Skip to contentSkip to conversation

netfilter: Exclude LEGACY TABLES on PREEMPT_RT.

PATCH Applied Sebastian Andrzej Siewior 3 participants 8 messages Apr 4, 2025 – Jun 30, 2025

Subsystems (11)
Networking Kselftest netfilter BPF +7 more subsystems

v4 of 5 Applied Latest is v5 →

lore
Share & tools
lore

Sebastian Andrzej Siewior

author maintainer
#0 lore raw

PATCH v4 1/1 netfilter: Exclude LEGACY TABLES on PREEMPT_RT.

From: Pablo Neira Ayuso <[email protected]>

The seqcount xt_recseq is used to synchronize the replacement of xt_table::private in xt_replace_table() against all readers such as ipt_do_table()

To ensure that there is only one writer, the writing side disables bottom halves. The sequence counter can be acquired recursively. Only the first invocation modifies the sequence counter (signaling that a writer is in progress) while the following (recursive) writer does not modify the counter.
The lack of a proper locking mechanism for the sequence counter can lead to live lock on PREEMPT_RT if the high prior reader preempts the writer. Additionally if the per-CPU lock on PREEMPT_RT is removed from local_bh_disable() then there is no synchronisation for the per-CPU sequence counter.

The affected code is "just" the legacy netfilter code which is replaced by "netfilter tables". That code can be disabled without sacrificing functionality because everything is provided by the newer implementation. This will only requires the usage of the "-nft" tools instead of the "-legacy" ones.
The long term plan is to remove the legacy code so lets accelerate the progress.

Relax dependencies on iptables legacy, replace select with depends on, this should cause no harm to existing kernel configs and users can still toggle IP{6}_NF_IPTABLES_LEGACY in any case.
Make EBTABLES_LEGACY, IPTABLES_LEGACY and ARPTABLES depend on NETFILTER_LEGACY. Hide xt_recseq and its users, xt_register_table() and xt_percpu_counter_alloc() behind NETFILTER_LEGACY. Let NETFILTER_LEGACY depend on !PREEMPT_RT.

Replace CONFIG_IP6_NF_MANGLE->CONFIG_IP6_NF_IPTABLES for TCPOPTSTRIP and add CONFIG_NFT_COMPAT_ARP to the MARK target for the IPv6 and ARP target to keep it enabled without the LEGACY code for NFT.

---

v3…v4 https://lore.kernel.org/all/[email protected]/
- Merge all three patches into one.
- CONFIG_IP6_NF_MANGLE -> CONFIG_IP6_NF_IPTABLES in xt_TCPOPTSTRIP and
+ CONFIG_NFT_COMPAT_ARP xt_mark to allow the modules without LEGACY as per Florian.

v2…v3 https://lore.kernel.org/all/[email protected]/
- Instead of getting LEGACY code to work for PREEMPT_RT the code is now disabled on PREEMPT_RT. Since the long term plan is to get rid of it anyway, it might be less painful for everyone.

v1…v2 https://lore.kernel.org/all/[email protected]/
- Updated kerneldoc in 2/3 so that the renamed parameter is part of

    it.

- Updated description 1/3 in case there are complains regarding the synchronize_rcu(). The suggested course of action is to motivate people to move away from "legacy" towards "nft" tooling. Last resort is not to wait for the in-flight counter and just copy what is there.

 net/Kconfig                    | 10 ++++++++++
 net/bridge/netfilter/Kconfig   |  8 ++++----
 net/ipv4/netfilter/Kconfig     | 15 ++++++++-------
 net/ipv6/netfilter/Kconfig     | 13 +++++++------
 net/netfilter/x_tables.c       | 16 +++++++++++-----
 net/netfilter/xt_TCPOPTSTRIP.c |  4 ++--
 net/netfilter/xt_mark.c        |  2 +-
 7 files changed, 43 insertions(+), 25 deletions(-)
net/Kconfig +10 -0
index c3fca69a7c834..e5d5bcafa0e18 100644
@@ -211,6 +211,16 @@ menuconfig NETFILTER
 
 if NETFILTER
 
+config NETFILTER_LEGACY
+	bool "Netfilter legacy tables support"
+	depends on NETFILTER && !PREEMPT_RT
+	help
+	  Say Y here if you still require support for legacy tables. This is
+	  required by the legacy tools (iptables-legacy) and is not needed if
+	  you use iptables over nftables (iptales-nft).
+	  Legacy support is not limited to IP, it also includes EBTABLES and
+	  ARPTABLES.
+
 config NETFILTER_ADVANCED
 	bool "Advanced netfilter configuration"
 	depends on NETFILTER
net/bridge/netfilter/Kconfig +4 -4
index f16bbbbb94817..0080127421882 100644
@@ -42,7 +42,7 @@ config NF_CONNTRACK_BRIDGE
 # old sockopt interface and eval loop
 config BRIDGE_NF_EBTABLES_LEGACY
 	tristate "Legacy EBTABLES support"
-	depends on BRIDGE && NETFILTER_XTABLES
+	depends on BRIDGE && NETFILTER_XTABLES && NETFILTER_LEGACY
 	default n
 	help
 	 Legacy ebtables packet/frame classifier.
@@ -65,7 +65,7 @@ if BRIDGE_NF_EBTABLES
 #
 config BRIDGE_EBT_BROUTE
 	tristate "ebt: broute table support"
-	select BRIDGE_NF_EBTABLES_LEGACY
+	depends on BRIDGE_NF_EBTABLES_LEGACY
 	help
 	  The ebtables broute table is used to define rules that decide between
 	  bridging and routing frames, giving Linux the functionality of a
@@ -76,7 +76,7 @@ config BRIDGE_EBT_BROUTE
 
 config BRIDGE_EBT_T_FILTER
 	tristate "ebt: filter table support"
-	select BRIDGE_NF_EBTABLES_LEGACY
+	depends on BRIDGE_NF_EBTABLES_LEGACY
 	help
 	  The ebtables filter table is used to define frame filtering rules at
 	  local input, forwarding and local output. See the man page for
@@ -86,7 +86,7 @@ config BRIDGE_EBT_T_FILTER
 
 config BRIDGE_EBT_T_NAT
 	tristate "ebt: nat table support"
-	select BRIDGE_NF_EBTABLES_LEGACY
+	depends on BRIDGE_NF_EBTABLES_LEGACY
 	help
 	  The ebtables nat table is used to define rules that alter the MAC
 	  source address (MAC SNAT) or the MAC destination address (MAC DNAT).
net/ipv4/netfilter/Kconfig +8 -7
index ef8009281da5c..dcf015e0d4266 100644
@@ -14,6 +14,7 @@ config NF_DEFRAG_IPV4
 config IP_NF_IPTABLES_LEGACY
 	tristate "Legacy IP tables support"
 	default	n
+	depends on NETFILTER_LEGACY
 	select NETFILTER_XTABLES
 	help
 	  iptables is a legacy packet classifier.
@@ -183,7 +184,7 @@ config IP_NF_MATCH_TTL
 config IP_NF_FILTER
 	tristate "Packet filtering"
 	default m if NETFILTER_ADVANCED=n
-	select IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  Packet filtering defines a table `filter', which has a series of
 	  rules for simple packet filtering at local input, forwarding and
@@ -220,10 +221,10 @@ config IP_NF_TARGET_SYNPROXY
 config IP_NF_NAT
 	tristate "iptables NAT support"
 	depends on NF_CONNTRACK
+	depends on IP_NF_IPTABLES_LEGACY
 	default m if NETFILTER_ADVANCED=n
 	select NF_NAT
 	select NETFILTER_XT_NAT
-	select IP_NF_IPTABLES_LEGACY
 	help
 	  This enables the `nat' table in iptables. This allows masquerading,
 	  port forwarding and other forms of full Network Address Port
@@ -264,7 +265,7 @@ endif # IP_NF_NAT
 config IP_NF_MANGLE
 	tristate "Packet mangling"
 	default m if NETFILTER_ADVANCED=n
-	select IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `mangle' table to iptables: see the man page for
 	  iptables(8).  This table is used for various packet alterations
@@ -299,7 +300,7 @@ config IP_NF_TARGET_TTL
 # raw + specific targets
 config IP_NF_RAW
 	tristate  'raw table support (required for NOTRACK/TRACE)'
-	select IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `raw' table to iptables. This table is the very
 	  first in the netfilter framework and hooks in at the PREROUTING
@@ -313,7 +314,7 @@ config IP_NF_SECURITY
 	tristate "Security table"
 	depends on SECURITY
 	depends on NETFILTER_ADVANCED
-	select IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `security' table to iptables, for use
 	  with Mandatory Access Control (MAC) policy.
@@ -325,7 +326,7 @@ endif # IP_NF_IPTABLES
 # ARP tables
 config IP_NF_ARPTABLES
 	tristate "Legacy ARPTABLES support"
-	depends on NETFILTER_XTABLES
+	depends on NETFILTER_XTABLES && NETFILTER_LEGACY
 	default n
 	help
 	  arptables is a legacy packet classifier.
@@ -342,7 +343,7 @@ config IP_NF_ARPFILTER
 	tristate "arptables-legacy packet filtering support"
 	select IP_NF_ARPTABLES
 	select NETFILTER_FAMILY_ARP
-	depends on NETFILTER_XTABLES
+	depends on NETFILTER_XTABLES && NETFILTER_LEGACY
 	help
 	  ARP packet filtering defines a table `filter', which has a series of
 	  rules for simple ARP packet filtering at local input and
net/ipv6/netfilter/Kconfig +7 -6
index e087a8e97ba78..303942174b5d5 100644
@@ -9,8 +9,9 @@ menu "IPv6: Netfilter Configuration"
 # old sockopt interface and eval loop
 config IP6_NF_IPTABLES_LEGACY
 	tristate "Legacy IP6 tables support"
-	depends on INET && IPV6
+	depends on INET && IPV6 && NETFILTER_LEGACY
 	select NETFILTER_XTABLES
+	select IP6_NF_IPTABLES
 	default n
 	help
 	  ip6tables is a legacy packet classifier.
@@ -197,7 +198,7 @@ config IP6_NF_TARGET_HL
 config IP6_NF_FILTER
 	tristate "Packet filtering"
 	default m if NETFILTER_ADVANCED=n
-	select IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	tristate
 	help
 	  Packet filtering defines a table `filter', which has a series of
@@ -234,7 +235,7 @@ config IP6_NF_TARGET_SYNPROXY
 config IP6_NF_MANGLE
 	tristate "Packet mangling"
 	default m if NETFILTER_ADVANCED=n
-	select IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `mangle' table to iptables: see the man page for
 	  iptables(8).  This table is used for various packet alterations
@@ -244,7 +245,7 @@ config IP6_NF_MANGLE
 
 config IP6_NF_RAW
 	tristate  'raw table support (required for TRACE)'
-	select IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `raw' table to ip6tables. This table is the very
 	  first in the netfilter framework and hooks in at the PREROUTING
@@ -258,7 +259,7 @@ config IP6_NF_SECURITY
 	tristate "Security table"
 	depends on SECURITY
 	depends on NETFILTER_ADVANCED
-	select IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `security' table to iptables, for use
 	  with Mandatory Access Control (MAC) policy.
@@ -269,8 +270,8 @@ config IP6_NF_NAT
 	tristate "ip6tables NAT support"
 	depends on NF_CONNTRACK
 	depends on NETFILTER_ADVANCED
+	depends on IP6_NF_IPTABLES_LEGACY
 	select NF_NAT
-	select IP6_NF_IPTABLES_LEGACY
 	select NETFILTER_XT_NAT
 	help
 	  This enables the `nat' table in ip6tables. This allows masquerading,
net/netfilter/x_tables.c +11 -5
index 709840612f0df..24788bd3cbcb6 100644
@@ -1317,12 +1317,13 @@ void xt_compat_unlock(u_int8_t af)
 EXPORT_SYMBOL_GPL(xt_compat_unlock);
 #endif
 
-DEFINE_PER_CPU(seqcount_t, xt_recseq);
-EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
-
 struct static_key xt_tee_enabled __read_mostly;
 EXPORT_SYMBOL_GPL(xt_tee_enabled);
 
+#ifdef CONFIG_NETFILTER_LEGACY
+DEFINE_PER_CPU(seqcount_t, xt_recseq);
+EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
+
 static int xt_jumpstack_alloc(struct xt_table_info *i)
 {
 	unsigned int size;
@@ -1514,6 +1515,7 @@ void *xt_unregister_table(struct xt_table *table)
 	return private;
 }
 EXPORT_SYMBOL_GPL(xt_unregister_table);
+#endif
 
 #ifdef CONFIG_PROC_FS
 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
@@ -1897,6 +1899,7 @@ void xt_proto_fini(struct net *net, u_int8_t af)
 }
 EXPORT_SYMBOL_GPL(xt_proto_fini);
 
+#ifdef CONFIG_NETFILTER_LEGACY
 /**
  * xt_percpu_counter_alloc - allocate x_tables rule counter
  *
@@ -1951,6 +1954,7 @@ void xt_percpu_counter_free(struct xt_counters *counters)
 		free_percpu((void __percpu *)pcnt);
 }
 EXPORT_SYMBOL_GPL(xt_percpu_counter_free);
+#endif
 
 static int __net_init xt_net_init(struct net *net)
 {
@@ -1983,8 +1987,10 @@ static int __init xt_init(void)
 	unsigned int i;
 	int rv;
 
-	for_each_possible_cpu(i) {
-		seqcount_init(&per_cpu(xt_recseq, i));
+	if (IS_ENABLED(CONFIG_NETFILTER_LEGACY)) {
+		for_each_possible_cpu(i) {
+			seqcount_init(&per_cpu(xt_recseq, i));
+		}
 	}
 
 	xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL);
net/netfilter/xt_TCPOPTSTRIP.c +2 -2
index 30e99464171b7..93f064306901c 100644
@@ -91,7 +91,7 @@ tcpoptstrip_tg4(struct sk_buff *skb, const struct xt_action_param *par)
 	return tcpoptstrip_mangle_packet(skb, par, ip_hdrlen(skb));
 }
 
-#if IS_ENABLED(CONFIG_IP6_NF_MANGLE)
+#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
 static unsigned int
 tcpoptstrip_tg6(struct sk_buff *skb, const struct xt_action_param *par)
 {
@@ -119,7 +119,7 @@ static struct xt_target tcpoptstrip_tg_reg[] __read_mostly = {
 		.targetsize = sizeof(struct xt_tcpoptstrip_target_info),
 		.me         = THIS_MODULE,
 	},
-#if IS_ENABLED(CONFIG_IP6_NF_MANGLE)
+#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
 	{
 		.name       = "TCPOPTSTRIP",
 		.family     = NFPROTO_IPV6,
net/netfilter/xt_mark.c +1 -1
index 65b965ca40ea7..59b9d04400cac 100644
@@ -48,7 +48,7 @@ static struct xt_target mark_tg_reg[] __read_mostly = {
 		.targetsize     = sizeof(struct xt_mark_tginfo2),
 		.me             = THIS_MODULE,
 	},
-#if IS_ENABLED(CONFIG_IP_NF_ARPTABLES)
+#if IS_ENABLED(CONFIG_IP_NF_ARPTABLES) || IS_ENABLED(CONFIG_NFT_COMPAT_ARP)
 	{
 		.name           = "MARK",
 		.revision       = 2,
Signature · 1 line
2.49.0
maintainer Applied
#1 lore raw
Quoted 33 lines · Sebastian Andrzej Siewior

From: Pablo Neira Ayuso <[email protected]>

The seqcount xt_recseq is used to synchronize the replacement of xt_table::private in xt_replace_table() against all readers such as ipt_do_table()

To ensure that there is only one writer, the writing side disables bottom halves. The sequence counter can be acquired recursively. Only the first invocation modifies the sequence counter (signaling that a writer is in progress) while the following (recursive) writer does not modify the counter.
The lack of a proper locking mechanism for the sequence counter can lead to live lock on PREEMPT_RT if the high prior reader preempts the writer. Additionally if the per-CPU lock on PREEMPT_RT is removed from local_bh_disable() then there is no synchronisation for the per-CPU sequence counter.

The affected code is "just" the legacy netfilter code which is replaced by "netfilter tables". That code can be disabled without sacrificing functionality because everything is provided by the newer implementation. This will only requires the usage of the "-nft" tools instead of the "-legacy" ones.
The long term plan is to remove the legacy code so lets accelerate the progress.

Relax dependencies on iptables legacy, replace select with depends on, this should cause no harm to existing kernel configs and users can still toggle IP{6}_NF_IPTABLES_LEGACY in any case.
Make EBTABLES_LEGACY, IPTABLES_LEGACY and ARPTABLES depend on NETFILTER_LEGACY. Hide xt_recseq and its users, xt_register_table() and xt_percpu_counter_alloc() behind NETFILTER_LEGACY. Let NETFILTER_LEGACY depend on !PREEMPT_RT.

Quoted 3 lines

Replace CONFIG_IP6_NF_MANGLE->CONFIG_IP6_NF_IPTABLES for TCPOPTSTRIP and add CONFIG_NFT_COMPAT_ARP to the MARK target for the IPv6 and ARP target to keep it enabled without the LEGACY code for NFT.

Applied to nf-next.

Thanks for keeping me as author, I don't deserve it.

author maintainer
#2 lore raw
Quoted 3 lines · Pablo Neira Ayuso

Applied to nf-next.

Thanks for keeping me as author, I don't deserve it.

I've been rebasing my trees on top of v6.16-rc1 and noticed that this patch remained (because it still applies). My other nf patches were dropped because they made it into v6.16-rc1.

Did something happen to this one?

Sebastian

maintainer
#3 lore raw
Quoted 5 lines · Sebastian Andrzej Siewior

I've been rebasing my trees on top of v6.16-rc1 and noticed that this patch remained (because it still applies). My other nf patches were dropped because they made it into v6.16-rc1.

Did something happen to this one?

It had to be dropped due to fallout in net and bpf CI pipelines.

There are problems with kconfig settings.

A small subset of this patch has been upstreamed
c38eb2973c18 ("netfilter: xtables: support arpt_mark and ipv6 optstrip for iptables-nft only builds").

My plan was to zap some of the backwards-compat kconfig knobs that we have and update various selftest config files, then rebase this and retry.

author maintainer
#4 lore raw
Quoted 9 lines · Florian Westphal
Quoted 5 lines · Sebastian Andrzej Siewior

I've been rebasing my trees on top of v6.16-rc1 and noticed that this patch remained (because it still applies). My other nf patches were dropped because they made it into v6.16-rc1.

Did something happen to this one?

It had to be dropped due to fallout in net and bpf CI pipelines.

Oh.

Quoted 8 lines

There are problems with kconfig settings.

A small subset of this patch has been upstreamed
c38eb2973c18 ("netfilter: xtables: support arpt_mark and ipv6 optstrip for iptables-nft only builds").

My plan was to zap some of the backwards-compat kconfig knobs that we have and update various selftest config files, then rebase this and retry.

Anything that I can help with?

Sebastian

maintainer
#5 lore raw
Quoted 8 lines · Sebastian Andrzej Siewior
Quoted 6 lines

A small subset of this patch has been upstreamed
c38eb2973c18 ("netfilter: xtables: support arpt_mark and ipv6 optstrip for iptables-nft only builds").

My plan was to zap some of the backwards-compat kconfig knobs that we have and update various selftest config files, then rebase this and retry.

Anything that I can help with?

So some of the problems with CI pipelines are caused by 'config' settings having something like:

IP_NF_TARGET_TTL=m

... but if you look at net/ipv4/netfilter/Kconfig this is:
config IP_NF_TARGET_TTL
tristate '"TTL" target support'
depends on NETFILTER_ADVANCED && IP_NF_MANGLE
select NETFILTER_XT_TARGET_HL
help
This is a backwards-compatible option for the user's convenience (e.g. when running oldconfig). It selects CONFIG_NETFILTER_XT_TARGET_HL.

... and that doesn't do anything anymore due to IP_NF_MANGLE dependency (thats a legacy thing, so it will be off).

So my plan was to zap those old backwards hints first and update the configs to make sure none of the old symbols remain.

OTOH one could just add the correct config settings.

I need to re-test but the attached updated patch should not omit any of the required features even with legacy=n at least for the net ci.

I suspect that it would make sense to split the config tweaks into a distinct patch, however.

If you have cycles please feel free to work on it, I can most likey not get back to it until 2nd week of July.

From be2d4190d7935d3cecec04e76f93fb01f745c32e Mon Sep 17 00:00:00 2001
From: Pablo Neira Ayuso <[email protected]>
Date: Fri, 4 Apr 2025 17:28:15 +0200
Subject: [PATCH 1/1] netfilter: Exclude LEGACY TABLES on PREEMPT_RT.

The seqcount xt_recseq is used to synchronize the replacement of xt_table::private in xt_replace_table() against all readers such as ipt_do_table()

To ensure that there is only one writer, the writing side disables bottom halves. The sequence counter can be acquired recursively. Only the first invocation modifies the sequence counter (signaling that a writer is in progress) while the following (recursive) writer does not modify the counter.
The lack of a proper locking mechanism for the sequence counter can lead to live lock on PREEMPT_RT if the high prior reader preempts the writer. Additionally if the per-CPU lock on PREEMPT_RT is removed from local_bh_disable() then there is no synchronisation for the per-CPU sequence counter.

The affected code is "just" the legacy netfilter code which is replaced by "netfilter tables". That code can be disabled without sacrificing functionality because everything is provided by the newer implementation. This will only requires the usage of the "-nft" tools instead of the "-legacy" ones.
The long term plan is to remove the legacy code so lets accelerate the progress.

Relax dependencies on iptables legacy, replace select with depends on, this should cause no harm to existing kernel configs and users can still toggle IP{6}_NF_IPTABLES_LEGACY in any case.
Make EBTABLES_LEGACY, IPTABLES_LEGACY and ARPTABLES depend on NETFILTER_LEGACY. Hide xt_recseq and its users, xt_register_table() and xt_percpu_counter_alloc() behind NETFILTER_LEGACY. Let NETFILTER_LEGACY depend on !PREEMPT_RT.

Replace CONFIG_IP6_NF_MANGLE->CONFIG_IP6_NF_IPTABLES for TCPOPTSTRIP and add CONFIG_NFT_COMPAT_ARP to the MARK target for the IPv6 and ARP target to keep it enabled without the LEGACY code for NFT.

Regarding selftests, prepare for non-default IP_TABLES_LEGACY.

Enable relevant iptables config options explicitly, this is needed to avoid breakage when symbols related to iptables-legacy will depend on NETFILTER_LEGACY resp. IP_TABLES_LEGACY.

This also means that the classic tables (Kernel modules) will not be enabled by default, so enable them too.

---
 net/bridge/netfilter/Kconfig                  | 10 ++++----
 net/ipv4/netfilter/Kconfig                    | 24 +++++++++----------
 net/ipv6/netfilter/Kconfig                    | 19 +++++++--------
 net/netfilter/Kconfig                         | 10 ++++++++
 net/netfilter/x_tables.c                      | 16 +++++++++----
 tools/testing/selftests/bpf/config            |  1 +
 tools/testing/selftests/hid/config.common     |  1 +
 tools/testing/selftests/net/config            | 11 +++++++++
 tools/testing/selftests/net/mptcp/config      |  2 ++
 tools/testing/selftests/net/netfilter/config  |  5 ++++
 .../selftests/wireguard/qemu/kernel.config    |  4 ++++
 11 files changed, 71 insertions(+), 32 deletions(-)
net/bridge/netfilter/Kconfig +5 -5
index f16bbbbb9481..60f28e4fb5c0 100644
@@ -42,8 +42,8 @@ config NF_CONNTRACK_BRIDGE
 # old sockopt interface and eval loop
 config BRIDGE_NF_EBTABLES_LEGACY
 	tristate "Legacy EBTABLES support"
-	depends on BRIDGE && NETFILTER_XTABLES
-	default n
+	depends on BRIDGE && NETFILTER_XTABLES_LEGACY
+	default	n
 	help
 	 Legacy ebtables packet/frame classifier.
 	 This is not needed if you are using ebtables over nftables
@@ -65,7 +65,7 @@ if BRIDGE_NF_EBTABLES
 #
 config BRIDGE_EBT_BROUTE
 	tristate "ebt: broute table support"
-	select BRIDGE_NF_EBTABLES_LEGACY
+	depends on BRIDGE_NF_EBTABLES_LEGACY
 	help
 	  The ebtables broute table is used to define rules that decide between
 	  bridging and routing frames, giving Linux the functionality of a
@@ -76,7 +76,7 @@ config BRIDGE_EBT_BROUTE
 
 config BRIDGE_EBT_T_FILTER
 	tristate "ebt: filter table support"
-	select BRIDGE_NF_EBTABLES_LEGACY
+	depends on BRIDGE_NF_EBTABLES_LEGACY
 	help
 	  The ebtables filter table is used to define frame filtering rules at
 	  local input, forwarding and local output. See the man page for
@@ -86,7 +86,7 @@ config BRIDGE_EBT_T_FILTER
 
 config BRIDGE_EBT_T_NAT
 	tristate "ebt: nat table support"
-	select BRIDGE_NF_EBTABLES_LEGACY
+	depends on BRIDGE_NF_EBTABLES_LEGACY
 	help
 	  The ebtables nat table is used to define rules that alter the MAC
 	  source address (MAC SNAT) or the MAC destination address (MAC DNAT).
net/ipv4/netfilter/Kconfig +12 -12
index ef8009281da5..2c438b140e88 100644
@@ -13,8 +13,8 @@ config NF_DEFRAG_IPV4
 # old sockopt interface and eval loop
 config IP_NF_IPTABLES_LEGACY
 	tristate "Legacy IP tables support"
-	default	n
-	select NETFILTER_XTABLES
+	depends on NETFILTER_XTABLES_LEGACY
+	default	m if NETFILTER_XTABLES_LEGACY
 	help
 	  iptables is a legacy packet classifier.
 	  This is not needed if you are using iptables over nftables
@@ -182,8 +182,8 @@ config IP_NF_MATCH_TTL
 # `filter', generic and specific targets
 config IP_NF_FILTER
 	tristate "Packet filtering"
-	default m if NETFILTER_ADVANCED=n
-	select IP_NF_IPTABLES_LEGACY
+	default m if NETFILTER_ADVANCED=n || IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  Packet filtering defines a table `filter', which has a series of
 	  rules for simple packet filtering at local input, forwarding and
@@ -220,10 +220,10 @@ config IP_NF_TARGET_SYNPROXY
 config IP_NF_NAT
 	tristate "iptables NAT support"
 	depends on NF_CONNTRACK
+	depends on IP_NF_IPTABLES_LEGACY
 	default m if NETFILTER_ADVANCED=n
 	select NF_NAT
 	select NETFILTER_XT_NAT
-	select IP_NF_IPTABLES_LEGACY
 	help
 	  This enables the `nat' table in iptables. This allows masquerading,
 	  port forwarding and other forms of full Network Address Port
@@ -263,8 +263,8 @@ endif # IP_NF_NAT
 # mangle + specific targets
 config IP_NF_MANGLE
 	tristate "Packet mangling"
-	default m if NETFILTER_ADVANCED=n
-	select IP_NF_IPTABLES_LEGACY
+	default m if NETFILTER_ADVANCED=n || IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `mangle' table to iptables: see the man page for
 	  iptables(8).  This table is used for various packet alterations
@@ -299,7 +299,7 @@ config IP_NF_TARGET_TTL
 # raw + specific targets
 config IP_NF_RAW
 	tristate  'raw table support (required for NOTRACK/TRACE)'
-	select IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `raw' table to iptables. This table is the very
 	  first in the netfilter framework and hooks in at the PREROUTING
@@ -313,7 +313,7 @@ config IP_NF_SECURITY
 	tristate "Security table"
 	depends on SECURITY
 	depends on NETFILTER_ADVANCED
-	select IP_NF_IPTABLES_LEGACY
+	depends on IP_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `security' table to iptables, for use
 	  with Mandatory Access Control (MAC) policy.
@@ -325,8 +325,8 @@ endif # IP_NF_IPTABLES
 # ARP tables
 config IP_NF_ARPTABLES
 	tristate "Legacy ARPTABLES support"
-	depends on NETFILTER_XTABLES
-	default n
+	depends on NETFILTER_XTABLES_LEGACY
+	default	n
 	help
 	  arptables is a legacy packet classifier.
 	  This is not needed if you are using arptables over nftables
@@ -342,7 +342,7 @@ config IP_NF_ARPFILTER
 	tristate "arptables-legacy packet filtering support"
 	select IP_NF_ARPTABLES
 	select NETFILTER_FAMILY_ARP
-	depends on NETFILTER_XTABLES
+	depends on NETFILTER_XTABLES_LEGACY
 	help
 	  ARP packet filtering defines a table `filter', which has a series of
 	  rules for simple ARP packet filtering at local input and
net/ipv6/netfilter/Kconfig +9 -10
index e087a8e97ba7..276860f65baa 100644
@@ -9,9 +9,8 @@ menu "IPv6: Netfilter Configuration"
 # old sockopt interface and eval loop
 config IP6_NF_IPTABLES_LEGACY
 	tristate "Legacy IP6 tables support"
-	depends on INET && IPV6
-	select NETFILTER_XTABLES
-	default n
+	depends on INET && IPV6 && NETFILTER_XTABLES_LEGACY
+	default	m if NETFILTER_XTABLES_LEGACY
 	help
 	  ip6tables is a legacy packet classifier.
 	  This is not needed if you are using iptables over nftables
@@ -196,8 +195,8 @@ config IP6_NF_TARGET_HL
 
 config IP6_NF_FILTER
 	tristate "Packet filtering"
-	default m if NETFILTER_ADVANCED=n
-	select IP6_NF_IPTABLES_LEGACY
+	default m if NETFILTER_ADVANCED=n || IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	tristate
 	help
 	  Packet filtering defines a table `filter', which has a series of
@@ -233,8 +232,8 @@ config IP6_NF_TARGET_SYNPROXY
 
 config IP6_NF_MANGLE
 	tristate "Packet mangling"
-	default m if NETFILTER_ADVANCED=n
-	select IP6_NF_IPTABLES_LEGACY
+	default m if NETFILTER_ADVANCED=n || IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `mangle' table to iptables: see the man page for
 	  iptables(8).  This table is used for various packet alterations
@@ -244,7 +243,7 @@ config IP6_NF_MANGLE
 
 config IP6_NF_RAW
 	tristate  'raw table support (required for TRACE)'
-	select IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `raw' table to ip6tables. This table is the very
 	  first in the netfilter framework and hooks in at the PREROUTING
@@ -258,7 +257,7 @@ config IP6_NF_SECURITY
 	tristate "Security table"
 	depends on SECURITY
 	depends on NETFILTER_ADVANCED
-	select IP6_NF_IPTABLES_LEGACY
+	depends on IP6_NF_IPTABLES_LEGACY
 	help
 	  This option adds a `security' table to iptables, for use
 	  with Mandatory Access Control (MAC) policy.
@@ -269,8 +268,8 @@ config IP6_NF_NAT
 	tristate "ip6tables NAT support"
 	depends on NF_CONNTRACK
 	depends on NETFILTER_ADVANCED
+	depends on IP6_NF_IPTABLES_LEGACY
 	select NF_NAT
-	select IP6_NF_IPTABLES_LEGACY
 	select NETFILTER_XT_NAT
 	help
 	  This enables the `nat' table in ip6tables. This allows masquerading,
net/netfilter/Kconfig +10 -0
index 2560416218d0..fe114607234e 100644
@@ -762,6 +762,16 @@ config NETFILTER_XTABLES_COMPAT
 
 	   If unsure, say N.
 
+config NETFILTER_XTABLES_LEGACY
+	bool "Netfilter legacy tables support"
+	depends on !PREEMPT_RT
+	help
+	  Say Y here if you still require support for legacy tables. This is
+	  required by the legacy tools (iptables-legacy) and is not needed if
+	  you use iptables over nftables (iptables-nft).
+	  Legacy support is not limited to IP, it also includes EBTABLES and
+	  ARPTABLES.
+
 comment "Xtables combined modules"
 
 config NETFILTER_XT_MARK
net/netfilter/x_tables.c +11 -5
index 709840612f0d..90b7630421c4 100644
@@ -1317,12 +1317,13 @@ void xt_compat_unlock(u_int8_t af)
 EXPORT_SYMBOL_GPL(xt_compat_unlock);
 #endif
 
-DEFINE_PER_CPU(seqcount_t, xt_recseq);
-EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
-
 struct static_key xt_tee_enabled __read_mostly;
 EXPORT_SYMBOL_GPL(xt_tee_enabled);
 
+#ifdef CONFIG_NETFILTER_XTABLES_LEGACY
+DEFINE_PER_CPU(seqcount_t, xt_recseq);
+EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
+
 static int xt_jumpstack_alloc(struct xt_table_info *i)
 {
 	unsigned int size;
@@ -1514,6 +1515,7 @@ void *xt_unregister_table(struct xt_table *table)
 	return private;
 }
 EXPORT_SYMBOL_GPL(xt_unregister_table);
+#endif
 
 #ifdef CONFIG_PROC_FS
 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
@@ -1897,6 +1899,7 @@ void xt_proto_fini(struct net *net, u_int8_t af)
 }
 EXPORT_SYMBOL_GPL(xt_proto_fini);
 
+#ifdef CONFIG_NETFILTER_XTABLES_LEGACY
 /**
  * xt_percpu_counter_alloc - allocate x_tables rule counter
  *
@@ -1951,6 +1954,7 @@ void xt_percpu_counter_free(struct xt_counters *counters)
 		free_percpu((void __percpu *)pcnt);
 }
 EXPORT_SYMBOL_GPL(xt_percpu_counter_free);
+#endif
 
 static int __net_init xt_net_init(struct net *net)
 {
@@ -1983,8 +1987,10 @@ static int __init xt_init(void)
 	unsigned int i;
 	int rv;
 
-	for_each_possible_cpu(i) {
-		seqcount_init(&per_cpu(xt_recseq, i));
+	if (IS_ENABLED(CONFIG_NETFILTER_XTABLES_LEGACY)) {
+		for_each_possible_cpu(i) {
+			seqcount_init(&per_cpu(xt_recseq, i));
+		}
 	}
 
 	xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL);
tools/testing/selftests/bpf/config +1 -0
index f74e1ea0ad3b..521836776733 100644
@@ -97,6 +97,7 @@ CONFIG_NF_TABLES_NETDEV=y
 CONFIG_NF_TABLES_IPV4=y
 CONFIG_NF_TABLES_IPV6=y
 CONFIG_NETFILTER_INGRESS=y
+CONFIG_NETFILTER_XTABLES_LEGACY=y
 CONFIG_NF_FLOW_TABLE=y
 CONFIG_NF_FLOW_TABLE_INET=y
 CONFIG_NETFILTER_NETLINK=y
tools/testing/selftests/hid/config.common +1 -0
index b1f40857307d..38c51158adf8 100644
@@ -135,6 +135,7 @@ CONFIG_NET_EMATCH=y
 CONFIG_NETFILTER_NETLINK_LOG=y
 CONFIG_NETFILTER_NETLINK_QUEUE=y
 CONFIG_NETFILTER_XTABLES=y
+CONFIG_NETFILTER_XTABLES_LEGACY=y
 CONFIG_NETFILTER_XT_MATCH_ADDRTYPE=y
 CONFIG_NETFILTER_XT_MATCH_BPF=y
 CONFIG_NETFILTER_XT_MATCH_COMMENT=y
tools/testing/selftests/net/config +11 -0
index 3cfef5153823..c24417d0047b 100644
@@ -30,16 +30,25 @@ CONFIG_NET_FOU=y
 CONFIG_NET_FOU_IP_TUNNELS=y
 CONFIG_NETFILTER=y
 CONFIG_NETFILTER_ADVANCED=y
+CONFIG_NETFILTER_XTABLES_LEGACY=y
 CONFIG_NF_CONNTRACK=m
 CONFIG_IPV6_MROUTE=y
 CONFIG_IPV6_SIT=y
 CONFIG_NF_NAT=m
 CONFIG_IP6_NF_IPTABLES=m
+CONFIG_IP6_NF_IPTABLES_LEGACY=m
 CONFIG_IP_NF_IPTABLES=m
+CONFIG_IP_NF_IPTABLES_LEGACY=m
+CONFIG_IP6_NF_MANGLE=m
+CONFIG_IP6_NF_FILTER=m
 CONFIG_IP6_NF_NAT=m
 CONFIG_IP6_NF_RAW=m
+CONFIG_IP_NF_MANGLE=m
+CONFIG_IP_NF_FILTER=m
 CONFIG_IP_NF_NAT=m
 CONFIG_IP_NF_RAW=m
+CONFIG_IP_NF_TARGET_REJECT=m
+CONFIG_IP6_NF_TARGET_REJECT=m
 CONFIG_IP_NF_TARGET_TTL=m
 CONFIG_IPV6_GRE=m
 CONFIG_IPV6_SEG6_LWTUNNEL=y
@@ -57,6 +66,8 @@ CONFIG_NF_TABLES_IPV6=y
 CONFIG_NF_TABLES_IPV4=y
 CONFIG_NFT_NAT=m
 CONFIG_NETFILTER_XT_MATCH_LENGTH=m
+CONFIG_NETFILTER_XT_TARGET_HL=m
+CONFIG_NETFILTER_XT_NAT=m
 CONFIG_NET_ACT_CSUM=m
 CONFIG_NET_ACT_CT=m
 CONFIG_NET_ACT_GACT=m
tools/testing/selftests/net/mptcp/config +2 -0
index 4f80014cae49..968d440c03fe 100644
@@ -13,6 +13,7 @@ CONFIG_NETFILTER_NETLINK=m
 CONFIG_NF_TABLES=m
 CONFIG_NFT_COMPAT=m
 CONFIG_NETFILTER_XTABLES=m
+CONFIG_NETFILTER_XTABLES_LEGACY=y
 CONFIG_NETFILTER_XT_MATCH_BPF=m
 CONFIG_NETFILTER_XT_MATCH_LENGTH=m
 CONFIG_NETFILTER_XT_MATCH_STATISTIC=m
@@ -25,6 +26,7 @@ CONFIG_IP_MULTIPLE_TABLES=y
 CONFIG_IP_NF_FILTER=m
 CONFIG_IP_NF_MANGLE=m
 CONFIG_IP_NF_TARGET_REJECT=m
+CONFIG_IP6_NF_TARGET_REJECT=m
 CONFIG_IPV6_MULTIPLE_TABLES=y
 CONFIG_IP6_NF_FILTER=m
 CONFIG_NET_ACT_CSUM=m
tools/testing/selftests/net/netfilter/config +5 -0
index 363646f4fefe..c981d2a38ed6 100644
@@ -1,6 +1,8 @@
 CONFIG_AUDIT=y
 CONFIG_BPF_SYSCALL=y
 CONFIG_BRIDGE=m
+CONFIG_NETFILTER_XTABLES_LEGACY=y
+CONFIG_BRIDGE_NF_EBTABLES_LEGACY=m
 CONFIG_BRIDGE_EBT_BROUTE=m
 CONFIG_BRIDGE_EBT_IP=m
 CONFIG_BRIDGE_EBT_REDIRECT=m
@@ -14,7 +16,10 @@ CONFIG_INET_ESP=m
 CONFIG_IP_NF_MATCH_RPFILTER=m
 CONFIG_IP6_NF_MATCH_RPFILTER=m
 CONFIG_IP_NF_IPTABLES=m
+CONFIG_IP_NF_IPTABLES_LEGACY=m
 CONFIG_IP6_NF_IPTABLES=m
+CONFIG_IP6_NF_IPTABLES_LEGACY=m
+CONFIG_IP_NF_NAT=m
 CONFIG_IP_NF_FILTER=m
 CONFIG_IP6_NF_FILTER=m
 CONFIG_IP_NF_RAW=m
tools/testing/selftests/wireguard/qemu/kernel.config +4 -0
index f314d3789f17..0a5381717e9f 100644
@@ -16,9 +16,13 @@ CONFIG_NETFILTER_ADVANCED=y
 CONFIG_NF_CONNTRACK=y
 CONFIG_NF_NAT=y
 CONFIG_NETFILTER_XTABLES=y
+CONFIG_NETFILTER_XTABLES_LEGACY=y
 CONFIG_NETFILTER_XT_NAT=y
 CONFIG_NETFILTER_XT_MATCH_LENGTH=y
 CONFIG_NETFILTER_XT_MARK=y
+CONFIG_NETFILTER_XT_TARGET_MASQUERADE=m
+CONFIG_IP_NF_TARGET_REJECT=m
+CONFIG_IP6_NF_TARGET_REJECT=m
 CONFIG_IP_NF_IPTABLES=y
 CONFIG_IP_NF_FILTER=y
 CONFIG_IP_NF_MANGLE=y
Signature · 1 line
2.49.0
author maintainer
#6 lore raw
Quoted 15 lines · Florian Westphal

So some of the problems with CI pipelines are caused by 'config' settings having something like:

IP_NF_TARGET_TTL=m

... but if you look at net/ipv4/netfilter/Kconfig this is:
config IP_NF_TARGET_TTL
tristate '"TTL" target support'
depends on NETFILTER_ADVANCED && IP_NF_MANGLE
select NETFILTER_XT_TARGET_HL
help
This is a backwards-compatible option for the user's convenience (e.g. when running oldconfig). It selects CONFIG_NETFILTER_XT_TARGET_HL.

Quoted 5 lines

... and that doesn't do anything anymore due to IP_NF_MANGLE dependency (thats a legacy thing, so it will be off).

So my plan was to zap those old backwards hints first and update the configs to make sure none of the old symbols remain.

Oh.

Quoted 4 lines

OTOH one could just add the correct config settings.

I need to re-test but the attached updated patch should not omit any of the required features even with legacy=n at least for the net ci.

I tested the defconfig + the individual config file from tools/testing/selftests/ and compared the results from -rc4 vs -rc4 and the patch. Additionally checked defconfig + kselftest-merge.

I don't see any "missing" options which are NF related with this patch now.

Two options were always missing:

tools/testing/selftests/net/netfilter/config +1 -1
index c981d2a38ed68..79d5b33966ba1 100644
@@ -97,4 +97,4 @@ CONFIG_XFRM_STATISTICS=y
 CONFIG_NET_PKTGEN=m
 CONFIG_TUN=m
 CONFIG_INET_DIAG=m
-CONFIG_SCTP_DIAG=m
+CONFIG_INET_SCTP_DIAG=m
tools/testing/selftests/tc-testing/config +1 -0
index db176fe7d0c3f..8e902f7f1a181 100644
@@ -21,6 +21,7 @@ CONFIG_NF_NAT=m
 CONFIG_NETFILTER_XT_TARGET_LOG=m
 
 CONFIG_NET_SCHED=y
+CONFIG_IP_SET=m
 
 #
 # Queueing/Scheduling

The CONFIG_SCTP_DIAG switch probably never existed. tc-testing asks for NET_EMATCH_IPSET but this one always required IP_SET. We can either remove them or fix them as suggested. Preferences?

Quoted 2 lines

I suspect that it would make sense to split the config tweaks into a distinct patch, however.

I can split the config tweaks out of it if you want.

Quoted 2 lines

If you have cycles please feel free to work on it, I can most likey not get back to it until 2nd week of July.

If we are not making the legacy bits default, because we want to get rid of them eventually, and therefore we fix the configs for the testsuite:
What about those under arch/*/configs?

Sebastian

maintainer
#7 lore raw
Quoted 23 lines · Sebastian Andrzej Siewior
tools/testing/selftests/net/netfilter/config +1 -1
index c981d2a38ed68..79d5b33966ba1 100644
@@ -97,4 +97,4 @@ CONFIG_XFRM_STATISTICS=y
 CONFIG_NET_PKTGEN=m
 CONFIG_TUN=m
 CONFIG_INET_DIAG=m
-CONFIG_SCTP_DIAG=m
+CONFIG_INET_SCTP_DIAG=m
tools/testing/selftests/tc-testing/config +1 -0
index db176fe7d0c3f..8e902f7f1a181 100644
@@ -21,6 +21,7 @@ CONFIG_NF_NAT=m
 CONFIG_NETFILTER_XT_TARGET_LOG=m
 
 CONFIG_NET_SCHED=y
+CONFIG_IP_SET=m
 
 #
 # Queueing/Scheduling
Quoted 4 lines

The CONFIG_SCTP_DIAG switch probably never existed. tc-testing asks for NET_EMATCH_IPSET but this one always required IP_SET. We can either remove them or fix them as suggested. Preferences?

Fix them as suggested.

Quoted 4 lines
Quoted 2 lines

I suspect that it would make sense to split the config tweaks into a distinct patch, however.

I can split the config tweaks out of it if you want.

That would be my preference, will also help to root-cause in case something breaks.

Quoted 3 lines

If we are not making the legacy bits default, because we want to get rid of them eventually, and therefore we fix the configs for the testsuite:
What about those under arch/*/configs?

I don't see many changes to those, so looks like they can be left alone.