netfilter: nf_tables: Implement jump limit for nft_table_validate
Subsystems (4)
PATCH v3 1/1 netfilter: nf_tables: Implement jump limit for nft_table_validate
Observing https://bugzilla.netfilter.org/show_bug.cgi?id=1665, I was able to reproduce the bug using linux-stable. Summarized, when adding large/repeated jump chains, while still staying under the NFT_JUMP_STACK_SIZE (currently 16), the kernel soons locks up.
From the bug report:
table='loop-test'
nft add table inet $table
nft add chain inet $table test0 '{type filter hook input priority 0; policy accept;}' for((i=1;i<16;i++));do nft add chain inet $table test$i;done
nft add rule inet $table test0 jump test1
for((i=1;i<15;i++));do nft add rule inet $table test$i jump test$((i+1));done nft add rule inet $table test15 tcp dport 8080 drop
After the jump rule is added for 3 to 5 times, the system freezes and even softlockup occurs. for((i=1;i<15;i++));do nft add rule inet $table test$i jump test$((i+1));done for((i=1;i<15;i++));do nft add rule inet $table test$i jump test$((i+1));done for((i=1;i<15;i++));do nft add rule inet $table test$i jump test$((i+1));done
This patch is a different approach than the original proposed approach found in the bug report. Additionally, the limit, namespace specific, is only applied to non-init-net namespaces, with the common use case being to protect against rogue containers.
Add a new counter, total_jump_counter, to nft_ctx. On every call to nft_table_validate() (rule addition time, versus packet inspection time) start the counter at the current sum of all jump counts in all other tables with the same family, as well as netdev.
Increment said counter for every jump encountered during table validation. If the counter ever exceeds the namespaces jump limit *during validation*, gracefully reject the rule with -EMLINK (the same behavior as exceeding NFT_JUMP_STACK_SIZE).
This allows immediate feedback to the user about a bad chain, versus the original idea (from the bug report) of allowing the addition to the table. It keeps the in memory ruleset consistent, versus catching the failure during packet inspection at some unknown point in the future and arbitrarily denying the packet. Most importantly, it removes the original problem of a kernel crash.
The compile time limit NFT_DEFAULT_MAX_TABLE_JUMPS of 65536 was chosen to account for any normal use case, and when this value (and associated stressing loop table) was tested against a 1CPU/256MB machine, the system remained functional.
A sysctl entry net/netfilter/nf_max_table_jumps_netns for the limit was also added for any use cases that may exceed this limit. It is network namespace specific. As it is a control limit, for namespaces with an owner that is non-init_user_ns, this sysctl is read only.
Example output of nft when patch is applied (and count is reached):
Error: Could not process rule: Too many links
add rule inet loop-test test14 jump test15
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
v2: nf_max_table_jumps renamed to nf_max_table_jumps_netns;
Limit namespace specific, only applies to non-init netns; Limit raised to 16k
v3: Fix improper conditional compile; removed unneeded pernet subsys;
Include NFPROTO_IPV4|6 for _INET matches; add selftest script
--- Documentation/networking/netfilter-sysctl.rst | 9 + include/net/netfilter/nf_tables.h | 4 + include/net/netns/netfilter.h | 2 + net/netfilter/nf_tables_api.c | 100 +++++++- net/netfilter/nft_immediate.c | 1 + .../testing/selftests/net/netfilter/Makefile | 1 + .../netfilter/nft_max_table_jumps_netns.sh | 218 ++++++++++++++++++ 7 files changed, 333 insertions(+), 2 deletions(-) create mode 100755 tools/testing/selftests/net/netfilter/nft_max_table_jumps_netns.sh
Documentation/networking/netfilter-sysctl.rst +9 -0
@@ -15,3 +15,12 @@ nf_log_all_netns - BOOLEAN with LOG target; this aims to prevent containers from flooding host kernel log. If enabled, this target also works in other network namespaces. This variable is only accessible from init_net. + +nf_max_table_jumps_netns - INTEGER (count) + default 65536 + + The maximum numbers of jumps a table family (+ netdev) can + have. This only applies to non-init_net namespaces, and is read + only for non-init_user_ns namespaces. Meeting or exceeding this + value will cause additional rules to not be added, with EMLINK being + return to the user.
include/net/netfilter/nf_tables.h +4 -0
@@ -205,6 +205,7 @@ static inline void nft_data_copy(u32 *dst, const struct nft_data *src,* @nla: netlink attributes * @portid: netlink portID of the original message * @seq: netlink sequence number +* @total_jump_count: Found jumps for family set + netdev * @flags: modifiers to new request * @family: protocol family * @level: depth of the chains
@@ -218,6 +219,7 @@ struct nft_ctx {const struct nlattr *const *nla;u32 portid;u32 seq; +u32 total_jump_count;u16 flags;u8 family;u8 level;
@@ -1275,6 +1277,7 @@ static inline void nft_use_inc_restore(u32 *use)* @hgenerator: handle generator state * @handle: table handle * @use: number of chain references to this table +* @total_jump_count: jumps as per last validate * @family:address family * @flags: table flag (see enum nft_table_flags) * @genmask: generation mask
@@ -1294,6 +1297,7 @@ struct nft_table {u64 hgenerator;u64 handle;u32 use; +u32 total_jump_count; u16 family:6, flags:8, genmask:2;
include/net/netns/netfilter.h +2 -0
@@ -15,6 +15,7 @@ struct netns_nf {const struct nf_logger __rcu *nf_loggers[NFPROTO_NUMPROTO ];#ifdef CONFIG_SYSCTL struct ctl_table_header *nf_log_dir_header; +struct ctl_table_header *nf_limit_control_dir_header;#ifdef CONFIG_LWTUNNEL struct ctl_table_header *nf_lwtnl_dir_header;#endif
@@ -33,5 +34,6 @@ struct netns_nf {#if IS_ENABLED (CONFIG_NF_DEFRAG_IPV6 )unsigned int defrag_ipv6_users;#endif + u32 nf_max_table_jumps_netns __maybe_unused; };#endif
net/netfilter/nf_tables_api.c +98 -2
166 lines not on this page · raw message
net/netfilter/nft_immediate.c +1 -0
@@ -260,6 +260,7 @@ static int nft_immediate_validate(const struct nft_ctx *ctx,case NFT_JUMP :case NFT_GOTO : pctx->level++; + pctx->total_jump_count++; err =nft_chain_validate (ctx, data->verdict.chain);if (err <0 )return err;
tools/testing/selftests/net/netfilter/Makefile +1 -0
@@ -23,6 +23,7 @@ TEST_PROGS += nft_concat_range.sh TEST_PROGS += nft_conntrack_helper.sh TEST_PROGS += nft_fib.sh TEST_PROGS += nft_flowtable.sh +TEST_PROGS += nft_max_table_jumps_netns.sh TEST_PROGS += nft_meta.sh TEST_PROGS += nft_nat.sh TEST_PROGS += nft_nat_zones.sh
tools/testing/selftests/net/netfilter/nft_max_table_jumps_netns.sh new file +218 -0
218 lines not on this page · raw message
Signature · 1 line
2.49.0
Pablo Neira Ayuso
maintainerHi,
On Mon, May 12, 2025 at 10:08:56PM -0400, Shaun Brady wrote:
[...]
Quoted 4 lines
Add a new counter,
total_jump_counter, tonft_ctx. On every call tonft_table_validate()(rule addition time, versus packet inspection time) start the counter at the current sum of all jump counts in all other tables with the same family, as well as netdev.
What about the bridge family? If bridged frames are passed up to the IP stack, then these hooks can have basechains with jumps too.
Maybe it is better to have a global limit for all tables, regardless the family, in a non-init-netns?
Quoted 10 lines
Increment said counter for every jump encountered during table validation. If the counter ever exceeds the namespaces jump limit *during validation*, gracefully reject the rule with
-EMLINK(the same behavior as exceedingNFT_JUMP_STACK_SIZE).This allows immediate feedback to the user about a bad chain, versus the original idea (from the bug report) of allowing the addition to the table. It keeps the in memory ruleset consistent, versus catching the failure during packet inspection at some unknown point in the future and arbitrarily denying the packet.
Agreed, I also prefer to enforce this limit from control plane.
Thanks
Florian Westphal
maintainerQuoted 9 lines · Pablo Neira Ayuso
On Mon, May 12, 2025 at 10:08:56PM -0400, Shaun Brady wrote:
[...]Quoted 4 lines
Add a new counter,
total_jump_counter, tonft_ctx. On every call tonft_table_validate()(rule addition time, versus packet inspection time) start the counter at the current sum of all jump counts in all other tables with the same family, as well as netdev.What about the bridge family? If bridged frames are passed up to the IP stack, then these hooks can have basechains with jumps too.
Good point, I forgot about this.
Quoted 2 lines
Maybe it is better to have a global limit for all tables, regardless the family, in a non-init-netns?
Looks like it would be simpler.
The only cases where processing is disjunct is ipv4 vs. ipv6.
And arp. But large arp rulesets are unicorns so we should not bother with that.
Quoted 9 lines · Florian Westphal
Quoted 2 lines
Maybe it is better to have a global limit for all tables, regardless the family, in a non-init-netns?
Looks like it would be simpler.
The only cases where processing is disjunct is ipv4 vs. ipv6.
And arp. But large arp rulesets are unicorns so we should not bother with that.
I'm good with the simpler condition. Let me double check my understanding:
Each table will keep a running jump count. When validating a table
(rule modification), sum every other table in the netns, with the single condition set being
if sibling_table->family == NFPROTO_IPV4 && table->family == NFPROTO_IPV6 ||
sibling_table->family == NFPROTO_IPV6 && table->family == NFPROTO_IPV4
do not include in total for netns (break).
Assuming this logic is sound, I'll do v4.
Thanks!
SB