xen: add block device backend driver. (Gerd Hoffmann)

This patch adds a block device backend driver to qemu.  It is a pure
userspace implemention using the gntdev interface.  It uses "qdisk" as
backend name in xenstore so it doesn't interfere with the other existing
backends (blkback aka "vbd" and tapdisk aka "tap").

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>


git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7223 c046a42c-6fe2-441c-8c8c-71466251a162

7 files changed, 892 insertions, 3 deletions

diff --git a/Makefile.target b/Makefile.target
index 8734190279..ea3b9d5f65 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -562,7 +562,7 @@ endif
 
 # xen backend driver support
 XEN_OBJS := xen_machine_pv.o xen_backend.o
-XEN_OBJS += xen_console.o xenfb.o
+XEN_OBJS += xen_console.o xenfb.o xen_disk.o
 ifeq ($(CONFIG_XEN), yes)
   OBJS += $(XEN_OBJS)
   LIBS += $(XEN_LIBS)
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
index e9a4e2dd2d..dd426ddb4f 100644
--- a/hw/xen_backend.h
+++ b/hw/xen_backend.h
@@ -2,6 +2,7 @@
 #define QEMU_HW_XEN_BACKEND_H 1
 
 #include "xen_common.h"
+#include "sysemu.h"
 
 /* ------------------------------------------------------------- */
 
@@ -87,6 +88,7 @@ void xen_be_printf(struct XenDevice *xendev, int msg_level, const char *fmt, ...
 extern struct XenDevOps xen_console_ops;      /* xen_console.c     */
 extern struct XenDevOps xen_kbdmouse_ops;     /* xen_framebuffer.c */
 extern struct XenDevOps xen_framebuffer_ops;  /* xen_framebuffer.c */
+extern struct XenDevOps xen_blkdev_ops;       /* xen_disk.c        */
 
 void xen_init_display(int domid);
 
diff --git a/hw/xen_blkif.h b/hw/xen_blkif.h
new file mode 100644
index 0000000000..254a5fd50c
--- /dev/null
+++ b/hw/xen_blkif.h
@@ -0,0 +1,103 @@
+#ifndef __XEN_BLKIF_H__
+#define __XEN_BLKIF_H__
+
+#include <xen/io/ring.h>
+#include <xen/io/blkif.h>
+#include <xen/io/protocols.h>
+
+/* Not a real protocol.  Used to generate ring structs which contain
+ * the elements common to all protocols only.  This way we get a
+ * compiler-checkable way to use common struct elements, so we can
+ * avoid using switch(protocol) in a number of places.  */
+struct blkif_common_request {
+	char dummy;
+};
+struct blkif_common_response {
+	char dummy;
+};
+
+/* i386 protocol version */
+#pragma pack(push, 4)
+struct blkif_x86_32_request {
+	uint8_t        operation;    /* BLKIF_OP_???                         */
+	uint8_t        nr_segments;  /* number of segments                   */
+	blkif_vdev_t   handle;       /* only for read/write requests         */
+	uint64_t       id;           /* private guest value, echoed in resp  */
+	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
+	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+};
+struct blkif_x86_32_response {
+	uint64_t        id;              /* copied from request */
+	uint8_t         operation;       /* copied from request */
+	int16_t         status;          /* BLKIF_RSP_???       */
+};
+typedef struct blkif_x86_32_request blkif_x86_32_request_t;
+typedef struct blkif_x86_32_response blkif_x86_32_response_t;
+#pragma pack(pop)
+
+/* x86_64 protocol version */
+struct blkif_x86_64_request {
+	uint8_t        operation;    /* BLKIF_OP_???                         */
+	uint8_t        nr_segments;  /* number of segments                   */
+	blkif_vdev_t   handle;       /* only for read/write requests         */
+	uint64_t       __attribute__((__aligned__(8))) id;
+	blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
+	struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+};
+struct blkif_x86_64_response {
+	uint64_t       __attribute__((__aligned__(8))) id;
+	uint8_t         operation;       /* copied from request */
+	int16_t         status;          /* BLKIF_RSP_???       */
+};
+typedef struct blkif_x86_64_request blkif_x86_64_request_t;
+typedef struct blkif_x86_64_response blkif_x86_64_response_t;
+
+DEFINE_RING_TYPES(blkif_common, struct blkif_common_request, struct blkif_common_response);
+DEFINE_RING_TYPES(blkif_x86_32, struct blkif_x86_32_request, struct blkif_x86_32_response);
+DEFINE_RING_TYPES(blkif_x86_64, struct blkif_x86_64_request, struct blkif_x86_64_response);
+
+union blkif_back_rings {
+	blkif_back_ring_t        native;
+	blkif_common_back_ring_t common;
+	blkif_x86_32_back_ring_t x86_32;
+	blkif_x86_64_back_ring_t x86_64;
+};
+typedef union blkif_back_rings blkif_back_rings_t;
+
+enum blkif_protocol {
+	BLKIF_PROTOCOL_NATIVE = 1,
+	BLKIF_PROTOCOL_X86_32 = 2,
+	BLKIF_PROTOCOL_X86_64 = 3,
+};
+
+static void inline blkif_get_x86_32_req(blkif_request_t *dst, blkif_x86_32_request_t *src)
+{
+	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
+
+	dst->operation = src->operation;
+	dst->nr_segments = src->nr_segments;
+	dst->handle = src->handle;
+	dst->id = src->id;
+	dst->sector_number = src->sector_number;
+	if (n > src->nr_segments)
+		n = src->nr_segments;
+	for (i = 0; i < n; i++)
+		dst->seg[i] = src->seg[i];
+}
+
+static void inline blkif_get_x86_64_req(blkif_request_t *dst, blkif_x86_64_request_t *src)
+{
+	int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
+
+	dst->operation = src->operation;
+	dst->nr_segments = src->nr_segments;
+	dst->handle = src->handle;
+	dst->id = src->id;
+	dst->sector_number = src->sector_number;
+	if (n > src->nr_segments)
+		n = src->nr_segments;
+	for (i = 0; i < n; i++)
+		dst->seg[i] = src->seg[i];
+}
+
+#endif /* __XEN_BLKIF_H__ */
diff --git a/hw/xen_disk.c b/hw/xen_disk.c
new file mode 100644
index 0000000000..527f84b3e5
--- /dev/null
+++ b/hw/xen_disk.c
@@ -0,0 +1,779 @@
+/*
+ *  xen paravirt block device backend
+ *
+ *  (c) Gerd Hoffmann <kraxel@redhat.com>
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; under version 2 of the License.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License along
+ *  with this program; if not, write to the Free Software Foundation, Inc.,
+ *  51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <unistd.h>
+#include <signal.h>
+#include <inttypes.h>
+#include <time.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <sys/uio.h>
+
+#include <xs.h>
+#include <xenctrl.h>
+#include <xen/io/xenbus.h>
+
+#include "hw.h"
+#include "block_int.h"
+#include "qemu-char.h"
+#include "xen_blkif.h"
+#include "xen_backend.h"
+
+/* ------------------------------------------------------------- */
+
+static int syncwrite    = 0;
+static int batch_maps   = 0;
+
+static int max_requests = 32;
+static int use_aio      = 1;
+
+/* ------------------------------------------------------------- */
+
+#define BLOCK_SIZE  512
+#define IOCB_COUNT  (BLKIF_MAX_SEGMENTS_PER_REQUEST + 2)
+
+struct ioreq {
+    blkif_request_t     req;
+    int16_t             status;
+
+    /* parsed request */
+    off_t               start;
+    QEMUIOVector        v;
+    int                 presync;
+    int                 postsync;
+
+    /* grant mapping */
+    uint32_t            domids[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+    uint32_t            refs[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+    int                 prot;
+    void                *page[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+    void                *pages;
+
+    /* aio status */
+    int                 aio_inflight;
+    int                 aio_errors;
+
+    struct XenBlkDev    *blkdev;
+    LIST_ENTRY(ioreq)   list;
+};
+
+struct XenBlkDev {
+    struct XenDevice    xendev;  /* must be first */
+    char                *params;
+    char                *mode;
+    char                *type;
+    char                *dev;
+    char                *devtype;
+    const char          *fileproto;
+    const char          *filename;
+    int                 ring_ref;
+    void                *sring;
+    int64_t             file_blk;
+    int64_t             file_size;
+    int                 protocol;
+    blkif_back_rings_t  rings;
+    int                 more_work;
+    int                 cnt_map;
+
+    /* request lists */
+    LIST_HEAD(inflight_head, ioreq) inflight;
+    LIST_HEAD(finished_head, ioreq) finished;
+    LIST_HEAD(freelist_head, ioreq) freelist;
+    int                 requests_total;
+    int                 requests_inflight;
+    int                 requests_finished;
+
+    /* qemu block driver */
+    int                 index;
+    BlockDriverState    *bs;
+    QEMUBH              *bh;
+};
+
+/* ------------------------------------------------------------- */
+
+static struct ioreq *ioreq_start(struct XenBlkDev *blkdev)
+{
+    struct ioreq *ioreq = NULL;
+
+    if (LIST_EMPTY(&blkdev->freelist)) {
+	if (blkdev->requests_total >= max_requests)
+	    goto out;
+	/* allocate new struct */
+	ioreq = qemu_mallocz(sizeof(*ioreq));
+	ioreq->blkdev = blkdev;
+	blkdev->requests_total++;
+        qemu_iovec_init(&ioreq->v, BLKIF_MAX_SEGMENTS_PER_REQUEST);
+    } else {
+	/* get one from freelist */
+	ioreq = LIST_FIRST(&blkdev->freelist);
+	LIST_REMOVE(ioreq, list);
+        qemu_iovec_reset(&ioreq->v);
+    }
+    LIST_INSERT_HEAD(&blkdev->inflight, ioreq, list);
+    blkdev->requests_inflight++;
+
+out:
+    return ioreq;
+}
+
+static void ioreq_finish(struct ioreq *ioreq)
+{
+    struct XenBlkDev *blkdev = ioreq->blkdev;
+
+    LIST_REMOVE(ioreq, list);
+    LIST_INSERT_HEAD(&blkdev->finished, ioreq, list);
+    blkdev->requests_inflight--;
+    blkdev->requests_finished++;
+}
+
+static void ioreq_release(struct ioreq *ioreq)
+{
+    struct XenBlkDev *blkdev = ioreq->blkdev;
+
+    LIST_REMOVE(ioreq, list);
+    memset(ioreq, 0, sizeof(*ioreq));
+    ioreq->blkdev = blkdev;
+    LIST_INSERT_HEAD(&blkdev->freelist, ioreq, list);
+    blkdev->requests_finished--;
+}
+
+/*
+ * translate request into iovec + start offset
+ * do sanity checks along the way
+ */
+static int ioreq_parse(struct ioreq *ioreq)
+{
+    struct XenBlkDev *blkdev = ioreq->blkdev;
+    uintptr_t mem;
+    size_t len;
+    int i;
+
+    xen_be_printf(&blkdev->xendev, 3,
+		  "op %d, nr %d, handle %d, id %" PRId64 ", sector %" PRId64 "\n",
+		  ioreq->req.operation, ioreq->req.nr_segments,
+		  ioreq->req.handle, ioreq->req.id, ioreq->req.sector_number);
+    switch (ioreq->req.operation) {
+    case BLKIF_OP_READ:
+	ioreq->prot = PROT_WRITE; /* to memory */
+	if (BLKIF_OP_READ != ioreq->req.operation && blkdev->mode[0] != 'w') {
+	    xen_be_printf(&blkdev->xendev, 0, "error: write req for ro device\n");
+	    goto err;
+	}
+	break;
+    case BLKIF_OP_WRITE_BARRIER:
+	if (!syncwrite)
+	    ioreq->presync = ioreq->postsync = 1;
+	/* fall through */
+    case BLKIF_OP_WRITE:
+	ioreq->prot = PROT_READ; /* from memory */
+	if (syncwrite)
+	    ioreq->postsync = 1;
+	break;
+    default:
+	xen_be_printf(&blkdev->xendev, 0, "error: unknown operation (%d)\n",
+		      ioreq->req.operation);
+	goto err;
+    };
+
+    ioreq->start = ioreq->req.sector_number * blkdev->file_blk;
+    for (i = 0; i < ioreq->req.nr_segments; i++) {
+	if (i == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
+	    xen_be_printf(&blkdev->xendev, 0, "error: nr_segments too big\n");
+	    goto err;
+	}
+	if (ioreq->req.seg[i].first_sect > ioreq->req.seg[i].last_sect) {
+	    xen_be_printf(&blkdev->xendev, 0, "error: first > last sector\n");
+	    goto err;
+	}
+	if (ioreq->req.seg[i].last_sect * BLOCK_SIZE >= XC_PAGE_SIZE) {
+	    xen_be_printf(&blkdev->xendev, 0, "error: page crossing\n");
+	    goto err;
+	}
+
+	ioreq->domids[i] = blkdev->xendev.dom;
+	ioreq->refs[i]   = ioreq->req.seg[i].gref;
+
+	mem = ioreq->req.seg[i].first_sect * blkdev->file_blk;
+	len = (ioreq->req.seg[i].last_sect - ioreq->req.seg[i].first_sect + 1) * blkdev->file_blk;
+        qemu_iovec_add(&ioreq->v, (void*)mem, len);
+    }
+    if (ioreq->start + ioreq->v.size > blkdev->file_size) {
+	xen_be_printf(&blkdev->xendev, 0, "error: access beyond end of file\n");
+	goto err;
+    }
+    return 0;
+
+err:
+    ioreq->status = BLKIF_RSP_ERROR;
+    return -1;
+}
+
+static void ioreq_unmap(struct ioreq *ioreq)
+{
+    int gnt = ioreq->blkdev->xendev.gnttabdev;
+    int i;
+
+    if (ioreq->v.niov == 0)
+        return;
+    if (batch_maps) {
+	if (!ioreq->pages)
+	    return;
+	if (xc_gnttab_munmap(gnt, ioreq->pages, ioreq->v.niov) != 0)
+	    xen_be_printf(&ioreq->blkdev->xendev, 0, "xc_gnttab_munmap failed: %s\n",
+			  strerror(errno));
+	ioreq->blkdev->cnt_map -= ioreq->v.niov;
+	ioreq->pages = NULL;
+    } else {
+	for (i = 0; i < ioreq->v.niov; i++) {
+	    if (!ioreq->page[i])
+		continue;
+	    if (xc_gnttab_munmap(gnt, ioreq->page[i], 1) != 0)
+		xen_be_printf(&ioreq->blkdev->xendev, 0, "xc_gnttab_munmap failed: %s\n",
+			      strerror(errno));
+	    ioreq->blkdev->cnt_map--;
+	    ioreq->page[i] = NULL;
+	}
+    }
+}
+
+static int ioreq_map(struct ioreq *ioreq)
+{
+    int gnt = ioreq->blkdev->xendev.gnttabdev;
+    int i;
+
+    if (ioreq->v.niov == 0)
+        return 0;
+    if (batch_maps) {
+	ioreq->pages = xc_gnttab_map_grant_refs
+	    (gnt, ioreq->v.niov, ioreq->domids, ioreq->refs, ioreq->prot);
+	if (ioreq->pages == NULL) {
+	    xen_be_printf(&ioreq->blkdev->xendev, 0,
+			  "can't map %d grant refs (%s, %d maps)\n",
+			  ioreq->v.niov, strerror(errno), ioreq->blkdev->cnt_map);
+	    return -1;
+	}
+	for (i = 0; i < ioreq->v.niov; i++)
+	    ioreq->v.iov[i].iov_base = ioreq->pages + i * XC_PAGE_SIZE +
+		(uintptr_t)ioreq->v.iov[i].iov_base;
+	ioreq->blkdev->cnt_map += ioreq->v.niov;
+    } else  {
+	for (i = 0; i < ioreq->v.niov; i++) {
+	    ioreq->page[i] = xc_gnttab_map_grant_ref
+		(gnt, ioreq->domids[i], ioreq->refs[i], ioreq->prot);
+	    if (ioreq->page[i] == NULL) {
+		xen_be_printf(&ioreq->blkdev->xendev, 0,
+			      "can't map grant ref %d (%s, %d maps)\n",
+			      ioreq->refs[i], strerror(errno), ioreq->blkdev->cnt_map);
+		ioreq_unmap(ioreq);
+		return -1;
+	    }
+	    ioreq->v.iov[i].iov_base = ioreq->page[i] + (uintptr_t)ioreq->v.iov[i].iov_base;
+	    ioreq->blkdev->cnt_map++;
+	}
+    }
+    return 0;
+}
+
+static int ioreq_runio_qemu_sync(struct ioreq *ioreq)
+{
+    struct XenBlkDev *blkdev = ioreq->blkdev;
+    int i, rc, len = 0;
+    off_t pos;
+
+    if (ioreq_map(ioreq) == -1)
+	goto err;
+    if (ioreq->presync)
+	bdrv_flush(blkdev->bs);
+
+    switch (ioreq->req.operation) {
+    case BLKIF_OP_READ:
+	pos = ioreq->start;
+	for (i = 0; i < ioreq->v.niov; i++) {
+	    rc = bdrv_read(blkdev->bs, pos / BLOCK_SIZE,
+			   ioreq->v.iov[i].iov_base,
+			   ioreq->v.iov[i].iov_len / BLOCK_SIZE);
+	    if (rc != 0) {
+		xen_be_printf(&blkdev->xendev, 0, "rd I/O error (%p, len %zd)\n",
+			      ioreq->v.iov[i].iov_base,
+			      ioreq->v.iov[i].iov_len);
+		goto err;
+	    }
+	    len += ioreq->v.iov[i].iov_len;
+	    pos += ioreq->v.iov[i].iov_len;
+	}
+	break;
+    case BLKIF_OP_WRITE:
+    case BLKIF_OP_WRITE_BARRIER:
+	pos = ioreq->start;
+	for (i = 0; i < ioreq->v.niov; i++) {
+	    rc = bdrv_write(blkdev->bs, pos / BLOCK_SIZE,
+			    ioreq->v.iov[i].iov_base,
+			    ioreq->v.iov[i].iov_len / BLOCK_SIZE);
+	    if (rc != 0) {
+		xen_be_printf(&blkdev->xendev, 0, "wr I/O error (%p, len %zd)\n",
+			      ioreq->v.iov[i].iov_base,
+			      ioreq->v.iov[i].iov_len);
+		goto err;
+	    }
+	    len += ioreq->v.iov[i].iov_len;
+	    pos += ioreq->v.iov[i].iov_len;
+	}
+	break;
+    default:
+	/* unknown operation (shouldn't happen -- parse catches this) */
+	goto err;
+    }
+
+    if (ioreq->postsync)
+	bdrv_flush(blkdev->bs);
+    ioreq->status = BLKIF_RSP_OKAY;
+
+    ioreq_unmap(ioreq);
+    ioreq_finish(ioreq);
+    return 0;
+
+err:
+    ioreq->status = BLKIF_RSP_ERROR;
+    return -1;
+}
+
+static void qemu_aio_complete(void *opaque, int ret)
+{
+    struct ioreq *ioreq = opaque;
+
+    if (ret != 0) {
+        xen_be_printf(&ioreq->blkdev->xendev, 0, "%s I/O error\n",
+                      ioreq->req.operation == BLKIF_OP_READ ? "read" : "write");
+        ioreq->aio_errors++;
+    }
+
+    ioreq->aio_inflight--;
+    if (ioreq->aio_inflight > 0)
+        return;
+
+    ioreq->status = ioreq->aio_errors ? BLKIF_RSP_ERROR : BLKIF_RSP_OKAY;
+    ioreq_unmap(ioreq);
+    ioreq_finish(ioreq);
+    qemu_bh_schedule(ioreq->blkdev->bh);
+}
+
+static int ioreq_runio_qemu_aio(struct ioreq *ioreq)
+{
+    struct XenBlkDev *blkdev = ioreq->blkdev;
+
+    if (ioreq_map(ioreq) == -1)
+	goto err;
+
+    ioreq->aio_inflight++;
+    if (ioreq->presync)
+	bdrv_flush(blkdev->bs); /* FIXME: aio_flush() ??? */
+
+    switch (ioreq->req.operation) {
+    case BLKIF_OP_READ:
+        ioreq->aio_inflight++;
+        bdrv_aio_readv(blkdev->bs, ioreq->start / BLOCK_SIZE,
+                       &ioreq->v, ioreq->v.size / BLOCK_SIZE,
+                       qemu_aio_complete, ioreq);
+	break;
+    case BLKIF_OP_WRITE:
+    case BLKIF_OP_WRITE_BARRIER:
+        ioreq->aio_inflight++;
+        bdrv_aio_writev(blkdev->bs, ioreq->start / BLOCK_SIZE,
+                        &ioreq->v, ioreq->v.size / BLOCK_SIZE,
+                        qemu_aio_complete, ioreq);
+	break;
+    default:
+	/* unknown operation (shouldn't happen -- parse catches this) */
+	goto err;
+    }
+
+    if (ioreq->postsync)
+	bdrv_flush(blkdev->bs); /* FIXME: aio_flush() ??? */
+    qemu_aio_complete(ioreq, 0);
+
+    return 0;
+
+err:
+    ioreq->status = BLKIF_RSP_ERROR;
+    return -1;
+}
+
+static int blk_send_response_one(struct ioreq *ioreq)
+{
+    struct XenBlkDev  *blkdev = ioreq->blkdev;
+    int               send_notify   = 0;
+    int               have_requests = 0;
+    blkif_response_t  resp;
+    void              *dst;
+
+    resp.id        = ioreq->req.id;
+    resp.operation = ioreq->req.operation;
+    resp.status    = ioreq->status;
+
+    /* Place on the response ring for the relevant domain. */
+    switch (blkdev->protocol) {
+    case BLKIF_PROTOCOL_NATIVE:
+	dst = RING_GET_RESPONSE(&blkdev->rings.native, blkdev->rings.native.rsp_prod_pvt);
+	break;
+    case BLKIF_PROTOCOL_X86_32:
+	dst = RING_GET_RESPONSE(&blkdev->rings.x86_32, blkdev->rings.x86_32.rsp_prod_pvt);
+	break;
+    case BLKIF_PROTOCOL_X86_64:
+	dst = RING_GET_RESPONSE(&blkdev->rings.x86_64, blkdev->rings.x86_64.rsp_prod_pvt);
+	break;
+    default:
+	dst = NULL;
+    }
+    memcpy(dst, &resp, sizeof(resp));
+    blkdev->rings.common.rsp_prod_pvt++;
+
+    RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&blkdev->rings.common, send_notify);
+    if (blkdev->rings.common.rsp_prod_pvt == blkdev->rings.common.req_cons) {
+	/*
+	 * Tail check for pending requests. Allows frontend to avoid
+	 * notifications if requests are already in flight (lower
+	 * overheads and promotes batching).
+	 */
+	RING_FINAL_CHECK_FOR_REQUESTS(&blkdev->rings.common, have_requests);
+    } else if (RING_HAS_UNCONSUMED_REQUESTS(&blkdev->rings.common)) {
+	have_requests = 1;
+    }
+
+    if (have_requests)
+	blkdev->more_work++;
+    return send_notify;
+}
+
+/* walk finished list, send outstanding responses, free requests */
+static void blk_send_response_all(struct XenBlkDev *blkdev)
+{
+    struct ioreq *ioreq;
+    int send_notify = 0;
+
+    while (!LIST_EMPTY(&blkdev->finished)) {
+        ioreq = LIST_FIRST(&blkdev->finished);
+	send_notify += blk_send_response_one(ioreq);
+	ioreq_release(ioreq);
+    }
+    if (send_notify)
+	xen_be_send_notify(&blkdev->xendev);
+}
+
+static int blk_get_request(struct XenBlkDev *blkdev, struct ioreq *ioreq, RING_IDX rc)
+{
+    switch (blkdev->protocol) {
+    case BLKIF_PROTOCOL_NATIVE:
+	memcpy(&ioreq->req, RING_GET_REQUEST(&blkdev->rings.native, rc),
+	       sizeof(ioreq->req));
+	break;
+    case BLKIF_PROTOCOL_X86_32:
+	blkif_get_x86_32_req(&ioreq->req, RING_GET_REQUEST(&blkdev->rings.x86_32, rc));
+	break;
+    case BLKIF_PROTOCOL_X86_64:
+	blkif_get_x86_64_req(&ioreq->req, RING_GET_REQUEST(&blkdev->rings.x86_64, rc));
+	break;
+    }
+    return 0;
+}
+
+static void blk_handle_requests(struct XenBlkDev *blkdev)
+{
+    RING_IDX rc, rp;
+    struct ioreq *ioreq;
+
+    blkdev->more_work = 0;
+
+    rc = blkdev->rings.common.req_cons;
+    rp = blkdev->rings.common.sring->req_prod;
+    xen_rmb(); /* Ensure we see queued requests up to 'rp'. */
+
+    if (use_aio)
+        blk_send_response_all(blkdev);
+    while ((rc != rp)) {
+        /* pull request from ring */
+        if (RING_REQUEST_CONS_OVERFLOW(&blkdev->rings.common, rc))
+            break;
+        ioreq = ioreq_start(blkdev);
+        if (ioreq == NULL) {
+            blkdev->more_work++;
+            break;
+        }
+        blk_get_request(blkdev, ioreq, rc);
+        blkdev->rings.common.req_cons = ++rc;
+
+        /* parse them */
+        if (ioreq_parse(ioreq) != 0) {
+            if (blk_send_response_one(ioreq))
+                xen_be_send_notify(&blkdev->xendev);
+            ioreq_release(ioreq);
+            continue;
+        }
+
+        if (use_aio) {
+            /* run i/o in aio mode */
+            ioreq_runio_qemu_aio(ioreq);
+        } else {
+            /* run i/o in sync mode */
+            ioreq_runio_qemu_sync(ioreq);
+        }
+    }
+    if (!use_aio)
+        blk_send_response_all(blkdev);
+
+    if (blkdev->more_work && blkdev->requests_inflight < max_requests)
+        qemu_bh_schedule(blkdev->bh);
+}
+
+/* ------------------------------------------------------------- */
+
+static void blk_bh(void *opaque)
+{
+    struct XenBlkDev *blkdev = opaque;
+    blk_handle_requests(blkdev);
+}
+
+static void blk_alloc(struct XenDevice *xendev)
+{
+    struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+    LIST_INIT(&blkdev->inflight);
+    LIST_INIT(&blkdev->finished);
+    LIST_INIT(&blkdev->freelist);
+    blkdev->bh = qemu_bh_new(blk_bh, blkdev);
+    if (xen_mode != XEN_EMULATE)
+        batch_maps = 1;
+}
+
+static int blk_init(struct XenDevice *xendev)
+{
+    struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+    int mode, qflags, have_barriers, info = 0;
+    char *h;
+
+    /* read xenstore entries */
+    if (blkdev->params == NULL) {
+	blkdev->params = xenstore_read_be_str(&blkdev->xendev, "params");
+        h = strchr(blkdev->params, ':');
+	if (h != NULL) {
+	    blkdev->fileproto = blkdev->params;
+	    blkdev->filename  = h+1;
+	    *h = 0;
+	} else {
+	    blkdev->fileproto = "<unset>";
+	    blkdev->filename  = blkdev->params;
+	}
+    }
+    if (blkdev->mode == NULL)
+	blkdev->mode = xenstore_read_be_str(&blkdev->xendev, "mode");
+    if (blkdev->type == NULL)
+	blkdev->type = xenstore_read_be_str(&blkdev->xendev, "type");
+    if (blkdev->dev == NULL)
+	blkdev->dev = xenstore_read_be_str(&blkdev->xendev, "dev");
+    if (blkdev->devtype == NULL)
+	blkdev->devtype = xenstore_read_be_str(&blkdev->xendev, "device-type");
+
+    /* do we have all we need? */
+    if (blkdev->params == NULL ||
+	blkdev->mode == NULL   ||
+	blkdev->type == NULL   ||
+	blkdev->dev == NULL)
+	return -1;
+
+    /* read-only ? */
+    if (strcmp(blkdev->mode, "w") == 0) {
+	mode   = O_RDWR;
+	qflags = BDRV_O_RDWR;
+    } else {
+	mode   = O_RDONLY;
+	qflags = BDRV_O_RDONLY;
+	info  |= VDISK_READONLY;
+    }
+
+    /* cdrom ? */
+    if (blkdev->devtype && !strcmp(blkdev->devtype, "cdrom"))
+	info  |= VDISK_CDROM;
+
+    /* init qemu block driver */
+    blkdev->index = (blkdev->xendev.dev - 202 * 256) / 16;
+    blkdev->index = drive_get_index(IF_XEN, 0, blkdev->index);
+    if (blkdev->index == -1) {
+        /* setup via xenbus -> create new block driver instance */
+        xen_be_printf(&blkdev->xendev, 2, "create new bdrv (xenbus setup)\n");
+	blkdev->bs = bdrv_new(blkdev->dev);
+	if (blkdev->bs) {
+	    if (bdrv_open2(blkdev->bs, blkdev->filename, qflags,
+                           bdrv_find_format(blkdev->fileproto)) != 0) {
+		bdrv_delete(blkdev->bs);
+		blkdev->bs = NULL;
+	    }
+	}
+	if (!blkdev->bs)
+	    return -1;
+    } else {
+        /* setup via qemu cmdline -> already setup for us */
+        xen_be_printf(&blkdev->xendev, 2, "get configured bdrv (cmdline setup)\n");
+	blkdev->bs = drives_table[blkdev->index].bdrv;
+    }
+    blkdev->file_blk  = BLOCK_SIZE;
+    blkdev->file_size = bdrv_getlength(blkdev->bs);
+    if (blkdev->file_size < 0) {
+        xen_be_printf(&blkdev->xendev, 1, "bdrv_getlength: %d (%s) | drv %s\n",
+                      (int)blkdev->file_size, strerror(-blkdev->file_size),
+                      blkdev->bs->drv ? blkdev->bs->drv->format_name : "-");
+	blkdev->file_size = 0;
+    }
+    have_barriers = blkdev->bs->drv && blkdev->bs->drv->bdrv_flush ? 1 : 0;
+
+    xen_be_printf(xendev, 1, "type \"%s\", fileproto \"%s\", filename \"%s\","
+		  " size %" PRId64 " (%" PRId64 " MB)\n",
+		  blkdev->type, blkdev->fileproto, blkdev->filename,
+		  blkdev->file_size, blkdev->file_size >> 20);
+
+    /* fill info */
+    xenstore_write_be_int(&blkdev->xendev, "feature-barrier", have_barriers);
+    xenstore_write_be_int(&blkdev->xendev, "info",            info);
+    xenstore_write_be_int(&blkdev->xendev, "sector-size",     blkdev->file_blk);
+    xenstore_write_be_int(&blkdev->xendev, "sectors",
+			  blkdev->file_size / blkdev->file_blk);
+    return 0;
+}
+
+static int blk_connect(struct XenDevice *xendev)
+{
+    struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+    if (xenstore_read_fe_int(&blkdev->xendev, "ring-ref", &blkdev->ring_ref) == -1)
+	return -1;
+    if (xenstore_read_fe_int(&blkdev->xendev, "event-channel",
+                             &blkdev->xendev.remote_port) == -1)
+	return -1;
+
+    blkdev->protocol = BLKIF_PROTOCOL_NATIVE;
+    if (blkdev->xendev.protocol) {
+        if (strcmp(blkdev->xendev.protocol, XEN_IO_PROTO_ABI_X86_32) == 0)
+            blkdev->protocol = BLKIF_PROTOCOL_X86_32;
+        if (strcmp(blkdev->xendev.protocol, XEN_IO_PROTO_ABI_X86_64) == 0)
+            blkdev->protocol = BLKIF_PROTOCOL_X86_64;
+    }
+
+    blkdev->sring = xc_gnttab_map_grant_ref(blkdev->xendev.gnttabdev,
+					    blkdev->xendev.dom,
+					    blkdev->ring_ref,
+					    PROT_READ | PROT_WRITE);
+    if (!blkdev->sring)
+	return -1;
+    blkdev->cnt_map++;
+
+    switch (blkdev->protocol) {
+    case BLKIF_PROTOCOL_NATIVE:
+    {
+	blkif_sring_t *sring_native = blkdev->sring;
+	BACK_RING_INIT(&blkdev->rings.native, sring_native, XC_PAGE_SIZE);
+	break;
+    }
+    case BLKIF_PROTOCOL_X86_32:
+    {
+	blkif_x86_32_sring_t *sring_x86_32 = blkdev->sring;
+	BACK_RING_INIT(&blkdev->rings.x86_32, sring_x86_32, XC_PAGE_SIZE);
+	break;
+    }
+    case BLKIF_PROTOCOL_X86_64:
+    {
+	blkif_x86_64_sring_t *sring_x86_64 = blkdev->sring;
+	BACK_RING_INIT(&blkdev->rings.x86_64, sring_x86_64, XC_PAGE_SIZE);
+	break;
+    }
+    }
+
+    xen_be_bind_evtchn(&blkdev->xendev);
+
+    xen_be_printf(&blkdev->xendev, 1, "ok: proto %s, ring-ref %d, "
+		  "remote port %d, local port %d\n",
+		  blkdev->xendev.protocol, blkdev->ring_ref,
+		  blkdev->xendev.remote_port, blkdev->xendev.local_port);
+    return 0;
+}
+
+static void blk_disconnect(struct XenDevice *xendev)
+{
+    struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+    if (blkdev->bs) {
+        if (blkdev->index == -1) {
+            /* close/delete only if we created it ourself */
+            bdrv_close(blkdev->bs);
+            bdrv_delete(blkdev->bs);
+        }
+	blkdev->bs = NULL;
+    }
+    xen_be_unbind_evtchn(&blkdev->xendev);
+
+    if (blkdev->sring) {
+	xc_gnttab_munmap(blkdev->xendev.gnttabdev, blkdev->sring, 1);
+	blkdev->cnt_map--;
+	blkdev->sring = NULL;
+    }
+}
+
+static int blk_free(struct XenDevice *xendev)
+{
+    struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+    struct ioreq *ioreq;
+
+    while (!LIST_EMPTY(&blkdev->freelist)) {
+	ioreq = LIST_FIRST(&blkdev->freelist);
+        LIST_REMOVE(ioreq, list);
+        qemu_iovec_destroy(&ioreq->v);
+	qemu_free(ioreq);
+    }
+
+    qemu_free(blkdev->params);
+    qemu_free(blkdev->mode);
+    qemu_free(blkdev->type);
+    qemu_free(blkdev->dev);
+    qemu_free(blkdev->devtype);
+    qemu_bh_delete(blkdev->bh);
+    return 0;
+}
+
+static void blk_event(struct XenDevice *xendev)
+{
+    struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+    qemu_bh_schedule(blkdev->bh);
+}
+
+struct XenDevOps xen_blkdev_ops = {
+    .size       = sizeof(struct XenBlkDev),
+    .flags      = DEVOPS_FLAG_NEED_GNTDEV,
+    .alloc      = blk_alloc,
+    .init       = blk_init,
+    .connect    = blk_connect,
+    .disconnect = blk_disconnect,
+    .event      = blk_event,
+    .free       = blk_free,
+};
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index 1f0b3ec8fd..c764beae43 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -59,6 +59,7 @@ static void xen_init_pv(ram_addr_t ram_size, int vga_ram_size,
     xen_be_register("console", &xen_console_ops);
     xen_be_register("vkbd", &xen_kbdmouse_ops);
     xen_be_register("vfb", &xen_framebuffer_ops);
+    xen_be_register("qdisk", &xen_blkdev_ops);
 
     /* setup framebuffer */
     xen_init_display(xen_domid);
diff --git a/sysemu.h b/sysemu.h
index cbfbb8e6d1..e97fc010bf 100644
--- a/sysemu.h
+++ b/sysemu.h
@@ -131,7 +131,7 @@ extern unsigned int nb_prom_envs;
 #endif
 
 typedef enum {
-    IF_IDE, IF_SCSI, IF_FLOPPY, IF_PFLASH, IF_MTD, IF_SD, IF_VIRTIO
+    IF_IDE, IF_SCSI, IF_FLOPPY, IF_PFLASH, IF_MTD, IF_SD, IF_VIRTIO, IF_XEN
 } BlockInterfaceType;
 
 typedef enum {
diff --git a/vl.c b/vl.c
index 7b0e0d5fe9..13426653b5 100644
--- a/vl.c
+++ b/vl.c
@@ -2373,7 +2373,10 @@ int drive_init(struct drive_opt *arg, int snapshot, void *opaque)
         } else if (!strcmp(buf, "virtio")) {
             type = IF_VIRTIO;
             max_devs = 0;
-        } else {
+	} else if (!strcmp(buf, "xen")) {
+	    type = IF_XEN;
+            max_devs = 0;
+	} else {
             fprintf(stderr, "qemu: '%s' unsupported bus type '%s'\n", str, buf);
             return -1;
 	}
@@ -2587,6 +2590,7 @@ int drive_init(struct drive_opt *arg, int snapshot, void *opaque)
     switch(type) {
     case IF_IDE:
     case IF_SCSI:
+    case IF_XEN:
         switch(media) {
 	case MEDIA_DISK:
             if (cyls != 0) {