From 2651726defb7385669fc0538fa7185d290342035 Mon Sep 17 00:00:00 2001
From: Junjie Chen <junjie.j.chen@intel.com>
Date: Mon, 15 Jan 2018 06:32:19 -0500
Subject: [PATCH] vhost: do deep copy while reallocating queue
When vhost reallocate dev and vq for NUMA enabled case, it doesn't perform
deep copy, which lead to 1) zmbuf list not valid 2) remote memory access.
This patch is to re-initlize the zmbuf list and also do the deep copy.
Signed-off-by: Junjie Chen <junjie.j.chen@intel.com>
Reviewed-by: Zhiyong Yang <zhiyong.yang@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Yuanhan Liu <yliu@fridaylinux.org>
---
lib/librte_vhost/vhost_user.c | 31 +++++++++++++++++++++++++++++++
1 file changed, 31 insertions(+)
diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index e54795a412..0ea28ecee9 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c
@@ -232,6 +232,7 @@ vhost_user_set_vring_num(struct virtio_net *dev,
"zero copy is force disabled\n");
dev->dequeue_zero_copy = 0;
}
+ TAILQ_INIT(&vq->zmbuf_list);
}
vq->shadow_used_ring = rte_malloc(NULL,
@@ -266,6 +267,9 @@ numa_realloc(struct virtio_net *dev, int index)
int oldnode, newnode;
struct virtio_net *old_dev;
struct vhost_virtqueue *old_vq, *vq;
+ struct zcopy_mbuf *new_zmbuf;
+ struct vring_used_elem *new_shadow_used_ring;
+ struct batch_copy_elem *new_batch_copy_elems;
int ret;
old_dev = dev;
@@ -290,6 +294,33 @@ numa_realloc(struct virtio_net *dev, int index)
return dev;
memcpy(vq, old_vq, sizeof(*vq));
+ TAILQ_INIT(&vq->zmbuf_list);
+
+ new_zmbuf = rte_malloc_socket(NULL, vq->zmbuf_size *
+ sizeof(struct zcopy_mbuf), 0, newnode);
+ if (new_zmbuf) {
+ rte_free(vq->zmbufs);
+ vq->zmbufs = new_zmbuf;
+ }
+
+ new_shadow_used_ring = rte_malloc_socket(NULL,
+ vq->size * sizeof(struct vring_used_elem),
+ RTE_CACHE_LINE_SIZE,
+ newnode);
+ if (new_shadow_used_ring) {
+ rte_free(vq->shadow_used_ring);
+ vq->shadow_used_ring = new_shadow_used_ring;
+ }
+
+ new_batch_copy_elems = rte_malloc_socket(NULL,
+ vq->size * sizeof(struct batch_copy_elem),
+ RTE_CACHE_LINE_SIZE,
+ newnode);
+ if (new_batch_copy_elems) {
+ rte_free(vq->batch_copy_elems);
+ vq->batch_copy_elems = new_batch_copy_elems;
+ }
+
rte_free(old_vq);
}
--
2.39.5