Due to races clean blocks can remain cached, remove a conditional from

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

/*
 * Copyright (c) 2003 Matthew Dillon <dillon@backplane.com> All rights reserved.
 * cdevsw from kern/kern_conf.c Copyright (c) 1995 Terrence R. Lambert
 * cdevsw from kern/kern_conf.c Copyright (c) 1995 Julian R. Elishcer,
 *                                                      All rights reserved.
 * Copyright (c) 1982, 1986, 1991, 1993
 *      The Regents of the University of California.  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 AUTHOR 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 AUTHOR 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.
 *
 * $DragonFly: src/sys/kern/kern_device.c,v 1.27 2007/07/23 18:59:50 dillon Exp $
 */
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/module.h>
#include <sys/malloc.h>
#include <sys/conf.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/vnode.h>
#include <sys/queue.h>
#include <sys/device.h>
#include <sys/tree.h>
#include <sys/syslink_rpc.h>
#include <sys/proc.h>
#include <machine/stdarg.h>
#include <sys/thread2.h>

/*
 * system link descriptors identify the command in the
 * arguments structure.
 */
#define DDESCNAME(name) __CONCAT(__CONCAT(dev_,name),_desc)

#define DEVOP_DESC_INIT(name)                                           \
            struct syslink_desc DDESCNAME(name) = {                     \
                __offsetof(struct dev_ops, __CONCAT(d_, name)), \
            #name }

DEVOP_DESC_INIT(default);
DEVOP_DESC_INIT(open);
DEVOP_DESC_INIT(close);
DEVOP_DESC_INIT(read);
DEVOP_DESC_INIT(write);
DEVOP_DESC_INIT(ioctl);
DEVOP_DESC_INIT(dump);
DEVOP_DESC_INIT(psize);
DEVOP_DESC_INIT(poll);
DEVOP_DESC_INIT(mmap);
DEVOP_DESC_INIT(strategy);
DEVOP_DESC_INIT(kqfilter);
DEVOP_DESC_INIT(clone);

/*
 * Misc default ops
 */
struct dev_ops dead_dev_ops;

struct dev_ops default_dev_ops = {
        { "null" },
        .d_default = NULL,      /* must be NULL */
        .d_open = noopen,
        .d_close = noclose,
        .d_read = noread,
        .d_write = nowrite,
        .d_ioctl = noioctl,
        .d_poll = nopoll,
        .d_mmap = nommap,
        .d_strategy = nostrategy,
        .d_dump = nodump,
        .d_psize = nopsize,
        .d_kqfilter = nokqfilter,
        .d_clone = noclone
};
    
/************************************************************************
 *                      GENERAL DEVICE API FUNCTIONS                    *
 ************************************************************************/

int
dev_dopen(cdev_t dev, int oflags, int devtype, struct ucred *cred)
{
        struct dev_open_args ap;

        ap.a_head.a_desc = &dev_open_desc;
        ap.a_head.a_dev = dev;
        ap.a_oflags = oflags;
        ap.a_devtype = devtype;
        ap.a_cred = cred;
        return(dev->si_ops->d_open(&ap));
}

int
dev_dclose(cdev_t dev, int fflag, int devtype)
{
        struct dev_close_args ap;

        ap.a_head.a_desc = &dev_close_desc;
        ap.a_head.a_dev = dev;
        ap.a_fflag = fflag;
        ap.a_devtype = devtype;
        return(dev->si_ops->d_close(&ap));
}

int
dev_dread(cdev_t dev, struct uio *uio, int ioflag)
{
        struct dev_read_args ap;
        int error;

        ap.a_head.a_desc = &dev_read_desc;
        ap.a_head.a_dev = dev;
        ap.a_uio = uio;
        ap.a_ioflag = ioflag;
        error = dev->si_ops->d_read(&ap);
        if (error == 0)
                dev->si_lastread = time_second;
        return (error);
}

int
dev_dwrite(cdev_t dev, struct uio *uio, int ioflag)
{
        struct dev_write_args ap;
        int error;

        dev->si_lastwrite = time_second;
        ap.a_head.a_desc = &dev_write_desc;
        ap.a_head.a_dev = dev;
        ap.a_uio = uio;
        ap.a_ioflag = ioflag;
        error = dev->si_ops->d_write(&ap);
        return (error);
}

int
dev_dioctl(cdev_t dev, u_long cmd, caddr_t data, int fflag, struct ucred *cred)
{
        struct dev_ioctl_args ap;

        ap.a_head.a_desc = &dev_ioctl_desc;
        ap.a_head.a_dev = dev;
        ap.a_cmd = cmd;
        ap.a_data = data;
        ap.a_fflag = fflag;
        ap.a_cred = cred;
        return(dev->si_ops->d_ioctl(&ap));
}

int
dev_dpoll(cdev_t dev, int events)
{
        struct dev_poll_args ap;
        int error;

        ap.a_head.a_desc = &dev_poll_desc;
        ap.a_head.a_dev = dev;
        ap.a_events = events;
        error = dev->si_ops->d_poll(&ap);
        if (error == 0)
                return(ap.a_events);
        return (seltrue(dev, events));
}

int
dev_dmmap(cdev_t dev, vm_offset_t offset, int nprot)
{
        struct dev_mmap_args ap;
        int error;

        ap.a_head.a_desc = &dev_mmap_desc;
        ap.a_head.a_dev = dev;
        ap.a_offset = offset;
        ap.a_nprot = nprot;
        error = dev->si_ops->d_mmap(&ap);
        if (error == 0)
                return(ap.a_result);
        return(-1);
}

int
dev_dclone(cdev_t dev)
{
        struct dev_clone_args ap;

        ap.a_head.a_desc = &dev_clone_desc;
        ap.a_head.a_dev = dev;
        return (dev->si_ops->d_clone(&ap));
}

/*
 * Core device strategy call, used to issue I/O on a device.  There are
 * two versions, a non-chained version and a chained version.  The chained
 * version reuses a BIO set up by vn_strategy().  The only difference is
 * that, for now, we do not push a new tracking structure when chaining
 * from vn_strategy.  XXX this will ultimately have to change.
 */
void
dev_dstrategy(cdev_t dev, struct bio *bio)
{
        struct dev_strategy_args ap;
        struct bio_track *track;

        ap.a_head.a_desc = &dev_strategy_desc;
        ap.a_head.a_dev = dev;
        ap.a_bio = bio;

        KKASSERT(bio->bio_track == NULL);
        KKASSERT(bio->bio_buf->b_cmd != BUF_CMD_DONE);
        if (bio->bio_buf->b_cmd == BUF_CMD_READ)
            track = &dev->si_track_read;
        else
            track = &dev->si_track_write;
        atomic_add_int(&track->bk_active, 1);
        bio->bio_track = track;
        (void)dev->si_ops->d_strategy(&ap);
}

void
dev_dstrategy_chain(cdev_t dev, struct bio *bio)
{
        struct dev_strategy_args ap;

        KKASSERT(bio->bio_track != NULL);
        ap.a_head.a_desc = &dev_strategy_desc;
        ap.a_head.a_dev = dev;
        ap.a_bio = bio;
        (void)dev->si_ops->d_strategy(&ap);
}

/*
 * note: the disk layer is expected to set count, blkno, and secsize before
 * forwarding the message.
 */
int
dev_ddump(cdev_t dev)
{
        struct dev_dump_args ap;

        ap.a_head.a_desc = &dev_dump_desc;
        ap.a_head.a_dev = dev;
        ap.a_count = 0;
        ap.a_blkno = 0;
        ap.a_secsize = 0;
        return(dev->si_ops->d_dump(&ap));
}

int64_t
dev_dpsize(cdev_t dev)
{
        struct dev_psize_args ap;
        int error;

        ap.a_head.a_desc = &dev_psize_desc;
        ap.a_head.a_dev = dev;
        error = dev->si_ops->d_psize(&ap);
        if (error == 0)
                return (ap.a_result);
        return(-1);
}

int
dev_dkqfilter(cdev_t dev, struct knote *kn)
{
        struct dev_kqfilter_args ap;
        int error;

        ap.a_head.a_desc = &dev_kqfilter_desc;
        ap.a_head.a_dev = dev;
        ap.a_kn = kn;
        error = dev->si_ops->d_kqfilter(&ap);
        if (error == 0)
                return(ap.a_result);
        return(ENODEV);
}

/************************************************************************
 *                      DEVICE HELPER FUNCTIONS                         *
 ************************************************************************/

int
dev_drefs(cdev_t dev)
{
    return(dev->si_sysref.refcnt);
}

const char *
dev_dname(cdev_t dev)
{
    return(dev->si_ops->head.name);
}

int
dev_dflags(cdev_t dev)
{
    return(dev->si_ops->head.flags);
}

int
dev_dmaj(cdev_t dev)
{
    return(dev->si_ops->head.maj);
}

/*
 * Used when forwarding a request through layers.  The caller adjusts
 * ap->a_head.a_dev and then calls this function.
 */
int
dev_doperate(struct dev_generic_args *ap)
{
    int (*func)(struct dev_generic_args *);

    func = *(void **)((char *)ap->a_dev->si_ops + ap->a_desc->sd_offset);
    return (func(ap));
}

/*
 * Used by the console intercept code only.  Issue an operation through
 * a foreign ops structure allowing the ops structure associated
 * with the device to remain intact.
 */
int
dev_doperate_ops(struct dev_ops *ops, struct dev_generic_args *ap)
{
    int (*func)(struct dev_generic_args *);

    func = *(void **)((char *)ops + ap->a_desc->sd_offset);
    return (func(ap));
}

/*
 * Convert a template dev_ops into the real thing by filling in 
 * uninitialized fields.
 */
void
compile_dev_ops(struct dev_ops *ops)
{
        int offset;

        for (offset = offsetof(struct dev_ops, dev_ops_first_field);
             offset <= offsetof(struct dev_ops, dev_ops_last_field);
             offset += sizeof(void *)
        ) {
                void **func_p = (void **)((char *)ops + offset);
                void **def_p = (void **)((char *)&default_dev_ops + offset);
                if (*func_p == NULL) {
                        if (ops->d_default)
                                *func_p = ops->d_default;
                        else
                                *func_p = *def_p;
                }
        }
}

/************************************************************************
 *                      MAJOR/MINOR SPACE FUNCTION                      *
 ************************************************************************/

/*
 * This makes a dev_ops entry visible to userland (e.g /dev/<blah>).
 *
 * The kernel can overload a data space by making multiple dev_ops_add()
 * calls, but only the most recent one in the list matching the mask/match
 * will be visible to userland.
 *
 * make_dev() does not automatically call dev_ops_add() (nor do we want it
 * to, since partition-managed disk devices are overloaded on top of the
 * raw device).
 *
 * Disk devices typically register their major, e.g. 'ad0', and then call
 * into the disk label management code which overloads its own onto e.g. 'ad0'
 * to support all the various slice and partition combinations.
 *
 * The mask/match supplied in this call are a full 32 bits and the same
 * mask and match must be specified in a later dev_ops_remove() call to
 * match this add.  However, the match value for the minor number should never
 * have any bits set in the major number's bit range (8-15).  The mask value
 * may be conveniently specified as -1 without creating any major number
 * interference.
 */

static
int
rb_dev_ops_compare(struct dev_ops_maj *a, struct dev_ops_maj *b)
{
    if (a->maj < b->maj)
        return(-1);
    else if (a->maj > b->maj)
        return(1);
    return(0);
}

RB_GENERATE2(dev_ops_rb_tree, dev_ops_maj, rbnode, rb_dev_ops_compare, int, maj);

struct dev_ops_rb_tree dev_ops_rbhead = RB_INITIALIZER(dev_ops_rbhead);

int
dev_ops_add(struct dev_ops *ops, u_int mask, u_int match)
{
    static int next_maj = 256;          /* first dynamic major number */
    struct dev_ops_maj *rbmaj;
    struct dev_ops_link *link;

    compile_dev_ops(ops);
    if (ops->head.maj < 0) {
        while (dev_ops_rb_tree_RB_LOOKUP(&dev_ops_rbhead, next_maj) != NULL) {
                if (++next_maj <= 0)
                        next_maj = 256;
        }
        ops->head.maj = next_maj;
    }
    rbmaj = dev_ops_rb_tree_RB_LOOKUP(&dev_ops_rbhead, ops->head.maj);
    if (rbmaj == NULL) {
        rbmaj = kmalloc(sizeof(*rbmaj), M_DEVBUF, M_INTWAIT | M_ZERO);
        rbmaj->maj = ops->head.maj;
        dev_ops_rb_tree_RB_INSERT(&dev_ops_rbhead, rbmaj);
    }
    for (link = rbmaj->link; link; link = link->next) {
            /*
             * If we get an exact match we usurp the target, but we only print
             * a warning message if a different device switch is installed.
             */
            if (link->mask == mask && link->match == match) {
                    if (link->ops != ops) {
                            kprintf("WARNING: \"%s\" (%p) is usurping \"%s\"'s"
                                " (%p)\n",
                                ops->head.name, ops, 
                                link->ops->head.name, link->ops);
                            link->ops = ops;
                            ++ops->head.refs;
                    }
                    return(0);
            }
            /*
             * XXX add additional warnings for overlaps
             */
    }

    link = kmalloc(sizeof(struct dev_ops_link), M_DEVBUF, M_INTWAIT|M_ZERO);
    link->mask = mask;
    link->match = match;
    link->ops = ops;
    link->next = rbmaj->link;
    rbmaj->link = link;
    ++ops->head.refs;
    return(0);
}

/*
 * Should only be used by udev2dev().
 *
 * If the minor number is -1, we match the first ops we find for this
 * major.   If the mask is not -1 then multiple minor numbers can match
 * the same ops.
 *
 * Note that this function will return NULL if the minor number is not within
 * the bounds of the installed mask(s).
 *
 * The specified minor number should NOT include any major bits.
 */
struct dev_ops *
dev_ops_get(int x, int y)
{
        struct dev_ops_maj *rbmaj;
        struct dev_ops_link *link;

        rbmaj = dev_ops_rb_tree_RB_LOOKUP(&dev_ops_rbhead, x);
        if (rbmaj == NULL)
                return(NULL);
        for (link = rbmaj->link; link; link = link->next) {
                if (y == -1 || (link->mask & y) == link->match)
                        return(link->ops);
        }
        return(NULL);
}

/*
 * Take a cookie cutter to the major/minor device space for the passed
 * device and generate a new dev_ops visible to userland which the caller
 * can then modify.  The original device is not modified but portions of
 * its major/minor space will no longer be visible to userland.
 */
struct dev_ops *
dev_ops_add_override(cdev_t backing_dev, struct dev_ops *template,
                     u_int mask, u_int match)
{
        struct dev_ops *ops;
        struct dev_ops *backing_ops = backing_dev->si_ops;

        ops = kmalloc(sizeof(struct dev_ops), M_DEVBUF, M_INTWAIT);
        *ops = *template;
        ops->head.name = backing_ops->head.name;
        ops->head.maj = backing_ops->head.maj;
        ops->head.flags |= backing_ops->head.flags & ~D_TRACKCLOSE;
        compile_dev_ops(ops);
        dev_ops_add(ops, mask, match);

        return(ops);
}

/*
 * Remove all matching dev_ops entries from the dev_ops_array[] major
 * array so no new user opens can be performed, and destroy all devices
 * installed in the hash table that are associated with this dev_ops.  (see
 * destroy_all_devs()).
 *
 * The mask and match should match a previous call to dev_ops_add*().
 */
int
dev_ops_remove(struct dev_ops *ops, u_int mask, u_int match)
{
        struct dev_ops_maj *rbmaj;
        struct dev_ops_link *link;
        struct dev_ops_link **plink;
     
        if (ops != &dead_dev_ops)
                destroy_all_devs(ops, mask, match);

        rbmaj = dev_ops_rb_tree_RB_LOOKUP(&dev_ops_rbhead, ops->head.maj);
        if (rbmaj == NULL) {
                kprintf("double-remove of dev_ops %p for %s(%d)\n",
                        ops, ops->head.name, ops->head.maj);
                return(0);
        }
        for (plink = &rbmaj->link; (link = *plink) != NULL;
             plink = &link->next) {
                if (link->mask == mask && link->match == match) {
                        if (link->ops == ops)
                                break;
                        kprintf("%s: ERROR: cannot remove dev_ops, "
                               "its major number %d was stolen by %s\n",
                                ops->head.name, ops->head.maj,
                                link->ops->head.name
                        );
                }
        }
        if (link == NULL) {
                kprintf("%s(%d)[%08x/%08x]: WARNING: ops removed "
                       "multiple times!\n",
                       ops->head.name, ops->head.maj, mask, match);
        } else {
                *plink = link->next;
                --ops->head.refs; /* XXX ops_release() / record refs */
                kfree(link, M_DEVBUF);
        }

        /*
         * Scrap the RB tree node for the major number if no ops are
         * installed any longer.
         */
        if (rbmaj->link == NULL) {
                dev_ops_rb_tree_RB_REMOVE(&dev_ops_rbhead, rbmaj);
                kfree(rbmaj, M_DEVBUF);
        }

        if (ops->head.refs != 0) {
                kprintf("%s(%d)[%08x/%08x]: Warning: dev_ops_remove() called "
                        "while %d device refs still exist!\n", 
                        ops->head.name, ops->head.maj, mask, match,
                        ops->head.refs);
        } else {
                if (bootverbose)
                        kprintf("%s: ops removed\n", ops->head.name);
        }
        return 0;
}

/*
 * dev_ops_scan() - Issue a callback for all installed dev_ops structures.
 *
 * The scan will terminate if a callback returns a negative number.
 */
struct dev_ops_scan_info {
        int     (*callback)(struct dev_ops *, void *);
        void    *arg;
};

static
int
dev_ops_scan_callback(struct dev_ops_maj *rbmaj, void *arg)
{
        struct dev_ops_scan_info *info = arg;
        struct dev_ops_link *link;
        int count = 0;
        int r;

        for (link = rbmaj->link; link; link = link->next) {
                r = info->callback(link->ops, info->arg);
                if (r < 0)
                        return(r);
                count += r;
        }
        return(count);
}

int
dev_ops_scan(int (*callback)(struct dev_ops *, void *), void *arg)
{
        struct dev_ops_scan_info info = { callback, arg };

        return (dev_ops_rb_tree_RB_SCAN(&dev_ops_rbhead, NULL,
                                        dev_ops_scan_callback, &info));
}


/*
 * Release a ops entry.  When the ref count reaches zero, recurse
 * through the stack.
 */
void
dev_ops_release(struct dev_ops *ops)
{
        --ops->head.refs;
        if (ops->head.refs == 0) {
                /* XXX */
        }
}

struct dev_ops *
dev_ops_intercept(cdev_t dev, struct dev_ops *iops)
{
        struct dev_ops *oops = dev->si_ops;

        compile_dev_ops(iops);
        iops->head.maj = oops->head.maj;
        iops->head.data = oops->head.data;
        iops->head.flags = oops->head.flags;
        dev->si_ops = iops;
        dev->si_flags |= SI_INTERCEPTED;

        return (oops);
}

void
dev_ops_restore(cdev_t dev, struct dev_ops *oops)
{
        struct dev_ops *iops = dev->si_ops;

        dev->si_ops = oops;
        dev->si_flags &= ~SI_INTERCEPTED;
        iops->head.maj = 0;
        iops->head.data = NULL;
        iops->head.flags = 0;
}

/************************************************************************
 *                      DEFAULT DEV OPS FUNCTIONS                       *
 ************************************************************************/


/*
 * Unsupported devswitch functions (e.g. for writing to read-only device).
 * XXX may belong elsewhere.
 */

int
noclone(struct dev_clone_args *ap)
{
        /* take no action */
        return (0);     /* allow the clone */
}

int
noopen(struct dev_open_args *ap)
{
        return (ENODEV);
}

int
noclose(struct dev_close_args *ap)
{
        return (ENODEV);
}

int
noread(struct dev_read_args *ap)
{
        return (ENODEV);
}

int
nowrite(struct dev_write_args *ap)
{
        return (ENODEV);
}

int
noioctl(struct dev_ioctl_args *ap)
{
        return (ENODEV);
}

int
nokqfilter(struct dev_kqfilter_args *ap)
{
        return (ENODEV);
}

int
nommap(struct dev_mmap_args *ap)
{
        return (ENODEV);
}

int
nopoll(struct dev_poll_args *ap)
{
        ap->a_events = 0;
        return(0);
}

int
nostrategy(struct dev_strategy_args *ap)
{
        struct bio *bio = ap->a_bio;

        bio->bio_buf->b_flags |= B_ERROR;
        bio->bio_buf->b_error = EOPNOTSUPP;
        biodone(bio);
        return(0);
}

int
nopsize(struct dev_psize_args *ap)
{
        ap->a_result = 0;
        return(0);
}

int
nodump(struct dev_dump_args *ap)
{
        return (ENODEV);
}

/*
 * XXX this is probably bogus.  Any device that uses it isn't checking the
 * minor number.
 */
int
nullopen(struct dev_open_args *ap)
{
        return (0);
}

int
nullclose(struct dev_close_args *ap)
{
        return (0);
}