From 82fa3c776e5abba7ed6e4b4f4983d14731c37d6a Mon Sep 17 00:00:00 2001 From: KY Srinivasan Date: Mon, 11 May 2015 15:39:46 -0700 Subject: hv_netvsc: Use the xmit_more skb flag to optimize signaling the host Based on the information given to this driver (via the xmit_more skb flag), we can defer signaling the host if more packets are on the way. This will help make the host more efficient since it can potentially process a larger batch of packets. Implement this optimization. Signed-off-by: K. Y. Srinivasan Signed-off-by: David S. Miller --- drivers/net/hyperv/netvsc.c | 43 +++++++++++++++++++++++++++---------------- 1 file changed, 27 insertions(+), 16 deletions(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c index 2d9ef533cc48..1c4f265f4e7c 100644 --- a/drivers/net/hyperv/netvsc.c +++ b/drivers/net/hyperv/netvsc.c @@ -743,6 +743,7 @@ static inline int netvsc_send_pkt( u64 req_id; int ret; struct hv_page_buffer *pgbuf; + u32 ring_avail = hv_ringbuf_avail_percent(&out_channel->outbound); nvmsg.hdr.msg_type = NVSP_MSG1_TYPE_SEND_RNDIS_PKT; if (packet->is_data_pkt) { @@ -769,32 +770,42 @@ static inline int netvsc_send_pkt( if (out_channel->rescind) return -ENODEV; + /* + * It is possible that once we successfully place this packet + * on the ringbuffer, we may stop the queue. In that case, we want + * to notify the host independent of the xmit_more flag. We don't + * need to be precise here; in the worst case we may signal the host + * unnecessarily. + */ + if (ring_avail < (RING_AVAIL_PERCENT_LOWATER + 1)) + packet->xmit_more = false; + if (packet->page_buf_cnt) { pgbuf = packet->cp_partial ? packet->page_buf + packet->rmsg_pgcnt : packet->page_buf; - ret = vmbus_sendpacket_pagebuffer(out_channel, - pgbuf, - packet->page_buf_cnt, - &nvmsg, - sizeof(struct nvsp_message), - req_id); + ret = vmbus_sendpacket_pagebuffer_ctl(out_channel, + pgbuf, + packet->page_buf_cnt, + &nvmsg, + sizeof(struct nvsp_message), + req_id, + VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED, + !packet->xmit_more); } else { - ret = vmbus_sendpacket( - out_channel, &nvmsg, - sizeof(struct nvsp_message), - req_id, - VM_PKT_DATA_INBAND, - VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED); + ret = vmbus_sendpacket_ctl(out_channel, &nvmsg, + sizeof(struct nvsp_message), + req_id, + VM_PKT_DATA_INBAND, + VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED, + !packet->xmit_more); } if (ret == 0) { atomic_inc(&net_device->num_outstanding_sends); atomic_inc(&net_device->queue_sends[q_idx]); - if (hv_ringbuf_avail_percent(&out_channel->outbound) < - RING_AVAIL_PERCENT_LOWATER) { - netif_tx_stop_queue(netdev_get_tx_queue( - ndev, q_idx)); + if (ring_avail < RING_AVAIL_PERCENT_LOWATER) { + netif_tx_stop_queue(netdev_get_tx_queue(ndev, q_idx)); if (atomic_read(&net_device-> queue_sends[q_idx]) < 1) -- cgit v1.2.3 From 06635a35d13d42b95422bba6633f175245cc644e Mon Sep 17 00:00:00 2001 From: Jiri Pirko Date: Tue, 12 May 2015 14:56:16 +0200 Subject: flow_dissect: use programable dissector in skb_flow_dissect and friends Signed-off-by: Jiri Pirko Signed-off-by: David S. Miller --- drivers/net/bonding/bond_main.c | 18 +-- drivers/net/ethernet/cisco/enic/enic_clsf.c | 27 ++-- drivers/net/ethernet/cisco/enic/enic_ethtool.c | 10 +- drivers/net/hyperv/netvsc_drv.c | 8 +- include/linux/skbuff.h | 4 +- include/net/flow_dissector.h | 63 ++++---- include/net/ip.h | 8 +- include/net/ipv6.h | 8 +- net/core/flow_dissector.c | 199 +++++++++++++++++-------- net/ethernet/eth.c | 6 +- net/sched/cls_flow.c | 20 +-- net/sched/sch_choke.c | 4 +- 12 files changed, 229 insertions(+), 146 deletions(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index fe7c38721775..ef26e0147050 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -3051,16 +3051,16 @@ static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, int noff, proto = -1; if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) - return skb_flow_dissect(skb, fk); + return skb_flow_dissect_flow_keys(skb, fk); - fk->ports = 0; + fk->ports.ports = 0; noff = skb_network_offset(skb); if (skb->protocol == htons(ETH_P_IP)) { if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) return false; iph = ip_hdr(skb); - fk->src = iph->saddr; - fk->dst = iph->daddr; + fk->addrs.src = iph->saddr; + fk->addrs.dst = iph->daddr; noff += iph->ihl << 2; if (!ip_is_fragment(iph)) proto = iph->protocol; @@ -3068,15 +3068,15 @@ static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6)))) return false; iph6 = ipv6_hdr(skb); - fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr); - fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr); + fk->addrs.src = (__force __be32)ipv6_addr_hash(&iph6->saddr); + fk->addrs.dst = (__force __be32)ipv6_addr_hash(&iph6->daddr); noff += sizeof(*iph6); proto = iph6->nexthdr; } else { return false; } if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) - fk->ports = skb_flow_get_ports(skb, noff, proto); + fk->ports.ports = skb_flow_get_ports(skb, noff, proto); return true; } @@ -3102,8 +3102,8 @@ u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) hash = bond_eth_hash(skb); else - hash = (__force u32)flow.ports; - hash ^= (__force u32)flow.dst ^ (__force u32)flow.src; + hash = (__force u32)flow.ports.ports; + hash ^= (__force u32)flow.addrs.dst ^ (__force u32)flow.addrs.src; hash ^= (hash >> 16); hash ^= (hash >> 8); diff --git a/drivers/net/ethernet/cisco/enic/enic_clsf.c b/drivers/net/ethernet/cisco/enic/enic_clsf.c index 380f04903a36..d3d25c7e5b96 100644 --- a/drivers/net/ethernet/cisco/enic/enic_clsf.c +++ b/drivers/net/ethernet/cisco/enic/enic_clsf.c @@ -22,7 +22,7 @@ int enic_addfltr_5t(struct enic *enic, struct flow_keys *keys, u16 rq) int res; struct filter data; - switch (keys->ip_proto) { + switch (keys->basic.ip_proto) { case IPPROTO_TCP: data.u.ipv4.protocol = PROTO_TCP; break; @@ -33,10 +33,10 @@ int enic_addfltr_5t(struct enic *enic, struct flow_keys *keys, u16 rq) return -EPROTONOSUPPORT; }; data.type = FILTER_IPV4_5TUPLE; - data.u.ipv4.src_addr = ntohl(keys->src); - data.u.ipv4.dst_addr = ntohl(keys->dst); - data.u.ipv4.src_port = ntohs(keys->port16[0]); - data.u.ipv4.dst_port = ntohs(keys->port16[1]); + data.u.ipv4.src_addr = ntohl(keys->addrs.src); + data.u.ipv4.dst_addr = ntohl(keys->addrs.dst); + data.u.ipv4.src_port = ntohs(keys->ports.port16[0]); + data.u.ipv4.dst_port = ntohs(keys->ports.port16[1]); data.u.ipv4.flags = FILTER_FIELDS_IPV4_5TUPLE; spin_lock_bh(&enic->devcmd_lock); @@ -158,11 +158,11 @@ static struct enic_rfs_fltr_node *htbl_key_search(struct hlist_head *h, struct enic_rfs_fltr_node *tpos; hlist_for_each_entry(tpos, h, node) - if (tpos->keys.src == k->src && - tpos->keys.dst == k->dst && - tpos->keys.ports == k->ports && - tpos->keys.ip_proto == k->ip_proto && - tpos->keys.n_proto == k->n_proto) + if (tpos->keys.addrs.src == k->addrs.src && + tpos->keys.addrs.dst == k->addrs.dst && + tpos->keys.ports.ports == k->ports.ports && + tpos->keys.basic.ip_proto == k->basic.ip_proto && + tpos->keys.basic.n_proto == k->basic.n_proto) return tpos; return NULL; } @@ -177,9 +177,10 @@ int enic_rx_flow_steer(struct net_device *dev, const struct sk_buff *skb, int res, i; enic = netdev_priv(dev); - res = skb_flow_dissect(skb, &keys); - if (!res || keys.n_proto != htons(ETH_P_IP) || - (keys.ip_proto != IPPROTO_TCP && keys.ip_proto != IPPROTO_UDP)) + res = skb_flow_dissect_flow_keys(skb, &keys); + if (!res || keys.basic.n_proto != htons(ETH_P_IP) || + (keys.basic.ip_proto != IPPROTO_TCP && + keys.basic.ip_proto != IPPROTO_UDP)) return -EPROTONOSUPPORT; tbl_idx = skb_get_hash_raw(skb) & ENIC_RFS_FLW_MASK; diff --git a/drivers/net/ethernet/cisco/enic/enic_ethtool.c b/drivers/net/ethernet/cisco/enic/enic_ethtool.c index 28d9ca675a27..7588f8dcc890 100644 --- a/drivers/net/ethernet/cisco/enic/enic_ethtool.c +++ b/drivers/net/ethernet/cisco/enic/enic_ethtool.c @@ -334,7 +334,7 @@ static int enic_grxclsrule(struct enic *enic, struct ethtool_rxnfc *cmd) n = htbl_fltr_search(enic, (u16)fsp->location); if (!n) return -EINVAL; - switch (n->keys.ip_proto) { + switch (n->keys.basic.ip_proto) { case IPPROTO_TCP: fsp->flow_type = TCP_V4_FLOW; break; @@ -346,16 +346,16 @@ static int enic_grxclsrule(struct enic *enic, struct ethtool_rxnfc *cmd) break; } - fsp->h_u.tcp_ip4_spec.ip4src = n->keys.src; + fsp->h_u.tcp_ip4_spec.ip4src = n->keys.addrs.src; fsp->m_u.tcp_ip4_spec.ip4src = (__u32)~0; - fsp->h_u.tcp_ip4_spec.ip4dst = n->keys.dst; + fsp->h_u.tcp_ip4_spec.ip4dst = n->keys.addrs.dst; fsp->m_u.tcp_ip4_spec.ip4dst = (__u32)~0; - fsp->h_u.tcp_ip4_spec.psrc = n->keys.port16[0]; + fsp->h_u.tcp_ip4_spec.psrc = n->keys.ports.port16[0]; fsp->m_u.tcp_ip4_spec.psrc = (__u16)~0; - fsp->h_u.tcp_ip4_spec.pdst = n->keys.port16[1]; + fsp->h_u.tcp_ip4_spec.pdst = n->keys.ports.port16[1]; fsp->m_u.tcp_ip4_spec.pdst = (__u16)~0; fsp->ring_cookie = n->rq_id; diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c index 5993c7e2d723..8e5fe888a0ec 100644 --- a/drivers/net/hyperv/netvsc_drv.c +++ b/drivers/net/hyperv/netvsc_drv.c @@ -196,12 +196,12 @@ static bool netvsc_set_hash(u32 *hash, struct sk_buff *skb) struct flow_keys flow; int data_len; - if (!skb_flow_dissect(skb, &flow) || - !(flow.n_proto == htons(ETH_P_IP) || - flow.n_proto == htons(ETH_P_IPV6))) + if (!skb_flow_dissect_flow_keys(skb, &flow) || + !(flow.basic.n_proto == htons(ETH_P_IP) || + flow.basic.n_proto == htons(ETH_P_IPV6))) return false; - if (flow.ip_proto == IPPROTO_TCP) + if (flow.basic.ip_proto == IPPROTO_TCP) data_len = 12; else data_len = 8; diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index b01c7fba7c17..f83aa6568cbf 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -1935,8 +1935,8 @@ static inline void skb_probe_transport_header(struct sk_buff *skb, if (skb_transport_header_was_set(skb)) return; - else if (skb_flow_dissect(skb, &keys)) - skb_set_transport_header(skb, keys.thoff); + else if (skb_flow_dissect_flow_keys(skb, &keys)) + skb_set_transport_header(skb, keys.basic.thoff); else skb_set_transport_header(skb, offset_hint); } diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h index 20239e807f77..0c8d406fb730 100644 --- a/include/net/flow_dissector.h +++ b/include/net/flow_dissector.h @@ -59,42 +59,47 @@ struct flow_dissector { unsigned short int offset[FLOW_DISSECTOR_KEY_MAX]; }; -/* struct flow_keys: - * @src: source ip address in case of IPv4 - * For IPv6 it contains 32bit hash of src address - * @dst: destination ip address in case of IPv4 - * For IPv6 it contains 32bit hash of dst address - * @ports: port numbers of Transport header - * port16[0]: src port number - * port16[1]: dst port number - * @thoff: Transport header offset - * @n_proto: Network header protocol (eg. IPv4/IPv6) - * @ip_proto: Transport header protocol (eg. TCP/UDP) - * All the members, except thoff, are in network byte order. - */ -struct flow_keys { - /* (src,dst) must be grouped, in the same way than in IP header */ - __be32 src; - __be32 dst; - union { - __be32 ports; - __be16 port16[2]; - }; - u16 thoff; - __be16 n_proto; - u8 ip_proto; -}; - void skb_flow_dissector_init(struct flow_dissector *flow_dissector, const struct flow_dissector_key *key, unsigned int key_count); -bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow, + +bool __skb_flow_dissect(const struct sk_buff *skb, + struct flow_dissector *flow_dissector, + void *target_container, void *data, __be16 proto, int nhoff, int hlen); static inline bool skb_flow_dissect(const struct sk_buff *skb, - struct flow_keys *flow) + struct flow_dissector *flow_dissector, + void *target_container) +{ + return __skb_flow_dissect(skb, flow_dissector, target_container, + NULL, 0, 0, 0); +} + +struct flow_keys { + struct flow_dissector_key_addrs addrs; + struct flow_dissector_key_ports ports; + struct flow_dissector_key_basic basic; +}; + +extern struct flow_dissector flow_keys_dissector; +extern struct flow_dissector flow_keys_buf_dissector; + +static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, + struct flow_keys *flow) +{ + memset(flow, 0, sizeof(*flow)); + return __skb_flow_dissect(skb, &flow_keys_dissector, flow, + NULL, 0, 0, 0); +} + +static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow, + void *data, __be16 proto, + int nhoff, int hlen) { - return __skb_flow_dissect(skb, flow, NULL, 0, 0, 0); + memset(flow, 0, sizeof(*flow)); + return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow, + data, proto, nhoff, hlen); } __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, diff --git a/include/net/ip.h b/include/net/ip.h index 562eb653aebc..b0443d4fe13f 100644 --- a/include/net/ip.h +++ b/include/net/ip.h @@ -360,10 +360,10 @@ static inline void inet_set_txhash(struct sock *sk) struct inet_sock *inet = inet_sk(sk); struct flow_keys keys; - keys.src = inet->inet_saddr; - keys.dst = inet->inet_daddr; - keys.port16[0] = inet->inet_sport; - keys.port16[1] = inet->inet_dport; + keys.addrs.src = inet->inet_saddr; + keys.addrs.dst = inet->inet_daddr; + keys.ports.port16[0] = inet->inet_sport; + keys.ports.port16[1] = inet->inet_dport; sk->sk_txhash = flow_hash_from_keys(&keys); } diff --git a/include/net/ipv6.h b/include/net/ipv6.h index 9932b86394a9..9eed9761dfce 100644 --- a/include/net/ipv6.h +++ b/include/net/ipv6.h @@ -698,10 +698,10 @@ static inline void ip6_set_txhash(struct sock *sk) struct ipv6_pinfo *np = inet6_sk(sk); struct flow_keys keys; - keys.src = (__force __be32)ipv6_addr_hash(&np->saddr); - keys.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr); - keys.port16[0] = inet->inet_sport; - keys.port16[1] = inet->inet_dport; + keys.addrs.src = (__force __be32)ipv6_addr_hash(&np->saddr); + keys.addrs.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr); + keys.ports.port16[0] = inet->inet_sport; + keys.ports.port16[1] = inet->inet_dport; sk->sk_txhash = flow_hash_from_keys(&keys); } diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c index 6883e22b8fc4..6a49acaa6651 100644 --- a/net/core/flow_dissector.c +++ b/net/core/flow_dissector.c @@ -13,6 +13,7 @@ #include #include #include +#include #include #include @@ -63,17 +64,6 @@ void skb_flow_dissector_init(struct flow_dissector *flow_dissector, } EXPORT_SYMBOL(skb_flow_dissector_init); -/* copy saddr & daddr, possibly using 64bit load/store - * Equivalent to : flow->src = iph->saddr; - * flow->dst = iph->daddr; - */ -static void iph_to_flow_copy_addrs(struct flow_keys *flow, const struct iphdr *iph) -{ - BUILD_BUG_ON(offsetof(typeof(*flow), dst) != - offsetof(typeof(*flow), src) + sizeof(flow->src)); - memcpy(&flow->src, &iph->saddr, sizeof(flow->src) + sizeof(flow->dst)); -} - /** * __skb_flow_get_ports - extract the upper layer ports and return them * @skb: sk_buff to extract the ports from @@ -111,17 +101,27 @@ EXPORT_SYMBOL(__skb_flow_get_ports); /** * __skb_flow_dissect - extract the flow_keys struct and return it * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified + * @flow_dissector: list of keys to dissect + * @target_container: target structure to put dissected values into * @data: raw buffer pointer to the packet, if NULL use skb->data * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb) * @hlen: packet header length, if @data is NULL use skb_headlen(skb) * - * The function will try to retrieve the struct flow_keys from either the skbuff - * or a raw buffer specified by the rest parameters + * The function will try to retrieve individual keys into target specified + * by flow_dissector from either the skbuff or a raw buffer specified by the + * rest parameters. + * + * Caller must take care of zeroing target container memory. */ -bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow, +bool __skb_flow_dissect(const struct sk_buff *skb, + struct flow_dissector *flow_dissector, + void *target_container, void *data, __be16 proto, int nhoff, int hlen) { + struct flow_dissector_key_basic *key_basic; + struct flow_dissector_key_addrs *key_addrs; + struct flow_dissector_key_ports *key_ports; u8 ip_proto; if (!data) { @@ -131,7 +131,12 @@ bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow, hlen = skb_headlen(skb); } - memset(flow, 0, sizeof(*flow)); + /* It is ensured by skb_flow_dissector_init() that basic key will + * be always present. + */ + key_basic = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_BASIC, + target_container); again: switch (proto) { @@ -148,14 +153,13 @@ ip: if (ip_is_fragment(iph)) ip_proto = 0; - /* skip the address processing if skb is NULL. The assumption - * here is that if there is no skb we are not looking for flow - * info but lengths and protocols. - */ - if (!skb) + if (!skb_flow_dissector_uses_key(flow_dissector, + FLOW_DISSECTOR_KEY_IPV4_ADDRS)) break; - - iph_to_flow_copy_addrs(flow, iph); + key_addrs = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_IPV4_ADDRS, + target_container); + memcpy(key_addrs, &iph->saddr, sizeof(*key_addrs)); break; } case htons(ETH_P_IPV6): { @@ -171,12 +175,15 @@ ipv6: ip_proto = iph->nexthdr; nhoff += sizeof(struct ipv6hdr); - /* see comment above in IPv4 section */ - if (!skb) + if (!skb_flow_dissector_uses_key(flow_dissector, + FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS)) break; + key_addrs = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS, + target_container); - flow->src = (__force __be32)ipv6_addr_hash(&iph->saddr); - flow->dst = (__force __be32)ipv6_addr_hash(&iph->daddr); + key_addrs->src = (__force __be32)ipv6_addr_hash(&iph->saddr); + key_addrs->dst = (__force __be32)ipv6_addr_hash(&iph->daddr); flow_label = ip6_flowlabel(iph); if (flow_label) { @@ -184,10 +191,18 @@ ipv6: * use that to represent the ports without any * further dissection. */ - flow->n_proto = proto; - flow->ip_proto = ip_proto; - flow->ports = flow_label; - flow->thoff = (u16)nhoff; + + key_basic->n_proto = proto; + key_basic->ip_proto = ip_proto; + key_basic->thoff = (u16)nhoff; + + if (!skb_flow_dissector_uses_key(flow_dissector, + FLOW_DISSECTOR_KEY_PORTS)) + break; + key_ports = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_PORTS, + target_container); + key_ports->ports = flow_label; return true; } @@ -234,14 +249,22 @@ ipv6: hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr); if (!hdr) return false; - flow->src = hdr->srcnode; - flow->dst = 0; - flow->n_proto = proto; - flow->thoff = (u16)nhoff; + key_basic->n_proto = proto; + key_basic->thoff = (u16)nhoff; + + if (skb_flow_dissector_uses_key(flow_dissector, + FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS)) { + return true; + key_addrs = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS, + target_container); + key_addrs->src = hdr->srcnode; + key_addrs->dst = 0; + } return true; } case htons(ETH_P_FCOE): - flow->thoff = (u16)(nhoff + FCOE_HEADER_LEN); + key_basic->thoff = (u16)(nhoff + FCOE_HEADER_LEN); /* fall through */ default: return false; @@ -296,14 +319,24 @@ ipv6: break; } - flow->n_proto = proto; - flow->ip_proto = ip_proto; - flow->thoff = (u16) nhoff; - - /* unless skb is set we don't need to record port info */ - if (skb) - flow->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, - data, hlen); + /* It is ensured by skb_flow_dissector_init() that basic key will + * be always present. + */ + key_basic = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_BASIC, + target_container); + key_basic->n_proto = proto; + key_basic->ip_proto = ip_proto; + key_basic->thoff = (u16) nhoff; + + if (skb_flow_dissector_uses_key(flow_dissector, + FLOW_DISSECTOR_KEY_PORTS)) { + key_ports = skb_flow_dissector_target(flow_dissector, + FLOW_DISSECTOR_KEY_PORTS, + target_container); + key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto, + data, hlen); + } return true; } @@ -325,16 +358,16 @@ static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval) u32 hash; /* get a consistent hash (same value on both flow directions) */ - if (((__force u32)keys->dst < (__force u32)keys->src) || - (((__force u32)keys->dst == (__force u32)keys->src) && - ((__force u16)keys->port16[1] < (__force u16)keys->port16[0]))) { - swap(keys->dst, keys->src); - swap(keys->port16[0], keys->port16[1]); + if (((__force u32)keys->addrs.dst < (__force u32)keys->addrs.src) || + (((__force u32)keys->addrs.dst == (__force u32)keys->addrs.src) && + ((__force u16)keys->ports.port16[1] < (__force u16)keys->ports.port16[0]))) { + swap(keys->addrs.dst, keys->addrs.src); + swap(keys->ports.port16[0], keys->ports.port16[1]); } - hash = __flow_hash_3words((__force u32)keys->dst, - (__force u32)keys->src, - (__force u32)keys->ports, + hash = __flow_hash_3words((__force u32)keys->addrs.dst, + (__force u32)keys->addrs.src, + (__force u32)keys->ports.ports, keyval); if (!hash) hash = 1; @@ -352,7 +385,7 @@ EXPORT_SYMBOL(flow_hash_from_keys); static inline u32 ___skb_get_hash(const struct sk_buff *skb, struct flow_keys *keys, u32 keyval) { - if (!skb_flow_dissect(skb, keys)) + if (!skb_flow_dissect_flow_keys(skb, keys)) return 0; return __flow_hash_from_keys(keys, keyval); @@ -377,11 +410,11 @@ void make_flow_keys_digest(struct flow_keys_digest *digest, memset(digest, 0, sizeof(*digest)); - data->n_proto = flow->n_proto; - data->ip_proto = flow->ip_proto; - data->ports = flow->ports; - data->src = flow->src; - data->dst = flow->dst; + data->n_proto = flow->basic.n_proto; + data->ip_proto = flow->basic.ip_proto; + data->ports = flow->ports.ports; + data->src = flow->addrs.src; + data->dst = flow->addrs.dst; } EXPORT_SYMBOL(make_flow_keys_digest); @@ -404,7 +437,7 @@ void __skb_get_hash(struct sk_buff *skb) hash = ___skb_get_hash(skb, &keys, hashrnd); if (!hash) return; - if (keys.ports) + if (keys.ports.ports) skb->l4_hash = 1; skb->sw_hash = 1; skb->hash = hash; @@ -422,9 +455,9 @@ EXPORT_SYMBOL(skb_get_hash_perturb); u32 __skb_get_poff(const struct sk_buff *skb, void *data, const struct flow_keys *keys, int hlen) { - u32 poff = keys->thoff; + u32 poff = keys->basic.thoff; - switch (keys->ip_proto) { + switch (keys->basic.ip_proto) { case IPPROTO_TCP: { /* access doff as u8 to avoid unaligned access */ const u8 *doff; @@ -478,8 +511,52 @@ u32 skb_get_poff(const struct sk_buff *skb) { struct flow_keys keys; - if (!skb_flow_dissect(skb, &keys)) + if (!skb_flow_dissect_flow_keys(skb, &keys)) return 0; return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb)); } + +static const struct flow_dissector_key flow_keys_dissector_keys[] = { + { + .key_id = FLOW_DISSECTOR_KEY_BASIC, + .offset = offsetof(struct flow_keys, basic), + }, + { + .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, + .offset = offsetof(struct flow_keys, addrs), + }, + { + .key_id = FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS, + .offset = offsetof(struct flow_keys, addrs), + }, + { + .key_id = FLOW_DISSECTOR_KEY_PORTS, + .offset = offsetof(struct flow_keys, ports), + }, +}; + +static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = { + { + .key_id = FLOW_DISSECTOR_KEY_BASIC, + .offset = offsetof(struct flow_keys, basic), + }, +}; + +struct flow_dissector flow_keys_dissector __read_mostly; +EXPORT_SYMBOL(flow_keys_dissector); + +struct flow_dissector flow_keys_buf_dissector __read_mostly; + +static int __init init_default_flow_dissectors(void) +{ + skb_flow_dissector_init(&flow_keys_dissector, + flow_keys_dissector_keys, + ARRAY_SIZE(flow_keys_dissector_keys)); + skb_flow_dissector_init(&flow_keys_buf_dissector, + flow_keys_buf_dissector_keys, + ARRAY_SIZE(flow_keys_buf_dissector_keys)); + return 0; +} + +late_initcall_sync(init_default_flow_dissectors); diff --git a/net/ethernet/eth.c b/net/ethernet/eth.c index 9332a0ab0698..c3325bd2f3fb 100644 --- a/net/ethernet/eth.c +++ b/net/ethernet/eth.c @@ -131,9 +131,9 @@ u32 eth_get_headlen(void *data, unsigned int len) return len; /* parse any remaining L2/L3 headers, check for L4 */ - if (!__skb_flow_dissect(NULL, &keys, data, - eth->h_proto, sizeof(*eth), len)) - return max_t(u32, keys.thoff, sizeof(*eth)); + if (!skb_flow_dissect_flow_keys_buf(&keys, data, eth->h_proto, + sizeof(*eth), len)) + return max_t(u32, keys.basic.thoff, sizeof(*eth)); /* parse for any L4 headers */ return min_t(u32, __skb_get_poff(NULL, data, &keys, len), len); diff --git a/net/sched/cls_flow.c b/net/sched/cls_flow.c index 4c5a92298906..4b3e3e30bf4d 100644 --- a/net/sched/cls_flow.c +++ b/net/sched/cls_flow.c @@ -68,35 +68,35 @@ static inline u32 addr_fold(void *addr) static u32 flow_get_src(const struct sk_buff *skb, const struct flow_keys *flow) { - if (flow->src) - return ntohl(flow->src); + if (flow->addrs.src) + return ntohl(flow->addrs.src); return addr_fold(skb->sk); } static u32 flow_get_dst(const struct sk_buff *skb, const struct flow_keys *flow) { - if (flow->dst) - return ntohl(flow->dst); + if (flow->addrs.dst) + return ntohl(flow->addrs.dst); return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb); } static u32 flow_get_proto(const struct sk_buff *skb, const struct flow_keys *flow) { - return flow->ip_proto; + return flow->basic.ip_proto; } static u32 flow_get_proto_src(const struct sk_buff *skb, const struct flow_keys *flow) { - if (flow->ports) - return ntohs(flow->port16[0]); + if (flow->ports.ports) + return ntohs(flow->ports.port16[0]); return addr_fold(skb->sk); } static u32 flow_get_proto_dst(const struct sk_buff *skb, const struct flow_keys *flow) { - if (flow->ports) - return ntohs(flow->port16[1]); + if (flow->ports.ports) + return ntohs(flow->ports.port16[1]); return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb); } @@ -295,7 +295,7 @@ static int flow_classify(struct sk_buff *skb, const struct tcf_proto *tp, keymask = f->keymask; if (keymask & FLOW_KEYS_NEEDED) - skb_flow_dissect(skb, &flow_keys); + skb_flow_dissect_flow_keys(skb, &flow_keys); for (n = 0; n < f->nkeys; n++) { key = ffs(keymask) - 1; diff --git a/net/sched/sch_choke.c b/net/sched/sch_choke.c index 2624e991a2d1..93d5742dc7e0 100644 --- a/net/sched/sch_choke.c +++ b/net/sched/sch_choke.c @@ -170,13 +170,13 @@ static bool choke_match_flow(struct sk_buff *skb1, if (!choke_skb_cb(skb1)->keys_valid) { choke_skb_cb(skb1)->keys_valid = 1; - skb_flow_dissect(skb1, &temp); + skb_flow_dissect_flow_keys(skb1, &temp); make_flow_keys_digest(&choke_skb_cb(skb1)->keys, &temp); } if (!choke_skb_cb(skb2)->keys_valid) { choke_skb_cb(skb2)->keys_valid = 1; - skb_flow_dissect(skb2, &temp); + skb_flow_dissect_flow_keys(skb2, &temp); make_flow_keys_digest(&choke_skb_cb(skb2)->keys, &temp); } -- cgit v1.2.3 From 7eafd9b4005643cfc24f1daf78f4dd56ff71f559 Mon Sep 17 00:00:00 2001 From: sixiao@microsoft.com Date: Thu, 14 May 2015 01:00:25 -0700 Subject: hv_netvsc: use per_cpu stats to calculate TX/RX data Current code does not lock anything when calculating the TX and RX stats. As a result, the RX and TX data reported by ifconfig are not accuracy in a system with high network throughput and multiple CPUs (in my test, RX/TX = 83% between 2 HyperV VM nodes which have 8 vCPUs and 40G Ethernet). This patch fixed the above issue by using per_cpu stats. netvsc_get_stats64() summarizes TX and RX data by iterating over all CPUs to get their respective stats. This v2 patch addressed David's comments on the cleanup path when netdev_alloc_pcpu_stats() failed. Signed-off-by: Simon Xiao Reviewed-by: K. Y. Srinivasan Reviewed-by: Haiyang Zhang Signed-off-by: David S. Miller --- drivers/net/hyperv/hyperv_net.h | 9 +++++ drivers/net/hyperv/netvsc_drv.c | 85 +++++++++++++++++++++++++++++++++++++---- 2 files changed, 86 insertions(+), 8 deletions(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h index 41071d32bc8e..5a92b36daed6 100644 --- a/drivers/net/hyperv/hyperv_net.h +++ b/drivers/net/hyperv/hyperv_net.h @@ -611,6 +611,12 @@ struct multi_send_data { u32 count; /* counter of batched packets */ }; +struct netvsc_stats { + u64 packets; + u64 bytes; + struct u64_stats_sync s_sync; +}; + /* The context of the netvsc device */ struct net_device_context { /* point back to our device context */ @@ -618,6 +624,9 @@ struct net_device_context { struct delayed_work dwork; struct work_struct work; u32 msg_enable; /* debug level */ + + struct netvsc_stats __percpu *tx_stats; + struct netvsc_stats __percpu *rx_stats; }; /* Per netvsc device */ diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c index 8e5fe888a0ec..0c8587240ff3 100644 --- a/drivers/net/hyperv/netvsc_drv.c +++ b/drivers/net/hyperv/netvsc_drv.c @@ -391,7 +391,7 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net) u32 skb_length; u32 pkt_sz; struct hv_page_buffer page_buf[MAX_PAGE_BUFFER_COUNT]; - + struct netvsc_stats *tx_stats = this_cpu_ptr(net_device_ctx->tx_stats); /* We will atmost need two pages to describe the rndis * header. We can only transmit MAX_PAGE_BUFFER_COUNT number @@ -580,8 +580,10 @@ do_send: drop: if (ret == 0) { - net->stats.tx_bytes += skb_length; - net->stats.tx_packets++; + u64_stats_update_begin(&tx_stats->s_sync); + tx_stats->packets++; + tx_stats->bytes += skb_length; + u64_stats_update_end(&tx_stats->s_sync); } else { if (ret != -EAGAIN) { dev_kfree_skb_any(skb); @@ -644,13 +646,17 @@ int netvsc_recv_callback(struct hv_device *device_obj, struct ndis_tcp_ip_checksum_info *csum_info) { struct net_device *net; + struct net_device_context *net_device_ctx; struct sk_buff *skb; + struct netvsc_stats *rx_stats; net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev; if (!net || net->reg_state != NETREG_REGISTERED) { packet->status = NVSP_STAT_FAIL; return 0; } + net_device_ctx = netdev_priv(net); + rx_stats = this_cpu_ptr(net_device_ctx->rx_stats); /* Allocate a skb - TODO direct I/O to pages? */ skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen); @@ -686,8 +692,10 @@ int netvsc_recv_callback(struct hv_device *device_obj, skb_record_rx_queue(skb, packet->channel-> offermsg.offer.sub_channel_index); - net->stats.rx_packets++; - net->stats.rx_bytes += packet->total_data_buflen; + u64_stats_update_begin(&rx_stats->s_sync); + rx_stats->packets++; + rx_stats->bytes += packet->total_data_buflen; + u64_stats_update_end(&rx_stats->s_sync); /* * Pass the skb back up. Network stack will deallocate the skb when it @@ -753,6 +761,46 @@ static int netvsc_change_mtu(struct net_device *ndev, int mtu) return 0; } +static struct rtnl_link_stats64 *netvsc_get_stats64(struct net_device *net, + struct rtnl_link_stats64 *t) +{ + struct net_device_context *ndev_ctx = netdev_priv(net); + int cpu; + + for_each_possible_cpu(cpu) { + struct netvsc_stats *tx_stats = per_cpu_ptr(ndev_ctx->tx_stats, + cpu); + struct netvsc_stats *rx_stats = per_cpu_ptr(ndev_ctx->rx_stats, + cpu); + u64 tx_packets, tx_bytes, rx_packets, rx_bytes; + unsigned int start; + + do { + start = u64_stats_fetch_begin_irq(&tx_stats->s_sync); + tx_packets = tx_stats->packets; + tx_bytes = tx_stats->bytes; + } while (u64_stats_fetch_retry_irq(&tx_stats->s_sync, start)); + + do { + start = u64_stats_fetch_begin_irq(&rx_stats->s_sync); + rx_packets = rx_stats->packets; + rx_bytes = rx_stats->bytes; + } while (u64_stats_fetch_retry_irq(&rx_stats->s_sync, start)); + + t->tx_bytes += tx_bytes; + t->tx_packets += tx_packets; + t->rx_bytes += rx_bytes; + t->rx_packets += rx_packets; + } + + t->tx_dropped = net->stats.tx_dropped; + t->tx_errors = net->stats.tx_dropped; + + t->rx_dropped = net->stats.rx_dropped; + t->rx_errors = net->stats.rx_errors; + + return t; +} static int netvsc_set_mac_addr(struct net_device *ndev, void *p) { @@ -804,6 +852,7 @@ static const struct net_device_ops device_ops = { .ndo_validate_addr = eth_validate_addr, .ndo_set_mac_address = netvsc_set_mac_addr, .ndo_select_queue = netvsc_select_queue, + .ndo_get_stats64 = netvsc_get_stats64, #ifdef CONFIG_NET_POLL_CONTROLLER .ndo_poll_controller = netvsc_poll_controller, #endif @@ -855,6 +904,14 @@ static void netvsc_link_change(struct work_struct *w) netdev_notify_peers(net); } +static void netvsc_free_netdev(struct net_device *netdev) +{ + struct net_device_context *net_device_ctx = netdev_priv(netdev); + + free_percpu(net_device_ctx->tx_stats); + free_percpu(net_device_ctx->rx_stats); + free_netdev(netdev); +} static int netvsc_probe(struct hv_device *dev, const struct hv_vmbus_device_id *dev_id) @@ -883,6 +940,18 @@ static int netvsc_probe(struct hv_device *dev, netdev_dbg(net, "netvsc msg_enable: %d\n", net_device_ctx->msg_enable); + net_device_ctx->tx_stats = netdev_alloc_pcpu_stats(struct netvsc_stats); + if (!net_device_ctx->tx_stats) { + free_netdev(net); + return -ENOMEM; + } + net_device_ctx->rx_stats = netdev_alloc_pcpu_stats(struct netvsc_stats); + if (!net_device_ctx->rx_stats) { + free_percpu(net_device_ctx->tx_stats); + free_netdev(net); + return -ENOMEM; + } + hv_set_drvdata(dev, net); INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_link_change); INIT_WORK(&net_device_ctx->work, do_set_multicast); @@ -909,7 +978,7 @@ static int netvsc_probe(struct hv_device *dev, ret = rndis_filter_device_add(dev, &device_info); if (ret != 0) { netdev_err(net, "unable to add netvsc device (ret %d)\n", ret); - free_netdev(net); + netvsc_free_netdev(net); hv_set_drvdata(dev, NULL); return ret; } @@ -923,7 +992,7 @@ static int netvsc_probe(struct hv_device *dev, if (ret != 0) { pr_err("Unable to register netdev.\n"); rndis_filter_device_remove(dev); - free_netdev(net); + netvsc_free_netdev(net); } else { schedule_delayed_work(&net_device_ctx->dwork, 0); } @@ -962,7 +1031,7 @@ static int netvsc_remove(struct hv_device *dev) */ rndis_filter_device_remove(dev); - free_netdev(net); + netvsc_free_netdev(net); return 0; } -- cgit v1.2.3 From 4b02b58b52fab385e120c4b4f0d90f5f3076352d Mon Sep 17 00:00:00 2001 From: sixiao@microsoft.com Date: Fri, 15 May 2015 02:33:03 -0700 Subject: hv_netvsc: change member name of struct netvsc_stats Currently the struct netvsc_stats has a member s_sync of type u64_stats_sync. This definition will break kernel build as the macro netdev_alloc_pcpu_stats requires this member name to be syncp. (see netdev_alloc_pcpu_stats definition in ./include/linux/netdevice.h) This patch changes netvsc_stats's member name from s_sync to syncp to fix the build break. Signed-off-by: Simon Xiao Signed-off-by: David S. Miller --- drivers/net/hyperv/hyperv_net.h | 2 +- drivers/net/hyperv/netvsc_drv.c | 16 ++++++++-------- 2 files changed, 9 insertions(+), 9 deletions(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h index 5a92b36daed6..ddcc7f8d22b4 100644 --- a/drivers/net/hyperv/hyperv_net.h +++ b/drivers/net/hyperv/hyperv_net.h @@ -614,7 +614,7 @@ struct multi_send_data { struct netvsc_stats { u64 packets; u64 bytes; - struct u64_stats_sync s_sync; + struct u64_stats_sync syncp; }; /* The context of the netvsc device */ diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c index 0c8587240ff3..d9c88bc09f45 100644 --- a/drivers/net/hyperv/netvsc_drv.c +++ b/drivers/net/hyperv/netvsc_drv.c @@ -580,10 +580,10 @@ do_send: drop: if (ret == 0) { - u64_stats_update_begin(&tx_stats->s_sync); + u64_stats_update_begin(&tx_stats->syncp); tx_stats->packets++; tx_stats->bytes += skb_length; - u64_stats_update_end(&tx_stats->s_sync); + u64_stats_update_end(&tx_stats->syncp); } else { if (ret != -EAGAIN) { dev_kfree_skb_any(skb); @@ -692,10 +692,10 @@ int netvsc_recv_callback(struct hv_device *device_obj, skb_record_rx_queue(skb, packet->channel-> offermsg.offer.sub_channel_index); - u64_stats_update_begin(&rx_stats->s_sync); + u64_stats_update_begin(&rx_stats->syncp); rx_stats->packets++; rx_stats->bytes += packet->total_data_buflen; - u64_stats_update_end(&rx_stats->s_sync); + u64_stats_update_end(&rx_stats->syncp); /* * Pass the skb back up. Network stack will deallocate the skb when it @@ -776,16 +776,16 @@ static struct rtnl_link_stats64 *netvsc_get_stats64(struct net_device *net, unsigned int start; do { - start = u64_stats_fetch_begin_irq(&tx_stats->s_sync); + start = u64_stats_fetch_begin_irq(&tx_stats->syncp); tx_packets = tx_stats->packets; tx_bytes = tx_stats->bytes; - } while (u64_stats_fetch_retry_irq(&tx_stats->s_sync, start)); + } while (u64_stats_fetch_retry_irq(&tx_stats->syncp, start)); do { - start = u64_stats_fetch_begin_irq(&rx_stats->s_sync); + start = u64_stats_fetch_begin_irq(&rx_stats->syncp); rx_packets = rx_stats->packets; rx_bytes = rx_stats->bytes; - } while (u64_stats_fetch_retry_irq(&rx_stats->s_sync, start)); + } while (u64_stats_fetch_retry_irq(&rx_stats->syncp, start)); t->tx_bytes += tx_bytes; t->tx_packets += tx_packets; -- cgit v1.2.3 From e01ec2199ef22e2cabd7d6e68a192f3eb728029f Mon Sep 17 00:00:00 2001 From: KY Srinivasan Date: Wed, 27 May 2015 13:16:57 -0700 Subject: hv_netvsc: Properly size the vrss queues The current algorithm for deciding on the number of VRSS channels is not optimal since we open up the min of number of CPUs online and the number of VRSS channels the host is offering. So on a 32 VCPU guest we could potentially open 32 VRSS subchannels. Experimentation has shown that it is best to limit the number of VRSS channels to the number of CPUs within a NUMA node. Here is the new algorithm for deciding on the number of sub-channels we would open up: 1) Pick the minimum of what the host is offering and what the driver in the guest is specifying as the default value. 2) Pick the minimum of (1) and the numbers of CPUs in the NUMA node the primary channel is bound to. Signed-off-by: K. Y. Srinivasan Signed-off-by: David S. Miller --- drivers/net/hyperv/hyperv_net.h | 1 + drivers/net/hyperv/netvsc_drv.c | 4 ++++ drivers/net/hyperv/rndis_filter.c | 16 ++++++++++++++-- 3 files changed, 19 insertions(+), 2 deletions(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h index ddcc7f8d22b4..dd4544085db3 100644 --- a/drivers/net/hyperv/hyperv_net.h +++ b/drivers/net/hyperv/hyperv_net.h @@ -161,6 +161,7 @@ struct netvsc_device_info { unsigned char mac_adr[ETH_ALEN]; bool link_state; /* 0 - link up, 1 - link down */ int ring_size; + u32 max_num_vrss_chns; }; enum rndis_device_state { diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c index d9c88bc09f45..358475ed9b59 100644 --- a/drivers/net/hyperv/netvsc_drv.c +++ b/drivers/net/hyperv/netvsc_drv.c @@ -46,6 +46,8 @@ static int ring_size = 128; module_param(ring_size, int, S_IRUGO); MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)"); +static int max_num_vrss_chns = 8; + static const u32 default_msg = NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN | NETIF_MSG_RX_ERR | @@ -755,6 +757,7 @@ static int netvsc_change_mtu(struct net_device *ndev, int mtu) ndevctx->device_ctx = hdev; hv_set_drvdata(hdev, ndev); device_info.ring_size = ring_size; + device_info.max_num_vrss_chns = max_num_vrss_chns; rndis_filter_device_add(hdev, &device_info); netif_tx_wake_all_queues(ndev); @@ -975,6 +978,7 @@ static int netvsc_probe(struct hv_device *dev, /* Notify the netvsc driver of the new device */ device_info.ring_size = ring_size; + device_info.max_num_vrss_chns = max_num_vrss_chns; ret = rndis_filter_device_add(dev, &device_info); if (ret != 0) { netdev_err(net, "unable to add netvsc device (ret %d)\n", ret); diff --git a/drivers/net/hyperv/rndis_filter.c b/drivers/net/hyperv/rndis_filter.c index 9118cea91882..006c1b8c2385 100644 --- a/drivers/net/hyperv/rndis_filter.c +++ b/drivers/net/hyperv/rndis_filter.c @@ -1013,6 +1013,9 @@ int rndis_filter_device_add(struct hv_device *dev, struct ndis_recv_scale_cap rsscap; u32 rsscap_size = sizeof(struct ndis_recv_scale_cap); u32 mtu, size; + u32 num_rss_qs; + const struct cpumask *node_cpu_mask; + u32 num_possible_rss_qs; rndis_device = get_rndis_device(); if (!rndis_device) @@ -1100,9 +1103,18 @@ int rndis_filter_device_add(struct hv_device *dev, if (ret || rsscap.num_recv_que < 2) goto out; + num_rss_qs = min(device_info->max_num_vrss_chns, rsscap.num_recv_que); + net_device->max_chn = rsscap.num_recv_que; - net_device->num_chn = (num_online_cpus() < rsscap.num_recv_que) ? - num_online_cpus() : rsscap.num_recv_que; + + /* + * We will limit the VRSS channels to the number CPUs in the NUMA node + * the primary channel is currently bound to. + */ + node_cpu_mask = cpumask_of_node(cpu_to_node(dev->channel->target_cpu)); + num_possible_rss_qs = cpumask_weight(node_cpu_mask); + net_device->num_chn = min(num_possible_rss_qs, num_rss_qs); + if (net_device->num_chn == 1) goto out; -- cgit v1.2.3 From 0a726c2b499e390b1c1fc3092bd789f2192a2d03 Mon Sep 17 00:00:00 2001 From: K. Y. Srinivasan Date: Thu, 28 May 2015 17:08:06 -0700 Subject: hv_netvsc: Allocate the receive buffer from the correct NUMA node Allocate the receive bufer from the NUMA node assigned to the primary channel. Signed-off-by: K. Y. Srinivasan Signed-off-by: David S. Miller --- drivers/net/hyperv/netvsc.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c index b0249685139c..d187965eba36 100644 --- a/drivers/net/hyperv/netvsc.c +++ b/drivers/net/hyperv/netvsc.c @@ -227,13 +227,18 @@ static int netvsc_init_buf(struct hv_device *device) struct netvsc_device *net_device; struct nvsp_message *init_packet; struct net_device *ndev; + int node; net_device = get_outbound_net_device(device); if (!net_device) return -ENODEV; ndev = net_device->ndev; - net_device->recv_buf = vzalloc(net_device->recv_buf_size); + node = cpu_to_node(device->channel->target_cpu); + net_device->recv_buf = vzalloc_node(net_device->recv_buf_size, node); + if (!net_device->recv_buf) + net_device->recv_buf = vzalloc(net_device->recv_buf_size); + if (!net_device->recv_buf) { netdev_err(ndev, "unable to allocate receive " "buffer of size %d\n", net_device->recv_buf_size); -- cgit v1.2.3 From 5defde5946676ee23cd6a9d0e1de899410f4a33f Mon Sep 17 00:00:00 2001 From: K. Y. Srinivasan Date: Thu, 28 May 2015 17:08:07 -0700 Subject: hv_netvsc: Allocate the sendbuf in a NUMA aware way Allocate the send buffer in a NUMA aware way. Signed-off-by: K. Y. Srinivasan Signed-off-by: David S. Miller --- drivers/net/hyperv/netvsc.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/net/hyperv') diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c index d187965eba36..06de98a05622 100644 --- a/drivers/net/hyperv/netvsc.c +++ b/drivers/net/hyperv/netvsc.c @@ -326,7 +326,9 @@ static int netvsc_init_buf(struct hv_device *device) /* Now setup the send buffer. */ - net_device->send_buf = vzalloc(net_device->send_buf_size); + net_device->send_buf = vzalloc_node(net_device->send_buf_size, node); + if (!net_device->send_buf) + net_device->send_buf = vzalloc(net_device->send_buf_size); if (!net_device->send_buf) { netdev_err(ndev, "unable to allocate send " "buffer of size %d\n", net_device->send_buf_size); -- cgit v1.2.3