From f3a8505656935cde32e28c1c6317f725084da1e0 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:48:55 +0100 Subject: [PATCH 01/14] qdev/qbus: add hidden device support This adds support for hiding a device to the qbus and qdev APIs. The first user of this will be the virtio-net failover feature but the API introduced with this patch could be used to implement other features as well, for example hiding pci devices when a pci bus is powered off. qdev_device_add() is modified to check for a failover_pair_id argument in the option string. A DeviceListener callback should_be_hidden() is added. It can be used by a standby device to inform qdev that this device should not be added now. The standby device handler can store the device options to plug the device in at a later point in time. One reason for hiding the device is that we don't want to expose both devices to the guest kernel until the respective virtio feature bit VIRTIO_NET_F_STANDBY was negotiated and we know that the devices will be handled correctly by the guest. More information on the kernel feature this is using: https://www.kernel.org/doc/html/latest/networking/net_failover.html An example where the primary device is a vfio-pci device and the standby device is a virtio-net device: A device is hidden when it has an "failover_pair_id" option, e.g. -device virtio-net-pci,...,failover=on,... -device vfio-pci,...,failover_pair_id=net1,... Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-2-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/core/qdev.c | 24 ++++++++++++++++++++++++ include/hw/qdev-core.h | 29 +++++++++++++++++++++++++++++ qdev-monitor.c | 41 +++++++++++++++++++++++++++++++++++++---- vl.c | 6 ++++-- 4 files changed, 94 insertions(+), 6 deletions(-) diff --git a/hw/core/qdev.c b/hw/core/qdev.c index cbad6c1d55..3b8d43d0fd 100644 --- a/hw/core/qdev.c +++ b/hw/core/qdev.c @@ -212,6 +212,30 @@ void device_listener_unregister(DeviceListener *listener) QTAILQ_REMOVE(&device_listeners, listener, link); } +bool qdev_should_hide_device(QemuOpts *opts) +{ + int rc = -1; + DeviceListener *listener; + + QTAILQ_FOREACH(listener, &device_listeners, link) { + if (listener->should_be_hidden) { + /* + * should_be_hidden_will return + * 1 if device matches opts and it should be hidden + * 0 if device matches opts and should not be hidden + * -1 if device doesn't match ops + */ + rc = listener->should_be_hidden(listener, opts); + } + + if (rc > 0) { + break; + } + } + + return rc > 0; +} + void qdev_set_legacy_instance_id(DeviceState *dev, int alias_id, int required_for_version) { diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h index aa123f88cb..710981af36 100644 --- a/include/hw/qdev-core.h +++ b/include/hw/qdev-core.h @@ -78,6 +78,19 @@ typedef void (*BusUnrealize)(BusState *bus, Error **errp); * respective parent types. * * + * + * # Hiding a device # + * To hide a device, a DeviceListener function should_be_hidden() needs to + * be registered. + * It can be used to defer adding a device and therefore hide it from the + * guest. The handler registering to this DeviceListener can save the QOpts + * passed to it for re-using it later and must return that it wants the device + * to be/remain hidden or not. When the handler function decides the device + * shall not be hidden it will be added in qdev_device_add() and + * realized as any other device. Otherwise qdev_device_add() will return early + * without adding the device. The guest will not see a "hidden" device + * until it was marked don't hide and qdev_device_add called again. + * */ typedef struct DeviceClass { /*< private >*/ @@ -154,6 +167,12 @@ struct DeviceState { struct DeviceListener { void (*realize)(DeviceListener *listener, DeviceState *dev); void (*unrealize)(DeviceListener *listener, DeviceState *dev); + /* + * This callback is called upon init of the DeviceState and allows to + * inform qdev that a device should be hidden, depending on the device + * opts, for example, to hide a standby device. + */ + int (*should_be_hidden)(DeviceListener *listener, QemuOpts *device_opts); QTAILQ_ENTRY(DeviceListener) link; }; @@ -451,4 +470,14 @@ static inline bool qbus_is_hotpluggable(BusState *bus) void device_listener_register(DeviceListener *listener); void device_listener_unregister(DeviceListener *listener); +/** + * @qdev_should_hide_device: + * @opts: QemuOpts as passed on cmdline. + * + * Check if a device should be added. + * When a device is added via qdev_device_add() this will be called, + * and return if the device should be added now or not. + */ +bool qdev_should_hide_device(QemuOpts *opts); + #endif diff --git a/qdev-monitor.c b/qdev-monitor.c index 148df9cacf..ffa08c670f 100644 --- a/qdev-monitor.c +++ b/qdev-monitor.c @@ -32,9 +32,11 @@ #include "qemu/help_option.h" #include "qemu/option.h" #include "qemu/qemu-print.h" +#include "qemu/option_int.h" #include "sysemu/block-backend.h" #include "sysemu/sysemu.h" #include "migration/misc.h" +#include "migration/migration.h" /* * Aliases were a bad idea from the start. Let's keep them @@ -562,13 +564,36 @@ void qdev_set_id(DeviceState *dev, const char *id) } } +static int is_failover_device(void *opaque, const char *name, const char *value, + Error **errp) +{ + if (strcmp(name, "failover_pair_id") == 0) { + QemuOpts *opts = (QemuOpts *)opaque; + + if (qdev_should_hide_device(opts)) { + return 1; + } + } + + return 0; +} + +static bool should_hide_device(QemuOpts *opts) +{ + if (qemu_opt_foreach(opts, is_failover_device, opts, NULL) == 0) { + return false; + } + return true; +} + DeviceState *qdev_device_add(QemuOpts *opts, Error **errp) { DeviceClass *dc; const char *driver, *path; - DeviceState *dev; + DeviceState *dev = NULL; BusState *bus = NULL; Error *err = NULL; + bool hide; driver = qemu_opt_get(opts, "driver"); if (!driver) { @@ -602,11 +627,17 @@ DeviceState *qdev_device_add(QemuOpts *opts, Error **errp) return NULL; } } - if (qdev_hotplug && bus && !qbus_is_hotpluggable(bus)) { + hide = should_hide_device(opts); + + if ((hide || qdev_hotplug) && bus && !qbus_is_hotpluggable(bus)) { error_setg(errp, QERR_BUS_NO_HOTPLUG, bus->name); return NULL; } + if (hide) { + return NULL; + } + if (!migration_is_idle()) { error_setg(errp, "device_add not allowed while migrating"); return NULL; @@ -648,8 +679,10 @@ DeviceState *qdev_device_add(QemuOpts *opts, Error **errp) err_del_dev: error_propagate(errp, err); - object_unparent(OBJECT(dev)); - object_unref(OBJECT(dev)); + if (dev) { + object_unparent(OBJECT(dev)); + object_unref(OBJECT(dev)); + } return NULL; } diff --git a/vl.c b/vl.c index 49778e7399..c389d24b2c 100644 --- a/vl.c +++ b/vl.c @@ -2207,10 +2207,12 @@ static int device_init_func(void *opaque, QemuOpts *opts, Error **errp) DeviceState *dev; dev = qdev_device_add(opts, errp); - if (!dev) { + if (!dev && *errp) { + error_report_err(*errp); return -1; + } else if (dev) { + object_unref(OBJECT(dev)); } - object_unref(OBJECT(dev)); return 0; } From 4f5b6a05a4e76892d8d03838a36aca735e9bcefb Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:48:56 +0100 Subject: [PATCH 02/14] pci: add option for net failover This patch adds a failover_pair_id property to PCIDev which is used to link the primary device in a failover pair (the PCI dev) to a standby (a virtio-net-pci) device. It only supports ethernet devices. Also currently it only supports PCIe devices. The requirement for PCIe is because it doesn't support other hotplug controllers at the moment. The failover functionality can be added to other hotplug controllers like ACPI, SHCP,... later on. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-3-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/pci/pci.c | 31 +++++++++++++++++++++++++++++++ include/hw/pci/pci.h | 3 +++ 2 files changed, 34 insertions(+) diff --git a/hw/pci/pci.c b/hw/pci/pci.c index aa05c2b9b2..824ab4ed7b 100644 --- a/hw/pci/pci.c +++ b/hw/pci/pci.c @@ -75,6 +75,8 @@ static Property pci_props[] = { QEMU_PCIE_LNKSTA_DLLLA_BITNR, true), DEFINE_PROP_BIT("x-pcie-extcap-init", PCIDevice, cap_present, QEMU_PCIE_EXTCAP_INIT_BITNR, true), + DEFINE_PROP_STRING("failover_pair_id", PCIDevice, + failover_pair_id), DEFINE_PROP_END_OF_LIST() }; @@ -2077,6 +2079,7 @@ static void pci_qdev_realize(DeviceState *qdev, Error **errp) ObjectClass *klass = OBJECT_CLASS(pc); Error *local_err = NULL; bool is_default_rom; + uint16_t class_id; /* initialize cap_present for pci_is_express() and pci_config_size(), * Note that hybrid PCIs are not set automatically and need to manage @@ -2101,6 +2104,34 @@ static void pci_qdev_realize(DeviceState *qdev, Error **errp) } } + if (pci_dev->failover_pair_id) { + if (!pci_bus_is_express(pci_get_bus(pci_dev))) { + error_setg(errp, "failover primary device must be on " + "PCIExpress bus"); + error_propagate(errp, local_err); + pci_qdev_unrealize(DEVICE(pci_dev), NULL); + return; + } + class_id = pci_get_word(pci_dev->config + PCI_CLASS_DEVICE); + if (class_id != PCI_CLASS_NETWORK_ETHERNET) { + error_setg(errp, "failover primary device is not an " + "Ethernet device"); + error_propagate(errp, local_err); + pci_qdev_unrealize(DEVICE(pci_dev), NULL); + return; + } + if (!(pci_dev->cap_present & QEMU_PCI_CAP_MULTIFUNCTION) + && (PCI_FUNC(pci_dev->devfn) == 0)) { + qdev->allow_unplug_during_migration = true; + } else { + error_setg(errp, "failover: primary device must be in its own " + "PCI slot"); + error_propagate(errp, local_err); + pci_qdev_unrealize(DEVICE(pci_dev), NULL); + return; + } + } + /* rom loading */ is_default_rom = false; if (pci_dev->romfile == NULL && pc->romfile != NULL) { diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h index f3f0ffd5fb..69d1f0228b 100644 --- a/include/hw/pci/pci.h +++ b/include/hw/pci/pci.h @@ -352,6 +352,9 @@ struct PCIDevice { MSIVectorUseNotifier msix_vector_use_notifier; MSIVectorReleaseNotifier msix_vector_release_notifier; MSIVectorPollNotifier msix_vector_poll_notifier; + + /* ID of standby device in net_failover pair */ + char *failover_pair_id; }; void pci_register_bar(PCIDevice *pci_dev, int region_num, From a99c4da9fc2a39847cfebbc0c6802ee122dcca39 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:48:57 +0100 Subject: [PATCH 03/14] pci: mark devices partially unplugged Only the guest unplug request was triggered. This is needed for the failover feature. In case of a failed migration we need to plug the device back to the guest. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-4-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/pci/pcie.c | 3 +++ include/hw/pci/pci.h | 1 + 2 files changed, 4 insertions(+) diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c index a6beb567bd..19363ff8ce 100644 --- a/hw/pci/pcie.c +++ b/hw/pci/pcie.c @@ -456,6 +456,9 @@ static void pcie_unplug_device(PCIBus *bus, PCIDevice *dev, void *opaque) { HotplugHandler *hotplug_ctrl = qdev_get_hotplug_handler(DEVICE(dev)); + if (dev->partially_hotplugged) { + return; + } hotplug_handler_unplug(hotplug_ctrl, DEVICE(dev), &error_abort); object_unparent(OBJECT(dev)); } diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h index 69d1f0228b..db75c6dfd0 100644 --- a/include/hw/pci/pci.h +++ b/include/hw/pci/pci.h @@ -265,6 +265,7 @@ typedef struct PCIReqIDCache PCIReqIDCache; struct PCIDevice { DeviceState qdev; + bool partially_hotplugged; /* PCI config space */ uint8_t *config; From c000a9bd06ea79ea5c2c1f894a47410be48c0cc1 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:48:58 +0100 Subject: [PATCH 04/14] pci: mark device having guest unplug request pending Set pending_deleted_event in DeviceState for failover primary devices that were successfully unplugged by the Guest OS. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-5-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/pci/pcie.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c index 19363ff8ce..08718188bb 100644 --- a/hw/pci/pcie.c +++ b/hw/pci/pcie.c @@ -457,6 +457,7 @@ static void pcie_unplug_device(PCIBus *bus, PCIDevice *dev, void *opaque) HotplugHandler *hotplug_ctrl = qdev_get_hotplug_handler(DEVICE(dev)); if (dev->partially_hotplugged) { + dev->qdev.pending_deleted_event = false; return; } hotplug_handler_unplug(hotplug_ctrl, DEVICE(dev), &error_abort); @@ -476,6 +477,8 @@ void pcie_cap_slot_unplug_request_cb(HotplugHandler *hotplug_dev, return; } + dev->pending_deleted_event = true; + /* In case user cancel the operation of multi-function hot-add, * remove the function that is unexposed to guest individually, * without interaction with guest. From d328e6f37229e86a5e0059a98c4acfa9be9c8783 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:48:59 +0100 Subject: [PATCH 05/14] qapi: add unplug primary event This event is emitted when we sent a request to unplug a failover primary device from the Guest OS and it includes the device id of the primary device. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-6-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin Acked-by: Markus Armbruster --- qapi/migration.json | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/qapi/migration.json b/qapi/migration.json index 82feb5bd39..e9e7a97c03 100644 --- a/qapi/migration.json +++ b/qapi/migration.json @@ -1448,3 +1448,22 @@ # Since: 3.0 ## { 'command': 'migrate-pause', 'allow-oob': true } + +## +# @UNPLUG_PRIMARY: +# +# Emitted from source side of a migration when migration state is +# WAIT_UNPLUG. Device was unplugged by guest operating system. +# Device resources in QEMU are kept on standby to be able to re-plug it in case +# of migration failure. +# +# @device-id: QEMU device id of the unplugged device +# +# Since: 4.2 +# +# Example: +# {"event": "UNPLUG_PRIMARY", "data": {"device-id": "hostdev0"} } +# +## +{ 'event': 'UNPLUG_PRIMARY', + 'data': { 'device-id': 'str' } } From 70d04971f1ec71ba01c3dc5d4c760fa2c87e2612 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:49:00 +0100 Subject: [PATCH 06/14] qapi: add failover negotiated event This event is sent to let libvirt know that VIRTIO_NET_F_STANDBY feature is enabled. The primary device this virtio-net (standby) device is associated with, is now hotplugged by the virtio-net device. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-7-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin Acked-by: Markus Armbruster --- qapi/net.json | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/qapi/net.json b/qapi/net.json index 4c96137811..335295be50 100644 --- a/qapi/net.json +++ b/qapi/net.json @@ -735,3 +735,22 @@ ## { 'command': 'announce-self', 'boxed': true, 'data' : 'AnnounceParameters'} + +## +# @FAILOVER_NEGOTIATED: +# +# Emitted when VIRTIO_NET_F_STANDBY was enabled during feature negotiation. +# Failover primary devices which were hidden (not hotplugged when requested) +# before will now be hotplugged by the virtio-net standby device. +# +# device-id: QEMU device id of the unplugged device +# Since: 4.2 +# +# Example: +# +# <- { "event": "FAILOVER_NEGOTIATED", +# "data": "net1" } +# +## +{ 'event': 'FAILOVER_NEGOTIATED', + 'data': {'device-id': 'str'} } From a1190ab628c0e2816eae42786cd7396d6638aa48 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:49:01 +0100 Subject: [PATCH 07/14] migration: allow unplug during migration for failover devices In "b06424de62 migration: Disable hotplug/unplug during migration" we added a check to disable unplug for all devices until we have figured out what works. For failover primary devices qdev_unplug() is called from the migration handler, i.e. during migration. This patch adds a flag to DeviceState which is set to false for all devices and makes an exception for PCI devices that are also primary devices in a failover pair. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-8-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/core/qdev.c | 1 + hw/pci/pci.c | 1 + include/hw/qdev-core.h | 1 + qdev-monitor.c | 2 +- 4 files changed, 4 insertions(+), 1 deletion(-) diff --git a/hw/core/qdev.c b/hw/core/qdev.c index 3b8d43d0fd..cf1ba28fe3 100644 --- a/hw/core/qdev.c +++ b/hw/core/qdev.c @@ -996,6 +996,7 @@ static void device_initfn(Object *obj) dev->instance_id_alias = -1; dev->realized = false; + dev->allow_unplug_during_migration = false; object_property_add_bool(obj, "realized", device_get_realized, device_set_realized, NULL); diff --git a/hw/pci/pci.c b/hw/pci/pci.c index 824ab4ed7b..c68498c0de 100644 --- a/hw/pci/pci.c +++ b/hw/pci/pci.c @@ -2130,6 +2130,7 @@ static void pci_qdev_realize(DeviceState *qdev, Error **errp) pci_qdev_unrealize(DEVICE(pci_dev), NULL); return; } + qdev->allow_unplug_during_migration = true; } /* rom loading */ diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h index 710981af36..1518495b1e 100644 --- a/include/hw/qdev-core.h +++ b/include/hw/qdev-core.h @@ -156,6 +156,7 @@ struct DeviceState { bool pending_deleted_event; QemuOpts *opts; int hotplugged; + bool allow_unplug_during_migration; BusState *parent_bus; QLIST_HEAD(, NamedGPIOList) gpios; QLIST_HEAD(, BusState) child_bus; diff --git a/qdev-monitor.c b/qdev-monitor.c index ffa08c670f..e6b112eb0a 100644 --- a/qdev-monitor.c +++ b/qdev-monitor.c @@ -851,7 +851,7 @@ void qdev_unplug(DeviceState *dev, Error **errp) return; } - if (!migration_is_idle()) { + if (!migration_is_idle() && !dev->allow_unplug_during_migration) { error_setg(errp, "device_del not allowed while migrating"); return; } From c7e0acd5a3f87d4aabb7a702aff87f83fc3ce151 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:49:02 +0100 Subject: [PATCH 08/14] migration: add new migration state wait-unplug This patch adds a new migration state called wait-unplug. It is entered after the SETUP state if failover devices are present. It will transition into ACTIVE once all devices were succesfully unplugged from the guest. So if a guest doesn't respond or takes long to honor the unplug request the user will see the migration state 'wait-unplug'. In the migration thread we query failover devices if they're are still pending the guest unplug. When all are unplugged the migration continues. If one device won't unplug migration will stay in wait_unplug state. Signed-off-by: Jens Freimann Reviewed-by: Dr. David Alan Gilbert Message-Id: <20191029114905.6856-9-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- include/migration/vmstate.h | 2 ++ migration/migration.c | 21 +++++++++++++++++++++ migration/migration.h | 3 +++ migration/savevm.c | 31 +++++++++++++++++++++++++++++++ migration/savevm.h | 2 ++ qapi/migration.json | 5 ++++- 6 files changed, 63 insertions(+), 1 deletion(-) diff --git a/include/migration/vmstate.h b/include/migration/vmstate.h index b9ee563aa4..ac4f46a67d 100644 --- a/include/migration/vmstate.h +++ b/include/migration/vmstate.h @@ -186,6 +186,8 @@ struct VMStateDescription { int (*pre_save)(void *opaque); int (*post_save)(void *opaque); bool (*needed)(void *opaque); + bool (*dev_unplug_pending)(void *opaque); + const VMStateField *fields; const VMStateDescription **subsections; }; diff --git a/migration/migration.c b/migration/migration.c index 4133ed2684..354ad072fa 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -52,6 +52,7 @@ #include "hw/qdev-properties.h" #include "monitor/monitor.h" #include "net/announce.h" +#include "qemu/queue.h" #define MAX_THROTTLE (32 << 20) /* Migration transfer speed throttling */ @@ -819,6 +820,7 @@ bool migration_is_setup_or_active(int state) case MIGRATION_STATUS_SETUP: case MIGRATION_STATUS_PRE_SWITCHOVER: case MIGRATION_STATUS_DEVICE: + case MIGRATION_STATUS_WAIT_UNPLUG: return true; default: @@ -954,6 +956,9 @@ static void fill_source_migration_info(MigrationInfo *info) case MIGRATION_STATUS_CANCELLED: info->has_status = true; break; + case MIGRATION_STATUS_WAIT_UNPLUG: + info->has_status = true; + break; } info->status = s->state; } @@ -1694,6 +1699,7 @@ bool migration_is_idle(void) case MIGRATION_STATUS_COLO: case MIGRATION_STATUS_PRE_SWITCHOVER: case MIGRATION_STATUS_DEVICE: + case MIGRATION_STATUS_WAIT_UNPLUG: return false; case MIGRATION_STATUS__MAX: g_assert_not_reached(); @@ -3264,6 +3270,19 @@ static void *migration_thread(void *opaque) qemu_savevm_state_setup(s->to_dst_file); + if (qemu_savevm_nr_failover_devices()) { + migrate_set_state(&s->state, MIGRATION_STATUS_SETUP, + MIGRATION_STATUS_WAIT_UNPLUG); + + while (s->state == MIGRATION_STATUS_WAIT_UNPLUG && + qemu_savevm_state_guest_unplug_pending()) { + qemu_sem_timedwait(&s->wait_unplug_sem, 250); + } + + migrate_set_state(&s->state, MIGRATION_STATUS_WAIT_UNPLUG, + MIGRATION_STATUS_ACTIVE); + } + s->setup_time = qemu_clock_get_ms(QEMU_CLOCK_HOST) - setup_start; migrate_set_state(&s->state, MIGRATION_STATUS_SETUP, MIGRATION_STATUS_ACTIVE); @@ -3511,6 +3530,7 @@ static void migration_instance_finalize(Object *obj) qemu_mutex_destroy(&ms->qemu_file_lock); g_free(params->tls_hostname); g_free(params->tls_creds); + qemu_sem_destroy(&ms->wait_unplug_sem); qemu_sem_destroy(&ms->rate_limit_sem); qemu_sem_destroy(&ms->pause_sem); qemu_sem_destroy(&ms->postcopy_pause_sem); @@ -3556,6 +3576,7 @@ static void migration_instance_init(Object *obj) qemu_sem_init(&ms->postcopy_pause_rp_sem, 0); qemu_sem_init(&ms->rp_state.rp_sem, 0); qemu_sem_init(&ms->rate_limit_sem, 0); + qemu_sem_init(&ms->wait_unplug_sem, 0); qemu_mutex_init(&ms->qemu_file_lock); } diff --git a/migration/migration.h b/migration/migration.h index 4f2fe193dc..79b3dda146 100644 --- a/migration/migration.h +++ b/migration/migration.h @@ -206,6 +206,9 @@ struct MigrationState /* Flag set once the migration thread called bdrv_inactivate_all */ bool block_inactive; + /* Migration is waiting for guest to unplug device */ + QemuSemaphore wait_unplug_sem; + /* Migration is paused due to pause-before-switchover */ QemuSemaphore pause_sem; diff --git a/migration/savevm.c b/migration/savevm.c index 8d95e261f6..966a9c3bdb 100644 --- a/migration/savevm.c +++ b/migration/savevm.c @@ -1113,6 +1113,37 @@ void qemu_savevm_state_header(QEMUFile *f) } } +int qemu_savevm_nr_failover_devices(void) +{ + SaveStateEntry *se; + int n = 0; + + QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { + if (se->vmsd && se->vmsd->dev_unplug_pending) { + n++; + } + } + + return n; +} + +bool qemu_savevm_state_guest_unplug_pending(void) +{ + SaveStateEntry *se; + int n = 0; + + QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { + if (!se->vmsd || !se->vmsd->dev_unplug_pending) { + continue; + } + if (se->vmsd->dev_unplug_pending(se->opaque)) { + n++; + } + } + + return n > 0; +} + void qemu_savevm_state_setup(QEMUFile *f) { SaveStateEntry *se; diff --git a/migration/savevm.h b/migration/savevm.h index 51a4b9caa8..c42b9c80ee 100644 --- a/migration/savevm.h +++ b/migration/savevm.h @@ -31,6 +31,8 @@ bool qemu_savevm_state_blocked(Error **errp); void qemu_savevm_state_setup(QEMUFile *f); +int qemu_savevm_nr_failover_devices(void); +bool qemu_savevm_state_guest_unplug_pending(void); int qemu_savevm_state_resume_prepare(MigrationState *s); void qemu_savevm_state_header(QEMUFile *f); int qemu_savevm_state_iterate(QEMUFile *f, bool postcopy); diff --git a/qapi/migration.json b/qapi/migration.json index e9e7a97c03..b7348d0c8b 100644 --- a/qapi/migration.json +++ b/qapi/migration.json @@ -133,6 +133,9 @@ # @device: During device serialisation when pause-before-switchover is enabled # (since 2.11) # +# @wait-unplug: wait for device unplug request by guest OS to be completed. +# (since 4.2) +# # Since: 2.3 # ## @@ -140,7 +143,7 @@ 'data': [ 'none', 'setup', 'cancelling', 'cancelled', 'active', 'postcopy-active', 'postcopy-paused', 'postcopy-recover', 'completed', 'failed', 'colo', - 'pre-switchover', 'device' ] } + 'pre-switchover', 'device', 'wait-unplug' ] } ## # @MigrationInfo: From ea45cb8d84d25c3a3103932def12558de1bbe208 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:49:03 +0100 Subject: [PATCH 09/14] libqos: tolerate wait-unplug migration state Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-10-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- tests/libqos/libqos.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tests/libqos/libqos.c b/tests/libqos/libqos.c index d71557c5cb..f229eb2cb8 100644 --- a/tests/libqos/libqos.c +++ b/tests/libqos/libqos.c @@ -125,7 +125,8 @@ void migrate(QOSState *from, QOSState *to, const char *uri) break; } - if ((strcmp(st, "setup") == 0) || (strcmp(st, "active") == 0)) { + if ((strcmp(st, "setup") == 0) || (strcmp(st, "active") == 0) + || (strcmp(st, "wait-unplug") == 0)) { qobject_unref(rsp); g_usleep(5000); continue; From 9711cd0dfc3fa414f7f64935713c07134ae67971 Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:49:04 +0100 Subject: [PATCH 10/14] net/virtio: add failover support This patch adds support to handle failover device pairs of a virtio-net device and a (vfio-)pci device, where the virtio-net acts as the standby device and the (vfio-)pci device as the primary. The general idea is that we have a pair of devices, a (vfio-)pci and a emulated (virtio-net) device. Before migration the vfio device is unplugged and data flows to the emulated device, on the target side another (vfio-)pci device is plugged in to take over the data-path. In the guest the net_failover module will pair net devices with the same MAC address. To achieve this we need: 1. Provide a callback function for the should_be_hidden DeviceListener. It is called when the primary device is plugged in. Evaluate the QOpt passed in to check if it is the matching primary device. It returns if the device should be hidden or not. When it should be hidden it stores the device options in the VirtioNet struct and the device is added once the VIRTIO_NET_F_STANDBY feature is negotiated during virtio feature negotiation. If the virtio-net devices are not realized at the time the (vfio-)pci devices are realized, we need to connect the devices later. This way we make sure primary and standby devices can be specified in any order. 2. Register a callback for migration status notifier. When called it will unplug its primary device before the migration happens. 3. Register a callback for the migration code that checks if a device needs to be unplugged from the guest. Signed-off-by: Jens Freimann Message-Id: <20191029114905.6856-11-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- MAINTAINERS | 1 + docs/virtio-net-failover.rst | 68 ++++++++ hw/net/virtio-net.c | 302 +++++++++++++++++++++++++++++++++ include/hw/virtio/virtio-net.h | 12 ++ include/hw/virtio/virtio.h | 1 + 5 files changed, 384 insertions(+) create mode 100644 docs/virtio-net-failover.rst diff --git a/MAINTAINERS b/MAINTAINERS index 325e67a04e..c958fc71ea 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1435,6 +1435,7 @@ S: Odd Fixes F: hw/net/ F: include/hw/net/ F: tests/virtio-net-test.c +F: docs/virtio-net-failover.rst T: git https://github.com/jasowang/qemu.git net Parallel NOR Flash devices diff --git a/docs/virtio-net-failover.rst b/docs/virtio-net-failover.rst new file mode 100644 index 0000000000..22f64c7bc8 --- /dev/null +++ b/docs/virtio-net-failover.rst @@ -0,0 +1,68 @@ +======================== +QEMU virtio-net standby (net_failover) +======================== + +This document explains the setup and usage of virtio-net standby feature which +is used to create a net_failover pair of devices. + +The general idea is that we have a pair of devices, a (vfio-)pci and a +virtio-net device. Before migration the vfio device is unplugged and data flows +through the virtio-net device, on the target side another vfio-pci device is +plugged in to take over the data-path. In the guest the net_failover kernel +module will pair net devices with the same MAC address. + +The two devices are called primary and standby device. The fast hardware based +networking device is called the primary device and the virtio-net device is the +standby device. + +Restrictions +------------ + +Currently only PCIe devices are allowed as primary devices, this restriction +can be lifted in the future with enhanced QEMU support. Also, only networking +devices are allowed as primary device. The user needs to ensure that primary +and standby devices are not plugged into the same PCIe slot. + +Usecase +------- + + Virtio-net standby allows easy migration while using a passed-through fast + networking device by falling back to a virtio-net device for the duration of + the migration. It is like a simple version of a bond, the difference is that it + requires no configuration in the guest. When a guest is live-migrated to + another host QEMU will unplug the primary device via the PCIe based hotplug + handler and traffic will go through the virtio-net device. On the target + system the primary device will be automatically plugged back and the + net_failover module registers it again as the primary device. + +Usage +----- + + The primary device can be hotplugged or be part of the startup configuration + + -device virtio-net-pci,netdev=hostnet1,id=net1,mac=52:54:00:6f:55:cc, \ + bus=root2,failover=on + + With the parameter failover=on the VIRTIO_NET_F_STANDBY feature will be enabled. + + -device vfio-pci,host=5e:00.2,id=hostdev0,bus=root1,failover_pair_id=net1 + + failover_pair_id references the id of the virtio-net standby device. This + is only for pairing the devices within QEMU. The guest kernel module + net_failover will match devices with identical MAC addresses. + +Hotplug +------- + + Both primary and standby device can be hotplugged via the QEMU monitor. Note + that if the virtio-net device is plugged first a warning will be issued that it + couldn't find the primary device. + +Migration +--------- + + A new migration state wait-unplug was added for this feature. If failover primary + devices are present in the configuration, migration will go into this state. + It will wait until the device unplug is completed in the guest and then move into + active state. On the target system the primary devices will be automatically hotplugged + when the feature bit was negotiated for the virtio-net standby device. diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c index d3d688a5f8..314f365e7c 100644 --- a/hw/net/virtio-net.c +++ b/hw/net/virtio-net.c @@ -12,6 +12,7 @@ */ #include "qemu/osdep.h" +#include "qemu/atomic.h" #include "qemu/iov.h" #include "qemu/main-loop.h" #include "qemu/module.h" @@ -21,6 +22,10 @@ #include "net/tap.h" #include "qemu/error-report.h" #include "qemu/timer.h" +#include "qemu/option.h" +#include "qemu/option_int.h" +#include "qemu/config-file.h" +#include "qapi/qmp/qdict.h" #include "hw/virtio/virtio-net.h" #include "net/vhost_net.h" #include "net/announce.h" @@ -28,11 +33,15 @@ #include "qapi/error.h" #include "qapi/qapi-events-net.h" #include "hw/qdev-properties.h" +#include "qapi/qapi-types-migration.h" +#include "qapi/qapi-events-migration.h" #include "hw/virtio/virtio-access.h" #include "migration/misc.h" #include "standard-headers/linux/ethtool.h" #include "sysemu/sysemu.h" #include "trace.h" +#include "monitor/qdev.h" +#include "hw/pci/pci.h" #define VIRTIO_NET_VM_VERSION 11 @@ -746,9 +755,99 @@ static inline uint64_t virtio_net_supported_guest_offloads(VirtIONet *n) return virtio_net_guest_offloads_by_features(vdev->guest_features); } +static void failover_add_primary(VirtIONet *n, Error **errp) +{ + Error *err = NULL; + + n->primary_device_opts = qemu_opts_find(qemu_find_opts("device"), + n->primary_device_id); + if (n->primary_device_opts) { + n->primary_dev = qdev_device_add(n->primary_device_opts, &err); + if (err) { + qemu_opts_del(n->primary_device_opts); + } + if (n->primary_dev) { + n->primary_bus = n->primary_dev->parent_bus; + if (err) { + qdev_unplug(n->primary_dev, &err); + qdev_set_id(n->primary_dev, ""); + + } + } + } else { + error_setg(errp, "Primary device not found"); + error_append_hint(errp, "Virtio-net failover will not work. Make " + "sure primary device has parameter" + " failover_pair_id=\n"); +} + if (err) { + error_propagate(errp, err); + } +} + +static int is_my_primary(void *opaque, QemuOpts *opts, Error **errp) +{ + VirtIONet *n = opaque; + int ret = 0; + + const char *standby_id = qemu_opt_get(opts, "failover_pair_id"); + + if (standby_id != NULL && (g_strcmp0(standby_id, n->netclient_name) == 0)) { + n->primary_device_id = g_strdup(opts->id); + ret = 1; + } + + return ret; +} + +static DeviceState *virtio_net_find_primary(VirtIONet *n, Error **errp) +{ + DeviceState *dev = NULL; + Error *err = NULL; + + if (qemu_opts_foreach(qemu_find_opts("device"), + is_my_primary, n, &err)) { + if (err) { + error_propagate(errp, err); + return NULL; + } + if (n->primary_device_id) { + dev = qdev_find_recursive(sysbus_get_default(), + n->primary_device_id); + } else { + error_setg(errp, "Primary device id not found"); + return NULL; + } + } + return dev; +} + + + +static DeviceState *virtio_connect_failover_devices(VirtIONet *n, + DeviceState *dev, + Error **errp) +{ + DeviceState *prim_dev = NULL; + Error *err = NULL; + + prim_dev = virtio_net_find_primary(n, &err); + if (prim_dev) { + n->primary_device_id = g_strdup(prim_dev->id); + n->primary_device_opts = prim_dev->opts; + } else { + if (err) { + error_propagate(errp, err); + } + } + + return prim_dev; +} + static void virtio_net_set_features(VirtIODevice *vdev, uint64_t features) { VirtIONet *n = VIRTIO_NET(vdev); + Error *err = NULL; int i; if (n->mtu_bypass_backend && @@ -790,6 +889,28 @@ static void virtio_net_set_features(VirtIODevice *vdev, uint64_t features) } else { memset(n->vlans, 0xff, MAX_VLAN >> 3); } + + if (virtio_has_feature(features, VIRTIO_NET_F_STANDBY)) { + qapi_event_send_failover_negotiated(n->netclient_name); + atomic_set(&n->primary_should_be_hidden, false); + failover_add_primary(n, &err); + if (err) { + n->primary_dev = virtio_connect_failover_devices(n, n->qdev, &err); + if (err) { + goto out_err; + } + failover_add_primary(n, &err); + if (err) { + goto out_err; + } + } + } + return; + +out_err: + if (err) { + warn_report_err(err); + } } static int virtio_net_handle_rx_mode(VirtIONet *n, uint8_t cmd, @@ -2650,6 +2771,150 @@ void virtio_net_set_netclient_name(VirtIONet *n, const char *name, n->netclient_type = g_strdup(type); } +static bool failover_unplug_primary(VirtIONet *n) +{ + HotplugHandler *hotplug_ctrl; + PCIDevice *pci_dev; + Error *err = NULL; + + hotplug_ctrl = qdev_get_hotplug_handler(n->primary_dev); + if (hotplug_ctrl) { + pci_dev = PCI_DEVICE(n->primary_dev); + pci_dev->partially_hotplugged = true; + hotplug_handler_unplug_request(hotplug_ctrl, n->primary_dev, &err); + if (err) { + error_report_err(err); + return false; + } + } else { + return false; + } + return true; +} + +static bool failover_replug_primary(VirtIONet *n, Error **errp) +{ + HotplugHandler *hotplug_ctrl; + PCIDevice *pdev = PCI_DEVICE(n->primary_dev); + + if (!pdev->partially_hotplugged) { + return true; + } + if (!n->primary_device_opts) { + n->primary_device_opts = qemu_opts_from_qdict( + qemu_find_opts("device"), + n->primary_device_dict, errp); + } + if (n->primary_device_opts) { + if (n->primary_dev) { + n->primary_bus = n->primary_dev->parent_bus; + } + qdev_set_parent_bus(n->primary_dev, n->primary_bus); + n->primary_should_be_hidden = false; + qemu_opt_set_bool(n->primary_device_opts, + "partially_hotplugged", true, errp); + hotplug_ctrl = qdev_get_hotplug_handler(n->primary_dev); + if (hotplug_ctrl) { + hotplug_handler_pre_plug(hotplug_ctrl, n->primary_dev, errp); + hotplug_handler_plug(hotplug_ctrl, n->primary_dev, errp); + } + if (!n->primary_dev) { + error_setg(errp, "virtio_net: couldn't find primary device"); + } + } + return *errp != NULL; +} + +static void virtio_net_handle_migration_primary(VirtIONet *n, + MigrationState *s) +{ + bool should_be_hidden; + Error *err = NULL; + + should_be_hidden = atomic_read(&n->primary_should_be_hidden); + + if (!n->primary_dev) { + n->primary_dev = virtio_connect_failover_devices(n, n->qdev, &err); + if (!n->primary_dev) { + return; + } + } + + if (migration_in_setup(s) && !should_be_hidden && + n->primary_dev) { + if (failover_unplug_primary(n)) { + vmstate_unregister(n->primary_dev, qdev_get_vmsd(n->primary_dev), + n->primary_dev); + qapi_event_send_unplug_primary(n->primary_device_id); + atomic_set(&n->primary_should_be_hidden, true); + } else { + warn_report("couldn't unplug primary device"); + } + } else if (migration_has_failed(s)) { + /* We already unplugged the device let's plugged it back */ + if (!failover_replug_primary(n, &err)) { + if (err) { + error_report_err(err); + } + } + } +} + +static void virtio_net_migration_state_notifier(Notifier *notifier, void *data) +{ + MigrationState *s = data; + VirtIONet *n = container_of(notifier, VirtIONet, migration_state); + virtio_net_handle_migration_primary(n, s); +} + +static int virtio_net_primary_should_be_hidden(DeviceListener *listener, + QemuOpts *device_opts) +{ + VirtIONet *n = container_of(listener, VirtIONet, primary_listener); + bool match_found; + bool hide; + + n->primary_device_dict = qemu_opts_to_qdict(device_opts, + n->primary_device_dict); + if (n->primary_device_dict) { + g_free(n->standby_id); + n->standby_id = g_strdup(qdict_get_try_str(n->primary_device_dict, + "failover_pair_id")); + } + if (device_opts && g_strcmp0(n->standby_id, n->netclient_name) == 0) { + match_found = true; + } else { + match_found = false; + hide = false; + g_free(n->standby_id); + n->primary_device_dict = NULL; + goto out; + } + + n->primary_device_opts = device_opts; + + /* primary_should_be_hidden is set during feature negotiation */ + hide = atomic_read(&n->primary_should_be_hidden); + + if (n->primary_device_dict) { + g_free(n->primary_device_id); + n->primary_device_id = g_strdup(qdict_get_try_str( + n->primary_device_dict, "id")); + if (!n->primary_device_id) { + warn_report("primary_device_id not set"); + } + } + +out: + if (match_found && hide) { + return 1; + } else if (match_found && !hide) { + return 0; + } else { + return -1; + } +} + static void virtio_net_device_realize(DeviceState *dev, Error **errp) { VirtIODevice *vdev = VIRTIO_DEVICE(dev); @@ -2680,6 +2945,16 @@ static void virtio_net_device_realize(DeviceState *dev, Error **errp) n->host_features |= (1ULL << VIRTIO_NET_F_SPEED_DUPLEX); } + if (n->failover) { + n->primary_listener.should_be_hidden = + virtio_net_primary_should_be_hidden; + atomic_set(&n->primary_should_be_hidden, true); + device_listener_register(&n->primary_listener); + n->migration_state.notify = virtio_net_migration_state_notifier; + add_migration_state_change_notifier(&n->migration_state); + n->host_features |= (1ULL << VIRTIO_NET_F_STANDBY); + } + virtio_net_set_config_size(n, n->host_features); virtio_init(vdev, "virtio-net", VIRTIO_ID_NET, n->config_size); @@ -2802,6 +3077,13 @@ static void virtio_net_device_unrealize(DeviceState *dev, Error **errp) g_free(n->mac_table.macs); g_free(n->vlans); + if (n->failover) { + g_free(n->primary_device_id); + g_free(n->standby_id); + qobject_unref(n->primary_device_dict); + n->primary_device_dict = NULL; + } + max_queues = n->multiqueue ? n->max_queues : 1; for (i = 0; i < max_queues; i++) { virtio_net_del_queue(n, i); @@ -2839,6 +3121,23 @@ static int virtio_net_pre_save(void *opaque) return 0; } +static bool primary_unplug_pending(void *opaque) +{ + DeviceState *dev = opaque; + VirtIODevice *vdev = VIRTIO_DEVICE(dev); + VirtIONet *n = VIRTIO_NET(vdev); + + return n->primary_dev ? n->primary_dev->pending_deleted_event : false; +} + +static bool dev_unplug_pending(void *opaque) +{ + DeviceState *dev = opaque; + VirtioDeviceClass *vdc = VIRTIO_DEVICE_GET_CLASS(dev); + + return vdc->primary_unplug_pending(dev); +} + static const VMStateDescription vmstate_virtio_net = { .name = "virtio-net", .minimum_version_id = VIRTIO_NET_VM_VERSION, @@ -2848,6 +3147,7 @@ static const VMStateDescription vmstate_virtio_net = { VMSTATE_END_OF_LIST() }, .pre_save = virtio_net_pre_save, + .dev_unplug_pending = dev_unplug_pending, }; static Property virtio_net_properties[] = { @@ -2909,6 +3209,7 @@ static Property virtio_net_properties[] = { true), DEFINE_PROP_INT32("speed", VirtIONet, net_conf.speed, SPEED_UNKNOWN), DEFINE_PROP_STRING("duplex", VirtIONet, net_conf.duplex_str), + DEFINE_PROP_BOOL("failover", VirtIONet, failover, false), DEFINE_PROP_END_OF_LIST(), }; @@ -2934,6 +3235,7 @@ static void virtio_net_class_init(ObjectClass *klass, void *data) vdc->legacy_features |= (0x1 << VIRTIO_NET_F_GSO); vdc->post_load = virtio_net_post_load_virtio; vdc->vmsd = &vmstate_virtio_net_device; + vdc->primary_unplug_pending = primary_unplug_pending; } static const TypeInfo virtio_net_info = { diff --git a/include/hw/virtio/virtio-net.h b/include/hw/virtio/virtio-net.h index 07a9319f4b..96c68d4a92 100644 --- a/include/hw/virtio/virtio-net.h +++ b/include/hw/virtio/virtio-net.h @@ -18,6 +18,7 @@ #include "standard-headers/linux/virtio_net.h" #include "hw/virtio/virtio.h" #include "net/announce.h" +#include "qemu/option_int.h" #define TYPE_VIRTIO_NET "virtio-net-device" #define VIRTIO_NET(obj) \ @@ -43,6 +44,7 @@ typedef struct virtio_net_conf int32_t speed; char *duplex_str; uint8_t duplex; + char *primary_id_str; } virtio_net_conf; /* Coalesced packets type & status */ @@ -187,6 +189,16 @@ struct VirtIONet { AnnounceTimer announce_timer; bool needs_vnet_hdr_swap; bool mtu_bypass_backend; + QemuOpts *primary_device_opts; + QDict *primary_device_dict; + DeviceState *primary_dev; + BusState *primary_bus; + char *primary_device_id; + char *standby_id; + bool primary_should_be_hidden; + bool failover; + DeviceListener primary_listener; + Notifier migration_state; }; void virtio_net_set_netclient_name(VirtIONet *n, const char *name, diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h index 60bbc0e18a..3448d67d2a 100644 --- a/include/hw/virtio/virtio.h +++ b/include/hw/virtio/virtio.h @@ -160,6 +160,7 @@ typedef struct VirtioDeviceClass { */ int (*post_load)(VirtIODevice *vdev); const VMStateDescription *vmsd; + bool (*primary_unplug_pending)(void *opaque); } VirtioDeviceClass; void virtio_instance_init_common(Object *proxy_obj, void *data, From f045a0104c8ce38da7f42141a6b3ebf3d74b474b Mon Sep 17 00:00:00 2001 From: Jens Freimann Date: Tue, 29 Oct 2019 12:49:05 +0100 Subject: [PATCH 11/14] vfio: unplug failover primary device before migration As usual block all vfio-pci devices from being migrated, but make an exception for failover primary devices. This is achieved by setting unmigratable to 0 but also add a migration blocker for all vfio-pci devices except failover primary devices. These will be unplugged before migration happens by the migration handler of the corresponding virtio-net standby device. Signed-off-by: Jens Freimann Acked-by: Alex Williamson Message-Id: <20191029114905.6856-12-jfreimann@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/vfio/pci.c | 26 ++++++++++++++++++++------ hw/vfio/pci.h | 1 + 2 files changed, 21 insertions(+), 6 deletions(-) diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c index 12fac39804..e6569a7968 100644 --- a/hw/vfio/pci.c +++ b/hw/vfio/pci.c @@ -40,6 +40,7 @@ #include "pci.h" #include "trace.h" #include "qapi/error.h" +#include "migration/blocker.h" #define TYPE_VFIO_PCI "vfio-pci" #define PCI_VFIO(obj) OBJECT_CHECK(VFIOPCIDevice, obj, TYPE_VFIO_PCI) @@ -2732,6 +2733,17 @@ static void vfio_realize(PCIDevice *pdev, Error **errp) return; } + if (!pdev->failover_pair_id) { + error_setg(&vdev->migration_blocker, + "VFIO device doesn't support migration"); + ret = migrate_add_blocker(vdev->migration_blocker, &err); + if (err) { + error_propagate(errp, err); + error_free(vdev->migration_blocker); + return; + } + } + vdev->vbasedev.name = g_path_get_basename(vdev->vbasedev.sysfsdev); vdev->vbasedev.ops = &vfio_pci_ops; vdev->vbasedev.type = VFIO_DEVICE_TYPE_PCI; @@ -3008,6 +3020,10 @@ out_teardown: vfio_bars_exit(vdev); error: error_prepend(errp, VFIO_MSG_PREFIX, vdev->vbasedev.name); + if (vdev->migration_blocker) { + migrate_del_blocker(vdev->migration_blocker); + error_free(vdev->migration_blocker); + } } static void vfio_instance_finalize(Object *obj) @@ -3019,6 +3035,10 @@ static void vfio_instance_finalize(Object *obj) vfio_bars_finalize(vdev); g_free(vdev->emulated_config_bits); g_free(vdev->rom); + if (vdev->migration_blocker) { + migrate_del_blocker(vdev->migration_blocker); + error_free(vdev->migration_blocker); + } /* * XXX Leaking igd_opregion is not an oversight, we can't remove the * fw_cfg entry therefore leaking this allocation seems like the safest @@ -3151,11 +3171,6 @@ static Property vfio_pci_dev_properties[] = { DEFINE_PROP_END_OF_LIST(), }; -static const VMStateDescription vfio_pci_vmstate = { - .name = "vfio-pci", - .unmigratable = 1, -}; - static void vfio_pci_dev_class_init(ObjectClass *klass, void *data) { DeviceClass *dc = DEVICE_CLASS(klass); @@ -3163,7 +3178,6 @@ static void vfio_pci_dev_class_init(ObjectClass *klass, void *data) dc->reset = vfio_pci_reset; dc->props = vfio_pci_dev_properties; - dc->vmsd = &vfio_pci_vmstate; dc->desc = "VFIO-based PCI device assignment"; set_bit(DEVICE_CATEGORY_MISC, dc->categories); pdc->realize = vfio_realize; diff --git a/hw/vfio/pci.h b/hw/vfio/pci.h index 834a90d646..b329d50338 100644 --- a/hw/vfio/pci.h +++ b/hw/vfio/pci.h @@ -168,6 +168,7 @@ typedef struct VFIOPCIDevice { bool no_vfio_ioeventfd; bool enable_ramfb; VFIODisplay *dpy; + Error *migration_blocker; } VFIOPCIDevice; uint32_t vfio_pci_read_config(PCIDevice *pdev, uint32_t addr, int len); From 7a064bcc6699151311c2067eea6546da6498ebe7 Mon Sep 17 00:00:00 2001 From: "Dr. David Alan Gilbert" Date: Fri, 25 Oct 2019 11:34:02 +0100 Subject: [PATCH 12/14] virtio/vhost: Use auto_rcu_read macros Use RCU_READ_LOCK_GUARD instead of manual rcu_read_(un)lock Signed-off-by: Dr. David Alan Gilbert Message-Id: <20191025103403.120616-2-dgilbert@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/virtio/vhost.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c index 99de5f196f..4da0d5a6c5 100644 --- a/hw/virtio/vhost.c +++ b/hw/virtio/vhost.c @@ -924,7 +924,7 @@ int vhost_device_iotlb_miss(struct vhost_dev *dev, uint64_t iova, int write) uint64_t uaddr, len; int ret = -EFAULT; - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); trace_vhost_iotlb_miss(dev, 1); @@ -956,8 +956,6 @@ int vhost_device_iotlb_miss(struct vhost_dev *dev, uint64_t iova, int write) trace_vhost_iotlb_miss(dev, 2); out: - rcu_read_unlock(); - return ret; } From 068ddfa970063730c5a61b06bae336d68de114e1 Mon Sep 17 00:00:00 2001 From: "Dr. David Alan Gilbert" Date: Fri, 25 Oct 2019 11:34:03 +0100 Subject: [PATCH 13/14] virtio_net: use RCU_READ_LOCK_GUARD Use RCU_READ_LOCK_GUARD rather than the manual rcu_read_(un)lock call. Signed-off-by: Dr. David Alan Gilbert Message-Id: <20191025103403.120616-3-dgilbert@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/net/virtio-net.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c index 314f365e7c..97a5113f7e 100644 --- a/hw/net/virtio-net.c +++ b/hw/net/virtio-net.c @@ -1490,12 +1490,9 @@ static ssize_t virtio_net_receive_rcu(NetClientState *nc, const uint8_t *buf, static ssize_t virtio_net_do_receive(NetClientState *nc, const uint8_t *buf, size_t size) { - ssize_t r; + RCU_READ_LOCK_GUARD(); - rcu_read_lock(); - r = virtio_net_receive_rcu(nc, buf, size); - rcu_read_unlock(); - return r; + return virtio_net_receive_rcu(nc, buf, size); } static void virtio_net_rsc_extract_unit4(VirtioNetRscChain *chain, From b5f53d04a5a567ac70d33ec95628d35583eba600 Mon Sep 17 00:00:00 2001 From: "Dr. David Alan Gilbert" Date: Mon, 28 Oct 2019 16:11:09 +0000 Subject: [PATCH 14/14] virtio: Use auto rcu_read macros Use RCU_READ_LOCK_GUARD and WITH_RCU_READ_LOCK_GUARD to replace the manual rcu_read_(un)lock calls. I think the only change is virtio_load which was missing unlocks in error paths; those end up being fatal errors so it's not that important anyway. Signed-off-by: Dr. David Alan Gilbert Message-Id: <20191028161109.60205-1-dgilbert@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/virtio/virtio.c | 65 ++++++++++++++++------------------------------ 1 file changed, 23 insertions(+), 42 deletions(-) diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c index 6348cd30e9..762df12f4c 100644 --- a/hw/virtio/virtio.c +++ b/hw/virtio/virtio.c @@ -387,7 +387,8 @@ static inline void vring_set_avail_event(VirtQueue *vq, uint16_t val) static void virtio_queue_split_set_notification(VirtQueue *vq, int enable) { - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_RING_F_EVENT_IDX)) { vring_set_avail_event(vq, vring_avail_idx(vq)); } else if (enable) { @@ -399,7 +400,6 @@ static void virtio_queue_split_set_notification(VirtQueue *vq, int enable) /* Expose avail event/used flags before caller checks the avail idx. */ smp_mb(); } - rcu_read_unlock(); } static void virtio_queue_packed_set_notification(VirtQueue *vq, int enable) @@ -408,7 +408,7 @@ static void virtio_queue_packed_set_notification(VirtQueue *vq, int enable) VRingPackedDescEvent e; VRingMemoryRegionCaches *caches; - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); caches = vring_get_region_caches(vq); vring_packed_event_read(vq->vdev, &caches->used, &e); @@ -429,7 +429,6 @@ static void virtio_queue_packed_set_notification(VirtQueue *vq, int enable) /* Expose avail event/used flags before caller checks the avail idx. */ smp_mb(); } - rcu_read_unlock(); } void virtio_queue_set_notification(VirtQueue *vq, int enable) @@ -577,9 +576,8 @@ static int virtio_queue_split_empty(VirtQueue *vq) return 0; } - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); empty = vring_avail_idx(vq) == vq->last_avail_idx; - rcu_read_unlock(); return empty; } @@ -601,12 +599,8 @@ static int virtio_queue_packed_empty_rcu(VirtQueue *vq) static int virtio_queue_packed_empty(VirtQueue *vq) { - bool empty; - - rcu_read_lock(); - empty = virtio_queue_packed_empty_rcu(vq); - rcu_read_unlock(); - return empty; + RCU_READ_LOCK_GUARD(); + return virtio_queue_packed_empty_rcu(vq); } int virtio_queue_empty(VirtQueue *vq) @@ -859,10 +853,9 @@ void virtqueue_flush(VirtQueue *vq, unsigned int count) void virtqueue_push(VirtQueue *vq, const VirtQueueElement *elem, unsigned int len) { - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); virtqueue_fill(vq, elem, len, 0); virtqueue_flush(vq, 1); - rcu_read_unlock(); } /* Called within rcu_read_lock(). */ @@ -943,7 +936,8 @@ static void virtqueue_split_get_avail_bytes(VirtQueue *vq, int64_t len = 0; int rc; - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); + idx = vq->last_avail_idx; total_bufs = in_total = out_total = 0; @@ -1033,7 +1027,6 @@ done: if (out_bytes) { *out_bytes = out_total; } - rcu_read_unlock(); return; err: @@ -1083,7 +1076,7 @@ static void virtqueue_packed_get_avail_bytes(VirtQueue *vq, VRingPackedDesc desc; bool wrap_counter; - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); idx = vq->last_avail_idx; wrap_counter = vq->last_avail_wrap_counter; total_bufs = in_total = out_total = 0; @@ -1176,7 +1169,6 @@ done: if (out_bytes) { *out_bytes = out_total; } - rcu_read_unlock(); return; err: @@ -1360,7 +1352,7 @@ static void *virtqueue_split_pop(VirtQueue *vq, size_t sz) VRingDesc desc; int rc; - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); if (virtio_queue_empty_rcu(vq)) { goto done; } @@ -1469,7 +1461,6 @@ static void *virtqueue_split_pop(VirtQueue *vq, size_t sz) trace_virtqueue_pop(vq, elem, elem->in_num, elem->out_num); done: address_space_cache_destroy(&indirect_desc_cache); - rcu_read_unlock(); return elem; @@ -1494,7 +1485,7 @@ static void *virtqueue_packed_pop(VirtQueue *vq, size_t sz) uint16_t id; int rc; - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); if (virtio_queue_packed_empty_rcu(vq)) { goto done; } @@ -1600,7 +1591,6 @@ static void *virtqueue_packed_pop(VirtQueue *vq, size_t sz) trace_virtqueue_pop(vq, elem, elem->in_num, elem->out_num); done: address_space_cache_destroy(&indirect_desc_cache); - rcu_read_unlock(); return elem; @@ -2437,13 +2427,10 @@ static bool virtio_should_notify(VirtIODevice *vdev, VirtQueue *vq) void virtio_notify_irqfd(VirtIODevice *vdev, VirtQueue *vq) { - bool should_notify; - rcu_read_lock(); - should_notify = virtio_should_notify(vdev, vq); - rcu_read_unlock(); - - if (!should_notify) { - return; + WITH_RCU_READ_LOCK_GUARD() { + if (!virtio_should_notify(vdev, vq)) { + return; + } } trace_virtio_notify_irqfd(vdev, vq); @@ -2475,13 +2462,10 @@ static void virtio_irq(VirtQueue *vq) void virtio_notify(VirtIODevice *vdev, VirtQueue *vq) { - bool should_notify; - rcu_read_lock(); - should_notify = virtio_should_notify(vdev, vq); - rcu_read_unlock(); - - if (!should_notify) { - return; + WITH_RCU_READ_LOCK_GUARD() { + if (!virtio_should_notify(vdev, vq)) { + return; + } } trace_virtio_notify(vdev, vq); @@ -3032,7 +3016,7 @@ int virtio_load(VirtIODevice *vdev, QEMUFile *f, int version_id) vdev->start_on_kick = true; } - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); for (i = 0; i < num; i++) { if (vdev->vq[i].vring.desc) { uint16_t nheads; @@ -3087,7 +3071,6 @@ int virtio_load(VirtIODevice *vdev, QEMUFile *f, int version_id) } } } - rcu_read_unlock(); if (vdc->post_load) { ret = vdc->post_load(vdev); @@ -3297,12 +3280,11 @@ static void virtio_queue_packed_restore_last_avail_idx(VirtIODevice *vdev, static void virtio_queue_split_restore_last_avail_idx(VirtIODevice *vdev, int n) { - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); if (vdev->vq[n].vring.desc) { vdev->vq[n].last_avail_idx = vring_used_idx(&vdev->vq[n]); vdev->vq[n].shadow_avail_idx = vdev->vq[n].last_avail_idx; } - rcu_read_unlock(); } void virtio_queue_restore_last_avail_idx(VirtIODevice *vdev, int n) @@ -3322,11 +3304,10 @@ static void virtio_queue_packed_update_used_idx(VirtIODevice *vdev, int n) static void virtio_split_packed_update_used_idx(VirtIODevice *vdev, int n) { - rcu_read_lock(); + RCU_READ_LOCK_GUARD(); if (vdev->vq[n].vring.desc) { vdev->vq[n].used_idx = vring_used_idx(&vdev->vq[n]); } - rcu_read_unlock(); } void virtio_queue_update_used_idx(VirtIODevice *vdev, int n)