Merge remote-tracking branch 'bonzini/virtio-scsi' into staging

* bonzini/virtio-scsi:
  scsi-block: always use scsi_generic_ops for cache != none
  scsi: fix searching for an empty id
  scsi: fix wrong return for target INQUIRY
  virtio-scsi: add migration support
  virtio-scsi: process control queue requests
  virtio-scsi: add basic SCSI bus operation
  virtio-scsi: Add basic request processing infrastructure
  virtio-scsi: Add virtio-scsi stub device
  scsi-disk: add migration support
  scsi-generic: add migration support
  scsi: add SCSIDevice vmstate definitions
  scsi-disk: enable scatter/gather functionality
  scsi: add scatter/gather functionality
  scsi: pass residual amount to command_complete
  ahci: use new DMA helpers
  dma-helpers: add accounting wrappers
  dma-helpers: add dma_buf_read and dma_buf_write
  dma-helpers: make QEMUSGList target independent

17 files changed, 1079 insertions, 106 deletions

diff --git a/hw/esp.c b/hw/esp.c
index 2dda8e358e..8d73e56886 100644
--- a/hw/esp.c
+++ b/hw/esp.c
@@ -390,7 +390,8 @@ static void esp_do_dma(ESPState *s)
     esp_dma_done(s);
 }
 
-static void esp_command_complete(SCSIRequest *req, uint32_t status)
+static void esp_command_complete(SCSIRequest *req, uint32_t status,
+                                 size_t resid)
 {
     ESPState *s = DO_UPCAST(ESPState, busdev.qdev, req->bus->qbus.parent);
 
diff --git a/hw/ide/ahci.c b/hw/ide/ahci.c
index b515f417b1..041ce1e89f 100644
--- a/hw/ide/ahci.c
+++ b/hw/ide/ahci.c
@@ -428,55 +428,6 @@ static void ahci_reg_init(AHCIState *s)
     }
 }
 
-static uint32_t read_from_sglist(uint8_t *buffer, uint32_t len,
-                                 QEMUSGList *sglist)
-{
-    uint32_t i = 0;
-    uint32_t total = 0, once;
-    ScatterGatherEntry *cur_prd;
-    uint32_t sgcount;
-
-    cur_prd = sglist->sg;
-    sgcount = sglist->nsg;
-    for (i = 0; len && sgcount; i++) {
-        once = MIN(cur_prd->len, len);
-        cpu_physical_memory_read(cur_prd->base, buffer, once);
-        cur_prd++;
-        sgcount--;
-        len -= once;
-        buffer += once;
-        total += once;
-    }
-
-    return total;
-}
-
-static uint32_t write_to_sglist(uint8_t *buffer, uint32_t len,
-                                QEMUSGList *sglist)
-{
-    uint32_t i = 0;
-    uint32_t total = 0, once;
-    ScatterGatherEntry *cur_prd;
-    uint32_t sgcount;
-
-    DPRINTF(-1, "total: 0x%x bytes\n", len);
-
-    cur_prd = sglist->sg;
-    sgcount = sglist->nsg;
-    for (i = 0; len && sgcount; i++) {
-        once = MIN(cur_prd->len, len);
-        DPRINTF(-1, "write 0x%x bytes to 0x%lx\n", once, (long)cur_prd->base);
-        cpu_physical_memory_write(cur_prd->base, buffer, once);
-        cur_prd++;
-        sgcount--;
-        len -= once;
-        buffer += once;
-        total += once;
-    }
-
-    return total;
-}
-
 static void check_cmd(AHCIState *s, int port)
 {
     AHCIPortRegs *pr = &s->dev[port].port_regs;
@@ -802,9 +753,8 @@ static void process_ncq_command(AHCIState *s, int port, uint8_t *cmd_fis,
             DPRINTF(port, "tag %d aio read %"PRId64"\n",
                     ncq_tfs->tag, ncq_tfs->lba);
 
-            bdrv_acct_start(ncq_tfs->drive->port.ifs[0].bs, &ncq_tfs->acct,
-                            (ncq_tfs->sector_count-1) * BDRV_SECTOR_SIZE,
-                            BDRV_ACCT_READ);
+            dma_acct_start(ncq_tfs->drive->port.ifs[0].bs, &ncq_tfs->acct,
+                           &ncq_tfs->sglist, BDRV_ACCT_READ);
             ncq_tfs->aiocb = dma_bdrv_read(ncq_tfs->drive->port.ifs[0].bs,
                                            &ncq_tfs->sglist, ncq_tfs->lba,
                                            ncq_cb, ncq_tfs);
@@ -816,9 +766,8 @@ static void process_ncq_command(AHCIState *s, int port, uint8_t *cmd_fis,
             DPRINTF(port, "tag %d aio write %"PRId64"\n",
                     ncq_tfs->tag, ncq_tfs->lba);
 
-            bdrv_acct_start(ncq_tfs->drive->port.ifs[0].bs, &ncq_tfs->acct,
-                            (ncq_tfs->sector_count-1) * BDRV_SECTOR_SIZE,
-                            BDRV_ACCT_WRITE);
+            dma_acct_start(ncq_tfs->drive->port.ifs[0].bs, &ncq_tfs->acct,
+                           &ncq_tfs->sglist, BDRV_ACCT_WRITE);
             ncq_tfs->aiocb = dma_bdrv_write(ncq_tfs->drive->port.ifs[0].bs,
                                             &ncq_tfs->sglist, ncq_tfs->lba,
                                             ncq_cb, ncq_tfs);
@@ -1023,12 +972,12 @@ static int ahci_start_transfer(IDEDMA *dma)
             is_write ? "writ" : "read", size, is_atapi ? "atapi" : "ata",
             has_sglist ? "" : "o");
 
-    if (is_write && has_sglist && (s->data_ptr < s->data_end)) {
-        read_from_sglist(s->data_ptr, size, &s->sg);
-    }
-
-    if (!is_write && has_sglist && (s->data_ptr < s->data_end)) {
-        write_to_sglist(s->data_ptr, size, &s->sg);
+    if (has_sglist && size) {
+        if (is_write) {
+            dma_buf_write(s->data_ptr, size, &s->sg);
+        } else {
+            dma_buf_read(s->data_ptr, size, &s->sg);
+        }
     }
 
     /* update number of transferred bytes */
@@ -1067,14 +1016,9 @@ static int ahci_dma_prepare_buf(IDEDMA *dma, int is_write)
 {
     AHCIDevice *ad = DO_UPCAST(AHCIDevice, dma, dma);
     IDEState *s = &ad->port.ifs[0];
-    int i;
 
     ahci_populate_sglist(ad, &s->sg);
-
-    s->io_buffer_size = 0;
-    for (i = 0; i < s->sg.nsg; i++) {
-        s->io_buffer_size += s->sg.sg[i].len;
-    }
+    s->io_buffer_size = s->sg.size;
 
     DPRINTF(ad->port_no, "len=%#x\n", s->io_buffer_size);
     return s->io_buffer_size != 0;
@@ -1092,9 +1036,9 @@ static int ahci_dma_rw_buf(IDEDMA *dma, int is_write)
     }
 
     if (is_write) {
-        write_to_sglist(p, l, &s->sg);
+        dma_buf_read(p, l, &s->sg);
     } else {
-        read_from_sglist(p, l, &s->sg);
+        dma_buf_write(p, l, &s->sg);
     }
 
     /* update number of transferred bytes */
diff --git a/hw/lsi53c895a.c b/hw/lsi53c895a.c
index 0acd1d06df..edc09b7307 100644
--- a/hw/lsi53c895a.c
+++ b/hw/lsi53c895a.c
@@ -699,7 +699,7 @@ static int lsi_queue_req(LSIState *s, SCSIRequest *req, uint32_t len)
 }
 
  /* Callback to indicate that the SCSI layer has completed a command.  */
-static void lsi_command_complete(SCSIRequest *req, uint32_t status)
+static void lsi_command_complete(SCSIRequest *req, uint32_t status, size_t resid)
 {
     LSIState *s = DO_UPCAST(LSIState, dev.qdev, req->bus->qbus.parent);
     int out;
diff --git a/hw/pci.h b/hw/pci.h
index ee53f26bb1..4f19fdbd89 100644
--- a/hw/pci.h
+++ b/hw/pci.h
@@ -75,6 +75,7 @@
 #define PCI_DEVICE_ID_VIRTIO_BLOCK       0x1001
 #define PCI_DEVICE_ID_VIRTIO_BALLOON     0x1002
 #define PCI_DEVICE_ID_VIRTIO_CONSOLE     0x1003
+#define PCI_DEVICE_ID_VIRTIO_SCSI        0x1004
 
 #define FMT_PCIBUS                      PRIx64
 
diff --git a/hw/s390-virtio-bus.c b/hw/s390-virtio-bus.c
index 9d480564d7..c450e4bb5b 100644
--- a/hw/s390-virtio-bus.c
+++ b/hw/s390-virtio-bus.c
@@ -169,6 +169,18 @@ static int s390_virtio_serial_init(VirtIOS390Device *dev)
     return r;
 }
 
+static int s390_virtio_scsi_init(VirtIOS390Device *dev)
+{
+    VirtIODevice *vdev;
+
+    vdev = virtio_scsi_init((DeviceState *)dev, &dev->scsi);
+    if (!vdev) {
+        return -1;
+    }
+
+    return s390_virtio_device_init(dev, vdev);
+}
+
 static uint64_t s390_virtio_device_vq_token(VirtIOS390Device *dev, int vq)
 {
     ram_addr_t token_off;
@@ -433,6 +445,26 @@ static TypeInfo virtio_s390_device_info = {
     .abstract = true,
 };
 
+static Property s390_virtio_scsi_properties[] = {
+    DEFINE_VIRTIO_SCSI_PROPERTIES(VirtIOS390Device, host_features, scsi),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void s390_virtio_scsi_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    VirtIOS390DeviceClass *k = VIRTIO_S390_DEVICE_CLASS(klass);
+
+    k->init = s390_virtio_scsi_init;
+    dc->props = s390_virtio_scsi_properties;
+}
+
+static TypeInfo s390_virtio_scsi = {
+    .name          = "virtio-scsi-s390",
+    .parent        = TYPE_VIRTIO_S390_DEVICE,
+    .instance_size = sizeof(VirtIOS390Device),
+    .class_init    = s390_virtio_scsi_class_init,
+};
 
 /***************** S390 Virtio Bus Bridge Device *******************/
 /* Only required to have the virtio bus as child in the system bus */
@@ -465,6 +497,7 @@ static void s390_virtio_register_types(void)
     type_register_static(&s390_virtio_serial);
     type_register_static(&s390_virtio_blk);
     type_register_static(&s390_virtio_net);
+    type_register_static(&s390_virtio_scsi);
     type_register_static(&s390_virtio_bridge_info);
 }
 
diff --git a/hw/s390-virtio-bus.h b/hw/s390-virtio-bus.h
index b5e59b7d4b..0e60bc0fa2 100644
--- a/hw/s390-virtio-bus.h
+++ b/hw/s390-virtio-bus.h
@@ -19,6 +19,7 @@
 
 #include "virtio-net.h"
 #include "virtio-serial.h"
+#include "virtio-scsi.h"
 
 #define VIRTIO_DEV_OFFS_TYPE		0	/* 8 bits */
 #define VIRTIO_DEV_OFFS_NUM_VQ		1	/* 8 bits */
@@ -67,6 +68,7 @@ struct VirtIOS390Device {
     uint32_t host_features;
     virtio_serial_conf serial;
     virtio_net_conf net;
+    VirtIOSCSIConf scsi;
 };
 
 typedef struct VirtIOS390Bus {
diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
index b3e97ceeec..2cb5a18da2 100644
--- a/hw/scsi-bus.c
+++ b/hw/scsi-bus.c
@@ -5,6 +5,7 @@
 #include "qdev.h"
 #include "blockdev.h"
 #include "trace.h"
+#include "dma.h"
 
 static char *scsibus_get_fw_dev_path(DeviceState *dev);
 static int scsi_req_parse(SCSICommand *cmd, SCSIDevice *dev, uint8_t *buf);
@@ -86,6 +87,7 @@ static void scsi_dma_restart_bh(void *opaque)
                 scsi_req_continue(req);
                 break;
             case SCSI_XFER_NONE:
+                assert(!req->sg);
                 scsi_req_dequeue(req);
                 scsi_req_enqueue(req);
                 break;
@@ -130,6 +132,10 @@ static int scsi_qdev_init(DeviceState *qdev)
         error_report("bad scsi device id: %d", dev->id);
         goto err;
     }
+    if (dev->lun != -1 && dev->lun > bus->info->max_lun) {
+        error_report("bad scsi device lun: %d", dev->lun);
+        goto err;
+    }
 
     if (dev->id == -1) {
         int id = -1;
@@ -138,8 +144,8 @@ static int scsi_qdev_init(DeviceState *qdev)
         }
         do {
             d = scsi_device_find(bus, dev->channel, ++id, dev->lun);
-        } while (d && d->lun == dev->lun && id <= bus->info->max_target);
-        if (id > bus->info->max_target) {
+        } while (d && d->lun == dev->lun && id < bus->info->max_target);
+        if (d && d->lun == dev->lun) {
             error_report("no free target");
             goto err;
         }
@@ -149,14 +155,15 @@ static int scsi_qdev_init(DeviceState *qdev)
         do {
             d = scsi_device_find(bus, dev->channel, dev->id, ++lun);
         } while (d && d->lun == lun && lun < bus->info->max_lun);
-        if (lun > bus->info->max_lun) {
+        if (d && d->lun == lun) {
             error_report("no free lun");
             goto err;
         }
         dev->lun = lun;
     } else {
         d = scsi_device_find(bus, dev->channel, dev->id, dev->lun);
-        if (dev->lun == d->lun && dev != d) {
+        assert(d);
+        if (d->lun == dev->lun && dev != d) {
             qdev_free(&d->qdev);
         }
     }
@@ -215,7 +222,7 @@ int scsi_bus_legacy_handle_cmdline(SCSIBus *bus)
     int res = 0, unit;
 
     loc_push_none(&loc);
-    for (unit = 0; unit < bus->info->max_target; unit++) {
+    for (unit = 0; unit <= bus->info->max_target; unit++) {
         dinfo = drive_get(IF_SCSI, bus->busnr, unit);
         if (dinfo == NULL) {
             continue;
@@ -378,7 +385,7 @@ static bool scsi_target_emulate_inquiry(SCSITargetReq *r)
 
     /* PAGE CODE == 0 */
     if (r->req.cmd.xfer < 5) {
-        return -1;
+        return false;
     }
 
     r->len = MIN(r->req.cmd.xfer, 36);
@@ -533,6 +540,8 @@ SCSIRequest *scsi_req_new(SCSIDevice *d, uint32_t tag, uint32_t lun,
     }
 
     req->cmd = cmd;
+    req->resid = req->cmd.xfer;
+
     switch (buf[0]) {
     case INQUIRY:
         trace_scsi_inquiry(d->id, lun, tag, cmd.buf[1], cmd.buf[2]);
@@ -643,15 +652,25 @@ void scsi_req_build_sense(SCSIRequest *req, SCSISense sense)
     req->sense_len = 18;
 }
 
-int32_t scsi_req_enqueue(SCSIRequest *req)
+static void scsi_req_enqueue_internal(SCSIRequest *req)
 {
-    int32_t rc;
-
     assert(!req->enqueued);
     scsi_req_ref(req);
+    if (req->bus->info->get_sg_list) {
+        req->sg = req->bus->info->get_sg_list(req);
+    } else {
+        req->sg = NULL;
+    }
     req->enqueued = true;
     QTAILQ_INSERT_TAIL(&req->dev->requests, req, next);
+}
 
+int32_t scsi_req_enqueue(SCSIRequest *req)
+{
+    int32_t rc;
+
+    assert(!req->retry);
+    scsi_req_enqueue_internal(req);
     scsi_req_ref(req);
     rc = req->ops->send_command(req, req->cmd.buf);
     scsi_req_unref(req);
@@ -1273,12 +1292,32 @@ void scsi_req_continue(SCSIRequest *req)
    Once it completes, calling scsi_req_continue will restart I/O.  */
 void scsi_req_data(SCSIRequest *req, int len)
 {
+    uint8_t *buf;
     if (req->io_canceled) {
         trace_scsi_req_data_canceled(req->dev->id, req->lun, req->tag, len);
-    } else {
-        trace_scsi_req_data(req->dev->id, req->lun, req->tag, len);
+        return;
+    }
+    trace_scsi_req_data(req->dev->id, req->lun, req->tag, len);
+    assert(req->cmd.mode != SCSI_XFER_NONE);
+    if (!req->sg) {
+        req->resid -= len;
         req->bus->info->transfer_data(req, len);
+        return;
     }
+
+    /* If the device calls scsi_req_data and the HBA specified a
+     * scatter/gather list, the transfer has to happen in a single
+     * step.  */
+    assert(!req->dma_started);
+    req->dma_started = true;
+
+    buf = scsi_req_get_buf(req);
+    if (req->cmd.mode == SCSI_XFER_FROM_DEV) {
+        req->resid = dma_buf_read(buf, len, req->sg);
+    } else {
+        req->resid = dma_buf_write(buf, len, req->sg);
+    }
+    scsi_req_continue(req);
 }
 
 void scsi_req_print(SCSIRequest *req)
@@ -1337,7 +1376,7 @@ void scsi_req_complete(SCSIRequest *req, int status)
 
     scsi_req_ref(req);
     scsi_req_dequeue(req);
-    req->bus->info->complete(req, req->status);
+    req->bus->info->complete(req, req->status, req->resid);
     scsi_req_unref(req);
 }
 
@@ -1413,6 +1452,102 @@ SCSIDevice *scsi_device_find(SCSIBus *bus, int channel, int id, int lun)
     return target_dev;
 }
 
+/* SCSI request list.  For simplicity, pv points to the whole device */
+
+static void put_scsi_requests(QEMUFile *f, void *pv, size_t size)
+{
+    SCSIDevice *s = pv;
+    SCSIBus *bus = DO_UPCAST(SCSIBus, qbus, s->qdev.parent_bus);
+    SCSIRequest *req;
+
+    QTAILQ_FOREACH(req, &s->requests, next) {
+        assert(!req->io_canceled);
+        assert(req->status == -1);
+        assert(req->retry);
+        assert(req->enqueued);
+
+        qemu_put_sbyte(f, 1);
+        qemu_put_buffer(f, req->cmd.buf, sizeof(req->cmd.buf));
+        qemu_put_be32s(f, &req->tag);
+        qemu_put_be32s(f, &req->lun);
+        if (bus->info->save_request) {
+            bus->info->save_request(f, req);
+        }
+        if (req->ops->save_request) {
+            req->ops->save_request(f, req);
+        }
+    }
+    qemu_put_sbyte(f, 0);
+}
+
+static int get_scsi_requests(QEMUFile *f, void *pv, size_t size)
+{
+    SCSIDevice *s = pv;
+    SCSIBus *bus = DO_UPCAST(SCSIBus, qbus, s->qdev.parent_bus);
+
+    while (qemu_get_sbyte(f)) {
+        uint8_t buf[SCSI_CMD_BUF_SIZE];
+        uint32_t tag;
+        uint32_t lun;
+        SCSIRequest *req;
+
+        qemu_get_buffer(f, buf, sizeof(buf));
+        qemu_get_be32s(f, &tag);
+        qemu_get_be32s(f, &lun);
+        req = scsi_req_new(s, tag, lun, buf, NULL);
+        if (bus->info->load_request) {
+            req->hba_private = bus->info->load_request(f, req);
+        }
+        if (req->ops->load_request) {
+            req->ops->load_request(f, req);
+        }
+
+        /* Just restart it later.  */
+        req->retry = true;
+        scsi_req_enqueue_internal(req);
+
+        /* At this point, the request will be kept alive by the reference
+         * added by scsi_req_enqueue_internal, so we can release our reference.
+         * The HBA of course will add its own reference in the load_request
+         * callback if it needs to hold on the SCSIRequest.
+         */
+        scsi_req_unref(req);
+    }
+
+    return 0;
+}
+
+const VMStateInfo vmstate_info_scsi_requests = {
+    .name = "scsi-requests",
+    .get  = get_scsi_requests,
+    .put  = put_scsi_requests,
+};
+
+const VMStateDescription vmstate_scsi_device = {
+    .name = "SCSIDevice",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_UINT8(unit_attention.key, SCSIDevice),
+        VMSTATE_UINT8(unit_attention.asc, SCSIDevice),
+        VMSTATE_UINT8(unit_attention.ascq, SCSIDevice),
+        VMSTATE_BOOL(sense_is_ua, SCSIDevice),
+        VMSTATE_UINT8_ARRAY(sense, SCSIDevice, SCSI_SENSE_BUF_SIZE),
+        VMSTATE_UINT32(sense_len, SCSIDevice),
+        {
+            .name         = "requests",
+            .version_id   = 0,
+            .field_exists = NULL,
+            .size         = 0,   /* ouch */
+            .info         = &vmstate_info_scsi_requests,
+            .flags        = VMS_SINGLE,
+            .offset       = 0,
+        },
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 static void scsi_device_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *k = DEVICE_CLASS(klass);
diff --git a/hw/scsi-disk.c b/hw/scsi-disk.c
index c12e3a6cb3..4101a59d34 100644
--- a/hw/scsi-disk.c
+++ b/hw/scsi-disk.c
@@ -38,6 +38,7 @@ do { fprintf(stderr, "scsi-disk: " fmt , ## __VA_ARGS__); } while (0)
 #include "sysemu.h"
 #include "blockdev.h"
 #include "block_int.h"
+#include "dma.h"
 
 #ifdef __linux
 #include <scsi/sg.h>
@@ -110,12 +111,12 @@ static void scsi_cancel_io(SCSIRequest *req)
     r->req.aiocb = NULL;
 }
 
-static uint32_t scsi_init_iovec(SCSIDiskReq *r)
+static uint32_t scsi_init_iovec(SCSIDiskReq *r, size_t size)
 {
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, r->req.dev);
 
     if (!r->iov.iov_base) {
-        r->buflen = SCSI_DMA_BUF_SIZE;
+        r->buflen = size;
         r->iov.iov_base = qemu_blockalign(s->qdev.conf.bs, r->buflen);
     }
     r->iov.iov_len = MIN(r->sector_count * 512, r->buflen);
@@ -123,6 +124,56 @@ static uint32_t scsi_init_iovec(SCSIDiskReq *r)
     return r->qiov.size / 512;
 }
 
+static void scsi_disk_save_request(QEMUFile *f, SCSIRequest *req)
+{
+    SCSIDiskReq *r = DO_UPCAST(SCSIDiskReq, req, req);
+
+    qemu_put_be64s(f, &r->sector);
+    qemu_put_be32s(f, &r->sector_count);
+    qemu_put_be32s(f, &r->buflen);
+    if (r->buflen && r->req.cmd.mode == SCSI_XFER_TO_DEV) {
+        qemu_put_buffer(f, r->iov.iov_base, r->iov.iov_len);
+    }
+}
+
+static void scsi_disk_load_request(QEMUFile *f, SCSIRequest *req)
+{
+    SCSIDiskReq *r = DO_UPCAST(SCSIDiskReq, req, req);
+
+    qemu_get_be64s(f, &r->sector);
+    qemu_get_be32s(f, &r->sector_count);
+    qemu_get_be32s(f, &r->buflen);
+    if (r->buflen) {
+        scsi_init_iovec(r, r->buflen);
+        if (r->req.cmd.mode == SCSI_XFER_TO_DEV) {
+            qemu_get_buffer(f, r->iov.iov_base, r->iov.iov_len);
+        }
+    }
+
+    qemu_iovec_init_external(&r->qiov, &r->iov, 1);
+}
+
+static void scsi_dma_complete(void *opaque, int ret)
+{
+    SCSIDiskReq *r = (SCSIDiskReq *)opaque;
+    SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, r->req.dev);
+
+    bdrv_acct_done(s->qdev.conf.bs, &r->acct);
+
+    if (ret) {
+        if (scsi_handle_rw_error(r, -ret)) {
+            goto done;
+        }
+    }
+
+    r->sector += r->sector_count;
+    r->sector_count = 0;
+    scsi_req_complete(&r->req, GOOD);
+
+done:
+    scsi_req_unref(&r->req);
+}
+
 static void scsi_read_complete(void * opaque, int ret)
 {
     SCSIDiskReq *r = (SCSIDiskReq *)opaque;
@@ -213,10 +264,17 @@ static void scsi_read_data(SCSIRequest *req)
         return;
     }
 
-    n = scsi_init_iovec(r);
-    bdrv_acct_start(s->qdev.conf.bs, &r->acct, n * BDRV_SECTOR_SIZE, BDRV_ACCT_READ);
-    r->req.aiocb = bdrv_aio_readv(s->qdev.conf.bs, r->sector, &r->qiov, n,
-                              scsi_read_complete, r);
+    if (r->req.sg) {
+        dma_acct_start(s->qdev.conf.bs, &r->acct, r->req.sg, BDRV_ACCT_READ);
+        r->req.resid -= r->req.sg->size;
+        r->req.aiocb = dma_bdrv_read(s->qdev.conf.bs, r->req.sg, r->sector,
+                                     scsi_dma_complete, r);
+    } else {
+        n = scsi_init_iovec(r, SCSI_DMA_BUF_SIZE);
+        bdrv_acct_start(s->qdev.conf.bs, &r->acct, n * BDRV_SECTOR_SIZE, BDRV_ACCT_READ);
+        r->req.aiocb = bdrv_aio_readv(s->qdev.conf.bs, r->sector, &r->qiov, n,
+                                      scsi_read_complete, r);
+    }
 }
 
 /*
@@ -287,7 +345,7 @@ static void scsi_write_complete(void * opaque, int ret)
     if (r->sector_count == 0) {
         scsi_req_complete(&r->req, GOOD);
     } else {
-        scsi_init_iovec(r);
+        scsi_init_iovec(r, SCSI_DMA_BUF_SIZE);
         DPRINTF("Write complete tag=0x%x more=%d\n", r->req.tag, r->qiov.size);
         scsi_req_data(&r->req, r->qiov.size);
     }
@@ -315,18 +373,26 @@ static void scsi_write_data(SCSIRequest *req)
         return;
     }
 
-    n = r->qiov.size / 512;
-    if (n) {
-        if (s->tray_open) {
-            scsi_write_complete(r, -ENOMEDIUM);
-            return;
-        }
+    if (!r->req.sg && !r->qiov.size) {
+        /* Called for the first time.  Ask the driver to send us more data.  */
+        scsi_write_complete(r, 0);
+        return;
+    }
+    if (s->tray_open) {
+        scsi_write_complete(r, -ENOMEDIUM);
+        return;
+    }
+
+    if (r->req.sg) {
+        dma_acct_start(s->qdev.conf.bs, &r->acct, r->req.sg, BDRV_ACCT_WRITE);
+        r->req.resid -= r->req.sg->size;
+        r->req.aiocb = dma_bdrv_write(s->qdev.conf.bs, r->req.sg, r->sector,
+                                      scsi_dma_complete, r);
+    } else {
+        n = r->qiov.size / 512;
         bdrv_acct_start(s->qdev.conf.bs, &r->acct, n * BDRV_SECTOR_SIZE, BDRV_ACCT_WRITE);
         r->req.aiocb = bdrv_aio_writev(s->qdev.conf.bs, r->sector, &r->qiov, n,
                                        scsi_write_complete, r);
-    } else {
-        /* Called for the first time.  Ask the driver to send us more data.  */
-        scsi_write_complete(r, 0);
     }
 }
 
@@ -1584,6 +1650,8 @@ static const SCSIReqOps scsi_disk_reqops = {
     .write_data   = scsi_write_data,
     .cancel_io    = scsi_cancel_io,
     .get_buf      = scsi_get_buf,
+    .load_request = scsi_disk_load_request,
+    .save_request = scsi_disk_save_request,
 };
 
 static SCSIRequest *scsi_new_request(SCSIDevice *d, uint32_t tag, uint32_t lun,
@@ -1686,6 +1754,15 @@ static SCSIRequest *scsi_block_new_request(SCSIDevice *d, uint32_t tag,
     case WRITE_VERIFY_10:
     case WRITE_VERIFY_12:
     case WRITE_VERIFY_16:
+        /* If we are not using O_DIRECT, we might read stale data from the
+	 * host cache if writes were made using other commands than these
+	 * ones (such as WRITE SAME or EXTENDED COPY, etc.).  So, without
+	 * O_DIRECT everything must go through SG_IO.
+         */
+        if (!(s->qdev.conf.bs->open_flags & BDRV_O_NOCACHE)) {
+            break;
+        }
+
         /* MMC writing cannot be done via pread/pwrite, because it sometimes
          * involves writing beyond the maximum LBA or to negative LBA (lead-in).
          * And once you do these writes, reading from the block device is
@@ -1696,10 +1773,11 @@ static SCSIRequest *scsi_block_new_request(SCSIDevice *d, uint32_t tag,
          * seen, but performance usually isn't paramount on optical media.  So,
          * just make scsi-block operate the same as scsi-generic for them.
          */
-        if (s->qdev.type != TYPE_ROM) {
-            return scsi_req_alloc(&scsi_disk_reqops, &s->qdev, tag, lun,
-                                  hba_private);
-        }
+        if (s->qdev.type == TYPE_ROM) {
+            break;
+	}
+        return scsi_req_alloc(&scsi_disk_reqops, &s->qdev, tag, lun,
+                              hba_private);
     }
 
     return scsi_req_alloc(&scsi_generic_req_ops, &s->qdev, tag, lun,
@@ -1718,6 +1796,22 @@ static Property scsi_hd_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static const VMStateDescription vmstate_scsi_disk_state = {
+    .name = "scsi-disk",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_SCSI_DEVICE(qdev, SCSIDiskState),
+        VMSTATE_BOOL(media_changed, SCSIDiskState),
+        VMSTATE_BOOL(media_event, SCSIDiskState),
+        VMSTATE_BOOL(eject_request, SCSIDiskState),
+        VMSTATE_BOOL(tray_open, SCSIDiskState),
+        VMSTATE_BOOL(tray_locked, SCSIDiskState),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 static void scsi_hd_class_initfn(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
@@ -1731,6 +1825,7 @@ static void scsi_hd_class_initfn(ObjectClass *klass, void *data)
     dc->desc = "virtual SCSI disk";
     dc->reset = scsi_disk_reset;
     dc->props = scsi_hd_properties;
+    dc->vmsd  = &vmstate_scsi_disk_state;
 }
 
 static TypeInfo scsi_hd_info = {
@@ -1758,6 +1853,7 @@ static void scsi_cd_class_initfn(ObjectClass *klass, void *data)
     dc->desc = "virtual SCSI CD-ROM";
     dc->reset = scsi_disk_reset;
     dc->props = scsi_cd_properties;
+    dc->vmsd  = &vmstate_scsi_disk_state;
 }
 
 static TypeInfo scsi_cd_info = {
@@ -1785,6 +1881,7 @@ static void scsi_block_class_initfn(ObjectClass *klass, void *data)
     dc->desc = "SCSI block device passthrough";
     dc->reset = scsi_disk_reset;
     dc->props = scsi_block_properties;
+    dc->vmsd  = &vmstate_scsi_disk_state;
 }
 
 static TypeInfo scsi_block_info = {
@@ -1814,6 +1911,7 @@ static void scsi_disk_class_initfn(ObjectClass *klass, void *data)
     dc->desc = "virtual SCSI disk or CD-ROM (legacy)";
     dc->reset = scsi_disk_reset;
     dc->props = scsi_disk_properties;
+    dc->vmsd  = &vmstate_scsi_disk_state;
 }
 
 static TypeInfo scsi_disk_info = {
diff --git a/hw/scsi-generic.c b/hw/scsi-generic.c
index 86014aa893..d856d23b3b 100644
--- a/hw/scsi-generic.c
+++ b/hw/scsi-generic.c
@@ -59,6 +59,28 @@ typedef struct SCSIGenericReq {
     sg_io_hdr_t io_header;
 } SCSIGenericReq;
 
+static void scsi_generic_save_request(QEMUFile *f, SCSIRequest *req)
+{
+    SCSIGenericReq *r = DO_UPCAST(SCSIGenericReq, req, req);
+
+    qemu_put_sbe32s(f, &r->buflen);
+    if (r->buflen && r->req.cmd.mode == SCSI_XFER_TO_DEV) {
+        assert(!r->req.sg);
+        qemu_put_buffer(f, r->buf, r->req.cmd.xfer);
+    }
+}
+
+static void scsi_generic_load_request(QEMUFile *f, SCSIRequest *req)
+{
+    SCSIGenericReq *r = DO_UPCAST(SCSIGenericReq, req, req);
+
+    qemu_get_sbe32s(f, &r->buflen);
+    if (r->buflen && r->req.cmd.mode == SCSI_XFER_TO_DEV) {
+        assert(!r->req.sg);
+        qemu_get_buffer(f, r->buf, r->req.cmd.xfer);
+    }
+}
+
 static void scsi_free_request(SCSIRequest *req)
 {
     SCSIGenericReq *r = DO_UPCAST(SCSIGenericReq, req, req);
@@ -446,6 +468,8 @@ const SCSIReqOps scsi_generic_req_ops = {
     .write_data   = scsi_write_data,
     .cancel_io    = scsi_cancel_io,
     .get_buf      = scsi_get_buf,
+    .load_request = scsi_generic_load_request,
+    .save_request = scsi_generic_save_request,
 };
 
 static SCSIRequest *scsi_new_request(SCSIDevice *d, uint32_t tag, uint32_t lun,
@@ -474,6 +498,7 @@ static void scsi_generic_class_initfn(ObjectClass *klass, void *data)
     dc->desc = "pass through generic scsi device (/dev/sg*)";
     dc->reset = scsi_generic_reset;
     dc->props = scsi_generic_properties;
+    dc->vmsd  = &vmstate_scsi_device;
 }
 
 static TypeInfo scsi_generic_info = {
diff --git a/hw/scsi.h b/hw/scsi.h
index dc72b6fc1e..2eb66f7393 100644
--- a/hw/scsi.h
+++ b/hw/scsi.h
@@ -46,8 +46,11 @@ struct SCSIRequest {
     uint32_t          tag;
     uint32_t          lun;
     uint32_t          status;
+    size_t            resid;
     SCSICommand       cmd;
     BlockDriverAIOCB  *aiocb;
+    QEMUSGList        *sg;
+    bool              dma_started;
     uint8_t sense[SCSI_SENSE_BUF_SIZE];
     uint32_t sense_len;
     bool enqueued;
@@ -93,6 +96,16 @@ struct SCSIDevice
     uint64_t max_lba;
 };
 
+extern const VMStateDescription vmstate_scsi_device;
+
+#define VMSTATE_SCSI_DEVICE(_field, _state) {                        \
+    .name       = (stringify(_field)),                               \
+    .size       = sizeof(SCSIDevice),                                \
+    .vmsd       = &vmstate_scsi_device,                              \
+    .flags      = VMS_STRUCT,                                        \
+    .offset     = vmstate_offset_value(_state, _field, SCSIDevice),  \
+}
+
 /* cdrom.c */
 int cdrom_read_toc(int nb_sectors, uint8_t *buf, int msf, int start_track);
 int cdrom_read_toc_raw(int nb_sectors, uint8_t *buf, int msf, int session_num);
@@ -106,14 +119,21 @@ struct SCSIReqOps {
     void (*write_data)(SCSIRequest *req);
     void (*cancel_io)(SCSIRequest *req);
     uint8_t *(*get_buf)(SCSIRequest *req);
+
+    void (*save_request)(QEMUFile *f, SCSIRequest *req);
+    void (*load_request)(QEMUFile *f, SCSIRequest *req);
 };
 
 struct SCSIBusInfo {
     int tcq;
     int max_channel, max_target, max_lun;
     void (*transfer_data)(SCSIRequest *req, uint32_t arg);
-    void (*complete)(SCSIRequest *req, uint32_t arg);
+    void (*complete)(SCSIRequest *req, uint32_t arg, size_t resid);
     void (*cancel)(SCSIRequest *req);
+    QEMUSGList *(*get_sg_list)(SCSIRequest *req);
+
+    void (*save_request)(QEMUFile *f, SCSIRequest *req);
+    void *(*load_request)(QEMUFile *f, SCSIRequest *req);
 };
 
 struct SCSIBus {
diff --git a/hw/spapr_vscsi.c b/hw/spapr_vscsi.c
index ffce261f98..21670170e8 100644
--- a/hw/spapr_vscsi.c
+++ b/hw/spapr_vscsi.c
@@ -494,7 +494,7 @@ static void vscsi_transfer_data(SCSIRequest *sreq, uint32_t len)
 }
 
 /* Callback to indicate that the SCSI layer has completed a transfer.  */
-static void vscsi_command_complete(SCSIRequest *sreq, uint32_t status)
+static void vscsi_command_complete(SCSIRequest *sreq, uint32_t status, size_t resid)
 {
     VSCSIState *s = DO_UPCAST(VSCSIState, vdev.qdev, sreq->bus->qbus.parent);
     vscsi_req *req = sreq->hba_private;
diff --git a/hw/usb-msd.c b/hw/usb-msd.c
index c933efe19a..5fbd2d021b 100644
--- a/hw/usb-msd.c
+++ b/hw/usb-msd.c
@@ -223,7 +223,7 @@ static void usb_msd_transfer_data(SCSIRequest *req, uint32_t len)
     }
 }
 
-static void usb_msd_command_complete(SCSIRequest *req, uint32_t status)
+static void usb_msd_command_complete(SCSIRequest *req, uint32_t status, size_t resid)
 {
     MSDState *s = DO_UPCAST(MSDState, dev.qdev, req->bus->qbus.parent);
     USBPacket *p = s->packet;
diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
index 907b52a25d..a0fb7c1b9c 100644
--- a/hw/virtio-pci.c
+++ b/hw/virtio-pci.c
@@ -21,6 +21,7 @@
 #include "virtio-blk.h"
 #include "virtio-net.h"
 #include "virtio-serial.h"
+#include "virtio-scsi.h"
 #include "pci.h"
 #include "qemu-error.h"
 #include "msix.h"
@@ -930,12 +931,67 @@ static TypeInfo virtio_balloon_info = {
     .class_init    = virtio_balloon_class_init,
 };
 
+static int virtio_scsi_init_pci(PCIDevice *pci_dev)
+{
+    VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
+    VirtIODevice *vdev;
+
+    vdev = virtio_scsi_init(&pci_dev->qdev, &proxy->scsi);
+    if (!vdev) {
+        return -EINVAL;
+    }
+
+    vdev->nvectors = proxy->nvectors;
+    virtio_init_pci(proxy, vdev);
+
+    /* make the actual value visible */
+    proxy->nvectors = vdev->nvectors;
+    return 0;
+}
+
+static int virtio_scsi_exit_pci(PCIDevice *pci_dev)
+{
+    VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
+
+    virtio_scsi_exit(proxy->vdev);
+    return virtio_exit_pci(pci_dev);
+}
+
+static Property virtio_scsi_properties[] = {
+    DEFINE_PROP_UINT32("vectors", VirtIOPCIProxy, nvectors, 2),
+    DEFINE_VIRTIO_SCSI_PROPERTIES(VirtIOPCIProxy, host_features, scsi),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void virtio_scsi_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+
+    k->init = virtio_scsi_init_pci;
+    k->exit = virtio_scsi_exit_pci;
+    k->vendor_id = PCI_VENDOR_ID_REDHAT_QUMRANET;
+    k->device_id = PCI_DEVICE_ID_VIRTIO_SCSI;
+    k->revision = 0x00;
+    k->class_id = PCI_CLASS_STORAGE_SCSI;
+    dc->reset = virtio_pci_reset;
+    dc->props = virtio_scsi_properties;
+}
+
+static TypeInfo virtio_scsi_info = {
+    .name          = "virtio-scsi-pci",
+    .parent        = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(VirtIOPCIProxy),
+    .class_init    = virtio_scsi_class_init,
+};
+
 static void virtio_pci_register_types(void)
 {
     type_register_static(&virtio_blk_info);
     type_register_static(&virtio_net_info);
     type_register_static(&virtio_serial_info);
     type_register_static(&virtio_balloon_info);
+    type_register_static(&virtio_scsi_info);
 }
 
 type_init(virtio_pci_register_types)
diff --git a/hw/virtio-pci.h b/hw/virtio-pci.h
index 344c22b68f..e5604282e5 100644
--- a/hw/virtio-pci.h
+++ b/hw/virtio-pci.h
@@ -17,6 +17,7 @@
 
 #include "virtio-net.h"
 #include "virtio-serial.h"
+#include "virtio-scsi.h"
 
 /* Performance improves when virtqueue kick processing is decoupled from the
  * vcpu thread using ioeventfd for some devices. */
@@ -40,6 +41,7 @@ typedef struct {
 #endif
     virtio_serial_conf serial;
     virtio_net_conf net;
+    VirtIOSCSIConf scsi;
     bool ioeventfd_disabled;
     bool ioeventfd_started;
 } VirtIOPCIProxy;
diff --git a/hw/virtio-scsi.c b/hw/virtio-scsi.c
new file mode 100644
index 0000000000..e607edc915
--- /dev/null
+++ b/hw/virtio-scsi.c
@@ -0,0 +1,617 @@
+/*
+ * Virtio SCSI HBA
+ *
+ * Copyright IBM, Corp. 2010
+ * Copyright Red Hat, Inc. 2011
+ *
+ * Authors:
+ *   Stefan Hajnoczi    <stefanha@linux.vnet.ibm.com>
+ *   Paolo Bonzini      <pbonzini@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "virtio-scsi.h"
+#include <hw/scsi.h>
+#include <hw/scsi-defs.h>
+
+#define VIRTIO_SCSI_VQ_SIZE     128
+#define VIRTIO_SCSI_CDB_SIZE    32
+#define VIRTIO_SCSI_SENSE_SIZE  96
+#define VIRTIO_SCSI_MAX_CHANNEL 0
+#define VIRTIO_SCSI_MAX_TARGET  255
+#define VIRTIO_SCSI_MAX_LUN     16383
+
+/* Response codes */
+#define VIRTIO_SCSI_S_OK                       0
+#define VIRTIO_SCSI_S_OVERRUN                  1
+#define VIRTIO_SCSI_S_ABORTED                  2
+#define VIRTIO_SCSI_S_BAD_TARGET               3
+#define VIRTIO_SCSI_S_RESET                    4
+#define VIRTIO_SCSI_S_BUSY                     5
+#define VIRTIO_SCSI_S_TRANSPORT_FAILURE        6
+#define VIRTIO_SCSI_S_TARGET_FAILURE           7
+#define VIRTIO_SCSI_S_NEXUS_FAILURE            8
+#define VIRTIO_SCSI_S_FAILURE                  9
+#define VIRTIO_SCSI_S_FUNCTION_SUCCEEDED       10
+#define VIRTIO_SCSI_S_FUNCTION_REJECTED        11
+#define VIRTIO_SCSI_S_INCORRECT_LUN            12
+
+/* Controlq type codes.  */
+#define VIRTIO_SCSI_T_TMF                      0
+#define VIRTIO_SCSI_T_AN_QUERY                 1
+#define VIRTIO_SCSI_T_AN_SUBSCRIBE             2
+
+/* Valid TMF subtypes.  */
+#define VIRTIO_SCSI_T_TMF_ABORT_TASK           0
+#define VIRTIO_SCSI_T_TMF_ABORT_TASK_SET       1
+#define VIRTIO_SCSI_T_TMF_CLEAR_ACA            2
+#define VIRTIO_SCSI_T_TMF_CLEAR_TASK_SET       3
+#define VIRTIO_SCSI_T_TMF_I_T_NEXUS_RESET      4
+#define VIRTIO_SCSI_T_TMF_LOGICAL_UNIT_RESET   5
+#define VIRTIO_SCSI_T_TMF_QUERY_TASK           6
+#define VIRTIO_SCSI_T_TMF_QUERY_TASK_SET       7
+
+/* Events.  */
+#define VIRTIO_SCSI_T_EVENTS_MISSED            0x80000000
+#define VIRTIO_SCSI_T_NO_EVENT                 0
+#define VIRTIO_SCSI_T_TRANSPORT_RESET          1
+#define VIRTIO_SCSI_T_ASYNC_NOTIFY             2
+
+/* SCSI command request, followed by data-out */
+typedef struct {
+    uint8_t lun[8];              /* Logical Unit Number */
+    uint64_t tag;                /* Command identifier */
+    uint8_t task_attr;           /* Task attribute */
+    uint8_t prio;
+    uint8_t crn;
+    uint8_t cdb[];
+} QEMU_PACKED VirtIOSCSICmdReq;
+
+/* Response, followed by sense data and data-in */
+typedef struct {
+    uint32_t sense_len;          /* Sense data length */
+    uint32_t resid;              /* Residual bytes in data buffer */
+    uint16_t status_qualifier;   /* Status qualifier */
+    uint8_t status;              /* Command completion status */
+    uint8_t response;            /* Response values */
+    uint8_t sense[];
+} QEMU_PACKED VirtIOSCSICmdResp;
+
+/* Task Management Request */
+typedef struct {
+    uint32_t type;
+    uint32_t subtype;
+    uint8_t lun[8];
+    uint64_t tag;
+} QEMU_PACKED VirtIOSCSICtrlTMFReq;
+
+typedef struct {
+    uint8_t response;
+} QEMU_PACKED VirtIOSCSICtrlTMFResp;
+
+/* Asynchronous notification query/subscription */
+typedef struct {
+    uint32_t type;
+    uint8_t lun[8];
+    uint32_t event_requested;
+} QEMU_PACKED VirtIOSCSICtrlANReq;
+
+typedef struct {
+    uint32_t event_actual;
+    uint8_t response;
+} QEMU_PACKED VirtIOSCSICtrlANResp;
+
+typedef struct {
+    uint32_t event;
+    uint8_t lun[8];
+    uint32_t reason;
+} QEMU_PACKED VirtIOSCSIEvent;
+
+typedef struct {
+    uint32_t num_queues;
+    uint32_t seg_max;
+    uint32_t max_sectors;
+    uint32_t cmd_per_lun;
+    uint32_t event_info_size;
+    uint32_t sense_size;
+    uint32_t cdb_size;
+    uint16_t max_channel;
+    uint16_t max_target;
+    uint32_t max_lun;
+} QEMU_PACKED VirtIOSCSIConfig;
+
+typedef struct {
+    VirtIODevice vdev;
+    DeviceState *qdev;
+    VirtIOSCSIConf *conf;
+
+    SCSIBus bus;
+    VirtQueue *ctrl_vq;
+    VirtQueue *event_vq;
+    VirtQueue *cmd_vq;
+    uint32_t sense_size;
+    uint32_t cdb_size;
+    int resetting;
+} VirtIOSCSI;
+
+typedef struct VirtIOSCSIReq {
+    VirtIOSCSI *dev;
+    VirtQueue *vq;
+    VirtQueueElement elem;
+    QEMUSGList qsgl;
+    SCSIRequest *sreq;
+    union {
+        char                  *buf;
+        VirtIOSCSICmdReq      *cmd;
+        VirtIOSCSICtrlTMFReq  *tmf;
+        VirtIOSCSICtrlANReq   *an;
+    } req;
+    union {
+        char                  *buf;
+        VirtIOSCSICmdResp     *cmd;
+        VirtIOSCSICtrlTMFResp *tmf;
+        VirtIOSCSICtrlANResp  *an;
+        VirtIOSCSIEvent       *event;
+    } resp;
+} VirtIOSCSIReq;
+
+static inline int virtio_scsi_get_lun(uint8_t *lun)
+{
+    return ((lun[2] << 8) | lun[3]) & 0x3FFF;
+}
+
+static inline SCSIDevice *virtio_scsi_device_find(VirtIOSCSI *s, uint8_t *lun)
+{
+    if (lun[0] != 1) {
+        return NULL;
+    }
+    if (lun[2] != 0 && !(lun[2] >= 0x40 && lun[2] < 0x80)) {
+        return NULL;
+    }
+    return scsi_device_find(&s->bus, 0, lun[1], virtio_scsi_get_lun(lun));
+}
+
+static void virtio_scsi_complete_req(VirtIOSCSIReq *req)
+{
+    VirtIOSCSI *s = req->dev;
+    VirtQueue *vq = req->vq;
+    virtqueue_push(vq, &req->elem, req->qsgl.size + req->elem.in_sg[0].iov_len);
+    qemu_sglist_destroy(&req->qsgl);
+    if (req->sreq) {
+        req->sreq->hba_private = NULL;
+        scsi_req_unref(req->sreq);
+    }
+    g_free(req);
+    virtio_notify(&s->vdev, vq);
+}
+
+static void virtio_scsi_bad_req(void)
+{
+    error_report("wrong size for virtio-scsi headers");
+    exit(1);
+}
+
+static void qemu_sgl_init_external(QEMUSGList *qsgl, struct iovec *sg,
+                                   target_phys_addr_t *addr, int num)
+{
+    memset(qsgl, 0, sizeof(*qsgl));
+    while (num--) {
+        qemu_sglist_add(qsgl, *(addr++), (sg++)->iov_len);
+    }
+}
+
+static void virtio_scsi_parse_req(VirtIOSCSI *s, VirtQueue *vq,
+                                  VirtIOSCSIReq *req)
+{
+    assert(req->elem.out_num && req->elem.in_num);
+    req->vq = vq;
+    req->dev = s;
+    req->sreq = NULL;
+    req->req.buf = req->elem.out_sg[0].iov_base;
+    req->resp.buf = req->elem.in_sg[0].iov_base;
+
+    if (req->elem.out_num > 1) {
+        qemu_sgl_init_external(&req->qsgl, &req->elem.out_sg[1],
+                               &req->elem.out_addr[1],
+                               req->elem.out_num - 1);
+    } else {
+        qemu_sgl_init_external(&req->qsgl, &req->elem.in_sg[1],
+                               &req->elem.in_addr[1],
+                               req->elem.in_num - 1);
+    }
+}
+
+static VirtIOSCSIReq *virtio_scsi_pop_req(VirtIOSCSI *s, VirtQueue *vq)
+{
+    VirtIOSCSIReq *req;
+    req = g_malloc(sizeof(*req));
+    if (!virtqueue_pop(vq, &req->elem)) {
+        g_free(req);
+        return NULL;
+    }
+
+    virtio_scsi_parse_req(s, vq, req);
+    return req;
+}
+
+static void virtio_scsi_save_request(QEMUFile *f, SCSIRequest *sreq)
+{
+    VirtIOSCSIReq *req = sreq->hba_private;
+
+    qemu_put_buffer(f, (unsigned char *)&req->elem, sizeof(req->elem));
+}
+
+static void *virtio_scsi_load_request(QEMUFile *f, SCSIRequest *sreq)
+{
+    SCSIBus *bus = sreq->bus;
+    VirtIOSCSI *s = container_of(bus, VirtIOSCSI, bus);
+    VirtIOSCSIReq *req;
+
+    req = g_malloc(sizeof(*req));
+    qemu_get_buffer(f, (unsigned char *)&req->elem, sizeof(req->elem));
+    virtio_scsi_parse_req(s, s->cmd_vq, req);
+
+    scsi_req_ref(sreq);
+    req->sreq = sreq;
+    if (req->sreq->cmd.mode != SCSI_XFER_NONE) {
+        int req_mode =
+            (req->elem.in_num > 1 ? SCSI_XFER_FROM_DEV : SCSI_XFER_TO_DEV);
+
+        assert(req->sreq->cmd.mode == req_mode);
+    }
+    return req;
+}
+
+static void virtio_scsi_do_tmf(VirtIOSCSI *s, VirtIOSCSIReq *req)
+{
+    SCSIDevice *d = virtio_scsi_device_find(s, req->req.tmf->lun);
+    SCSIRequest *r, *next;
+    DeviceState *qdev;
+    int target;
+
+    /* Here VIRTIO_SCSI_S_OK means "FUNCTION COMPLETE".  */
+    req->resp.tmf->response = VIRTIO_SCSI_S_OK;
+
+    switch (req->req.tmf->subtype) {
+    case VIRTIO_SCSI_T_TMF_ABORT_TASK:
+    case VIRTIO_SCSI_T_TMF_QUERY_TASK:
+        if (!d) {
+            goto fail;
+        }
+        if (d->lun != virtio_scsi_get_lun(req->req.tmf->lun)) {
+            goto incorrect_lun;
+        }
+        QTAILQ_FOREACH_SAFE(r, &d->requests, next, next) {
+            if (r->tag == req->req.tmf->tag) {
+                break;
+            }
+        }
+        if (r && r->hba_private) {
+            if (req->req.tmf->subtype == VIRTIO_SCSI_T_TMF_QUERY_TASK) {
+                /* "If the specified command is present in the task set, then
+                 * return a service response set to FUNCTION SUCCEEDED".
+                 */
+                req->resp.tmf->response = VIRTIO_SCSI_S_FUNCTION_SUCCEEDED;
+            } else {
+                scsi_req_cancel(r);
+            }
+        }
+        break;
+
+    case VIRTIO_SCSI_T_TMF_LOGICAL_UNIT_RESET:
+        if (!d) {
+            goto fail;
+        }
+        if (d->lun != virtio_scsi_get_lun(req->req.tmf->lun)) {
+            goto incorrect_lun;
+        }
+        s->resetting++;
+        qdev_reset_all(&d->qdev);
+        s->resetting--;
+        break;
+
+    case VIRTIO_SCSI_T_TMF_ABORT_TASK_SET:
+    case VIRTIO_SCSI_T_TMF_CLEAR_TASK_SET:
+    case VIRTIO_SCSI_T_TMF_QUERY_TASK_SET:
+        if (!d) {
+            goto fail;
+        }
+        if (d->lun != virtio_scsi_get_lun(req->req.tmf->lun)) {
+            goto incorrect_lun;
+        }
+        QTAILQ_FOREACH_SAFE(r, &d->requests, next, next) {
+            if (r->hba_private) {
+                if (req->req.tmf->subtype == VIRTIO_SCSI_T_TMF_QUERY_TASK_SET) {
+                    /* "If there is any command present in the task set, then
+                     * return a service response set to FUNCTION SUCCEEDED".
+                     */
+                    req->resp.tmf->response = VIRTIO_SCSI_S_FUNCTION_SUCCEEDED;
+                    break;
+                } else {
+                    scsi_req_cancel(r);
+                }
+            }
+        }
+        break;
+
+    case VIRTIO_SCSI_T_TMF_I_T_NEXUS_RESET:
+        target = req->req.tmf->lun[1];
+        s->resetting++;
+        QTAILQ_FOREACH(qdev, &s->bus.qbus.children, sibling) {
+             d = DO_UPCAST(SCSIDevice, qdev, qdev);
+             if (d->channel == 0 && d->id == target) {
+                qdev_reset_all(&d->qdev);
+             }
+        }
+        s->resetting--;
+        break;
+
+    case VIRTIO_SCSI_T_TMF_CLEAR_ACA:
+    default:
+        req->resp.tmf->response = VIRTIO_SCSI_S_FUNCTION_REJECTED;
+        break;
+    }
+
+    return;
+
+incorrect_lun:
+    req->resp.tmf->response = VIRTIO_SCSI_S_INCORRECT_LUN;
+    return;
+
+fail:
+    req->resp.tmf->response = VIRTIO_SCSI_S_BAD_TARGET;
+}
+
+static void virtio_scsi_handle_ctrl(VirtIODevice *vdev, VirtQueue *vq)
+{
+    VirtIOSCSI *s = (VirtIOSCSI *)vdev;
+    VirtIOSCSIReq *req;
+
+    while ((req = virtio_scsi_pop_req(s, vq))) {
+        int out_size, in_size;
+        if (req->elem.out_num < 1 || req->elem.in_num < 1) {
+            virtio_scsi_bad_req();
+            continue;
+        }
+
+        out_size = req->elem.out_sg[0].iov_len;
+        in_size = req->elem.in_sg[0].iov_len;
+        if (req->req.tmf->type == VIRTIO_SCSI_T_TMF) {
+            if (out_size < sizeof(VirtIOSCSICtrlTMFReq) ||
+                in_size < sizeof(VirtIOSCSICtrlTMFResp)) {
+                virtio_scsi_bad_req();
+            }
+            virtio_scsi_do_tmf(s, req);
+
+        } else if (req->req.tmf->type == VIRTIO_SCSI_T_AN_QUERY ||
+                   req->req.tmf->type == VIRTIO_SCSI_T_AN_SUBSCRIBE) {
+            if (out_size < sizeof(VirtIOSCSICtrlANReq) ||
+                in_size < sizeof(VirtIOSCSICtrlANResp)) {
+                virtio_scsi_bad_req();
+            }
+            req->resp.an->event_actual = 0;
+            req->resp.an->response = VIRTIO_SCSI_S_OK;
+        }
+        virtio_scsi_complete_req(req);
+    }
+}
+
+static void virtio_scsi_command_complete(SCSIRequest *r, uint32_t status,
+                                         size_t resid)
+{
+    VirtIOSCSIReq *req = r->hba_private;
+
+    req->resp.cmd->response = VIRTIO_SCSI_S_OK;
+    req->resp.cmd->status = status;
+    if (req->resp.cmd->status == GOOD) {
+        req->resp.cmd->resid = resid;
+    } else {
+        req->resp.cmd->resid = 0;
+        req->resp.cmd->sense_len =
+            scsi_req_get_sense(r, req->resp.cmd->sense, VIRTIO_SCSI_SENSE_SIZE);
+    }
+    virtio_scsi_complete_req(req);
+}
+
+static QEMUSGList *virtio_scsi_get_sg_list(SCSIRequest *r)
+{
+    VirtIOSCSIReq *req = r->hba_private;
+
+    return &req->qsgl;
+}
+
+static void virtio_scsi_request_cancelled(SCSIRequest *r)
+{
+    VirtIOSCSIReq *req = r->hba_private;
+
+    if (!req) {
+        return;
+    }
+    if (req->dev->resetting) {
+        req->resp.cmd->response = VIRTIO_SCSI_S_RESET;
+    } else {
+        req->resp.cmd->response = VIRTIO_SCSI_S_ABORTED;
+    }
+    virtio_scsi_complete_req(req);
+}
+
+static void virtio_scsi_fail_cmd_req(VirtIOSCSIReq *req)
+{
+    req->resp.cmd->response = VIRTIO_SCSI_S_FAILURE;
+    virtio_scsi_complete_req(req);
+}
+
+static void virtio_scsi_handle_cmd(VirtIODevice *vdev, VirtQueue *vq)
+{
+    VirtIOSCSI *s = (VirtIOSCSI *)vdev;
+    VirtIOSCSIReq *req;
+    int n;
+
+    while ((req = virtio_scsi_pop_req(s, vq))) {
+        SCSIDevice *d;
+        int out_size, in_size;
+        if (req->elem.out_num < 1 || req->elem.in_num < 1) {
+            virtio_scsi_bad_req();
+        }
+
+        out_size = req->elem.out_sg[0].iov_len;
+        in_size = req->elem.in_sg[0].iov_len;
+        if (out_size < sizeof(VirtIOSCSICmdReq) + s->cdb_size ||
+            in_size < sizeof(VirtIOSCSICmdResp) + s->sense_size) {
+            virtio_scsi_bad_req();
+        }
+
+        if (req->elem.out_num > 1 && req->elem.in_num > 1) {
+            virtio_scsi_fail_cmd_req(req);
+            continue;
+        }
+
+        d = virtio_scsi_device_find(s, req->req.cmd->lun);
+        if (!d) {
+            req->resp.cmd->response = VIRTIO_SCSI_S_BAD_TARGET;
+            virtio_scsi_complete_req(req);
+            continue;
+        }
+        req->sreq = scsi_req_new(d, req->req.cmd->tag,
+                                 virtio_scsi_get_lun(req->req.cmd->lun),
+                                 req->req.cmd->cdb, req);
+
+        if (req->sreq->cmd.mode != SCSI_XFER_NONE) {
+            int req_mode =
+                (req->elem.in_num > 1 ? SCSI_XFER_FROM_DEV : SCSI_XFER_TO_DEV);
+
+            if (req->sreq->cmd.mode != req_mode ||
+                req->sreq->cmd.xfer > req->qsgl.size) {
+                req->resp.cmd->response = VIRTIO_SCSI_S_OVERRUN;
+                virtio_scsi_complete_req(req);
+                continue;
+            }
+        }
+
+        n = scsi_req_enqueue(req->sreq);
+        if (n) {
+            scsi_req_continue(req->sreq);
+        }
+    }
+}
+
+static void virtio_scsi_get_config(VirtIODevice *vdev,
+                                   uint8_t *config)
+{
+    VirtIOSCSIConfig *scsiconf = (VirtIOSCSIConfig *)config;
+    VirtIOSCSI *s = (VirtIOSCSI *)vdev;
+
+    stl_raw(&scsiconf->num_queues, s->conf->num_queues);
+    stl_raw(&scsiconf->seg_max, 128 - 2);
+    stl_raw(&scsiconf->max_sectors, s->conf->max_sectors);
+    stl_raw(&scsiconf->cmd_per_lun, s->conf->cmd_per_lun);
+    stl_raw(&scsiconf->event_info_size, sizeof(VirtIOSCSIEvent));
+    stl_raw(&scsiconf->sense_size, s->sense_size);
+    stl_raw(&scsiconf->cdb_size, s->cdb_size);
+    stl_raw(&scsiconf->max_channel, VIRTIO_SCSI_MAX_CHANNEL);
+    stl_raw(&scsiconf->max_target, VIRTIO_SCSI_MAX_TARGET);
+    stl_raw(&scsiconf->max_lun, VIRTIO_SCSI_MAX_LUN);
+}
+
+static void virtio_scsi_set_config(VirtIODevice *vdev,
+                                   const uint8_t *config)
+{
+    VirtIOSCSIConfig *scsiconf = (VirtIOSCSIConfig *)config;
+    VirtIOSCSI *s = (VirtIOSCSI *)vdev;
+
+    if ((uint32_t) ldl_raw(&scsiconf->sense_size) >= 65536 ||
+        (uint32_t) ldl_raw(&scsiconf->cdb_size) >= 256) {
+        error_report("bad data written to virtio-scsi configuration space");
+        exit(1);
+    }
+
+    s->sense_size = ldl_raw(&scsiconf->sense_size);
+    s->cdb_size = ldl_raw(&scsiconf->cdb_size);
+}
+
+static uint32_t virtio_scsi_get_features(VirtIODevice *vdev,
+                                         uint32_t requested_features)
+{
+    return requested_features;
+}
+
+static void virtio_scsi_reset(VirtIODevice *vdev)
+{
+    VirtIOSCSI *s = (VirtIOSCSI *)vdev;
+
+    s->sense_size = VIRTIO_SCSI_SENSE_SIZE;
+    s->cdb_size = VIRTIO_SCSI_CDB_SIZE;
+}
+
+/* The device does not have anything to save beyond the virtio data.
+ * Request data is saved with callbacks from SCSI devices.
+ */
+static void virtio_scsi_save(QEMUFile *f, void *opaque)
+{
+    VirtIOSCSI *s = opaque;
+    virtio_save(&s->vdev, f);
+}
+
+static int virtio_scsi_load(QEMUFile *f, void *opaque, int version_id)
+{
+    VirtIOSCSI *s = opaque;
+    virtio_load(&s->vdev, f);
+    return 0;
+}
+
+static struct SCSIBusInfo virtio_scsi_scsi_info = {
+    .tcq = true,
+    .max_channel = VIRTIO_SCSI_MAX_CHANNEL,
+    .max_target = VIRTIO_SCSI_MAX_TARGET,
+    .max_lun = VIRTIO_SCSI_MAX_LUN,
+
+    .complete = virtio_scsi_command_complete,
+    .cancel = virtio_scsi_request_cancelled,
+    .get_sg_list = virtio_scsi_get_sg_list,
+    .save_request = virtio_scsi_save_request,
+    .load_request = virtio_scsi_load_request,
+};
+
+VirtIODevice *virtio_scsi_init(DeviceState *dev, VirtIOSCSIConf *proxyconf)
+{
+    VirtIOSCSI *s;
+    static int virtio_scsi_id;
+
+    s = (VirtIOSCSI *)virtio_common_init("virtio-scsi", VIRTIO_ID_SCSI,
+                                         sizeof(VirtIOSCSIConfig),
+                                         sizeof(VirtIOSCSI));
+
+    s->qdev = dev;
+    s->conf = proxyconf;
+
+    /* TODO set up vdev function pointers */
+    s->vdev.get_config = virtio_scsi_get_config;
+    s->vdev.set_config = virtio_scsi_set_config;
+    s->vdev.get_features = virtio_scsi_get_features;
+    s->vdev.reset = virtio_scsi_reset;
+
+    s->ctrl_vq = virtio_add_queue(&s->vdev, VIRTIO_SCSI_VQ_SIZE,
+                                   virtio_scsi_handle_ctrl);
+    s->event_vq = virtio_add_queue(&s->vdev, VIRTIO_SCSI_VQ_SIZE,
+                                   NULL);
+    s->cmd_vq = virtio_add_queue(&s->vdev, VIRTIO_SCSI_VQ_SIZE,
+                                   virtio_scsi_handle_cmd);
+
+    scsi_bus_new(&s->bus, dev, &virtio_scsi_scsi_info);
+    if (!dev->hotplugged) {
+        scsi_bus_legacy_handle_cmdline(&s->bus);
+    }
+
+    register_savevm(dev, "virtio-scsi", virtio_scsi_id++, 1,
+                    virtio_scsi_save, virtio_scsi_load, s);
+
+    return &s->vdev;
+}
+
+void virtio_scsi_exit(VirtIODevice *vdev)
+{
+    virtio_cleanup(vdev);
+}
diff --git a/hw/virtio-scsi.h b/hw/virtio-scsi.h
new file mode 100644
index 0000000000..4bc889de02
--- /dev/null
+++ b/hw/virtio-scsi.h
@@ -0,0 +1,36 @@
+/*
+ * Virtio SCSI HBA
+ *
+ * Copyright IBM, Corp. 2010
+ *
+ * Authors:
+ *  Stefan Hajnoczi    <stefanha@linux.vnet.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef _QEMU_VIRTIO_SCSI_H
+#define _QEMU_VIRTIO_SCSI_H
+
+#include "virtio.h"
+#include "net.h"
+#include "pci.h"
+
+/* The ID for virtio_scsi */
+#define VIRTIO_ID_SCSI  8
+
+struct VirtIOSCSIConf {
+    uint32_t num_queues;
+    uint32_t max_sectors;
+    uint32_t cmd_per_lun;
+};
+
+#define DEFINE_VIRTIO_SCSI_PROPERTIES(_state, _features_field, _conf_field) \
+    DEFINE_VIRTIO_COMMON_FEATURES(_state, _features_field), \
+    DEFINE_PROP_UINT32("num_queues", _state, _conf_field.num_queues, 1), \
+    DEFINE_PROP_UINT32("max_sectors", _state, _conf_field.max_sectors, 0xFFFF), \
+    DEFINE_PROP_UINT32("cmd_per_lun", _state, _conf_field.cmd_per_lun, 128)
+
+#endif /* _QEMU_VIRTIO_SCSI_H */
diff --git a/hw/virtio.h b/hw/virtio.h
index 25f55647b4..400c092c95 100644
--- a/hw/virtio.h
+++ b/hw/virtio.h
@@ -199,6 +199,8 @@ VirtIODevice *virtio_net_init(DeviceState *dev, NICConf *conf,
 typedef struct virtio_serial_conf virtio_serial_conf;
 VirtIODevice *virtio_serial_init(DeviceState *dev, virtio_serial_conf *serial);
 VirtIODevice *virtio_balloon_init(DeviceState *dev);
+typedef struct VirtIOSCSIConf VirtIOSCSIConf;
+VirtIODevice *virtio_scsi_init(DeviceState *dev, VirtIOSCSIConf *conf);
 #ifdef CONFIG_LINUX
 VirtIODevice *virtio_9p_init(DeviceState *dev, V9fsConf *conf);
 #endif
@@ -208,6 +210,7 @@ void virtio_net_exit(VirtIODevice *vdev);
 void virtio_blk_exit(VirtIODevice *vdev);
 void virtio_serial_exit(VirtIODevice *vdev);
 void virtio_balloon_exit(VirtIODevice *vdev);
+void virtio_scsi_exit(VirtIODevice *vdev);
 
 #define DEFINE_VIRTIO_COMMON_FEATURES(_state, _field) \
 	DEFINE_PROP_BIT("indirect_desc", _state, _field, \