From 310837de6c1e0badfd736b1b316b1698c53120a7 Mon Sep 17 00:00:00 2001 From: Paolo Bonzini Date: Fri, 18 Nov 2016 16:07:00 +0100 Subject: virtio: introduce grab/release_ioeventfd to fix vhost Following the recent refactoring of virtio notifiers [1], more specifically the patch ed08a2a0b ("virtio: use virtio_bus_set_host_notifier to start/stop ioeventfd") that uses virtio_bus_set_host_notifier [2] by default, core virtio code requires 'ioeventfd_started' to be set to true/false when the host notifiers are configured. When vhost is stopped and started, however, there is a stop followed by another start. Since ioeventfd_started was never set to true, the 'stop' operation triggered by virtio_bus_set_host_notifier() will not result in a call to virtio_pci_ioeventfd_assign(assign=false). This leaves the memory regions with stale notifiers and results on the next start triggering the following assertion: kvm_mem_ioeventfd_add: error adding ioeventfd: File exists Aborted This patch reintroduces (hopefully in a cleaner way) the concept that was present with ioeventfd_disabled before the refactoring. When ioeventfd_grabbed>0, ioeventfd_started tracks whether ioeventfd should be enabled or not, but ioeventfd is actually not started at all until vhost releases the host notifiers. [1] http://lists.nongnu.org/archive/html/qemu-devel/2016-10/msg07748.html [2] http://lists.nongnu.org/archive/html/qemu-devel/2016-10/msg07760.html Reported-by: Felipe Franciosi Reported-by: Christian Borntraeger Reported-by: Alex Williamson Fixes: ed08a2a0b ("virtio: use virtio_bus_set_host_notifier to start/stop ioeventfd") Reviewed-by: Cornelia Huck Reviewed-by: Stefan Hajnoczi Tested-by: Alexey Kardashevskiy Tested-by: Farhan Ali Tested-by: Alex Williamson Signed-off-by: Paolo Bonzini Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/virtio/virtio.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'hw/virtio/virtio.c') diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c index 55a00cdf9e..b7d5828601 100644 --- a/hw/virtio/virtio.c +++ b/hw/virtio/virtio.c @@ -2191,6 +2191,22 @@ void virtio_device_stop_ioeventfd(VirtIODevice *vdev) virtio_bus_stop_ioeventfd(vbus); } +int virtio_device_grab_ioeventfd(VirtIODevice *vdev) +{ + BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); + VirtioBusState *vbus = VIRTIO_BUS(qbus); + + return virtio_bus_grab_ioeventfd(vbus); +} + +void virtio_device_release_ioeventfd(VirtIODevice *vdev) +{ + BusState *qbus = qdev_get_parent_bus(DEVICE(vdev)); + VirtioBusState *vbus = VIRTIO_BUS(qbus); + + virtio_bus_release_ioeventfd(vbus); +} + static void virtio_device_class_init(ObjectClass *klass, void *data) { /* Set the default value here. */ -- cgit 1.4.1 From 0687c37c5eeef8580b31cc6e1202d874833ae38a Mon Sep 17 00:00:00 2001 From: Paolo Bonzini Date: Fri, 18 Nov 2016 16:07:01 +0100 Subject: virtio: access ISR atomically This will be needed once dataplane will be able to set it outside the big QEMU lock. Reviewed-by: Stefan Hajnoczi Tested-by: Farhan Ali Tested-by: Alex Williamson Signed-off-by: Paolo Bonzini Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/virtio/virtio-mmio.c | 6 +++--- hw/virtio/virtio-pci.c | 9 +++------ hw/virtio/virtio.c | 22 +++++++++++++++++----- 3 files changed, 23 insertions(+), 14 deletions(-) (limited to 'hw/virtio/virtio.c') diff --git a/hw/virtio/virtio-mmio.c b/hw/virtio/virtio-mmio.c index a30270f902..17412cb7b5 100644 --- a/hw/virtio/virtio-mmio.c +++ b/hw/virtio/virtio-mmio.c @@ -191,7 +191,7 @@ static uint64_t virtio_mmio_read(void *opaque, hwaddr offset, unsigned size) return virtio_queue_get_addr(vdev, vdev->queue_sel) >> proxy->guest_page_shift; case VIRTIO_MMIO_INTERRUPTSTATUS: - return vdev->isr; + return atomic_read(&vdev->isr); case VIRTIO_MMIO_STATUS: return vdev->status; case VIRTIO_MMIO_HOSTFEATURESSEL: @@ -299,7 +299,7 @@ static void virtio_mmio_write(void *opaque, hwaddr offset, uint64_t value, } break; case VIRTIO_MMIO_INTERRUPTACK: - vdev->isr &= ~value; + atomic_and(&vdev->isr, ~value); virtio_update_irq(vdev); break; case VIRTIO_MMIO_STATUS: @@ -347,7 +347,7 @@ static void virtio_mmio_update_irq(DeviceState *opaque, uint16_t vector) if (!vdev) { return; } - level = (vdev->isr != 0); + level = (atomic_read(&vdev->isr) != 0); DPRINTF("virtio_mmio setting IRQ %d\n", level); qemu_set_irq(proxy->irq, level); } diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c index 97b32febaf..521ba0b415 100644 --- a/hw/virtio/virtio-pci.c +++ b/hw/virtio/virtio-pci.c @@ -73,7 +73,7 @@ static void virtio_pci_notify(DeviceState *d, uint16_t vector) msix_notify(&proxy->pci_dev, vector); else { VirtIODevice *vdev = virtio_bus_get_device(&proxy->bus); - pci_set_irq(&proxy->pci_dev, vdev->isr & 1); + pci_set_irq(&proxy->pci_dev, atomic_read(&vdev->isr) & 1); } } @@ -449,8 +449,7 @@ static uint32_t virtio_ioport_read(VirtIOPCIProxy *proxy, uint32_t addr) break; case VIRTIO_PCI_ISR: /* reading from the ISR also clears it. */ - ret = vdev->isr; - vdev->isr = 0; + ret = atomic_xchg(&vdev->isr, 0); pci_irq_deassert(&proxy->pci_dev); break; case VIRTIO_MSI_CONFIG_VECTOR: @@ -1379,9 +1378,7 @@ static uint64_t virtio_pci_isr_read(void *opaque, hwaddr addr, { VirtIOPCIProxy *proxy = opaque; VirtIODevice *vdev = virtio_bus_get_device(&proxy->bus); - uint64_t val = vdev->isr; - - vdev->isr = 0; + uint64_t val = atomic_xchg(&vdev->isr, 0); pci_irq_deassert(&proxy->pci_dev); return val; diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c index b7d5828601..138a414cbe 100644 --- a/hw/virtio/virtio.c +++ b/hw/virtio/virtio.c @@ -945,7 +945,7 @@ void virtio_reset(void *opaque) vdev->guest_features = 0; vdev->queue_sel = 0; vdev->status = 0; - vdev->isr = 0; + atomic_set(&vdev->isr, 0); vdev->config_vector = VIRTIO_NO_VECTOR; virtio_notify_vector(vdev, vdev->config_vector); @@ -1318,10 +1318,22 @@ void virtio_del_queue(VirtIODevice *vdev, int n) vdev->vq[n].vring.num_default = 0; } +static void virtio_set_isr(VirtIODevice *vdev, int value) +{ + uint8_t old = atomic_read(&vdev->isr); + + /* Do not write ISR if it does not change, so that its cacheline remains + * shared in the common case where the guest does not read it. + */ + if ((old & value) != value) { + atomic_or(&vdev->isr, value); + } +} + void virtio_irq(VirtQueue *vq) { trace_virtio_irq(vq); - vq->vdev->isr |= 0x01; + virtio_set_isr(vq->vdev, 0x1); virtio_notify_vector(vq->vdev, vq->vector); } @@ -1355,7 +1367,7 @@ void virtio_notify(VirtIODevice *vdev, VirtQueue *vq) } trace_virtio_notify(vdev, vq); - vdev->isr |= 0x01; + virtio_set_isr(vq->vdev, 0x1); virtio_notify_vector(vdev, vq->vector); } @@ -1364,7 +1376,7 @@ void virtio_notify_config(VirtIODevice *vdev) if (!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) return; - vdev->isr |= 0x03; + virtio_set_isr(vdev, 0x3); vdev->generation++; virtio_notify_vector(vdev, vdev->config_vector); } @@ -1895,7 +1907,7 @@ void virtio_init(VirtIODevice *vdev, const char *name, vdev->device_id = device_id; vdev->status = 0; - vdev->isr = 0; + atomic_set(&vdev->isr, 0); vdev->queue_sel = 0; vdev->config_vector = VIRTIO_NO_VECTOR; vdev->vq = g_malloc0(sizeof(VirtQueue) * VIRTIO_QUEUE_MAX); -- cgit 1.4.1 From 83d768b5640946b7da55ce8335509df297e2c7cd Mon Sep 17 00:00:00 2001 From: Paolo Bonzini Date: Fri, 18 Nov 2016 16:07:02 +0100 Subject: virtio: set ISR on dataplane notifications Dataplane has been omitting forever the step of setting ISR when an interrupt is raised. This caused little breakage, because the specification actually says that ISR may not be updated in MSI mode. Some versions of the Windows drivers however didn't clear MSI mode correctly, and proceeded using polling mode (using ISR, not the used ring index!) for crashdump and hibernation. If it were just crashdump and hibernation it would not be a big deal, but recent releases of Windows do not really shut down, but rather log out and hibernate to make the next startup faster. Hence, this manifested as a more serious hang during shutdown with e.g. Windows 8.1 and virtio-win 1.8.0 RPMs. Newer versions fixed this, while older versions do not use MSI at all. The failure has always been there for virtio dataplane, but it became visible after commits 9ffe337 ("virtio-blk: always use dataplane path if ioeventfd is active", 2016-10-30) and ad07cd6 ("virtio-scsi: always use dataplane path if ioeventfd is active", 2016-10-30) made virtio-blk and virtio-scsi always use the dataplane code under KVM. The good news therefore is that it was not a bug in the patches---they were doing exactly what they were meant for, i.e. shake out remaining dataplane bugs. The fix is not hard, so it's worth arranging for the broken drivers. The virtio_should_notify+event_notifier_set pair that is common to virtio-blk and virtio-scsi dataplane is replaced with a new public function virtio_notify_irqfd that also sets ISR. The irqfd emulation code now need not set ISR anymore, so virtio_irq is removed. Reviewed-by: Stefan Hajnoczi Tested-by: Farhan Ali Tested-by: Alex Williamson Signed-off-by: Paolo Bonzini Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin --- hw/block/dataplane/virtio-blk.c | 4 +--- hw/scsi/virtio-scsi-dataplane.c | 7 ------- hw/scsi/virtio-scsi.c | 2 +- hw/virtio/trace-events | 2 +- hw/virtio/virtio.c | 36 ++++++++++++++++++++++++++++-------- include/hw/virtio/virtio-scsi.h | 1 - include/hw/virtio/virtio.h | 2 +- 7 files changed, 32 insertions(+), 22 deletions(-) (limited to 'hw/virtio/virtio.c') diff --git a/hw/block/dataplane/virtio-blk.c b/hw/block/dataplane/virtio-blk.c index 90ef557c8c..d1f9f63eaf 100644 --- a/hw/block/dataplane/virtio-blk.c +++ b/hw/block/dataplane/virtio-blk.c @@ -68,9 +68,7 @@ static void notify_guest_bh(void *opaque) unsigned i = j + ctzl(bits); VirtQueue *vq = virtio_get_queue(s->vdev, i); - if (virtio_should_notify(s->vdev, vq)) { - event_notifier_set(virtio_queue_get_guest_notifier(vq)); - } + virtio_notify_irqfd(s->vdev, vq); bits &= bits - 1; /* clear right-most bit */ } diff --git a/hw/scsi/virtio-scsi-dataplane.c b/hw/scsi/virtio-scsi-dataplane.c index f2ea29dbc3..6b8d0f0024 100644 --- a/hw/scsi/virtio-scsi-dataplane.c +++ b/hw/scsi/virtio-scsi-dataplane.c @@ -95,13 +95,6 @@ static int virtio_scsi_vring_init(VirtIOSCSI *s, VirtQueue *vq, int n, return 0; } -void virtio_scsi_dataplane_notify(VirtIODevice *vdev, VirtIOSCSIReq *req) -{ - if (virtio_should_notify(vdev, req->vq)) { - event_notifier_set(virtio_queue_get_guest_notifier(req->vq)); - } -} - /* assumes s->ctx held */ static void virtio_scsi_clear_aio(VirtIOSCSI *s) { diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c index 3e5ae6ac0f..10fd687193 100644 --- a/hw/scsi/virtio-scsi.c +++ b/hw/scsi/virtio-scsi.c @@ -69,7 +69,7 @@ static void virtio_scsi_complete_req(VirtIOSCSIReq *req) qemu_iovec_from_buf(&req->resp_iov, 0, &req->resp, req->resp_size); virtqueue_push(vq, &req->elem, req->qsgl.size + req->resp_iov.size); if (s->dataplane_started && !s->dataplane_fenced) { - virtio_scsi_dataplane_notify(vdev, req); + virtio_notify_irqfd(vdev, vq); } else { virtio_notify(vdev, vq); } diff --git a/hw/virtio/trace-events b/hw/virtio/trace-events index 8756cefa79..7b6f55e70e 100644 --- a/hw/virtio/trace-events +++ b/hw/virtio/trace-events @@ -5,7 +5,7 @@ virtqueue_fill(void *vq, const void *elem, unsigned int len, unsigned int idx) " virtqueue_flush(void *vq, unsigned int count) "vq %p count %u" virtqueue_pop(void *vq, void *elem, unsigned int in_num, unsigned int out_num) "vq %p elem %p in_num %u out_num %u" virtio_queue_notify(void *vdev, int n, void *vq) "vdev %p n %d vq %p" -virtio_irq(void *vq) "vq %p" +virtio_notify_irqfd(void *vdev, void *vq) "vdev %p vq %p" virtio_notify(void *vdev, void *vq) "vdev %p vq %p" virtio_set_status(void *vdev, uint8_t val) "vdev %p val %u" diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c index 138a414cbe..1af2de2714 100644 --- a/hw/virtio/virtio.c +++ b/hw/virtio/virtio.c @@ -1330,13 +1330,6 @@ static void virtio_set_isr(VirtIODevice *vdev, int value) } } -void virtio_irq(VirtQueue *vq) -{ - trace_virtio_irq(vq); - virtio_set_isr(vq->vdev, 0x1); - virtio_notify_vector(vq->vdev, vq->vector); -} - bool virtio_should_notify(VirtIODevice *vdev, VirtQueue *vq) { uint16_t old, new; @@ -1360,6 +1353,33 @@ bool virtio_should_notify(VirtIODevice *vdev, VirtQueue *vq) return !v || vring_need_event(vring_get_used_event(vq), new, old); } +void virtio_notify_irqfd(VirtIODevice *vdev, VirtQueue *vq) +{ + if (!virtio_should_notify(vdev, vq)) { + return; + } + + trace_virtio_notify_irqfd(vdev, vq); + + /* + * virtio spec 1.0 says ISR bit 0 should be ignored with MSI, but + * windows drivers included in virtio-win 1.8.0 (circa 2015) are + * incorrectly polling this bit during crashdump and hibernation + * in MSI mode, causing a hang if this bit is never updated. + * Recent releases of Windows do not really shut down, but rather + * log out and hibernate to make the next startup faster. Hence, + * this manifested as a more serious hang during shutdown with + * + * Next driver release from 2016 fixed this problem, so working around it + * is not a must, but it's easy to do so let's do it here. + * + * Note: it's safe to update ISR from any thread as it was switched + * to an atomic operation. + */ + virtio_set_isr(vq->vdev, 0x1); + event_notifier_set(&vq->guest_notifier); +} + void virtio_notify(VirtIODevice *vdev, VirtQueue *vq) { if (!virtio_should_notify(vdev, vq)) { @@ -1994,7 +2014,7 @@ static void virtio_queue_guest_notifier_read(EventNotifier *n) { VirtQueue *vq = container_of(n, VirtQueue, guest_notifier); if (event_notifier_test_and_clear(n)) { - virtio_irq(vq); + virtio_notify_vector(vq->vdev, vq->vector); } } diff --git a/include/hw/virtio/virtio-scsi.h b/include/hw/virtio/virtio-scsi.h index 9fbc7d7475..73751969ba 100644 --- a/include/hw/virtio/virtio-scsi.h +++ b/include/hw/virtio/virtio-scsi.h @@ -137,6 +137,5 @@ void virtio_scsi_push_event(VirtIOSCSI *s, SCSIDevice *dev, void virtio_scsi_dataplane_setup(VirtIOSCSI *s, Error **errp); int virtio_scsi_dataplane_start(VirtIODevice *s); void virtio_scsi_dataplane_stop(VirtIODevice *s); -void virtio_scsi_dataplane_notify(VirtIODevice *vdev, VirtIOSCSIReq *req); #endif /* QEMU_VIRTIO_SCSI_H */ diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h index 835b085d11..ab0e030cc4 100644 --- a/include/hw/virtio/virtio.h +++ b/include/hw/virtio/virtio.h @@ -181,6 +181,7 @@ void virtqueue_get_avail_bytes(VirtQueue *vq, unsigned int *in_bytes, unsigned max_in_bytes, unsigned max_out_bytes); bool virtio_should_notify(VirtIODevice *vdev, VirtQueue *vq); +void virtio_notify_irqfd(VirtIODevice *vdev, VirtQueue *vq); void virtio_notify(VirtIODevice *vdev, VirtQueue *vq); void virtio_save(VirtIODevice *vdev, QEMUFile *f); @@ -280,7 +281,6 @@ void virtio_queue_host_notifier_read(EventNotifier *n); void virtio_queue_aio_set_host_notifier_handler(VirtQueue *vq, AioContext *ctx, void (*fn)(VirtIODevice *, VirtQueue *)); -void virtio_irq(VirtQueue *vq); VirtQueue *virtio_vector_first_queue(VirtIODevice *vdev, uint16_t vector); VirtQueue *virtio_vector_next_queue(VirtQueue *vq); -- cgit 1.4.1