From f1627abea333781d3e2a61bac4c7fd4502395741 Mon Sep 17 00:00:00 2001 From: Kevin Darbyshire-Bryant Date: Sat, 23 Mar 2019 09:29:49 +0000 Subject: [PATCH] netfilter: connmark: introduce set-dscpmark set-dscpmark is a method of storing the DSCP of an ip packet into conntrack mark. In combination with a suitable tc filter action (act_ctinfo) DSCP values are able to be stored in the mark on egress and restored on ingress across links that otherwise alter or bleach DSCP. This is useful for qdiscs such as CAKE which are able to shape according to policies based on DSCP. Ingress classification is traditionally a challenging task since iptables rules haven't yet run and tc filter/eBPF programs are pre-NAT lookups, hence are unable to see internal IPv4 addresses as used on the typical home masquerading gateway. x_tables CONNMARK set-dscpmark target solves the problem of storing the DSCP to the conntrack mark in a way suitable for the new act_ctinfo tc action to restore. The set-dscpmark option accepts 2 parameters, a 32bit 'dscpmask' and a 32bit 'statemask'. The dscp mask must be 6 contiguous bits and represents the area where the DSCP will be stored in the connmark. The state mask is a minimum 1 bit length mask that must not overlap with the dscpmask. It represents a flag which is set when the DSCP has been stored in the conntrack mark. This is useful to implement a 'one shot' iptables based classification where the 'complicated' iptables rules are only run once to classify the connection on initial (egress) packet and subsequent packets are all marked/restored with the same DSCP. A state mask of zero disables the setting of a status bit/s. example syntax with a suitably modified iptables user space application: iptables -A QOS_MARK_eth0 -t mangle -j CONNMARK --set-dscpmark 0xfc000000/0x01000000 Would store the DSCP in the top 6 bits of the 32bit mark field, and use the LSB of the top byte as the 'DSCP has been stored' marker. |----0xFC----conntrack mark----000000---| | Bits 31-26 | bit 25 | bit24 |~~~ Bit 0| | DSCP | unused | flag |unused | |-----------------------0x01---000000---| ^ ^ | | ---| Conditional flag | set this when dscp |-ip diffserv-| stored in mark | 6 bits | |-------------| an identically configured tc action to restore looks like: tc filter show dev eth0 ingress filter parent ffff: protocol all pref 10 u32 chain 0 filter parent ffff: protocol all pref 10 u32 chain 0 fh 800: ht divisor 1 filter parent ffff: protocol all pref 10 u32 chain 0 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1: not_in_hw match 00000000/00000000 at 0 action order 1: ctinfo zone 0 pipe index 2 ref 1 bind 1 dscp 0xfc000000/0x1000000 action order 2: mirred (Egress Redirect to device ifb4eth0) stolen index 1 ref 1 bind 1 |----0xFC----conntrack mark----000000---| | Bits 31-26 | bit 25 | bit24 |~~~ Bit 0| | DSCP | unused | flag |unused | |-----------------------0x01---000000---| | | | | ---| Conditional flag v only restore if set |-ip diffserv-| | 6 bits | |-------------| Signed-off-by: Kevin Darbyshire-Bryant --- net/netfilter/xt_connmark.c | 19 +++++++++++++++++-- 1 file changed, 17 insertions(+), 2 deletions(-) --- a/net/netfilter/xt_connmark.c +++ b/net/netfilter/xt_connmark.c @@ -42,6 +42,7 @@ connmark_tg(struct sk_buff *skb, const s enum ip_conntrack_info ctinfo; struct nf_conn *ct; u_int32_t newmark; + u_int8_t dscp, maskshift; ct = nf_ct_get(skb, &ctinfo); if (ct == NULL) @@ -49,7 +50,21 @@ connmark_tg(struct sk_buff *skb, const s switch (info->mode) { case XT_CONNMARK_SET: - newmark = (ct->mark & ~info->ctmask) ^ info->ctmark; + if (info->nfmask & 0x80000000) { + newmark = (ct->mark & ~info->ctmask) ^ info->ctmark; + } else { + if (skb->protocol == htons(ETH_P_IP)) + dscp = ipv4_get_dsfield(ip_hdr(skb)) >> 2; + else if (skb->protocol == htons(ETH_P_IPV6)) + dscp = ipv6_get_dsfield(ipv6_hdr(skb)) >> 2; + else /* protocol doesn't have diffserv */ + break; + + /* nfmask contains the mask shift value */ + maskshift = info->nfmask & 0x1f; + newmark = (ct->mark & ~info->ctmark) | + (info->ctmask | (dscp << maskshift)); + } if (ct->mark != newmark) { ct->mark = newmark; nf_conntrack_event_cache(IPCT_MARK, ct); @@ -57,7 +72,7 @@ connmark_tg(struct sk_buff *skb, const s break; case XT_CONNMARK_SAVE: newmark = (ct->mark & ~info->ctmask) ^ - (skb->mark & info->nfmask); + (skb->mark & info->nfmask); if (ct->mark != newmark) { ct->mark = newmark; nf_conntrack_event_cache(IPCT_MARK, ct);