Date: Tue, 22 Mar 2011 16:54:22 +0000 (UTC) From: Alexander Motin <mav@FreeBSD.org> To: src-committers@freebsd.org, svn-src-projects@freebsd.org Subject: svn commit: r219874 - in projects/graid/head: sbin/geom/class/raid sys/conf sys/geom/raid sys/modules/geom/geom_raid Message-ID: <201103221654.p2MGsMGK065527@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mav Date: Tue Mar 22 16:54:21 2011 New Revision: 219874 URL: http://svn.freebsd.org/changeset/base/219874 Log: Add initial support for NVidia MediaShield metadata format. Added: projects/graid/head/sys/geom/raid/md_nvidia.c (contents, props changed) Modified: projects/graid/head/sbin/geom/class/raid/graid.8 projects/graid/head/sys/conf/files projects/graid/head/sys/modules/geom/geom_raid/Makefile Modified: projects/graid/head/sbin/geom/class/raid/graid.8 ============================================================================== --- projects/graid/head/sbin/geom/class/raid/graid.8 Tue Mar 22 16:21:11 2011 (r219873) +++ projects/graid/head/sbin/geom/class/raid/graid.8 Tue Mar 22 16:54:21 2011 (r219874) @@ -24,7 +24,7 @@ .\" .\" $FreeBSD$ .\" -.Dd March 18, 2011 +.Dd March 22, 2011 .Dt GRAID 8 .Os .Sh NAME @@ -214,6 +214,13 @@ Supports configurations: RAID0 (2+ disks RAID10 (4 disks), CONCAT (2+ disks). Configurations not supported by JMicron RAID BIOS, but enforceable on your own risk: RAID1 (3+ disks), RAID1E (3+ disks), RAID10 (6+ disks), RAID5 (3+ disks). +.It NVidia +The format used by NVidia MediaShield RAID BIOS. +Supports one volume per array. +Supports configurations: RAID0 (2+ disks), RAID1 (2 disks), +RAID5 (3+ disks), RAID10 (4+ disks), SINGLE (1 disk), CONCAT (2+ disks). +Configurations not supported by NVidia MediaShield RAID BIOS, but enforceable +on your own risk: RAID1 (3+ disks). .It Promise The format used by Promise and AMD/ATI RAID BIOSes and FreeBSD ataraid(4) driver. Modified: projects/graid/head/sys/conf/files ============================================================================== --- projects/graid/head/sys/conf/files Tue Mar 22 16:21:11 2011 (r219873) +++ projects/graid/head/sys/conf/files Tue Mar 22 16:54:21 2011 (r219874) @@ -2114,6 +2114,7 @@ geom/raid/g_raid_md_if.m optional geom_r geom/raid/g_raid_tr_if.m optional geom_raid geom/raid/md_intel.c optional geom_raid geom/raid/md_jmicron.c optional geom_raid +geom/raid/md_nvidia.c optional geom_raid geom/raid/md_promise.c optional geom_raid geom/raid/md_sii.c optional geom_raid geom/raid/tr_concat.c optional geom_raid Added: projects/graid/head/sys/geom/raid/md_nvidia.c ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ projects/graid/head/sys/geom/raid/md_nvidia.c Tue Mar 22 16:54:21 2011 (r219874) @@ -0,0 +1,1650 @@ +/*- + * Copyright (c) 2011 Alexander Motin <mav@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/bio.h> +#include <sys/endian.h> +#include <sys/kernel.h> +#include <sys/kobj.h> +#include <sys/limits.h> +#include <sys/lock.h> +#include <sys/malloc.h> +#include <sys/mutex.h> +#include <sys/systm.h> +#include <sys/taskqueue.h> +#include <geom/geom.h> +#include "geom/raid/g_raid.h" +#include "g_raid_md_if.h" + +static MALLOC_DEFINE(M_MD_NVIDIA, "md_nvidia_data", "GEOM_RAID NVidia metadata"); + +struct nvidia_raid_conf { + uint8_t nvidia_id[8]; +#define NVIDIA_MAGIC "NVIDIA " + + uint32_t config_size; + uint32_t checksum; + uint16_t version; + uint8_t disk_number; + uint8_t dummy_0; + uint32_t total_sectors; + uint32_t sector_size; + uint8_t name[16]; + uint8_t revision[4]; + uint32_t dummy_1; + + uint32_t magic_0; +#define NVIDIA_MAGIC0 0x00640044 + + uint64_t volume_id[2]; + uint8_t flags; + uint8_t array_width; + uint8_t total_disks; + uint8_t dummy_2; + uint16_t type; +#define NVIDIA_T_RAID0 0x00000080 +#define NVIDIA_T_RAID1 0x00000081 +#define NVIDIA_T_RAID3 0x00000083 +#define NVIDIA_T_RAID5 0x00000085 +#define NVIDIA_T_RAID10 0x00008180 +#define NVIDIA_T_CONCAT 0x000000ff + + uint16_t dummy_3; + uint32_t strip_sectors; + uint32_t strip_bytes; + uint32_t strip_shift; + uint32_t strip_mask; + uint32_t strip_sizesectors; + uint32_t strip_sizebytes; + uint32_t rebuild_lba; + uint32_t orig_type; + uint32_t orig_total_sectors; + uint32_t status; +#define NVIDIA_S_BOOTABLE 0x00000001 +#define NVIDIA_S_DEGRADED 0x00000002 + + uint32_t filler[98]; +} __packed; + +struct g_raid_md_nvidia_perdisk { + struct nvidia_raid_conf *pd_meta; + int pd_disk_pos; + off_t pd_disk_size; +}; + +struct g_raid_md_nvidia_object { + struct g_raid_md_object mdio_base; + uint64_t mdio_volume_id[2]; + uint8_t mdio_location; + uint32_t mdio_generation; + struct nvidia_raid_conf *mdio_meta; + struct callout mdio_start_co; /* STARTING state timer. */ + int mdio_total_disks; + int mdio_disks_present; + int mdio_started; + int mdio_incomplete; + struct root_hold_token *mdio_rootmount; /* Root mount delay token. */ +}; + +static g_raid_md_create_t g_raid_md_create_nvidia; +static g_raid_md_taste_t g_raid_md_taste_nvidia; +static g_raid_md_event_t g_raid_md_event_nvidia; +static g_raid_md_ctl_t g_raid_md_ctl_nvidia; +static g_raid_md_write_t g_raid_md_write_nvidia; +static g_raid_md_fail_disk_t g_raid_md_fail_disk_nvidia; +static g_raid_md_free_disk_t g_raid_md_free_disk_nvidia; +static g_raid_md_free_t g_raid_md_free_nvidia; + +static kobj_method_t g_raid_md_nvidia_methods[] = { + KOBJMETHOD(g_raid_md_create, g_raid_md_create_nvidia), + KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_nvidia), + KOBJMETHOD(g_raid_md_event, g_raid_md_event_nvidia), + KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_nvidia), + KOBJMETHOD(g_raid_md_write, g_raid_md_write_nvidia), + KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_nvidia), + KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_nvidia), + KOBJMETHOD(g_raid_md_free, g_raid_md_free_nvidia), + { 0, 0 } +}; + +static struct g_raid_md_class g_raid_md_nvidia_class = { + "NVidia", + g_raid_md_nvidia_methods, + sizeof(struct g_raid_md_nvidia_object), + .mdc_priority = 100 +}; + +static void +g_raid_md_nvidia_print(struct nvidia_raid_conf *meta) +{ + + if (g_raid_debug < 1) + return; + + printf("********* ATA NVidia RAID Metadata *********\n"); + printf("nvidia_id <%.8s>\n", meta->nvidia_id); + printf("config_size 0x%08x\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 %u\n", meta->total_sectors); + printf("sector_size %u\n", meta->sector_size); + printf("name <%.16s>\n", meta->name); + printf("revision 0x%02x%02x%02x%02x\n", + meta->revision[0], meta->revision[1], + meta->revision[2], meta->revision[3]); + printf("dummy_1 0x%08x\n", meta->dummy_1); + printf("magic_0 0x%08x\n", meta->magic_0); + printf("volume_id 0x%016jx%016jx\n", + meta->volume_id[0], meta->volume_id[1]); + printf("flags 0x%02x\n", meta->flags); + printf("array_width %u\n", meta->array_width); + printf("total_disks %u\n", meta->total_disks); + printf("dummy_2 0x%02x\n", meta->dummy_2); + printf("type 0x%04x\n", meta->type); + printf("dummy_3 0x%04x\n", meta->dummy_3); + printf("strip_sectors %u\n", meta->strip_sectors); + printf("strip_bytes %u\n", meta->strip_bytes); + printf("strip_shift %u\n", meta->strip_shift); + printf("strip_mask 0x%08x\n", meta->strip_mask); + printf("strip_sizesectors %u\n", meta->strip_sizesectors); + printf("strip_sizebytes %u\n", meta->strip_sizebytes); + printf("rebuild_lba %u\n", meta->rebuild_lba); + printf("orig_type 0x%04x\n", meta->orig_type); + printf("orig_total_sectors %u\n", meta->orig_total_sectors); + printf("status 0x%08x\n", meta->status); + printf("=================================================\n"); +} + +static struct nvidia_raid_conf * +nvidia_meta_copy(struct nvidia_raid_conf *meta) +{ + struct nvidia_raid_conf *nmeta; + + nmeta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK); + memcpy(nmeta, meta, sizeof(*meta)); + return (nmeta); +} + +static void +nvidia_meta_get_name(struct nvidia_raid_conf *meta, char *buf) +{ + int i; + + strncpy(buf, meta->name, 16); + buf[16] = 0; + for (i = 15; i >= 0; i--) { + if (buf[i] > 0x20) + break; + buf[i] = 0; + } +} + +#if 0 +static void +nvidia_meta_put_name(struct nvidia_raid_conf *meta, char *buf) +{ + + memset(meta->name, 0x20, 16); + memcpy(meta->name, buf, MIN(strlen(buf), 16)); +} +#endif + +static struct nvidia_raid_conf * +nvidia_meta_read(struct g_consumer *cp) +{ + struct g_provider *pp; + struct nvidia_raid_conf *meta; + char *buf; + int error, i; + uint32_t checksum, *ptr; + + pp = cp->provider; + + /* Read the anchor sector. */ + buf = g_read_data(cp, + pp->mediasize - 2 * pp->sectorsize, pp->sectorsize, &error); + if (buf == NULL) { + G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).", + pp->name, error); + return (NULL); + } + meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK); + memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize)); + g_free(buf); + + /* Check if this is an NVidia RAID struct */ + if (strncmp(meta->nvidia_id, NVIDIA_MAGIC, strlen(NVIDIA_MAGIC))) { + G_RAID_DEBUG(1, "NVidia signature check failed on %s", pp->name); + free(meta, M_MD_NVIDIA); + return (NULL); + } + if (meta->config_size > 128 || + meta->config_size < 10) { + G_RAID_DEBUG(1, "NVidia metadata size looks wrong: %d", + meta->config_size); + free(meta, M_MD_NVIDIA); + return (NULL); + } + /* Check metadata major version. */ +/* if (meta->version_major != 2) { + G_RAID_DEBUG(1, "NVidia version check failed on %s (%d.%d)", + pp->name, meta->version_major, meta->version_minor); + free(meta, M_MD_NVIDIA); + return (NULL); + } +*/ + /* Check metadata checksum. */ + for (checksum = 0, ptr = (uint32_t *)meta, + i = 0; i < meta->config_size; i++) + checksum += *ptr++; + if (checksum != 0) { + G_RAID_DEBUG(1, "NVidia checksum check failed on %s", pp->name); + free(meta, M_MD_NVIDIA); + return (NULL); + } + + /* Check raid type. */ + if (meta->type != NVIDIA_T_RAID0 && meta->type != NVIDIA_T_RAID1 && + meta->type != NVIDIA_T_RAID3 && meta->type != NVIDIA_T_RAID5 && + meta->type != NVIDIA_T_RAID10 && meta->type != NVIDIA_T_CONCAT) { + G_RAID_DEBUG(1, "NVidia unknown RAID level on %s (0x%02x)", + pp->name, meta->type); + free(meta, M_MD_NVIDIA); + return (NULL); + } + + return (meta); +} + +#if 0 +static int +nvidia_meta_write(struct g_consumer *cp, struct nvidia_raid_conf *meta) +{ + struct g_provider *pp; + char *buf; + int error, i; + uint16_t checksum, *ptr; + + pp = cp->provider; + + /* Recalculate checksum for case if metadata were changed. */ + meta->checksum = 0; + for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 159; i++) + checksum += *ptr++; + meta->checksum -= checksum; + + /* Create and fill buffer. */ + buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO); + memcpy(buf, meta, sizeof(*meta)); + + /* Write 4 copies of metadata. */ + for (i = 0; i < 4; i++) { + error = g_write_data(cp, + pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)), + buf, pp->sectorsize); + if (error != 0) { + G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).", + pp->name, error); + break; + } + } + + free(buf, M_MD_NVIDIA); + return (error); +} +#endif + +static int +nvidia_meta_erase(struct g_consumer *cp) +{ + struct g_provider *pp; + char *buf; + int error, i; + + pp = cp->provider; + buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO); + /* Write 4 copies of metadata. */ + for (i = 0; i < 4; i++) { + error = g_write_data(cp, + pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)), + buf, pp->sectorsize); + if (error != 0) { + G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).", + pp->name, error); + } + } + free(buf, M_MD_NVIDIA); + return (error); +} + +static int +nvidia_meta_write_spare(struct g_consumer *cp) +{ +#if 0 + struct nvidia_raid_conf *meta; + int error; + + meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK | M_ZERO); + meta->total_sectors = cp->provider->mediasize / + cp->provider->sectorsize - 0x800; + meta->vendor_id = 0x1095; + meta->version_minor = 0; + meta->version_major = 2; + meta->volume_id[0] = arc4random(); + meta->volume_id[1] = arc4random(); + meta->type = NVIDIA_T_SPARE; + meta->generation = 1; + meta->raid1_ident = 0xff; + meta->raid_location = arc4random(); + error = nvidia_meta_write(cp, meta); + free(meta, M_MD_NVIDIA); + return (error); +#endif + return (0); +} + +static struct g_raid_disk * +g_raid_md_nvidia_get_disk(struct g_raid_softc *sc, int id) +{ + struct g_raid_disk *disk; + struct g_raid_md_nvidia_perdisk *pd; + + TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { + pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data; + if (pd->pd_disk_pos == id) + break; + } + return (disk); +} + +static int +g_raid_md_nvidia_supported(int level, int qual, int disks, int force) +{ + + switch (level) { + case G_RAID_VOLUME_RL_RAID0: + if (disks < 1) + return (0); + if (!force && (disks < 2 || disks > 6)) + return (0); + break; + case G_RAID_VOLUME_RL_RAID1: + if (disks < 1) + return (0); + if (!force && (disks != 2)) + return (0); + break; + case G_RAID_VOLUME_RL_RAID1E: + if (disks < 2) + return (0); + if (disks % 2 != 0) + return (0); + if (!force && (disks < 4)) + return (0); + break; + case G_RAID_VOLUME_RL_SINGLE: + if (disks != 1) + return (0); + break; + case G_RAID_VOLUME_RL_CONCAT: + if (disks < 2) + return (0); + break; + case G_RAID_VOLUME_RL_RAID5: + if (disks < 3) + return (0); + break; + default: + return (0); + } + if (qual != G_RAID_VOLUME_RLQ_NONE) + return (0); + return (1); +} + +static int +g_raid_md_nvidia_start_disk(struct g_raid_disk *disk) +{ + struct g_raid_softc *sc; + struct g_raid_subdisk *sd, *tmpsd; + struct g_raid_disk *olddisk, *tmpdisk; + struct g_raid_md_object *md; + struct g_raid_md_nvidia_object *mdi; + struct g_raid_md_nvidia_perdisk *pd, *oldpd; + struct nvidia_raid_conf *meta; + int disk_pos, resurrection = 0; + + sc = disk->d_softc; + md = sc->sc_md; + mdi = (struct g_raid_md_nvidia_object *)md; + meta = mdi->mdio_meta; + pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data; + olddisk = NULL; + + /* Find disk position in metadata by it's serial. */ + if (pd->pd_meta != NULL) + disk_pos = pd->pd_meta->disk_number; + else + disk_pos = -3; + if (disk_pos < 0) { + G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk"); + /* If we are in the start process, that's all for now. */ + if (!mdi->mdio_started) + goto nofit; + /* + * If we have already started - try to get use of the disk. + * Try to replace OFFLINE disks first, then FAILED. + */ + TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) { + if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE && + tmpdisk->d_state != G_RAID_DISK_S_FAILED) + continue; + /* Make sure this disk is big enough. */ + TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) { + if (sd->sd_offset + sd->sd_size + 512 > + pd->pd_disk_size) { + G_RAID_DEBUG1(1, sc, + "Disk too small (%ju < %ju)", + pd->pd_disk_size, + sd->sd_offset + sd->sd_size + 512); + break; + } + } + if (sd != NULL) + continue; + if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) { + olddisk = tmpdisk; + break; + } else if (olddisk == NULL) + olddisk = tmpdisk; + } + if (olddisk == NULL) { +nofit: + if (disk_pos == -3 || pd->pd_disk_pos == -3) { + g_raid_change_disk_state(disk, + G_RAID_DISK_S_SPARE); + return (1); + } else { + g_raid_change_disk_state(disk, + G_RAID_DISK_S_STALE); + return (0); + } + } + oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data; + disk_pos = oldpd->pd_disk_pos; + resurrection = 1; + } + + if (olddisk == NULL) { + /* Find placeholder by position. */ + olddisk = g_raid_md_nvidia_get_disk(sc, disk_pos); + if (olddisk == NULL) + panic("No disk at position %d!", disk_pos); + if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) { + G_RAID_DEBUG1(1, sc, "More then one disk for pos %d", + disk_pos); + g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE); + return (0); + } + oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data; + } + + /* Replace failed disk or placeholder with new disk. */ + TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) { + TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next); + TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); + sd->sd_disk = disk; + } + oldpd->pd_disk_pos = -2; + pd->pd_disk_pos = disk_pos; + + /* If it was placeholder -- destroy it. */ + if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) { + g_raid_destroy_disk(olddisk); + } else { + /* Otherwise, make it STALE_FAILED. */ + g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED); + } + + /* Welcome the new disk. */ + if (resurrection) + g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); + else// if (pd->pd_meta->disk_status == NVIDIA_S_CURRENT || + //pd->pd_meta->disk_status == NVIDIA_S_REBUILD) + g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); +// else +// g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED); + TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { + + /* + * Different disks may have different sizes, + * in concat mode. Update from real disk size. + */ + if (meta->type == NVIDIA_T_CONCAT) + sd->sd_size = pd->pd_disk_size - 0x800 * 512; + + if (resurrection) { + /* New or ex-spare disk. */ + g_raid_change_subdisk_state(sd, + G_RAID_SUBDISK_S_NEW); +#if 0 + } else if (pd->pd_meta->disk_status == NVIDIA_S_REBUILD) { + /* Rebuilding disk. */ + g_raid_change_subdisk_state(sd, + G_RAID_SUBDISK_S_REBUILD); + if (pd->pd_meta->generation == meta->generation) + sd->sd_rebuild_pos = pd->pd_meta->rebuild_lba * 512; + else + sd->sd_rebuild_pos = 0; + } else if (pd->pd_meta->disk_status == NVIDIA_S_CURRENT) { + if (pd->pd_meta->raid_status == NVIDIA_S_ONLINE || + pd->pd_meta->generation != meta->generation) { + /* Dirty or resyncing disk. */ + g_raid_change_subdisk_state(sd, + G_RAID_SUBDISK_S_STALE); +#endif + } else { + /* Up to date disk. */ + g_raid_change_subdisk_state(sd, + G_RAID_SUBDISK_S_ACTIVE); +#if 0 + } + } else { + g_raid_change_subdisk_state(sd, + G_RAID_SUBDISK_S_FAILED); +#endif + } + g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, + G_RAID_EVENT_SUBDISK); + } + + /* Update status of our need for spare. */ + if (mdi->mdio_started) { + mdi->mdio_incomplete = + (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < + mdi->mdio_total_disks); + } + + return (resurrection); +} + +static void +g_disk_md_nvidia_retaste(void *arg, int pending) +{ + + G_RAID_DEBUG(1, "Array is not complete, trying to retaste."); + g_retaste(&g_raid_class); + free(arg, M_MD_NVIDIA); +} + +static void +g_raid_md_nvidia_refill(struct g_raid_softc *sc) +{ + struct g_raid_md_object *md; + struct g_raid_md_nvidia_object *mdi; + struct nvidia_raid_conf *meta; + struct g_raid_disk *disk; + struct task *task; + int update, na; + + md = sc->sc_md; + mdi = (struct g_raid_md_nvidia_object *)md; + meta = mdi->mdio_meta; + update = 0; + do { + /* Make sure we miss anything. */ + na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE); + if (na == mdi->mdio_total_disks) + break; + + G_RAID_DEBUG1(1, md->mdo_softc, + "Array is not complete (%d of %d), " + "trying to refill.", na, mdi->mdio_total_disks); + + /* Try to get use some of STALE disks. */ + TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { + if (disk->d_state == G_RAID_DISK_S_STALE) { + update += g_raid_md_nvidia_start_disk(disk); + if (disk->d_state == G_RAID_DISK_S_ACTIVE) + break; + } + } + if (disk != NULL) + continue; + + /* Try to get use some of SPARE disks. */ + TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { + if (disk->d_state == G_RAID_DISK_S_SPARE) { + update += g_raid_md_nvidia_start_disk(disk); + if (disk->d_state == G_RAID_DISK_S_ACTIVE) + break; + } + } + } while (disk != NULL); + + /* Write new metadata if we changed something. */ + if (update) { + g_raid_md_write_nvidia(md, NULL, NULL, NULL); + meta = mdi->mdio_meta; + } + + /* Update status of our need for spare. */ + mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < + mdi->mdio_total_disks); + + /* Request retaste hoping to find spare. */ + if (mdi->mdio_incomplete) { + task = malloc(sizeof(struct task), + M_MD_NVIDIA, M_WAITOK | M_ZERO); + TASK_INIT(task, 0, g_disk_md_nvidia_retaste, task); + taskqueue_enqueue(taskqueue_swi, task); + } +} + +static void +g_raid_md_nvidia_start(struct g_raid_softc *sc) +{ + struct g_raid_md_object *md; + struct g_raid_md_nvidia_object *mdi; + struct g_raid_md_nvidia_perdisk *pd; + struct nvidia_raid_conf *meta; + struct g_raid_volume *vol; + struct g_raid_subdisk *sd; + struct g_raid_disk *disk, *best; + off_t size; + int j, disk_pos; + uint32_t gendiff, bestgendiff; + char buf[17]; + + md = sc->sc_md; + mdi = (struct g_raid_md_nvidia_object *)md; + meta = mdi->mdio_meta; + + /* Create volumes and subdisks. */ + nvidia_meta_get_name(meta, buf); + vol = g_raid_create_volume(sc, buf, -1); + vol->v_mediasize = (off_t)meta->total_sectors * 512; + if (meta->type == NVIDIA_T_RAID0) { + vol->v_raid_level = G_RAID_VOLUME_RL_RAID0; + size = vol->v_mediasize / mdi->mdio_total_disks; + } else if (meta->type == NVIDIA_T_RAID1) { + vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; + size = vol->v_mediasize; + } else if (meta->type == NVIDIA_T_RAID10) { + vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; + size = vol->v_mediasize / (mdi->mdio_total_disks / 2); + } else if (meta->type == NVIDIA_T_CONCAT) { + if (mdi->mdio_total_disks == 1) + vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE; + else + vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT; + size = 0; + } else if (meta->type == NVIDIA_T_RAID5) { + vol->v_raid_level = G_RAID_VOLUME_RL_RAID5; + size = vol->v_mediasize / (mdi->mdio_total_disks - 1); + } else { + vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN; + size = 0; + } + vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; + vol->v_strip_size = meta->strip_sectors * 512; //ZZZ + vol->v_disks_count = mdi->mdio_total_disks; + vol->v_sectorsize = 512; //ZZZ + for (j = 0; j < vol->v_disks_count; j++) { + sd = &vol->v_subdisks[j]; + sd->sd_offset = 0; + sd->sd_size = size; + } + g_raid_start_volume(vol); + + /* Create disk placeholders to store data for later writing. */ + for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) { + pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO); + pd->pd_disk_pos = disk_pos; + disk = g_raid_create_disk(sc); + disk->d_md_data = (void *)pd; + disk->d_state = G_RAID_DISK_S_OFFLINE; + sd = &vol->v_subdisks[disk_pos]; + sd->sd_disk = disk; + TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); + } + + /* + * Make all disks found till the moment take their places + * in order of their generation numbers. + */ + do { + best = NULL; + bestgendiff = 0xffffffff; + TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { + if (disk->d_state != G_RAID_DISK_S_NONE) + continue; + pd = disk->d_md_data; + if (pd->pd_meta == NULL) + gendiff = 0xfffffffe; + else + gendiff = 0;//meta->generation - + //pd->pd_meta->generation; + if (gendiff < bestgendiff) { + best = disk; + bestgendiff = gendiff; + } + } + if (best != NULL) + g_raid_md_nvidia_start_disk(best); + } while (best != NULL); + + mdi->mdio_started = 1; + G_RAID_DEBUG1(0, sc, "Array started."); + g_raid_md_write_nvidia(md, NULL, NULL, NULL); + + /* Pickup any STALE/SPARE disks to refill array if needed. */ + g_raid_md_nvidia_refill(sc); + + g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME); + + callout_stop(&mdi->mdio_start_co); + G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount); + root_mount_rel(mdi->mdio_rootmount); + mdi->mdio_rootmount = NULL; +} + +static void +g_raid_md_nvidia_new_disk(struct g_raid_disk *disk) +{ + struct g_raid_softc *sc; + struct g_raid_md_object *md; + struct g_raid_md_nvidia_object *mdi; + struct nvidia_raid_conf *pdmeta; + struct g_raid_md_nvidia_perdisk *pd; + + sc = disk->d_softc; + md = sc->sc_md; + mdi = (struct g_raid_md_nvidia_object *)md; + pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data; + pdmeta = pd->pd_meta; + + if (mdi->mdio_started) { + if (g_raid_md_nvidia_start_disk(disk)) + g_raid_md_write_nvidia(md, NULL, NULL, NULL); + } else { + if (mdi->mdio_meta == NULL/* || + ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0*/) { + G_RAID_DEBUG1(1, sc, "Newer disk"); + if (mdi->mdio_meta != NULL) + free(mdi->mdio_meta, M_MD_NVIDIA); + mdi->mdio_meta = nvidia_meta_copy(pdmeta); +// mdi->mdio_generation = mdi->mdio_meta->generation; + mdi->mdio_total_disks = pdmeta->total_disks; + mdi->mdio_disks_present = 1; + } else /*if (pdmeta->generation == mdi->mdio_generation)*/ { + mdi->mdio_disks_present++; + G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)", + mdi->mdio_disks_present, + mdi->mdio_total_disks); +// } else { +// G_RAID_DEBUG1(1, sc, "Older disk"); + } + + /* If we collected all needed disks - start array. */ + if (mdi->mdio_disks_present == mdi->mdio_total_disks) + g_raid_md_nvidia_start(sc); + } +} + +static void +g_raid_nvidia_go(void *arg) +{ + struct g_raid_softc *sc; + struct g_raid_md_object *md; + struct g_raid_md_nvidia_object *mdi; + + sc = arg; + md = sc->sc_md; + mdi = (struct g_raid_md_nvidia_object *)md; + if (!mdi->mdio_started) { + G_RAID_DEBUG1(0, sc, "Force array start due to timeout."); + g_raid_event_send(sc, G_RAID_NODE_E_START, 0); + } +} + +static int +g_raid_md_create_nvidia(struct g_raid_md_object *md, struct g_class *mp, + struct g_geom **gp) +{ + struct g_raid_softc *sc; + struct g_raid_md_nvidia_object *mdi; + char name[32]; + + mdi = (struct g_raid_md_nvidia_object *)md; + arc4rand(&mdi->mdio_volume_id, 16, 0); + mdi->mdio_generation = 0; + snprintf(name, sizeof(name), "NVidia-%016jx%016jx", + mdi->mdio_volume_id[0], mdi->mdio_volume_id[1]); + sc = g_raid_create_node(mp, name, md); + if (sc == NULL) + return (G_RAID_MD_TASTE_FAIL); + md->mdo_softc = sc; + *gp = sc->sc_geom; + return (G_RAID_MD_TASTE_NEW); +} + +static int +g_raid_md_taste_nvidia(struct g_raid_md_object *md, struct g_class *mp, + struct g_consumer *cp, struct g_geom **gp) +{ + struct g_consumer *rcp; + struct g_provider *pp; + struct g_raid_md_nvidia_object *mdi, *mdi1; + struct g_raid_softc *sc; + struct g_raid_disk *disk; + struct nvidia_raid_conf *meta; + struct g_raid_md_nvidia_perdisk *pd; + struct g_geom *geom; + int error, disk_pos, result, spare, len; + char name[32]; + uint16_t vendor; + + G_RAID_DEBUG(1, "Tasting NVidia on %s", cp->provider->name); + mdi = (struct g_raid_md_nvidia_object *)md; + pp = cp->provider; + + /* Read metadata from device. */ + meta = NULL; + spare = 0; + vendor = 0xffff; + disk_pos = 0; + if (g_access(cp, 1, 0, 0) != 0) + return (G_RAID_MD_TASTE_FAIL); + g_topology_unlock(); + len = 2; + if (pp->geom->rank == 1) + g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor); + meta = nvidia_meta_read(cp); + g_topology_lock(); + g_access(cp, -1, 0, 0); + if (meta == NULL) { + if (g_raid_aggressive_spare) { + if (vendor == 0x1095) { + G_RAID_DEBUG(1, + "No NVidia metadata, forcing spare."); + spare = 2; + goto search; + } else { + G_RAID_DEBUG(1, + "NVidia vendor mismatch 0x%04x != 0x1095", + vendor); + } + } + return (G_RAID_MD_TASTE_FAIL); + } + + /* Check this disk position in obtained metadata. */ + disk_pos = meta->disk_number; + if (disk_pos == -1) { + G_RAID_DEBUG(1, "NVidia disk position not found"); + goto fail1; + } + + /* Metadata valid. Print it. */ + g_raid_md_nvidia_print(meta); + G_RAID_DEBUG(1, "NVidia disk position %d", disk_pos); + spare = 0;//(meta->type == NVIDIA_T_SPARE) ? 1 : 0; + +search: + /* Search for matching node. */ + sc = NULL; + mdi1 = NULL; + LIST_FOREACH(geom, &mp->geom, geom) { + sc = geom->softc; + if (sc == NULL) + continue; + if (sc->sc_stopping != 0) + continue; + if (sc->sc_md->mdo_class != md->mdo_class) + continue; + mdi1 = (struct g_raid_md_nvidia_object *)sc->sc_md; + if (spare) { + if (mdi1->mdio_incomplete) + break; + } else { + if (memcmp(&mdi1->mdio_volume_id, + &meta->volume_id, 16) == 0) + break; + } + } + + /* Found matching node. */ + if (geom != NULL) { + G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name); + result = G_RAID_MD_TASTE_EXISTING; + + } else if (spare) { /* Not found needy node -- left for later. */ + G_RAID_DEBUG(1, "Spare is not needed at this time"); + goto fail1; + + } else { /* Not found matching node -- create one. */ + result = G_RAID_MD_TASTE_NEW; + memcpy(&mdi->mdio_volume_id, &meta->volume_id, 16); + snprintf(name, sizeof(name), "NVidia-%016jx%016jx", + mdi->mdio_volume_id[0], mdi->mdio_volume_id[1]); + sc = g_raid_create_node(mp, name, md); + md->mdo_softc = sc; + geom = sc->sc_geom; + callout_init(&mdi->mdio_start_co, 1); *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201103221654.p2MGsMGK065527>