array.
389 */
390struct ddf_super {
391 struct ddf_header anchor, primary, secondary, *active;
392 struct ddf_controller_data controller;
393 struct phys_disk *phys;
394 struct virtual_disk *virt;
395 int pdsize, vdsize;
396 int max_part;
397 struct vcl {
398 struct vcl *next;
399 __u64 *lba_offset; /* location in 'conf' of
400 * the lba table */
401 struct vd_config conf;
402 } *conflist, *newconf;
403 struct dl {
404 struct dl *next;
405 struct disk_data disk;
406 int major, minor;
407 char *devname;
408 int fd;
409 struct vcl *vlist[0]; /* max_part+1 in size */
410 } *dlist;
411};
412
413#ifndef offsetof
414#define offsetof(t,f) ((size_t)&(((t*)0)->f))
415#endif
416
417struct superswitch super_ddf_container, super_ddf_bvd;
418
419static int calc_crc(void *buf, int len)
420{
421 /* crcs are always at the same place as in the ddf_header */
422 struct ddf_header *ddf = buf;
423 __u32 oldcrc = ddf->crc;
424 __u32 newcrc;
425 ddf->crc = 0xffffffff;
426
427 newcrc = crc32(0, buf, len);
428 ddf->crc = oldcrc;
429 return newcrc;
430}
431
432static int load_ddf_header(int fd, unsigned long long lba,
433 unsigned long long size,
434 int type,
435 struct ddf_header *hdr, struct ddf_header *anchor)
436{
437 /* read a ddf header (primary or secondary) from fd/lba
438 * and check that it is consistent with anchor
439 * Need to check:
440 * magic, crc, guid, rev, and LBA's header_type, and
441 * everything after header_type must be the same
442 */
443 if (lba >= size-1)
444 return 0;
445
446 if (lseek64(fd, lba<<9, 0) < 0)
447 return 0;
448
449 if (read(fd, hdr, 512) != 512)
450 return 0;
451
452 if (hdr->magic != DDF_HEADER_MAGIC)
453 return 0;
454 if (calc_crc(hdr, 512) != hdr->crc)
455 return 0;
456 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
457 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
458 anchor->primary_lba != hdr->primary_lba ||
459 anchor->secondary_lba != hdr->secondary_lba ||
460 hdr->type != type ||
461 memcmp(anchor->pad2, hdr->pad2, 512 -
462 offsetof(struct ddf_header, pad2)) != 0)
463 return 0;
464
465 /* Looks good enough to me... */
466 return 1;
467}
468
469static void *load_section(int fd, struct ddf_super *super, void *buf,
470 __u32 offset_be, __u32 len_be, int check)
471{
472 unsigned long long offset = __be32_to_cpu(offset_be);
473 unsigned long long len = __be32_to_cpu(len_be);
474 int dofree = (buf == NULL);
475
476 if (check)
477 if (len != 2 && len != 8 && len != 32
478 && len != 128 && len != 512)
479 return NULL;
480
481 if (len > 1024)
482 return NULL;
483 if (buf) {
484 /* All pre-allocated sections are a single block */
485 if (len != 1)
486 return NULL;
487 } else
488 buf = malloc(len<<9);
489 if (!buf)
490 return NULL;
491
492 if (super->active->type == 1)
493 offset += __be64_to_cpu(super->active->primary_lba);
494 else
495 offset += __be64_to_cpu(super->active->secondary_lba);
496
497 if (lseek64(fd, offset<<9, 0) != (offset<<9)) {
498 if (dofree)
499 free(buf);
500 return NULL;
501 }
502 if (read(fd, buf, len<<9) != (len<<9)) {
503 if (dofree)
504 free(buf);
505 return NULL;
506 }
507 return buf;
508}
509
510static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
511{
512 unsigned long long dsize;
513
514 get_dev_size(fd, NULL, &dsize);
515
516 if (lseek64(fd, dsize-512, 0) < 0) {
517 if (devname)
518 fprintf(stderr,
519 Name": Cannot seek to anchor block on %s: %s\n",
520 devname, strerror(errno));
521 return 1;
522 }
523 if (read(fd, &super->anchor, 512) != 512) {
524 if (devname)
525 fprintf(stderr,
526 Name ": Cannot read anchor block on %s: %s\n",
527 devname, strerror(errno));
528 return 1;
529 }
530 if (super->anchor.magic != DDF_HEADER_MAGIC) {
531 if (devname)
532 fprintf(stderr, Name ": no DDF anchor found on %s\n",
533 devname);
534 return 2;
535 }
536 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
537 if (devname)
538 fprintf(stderr, Name ": bad CRC on anchor on %s\n",
539 devname);
540 return 2;
541 }
542 if (memcmp(super->anchor.revision, DDF_REVISION, 8) != 0) {
543 if (devname)
544 fprintf(stderr, Name ": can only support super revision"
545 " %.8s, not %.8s on %s\n",
546 DDF_REVISION, super->anchor.revision, devname);
547 return 2;
548 }
549 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
550 dsize >> 9, 1,
551 &super->primary, &super->anchor) == 0) {
552 if (devname)
553 fprintf(stderr,
554 Name ": Failed to load primary DDF header "
555 "on %s\n", devname);
556 return 2;
557 }
558 super->active = &super->primary;
559 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
560 dsize >> 9, 2,
561 &super->secondary, &super->anchor)) {
562 if ((__be32_to_cpu(super->primary.seq)
563 < __be32_to_cpu(super->secondary.seq) &&
564 !super->secondary.openflag)
565 || (__be32_to_cpu(super->primary.seq)
566 == __be32_to_cpu(super->secondary.seq) &&
567 super->primary.openflag && !super->secondary.openflag)
568 )
569 super->active = &super->secondary;
570 }
571 return 0;
572}
573
574static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
575{
576 void *ok;
577 ok = load_section(fd, super, &super->controller,
578 super->active->controller_section_offset,
579 super->active->controller_section_length,
580 0);
581 super->phys = load_section(fd, super, NULL,
582 super->active->phys_section_offset,
583 super->active->phys_section_length,
584 1);
585 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
586
587 super->virt = load_section(fd, super, NULL,
588 super->active->virt_section_offset,
589 super->active->virt_section_length,
590 1);
591 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
592 if (!ok ||
593 !super->phys ||
594 !super->virt) {
595 free(super->phys);
596 free(super->virt);
597 return 2;
598 }
599 super->conflist = NULL;
600 super->dlist = NULL;
601 return 0;
602}
603
604static int load_ddf_local(int fd, struct ddf_super *super,
605 char *devname, int keep)
606{
607 struct dl *dl;
608 struct stat stb;
609 char *conf;
610 int i;
611 int conflen;
612
613 /* First the local disk info */
614 super->max_part = __be16_to_cpu(super->active->max_partitions);
615 dl = malloc(sizeof(*dl) +
616 (super->max_part+1) * sizeof(dl->vlist[0]));
617
618 load_section(fd, super, &dl->disk,
619 super->active->data_section_offset,
620 super->active->data_section_length,
621 0);
622 dl->devname = devname ? strdup(devname) : NULL;
623 fstat(fd, &stb);
624 dl->major = major(stb.st_rdev);
625 dl->minor = minor(stb.st_rdev);
626 dl->next = super->dlist;
627 dl->fd = keep ? fd : -1;
628 for (i=0 ; i < super->max_part + 1 ; i++)
629 dl->vlist[i] = NULL;
630 super->dlist = dl;
631
632 /* Now the config list. */
633 /* 'conf' is an array of config entries, some of which are
634 * probably invalid. Those which are good need to be copied into
635 * the conflist
636 */
637 conflen = __be16_to_cpu(super->active->config_record_len);
638
639 conf = load_section(fd, super, NULL,
640 super->active->config_section_offset,
641 super->active->config_section_length,
642 0);
643
644 for (i = 0;
645 i < __be32_to_cpu(super->active->config_section_length);
646 i += conflen) {
647 struct vd_config *vd =
648 (struct vd_config *)((char*)conf + i*512);
649 struct vcl *vcl;
650
651 if (vd->magic != DDF_VD_CONF_MAGIC)
652 continue;
653 for (vcl = super->conflist; vcl; vcl = vcl->next) {
654 if (memcmp(vcl->conf.guid,
655 vd->guid, DDF_GUID_LEN) == 0)
656 break;
657 }
658
659 if (vcl) {
660 dl->vlist[i/conflen] = vcl;
661 if (__be32_to_cpu(vd->seqnum) <=
662 __be32_to_cpu(vcl->conf.seqnum))
663 continue;
664 } else {
665 vcl = malloc(conflen*512 + offsetof(struct vcl, conf));
666 vcl->next = super->conflist;
667 super->conflist = vcl;
668 }
669 memcpy(&vcl->conf, vd, conflen*512);
670 vcl->lba_offset = (__u64*)
671 &vcl->conf.phys_refnum[super->max_part+1];
672 dl->vlist[i/conflen] = vcl;
673 }
674 free(conf);
675
676 return 0;
677}
678
679#ifndef MDASSEMBLE
680static int load_super_ddf_all(struct supertype *st, int fd,
681 void **sbp, char *devname, int keep_fd);
682#endif
683static int load_super_ddf(struct supertype *st, int fd,
684 char *devname)
685{
686 unsigned long long dsize;
687 struct ddf_super *super;
688 int rv;
689
690#ifndef MDASSEMBLE
691 if (load_super_ddf_all(st, fd, &st->sb, devname, 0) == 0)
692 return 0;
693#endif
694
695 if (get_dev_size(fd, devname, &dsize) == 0)
696 return 1;
697
698 /* 32M is a lower bound */
699 if (dsize <= 32*1024*1024) {
700 if (devname) {
701 fprintf(stderr,
702 Name ": %s is too small for ddf: "
703 "size is %llu sectors.\n",
704 devname, dsize>>9);
705 return 1;
706 }
707 }
708 if (dsize & 511) {
709 if (devname) {
710 fprintf(stderr,
711 Name ": %s is an odd size for ddf: "
712 "size is %llu bytes.\n",
713 devname, dsize);
714 return 1;
715 }
716 }
717
718 super = malloc(sizeof(*super));
719 if (!super) {
720 fprintf(stderr, Name ": malloc of %zu failed.\n",
721 sizeof(*super));
722 return 1;
723 }
724
725 rv = load_ddf_headers(fd, super, devname);
726 if (rv) {
727 free(super);
728 return rv;
729 }
730
731 /* Have valid headers and have chosen the best. Let's read in the rest*/
732
733 rv = load_ddf_global(fd, super, devname);
734
735 if (rv) {
736 if (devname)
737 fprintf(stderr,
738 Name ": Failed to load all information "
739 "sections on %s\n", devname);
740 free(super);
741 return rv;
742 }
743
744 load_ddf_local(fd, super, devname, 0);
745
746 /* Should possibly check the sections .... */
747
748 st->sb = super;
749 if (st->ss == NULL) {
750 st->ss = &super_ddf;
751 st->minor_version = 0;
752 st->max_devs = 512;
753 }
754 return 0;
755
756}
757
758static void free_super_ddf(struct supertype *st)
759{
760 struct ddf_super *ddf = st->sb;
761 if (ddf == NULL)
762 return;
763 free(ddf->phys);
764 free(ddf->virt);
765 while (ddf->conflist) {
766 struct vcl *v = ddf->conflist;
767 ddf->conflist = v->next;
768 free(v);
769 }
770 while (ddf->dlist) {
771 struct dl *d = ddf->dlist;
772 ddf->dlist = d->next;
773 if (d->fd >= 0)
774 close(d->fd);
775 free(d);
776 }
777 free(ddf);
778 st->sb = NULL;
779}
780
781static struct supertype *match_metadata_desc_ddf(char *arg)
782{
783 /* 'ddf' only support containers */
784 struct supertype *st;
785 if (strcmp(arg, "ddf") != 0 &&
786 strcmp(arg, "default") != 0
787 )
788 return NULL;
789
790 st = malloc(sizeof(*st));
791 st->ss = &super_ddf;
792 st->max_devs = 512;
793 st->minor_version = 0;
794 st->sb = NULL;
795 return st;
796}
797
798static struct supertype *match_metadata_desc_ddf_bvd(char *arg)
799{
800 struct supertype *st;
801 if (strcmp(arg, "ddf/bvd") != 0 &&
802 strcmp(arg, "bvd") != 0 &&
803 strcmp(arg, "default") != 0
804 )
805 return NULL;
806
807 st = malloc(sizeof(*st));
808 st->ss = &super_ddf_bvd;
809 st->max_devs = 512;
810 st->minor_version = 0;
811 st->sb = NULL;
812 return st;
813}
814static struct supertype *match_metadata_desc_ddf_svd(char *arg)
815{
816 struct supertype *st;
817 if (strcmp(arg, "ddf/svd") != 0 &&
818 strcmp(arg, "svd") != 0 &&
819 strcmp(arg, "default") != 0
820 )
821 return NULL;
822
823 st = malloc(sizeof(*st));
824 st->ss = &super_ddf_svd;
825 st->max_devs = 512;
826 st->minor_version = 0;
827 st->sb = NULL;
828 return st;
829}
830
831#ifndef MDASSEMBLE
832
833static mapping_t ddf_state[] = {
834 { "Optimal", 0},
835 { "Degraded", 1},
836 { "Deleted", 2},
837 { "Missing", 3},
838 { "Failed", 4},
839 { "Partially Optimal", 5},
840 { "-reserved-", 6},
841 { "-reserved-", 7},
842 { NULL, 0}
843};
844
845static mapping_t ddf_init_state[] = {
846 { "Not Initialised", 0},
847 { "QuickInit in Progress", 1},
848 { "Fully Initialised", 2},
849 { "*UNKNOWN*", 3},
850 { NULL, 0}
851};
852static mapping_t ddf_access[] = {
853 { "Read/Write", 0},
854 { "Reserved", 1},
855 { "Read Only", 2},
856 { "Blocked (no access)", 3},
857 { NULL ,0}
858};
859
860static mapping_t ddf_level[] = {
861 { "RAID0", DDF_RAID0},
862 { "RAID1", DDF_RAID1},
863 { "RAID3", DDF_RAID3},
864 { "RAID4", DDF_RAID4},
865 { "RAID5", DDF_RAID5},
866 { "RAID1E",DDF_RAID1E},
867 { "JBOD", DDF_JBOD},
868 { "CONCAT",DDF_CONCAT},
869 { "RAID5E",DDF_RAID5E},
870 { "RAID5EE",DDF_RAID5EE},
871 { "RAID6", DDF_RAID6},
872 { NULL, 0}
873};
874static mapping_t ddf_sec_level[] = {
875 { "Striped", DDF_2STRIPED},
876 { "Mirrored", DDF_2MIRRORED},
877 { "Concat", DDF_2CONCAT},
878 { "Spanned", DDF_2SPANNED},
879 { NULL, 0}
880};
881#endif
882
883struct num_mapping {
884 int num1, num2;
885};
886static struct num_mapping ddf_level_num[] = {
887 { DDF_RAID0, 0 },
888 { DDF_RAID1, 1 },
889 { DDF_RAID3, LEVEL_UNSUPPORTED },
890 { DDF_RAID5, 4 },
891 { DDF_RAID1E, LEVEL_UNSUPPORTED },
892 { DDF_JBOD, LEVEL_UNSUPPORTED },
893 { DDF_CONCAT, LEVEL_LINEAR },
894 { DDF_RAID5E, LEVEL_UNSUPPORTED },
895 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
896 { DDF_RAID6, 6},
897 { MAXINT, MAXINT }
898};
899
900static int map_num1(struct num_mapping *map, int num)
901{
902 int i;
903 for (i=0 ; map[i].num1 != MAXINT; i++)
904 if (map[i].num1 == num)
905 break;
906 return map[i].num2;
907}
908
909#ifndef MDASSEMBLE
910static void print_guid(char *guid, int tstamp)
911{
912 /* A GUIDs are part (or all) ASCII and part binary.
913 * They tend to be space padded.
914 * We ignore trailing spaces and print numbers
915 * <0x20 and >=0x7f as \xXX
916 * Some GUIDs have a time stamp in bytes 16-19.
917 * We print that if appropriate
918 */
919 int l = DDF_GUID_LEN;
920 int i;
921 while (l && guid[l-1] == ' ')
922 l--;
923 for (i=0 ; i<l ; i++) {
924 if (guid[i] >= 0x20 && guid[i] < 0x7f)
925 fputc(guid[i], stdout);
926 else
927 fprintf(stdout, "\\x%02x", guid[i]&255);
928 }
929 if (tstamp) {
930 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
931 char tbuf[100];
932 struct tm *tm;
933 tm = localtime(&then);
934 strftime(tbuf, 100, " (%D %T)",tm);
935 fputs(tbuf, stdout);
936 }
937}
938
939static void examine_vd(int n, struct ddf_super *sb, char *guid)
940{
941 int crl = __be16_to_cpu(sb->anchor.config_record_len);
942 struct vcl *vcl;
943
944 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
945 struct vd_config *vc = &vcl->conf;
946
947 if (calc_crc(vc, crl*512) != vc->crc)
948 continue;
949 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
950 continue;
951
952 /* Ok, we know about this VD, let's give more details */
953 printf(" Raid Devices[%d] : %d\n", n,
954 __be16_to_cpu(vc->prim_elmnt_count));
955 printf(" Chunk Size[%d] : %d sectors\n", n,
956 1 << vc->chunk_shift);
957 printf(" Raid Level[%d] : %s\n", n,
958 map_num(ddf_level, vc->prl)?:"-unknown-");
959 if (vc->sec_elmnt_count != 1) {
960 printf(" Secondary Position[%d] : %d of %d\n", n,
961 vc->sec_elmnt_seq, vc->sec_elmnt_count);
962 printf(" Secondary Level[%d] : %s\n", n,
963 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
964 }
965 printf(" Device Size[%d] : %llu\n", n,
966 __be64_to_cpu(vc->blocks)/2);
967 printf(" Array Size[%d] : %llu\n", n,
968 __be64_to_cpu(vc->array_blocks)/2);
969 }
970}
971
972static void examine_vds(struct ddf_super *sb)
973{
974 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
975 int i;
976 printf(" Virtual Disks : %d\n", cnt);
977
978 for (i=0; i<cnt; i++) {
979 struct virtual_entry *ve = &sb->virt->entries[i];
980 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
981 printf("\n");
982 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
983 printf(" state[%d] : %s, %s%s\n", i,
984 map_num(ddf_state, ve->state & 7),
985 (ve->state & 8) ? "Morphing, ": "",
986 (ve->state & 16)? "Not Consistent" : "Consistent");
987 printf(" init state[%d] : %s\n", i,
988 map_num(ddf_init_state, ve->init_state&3));
989 printf(" access[%d] : %s\n", i,
990 map_num(ddf_access, (ve->init_state>>6) & 3));
991 printf(" Name[%d] : %.16s\n", i, ve->name);
992 examine_vd(i, sb, ve->guid);
993 }
994 if (cnt) printf("\n");
995}
996
997static void examine_pds(struct ddf_super *sb)
998{
999 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1000 int i;
1001 struct dl *dl;
1002 printf(" Physical Disks : %d\n", cnt);
1003
1004 for (i=0 ; i<cnt ; i++) {
1005 struct phys_disk_entry *pd = &sb->phys->entries[i];
1006 int type = __be16_to_cpu(pd->type);
1007 int state = __be16_to_cpu(pd->state);
1008
1009 printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1010 printf("\n");
1011 printf(" ref[%d] : %08x\n", i,
1012 __be32_to_cpu(pd->refnum));
1013 printf(" mode[%d] : %s%s%s%s%s\n", i,
1014 (type&2) ? "active":"",
1015 (type&4) ? "Global Spare":"",
1016 (type&8) ? "spare" : "",
1017 (type&16)? ", foreign" : "",
1018 (type&32)? "pass-through" : "");
1019 printf(" state[%d] : %s%s%s%s%s%s%s\n", i,
1020 (state&1)? "Online": "Offline",
1021 (state&2)? ", Failed": "",
1022 (state&4)? ", Rebuilding": "",
1023 (state&8)? ", in-transition": "",
1024 (state&16)? ", SMART errors": "",
1025 (state&32)? ", Unrecovered Read Errors": "",
1026 (state&64)? ", Missing" : "");
1027 printf(" Avail Size[%d] : %llu K\n", i,
1028 __be64_to_cpu(pd->config_size)>>1);
1029 for (dl = sb->dlist; dl ; dl = dl->next) {
1030 if (dl->disk.refnum == pd->refnum) {
1031 char *dv = map_dev(dl->major, dl->minor, 0);
1032 if (dv)
1033 printf(" Device[%d] : %s\n",
1034 i, dv);
1035 }
1036 }
1037 printf("\n");
1038 }
1039}
1040
1041static void examine_super_ddf(struct supertype *st, char *homehost)
1042{
1043 struct ddf_super *sb = st->sb;
1044
1045 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1046 printf(" Version : %.8s\n", sb->anchor.revision);
1047 printf("Controller GUID : "); print_guid(sb->anchor.guid, 1);
1048 printf("\n");
1049 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1050 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1051 ?"yes" : "no");
1052 examine_vds(sb);
1053 examine_pds(sb);
1054}
1055
1056static void brief_examine_super_ddf(struct supertype *st)
1057{
1058 /* We just write a generic DDF ARRAY entry
1059 * The uuid is all hex, 6 groups of 4 bytes
1060 */
1061 struct ddf_super *ddf = st->sb;
1062 int i;
1063 printf("ARRAY /dev/ddf UUID=");
1064 for (i = 0; i < DDF_GUID_LEN; i++) {
1065 printf("%02x", ddf->anchor.guid[i]);
1066 if ((i&3) == 0 && i != 0)
1067 printf(":");
1068 }
1069 printf("\n");
1070}
1071
1072static void detail_super_ddf(struct supertype *st, char *homehost)
1073{
1074 /* FIXME later
1075 * Could print DDF GUID
1076 * Need to find which array
1077 * If whole, briefly list all arrays
1078 * If one, give name
1079 */
1080}
1081
1082static void brief_detail_super_ddf(struct supertype *st)
1083{
1084 /* FIXME I really need to know which array we are detailing.
1085 * Can that be stored in ddf_super??
1086 */
1087// struct ddf_super *ddf = st->sb;
1088}
1089
1090
1091#endif
1092
1093static int match_home_ddf(struct supertype *st, char *homehost)
1094{
1095 /* It matches 'this' host if the controller is a
1096 * Linux-MD controller with vendor_data matching
1097 * the hostname
1098 */
1099 struct ddf_super *ddf = st->sb;
1100 int len = strlen(homehost);
1101
1102 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1103 len < sizeof(ddf->controller.vendor_data) &&
1104 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1105 ddf->controller.vendor_data[len] == 0);
1106}
1107
1108static struct vd_config *find_vdcr(struct ddf_super *ddf)
1109{
1110 /* FIXME this just picks off the first one */
1111 return &ddf->conflist->conf;
1112}
1113
1114static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1115{
1116 /* The uuid returned here is used for:
1117 * uuid to put into bitmap file (Create, Grow)
1118 * uuid for backup header when saving critical section (Grow)
1119 * comparing uuids when re-adding a device into an array
1120 * For each of these we can make do with a truncated
1121 * or hashed uuid rather than the original, as long as
1122 * everyone agrees.
1123 * In each case the uuid required is that of the data-array,
1124 * not the device-set.
1125 * In the case of SVD we assume the BVD is of interest,
1126 * though that might be the case if a bitmap were made for
1127 * a mirrored SVD - worry about that later.
1128 * So we need to find the VD configuration record for the
1129 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1130 * The first 16 bytes of the sha1 of these is used.
1131 */
1132 struct ddf_super *ddf = st->sb;
1133 struct vd_config *vd = find_vdcr(ddf);
1134
1135 if (!vd)
1136 memset(uuid, 0, sizeof (uuid));
1137 else {
1138 char buf[20];
1139 struct sha1_ctx ctx;
1140 sha1_init_ctx(&ctx);
1141 sha1_process_bytes(&vd->guid, DDF_GUID_LEN, &ctx);
1142 if (vd->sec_elmnt_count > 1)
1143 sha1_process_bytes(&vd->sec_elmnt_seq, 1, &ctx);
1144 sha1_finish_ctx(&ctx, buf);
1145 memcpy(uuid, buf, sizeof(uuid));
1146 }
1147}
1148
1149static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1150{
1151 struct ddf_super *ddf = st->sb;
1152
1153 info->array.major_version = 1000;
1154 info->array.minor_version = 0; /* FIXME use ddf->revision somehow */
1155 info->array.patch_version = 0;
1156 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1157 info->array.level = LEVEL_CONTAINER;
1158 info->array.layout = 0;
1159 info->array.md_minor = -1;
1160 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1161 (ddf->anchor.guid+16));
1162 info->array.utime = 0;
1163 info->array.chunk_size = 0;
1164
1165// info->data_offset = ???;
1166// info->component_size = ???;
1167
1168 info->disk.major = 0;
1169 info->disk.minor = 0;
1170// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1171// info->disk.raid_disk = find refnum in the table and use index;
1172// info->disk.state = ???;
1173
1174// uuid_from_super_ddf(info->uuid, sbv);
1175
1176// info->name[] ?? ;
1177}
1178
1179static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1180{
1181 struct ddf_super *ddf = st->sb;
1182 struct vd_config *vd = find_vdcr(ddf);
1183
1184 /* FIXME this returns BVD info - what if we want SVD ?? */
1185
1186 info->array.major_version = 1000;
1187 info->array.minor_version = 0; /* FIXME use ddf->revision somehow */
1188 info->array.patch_version = 0;
1189 info->array.raid_disks = __be16_to_cpu(vd->prim_elmnt_count);
1190 info->array.level = map_num1(ddf_level_num, vd->prl);
1191 info->array.layout = vd->rlq; /* FIXME should this be mapped */
1192 info->array.md_minor = -1;
1193 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)(vd->guid+16));
1194 info->array.utime = DECADE + __be32_to_cpu(vd->timestamp);
1195 info->array.chunk_size = 512 << vd->chunk_shift;
1196
1197// info->data_offset = ???;
1198// info->component_size = ???;
1199
1200 info->disk.major = 0;
1201 info->disk.minor = 0;
1202// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1203// info->disk.raid_disk = find refnum in the table and use index;
1204// info->disk.state = ???;
1205
1206 uuid_from_super_ddf(st, info->uuid);
1207
1208// info->name[] ?? ;
1209}
1210
1211static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1212 char *update,
1213 char *devname, int verbose,
1214 int uuid_set, char *homehost)
1215{
1216 /* For 'assemble' and 'force' we need to return non-zero if any
1217 * change was made. For others, the return value is ignored.
1218 * Update options are:
1219 * force-one : This device looks a bit old but needs to be included,
1220 * update age info appropriately.
1221 * assemble: clear any 'faulty' flag to allow this device to
1222 * be assembled.
1223 * force-array: Array is degraded but being forced, mark it clean
1224 * if that will be needed to assemble it.
1225 *
1226 * newdev: not used ????
1227 * grow: Array has gained a new device - this is currently for
1228 * linear only
1229 * resync: mark as dirty so a resync will happen.
1230 * uuid: Change the uuid of the array to match watch is given
1231 * homehost: update the recorded homehost
1232 * name: update the name - preserving the homehost
1233 * _reshape_progress: record new reshape_progress position.
1234 *
1235 * Following are not relevant for this version:
1236 * sparc2.2 : update from old dodgey metadata
1237 * super-minor: change the preferred_minor number
1238 * summaries: update redundant counters.
1239 */
1240 int rv = 0;
1241// struct ddf_super *ddf = st->sb;
1242// struct vd_config *vd = find_vdcr(ddf);
1243// struct virtual_entry *ve = find_ve(ddf);
1244
1245
1246 /* we don't need to handle "force-*" or "assemble" as
1247 * there is no need to 'trick' the kernel. We the metadata is
1248 * first updated to activate the array, all the implied modifications
1249 * will just happen.
1250 */
1251
1252 if (strcmp(update, "grow") == 0) {
1253 /* FIXME */
1254 }
1255 if (strcmp(update, "resync") == 0) {
1256// info->resync_checkpoint = 0;
1257 }
1258 /* We ignore UUID updates as they make even less sense
1259 * with DDF
1260 */
1261 if (strcmp(update, "homehost") == 0) {
1262 /* homehost is stored in controller->vendor_data,
1263 * or it is when we are the vendor
1264 */
1265// if (info->vendor_is_local)
1266// strcpy(ddf->controller.vendor_data, homehost);
1267 }
1268 if (strcmp(update, "name") == 0) {
1269 /* name is stored in virtual_entry->name */
1270// memset(ve->name, ' ', 16);
1271// strncpy(ve->name, info->name, 16);
1272 }
1273 if (strcmp(update, "_reshape_progress") == 0) {
1274 /* We don't support reshape yet */
1275 }
1276
1277// update_all_csum(ddf);
1278
1279 return rv;
1280}
1281
5f8097be
NB
1282static void make_header_guid(char *guid)
1283{
1284 __u32 stamp;
1285 int rfd;
1286 /* Create a DDF Header of Virtual Disk GUID */
1287
1288 /* 24 bytes of fiction required.
1289 * first 8 are a 'vendor-id' - "Linux-MD"
1290 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1291 * Remaining 8 random number plus timestamp
1292 */
1293 memcpy(guid, T10, sizeof(T10));
1294 stamp = __cpu_to_be32(0xdeadbeef);
1295 memcpy(guid+8, &stamp, 4);
1296 stamp = __cpu_to_be32(0);
1297 memcpy(guid+12, &stamp, 4);
1298 stamp = __cpu_to_be32(time(0) - DECADE);
1299 memcpy(guid+16, &stamp, 4);
1300 rfd = open("/dev/urandom", O_RDONLY);
1301 if (rfd < 0 || read(rfd, &stamp, 4) != 4)
1302 stamp = random();
1303 memcpy(guid+20, &stamp, 4);
1304 if (rfd >= 0) close(rfd);
1305}
a322f70c
DW
1306static int init_super_ddf(struct supertype *st,
1307 mdu_array_info_t *info,
1308 unsigned long long size, char *name, char *homehost,
1309 int *uuid)
1310{
1311 /* This is primarily called by Create when creating a new array.
1312 * We will then get add_to_super called for each component, and then
1313 * write_init_super called to write it out to each device.
1314 * For DDF, Create can create on fresh devices or on a pre-existing
1315 * array.
1316 * To create on a pre-existing array a different method will be called.
1317 * This one is just for fresh drives.
1318 *
1319 * We need to create the entire 'ddf' structure which includes:
1320 * DDF headers - these are easy.
1321 * Controller data - a Sector describing this controller .. not that
1322 * this is a controller exactly.
1323 * Physical Disk Record - one entry per device, so
1324 * leave plenty of space.
1325 * Virtual Disk Records - again, just leave plenty of space.
1326 * This just lists VDs, doesn't give details
1327 * Config records - describes the VDs that use this disk
1328 * DiskData - describes 'this' device.
1329 * BadBlockManagement - empty
1330 * Diag Space - empty
1331 * Vendor Logs - Could we put bitmaps here?
1332 *
1333 */
1334 struct ddf_super *ddf;
1335 char hostname[17];
1336 int hostlen;
a322f70c
DW
1337 int max_phys_disks, max_virt_disks;
1338 unsigned long long sector;
1339 int clen;
1340 int i;
1341 int pdsize, vdsize;
1342 struct phys_disk *pd;
1343 struct virtual_disk *vd;
1344
1345 ddf = malloc(sizeof(*ddf));
1346 ddf->dlist = NULL; /* no physical disks yet */
1347 ddf->conflist = NULL; /* No virtual disks yet */
1348
1349 /* At least 32MB *must* be reserved for the ddf. So let's just
1350 * start 32MB from the end, and put the primary header there.
1351 * Don't do secondary for now.
1352 * We don't know exactly where that will be yet as it could be
1353 * different on each device. To just set up the lengths.
1354 *
1355 */
1356
1357 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1358 make_header_guid(ddf->anchor.guid);
a322f70c
DW
1359
1360 memcpy(ddf->anchor.revision, DDF_REVISION, 8);
1361 ddf->anchor.seq = __cpu_to_be32(1);
1362 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1363 ddf->anchor.openflag = 0xFF;
1364 ddf->anchor.foreignflag = 0;
1365 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1366 ddf->anchor.pad0 = 0xff;
1367 memset(ddf->anchor.pad1, 0xff, 12);
1368 memset(ddf->anchor.header_ext, 0xff, 32);
1369 ddf->anchor.primary_lba = ~(__u64)0;
1370 ddf->anchor.secondary_lba = ~(__u64)0;
1371 ddf->anchor.type = DDF_HEADER_ANCHOR;
1372 memset(ddf->anchor.pad2, 0xff, 3);
1373 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1374 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1375 of 32M reserved.. */
1376 max_phys_disks = 1023; /* Should be enough */
1377 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1378 max_virt_disks = 255;
1379 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1380 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1381 ddf->max_part = 64;
1382 ddf->anchor.config_record_len = __cpu_to_be16(1 + 256*12/512);
1383 ddf->anchor.max_primary_element_entries = __cpu_to_be16(256);
1384 memset(ddf->anchor.pad3, 0xff, 54);
1385
1386 /* controller sections is one sector long immediately
1387 * after the ddf header */
1388 sector = 1;
1389 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1390 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1391 sector += 1;
1392
1393 /* phys is 8 sectors after that */
1394 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1395 sizeof(struct phys_disk_entry)*max_phys_disks,
1396 512);
1397 switch(pdsize/512) {
1398 case 2: case 8: case 32: case 128: case 512: break;
1399 default: abort();
1400 }
1401 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1402 ddf->anchor.phys_section_length =
1403 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1404 sector += pdsize/512;
1405
1406 /* virt is another 32 sectors */
1407 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1408 sizeof(struct virtual_entry) * max_virt_disks,
1409 512);
1410 switch(vdsize/512) {
1411 case 2: case 8: case 32: case 128: case 512: break;
1412 default: abort();
1413 }
1414 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1415 ddf->anchor.virt_section_length =
1416 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1417 sector += vdsize/512;
1418
1419 clen = (1 + 256*12/512) * (64+1);
1420 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1421 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1422 sector += clen;
1423
1424 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1425 ddf->anchor.data_section_length = __cpu_to_be32(1);
1426 sector += 1;
1427
1428 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1429 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1430 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1431 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1432 ddf->anchor.vendor_length = __cpu_to_be32(0);
1433 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1434
1435 memset(ddf->anchor.pad4, 0xff, 256);
1436
1437 memcpy(&ddf->primary, &ddf->anchor, 512);
1438 memcpy(&ddf->secondary, &ddf->anchor, 512);
1439
1440 ddf->primary.openflag = 1; /* I guess.. */
1441 ddf->primary.type = DDF_HEADER_PRIMARY;
1442
1443 ddf->secondary.openflag = 1; /* I guess.. */
1444 ddf->secondary.type = DDF_HEADER_SECONDARY;
1445
1446 ddf->active = &ddf->primary;
1447
1448 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1449
1450 /* 24 more bytes of fiction required.
1451 * first 8 are a 'vendor-id' - "Linux-MD"
1452 * Remaining 16 are serial number.... maybe a hostname would do?
1453 */
1454 memcpy(ddf->controller.guid, T10, sizeof(T10));
1455 gethostname(hostname, 17);
1456 hostname[17] = 0;
1457 hostlen = strlen(hostname);
1458 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1459 for (i = strlen(T10) ; i+hostlen < 24; i++)
1460 ddf->controller.guid[i] = ' ';
1461
1462 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1463 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1464 ddf->controller.type.sub_vendor_id = 0;
1465 ddf->controller.type.sub_device_id = 0;
1466 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1467 memset(ddf->controller.pad, 0xff, 8);
1468 memset(ddf->controller.vendor_data, 0xff, 448);
1469
1470 pd = ddf->phys = malloc(pdsize);
1471 ddf->pdsize = pdsize;
1472
1473 memset(pd, 0xff, pdsize);
1474 memset(pd, 0, sizeof(*pd));
1475 pd->magic = DDF_PHYS_DATA_MAGIC;
1476 pd->used_pdes = __cpu_to_be16(0);
1477 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1478 memset(pd->pad, 0xff, 52);
1479
1480 vd = ddf->virt = malloc(vdsize);
1481 ddf->vdsize = vdsize;
1482 memset(vd, 0, vdsize);
1483 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1484 vd->populated_vdes = __cpu_to_be16(0);
1485 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1486 memset(vd->pad, 0xff, 52);
1487
5f8097be
NB
1488 for (i=0; i<max_virt_disks; i++)
1489 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1490
a322f70c
DW
1491 st->sb = ddf;
1492 return 1;
1493}
1494
5f8097be
NB
1495static int all_ff(char *guid)
1496{
1497 int i;
1498 for (i = 0; i < DDF_GUID_LEN; i++)
1499 if (guid[i] != (char)0xff)
1500 return 0;
1501 return 1;
1502}
1503static int chunk_to_shift(int chunksize)
1504{
1505 return ffs(chunksize/512)-1;
1506}
1507
1508static int level_to_prl(int level)
1509{
1510 switch (level) {
1511 case LEVEL_LINEAR: return DDF_CONCAT;
1512 case 0: return DDF_RAID0;
1513 case 1: return DDF_RAID1;
1514 case 4: return DDF_RAID4;
1515 case 5: return DDF_RAID5;
1516 case 6: return DDF_RAID6;
1517 default: return -1;
1518 }
1519}
1520static int layout_to_rlq(int level, int layout, int raiddisks)
1521{
1522 switch(level) {
1523 case 0:
1524 return DDF_RAID0_SIMPLE;
1525 case 1:
1526 switch(raiddisks) {
1527 case 2: return DDF_RAID1_SIMPLE;
1528 case 3: return DDF_RAID1_MULTI;
1529 default: return -1;
1530 }
1531 case 4:
1532 switch(layout) {
1533 case 0: return DDF_RAID4_N;
1534 }
1535 break;
1536 case 5:
1537 case 6:
1538 switch(layout) {
1539 case ALGORITHM_LEFT_ASYMMETRIC:
1540 return DDF_RAID5_N_RESTART;
1541 case ALGORITHM_RIGHT_ASYMMETRIC:
1542 return DDF_RAID5_0_RESTART;
1543 case ALGORITHM_LEFT_SYMMETRIC:
1544 return DDF_RAID5_N_CONTINUE;
1545 case ALGORITHM_RIGHT_SYMMETRIC:
1546 return -1; /* not mentioned in standard */
1547 }
1548 }
1549 return -1;
1550}
1551
1552static int init_super_ddf_bvd(struct supertype *st,
1553 mdu_array_info_t *info,
1554 unsigned long long size,
1555 char *name, char *homehost,
1556 int *uuid)
1557{
1558 /* We are creating a BVD inside a pre-existing container.
1559 * so st->sb is already set.
1560 * We need to create a new vd_config and a new virtual_entry
1561 */
1562 struct ddf_super *ddf = st->sb;
1563 int venum;
1564 struct virtual_entry *ve;
1565 struct vcl *vcl;
1566 struct vd_config *vc;
1567 int mppe;
1568 int conflen;
1569
1570 if (__be16_to_cpu(ddf->virt->populated_vdes)
1571 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1572 fprintf(stderr, Name": This ddf already has the "
1573 "maximum of %d virtual devices\n",
1574 __be16_to_cpu(ddf->virt->max_vdes));
1575 return 0;
1576 }
1577
1578 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1579 if (all_ff(ddf->virt->entries[venum].guid))
1580 break;
1581 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1582 fprintf(stderr, Name ": Cannot find spare slot for "
1583 "virtual disk - DDF is corrupt\n");
1584 return 0;
1585 }
1586 ve = &ddf->virt->entries[venum];
1587
1588 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1589 * timestamp, random number
1590 */
1591 make_header_guid(ve->guid);
1592 ve->unit = __cpu_to_be16(info->md_minor);
1593 ve->pad0 = 0xFFFF;
1594 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1595 ve->type = 0;
1596 ve->state = 0;
1597 ve->init_state = 0;
1598 if (!(info->state & 1))
1599 ve->init_state = DDF_state_inconsistent;
1600 memset(ve->pad1, 0xff, 14);
1601 memset(ve->name, ' ', 16);
1602 if (name)
1603 strncpy(ve->name, name, 16);
1604 ddf->virt->populated_vdes =
1605 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1606
1607 /* Now create a new vd_config */
1608 conflen = __be16_to_cpu(ddf->active->config_record_len);
1609 vcl = malloc(offsetof(struct vcl, conf) + conflen * 512);
1610 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->max_part+1];
1611
1612 vc = &vcl->conf;
1613
1614 vc->magic = DDF_VD_CONF_MAGIC;
1615 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
1616 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
1617 vc->seqnum = __cpu_to_be32(1);
1618 memset(vc->pad0, 0xff, 24);
1619 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
1620 vc->chunk_shift = chunk_to_shift(info->chunk_size);
1621 vc->prl = level_to_prl(info->level);
1622 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
1623 vc->sec_elmnt_count = 1;
1624 vc->sec_elmnt_seq = 0;
1625 vc->srl = 0;
1626 vc->blocks = __cpu_to_be64(info->size * 2);
1627 vc->array_blocks = __cpu_to_be64(
1628 calc_array_size(info->level, info->raid_disks, info->layout,
1629 info->chunk_size, info->size*2));
1630 memset(vc->pad1, 0xff, 8);
1631 vc->spare_refs[0] = 0xffffffff;
1632 vc->spare_refs[1] = 0xffffffff;
1633 vc->spare_refs[2] = 0xffffffff;
1634 vc->spare_refs[3] = 0xffffffff;
1635 vc->spare_refs[4] = 0xffffffff;
1636 vc->spare_refs[5] = 0xffffffff;
1637 vc->spare_refs[6] = 0xffffffff;
1638 vc->spare_refs[7] = 0xffffffff;
1639 memset(vc->cache_pol, 0, 8);
1640 vc->bg_rate = 0x80;
1641 memset(vc->pad2, 0xff, 3);
1642 memset(vc->pad3, 0xff, 52);
1643 memset(vc->pad4, 0xff, 192);
1644 memset(vc->v0, 0xff, 32);
1645 memset(vc->v1, 0xff, 32);
1646 memset(vc->v2, 0xff, 16);
1647 memset(vc->v3, 0xff, 16);
1648 memset(vc->vendor, 0xff, 32);
1649 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
1650 memset(vc->phys_refnum, 0xff, 4*mppe);
1651 memset(vc->phys_refnum+mppe, 0x00, 8*mppe);
1652
1653 vcl->next = ddf->conflist;
1654 ddf->conflist = vcl;
1655 ddf->newconf = vcl;
1656 return 1;
1657}
1658
1659static void add_to_super_ddf_bvd(struct supertype *st,
1660 mdu_disk_info_t *dk, int fd, char *devname)
1661{
1662 /* fd and devname identify a device with-in the ddf container (st).
1663 * dk identifies a location in the new BVD.
1664 * We need to find suitable free space in that device and update
1665 * the phys_refnum and lba_offset for the newly created vd_config.
1666 * We might also want to update the type in the phys_disk
1667 * section. FIXME
1668 */
1669 struct dl *dl;
1670 struct ddf_super *ddf = st->sb;
1671 struct vd_config *vc;
1672 __u64 *lba_offset;
1673 int mppe;
1674
1675 for (dl = ddf->dlist; dl ; dl = dl->next)
1676 if (dl->major == dk->major &&
1677 dl->minor == dk->minor)
1678 break;
1679 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
1680 return;
1681
1682 vc = &ddf->newconf->conf;
1683 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
1684 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
1685 lba_offset = (__u64*)(vc->phys_refnum + mppe);
1686 lba_offset[dk->raid_disk] = 0; /* FIXME */
1687
1688 dl->vlist[0] =ddf->newconf; /* FIXME */
1689
1690 dl->fd = fd;
1691 dl->devname = devname;
1692}
1693
a322f70c
DW
1694/* add a device to a container, either while creating it or while
1695 * expanding a pre-existing container
1696 */
1697static void add_to_super_ddf(struct supertype *st,
1698 mdu_disk_info_t *dk, int fd, char *devname)
1699{
1700 struct ddf_super *ddf = st->sb;
1701 struct dl *dd;
1702 time_t now;
1703 struct tm *tm;
1704 unsigned long long size;
1705 struct phys_disk_entry *pde;
1706 int n, i;
1707 struct stat stb;
1708
1709 /* This is device numbered dk->number. We need to create
1710 * a phys_disk entry and a more detailed disk_data entry.
1711 */
1712 fstat(fd, &stb);
1713 dd = malloc(sizeof(*dd) + sizeof(dd->vlist[0]) * (ddf->max_part+1));
1714 dd->major = major(stb.st_rdev);
1715 dd->minor = minor(stb.st_rdev);
1716 dd->devname = devname;
1717 dd->next = ddf->dlist;
1718 dd->fd = fd;
1719
1720 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
1721 now = time(0);
1722 tm = localtime(&now);
1723 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
1724 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
1725 *(__u32*)(dd->disk.guid + 16) = random();
1726 *(__u32*)(dd->disk.guid + 20) = random();
1727
5f8097be 1728 dd->disk.refnum = random(); /* and hope for the best FIXME check this is unique!!*/
a322f70c
DW
1729 dd->disk.forced_ref = 1;
1730 dd->disk.forced_guid = 1;
1731 memset(dd->disk.vendor, ' ', 32);
1732 memcpy(dd->disk.vendor, "Linux", 5);
1733 memset(dd->disk.pad, 0xff, 442);
1734 for (i = 0; i < ddf->max_part+1 ; i++)
1735 dd->vlist[i] = NULL;
1736
1737 n = __be16_to_cpu(ddf->phys->used_pdes);
1738 pde = &ddf->phys->entries[n];
1739 n++;
1740 ddf->phys->used_pdes = __cpu_to_be16(n);
1741
1742 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
1743 pde->refnum = dd->disk.refnum;
1744 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID |DDF_Global_Spare);
1745 pde->state = __cpu_to_be16(DDF_Online);
1746 get_dev_size(fd, NULL, &size);
1747 /* We are required to reserve 32Meg, and record the size in sectors */
1748 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
1749 sprintf(pde->path, "%17.17s","Information: nil") ;
1750 memset(pde->pad, 0xff, 6);
1751
1752 ddf->dlist = dd;
1753}
1754
1755/*
1756 * This is the write_init_super method for a ddf container. It is
1757 * called when creating a container or adding another device to a
1758 * container.
1759 */
1760
1761#ifndef MDASSEMBLE
1762static int write_init_super_ddf(struct supertype *st)
1763{
1764
1765 struct ddf_super *ddf = st->sb;
1766 int i;
1767 struct dl *d;
1768 int n_config;
1769 int conf_size;
1770
1771 unsigned long long size, sector;
1772
1773 for (d = ddf->dlist; d; d=d->next) {
1774 int fd = d->fd;
1775
1776 if (fd < 0)
1777 continue;
1778
1779 /* We need to fill in the primary, (secondary) and workspace
1780 * lba's in the headers, set their checksums,
1781 * Also checksum phys, virt....
1782 *
1783 * Then write everything out, finally the anchor is written.
1784 */
1785 get_dev_size(fd, NULL, &size);
1786 size /= 512;
1787 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
1788 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
1789 ddf->anchor.seq = __cpu_to_be32(1);
1790 memcpy(&ddf->primary, &ddf->anchor, 512);
1791 memcpy(&ddf->secondary, &ddf->anchor, 512);
1792
1793 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
1794 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
1795 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
1796
1797 ddf->primary.openflag = 0;
1798 ddf->primary.type = DDF_HEADER_PRIMARY;
1799
1800 ddf->secondary.openflag = 0;
1801 ddf->secondary.type = DDF_HEADER_SECONDARY;
1802
1803 ddf->primary.crc = calc_crc(&ddf->primary, 512);
1804 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
1805
1806 sector = size - 16*1024*2;
1807 lseek64(fd, sector<<9, 0);
1808 write(fd, &ddf->primary, 512);
1809
1810 ddf->controller.crc = calc_crc(&ddf->controller, 512);
1811 write(fd, &ddf->controller, 512);
1812
1813 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
1814
1815 write(fd, ddf->phys, ddf->pdsize);
1816
1817 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
1818 write(fd, ddf->virt, ddf->vdsize);
1819
1820 /* Now write lots of config records. */
1821 n_config = __be16_to_cpu(ddf->active->max_partitions);
1822 conf_size = __be16_to_cpu(ddf->active->config_record_len) * 512;
1823 for (i = 0 ; i <= n_config ; i++) {
1824 struct vcl *c = d->vlist[i];
1825
1826 if (c) {
1827 c->conf.crc = calc_crc(&c->conf, conf_size);
1828 write(fd, &c->conf, conf_size);
1829 } else {
1830 __u32 sig = 0xffffffff;
1831 write(fd, &sig, 4);
1832 lseek64(fd, conf_size-4, SEEK_CUR);
1833 }
1834 }
1835 d->disk.crc = calc_crc(&d->disk, 512);
1836 write(fd, &d->disk, 512);
1837
1838 /* Maybe do the same for secondary */
1839
1840 lseek64(fd, (size-1)*512, SEEK_SET);
1841 write(fd, &ddf->anchor, 512);
1842 close(fd);
1843 }
1844 return 1;
1845}
1846#endif
1847
1848static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
1849{
1850 /* We must reserve the last 32Meg */
1851 if (devsize <= 32*1024*2)
1852 return 0;
1853 return devsize - 32*1024*2;
1854}
1855
1856#ifndef MDASSEMBLE
1857int validate_geometry_ddf(struct supertype *st,
1858 int level, int layout, int raiddisks,
1859 int chunk, unsigned long long size,
1860 char *dev, unsigned long long *freesize)
1861{
1862 int fd;
1863 struct mdinfo *sra;
1864 int cfd;
1865
1866 /* ddf potentially supports lots of things, but it depends on
1867 * what devices are offered (and maybe kernel version?)
1868 * If given unused devices, we will make a container.
1869 * If given devices in a container, we will make a BVD.
1870 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
1871 */
1872
1873 if (level == LEVEL_CONTAINER) {
1874 st->ss = &super_ddf_container;
5f8097be
NB
1875 if (dev) {
1876 int rv =st->ss->validate_geometry(st, level, layout,
1877 raiddisks, chunk,
1878 size,
1879 NULL, freesize);
1880 if (rv)
1881 return rv;
1882 }
1883 return st->ss->validate_geometry(st, level, layout, raiddisks,
1884 chunk, size, dev, freesize);
1885 }
1886
1887 if (st->sb) {
1888 /* creating in a given container */
1889 st->ss = &super_ddf_bvd;
1890 if (dev) {
1891 int rv =st->ss->validate_geometry(st, level, layout,
1892 raiddisks, chunk,
1893 size,
1894 NULL, freesize);
1895 if (rv)
1896 return rv;
1897 }
a322f70c
DW
1898 return st->ss->validate_geometry(st, level, layout, raiddisks,
1899 chunk, size, dev, freesize);
1900 }
5f8097be
NB
1901 /* FIXME should exclude MULTIPATH, or more appropriately, allow
1902 * only known levels.
1903 */
a322f70c
DW
1904 if (!dev)
1905 return 1;
1906
1907 /* This device needs to be either a device in a 'ddf' container,
5f8097be 1908 * or it needs to be a 'ddf-bvd' array.
a322f70c
DW
1909 */
1910
1911 fd = open(dev, O_RDONLY|O_EXCL, 0);
1912 if (fd >= 0) {
1913 sra = sysfs_read(fd, 0, GET_VERSION);
1914 close(fd);
1915 if (sra && sra->array.major_version == -1 &&
1916 strcmp(sra->text_version, "ddf-bvd") == 0) {
1917 st->ss = &super_ddf_svd;
1918 return st->ss->validate_geometry(st, level, layout,
1919 raiddisks, chunk, size,
1920 dev, freesize);
1921 }
1922
1923 fprintf(stderr,
1924 Name ": Cannot create this array on device %s\n",
1925 dev);
1926 return 0;
1927 }
1928 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
1929 fprintf(stderr, Name ": Cannot open %s: %s\n",
1930 dev, strerror(errno));
1931 return 0;
1932 }
1933 /* Well, it is in use by someone, maybe a 'ddf' container. */
1934 cfd = open_container(fd);
1935 if (cfd < 0) {
1936 close(fd);
1937 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
1938 dev);
1939 return 0;
1940 }
1941 sra = sysfs_read(cfd, 0, GET_VERSION);
1942 close(fd);
1943 if (sra && sra->array.major_version == -1 &&
1944 strcmp(sra->text_version, "ddf") == 0) {
1945 /* This is a member of a ddf container. Load the container
1946 * and try to create a bvd
1947 */
1948 struct ddf_super *ddf;
1949 st->ss = &super_ddf_bvd;
1950 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 1951 st->sb = ddf;
a322f70c
DW
1952 close(cfd);
1953 return st->ss->validate_geometry(st, level, layout,
1954 raiddisks, chunk, size,
1955 dev, freesize);
1956 }
1957 close(cfd);
1958 }
1959 fprintf(stderr, Name ": Cannot use %s: Already in use\n",
1960 dev);
1961 return 1;
1962}
1963
1964int validate_geometry_ddf_container(struct supertype *st,
1965 int level, int layout, int raiddisks,
1966 int chunk, unsigned long long size,
1967 char *dev, unsigned long long *freesize)
1968{
1969 int fd;
1970 unsigned long long ldsize;
1971
1972 if (level != LEVEL_CONTAINER)
1973 return 0;
1974 if (!dev)
1975 return 1;
1976
1977 fd = open(dev, O_RDONLY|O_EXCL, 0);
1978 if (fd < 0) {
1979 fprintf(stderr, Name ": Cannot open %s: %s\n",
1980 dev, strerror(errno));
1981 return 0;
1982 }
1983 if (!get_dev_size(fd, dev, &ldsize)) {
1984 close(fd);
1985 return 0;
1986 }
1987 close(fd);
1988
1989 *freesize = avail_size_ddf(st, ldsize);
1990
1991 return 1;
1992}
1993
5f8097be
NB
1994struct extent {
1995 unsigned long long start, size;
1996};
1997int cmp_extent(const void *av, const void *bv)
1998{
1999 const struct extent *a = av;
2000 const struct extent *b = bv;
2001 if (a->start < b->start)
2002 return -1;
2003 if (a->start > b->start)
2004 return 1;
2005 return 0;
2006}
2007
2008struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
2009{
2010 /* find a list of used extents on the give physical device
2011 * (dnum) or the given ddf.
2012 * Return a malloced array of 'struct extent'
2013
2014FIXME ignore DDF_Legacy devices?
2015
2016 */
2017 struct extent *rv;
2018 int n = 0;
2019 int dnum;
2020 int i, j;
2021
2022 for (dnum = 0; dnum < ddf->phys->used_pdes; dnum++)
2023 if (memcmp(dl->disk.guid,
2024 ddf->phys->entries[dnum].guid,
2025 DDF_GUID_LEN) == 0)
2026 break;
2027
2028 if (dnum == ddf->phys->used_pdes)
2029 return NULL;
2030
2031 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
2032 if (!rv)
2033 return NULL;
2034
2035 for (i = 0; i < ddf->max_part+1; i++) {
2036 struct vcl *v = dl->vlist[i];
2037 if (v == NULL)
2038 continue;
2039 for (j=0; j < v->conf.prim_elmnt_count; j++)
2040 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
2041 /* This device plays role 'j' in 'v'. */
2042 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
2043 rv[n].size = __be64_to_cpu(v->conf.blocks);
2044 n++;
2045 break;
2046 }
2047 }
2048 qsort(rv, n, sizeof(*rv), cmp_extent);
2049
2050 rv[n].start = __be64_to_cpu(ddf->phys->entries[dnum].config_size);
2051 rv[n].size = 0;
2052 return rv;
2053}
2054
a322f70c
DW
2055int validate_geometry_ddf_bvd(struct supertype *st,
2056 int level, int layout, int raiddisks,
2057 int chunk, unsigned long long size,
2058 char *dev, unsigned long long *freesize)
2059{
2060 struct stat stb;
2061 struct ddf_super *ddf = st->sb;
2062 struct dl *dl;
5f8097be
NB
2063 unsigned long long pos = 0;
2064 unsigned long long maxsize;
2065 struct extent *e;
2066 int i;
a322f70c
DW
2067 /* ddf/bvd supports lots of things, but not containers */
2068 if (level == LEVEL_CONTAINER)
2069 return 0;
2070 /* We must have the container info already read in. */
2071 if (!ddf)
2072 return 0;
2073
5f8097be
NB
2074 if (!dev) {
2075 /* General test: make sure there is space for
2076 * 'raiddisks' device extents of size 'size'.
2077 */
2078 unsigned long long minsize = size;
2079 int dcnt = 0;
2080 if (minsize == 0)
2081 minsize = 8;
2082 for (dl = ddf->dlist; dl ; dl = dl->next)
2083 {
2084 int found = 0;
2085
2086 i = 0;
2087 e = get_extents(ddf, dl);
2088 if (!e) continue;
2089 do {
2090 unsigned long long esize;
2091 esize = e[i].start - pos;
2092 if (esize >= minsize)
2093 found = 1;
2094 pos = e[i].start + e[i].size;
2095 i++;
2096 } while (e[i-1].size);
2097 if (found)
2098 dcnt++;
2099 free(e);
2100 }
2101 if (dcnt < raiddisks) {
2102 fprintf(stderr, Name ": Not enough devices with space "
2103 "for this array (%d < %d)\n",
2104 dcnt, raiddisks);
2105 return 0;
2106 }
2107 return 1;
2108 }
a322f70c
DW
2109 /* This device must be a member of the set */
2110 if (stat(dev, &stb) < 0)
2111 return 0;
2112 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2113 return 0;
2114 for (dl = ddf->dlist ; dl ; dl = dl->next) {
2115 if (dl->major == major(stb.st_rdev) &&
2116 dl->minor == minor(stb.st_rdev))
2117 break;
2118 }
5f8097be
NB
2119 if (!dl) {
2120 fprintf(stderr, Name ": %s is not in the same DDF set\n",
2121 dev);
2122 return 0;
2123 }
2124 e = get_extents(ddf, dl);
2125 maxsize = 0;
2126 i = 0;
2127 if (e) do {
2128 unsigned long long esize;
2129 esize = e[i].start - pos;
2130 if (esize >= maxsize)
2131 maxsize = esize;
2132 pos = e[i].start + e[i].size;
2133 i++;
2134 } while (e[i-1].size);
2135 *freesize = maxsize;
a322f70c
DW
2136 // FIXME here I am
2137
2138 return 1;
2139}
2140int validate_geometry_ddf_svd(struct supertype *st,
2141 int level, int layout, int raiddisks,
2142 int chunk, unsigned long long size,
2143 char *dev, unsigned long long *freesize)
2144{
2145 /* dd/svd only supports striped, mirrored, concat, spanned... */
2146 if (level != LEVEL_LINEAR &&
2147 level != 0 &&
2148 level != 1)
2149 return 0;
2150 return 1;
2151}
2152
2153
2154static int load_super_ddf_all(struct supertype *st, int fd,
2155 void **sbp, char *devname, int keep_fd)
2156{
2157 struct mdinfo *sra;
2158 struct ddf_super *super;
2159 struct mdinfo *sd, *best = NULL;
2160 int bestseq = 0;
2161 int seq;
2162 char nm[20];
2163 int dfd;
2164
2165 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
2166 if (!sra)
2167 return 1;
2168 if (sra->array.major_version != -1 ||
2169 sra->array.minor_version != -2 ||
2170 strcmp(sra->text_version, "ddf") != 0)
2171 return 1;
2172
2173 super = malloc(sizeof(*super));
2174 if (!super)
2175 return 1;
2176
2177 /* first, try each device, and choose the best ddf */
2178 for (sd = sra->devs ; sd ; sd = sd->next) {
2179 int rv;
2180 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2181 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
2182 if (!dfd)
2183 return 2;
2184 rv = load_ddf_headers(dfd, super, NULL);
2185 if (!keep_fd) close(dfd);
2186 if (rv == 0) {
2187 seq = __be32_to_cpu(super->active->seq);
2188 if (super->active->openflag)
2189 seq--;
2190 if (!best || seq > bestseq) {
2191 bestseq = seq;
2192 best = sd;
2193 }
2194 }
2195 }
2196 if (!best)
2197 return 1;
2198 /* OK, load this ddf */
2199 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2200 dfd = dev_open(nm, O_RDONLY);
2201 if (!dfd)
2202 return 1;
2203 load_ddf_headers(dfd, super, NULL);
2204 load_ddf_global(dfd, super, NULL);
2205 close(dfd);
2206 /* Now we need the device-local bits */
2207 for (sd = sra->devs ; sd ; sd = sd->next) {
2208 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2209 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
a322f70c
DW
2210 if (!dfd)
2211 return 2;
2212 seq = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2213 if (!keep_fd) close(dfd);
a322f70c
DW
2214 }
2215 *sbp = super;
2216 if (st->ss == NULL) {
2217 st->ss = &super_ddf;
2218 st->minor_version = 0;
2219 st->max_devs = 512;
2220 }
2221 return 0;
2222}
2223#endif
2224
2225
2226
2227static int init_zero_ddf(struct supertype *st,
2228 mdu_array_info_t *info,
2229 unsigned long long size, char *name,
2230 char *homehost, int *uuid)
2231{
2232 st->sb = NULL;
2233 return 0;
2234}
2235
2236static int store_zero_ddf(struct supertype *st, int fd)
2237{
2238 unsigned long long dsize;
2239 char buf[512];
2240 memset(buf, 0, 512);
2241
2242
2243 if (!get_dev_size(fd, NULL, &dsize))
2244 return 1;
2245
2246 lseek64(fd, dsize-512, 0);
2247 write(fd, buf, 512);
2248 return 0;
2249}
2250
2251struct superswitch super_ddf = {
2252#ifndef MDASSEMBLE
2253 .examine_super = examine_super_ddf,
2254 .brief_examine_super = brief_examine_super_ddf,
2255 .detail_super = detail_super_ddf,
2256 .brief_detail_super = brief_detail_super_ddf,
2257 .validate_geometry = validate_geometry_ddf,
2258#endif
2259 .match_home = match_home_ddf,
2260 .uuid_from_super= uuid_from_super_ddf,
2261 .getinfo_super = getinfo_super_ddf,
2262 .update_super = update_super_ddf,
2263
2264 .avail_size = avail_size_ddf,
2265
2266 .load_super = load_super_ddf,
2267 .init_super = init_zero_ddf,
2268 .store_super = store_zero_ddf,
2269 .free_super = free_super_ddf,
2270 .match_metadata_desc = match_metadata_desc_ddf,
2271
2272
2273 .major = 1000,
2274 .swapuuid = 0,
2275 .external = 1,
2276 .text_version = "ddf",
2277};
2278
2279/* Super_ddf_container is set by validate_geometry_ddf when given a
2280 * device that is not part of any array
2281 */
2282struct superswitch super_ddf_container = {
2283#ifndef MDASSEMBLE
2284 .validate_geometry = validate_geometry_ddf_container,
2285 .write_init_super = write_init_super_ddf,
2286#endif
2287
2288 .init_super = init_super_ddf,
2289 .add_to_super = add_to_super_ddf,
2290
2291 .free_super = free_super_ddf,
2292
2293 .major = 1000,
2294 .swapuuid = 0,
2295 .external = 1,
2296 .text_version = "ddf",
2297};
2298
2299struct superswitch super_ddf_bvd = {
2300#ifndef MDASSEMBLE
2301// .detail_super = detail_super_ddf_bvd,
2302// .brief_detail_super = brief_detail_super_ddf_bvd,
2303 .validate_geometry = validate_geometry_ddf_bvd,
5f8097be 2304 .write_init_super = write_init_super_ddf,
a322f70c
DW
2305#endif
2306 .update_super = update_super_ddf,
5f8097be
NB
2307 .init_super = init_super_ddf_bvd,
2308 .add_to_super = add_to_super_ddf_bvd,
a322f70c
DW
2309 .getinfo_super = getinfo_super_ddf_bvd,
2310
2311 .load_super = load_super_ddf,
2312 .free_super = free_super_ddf,
2313 .match_metadata_desc = match_metadata_desc_ddf_bvd,
2314
2315
2316 .major = 1001,
2317 .swapuuid = 0,
5f8097be 2318 .external = 2,
a322f70c
DW
2319 .text_version = "ddf",
2320};
2321
2322struct superswitch super_ddf_svd = {
2323#ifndef MDASSEMBLE
2324// .detail_super = detail_super_ddf_svd,
2325// .brief_detail_super = brief_detail_super_ddf_svd,
5f8097be 2326 .validate_geometry = validate_geometry_ddf_svd,
a322f70c
DW
2327#endif
2328 .update_super = update_super_ddf,
2329 .init_super = init_super_ddf,
2330
2331 .load_super = load_super_ddf,
2332 .free_super = free_super_ddf,
2333 .match_metadata_desc = match_metadata_desc_ddf_svd,
2334
a322f70c
DW
2335 .major = 1002,
2336 .swapuuid = 0,
5f8097be 2337 .external = 2,
a322f70c
DW
2338 .text_version = "ddf",
2339};