xref: /freebsd/sys/dev/xen/blkfront/blkfront.c (revision a4660d594f0f1d4687fd71750af95bd7b61c679e)

/*
 * XenBSD block device driver
 *
 * Copyright (c) 2009 Scott Long, Yahoo!
 * Copyright (c) 2009 Frank Suchomel, Citrix
 * Copyright (c) 2009 Doug F. Rabson, Citrix
 * Copyright (c) 2005 Kip Macy
 * Copyright (c) 2003-2004, Keir Fraser & Steve Hand
 * Modifications by Mark A. Williamson are (c) Intel Research Cambridge
 *
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to
 * deal in the Software without restriction, including without limitation the
 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
 * sell copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <vm/vm.h>
#include <vm/pmap.h>

#include <sys/bio.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/module.h>

#include <machine/bus.h>
#include <sys/rman.h>
#include <machine/resource.h>
#include <machine/intr_machdep.h>
#include <machine/vmparam.h>
#include <sys/bus_dma.h>

#include <machine/xen/xen-os.h>
#include <machine/xen/xenfunc.h>
#include <xen/hypervisor.h>
#include <xen/xen_intr.h>
#include <xen/evtchn.h>
#include <xen/gnttab.h>
#include <xen/interface/grant_table.h>
#include <xen/interface/io/protocols.h>
#include <xen/xenbus/xenbusvar.h>

#include <geom/geom_disk.h>

#include <dev/xen/blkfront/block.h>

#include "xenbus_if.h"

/* prototypes */
static void xb_free_command(struct xb_command *cm);
static void xb_startio(struct xb_softc *sc);
static void connect(struct xb_softc *);
static void blkfront_closing(device_t);
static int blkfront_detach(device_t);
static int talk_to_backend(struct xb_softc *);
static int setup_blkring(struct xb_softc *);
static void blkif_int(void *);
static void blkif_recover(struct xb_softc *);
static void blkif_completion(struct xb_command *);
static void blkif_free(struct xb_softc *, int);
static void blkif_queue_cb(void *, bus_dma_segment_t *, int, int);

#define GRANT_INVALID_REF 0

/* Control whether runtime update of vbds is enabled. */
#define ENABLE_VBD_UPDATE 0

#if ENABLE_VBD_UPDATE
static void vbd_update(void);
#endif

#define BLKIF_STATE_DISCONNECTED 0
#define BLKIF_STATE_CONNECTED    1
#define BLKIF_STATE_SUSPENDED    2

#ifdef notyet
static char *blkif_state_name[] = {
	[BLKIF_STATE_DISCONNECTED] = "disconnected",
	[BLKIF_STATE_CONNECTED]    = "connected",
	[BLKIF_STATE_SUSPENDED]    = "closed",
};

static char * blkif_status_name[] = {
	[BLKIF_INTERFACE_STATUS_CLOSED]       = "closed",
	[BLKIF_INTERFACE_STATUS_DISCONNECTED] = "disconnected",
	[BLKIF_INTERFACE_STATUS_CONNECTED]    = "connected",
	[BLKIF_INTERFACE_STATUS_CHANGED]      = "changed",
};
#endif

#if 0
#define DPRINTK(fmt, args...) printf("[XEN] %s:%d: " fmt ".\n", __func__, __LINE__, ##args)
#else
#define DPRINTK(fmt, args...)
#endif

#define MAXIMUM_OUTSTANDING_BLOCK_REQS \
    (BLKIF_MAX_SEGMENTS_PER_REQUEST * BLK_RING_SIZE)

#define BLKIF_MAXIO	(32 * 1024)

static int blkif_open(struct disk *dp);
static int blkif_close(struct disk *dp);
static int blkif_ioctl(struct disk *dp, u_long cmd, void *addr, int flag, struct thread *td);
static int blkif_queue_request(struct xb_softc *sc, struct xb_command *cm);
static void xb_strategy(struct bio *bp);

// In order to quiesce the device during kernel dumps, outstanding requests to
// DOM0 for disk reads/writes need to be accounted for.
static	int	xb_dump(void *, void *, vm_offset_t, off_t, size_t);

/* XXX move to xb_vbd.c when VBD update support is added */
#define MAX_VBDS 64

#define XBD_SECTOR_SIZE		512	/* XXX: assume for now */
#define XBD_SECTOR_SHFT		9

/*
 * Translate Linux major/minor to an appropriate name and unit
 * number. For HVM guests, this allows us to use the same drive names
 * with blkfront as the emulated drives, easing transition slightly.
 */
static void
blkfront_vdevice_to_unit(int vdevice, int *unit, const char **name)
{
	static struct vdev_info {
		int major;
		int shift;
		int base;
		const char *name;
	} info[] = {
		{3,	6,	0,	"ad"},	/* ide0 */
		{22,	6,	2,	"ad"},	/* ide1 */
		{33,	6,	4,	"ad"},	/* ide2 */
		{34,	6,	6,	"ad"},	/* ide3 */
		{56,	6,	8,	"ad"},	/* ide4 */
		{57,	6,	10,	"ad"},	/* ide5 */
		{88,	6,	12,	"ad"},	/* ide6 */
		{89,	6,	14,	"ad"},	/* ide7 */
		{90,	6,	16,	"ad"},	/* ide8 */
		{91,	6,	18,	"ad"},	/* ide9 */

		{8,	4,	0,	"da"},	/* scsi disk0 */
		{65,	4,	16,	"da"},	/* scsi disk1 */
		{66,	4,	32,	"da"},	/* scsi disk2 */
		{67,	4,	48,	"da"},	/* scsi disk3 */
		{68,	4,	64,	"da"},	/* scsi disk4 */
		{69,	4,	80,	"da"},	/* scsi disk5 */
		{70,	4,	96,	"da"},	/* scsi disk6 */
		{71,	4,	112,	"da"},	/* scsi disk7 */
		{128,	4,	128,	"da"},	/* scsi disk8 */
		{129,	4,	144,	"da"},	/* scsi disk9 */
		{130,	4,	160,	"da"},	/* scsi disk10 */
		{131,	4,	176,	"da"},	/* scsi disk11 */
		{132,	4,	192,	"da"},	/* scsi disk12 */
		{133,	4,	208,	"da"},	/* scsi disk13 */
		{134,	4,	224,	"da"},	/* scsi disk14 */
		{135,	4,	240,	"da"},	/* scsi disk15 */

		{202,	4,	0,	"xbd"},	/* xbd */

		{0,	0,	0,	NULL},
	};
	int major = vdevice >> 8;
	int minor = vdevice & 0xff;
	int i;

	if (vdevice & (1 << 28)) {
		*unit = (vdevice & ((1 << 28) - 1)) >> 8;
		*name = "xbd";
	}

	for (i = 0; info[i].major; i++) {
		if (info[i].major == major) {
			*unit = info[i].base + (minor >> info[i].shift);
			*name = info[i].name;
			return;
		}
	}

	*unit = minor >> 4;
	*name = "xbd";
}

int
xlvbd_add(struct xb_softc *sc, blkif_sector_t capacity,
    int vdevice, uint16_t vdisk_info, uint16_t sector_size)
{
	int	unit, error = 0;
	const char *name;

	blkfront_vdevice_to_unit(vdevice, &unit, &name);

	sc->xb_unit = unit;

	if (strcmp(name, "xbd"))
		device_printf(sc->xb_dev, "attaching as %s%d\n", name, unit);

	memset(&sc->xb_disk, 0, sizeof(sc->xb_disk));
	sc->xb_disk = disk_alloc();
	sc->xb_disk->d_unit = sc->xb_unit;
	sc->xb_disk->d_open = blkif_open;
	sc->xb_disk->d_close = blkif_close;
	sc->xb_disk->d_ioctl = blkif_ioctl;
	sc->xb_disk->d_strategy = xb_strategy;
	sc->xb_disk->d_dump = xb_dump;
	sc->xb_disk->d_name = name;
	sc->xb_disk->d_drv1 = sc;
	sc->xb_disk->d_sectorsize = sector_size;

	sc->xb_disk->d_mediasize = capacity << XBD_SECTOR_SHFT;
	sc->xb_disk->d_maxsize = BLKIF_MAXIO;
	sc->xb_disk->d_flags = 0;
	disk_create(sc->xb_disk, DISK_VERSION_00);

	return error;
}

void
xlvbd_del(struct xb_softc *sc)
{

	disk_destroy(sc->xb_disk);
}
/************************ end VBD support *****************/

/*
 * Read/write routine for a buffer.  Finds the proper unit, place it on
 * the sortq and kick the controller.
 */
static void
xb_strategy(struct bio *bp)
{
	struct xb_softc	*sc = (struct xb_softc *)bp->bio_disk->d_drv1;

	/* bogus disk? */
	if (sc == NULL) {
		bp->bio_error = EINVAL;
		bp->bio_flags |= BIO_ERROR;
		bp->bio_resid = bp->bio_bcount;
		biodone(bp);
		return;
	}

	/*
	 * Place it in the queue of disk activities for this disk
	 */
	mtx_lock(&sc->xb_io_lock);

	xb_enqueue_bio(sc, bp);
	xb_startio(sc);

	mtx_unlock(&sc->xb_io_lock);
	return;
}

static void
xb_bio_complete(struct xb_softc *sc, struct xb_command *cm)
{
	struct bio *bp;

	bp = cm->bp;

	if ( unlikely(cm->status != BLKIF_RSP_OKAY) ) {
		disk_err(bp, "disk error" , -1, 0);
		printf(" status: %x\n", cm->status);
		bp->bio_flags |= BIO_ERROR;
	}

	if (bp->bio_flags & BIO_ERROR)
		bp->bio_error = EIO;
	else
		bp->bio_resid = 0;

	xb_free_command(cm);
	biodone(bp);
}

// Quiesce the disk writes for a dump file before allowing the next buffer.
static void
xb_quiesce(struct xb_softc *sc)
{
	int		mtd;

	// While there are outstanding requests
	while (!TAILQ_EMPTY(&sc->cm_busy)) {
		RING_FINAL_CHECK_FOR_RESPONSES(&sc->ring, mtd);
		if (mtd) {
			/* Recieved request completions, update queue. */
			blkif_int(sc);
		}
		if (!TAILQ_EMPTY(&sc->cm_busy)) {
			/*
			 * Still pending requests, wait for the disk i/o
			 * to complete.
			 */
			HYPERVISOR_yield();
		}
	}
}

/* Kernel dump function for a paravirtualized disk device */
static void
xb_dump_complete(struct xb_command *cm)
{

	xb_enqueue_complete(cm);
}

static int
xb_dump(void *arg, void *virtual, vm_offset_t physical, off_t offset,
        size_t length)
{
	struct	disk   	*dp = arg;
	struct xb_softc	*sc = (struct xb_softc *) dp->d_drv1;
	struct xb_command *cm;
	size_t		chunk;
	int		sbp;
	int		rc = 0;

	if (length <= 0)
		return (rc);

	xb_quiesce(sc);	/* All quiet on the western front. */

	/*
	 * If this lock is held, then this module is failing, and a
	 * successful kernel dump is highly unlikely anyway.
	 */
	mtx_lock(&sc->xb_io_lock);

	/* Split the 64KB block as needed */
	for (sbp=0; length > 0; sbp++) {
		cm = xb_dequeue_free(sc);
		if (cm == NULL) {
			mtx_unlock(&sc->xb_io_lock);
			device_printf(sc->xb_dev, "dump: no more commands?\n");
			return (EBUSY);
		}

		if (gnttab_alloc_grant_references(
		    BLKIF_MAX_SEGMENTS_PER_REQUEST, &cm->gref_head) < 0) {
			xb_free_command(cm);
			mtx_unlock(&sc->xb_io_lock);
			device_printf(sc->xb_dev, "no more grant allocs?\n");
			return (EBUSY);
		}

		chunk = length > BLKIF_MAXIO ? BLKIF_MAXIO : length;
		cm->data = virtual;
		cm->datalen = chunk;
		cm->operation = BLKIF_OP_WRITE;
		cm->sector_number = offset / dp->d_sectorsize;
		cm->cm_complete = xb_dump_complete;

		xb_enqueue_ready(cm);

		length -= chunk;
		offset += chunk;
		virtual = (char *) virtual + chunk;
	}

	/* Tell DOM0 to do the I/O */
	xb_startio(sc);
	mtx_unlock(&sc->xb_io_lock);

	/* Poll for the completion. */
	xb_quiesce(sc);	/* All quite on the eastern front */

	/* If there were any errors, bail out... */
	while ((cm = xb_dequeue_complete(sc)) != NULL) {
		if (cm->status != BLKIF_RSP_OKAY) {
			device_printf(sc->xb_dev,
			    "Dump I/O failed at sector %jd\n",
			    cm->sector_number);
			rc = EIO;
		}
		xb_free_command(cm);
	}

	return (rc);
}


static int
blkfront_probe(device_t dev)
{

	if (!strcmp(xenbus_get_type(dev), "vbd")) {
		device_set_desc(dev, "Virtual Block Device");
		device_quiet(dev);
		return (0);
	}

	return (ENXIO);
}

/*
 * Setup supplies the backend dir, virtual device.  We place an event
 * channel and shared frame entries.  We watch backend to wait if it's
 * ok.
 */
static int
blkfront_attach(device_t dev)
{
	struct xb_softc *sc;
	struct xb_command *cm;
	const char *name;
	int error, vdevice, i, unit;

	/* FIXME: Use dynamic device id if this is not set. */
	error = xenbus_scanf(XBT_NIL, xenbus_get_node(dev),
	    "virtual-device", NULL, "%i", &vdevice);
	if (error) {
		xenbus_dev_fatal(dev, error, "reading virtual-device");
		printf("couldn't find virtual device");
		return (error);
	}

	blkfront_vdevice_to_unit(vdevice, &unit, &name);
	if (!strcmp(name, "xbd"))
		device_set_unit(dev, unit);

	sc = device_get_softc(dev);
	mtx_init(&sc->xb_io_lock, "blkfront i/o lock", NULL, MTX_DEF);
	xb_initq_free(sc);
	xb_initq_busy(sc);
	xb_initq_ready(sc);
	xb_initq_complete(sc);
	xb_initq_bio(sc);

	/* Allocate parent DMA tag */
	if (bus_dma_tag_create(	NULL,			/* parent */
				512, 4096,		/* algnmnt, boundary */
				BUS_SPACE_MAXADDR,	/* lowaddr */
				BUS_SPACE_MAXADDR,	/* highaddr */
				NULL, NULL,		/* filter, filterarg */
				BLKIF_MAXIO,		/* maxsize */
				BLKIF_MAX_SEGMENTS_PER_REQUEST,	/* nsegments */
				PAGE_SIZE,		/* maxsegsize */
				BUS_DMA_ALLOCNOW,	/* flags */
				busdma_lock_mutex,	/* lockfunc */
				&sc->xb_io_lock,	/* lockarg */
				&sc->xb_io_dmat)) {
		device_printf(dev, "Cannot allocate parent DMA tag\n");
		return (ENOMEM);
	}
#ifdef notyet
	if (bus_dma_tag_set(sc->xb_io_dmat, BUS_DMA_SET_MINSEGSZ,
		XBD_SECTOR_SIZE)) {
		device_printf(dev, "Cannot set sector size\n");
		return (EINVAL);
	}
#endif

	sc->xb_dev = dev;
	sc->vdevice = vdevice;
	sc->connected = BLKIF_STATE_DISCONNECTED;

	/* work queue needed ? */
	for (i = 0; i < BLK_RING_SIZE; i++) {
		cm = &sc->shadow[i];
		cm->req.id = i;
		cm->cm_sc = sc;
		if (bus_dmamap_create(sc->xb_io_dmat, 0, &cm->map) != 0)
			break;
		xb_free_command(cm);
	}

	/* Front end dir is a number, which is used as the id. */
	sc->handle = strtoul(strrchr(xenbus_get_node(dev),'/')+1, NULL, 0);

	error = talk_to_backend(sc);
	if (error)
		return (error);

	return (0);
}

static int
blkfront_suspend(device_t dev)
{
	struct xb_softc *sc = device_get_softc(dev);

	/* Prevent new requests being issued until we fix things up. */
	mtx_lock(&sc->xb_io_lock);
	sc->connected = BLKIF_STATE_SUSPENDED;
	mtx_unlock(&sc->xb_io_lock);

	return (0);
}

static int
blkfront_resume(device_t dev)
{
	struct xb_softc *sc = device_get_softc(dev);
	int err;

	DPRINTK("blkfront_resume: %s\n", xenbus_get_node(dev));

	blkif_free(sc, 1);
	err = talk_to_backend(sc);
	if (sc->connected == BLKIF_STATE_SUSPENDED && !err)
		blkif_recover(sc);

	return (err);
}

/* Common code used when first setting up, and when resuming. */
static int
talk_to_backend(struct xb_softc *sc)
{
	device_t dev;
	struct xenbus_transaction xbt;
	const char *message = NULL;
	int err;

	/* Create shared ring, alloc event channel. */
	dev = sc->xb_dev;
	err = setup_blkring(sc);
	if (err)
		goto out;

 again:
	err = xenbus_transaction_start(&xbt);
	if (err) {
		xenbus_dev_fatal(dev, err, "starting transaction");
		goto destroy_blkring;
	}

	err = xenbus_printf(xbt, xenbus_get_node(dev),
			    "ring-ref","%u", sc->ring_ref);
	if (err) {
		message = "writing ring-ref";
		goto abort_transaction;
	}
	err = xenbus_printf(xbt, xenbus_get_node(dev),
		"event-channel", "%u", irq_to_evtchn_port(sc->irq));
	if (err) {
		message = "writing event-channel";
		goto abort_transaction;
	}
	err = xenbus_printf(xbt, xenbus_get_node(dev),
		"protocol", "%s", XEN_IO_PROTO_ABI_NATIVE);
	if (err) {
		message = "writing protocol";
		goto abort_transaction;
	}

	err = xenbus_transaction_end(xbt, 0);
	if (err) {
		if (err == EAGAIN)
			goto again;
		xenbus_dev_fatal(dev, err, "completing transaction");
		goto destroy_blkring;
	}
	xenbus_set_state(dev, XenbusStateInitialised);

	return 0;

 abort_transaction:
	xenbus_transaction_end(xbt, 1);
	if (message)
		xenbus_dev_fatal(dev, err, "%s", message);
 destroy_blkring:
	blkif_free(sc, 0);
 out:
	return err;
}

static int
setup_blkring(struct xb_softc *sc)
{
	blkif_sring_t *sring;
	int error;

	sc->ring_ref = GRANT_INVALID_REF;

	sring = (blkif_sring_t *)malloc(PAGE_SIZE, M_DEVBUF, M_NOWAIT|M_ZERO);
	if (sring == NULL) {
		xenbus_dev_fatal(sc->xb_dev, ENOMEM, "allocating shared ring");
		return ENOMEM;
	}
	SHARED_RING_INIT(sring);
	FRONT_RING_INIT(&sc->ring, sring, PAGE_SIZE);

	error = xenbus_grant_ring(sc->xb_dev,
	    (vtomach(sc->ring.sring) >> PAGE_SHIFT), &sc->ring_ref);
	if (error) {
		free(sring, M_DEVBUF);
		sc->ring.sring = NULL;
		goto fail;
	}

	error = bind_listening_port_to_irqhandler(xenbus_get_otherend_id(sc->xb_dev),
	    "xbd", (driver_intr_t *)blkif_int, sc,
	    INTR_TYPE_BIO | INTR_MPSAFE, &sc->irq);
	if (error) {
		xenbus_dev_fatal(sc->xb_dev, error,
		    "bind_evtchn_to_irqhandler failed");
		goto fail;
	}

	return (0);
 fail:
	blkif_free(sc, 0);
	return (error);
}


/**
 * Callback received when the backend's state changes.
 */
static int
blkfront_backend_changed(device_t dev, XenbusState backend_state)
{
	struct xb_softc *sc = device_get_softc(dev);

	DPRINTK("backend_state=%d\n", backend_state);

	switch (backend_state) {
	case XenbusStateUnknown:
	case XenbusStateInitialising:
	case XenbusStateInitWait:
	case XenbusStateInitialised:
	case XenbusStateClosed:
	case XenbusStateReconfigured:
	case XenbusStateReconfiguring:
		break;

	case XenbusStateConnected:
		connect(sc);
		break;

	case XenbusStateClosing:
		if (sc->users > 0)
			xenbus_dev_error(dev, -EBUSY,
					 "Device in use; refusing to close");
		else
			blkfront_closing(dev);
#ifdef notyet
		bd = bdget(sc->dev);
		if (bd == NULL)
			xenbus_dev_fatal(dev, -ENODEV, "bdget failed");

		down(&bd->bd_sem);
		if (sc->users > 0)
			xenbus_dev_error(dev, -EBUSY,
					 "Device in use; refusing to close");
		else
			blkfront_closing(dev);
		up(&bd->bd_sem);
		bdput(bd);
#endif
	}

	return (0);
}

/*
** Invoked when the backend is finally 'ready' (and has told produced
** the details about the physical device - #sectors, size, etc).
*/
static void
connect(struct xb_softc *sc)
{
	device_t dev = sc->xb_dev;
	unsigned long sectors, sector_size;
	unsigned int binfo;
	int err, feature_barrier;

        if( (sc->connected == BLKIF_STATE_CONNECTED) ||
	    (sc->connected == BLKIF_STATE_SUSPENDED) )
		return;

	DPRINTK("blkfront.c:connect:%s.\n", xenbus_get_otherend_path(dev));

	err = xenbus_gather(XBT_NIL, xenbus_get_otherend_path(dev),
			    "sectors", "%lu", &sectors,
			    "info", "%u", &binfo,
			    "sector-size", "%lu", &sector_size,
			    NULL);
	if (err) {
		xenbus_dev_fatal(dev, err,
		    "reading backend fields at %s",
		    xenbus_get_otherend_path(dev));
		return;
	}
	err = xenbus_gather(XBT_NIL, xenbus_get_otherend_path(dev),
			    "feature-barrier", "%lu", &feature_barrier,
			    NULL);
	if (!err || feature_barrier)
		sc->xb_flags |= XB_BARRIER;

	device_printf(dev, "%juMB <%s> at %s",
	    (uintmax_t) sectors / (1048576 / sector_size),
	    device_get_desc(dev),
	    xenbus_get_node(dev));
	bus_print_child_footer(device_get_parent(dev), dev);

	xlvbd_add(sc, sectors, sc->vdevice, binfo, sector_size);

	(void)xenbus_set_state(dev, XenbusStateConnected);

	/* Kick pending requests. */
	mtx_lock(&sc->xb_io_lock);
	sc->connected = BLKIF_STATE_CONNECTED;
	xb_startio(sc);
	sc->xb_flags |= XB_READY;
	mtx_unlock(&sc->xb_io_lock);

}

/**
 * Handle the change of state of the backend to Closing.  We must delete our
 * device-layer structures now, to ensure that writes are flushed through to
 * the backend.  Once this is done, we can switch to Closed in
 * acknowledgement.
 */
static void
blkfront_closing(device_t dev)
{
	struct xb_softc *sc = device_get_softc(dev);

	DPRINTK("blkfront_closing: %s removed\n", xenbus_get_node(dev));

	if (sc->mi) {
		DPRINTK("Calling xlvbd_del\n");
		xlvbd_del(sc);
		sc->mi = NULL;
	}

	xenbus_set_state(dev, XenbusStateClosed);
}


static int
blkfront_detach(device_t dev)
{
	struct xb_softc *sc = device_get_softc(dev);

	DPRINTK("blkfront_remove: %s removed\n", xenbus_get_node(dev));

	blkif_free(sc, 0);
	mtx_destroy(&sc->xb_io_lock);

	return 0;
}


static inline void
flush_requests(struct xb_softc *sc)
{
	int notify;

	RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->ring, notify);

	if (notify)
		notify_remote_via_irq(sc->irq);
}

static void blkif_restart_queue_callback(void *arg)
{
	struct xb_softc *sc = arg;

	xb_startio(sc);
}

static int
blkif_open(struct disk *dp)
{
	struct xb_softc	*sc = (struct xb_softc *)dp->d_drv1;

	if (sc == NULL) {
		printf("xb%d: not found", sc->xb_unit);
		return (ENXIO);
	}

	sc->xb_flags |= XB_OPEN;
	sc->users++;
	return (0);
}

static int
blkif_close(struct disk *dp)
{
	struct xb_softc	*sc = (struct xb_softc *)dp->d_drv1;

	if (sc == NULL)
		return (ENXIO);
	sc->xb_flags &= ~XB_OPEN;
	if (--(sc->users) == 0) {
		/* Check whether we have been instructed to close.  We will
		   have ignored this request initially, as the device was
		   still mounted. */
		device_t dev = sc->xb_dev;
		XenbusState state =
			xenbus_read_driver_state(xenbus_get_otherend_path(dev));

		if (state == XenbusStateClosing)
			blkfront_closing(dev);
	}
	return (0);
}

static int
blkif_ioctl(struct disk *dp, u_long cmd, void *addr, int flag, struct thread *td)
{
	struct xb_softc	*sc = (struct xb_softc *)dp->d_drv1;

	if (sc == NULL)
		return (ENXIO);

	return (ENOTTY);
}

static void
xb_free_command(struct xb_command *cm)
{

	KASSERT((cm->cm_flags & XB_ON_XBQ_MASK) == 0,
	    ("Freeing command that is still on a queue\n"));

	cm->cm_flags = 0;
	cm->bp = NULL;
	cm->cm_complete = NULL;
	xb_enqueue_free(cm);
}

/*
 * blkif_queue_request
 *
 * request block io
 *
 * id: for guest use only.
 * operation: BLKIF_OP_{READ,WRITE,PROBE}
 * buffer: buffer to read/write into. this should be a
 *   virtual address in the guest os.
 */
static struct xb_command *
xb_bio_command(struct xb_softc *sc)
{
	struct xb_command *cm;
	struct bio *bp;

	if (unlikely(sc->connected != BLKIF_STATE_CONNECTED))
		return (NULL);

	bp = xb_dequeue_bio(sc);
	if (bp == NULL)
		return (NULL);

	if ((cm = xb_dequeue_free(sc)) == NULL) {
		xb_requeue_bio(sc, bp);
		return (NULL);
	}

	if (gnttab_alloc_grant_references(BLKIF_MAX_SEGMENTS_PER_REQUEST,
	    &cm->gref_head) < 0) {
		gnttab_request_free_callback(&sc->callback,
			blkif_restart_queue_callback, sc,
			BLKIF_MAX_SEGMENTS_PER_REQUEST);
		xb_requeue_bio(sc, bp);
		xb_enqueue_free(cm);
		sc->xb_flags |= XB_FROZEN;
		return (NULL);
	}

	/* XXX Can we grab refs before doing the load so that the ref can
	 * be filled out here?
	 */
	cm->bp = bp;
	cm->data = bp->bio_data;
	cm->datalen = bp->bio_bcount;
	cm->operation = (bp->bio_cmd == BIO_READ) ? BLKIF_OP_READ :
	    BLKIF_OP_WRITE;
	cm->sector_number = (blkif_sector_t)bp->bio_pblkno;

	return (cm);
}

static int
blkif_queue_request(struct xb_softc *sc, struct xb_command *cm)
{
	int	error;

	error = bus_dmamap_load(sc->xb_io_dmat, cm->map, cm->data, cm->datalen,
	    blkif_queue_cb, cm, 0);
	if (error == EINPROGRESS) {
		printf("EINPROGRESS\n");
		sc->xb_flags |= XB_FROZEN;
		cm->cm_flags |= XB_CMD_FROZEN;
		return (0);
	}

	return (error);
}

static void
blkif_queue_cb(void *arg, bus_dma_segment_t *segs, int nsegs, int error)
{
	struct xb_softc *sc;
	struct xb_command *cm;
	blkif_request_t	*ring_req;
	vm_paddr_t buffer_ma;
	uint64_t fsect, lsect;
	int ref, i, op;

	cm = arg;
	sc = cm->cm_sc;

	if (error) {
		printf("error %d in blkif_queue_cb\n", error);
		cm->bp->bio_error = EIO;
		biodone(cm->bp);
		xb_free_command(cm);
		return;
	}

	/* Fill out a communications ring structure. */
	ring_req = RING_GET_REQUEST(&sc->ring, sc->ring.req_prod_pvt);
	if (ring_req == NULL) {
		/* XXX Is this possible? */
		printf("ring_req NULL, requeuing\n");
		xb_enqueue_ready(cm);
		return;
	}
	ring_req->id = cm->req.id;
	ring_req->operation = cm->operation;
	ring_req->sector_number = cm->sector_number;
	ring_req->handle = (blkif_vdev_t)(uintptr_t)sc->xb_disk;
	ring_req->nr_segments = nsegs;

	for (i = 0; i < nsegs; i++) {
		buffer_ma = segs[i].ds_addr;
		fsect = (buffer_ma & PAGE_MASK) >> XBD_SECTOR_SHFT;
		lsect = fsect + (segs[i].ds_len  >> XBD_SECTOR_SHFT) - 1;

		KASSERT(lsect <= 7,
		    ("XEN disk driver data cannot cross a page boundary"));

		/* install a grant reference. */
		ref = gnttab_claim_grant_reference(&cm->gref_head);
		KASSERT( ref != ENOSPC, ("grant_reference failed") );

		gnttab_grant_foreign_access_ref(
			ref,
			xenbus_get_otherend_id(sc->xb_dev),
			buffer_ma >> PAGE_SHIFT,
			ring_req->operation & 1 ); /* ??? */

		ring_req->seg[i] =
			(struct blkif_request_segment) {
				.gref       = ref,
				.first_sect = fsect,
				.last_sect  = lsect };
	}


	if (cm->operation == BLKIF_OP_READ)
		op = BUS_DMASYNC_PREREAD;
	else if (cm->operation == BLKIF_OP_WRITE)
		op = BUS_DMASYNC_PREWRITE;
	else
		op = 0;
	bus_dmamap_sync(sc->xb_io_dmat, cm->map, op);

	sc->ring.req_prod_pvt++;

	/* Keep a private copy so we can reissue requests when recovering. */
	cm->req = *ring_req;

	xb_enqueue_busy(cm);

	gnttab_free_grant_references(cm->gref_head);

	/*
	 * This flag means that we're probably executing in the busdma swi
	 * instead of in the startio context, so an explicit flush is needed.
	 */
	if (cm->cm_flags & XB_CMD_FROZEN)
		flush_requests(sc);

	return;
}

/*
 * Dequeue buffers and place them in the shared communication ring.
 * Return when no more requests can be accepted or all buffers have
 * been queued.
 *
 * Signal XEN once the ring has been filled out.
 */
static void
xb_startio(struct xb_softc *sc)
{
	struct xb_command *cm;
	int error, queued = 0;

	mtx_assert(&sc->xb_io_lock, MA_OWNED);

	while (!RING_FULL(&sc->ring)) {
		if (sc->xb_flags & XB_FROZEN)
			break;

		cm = xb_dequeue_ready(sc);

		if (cm == NULL)
		    cm = xb_bio_command(sc);

		if (cm == NULL)
			break;

		if ((error = blkif_queue_request(sc, cm)) != 0) {
			printf("blkif_queue_request returned %d\n", error);
			break;
		}
		queued++;
	}

	if (queued != 0)
		flush_requests(sc);
}

static void
blkif_int(void *xsc)
{
	struct xb_softc *sc = xsc;
	struct xb_command *cm;
	blkif_response_t *bret;
	RING_IDX i, rp;
	int op;

	mtx_lock(&sc->xb_io_lock);

	if (unlikely(sc->connected != BLKIF_STATE_CONNECTED)) {
		mtx_unlock(&sc->xb_io_lock);
		return;
	}

 again:
	rp = sc->ring.sring->rsp_prod;
	rmb(); /* Ensure we see queued responses up to 'rp'. */

	for (i = sc->ring.rsp_cons; i != rp; i++) {
		bret = RING_GET_RESPONSE(&sc->ring, i);
		cm   = &sc->shadow[bret->id];

		xb_remove_busy(cm);
		blkif_completion(cm);

		if (cm->operation == BLKIF_OP_READ)
			op = BUS_DMASYNC_POSTREAD;
		else if (cm->operation == BLKIF_OP_WRITE)
			op = BUS_DMASYNC_POSTWRITE;
		else
			op = 0;
		bus_dmamap_sync(sc->xb_io_dmat, cm->map, op);
		bus_dmamap_unload(sc->xb_io_dmat, cm->map);

		/*
		 * If commands are completing then resources are probably
		 * being freed as well.  It's a cheap assumption even when
		 * wrong.
		 */
		sc->xb_flags &= ~XB_FROZEN;

		/*
		 * Directly call the i/o complete routine to save an
		 * an indirection in the common case.
		 */
		cm->status = bret->status;
		if (cm->bp)
			xb_bio_complete(sc, cm);
		else if (cm->cm_complete)
			(cm->cm_complete)(cm);
		else
			xb_free_command(cm);
	}

	sc->ring.rsp_cons = i;

	if (i != sc->ring.req_prod_pvt) {
		int more_to_do;
		RING_FINAL_CHECK_FOR_RESPONSES(&sc->ring, more_to_do);
		if (more_to_do)
			goto again;
	} else {
		sc->ring.sring->rsp_event = i + 1;
	}

	xb_startio(sc);

	mtx_unlock(&sc->xb_io_lock);
}

static void
blkif_free(struct xb_softc *sc, int suspend)
{

/* Prevent new requests being issued until we fix things up. */
	mtx_lock(&sc->xb_io_lock);
	sc->connected = suspend ?
		BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED;
	mtx_unlock(&sc->xb_io_lock);

	/* Free resources associated with old device channel. */
	if (sc->ring_ref != GRANT_INVALID_REF) {
		gnttab_end_foreign_access(sc->ring_ref,
					  sc->ring.sring);
		sc->ring_ref = GRANT_INVALID_REF;
		sc->ring.sring = NULL;
	}
	if (sc->irq)
		unbind_from_irqhandler(sc->irq);
	sc->irq = 0;

}

static void
blkif_completion(struct xb_command *s)
{
	int i;

	for (i = 0; i < s->req.nr_segments; i++)
		gnttab_end_foreign_access(s->req.seg[i].gref, 0UL);
}

static void
blkif_recover(struct xb_softc *sc)
{
	/*
	 * XXX The whole concept of not quiescing and completing all i/o
	 * during suspend, and then hoping to recover and replay the
	 * resulting abandoned I/O during resume, is laughable.  At best,
	 * it invalidates the i/o ordering rules required by just about
	 * every filesystem, and at worst it'll corrupt data.  The code
	 * has been removed until further notice.
	 */
}

/* ** Driver registration ** */
static device_method_t blkfront_methods[] = {
	/* Device interface */
	DEVMETHOD(device_probe,         blkfront_probe),
	DEVMETHOD(device_attach,        blkfront_attach),
	DEVMETHOD(device_detach,        blkfront_detach),
	DEVMETHOD(device_shutdown,      bus_generic_shutdown),
	DEVMETHOD(device_suspend,       blkfront_suspend),
	DEVMETHOD(device_resume,        blkfront_resume),

	/* Xenbus interface */
	DEVMETHOD(xenbus_backend_changed, blkfront_backend_changed),

	{ 0, 0 }
};

static driver_t blkfront_driver = {
	"xbd",
	blkfront_methods,
	sizeof(struct xb_softc),
};
devclass_t blkfront_devclass;

DRIVER_MODULE(xbd, xenbus, blkfront_driver, blkfront_devclass, 0, 0);