* [Qemu-devel] [PATCH 0/2] Introduce vhost-user-scsi and sample application
@ 2016-10-26 15:26 Felipe Franciosi
2016-10-26 15:26 ` [Qemu-devel] [PATCH 1/2] vus: Introduce vhost-user-scsi host device Felipe Franciosi
2016-10-26 15:26 ` [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application Felipe Franciosi
0 siblings, 2 replies; 7+ messages in thread
From: Felipe Franciosi @ 2016-10-26 15:26 UTC (permalink / raw)
To: Paolo Bonzini, Stefan Hajnoczi, Marc-Andre Lureau, Michael S. Tsirkin
Cc: qemu-devel, Felipe Franciosi
Based on various discussions on the 2016 KVM Forum, I'm sending over a
vhost-user-scsi implementation for your consideration.
This patch set introduces a new vhost-user SCSI device. While heavily
based on vhost-scsi, it is implemented using vhost's userspace
counterpart. The device has been coded and tested to work with live
migration.
A sample application based on the newly introduced libvhost-user is
also included. It makes use of libiscsi for simplicity.
For convenience, I'm maintaining an up-to-date version of these
patches (including some necessary fixes for libvhost-user still under
discussion) on:
https://github.com/franciozzy/qemu/tree/vus-upstream
See the individual patches for build and use instructions.
Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
Felipe Franciosi (2):
vus: Introduce vhost-user-scsi host device
vus: Introduce a vhost-user-scsi sample application
configure | 10 +
hw/scsi/Makefile.objs | 1 +
hw/scsi/vhost-user-scsi.c | 299 +++++++++++++
hw/virtio/virtio-pci.c | 58 +++
hw/virtio/virtio-pci.h | 15 +
include/hw/virtio/vhost-user-scsi.h | 45 ++
include/hw/virtio/virtio-scsi.h | 5 +
tests/Makefile.include | 2 +
tests/vhost-user-scsi.c | 862 ++++++++++++++++++++++++++++++++++++
9 files changed, 1297 insertions(+)
create mode 100644 hw/scsi/vhost-user-scsi.c
create mode 100644 include/hw/virtio/vhost-user-scsi.h
create mode 100644 tests/vhost-user-scsi.c
--
1.9.4
^ permalink raw reply [flat|nested] 7+ messages in thread
* [Qemu-devel] [PATCH 1/2] vus: Introduce vhost-user-scsi host device
2016-10-26 15:26 [Qemu-devel] [PATCH 0/2] Introduce vhost-user-scsi and sample application Felipe Franciosi
@ 2016-10-26 15:26 ` Felipe Franciosi
2016-10-27 12:12 ` Paolo Bonzini
2016-10-26 15:26 ` [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application Felipe Franciosi
1 sibling, 1 reply; 7+ messages in thread
From: Felipe Franciosi @ 2016-10-26 15:26 UTC (permalink / raw)
To: Paolo Bonzini, Stefan Hajnoczi, Marc-Andre Lureau, Michael S. Tsirkin
Cc: qemu-devel, Felipe Franciosi
This commit introduces a vhost-user device for SCSI. This is based
on the existing vhost-scsi implementation, but done over vhost-user
instead. It also uses a chardev to connect to the backend. Unlike
vhost-scsi (today), VMs using vhost-user-scsi can be live migrated.
To use it, one must configure Qemu with --enable-vhost-user-scsi and
start Qemu with a command line equivalent to:
qemu-system-x86_64 \
-chardev socket,id=vus0,path=/tmp/vus.sock \
-device vhost-user-scsi-pci,chardev=vus0,bus=pci.0,addr=0x4
A separate commit presents a sample application linked with libiscsi to
provide a backend for vhost-user-scsi.
Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
---
configure | 10 ++
hw/scsi/Makefile.objs | 1 +
hw/scsi/vhost-user-scsi.c | 299 ++++++++++++++++++++++++++++++++++++
hw/virtio/virtio-pci.c | 58 +++++++
hw/virtio/virtio-pci.h | 15 ++
include/hw/virtio/vhost-user-scsi.h | 45 ++++++
include/hw/virtio/virtio-scsi.h | 5 +
7 files changed, 433 insertions(+)
create mode 100644 hw/scsi/vhost-user-scsi.c
create mode 100644 include/hw/virtio/vhost-user-scsi.h
diff --git a/configure b/configure
index d3dafcb..0574ff2 100755
--- a/configure
+++ b/configure
@@ -228,6 +228,7 @@ xfs=""
vhost_net="no"
vhost_scsi="no"
+vhost_user_scsi="no"
vhost_vsock="no"
kvm="no"
rdma=""
@@ -677,6 +678,7 @@ Haiku)
kvm="yes"
vhost_net="yes"
vhost_scsi="yes"
+ vhost_user_scsi="yes"
vhost_vsock="yes"
QEMU_INCLUDES="-I\$(SRC_PATH)/linux-headers -I$(pwd)/linux-headers $QEMU_INCLUDES"
;;
@@ -1019,6 +1021,10 @@ for opt do
;;
--enable-vhost-scsi) vhost_scsi="yes"
;;
+ --disable-vhost-user-scsi) vhost_user_scsi="no"
+ ;;
+ --enable-vhost-user-scsi) vhost_user_scsi="yes"
+ ;;
--disable-vhost-vsock) vhost_vsock="no"
;;
--enable-vhost-vsock) vhost_vsock="yes"
@@ -4951,6 +4957,7 @@ echo "posix_madvise $posix_madvise"
echo "libcap-ng support $cap_ng"
echo "vhost-net support $vhost_net"
echo "vhost-scsi support $vhost_scsi"
+echo "vhost-user-scsi support $vhost_user_scsi"
echo "vhost-vsock support $vhost_vsock"
echo "Trace backends $trace_backends"
if have_backend "simple"; then
@@ -5336,6 +5343,9 @@ fi
if test "$vhost_scsi" = "yes" ; then
echo "CONFIG_VHOST_SCSI=y" >> $config_host_mak
fi
+if test "$vhost_user_scsi" = "yes" ; then
+ echo "CONFIG_VHOST_USER_SCSI=y" >> $config_host_mak
+fi
if test "$vhost_net" = "yes" ; then
echo "CONFIG_VHOST_NET_USED=y" >> $config_host_mak
fi
diff --git a/hw/scsi/Makefile.objs b/hw/scsi/Makefile.objs
index 5a2248b..3338aad 100644
--- a/hw/scsi/Makefile.objs
+++ b/hw/scsi/Makefile.objs
@@ -11,4 +11,5 @@ obj-$(CONFIG_PSERIES) += spapr_vscsi.o
ifeq ($(CONFIG_VIRTIO),y)
obj-y += virtio-scsi.o virtio-scsi-dataplane.o
obj-$(CONFIG_VHOST_SCSI) += vhost-scsi.o
+obj-$(CONFIG_VHOST_USER_SCSI) += vhost-user-scsi.o
endif
diff --git a/hw/scsi/vhost-user-scsi.c b/hw/scsi/vhost-user-scsi.c
new file mode 100644
index 0000000..a7b9839
--- /dev/null
+++ b/hw/scsi/vhost-user-scsi.c
@@ -0,0 +1,299 @@
+/*
+ * vhost-user-scsi host device
+ *
+ * Copyright (c) 2016 Nutanix Inc. All rights reserved.
+ *
+ * Author:
+ * Felipe Franciosi <felipe@nutanix.com>
+ *
+ * This work is largely based on the "vhost-scsi" implementation by:
+ * Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
+ * Nicholas Bellinger <nab@risingtidesystems.com>
+ *
+ * This work is licensed under the terms of the GNU LGPL, version 2 or later.
+ * See the COPYING.LIB file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "migration/vmstate.h"
+#include "qapi/error.h"
+#include "qemu/error-report.h"
+#include "qemu/typedefs.h"
+#include "qom/object.h"
+#include "hw/fw-path-provider.h"
+#include "hw/qdev-core.h"
+#include "hw/virtio/vhost.h"
+#include "hw/virtio/vhost-backend.h"
+#include "hw/virtio/vhost-user-scsi.h"
+#include "hw/virtio/virtio.h"
+#include "hw/virtio/virtio-access.h"
+#include "hw/virtio/virtio-bus.h"
+#include "sysemu/char.h"
+
+/* Features supported by the host application */
+static const int user_feature_bits[] = {
+ VIRTIO_F_NOTIFY_ON_EMPTY,
+ VIRTIO_RING_F_INDIRECT_DESC,
+ VIRTIO_RING_F_EVENT_IDX,
+ VIRTIO_SCSI_F_HOTPLUG,
+ VHOST_INVALID_FEATURE_BIT
+};
+
+static int vhost_user_scsi_start(VHostUserSCSI *s)
+{
+ int ret, i;
+ VirtIODevice *vdev = VIRTIO_DEVICE(s);
+ BusState *qbus = BUS(qdev_get_parent_bus(DEVICE(vdev)));
+ VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus);
+
+ if (!k->set_guest_notifiers) {
+ error_report("binding does not support guest notifiers");
+ return -ENOSYS;
+ }
+
+ ret = vhost_dev_enable_notifiers(&s->dev, vdev);
+ if (ret < 0) {
+ return ret;
+ }
+
+ s->dev.acked_features = vdev->guest_features;
+ ret = vhost_dev_start(&s->dev, vdev);
+ if (ret < 0) {
+ error_report("Error starting vhost-user device");
+ goto err_notifiers;
+ }
+
+ ret = k->set_guest_notifiers(qbus->parent, s->dev.nvqs, true);
+ if (ret < 0) {
+ error_report("Error binding guest notifier");
+ goto err_vhost_stop;
+ }
+
+ /* guest_notifier_mask/pending not used yet, so just unmask
+ * everything here. virtio-pci will do the right thing by
+ * enabling/disabling irqfd.
+ */
+ for (i = 0; i < s->dev.nvqs; i++) {
+ vhost_virtqueue_mask(&s->dev, vdev, s->dev.vq_index + i, false);
+ }
+
+ return ret;
+
+err_vhost_stop:
+ vhost_dev_stop(&s->dev, vdev);
+err_notifiers:
+ vhost_dev_disable_notifiers(&s->dev, vdev);
+ return ret;
+}
+
+static void vhost_user_scsi_stop(VHostUserSCSI *s)
+{
+ VirtIODevice *vdev = VIRTIO_DEVICE(s);
+ BusState *qbus = BUS(qdev_get_parent_bus(DEVICE(vdev)));
+ VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus);
+ int ret = 0;
+
+ vhost_dev_stop(&s->dev, vdev);
+ vhost_dev_disable_notifiers(&s->dev, vdev);
+
+ if (k->set_guest_notifiers) {
+ ret = k->set_guest_notifiers(qbus->parent, s->dev.nvqs, false);
+ if (ret < 0) {
+ error_report("vhost guest notifier cleanup failed: %d", ret);
+ }
+ }
+ assert(ret >= 0);
+}
+
+static uint64_t vhost_user_scsi_get_features(VirtIODevice *vdev,
+ uint64_t features,
+ Error **errp)
+{
+ VHostUserSCSI *s = VHOST_USER_SCSI(vdev);
+
+ return vhost_get_features(&s->dev, user_feature_bits, features);
+}
+
+static void vhost_user_scsi_set_config(VirtIODevice *vdev,
+ const uint8_t *config)
+{
+ VirtIOSCSIConfig *scsiconf = (VirtIOSCSIConfig *)config;
+ VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(vdev);
+
+ if ((uint32_t)virtio_ldl_p(vdev, &scsiconf->sense_size) != vs->sense_size ||
+ (uint32_t)virtio_ldl_p(vdev, &scsiconf->cdb_size) != vs->cdb_size) {
+ error_report("vhost-user-scsi doesn't allow sense or CDB sizes change");
+ exit(1);
+ }
+}
+
+static void vhost_user_scsi_set_status(VirtIODevice *vdev, uint8_t status)
+{
+ VHostUserSCSI *s = (VHostUserSCSI *)vdev;
+ bool start = (status & VIRTIO_CONFIG_S_DRIVER_OK) && vdev->vm_running;
+
+ if (s->dev.started == start) {
+ return;
+ }
+
+ if (start) {
+ int ret;
+
+ ret = vhost_user_scsi_start(s);
+ if (ret < 0) {
+ error_report("unable to start vhost-user-scsi: %s", strerror(-ret));
+ exit(1);
+ }
+ } else {
+ vhost_user_scsi_stop(s);
+ }
+}
+
+static void vhost_dummy_handle_output(VirtIODevice *vdev, VirtQueue *vq)
+{
+}
+
+static void vhost_user_scsi_save(QEMUFile *f, void *opaque)
+{
+ VirtIODevice *vdev = VIRTIO_DEVICE(opaque);
+ virtio_save(vdev, f);
+}
+
+static int vhost_user_scsi_load(QEMUFile *f, void *opaque, int version_id)
+{
+ VirtIODevice *vdev = VIRTIO_DEVICE(opaque);
+ return virtio_load(vdev, f, version_id);
+}
+
+static void vhost_user_scsi_realize(DeviceState *dev, Error **errp)
+{
+ VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(dev);
+ VHostUserSCSI *s = VHOST_USER_SCSI(dev);
+ static int vhost_user_scsi_id;
+ Error *err = NULL;
+ int ret;
+
+ if (!vs->conf.chardev.chr) {
+ error_setg(errp, "vhost-user-scsi: missing chardev");
+ return;
+ }
+
+ virtio_scsi_common_realize(dev, &err, vhost_dummy_handle_output,
+ vhost_dummy_handle_output,
+ vhost_dummy_handle_output);
+ if (err != NULL) {
+ error_propagate(errp, err);
+ return;
+ }
+
+ s->dev.nvqs = VHOST_USER_SCSI_VQ_NUM_FIXED + vs->conf.num_queues;
+ s->dev.vqs = g_new(struct vhost_virtqueue, s->dev.nvqs);
+ s->dev.vq_index = 0;
+ s->dev.backend_features = 0;
+
+ ret = vhost_dev_init(&s->dev, (void *)&vs->conf.chardev,
+ VHOST_BACKEND_TYPE_USER, 0);
+ if (ret < 0) {
+ error_setg(errp, "vhost-user-scsi: vhost initialization failed: %s",
+ strerror(-ret));
+ return;
+ }
+
+ /* Channel and lun both are 0 for bootable vhost-user-scsi disk */
+ s->channel = 0;
+ s->lun = 0;
+ s->target = vs->conf.boot_tpgt;
+
+ register_savevm(dev, "vhost-user-scsi", vhost_user_scsi_id++, 1,
+ vhost_user_scsi_save, vhost_user_scsi_load, s);
+}
+
+static void vhost_user_scsi_unrealize(DeviceState *dev, Error **errp)
+{
+ VirtIODevice *vdev = VIRTIO_DEVICE(dev);
+ VHostUserSCSI *s = VHOST_USER_SCSI(dev);
+
+ /* This will stop the vhost backend. */
+ vhost_user_scsi_set_status(vdev, 0);
+
+ vhost_dev_cleanup(&s->dev);
+ g_free(s->dev.vqs);
+
+ virtio_scsi_common_unrealize(dev, errp);
+}
+
+/*
+ * Implementation of an interface to adjust firmware path
+ * for the bootindex property handling.
+ */
+static char *vhost_user_scsi_get_fw_dev_path(FWPathProvider *p, BusState *bus,
+ DeviceState *dev)
+{
+ VHostUserSCSI *s = VHOST_USER_SCSI(dev);
+ /* format: /channel@<channel>/vhost-user-scsi@<target>,<lun> */
+ return g_strdup_printf("/channel@%x/%s@%x,%x", s->channel,
+ qdev_fw_name(dev), s->target, s->lun);
+}
+
+static Property vhost_user_scsi_properties[] = {
+ DEFINE_PROP_CHR("chardev", VHostUserSCSI,
+ parent_obj.conf.chardev),
+ DEFINE_PROP_UINT32("boot_tpgt", VHostUserSCSI,
+ parent_obj.conf.boot_tpgt, 0),
+ DEFINE_PROP_UINT32("num_queues", VHostUserSCSI,
+ parent_obj.conf.num_queues, 1),
+ DEFINE_PROP_UINT32("max_sectors", VHostUserSCSI,
+ parent_obj.conf.max_sectors, 0xFFFF),
+ DEFINE_PROP_UINT32("cmd_per_lun", VHostUserSCSI,
+ parent_obj.conf.cmd_per_lun, 128),
+ DEFINE_PROP_END_OF_LIST(),
+};
+
+static void vhost_user_scsi_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ VirtioDeviceClass *vdc = VIRTIO_DEVICE_CLASS(klass);
+ FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(klass);
+
+ dc->props = vhost_user_scsi_properties;
+ set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
+ vdc->realize = vhost_user_scsi_realize;
+ vdc->unrealize = vhost_user_scsi_unrealize;
+ vdc->get_features = vhost_user_scsi_get_features;
+ vdc->set_config = vhost_user_scsi_set_config;
+ vdc->set_status = vhost_user_scsi_set_status;
+ fwc->get_dev_path = vhost_user_scsi_get_fw_dev_path;
+}
+
+static void vhost_user_scsi_instance_init(Object *obj)
+{
+ VHostUserSCSI *dev = VHOST_USER_SCSI(obj);
+ VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(obj);
+
+ // Add the bootindex property for this object
+ device_add_bootindex_property(obj, &dev->bootindex, "bootindex", NULL,
+ DEVICE(dev), NULL);
+
+ // Set boot index according the the device config
+ object_property_set_int(obj, vs->conf.bootindex, "bootindex", NULL);
+}
+
+static const TypeInfo vhost_user_scsi_info = {
+ .name = TYPE_VHOST_USER_SCSI,
+ .parent = TYPE_VIRTIO_SCSI_COMMON,
+ .instance_size = sizeof(VHostUserSCSI),
+ .class_init = vhost_user_scsi_class_init,
+ .instance_init = vhost_user_scsi_instance_init,
+ .interfaces = (InterfaceInfo[]) {
+ { TYPE_FW_PATH_PROVIDER },
+ { }
+ },
+};
+
+static void virtio_register_types(void)
+{
+ type_register_static(&vhost_user_scsi_info);
+}
+
+type_init(virtio_register_types)
diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
index 06831de..b996d37 100644
--- a/hw/virtio/virtio-pci.c
+++ b/hw/virtio/virtio-pci.c
@@ -2098,6 +2098,61 @@ static const TypeInfo vhost_scsi_pci_info = {
};
#endif
+/* vhost-user-scsi-pci */
+#ifdef CONFIG_VHOST_USER_SCSI
+static Property vhost_user_scsi_pci_properties[] = {
+ DEFINE_PROP_UINT32("vectors", VirtIOPCIProxy, nvectors,
+ DEV_NVECTORS_UNSPECIFIED),
+ DEFINE_PROP_END_OF_LIST(),
+};
+
+static void vhost_user_scsi_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
+{
+ VHostUserSCSIPCI *dev = VHOST_USER_SCSI_PCI(vpci_dev);
+ DeviceState *vdev = DEVICE(&dev->vdev);
+ VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(vdev);
+
+ if (vpci_dev->nvectors == DEV_NVECTORS_UNSPECIFIED) {
+ vpci_dev->nvectors = vs->conf.num_queues + 3;
+ }
+
+ qdev_set_parent_bus(vdev, BUS(&vpci_dev->bus));
+ object_property_set_bool(OBJECT(vdev), true, "realized", errp);
+}
+
+static void vhost_user_scsi_pci_class_init(ObjectClass *klass, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(klass);
+ VirtioPCIClass *k = VIRTIO_PCI_CLASS(klass);
+ PCIDeviceClass *pcidev_k = PCI_DEVICE_CLASS(klass);
+ k->realize = vhost_user_scsi_pci_realize;
+ set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
+ dc->props = vhost_user_scsi_pci_properties;
+ pcidev_k->vendor_id = PCI_VENDOR_ID_REDHAT_QUMRANET;
+ pcidev_k->device_id = PCI_DEVICE_ID_VIRTIO_SCSI;
+ pcidev_k->revision = 0x00;
+ pcidev_k->class_id = PCI_CLASS_STORAGE_SCSI;
+}
+
+static void vhost_user_scsi_pci_instance_init(Object *obj)
+{
+ VHostUserSCSIPCI *dev = VHOST_USER_SCSI_PCI(obj);
+
+ virtio_instance_init_common(obj, &dev->vdev, sizeof(dev->vdev),
+ TYPE_VHOST_USER_SCSI);
+ object_property_add_alias(obj, "bootindex", OBJECT(&dev->vdev),
+ "bootindex", &error_abort);
+}
+
+static const TypeInfo vhost_user_scsi_pci_info = {
+ .name = TYPE_VHOST_USER_SCSI_PCI,
+ .parent = TYPE_VIRTIO_PCI,
+ .instance_size = sizeof(VHostUserSCSIPCI),
+ .instance_init = vhost_user_scsi_pci_instance_init,
+ .class_init = vhost_user_scsi_pci_class_init,
+};
+#endif /* CONFIG_VHOST_USER_SCSI */
+
/* vhost-vsock-pci */
#ifdef CONFIG_VHOST_VSOCK
@@ -2577,6 +2632,9 @@ static void virtio_pci_register_types(void)
#ifdef CONFIG_VHOST_SCSI
type_register_static(&vhost_scsi_pci_info);
#endif
+#ifdef CONFIG_VHOST_USER_SCSI
+ type_register_static(&vhost_user_scsi_pci_info);
+#endif
#ifdef CONFIG_VHOST_VSOCK
type_register_static(&vhost_vsock_pci_info);
#endif
diff --git a/hw/virtio/virtio-pci.h b/hw/virtio/virtio-pci.h
index b4edea6..404b2c4 100644
--- a/hw/virtio/virtio-pci.h
+++ b/hw/virtio/virtio-pci.h
@@ -31,6 +31,9 @@
#ifdef CONFIG_VHOST_SCSI
#include "hw/virtio/vhost-scsi.h"
#endif
+#ifdef CONFIG_VHOST_USER_SCSI
+#include "hw/virtio/vhost-user-scsi.h"
+#endif
#ifdef CONFIG_VHOST_VSOCK
#include "hw/virtio/vhost-vsock.h"
#endif
@@ -42,6 +45,7 @@ typedef struct VirtIOBalloonPCI VirtIOBalloonPCI;
typedef struct VirtIOSerialPCI VirtIOSerialPCI;
typedef struct VirtIONetPCI VirtIONetPCI;
typedef struct VHostSCSIPCI VHostSCSIPCI;
+typedef struct VHostUserSCSIPCI VHostUserSCSIPCI;
typedef struct VirtIORngPCI VirtIORngPCI;
typedef struct VirtIOInputPCI VirtIOInputPCI;
typedef struct VirtIOInputHIDPCI VirtIOInputHIDPCI;
@@ -212,6 +216,17 @@ struct VHostSCSIPCI {
};
#endif
+#ifdef CONFIG_VHOST_USER_SCSI
+#define TYPE_VHOST_USER_SCSI_PCI "vhost-user-scsi-pci"
+#define VHOST_USER_SCSI_PCI(obj) \
+ OBJECT_CHECK(VHostUserSCSIPCI, (obj), TYPE_VHOST_USER_SCSI_PCI)
+
+struct VHostUserSCSIPCI {
+ VirtIOPCIProxy parent_obj;
+ VHostUserSCSI vdev;
+};
+#endif
+
/*
* virtio-blk-pci: This extends VirtioPCIProxy.
*/
diff --git a/include/hw/virtio/vhost-user-scsi.h b/include/hw/virtio/vhost-user-scsi.h
new file mode 100644
index 0000000..4a16181
--- /dev/null
+++ b/include/hw/virtio/vhost-user-scsi.h
@@ -0,0 +1,45 @@
+/*
+ * vhost-user-scsi host device
+ *
+ * Copyright (c) 2016 Nutanix Inc. All rights reserved.
+ *
+ * Author:
+ * Felipe Franciosi <felipe@nutanix.com>
+ *
+ * This file is largely based on "vhost-scsi.h" by:
+ * Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU LGPL, version 2 or later.
+ * See the COPYING.LIB file in the top-level directory.
+ *
+ */
+
+#ifndef VHOST_USER_SCSI_H
+#define VHOST_USER_SCSI_H
+
+#include "qemu-common.h"
+#include "hw/qdev.h"
+#include "hw/virtio/virtio-scsi.h"
+#include "hw/virtio/vhost.h"
+
+enum vhost_user_scsi_vq_list {
+ VHOST_USER_SCSI_VQ_CONTROL = 0,
+ VHOST_USER_SCSI_VQ_EVENT = 1,
+ VHOST_USER_SCSI_VQ_NUM_FIXED = 2,
+};
+
+#define TYPE_VHOST_USER_SCSI "vhost-user-scsi"
+#define VHOST_USER_SCSI(obj) \
+ OBJECT_CHECK(VHostUserSCSI, (obj), TYPE_VHOST_USER_SCSI)
+
+typedef struct VHostUserSCSI {
+ VirtIOSCSICommon parent_obj;
+
+ struct vhost_dev dev;
+ int32_t bootindex;
+ int channel;
+ int target;
+ int lun;
+} VHostUserSCSI;
+
+#endif /* VHOST_USER_SCSI_H */
diff --git a/include/hw/virtio/virtio-scsi.h b/include/hw/virtio/virtio-scsi.h
index a1e0cfb..4b97386 100644
--- a/include/hw/virtio/virtio-scsi.h
+++ b/include/hw/virtio/virtio-scsi.h
@@ -21,6 +21,7 @@
#include "hw/virtio/virtio.h"
#include "hw/pci/pci.h"
#include "hw/scsi/scsi.h"
+#include "sysemu/char.h"
#include "sysemu/iothread.h"
#define TYPE_VIRTIO_SCSI_COMMON "virtio-scsi-common"
@@ -53,6 +54,10 @@ struct VirtIOSCSIConf {
char *wwpn;
uint32_t boot_tpgt;
IOThread *iothread;
+#ifdef CONFIG_VHOST_USER_SCSI
+ CharBackend chardev;
+ int32_t bootindex;
+#endif
};
struct VirtIOSCSI;
--
1.9.4
^ permalink raw reply related [flat|nested] 7+ messages in thread
* [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application
2016-10-26 15:26 [Qemu-devel] [PATCH 0/2] Introduce vhost-user-scsi and sample application Felipe Franciosi
2016-10-26 15:26 ` [Qemu-devel] [PATCH 1/2] vus: Introduce vhost-user-scsi host device Felipe Franciosi
@ 2016-10-26 15:26 ` Felipe Franciosi
2016-10-27 12:16 ` Paolo Bonzini
1 sibling, 1 reply; 7+ messages in thread
From: Felipe Franciosi @ 2016-10-26 15:26 UTC (permalink / raw)
To: Paolo Bonzini, Stefan Hajnoczi, Marc-Andre Lureau, Michael S. Tsirkin
Cc: qemu-devel, Felipe Franciosi
This commit introduces a vhost-user-scsi backend sample application. It
must be linked with libiscsi and libvhost-user.
To use it, compile with:
make tests/vhost-user-scsi
And run as follows:
tests/vhost-user-scsi -u /tmp/vus.sock -i iscsi://uri_to_target/
The application is currently limited at one LUN only and it processes
requests synchronously (therefore only achieving QD1). The purpose of
the code is to show how a backend can be implemented and to test the
vhost-user-scsi Qemu implementation.
If a different instance of this vhost-user-scsi application is executed
at a remote host, a VM can be live migrated to such a host.
Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
---
tests/Makefile.include | 2 +
tests/vhost-user-scsi.c | 862 ++++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 864 insertions(+)
create mode 100644 tests/vhost-user-scsi.c
diff --git a/tests/Makefile.include b/tests/Makefile.include
index 7e6fd23..e61fe54 100644
--- a/tests/Makefile.include
+++ b/tests/Makefile.include
@@ -685,6 +685,8 @@ tests/test-filter-redirector$(EXESUF): tests/test-filter-redirector.o $(qtest-ob
tests/test-x86-cpuid-compat$(EXESUF): tests/test-x86-cpuid-compat.o $(qtest-obj-y)
tests/ivshmem-test$(EXESUF): tests/ivshmem-test.o contrib/ivshmem-server/ivshmem-server.o $(libqos-pc-obj-y)
tests/vhost-user-bridge$(EXESUF): tests/vhost-user-bridge.o contrib/libvhost-user/libvhost-user.o $(test-util-obj-y)
+tests/vhost-user-scsi.o-cflags := $(LIBISCSI_CFLAGS)
+tests/vhost-user-scsi$(EXESUF): tests/vhost-user-scsi.o contrib/libvhost-user/libvhost-user.o $(test-util-obj-y) $(test-block-obj-y)
tests/test-uuid$(EXESUF): tests/test-uuid.o $(test-util-obj-y)
tests/test-arm-mptimer$(EXESUF): tests/test-arm-mptimer.o
diff --git a/tests/vhost-user-scsi.c b/tests/vhost-user-scsi.c
new file mode 100644
index 0000000..c92b3b2
--- /dev/null
+++ b/tests/vhost-user-scsi.c
@@ -0,0 +1,862 @@
+/*
+ * vhost-user-scsi sample application
+ *
+ * Copyright (c) 2016 Nutanix Inc. All rights reserved.
+ *
+ * Author:
+ * Felipe Franciosi <felipe@nutanix.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 only.
+ * See the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "contrib/libvhost-user/libvhost-user.h"
+#include "hw/virtio/virtio-scsi.h"
+#include "iscsi/iscsi.h"
+
+#include <poll.h>
+
+#define VHOST_USER_SCSI_DEBUG 1
+
+/** Log helpers **/
+
+#define PPRE \
+ struct timespec ts; \
+ char timebuf[64]; \
+ struct tm tm; \
+ (void)clock_gettime(CLOCK_REALTIME, &ts); \
+ (void)strftime(timebuf, 64, "%Y%m%d %T", gmtime_r(&ts.tv_sec, &tm))
+
+#define PEXT(lvl, msg, ...) do { \
+ PPRE; \
+ fprintf(stderr, "%s.%06ld " lvl ": %s:%s():%d: " msg "\n", \
+ timebuf, ts.tv_nsec/1000, \
+ __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
+} while(0)
+
+#define PNOR(lvl, msg, ...) do { \
+ PPRE; \
+ fprintf(stderr, "%s.%06ld " lvl ": " msg "\n", \
+ timebuf, ts.tv_nsec/1000, ## __VA_ARGS__); \
+} while(0);
+
+#ifdef VHOST_USER_SCSI_DEBUG
+#define PDBG(msg, ...) PEXT("DBG", msg, ## __VA_ARGS__)
+#define PERR(msg, ...) PEXT("ERR", msg, ## __VA_ARGS__)
+#define PLOG(msg, ...) PEXT("LOG", msg, ## __VA_ARGS__)
+#else
+#define PDBG(msg, ...) { }
+#define PERR(msg, ...) PNOR("ERR", msg, ## __VA_ARGS__)
+#define PLOG(msg, ...) PNOR("LOG", msg, ## __VA_ARGS__)
+#endif
+
+/** vhost-user-scsi specific definitions **/
+
+/* TODO: MAX is defined at 8, should be 1024 */
+#define VUS_SCHED_MAX_FDS (1 + (2*VHOST_MAX_NR_VIRTQUEUE))
+
+#define VDEV_SCSI_MAX_LUNS 1 // Only 1 lun supported today
+#define VDEV_SCSI_MAX_DEVS 1 // Only 1 devices supported today
+
+#define ISCSI_INITIATOR "iqn.2016-10.com.nutanix:vhost-user-scsi"
+
+typedef void (*misc_cb) (short evt, void *pvt);
+
+typedef struct sched_data {
+ vu_watch_cb cb1;
+ misc_cb cb2;
+ void *pvt;
+ short evt;
+} sched_data_t;
+
+typedef struct sched {
+ VuDev *vu_dev;
+ nfds_t nfds;
+ struct pollfd fds[VUS_SCHED_MAX_FDS];
+ sched_data_t data[VUS_SCHED_MAX_FDS];
+ int quit;
+} sched_t;
+
+typedef struct iscsi_lun {
+ struct iscsi_context *iscsi_ctx;
+ int iscsi_lun;
+} iscsi_lun_t;
+
+typedef struct vhost_scsi_dev {
+ VuDev vu_dev;
+ int server_sock;
+ sched_t sched;
+ iscsi_lun_t luns[VDEV_SCSI_MAX_LUNS];
+} vhost_scsi_dev_t;
+
+static vhost_scsi_dev_t *vhost_scsi_devs[VDEV_SCSI_MAX_DEVS];
+
+static vhost_scsi_dev_t *vdev_scsi_find_by_vu(VuDev *vu_dev);
+
+/** poll-based scheduler for libvhost-user and misc callbacks **/
+
+static int sched_add(sched_t *sched, int fd, short int evt,
+ vu_watch_cb cb1, misc_cb cb2, void *pvt) {
+ int i;
+
+ assert(sched);
+ assert(fd >= 0);
+ assert(evt);
+ assert(cb1 || cb2);
+ assert(!(cb1 && cb2)); // only one of the cbs should be used
+
+ for (i=0; i<sched->nfds && i<VUS_SCHED_MAX_FDS; i++) {
+ if (sched->fds[i].fd == fd) {
+ break;
+ }
+ }
+ if (i == VUS_SCHED_MAX_FDS) {
+ PERR("Error adding fd: max number of fds reached");
+ return -1;
+ }
+
+ sched->fds[i].fd = fd;
+ sched->fds[i].events = evt;
+ sched->data[i].cb1 = cb1;
+ sched->data[i].cb2 = cb2;
+ sched->data[i].pvt = pvt;
+ sched->data[i].evt = evt;
+
+ if (sched->nfds <= i) {
+ sched->nfds = i+1;
+ }
+
+ PDBG("sched@%p: add fd %d to slot %i", sched, fd, i);
+
+ return 0;
+}
+
+static int sched_del(sched_t *sched, int fd) {
+ int i;
+
+ assert(sched);
+ assert(fd >= 0);
+
+ for (i=0; i<sched->nfds; i++) {
+ if (sched->fds[i].fd == fd) {
+ break;
+ }
+ }
+ if (sched->nfds == i) {
+#ifdef VUS_PEDANTIC_SCHEDULER
+ PERR("Error deleting fd %d: fd not found", fd);
+ return -1;
+#else
+ return 0;
+#endif
+ }
+
+ sched->nfds--;
+ if (sched->nfds > 0) {
+ // Overwrite deleted entry with last entry from scheduler
+ memcpy(&sched->fds[i], &sched->fds[sched->nfds],
+ sizeof(struct pollfd));
+ memcpy(&sched->data[i], &sched->data[sched->nfds],
+ sizeof(sched_data_t));
+ }
+ memset(&sched->fds[sched->nfds], 0, sizeof(struct pollfd));
+ memset(&sched->data[sched->nfds], 0, sizeof(sched_data_t));
+
+ PDBG("sched@%p: del fd %d from slot %i", sched, fd, i);
+
+ return 0;
+}
+
+static int sched_loop(sched_t *sched) {
+ int i, n;
+
+ assert(sched);
+ assert(sched->nfds > 0);
+
+ while (!sched->quit) {
+ n = poll(sched->fds, sched->nfds, -1);
+ if (n < 0) {
+ PERR("Error polling: %s", strerror(errno));
+ return -1;
+ }
+
+ for (i=0; i<sched->nfds && n; i++) {
+ if (sched->fds[i].revents != 0) {
+
+ if (sched->data[i].cb1) {
+ int vu_evt = 0;
+
+ if (sched->fds[i].revents & POLLIN) vu_evt |= VU_WATCH_IN;
+ if (sched->fds[i].revents & POLLOUT) vu_evt |= VU_WATCH_OUT;
+ if (sched->fds[i].revents & POLLPRI) vu_evt |= VU_WATCH_PRI;
+ if (sched->fds[i].revents & POLLERR) vu_evt |= VU_WATCH_ERR;
+ if (sched->fds[i].revents & POLLHUP) vu_evt |= VU_WATCH_HUP;
+
+ PDBG("sched@%p: fd[%d] (%d): cb1(%p, %d, %p)", sched, i,
+ sched->fds[i].fd, sched->vu_dev, vu_evt,
+ sched->data[i].pvt);
+
+ sched->data[i].cb1(sched->vu_dev, vu_evt,
+ sched->data[i].pvt);
+ } else {
+ PDBG("sched@%p: fd[%d] (%d): cbb(%hd, %p)", sched, i,
+ sched->fds[i].fd, sched->fds[i].revents,
+ sched->data[i].pvt);
+
+ sched->data[i].cb2(sched->fds[i].revents,
+ sched->data[i].pvt);
+ }
+
+ n--;
+ }
+ }
+ }
+
+ return 0;
+}
+
+/** from libiscsi's scsi-lowlevel.h **/
+
+#define SCSI_CDB_MAX_SIZE 16
+
+struct scsi_iovector {
+ struct scsi_iovec *iov;
+ int niov;
+ int nalloc;
+ size_t offset;
+ int consumed;
+};
+
+struct scsi_allocated_memory {
+ struct scsi_allocated_memory *next;
+ char buf[0];
+};
+
+struct scsi_data {
+ int size;
+ unsigned char *data;
+};
+
+enum scsi_sense_key {
+ SCSI_SENSE_NO_SENSE = 0x00,
+ SCSI_SENSE_RECOVERED_ERROR = 0x01,
+ SCSI_SENSE_NOT_READY = 0x02,
+ SCSI_SENSE_MEDIUM_ERROR = 0x03,
+ SCSI_SENSE_HARDWARE_ERROR = 0x04,
+ SCSI_SENSE_ILLEGAL_REQUEST = 0x05,
+ SCSI_SENSE_UNIT_ATTENTION = 0x06,
+ SCSI_SENSE_DATA_PROTECTION = 0x07,
+ SCSI_SENSE_BLANK_CHECK = 0x08,
+ SCSI_SENSE_VENDOR_SPECIFIC = 0x09,
+ SCSI_SENSE_COPY_ABORTED = 0x0a,
+ SCSI_SENSE_COMMAND_ABORTED = 0x0b,
+ SCSI_SENSE_OBSOLETE_ERROR_CODE = 0x0c,
+ SCSI_SENSE_OVERFLOW_COMMAND = 0x0d,
+ SCSI_SENSE_MISCOMPARE = 0x0e
+};
+
+struct scsi_sense {
+ unsigned char error_type;
+ enum scsi_sense_key key;
+ int ascq;
+ unsigned sense_specific:1;
+ unsigned ill_param_in_cdb:1;
+ unsigned bit_pointer_valid:1;
+ unsigned char bit_pointer;
+ uint16_t field_pointer;
+};
+
+enum scsi_residual {
+ SCSI_RESIDUAL_NO_RESIDUAL = 0,
+ SCSI_RESIDUAL_UNDERFLOW,
+ SCSI_RESIDUAL_OVERFLOW
+};
+
+struct scsi_task {
+ int status;
+ int cdb_size;
+ int xfer_dir;
+ int expxferlen;
+ unsigned char cdb[SCSI_CDB_MAX_SIZE];
+ enum scsi_residual residual_status;
+ size_t residual;
+ struct scsi_sense sense;
+ struct scsi_data datain;
+ struct scsi_allocated_memory *mem;
+ void *ptr;
+
+ uint32_t itt;
+ uint32_t cmdsn;
+ uint32_t lun;
+
+ struct scsi_iovector iovector_in;
+ struct scsi_iovector iovector_out;
+};
+
+/** libiscsi integration **/
+
+static int iscsi_add_lun(iscsi_lun_t *lun, char *iscsi_uri) {
+ struct iscsi_url *iscsi_url;
+ struct iscsi_context *iscsi_ctx;
+ int ret = 0;
+
+ assert(lun);
+
+ iscsi_ctx = iscsi_create_context(ISCSI_INITIATOR);
+ if (!iscsi_ctx) {
+ PERR("Unable to create iSCSI context");
+ return -1;
+ }
+
+ iscsi_url = iscsi_parse_full_url(iscsi_ctx, iscsi_uri);
+ if (!iscsi_url) {
+ PERR("Unable to parse iSCSI URL: %s", iscsi_get_error(iscsi_ctx));
+ goto fail;
+ }
+
+ iscsi_set_session_type(iscsi_ctx, ISCSI_SESSION_NORMAL);
+ iscsi_set_header_digest(iscsi_ctx, ISCSI_HEADER_DIGEST_NONE_CRC32C);
+ if (iscsi_full_connect_sync(iscsi_ctx, iscsi_url->portal, iscsi_url->lun)) {
+ PERR("Unable to login to iSCSI portal: %s", iscsi_get_error(iscsi_ctx));
+ goto fail;
+ }
+
+ lun->iscsi_ctx = iscsi_ctx;
+ lun->iscsi_lun = iscsi_url->lun;
+
+ PDBG("Context %p created for lun 0: %s", iscsi_ctx, iscsi_uri);
+
+out:
+ if (iscsi_url) {
+ iscsi_destroy_url(iscsi_url);
+ }
+ return ret;
+
+fail:
+ (void)iscsi_destroy_context(iscsi_ctx);
+ ret = -1;
+ goto out;
+}
+
+static struct scsi_task *scsi_task_new(int cdb_len, uint8_t *cdb, int dir,
+ int xfer_len) {
+ struct scsi_task *task;
+
+ assert(cdb_len > 0);
+ assert(cdb);
+
+ task = calloc(1, sizeof(struct scsi_task));
+ if (!task) {
+ PERR("Error allocating task: %s", strerror(errno));
+ return NULL;
+ }
+
+ memcpy(task->cdb, cdb, cdb_len);
+ task->cdb_size = cdb_len;
+ task->xfer_dir = dir;
+ task->expxferlen = xfer_len;
+
+ return task;
+}
+
+static int get_cdb_len(uint8_t *cdb) {
+ switch(cdb[0] >> 5){
+ case 0:
+ return 6;
+ case 1:
+ case 2:
+ return 10;
+ case 4:
+ return 16;
+ case 5:
+ return 12;
+ }
+ PERR("Unable to determine cdb len (0x%02hhX)", cdb[0]>>5);
+ return -1;
+}
+
+static int handle_cmd_sync(struct iscsi_context *ctx,
+ VirtIOSCSICmdReq *req,
+ struct iovec *out, unsigned int out_len,
+ VirtIOSCSICmdResp *rsp,
+ struct iovec *in, unsigned int in_len) {
+ struct scsi_task *task;
+ uint32_t dir;
+ uint32_t len;
+ int cdb_len;
+ int i;
+
+ if (!((!req->lun[1]) && (req->lun[2] == 0x40) && (!req->lun[3]))) {
+ // Ignore anything different than target=0, lun=0
+ PDBG("Ignoring unconnected lun (0x%hhX, 0x%hhX)",
+ req->lun[1], req->lun[3]);
+ rsp->status = SCSI_STATUS_CHECK_CONDITION;
+ memset(rsp->sense, 0, sizeof(rsp->sense));
+ rsp->sense_len = 18;
+ rsp->sense[0] = 0x70;
+ rsp->sense[2] = 0x05; // ILLEGAL_REQUEST
+ rsp->sense[7] = 10;
+ rsp->sense[12] = 0x24;
+
+ return 0;
+ }
+
+ cdb_len = get_cdb_len(req->cdb);
+ if (cdb_len == -1) {
+ return -1;
+ }
+
+ len = 0;
+ if (!out_len && !in_len) {
+ dir = SCSI_XFER_NONE;
+ } else if (out_len) {
+ dir = SCSI_XFER_TO_DEV;
+ for (i=0; i<out_len; i++) {
+ len += out[i].iov_len;
+ }
+ } else {
+ dir = SCSI_XFER_FROM_DEV;
+ for (i=0; i<in_len; i++) {
+ len += in[i].iov_len;
+ }
+ }
+
+ task = scsi_task_new(cdb_len, req->cdb, dir, len);
+ if (!task) {
+ PERR("Unable to create iscsi task");
+ return -1;
+ }
+
+ if (dir == SCSI_XFER_TO_DEV) {
+ task->iovector_out.iov = (struct scsi_iovec *)out;
+ task->iovector_out.niov = out_len;
+ } else if (dir == SCSI_XFER_FROM_DEV) {
+ task->iovector_in.iov = (struct scsi_iovec *)in;
+ task->iovector_in.niov = in_len;
+ }
+
+ PDBG("Sending iscsi cmd (cdb_len=%d, dir=%d, task=%p)",
+ cdb_len, dir, task);
+ if (!iscsi_scsi_command_sync(ctx, 0, task, NULL)) {
+ PERR("Error serving SCSI command");
+ free(task);
+ return -1;
+ }
+
+ memset(rsp, 0, sizeof(*rsp));
+
+ rsp->status = task->status;
+ rsp->resid = task->residual;
+
+ if (task->status == SCSI_STATUS_CHECK_CONDITION) {
+ rsp->response = VIRTIO_SCSI_S_FAILURE;
+ rsp->sense_len = task->datain.size - 2;
+ memcpy(rsp->sense, &task->datain.data[2], rsp->sense_len);
+ }
+
+ free(task);
+
+ PDBG("Filled in rsp: status=%hhX, resid=%u, response=%hhX, sense_len=%u",
+ rsp->status, rsp->resid, rsp->response, rsp->sense_len);
+
+ return 0;
+}
+
+/** libvhost-user callbacks **/
+
+static void vus_panic_cb(VuDev *vu_dev, const char *buf) {
+ vhost_scsi_dev_t *vdev_scsi;
+
+ assert(vu_dev);
+
+ vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
+
+ if (buf) {
+ PERR("vu_panic: %s", buf);
+ }
+
+ if (vdev_scsi) {
+ vdev_scsi->sched.quit = 1;
+ }
+}
+
+static void vus_add_watch_cb(VuDev *vu_dev, int fd, int vu_evt, vu_watch_cb cb,
+ void *pvt) {
+ vhost_scsi_dev_t *vdev_scsi;
+ int poll_evt = 0;
+
+ assert(vu_dev);
+ assert(fd >= 0);
+ assert(cb);
+
+ vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
+ if (!vdev_scsi) {
+ vus_panic_cb(vu_dev, NULL);
+ }
+
+ /* TODO: VU_WATCH_* should match POLL*, check it */
+ if (vu_evt & VU_WATCH_IN) poll_evt |= POLLIN;
+ if (vu_evt & VU_WATCH_OUT) poll_evt |= POLLOUT;
+ if (vu_evt & VU_WATCH_PRI) poll_evt |= POLLPRI;
+ if (vu_evt & VU_WATCH_ERR) poll_evt |= POLLERR;
+ if (vu_evt & VU_WATCH_HUP) poll_evt |= POLLHUP;
+
+ if (sched_add(&vdev_scsi->sched, fd, poll_evt, cb, NULL, pvt)) {
+ vus_panic_cb(vu_dev, NULL);
+ }
+}
+
+static void vus_del_watch_cb(VuDev *vu_dev, int fd) {
+ vhost_scsi_dev_t *vdev_scsi;
+
+ assert(vu_dev);
+ assert(fd >= 0);
+
+ vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
+ if (!vdev_scsi) {
+ vus_panic_cb(vu_dev, NULL);
+ return;
+ }
+
+ if (sched_del(&vdev_scsi->sched, fd)) {
+ vus_panic_cb(vu_dev, NULL);
+ }
+}
+
+static void vus_proc_ctl(VuDev *vu_dev, int idx) {
+ /* Control VQ not implemented */
+}
+
+static void vus_proc_evt(VuDev *vu_dev, int idx) {
+ /* Event VQ not implemented */
+}
+
+static void vus_proc_req(VuDev *vu_dev, int idx) {
+ vhost_scsi_dev_t *vdev_scsi;
+ VuVirtq *vq;
+
+ assert(vu_dev);
+
+ vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
+ if (!vdev_scsi) {
+ vus_panic_cb(vu_dev, NULL);
+ return;
+ }
+
+ if ((idx < 0) || (idx >= VHOST_MAX_NR_VIRTQUEUE)) {
+ PERR("VQ Index out of range: %d", idx);
+ vus_panic_cb(vu_dev, NULL);
+ return;
+ }
+
+ vq = vu_get_queue(vu_dev, idx);
+ if (!vq) {
+ PERR("Error fetching VQ (dev=%p, idx=%d)", vu_dev, idx);
+ vus_panic_cb(vu_dev, NULL);
+ return;
+ }
+
+ PDBG("Got kicked on vq[%d]@%p", idx, vq);
+
+ while(1) {
+ VuVirtqElement *elem;
+ VirtIOSCSICmdReq *req;
+ VirtIOSCSICmdResp *rsp;
+
+ elem = vu_queue_pop(vu_dev, vq, sizeof(VuVirtqElement));
+ if (!elem) {
+ PDBG("No more elements pending on vq[%d]@%p", idx, vq);
+ break;
+ }
+ PDBG("Popped elem@%p", elem);
+
+ assert(!((elem->out_num > 1) && (elem->in_num > 1)));
+ assert((elem->out_num > 0) && (elem->in_num > 0));
+
+ if (elem->out_sg[0].iov_len < sizeof(VirtIOSCSICmdReq)) {
+ PERR("Invalid virtio-scsi req header");
+ vus_panic_cb(vu_dev, NULL);
+ break;
+ }
+ req = (VirtIOSCSICmdReq *)elem->out_sg[0].iov_base;
+
+ if (elem->in_sg[0].iov_len < sizeof(VirtIOSCSICmdResp)) {
+ PERR("Invalid virtio-scsi rsp header");
+ vus_panic_cb(vu_dev, NULL);
+ break;
+ }
+ rsp = (VirtIOSCSICmdResp *)elem->in_sg[0].iov_base;
+
+ if (handle_cmd_sync(vdev_scsi->luns[0].iscsi_ctx,
+ req, &elem->out_sg[1], elem->out_num-1,
+ rsp, &elem->in_sg[1], elem->in_num-1) != 0) {
+ vus_panic_cb(vu_dev, NULL);
+ break;
+ }
+
+ vu_queue_push(vu_dev, vq, elem, 0);
+ vu_queue_notify(vu_dev, vq);
+
+ free(elem);
+ }
+
+}
+
+static void vus_queue_set_started(VuDev *vu_dev, int idx, bool started) {
+ VuVirtq *vq;
+
+ assert(vu_dev);
+
+ if ((idx < 0) || (idx >= VHOST_MAX_NR_VIRTQUEUE)) {
+ PERR("VQ Index out of range: %d", idx);
+ vus_panic_cb(vu_dev, NULL);
+ return;
+ }
+
+ vq = vu_get_queue(vu_dev, idx);
+
+ switch(idx) {
+ case 0:
+ vu_set_queue_handler(vu_dev, vq, started?vus_proc_ctl:NULL);
+ break;
+ case 1:
+ vu_set_queue_handler(vu_dev, vq, started?vus_proc_evt:NULL);
+ break;
+ default:
+ vu_set_queue_handler(vu_dev, vq, started?vus_proc_req:NULL);
+ }
+}
+
+static const VuDevIface vus_iface = {
+ .queue_set_started = vus_queue_set_started,
+};
+
+static void vus_vhost_cb(VuDev *vu_dev, int vu_evt, void *data) {
+ assert(vu_dev);
+
+ if (!vu_dispatch(vu_dev) != 0) {
+ PERR("Error processing vhost message");
+ vus_panic_cb(vu_dev, NULL);
+ }
+}
+
+/** util **/
+
+static int unix_sock_new(char *unix_fn) {
+ int sock;
+ struct sockaddr_un un;
+ size_t len;
+
+ assert(unix_fn);
+
+ sock = socket(AF_UNIX, SOCK_STREAM, 0);
+ if (sock <= 0) {
+ perror("socket");
+ return -1;
+ }
+
+ un.sun_family = AF_UNIX;
+ (void)snprintf(un.sun_path, sizeof(un.sun_path), "%s", unix_fn);
+ len = sizeof(un.sun_family) + strlen(un.sun_path);
+
+ (void)unlink(unix_fn);
+ if (bind(sock, (struct sockaddr *)&un, len) < 0) {
+ perror("bind");
+ goto fail;
+ }
+
+ if (listen(sock, 1) < 0) {
+ perror("listen");
+ goto fail;
+ }
+
+ return sock;
+
+fail:
+ (void)close(sock);
+
+ return -1;
+}
+
+/** vhost-user-scsi **/
+
+static vhost_scsi_dev_t *vdev_scsi_find_by_vu(VuDev *vu_dev) {
+ int i;
+
+ assert(vu_dev);
+
+ for (i=0; i<VDEV_SCSI_MAX_DEVS; i++) {
+ if (&vhost_scsi_devs[i]->vu_dev == vu_dev) {
+ return vhost_scsi_devs[i];
+ }
+ }
+
+ PERR("Unknown VuDev %p", vu_dev);
+ return NULL;
+}
+
+static void vdev_scsi_deinit(vhost_scsi_dev_t *vdev_scsi) {
+ if (!vdev_scsi) {
+ return;
+ }
+
+ if (vdev_scsi->server_sock >= 0) {
+ struct sockaddr_storage ss;
+ socklen_t sslen = sizeof(ss);
+
+ if (getsockname(vdev_scsi->server_sock, (struct sockaddr *)&ss,
+ &sslen) == 0) {
+ struct sockaddr_un *su = (struct sockaddr_un *)&ss;
+ (void)unlink(su->sun_path);
+ }
+
+ (void)close(vdev_scsi->server_sock);
+ }
+}
+
+static vhost_scsi_dev_t *vdev_scsi_new(char *unix_fn) {
+ vhost_scsi_dev_t *vdev_scsi;
+
+ assert(unix_fn);
+
+ vdev_scsi = calloc(1, sizeof(vhost_scsi_dev_t));
+ if (!vdev_scsi) {
+ perror("calloc");
+ return NULL;
+ }
+
+ vdev_scsi->server_sock = unix_sock_new(unix_fn);
+ if (vdev_scsi->server_sock < 0) {
+ free(vdev_scsi);
+ return NULL;
+ }
+
+ vdev_scsi->sched.vu_dev = &vdev_scsi->vu_dev;
+
+ return vdev_scsi;
+}
+
+static int vdev_scsi_iscsi_add_lun(vhost_scsi_dev_t *vdev_scsi,
+ char *iscsi_uri, uint32_t lun) {
+ assert(vdev_scsi);
+ assert(iscsi_uri);
+ assert(lun < VDEV_SCSI_MAX_LUNS);
+
+ if (vdev_scsi->luns[lun].iscsi_ctx) {
+ PERR("Lun %d already configured", lun);
+ return -1;
+ }
+
+ if (iscsi_add_lun(&vdev_scsi->luns[lun], iscsi_uri) != 0) {
+ return -1;
+ }
+
+ return 0;
+}
+
+static int vdev_scsi_run(vhost_scsi_dev_t *vdev_scsi) {
+ int cli_sock;
+ int ret = 0;
+
+ assert(vdev_scsi);
+ assert(vdev_scsi->server_sock >= 0);
+
+ cli_sock = accept(vdev_scsi->server_sock, (void *)0, (void *)0);
+ if (cli_sock < 0) {
+ perror("accept");
+ return -1;
+ }
+
+ vu_init(&vdev_scsi->vu_dev,
+ cli_sock,
+ vus_panic_cb,
+ vus_add_watch_cb,
+ vus_del_watch_cb,
+ &vus_iface);
+
+ ret = sched_add(&vdev_scsi->sched, cli_sock, POLLIN, vus_vhost_cb, NULL, 0);
+ if (ret) {
+ goto fail;
+ }
+
+ if (sched_loop(&vdev_scsi->sched) != 0) {
+ goto fail;
+ }
+
+out:
+ vu_deinit(&vdev_scsi->vu_dev);
+
+ return ret;
+
+fail:
+ ret = -1;
+ goto out;
+}
+
+int main(int argc, char **argv)
+{
+ vhost_scsi_dev_t *vdev_scsi = NULL;
+ char *unix_fn = NULL;
+ char *iscsi_uri = NULL;
+ int opt, err = EXIT_SUCCESS;
+
+ while ((opt = getopt(argc, argv, "u:i:")) != -1) {
+ switch (opt) {
+ case 'h':
+ goto help;
+ case 'u':
+ unix_fn = strdup(optarg);
+ break;
+ case 'i':
+ iscsi_uri = strdup(optarg);
+ break;
+ default:
+ goto help;
+ }
+ }
+ if (!unix_fn || !iscsi_uri) {
+ goto help;
+ }
+
+ vdev_scsi = vdev_scsi_new(unix_fn);
+ if (!vdev_scsi) {
+ goto err;
+ }
+ vhost_scsi_devs[0] = vdev_scsi;
+
+ if (vdev_scsi_iscsi_add_lun(vdev_scsi, iscsi_uri, 0) != 0) {
+ goto err;
+ }
+
+ if (vdev_scsi_run(vdev_scsi) != 0) {
+ goto err;
+ }
+
+out:
+ if (vdev_scsi) {
+ vdev_scsi_deinit(vdev_scsi);
+ free(vdev_scsi);
+ }
+ if (unix_fn) {
+ free(unix_fn);
+ }
+ if (iscsi_uri) {
+ free(iscsi_uri);
+ }
+
+ return err;
+
+err:
+ err = EXIT_FAILURE;
+ goto out;
+
+help:
+ fprintf(stderr, "Usage: %s [ -u unix_sock_path -i iscsi_uri ] | [ -h ]\n",
+ argv[0]);
+ fprintf(stderr, " -u path to unix socket\n");
+ fprintf(stderr, " -i iscsi uri for lun 0\n");
+ fprintf(stderr, " -h print help and quit\n");
+
+ goto err;
+}
--
1.9.4
^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [Qemu-devel] [PATCH 1/2] vus: Introduce vhost-user-scsi host device
2016-10-26 15:26 ` [Qemu-devel] [PATCH 1/2] vus: Introduce vhost-user-scsi host device Felipe Franciosi
@ 2016-10-27 12:12 ` Paolo Bonzini
0 siblings, 0 replies; 7+ messages in thread
From: Paolo Bonzini @ 2016-10-27 12:12 UTC (permalink / raw)
To: Felipe Franciosi, Stefan Hajnoczi, Marc-Andre Lureau, Michael S. Tsirkin
Cc: qemu-devel
On 26/10/2016 17:26, Felipe Franciosi wrote:
> This commit introduces a vhost-user device for SCSI. This is based
> on the existing vhost-scsi implementation, but done over vhost-user
> instead. It also uses a chardev to connect to the backend. Unlike
> vhost-scsi (today), VMs using vhost-user-scsi can be live migrated.
>
> To use it, one must configure Qemu with --enable-vhost-user-scsi and
> start Qemu with a command line equivalent to:
>
> qemu-system-x86_64 \
> -chardev socket,id=vus0,path=/tmp/vus.sock \
> -device vhost-user-scsi-pci,chardev=vus0,bus=pci.0,addr=0x4
>
> A separate commit presents a sample application linked with libiscsi to
> provide a backend for vhost-user-scsi.
Hi,
most of the code you copy can be kept in one place only, by using the
vhost_ops struct. Please make a hierarchy
virtio-scsi-common
vhost-scsi-common
vhost-scsi (adds vhostfd+wwpn)
vhost-user-scsi (adds chardev)
with abstract methods in vhost-scsi-common to abstract e.g. opening
/dev/scsi and setting/clearing the endpoint, which are only needed by
vhost-scsi's realize method. This should avoid the duplication.
Thanks!
Paolo
> Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
> ---
> configure | 10 ++
> hw/scsi/Makefile.objs | 1 +
> hw/scsi/vhost-user-scsi.c | 299 ++++++++++++++++++++++++++++++++++++
> hw/virtio/virtio-pci.c | 58 +++++++
> hw/virtio/virtio-pci.h | 15 ++
> include/hw/virtio/vhost-user-scsi.h | 45 ++++++
> include/hw/virtio/virtio-scsi.h | 5 +
> 7 files changed, 433 insertions(+)
> create mode 100644 hw/scsi/vhost-user-scsi.c
> create mode 100644 include/hw/virtio/vhost-user-scsi.h
>
> diff --git a/configure b/configure
> index d3dafcb..0574ff2 100755
> --- a/configure
> +++ b/configure
> @@ -228,6 +228,7 @@ xfs=""
>
> vhost_net="no"
> vhost_scsi="no"
> +vhost_user_scsi="no"
> vhost_vsock="no"
> kvm="no"
> rdma=""
> @@ -677,6 +678,7 @@ Haiku)
> kvm="yes"
> vhost_net="yes"
> vhost_scsi="yes"
> + vhost_user_scsi="yes"
> vhost_vsock="yes"
> QEMU_INCLUDES="-I\$(SRC_PATH)/linux-headers -I$(pwd)/linux-headers $QEMU_INCLUDES"
> ;;
> @@ -1019,6 +1021,10 @@ for opt do
> ;;
> --enable-vhost-scsi) vhost_scsi="yes"
> ;;
> + --disable-vhost-user-scsi) vhost_user_scsi="no"
> + ;;
> + --enable-vhost-user-scsi) vhost_user_scsi="yes"
> + ;;
> --disable-vhost-vsock) vhost_vsock="no"
> ;;
> --enable-vhost-vsock) vhost_vsock="yes"
> @@ -4951,6 +4957,7 @@ echo "posix_madvise $posix_madvise"
> echo "libcap-ng support $cap_ng"
> echo "vhost-net support $vhost_net"
> echo "vhost-scsi support $vhost_scsi"
> +echo "vhost-user-scsi support $vhost_user_scsi"
> echo "vhost-vsock support $vhost_vsock"
> echo "Trace backends $trace_backends"
> if have_backend "simple"; then
> @@ -5336,6 +5343,9 @@ fi
> if test "$vhost_scsi" = "yes" ; then
> echo "CONFIG_VHOST_SCSI=y" >> $config_host_mak
> fi
> +if test "$vhost_user_scsi" = "yes" ; then
> + echo "CONFIG_VHOST_USER_SCSI=y" >> $config_host_mak
> +fi
> if test "$vhost_net" = "yes" ; then
> echo "CONFIG_VHOST_NET_USED=y" >> $config_host_mak
> fi
> diff --git a/hw/scsi/Makefile.objs b/hw/scsi/Makefile.objs
> index 5a2248b..3338aad 100644
> --- a/hw/scsi/Makefile.objs
> +++ b/hw/scsi/Makefile.objs
> @@ -11,4 +11,5 @@ obj-$(CONFIG_PSERIES) += spapr_vscsi.o
> ifeq ($(CONFIG_VIRTIO),y)
> obj-y += virtio-scsi.o virtio-scsi-dataplane.o
> obj-$(CONFIG_VHOST_SCSI) += vhost-scsi.o
> +obj-$(CONFIG_VHOST_USER_SCSI) += vhost-user-scsi.o
> endif
> diff --git a/hw/scsi/vhost-user-scsi.c b/hw/scsi/vhost-user-scsi.c
> new file mode 100644
> index 0000000..a7b9839
> --- /dev/null
> +++ b/hw/scsi/vhost-user-scsi.c
> @@ -0,0 +1,299 @@
> +/*
> + * vhost-user-scsi host device
> + *
> + * Copyright (c) 2016 Nutanix Inc. All rights reserved.
> + *
> + * Author:
> + * Felipe Franciosi <felipe@nutanix.com>
> + *
> + * This work is largely based on the "vhost-scsi" implementation by:
> + * Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
> + * Nicholas Bellinger <nab@risingtidesystems.com>
> + *
> + * This work is licensed under the terms of the GNU LGPL, version 2 or later.
> + * See the COPYING.LIB file in the top-level directory.
> + *
> + */
> +
> +#include "qemu/osdep.h"
> +#include "migration/vmstate.h"
> +#include "qapi/error.h"
> +#include "qemu/error-report.h"
> +#include "qemu/typedefs.h"
> +#include "qom/object.h"
> +#include "hw/fw-path-provider.h"
> +#include "hw/qdev-core.h"
> +#include "hw/virtio/vhost.h"
> +#include "hw/virtio/vhost-backend.h"
> +#include "hw/virtio/vhost-user-scsi.h"
> +#include "hw/virtio/virtio.h"
> +#include "hw/virtio/virtio-access.h"
> +#include "hw/virtio/virtio-bus.h"
> +#include "sysemu/char.h"
> +
> +/* Features supported by the host application */
> +static const int user_feature_bits[] = {
> + VIRTIO_F_NOTIFY_ON_EMPTY,
> + VIRTIO_RING_F_INDIRECT_DESC,
> + VIRTIO_RING_F_EVENT_IDX,
> + VIRTIO_SCSI_F_HOTPLUG,
> + VHOST_INVALID_FEATURE_BIT
> +};
> +
> +static int vhost_user_scsi_start(VHostUserSCSI *s)
> +{
> + int ret, i;
> + VirtIODevice *vdev = VIRTIO_DEVICE(s);
> + BusState *qbus = BUS(qdev_get_parent_bus(DEVICE(vdev)));
> + VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus);
> +
> + if (!k->set_guest_notifiers) {
> + error_report("binding does not support guest notifiers");
> + return -ENOSYS;
> + }
> +
> + ret = vhost_dev_enable_notifiers(&s->dev, vdev);
> + if (ret < 0) {
> + return ret;
> + }
> +
> + s->dev.acked_features = vdev->guest_features;
> + ret = vhost_dev_start(&s->dev, vdev);
> + if (ret < 0) {
> + error_report("Error starting vhost-user device");
> + goto err_notifiers;
> + }
> +
> + ret = k->set_guest_notifiers(qbus->parent, s->dev.nvqs, true);
> + if (ret < 0) {
> + error_report("Error binding guest notifier");
> + goto err_vhost_stop;
> + }
> +
> + /* guest_notifier_mask/pending not used yet, so just unmask
> + * everything here. virtio-pci will do the right thing by
> + * enabling/disabling irqfd.
> + */
> + for (i = 0; i < s->dev.nvqs; i++) {
> + vhost_virtqueue_mask(&s->dev, vdev, s->dev.vq_index + i, false);
> + }
> +
> + return ret;
> +
> +err_vhost_stop:
> + vhost_dev_stop(&s->dev, vdev);
> +err_notifiers:
> + vhost_dev_disable_notifiers(&s->dev, vdev);
> + return ret;
> +}
> +
> +static void vhost_user_scsi_stop(VHostUserSCSI *s)
> +{
> + VirtIODevice *vdev = VIRTIO_DEVICE(s);
> + BusState *qbus = BUS(qdev_get_parent_bus(DEVICE(vdev)));
> + VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus);
> + int ret = 0;
> +
> + vhost_dev_stop(&s->dev, vdev);
> + vhost_dev_disable_notifiers(&s->dev, vdev);
> +
> + if (k->set_guest_notifiers) {
> + ret = k->set_guest_notifiers(qbus->parent, s->dev.nvqs, false);
> + if (ret < 0) {
> + error_report("vhost guest notifier cleanup failed: %d", ret);
> + }
> + }
> + assert(ret >= 0);
> +}
> +
> +static uint64_t vhost_user_scsi_get_features(VirtIODevice *vdev,
> + uint64_t features,
> + Error **errp)
> +{
> + VHostUserSCSI *s = VHOST_USER_SCSI(vdev);
> +
> + return vhost_get_features(&s->dev, user_feature_bits, features);
> +}
> +
> +static void vhost_user_scsi_set_config(VirtIODevice *vdev,
> + const uint8_t *config)
> +{
> + VirtIOSCSIConfig *scsiconf = (VirtIOSCSIConfig *)config;
> + VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(vdev);
> +
> + if ((uint32_t)virtio_ldl_p(vdev, &scsiconf->sense_size) != vs->sense_size ||
> + (uint32_t)virtio_ldl_p(vdev, &scsiconf->cdb_size) != vs->cdb_size) {
> + error_report("vhost-user-scsi doesn't allow sense or CDB sizes change");
> + exit(1);
> + }
> +}
> +
> +static void vhost_user_scsi_set_status(VirtIODevice *vdev, uint8_t status)
> +{
> + VHostUserSCSI *s = (VHostUserSCSI *)vdev;
> + bool start = (status & VIRTIO_CONFIG_S_DRIVER_OK) && vdev->vm_running;
> +
> + if (s->dev.started == start) {
> + return;
> + }
> +
> + if (start) {
> + int ret;
> +
> + ret = vhost_user_scsi_start(s);
> + if (ret < 0) {
> + error_report("unable to start vhost-user-scsi: %s", strerror(-ret));
> + exit(1);
> + }
> + } else {
> + vhost_user_scsi_stop(s);
> + }
> +}
> +
> +static void vhost_dummy_handle_output(VirtIODevice *vdev, VirtQueue *vq)
> +{
> +}
> +
> +static void vhost_user_scsi_save(QEMUFile *f, void *opaque)
> +{
> + VirtIODevice *vdev = VIRTIO_DEVICE(opaque);
> + virtio_save(vdev, f);
> +}
> +
> +static int vhost_user_scsi_load(QEMUFile *f, void *opaque, int version_id)
> +{
> + VirtIODevice *vdev = VIRTIO_DEVICE(opaque);
> + return virtio_load(vdev, f, version_id);
> +}
> +
> +static void vhost_user_scsi_realize(DeviceState *dev, Error **errp)
> +{
> + VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(dev);
> + VHostUserSCSI *s = VHOST_USER_SCSI(dev);
> + static int vhost_user_scsi_id;
> + Error *err = NULL;
> + int ret;
> +
> + if (!vs->conf.chardev.chr) {
> + error_setg(errp, "vhost-user-scsi: missing chardev");
> + return;
> + }
> +
> + virtio_scsi_common_realize(dev, &err, vhost_dummy_handle_output,
> + vhost_dummy_handle_output,
> + vhost_dummy_handle_output);
> + if (err != NULL) {
> + error_propagate(errp, err);
> + return;
> + }
> +
> + s->dev.nvqs = VHOST_USER_SCSI_VQ_NUM_FIXED + vs->conf.num_queues;
> + s->dev.vqs = g_new(struct vhost_virtqueue, s->dev.nvqs);
> + s->dev.vq_index = 0;
> + s->dev.backend_features = 0;
> +
> + ret = vhost_dev_init(&s->dev, (void *)&vs->conf.chardev,
> + VHOST_BACKEND_TYPE_USER, 0);
> + if (ret < 0) {
> + error_setg(errp, "vhost-user-scsi: vhost initialization failed: %s",
> + strerror(-ret));
> + return;
> + }
> +
> + /* Channel and lun both are 0 for bootable vhost-user-scsi disk */
> + s->channel = 0;
> + s->lun = 0;
> + s->target = vs->conf.boot_tpgt;
> +
> + register_savevm(dev, "vhost-user-scsi", vhost_user_scsi_id++, 1,
> + vhost_user_scsi_save, vhost_user_scsi_load, s);
> +}
> +
> +static void vhost_user_scsi_unrealize(DeviceState *dev, Error **errp)
> +{
> + VirtIODevice *vdev = VIRTIO_DEVICE(dev);
> + VHostUserSCSI *s = VHOST_USER_SCSI(dev);
> +
> + /* This will stop the vhost backend. */
> + vhost_user_scsi_set_status(vdev, 0);
> +
> + vhost_dev_cleanup(&s->dev);
> + g_free(s->dev.vqs);
> +
> + virtio_scsi_common_unrealize(dev, errp);
> +}
> +
> +/*
> + * Implementation of an interface to adjust firmware path
> + * for the bootindex property handling.
> + */
> +static char *vhost_user_scsi_get_fw_dev_path(FWPathProvider *p, BusState *bus,
> + DeviceState *dev)
> +{
> + VHostUserSCSI *s = VHOST_USER_SCSI(dev);
> + /* format: /channel@<channel>/vhost-user-scsi@<target>,<lun> */
> + return g_strdup_printf("/channel@%x/%s@%x,%x", s->channel,
> + qdev_fw_name(dev), s->target, s->lun);
> +}
> +
> +static Property vhost_user_scsi_properties[] = {
> + DEFINE_PROP_CHR("chardev", VHostUserSCSI,
> + parent_obj.conf.chardev),
> + DEFINE_PROP_UINT32("boot_tpgt", VHostUserSCSI,
> + parent_obj.conf.boot_tpgt, 0),
> + DEFINE_PROP_UINT32("num_queues", VHostUserSCSI,
> + parent_obj.conf.num_queues, 1),
> + DEFINE_PROP_UINT32("max_sectors", VHostUserSCSI,
> + parent_obj.conf.max_sectors, 0xFFFF),
> + DEFINE_PROP_UINT32("cmd_per_lun", VHostUserSCSI,
> + parent_obj.conf.cmd_per_lun, 128),
> + DEFINE_PROP_END_OF_LIST(),
> +};
> +
> +static void vhost_user_scsi_class_init(ObjectClass *klass, void *data)
> +{
> + DeviceClass *dc = DEVICE_CLASS(klass);
> + VirtioDeviceClass *vdc = VIRTIO_DEVICE_CLASS(klass);
> + FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(klass);
> +
> + dc->props = vhost_user_scsi_properties;
> + set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> + vdc->realize = vhost_user_scsi_realize;
> + vdc->unrealize = vhost_user_scsi_unrealize;
> + vdc->get_features = vhost_user_scsi_get_features;
> + vdc->set_config = vhost_user_scsi_set_config;
> + vdc->set_status = vhost_user_scsi_set_status;
> + fwc->get_dev_path = vhost_user_scsi_get_fw_dev_path;
> +}
> +
> +static void vhost_user_scsi_instance_init(Object *obj)
> +{
> + VHostUserSCSI *dev = VHOST_USER_SCSI(obj);
> + VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(obj);
> +
> + // Add the bootindex property for this object
> + device_add_bootindex_property(obj, &dev->bootindex, "bootindex", NULL,
> + DEVICE(dev), NULL);
> +
> + // Set boot index according the the device config
> + object_property_set_int(obj, vs->conf.bootindex, "bootindex", NULL);
> +}
> +
> +static const TypeInfo vhost_user_scsi_info = {
> + .name = TYPE_VHOST_USER_SCSI,
> + .parent = TYPE_VIRTIO_SCSI_COMMON,
> + .instance_size = sizeof(VHostUserSCSI),
> + .class_init = vhost_user_scsi_class_init,
> + .instance_init = vhost_user_scsi_instance_init,
> + .interfaces = (InterfaceInfo[]) {
> + { TYPE_FW_PATH_PROVIDER },
> + { }
> + },
> +};
> +
> +static void virtio_register_types(void)
> +{
> + type_register_static(&vhost_user_scsi_info);
> +}
> +
> +type_init(virtio_register_types)
> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
> index 06831de..b996d37 100644
> --- a/hw/virtio/virtio-pci.c
> +++ b/hw/virtio/virtio-pci.c
> @@ -2098,6 +2098,61 @@ static const TypeInfo vhost_scsi_pci_info = {
> };
> #endif
>
> +/* vhost-user-scsi-pci */
> +#ifdef CONFIG_VHOST_USER_SCSI
> +static Property vhost_user_scsi_pci_properties[] = {
> + DEFINE_PROP_UINT32("vectors", VirtIOPCIProxy, nvectors,
> + DEV_NVECTORS_UNSPECIFIED),
> + DEFINE_PROP_END_OF_LIST(),
> +};
> +
> +static void vhost_user_scsi_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
> +{
> + VHostUserSCSIPCI *dev = VHOST_USER_SCSI_PCI(vpci_dev);
> + DeviceState *vdev = DEVICE(&dev->vdev);
> + VirtIOSCSICommon *vs = VIRTIO_SCSI_COMMON(vdev);
> +
> + if (vpci_dev->nvectors == DEV_NVECTORS_UNSPECIFIED) {
> + vpci_dev->nvectors = vs->conf.num_queues + 3;
> + }
> +
> + qdev_set_parent_bus(vdev, BUS(&vpci_dev->bus));
> + object_property_set_bool(OBJECT(vdev), true, "realized", errp);
> +}
> +
> +static void vhost_user_scsi_pci_class_init(ObjectClass *klass, void *data)
> +{
> + DeviceClass *dc = DEVICE_CLASS(klass);
> + VirtioPCIClass *k = VIRTIO_PCI_CLASS(klass);
> + PCIDeviceClass *pcidev_k = PCI_DEVICE_CLASS(klass);
> + k->realize = vhost_user_scsi_pci_realize;
> + set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
> + dc->props = vhost_user_scsi_pci_properties;
> + pcidev_k->vendor_id = PCI_VENDOR_ID_REDHAT_QUMRANET;
> + pcidev_k->device_id = PCI_DEVICE_ID_VIRTIO_SCSI;
> + pcidev_k->revision = 0x00;
> + pcidev_k->class_id = PCI_CLASS_STORAGE_SCSI;
> +}
> +
> +static void vhost_user_scsi_pci_instance_init(Object *obj)
> +{
> + VHostUserSCSIPCI *dev = VHOST_USER_SCSI_PCI(obj);
> +
> + virtio_instance_init_common(obj, &dev->vdev, sizeof(dev->vdev),
> + TYPE_VHOST_USER_SCSI);
> + object_property_add_alias(obj, "bootindex", OBJECT(&dev->vdev),
> + "bootindex", &error_abort);
> +}
> +
> +static const TypeInfo vhost_user_scsi_pci_info = {
> + .name = TYPE_VHOST_USER_SCSI_PCI,
> + .parent = TYPE_VIRTIO_PCI,
> + .instance_size = sizeof(VHostUserSCSIPCI),
> + .instance_init = vhost_user_scsi_pci_instance_init,
> + .class_init = vhost_user_scsi_pci_class_init,
> +};
> +#endif /* CONFIG_VHOST_USER_SCSI */
> +
> /* vhost-vsock-pci */
>
> #ifdef CONFIG_VHOST_VSOCK
> @@ -2577,6 +2632,9 @@ static void virtio_pci_register_types(void)
> #ifdef CONFIG_VHOST_SCSI
> type_register_static(&vhost_scsi_pci_info);
> #endif
> +#ifdef CONFIG_VHOST_USER_SCSI
> + type_register_static(&vhost_user_scsi_pci_info);
> +#endif
> #ifdef CONFIG_VHOST_VSOCK
> type_register_static(&vhost_vsock_pci_info);
> #endif
> diff --git a/hw/virtio/virtio-pci.h b/hw/virtio/virtio-pci.h
> index b4edea6..404b2c4 100644
> --- a/hw/virtio/virtio-pci.h
> +++ b/hw/virtio/virtio-pci.h
> @@ -31,6 +31,9 @@
> #ifdef CONFIG_VHOST_SCSI
> #include "hw/virtio/vhost-scsi.h"
> #endif
> +#ifdef CONFIG_VHOST_USER_SCSI
> +#include "hw/virtio/vhost-user-scsi.h"
> +#endif
> #ifdef CONFIG_VHOST_VSOCK
> #include "hw/virtio/vhost-vsock.h"
> #endif
> @@ -42,6 +45,7 @@ typedef struct VirtIOBalloonPCI VirtIOBalloonPCI;
> typedef struct VirtIOSerialPCI VirtIOSerialPCI;
> typedef struct VirtIONetPCI VirtIONetPCI;
> typedef struct VHostSCSIPCI VHostSCSIPCI;
> +typedef struct VHostUserSCSIPCI VHostUserSCSIPCI;
> typedef struct VirtIORngPCI VirtIORngPCI;
> typedef struct VirtIOInputPCI VirtIOInputPCI;
> typedef struct VirtIOInputHIDPCI VirtIOInputHIDPCI;
> @@ -212,6 +216,17 @@ struct VHostSCSIPCI {
> };
> #endif
>
> +#ifdef CONFIG_VHOST_USER_SCSI
> +#define TYPE_VHOST_USER_SCSI_PCI "vhost-user-scsi-pci"
> +#define VHOST_USER_SCSI_PCI(obj) \
> + OBJECT_CHECK(VHostUserSCSIPCI, (obj), TYPE_VHOST_USER_SCSI_PCI)
> +
> +struct VHostUserSCSIPCI {
> + VirtIOPCIProxy parent_obj;
> + VHostUserSCSI vdev;
> +};
> +#endif
> +
> /*
> * virtio-blk-pci: This extends VirtioPCIProxy.
> */
> diff --git a/include/hw/virtio/vhost-user-scsi.h b/include/hw/virtio/vhost-user-scsi.h
> new file mode 100644
> index 0000000..4a16181
> --- /dev/null
> +++ b/include/hw/virtio/vhost-user-scsi.h
> @@ -0,0 +1,45 @@
> +/*
> + * vhost-user-scsi host device
> + *
> + * Copyright (c) 2016 Nutanix Inc. All rights reserved.
> + *
> + * Author:
> + * Felipe Franciosi <felipe@nutanix.com>
> + *
> + * This file is largely based on "vhost-scsi.h" by:
> + * Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
> + *
> + * This work is licensed under the terms of the GNU LGPL, version 2 or later.
> + * See the COPYING.LIB file in the top-level directory.
> + *
> + */
> +
> +#ifndef VHOST_USER_SCSI_H
> +#define VHOST_USER_SCSI_H
> +
> +#include "qemu-common.h"
> +#include "hw/qdev.h"
> +#include "hw/virtio/virtio-scsi.h"
> +#include "hw/virtio/vhost.h"
> +
> +enum vhost_user_scsi_vq_list {
> + VHOST_USER_SCSI_VQ_CONTROL = 0,
> + VHOST_USER_SCSI_VQ_EVENT = 1,
> + VHOST_USER_SCSI_VQ_NUM_FIXED = 2,
> +};
> +
> +#define TYPE_VHOST_USER_SCSI "vhost-user-scsi"
> +#define VHOST_USER_SCSI(obj) \
> + OBJECT_CHECK(VHostUserSCSI, (obj), TYPE_VHOST_USER_SCSI)
> +
> +typedef struct VHostUserSCSI {
> + VirtIOSCSICommon parent_obj;
> +
> + struct vhost_dev dev;
> + int32_t bootindex;
> + int channel;
> + int target;
> + int lun;
> +} VHostUserSCSI;
> +
> +#endif /* VHOST_USER_SCSI_H */
> diff --git a/include/hw/virtio/virtio-scsi.h b/include/hw/virtio/virtio-scsi.h
> index a1e0cfb..4b97386 100644
> --- a/include/hw/virtio/virtio-scsi.h
> +++ b/include/hw/virtio/virtio-scsi.h
> @@ -21,6 +21,7 @@
> #include "hw/virtio/virtio.h"
> #include "hw/pci/pci.h"
> #include "hw/scsi/scsi.h"
> +#include "sysemu/char.h"
> #include "sysemu/iothread.h"
>
> #define TYPE_VIRTIO_SCSI_COMMON "virtio-scsi-common"
> @@ -53,6 +54,10 @@ struct VirtIOSCSIConf {
> char *wwpn;
> uint32_t boot_tpgt;
> IOThread *iothread;
> +#ifdef CONFIG_VHOST_USER_SCSI
> + CharBackend chardev;
> + int32_t bootindex;
> +#endif
> };
>
> struct VirtIOSCSI;
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application
2016-10-26 15:26 ` [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application Felipe Franciosi
@ 2016-10-27 12:16 ` Paolo Bonzini
2016-10-27 12:48 ` Felipe Franciosi
0 siblings, 1 reply; 7+ messages in thread
From: Paolo Bonzini @ 2016-10-27 12:16 UTC (permalink / raw)
To: Felipe Franciosi, Stefan Hajnoczi, Marc-Andre Lureau, Michael S. Tsirkin
Cc: qemu-devel
On 26/10/2016 17:26, Felipe Franciosi wrote:
> This commit introduces a vhost-user-scsi backend sample application. It
> must be linked with libiscsi and libvhost-user.
>
> To use it, compile with:
> make tests/vhost-user-scsi
>
> And run as follows:
> tests/vhost-user-scsi -u /tmp/vus.sock -i iscsi://uri_to_target/
>
> The application is currently limited at one LUN only and it processes
> requests synchronously (therefore only achieving QD1). The purpose of
> the code is to show how a backend can be implemented and to test the
> vhost-user-scsi Qemu implementation.
>
> If a different instance of this vhost-user-scsi application is executed
> at a remote host, a VM can be live migrated to such a host.
Hi,
the right directory for this is contrib/.
Is it possible to use GSource and GIOChannel instead for the event loop?
There is some dead code (for example cb2 as far as I can see) and
having the millionth implementation of an event loop distracts from the
meat of the code. :)
Thanks,
Paolo
> Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
> ---
> tests/Makefile.include | 2 +
> tests/vhost-user-scsi.c | 862 ++++++++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 864 insertions(+)
> create mode 100644 tests/vhost-user-scsi.c
>
> diff --git a/tests/Makefile.include b/tests/Makefile.include
> index 7e6fd23..e61fe54 100644
> --- a/tests/Makefile.include
> +++ b/tests/Makefile.include
> @@ -685,6 +685,8 @@ tests/test-filter-redirector$(EXESUF): tests/test-filter-redirector.o $(qtest-ob
> tests/test-x86-cpuid-compat$(EXESUF): tests/test-x86-cpuid-compat.o $(qtest-obj-y)
> tests/ivshmem-test$(EXESUF): tests/ivshmem-test.o contrib/ivshmem-server/ivshmem-server.o $(libqos-pc-obj-y)
> tests/vhost-user-bridge$(EXESUF): tests/vhost-user-bridge.o contrib/libvhost-user/libvhost-user.o $(test-util-obj-y)
> +tests/vhost-user-scsi.o-cflags := $(LIBISCSI_CFLAGS)
> +tests/vhost-user-scsi$(EXESUF): tests/vhost-user-scsi.o contrib/libvhost-user/libvhost-user.o $(test-util-obj-y) $(test-block-obj-y)
> tests/test-uuid$(EXESUF): tests/test-uuid.o $(test-util-obj-y)
> tests/test-arm-mptimer$(EXESUF): tests/test-arm-mptimer.o
>
> diff --git a/tests/vhost-user-scsi.c b/tests/vhost-user-scsi.c
> new file mode 100644
> index 0000000..c92b3b2
> --- /dev/null
> +++ b/tests/vhost-user-scsi.c
> @@ -0,0 +1,862 @@
> +/*
> + * vhost-user-scsi sample application
> + *
> + * Copyright (c) 2016 Nutanix Inc. All rights reserved.
> + *
> + * Author:
> + * Felipe Franciosi <felipe@nutanix.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 only.
> + * See the COPYING file in the top-level directory.
> + */
> +
> +#include "qemu/osdep.h"
> +#include "contrib/libvhost-user/libvhost-user.h"
> +#include "hw/virtio/virtio-scsi.h"
> +#include "iscsi/iscsi.h"
> +
> +#include <poll.h>
> +
> +#define VHOST_USER_SCSI_DEBUG 1
> +
> +/** Log helpers **/
> +
> +#define PPRE \
> + struct timespec ts; \
> + char timebuf[64]; \
> + struct tm tm; \
> + (void)clock_gettime(CLOCK_REALTIME, &ts); \
> + (void)strftime(timebuf, 64, "%Y%m%d %T", gmtime_r(&ts.tv_sec, &tm))
> +
> +#define PEXT(lvl, msg, ...) do { \
> + PPRE; \
> + fprintf(stderr, "%s.%06ld " lvl ": %s:%s():%d: " msg "\n", \
> + timebuf, ts.tv_nsec/1000, \
> + __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
> +} while(0)
> +
> +#define PNOR(lvl, msg, ...) do { \
> + PPRE; \
> + fprintf(stderr, "%s.%06ld " lvl ": " msg "\n", \
> + timebuf, ts.tv_nsec/1000, ## __VA_ARGS__); \
> +} while(0);
> +
> +#ifdef VHOST_USER_SCSI_DEBUG
> +#define PDBG(msg, ...) PEXT("DBG", msg, ## __VA_ARGS__)
> +#define PERR(msg, ...) PEXT("ERR", msg, ## __VA_ARGS__)
> +#define PLOG(msg, ...) PEXT("LOG", msg, ## __VA_ARGS__)
> +#else
> +#define PDBG(msg, ...) { }
> +#define PERR(msg, ...) PNOR("ERR", msg, ## __VA_ARGS__)
> +#define PLOG(msg, ...) PNOR("LOG", msg, ## __VA_ARGS__)
> +#endif
> +
> +/** vhost-user-scsi specific definitions **/
> +
> +/* TODO: MAX is defined at 8, should be 1024 */
> +#define VUS_SCHED_MAX_FDS (1 + (2*VHOST_MAX_NR_VIRTQUEUE))
> +
> +#define VDEV_SCSI_MAX_LUNS 1 // Only 1 lun supported today
> +#define VDEV_SCSI_MAX_DEVS 1 // Only 1 devices supported today
> +
> +#define ISCSI_INITIATOR "iqn.2016-10.com.nutanix:vhost-user-scsi"
> +
> +typedef void (*misc_cb) (short evt, void *pvt);
> +
> +typedef struct sched_data {
> + vu_watch_cb cb1;
> + misc_cb cb2;
> + void *pvt;
> + short evt;
> +} sched_data_t;
> +
> +typedef struct sched {
> + VuDev *vu_dev;
> + nfds_t nfds;
> + struct pollfd fds[VUS_SCHED_MAX_FDS];
> + sched_data_t data[VUS_SCHED_MAX_FDS];
> + int quit;
> +} sched_t;
> +
> +typedef struct iscsi_lun {
> + struct iscsi_context *iscsi_ctx;
> + int iscsi_lun;
> +} iscsi_lun_t;
> +
> +typedef struct vhost_scsi_dev {
> + VuDev vu_dev;
> + int server_sock;
> + sched_t sched;
> + iscsi_lun_t luns[VDEV_SCSI_MAX_LUNS];
> +} vhost_scsi_dev_t;
> +
> +static vhost_scsi_dev_t *vhost_scsi_devs[VDEV_SCSI_MAX_DEVS];
> +
> +static vhost_scsi_dev_t *vdev_scsi_find_by_vu(VuDev *vu_dev);
> +
> +/** poll-based scheduler for libvhost-user and misc callbacks **/
> +
> +static int sched_add(sched_t *sched, int fd, short int evt,
> + vu_watch_cb cb1, misc_cb cb2, void *pvt) {
> + int i;
> +
> + assert(sched);
> + assert(fd >= 0);
> + assert(evt);
> + assert(cb1 || cb2);
> + assert(!(cb1 && cb2)); // only one of the cbs should be used
> +
> + for (i=0; i<sched->nfds && i<VUS_SCHED_MAX_FDS; i++) {
> + if (sched->fds[i].fd == fd) {
> + break;
> + }
> + }
> + if (i == VUS_SCHED_MAX_FDS) {
> + PERR("Error adding fd: max number of fds reached");
> + return -1;
> + }
> +
> + sched->fds[i].fd = fd;
> + sched->fds[i].events = evt;
> + sched->data[i].cb1 = cb1;
> + sched->data[i].cb2 = cb2;
> + sched->data[i].pvt = pvt;
> + sched->data[i].evt = evt;
> +
> + if (sched->nfds <= i) {
> + sched->nfds = i+1;
> + }
> +
> + PDBG("sched@%p: add fd %d to slot %i", sched, fd, i);
> +
> + return 0;
> +}
> +
> +static int sched_del(sched_t *sched, int fd) {
> + int i;
> +
> + assert(sched);
> + assert(fd >= 0);
> +
> + for (i=0; i<sched->nfds; i++) {
> + if (sched->fds[i].fd == fd) {
> + break;
> + }
> + }
> + if (sched->nfds == i) {
> +#ifdef VUS_PEDANTIC_SCHEDULER
> + PERR("Error deleting fd %d: fd not found", fd);
> + return -1;
> +#else
> + return 0;
> +#endif
> + }
> +
> + sched->nfds--;
> + if (sched->nfds > 0) {
> + // Overwrite deleted entry with last entry from scheduler
> + memcpy(&sched->fds[i], &sched->fds[sched->nfds],
> + sizeof(struct pollfd));
> + memcpy(&sched->data[i], &sched->data[sched->nfds],
> + sizeof(sched_data_t));
> + }
> + memset(&sched->fds[sched->nfds], 0, sizeof(struct pollfd));
> + memset(&sched->data[sched->nfds], 0, sizeof(sched_data_t));
> +
> + PDBG("sched@%p: del fd %d from slot %i", sched, fd, i);
> +
> + return 0;
> +}
> +
> +static int sched_loop(sched_t *sched) {
> + int i, n;
> +
> + assert(sched);
> + assert(sched->nfds > 0);
> +
> + while (!sched->quit) {
> + n = poll(sched->fds, sched->nfds, -1);
> + if (n < 0) {
> + PERR("Error polling: %s", strerror(errno));
> + return -1;
> + }
> +
> + for (i=0; i<sched->nfds && n; i++) {
> + if (sched->fds[i].revents != 0) {
> +
> + if (sched->data[i].cb1) {
> + int vu_evt = 0;
> +
> + if (sched->fds[i].revents & POLLIN) vu_evt |= VU_WATCH_IN;
> + if (sched->fds[i].revents & POLLOUT) vu_evt |= VU_WATCH_OUT;
> + if (sched->fds[i].revents & POLLPRI) vu_evt |= VU_WATCH_PRI;
> + if (sched->fds[i].revents & POLLERR) vu_evt |= VU_WATCH_ERR;
> + if (sched->fds[i].revents & POLLHUP) vu_evt |= VU_WATCH_HUP;
> +
> + PDBG("sched@%p: fd[%d] (%d): cb1(%p, %d, %p)", sched, i,
> + sched->fds[i].fd, sched->vu_dev, vu_evt,
> + sched->data[i].pvt);
> +
> + sched->data[i].cb1(sched->vu_dev, vu_evt,
> + sched->data[i].pvt);
> + } else {
> + PDBG("sched@%p: fd[%d] (%d): cbb(%hd, %p)", sched, i,
> + sched->fds[i].fd, sched->fds[i].revents,
> + sched->data[i].pvt);
> +
> + sched->data[i].cb2(sched->fds[i].revents,
> + sched->data[i].pvt);
> + }
> +
> + n--;
> + }
> + }
> + }
> +
> + return 0;
> +}
> +
> +/** from libiscsi's scsi-lowlevel.h **/
> +
> +#define SCSI_CDB_MAX_SIZE 16
> +
> +struct scsi_iovector {
> + struct scsi_iovec *iov;
> + int niov;
> + int nalloc;
> + size_t offset;
> + int consumed;
> +};
> +
> +struct scsi_allocated_memory {
> + struct scsi_allocated_memory *next;
> + char buf[0];
> +};
> +
> +struct scsi_data {
> + int size;
> + unsigned char *data;
> +};
> +
> +enum scsi_sense_key {
> + SCSI_SENSE_NO_SENSE = 0x00,
> + SCSI_SENSE_RECOVERED_ERROR = 0x01,
> + SCSI_SENSE_NOT_READY = 0x02,
> + SCSI_SENSE_MEDIUM_ERROR = 0x03,
> + SCSI_SENSE_HARDWARE_ERROR = 0x04,
> + SCSI_SENSE_ILLEGAL_REQUEST = 0x05,
> + SCSI_SENSE_UNIT_ATTENTION = 0x06,
> + SCSI_SENSE_DATA_PROTECTION = 0x07,
> + SCSI_SENSE_BLANK_CHECK = 0x08,
> + SCSI_SENSE_VENDOR_SPECIFIC = 0x09,
> + SCSI_SENSE_COPY_ABORTED = 0x0a,
> + SCSI_SENSE_COMMAND_ABORTED = 0x0b,
> + SCSI_SENSE_OBSOLETE_ERROR_CODE = 0x0c,
> + SCSI_SENSE_OVERFLOW_COMMAND = 0x0d,
> + SCSI_SENSE_MISCOMPARE = 0x0e
> +};
> +
> +struct scsi_sense {
> + unsigned char error_type;
> + enum scsi_sense_key key;
> + int ascq;
> + unsigned sense_specific:1;
> + unsigned ill_param_in_cdb:1;
> + unsigned bit_pointer_valid:1;
> + unsigned char bit_pointer;
> + uint16_t field_pointer;
> +};
> +
> +enum scsi_residual {
> + SCSI_RESIDUAL_NO_RESIDUAL = 0,
> + SCSI_RESIDUAL_UNDERFLOW,
> + SCSI_RESIDUAL_OVERFLOW
> +};
> +
> +struct scsi_task {
> + int status;
> + int cdb_size;
> + int xfer_dir;
> + int expxferlen;
> + unsigned char cdb[SCSI_CDB_MAX_SIZE];
> + enum scsi_residual residual_status;
> + size_t residual;
> + struct scsi_sense sense;
> + struct scsi_data datain;
> + struct scsi_allocated_memory *mem;
> + void *ptr;
> +
> + uint32_t itt;
> + uint32_t cmdsn;
> + uint32_t lun;
> +
> + struct scsi_iovector iovector_in;
> + struct scsi_iovector iovector_out;
> +};
> +
> +/** libiscsi integration **/
> +
> +static int iscsi_add_lun(iscsi_lun_t *lun, char *iscsi_uri) {
> + struct iscsi_url *iscsi_url;
> + struct iscsi_context *iscsi_ctx;
> + int ret = 0;
> +
> + assert(lun);
> +
> + iscsi_ctx = iscsi_create_context(ISCSI_INITIATOR);
> + if (!iscsi_ctx) {
> + PERR("Unable to create iSCSI context");
> + return -1;
> + }
> +
> + iscsi_url = iscsi_parse_full_url(iscsi_ctx, iscsi_uri);
> + if (!iscsi_url) {
> + PERR("Unable to parse iSCSI URL: %s", iscsi_get_error(iscsi_ctx));
> + goto fail;
> + }
> +
> + iscsi_set_session_type(iscsi_ctx, ISCSI_SESSION_NORMAL);
> + iscsi_set_header_digest(iscsi_ctx, ISCSI_HEADER_DIGEST_NONE_CRC32C);
> + if (iscsi_full_connect_sync(iscsi_ctx, iscsi_url->portal, iscsi_url->lun)) {
> + PERR("Unable to login to iSCSI portal: %s", iscsi_get_error(iscsi_ctx));
> + goto fail;
> + }
> +
> + lun->iscsi_ctx = iscsi_ctx;
> + lun->iscsi_lun = iscsi_url->lun;
> +
> + PDBG("Context %p created for lun 0: %s", iscsi_ctx, iscsi_uri);
> +
> +out:
> + if (iscsi_url) {
> + iscsi_destroy_url(iscsi_url);
> + }
> + return ret;
> +
> +fail:
> + (void)iscsi_destroy_context(iscsi_ctx);
> + ret = -1;
> + goto out;
> +}
> +
> +static struct scsi_task *scsi_task_new(int cdb_len, uint8_t *cdb, int dir,
> + int xfer_len) {
> + struct scsi_task *task;
> +
> + assert(cdb_len > 0);
> + assert(cdb);
> +
> + task = calloc(1, sizeof(struct scsi_task));
> + if (!task) {
> + PERR("Error allocating task: %s", strerror(errno));
> + return NULL;
> + }
> +
> + memcpy(task->cdb, cdb, cdb_len);
> + task->cdb_size = cdb_len;
> + task->xfer_dir = dir;
> + task->expxferlen = xfer_len;
> +
> + return task;
> +}
> +
> +static int get_cdb_len(uint8_t *cdb) {
> + switch(cdb[0] >> 5){
> + case 0:
> + return 6;
> + case 1:
> + case 2:
> + return 10;
> + case 4:
> + return 16;
> + case 5:
> + return 12;
> + }
> + PERR("Unable to determine cdb len (0x%02hhX)", cdb[0]>>5);
> + return -1;
> +}
> +
> +static int handle_cmd_sync(struct iscsi_context *ctx,
> + VirtIOSCSICmdReq *req,
> + struct iovec *out, unsigned int out_len,
> + VirtIOSCSICmdResp *rsp,
> + struct iovec *in, unsigned int in_len) {
> + struct scsi_task *task;
> + uint32_t dir;
> + uint32_t len;
> + int cdb_len;
> + int i;
> +
> + if (!((!req->lun[1]) && (req->lun[2] == 0x40) && (!req->lun[3]))) {
> + // Ignore anything different than target=0, lun=0
> + PDBG("Ignoring unconnected lun (0x%hhX, 0x%hhX)",
> + req->lun[1], req->lun[3]);
> + rsp->status = SCSI_STATUS_CHECK_CONDITION;
> + memset(rsp->sense, 0, sizeof(rsp->sense));
> + rsp->sense_len = 18;
> + rsp->sense[0] = 0x70;
> + rsp->sense[2] = 0x05; // ILLEGAL_REQUEST
> + rsp->sense[7] = 10;
> + rsp->sense[12] = 0x24;
> +
> + return 0;
> + }
> +
> + cdb_len = get_cdb_len(req->cdb);
> + if (cdb_len == -1) {
> + return -1;
> + }
> +
> + len = 0;
> + if (!out_len && !in_len) {
> + dir = SCSI_XFER_NONE;
> + } else if (out_len) {
> + dir = SCSI_XFER_TO_DEV;
> + for (i=0; i<out_len; i++) {
> + len += out[i].iov_len;
> + }
> + } else {
> + dir = SCSI_XFER_FROM_DEV;
> + for (i=0; i<in_len; i++) {
> + len += in[i].iov_len;
> + }
> + }
> +
> + task = scsi_task_new(cdb_len, req->cdb, dir, len);
> + if (!task) {
> + PERR("Unable to create iscsi task");
> + return -1;
> + }
> +
> + if (dir == SCSI_XFER_TO_DEV) {
> + task->iovector_out.iov = (struct scsi_iovec *)out;
> + task->iovector_out.niov = out_len;
> + } else if (dir == SCSI_XFER_FROM_DEV) {
> + task->iovector_in.iov = (struct scsi_iovec *)in;
> + task->iovector_in.niov = in_len;
> + }
> +
> + PDBG("Sending iscsi cmd (cdb_len=%d, dir=%d, task=%p)",
> + cdb_len, dir, task);
> + if (!iscsi_scsi_command_sync(ctx, 0, task, NULL)) {
> + PERR("Error serving SCSI command");
> + free(task);
> + return -1;
> + }
> +
> + memset(rsp, 0, sizeof(*rsp));
> +
> + rsp->status = task->status;
> + rsp->resid = task->residual;
> +
> + if (task->status == SCSI_STATUS_CHECK_CONDITION) {
> + rsp->response = VIRTIO_SCSI_S_FAILURE;
> + rsp->sense_len = task->datain.size - 2;
> + memcpy(rsp->sense, &task->datain.data[2], rsp->sense_len);
> + }
> +
> + free(task);
> +
> + PDBG("Filled in rsp: status=%hhX, resid=%u, response=%hhX, sense_len=%u",
> + rsp->status, rsp->resid, rsp->response, rsp->sense_len);
> +
> + return 0;
> +}
> +
> +/** libvhost-user callbacks **/
> +
> +static void vus_panic_cb(VuDev *vu_dev, const char *buf) {
> + vhost_scsi_dev_t *vdev_scsi;
> +
> + assert(vu_dev);
> +
> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
> +
> + if (buf) {
> + PERR("vu_panic: %s", buf);
> + }
> +
> + if (vdev_scsi) {
> + vdev_scsi->sched.quit = 1;
> + }
> +}
> +
> +static void vus_add_watch_cb(VuDev *vu_dev, int fd, int vu_evt, vu_watch_cb cb,
> + void *pvt) {
> + vhost_scsi_dev_t *vdev_scsi;
> + int poll_evt = 0;
> +
> + assert(vu_dev);
> + assert(fd >= 0);
> + assert(cb);
> +
> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
> + if (!vdev_scsi) {
> + vus_panic_cb(vu_dev, NULL);
> + }
> +
> + /* TODO: VU_WATCH_* should match POLL*, check it */
> + if (vu_evt & VU_WATCH_IN) poll_evt |= POLLIN;
> + if (vu_evt & VU_WATCH_OUT) poll_evt |= POLLOUT;
> + if (vu_evt & VU_WATCH_PRI) poll_evt |= POLLPRI;
> + if (vu_evt & VU_WATCH_ERR) poll_evt |= POLLERR;
> + if (vu_evt & VU_WATCH_HUP) poll_evt |= POLLHUP;
> +
> + if (sched_add(&vdev_scsi->sched, fd, poll_evt, cb, NULL, pvt)) {
> + vus_panic_cb(vu_dev, NULL);
> + }
> +}
> +
> +static void vus_del_watch_cb(VuDev *vu_dev, int fd) {
> + vhost_scsi_dev_t *vdev_scsi;
> +
> + assert(vu_dev);
> + assert(fd >= 0);
> +
> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
> + if (!vdev_scsi) {
> + vus_panic_cb(vu_dev, NULL);
> + return;
> + }
> +
> + if (sched_del(&vdev_scsi->sched, fd)) {
> + vus_panic_cb(vu_dev, NULL);
> + }
> +}
> +
> +static void vus_proc_ctl(VuDev *vu_dev, int idx) {
> + /* Control VQ not implemented */
> +}
> +
> +static void vus_proc_evt(VuDev *vu_dev, int idx) {
> + /* Event VQ not implemented */
> +}
> +
> +static void vus_proc_req(VuDev *vu_dev, int idx) {
> + vhost_scsi_dev_t *vdev_scsi;
> + VuVirtq *vq;
> +
> + assert(vu_dev);
> +
> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
> + if (!vdev_scsi) {
> + vus_panic_cb(vu_dev, NULL);
> + return;
> + }
> +
> + if ((idx < 0) || (idx >= VHOST_MAX_NR_VIRTQUEUE)) {
> + PERR("VQ Index out of range: %d", idx);
> + vus_panic_cb(vu_dev, NULL);
> + return;
> + }
> +
> + vq = vu_get_queue(vu_dev, idx);
> + if (!vq) {
> + PERR("Error fetching VQ (dev=%p, idx=%d)", vu_dev, idx);
> + vus_panic_cb(vu_dev, NULL);
> + return;
> + }
> +
> + PDBG("Got kicked on vq[%d]@%p", idx, vq);
> +
> + while(1) {
> + VuVirtqElement *elem;
> + VirtIOSCSICmdReq *req;
> + VirtIOSCSICmdResp *rsp;
> +
> + elem = vu_queue_pop(vu_dev, vq, sizeof(VuVirtqElement));
> + if (!elem) {
> + PDBG("No more elements pending on vq[%d]@%p", idx, vq);
> + break;
> + }
> + PDBG("Popped elem@%p", elem);
> +
> + assert(!((elem->out_num > 1) && (elem->in_num > 1)));
> + assert((elem->out_num > 0) && (elem->in_num > 0));
> +
> + if (elem->out_sg[0].iov_len < sizeof(VirtIOSCSICmdReq)) {
> + PERR("Invalid virtio-scsi req header");
> + vus_panic_cb(vu_dev, NULL);
> + break;
> + }
> + req = (VirtIOSCSICmdReq *)elem->out_sg[0].iov_base;
> +
> + if (elem->in_sg[0].iov_len < sizeof(VirtIOSCSICmdResp)) {
> + PERR("Invalid virtio-scsi rsp header");
> + vus_panic_cb(vu_dev, NULL);
> + break;
> + }
> + rsp = (VirtIOSCSICmdResp *)elem->in_sg[0].iov_base;
> +
> + if (handle_cmd_sync(vdev_scsi->luns[0].iscsi_ctx,
> + req, &elem->out_sg[1], elem->out_num-1,
> + rsp, &elem->in_sg[1], elem->in_num-1) != 0) {
> + vus_panic_cb(vu_dev, NULL);
> + break;
> + }
> +
> + vu_queue_push(vu_dev, vq, elem, 0);
> + vu_queue_notify(vu_dev, vq);
> +
> + free(elem);
> + }
> +
> +}
> +
> +static void vus_queue_set_started(VuDev *vu_dev, int idx, bool started) {
> + VuVirtq *vq;
> +
> + assert(vu_dev);
> +
> + if ((idx < 0) || (idx >= VHOST_MAX_NR_VIRTQUEUE)) {
> + PERR("VQ Index out of range: %d", idx);
> + vus_panic_cb(vu_dev, NULL);
> + return;
> + }
> +
> + vq = vu_get_queue(vu_dev, idx);
> +
> + switch(idx) {
> + case 0:
> + vu_set_queue_handler(vu_dev, vq, started?vus_proc_ctl:NULL);
> + break;
> + case 1:
> + vu_set_queue_handler(vu_dev, vq, started?vus_proc_evt:NULL);
> + break;
> + default:
> + vu_set_queue_handler(vu_dev, vq, started?vus_proc_req:NULL);
> + }
> +}
> +
> +static const VuDevIface vus_iface = {
> + .queue_set_started = vus_queue_set_started,
> +};
> +
> +static void vus_vhost_cb(VuDev *vu_dev, int vu_evt, void *data) {
> + assert(vu_dev);
> +
> + if (!vu_dispatch(vu_dev) != 0) {
> + PERR("Error processing vhost message");
> + vus_panic_cb(vu_dev, NULL);
> + }
> +}
> +
> +/** util **/
> +
> +static int unix_sock_new(char *unix_fn) {
> + int sock;
> + struct sockaddr_un un;
> + size_t len;
> +
> + assert(unix_fn);
> +
> + sock = socket(AF_UNIX, SOCK_STREAM, 0);
> + if (sock <= 0) {
> + perror("socket");
> + return -1;
> + }
> +
> + un.sun_family = AF_UNIX;
> + (void)snprintf(un.sun_path, sizeof(un.sun_path), "%s", unix_fn);
> + len = sizeof(un.sun_family) + strlen(un.sun_path);
> +
> + (void)unlink(unix_fn);
> + if (bind(sock, (struct sockaddr *)&un, len) < 0) {
> + perror("bind");
> + goto fail;
> + }
> +
> + if (listen(sock, 1) < 0) {
> + perror("listen");
> + goto fail;
> + }
> +
> + return sock;
> +
> +fail:
> + (void)close(sock);
> +
> + return -1;
> +}
> +
> +/** vhost-user-scsi **/
> +
> +static vhost_scsi_dev_t *vdev_scsi_find_by_vu(VuDev *vu_dev) {
> + int i;
> +
> + assert(vu_dev);
> +
> + for (i=0; i<VDEV_SCSI_MAX_DEVS; i++) {
> + if (&vhost_scsi_devs[i]->vu_dev == vu_dev) {
> + return vhost_scsi_devs[i];
> + }
> + }
> +
> + PERR("Unknown VuDev %p", vu_dev);
> + return NULL;
> +}
> +
> +static void vdev_scsi_deinit(vhost_scsi_dev_t *vdev_scsi) {
> + if (!vdev_scsi) {
> + return;
> + }
> +
> + if (vdev_scsi->server_sock >= 0) {
> + struct sockaddr_storage ss;
> + socklen_t sslen = sizeof(ss);
> +
> + if (getsockname(vdev_scsi->server_sock, (struct sockaddr *)&ss,
> + &sslen) == 0) {
> + struct sockaddr_un *su = (struct sockaddr_un *)&ss;
> + (void)unlink(su->sun_path);
> + }
> +
> + (void)close(vdev_scsi->server_sock);
> + }
> +}
> +
> +static vhost_scsi_dev_t *vdev_scsi_new(char *unix_fn) {
> + vhost_scsi_dev_t *vdev_scsi;
> +
> + assert(unix_fn);
> +
> + vdev_scsi = calloc(1, sizeof(vhost_scsi_dev_t));
> + if (!vdev_scsi) {
> + perror("calloc");
> + return NULL;
> + }
> +
> + vdev_scsi->server_sock = unix_sock_new(unix_fn);
> + if (vdev_scsi->server_sock < 0) {
> + free(vdev_scsi);
> + return NULL;
> + }
> +
> + vdev_scsi->sched.vu_dev = &vdev_scsi->vu_dev;
> +
> + return vdev_scsi;
> +}
> +
> +static int vdev_scsi_iscsi_add_lun(vhost_scsi_dev_t *vdev_scsi,
> + char *iscsi_uri, uint32_t lun) {
> + assert(vdev_scsi);
> + assert(iscsi_uri);
> + assert(lun < VDEV_SCSI_MAX_LUNS);
> +
> + if (vdev_scsi->luns[lun].iscsi_ctx) {
> + PERR("Lun %d already configured", lun);
> + return -1;
> + }
> +
> + if (iscsi_add_lun(&vdev_scsi->luns[lun], iscsi_uri) != 0) {
> + return -1;
> + }
> +
> + return 0;
> +}
> +
> +static int vdev_scsi_run(vhost_scsi_dev_t *vdev_scsi) {
> + int cli_sock;
> + int ret = 0;
> +
> + assert(vdev_scsi);
> + assert(vdev_scsi->server_sock >= 0);
> +
> + cli_sock = accept(vdev_scsi->server_sock, (void *)0, (void *)0);
> + if (cli_sock < 0) {
> + perror("accept");
> + return -1;
> + }
> +
> + vu_init(&vdev_scsi->vu_dev,
> + cli_sock,
> + vus_panic_cb,
> + vus_add_watch_cb,
> + vus_del_watch_cb,
> + &vus_iface);
> +
> + ret = sched_add(&vdev_scsi->sched, cli_sock, POLLIN, vus_vhost_cb, NULL, 0);
> + if (ret) {
> + goto fail;
> + }
> +
> + if (sched_loop(&vdev_scsi->sched) != 0) {
> + goto fail;
> + }
> +
> +out:
> + vu_deinit(&vdev_scsi->vu_dev);
> +
> + return ret;
> +
> +fail:
> + ret = -1;
> + goto out;
> +}
> +
> +int main(int argc, char **argv)
> +{
> + vhost_scsi_dev_t *vdev_scsi = NULL;
> + char *unix_fn = NULL;
> + char *iscsi_uri = NULL;
> + int opt, err = EXIT_SUCCESS;
> +
> + while ((opt = getopt(argc, argv, "u:i:")) != -1) {
> + switch (opt) {
> + case 'h':
> + goto help;
> + case 'u':
> + unix_fn = strdup(optarg);
> + break;
> + case 'i':
> + iscsi_uri = strdup(optarg);
> + break;
> + default:
> + goto help;
> + }
> + }
> + if (!unix_fn || !iscsi_uri) {
> + goto help;
> + }
> +
> + vdev_scsi = vdev_scsi_new(unix_fn);
> + if (!vdev_scsi) {
> + goto err;
> + }
> + vhost_scsi_devs[0] = vdev_scsi;
> +
> + if (vdev_scsi_iscsi_add_lun(vdev_scsi, iscsi_uri, 0) != 0) {
> + goto err;
> + }
> +
> + if (vdev_scsi_run(vdev_scsi) != 0) {
> + goto err;
> + }
> +
> +out:
> + if (vdev_scsi) {
> + vdev_scsi_deinit(vdev_scsi);
> + free(vdev_scsi);
> + }
> + if (unix_fn) {
> + free(unix_fn);
> + }
> + if (iscsi_uri) {
> + free(iscsi_uri);
> + }
> +
> + return err;
> +
> +err:
> + err = EXIT_FAILURE;
> + goto out;
> +
> +help:
> + fprintf(stderr, "Usage: %s [ -u unix_sock_path -i iscsi_uri ] | [ -h ]\n",
> + argv[0]);
> + fprintf(stderr, " -u path to unix socket\n");
> + fprintf(stderr, " -i iscsi uri for lun 0\n");
> + fprintf(stderr, " -h print help and quit\n");
> +
> + goto err;
> +}
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application
2016-10-27 12:16 ` Paolo Bonzini
@ 2016-10-27 12:48 ` Felipe Franciosi
2016-10-27 12:58 ` Paolo Bonzini
0 siblings, 1 reply; 7+ messages in thread
From: Felipe Franciosi @ 2016-10-27 12:48 UTC (permalink / raw)
To: Paolo Bonzini
Cc: Felipe Franciosi, Stefan Hajnoczi, Marc-Andre Lureau,
Michael S. Tsirkin, qemu-devel
Hello,
> On 27 Oct 2016, at 13:16, Paolo Bonzini <pbonzini@redhat.com> wrote:
>
>
>
> On 26/10/2016 17:26, Felipe Franciosi wrote:
>> This commit introduces a vhost-user-scsi backend sample application. It
>> must be linked with libiscsi and libvhost-user.
>>
>> To use it, compile with:
>> make tests/vhost-user-scsi
>>
>> And run as follows:
>> tests/vhost-user-scsi -u /tmp/vus.sock -i iscsi://uri_to_target/
>>
>> The application is currently limited at one LUN only and it processes
>> requests synchronously (therefore only achieving QD1). The purpose of
>> the code is to show how a backend can be implemented and to test the
>> vhost-user-scsi Qemu implementation.
>>
>> If a different instance of this vhost-user-scsi application is executed
>> at a remote host, a VM can be live migrated to such a host.
>
> Hi,
>
> the right directory for this is contrib/.
Cool. I was following suit from vhost-user-bridge which lives in tests/ today. To me, it makes more sense for these to be in contrib/. I'll place my sample application there for v2 and perhaps we should move vhost-user-bridge later?
>
> Is it possible to use GSource and GIOChannel instead for the event loop?
> There is some dead code (for example cb2 as far as I can see) and
> having the millionth implementation of an event loop distracts from the
> meat of the code. :)
That's true. I'll have a stab at using glib's event loop. The cb2 was meant to be used for libiscsi's async submission, but I ended up with QD1 for simplicity. You're right, it looks pretty dead at the minute. :)
Cheers,
Felipe
>
> Thanks,
>
> Paolo
>
>> Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
>> ---
>> tests/Makefile.include | 2 +
>> tests/vhost-user-scsi.c | 862 ++++++++++++++++++++++++++++++++++++++++++++++++
>> 2 files changed, 864 insertions(+)
>> create mode 100644 tests/vhost-user-scsi.c
>>
>> diff --git a/tests/Makefile.include b/tests/Makefile.include
>> index 7e6fd23..e61fe54 100644
>> --- a/tests/Makefile.include
>> +++ b/tests/Makefile.include
>> @@ -685,6 +685,8 @@ tests/test-filter-redirector$(EXESUF): tests/test-filter-redirector.o $(qtest-ob
>> tests/test-x86-cpuid-compat$(EXESUF): tests/test-x86-cpuid-compat.o $(qtest-obj-y)
>> tests/ivshmem-test$(EXESUF): tests/ivshmem-test.o contrib/ivshmem-server/ivshmem-server.o $(libqos-pc-obj-y)
>> tests/vhost-user-bridge$(EXESUF): tests/vhost-user-bridge.o contrib/libvhost-user/libvhost-user.o $(test-util-obj-y)
>> +tests/vhost-user-scsi.o-cflags := $(LIBISCSI_CFLAGS)
>> +tests/vhost-user-scsi$(EXESUF): tests/vhost-user-scsi.o contrib/libvhost-user/libvhost-user.o $(test-util-obj-y) $(test-block-obj-y)
>> tests/test-uuid$(EXESUF): tests/test-uuid.o $(test-util-obj-y)
>> tests/test-arm-mptimer$(EXESUF): tests/test-arm-mptimer.o
>>
>> diff --git a/tests/vhost-user-scsi.c b/tests/vhost-user-scsi.c
>> new file mode 100644
>> index 0000000..c92b3b2
>> --- /dev/null
>> +++ b/tests/vhost-user-scsi.c
>> @@ -0,0 +1,862 @@
>> +/*
>> + * vhost-user-scsi sample application
>> + *
>> + * Copyright (c) 2016 Nutanix Inc. All rights reserved.
>> + *
>> + * Author:
>> + * Felipe Franciosi <felipe@nutanix.com>
>> + *
>> + * This work is licensed under the terms of the GNU GPL, version 2 only.
>> + * See the COPYING file in the top-level directory.
>> + */
>> +
>> +#include "qemu/osdep.h"
>> +#include "contrib/libvhost-user/libvhost-user.h"
>> +#include "hw/virtio/virtio-scsi.h"
>> +#include "iscsi/iscsi.h"
>> +
>> +#include <poll.h>
>> +
>> +#define VHOST_USER_SCSI_DEBUG 1
>> +
>> +/** Log helpers **/
>> +
>> +#define PPRE \
>> + struct timespec ts; \
>> + char timebuf[64]; \
>> + struct tm tm; \
>> + (void)clock_gettime(CLOCK_REALTIME, &ts); \
>> + (void)strftime(timebuf, 64, "%Y%m%d %T", gmtime_r(&ts.tv_sec, &tm))
>> +
>> +#define PEXT(lvl, msg, ...) do { \
>> + PPRE; \
>> + fprintf(stderr, "%s.%06ld " lvl ": %s:%s():%d: " msg "\n", \
>> + timebuf, ts.tv_nsec/1000, \
>> + __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
>> +} while(0)
>> +
>> +#define PNOR(lvl, msg, ...) do { \
>> + PPRE; \
>> + fprintf(stderr, "%s.%06ld " lvl ": " msg "\n", \
>> + timebuf, ts.tv_nsec/1000, ## __VA_ARGS__); \
>> +} while(0);
>> +
>> +#ifdef VHOST_USER_SCSI_DEBUG
>> +#define PDBG(msg, ...) PEXT("DBG", msg, ## __VA_ARGS__)
>> +#define PERR(msg, ...) PEXT("ERR", msg, ## __VA_ARGS__)
>> +#define PLOG(msg, ...) PEXT("LOG", msg, ## __VA_ARGS__)
>> +#else
>> +#define PDBG(msg, ...) { }
>> +#define PERR(msg, ...) PNOR("ERR", msg, ## __VA_ARGS__)
>> +#define PLOG(msg, ...) PNOR("LOG", msg, ## __VA_ARGS__)
>> +#endif
>> +
>> +/** vhost-user-scsi specific definitions **/
>> +
>> +/* TODO: MAX is defined at 8, should be 1024 */
>> +#define VUS_SCHED_MAX_FDS (1 + (2*VHOST_MAX_NR_VIRTQUEUE))
>> +
>> +#define VDEV_SCSI_MAX_LUNS 1 // Only 1 lun supported today
>> +#define VDEV_SCSI_MAX_DEVS 1 // Only 1 devices supported today
>> +
>> +#define ISCSI_INITIATOR "iqn.2016-10.com.nutanix:vhost-user-scsi"
>> +
>> +typedef void (*misc_cb) (short evt, void *pvt);
>> +
>> +typedef struct sched_data {
>> + vu_watch_cb cb1;
>> + misc_cb cb2;
>> + void *pvt;
>> + short evt;
>> +} sched_data_t;
>> +
>> +typedef struct sched {
>> + VuDev *vu_dev;
>> + nfds_t nfds;
>> + struct pollfd fds[VUS_SCHED_MAX_FDS];
>> + sched_data_t data[VUS_SCHED_MAX_FDS];
>> + int quit;
>> +} sched_t;
>> +
>> +typedef struct iscsi_lun {
>> + struct iscsi_context *iscsi_ctx;
>> + int iscsi_lun;
>> +} iscsi_lun_t;
>> +
>> +typedef struct vhost_scsi_dev {
>> + VuDev vu_dev;
>> + int server_sock;
>> + sched_t sched;
>> + iscsi_lun_t luns[VDEV_SCSI_MAX_LUNS];
>> +} vhost_scsi_dev_t;
>> +
>> +static vhost_scsi_dev_t *vhost_scsi_devs[VDEV_SCSI_MAX_DEVS];
>> +
>> +static vhost_scsi_dev_t *vdev_scsi_find_by_vu(VuDev *vu_dev);
>> +
>> +/** poll-based scheduler for libvhost-user and misc callbacks **/
>> +
>> +static int sched_add(sched_t *sched, int fd, short int evt,
>> + vu_watch_cb cb1, misc_cb cb2, void *pvt) {
>> + int i;
>> +
>> + assert(sched);
>> + assert(fd >= 0);
>> + assert(evt);
>> + assert(cb1 || cb2);
>> + assert(!(cb1 && cb2)); // only one of the cbs should be used
>> +
>> + for (i=0; i<sched->nfds && i<VUS_SCHED_MAX_FDS; i++) {
>> + if (sched->fds[i].fd == fd) {
>> + break;
>> + }
>> + }
>> + if (i == VUS_SCHED_MAX_FDS) {
>> + PERR("Error adding fd: max number of fds reached");
>> + return -1;
>> + }
>> +
>> + sched->fds[i].fd = fd;
>> + sched->fds[i].events = evt;
>> + sched->data[i].cb1 = cb1;
>> + sched->data[i].cb2 = cb2;
>> + sched->data[i].pvt = pvt;
>> + sched->data[i].evt = evt;
>> +
>> + if (sched->nfds <= i) {
>> + sched->nfds = i+1;
>> + }
>> +
>> + PDBG("sched@%p: add fd %d to slot %i", sched, fd, i);
>> +
>> + return 0;
>> +}
>> +
>> +static int sched_del(sched_t *sched, int fd) {
>> + int i;
>> +
>> + assert(sched);
>> + assert(fd >= 0);
>> +
>> + for (i=0; i<sched->nfds; i++) {
>> + if (sched->fds[i].fd == fd) {
>> + break;
>> + }
>> + }
>> + if (sched->nfds == i) {
>> +#ifdef VUS_PEDANTIC_SCHEDULER
>> + PERR("Error deleting fd %d: fd not found", fd);
>> + return -1;
>> +#else
>> + return 0;
>> +#endif
>> + }
>> +
>> + sched->nfds--;
>> + if (sched->nfds > 0) {
>> + // Overwrite deleted entry with last entry from scheduler
>> + memcpy(&sched->fds[i], &sched->fds[sched->nfds],
>> + sizeof(struct pollfd));
>> + memcpy(&sched->data[i], &sched->data[sched->nfds],
>> + sizeof(sched_data_t));
>> + }
>> + memset(&sched->fds[sched->nfds], 0, sizeof(struct pollfd));
>> + memset(&sched->data[sched->nfds], 0, sizeof(sched_data_t));
>> +
>> + PDBG("sched@%p: del fd %d from slot %i", sched, fd, i);
>> +
>> + return 0;
>> +}
>> +
>> +static int sched_loop(sched_t *sched) {
>> + int i, n;
>> +
>> + assert(sched);
>> + assert(sched->nfds > 0);
>> +
>> + while (!sched->quit) {
>> + n = poll(sched->fds, sched->nfds, -1);
>> + if (n < 0) {
>> + PERR("Error polling: %s", strerror(errno));
>> + return -1;
>> + }
>> +
>> + for (i=0; i<sched->nfds && n; i++) {
>> + if (sched->fds[i].revents != 0) {
>> +
>> + if (sched->data[i].cb1) {
>> + int vu_evt = 0;
>> +
>> + if (sched->fds[i].revents & POLLIN) vu_evt |= VU_WATCH_IN;
>> + if (sched->fds[i].revents & POLLOUT) vu_evt |= VU_WATCH_OUT;
>> + if (sched->fds[i].revents & POLLPRI) vu_evt |= VU_WATCH_PRI;
>> + if (sched->fds[i].revents & POLLERR) vu_evt |= VU_WATCH_ERR;
>> + if (sched->fds[i].revents & POLLHUP) vu_evt |= VU_WATCH_HUP;
>> +
>> + PDBG("sched@%p: fd[%d] (%d): cb1(%p, %d, %p)", sched, i,
>> + sched->fds[i].fd, sched->vu_dev, vu_evt,
>> + sched->data[i].pvt);
>> +
>> + sched->data[i].cb1(sched->vu_dev, vu_evt,
>> + sched->data[i].pvt);
>> + } else {
>> + PDBG("sched@%p: fd[%d] (%d): cbb(%hd, %p)", sched, i,
>> + sched->fds[i].fd, sched->fds[i].revents,
>> + sched->data[i].pvt);
>> +
>> + sched->data[i].cb2(sched->fds[i].revents,
>> + sched->data[i].pvt);
>> + }
>> +
>> + n--;
>> + }
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +/** from libiscsi's scsi-lowlevel.h **/
>> +
>> +#define SCSI_CDB_MAX_SIZE 16
>> +
>> +struct scsi_iovector {
>> + struct scsi_iovec *iov;
>> + int niov;
>> + int nalloc;
>> + size_t offset;
>> + int consumed;
>> +};
>> +
>> +struct scsi_allocated_memory {
>> + struct scsi_allocated_memory *next;
>> + char buf[0];
>> +};
>> +
>> +struct scsi_data {
>> + int size;
>> + unsigned char *data;
>> +};
>> +
>> +enum scsi_sense_key {
>> + SCSI_SENSE_NO_SENSE = 0x00,
>> + SCSI_SENSE_RECOVERED_ERROR = 0x01,
>> + SCSI_SENSE_NOT_READY = 0x02,
>> + SCSI_SENSE_MEDIUM_ERROR = 0x03,
>> + SCSI_SENSE_HARDWARE_ERROR = 0x04,
>> + SCSI_SENSE_ILLEGAL_REQUEST = 0x05,
>> + SCSI_SENSE_UNIT_ATTENTION = 0x06,
>> + SCSI_SENSE_DATA_PROTECTION = 0x07,
>> + SCSI_SENSE_BLANK_CHECK = 0x08,
>> + SCSI_SENSE_VENDOR_SPECIFIC = 0x09,
>> + SCSI_SENSE_COPY_ABORTED = 0x0a,
>> + SCSI_SENSE_COMMAND_ABORTED = 0x0b,
>> + SCSI_SENSE_OBSOLETE_ERROR_CODE = 0x0c,
>> + SCSI_SENSE_OVERFLOW_COMMAND = 0x0d,
>> + SCSI_SENSE_MISCOMPARE = 0x0e
>> +};
>> +
>> +struct scsi_sense {
>> + unsigned char error_type;
>> + enum scsi_sense_key key;
>> + int ascq;
>> + unsigned sense_specific:1;
>> + unsigned ill_param_in_cdb:1;
>> + unsigned bit_pointer_valid:1;
>> + unsigned char bit_pointer;
>> + uint16_t field_pointer;
>> +};
>> +
>> +enum scsi_residual {
>> + SCSI_RESIDUAL_NO_RESIDUAL = 0,
>> + SCSI_RESIDUAL_UNDERFLOW,
>> + SCSI_RESIDUAL_OVERFLOW
>> +};
>> +
>> +struct scsi_task {
>> + int status;
>> + int cdb_size;
>> + int xfer_dir;
>> + int expxferlen;
>> + unsigned char cdb[SCSI_CDB_MAX_SIZE];
>> + enum scsi_residual residual_status;
>> + size_t residual;
>> + struct scsi_sense sense;
>> + struct scsi_data datain;
>> + struct scsi_allocated_memory *mem;
>> + void *ptr;
>> +
>> + uint32_t itt;
>> + uint32_t cmdsn;
>> + uint32_t lun;
>> +
>> + struct scsi_iovector iovector_in;
>> + struct scsi_iovector iovector_out;
>> +};
>> +
>> +/** libiscsi integration **/
>> +
>> +static int iscsi_add_lun(iscsi_lun_t *lun, char *iscsi_uri) {
>> + struct iscsi_url *iscsi_url;
>> + struct iscsi_context *iscsi_ctx;
>> + int ret = 0;
>> +
>> + assert(lun);
>> +
>> + iscsi_ctx = iscsi_create_context(ISCSI_INITIATOR);
>> + if (!iscsi_ctx) {
>> + PERR("Unable to create iSCSI context");
>> + return -1;
>> + }
>> +
>> + iscsi_url = iscsi_parse_full_url(iscsi_ctx, iscsi_uri);
>> + if (!iscsi_url) {
>> + PERR("Unable to parse iSCSI URL: %s", iscsi_get_error(iscsi_ctx));
>> + goto fail;
>> + }
>> +
>> + iscsi_set_session_type(iscsi_ctx, ISCSI_SESSION_NORMAL);
>> + iscsi_set_header_digest(iscsi_ctx, ISCSI_HEADER_DIGEST_NONE_CRC32C);
>> + if (iscsi_full_connect_sync(iscsi_ctx, iscsi_url->portal, iscsi_url->lun)) {
>> + PERR("Unable to login to iSCSI portal: %s", iscsi_get_error(iscsi_ctx));
>> + goto fail;
>> + }
>> +
>> + lun->iscsi_ctx = iscsi_ctx;
>> + lun->iscsi_lun = iscsi_url->lun;
>> +
>> + PDBG("Context %p created for lun 0: %s", iscsi_ctx, iscsi_uri);
>> +
>> +out:
>> + if (iscsi_url) {
>> + iscsi_destroy_url(iscsi_url);
>> + }
>> + return ret;
>> +
>> +fail:
>> + (void)iscsi_destroy_context(iscsi_ctx);
>> + ret = -1;
>> + goto out;
>> +}
>> +
>> +static struct scsi_task *scsi_task_new(int cdb_len, uint8_t *cdb, int dir,
>> + int xfer_len) {
>> + struct scsi_task *task;
>> +
>> + assert(cdb_len > 0);
>> + assert(cdb);
>> +
>> + task = calloc(1, sizeof(struct scsi_task));
>> + if (!task) {
>> + PERR("Error allocating task: %s", strerror(errno));
>> + return NULL;
>> + }
>> +
>> + memcpy(task->cdb, cdb, cdb_len);
>> + task->cdb_size = cdb_len;
>> + task->xfer_dir = dir;
>> + task->expxferlen = xfer_len;
>> +
>> + return task;
>> +}
>> +
>> +static int get_cdb_len(uint8_t *cdb) {
>> + switch(cdb[0] >> 5){
>> + case 0:
>> + return 6;
>> + case 1:
>> + case 2:
>> + return 10;
>> + case 4:
>> + return 16;
>> + case 5:
>> + return 12;
>> + }
>> + PERR("Unable to determine cdb len (0x%02hhX)", cdb[0]>>5);
>> + return -1;
>> +}
>> +
>> +static int handle_cmd_sync(struct iscsi_context *ctx,
>> + VirtIOSCSICmdReq *req,
>> + struct iovec *out, unsigned int out_len,
>> + VirtIOSCSICmdResp *rsp,
>> + struct iovec *in, unsigned int in_len) {
>> + struct scsi_task *task;
>> + uint32_t dir;
>> + uint32_t len;
>> + int cdb_len;
>> + int i;
>> +
>> + if (!((!req->lun[1]) && (req->lun[2] == 0x40) && (!req->lun[3]))) {
>> + // Ignore anything different than target=0, lun=0
>> + PDBG("Ignoring unconnected lun (0x%hhX, 0x%hhX)",
>> + req->lun[1], req->lun[3]);
>> + rsp->status = SCSI_STATUS_CHECK_CONDITION;
>> + memset(rsp->sense, 0, sizeof(rsp->sense));
>> + rsp->sense_len = 18;
>> + rsp->sense[0] = 0x70;
>> + rsp->sense[2] = 0x05; // ILLEGAL_REQUEST
>> + rsp->sense[7] = 10;
>> + rsp->sense[12] = 0x24;
>> +
>> + return 0;
>> + }
>> +
>> + cdb_len = get_cdb_len(req->cdb);
>> + if (cdb_len == -1) {
>> + return -1;
>> + }
>> +
>> + len = 0;
>> + if (!out_len && !in_len) {
>> + dir = SCSI_XFER_NONE;
>> + } else if (out_len) {
>> + dir = SCSI_XFER_TO_DEV;
>> + for (i=0; i<out_len; i++) {
>> + len += out[i].iov_len;
>> + }
>> + } else {
>> + dir = SCSI_XFER_FROM_DEV;
>> + for (i=0; i<in_len; i++) {
>> + len += in[i].iov_len;
>> + }
>> + }
>> +
>> + task = scsi_task_new(cdb_len, req->cdb, dir, len);
>> + if (!task) {
>> + PERR("Unable to create iscsi task");
>> + return -1;
>> + }
>> +
>> + if (dir == SCSI_XFER_TO_DEV) {
>> + task->iovector_out.iov = (struct scsi_iovec *)out;
>> + task->iovector_out.niov = out_len;
>> + } else if (dir == SCSI_XFER_FROM_DEV) {
>> + task->iovector_in.iov = (struct scsi_iovec *)in;
>> + task->iovector_in.niov = in_len;
>> + }
>> +
>> + PDBG("Sending iscsi cmd (cdb_len=%d, dir=%d, task=%p)",
>> + cdb_len, dir, task);
>> + if (!iscsi_scsi_command_sync(ctx, 0, task, NULL)) {
>> + PERR("Error serving SCSI command");
>> + free(task);
>> + return -1;
>> + }
>> +
>> + memset(rsp, 0, sizeof(*rsp));
>> +
>> + rsp->status = task->status;
>> + rsp->resid = task->residual;
>> +
>> + if (task->status == SCSI_STATUS_CHECK_CONDITION) {
>> + rsp->response = VIRTIO_SCSI_S_FAILURE;
>> + rsp->sense_len = task->datain.size - 2;
>> + memcpy(rsp->sense, &task->datain.data[2], rsp->sense_len);
>> + }
>> +
>> + free(task);
>> +
>> + PDBG("Filled in rsp: status=%hhX, resid=%u, response=%hhX, sense_len=%u",
>> + rsp->status, rsp->resid, rsp->response, rsp->sense_len);
>> +
>> + return 0;
>> +}
>> +
>> +/** libvhost-user callbacks **/
>> +
>> +static void vus_panic_cb(VuDev *vu_dev, const char *buf) {
>> + vhost_scsi_dev_t *vdev_scsi;
>> +
>> + assert(vu_dev);
>> +
>> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
>> +
>> + if (buf) {
>> + PERR("vu_panic: %s", buf);
>> + }
>> +
>> + if (vdev_scsi) {
>> + vdev_scsi->sched.quit = 1;
>> + }
>> +}
>> +
>> +static void vus_add_watch_cb(VuDev *vu_dev, int fd, int vu_evt, vu_watch_cb cb,
>> + void *pvt) {
>> + vhost_scsi_dev_t *vdev_scsi;
>> + int poll_evt = 0;
>> +
>> + assert(vu_dev);
>> + assert(fd >= 0);
>> + assert(cb);
>> +
>> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
>> + if (!vdev_scsi) {
>> + vus_panic_cb(vu_dev, NULL);
>> + }
>> +
>> + /* TODO: VU_WATCH_* should match POLL*, check it */
>> + if (vu_evt & VU_WATCH_IN) poll_evt |= POLLIN;
>> + if (vu_evt & VU_WATCH_OUT) poll_evt |= POLLOUT;
>> + if (vu_evt & VU_WATCH_PRI) poll_evt |= POLLPRI;
>> + if (vu_evt & VU_WATCH_ERR) poll_evt |= POLLERR;
>> + if (vu_evt & VU_WATCH_HUP) poll_evt |= POLLHUP;
>> +
>> + if (sched_add(&vdev_scsi->sched, fd, poll_evt, cb, NULL, pvt)) {
>> + vus_panic_cb(vu_dev, NULL);
>> + }
>> +}
>> +
>> +static void vus_del_watch_cb(VuDev *vu_dev, int fd) {
>> + vhost_scsi_dev_t *vdev_scsi;
>> +
>> + assert(vu_dev);
>> + assert(fd >= 0);
>> +
>> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
>> + if (!vdev_scsi) {
>> + vus_panic_cb(vu_dev, NULL);
>> + return;
>> + }
>> +
>> + if (sched_del(&vdev_scsi->sched, fd)) {
>> + vus_panic_cb(vu_dev, NULL);
>> + }
>> +}
>> +
>> +static void vus_proc_ctl(VuDev *vu_dev, int idx) {
>> + /* Control VQ not implemented */
>> +}
>> +
>> +static void vus_proc_evt(VuDev *vu_dev, int idx) {
>> + /* Event VQ not implemented */
>> +}
>> +
>> +static void vus_proc_req(VuDev *vu_dev, int idx) {
>> + vhost_scsi_dev_t *vdev_scsi;
>> + VuVirtq *vq;
>> +
>> + assert(vu_dev);
>> +
>> + vdev_scsi = vdev_scsi_find_by_vu(vu_dev);
>> + if (!vdev_scsi) {
>> + vus_panic_cb(vu_dev, NULL);
>> + return;
>> + }
>> +
>> + if ((idx < 0) || (idx >= VHOST_MAX_NR_VIRTQUEUE)) {
>> + PERR("VQ Index out of range: %d", idx);
>> + vus_panic_cb(vu_dev, NULL);
>> + return;
>> + }
>> +
>> + vq = vu_get_queue(vu_dev, idx);
>> + if (!vq) {
>> + PERR("Error fetching VQ (dev=%p, idx=%d)", vu_dev, idx);
>> + vus_panic_cb(vu_dev, NULL);
>> + return;
>> + }
>> +
>> + PDBG("Got kicked on vq[%d]@%p", idx, vq);
>> +
>> + while(1) {
>> + VuVirtqElement *elem;
>> + VirtIOSCSICmdReq *req;
>> + VirtIOSCSICmdResp *rsp;
>> +
>> + elem = vu_queue_pop(vu_dev, vq, sizeof(VuVirtqElement));
>> + if (!elem) {
>> + PDBG("No more elements pending on vq[%d]@%p", idx, vq);
>> + break;
>> + }
>> + PDBG("Popped elem@%p", elem);
>> +
>> + assert(!((elem->out_num > 1) && (elem->in_num > 1)));
>> + assert((elem->out_num > 0) && (elem->in_num > 0));
>> +
>> + if (elem->out_sg[0].iov_len < sizeof(VirtIOSCSICmdReq)) {
>> + PERR("Invalid virtio-scsi req header");
>> + vus_panic_cb(vu_dev, NULL);
>> + break;
>> + }
>> + req = (VirtIOSCSICmdReq *)elem->out_sg[0].iov_base;
>> +
>> + if (elem->in_sg[0].iov_len < sizeof(VirtIOSCSICmdResp)) {
>> + PERR("Invalid virtio-scsi rsp header");
>> + vus_panic_cb(vu_dev, NULL);
>> + break;
>> + }
>> + rsp = (VirtIOSCSICmdResp *)elem->in_sg[0].iov_base;
>> +
>> + if (handle_cmd_sync(vdev_scsi->luns[0].iscsi_ctx,
>> + req, &elem->out_sg[1], elem->out_num-1,
>> + rsp, &elem->in_sg[1], elem->in_num-1) != 0) {
>> + vus_panic_cb(vu_dev, NULL);
>> + break;
>> + }
>> +
>> + vu_queue_push(vu_dev, vq, elem, 0);
>> + vu_queue_notify(vu_dev, vq);
>> +
>> + free(elem);
>> + }
>> +
>> +}
>> +
>> +static void vus_queue_set_started(VuDev *vu_dev, int idx, bool started) {
>> + VuVirtq *vq;
>> +
>> + assert(vu_dev);
>> +
>> + if ((idx < 0) || (idx >= VHOST_MAX_NR_VIRTQUEUE)) {
>> + PERR("VQ Index out of range: %d", idx);
>> + vus_panic_cb(vu_dev, NULL);
>> + return;
>> + }
>> +
>> + vq = vu_get_queue(vu_dev, idx);
>> +
>> + switch(idx) {
>> + case 0:
>> + vu_set_queue_handler(vu_dev, vq, started?vus_proc_ctl:NULL);
>> + break;
>> + case 1:
>> + vu_set_queue_handler(vu_dev, vq, started?vus_proc_evt:NULL);
>> + break;
>> + default:
>> + vu_set_queue_handler(vu_dev, vq, started?vus_proc_req:NULL);
>> + }
>> +}
>> +
>> +static const VuDevIface vus_iface = {
>> + .queue_set_started = vus_queue_set_started,
>> +};
>> +
>> +static void vus_vhost_cb(VuDev *vu_dev, int vu_evt, void *data) {
>> + assert(vu_dev);
>> +
>> + if (!vu_dispatch(vu_dev) != 0) {
>> + PERR("Error processing vhost message");
>> + vus_panic_cb(vu_dev, NULL);
>> + }
>> +}
>> +
>> +/** util **/
>> +
>> +static int unix_sock_new(char *unix_fn) {
>> + int sock;
>> + struct sockaddr_un un;
>> + size_t len;
>> +
>> + assert(unix_fn);
>> +
>> + sock = socket(AF_UNIX, SOCK_STREAM, 0);
>> + if (sock <= 0) {
>> + perror("socket");
>> + return -1;
>> + }
>> +
>> + un.sun_family = AF_UNIX;
>> + (void)snprintf(un.sun_path, sizeof(un.sun_path), "%s", unix_fn);
>> + len = sizeof(un.sun_family) + strlen(un.sun_path);
>> +
>> + (void)unlink(unix_fn);
>> + if (bind(sock, (struct sockaddr *)&un, len) < 0) {
>> + perror("bind");
>> + goto fail;
>> + }
>> +
>> + if (listen(sock, 1) < 0) {
>> + perror("listen");
>> + goto fail;
>> + }
>> +
>> + return sock;
>> +
>> +fail:
>> + (void)close(sock);
>> +
>> + return -1;
>> +}
>> +
>> +/** vhost-user-scsi **/
>> +
>> +static vhost_scsi_dev_t *vdev_scsi_find_by_vu(VuDev *vu_dev) {
>> + int i;
>> +
>> + assert(vu_dev);
>> +
>> + for (i=0; i<VDEV_SCSI_MAX_DEVS; i++) {
>> + if (&vhost_scsi_devs[i]->vu_dev == vu_dev) {
>> + return vhost_scsi_devs[i];
>> + }
>> + }
>> +
>> + PERR("Unknown VuDev %p", vu_dev);
>> + return NULL;
>> +}
>> +
>> +static void vdev_scsi_deinit(vhost_scsi_dev_t *vdev_scsi) {
>> + if (!vdev_scsi) {
>> + return;
>> + }
>> +
>> + if (vdev_scsi->server_sock >= 0) {
>> + struct sockaddr_storage ss;
>> + socklen_t sslen = sizeof(ss);
>> +
>> + if (getsockname(vdev_scsi->server_sock, (struct sockaddr *)&ss,
>> + &sslen) == 0) {
>> + struct sockaddr_un *su = (struct sockaddr_un *)&ss;
>> + (void)unlink(su->sun_path);
>> + }
>> +
>> + (void)close(vdev_scsi->server_sock);
>> + }
>> +}
>> +
>> +static vhost_scsi_dev_t *vdev_scsi_new(char *unix_fn) {
>> + vhost_scsi_dev_t *vdev_scsi;
>> +
>> + assert(unix_fn);
>> +
>> + vdev_scsi = calloc(1, sizeof(vhost_scsi_dev_t));
>> + if (!vdev_scsi) {
>> + perror("calloc");
>> + return NULL;
>> + }
>> +
>> + vdev_scsi->server_sock = unix_sock_new(unix_fn);
>> + if (vdev_scsi->server_sock < 0) {
>> + free(vdev_scsi);
>> + return NULL;
>> + }
>> +
>> + vdev_scsi->sched.vu_dev = &vdev_scsi->vu_dev;
>> +
>> + return vdev_scsi;
>> +}
>> +
>> +static int vdev_scsi_iscsi_add_lun(vhost_scsi_dev_t *vdev_scsi,
>> + char *iscsi_uri, uint32_t lun) {
>> + assert(vdev_scsi);
>> + assert(iscsi_uri);
>> + assert(lun < VDEV_SCSI_MAX_LUNS);
>> +
>> + if (vdev_scsi->luns[lun].iscsi_ctx) {
>> + PERR("Lun %d already configured", lun);
>> + return -1;
>> + }
>> +
>> + if (iscsi_add_lun(&vdev_scsi->luns[lun], iscsi_uri) != 0) {
>> + return -1;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int vdev_scsi_run(vhost_scsi_dev_t *vdev_scsi) {
>> + int cli_sock;
>> + int ret = 0;
>> +
>> + assert(vdev_scsi);
>> + assert(vdev_scsi->server_sock >= 0);
>> +
>> + cli_sock = accept(vdev_scsi->server_sock, (void *)0, (void *)0);
>> + if (cli_sock < 0) {
>> + perror("accept");
>> + return -1;
>> + }
>> +
>> + vu_init(&vdev_scsi->vu_dev,
>> + cli_sock,
>> + vus_panic_cb,
>> + vus_add_watch_cb,
>> + vus_del_watch_cb,
>> + &vus_iface);
>> +
>> + ret = sched_add(&vdev_scsi->sched, cli_sock, POLLIN, vus_vhost_cb, NULL, 0);
>> + if (ret) {
>> + goto fail;
>> + }
>> +
>> + if (sched_loop(&vdev_scsi->sched) != 0) {
>> + goto fail;
>> + }
>> +
>> +out:
>> + vu_deinit(&vdev_scsi->vu_dev);
>> +
>> + return ret;
>> +
>> +fail:
>> + ret = -1;
>> + goto out;
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> + vhost_scsi_dev_t *vdev_scsi = NULL;
>> + char *unix_fn = NULL;
>> + char *iscsi_uri = NULL;
>> + int opt, err = EXIT_SUCCESS;
>> +
>> + while ((opt = getopt(argc, argv, "u:i:")) != -1) {
>> + switch (opt) {
>> + case 'h':
>> + goto help;
>> + case 'u':
>> + unix_fn = strdup(optarg);
>> + break;
>> + case 'i':
>> + iscsi_uri = strdup(optarg);
>> + break;
>> + default:
>> + goto help;
>> + }
>> + }
>> + if (!unix_fn || !iscsi_uri) {
>> + goto help;
>> + }
>> +
>> + vdev_scsi = vdev_scsi_new(unix_fn);
>> + if (!vdev_scsi) {
>> + goto err;
>> + }
>> + vhost_scsi_devs[0] = vdev_scsi;
>> +
>> + if (vdev_scsi_iscsi_add_lun(vdev_scsi, iscsi_uri, 0) != 0) {
>> + goto err;
>> + }
>> +
>> + if (vdev_scsi_run(vdev_scsi) != 0) {
>> + goto err;
>> + }
>> +
>> +out:
>> + if (vdev_scsi) {
>> + vdev_scsi_deinit(vdev_scsi);
>> + free(vdev_scsi);
>> + }
>> + if (unix_fn) {
>> + free(unix_fn);
>> + }
>> + if (iscsi_uri) {
>> + free(iscsi_uri);
>> + }
>> +
>> + return err;
>> +
>> +err:
>> + err = EXIT_FAILURE;
>> + goto out;
>> +
>> +help:
>> + fprintf(stderr, "Usage: %s [ -u unix_sock_path -i iscsi_uri ] | [ -h ]\n",
>> + argv[0]);
>> + fprintf(stderr, " -u path to unix socket\n");
>> + fprintf(stderr, " -i iscsi uri for lun 0\n");
>> + fprintf(stderr, " -h print help and quit\n");
>> +
>> + goto err;
>> +}
>>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application
2016-10-27 12:48 ` Felipe Franciosi
@ 2016-10-27 12:58 ` Paolo Bonzini
0 siblings, 0 replies; 7+ messages in thread
From: Paolo Bonzini @ 2016-10-27 12:58 UTC (permalink / raw)
To: Felipe Franciosi
Cc: Stefan Hajnoczi, Marc-Andre Lureau, Michael S. Tsirkin,
qemu-devel, Victor Kaplansky
On 27/10/2016 14:48, Felipe Franciosi wrote:
> Hello,
>
>> On 27 Oct 2016, at 13:16, Paolo Bonzini <pbonzini@redhat.com> wrote:
>>
>>
>>
>> On 26/10/2016 17:26, Felipe Franciosi wrote:
>>> This commit introduces a vhost-user-scsi backend sample application. It
>>> must be linked with libiscsi and libvhost-user.
>>>
>>> To use it, compile with:
>>> make tests/vhost-user-scsi
>>>
>>> And run as follows:
>>> tests/vhost-user-scsi -u /tmp/vus.sock -i iscsi://uri_to_target/
>>>
>>> The application is currently limited at one LUN only and it processes
>>> requests synchronously (therefore only achieving QD1). The purpose of
>>> the code is to show how a backend can be implemented and to test the
>>> vhost-user-scsi Qemu implementation.
>>>
>>> If a different instance of this vhost-user-scsi application is executed
>>> at a remote host, a VM can be live migrated to such a host.
>>
>> Hi,
>>
>> the right directory for this is contrib/.
>
> Cool. I was following suit from vhost-user-bridge which lives in
> tests/ today. To me, it makes more sense for these to be in contrib/.
> I'll place my sample application there for v2 and perhaps we should move
> vhost-user-bridge later?
Yes, that would make sense. Adding Victor in Cc.
Paolo
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2016-10-27 12:58 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-10-26 15:26 [Qemu-devel] [PATCH 0/2] Introduce vhost-user-scsi and sample application Felipe Franciosi
2016-10-26 15:26 ` [Qemu-devel] [PATCH 1/2] vus: Introduce vhost-user-scsi host device Felipe Franciosi
2016-10-27 12:12 ` Paolo Bonzini
2016-10-26 15:26 ` [Qemu-devel] [PATCH 2/2] vus: Introduce a vhost-user-scsi sample application Felipe Franciosi
2016-10-27 12:16 ` Paolo Bonzini
2016-10-27 12:48 ` Felipe Franciosi
2016-10-27 12:58 ` Paolo Bonzini
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.