getdevpath.3: Fix reference and sort.

[dragonfly.git] / sys / kern / subr_disk.c

/*
 * Copyright (c) 2003,2004,2009 The DragonFly Project.  All rights reserved.
 *
 * This code is derived from software contributed to The DragonFly Project
 * by Matthew Dillon <dillon@backplane.com>
 *
 * 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.
 * 3. Neither the name of The DragonFly Project nor the names of its
 *    contributors may be used to endorse or promote products derived
 *    from this software without specific, prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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
 * COPYRIGHT HOLDERS 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.
 *
 * ----------------------------------------------------------------------------
 * "THE BEER-WARE LICENSE" (Revision 42):
 * <phk@FreeBSD.ORG> wrote this file.  As long as you retain this notice you
 * can do whatever you want with this stuff. If we meet some day, and you think
 * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
 * ----------------------------------------------------------------------------
 *
 * Copyright (c) 1982, 1986, 1988, 1993
 *      The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * 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.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *      This product includes software developed by the University of
 *      California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
 *
 *      @(#)ufs_disksubr.c      8.5 (Berkeley) 1/21/94
 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
 * $DragonFly: src/sys/kern/subr_disk.c,v 1.40 2008/06/05 18:06:32 swildner Exp $
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/sysctl.h>
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/disklabel.h>
#include <sys/disklabel32.h>
#include <sys/disklabel64.h>
#include <sys/diskslice.h>
#include <sys/diskmbr.h>
#include <sys/disk.h>
#include <sys/malloc.h>
#include <sys/sysctl.h>
#include <machine/md_var.h>
#include <sys/ctype.h>
#include <sys/syslog.h>
#include <sys/device.h>
#include <sys/msgport.h>
#include <sys/msgport2.h>
#include <sys/buf2.h>
#include <sys/devfs.h>
#include <sys/thread.h>
#include <sys/thread2.h>

#include <sys/queue.h>
#include <sys/lock.h>

static MALLOC_DEFINE(M_DISK, "disk", "disk data");
static int disk_debug_enable = 0;

static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
static void disk_msg_core(void *);
static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
static void disk_probe(struct disk *dp, int reprobe);
static void _setdiskinfo(struct disk *disk, struct disk_info *info);
static void bioqwritereorder(struct bio_queue_head *bioq);
static void disk_cleanserial(char *serno);

static d_open_t diskopen;
static d_close_t diskclose;
static d_ioctl_t diskioctl;
static d_strategy_t diskstrategy;
static d_psize_t diskpsize;
static d_clone_t diskclone;
static d_dump_t diskdump;

static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
static struct lwkt_token disklist_token;

static struct dev_ops disk_ops = {
        { "disk", 0, D_DISK },
        .d_open = diskopen,
        .d_close = diskclose,
        .d_read = physread,
        .d_write = physwrite,
        .d_ioctl = diskioctl,
        .d_strategy = diskstrategy,
        .d_dump = diskdump,
        .d_psize = diskpsize,
        .d_clone = diskclone
};

static struct objcache  *disk_msg_cache;

struct objcache_malloc_args disk_msg_malloc_args = {
        sizeof(struct disk_msg), M_DISK };

static struct lwkt_port disk_dispose_port;
static struct lwkt_port disk_msg_port;

static int
disk_debug(int level, char *fmt, ...)
{
        __va_list ap;

        __va_start(ap, fmt);
        if (level <= disk_debug_enable)
                kvprintf(fmt, ap);
        __va_end(ap);

        return 0;
}

static int
disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
{
        struct disk_info *info = &dp->d_info;
        struct diskslice *sp = &dp->d_slice->dss_slices[slice];
        disklabel_ops_t ops;
        struct partinfo part;
        const char *msg;
        cdev_t ndev;
        int sno;
        u_int i;

        disk_debug(2,
                    "disk_probe_slice (begin): %s (%s)\n",
                        dev->si_name, dp->d_cdev->si_name);

        sno = slice ? slice - 1 : 0;

        ops = &disklabel32_ops;
        msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
        if (msg && !strcmp(msg, "no disk label")) {
                ops = &disklabel64_ops;
                msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
        }
        if (msg == NULL) {
                if (slice != WHOLE_DISK_SLICE)
                        ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
                else
                        sp->ds_reserved = 0;

                sp->ds_ops = ops;
                for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
                        ops->op_loadpartinfo(sp->ds_label, i, &part);
                        if (part.fstype) {
                                if (reprobe &&
                                    (ndev = devfs_find_device_by_name("%s%c",
                                                dev->si_name, 'a' + i))
                                ) {
                                        /*
                                         * Device already exists and
                                         * is still valid.
                                         */
                                        ndev->si_flags |= SI_REPROBE_TEST;
                                } else {
                                        ndev = make_dev_covering(&disk_ops, dp->d_rawdev,
                                                dkmakeminor(dkunit(dp->d_cdev),
                                                            slice, i),
                                                UID_ROOT, GID_OPERATOR, 0640,
                                                "%s%c", dev->si_name, 'a'+ i);
                                        ndev->si_disk = dp;
                                        if (dp->d_info.d_serialno) {
                                                make_dev_alias(ndev,
                                                    "serno/%s.s%d%c",
                                                    dp->d_info.d_serialno,
                                                    sno, 'a' + i);
                                        }
                                        ndev->si_flags |= SI_REPROBE_TEST;
                                }
                        }
                }
        } else if (info->d_dsflags & DSO_COMPATLABEL) {
                msg = NULL;
                if (sp->ds_size >= 0x100000000ULL)
                        ops = &disklabel64_ops;
                else
                        ops = &disklabel32_ops;
                sp->ds_label = ops->op_clone_label(info, sp);
        } else {
                if (sp->ds_type == DOSPTYP_386BSD /* XXX */) {
                        log(LOG_WARNING, "%s: cannot find label (%s)\n",
                            dev->si_name, msg);
                }
        }

        if (msg == NULL) {
                sp->ds_wlabel = FALSE;
        }

        return (msg ? EINVAL : 0);
}


static void
disk_probe(struct disk *dp, int reprobe)
{
        struct disk_info *info = &dp->d_info;
        cdev_t dev = dp->d_cdev;
        cdev_t ndev;
        int error, i, sno;
        struct diskslice *sp;

        KKASSERT (info->d_media_blksize != 0);

        dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
        disk_debug(1,
                    "disk_probe (begin): %s\n",
                        dp->d_cdev->si_name);

        error = mbrinit(dev, info, &(dp->d_slice));
        if (error)
                return;

        for (i = 0; i < dp->d_slice->dss_nslices; i++) {
                /*
                 * Ignore the whole-disk slice, it has already been created.
                 */
                if (i == WHOLE_DISK_SLICE)
                        continue;
                sp = &dp->d_slice->dss_slices[i];

                /*
                 * Handle s0.  s0 is a compatibility slice if there are no
                 * other slices and it has not otherwise been set up, else
                 * we ignore it.
                 */
                if (i == COMPATIBILITY_SLICE) {
                        sno = 0;
                        if (sp->ds_type == 0 &&
                            dp->d_slice->dss_nslices == BASE_SLICE) {
                                sp->ds_size = info->d_media_blocks;
                                sp->ds_reserved = 0;
                        }
                } else {
                        sno = i - 1;
                        sp->ds_reserved = 0;
                }

                /*
                 * Ignore 0-length slices
                 */
                if (sp->ds_size == 0)
                        continue;

                if (reprobe &&
                    (ndev = devfs_find_device_by_name("%ss%d",
                                                      dev->si_name, sno))) {
                        /*
                         * Device already exists and is still valid
                         */
                        ndev->si_flags |= SI_REPROBE_TEST;
                } else {
                        /*
                         * Else create new device
                         */
                        ndev = make_dev_covering(&disk_ops, dp->d_rawdev,
                                        dkmakewholeslice(dkunit(dev), i),
                                        UID_ROOT, GID_OPERATOR, 0640,
                                        "%ss%d", dev->si_name, sno);
                        if (dp->d_info.d_serialno) {
                                make_dev_alias(ndev, "serno/%s.s%d",
                                               dp->d_info.d_serialno, sno);
                        }
                        ndev->si_disk = dp;
                        ndev->si_flags |= SI_REPROBE_TEST;
                }
                sp->ds_dev = ndev;

                /*
                 * Probe appropriate slices for a disklabel
                 *
                 * XXX slice type 1 used by our gpt probe code.
                 * XXX slice type 0 used by mbr compat slice.
                 */
                if (sp->ds_type == DOSPTYP_386BSD || sp->ds_type == 0 ||
                        sp->ds_type == 1) {
                        if (dp->d_slice->dss_first_bsd_slice == 0)
                                dp->d_slice->dss_first_bsd_slice = i;
                        disk_probe_slice(dp, ndev, i, reprobe);
                }
        }
        disk_debug(1,
                    "disk_probe (end): %s\n",
                        dp->d_cdev->si_name);
}


static void
disk_msg_core(void *arg)
{
        struct disk     *dp;
        struct diskslice *sp;
        lwkt_tokref ilock;
        disk_msg_t msg;
        int run;

        lwkt_initport_thread(&disk_msg_port, curthread);
        wakeup(curthread);
        run = 1;

        while (run) {
                msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);

                switch (msg->hdr.u.ms_result) {
                case DISK_DISK_PROBE:
                        dp = (struct disk *)msg->load;
                        disk_debug(1,
                                    "DISK_DISK_PROBE: %s\n",
                                        dp->d_cdev->si_name);
                        disk_probe(dp, 0);
                        break;
                case DISK_DISK_DESTROY:
                        dp = (struct disk *)msg->load;
                        disk_debug(1,
                                    "DISK_DISK_DESTROY: %s\n",
                                        dp->d_cdev->si_name);
                        devfs_destroy_subnames(dp->d_cdev->si_name);
                        devfs_destroy_dev(dp->d_cdev);
                        lwkt_gettoken(&ilock, &disklist_token);
                        LIST_REMOVE(dp, d_list);
                        lwkt_reltoken(&ilock);
                        if (dp->d_info.d_serialno) {
                                kfree(dp->d_info.d_serialno, M_TEMP);
                                dp->d_info.d_serialno = NULL;
                        }
                        break;
                case DISK_UNPROBE:
                        dp = (struct disk *)msg->load;
                        disk_debug(1,
                                    "DISK_DISK_UNPROBE: %s\n",
                                        dp->d_cdev->si_name);
                        devfs_destroy_subnames(dp->d_cdev->si_name);
                        break;
                case DISK_SLICE_REPROBE:
                        dp = (struct disk *)msg->load;
                        sp = (struct diskslice *)msg->load2;
                        devfs_clr_subnames_flag(sp->ds_dev->si_name,
                                                SI_REPROBE_TEST);
                        disk_debug(1,
                                    "DISK_SLICE_REPROBE: %s\n",
                                    sp->ds_dev->si_name);
                        disk_probe_slice(dp, sp->ds_dev,
                                         dkslice(sp->ds_dev), 1);
                        devfs_destroy_subnames_without_flag(
                                        sp->ds_dev->si_name, SI_REPROBE_TEST);
                        break;
                case DISK_DISK_REPROBE:
                        dp = (struct disk *)msg->load;
                        devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
                        disk_debug(1,
                                    "DISK_DISK_REPROBE: %s\n",
                                    dp->d_cdev->si_name);
                        disk_probe(dp, 1);
                        devfs_destroy_subnames_without_flag(
                                        dp->d_cdev->si_name, SI_REPROBE_TEST);
                        break;
                case DISK_SYNC:
                        disk_debug(1, "DISK_SYNC\n");
                        break;
                default:
                        devfs_debug(DEVFS_DEBUG_WARNING,
                                    "disk_msg_core: unknown message "
                                    "received at core\n");
                        break;
                }
                lwkt_replymsg((lwkt_msg_t)msg, 0);
        }
        lwkt_exit();
}


/*
 * Acts as a message drain. Any message that is replied to here gets
 * destroyed and the memory freed.
 */
static void
disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
{
        objcache_put(disk_msg_cache, msg);
}


void
disk_msg_send(uint32_t cmd, void *load, void *load2)
{
        disk_msg_t disk_msg;
        lwkt_port_t port = &disk_msg_port;

        disk_msg = objcache_get(disk_msg_cache, M_WAITOK);

        lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);

        disk_msg->hdr.u.ms_result = cmd;
        disk_msg->load = load;
        disk_msg->load2 = load2;
        KKASSERT(port);
        lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
}

void
disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
{
        struct lwkt_port rep_port;
        disk_msg_t disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
        disk_msg_t      msg_incoming;
        lwkt_port_t port = &disk_msg_port;

        lwkt_initport_thread(&rep_port, curthread);
        lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);

        disk_msg->hdr.u.ms_result = cmd;
        disk_msg->load = load;
        disk_msg->load2 = load2;

        KKASSERT(port);
        lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
        msg_incoming = lwkt_waitport(&rep_port, 0);
}

/*
 * Create a raw device for the dev_ops template (which is returned).  Also
 * create a slice and unit managed disk and overload the user visible
 * device space with it.
 *
 * NOTE: The returned raw device is NOT a slice and unit managed device.
 * It is an actual raw device representing the raw disk as specified by
 * the passed dev_ops.  The disk layer not only returns such a raw device,
 * it also uses it internally when passing (modified) commands through.
 */
cdev_t
disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
{
        lwkt_tokref ilock;
        cdev_t rawdev;

        disk_debug(1,
                    "disk_create (begin): %s%d\n",
                        raw_ops->head.name, unit);

        rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
                            UID_ROOT, GID_OPERATOR, 0640,
                            "%s%d", raw_ops->head.name, unit);

        bzero(dp, sizeof(*dp));

        dp->d_rawdev = rawdev;
        dp->d_raw_ops = raw_ops;
        dp->d_dev_ops = &disk_ops;
        dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev,
                            dkmakewholedisk(unit),
                            UID_ROOT, GID_OPERATOR, 0640,
                            "%s%d", raw_ops->head.name, unit);

        dp->d_cdev->si_disk = dp;

        lwkt_gettoken(&ilock, &disklist_token);
        LIST_INSERT_HEAD(&disklist, dp, d_list);
        lwkt_reltoken(&ilock);

        disk_debug(1,
                    "disk_create (end): %s%d\n",
                        raw_ops->head.name, unit);

        return (dp->d_rawdev);
}


static void
_setdiskinfo(struct disk *disk, struct disk_info *info)
{
        char *oldserialno;

        oldserialno = disk->d_info.d_serialno;
        bcopy(info, &disk->d_info, sizeof(disk->d_info));
        info = &disk->d_info;

        disk_debug(1,
                    "_setdiskinfo: %s\n",
                        disk->d_cdev->si_name);

        /*
         * The serial number is duplicated so the caller can throw
         * their copy away.
         */
        if (info->d_serialno && info->d_serialno[0]) {
                info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
                disk_cleanserial(info->d_serialno);
                if (disk->d_cdev) {
                        make_dev_alias(disk->d_cdev, "serno/%s",
                                        info->d_serialno);
                }
        } else {
                info->d_serialno = NULL;
        }
        if (oldserialno)
                kfree(oldserialno, M_TEMP);

        /*
         * The caller may set d_media_size or d_media_blocks and we
         * calculate the other.
         */
        KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
        if (info->d_media_size == 0 && info->d_media_blocks) {
                info->d_media_size = (u_int64_t)info->d_media_blocks *
                                     info->d_media_blksize;
        } else if (info->d_media_size && info->d_media_blocks == 0 &&
                   info->d_media_blksize) {
                info->d_media_blocks = info->d_media_size /
                                       info->d_media_blksize;
        }

        /*
         * The si_* fields for rawdev are not set until after the
         * disk_create() call, so someone using the cooked version
         * of the raw device (i.e. da0s0) will not get the right
         * si_iosize_max unless we fix it up here.
         */
        if (disk->d_cdev && disk->d_rawdev &&
            disk->d_cdev->si_iosize_max == 0) {
                disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
                disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
                disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
        }
}

/*
 * Disk drivers must call this routine when media parameters are available
 * or have changed.
 */
void
disk_setdiskinfo(struct disk *disk, struct disk_info *info)
{
        _setdiskinfo(disk, info);
        disk_msg_send(DISK_DISK_PROBE, disk, NULL);
        disk_debug(1,
                    "disk_setdiskinfo: sent probe for %s\n",
                        disk->d_cdev->si_name);
}

void
disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
{
        _setdiskinfo(disk, info);
        disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
        disk_debug(1,
                    "disk_setdiskinfo_sync: sent probe for %s\n",
                        disk->d_cdev->si_name);
}

/*
 * This routine is called when an adapter detaches.  The higher level
 * managed disk device is destroyed while the lower level raw device is
 * released.
 */
void
disk_destroy(struct disk *disk)
{
        disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
        return;
}

int
disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
{
        struct partinfo pinfo;
        int error;

        bzero(&pinfo, sizeof(pinfo));
        error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0,
                           proc0.p_ucred, NULL);
        if (error)
                return (error);
        if (pinfo.media_blksize == 0)
                return (ENXIO);
        *count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
        if (dumplo64 < pinfo.reserved_blocks ||
            dumplo64 + *count > pinfo.media_blocks) {
                return (ENOSPC);
        }
        *blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
        *secsize = pinfo.media_blksize;
        return (0);
}

void
disk_unprobe(struct disk *disk)
{
        if (disk == NULL)
                return;

        disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
}

void
disk_invalidate (struct disk *disk)
{
        if (disk->d_slice)
                dsgone(&disk->d_slice);
}

struct disk *
disk_enumerate(struct disk *disk)
{
        struct disk *dp;
        lwkt_tokref ilock;

        lwkt_gettoken(&ilock, &disklist_token);
        if (!disk)
                dp = (LIST_FIRST(&disklist));
        else
                dp = (LIST_NEXT(disk, d_list));
        lwkt_reltoken(&ilock);

        return dp;
}

static
int
sysctl_disks(SYSCTL_HANDLER_ARGS)
{
        struct disk *disk;
        int error, first;

        disk = NULL;
        first = 1;

        while ((disk = disk_enumerate(disk))) {
                if (!first) {
                        error = SYSCTL_OUT(req, " ", 1);
                        if (error)
                                return error;
                } else {
                        first = 0;
                }
                error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
                                   strlen(disk->d_rawdev->si_name));
                if (error)
                        return error;
        }
        error = SYSCTL_OUT(req, "", 1);
        return error;
}

SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
    sysctl_disks, "A", "names of available disks");

/*
 * Open a disk device or partition.
 */
static
int
diskopen(struct dev_open_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct disk *dp;
        int error;

        /*
         * dp can't be NULL here XXX.
         *
         * d_slice will be NULL if setdiskinfo() has not been called yet.
         * setdiskinfo() is typically called whether the disk is present
         * or not (e.g. CD), but the base disk device is created first
         * and there may be a race.
         */
        dp = dev->si_disk;
        if (dp == NULL || dp->d_slice == NULL)
                return (ENXIO);
        error = 0;

        /*
         * Deal with open races
         */
        while (dp->d_flags & DISKFLAG_LOCK) {
                dp->d_flags |= DISKFLAG_WANTED;
                error = tsleep(dp, PCATCH, "diskopen", hz);
                if (error)
                        return (error);
        }
        dp->d_flags |= DISKFLAG_LOCK;

        /*
         * Open the underlying raw device.
         */
        if (!dsisopen(dp->d_slice)) {
#if 0
                if (!pdev->si_iosize_max)
                        pdev->si_iosize_max = dev->si_iosize_max;
#endif
                error = dev_dopen(dp->d_rawdev, ap->a_oflags,
                                  ap->a_devtype, ap->a_cred);
        }
#if 0
        /*
         * Inherit properties from the underlying device now that it is
         * open.
         */
        dev_dclone(dev);
#endif

        if (error)
                goto out;
        error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
                       &dp->d_slice, &dp->d_info);
        if (!dsisopen(dp->d_slice)) {
                dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
        }
out:
        dp->d_flags &= ~DISKFLAG_LOCK;
        if (dp->d_flags & DISKFLAG_WANTED) {
                dp->d_flags &= ~DISKFLAG_WANTED;
                wakeup(dp);
        }

        return(error);
}

/*
 * Close a disk device or partition
 */
static
int
diskclose(struct dev_close_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct disk *dp;
        int error;

        error = 0;
        dp = dev->si_disk;

        dsclose(dev, ap->a_devtype, dp->d_slice);
        if (!dsisopen(dp->d_slice)) {
                error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
        }
        return (error);
}

/*
 * First execute the ioctl on the disk device, and if it isn't supported
 * try running it on the backing device.
 */
static
int
diskioctl(struct dev_ioctl_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct disk *dp;
        int error;

        dp = dev->si_disk;
        if (dp == NULL)
                return (ENXIO);

        devfs_debug(DEVFS_DEBUG_DEBUG,
                    "diskioctl: cmd is: %x (name: %s)\n",
                    ap->a_cmd, dev->si_name);
        devfs_debug(DEVFS_DEBUG_DEBUG,
                    "diskioctl: &dp->d_slice is: %x, %x\n",
                    &dp->d_slice, dp->d_slice);

        error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
                        &dp->d_slice, &dp->d_info);

        if (error == ENOIOCTL) {
                error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
                                   ap->a_fflag, ap->a_cred, NULL);
        }
        return (error);
}

/*
 * Execute strategy routine
 */
static
int
diskstrategy(struct dev_strategy_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct bio *bio = ap->a_bio;
        struct bio *nbio;
        struct disk *dp;

        dp = dev->si_disk;

        if (dp == NULL) {
                bio->bio_buf->b_error = ENXIO;
                bio->bio_buf->b_flags |= B_ERROR;
                biodone(bio);
                return(0);
        }
        KKASSERT(dev->si_disk == dp);

        /*
         * The dscheck() function will also transform the slice relative
         * block number i.e. bio->bio_offset into a block number that can be
         * passed directly to the underlying raw device.  If dscheck()
         * returns NULL it will have handled the bio for us (e.g. EOF
         * or error due to being beyond the device size).
         */
        if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
                dev_dstrategy(dp->d_rawdev, nbio);
        } else {
                biodone(bio);
        }
        return(0);
}

/*
 * Return the partition size in ?blocks?
 */
static
int
diskpsize(struct dev_psize_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct disk *dp;

        dp = dev->si_disk;
        if (dp == NULL)
                return(ENODEV);
        ap->a_result = dssize(dev, &dp->d_slice);
        return(0);
}

/*
 * When new device entries are instantiated, make sure they inherit our
 * si_disk structure and block and iosize limits from the raw device.
 *
 * This routine is always called synchronously in the context of the
 * client.
 *
 * XXX The various io and block size constraints are not always initialized
 * properly by devices.
 */
static
int
diskclone(struct dev_clone_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct disk *dp;
        dp = dev->si_disk;

        KKASSERT(dp != NULL);
        dev->si_disk = dp;
        dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
        dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
        dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
        return(0);
}

int
diskdump(struct dev_dump_args *ap)
{
        cdev_t dev = ap->a_head.a_dev;
        struct disk *dp = dev->si_disk;
        int error;

        error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
        if (error == 0) {
                ap->a_head.a_dev = dp->d_rawdev;
                error = dev_doperate(&ap->a_head);
        }

        return(error);
}


SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
    0, sizeof(struct diskslices), "sizeof(struct diskslices)");

SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
    0, sizeof(struct disk), "sizeof(struct disk)");

/*
 * Reorder interval for burst write allowance and minor write
 * allowance.
 *
 * We always want to trickle some writes in to make use of the
 * disk's zone cache.  Bursting occurs on a longer interval and only
 * runningbufspace is well over the hirunningspace limit.
 */
int bioq_reorder_burst_interval = 60;   /* should be multiple of minor */
SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval,
           CTLFLAG_RW, &bioq_reorder_burst_interval, 0, "");
int bioq_reorder_minor_interval = 5;
SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval,
           CTLFLAG_RW, &bioq_reorder_minor_interval, 0, "");

int bioq_reorder_burst_bytes = 3000000;
SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes,
           CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, "");
int bioq_reorder_minor_bytes = 262144;
SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes,
           CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, "");


/*
 * Order I/Os.  Generally speaking this code is designed to make better
 * use of drive zone caches.  A drive zone cache can typically track linear
 * reads or writes for around 16 zones simultaniously.
 *
 * Read prioritization issues:  It is possible for hundreds of megabytes worth
 * of writes to be queued asynchronously.  This creates a huge bottleneck
 * for reads which reduce read bandwidth to a trickle.
 *
 * To solve this problem we generally reorder reads before writes.
 *
 * However, a large number of random reads can also starve writes and
 * make poor use of the drive zone cache so we allow writes to trickle
 * in every N reads.
 */
void
bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
{
        /*
         * The BIO wants to be ordered.  Adding to the tail also
         * causes transition to be set to NULL, forcing the ordering
         * of all prior I/O's.
         */
        if (bio->bio_buf->b_flags & B_ORDERED) {
                bioq_insert_tail(bioq, bio);
                return;
        }

        switch(bio->bio_buf->b_cmd) {
        case BUF_CMD_READ:
                if (bioq->transition) {
                        /*
                         * Insert before the first write.  Bleedover writes
                         * based on reorder intervals to prevent starvation.
                         */
                        TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act);
                        ++bioq->reorder;
                        if (bioq->reorder % bioq_reorder_minor_interval == 0) {
                                bioqwritereorder(bioq);
                                if (bioq->reorder >=
                                    bioq_reorder_burst_interval) {
                                        bioq->reorder = 0;
                                }
                        }
                } else {
                        /*
                         * No writes queued (or ordering was forced),
                         * insert at tail.
                         */
                        TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
                }
                break;
        case BUF_CMD_WRITE:
                /*
                 * Writes are always appended.  If no writes were previously
                 * queued or an ordered tail insertion occured the transition
                 * field will be NULL.
                 */
                TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
                if (bioq->transition == NULL)
                        bioq->transition = bio;
                break;
        default:
                /*
                 * All other request types are forced to be ordered.
                 */
                bioq_insert_tail(bioq, bio);
                break;
        }
}

/*
 * Move the read-write transition point to prevent reads from
 * completely starving our writes.  This brings a number of writes into
 * the fold every N reads.
 *
 * We bring a few linear writes into the fold on a minor interval
 * and we bring a non-linear burst of writes into the fold on a major
 * interval.  Bursting only occurs if runningbufspace is really high
 * (typically from syncs, fsyncs, or HAMMER flushes).
 */
static
void
bioqwritereorder(struct bio_queue_head *bioq)
{
        struct bio *bio;
        off_t next_offset;
        size_t left;
        size_t n;
        int check_off;

        if (bioq->reorder < bioq_reorder_burst_interval ||
            !buf_runningbufspace_severe()) {
                left = (size_t)bioq_reorder_minor_bytes;
                check_off = 1;
        } else {
                left = (size_t)bioq_reorder_burst_bytes;
                check_off = 0;
        }

        next_offset = bioq->transition->bio_offset;
        while ((bio = bioq->transition) != NULL &&
               (check_off == 0 || next_offset == bio->bio_offset)
        ) {
                n = bio->bio_buf->b_bcount;
                next_offset = bio->bio_offset + n;
                bioq->transition = TAILQ_NEXT(bio, bio_act);
                if (left < n)
                        break;
                left -= n;
        }
}

/*
 * Disk error is the preface to plaintive error messages
 * about failing disk transfers.  It prints messages of the form

hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)

 * if the offset of the error in the transfer and a disk label
 * are both available.  blkdone should be -1 if the position of the error
 * is unknown; the disklabel pointer may be null from drivers that have not
 * been converted to use them.  The message is printed with kprintf
 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
 * The message should be completed (with at least a newline) with kprintf
 * or log(-1, ...), respectively.  There is no trailing space.
 */
void
diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
{
        struct buf *bp = bio->bio_buf;
        const char *term;

        switch(bp->b_cmd) {
        case BUF_CMD_READ:
                term = "read";
                break;
        case BUF_CMD_WRITE:
                term = "write";
                break;
        default:
                term = "access";
                break;
        }
        kprintf("%s: %s %sing ", dev->si_name, what, term);
        kprintf("offset %012llx for %d",
                (long long)bio->bio_offset,
                bp->b_bcount);

        if (donecnt)
                kprintf(" (%d bytes completed)", donecnt);
}

/*
 * Locate a disk device
 */
cdev_t
disk_locate(const char *devname)
{
        return devfs_find_device_by_name(devname);
}

void
disk_config(void *arg)
{
        disk_msg_send_sync(DISK_SYNC, NULL, NULL);
}

static void
disk_init(void)
{
        struct thread* td_core;

        disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
                                         NULL, NULL, NULL,
                                         objcache_malloc_alloc,
                                         objcache_malloc_free,
                                         &disk_msg_malloc_args);

        lwkt_token_init(&disklist_token);

        /*
         * Initialize the reply-only port which acts as a message drain
         */
        lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);

        lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
                    0, 0, "disk_msg_core");

        tsleep(td_core, 0, "diskcore", 0);
}

static void
disk_uninit(void)
{
        objcache_destroy(disk_msg_cache);
}

/*
 * Clean out illegal characters in serial numbers.
 */
static void
disk_cleanserial(char *serno)
{
        char c;

        while ((c = *serno) != 0) {
                if (c >= 'a' && c <= 'z')
                        ;
                else if (c >= 'A' && c <= 'Z')
                        ;
                else if (c >= '0' && c <= '9')
                        ;
                else if (c == '-' || c == '@' || c == '+' || c == '.')
                        ;
                else
                        c = '_';
                *serno++= c;
        }
}

TUNABLE_INT("kern.disk_debug", &disk_debug_enable);
SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable,
                0, "Enable subr_disk debugging");

SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);