virtio-net: implement RX RSS processing

If VIRTIO_NET_F_RSS negotiated and RSS is enabled, process
incoming packets, calculate packet's hash and place the
packet into respective RX virtqueue.

Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

2 files changed, 87 insertions, 2 deletions

diff --git a/hw/net/Makefile.objs b/hw/net/Makefile.objs
index f2b73983ee..7ccbf72ea7 100644
--- a/hw/net/Makefile.objs
+++ b/hw/net/Makefile.objs
@@ -41,6 +41,7 @@ obj-$(CONFIG_MILKYMIST) += milkymist-minimac2.o
 obj-$(CONFIG_PSERIES) += spapr_llan.o
 obj-$(CONFIG_XILINX_ETHLITE) += xilinx_ethlite.o
 
+common-obj-$(CONFIG_VIRTIO_NET) += net_rx_pkt.o
 obj-$(CONFIG_VIRTIO_NET) += virtio-net.o
 common-obj-$(call land,$(CONFIG_VIRTIO_NET),$(CONFIG_VHOST_NET)) += vhost_net.o
 common-obj-$(call lnot,$(call land,$(CONFIG_VIRTIO_NET),$(CONFIG_VHOST_NET))) += vhost_net-stub.o
diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
index e803b0a26f..556f221669 100644
--- a/hw/net/virtio-net.c
+++ b/hw/net/virtio-net.c
@@ -42,6 +42,7 @@
 #include "trace.h"
 #include "monitor/qdev.h"
 #include "hw/pci/pci.h"
+#include "net_rx_pkt.h"
 
 #define VIRTIO_NET_VM_VERSION    11
 
@@ -1533,8 +1534,80 @@ static int receive_filter(VirtIONet *n, const uint8_t *buf, int size)
     return 0;
 }
 
+static uint8_t virtio_net_get_hash_type(bool isip4,
+                                        bool isip6,
+                                        bool isudp,
+                                        bool istcp,
+                                        uint32_t types)
+{
+    if (isip4) {
+        if (istcp && (types & VIRTIO_NET_RSS_HASH_TYPE_TCPv4)) {
+            return NetPktRssIpV4Tcp;
+        }
+        if (isudp && (types & VIRTIO_NET_RSS_HASH_TYPE_UDPv4)) {
+            return NetPktRssIpV4Udp;
+        }
+        if (types & VIRTIO_NET_RSS_HASH_TYPE_IPv4) {
+            return NetPktRssIpV4;
+        }
+    } else if (isip6) {
+        uint32_t mask = VIRTIO_NET_RSS_HASH_TYPE_TCP_EX |
+                        VIRTIO_NET_RSS_HASH_TYPE_TCPv6;
+
+        if (istcp && (types & mask)) {
+            return (types & VIRTIO_NET_RSS_HASH_TYPE_TCP_EX) ?
+                NetPktRssIpV6TcpEx : NetPktRssIpV6Tcp;
+        }
+        mask = VIRTIO_NET_RSS_HASH_TYPE_UDP_EX | VIRTIO_NET_RSS_HASH_TYPE_UDPv6;
+        if (isudp && (types & mask)) {
+            return (types & VIRTIO_NET_RSS_HASH_TYPE_UDP_EX) ?
+                NetPktRssIpV6UdpEx : NetPktRssIpV6Udp;
+        }
+        mask = VIRTIO_NET_RSS_HASH_TYPE_IP_EX | VIRTIO_NET_RSS_HASH_TYPE_IPv6;
+        if (types & mask) {
+            return (types & VIRTIO_NET_RSS_HASH_TYPE_IP_EX) ?
+                NetPktRssIpV6Ex : NetPktRssIpV6;
+        }
+    }
+    return 0xff;
+}
+
+static int virtio_net_process_rss(NetClientState *nc, const uint8_t *buf,
+                                  size_t size)
+{
+    VirtIONet *n = qemu_get_nic_opaque(nc);
+    unsigned int index = nc->queue_index, new_index;
+    struct NetRxPkt *pkt = n->rx_pkt;
+    uint8_t net_hash_type;
+    uint32_t hash;
+    bool isip4, isip6, isudp, istcp;
+
+    net_rx_pkt_set_protocols(pkt, buf + n->host_hdr_len,
+                             size - n->host_hdr_len);
+    net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
+    if (isip4 && (net_rx_pkt_get_ip4_info(pkt)->fragment)) {
+        istcp = isudp = false;
+    }
+    if (isip6 && (net_rx_pkt_get_ip6_info(pkt)->fragment)) {
+        istcp = isudp = false;
+    }
+    net_hash_type = virtio_net_get_hash_type(isip4, isip6, isudp, istcp,
+                                             n->rss_data.hash_types);
+    if (net_hash_type > NetPktRssIpV6UdpEx) {
+        return n->rss_data.default_queue;
+    }
+
+    hash = net_rx_pkt_calc_rss_hash(pkt, net_hash_type, n->rss_data.key);
+    new_index = hash & (n->rss_data.indirections_len - 1);
+    new_index = n->rss_data.indirections_table[new_index];
+    if (index == new_index) {
+        return -1;
+    }
+    return new_index;
+}
+
 static ssize_t virtio_net_receive_rcu(NetClientState *nc, const uint8_t *buf,
-                                      size_t size)
+                                      size_t size, bool no_rss)
 {
     VirtIONet *n = qemu_get_nic_opaque(nc);
     VirtIONetQueue *q = virtio_net_get_subqueue(nc);
@@ -1548,6 +1621,14 @@ static ssize_t virtio_net_receive_rcu(NetClientState *nc, const uint8_t *buf,
         return -1;
     }
 
+    if (!no_rss && n->rss_data.enabled) {
+        int index = virtio_net_process_rss(nc, buf, size);
+        if (index >= 0) {
+            NetClientState *nc2 = qemu_get_subqueue(n->nic, index);
+            return virtio_net_receive_rcu(nc2, buf, size, true);
+        }
+    }
+
     /* hdr_len refers to the header we supply to the guest */
     if (!virtio_net_has_buffers(q, size + n->guest_hdr_len - n->host_hdr_len)) {
         return 0;
@@ -1642,7 +1723,7 @@ static ssize_t virtio_net_do_receive(NetClientState *nc, const uint8_t *buf,
 {
     RCU_READ_LOCK_GUARD();
 
-    return virtio_net_receive_rcu(nc, buf, size);
+    return virtio_net_receive_rcu(nc, buf, size, false);
 }
 
 static void virtio_net_rsc_extract_unit4(VirtioNetRscChain *chain,
@@ -3221,6 +3302,8 @@ static void virtio_net_device_realize(DeviceState *dev, Error **errp)
 
     QTAILQ_INIT(&n->rsc_chains);
     n->qdev = dev;
+
+    net_rx_pkt_init(&n->rx_pkt, false);
 }
 
 static void virtio_net_device_unrealize(DeviceState *dev)
@@ -3258,6 +3341,7 @@ static void virtio_net_device_unrealize(DeviceState *dev)
     qemu_del_nic(n->nic);
     virtio_net_rsc_cleanup(n);
     g_free(n->rss_data.indirections_table);
+    net_rx_pkt_uninit(n->rx_pkt);
     virtio_cleanup(vdev);
 }