freebsd-skq/sys/dev/ata/ata-raid.c
Søren Schmidt 9e34f8233b Properly initialize and destroy the RAID lock.
Also dont mess with RAID's thats not attached yet and avoid panic.
2006-08-25 09:33:56 +00:00

4951 lines
148 KiB
C

/*-
* Copyright (c) 2000 - 2006 Søren Schmidt <sos@FreeBSD.org>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer,
* without modification, immediately at the beginning of the file.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ata.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/ata.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/endian.h>
#include <sys/bio.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/disk.h>
#include <sys/cons.h>
#include <sys/sema.h>
#include <sys/taskqueue.h>
#include <vm/uma.h>
#include <machine/bus.h>
#include <sys/rman.h>
#include <dev/pci/pcivar.h>
#include <geom/geom_disk.h>
#include <dev/ata/ata-all.h>
#include <dev/ata/ata-disk.h>
#include <dev/ata/ata-raid.h>
#include <dev/ata/ata-pci.h>
#include <ata_if.h>
/* prototypes */
static void ata_raid_done(struct ata_request *request);
static void ata_raid_config_changed(struct ar_softc *rdp, int writeback);
static int ata_raid_status(struct ata_ioc_raid_config *config);
static int ata_raid_create(struct ata_ioc_raid_config *config);
static int ata_raid_delete(int array);
static int ata_raid_addspare(struct ata_ioc_raid_config *config);
static int ata_raid_rebuild(int array);
static int ata_raid_read_metadata(device_t subdisk);
static int ata_raid_write_metadata(struct ar_softc *rdp);
static int ata_raid_wipe_metadata(struct ar_softc *rdp);
static int ata_raid_adaptec_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_hptv2_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_hptv2_write_meta(struct ar_softc *rdp);
static int ata_raid_hptv3_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_intel_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_intel_write_meta(struct ar_softc *rdp);
static int ata_raid_ite_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_jmicron_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_jmicron_write_meta(struct ar_softc *rdp);
static int ata_raid_lsiv2_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_lsiv3_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_nvidia_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_promise_read_meta(device_t dev, struct ar_softc **raidp, int native);
static int ata_raid_promise_write_meta(struct ar_softc *rdp);
static int ata_raid_sii_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_sis_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_sis_write_meta(struct ar_softc *rdp);
static int ata_raid_via_read_meta(device_t dev, struct ar_softc **raidp);
static int ata_raid_via_write_meta(struct ar_softc *rdp);
static struct ata_request *ata_raid_init_request(struct ar_softc *rdp, struct bio *bio);
static int ata_raid_send_request(struct ata_request *request);
static int ata_raid_rw(device_t dev, u_int64_t lba, void *data, u_int bcount, int flags);
static char * ata_raid_format(struct ar_softc *rdp);
static char * ata_raid_type(struct ar_softc *rdp);
static char * ata_raid_flags(struct ar_softc *rdp);
/* debugging only */
static void ata_raid_print_meta(struct ar_softc *meta);
static void ata_raid_adaptec_print_meta(struct adaptec_raid_conf *meta);
static void ata_raid_hptv2_print_meta(struct hptv2_raid_conf *meta);
static void ata_raid_hptv3_print_meta(struct hptv3_raid_conf *meta);
static void ata_raid_intel_print_meta(struct intel_raid_conf *meta);
static void ata_raid_ite_print_meta(struct ite_raid_conf *meta);
static void ata_raid_jmicron_print_meta(struct jmicron_raid_conf *meta);
static void ata_raid_lsiv2_print_meta(struct lsiv2_raid_conf *meta);
static void ata_raid_lsiv3_print_meta(struct lsiv3_raid_conf *meta);
static void ata_raid_nvidia_print_meta(struct nvidia_raid_conf *meta);
static void ata_raid_promise_print_meta(struct promise_raid_conf *meta);
static void ata_raid_sii_print_meta(struct sii_raid_conf *meta);
static void ata_raid_sis_print_meta(struct sis_raid_conf *meta);
static void ata_raid_via_print_meta(struct via_raid_conf *meta);
/* internal vars */
static struct ar_softc *ata_raid_arrays[MAX_ARRAYS];
static MALLOC_DEFINE(M_AR, "ar_driver", "ATA PseudoRAID driver");
static devclass_t ata_raid_sub_devclass;
static int testing = 0;
/* device structures */
static disk_strategy_t ata_raid_strategy;
static dumper_t ata_raid_dump;
static void
ata_raid_attach(struct ar_softc *rdp, int writeback)
{
char buffer[32];
int disk;
mtx_init(&rdp->lock, "ATA PseudoRAID metadata lock", NULL, MTX_DEF);
ata_raid_config_changed(rdp, writeback);
/* sanitize arrays total_size % (width * interleave) == 0 */
if (rdp->type == AR_T_RAID0 || rdp->type == AR_T_RAID01 ||
rdp->type == AR_T_RAID5) {
rdp->total_sectors = (rdp->total_sectors/(rdp->interleave*rdp->width))*
(rdp->interleave * rdp->width);
sprintf(buffer, " (stripe %d KB)",
(rdp->interleave * DEV_BSIZE) / 1024);
}
else
buffer[0] = '\0';
rdp->disk = disk_alloc();
rdp->disk->d_strategy = ata_raid_strategy;
rdp->disk->d_dump = ata_raid_dump;
rdp->disk->d_name = "ar";
rdp->disk->d_sectorsize = DEV_BSIZE;
rdp->disk->d_mediasize = (off_t)rdp->total_sectors * DEV_BSIZE;
rdp->disk->d_fwsectors = rdp->sectors;
rdp->disk->d_fwheads = rdp->heads;
rdp->disk->d_maxsize = 128 * DEV_BSIZE;
rdp->disk->d_drv1 = rdp;
rdp->disk->d_unit = rdp->lun;
disk_create(rdp->disk, DISK_VERSION);
printf("ar%d: %juMB <%s %s%s> status: %s\n", rdp->lun,
rdp->total_sectors / ((1024L * 1024L) / DEV_BSIZE),
ata_raid_format(rdp), ata_raid_type(rdp),
buffer, ata_raid_flags(rdp));
if (testing || bootverbose)
printf("ar%d: %ju sectors [%dC/%dH/%dS] <%s> subdisks defined as:\n",
rdp->lun, rdp->total_sectors,
rdp->cylinders, rdp->heads, rdp->sectors, rdp->name);
for (disk = 0; disk < rdp->total_disks; disk++) {
printf("ar%d: disk%d ", rdp->lun, disk);
if (rdp->disks[disk].dev) {
if (rdp->disks[disk].flags & AR_DF_PRESENT) {
/* status of this disk in the array */
if (rdp->disks[disk].flags & AR_DF_ONLINE)
printf("READY ");
else if (rdp->disks[disk].flags & AR_DF_SPARE)
printf("SPARE ");
else
printf("FREE ");
/* what type of disk is this in the array */
switch (rdp->type) {
case AR_T_RAID1:
case AR_T_RAID01:
if (disk < rdp->width)
printf("(master) ");
else
printf("(mirror) ");
}
/* which physical disk is used */
printf("using %s at ata%d-%s\n",
device_get_nameunit(rdp->disks[disk].dev),
device_get_unit(device_get_parent(rdp->disks[disk].dev)),
(((struct ata_device *)
device_get_softc(rdp->disks[disk].dev))->unit ==
ATA_MASTER) ? "master" : "slave");
}
else if (rdp->disks[disk].flags & AR_DF_ASSIGNED)
printf("DOWN\n");
else
printf("INVALID no RAID config on this subdisk\n");
}
else
printf("DOWN no device found for this subdisk\n");
}
}
static int
ata_raid_ioctl(u_long cmd, caddr_t data)
{
struct ata_ioc_raid_config *config = (struct ata_ioc_raid_config *)data;
int *lun = (int *)data;
int error = EOPNOTSUPP;
switch (cmd) {
case IOCATARAIDSTATUS:
error = ata_raid_status(config);
break;
case IOCATARAIDCREATE:
error = ata_raid_create(config);
break;
case IOCATARAIDDELETE:
error = ata_raid_delete(*lun);
break;
case IOCATARAIDADDSPARE:
error = ata_raid_addspare(config);
break;
case IOCATARAIDREBUILD:
error = ata_raid_rebuild(*lun);
break;
}
return error;
}
static void
ata_raid_strategy(struct bio *bp)
{
struct ar_softc *rdp = bp->bio_disk->d_drv1;
struct ata_request *request;
caddr_t data;
u_int64_t blkno, lba, blk = 0;
int count, chunk, drv, par = 0, change = 0;
if (!(rdp->status & AR_S_READY) ||
(bp->bio_cmd != BIO_READ && bp->bio_cmd != BIO_WRITE)) {
biofinish(bp, NULL, EIO);
return;
}
bp->bio_resid = bp->bio_bcount;
for (count = howmany(bp->bio_bcount, DEV_BSIZE),
blkno = bp->bio_pblkno, data = bp->bio_data;
count > 0;
count -= chunk, blkno += chunk, data += (chunk * DEV_BSIZE)) {
switch (rdp->type) {
case AR_T_RAID1:
drv = 0;
lba = blkno;
chunk = count;
break;
case AR_T_JBOD:
case AR_T_SPAN:
drv = 0;
lba = blkno;
while (lba >= rdp->disks[drv].sectors)
lba -= rdp->disks[drv++].sectors;
chunk = min(rdp->disks[drv].sectors - lba, count);
break;
case AR_T_RAID0:
case AR_T_RAID01:
chunk = blkno % rdp->interleave;
drv = (blkno / rdp->interleave) % rdp->width;
lba = (((blkno/rdp->interleave)/rdp->width)*rdp->interleave)+chunk;
chunk = min(count, rdp->interleave - chunk);
break;
case AR_T_RAID5:
drv = (blkno / rdp->interleave) % (rdp->width - 1);
par = rdp->width - 1 -
(blkno / (rdp->interleave * (rdp->width - 1))) % rdp->width;
if (drv >= par)
drv++;
lba = ((blkno/rdp->interleave)/(rdp->width-1))*(rdp->interleave) +
((blkno%(rdp->interleave*(rdp->width-1)))%rdp->interleave);
chunk = min(count, rdp->interleave - (lba % rdp->interleave));
break;
default:
printf("ar%d: unknown array type in ata_raid_strategy\n", rdp->lun);
biofinish(bp, NULL, EIO);
return;
}
/* offset on all but "first on HPTv2" */
if (!(drv == 0 && rdp->format == AR_F_HPTV2_RAID))
lba += rdp->offset_sectors;
if (!(request = ata_raid_init_request(rdp, bp))) {
biofinish(bp, NULL, EIO);
return;
}
request->data = data;
request->bytecount = chunk * DEV_BSIZE;
request->u.ata.lba = lba;
request->u.ata.count = request->bytecount / DEV_BSIZE;
switch (rdp->type) {
case AR_T_JBOD:
case AR_T_SPAN:
case AR_T_RAID0:
if (((rdp->disks[drv].flags & (AR_DF_PRESENT|AR_DF_ONLINE)) ==
(AR_DF_PRESENT|AR_DF_ONLINE) && !rdp->disks[drv].dev)) {
rdp->disks[drv].flags &= ~AR_DF_ONLINE;
ata_raid_config_changed(rdp, 1);
ata_free_request(request);
biofinish(bp, NULL, EIO);
return;
}
request->this = drv;
request->dev = rdp->disks[request->this].dev;
ata_raid_send_request(request);
break;
case AR_T_RAID1:
case AR_T_RAID01:
if ((rdp->disks[drv].flags &
(AR_DF_PRESENT|AR_DF_ONLINE))==(AR_DF_PRESENT|AR_DF_ONLINE) &&
!rdp->disks[drv].dev) {
rdp->disks[drv].flags &= ~AR_DF_ONLINE;
change = 1;
}
if ((rdp->disks[drv + rdp->width].flags &
(AR_DF_PRESENT|AR_DF_ONLINE))==(AR_DF_PRESENT|AR_DF_ONLINE) &&
!rdp->disks[drv + rdp->width].dev) {
rdp->disks[drv + rdp->width].flags &= ~AR_DF_ONLINE;
change = 1;
}
if (change)
ata_raid_config_changed(rdp, 1);
if (!(rdp->status & AR_S_READY)) {
ata_free_request(request);
biofinish(bp, NULL, EIO);
return;
}
if (rdp->status & AR_S_REBUILDING)
blk = ((lba / rdp->interleave) * rdp->width) * rdp->interleave +
(rdp->interleave * (drv % rdp->width)) +
lba % rdp->interleave;;
if (bp->bio_cmd == BIO_READ) {
int src_online =
(rdp->disks[drv].flags & AR_DF_ONLINE);
int mir_online =
(rdp->disks[drv+rdp->width].flags & AR_DF_ONLINE);
/* if mirror gone or close to last access on source */
if (!mir_online ||
((src_online) &&
bp->bio_pblkno >=
(rdp->disks[drv].last_lba - AR_PROXIMITY) &&
bp->bio_pblkno <=
(rdp->disks[drv].last_lba + AR_PROXIMITY))) {
rdp->toggle = 0;
}
/* if source gone or close to last access on mirror */
else if (!src_online ||
((mir_online) &&
bp->bio_pblkno >=
(rdp->disks[drv+rdp->width].last_lba-AR_PROXIMITY) &&
bp->bio_pblkno <=
(rdp->disks[drv+rdp->width].last_lba+AR_PROXIMITY))) {
drv += rdp->width;
rdp->toggle = 1;
}
/* not close to any previous access, toggle */
else {
if (rdp->toggle)
rdp->toggle = 0;
else {
drv += rdp->width;
rdp->toggle = 1;
}
}
if ((rdp->status & AR_S_REBUILDING) &&
(blk <= rdp->rebuild_lba) &&
((blk + chunk) > rdp->rebuild_lba)) {
struct ata_composite *composite;
struct ata_request *rebuild;
int this;
/* figure out what part to rebuild */
if (drv < rdp->width)
this = drv + rdp->width;
else
this = drv - rdp->width;
/* do we have a spare to rebuild on ? */
if (rdp->disks[this].flags & AR_DF_SPARE) {
if ((composite = ata_alloc_composite())) {
if ((rebuild = ata_alloc_request())) {
rdp->rebuild_lba = blk + chunk;
bcopy(request, rebuild,
sizeof(struct ata_request));
rebuild->this = this;
rebuild->dev = rdp->disks[this].dev;
rebuild->flags &= ~ATA_R_READ;
rebuild->flags |= ATA_R_WRITE;
mtx_init(&composite->lock,
"ATA PseudoRAID rebuild lock",
NULL, MTX_DEF);
composite->residual = request->bytecount;
composite->rd_needed |= (1 << drv);
composite->wr_depend |= (1 << drv);
composite->wr_needed |= (1 << this);
composite->request[drv] = request;
composite->request[this] = rebuild;
request->composite = composite;
rebuild->composite = composite;
ata_raid_send_request(rebuild);
}
else {
ata_free_composite(composite);
printf("DOH! ata_alloc_request failed!\n");
}
}
else {
printf("DOH! ata_alloc_composite failed!\n");
}
}
else if (rdp->disks[this].flags & AR_DF_ONLINE) {
/*
* if we got here we are a chunk of a RAID01 that
* does not need a rebuild, but we need to increment
* the rebuild_lba address to get the rebuild to
* move to the next chunk correctly
*/
rdp->rebuild_lba = blk + chunk;
}
else
printf("DOH! we didn't find the rebuild part\n");
}
}
if (bp->bio_cmd == BIO_WRITE) {
if ((rdp->disks[drv+rdp->width].flags & AR_DF_ONLINE) ||
((rdp->status & AR_S_REBUILDING) &&
(rdp->disks[drv+rdp->width].flags & AR_DF_SPARE) &&
((blk < rdp->rebuild_lba) ||
((blk <= rdp->rebuild_lba) &&
((blk + chunk) > rdp->rebuild_lba))))) {
if ((rdp->disks[drv].flags & AR_DF_ONLINE) ||
((rdp->status & AR_S_REBUILDING) &&
(rdp->disks[drv].flags & AR_DF_SPARE) &&
((blk < rdp->rebuild_lba) ||
((blk <= rdp->rebuild_lba) &&
((blk + chunk) > rdp->rebuild_lba))))) {
struct ata_request *mirror;
struct ata_composite *composite;
int this = drv + rdp->width;
if ((composite = ata_alloc_composite())) {
if ((mirror = ata_alloc_request())) {
if ((blk <= rdp->rebuild_lba) &&
((blk + chunk) > rdp->rebuild_lba))
rdp->rebuild_lba = blk + chunk;
bcopy(request, mirror,
sizeof(struct ata_request));
mirror->this = this;
mirror->dev = rdp->disks[this].dev;
mtx_init(&composite->lock,
"ATA PseudoRAID mirror lock",
NULL, MTX_DEF);
composite->residual = request->bytecount;
composite->wr_needed |= (1 << drv);
composite->wr_needed |= (1 << this);
composite->request[drv] = request;
composite->request[this] = mirror;
request->composite = composite;
mirror->composite = composite;
ata_raid_send_request(mirror);
rdp->disks[this].last_lba =
bp->bio_pblkno + chunk;
}
else {
ata_free_composite(composite);
printf("DOH! ata_alloc_request failed!\n");
}
}
else {
printf("DOH! ata_alloc_composite failed!\n");
}
}
else
drv += rdp->width;
}
}
request->this = drv;
request->dev = rdp->disks[request->this].dev;
ata_raid_send_request(request);
rdp->disks[request->this].last_lba = bp->bio_pblkno + chunk;
break;
case AR_T_RAID5:
if (((rdp->disks[drv].flags & (AR_DF_PRESENT|AR_DF_ONLINE)) ==
(AR_DF_PRESENT|AR_DF_ONLINE) && !rdp->disks[drv].dev)) {
rdp->disks[drv].flags &= ~AR_DF_ONLINE;
change = 1;
}
if (((rdp->disks[par].flags & (AR_DF_PRESENT|AR_DF_ONLINE)) ==
(AR_DF_PRESENT|AR_DF_ONLINE) && !rdp->disks[par].dev)) {
rdp->disks[par].flags &= ~AR_DF_ONLINE;
change = 1;
}
if (change)
ata_raid_config_changed(rdp, 1);
if (!(rdp->status & AR_S_READY)) {
ata_free_request(request);
biofinish(bp, NULL, EIO);
return;
}
if (rdp->status & AR_S_DEGRADED) {
/* do the XOR game if possible */
}
else {
request->this = drv;
request->dev = rdp->disks[request->this].dev;
if (bp->bio_cmd == BIO_READ) {
ata_raid_send_request(request);
}
if (bp->bio_cmd == BIO_WRITE) {
ata_raid_send_request(request);
// sikre at læs-modify-skriv til hver disk er atomarisk.
// par kopi af request
// læse orgdata fra drv
// skriv nydata til drv
// læse parorgdata fra par
// skriv orgdata xor parorgdata xor nydata til par
}
}
break;
default:
printf("ar%d: unknown array type in ata_raid_strategy\n", rdp->lun);
}
}
}
static void
ata_raid_done(struct ata_request *request)
{
struct ar_softc *rdp = request->driver;
struct ata_composite *composite = NULL;
struct bio *bp = request->bio;
int i, mirror, finished = 0;
switch (rdp->type) {
case AR_T_JBOD:
case AR_T_SPAN:
case AR_T_RAID0:
if (request->result) {
rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
ata_raid_config_changed(rdp, 1);
bp->bio_error = request->result;
finished = 1;
}
else {
bp->bio_resid -= request->donecount;
if (!bp->bio_resid)
finished = 1;
}
break;
case AR_T_RAID1:
case AR_T_RAID01:
if (request->this < rdp->width)
mirror = request->this + rdp->width;
else
mirror = request->this - rdp->width;
if (request->result) {
rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
ata_raid_config_changed(rdp, 1);
}
if (rdp->status & AR_S_READY) {
u_int64_t blk = 0;
if (rdp->status & AR_S_REBUILDING)
blk = ((request->u.ata.lba / rdp->interleave) * rdp->width) *
rdp->interleave + (rdp->interleave *
(request->this % rdp->width)) +
request->u.ata.lba % rdp->interleave;
if (bp->bio_cmd == BIO_READ) {
/* is this a rebuild composite */
if ((composite = request->composite)) {
mtx_lock(&composite->lock);
/* handle the read part of a rebuild composite */
if (request->flags & ATA_R_READ) {
/* if read failed array is now broken */
if (request->result) {
rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
ata_raid_config_changed(rdp, 1);
bp->bio_error = request->result;
rdp->rebuild_lba = blk;
finished = 1;
}
/* good data, update how far we've gotten */
else {
bp->bio_resid -= request->donecount;
composite->residual -= request->donecount;
if (!composite->residual) {
if (composite->wr_done & (1 << mirror))
finished = 1;
}
}
}
/* handle the write part of a rebuild composite */
else if (request->flags & ATA_R_WRITE) {
if (composite->rd_done & (1 << mirror)) {
if (request->result) {
printf("DOH! rebuild failed\n"); /* XXX SOS */
rdp->rebuild_lba = blk;
}
if (!composite->residual)
finished = 1;
}
}
mtx_unlock(&composite->lock);
}
/* if read failed retry on the mirror */
else if (request->result) {
request->dev = rdp->disks[mirror].dev;
request->flags &= ~ATA_R_TIMEOUT;
ata_raid_send_request(request);
return;
}
/* we have good data */
else {
bp->bio_resid -= request->donecount;
if (!bp->bio_resid)
finished = 1;
}
}
else if (bp->bio_cmd == BIO_WRITE) {
/* do we have a mirror or rebuild to deal with ? */
if ((composite = request->composite)) {
mtx_lock(&composite->lock);
if (composite->wr_done & (1 << mirror)) {
if (request->result) {
if (composite->request[mirror]->result) {
printf("DOH! all disks failed and got here\n");
bp->bio_error = EIO;
}
if (rdp->status & AR_S_REBUILDING) {
rdp->rebuild_lba = blk;
printf("DOH! rebuild failed\n"); /* XXX SOS */
}
bp->bio_resid -=
composite->request[mirror]->donecount;
composite->residual -=
composite->request[mirror]->donecount;
}
else {
bp->bio_resid -= request->donecount;
composite->residual -= request->donecount;
}
if (!composite->residual)
finished = 1;
}
mtx_unlock(&composite->lock);
}
/* no mirror we are done */
else {
bp->bio_resid -= request->donecount;
if (!bp->bio_resid)
finished = 1;
}
}
}
else
biofinish(bp, NULL, request->result);
break;
case AR_T_RAID5:
if (request->result) {
rdp->disks[request->this].flags &= ~AR_DF_ONLINE;
ata_raid_config_changed(rdp, 1);
if (rdp->status & AR_S_READY) {
if (bp->bio_cmd == BIO_READ) {
/* do the XOR game to recover data */
}
if (bp->bio_cmd == BIO_WRITE) {
/* if the parity failed we're OK sortof */
/* otherwise wee need to do the XOR long dance */
}
finished = 1;
}
else
biofinish(bp, NULL, request->result);
}
else {
// did we have an XOR game going ??
bp->bio_resid -= request->donecount;
if (!bp->bio_resid)
finished = 1;
}
break;
default:
printf("ar%d: unknown array type in ata_raid_done\n", rdp->lun);
}
if (finished) {
if ((rdp->status & AR_S_REBUILDING) &&
rdp->rebuild_lba >= rdp->total_sectors) {
int disk;
for (disk = 0; disk < rdp->total_disks; disk++) {
if ((rdp->disks[disk].flags &
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE)) ==
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE)) {
rdp->disks[disk].flags &= ~AR_DF_SPARE;
rdp->disks[disk].flags |= AR_DF_ONLINE;
}
}
rdp->status &= ~AR_S_REBUILDING;
ata_raid_config_changed(rdp, 1);
}
if (!bp->bio_resid)
biodone(bp);
}
if (composite) {
if (finished) {
/* we are done with this composite, free all resources */
for (i = 0; i < 32; i++) {
if (composite->rd_needed & (1 << i) ||
composite->wr_needed & (1 << i)) {
ata_free_request(composite->request[i]);
}
}
mtx_destroy(&composite->lock);
ata_free_composite(composite);
}
}
else
ata_free_request(request);
}
static int
ata_raid_dump(void *arg, void *virtual, vm_offset_t physical,
off_t offset, size_t length)
{
struct disk *dp = arg;
struct ar_softc *rdp = dp->d_drv1;
struct bio bp;
/* length zero is special and really means flush buffers to media */
if (!length) {
int disk, error;
for (disk = 0, error = 0; disk < rdp->total_disks; disk++)
if (rdp->disks[disk].dev)
error |= ata_controlcmd(rdp->disks[disk].dev,
ATA_FLUSHCACHE, 0, 0, 0);
return (error ? EIO : 0);
}
bzero(&bp, sizeof(struct bio));
bp.bio_disk = dp;
bp.bio_pblkno = offset / DEV_BSIZE;
bp.bio_bcount = length;
bp.bio_data = virtual;
bp.bio_cmd = BIO_WRITE;
ata_raid_strategy(&bp);
return bp.bio_error;
}
static void
ata_raid_config_changed(struct ar_softc *rdp, int writeback)
{
int disk, count, status;
mtx_lock(&rdp->lock);
/* set default all working mode */
status = rdp->status;
rdp->status &= ~AR_S_DEGRADED;
rdp->status |= AR_S_READY;
/* make sure all lost drives are accounted for */
for (disk = 0; disk < rdp->total_disks; disk++) {
if (!(rdp->disks[disk].flags & AR_DF_PRESENT))
rdp->disks[disk].flags &= ~AR_DF_ONLINE;
}
/* depending on RAID type figure out our health status */
switch (rdp->type) {
case AR_T_JBOD:
case AR_T_SPAN:
case AR_T_RAID0:
for (disk = 0; disk < rdp->total_disks; disk++)
if (!(rdp->disks[disk].flags & AR_DF_ONLINE))
rdp->status &= ~AR_S_READY;
break;
case AR_T_RAID1:
case AR_T_RAID01:
for (disk = 0; disk < rdp->width; disk++) {
if (!(rdp->disks[disk].flags & AR_DF_ONLINE) &&
!(rdp->disks[disk + rdp->width].flags & AR_DF_ONLINE)) {
rdp->status &= ~AR_S_READY;
}
else if (((rdp->disks[disk].flags & AR_DF_ONLINE) &&
!(rdp->disks[disk + rdp->width].flags & AR_DF_ONLINE)) ||
(!(rdp->disks[disk].flags & AR_DF_ONLINE) &&
(rdp->disks [disk + rdp->width].flags & AR_DF_ONLINE))) {
rdp->status |= AR_S_DEGRADED;
}
}
break;
case AR_T_RAID5:
for (count = 0, disk = 0; disk < rdp->total_disks; disk++) {
if (!(rdp->disks[disk].flags & AR_DF_ONLINE))
count++;
}
if (count) {
if (count > 1)
rdp->status &= ~AR_S_READY;
else
rdp->status |= AR_S_DEGRADED;
}
break;
default:
rdp->status &= ~AR_S_READY;
}
if (rdp->status != status) {
if (!(rdp->status & AR_S_READY)) {
printf("ar%d: FAILURE - %s array broken\n",
rdp->lun, ata_raid_type(rdp));
}
else if (rdp->status & AR_S_DEGRADED) {
if (rdp->type & (AR_T_RAID1 | AR_T_RAID01))
printf("ar%d: WARNING - mirror", rdp->lun);
else
printf("ar%d: WARNING - parity", rdp->lun);
printf(" protection lost. %s array in DEGRADED mode\n",
ata_raid_type(rdp));
}
}
mtx_unlock(&rdp->lock);
if (writeback)
ata_raid_write_metadata(rdp);
}
static int
ata_raid_status(struct ata_ioc_raid_config *config)
{
struct ar_softc *rdp;
int i;
if (!(rdp = ata_raid_arrays[config->lun]))
return ENXIO;
config->type = rdp->type;
config->total_disks = rdp->total_disks;
for (i = 0; i < rdp->total_disks; i++ ) {
if ((rdp->disks[i].flags & AR_DF_PRESENT) && rdp->disks[i].dev)
config->disks[i] = device_get_unit(rdp->disks[i].dev);
else
config->disks[i] = -1;
}
config->interleave = rdp->interleave;
config->status = rdp->status;
config->progress = 100 * rdp->rebuild_lba / rdp->total_sectors;
return 0;
}
static int
ata_raid_create(struct ata_ioc_raid_config *config)
{
struct ar_softc *rdp;
device_t subdisk;
int array, disk;
int ctlr = 0, disk_size = 0, total_disks = 0;
for (array = 0; array < MAX_ARRAYS; array++) {
if (!ata_raid_arrays[array])
break;
}
if (array >= MAX_ARRAYS)
return ENOSPC;
if (!(rdp = (struct ar_softc*)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO))) {
printf("ar%d: no memory for metadata storage\n", array);
return ENOMEM;
}
for (disk = 0; disk < config->total_disks; disk++) {
if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
config->disks[disk]))) {
struct ata_raid_subdisk *ars = device_get_softc(subdisk);
/* is device already assigned to another array ? */
if (ars->raid[rdp->volume]) {
config->disks[disk] = -1;
free(rdp, M_AR);
return EBUSY;
}
rdp->disks[disk].dev = device_get_parent(subdisk);
switch (pci_get_vendor(GRANDPARENT(rdp->disks[disk].dev))) {
case ATA_HIGHPOINT_ID:
/*
* we need some way to decide if it should be v2 or v3
* for now just use v2 since the v3 BIOS knows how to
* handle that as well.
*/
ctlr = AR_F_HPTV2_RAID;
rdp->disks[disk].sectors = HPTV3_LBA(rdp->disks[disk].dev);
break;
case ATA_INTEL_ID:
ctlr = AR_F_INTEL_RAID;
rdp->disks[disk].sectors = INTEL_LBA(rdp->disks[disk].dev);
break;
case ATA_ITE_ID:
ctlr = AR_F_ITE_RAID;
rdp->disks[disk].sectors = ITE_LBA(rdp->disks[disk].dev);
break;
case ATA_JMICRON_ID:
ctlr = AR_F_JMICRON_RAID;
rdp->disks[disk].sectors = JMICRON_LBA(rdp->disks[disk].dev);
break;
case 0: /* XXX SOS cover up for bug in our PCI code */
case ATA_PROMISE_ID:
ctlr = AR_F_PROMISE_RAID;
rdp->disks[disk].sectors = PROMISE_LBA(rdp->disks[disk].dev);
break;
case ATA_SIS_ID:
ctlr = AR_F_SIS_RAID;
rdp->disks[disk].sectors = SIS_LBA(rdp->disks[disk].dev);
break;
case ATA_ATI_ID:
case ATA_VIA_ID:
ctlr = AR_F_VIA_RAID;
rdp->disks[disk].sectors = VIA_LBA(rdp->disks[disk].dev);
break;
default:
/* XXX SOS
* right, so here we are, we have an ATA chip and we want
* to create a RAID and store the metadata.
* we need to find a way to tell what kind of metadata this
* hardware's BIOS might be using (good ideas are welcomed)
* for now we just use our own native FreeBSD format.
* the only way to get support for the BIOS format is to
* setup the RAID from there, in that case we pickup the
* metadata format from the disks (if we support it).
*/
printf("WARNING!! - not able to determine metadata format\n"
"WARNING!! - Using FreeBSD PseudoRAID metadata\n"
"If that is not what you want, use the BIOS to "
"create the array\n");
ctlr = AR_F_FREEBSD_RAID;
rdp->disks[disk].sectors = PROMISE_LBA(rdp->disks[disk].dev);
break;
}
/* we need all disks to be of the same format */
if ((rdp->format & AR_F_FORMAT_MASK) &&
(rdp->format & AR_F_FORMAT_MASK) != (ctlr & AR_F_FORMAT_MASK)) {
free(rdp, M_AR);
return EXDEV;
}
else
rdp->format = ctlr;
/* use the smallest disk of the lots size */
/* gigabyte boundry ??? XXX SOS */
if (disk_size)
disk_size = min(rdp->disks[disk].sectors, disk_size);
else
disk_size = rdp->disks[disk].sectors;
rdp->disks[disk].flags =
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
total_disks++;
}
else {
config->disks[disk] = -1;
free(rdp, M_AR);
return ENXIO;
}
}
if (total_disks != config->total_disks) {
free(rdp, M_AR);
return ENODEV;
}
switch (config->type) {
case AR_T_JBOD:
case AR_T_SPAN:
case AR_T_RAID0:
break;
case AR_T_RAID1:
if (total_disks != 2) {
free(rdp, M_AR);
return EPERM;
}
break;
case AR_T_RAID01:
if (total_disks % 2 != 0) {
free(rdp, M_AR);
return EPERM;
}
break;
case AR_T_RAID5:
if (total_disks < 3) {
free(rdp, M_AR);
return EPERM;
}
break;
default:
free(rdp, M_AR);
return EOPNOTSUPP;
}
rdp->type = config->type;
rdp->lun = array;
if (rdp->type == AR_T_RAID0 || rdp->type == AR_T_RAID01 ||
rdp->type == AR_T_RAID5) {
int bit = 0;
while (config->interleave >>= 1)
bit++;
rdp->interleave = 1 << bit;
}
rdp->offset_sectors = 0;
/* values that depend on metadata format */
switch (rdp->format) {
case AR_F_ADAPTEC_RAID:
rdp->interleave = min(max(32, rdp->interleave), 128); /*+*/
break;
case AR_F_HPTV2_RAID:
rdp->interleave = min(max(8, rdp->interleave), 128); /*+*/
rdp->offset_sectors = HPTV2_LBA(x) + 1;
break;
case AR_F_HPTV3_RAID:
rdp->interleave = min(max(32, rdp->interleave), 4096); /*+*/
break;
case AR_F_INTEL_RAID:
rdp->interleave = min(max(8, rdp->interleave), 256); /*+*/
break;
case AR_F_ITE_RAID:
rdp->interleave = min(max(2, rdp->interleave), 128); /*+*/
break;
case AR_F_JMICRON_RAID:
rdp->interleave = min(max(8, rdp->interleave), 256); /*+*/
break;
case AR_F_LSIV2_RAID:
rdp->interleave = min(max(2, rdp->interleave), 4096);
break;
case AR_F_LSIV3_RAID:
rdp->interleave = min(max(2, rdp->interleave), 256);
break;
case AR_F_PROMISE_RAID:
rdp->interleave = min(max(2, rdp->interleave), 2048); /*+*/
break;
case AR_F_SII_RAID:
rdp->interleave = min(max(8, rdp->interleave), 256); /*+*/
break;
case AR_F_SIS_RAID:
rdp->interleave = min(max(32, rdp->interleave), 512); /*+*/
break;
case AR_F_VIA_RAID:
rdp->interleave = min(max(8, rdp->interleave), 128); /*+*/
break;
}
rdp->total_disks = total_disks;
rdp->width = total_disks / (rdp->type & (AR_RAID1 | AR_T_RAID01) ? 2 : 1);
rdp->total_sectors = disk_size * (rdp->width - (rdp->type == AR_RAID5));
rdp->heads = 255;
rdp->sectors = 63;
rdp->cylinders = rdp->total_sectors / (255 * 63);
rdp->rebuild_lba = 0;
rdp->status |= AR_S_READY;
/* we are committed to this array, grap the subdisks */
for (disk = 0; disk < config->total_disks; disk++) {
if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
config->disks[disk]))) {
struct ata_raid_subdisk *ars = device_get_softc(subdisk);
ars->raid[rdp->volume] = rdp;
ars->disk_number[rdp->volume] = disk;
}
}
ata_raid_attach(rdp, 1);
ata_raid_arrays[array] = rdp;
config->lun = array;
return 0;
}
static int
ata_raid_delete(int array)
{
struct ar_softc *rdp;
device_t subdisk;
int disk;
if (!(rdp = ata_raid_arrays[array]))
return ENXIO;
rdp->status &= ~AR_S_READY;
if (rdp->disk)
disk_destroy(rdp->disk);
for (disk = 0; disk < rdp->total_disks; disk++) {
if ((rdp->disks[disk].flags & AR_DF_PRESENT) && rdp->disks[disk].dev) {
if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
device_get_unit(rdp->disks[disk].dev)))) {
struct ata_raid_subdisk *ars = device_get_softc(subdisk);
if (ars->raid[rdp->volume] != rdp) /* XXX SOS */
device_printf(subdisk, "DOH! this disk doesn't belong\n");
if (ars->disk_number[rdp->volume] != disk) /* XXX SOS */
device_printf(subdisk, "DOH! this disk number is wrong\n");
ars->raid[rdp->volume] = NULL;
ars->disk_number[rdp->volume] = -1;
}
rdp->disks[disk].flags = 0;
}
}
ata_raid_wipe_metadata(rdp);
ata_raid_arrays[array] = NULL;
free(rdp, M_AR);
return 0;
}
static int
ata_raid_addspare(struct ata_ioc_raid_config *config)
{
struct ar_softc *rdp;
device_t subdisk;
int disk;
if (!(rdp = ata_raid_arrays[config->lun]))
return ENXIO;
if (!(rdp->status & AR_S_DEGRADED) || !(rdp->status & AR_S_READY))
return ENXIO;
if (rdp->status & AR_S_REBUILDING)
return EBUSY;
switch (rdp->type) {
case AR_T_RAID1:
case AR_T_RAID01:
case AR_T_RAID5:
for (disk = 0; disk < rdp->total_disks; disk++ ) {
if (((rdp->disks[disk].flags & (AR_DF_PRESENT | AR_DF_ONLINE)) ==
(AR_DF_PRESENT | AR_DF_ONLINE)) && rdp->disks[disk].dev)
continue;
if ((subdisk = devclass_get_device(ata_raid_sub_devclass,
config->disks[0] ))) {
struct ata_raid_subdisk *ars = device_get_softc(subdisk);
if (ars->raid[rdp->volume])
return EBUSY;
/* XXX SOS validate size etc etc */
ars->raid[rdp->volume] = rdp;
ars->disk_number[rdp->volume] = disk;
rdp->disks[disk].dev = device_get_parent(subdisk);
rdp->disks[disk].flags =
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE);
device_printf(rdp->disks[disk].dev,
"inserted into ar%d disk%d as spare\n",
rdp->lun, disk);
ata_raid_config_changed(rdp, 1);
return 0;
}
}
return ENXIO;
default:
return EPERM;
}
}
static int
ata_raid_rebuild(int array)
{
struct ar_softc *rdp;
int disk, count;
if (!(rdp = ata_raid_arrays[array]))
return ENXIO;
/* XXX SOS we should lock the rdp softc here */
if (!(rdp->status & AR_S_DEGRADED) || !(rdp->status & AR_S_READY))
return ENXIO;
if (rdp->status & AR_S_REBUILDING)
return EBUSY;
switch (rdp->type) {
case AR_T_RAID1:
case AR_T_RAID01:
case AR_T_RAID5:
for (count = 0, disk = 0; disk < rdp->total_disks; disk++ ) {
if (((rdp->disks[disk].flags &
(AR_DF_PRESENT|AR_DF_ASSIGNED|AR_DF_ONLINE|AR_DF_SPARE)) ==
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_SPARE)) &&
rdp->disks[disk].dev) {
count++;
}
}
if (count) {
rdp->rebuild_lba = 0;
rdp->status |= AR_S_REBUILDING;
return 0;
}
return EIO;
default:
return EPERM;
}
}
static int
ata_raid_read_metadata(device_t subdisk)
{
devclass_t pci_devclass = devclass_find("pci");
devclass_t devclass=device_get_devclass(GRANDPARENT(GRANDPARENT(subdisk)));
/* prioritize vendor native metadata layout if possible */
if (devclass == pci_devclass) {
switch (pci_get_vendor(GRANDPARENT(device_get_parent(subdisk)))) {
case ATA_HIGHPOINT_ID:
if (ata_raid_hptv3_read_meta(subdisk, ata_raid_arrays))
return 0;
if (ata_raid_hptv2_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case ATA_INTEL_ID:
if (ata_raid_intel_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case ATA_ITE_ID:
if (ata_raid_ite_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case ATA_JMICRON_ID:
if (ata_raid_jmicron_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case ATA_NVIDIA_ID:
if (ata_raid_nvidia_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case 0: /* XXX SOS cover up for bug in our PCI code */
case ATA_PROMISE_ID:
if (ata_raid_promise_read_meta(subdisk, ata_raid_arrays, 0))
return 0;
break;
case ATA_ATI_ID:
case ATA_SILICON_IMAGE_ID:
if (ata_raid_sii_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case ATA_SIS_ID:
if (ata_raid_sis_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
case ATA_VIA_ID:
if (ata_raid_via_read_meta(subdisk, ata_raid_arrays))
return 0;
break;
}
}
/* handle controllers that have multiple layout possibilities */
/* NOTE: the order of these are not insignificant */
/* Adaptec HostRAID */
if (ata_raid_adaptec_read_meta(subdisk, ata_raid_arrays))
return 0;
/* LSILogic v3 and v2 */
if (ata_raid_lsiv3_read_meta(subdisk, ata_raid_arrays))
return 0;
if (ata_raid_lsiv2_read_meta(subdisk, ata_raid_arrays))
return 0;
/* if none of the above matched, try FreeBSD native format */
return ata_raid_promise_read_meta(subdisk, ata_raid_arrays, 1);
}
static int
ata_raid_write_metadata(struct ar_softc *rdp)
{
switch (rdp->format) {
case AR_F_FREEBSD_RAID:
case AR_F_PROMISE_RAID:
return ata_raid_promise_write_meta(rdp);
case AR_F_HPTV3_RAID:
case AR_F_HPTV2_RAID:
/*
* always write HPT v2 metadata, the v3 BIOS knows it as well.
* this is handy since we cannot know what version BIOS is on there
*/
return ata_raid_hptv2_write_meta(rdp);
case AR_F_INTEL_RAID:
return ata_raid_intel_write_meta(rdp);
case AR_F_JMICRON_RAID:
return ata_raid_jmicron_write_meta(rdp);
case AR_F_SIS_RAID:
return ata_raid_sis_write_meta(rdp);
case AR_F_VIA_RAID:
return ata_raid_via_write_meta(rdp);
#if 0
case AR_F_HPTV3_RAID:
return ata_raid_hptv3_write_meta(rdp);
case AR_F_ADAPTEC_RAID:
return ata_raid_adaptec_write_meta(rdp);
case AR_F_ITE_RAID:
return ata_raid_ite_write_meta(rdp);
case AR_F_LSIV2_RAID:
return ata_raid_lsiv2_write_meta(rdp);
case AR_F_LSIV3_RAID:
return ata_raid_lsiv3_write_meta(rdp);
case AR_F_NVIDIA_RAID:
return ata_raid_nvidia_write_meta(rdp);
case AR_F_SII_RAID:
return ata_raid_sii_write_meta(rdp);
#endif
default:
printf("ar%d: writing of %s metadata is NOT supported yet\n",
rdp->lun, ata_raid_format(rdp));
}
return -1;
}
static int
ata_raid_wipe_metadata(struct ar_softc *rdp)
{
int disk, error = 0;
u_int64_t lba;
u_int32_t size;
u_int8_t *meta;
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
switch (rdp->format) {
case AR_F_ADAPTEC_RAID:
lba = ADP_LBA(rdp->disks[disk].dev);
size = sizeof(struct adaptec_raid_conf);
break;
case AR_F_HPTV2_RAID:
lba = HPTV2_LBA(rdp->disks[disk].dev);
size = sizeof(struct hptv2_raid_conf);
break;
case AR_F_HPTV3_RAID:
lba = HPTV3_LBA(rdp->disks[disk].dev);
size = sizeof(struct hptv3_raid_conf);
break;
case AR_F_INTEL_RAID:
lba = INTEL_LBA(rdp->disks[disk].dev);
size = 3 * 512; /* XXX SOS */
break;
case AR_F_ITE_RAID:
lba = ITE_LBA(rdp->disks[disk].dev);
size = sizeof(struct ite_raid_conf);
break;
case AR_F_JMICRON_RAID:
lba = JMICRON_LBA(rdp->disks[disk].dev);
size = sizeof(struct jmicron_raid_conf);
break;
case AR_F_LSIV2_RAID:
lba = LSIV2_LBA(rdp->disks[disk].dev);
size = sizeof(struct lsiv2_raid_conf);
break;
case AR_F_LSIV3_RAID:
lba = LSIV3_LBA(rdp->disks[disk].dev);
size = sizeof(struct lsiv3_raid_conf);
break;
case AR_F_NVIDIA_RAID:
lba = NVIDIA_LBA(rdp->disks[disk].dev);
size = sizeof(struct nvidia_raid_conf);
break;
case AR_F_FREEBSD_RAID:
case AR_F_PROMISE_RAID:
lba = PROMISE_LBA(rdp->disks[disk].dev);
size = sizeof(struct promise_raid_conf);
break;
case AR_F_SII_RAID:
lba = SII_LBA(rdp->disks[disk].dev);
size = sizeof(struct sii_raid_conf);
break;
case AR_F_SIS_RAID:
lba = SIS_LBA(rdp->disks[disk].dev);
size = sizeof(struct sis_raid_conf);
break;
case AR_F_VIA_RAID:
lba = VIA_LBA(rdp->disks[disk].dev);
size = sizeof(struct via_raid_conf);
break;
default:
printf("ar%d: wiping of %s metadata is NOT supported yet\n",
rdp->lun, ata_raid_format(rdp));
return ENXIO;
}
if (!(meta = malloc(size, M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(rdp->disks[disk].dev, lba, meta, size,
ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "wipe metadata failed\n");
error = EIO;
}
free(meta, M_AR);
}
}
return error;
}
/* Adaptec HostRAID Metadata */
static int
ata_raid_adaptec_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct adaptec_raid_conf *meta;
struct ar_softc *raid;
int array, disk, retval = 0;
if (!(meta = (struct adaptec_raid_conf *)
malloc(sizeof(struct adaptec_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, ADP_LBA(parent),
meta, sizeof(struct adaptec_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "Adaptec read metadata failed\n");
goto adaptec_out;
}
/* check if this is a Adaptec RAID struct */
if (meta->magic_0 != ADP_MAGIC_0 || meta->magic_3 != ADP_MAGIC_3) {
if (testing || bootverbose)
device_printf(parent, "Adaptec check1 failed\n");
goto adaptec_out;
}
if (testing || bootverbose)
ata_raid_adaptec_print_meta(meta);
/* now convert Adaptec metadata into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto adaptec_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_ADAPTEC_RAID))
continue;
if (raid->magic_0 && raid->magic_0 != meta->configs[0].magic_0)
continue;
if (!meta->generation || be32toh(meta->generation) > raid->generation) {
switch (meta->configs[0].type) {
case ADP_T_RAID0:
raid->magic_0 = meta->configs[0].magic_0;
raid->type = AR_T_RAID0;
raid->interleave = 1 << (meta->configs[0].stripe_shift >> 1);
raid->width = be16toh(meta->configs[0].total_disks);
break;
case ADP_T_RAID1:
raid->magic_0 = meta->configs[0].magic_0;
raid->type = AR_T_RAID1;
raid->width = be16toh(meta->configs[0].total_disks) / 2;
break;
default:
device_printf(parent, "Adaptec unknown RAID type 0x%02x\n",
meta->configs[0].type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto adaptec_out;
}
raid->format = AR_F_ADAPTEC_RAID;
raid->generation = be32toh(meta->generation);
raid->total_disks = be16toh(meta->configs[0].total_disks);
raid->total_sectors = be32toh(meta->configs[0].sectors);
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = 0;
raid->lun = array;
strncpy(raid->name, meta->configs[0].name,
min(sizeof(raid->name), sizeof(meta->configs[0].name)));
/* clear out any old info */
if (raid->generation) {
for (disk = 0; disk < raid->total_disks; disk++) {
raid->disks[disk].dev = NULL;
raid->disks[disk].flags = 0;
}
}
}
if (be32toh(meta->generation) >= raid->generation) {
struct ata_device *atadev = device_get_softc(parent);
struct ata_channel *ch = device_get_softc(GRANDPARENT(dev));
int disk_number = (ch->unit << !(ch->flags & ATA_NO_SLAVE)) +
ATA_DEV(atadev->unit);
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].sectors =
be32toh(meta->configs[disk_number + 1].sectors);
raid->disks[disk_number].flags =
(AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
}
break;
}
adaptec_out:
free(meta, M_AR);
return retval;
}
/* Highpoint V2 RocketRAID Metadata */
static int
ata_raid_hptv2_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct hptv2_raid_conf *meta;
struct ar_softc *raid = NULL;
int array, disk_number = 0, retval = 0;
if (!(meta = (struct hptv2_raid_conf *)
malloc(sizeof(struct hptv2_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, HPTV2_LBA(parent),
meta, sizeof(struct hptv2_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "HighPoint (v2) read metadata failed\n");
goto hptv2_out;
}
/* check if this is a HighPoint v2 RAID struct */
if (meta->magic != HPTV2_MAGIC_OK && meta->magic != HPTV2_MAGIC_BAD) {
if (testing || bootverbose)
device_printf(parent, "HighPoint (v2) check1 failed\n");
goto hptv2_out;
}
/* is this disk defined, or an old leftover/spare ? */
if (!meta->magic_0) {
if (testing || bootverbose)
device_printf(parent, "HighPoint (v2) check2 failed\n");
goto hptv2_out;
}
if (testing || bootverbose)
ata_raid_hptv2_print_meta(meta);
/* now convert HighPoint (v2) metadata into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto hptv2_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_HPTV2_RAID))
continue;
switch (meta->type) {
case HPTV2_T_RAID0:
if ((meta->order & (HPTV2_O_RAID0|HPTV2_O_OK)) ==
(HPTV2_O_RAID0|HPTV2_O_OK))
goto highpoint_raid1;
if (meta->order & (HPTV2_O_RAID0 | HPTV2_O_RAID1))
goto highpoint_raid01;
if (raid->magic_0 && raid->magic_0 != meta->magic_0)
continue;
raid->magic_0 = meta->magic_0;
raid->type = AR_T_RAID0;
raid->interleave = 1 << meta->stripe_shift;
disk_number = meta->disk_number;
if (!(meta->order & HPTV2_O_OK))
meta->magic = 0; /* mark bad */
break;
case HPTV2_T_RAID1:
highpoint_raid1:
if (raid->magic_0 && raid->magic_0 != meta->magic_0)
continue;
raid->magic_0 = meta->magic_0;
raid->type = AR_T_RAID1;
disk_number = (meta->disk_number > 0);
break;
case HPTV2_T_RAID01_RAID0:
highpoint_raid01:
if (meta->order & HPTV2_O_RAID0) {
if ((raid->magic_0 && raid->magic_0 != meta->magic_0) ||
(raid->magic_1 && raid->magic_1 != meta->magic_1))
continue;
raid->magic_0 = meta->magic_0;
raid->magic_1 = meta->magic_1;
raid->type = AR_T_RAID01;
raid->interleave = 1 << meta->stripe_shift;
disk_number = meta->disk_number;
}
else {
if (raid->magic_1 && raid->magic_1 != meta->magic_1)
continue;
raid->magic_1 = meta->magic_1;
raid->type = AR_T_RAID01;
raid->interleave = 1 << meta->stripe_shift;
disk_number = meta->disk_number + meta->array_width;
if (!(meta->order & HPTV2_O_RAID1))
meta->magic = 0; /* mark bad */
}
break;
case HPTV2_T_SPAN:
if (raid->magic_0 && raid->magic_0 != meta->magic_0)
continue;
raid->magic_0 = meta->magic_0;
raid->type = AR_T_SPAN;
disk_number = meta->disk_number;
break;
default:
device_printf(parent, "Highpoint (v2) unknown RAID type 0x%02x\n",
meta->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto hptv2_out;
}
raid->format |= AR_F_HPTV2_RAID;
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].flags = (AR_DF_PRESENT | AR_DF_ASSIGNED);
raid->lun = array;
strncpy(raid->name, meta->name_1,
min(sizeof(raid->name), sizeof(meta->name_1)));
if (meta->magic == HPTV2_MAGIC_OK) {
raid->disks[disk_number].flags |= AR_DF_ONLINE;
raid->width = meta->array_width;
raid->total_sectors = meta->total_sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = HPTV2_LBA(parent) + 1;
raid->rebuild_lba = meta->rebuild_lba;
raid->disks[disk_number].sectors =
raid->total_sectors / raid->width;
}
else
raid->disks[disk_number].flags &= ~AR_DF_ONLINE;
if ((raid->type & AR_T_RAID0) && (raid->total_disks < raid->width))
raid->total_disks = raid->width;
if (disk_number >= raid->total_disks)
raid->total_disks = disk_number + 1;
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
break;
}
hptv2_out:
free(meta, M_AR);
return retval;
}
static int
ata_raid_hptv2_write_meta(struct ar_softc *rdp)
{
struct hptv2_raid_conf *meta;
struct timeval timestamp;
int disk, error = 0;
if (!(meta = (struct hptv2_raid_conf *)
malloc(sizeof(struct hptv2_raid_conf), M_AR, M_NOWAIT | M_ZERO))) {
printf("ar%d: failed to allocate metadata storage\n", rdp->lun);
return ENOMEM;
}
microtime(&timestamp);
rdp->magic_0 = timestamp.tv_sec + 2;
rdp->magic_1 = timestamp.tv_sec;
for (disk = 0; disk < rdp->total_disks; disk++) {
if ((rdp->disks[disk].flags & (AR_DF_PRESENT | AR_DF_ONLINE)) ==
(AR_DF_PRESENT | AR_DF_ONLINE))
meta->magic = HPTV2_MAGIC_OK;
if (rdp->disks[disk].flags & AR_DF_ASSIGNED) {
meta->magic_0 = rdp->magic_0;
if (strlen(rdp->name))
strncpy(meta->name_1, rdp->name, sizeof(meta->name_1));
else
strcpy(meta->name_1, "FreeBSD");
}
meta->disk_number = disk;
switch (rdp->type) {
case AR_T_RAID0:
meta->type = HPTV2_T_RAID0;
strcpy(meta->name_2, "RAID 0");
if (rdp->disks[disk].flags & AR_DF_ONLINE)
meta->order = HPTV2_O_OK;
break;
case AR_T_RAID1:
meta->type = HPTV2_T_RAID0;
strcpy(meta->name_2, "RAID 1");
meta->disk_number = (disk < rdp->width) ? disk : disk + 5;
meta->order = HPTV2_O_RAID0 | HPTV2_O_OK;
break;
case AR_T_RAID01:
meta->type = HPTV2_T_RAID01_RAID0;
strcpy(meta->name_2, "RAID 0+1");
if (rdp->disks[disk].flags & AR_DF_ONLINE) {
if (disk < rdp->width) {
meta->order = (HPTV2_O_RAID0 | HPTV2_O_RAID1);
meta->magic_0 = rdp->magic_0 - 1;
}
else {
meta->order = HPTV2_O_RAID1;
meta->disk_number -= rdp->width;
}
}
else
meta->magic_0 = rdp->magic_0 - 1;
meta->magic_1 = rdp->magic_1;
break;
case AR_T_SPAN:
meta->type = HPTV2_T_SPAN;
strcpy(meta->name_2, "SPAN");
break;
default:
free(meta, M_AR);
return ENODEV;
}
meta->array_width = rdp->width;
meta->stripe_shift = (rdp->width > 1) ? (ffs(rdp->interleave)-1) : 0;
meta->total_sectors = rdp->total_sectors;
meta->rebuild_lba = rdp->rebuild_lba;
if (testing || bootverbose)
ata_raid_hptv2_print_meta(meta);
if (rdp->disks[disk].dev) {
if (ata_raid_rw(rdp->disks[disk].dev,
HPTV2_LBA(rdp->disks[disk].dev), meta,
sizeof(struct promise_raid_conf),
ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "write metadata failed\n");
error = EIO;
}
}
}
free(meta, M_AR);
return error;
}
/* Highpoint V3 RocketRAID Metadata */
static int
ata_raid_hptv3_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct hptv3_raid_conf *meta;
struct ar_softc *raid = NULL;
int array, disk_number, retval = 0;
if (!(meta = (struct hptv3_raid_conf *)
malloc(sizeof(struct hptv3_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, HPTV3_LBA(parent),
meta, sizeof(struct hptv3_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "HighPoint (v3) read metadata failed\n");
goto hptv3_out;
}
/* check if this is a HighPoint v3 RAID struct */
if (meta->magic != HPTV3_MAGIC) {
if (testing || bootverbose)
device_printf(parent, "HighPoint (v3) check1 failed\n");
goto hptv3_out;
}
/* check if there are any config_entries */
if (meta->config_entries < 1) {
if (testing || bootverbose)
device_printf(parent, "HighPoint (v3) check2 failed\n");
goto hptv3_out;
}
if (testing || bootverbose)
ata_raid_hptv3_print_meta(meta);
/* now convert HighPoint (v3) metadata into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto hptv3_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_HPTV3_RAID))
continue;
if ((raid->format & AR_F_HPTV3_RAID) && raid->magic_0 != meta->magic_0)
continue;
switch (meta->configs[0].type) {
case HPTV3_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = meta->configs[0].total_disks;
disk_number = meta->configs[0].disk_number;
break;
case HPTV3_T_RAID1:
raid->type = AR_T_RAID1;
raid->width = meta->configs[0].total_disks / 2;
disk_number = meta->configs[0].disk_number;
break;
case HPTV3_T_RAID5:
raid->type = AR_T_RAID5;
raid->width = meta->configs[0].total_disks;
disk_number = meta->configs[0].disk_number;
break;
case HPTV3_T_SPAN:
raid->type = AR_T_SPAN;
raid->width = meta->configs[0].total_disks;
disk_number = meta->configs[0].disk_number;
break;
default:
device_printf(parent, "Highpoint (v3) unknown RAID type 0x%02x\n",
meta->configs[0].type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto hptv3_out;
}
if (meta->config_entries == 2) {
switch (meta->configs[1].type) {
case HPTV3_T_RAID1:
if (raid->type == AR_T_RAID0) {
raid->type = AR_T_RAID01;
disk_number = meta->configs[1].disk_number +
(meta->configs[0].disk_number << 1);
break;
}
default:
device_printf(parent, "Highpoint (v3) unknown level 2 0x%02x\n",
meta->configs[1].type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto hptv3_out;
}
}
raid->magic_0 = meta->magic_0;
raid->format = AR_F_HPTV3_RAID;
raid->generation = meta->timestamp;
raid->interleave = 1 << meta->configs[0].stripe_shift;
raid->total_disks = meta->configs[0].total_disks +
meta->configs[1].total_disks;
raid->total_sectors = meta->configs[0].total_sectors +
((u_int64_t)meta->configs_high[0].total_sectors << 32);
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = meta->configs[0].rebuild_lba +
((u_int64_t)meta->configs_high[0].rebuild_lba << 32);
raid->lun = array;
strncpy(raid->name, meta->name,
min(sizeof(raid->name), sizeof(meta->name)));
raid->disks[disk_number].sectors = raid->total_sectors /
(raid->type == AR_T_RAID5 ? raid->width - 1 : raid->width);
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].flags =
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
break;
}
hptv3_out:
free(meta, M_AR);
return retval;
}
/* Intel MatrixRAID Metadata */
static int
ata_raid_intel_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct intel_raid_conf *meta;
struct intel_raid_mapping *map;
struct ar_softc *raid = NULL;
u_int32_t checksum, *ptr;
int array, count, disk, volume = 1, retval = 0;
char *tmp;
if (!(meta = (struct intel_raid_conf *)
malloc(1536, M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, INTEL_LBA(parent), meta, 1024, ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "Intel read metadata failed\n");
goto intel_out;
}
tmp = (char *)meta;
bcopy(tmp, tmp+1024, 512);
bcopy(tmp+512, tmp, 1024);
bzero(tmp+1024, 512);
/* check if this is a Intel RAID struct */
if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "Intel check1 failed\n");
goto intel_out;
}
for (checksum = 0, ptr = (u_int32_t *)meta, count = 0;
count < (meta->config_size / sizeof(u_int32_t)); count++) {
checksum += *ptr++;
}
checksum -= meta->checksum;
if (checksum != meta->checksum) {
if (testing || bootverbose)
device_printf(parent, "Intel check2 failed\n");
goto intel_out;
}
if (testing || bootverbose)
ata_raid_intel_print_meta(meta);
map = (struct intel_raid_mapping *)&meta->disk[meta->total_disks];
/* now convert Intel metadata into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto intel_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_INTEL_RAID))
continue;
if ((raid->format & AR_F_INTEL_RAID) &&
(raid->magic_0 != meta->config_id))
continue;
/*
* update our knowledge about the array config based on generation
* NOTE: there can be multiple volumes on a disk set
*/
if (!meta->generation || meta->generation > raid->generation) {
switch (map->type) {
case INTEL_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = map->total_disks;
break;
case INTEL_T_RAID1:
if (map->total_disks == 4)
raid->type = AR_T_RAID01;
else
raid->type = AR_T_RAID1;
raid->width = map->total_disks / 2;
break;
case INTEL_T_RAID5:
raid->type = AR_T_RAID5;
raid->width = map->total_disks;
break;
default:
device_printf(parent, "Intel unknown RAID type 0x%02x\n",
map->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto intel_out;
}
switch (map->status) {
case INTEL_S_READY:
raid->status = AR_S_READY;
break;
case INTEL_S_DEGRADED:
raid->status |= AR_S_DEGRADED;
break;
case INTEL_S_DISABLED:
case INTEL_S_FAILURE:
raid->status = 0;
}
raid->magic_0 = meta->config_id;
raid->format = AR_F_INTEL_RAID;
raid->generation = meta->generation;
raid->interleave = map->stripe_sectors;
raid->total_disks = map->total_disks;
raid->total_sectors = map->total_sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = map->offset;
raid->rebuild_lba = 0;
raid->lun = array;
raid->volume = volume - 1;
strncpy(raid->name, map->name,
min(sizeof(raid->name), sizeof(map->name)));
/* clear out any old info */
for (disk = 0; disk < raid->total_disks; disk++) {
raid->disks[disk].dev = NULL;
bcopy(meta->disk[map->disk_idx[disk]].serial,
raid->disks[disk].serial,
sizeof(raid->disks[disk].serial));
raid->disks[disk].sectors =
meta->disk[map->disk_idx[disk]].sectors;
raid->disks[disk].flags = 0;
if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_ONLINE)
raid->disks[disk].flags |= AR_DF_ONLINE;
if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_ASSIGNED)
raid->disks[disk].flags |= AR_DF_ASSIGNED;
if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_SPARE) {
raid->disks[disk].flags &= ~(AR_DF_ONLINE | AR_DF_ASSIGNED);
raid->disks[disk].flags |= AR_DF_SPARE;
}
if (meta->disk[map->disk_idx[disk]].flags & INTEL_F_DOWN)
raid->disks[disk].flags &= ~AR_DF_ONLINE;
}
}
if (meta->generation >= raid->generation) {
for (disk = 0; disk < raid->total_disks; disk++) {
struct ata_device *atadev = device_get_softc(parent);
if (!strncmp(raid->disks[disk].serial, atadev->param.serial,
sizeof(raid->disks[disk].serial))) {
raid->disks[disk].dev = parent;
raid->disks[disk].flags |= (AR_DF_PRESENT | AR_DF_ONLINE);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk;
retval = 1;
}
}
}
else
goto intel_out;
if (retval) {
if (volume < meta->total_volumes) {
map = (struct intel_raid_mapping *)
&map->disk_idx[map->total_disks];
volume++;
retval = 0;
continue;
}
break;
}
else {
free(raidp[array], M_AR);
raidp[array] = NULL;
if (volume == 2)
retval = 1;
}
}
intel_out:
free(meta, M_AR);
return retval;
}
static int
ata_raid_intel_write_meta(struct ar_softc *rdp)
{
struct intel_raid_conf *meta;
struct intel_raid_mapping *map;
struct timeval timestamp;
u_int32_t checksum, *ptr;
int count, disk, error = 0;
char *tmp;
if (!(meta = (struct intel_raid_conf *)
malloc(1536, M_AR, M_NOWAIT | M_ZERO))) {
printf("ar%d: failed to allocate metadata storage\n", rdp->lun);
return ENOMEM;
}
rdp->generation++;
microtime(&timestamp);
bcopy(INTEL_MAGIC, meta->intel_id, sizeof(meta->intel_id));
bcopy(INTEL_VERSION_1100, meta->version, sizeof(meta->version));
meta->config_id = timestamp.tv_sec;
meta->generation = rdp->generation;
meta->total_disks = rdp->total_disks;
meta->total_volumes = 1; /* XXX SOS */
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
struct ata_channel *ch =
device_get_softc(device_get_parent(rdp->disks[disk].dev));
struct ata_device *atadev =
device_get_softc(rdp->disks[disk].dev);
bcopy(atadev->param.serial, meta->disk[disk].serial,
sizeof(rdp->disks[disk].serial));
meta->disk[disk].sectors = rdp->disks[disk].sectors;
meta->disk[disk].id = (ch->unit << 16) | ATA_DEV(atadev->unit);
}
else
meta->disk[disk].sectors = rdp->total_sectors / rdp->width;
meta->disk[disk].flags = 0;
if (rdp->disks[disk].flags & AR_DF_SPARE)
meta->disk[disk].flags |= INTEL_F_SPARE;
else {
if (rdp->disks[disk].flags & AR_DF_ONLINE)
meta->disk[disk].flags |= INTEL_F_ONLINE;
else
meta->disk[disk].flags |= INTEL_F_DOWN;
if (rdp->disks[disk].flags & AR_DF_ASSIGNED)
meta->disk[disk].flags |= INTEL_F_ASSIGNED;
}
}
map = (struct intel_raid_mapping *)&meta->disk[meta->total_disks];
bcopy(rdp->name, map->name, sizeof(rdp->name));
map->total_sectors = rdp->total_sectors;
map->state = 12; /* XXX SOS */
map->offset = rdp->offset_sectors;
map->stripe_count = rdp->total_sectors / (rdp->interleave*rdp->total_disks);
map->stripe_sectors = rdp->interleave;
map->disk_sectors = rdp->total_sectors / rdp->width;
map->status = INTEL_S_READY; /* XXX SOS */
switch (rdp->type) {
case AR_T_RAID0:
map->type = INTEL_T_RAID0;
break;
case AR_T_RAID1:
map->type = INTEL_T_RAID1;
break;
case AR_T_RAID01:
map->type = INTEL_T_RAID1;
break;
case AR_T_RAID5:
map->type = INTEL_T_RAID5;
break;
default:
free(meta, M_AR);
return ENODEV;
}
map->total_disks = rdp->total_disks;
map->magic[0] = 0x02;
map->magic[1] = 0xff;
map->magic[2] = 0x01;
for (disk = 0; disk < rdp->total_disks; disk++)
map->disk_idx[disk] = disk;
meta->config_size = (char *)&map->disk_idx[disk] - (char *)meta;
for (checksum = 0, ptr = (u_int32_t *)meta, count = 0;
count < (meta->config_size / sizeof(u_int32_t)); count++) {
checksum += *ptr++;
}
meta->checksum = checksum;
if (testing || bootverbose)
ata_raid_intel_print_meta(meta);
tmp = (char *)meta;
bcopy(tmp, tmp+1024, 512);
bcopy(tmp+512, tmp, 1024);
bzero(tmp+1024, 512);
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
if (ata_raid_rw(rdp->disks[disk].dev,
INTEL_LBA(rdp->disks[disk].dev),
meta, 1024, ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "write metadata failed\n");
error = EIO;
}
}
}
free(meta, M_AR);
return error;
}
/* Integrated Technology Express Metadata */
static int
ata_raid_ite_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct ite_raid_conf *meta;
struct ar_softc *raid = NULL;
int array, disk_number, count, retval = 0;
u_int16_t *ptr;
if (!(meta = (struct ite_raid_conf *)
malloc(sizeof(struct ite_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, ITE_LBA(parent),
meta, sizeof(struct ite_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "ITE read metadata failed\n");
goto ite_out;
}
/* check if this is a ITE RAID struct */
for (ptr = (u_int16_t *)meta->ite_id, count = 0;
count < sizeof(meta->ite_id)/sizeof(uint16_t); count++)
ptr[count] = be16toh(ptr[count]);
if (strncmp(meta->ite_id, ITE_MAGIC, strlen(ITE_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "ITE check1 failed\n");
goto ite_out;
}
if (testing || bootverbose)
ata_raid_ite_print_meta(meta);
/* now convert ITE metadata into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if ((raid = raidp[array])) {
if (raid->format != AR_F_ITE_RAID)
continue;
if (raid->magic_0 != *((u_int64_t *)meta->timestamp_0))
continue;
}
/* if we dont have a disks timestamp the RAID is invalidated */
if (*((u_int64_t *)meta->timestamp_1) == 0)
goto ite_out;
if (!raid) {
raidp[array] = (struct ar_softc *)malloc(sizeof(struct ar_softc),
M_AR, M_NOWAIT | M_ZERO);
if (!(raid = raidp[array])) {
device_printf(parent, "failed to allocate metadata storage\n");
goto ite_out;
}
}
switch (meta->type) {
case ITE_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = meta->array_width;
raid->total_disks = meta->array_width;
disk_number = meta->disk_number;
break;
case ITE_T_RAID1:
raid->type = AR_T_RAID1;
raid->width = 1;
raid->total_disks = 2;
disk_number = meta->disk_number;
break;
case ITE_T_RAID01:
raid->type = AR_T_RAID01;
raid->width = meta->array_width;
raid->total_disks = 4;
disk_number = ((meta->disk_number & 0x02) >> 1) |
((meta->disk_number & 0x01) << 1);
break;
case ITE_T_SPAN:
raid->type = AR_T_SPAN;
raid->width = 1;
raid->total_disks = meta->array_width;
disk_number = meta->disk_number;
break;
default:
device_printf(parent, "ITE unknown RAID type 0x%02x\n", meta->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto ite_out;
}
raid->magic_0 = *((u_int64_t *)meta->timestamp_0);
raid->format = AR_F_ITE_RAID;
raid->generation = 0;
raid->interleave = meta->stripe_sectors;
raid->total_sectors = meta->total_sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = 0;
raid->lun = array;
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].sectors = raid->total_sectors / raid->width;
raid->disks[disk_number].flags =
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
break;
}
ite_out:
free(meta, M_AR);
return retval;
}
/* JMicron Technology Corp Metadata */
static int
ata_raid_jmicron_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct jmicron_raid_conf *meta;
struct ar_softc *raid = NULL;
u_int16_t checksum, *ptr;
u_int64_t disk_size;
int count, array, disk, total_disks, retval = 0;
if (!(meta = (struct jmicron_raid_conf *)
malloc(sizeof(struct jmicron_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, JMICRON_LBA(parent),
meta, sizeof(struct jmicron_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent,
"JMicron read metadata failed\n");
}
/* check for JMicron signature */
if (strncmp(meta->signature, JMICRON_MAGIC, 2)) {
if (testing || bootverbose)
device_printf(parent, "JMicron check1 failed\n");
goto jmicron_out;
}
/* calculate checksum and compare for valid */
for (checksum = 0, ptr = (u_int16_t *)meta, count = 0; count < 64; count++)
checksum += *ptr++;
if (checksum) {
if (testing || bootverbose)
device_printf(parent, "JMicron check2 failed\n");
goto jmicron_out;
}
if (testing || bootverbose)
ata_raid_jmicron_print_meta(meta);
/* now convert JMicron meta into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
jmicron_next:
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto jmicron_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_JMICRON_RAID))
continue;
for (total_disks = 0, disk = 0; disk < JM_MAX_DISKS; disk++) {
if (meta->disks[disk]) {
if (raid->format == AR_F_JMICRON_RAID) {
if (bcmp(&meta->disks[disk],
raid->disks[disk].serial, sizeof(u_int32_t))) {
array++;
goto jmicron_next;
}
}
else
bcopy(&meta->disks[disk],
raid->disks[disk].serial, sizeof(u_int32_t));
total_disks++;
}
}
/* handle spares XXX SOS */
switch (meta->type) {
case JM_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = total_disks;
break;
case JM_T_RAID1:
raid->type = AR_T_RAID1;
raid->width = 1;
break;
case JM_T_RAID01:
raid->type = AR_T_RAID01;
raid->width = total_disks / 2;
break;
case JM_T_RAID5:
raid->type = AR_T_RAID5;
raid->width = total_disks;
break;
case JM_T_JBOD:
raid->type = AR_T_SPAN;
raid->width = 1;
break;
default:
device_printf(parent,
"JMicron unknown RAID type 0x%02x\n", meta->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto jmicron_out;
}
disk_size = (meta->disk_sectors_high << 16) + meta->disk_sectors_low;
raid->format = AR_F_JMICRON_RAID;
strncpy(raid->name, meta->name, sizeof(meta->name));
raid->generation = 0;
raid->interleave = 2 << meta->stripe_shift;
raid->total_disks = total_disks;
raid->total_sectors = disk_size * (raid->width-(raid->type==AR_RAID5));
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = meta->offset * 16;
raid->rebuild_lba = 0;
raid->lun = array;
for (disk = 0; disk < raid->total_disks; disk++) {
if (meta->disks[disk] == meta->disk_id) {
raid->disks[disk].dev = parent;
raid->disks[disk].sectors = disk_size;
raid->disks[disk].flags =
(AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk;
retval = 1;
break;
}
}
break;
}
jmicron_out:
free(meta, M_AR);
return retval;
}
static int
ata_raid_jmicron_write_meta(struct ar_softc *rdp)
{
struct jmicron_raid_conf *meta;
u_int64_t disk_sectors;
int disk, error = 0;
if (!(meta = (struct jmicron_raid_conf *)
malloc(sizeof(struct jmicron_raid_conf), M_AR, M_NOWAIT | M_ZERO))) {
printf("ar%d: failed to allocate metadata storage\n", rdp->lun);
return ENOMEM;
}
rdp->generation++;
switch (rdp->type) {
case AR_T_JBOD:
meta->type = JM_T_JBOD;
break;
case AR_T_RAID0:
meta->type = JM_T_RAID0;
break;
case AR_T_RAID1:
meta->type = JM_T_RAID1;
break;
case AR_T_RAID5:
meta->type = JM_T_RAID5;
break;
case AR_T_RAID01:
meta->type = JM_T_RAID01;
break;
default:
free(meta, M_AR);
return ENODEV;
}
bcopy(JMICRON_MAGIC, meta->signature, sizeof(JMICRON_MAGIC));
meta->version = JMICRON_VERSION;
meta->offset = rdp->offset_sectors / 16;
disk_sectors = rdp->total_sectors / (rdp->width - (rdp->type == AR_RAID5));
meta->disk_sectors_low = disk_sectors & 0xffff;
meta->disk_sectors_high = disk_sectors >> 16;
strncpy(meta->name, rdp->name, sizeof(meta->name));
meta->stripe_shift = ffs(rdp->interleave) - 2;
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].serial[0])
bcopy(rdp->disks[disk].serial,&meta->disks[disk],sizeof(u_int32_t));
else
meta->disks[disk] = (u_int32_t)(uintptr_t)rdp->disks[disk].dev;
}
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
u_int16_t checksum = 0, *ptr;
int count;
meta->disk_id = meta->disks[disk];
meta->checksum = 0;
for (ptr = (u_int16_t *)meta, count = 0; count < 64; count++)
checksum += *ptr++;
meta->checksum -= checksum;
if (testing || bootverbose)
ata_raid_jmicron_print_meta(meta);
if (ata_raid_rw(rdp->disks[disk].dev,
JMICRON_LBA(rdp->disks[disk].dev),
meta, sizeof(struct jmicron_raid_conf),
ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "write metadata failed\n");
error = EIO;
}
}
}
/* handle spares XXX SOS */
free(meta, M_AR);
return error;
}
/* LSILogic V2 MegaRAID Metadata */
static int
ata_raid_lsiv2_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct lsiv2_raid_conf *meta;
struct ar_softc *raid = NULL;
int array, retval = 0;
if (!(meta = (struct lsiv2_raid_conf *)
malloc(sizeof(struct lsiv2_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, LSIV2_LBA(parent),
meta, sizeof(struct lsiv2_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "LSI (v2) read metadata failed\n");
goto lsiv2_out;
}
/* check if this is a LSI RAID struct */
if (strncmp(meta->lsi_id, LSIV2_MAGIC, strlen(LSIV2_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "LSI (v2) check1 failed\n");
goto lsiv2_out;
}
if (testing || bootverbose)
ata_raid_lsiv2_print_meta(meta);
/* now convert LSI (v2) config meta into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
int raid_entry, conf_entry;
if (!raidp[array + meta->raid_number]) {
raidp[array + meta->raid_number] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array + meta->raid_number]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto lsiv2_out;
}
}
raid = raidp[array + meta->raid_number];
if (raid->format && (raid->format != AR_F_LSIV2_RAID))
continue;
if (raid->magic_0 &&
((raid->magic_0 != meta->timestamp) ||
(raid->magic_1 != meta->raid_number)))
continue;
array += meta->raid_number;
raid_entry = meta->raid_number;
conf_entry = (meta->configs[raid_entry].raid.config_offset >> 4) +
meta->disk_number - 1;
switch (meta->configs[raid_entry].raid.type) {
case LSIV2_T_RAID0:
raid->magic_0 = meta->timestamp;
raid->magic_1 = meta->raid_number;
raid->type = AR_T_RAID0;
raid->interleave = meta->configs[raid_entry].raid.stripe_sectors;
raid->width = meta->configs[raid_entry].raid.array_width;
break;
case LSIV2_T_RAID1:
raid->magic_0 = meta->timestamp;
raid->magic_1 = meta->raid_number;
raid->type = AR_T_RAID1;
raid->width = meta->configs[raid_entry].raid.array_width;
break;
case LSIV2_T_RAID0 | LSIV2_T_RAID1:
raid->magic_0 = meta->timestamp;
raid->magic_1 = meta->raid_number;
raid->type = AR_T_RAID01;
raid->interleave = meta->configs[raid_entry].raid.stripe_sectors;
raid->width = meta->configs[raid_entry].raid.array_width;
break;
default:
device_printf(parent, "LSI v2 unknown RAID type 0x%02x\n",
meta->configs[raid_entry].raid.type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto lsiv2_out;
}
raid->format = AR_F_LSIV2_RAID;
raid->generation = 0;
raid->total_disks = meta->configs[raid_entry].raid.disk_count;
raid->total_sectors = meta->configs[raid_entry].raid.total_sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = 0;
raid->lun = array;
if (meta->configs[conf_entry].disk.device != LSIV2_D_NONE) {
raid->disks[meta->disk_number].dev = parent;
raid->disks[meta->disk_number].sectors =
meta->configs[conf_entry].disk.disk_sectors;
raid->disks[meta->disk_number].flags =
(AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = meta->disk_number;
retval = 1;
}
else
raid->disks[meta->disk_number].flags &= ~AR_DF_ONLINE;
break;
}
lsiv2_out:
free(meta, M_AR);
return retval;
}
/* LSILogic V3 MegaRAID Metadata */
static int
ata_raid_lsiv3_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct lsiv3_raid_conf *meta;
struct ar_softc *raid = NULL;
u_int8_t checksum, *ptr;
int array, entry, count, disk_number, retval = 0;
if (!(meta = (struct lsiv3_raid_conf *)
malloc(sizeof(struct lsiv3_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, LSIV3_LBA(parent),
meta, sizeof(struct lsiv3_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "LSI (v3) read metadata failed\n");
goto lsiv3_out;
}
/* check if this is a LSI RAID struct */
if (strncmp(meta->lsi_id, LSIV3_MAGIC, strlen(LSIV3_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "LSI (v3) check1 failed\n");
goto lsiv3_out;
}
/* check if the checksum is OK */
for (checksum = 0, ptr = meta->lsi_id, count = 0; count < 512; count++)
checksum += *ptr++;
if (checksum) {
if (testing || bootverbose)
device_printf(parent, "LSI (v3) check2 failed\n");
goto lsiv3_out;
}
if (testing || bootverbose)
ata_raid_lsiv3_print_meta(meta);
/* now convert LSI (v3) config meta into our generic form */
for (array = 0, entry = 0; array < MAX_ARRAYS && entry < 8;) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto lsiv3_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_LSIV3_RAID)) {
array++;
continue;
}
if ((raid->format == AR_F_LSIV3_RAID) &&
(raid->magic_0 != meta->timestamp)) {
array++;
continue;
}
switch (meta->raid[entry].total_disks) {
case 0:
entry++;
continue;
case 1:
if (meta->raid[entry].device == meta->device) {
disk_number = 0;
break;
}
if (raid->format)
array++;
entry++;
continue;
case 2:
disk_number = (meta->device & (LSIV3_D_DEVICE|LSIV3_D_CHANNEL))?1:0;
break;
default:
device_printf(parent, "lsiv3 > 2 disk support untested!!\n");
disk_number = (meta->device & LSIV3_D_DEVICE ? 1 : 0) +
(meta->device & LSIV3_D_CHANNEL ? 2 : 0);
break;
}
switch (meta->raid[entry].type) {
case LSIV3_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = meta->raid[entry].total_disks;
break;
case LSIV3_T_RAID1:
raid->type = AR_T_RAID1;
raid->width = meta->raid[entry].array_width;
break;
default:
device_printf(parent, "LSI v3 unknown RAID type 0x%02x\n",
meta->raid[entry].type);
free(raidp[array], M_AR);
raidp[array] = NULL;
entry++;
continue;
}
raid->magic_0 = meta->timestamp;
raid->format = AR_F_LSIV3_RAID;
raid->generation = 0;
raid->interleave = meta->raid[entry].stripe_pages * 8;
raid->total_disks = meta->raid[entry].total_disks;
raid->total_sectors = raid->width * meta->raid[entry].sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = meta->raid[entry].offset;
raid->rebuild_lba = 0;
raid->lun = array;
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].sectors = raid->total_sectors / raid->width;
raid->disks[disk_number].flags =
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
entry++;
array++;
}
lsiv3_out:
free(meta, M_AR);
return retval;
}
/* nVidia MediaShield Metadata */
static int
ata_raid_nvidia_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct nvidia_raid_conf *meta;
struct ar_softc *raid = NULL;
u_int32_t checksum, *ptr;
int array, count, retval = 0;
if (!(meta = (struct nvidia_raid_conf *)
malloc(sizeof(struct nvidia_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, NVIDIA_LBA(parent),
meta, sizeof(struct nvidia_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "nVidia read metadata failed\n");
goto nvidia_out;
}
/* check if this is a nVidia RAID struct */
if (strncmp(meta->nvidia_id, NV_MAGIC, strlen(NV_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "nVidia check1 failed\n");
goto nvidia_out;
}
/* check if the checksum is OK */
for (checksum = 0, ptr = (u_int32_t*)meta, count = 0;
count < meta->config_size; count++)
checksum += *ptr++;
if (checksum) {
if (testing || bootverbose)
device_printf(parent, "nVidia check2 failed\n");
goto nvidia_out;
}
if (testing || bootverbose)
ata_raid_nvidia_print_meta(meta);
/* now convert nVidia meta into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto nvidia_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_NVIDIA_RAID))
continue;
if (raid->format == AR_F_NVIDIA_RAID &&
((raid->magic_0 != meta->magic_1) ||
(raid->magic_1 != meta->magic_2))) {
continue;
}
switch (meta->type) {
case NV_T_SPAN:
raid->type = AR_T_SPAN;
break;
case NV_T_RAID0:
raid->type = AR_T_RAID0;
break;
case NV_T_RAID1:
raid->type = AR_T_RAID1;
break;
case NV_T_RAID5:
raid->type = AR_T_RAID5;
break;
case NV_T_RAID01:
raid->type = AR_T_RAID01;
break;
default:
device_printf(parent, "nVidia unknown RAID type 0x%02x\n",
meta->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto nvidia_out;
}
raid->magic_0 = meta->magic_1;
raid->magic_1 = meta->magic_2;
raid->format = AR_F_NVIDIA_RAID;
raid->generation = 0;
raid->interleave = meta->stripe_sectors;
raid->width = meta->array_width;
raid->total_disks = meta->total_disks;
raid->total_sectors = meta->total_sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = meta->rebuild_lba;
raid->lun = array;
raid->status = AR_S_READY;
if (meta->status & NV_S_DEGRADED)
raid->status |= AR_S_DEGRADED;
raid->disks[meta->disk_number].dev = parent;
raid->disks[meta->disk_number].sectors =
raid->total_sectors / raid->width;
raid->disks[meta->disk_number].flags =
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = meta->disk_number;
retval = 1;
break;
}
nvidia_out:
free(meta, M_AR);
return retval;
}
/* Promise FastTrak Metadata */
static int
ata_raid_promise_read_meta(device_t dev, struct ar_softc **raidp, int native)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct promise_raid_conf *meta;
struct ar_softc *raid;
u_int32_t checksum, *ptr;
int array, count, disk, disksum = 0, retval = 0;
if (!(meta = (struct promise_raid_conf *)
malloc(sizeof(struct promise_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, PROMISE_LBA(parent),
meta, sizeof(struct promise_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "%s read metadata failed\n",
native ? "FreeBSD" : "Promise");
goto promise_out;
}
/* check the signature */
if (native) {
if (strncmp(meta->promise_id, ATA_MAGIC, strlen(ATA_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "FreeBSD check1 failed\n");
goto promise_out;
}
}
else {
if (strncmp(meta->promise_id, PR_MAGIC, strlen(PR_MAGIC))) {
if (testing || bootverbose)
device_printf(parent, "Promise check1 failed\n");
goto promise_out;
}
}
/* check if the checksum is OK */
for (checksum = 0, ptr = (u_int32_t *)meta, count = 0; count < 511; count++)
checksum += *ptr++;
if (checksum != *ptr) {
if (testing || bootverbose)
device_printf(parent, "%s check2 failed\n",
native ? "FreeBSD" : "Promise");
goto promise_out;
}
/* check on disk integrity status */
if (meta->raid.integrity != PR_I_VALID) {
if (testing || bootverbose)
device_printf(parent, "%s check3 failed\n",
native ? "FreeBSD" : "Promise");
goto promise_out;
}
if (testing || bootverbose)
ata_raid_promise_print_meta(meta);
/* now convert Promise metadata into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto promise_out;
}
}
raid = raidp[array];
if (raid->format &&
(raid->format != (native ? AR_F_FREEBSD_RAID : AR_F_PROMISE_RAID)))
continue;
if ((raid->format == (native ? AR_F_FREEBSD_RAID : AR_F_PROMISE_RAID))&&
!(meta->raid.magic_1 == (raid->magic_1)))
continue;
/* update our knowledge about the array config based on generation */
if (!meta->raid.generation || meta->raid.generation > raid->generation){
switch (meta->raid.type) {
case PR_T_SPAN:
raid->type = AR_T_SPAN;
break;
case PR_T_JBOD:
raid->type = AR_T_JBOD;
break;
case PR_T_RAID0:
raid->type = AR_T_RAID0;
break;
case PR_T_RAID1:
raid->type = AR_T_RAID1;
if (meta->raid.array_width > 1)
raid->type = AR_T_RAID01;
break;
case PR_T_RAID5:
raid->type = AR_T_RAID5;
break;
default:
device_printf(parent, "%s unknown RAID type 0x%02x\n",
native ? "FreeBSD" : "Promise", meta->raid.type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto promise_out;
}
raid->magic_1 = meta->raid.magic_1;
raid->format = (native ? AR_F_FREEBSD_RAID : AR_F_PROMISE_RAID);
raid->generation = meta->raid.generation;
raid->interleave = 1 << meta->raid.stripe_shift;
raid->width = meta->raid.array_width;
raid->total_disks = meta->raid.total_disks;
raid->heads = meta->raid.heads + 1;
raid->sectors = meta->raid.sectors;
raid->cylinders = meta->raid.cylinders + 1;
raid->total_sectors = meta->raid.total_sectors;
raid->offset_sectors = 0;
raid->rebuild_lba = meta->raid.rebuild_lba;
raid->lun = array;
if ((meta->raid.status &
(PR_S_VALID | PR_S_ONLINE | PR_S_INITED | PR_S_READY)) ==
(PR_S_VALID | PR_S_ONLINE | PR_S_INITED | PR_S_READY)) {
raid->status |= AR_S_READY;
if (meta->raid.status & PR_S_DEGRADED)
raid->status |= AR_S_DEGRADED;
}
else
raid->status &= ~AR_S_READY;
/* convert disk flags to our internal types */
for (disk = 0; disk < meta->raid.total_disks; disk++) {
raid->disks[disk].dev = NULL;
raid->disks[disk].flags = 0;
*((u_int64_t *)(raid->disks[disk].serial)) =
meta->raid.disk[disk].magic_0;
disksum += meta->raid.disk[disk].flags;
if (meta->raid.disk[disk].flags & PR_F_ONLINE)
raid->disks[disk].flags |= AR_DF_ONLINE;
if (meta->raid.disk[disk].flags & PR_F_ASSIGNED)
raid->disks[disk].flags |= AR_DF_ASSIGNED;
if (meta->raid.disk[disk].flags & PR_F_SPARE) {
raid->disks[disk].flags &= ~(AR_DF_ONLINE | AR_DF_ASSIGNED);
raid->disks[disk].flags |= AR_DF_SPARE;
}
if (meta->raid.disk[disk].flags & (PR_F_REDIR | PR_F_DOWN))
raid->disks[disk].flags &= ~AR_DF_ONLINE;
}
if (!disksum) {
device_printf(parent, "%s subdisks has no flags\n",
native ? "FreeBSD" : "Promise");
free(raidp[array], M_AR);
raidp[array] = NULL;
goto promise_out;
}
}
if (meta->raid.generation >= raid->generation) {
int disk_number = meta->raid.disk_number;
if (raid->disks[disk_number].flags && (meta->magic_0 ==
*((u_int64_t *)(raid->disks[disk_number].serial)))) {
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].flags |= AR_DF_PRESENT;
raid->disks[disk_number].sectors = meta->raid.disk_sectors;
if ((raid->disks[disk_number].flags &
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE)) ==
(AR_DF_PRESENT | AR_DF_ASSIGNED | AR_DF_ONLINE)) {
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
}
}
}
break;
}
promise_out:
free(meta, M_AR);
return retval;
}
static int
ata_raid_promise_write_meta(struct ar_softc *rdp)
{
struct promise_raid_conf *meta;
struct timeval timestamp;
u_int32_t *ckptr;
int count, disk, drive, error = 0;
if (!(meta = (struct promise_raid_conf *)
malloc(sizeof(struct promise_raid_conf), M_AR, M_NOWAIT))) {
printf("ar%d: failed to allocate metadata storage\n", rdp->lun);
return ENOMEM;
}
rdp->generation++;
microtime(&timestamp);
for (disk = 0; disk < rdp->total_disks; disk++) {
for (count = 0; count < sizeof(struct promise_raid_conf); count++)
*(((u_int8_t *)meta) + count) = 255 - (count % 256);
meta->dummy_0 = 0x00020000;
meta->raid.disk_number = disk;
if (rdp->disks[disk].dev) {
struct ata_device *atadev = device_get_softc(rdp->disks[disk].dev);
struct ata_channel *ch =
device_get_softc(device_get_parent(rdp->disks[disk].dev));
meta->raid.channel = ch->unit;
meta->raid.device = ATA_DEV(atadev->unit);
meta->raid.disk_sectors = rdp->disks[disk].sectors;
meta->raid.disk_offset = rdp->offset_sectors;
}
else {
meta->raid.channel = 0;
meta->raid.device = 0;
meta->raid.disk_sectors = 0;
meta->raid.disk_offset = 0;
}
meta->magic_0 = PR_MAGIC0(meta->raid) | timestamp.tv_sec;
meta->magic_1 = timestamp.tv_sec >> 16;
meta->magic_2 = timestamp.tv_sec;
meta->raid.integrity = PR_I_VALID;
meta->raid.magic_0 = meta->magic_0;
meta->raid.rebuild_lba = rdp->rebuild_lba;
meta->raid.generation = rdp->generation;
if (rdp->status & AR_S_READY) {
meta->raid.flags = (PR_F_VALID | PR_F_ASSIGNED | PR_F_ONLINE);
meta->raid.status =
(PR_S_VALID | PR_S_ONLINE | PR_S_INITED | PR_S_READY);
if (rdp->status & AR_S_DEGRADED)
meta->raid.status |= PR_S_DEGRADED;
else
meta->raid.status |= PR_S_FUNCTIONAL;
}
else {
meta->raid.flags = PR_F_DOWN;
meta->raid.status = 0;
}
switch (rdp->type) {
case AR_T_RAID0:
meta->raid.type = PR_T_RAID0;
break;
case AR_T_RAID1:
meta->raid.type = PR_T_RAID1;
break;
case AR_T_RAID01:
meta->raid.type = PR_T_RAID1;
break;
case AR_T_RAID5:
meta->raid.type = PR_T_RAID5;
break;
case AR_T_SPAN:
meta->raid.type = PR_T_SPAN;
break;
case AR_T_JBOD:
meta->raid.type = PR_T_JBOD;
break;
default:
free(meta, M_AR);
return ENODEV;
}
meta->raid.total_disks = rdp->total_disks;
meta->raid.stripe_shift = ffs(rdp->interleave) - 1;
meta->raid.array_width = rdp->width;
meta->raid.array_number = rdp->lun;
meta->raid.total_sectors = rdp->total_sectors;
meta->raid.cylinders = rdp->cylinders - 1;
meta->raid.heads = rdp->heads - 1;
meta->raid.sectors = rdp->sectors;
meta->raid.magic_1 = (u_int64_t)meta->magic_2<<16 | meta->magic_1;
bzero(&meta->raid.disk, 8 * 12);
for (drive = 0; drive < rdp->total_disks; drive++) {
meta->raid.disk[drive].flags = 0;
if (rdp->disks[drive].flags & AR_DF_PRESENT)
meta->raid.disk[drive].flags |= PR_F_VALID;
if (rdp->disks[drive].flags & AR_DF_ASSIGNED)
meta->raid.disk[drive].flags |= PR_F_ASSIGNED;
if (rdp->disks[drive].flags & AR_DF_ONLINE)
meta->raid.disk[drive].flags |= PR_F_ONLINE;
else
if (rdp->disks[drive].flags & AR_DF_PRESENT)
meta->raid.disk[drive].flags = (PR_F_REDIR | PR_F_DOWN);
if (rdp->disks[drive].flags & AR_DF_SPARE)
meta->raid.disk[drive].flags |= PR_F_SPARE;
meta->raid.disk[drive].dummy_0 = 0x0;
if (rdp->disks[drive].dev) {
struct ata_channel *ch =
device_get_softc(device_get_parent(rdp->disks[drive].dev));
struct ata_device *atadev =
device_get_softc(rdp->disks[drive].dev);
meta->raid.disk[drive].channel = ch->unit;
meta->raid.disk[drive].device = ATA_DEV(atadev->unit);
}
meta->raid.disk[drive].magic_0 =
PR_MAGIC0(meta->raid.disk[drive]) | timestamp.tv_sec;
}
if (rdp->disks[disk].dev) {
if ((rdp->disks[disk].flags & (AR_DF_PRESENT | AR_DF_ONLINE)) ==
(AR_DF_PRESENT | AR_DF_ONLINE)) {
if (rdp->format == AR_F_FREEBSD_RAID)
bcopy(ATA_MAGIC, meta->promise_id, sizeof(ATA_MAGIC));
else
bcopy(PR_MAGIC, meta->promise_id, sizeof(PR_MAGIC));
}
else
bzero(meta->promise_id, sizeof(meta->promise_id));
meta->checksum = 0;
for (ckptr = (int32_t *)meta, count = 0; count < 511; count++)
meta->checksum += *ckptr++;
if (testing || bootverbose)
ata_raid_promise_print_meta(meta);
if (ata_raid_rw(rdp->disks[disk].dev,
PROMISE_LBA(rdp->disks[disk].dev),
meta, sizeof(struct promise_raid_conf),
ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "write metadata failed\n");
error = EIO;
}
}
}
free(meta, M_AR);
return error;
}
/* Silicon Image Medley Metadata */
static int
ata_raid_sii_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct sii_raid_conf *meta;
struct ar_softc *raid = NULL;
u_int16_t checksum, *ptr;
int array, count, disk, retval = 0;
if (!(meta = (struct sii_raid_conf *)
malloc(sizeof(struct sii_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, SII_LBA(parent),
meta, sizeof(struct sii_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "Silicon Image read metadata failed\n");
goto sii_out;
}
/* check if this is a Silicon Image (Medley) RAID struct */
for (checksum = 0, ptr = (u_int16_t *)meta, count = 0; count < 160; count++)
checksum += *ptr++;
if (checksum) {
if (testing || bootverbose)
device_printf(parent, "Silicon Image check1 failed\n");
goto sii_out;
}
for (checksum = 0, ptr = (u_int16_t *)meta, count = 0; count < 256; count++)
checksum += *ptr++;
if (checksum != meta->checksum_1) {
if (testing || bootverbose)
device_printf(parent, "Silicon Image check2 failed\n");
goto sii_out;
}
/* check verison */
if (meta->version_major != 0x0002 ||
(meta->version_minor != 0x0000 && meta->version_minor != 0x0001)) {
if (testing || bootverbose)
device_printf(parent, "Silicon Image check3 failed\n");
goto sii_out;
}
if (testing || bootverbose)
ata_raid_sii_print_meta(meta);
/* now convert Silicon Image meta into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto sii_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_SII_RAID))
continue;
if (raid->format == AR_F_SII_RAID &&
(raid->magic_0 != *((u_int64_t *)meta->timestamp))) {
continue;
}
/* update our knowledge about the array config based on generation */
if (!meta->generation || meta->generation > raid->generation) {
switch (meta->type) {
case SII_T_RAID0:
raid->type = AR_T_RAID0;
break;
case SII_T_RAID1:
raid->type = AR_T_RAID1;
break;
case SII_T_RAID01:
raid->type = AR_T_RAID01;
break;
case SII_T_SPARE:
device_printf(parent, "Silicon Image SPARE disk\n");
free(raidp[array], M_AR);
raidp[array] = NULL;
goto sii_out;
default:
device_printf(parent,"Silicon Image unknown RAID type 0x%02x\n",
meta->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto sii_out;
}
raid->magic_0 = *((u_int64_t *)meta->timestamp);
raid->format = AR_F_SII_RAID;
raid->generation = meta->generation;
raid->interleave = meta->stripe_sectors;
raid->width = (meta->raid0_disks != 0xff) ? meta->raid0_disks : 1;
raid->total_disks =
((meta->raid0_disks != 0xff) ? meta->raid0_disks : 0) +
((meta->raid1_disks != 0xff) ? meta->raid1_disks : 0);
raid->total_sectors = meta->total_sectors;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = meta->rebuild_lba;
raid->lun = array;
strncpy(raid->name, meta->name,
min(sizeof(raid->name), sizeof(meta->name)));
/* clear out any old info */
if (raid->generation) {
for (disk = 0; disk < raid->total_disks; disk++) {
raid->disks[disk].dev = NULL;
raid->disks[disk].flags = 0;
}
}
}
if (meta->generation >= raid->generation) {
/* XXX SOS add check for the right physical disk by serial# */
if (meta->status & SII_S_READY) {
int disk_number = (raid->type == AR_T_RAID01) ?
meta->raid1_ident + (meta->raid0_ident << 1) :
meta->disk_number;
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].sectors =
raid->total_sectors / raid->width;
raid->disks[disk_number].flags =
(AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
retval = 1;
}
}
break;
}
sii_out:
free(meta, M_AR);
return retval;
}
/* Silicon Integrated Systems Metadata */
static int
ata_raid_sis_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct sis_raid_conf *meta;
struct ar_softc *raid = NULL;
int array, disk_number, drive, retval = 0;
if (!(meta = (struct sis_raid_conf *)
malloc(sizeof(struct sis_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, SIS_LBA(parent),
meta, sizeof(struct sis_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent,
"Silicon Integrated Systems read metadata failed\n");
}
/* check for SiS magic */
if (meta->magic != SIS_MAGIC) {
if (testing || bootverbose)
device_printf(parent,
"Silicon Integrated Systems check1 failed\n");
goto sis_out;
}
if (testing || bootverbose)
ata_raid_sis_print_meta(meta);
/* now convert SiS meta into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto sis_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_SIS_RAID))
continue;
if ((raid->format == AR_F_SIS_RAID) &&
((raid->magic_0 != meta->controller_pci_id) ||
(raid->magic_1 != meta->timestamp))) {
continue;
}
switch (meta->type_total_disks & SIS_T_MASK) {
case SIS_T_JBOD:
raid->type = AR_T_JBOD;
raid->width = (meta->type_total_disks & SIS_D_MASK);
raid->total_sectors += SIS_LBA(parent);
break;
case SIS_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = (meta->type_total_disks & SIS_D_MASK);
if (!raid->total_sectors ||
(raid->total_sectors > (raid->width * SIS_LBA(parent))))
raid->total_sectors = raid->width * SIS_LBA(parent);
break;
case SIS_T_RAID1:
raid->type = AR_T_RAID1;
raid->width = 1;
if (!raid->total_sectors || (raid->total_sectors > SIS_LBA(parent)))
raid->total_sectors = SIS_LBA(parent);
break;
default:
device_printf(parent, "Silicon Integrated Systems "
"unknown RAID type 0x%08x\n", meta->magic);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto sis_out;
}
raid->magic_0 = meta->controller_pci_id;
raid->magic_1 = meta->timestamp;
raid->format = AR_F_SIS_RAID;
raid->generation = 0;
raid->interleave = meta->stripe_sectors;
raid->total_disks = (meta->type_total_disks & SIS_D_MASK);
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = 0;
raid->lun = array;
/* XXX SOS if total_disks > 2 this doesn't float */
if (((meta->disks & SIS_D_MASTER) >> 4) == meta->disk_number)
disk_number = 0;
else
disk_number = 1;
for (drive = 0; drive < raid->total_disks; drive++) {
raid->disks[drive].sectors = raid->total_sectors/raid->width;
if (drive == disk_number) {
raid->disks[disk_number].dev = parent;
raid->disks[disk_number].flags =
(AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk_number;
}
}
retval = 1;
break;
}
sis_out:
free(meta, M_AR);
return retval;
}
static int
ata_raid_sis_write_meta(struct ar_softc *rdp)
{
struct sis_raid_conf *meta;
struct timeval timestamp;
int disk, error = 0;
if (!(meta = (struct sis_raid_conf *)
malloc(sizeof(struct sis_raid_conf), M_AR, M_NOWAIT | M_ZERO))) {
printf("ar%d: failed to allocate metadata storage\n", rdp->lun);
return ENOMEM;
}
rdp->generation++;
microtime(&timestamp);
meta->magic = SIS_MAGIC;
/* XXX SOS if total_disks > 2 this doesn't float */
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
struct ata_channel *ch =
device_get_softc(device_get_parent(rdp->disks[disk].dev));
struct ata_device *atadev = device_get_softc(rdp->disks[disk].dev);
int disk_number = 1 + ATA_DEV(atadev->unit) + (ch->unit << 1);
meta->disks |= disk_number << ((1 - disk) << 2);
}
}
switch (rdp->type) {
case AR_T_JBOD:
meta->type_total_disks = SIS_T_JBOD;
break;
case AR_T_RAID0:
meta->type_total_disks = SIS_T_RAID0;
break;
case AR_T_RAID1:
meta->type_total_disks = SIS_T_RAID1;
break;
default:
free(meta, M_AR);
return ENODEV;
}
meta->type_total_disks |= (rdp->total_disks & SIS_D_MASK);
meta->stripe_sectors = rdp->interleave;
meta->timestamp = timestamp.tv_sec;
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
struct ata_channel *ch =
device_get_softc(device_get_parent(rdp->disks[disk].dev));
struct ata_device *atadev = device_get_softc(rdp->disks[disk].dev);
meta->controller_pci_id =
(pci_get_vendor(GRANDPARENT(rdp->disks[disk].dev)) << 16) |
pci_get_device(GRANDPARENT(rdp->disks[disk].dev));
bcopy(atadev->param.model, meta->model, sizeof(meta->model));
/* XXX SOS if total_disks > 2 this may not float */
meta->disk_number = 1 + ATA_DEV(atadev->unit) + (ch->unit << 1);
if (testing || bootverbose)
ata_raid_sis_print_meta(meta);
if (ata_raid_rw(rdp->disks[disk].dev,
SIS_LBA(rdp->disks[disk].dev),
meta, sizeof(struct sis_raid_conf),
ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "write metadata failed\n");
error = EIO;
}
}
}
free(meta, M_AR);
return error;
}
/* VIA Tech V-RAID Metadata */
static int
ata_raid_via_read_meta(device_t dev, struct ar_softc **raidp)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
device_t parent = device_get_parent(dev);
struct via_raid_conf *meta;
struct ar_softc *raid = NULL;
u_int8_t checksum, *ptr;
int array, count, disk, retval = 0;
if (!(meta = (struct via_raid_conf *)
malloc(sizeof(struct via_raid_conf), M_AR, M_NOWAIT | M_ZERO)))
return ENOMEM;
if (ata_raid_rw(parent, VIA_LBA(parent),
meta, sizeof(struct via_raid_conf), ATA_R_READ)) {
if (testing || bootverbose)
device_printf(parent, "VIA read metadata failed\n");
goto via_out;
}
/* check if this is a VIA RAID struct */
if (meta->magic != VIA_MAGIC) {
if (testing || bootverbose)
device_printf(parent, "VIA check1 failed\n");
goto via_out;
}
/* calculate checksum and compare for valid */
for (checksum = 0, ptr = (u_int8_t *)meta, count = 0; count < 50; count++)
checksum += *ptr++;
if (checksum != meta->checksum) {
if (testing || bootverbose)
device_printf(parent, "VIA check2 failed\n");
goto via_out;
}
if (testing || bootverbose)
ata_raid_via_print_meta(meta);
/* now convert VIA meta into our generic form */
for (array = 0; array < MAX_ARRAYS; array++) {
if (!raidp[array]) {
raidp[array] =
(struct ar_softc *)malloc(sizeof(struct ar_softc), M_AR,
M_NOWAIT | M_ZERO);
if (!raidp[array]) {
device_printf(parent, "failed to allocate metadata storage\n");
goto via_out;
}
}
raid = raidp[array];
if (raid->format && (raid->format != AR_F_VIA_RAID))
continue;
if (raid->format == AR_F_VIA_RAID && (raid->magic_0 != meta->disks[0]))
continue;
switch (meta->type & VIA_T_MASK) {
case VIA_T_RAID0:
raid->type = AR_T_RAID0;
raid->width = meta->stripe_layout & VIA_L_DISKS;
if (!raid->total_sectors ||
(raid->total_sectors > (raid->width * meta->disk_sectors)))
raid->total_sectors = raid->width * meta->disk_sectors;
break;
case VIA_T_RAID1:
raid->type = AR_T_RAID1;
raid->width = 1;
raid->total_sectors = meta->disk_sectors;
break;
case VIA_T_RAID01:
raid->type = AR_T_RAID01;
raid->width = meta->stripe_layout & VIA_L_DISKS;
if (!raid->total_sectors ||
(raid->total_sectors > (raid->width * meta->disk_sectors)))
raid->total_sectors = raid->width * meta->disk_sectors;
break;
case VIA_T_RAID5:
raid->type = AR_T_RAID5;
raid->width = meta->stripe_layout & VIA_L_DISKS;
if (!raid->total_sectors ||
(raid->total_sectors > ((raid->width - 1)*meta->disk_sectors)))
raid->total_sectors = (raid->width - 1) * meta->disk_sectors;
break;
case VIA_T_SPAN:
raid->type = AR_T_SPAN;
raid->width = 1;
raid->total_sectors += meta->disk_sectors;
break;
default:
device_printf(parent,"VIA unknown RAID type 0x%02x\n", meta->type);
free(raidp[array], M_AR);
raidp[array] = NULL;
goto via_out;
}
raid->magic_0 = meta->disks[0];
raid->format = AR_F_VIA_RAID;
raid->generation = 0;
raid->interleave =
0x08 << ((meta->stripe_layout & VIA_L_MASK) >> VIA_L_SHIFT);
for (count = 0, disk = 0; disk < 8; disk++)
if (meta->disks[disk])
count++;
raid->total_disks = count;
raid->heads = 255;
raid->sectors = 63;
raid->cylinders = raid->total_sectors / (63 * 255);
raid->offset_sectors = 0;
raid->rebuild_lba = 0;
raid->lun = array;
for (disk = 0; disk < raid->total_disks; disk++) {
if (meta->disks[disk] == meta->disk_id) {
raid->disks[disk].dev = parent;
bcopy(&meta->disk_id, raid->disks[disk].serial,
sizeof(u_int32_t));
raid->disks[disk].sectors = meta->disk_sectors;
raid->disks[disk].flags =
(AR_DF_ONLINE | AR_DF_PRESENT | AR_DF_ASSIGNED);
ars->raid[raid->volume] = raid;
ars->disk_number[raid->volume] = disk;
retval = 1;
break;
}
}
break;
}
via_out:
free(meta, M_AR);
return retval;
}
static int
ata_raid_via_write_meta(struct ar_softc *rdp)
{
struct via_raid_conf *meta;
int disk, error = 0;
if (!(meta = (struct via_raid_conf *)
malloc(sizeof(struct via_raid_conf), M_AR, M_NOWAIT | M_ZERO))) {
printf("ar%d: failed to allocate metadata storage\n", rdp->lun);
return ENOMEM;
}
rdp->generation++;
meta->magic = VIA_MAGIC;
meta->dummy_0 = 0x02;
switch (rdp->type) {
case AR_T_SPAN:
meta->type = VIA_T_SPAN;
meta->stripe_layout = (rdp->total_disks & VIA_L_DISKS);
break;
case AR_T_RAID0:
meta->type = VIA_T_RAID0;
meta->stripe_layout = ((rdp->interleave >> 1) & VIA_L_MASK);
meta->stripe_layout |= (rdp->total_disks & VIA_L_DISKS);
break;
case AR_T_RAID1:
meta->type = VIA_T_RAID1;
meta->stripe_layout = (rdp->total_disks & VIA_L_DISKS);
break;
case AR_T_RAID5:
meta->type = VIA_T_RAID5;
meta->stripe_layout = ((rdp->interleave >> 1) & VIA_L_MASK);
meta->stripe_layout |= (rdp->total_disks & VIA_L_DISKS);
break;
case AR_T_RAID01:
meta->type = VIA_T_RAID01;
meta->stripe_layout = ((rdp->interleave >> 1) & VIA_L_MASK);
meta->stripe_layout |= (rdp->width & VIA_L_DISKS);
break;
default:
free(meta, M_AR);
return ENODEV;
}
meta->type |= VIA_T_BOOTABLE; /* XXX SOS */
meta->disk_sectors =
rdp->total_sectors / (rdp->width - (rdp->type == AR_RAID5));
for (disk = 0; disk < rdp->total_disks; disk++)
meta->disks[disk] = (u_int32_t)(uintptr_t)rdp->disks[disk].dev;
for (disk = 0; disk < rdp->total_disks; disk++) {
if (rdp->disks[disk].dev) {
u_int8_t *ptr;
int count;
meta->disk_index = disk * sizeof(u_int32_t);
if (rdp->type == AR_T_RAID01)
meta->disk_index = ((meta->disk_index & 0x08) << 2) |
(meta->disk_index & ~0x08);
meta->disk_id = meta->disks[disk];
meta->checksum = 0;
for (ptr = (u_int8_t *)meta, count = 0; count < 50; count++)
meta->checksum += *ptr++;
if (testing || bootverbose)
ata_raid_via_print_meta(meta);
if (ata_raid_rw(rdp->disks[disk].dev,
VIA_LBA(rdp->disks[disk].dev),
meta, sizeof(struct via_raid_conf),
ATA_R_WRITE | ATA_R_DIRECT)) {
device_printf(rdp->disks[disk].dev, "write metadata failed\n");
error = EIO;
}
}
}
free(meta, M_AR);
return error;
}
static struct ata_request *
ata_raid_init_request(struct ar_softc *rdp, struct bio *bio)
{
struct ata_request *request;
if (!(request = ata_alloc_request())) {
printf("FAILURE - out of memory in ata_raid_init_request\n");
return NULL;
}
request->timeout = 5;
request->retries = 2;
request->callback = ata_raid_done;
request->driver = rdp;
request->bio = bio;
switch (request->bio->bio_cmd) {
case BIO_READ:
request->flags = ATA_R_READ;
break;
case BIO_WRITE:
request->flags = ATA_R_WRITE;
break;
}
return request;
}
static int
ata_raid_send_request(struct ata_request *request)
{
struct ata_device *atadev = device_get_softc(request->dev);
request->transfersize = min(request->bytecount, atadev->max_iosize);
if (request->flags & ATA_R_READ) {
if (atadev->mode >= ATA_DMA) {
request->flags |= ATA_R_DMA;
request->u.ata.command = ATA_READ_DMA;
}
else if (atadev->max_iosize > DEV_BSIZE)
request->u.ata.command = ATA_READ_MUL;
else
request->u.ata.command = ATA_READ;
}
else if (request->flags & ATA_R_WRITE) {
if (atadev->mode >= ATA_DMA) {
request->flags |= ATA_R_DMA;
request->u.ata.command = ATA_WRITE_DMA;
}
else if (atadev->max_iosize > DEV_BSIZE)
request->u.ata.command = ATA_WRITE_MUL;
else
request->u.ata.command = ATA_WRITE;
}
else {
device_printf(request->dev, "FAILURE - unknown IO operation\n");
ata_free_request(request);
return EIO;
}
request->flags |= (ATA_R_ORDERED | ATA_R_THREAD);
ata_queue_request(request);
return 0;
}
static int
ata_raid_rw(device_t dev, u_int64_t lba, void *data, u_int bcount, int flags)
{
struct ata_device *atadev = device_get_softc(dev);
struct ata_request *request;
int error;
if (bcount % DEV_BSIZE) {
device_printf(dev, "FAILURE - transfers must be modulo sectorsize\n");
return ENOMEM;
}
if (!(request = ata_alloc_request())) {
device_printf(dev, "FAILURE - out of memory in ata_raid_rw\n");
return ENOMEM;
}
/* setup request */
request->dev = dev;
request->timeout = 10;
request->retries = 0;
request->data = data;
request->bytecount = bcount;
request->transfersize = DEV_BSIZE;
request->u.ata.lba = lba;
request->u.ata.count = request->bytecount / DEV_BSIZE;
request->flags = flags;
if (flags & ATA_R_READ) {
if (atadev->mode >= ATA_DMA) {
request->u.ata.command = ATA_READ_DMA;
request->flags |= ATA_R_DMA;
}
else
request->u.ata.command = ATA_READ;
ata_queue_request(request);
}
else if (flags & ATA_R_WRITE) {
if (atadev->mode >= ATA_DMA) {
request->u.ata.command = ATA_WRITE_DMA;
request->flags |= ATA_R_DMA;
}
else
request->u.ata.command = ATA_WRITE;
ata_queue_request(request);
}
else {
device_printf(dev, "FAILURE - unknown IO operation\n");
request->result = EIO;
}
error = request->result;
ata_free_request(request);
return error;
}
/*
* module handeling
*/
static int
ata_raid_subdisk_probe(device_t dev)
{
device_quiet(dev);
return 0;
}
static int
ata_raid_subdisk_attach(device_t dev)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
int volume;
for (volume = 0; volume < MAX_VOLUMES; volume++) {
ars->raid[volume] = NULL;
ars->disk_number[volume] = -1;
}
ata_raid_read_metadata(dev);
return 0;
}
static int
ata_raid_subdisk_detach(device_t dev)
{
struct ata_raid_subdisk *ars = device_get_softc(dev);
int volume;
for (volume = 0; volume < MAX_VOLUMES; volume++) {
if (ars->raid[volume]) {
ars->raid[volume]->disks[ars->disk_number[volume]].flags &=
~(AR_DF_PRESENT | AR_DF_ONLINE);
ars->raid[volume]->disks[ars->disk_number[volume]].dev = NULL;
if (mtx_initialized(&ars->raid[volume]->lock))
ata_raid_config_changed(ars->raid[volume], 1);
ars->raid[volume] = NULL;
ars->disk_number[volume] = -1;
}
}
return 0;
}
static device_method_t ata_raid_sub_methods[] = {
/* device interface */
DEVMETHOD(device_probe, ata_raid_subdisk_probe),
DEVMETHOD(device_attach, ata_raid_subdisk_attach),
DEVMETHOD(device_detach, ata_raid_subdisk_detach),
{ 0, 0 }
};
static driver_t ata_raid_sub_driver = {
"subdisk",
ata_raid_sub_methods,
sizeof(struct ata_raid_subdisk)
};
DRIVER_MODULE(subdisk, ad, ata_raid_sub_driver, ata_raid_sub_devclass, NULL, NULL);
static int
ata_raid_module_event_handler(module_t mod, int what, void *arg)
{
int i;
switch (what) {
case MOD_LOAD:
if (testing || bootverbose)
printf("ATA PseudoRAID loaded\n");
#if 0
/* setup table to hold metadata for all ATA PseudoRAID arrays */
ata_raid_arrays = malloc(sizeof(struct ar_soft *) * MAX_ARRAYS,
M_AR, M_NOWAIT | M_ZERO);
if (!ata_raid_arrays) {
printf("ataraid: no memory for metadata storage\n");
return ENOMEM;
}
#endif
/* attach found PseudoRAID arrays */
for (i = 0; i < MAX_ARRAYS; i++) {
struct ar_softc *rdp = ata_raid_arrays[i];
if (!rdp || !rdp->format)
continue;
if (testing || bootverbose)
ata_raid_print_meta(rdp);
ata_raid_attach(rdp, 0);
}
ata_raid_ioctl_func = ata_raid_ioctl;
return 0;
case MOD_UNLOAD:
/* detach found PseudoRAID arrays */
for (i = 0; i < MAX_ARRAYS; i++) {
struct ar_softc *rdp = ata_raid_arrays[i];
if (!rdp || !rdp->status)
continue;
if (mtx_initialized(&rdp->lock))
mtx_destroy(&rdp->lock);
if (rdp->disk)
disk_destroy(rdp->disk);
}
if (testing || bootverbose)
printf("ATA PseudoRAID unloaded\n");
#if 0
free(ata_raid_arrays, M_AR);
#endif
ata_raid_ioctl_func = NULL;
return 0;
default:
return EOPNOTSUPP;
}
}
static moduledata_t ata_raid_moduledata =
{ "ataraid", ata_raid_module_event_handler, NULL };
DECLARE_MODULE(ata, ata_raid_moduledata, SI_SUB_RAID, SI_ORDER_FIRST);
MODULE_VERSION(ataraid, 1);
MODULE_DEPEND(ataraid, ata, 1, 1, 1);
MODULE_DEPEND(ataraid, ad, 1, 1, 1);
static char *
ata_raid_format(struct ar_softc *rdp)
{
switch (rdp->format) {
case AR_F_FREEBSD_RAID: return "FreeBSD PseudoRAID";
case AR_F_ADAPTEC_RAID: return "Adaptec HostRAID";
case AR_F_HPTV2_RAID: return "HighPoint v2 RocketRAID";
case AR_F_HPTV3_RAID: return "HighPoint v3 RocketRAID";
case AR_F_INTEL_RAID: return "Intel MatrixRAID";
case AR_F_ITE_RAID: return "Integrated Technology Express";
case AR_F_JMICRON_RAID: return "JMicron Technology Corp";
case AR_F_LSIV2_RAID: return "LSILogic v2 MegaRAID";
case AR_F_LSIV3_RAID: return "LSILogic v3 MegaRAID";
case AR_F_NVIDIA_RAID: return "nVidia MediaShield";
case AR_F_PROMISE_RAID: return "Promise Fasttrak";
case AR_F_SII_RAID: return "Silicon Image Medley";
case AR_F_SIS_RAID: return "Silicon Integrated Systems";
case AR_F_VIA_RAID: return "VIA Tech V-RAID";
default: return "UNKNOWN";
}
}
static char *
ata_raid_type(struct ar_softc *rdp)
{
switch (rdp->type) {
case AR_T_JBOD: return "JBOD";
case AR_T_SPAN: return "SPAN";
case AR_T_RAID0: return "RAID0";
case AR_T_RAID1: return "RAID1";
case AR_T_RAID3: return "RAID3";
case AR_T_RAID4: return "RAID4";
case AR_T_RAID5: return "RAID5";
case AR_T_RAID01: return "RAID0+1";
default: return "UNKNOWN";
}
}
static char *
ata_raid_flags(struct ar_softc *rdp)
{
switch (rdp->status & (AR_S_READY | AR_S_DEGRADED | AR_S_REBUILDING)) {
case AR_S_READY: return "READY";
case AR_S_READY | AR_S_DEGRADED: return "DEGRADED";
case AR_S_READY | AR_S_REBUILDING:
case AR_S_READY | AR_S_DEGRADED | AR_S_REBUILDING: return "REBUILDING";
default: return "BROKEN";
}
}
/* debugging gunk */
static void
ata_raid_print_meta(struct ar_softc *raid)
{
int i;
printf("********** ATA PseudoRAID ar%d Metadata **********\n", raid->lun);
printf("=================================================\n");
printf("format %s\n", ata_raid_format(raid));
printf("type %s\n", ata_raid_type(raid));
printf("flags 0x%02x %b\n", raid->status, raid->status,
"\20\3REBUILDING\2DEGRADED\1READY\n");
printf("magic_0 0x%016jx\n", raid->magic_0);
printf("magic_1 0x%016jx\n",raid->magic_1);
printf("generation %u\n", raid->generation);
printf("total_sectors %ju\n", raid->total_sectors);
printf("offset_sectors %ju\n", raid->offset_sectors);
printf("heads %u\n", raid->heads);
printf("sectors %u\n", raid->sectors);
printf("cylinders %u\n", raid->cylinders);
printf("width %u\n", raid->width);
printf("interleave %u\n", raid->interleave);
printf("total_disks %u\n", raid->total_disks);
for (i = 0; i < raid->total_disks; i++) {
printf(" disk %d: flags = 0x%02x %b\n", i, raid->disks[i].flags,
raid->disks[i].flags, "\20\4ONLINE\3SPARE\2ASSIGNED\1PRESENT\n");
if (raid->disks[i].dev) {
printf(" ");
device_printf(raid->disks[i].dev, " sectors %jd\n",
raid->disks[i].sectors);
}
}
printf("=================================================\n");
}
static char *
ata_raid_adaptec_type(int type)
{
static char buffer[16];
switch (type) {
case ADP_T_RAID0: return "RAID0";
case ADP_T_RAID1: return "RAID1";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_adaptec_print_meta(struct adaptec_raid_conf *meta)
{
int i;
printf("********* ATA Adaptec HostRAID Metadata *********\n");
printf("magic_0 <0x%08x>\n", be32toh(meta->magic_0));
printf("generation 0x%08x\n", be32toh(meta->generation));
printf("dummy_0 0x%04x\n", be16toh(meta->dummy_0));
printf("total_configs %u\n", be16toh(meta->total_configs));
printf("dummy_1 0x%04x\n", be16toh(meta->dummy_1));
printf("checksum 0x%04x\n", be16toh(meta->checksum));
printf("dummy_2 0x%08x\n", be32toh(meta->dummy_2));
printf("dummy_3 0x%08x\n", be32toh(meta->dummy_3));
printf("flags 0x%08x\n", be32toh(meta->flags));
printf("timestamp 0x%08x\n", be32toh(meta->timestamp));
printf("dummy_4 0x%08x 0x%08x 0x%08x 0x%08x\n",
be32toh(meta->dummy_4[0]), be32toh(meta->dummy_4[1]),
be32toh(meta->dummy_4[2]), be32toh(meta->dummy_4[3]));
printf("dummy_5 0x%08x 0x%08x 0x%08x 0x%08x\n",
be32toh(meta->dummy_5[0]), be32toh(meta->dummy_5[1]),
be32toh(meta->dummy_5[2]), be32toh(meta->dummy_5[3]));
for (i = 0; i < be16toh(meta->total_configs); i++) {
printf(" %d total_disks %u\n", i,
be16toh(meta->configs[i].disk_number));
printf(" %d generation %u\n", i,
be16toh(meta->configs[i].generation));
printf(" %d magic_0 0x%08x\n", i,
be32toh(meta->configs[i].magic_0));
printf(" %d dummy_0 0x%02x\n", i, meta->configs[i].dummy_0);
printf(" %d type %s\n", i,
ata_raid_adaptec_type(meta->configs[i].type));
printf(" %d dummy_1 0x%02x\n", i, meta->configs[i].dummy_1);
printf(" %d flags %d\n", i,
be32toh(meta->configs[i].flags));
printf(" %d dummy_2 0x%02x\n", i, meta->configs[i].dummy_2);
printf(" %d dummy_3 0x%02x\n", i, meta->configs[i].dummy_3);
printf(" %d dummy_4 0x%02x\n", i, meta->configs[i].dummy_4);
printf(" %d dummy_5 0x%02x\n", i, meta->configs[i].dummy_5);
printf(" %d disk_number %u\n", i,
be32toh(meta->configs[i].disk_number));
printf(" %d dummy_6 0x%08x\n", i,
be32toh(meta->configs[i].dummy_6));
printf(" %d sectors %u\n", i,
be32toh(meta->configs[i].sectors));
printf(" %d stripe_shift %u\n", i,
be16toh(meta->configs[i].stripe_shift));
printf(" %d dummy_7 0x%08x\n", i,
be32toh(meta->configs[i].dummy_7));
printf(" %d dummy_8 0x%08x 0x%08x 0x%08x 0x%08x\n", i,
be32toh(meta->configs[i].dummy_8[0]),
be32toh(meta->configs[i].dummy_8[1]),
be32toh(meta->configs[i].dummy_8[2]),
be32toh(meta->configs[i].dummy_8[3]));
printf(" %d name <%s>\n", i, meta->configs[i].name);
}
printf("magic_1 <0x%08x>\n", be32toh(meta->magic_1));
printf("magic_2 <0x%08x>\n", be32toh(meta->magic_2));
printf("magic_3 <0x%08x>\n", be32toh(meta->magic_3));
printf("magic_4 <0x%08x>\n", be32toh(meta->magic_4));
printf("=================================================\n");
}
static char *
ata_raid_hptv2_type(int type)
{
static char buffer[16];
switch (type) {
case HPTV2_T_RAID0: return "RAID0";
case HPTV2_T_RAID1: return "RAID1";
case HPTV2_T_RAID01_RAID0: return "RAID01_RAID0";
case HPTV2_T_SPAN: return "SPAN";
case HPTV2_T_RAID_3: return "RAID3";
case HPTV2_T_RAID_5: return "RAID5";
case HPTV2_T_JBOD: return "JBOD";
case HPTV2_T_RAID01_RAID1: return "RAID01_RAID1";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_hptv2_print_meta(struct hptv2_raid_conf *meta)
{
int i;
printf("****** ATA Highpoint V2 RocketRAID Metadata *****\n");
printf("magic 0x%08x\n", meta->magic);
printf("magic_0 0x%08x\n", meta->magic_0);
printf("magic_1 0x%08x\n", meta->magic_1);
printf("order 0x%08x\n", meta->order);
printf("array_width %u\n", meta->array_width);
printf("stripe_shift %u\n", meta->stripe_shift);
printf("type %s\n", ata_raid_hptv2_type(meta->type));
printf("disk_number %u\n", meta->disk_number);
printf("total_sectors %u\n", meta->total_sectors);
printf("disk_mode 0x%08x\n", meta->disk_mode);
printf("boot_mode 0x%08x\n", meta->boot_mode);
printf("boot_disk 0x%02x\n", meta->boot_disk);
printf("boot_protect 0x%02x\n", meta->boot_protect);
printf("log_entries 0x%02x\n", meta->error_log_entries);
printf("log_index 0x%02x\n", meta->error_log_index);
if (meta->error_log_entries) {
printf(" timestamp reason disk status sectors lba\n");
for (i = meta->error_log_index;
i < meta->error_log_index + meta->error_log_entries; i++)
printf(" 0x%08x 0x%02x 0x%02x 0x%02x 0x%02x 0x%08x\n",
meta->errorlog[i%32].timestamp,
meta->errorlog[i%32].reason,
meta->errorlog[i%32].disk, meta->errorlog[i%32].status,
meta->errorlog[i%32].sectors, meta->errorlog[i%32].lba);
}
printf("rebuild_lba 0x%08x\n", meta->rebuild_lba);
printf("dummy_1 0x%02x\n", meta->dummy_1);
printf("name_1 <%.15s>\n", meta->name_1);
printf("dummy_2 0x%02x\n", meta->dummy_2);
printf("name_2 <%.15s>\n", meta->name_2);
printf("=================================================\n");
}
static char *
ata_raid_hptv3_type(int type)
{
static char buffer[16];
switch (type) {
case HPTV3_T_SPARE: return "SPARE";
case HPTV3_T_JBOD: return "JBOD";
case HPTV3_T_SPAN: return "SPAN";
case HPTV3_T_RAID0: return "RAID0";
case HPTV3_T_RAID1: return "RAID1";
case HPTV3_T_RAID3: return "RAID3";
case HPTV3_T_RAID5: return "RAID5";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_hptv3_print_meta(struct hptv3_raid_conf *meta)
{
int i;
printf("****** ATA Highpoint V3 RocketRAID Metadata *****\n");
printf("magic 0x%08x\n", meta->magic);
printf("magic_0 0x%08x\n", meta->magic_0);
printf("checksum_0 0x%02x\n", meta->checksum_0);
printf("mode 0x%02x\n", meta->mode);
printf("user_mode 0x%02x\n", meta->user_mode);
printf("config_entries 0x%02x\n", meta->config_entries);
for (i = 0; i < meta->config_entries; i++) {
printf("config %d:\n", i);
printf(" total_sectors %ju\n",
meta->configs[0].total_sectors +
((u_int64_t)meta->configs_high[0].total_sectors << 32));
printf(" type %s\n",
ata_raid_hptv3_type(meta->configs[i].type));
printf(" total_disks %u\n", meta->configs[i].total_disks);
printf(" disk_number %u\n", meta->configs[i].disk_number);
printf(" stripe_shift %u\n", meta->configs[i].stripe_shift);
printf(" status %b\n", meta->configs[i].status,
"\20\2RAID5\1NEED_REBUILD\n");
printf(" critical_disks %u\n", meta->configs[i].critical_disks);
printf(" rebuild_lba %ju\n",
meta->configs_high[0].rebuild_lba +
((u_int64_t)meta->configs_high[0].rebuild_lba << 32));
}
printf("name <%.16s>\n", meta->name);
printf("timestamp 0x%08x\n", meta->timestamp);
printf("description <%.16s>\n", meta->description);
printf("creator <%.16s>\n", meta->creator);
printf("checksum_1 0x%02x\n", meta->checksum_1);
printf("dummy_0 0x%02x\n", meta->dummy_0);
printf("dummy_1 0x%02x\n", meta->dummy_1);
printf("flags %b\n", meta->flags,
"\20\4RCACHE\3WCACHE\2NCQ\1TCQ\n");
printf("=================================================\n");
}
static char *
ata_raid_intel_type(int type)
{
static char buffer[16];
switch (type) {
case INTEL_T_RAID0: return "RAID0";
case INTEL_T_RAID1: return "RAID1";
case INTEL_T_RAID5: return "RAID5";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_intel_print_meta(struct intel_raid_conf *meta)
{
struct intel_raid_mapping *map;
int i, j;
printf("********* ATA Intel MatrixRAID Metadata *********\n");
printf("intel_id <%.24s>\n", meta->intel_id);
printf("version <%.6s>\n", meta->version);
printf("checksum 0x%08x\n", meta->checksum);
printf("config_size 0x%08x\n", meta->config_size);
printf("config_id 0x%08x\n", meta->config_id);
printf("generation 0x%08x\n", meta->generation);
printf("total_disks %u\n", meta->total_disks);
printf("total_volumes %u\n", meta->total_volumes);
printf("DISK# serial disk_sectors disk_id flags\n");
for (i = 0; i < meta->total_disks; i++ ) {
printf(" %d <%.16s> %u 0x%08x 0x%08x\n", i,
meta->disk[i].serial, meta->disk[i].sectors,
meta->disk[i].id, meta->disk[i].flags);
}
map = (struct intel_raid_mapping *)&meta->disk[meta->total_disks];
for (j = 0; j < meta->total_volumes; j++) {
printf("name %.16s\n", map->name);
printf("total_sectors %ju\n", map->total_sectors);
printf("state %u\n", map->state);
printf("reserved %u\n", map->reserved);
printf("offset %u\n", map->offset);
printf("disk_sectors %u\n", map->disk_sectors);
printf("stripe_count %u\n", map->stripe_count);
printf("stripe_sectors %u\n", map->stripe_sectors);
printf("status %u\n", map->status);
printf("type %s\n", ata_raid_intel_type(map->type));
printf("total_disks %u\n", map->total_disks);
printf("magic[0] 0x%02x\n", map->magic[0]);
printf("magic[1] 0x%02x\n", map->magic[1]);
printf("magic[2] 0x%02x\n", map->magic[2]);
for (i = 0; i < map->total_disks; i++ ) {
printf(" disk %d at disk_idx 0x%08x\n", i, map->disk_idx[i]);
}
map = (struct intel_raid_mapping *)&map->disk_idx[map->total_disks];
}
printf("=================================================\n");
}
static char *
ata_raid_ite_type(int type)
{
static char buffer[16];
switch (type) {
case ITE_T_RAID0: return "RAID0";
case ITE_T_RAID1: return "RAID1";
case ITE_T_RAID01: return "RAID0+1";
case ITE_T_SPAN: return "SPAN";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_ite_print_meta(struct ite_raid_conf *meta)
{
printf("*** ATA Integrated Technology Express Metadata **\n");
printf("ite_id <%.40s>\n", meta->ite_id);
printf("timestamp_0 %04x/%02x/%02x %02x:%02x:%02x.%02x\n",
*((u_int16_t *)meta->timestamp_0), meta->timestamp_0[2],
meta->timestamp_0[3], meta->timestamp_0[5], meta->timestamp_0[4],
meta->timestamp_0[7], meta->timestamp_0[6]);
printf("total_sectors %jd\n", meta->total_sectors);
printf("type %s\n", ata_raid_ite_type(meta->type));
printf("stripe_1kblocks %u\n", meta->stripe_1kblocks);
printf("timestamp_1 %04x/%02x/%02x %02x:%02x:%02x.%02x\n",
*((u_int16_t *)meta->timestamp_1), meta->timestamp_1[2],
meta->timestamp_1[3], meta->timestamp_1[5], meta->timestamp_1[4],
meta->timestamp_1[7], meta->timestamp_1[6]);
printf("stripe_sectors %u\n", meta->stripe_sectors);
printf("array_width %u\n", meta->array_width);
printf("disk_number %u\n", meta->disk_number);
printf("disk_sectors %u\n", meta->disk_sectors);
printf("=================================================\n");
}
static char *
ata_raid_jmicron_type(int type)
{
static char buffer[16];
switch (type) {
case JM_T_RAID0: return "RAID0";
case JM_T_RAID1: return "RAID1";
case JM_T_RAID01: return "RAID0+1";
case JM_T_JBOD: return "JBOD";
case JM_T_RAID5: return "RAID5";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_jmicron_print_meta(struct jmicron_raid_conf *meta)
{
int i;
printf("***** ATA JMicron Technology Corp Metadata ******\n");
printf("signature %.2s\n", meta->signature);
printf("version 0x%04x\n", meta->version);
printf("checksum 0x%04x\n", meta->checksum);
printf("disk_id 0x%08x\n", meta->disk_id);
printf("offset 0x%08x\n", meta->offset);
printf("disk_sectors_low 0x%08x\n", meta->disk_sectors_low);
printf("disk_sectors_high 0x%08x\n", meta->disk_sectors_high);
printf("name %.16s\n", meta->name);
printf("type %s\n", ata_raid_jmicron_type(meta->type));
printf("stripe_shift %d\n", meta->stripe_shift);
printf("flags 0x%04x\n", meta->flags);
printf("spare:\n");
for (i=0; i < 2 && meta->spare[i]; i++)
printf(" %d 0x%08x\n", i, meta->spare[i]);
printf("disks:\n");
for (i=0; i < 8 && meta->disks[i]; i++)
printf(" %d 0x%08x\n", i, meta->disks[i]);
printf("=================================================\n");
}
static char *
ata_raid_lsiv2_type(int type)
{
static char buffer[16];
switch (type) {
case LSIV2_T_RAID0: return "RAID0";
case LSIV2_T_RAID1: return "RAID1";
case LSIV2_T_SPARE: return "SPARE";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_lsiv2_print_meta(struct lsiv2_raid_conf *meta)
{
int i;
printf("******* ATA LSILogic V2 MegaRAID Metadata *******\n");
printf("lsi_id <%s>\n", meta->lsi_id);
printf("dummy_0 0x%02x\n", meta->dummy_0);
printf("flags 0x%02x\n", meta->flags);
printf("version 0x%04x\n", meta->version);
printf("config_entries 0x%02x\n", meta->config_entries);
printf("raid_count 0x%02x\n", meta->raid_count);
printf("total_disks 0x%02x\n", meta->total_disks);
printf("dummy_1 0x%02x\n", meta->dummy_1);
printf("dummy_2 0x%04x\n", meta->dummy_2);
for (i = 0; i < meta->config_entries; i++) {
printf(" type %s\n",
ata_raid_lsiv2_type(meta->configs[i].raid.type));
printf(" dummy_0 %02x\n", meta->configs[i].raid.dummy_0);
printf(" stripe_sectors %u\n",
meta->configs[i].raid.stripe_sectors);
printf(" array_width %u\n",
meta->configs[i].raid.array_width);
printf(" disk_count %u\n", meta->configs[i].raid.disk_count);
printf(" config_offset %u\n",
meta->configs[i].raid.config_offset);
printf(" dummy_1 %u\n", meta->configs[i].raid.dummy_1);
printf(" flags %02x\n", meta->configs[i].raid.flags);
printf(" total_sectors %u\n",
meta->configs[i].raid.total_sectors);
}
printf("disk_number 0x%02x\n", meta->disk_number);
printf("raid_number 0x%02x\n", meta->raid_number);
printf("timestamp 0x%08x\n", meta->timestamp);
printf("=================================================\n");
}
static char *
ata_raid_lsiv3_type(int type)
{
static char buffer[16];
switch (type) {
case LSIV3_T_RAID0: return "RAID0";
case LSIV3_T_RAID1: return "RAID1";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_lsiv3_print_meta(struct lsiv3_raid_conf *meta)
{
int i;
printf("******* ATA LSILogic V3 MegaRAID Metadata *******\n");
printf("lsi_id <%.6s>\n", meta->lsi_id);
printf("dummy_0 0x%04x\n", meta->dummy_0);
printf("version 0x%04x\n", meta->version);
printf("dummy_0 0x%04x\n", meta->dummy_1);
printf("RAID configs:\n");
for (i = 0; i < 8; i++) {
if (meta->raid[i].total_disks) {
printf("%02d stripe_pages %u\n", i,
meta->raid[i].stripe_pages);
printf("%02d type %s\n", i,
ata_raid_lsiv3_type(meta->raid[i].type));
printf("%02d total_disks %u\n", i,
meta->raid[i].total_disks);
printf("%02d array_width %u\n", i,
meta->raid[i].array_width);
printf("%02d sectors %u\n", i, meta->raid[i].sectors);
printf("%02d offset %u\n", i, meta->raid[i].offset);
printf("%02d device 0x%02x\n", i,
meta->raid[i].device);
}
}
printf("DISK configs:\n");
for (i = 0; i < 6; i++) {
if (meta->disk[i].disk_sectors) {
printf("%02d disk_sectors %u\n", i,
meta->disk[i].disk_sectors);
printf("%02d flags 0x%02x\n", i, meta->disk[i].flags);
}
}
printf("device 0x%02x\n", meta->device);
printf("timestamp 0x%08x\n", meta->timestamp);
printf("checksum_1 0x%02x\n", meta->checksum_1);
printf("=================================================\n");
}
static char *
ata_raid_nvidia_type(int type)
{
static char buffer[16];
switch (type) {
case NV_T_SPAN: return "SPAN";
case NV_T_RAID0: return "RAID0";
case NV_T_RAID1: return "RAID1";
case NV_T_RAID3: return "RAID3";
case NV_T_RAID5: return "RAID5";
case NV_T_RAID01: return "RAID0+1";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_nvidia_print_meta(struct nvidia_raid_conf *meta)
{
printf("******** ATA nVidia MediaShield Metadata ********\n");
printf("nvidia_id <%.8s>\n", meta->nvidia_id);
printf("config_size %d\n", meta->config_size);
printf("checksum 0x%08x\n", meta->checksum);
printf("version 0x%04x\n", meta->version);
printf("disk_number %d\n", meta->disk_number);
printf("dummy_0 0x%02x\n", meta->dummy_0);
printf("total_sectors %d\n", meta->total_sectors);
printf("sectors_size %d\n", meta->sector_size);
printf("serial %.16s\n", meta->serial);
printf("revision %.4s\n", meta->revision);
printf("dummy_1 0x%08x\n", meta->dummy_1);
printf("magic_0 0x%08x\n", meta->magic_0);
printf("magic_1 0x%016jx\n", meta->magic_1);
printf("magic_2 0x%016jx\n", meta->magic_2);
printf("flags 0x%02x\n", meta->flags);
printf("array_width %d\n", meta->array_width);
printf("total_disks %d\n", meta->total_disks);
printf("dummy_2 0x%02x\n", meta->dummy_2);
printf("type %s\n", ata_raid_nvidia_type(meta->type));
printf("dummy_3 0x%04x\n", meta->dummy_3);
printf("stripe_sectors %d\n", meta->stripe_sectors);
printf("stripe_bytes %d\n", meta->stripe_bytes);
printf("stripe_shift %d\n", meta->stripe_shift);
printf("stripe_mask 0x%08x\n", meta->stripe_mask);
printf("stripe_sizesectors %d\n", meta->stripe_sizesectors);
printf("stripe_sizebytes %d\n", meta->stripe_sizebytes);
printf("rebuild_lba %d\n", meta->rebuild_lba);
printf("dummy_4 0x%08x\n", meta->dummy_4);
printf("dummy_5 0x%08x\n", meta->dummy_5);
printf("status 0x%08x\n", meta->status);
printf("=================================================\n");
}
static char *
ata_raid_promise_type(int type)
{
static char buffer[16];
switch (type) {
case PR_T_RAID0: return "RAID0";
case PR_T_RAID1: return "RAID1";
case PR_T_RAID3: return "RAID3";
case PR_T_RAID5: return "RAID5";
case PR_T_SPAN: return "SPAN";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_promise_print_meta(struct promise_raid_conf *meta)
{
int i;
printf("********* ATA Promise FastTrak Metadata *********\n");
printf("promise_id <%s>\n", meta->promise_id);
printf("dummy_0 0x%08x\n", meta->dummy_0);
printf("magic_0 0x%016jx\n", meta->magic_0);
printf("magic_1 0x%04x\n", meta->magic_1);
printf("magic_2 0x%08x\n", meta->magic_2);
printf("integrity 0x%08x %b\n", meta->raid.integrity,
meta->raid.integrity, "\20\10VALID\n" );
printf("flags 0x%02x %b\n",
meta->raid.flags, meta->raid.flags,
"\20\10READY\7DOWN\6REDIR\5DUPLICATE\4SPARE"
"\3ASSIGNED\2ONLINE\1VALID\n");
printf("disk_number %d\n", meta->raid.disk_number);
printf("channel 0x%02x\n", meta->raid.channel);
printf("device 0x%02x\n", meta->raid.device);
printf("magic_0 0x%016jx\n", meta->raid.magic_0);
printf("disk_offset %u\n", meta->raid.disk_offset);
printf("disk_sectors %u\n", meta->raid.disk_sectors);
printf("rebuild_lba 0x%08x\n", meta->raid.rebuild_lba);
printf("generation 0x%04x\n", meta->raid.generation);
printf("status 0x%02x %b\n",
meta->raid.status, meta->raid.status,
"\20\6MARKED\5DEGRADED\4READY\3INITED\2ONLINE\1VALID\n");
printf("type %s\n", ata_raid_promise_type(meta->raid.type));
printf("total_disks %u\n", meta->raid.total_disks);
printf("stripe_shift %u\n", meta->raid.stripe_shift);
printf("array_width %u\n", meta->raid.array_width);
printf("array_number %u\n", meta->raid.array_number);
printf("total_sectors %u\n", meta->raid.total_sectors);
printf("cylinders %u\n", meta->raid.cylinders);
printf("heads %u\n", meta->raid.heads);
printf("sectors %u\n", meta->raid.sectors);
printf("magic_1 0x%016jx\n", meta->raid.magic_1);
printf("DISK# flags dummy_0 channel device magic_0\n");
for (i = 0; i < 8; i++) {
printf(" %d %b 0x%02x 0x%02x 0x%02x ",
i, meta->raid.disk[i].flags,
"\20\10READY\7DOWN\6REDIR\5DUPLICATE\4SPARE"
"\3ASSIGNED\2ONLINE\1VALID\n", meta->raid.disk[i].dummy_0,
meta->raid.disk[i].channel, meta->raid.disk[i].device);
printf("0x%016jx\n", meta->raid.disk[i].magic_0);
}
printf("checksum 0x%08x\n", meta->checksum);
printf("=================================================\n");
}
static char *
ata_raid_sii_type(int type)
{
static char buffer[16];
switch (type) {
case SII_T_RAID0: return "RAID0";
case SII_T_RAID1: return "RAID1";
case SII_T_RAID01: return "RAID0+1";
case SII_T_SPARE: return "SPARE";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_sii_print_meta(struct sii_raid_conf *meta)
{
printf("******* ATA Silicon Image Medley Metadata *******\n");
printf("total_sectors %ju\n", meta->total_sectors);
printf("dummy_0 0x%04x\n", meta->dummy_0);
printf("dummy_1 0x%04x\n", meta->dummy_1);
printf("controller_pci_id 0x%08x\n", meta->controller_pci_id);
printf("version_minor 0x%04x\n", meta->version_minor);
printf("version_major 0x%04x\n", meta->version_major);
printf("timestamp 20%02x/%02x/%02x %02x:%02x:%02x\n",
meta->timestamp[5], meta->timestamp[4], meta->timestamp[3],
meta->timestamp[2], meta->timestamp[1], meta->timestamp[0]);
printf("stripe_sectors %u\n", meta->stripe_sectors);
printf("dummy_2 0x%04x\n", meta->dummy_2);
printf("disk_number %u\n", meta->disk_number);
printf("type %s\n", ata_raid_sii_type(meta->type));
printf("raid0_disks %u\n", meta->raid0_disks);
printf("raid0_ident %u\n", meta->raid0_ident);
printf("raid1_disks %u\n", meta->raid1_disks);
printf("raid1_ident %u\n", meta->raid1_ident);
printf("rebuild_lba %ju\n", meta->rebuild_lba);
printf("generation 0x%08x\n", meta->generation);
printf("status 0x%02x %b\n",
meta->status, meta->status,
"\20\1READY\n");
printf("base_raid1_position %02x\n", meta->base_raid1_position);
printf("base_raid0_position %02x\n", meta->base_raid0_position);
printf("position %02x\n", meta->position);
printf("dummy_3 %04x\n", meta->dummy_3);
printf("name <%.16s>\n", meta->name);
printf("checksum_0 0x%04x\n", meta->checksum_0);
printf("checksum_1 0x%04x\n", meta->checksum_1);
printf("=================================================\n");
}
static char *
ata_raid_sis_type(int type)
{
static char buffer[16];
switch (type) {
case SIS_T_JBOD: return "JBOD";
case SIS_T_RAID0: return "RAID0";
case SIS_T_RAID1: return "RAID1";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_sis_print_meta(struct sis_raid_conf *meta)
{
printf("**** ATA Silicon Integrated Systems Metadata ****\n");
printf("magic 0x%04x\n", meta->magic);
printf("disks 0x%02x\n", meta->disks);
printf("type %s\n",
ata_raid_sis_type(meta->type_total_disks & SIS_T_MASK));
printf("total_disks %u\n", meta->type_total_disks & SIS_D_MASK);
printf("dummy_0 0x%08x\n", meta->dummy_0);
printf("controller_pci_id 0x%08x\n", meta->controller_pci_id);
printf("stripe_sectors %u\n", meta->stripe_sectors);
printf("dummy_1 0x%04x\n", meta->dummy_1);
printf("timestamp 0x%08x\n", meta->timestamp);
printf("model %.40s\n", meta->model);
printf("disk_number %u\n", meta->disk_number);
printf("dummy_2 0x%02x 0x%02x 0x%02x\n",
meta->dummy_2[0], meta->dummy_2[1], meta->dummy_2[2]);
printf("=================================================\n");
}
static char *
ata_raid_via_type(int type)
{
static char buffer[16];
switch (type) {
case VIA_T_RAID0: return "RAID0";
case VIA_T_RAID1: return "RAID1";
case VIA_T_RAID5: return "RAID5";
case VIA_T_RAID01: return "RAID0+1";
case VIA_T_SPAN: return "SPAN";
default: sprintf(buffer, "UNKNOWN 0x%02x", type);
return buffer;
}
}
static void
ata_raid_via_print_meta(struct via_raid_conf *meta)
{
int i;
printf("*************** ATA VIA Metadata ****************\n");
printf("magic 0x%02x\n", meta->magic);
printf("dummy_0 0x%02x\n", meta->dummy_0);
printf("type %s\n",
ata_raid_via_type(meta->type & VIA_T_MASK));
printf("bootable %d\n", meta->type & VIA_T_BOOTABLE);
printf("unknown %d\n", meta->type & VIA_T_UNKNOWN);
printf("disk_index 0x%02x\n", meta->disk_index);
printf("stripe_layout 0x%02x\n", meta->stripe_layout);
printf(" stripe_disks %d\n", meta->stripe_layout & VIA_L_DISKS);
printf(" stripe_sectors %d\n",
0x08 << ((meta->stripe_layout & VIA_L_MASK) >> VIA_L_SHIFT));
printf("disk_sectors %ju\n", meta->disk_sectors);
printf("disk_id 0x%08x\n", meta->disk_id);
printf("DISK# disk_id\n");
for (i = 0; i < 8; i++) {
if (meta->disks[i])
printf(" %d 0x%08x\n", i, meta->disks[i]);
}
printf("checksum 0x%02x\n", meta->checksum);
printf("=================================================\n");
}