]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/geom/raid/md_ddf.c
zfs: merge openzfs/zfs@03e9caaec
[FreeBSD/FreeBSD.git] / sys / geom / raid / md_ddf.c
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2012 Alexander Motin <mav@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28
29 #include <sys/cdefs.h>
30 #include <sys/param.h>
31 #include <sys/bio.h>
32 #include <sys/gsb_crc32.h>
33 #include <sys/endian.h>
34 #include <sys/kernel.h>
35 #include <sys/kobj.h>
36 #include <sys/limits.h>
37 #include <sys/lock.h>
38 #include <sys/malloc.h>
39 #include <sys/mutex.h>
40 #include <sys/systm.h>
41 #include <sys/time.h>
42 #include <sys/clock.h>
43 #include <sys/disk.h>
44 #include <geom/geom.h>
45 #include <geom/geom_dbg.h>
46 #include "geom/raid/g_raid.h"
47 #include "geom/raid/md_ddf.h"
48 #include "g_raid_md_if.h"
49
50 static MALLOC_DEFINE(M_MD_DDF, "md_ddf_data", "GEOM_RAID DDF metadata");
51
52 #define DDF_MAX_DISKS_HARD      128
53
54 #define DDF_MAX_DISKS   16
55 #define DDF_MAX_VDISKS  7
56 #define DDF_MAX_PARTITIONS      1
57
58 #define DECADE (3600*24*(365*10+2))     /* 10 years in seconds. */
59
60 struct ddf_meta {
61         u_int   sectorsize;
62         u_int   bigendian;
63         struct ddf_header *hdr;
64         struct ddf_cd_record *cdr;
65         struct ddf_pd_record *pdr;
66         struct ddf_vd_record *vdr;
67         void *cr;
68         struct ddf_pdd_record *pdd;
69         struct ddf_bbm_log *bbm;
70 };
71
72 struct ddf_vol_meta {
73         u_int   sectorsize;
74         u_int   bigendian;
75         struct ddf_header *hdr;
76         struct ddf_cd_record *cdr;
77         struct ddf_vd_entry *vde;
78         struct ddf_vdc_record *vdc;
79         struct ddf_vdc_record *bvdc[DDF_MAX_DISKS_HARD];
80 };
81
82 struct g_raid_md_ddf_perdisk {
83         struct ddf_meta  pd_meta;
84 };
85
86 struct g_raid_md_ddf_pervolume {
87         struct ddf_vol_meta              pv_meta;
88         int                              pv_started;
89         struct callout                   pv_start_co;   /* STARTING state timer. */
90 };
91
92 struct g_raid_md_ddf_object {
93         struct g_raid_md_object  mdio_base;
94         u_int                    mdio_bigendian;
95         struct ddf_meta          mdio_meta;
96         int                      mdio_starting;
97         struct callout           mdio_start_co; /* STARTING state timer. */
98         int                      mdio_started;
99         struct root_hold_token  *mdio_rootmount; /* Root mount delay token. */
100 };
101
102 static g_raid_md_create_req_t g_raid_md_create_req_ddf;
103 static g_raid_md_taste_t g_raid_md_taste_ddf;
104 static g_raid_md_event_t g_raid_md_event_ddf;
105 static g_raid_md_volume_event_t g_raid_md_volume_event_ddf;
106 static g_raid_md_ctl_t g_raid_md_ctl_ddf;
107 static g_raid_md_write_t g_raid_md_write_ddf;
108 static g_raid_md_fail_disk_t g_raid_md_fail_disk_ddf;
109 static g_raid_md_free_disk_t g_raid_md_free_disk_ddf;
110 static g_raid_md_free_volume_t g_raid_md_free_volume_ddf;
111 static g_raid_md_free_t g_raid_md_free_ddf;
112
113 static kobj_method_t g_raid_md_ddf_methods[] = {
114         KOBJMETHOD(g_raid_md_create_req,        g_raid_md_create_req_ddf),
115         KOBJMETHOD(g_raid_md_taste,     g_raid_md_taste_ddf),
116         KOBJMETHOD(g_raid_md_event,     g_raid_md_event_ddf),
117         KOBJMETHOD(g_raid_md_volume_event,      g_raid_md_volume_event_ddf),
118         KOBJMETHOD(g_raid_md_ctl,       g_raid_md_ctl_ddf),
119         KOBJMETHOD(g_raid_md_write,     g_raid_md_write_ddf),
120         KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_ddf),
121         KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_ddf),
122         KOBJMETHOD(g_raid_md_free_volume,       g_raid_md_free_volume_ddf),
123         KOBJMETHOD(g_raid_md_free,      g_raid_md_free_ddf),
124         { 0, 0 }
125 };
126
127 static struct g_raid_md_class g_raid_md_ddf_class = {
128         "DDF",
129         g_raid_md_ddf_methods,
130         sizeof(struct g_raid_md_ddf_object),
131         .mdc_enable = 1,
132         .mdc_priority = 100
133 };
134
135 #define GET8(m, f)      ((m)->f)
136 #define GET16(m, f)     ((m)->bigendian ? be16dec(&(m)->f) : le16dec(&(m)->f))
137 #define GET32(m, f)     ((m)->bigendian ? be32dec(&(m)->f) : le32dec(&(m)->f))
138 #define GET64(m, f)     ((m)->bigendian ? be64dec(&(m)->f) : le64dec(&(m)->f))
139 #define GET8D(m, f)     (f)
140 #define GET16D(m, f)    ((m)->bigendian ? be16dec(&f) : le16dec(&f))
141 #define GET32D(m, f)    ((m)->bigendian ? be32dec(&f) : le32dec(&f))
142 #define GET64D(m, f)    ((m)->bigendian ? be64dec(&f) : le64dec(&f))
143 #define GET8P(m, f)     (*(f))
144 #define GET16P(m, f)    ((m)->bigendian ? be16dec(f) : le16dec(f))
145 #define GET32P(m, f)    ((m)->bigendian ? be32dec(f) : le32dec(f))
146 #define GET64P(m, f)    ((m)->bigendian ? be64dec(f) : le64dec(f))
147
148 #define SET8P(m, f, v)                                                  \
149         (*(f) = (v))
150 #define SET16P(m, f, v)                                                 \
151         do {                                                            \
152                 if ((m)->bigendian)                                     \
153                         be16enc((f), (v));                              \
154                 else                                                    \
155                         le16enc((f), (v));                              \
156         } while (0)
157 #define SET32P(m, f, v)                                                 \
158         do {                                                            \
159                 if ((m)->bigendian)                                     \
160                         be32enc((f), (v));                              \
161                 else                                                    \
162                         le32enc((f), (v));                              \
163         } while (0)
164 #define SET64P(m, f, v)                                                 \
165         do {                                                            \
166                 if ((m)->bigendian)                                     \
167                         be64enc((f), (v));                              \
168                 else                                                    \
169                         le64enc((f), (v));                              \
170         } while (0)
171 #define SET8(m, f, v)   SET8P((m), &((m)->f), (v))
172 #define SET16(m, f, v)  SET16P((m), &((m)->f), (v))
173 #define SET32(m, f, v)  SET32P((m), &((m)->f), (v))
174 #define SET64(m, f, v)  SET64P((m), &((m)->f), (v))
175 #define SET8D(m, f, v)  SET8P((m), &(f), (v))
176 #define SET16D(m, f, v) SET16P((m), &(f), (v))
177 #define SET32D(m, f, v) SET32P((m), &(f), (v))
178 #define SET64D(m, f, v) SET64P((m), &(f), (v))
179
180 #define GETCRNUM(m)     (GET32((m), hdr->cr_length) /                   \
181         GET16((m), hdr->Configuration_Record_Length))
182
183 #define GETVDCPTR(m, n) ((struct ddf_vdc_record *)((uint8_t *)(m)->cr + \
184         (n) * GET16((m), hdr->Configuration_Record_Length) *            \
185         (m)->sectorsize))
186
187 #define GETSAPTR(m, n)  ((struct ddf_sa_record *)((uint8_t *)(m)->cr +  \
188         (n) * GET16((m), hdr->Configuration_Record_Length) *            \
189         (m)->sectorsize))
190
191 static int
192 isff(uint8_t *buf, int size)
193 {
194         int i;
195
196         for (i = 0; i < size; i++)
197                 if (buf[i] != 0xff)
198                         return (0);
199         return (1);
200 }
201
202 static void
203 print_guid(uint8_t *buf)
204 {
205         int i, ascii;
206
207         ascii = 1;
208         for (i = 0; i < 24; i++) {
209                 if (buf[i] != 0 && (buf[i] < ' ' || buf[i] > 127)) {
210                         ascii = 0;
211                         break;
212                 }
213         }
214         if (ascii) {
215                 printf("'%.24s'", buf);
216         } else {
217                 for (i = 0; i < 24; i++)
218                         printf("%02x", buf[i]);
219         }
220 }
221
222 static void
223 g_raid_md_ddf_print(struct ddf_meta *meta)
224 {
225         struct ddf_vdc_record *vdc;
226         struct ddf_vuc_record *vuc;
227         struct ddf_sa_record *sa;
228         uint64_t *val2;
229         uint32_t val;
230         int i, j, k, num, num2;
231
232         if (g_raid_debug < 1)
233                 return;
234
235         printf("********* DDF Metadata *********\n");
236         printf("**** Header ****\n");
237         printf("DDF_Header_GUID      ");
238         print_guid(meta->hdr->DDF_Header_GUID);
239         printf("\n");
240         printf("DDF_rev              %8.8s\n", (char *)&meta->hdr->DDF_rev[0]);
241         printf("Sequence_Number      0x%08x\n", GET32(meta, hdr->Sequence_Number));
242         printf("TimeStamp            0x%08x\n", GET32(meta, hdr->TimeStamp));
243         printf("Open_Flag            0x%02x\n", GET16(meta, hdr->Open_Flag));
244         printf("Foreign_Flag         0x%02x\n", GET16(meta, hdr->Foreign_Flag));
245         printf("Diskgrouping         0x%02x\n", GET16(meta, hdr->Diskgrouping));
246         printf("Primary_Header_LBA   %ju\n", GET64(meta, hdr->Primary_Header_LBA));
247         printf("Secondary_Header_LBA %ju\n", GET64(meta, hdr->Secondary_Header_LBA));
248         printf("WorkSpace_Length     %u\n", GET32(meta, hdr->WorkSpace_Length));
249         printf("WorkSpace_LBA        %ju\n", GET64(meta, hdr->WorkSpace_LBA));
250         printf("Max_PD_Entries       %u\n", GET16(meta, hdr->Max_PD_Entries));
251         printf("Max_VD_Entries       %u\n", GET16(meta, hdr->Max_VD_Entries));
252         printf("Max_Partitions       %u\n", GET16(meta, hdr->Max_Partitions));
253         printf("Configuration_Record_Length %u\n", GET16(meta, hdr->Configuration_Record_Length));
254         printf("Max_Primary_Element_Entries %u\n", GET16(meta, hdr->Max_Primary_Element_Entries));
255         printf("Controller Data      %u:%u\n", GET32(meta, hdr->cd_section), GET32(meta, hdr->cd_length));
256         printf("Physical Disk        %u:%u\n", GET32(meta, hdr->pdr_section), GET32(meta, hdr->pdr_length));
257         printf("Virtual Disk         %u:%u\n", GET32(meta, hdr->vdr_section), GET32(meta, hdr->vdr_length));
258         printf("Configuration Recs   %u:%u\n", GET32(meta, hdr->cr_section), GET32(meta, hdr->cr_length));
259         printf("Physical Disk Recs   %u:%u\n", GET32(meta, hdr->pdd_section), GET32(meta, hdr->pdd_length));
260         printf("BBM Log              %u:%u\n", GET32(meta, hdr->bbmlog_section), GET32(meta, hdr->bbmlog_length));
261         printf("Diagnostic Space     %u:%u\n", GET32(meta, hdr->Diagnostic_Space), GET32(meta, hdr->Diagnostic_Space_Length));
262         printf("Vendor_Specific_Logs %u:%u\n", GET32(meta, hdr->Vendor_Specific_Logs), GET32(meta, hdr->Vendor_Specific_Logs_Length));
263         printf("**** Controller Data ****\n");
264         printf("Controller_GUID      ");
265         print_guid(meta->cdr->Controller_GUID);
266         printf("\n");
267         printf("Controller_Type      0x%04x%04x 0x%04x%04x\n",
268             GET16(meta, cdr->Controller_Type.Vendor_ID),
269             GET16(meta, cdr->Controller_Type.Device_ID),
270             GET16(meta, cdr->Controller_Type.SubVendor_ID),
271             GET16(meta, cdr->Controller_Type.SubDevice_ID));
272         printf("Product_ID           '%.16s'\n", (char *)&meta->cdr->Product_ID[0]);
273         printf("**** Physical Disk Records ****\n");
274         printf("Populated_PDEs       %u\n", GET16(meta, pdr->Populated_PDEs));
275         printf("Max_PDE_Supported    %u\n", GET16(meta, pdr->Max_PDE_Supported));
276         for (j = 0; j < GET16(meta, pdr->Populated_PDEs); j++) {
277                 if (isff(meta->pdr->entry[j].PD_GUID, 24))
278                         continue;
279                 if (GET32(meta, pdr->entry[j].PD_Reference) == 0xffffffff)
280                         continue;
281                 printf("PD_GUID              ");
282                 print_guid(meta->pdr->entry[j].PD_GUID);
283                 printf("\n");
284                 printf("PD_Reference         0x%08x\n",
285                     GET32(meta, pdr->entry[j].PD_Reference));
286                 printf("PD_Type              0x%04x\n",
287                     GET16(meta, pdr->entry[j].PD_Type));
288                 printf("PD_State             0x%04x\n",
289                     GET16(meta, pdr->entry[j].PD_State));
290                 printf("Configured_Size      %ju\n",
291                     GET64(meta, pdr->entry[j].Configured_Size));
292                 printf("Block_Size           %u\n",
293                     GET16(meta, pdr->entry[j].Block_Size));
294         }
295         printf("**** Virtual Disk Records ****\n");
296         printf("Populated_VDEs       %u\n", GET16(meta, vdr->Populated_VDEs));
297         printf("Max_VDE_Supported    %u\n", GET16(meta, vdr->Max_VDE_Supported));
298         for (j = 0; j < GET16(meta, vdr->Populated_VDEs); j++) {
299                 if (isff(meta->vdr->entry[j].VD_GUID, 24))
300                         continue;
301                 printf("VD_GUID              ");
302                 print_guid(meta->vdr->entry[j].VD_GUID);
303                 printf("\n");
304                 printf("VD_Number            0x%04x\n",
305                     GET16(meta, vdr->entry[j].VD_Number));
306                 printf("VD_Type              0x%04x\n",
307                     GET16(meta, vdr->entry[j].VD_Type));
308                 printf("VD_State             0x%02x\n",
309                     GET8(meta, vdr->entry[j].VD_State));
310                 printf("Init_State           0x%02x\n",
311                     GET8(meta, vdr->entry[j].Init_State));
312                 printf("Drive_Failures_Remaining %u\n",
313                     GET8(meta, vdr->entry[j].Drive_Failures_Remaining));
314                 printf("VD_Name              '%.16s'\n",
315                     (char *)&meta->vdr->entry[j].VD_Name);
316         }
317         printf("**** Configuration Records ****\n");
318         num = GETCRNUM(meta);
319         for (j = 0; j < num; j++) {
320                 vdc = GETVDCPTR(meta, j);
321                 val = GET32D(meta, vdc->Signature);
322                 switch (val) {
323                 case DDF_VDCR_SIGNATURE:
324                         printf("** Virtual Disk Configuration **\n");
325                         printf("VD_GUID              ");
326                         print_guid(vdc->VD_GUID);
327                         printf("\n");
328                         printf("Timestamp            0x%08x\n",
329                             GET32D(meta, vdc->Timestamp));
330                         printf("Sequence_Number      0x%08x\n",
331                             GET32D(meta, vdc->Sequence_Number));
332                         printf("Primary_Element_Count %u\n",
333                             GET16D(meta, vdc->Primary_Element_Count));
334                         printf("Stripe_Size          %u\n",
335                             GET8D(meta, vdc->Stripe_Size));
336                         printf("Primary_RAID_Level   0x%02x\n",
337                             GET8D(meta, vdc->Primary_RAID_Level));
338                         printf("RLQ                  0x%02x\n",
339                             GET8D(meta, vdc->RLQ));
340                         printf("Secondary_Element_Count %u\n",
341                             GET8D(meta, vdc->Secondary_Element_Count));
342                         printf("Secondary_Element_Seq %u\n",
343                             GET8D(meta, vdc->Secondary_Element_Seq));
344                         printf("Secondary_RAID_Level 0x%02x\n",
345                             GET8D(meta, vdc->Secondary_RAID_Level));
346                         printf("Block_Count          %ju\n",
347                             GET64D(meta, vdc->Block_Count));
348                         printf("VD_Size              %ju\n",
349                             GET64D(meta, vdc->VD_Size));
350                         printf("Block_Size           %u\n",
351                             GET16D(meta, vdc->Block_Size));
352                         printf("Rotate_Parity_count  %u\n",
353                             GET8D(meta, vdc->Rotate_Parity_count));
354                         printf("Associated_Spare_Disks");
355                         for (i = 0; i < 8; i++) {
356                                 if (GET32D(meta, vdc->Associated_Spares[i]) != 0xffffffff)
357                                         printf(" 0x%08x", GET32D(meta, vdc->Associated_Spares[i]));
358                         }
359                         printf("\n");
360                         printf("Cache_Flags          %016jx\n",
361                             GET64D(meta, vdc->Cache_Flags));
362                         printf("BG_Rate              %u\n",
363                             GET8D(meta, vdc->BG_Rate));
364                         printf("MDF_Parity_Disks     %u\n",
365                             GET8D(meta, vdc->MDF_Parity_Disks));
366                         printf("MDF_Parity_Generator_Polynomial 0x%04x\n",
367                             GET16D(meta, vdc->MDF_Parity_Generator_Polynomial));
368                         printf("MDF_Constant_Generation_Method 0x%02x\n",
369                             GET8D(meta, vdc->MDF_Constant_Generation_Method));
370                         printf("Physical_Disks      ");
371                         num2 = GET16D(meta, vdc->Primary_Element_Count);
372                         val2 = (uint64_t *)&(vdc->Physical_Disk_Sequence[GET16(meta, hdr->Max_Primary_Element_Entries)]);
373                         for (i = 0; i < num2; i++)
374                                 printf(" 0x%08x @ %ju",
375                                     GET32D(meta, vdc->Physical_Disk_Sequence[i]),
376                                     GET64P(meta, val2 + i));
377                         printf("\n");
378                         break;
379                 case DDF_VUCR_SIGNATURE:
380                         printf("** Vendor Unique Configuration **\n");
381                         vuc = (struct ddf_vuc_record *)vdc;
382                         printf("VD_GUID              ");
383                         print_guid(vuc->VD_GUID);
384                         printf("\n");
385                         break;
386                 case DDF_SA_SIGNATURE:
387                         printf("** Spare Assignment Configuration **\n");
388                         sa = (struct ddf_sa_record *)vdc;
389                         printf("Timestamp            0x%08x\n",
390                             GET32D(meta, sa->Timestamp));
391                         printf("Spare_Type           0x%02x\n",
392                             GET8D(meta, sa->Spare_Type));
393                         printf("Populated_SAEs       %u\n",
394                             GET16D(meta, sa->Populated_SAEs));
395                         printf("MAX_SAE_Supported    %u\n",
396                             GET16D(meta, sa->MAX_SAE_Supported));
397                         for (i = 0; i < GET16D(meta, sa->Populated_SAEs); i++) {
398                                 if (isff(sa->entry[i].VD_GUID, 24))
399                                         continue;
400                                 printf("VD_GUID             ");
401                                 for (k = 0; k < 24; k++)
402                                         printf("%02x", sa->entry[i].VD_GUID[k]);
403                                 printf("\n");
404                                 printf("Secondary_Element   %u\n",
405                                     GET16D(meta, sa->entry[i].Secondary_Element));
406                         }
407                         break;
408                 case 0x00000000:
409                 case 0xFFFFFFFF:
410                         break;
411                 default:
412                         printf("Unknown configuration signature %08x\n", val);
413                         break;
414                 }
415         }
416         printf("**** Physical Disk Data ****\n");
417         printf("PD_GUID              ");
418         print_guid(meta->pdd->PD_GUID);
419         printf("\n");
420         printf("PD_Reference         0x%08x\n",
421             GET32(meta, pdd->PD_Reference));
422         printf("Forced_Ref_Flag      0x%02x\n",
423             GET8(meta, pdd->Forced_Ref_Flag));
424         printf("Forced_PD_GUID_Flag  0x%02x\n",
425             GET8(meta, pdd->Forced_PD_GUID_Flag));
426 }
427
428 static int
429 ddf_meta_find_pd(struct ddf_meta *meta, uint8_t *GUID, uint32_t PD_Reference)
430 {
431         int i;
432
433         for (i = 0; i < GET16(meta, pdr->Populated_PDEs); i++) {
434                 if (GUID != NULL) {
435                         if (memcmp(meta->pdr->entry[i].PD_GUID, GUID, 24) == 0)
436                                 return (i);
437                 } else if (PD_Reference != 0xffffffff) {
438                         if (GET32(meta, pdr->entry[i].PD_Reference) == PD_Reference)
439                                 return (i);
440                 } else
441                         if (isff(meta->pdr->entry[i].PD_GUID, 24))
442                                 return (i);
443         }
444         if (GUID == NULL && PD_Reference == 0xffffffff) {
445                 if (i >= GET16(meta, pdr->Max_PDE_Supported))
446                         return (-1);
447                 SET16(meta, pdr->Populated_PDEs, i + 1);
448                 return (i);
449         }
450         return (-1);
451 }
452
453 static int
454 ddf_meta_find_vd(struct ddf_meta *meta, uint8_t *GUID)
455 {
456         int i;
457
458         for (i = 0; i < GET16(meta, vdr->Populated_VDEs); i++) {
459                 if (GUID != NULL) {
460                         if (memcmp(meta->vdr->entry[i].VD_GUID, GUID, 24) == 0)
461                                 return (i);
462                 } else
463                         if (isff(meta->vdr->entry[i].VD_GUID, 24))
464                                 return (i);
465         }
466         if (GUID == NULL) {
467                 if (i >= GET16(meta, vdr->Max_VDE_Supported))
468                         return (-1);
469                 SET16(meta, vdr->Populated_VDEs, i + 1);
470                 return (i);
471         }
472         return (-1);
473 }
474
475 static struct ddf_vdc_record *
476 ddf_meta_find_vdc(struct ddf_meta *meta, uint8_t *GUID)
477 {
478         struct ddf_vdc_record *vdc;
479         int i, num;
480
481         num = GETCRNUM(meta);
482         for (i = 0; i < num; i++) {
483                 vdc = GETVDCPTR(meta, i);
484                 if (GUID != NULL) {
485                         if (GET32D(meta, vdc->Signature) == DDF_VDCR_SIGNATURE &&
486                             memcmp(vdc->VD_GUID, GUID, 24) == 0)
487                                 return (vdc);
488                 } else
489                         if (GET32D(meta, vdc->Signature) == 0xffffffff ||
490                             GET32D(meta, vdc->Signature) == 0)
491                                 return (vdc);
492         }
493         return (NULL);
494 }
495
496 static int
497 ddf_meta_count_vdc(struct ddf_meta *meta, uint8_t *GUID)
498 {
499         struct ddf_vdc_record *vdc;
500         int i, num, cnt;
501
502         cnt = 0;
503         num = GETCRNUM(meta);
504         for (i = 0; i < num; i++) {
505                 vdc = GETVDCPTR(meta, i);
506                 if (GET32D(meta, vdc->Signature) != DDF_VDCR_SIGNATURE)
507                         continue;
508                 if (GUID == NULL || memcmp(vdc->VD_GUID, GUID, 24) == 0)
509                         cnt++;
510         }
511         return (cnt);
512 }
513
514 static int
515 ddf_meta_find_disk(struct ddf_vol_meta *vmeta, uint32_t PD_Reference,
516     int *bvdp, int *posp)
517 {
518         int i, bvd, pos;
519
520         i = 0;
521         for (bvd = 0; bvd < GET8(vmeta, vdc->Secondary_Element_Count); bvd++) {
522                 if (vmeta->bvdc[bvd] == NULL) {
523                         i += GET16(vmeta, vdc->Primary_Element_Count); // XXX
524                         continue;
525                 }
526                 for (pos = 0; pos < GET16(vmeta, bvdc[bvd]->Primary_Element_Count);
527                     pos++, i++) {
528                         if (GET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos]) ==
529                             PD_Reference) {
530                                 if (bvdp != NULL)
531                                         *bvdp = bvd;
532                                 if (posp != NULL)
533                                         *posp = pos;
534                                 return (i);
535                         }
536                 }
537         }
538         return (-1);
539 }
540
541 static struct ddf_sa_record *
542 ddf_meta_find_sa(struct ddf_meta *meta, int create)
543 {
544         struct ddf_sa_record *sa;
545         int i, num;
546
547         num = GETCRNUM(meta);
548         for (i = 0; i < num; i++) {
549                 sa = GETSAPTR(meta, i);
550                 if (GET32D(meta, sa->Signature) == DDF_SA_SIGNATURE)
551                         return (sa);
552         }
553         if (create) {
554                 for (i = 0; i < num; i++) {
555                         sa = GETSAPTR(meta, i);
556                         if (GET32D(meta, sa->Signature) == 0xffffffff ||
557                             GET32D(meta, sa->Signature) == 0)
558                                 return (sa);
559                 }
560         }
561         return (NULL);
562 }
563
564 static void
565 ddf_meta_create(struct g_raid_disk *disk, struct ddf_meta *sample)
566 {
567         struct timespec ts;
568         struct clocktime ct;
569         struct g_raid_md_ddf_perdisk *pd;
570         struct g_raid_md_ddf_object *mdi;
571         struct ddf_meta *meta;
572         struct ddf_pd_entry *pde;
573         off_t anchorlba;
574         u_int ss, pos, size;
575         int len, error;
576         char serial_buffer[DISK_IDENT_SIZE];
577
578         if (sample->hdr == NULL)
579                 sample = NULL;
580
581         mdi = (struct g_raid_md_ddf_object *)disk->d_softc->sc_md;
582         pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
583         meta = &pd->pd_meta;
584         ss = disk->d_consumer->provider->sectorsize;
585         anchorlba = disk->d_consumer->provider->mediasize / ss - 1;
586
587         meta->sectorsize = ss;
588         meta->bigendian = sample ? sample->bigendian : mdi->mdio_bigendian;
589         getnanotime(&ts);
590         clock_ts_to_ct(&ts, &ct);
591
592         /* Header */
593         meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
594         memset(meta->hdr, 0xff, ss);
595         if (sample) {
596                 memcpy(meta->hdr, sample->hdr, sizeof(struct ddf_header));
597                 if (ss != sample->sectorsize) {
598                         SET32(meta, hdr->WorkSpace_Length,
599                             howmany(GET32(sample, hdr->WorkSpace_Length) *
600                                 sample->sectorsize, ss));
601                         SET16(meta, hdr->Configuration_Record_Length,
602                             howmany(GET16(sample,
603                                 hdr->Configuration_Record_Length) *
604                                 sample->sectorsize, ss));
605                         SET32(meta, hdr->cd_length,
606                             howmany(GET32(sample, hdr->cd_length) *
607                                 sample->sectorsize, ss));
608                         SET32(meta, hdr->pdr_length,
609                             howmany(GET32(sample, hdr->pdr_length) *
610                                 sample->sectorsize, ss));
611                         SET32(meta, hdr->vdr_length,
612                             howmany(GET32(sample, hdr->vdr_length) *
613                                 sample->sectorsize, ss));
614                         SET32(meta, hdr->cr_length,
615                             howmany(GET32(sample, hdr->cr_length) *
616                                 sample->sectorsize, ss));
617                         SET32(meta, hdr->pdd_length,
618                             howmany(GET32(sample, hdr->pdd_length) *
619                                 sample->sectorsize, ss));
620                         SET32(meta, hdr->bbmlog_length,
621                             howmany(GET32(sample, hdr->bbmlog_length) *
622                                 sample->sectorsize, ss));
623                         SET32(meta, hdr->Diagnostic_Space,
624                             howmany(GET32(sample, hdr->bbmlog_length) *
625                                 sample->sectorsize, ss));
626                         SET32(meta, hdr->Vendor_Specific_Logs,
627                             howmany(GET32(sample, hdr->bbmlog_length) *
628                                 sample->sectorsize, ss));
629                 }
630         } else {
631                 SET32(meta, hdr->Signature, DDF_HEADER_SIGNATURE);
632                 snprintf(meta->hdr->DDF_Header_GUID, 25, "FreeBSD %08x%08x",
633                     (u_int)(ts.tv_sec - DECADE), arc4random());
634                 memcpy(meta->hdr->DDF_rev, "02.00.00", 8);
635                 SET32(meta, hdr->TimeStamp, (ts.tv_sec - DECADE));
636                 SET32(meta, hdr->WorkSpace_Length, 16 * 1024 * 1024 / ss);
637                 SET16(meta, hdr->Max_PD_Entries, DDF_MAX_DISKS - 1);
638                 SET16(meta, hdr->Max_VD_Entries, DDF_MAX_VDISKS);
639                 SET16(meta, hdr->Max_Partitions, DDF_MAX_PARTITIONS);
640                 SET16(meta, hdr->Max_Primary_Element_Entries, DDF_MAX_DISKS);
641                 SET16(meta, hdr->Configuration_Record_Length,
642                     howmany(sizeof(struct ddf_vdc_record) + (4 + 8) *
643                         GET16(meta, hdr->Max_Primary_Element_Entries), ss));
644                 SET32(meta, hdr->cd_length,
645                     howmany(sizeof(struct ddf_cd_record), ss));
646                 SET32(meta, hdr->pdr_length,
647                     howmany(sizeof(struct ddf_pd_record) +
648                         sizeof(struct ddf_pd_entry) * GET16(meta,
649                         hdr->Max_PD_Entries), ss));
650                 SET32(meta, hdr->vdr_length,
651                     howmany(sizeof(struct ddf_vd_record) +
652                         sizeof(struct ddf_vd_entry) *
653                         GET16(meta, hdr->Max_VD_Entries), ss));
654                 SET32(meta, hdr->cr_length,
655                     GET16(meta, hdr->Configuration_Record_Length) *
656                     (GET16(meta, hdr->Max_Partitions) + 1));
657                 SET32(meta, hdr->pdd_length,
658                     howmany(sizeof(struct ddf_pdd_record), ss));
659                 SET32(meta, hdr->bbmlog_length, 0);
660                 SET32(meta, hdr->Diagnostic_Space_Length, 0);
661                 SET32(meta, hdr->Vendor_Specific_Logs_Length, 0);
662         }
663         pos = 1;
664         SET32(meta, hdr->cd_section, pos);
665         pos += GET32(meta, hdr->cd_length);
666         SET32(meta, hdr->pdr_section, pos);
667         pos += GET32(meta, hdr->pdr_length);
668         SET32(meta, hdr->vdr_section, pos);
669         pos += GET32(meta, hdr->vdr_length);
670         SET32(meta, hdr->cr_section, pos);
671         pos += GET32(meta, hdr->cr_length);
672         SET32(meta, hdr->pdd_section, pos);
673         pos += GET32(meta, hdr->pdd_length);
674         SET32(meta, hdr->bbmlog_section,
675             GET32(meta, hdr->bbmlog_length) != 0 ? pos : 0xffffffff);
676         pos += GET32(meta, hdr->bbmlog_length);
677         SET32(meta, hdr->Diagnostic_Space,
678             GET32(meta, hdr->Diagnostic_Space_Length) != 0 ? pos : 0xffffffff);
679         pos += GET32(meta, hdr->Diagnostic_Space_Length);
680         SET32(meta, hdr->Vendor_Specific_Logs,
681             GET32(meta, hdr->Vendor_Specific_Logs_Length) != 0 ? pos : 0xffffffff);
682         pos += min(GET32(meta, hdr->Vendor_Specific_Logs_Length), 1);
683         SET64(meta, hdr->Primary_Header_LBA,
684             anchorlba - pos);
685         SET64(meta, hdr->Secondary_Header_LBA,
686             0xffffffffffffffffULL);
687         SET64(meta, hdr->WorkSpace_LBA,
688             anchorlba + 1 - 32 * 1024 * 1024 / ss);
689
690         /* Controller Data */
691         size = GET32(meta, hdr->cd_length) * ss;
692         meta->cdr = malloc(size, M_MD_DDF, M_WAITOK);
693         memset(meta->cdr, 0xff, size);
694         SET32(meta, cdr->Signature, DDF_CONTROLLER_DATA_SIGNATURE);
695         memcpy(meta->cdr->Controller_GUID, "FreeBSD GEOM RAID SERIAL", 24);
696         memcpy(meta->cdr->Product_ID, "FreeBSD GEOMRAID", 16);
697
698         /* Physical Drive Records. */
699         size = GET32(meta, hdr->pdr_length) * ss;
700         meta->pdr = malloc(size, M_MD_DDF, M_WAITOK);
701         memset(meta->pdr, 0xff, size);
702         SET32(meta, pdr->Signature, DDF_PDR_SIGNATURE);
703         SET16(meta, pdr->Populated_PDEs, 1);
704         SET16(meta, pdr->Max_PDE_Supported,
705             GET16(meta, hdr->Max_PD_Entries));
706
707         pde = &meta->pdr->entry[0];
708         len = sizeof(serial_buffer);
709         error = g_io_getattr("GEOM::ident", disk->d_consumer, &len, serial_buffer);
710         if (error == 0 && (len = strlen (serial_buffer)) >= 6 && len <= 20)
711                 snprintf(pde->PD_GUID, 25, "DISK%20s", serial_buffer);
712         else
713                 snprintf(pde->PD_GUID, 25, "DISK%04d%02d%02d%08x%04x",
714                     ct.year, ct.mon, ct.day,
715                     arc4random(), arc4random() & 0xffff);
716         SET32D(meta, pde->PD_Reference, arc4random());
717         SET16D(meta, pde->PD_Type, DDF_PDE_GUID_FORCE);
718         SET16D(meta, pde->PD_State, 0);
719         SET64D(meta, pde->Configured_Size,
720             anchorlba + 1 - 32 * 1024 * 1024 / ss);
721         SET16D(meta, pde->Block_Size, ss);
722
723         /* Virtual Drive Records. */
724         size = GET32(meta, hdr->vdr_length) * ss;
725         meta->vdr = malloc(size, M_MD_DDF, M_WAITOK);
726         memset(meta->vdr, 0xff, size);
727         SET32(meta, vdr->Signature, DDF_VD_RECORD_SIGNATURE);
728         SET32(meta, vdr->Populated_VDEs, 0);
729         SET16(meta, vdr->Max_VDE_Supported,
730             GET16(meta, hdr->Max_VD_Entries));
731
732         /* Configuration Records. */
733         size = GET32(meta, hdr->cr_length) * ss;
734         meta->cr = malloc(size, M_MD_DDF, M_WAITOK);
735         memset(meta->cr, 0xff, size);
736
737         /* Physical Disk Data. */
738         size = GET32(meta, hdr->pdd_length) * ss;
739         meta->pdd = malloc(size, M_MD_DDF, M_WAITOK);
740         memset(meta->pdd, 0xff, size);
741         SET32(meta, pdd->Signature, DDF_PDD_SIGNATURE);
742         memcpy(meta->pdd->PD_GUID, pde->PD_GUID, 24);
743         SET32(meta, pdd->PD_Reference, GET32D(meta, pde->PD_Reference));
744         SET8(meta, pdd->Forced_Ref_Flag, DDF_PDD_FORCED_REF);
745         SET8(meta, pdd->Forced_PD_GUID_Flag, DDF_PDD_FORCED_GUID);
746
747         /* Bad Block Management Log. */
748         if (GET32(meta, hdr->bbmlog_length) != 0) {
749                 size = GET32(meta, hdr->bbmlog_length) * ss;
750                 meta->bbm = malloc(size, M_MD_DDF, M_WAITOK);
751                 memset(meta->bbm, 0xff, size);
752                 SET32(meta, bbm->Signature, DDF_BBML_SIGNATURE);
753                 SET32(meta, bbm->Entry_Count, 0);
754                 SET32(meta, bbm->Spare_Block_Count, 0);
755         }
756 }
757
758 static void
759 ddf_meta_copy(struct ddf_meta *dst, struct ddf_meta *src)
760 {
761         u_int ss;
762
763         dst->bigendian = src->bigendian;
764         ss = dst->sectorsize = src->sectorsize;
765         dst->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
766         memcpy(dst->hdr, src->hdr, ss);
767         dst->cdr = malloc(GET32(src, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
768         memcpy(dst->cdr, src->cdr, GET32(src, hdr->cd_length) * ss);
769         dst->pdr = malloc(GET32(src, hdr->pdr_length) * ss, M_MD_DDF, M_WAITOK);
770         memcpy(dst->pdr, src->pdr, GET32(src, hdr->pdr_length) * ss);
771         dst->vdr = malloc(GET32(src, hdr->vdr_length) * ss, M_MD_DDF, M_WAITOK);
772         memcpy(dst->vdr, src->vdr, GET32(src, hdr->vdr_length) * ss);
773         dst->cr = malloc(GET32(src, hdr->cr_length) * ss, M_MD_DDF, M_WAITOK);
774         memcpy(dst->cr, src->cr, GET32(src, hdr->cr_length) * ss);
775         dst->pdd = malloc(GET32(src, hdr->pdd_length) * ss, M_MD_DDF, M_WAITOK);
776         memcpy(dst->pdd, src->pdd, GET32(src, hdr->pdd_length) * ss);
777         if (src->bbm != NULL) {
778                 dst->bbm = malloc(GET32(src, hdr->bbmlog_length) * ss, M_MD_DDF, M_WAITOK);
779                 memcpy(dst->bbm, src->bbm, GET32(src, hdr->bbmlog_length) * ss);
780         }
781 }
782
783 static void
784 ddf_meta_update(struct ddf_meta *meta, struct ddf_meta *src)
785 {
786         struct ddf_pd_entry *pde, *spde;
787         int i, j;
788
789         for (i = 0; i < GET16(src, pdr->Populated_PDEs); i++) {
790                 spde = &src->pdr->entry[i];
791                 if (isff(spde->PD_GUID, 24))
792                         continue;
793                 j = ddf_meta_find_pd(meta, NULL,
794                     GET32(src, pdr->entry[i].PD_Reference));
795                 if (j < 0) {
796                         j = ddf_meta_find_pd(meta, NULL, 0xffffffff);
797                         pde = &meta->pdr->entry[j];
798                         memcpy(pde, spde, sizeof(*pde));
799                 } else {
800                         pde = &meta->pdr->entry[j];
801                         SET16D(meta, pde->PD_State,
802                             GET16D(meta, pde->PD_State) |
803                             GET16D(src, pde->PD_State));
804                 }
805         }
806 }
807
808 static void
809 ddf_meta_free(struct ddf_meta *meta)
810 {
811
812         if (meta->hdr != NULL) {
813                 free(meta->hdr, M_MD_DDF);
814                 meta->hdr = NULL;
815         }
816         if (meta->cdr != NULL) {
817                 free(meta->cdr, M_MD_DDF);
818                 meta->cdr = NULL;
819         }
820         if (meta->pdr != NULL) {
821                 free(meta->pdr, M_MD_DDF);
822                 meta->pdr = NULL;
823         }
824         if (meta->vdr != NULL) {
825                 free(meta->vdr, M_MD_DDF);
826                 meta->vdr = NULL;
827         }
828         if (meta->cr != NULL) {
829                 free(meta->cr, M_MD_DDF);
830                 meta->cr = NULL;
831         }
832         if (meta->pdd != NULL) {
833                 free(meta->pdd, M_MD_DDF);
834                 meta->pdd = NULL;
835         }
836         if (meta->bbm != NULL) {
837                 free(meta->bbm, M_MD_DDF);
838                 meta->bbm = NULL;
839         }
840 }
841
842 static void
843 ddf_vol_meta_create(struct ddf_vol_meta *meta, struct ddf_meta *sample)
844 {
845         struct timespec ts;
846         struct clocktime ct;
847         u_int ss, size;
848
849         meta->bigendian = sample->bigendian;
850         ss = meta->sectorsize = sample->sectorsize;
851         meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
852         memcpy(meta->hdr, sample->hdr, ss);
853         meta->cdr = malloc(GET32(sample, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
854         memcpy(meta->cdr, sample->cdr, GET32(sample, hdr->cd_length) * ss);
855         meta->vde = malloc(sizeof(struct ddf_vd_entry), M_MD_DDF, M_WAITOK);
856         memset(meta->vde, 0xff, sizeof(struct ddf_vd_entry));
857         getnanotime(&ts);
858         clock_ts_to_ct(&ts, &ct);
859         snprintf(meta->vde->VD_GUID, 25, "FreeBSD%04d%02d%02d%08x%01x",
860             ct.year, ct.mon, ct.day,
861             arc4random(), arc4random() & 0xf);
862         size = GET16(sample, hdr->Configuration_Record_Length) * ss;
863         meta->vdc = malloc(size, M_MD_DDF, M_WAITOK);
864         memset(meta->vdc, 0xff, size);
865         SET32(meta, vdc->Signature, DDF_VDCR_SIGNATURE);
866         memcpy(meta->vdc->VD_GUID, meta->vde->VD_GUID, 24);
867         SET32(meta, vdc->Sequence_Number, 0);
868 }
869
870 static void
871 ddf_vol_meta_update(struct ddf_vol_meta *dst, struct ddf_meta *src,
872     uint8_t *GUID, int started)
873 {
874         struct ddf_vd_entry *vde;
875         struct ddf_vdc_record *vdc;
876         int vnew, bvnew, bvd, size;
877         u_int ss;
878
879         vde = &src->vdr->entry[ddf_meta_find_vd(src, GUID)];
880         vdc = ddf_meta_find_vdc(src, GUID);
881         if (GET8D(src, vdc->Secondary_Element_Count) == 1)
882                 bvd = 0;
883         else
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
929 static void
930 ddf_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
958 static int
959 ddf_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
1011 static void
1012 ddf_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
1025 static void
1026 ddf_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
1035 static int
1036 ddf_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
1048         pp = cp->provider;
1049         ss = meta->sectorsize = pp->sectorsize;
1050         if (ss < sizeof(*hdr))
1051                 return (ENXIO);
1052         /* Read anchor block. */
1053         abuf = g_read_data(cp, pp->mediasize - ss, ss, &error);
1054         if (abuf == NULL) {
1055                 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
1056                     pp->name, error);
1057                 return (error);
1058         }
1059         ahdr = (struct ddf_header *)abuf;
1060
1061         /* Check if this is an DDF RAID struct */
1062         if (be32dec(&ahdr->Signature) == DDF_HEADER_SIGNATURE)
1063                 meta->bigendian = 1;
1064         else if (le32dec(&ahdr->Signature) == DDF_HEADER_SIGNATURE)
1065                 meta->bigendian = 0;
1066         else {
1067                 G_RAID_DEBUG(1, "DDF signature check failed on %s", pp->name);
1068                 error = EINVAL;
1069                 goto done;
1070         }
1071         if (ahdr->Header_Type != DDF_HEADER_ANCHOR) {
1072                 G_RAID_DEBUG(1, "DDF header type check failed on %s", pp->name);
1073                 error = EINVAL;
1074                 goto done;
1075         }
1076         meta->hdr = ahdr;
1077         plba = GET64(meta, hdr->Primary_Header_LBA);
1078         slba = GET64(meta, hdr->Secondary_Header_LBA);
1079         val = GET32(meta, hdr->CRC);
1080         SET32(meta, hdr->CRC, 0xffffffff);
1081         meta->hdr = NULL;
1082         if (crc32(ahdr, ss) != val) {
1083                 G_RAID_DEBUG(1, "DDF CRC mismatch on %s", pp->name);
1084                 error = EINVAL;
1085                 goto done;
1086         }
1087         if ((plba + 6) * ss >= pp->mediasize) {
1088                 G_RAID_DEBUG(1, "DDF primary header LBA is wrong on %s", pp->name);
1089                 error = EINVAL;
1090                 goto done;
1091         }
1092         if (slba != -1 && (slba + 6) * ss >= pp->mediasize) {
1093                 G_RAID_DEBUG(1, "DDF secondary header LBA is wrong on %s", pp->name);
1094                 error = EINVAL;
1095                 goto done;
1096         }
1097         lba = plba;
1098
1099 doread:
1100         error = 0;
1101         ddf_meta_free(meta);
1102
1103         /* Read header block. */
1104         buf = g_read_data(cp, lba * ss, ss, &error);
1105         if (buf == NULL) {
1106 readerror:
1107                 G_RAID_DEBUG(1, "DDF %s metadata read error on %s (error=%d).",
1108                     (lba == plba) ? "primary" : "secondary", pp->name, error);
1109                 if (lba == plba && slba != -1) {
1110                         lba = slba;
1111                         goto doread;
1112                 }
1113                 G_RAID_DEBUG(1, "DDF metadata read error on %s.", pp->name);
1114                 goto done;
1115         }
1116         meta->hdr = malloc(ss, M_MD_DDF, M_WAITOK);
1117         memcpy(meta->hdr, buf, ss);
1118         g_free(buf);
1119         hdr = meta->hdr;
1120         val = GET32(meta, hdr->CRC);
1121         SET32(meta, hdr->CRC, 0xffffffff);
1122         if (hdr->Signature != ahdr->Signature ||
1123             crc32(meta->hdr, ss) != val ||
1124             memcmp(hdr->DDF_Header_GUID, ahdr->DDF_Header_GUID, 24) ||
1125             GET64(meta, hdr->Primary_Header_LBA) != plba ||
1126             GET64(meta, hdr->Secondary_Header_LBA) != slba) {
1127 hdrerror:
1128                 G_RAID_DEBUG(1, "DDF %s metadata check failed on %s",
1129                     (lba == plba) ? "primary" : "secondary", pp->name);
1130                 if (lba == plba && slba != -1) {
1131                         lba = slba;
1132                         goto doread;
1133                 }
1134                 G_RAID_DEBUG(1, "DDF metadata check failed on %s", pp->name);
1135                 error = EINVAL;
1136                 goto done;
1137         }
1138         if ((lba == plba && hdr->Header_Type != DDF_HEADER_PRIMARY) ||
1139             (lba == slba && hdr->Header_Type != DDF_HEADER_SECONDARY))
1140                 goto hdrerror;
1141         len = 1;
1142         len = max(len, GET32(meta, hdr->cd_section) + GET32(meta, hdr->cd_length));
1143         len = max(len, GET32(meta, hdr->pdr_section) + GET32(meta, hdr->pdr_length));
1144         len = max(len, GET32(meta, hdr->vdr_section) + GET32(meta, hdr->vdr_length));
1145         len = max(len, GET32(meta, hdr->cr_section) + GET32(meta, hdr->cr_length));
1146         len = max(len, GET32(meta, hdr->pdd_section) + GET32(meta, hdr->pdd_length));
1147         if ((val = GET32(meta, hdr->bbmlog_section)) != 0xffffffff)
1148                 len = max(len, val + GET32(meta, hdr->bbmlog_length));
1149         if ((val = GET32(meta, hdr->Diagnostic_Space)) != 0xffffffff)
1150                 len = max(len, val + GET32(meta, hdr->Diagnostic_Space_Length));
1151         if ((val = GET32(meta, hdr->Vendor_Specific_Logs)) != 0xffffffff)
1152                 len = max(len, val + GET32(meta, hdr->Vendor_Specific_Logs_Length));
1153         if ((plba + len) * ss >= pp->mediasize)
1154                 goto hdrerror;
1155         if (slba != -1 && (slba + len) * ss >= pp->mediasize)
1156                 goto hdrerror;
1157         /* Workaround for Adaptec implementation. */
1158         if (GET16(meta, hdr->Max_Primary_Element_Entries) == 0xffff) {
1159                 SET16(meta, hdr->Max_Primary_Element_Entries,
1160                     min(GET16(meta, hdr->Max_PD_Entries),
1161                     (GET16(meta, hdr->Configuration_Record_Length) * ss - 512) / 12));
1162         }
1163
1164         if (GET32(meta, hdr->cd_length) * ss >= maxphys ||
1165             GET32(meta, hdr->pdr_length) * ss >= maxphys ||
1166             GET32(meta, hdr->vdr_length) * ss >= maxphys ||
1167             GET32(meta, hdr->cr_length) * ss >= maxphys ||
1168             GET32(meta, hdr->pdd_length) * ss >= maxphys ||
1169             GET32(meta, hdr->bbmlog_length) * ss >= maxphys) {
1170                 G_RAID_DEBUG(1, "%s: Blocksize is too big.", pp->name);
1171                 goto hdrerror;
1172         }
1173
1174         /* Read controller data. */
1175         buf = g_read_data(cp, (lba + GET32(meta, hdr->cd_section)) * ss,
1176             GET32(meta, hdr->cd_length) * ss, &error);
1177         if (buf == NULL)
1178                 goto readerror;
1179         meta->cdr = malloc(GET32(meta, hdr->cd_length) * ss, M_MD_DDF, M_WAITOK);
1180         memcpy(meta->cdr, buf, GET32(meta, hdr->cd_length) * ss);
1181         g_free(buf);
1182         if (GET32(meta, cdr->Signature) != DDF_CONTROLLER_DATA_SIGNATURE)
1183                 goto hdrerror;
1184
1185         /* Read physical disk records. */
1186         buf = g_read_data(cp, (lba + GET32(meta, hdr->pdr_section)) * ss,
1187             GET32(meta, hdr->pdr_length) * ss, &error);
1188         if (buf == NULL)
1189                 goto readerror;
1190         meta->pdr = malloc(GET32(meta, hdr->pdr_length) * ss, M_MD_DDF, M_WAITOK);
1191         memcpy(meta->pdr, buf, GET32(meta, hdr->pdr_length) * ss);
1192         g_free(buf);
1193         if (GET32(meta, pdr->Signature) != DDF_PDR_SIGNATURE)
1194                 goto hdrerror;
1195         /*
1196          * Workaround for reading metadata corrupted due to graid bug.
1197          * XXX: Remove this before we have disks above 128PB. :)
1198          */
1199         if (meta->bigendian) {
1200                 for (i = 0; i < GET16(meta, pdr->Populated_PDEs); i++) {
1201                         if (isff(meta->pdr->entry[i].PD_GUID, 24))
1202                                 continue;
1203                         if (GET32(meta, pdr->entry[i].PD_Reference) ==
1204                             0xffffffff)
1205                                 continue;
1206                         if (GET64(meta, pdr->entry[i].Configured_Size) >=
1207                              (1ULL << 48)) {
1208                                 SET16(meta, pdr->entry[i].PD_State,
1209                                     GET16(meta, pdr->entry[i].PD_State) &
1210                                     ~DDF_PDE_FAILED);
1211                                 SET64(meta, pdr->entry[i].Configured_Size,
1212                                     GET64(meta, pdr->entry[i].Configured_Size) &
1213                                     ((1ULL << 48) - 1));
1214                         }
1215                 }
1216         }
1217
1218         /* Read virtual disk records. */
1219         buf = g_read_data(cp, (lba + GET32(meta, hdr->vdr_section)) * ss,
1220             GET32(meta, hdr->vdr_length) * ss, &error);
1221         if (buf == NULL)
1222                 goto readerror;
1223         meta->vdr = malloc(GET32(meta, hdr->vdr_length) * ss, M_MD_DDF, M_WAITOK);
1224         memcpy(meta->vdr, buf, GET32(meta, hdr->vdr_length) * ss);
1225         g_free(buf);
1226         if (GET32(meta, vdr->Signature) != DDF_VD_RECORD_SIGNATURE)
1227                 goto hdrerror;
1228
1229         /* Read configuration records. */
1230         buf = g_read_data(cp, (lba + GET32(meta, hdr->cr_section)) * ss,
1231             GET32(meta, hdr->cr_length) * ss, &error);
1232         if (buf == NULL)
1233                 goto readerror;
1234         meta->cr = malloc(GET32(meta, hdr->cr_length) * ss, M_MD_DDF, M_WAITOK);
1235         memcpy(meta->cr, buf, GET32(meta, hdr->cr_length) * ss);
1236         g_free(buf);
1237
1238         /* Read physical disk data. */
1239         buf = g_read_data(cp, (lba + GET32(meta, hdr->pdd_section)) * ss,
1240             GET32(meta, hdr->pdd_length) * ss, &error);
1241         if (buf == NULL)
1242                 goto readerror;
1243         meta->pdd = malloc(GET32(meta, hdr->pdd_length) * ss, M_MD_DDF, M_WAITOK);
1244         memcpy(meta->pdd, buf, GET32(meta, hdr->pdd_length) * ss);
1245         g_free(buf);
1246         if (GET32(meta, pdd->Signature) != DDF_PDD_SIGNATURE)
1247                 goto hdrerror;
1248         i = ddf_meta_find_pd(meta, NULL, GET32(meta, pdd->PD_Reference));
1249         if (i < 0)
1250                 goto hdrerror;
1251
1252         /* Read BBM Log. */
1253         if (GET32(meta, hdr->bbmlog_section) != 0xffffffff &&
1254             GET32(meta, hdr->bbmlog_length) != 0) {
1255                 buf = g_read_data(cp, (lba + GET32(meta, hdr->bbmlog_section)) * ss,
1256                     GET32(meta, hdr->bbmlog_length) * ss, &error);
1257                 if (buf == NULL)
1258                         goto readerror;
1259                 meta->bbm = malloc(GET32(meta, hdr->bbmlog_length) * ss, M_MD_DDF, M_WAITOK);
1260                 memcpy(meta->bbm, buf, GET32(meta, hdr->bbmlog_length) * ss);
1261                 g_free(buf);
1262                 if (GET32(meta, bbm->Signature) != DDF_BBML_SIGNATURE)
1263                         goto hdrerror;
1264         }
1265
1266 done:
1267         g_free(abuf);
1268         if (error != 0)
1269                 ddf_meta_free(meta);
1270         return (error);
1271 }
1272
1273 static int
1274 ddf_meta_write(struct g_consumer *cp, struct ddf_meta *meta)
1275 {
1276         struct g_provider *pp;
1277         struct ddf_vdc_record *vdc;
1278         off_t alba, plba, slba, lba;
1279         u_int ss, size;
1280         int error, i, num;
1281
1282         pp = cp->provider;
1283         ss = pp->sectorsize;
1284         lba = alba = pp->mediasize / ss - 1;
1285         plba = GET64(meta, hdr->Primary_Header_LBA);
1286         slba = GET64(meta, hdr->Secondary_Header_LBA);
1287
1288 next:
1289         SET8(meta, hdr->Header_Type, (lba == alba) ? DDF_HEADER_ANCHOR :
1290             (lba == plba) ? DDF_HEADER_PRIMARY : DDF_HEADER_SECONDARY);
1291         SET32(meta, hdr->CRC, 0xffffffff);
1292         SET32(meta, hdr->CRC, crc32(meta->hdr, ss));
1293         error = g_write_data(cp, lba * ss, meta->hdr, ss);
1294         if (error != 0) {
1295 err:
1296                 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
1297                     pp->name, error);
1298                 if (lba != alba)
1299                         goto done;
1300         }
1301         if (lba == alba) {
1302                 lba = plba;
1303                 goto next;
1304         }
1305
1306         size = GET32(meta, hdr->cd_length) * ss;
1307         SET32(meta, cdr->CRC, 0xffffffff);
1308         SET32(meta, cdr->CRC, crc32(meta->cdr, size));
1309         error = g_write_data(cp, (lba + GET32(meta, hdr->cd_section)) * ss,
1310             meta->cdr, size);
1311         if (error != 0)
1312                 goto err;
1313
1314         size = GET32(meta, hdr->pdr_length) * ss;
1315         SET32(meta, pdr->CRC, 0xffffffff);
1316         SET32(meta, pdr->CRC, crc32(meta->pdr, size));
1317         error = g_write_data(cp, (lba + GET32(meta, hdr->pdr_section)) * ss,
1318             meta->pdr, size);
1319         if (error != 0)
1320                 goto err;
1321
1322         size = GET32(meta, hdr->vdr_length) * ss;
1323         SET32(meta, vdr->CRC, 0xffffffff);
1324         SET32(meta, vdr->CRC, crc32(meta->vdr, size));
1325         error = g_write_data(cp, (lba + GET32(meta, hdr->vdr_section)) * ss,
1326             meta->vdr, size);
1327         if (error != 0)
1328                 goto err;
1329
1330         size = GET16(meta, hdr->Configuration_Record_Length) * ss;
1331         num = GETCRNUM(meta);
1332         for (i = 0; i < num; i++) {
1333                 vdc = GETVDCPTR(meta, i);
1334                 SET32D(meta, vdc->CRC, 0xffffffff);
1335                 SET32D(meta, vdc->CRC, crc32(vdc, size));
1336         }
1337         error = g_write_data(cp, (lba + GET32(meta, hdr->cr_section)) * ss,
1338             meta->cr, size * num);
1339         if (error != 0)
1340                 goto err;
1341
1342         size = GET32(meta, hdr->pdd_length) * ss;
1343         SET32(meta, pdd->CRC, 0xffffffff);
1344         SET32(meta, pdd->CRC, crc32(meta->pdd, size));
1345         error = g_write_data(cp, (lba + GET32(meta, hdr->pdd_section)) * ss,
1346             meta->pdd, size);
1347         if (error != 0)
1348                 goto err;
1349
1350         if (GET32(meta, hdr->bbmlog_length) != 0) {
1351                 size = GET32(meta, hdr->bbmlog_length) * ss;
1352                 SET32(meta, bbm->CRC, 0xffffffff);
1353                 SET32(meta, bbm->CRC, crc32(meta->bbm, size));
1354                 error = g_write_data(cp,
1355                     (lba + GET32(meta, hdr->bbmlog_section)) * ss,
1356                     meta->bbm, size);
1357                 if (error != 0)
1358                         goto err;
1359         }
1360
1361 done:
1362         if (lba == plba && slba != -1) {
1363                 lba = slba;
1364                 goto next;
1365         }
1366
1367         return (error);
1368 }
1369
1370 static int
1371 ddf_meta_erase(struct g_consumer *cp)
1372 {
1373         struct g_provider *pp;
1374         char *buf;
1375         int error;
1376
1377         pp = cp->provider;
1378         buf = malloc(pp->sectorsize, M_MD_DDF, M_WAITOK | M_ZERO);
1379         error = g_write_data(cp, pp->mediasize - pp->sectorsize,
1380             buf, pp->sectorsize);
1381         if (error != 0) {
1382                 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
1383                     pp->name, error);
1384         }
1385         free(buf, M_MD_DDF);
1386         return (error);
1387 }
1388
1389 static struct g_raid_volume *
1390 g_raid_md_ddf_get_volume(struct g_raid_softc *sc, uint8_t *GUID)
1391 {
1392         struct g_raid_volume    *vol;
1393         struct g_raid_md_ddf_pervolume *pv;
1394
1395         TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1396                 pv = vol->v_md_data;
1397                 if (memcmp(pv->pv_meta.vde->VD_GUID, GUID, 24) == 0)
1398                         break;
1399         }
1400         return (vol);
1401 }
1402
1403 static struct g_raid_disk *
1404 g_raid_md_ddf_get_disk(struct g_raid_softc *sc, uint8_t *GUID, uint32_t id)
1405 {
1406         struct g_raid_disk      *disk;
1407         struct g_raid_md_ddf_perdisk *pd;
1408         struct ddf_meta *meta;
1409
1410         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1411                 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1412                 meta = &pd->pd_meta;
1413                 if (GUID != NULL) {
1414                         if (memcmp(meta->pdd->PD_GUID, GUID, 24) == 0)
1415                                 break;
1416                 } else {
1417                         if (GET32(meta, pdd->PD_Reference) == id)
1418                                 break;
1419                 }
1420         }
1421         return (disk);
1422 }
1423
1424 static int
1425 g_raid_md_ddf_purge_volumes(struct g_raid_softc *sc)
1426 {
1427         struct g_raid_volume    *vol, *tvol;
1428         int i, res;
1429
1430         res = 0;
1431         TAILQ_FOREACH_SAFE(vol, &sc->sc_volumes, v_next, tvol) {
1432                 if (vol->v_stopping)
1433                         continue;
1434                 for (i = 0; i < vol->v_disks_count; i++) {
1435                         if (vol->v_subdisks[i].sd_state != G_RAID_SUBDISK_S_NONE)
1436                                 break;
1437                 }
1438                 if (i >= vol->v_disks_count) {
1439                         g_raid_destroy_volume(vol);
1440                         res = 1;
1441                 }
1442         }
1443         return (res);
1444 }
1445
1446 static int
1447 g_raid_md_ddf_purge_disks(struct g_raid_softc *sc)
1448 {
1449 #if 0
1450         struct g_raid_disk      *disk, *tdisk;
1451         struct g_raid_volume    *vol;
1452         struct g_raid_md_ddf_perdisk *pd;
1453         int i, j, res;
1454
1455         res = 0;
1456         TAILQ_FOREACH_SAFE(disk, &sc->sc_disks, d_next, tdisk) {
1457                 if (disk->d_state == G_RAID_DISK_S_SPARE)
1458                         continue;
1459                 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1460
1461                 /* Scan for deleted volumes. */
1462                 for (i = 0; i < pd->pd_subdisks; ) {
1463                         vol = g_raid_md_ddf_get_volume(sc,
1464                             pd->pd_meta[i]->volume_id);
1465                         if (vol != NULL && !vol->v_stopping) {
1466                                 i++;
1467                                 continue;
1468                         }
1469                         free(pd->pd_meta[i], M_MD_DDF);
1470                         for (j = i; j < pd->pd_subdisks - 1; j++)
1471                                 pd->pd_meta[j] = pd->pd_meta[j + 1];
1472                         pd->pd_meta[DDF_MAX_SUBDISKS - 1] = NULL;
1473                         pd->pd_subdisks--;
1474                         pd->pd_updated = 1;
1475                 }
1476
1477                 /* If there is no metadata left - erase and delete disk. */
1478                 if (pd->pd_subdisks == 0) {
1479                         ddf_meta_erase(disk->d_consumer);
1480                         g_raid_destroy_disk(disk);
1481                         res = 1;
1482                 }
1483         }
1484         return (res);
1485 #endif
1486         return (0);
1487 }
1488
1489 static int
1490 g_raid_md_ddf_supported(int level, int qual, int disks, int force)
1491 {
1492
1493         if (disks > DDF_MAX_DISKS_HARD)
1494                 return (0);
1495         switch (level) {
1496         case G_RAID_VOLUME_RL_RAID0:
1497                 if (qual != G_RAID_VOLUME_RLQ_NONE)
1498                         return (0);
1499                 if (disks < 1)
1500                         return (0);
1501                 if (!force && disks < 2)
1502                         return (0);
1503                 break;
1504         case G_RAID_VOLUME_RL_RAID1:
1505                 if (disks < 1)
1506                         return (0);
1507                 if (qual == G_RAID_VOLUME_RLQ_R1SM) {
1508                         if (!force && disks != 2)
1509                                 return (0);
1510                 } else if (qual == G_RAID_VOLUME_RLQ_R1MM) {
1511                         if (!force && disks != 3)
1512                                 return (0);
1513                 } else 
1514                         return (0);
1515                 break;
1516         case G_RAID_VOLUME_RL_RAID3:
1517                 if (qual != G_RAID_VOLUME_RLQ_R3P0 &&
1518                     qual != G_RAID_VOLUME_RLQ_R3PN)
1519                         return (0);
1520                 if (disks < 3)
1521                         return (0);
1522                 break;
1523         case G_RAID_VOLUME_RL_RAID4:
1524                 if (qual != G_RAID_VOLUME_RLQ_R4P0 &&
1525                     qual != G_RAID_VOLUME_RLQ_R4PN)
1526                         return (0);
1527                 if (disks < 3)
1528                         return (0);
1529                 break;
1530         case G_RAID_VOLUME_RL_RAID5:
1531                 if (qual != G_RAID_VOLUME_RLQ_R5RA &&
1532                     qual != G_RAID_VOLUME_RLQ_R5RS &&
1533                     qual != G_RAID_VOLUME_RLQ_R5LA &&
1534                     qual != G_RAID_VOLUME_RLQ_R5LS)
1535                         return (0);
1536                 if (disks < 3)
1537                         return (0);
1538                 break;
1539         case G_RAID_VOLUME_RL_RAID6:
1540                 if (qual != G_RAID_VOLUME_RLQ_R6RA &&
1541                     qual != G_RAID_VOLUME_RLQ_R6RS &&
1542                     qual != G_RAID_VOLUME_RLQ_R6LA &&
1543                     qual != G_RAID_VOLUME_RLQ_R6LS)
1544                         return (0);
1545                 if (disks < 4)
1546                         return (0);
1547                 break;
1548         case G_RAID_VOLUME_RL_RAIDMDF:
1549                 if (qual != G_RAID_VOLUME_RLQ_RMDFRA &&
1550                     qual != G_RAID_VOLUME_RLQ_RMDFRS &&
1551                     qual != G_RAID_VOLUME_RLQ_RMDFLA &&
1552                     qual != G_RAID_VOLUME_RLQ_RMDFLS)
1553                         return (0);
1554                 if (disks < 4)
1555                         return (0);
1556                 break;
1557         case G_RAID_VOLUME_RL_RAID1E:
1558                 if (qual != G_RAID_VOLUME_RLQ_R1EA &&
1559                     qual != G_RAID_VOLUME_RLQ_R1EO)
1560                         return (0);
1561                 if (disks < 3)
1562                         return (0);
1563                 break;
1564         case G_RAID_VOLUME_RL_SINGLE:
1565                 if (qual != G_RAID_VOLUME_RLQ_NONE)
1566                         return (0);
1567                 if (disks != 1)
1568                         return (0);
1569                 break;
1570         case G_RAID_VOLUME_RL_CONCAT:
1571                 if (qual != G_RAID_VOLUME_RLQ_NONE)
1572                         return (0);
1573                 if (disks < 2)
1574                         return (0);
1575                 break;
1576         case G_RAID_VOLUME_RL_RAID5E:
1577                 if (qual != G_RAID_VOLUME_RLQ_R5ERA &&
1578                     qual != G_RAID_VOLUME_RLQ_R5ERS &&
1579                     qual != G_RAID_VOLUME_RLQ_R5ELA &&
1580                     qual != G_RAID_VOLUME_RLQ_R5ELS)
1581                         return (0);
1582                 if (disks < 4)
1583                         return (0);
1584                 break;
1585         case G_RAID_VOLUME_RL_RAID5EE:
1586                 if (qual != G_RAID_VOLUME_RLQ_R5EERA &&
1587                     qual != G_RAID_VOLUME_RLQ_R5EERS &&
1588                     qual != G_RAID_VOLUME_RLQ_R5EELA &&
1589                     qual != G_RAID_VOLUME_RLQ_R5EELS)
1590                         return (0);
1591                 if (disks < 4)
1592                         return (0);
1593                 break;
1594         case G_RAID_VOLUME_RL_RAID5R:
1595                 if (qual != G_RAID_VOLUME_RLQ_R5RRA &&
1596                     qual != G_RAID_VOLUME_RLQ_R5RRS &&
1597                     qual != G_RAID_VOLUME_RLQ_R5RLA &&
1598                     qual != G_RAID_VOLUME_RLQ_R5RLS)
1599                         return (0);
1600                 if (disks < 3)
1601                         return (0);
1602                 break;
1603         default:
1604                 return (0);
1605         }
1606         return (1);
1607 }
1608
1609 static int
1610 g_raid_md_ddf_start_disk(struct g_raid_disk *disk, struct g_raid_volume *vol)
1611 {
1612         struct g_raid_softc *sc;
1613         struct g_raid_subdisk *sd;
1614         struct g_raid_md_ddf_perdisk *pd;
1615         struct g_raid_md_ddf_pervolume *pv;
1616         struct g_raid_md_ddf_object *mdi;
1617         struct ddf_vol_meta *vmeta;
1618         struct ddf_meta *pdmeta, *gmeta;
1619         struct ddf_vdc_record *vdc1;
1620         struct ddf_sa_record *sa;
1621         off_t size, eoff = 0, esize = 0;
1622         uint64_t *val2;
1623         int disk_pos, md_disk_bvd = -1, md_disk_pos = -1, md_pde_pos;
1624         int i, resurrection = 0;
1625         uint32_t reference;
1626
1627         sc = disk->d_softc;
1628         mdi = (struct g_raid_md_ddf_object *)sc->sc_md;
1629         pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1630         pdmeta = &pd->pd_meta;
1631         reference = GET32(&pd->pd_meta, pdd->PD_Reference);
1632
1633         pv = vol->v_md_data;
1634         vmeta = &pv->pv_meta;
1635         gmeta = &mdi->mdio_meta;
1636
1637         /* Find disk position in metadata by its reference. */
1638         disk_pos = ddf_meta_find_disk(vmeta, reference,
1639             &md_disk_bvd, &md_disk_pos);
1640         md_pde_pos = ddf_meta_find_pd(gmeta, NULL, reference);
1641
1642         if (disk_pos < 0) {
1643                 G_RAID_DEBUG1(1, sc,
1644                     "Disk %s is not a present part of the volume %s",
1645                     g_raid_get_diskname(disk), vol->v_name);
1646
1647                 /* Failed stale disk is useless for us. */
1648                 if ((GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA) != 0) {
1649                         g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
1650                         return (0);
1651                 }
1652
1653                 /* If disk has some metadata for this volume - erase. */
1654                 if ((vdc1 = ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID)) != NULL)
1655                         SET32D(pdmeta, vdc1->Signature, 0xffffffff);
1656
1657                 /* If we are in the start process, that's all for now. */
1658                 if (!pv->pv_started)
1659                         goto nofit;
1660                 /*
1661                  * If we have already started - try to get use of the disk.
1662                  * Try to replace OFFLINE disks first, then FAILED.
1663                  */
1664                 if (ddf_meta_count_vdc(&pd->pd_meta, NULL) >=
1665                         GET16(&pd->pd_meta, hdr->Max_Partitions)) {
1666                         G_RAID_DEBUG1(1, sc, "No free partitions on disk %s",
1667                             g_raid_get_diskname(disk));
1668                         goto nofit;
1669                 }
1670                 ddf_meta_unused_range(&pd->pd_meta, &eoff, &esize);
1671                 if (esize == 0) {
1672                         G_RAID_DEBUG1(1, sc, "No free space on disk %s",
1673                             g_raid_get_diskname(disk));
1674                         goto nofit;
1675                 }
1676                 eoff *= pd->pd_meta.sectorsize;
1677                 esize *= pd->pd_meta.sectorsize;
1678                 size = INT64_MAX;
1679                 for (i = 0; i < vol->v_disks_count; i++) {
1680                         sd = &vol->v_subdisks[i];
1681                         if (sd->sd_state != G_RAID_SUBDISK_S_NONE)
1682                                 size = sd->sd_size;
1683                         if (sd->sd_state <= G_RAID_SUBDISK_S_FAILED &&
1684                             (disk_pos < 0 ||
1685                              vol->v_subdisks[i].sd_state < sd->sd_state))
1686                                 disk_pos = i;
1687                 }
1688                 if (disk_pos >= 0 &&
1689                     vol->v_raid_level != G_RAID_VOLUME_RL_CONCAT &&
1690                     esize < size) {
1691                         G_RAID_DEBUG1(1, sc, "Disk %s free space "
1692                             "is too small (%ju < %ju)",
1693                             g_raid_get_diskname(disk), esize, size);
1694                         disk_pos = -1;
1695                 }
1696                 if (disk_pos >= 0) {
1697                         if (vol->v_raid_level != G_RAID_VOLUME_RL_CONCAT)
1698                                 esize = size;
1699                         md_disk_bvd = disk_pos / GET16(vmeta, vdc->Primary_Element_Count); // XXX
1700                         md_disk_pos = disk_pos % GET16(vmeta, vdc->Primary_Element_Count); // XXX
1701                 } else {
1702 nofit:
1703                         if (disk->d_state == G_RAID_DISK_S_NONE)
1704                                 g_raid_change_disk_state(disk,
1705                                     G_RAID_DISK_S_STALE);
1706                         return (0);
1707                 }
1708
1709                 /*
1710                  * If spare is committable, delete spare record.
1711                  * Othersize, mark it active and leave there.
1712                  */
1713                 sa = ddf_meta_find_sa(&pd->pd_meta, 0);
1714                 if (sa != NULL) {
1715                         if ((GET8D(&pd->pd_meta, sa->Spare_Type) &
1716                             DDF_SAR_TYPE_REVERTIBLE) == 0) {
1717                                 SET32D(&pd->pd_meta, sa->Signature, 0xffffffff);
1718                         } else {
1719                                 SET8D(&pd->pd_meta, sa->Spare_Type,
1720                                     GET8D(&pd->pd_meta, sa->Spare_Type) |
1721                                     DDF_SAR_TYPE_ACTIVE);
1722                         }
1723                 }
1724
1725                 G_RAID_DEBUG1(1, sc, "Disk %s takes pos %d in the volume %s",
1726                     g_raid_get_diskname(disk), disk_pos, vol->v_name);
1727                 resurrection = 1;
1728         }
1729
1730         sd = &vol->v_subdisks[disk_pos];
1731
1732         if (resurrection && sd->sd_disk != NULL) {
1733                 g_raid_change_disk_state(sd->sd_disk,
1734                     G_RAID_DISK_S_STALE_FAILED);
1735                 TAILQ_REMOVE(&sd->sd_disk->d_subdisks,
1736                     sd, sd_next);
1737         }
1738         vol->v_subdisks[disk_pos].sd_disk = disk;
1739         TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1740
1741         /* Welcome the new disk. */
1742         if (resurrection)
1743                 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1744         else if (GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA)
1745                 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1746         else
1747                 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1748
1749         if (resurrection) {
1750                 sd->sd_offset = eoff;
1751                 sd->sd_size = esize;
1752         } else if (pdmeta->cr != NULL &&
1753             (vdc1 = ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID)) != NULL) {
1754                 val2 = (uint64_t *)&(vdc1->Physical_Disk_Sequence[GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
1755                 sd->sd_offset = (off_t)GET64P(pdmeta, val2 + md_disk_pos) * 512;
1756                 sd->sd_size = (off_t)GET64D(pdmeta, vdc1->Block_Count) * 512;
1757         }
1758
1759         if (resurrection) {
1760                 /* Stale disk, almost same as new. */
1761                 g_raid_change_subdisk_state(sd,
1762                     G_RAID_SUBDISK_S_NEW);
1763         } else if (GET16(gmeta, pdr->entry[md_pde_pos].PD_State) & DDF_PDE_PFA) {
1764                 /* Failed disk. */
1765                 g_raid_change_subdisk_state(sd,
1766                     G_RAID_SUBDISK_S_FAILED);
1767         } else if ((GET16(gmeta, pdr->entry[md_pde_pos].PD_State) &
1768              (DDF_PDE_FAILED | DDF_PDE_REBUILD)) != 0) {
1769                 /* Rebuilding disk. */
1770                 g_raid_change_subdisk_state(sd,
1771                     G_RAID_SUBDISK_S_REBUILD);
1772                 sd->sd_rebuild_pos = 0;
1773         } else if ((GET8(vmeta, vde->VD_State) & DDF_VDE_DIRTY) != 0 ||
1774             (GET8(vmeta, vde->Init_State) & DDF_VDE_INIT_MASK) !=
1775              DDF_VDE_INIT_FULL) {
1776                 /* Stale disk or dirty volume (unclean shutdown). */
1777                 g_raid_change_subdisk_state(sd,
1778                     G_RAID_SUBDISK_S_STALE);
1779         } else {
1780                 /* Up to date disk. */
1781                 g_raid_change_subdisk_state(sd,
1782                     G_RAID_SUBDISK_S_ACTIVE);
1783         }
1784         g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1785             G_RAID_EVENT_SUBDISK);
1786
1787         return (resurrection);
1788 }
1789
1790 static void
1791 g_raid_md_ddf_refill(struct g_raid_softc *sc)
1792 {
1793         struct g_raid_volume *vol;
1794         struct g_raid_subdisk *sd;
1795         struct g_raid_disk *disk;
1796         struct g_raid_md_object *md;
1797         struct g_raid_md_ddf_perdisk *pd;
1798         struct g_raid_md_ddf_pervolume *pv;
1799         int update, updated, i, bad;
1800
1801         md = sc->sc_md;
1802 restart:
1803         updated = 0;
1804         TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1805                 pv = vol->v_md_data;
1806                 if (!pv->pv_started || vol->v_stopping)
1807                         continue;
1808
1809                 /* Search for subdisk that needs replacement. */
1810                 bad = 0;
1811                 for (i = 0; i < vol->v_disks_count; i++) {
1812                         sd = &vol->v_subdisks[i];
1813                         if (sd->sd_state == G_RAID_SUBDISK_S_NONE ||
1814                             sd->sd_state == G_RAID_SUBDISK_S_FAILED)
1815                                 bad = 1;
1816                 }
1817                 if (!bad)
1818                         continue;
1819
1820                 G_RAID_DEBUG1(1, sc, "Volume %s is not complete, "
1821                     "trying to refill.", vol->v_name);
1822
1823                 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1824                         /* Skip failed. */
1825                         if (disk->d_state < G_RAID_DISK_S_SPARE)
1826                                 continue;
1827                         /* Skip already used by this volume. */
1828                         for (i = 0; i < vol->v_disks_count; i++) {
1829                                 sd = &vol->v_subdisks[i];
1830                                 if (sd->sd_disk == disk)
1831                                         break;
1832                         }
1833                         if (i < vol->v_disks_count)
1834                                 continue;
1835
1836                         /* Try to use disk if it has empty extents. */
1837                         pd = disk->d_md_data;
1838                         if (ddf_meta_count_vdc(&pd->pd_meta, NULL) <
1839                             GET16(&pd->pd_meta, hdr->Max_Partitions)) {
1840                                 update = g_raid_md_ddf_start_disk(disk, vol);
1841                         } else
1842                                 update = 0;
1843                         if (update) {
1844                                 updated = 1;
1845                                 g_raid_md_write_ddf(md, vol, NULL, disk);
1846                                 break;
1847                         }
1848                 }
1849         }
1850         if (updated)
1851                 goto restart;
1852 }
1853
1854 static void
1855 g_raid_md_ddf_start(struct g_raid_volume *vol)
1856 {
1857         struct g_raid_softc *sc;
1858         struct g_raid_subdisk *sd;
1859         struct g_raid_disk *disk;
1860         struct g_raid_md_object *md;
1861         struct g_raid_md_ddf_perdisk *pd;
1862         struct g_raid_md_ddf_pervolume *pv;
1863         struct g_raid_md_ddf_object *mdi;
1864         struct ddf_vol_meta *vmeta;
1865         uint64_t *val2;
1866         int i, j, bvd;
1867
1868         sc = vol->v_softc;
1869         md = sc->sc_md;
1870         mdi = (struct g_raid_md_ddf_object *)md;
1871         pv = vol->v_md_data;
1872         vmeta = &pv->pv_meta;
1873
1874         vol->v_raid_level = GET8(vmeta, vdc->Primary_RAID_Level);
1875         vol->v_raid_level_qualifier = GET8(vmeta, vdc->RLQ);
1876         if (GET8(vmeta, vdc->Secondary_Element_Count) > 1 &&
1877             vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 &&
1878             GET8(vmeta, vdc->Secondary_RAID_Level) == 0)
1879                 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1880         vol->v_sectorsize = GET16(vmeta, vdc->Block_Size);
1881         if (vol->v_sectorsize == 0xffff)
1882                 vol->v_sectorsize = vmeta->sectorsize;
1883         vol->v_strip_size = vol->v_sectorsize << GET8(vmeta, vdc->Stripe_Size);
1884         vol->v_disks_count = GET16(vmeta, vdc->Primary_Element_Count) *
1885             GET8(vmeta, vdc->Secondary_Element_Count);
1886         vol->v_mdf_pdisks = GET8(vmeta, vdc->MDF_Parity_Disks);
1887         vol->v_mdf_polynomial = GET16(vmeta, vdc->MDF_Parity_Generator_Polynomial);
1888         vol->v_mdf_method = GET8(vmeta, vdc->MDF_Constant_Generation_Method);
1889         if (GET8(vmeta, vdc->Rotate_Parity_count) > 31)
1890                 vol->v_rotate_parity = 1;
1891         else
1892                 vol->v_rotate_parity = 1 << GET8(vmeta, vdc->Rotate_Parity_count);
1893         vol->v_mediasize = GET64(vmeta, vdc->VD_Size) * vol->v_sectorsize;
1894         for (i = 0, j = 0, bvd = 0; i < vol->v_disks_count; i++, j++) {
1895                 if (j == GET16(vmeta, vdc->Primary_Element_Count)) {
1896                         j = 0;
1897                         bvd++;
1898                 }
1899                 sd = &vol->v_subdisks[i];
1900                 if (vmeta->bvdc[bvd] == NULL) {
1901                         sd->sd_offset = 0;
1902                         sd->sd_size = GET64(vmeta, vdc->Block_Count) *
1903                             vol->v_sectorsize;
1904                         continue;
1905                 }
1906                 val2 = (uint64_t *)&(vmeta->bvdc[bvd]->Physical_Disk_Sequence[
1907                     GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
1908                 sd->sd_offset = GET64P(vmeta, val2 + j) * vol->v_sectorsize;
1909                 sd->sd_size = GET64(vmeta, bvdc[bvd]->Block_Count) *
1910                     vol->v_sectorsize;
1911         }
1912         g_raid_start_volume(vol);
1913
1914         /* Make all disks found till the moment take their places. */
1915         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1916                 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1917                 if (ddf_meta_find_vdc(&pd->pd_meta, vmeta->vdc->VD_GUID) != NULL)
1918                         g_raid_md_ddf_start_disk(disk, vol);
1919         }
1920
1921         pv->pv_started = 1;
1922         mdi->mdio_starting--;
1923         callout_stop(&pv->pv_start_co);
1924         G_RAID_DEBUG1(0, sc, "Volume started.");
1925         g_raid_md_write_ddf(md, vol, NULL, NULL);
1926
1927         /* Pickup any STALE/SPARE disks to refill array if needed. */
1928         g_raid_md_ddf_refill(sc);
1929
1930         g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
1931 }
1932
1933 static void
1934 g_raid_ddf_go(void *arg)
1935 {
1936         struct g_raid_volume *vol;
1937         struct g_raid_softc *sc;
1938         struct g_raid_md_ddf_pervolume *pv;
1939
1940         vol = arg;
1941         pv = vol->v_md_data;
1942         sc = vol->v_softc;
1943         if (!pv->pv_started) {
1944                 G_RAID_DEBUG1(0, sc, "Force volume start due to timeout.");
1945                 g_raid_event_send(vol, G_RAID_VOLUME_E_STARTMD,
1946                     G_RAID_EVENT_VOLUME);
1947         }
1948 }
1949
1950 static void
1951 g_raid_md_ddf_new_disk(struct g_raid_disk *disk)
1952 {
1953         struct g_raid_softc *sc;
1954         struct g_raid_md_object *md;
1955         struct g_raid_md_ddf_perdisk *pd;
1956         struct g_raid_md_ddf_pervolume *pv;
1957         struct g_raid_md_ddf_object *mdi;
1958         struct g_raid_volume *vol;
1959         struct ddf_meta *pdmeta;
1960         struct ddf_vol_meta *vmeta;
1961         struct ddf_vdc_record *vdc;
1962         struct ddf_vd_entry *vde;
1963         int i, j, k, num, have, need, cnt, spare;
1964         uint32_t val;
1965         char buf[17];
1966
1967         sc = disk->d_softc;
1968         md = sc->sc_md;
1969         mdi = (struct g_raid_md_ddf_object *)md;
1970         pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
1971         pdmeta = &pd->pd_meta;
1972         spare = -1;
1973
1974         if (mdi->mdio_meta.hdr == NULL)
1975                 ddf_meta_copy(&mdi->mdio_meta, pdmeta);
1976         else
1977                 ddf_meta_update(&mdi->mdio_meta, pdmeta);
1978
1979         num = GETCRNUM(pdmeta);
1980         for (j = 0; j < num; j++) {
1981                 vdc = GETVDCPTR(pdmeta, j);
1982                 val = GET32D(pdmeta, vdc->Signature);
1983
1984                 if (val == DDF_SA_SIGNATURE && spare == -1)
1985                         spare = 1;
1986
1987                 if (val != DDF_VDCR_SIGNATURE)
1988                         continue;
1989                 spare = 0;
1990                 k = ddf_meta_find_vd(pdmeta, vdc->VD_GUID);
1991                 if (k < 0)
1992                         continue;
1993                 vde = &pdmeta->vdr->entry[k];
1994
1995                 /* Look for volume with matching ID. */
1996                 vol = g_raid_md_ddf_get_volume(sc, vdc->VD_GUID);
1997                 if (vol == NULL) {
1998                         ddf_meta_get_name(pdmeta, k, buf);
1999                         vol = g_raid_create_volume(sc, buf,
2000                             GET16D(pdmeta, vde->VD_Number));
2001                         pv = malloc(sizeof(*pv), M_MD_DDF, M_WAITOK | M_ZERO);
2002                         vol->v_md_data = pv;
2003                         callout_init(&pv->pv_start_co, 1);
2004                         callout_reset(&pv->pv_start_co,
2005                             g_raid_start_timeout * hz,
2006                             g_raid_ddf_go, vol);
2007                         mdi->mdio_starting++;
2008                 } else
2009                         pv = vol->v_md_data;
2010
2011                 /* If we haven't started yet - check metadata freshness. */
2012                 vmeta = &pv->pv_meta;
2013                 ddf_vol_meta_update(vmeta, pdmeta, vdc->VD_GUID, pv->pv_started);
2014         }
2015
2016         if (spare == 1) {
2017                 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
2018                 g_raid_md_ddf_refill(sc);
2019         }
2020
2021         TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2022                 pv = vol->v_md_data;
2023                 vmeta = &pv->pv_meta;
2024
2025                 if (ddf_meta_find_vdc(pdmeta, vmeta->vdc->VD_GUID) == NULL)
2026                         continue;
2027
2028                 if (pv->pv_started) {
2029                         if (g_raid_md_ddf_start_disk(disk, vol))
2030                                 g_raid_md_write_ddf(md, vol, NULL, NULL);
2031                         continue;
2032                 }
2033
2034                 /* If we collected all needed disks - start array. */
2035                 need = 0;
2036                 have = 0;
2037                 for (k = 0; k < GET8(vmeta, vdc->Secondary_Element_Count); k++) {
2038                         if (vmeta->bvdc[k] == NULL) {
2039                                 need += GET16(vmeta, vdc->Primary_Element_Count);
2040                                 continue;
2041                         }
2042                         cnt = GET16(vmeta, bvdc[k]->Primary_Element_Count);
2043                         need += cnt;
2044                         for (i = 0; i < cnt; i++) {
2045                                 val = GET32(vmeta, bvdc[k]->Physical_Disk_Sequence[i]);
2046                                 if (g_raid_md_ddf_get_disk(sc, NULL, val) != NULL)
2047                                         have++;
2048                         }
2049                 }
2050                 G_RAID_DEBUG1(1, sc, "Volume %s now has %d of %d disks",
2051                     vol->v_name, have, need);
2052                 if (have == need)
2053                         g_raid_md_ddf_start(vol);
2054         }
2055 }
2056
2057 static int
2058 g_raid_md_create_req_ddf(struct g_raid_md_object *md, struct g_class *mp,
2059     struct gctl_req *req, struct g_geom **gp)
2060 {
2061         struct g_geom *geom;
2062         struct g_raid_softc *sc;
2063         struct g_raid_md_ddf_object *mdi, *mdi1;
2064         char name[16];
2065         const char *fmtopt;
2066         int be = 1;
2067
2068         mdi = (struct g_raid_md_ddf_object *)md;
2069         fmtopt = gctl_get_asciiparam(req, "fmtopt");
2070         if (fmtopt == NULL || strcasecmp(fmtopt, "BE") == 0)
2071                 be = 1;
2072         else if (strcasecmp(fmtopt, "LE") == 0)
2073                 be = 0;
2074         else {
2075                 gctl_error(req, "Incorrect fmtopt argument.");
2076                 return (G_RAID_MD_TASTE_FAIL);
2077         }
2078
2079         /* Search for existing node. */
2080         LIST_FOREACH(geom, &mp->geom, geom) {
2081                 sc = geom->softc;
2082                 if (sc == NULL)
2083                         continue;
2084                 if (sc->sc_stopping != 0)
2085                         continue;
2086                 if (sc->sc_md->mdo_class != md->mdo_class)
2087                         continue;
2088                 mdi1 = (struct g_raid_md_ddf_object *)sc->sc_md;
2089                 if (mdi1->mdio_bigendian != be)
2090                         continue;
2091                 break;
2092         }
2093         if (geom != NULL) {
2094                 *gp = geom;
2095                 return (G_RAID_MD_TASTE_EXISTING);
2096         }
2097
2098         /* Create new one if not found. */
2099         mdi->mdio_bigendian = be;
2100         snprintf(name, sizeof(name), "DDF%s", be ? "" : "-LE");
2101         sc = g_raid_create_node(mp, name, md);
2102         if (sc == NULL)
2103                 return (G_RAID_MD_TASTE_FAIL);
2104         md->mdo_softc = sc;
2105         *gp = sc->sc_geom;
2106         return (G_RAID_MD_TASTE_NEW);
2107 }
2108
2109 static int
2110 g_raid_md_taste_ddf(struct g_raid_md_object *md, struct g_class *mp,
2111                               struct g_consumer *cp, struct g_geom **gp)
2112 {
2113         struct g_consumer *rcp;
2114         struct g_provider *pp;
2115         struct g_raid_softc *sc;
2116         struct g_raid_disk *disk;
2117         struct ddf_meta meta;
2118         struct g_raid_md_ddf_perdisk *pd;
2119         struct g_raid_md_ddf_object *mdi;
2120         struct g_geom *geom;
2121         int error, result, be;
2122         char name[16];
2123
2124         G_RAID_DEBUG(1, "Tasting DDF on %s", cp->provider->name);
2125         mdi = (struct g_raid_md_ddf_object *)md;
2126         pp = cp->provider;
2127
2128         /* Read metadata from device. */
2129         g_topology_unlock();
2130         bzero(&meta, sizeof(meta));
2131         error = ddf_meta_read(cp, &meta);
2132         g_topology_lock();
2133         if (error != 0)
2134                 return (G_RAID_MD_TASTE_FAIL);
2135         be = meta.bigendian;
2136
2137         /* Metadata valid. Print it. */
2138         g_raid_md_ddf_print(&meta);
2139
2140         /* Search for matching node. */
2141         sc = NULL;
2142         LIST_FOREACH(geom, &mp->geom, geom) {
2143                 sc = geom->softc;
2144                 if (sc == NULL)
2145                         continue;
2146                 if (sc->sc_stopping != 0)
2147                         continue;
2148                 if (sc->sc_md->mdo_class != md->mdo_class)
2149                         continue;
2150                 mdi = (struct g_raid_md_ddf_object *)sc->sc_md;
2151                 if (mdi->mdio_bigendian != be)
2152                         continue;
2153                 break;
2154         }
2155
2156         /* Found matching node. */
2157         if (geom != NULL) {
2158                 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
2159                 result = G_RAID_MD_TASTE_EXISTING;
2160
2161         } else { /* Not found matching node -- create one. */
2162                 result = G_RAID_MD_TASTE_NEW;
2163                 mdi->mdio_bigendian = be;
2164                 snprintf(name, sizeof(name), "DDF%s", be ? "" : "-LE");
2165                 sc = g_raid_create_node(mp, name, md);
2166                 md->mdo_softc = sc;
2167                 geom = sc->sc_geom;
2168         }
2169
2170         /* There is no return after this point, so we close passed consumer. */
2171         g_access(cp, -1, 0, 0);
2172
2173         rcp = g_new_consumer(geom);
2174         rcp->flags |= G_CF_DIRECT_RECEIVE;
2175         g_attach(rcp, pp);
2176         if (g_access(rcp, 1, 1, 1) != 0)
2177                 ; //goto fail1;
2178
2179         g_topology_unlock();
2180         sx_xlock(&sc->sc_lock);
2181
2182         pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2183         pd->pd_meta = meta;
2184         disk = g_raid_create_disk(sc);
2185         disk->d_md_data = (void *)pd;
2186         disk->d_consumer = rcp;
2187         rcp->private = disk;
2188
2189         g_raid_get_disk_info(disk);
2190
2191         g_raid_md_ddf_new_disk(disk);
2192
2193         sx_xunlock(&sc->sc_lock);
2194         g_topology_lock();
2195         *gp = geom;
2196         return (result);
2197 }
2198
2199 static int
2200 g_raid_md_event_ddf(struct g_raid_md_object *md,
2201     struct g_raid_disk *disk, u_int event)
2202 {
2203         struct g_raid_softc *sc;
2204
2205         sc = md->mdo_softc;
2206         if (disk == NULL)
2207                 return (-1);
2208         switch (event) {
2209         case G_RAID_DISK_E_DISCONNECTED:
2210                 /* Delete disk. */
2211                 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2212                 g_raid_destroy_disk(disk);
2213                 g_raid_md_ddf_purge_volumes(sc);
2214
2215                 /* Write updated metadata to all disks. */
2216                 g_raid_md_write_ddf(md, NULL, NULL, NULL);
2217
2218                 /* Check if anything left. */
2219                 if (g_raid_ndisks(sc, -1) == 0)
2220                         g_raid_destroy_node(sc, 0);
2221                 else
2222                         g_raid_md_ddf_refill(sc);
2223                 return (0);
2224         }
2225         return (-2);
2226 }
2227
2228 static int
2229 g_raid_md_volume_event_ddf(struct g_raid_md_object *md,
2230     struct g_raid_volume *vol, u_int event)
2231 {
2232         struct g_raid_md_ddf_pervolume *pv;
2233
2234         pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2235         switch (event) {
2236         case G_RAID_VOLUME_E_STARTMD:
2237                 if (!pv->pv_started)
2238                         g_raid_md_ddf_start(vol);
2239                 return (0);
2240         }
2241         return (-2);
2242 }
2243
2244 static int
2245 g_raid_md_ctl_ddf(struct g_raid_md_object *md,
2246     struct gctl_req *req)
2247 {
2248         struct g_raid_softc *sc;
2249         struct g_raid_volume *vol, *vol1;
2250         struct g_raid_subdisk *sd;
2251         struct g_raid_disk *disk, *disks[DDF_MAX_DISKS_HARD];
2252         struct g_raid_md_ddf_perdisk *pd;
2253         struct g_raid_md_ddf_pervolume *pv;
2254         struct g_raid_md_ddf_object *mdi;
2255         struct ddf_sa_record *sa;
2256         struct g_consumer *cp;
2257         struct g_provider *pp;
2258         char arg[16];
2259         const char *nodename, *verb, *volname, *levelname, *diskname;
2260         char *tmp;
2261         int *nargs, *force;
2262         off_t size, sectorsize, strip, offs[DDF_MAX_DISKS_HARD], esize;
2263         intmax_t *sizearg, *striparg;
2264         int i, numdisks, len, level, qual;
2265         int error;
2266
2267         sc = md->mdo_softc;
2268         mdi = (struct g_raid_md_ddf_object *)md;
2269         verb = gctl_get_param(req, "verb", NULL);
2270         nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
2271         error = 0;
2272
2273         if (strcmp(verb, "label") == 0) {
2274                 if (*nargs < 4) {
2275                         gctl_error(req, "Invalid number of arguments.");
2276                         return (-1);
2277                 }
2278                 volname = gctl_get_asciiparam(req, "arg1");
2279                 if (volname == NULL) {
2280                         gctl_error(req, "No volume name.");
2281                         return (-2);
2282                 }
2283                 levelname = gctl_get_asciiparam(req, "arg2");
2284                 if (levelname == NULL) {
2285                         gctl_error(req, "No RAID level.");
2286                         return (-3);
2287                 }
2288                 if (g_raid_volume_str2level(levelname, &level, &qual)) {
2289                         gctl_error(req, "Unknown RAID level '%s'.", levelname);
2290                         return (-4);
2291                 }
2292                 numdisks = *nargs - 3;
2293                 force = gctl_get_paraml(req, "force", sizeof(*force));
2294                 if (!g_raid_md_ddf_supported(level, qual, numdisks,
2295                     force ? *force : 0)) {
2296                         gctl_error(req, "Unsupported RAID level "
2297                             "(0x%02x/0x%02x), or number of disks (%d).",
2298                             level, qual, numdisks);
2299                         return (-5);
2300                 }
2301
2302                 /* Search for disks, connect them and probe. */
2303                 size = INT64_MAX;
2304                 sectorsize = 0;
2305                 bzero(disks, sizeof(disks));
2306                 bzero(offs, sizeof(offs));
2307                 for (i = 0; i < numdisks; i++) {
2308                         snprintf(arg, sizeof(arg), "arg%d", i + 3);
2309                         diskname = gctl_get_asciiparam(req, arg);
2310                         if (diskname == NULL) {
2311                                 gctl_error(req, "No disk name (%s).", arg);
2312                                 error = -6;
2313                                 break;
2314                         }
2315                         if (strcmp(diskname, "NONE") == 0)
2316                                 continue;
2317
2318                         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2319                                 if (disk->d_consumer != NULL && 
2320                                     disk->d_consumer->provider != NULL &&
2321                                     strcmp(disk->d_consumer->provider->name,
2322                                      diskname) == 0)
2323                                         break;
2324                         }
2325                         if (disk != NULL) {
2326                                 if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2327                                         gctl_error(req, "Disk '%s' is in a "
2328                                             "wrong state (%s).", diskname,
2329                                             g_raid_disk_state2str(disk->d_state));
2330                                         error = -7;
2331                                         break;
2332                                 }
2333                                 pd = disk->d_md_data;
2334                                 if (ddf_meta_count_vdc(&pd->pd_meta, NULL) >=
2335                                     GET16(&pd->pd_meta, hdr->Max_Partitions)) {
2336                                         gctl_error(req, "No free partitions "
2337                                             "on disk '%s'.",
2338                                             diskname);
2339                                         error = -7;
2340                                         break;
2341                                 }
2342                                 pp = disk->d_consumer->provider;
2343                                 disks[i] = disk;
2344                                 ddf_meta_unused_range(&pd->pd_meta,
2345                                     &offs[i], &esize);
2346                                 offs[i] *= pp->sectorsize;
2347                                 size = MIN(size, (off_t)esize * pp->sectorsize);
2348                                 sectorsize = MAX(sectorsize, pp->sectorsize);
2349                                 continue;
2350                         }
2351
2352                         g_topology_lock();
2353                         cp = g_raid_open_consumer(sc, diskname);
2354                         if (cp == NULL) {
2355                                 gctl_error(req, "Can't open disk '%s'.",
2356                                     diskname);
2357                                 g_topology_unlock();
2358                                 error = -8;
2359                                 break;
2360                         }
2361                         pp = cp->provider;
2362                         pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2363                         disk = g_raid_create_disk(sc);
2364                         disk->d_md_data = (void *)pd;
2365                         disk->d_consumer = cp;
2366                         disks[i] = disk;
2367                         cp->private = disk;
2368                         ddf_meta_create(disk, &mdi->mdio_meta);
2369                         if (mdi->mdio_meta.hdr == NULL)
2370                                 ddf_meta_copy(&mdi->mdio_meta, &pd->pd_meta);
2371                         else
2372                                 ddf_meta_update(&mdi->mdio_meta, &pd->pd_meta);
2373                         g_topology_unlock();
2374
2375                         g_raid_get_disk_info(disk);
2376
2377                         /* Reserve some space for metadata. */
2378                         size = MIN(size, GET64(&pd->pd_meta,
2379                             pdr->entry[0].Configured_Size) * pp->sectorsize);
2380                         sectorsize = MAX(sectorsize, pp->sectorsize);
2381                 }
2382                 if (error != 0) {
2383                         for (i = 0; i < numdisks; i++) {
2384                                 if (disks[i] != NULL &&
2385                                     disks[i]->d_state == G_RAID_DISK_S_NONE)
2386                                         g_raid_destroy_disk(disks[i]);
2387                         }
2388                         return (error);
2389                 }
2390
2391                 if (sectorsize <= 0) {
2392                         gctl_error(req, "Can't get sector size.");
2393                         return (-8);
2394                 }
2395
2396                 /* Handle size argument. */
2397                 len = sizeof(*sizearg);
2398                 sizearg = gctl_get_param(req, "size", &len);
2399                 if (sizearg != NULL && len == sizeof(*sizearg) &&
2400                     *sizearg > 0) {
2401                         if (*sizearg > size) {
2402                                 gctl_error(req, "Size too big %lld > %lld.",
2403                                     (long long)*sizearg, (long long)size);
2404                                 return (-9);
2405                         }
2406                         size = *sizearg;
2407                 }
2408
2409                 /* Handle strip argument. */
2410                 strip = 131072;
2411                 len = sizeof(*striparg);
2412                 striparg = gctl_get_param(req, "strip", &len);
2413                 if (striparg != NULL && len == sizeof(*striparg) &&
2414                     *striparg > 0) {
2415                         if (*striparg < sectorsize) {
2416                                 gctl_error(req, "Strip size too small.");
2417                                 return (-10);
2418                         }
2419                         if (*striparg % sectorsize != 0) {
2420                                 gctl_error(req, "Incorrect strip size.");
2421                                 return (-11);
2422                         }
2423                         strip = *striparg;
2424                 }
2425
2426                 /* Round size down to strip or sector. */
2427                 if (level == G_RAID_VOLUME_RL_RAID1 ||
2428                     level == G_RAID_VOLUME_RL_RAID3 ||
2429                     level == G_RAID_VOLUME_RL_SINGLE ||
2430                     level == G_RAID_VOLUME_RL_CONCAT)
2431                         size -= (size % sectorsize);
2432                 else if (level == G_RAID_VOLUME_RL_RAID1E &&
2433                     (numdisks & 1) != 0)
2434                         size -= (size % (2 * strip));
2435                 else
2436                         size -= (size % strip);
2437                 if (size <= 0) {
2438                         gctl_error(req, "Size too small.");
2439                         return (-13);
2440                 }
2441
2442                 /* We have all we need, create things: volume, ... */
2443                 pv = malloc(sizeof(*pv), M_MD_DDF, M_WAITOK | M_ZERO);
2444                 ddf_vol_meta_create(&pv->pv_meta, &mdi->mdio_meta);
2445                 pv->pv_started = 1;
2446                 vol = g_raid_create_volume(sc, volname, -1);
2447                 vol->v_md_data = pv;
2448                 vol->v_raid_level = level;
2449                 vol->v_raid_level_qualifier = qual;
2450                 vol->v_strip_size = strip;
2451                 vol->v_disks_count = numdisks;
2452                 if (level == G_RAID_VOLUME_RL_RAID0 ||
2453                     level == G_RAID_VOLUME_RL_CONCAT ||
2454                     level == G_RAID_VOLUME_RL_SINGLE)
2455                         vol->v_mediasize = size * numdisks;
2456                 else if (level == G_RAID_VOLUME_RL_RAID1)
2457                         vol->v_mediasize = size;
2458                 else if (level == G_RAID_VOLUME_RL_RAID3 ||
2459                     level == G_RAID_VOLUME_RL_RAID4 ||
2460                     level == G_RAID_VOLUME_RL_RAID5)
2461                         vol->v_mediasize = size * (numdisks - 1);
2462                 else if (level == G_RAID_VOLUME_RL_RAID5R) {
2463                         vol->v_mediasize = size * (numdisks - 1);
2464                         vol->v_rotate_parity = 1024;
2465                 } else if (level == G_RAID_VOLUME_RL_RAID6 ||
2466                     level == G_RAID_VOLUME_RL_RAID5E ||
2467                     level == G_RAID_VOLUME_RL_RAID5EE)
2468                         vol->v_mediasize = size * (numdisks - 2);
2469                 else if (level == G_RAID_VOLUME_RL_RAIDMDF) {
2470                         if (numdisks < 5)
2471                                 vol->v_mdf_pdisks = 2;
2472                         else
2473                                 vol->v_mdf_pdisks = 3;
2474                         vol->v_mdf_polynomial = 0x11d;
2475                         vol->v_mdf_method = 0x00;
2476                         vol->v_mediasize = size * (numdisks - vol->v_mdf_pdisks);
2477                 } else { /* RAID1E */
2478                         vol->v_mediasize = ((size * numdisks) / strip / 2) *
2479                             strip;
2480                 }
2481                 vol->v_sectorsize = sectorsize;
2482                 g_raid_start_volume(vol);
2483
2484                 /* , and subdisks. */
2485                 for (i = 0; i < numdisks; i++) {
2486                         disk = disks[i];
2487                         sd = &vol->v_subdisks[i];
2488                         sd->sd_disk = disk;
2489                         sd->sd_offset = offs[i];
2490                         sd->sd_size = size;
2491                         if (disk == NULL)
2492                                 continue;
2493                         TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
2494                         g_raid_change_disk_state(disk,
2495                             G_RAID_DISK_S_ACTIVE);
2496                         g_raid_change_subdisk_state(sd,
2497                             G_RAID_SUBDISK_S_ACTIVE);
2498                         g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
2499                             G_RAID_EVENT_SUBDISK);
2500                 }
2501
2502                 /* Write metadata based on created entities. */
2503                 G_RAID_DEBUG1(0, sc, "Array started.");
2504                 g_raid_md_write_ddf(md, vol, NULL, NULL);
2505
2506                 /* Pickup any STALE/SPARE disks to refill array if needed. */
2507                 g_raid_md_ddf_refill(sc);
2508
2509                 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
2510                     G_RAID_EVENT_VOLUME);
2511                 return (0);
2512         }
2513         if (strcmp(verb, "add") == 0) {
2514                 gctl_error(req, "`add` command is not applicable, "
2515                     "use `label` instead.");
2516                 return (-99);
2517         }
2518         if (strcmp(verb, "delete") == 0) {
2519                 nodename = gctl_get_asciiparam(req, "arg0");
2520                 if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2521                         nodename = NULL;
2522
2523                 /* Full node destruction. */
2524                 if (*nargs == 1 && nodename != NULL) {
2525                         /* Check if some volume is still open. */
2526                         force = gctl_get_paraml(req, "force", sizeof(*force));
2527                         if (force != NULL && *force == 0 &&
2528                             g_raid_nopens(sc) != 0) {
2529                                 gctl_error(req, "Some volume is still open.");
2530                                 return (-4);
2531                         }
2532
2533                         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2534                                 if (disk->d_consumer)
2535                                         ddf_meta_erase(disk->d_consumer);
2536                         }
2537                         g_raid_destroy_node(sc, 0);
2538                         return (0);
2539                 }
2540
2541                 /* Destroy specified volume. If it was last - all node. */
2542                 if (*nargs > 2) {
2543                         gctl_error(req, "Invalid number of arguments.");
2544                         return (-1);
2545                 }
2546                 volname = gctl_get_asciiparam(req,
2547                     nodename != NULL ? "arg1" : "arg0");
2548                 if (volname == NULL) {
2549                         gctl_error(req, "No volume name.");
2550                         return (-2);
2551                 }
2552
2553                 /* Search for volume. */
2554                 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2555                         if (strcmp(vol->v_name, volname) == 0)
2556                                 break;
2557                         pp = vol->v_provider;
2558                         if (pp == NULL)
2559                                 continue;
2560                         if (strcmp(pp->name, volname) == 0)
2561                                 break;
2562                         if (strncmp(pp->name, "raid/", 5) == 0 &&
2563                             strcmp(pp->name + 5, volname) == 0)
2564                                 break;
2565                 }
2566                 if (vol == NULL) {
2567                         i = strtol(volname, &tmp, 10);
2568                         if (verb != volname && tmp[0] == 0) {
2569                                 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2570                                         if (vol->v_global_id == i)
2571                                                 break;
2572                                 }
2573                         }
2574                 }
2575                 if (vol == NULL) {
2576                         gctl_error(req, "Volume '%s' not found.", volname);
2577                         return (-3);
2578                 }
2579
2580                 /* Check if volume is still open. */
2581                 force = gctl_get_paraml(req, "force", sizeof(*force));
2582                 if (force != NULL && *force == 0 &&
2583                     vol->v_provider_open != 0) {
2584                         gctl_error(req, "Volume is still open.");
2585                         return (-4);
2586                 }
2587
2588                 /* Destroy volume and potentially node. */
2589                 i = 0;
2590                 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2591                         i++;
2592                 if (i >= 2) {
2593                         g_raid_destroy_volume(vol);
2594                         g_raid_md_ddf_purge_disks(sc);
2595                         g_raid_md_write_ddf(md, NULL, NULL, NULL);
2596                 } else {
2597                         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2598                                 if (disk->d_consumer)
2599                                         ddf_meta_erase(disk->d_consumer);
2600                         }
2601                         g_raid_destroy_node(sc, 0);
2602                 }
2603                 return (0);
2604         }
2605         if (strcmp(verb, "remove") == 0 ||
2606             strcmp(verb, "fail") == 0) {
2607                 if (*nargs < 2) {
2608                         gctl_error(req, "Invalid number of arguments.");
2609                         return (-1);
2610                 }
2611                 for (i = 1; i < *nargs; i++) {
2612                         snprintf(arg, sizeof(arg), "arg%d", i);
2613                         diskname = gctl_get_asciiparam(req, arg);
2614                         if (diskname == NULL) {
2615                                 gctl_error(req, "No disk name (%s).", arg);
2616                                 error = -2;
2617                                 break;
2618                         }
2619                         if (strncmp(diskname, _PATH_DEV, 5) == 0)
2620                                 diskname += 5;
2621
2622                         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2623                                 if (disk->d_consumer != NULL && 
2624                                     disk->d_consumer->provider != NULL &&
2625                                     strcmp(disk->d_consumer->provider->name,
2626                                      diskname) == 0)
2627                                         break;
2628                         }
2629                         if (disk == NULL) {
2630                                 gctl_error(req, "Disk '%s' not found.",
2631                                     diskname);
2632                                 error = -3;
2633                                 break;
2634                         }
2635
2636                         if (strcmp(verb, "fail") == 0) {
2637                                 g_raid_md_fail_disk_ddf(md, NULL, disk);
2638                                 continue;
2639                         }
2640
2641                         /* Erase metadata on deleting disk and destroy it. */
2642                         ddf_meta_erase(disk->d_consumer);
2643                         g_raid_destroy_disk(disk);
2644                 }
2645                 g_raid_md_ddf_purge_volumes(sc);
2646
2647                 /* Write updated metadata to remaining disks. */
2648                 g_raid_md_write_ddf(md, NULL, NULL, NULL);
2649
2650                 /* Check if anything left. */
2651                 if (g_raid_ndisks(sc, -1) == 0)
2652                         g_raid_destroy_node(sc, 0);
2653                 else
2654                         g_raid_md_ddf_refill(sc);
2655                 return (error);
2656         }
2657         if (strcmp(verb, "insert") == 0) {
2658                 if (*nargs < 2) {
2659                         gctl_error(req, "Invalid number of arguments.");
2660                         return (-1);
2661                 }
2662                 for (i = 1; i < *nargs; i++) {
2663                         /* Get disk name. */
2664                         snprintf(arg, sizeof(arg), "arg%d", i);
2665                         diskname = gctl_get_asciiparam(req, arg);
2666                         if (diskname == NULL) {
2667                                 gctl_error(req, "No disk name (%s).", arg);
2668                                 error = -3;
2669                                 break;
2670                         }
2671
2672                         /* Try to find provider with specified name. */
2673                         g_topology_lock();
2674                         cp = g_raid_open_consumer(sc, diskname);
2675                         if (cp == NULL) {
2676                                 gctl_error(req, "Can't open disk '%s'.",
2677                                     diskname);
2678                                 g_topology_unlock();
2679                                 error = -4;
2680                                 break;
2681                         }
2682                         pp = cp->provider;
2683                         g_topology_unlock();
2684
2685                         pd = malloc(sizeof(*pd), M_MD_DDF, M_WAITOK | M_ZERO);
2686
2687                         disk = g_raid_create_disk(sc);
2688                         disk->d_consumer = cp;
2689                         disk->d_md_data = (void *)pd;
2690                         cp->private = disk;
2691
2692                         g_raid_get_disk_info(disk);
2693
2694                         /* Welcome the "new" disk. */
2695                         g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
2696                         ddf_meta_create(disk, &mdi->mdio_meta);
2697                         sa = ddf_meta_find_sa(&pd->pd_meta, 1);
2698                         if (sa != NULL) {
2699                                 SET32D(&pd->pd_meta, sa->Signature,
2700                                     DDF_SA_SIGNATURE);
2701                                 SET8D(&pd->pd_meta, sa->Spare_Type, 0);
2702                                 SET16D(&pd->pd_meta, sa->Populated_SAEs, 0);
2703                                 SET16D(&pd->pd_meta, sa->MAX_SAE_Supported,
2704                                     (GET16(&pd->pd_meta, hdr->Configuration_Record_Length) *
2705                                      pd->pd_meta.sectorsize -
2706                                      sizeof(struct ddf_sa_record)) /
2707                                     sizeof(struct ddf_sa_entry));
2708                         }
2709                         if (mdi->mdio_meta.hdr == NULL)
2710                                 ddf_meta_copy(&mdi->mdio_meta, &pd->pd_meta);
2711                         else
2712                                 ddf_meta_update(&mdi->mdio_meta, &pd->pd_meta);
2713                         g_raid_md_write_ddf(md, NULL, NULL, NULL);
2714                         g_raid_md_ddf_refill(sc);
2715                 }
2716                 return (error);
2717         }
2718         return (-100);
2719 }
2720
2721 static int
2722 g_raid_md_write_ddf(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2723     struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2724 {
2725         struct g_raid_softc *sc;
2726         struct g_raid_volume *vol;
2727         struct g_raid_subdisk *sd;
2728         struct g_raid_disk *disk;
2729         struct g_raid_md_ddf_perdisk *pd;
2730         struct g_raid_md_ddf_pervolume *pv;
2731         struct g_raid_md_ddf_object *mdi;
2732         struct ddf_meta *gmeta;
2733         struct ddf_vol_meta *vmeta;
2734         struct ddf_vdc_record *vdc;
2735         struct ddf_sa_record *sa;
2736         uint64_t *val2;
2737         int i, j, pos, bvd, size;
2738
2739         sc = md->mdo_softc;
2740         mdi = (struct g_raid_md_ddf_object *)md;
2741         gmeta = &mdi->mdio_meta;
2742
2743         if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2744                 return (0);
2745
2746         /*
2747          * Clear disk flags to let only really needed ones to be reset.
2748          * Do it only if there are no volumes in starting state now,
2749          * as they can update disk statuses yet and we may kill innocent.
2750          */
2751         if (mdi->mdio_starting == 0) {
2752                 for (i = 0; i < GET16(gmeta, pdr->Populated_PDEs); i++) {
2753                         if (isff(gmeta->pdr->entry[i].PD_GUID, 24))
2754                                 continue;
2755                         SET16(gmeta, pdr->entry[i].PD_Type,
2756                             GET16(gmeta, pdr->entry[i].PD_Type) &
2757                             ~(DDF_PDE_PARTICIPATING |
2758                               DDF_PDE_GLOBAL_SPARE | DDF_PDE_CONFIG_SPARE));
2759                         if ((GET16(gmeta, pdr->entry[i].PD_State) &
2760                             DDF_PDE_PFA) == 0)
2761                                 SET16(gmeta, pdr->entry[i].PD_State, 0);
2762                 }
2763         }
2764
2765         /* Generate/update new per-volume metadata. */
2766         TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2767                 pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2768                 if (vol->v_stopping || !pv->pv_started)
2769                         continue;
2770                 vmeta = &pv->pv_meta;
2771
2772                 SET32(vmeta, vdc->Sequence_Number,
2773                     GET32(vmeta, vdc->Sequence_Number) + 1);
2774                 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E &&
2775                     vol->v_disks_count % 2 == 0)
2776                         SET16(vmeta, vdc->Primary_Element_Count, 2);
2777                 else
2778                         SET16(vmeta, vdc->Primary_Element_Count,
2779                             vol->v_disks_count);
2780                 SET8(vmeta, vdc->Stripe_Size,
2781                     ffs(vol->v_strip_size / vol->v_sectorsize) - 1);
2782                 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E &&
2783                     vol->v_disks_count % 2 == 0) {
2784                         SET8(vmeta, vdc->Primary_RAID_Level,
2785                             DDF_VDCR_RAID1);
2786                         SET8(vmeta, vdc->RLQ, 0);
2787                         SET8(vmeta, vdc->Secondary_Element_Count,
2788                             vol->v_disks_count / 2);
2789                         SET8(vmeta, vdc->Secondary_RAID_Level, 0);
2790                 } else {
2791                         SET8(vmeta, vdc->Primary_RAID_Level,
2792                             vol->v_raid_level);
2793                         SET8(vmeta, vdc->RLQ,
2794                             vol->v_raid_level_qualifier);
2795                         SET8(vmeta, vdc->Secondary_Element_Count, 1);
2796                         SET8(vmeta, vdc->Secondary_RAID_Level, 0);
2797                 }
2798                 SET8(vmeta, vdc->Secondary_Element_Seq, 0);
2799                 SET64(vmeta, vdc->Block_Count, 0);
2800                 SET64(vmeta, vdc->VD_Size, vol->v_mediasize / vol->v_sectorsize);
2801                 SET16(vmeta, vdc->Block_Size, vol->v_sectorsize);
2802                 SET8(vmeta, vdc->Rotate_Parity_count,
2803                     fls(vol->v_rotate_parity) - 1);
2804                 SET8(vmeta, vdc->MDF_Parity_Disks, vol->v_mdf_pdisks);
2805                 SET16(vmeta, vdc->MDF_Parity_Generator_Polynomial,
2806                     vol->v_mdf_polynomial);
2807                 SET8(vmeta, vdc->MDF_Constant_Generation_Method,
2808                     vol->v_mdf_method);
2809
2810                 SET16(vmeta, vde->VD_Number, vol->v_global_id);
2811                 if (vol->v_state <= G_RAID_VOLUME_S_BROKEN)
2812                         SET8(vmeta, vde->VD_State, DDF_VDE_FAILED);
2813                 else if (vol->v_state <= G_RAID_VOLUME_S_DEGRADED)
2814                         SET8(vmeta, vde->VD_State, DDF_VDE_DEGRADED);
2815                 else if (vol->v_state <= G_RAID_VOLUME_S_SUBOPTIMAL)
2816                         SET8(vmeta, vde->VD_State, DDF_VDE_PARTIAL);
2817                 else
2818                         SET8(vmeta, vde->VD_State, DDF_VDE_OPTIMAL);
2819                 if (vol->v_dirty ||
2820                     g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_STALE) > 0 ||
2821                     g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_RESYNC) > 0)
2822                         SET8(vmeta, vde->VD_State,
2823                             GET8(vmeta, vde->VD_State) | DDF_VDE_DIRTY);
2824                 SET8(vmeta, vde->Init_State, DDF_VDE_INIT_FULL); // XXX
2825                 ddf_meta_put_name(vmeta, vol->v_name);
2826
2827                 for (i = 0; i < vol->v_disks_count; i++) {
2828                         sd = &vol->v_subdisks[i];
2829                         bvd = i / GET16(vmeta, vdc->Primary_Element_Count);
2830                         pos = i % GET16(vmeta, vdc->Primary_Element_Count);
2831                         disk = sd->sd_disk;
2832                         if (disk != NULL) {
2833                                 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2834                                 if (vmeta->bvdc[bvd] == NULL) {
2835                                         size = GET16(vmeta,
2836                                             hdr->Configuration_Record_Length) *
2837                                             vmeta->sectorsize;
2838                                         vmeta->bvdc[bvd] = malloc(size,
2839                                             M_MD_DDF, M_WAITOK);
2840                                         memset(vmeta->bvdc[bvd], 0xff, size);
2841                                 }
2842                                 memcpy(vmeta->bvdc[bvd], vmeta->vdc,
2843                                     sizeof(struct ddf_vdc_record));
2844                                 SET8(vmeta, bvdc[bvd]->Secondary_Element_Seq, bvd);
2845                                 SET64(vmeta, bvdc[bvd]->Block_Count,
2846                                     sd->sd_size / vol->v_sectorsize);
2847                                 SET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos],
2848                                     GET32(&pd->pd_meta, pdd->PD_Reference));
2849                                 val2 = (uint64_t *)&(vmeta->bvdc[bvd]->Physical_Disk_Sequence[
2850                                     GET16(vmeta, hdr->Max_Primary_Element_Entries)]);
2851                                 SET64P(vmeta, val2 + pos,
2852                                     sd->sd_offset / vol->v_sectorsize);
2853                         }
2854                         if (vmeta->bvdc[bvd] == NULL)
2855                                 continue;
2856
2857                         j = ddf_meta_find_pd(gmeta, NULL,
2858                             GET32(vmeta, bvdc[bvd]->Physical_Disk_Sequence[pos]));
2859                         if (j < 0)
2860                                 continue;
2861                         SET16(gmeta, pdr->entry[j].PD_Type,
2862                             GET16(gmeta, pdr->entry[j].PD_Type) |
2863                             DDF_PDE_PARTICIPATING);
2864                         if (sd->sd_state == G_RAID_SUBDISK_S_NONE)
2865                                 SET16(gmeta, pdr->entry[j].PD_State,
2866                                     GET16(gmeta, pdr->entry[j].PD_State) |
2867                                     (DDF_PDE_FAILED | DDF_PDE_MISSING));
2868                         else if (sd->sd_state == G_RAID_SUBDISK_S_FAILED)
2869                                 SET16(gmeta, pdr->entry[j].PD_State,
2870                                     GET16(gmeta, pdr->entry[j].PD_State) |
2871                                     (DDF_PDE_FAILED | DDF_PDE_PFA));
2872                         else if (sd->sd_state <= G_RAID_SUBDISK_S_REBUILD)
2873                                 SET16(gmeta, pdr->entry[j].PD_State,
2874                                     GET16(gmeta, pdr->entry[j].PD_State) |
2875                                     DDF_PDE_REBUILD);
2876                         else
2877                                 SET16(gmeta, pdr->entry[j].PD_State,
2878                                     GET16(gmeta, pdr->entry[j].PD_State) |
2879                                     DDF_PDE_ONLINE);
2880                 }
2881         }
2882
2883         /* Mark spare and failed disks as such. */
2884         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2885                 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2886                 i = ddf_meta_find_pd(gmeta, NULL,
2887                     GET32(&pd->pd_meta, pdd->PD_Reference));
2888                 if (i < 0)
2889                         continue;
2890                 if (disk->d_state == G_RAID_DISK_S_FAILED) {
2891                         SET16(gmeta, pdr->entry[i].PD_State,
2892                             GET16(gmeta, pdr->entry[i].PD_State) |
2893                             (DDF_PDE_FAILED | DDF_PDE_PFA));
2894                 }
2895                 if (disk->d_state != G_RAID_DISK_S_SPARE)
2896                         continue;
2897                 sa = ddf_meta_find_sa(&pd->pd_meta, 0);
2898                 if (sa == NULL ||
2899                     (GET8D(&pd->pd_meta, sa->Spare_Type) &
2900                      DDF_SAR_TYPE_DEDICATED) == 0) {
2901                         SET16(gmeta, pdr->entry[i].PD_Type,
2902                             GET16(gmeta, pdr->entry[i].PD_Type) |
2903                             DDF_PDE_GLOBAL_SPARE);
2904                 } else {
2905                         SET16(gmeta, pdr->entry[i].PD_Type,
2906                             GET16(gmeta, pdr->entry[i].PD_Type) |
2907                             DDF_PDE_CONFIG_SPARE);
2908                 }
2909                 SET16(gmeta, pdr->entry[i].PD_State,
2910                     GET16(gmeta, pdr->entry[i].PD_State) |
2911                     DDF_PDE_ONLINE);
2912         }
2913
2914         /* Remove disks without "participating" flag (unused). */
2915         for (i = 0, j = -1; i < GET16(gmeta, pdr->Populated_PDEs); i++) {
2916                 if (isff(gmeta->pdr->entry[i].PD_GUID, 24))
2917                         continue;
2918                 if ((GET16(gmeta, pdr->entry[i].PD_Type) &
2919                     (DDF_PDE_PARTICIPATING |
2920                      DDF_PDE_GLOBAL_SPARE | DDF_PDE_CONFIG_SPARE)) != 0 ||
2921                     g_raid_md_ddf_get_disk(sc,
2922                      NULL, GET32(gmeta, pdr->entry[i].PD_Reference)) != NULL)
2923                         j = i;
2924                 else
2925                         memset(&gmeta->pdr->entry[i], 0xff,
2926                             sizeof(struct ddf_pd_entry));
2927         }
2928         SET16(gmeta, pdr->Populated_PDEs, j + 1);
2929
2930         /* Update per-disk metadata and write them. */
2931         TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2932                 pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
2933                 if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
2934                     disk->d_state != G_RAID_DISK_S_SPARE)
2935                         continue;
2936                 /* Update PDR. */
2937                 memcpy(pd->pd_meta.pdr, gmeta->pdr,
2938                     GET32(&pd->pd_meta, hdr->pdr_length) *
2939                     pd->pd_meta.sectorsize);
2940                 /* Update VDR. */
2941                 SET16(&pd->pd_meta, vdr->Populated_VDEs, 0);
2942                 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2943                         if (vol->v_stopping)
2944                                 continue;
2945                         pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2946                         i = ddf_meta_find_vd(&pd->pd_meta,
2947                             pv->pv_meta.vde->VD_GUID);
2948                         if (i < 0)
2949                                 i = ddf_meta_find_vd(&pd->pd_meta, NULL);
2950                         if (i >= 0)
2951                                 memcpy(&pd->pd_meta.vdr->entry[i],
2952                                     pv->pv_meta.vde,
2953                                     sizeof(struct ddf_vd_entry));
2954                 }
2955                 /* Update VDC. */
2956                 if (mdi->mdio_starting == 0) {
2957                         /* Remove all VDCs to restore needed later. */
2958                         j = GETCRNUM(&pd->pd_meta);
2959                         for (i = 0; i < j; i++) {
2960                                 vdc = GETVDCPTR(&pd->pd_meta, i);
2961                                 if (GET32D(&pd->pd_meta, vdc->Signature) !=
2962                                     DDF_VDCR_SIGNATURE)
2963                                         continue;
2964                                 SET32D(&pd->pd_meta, vdc->Signature, 0xffffffff);
2965                         }
2966                 }
2967                 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2968                         vol = sd->sd_volume;
2969                         if (vol->v_stopping)
2970                                 continue;
2971                         pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
2972                         vmeta = &pv->pv_meta;
2973                         vdc = ddf_meta_find_vdc(&pd->pd_meta,
2974                             vmeta->vde->VD_GUID);
2975                         if (vdc == NULL)
2976                                 vdc = ddf_meta_find_vdc(&pd->pd_meta, NULL);
2977                         if (vdc != NULL) {
2978                                 bvd = sd->sd_pos / GET16(vmeta,
2979                                     vdc->Primary_Element_Count);
2980                                 memcpy(vdc, vmeta->bvdc[bvd],
2981                                     GET16(&pd->pd_meta,
2982                                     hdr->Configuration_Record_Length) *
2983                                     pd->pd_meta.sectorsize);
2984                         }
2985                 }
2986                 G_RAID_DEBUG(1, "Writing DDF metadata to %s",
2987                     g_raid_get_diskname(disk));
2988                 g_raid_md_ddf_print(&pd->pd_meta);
2989                 ddf_meta_write(disk->d_consumer, &pd->pd_meta);
2990         }
2991         return (0);
2992 }
2993
2994 static int
2995 g_raid_md_fail_disk_ddf(struct g_raid_md_object *md,
2996     struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2997 {
2998         struct g_raid_softc *sc;
2999         struct g_raid_md_ddf_perdisk *pd;
3000         struct g_raid_subdisk *sd;
3001         int i;
3002
3003         sc = md->mdo_softc;
3004         pd = (struct g_raid_md_ddf_perdisk *)tdisk->d_md_data;
3005
3006         /* We can't fail disk that is not a part of array now. */
3007         if (tdisk->d_state != G_RAID_DISK_S_ACTIVE)
3008                 return (-1);
3009
3010         /*
3011          * Mark disk as failed in metadata and try to write that metadata
3012          * to the disk itself to prevent it's later resurrection as STALE.
3013          */
3014         G_RAID_DEBUG(1, "Writing DDF metadata to %s",
3015             g_raid_get_diskname(tdisk));
3016         i = ddf_meta_find_pd(&pd->pd_meta, NULL, GET32(&pd->pd_meta, pdd->PD_Reference));
3017         SET16(&pd->pd_meta, pdr->entry[i].PD_State, DDF_PDE_FAILED | DDF_PDE_PFA);
3018         if (tdisk->d_consumer != NULL)
3019                 ddf_meta_write(tdisk->d_consumer, &pd->pd_meta);
3020
3021         /* Change states. */
3022         g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
3023         TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
3024                 g_raid_change_subdisk_state(sd,
3025                     G_RAID_SUBDISK_S_FAILED);
3026                 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
3027                     G_RAID_EVENT_SUBDISK);
3028         }
3029
3030         /* Write updated metadata to remaining disks. */
3031         g_raid_md_write_ddf(md, NULL, NULL, tdisk);
3032
3033         g_raid_md_ddf_refill(sc);
3034         return (0);
3035 }
3036
3037 static int
3038 g_raid_md_free_disk_ddf(struct g_raid_md_object *md,
3039     struct g_raid_disk *disk)
3040 {
3041         struct g_raid_md_ddf_perdisk *pd;
3042
3043         pd = (struct g_raid_md_ddf_perdisk *)disk->d_md_data;
3044         ddf_meta_free(&pd->pd_meta);
3045         free(pd, M_MD_DDF);
3046         disk->d_md_data = NULL;
3047         return (0);
3048 }
3049
3050 static int
3051 g_raid_md_free_volume_ddf(struct g_raid_md_object *md,
3052     struct g_raid_volume *vol)
3053 {
3054         struct g_raid_md_ddf_object *mdi;
3055         struct g_raid_md_ddf_pervolume *pv;
3056
3057         mdi = (struct g_raid_md_ddf_object *)md;
3058         pv = (struct g_raid_md_ddf_pervolume *)vol->v_md_data;
3059         ddf_vol_meta_free(&pv->pv_meta);
3060         if (!pv->pv_started) {
3061                 pv->pv_started = 1;
3062                 mdi->mdio_starting--;
3063                 callout_stop(&pv->pv_start_co);
3064         }
3065         free(pv, M_MD_DDF);
3066         vol->v_md_data = NULL;
3067         return (0);
3068 }
3069
3070 static int
3071 g_raid_md_free_ddf(struct g_raid_md_object *md)
3072 {
3073         struct g_raid_md_ddf_object *mdi;
3074
3075         mdi = (struct g_raid_md_ddf_object *)md;
3076         if (!mdi->mdio_started) {
3077                 mdi->mdio_started = 0;
3078                 callout_stop(&mdi->mdio_start_co);
3079                 G_RAID_DEBUG1(1, md->mdo_softc,
3080                     "root_mount_rel %p", mdi->mdio_rootmount);
3081                 root_mount_rel(mdi->mdio_rootmount);
3082                 mdi->mdio_rootmount = NULL;
3083         }
3084         ddf_meta_free(&mdi->mdio_meta);
3085         return (0);
3086 }
3087
3088 G_RAID_MD_DECLARE(ddf, "DDF");