Skip to contentSkip to conversation

Some fixes for vhost

PATCH Applied Tiwei Bie 2 participants 8 messages Aug 19, 2019 – Sep 27, 2019

lore
Share & tools
lore

Tiwei Bie

author
#0 lore raw

cover Some fixes for vhost

Tiwei Bie (3):
vhost: do not realloc device and queues during running
vhost: fix vring addr handling during live migration
vhost: protect vring access done by application

 lib/librte_vhost/vhost.c      | 53 ++++++++++++++++++++++++++---------
 lib/librte_vhost/vhost_user.c | 15 ++++++++--
 2 files changed, 53 insertions(+), 15 deletions(-)
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH 1/3 vhost: do not realloc device and queues during running

When the device has been started, don't do the reallocation anymore. Otherwise the pointers used in application threads can be invalidated without proper protection. Instead of introducing a global lock to protect the change of device pointers which will hurt the performance, let's just do the reallocation during setup.

Fixes af295ad4698c ("vhost: realloc device and queues to same numa node as vring desc")
---
 lib/librte_vhost/vhost_user.c | 3 +++
 1 file changed, 3 insertions(+)
lib/librte_vhost/vhost_user.c +3 -0
index 0b72648a5..ea43ab139 100644
@@ -410,6 +410,9 @@ numa_realloc(struct virtio_net *dev, int index)
 	struct batch_copy_elem *new_batch_copy_elems;
 	int ret;
 
+	if (dev->flags & VIRTIO_DEV_RUNNING)
+		return dev;
+
 	old_dev = dev;
 	vq = old_vq = dev->virtqueue[index];
 
Signature · 1 line
2.17.1
Reviewed-by on v1 1/3
#2 lore raw

Re: [dpdk-dev] [PATCH 1/3] vhost: do not realloc device and queues during running

Quoted 6 lines · Tiwei Bie

When the device has been started, don't do the reallocation anymore. Otherwise the pointers used in application threads can be invalidated without proper protection. Instead of introducing a global lock to protect the change of device pointers which will hurt the performance, let's just do the reallocation during setup.

Quoted 8 lines
Fixes af295ad4698c ("vhost: realloc device and queues to same numa node as vring desc")
---
 lib/librte_vhost/vhost_user.c | 3 +++
 1 file changed, 3 insertions(+)

Thanks,
Maxime

author
#3 lore raw

PATCH 2/3 vhost: fix vring addr handling during live migration

In reply to Tiwei Bie

When live migration starts, QEMU will set ring addrs again for each virtqueue. In this case, we should try to translate ring addrs after we invalidating the ring, otherwise virtqueues can be enabled with the addrs untranslated. Besides, also leverage the access_ok flag in non-IOMMU case to prevent the data path accessing invalidated virtqueues.

Fixes 5a4933e56be4 ("vhost: postpone ring address translations at kick time only")
---
 lib/librte_vhost/vhost.c      |  3 +--
 lib/librte_vhost/vhost_user.c | 12 ++++++++++--
 2 files changed, 11 insertions(+), 4 deletions(-)
lib/librte_vhost/vhost.c +1 -2
index 981837b5d..77be16069 100644
@@ -358,7 +358,7 @@ vring_translate(struct virtio_net *dev, struct vhost_virtqueue *vq)
 {
 
 	if (!(dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)))
-		goto out;
+		return -1;
 
 	if (vq_is_packed(dev)) {
 		if (vring_translate_packed(dev, vq) < 0)
@@ -367,7 +367,6 @@ vring_translate(struct virtio_net *dev, struct vhost_virtqueue *vq)
 		if (vring_translate_split(dev, vq) < 0)
 			return -1;
 	}
-out:
 	vq->access_ok = 1;
 
 	return 0;
lib/librte_vhost/vhost_user.c +10 -2
index ea43ab139..381250cbc 100644
@@ -622,6 +622,7 @@ translate_ring_addresses(struct virtio_net *dev, int vq_index)
 			return dev;
 		}
 
+		vq->access_ok = 1;
 		return dev;
 	}
 
@@ -680,6 +681,7 @@ translate_ring_addresses(struct virtio_net *dev, int vq_index)
 	}
 
 	vq->log_guest_addr = addr->log_guest_addr;
+	vq->access_ok = 1;
 
 	VHOST_LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address desc: %p\n",
 			dev->vid, vq->desc);
@@ -704,6 +706,7 @@ vhost_user_set_vring_addr(struct virtio_net **pdev, struct VhostUserMsg *msg,
 	struct virtio_net *dev = *pdev;
 	struct vhost_virtqueue *vq;
 	struct vhost_vring_addr *addr = &msg->payload.addr;
+	bool access_ok;
 
 	if (dev->mem == NULL)
 		return RTE_VHOST_MSG_RESULT_ERR;
@@ -711,6 +714,8 @@ vhost_user_set_vring_addr(struct virtio_net **pdev, struct VhostUserMsg *msg,
 	/* addr->index refers to the queue index. The txq 1, rxq is 0. */
 	vq = dev->virtqueue[msg->payload.addr.index];
 
+	access_ok = vq->access_ok;
+
 	/*
 	 * Rings addresses should not be interpreted as long as the ring is not
 	 * started and enabled
@@ -719,8 +724,9 @@ vhost_user_set_vring_addr(struct virtio_net **pdev, struct VhostUserMsg *msg,
 
 	vring_invalidate(dev, vq);
 
-	if (vq->enabled && (dev->features &
-				(1ULL << VHOST_USER_F_PROTOCOL_FEATURES))) {
+	if ((vq->enabled && (dev->features &
+				(1ULL << VHOST_USER_F_PROTOCOL_FEATURES))) ||
+			access_ok) {
 		dev = translate_ring_addresses(dev, msg->payload.addr.index);
 		if (!dev)
 			return RTE_VHOST_MSG_RESULT_ERR;
@@ -1325,6 +1331,8 @@ vhost_user_get_vring_base(struct virtio_net **pdev,
 	msg->size = sizeof(msg->payload.state);
 	msg->fd_num = 0;
 
+	vring_invalidate(dev, vq);
+
 	return RTE_VHOST_MSG_RESULT_REPLY;
 }
 
Signature · 1 line
2.17.1
Reviewed-by on v1 2/3
#4 lore raw

Re: [dpdk-dev] [PATCH 2/3] vhost: fix vring addr handling during live migration

Quoted 10 lines · Tiwei Bie

When live migration starts, QEMU will set ring addrs again for each virtqueue. In this case, we should try to translate ring addrs after we invalidating the ring, otherwise virtqueues can be enabled with the addrs untranslated. Besides, also leverage the access_ok flag in non-IOMMU case to prevent the data path accessing invalidated virtqueues.

Fixes 5a4933e56be4 ("vhost: postpone ring address translations at kick time only")
Quoted 7 lines
---
 lib/librte_vhost/vhost.c      |  3 +--
 lib/librte_vhost/vhost_user.c | 12 ++++++++++--
 2 files changed, 11 insertions(+), 4 deletions(-)

Thanks,
Maxime

author
#5 lore raw

PATCH 3/3 vhost: protect vring access done by application

In reply to Tiwei Bie

Besides the enqueue/dequeue API, other APIs of the builtin net backend should also be protected.

Fixes a3688046995f ("vhost: protect active rings from async ring changes")
---
 lib/librte_vhost/vhost.c | 50 +++++++++++++++++++++++++++++++---------
 1 file changed, 39 insertions(+), 11 deletions(-)
lib/librte_vhost/vhost.c +39 -11
index 77be16069..cea44df8c 100644
@@ -785,22 +785,33 @@ rte_vhost_avail_entries(int vid, uint16_t queue_id)
 {
 	struct virtio_net *dev;
 	struct vhost_virtqueue *vq;
+	uint16_t ret = 0;
 
 	dev = get_device(vid);
 	if (!dev)
 		return 0;
 
 	vq = dev->virtqueue[queue_id];
-	if (!vq->enabled)
-		return 0;
 
-	return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx;
+	rte_spinlock_lock(&vq->access_lock);
+
+	if (unlikely(!vq->enabled || vq->avail == NULL))
+		goto out;
+
+	ret = *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx;
+
+out:
+	rte_spinlock_unlock(&vq->access_lock);
+	return ret;
 }
 
-static inline void
+static inline int
 vhost_enable_notify_split(struct virtio_net *dev,
 		struct vhost_virtqueue *vq, int enable)
 {
+	if (vq->used == NULL)
+		return -1;
+
 	if (!(dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))) {
 		if (enable)
 			vq->used->flags &= ~VRING_USED_F_NO_NOTIFY;
@@ -810,17 +821,21 @@ vhost_enable_notify_split(struct virtio_net *dev,
 		if (enable)
 			vhost_avail_event(vq) = vq->last_avail_idx;
 	}
+	return 0;
 }
 
-static inline void
+static inline int
 vhost_enable_notify_packed(struct virtio_net *dev,
 		struct vhost_virtqueue *vq, int enable)
 {
 	uint16_t flags;
 
+	if (vq->device_event == NULL)
+		return -1;
+
 	if (!enable) {
 		vq->device_event->flags = VRING_EVENT_F_DISABLE;
-		return;
+		return 0;
 	}
 
 	flags = VRING_EVENT_F_ENABLE;
@@ -833,6 +848,7 @@ vhost_enable_notify_packed(struct virtio_net *dev,
 	rte_smp_wmb();
 
 	vq->device_event->flags = flags;
+	return 0;
 }
 
 int
@@ -840,18 +856,23 @@ rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
 {
 	struct virtio_net *dev = get_device(vid);
 	struct vhost_virtqueue *vq;
+	int ret;
 
 	if (!dev)
 		return -1;
 
 	vq = dev->virtqueue[queue_id];
 
+	rte_spinlock_lock(&vq->access_lock);
+
 	if (vq_is_packed(dev))
-		vhost_enable_notify_packed(dev, vq, enable);
+		ret = vhost_enable_notify_packed(dev, vq, enable);
 	else
-		vhost_enable_notify_split(dev, vq, enable);
+		ret = vhost_enable_notify_split(dev, vq, enable);
 
-	return 0;
+	rte_spinlock_unlock(&vq->access_lock);
+
+	return ret;
 }
 
 void
@@ -890,6 +911,7 @@ rte_vhost_rx_queue_count(int vid, uint16_t qid)
 {
 	struct virtio_net *dev;
 	struct vhost_virtqueue *vq;
+	uint32_t ret = 0;
 
 	dev = get_device(vid);
 	if (dev == NULL)
@@ -905,10 +927,16 @@ rte_vhost_rx_queue_count(int vid, uint16_t qid)
 	if (vq == NULL)
 		return 0;
 
+	rte_spinlock_lock(&vq->access_lock);
+
 	if (unlikely(vq->enabled == 0 || vq->avail == NULL))
-		return 0;
+		goto out;
 
-	return *((volatile uint16_t *)&vq->avail->idx) - vq->last_avail_idx;
+	ret = *((volatile uint16_t *)&vq->avail->idx) - vq->last_avail_idx;
+
+out:
+	rte_spinlock_unlock(&vq->access_lock);
+	return ret;
 }
 
 int rte_vhost_get_vdpa_device_id(int vid)
Signature · 1 line
2.17.1
Reviewed-by on v1 3/3
#6 lore raw

Re: [dpdk-dev] [PATCH 3/3] vhost: protect vring access done by application

Quoted 12 lines · Tiwei Bie

Besides the enqueue/dequeue API, other APIs of the builtin net backend should also be protected.

Fixes a3688046995f ("vhost: protect active rings from async ring changes")
---
 lib/librte_vhost/vhost.c | 50 +++++++++++++++++++++++++++++++---------
 1 file changed, 39 insertions(+), 11 deletions(-)

Thanks,
Maxime

#7 lore raw
In reply to Tiwei Bie
Quoted 9 lines · Tiwei Bie

Tiwei Bie (3):
vhost: do not realloc device and queues during running
vhost: fix vring addr handling during live migration
vhost: protect vring access done by application

 lib/librte_vhost/vhost.c      | 53 ++++++++++++++++++++++++++---------
 lib/librte_vhost/vhost_user.c | 15 ++++++++--
 2 files changed, 53 insertions(+), 15 deletions(-)

Applied to dpdk-next-virtio/master.

Thanks,
Maxime