On Thu, Dec 11, 2025 at 03:00:49PM +0100, Dion Bosschieter wrote:
This series aims to implement nftables as a backend driver for the nwfilter feature. The idea is that eventually it will replace the ebiptables driver and provide an easy way for users to switch from one driver to another.
The first 2 patches are moving of functions and renames, meant to decouple nwfilter from the currently only existing ebiptables driver.
The 3rd patch introduces the new nwfilter driver. After which nwfilter allows users to choose it in the 4th patch.
The last patch introduces unit testing of the new nftables driver.
- Resolves issue https://gitlab.com/libvirt/libvirt/-/issues/603 benchmarks showed that the amount of iifname jumps for each interface with is the cause for this. Switched the nftables driver towards a vmap (verdict map) so we can have 1 rule that jumps to the correct root input/output chain per interface. Which improves throughput as when the number of interface check and jump rules increases the throughput decreases. The issue describes the interface matching works using the interface name and the majority of the effort is the strncpy, this commit also switches nftables to an interface_index compare instead. However, just using the interface_index is not enough, the amount of oif and iif jump rules causes quite a performance issue, the vmap instead solves this.
That's good.
- Split rules into separate tables: "libvirt-nwfilter-ethernet" and "libvirt-nwfilter-other" to preserve existing firewall behavior.
- Stuck with prerouting and postrouting as hooks for input / output on the -ethernet and -other table. This makes it easier to merge the tables in the future. Saving management of two tables and decreasing the amount of tables a packet sees. Currently ebtables filtering happens via PREROUTING and POSTROUTING hooks, while ip/ip6tables filtering happens in the output/forward hooks. - Stuck with 2 tables for compatibility reasons with eb iptables, unifying into 1 table will break users firewall definitions, which depend on being able to do accepts on ethernet rules (which currently get defined via ebtables) and additional filtering via the ip rules (which currently get defined via ip(6)tables). The nwfilter_nftables_driver keeps splitting the ethernet and non ethernet (other) rules in seperate tables "libvirt-nwfilter-ethernet" and "libvirt-nwfilter-other".
I guess with xtables, we would have effectively three - ebtables, iptables and ip6tables. "other" here covers both iptables and ip6tables which is fine because rules for those are mutually exclusive for any single packet. Perhaps call it "-inet" instead of '-other' ?
Unsupported nwfilter features (for now): - STP filtering - Gratuitous ARP filtering - IPSets (potential future support via nft sets) - Reject due to filtering in pre/postrouting, using drop instead of reject, copying logic from existing ebiptables ebtables actions
I don't know if it is related, but when I tried this patch series on an existing VM configured with the "clean-traffic" filter I get a failure: 2026-01-16 16:24:14.492+0000: 749946: error : virFirewallCmdNftablesApply:748 : internal error: Failed to apply firewall command 'nft add rule bridge libvirt-nwfilter-ethernet n-vnet0-rarp-out ether saddr == 52:54:00:5e:58:55 ether daddr == ff:ff:ff:ff:ff:ff ether type 0x8035 'arp operation' 3 arp saddr ip 0.0.0.0/32 arp daddr ip 0.0.0.0/32 'ether saddr' 52:54:00:5e:58:55 'ether daddr' 52:54:00:5e:58:55 accept comment '"priority=500"'': Error: conflicting statements nft add rule bridge libvirt-nwfilter-ethernet n-vnet0-rarp-out \ ether saddr == 52:54:00:5e:58:55 \ ether daddr == ff:ff:ff:ff:ff:ff \ ether type 0x8035 \ 'arp operation' 3 \ arp saddr ip 0.0.0.0/32 \ arp daddr ip 0.0.0.0/32 \ 'ether saddr' 52:54:00:5e:58:55 \ 'ether daddr' 52:54:00:5e:58:55 \ accept comment '"priority=500"' We seem to have generated the MAC address matches twice resulting in the 'conflicting statements' message. Also those extra single quotes are a sign of a bug in passing the args. Two nft args are being passed as a single argv element. Interestingly that didn't seem to cause any problem, making me think nft is munging all argv back to a single string and then parsing it. The following would fix it diff --git a/src/nwfilter/nwfilter_nftables_driver.c b/src/nwfilter/nwfilter_nftables_driver.c index 36a6c63f22..0ccb765a4e 100644 --- a/src/nwfilter/nwfilter_nftables_driver.c +++ b/src/nwfilter/nwfilter_nftables_driver.c @@ -460,7 +460,8 @@ insertRuleArg2Param(virFirewall *fw, virNWFilterVarCombIter *vars, nwItemDesc *itemLow, nwItemDesc *itemHigh, - const char *argument, + const char *argument1, + const char *argument2, const char *seperator) { char field[VIR_INT64_STR_BUFLEN]; @@ -471,7 +472,8 @@ insertRuleArg2Param(virFirewall *fw, field, sizeof(field), itemLow) < 0) return -1; - virFirewallCmdAddArg(fw, fwrule, argument); + virFirewallCmdAddArg(fw, fwrule, argument1); + virFirewallCmdAddArg(fw, fwrule, argument2); if (ENTRY_WANT_NEG_SIGN(itemLow)) virFirewallCmdAddArg(fw, fwrule, "!="); if (HAS_ENTRY_ITEM(itemHigh)) { @@ -497,21 +499,15 @@ nftablesHandlePortData(virFirewall *fw, portDataDef *portData, bool reverseRule) { - char dport[VIR_INT64_STR_BUFLEN]; - char sport[VIR_INT64_STR_BUFLEN]; - - g_snprintf(dport, sizeof(dport), reverseRule ? "%s sport" : "%s dport", - protocol); - g_snprintf(sport, sizeof(sport), reverseRule ? "%s dport": "%s sport", - protocol); - if (insertRuleArg2Param(fw, fwrule, vars, &portData->dataDstPortStart, - &portData->dataDstPortEnd, dport, "-") < 0) + &portData->dataDstPortEnd, protocol, + reverseRule ? "sport" : "dport", "-") < 0) return -1; if (insertRuleArg2Param(fw, fwrule, vars, &portData->dataSrcPortStart, - &portData->dataSrcPortEnd, sport, "-") < 0) + &portData->dataSrcPortEnd, protocol, + reverseRule ? "dport" : "sport", "-") < 0) return -1; return 0; @@ -522,7 +518,8 @@ nftablesHandleMacAddr(virFirewall *fw, virFirewallCmd *fwrule, virNWFilterVarCombIter *vars, nwItemDesc *macaddr, - const char *argument) + const char *argument1, + const char *argument2) { char macstr[VIR_MAC_STRING_BUFLEN]; @@ -532,7 +529,8 @@ nftablesHandleMacAddr(virFirewall *fw, macaddr) < 0) return -1; - virFirewallCmdAddArg(fw, fwrule, argument); + virFirewallCmdAddArg(fw, fwrule, argument1); + virFirewallCmdAddArg(fw, fwrule, argument2); if (ENTRY_WANT_NEG_SIGN(macaddr)) virFirewallCmdAddArg(fw, fwrule, "!="); virFirewallCmdAddArg(fw, fwrule, macstr); @@ -547,7 +545,7 @@ nftablesHandleSrcMacAddr(virFirewall *fw, virNWFilterVarCombIter *vars, nwItemDesc *srcMacAddr) { - return nftablesHandleMacAddr(fw, fwrule, vars, srcMacAddr, "ether saddr"); + return nftablesHandleMacAddr(fw, fwrule, vars, srcMacAddr, "ether", "saddr"); } static void @@ -893,7 +891,8 @@ insertRuleArgParam(virFirewall *fw, virFirewallCmd *fwrule, virNWFilterVarCombIter *vars, nwItemDesc *item, - const char *argument) + const char *argument1, + const char *argument2) { char field[VIR_INT64_STR_BUFLEN]; @@ -902,7 +901,8 @@ insertRuleArgParam(virFirewall *fw, field, sizeof(field), item) < 0) return -1; - virFirewallCmdAddArg(fw, fwrule, argument); + virFirewallCmdAddArg(fw, fwrule, argument1); + virFirewallCmdAddArg(fw, fwrule, argument2); if (ENTRY_WANT_NEG_SIGN(item)) virFirewallCmdAddArg(fw, fwrule, "!="); @@ -917,7 +917,8 @@ insertRuleArgParamHex(virFirewall *fw, virFirewallCmd *fwrule, virNWFilterVarCombIter *vars, nwItemDesc *item, - const char *argument) + const char *argument1, + const char *argument2) { char field[VIR_INT64_STR_BUFLEN]; @@ -926,7 +927,8 @@ insertRuleArgParamHex(virFirewall *fw, field, sizeof(field), item) < 0) return -1; - virFirewallCmdAddArg(fw, fwrule, argument); + virFirewallCmdAddArg(fw, fwrule, argument1); + virFirewallCmdAddArg(fw, fwrule, argument2); if (ENTRY_WANT_NEG_SIGN(item)) virFirewallCmdAddArg(fw, fwrule, "!="); @@ -971,7 +973,7 @@ nftablesHandleEthernetRule(virFirewall *fw, if (insertRuleArgParamHex(fw, fwrule, vars, &rule->p.ethHdrFilter.dataProtocolID, - "ether type") < 0) + "ether", "type") < 0) return -1; break; case VIR_NWFILTER_RULE_PROTOCOL_IP: @@ -1031,21 +1033,21 @@ nftablesHandleEthernetRule(virFirewall *fw, if (insertRuleArgParam(fw, fwrule, vars, &rule->p.ipHdrFilter.ipHdr.dataProtocolID, - "ip protocol") < 0) + "ip", "protocol") < 0) return -1; if (insertRuleArg2Param(fw, fwrule, vars, &rule->p.ipHdrFilter.portData.dataSrcPortStart, &rule->p.ipHdrFilter.portData.dataSrcPortEnd, - reverseRule ? "th dport" : "th sport", "-") < 0) + "th", reverseRule ? "dport" : "sport", "-") < 0) return -1; if (insertRuleArg2Param(fw, fwrule, vars, &rule->p.ipHdrFilter.portData.dataDstPortStart, &rule->p.ipHdrFilter.portData.dataDstPortEnd, - reverseRule ? "th sport" : "th dport", "-") < 0) + "th", reverseRule ? "sport" : "dport", "-") < 0) return -1; if (insertRuleArgParamHex(fw, fwrule, vars, &rule->p.ipHdrFilter.ipHdr.dataDSCP, - "ip dscp") < 0) + "ip", "dscp") < 0) return -1; break; case VIR_NWFILTER_RULE_PROTOCOL_ARP: @@ -1063,15 +1065,15 @@ nftablesHandleEthernetRule(virFirewall *fw, if (insertRuleArgParam(fw, fwrule, vars, &rule->p.arpHdrFilter.dataHWType, - "arp htype") < 0) + "arp", "htype") < 0) return -1; if (insertRuleArgParam(fw, fwrule, vars, &rule->p.arpHdrFilter.dataOpcode, - "arp operation") < 0) + "arp", "operation") < 0) return -1; if (insertRuleArgParamHex(fw, fwrule, vars, &rule->p.arpHdrFilter.dataProtocolType, - "arp ptype") < 0) + "arp", "ptype") < 0) return -1; if (HAS_ENTRY_ITEM(&rule->p.arpHdrFilter.dataARPSrcIPAddr)) { @@ -1118,11 +1120,11 @@ nftablesHandleEthernetRule(virFirewall *fw, if (nftablesHandleMacAddr(fw, fwrule, vars, &rule->p.arpHdrFilter.dataARPSrcMACAddr, - reverseRule ? "ether daddr": "ether saddr") < 0) + "ether", reverseRule ? "addr": "saddr") < 0) return -1; if (nftablesHandleMacAddr(fw, fwrule, vars, &rule->p.arpHdrFilter.dataARPDstMACAddr, - reverseRule ? "ether saddr": "ether daddr") < 0) + "ether", reverseRule ? "saddr": "daddr") < 0) return -1; if (HAS_ENTRY_ITEM(&rule->p.arpHdrFilter.dataGratuitousARP) && @@ -1187,28 +1189,28 @@ nftablesHandleEthernetRule(virFirewall *fw, if (insertRuleArgParam(fw, fwrule, vars, &rule->p.ipv6HdrFilter.ipHdr.dataProtocolID, - "ip6 nexthdr") < 0) + "ip6", "nexthdr") < 0) return -1; if (insertRuleArg2Param(fw, fwrule, vars, &rule->p.ipv6HdrFilter.portData.dataSrcPortStart, &rule->p.ipv6HdrFilter.portData.dataSrcPortEnd, - reverseRule ? "th dport" : "th sport", "-") < 0) + "th", reverseRule ? "dport" : "sport", "-") < 0) return -1; if (insertRuleArg2Param(fw, fwrule, vars, &rule->p.ipv6HdrFilter.portData.dataDstPortStart, &rule->p.ipv6HdrFilter.portData.dataDstPortEnd, - reverseRule ? "th sport" : "th dport", "-") < 0) + "th", reverseRule ? "sport" : "dport", "-") < 0) return -1; if (HAS_ENTRY_ITEM(&rule->p.ipv6HdrFilter.dataICMPTypeStart) || HAS_ENTRY_ITEM(&rule->p.ipv6HdrFilter.dataICMPCodeStart)) { if (insertRuleArgParam(fw, fwrule, vars, &rule->p.ipv6HdrFilter.dataICMPTypeStart, - "icmpv6 type") < 0) + "icmpv6", "type") < 0) return -1; if (insertRuleArgParam(fw, fwrule, vars, &rule->p.ipv6HdrFilter.dataICMPCodeStart, - "icmpv6 code") < 0) + "icmpv6", "code") < 0) return -1; } break; @@ -1222,11 +1224,11 @@ nftablesHandleEthernetRule(virFirewall *fw, if (insertRuleArgParam(fw, fwrule, vars, &rule->p.vlanHdrFilter.dataVlanID, - "vlan id") < 0) + "vlan", "id") < 0) return -1; if (insertRuleArgParam(fw, fwrule, vars, &rule->p.vlanHdrFilter.dataVlanEncap, - "vlan type") < 0) + "vlan", "type") < 0) return -1; break; case VIR_NWFILTER_RULE_PROTOCOL_STP: With regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|