sfe_ipv4_gre.c

/*
 * sfe_ipv4_gre.c
 *	Shortcut forwarding engine file for IPv4 GRE
 *
 * Copyright (c) 2022 Qualcomm Innovation Center, Inc. All rights reserved.
 *
 * Permission to use, copy, modify, and/or distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */

#include <linux/skbuff.h>
#include <net/protocol.h>
#include <linux/etherdevice.h>
#include <linux/lockdep.h>
#include <net/gre.h>
#include <net/pptp.h>

#include "sfe_debug.h"
#include "sfe_api.h"
#include "sfe.h"
#include "sfe_flow_cookie.h"
#include "sfe_ipv4.h"
#include "sfe_vlan.h"

/*
 * sfe_ipv4_recv_gre()
 *	GRE tunnel packet receive and forwarding.
 */
int sfe_ipv4_recv_gre(struct sfe_ipv4 *si, struct sk_buff *skb, struct net_device *dev,
		      unsigned int len, struct iphdr *iph, unsigned int ihl, bool sync_on_find,
		      struct sfe_l2_info *l2_info, bool tun_outer)
{
	struct sfe_ipv4_connection_match *cm;
	struct pptp_gre_header *pptp_hdr;
	struct gre_base_hdr *gre_hdr;
	struct net_device *xmit_dev;
	__be16 dest_port = 0;
	bool passthrough;
	bool bridge_flow;
	__be32 dest_ip;
	__be32 src_ip;
	bool hw_csum;
	bool ret;
	u8 ttl;

	/*
	 * Is our packet too short to contain a valid GRE header?
	 */
	if (unlikely(!pskb_may_pull(skb, sizeof(*gre_hdr) + ihl))) {
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_GRE_HEADER_INCOMPLETE);
		DEBUG_TRACE("packet too short for GRE header\n");
		return 0;
	}


	/*
	 * Read the source and destination IP address.
	 */
	src_ip = iph->saddr;
	dest_ip = iph->daddr;

	rcu_read_lock();

	/*
	 * Look for a connection match with 4 tuple if it is PPTP
	 */
	gre_hdr = (struct gre_base_hdr *)(skb->data + ihl);

	if ((gre_hdr->protocol == GRE_PROTO_PPP) && likely(pskb_may_pull(skb, (sizeof(*pptp_hdr) - 8) + ihl))) {
		pptp_hdr = (struct pptp_gre_header *)(skb->data + ihl);
		dest_port = pptp_hdr->call_id;
	}

#ifdef CONFIG_NF_FLOW_COOKIE
	cm = si->sfe_flow_cookie_table[skb->flow_cookie & SFE_FLOW_COOKIE_MASK].match;
	if (unlikely(!cm)) {
		cm = sfe_ipv4_find_connection_match_rcu(si, dev, IPPROTO_GRE, src_ip, 0, dest_ip, dest_port);
	}
#else
	cm = sfe_ipv4_find_connection_match_rcu(si, dev, IPPROTO_GRE, src_ip, 0, dest_ip, dest_port);
#endif

	if (unlikely(!cm)) {
		rcu_read_unlock();
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_GRE_NO_CONNECTION);
		DEBUG_INFO("no GRE connection match found dev %s src ip %pI4 dest ip %pI4 port %d\n", dev->name, &src_ip, &dest_ip, ntohs(dest_port));
		return 0;
	}

	/*
	 * Source interface validate.
	 */
	if (unlikely((cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_SRC_INTERFACE_CHECK) && (cm->match_dev != dev))) {
		struct sfe_ipv4_connection *c = cm->connection;
		int ret;

		spin_lock_bh(&si->lock);
		ret = sfe_ipv4_remove_connection(si, c);
		spin_unlock_bh(&si->lock);

		if (ret) {
			sfe_ipv4_flush_connection(si, c, SFE_SYNC_REASON_FLUSH);
		}
		rcu_read_unlock();
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_INVALID_SRC_IFACE);
		DEBUG_TRACE("flush on wrong source interface check failure\n");
		return 0;
	}

	passthrough = cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_PASSTHROUGH;

	/*
	 * If our packet has been marked as "sync on find" we can't actually
	 * forward it in the fast path, but now that we've found an associated
	 * connection we need sync its status before exception it to slow path unless
	 * it is passthrough (packets not directed to DUT) packet.
	 * TODO: revisit to ensure that pass through traffic is not bypassing firewall for fragmented cases
	 */
	if (unlikely(sync_on_find) && !passthrough) {
		sfe_ipv4_sync_status(si, cm->connection, SFE_SYNC_REASON_STATS);
		rcu_read_unlock();
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_GRE_IP_OPTIONS_OR_INITIAL_FRAGMENT);
		DEBUG_TRACE("%px: sfe: sync on find\n", cm);
		return 0;
	}

	/*
	 * Do we expect an ingress VLAN tag for this flow?
	 */
	if (unlikely(!sfe_vlan_validate_ingress_tag(skb, cm->ingress_vlan_hdr_cnt, cm->ingress_vlan_hdr, l2_info))) {
		rcu_read_unlock();
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_INGRESS_VLAN_TAG_MISMATCH);
		DEBUG_TRACE("VLAN tag mismatch. skb=%px\n", skb);
		return 0;
	}

	bridge_flow = !!(cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_BRIDGE_FLOW);

	/*
	 * Does our TTL allow forwarding?
	 */
	ttl = iph->ttl;
	if (!bridge_flow && (ttl < 2) && passthrough) {
		sfe_ipv4_sync_status(si, cm->connection, SFE_SYNC_REASON_STATS);
		rcu_read_unlock();

		DEBUG_TRACE("%px: sfe: TTL too low\n", skb);
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_GRE_SMALL_TTL);
		return 0;
	}

	/*
	 * From this point on we're good to modify the packet.
	 */

	/*
	 * Check if skb was cloned. If it was, unshare it. Because
	 * the data area is going to be written in this path and we don't want to
	 * change the cloned skb's data section.
	 */
	if (unlikely(skb_cloned(skb))) {
		DEBUG_TRACE("%px: skb is a cloned skb\n", skb);
		skb = skb_unshare(skb, GFP_ATOMIC);
		if (!skb) {
			DEBUG_WARN("Failed to unshare the cloned skb\n");
			rcu_read_unlock();
			return 1;
		}

		/*
		 * Update the iph and udph pointers with the unshared skb's data area.
		 */
		iph = (struct iphdr *)skb->data;
	}

	/*
	 * Check if skb has enough headroom to write L2 headers
	 */
	if (unlikely(skb_headroom(skb) < cm->l2_hdr_size)) {
		rcu_read_unlock();
		DEBUG_WARN("%px: Not enough headroom: %u\n", skb, skb_headroom(skb));
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_NO_HEADROOM);
		return 0;
	}

	/*
	 * Enable HW csum if rx checksum is verified and xmit interface is CSUM offload capable.
	 */
	hw_csum = !!(cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_CSUM_OFFLOAD) && (skb->ip_summed == CHECKSUM_UNNECESSARY);

	/*
	 * protocol handler will be valid only in decap-path.
	 */
	if (cm->proto) {
		struct net_protocol *ipprot = cm->proto;
		skb_reset_network_header(skb);
		skb_pull(skb, ihl);
		skb_reset_transport_header(skb);
		skb->fast_forwarded = 1;

		ret = ipprot->handler(skb);
		if (ret) {
			this_cpu_inc(si->stats_pcpu->packets_not_forwarded64);
			rcu_read_unlock();
			DEBUG_TRACE("GRE handler returned error %u\n", ret);
			return 1;
		}

		/*
		 * Update traffic stats
		 */
		atomic_inc(&cm->rx_packet_count);
		atomic_add(len, &cm->rx_byte_count);

		this_cpu_inc(si->stats_pcpu->packets_forwarded64);
		rcu_read_unlock();
		return 1;
	}

	/*
	 * If our packet is larger than the MTU of the transmit interface then
	 * we can't forward it easily.
	 */
	if (unlikely(len > cm->xmit_dev_mtu)) {
		sfe_ipv4_sync_status(si, cm->connection, SFE_SYNC_REASON_STATS);
		rcu_read_unlock();
		sfe_ipv4_exception_stats_inc(si, SFE_IPV4_EXCEPTION_EVENT_GRE_NEEDS_FRAGMENTATION);
		DEBUG_TRACE("%px: sfe: larger than MTU\n", cm);
		return 0;
	}

	/*
	 * Decrement our TTL
	 */
	iph->ttl = (ttl - (u8)(!bridge_flow && !tun_outer));

	/*
	 * Update DSCP
	 */
	if (unlikely(cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_DSCP_REMARK)) {
		iph->tos = (iph->tos & SFE_IPV4_DSCP_MASK) | cm->dscp;
	}

	/*
	 * Replace the IP checksum.
	 */
	if (likely(hw_csum)) {
		skb->ip_summed = CHECKSUM_PARTIAL;
	} else {
		iph->check = sfe_ipv4_gen_ip_csum(iph);
	}

	/*
	 * Update traffic stats
	 */
	atomic_inc(&cm->rx_packet_count);
	atomic_add(len, &cm->rx_byte_count);

	xmit_dev = cm->xmit_dev;
	skb->dev = xmit_dev;

	/*
	 * Check to see if we need to add VLAN tags
	 */
	if (unlikely(cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_INSERT_EGRESS_VLAN_TAG)) {
		sfe_vlan_add_tag(skb, cm->egress_vlan_hdr_cnt, cm->egress_vlan_hdr);
	}

	/*
	 * For the simple case we write this really fast.
	 */
	if (cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_WRITE_FAST_ETH_HDR) {
		struct ethhdr *eth = (struct ethhdr *)__skb_push(skb, ETH_HLEN);
		eth->h_proto = htons(ETH_P_IP);
		ether_addr_copy((u8 *)eth->h_dest, (u8 *)cm->xmit_dest_mac);
		ether_addr_copy((u8 *)eth->h_source, (u8 *)cm->xmit_src_mac);
	} else if (cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_WRITE_L2_HDR) {
		dev_hard_header(skb, xmit_dev, ETH_P_IP, cm->xmit_dest_mac, cm->xmit_src_mac, len);
	}

	/*
	 * Update priority of skb.
	 */
	if (unlikely(cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_PRIORITY_REMARK)) {
		skb->priority = cm->priority;
	}

	/*
	 * Mark outgoing packet.
	 */
	if (unlikely(cm->flags & SFE_IPV4_CONNECTION_MATCH_FLAG_MARK)) {
		skb->mark = cm->mark;
	}

	this_cpu_inc(si->stats_pcpu->packets_forwarded64);

	rcu_read_unlock();

	/*
	 * We're going to check for GSO flags when we transmit the packet so
	 * start fetching the necessary cache line now.
	 */
	prefetch(skb_shinfo(skb));

	/*
	 * Mark that this packet has been fast forwarded.
	 */
	skb->fast_forwarded = 1;

	/*
	 * Send the packet on its way.
	 */
	dev_queue_xmit(skb);

	return 1;
}