2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2012 Alexander Motin <mav@FreeBSD.org>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
34 #include <sys/gsb_crc32.h>
35 #include <sys/endian.h>
36 #include <sys/kernel.h>
38 #include <sys/limits.h>
40 #include <sys/malloc.h>
41 #include <sys/mutex.h>
42 #include <sys/systm.h>
44 #include <sys/clock.h>
46 #include <geom/geom.h>
47 #include <geom/geom_dbg.h>
48 #include "geom/raid/g_raid.h"
49 #include "geom/raid/md_ddf.h"
50 #include "g_raid_md_if.h"
52 static MALLOC_DEFINE(M_MD_DDF, "md_ddf_data", "GEOM_RAID DDF metadata");
54 #define DDF_MAX_DISKS_HARD 128
56 #define DDF_MAX_DISKS 16
57 #define DDF_MAX_VDISKS 7
58 #define DDF_MAX_PARTITIONS 1
60 #define DECADE (3600*24*(365*10+2)) /* 10 years in seconds. */
65 struct ddf_header *hdr;
66 struct ddf_cd_record *cdr;
67 struct ddf_pd_record *pdr;
68 struct ddf_vd_record *vdr;
70 struct ddf_pdd_record *pdd;
71 struct ddf_bbm_log *bbm;
77 struct ddf_header *hdr;
78 struct ddf_cd_record *cdr;
79 struct ddf_vd_entry *vde;
80 struct ddf_vdc_record *vdc;
81 struct ddf_vdc_record *bvdc[DDF_MAX_DISKS_HARD];
84 struct g_raid_md_ddf_perdisk {
85 struct ddf_meta pd_meta;
88 struct g_raid_md_ddf_pervolume {
89 struct ddf_vol_meta pv_meta;
91 struct callout pv_start_co; /* STARTING state timer. */
94 struct g_raid_md_ddf_object {
95 struct g_raid_md_object mdio_base;
97 struct ddf_meta mdio_meta;
99 struct callout mdio_start_co; /* STARTING state timer. */
101 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
104 static g_raid_md_create_req_t g_raid_md_create_req_ddf;
105 static g_raid_md_taste_t g_raid_md_taste_ddf;
106 static g_raid_md_event_t g_raid_md_event_ddf;
107 static g_raid_md_volume_event_t g_raid_md_volume_event_ddf;
108 static g_raid_md_ctl_t g_raid_md_ctl_ddf;
109 static g_raid_md_write_t g_raid_md_write_ddf;
110 static g_raid_md_fail_disk_t g_raid_md_fail_disk_ddf;
111 static g_raid_md_free_disk_t g_raid_md_free_disk_ddf;
112 static g_raid_md_free_volume_t g_raid_md_free_volume_ddf;
113 static g_raid_md_free_t g_raid_md_free_ddf;
115 static kobj_method_t g_raid_md_ddf_methods[] = {
116 KOBJMETHOD(g_raid_md_create_req, g_raid_md_create_req_ddf),
117 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_ddf),
118 KOBJMETHOD(g_raid_md_event, g_raid_md_event_ddf),
119 KOBJMETHOD(g_raid_md_volume_event, g_raid_md_volume_event_ddf),
120 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_ddf),
121 KOBJMETHOD(g_raid_md_write, g_raid_md_write_ddf),
122 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_ddf),
123 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_ddf),
124 KOBJMETHOD(g_raid_md_free_volume, g_raid_md_free_volume_ddf),
125 KOBJMETHOD(g_raid_md_free, g_raid_md_free_ddf),
129 static struct g_raid_md_class g_raid_md_ddf_class = {
131 g_raid_md_ddf_methods,
132 sizeof(struct g_raid_md_ddf_object),
137 #define GET8(m, f) ((m)->f)
138 #define GET16(m, f) ((m)->bigendian ? be16dec(&(m)->f) : le16dec(&(m)->f))
139 #define GET32(m, f) ((m)->bigendian ? be32dec(&(m)->f) : le32dec(&(m)->f))
140 #define GET64(m, f) ((m)->bigendian ? be64dec(&(m)->f) : le64dec(&(m)->f))
141 #define GET8D(m, f) (f)
142 #define GET16D(m, f) ((m)->bigendian ? be16dec(&f) : le16dec(&f))
143 #define GET32D(m, f) ((m)->bigendian ? be32dec(&f) : le32dec(&f))
144 #define GET64D(m, f) ((m)->bigendian ? be64dec(&f) : le64dec(&f))
145 #define GET8P(m, f) (*(f))
146 #define GET16P(m, f) ((m)->bigendian ? be16dec(f) : le16dec(f))
147 #define GET32P(m, f) ((m)->bigendian ? be32dec(f) : le32dec(f))
148 #define GET64P(m, f) ((m)->bigendian ? be64dec(f) : le64dec(f))
150 #define SET8P(m, f, v) \
152 #define SET16P(m, f, v) \
154 if ((m)->bigendian) \
159 #define SET32P(m, f, v) \
161 if ((m)->bigendian) \
166 #define SET64P(m, f, v) \
168 if ((m)->bigendian) \
173 #define SET8(m, f, v) SET8P((m), &((m)->f), (v))
174 #define SET16(m, f, v) SET16P((m), &((m)->f), (v))
175 #define SET32(m, f, v) SET32P((m), &((m)->f), (v))
176 #define SET64(m, f, v) SET64P((m), &((m)->f), (v))
177 #define SET8D(m, f, v) SET8P((m), &(f), (v))
178 #define SET16D(m, f, v) SET16P((m), &(f), (v))
179 #define SET32D(m, f, v) SET32P((m), &(f), (v))
180 #define SET64D(m, f, v) SET64P((m), &(f), (v))
182 #define GETCRNUM(m) (GET32((m), hdr->cr_length) / \
183 GET16((m), hdr->Configuration_Record_Length))
185 #define GETVDCPTR(m, n) ((struct ddf_vdc_record *)((uint8_t *)(m)->cr + \
186 (n) * GET16((m), hdr->Configuration_Record_Length) * \
189 #define GETSAPTR(m, n) ((struct ddf_sa_record *)((uint8_t *)(m)->cr + \
190 (n) * GET16((m), hdr->Configuration_Record_Length) * \
194 isff(uint8_t *buf, int size)
198 for (i = 0; i < size; i++)
205 print_guid(uint8_t *buf)
210 for (i = 0; i < 24; i++) {
211 if (buf[i] != 0 && (buf[i] < ' ' || buf[i] > 127)) {
217 printf("'%.24s'", buf);
219 for (i = 0; i < 24; i++)
220 printf("%02x", buf[i]);
225 g_raid_md_ddf_print(struct ddf_meta *meta)
227 struct ddf_vdc_record *vdc;
228 struct ddf_vuc_record *vuc;
229 struct ddf_sa_record *sa;
232 int i, j, k, num, num2;
234 if (g_raid_debug < 1)
237 printf("********* DDF Metadata *********\n");
238 printf("**** Header ****\n");
239 printf("DDF_Header_GUID ");
240 print_guid(meta->hdr->DDF_Header_GUID);
242 printf("DDF_rev %8.8s\n", (char *)&meta->hdr->DDF_rev[0]);
243 printf("Sequence_Number 0x%08x\n", GET32(meta, hdr->Sequence_Number));
244 printf("TimeStamp 0x%08x\n", GET32(meta, hdr->TimeStamp));
245 printf("Open_Flag 0x%02x\n", GET16(meta, hdr->Open_Flag));
246 printf("Foreign_Flag 0x%02x\n", GET16(meta, hdr->Foreign_Flag));
247 printf("Diskgrouping 0x%02x\n", GET16(meta, hdr->Diskgrouping));
248 printf("Primary_Header_LBA %ju\n", GET64(meta, hdr->Primary_Header_LBA));
249 printf("Secondary_Header_LBA %ju\n", GET64(meta, hdr->Secondary_Header_LBA));
250 printf("WorkSpace_Length %u\n", GET32(meta, hdr->WorkSpace_Length));
251 printf("WorkSpace_LBA %ju\n", GET64(meta, hdr->WorkSpace_LBA));
252 printf("Max_PD_Entries %u\n", GET16(meta, hdr->Max_PD_Entries));
253 printf("Max_VD_Entries %u\n", GET16(meta, hdr->Max_VD_Entries));
254 printf("Max_Partitions %u\n", GET16(meta, hdr->Max_Partitions));
255 printf("Configuration_Record_Length %u\n", GET16(meta, hdr->Configuration_Record_Length));
256 printf("Max_Primary_Element_Entries %u\n", GET16(meta, hdr->Max_Primary_Element_Entries));
257 printf("Controller Data %u:%u\n", GET32(meta, hdr->cd_section), GET32(meta, hdr->cd_length));
258 printf("Physical Disk %u:%u\n", GET32(meta, hdr->pdr_section), GET32(meta, hdr->pdr_length));
259 printf("Virtual Disk %u:%u\n", GET32(meta, hdr->vdr_section), GET32(meta, hdr->vdr_length));
260 printf("Configuration Recs %u:%u\n", GET32(meta, hdr->cr_section), GET32(meta, hdr->cr_length));
261 printf("Physical Disk Recs %u:%u\n", GET32(meta, hdr->pdd_section), GET32(meta, hdr->pdd_length));
262 printf("BBM Log %u:%u\n", GET32(meta, hdr->bbmlog_section), GET32(meta, hdr->bbmlog_length));
263 printf("Diagnostic Space %u:%u\n", GET32(meta, hdr->Diagnostic_Space), GET32(meta, hdr->Diagnostic_Space_Length));
264 printf("Vendor_Specific_Logs %u:%u\n", GET32(meta, hdr->Vendor_Specific_Logs), GET32(meta, hdr->Vendor_Specific_Logs_Length));
265 printf("**** Controller Data ****\n");
266 printf("Controller_GUID ");
267 print_guid(meta->cdr->Controller_GUID);
269 printf("Controller_Type 0x%04x%04x 0x%04x%04x\n",
270 GET16(meta, cdr->Controller_Type.Vendor_ID),
271 GET16(meta, cdr->Controller_Type.Device_ID),
272 GET16(meta, cdr->Controller_Type.SubVendor_ID),
273 GET16(meta, cdr->Controller_Type.SubDevice_ID));
274 printf("Product_ID '%.16s'\n", (char *)&meta->cdr->Product_ID[0]);
275 printf("**** Physical Disk Records ****\n");
276 printf("Populated_PDEs %u\n", GET16(meta, pdr->Populated_PDEs));
277 printf("Max_PDE_Supported %u\n", GET16(meta, pdr->Max_PDE_Supported));
278 for (j = 0; j < GET16(meta, pdr->Populated_PDEs); j++) {
279 if (isff(meta->pdr->entry[j].PD_GUID, 24))
281 if (GET32(meta, pdr->entry[j].PD_Reference) == 0xffffffff)
284 print_guid(meta->pdr->entry[j].PD_GUID);
286 printf("PD_Reference 0x%08x\n",
287 GET32(meta, pdr->entry[j].PD_Reference));
288 printf("PD_Type 0x%04x\n",
289 GET16(meta, pdr->entry[j].PD_Type));
290 printf("PD_State 0x%04x\n",
291 GET16(meta, pdr->entry[j].PD_State));
292 printf("Configured_Size %ju\n",
293 GET64(meta, pdr->entry[j].Configured_Size));
294 printf("Block_Size %u\n",
295 GET16(meta, pdr->entry[j].Block_Size));
297 printf("**** Virtual Disk Records ****\n");
298 printf("Populated_VDEs %u\n", GET16(meta, vdr->Populated_VDEs));
299 printf("Max_VDE_Supported %u\n", GET16(meta, vdr->Max_VDE_Supported));
300 for (j = 0; j < GET16(meta, vdr->Populated_VDEs); j++) {
301 if (isff(meta->vdr->entry[j].VD_GUID, 24))
304 print_guid(meta->vdr->entry[j].VD_GUID);
306 printf("VD_Number 0x%04x\n",
307 GET16(meta, vdr->entry[j].VD_Number));
308 printf("VD_Type 0x%04x\n",
309 GET16(meta, vdr->entry[j].VD_Type));
310 printf("VD_State 0x%02x\n",
311 GET8(meta, vdr->entry[j].VD_State));
312 printf("Init_State 0x%02x\n",
313 GET8(meta, vdr->entry[j].Init_State));
314 printf("Drive_Failures_Remaining %u\n",
315 GET8(meta, vdr->entry[j].Drive_Failures_Remaining));
316 printf("VD_Name '%.16s'\n",
317 (char *)&meta->vdr->entry[j].VD_Name);
319 printf("**** Configuration Records ****\n");
320 num = GETCRNUM(meta);
321 for (j = 0; j < num; j++) {
322 vdc = GETVDCPTR(meta, j);
323 val = GET32D(meta, vdc->Signature);
325 case DDF_VDCR_SIGNATURE:
326 printf("** Virtual Disk Configuration **\n");
328 print_guid(vdc->VD_GUID);
330 printf("Timestamp 0x%08x\n",
331 GET32D(meta, vdc->Timestamp));
332 printf("Sequence_Number 0x%08x\n",
333 GET32D(meta, vdc->Sequence_Number));
334 printf("Primary_Element_Count %u\n",
335 GET16D(meta, vdc->Primary_Element_Count));
336 printf("Stripe_Size %u\n",
337 GET8D(meta, vdc->Stripe_Size));
338 printf("Primary_RAID_Level 0x%02x\n",
339 GET8D(meta, vdc->Primary_RAID_Level));
340 printf("RLQ 0x%02x\n",
341 GET8D(meta, vdc->RLQ));
342 printf("Secondary_Element_Count %u\n",
343 GET8D(meta, vdc->Secondary_Element_Count));
344 printf("Secondary_Element_Seq %u\n",
345 GET8D(meta, vdc->Secondary_Element_Seq));
346 printf("Secondary_RAID_Level 0x%02x\n",
347 GET8D(meta, vdc->Secondary_RAID_Level));
348 printf("Block_Count %ju\n",
349 GET64D(meta, vdc->Block_Count));
350 printf("VD_Size %ju\n",
351 GET64D(meta, vdc->VD_Size));
352 printf("Block_Size %u\n",
353 GET16D(meta, vdc->Block_Size));
354 printf("Rotate_Parity_count %u\n",
355 GET8D(meta, vdc->Rotate_Parity_count));
356 printf("Associated_Spare_Disks");
357 for (i = 0; i < 8; i++) {
358 if (GET32D(meta, vdc->Associated_Spares[i]) != 0xffffffff)
359 printf(" 0x%08x", GET32D(meta, vdc->Associated_Spares[i]));
362 printf("Cache_Flags %016jx\n",
363 GET64D(meta, vdc->Cache_Flags));
364 printf("BG_Rate %u\n",
365 GET8D(meta, vdc->BG_Rate));
366 printf("MDF_Parity_Disks %u\n",
367 GET8D(meta, vdc->MDF_Parity_Disks));
368 printf("MDF_Parity_Generator_Polynomial 0x%04x\n",
369 GET16D(meta, vdc->MDF_Parity_Generator_Polynomial));
370 printf("MDF_Constant_Generation_Method 0x%02x\n",
371 GET8D(meta, vdc->MDF_Constant_Generation_Method));
372 printf("Physical_Disks ");
373 num2 = GET16D(meta, vdc->Primary_Element_Count);
374 val2 = (uint64_t *)&(vdc->Physical_Disk_Sequence[GET16(meta, hdr->Max_Primary_Element_Entries)]);
375 for (i = 0; i < num2; i++)
376 printf(" 0x%08x @ %ju",
377 GET32D(meta, vdc->Physical_Disk_Sequence[i]),
378 GET64P(meta, val2 + i));
381 case DDF_VUCR_SIGNATURE:
382 printf("** Vendor Unique Configuration **\n");
383 vuc = (struct ddf_vuc_record *)vdc;
385 print_guid(vuc->VD_GUID);
388 case DDF_SA_SIGNATURE:
389 printf("** Spare Assignment Configuration **\n");
390 sa = (struct ddf_sa_record *)vdc;
391 printf("Timestamp 0x%08x\n",
392 GET32D(meta, sa->Timestamp));
393 printf("Spare_Type 0x%02x\n",
394 GET8D(meta, sa->Spare_Type));
395 printf("Populated_SAEs %u\n",
396 GET16D(meta, sa->Populated_SAEs));
397 printf("MAX_SAE_Supported %u\n",
398 GET16D(meta, sa->MAX_SAE_Supported));
399 for (i = 0; i < GET16D(meta, sa->Populated_SAEs); i++) {
400 if (isff(sa->entry[i].VD_GUID, 24))
403 for (k = 0; k < 24; k++)
404 printf("%02x", sa->entry[i].VD_GUID[k]);
406 printf("Secondary_Element %u\n",
407 GET16D(meta, sa->entry[i].Secondary_Element));
414 printf("Unknown configuration signature %08x\n", val);
418 printf("**** Physical Disk Data ****\n");
420 print_guid(meta->pdd->PD_GUID);
422 printf("PD_Reference 0x%08x\n",
423 GET32(meta, pdd->PD_Reference));
424 printf("Forced_Ref_Flag 0x%02x\n",
425 GET8(meta, pdd->Forced_Ref_Flag));
426 printf("Forced_PD_GUID_Flag 0x%02x\n",
427 GET8(meta, pdd->Forced_PD_GUID_Flag));
431 ddf_meta_find_pd(struct ddf_meta *meta, uint8_t *GUID, uint32_t PD_Reference)
435 for (i = 0; i < GET16(meta, pdr->Populated_PDEs); i++) {
437 if (memcmp(meta->pdr->entry[i].PD_GUID, GUID, 24) == 0)
439 } else if (PD_Reference != 0xffffffff) {
440 if (GET32(meta, pdr->entry[i].PD_Reference) == PD_Reference)
443 if (isff(meta->pdr->entry[i].PD_GUID, 24))
446 if (GUID == NULL && PD_Reference == 0xffffffff) {
447 if (i >= GET16(meta, pdr->Max_PDE_Supported))
449 SET16(meta, pdr->Populated_PDEs, i + 1);
456 ddf_meta_find_vd(struct ddf_meta *meta, uint8_t *GUID)
460 for (i = 0; i < GET16(meta, vdr->Populated_VDEs); i++) {
462 if (memcmp(meta->vdr->entry[i].VD_GUID, GUID, 24) == 0)
465 if (isff(meta->vdr->entry[i].VD_GUID, 24))
469 if (i >= GET16(meta, vdr->Max_VDE_Supported))
471 SET16(meta, vdr->Populated_VDEs, i + 1);
477 static struct ddf_vdc_record *
478 ddf_meta_find_vdc(struct ddf_meta *meta, uint8_t *GUID)
480 struct ddf_vdc_record *vdc;
483 num = GETCRNUM(meta);
484 for (i = 0; i < num; i++) {
485 vdc = GETVDCPTR(meta, i);
487 if (GET32D(meta, vdc->Signature) == DDF_VDCR_SIGNATURE &&
488 memcmp(vdc->VD_GUID, GUID, 24) == 0)
491 if (GET32D(meta, vdc->Signature) == 0xffffffff ||
492 GET32D(meta, vdc->Signature) == 0)
499 ddf_meta_count_vdc(struct ddf_meta *meta, uint8_t *GUID)
501 struct ddf_vdc_record *vdc;
505 num = GETCRNUM(meta);
506 for (i = 0; i < num; i++) {
507 vdc = GETVDCPTR(meta, i);
508 if (GET32D(meta, vdc->Signature) != DDF_VDCR_SIGNATURE)
510 if (GUID == NULL || memcmp(vdc->VD_GUID, GUID, 24) == 0)
517 ddf_meta_find_disk(struct ddf_vol_meta *vmeta, uint32_t PD_Reference,
518 int *bvdp, int *posp)
523 for (bvd = 0; bvd < GET8(vmeta, vdc->Secondary_Element_Count); bvd++) {
524 if (vmeta->bvdc[bvd] == NULL) {
525 i += GET16(vmeta, vdc->Primary_Element_Count); // XXX
528 for (pos = 0; pos < GET16(vmeta, bvdc[bvd]->Primary_Element_Count);
530 if (GET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos]) ==
543 static struct ddf_sa_record *
544 ddf_meta_find_sa(struct ddf_meta *meta, int create)
546 struct ddf_sa_record *sa;
549 num = GETCRNUM(meta);
550 for (i = 0; i < num; i++) {
551 sa = GETSAPTR(meta, i);
552 if (GET32D(meta, sa->Signature) == DDF_SA_SIGNATURE)
556 for (i = 0; i < num; i++) {
557 sa = GETSAPTR(meta, i);
558 if (GET32D(meta, sa->Signature) == 0xffffffff ||
559 GET32D(meta, sa->Signature) == 0)
567 ddf_meta_create(struct g_raid_disk *disk, struct ddf_meta *sample)
571 struct g_raid_md_ddf_perdisk *pd;
572 struct g_raid_md_ddf_object *mdi;
573 struct ddf_meta *meta;
574 struct ddf_pd_entry *pde;
578 char serial_buffer[DISK_IDENT_SIZE];
580 if (sample->hdr == NULL)
583 mdi = (struct g_raid_md_ddf_object *)disk->d_softc->sc_md;
584 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
586 ss = disk->d_consumer->provider->sectorsize;
587 anchorlba = disk->d_consumer->provider->mediasize / ss - 1;
589 meta->sectorsize = ss;
590 meta->bigendian = sample ? sample->bigendian : mdi->mdio_bigendian;
592 clock_ts_to_ct(&ts, &ct);
595 meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
596 memset(meta->hdr, 0xff, ss);
598 memcpy(meta->hdr, sample->hdr, sizeof(struct ddf_header));
599 if (ss != sample->sectorsize) {
600 SET32(meta, hdr->WorkSpace_Length,
601 howmany(GET32(sample, hdr->WorkSpace_Length) *
602 sample->sectorsize, ss));
603 SET16(meta, hdr->Configuration_Record_Length,
604 howmany(GET16(sample,
605 hdr->Configuration_Record_Length) *
606 sample->sectorsize, ss));
607 SET32(meta, hdr->cd_length,
608 howmany(GET32(sample, hdr->cd_length) *
609 sample->sectorsize, ss));
610 SET32(meta, hdr->pdr_length,
611 howmany(GET32(sample, hdr->pdr_length) *
612 sample->sectorsize, ss));
613 SET32(meta, hdr->vdr_length,
614 howmany(GET32(sample, hdr->vdr_length) *
615 sample->sectorsize, ss));
616 SET32(meta, hdr->cr_length,
617 howmany(GET32(sample, hdr->cr_length) *
618 sample->sectorsize, ss));
619 SET32(meta, hdr->pdd_length,
620 howmany(GET32(sample, hdr->pdd_length) *
621 sample->sectorsize, ss));
622 SET32(meta, hdr->bbmlog_length,
623 howmany(GET32(sample, hdr->bbmlog_length) *
624 sample->sectorsize, ss));
625 SET32(meta, hdr->Diagnostic_Space,
626 howmany(GET32(sample, hdr->bbmlog_length) *
627 sample->sectorsize, ss));
628 SET32(meta, hdr->Vendor_Specific_Logs,
629 howmany(GET32(sample, hdr->bbmlog_length) *
630 sample->sectorsize, ss));
633 SET32(meta, hdr->Signature, DDF_HEADER_SIGNATURE);
634 snprintf(meta->hdr->DDF_Header_GUID, 25, "FreeBSD %08x%08x",
635 (u_int)(ts.tv_sec - DECADE), arc4random());
636 memcpy(meta->hdr->DDF_rev, "02.00.00", 8);
637 SET32(meta, hdr->TimeStamp, (ts.tv_sec - DECADE));
638 SET32(meta, hdr->WorkSpace_Length, 16 * 1024 * 1024 / ss);
639 SET16(meta, hdr->Max_PD_Entries, DDF_MAX_DISKS - 1);
640 SET16(meta, hdr->Max_VD_Entries, DDF_MAX_VDISKS);
641 SET16(meta, hdr->Max_Partitions, DDF_MAX_PARTITIONS);
642 SET16(meta, hdr->Max_Primary_Element_Entries, DDF_MAX_DISKS);
643 SET16(meta, hdr->Configuration_Record_Length,
644 howmany(sizeof(struct ddf_vdc_record) + (4 + 8) *
645 GET16(meta, hdr->Max_Primary_Element_Entries), ss));
646 SET32(meta, hdr->cd_length,
647 howmany(sizeof(struct ddf_cd_record), ss));
648 SET32(meta, hdr->pdr_length,
649 howmany(sizeof(struct ddf_pd_record) +
650 sizeof(struct ddf_pd_entry) * GET16(meta,
651 hdr->Max_PD_Entries), ss));
652 SET32(meta, hdr->vdr_length,
653 howmany(sizeof(struct ddf_vd_record) +
654 sizeof(struct ddf_vd_entry) *
655 GET16(meta, hdr->Max_VD_Entries), ss));
656 SET32(meta, hdr->cr_length,
657 GET16(meta, hdr->Configuration_Record_Length) *
658 (GET16(meta, hdr->Max_Partitions) + 1));
659 SET32(meta, hdr->pdd_length,
660 howmany(sizeof(struct ddf_pdd_record), ss));
661 SET32(meta, hdr->bbmlog_length, 0);
662 SET32(meta, hdr->Diagnostic_Space_Length, 0);
663 SET32(meta, hdr->Vendor_Specific_Logs_Length, 0);
666 SET32(meta, hdr->cd_section, pos);
667 pos += GET32(meta, hdr->cd_length);
668 SET32(meta, hdr->pdr_section, pos);
669 pos += GET32(meta, hdr->pdr_length);
670 SET32(meta, hdr->vdr_section, pos);
671 pos += GET32(meta, hdr->vdr_length);
672 SET32(meta, hdr->cr_section, pos);
673 pos += GET32(meta, hdr->cr_length);
674 SET32(meta, hdr->pdd_section, pos);
675 pos += GET32(meta, hdr->pdd_length);
676 SET32(meta, hdr->bbmlog_section,
677 GET32(meta, hdr->bbmlog_length) != 0 ? pos : 0xffffffff);
678 pos += GET32(meta, hdr->bbmlog_length);
679 SET32(meta, hdr->Diagnostic_Space,
680 GET32(meta, hdr->Diagnostic_Space_Length) != 0 ? pos : 0xffffffff);
681 pos += GET32(meta, hdr->Diagnostic_Space_Length);
682 SET32(meta, hdr->Vendor_Specific_Logs,
683 GET32(meta, hdr->Vendor_Specific_Logs_Length) != 0 ? pos : 0xffffffff);
684 pos += min(GET32(meta, hdr->Vendor_Specific_Logs_Length), 1);
685 SET64(meta, hdr->Primary_Header_LBA,
687 SET64(meta, hdr->Secondary_Header_LBA,
688 0xffffffffffffffffULL);
689 SET64(meta, hdr->WorkSpace_LBA,
690 anchorlba + 1 - 32 * 1024 * 1024 / ss);
692 /* Controller Data */
693 size = GET32(meta, hdr->cd_length) * ss;
694 meta->cdr = malloc(size, M_MD_DDF, M_WAITOK);
695 memset(meta->cdr, 0xff, size);
696 SET32(meta, cdr->Signature, DDF_CONTROLLER_DATA_SIGNATURE);
697 memcpy(meta->cdr->Controller_GUID, "FreeBSD GEOM RAID SERIAL", 24);
698 memcpy(meta->cdr->Product_ID, "FreeBSD GEOMRAID", 16);
700 /* Physical Drive Records. */
701 size = GET32(meta, hdr->pdr_length) * ss;
702 meta->pdr = malloc(size, M_MD_DDF, M_WAITOK);
703 memset(meta->pdr, 0xff, size);
704 SET32(meta, pdr->Signature, DDF_PDR_SIGNATURE);
705 SET16(meta, pdr->Populated_PDEs, 1);
706 SET16(meta, pdr->Max_PDE_Supported,
707 GET16(meta, hdr->Max_PD_Entries));
709 pde = &meta->pdr->entry[0];
710 len = sizeof(serial_buffer);
711 error = g_io_getattr("GEOM::ident", disk->d_consumer, &len, serial_buffer);
712 if (error == 0 && (len = strlen (serial_buffer)) >= 6 && len <= 20)
713 snprintf(pde->PD_GUID, 25, "DISK%20s", serial_buffer);
715 snprintf(pde->PD_GUID, 25, "DISK%04d%02d%02d%08x%04x",
716 ct.year, ct.mon, ct.day,
717 arc4random(), arc4random() & 0xffff);
718 SET32D(meta, pde->PD_Reference, arc4random());
719 SET16D(meta, pde->PD_Type, DDF_PDE_GUID_FORCE);
720 SET16D(meta, pde->PD_State, 0);
721 SET64D(meta, pde->Configured_Size,
722 anchorlba + 1 - 32 * 1024 * 1024 / ss);
723 SET16D(meta, pde->Block_Size, ss);
725 /* Virtual Drive Records. */
726 size = GET32(meta, hdr->vdr_length) * ss;
727 meta->vdr = malloc(size, M_MD_DDF, M_WAITOK);
728 memset(meta->vdr, 0xff, size);
729 SET32(meta, vdr->Signature, DDF_VD_RECORD_SIGNATURE);
730 SET32(meta, vdr->Populated_VDEs, 0);
731 SET16(meta, vdr->Max_VDE_Supported,
732 GET16(meta, hdr->Max_VD_Entries));
734 /* Configuration Records. */
735 size = GET32(meta, hdr->cr_length) * ss;
736 meta->cr = malloc(size, M_MD_DDF, M_WAITOK);
737 memset(meta->cr, 0xff, size);
739 /* Physical Disk Data. */
740 size = GET32(meta, hdr->pdd_length) * ss;
741 meta->pdd = malloc(size, M_MD_DDF, M_WAITOK);
742 memset(meta->pdd, 0xff, size);
743 SET32(meta, pdd->Signature, DDF_PDD_SIGNATURE);
744 memcpy(meta->pdd->PD_GUID, pde->PD_GUID, 24);
745 SET32(meta, pdd->PD_Reference, GET32D(meta, pde->PD_Reference));
746 SET8(meta, pdd->Forced_Ref_Flag, DDF_PDD_FORCED_REF);
747 SET8(meta, pdd->Forced_PD_GUID_Flag, DDF_PDD_FORCED_GUID);
749 /* Bad Block Management Log. */
750 if (GET32(meta, hdr->bbmlog_length) != 0) {
751 size = GET32(meta, hdr->bbmlog_length) * ss;
752 meta->bbm = malloc(size, M_MD_DDF, M_WAITOK);
753 memset(meta->bbm, 0xff, size);
754 SET32(meta, bbm->Signature, DDF_BBML_SIGNATURE);
755 SET32(meta, bbm->Entry_Count, 0);
756 SET32(meta, bbm->Spare_Block_Count, 0);
761 ddf_meta_copy(struct ddf_meta *dst, struct ddf_meta *src)
765 dst->bigendian = src->bigendian;
766 ss = dst->sectorsize = src->sectorsize;
767 dst->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
768 memcpy(dst->hdr, src->hdr, ss);
769 dst->cdr = malloc(GET32(src, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
770 memcpy(dst->cdr, src->cdr, GET32(src, hdr->cd_length) * ss);
771 dst->pdr = malloc(GET32(src, hdr->pdr_length) * ss, M_MD_DDF, M_WAITOK);
772 memcpy(dst->pdr, src->pdr, GET32(src, hdr->pdr_length) * ss);
773 dst->vdr = malloc(GET32(src, hdr->vdr_length) * ss, M_MD_DDF, M_WAITOK);
774 memcpy(dst->vdr, src->vdr, GET32(src, hdr->vdr_length) * ss);
775 dst->cr = malloc(GET32(src, hdr->cr_length) * ss, M_MD_DDF, M_WAITOK);
776 memcpy(dst->cr, src->cr, GET32(src, hdr->cr_length) * ss);
777 dst->pdd = malloc(GET32(src, hdr->pdd_length) * ss, M_MD_DDF, M_WAITOK);
778 memcpy(dst->pdd, src->pdd, GET32(src, hdr->pdd_length) * ss);
779 if (src->bbm != NULL) {
780 dst->bbm = malloc(GET32(src, hdr->bbmlog_length) * ss, M_MD_DDF, M_WAITOK);
781 memcpy(dst->bbm, src->bbm, GET32(src, hdr->bbmlog_length) * ss);
786 ddf_meta_update(struct ddf_meta *meta, struct ddf_meta *src)
788 struct ddf_pd_entry *pde, *spde;
791 for (i = 0; i < GET16(src, pdr->Populated_PDEs); i++) {
792 spde = &src->pdr->entry[i];
793 if (isff(spde->PD_GUID, 24))
795 j = ddf_meta_find_pd(meta, NULL,
796 GET32(src, pdr->entry[i].PD_Reference));
798 j = ddf_meta_find_pd(meta, NULL, 0xffffffff);
799 pde = &meta->pdr->entry[j];
800 memcpy(pde, spde, sizeof(*pde));
802 pde = &meta->pdr->entry[j];
803 SET16D(meta, pde->PD_State,
804 GET16D(meta, pde->PD_State) |
805 GET16D(src, pde->PD_State));
811 ddf_meta_free(struct ddf_meta *meta)
814 if (meta->hdr != NULL) {
815 free(meta->hdr, M_MD_DDF);
818 if (meta->cdr != NULL) {
819 free(meta->cdr, M_MD_DDF);
822 if (meta->pdr != NULL) {
823 free(meta->pdr, M_MD_DDF);
826 if (meta->vdr != NULL) {
827 free(meta->vdr, M_MD_DDF);
830 if (meta->cr != NULL) {
831 free(meta->cr, M_MD_DDF);
834 if (meta->pdd != NULL) {
835 free(meta->pdd, M_MD_DDF);
838 if (meta->bbm != NULL) {
839 free(meta->bbm, M_MD_DDF);
845 ddf_vol_meta_create(struct ddf_vol_meta *meta, struct ddf_meta *sample)
851 meta->bigendian = sample->bigendian;
852 ss = meta->sectorsize = sample->sectorsize;
853 meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
854 memcpy(meta->hdr, sample->hdr, ss);
855 meta->cdr = malloc(GET32(sample, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
856 memcpy(meta->cdr, sample->cdr, GET32(sample, hdr->cd_length) * ss);
857 meta->vde = malloc(sizeof(struct ddf_vd_entry), M_MD_DDF, M_WAITOK);
858 memset(meta->vde, 0xff, sizeof(struct ddf_vd_entry));
860 clock_ts_to_ct(&ts, &ct);
861 snprintf(meta->vde->VD_GUID, 25, "FreeBSD%04d%02d%02d%08x%01x",
862 ct.year, ct.mon, ct.day,
863 arc4random(), arc4random() & 0xf);
864 size = GET16(sample, hdr->Configuration_Record_Length) * ss;
865 meta->vdc = malloc(size, M_MD_DDF, M_WAITOK);
866 memset(meta->vdc, 0xff, size);
867 SET32(meta, vdc->Signature, DDF_VDCR_SIGNATURE);
868 memcpy(meta->vdc->VD_GUID, meta->vde->VD_GUID, 24);
869 SET32(meta, vdc->Sequence_Number, 0);
873 ddf_vol_meta_update(struct ddf_vol_meta *dst, struct ddf_meta *src,
874 uint8_t *GUID, int started)
876 struct ddf_vd_entry *vde;
877 struct ddf_vdc_record *vdc;
878 int vnew, bvnew, bvd, size;
881 vde = &src->vdr->entry[ddf_meta_find_vd(src, GUID)];
882 vdc = ddf_meta_find_vdc(src, GUID);
883 if (GET8D(src, vdc->Secondary_Element_Count) == 1)
886 bvd = GET8D(src, vdc->Secondary_Element_Seq);
887 size = GET16(src, hdr->Configuration_Record_Length) * src->sectorsize;
889 if (dst->vdc == NULL ||
890 (!started && ((int32_t)(GET32D(src, vdc->Sequence_Number) -
891 GET32(dst, vdc->Sequence_Number))) > 0))
896 if (dst->bvdc[bvd] == NULL ||
897 (!started && ((int32_t)(GET32D(src, vdc->Sequence_Number) -
898 GET32(dst, bvdc[bvd]->Sequence_Number))) > 0))
904 dst->bigendian = src->bigendian;
905 ss = dst->sectorsize = src->sectorsize;
906 if (dst->hdr != NULL)
907 free(dst->hdr, M_MD_DDF);
908 dst->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
909 memcpy(dst->hdr, src->hdr, ss);
910 if (dst->cdr != NULL)
911 free(dst->cdr, M_MD_DDF);
912 dst->cdr = malloc(GET32(src, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
913 memcpy(dst->cdr, src->cdr, GET32(src, hdr->cd_length) * ss);
914 if (dst->vde != NULL)
915 free(dst->vde, M_MD_DDF);
916 dst->vde = malloc(sizeof(struct ddf_vd_entry), M_MD_DDF, M_WAITOK);
917 memcpy(dst->vde, vde, sizeof(struct ddf_vd_entry));
918 if (dst->vdc != NULL)
919 free(dst->vdc, M_MD_DDF);
920 dst->vdc = malloc(size, M_MD_DDF, M_WAITOK);
921 memcpy(dst->vdc, vdc, size);
924 if (dst->bvdc[bvd] != NULL)
925 free(dst->bvdc[bvd], M_MD_DDF);
926 dst->bvdc[bvd] = malloc(size, M_MD_DDF, M_WAITOK);
927 memcpy(dst->bvdc[bvd], vdc, size);
932 ddf_vol_meta_free(struct ddf_vol_meta *meta)
936 if (meta->hdr != NULL) {
937 free(meta->hdr, M_MD_DDF);
940 if (meta->cdr != NULL) {
941 free(meta->cdr, M_MD_DDF);
944 if (meta->vde != NULL) {
945 free(meta->vde, M_MD_DDF);
948 if (meta->vdc != NULL) {
949 free(meta->vdc, M_MD_DDF);
952 for (i = 0; i < DDF_MAX_DISKS_HARD; i++) {
953 if (meta->bvdc[i] != NULL) {
954 free(meta->bvdc[i], M_MD_DDF);
955 meta->bvdc[i] = NULL;
961 ddf_meta_unused_range(struct ddf_meta *meta, off_t *off, off_t *size)
963 struct ddf_vdc_record *vdc;
964 off_t beg[32], end[32], beg1, end1;
966 int i, j, n, num, pos;
971 ref = GET32(meta, pdd->PD_Reference);
972 pos = ddf_meta_find_pd(meta, NULL, ref);
974 end[0] = GET64(meta, pdr->entry[pos].Configured_Size);
976 num = GETCRNUM(meta);
977 for (i = 0; i < num; i++) {
978 vdc = GETVDCPTR(meta, i);
979 if (GET32D(meta, vdc->Signature) != DDF_VDCR_SIGNATURE)
981 for (pos = 0; pos < GET16D(meta, vdc->Primary_Element_Count); pos++)
982 if (GET32D(meta, vdc->Physical_Disk_Sequence[pos]) == ref)
984 if (pos == GET16D(meta, vdc->Primary_Element_Count))
986 offp = (uint64_t *)&(vdc->Physical_Disk_Sequence[
987 GET16(meta, hdr->Max_Primary_Element_Entries)]);
988 beg1 = GET64P(meta, offp + pos);
989 end1 = beg1 + GET64D(meta, vdc->Block_Count);
990 for (j = 0; j < n; j++) {
991 if (beg[j] >= end1 || end[j] <= beg1 )
993 if (beg[j] < beg1 && end[j] > end1) {
998 } else if (beg[j] < beg1)
1004 for (j = 0; j < n; j++) {
1005 if (end[j] - beg[j] > *size) {
1007 *size = end[j] - beg[j];
1010 return ((*size > 0) ? 1 : 0);
1014 ddf_meta_get_name(struct ddf_meta *meta, int num, char *buf)
1019 b = meta->vdr->entry[num].VD_Name;
1020 for (i = 15; i >= 0; i--)
1023 memcpy(buf, b, i + 1);
1028 ddf_meta_put_name(struct ddf_vol_meta *meta, char *buf)
1032 len = min(strlen(buf), 16);
1033 memset(meta->vde->VD_Name, 0x20, 16);
1034 memcpy(meta->vde->VD_Name, buf, len);
1038 ddf_meta_read(struct g_consumer *cp, struct ddf_meta *meta)
1040 struct g_provider *pp;
1041 struct ddf_header *ahdr, *hdr;
1043 off_t plba, slba, lba;
1048 ddf_meta_free(meta);
1051 ss = meta->sectorsize = pp->sectorsize;
1052 if (ss < sizeof(*hdr))
1054 /* Read anchor block. */
1055 abuf = g_read_data(cp, pp->mediasize - ss, ss, &error);
1057 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
1061 ahdr = (struct ddf_header *)abuf;
1063 /* Check if this is an DDF RAID struct */
1064 if (be32dec(&ahdr->Signature) == DDF_HEADER_SIGNATURE)
1065 meta->bigendian = 1;
1066 else if (le32dec(&ahdr->Signature) == DDF_HEADER_SIGNATURE)
1067 meta->bigendian = 0;
1069 G_RAID_DEBUG(1, "DDF signature check failed on %s", pp->name);
1073 if (ahdr->Header_Type != DDF_HEADER_ANCHOR) {
1074 G_RAID_DEBUG(1, "DDF header type check failed on %s", pp->name);
1079 plba = GET64(meta, hdr->Primary_Header_LBA);
1080 slba = GET64(meta, hdr->Secondary_Header_LBA);
1081 val = GET32(meta, hdr->CRC);
1082 SET32(meta, hdr->CRC, 0xffffffff);
1084 if (crc32(ahdr, ss) != val) {
1085 G_RAID_DEBUG(1, "DDF CRC mismatch on %s", pp->name);
1089 if ((plba + 6) * ss >= pp->mediasize) {
1090 G_RAID_DEBUG(1, "DDF primary header LBA is wrong on %s", pp->name);
1094 if (slba != -1 && (slba + 6) * ss >= pp->mediasize) {
1095 G_RAID_DEBUG(1, "DDF secondary header LBA is wrong on %s", pp->name);
1103 ddf_meta_free(meta);
1105 /* Read header block. */
1106 buf = g_read_data(cp, lba * ss, ss, &error);
1109 G_RAID_DEBUG(1, "DDF %s metadata read error on %s (error=%d).",
1110 (lba == plba) ? "primary" : "secondary", pp->name, error);
1111 if (lba == plba && slba != -1) {
1115 G_RAID_DEBUG(1, "DDF metadata read error on %s.", pp->name);
1118 meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
1119 memcpy(meta->hdr, buf, ss);
1122 val = GET32(meta, hdr->CRC);
1123 SET32(meta, hdr->CRC, 0xffffffff);
1124 if (hdr->Signature != ahdr->Signature ||
1125 crc32(meta->hdr, ss) != val ||
1126 memcmp(hdr->DDF_Header_GUID, ahdr->DDF_Header_GUID, 24) ||
1127 GET64(meta, hdr->Primary_Header_LBA) != plba ||
1128 GET64(meta, hdr->Secondary_Header_LBA) != slba) {
1130 G_RAID_DEBUG(1, "DDF %s metadata check failed on %s",
1131 (lba == plba) ? "primary" : "secondary", pp->name);
1132 if (lba == plba && slba != -1) {
1136 G_RAID_DEBUG(1, "DDF metadata check failed on %s", pp->name);
1140 if ((lba == plba && hdr->Header_Type != DDF_HEADER_PRIMARY) ||
1141 (lba == slba && hdr->Header_Type != DDF_HEADER_SECONDARY))
1144 len = max(len, GET32(meta, hdr->cd_section) + GET32(meta, hdr->cd_length));
1145 len = max(len, GET32(meta, hdr->pdr_section) + GET32(meta, hdr->pdr_length));
1146 len = max(len, GET32(meta, hdr->vdr_section) + GET32(meta, hdr->vdr_length));
1147 len = max(len, GET32(meta, hdr->cr_section) + GET32(meta, hdr->cr_length));
1148 len = max(len, GET32(meta, hdr->pdd_section) + GET32(meta, hdr->pdd_length));
1149 if ((val = GET32(meta, hdr->bbmlog_section)) != 0xffffffff)
1150 len = max(len, val + GET32(meta, hdr->bbmlog_length));
1151 if ((val = GET32(meta, hdr->Diagnostic_Space)) != 0xffffffff)
1152 len = max(len, val + GET32(meta, hdr->Diagnostic_Space_Length));
1153 if ((val = GET32(meta, hdr->Vendor_Specific_Logs)) != 0xffffffff)
1154 len = max(len, val + GET32(meta, hdr->Vendor_Specific_Logs_Length));
1155 if ((plba + len) * ss >= pp->mediasize)
1157 if (slba != -1 && (slba + len) * ss >= pp->mediasize)
1159 /* Workaround for Adaptec implementation. */
1160 if (GET16(meta, hdr->Max_Primary_Element_Entries) == 0xffff) {
1161 SET16(meta, hdr->Max_Primary_Element_Entries,
1162 min(GET16(meta, hdr->Max_PD_Entries),
1163 (GET16(meta, hdr->Configuration_Record_Length) * ss - 512) / 12));
1166 if (GET32(meta, hdr->cd_length) * ss >= maxphys ||
1167 GET32(meta, hdr->pdr_length) * ss >= maxphys ||
1168 GET32(meta, hdr->vdr_length) * ss >= maxphys ||
1169 GET32(meta, hdr->cr_length) * ss >= maxphys ||
1170 GET32(meta, hdr->pdd_length) * ss >= maxphys ||
1171 GET32(meta, hdr->bbmlog_length) * ss >= maxphys) {
1172 G_RAID_DEBUG(1, "%s: Blocksize is too big.", pp->name);
1176 /* Read controller data. */
1177 buf = g_read_data(cp, (lba + GET32(meta, hdr->cd_section)) * ss,
1178 GET32(meta, hdr->cd_length) * ss, &error);
1181 meta->cdr = malloc(GET32(meta, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
1182 memcpy(meta->cdr, buf, GET32(meta, hdr->cd_length) * ss);
1184 if (GET32(meta, cdr->Signature) != DDF_CONTROLLER_DATA_SIGNATURE)
1187 /* Read physical disk records. */
1188 buf = g_read_data(cp, (lba + GET32(meta, hdr->pdr_section)) * ss,
1189 GET32(meta, hdr->pdr_length) * ss, &error);
1192 meta->pdr = malloc(GET32(meta, hdr->pdr_length) * ss, M_MD_DDF, M_WAITOK);
1193 memcpy(meta->pdr, buf, GET32(meta, hdr->pdr_length) * ss);
1195 if (GET32(meta, pdr->Signature) != DDF_PDR_SIGNATURE)
1198 * Workaround for reading metadata corrupted due to graid bug.
1199 * XXX: Remove this before we have disks above 128PB. :)
1201 if (meta->bigendian) {
1202 for (i = 0; i < GET16(meta, pdr->Populated_PDEs); i++) {
1203 if (isff(meta->pdr->entry[i].PD_GUID, 24))
1205 if (GET32(meta, pdr->entry[i].PD_Reference) ==
1208 if (GET64(meta, pdr->entry[i].Configured_Size) >=
1210 SET16(meta, pdr->entry[i].PD_State,
1211 GET16(meta, pdr->entry[i].PD_State) &
1213 SET64(meta, pdr->entry[i].Configured_Size,
1214 GET64(meta, pdr->entry[i].Configured_Size) &
1215 ((1ULL << 48) - 1));
1220 /* Read virtual disk records. */
1221 buf = g_read_data(cp, (lba + GET32(meta, hdr->vdr_section)) * ss,
1222 GET32(meta, hdr->vdr_length) * ss, &error);
1225 meta->vdr = malloc(GET32(meta, hdr->vdr_length) * ss, M_MD_DDF, M_WAITOK);
1226 memcpy(meta->vdr, buf, GET32(meta, hdr->vdr_length) * ss);
1228 if (GET32(meta, vdr->Signature) != DDF_VD_RECORD_SIGNATURE)
1231 /* Read configuration records. */
1232 buf = g_read_data(cp, (lba + GET32(meta, hdr->cr_section)) * ss,
1233 GET32(meta, hdr->cr_length) * ss, &error);
1236 meta->cr = malloc(GET32(meta, hdr->cr_length) * ss, M_MD_DDF, M_WAITOK);
1237 memcpy(meta->cr, buf, GET32(meta, hdr->cr_length) * ss);
1240 /* Read physical disk data. */
1241 buf = g_read_data(cp, (lba + GET32(meta, hdr->pdd_section)) * ss,
1242 GET32(meta, hdr->pdd_length) * ss, &error);
1245 meta->pdd = malloc(GET32(meta, hdr->pdd_length) * ss, M_MD_DDF, M_WAITOK);
1246 memcpy(meta->pdd, buf, GET32(meta, hdr->pdd_length) * ss);
1248 if (GET32(meta, pdd->Signature) != DDF_PDD_SIGNATURE)
1250 i = ddf_meta_find_pd(meta, NULL, GET32(meta, pdd->PD_Reference));
1255 if (GET32(meta, hdr->bbmlog_section) != 0xffffffff &&
1256 GET32(meta, hdr->bbmlog_length) != 0) {
1257 buf = g_read_data(cp, (lba + GET32(meta, hdr->bbmlog_section)) * ss,
1258 GET32(meta, hdr->bbmlog_length) * ss, &error);
1261 meta->bbm = malloc(GET32(meta, hdr->bbmlog_length) * ss, M_MD_DDF, M_WAITOK);
1262 memcpy(meta->bbm, buf, GET32(meta, hdr->bbmlog_length) * ss);
1264 if (GET32(meta, bbm->Signature) != DDF_BBML_SIGNATURE)
1271 ddf_meta_free(meta);
1276 ddf_meta_write(struct g_consumer *cp, struct ddf_meta *meta)
1278 struct g_provider *pp;
1279 struct ddf_vdc_record *vdc;
1280 off_t alba, plba, slba, lba;
1285 ss = pp->sectorsize;
1286 lba = alba = pp->mediasize / ss - 1;
1287 plba = GET64(meta, hdr->Primary_Header_LBA);
1288 slba = GET64(meta, hdr->Secondary_Header_LBA);
1291 SET8(meta, hdr->Header_Type, (lba == alba) ? DDF_HEADER_ANCHOR :
1292 (lba == plba) ? DDF_HEADER_PRIMARY : DDF_HEADER_SECONDARY);
1293 SET32(meta, hdr->CRC, 0xffffffff);
1294 SET32(meta, hdr->CRC, crc32(meta->hdr, ss));
1295 error = g_write_data(cp, lba * ss, meta->hdr, ss);
1298 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
1308 size = GET32(meta, hdr->cd_length) * ss;
1309 SET32(meta, cdr->CRC, 0xffffffff);
1310 SET32(meta, cdr->CRC, crc32(meta->cdr, size));
1311 error = g_write_data(cp, (lba + GET32(meta, hdr->cd_section)) * ss,
1316 size = GET32(meta, hdr->pdr_length) * ss;
1317 SET32(meta, pdr->CRC, 0xffffffff);
1318 SET32(meta, pdr->CRC, crc32(meta->pdr, size));
1319 error = g_write_data(cp, (lba + GET32(meta, hdr->pdr_section)) * ss,
1324 size = GET32(meta, hdr->vdr_length) * ss;
1325 SET32(meta, vdr->CRC, 0xffffffff);
1326 SET32(meta, vdr->CRC, crc32(meta->vdr, size));
1327 error = g_write_data(cp, (lba + GET32(meta, hdr->vdr_section)) * ss,
1332 size = GET16(meta, hdr->Configuration_Record_Length) * ss;
1333 num = GETCRNUM(meta);
1334 for (i = 0; i < num; i++) {
1335 vdc = GETVDCPTR(meta, i);
1336 SET32D(meta, vdc->CRC, 0xffffffff);
1337 SET32D(meta, vdc->CRC, crc32(vdc, size));
1339 error = g_write_data(cp, (lba + GET32(meta, hdr->cr_section)) * ss,
1340 meta->cr, size * num);
1344 size = GET32(meta, hdr->pdd_length) * ss;
1345 SET32(meta, pdd->CRC, 0xffffffff);
1346 SET32(meta, pdd->CRC, crc32(meta->pdd, size));
1347 error = g_write_data(cp, (lba + GET32(meta, hdr->pdd_section)) * ss,
1352 if (GET32(meta, hdr->bbmlog_length) != 0) {
1353 size = GET32(meta, hdr->bbmlog_length) * ss;
1354 SET32(meta, bbm->CRC, 0xffffffff);
1355 SET32(meta, bbm->CRC, crc32(meta->bbm, size));
1356 error = g_write_data(cp,
1357 (lba + GET32(meta, hdr->bbmlog_section)) * ss,
1364 if (lba == plba && slba != -1) {
1373 ddf_meta_erase(struct g_consumer *cp)
1375 struct g_provider *pp;
1380 buf = malloc(pp->sectorsize, M_MD_DDF, M_WAITOK | M_ZERO);
1381 error = g_write_data(cp, pp->mediasize - pp->sectorsize,
1382 buf, pp->sectorsize);
1384 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
1387 free(buf, M_MD_DDF);
1391 static struct g_raid_volume *
1392 g_raid_md_ddf_get_volume(struct g_raid_softc *sc, uint8_t *GUID)
1394 struct g_raid_volume *vol;
1395 struct g_raid_md_ddf_pervolume *pv;
1397 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1398 pv = vol->v_md_data;
1399 if (memcmp(pv->pv_meta.vde->VD_GUID, GUID, 24) == 0)
1405 static struct g_raid_disk *
1406 g_raid_md_ddf_get_disk(struct g_raid_softc *sc, uint8_t *GUID, uint32_t id)
1408 struct g_raid_disk *disk;
1409 struct g_raid_md_ddf_perdisk *pd;
1410 struct ddf_meta *meta;
1412 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1413 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1414 meta = &pd->pd_meta;
1416 if (memcmp(meta->pdd->PD_GUID, GUID, 24) == 0)
1419 if (GET32(meta, pdd->PD_Reference) == id)
1427 g_raid_md_ddf_purge_volumes(struct g_raid_softc *sc)
1429 struct g_raid_volume *vol, *tvol;
1433 TAILQ_FOREACH_SAFE(vol, &sc->sc_volumes, v_next, tvol) {
1434 if (vol->v_stopping)
1436 for (i = 0; i < vol->v_disks_count; i++) {
1437 if (vol->v_subdisks[i].sd_state != G_RAID_SUBDISK_S_NONE)
1440 if (i >= vol->v_disks_count) {
1441 g_raid_destroy_volume(vol);
1449 g_raid_md_ddf_purge_disks(struct g_raid_softc *sc)
1452 struct g_raid_disk *disk, *tdisk;
1453 struct g_raid_volume *vol;
1454 struct g_raid_md_ddf_perdisk *pd;
1458 TAILQ_FOREACH_SAFE(disk, &sc->sc_disks, d_next, tdisk) {
1459 if (disk->d_state == G_RAID_DISK_S_SPARE)
1461 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1463 /* Scan for deleted volumes. */
1464 for (i = 0; i < pd->pd_subdisks; ) {
1465 vol = g_raid_md_ddf_get_volume(sc,
1466 pd->pd_meta[i]->volume_id);
1467 if (vol != NULL && !vol->v_stopping) {
1471 free(pd->pd_meta[i], M_MD_DDF);
1472 for (j = i; j < pd->pd_subdisks - 1; j++)
1473 pd->pd_meta[j] = pd->pd_meta[j + 1];
1474 pd->pd_meta[DDF_MAX_SUBDISKS - 1] = NULL;
1479 /* If there is no metadata left - erase and delete disk. */
1480 if (pd->pd_subdisks == 0) {
1481 ddf_meta_erase(disk->d_consumer);
1482 g_raid_destroy_disk(disk);
1492 g_raid_md_ddf_supported(int level, int qual, int disks, int force)
1495 if (disks > DDF_MAX_DISKS_HARD)
1498 case G_RAID_VOLUME_RL_RAID0:
1499 if (qual != G_RAID_VOLUME_RLQ_NONE)
1503 if (!force && disks < 2)
1506 case G_RAID_VOLUME_RL_RAID1:
1509 if (qual == G_RAID_VOLUME_RLQ_R1SM) {
1510 if (!force && disks != 2)
1512 } else if (qual == G_RAID_VOLUME_RLQ_R1MM) {
1513 if (!force && disks != 3)
1518 case G_RAID_VOLUME_RL_RAID3:
1519 if (qual != G_RAID_VOLUME_RLQ_R3P0 &&
1520 qual != G_RAID_VOLUME_RLQ_R3PN)
1525 case G_RAID_VOLUME_RL_RAID4:
1526 if (qual != G_RAID_VOLUME_RLQ_R4P0 &&
1527 qual != G_RAID_VOLUME_RLQ_R4PN)
1532 case G_RAID_VOLUME_RL_RAID5:
1533 if (qual != G_RAID_VOLUME_RLQ_R5RA &&
1534 qual != G_RAID_VOLUME_RLQ_R5RS &&
1535 qual != G_RAID_VOLUME_RLQ_R5LA &&
1536 qual != G_RAID_VOLUME_RLQ_R5LS)
1541 case G_RAID_VOLUME_RL_RAID6:
1542 if (qual != G_RAID_VOLUME_RLQ_R6RA &&
1543 qual != G_RAID_VOLUME_RLQ_R6RS &&
1544 qual != G_RAID_VOLUME_RLQ_R6LA &&
1545 qual != G_RAID_VOLUME_RLQ_R6LS)
1550 case G_RAID_VOLUME_RL_RAIDMDF:
1551 if (qual != G_RAID_VOLUME_RLQ_RMDFRA &&
1552 qual != G_RAID_VOLUME_RLQ_RMDFRS &&
1553 qual != G_RAID_VOLUME_RLQ_RMDFLA &&
1554 qual != G_RAID_VOLUME_RLQ_RMDFLS)
1559 case G_RAID_VOLUME_RL_RAID1E:
1560 if (qual != G_RAID_VOLUME_RLQ_R1EA &&
1561 qual != G_RAID_VOLUME_RLQ_R1EO)
1566 case G_RAID_VOLUME_RL_SINGLE:
1567 if (qual != G_RAID_VOLUME_RLQ_NONE)
1572 case G_RAID_VOLUME_RL_CONCAT:
1573 if (qual != G_RAID_VOLUME_RLQ_NONE)
1578 case G_RAID_VOLUME_RL_RAID5E:
1579 if (qual != G_RAID_VOLUME_RLQ_R5ERA &&
1580 qual != G_RAID_VOLUME_RLQ_R5ERS &&
1581 qual != G_RAID_VOLUME_RLQ_R5ELA &&
1582 qual != G_RAID_VOLUME_RLQ_R5ELS)
1587 case G_RAID_VOLUME_RL_RAID5EE:
1588 if (qual != G_RAID_VOLUME_RLQ_R5EERA &&
1589 qual != G_RAID_VOLUME_RLQ_R5EERS &&
1590 qual != G_RAID_VOLUME_RLQ_R5EELA &&
1591 qual != G_RAID_VOLUME_RLQ_R5EELS)
1596 case G_RAID_VOLUME_RL_RAID5R:
1597 if (qual != G_RAID_VOLUME_RLQ_R5RRA &&
1598 qual != G_RAID_VOLUME_RLQ_R5RRS &&
1599 qual != G_RAID_VOLUME_RLQ_R5RLA &&
1600 qual != G_RAID_VOLUME_RLQ_R5RLS)
1612 g_raid_md_ddf_start_disk(struct g_raid_disk *disk, struct g_raid_volume *vol)
1614 struct g_raid_softc *sc;
1615 struct g_raid_subdisk *sd;
1616 struct g_raid_md_ddf_perdisk *pd;
1617 struct g_raid_md_ddf_pervolume *pv;
1618 struct g_raid_md_ddf_object *mdi;
1619 struct ddf_vol_meta *vmeta;
1620 struct ddf_meta *pdmeta, *gmeta;
1621 struct ddf_vdc_record *vdc1;
1622 struct ddf_sa_record *sa;
1623 off_t size, eoff = 0, esize = 0;
1625 int disk_pos, md_disk_bvd = -1, md_disk_pos = -1, md_pde_pos;
1626 int i, resurrection = 0;
1630 mdi = (struct g_raid_md_ddf_object *)sc->sc_md;
1631 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1632 pdmeta = &pd->pd_meta;
1633 reference = GET32(&pd->pd_meta, pdd->PD_Reference);
1635 pv = vol->v_md_data;
1636 vmeta = &pv->pv_meta;
1637 gmeta = &mdi->mdio_meta;
1639 /* Find disk position in metadata by its reference. */
1640 disk_pos = ddf_meta_find_disk(vmeta, reference,
1641 &md_disk_bvd, &md_disk_pos);
1642 md_pde_pos = ddf_meta_find_pd(gmeta, NULL, reference);
1645 G_RAID_DEBUG1(1, sc,
1646 "Disk %s is not a present part of the volume %s",
1647 g_raid_get_diskname(disk), vol->v_name);
1649 /* Failed stale disk is useless for us. */
1650 if ((GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA) != 0) {
1651 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
1655 /* If disk has some metadata for this volume - erase. */
1656 if ((vdc1 = ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID)) != NULL)
1657 SET32D(pdmeta, vdc1->Signature, 0xffffffff);
1659 /* If we are in the start process, that's all for now. */
1660 if (!pv->pv_started)
1663 * If we have already started - try to get use of the disk.
1664 * Try to replace OFFLINE disks first, then FAILED.
1666 if (ddf_meta_count_vdc(&pd->pd_meta, NULL) >=
1667 GET16(&pd->pd_meta, hdr->Max_Partitions)) {
1668 G_RAID_DEBUG1(1, sc, "No free partitions on disk %s",
1669 g_raid_get_diskname(disk));
1672 ddf_meta_unused_range(&pd->pd_meta, &eoff, &esize);
1674 G_RAID_DEBUG1(1, sc, "No free space on disk %s",
1675 g_raid_get_diskname(disk));
1678 eoff *= pd->pd_meta.sectorsize;
1679 esize *= pd->pd_meta.sectorsize;
1681 for (i = 0; i < vol->v_disks_count; i++) {
1682 sd = &vol->v_subdisks[i];
1683 if (sd->sd_state != G_RAID_SUBDISK_S_NONE)
1685 if (sd->sd_state <= G_RAID_SUBDISK_S_FAILED &&
1687 vol->v_subdisks[i].sd_state < sd->sd_state))
1690 if (disk_pos >= 0 &&
1691 vol->v_raid_level != G_RAID_VOLUME_RL_CONCAT &&
1693 G_RAID_DEBUG1(1, sc, "Disk %s free space "
1694 "is too small (%ju < %ju)",
1695 g_raid_get_diskname(disk), esize, size);
1698 if (disk_pos >= 0) {
1699 if (vol->v_raid_level != G_RAID_VOLUME_RL_CONCAT)
1701 md_disk_bvd = disk_pos / GET16(vmeta, vdc->Primary_Element_Count); // XXX
1702 md_disk_pos = disk_pos % GET16(vmeta, vdc->Primary_Element_Count); // XXX
1705 if (disk->d_state == G_RAID_DISK_S_NONE)
1706 g_raid_change_disk_state(disk,
1707 G_RAID_DISK_S_STALE);
1712 * If spare is committable, delete spare record.
1713 * Othersize, mark it active and leave there.
1715 sa = ddf_meta_find_sa(&pd->pd_meta, 0);
1717 if ((GET8D(&pd->pd_meta, sa->Spare_Type) &
1718 DDF_SAR_TYPE_REVERTIBLE) == 0) {
1719 SET32D(&pd->pd_meta, sa->Signature, 0xffffffff);
1721 SET8D(&pd->pd_meta, sa->Spare_Type,
1722 GET8D(&pd->pd_meta, sa->Spare_Type) |
1723 DDF_SAR_TYPE_ACTIVE);
1727 G_RAID_DEBUG1(1, sc, "Disk %s takes pos %d in the volume %s",
1728 g_raid_get_diskname(disk), disk_pos, vol->v_name);
1732 sd = &vol->v_subdisks[disk_pos];
1734 if (resurrection && sd->sd_disk != NULL) {
1735 g_raid_change_disk_state(sd->sd_disk,
1736 G_RAID_DISK_S_STALE_FAILED);
1737 TAILQ_REMOVE(&sd->sd_disk->d_subdisks,
1740 vol->v_subdisks[disk_pos].sd_disk = disk;
1741 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1743 /* Welcome the new disk. */
1745 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1746 else if (GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA)
1747 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1749 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1752 sd->sd_offset = eoff;
1753 sd->sd_size = esize;
1754 } else if (pdmeta->cr != NULL &&
1755 (vdc1 = ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID)) != NULL) {
1756 val2 = (uint64_t *)&(vdc1->Physical_Disk_Sequence[GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
1757 sd->sd_offset = (off_t)GET64P(pdmeta, val2 + md_disk_pos) * 512;
1758 sd->sd_size = (off_t)GET64D(pdmeta, vdc1->Block_Count) * 512;
1762 /* Stale disk, almost same as new. */
1763 g_raid_change_subdisk_state(sd,
1764 G_RAID_SUBDISK_S_NEW);
1765 } else if (GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA) {
1767 g_raid_change_subdisk_state(sd,
1768 G_RAID_SUBDISK_S_FAILED);
1769 } else if ((GET16(gmeta, pdr->entry[md_pde_pos].PD_State) &
1770 (DDF_PDE_FAILED | DDF_PDE_REBUILD)) != 0) {
1771 /* Rebuilding disk. */
1772 g_raid_change_subdisk_state(sd,
1773 G_RAID_SUBDISK_S_REBUILD);
1774 sd->sd_rebuild_pos = 0;
1775 } else if ((GET8(vmeta, vde->VD_State) & DDF_VDE_DIRTY) != 0 ||
1776 (GET8(vmeta, vde->Init_State) & DDF_VDE_INIT_MASK) !=
1777 DDF_VDE_INIT_FULL) {
1778 /* Stale disk or dirty volume (unclean shutdown). */
1779 g_raid_change_subdisk_state(sd,
1780 G_RAID_SUBDISK_S_STALE);
1782 /* Up to date disk. */
1783 g_raid_change_subdisk_state(sd,
1784 G_RAID_SUBDISK_S_ACTIVE);
1786 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1787 G_RAID_EVENT_SUBDISK);
1789 return (resurrection);
1793 g_raid_md_ddf_refill(struct g_raid_softc *sc)
1795 struct g_raid_volume *vol;
1796 struct g_raid_subdisk *sd;
1797 struct g_raid_disk *disk;
1798 struct g_raid_md_object *md;
1799 struct g_raid_md_ddf_perdisk *pd;
1800 struct g_raid_md_ddf_pervolume *pv;
1801 int update, updated, i, bad;
1806 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1807 pv = vol->v_md_data;
1808 if (!pv->pv_started || vol->v_stopping)
1811 /* Search for subdisk that needs replacement. */
1813 for (i = 0; i < vol->v_disks_count; i++) {
1814 sd = &vol->v_subdisks[i];
1815 if (sd->sd_state == G_RAID_SUBDISK_S_NONE ||
1816 sd->sd_state == G_RAID_SUBDISK_S_FAILED)
1822 G_RAID_DEBUG1(1, sc, "Volume %s is not complete, "
1823 "trying to refill.", vol->v_name);
1825 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1827 if (disk->d_state < G_RAID_DISK_S_SPARE)
1829 /* Skip already used by this volume. */
1830 for (i = 0; i < vol->v_disks_count; i++) {
1831 sd = &vol->v_subdisks[i];
1832 if (sd->sd_disk == disk)
1835 if (i < vol->v_disks_count)
1838 /* Try to use disk if it has empty extents. */
1839 pd = disk->d_md_data;
1840 if (ddf_meta_count_vdc(&pd->pd_meta, NULL) <
1841 GET16(&pd->pd_meta, hdr->Max_Partitions)) {
1842 update = g_raid_md_ddf_start_disk(disk, vol);
1847 g_raid_md_write_ddf(md, vol, NULL, disk);
1857 g_raid_md_ddf_start(struct g_raid_volume *vol)
1859 struct g_raid_softc *sc;
1860 struct g_raid_subdisk *sd;
1861 struct g_raid_disk *disk;
1862 struct g_raid_md_object *md;
1863 struct g_raid_md_ddf_perdisk *pd;
1864 struct g_raid_md_ddf_pervolume *pv;
1865 struct g_raid_md_ddf_object *mdi;
1866 struct ddf_vol_meta *vmeta;
1872 mdi = (struct g_raid_md_ddf_object *)md;
1873 pv = vol->v_md_data;
1874 vmeta = &pv->pv_meta;
1876 vol->v_raid_level = GET8(vmeta, vdc->Primary_RAID_Level);
1877 vol->v_raid_level_qualifier = GET8(vmeta, vdc->RLQ);
1878 if (GET8(vmeta, vdc->Secondary_Element_Count) > 1 &&
1879 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 &&
1880 GET8(vmeta, vdc->Secondary_RAID_Level) == 0)
1881 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1882 vol->v_sectorsize = GET16(vmeta, vdc->Block_Size);
1883 if (vol->v_sectorsize == 0xffff)
1884 vol->v_sectorsize = vmeta->sectorsize;
1885 vol->v_strip_size = vol->v_sectorsize << GET8(vmeta, vdc->Stripe_Size);
1886 vol->v_disks_count = GET16(vmeta, vdc->Primary_Element_Count) *
1887 GET8(vmeta, vdc->Secondary_Element_Count);
1888 vol->v_mdf_pdisks = GET8(vmeta, vdc->MDF_Parity_Disks);
1889 vol->v_mdf_polynomial = GET16(vmeta, vdc->MDF_Parity_Generator_Polynomial);
1890 vol->v_mdf_method = GET8(vmeta, vdc->MDF_Constant_Generation_Method);
1891 if (GET8(vmeta, vdc->Rotate_Parity_count) > 31)
1892 vol->v_rotate_parity = 1;
1894 vol->v_rotate_parity = 1 << GET8(vmeta, vdc->Rotate_Parity_count);
1895 vol->v_mediasize = GET64(vmeta, vdc->VD_Size) * vol->v_sectorsize;
1896 for (i = 0, j = 0, bvd = 0; i < vol->v_disks_count; i++, j++) {
1897 if (j == GET16(vmeta, vdc->Primary_Element_Count)) {
1901 sd = &vol->v_subdisks[i];
1902 if (vmeta->bvdc[bvd] == NULL) {
1904 sd->sd_size = GET64(vmeta, vdc->Block_Count) *
1908 val2 = (uint64_t *)&(vmeta->bvdc[bvd]->Physical_Disk_Sequence[
1909 GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
1910 sd->sd_offset = GET64P(vmeta, val2 + j) * vol->v_sectorsize;
1911 sd->sd_size = GET64(vmeta, bvdc[bvd]->Block_Count) *
1914 g_raid_start_volume(vol);
1916 /* Make all disks found till the moment take their places. */
1917 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1918 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1919 if (ddf_meta_find_vdc(&pd->pd_meta, vmeta->vdc->VD_GUID) != NULL)
1920 g_raid_md_ddf_start_disk(disk, vol);
1924 mdi->mdio_starting--;
1925 callout_stop(&pv->pv_start_co);
1926 G_RAID_DEBUG1(0, sc, "Volume started.");
1927 g_raid_md_write_ddf(md, vol, NULL, NULL);
1929 /* Pickup any STALE/SPARE disks to refill array if needed. */
1930 g_raid_md_ddf_refill(sc);
1932 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
1936 g_raid_ddf_go(void *arg)
1938 struct g_raid_volume *vol;
1939 struct g_raid_softc *sc;
1940 struct g_raid_md_ddf_pervolume *pv;
1943 pv = vol->v_md_data;
1945 if (!pv->pv_started) {
1946 G_RAID_DEBUG1(0, sc, "Force volume start due to timeout.");
1947 g_raid_event_send(vol, G_RAID_VOLUME_E_STARTMD,
1948 G_RAID_EVENT_VOLUME);
1953 g_raid_md_ddf_new_disk(struct g_raid_disk *disk)
1955 struct g_raid_softc *sc;
1956 struct g_raid_md_object *md;
1957 struct g_raid_md_ddf_perdisk *pd;
1958 struct g_raid_md_ddf_pervolume *pv;
1959 struct g_raid_md_ddf_object *mdi;
1960 struct g_raid_volume *vol;
1961 struct ddf_meta *pdmeta;
1962 struct ddf_vol_meta *vmeta;
1963 struct ddf_vdc_record *vdc;
1964 struct ddf_vd_entry *vde;
1965 int i, j, k, num, have, need, cnt, spare;
1971 mdi = (struct g_raid_md_ddf_object *)md;
1972 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1973 pdmeta = &pd->pd_meta;
1976 if (mdi->mdio_meta.hdr == NULL)
1977 ddf_meta_copy(&mdi->mdio_meta, pdmeta);
1979 ddf_meta_update(&mdi->mdio_meta, pdmeta);
1981 num = GETCRNUM(pdmeta);
1982 for (j = 0; j < num; j++) {
1983 vdc = GETVDCPTR(pdmeta, j);
1984 val = GET32D(pdmeta, vdc->Signature);
1986 if (val == DDF_SA_SIGNATURE && spare == -1)
1989 if (val != DDF_VDCR_SIGNATURE)
1992 k = ddf_meta_find_vd(pdmeta, vdc->VD_GUID);
1995 vde = &pdmeta->vdr->entry[k];
1997 /* Look for volume with matching ID. */
1998 vol = g_raid_md_ddf_get_volume(sc, vdc->VD_GUID);
2000 ddf_meta_get_name(pdmeta, k, buf);
2001 vol = g_raid_create_volume(sc, buf,
2002 GET16D(pdmeta, vde->VD_Number));
2003 pv = malloc(sizeof(*pv), M_MD_DDF, M_WAITOK | M_ZERO);
2004 vol->v_md_data = pv;
2005 callout_init(&pv->pv_start_co, 1);
2006 callout_reset(&pv->pv_start_co,
2007 g_raid_start_timeout * hz,
2008 g_raid_ddf_go, vol);
2009 mdi->mdio_starting++;
2011 pv = vol->v_md_data;
2013 /* If we haven't started yet - check metadata freshness. */
2014 vmeta = &pv->pv_meta;
2015 ddf_vol_meta_update(vmeta, pdmeta, vdc->VD_GUID, pv->pv_started);
2019 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
2020 g_raid_md_ddf_refill(sc);
2023 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2024 pv = vol->v_md_data;
2025 vmeta = &pv->pv_meta;
2027 if (ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID) == NULL)
2030 if (pv->pv_started) {
2031 if (g_raid_md_ddf_start_disk(disk, vol))
2032 g_raid_md_write_ddf(md, vol, NULL, NULL);
2036 /* If we collected all needed disks - start array. */
2039 for (k = 0; k < GET8(vmeta, vdc->Secondary_Element_Count); k++) {
2040 if (vmeta->bvdc[k] == NULL) {
2041 need += GET16(vmeta, vdc->Primary_Element_Count);
2044 cnt = GET16(vmeta, bvdc[k]->Primary_Element_Count);
2046 for (i = 0; i < cnt; i++) {
2047 val = GET32(vmeta, bvdc[k]->Physical_Disk_Sequence[i]);
2048 if (g_raid_md_ddf_get_disk(sc, NULL, val) != NULL)
2052 G_RAID_DEBUG1(1, sc, "Volume %s now has %d of %d disks",
2053 vol->v_name, have, need);
2055 g_raid_md_ddf_start(vol);
2060 g_raid_md_create_req_ddf(struct g_raid_md_object *md, struct g_class *mp,
2061 struct gctl_req *req, struct g_geom **gp)
2063 struct g_geom *geom;
2064 struct g_raid_softc *sc;
2065 struct g_raid_md_ddf_object *mdi, *mdi1;
2070 mdi = (struct g_raid_md_ddf_object *)md;
2071 fmtopt = gctl_get_asciiparam(req, "fmtopt");
2072 if (fmtopt == NULL || strcasecmp(fmtopt, "BE") == 0)
2074 else if (strcasecmp(fmtopt, "LE") == 0)
2077 gctl_error(req, "Incorrect fmtopt argument.");
2078 return (G_RAID_MD_TASTE_FAIL);
2081 /* Search for existing node. */
2082 LIST_FOREACH(geom, &mp->geom, geom) {
2086 if (sc->sc_stopping != 0)
2088 if (sc->sc_md->mdo_class != md->mdo_class)
2090 mdi1 = (struct g_raid_md_ddf_object *)sc->sc_md;
2091 if (mdi1->mdio_bigendian != be)
2097 return (G_RAID_MD_TASTE_EXISTING);
2100 /* Create new one if not found. */
2101 mdi->mdio_bigendian = be;
2102 snprintf(name, sizeof(name), "DDF%s", be ? "" : "-LE");
2103 sc = g_raid_create_node(mp, name, md);
2105 return (G_RAID_MD_TASTE_FAIL);
2108 return (G_RAID_MD_TASTE_NEW);
2112 g_raid_md_taste_ddf(struct g_raid_md_object *md, struct g_class *mp,
2113 struct g_consumer *cp, struct g_geom **gp)
2115 struct g_consumer *rcp;
2116 struct g_provider *pp;
2117 struct g_raid_softc *sc;
2118 struct g_raid_disk *disk;
2119 struct ddf_meta meta;
2120 struct g_raid_md_ddf_perdisk *pd;
2121 struct g_raid_md_ddf_object *mdi;
2122 struct g_geom *geom;
2123 int error, result, be;
2126 G_RAID_DEBUG(1, "Tasting DDF on %s", cp->provider->name);
2127 mdi = (struct g_raid_md_ddf_object *)md;
2130 /* Read metadata from device. */
2131 g_topology_unlock();
2132 bzero(&meta, sizeof(meta));
2133 error = ddf_meta_read(cp, &meta);
2136 return (G_RAID_MD_TASTE_FAIL);
2137 be = meta.bigendian;
2139 /* Metadata valid. Print it. */
2140 g_raid_md_ddf_print(&meta);
2142 /* Search for matching node. */
2144 LIST_FOREACH(geom, &mp->geom, geom) {
2148 if (sc->sc_stopping != 0)
2150 if (sc->sc_md->mdo_class != md->mdo_class)
2152 mdi = (struct g_raid_md_ddf_object *)sc->sc_md;
2153 if (mdi->mdio_bigendian != be)
2158 /* Found matching node. */
2160 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
2161 result = G_RAID_MD_TASTE_EXISTING;
2163 } else { /* Not found matching node -- create one. */
2164 result = G_RAID_MD_TASTE_NEW;
2165 mdi->mdio_bigendian = be;
2166 snprintf(name, sizeof(name), "DDF%s", be ? "" : "-LE");
2167 sc = g_raid_create_node(mp, name, md);
2172 /* There is no return after this point, so we close passed consumer. */
2173 g_access(cp, -1, 0, 0);
2175 rcp = g_new_consumer(geom);
2176 rcp->flags |= G_CF_DIRECT_RECEIVE;
2178 if (g_access(rcp, 1, 1, 1) != 0)
2181 g_topology_unlock();
2182 sx_xlock(&sc->sc_lock);
2184 pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2186 disk = g_raid_create_disk(sc);
2187 disk->d_md_data = (void *)pd;
2188 disk->d_consumer = rcp;
2189 rcp->private = disk;
2191 g_raid_get_disk_info(disk);
2193 g_raid_md_ddf_new_disk(disk);
2195 sx_xunlock(&sc->sc_lock);
2202 g_raid_md_event_ddf(struct g_raid_md_object *md,
2203 struct g_raid_disk *disk, u_int event)
2205 struct g_raid_softc *sc;
2211 case G_RAID_DISK_E_DISCONNECTED:
2213 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2214 g_raid_destroy_disk(disk);
2215 g_raid_md_ddf_purge_volumes(sc);
2217 /* Write updated metadata to all disks. */
2218 g_raid_md_write_ddf(md, NULL, NULL, NULL);
2220 /* Check if anything left. */
2221 if (g_raid_ndisks(sc, -1) == 0)
2222 g_raid_destroy_node(sc, 0);
2224 g_raid_md_ddf_refill(sc);
2231 g_raid_md_volume_event_ddf(struct g_raid_md_object *md,
2232 struct g_raid_volume *vol, u_int event)
2234 struct g_raid_md_ddf_pervolume *pv;
2236 pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2238 case G_RAID_VOLUME_E_STARTMD:
2239 if (!pv->pv_started)
2240 g_raid_md_ddf_start(vol);
2247 g_raid_md_ctl_ddf(struct g_raid_md_object *md,
2248 struct gctl_req *req)
2250 struct g_raid_softc *sc;
2251 struct g_raid_volume *vol, *vol1;
2252 struct g_raid_subdisk *sd;
2253 struct g_raid_disk *disk, *disks[DDF_MAX_DISKS_HARD];
2254 struct g_raid_md_ddf_perdisk *pd;
2255 struct g_raid_md_ddf_pervolume *pv;
2256 struct g_raid_md_ddf_object *mdi;
2257 struct ddf_sa_record *sa;
2258 struct g_consumer *cp;
2259 struct g_provider *pp;
2261 const char *nodename, *verb, *volname, *levelname, *diskname;
2264 off_t size, sectorsize, strip, offs[DDF_MAX_DISKS_HARD], esize;
2265 intmax_t *sizearg, *striparg;
2266 int i, numdisks, len, level, qual;
2270 mdi = (struct g_raid_md_ddf_object *)md;
2271 verb = gctl_get_param(req, "verb", NULL);
2272 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
2275 if (strcmp(verb, "label") == 0) {
2277 gctl_error(req, "Invalid number of arguments.");
2280 volname = gctl_get_asciiparam(req, "arg1");
2281 if (volname == NULL) {
2282 gctl_error(req, "No volume name.");
2285 levelname = gctl_get_asciiparam(req, "arg2");
2286 if (levelname == NULL) {
2287 gctl_error(req, "No RAID level.");
2290 if (g_raid_volume_str2level(levelname, &level, &qual)) {
2291 gctl_error(req, "Unknown RAID level '%s'.", levelname);
2294 numdisks = *nargs - 3;
2295 force = gctl_get_paraml(req, "force", sizeof(*force));
2296 if (!g_raid_md_ddf_supported(level, qual, numdisks,
2297 force ? *force : 0)) {
2298 gctl_error(req, "Unsupported RAID level "
2299 "(0x%02x/0x%02x), or number of disks (%d).",
2300 level, qual, numdisks);
2304 /* Search for disks, connect them and probe. */
2307 bzero(disks, sizeof(disks));
2308 bzero(offs, sizeof(offs));
2309 for (i = 0; i < numdisks; i++) {
2310 snprintf(arg, sizeof(arg), "arg%d", i + 3);
2311 diskname = gctl_get_asciiparam(req, arg);
2312 if (diskname == NULL) {
2313 gctl_error(req, "No disk name (%s).", arg);
2317 if (strcmp(diskname, "NONE") == 0)
2320 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2321 if (disk->d_consumer != NULL &&
2322 disk->d_consumer->provider != NULL &&
2323 strcmp(disk->d_consumer->provider->name,
2328 if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2329 gctl_error(req, "Disk '%s' is in a "
2330 "wrong state (%s).", diskname,
2331 g_raid_disk_state2str(disk->d_state));
2335 pd = disk->d_md_data;
2336 if (ddf_meta_count_vdc(&pd->pd_meta, NULL) >=
2337 GET16(&pd->pd_meta, hdr->Max_Partitions)) {
2338 gctl_error(req, "No free partitions "
2344 pp = disk->d_consumer->provider;
2346 ddf_meta_unused_range(&pd->pd_meta,
2348 offs[i] *= pp->sectorsize;
2349 size = MIN(size, (off_t)esize * pp->sectorsize);
2350 sectorsize = MAX(sectorsize, pp->sectorsize);
2355 cp = g_raid_open_consumer(sc, diskname);
2357 gctl_error(req, "Can't open disk '%s'.",
2359 g_topology_unlock();
2364 pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2365 disk = g_raid_create_disk(sc);
2366 disk->d_md_data = (void *)pd;
2367 disk->d_consumer = cp;
2370 ddf_meta_create(disk, &mdi->mdio_meta);
2371 if (mdi->mdio_meta.hdr == NULL)
2372 ddf_meta_copy(&mdi->mdio_meta, &pd->pd_meta);
2374 ddf_meta_update(&mdi->mdio_meta, &pd->pd_meta);
2375 g_topology_unlock();
2377 g_raid_get_disk_info(disk);
2379 /* Reserve some space for metadata. */
2380 size = MIN(size, GET64(&pd->pd_meta,
2381 pdr->entry[0].Configured_Size) * pp->sectorsize);
2382 sectorsize = MAX(sectorsize, pp->sectorsize);
2385 for (i = 0; i < numdisks; i++) {
2386 if (disks[i] != NULL &&
2387 disks[i]->d_state == G_RAID_DISK_S_NONE)
2388 g_raid_destroy_disk(disks[i]);
2393 if (sectorsize <= 0) {
2394 gctl_error(req, "Can't get sector size.");
2398 /* Handle size argument. */
2399 len = sizeof(*sizearg);
2400 sizearg = gctl_get_param(req, "size", &len);
2401 if (sizearg != NULL && len == sizeof(*sizearg) &&
2403 if (*sizearg > size) {
2404 gctl_error(req, "Size too big %lld > %lld.",
2405 (long long)*sizearg, (long long)size);
2411 /* Handle strip argument. */
2413 len = sizeof(*striparg);
2414 striparg = gctl_get_param(req, "strip", &len);
2415 if (striparg != NULL && len == sizeof(*striparg) &&
2417 if (*striparg < sectorsize) {
2418 gctl_error(req, "Strip size too small.");
2421 if (*striparg % sectorsize != 0) {
2422 gctl_error(req, "Incorrect strip size.");
2428 /* Round size down to strip or sector. */
2429 if (level == G_RAID_VOLUME_RL_RAID1 ||
2430 level == G_RAID_VOLUME_RL_RAID3 ||
2431 level == G_RAID_VOLUME_RL_SINGLE ||
2432 level == G_RAID_VOLUME_RL_CONCAT)
2433 size -= (size % sectorsize);
2434 else if (level == G_RAID_VOLUME_RL_RAID1E &&
2435 (numdisks & 1) != 0)
2436 size -= (size % (2 * strip));
2438 size -= (size % strip);
2440 gctl_error(req, "Size too small.");
2444 /* We have all we need, create things: volume, ... */
2445 pv = malloc(sizeof(*pv), M_MD_DDF, M_WAITOK | M_ZERO);
2446 ddf_vol_meta_create(&pv->pv_meta, &mdi->mdio_meta);
2448 vol = g_raid_create_volume(sc, volname, -1);
2449 vol->v_md_data = pv;
2450 vol->v_raid_level = level;
2451 vol->v_raid_level_qualifier = qual;
2452 vol->v_strip_size = strip;
2453 vol->v_disks_count = numdisks;
2454 if (level == G_RAID_VOLUME_RL_RAID0 ||
2455 level == G_RAID_VOLUME_RL_CONCAT ||
2456 level == G_RAID_VOLUME_RL_SINGLE)
2457 vol->v_mediasize = size * numdisks;
2458 else if (level == G_RAID_VOLUME_RL_RAID1)
2459 vol->v_mediasize = size;
2460 else if (level == G_RAID_VOLUME_RL_RAID3 ||
2461 level == G_RAID_VOLUME_RL_RAID4 ||
2462 level == G_RAID_VOLUME_RL_RAID5)
2463 vol->v_mediasize = size * (numdisks - 1);
2464 else if (level == G_RAID_VOLUME_RL_RAID5R) {
2465 vol->v_mediasize = size * (numdisks - 1);
2466 vol->v_rotate_parity = 1024;
2467 } else if (level == G_RAID_VOLUME_RL_RAID6 ||
2468 level == G_RAID_VOLUME_RL_RAID5E ||
2469 level == G_RAID_VOLUME_RL_RAID5EE)
2470 vol->v_mediasize = size * (numdisks - 2);
2471 else if (level == G_RAID_VOLUME_RL_RAIDMDF) {
2473 vol->v_mdf_pdisks = 2;
2475 vol->v_mdf_pdisks = 3;
2476 vol->v_mdf_polynomial = 0x11d;
2477 vol->v_mdf_method = 0x00;
2478 vol->v_mediasize = size * (numdisks - vol->v_mdf_pdisks);
2479 } else { /* RAID1E */
2480 vol->v_mediasize = ((size * numdisks) / strip / 2) *
2483 vol->v_sectorsize = sectorsize;
2484 g_raid_start_volume(vol);
2486 /* , and subdisks. */
2487 for (i = 0; i < numdisks; i++) {
2489 sd = &vol->v_subdisks[i];
2491 sd->sd_offset = offs[i];
2495 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
2496 g_raid_change_disk_state(disk,
2497 G_RAID_DISK_S_ACTIVE);
2498 g_raid_change_subdisk_state(sd,
2499 G_RAID_SUBDISK_S_ACTIVE);
2500 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
2501 G_RAID_EVENT_SUBDISK);
2504 /* Write metadata based on created entities. */
2505 G_RAID_DEBUG1(0, sc, "Array started.");
2506 g_raid_md_write_ddf(md, vol, NULL, NULL);
2508 /* Pickup any STALE/SPARE disks to refill array if needed. */
2509 g_raid_md_ddf_refill(sc);
2511 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
2512 G_RAID_EVENT_VOLUME);
2515 if (strcmp(verb, "add") == 0) {
2516 gctl_error(req, "`add` command is not applicable, "
2517 "use `label` instead.");
2520 if (strcmp(verb, "delete") == 0) {
2521 nodename = gctl_get_asciiparam(req, "arg0");
2522 if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2525 /* Full node destruction. */
2526 if (*nargs == 1 && nodename != NULL) {
2527 /* Check if some volume is still open. */
2528 force = gctl_get_paraml(req, "force", sizeof(*force));
2529 if (force != NULL && *force == 0 &&
2530 g_raid_nopens(sc) != 0) {
2531 gctl_error(req, "Some volume is still open.");
2535 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2536 if (disk->d_consumer)
2537 ddf_meta_erase(disk->d_consumer);
2539 g_raid_destroy_node(sc, 0);
2543 /* Destroy specified volume. If it was last - all node. */
2545 gctl_error(req, "Invalid number of arguments.");
2548 volname = gctl_get_asciiparam(req,
2549 nodename != NULL ? "arg1" : "arg0");
2550 if (volname == NULL) {
2551 gctl_error(req, "No volume name.");
2555 /* Search for volume. */
2556 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2557 if (strcmp(vol->v_name, volname) == 0)
2559 pp = vol->v_provider;
2562 if (strcmp(pp->name, volname) == 0)
2564 if (strncmp(pp->name, "raid/", 5) == 0 &&
2565 strcmp(pp->name + 5, volname) == 0)
2569 i = strtol(volname, &tmp, 10);
2570 if (verb != volname && tmp[0] == 0) {
2571 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2572 if (vol->v_global_id == i)
2578 gctl_error(req, "Volume '%s' not found.", volname);
2582 /* Check if volume is still open. */
2583 force = gctl_get_paraml(req, "force", sizeof(*force));
2584 if (force != NULL && *force == 0 &&
2585 vol->v_provider_open != 0) {
2586 gctl_error(req, "Volume is still open.");
2590 /* Destroy volume and potentially node. */
2592 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2595 g_raid_destroy_volume(vol);
2596 g_raid_md_ddf_purge_disks(sc);
2597 g_raid_md_write_ddf(md, NULL, NULL, NULL);
2599 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2600 if (disk->d_consumer)
2601 ddf_meta_erase(disk->d_consumer);
2603 g_raid_destroy_node(sc, 0);
2607 if (strcmp(verb, "remove") == 0 ||
2608 strcmp(verb, "fail") == 0) {
2610 gctl_error(req, "Invalid number of arguments.");
2613 for (i = 1; i < *nargs; i++) {
2614 snprintf(arg, sizeof(arg), "arg%d", i);
2615 diskname = gctl_get_asciiparam(req, arg);
2616 if (diskname == NULL) {
2617 gctl_error(req, "No disk name (%s).", arg);
2621 if (strncmp(diskname, _PATH_DEV, 5) == 0)
2624 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2625 if (disk->d_consumer != NULL &&
2626 disk->d_consumer->provider != NULL &&
2627 strcmp(disk->d_consumer->provider->name,
2632 gctl_error(req, "Disk '%s' not found.",
2638 if (strcmp(verb, "fail") == 0) {
2639 g_raid_md_fail_disk_ddf(md, NULL, disk);
2643 /* Erase metadata on deleting disk and destroy it. */
2644 ddf_meta_erase(disk->d_consumer);
2645 g_raid_destroy_disk(disk);
2647 g_raid_md_ddf_purge_volumes(sc);
2649 /* Write updated metadata to remaining disks. */
2650 g_raid_md_write_ddf(md, NULL, NULL, NULL);
2652 /* Check if anything left. */
2653 if (g_raid_ndisks(sc, -1) == 0)
2654 g_raid_destroy_node(sc, 0);
2656 g_raid_md_ddf_refill(sc);
2659 if (strcmp(verb, "insert") == 0) {
2661 gctl_error(req, "Invalid number of arguments.");
2664 for (i = 1; i < *nargs; i++) {
2665 /* Get disk name. */
2666 snprintf(arg, sizeof(arg), "arg%d", i);
2667 diskname = gctl_get_asciiparam(req, arg);
2668 if (diskname == NULL) {
2669 gctl_error(req, "No disk name (%s).", arg);
2674 /* Try to find provider with specified name. */
2676 cp = g_raid_open_consumer(sc, diskname);
2678 gctl_error(req, "Can't open disk '%s'.",
2680 g_topology_unlock();
2685 g_topology_unlock();
2687 pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2689 disk = g_raid_create_disk(sc);
2690 disk->d_consumer = cp;
2691 disk->d_md_data = (void *)pd;
2694 g_raid_get_disk_info(disk);
2696 /* Welcome the "new" disk. */
2697 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
2698 ddf_meta_create(disk, &mdi->mdio_meta);
2699 sa = ddf_meta_find_sa(&pd->pd_meta, 1);
2701 SET32D(&pd->pd_meta, sa->Signature,
2703 SET8D(&pd->pd_meta, sa->Spare_Type, 0);
2704 SET16D(&pd->pd_meta, sa->Populated_SAEs, 0);
2705 SET16D(&pd->pd_meta, sa->MAX_SAE_Supported,
2706 (GET16(&pd->pd_meta, hdr->Configuration_Record_Length) *
2707 pd->pd_meta.sectorsize -
2708 sizeof(struct ddf_sa_record)) /
2709 sizeof(struct ddf_sa_entry));
2711 if (mdi->mdio_meta.hdr == NULL)
2712 ddf_meta_copy(&mdi->mdio_meta, &pd->pd_meta);
2714 ddf_meta_update(&mdi->mdio_meta, &pd->pd_meta);
2715 g_raid_md_write_ddf(md, NULL, NULL, NULL);
2716 g_raid_md_ddf_refill(sc);
2724 g_raid_md_write_ddf(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2725 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2727 struct g_raid_softc *sc;
2728 struct g_raid_volume *vol;
2729 struct g_raid_subdisk *sd;
2730 struct g_raid_disk *disk;
2731 struct g_raid_md_ddf_perdisk *pd;
2732 struct g_raid_md_ddf_pervolume *pv;
2733 struct g_raid_md_ddf_object *mdi;
2734 struct ddf_meta *gmeta;
2735 struct ddf_vol_meta *vmeta;
2736 struct ddf_vdc_record *vdc;
2737 struct ddf_sa_record *sa;
2739 int i, j, pos, bvd, size;
2742 mdi = (struct g_raid_md_ddf_object *)md;
2743 gmeta = &mdi->mdio_meta;
2745 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2749 * Clear disk flags to let only really needed ones to be reset.
2750 * Do it only if there are no volumes in starting state now,
2751 * as they can update disk statuses yet and we may kill innocent.
2753 if (mdi->mdio_starting == 0) {
2754 for (i = 0; i < GET16(gmeta, pdr->Populated_PDEs); i++) {
2755 if (isff(gmeta->pdr->entry[i].PD_GUID, 24))
2757 SET16(gmeta, pdr->entry[i].PD_Type,
2758 GET16(gmeta, pdr->entry[i].PD_Type) &
2759 ~(DDF_PDE_PARTICIPATING |
2760 DDF_PDE_GLOBAL_SPARE | DDF_PDE_CONFIG_SPARE));
2761 if ((GET16(gmeta, pdr->entry[i].PD_State) &
2763 SET16(gmeta, pdr->entry[i].PD_State, 0);
2767 /* Generate/update new per-volume metadata. */
2768 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2769 pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2770 if (vol->v_stopping || !pv->pv_started)
2772 vmeta = &pv->pv_meta;
2774 SET32(vmeta, vdc->Sequence_Number,
2775 GET32(vmeta, vdc->Sequence_Number) + 1);
2776 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E &&
2777 vol->v_disks_count % 2 == 0)
2778 SET16(vmeta, vdc->Primary_Element_Count, 2);
2780 SET16(vmeta, vdc->Primary_Element_Count,
2781 vol->v_disks_count);
2782 SET8(vmeta, vdc->Stripe_Size,
2783 ffs(vol->v_strip_size / vol->v_sectorsize) - 1);
2784 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E &&
2785 vol->v_disks_count % 2 == 0) {
2786 SET8(vmeta, vdc->Primary_RAID_Level,
2788 SET8(vmeta, vdc->RLQ, 0);
2789 SET8(vmeta, vdc->Secondary_Element_Count,
2790 vol->v_disks_count / 2);
2791 SET8(vmeta, vdc->Secondary_RAID_Level, 0);
2793 SET8(vmeta, vdc->Primary_RAID_Level,
2795 SET8(vmeta, vdc->RLQ,
2796 vol->v_raid_level_qualifier);
2797 SET8(vmeta, vdc->Secondary_Element_Count, 1);
2798 SET8(vmeta, vdc->Secondary_RAID_Level, 0);
2800 SET8(vmeta, vdc->Secondary_Element_Seq, 0);
2801 SET64(vmeta, vdc->Block_Count, 0);
2802 SET64(vmeta, vdc->VD_Size, vol->v_mediasize / vol->v_sectorsize);
2803 SET16(vmeta, vdc->Block_Size, vol->v_sectorsize);
2804 SET8(vmeta, vdc->Rotate_Parity_count,
2805 fls(vol->v_rotate_parity) - 1);
2806 SET8(vmeta, vdc->MDF_Parity_Disks, vol->v_mdf_pdisks);
2807 SET16(vmeta, vdc->MDF_Parity_Generator_Polynomial,
2808 vol->v_mdf_polynomial);
2809 SET8(vmeta, vdc->MDF_Constant_Generation_Method,
2812 SET16(vmeta, vde->VD_Number, vol->v_global_id);
2813 if (vol->v_state <= G_RAID_VOLUME_S_BROKEN)
2814 SET8(vmeta, vde->VD_State, DDF_VDE_FAILED);
2815 else if (vol->v_state <= G_RAID_VOLUME_S_DEGRADED)
2816 SET8(vmeta, vde->VD_State, DDF_VDE_DEGRADED);
2817 else if (vol->v_state <= G_RAID_VOLUME_S_SUBOPTIMAL)
2818 SET8(vmeta, vde->VD_State, DDF_VDE_PARTIAL);
2820 SET8(vmeta, vde->VD_State, DDF_VDE_OPTIMAL);
2822 g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_STALE) > 0 ||
2823 g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_RESYNC) > 0)
2824 SET8(vmeta, vde->VD_State,
2825 GET8(vmeta, vde->VD_State) | DDF_VDE_DIRTY);
2826 SET8(vmeta, vde->Init_State, DDF_VDE_INIT_FULL); // XXX
2827 ddf_meta_put_name(vmeta, vol->v_name);
2829 for (i = 0; i < vol->v_disks_count; i++) {
2830 sd = &vol->v_subdisks[i];
2831 bvd = i / GET16(vmeta, vdc->Primary_Element_Count);
2832 pos = i % GET16(vmeta, vdc->Primary_Element_Count);
2835 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2836 if (vmeta->bvdc[bvd] == NULL) {
2838 hdr->Configuration_Record_Length) *
2840 vmeta->bvdc[bvd] = malloc(size,
2841 M_MD_DDF, M_WAITOK);
2842 memset(vmeta->bvdc[bvd], 0xff, size);
2844 memcpy(vmeta->bvdc[bvd], vmeta->vdc,
2845 sizeof(struct ddf_vdc_record));
2846 SET8(vmeta, bvdc[bvd]->Secondary_Element_Seq, bvd);
2847 SET64(vmeta, bvdc[bvd]->Block_Count,
2848 sd->sd_size / vol->v_sectorsize);
2849 SET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos],
2850 GET32(&pd->pd_meta, pdd->PD_Reference));
2851 val2 = (uint64_t *)&(vmeta->bvdc[bvd]->Physical_Disk_Sequence[
2852 GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
2853 SET64P(vmeta, val2 + pos,
2854 sd->sd_offset / vol->v_sectorsize);
2856 if (vmeta->bvdc[bvd] == NULL)
2859 j = ddf_meta_find_pd(gmeta, NULL,
2860 GET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos]));
2863 SET16(gmeta, pdr->entry[j].PD_Type,
2864 GET16(gmeta, pdr->entry[j].PD_Type) |
2865 DDF_PDE_PARTICIPATING);
2866 if (sd->sd_state == G_RAID_SUBDISK_S_NONE)
2867 SET16(gmeta, pdr->entry[j].PD_State,
2868 GET16(gmeta, pdr->entry[j].PD_State) |
2869 (DDF_PDE_FAILED | DDF_PDE_MISSING));
2870 else if (sd->sd_state == G_RAID_SUBDISK_S_FAILED)
2871 SET16(gmeta, pdr->entry[j].PD_State,
2872 GET16(gmeta, pdr->entry[j].PD_State) |
2873 (DDF_PDE_FAILED | DDF_PDE_PFA));
2874 else if (sd->sd_state <= G_RAID_SUBDISK_S_REBUILD)
2875 SET16(gmeta, pdr->entry[j].PD_State,
2876 GET16(gmeta, pdr->entry[j].PD_State) |
2879 SET16(gmeta, pdr->entry[j].PD_State,
2880 GET16(gmeta, pdr->entry[j].PD_State) |
2885 /* Mark spare and failed disks as such. */
2886 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2887 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2888 i = ddf_meta_find_pd(gmeta, NULL,
2889 GET32(&pd->pd_meta, pdd->PD_Reference));
2892 if (disk->d_state == G_RAID_DISK_S_FAILED) {
2893 SET16(gmeta, pdr->entry[i].PD_State,
2894 GET16(gmeta, pdr->entry[i].PD_State) |
2895 (DDF_PDE_FAILED | DDF_PDE_PFA));
2897 if (disk->d_state != G_RAID_DISK_S_SPARE)
2899 sa = ddf_meta_find_sa(&pd->pd_meta, 0);
2901 (GET8D(&pd->pd_meta, sa->Spare_Type) &
2902 DDF_SAR_TYPE_DEDICATED) == 0) {
2903 SET16(gmeta, pdr->entry[i].PD_Type,
2904 GET16(gmeta, pdr->entry[i].PD_Type) |
2905 DDF_PDE_GLOBAL_SPARE);
2907 SET16(gmeta, pdr->entry[i].PD_Type,
2908 GET16(gmeta, pdr->entry[i].PD_Type) |
2909 DDF_PDE_CONFIG_SPARE);
2911 SET16(gmeta, pdr->entry[i].PD_State,
2912 GET16(gmeta, pdr->entry[i].PD_State) |
2916 /* Remove disks without "participating" flag (unused). */
2917 for (i = 0, j = -1; i < GET16(gmeta, pdr->Populated_PDEs); i++) {
2918 if (isff(gmeta->pdr->entry[i].PD_GUID, 24))
2920 if ((GET16(gmeta, pdr->entry[i].PD_Type) &
2921 (DDF_PDE_PARTICIPATING |
2922 DDF_PDE_GLOBAL_SPARE | DDF_PDE_CONFIG_SPARE)) != 0 ||
2923 g_raid_md_ddf_get_disk(sc,
2924 NULL, GET32(gmeta, pdr->entry[i].PD_Reference)) != NULL)
2927 memset(&gmeta->pdr->entry[i], 0xff,
2928 sizeof(struct ddf_pd_entry));
2930 SET16(gmeta, pdr->Populated_PDEs, j + 1);
2932 /* Update per-disk metadata and write them. */
2933 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2934 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2935 if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
2936 disk->d_state != G_RAID_DISK_S_SPARE)
2939 memcpy(pd->pd_meta.pdr, gmeta->pdr,
2940 GET32(&pd->pd_meta, hdr->pdr_length) *
2941 pd->pd_meta.sectorsize);
2943 SET16(&pd->pd_meta, vdr->Populated_VDEs, 0);
2944 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2945 if (vol->v_stopping)
2947 pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2948 i = ddf_meta_find_vd(&pd->pd_meta,
2949 pv->pv_meta.vde->VD_GUID);
2951 i = ddf_meta_find_vd(&pd->pd_meta, NULL);
2953 memcpy(&pd->pd_meta.vdr->entry[i],
2955 sizeof(struct ddf_vd_entry));
2958 if (mdi->mdio_starting == 0) {
2959 /* Remove all VDCs to restore needed later. */
2960 j = GETCRNUM(&pd->pd_meta);
2961 for (i = 0; i < j; i++) {
2962 vdc = GETVDCPTR(&pd->pd_meta, i);
2963 if (GET32D(&pd->pd_meta, vdc->Signature) !=
2966 SET32D(&pd->pd_meta, vdc->Signature, 0xffffffff);
2969 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2970 vol = sd->sd_volume;
2971 if (vol->v_stopping)
2973 pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2974 vmeta = &pv->pv_meta;
2975 vdc = ddf_meta_find_vdc(&pd->pd_meta,
2976 vmeta->vde->VD_GUID);
2978 vdc = ddf_meta_find_vdc(&pd->pd_meta, NULL);
2980 bvd = sd->sd_pos / GET16(vmeta,
2981 vdc->Primary_Element_Count);
2982 memcpy(vdc, vmeta->bvdc[bvd],
2984 hdr->Configuration_Record_Length) *
2985 pd->pd_meta.sectorsize);
2988 G_RAID_DEBUG(1, "Writing DDF metadata to %s",
2989 g_raid_get_diskname(disk));
2990 g_raid_md_ddf_print(&pd->pd_meta);
2991 ddf_meta_write(disk->d_consumer, &pd->pd_meta);
2997 g_raid_md_fail_disk_ddf(struct g_raid_md_object *md,
2998 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
3000 struct g_raid_softc *sc;
3001 struct g_raid_md_ddf_perdisk *pd;
3002 struct g_raid_subdisk *sd;
3006 pd = (struct g_raid_md_ddf_perdisk *)tdisk->d_md_data;
3008 /* We can't fail disk that is not a part of array now. */
3009 if (tdisk->d_state != G_RAID_DISK_S_ACTIVE)
3013 * Mark disk as failed in metadata and try to write that metadata
3014 * to the disk itself to prevent it's later resurrection as STALE.
3016 G_RAID_DEBUG(1, "Writing DDF metadata to %s",
3017 g_raid_get_diskname(tdisk));
3018 i = ddf_meta_find_pd(&pd->pd_meta, NULL, GET32(&pd->pd_meta, pdd->PD_Reference));
3019 SET16(&pd->pd_meta, pdr->entry[i].PD_State, DDF_PDE_FAILED | DDF_PDE_PFA);
3020 if (tdisk->d_consumer != NULL)
3021 ddf_meta_write(tdisk->d_consumer, &pd->pd_meta);
3023 /* Change states. */
3024 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
3025 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
3026 g_raid_change_subdisk_state(sd,
3027 G_RAID_SUBDISK_S_FAILED);
3028 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
3029 G_RAID_EVENT_SUBDISK);
3032 /* Write updated metadata to remaining disks. */
3033 g_raid_md_write_ddf(md, NULL, NULL, tdisk);
3035 g_raid_md_ddf_refill(sc);
3040 g_raid_md_free_disk_ddf(struct g_raid_md_object *md,
3041 struct g_raid_disk *disk)
3043 struct g_raid_md_ddf_perdisk *pd;
3045 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
3046 ddf_meta_free(&pd->pd_meta);
3048 disk->d_md_data = NULL;
3053 g_raid_md_free_volume_ddf(struct g_raid_md_object *md,
3054 struct g_raid_volume *vol)
3056 struct g_raid_md_ddf_object *mdi;
3057 struct g_raid_md_ddf_pervolume *pv;
3059 mdi = (struct g_raid_md_ddf_object *)md;
3060 pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
3061 ddf_vol_meta_free(&pv->pv_meta);
3062 if (!pv->pv_started) {
3064 mdi->mdio_starting--;
3065 callout_stop(&pv->pv_start_co);
3068 vol->v_md_data = NULL;
3073 g_raid_md_free_ddf(struct g_raid_md_object *md)
3075 struct g_raid_md_ddf_object *mdi;
3077 mdi = (struct g_raid_md_ddf_object *)md;
3078 if (!mdi->mdio_started) {
3079 mdi->mdio_started = 0;
3080 callout_stop(&mdi->mdio_start_co);
3081 G_RAID_DEBUG1(1, md->mdo_softc,
3082 "root_mount_rel %p", mdi->mdio_rootmount);
3083 root_mount_rel(mdi->mdio_rootmount);
3084 mdi->mdio_rootmount = NULL;
3086 ddf_meta_free(&mdi->mdio_meta);
3090 G_RAID_MD_DECLARE(ddf, "DDF");