From dba5a01358d98cf5a83826f1890f9aa31a49a4d4 Mon Sep 17 00:00:00 2001 From: Kevin Darbyshire-Bryant Date: Wed, 5 Jun 2019 18:31:47 +0100 Subject: [PATCH] kernel: netfilter add connmark savedscp support savedscp is a method of storing the DSCP of an ip packet into conntrack mark. In combination with a suitable tc filter action (conndscp but may end up being integrated into connmark) DSCP values are able to be stored 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. The ingress problem is solved by the tc filter, but the tc people didn't like the idea of tc setting conntrack mark values, though they are ok with reading conntrack values and hence restoring DSCP from conntrack marks. x_tables CONNMARK with the new savedscp action solves the problem of storing the DSCP to the conntrack mark. It accepts 2 parameters. The mark is a 32bit value with usually one 1 bit set. This bit is set when savedscp saves the DSCP to the 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 mark of zero disables the setting of a status bit/s. The mask is a 32bit value of at least 6 contiguous bits and represents the area where the DSCP will be stored. e.g. iptables -A QOS_MARK_eth0 -t mangle -j CONNMARK --savedscp-mark 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. Signed-off-by: Kevin Darbyshire-Bryant --- ...etfilter-connmark-introduce-savedscp.patch | 107 ++++++++++++++ ...etfilter-connmark-introduce-savedscp.patch | 135 ++++++++++++++++++ 2 files changed, 242 insertions(+) create mode 100644 target/linux/generic/hack-4.14/645-netfilter-connmark-introduce-savedscp.patch create mode 100644 target/linux/generic/hack-4.19/645-netfilter-connmark-introduce-savedscp.patch diff --git a/target/linux/generic/hack-4.14/645-netfilter-connmark-introduce-savedscp.patch b/target/linux/generic/hack-4.14/645-netfilter-connmark-introduce-savedscp.patch new file mode 100644 index 0000000000..1681ed5db8 --- /dev/null +++ b/target/linux/generic/hack-4.14/645-netfilter-connmark-introduce-savedscp.patch @@ -0,0 +1,107 @@ +From 5a4d7714faa28c03e85d696fba82716fbda5c432 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 savedscp + +savedscp is a method of storing the DSCP of an ip packet into conntrack +mark. In combination with a suitable tc filter action (conndscp but may +end up being integrated into connmark) DSCP values are able to be stored +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. + +The ingress problem is solved by the tc filter, but the tc people didn't +like the idea of tc setting conntrack mark values, though they are ok +with reading conntrack values and hence restoring DSCP from conntrack +marks. + +x_tables CONNMARK with the new savedscp action solves the problem of +storing the DSCP to the conntrack mark. + +It accepts 2 parameters. The mark is a 32bit value with usually one 1 +bit set. This bit is set when savedscp saves the DSCP to the 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 mark +of zero disables the setting of a status bit/s. + +The mask is a 32bit value of at least 6 contiguous bits and represents +the area where the DSCP will be stored. + +e.g. + +iptables -A QOS_MARK_eth0 -t mangle -j CONNMARK --savedscp-mark 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. + +Signed-off-by: Kevin Darbyshire-Bryant +--- + include/uapi/linux/netfilter/xt_connmark.h | 3 ++- + net/netfilter/xt_connmark.c | 21 ++++++++++++++++++++- + 2 files changed, 22 insertions(+), 2 deletions(-) + +diff --git a/include/uapi/linux/netfilter/xt_connmark.h b/include/uapi/linux/netfilter/xt_connmark.h +index 408a9654f05c..e63ad3c89b92 100644 +--- a/include/uapi/linux/netfilter/xt_connmark.h ++++ b/include/uapi/linux/netfilter/xt_connmark.h +@@ -16,7 +16,8 @@ + enum { + XT_CONNMARK_SET = 0, + XT_CONNMARK_SAVE, +- XT_CONNMARK_RESTORE ++ XT_CONNMARK_RESTORE, ++ XT_CONNMARK_SAVEDSCP + }; + + struct xt_connmark_tginfo1 { +diff --git a/net/netfilter/xt_connmark.c b/net/netfilter/xt_connmark.c +index ec377cc6a369..aa5bcea8f9ba 100644 +--- a/net/netfilter/xt_connmark.c ++++ b/net/netfilter/xt_connmark.c +@@ -42,6 +42,7 @@ connmark_tg(struct sk_buff *skb, const struct xt_action_param *par) + 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) +@@ -57,7 +58,25 @@ connmark_tg(struct sk_buff *skb, const struct xt_action_param *par) + 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); ++ } ++ break; ++ case XT_CONNMARK_SAVEDSCP: ++ 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); +-- +2.20.1 (Apple Git-117) + diff --git a/target/linux/generic/hack-4.19/645-netfilter-connmark-introduce-savedscp.patch b/target/linux/generic/hack-4.19/645-netfilter-connmark-introduce-savedscp.patch new file mode 100644 index 0000000000..0c4ef92c00 --- /dev/null +++ b/target/linux/generic/hack-4.19/645-netfilter-connmark-introduce-savedscp.patch @@ -0,0 +1,135 @@ +From f171924dcf1d0b31fb7bd1cff113d7a1f7f05ec2 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 savedscp + +savedscp 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 savedscp action 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 savedsp option accepts 2 parameters, a 32bit 'dscpmask' and a 32bit +'statemask'. The dscp mask must be a minimum of 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 --savedscp-mark 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 1:1 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 +--- + include/uapi/linux/netfilter/xt_connmark.h | 3 ++- + net/netfilter/xt_connmark.c | 17 +++++++++++++++++ + 2 files changed, 19 insertions(+), 1 deletion(-) + +--- a/include/uapi/linux/netfilter/xt_connmark.h ++++ b/include/uapi/linux/netfilter/xt_connmark.h +@@ -16,7 +16,8 @@ + enum { + XT_CONNMARK_SET = 0, + XT_CONNMARK_SAVE, +- XT_CONNMARK_RESTORE ++ XT_CONNMARK_RESTORE, ++ XT_CONNMARK_SAVEDSCP + }; + + enum { +--- a/net/netfilter/xt_connmark.c ++++ b/net/netfilter/xt_connmark.c +@@ -42,6 +42,7 @@ connmark_tg_shift(struct sk_buff *skb, c + u_int32_t new_targetmark; + struct nf_conn *ct; + u_int32_t newmark; ++ u_int8_t dscp; + + ct = nf_ct_get(skb, &ctinfo); + if (ct == NULL) +@@ -74,6 +75,21 @@ connmark_tg_shift(struct sk_buff *skb, c + nf_conntrack_event_cache(IPCT_MARK, ct); + } + break; ++ case XT_CONNMARK_SAVEDSCP: ++ 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; ++ ++ newmark = (ct->mark & ~info->ctmark) | ++ (info->ctmask | (dscp << info->shift_bits)); ++ if (ct->mark != newmark) { ++ ct->mark = newmark; ++ nf_conntrack_event_cache(IPCT_MARK, ct); ++ } ++ break; + case XT_CONNMARK_RESTORE: + new_targetmark = (ct->mark & info->ctmask); + if (info->shift_dir == D_SHIFT_RIGHT) +@@ -86,6 +102,7 @@ connmark_tg_shift(struct sk_buff *skb, c + skb->mark = newmark; + break; + } ++out: + return XT_CONTINUE; + } + -- 2.25.1