md_ddf.c revision bfb53c205cc5070309c8d61379cb48a0522ecade
1/*-
2 * Copyright (c) 2012 Alexander Motin <mav@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD$");
29
30#include <sys/param.h>
31#include <sys/bio.h>
32#include <sys/endian.h>
33#include <sys/kernel.h>
34#include <sys/kobj.h>
35#include <sys/limits.h>
36#include <sys/lock.h>
37#include <sys/malloc.h>
38#include <sys/mutex.h>
39#include <sys/systm.h>
40#include <sys/time.h>
41#include <sys/clock.h>
42#include <geom/geom.h>
43#include "geom/raid/g_raid.h"
44#include "geom/raid/md_ddf.h"
45#include "g_raid_md_if.h"
46
47static MALLOC_DEFINE(M_MD_DDF, "md_ddf_data", "GEOM_RAID DDF metadata");
48
49#define	DDF_MAX_DISKS_HARD	128
50
51#define	DDF_MAX_DISKS	16
52#define	DDF_MAX_VDISKS	7
53#define	DDF_MAX_PARTITIONS	1
54
55#define DECADE (3600*24*(365*10+2))	/* 10 years in seconds. */
56
57struct ddf_meta {
58	u_int	sectorsize;
59	u_int	bigendian;
60	struct ddf_header *hdr;
61	struct ddf_cd_record *cdr;
62	struct ddf_pd_record *pdr;
63	struct ddf_vd_record *vdr;
64	void *cr;
65	struct ddf_pdd_record *pdd;
66	struct ddf_bbm_log *bbm;
67};
68
69struct ddf_vol_meta {
70	u_int	sectorsize;
71	u_int	bigendian;
72	struct ddf_header *hdr;
73	struct ddf_cd_record *cdr;
74	struct ddf_vd_entry *vde;
75	struct ddf_vdc_record *vdc;
76	struct ddf_vdc_record *bvdc[DDF_MAX_DISKS_HARD];
77};
78
79struct g_raid_md_ddf_perdisk {
80	struct ddf_meta	 pd_meta;
81};
82
83struct g_raid_md_ddf_pervolume {
84	struct ddf_vol_meta		 pv_meta;
85	int				 pv_started;
86	struct callout			 pv_start_co;	/* STARTING state timer. */
87};
88
89struct g_raid_md_ddf_object {
90	struct g_raid_md_object	 mdio_base;
91	u_int			 mdio_bigendian;
92	struct ddf_meta		 mdio_meta;
93	int			 mdio_starting;
94	struct callout		 mdio_start_co;	/* STARTING state timer. */
95	int			 mdio_started;
96	struct root_hold_token	*mdio_rootmount; /* Root mount delay token. */
97};
98
99static g_raid_md_create_req_t g_raid_md_create_req_ddf;
100static g_raid_md_taste_t g_raid_md_taste_ddf;
101static g_raid_md_event_t g_raid_md_event_ddf;
102static g_raid_md_volume_event_t g_raid_md_volume_event_ddf;
103static g_raid_md_ctl_t g_raid_md_ctl_ddf;
104static g_raid_md_write_t g_raid_md_write_ddf;
105static g_raid_md_fail_disk_t g_raid_md_fail_disk_ddf;
106static g_raid_md_free_disk_t g_raid_md_free_disk_ddf;
107static g_raid_md_free_volume_t g_raid_md_free_volume_ddf;
108static g_raid_md_free_t g_raid_md_free_ddf;
109
110static kobj_method_t g_raid_md_ddf_methods[] = {
111	KOBJMETHOD(g_raid_md_create_req,	g_raid_md_create_req_ddf),
112	KOBJMETHOD(g_raid_md_taste,	g_raid_md_taste_ddf),
113	KOBJMETHOD(g_raid_md_event,	g_raid_md_event_ddf),
114	KOBJMETHOD(g_raid_md_volume_event,	g_raid_md_volume_event_ddf),
115	KOBJMETHOD(g_raid_md_ctl,	g_raid_md_ctl_ddf),
116	KOBJMETHOD(g_raid_md_write,	g_raid_md_write_ddf),
117	KOBJMETHOD(g_raid_md_fail_disk,	g_raid_md_fail_disk_ddf),
118	KOBJMETHOD(g_raid_md_free_disk,	g_raid_md_free_disk_ddf),
119	KOBJMETHOD(g_raid_md_free_volume,	g_raid_md_free_volume_ddf),
120	KOBJMETHOD(g_raid_md_free,	g_raid_md_free_ddf),
121	{ 0, 0 }
122};
123
124static struct g_raid_md_class g_raid_md_ddf_class = {
125	"DDF",
126	g_raid_md_ddf_methods,
127	sizeof(struct g_raid_md_ddf_object),
128	.mdc_enable = 1,
129	.mdc_priority = 100
130};
131
132#define GET8(m, f)	((m)->f)
133#define GET16(m, f)	((m)->bigendian ? be16dec(&(m)->f) : le16dec(&(m)->f))
134#define GET32(m, f)	((m)->bigendian ? be32dec(&(m)->f) : le32dec(&(m)->f))
135#define GET64(m, f)	((m)->bigendian ? be64dec(&(m)->f) : le64dec(&(m)->f))
136#define GET8D(m, f)	(f)
137#define GET16D(m, f)	((m)->bigendian ? be16dec(&f) : le16dec(&f))
138#define GET32D(m, f)	((m)->bigendian ? be32dec(&f) : le32dec(&f))
139#define GET64D(m, f)	((m)->bigendian ? be64dec(&f) : le64dec(&f))
140#define GET8P(m, f)	(*(f))
141#define GET16P(m, f)	((m)->bigendian ? be16dec(f) : le16dec(f))
142#define GET32P(m, f)	((m)->bigendian ? be32dec(f) : le32dec(f))
143#define GET64P(m, f)	((m)->bigendian ? be64dec(f) : le64dec(f))
144
145#define SET8P(m, f, v)							\
146	(*(f) = (v))
147#define SET16P(m, f, v)							\
148	do {								\
149		if ((m)->bigendian)					\
150			be16enc((f), (v));				\
151		else							\
152			le16enc((f), (v));				\
153	} while (0)
154#define SET32P(m, f, v)							\
155	do {								\
156		if ((m)->bigendian)					\
157			be32enc((f), (v));				\
158		else							\
159			le32enc((f), (v));				\
160	} while (0)
161#define SET64P(m, f, v)							\
162	do {								\
163		if ((m)->bigendian)					\
164			be64enc((f), (v));				\
165		else							\
166			le64enc((f), (v));				\
167	} while (0)
168#define SET8(m, f, v)	SET8P((m), &((m)->f), (v))
169#define SET16(m, f, v)	SET16P((m), &((m)->f), (v))
170#define SET32(m, f, v)	SET32P((m), &((m)->f), (v))
171#define SET64(m, f, v)	SET64P((m), &((m)->f), (v))
172#define SET8D(m, f, v)	SET8P((m), &(f), (v))
173#define SET16D(m, f, v)	SET16P((m), &(f), (v))
174#define SET32D(m, f, v)	SET32P((m), &(f), (v))
175#define SET64D(m, f, v)	SET64P((m), &(f), (v))
176
177#define GETCRNUM(m)	(GET32((m), hdr->cr_length) /			\
178	GET16((m), hdr->Configuration_Record_Length))
179
180#define GETVDCPTR(m, n)	((struct ddf_vdc_record *)((uint8_t *)(m)->cr +	\
181	(n) * GET16((m), hdr->Configuration_Record_Length) *		\
182	(m)->sectorsize))
183
184#define GETSAPTR(m, n)	((struct ddf_sa_record *)((uint8_t *)(m)->cr +	\
185	(n) * GET16((m), hdr->Configuration_Record_Length) *		\
186	(m)->sectorsize))
187
188static int
189isff(uint8_t *buf, int size)
190{
191	int i;
192
193	for (i = 0; i < size; i++)
194		if (buf[i] != 0xff)
195			return (0);
196	return (1);
197}
198
199static void
200print_guid(uint8_t *buf)
201{
202	int i, ascii;
203
204	ascii = 1;
205	for (i = 0; i < 24; i++) {
206		if (buf[i] != 0 && (buf[i] < ' ' || buf[i] > 127)) {
207			ascii = 0;
208			break;
209		}
210	}
211	if (ascii) {
212		printf("'%.24s'", buf);
213	} else {
214		for (i = 0; i < 24; i++)
215			printf("%02x", buf[i]);
216	}
217}
218
219static void
220g_raid_md_ddf_print(struct ddf_meta *meta)
221{
222	struct ddf_vdc_record *vdc;
223	struct ddf_vuc_record *vuc;
224	struct ddf_sa_record *sa;
225	uint64_t *val2;
226	uint32_t val;
227	int i, j, k, num, num2;
228
229	if (g_raid_debug < 1)
230		return;
231
232	printf("********* DDF Metadata *********\n");
233	printf("**** Header ****\n");
234	printf("DDF_Header_GUID      ");
235	print_guid(meta->hdr->DDF_Header_GUID);
236	printf("\n");
237	printf("DDF_rev              %8.8s\n", (char *)&meta->hdr->DDF_rev[0]);
238	printf("Sequence_Number      0x%08x\n", GET32(meta, hdr->Sequence_Number));
239	printf("TimeStamp            0x%08x\n", GET32(meta, hdr->TimeStamp));
240	printf("Open_Flag            0x%02x\n", GET16(meta, hdr->Open_Flag));
241	printf("Foreign_Flag         0x%02x\n", GET16(meta, hdr->Foreign_Flag));
242	printf("Diskgrouping         0x%02x\n", GET16(meta, hdr->Diskgrouping));
243	printf("Primary_Header_LBA   %ju\n", GET64(meta, hdr->Primary_Header_LBA));
244	printf("Secondary_Header_LBA %ju\n", GET64(meta, hdr->Secondary_Header_LBA));
245	printf("WorkSpace_Length     %u\n", GET32(meta, hdr->WorkSpace_Length));
246	printf("WorkSpace_LBA        %ju\n", GET64(meta, hdr->WorkSpace_LBA));
247	printf("Max_PD_Entries       %u\n", GET16(meta, hdr->Max_PD_Entries));
248	printf("Max_VD_Entries       %u\n", GET16(meta, hdr->Max_VD_Entries));
249	printf("Max_Partitions       %u\n", GET16(meta, hdr->Max_Partitions));
250	printf("Configuration_Record_Length %u\n", GET16(meta, hdr->Configuration_Record_Length));
251	printf("Max_Primary_Element_Entries %u\n", GET16(meta, hdr->Max_Primary_Element_Entries));
252	printf("Controller Data      %u:%u\n", GET32(meta, hdr->cd_section), GET32(meta, hdr->cd_length));
253	printf("Physical Disk        %u:%u\n", GET32(meta, hdr->pdr_section), GET32(meta, hdr->pdr_length));
254	printf("Virtual Disk         %u:%u\n", GET32(meta, hdr->vdr_section), GET32(meta, hdr->vdr_length));
255	printf("Configuration Recs   %u:%u\n", GET32(meta, hdr->cr_section), GET32(meta, hdr->cr_length));
256	printf("Physical Disk Recs   %u:%u\n", GET32(meta, hdr->pdd_section), GET32(meta, hdr->pdd_length));
257	printf("BBM Log              %u:%u\n", GET32(meta, hdr->bbmlog_section), GET32(meta, hdr->bbmlog_length));
258	printf("Diagnostic Space     %u:%u\n", GET32(meta, hdr->Diagnostic_Space), GET32(meta, hdr->Diagnostic_Space_Length));
259	printf("Vendor_Specific_Logs %u:%u\n", GET32(meta, hdr->Vendor_Specific_Logs), GET32(meta, hdr->Vendor_Specific_Logs_Length));
260	printf("**** Controler Data ****\n");
261	printf("Controller_GUID      ");
262	print_guid(meta->cdr->Controller_GUID);
263	printf("\n");
264	printf("Controller_Type      0x%04x%04x 0x%04x%04x\n",
265	    GET16(meta, cdr->Controller_Type.Vendor_ID),
266	    GET16(meta, cdr->Controller_Type.Device_ID),
267	    GET16(meta, cdr->Controller_Type.SubVendor_ID),
268	    GET16(meta, cdr->Controller_Type.SubDevice_ID));
269	printf("Product_ID           '%.16s'\n", (char *)&meta->cdr->Product_ID[0]);
270	printf("**** Physical Disk Records ****\n");
271	printf("Populated_PDEs       %u\n", GET16(meta, pdr->Populated_PDEs));
272	printf("Max_PDE_Supported    %u\n", GET16(meta, pdr->Max_PDE_Supported));
273	for (j = 0; j < GET16(meta, pdr->Populated_PDEs); j++) {
274		if (isff(meta->pdr->entry[j].PD_GUID, 24))
275			continue;
276		if (GET32(meta, pdr->entry[j].PD_Reference) == 0xffffffff)
277			continue;
278		printf("PD_GUID              ");
279		print_guid(meta->pdr->entry[j].PD_GUID);
280		printf("\n");
281		printf("PD_Reference         0x%08x\n",
282		    GET32(meta, pdr->entry[j].PD_Reference));
283		printf("PD_Type              0x%04x\n",
284		    GET16(meta, pdr->entry[j].PD_Type));
285		printf("PD_State             0x%04x\n",
286		    GET16(meta, pdr->entry[j].PD_State));
287		printf("Configured_Size      %ju\n",
288		    GET64(meta, pdr->entry[j].Configured_Size));
289		printf("Block_Size           %u\n",
290		    GET16(meta, pdr->entry[j].Block_Size));
291	}
292	printf("**** Virtual Disk Records ****\n");
293	printf("Populated_VDEs       %u\n", GET16(meta, vdr->Populated_VDEs));
294	printf("Max_VDE_Supported    %u\n", GET16(meta, vdr->Max_VDE_Supported));
295	for (j = 0; j < GET16(meta, vdr->Populated_VDEs); j++) {
296		if (isff(meta->vdr->entry[j].VD_GUID, 24))
297			continue;
298		printf("VD_GUID              ");
299		print_guid(meta->vdr->entry[j].VD_GUID);
300		printf("\n");
301		printf("VD_Number            0x%04x\n",
302		    GET16(meta, vdr->entry[j].VD_Number));
303		printf("VD_Type              0x%04x\n",
304		    GET16(meta, vdr->entry[j].VD_Type));
305		printf("VD_State             0x%02x\n",
306		    GET8(meta, vdr->entry[j].VD_State));
307		printf("Init_State           0x%02x\n",
308		    GET8(meta, vdr->entry[j].Init_State));
309		printf("Drive_Failures_Remaining %u\n",
310		    GET8(meta, vdr->entry[j].Drive_Failures_Remaining));
311		printf("VD_Name              '%.16s'\n",
312		    (char *)&meta->vdr->entry[j].VD_Name);
313	}
314	printf("**** Configuration Records ****\n");
315	num = GETCRNUM(meta);
316	for (j = 0; j < num; j++) {
317		vdc = GETVDCPTR(meta, j);
318		val = GET32D(meta, vdc->Signature);
319		switch (val) {
320		case DDF_VDCR_SIGNATURE:
321			printf("** Virtual Disk Configuration **\n");
322			printf("VD_GUID              ");
323			print_guid(vdc->VD_GUID);
324			printf("\n");
325			printf("Timestamp            0x%08x\n",
326			    GET32D(meta, vdc->Timestamp));
327			printf("Sequence_Number      0x%08x\n",
328			    GET32D(meta, vdc->Sequence_Number));
329			printf("Primary_Element_Count %u\n",
330			    GET16D(meta, vdc->Primary_Element_Count));
331			printf("Stripe_Size          %u\n",
332			    GET8D(meta, vdc->Stripe_Size));
333			printf("Primary_RAID_Level   0x%02x\n",
334			    GET8D(meta, vdc->Primary_RAID_Level));
335			printf("RLQ                  0x%02x\n",
336			    GET8D(meta, vdc->RLQ));
337			printf("Secondary_Element_Count %u\n",
338			    GET8D(meta, vdc->Secondary_Element_Count));
339			printf("Secondary_Element_Seq %u\n",
340			    GET8D(meta, vdc->Secondary_Element_Seq));
341			printf("Secondary_RAID_Level 0x%02x\n",
342			    GET8D(meta, vdc->Secondary_RAID_Level));
343			printf("Block_Count          %ju\n",
344			    GET64D(meta, vdc->Block_Count));
345			printf("VD_Size              %ju\n",
346			    GET64D(meta, vdc->VD_Size));
347			printf("Block_Size           %u\n",
348			    GET16D(meta, vdc->Block_Size));
349			printf("Rotate_Parity_count  %u\n",
350			    GET8D(meta, vdc->Rotate_Parity_count));
351			printf("Associated_Spare_Disks");
352			for (i = 0; i < 8; i++) {
353				if (GET32D(meta, vdc->Associated_Spares[i]) != 0xffffffff)
354					printf(" 0x%08x", GET32D(meta, vdc->Associated_Spares[i]));
355			}
356			printf("\n");
357			printf("Cache_Flags          %016jx\n",
358			    GET64D(meta, vdc->Cache_Flags));
359			printf("BG_Rate              %u\n",
360			    GET8D(meta, vdc->BG_Rate));
361			printf("MDF_Parity_Disks     %u\n",
362			    GET8D(meta, vdc->MDF_Parity_Disks));
363			printf("MDF_Parity_Generator_Polynomial 0x%04x\n",
364			    GET16D(meta, vdc->MDF_Parity_Generator_Polynomial));
365			printf("MDF_Constant_Generation_Method 0x%02x\n",
366			    GET8D(meta, vdc->MDF_Constant_Generation_Method));
367			printf("Physical_Disks      ");
368			num2 = GET16D(meta, vdc->Primary_Element_Count);
369			val2 = (uint64_t *)&(vdc->Physical_Disk_Sequence[GET16(meta, hdr->Max_Primary_Element_Entries)]);
370			for (i = 0; i < num2; i++)
371				printf(" 0x%08x @ %ju",
372				    GET32D(meta, vdc->Physical_Disk_Sequence[i]),
373				    GET64P(meta, val2 + i));
374			printf("\n");
375			break;
376		case DDF_VUCR_SIGNATURE:
377			printf("** Vendor Unique Configuration **\n");
378			vuc = (struct ddf_vuc_record *)vdc;
379			printf("VD_GUID              ");
380			print_guid(vuc->VD_GUID);
381			printf("\n");
382			break;
383		case DDF_SA_SIGNATURE:
384			printf("** Spare Assignment Configuration **\n");
385			sa = (struct ddf_sa_record *)vdc;
386			printf("Timestamp            0x%08x\n",
387			    GET32D(meta, sa->Timestamp));
388			printf("Spare_Type           0x%02x\n",
389			    GET8D(meta, sa->Spare_Type));
390			printf("Populated_SAEs       %u\n",
391			    GET16D(meta, sa->Populated_SAEs));
392			printf("MAX_SAE_Supported    %u\n",
393			    GET16D(meta, sa->MAX_SAE_Supported));
394			for (i = 0; i < GET16D(meta, sa->Populated_SAEs); i++) {
395				if (isff(sa->entry[i].VD_GUID, 24))
396					continue;
397				printf("VD_GUID             ");
398				for (k = 0; k < 24; k++)
399					printf("%02x", sa->entry[i].VD_GUID[k]);
400				printf("\n");
401				printf("Secondary_Element   %u\n",
402				    GET16D(meta, sa->entry[i].Secondary_Element));
403			}
404			break;
405		case 0x00000000:
406		case 0xFFFFFFFF:
407			break;
408		default:
409			printf("Unknown configuration signature %08x\n", val);
410			break;
411		}
412	}
413	printf("**** Physical Disk Data ****\n");
414	printf("PD_GUID              ");
415	print_guid(meta->pdd->PD_GUID);
416	printf("\n");
417	printf("PD_Reference         0x%08x\n",
418	    GET32(meta, pdd->PD_Reference));
419	printf("Forced_Ref_Flag      0x%02x\n",
420	    GET8(meta, pdd->Forced_Ref_Flag));
421	printf("Forced_PD_GUID_Flag  0x%02x\n",
422	    GET8(meta, pdd->Forced_PD_GUID_Flag));
423}
424
425static int
426ddf_meta_find_pd(struct ddf_meta *meta, uint8_t *GUID, uint32_t PD_Reference)
427{
428	int i;
429
430	for (i = 0; i < GET16(meta, pdr->Populated_PDEs); i++) {
431		if (GUID != NULL) {
432			if (memcmp(meta->pdr->entry[i].PD_GUID, GUID, 24) == 0)
433				return (i);
434		} else if (PD_Reference != 0xffffffff) {
435			if (GET32(meta, pdr->entry[i].PD_Reference) == PD_Reference)
436				return (i);
437		} else
438			if (isff(meta->pdr->entry[i].PD_GUID, 24))
439				return (i);
440	}
441	if (GUID == NULL && PD_Reference == 0xffffffff) {
442		if (i >= GET16(meta, pdr->Max_PDE_Supported))
443			return (-1);
444		SET16(meta, pdr->Populated_PDEs, i + 1);
445		return (i);
446	}
447	return (-1);
448}
449
450static int
451ddf_meta_find_vd(struct ddf_meta *meta, uint8_t *GUID)
452{
453	int i;
454
455	for (i = 0; i < GET16(meta, vdr->Populated_VDEs); i++) {
456		if (GUID != NULL) {
457			if (memcmp(meta->vdr->entry[i].VD_GUID, GUID, 24) == 0)
458				return (i);
459		} else
460			if (isff(meta->vdr->entry[i].VD_GUID, 24))
461				return (i);
462	}
463	if (GUID == NULL) {
464		if (i >= GET16(meta, vdr->Max_VDE_Supported))
465			return (-1);
466		SET16(meta, vdr->Populated_VDEs, i + 1);
467		return (i);
468	}
469	return (-1);
470}
471
472static struct ddf_vdc_record *
473ddf_meta_find_vdc(struct ddf_meta *meta, uint8_t *GUID)
474{
475	struct ddf_vdc_record *vdc;
476	int i, num;
477
478	num = GETCRNUM(meta);
479	for (i = 0; i < num; i++) {
480		vdc = GETVDCPTR(meta, i);
481		if (GUID != NULL) {
482			if (GET32D(meta, vdc->Signature) == DDF_VDCR_SIGNATURE &&
483			    memcmp(vdc->VD_GUID, GUID, 24) == 0)
484				return (vdc);
485		} else
486			if (GET32D(meta, vdc->Signature) == 0xffffffff ||
487			    GET32D(meta, vdc->Signature) == 0)
488				return (vdc);
489	}
490	return (NULL);
491}
492
493static int
494ddf_meta_count_vdc(struct ddf_meta *meta, uint8_t *GUID)
495{
496	struct ddf_vdc_record *vdc;
497	int i, num, cnt;
498
499	cnt = 0;
500	num = GETCRNUM(meta);
501	for (i = 0; i < num; i++) {
502		vdc = GETVDCPTR(meta, i);
503		if (GET32D(meta, vdc->Signature) != DDF_VDCR_SIGNATURE)
504			continue;
505		if (GUID == NULL || memcmp(vdc->VD_GUID, GUID, 24) == 0)
506			cnt++;
507	}
508	return (cnt);
509}
510
511static int
512ddf_meta_find_disk(struct ddf_vol_meta *vmeta, uint32_t PD_Reference,
513    int *bvdp, int *posp)
514{
515	int i, bvd, pos;
516
517	i = 0;
518	for (bvd = 0; bvd < GET16(vmeta, vdc->Secondary_Element_Count); bvd++) {
519		if (vmeta->bvdc[bvd] == NULL) {
520			i += GET16(vmeta, vdc->Primary_Element_Count); // XXX
521			continue;
522		}
523		for (pos = 0; pos < GET16(vmeta, bvdc[bvd]->Primary_Element_Count);
524		    pos++, i++) {
525			if (GET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos]) ==
526			    PD_Reference) {
527				if (bvdp != NULL)
528					*bvdp = bvd;
529				if (posp != NULL)
530					*posp = pos;
531				return (i);
532			}
533		}
534	}
535	return (-1);
536}
537
538static struct ddf_sa_record *
539ddf_meta_find_sa(struct ddf_meta *meta, int create)
540{
541	struct ddf_sa_record *sa;
542	int i, num;
543
544	num = GETCRNUM(meta);
545	for (i = 0; i < num; i++) {
546		sa = GETSAPTR(meta, i);
547		if (GET32D(meta, sa->Signature) == DDF_SA_SIGNATURE)
548			return (sa);
549	}
550	if (create) {
551		for (i = 0; i < num; i++) {
552			sa = GETSAPTR(meta, i);
553			if (GET32D(meta, sa->Signature) == 0xffffffff ||
554			    GET32D(meta, sa->Signature) == 0)
555				return (sa);
556		}
557	}
558	return (NULL);
559}
560
561static void
562ddf_meta_create(struct g_raid_disk *disk, struct ddf_meta *sample)
563{
564	struct timespec ts;
565	struct clocktime ct;
566	struct g_raid_md_ddf_perdisk *pd;
567	struct g_raid_md_ddf_object *mdi;
568	struct ddf_meta *meta;
569	struct ddf_pd_entry *pde;
570	off_t anchorlba;
571	u_int ss, pos, size;
572	int len, error;
573	char serial_buffer[24];
574
575	if (sample->hdr == NULL)
576		sample = NULL;
577
578	mdi = (struct g_raid_md_ddf_object *)disk->d_softc->sc_md;
579	pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
580	meta = &pd->pd_meta;
581	ss = disk->d_consumer->provider->sectorsize;
582	anchorlba = disk->d_consumer->provider->mediasize / ss - 1;
583
584	meta->sectorsize = ss;
585	meta->bigendian = sample ? sample->bigendian : mdi->mdio_bigendian;
586	getnanotime(&ts);
587	clock_ts_to_ct(&ts, &ct);
588
589	/* Header */
590	meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
591	memset(meta->hdr, 0xff, ss);
592	if (sample) {
593		memcpy(meta->hdr, sample->hdr, sizeof(struct ddf_header));
594		if (ss != sample->sectorsize) {
595			SET32(meta, hdr->WorkSpace_Length,
596			    (GET32(sample, hdr->WorkSpace_Length) *
597			    sample->sectorsize + ss - 1) / ss);
598			SET16(meta, hdr->Configuration_Record_Length,
599			    (GET16(sample, hdr->Configuration_Record_Length) *
600			    sample->sectorsize + ss - 1) / ss);
601			SET32(meta, hdr->cd_length,
602			    (GET32(sample, hdr->cd_length) *
603			    sample->sectorsize + ss - 1) / ss);
604			SET32(meta, hdr->pdr_length,
605			    (GET32(sample, hdr->pdr_length) *
606			    sample->sectorsize + ss - 1) / ss);
607			SET32(meta, hdr->vdr_length,
608			    (GET32(sample, hdr->vdr_length) *
609			    sample->sectorsize + ss - 1) / ss);
610			SET32(meta, hdr->cr_length,
611			    (GET32(sample, hdr->cr_length) *
612			    sample->sectorsize + ss - 1) / ss);
613			SET32(meta, hdr->pdd_length,
614			    (GET32(sample, hdr->pdd_length) *
615			    sample->sectorsize + ss - 1) / ss);
616			SET32(meta, hdr->bbmlog_length,
617			    (GET32(sample, hdr->bbmlog_length) *
618			    sample->sectorsize + ss - 1) / ss);
619			SET32(meta, hdr->Diagnostic_Space,
620			    (GET32(sample, hdr->bbmlog_length) *
621			    sample->sectorsize + ss - 1) / ss);
622			SET32(meta, hdr->Vendor_Specific_Logs,
623			    (GET32(sample, hdr->bbmlog_length) *
624			    sample->sectorsize + ss - 1) / ss);
625		}
626	} else {
627		SET32(meta, hdr->Signature, DDF_HEADER_SIGNATURE);
628		snprintf(meta->hdr->DDF_Header_GUID, 25, "FreeBSD %08x%08x",
629		    (u_int)(ts.tv_sec - DECADE), arc4random());
630		memcpy(meta->hdr->DDF_rev, "02.00.00", 8);
631		SET32(meta, hdr->TimeStamp, (ts.tv_sec - DECADE));
632		SET32(meta, hdr->WorkSpace_Length, 16 * 1024 * 1024 / ss);
633		SET16(meta, hdr->Max_PD_Entries, DDF_MAX_DISKS - 1);
634		SET16(meta, hdr->Max_VD_Entries, DDF_MAX_VDISKS);
635		SET16(meta, hdr->Max_Partitions, DDF_MAX_PARTITIONS);
636		SET16(meta, hdr->Max_Primary_Element_Entries, DDF_MAX_DISKS);
637		SET16(meta, hdr->Configuration_Record_Length,
638		    (sizeof(struct ddf_vdc_record) +
639		     (4 + 8) * GET16(meta, hdr->Max_Primary_Element_Entries) +
640		     ss - 1) / ss);
641		SET32(meta, hdr->cd_length,
642		    (sizeof(struct ddf_cd_record) + ss - 1) / ss);
643		SET32(meta, hdr->pdr_length,
644		    (sizeof(struct ddf_pd_record) +
645		     sizeof(struct ddf_pd_entry) *
646		     GET16(meta, hdr->Max_PD_Entries) + ss - 1) / ss);
647		SET32(meta, hdr->vdr_length,
648		    (sizeof(struct ddf_vd_record) +
649		     sizeof(struct ddf_vd_entry) *
650		     GET16(meta, hdr->Max_VD_Entries) + ss - 1) / ss);
651		SET32(meta, hdr->cr_length,
652		    GET16(meta, hdr->Configuration_Record_Length) *
653		    (GET16(meta, hdr->Max_Partitions) + 1));
654		SET32(meta, hdr->pdd_length,
655		    (sizeof(struct ddf_pdd_record) + ss - 1) / ss);
656		SET32(meta, hdr->bbmlog_length, 0);
657		SET32(meta, hdr->Diagnostic_Space_Length, 0);
658		SET32(meta, hdr->Vendor_Specific_Logs_Length, 0);
659	}
660	pos = 1;
661	SET32(meta, hdr->cd_section, pos);
662	pos += GET32(meta, hdr->cd_length);
663	SET32(meta, hdr->pdr_section, pos);
664	pos += GET32(meta, hdr->pdr_length);
665	SET32(meta, hdr->vdr_section, pos);
666	pos += GET32(meta, hdr->vdr_length);
667	SET32(meta, hdr->cr_section, pos);
668	pos += GET32(meta, hdr->cr_length);
669	SET32(meta, hdr->pdd_section, pos);
670	pos += GET32(meta, hdr->pdd_length);
671	SET32(meta, hdr->bbmlog_section,
672	    GET32(meta, hdr->bbmlog_length) != 0 ? pos : 0xffffffff);
673	pos += GET32(meta, hdr->bbmlog_length);
674	SET32(meta, hdr->Diagnostic_Space,
675	    GET32(meta, hdr->Diagnostic_Space_Length) != 0 ? pos : 0xffffffff);
676	pos += GET32(meta, hdr->Diagnostic_Space_Length);
677	SET32(meta, hdr->Vendor_Specific_Logs,
678	    GET32(meta, hdr->Vendor_Specific_Logs_Length) != 0 ? pos : 0xffffffff);
679	pos += min(GET32(meta, hdr->Vendor_Specific_Logs_Length), 1);
680	SET64(meta, hdr->Primary_Header_LBA,
681	    anchorlba - pos);
682	SET64(meta, hdr->Secondary_Header_LBA,
683	    0xffffffffffffffffULL);
684	SET64(meta, hdr->WorkSpace_LBA,
685	    anchorlba + 1 - 32 * 1024 * 1024 / ss);
686
687	/* Controller Data */
688	size = GET32(meta, hdr->cd_length) * ss;
689	meta->cdr = malloc(size, M_MD_DDF, M_WAITOK);
690	memset(meta->cdr, 0xff, size);
691	SET32(meta, cdr->Signature, DDF_CONTROLLER_DATA_SIGNATURE);
692	memcpy(meta->cdr->Controller_GUID, "FreeBSD GEOM RAID SERIAL", 24);
693	memcpy(meta->cdr->Product_ID, "FreeBSD GEOMRAID", 16);
694
695	/* Physical Drive Records. */
696	size = GET32(meta, hdr->pdr_length) * ss;
697	meta->pdr = malloc(size, M_MD_DDF, M_WAITOK);
698	memset(meta->pdr, 0xff, size);
699	SET32(meta, pdr->Signature, DDF_PDR_SIGNATURE);
700	SET16(meta, pdr->Populated_PDEs, 1);
701	SET16(meta, pdr->Max_PDE_Supported,
702	    GET16(meta, hdr->Max_PD_Entries));
703
704	pde = &meta->pdr->entry[0];
705	len = sizeof(serial_buffer);
706	error = g_io_getattr("GEOM::ident", disk->d_consumer, &len, serial_buffer);
707	if (error == 0 && (len = strlen (serial_buffer)) >= 6 && len <= 20)
708		snprintf(pde->PD_GUID, 25, "DISK%20s", serial_buffer);
709	else
710		snprintf(pde->PD_GUID, 25, "DISK%04d%02d%02d%08x%04x",
711		    ct.year, ct.mon, ct.day,
712		    arc4random(), arc4random() & 0xffff);
713	SET32D(meta, pde->PD_Reference, arc4random());
714	SET16D(meta, pde->PD_Type, DDF_PDE_GUID_FORCE);
715	SET16D(meta, pde->PD_State, 0);
716	SET64D(meta, pde->Configured_Size,
717	    anchorlba + 1 - 32 * 1024 * 1024 / ss);
718	SET16D(meta, pde->Block_Size, ss);
719
720	/* Virtual Drive Records. */
721	size = GET32(meta, hdr->vdr_length) * ss;
722	meta->vdr = malloc(size, M_MD_DDF, M_WAITOK);
723	memset(meta->vdr, 0xff, size);
724	SET32(meta, vdr->Signature, DDF_VD_RECORD_SIGNATURE);
725	SET32(meta, vdr->Populated_VDEs, 0);
726	SET16(meta, vdr->Max_VDE_Supported,
727	    GET16(meta, hdr->Max_VD_Entries));
728
729	/* Configuration Records. */
730	size = GET32(meta, hdr->cr_length) * ss;
731	meta->cr = malloc(size, M_MD_DDF, M_WAITOK);
732	memset(meta->cr, 0xff, size);
733
734	/* Physical Disk Data. */
735	size = GET32(meta, hdr->pdd_length) * ss;
736	meta->pdd = malloc(size, M_MD_DDF, M_WAITOK);
737	memset(meta->pdd, 0xff, size);
738	SET32(meta, pdd->Signature, DDF_PDD_SIGNATURE);
739	memcpy(meta->pdd->PD_GUID, pde->PD_GUID, 24);
740	SET32(meta, pdd->PD_Reference, GET32D(meta, pde->PD_Reference));
741	SET8(meta, pdd->Forced_Ref_Flag, DDF_PDD_FORCED_REF);
742	SET8(meta, pdd->Forced_PD_GUID_Flag, DDF_PDD_FORCED_GUID);
743
744	/* Bad Block Management Log. */
745	if (GET32(meta, hdr->bbmlog_length) != 0) {
746		size = GET32(meta, hdr->bbmlog_length) * ss;
747		meta->bbm = malloc(size, M_MD_DDF, M_WAITOK);
748		memset(meta->bbm, 0xff, size);
749		SET32(meta, bbm->Signature, DDF_BBML_SIGNATURE);
750		SET32(meta, bbm->Entry_Count, 0);
751		SET32(meta, bbm->Spare_Block_Count, 0);
752	}
753}
754
755static void
756ddf_meta_copy(struct ddf_meta *dst, struct ddf_meta *src)
757{
758	struct ddf_header *hdr;
759	u_int ss;
760
761	hdr = src->hdr;
762	dst->bigendian = src->bigendian;
763	ss = dst->sectorsize = src->sectorsize;
764	dst->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
765	memcpy(dst->hdr, src->hdr, ss);
766	dst->cdr = malloc(GET32(src, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
767	memcpy(dst->cdr, src->cdr, GET32(src, hdr->cd_length) * ss);
768	dst->pdr = malloc(GET32(src, hdr->pdr_length) * ss, M_MD_DDF, M_WAITOK);
769	memcpy(dst->pdr, src->pdr, GET32(src, hdr->pdr_length) * ss);
770	dst->vdr = malloc(GET32(src, hdr->vdr_length) * ss, M_MD_DDF, M_WAITOK);
771	memcpy(dst->vdr, src->vdr, GET32(src, hdr->vdr_length) * ss);
772	dst->cr = malloc(GET32(src, hdr->cr_length) * ss, M_MD_DDF, M_WAITOK);
773	memcpy(dst->cr, src->cr, GET32(src, hdr->cr_length) * ss);
774	dst->pdd = malloc(GET32(src, hdr->pdd_length) * ss, M_MD_DDF, M_WAITOK);
775	memcpy(dst->pdd, src->pdd, GET32(src, hdr->pdd_length) * ss);
776	if (src->bbm != NULL) {
777		dst->bbm = malloc(GET32(src, hdr->bbmlog_length) * ss, M_MD_DDF, M_WAITOK);
778		memcpy(dst->bbm, src->bbm, GET32(src, hdr->bbmlog_length) * ss);
779	}
780}
781
782static void
783ddf_meta_update(struct ddf_meta *meta, struct ddf_meta *src)
784{
785	struct ddf_pd_entry *pde, *spde;
786	int i, j;
787
788	for (i = 0; i < GET16(src, pdr->Populated_PDEs); i++) {
789		spde = &src->pdr->entry[i];
790		if (isff(spde->PD_GUID, 24))
791			continue;
792		j = ddf_meta_find_pd(meta, NULL,
793		    GET32(src, pdr->entry[i].PD_Reference));
794		if (j < 0) {
795			j = ddf_meta_find_pd(meta, NULL, 0xffffffff);
796			pde = &meta->pdr->entry[j];
797			memcpy(pde, spde, sizeof(*pde));
798		} else {
799			pde = &meta->pdr->entry[j];
800			SET16D(meta, pde->PD_State,
801			    GET16D(meta, pde->PD_State) |
802			    GET16D(src, pde->PD_State));
803		}
804	}
805}
806
807static void
808ddf_meta_free(struct ddf_meta *meta)
809{
810
811	if (meta->hdr != NULL) {
812		free(meta->hdr, M_MD_DDF);
813		meta->hdr = NULL;
814	}
815	if (meta->cdr != NULL) {
816		free(meta->cdr, M_MD_DDF);
817		meta->cdr = NULL;
818	}
819	if (meta->pdr != NULL) {
820		free(meta->pdr, M_MD_DDF);
821		meta->pdr = NULL;
822	}
823	if (meta->vdr != NULL) {
824		free(meta->vdr, M_MD_DDF);
825		meta->vdr = NULL;
826	}
827	if (meta->cr != NULL) {
828		free(meta->cr, M_MD_DDF);
829		meta->cr = NULL;
830	}
831	if (meta->pdd != NULL) {
832		free(meta->pdd, M_MD_DDF);
833		meta->pdd = NULL;
834	}
835	if (meta->bbm != NULL) {
836		free(meta->bbm, M_MD_DDF);
837		meta->bbm = NULL;
838	}
839}
840
841static void
842ddf_vol_meta_create(struct ddf_vol_meta *meta, struct ddf_meta *sample)
843{
844	struct timespec ts;
845	struct clocktime ct;
846	struct ddf_header *hdr;
847	u_int ss, size;
848
849	hdr = sample->hdr;
850	meta->bigendian = sample->bigendian;
851	ss = meta->sectorsize = sample->sectorsize;
852	meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
853	memcpy(meta->hdr, sample->hdr, ss);
854	meta->cdr = malloc(GET32(sample, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
855	memcpy(meta->cdr, sample->cdr, GET32(sample, hdr->cd_length) * ss);
856	meta->vde = malloc(sizeof(struct ddf_vd_entry), M_MD_DDF, M_WAITOK);
857	memset(meta->vde, 0xff, sizeof(struct ddf_vd_entry));
858	getnanotime(&ts);
859	clock_ts_to_ct(&ts, &ct);
860	snprintf(meta->vde->VD_GUID, 25, "FreeBSD%04d%02d%02d%08x%01x",
861	    ct.year, ct.mon, ct.day,
862	    arc4random(), arc4random() & 0xf);
863	size = GET16(sample, hdr->Configuration_Record_Length) * ss;
864	meta->vdc = malloc(size, M_MD_DDF, M_WAITOK);
865	memset(meta->vdc, 0xff, size);
866	SET32(meta, vdc->Signature, DDF_VDCR_SIGNATURE);
867	memcpy(meta->vdc->VD_GUID, meta->vde->VD_GUID, 24);
868	SET32(meta, vdc->Sequence_Number, 0);
869}
870
871static void
872ddf_vol_meta_update(struct ddf_vol_meta *dst, struct ddf_meta *src,
873    uint8_t *GUID, int started)
874{
875	struct ddf_header *hdr;
876	struct ddf_vd_entry *vde;
877	struct ddf_vdc_record *vdc;
878	int vnew, bvnew, bvd, size;
879	u_int ss;
880
881	hdr = src->hdr;
882	vde = &src->vdr->entry[ddf_meta_find_vd(src, GUID)];
883	vdc = ddf_meta_find_vdc(src, GUID);
884	bvd = GET8D(src, vdc->Secondary_Element_Seq);
885	size = GET16(src, hdr->Configuration_Record_Length) * src->sectorsize;
886
887	if (dst->vdc == NULL ||
888	    (!started && ((int32_t)(GET32D(src, vdc->Sequence_Number) -
889	    GET32(dst, vdc->Sequence_Number))) > 0))
890		vnew = 1;
891	else
892		vnew = 0;
893
894	if (dst->bvdc[bvd] == NULL ||
895	    (!started && ((int32_t)(GET32D(src, vdc->Sequence_Number) -
896	    GET32(dst, bvdc[bvd]->Sequence_Number))) > 0))
897		bvnew = 1;
898	else
899		bvnew = 0;
900
901	if (vnew) {
902		dst->bigendian = src->bigendian;
903		ss = dst->sectorsize = src->sectorsize;
904		if (dst->hdr != NULL)
905			free(dst->hdr, M_MD_DDF);
906		dst->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
907		memcpy(dst->hdr, src->hdr, ss);
908		if (dst->cdr != NULL)
909			free(dst->cdr, M_MD_DDF);
910		dst->cdr = malloc(GET32(src, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
911		memcpy(dst->cdr, src->cdr, GET32(src, hdr->cd_length) * ss);
912		if (dst->vde != NULL)
913			free(dst->vde, M_MD_DDF);
914		dst->vde = malloc(sizeof(struct ddf_vd_entry), M_MD_DDF, M_WAITOK);
915		memcpy(dst->vde, vde, sizeof(struct ddf_vd_entry));
916		if (dst->vdc != NULL)
917			free(dst->vdc, M_MD_DDF);
918		dst->vdc = malloc(size, M_MD_DDF, M_WAITOK);
919		memcpy(dst->vdc, vdc, size);
920	}
921	if (bvnew) {
922		if (dst->bvdc[bvd] != NULL)
923			free(dst->bvdc[bvd], M_MD_DDF);
924		dst->bvdc[bvd] = malloc(size, M_MD_DDF, M_WAITOK);
925		memcpy(dst->bvdc[bvd], vdc, size);
926	}
927}
928
929static void
930ddf_vol_meta_free(struct ddf_vol_meta *meta)
931{
932	int i;
933
934	if (meta->hdr != NULL) {
935		free(meta->hdr, M_MD_DDF);
936		meta->hdr = NULL;
937	}
938	if (meta->cdr != NULL) {
939		free(meta->cdr, M_MD_DDF);
940		meta->cdr = NULL;
941	}
942	if (meta->vde != NULL) {
943		free(meta->vde, M_MD_DDF);
944		meta->vde = NULL;
945	}
946	if (meta->vdc != NULL) {
947		free(meta->vdc, M_MD_DDF);
948		meta->vdc = NULL;
949	}
950	for (i = 0; i < DDF_MAX_DISKS_HARD; i++) {
951		if (meta->bvdc[i] != NULL) {
952			free(meta->bvdc[i], M_MD_DDF);
953			meta->bvdc[i] = NULL;
954		}
955	}
956}
957
958static int
959ddf_meta_unused_range(struct ddf_meta *meta, off_t *off, off_t *size)
960{
961	struct ddf_vdc_record *vdc;
962	off_t beg[32], end[32], beg1, end1;
963	uint64_t *offp;
964	int i, j, n, num, pos;
965	uint32_t ref;
966
967	*off = 0;
968	*size = 0;
969	ref = GET32(meta, pdd->PD_Reference);
970	pos = ddf_meta_find_pd(meta, NULL, ref);
971	beg[0] = 0;
972	end[0] = GET64(meta, pdr->entry[pos].Configured_Size);
973	n = 1;
974	num = GETCRNUM(meta);
975	for (i = 0; i < num; i++) {
976		vdc = GETVDCPTR(meta, i);
977		if (GET32D(meta, vdc->Signature) != DDF_VDCR_SIGNATURE)
978			continue;
979		for (pos = 0; pos < GET16D(meta, vdc->Primary_Element_Count); pos++)
980			if (GET32D(meta, vdc->Physical_Disk_Sequence[pos]) == ref)
981				break;
982		if (pos == GET16D(meta, vdc->Primary_Element_Count))
983			continue;
984		offp = (uint64_t *)&(vdc->Physical_Disk_Sequence[
985		    GET16(meta, hdr->Max_Primary_Element_Entries)]);
986		beg1 = GET64P(meta, offp + pos);
987		end1 = beg1 + GET64D(meta, vdc->Block_Count);
988		for (j = 0; j < n; j++) {
989			if (beg[j] >= end1 || end[j] <= beg1 )
990				continue;
991			if (beg[j] < beg1 && end[j] > end1) {
992				beg[n] = end1;
993				end[n] = end[j];
994				end[j] = beg1;
995				n++;
996			} else if (beg[j] < beg1)
997				end[j] = beg1;
998			else
999				beg[j] = end1;
1000		}
1001	}
1002	for (j = 0; j < n; j++) {
1003		if (end[j] - beg[j] > *size) {
1004			*off = beg[j];
1005			*size = end[j] - beg[j];
1006		}
1007	}
1008	return ((*size > 0) ? 1 : 0);
1009}
1010
1011static void
1012ddf_meta_get_name(struct ddf_meta *meta, int num, char *buf)
1013{
1014	const char *b;
1015	int i;
1016
1017	b = meta->vdr->entry[num].VD_Name;
1018	for (i = 15; i >= 0; i--)
1019		if (b[i] != 0x20)
1020			break;
1021	memcpy(buf, b, i + 1);
1022	buf[i + 1] = 0;
1023}
1024
1025static void
1026ddf_meta_put_name(struct ddf_vol_meta *meta, char *buf)
1027{
1028	int len;
1029
1030	len = min(strlen(buf), 16);
1031	memset(meta->vde->VD_Name, 0x20, 16);
1032	memcpy(meta->vde->VD_Name, buf, len);
1033}
1034
1035static int
1036ddf_meta_read(struct g_consumer *cp, struct ddf_meta *meta)
1037{
1038	struct g_provider *pp;
1039	struct ddf_header *ahdr, *hdr;
1040	char *abuf, *buf;
1041	off_t plba, slba, lba;
1042	int error, len, i;
1043	u_int ss;
1044	uint32_t val;
1045
1046	ddf_meta_free(meta);
1047	pp = cp->provider;
1048	ss = meta->sectorsize = pp->sectorsize;
1049	/* Read anchor block. */
1050	abuf = g_read_data(cp, pp->mediasize - ss, ss, &error);
1051	if (abuf == NULL) {
1052		G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
1053		    pp->name, error);
1054		return (error);
1055	}
1056	ahdr = (struct ddf_header *)abuf;
1057
1058	/* Check if this is an DDF RAID struct */
1059	if (be32dec(&ahdr->Signature) == DDF_HEADER_SIGNATURE)
1060		meta->bigendian = 1;
1061	else if (le32dec(&ahdr->Signature) == DDF_HEADER_SIGNATURE)
1062		meta->bigendian = 0;
1063	else {
1064		G_RAID_DEBUG(1, "DDF signature check failed on %s", pp->name);
1065		error = EINVAL;
1066		goto done;
1067	}
1068	if (ahdr->Header_Type != DDF_HEADER_ANCHOR) {
1069		G_RAID_DEBUG(1, "DDF header type check failed on %s", pp->name);
1070		error = EINVAL;
1071		goto done;
1072	}
1073	meta->hdr = ahdr;
1074	plba = GET64(meta, hdr->Primary_Header_LBA);
1075	slba = GET64(meta, hdr->Secondary_Header_LBA);
1076	val = GET32(meta, hdr->CRC);
1077	SET32(meta, hdr->CRC, 0xffffffff);
1078	meta->hdr = NULL;
1079	if (crc32(ahdr, ss) != val) {
1080		G_RAID_DEBUG(1, "DDF CRC mismatch on %s", pp->name);
1081		error = EINVAL;
1082		goto done;
1083	}
1084	if ((plba + 6) * ss >= pp->mediasize) {
1085		G_RAID_DEBUG(1, "DDF primary header LBA is wrong on %s", pp->name);
1086		error = EINVAL;
1087		goto done;
1088	}
1089	if (slba != -1 && (slba + 6) * ss >= pp->mediasize) {
1090		G_RAID_DEBUG(1, "DDF secondary header LBA is wrong on %s", pp->name);
1091		error = EINVAL;
1092		goto done;
1093	}
1094	lba = plba;
1095
1096doread:
1097	error = 0;
1098	ddf_meta_free(meta);
1099
1100	/* Read header block. */
1101	buf = g_read_data(cp, lba * ss, ss, &error);
1102	if (buf == NULL) {
1103readerror:
1104		G_RAID_DEBUG(1, "DDF %s metadata read error on %s (error=%d).",
1105		    (lba == plba) ? "primary" : "secondary", pp->name, error);
1106		if (lba == plba && slba != -1) {
1107			lba = slba;
1108			goto doread;
1109		}
1110		G_RAID_DEBUG(1, "DDF metadata read error on %s.", pp->name);
1111		goto done;
1112	}
1113	meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
1114	memcpy(meta->hdr, buf, ss);
1115	g_free(buf);
1116	hdr = meta->hdr;
1117	val = GET32(meta, hdr->CRC);
1118	SET32(meta, hdr->CRC, 0xffffffff);
1119	if (hdr->Signature != ahdr->Signature ||
1120	    crc32(meta->hdr, ss) != val ||
1121	    memcmp(hdr->DDF_Header_GUID, ahdr->DDF_Header_GUID, 24) ||
1122	    GET64(meta, hdr->Primary_Header_LBA) != plba ||
1123	    GET64(meta, hdr->Secondary_Header_LBA) != slba) {
1124hdrerror:
1125		G_RAID_DEBUG(1, "DDF %s metadata check failed on %s",
1126		    (lba == plba) ? "primary" : "secondary", pp->name);
1127		if (lba == plba && slba != -1) {
1128			lba = slba;
1129			goto doread;
1130		}
1131		G_RAID_DEBUG(1, "DDF metadata check failed on %s", pp->name);
1132		error = EINVAL;
1133		goto done;
1134	}
1135	if ((lba == plba && hdr->Header_Type != DDF_HEADER_PRIMARY) ||
1136	    (lba == slba && hdr->Header_Type != DDF_HEADER_SECONDARY))
1137		goto hdrerror;
1138	len = 1;
1139	len = max(len, GET32(meta, hdr->cd_section) + GET32(meta, hdr->cd_length));
1140	len = max(len, GET32(meta, hdr->pdr_section) + GET32(meta, hdr->pdr_length));
1141	len = max(len, GET32(meta, hdr->vdr_section) + GET32(meta, hdr->vdr_length));
1142	len = max(len, GET32(meta, hdr->cr_section) + GET32(meta, hdr->cr_length));
1143	len = max(len, GET32(meta, hdr->pdd_section) + GET32(meta, hdr->pdd_length));
1144	if ((val = GET32(meta, hdr->bbmlog_section)) != 0xffffffff)
1145		len = max(len, val + GET32(meta, hdr->bbmlog_length));
1146	if ((val = GET32(meta, hdr->Diagnostic_Space)) != 0xffffffff)
1147		len = max(len, val + GET32(meta, hdr->Diagnostic_Space_Length));
1148	if ((val = GET32(meta, hdr->Vendor_Specific_Logs)) != 0xffffffff)
1149		len = max(len, val + GET32(meta, hdr->Vendor_Specific_Logs_Length));
1150	if ((plba + len) * ss >= pp->mediasize)
1151		goto hdrerror;
1152	if (slba != -1 && (slba + len) * ss >= pp->mediasize)
1153		goto hdrerror;
1154	/* Workaround for Adaptec implementation. */
1155	if (GET16(meta, hdr->Max_Primary_Element_Entries) == 0xffff) {
1156		SET16(meta, hdr->Max_Primary_Element_Entries,
1157		    min(GET16(meta, hdr->Max_PD_Entries),
1158		    (GET16(meta, hdr->Configuration_Record_Length) * ss - 512) / 12));
1159	}
1160
1161	/* Read controller data. */
1162	buf = g_read_data(cp, (lba + GET32(meta, hdr->cd_section)) * ss,
1163	    GET32(meta, hdr->cd_length) * ss, &error);
1164	if (buf == NULL)
1165		goto readerror;
1166	meta->cdr = malloc(GET32(meta, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
1167	memcpy(meta->cdr, buf, GET32(meta, hdr->cd_length) * ss);
1168	g_free(buf);
1169	if (GET32(meta, cdr->Signature) != DDF_CONTROLLER_DATA_SIGNATURE)
1170		goto hdrerror;
1171
1172	/* Read physical disk records. */
1173	buf = g_read_data(cp, (lba + GET32(meta, hdr->pdr_section)) * ss,
1174	    GET32(meta, hdr->pdr_length) * ss, &error);
1175	if (buf == NULL)
1176		goto readerror;
1177	meta->pdr = malloc(GET32(meta, hdr->pdr_length) * ss, M_MD_DDF, M_WAITOK);
1178	memcpy(meta->pdr, buf, GET32(meta, hdr->pdr_length) * ss);
1179	g_free(buf);
1180	if (GET32(meta, pdr->Signature) != DDF_PDR_SIGNATURE)
1181		goto hdrerror;
1182
1183	/* Read virtual disk records. */
1184	buf = g_read_data(cp, (lba + GET32(meta, hdr->vdr_section)) * ss,
1185	    GET32(meta, hdr->vdr_length) * ss, &error);
1186	if (buf == NULL)
1187		goto readerror;
1188	meta->vdr = malloc(GET32(meta, hdr->vdr_length) * ss, M_MD_DDF, M_WAITOK);
1189	memcpy(meta->vdr, buf, GET32(meta, hdr->vdr_length) * ss);
1190	g_free(buf);
1191	if (GET32(meta, vdr->Signature) != DDF_VD_RECORD_SIGNATURE)
1192		goto hdrerror;
1193
1194	/* Read configuration records. */
1195	buf = g_read_data(cp, (lba + GET32(meta, hdr->cr_section)) * ss,
1196	    GET32(meta, hdr->cr_length) * ss, &error);
1197	if (buf == NULL)
1198		goto readerror;
1199	meta->cr = malloc(GET32(meta, hdr->cr_length) * ss, M_MD_DDF, M_WAITOK);
1200	memcpy(meta->cr, buf, GET32(meta, hdr->cr_length) * ss);
1201	g_free(buf);
1202
1203	/* Read physical disk data. */
1204	buf = g_read_data(cp, (lba + GET32(meta, hdr->pdd_section)) * ss,
1205	    GET32(meta, hdr->pdd_length) * ss, &error);
1206	if (buf == NULL)
1207		goto readerror;
1208	meta->pdd = malloc(GET32(meta, hdr->pdd_length) * ss, M_MD_DDF, M_WAITOK);
1209	memcpy(meta->pdd, buf, GET32(meta, hdr->pdd_length) * ss);
1210	g_free(buf);
1211	if (GET32(meta, pdd->Signature) != DDF_PDD_SIGNATURE)
1212		goto hdrerror;
1213	i = ddf_meta_find_pd(meta, NULL, GET32(meta, pdd->PD_Reference));
1214	if (i < 0)
1215		goto hdrerror;
1216
1217	/* Read BBM Log. */
1218	if (GET32(meta, hdr->bbmlog_section) != 0xffffffff &&
1219	    GET32(meta, hdr->bbmlog_length) != 0) {
1220		buf = g_read_data(cp, (lba + GET32(meta, hdr->bbmlog_section)) * ss,
1221		    GET32(meta, hdr->bbmlog_length) * ss, &error);
1222		if (buf == NULL)
1223			goto readerror;
1224		meta->bbm = malloc(GET32(meta, hdr->bbmlog_length) * ss, M_MD_DDF, M_WAITOK);
1225		memcpy(meta->bbm, buf, GET32(meta, hdr->bbmlog_length) * ss);
1226		g_free(buf);
1227		if (GET32(meta, bbm->Signature) != DDF_BBML_SIGNATURE)
1228			goto hdrerror;
1229	}
1230
1231done:
1232	g_free(abuf);
1233	if (error != 0)
1234		ddf_meta_free(meta);
1235	return (error);
1236}
1237
1238static int
1239ddf_meta_write(struct g_consumer *cp, struct ddf_meta *meta)
1240{
1241	struct g_provider *pp;
1242	struct ddf_vdc_record *vdc;
1243	off_t alba, plba, slba, lba;
1244	u_int ss, size;
1245	int error, i, num;
1246
1247	pp = cp->provider;
1248	ss = pp->sectorsize;
1249	lba = alba = pp->mediasize / ss - 1;
1250	plba = GET64(meta, hdr->Primary_Header_LBA);
1251	slba = GET64(meta, hdr->Secondary_Header_LBA);
1252
1253next:
1254	SET8(meta, hdr->Header_Type, (lba == alba) ? DDF_HEADER_ANCHOR :
1255	    (lba == plba) ? DDF_HEADER_PRIMARY : DDF_HEADER_SECONDARY);
1256	SET32(meta, hdr->CRC, 0xffffffff);
1257	SET32(meta, hdr->CRC, crc32(meta->hdr, ss));
1258	error = g_write_data(cp, lba * ss, meta->hdr, ss);
1259	if (error != 0) {
1260err:
1261		G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
1262		    pp->name, error);
1263		if (lba != alba)
1264			goto done;
1265	}
1266	if (lba == alba) {
1267		lba = plba;
1268		goto next;
1269	}
1270
1271	size = GET32(meta, hdr->cd_length) * ss;
1272	SET32(meta, cdr->CRC, 0xffffffff);
1273	SET32(meta, cdr->CRC, crc32(meta->cdr, size));
1274	error = g_write_data(cp, (lba + GET32(meta, hdr->cd_section)) * ss,
1275	    meta->cdr, size);
1276	if (error != 0)
1277		goto err;
1278
1279	size = GET32(meta, hdr->pdr_length) * ss;
1280	SET32(meta, pdr->CRC, 0xffffffff);
1281	SET32(meta, pdr->CRC, crc32(meta->pdr, size));
1282	error = g_write_data(cp, (lba + GET32(meta, hdr->pdr_section)) * ss,
1283	    meta->pdr, size);
1284	if (error != 0)
1285		goto err;
1286
1287	size = GET32(meta, hdr->vdr_length) * ss;
1288	SET32(meta, vdr->CRC, 0xffffffff);
1289	SET32(meta, vdr->CRC, crc32(meta->vdr, size));
1290	error = g_write_data(cp, (lba + GET32(meta, hdr->vdr_section)) * ss,
1291	    meta->vdr, size);
1292	if (error != 0)
1293		goto err;
1294
1295	size = GET16(meta, hdr->Configuration_Record_Length) * ss;
1296	num = GETCRNUM(meta);
1297	for (i = 0; i < num; i++) {
1298		vdc = GETVDCPTR(meta, i);
1299		SET32D(meta, vdc->CRC, 0xffffffff);
1300		SET32D(meta, vdc->CRC, crc32(vdc, size));
1301	}
1302	error = g_write_data(cp, (lba + GET32(meta, hdr->cr_section)) * ss,
1303	    meta->cr, size * num);
1304	if (error != 0)
1305		goto err;
1306
1307	size = GET32(meta, hdr->pdd_length) * ss;
1308	SET32(meta, pdd->CRC, 0xffffffff);
1309	SET32(meta, pdd->CRC, crc32(meta->pdd, size));
1310	error = g_write_data(cp, (lba + GET32(meta, hdr->pdd_section)) * ss,
1311	    meta->pdd, size);
1312	if (error != 0)
1313		goto err;
1314
1315	if (GET32(meta, hdr->bbmlog_length) != 0) {
1316		size = GET32(meta, hdr->bbmlog_length) * ss;
1317		SET32(meta, bbm->CRC, 0xffffffff);
1318		SET32(meta, bbm->CRC, crc32(meta->bbm, size));
1319		error = g_write_data(cp,
1320		    (lba + GET32(meta, hdr->bbmlog_section)) * ss,
1321		    meta->bbm, size);
1322		if (error != 0)
1323			goto err;
1324	}
1325
1326done:
1327	if (lba == plba && slba != -1) {
1328		lba = slba;
1329		goto next;
1330	}
1331
1332	return (error);
1333}
1334
1335static int
1336ddf_meta_erase(struct g_consumer *cp)
1337{
1338	struct g_provider *pp;
1339	char *buf;
1340	int error;
1341
1342	pp = cp->provider;
1343	buf = malloc(pp->sectorsize, M_MD_DDF, M_WAITOK | M_ZERO);
1344	error = g_write_data(cp, pp->mediasize - pp->sectorsize,
1345	    buf, pp->sectorsize);
1346	if (error != 0) {
1347		G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
1348		    pp->name, error);
1349	}
1350	free(buf, M_MD_DDF);
1351	return (error);
1352}
1353
1354static struct g_raid_volume *
1355g_raid_md_ddf_get_volume(struct g_raid_softc *sc, uint8_t *GUID)
1356{
1357	struct g_raid_volume	*vol;
1358	struct g_raid_md_ddf_pervolume *pv;
1359
1360	TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1361		pv = vol->v_md_data;
1362		if (memcmp(pv->pv_meta.vde->VD_GUID, GUID, 24) == 0)
1363			break;
1364	}
1365	return (vol);
1366}
1367
1368static struct g_raid_disk *
1369g_raid_md_ddf_get_disk(struct g_raid_softc *sc, uint8_t *GUID, uint32_t id)
1370{
1371	struct g_raid_disk	*disk;
1372	struct g_raid_md_ddf_perdisk *pd;
1373	struct ddf_meta *meta;
1374
1375	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1376		pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1377		meta = &pd->pd_meta;
1378		if (GUID != NULL) {
1379			if (memcmp(meta->pdd->PD_GUID, GUID, 24) == 0)
1380				break;
1381		} else {
1382			if (GET32(meta, pdd->PD_Reference) == id)
1383				break;
1384		}
1385	}
1386	return (disk);
1387}
1388
1389static int
1390g_raid_md_ddf_purge_volumes(struct g_raid_softc *sc)
1391{
1392	struct g_raid_volume	*vol, *tvol;
1393	struct g_raid_md_ddf_pervolume *pv;
1394	int i, res;
1395
1396	res = 0;
1397	TAILQ_FOREACH_SAFE(vol, &sc->sc_volumes, v_next, tvol) {
1398		pv = vol->v_md_data;
1399		if (vol->v_stopping)
1400			continue;
1401		for (i = 0; i < vol->v_disks_count; i++) {
1402			if (vol->v_subdisks[i].sd_state != G_RAID_SUBDISK_S_NONE)
1403				break;
1404		}
1405		if (i >= vol->v_disks_count) {
1406			g_raid_destroy_volume(vol);
1407			res = 1;
1408		}
1409	}
1410	return (res);
1411}
1412
1413static int
1414g_raid_md_ddf_purge_disks(struct g_raid_softc *sc)
1415{
1416#if 0
1417	struct g_raid_disk	*disk, *tdisk;
1418	struct g_raid_volume	*vol;
1419	struct g_raid_md_ddf_perdisk *pd;
1420	int i, j, res;
1421
1422	res = 0;
1423	TAILQ_FOREACH_SAFE(disk, &sc->sc_disks, d_next, tdisk) {
1424		if (disk->d_state == G_RAID_DISK_S_SPARE)
1425			continue;
1426		pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1427
1428		/* Scan for deleted volumes. */
1429		for (i = 0; i < pd->pd_subdisks; ) {
1430			vol = g_raid_md_ddf_get_volume(sc,
1431			    pd->pd_meta[i]->volume_id);
1432			if (vol != NULL && !vol->v_stopping) {
1433				i++;
1434				continue;
1435			}
1436			free(pd->pd_meta[i], M_MD_DDF);
1437			for (j = i; j < pd->pd_subdisks - 1; j++)
1438				pd->pd_meta[j] = pd->pd_meta[j + 1];
1439			pd->pd_meta[DDF_MAX_SUBDISKS - 1] = NULL;
1440			pd->pd_subdisks--;
1441			pd->pd_updated = 1;
1442		}
1443
1444		/* If there is no metadata left - erase and delete disk. */
1445		if (pd->pd_subdisks == 0) {
1446			ddf_meta_erase(disk->d_consumer);
1447			g_raid_destroy_disk(disk);
1448			res = 1;
1449		}
1450	}
1451	return (res);
1452#endif
1453	return (0);
1454}
1455
1456static int
1457g_raid_md_ddf_supported(int level, int qual, int disks, int force)
1458{
1459
1460	if (disks > DDF_MAX_DISKS_HARD)
1461		return (0);
1462	switch (level) {
1463	case G_RAID_VOLUME_RL_RAID0:
1464		if (qual != G_RAID_VOLUME_RLQ_NONE)
1465			return (0);
1466		if (disks < 1)
1467			return (0);
1468		if (!force && disks < 2)
1469			return (0);
1470		break;
1471	case G_RAID_VOLUME_RL_RAID1:
1472		if (disks < 1)
1473			return (0);
1474		if (qual == G_RAID_VOLUME_RLQ_R1SM) {
1475			if (!force && disks != 2)
1476				return (0);
1477		} else if (qual == G_RAID_VOLUME_RLQ_R1MM) {
1478			if (!force && disks != 3)
1479				return (0);
1480		} else
1481			return (0);
1482		break;
1483	case G_RAID_VOLUME_RL_RAID3:
1484		if (qual != G_RAID_VOLUME_RLQ_R3P0 &&
1485		    qual != G_RAID_VOLUME_RLQ_R3PN)
1486			return (0);
1487		if (disks < 3)
1488			return (0);
1489		break;
1490	case G_RAID_VOLUME_RL_RAID4:
1491		if (qual != G_RAID_VOLUME_RLQ_R4P0 &&
1492		    qual != G_RAID_VOLUME_RLQ_R4PN)
1493			return (0);
1494		if (disks < 3)
1495			return (0);
1496		break;
1497	case G_RAID_VOLUME_RL_RAID5:
1498		if (qual != G_RAID_VOLUME_RLQ_R5RA &&
1499		    qual != G_RAID_VOLUME_RLQ_R5RS &&
1500		    qual != G_RAID_VOLUME_RLQ_R5LA &&
1501		    qual != G_RAID_VOLUME_RLQ_R5LS)
1502			return (0);
1503		if (disks < 3)
1504			return (0);
1505		break;
1506	case G_RAID_VOLUME_RL_RAID6:
1507		if (qual != G_RAID_VOLUME_RLQ_R6RA &&
1508		    qual != G_RAID_VOLUME_RLQ_R6RS &&
1509		    qual != G_RAID_VOLUME_RLQ_R6LA &&
1510		    qual != G_RAID_VOLUME_RLQ_R6LS)
1511			return (0);
1512		if (disks < 4)
1513			return (0);
1514		break;
1515	case G_RAID_VOLUME_RL_RAIDMDF:
1516		if (qual != G_RAID_VOLUME_RLQ_RMDFRA &&
1517		    qual != G_RAID_VOLUME_RLQ_RMDFRS &&
1518		    qual != G_RAID_VOLUME_RLQ_RMDFLA &&
1519		    qual != G_RAID_VOLUME_RLQ_RMDFLS)
1520			return (0);
1521		if (disks < 4)
1522			return (0);
1523		break;
1524	case G_RAID_VOLUME_RL_RAID1E:
1525		if (qual != G_RAID_VOLUME_RLQ_R1EA &&
1526		    qual != G_RAID_VOLUME_RLQ_R1EO)
1527			return (0);
1528		if (disks < 3)
1529			return (0);
1530		break;
1531	case G_RAID_VOLUME_RL_SINGLE:
1532		if (qual != G_RAID_VOLUME_RLQ_NONE)
1533			return (0);
1534		if (disks != 1)
1535			return (0);
1536		break;
1537	case G_RAID_VOLUME_RL_CONCAT:
1538		if (qual != G_RAID_VOLUME_RLQ_NONE)
1539			return (0);
1540		if (disks < 2)
1541			return (0);
1542		break;
1543	case G_RAID_VOLUME_RL_RAID5E:
1544		if (qual != G_RAID_VOLUME_RLQ_R5ERA &&
1545		    qual != G_RAID_VOLUME_RLQ_R5ERS &&
1546		    qual != G_RAID_VOLUME_RLQ_R5ELA &&
1547		    qual != G_RAID_VOLUME_RLQ_R5ELS)
1548			return (0);
1549		if (disks < 4)
1550			return (0);
1551		break;
1552	case G_RAID_VOLUME_RL_RAID5EE:
1553		if (qual != G_RAID_VOLUME_RLQ_R5EERA &&
1554		    qual != G_RAID_VOLUME_RLQ_R5EERS &&
1555		    qual != G_RAID_VOLUME_RLQ_R5EELA &&
1556		    qual != G_RAID_VOLUME_RLQ_R5EELS)
1557			return (0);
1558		if (disks < 4)
1559			return (0);
1560		break;
1561	case G_RAID_VOLUME_RL_RAID5R:
1562		if (qual != G_RAID_VOLUME_RLQ_R5RRA &&
1563		    qual != G_RAID_VOLUME_RLQ_R5RRS &&
1564		    qual != G_RAID_VOLUME_RLQ_R5RLA &&
1565		    qual != G_RAID_VOLUME_RLQ_R5RLS)
1566			return (0);
1567		if (disks < 3)
1568			return (0);
1569		break;
1570	default:
1571		return (0);
1572	}
1573	return (1);
1574}
1575
1576static int
1577g_raid_md_ddf_start_disk(struct g_raid_disk *disk, struct g_raid_volume *vol)
1578{
1579	struct g_raid_softc *sc;
1580	struct g_raid_subdisk *sd;
1581	struct g_raid_md_ddf_perdisk *pd;
1582	struct g_raid_md_ddf_pervolume *pv;
1583	struct g_raid_md_ddf_object *mdi;
1584	struct ddf_vol_meta *vmeta;
1585	struct ddf_meta *pdmeta, *gmeta;
1586	struct ddf_vdc_record *vdc1;
1587	struct ddf_sa_record *sa;
1588	off_t size, eoff = 0, esize = 0;
1589	uint64_t *val2;
1590	int disk_pos, md_disk_bvd = -1, md_disk_pos = -1, md_pde_pos;
1591	int i, resurrection = 0;
1592	uint32_t reference;
1593
1594	sc = disk->d_softc;
1595	mdi = (struct g_raid_md_ddf_object *)sc->sc_md;
1596	pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1597	pdmeta = &pd->pd_meta;
1598	reference = GET32(&pd->pd_meta, pdd->PD_Reference);
1599
1600	pv = vol->v_md_data;
1601	vmeta = &pv->pv_meta;
1602	gmeta = &mdi->mdio_meta;
1603
1604	/* Find disk position in metadata by it's reference. */
1605	disk_pos = ddf_meta_find_disk(vmeta, reference,
1606	    &md_disk_bvd, &md_disk_pos);
1607	md_pde_pos = ddf_meta_find_pd(gmeta, NULL, reference);
1608
1609	if (disk_pos < 0) {
1610		G_RAID_DEBUG1(1, sc,
1611		    "Disk %s is not a present part of the volume %s",
1612		    g_raid_get_diskname(disk), vol->v_name);
1613
1614		/* Failed stale disk is useless for us. */
1615		if ((GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA) != 0) {
1616			g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
1617			return (0);
1618		}
1619
1620		/* If disk has some metadata for this volume - erase. */
1621		if ((vdc1 = ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID)) != NULL)
1622			SET32D(pdmeta, vdc1->Signature, 0xffffffff);
1623
1624		/* If we are in the start process, that's all for now. */
1625		if (!pv->pv_started)
1626			goto nofit;
1627		/*
1628		 * If we have already started - try to get use of the disk.
1629		 * Try to replace OFFLINE disks first, then FAILED.
1630		 */
1631		if (ddf_meta_count_vdc(&pd->pd_meta, NULL) >=
1632			GET16(&pd->pd_meta, hdr->Max_Partitions)) {
1633			G_RAID_DEBUG1(1, sc, "No free partitions on disk %s",
1634			    g_raid_get_diskname(disk));
1635			goto nofit;
1636		}
1637		ddf_meta_unused_range(&pd->pd_meta, &eoff, &esize);
1638		if (esize == 0) {
1639			G_RAID_DEBUG1(1, sc, "No free space on disk %s",
1640			    g_raid_get_diskname(disk));
1641			goto nofit;
1642		}
1643		eoff *= pd->pd_meta.sectorsize;
1644		esize *= pd->pd_meta.sectorsize;
1645		size = INT64_MAX;
1646		for (i = 0; i < vol->v_disks_count; i++) {
1647			sd = &vol->v_subdisks[i];
1648			if (sd->sd_state != G_RAID_SUBDISK_S_NONE)
1649				size = sd->sd_size;
1650			if (sd->sd_state <= G_RAID_SUBDISK_S_FAILED &&
1651			    (disk_pos < 0 ||
1652			     vol->v_subdisks[i].sd_state < sd->sd_state))
1653				disk_pos = i;
1654		}
1655		if (disk_pos >= 0 &&
1656		    vol->v_raid_level != G_RAID_VOLUME_RL_CONCAT &&
1657		    esize < size) {
1658			G_RAID_DEBUG1(1, sc, "Disk %s free space "
1659			    "is too small (%ju < %ju)",
1660			    g_raid_get_diskname(disk), esize, size);
1661			disk_pos = -1;
1662		}
1663		if (disk_pos >= 0) {
1664			if (vol->v_raid_level != G_RAID_VOLUME_RL_CONCAT)
1665				esize = size;
1666			md_disk_bvd = disk_pos / GET16(vmeta, vdc->Primary_Element_Count); // XXX
1667			md_disk_pos = disk_pos % GET16(vmeta, vdc->Primary_Element_Count); // XXX
1668		} else {
1669nofit:
1670			if (disk->d_state == G_RAID_DISK_S_NONE)
1671				g_raid_change_disk_state(disk,
1672				    G_RAID_DISK_S_STALE);
1673			return (0);
1674		}
1675
1676		/*
1677		 * If spare is committable, delete spare record.
1678		 * Othersize, mark it active and leave there.
1679		 */
1680		sa = ddf_meta_find_sa(&pd->pd_meta, 0);
1681		if (sa != NULL) {
1682			if ((GET8D(&pd->pd_meta, sa->Spare_Type) &
1683			    DDF_SAR_TYPE_REVERTIBLE) == 0) {
1684				SET32D(&pd->pd_meta, sa->Signature, 0xffffffff);
1685			} else {
1686				SET8D(&pd->pd_meta, sa->Spare_Type,
1687				    GET8D(&pd->pd_meta, sa->Spare_Type) |
1688				    DDF_SAR_TYPE_ACTIVE);
1689			}
1690		}
1691
1692		G_RAID_DEBUG1(1, sc, "Disk %s takes pos %d in the volume %s",
1693		    g_raid_get_diskname(disk), disk_pos, vol->v_name);
1694		resurrection = 1;
1695	}
1696
1697	sd = &vol->v_subdisks[disk_pos];
1698
1699	if (resurrection && sd->sd_disk != NULL) {
1700		g_raid_change_disk_state(sd->sd_disk,
1701		    G_RAID_DISK_S_STALE_FAILED);
1702		TAILQ_REMOVE(&sd->sd_disk->d_subdisks,
1703		    sd, sd_next);
1704	}
1705	vol->v_subdisks[disk_pos].sd_disk = disk;
1706	TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1707
1708	/* Welcome the new disk. */
1709	if (resurrection)
1710		g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1711	else if (GET8(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA)
1712		g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1713	else
1714		g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1715
1716	if (resurrection) {
1717		sd->sd_offset = eoff;
1718		sd->sd_size = esize;
1719	} else if (pdmeta->cr != NULL &&
1720	    (vdc1 = ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID)) != NULL) {
1721		val2 = (uint64_t *)&(vdc1->Physical_Disk_Sequence[GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
1722		sd->sd_offset = (off_t)GET64P(pdmeta, val2 + md_disk_pos) * 512;
1723		sd->sd_size = (off_t)GET64D(pdmeta, vdc1->Block_Count) * 512;
1724	}
1725
1726	if (resurrection) {
1727		/* Stale disk, almost same as new. */
1728		g_raid_change_subdisk_state(sd,
1729		    G_RAID_SUBDISK_S_NEW);
1730	} else if (GET8(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA) {
1731		/* Failed disk. */
1732		g_raid_change_subdisk_state(sd,
1733		    G_RAID_SUBDISK_S_FAILED);
1734	} else if ((GET8(gmeta, pdr->entry[md_pde_pos].PD_State) &
1735	     (DDF_PDE_FAILED | DDF_PDE_REBUILD)) != 0) {
1736		/* Rebuilding disk. */
1737		g_raid_change_subdisk_state(sd,
1738		    G_RAID_SUBDISK_S_REBUILD);
1739		sd->sd_rebuild_pos = 0;
1740	} else if ((GET8(vmeta, vde->VD_State) & DDF_VDE_DIRTY) != 0 ||
1741	    (GET8(vmeta, vde->Init_State) & DDF_VDE_INIT_MASK) !=
1742	     DDF_VDE_INIT_FULL) {
1743		/* Stale disk or dirty volume (unclean shutdown). */
1744		g_raid_change_subdisk_state(sd,
1745		    G_RAID_SUBDISK_S_STALE);
1746	} else {
1747		/* Up to date disk. */
1748		g_raid_change_subdisk_state(sd,
1749		    G_RAID_SUBDISK_S_ACTIVE);
1750	}
1751	g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1752	    G_RAID_EVENT_SUBDISK);
1753
1754	return (resurrection);
1755}
1756
1757static void
1758g_raid_md_ddf_refill(struct g_raid_softc *sc)
1759{
1760	struct g_raid_volume *vol;
1761	struct g_raid_subdisk *sd;
1762	struct g_raid_disk *disk;
1763	struct g_raid_md_object *md;
1764	struct g_raid_md_ddf_perdisk *pd;
1765	struct g_raid_md_ddf_pervolume *pv;
1766	int update, updated, i, bad;
1767
1768	md = sc->sc_md;
1769restart:
1770	updated = 0;
1771	TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1772		pv = vol->v_md_data;
1773		if (!pv->pv_started || vol->v_stopping)
1774			continue;
1775
1776		/* Search for subdisk that needs replacement. */
1777		bad = 0;
1778		for (i = 0; i < vol->v_disks_count; i++) {
1779			sd = &vol->v_subdisks[i];
1780			if (sd->sd_state == G_RAID_SUBDISK_S_NONE ||
1781			    sd->sd_state == G_RAID_SUBDISK_S_FAILED)
1782			        bad = 1;
1783		}
1784		if (!bad)
1785			continue;
1786
1787		G_RAID_DEBUG1(1, sc, "Volume %s is not complete, "
1788		    "trying to refill.", vol->v_name);
1789
1790		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1791			/* Skip failed. */
1792			if (disk->d_state < G_RAID_DISK_S_SPARE)
1793				continue;
1794			/* Skip already used by this volume. */
1795			for (i = 0; i < vol->v_disks_count; i++) {
1796				sd = &vol->v_subdisks[i];
1797				if (sd->sd_disk == disk)
1798					break;
1799			}
1800			if (i < vol->v_disks_count)
1801				continue;
1802
1803			/* Try to use disk if it has empty extents. */
1804			pd = disk->d_md_data;
1805			if (ddf_meta_count_vdc(&pd->pd_meta, NULL) <
1806			    GET16(&pd->pd_meta, hdr->Max_Partitions)) {
1807				update = g_raid_md_ddf_start_disk(disk, vol);
1808			} else
1809				update = 0;
1810			if (update) {
1811				updated = 1;
1812				g_raid_md_write_ddf(md, vol, NULL, disk);
1813				break;
1814			}
1815		}
1816	}
1817	if (updated)
1818		goto restart;
1819}
1820
1821static void
1822g_raid_md_ddf_start(struct g_raid_volume *vol)
1823{
1824	struct g_raid_softc *sc;
1825	struct g_raid_subdisk *sd;
1826	struct g_raid_disk *disk;
1827	struct g_raid_md_object *md;
1828	struct g_raid_md_ddf_perdisk *pd;
1829	struct g_raid_md_ddf_pervolume *pv;
1830	struct g_raid_md_ddf_object *mdi;
1831	struct ddf_vol_meta *vmeta;
1832	struct ddf_vdc_record *vdc;
1833	uint64_t *val2;
1834	int i, j, bvd;
1835
1836	sc = vol->v_softc;
1837	md = sc->sc_md;
1838	mdi = (struct g_raid_md_ddf_object *)md;
1839	pv = vol->v_md_data;
1840	vmeta = &pv->pv_meta;
1841	vdc = vmeta->vdc;
1842
1843	vol->v_raid_level = GET8(vmeta, vdc->Primary_RAID_Level);
1844	vol->v_raid_level_qualifier = GET8(vmeta, vdc->RLQ);
1845	if (GET8(vmeta, vdc->Secondary_Element_Count) > 1 &&
1846	    vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 &&
1847	    GET8(vmeta, vdc->Secondary_RAID_Level) == 0)
1848		vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1849	vol->v_sectorsize = GET16(vmeta, vdc->Block_Size);
1850	if (vol->v_sectorsize == 0xffff)
1851		vol->v_sectorsize = vmeta->sectorsize;
1852	vol->v_strip_size = vol->v_sectorsize << GET8(vmeta, vdc->Stripe_Size);
1853	vol->v_disks_count = GET16(vmeta, vdc->Primary_Element_Count) *
1854	    GET8(vmeta, vdc->Secondary_Element_Count);
1855	vol->v_mdf_pdisks = GET8(vmeta, vdc->MDF_Parity_Disks);
1856	vol->v_mdf_polynomial = GET16(vmeta, vdc->MDF_Parity_Generator_Polynomial);
1857	vol->v_mdf_method = GET8(vmeta, vdc->MDF_Constant_Generation_Method);
1858	if (GET8(vmeta, vdc->Rotate_Parity_count) > 31)
1859		vol->v_rotate_parity = 1;
1860	else
1861		vol->v_rotate_parity = 1 << GET8(vmeta, vdc->Rotate_Parity_count);
1862	vol->v_mediasize = GET64(vmeta, vdc->VD_Size) * vol->v_sectorsize;
1863	for (i = 0, j = 0, bvd = 0; i < vol->v_disks_count; i++, j++) {
1864		if (j == GET16(vmeta, vdc->Primary_Element_Count)) {
1865			j = 0;
1866			bvd++;
1867		}
1868		sd = &vol->v_subdisks[i];
1869		if (vmeta->bvdc[bvd] == NULL) {
1870			sd->sd_offset = 0;
1871			sd->sd_size = GET64(vmeta, vdc->Block_Count) *
1872			    vol->v_sectorsize;
1873			continue;
1874		}
1875		val2 = (uint64_t *)&(vmeta->bvdc[bvd]->Physical_Disk_Sequence[
1876		    GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
1877		sd->sd_offset = GET64P(vmeta, val2 + j) * vol->v_sectorsize;
1878		sd->sd_size = GET64(vmeta, bvdc[bvd]->Block_Count) *
1879		    vol->v_sectorsize;
1880	}
1881	g_raid_start_volume(vol);
1882
1883	/* Make all disks found till the moment take their places. */
1884	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1885		pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1886		if (ddf_meta_find_vdc(&pd->pd_meta, vmeta->vdc->VD_GUID) != NULL)
1887			g_raid_md_ddf_start_disk(disk, vol);
1888	}
1889
1890	pv->pv_started = 1;
1891	mdi->mdio_starting--;
1892	callout_stop(&pv->pv_start_co);
1893	G_RAID_DEBUG1(0, sc, "Volume started.");
1894	g_raid_md_write_ddf(md, vol, NULL, NULL);
1895
1896	/* Pickup any STALE/SPARE disks to refill array if needed. */
1897	g_raid_md_ddf_refill(sc);
1898
1899	g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
1900}
1901
1902static void
1903g_raid_ddf_go(void *arg)
1904{
1905	struct g_raid_volume *vol;
1906	struct g_raid_softc *sc;
1907	struct g_raid_md_ddf_pervolume *pv;
1908
1909	vol = arg;
1910	pv = vol->v_md_data;
1911	sc = vol->v_softc;
1912	if (!pv->pv_started) {
1913		G_RAID_DEBUG1(0, sc, "Force volume start due to timeout.");
1914		g_raid_event_send(vol, G_RAID_VOLUME_E_STARTMD,
1915		    G_RAID_EVENT_VOLUME);
1916	}
1917}
1918
1919static void
1920g_raid_md_ddf_new_disk(struct g_raid_disk *disk)
1921{
1922	struct g_raid_softc *sc;
1923	struct g_raid_md_object *md;
1924	struct g_raid_md_ddf_perdisk *pd;
1925	struct g_raid_md_ddf_pervolume *pv;
1926	struct g_raid_md_ddf_object *mdi;
1927	struct g_raid_volume *vol;
1928	struct ddf_meta *pdmeta;
1929	struct ddf_vol_meta *vmeta;
1930	struct ddf_vdc_record *vdc;
1931	struct ddf_vd_entry *vde;
1932	int i, j, k, num, have, need, cnt, spare;
1933	uint32_t val;
1934	char buf[17];
1935
1936	sc = disk->d_softc;
1937	md = sc->sc_md;
1938	mdi = (struct g_raid_md_ddf_object *)md;
1939	pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1940	pdmeta = &pd->pd_meta;
1941	spare = -1;
1942
1943	if (mdi->mdio_meta.hdr == NULL)
1944		ddf_meta_copy(&mdi->mdio_meta, pdmeta);
1945	else
1946		ddf_meta_update(&mdi->mdio_meta, pdmeta);
1947
1948	num = GETCRNUM(pdmeta);
1949	for (j = 0; j < num; j++) {
1950		vdc = GETVDCPTR(pdmeta, j);
1951		val = GET32D(pdmeta, vdc->Signature);
1952
1953		if (val == DDF_SA_SIGNATURE && spare == -1)
1954			spare = 1;
1955
1956		if (val != DDF_VDCR_SIGNATURE)
1957			continue;
1958		spare = 0;
1959		k = ddf_meta_find_vd(pdmeta, vdc->VD_GUID);
1960		if (k < 0)
1961			continue;
1962		vde = &pdmeta->vdr->entry[k];
1963
1964		/* Look for volume with matching ID. */
1965		vol = g_raid_md_ddf_get_volume(sc, vdc->VD_GUID);
1966		if (vol == NULL) {
1967			ddf_meta_get_name(pdmeta, k, buf);
1968			vol = g_raid_create_volume(sc, buf,
1969			    GET16D(pdmeta, vde->VD_Number));
1970			pv = malloc(sizeof(*pv), M_MD_DDF, M_WAITOK | M_ZERO);
1971			vol->v_md_data = pv;
1972			callout_init(&pv->pv_start_co, 1);
1973			callout_reset(&pv->pv_start_co,
1974			    g_raid_start_timeout * hz,
1975			    g_raid_ddf_go, vol);
1976			mdi->mdio_starting++;
1977		} else
1978			pv = vol->v_md_data;
1979
1980		/* If we haven't started yet - check metadata freshness. */
1981		vmeta = &pv->pv_meta;
1982		ddf_vol_meta_update(vmeta, pdmeta, vdc->VD_GUID, pv->pv_started);
1983	}
1984
1985	if (spare == 1) {
1986		g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
1987		g_raid_md_ddf_refill(sc);
1988	}
1989
1990	TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1991		pv = vol->v_md_data;
1992		vmeta = &pv->pv_meta;
1993
1994		if (ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID) == NULL)
1995			continue;
1996
1997		if (pv->pv_started) {
1998			if (g_raid_md_ddf_start_disk(disk, vol))
1999				g_raid_md_write_ddf(md, vol, NULL, NULL);
2000			continue;
2001		}
2002
2003		/* If we collected all needed disks - start array. */
2004		need = 0;
2005		have = 0;
2006		for (k = 0; k < GET8(vmeta, vdc->Secondary_Element_Count); k++) {
2007			if (vmeta->bvdc[k] == NULL) {
2008				need += GET16(vmeta, vdc->Primary_Element_Count);
2009				continue;
2010			}
2011			cnt = GET16(vmeta, bvdc[k]->Primary_Element_Count);
2012			need += cnt;
2013			for (i = 0; i < cnt; i++) {
2014				val = GET32(vmeta, bvdc[k]->Physical_Disk_Sequence[i]);
2015				if (g_raid_md_ddf_get_disk(sc, NULL, val) != NULL)
2016					have++;
2017			}
2018		}
2019		G_RAID_DEBUG1(1, sc, "Volume %s now has %d of %d disks",
2020		    vol->v_name, have, need);
2021		if (have == need)
2022			g_raid_md_ddf_start(vol);
2023	}
2024}
2025
2026static int
2027g_raid_md_create_req_ddf(struct g_raid_md_object *md, struct g_class *mp,
2028    struct gctl_req *req, struct g_geom **gp)
2029{
2030	struct g_geom *geom;
2031	struct g_raid_softc *sc;
2032	struct g_raid_md_ddf_object *mdi, *mdi1;
2033	char name[16];
2034	const char *fmtopt;
2035	int be = 1;
2036
2037	mdi = (struct g_raid_md_ddf_object *)md;
2038	fmtopt = gctl_get_asciiparam(req, "fmtopt");
2039	if (fmtopt == NULL || strcasecmp(fmtopt, "BE") == 0)
2040		be = 1;
2041	else if (strcasecmp(fmtopt, "LE") == 0)
2042		be = 0;
2043	else {
2044		gctl_error(req, "Incorrect fmtopt argument.");
2045		return (G_RAID_MD_TASTE_FAIL);
2046	}
2047
2048	/* Search for existing node. */
2049	LIST_FOREACH(geom, &mp->geom, geom) {
2050		sc = geom->softc;
2051		if (sc == NULL)
2052			continue;
2053		if (sc->sc_stopping != 0)
2054			continue;
2055		if (sc->sc_md->mdo_class != md->mdo_class)
2056			continue;
2057		mdi1 = (struct g_raid_md_ddf_object *)sc->sc_md;
2058		if (mdi1->mdio_bigendian != be)
2059			continue;
2060		break;
2061	}
2062	if (geom != NULL) {
2063		*gp = geom;
2064		return (G_RAID_MD_TASTE_EXISTING);
2065	}
2066
2067	/* Create new one if not found. */
2068	mdi->mdio_bigendian = be;
2069	snprintf(name, sizeof(name), "DDF%s", be ? "" : "-LE");
2070	sc = g_raid_create_node(mp, name, md);
2071	if (sc == NULL)
2072		return (G_RAID_MD_TASTE_FAIL);
2073	md->mdo_softc = sc;
2074	*gp = sc->sc_geom;
2075	return (G_RAID_MD_TASTE_NEW);
2076}
2077
2078static int
2079g_raid_md_taste_ddf(struct g_raid_md_object *md, struct g_class *mp,
2080                              struct g_consumer *cp, struct g_geom **gp)
2081{
2082	struct g_consumer *rcp;
2083	struct g_provider *pp;
2084	struct g_raid_softc *sc;
2085	struct g_raid_disk *disk;
2086	struct ddf_meta meta;
2087	struct g_raid_md_ddf_perdisk *pd;
2088	struct g_raid_md_ddf_object *mdi;
2089	struct g_geom *geom;
2090	int error, result, len, be;
2091	char name[16];
2092
2093	G_RAID_DEBUG(1, "Tasting DDF on %s", cp->provider->name);
2094	mdi = (struct g_raid_md_ddf_object *)md;
2095	pp = cp->provider;
2096
2097	/* Read metadata from device. */
2098	if (g_access(cp, 1, 0, 0) != 0)
2099		return (G_RAID_MD_TASTE_FAIL);
2100	g_topology_unlock();
2101	bzero(&meta, sizeof(meta));
2102	error = ddf_meta_read(cp, &meta);
2103	g_topology_lock();
2104	g_access(cp, -1, 0, 0);
2105	if (error != 0)
2106		return (G_RAID_MD_TASTE_FAIL);
2107	be = meta.bigendian;
2108
2109	/* Metadata valid. Print it. */
2110	g_raid_md_ddf_print(&meta);
2111
2112	/* Search for matching node. */
2113	sc = NULL;
2114	LIST_FOREACH(geom, &mp->geom, geom) {
2115		sc = geom->softc;
2116		if (sc == NULL)
2117			continue;
2118		if (sc->sc_stopping != 0)
2119			continue;
2120		if (sc->sc_md->mdo_class != md->mdo_class)
2121			continue;
2122		mdi = (struct g_raid_md_ddf_object *)sc->sc_md;
2123		if (mdi->mdio_bigendian != be)
2124			continue;
2125		break;
2126	}
2127
2128	/* Found matching node. */
2129	if (geom != NULL) {
2130		G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
2131		result = G_RAID_MD_TASTE_EXISTING;
2132
2133	} else { /* Not found matching node -- create one. */
2134		result = G_RAID_MD_TASTE_NEW;
2135		mdi->mdio_bigendian = be;
2136		snprintf(name, sizeof(name), "DDF%s", be ? "" : "-LE");
2137		sc = g_raid_create_node(mp, name, md);
2138		md->mdo_softc = sc;
2139		geom = sc->sc_geom;
2140	}
2141
2142	rcp = g_new_consumer(geom);
2143	g_attach(rcp, pp);
2144	if (g_access(rcp, 1, 1, 1) != 0)
2145		; //goto fail1;
2146
2147	g_topology_unlock();
2148	sx_xlock(&sc->sc_lock);
2149
2150	pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2151	pd->pd_meta = meta;
2152	disk = g_raid_create_disk(sc);
2153	disk->d_md_data = (void *)pd;
2154	disk->d_consumer = rcp;
2155	rcp->private = disk;
2156
2157	/* Read kernel dumping information. */
2158	disk->d_kd.offset = 0;
2159	disk->d_kd.length = OFF_MAX;
2160	len = sizeof(disk->d_kd);
2161	error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd);
2162	if (disk->d_kd.di.dumper == NULL)
2163		G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.",
2164		    rcp->provider->name, error);
2165
2166	g_raid_md_ddf_new_disk(disk);
2167
2168	sx_xunlock(&sc->sc_lock);
2169	g_topology_lock();
2170	*gp = geom;
2171	return (result);
2172}
2173
2174static int
2175g_raid_md_event_ddf(struct g_raid_md_object *md,
2176    struct g_raid_disk *disk, u_int event)
2177{
2178	struct g_raid_softc *sc;
2179
2180	sc = md->mdo_softc;
2181	if (disk == NULL)
2182		return (-1);
2183	switch (event) {
2184	case G_RAID_DISK_E_DISCONNECTED:
2185		/* Delete disk. */
2186		g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2187		g_raid_destroy_disk(disk);
2188		g_raid_md_ddf_purge_volumes(sc);
2189
2190		/* Write updated metadata to all disks. */
2191		g_raid_md_write_ddf(md, NULL, NULL, NULL);
2192
2193		/* Check if anything left. */
2194		if (g_raid_ndisks(sc, -1) == 0)
2195			g_raid_destroy_node(sc, 0);
2196		else
2197			g_raid_md_ddf_refill(sc);
2198		return (0);
2199	}
2200	return (-2);
2201}
2202
2203static int
2204g_raid_md_volume_event_ddf(struct g_raid_md_object *md,
2205    struct g_raid_volume *vol, u_int event)
2206{
2207	struct g_raid_md_ddf_pervolume *pv;
2208
2209	pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2210	switch (event) {
2211	case G_RAID_VOLUME_E_STARTMD:
2212		if (!pv->pv_started)
2213			g_raid_md_ddf_start(vol);
2214		return (0);
2215	}
2216	return (-2);
2217}
2218
2219static int
2220g_raid_md_ctl_ddf(struct g_raid_md_object *md,
2221    struct gctl_req *req)
2222{
2223	struct g_raid_softc *sc;
2224	struct g_raid_volume *vol, *vol1;
2225	struct g_raid_subdisk *sd;
2226	struct g_raid_disk *disk, *disks[DDF_MAX_DISKS_HARD];
2227	struct g_raid_md_ddf_perdisk *pd;
2228	struct g_raid_md_ddf_pervolume *pv;
2229	struct g_raid_md_ddf_object *mdi;
2230	struct ddf_sa_record *sa;
2231	struct g_consumer *cp;
2232	struct g_provider *pp;
2233	char arg[16];
2234	const char *nodename, *verb, *volname, *levelname, *diskname;
2235	char *tmp;
2236	int *nargs, *force;
2237	off_t size, sectorsize, strip, offs[DDF_MAX_DISKS_HARD], esize;
2238	intmax_t *sizearg, *striparg;
2239	int i, numdisks, len, level, qual;
2240	int error;
2241
2242	sc = md->mdo_softc;
2243	mdi = (struct g_raid_md_ddf_object *)md;
2244	verb = gctl_get_param(req, "verb", NULL);
2245	nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
2246	error = 0;
2247
2248	if (strcmp(verb, "label") == 0) {
2249
2250		if (*nargs < 4) {
2251			gctl_error(req, "Invalid number of arguments.");
2252			return (-1);
2253		}
2254		volname = gctl_get_asciiparam(req, "arg1");
2255		if (volname == NULL) {
2256			gctl_error(req, "No volume name.");
2257			return (-2);
2258		}
2259		levelname = gctl_get_asciiparam(req, "arg2");
2260		if (levelname == NULL) {
2261			gctl_error(req, "No RAID level.");
2262			return (-3);
2263		}
2264		if (g_raid_volume_str2level(levelname, &level, &qual)) {
2265			gctl_error(req, "Unknown RAID level '%s'.", levelname);
2266			return (-4);
2267		}
2268		numdisks = *nargs - 3;
2269		force = gctl_get_paraml(req, "force", sizeof(*force));
2270		if (!g_raid_md_ddf_supported(level, qual, numdisks,
2271		    force ? *force : 0)) {
2272			gctl_error(req, "Unsupported RAID level "
2273			    "(0x%02x/0x%02x), or number of disks (%d).",
2274			    level, qual, numdisks);
2275			return (-5);
2276		}
2277
2278		/* Search for disks, connect them and probe. */
2279		size = INT64_MAX;
2280		sectorsize = 0;
2281		bzero(disks, sizeof(disks));
2282		bzero(offs, sizeof(offs));
2283		for (i = 0; i < numdisks; i++) {
2284			snprintf(arg, sizeof(arg), "arg%d", i + 3);
2285			diskname = gctl_get_asciiparam(req, arg);
2286			if (diskname == NULL) {
2287				gctl_error(req, "No disk name (%s).", arg);
2288				error = -6;
2289				break;
2290			}
2291			if (strcmp(diskname, "NONE") == 0)
2292				continue;
2293
2294			TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2295				if (disk->d_consumer != NULL &&
2296				    disk->d_consumer->provider != NULL &&
2297				    strcmp(disk->d_consumer->provider->name,
2298				     diskname) == 0)
2299					break;
2300			}
2301			if (disk != NULL) {
2302				if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2303					gctl_error(req, "Disk '%s' is in a "
2304					    "wrong state (%s).", diskname,
2305					    g_raid_disk_state2str(disk->d_state));
2306					error = -7;
2307					break;
2308				}
2309				pd = disk->d_md_data;
2310				if (ddf_meta_count_vdc(&pd->pd_meta, NULL) >=
2311				    GET16(&pd->pd_meta, hdr->Max_Partitions)) {
2312					gctl_error(req, "No free partitions "
2313					    "on disk '%s'.",
2314					    diskname);
2315					error = -7;
2316					break;
2317				}
2318				pp = disk->d_consumer->provider;
2319				disks[i] = disk;
2320				ddf_meta_unused_range(&pd->pd_meta,
2321				    &offs[i], &esize);
2322				offs[i] *= pp->sectorsize;
2323				size = MIN(size, (off_t)esize * pp->sectorsize);
2324				sectorsize = MAX(sectorsize, pp->sectorsize);
2325				continue;
2326			}
2327
2328			g_topology_lock();
2329			cp = g_raid_open_consumer(sc, diskname);
2330			if (cp == NULL) {
2331				gctl_error(req, "Can't open disk '%s'.",
2332				    diskname);
2333				g_topology_unlock();
2334				error = -8;
2335				break;
2336			}
2337			pp = cp->provider;
2338			pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2339			disk = g_raid_create_disk(sc);
2340			disk->d_md_data = (void *)pd;
2341			disk->d_consumer = cp;
2342			disks[i] = disk;
2343			cp->private = disk;
2344			ddf_meta_create(disk, &mdi->mdio_meta);
2345			if (mdi->mdio_meta.hdr == NULL)
2346				ddf_meta_copy(&mdi->mdio_meta, &pd->pd_meta);
2347			else
2348				ddf_meta_update(&mdi->mdio_meta, &pd->pd_meta);
2349			g_topology_unlock();
2350
2351			/* Read kernel dumping information. */
2352			disk->d_kd.offset = 0;
2353			disk->d_kd.length = OFF_MAX;
2354			len = sizeof(disk->d_kd);
2355			g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
2356			if (disk->d_kd.di.dumper == NULL)
2357				G_RAID_DEBUG1(2, sc,
2358				    "Dumping not supported by %s.",
2359				    cp->provider->name);
2360
2361			/* Reserve some space for metadata. */
2362			size = MIN(size, GET64(&pd->pd_meta,
2363			    pdr->entry[0].Configured_Size) * pp->sectorsize);
2364			sectorsize = MAX(sectorsize, pp->sectorsize);
2365		}
2366		if (error != 0) {
2367			for (i = 0; i < numdisks; i++) {
2368				if (disks[i] != NULL &&
2369				    disks[i]->d_state == G_RAID_DISK_S_NONE)
2370					g_raid_destroy_disk(disks[i]);
2371			}
2372			return (error);
2373		}
2374
2375		if (sectorsize <= 0) {
2376			gctl_error(req, "Can't get sector size.");
2377			return (-8);
2378		}
2379
2380		/* Handle size argument. */
2381		len = sizeof(*sizearg);
2382		sizearg = gctl_get_param(req, "size", &len);
2383		if (sizearg != NULL && len == sizeof(*sizearg) &&
2384		    *sizearg > 0) {
2385			if (*sizearg > size) {
2386				gctl_error(req, "Size too big %lld > %lld.",
2387				    (long long)*sizearg, (long long)size);
2388				return (-9);
2389			}
2390			size = *sizearg;
2391		}
2392
2393		/* Handle strip argument. */
2394		strip = 131072;
2395		len = sizeof(*striparg);
2396		striparg = gctl_get_param(req, "strip", &len);
2397		if (striparg != NULL && len == sizeof(*striparg) &&
2398		    *striparg > 0) {
2399			if (*striparg < sectorsize) {
2400				gctl_error(req, "Strip size too small.");
2401				return (-10);
2402			}
2403			if (*striparg % sectorsize != 0) {
2404				gctl_error(req, "Incorrect strip size.");
2405				return (-11);
2406			}
2407			strip = *striparg;
2408		}
2409
2410		/* Round size down to strip or sector. */
2411		if (level == G_RAID_VOLUME_RL_RAID1 ||
2412		    level == G_RAID_VOLUME_RL_RAID3 ||
2413		    level == G_RAID_VOLUME_RL_SINGLE ||
2414		    level == G_RAID_VOLUME_RL_CONCAT)
2415			size -= (size % sectorsize);
2416		else if (level == G_RAID_VOLUME_RL_RAID1E &&
2417		    (numdisks & 1) != 0)
2418			size -= (size % (2 * strip));
2419		else
2420			size -= (size % strip);
2421		if (size <= 0) {
2422			gctl_error(req, "Size too small.");
2423			return (-13);
2424		}
2425
2426		/* We have all we need, create things: volume, ... */
2427		pv = malloc(sizeof(*pv), M_MD_DDF, M_WAITOK | M_ZERO);
2428		ddf_vol_meta_create(&pv->pv_meta, &mdi->mdio_meta);
2429		pv->pv_started = 1;
2430		vol = g_raid_create_volume(sc, volname, -1);
2431		vol->v_md_data = pv;
2432		vol->v_raid_level = level;
2433		vol->v_raid_level_qualifier = qual;
2434		vol->v_strip_size = strip;
2435		vol->v_disks_count = numdisks;
2436		if (level == G_RAID_VOLUME_RL_RAID0 ||
2437		    level == G_RAID_VOLUME_RL_CONCAT ||
2438		    level == G_RAID_VOLUME_RL_SINGLE)
2439			vol->v_mediasize = size * numdisks;
2440		else if (level == G_RAID_VOLUME_RL_RAID1)
2441			vol->v_mediasize = size;
2442		else if (level == G_RAID_VOLUME_RL_RAID3 ||
2443		    level == G_RAID_VOLUME_RL_RAID4 ||
2444		    level == G_RAID_VOLUME_RL_RAID5)
2445			vol->v_mediasize = size * (numdisks - 1);
2446		else if (level == G_RAID_VOLUME_RL_RAID5R) {
2447			vol->v_mediasize = size * (numdisks - 1);
2448			vol->v_rotate_parity = 1024;
2449		} else if (level == G_RAID_VOLUME_RL_RAID6 ||
2450		    level == G_RAID_VOLUME_RL_RAID5E ||
2451		    level == G_RAID_VOLUME_RL_RAID5EE)
2452			vol->v_mediasize = size * (numdisks - 2);
2453		else if (level == G_RAID_VOLUME_RL_RAIDMDF) {
2454			if (numdisks < 5)
2455				vol->v_mdf_pdisks = 2;
2456			else
2457				vol->v_mdf_pdisks = 3;
2458			vol->v_mdf_polynomial = 0x11d;
2459			vol->v_mdf_method = 0x00;
2460			vol->v_mediasize = size * (numdisks - vol->v_mdf_pdisks);
2461		} else { /* RAID1E */
2462			vol->v_mediasize = ((size * numdisks) / strip / 2) *
2463			    strip;
2464		}
2465		vol->v_sectorsize = sectorsize;
2466		g_raid_start_volume(vol);
2467
2468		/* , and subdisks. */
2469		for (i = 0; i < numdisks; i++) {
2470			disk = disks[i];
2471			sd = &vol->v_subdisks[i];
2472			sd->sd_disk = disk;
2473			sd->sd_offset = offs[i];
2474			sd->sd_size = size;
2475			if (disk == NULL)
2476				continue;
2477			TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
2478			g_raid_change_disk_state(disk,
2479			    G_RAID_DISK_S_ACTIVE);
2480			g_raid_change_subdisk_state(sd,
2481			    G_RAID_SUBDISK_S_ACTIVE);
2482			g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
2483			    G_RAID_EVENT_SUBDISK);
2484		}
2485
2486		/* Write metadata based on created entities. */
2487		G_RAID_DEBUG1(0, sc, "Array started.");
2488		g_raid_md_write_ddf(md, vol, NULL, NULL);
2489
2490		/* Pickup any STALE/SPARE disks to refill array if needed. */
2491		g_raid_md_ddf_refill(sc);
2492
2493		g_raid_event_send(vol, G_RAID_VOLUME_E_START,
2494		    G_RAID_EVENT_VOLUME);
2495		return (0);
2496	}
2497	if (strcmp(verb, "add") == 0) {
2498
2499		gctl_error(req, "`add` command is not applicable, "
2500		    "use `label` instead.");
2501		return (-99);
2502	}
2503	if (strcmp(verb, "delete") == 0) {
2504
2505		nodename = gctl_get_asciiparam(req, "arg0");
2506		if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2507			nodename = NULL;
2508
2509		/* Full node destruction. */
2510		if (*nargs == 1 && nodename != NULL) {
2511			/* Check if some volume is still open. */
2512			force = gctl_get_paraml(req, "force", sizeof(*force));
2513			if (force != NULL && *force == 0 &&
2514			    g_raid_nopens(sc) != 0) {
2515				gctl_error(req, "Some volume is still open.");
2516				return (-4);
2517			}
2518
2519			TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2520				if (disk->d_consumer)
2521					ddf_meta_erase(disk->d_consumer);
2522			}
2523			g_raid_destroy_node(sc, 0);
2524			return (0);
2525		}
2526
2527		/* Destroy specified volume. If it was last - all node. */
2528		if (*nargs > 2) {
2529			gctl_error(req, "Invalid number of arguments.");
2530			return (-1);
2531		}
2532		volname = gctl_get_asciiparam(req,
2533		    nodename != NULL ? "arg1" : "arg0");
2534		if (volname == NULL) {
2535			gctl_error(req, "No volume name.");
2536			return (-2);
2537		}
2538
2539		/* Search for volume. */
2540		TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2541			if (strcmp(vol->v_name, volname) == 0)
2542				break;
2543			pp = vol->v_provider;
2544			if (pp == NULL)
2545				continue;
2546			if (strcmp(pp->name, volname) == 0)
2547				break;
2548			if (strncmp(pp->name, "raid/", 5) == 0 &&
2549			    strcmp(pp->name + 5, volname) == 0)
2550				break;
2551		}
2552		if (vol == NULL) {
2553			i = strtol(volname, &tmp, 10);
2554			if (verb != volname && tmp[0] == 0) {
2555				TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2556					if (vol->v_global_id == i)
2557						break;
2558				}
2559			}
2560		}
2561		if (vol == NULL) {
2562			gctl_error(req, "Volume '%s' not found.", volname);
2563			return (-3);
2564		}
2565
2566		/* Check if volume is still open. */
2567		force = gctl_get_paraml(req, "force", sizeof(*force));
2568		if (force != NULL && *force == 0 &&
2569		    vol->v_provider_open != 0) {
2570			gctl_error(req, "Volume is still open.");
2571			return (-4);
2572		}
2573
2574		/* Destroy volume and potentially node. */
2575		i = 0;
2576		TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2577			i++;
2578		if (i >= 2) {
2579			g_raid_destroy_volume(vol);
2580			g_raid_md_ddf_purge_disks(sc);
2581			g_raid_md_write_ddf(md, NULL, NULL, NULL);
2582		} else {
2583			TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2584				if (disk->d_consumer)
2585					ddf_meta_erase(disk->d_consumer);
2586			}
2587			g_raid_destroy_node(sc, 0);
2588		}
2589		return (0);
2590	}
2591	if (strcmp(verb, "remove") == 0 ||
2592	    strcmp(verb, "fail") == 0) {
2593		if (*nargs < 2) {
2594			gctl_error(req, "Invalid number of arguments.");
2595			return (-1);
2596		}
2597		for (i = 1; i < *nargs; i++) {
2598			snprintf(arg, sizeof(arg), "arg%d", i);
2599			diskname = gctl_get_asciiparam(req, arg);
2600			if (diskname == NULL) {
2601				gctl_error(req, "No disk name (%s).", arg);
2602				error = -2;
2603				break;
2604			}
2605			if (strncmp(diskname, "/dev/", 5) == 0)
2606				diskname += 5;
2607
2608			TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2609				if (disk->d_consumer != NULL &&
2610				    disk->d_consumer->provider != NULL &&
2611				    strcmp(disk->d_consumer->provider->name,
2612				     diskname) == 0)
2613					break;
2614			}
2615			if (disk == NULL) {
2616				gctl_error(req, "Disk '%s' not found.",
2617				    diskname);
2618				error = -3;
2619				break;
2620			}
2621
2622			if (strcmp(verb, "fail") == 0) {
2623				g_raid_md_fail_disk_ddf(md, NULL, disk);
2624				continue;
2625			}
2626
2627			/* Erase metadata on deleting disk and destroy it. */
2628			ddf_meta_erase(disk->d_consumer);
2629			g_raid_destroy_disk(disk);
2630		}
2631		g_raid_md_ddf_purge_volumes(sc);
2632
2633		/* Write updated metadata to remaining disks. */
2634		g_raid_md_write_ddf(md, NULL, NULL, NULL);
2635
2636		/* Check if anything left. */
2637		if (g_raid_ndisks(sc, -1) == 0)
2638			g_raid_destroy_node(sc, 0);
2639		else
2640			g_raid_md_ddf_refill(sc);
2641		return (error);
2642	}
2643	if (strcmp(verb, "insert") == 0) {
2644		if (*nargs < 2) {
2645			gctl_error(req, "Invalid number of arguments.");
2646			return (-1);
2647		}
2648		for (i = 1; i < *nargs; i++) {
2649			/* Get disk name. */
2650			snprintf(arg, sizeof(arg), "arg%d", i);
2651			diskname = gctl_get_asciiparam(req, arg);
2652			if (diskname == NULL) {
2653				gctl_error(req, "No disk name (%s).", arg);
2654				error = -3;
2655				break;
2656			}
2657
2658			/* Try to find provider with specified name. */
2659			g_topology_lock();
2660			cp = g_raid_open_consumer(sc, diskname);
2661			if (cp == NULL) {
2662				gctl_error(req, "Can't open disk '%s'.",
2663				    diskname);
2664				g_topology_unlock();
2665				error = -4;
2666				break;
2667			}
2668			pp = cp->provider;
2669			g_topology_unlock();
2670
2671			pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2672
2673			disk = g_raid_create_disk(sc);
2674			disk->d_consumer = cp;
2675			disk->d_md_data = (void *)pd;
2676			cp->private = disk;
2677
2678			/* Read kernel dumping information. */
2679			disk->d_kd.offset = 0;
2680			disk->d_kd.length = OFF_MAX;
2681			len = sizeof(disk->d_kd);
2682			g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
2683			if (disk->d_kd.di.dumper == NULL)
2684				G_RAID_DEBUG1(2, sc,
2685				    "Dumping not supported by %s.",
2686				    cp->provider->name);
2687
2688			/* Welcome the "new" disk. */
2689			g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
2690			ddf_meta_create(disk, &mdi->mdio_meta);
2691			sa = ddf_meta_find_sa(&pd->pd_meta, 1);
2692			if (sa != NULL) {
2693				SET32D(&pd->pd_meta, sa->Signature,
2694				    DDF_SA_SIGNATURE);
2695				SET8D(&pd->pd_meta, sa->Spare_Type, 0);
2696				SET16D(&pd->pd_meta, sa->Populated_SAEs, 0);
2697				SET16D(&pd->pd_meta, sa->MAX_SAE_Supported,
2698				    (GET16(&pd->pd_meta, hdr->Configuration_Record_Length) *
2699				     pd->pd_meta.sectorsize -
2700				     sizeof(struct ddf_sa_record)) /
2701				    sizeof(struct ddf_sa_entry));
2702			}
2703			if (mdi->mdio_meta.hdr == NULL)
2704				ddf_meta_copy(&mdi->mdio_meta, &pd->pd_meta);
2705			else
2706				ddf_meta_update(&mdi->mdio_meta, &pd->pd_meta);
2707			g_raid_md_write_ddf(md, NULL, NULL, NULL);
2708			g_raid_md_ddf_refill(sc);
2709		}
2710		return (error);
2711	}
2712	return (-100);
2713}
2714
2715static int
2716g_raid_md_write_ddf(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2717    struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2718{
2719	struct g_raid_softc *sc;
2720	struct g_raid_volume *vol;
2721	struct g_raid_subdisk *sd;
2722	struct g_raid_disk *disk;
2723	struct g_raid_md_ddf_perdisk *pd;
2724	struct g_raid_md_ddf_pervolume *pv;
2725	struct g_raid_md_ddf_object *mdi;
2726	struct ddf_meta *gmeta;
2727	struct ddf_vol_meta *vmeta;
2728	struct ddf_vdc_record *vdc;
2729	struct ddf_sa_record *sa;
2730	uint64_t *val2;
2731	int i, j, pos, bvd, size;
2732
2733	sc = md->mdo_softc;
2734	mdi = (struct g_raid_md_ddf_object *)md;
2735	gmeta = &mdi->mdio_meta;
2736
2737	if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2738		return (0);
2739
2740	/*
2741	 * Clear disk flags to let only really needed ones to be reset.
2742	 * Do it only if there are no volumes in starting state now,
2743	 * as they can update disk statuses yet and we may kill innocent.
2744	 */
2745	if (mdi->mdio_starting == 0) {
2746		for (i = 0; i < GET16(gmeta, pdr->Populated_PDEs); i++) {
2747			if (isff(gmeta->pdr->entry[i].PD_GUID, 24))
2748				continue;
2749			SET16(gmeta, pdr->entry[i].PD_Type,
2750			    GET16(gmeta, pdr->entry[i].PD_Type) &
2751			    ~(DDF_PDE_PARTICIPATING |
2752			      DDF_PDE_GLOBAL_SPARE | DDF_PDE_CONFIG_SPARE));
2753			if ((GET16(gmeta, pdr->entry[i].PD_State) &
2754			    DDF_PDE_PFA) == 0)
2755				SET16(gmeta, pdr->entry[i].PD_State, 0);
2756		}
2757	}
2758
2759	/* Generate/update new per-volume metadata. */
2760	TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2761		pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2762		if (vol->v_stopping || !pv->pv_started)
2763			continue;
2764		vmeta = &pv->pv_meta;
2765
2766		SET32(vmeta, vdc->Sequence_Number,
2767		    GET32(vmeta, vdc->Sequence_Number) + 1);
2768		if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E &&
2769		    vol->v_disks_count % 2 == 0)
2770			SET16(vmeta, vdc->Primary_Element_Count, 2);
2771		else
2772			SET16(vmeta, vdc->Primary_Element_Count,
2773			    vol->v_disks_count);
2774		SET8(vmeta, vdc->Stripe_Size,
2775		    ffs(vol->v_strip_size / vol->v_sectorsize) - 1);
2776		if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E &&
2777		    vol->v_disks_count % 2 == 0) {
2778			SET8(vmeta, vdc->Primary_RAID_Level,
2779			    DDF_VDCR_RAID1);
2780			SET8(vmeta, vdc->RLQ, 0);
2781			SET8(vmeta, vdc->Secondary_Element_Count,
2782			    vol->v_disks_count / 2);
2783			SET8(vmeta, vdc->Secondary_RAID_Level, 0);
2784		} else {
2785			SET8(vmeta, vdc->Primary_RAID_Level,
2786			    vol->v_raid_level);
2787			SET8(vmeta, vdc->RLQ,
2788			    vol->v_raid_level_qualifier);
2789			SET8(vmeta, vdc->Secondary_Element_Count, 1);
2790			SET8(vmeta, vdc->Secondary_RAID_Level, 0);
2791		}
2792		SET8(vmeta, vdc->Secondary_Element_Seq, 0);
2793		SET64(vmeta, vdc->Block_Count, 0);
2794		SET64(vmeta, vdc->VD_Size, vol->v_mediasize / vol->v_sectorsize);
2795		SET16(vmeta, vdc->Block_Size, vol->v_sectorsize);
2796		SET8(vmeta, vdc->Rotate_Parity_count,
2797		    fls(vol->v_rotate_parity) - 1);
2798		SET8(vmeta, vdc->MDF_Parity_Disks, vol->v_mdf_pdisks);
2799		SET16(vmeta, vdc->MDF_Parity_Generator_Polynomial,
2800		    vol->v_mdf_polynomial);
2801		SET8(vmeta, vdc->MDF_Constant_Generation_Method,
2802		    vol->v_mdf_method);
2803
2804		SET16(vmeta, vde->VD_Number, vol->v_global_id);
2805		if (vol->v_state <= G_RAID_VOLUME_S_BROKEN)
2806			SET8(vmeta, vde->VD_State, DDF_VDE_FAILED);
2807		else if (vol->v_state <= G_RAID_VOLUME_S_DEGRADED)
2808			SET8(vmeta, vde->VD_State, DDF_VDE_DEGRADED);
2809		else if (vol->v_state <= G_RAID_VOLUME_S_SUBOPTIMAL)
2810			SET8(vmeta, vde->VD_State, DDF_VDE_PARTIAL);
2811		else
2812			SET8(vmeta, vde->VD_State, DDF_VDE_OPTIMAL);
2813		if (vol->v_dirty ||
2814		    g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_STALE) > 0 ||
2815		    g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_RESYNC) > 0)
2816			SET8(vmeta, vde->VD_State,
2817			    GET8(vmeta, vde->VD_State) | DDF_VDE_DIRTY);
2818		SET8(vmeta, vde->Init_State, DDF_VDE_INIT_FULL); // XXX
2819		ddf_meta_put_name(vmeta, vol->v_name);
2820
2821		for (i = 0; i < vol->v_disks_count; i++) {
2822			sd = &vol->v_subdisks[i];
2823			bvd = i / GET16(vmeta, vdc->Primary_Element_Count);
2824			pos = i % GET16(vmeta, vdc->Primary_Element_Count);
2825			disk = sd->sd_disk;
2826			if (disk != NULL) {
2827				pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2828				if (vmeta->bvdc[bvd] == NULL) {
2829					size = GET16(vmeta,
2830					    hdr->Configuration_Record_Length) *
2831					    vmeta->sectorsize;
2832					vmeta->bvdc[bvd] = malloc(size,
2833					    M_MD_DDF, M_WAITOK);
2834					memset(vmeta->bvdc[bvd], 0xff, size);
2835				}
2836				memcpy(vmeta->bvdc[bvd], vmeta->vdc,
2837				    sizeof(struct ddf_vdc_record));
2838				SET8(vmeta, bvdc[bvd]->Secondary_Element_Seq, bvd);
2839				SET64(vmeta, bvdc[bvd]->Block_Count,
2840				    sd->sd_size / vol->v_sectorsize);
2841				SET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos],
2842				    GET32(&pd->pd_meta, pdd->PD_Reference));
2843				val2 = (uint64_t *)&(vmeta->bvdc[bvd]->Physical_Disk_Sequence[
2844				    GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
2845				SET64P(vmeta, val2 + pos,
2846				    sd->sd_offset / vol->v_sectorsize);
2847			}
2848			if (vmeta->bvdc[bvd] == NULL)
2849				continue;
2850
2851			j = ddf_meta_find_pd(gmeta, NULL,
2852			    GET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos]));
2853			if (j < 0)
2854				continue;
2855			SET32(gmeta, pdr->entry[j].PD_Type,
2856			    GET32(gmeta, pdr->entry[j].PD_Type) |
2857			    DDF_PDE_PARTICIPATING);
2858			if (sd->sd_state == G_RAID_SUBDISK_S_NONE)
2859				SET32(gmeta, pdr->entry[j].PD_State,
2860				    GET32(gmeta, pdr->entry[j].PD_State) |
2861				    (DDF_PDE_FAILED | DDF_PDE_MISSING));
2862			else if (sd->sd_state == G_RAID_SUBDISK_S_FAILED)
2863				SET32(gmeta, pdr->entry[j].PD_State,
2864				    GET32(gmeta, pdr->entry[j].PD_State) |
2865				    (DDF_PDE_FAILED | DDF_PDE_PFA));
2866			else if (sd->sd_state <= G_RAID_SUBDISK_S_REBUILD)
2867				SET32(gmeta, pdr->entry[j].PD_State,
2868				    GET32(gmeta, pdr->entry[j].PD_State) |
2869				    DDF_PDE_REBUILD);
2870			else
2871				SET32(gmeta, pdr->entry[j].PD_State,
2872				    GET32(gmeta, pdr->entry[j].PD_State) |
2873				    DDF_PDE_ONLINE);
2874		}
2875	}
2876
2877	/* Mark spare and failed disks as such. */
2878	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2879		pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2880		i = ddf_meta_find_pd(gmeta, NULL,
2881		    GET32(&pd->pd_meta, pdd->PD_Reference));
2882		if (i < 0)
2883			continue;
2884		if (disk->d_state == G_RAID_DISK_S_FAILED) {
2885			SET32(gmeta, pdr->entry[i].PD_State,
2886			    GET32(gmeta, pdr->entry[i].PD_State) |
2887			    (DDF_PDE_FAILED | DDF_PDE_PFA));
2888		}
2889		if (disk->d_state != G_RAID_DISK_S_SPARE)
2890			continue;
2891		sa = ddf_meta_find_sa(&pd->pd_meta, 0);
2892		if (sa == NULL ||
2893		    (GET8D(&pd->pd_meta, sa->Spare_Type) &
2894		     DDF_SAR_TYPE_DEDICATED) == 0) {
2895			SET16(gmeta, pdr->entry[i].PD_Type,
2896			    GET16(gmeta, pdr->entry[i].PD_Type) |
2897			    DDF_PDE_GLOBAL_SPARE);
2898		} else {
2899			SET16(gmeta, pdr->entry[i].PD_Type,
2900			    GET16(gmeta, pdr->entry[i].PD_Type) |
2901			    DDF_PDE_CONFIG_SPARE);
2902		}
2903		SET32(gmeta, pdr->entry[i].PD_State,
2904		    GET32(gmeta, pdr->entry[i].PD_State) |
2905		    DDF_PDE_ONLINE);
2906	}
2907
2908	/* Remove disks without "participating" flag (unused). */
2909	for (i = 0, j = -1; i < GET16(gmeta, pdr->Populated_PDEs); i++) {
2910		if (isff(gmeta->pdr->entry[i].PD_GUID, 24))
2911			continue;
2912		if ((GET16(gmeta, pdr->entry[i].PD_Type) &
2913		    (DDF_PDE_PARTICIPATING |
2914		     DDF_PDE_GLOBAL_SPARE | DDF_PDE_CONFIG_SPARE)) != 0 ||
2915		    g_raid_md_ddf_get_disk(sc,
2916		     NULL, GET32(gmeta, pdr->entry[i].PD_Reference)) != NULL)
2917			j = i;
2918		else
2919			memset(&gmeta->pdr->entry[i], 0xff,
2920			    sizeof(struct ddf_pd_entry));
2921	}
2922	SET16(gmeta, pdr->Populated_PDEs, j + 1);
2923
2924	/* Update per-disk metadata and write them. */
2925	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2926		pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2927		if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
2928		    disk->d_state != G_RAID_DISK_S_SPARE)
2929			continue;
2930		/* Update PDR. */
2931		memcpy(pd->pd_meta.pdr, gmeta->pdr,
2932		    GET32(&pd->pd_meta, hdr->pdr_length) *
2933		    pd->pd_meta.sectorsize);
2934		/* Update VDR. */
2935		SET16(&pd->pd_meta, vdr->Populated_VDEs, 0);
2936		TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2937			if (vol->v_stopping)
2938				continue;
2939			pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2940			i = ddf_meta_find_vd(&pd->pd_meta,
2941			    pv->pv_meta.vde->VD_GUID);
2942			if (i < 0)
2943				i = ddf_meta_find_vd(&pd->pd_meta, NULL);
2944			if (i >= 0)
2945				memcpy(&pd->pd_meta.vdr->entry[i],
2946				    pv->pv_meta.vde,
2947				    sizeof(struct ddf_vd_entry));
2948		}
2949		/* Update VDC. */
2950		if (mdi->mdio_starting == 0) {
2951			/* Remove all VDCs to restore needed later. */
2952			j = GETCRNUM(&pd->pd_meta);
2953			for (i = 0; i < j; i++) {
2954				vdc = GETVDCPTR(&pd->pd_meta, i);
2955				if (GET32D(&pd->pd_meta, vdc->Signature) !=
2956				    DDF_VDCR_SIGNATURE)
2957					continue;
2958				SET32D(&pd->pd_meta, vdc->Signature, 0xffffffff);
2959			}
2960		}
2961		TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2962			vol = sd->sd_volume;
2963			if (vol->v_stopping)
2964				continue;
2965			pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2966			vmeta = &pv->pv_meta;
2967			vdc = ddf_meta_find_vdc(&pd->pd_meta,
2968			    vmeta->vde->VD_GUID);
2969			if (vdc == NULL)
2970				vdc = ddf_meta_find_vdc(&pd->pd_meta, NULL);
2971			if (vdc != NULL) {
2972				bvd = sd->sd_pos / GET16(vmeta,
2973				    vdc->Primary_Element_Count);
2974				memcpy(vdc, vmeta->bvdc[bvd],
2975				    GET16(&pd->pd_meta,
2976				    hdr->Configuration_Record_Length) *
2977				    pd->pd_meta.sectorsize);
2978			}
2979		}
2980		G_RAID_DEBUG(1, "Writing DDF metadata to %s",
2981		    g_raid_get_diskname(disk));
2982		g_raid_md_ddf_print(&pd->pd_meta);
2983		ddf_meta_write(disk->d_consumer, &pd->pd_meta);
2984	}
2985	return (0);
2986}
2987
2988static int
2989g_raid_md_fail_disk_ddf(struct g_raid_md_object *md,
2990    struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2991{
2992	struct g_raid_softc *sc;
2993	struct g_raid_md_ddf_perdisk *pd;
2994	struct g_raid_subdisk *sd;
2995	int i;
2996
2997	sc = md->mdo_softc;
2998	pd = (struct g_raid_md_ddf_perdisk *)tdisk->d_md_data;
2999
3000	/* We can't fail disk that is not a part of array now. */
3001	if (tdisk->d_state != G_RAID_DISK_S_ACTIVE)
3002		return (-1);
3003
3004	/*
3005	 * Mark disk as failed in metadata and try to write that metadata
3006	 * to the disk itself to prevent it's later resurrection as STALE.
3007	 */
3008	G_RAID_DEBUG(1, "Writing DDF metadata to %s",
3009	    g_raid_get_diskname(tdisk));
3010	i = ddf_meta_find_pd(&pd->pd_meta, NULL, GET32(&pd->pd_meta, pdd->PD_Reference));
3011	SET16(&pd->pd_meta, pdr->entry[i].PD_State, DDF_PDE_FAILED | DDF_PDE_PFA);
3012	if (tdisk->d_consumer != NULL)
3013		ddf_meta_write(tdisk->d_consumer, &pd->pd_meta);
3014
3015	/* Change states. */
3016	g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
3017	TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
3018		g_raid_change_subdisk_state(sd,
3019		    G_RAID_SUBDISK_S_FAILED);
3020		g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
3021		    G_RAID_EVENT_SUBDISK);
3022	}
3023
3024	/* Write updated metadata to remaining disks. */
3025	g_raid_md_write_ddf(md, NULL, NULL, tdisk);
3026
3027	g_raid_md_ddf_refill(sc);
3028	return (0);
3029}
3030
3031static int
3032g_raid_md_free_disk_ddf(struct g_raid_md_object *md,
3033    struct g_raid_disk *disk)
3034{
3035	struct g_raid_md_ddf_perdisk *pd;
3036
3037	pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
3038	ddf_meta_free(&pd->pd_meta);
3039	free(pd, M_MD_DDF);
3040	disk->d_md_data = NULL;
3041	return (0);
3042}
3043
3044static int
3045g_raid_md_free_volume_ddf(struct g_raid_md_object *md,
3046    struct g_raid_volume *vol)
3047{
3048	struct g_raid_md_ddf_object *mdi;
3049	struct g_raid_md_ddf_pervolume *pv;
3050
3051	mdi = (struct g_raid_md_ddf_object *)md;
3052	pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
3053	ddf_vol_meta_free(&pv->pv_meta);
3054	if (!pv->pv_started) {
3055		pv->pv_started = 1;
3056		mdi->mdio_starting--;
3057		callout_stop(&pv->pv_start_co);
3058	}
3059	free(pv, M_MD_DDF);
3060	vol->v_md_data = NULL;
3061	return (0);
3062}
3063
3064static int
3065g_raid_md_free_ddf(struct g_raid_md_object *md)
3066{
3067	struct g_raid_md_ddf_object *mdi;
3068
3069	mdi = (struct g_raid_md_ddf_object *)md;
3070	if (!mdi->mdio_started) {
3071		mdi->mdio_started = 0;
3072		callout_stop(&mdi->mdio_start_co);
3073		G_RAID_DEBUG1(1, md->mdo_softc,
3074		    "root_mount_rel %p", mdi->mdio_rootmount);
3075		root_mount_rel(mdi->mdio_rootmount);
3076		mdi->mdio_rootmount = NULL;
3077	}
3078	ddf_meta_free(&mdi->mdio_meta);
3079	return (0);
3080}
3081
3082G_RAID_MD_DECLARE(ddf, "DDF");
3083