Introduce GEOM RAID3 class, i.e. kernel module, which implements RAID3

transformation and graid3(8) userland utility, which can be used for
configuration. No manual page yet, sorry.

Hardware provided by:	Daniel Seuffert
This commit is contained in:
Pawel Jakub Dawidek 2004-08-16 06:23:14 +00:00
parent c1fbc251cd
commit 2d1661a5b6
Notes: svn2git 2020-12-20 02:59:44 +00:00
svn path=/head/; revision=133808
6 changed files with 3913 additions and 0 deletions

View File

@ -0,0 +1,11 @@
# $FreeBSD$
.PATH: ${.CURDIR}/../../misc
CLASS= raid3
NOMAN= notyet
DPADD= ${LIBMD}
LDADD= -lmd
.include <bsd.lib.mk>

View File

@ -0,0 +1,340 @@
/*-
* Copyright (c) 2004 Pawel Jakub Dawidek <pjd@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.
* 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 AUTHORS AND CONTRIBUTORS ``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 AUTHORS OR CONTRIBUTORS 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 <sys/param.h>
#include <errno.h>
#include <paths.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <strings.h>
#include <assert.h>
#include <libgeom.h>
#include <geom/raid3/g_raid3.h>
#include <core/geom.h>
#include <misc/subr.h>
uint32_t lib_version = G_LIB_VERSION;
uint32_t version = G_RAID3_VERSION;
static void raid3_main(struct gctl_req *req, unsigned f);
static void raid3_clear(struct gctl_req *req);
static void raid3_dump(struct gctl_req *req);
static void raid3_label(struct gctl_req *req);
struct g_command class_commands[] = {
{ "clear", G_FLAG_VERBOSE, raid3_main, G_NULL_OPTS },
{ "configure", G_FLAG_VERBOSE, NULL,
{
{ 'a', "autosync", NULL, G_TYPE_NONE },
{ 'd', "dynamic", NULL, G_TYPE_NONE },
{ 'h', "hardcode", NULL, G_TYPE_NONE },
{ 'n', "noautosync", NULL, G_TYPE_NONE },
G_OPT_SENTINEL
}
},
{ "dump", 0, raid3_main, G_NULL_OPTS },
{ "insert", G_FLAG_VERBOSE, NULL,
{
{ 'h', "hardcode", NULL, G_TYPE_NONE },
{ 'n', "number", NULL, G_TYPE_NUMBER },
G_OPT_SENTINEL
}
},
{ "label", G_FLAG_VERBOSE, raid3_main,
{
{ 'h', "hardcode", NULL, G_TYPE_NONE },
{ 'n', "noautosync", NULL, G_TYPE_NONE },
G_OPT_SENTINEL
}
},
{ "rebuild", G_FLAG_VERBOSE, NULL, G_NULL_OPTS },
{ "remove", G_FLAG_VERBOSE, NULL,
{
{ 'n', "number", NULL, G_TYPE_NUMBER },
G_OPT_SENTINEL
}
},
{ "stop", G_FLAG_VERBOSE, NULL,
{
{ 'f', "force", NULL, G_TYPE_NONE },
G_OPT_SENTINEL
}
},
G_CMD_SENTINEL
};
static int verbose = 0;
void usage(const char *);
void
usage(const char *comm)
{
fprintf(stderr,
"usage: %s label [-hnv] name prov prov prov [prov [...]]\n"
" %s clear [-v] prov [prov [...]]\n"
" %s dump prov [prov [...]]\n"
" %s configure [-adhnv] name\n"
" %s rebuild [-v] name prov\n"
" %s insert [-hv] <-n number> name prov\n"
" %s remove [-v] <-n number> name\n"
" %s stop [-fv] name\n",
comm, comm, comm, comm, comm, comm, comm, comm);
exit(EXIT_FAILURE);
}
static void
raid3_main(struct gctl_req *req, unsigned flags)
{
const char *name;
if ((flags & G_FLAG_VERBOSE) != 0)
verbose = 1;
name = gctl_get_asciiparam(req, "verb");
if (name == NULL) {
gctl_error(req, "No '%s' argument.", "verb");
return;
}
if (strcmp(name, "label") == 0)
raid3_label(req);
else if (strcmp(name, "clear") == 0)
raid3_clear(req);
else if (strcmp(name, "dump") == 0)
raid3_dump(req);
else
gctl_error(req, "Unknown command: %s.", name);
}
static void
raid3_label(struct gctl_req *req)
{
struct g_raid3_metadata md;
u_char sector[512];
const char *str;
char param[16];
int *hardcode, *nargs, *noautosync, error, i;
unsigned sectorsize;
off_t mediasize;
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs < 4) {
gctl_error(req, "Too few arguments.");
return;
}
#ifndef BITCOUNT
#define BITCOUNT(x) (((BX_(x) + (BX_(x) >> 4)) & 0x0F0F0F0F) % 255)
#define BX_(x) ((x) - (((x) >> 1) & 0x77777777) - \
(((x) >> 2) & 0x33333333) - (((x) >> 3) & 0x11111111))
#endif
if (BITCOUNT(*nargs - 2) != 1) {
gctl_error(req, "Invalid number of components.");
return;
}
strlcpy(md.md_magic, G_RAID3_MAGIC, sizeof(md.md_magic));
md.md_version = G_RAID3_VERSION;
str = gctl_get_asciiparam(req, "arg0");
if (str == NULL) {
gctl_error(req, "No 'arg%u' argument.", 0);
return;
}
strlcpy(md.md_name, str, sizeof(md.md_name));
md.md_all = *nargs - 1;
md.md_mflags = 0;
md.md_dflags = 0;
md.md_syncid = 1;
md.md_sync_offset = 0;
noautosync = gctl_get_paraml(req, "noautosync", sizeof(*noautosync));
if (noautosync == NULL) {
gctl_error(req, "No '%s' argument.", "noautosync");
return;
}
if (*noautosync)
md.md_mflags |= G_RAID3_DEVICE_FLAG_NOAUTOSYNC;
hardcode = gctl_get_paraml(req, "hardcode", sizeof(*hardcode));
if (hardcode == NULL) {
gctl_error(req, "No '%s' argument.", "hardcode");
return;
}
/*
* Calculate sectorsize by finding least common multiple from
* sectorsizes of every disk and find the smallest mediasize.
*/
mediasize = 0;
sectorsize = 0;
for (i = 1; i < *nargs; i++) {
unsigned ssize;
off_t msize;
snprintf(param, sizeof(param), "arg%u", i);
str = gctl_get_asciiparam(req, param);
msize = g_get_mediasize(str);
ssize = g_get_sectorsize(str);
if (msize == 0 || ssize == 0) {
gctl_error(req, "Can't get informations about %s: %s.",
str, strerror(errno));
return;
}
msize -= ssize;
if (mediasize == 0 || (mediasize > 0 && msize < mediasize))
mediasize = msize;
if (sectorsize == 0)
sectorsize = ssize;
else
sectorsize = g_lcm(sectorsize, ssize);
}
md.md_mediasize = mediasize * (*nargs - 2);
md.md_sectorsize = sectorsize * (*nargs - 2);
/*
* Clear last sector first, to spoil all components if device exists.
*/
for (i = 1; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%u", i);
str = gctl_get_asciiparam(req, param);
error = g_metadata_clear(str, NULL);
if (error != 0) {
gctl_error(req, "Can't store metadata on %s: %s.", str,
strerror(error));
return;
}
}
/*
* Ok, store metadata (use disk number as priority).
*/
for (i = 1; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%u", i);
str = gctl_get_asciiparam(req, param);
md.md_no = i - 1;
if (!*hardcode)
bzero(md.md_provider, sizeof(md.md_provider));
else {
if (strncmp(str, _PATH_DEV, strlen(_PATH_DEV)) == 0)
str += strlen(_PATH_DEV);
strlcpy(md.md_provider, str, sizeof(md.md_provider));
}
raid3_metadata_encode(&md, sector);
error = g_metadata_store(str, sector, sizeof(sector));
if (error != 0) {
fprintf(stderr, "Can't store metadata on %s: %s.\n",
str, strerror(error));
gctl_error(req, "Not fully done.");
continue;
}
if (verbose)
printf("Metadata value stored on %s.\n", str);
}
}
static void
raid3_clear(struct gctl_req *req)
{
const char *name;
char param[16];
int *nargs, error, i;
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs < 1) {
gctl_error(req, "Too few arguments.");
return;
}
for (i = 0; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%u", i);
name = gctl_get_asciiparam(req, param);
error = g_metadata_clear(name, G_RAID3_MAGIC);
if (error != 0) {
fprintf(stderr, "Can't clear metadata on %s: %s.\n",
name, strerror(error));
gctl_error(req, "Not fully done.");
continue;
}
if (verbose)
printf("Metadata cleared on %s.\n", name);
}
}
static void
raid3_dump(struct gctl_req *req)
{
struct g_raid3_metadata md, tmpmd;
const char *name;
char param[16];
int *nargs, error, i;
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs < 1) {
gctl_error(req, "Too few arguments.");
return;
}
for (i = 0; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%u", i);
name = gctl_get_asciiparam(req, param);
error = g_metadata_read(name, (u_char *)&tmpmd, sizeof(tmpmd),
G_RAID3_MAGIC);
if (error != 0) {
fprintf(stderr, "Can't read metadata from %s: %s.\n",
name, strerror(error));
gctl_error(req, "Not fully done.");
continue;
}
if (raid3_metadata_decode((u_char *)&tmpmd, &md) != 0) {
fprintf(stderr, "MD5 hash mismatch for %s, skipping.\n",
name);
gctl_error(req, "Not fully done.");
continue;
}
printf("Metadata on %s:\n", name);
raid3_metadata_dump(&md);
printf("\n");
}
}

2763
sys/geom/raid3/g_raid3.c Normal file

File diff suppressed because it is too large Load Diff

306
sys/geom/raid3/g_raid3.h Normal file
View File

@ -0,0 +1,306 @@
/*-
* Copyright (c) 2004 Pawel Jakub Dawidek <pjd@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.
* 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 AUTHORS AND CONTRIBUTORS ``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 AUTHORS OR CONTRIBUTORS 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.
*
* $FreeBSD$
*/
#ifndef _G_RAID3_H_
#define _G_RAID3_H_
#include <sys/endian.h>
#include <sys/md5.h>
#define G_RAID3_CLASS_NAME "RAID3"
#define G_RAID3_MAGIC "GEOM::RAID3"
#define G_RAID3_VERSION 0
#define G_RAID3_DISK_FLAG_DIRTY 0x0000000000000001ULL
#define G_RAID3_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL
#define G_RAID3_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL
#define G_RAID3_DISK_FLAG_HARDCODED 0x0000000000000008ULL
#define G_RAID3_DISK_FLAG_MASK (G_RAID3_DISK_FLAG_DIRTY | \
G_RAID3_DISK_FLAG_SYNCHRONIZING | \
G_RAID3_DISK_FLAG_FORCE_SYNC)
#define G_RAID3_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL
#define G_RAID3_DEVICE_FLAG_MASK (G_RAID3_DEVICE_FLAG_NOAUTOSYNC)
#ifdef _KERNEL
extern u_int g_raid3_debug;
#define G_RAID3_DEBUG(lvl, ...) do { \
if (g_raid3_debug >= (lvl)) { \
printf("GEOM_RAID3"); \
if (g_raid3_debug > 0) \
printf("[%u]", lvl); \
printf(": "); \
printf(__VA_ARGS__); \
printf("\n"); \
} \
} while (0)
#define G_RAID3_LOGREQ(lvl, bp, ...) do { \
if (g_raid3_debug >= (lvl)) { \
printf("GEOM_RAID3"); \
if (g_raid3_debug > 0) \
printf("[%u]", lvl); \
printf(": "); \
printf(__VA_ARGS__); \
printf(" "); \
g_print_bio(bp); \
printf("\n"); \
} \
} while (0)
#define G_RAID3_MAX_IO_SIZE (DFLTPHYS * 2)
#define G_RAID3_BIO_CFLAG_REGULAR 0x01
#define G_RAID3_BIO_CFLAG_SYNC 0x02
#define G_RAID3_BIO_CFLAG_PARITY 0x04
#define G_RAID3_BIO_CFLAG_NODISK 0x08
#define G_RAID3_BIO_CFLAG_REGSYNC 0x10
#define G_RAID3_BIO_PFLAG_DEGRADED 0x01
#define G_RAID3_BIO_PFLAG_NOPARITY 0x02
/*
* Informations needed for synchronization.
*/
struct g_raid3_disk_sync {
struct g_consumer *ds_consumer; /* Consumer connected to our device. */
off_t ds_offset; /* Offset of next request to send. */
off_t ds_offset_done; /* Offset of already synchronized
region. */
u_int ds_syncid; /* Disk's synchronization ID. */
u_char *ds_data;
};
/*
* Informations needed for synchronization.
*/
struct g_raid3_device_sync {
struct g_geom *ds_geom; /* Synchronization geom. */
};
#define G_RAID3_DISK_STATE_NODISK 0
#define G_RAID3_DISK_STATE_NONE 1
#define G_RAID3_DISK_STATE_NEW 2
#define G_RAID3_DISK_STATE_ACTIVE 3
#define G_RAID3_DISK_STATE_STALE 4
#define G_RAID3_DISK_STATE_SYNCHRONIZING 5
#define G_RAID3_DISK_STATE_DISCONNECTED 6
#define G_RAID3_DISK_STATE_DESTROY 7
struct g_raid3_disk {
u_int d_no; /* Disk number. */
struct g_consumer *d_consumer; /* Consumer. */
struct g_raid3_softc *d_softc; /* Back-pointer to softc. */
int d_state; /* Disk state. */
uint64_t d_flags; /* Additional flags. */
struct g_raid3_disk_sync d_sync; /* Sync information. */
LIST_ENTRY(g_raid3_disk) d_next;
};
#define d_name d_consumer->provider->name
#define G_RAID3_EVENT_DONTWAIT 0x1
#define G_RAID3_EVENT_WAIT 0x2
#define G_RAID3_EVENT_DEVICE 0x4
#define G_RAID3_EVENT_DONE 0x8
struct g_raid3_event {
struct g_raid3_disk *e_disk;
int e_state;
int e_flags;
int e_error;
TAILQ_ENTRY(g_raid3_event) e_next;
};
#define G_RAID3_DEVICE_FLAG_DESTROY 0x0100000000000000ULL
#define G_RAID3_DEVICE_FLAG_WAIT 0x0200000000000000ULL
#define G_RAID3_DEVICE_STATE_STARTING 0
#define G_RAID3_DEVICE_STATE_DEGRADED 1
#define G_RAID3_DEVICE_STATE_COMPLETE 2
#define G_RAID3_BUMP_ON_FIRST_WRITE 1
#define G_RAID3_BUMP_IMMEDIATELY 2
struct g_raid3_softc {
u_int sc_state; /* Device state. */
uint64_t sc_mediasize; /* Device size. */
uint32_t sc_sectorsize; /* Sector size. */
uint64_t sc_flags; /* Additional flags. */
struct g_geom *sc_geom;
struct g_provider *sc_provider;
uint32_t sc_id; /* Device unique ID. */
struct bio_queue_head sc_queue;
struct mtx sc_queue_mtx;
struct proc *sc_worker;
struct g_raid3_disk *sc_disks;
u_int sc_ndisks; /* Number of disks. */
struct g_raid3_disk *sc_syncdisk;
uma_zone_t sc_zone_64k;
uma_zone_t sc_zone_16k;
uma_zone_t sc_zone_4k;
u_int sc_syncid; /* Synchronization ID. */
int sc_bump_syncid;
struct g_raid3_device_sync sc_sync;
TAILQ_HEAD(, g_raid3_event) sc_events;
struct mtx sc_events_mtx;
struct callout sc_callout;
};
#define sc_name sc_geom->name
const char *g_raid3_get_diskname(struct g_raid3_disk *disk);
u_int g_raid3_ndisks(struct g_raid3_softc *sc, int state);
int g_raid3_destroy(struct g_raid3_softc *sc, boolean_t force);
int g_raid3_event_send(void *arg, int state, int flags);
struct g_raid3_metadata;
void g_raid3_fill_metadata(struct g_raid3_disk *disk,
struct g_raid3_metadata *md);
int g_raid3_clear_metadata(struct g_raid3_disk *disk);
void g_raid3_update_metadata(struct g_raid3_disk *disk);
g_ctl_req_t g_raid3_config;
#endif /* _KERNEL */
struct g_raid3_metadata {
char md_magic[16]; /* Magic value. */
uint32_t md_version; /* Version number. */
char md_name[16]; /* Device name. */
uint32_t md_id; /* Device unique ID. */
uint16_t md_no; /* Component number. */
uint16_t md_all; /* Number of disks in device. */
uint32_t md_syncid; /* Synchronization ID. */
uint64_t md_mediasize; /* Size of whole device. */
uint32_t md_sectorsize; /* Sector size. */
uint64_t md_sync_offset; /* Synchronized offset. */
uint64_t md_mflags; /* Additional device flags. */
uint64_t md_dflags; /* Additional disk flags. */
char md_provider[16]; /* Hardcoded provider. */
u_char md_hash[16]; /* MD5 hash. */
};
static __inline void
raid3_metadata_encode(struct g_raid3_metadata *md, u_char *data)
{
MD5_CTX ctx;
bcopy(md->md_magic, data, 16);
le32enc(data + 16, md->md_version);
bcopy(md->md_name, data + 20, 16);
le32enc(data + 36, md->md_id);
le16enc(data + 40, md->md_no);
le16enc(data + 42, md->md_all);
le32enc(data + 44, md->md_syncid);
le64enc(data + 48, md->md_mediasize);
le32enc(data + 56, md->md_sectorsize);
le64enc(data + 60, md->md_sync_offset);
le64enc(data + 68, md->md_mflags);
le64enc(data + 76, md->md_dflags);
bcopy(md->md_provider, data + 84, 16);
MD5Init(&ctx);
MD5Update(&ctx, data, 100);
MD5Final(md->md_hash, &ctx);
bcopy(md->md_hash, data + 100, 16);
}
static __inline int
raid3_metadata_decode(const u_char *data, struct g_raid3_metadata *md)
{
MD5_CTX ctx;
bcopy(data, md->md_magic, 16);
md->md_version = le32dec(data + 16);
bcopy(data + 20, md->md_name, 16);
md->md_id = le32dec(data + 36);
md->md_no = le16dec(data + 40);
md->md_all = le16dec(data + 42);
md->md_syncid = le32dec(data + 44);
md->md_mediasize = le64dec(data + 48);
md->md_sectorsize = le32dec(data + 56);
md->md_sync_offset = le64dec(data + 60);
md->md_mflags = le64dec(data + 68);
md->md_dflags = le64dec(data + 76);
bcopy(data + 84, md->md_provider, 16);
bcopy(data + 100, md->md_hash, 16);
MD5Init(&ctx);
MD5Update(&ctx, data, 100);
MD5Final(md->md_hash, &ctx);
if (bcmp(md->md_hash, data + 100, 16) != 0)
return (EINVAL);
return (0);
}
static __inline void
raid3_metadata_dump(const struct g_raid3_metadata *md)
{
static const char hex[] = "0123456789abcdef";
char hash[16 * 2 + 1];
u_int i;
printf(" magic: %s\n", md->md_magic);
printf(" version: %u\n", (u_int)md->md_version);
printf(" name: %s\n", md->md_name);
printf(" id: %u\n", (u_int)md->md_id);
printf(" no: %u\n", (u_int)md->md_no);
printf(" all: %u\n", (u_int)md->md_all);
printf(" syncid: %u\n", (u_int)md->md_syncid);
printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize);
printf("sectorsize: %u\n", (u_int)md->md_sectorsize);
printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset);
printf(" mflags:");
if (md->md_mflags == 0)
printf(" NONE");
else {
if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0)
printf(" NOAUTOSYNC");
}
printf("\n");
printf(" dflags:");
if (md->md_dflags == 0)
printf(" NONE");
else {
if ((md->md_dflags & G_RAID3_DISK_FLAG_DIRTY) != 0)
printf(" DIRTY");
if ((md->md_dflags & G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0)
printf(" SYNCHRONIZING");
if ((md->md_dflags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0)
printf(" FORCE_SYNC");
}
printf("\n");
printf("hcprovider: %s\n", md->md_provider);
bzero(hash, sizeof(hash));
for (i = 0; i < 16; i++) {
hash[i * 2] = hex[md->md_hash[i] >> 4];
hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f];
}
printf(" MD5 hash: %s\n", hash);
}
#endif /* !_G_RAID3_H_ */

View File

@ -0,0 +1,484 @@
/*-
* Copyright (c) 2004 Pawel Jakub Dawidek <pjd@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.
* 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 AUTHORS AND CONTRIBUTORS ``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 AUTHORS OR CONTRIBUTORS 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 <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/module.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/bio.h>
#include <sys/sysctl.h>
#include <sys/malloc.h>
#include <sys/bitstring.h>
#include <vm/uma.h>
#include <machine/atomic.h>
#include <geom/geom.h>
#include <sys/proc.h>
#include <sys/kthread.h>
#include <geom/raid3/g_raid3.h>
static struct g_raid3_softc *
g_raid3_find_device(struct g_class *mp, const char *name)
{
struct g_raid3_softc *sc;
struct g_geom *gp;
g_topology_assert();
LIST_FOREACH(gp, &mp->geom, geom) {
sc = gp->softc;
if (sc == NULL)
continue;
if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROY) != 0)
continue;
if (strcmp(gp->name, name) == 0 ||
strcmp(sc->sc_name, name) == 0) {
return (sc);
}
}
return (NULL);
}
static struct g_raid3_disk *
g_raid3_find_disk(struct g_raid3_softc *sc, const char *name)
{
struct g_raid3_disk *disk;
u_int n;
g_topology_assert();
for (n = 0; n < sc->sc_ndisks; n++) {
disk = &sc->sc_disks[n];
if (disk->d_state == G_RAID3_DISK_STATE_NODISK)
continue;
if (disk->d_consumer == NULL)
continue;
if (disk->d_consumer->provider == NULL)
continue;
if (strcmp(disk->d_consumer->provider->name, name) == 0)
return (disk);
}
return (NULL);
}
static void
g_raid3_ctl_configure(struct gctl_req *req, struct g_class *mp)
{
struct g_raid3_softc *sc;
struct g_raid3_disk *disk;
const char *name;
int *nargs, *autosync, *noautosync, do_sync = 0;
u_int n;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (*nargs != 1) {
gctl_error(req, "Invalid number of arguments.");
return;
}
name = gctl_get_asciiparam(req, "arg0");
sc = g_raid3_find_device(mp, name);
if (sc == NULL) {
gctl_error(req, "No such device: %s.", name);
return;
}
if (g_raid3_ndisks(sc, -1) < sc->sc_ndisks) {
gctl_error(req, "Not all disks connected.");
return;
}
autosync = gctl_get_paraml(req, "autosync", sizeof(*autosync));
if (autosync == NULL) {
gctl_error(req, "No '%s' argument.", "autosync");
return;
}
noautosync = gctl_get_paraml(req, "noautosync", sizeof(*noautosync));
if (noautosync == NULL) {
gctl_error(req, "No '%s' argument.", "noautosync");
return;
}
if (!*autosync && !*noautosync) {
gctl_error(req, "Nothing has changed.");
return;
}
if (*autosync && *noautosync) {
gctl_error(req, "'%s' and '%s' specified.", "autosync",
"noautosync");
return;
}
if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0) {
if (*autosync) {
sc->sc_flags &= ~G_RAID3_DEVICE_FLAG_NOAUTOSYNC;
do_sync = 1;
}
} else {
if (*noautosync)
sc->sc_flags |= G_RAID3_DEVICE_FLAG_NOAUTOSYNC;
}
for (n = 0; n < sc->sc_ndisks; n++) {
disk = &sc->sc_disks[n];
if (do_sync) {
if (disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING)
disk->d_flags &= ~G_RAID3_DISK_FLAG_FORCE_SYNC;
}
g_raid3_update_metadata(disk);
if (do_sync) {
if (disk->d_state == G_RAID3_DISK_STATE_STALE) {
/*
* XXX: This is probably possible that this
* component will not be retasted.
*/
g_raid3_event_send(disk,
G_RAID3_DISK_STATE_DISCONNECTED,
G_RAID3_EVENT_DONTWAIT);
}
}
}
}
static void
g_raid3_ctl_rebuild(struct gctl_req *req, struct g_class *mp)
{
struct g_raid3_softc *sc;
struct g_raid3_disk *disk;
const char *name;
int *nargs;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs != 2) {
gctl_error(req, "Invalid number of arguments.");
return;
}
name = gctl_get_asciiparam(req, "arg0");
if (name == NULL) {
gctl_error(req, "No 'arg%u' argument.", 0);
return;
}
sc = g_raid3_find_device(mp, name);
if (sc == NULL) {
gctl_error(req, "No such device: %s.", name);
return;
}
name = gctl_get_asciiparam(req, "arg1");
if (name == NULL) {
gctl_error(req, "No 'arg%u' argument.", 1);
return;
}
disk = g_raid3_find_disk(sc, name);
if (disk == NULL) {
gctl_error(req, "No such provider: %s.", name);
return;
}
if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE &&
g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) < sc->sc_ndisks) {
gctl_error(req, "There is one stale disk already.", name);
return;
}
/*
* Do rebuild by resetting syncid and disconnecting disk.
* It'll be retasted, connected to the device and synchronized.
*/
disk->d_sync.ds_syncid = 0;
if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0)
disk->d_flags |= G_RAID3_DISK_FLAG_FORCE_SYNC;
g_raid3_update_metadata(disk);
g_raid3_event_send(disk, G_RAID3_DISK_STATE_DISCONNECTED,
G_RAID3_EVENT_WAIT);
}
static void
g_raid3_ctl_stop(struct gctl_req *req, struct g_class *mp)
{
struct g_raid3_softc *sc;
int *force, *nargs, error;
const char *name;
char param[16];
u_int i;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs < 1) {
gctl_error(req, "Missing device(s).");
return;
}
force = gctl_get_paraml(req, "force", sizeof(*force));
if (force == NULL) {
gctl_error(req, "No '%s' argument.", "force");
return;
}
for (i = 0; i < (u_int)*nargs; i++) {
snprintf(param, sizeof(param), "arg%u", i);
name = gctl_get_asciiparam(req, param);
if (name == NULL) {
gctl_error(req, "No 'arg%u' argument.", i);
return;
}
sc = g_raid3_find_device(mp, name);
if (sc == NULL) {
gctl_error(req, "No such device: %s.", name);
return;
}
error = g_raid3_destroy(sc, *force);
if (error != 0) {
gctl_error(req, "Cannot destroy device %s (error=%d).",
sc->sc_geom->name, error);
return;
}
}
}
static void
g_raid3_ctl_insert_orphan(struct g_consumer *cp)
{
KASSERT(1 == 0, ("%s called while inserting %s.", __func__,
cp->provider->name));
}
static void
g_raid3_ctl_insert(struct gctl_req *req, struct g_class *mp)
{
struct g_raid3_metadata md;
struct g_raid3_softc *sc;
struct g_raid3_disk *disk;
struct g_geom *gp;
struct g_provider *pp;
struct g_consumer *cp;
const char *name;
u_char *sector;
intmax_t *no;
int *hardcode, *nargs, error;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs != 2) {
gctl_error(req, "Invalid number of arguments.");
return;
}
name = gctl_get_asciiparam(req, "arg0");
if (name == NULL) {
gctl_error(req, "No 'arg%u' argument.", 0);
return;
}
sc = g_raid3_find_device(mp, name);
if (sc == NULL) {
gctl_error(req, "No such device: %s.", name);
return;
}
no = gctl_get_paraml(req, "number", sizeof(*no));
if (no == NULL) {
gctl_error(req, "No '%s' argument.", "no");
return;
}
if (*no >= sc->sc_ndisks) {
gctl_error(req, "Invalid component number.");
return;
}
hardcode = gctl_get_paraml(req, "hardcode", sizeof(*hardcode));
if (hardcode == NULL) {
gctl_error(req, "No '%s' argument.", "hardcode");
return;
}
disk = &sc->sc_disks[*no];
if (disk->d_state != G_RAID3_DISK_STATE_NODISK) {
gctl_error(req, "Component %u is already connected.", *no);
return;
}
name = gctl_get_asciiparam(req, "arg1");
if (name == NULL) {
gctl_error(req, "No 'arg%u' argument.", 1);
return;
}
pp = g_provider_by_name(name);
if (pp == NULL) {
gctl_error(req, "Invalid provider.");
return;
}
if (((sc->sc_sectorsize / (sc->sc_ndisks - 1)) % pp->sectorsize) != 0) {
gctl_error(req,
"Cannot insert provider %s, because of its sector size.",
pp->name);
return;
}
gp = g_new_geomf(mp, "raid3:insert");
gp->orphan = g_raid3_ctl_insert_orphan;
cp = g_new_consumer(gp);
error = g_attach(cp, pp);
if (error != 0) {
gctl_error(req, "Cannot attach to %s.", pp->name);
goto end;
}
error = g_access(cp, 0, 1, 1);
if (error != 0) {
gctl_error(req, "Cannot access %s.", pp->name);
goto end;
}
g_raid3_fill_metadata(disk, &md);
md.md_syncid = 0;
md.md_dflags = 0;
if (*hardcode)
strlcpy(md.md_provider, pp->name, sizeof(md.md_provider));
else
bzero(md.md_provider, sizeof(md.md_provider));
sector = g_malloc(pp->sectorsize, M_WAITOK);
raid3_metadata_encode(&md, sector);
g_topology_unlock();
error = g_write_data(cp, pp->mediasize - pp->sectorsize, sector,
pp->sectorsize);
g_topology_lock();
g_free(sector);
if (error != 0)
gctl_error(req, "Cannot store metadata on %s.", pp->name);
end:
if (gp != NULL) {
if (cp != NULL) {
if (cp->acw > 0)
g_access(cp, 0, -1, -1);
if (cp->provider != NULL)
g_detach(cp);
g_destroy_consumer(cp);
}
g_destroy_geom(gp);
}
}
static void
g_raid3_ctl_remove(struct gctl_req *req, struct g_class *mp)
{
struct g_raid3_softc *sc;
struct g_raid3_disk *disk;
const char *name;
intmax_t *no;
int *nargs;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument.", "nargs");
return;
}
if (*nargs != 1) {
gctl_error(req, "Invalid number of arguments.");
return;
}
name = gctl_get_asciiparam(req, "arg0");
if (name == NULL) {
gctl_error(req, "No 'arg%u' argument.", 0);
return;
}
sc = g_raid3_find_device(mp, name);
if (sc == NULL) {
gctl_error(req, "No such device: %s.", name);
return;
}
no = gctl_get_paraml(req, "number", sizeof(*no));
if (no == NULL) {
gctl_error(req, "No '%s' argument.", "no");
return;
}
if (*no >= sc->sc_ndisks) {
gctl_error(req, "Invalid component number.");
return;
}
disk = &sc->sc_disks[*no];
switch (disk->d_state) {
case G_RAID3_DISK_STATE_ACTIVE:
/*
* When replacing ACTIVE component, all the rest has to be also
* ACTIVE.
*/
if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) <
sc->sc_ndisks) {
gctl_error(req, "Cannot replace component number %u.",
*no);
return;
}
/* FALLTHROUGH */
case G_RAID3_DISK_STATE_STALE:
case G_RAID3_DISK_STATE_SYNCHRONIZING:
if (g_raid3_clear_metadata(disk) != 0) {
gctl_error(req, "Cannot clear metadata on %s.",
g_raid3_get_diskname(disk));
sc->sc_bump_syncid = G_RAID3_BUMP_IMMEDIATELY;
}
g_raid3_event_send(disk, G_RAID3_DISK_STATE_DISCONNECTED,
G_RAID3_EVENT_WAIT);
break;
case G_RAID3_DISK_STATE_NODISK:
break;
default:
gctl_error(req, "Cannot replace component number %u.", *no);
return;
}
}
void
g_raid3_config(struct gctl_req *req, struct g_class *mp, const char *verb)
{
uint32_t *version;
g_topology_assert();
version = gctl_get_paraml(req, "version", sizeof(*version));
if (version == NULL) {
gctl_error(req, "No '%s' argument.", "version");
return;
}
if (*version != G_RAID3_VERSION) {
gctl_error(req, "Userland and kernel parts are out of sync.");
return;
}
if (strcmp(verb, "configure") == 0)
g_raid3_ctl_configure(req, mp);
else if (strcmp(verb, "insert") == 0)
g_raid3_ctl_insert(req, mp);
else if (strcmp(verb, "rebuild") == 0)
g_raid3_ctl_rebuild(req, mp);
else if (strcmp(verb, "remove") == 0)
g_raid3_ctl_remove(req, mp);
else if (strcmp(verb, "stop") == 0)
g_raid3_ctl_stop(req, mp);
else
gctl_error(req, "Unknown verb.");
}

View File

@ -0,0 +1,9 @@
# $FreeBSD$
.PATH: ${.CURDIR}/../../../geom/raid3
KMOD= geom_raid3
SRCS= g_raid3.c
SRCS+= g_raid3_ctl.c
.include <bsd.kmod.mk>