projects / dragonfly.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | history | HEAD
Revert "nataraid(4): Add devstat support."
[dragonfly.git] / sys / dev / disk / nata / ata-raid.c
CommitLineData
c1b3d7c5
TS		 1/*-
2 * Copyright (c) 2000 - 2006 Søren Schmidt <sos@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer,
10 * without modification, immediately at the beginning of the file.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 *
26 * $FreeBSD: src/sys/dev/ata/ata-raid.c,v 1.120 2006/04/15 10:27:41 maxim Exp $
c1b3d7c5
TS		 27 */
28
c1b3d7c5 29#include "opt_ata.h"
5e8604ce 30
c1b3d7c5 31#include <sys/param.h>
c1b3d7c5 32#include <sys/bio.h>
5e8604ce 33#include <sys/buf.h>
d438c7c2 34#include <sys/buf2.h>
c1b3d7c5
TS		 35#include <sys/bus.h>
36#include <sys/conf.h>
5e8604ce 37#include <sys/device.h>
c1b3d7c5 38#include <sys/disk.h>
5e8604ce
TS		 39#include <sys/endian.h>
40#include <sys/libkern.h>
41#include <sys/malloc.h>
42#include <sys/module.h>
43#include <sys/nata.h>
d438c7c2 44#include <sys/spinlock2.h>
5e8604ce
TS		 45#include <sys/systm.h>
46
47#include <vm/pmap.h>
48
49#include <machine/md_var.h>
50
51#include <bus/pci/pcivar.h>
52
53#include "ata-all.h"
54#include "ata-disk.h"
55#include "ata-raid.h"
56#include "ata-pci.h"
57#include "ata_if.h"
58
c1b3d7c5
TS		 59
60/* device structure */
61static d_strategy_t ata_raid_strategy;
62static d_dump_t ata_raid_dump;
5e8604ce 63static struct dev_ops ar_ops = {
47f4bca5 64 { "ar", 0, D_DISK },
c1b3d7c5
TS		 65 .d_open = nullopen,
66 .d_close = nullclose,
67 .d_read = physread,
68 .d_write = physwrite,
69 .d_strategy = ata_raid_strategy,
70 .d_dump = ata_raid_dump,
71};
72
73/* prototypes */
74static void ata_raid_done(struct ata_request *request);
75static void ata_raid_config_changed(struct ar_softc *rdp, int writeback);
76static int ata_raid_status(struct ata_ioc_raid_config *config);
77static int ata_raid_create(struct ata_ioc_raid_config *config);
78static int ata_raid_delete(int array);
79static int ata_raid_addspare(struct ata_ioc_raid_config *config);
80static int ata_raid_rebuild(int array);
81static int ata_raid_read_metadata(device_t subdisk);
82static int ata_raid_write_metadata(struct ar_softc *rdp);
83static int ata_raid_wipe_metadata(struct ar_softc *rdp);
84static int ata_raid_adaptec_read_meta(device_t dev, struct ar_softc **raidp);
85static int ata_raid_hptv2_read_meta(device_t dev, struct ar_softc **raidp);
86static int ata_raid_hptv2_write_meta(struct ar_softc *rdp);
87static int ata_raid_hptv3_read_meta(device_t dev, struct ar_softc **raidp);
88static int ata_raid_intel_read_meta(device_t dev, struct ar_softc **raidp);
89static int ata_raid_intel_write_meta(struct ar_softc *rdp);
90static int ata_raid_ite_read_meta(device_t dev, struct ar_softc **raidp);
91static int ata_raid_jmicron_read_meta(device_t dev, struct ar_softc **raidp);
92static int ata_raid_jmicron_write_meta(struct ar_softc *rdp);
93static int ata_raid_lsiv2_read_meta(device_t dev, struct ar_softc **raidp);
94static int ata_raid_lsiv3_read_meta(device_t dev, struct ar_softc **raidp);
95static int ata_raid_nvidia_read_meta(device_t dev, struct ar_softc **raidp);
96static int ata_raid_promise_read_meta(device_t dev, struct ar_softc **raidp, int native);
97static int ata_raid_promise_write_meta(struct ar_softc *rdp);
98static int ata_raid_sii_read_meta(device_t dev, struct ar_softc **raidp);
99static int ata_raid_sis_read_meta(device_t dev, struct ar_softc **raidp);
100static int ata_raid_sis_write_meta(struct ar_softc *rdp);
101static int ata_raid_via_read_meta(device_t dev, struct ar_softc **raidp);
102static int ata_raid_via_write_meta(struct ar_softc *rdp);
103static struct ata_request *ata_raid_init_request(struct ar_softc *rdp, struct bio *bio);
104static int ata_raid_send_request(struct ata_request *request);
105static int ata_raid_rw(device_t dev, u_int64_t lba, void *data, u_int bcount, int flags);
106static char * ata_raid_format(struct ar_softc *rdp);
107static char * ata_raid_type(struct ar_softc *rdp);
108static char * ata_raid_flags(struct ar_softc *rdp);
109
110/* debugging only */
111static void ata_raid_print_meta(struct ar_softc *meta);
112static void ata_raid_adaptec_print_meta(struct adaptec_raid_conf *meta);
113static void ata_raid_hptv2_print_meta(struct hptv2_raid_conf *meta);
114static void ata_raid_hptv3_print_meta(struct hptv3_raid_conf *meta);
115static void ata_raid_intel_print_meta(struct intel_raid_conf *meta);
116static void ata_raid_ite_print_meta(struct ite_raid_conf *meta);
117static void ata_raid_jmicron_print_meta(struct jmicron_raid_conf *meta);
118static void ata_raid_lsiv2_print_meta(struct lsiv2_raid_conf *meta);
119static void ata_raid_lsiv3_print_meta(struct lsiv3_raid_conf *meta);
120static void ata_raid_nvidia_print_meta(struct nvidia_raid_conf *meta);
121static void ata_raid_promise_print_meta(struct promise_raid_conf *meta);
122static void ata_raid_sii_print_meta(struct sii_raid_conf *meta);
123static void ata_raid_sis_print_meta(struct sis_raid_conf *meta);
124static void ata_raid_via_print_meta(struct via_raid_conf *meta);
125
126/* internal vars */
127static struct ar_softc *ata_raid_arrays[MAX_ARRAYS];
128static MALLOC_DEFINE(M_AR, "ar_driver", "ATA PseudoRAID driver");
129static devclass_t ata_raid_sub_devclass;
130static int testing = 0;
131
132static void
133ata_raid_attach(struct ar_softc *rdp, int writeback)
134{
a688b15c 135 struct disk_info info;
5e8604ce 136 cdev_t cdev;
c1b3d7c5
TS		 137 char buffer[32];
138 int disk;
139
140 spin_init(&rdp->lock);
141 ata_raid_config_changed(rdp, writeback);
142
143 /* sanitize arrays total_size % (width * interleave) == 0 */
144 if (rdp->type == AR_T_RAID0 || rdp->type == AR_T_RAID01 ||
145 rdp->type == AR_T_RAID5) {
146 rdp->total_sectors = (rdp->total_sectors/(rdp->interleave*rdp->width))*
147 (rdp->interleave * rdp->width);
f8c7a42d 148 ksprintf(buffer, " (stripe %d KB)",
c1b3d7c5
TS		 149 (rdp->interleave * DEV_BSIZE) / 1024);
150 }
151 else
152 buffer[0] = '\0';
603e3a1c 153 /* XXX TGEN add devstats? */
a688b15c 154 cdev = disk_create(rdp->lun, &rdp->disk, &ar_ops);
5e8604ce
TS		 155 cdev->si_drv1 = rdp;
156 cdev->si_iosize_max = 128 * DEV_BSIZE;
157 rdp->cdev = cdev;
a688b15c
MD		 158
159 bzero(&info, sizeof(info));
160 info.d_media_blksize = DEV_BSIZE; /* mandatory */
161 info.d_media_blocks = rdp->total_sectors;
162
163 info.d_secpertrack = rdp->sectors; /* optional */
164 info.d_nheads = rdp->heads;
165 info.d_ncylinders = rdp->total_sectors/(rdp->heads*rdp->sectors);
166 info.d_secpercyl = rdp->sectors * rdp->heads;
c1b3d7c5 167
e3869ec7 168 kprintf("ar%d: %juMB <%s %s%s> status: %s\n", rdp->lun,
c1b3d7c5
TS		 169 rdp->total_sectors / ((1024L * 1024L) / DEV_BSIZE),
170 ata_raid_format(rdp), ata_raid_type(rdp),
171 buffer, ata_raid_flags(rdp));
172
173 if (testing || bootverbose)
e3869ec7 174 kprintf("ar%d: %ju sectors [%dC/%dH/%dS] <%s> subdisks defined as:\n",
c1b3d7c5
TS		 175 rdp->lun, rdp->total_sectors,
176 rdp->cylinders, rdp->heads, rdp->sectors, rdp->name);
177
178 for (disk = 0; disk < rdp->total_disks; disk++) {
e3869ec7 179 kprintf("ar%d: disk%d ", rdp->lun, disk);
c1b3d7c5
TS		 180 if (rdp->disks[disk].dev) {
181 if (rdp->disks[disk].flags & AR_DF_PRESENT) {
182 /* status of this disk in the array */
183 if (rdp->disks[disk].flags & AR_DF_ONLINE)
e3869ec7 184 kprintf("READY ");
c1b3d7c5 185 else if (rdp->disks[disk].flags & AR_DF_SPARE)
e3869ec7 186 kprintf("SPARE ");
c1b3d7c5 187 else
e3869ec7 188 kprintf("FREE ");
c1b3d7c5
TS		 189
190 /* what type of disk is this in the array */
191 switch (rdp->type) {
192 case AR_T_RAID1:
193 case AR_T_RAID01:
194 if (disk < rdp->width)
e3869ec7 195 kprintf("(master) ");
c1b3d7c5 196 else
e3869ec7 197 kprintf("(mirror) ");
c1b3d7c5
TS		 198 }
199
200 /* which physical disk is used */
e3869ec7 201 kprintf("using %s at ata%d-%s\n",
c1b3d7c5
TS		 202 device_get_nameunit(rdp->disks[disk].dev),
203 device_get_unit(device_get_parent(rdp->disks[disk].dev)),
204 (((struct ata_device *)
205 device_get_softc(rdp->disks[disk].dev))->unit ==
206 ATA_MASTER) ? "master" : "slave");
207 }
208 else if (rdp->disks[disk].flags & AR_DF_ASSIGNED)
e3869ec7 209 kprintf("DOWN\n");
c1b3d7c5 210 else
e3869ec7 211 kprintf("INVALID no RAID config on this subdisk\n");
c1b3d7c5
TS		 212 }
213 else
e3869ec7 214 kprintf("DOWN no device found for this subdisk\n");
c1b3d7c5 215 }
a688b15c
MD		 216
217 disk_setdiskinfo(&rdp->disk, &info);
c1b3d7c5
TS		 218}
219
220/*
221 * ATA PseudoRAID ioctl function. Note that this does not need to be adjusted
222 * to the dev_ops way, because it's just chained from the generic ata ioctl.
223 */
224static int
225ata_raid_ioctl(u_long cmd, caddr_t data)
226{
227 struct ata_ioc_raid_config *config = (struct ata_ioc_raid_config *)data;
228 int *lun = (int *)data;
229 int error = EOPNOTSUPP;
230
231 switch (cmd) {
232 case IOCATARAIDSTATUS:
233 error = ata_raid_status(config);
234 break;
235
236 case IOCATARAIDCREATE:
237 error = ata_raid_create(config);
238 break;
239
240 case IOCATARAIDDELETE:
241 error = ata_raid_delete(*lun);
242 break;
243
244 case IOCATARAIDADDSPARE:
245 error = ata_raid_addspare(config);
246 break;
247
248 case IOCATARAIDREBUILD:
249 error = ata_raid_rebuild(*lun);
250 break;
251 }
252 return error;
253}
254
b106cb48
MD		 255static int
256ata_raid_flush(struct ar_softc *rdp, struct bio *bp)
257{
258 struct ata_request *request;
259 device_t dev;
260 int disk, error;
261
262 error = 0;
60233e58 263 bp->bio_driver_info = NULL;
b106cb48
MD		 264
265 for (disk = 0; disk < rdp->total_disks; disk++) {
266 if ((dev = rdp->disks[disk].dev) != NULL)
267 bp->bio_driver_info = (void *)((intptr_t)bp->bio_driver_info + 1);
268 }
269 for (disk = 0; disk < rdp->total_disks; disk++) {
270 if ((dev = rdp->disks[disk].dev) == NULL)
271 continue;
272 if (!(request = ata_raid_init_request(rdp, bp)))
273 return ENOMEM;
274 request->dev = dev;
275 request->u.ata.command = ATA_FLUSHCACHE;
276 request->u.ata.lba = 0;
277 request->u.ata.count = 0;
278 request->u.ata.feature = 0;
279 request->timeout = 1;
280 request->retries = 0;
281 request->flags |= ATA_R_ORDERED | ATA_R_DIRECT;
282 ata_queue_request(request);
283 }
284 return 0;
285}
286
c1b3d7c5
TS		 287/*
288 * XXX TGEN there are a lot of offset -> block number conversions going on
289 * here, which is suboptimal.
290 */
291static int
292ata_raid_strategy(struct dev_strategy_args *ap)
293{
5e8604ce 294 struct ar_softc *rdp = ap->a_head.a_dev->si_drv1;
c1b3d7c5
TS		 295 struct bio *bp = ap->a_bio;
296 struct buf *bbp = bp->bio_buf;
c1b3d7c5
TS		 297 struct ata_request *request;
298 caddr_t data;
299 u_int64_t blkno, lba, blk = 0;
300 int count, chunk, drv, par = 0, change = 0;
301
b106cb48
MD		 302 if (bbp->b_cmd == BUF_CMD_FLUSH) {
303 int error;
304
305 error = ata_raid_flush(rdp, bp);
306 if (error != 0) {
307 bbp->b_flags |= B_ERROR;
308 bbp->b_error = error;
309 biodone(bp);
310 }
311 return(0);
312 }
313
c1b3d7c5
TS		 314 if (!(rdp->status & AR_S_READY) ||
315 (bbp->b_cmd != BUF_CMD_READ && bbp->b_cmd != BUF_CMD_WRITE)) {
316 bbp->b_flags |= B_ERROR;
317 bbp->b_error = EIO;
318 biodone(bp);
5e8604ce 319 return(0);
c1b3d7c5
TS		 320 }
321
322 bbp->b_resid = bbp->b_bcount;
323 for (count = howmany(bbp->b_bcount, DEV_BSIZE),
324 /* bio_offset is byte granularity, convert */
325 blkno = (u_int64_t)(bp->bio_offset >> DEV_BSHIFT),
326 data = bbp->b_data;
327 count > 0;
328 count -= chunk, blkno += chunk, data += (chunk * DEV_BSIZE)) {
329
330 switch (rdp->type) {
331 case AR_T_RAID1:
332 drv = 0;
333 lba = blkno;
334 chunk = count;
335 break;
336
337 case AR_T_JBOD:
338 case AR_T_SPAN:
339 drv = 0;
340 lba = blkno;
341 while (lba >= rdp->disks[drv].sectors)
342 lba -= rdp->disks[drv++].sectors;
343 chunk = min(rdp->disks[drv].sectors - lba, count);
344 break;
345
346 case AR_T_RAID0:
347 case AR_T_RAID01:
348 chunk = blkno % rdp->interleave;
349 drv = (blkno / rdp->interleave) % rdp->width;
350 lba = (((blkno/rdp->interleave)/rdp->width)*rdp->interleave)+chunk;
351 chunk = min(count, rdp->interleave - chunk);
352 break;
353
354 case AR_T_RAID5:
355 drv = (blkno / rdp->interleave) % (rdp->width - 1);
356 par = rdp->width - 1 -
357 (blkno / (rdp->interleave * (rdp->width - 1))) % rdp->width;
358 if (drv >= par)
359 drv++;
360 lba = ((blkno/rdp->interleave)/(rdp->width-1))*(rdp->interleave) +
361 ((blkno%(rdp->interleave*(rdp->width-1)))%rdp->interleave);
362 chunk = min(count, rdp->interleave - (lba % rdp->interleave));
363 break;
364
365 default:
e3869ec7 366 kprintf("ar%d: unknown array type in ata_raid_strategy\n", rdp->lun);
c1b3d7c5
TS		 367 bbp->b_flags |= B_ERROR;
368 bbp->b_error = EIO;
369 biodone(bp);
5e8604ce 370 return(0);
c1b3d7c5
TS		 371 }
372
373 /* offset on all but "first on HPTv2" */
374 if (!(drv == 0 && rdp->format == AR_F_HPTV2_RAID))
375 lba += rdp->offset_sectors;
376
377 if (!(request = ata_raid_init_request(rdp, bp))) {
378 bbp->b_flags |= B_ERROR;
379 bbp->b_error = EIO;
380 biodone(bp);
5e8604ce 381 return(0);
c1b3d7c5
TS		 382 }
383 request->data = data;
384 request->bytecount = chunk * DEV_BSIZE;
385 request->u.ata.lba = lba;
386 request->u.ata.count = request->bytecount / DEV_BSIZE;
387
388 switch (rdp->type) {
389 case AR_T_JBOD:
390 case AR_T_SPAN:
391 case AR_T_RAID0:
392 if (((rdp->disks[drv].flags & (AR_DF_PRESENT|AR_DF_ONLINE)) ==
393 (AR_DF_PRESENT|AR_DF_ONLINE) && !rdp->disks[drv].dev)) {
394 rdp->disks[drv].flags &= ~AR_DF_ONLINE;
395 ata_raid_config_changed(rdp, 1);
396 ata_free_request(request);
397 bbp->b_flags |= B_ERROR;
398 bbp->b_error = EIO;
399 biodone(bp);
5e8604ce 400 return(0);
c1b3d7c5
TS		 401 }
402 request->this = drv;
403 request->dev = rdp->disks[request->this].dev;
404 ata_raid_send_request(request);
405 break;
406
407 case AR_T_RAID1:
408 case AR_T_RAID01:
409 if ((rdp->disks[drv].flags &
410 (AR_DF_PRESENT|AR_DF_ONLINE))==(AR_DF_PRESENT|AR_DF_ONLINE) &&
411 !rdp->disks[drv].dev) {
412 rdp->disks[drv].flags &= ~AR_DF_ONLINE;
413 change = 1;
414 }
415 if ((rdp->disks[drv + rdp->width].flags &
416 (AR_DF_PRESENT|AR_DF_ONLINE))==(AR_DF_PRESENT|AR_DF_ONLINE) &&
417 !rdp->disks[drv + rdp->width].dev) {
418 rdp->disks[drv + rdp->width].flags &= ~AR_DF_ONLINE;
419 change = 1;
420 }
421 if (change)
422 ata_raid_config_changed(rdp, 1);
423 if (!(rdp->status & AR_S_READY)) {
424 ata_free_request(request);
425 bbp->b_flags |= B_ERROR;
426 bbp->b_error = EIO;
427 biodone(bp);
5e8604ce 428 return(0);
c1b3d7c5
TS		 429 }
430
431 if (rdp->status & AR_S_REBUILDING)
432 blk = ((lba / rdp->interleave) * rdp->width) * rdp->interleave +
433 (rdp->interleave * (drv % rdp->width)) +
434 lba % rdp->interleave;;
435
436 if (bbp->b_cmd == BUF_CMD_READ) {
437 int src_online =
438 (rdp->disks[drv].flags & AR_DF_ONLINE);
439 int mir_online =
440 (rdp->disks[drv+rdp->width].flags & AR_DF_ONLINE);
441
442 /* if mirror gone or close to last access on source */
443 if (!mir_online ||
444 ((src_online) &&
445 ((u_int64_t)(bp->bio_offset >> DEV_BSHIFT)) >=
446 (rdp->disks[drv].last_lba - AR_PROXIMITY) &&
447 ((u_int64_t)(bp->bio_offset >> DEV_BSHIFT)) <=
448 (rdp->disks[drv].last_lba + AR_PROXIMITY))) {
449 rdp->toggle = 0;
450 }
451 /* if source gone or close to last access on mirror */
452 else if (!src_online ||
453 ((mir_online) &&
454 ((u_int64_t)(bp->bio_offset >> DEV_BSHIFT)) >=
455 (rdp->disks[drv+rdp->width].last_lba-AR_PROXIMITY) &&
456 ((u_int64_t)(bp->bio_offset >> DEV_BSHIFT)) <=
457 (rdp->disks[drv+rdp->width].last_lba+AR_PROXIMITY))) {
458 drv += rdp->width;
459 rdp->toggle = 1;
460 }
461 /* not close to any previous access, toggle */
462 else {
463 if (rdp->toggle)
464 rdp->toggle = 0;
465 else {
466 drv += rdp->width;
467 rdp->toggle = 1;
468 }
469 }
470
471 if ((rdp->status & AR_S_REBUILDING) &&
472 (blk <= rdp->rebuild_lba) &&
473 ((blk + chunk) > rdp->rebuild_lba)) {
474 struct ata_composite *composite;
475 struct ata_request *rebuild;
476 int this;
477
478 /* figure out what part to rebuild */
479 if (drv < rdp->width)
480 this = drv + rdp->width;
481 else
482 this = drv - rdp->width;
483
484 /* do we have a spare to rebuild on ? */
485 if (rdp->disks[this].flags & AR_DF_SPARE) {
486 if ((composite = ata_alloc_composite())) {
487 if ((rebuild = ata_alloc_request())) {
488 rdp->rebuild_lba = blk + chunk;
489 bcopy(request, rebuild,
490 sizeof(struct ata_request));
491 rebuild->this = this;
492 rebuild->dev = rdp->disks[this].dev;
493 rebuild->flags &= ~ATA_R_READ;
494 rebuild->flags |= ATA_R_WRITE;
495 spin_init(&composite->lock);
496 composite->residual = request->bytecount;
497 composite->rd_needed |= (1 << drv);
498 composite->wr_depend |= (1 << drv);
499 composite->wr_needed |= (1 << this);
500 composite->request[drv] = request;
501 composite->request[this] = rebuild;
502 request->composite = composite;
503 rebuild->composite = composite;
504 ata_raid_send_request(rebuild);
505 }
506 else {
507 ata_free_composite(composite);
e3869ec7 508 kprintf("DOH! ata_alloc_request failed!\n");
c1b3d7c5
TS		 509 }
510 }
511 else {
e3869ec7 512 kprintf("DOH! ata_alloc_composite failed!\n");
c1b3d7c5
TS		 513 }
514 }
515 else if (rdp->disks[this].flags & AR_DF_ONLINE) {
516 /*
517 * if we got here we are a chunk of a RAID01 that
518 * does not need a rebuild, but we need to increment
519 * the rebuild_lba address to get the rebuild to
520 * move to the next chunk correctly
521 */
522 rdp->rebuild_lba = blk + chunk;
523 }
524 else
e3869ec7 525 kprintf("DOH! we didn't find the rebuild part\n");
c1b3d7c5
TS		 526 }
527 }
528 if (bbp->b_cmd == BUF_CMD_WRITE) {
529 if ((rdp->disks[drv+rdp->width].flags & AR_DF_ONLINE) ||
530 ((rdp->status & AR_S_REBUILDING) &&
531 (rdp->disks[drv+rdp->width].flags & AR_DF_SPARE) &&
532 ((blk < rdp->rebuild_lba) ||
533 ((blk <= rdp->rebuild_lba) &&
534 ((blk + chunk) > rdp->rebuild_lba))))) {
535 if ((rdp->disks[drv].flags & AR_DF_ONLINE) ||
536 ((rdp->status & AR_S_REBUILDING) &&
537 (rdp->disks[drv].flags & AR_DF_SPARE) &&
538 ((blk < rdp->rebuild_lba) ||
539 ((blk <= rdp->rebuild_lba) &&
540 ((blk + chunk) > rdp->rebuild_lba))))) {
541 struct ata_request *mirror;
542 struct ata_composite *composite;
543 int this = drv + rdp->width;
544
545 if ((composite = ata_alloc_composite())) {
546 if ((mirror = ata_alloc_request())) {
547 if ((blk <= rdp->rebuild_lba) &&
548 ((blk + chunk) > rdp->rebuild_lba))
549 rdp->rebuild_lba = blk + chunk;
550 bcopy(request, mirror,
551 sizeof(struct ata_request));
552 mirror->this = this;
553 mirror->dev = rdp->disks[this].dev;
554 spin_init(&composite->lock);
555 composite->residual = request->bytecount;
556 composite->wr_needed |= (1 << drv);
557 composite->wr_needed |= (1 << this);
558 composite->request[drv] = request;
559 composite->request[this] = mirror;
560 request->composite = composite;
561 mirror->composite = composite;
562 ata_raid_send_request(mirror);
563 rdp->disks[this].last_lba =
564 (u_int64_t)(bp->bio_offset >> DEV_BSHIFT) +
565 chunk;
566 }
567 else {
568 ata_free_composite(composite);
e3869ec7 569 kprintf("DOH! ata_alloc_request failed!\n");
c1b3d7c5
TS		 570 }
571 }
572 else {
e3869ec7 573 kprintf("DOH! ata_alloc_composite failed!\n");
c1b3d7c5
TS		 574 }
575 }
576 else
577 drv += rdp->width;
578 }
579 }
580 request->this = drv;
581 request->dev = rdp->disks[request->this].dev;
582 ata_raid_send_request(request);
d438c7c2 583 rdp->disks[request->this].last_lba =
1a215404 584 ((u_int64_t)(bp->bio_offset) >> DEV_BSHIFT) + chunk;
c1b3d7c5
TS		 585 break;
586
587 case AR_T_RAID5:
588 if (((rdp->disks[drv].flags & (AR_DF_PRESENT|AR_DF_ONLINE)) ==
589 (AR_DF_PRESENT|AR_DF_ONLINE) && !rdp->disks[drv].dev)) {
590 rdp->disks[drv].flags &= ~AR_DF_ONLINE;
591 change = 1;
592 }
593 if (((rdp->disks[par].flags & (AR_DF_PRESENT|AR_DF_ONLINE)) ==
594 (AR_DF_PRESENT|AR_DF_ONLINE) && !rdp->disks[par].dev)) {
595 rdp->disks[par].flags &= ~AR_DF_ONLINE;
596 change = 1;
597 }
598 if (change)
599 ata_raid_config_changed(rdp, 1);
600 if (!(rdp->status & AR_S_READY)) {
601 ata_free_request(request);
602 bbp->b_flags |= B_ERROR;
603 bbp->b_error = EIO;
604 biodone(bp);
5e8604ce 605 return(0);
c1b3d7c5
TS		 606 }
607 if (rdp->status & AR_S_DEGRADED) {
608 /* do the XOR game if possible */
609 }
610 else {
611 request->this = drv;
612 request->dev = rdp->disks[request->this].dev;
613 if (bbp->b_cmd == BUF_CMD_READ) {
614 ata_raid_send_request(request);
615 }
616 if (bbp->b_cmd == BUF_CMD_WRITE) {
617 ata_raid_send_request(request);
5e8604ce 618 /* XXX TGEN no, I don't speak Danish either */
d438c7c2
TS		 619 /*
620 * sikre at læs-modify-skriv til hver disk er atomarisk.
621 * par kopi af request
622 * læse orgdata fra drv
623 * skriv nydata til drv
624 * læse parorgdata fra par
625 * skriv orgdata xor parorgdata xor nydata til par
626 */
c1b3d7c5
TS		 627 }
628 }
629 break;
630
631 default:
e3869ec7 632 kprintf("ar%d: unknown array type in ata_raid_strategy\n", rdp->lun);
c1b3d7c5
TS		 633 }
634 }
5e8604ce
TS		 635
636 return(0);
c1b3d7c5
TS		 637}
638
639static void
640ata_raid_done(struct ata_request *request)
641{
642 struct ar_softc *rdp = request->driver;
643 struct ata_composite *composite = NULL;
644 struct bio *bp = request->bio;
645 struct buf *bbp = bp->bio_buf;
646 int i, mirror, finished = 0;
647
b106cb48
MD		 648 if (bbp->b_cmd == BUF_CMD_FLUSH) {
649 if (bbp->b_error == 0)
650 bbp->b_error = request->result;
651 ata_free_request(request);
652 bp->bio_driver_info = (void *)((intptr_t)bp->bio_driver_info - 1);
653 if ((intptr_t)bp->bio_driver_info == 0) {
654 if (bbp->b_error)
655 bbp->b_flags |= B_ERROR;
656 biodone(bp);
657 }
658 return;
659 }
660
c1b3d7c5
TS		 661 switch (rdp->type) {
662 case AR_T_JBOD:
663 case AR_T_SPAN:
664 case AR_T_RAID0:
665 if (request->result) {
666 rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
667 ata_raid_config_changed(rdp, 1);
668 bbp->b_error = request->result;
669 finished = 1;
670 }
671 else {
672 bbp->b_resid -= request->donecount;
673 if (!bbp->b_resid)
674 finished = 1;
675 }
676 break;
677
678 case AR_T_RAID1:
679 case AR_T_RAID01:
680 if (request->this < rdp->width)
681 mirror = request->this + rdp->width;
682 else
683 mirror = request->this - rdp->width;
684 if (request->result) {
685 rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
686 ata_raid_config_changed(rdp, 1);
687 }
688 if (rdp->status & AR_S_READY) {
689 u_int64_t blk = 0;
690
691 if (rdp->status & AR_S_REBUILDING)
692 blk = ((request->u.ata.lba / rdp->interleave) * rdp->width) *
693 rdp->interleave + (rdp->interleave *
694 (request->this % rdp->width)) +
695 request->u.ata.lba % rdp->interleave;
696
697 if (bbp->b_cmd == BUF_CMD_READ) {
698
699 /* is this a rebuild composite */
700 if ((composite = request->composite)) {
287a8577 701 spin_lock(&composite->lock);
c1b3d7c5
TS		 702
703 /* handle the read part of a rebuild composite */
704 if (request->flags & ATA_R_READ) {
705
706 /* if read failed array is now broken */
707 if (request->result) {
708 rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
709 ata_raid_config_changed(rdp, 1);
710 bbp->b_error = request->result;
711 rdp->rebuild_lba = blk;
712 finished = 1;
713 }
714
715 /* good data, update how far we've gotten */
716 else {
717 bbp->b_resid -= request->donecount;
718 composite->residual -= request->donecount;
719 if (!composite->residual) {
720 if (composite->wr_done & (1 << mirror))
721 finished = 1;
722 }
723 }
724 }
725
726 /* handle the write part of a rebuild composite */
727 else if (request->flags & ATA_R_WRITE) {
728 if (composite->rd_done & (1 << mirror)) {
729 if (request->result) {
e3869ec7 730 kprintf("DOH! rebuild failed\n"); /* XXX SOS */
c1b3d7c5
TS		 731 rdp->rebuild_lba = blk;
732 }
733 if (!composite->residual)
734 finished = 1;
735 }
736 }
287a8577 737 spin_unlock(&composite->lock);
c1b3d7c5
TS		 738 }
739
740 /* if read failed retry on the mirror */
741 else if (request->result) {
742 request->dev = rdp->disks[mirror].dev;
743 request->flags &= ~ATA_R_TIMEOUT;
744 ata_raid_send_request(request);
745 return;
746 }
747
748 /* we have good data */
749 else {
750 bbp->b_resid -= request->donecount;
751 if (!bbp->b_resid)
752 finished = 1;
753 }
754 }
755 else if (bbp->b_cmd == BUF_CMD_WRITE) {
756 /* do we have a mirror or rebuild to deal with ? */
757 if ((composite = request->composite)) {
287a8577 758 spin_lock(&composite->lock);
c1b3d7c5
TS		 759 if (composite->wr_done & (1 << mirror)) {
760 if (request->result) {
761 if (composite->request[mirror]->result) {
e3869ec7 762 kprintf("DOH! all disks failed and got here\n");
c1b3d7c5
TS		 763 bbp->b_error = EIO;
764 }
765 if (rdp->status & AR_S_REBUILDING) {
766 rdp->rebuild_lba = blk;
e3869ec7 767 kprintf("DOH! rebuild failed\n"); /* XXX SOS */
c1b3d7c5
TS		 768 }
769 bbp->b_resid -=
770 composite->request[mirror]->donecount;
771 composite->residual -=
772 composite->request[mirror]->donecount;
773 }
774 else {
775 bbp->b_resid -= request->donecount;
776 composite->residual -= request->donecount;
777 }
778 if (!composite->residual)
779 finished = 1;
780 }
287a8577 781 spin_unlock(&composite->lock);
c1b3d7c5
TS		 782 }
783 /* no mirror we are done */
784 else {
785 bbp->b_resid -= request->donecount;
786 if (!bbp->b_resid)
787 finished = 1;
788 }
789 }
790 }
791 else {
792 /* XXX TGEN bbp->b_flags |= B_ERROR; */
793 bbp->b_error = request->result;
794 biodone(bp);
795 }
796 break;
797
798 case AR_T_RAID5:
799 if (request->result) {
800 rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
801 ata_raid_config_changed(rdp, 1);
802 if (rdp->status & AR_S_READY) {
803 if (bbp->b_cmd == BUF_CMD_READ) {
804 /* do the XOR game to recover data */
805 }
806 if (bbp->b_cmd == BUF_CMD_WRITE) {
807 /* if the parity failed we're OK sortof */
808 /* otherwise wee need to do the XOR long dance */
809 }
810 finished = 1;
811 }
812 else {
813 /* XXX TGEN bbp->b_flags |= B_ERROR; */
814 bbp->b_error = request->result;
815 biodone(bp);
816 }
817 }
818 else {
d438c7c2 819 /* did we have an XOR game going ?? */
c1b3d7c5
TS		 820 bbp->b_resid -= request->donecount;
821 if (!bbp->b_resid)
822 finished = 1;
823 }
824 break;
825
826 default:
e3869ec7 827 kprintf("ar%d: unknown array type in ata_raid_done\n", rdp->lun);
c1b3d7c5
TS		 828 }
829
830 if (finished) {
831 if ((rdp->status & AR_S_REBUILDING) &&
832 rdp->rebuild_lba >= rdp->total_sectors) {
833 int disk;
834
835 for (disk = 0; disk < rdp->total_disks; disk++) {
836 if ((rdp->disks[disk].flags &
837 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE)) ==
838 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE)) {
839 rdp->disks[disk].flags &= ~AR_DF_SPARE;
840 rdp->disks[disk].flags |= AR_DF_ONLINE;
841 }
842 }
843 rdp->status &= ~AR_S_REBUILDING;
844 ata_raid_config_changed(rdp, 1);
845 }
846 if (!bbp->b_resid)
847 biodone(bp);
848 }
849
850 if (composite) {
851 if (finished) {
852 /* we are done with this composite, free all resources */
853 for (i = 0; i < 32; i++) {
854 if (composite->rd_needed & (1 << i) ||
855 composite->wr_needed & (1 << i)) {
856 ata_free_request(composite->request[i]);
857 }
858 }
859 spin_uninit(&composite->lock);
860 ata_free_composite(composite);
861 }
862 }
863 else
864 ata_free_request(request);
865}
866
867static int
868ata_raid_dump(struct dev_dump_args *ap)
869{
b24cd69c
AH		 870 struct ar_softc *rdp = ap->a_head.a_dev->si_drv1;
871 struct buf dbuf;
872 int error = 0;
873 int disk;
874
875 if (ap->a_length == 0) {
876 /* flush subdisk buffers to media */
877 for (disk = 0, error = 0; disk < rdp->total_disks; disk++) {
878 if (rdp->disks[disk].dev) {
879 error |= ata_controlcmd(rdp->disks[disk].dev,
880 ATA_FLUSHCACHE, 0, 0, 0);
881 }
882 }
883 return (error ? EIO : 0);
c1b3d7c5
TS		 884 }
885
5e8604ce 886 bzero(&dbuf, sizeof(struct buf));
5e8604ce 887 initbufbio(&dbuf);
b5d7061d 888 BUF_LOCK(&dbuf, LK_EXCLUSIVE);
c1b3d7c5 889 /* bio_offset is byte granularity, convert block granularity a_blkno */
b24cd69c 890 dbuf.b_bio1.bio_offset = ap->a_offset;
5e8604ce 891 dbuf.b_bio1.bio_caller_info1.ptr = (void *)rdp;
ae8e83e6
MD		 892 dbuf.b_bio1.bio_flags |= BIO_SYNC;
893 dbuf.b_bio1.bio_done = biodone_sync;
b24cd69c
AH		 894 dbuf.b_bcount = ap->a_length;
895 dbuf.b_data = ap->a_virtual;
5e8604ce
TS		 896 dbuf.b_cmd = BUF_CMD_WRITE;
897 dev_dstrategy(rdp->cdev, &dbuf.b_bio1);
898 /* wait for completion, unlock the buffer, check status */
ae8e83e6 899 if (biowait(&dbuf.b_bio1, "dumpw")) {
5e8604ce
TS		 900 BUF_UNLOCK(&dbuf);
901 return(dbuf.b_error ? dbuf.b_error : EIO);
902 }
903 BUF_UNLOCK(&dbuf);
b5d7061d 904 uninitbufbio(&dbuf);
c1b3d7c5 905
b24cd69c 906 return 0;
c1b3d7c5
TS		 907}
908
909static void
910ata_raid_config_changed(struct ar_softc *rdp, int writeback)
911{
912 int disk, count, status;
913
287a8577 914 spin_lock(&rdp->lock);
c1b3d7c5
TS		 915 /* set default all working mode */
916 status = rdp->status;
917 rdp->status &= ~AR_S_DEGRADED;
918 rdp->status |= AR_S_READY;
919
920 /* make sure all lost drives are accounted for */
921 for (disk = 0; disk < rdp->total_disks; disk++) {
922 if (!(rdp->disks[disk].flags & AR_DF_PRESENT))
923 rdp->disks[disk].flags &= ~AR_DF_ONLINE;
924 }
925
926 /* depending on RAID type figure out our health status */
927 switch (rdp->type) {
928 case AR_T_JBOD:
929 case AR_T_SPAN:
930 case AR_T_RAID0:
931 for (disk = 0; disk < rdp->total_disks; disk++)
932 if (!(rdp->disks[disk].flags & AR_DF_ONLINE))
933 rdp->status &= ~AR_S_READY;
934 break;
935
936 case AR_T_RAID1:
937 case AR_T_RAID01:
938 for (disk = 0; disk < rdp->width; disk++) {
939 if (!(rdp->disks[disk].flags & AR_DF_ONLINE) &&
940 !(rdp->disks[disk + rdp->width].flags & AR_DF_ONLINE)) {
941 rdp->status &= ~AR_S_READY;
942 }
943 else if (((rdp->disks[disk].flags & AR_DF_ONLINE) &&
944 !(rdp->disks[disk + rdp->width].flags & AR_DF_ONLINE)) ||
945 (!(rdp->disks[disk].flags & AR_DF_ONLINE) &&
946 (rdp->disks [disk + rdp->width].flags & AR_DF_ONLINE))) {
947 rdp->status |= AR_S_DEGRADED;
948 }
949 }
950 break;
951
952 case AR_T_RAID5:
953 for (count = 0, disk = 0; disk < rdp->total_disks; disk++) {
954 if (!(rdp->disks[disk].flags & AR_DF_ONLINE))
955 count++;
956 }
957 if (count) {
958 if (count > 1)
959 rdp->status &= ~AR_S_READY;
960 else
961 rdp->status |= AR_S_DEGRADED;
962 }
963 break;
964 default:
965 rdp->status &= ~AR_S_READY;
966 }
967
6b7bbefa
MD		 968 /*
969 * Note that when the array breaks so comes up broken we
970 * force a write of the array config to the remaining
971 * drives so that the generation will be incremented past
972 * those of the missing or failed drives (in all cases).
973 */
c1b3d7c5
TS		 974 if (rdp->status != status) {
975 if (!(rdp->status & AR_S_READY)) {
e3869ec7 976 kprintf("ar%d: FAILURE - %s array broken\n",
c1b3d7c5 977 rdp->lun, ata_raid_type(rdp));
6b7bbefa 978 writeback = 1;
c1b3d7c5
TS		 979 }
980 else if (rdp->status & AR_S_DEGRADED) {
981 if (rdp->type & (AR_T_RAID1 | AR_T_RAID01))
e3869ec7 982 kprintf("ar%d: WARNING - mirror", rdp->lun);
c1b3d7c5 983 else
e3869ec7
SW		 984 kprintf("ar%d: WARNING - parity", rdp->lun);
985 kprintf(" protection lost. %s array in DEGRADED mode\n",
c1b3d7c5 986 ata_raid_type(rdp));
6b7bbefa 987 writeback = 1;
c1b3d7c5
TS		 988 }
989 }
287a8577 990 spin_unlock(&rdp->lock);
c1b3d7c5
TS		 991 if (writeback)
992 ata_raid_write_metadata(rdp);
993
994}
995
996static int
997ata_raid_status(struct ata_ioc_raid_config *config)
998{
999 struct ar_softc *rdp;
1000 int i;
1001
1002 if (!(rdp = ata_raid_arrays[config->lun]))
1003 return ENXIO;
1004
1005 config->type = rdp->type;
1006 config->total_disks = rdp->total_disks;
1007 for (i = 0; i < rdp->total_disks; i++ ) {
1008 if ((rdp->disks[i].flags & AR_DF_PRESENT) && rdp->disks[i].dev)
1009 config->disks[i] = device_get_unit(rdp->disks[i].dev);
1010 else
1011 config->disks[i] = -1;
1012 }
1013 config->interleave = rdp->interleave;
1014 config->status = rdp->status;
1015 config->progress = 100 * rdp->rebuild_lba / rdp->total_sectors;
1016 return 0;
1017}
1018
1019static int
1020ata_raid_create(struct ata_ioc_raid_config *config)
1021{
1022 struct ar_softc *rdp;
1023 device_t subdisk;
1024 int array, disk;
1025 int ctlr = 0, disk_size = 0, total_disks = 0;
da44240f 1026 device_t gpdev;
c1b3d7c5
TS		 1027
1028 for (array = 0; array < MAX_ARRAYS; array++) {
1029 if (!ata_raid_arrays[array])
1030 break;
1031 }
1032 if (array >= MAX_ARRAYS)
1033 return ENOSPC;
1034
978400d3
SW		 1035 rdp = (struct ar_softc*)kmalloc(sizeof(struct ar_softc), M_AR,
1036 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1037
1038 for (disk = 0; disk < config->total_disks; disk++) {
1039 if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
1040 config->disks[disk]))) {
1041 struct ata_raid_subdisk *ars = device_get_softc(subdisk);
1042
1043 /* is device already assigned to another array ? */
1044 if (ars->raid[rdp->volume]) {
1045 config->disks[disk] = -1;
d438c7c2 1046 kfree(rdp, M_AR);
c1b3d7c5
TS		 1047 return EBUSY;
1048 }
1049 rdp->disks[disk].dev = device_get_parent(subdisk);
1050
da44240f
MD		 1051 gpdev = GRANDPARENT(rdp->disks[disk].dev);
1052
1053 switch (pci_get_vendor(gpdev)) {
c1b3d7c5
TS		 1054 case ATA_HIGHPOINT_ID:
1055 /*
1056 * we need some way to decide if it should be v2 or v3
1057 * for now just use v2 since the v3 BIOS knows how to
1058 * handle that as well.
1059 */
1060 ctlr = AR_F_HPTV2_RAID;
1061 rdp->disks[disk].sectors = HPTV3_LBA(rdp->disks[disk].dev);
1062 break;
1063
1064 case ATA_INTEL_ID:
1065 ctlr = AR_F_INTEL_RAID;
1066 rdp->disks[disk].sectors = INTEL_LBA(rdp->disks[disk].dev);
1067 break;
1068
1069 case ATA_ITE_ID:
1070 ctlr = AR_F_ITE_RAID;
1071 rdp->disks[disk].sectors = ITE_LBA(rdp->disks[disk].dev);
1072 break;
1073
1074 case ATA_JMICRON_ID:
1075 ctlr = AR_F_JMICRON_RAID;
1076 rdp->disks[disk].sectors = JMICRON_LBA(rdp->disks[disk].dev);
1077 break;
1078
1079 case 0: /* XXX SOS cover up for bug in our PCI code */
1080 case ATA_PROMISE_ID:
1081 ctlr = AR_F_PROMISE_RAID;
1082 rdp->disks[disk].sectors = PROMISE_LBA(rdp->disks[disk].dev);
1083 break;
1084
1085 case ATA_SIS_ID:
1086 ctlr = AR_F_SIS_RAID;
1087 rdp->disks[disk].sectors = SIS_LBA(rdp->disks[disk].dev);
1088 break;
1089
1090 case ATA_ATI_ID:
1091 case ATA_VIA_ID:
1092 ctlr = AR_F_VIA_RAID;
1093 rdp->disks[disk].sectors = VIA_LBA(rdp->disks[disk].dev);
1094 break;
1095
1096 default:
1097 /* XXX SOS
1098 * right, so here we are, we have an ATA chip and we want
1099 * to create a RAID and store the metadata.
1100 * we need to find a way to tell what kind of metadata this
1101 * hardware's BIOS might be using (good ideas are welcomed)
1102 * for now we just use our own native FreeBSD format.
1103 * the only way to get support for the BIOS format is to
1104 * setup the RAID from there, in that case we pickup the
1105 * metadata format from the disks (if we support it).
1106 */
e3869ec7 1107 kprintf("WARNING!! - not able to determine metadata format\n"
c1b3d7c5
TS		 1108 "WARNING!! - Using FreeBSD PseudoRAID metadata\n"
1109 "If that is not what you want, use the BIOS to "
1110 "create the array\n");
1111 ctlr = AR_F_FREEBSD_RAID;
1112 rdp->disks[disk].sectors = PROMISE_LBA(rdp->disks[disk].dev);
1113 break;
1114 }
1115
1116 /* we need all disks to be of the same format */
1117 if ((rdp->format & AR_F_FORMAT_MASK) &&
1118 (rdp->format & AR_F_FORMAT_MASK) != (ctlr & AR_F_FORMAT_MASK)) {
d438c7c2 1119 kfree(rdp, M_AR);
c1b3d7c5
TS		 1120 return EXDEV;
1121 }
1122 else
1123 rdp->format = ctlr;
1124
1125 /* use the smallest disk of the lots size */
1126 /* gigabyte boundry ??? XXX SOS */
1127 if (disk_size)
1128 disk_size = min(rdp->disks[disk].sectors, disk_size);
1129 else
1130 disk_size = rdp->disks[disk].sectors;
1131 rdp->disks[disk].flags =
1132 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
1133
1134 total_disks++;
1135 }
1136 else {
1137 config->disks[disk] = -1;
d438c7c2 1138 kfree(rdp, M_AR);
c1b3d7c5
TS		 1139 return ENXIO;
1140 }
1141 }
1142
1143 if (total_disks != config->total_disks) {
d438c7c2 1144 kfree(rdp, M_AR);
c1b3d7c5
TS		 1145 return ENODEV;
1146 }
1147
1148 switch (config->type) {
1149 case AR_T_JBOD:
1150 case AR_T_SPAN:
1151 case AR_T_RAID0:
1152 break;
1153
1154 case AR_T_RAID1:
1155 if (total_disks != 2) {
d438c7c2 1156 kfree(rdp, M_AR);
c1b3d7c5
TS		 1157 return EPERM;
1158 }
1159 break;
1160
1161 case AR_T_RAID01:
1162 if (total_disks % 2 != 0) {
d438c7c2 1163 kfree(rdp, M_AR);
c1b3d7c5
TS		 1164 return EPERM;
1165 }
1166 break;
1167
1168 case AR_T_RAID5:
1169 if (total_disks < 3) {
d438c7c2 1170 kfree(rdp, M_AR);
c1b3d7c5
TS		 1171 return EPERM;
1172 }
1173 break;
1174
1175 default:
d438c7c2 1176 kfree(rdp, M_AR);
c1b3d7c5
TS		 1177 return EOPNOTSUPP;
1178 }
1179 rdp->type = config->type;
1180 rdp->lun = array;
1181 if (rdp->type == AR_T_RAID0 || rdp->type == AR_T_RAID01 ||
1182 rdp->type == AR_T_RAID5) {
1183 int bit = 0;
1184
1185 while (config->interleave >>= 1)
1186 bit++;
1187 rdp->interleave = 1 << bit;
1188 }
1189 rdp->offset_sectors = 0;
1190
1191 /* values that depend on metadata format */
1192 switch (rdp->format) {
1193 case AR_F_ADAPTEC_RAID:
1194 rdp->interleave = min(max(32, rdp->interleave), 128); /*+*/
1195 break;
1196
1197 case AR_F_HPTV2_RAID:
1198 rdp->interleave = min(max(8, rdp->interleave), 128); /*+*/
1199 rdp->offset_sectors = HPTV2_LBA(x) + 1;
1200 break;
1201
1202 case AR_F_HPTV3_RAID:
1203 rdp->interleave = min(max(32, rdp->interleave), 4096); /*+*/
1204 break;
1205
1206 case AR_F_INTEL_RAID:
1207 rdp->interleave = min(max(8, rdp->interleave), 256); /*+*/
1208 break;
1209
1210 case AR_F_ITE_RAID:
1211 rdp->interleave = min(max(2, rdp->interleave), 128); /*+*/
1212 break;
1213
1214 case AR_F_JMICRON_RAID:
1215 rdp->interleave = min(max(8, rdp->interleave), 256); /*+*/
1216 break;
1217
1218 case AR_F_LSIV2_RAID:
1219 rdp->interleave = min(max(2, rdp->interleave), 4096);
1220 break;
1221
1222 case AR_F_LSIV3_RAID:
1223 rdp->interleave = min(max(2, rdp->interleave), 256);
1224 break;
1225
1226 case AR_F_PROMISE_RAID:
1227 rdp->interleave = min(max(2, rdp->interleave), 2048); /*+*/
1228 break;
1229
1230 case AR_F_SII_RAID:
1231 rdp->interleave = min(max(8, rdp->interleave), 256); /*+*/
1232 break;
1233
1234 case AR_F_SIS_RAID:
1235 rdp->interleave = min(max(32, rdp->interleave), 512); /*+*/
1236 break;
1237
1238 case AR_F_VIA_RAID:
1239 rdp->interleave = min(max(8, rdp->interleave), 128); /*+*/
1240 break;
1241 }
1242
1243 rdp->total_disks = total_disks;
1244 rdp->width = total_disks / (rdp->type & (AR_RAID1 | AR_T_RAID01) ? 2 : 1);
1245 rdp->total_sectors = disk_size * (rdp->width - (rdp->type == AR_RAID5));
1246 rdp->heads = 255;
1247 rdp->sectors = 63;
1248 rdp->cylinders = rdp->total_sectors / (255 * 63);
1249 rdp->rebuild_lba = 0;
1250 rdp->status |= AR_S_READY;
1251
1252 /* we are committed to this array, grap the subdisks */
1253 for (disk = 0; disk < config->total_disks; disk++) {
1254 if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
1255 config->disks[disk]))) {
1256 struct ata_raid_subdisk *ars = device_get_softc(subdisk);
1257
1258 ars->raid[rdp->volume] = rdp;
1259 ars->disk_number[rdp->volume] = disk;
1260 }
1261 }
1262 ata_raid_attach(rdp, 1);
1263 ata_raid_arrays[array] = rdp;
1264 config->lun = array;
1265 return 0;
1266}
1267
1268static int
1269ata_raid_delete(int array)
1270{
1271 struct ar_softc *rdp;
1272 device_t subdisk;
1273 int disk;
1274
1275 if (!(rdp = ata_raid_arrays[array]))
1276 return ENXIO;
1277
1278 rdp->status &= ~AR_S_READY;
d438c7c2 1279 disk_destroy(&rdp->disk);
c1b3d7c5
TS		 1280
1281 for (disk = 0; disk < rdp->total_disks; disk++) {
1282 if ((rdp->disks[disk].flags & AR_DF_PRESENT) && rdp->disks[disk].dev) {
1283 if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
1284 device_get_unit(rdp->disks[disk].dev)))) {
1285 struct ata_raid_subdisk *ars = device_get_softc(subdisk);
1286
1287 if (ars->raid[rdp->volume] != rdp) /* XXX SOS */
1288 device_printf(subdisk, "DOH! this disk doesn't belong\n");
1289 if (ars->disk_number[rdp->volume] != disk) /* XXX SOS */
1290 device_printf(subdisk, "DOH! this disk number is wrong\n");
1291 ars->raid[rdp->volume] = NULL;
1292 ars->disk_number[rdp->volume] = -1;
1293 }
1294 rdp->disks[disk].flags = 0;
1295 }
1296 }
1297 ata_raid_wipe_metadata(rdp);
1298 ata_raid_arrays[array] = NULL;
d438c7c2 1299 kfree(rdp, M_AR);
c1b3d7c5
TS		 1300 return 0;
1301}
1302
1303static int
1304ata_raid_addspare(struct ata_ioc_raid_config *config)
1305{
1306 struct ar_softc *rdp;
1307 device_t subdisk;
1308 int disk;
1309
1310 if (!(rdp = ata_raid_arrays[config->lun]))
1311 return ENXIO;
1312 if (!(rdp->status & AR_S_DEGRADED) || !(rdp->status & AR_S_READY))
1313 return ENXIO;
1314 if (rdp->status & AR_S_REBUILDING)
1315 return EBUSY;
1316 switch (rdp->type) {
1317 case AR_T_RAID1:
1318 case AR_T_RAID01:
1319 case AR_T_RAID5:
1320 for (disk = 0; disk < rdp->total_disks; disk++ ) {
1321
1322 if (((rdp->disks[disk].flags & (AR_DF_PRESENT | AR_DF_ONLINE)) ==
1323 (AR_DF_PRESENT | AR_DF_ONLINE)) && rdp->disks[disk].dev)
1324 continue;
1325
1326 if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
1327 config->disks[0] ))) {
1328 struct ata_raid_subdisk *ars = device_get_softc(subdisk);
1329
1330 if (ars->raid[rdp->volume])
1331 return EBUSY;
1332
1333 /* XXX SOS validate size etc etc */
1334 ars->raid[rdp->volume] = rdp;
1335 ars->disk_number[rdp->volume] = disk;
1336 rdp->disks[disk].dev = device_get_parent(subdisk);
1337 rdp->disks[disk].flags =
1338 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE);
1339
1340 device_printf(rdp->disks[disk].dev,
1341 "inserted into ar%d disk%d as spare\n",
1342 rdp->lun, disk);
1343 ata_raid_config_changed(rdp, 1);
1344 return 0;
1345 }
1346 }
1347 return ENXIO;
1348
1349 default:
1350 return EPERM;
1351 }
1352}
1353
1354static int
1355ata_raid_rebuild(int array)
1356{
1357 struct ar_softc *rdp;
1358 int disk, count;
1359
1360 if (!(rdp = ata_raid_arrays[array]))
1361 return ENXIO;
1362 /* XXX SOS we should lock the rdp softc here */
1363 if (!(rdp->status & AR_S_DEGRADED) || !(rdp->status & AR_S_READY))
1364 return ENXIO;
1365 if (rdp->status & AR_S_REBUILDING)
1366 return EBUSY;
1367
1368 switch (rdp->type) {
1369 case AR_T_RAID1:
1370 case AR_T_RAID01:
1371 case AR_T_RAID5:
1372 for (count = 0, disk = 0; disk < rdp->total_disks; disk++ ) {
1373 if (((rdp->disks[disk].flags &
1374 (AR_DF_PRESENT|AR_DF_ASSIGNED|AR_DF_ONLINE|AR_DF_SPARE)) ==
1375 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE)) &&
1376 rdp->disks[disk].dev) {
1377 count++;
1378 }
1379 }
1380
1381 if (count) {
1382 rdp->rebuild_lba = 0;
1383 rdp->status |= AR_S_REBUILDING;
1384 return 0;
1385 }
1386 return EIO;
1387
1388 default:
1389 return EPERM;
1390 }
1391}
1392
1393static int
1394ata_raid_read_metadata(device_t subdisk)
1395{
1396 devclass_t pci_devclass = devclass_find("pci");
1397 devclass_t devclass=device_get_devclass(GRANDPARENT(GRANDPARENT(subdisk)));
da44240f
MD		 1398 device_t gpdev;
1399 uint16_t vendor;
c1b3d7c5
TS		 1400
1401 /* prioritize vendor native metadata layout if possible */
1402 if (devclass == pci_devclass) {
da44240f
MD		 1403 gpdev = device_get_parent(subdisk);
1404 gpdev = GRANDPARENT(gpdev);
1405 vendor = pci_get_vendor(gpdev);
1406
1407 switch (vendor) {
c1b3d7c5
TS		 1408 case ATA_HIGHPOINT_ID:
1409 if (ata_raid_hptv3_read_meta(subdisk, ata_raid_arrays))
1410 return 0;
1411 if (ata_raid_hptv2_read_meta(subdisk, ata_raid_arrays))
1412 return 0;
1413 break;
1414
1415 case ATA_INTEL_ID:
1416 if (ata_raid_intel_read_meta(subdisk, ata_raid_arrays))
1417 return 0;
1418 break;
1419
1420 case ATA_ITE_ID:
1421 if (ata_raid_ite_read_meta(subdisk, ata_raid_arrays))
1422 return 0;
1423 break;
1424
1425 case ATA_JMICRON_ID:
1426 if (ata_raid_jmicron_read_meta(subdisk, ata_raid_arrays))
1427 return 0;
1428 break;
1429
1430 case ATA_NVIDIA_ID:
1431 if (ata_raid_nvidia_read_meta(subdisk, ata_raid_arrays))
1432 return 0;
1433 break;
1434
1435 case 0: /* XXX SOS cover up for bug in our PCI code */
1436 case ATA_PROMISE_ID:
1437 if (ata_raid_promise_read_meta(subdisk, ata_raid_arrays, 0))
1438 return 0;
1439 break;
1440
1441 case ATA_ATI_ID:
1442 case ATA_SILICON_IMAGE_ID:
1443 if (ata_raid_sii_read_meta(subdisk, ata_raid_arrays))
1444 return 0;
1445 break;
1446
1447 case ATA_SIS_ID:
1448 if (ata_raid_sis_read_meta(subdisk, ata_raid_arrays))
1449 return 0;
1450 break;
1451
1452 case ATA_VIA_ID:
1453 if (ata_raid_via_read_meta(subdisk, ata_raid_arrays))
1454 return 0;
1455 break;
1456 }
1457 }
1458
1459 /* handle controllers that have multiple layout possibilities */
1460 /* NOTE: the order of these are not insignificant */
1461
1462 /* Adaptec HostRAID */
1463 if (ata_raid_adaptec_read_meta(subdisk, ata_raid_arrays))
1464 return 0;
1465
1466 /* LSILogic v3 and v2 */
1467 if (ata_raid_lsiv3_read_meta(subdisk, ata_raid_arrays))
1468 return 0;
1469 if (ata_raid_lsiv2_read_meta(subdisk, ata_raid_arrays))
1470 return 0;
1471
1472 /* if none of the above matched, try FreeBSD native format */
1473 return ata_raid_promise_read_meta(subdisk, ata_raid_arrays, 1);
1474}
1475
1476static int
1477ata_raid_write_metadata(struct ar_softc *rdp)
1478{
1479 switch (rdp->format) {
1480 case AR_F_FREEBSD_RAID:
1481 case AR_F_PROMISE_RAID:
1482 return ata_raid_promise_write_meta(rdp);
1483
1484 case AR_F_HPTV3_RAID:
1485 case AR_F_HPTV2_RAID:
1486 /*
1487 * always write HPT v2 metadata, the v3 BIOS knows it as well.
1488 * this is handy since we cannot know what version BIOS is on there
1489 */
1490 return ata_raid_hptv2_write_meta(rdp);
1491
1492 case AR_F_INTEL_RAID:
1493 return ata_raid_intel_write_meta(rdp);
1494
1495 case AR_F_JMICRON_RAID:
1496 return ata_raid_jmicron_write_meta(rdp);
1497
1498 case AR_F_SIS_RAID:
1499 return ata_raid_sis_write_meta(rdp);
1500
1501 case AR_F_VIA_RAID:
1502 return ata_raid_via_write_meta(rdp);
1503#if 0
1504 case AR_F_HPTV3_RAID:
1505 return ata_raid_hptv3_write_meta(rdp);
1506
1507 case AR_F_ADAPTEC_RAID:
1508 return ata_raid_adaptec_write_meta(rdp);
1509
1510 case AR_F_ITE_RAID:
1511 return ata_raid_ite_write_meta(rdp);
1512
1513 case AR_F_LSIV2_RAID:
1514 return ata_raid_lsiv2_write_meta(rdp);
1515
1516 case AR_F_LSIV3_RAID:
1517 return ata_raid_lsiv3_write_meta(rdp);
1518
1519 case AR_F_NVIDIA_RAID:
1520 return ata_raid_nvidia_write_meta(rdp);
1521
1522 case AR_F_SII_RAID:
1523 return ata_raid_sii_write_meta(rdp);
1524
1525#endif
1526 default:
e3869ec7 1527 kprintf("ar%d: writing of %s metadata is NOT supported yet\n",
c1b3d7c5
TS		 1528 rdp->lun, ata_raid_format(rdp));
1529 }
1530 return -1;
1531}
1532
1533static int
1534ata_raid_wipe_metadata(struct ar_softc *rdp)
1535{
1536 int disk, error = 0;
1537 u_int64_t lba;
1538 u_int32_t size;
1539 u_int8_t *meta;
1540
1541 for (disk = 0; disk < rdp->total_disks; disk++) {
1542 if (rdp->disks[disk].dev) {
1543 switch (rdp->format) {
1544 case AR_F_ADAPTEC_RAID:
1545 lba = ADP_LBA(rdp->disks[disk].dev);
1546 size = sizeof(struct adaptec_raid_conf);
1547 break;
1548
1549 case AR_F_HPTV2_RAID:
1550 lba = HPTV2_LBA(rdp->disks[disk].dev);
1551 size = sizeof(struct hptv2_raid_conf);
1552 break;
1553
1554 case AR_F_HPTV3_RAID:
1555 lba = HPTV3_LBA(rdp->disks[disk].dev);
1556 size = sizeof(struct hptv3_raid_conf);
1557 break;
1558
1559 case AR_F_INTEL_RAID:
1560 lba = INTEL_LBA(rdp->disks[disk].dev);
1561 size = 3 * 512; /* XXX SOS */
1562 break;
1563
1564 case AR_F_ITE_RAID:
1565 lba = ITE_LBA(rdp->disks[disk].dev);
1566 size = sizeof(struct ite_raid_conf);
1567 break;
1568
1569 case AR_F_JMICRON_RAID:
1570 lba = JMICRON_LBA(rdp->disks[disk].dev);
1571 size = sizeof(struct jmicron_raid_conf);
1572 break;
1573
1574 case AR_F_LSIV2_RAID:
1575 lba = LSIV2_LBA(rdp->disks[disk].dev);
1576 size = sizeof(struct lsiv2_raid_conf);
1577 break;
1578
1579 case AR_F_LSIV3_RAID:
1580 lba = LSIV3_LBA(rdp->disks[disk].dev);
1581 size = sizeof(struct lsiv3_raid_conf);
1582 break;
1583
1584 case AR_F_NVIDIA_RAID:
1585 lba = NVIDIA_LBA(rdp->disks[disk].dev);
1586 size = sizeof(struct nvidia_raid_conf);
1587 break;
1588
1589 case AR_F_FREEBSD_RAID:
1590 case AR_F_PROMISE_RAID:
1591 lba = PROMISE_LBA(rdp->disks[disk].dev);
1592 size = sizeof(struct promise_raid_conf);
1593 break;
1594
1595 case AR_F_SII_RAID:
1596 lba = SII_LBA(rdp->disks[disk].dev);
1597 size = sizeof(struct sii_raid_conf);
1598 break;
1599
1600 case AR_F_SIS_RAID:
1601 lba = SIS_LBA(rdp->disks[disk].dev);
1602 size = sizeof(struct sis_raid_conf);
1603 break;
1604
1605 case AR_F_VIA_RAID:
1606 lba = VIA_LBA(rdp->disks[disk].dev);
1607 size = sizeof(struct via_raid_conf);
1608 break;
1609
1610 default:
e3869ec7 1611 kprintf("ar%d: wiping of %s metadata is NOT supported yet\n",
c1b3d7c5
TS		 1612 rdp->lun, ata_raid_format(rdp));
1613 return ENXIO;
1614 }
978400d3 1615 meta = kmalloc(size, M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1616 if (ata_raid_rw(rdp->disks[disk].dev, lba, meta, size,
1617 ATA_R_WRITE | ATA_R_DIRECT)) {
1618 device_printf(rdp->disks[disk].dev, "wipe metadata failed\n");
1619 error = EIO;
1620 }
d438c7c2 1621 kfree(meta, M_AR);
c1b3d7c5
TS		 1622 }
1623 }
1624 return error;
1625}
1626
1627/* Adaptec HostRAID Metadata */
1628static int
1629ata_raid_adaptec_read_meta(device_t dev, struct ar_softc **raidp)
1630{
1631 struct ata_raid_subdisk *ars = device_get_softc(dev);
1632 device_t parent = device_get_parent(dev);
1633 struct adaptec_raid_conf *meta;
1634 struct ar_softc *raid;
1635 int array, disk, retval = 0;
1636
978400d3
SW		 1637 meta = (struct adaptec_raid_conf *)
1638 kmalloc(sizeof(struct adaptec_raid_conf), M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1639
1640 if (ata_raid_rw(parent, ADP_LBA(parent),
1641 meta, sizeof(struct adaptec_raid_conf), ATA_R_READ)) {
1642 if (testing || bootverbose)
1643 device_printf(parent, "Adaptec read metadata failed\n");
1644 goto adaptec_out;
1645 }
1646
1647 /* check if this is a Adaptec RAID struct */
1648 if (meta->magic_0 != ADP_MAGIC_0 || meta->magic_3 != ADP_MAGIC_3) {
1649 if (testing || bootverbose)
1650 device_printf(parent, "Adaptec check1 failed\n");
1651 goto adaptec_out;
1652 }
1653
1654 if (testing || bootverbose)
1655 ata_raid_adaptec_print_meta(meta);
1656
1657 /* now convert Adaptec metadata into our generic form */
1658 for (array = 0; array < MAX_ARRAYS; array++) {
1659 if (!raidp[array]) {
1660 raidp[array] =
d438c7c2
TS		 1661 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
1662 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1663 }
1664 raid = raidp[array];
1665 if (raid->format && (raid->format != AR_F_ADAPTEC_RAID))
1666 continue;
1667
1668 if (raid->magic_0 && raid->magic_0 != meta->configs[0].magic_0)
1669 continue;
1670
1671 if (!meta->generation || be32toh(meta->generation) > raid->generation) {
1672 switch (meta->configs[0].type) {
1673 case ADP_T_RAID0:
1674 raid->magic_0 = meta->configs[0].magic_0;
1675 raid->type = AR_T_RAID0;
1676 raid->interleave = 1 << (meta->configs[0].stripe_shift >> 1);
1677 raid->width = be16toh(meta->configs[0].total_disks);
1678 break;
1679
1680 case ADP_T_RAID1:
1681 raid->magic_0 = meta->configs[0].magic_0;
1682 raid->type = AR_T_RAID1;
1683 raid->width = be16toh(meta->configs[0].total_disks) / 2;
1684 break;
1685
1686 default:
1687 device_printf(parent, "Adaptec unknown RAID type 0x%02x\n",
1688 meta->configs[0].type);
d438c7c2 1689 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 1690 raidp[array] = NULL;
1691 goto adaptec_out;
1692 }
1693
1694 raid->format = AR_F_ADAPTEC_RAID;
1695 raid->generation = be32toh(meta->generation);
1696 raid->total_disks = be16toh(meta->configs[0].total_disks);
1697 raid->total_sectors = be32toh(meta->configs[0].sectors);
1698 raid->heads = 255;
1699 raid->sectors = 63;
1700 raid->cylinders = raid->total_sectors / (63 * 255);
1701 raid->offset_sectors = 0;
1702 raid->rebuild_lba = 0;
1703 raid->lun = array;
1704 strncpy(raid->name, meta->configs[0].name,
1705 min(sizeof(raid->name), sizeof(meta->configs[0].name)));
1706
1707 /* clear out any old info */
1708 if (raid->generation) {
1709 for (disk = 0; disk < raid->total_disks; disk++) {
1710 raid->disks[disk].dev = NULL;
1711 raid->disks[disk].flags = 0;
1712 }
1713 }
1714 }
1715 if (be32toh(meta->generation) >= raid->generation) {
1716 struct ata_device *atadev = device_get_softc(parent);
1717 struct ata_channel *ch = device_get_softc(GRANDPARENT(dev));
1718 int disk_number = (ch->unit << !(ch->flags & ATA_NO_SLAVE)) +
1719 ATA_DEV(atadev->unit);
1720
1721 raid->disks[disk_number].dev = parent;
1722 raid->disks[disk_number].sectors =
1723 be32toh(meta->configs[disk_number + 1].sectors);
1724 raid->disks[disk_number].flags =
1725 (AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
1726 ars->raid[raid->volume] = raid;
1727 ars->disk_number[raid->volume] = disk_number;
1728 retval = 1;
1729 }
1730 break;
1731 }
1732
1733adaptec_out:
d438c7c2 1734 kfree(meta, M_AR);
c1b3d7c5
TS		 1735 return retval;
1736}
1737
1738/* Highpoint V2 RocketRAID Metadata */
1739static int
1740ata_raid_hptv2_read_meta(device_t dev, struct ar_softc **raidp)
1741{
1742 struct ata_raid_subdisk *ars = device_get_softc(dev);
1743 device_t parent = device_get_parent(dev);
1744 struct hptv2_raid_conf *meta;
1745 struct ar_softc *raid = NULL;
1746 int array, disk_number = 0, retval = 0;
1747
978400d3
SW		 1748 meta = (struct hptv2_raid_conf *)kmalloc(sizeof(struct hptv2_raid_conf),
1749 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1750
1751 if (ata_raid_rw(parent, HPTV2_LBA(parent),
1752 meta, sizeof(struct hptv2_raid_conf), ATA_R_READ)) {
1753 if (testing || bootverbose)
1754 device_printf(parent, "HighPoint (v2) read metadata failed\n");
1755 goto hptv2_out;
1756 }
1757
1758 /* check if this is a HighPoint v2 RAID struct */
1759 if (meta->magic != HPTV2_MAGIC_OK && meta->magic != HPTV2_MAGIC_BAD) {
1760 if (testing || bootverbose)
1761 device_printf(parent, "HighPoint (v2) check1 failed\n");
1762 goto hptv2_out;
1763 }
1764
1765 /* is this disk defined, or an old leftover/spare ? */
1766 if (!meta->magic_0) {
1767 if (testing || bootverbose)
1768 device_printf(parent, "HighPoint (v2) check2 failed\n");
1769 goto hptv2_out;
1770 }
1771
1772 if (testing || bootverbose)
1773 ata_raid_hptv2_print_meta(meta);
1774
1775 /* now convert HighPoint (v2) metadata into our generic form */
1776 for (array = 0; array < MAX_ARRAYS; array++) {
1777 if (!raidp[array]) {
1778 raidp[array] =
d438c7c2
TS		 1779 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
1780 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1781 }
1782 raid = raidp[array];
1783 if (raid->format && (raid->format != AR_F_HPTV2_RAID))
1784 continue;
1785
1786 switch (meta->type) {
1787 case HPTV2_T_RAID0:
1788 if ((meta->order & (HPTV2_O_RAID0|HPTV2_O_OK)) ==
1789 (HPTV2_O_RAID0|HPTV2_O_OK))
1790 goto highpoint_raid1;
1791 if (meta->order & (HPTV2_O_RAID0 | HPTV2_O_RAID1))
1792 goto highpoint_raid01;
1793 if (raid->magic_0 && raid->magic_0 != meta->magic_0)
1794 continue;
1795 raid->magic_0 = meta->magic_0;
1796 raid->type = AR_T_RAID0;
1797 raid->interleave = 1 << meta->stripe_shift;
1798 disk_number = meta->disk_number;
1799 if (!(meta->order & HPTV2_O_OK))
1800 meta->magic = 0; /* mark bad */
1801 break;
1802
1803 case HPTV2_T_RAID1:
1804highpoint_raid1:
1805 if (raid->magic_0 && raid->magic_0 != meta->magic_0)
1806 continue;
1807 raid->magic_0 = meta->magic_0;
1808 raid->type = AR_T_RAID1;
1809 disk_number = (meta->disk_number > 0);
1810 break;
1811
1812 case HPTV2_T_RAID01_RAID0:
1813highpoint_raid01:
1814 if (meta->order & HPTV2_O_RAID0) {
1815 if ((raid->magic_0 && raid->magic_0 != meta->magic_0) ||
1816 (raid->magic_1 && raid->magic_1 != meta->magic_1))
1817 continue;
1818 raid->magic_0 = meta->magic_0;
1819 raid->magic_1 = meta->magic_1;
1820 raid->type = AR_T_RAID01;
1821 raid->interleave = 1 << meta->stripe_shift;
1822 disk_number = meta->disk_number;
1823 }
1824 else {
1825 if (raid->magic_1 && raid->magic_1 != meta->magic_1)
1826 continue;
1827 raid->magic_1 = meta->magic_1;
1828 raid->type = AR_T_RAID01;
1829 raid->interleave = 1 << meta->stripe_shift;
1830 disk_number = meta->disk_number + meta->array_width;
1831 if (!(meta->order & HPTV2_O_RAID1))
1832 meta->magic = 0; /* mark bad */
1833 }
1834 break;
1835
1836 case HPTV2_T_SPAN:
1837 if (raid->magic_0 && raid->magic_0 != meta->magic_0)
1838 continue;
1839 raid->magic_0 = meta->magic_0;
1840 raid->type = AR_T_SPAN;
1841 disk_number = meta->disk_number;
1842 break;
1843
1844 default:
1845 device_printf(parent, "Highpoint (v2) unknown RAID type 0x%02x\n",
1846 meta->type);
d438c7c2 1847 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 1848 raidp[array] = NULL;
1849 goto hptv2_out;
1850 }
1851
1852 raid->format |= AR_F_HPTV2_RAID;
1853 raid->disks[disk_number].dev = parent;
1854 raid->disks[disk_number].flags = (AR_DF_PRESENT | AR_DF_ASSIGNED);
1855 raid->lun = array;
1856 strncpy(raid->name, meta->name_1,
1857 min(sizeof(raid->name), sizeof(meta->name_1)));
1858 if (meta->magic == HPTV2_MAGIC_OK) {
1859 raid->disks[disk_number].flags |= AR_DF_ONLINE;
1860 raid->width = meta->array_width;
1861 raid->total_sectors = meta->total_sectors;
1862 raid->heads = 255;
1863 raid->sectors = 63;
1864 raid->cylinders = raid->total_sectors / (63 * 255);
1865 raid->offset_sectors = HPTV2_LBA(parent) + 1;
1866 raid->rebuild_lba = meta->rebuild_lba;
1867 raid->disks[disk_number].sectors =
1868 raid->total_sectors / raid->width;
1869 }
1870 else
1871 raid->disks[disk_number].flags &= ~AR_DF_ONLINE;
1872
1873 if ((raid->type & AR_T_RAID0) && (raid->total_disks < raid->width))
1874 raid->total_disks = raid->width;
1875 if (disk_number >= raid->total_disks)
1876 raid->total_disks = disk_number + 1;
1877 ars->raid[raid->volume] = raid;
1878 ars->disk_number[raid->volume] = disk_number;
1879 retval = 1;
1880 break;
1881 }
1882
1883hptv2_out:
d438c7c2 1884 kfree(meta, M_AR);
c1b3d7c5
TS		 1885 return retval;
1886}
1887
1888static int
1889ata_raid_hptv2_write_meta(struct ar_softc *rdp)
1890{
1891 struct hptv2_raid_conf *meta;
1892 struct timeval timestamp;
1893 int disk, error = 0;
1894
978400d3
SW		 1895 meta = (struct hptv2_raid_conf *)kmalloc(sizeof(struct hptv2_raid_conf),
1896 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1897
1898 microtime(&timestamp);
1899 rdp->magic_0 = timestamp.tv_sec + 2;
1900 rdp->magic_1 = timestamp.tv_sec;
1901
1902 for (disk = 0; disk < rdp->total_disks; disk++) {
1903 if ((rdp->disks[disk].flags & (AR_DF_PRESENT | AR_DF_ONLINE)) ==
1904 (AR_DF_PRESENT | AR_DF_ONLINE))
1905 meta->magic = HPTV2_MAGIC_OK;
1906 if (rdp->disks[disk].flags & AR_DF_ASSIGNED) {
1907 meta->magic_0 = rdp->magic_0;
1908 if (strlen(rdp->name))
1909 strncpy(meta->name_1, rdp->name, sizeof(meta->name_1));
1910 else
1911 strcpy(meta->name_1, "FreeBSD");
1912 }
1913 meta->disk_number = disk;
1914
1915 switch (rdp->type) {
1916 case AR_T_RAID0:
1917 meta->type = HPTV2_T_RAID0;
1918 strcpy(meta->name_2, "RAID 0");
1919 if (rdp->disks[disk].flags & AR_DF_ONLINE)
1920 meta->order = HPTV2_O_OK;
1921 break;
1922
1923 case AR_T_RAID1:
1924 meta->type = HPTV2_T_RAID0;
1925 strcpy(meta->name_2, "RAID 1");
1926 meta->disk_number = (disk < rdp->width) ? disk : disk + 5;
1927 meta->order = HPTV2_O_RAID0 | HPTV2_O_OK;
1928 break;
1929
1930 case AR_T_RAID01:
1931 meta->type = HPTV2_T_RAID01_RAID0;
1932 strcpy(meta->name_2, "RAID 0+1");
1933 if (rdp->disks[disk].flags & AR_DF_ONLINE) {
1934 if (disk < rdp->width) {
1935 meta->order = (HPTV2_O_RAID0 | HPTV2_O_RAID1);
1936 meta->magic_0 = rdp->magic_0 - 1;
1937 }
1938 else {
1939 meta->order = HPTV2_O_RAID1;
1940 meta->disk_number -= rdp->width;
1941 }
1942 }
1943 else
1944 meta->magic_0 = rdp->magic_0 - 1;
1945 meta->magic_1 = rdp->magic_1;
1946 break;
1947
1948 case AR_T_SPAN:
1949 meta->type = HPTV2_T_SPAN;
1950 strcpy(meta->name_2, "SPAN");
1951 break;
1952 default:
d438c7c2 1953 kfree(meta, M_AR);
c1b3d7c5
TS		 1954 return ENODEV;
1955 }
1956
1957 meta->array_width = rdp->width;
1958 meta->stripe_shift = (rdp->width > 1) ? (ffs(rdp->interleave)-1) : 0;
1959 meta->total_sectors = rdp->total_sectors;
1960 meta->rebuild_lba = rdp->rebuild_lba;
1961 if (testing || bootverbose)
1962 ata_raid_hptv2_print_meta(meta);
1963 if (rdp->disks[disk].dev) {
1964 if (ata_raid_rw(rdp->disks[disk].dev,
1965 HPTV2_LBA(rdp->disks[disk].dev), meta,
1966 sizeof(struct promise_raid_conf),
1967 ATA_R_WRITE | ATA_R_DIRECT)) {
1968 device_printf(rdp->disks[disk].dev, "write metadata failed\n");
1969 error = EIO;
1970 }
1971 }
1972 }
d438c7c2 1973 kfree(meta, M_AR);
c1b3d7c5
TS		 1974 return error;
1975}
1976
1977/* Highpoint V3 RocketRAID Metadata */
1978static int
1979ata_raid_hptv3_read_meta(device_t dev, struct ar_softc **raidp)
1980{
1981 struct ata_raid_subdisk *ars = device_get_softc(dev);
1982 device_t parent = device_get_parent(dev);
1983 struct hptv3_raid_conf *meta;
1984 struct ar_softc *raid = NULL;
1985 int array, disk_number, retval = 0;
1986
978400d3
SW		 1987 meta = (struct hptv3_raid_conf *)kmalloc(sizeof(struct hptv3_raid_conf),
1988 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 1989
1990 if (ata_raid_rw(parent, HPTV3_LBA(parent),
1991 meta, sizeof(struct hptv3_raid_conf), ATA_R_READ)) {
1992 if (testing || bootverbose)
1993 device_printf(parent, "HighPoint (v3) read metadata failed\n");
1994 goto hptv3_out;
1995 }
1996
1997 /* check if this is a HighPoint v3 RAID struct */
1998 if (meta->magic != HPTV3_MAGIC) {
1999 if (testing || bootverbose)
2000 device_printf(parent, "HighPoint (v3) check1 failed\n");
2001 goto hptv3_out;
2002 }
2003
2004 /* check if there are any config_entries */
2005 if (meta->config_entries < 1) {
2006 if (testing || bootverbose)
2007 device_printf(parent, "HighPoint (v3) check2 failed\n");
2008 goto hptv3_out;
2009 }
2010
2011 if (testing || bootverbose)
2012 ata_raid_hptv3_print_meta(meta);
2013
2014 /* now convert HighPoint (v3) metadata into our generic form */
2015 for (array = 0; array < MAX_ARRAYS; array++) {
2016 if (!raidp[array]) {
2017 raidp[array] =
d438c7c2
TS		 2018 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
2019 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2020 }
2021 raid = raidp[array];
2022 if (raid->format && (raid->format != AR_F_HPTV3_RAID))
2023 continue;
2024
2025 if ((raid->format & AR_F_HPTV3_RAID) && raid->magic_0 != meta->magic_0)
2026 continue;
2027
2028 switch (meta->configs[0].type) {
2029 case HPTV3_T_RAID0:
2030 raid->type = AR_T_RAID0;
2031 raid->width = meta->configs[0].total_disks;
2032 disk_number = meta->configs[0].disk_number;
2033 break;
2034
2035 case HPTV3_T_RAID1:
2036 raid->type = AR_T_RAID1;
2037 raid->width = meta->configs[0].total_disks / 2;
2038 disk_number = meta->configs[0].disk_number;
2039 break;
2040
2041 case HPTV3_T_RAID5:
2042 raid->type = AR_T_RAID5;
2043 raid->width = meta->configs[0].total_disks;
2044 disk_number = meta->configs[0].disk_number;
2045 break;
2046
2047 case HPTV3_T_SPAN:
2048 raid->type = AR_T_SPAN;
2049 raid->width = meta->configs[0].total_disks;
2050 disk_number = meta->configs[0].disk_number;
2051 break;
2052
2053 default:
2054 device_printf(parent, "Highpoint (v3) unknown RAID type 0x%02x\n",
2055 meta->configs[0].type);
d438c7c2 2056 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2057 raidp[array] = NULL;
2058 goto hptv3_out;
2059 }
2060 if (meta->config_entries == 2) {
2061 switch (meta->configs[1].type) {
2062 case HPTV3_T_RAID1:
2063 if (raid->type == AR_T_RAID0) {
2064 raid->type = AR_T_RAID01;
2065 disk_number = meta->configs[1].disk_number +
2066 (meta->configs[0].disk_number << 1);
2067 break;
2068 }
2069 default:
2070 device_printf(parent, "Highpoint (v3) unknown level 2 0x%02x\n",
2071 meta->configs[1].type);
d438c7c2 2072 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2073 raidp[array] = NULL;
2074 goto hptv3_out;
2075 }
2076 }
2077
2078 raid->magic_0 = meta->magic_0;
2079 raid->format = AR_F_HPTV3_RAID;
2080 raid->generation = meta->timestamp;
2081 raid->interleave = 1 << meta->configs[0].stripe_shift;
2082 raid->total_disks = meta->configs[0].total_disks +
2083 meta->configs[1].total_disks;
2084 raid->total_sectors = meta->configs[0].total_sectors +
2085 ((u_int64_t)meta->configs_high[0].total_sectors << 32);
2086 raid->heads = 255;
2087 raid->sectors = 63;
2088 raid->cylinders = raid->total_sectors / (63 * 255);
2089 raid->offset_sectors = 0;
2090 raid->rebuild_lba = meta->configs[0].rebuild_lba +
2091 ((u_int64_t)meta->configs_high[0].rebuild_lba << 32);
2092 raid->lun = array;
2093 strncpy(raid->name, meta->name,
2094 min(sizeof(raid->name), sizeof(meta->name)));
2095 raid->disks[disk_number].sectors = raid->total_sectors /
2096 (raid->type == AR_T_RAID5 ? raid->width - 1 : raid->width);
2097 raid->disks[disk_number].dev = parent;
2098 raid->disks[disk_number].flags =
2099 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
2100 ars->raid[raid->volume] = raid;
2101 ars->disk_number[raid->volume] = disk_number;
2102 retval = 1;
2103 break;
2104 }
2105
2106hptv3_out:
d438c7c2 2107 kfree(meta, M_AR);
c1b3d7c5
TS		 2108 return retval;
2109}
2110
2111/* Intel MatrixRAID Metadata */
2112static int
2113ata_raid_intel_read_meta(device_t dev, struct ar_softc **raidp)
2114{
2115 struct ata_raid_subdisk *ars = device_get_softc(dev);
2116 device_t parent = device_get_parent(dev);
2117 struct intel_raid_conf *meta;
2118 struct intel_raid_mapping *map;
2119 struct ar_softc *raid = NULL;
2120 u_int32_t checksum, *ptr;
2121 int array, count, disk, volume = 1, retval = 0;
2122 char *tmp;
2123
978400d3 2124 meta = (struct intel_raid_conf *)kmalloc(1536, M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2125
2126 if (ata_raid_rw(parent, INTEL_LBA(parent), meta, 1024, ATA_R_READ)) {
2127 if (testing || bootverbose)
2128 device_printf(parent, "Intel read metadata failed\n");
2129 goto intel_out;
2130 }
2131 tmp = (char *)meta;
2132 bcopy(tmp, tmp+1024, 512);
2133 bcopy(tmp+512, tmp, 1024);
2134 bzero(tmp+1024, 512);
2135
2136 /* check if this is a Intel RAID struct */
2137 if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) {
2138 if (testing || bootverbose)
2139 device_printf(parent, "Intel check1 failed\n");
2140 goto intel_out;
2141 }
2142
2143 for (checksum = 0, ptr = (u_int32_t *)meta, count = 0;
2144 count < (meta->config_size / sizeof(u_int32_t)); count++) {
2145 checksum += *ptr++;
2146 }
2147 checksum -= meta->checksum;
2148 if (checksum != meta->checksum) {
2149 if (testing || bootverbose)
2150 device_printf(parent, "Intel check2 failed\n");
2151 goto intel_out;
2152 }
2153
2154 if (testing || bootverbose)
2155 ata_raid_intel_print_meta(meta);
2156
2157 map = (struct intel_raid_mapping *)&meta->disk[meta->total_disks];
2158
2159 /* now convert Intel metadata into our generic form */
2160 for (array = 0; array < MAX_ARRAYS; array++) {
2161 if (!raidp[array]) {
2162 raidp[array] =
d438c7c2
TS		 2163 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
2164 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2165 }
2166 raid = raidp[array];
2167 if (raid->format && (raid->format != AR_F_INTEL_RAID))
2168 continue;
2169
2170 if ((raid->format & AR_F_INTEL_RAID) &&
2171 (raid->magic_0 != meta->config_id))
2172 continue;
2173
2174 /*
2175 * update our knowledge about the array config based on generation
2176 * NOTE: there can be multiple volumes on a disk set
2177 */
2178 if (!meta->generation || meta->generation > raid->generation) {
2179 switch (map->type) {
2180 case INTEL_T_RAID0:
2181 raid->type = AR_T_RAID0;
2182 raid->width = map->total_disks;
2183 break;
2184
2185 case INTEL_T_RAID1:
2186 if (map->total_disks == 4)
2187 raid->type = AR_T_RAID01;
2188 else
2189 raid->type = AR_T_RAID1;
2190 raid->width = map->total_disks / 2;
2191 break;
2192
2193 case INTEL_T_RAID5:
2194 raid->type = AR_T_RAID5;
2195 raid->width = map->total_disks;
2196 break;
2197
2198 default:
2199 device_printf(parent, "Intel unknown RAID type 0x%02x\n",
2200 map->type);
d438c7c2 2201 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2202 raidp[array] = NULL;
2203 goto intel_out;
2204 }
2205
2206 switch (map->status) {
2207 case INTEL_S_READY:
2208 raid->status = AR_S_READY;
2209 break;
2210 case INTEL_S_DEGRADED:
2211 raid->status |= AR_S_DEGRADED;
2212 break;
2213 case INTEL_S_DISABLED:
2214 case INTEL_S_FAILURE:
2215 raid->status = 0;
2216 }
2217
2218 raid->magic_0 = meta->config_id;
2219 raid->format = AR_F_INTEL_RAID;
2220 raid->generation = meta->generation;
2221 raid->interleave = map->stripe_sectors;
2222 raid->total_disks = map->total_disks;
2223 raid->total_sectors = map->total_sectors;
2224 raid->heads = 255;
2225 raid->sectors = 63;
2226 raid->cylinders = raid->total_sectors / (63 * 255);
2227 raid->offset_sectors = map->offset;
2228 raid->rebuild_lba = 0;
2229 raid->lun = array;
2230 raid->volume = volume - 1;
2231 strncpy(raid->name, map->name,
2232 min(sizeof(raid->name), sizeof(map->name)));
2233
2234 /* clear out any old info */
2235 for (disk = 0; disk < raid->total_disks; disk++) {
2236 raid->disks[disk].dev = NULL;
2237 bcopy(meta->disk[map->disk_idx[disk]].serial,
2238 raid->disks[disk].serial,
2239 sizeof(raid->disks[disk].serial));
2240 raid->disks[disk].sectors =
2241 meta->disk[map->disk_idx[disk]].sectors;
2242 raid->disks[disk].flags = 0;
2243 if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_ONLINE)
2244 raid->disks[disk].flags |= AR_DF_ONLINE;
2245 if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_ASSIGNED)
2246 raid->disks[disk].flags |= AR_DF_ASSIGNED;
2247 if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_SPARE) {
2248 raid->disks[disk].flags &= ~(AR_DF_ONLINE | AR_DF_ASSIGNED);
2249 raid->disks[disk].flags |= AR_DF_SPARE;
2250 }
2251 if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_DOWN)
2252 raid->disks[disk].flags &= ~AR_DF_ONLINE;
2253 }
2254 }
2255 if (meta->generation >= raid->generation) {
2256 for (disk = 0; disk < raid->total_disks; disk++) {
2257 struct ata_device *atadev = device_get_softc(parent);
2258
2259 if (!strncmp(raid->disks[disk].serial, atadev->param.serial,
2260 sizeof(raid->disks[disk].serial))) {
2261 raid->disks[disk].dev = parent;
2262 raid->disks[disk].flags |= (AR_DF_PRESENT | AR_DF_ONLINE);
2263 ars->raid[raid->volume] = raid;
2264 ars->disk_number[raid->volume] = disk;
2265 retval = 1;
2266 }
2267 }
2268 }
2269 else
2270 goto intel_out;
2271
2272 if (retval) {
2273 if (volume < meta->total_volumes) {
2274 map = (struct intel_raid_mapping *)
2275 &map->disk_idx[map->total_disks];
2276 volume++;
2277 retval = 0;
2278 continue;
2279 }
2280 break;
2281 }
2282 else {
d438c7c2 2283 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2284 raidp[array] = NULL;
2285 if (volume == 2)
2286 retval = 1;
2287 }
2288 }
2289
2290intel_out:
d438c7c2 2291 kfree(meta, M_AR);
c1b3d7c5
TS		 2292 return retval;
2293}
2294
2295static int
2296ata_raid_intel_write_meta(struct ar_softc *rdp)
2297{
2298 struct intel_raid_conf *meta;
2299 struct intel_raid_mapping *map;
2300 struct timeval timestamp;
2301 u_int32_t checksum, *ptr;
2302 int count, disk, error = 0;
2303 char *tmp;
2304
978400d3 2305 meta = (struct intel_raid_conf *)kmalloc(1536, M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2306
2307 rdp->generation++;
6b7bbefa
MD		 2308
2309 /* Generate a new config_id if none exists */
2310 if (!rdp->magic_0) {
2311 microtime(&timestamp);
2312 rdp->magic_0 = timestamp.tv_sec ^ timestamp.tv_usec;
2313 }
c1b3d7c5
TS		 2314
2315 bcopy(INTEL_MAGIC, meta->intel_id, sizeof(meta->intel_id));
2316 bcopy(INTEL_VERSION_1100, meta->version, sizeof(meta->version));
6b7bbefa 2317 meta->config_id = rdp->magic_0;
c1b3d7c5
TS		 2318 meta->generation = rdp->generation;
2319 meta->total_disks = rdp->total_disks;
2320 meta->total_volumes = 1; /* XXX SOS */
2321 for (disk = 0; disk < rdp->total_disks; disk++) {
2322 if (rdp->disks[disk].dev) {
2323 struct ata_channel *ch =
2324 device_get_softc(device_get_parent(rdp->disks[disk].dev));
2325 struct ata_device *atadev =
2326 device_get_softc(rdp->disks[disk].dev);
2327
2328 bcopy(atadev->param.serial, meta->disk[disk].serial,
2329 sizeof(rdp->disks[disk].serial));
2330 meta->disk[disk].sectors = rdp->disks[disk].sectors;
2331 meta->disk[disk].id = (ch->unit << 16) | ATA_DEV(atadev->unit);
2332 }
2333 else
2334 meta->disk[disk].sectors = rdp->total_sectors / rdp->width;
2335 meta->disk[disk].flags = 0;
2336 if (rdp->disks[disk].flags & AR_DF_SPARE)
2337 meta->disk[disk].flags |= INTEL_F_SPARE;
2338 else {
2339 if (rdp->disks[disk].flags & AR_DF_ONLINE)
2340 meta->disk[disk].flags |= INTEL_F_ONLINE;
2341 else
2342 meta->disk[disk].flags |= INTEL_F_DOWN;
2343 if (rdp->disks[disk].flags & AR_DF_ASSIGNED)
2344 meta->disk[disk].flags |= INTEL_F_ASSIGNED;
2345 }
2346 }
2347 map = (struct intel_raid_mapping *)&meta->disk[meta->total_disks];
2348
2349 bcopy(rdp->name, map->name, sizeof(rdp->name));
2350 map->total_sectors = rdp->total_sectors;
2351 map->state = 12; /* XXX SOS */
2352 map->offset = rdp->offset_sectors;
2353 map->stripe_count = rdp->total_sectors / (rdp->interleave*rdp->total_disks);
2354 map->stripe_sectors = rdp->interleave;
2355 map->disk_sectors = rdp->total_sectors / rdp->width;
2356 map->status = INTEL_S_READY; /* XXX SOS */
2357 switch (rdp->type) {
2358 case AR_T_RAID0:
2359 map->type = INTEL_T_RAID0;
2360 break;
2361 case AR_T_RAID1:
2362 map->type = INTEL_T_RAID1;
2363 break;
2364 case AR_T_RAID01:
2365 map->type = INTEL_T_RAID1;
2366 break;
2367 case AR_T_RAID5:
2368 map->type = INTEL_T_RAID5;
2369 break;
2370 default:
d438c7c2 2371 kfree(meta, M_AR);
c1b3d7c5
TS		 2372 return ENODEV;
2373 }
2374 map->total_disks = rdp->total_disks;
2375 map->magic[0] = 0x02;
2376 map->magic[1] = 0xff;
2377 map->magic[2] = 0x01;
2378 for (disk = 0; disk < rdp->total_disks; disk++)
2379 map->disk_idx[disk] = disk;
2380
2381 meta->config_size = (char *)&map->disk_idx[disk] - (char *)meta;
2382 for (checksum = 0, ptr = (u_int32_t *)meta, count = 0;
2383 count < (meta->config_size / sizeof(u_int32_t)); count++) {
2384 checksum += *ptr++;
2385 }
2386 meta->checksum = checksum;
2387
2388 if (testing || bootverbose)
2389 ata_raid_intel_print_meta(meta);
2390
2391 tmp = (char *)meta;
2392 bcopy(tmp, tmp+1024, 512);
2393 bcopy(tmp+512, tmp, 1024);
2394 bzero(tmp+1024, 512);
2395
2396 for (disk = 0; disk < rdp->total_disks; disk++) {
2397 if (rdp->disks[disk].dev) {
2398 if (ata_raid_rw(rdp->disks[disk].dev,
2399 INTEL_LBA(rdp->disks[disk].dev),
2400 meta, 1024, ATA_R_WRITE | ATA_R_DIRECT)) {
2401 device_printf(rdp->disks[disk].dev, "write metadata failed\n");
2402 error = EIO;
2403 }
2404 }
2405 }
d438c7c2 2406 kfree(meta, M_AR);
c1b3d7c5
TS		 2407 return error;
2408}
2409
2410
2411/* Integrated Technology Express Metadata */
2412static int
2413ata_raid_ite_read_meta(device_t dev, struct ar_softc **raidp)
2414{
2415 struct ata_raid_subdisk *ars = device_get_softc(dev);
2416 device_t parent = device_get_parent(dev);
2417 struct ite_raid_conf *meta;
2418 struct ar_softc *raid = NULL;
2419 int array, disk_number, count, retval = 0;
2420 u_int16_t *ptr;
2421
978400d3
SW		 2422 meta = (struct ite_raid_conf *)kmalloc(sizeof(struct ite_raid_conf), M_AR,
2423 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2424
2425 if (ata_raid_rw(parent, ITE_LBA(parent),
2426 meta, sizeof(struct ite_raid_conf), ATA_R_READ)) {
2427 if (testing || bootverbose)
2428 device_printf(parent, "ITE read metadata failed\n");
2429 goto ite_out;
2430 }
2431
2432 /* check if this is a ITE RAID struct */
2433 for (ptr = (u_int16_t *)meta->ite_id, count = 0;
2434 count < sizeof(meta->ite_id)/sizeof(uint16_t); count++)
2435 ptr[count] = be16toh(ptr[count]);
2436
2437 if (strncmp(meta->ite_id, ITE_MAGIC, strlen(ITE_MAGIC))) {
2438 if (testing || bootverbose)
2439 device_printf(parent, "ITE check1 failed\n");
2440 goto ite_out;
2441 }
2442
2443 if (testing || bootverbose)
2444 ata_raid_ite_print_meta(meta);
2445
2446 /* now convert ITE metadata into our generic form */
2447 for (array = 0; array < MAX_ARRAYS; array++) {
2448 if ((raid = raidp[array])) {
2449 if (raid->format != AR_F_ITE_RAID)
2450 continue;
2451 if (raid->magic_0 != *((u_int64_t *)meta->timestamp_0))
2452 continue;
2453 }
2454
2455 /* if we dont have a disks timestamp the RAID is invalidated */
2456 if (*((u_int64_t *)meta->timestamp_1) == 0)
2457 goto ite_out;
2458
2459 if (!raid) {
d438c7c2
TS		 2460 raidp[array] = (struct ar_softc *)kmalloc(sizeof(struct ar_softc),
2461 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2462 }
2463
2464 switch (meta->type) {
2465 case ITE_T_RAID0:
2466 raid->type = AR_T_RAID0;
2467 raid->width = meta->array_width;
2468 raid->total_disks = meta->array_width;
2469 disk_number = meta->disk_number;
2470 break;
2471
2472 case ITE_T_RAID1:
2473 raid->type = AR_T_RAID1;
2474 raid->width = 1;
2475 raid->total_disks = 2;
2476 disk_number = meta->disk_number;
2477 break;
2478
2479 case ITE_T_RAID01:
2480 raid->type = AR_T_RAID01;
2481 raid->width = meta->array_width;
2482 raid->total_disks = 4;
2483 disk_number = ((meta->disk_number & 0x02) >> 1) |
2484 ((meta->disk_number & 0x01) << 1);
2485 break;
2486
2487 case ITE_T_SPAN:
2488 raid->type = AR_T_SPAN;
2489 raid->width = 1;
2490 raid->total_disks = meta->array_width;
2491 disk_number = meta->disk_number;
2492 break;
2493
2494 default:
2495 device_printf(parent, "ITE unknown RAID type 0x%02x\n", meta->type);
d438c7c2 2496 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2497 raidp[array] = NULL;
2498 goto ite_out;
2499 }
2500
2501 raid->magic_0 = *((u_int64_t *)meta->timestamp_0);
2502 raid->format = AR_F_ITE_RAID;
2503 raid->generation = 0;
2504 raid->interleave = meta->stripe_sectors;
2505 raid->total_sectors = meta->total_sectors;
2506 raid->heads = 255;
2507 raid->sectors = 63;
2508 raid->cylinders = raid->total_sectors / (63 * 255);
2509 raid->offset_sectors = 0;
2510 raid->rebuild_lba = 0;
2511 raid->lun = array;
2512
2513 raid->disks[disk_number].dev = parent;
2514 raid->disks[disk_number].sectors = raid->total_sectors / raid->width;
2515 raid->disks[disk_number].flags =
2516 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
2517 ars->raid[raid->volume] = raid;
2518 ars->disk_number[raid->volume] = disk_number;
2519 retval = 1;
2520 break;
2521 }
2522ite_out:
d438c7c2 2523 kfree(meta, M_AR);
c1b3d7c5
TS		 2524 return retval;
2525}
2526
2527/* JMicron Technology Corp Metadata */
2528static int
2529ata_raid_jmicron_read_meta(device_t dev, struct ar_softc **raidp)
2530{
2531 struct ata_raid_subdisk *ars = device_get_softc(dev);
2532 device_t parent = device_get_parent(dev);
2533 struct jmicron_raid_conf *meta;
2534 struct ar_softc *raid = NULL;
2535 u_int16_t checksum, *ptr;
2536 u_int64_t disk_size;
2537 int count, array, disk, total_disks, retval = 0;
2538
978400d3
SW		 2539 meta = (struct jmicron_raid_conf *)
2540 kmalloc(sizeof(struct jmicron_raid_conf), M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2541
2542 if (ata_raid_rw(parent, JMICRON_LBA(parent),
2543 meta, sizeof(struct jmicron_raid_conf), ATA_R_READ)) {
2544 if (testing || bootverbose)
2545 device_printf(parent,
2546 "JMicron read metadata failed\n");
2547 }
2548
2549 /* check for JMicron signature */
2550 if (strncmp(meta->signature, JMICRON_MAGIC, 2)) {
2551 if (testing || bootverbose)
2552 device_printf(parent, "JMicron check1 failed\n");
2553 goto jmicron_out;
2554 }
2555
2556 /* calculate checksum and compare for valid */
2557 for (checksum = 0, ptr = (u_int16_t *)meta, count = 0; count < 64; count++)
2558 checksum += *ptr++;
2559 if (checksum) {
2560 if (testing || bootverbose)
2561 device_printf(parent, "JMicron check2 failed\n");
2562 goto jmicron_out;
2563 }
2564
2565 if (testing || bootverbose)
2566 ata_raid_jmicron_print_meta(meta);
2567
2568 /* now convert JMicron meta into our generic form */
2569 for (array = 0; array < MAX_ARRAYS; array++) {
2570jmicron_next:
2571 if (!raidp[array]) {
2572 raidp[array] =
d438c7c2
TS		 2573 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
2574 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2575 }
2576 raid = raidp[array];
2577 if (raid->format && (raid->format != AR_F_JMICRON_RAID))
2578 continue;
2579
2580 for (total_disks = 0, disk = 0; disk < JM_MAX_DISKS; disk++) {
2581 if (meta->disks[disk]) {
2582 if (raid->format == AR_F_JMICRON_RAID) {
2583 if (bcmp(&meta->disks[disk],
2584 raid->disks[disk].serial, sizeof(u_int32_t))) {
2585 array++;
2586 goto jmicron_next;
2587 }
2588 }
2589 else
2590 bcopy(&meta->disks[disk],
2591 raid->disks[disk].serial, sizeof(u_int32_t));
2592 total_disks++;
2593 }
2594 }
2595 /* handle spares XXX SOS */
2596
2597 switch (meta->type) {
2598 case JM_T_RAID0:
2599 raid->type = AR_T_RAID0;
2600 raid->width = total_disks;
2601 break;
2602
2603 case JM_T_RAID1:
2604 raid->type = AR_T_RAID1;
2605 raid->width = 1;
2606 break;
2607
2608 case JM_T_RAID01:
2609 raid->type = AR_T_RAID01;
2610 raid->width = total_disks / 2;
2611 break;
2612
2613 case JM_T_RAID5:
2614 raid->type = AR_T_RAID5;
2615 raid->width = total_disks;
2616 break;
2617
2618 case JM_T_JBOD:
2619 raid->type = AR_T_SPAN;
2620 raid->width = 1;
2621 break;
2622
2623 default:
2624 device_printf(parent,
2625 "JMicron unknown RAID type 0x%02x\n", meta->type);
d438c7c2 2626 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2627 raidp[array] = NULL;
2628 goto jmicron_out;
2629 }
2630 disk_size = (meta->disk_sectors_high << 16) + meta->disk_sectors_low;
2631 raid->format = AR_F_JMICRON_RAID;
2632 strncpy(raid->name, meta->name, sizeof(meta->name));
2633 raid->generation = 0;
2634 raid->interleave = 2 << meta->stripe_shift;
2635 raid->total_disks = total_disks;
2636 raid->total_sectors = disk_size * (raid->width-(raid->type==AR_RAID5));
2637 raid->heads = 255;
2638 raid->sectors = 63;
2639 raid->cylinders = raid->total_sectors / (63 * 255);
2640 raid->offset_sectors = meta->offset * 16;
2641 raid->rebuild_lba = 0;
2642 raid->lun = array;
2643
2644 for (disk = 0; disk < raid->total_disks; disk++) {
2645 if (meta->disks[disk] == meta->disk_id) {
2646 raid->disks[disk].dev = parent;
2647 raid->disks[disk].sectors = disk_size;
2648 raid->disks[disk].flags =
2649 (AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
2650 ars->raid[raid->volume] = raid;
2651 ars->disk_number[raid->volume] = disk;
2652 retval = 1;
2653 break;
2654 }
2655 }
2656 break;
2657 }
2658jmicron_out:
d438c7c2 2659 kfree(meta, M_AR);
c1b3d7c5
TS		 2660 return retval;
2661}
2662
2663static int
2664ata_raid_jmicron_write_meta(struct ar_softc *rdp)
2665{
2666 struct jmicron_raid_conf *meta;
2667 u_int64_t disk_sectors;
2668 int disk, error = 0;
2669
978400d3
SW		 2670 meta = (struct jmicron_raid_conf *)
2671 kmalloc(sizeof(struct jmicron_raid_conf), M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2672
2673 rdp->generation++;
2674 switch (rdp->type) {
2675 case AR_T_JBOD:
2676 meta->type = JM_T_JBOD;
2677 break;
2678
2679 case AR_T_RAID0:
2680 meta->type = JM_T_RAID0;
2681 break;
2682
2683 case AR_T_RAID1:
2684 meta->type = JM_T_RAID1;
2685 break;
2686
2687 case AR_T_RAID5:
2688 meta->type = JM_T_RAID5;
2689 break;
2690
2691 case AR_T_RAID01:
2692 meta->type = JM_T_RAID01;
2693 break;
2694
2695 default:
d438c7c2 2696 kfree(meta, M_AR);
c1b3d7c5
TS		 2697 return ENODEV;
2698 }
2699 bcopy(JMICRON_MAGIC, meta->signature, sizeof(JMICRON_MAGIC));
2700 meta->version = JMICRON_VERSION;
2701 meta->offset = rdp->offset_sectors / 16;
2702 disk_sectors = rdp->total_sectors / (rdp->width - (rdp->type == AR_RAID5));
2703 meta->disk_sectors_low = disk_sectors & 0xffff;
2704 meta->disk_sectors_high = disk_sectors >> 16;
2705 strncpy(meta->name, rdp->name, sizeof(meta->name));
2706 meta->stripe_shift = ffs(rdp->interleave) - 2;
2707
2708 for (disk = 0; disk < rdp->total_disks; disk++) {
2709 if (rdp->disks[disk].serial[0])
2710 bcopy(rdp->disks[disk].serial,&meta->disks[disk],sizeof(u_int32_t));
2711 else
2712 meta->disks[disk] = (u_int32_t)(uintptr_t)rdp->disks[disk].dev;
2713 }
2714
2715 for (disk = 0; disk < rdp->total_disks; disk++) {
2716 if (rdp->disks[disk].dev) {
2717 u_int16_t checksum = 0, *ptr;
2718 int count;
2719
2720 meta->disk_id = meta->disks[disk];
2721 meta->checksum = 0;
2722 for (ptr = (u_int16_t *)meta, count = 0; count < 64; count++)
2723 checksum += *ptr++;
2724 meta->checksum -= checksum;
2725
2726 if (testing || bootverbose)
2727 ata_raid_jmicron_print_meta(meta);
2728
2729 if (ata_raid_rw(rdp->disks[disk].dev,
2730 JMICRON_LBA(rdp->disks[disk].dev),
2731 meta, sizeof(struct jmicron_raid_conf),
2732 ATA_R_WRITE | ATA_R_DIRECT)) {
2733 device_printf(rdp->disks[disk].dev, "write metadata failed\n");
2734 error = EIO;
2735 }
2736 }
2737 }
2738 /* handle spares XXX SOS */
2739
d438c7c2 2740 kfree(meta, M_AR);
c1b3d7c5
TS		 2741 return error;
2742}
2743
2744/* LSILogic V2 MegaRAID Metadata */
2745static int
2746ata_raid_lsiv2_read_meta(device_t dev, struct ar_softc **raidp)
2747{
2748 struct ata_raid_subdisk *ars = device_get_softc(dev);
2749 device_t parent = device_get_parent(dev);
2750 struct lsiv2_raid_conf *meta;
2751 struct ar_softc *raid = NULL;
2752 int array, retval = 0;
2753
978400d3
SW		 2754 meta = (struct lsiv2_raid_conf *)kmalloc(sizeof(struct lsiv2_raid_conf),
2755 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2756
2757 if (ata_raid_rw(parent, LSIV2_LBA(parent),
2758 meta, sizeof(struct lsiv2_raid_conf), ATA_R_READ)) {
2759 if (testing || bootverbose)
2760 device_printf(parent, "LSI (v2) read metadata failed\n");
2761 goto lsiv2_out;
2762 }
2763
2764 /* check if this is a LSI RAID struct */
2765 if (strncmp(meta->lsi_id, LSIV2_MAGIC, strlen(LSIV2_MAGIC))) {
2766 if (testing || bootverbose)
2767 device_printf(parent, "LSI (v2) check1 failed\n");
2768 goto lsiv2_out;
2769 }
2770
2771 if (testing || bootverbose)
2772 ata_raid_lsiv2_print_meta(meta);
2773
2774 /* now convert LSI (v2) config meta into our generic form */
2775 for (array = 0; array < MAX_ARRAYS; array++) {
2776 int raid_entry, conf_entry;
2777
2778 if (!raidp[array + meta->raid_number]) {
2779 raidp[array + meta->raid_number] =
d438c7c2
TS		 2780 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
2781 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2782 }
2783 raid = raidp[array + meta->raid_number];
2784 if (raid->format && (raid->format != AR_F_LSIV2_RAID))
2785 continue;
2786
2787 if (raid->magic_0 &&
2788 ((raid->magic_0 != meta->timestamp) ||
2789 (raid->magic_1 != meta->raid_number)))
2790 continue;
2791
2792 array += meta->raid_number;
2793
2794 raid_entry = meta->raid_number;
2795 conf_entry = (meta->configs[raid_entry].raid.config_offset >> 4) +
2796 meta->disk_number - 1;
2797
2798 switch (meta->configs[raid_entry].raid.type) {
2799 case LSIV2_T_RAID0:
2800 raid->magic_0 = meta->timestamp;
2801 raid->magic_1 = meta->raid_number;
2802 raid->type = AR_T_RAID0;
2803 raid->interleave = meta->configs[raid_entry].raid.stripe_sectors;
2804 raid->width = meta->configs[raid_entry].raid.array_width;
2805 break;
2806
2807 case LSIV2_T_RAID1:
2808 raid->magic_0 = meta->timestamp;
2809 raid->magic_1 = meta->raid_number;
2810 raid->type = AR_T_RAID1;
2811 raid->width = meta->configs[raid_entry].raid.array_width;
2812 break;
2813
2814 case LSIV2_T_RAID0 | LSIV2_T_RAID1:
2815 raid->magic_0 = meta->timestamp;
2816 raid->magic_1 = meta->raid_number;
2817 raid->type = AR_T_RAID01;
2818 raid->interleave = meta->configs[raid_entry].raid.stripe_sectors;
2819 raid->width = meta->configs[raid_entry].raid.array_width;
2820 break;
2821
2822 default:
2823 device_printf(parent, "LSI v2 unknown RAID type 0x%02x\n",
2824 meta->configs[raid_entry].raid.type);
d438c7c2 2825 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2826 raidp[array] = NULL;
2827 goto lsiv2_out;
2828 }
2829
2830 raid->format = AR_F_LSIV2_RAID;
2831 raid->generation = 0;
2832 raid->total_disks = meta->configs[raid_entry].raid.disk_count;
2833 raid->total_sectors = meta->configs[raid_entry].raid.total_sectors;
2834 raid->heads = 255;
2835 raid->sectors = 63;
2836 raid->cylinders = raid->total_sectors / (63 * 255);
2837 raid->offset_sectors = 0;
2838 raid->rebuild_lba = 0;
2839 raid->lun = array;
2840
2841 if (meta->configs[conf_entry].disk.device != LSIV2_D_NONE) {
2842 raid->disks[meta->disk_number].dev = parent;
2843 raid->disks[meta->disk_number].sectors =
2844 meta->configs[conf_entry].disk.disk_sectors;
2845 raid->disks[meta->disk_number].flags =
2846 (AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
2847 ars->raid[raid->volume] = raid;
2848 ars->disk_number[raid->volume] = meta->disk_number;
2849 retval = 1;
2850 }
2851 else
2852 raid->disks[meta->disk_number].flags &= ~AR_DF_ONLINE;
2853
2854 break;
2855 }
2856
2857lsiv2_out:
d438c7c2 2858 kfree(meta, M_AR);
c1b3d7c5
TS		 2859 return retval;
2860}
2861
2862/* LSILogic V3 MegaRAID Metadata */
2863static int
2864ata_raid_lsiv3_read_meta(device_t dev, struct ar_softc **raidp)
2865{
2866 struct ata_raid_subdisk *ars = device_get_softc(dev);
2867 device_t parent = device_get_parent(dev);
2868 struct lsiv3_raid_conf *meta;
2869 struct ar_softc *raid = NULL;
2870 u_int8_t checksum, *ptr;
2871 int array, entry, count, disk_number, retval = 0;
2872
978400d3
SW		 2873 meta = (struct lsiv3_raid_conf *)kmalloc(sizeof(struct lsiv3_raid_conf),
2874 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2875
2876 if (ata_raid_rw(parent, LSIV3_LBA(parent),
2877 meta, sizeof(struct lsiv3_raid_conf), ATA_R_READ)) {
2878 if (testing || bootverbose)
2879 device_printf(parent, "LSI (v3) read metadata failed\n");
2880 goto lsiv3_out;
2881 }
2882
2883 /* check if this is a LSI RAID struct */
2884 if (strncmp(meta->lsi_id, LSIV3_MAGIC, strlen(LSIV3_MAGIC))) {
2885 if (testing || bootverbose)
2886 device_printf(parent, "LSI (v3) check1 failed\n");
2887 goto lsiv3_out;
2888 }
2889
2890 /* check if the checksum is OK */
2891 for (checksum = 0, ptr = meta->lsi_id, count = 0; count < 512; count++)
2892 checksum += *ptr++;
2893 if (checksum) {
2894 if (testing || bootverbose)
2895 device_printf(parent, "LSI (v3) check2 failed\n");
2896 goto lsiv3_out;
2897 }
2898
2899 if (testing || bootverbose)
2900 ata_raid_lsiv3_print_meta(meta);
2901
2902 /* now convert LSI (v3) config meta into our generic form */
2903 for (array = 0, entry = 0; array < MAX_ARRAYS && entry < 8;) {
2904 if (!raidp[array]) {
2905 raidp[array] =
d438c7c2
TS		 2906 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
2907 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 2908 }
2909 raid = raidp[array];
2910 if (raid->format && (raid->format != AR_F_LSIV3_RAID)) {
2911 array++;
2912 continue;
2913 }
2914
2915 if ((raid->format == AR_F_LSIV3_RAID) &&
2916 (raid->magic_0 != meta->timestamp)) {
2917 array++;
2918 continue;
2919 }
2920
2921 switch (meta->raid[entry].total_disks) {
2922 case 0:
2923 entry++;
2924 continue;
2925 case 1:
2926 if (meta->raid[entry].device == meta->device) {
2927 disk_number = 0;
2928 break;
2929 }
2930 if (raid->format)
2931 array++;
2932 entry++;
2933 continue;
2934 case 2:
2935 disk_number = (meta->device & (LSIV3_D_DEVICE|LSIV3_D_CHANNEL))?1:0;
2936 break;
2937 default:
2938 device_printf(parent, "lsiv3 > 2 disk support untested!!\n");
2939 disk_number = (meta->device & LSIV3_D_DEVICE ? 1 : 0) +
2940 (meta->device & LSIV3_D_CHANNEL ? 2 : 0);
2941 break;
2942 }
2943
2944 switch (meta->raid[entry].type) {
2945 case LSIV3_T_RAID0:
2946 raid->type = AR_T_RAID0;
2947 raid->width = meta->raid[entry].total_disks;
2948 break;
2949
2950 case LSIV3_T_RAID1:
2951 raid->type = AR_T_RAID1;
2952 raid->width = meta->raid[entry].array_width;
2953 break;
2954
2955 default:
2956 device_printf(parent, "LSI v3 unknown RAID type 0x%02x\n",
2957 meta->raid[entry].type);
d438c7c2 2958 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 2959 raidp[array] = NULL;
2960 entry++;
2961 continue;
2962 }
2963
2964 raid->magic_0 = meta->timestamp;
2965 raid->format = AR_F_LSIV3_RAID;
2966 raid->generation = 0;
2967 raid->interleave = meta->raid[entry].stripe_pages * 8;
2968 raid->total_disks = meta->raid[entry].total_disks;
2969 raid->total_sectors = raid->width * meta->raid[entry].sectors;
2970 raid->heads = 255;
2971 raid->sectors = 63;
2972 raid->cylinders = raid->total_sectors / (63 * 255);
2973 raid->offset_sectors = meta->raid[entry].offset;
2974 raid->rebuild_lba = 0;
2975 raid->lun = array;
2976
2977 raid->disks[disk_number].dev = parent;
2978 raid->disks[disk_number].sectors = raid->total_sectors / raid->width;
2979 raid->disks[disk_number].flags =
2980 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
2981 ars->raid[raid->volume] = raid;
2982 ars->disk_number[raid->volume] = disk_number;
2983 retval = 1;
2984 entry++;
2985 array++;
2986 }
2987
2988lsiv3_out:
d438c7c2 2989 kfree(meta, M_AR);
c1b3d7c5
TS		 2990 return retval;
2991}
2992
2993/* nVidia MediaShield Metadata */
2994static int
2995ata_raid_nvidia_read_meta(device_t dev, struct ar_softc **raidp)
2996{
2997 struct ata_raid_subdisk *ars = device_get_softc(dev);
2998 device_t parent = device_get_parent(dev);
2999 struct nvidia_raid_conf *meta;
3000 struct ar_softc *raid = NULL;
3001 u_int32_t checksum, *ptr;
3002 int array, count, retval = 0;
3003
978400d3
SW		 3004 meta = (struct nvidia_raid_conf *)kmalloc(sizeof(struct nvidia_raid_conf),
3005 M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3006
3007 if (ata_raid_rw(parent, NVIDIA_LBA(parent),
3008 meta, sizeof(struct nvidia_raid_conf), ATA_R_READ)) {
3009 if (testing || bootverbose)
3010 device_printf(parent, "nVidia read metadata failed\n");
3011 goto nvidia_out;
3012 }
3013
3014 /* check if this is a nVidia RAID struct */
3015 if (strncmp(meta->nvidia_id, NV_MAGIC, strlen(NV_MAGIC))) {
3016 if (testing || bootverbose)
3017 device_printf(parent, "nVidia check1 failed\n");
3018 goto nvidia_out;
3019 }
3020
3021 /* check if the checksum is OK */
3022 for (checksum = 0, ptr = (u_int32_t*)meta, count = 0;
3023 count < meta->config_size; count++)
3024 checksum += *ptr++;
3025 if (checksum) {
3026 if (testing || bootverbose)
3027 device_printf(parent, "nVidia check2 failed\n");
3028 goto nvidia_out;
3029 }
3030
3031 if (testing || bootverbose)
3032 ata_raid_nvidia_print_meta(meta);
3033
3034 /* now convert nVidia meta into our generic form */
3035 for (array = 0; array < MAX_ARRAYS; array++) {
3036 if (!raidp[array]) {
3037 raidp[array] =
d438c7c2
TS		 3038 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
3039 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3040 }
3041 raid = raidp[array];
3042 if (raid->format && (raid->format != AR_F_NVIDIA_RAID))
3043 continue;
3044
3045 if (raid->format == AR_F_NVIDIA_RAID &&
3046 ((raid->magic_0 != meta->magic_1) ||
3047 (raid->magic_1 != meta->magic_2))) {
3048 continue;
3049 }
3050
3051 switch (meta->type) {
3052 case NV_T_SPAN:
3053 raid->type = AR_T_SPAN;
3054 break;
3055
3056 case NV_T_RAID0:
3057 raid->type = AR_T_RAID0;
3058 break;
3059
3060 case NV_T_RAID1:
3061 raid->type = AR_T_RAID1;
3062 break;
3063
3064 case NV_T_RAID5:
3065 raid->type = AR_T_RAID5;
3066 break;
3067
3068 case NV_T_RAID01:
3069 raid->type = AR_T_RAID01;
3070 break;
3071
3072 default:
3073 device_printf(parent, "nVidia unknown RAID type 0x%02x\n",
3074 meta->type);
d438c7c2 3075 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 3076 raidp[array] = NULL;
3077 goto nvidia_out;
3078 }
3079 raid->magic_0 = meta->magic_1;
3080 raid->magic_1 = meta->magic_2;
3081 raid->format = AR_F_NVIDIA_RAID;
3082 raid->generation = 0;
3083 raid->interleave = meta->stripe_sectors;
3084 raid->width = meta->array_width;
3085 raid->total_disks = meta->total_disks;
3086 raid->total_sectors = meta->total_sectors;
3087 raid->heads = 255;
3088 raid->sectors = 63;
3089 raid->cylinders = raid->total_sectors / (63 * 255);
3090 raid->offset_sectors = 0;
3091 raid->rebuild_lba = meta->rebuild_lba;
3092 raid->lun = array;
3093 raid->status = AR_S_READY;
3094 if (meta->status & NV_S_DEGRADED)
3095 raid->status |= AR_S_DEGRADED;
3096
3097 raid->disks[meta->disk_number].dev = parent;
3098 raid->disks[meta->disk_number].sectors =
3099 raid->total_sectors / raid->width;
3100 raid->disks[meta->disk_number].flags =
3101 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
3102 ars->raid[raid->volume] = raid;
3103 ars->disk_number[raid->volume] = meta->disk_number;
3104 retval = 1;
3105 break;
3106 }
3107
3108nvidia_out:
d438c7c2 3109 kfree(meta, M_AR);
c1b3d7c5
TS		 3110 return retval;
3111}
3112
3113/* Promise FastTrak Metadata */
3114static int
3115ata_raid_promise_read_meta(device_t dev, struct ar_softc **raidp, int native)
3116{
3117 struct ata_raid_subdisk *ars = device_get_softc(dev);
3118 device_t parent = device_get_parent(dev);
3119 struct promise_raid_conf *meta;
3120 struct ar_softc *raid;
3121 u_int32_t checksum, *ptr;
3122 int array, count, disk, disksum = 0, retval = 0;
3123
978400d3
SW		 3124 meta = (struct promise_raid_conf *)
3125 kmalloc(sizeof(struct promise_raid_conf), M_AR, M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3126
3127 if (ata_raid_rw(parent, PROMISE_LBA(parent),
3128 meta, sizeof(struct promise_raid_conf), ATA_R_READ)) {
3129 if (testing || bootverbose)
3130 device_printf(parent, "%s read metadata failed\n",
3131 native ? "FreeBSD" : "Promise");
3132 goto promise_out;
3133 }
3134
3135 /* check the signature */
3136 if (native) {
3137 if (strncmp(meta->promise_id, ATA_MAGIC, strlen(ATA_MAGIC))) {
3138 if (testing || bootverbose)
3139 device_printf(parent, "FreeBSD check1 failed\n");
3140 goto promise_out;
3141 }
3142 }
3143 else {
3144 if (strncmp(meta->promise_id, PR_MAGIC, strlen(PR_MAGIC))) {
3145 if (testing || bootverbose)
3146 device_printf(parent, "Promise check1 failed\n");
3147 goto promise_out;
3148 }
3149 }
3150
3151 /* check if the checksum is OK */
3152 for (checksum = 0, ptr = (u_int32_t *)meta, count = 0; count < 511; count++)
3153 checksum += *ptr++;
3154 if (checksum != *ptr) {
3155 if (testing || bootverbose)
3156 device_printf(parent, "%s check2 failed\n",
3157 native ? "FreeBSD" : "Promise");
3158 goto promise_out;
3159 }
3160
3161 /* check on disk integrity status */
3162 if (meta->raid.integrity != PR_I_VALID) {
3163 if (testing || bootverbose)
3164 device_printf(parent, "%s check3 failed\n",
3165 native ? "FreeBSD" : "Promise");
3166 goto promise_out;
3167 }
3168
3169 if (testing || bootverbose)
3170 ata_raid_promise_print_meta(meta);
3171
3172 /* now convert Promise metadata into our generic form */
3173 for (array = 0; array < MAX_ARRAYS; array++) {
3174 if (!raidp[array]) {
3175 raidp[array] =
d438c7c2
TS		 3176 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
3177 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3178 }
3179 raid = raidp[array];
3180 if (raid->format &&
3181 (raid->format != (native ? AR_F_FREEBSD_RAID : AR_F_PROMISE_RAID)))
3182 continue;
3183
3184 if ((raid->format == (native ? AR_F_FREEBSD_RAID : AR_F_PROMISE_RAID))&&
3185 !(meta->raid.magic_1 == (raid->magic_1)))
3186 continue;
3187
3188 /* update our knowledge about the array config based on generation */
3189 if (!meta->raid.generation || meta->raid.generation > raid->generation){
3190 switch (meta->raid.type) {
3191 case PR_T_SPAN:
3192 raid->type = AR_T_SPAN;
3193 break;
3194
3195 case PR_T_JBOD:
3196 raid->type = AR_T_JBOD;
3197 break;
3198
3199 case PR_T_RAID0:
3200 raid->type = AR_T_RAID0;
3201 break;
3202
3203 case PR_T_RAID1:
3204 raid->type = AR_T_RAID1;
3205 if (meta->raid.array_width > 1)
3206 raid->type = AR_T_RAID01;
3207 break;
3208
3209 case PR_T_RAID5:
3210 raid->type = AR_T_RAID5;
3211 break;
3212
3213 default:
3214 device_printf(parent, "%s unknown RAID type 0x%02x\n",
3215 native ? "FreeBSD" : "Promise", meta->raid.type);
d438c7c2 3216 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 3217 raidp[array] = NULL;
3218 goto promise_out;
3219 }
3220 raid->magic_1 = meta->raid.magic_1;
3221 raid->format = (native ? AR_F_FREEBSD_RAID : AR_F_PROMISE_RAID);
3222 raid->generation = meta->raid.generation;
3223 raid->interleave = 1 << meta->raid.stripe_shift;
3224 raid->width = meta->raid.array_width;
3225 raid->total_disks = meta->raid.total_disks;
3226 raid->heads = meta->raid.heads + 1;
3227 raid->sectors = meta->raid.sectors;
3228 raid->cylinders = meta->raid.cylinders + 1;
3229 raid->total_sectors = meta->raid.total_sectors;
3230 raid->offset_sectors = 0;
3231 raid->rebuild_lba = meta->raid.rebuild_lba;
3232 raid->lun = array;
3233 if ((meta->raid.status &
3234 (PR_S_VALID | PR_S_ONLINE | PR_S_INITED | PR_S_READY)) ==
3235 (PR_S_VALID | PR_S_ONLINE | PR_S_INITED | PR_S_READY)) {
3236 raid->status |= AR_S_READY;
3237 if (meta->raid.status & PR_S_DEGRADED)
3238 raid->status |= AR_S_DEGRADED;
3239 }
3240 else
3241 raid->status &= ~AR_S_READY;
3242
3243 /* convert disk flags to our internal types */
3244 for (disk = 0; disk < meta->raid.total_disks; disk++) {
3245 raid->disks[disk].dev = NULL;
3246 raid->disks[disk].flags = 0;
3247 *((u_int64_t *)(raid->disks[disk].serial)) =
3248 meta->raid.disk[disk].magic_0;
3249 disksum += meta->raid.disk[disk].flags;
3250 if (meta->raid.disk[disk].flags & PR_F_ONLINE)
3251 raid->disks[disk].flags |= AR_DF_ONLINE;
3252 if (meta->raid.disk[disk].flags & PR_F_ASSIGNED)
3253 raid->disks[disk].flags |= AR_DF_ASSIGNED;
3254 if (meta->raid.disk[disk].flags & PR_F_SPARE) {
3255 raid->disks[disk].flags &= ~(AR_DF_ONLINE | AR_DF_ASSIGNED);
3256 raid->disks[disk].flags |= AR_DF_SPARE;
3257 }
3258 if (meta->raid.disk[disk].flags & (PR_F_REDIR | PR_F_DOWN))
3259 raid->disks[disk].flags &= ~AR_DF_ONLINE;
3260 }
3261 if (!disksum) {
3262 device_printf(parent, "%s subdisks has no flags\n",
3263 native ? "FreeBSD" : "Promise");
d438c7c2 3264 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 3265 raidp[array] = NULL;
3266 goto promise_out;
3267 }
3268 }
3269 if (meta->raid.generation >= raid->generation) {
3270 int disk_number = meta->raid.disk_number;
3271
3272 if (raid->disks[disk_number].flags && (meta->magic_0 ==
3273 *((u_int64_t *)(raid->disks[disk_number].serial)))) {
3274 raid->disks[disk_number].dev = parent;
3275 raid->disks[disk_number].flags |= AR_DF_PRESENT;
3276 raid->disks[disk_number].sectors = meta->raid.disk_sectors;
3277 if ((raid->disks[disk_number].flags &
3278 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE)) ==
3279 (AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE)) {
3280 ars->raid[raid->volume] = raid;
3281 ars->disk_number[raid->volume] = disk_number;
3282 retval = 1;
3283 }
3284 }
3285 }
3286 break;
3287 }
3288
3289promise_out:
d438c7c2 3290 kfree(meta, M_AR);
c1b3d7c5
TS		 3291 return retval;
3292}
3293
3294static int
3295ata_raid_promise_write_meta(struct ar_softc *rdp)
3296{
3297 struct promise_raid_conf *meta;
3298 struct timeval timestamp;
3299 u_int32_t *ckptr;
3300 int count, disk, drive, error = 0;
3301
978400d3
SW		 3302 meta = (struct promise_raid_conf *)
3303 kmalloc(sizeof(struct promise_raid_conf), M_AR, M_WAITOK);
c1b3d7c5
TS		 3304
3305 rdp->generation++;
3306 microtime(&timestamp);
3307
3308 for (disk = 0; disk < rdp->total_disks; disk++) {
3309 for (count = 0; count < sizeof(struct promise_raid_conf); count++)
3310 *(((u_int8_t *)meta) + count) = 255 - (count % 256);
3311 meta->dummy_0 = 0x00020000;
3312 meta->raid.disk_number = disk;
3313
3314 if (rdp->disks[disk].dev) {
3315 struct ata_device *atadev = device_get_softc(rdp->disks[disk].dev);
3316 struct ata_channel *ch =
3317 device_get_softc(device_get_parent(rdp->disks[disk].dev));
3318
3319 meta->raid.channel = ch->unit;
3320 meta->raid.device = ATA_DEV(atadev->unit);
3321 meta->raid.disk_sectors = rdp->disks[disk].sectors;
3322 meta->raid.disk_offset = rdp->offset_sectors;
3323 }
3324 else {
3325 meta->raid.channel = 0;
3326 meta->raid.device = 0;
3327 meta->raid.disk_sectors = 0;
3328 meta->raid.disk_offset = 0;
3329 }
3330 meta->magic_0 = PR_MAGIC0(meta->raid) | timestamp.tv_sec;
3331 meta->magic_1 = timestamp.tv_sec >> 16;
3332 meta->magic_2 = timestamp.tv_sec;
3333 meta->raid.integrity = PR_I_VALID;
3334 meta->raid.magic_0 = meta->magic_0;
3335 meta->raid.rebuild_lba = rdp->rebuild_lba;
3336 meta->raid.generation = rdp->generation;
3337
3338 if (rdp->status & AR_S_READY) {
3339 meta->raid.flags = (PR_F_VALID | PR_F_ASSIGNED | PR_F_ONLINE);
3340 meta->raid.status =
3341 (PR_S_VALID | PR_S_ONLINE | PR_S_INITED | PR_S_READY);
3342 if (rdp->status & AR_S_DEGRADED)
3343 meta->raid.status |= PR_S_DEGRADED;
3344 else
3345 meta->raid.status |= PR_S_FUNCTIONAL;
3346 }
3347 else {
3348 meta->raid.flags = PR_F_DOWN;
3349 meta->raid.status = 0;
3350 }
3351
3352 switch (rdp->type) {
3353 case AR_T_RAID0:
3354 meta->raid.type = PR_T_RAID0;
3355 break;
3356 case AR_T_RAID1:
3357 meta->raid.type = PR_T_RAID1;
3358 break;
3359 case AR_T_RAID01:
3360 meta->raid.type = PR_T_RAID1;
3361 break;
3362 case AR_T_RAID5:
3363 meta->raid.type = PR_T_RAID5;
3364 break;
3365 case AR_T_SPAN:
3366 meta->raid.type = PR_T_SPAN;
3367 break;
3368 case AR_T_JBOD:
3369 meta->raid.type = PR_T_JBOD;
3370 break;
3371 default:
d438c7c2 3372 kfree(meta, M_AR);
c1b3d7c5
TS		 3373 return ENODEV;
3374 }
3375
3376 meta->raid.total_disks = rdp->total_disks;
3377 meta->raid.stripe_shift = ffs(rdp->interleave) - 1;
3378 meta->raid.array_width = rdp->width;
3379 meta->raid.array_number = rdp->lun;
3380 meta->raid.total_sectors = rdp->total_sectors;
3381 meta->raid.cylinders = rdp->cylinders - 1;
3382 meta->raid.heads = rdp->heads - 1;
3383 meta->raid.sectors = rdp->sectors;
3384 meta->raid.magic_1 = (u_int64_t)meta->magic_2<<16 | meta->magic_1;
3385
3386 bzero(&meta->raid.disk, 8 * 12);
3387 for (drive = 0; drive < rdp->total_disks; drive++) {
3388 meta->raid.disk[drive].flags = 0;
3389 if (rdp->disks[drive].flags & AR_DF_PRESENT)
3390 meta->raid.disk[drive].flags |= PR_F_VALID;
3391 if (rdp->disks[drive].flags & AR_DF_ASSIGNED)
3392 meta->raid.disk[drive].flags |= PR_F_ASSIGNED;
3393 if (rdp->disks[drive].flags & AR_DF_ONLINE)
3394 meta->raid.disk[drive].flags |= PR_F_ONLINE;
3395 else
3396 if (rdp->disks[drive].flags & AR_DF_PRESENT)
3397 meta->raid.disk[drive].flags = (PR_F_REDIR | PR_F_DOWN);
3398 if (rdp->disks[drive].flags & AR_DF_SPARE)
3399 meta->raid.disk[drive].flags |= PR_F_SPARE;
3400 meta->raid.disk[drive].dummy_0 = 0x0;
3401 if (rdp->disks[drive].dev) {
3402 struct ata_channel *ch =
3403 device_get_softc(device_get_parent(rdp->disks[drive].dev));
3404 struct ata_device *atadev =
3405 device_get_softc(rdp->disks[drive].dev);
3406
3407 meta->raid.disk[drive].channel = ch->unit;
3408 meta->raid.disk[drive].device = ATA_DEV(atadev->unit);
3409 }
3410 meta->raid.disk[drive].magic_0 =
3411 PR_MAGIC0(meta->raid.disk[drive]) | timestamp.tv_sec;
3412 }
3413
3414 if (rdp->disks[disk].dev) {
3415 if ((rdp->disks[disk].flags & (AR_DF_PRESENT | AR_DF_ONLINE)) ==
3416 (AR_DF_PRESENT | AR_DF_ONLINE)) {
3417 if (rdp->format == AR_F_FREEBSD_RAID)
3418 bcopy(ATA_MAGIC, meta->promise_id, sizeof(ATA_MAGIC));
3419 else
3420 bcopy(PR_MAGIC, meta->promise_id, sizeof(PR_MAGIC));
3421 }
3422 else
3423 bzero(meta->promise_id, sizeof(meta->promise_id));
3424 meta->checksum = 0;
3425 for (ckptr = (int32_t *)meta, count = 0; count < 511; count++)
3426 meta->checksum += *ckptr++;
3427 if (testing || bootverbose)
3428 ata_raid_promise_print_meta(meta);
3429 if (ata_raid_rw(rdp->disks[disk].dev,
3430 PROMISE_LBA(rdp->disks[disk].dev),
3431 meta, sizeof(struct promise_raid_conf),
3432 ATA_R_WRITE | ATA_R_DIRECT)) {
3433 device_printf(rdp->disks[disk].dev, "write metadata failed\n");
3434 error = EIO;
3435 }
3436 }
3437 }
d438c7c2 3438 kfree(meta, M_AR);
c1b3d7c5
TS		 3439 return error;
3440}
3441
3442/* Silicon Image Medley Metadata */
3443static int
3444ata_raid_sii_read_meta(device_t dev, struct ar_softc **raidp)
3445{
3446 struct ata_raid_subdisk *ars = device_get_softc(dev);
3447 device_t parent = device_get_parent(dev);
3448 struct sii_raid_conf *meta;
3449 struct ar_softc *raid = NULL;
3450 u_int16_t checksum, *ptr;
3451 int array, count, disk, retval = 0;
3452
978400d3
SW		 3453 meta = (struct sii_raid_conf *)kmalloc(sizeof(struct sii_raid_conf), M_AR,
3454 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3455
3456 if (ata_raid_rw(parent, SII_LBA(parent),
3457 meta, sizeof(struct sii_raid_conf), ATA_R_READ)) {
3458 if (testing || bootverbose)
3459 device_printf(parent, "Silicon Image read metadata failed\n");
3460 goto sii_out;
3461 }
3462
3463 /* check if this is a Silicon Image (Medley) RAID struct */
3464 for (checksum = 0, ptr = (u_int16_t *)meta, count = 0; count < 160; count++)
3465 checksum += *ptr++;
3466 if (checksum) {
3467 if (testing || bootverbose)
3468 device_printf(parent, "Silicon Image check1 failed\n");
3469 goto sii_out;
3470 }
3471
3472 for (checksum = 0, ptr = (u_int16_t *)meta, count = 0; count < 256; count++)
3473 checksum += *ptr++;
3474 if (checksum != meta->checksum_1) {
3475 if (testing || bootverbose)
3476 device_printf(parent, "Silicon Image check2 failed\n");
3477 goto sii_out;
3478 }
3479
3480 /* check verison */
3481 if (meta->version_major != 0x0002 ||
3482 (meta->version_minor != 0x0000 && meta->version_minor != 0x0001)) {
3483 if (testing || bootverbose)
3484 device_printf(parent, "Silicon Image check3 failed\n");
3485 goto sii_out;
3486 }
3487
3488 if (testing || bootverbose)
3489 ata_raid_sii_print_meta(meta);
3490
3491 /* now convert Silicon Image meta into our generic form */
3492 for (array = 0; array < MAX_ARRAYS; array++) {
3493 if (!raidp[array]) {
3494 raidp[array] =
d438c7c2
TS		 3495 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
3496 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3497 }
3498 raid = raidp[array];
3499 if (raid->format && (raid->format != AR_F_SII_RAID))
3500 continue;
3501
3502 if (raid->format == AR_F_SII_RAID &&
3503 (raid->magic_0 != *((u_int64_t *)meta->timestamp))) {
3504 continue;
3505 }
3506
3507 /* update our knowledge about the array config based on generation */
3508 if (!meta->generation || meta->generation > raid->generation) {
3509 switch (meta->type) {
3510 case SII_T_RAID0:
3511 raid->type = AR_T_RAID0;
3512 break;
3513
3514 case SII_T_RAID1:
3515 raid->type = AR_T_RAID1;
3516 break;
3517
3518 case SII_T_RAID01:
3519 raid->type = AR_T_RAID01;
3520 break;
3521
3522 case SII_T_SPARE:
3523 device_printf(parent, "Silicon Image SPARE disk\n");
d438c7c2 3524 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 3525 raidp[array] = NULL;
3526 goto sii_out;
3527
3528 default:
3529 device_printf(parent,"Silicon Image unknown RAID type 0x%02x\n",
3530 meta->type);
d438c7c2 3531 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 3532 raidp[array] = NULL;
3533 goto sii_out;
3534 }
3535 raid->magic_0 = *((u_int64_t *)meta->timestamp);
3536 raid->format = AR_F_SII_RAID;
3537 raid->generation = meta->generation;
3538 raid->interleave = meta->stripe_sectors;
3539 raid->width = (meta->raid0_disks != 0xff) ? meta->raid0_disks : 1;
3540 raid->total_disks =
3541 ((meta->raid0_disks != 0xff) ? meta->raid0_disks : 0) +
3542 ((meta->raid1_disks != 0xff) ? meta->raid1_disks : 0);
3543 raid->total_sectors = meta->total_sectors;
3544 raid->heads = 255;
3545 raid->sectors = 63;
3546 raid->cylinders = raid->total_sectors / (63 * 255);
3547 raid->offset_sectors = 0;
3548 raid->rebuild_lba = meta->rebuild_lba;
3549 raid->lun = array;
3550 strncpy(raid->name, meta->name,
3551 min(sizeof(raid->name), sizeof(meta->name)));
3552
3553 /* clear out any old info */
3554 if (raid->generation) {
3555 for (disk = 0; disk < raid->total_disks; disk++) {
3556 raid->disks[disk].dev = NULL;
3557 raid->disks[disk].flags = 0;
3558 }
3559 }
3560 }
3561 if (meta->generation >= raid->generation) {
3562 /* XXX SOS add check for the right physical disk by serial# */
3563 if (meta->status & SII_S_READY) {
3564 int disk_number = (raid->type == AR_T_RAID01) ?
3565 meta->raid1_ident + (meta->raid0_ident << 1) :
3566 meta->disk_number;
3567
3568 raid->disks[disk_number].dev = parent;
3569 raid->disks[disk_number].sectors =
3570 raid->total_sectors / raid->width;
3571 raid->disks[disk_number].flags =
3572 (AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
3573 ars->raid[raid->volume] = raid;
3574 ars->disk_number[raid->volume] = disk_number;
3575 retval = 1;
3576 }
3577 }
3578 break;
3579 }
3580
3581sii_out:
d438c7c2 3582 kfree(meta, M_AR);
c1b3d7c5
TS		 3583 return retval;
3584}
3585
3586/* Silicon Integrated Systems Metadata */
3587static int
3588ata_raid_sis_read_meta(device_t dev, struct ar_softc **raidp)
3589{
3590 struct ata_raid_subdisk *ars = device_get_softc(dev);
3591 device_t parent = device_get_parent(dev);
3592 struct sis_raid_conf *meta;
3593 struct ar_softc *raid = NULL;
3594 int array, disk_number, drive, retval = 0;
3595
978400d3
SW		 3596 meta = (struct sis_raid_conf *)kmalloc(sizeof(struct sis_raid_conf), M_AR,
3597 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3598
3599 if (ata_raid_rw(parent, SIS_LBA(parent),
3600 meta, sizeof(struct sis_raid_conf), ATA_R_READ)) {
3601 if (testing || bootverbose)
3602 device_printf(parent,
3603 "Silicon Integrated Systems read metadata failed\n");
3604 }
3605
3606 /* check for SiS magic */
3607 if (meta->magic != SIS_MAGIC) {
3608 if (testing || bootverbose)
3609 device_printf(parent,
3610 "Silicon Integrated Systems check1 failed\n");
3611 goto sis_out;
3612 }
3613
3614 if (testing || bootverbose)
3615 ata_raid_sis_print_meta(meta);
3616
3617 /* now convert SiS meta into our generic form */
3618 for (array = 0; array < MAX_ARRAYS; array++) {
3619 if (!raidp[array]) {
3620 raidp[array] =
d438c7c2
TS		 3621 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
3622 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3623 }
3624
3625 raid = raidp[array];
3626 if (raid->format && (raid->format != AR_F_SIS_RAID))
3627 continue;
3628
3629 if ((raid->format == AR_F_SIS_RAID) &&
3630 ((raid->magic_0 != meta->controller_pci_id) ||
3631 (raid->magic_1 != meta->timestamp))) {
3632 continue;
3633 }
3634
3635 switch (meta->type_total_disks & SIS_T_MASK) {
3636 case SIS_T_JBOD:
3637 raid->type = AR_T_JBOD;
3638 raid->width = (meta->type_total_disks & SIS_D_MASK);
3639 raid->total_sectors += SIS_LBA(parent);
3640 break;
3641
3642 case SIS_T_RAID0:
3643 raid->type = AR_T_RAID0;
3644 raid->width = (meta->type_total_disks & SIS_D_MASK);
3645 if (!raid->total_sectors ||
3646 (raid->total_sectors > (raid->width * SIS_LBA(parent))))
3647 raid->total_sectors = raid->width * SIS_LBA(parent);
3648 break;
3649
3650 case SIS_T_RAID1:
3651 raid->type = AR_T_RAID1;
3652 raid->width = 1;
3653 if (!raid->total_sectors || (raid->total_sectors > SIS_LBA(parent)))
3654 raid->total_sectors = SIS_LBA(parent);
3655 break;
3656
3657 default:
3658 device_printf(parent, "Silicon Integrated Systems "
3659 "unknown RAID type 0x%08x\n", meta->magic);
d438c7c2 3660 kfree(raidp[array], M_AR);
c1b3d7c5
TS		 3661 raidp[array] = NULL;
3662 goto sis_out;
3663 }
3664 raid->magic_0 = meta->controller_pci_id;
3665 raid->magic_1 = meta->timestamp;
3666 raid->format = AR_F_SIS_RAID;
3667 raid->generation = 0;
3668 raid->interleave = meta->stripe_sectors;
3669 raid->total_disks = (meta->type_total_disks & SIS_D_MASK);
3670 raid->heads = 255;
3671 raid->sectors = 63;
3672 raid->cylinders = raid->total_sectors / (63 * 255);
3673 raid->offset_sectors = 0;
3674 raid->rebuild_lba = 0;
3675 raid->lun = array;
3676 /* XXX SOS if total_disks > 2 this doesn't float */
3677 if (((meta->disks & SIS_D_MASTER) >> 4) == meta->disk_number)
3678 disk_number = 0;
3679 else
3680 disk_number = 1;
3681
3682 for (drive = 0; drive < raid->total_disks; drive++) {
3683 raid->disks[drive].sectors = raid->total_sectors/raid->width;
3684 if (drive == disk_number) {
3685 raid->disks[disk_number].dev = parent;
3686 raid->disks[disk_number].flags =
3687 (AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
3688 ars->raid[raid->volume] = raid;
3689 ars->disk_number[raid->volume] = disk_number;
3690 }
3691 }
3692 retval = 1;
3693 break;
3694 }
3695
3696sis_out:
d438c7c2 3697 kfree(meta, M_AR);
c1b3d7c5
TS		 3698 return retval;
3699}
3700
3701static int
3702ata_raid_sis_write_meta(struct ar_softc *rdp)
3703{
3704 struct sis_raid_conf *meta;
3705 struct timeval timestamp;
3706 int disk, error = 0;
3707
978400d3
SW		 3708 meta = (struct sis_raid_conf *)kmalloc(sizeof(struct sis_raid_conf), M_AR,
3709 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3710
3711 rdp->generation++;
3712 microtime(&timestamp);
3713
3714 meta->magic = SIS_MAGIC;
3715 /* XXX SOS if total_disks > 2 this doesn't float */
3716 for (disk = 0; disk < rdp->total_disks; disk++) {
3717 if (rdp->disks[disk].dev) {
3718 struct ata_channel *ch =
3719 device_get_softc(device_get_parent(rdp->disks[disk].dev));
3720 struct ata_device *atadev = device_get_softc(rdp->disks[disk].dev);
3721 int disk_number = 1 + ATA_DEV(atadev->unit) + (ch->unit << 1);
3722
3723 meta->disks |= disk_number << ((1 - disk) << 2);
3724 }
3725 }
3726 switch (rdp->type) {
3727 case AR_T_JBOD:
3728 meta->type_total_disks = SIS_T_JBOD;
3729 break;
3730
3731 case AR_T_RAID0:
3732 meta->type_total_disks = SIS_T_RAID0;
3733 break;
3734
3735 case AR_T_RAID1:
3736 meta->type_total_disks = SIS_T_RAID1;
3737 break;
3738
3739 default:
d438c7c2 3740 kfree(meta, M_AR);
c1b3d7c5
TS		 3741 return ENODEV;
3742 }
3743 meta->type_total_disks |= (rdp->total_disks & SIS_D_MASK);
3744 meta->stripe_sectors = rdp->interleave;
3745 meta->timestamp = timestamp.tv_sec;
3746
3747 for (disk = 0; disk < rdp->total_disks; disk++) {
3748 if (rdp->disks[disk].dev) {
3749 struct ata_channel *ch =
3750 device_get_softc(device_get_parent(rdp->disks[disk].dev));
3751 struct ata_device *atadev = device_get_softc(rdp->disks[disk].dev);
3752
3753 meta->controller_pci_id =
3754 (pci_get_vendor(GRANDPARENT(rdp->disks[disk].dev)) << 16) |
3755 pci_get_device(GRANDPARENT(rdp->disks[disk].dev));
3756 bcopy(atadev->param.model, meta->model, sizeof(meta->model));
3757
3758 /* XXX SOS if total_disks > 2 this may not float */
3759 meta->disk_number = 1 + ATA_DEV(atadev->unit) + (ch->unit << 1);
3760
3761 if (testing || bootverbose)
3762 ata_raid_sis_print_meta(meta);
3763
3764 if (ata_raid_rw(rdp->disks[disk].dev,
3765 SIS_LBA(rdp->disks[disk].dev),
3766 meta, sizeof(struct sis_raid_conf),
3767 ATA_R_WRITE | ATA_R_DIRECT)) {
3768 device_printf(rdp->disks[disk].dev, "write metadata failed\n");
3769 error = EIO;
3770 }
3771 }
3772 }
d438c7c2 3773 kfree(meta, M_AR);
c1b3d7c5
TS		 3774 return error;
3775}
3776
3777/* VIA Tech V-RAID Metadata */
3778static int
3779ata_raid_via_read_meta(device_t dev, struct ar_softc **raidp)
3780{
3781 struct ata_raid_subdisk *ars = device_get_softc(dev);
3782 device_t parent = device_get_parent(dev);
3783 struct via_raid_conf *meta;
3784 struct ar_softc *raid = NULL;
3785 u_int8_t checksum, *ptr;
3786 int array, count, disk, retval = 0;
3787
978400d3
SW		 3788 meta = (struct via_raid_conf *)kmalloc(sizeof(struct via_raid_conf), M_AR,
3789 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3790
3791 if (ata_raid_rw(parent, VIA_LBA(parent),
3792 meta, sizeof(struct via_raid_conf), ATA_R_READ)) {
3793 if (testing || bootverbose)
3794 device_printf(parent, "VIA read metadata failed\n");
3795 goto via_out;
3796 }
3797
3798 /* check if this is a VIA RAID struct */
3799 if (meta->magic != VIA_MAGIC) {
3800 if (testing || bootverbose)
3801 device_printf(parent, "VIA check1 failed\n");
3802 goto via_out;
3803 }
3804
3805 /* calculate checksum and compare for valid */
3806 for (checksum = 0, ptr = (u_int8_t *)meta, count = 0; count < 50; count++)
3807 checksum += *ptr++;
3808 if (checksum != meta->checksum) {
3809 if (testing || bootverbose)
3810 device_printf(parent, "VIA check2 failed\n");
3811 goto via_out;
3812 }
3813
3814 if (testing || bootverbose)
3815 ata_raid_via_print_meta(meta);
3816
3817 /* now convert VIA meta into our generic form */
3818 for (array = 0; array < MAX_ARRAYS; array++) {
3819 if (!raidp[array]) {
3820 raidp[array] =
d438c7c2
TS		 3821 (struct ar_softc *)kmalloc(sizeof(struct ar_softc), M_AR,
3822 M_WAITOK | M_ZERO);
c1b3d7c5
TS		 3823 }
3824 raid = raidp[array];
3825 if (raid->format && (raid->format != AR_F_VIA_RAID))
3826 continue;
3827
3828 if (raid->format == AR_F_VIA_RAID && (raid->magic_0 != meta->disks[0]))
3829 continue;
3830
3831 switch (meta->type & VIA_T_MASK) {
3832 case VIA_T_RAID0:
3833 raid->type = AR_T_RAID0;
3834 raid->width = meta->stripe_layout & VIA_L_DISKS;
3835 if (!raid->total_sectors ||
3836 (raid->total_sectors > (raid->width * meta->disk_sectors)))
3837 raid->total_sectors = raid->width * meta->disk_sectors;
3838 break;
3839
3840 case VIA_T_RAID1:
3841 raid->type = AR_T_RAID1;
3842 raid->width = 1;
3843 raid->total_sectors = meta->disk_sectors;
3844 break;
3845
3846 case VIA_T_RAID01:
3847 raid->type = AR_T_RAID01;
3848 raid->width = meta->stripe_layout & VIA_L_DISKS;