aboutsummaryrefslogtreecommitdiffstatshomepage
path: root/net/netfilter
AgeCommit message (Collapse)AuthorFilesLines
2026-11-05netfilter: nat: zero newrange before filling it inFlorian Westphal2-4/+4
nf_nat_masquerade_ipv4(), nf_nat_masquerade_ipv6(), netmap_tg4() and netmap_tg6() each build a stack-local struct nf_nat_range2 from the caller-supplied range, inheriting range->flags unmodified. Two of them only memset() the min_addr/max_addr members, the other two don't memset() anything at all, so newrange.base_proto is left as stack garbage. If the inherited flags include NF_NAT_RANGE_PROTO_OFFSET, nf_nat_l4proto_unique_tuple() reads range->base_proto.all to compute the NAT port offset. Neither the NETMAP nor the MASQUERADE targets mask the flags field in their checkentry callback, so this uninitialized stack value can end up feeding the NAT port selection. The per-member memset()s are not enough: they have to enumerate every field that must be cleared, and base_proto was missed when NF_NAT_RANGE_PROTO_OFFSET/base_proto was added to the struct. memset() the whole struct instead, so every member is covered regardless of which ones get explicitly assigned afterwards. There are other places that miss this, but those hardcode .flags so nat core won't read from the uninitialized fields. C99 initializers would be preferrable, but this is a more intrusive change. Lets followup in nf-next and convert others as well. This is an LLM finding. Fixes: 2eb0f624b709 ("netfilter: add NAT support for shifted portmap ranges") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-05netfilter: xt_recent: do not call remove_proc_entry() under recent_lockCen Zhang (Microsoft)1-3/+4
recent_proc_net_exit() removes the /proc/net/xt_recent/<name> entries of a dying network namespace while holding spin_lock_bh(&recent_lock). remove_proc_entry() may sleep: proc_entry_rundown() waits until every task that is still inside the entry's file operations has left. A reader of such an entry is counted in by proc_reg_read() before recent_seq_start() takes recent_lock. So if a read is in flight when the namespace is torn down, cleanup_net() sleeps in atomic context holding the spinlock that the reader is spinning on, and neither side can ever make progress. An unprivileged user can trigger this from a user namespace by adding a rule with the "recent" match and reading its proc file while the namespace is being destroyed. As recent_lock is global, every later attempt to take it spins forever too: the spinning can spread to all CPUs, and network namespace teardown stops system-wide. BUG: sleeping function called from invalid context at kernel/sched/completion.c:116 #3: (recent_lock){+...}-{3:3}, at: recent_net_exit+0x2b/0x140 Workqueue: netns cleanup_net Call Trace: proc_entry_rundown+0x185/0x1b0 remove_proc_entry+0x21a/0x520 recent_net_exit+0x7f/0x140 cleanup_net+0x414/0x8f0 ... rcu: INFO: rcu_preempt self-detected stall on CPU Call Trace: queued_spin_lock_slowpath+0x313/0xc20 recent_seq_start+0x87/0x1b0 seq_read_iter+0x265/0x11e0 proc_reg_read+0x1ab/0x270 Fix by taking recent_mutex instead of recent_lock. According to commit a0e889bb1bdc ("[NETFILTER]: recent match: fix "sleeping function called from invalid context""), recent_mutex is the lock to use here: it protects recent_net->tables and was introduced so that proc entries are created and removed without recent_lock held, because those calls may sleep. hashlimit_proc_net_exit() does the same for xt_hashlimit. Fixes: 665e205c16c1 ("netfilter: xt_recent: fix namespace destroy path") Cc: stable@vger.kernel.org Cc: AutonomousCodeSecurity@microsoft.com Signed-off-by: Cen Zhang (Microsoft) <cenzhang@linux.microsoft.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-05netfilter: nf_tables: fix memleak when rule deletion fails with hardware offloadPablo Neira Ayuso2-12/+12
Allocate flow rule object after transaction has been allocation and rule deactivation has been performed, otherwise memleak is possible. Let the transaction unwind the existing rule deactivation from the abort path, this needs to update nft_flow_rule_destroy() to skip release if flow_rule is null. While at this, remove check for non-null flow_rule in newrule error path now that nft_flow_rule_destroy() is skipped if flow_rule is null. Fixes: 63b48c73ff56 ("netfilter: nf_tables_offload: undo updates if transaction fails") Reported-by: Xingyuan Mo <hdthky0@gmail.com> Link: https://patch.msgid.link/20261028062204.1079089-1-hdthky0@gmail.com/ Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-05netfilter: nf_conntrack_sip: fix OOB read in ct_sip_parse_transport()Joas Antonio dos Santos1-2/+2
ct_sip_parse_transport() compares the value of the "transport=" header parameter against "TCP"/"UDP" with strncasecmp(..., 3), but it never checks that the parsed value is at least 3 bytes long. ct_sip_parse_param() returns matchlen = end - start, where end is the next ';' or, when there is no ';', the end of the payload buffer (limit = dptr + datalen). If the "transport=" value is only 1 or 2 bytes long and runs exactly to the end of the payload (e.g. a REGISTER request whose Contact header ends in ";transport=T" with no trailing CRLF), matchlen is less than 3 while strncasecmp() still reads up to 3 bytes, reading 1-2 bytes past the end of the linearized skb data. Add a matchlen check before the comparison, mirroring the sdp_media_type() pattern a few lines below, so short values cannot reach strncasecmp(). The bug was found by LLM-assisted manual review of the SIP helper's parsers and confirmed with a userspace ASan harness built from the verbatim parser functions, which reports a heap-buffer-overflow read of size 2 in strncasecmp() called from ct_sip_parse_transport(). Fixes: ea45f12a2766d ("[NETFILTER]: nf_conntrack_sip: parse SIP headers properly") Link: https://patch.msgid.link/20260815133104.50369-1-joasantonio108@gmail.com/ Signed-off-by: Joas Antonio dos Santos <joasantonio108@gmail.com> Assisted-by: Claude:claude-opus-5-5 Reviewed-by: Fernando Fernandez Mancera <fmancera@suse.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-05netfilter: nft_set_rbtree: check catchall_sync return before queue_doneXingyuan Mo1-1/+2
nft_rbtree_gc_queue() passes the result of nft_trans_gc_catchall_sync() straight to nft_trans_gc_queue_sync_done(). When the catchall path flushes a full GC batch and nft_trans_gc_alloc() fails, catchall_sync returns NULL and queue_sync_done dereferences trans->net (NULL+0x10). Guard the call like the pipapo GC path already does. Fixes: 7d259f021aaa ("netfilter: nft_set_rbtree: prefer sync gc to async worker") Assisted-by: opencode:deepseek-v4 Signed-off-by: Xingyuan Mo <hdthky0@gmail.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-05netfilter: flowtable: extend ct timeout of offloaded GRE flowsMichail Andreas Michalopoulos1-0/+1
Since commit 03428ca5cee9 ("netfilter: conntrack: rework offload nf_conn timeout extension logic"), the flowtable gc worker extends the timeout of an offloaded conntrack entry from nf_flow_table_extend_ct_timeout() once it falls below 5 minutes. That function only knows about TCP and UDP and hits WARN_ON_ONCE() for any other layer 4 protocol. However, GRE flows can be offloaded too, both via nft_flow_offload and act_ct, since commit 4e8d9584d154 ("netfilter: flowtable: Support GRE"). The timeout of an offloaded GRE entry drops below that threshold as soon as a packet of the flow still goes through conntrack. With ip6gre tunnels using the default encapsulation limit this is ordinary traffic: the flowtable does not parse the Destination Options header in front of GRE, so every packet of the flow takes the classic forwarding path. It also happens for a packet with ttl <= 1, which the flowtable fast path passes up to the classic forwarding path. nf_conntrack_gre_packet() then refreshes ct->timeout to the GRE stream timeout (180 seconds by default), and the next gc run triggers, as reported by syzbot: WARNING: net/netfilter/nf_flow_table_core.c:553 at nf_flow_table_extend_ct_timeout+0x1ee/0x2f0 Workqueue: events_power_efficient nf_flow_offload_work_gc Call Trace: nf_flow_offload_gc_step+0x545/0x7b0 nf_flow_table_iterate+0x13c/0x1f0 nf_flow_offload_work_gc+0x22/0x70 process_scheduled_works+0xc3d/0x1630 worker_thread+0xa47/0xfb0 kthread+0x38b/0x480 Besides the splat, which panics the machine with panic_on_warn, the default branch sets the new timeout to 1 jiffy, so the conntrack entry expires right away and the offloaded flow is torn down. Handle GRE like UDP and extend the timeout to NF_CT_DAY. GRE conntrack has no protocol state to take into account, and NF_CT_DAY is also what flow_offload_add() already sets for every offloaded entry, GRE included. This also restores the behaviour from before commit 03428ca5cee9, when the conntrack gc worker extended offloaded entries of any protocol to NF_CT_DAY. Fixes: 03428ca5cee9 ("netfilter: conntrack: rework offload nf_conn timeout extension logic") Reported-by: syzbot+521f7e7e2b974abde99a@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=521f7e7e2b974abde99a Cc: stable@vger.kernel.org Assisted-by: LLM Signed-off-by: Michail Andreas Michalopoulos <mike.michalo@gmail.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-05netfilter: nf_dup: preserve guard across NFQUEUE reinjectionZihan Xi2-3/+15
NFQUEUE can suspend traversal of a duplicated packet while the task-local duplication guard is set. The output call then returns and clears the guard before a later verdict resumes hook traversal. Save the guard in the queue entry and restore it while reinjection continues hook traversal. Preserve and restore the verdict task's prior state, including nested synchronous reinjection. Fixes: cd58bcd9787e ("netfilter: xt_TEE: have cloned packet travel through Xtables too") Cc: stable@vger.kernel.org Reported-by: Vega <vega@nebusec.ai> Assisted-by: LLM Co-developed-by: Luxing Yin <root@tr0jan.top> Signed-off-by: Luxing Yin <root@tr0jan.top> Signed-off-by: Zihan Xi <zihanx@nebusec.ai> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-04ipvs: fix protocol data lifetime during namespace teardownChengfeng Ye1-6/+6
IPVS frees its per-net protocol data before control cleanup cancels defense work and unregisters the sysctl table. A defense worker can load a protocol data pointer, then namespace cleanup can unlink and free that object before the worker dereferences pd->pp or pd->next. The worker's securetcp_lock does not serialize with protocol cleanup. A concurrent defense-mode sysctl write can reach the same timeout update path. KASAN reported: BUG: KASAN: slab-use-after-free in ip_vs_protocol_timeout_change Workqueue: events_long defense_work_handler Call Trace: ip_vs_protocol_timeout_change+0x1b4/0x1e0 update_defense_level+0x63e/0xd80 defense_work_handler+0x1e/0xc0 Allocated by task 125: ip_vs_protocol_net_init+0xda/0x2f0 __ip_vs_init+0x16b/0x240 setup_net+0xfc/0x310 Freed by task 12: ip_vs_protocol_net_cleanup+0x1d6/0x2e0 __ip_vs_cleanup_batch+0x7d/0x100 cleanup_net+0x38c/0x770 Run control cleanup before protocol cleanup so that defense work and active sysctl handlers have finished before their protocol data is freed. Initialize protocols before exposing the control interface, and unwind in reverse order, to enforce the same lifetime on initialization failure. Protocol initialization and exit do not depend on control state. Fixes: 9330419d9aa4 ("IPVS: netns, use ip_vs_proto_data as param.") Cc: stable@vger.kernel.org Signed-off-by: Chengfeng Ye <nicoyip.dev@gmail.com> Acked-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-04ipvs: Defer application parent freeing until after RCU readersChengfeng Ye1-1/+1
unregister_ip_vs_app() removes each incarnation from its protocol list and uses call_rcu() to defer freeing it, but frees the parent application immediately. An RCU reader that already found an incarnation can still access the parent through inc->app in ip_vs_app_inc_get(). During FTP helper module unload, the following interleaving is possible: CPU 0 (RCU reader) CPU 1 (module unload) tcp_app_conn_bind() find inc in protocol list unregister_ip_vs_app() ip_vs_app_inc_release(inc) list_del_rcu(&inc->p_list) call_rcu(&inc->rcu_head, ...) kfree(a) ip_vs_app_inc_get(inc) try_module_get(inc->app->module) The helper module is already going away, so try_module_get() would fail, but evaluating its argument first reads the freed parent. The subsequent rcu_barrier() in pernet unregistration cannot protect this earlier free. KASAN reported: BUG: KASAN: slab-use-after-free in ip_vs_app_inc_get+0x7c/0x90 Call Trace: ip_vs_app_inc_get+0x7c/0x90 tcp_app_conn_bind+0x1bc/0x290 ip_vs_conn_new+0x1915/0x20d0 ip_vs_schedule+0x697/0xea0 tcp_conn_schedule+0x489/0x820 ip_vs_in_hook+0x7bf/0x1f40 Allocated by task 88: kmemdup_noprof+0x20/0x50 register_ip_vs_app+0x12d/0x2c0 __ip_vs_ftp_init+0x56/0x160 [ip_vs_ftp] Freed by task 104: kfree+0x131/0x3c0 unregister_ip_vs_app+0x2f4/0x5b0 unregister_pernet_operations+0x232/0x490 unregister_pernet_subsys+0x1c/0x30 __do_sys_delete_module+0x346/0x510 Use kfree_rcu() with the existing rcu_head to keep the parent alive until these readers finish. Successful helper references already prevent normal module unload, and incarnation RCU callbacks do not access the parent. Fixes: 363c97d7435e ("ipvs: convert app locks") Cc: stable@vger.kernel.org Signed-off-by: Chengfeng Ye <nicoyip.dev@gmail.com> Acked-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-11-01netfilter: nft_reject: prevent unbounded recursion in output hooksFernando Fernandez Mancera2-0/+2
nftables payload-mangling rules can rewrite packet contents (such as TCP flags) before they are evaluated by reject targets. This defeats the content-based loop guards, allowing unbounded hook recursion that leads to kernel stack exhaustion and panics. Fix this by extracting the per-CPU xmit recursion limit previously used by the netdev duplication/forwarding paths into a shared header, and apply it to the inet and inet6 reject transmit functions. It disables BH while transmitting the skb to address a possible migration to different CPU leading to imbalanced decrementation of the recursion counters. Note that the dev_queue_xmit() path used when CONFIG_BRIDGE_NETFILTER is enabled does not require this recursion guard. In that scenario, the generated reset packet is transmitted directly at Layer 2. Because the packet egresses from a different hook, it cannot synchronously loop. In addition nf_send_unreach() and nf_send_unreach6() are safe as the ICMP transmit functions are protected against recursion by the icmp_xmit_lock() and icmpv6_xmit_lock() guards. Reported-by: Yu Junzhe <junzheyu1@gmail.com> Closes: https://lore.kernel.org/netfilter-devel/20260921061320.543-1-junzheyu1@gmail.com/ Fixes: bee11dc78fc8 ("netfilter: nft_reject: support for IPv6 and TCP reset") Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-22Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/netPaolo Abeni21-73/+91
Merge in late fixes in preparation for the net-next PR. Conflicts: net/netfilter/nf_conntrack_netlink.c cd37ab22a74e0 ("netfilter: nf_conntrack_netlink: fix inverted IPv6 tuple filter") 1ae2d094b0200 ("netfilter: ctnetlink: fix inverted IPv6 address match in dump filter") Adjacent changes: drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c 8830e65ed46de ("net: use assign_bit() where applicable") 07bd391eb957 ("octeontx2-af: Fix CQ enable state tracking after AQ writes") net/core/sock.c 31c5ea8a3296 ("ipv6: Remove IPV6_ADDRFORM.") 31b564f70d88 ("net: use the full socket in sk_mc_loop()") net/mptcp/protocol.h 0cb6910130a3 ("mptcp: remove thmac from subflow ctx") 7538fa33bb2f ("mptcp: fix subflow bitfield misuse") Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2026-10-20Merge tag 'nf-26-10-19' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nfJakub Kicinski21-73/+91
Pablo Neira Ayuso says: ==================== Netfilter/IPVS fixes for net The following patchset contains Netfilter/IPVS for net: 1) Fix IPv6 nexthdr offset truncation in fragment reassembly, this has been uncovered by LLM, from Jeremy Jean. 2) Restrict checksums in nfqueue to known protocols and offset, after this patch the plan is to re-enable nfqueue in userns, this came via LLM report. from Florian Westhphal. 3) Fix races when handling overload flag in IPVS, this patch is not perfect but Julian considers this improves the situation. This was uncovered by LLM. From Julian Anastasov. 4) Error handling for IPVS sync_threshold sysctl, from Joel Granados. 5) Use skb_vlan_tag_get_id() to retrieve the VLAN id, discard tci bits, when handling VLAN in the flowtable, from Shuangfeng He. 6) Initialize IPCB for inet/ingress reject, uncovered by LLM. 7) Restrict completion of binary search array to anonymous set. 8) Restrict DCCP option matching to avoid out-of-bound read, from Sung Byeongchan. 9) Reset transport header offset in flowtable after tunnel decapsulation to fix issue with qdisc xmit path, from Lorenzo Bianconi. 10) Fix inverted matching in ctnetlink dump filtering, from Liu Chao. * tag 'nf-26-10-19' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf: netfilter: nf_conntrack_netlink: fix inverted IPv6 tuple filter netfilter: flowtable: reset transport header after nf_flow_encap_pop() netfilter: xt_dccp: handle reserved options as single-byte netfilter: nft_set_rbtree: narrow down range completion to anonymous sets netfilter: nf_reject: initialize IPCB at inet ingress netfilter: flowtable: use the vlan id, not the full tci, in the tuple key ipvs: skip committing sync_threshold when proc_dointvec fails ipvs: fix more races around the overload flag netfilter: nft_payload: restrict checksum offsets to known values netfilter: nf_conntrack: avoid truncating IPv6 nexthdr offset ==================== Link: https://patch.msgid.link/20261019150249.1879516-1-pablo@netfilter.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2026-10-20Merge tag 'nf-next-26-10-18' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-nextJakub Kicinski13-125/+272
Pablo Neira Ayuso says: ==================== Netfilter updates for net-next The following patchset contains Netfilter/IPVS updates for net-next: 1) Disable br_netfilter in userns and schedule it for removal in 2027, from Florian Westphal. 2) Update selftest nft_queue.sh to make it more stable on debug kernels, from Jakub Kicinski. 3) Fix a overflow in IPVS sync messages which can only happen in very old 2.6 kernels, from Julian Anastasov. 4) Finer grain packet drop in nfnetlink_queue when dependency objects are going away, from Antonio Ojea. 5) Add a selftest for the nfnetlink_queue packet drop case, also from Antonio Ojea. 6) Flush conntrack gc worker when existing netns to address a unlikely UaF, from Florian Westphal. 7) Switch nf_tables chain types to use rcu, otherwise module removal can trigger UaF, also from Florian. 8) Remove redundant parsing in nft_exthdr for the tcp options case, from Fernando F. Mancera. 9) Remove redundant skb_header_pointer() calls in nft_exthdr for tcp options, from Fernando F. Mancera. 10) Fix conntrack_vrf.sh selftest return value in zone subtest, from Florian Westphal. 11) Restrict raw table in nft_compat to specific family, hook and priority, from Florian Westphal. Prepare the flowtable to support for IPv4 over IPv6 tunneling and vice-versa, this includes 4 patches from Lorenzo Bianconi. 12) Explicitly set inner protocol when popping header, this is not needed for the existing IPIP and IP6IP6 case because skb->protocol remains the same. 13) Simplify tuple tunnel information fill up when parsing outer tunnel IP headers. 14) Add a new encap_proto which determines the outer tunnel protocol family. 15) Prepare flowtable forwarding path to deal with mixed IPv4-IPv6 tunnels. * tag 'nf-next-26-10-18' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next: netfilter: flowtable: unify tunnel push for IPv4 and IPv6 netfilter: flowtable: add encap_proto to flow_offload_tunnel netfilter: flowtable: populate tunnel tuple regardless of inner protocol netfilter: flowtable: set inner protocol when popping tunnel header netfilter: nft_compat: restrict raw table targets/matches selftests: netfilter: fix return value in zone subtest netfilter: nft_exthdr: avoid 60 bytes stack buffer in TCP option mangling netfilter: nft_exthdr: remove redundant op parsing in tcp_set_init netfilter: nf_tables: switch chain_types array to rcu netfilter: conntrack: fix unlikely UaF in gc worker selftests: netfilter: nft_queue: check the scope of the hook drop flush netfilter: nf_queue: limit the hook drop flush to the unregistered hook point ipvs: fix buffer overflow when sending sync messages selftests: netfilter: nft_queue.sh: only queue icmp echo request/reply netfilter: disable br_netfilter in user namespaces ==================== Link: https://patch.msgid.link/20261018005638.1776148-1-pablo@netfilter.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2026-10-19netfilter: nf_conntrack_netlink: fix inverted IPv6 tuple filterLiu Chao1-4/+4
ctnetlink_filter_match_tuple() returns 0 (no match) when the IPv6 source or destination address differs from the filter, because the !ipv6_addr_cmp() test is inverted. ipv6_addr_cmp() returns 0 when the addresses are equal, so the negation makes the branch trigger on a match instead of a mismatch, which is the opposite of the IPv4 path above. Drop the '!' so the IPv6 branch returns 0 only when the addresses differ, matching the IPv4 semantics and the caller's expectation. Fixes: cb8aa9a3affb ("netfilter: ctnetlink: add kernel side filtering for dump") Cc: stable@vger.kernel.org Signed-off-by: Liu Chao <liuc63@xiaopeng.com> Reviewed-by: Weibin Liu <liuwb@xiaopeng.com> Reviewed-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19netfilter: flowtable: reset transport header after nf_flow_encap_pop()Lorenzo Bianconi1-0/+2
nf_flow_encap_pop() pulls vlan/pppoe or ip tunnel headers and resets the network header, but leaves the transport header pointing into the removed outer header. Reset the transport header to the inner L4 header, otherwise later consumers (e.g. GSO segmentation) parse a bogus header. Fixes: 4cd91f7c290f ("netfilter: flowtable: add vlan support") Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@oss.qualcomm.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19netfilter: xt_dccp: handle reserved options as single-byteSung Byeongchan1-1/+1
The DCCP UAPI defines option numbers 3 through 31 as reserved. Like Padding and Mandatory, these options occupy one byte. dccp_find_option() only treats values below 2 as single-byte and interprets the byte after a reserved option as a TLV length. This can skip a later option and invert an XT_DCCP_OPTION rule result. A local production-hook fixture installed the same DROP rule for IPv4 and IPv6. In three fresh boots per kernel, a target-only packet was dropped 20/20 times, while adding reserved option 3 before the same target let 20/20 packets reach the protected receiver. The fixed kernel dropped all trigger packets in three fresh boots and preserved the normal controls. Treat every option up to DCCPO_MAX_RESERVED as single-byte, matching the protocol definition. Triggering requires CONFIG_NETFILTER_XTABLES_LEGACY, CONFIG_NETFILTER_XT_MATCH_DCCP, an option-based rule, and remote protocol-33 reachability. The touched source is unchanged at 0c2669a9f4a1d607e7591ae50ccf3c432a0aff08, where the patch applies and builds cleanly. Fixes: 1d3de414eb20 ("[NETFILTER]: New iptables DCCP protocol header match") Assisted-by: LLM Signed-off-by: Sung Byeongchan <tjdqudcks0424@naver.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19netfilter: nft_set_rbtree: narrow down range completion to anonymous setsPablo Neira Ayuso1-1/+2
For anonymous sets, when adjacent ranges are found, the end element is not added to the set to pack the set representation. Use next start element to complete this interval. Jérémy Jean says: Consider two ranges in a netfilter verdict map: * keys 10-19 accept packets, and * keys 30-39 jump to chain 'victim'. Now, delete the end marker at 20. The lookup array now uses element 30 as the first range's end, so an INTERVAL_END query for key 15 returns element 30, including its jump verdict and chain pointer. This can make KASAN reports a UAF after the GC releases reference chain for 30 that has expired: BUG: KASAN: slab-use-after-free in nft_verdict_dump+0xdd/0x110 Read of size 8 at addr ff110000009dfb58 by task repro/50 ... Call Trace: ... nft_verdict_dump+0xdd/0x110 nft_data_dump+0x4d/0x80 nf_tables_fill_setelem+0x30b/0x8a0 ? __asan_memcpy+0x3c/0x60 nf_tables_fill_setelem_info+0x1b0/0x200 nft_get_set_elem+0x3fd/0x4b0 ... Freed by task 24: ... kfree+0x1a5/0x280 nf_tables_trans_destroy_work+0x2cb/0x520 Fixes: 2aa34191f06f ("netfilter: nft_set_rbtree: use binary search array in get command") Reported-by: Jérémy Jean <Jeremy.Jean@oss.cyber.gouv.fr> Link: https://patch.msgid.link/netfilter-devel/20261005200404.3544062-2-Jeremy.Jean@oss.cyber.gouv.fr/ Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19netfilter: flowtable: use the vlan id, not the full tci, in the tuple keyShuangfeng He1-2/+2
nf_flow_tuple_encap() stores the skb vlan tag in the encap tuple id using skb_vlan_tag_get(), which returns the unmasked TCI including the 3 PCP priority bits. The flowtable entry, however, is programmed by nf_dev_path_info() with vlan_dev_vlan_id(), a plain 12-bit VID. When a NIC delivers RX vlan offload skbs with a non-zero priority (an RTL9617C EPON ONU was observed handing over vid 818 skbs with tci 0x6332, PCP 3), the runtime lookup key never matches the programmed entry, so every packet misses the flow table and software flow offload is silently defeated. Mask both parse sites to the 12-bit VID so the lookup key matches for any priority. The in-header variant has the same problem for frames received with PCP != 0 on a non-offload path. The egress side is unaffected: nf_flow_encap_push() restores tags from the programmed tuple id, which never carried priority bits. Measured on an RTL9617C ONU (PPPoE over VLAN 818 WAN, NATed 6-flow download over a 1G EPON line), before -> after: download throughput 377 Mbps -> 936 Mbps CPU0 %soft at that rate 91.5% -> 70.9% Fixes: 4cd91f7c290f ("netfilter: flowtable: add vlan support") Signed-off-by: Shuangfeng He <huangya90@gmail.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19ipvs: skip committing sync_threshold when proc_dointvec failsJoel Granados1-0/+3
Forward rc value and skip updating sysctl_sync_threshold when proc_dointvec fails. This targets a write such as "5 x" which would fail with -EINVAL but would store the first element anyway. Signed-off-by: Joel Granados <joel.granados@kernel.org> Acked-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19ipvs: fix more races around the overload flagJulian Anastasov16-58/+44
Sashiko warns that we still have races when the dest overload thresholds are updated. If we narrow the range between thresholds the lockless readers (ip_vs_bind_dest and ip_vs_unbind_dest) may use old threshold values and as result to skip the required flag change. To solve the problem ip_vs_bind_dest() and ip_vs_unbind_dest() will have the chance as before to change the overload state not only at the exact threshold positions but at any conn count below the lower and above the upper thresholds. To make it work without entering the locked region for every connection, we introduce l_threshold_val. Together with u_threshold_val we will switch between two states: 1. not overloaded: where l_threshold_val is set to 0 to close the door for more ip_vs_unbind_dest() calls to clear the overload state. 2. overloaded: where u_threshold_val is set to INT_MAX to close the door for more ip_vs_bind_dest() calls to set the overload state. If a race occurs, multiple CPUs can repeat the state change but this should be for short time window. With the introduction of l_threshold_val the IP_VS_DEST_F_OVERLOAD flag and the 'flags' field are not needed anymore. We will use the ip_vs_dest_is_overloaded() helper for all places that read the overload state. Do not check IP_VS_DEST_CF_AVAILABLE in ip_vs_dh.c because we do not hold reference to dest after it is removed and rely on RCU to access the available dests. To reduce the cacheline accesses while selecting destination, keep the mostly used fields (weight and l_threshold_val) at the first read-mostly cacheline together with the hash nodes. Fixes: 8f843441c4e7 ("ipvs: properly update the overload flag on dest edit") Link: https://sashiko.dev/#/patchset/20260810190621.894119-1-pablo%40netfilter.org Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-19netfilter: nft_payload: restrict checksum offsets to known valuesFlorian Westphal1-7/+33
We need to prevent userspace from corrupting e.g. tcp->doff, because many locations in conntrack and conntrack helpers rely on nf_conntrack_in() having validated the packet headers. nft_payload allows to alter headers later which invalidates this assumption. The 'Fixes' commit restricts writes to safe fields, but there is another side channel: the checksum location. Restrict this too. Reported via sashiko/gemini. Fixes: 112e447d17f7 ("netfilter: validate L4 headers after userspace packet writes") Assisted-by: Claude:claude-opus-4-6 Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: flowtable: unify tunnel push for IPv4 and IPv6Lorenzo Bianconi1-56/+87
Refactor nf_flow_tunnel_ipip_push() and nf_flow_tunnel_ip6ip6_push() into nf_flow_tunnel_ip_push() and nf_flow_tunnel_ip6_push(), keying the inner header handling off tuple->tun.inner_proto so both IP-in-IP and IPv6-in-IPv6 inner protocols are supported regardless of the outer address family. Replace nf_flow_tunnel_v4_push() and nf_flow_tunnel_v6_push() with a single nf_flow_tunnel_push() that dispatches on tuple->tun.encap_proto, and set skb->protocol explicitly after pushing the outer header. This is a preliminary patch to support IPv4 over IPv6 and SIT tunnel flowtable offload. Please note IPv4 over IPv6 and SIT tunnel flowtable offloading is not enabled yet. Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@oss.qualcomm.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: flowtable: add encap_proto to flow_offload_tunnelLorenzo Bianconi2-0/+4
Add encap_proto (AF_INET or AF_INET6) to struct flow_offload_tunnel to allow its use as part of the hash table key during flowtable entry lookup. This is a preliminary change to support IPv4 over IPv6 tunneling via the flowtable infrastructure for software acceleration. Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@oss.qualcomm.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: flowtable: populate tunnel tuple regardless of inner protocolLorenzo Bianconi1-10/+10
Store the outer tunnel addresses and inner protocol in the flow tuple unconditionally in nf_flow_tuple_encap(), instead of only when the tunnel is IP-in-IP or IPv6-in-IPv6. Bail out early when the inner protocol is neither IPPROTO_IPIP nor IPPROTO_IPV6. This makes the tuple usable for cross-family tunnels such as IPv4-over-IPv6 and SIT, where the inner protocol differs from the outer address family. This is a preliminary patch to support IPv4 over IPv6 and SIT tunnel flowtable offload. Please note IPv4 over IPv6 and SIT tunnel flowtable offloading is not enabled yet. Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@oss.qualcomm.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: flowtable: set inner protocol when popping tunnel headerLorenzo Bianconi1-0/+5
Pop the outer tunnel header and set skb->protocol to the inner protocol derived from ctx->tun.inner_proto in nf_flow_ip_tunnel_pop(), instead of bailing out unless the tunnel carries IP-in-IP or IPv6-in-IPv6. This is a preliminary patch to support IPv4 over IPv6 and SIT tunnel flowtable offload. Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@oss.qualcomm.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: nft_compat: restrict raw table targets/matchesFlorian Westphal1-0/+15
xt targets and matches declaring .table = "raw" (CT, NOTRACK) rely on only ever running in the legacy raw table's prerouting/output chains, ahead of conntrack. nft_compat_chain_validate_dependency() already enforces this kind of correspondence for "nat" by requiring the chain type to actually be NAT, but did nothing for "raw": a rule wrapping the CT target could be loaded into any hook and any priority via nft_compat, including after the real conntrack hook has already run. The CT target attaches a conntrack template to the skb when none is set yet. Require that a "raw" table dependency actually matches what xtables enforces via the implict/builtin table dependency. Fixes: 0ca743a55991 ("netfilter: nf_tables: add compatibility layer for x_tables") Assisted-by: LLM Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: nft_exthdr: avoid 60 bytes stack buffer in TCP option manglingFernando Fernandez Mancera1-6/+20
nft_exthdr_tcp_set_eval() and nft_exthdr_tcp_strip_eval() fetched the TCP header and options through nft_tcp_header_pointer(), which calls skb_header_pointer() a second time for the full tcphdr_len into a 60 bytes stack buffer. Since commit 28427f368f0e ("netfilter: nft_exthdr: Fix non-linear header modification") both functions then call skb_ensure_writable() and re-read the header from skb->data, so the buffer was never used. Only the 20 bytes fixed header is needed to learn tcphdr_len, so fetch just that, validate the length as before, and let skb_ensure_writable() pull the rest. This also drops the 60 bytes stack buffer from both functions. Comparison of performance in mpps: op linear/non-linear before after delta ================================================================== set, non-linear 4.193 4.348 +3.7% mpps strip, non-linear 4.184 4.434 +6.0% mpps set, linear 5.302 5.427 +2.4% mpps strip, linear 5.452 5.527 +1.4% mpss Comparison of performance in ns/expression: op linear/non-linear before after delta ================================================================== set, non-linear 34.6 25.3 -9.4 ns strip, non-linear 35.2 20.8 -14.3 ns set, linear 23.6 19.7 -3.9 ns strip, linear 22.9 19.8 -3.1 ns ./scripts/bloat-o-meter exthdr_before.o exthdr_after.o add/remove: 0/1 grow/shrink: 3/0 up/down: 233/-207 (26) Function old new delta nft_exthdr_tcp_eval 792 946 +154 nft_exthdr_tcp_strip_eval 825 891 +66 nft_exthdr_tcp_set_eval 602 615 +13 nft_tcp_header_pointer.part.constprop 207 - -207 Total: Before=6527, After=6553, chg +0.40% This also allow the compiler to inline nft_tcp_header_pointer inside nft_exthdr_tcp_eval() which could yield some minimal performance gain. This patch also updates the logic to drop malformed packets, before this patch NFT_BREAK was used as a verdict instead. Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: nft_exthdr: remove redundant op parsing in tcp_set_initFernando Fernandez Mancera1-6/+2
The nft_exthdr_tcp_set_init() function is exclusively used for TCP options. Therefore, parsing the NFTA_EXTHDR_OP attribute from user space is redundant, as the operation type must always be NFT_EXTHDR_OP_TCPOPT. Remove the unnecessary netlink attribute parsing and hardcode priv->op to NFT_EXTHDR_OP_TCPOPT, which simplifies the code and reduces the object size. ./scripts/bloat-o-meter exthdr_before.o exthdr_after.o add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-54 (-54) Function old new delta nft_exthdr_tcp_set_init 412 358 -54 Total: Before=6581, After=6527, chg -0.82% Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: nf_tables: switch chain_types array to rcuFlorian Westphal2-30/+65
LLM quote: The global chain_type array is cleared under the nfnetlink subsystem mutex, but transaction processing reads it while holding only the pernet transaction mutex. nf_tables_chain_type_lookup() and nft_chain_parse_hook() dereference type->name/hook_mask before try_module_get(type->owner). Concurrent unloading of a separate provider such as nft_chain_nat can unregister and free its static nft_chain_type while a transaction is in the unprotected interval, causing a module-memory use-after-free. LLM finding, not observed in practice. Move this under RCU and grab the module owner earlier. Both nft_chain_route and nft_chain_filter are built into nf_tables.ko: only nft_chain_nat needs an extra synchronize_rcu(). Reported-by: Vega <vega@nebusec.ai> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: conntrack: fix unlikely UaF in gc workerFlorian Westphal1-0/+5
When early-drop scan is enabled, gc_worker() reads tmp->ct_net, resolves cnet with nf_ct_pernet(), and reads cnet->count with RCU read lock held but before acquiring a reference to tmp. Its unlikely but possible that the worker gets preempted before reading cnet->count while nf_conntrack_cleanup_net_list manages to unlink all conntracks and returns in the mean time. This allows for the pernet memory to get free'd. On resume, use after free would occur when the worker reads cnet->count. This can be solved either via another synchronize_rcu() in nf_conntrack_cleanup_net_list or via cheaper flush_work(). If work wasn't running no race can happen as next call cannot observe already-unlinked conntrack objects. This is an LLM finding with no reproducer, but it appears to be correct. Fixes: c6dd940b1f74 ("netfilter: allow early drop of assured conntracks") Reported-by: Vega <vega@nebusec.ai> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-18netfilter: nf_queue: limit the hook drop flush to the unregistered hook pointAntonio Ojea6-8/+45
Unregistering a netfilter hook calls nf_queue_nf_hook_drop(), which makes nfqnl_nf_hook_drop() walk every nfqueue instance of the network namespace and flush it, reinjecting every packet that waits for a verdict with NF_DROP. Deleting an nftables base chain of any table, family and hook point therefore drops the packets queued by hooks of unrelated tables, families and hook points. The verdict that userspace sends for them afterwards fails with -ENOENT. Before commit 039b40ee5854 ("netfilter: nf_queue: only call synchronize_net twice if nf_queue is active") nf_queue_nf_hook_drop() received the removed hook and nfqueue only dropped the entries queued by that hook. The commit removed the argument and the comparison because base chain removal while nfqueue is in use was rare at the time. With base chains added and removed by userspace agents while another program holds packets in a queue this is no longer the case. Commit 960632ece694 ("netfilter: convert hook list to an array") then replaced the hook pointer in struct nf_queue_entry with an index into the hook array of the hook point, so a comparison per hook is no longer possible. A comparison per hook point is: the index of an entry is only stale when the array of its own hook point changed. Pass the nf_hook_ops being unregistered down to the queue handler and add nf_hook_cmp() to nfnetlink_queue, a cmpfn for nfqnl_flush() that matches the entries whose state.pf and state.hook are the hook point of that hook. The handler takes the nf_hook_ops rather than the family of the modified array to keep a single argument with NULL meaning "flush everything", which nf_ct_iterate_destroy() uses on module exit and __nft_ct_set_destroy() uses when a ct zone expression goes away and a queued packet may still reference its conntrack template. The cmpfn resolves the family itself. An NFPROTO_INET hook is registered in the IPv4 and the IPv6 array, so it matches the entries of both families; nf_unregister_net_hook() calls __nf_unregister_net_hook() for the two arrays back to back and the second flush finds nothing left. If the shrink of one array fails to allocate, that array keeps its dummy hook and its entries stay valid, so the flush for the other family drops them without need; no entry with a stale index survives. An NFPROTO_INET hook at NF_INET_INGRESS is registered in the netdev ingress array, the same mapping that nf_static_key_inc() applies. nfqueue does not accept the ingress and egress hooks (nft_queue_validate()), the mapping keeps the comparison consistent with the registration. The entries queued from the hook point whose array changed are still dropped, their hook_index is stale. Registering a hook in front of the one that queued a packet still makes nf_reinject() resume at a stale index, so the packet traverses the queueing hook a second time; this is not changed here. Tested with the reproducer script from the thread, unchanged: it queues 20 UDP packets from an inet output chain to a queue program that delays every verdict by 100 ms and changes the ruleset while they wait. Before this change deleting an unrelated ip6 base chain or an unrelated ip6 table delivers 3 of 20 packets and the queue program exits on ENOENT; with it all 20 are delivered. Deleting a base chain at the same hook point (inet output, another priority, in the same or another table) or an ip table with an output base chain still drops the queued packets; deleting an ip6 table with an output base chain or a netdev ingress chain does not. Adding a base chain in front of the queueing one still queues the packets twice. The nft_queue.sh selftest passes, with CONFIG_NETFILTER_NETLINK_QUEUE=y and built as a module. The following patch adds a test for both outcomes that holds the verdicts instead of delaying them. The code and this commit message were written with an LLM coding assistant from a task description that contained the report, the suggestion from the thread and the two commits above. The author reviewed the result, directed the changes and validated them with the reproducer and the selftest in virtme-ng before and after the change. Suggested-by: Florian Westphal <fw@strlen.de> Link: https://lore.kernel.org/netfilter-devel/CABhP=tYtCyr-qzL==98nkmDVZr6x+f8E26V1AAOpMXgZgt0vAQ@mail.gmail.com/ Assisted-by: LLM Signed-off-by: Antonio Ojea <aojea@google.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-14ipvs: fix buffer overflow when sending sync messagesJulian Anastasov1-9/+14
Sashiko reports for possible heap buffer overflow when generating sync message for the v0 and v1 message formats. We should read the cp->flags only once because another CPU can concurrently set the SEQ_MASK between the two reads. Note that IPVS does not set the SEQ_MASK anymore for the ip_vs_ftp.c helper starting from commit 7f1c40757951 ("IPVS: make FTP work with full NAT support") (2.6.36+), so the problem should not be reproducible. Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Link: https://sashiko.dev/#/patchset/20260903004149.1037028-1-pablo%40netfilter.org Link: https://sashiko.dev/#/patchset/20260909111338.44357-1-ja%40ssi.bg Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-10-08Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/netJakub Kicinski2-1/+12
Cross-merge networking fixes after downstream PR (net-7.3-rc7). Conflicts: include/net/strparser.h 0984ebc631792 ("strparser: make sure __strp_recv isn't running before tearing down the parser") 02fd0a2111374 ("net: strparser: removed the aborted bit and aborts counter") https://lore.kernel.org/asYfmLsfy-SCxXjr@sirena.co.uk Adjacent changes: Documentation/networking/strparser.rst 5f193cdabf25 ("docs: networking: strparser: remove general mode") 0984ebc63179 ("strparser: make sure __strp_recv isn't running before tearing down the parser") Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2026-10-08ipv4: validate checksum_start before completing checksumMichael S. Tsirkin2-1/+12
If a packet with bad checksum metadata gets into the ipv4 stack, skb_checksum_help can corrupt the network header and cause a bunch of mischief. This was discovered and reported by Paulos, and has been reporoduced by others independently since. We really shouldn't allow such packets in, but as a defence in depth measure, let's also check before we complete the checksum. A more complete validation at input is forthcoming, but needs more work. Cc: stable@vger.kernel.org Fixes: f43798c27684 ("tun: Allow GSO using virtio_net_hdr") Fixes: bfd5f4a3d605 ("packet: Add GSO/csum offload support.") Reported-by: Paulos Yibelo <habte.yibelo@gmail.com> Closes: https://lore.kernel.org/netdev/20260922030310.8684-2-habte.yibelo@gmail.com/ Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/0a012b4923e189c4c593ef4f471e5ff0edbe9030.1791412497.git.mst@redhat.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2026-10-01Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/netJakub Kicinski12-17/+69
Cross-merge networking fixes after downstream PR (net-7.3-rc6). Conflicts: net/mac80211/tx.c net/mac80211/ieee80211_i.h 8effab902fa34 ("wifi: mac80211: prevent AP VLAN tx from other interfaces") 78b843974fb83 ("wifi: mac80211: report multicast transmission from lookup_ra_sta()") https://lore.kernel.org/arJmcuhpJ69wAvYK@sirena.co.uk drivers/net/ethernet/realtek/r8169_main.c 3cdeaef1754ab ("r8169: disable EEE on RTL8168h/8111h") 8a3c76523e449 ("r8169: add support for phylink") https://lore.kernel.org/ar5vAxEqS8tO8aVT@sirena.org.uk Adjacent changes: rust/kernel/net/netlink.rs 5e5916923759 ("rust: net: netlink: Migrate to zerocopy's IntoBytes") 0923198be4ae ("rust: net: netlink: validate attribute length before casting to `c_int`") Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2026-09-30netfilter: flowtable: restore ieee80211 forward pathPablo Neira Ayuso1-0/+3
Before commit 871df5007eda ("netfilter: flowtable: bail out if forward path cannot be discovered"), there was a fallback to set up a forward path in case .ndo_fill_forward_path fails or DEV_PATH_MTK_WDMA was used. Such fallback was used by commit d787a3e38f01 ("mac80211: add support for .ndo_fill_forward_path"). One possibility is to handle DEV_PATH_MTK_WDMA from the flowtable forward path discovery. However, this is only used internally by drivers to retrieve mtk_wdma information to set up hardware offload. Felix decided to use the .fill_forward_path interface for this purpose due to the lack of a better interface at that time. Add a new DEV_PATH_IEEE80211 path which is offered if the new ieee80211 flag is set on in the struct net_device_path_ctx to restore the flowtable with a ieee80211 netdevice. Handle this new DEV_PATH_IEEE80211 path just like DEV_PATH_ETHERNET and DEV_PATH_DSA, ie. this is the last netdevice in the stack. This new ieee80211 flag is implicitly unset for mtk_ppe and airoha which call dev_fill_forward_path() to retrieve a DEV_PATH_MTK_WDMA path. Fixes: 871df5007eda ("netfilter: flowtable: bail out if forward path cannot be discovered") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30netfilter: flowtable: generalize pending status bitPablo Neira Ayuso2-10/+11
Rename NF_FLOW_HW_PENDING to NF_FLOW_PENDING and use it to inhibit the flowtable GC worker until pending hw offload work has been completed. Apparently, nf_flow_offload_stats() can schedule work to retrieve stats while the flow is being removed by GC. And this bit can also be used in a follow up patch to disable GC until the flow has been fully added in both directions. Revert the reordering done in commit d644b23afe1e ("netfilter: flowtable: publish HW_DEAD after worker is done") to prevent a race between GC and hw offload handler. Fixes: 2c8897953f3b ("netfilter: flowtable: Add pending bit for offload work") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30netfilter: bpf: reject invalid NAT manipulation typesFernando Fernandez Mancera2-2/+6
As bpf_ct_set_nat_info() is not validating the NAT manipulation type a wrong value can be passed directly to nf_nat_setup_info(). This triggers the WARN_ON() at nf_nat_setup_info() and if panic_on_warn isn't set, then IPS_SRC_NAT_DONE is set without adding nat_bysource and conntrack cleanup tries to unlink an uninitialized hlist node. Fix this by checking that NAT manipulation type is correct before calling nf_nat_setup_info(). In addition, if the WARN_ON is hit, return NF_DROP instead of continuing with the processing to avoid similar situations in the future. Reported-by: VEGA <vega@nebusec.ai> Fixes: 0fabd2aa199f ("net: netfilter: add bpf_ct_set_nat_info kfunc helper") Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30netfilter: nft_set_rbtree: skip transaction elements during GCWeiming Shi1-0/+2
Since nft_set_commit_update() runs set commit callbacks before processing NEWSETELEM transactions, nft_rbtree_gc_scan() can observe elements added by the transaction being committed. The scan records an interval end in rbe_end without checking the element's transaction state. A later, unrelated expired start then moves both elements to the expired list. The synchronous GC queue can free the new end element before the transaction subsequently activates it, causing a use-after-free. Only consider elements that are fully active in both generations. This keeps transaction-state elements out of the GC scan and preserves interval pairing across skipped elements. KASAN reports: BUG: KASAN: slab-use-after-free in nft_setelem_activate nft_setelem_activate net/netfilter/nf_tables_api.c:7047 nf_tables_commit net/netfilter/nf_tables_api.c:11137 Allocated by task 130: nft_set_elem_init net/netfilter/nf_tables_api.c:6794 nft_add_set_elem net/netfilter/nf_tables_api.c:7523 Freed by task 130: nft_trans_gc_trans_free net/netfilter/nf_tables_api.c:10506 rcu_core kernel/rcu/tree.c:2919 Fixes: 1e3b9e1c77fe ("netfilter: nf_tables: call set ops .commit when building new ruleset blob") Reported-by: <co+ee5e50ef2670e5f4@bugs.sh> Assisted-by: LLM Signed-off-by: Weiming Shi <bestswngs@gmail.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30ipvs: filter some flags received in the backup serverJulian Anastasov1-3/+30
While the IPVS SYNC protocol is not secure by design we can still protect the backup server from messages that can wreak havoc. This commit addresses problems from received connection flags or their combinations. We now drop messages as follows: 1. the NO_CPORT+TEMPLATE combination allows lookups for normal connections to hit template which can break in many ways. While the master does not sync connections with NO_CPORT flag, i.e. before they are established, we still accept NO_CPORT without TEMPLATE. 2. ONE_PACKET: it is not sent by master, so we do not expect it in backup. Before now it was ignored by IP_VS_CONN_F_BACKUP_MASK for protocol v1 while protocol v0 created connections that are not hashed and dropped immediately. Better to apply the IP_VS_CONN_F_BACKUP_MASK also to the flags from v0 messages for consistency with v1. Fixes: 87375ab47cd0 ("[IPVS]: ip_vs_ftp breaks connections using persistence") Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30ipvs: do not create invisible templatesJulian Anastasov1-0/+3
The IP_VS_CONN_F_ONE_PACKET flag was implemented for normal connections. When conn template inherits this flag from dest->conn_flags it will not be hashed. As result, we will create new template for every new normal connection. Fix it to allow one template to be used by many normal connections. Fixes: 26ec037f9841 ("IPVS: one-packet scheduling") Link: https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260916231652.127456-1-pablo%40netfilter.org Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30ipvs: bound LBLCR and LBLC cache growthZhiling Zou2-0/+6
ip_vs_lblcr_new() and ip_vs_lblc_new() create cache entries for every previously unseen destination address. The table max_size only tells the periodic collector to reclaim entries after the cache has already exceeded the limit. It does not reclaim entries that the attacker continues to use. Reject new cache entries once either table reaches max_size * 3 / 2. The extra headroom lets the periodic collector catch up while the existing scheduler fallback continues to use the selected destination when cache creation fails. New traffic therefore stays serviceable without growing the tables further. Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Cc: stable@vger.kernel.org Reported-by: Vega <vega@nebusec.ai> Suggested-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai> Acked-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30ipvs: fix missing counter decrement in lblcJulian Anastasov1-0/+1
LBLC may delete cache entries for destinations that are removed or overloaded and replace them with available ones. But ip_vs_lblc_new() forgets to decrement the tbl->entries counter after calling ip_vs_lblc_del(). This can lead to increased shrinking of the cache with every new garbage collection. Fixes: 2f3d771a35fe ("ipvs: do not use dest after ip_vs_dest_put in LBLC") Link: https://sashiko.dev/#/patchset/0bdd5abe9968ded7ca2b9cb6844ba83d94cc8d53.1787318053.git.zhilinz%40nebusec.ai Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-30netfilter: nft_flow_offload: drop flowtable reference on init error pathAohan Mei1-1/+6
nft_flow_offload_init() bumps the flowtable use count with nft_use_inc() before calling nf_ct_netns_get(). When the latter fails, the error is returned as-is and the reference is leaked. The upper layers do not balance it either: nf_tables_newexpr() clears expr->ops when the expression init callback fails, so the nft_expr_more() iteration in nft_rule_expr_deactivate() and nf_tables_rule_destroy() stops right before the failed expression and its ->destroy callback, which would drop the reference, never runs. Each failed rule addition therefore leaks one flowtable reference and the flowtable can no longer be removed: NFT_MSG_DELFLOWTABLE keeps reporting -EBUSY even though no rule references it. Save the nf_ct_netns_get() return value and undo the nft_use_inc() when it fails, restoring the inc/dec pairing within nft_flow_offload_init() itself. Fixes: a3c90f7a2323 ("netfilter: nf_tables: flow offload expression") Reported-by: TencentOS Corvus AI <corvus@tencent.com> Cc: stable@vger.kernel.org Assisted-by: CodeBuddy:Kimi-K3 Signed-off-by: Aohan Mei <henrymei@tencent.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-27netfilter: conntrack: make filtering by zone discoverableIlya Maximets1-0/+11
Conntrack flush supports filtering by zone using CTA_ZONE, but this attribute is really hard to use from user space applications. The reason is that it is not possible to tell if it's supported or not. Older kernels silently ignore CTA_ZONE. And in that case they just happily flush all the entries from all zones breaking all the existing connections. So, applications have to infer support from the kernel version. While it works in most cases, it's not a particularly reliable or desired way to check kernel capabilities from applications that aim to be portable. There should be a better way to probe or discover features in the kernel. The CTA_FILTER interface on the other hand is simple enough to probe. We can check for NLM_F_DUMP_FILTERED in the dump to see if filtering is supported. And unknown sub-attributes in CTA_FILTER are rejected explicitly since strict validation is in use there. Let's add new CTA_FILTER_ZONE that signals that CTA_ZONE should be filtered on. It is a flag, since everything in the CTA_FILTER is a bit mask, i.e., a form of a flag. If set, it means that CTA_ZONE must be present and be used for filtering. If the flag is not set however, the filtering on CTA_ZONE will still take place to ensure backwards compatibility. So, the flag doesn't really change the filtering behavior, but it allows user space applications to properly discover support for CTA_ZONE filtering without need to rely on kernel version parsing or risk accidental flushes of the entire conntrack table, and also without modifying the kernel state. A new test variant is added to test with and without the new flag. Since the setup code is moved into a shared function, expectations replaced with assertions to bail early if the base setup fails to avoid the cascade of secondary failures that can be misleading. Error return is only for the SKIP cases. Signed-off-by: Ilya Maximets <i.maximets@ovn.org> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-27netfilter: ctnetlink: fix inverted IPv6 address match in dump filterPiotr Kubik1-4/+4
ctnetlink_filter_match_tuple() rejects an entry when !ipv6_addr_cmp(filter, entry) is true. ipv6_addr_cmp() is a memcmp() and returns 0 for equal addresses, so the condition is true exactly when the addresses match: a CTA_FILTER dump with an IPv6 CTA_IP_SRC or CTA_IP_DST skips every entry that matches the requested address and returns all the others. The IPv4 branch compares with != and behaves correctly. Observed with libnetfilter_conntrack NFCT_FILTER_DUMP_TUPLE: a dump filtered on src 2001:db8::10 returned only the unrelated entries, and a dump filtered on ::1 returned every entry except the ::1 ones. conntrack-tools does not use the tuple filter for -L, which is why this went unnoticed. Use ipv6_addr_equal() so the IPv6 branch mirrors the IPv4 one. Fixes: cb8aa9a3affb ("netfilter: ctnetlink: add kernel side filtering for dump") Cc: stable@vger.kernel.org Signed-off-by: Piotr Kubik <piotr@kubik.pl> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-27netfilter: nfnetlink: Fix for interrupted hook dumpsPhil Sutter3-32/+71
Handling of concurrent hook changes with a dump in progress was problematic in nfnl_hook_dump and entirely broken in nfnl_hook_dump_nat. Address all issues in a single patch to please the review LLM. Introduce sequence numbers in struct netns_nf to replace pointer value-based modification detection which may fail due to memory buffer reuse. This also eliminates the need for most manual cb->seq updates and excessive array index value checks. Since nat hook updates are protected by a different mutex than others, they have to bump their own sequence number. nfnl_hook_dump_nat therefore open-codes what nl_dump_check_consistent does but bumps cb->seq instead of setting NLM_F_DUMP_INTR flag. Dumps of many nat hooks was entirely broken since the array index was not (re)stored. Use cb->args[1] for that and make sure it is reset upon completion as the same dump may loop over multiple arrays of nat hooks. In nfnl_hook_dump, don't signal NLM_F_DUMP_INTR if nfnl_hook_entries_head returns error: This is a permanent condition which does not change during a dump. It is already caught by nfnl_hook_dump_start though, so should not happen anyway. In general, access ops array pointer values using READ_ONCE since they are assigned to using WRITE_ONCE. Avoid setting NLM_F_DUMP_INTR flag in garbage memory by calling nl_dump_check_consistent only for non-empty skbs. If not a single netlink message was created, nfnetlink code will take care of setting the flag. Fixes: e2cf17d3774c ("netfilter: add new hook nfnl subsystem") Fixes: b010e2a4a9ac ("netfilter: nfnetlink_hook: Dump nat type chains") Signed-off-by: Phil Sutter <phil@nwl.cc> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-27netfilter: flowtable: check namespace before iterating flowsQingfang Deng1-10/+7
nf_flow_table_cleanup() walks every registered flow table, checking the network namespace for each flow in nf_flow_table_do_cleanup(). As a result, tables in other namespaces are still iterated and their cleanup work is flushed. Compare the flow table's namespace with the device's namespace in nf_flow_table_cleanup() and skip nonmatching tables. This avoids unnecessary iteration and work flushing, and leaves the per-flow cleanup callback to check only the interface index. Signed-off-by: Qingfang Deng <qingfang.deng@linux.dev> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-27netfilter: conntrack: Untangle drop and invalid countersPhil Sutter1-1/+2
In nf_conntrack_in, 'drop' counter increments if nf_conntrack_handle_packet returns NF_DROP. This is a special case with TCP packets (added by commit 6b69fe0c73c0 ("netfilter: nf_conntrack_tcp: fix endless loop") and not related to invalid packets which are responsible for all the other <=0 returns. So don't increment 'invalid' ounter in this case. Signed-off-by: Phil Sutter <phil@nwl.cc> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2026-09-27netfilter: conntrack: Untangle insert_failed counter from othersPhil Sutter1-2/+0
There is not much sense in incrementing multiple conntrack counters for the same situation. Defining insert_failed as a situation where a valid packet's conntrack can't be confirmed due to unresolvable clash sets it apart from 'drop' (ENOMEM situation) and 'chaintoolong'. Signed-off-by: Phil Sutter <phil@nwl.cc> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>