Since TCP BBR congestion control was introduced in
commit 0f8782ea14 ("tcp_bbr: add BBR congestion control")
it has always been offered as "Dual BSD/GPL":
MODULE_LICENSE("Dual BSD/GPL");
A GPL-2.0-only SPDX header was erroneously added in the recent
commit 2ed4b46b4f ("net: Add SPDX ids to some source files").
This commit revises the tcp_bbr.c SPDX-License-Identifier to note that
this file is licensed as "GPL-2.0 OR BSD-3-Clause".
Fixes: 2ed4b46b4f ("net: Add SPDX ids to some source files")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Van Jacobson <vanj@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: Tim Bird <tim.bird@sony.com>
Link: https://patch.msgid.link/20260531183558.2337381-1-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Conflicts:
drivers/net/ethernet/microsoft/mana/mana_en.c:
17bfe0a8c0 ("net: mana: Add NULL guards in teardown path to prevent panic on attach failure")
d07efe5a6e ("net: mana: Use per-queue allocation for tx_qp to reduce allocation size")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-----BEGIN PGP SIGNATURE-----
iQIzBAABCgAdFiEEH7ZpcWbFyOOp6OJbrB3Eaf9PW7cFAmoZWqgACgkQrB3Eaf9P
W7cz1A//RDEq8pvp1kefBC6YLM9nAEpiIS+gdBWjUty/zC2bpuvWPnEaDKXeZVVx
Vvo9ITV6BsgNsiUEOyM5ehsDknY9TZMFXSawQQWGiRZmGtP+wM3fesoklUDUz+QD
JBaPg7JEcGjFXPlr1X+MF+bvPVfyPaf/s8VEcatFfkPVV2JZPiENwLmxq/ZV3LWF
R5pB0Mz1AreRJQ3IZuUn8ae/UqUQ+GSP3VtI45lrNDWDBeVeP8zT3orm4Tv9ITYm
doNvbXWYhZNlXUcP0qZ887G2Kn6dbrUbsdp0dOnQDAQu2NR0+tYQWxhoCN5Ps3zl
OisDsNEp4aUzwFkwIE84E43rygD6wc7lx+BGgdFUM2FtmxRv7fUiIuvVuCtC87hv
CsK0SueSgog5x3Ltx/P5O+hn80wKAUqPMESb/7Oxja0rUXi251E7WLVNJdgV0t2y
OJMOMFm1uFwsckFBoSi54QNbJkFFK2lvdl+jQ068E7Cqf88LeqtNe56TOLr/Ut7I
UnQakEDnOgzi1HHcpOs/hycyqvPgvBqhRI6IwAtZZFUzQ/i+usmLUIP4AhQRsA9u
ffI/m+7uF4EJ4H+L/FxZds+AMGh28sL6a3muKpYgcHRJ/3bDPOGaL8NHyy+sTfFW
U6GpFqjv2sEWZM8bCN1g7ymNg+70a/xeFwu6/38+X3cP7bg+QgE=
=NQJ5
-----END PGP SIGNATURE-----
Merge tag 'ipsec-2026-05-29' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec
Steffen Klassert says:
====================
pull request (net): ipsec 2026-05-29
1) xfrm: route MIGRATE notifications to caller's netns
Thread the caller's netns through km_migrate() so that
MIGRATE notifications go to the issuing netns, fixing both the
init_net listener leak and MOBIKE notifications inside
non-init netns. From Maoyi Xie.
2) xfrm: ipcomp: Free destination pages on acomp errors
Move the out_free_req label up so that allocated destination
pages are released on decompression errors, not only on success.
From Herbert Xu.
3) xfrm: Check for underflow in xfrm_state_mtu
Reject configurations that cause xfrm_state_mtu() to underflow,
preventing a negative TFCPAD value from becoming a memset size
that triggers an out-of-bounds write of several terabytes.
From David Ahern.
4) xfrm: ah: use skb_to_full_sk in async output callbacks
Convert the possibly-incomplete skb->sk to a full socket pointer
in async AH callbacks so that a request_sock or timewait_sock
never reaches xfrm_output_resume() downstream consumers.
From Michael Bommarito.
5) Add and revert: esp: fix page frag reference leak on skb_to_sgvec failure
The patch does not fix te issue completely.
6) xfrm: esp: restore combined single-frag length gate
Check the aligned post-trailer combined length against a page limit
in the fast path, preventing skb_page_frag_refill() from falling
back to a page too small for the destination scatterlist.
From Jingguo Tan.
7) xfrm: iptfs: reset runtime state when cloning SAs
Reinitialise the clone's mode_data runtime objects before
publishing it, preventing queued skbs from being freed with
list state copied from the original SA when migration fails.
From Shaomin Chen.
8) xfrm: move policy_bydst RCU sync from per-netns .exit to .pre_exit
Flush policy tables and drain the workqueue in a .pre_exit handler
so that cleanup_net() pays one RCU grace period per batch instead
of one per namespace, fixing stalls at high CLONE_NEWNET rates.
From Usama Arif.
9) xfrm: input: hold netns during deferred transport reinjection
Take a netns reference when queueing deferred transport reinjection
work and drop it after the callback completes, keeping the skb->cb
net pointer valid until the deferred work runs.
From Zhengchuan Liang.
* tag 'ipsec-2026-05-29' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec:
Revert "esp: fix page frag reference leak on skb_to_sgvec failure"
xfrm: input: hold netns during deferred transport reinjection
xfrm: move policy_bydst RCU sync from per-netns .exit to .pre_exit
xfrm: iptfs: reset runtime state when cloning SAs
xfrm: esp: restore combined single-frag length gate
esp: fix page frag reference leak on skb_to_sgvec failure
xfrm: ah: use skb_to_full_sk in async output callbacks
xfrm: Check for underflow in xfrm_state_mtu
xfrm: ipcomp: Free destination pages on acomp errors
xfrm: route MIGRATE notifications to caller's netns
====================
Link: https://patch.msgid.link/20260529092648.3878973-1-steffen.klassert@secunet.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This reverts commit 2982e599ff.
The patch does not fully fix the issue and the Author does
not match the 'Signed-off-by:' tag, so revert it for now.
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
In some cases, iptunnel_pmtud_check_icmp() can be called while
skb transport header is not set.
This triggers an out-of-bound access, because
(typeof(skb->transport_header))~0U is 65535.
Access the icmp header based on IPv4 network header,
after making sure icmp->type is present in skb linear part.
Note that iptunnel_pmtud_check_icmpv6()) is fine.
Fixes: 4cb47a8644 ("tunnels: PMTU discovery support for directly bridged IP packets")
Reported-by: Damiano Melotti <melotti@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260522115512.1519110-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sashiko found that iptunnel_pmtud_build_icmp() and
iptunnel_pmtud_build_icmpv6() were caching ip_hdr() and ipv6_hdr()
before an skb_cow() call which can reallocate skb->head.
Fix this possible UAF by initializing the local variables
after the skb_cow() call.
Remove skb_reset_network_header() calls which were not needed.
Fixes: 4cb47a8644 ("tunnels: PMTU discovery support for directly bridged IP packets")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Link: https://patch.msgid.link/20260525201335.2361845-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----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=bkBh
-----END PGP SIGNATURE-----
Merge tag 'nf-next-26-05-25' of https://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next
Florian Westphal says:
====================
netfilter: updates for net-next
The following patchset contains Netfilter fixes and small enhancements:
1) Disable 32-bit x_tables compatibility (32bit binaries on 64bit
kernel) interface in user namespaces. This is 'last warning'
before this is removed for good.
2) Add a configuration toggle for netfilter GCOV profiling. Provide
dedicated toggles for ipset and ipvs.
3) Remove modular support for nfnetlink and restrict it to built-in only.
From Pablo Neira Ayuso.
4) Use per-rule hash initval in nf_conncount. This avoids unecessary
lock contention with short keys (e.g. conntrack zones) in different
namespaces.
5) Use nf_ct_exp_net() in ctnetlink expectation dumps.
From Pratham Gupta.
6) Remove a dead conditional in nft_set_rbtree.
7) Fix conntrack helper policy updates to apply per-class values correctly.
From David Carlier.
8) Fix an off-by-one OOB read in nf_conntrack_irc:parse_dcc(). Use strict
less-than comparison in the newline search loop to respect the
exclusive-end pointer convention. From Muhammad Bilal.
9) Fix typos in nf_conntrack_proto_tcp comments. From Avinash Duduskar.
10) Restore performance optimization in nft_set_pipapo_avx2 by passing
the next map index. Refactor lookup logic for clarity and add a
DEBUG_NET check to document this.
11) Avoid (harmless) u16 overflow in nf_conntrack_ftp when parsing FTP PORT
and EPRT commands. Ignore commands where single octet exceeds 255.
From Giuseppe Caruso.
Patch 12, which removes incorrect (and obviously unused) code from
nft_byteorder was kept back to avoid a net -> net-next merge conflict.
* tag 'nf-next-26-05-25' of https://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next:
netfilter: nf_conntrack_ftp: avoid u16 overflows
netfilter: nft_set_pipapo_avx2: restore performance optimization
netfilter: nf_conntrack_proto_tcp: fix typos in comments
netfilter: nf_conntrack_irc: fix parse_dcc() off-by-one OOB read
netfilter: nfnl_cthelper: apply per-class values when updating policies
netfilter: nft_set_rbtree: remove dead conditional
netfilter: ctnetlink: use nf_ct_exp_net() in expectation dump
netfilter: nf_conncount: use per-rule hash initval
netfilter: allow nfnetlink built-in only
netfilter: add option for GCOV profiling
netfilter: x_tables: disable 32bit compat interface in user namespaces
====================
Link: https://patch.msgid.link/20260525182924.28456-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
/proc/net/igmp walks IPv4 multicast memberships under RCU and
prints im->users without holding RTNL, while multicast join and leave
paths update the field while holding RTNL. Annotate this intentional
lockless snapshot with READ_ONCE() and the matching writers with
WRITE_ONCE().
Signed-off-by: Yuyang Huang <sigefriedhyy@gmail.com>
Reviewed-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260522093906.39764-1-sigefriedhyy@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Similar to a few other subsystems: add a new config toggle to
enable netfilter gcov profiling in netfilter, including ebtables,
arptables and so on.
ipset and ipvs gain their own, dedicated toggles.
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Florian Westphal <fw@strlen.de>
This feature is required to use 32bit arp/ip/ip6/ebtables binaries on
64bit kernels. I don't think there are many users left.
Support has been a compile-time option since 2021 and defaults to off
since 2023.
The XTABLES_COMPAT config option is already off in many distributions
including Debian and Fedora.
Give a few more months before complete removal but disable support in
user namespaces already.
Assisted-by: Claude:claude-sonnet-4-6
Signed-off-by: Florian Westphal <fw@strlen.de>
ipv4_sysctl_exit_net() is currently freeing net->ipv4.sysctl_local_reserved_ports
too soon.
Only after unregister_net_sysctl_table() we can be sure no threads can possibly
use the sysctls, including /proc/sys/net/ipv4/ip_local_reserved_ports.
Fixes: 122ff243f5 ("ipv4: make ip_local_reserved_ports per netns")
Reported-by: Ji'an Zhou <eilaimemedsnaimel@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Link: https://patch.msgid.link/20260521122147.3584624-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The ESP out-of-place fast path appends the trailer in esp_output_head()
before esp_output_tail() allocates the destination page frag. The
head-side gate currently checks skb->data_len and tailen separately, but
the tail code allocates a single destination frag from the combined
post-trailer skb->data_len.
Reject the page-frag fast path when the combined aligned length exceeds a
page. Otherwise skb_page_frag_refill() may fall back to a single page while
the destination sg still spans the combined skb->data_len.
Restore this combined-length page gate for both IPv4 and IPv6.
Fixes: 5bd8baab08 ("esp: limit skb_page_frag_refill use to a single page")
Cc: stable@vger.kernel.org
Signed-off-by: Lin Ma <malin89@huawei.com>
Signed-off-by: Chenyuan Mi <michenyuan@huawei.com>
Signed-off-by: Jingguo Tan <tanjingguo@huawei.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
In esp_output_tail(), when esp->inplace is false, the old skb page frags
are replaced with a new page from the xfrm page_frag cache. The source
scatterlist (sg) is built from the old frags before the replacement, and
esp_ssg_unref() is responsible for releasing the old page references
after the crypto operation completes.
However, if the second skb_to_sgvec() call (which builds the destination
scatterlist from the new page) fails, the code jumps to error_free which
only calls kfree(tmp). The old page frag references captured in the
source scatterlist are never released:
1. sg[] is built from old frags via skb_to_sgvec() (no extra get_page)
2. nr_frags is set to 1 and frag[0] is replaced with the new page
3. Second skb_to_sgvec() fails -> goto error_free
4. kfree(tmp) frees the sg[] memory but old frags are not unref'd
5. kfree_skb() only releases frag[0] (the new page), not the old ones
Fix this by adding a bool parameter to esp_ssg_unref() that, when true,
unconditionally unrefs the source scatterlist frags without checking
req->src and req->dst, since those fields are not yet initialized by
aead_request_set_crypt() at the point of the error. Existing callers
pass false to preserve the original behavior.
The same issue exists in both esp4 and esp6 as the code is identical.
Fixes: cac2661c53 ("esp4: Avoid skb_cow_data whenever possible")
Fixes: 03e2a30f6a ("esp6: Avoid skb_cow_data whenever possible")
Signed-off-by: Alessandro Schino <7991aleschino@gmail.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Two frag-transfer helpers (__pskb_copy_fclone() and skb_shift()) fail
to propagate the SKBFL_SHARED_FRAG bit in skb_shinfo()->flags when
moving frags from source to destination. __pskb_copy_fclone() defers
the rest of the shinfo metadata to skb_copy_header() after copying
frag descriptors, but that helper only carries over gso_{size,segs,
type} and never touches skb_shinfo()->flags; skb_shift() moves frag
descriptors directly and leaves flags untouched. As a result, the
destination skb keeps a reference to the same externally-owned or
page-cache-backed pages while reporting skb_has_shared_frag() as
false.
The mismatch is harmful in any in-place writer that uses
skb_has_shared_frag() to decide whether shared pages must be detoured
through skb_cow_data(). ESP input is one such writer (esp4.c,
esp6.c), and a single nft 'dup to <local>' rule -- or any other
nf_dup_ipv4() / xt_TEE caller -- is enough to land a pskb_copy()'d
skb in esp_input() with the marker stripped, letting an unprivileged
user write into the page cache of a root-owned read-only file via
authencesn-ESN stray writes.
Set SKBFL_SHARED_FRAG on the destination whenever frag descriptors
were actually moved from the source. skb_copy() and skb_copy_expand()
share skb_copy_header() too but linearize all paged data into freshly
allocated head storage and emerge with nr_frags == 0, so
skb_has_shared_frag() returns false on its own; they need no change.
The same omission exists in skb_gro_receive() and skb_gro_receive_list().
The former moves the incoming skb's frag descriptors into the
accumulator's last sub-skb via two paths (a direct frag-move loop and
the head_frag + memcpy path); the latter chains the incoming skb whole
onto p's frag_list. Downstream skb_segment() reads only
skb_shinfo(p)->flags, and skb_segment_list() reuses each sub-skb's
shinfo as the nskb -- both p and lp must carry the marker.
The same omission also exists in tcp_clone_payload(), which builds an
MTU probe skb by moving frag descriptors from skbs on sk_write_queue
into a freshly allocated nskb. The helper falls into the same family
and warrants the same fix for consistency; no TCP TX-side in-place
writer is currently known to reach a user page through this gap, but
a future consumer depending on the marker would regress silently.
The same omission exists in skb_segment(): the per-iteration flag
merge takes only head_skb's flag, and the inner switch that rebinds
frag_skb to list_skb on head_skb-frags exhaustion does not fold the
new frag_skb's flag into nskb. Fold frag_skb's flag at both sites
so segments drawing frags from frag_list members carry the marker.
Fixes: cef401de7b ("net: fix possible wrong checksum generation")
Fixes: f4c50a4034 ("xfrm: esp: avoid in-place decrypt on shared skb frags")
Suggested-by: Sabrina Dubroca <sd@queasysnail.net>
Suggested-by: Sultan Alsawaf <sultan@kerneltoast.com>
Suggested-by: Ben Hutchings <ben@decadent.org.uk>
Suggested-by: Lin Ma <malin89@huawei.com>
Suggested-by: Jingguo Tan <tanjingguo@huawei.com>
Suggested-by: Aaron Esau <aaron1esau@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: Hyunwoo Kim <imv4bel@gmail.com>
Tested-by: Rajat Gupta <rajat.gupta@oss.qualcomm.com>
Link: https://patch.msgid.link/ageeJfJHwgzmKXbh@v4bel
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Blamed commit moved the TIME_WAIT-derived ISN from the skb control
block to a per-CPU variable, assuming the value would always be consumed
by tcp_conn_request() for the same packet that wrote it. That assumption
is violated by multiple drop paths between the producer
(__this_cpu_write(tcp_tw_isn, isn) in tcp_v{4,6}_rcv()) and the consumer
(tcp_conn_request()):
- min_ttl / min_hopcount check
- xfrm policy check
- tcp_inbound_hash() MD5/AO mismatch
- tcp_filter() eBPF/SO_ATTACH_FILTER drop
- th->syn && th->fin discard in tcp_rcv_state_process() TCP_LISTEN
- psp_sk_rx_policy_check() in tcp_v{4,6}_do_rcv()
- tcp_checksum_complete() in tcp_v{4,6}_do_rcv()
- tcp_v{4,6}_cookie_check() returning NULL
When a packet is dropped on any of these paths, tcp_tw_isn is left set.
The next SYN processed on the same CPU then consumes the non zero value in
tcp_conn_request(), receiving a potentially predictable ISN.
This patch moves back tcp_tw_isn to skb->cb[], getting rid of the per-cpu
variable.
Note that tcp_v{4,6}_fill_cb() do not set it.
Very litle impact on overall code size/complexity:
$ scripts/bloat-o-meter -t vmlinux.old vmlinux.new
add/remove: 0/0 grow/shrink: 2/1 up/down: 8/-15 (-7)
Function old new delta
tcp_v6_rcv 3038 3042 +4
tcp_v4_rcv 3035 3039 +4
tcp_conn_request 2938 2923 -15
Total: Before=24436060, After=24436053, chg -0.00%
Fixes: 41eecbd712 ("tcp: replace TCP_SKB_CB(skb)->tcp_tw_isn with a per-cpu field")
Reported-by: Chris Mason <clm@meta.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260519084611.2485277-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It turns out ip_rt_bug() can be called more than expected.
syzbot will still panic (because of panic_on_warn=1), but non debug
kernels will no longer die while repeating stack traces on the console.
Fixes: c378a9c019 ("ipv4: Give backtrace in ip_rt_bug().")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Link: https://patch.msgid.link/20260519193248.4018872-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
syzbot was able to trigger ip_rt_bug() in a loop, using an IPv4 packet
with a crafted IPOPT_SSRR option:
options: ipv4_options {
options: array[ipv4_option] {
union ipv4_option {
ssrr: ipv4_option_route[IPOPT_SSRR] {
type: const = 0x89 (1 bytes)
length: len = 0x7 (1 bytes)
pointer: int8 = 0xa2 (1 bytes)
data: array[ipv4_addr] {
union ipv4_addr {
broadcast: const = 0xffffffff (4 bytes)
}
}
}
}
Change __icmp_send() to not send ICMP to broadcast/multicast destinations.
Fixes: c378a9c019 ("ipv4: Give backtrace in ip_rt_bug().")
Reported-by: syzbot+c13a57c2639c2c0d03a6@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/6a0cc169.170a0220.1f6c2d.0004.GAE@google.com/T/#u
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20260519200836.4141061-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Following the cited commit, __udp_gso_segment() writes single MSS length
in the UDP header.
The cited patch doesn't account for the fact that the last segment could
be a GSO skb by itself. This could happen when the size of the packet is
a multiple of MSS, hence the first segment is also the last one (there
is no need for a remainder skb).
When the post-loop segment is a GSO skb, assign the single MSS length in
the UDP header.
Fixes: b10b446ce7 ("udp: gso: Use single MSS length in UDP header for GSO_PARTIAL")
Reported-by: Matthew Schwartz <matthew.schwartz@linux.dev>
Closes: https://lore.kernel.org/all/6c3fb15e-711d-4b8d-b152-e03d9b05293f@linux.dev/
Tested-by: Matthew Schwartz <matthew.schwartz@linux.dev>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
Link: https://patch.msgid.link/20260518062250.3019914-3-gal@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The cited commit started using msslen for uh->len, but still uses newlen
to adjust uh->check. Although the checksum is ignored in most cases due
to the hardware offload, __udp_gso_segment attempts to maintain the
correct one. Fix uh->check and adjust it by the right value.
Additionally, after the fix, newlen becomes assigned and unused before
the loop. The code can be simplified a bit if mss adjustment is dropped,
so that newlen becomes equal to msslen before the loop, and msslen can
be also dropped, saving a few lines of code.
This brings us back to one variable, drops an unneeded arithmetic for
mss, and fixes the UDP checksum.
Fixes: b10b446ce7 ("udp: gso: Use single MSS length in UDP header for GSO_PARTIAL")
Signed-off-by: Alice Mikityanska <alice@isovalent.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
Link: https://patch.msgid.link/20260518062250.3019914-2-gal@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch continues the effort to refactor workqueue APIs, which has begun
with the changes introducing new workqueues and a new alloc_workqueue flag:
commit 128ea9f6cc ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566 ("workqueue: Add new WQ_PERCPU flag")
The point of the refactoring is to eventually alter the default behavior of
workqueues to become unbound by default so that their workload placement is
optimized by the scheduler.
Before that to happen, workqueue users must be converted to the better named
new workqueues with no intended behaviour changes:
system_wq -> system_percpu_wq
system_unbound_wq -> system_dfl_wq
This way the old obsolete workqueues (system_wq, system_unbound_wq) can be
removed in the future.
Cc: David Ahern <dsahern@kernel.org>
Cc: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260515135143.259669-2-marco.crivellari@suse.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Only handles packets where the pppoe header length field matches the exact
packet length. Significantly improves rx throughput.
When running NAT traffic through a MediaTek MT7621 devices from a host
behind PPPoE to a host directly connected via ethernet, the TCP throughput
that the device is able to handle improves from ~130 Mbit/s to ~630 Mbit/s,
using fraglist GRO.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Qingfang Deng <qingfang.deng@linux.dev>
Tested-by: Pablo Neira Ayuso <pablo@netfilter.org>
Link: https://patch.msgid.link/20260513013400.7467-1-qingfang.deng@linux.dev
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When AH output is offloaded to an asynchronous crypto provider
(hardware accelerators such as AMD CCP, or a forced-async software
shim used for testing), the digest completion fires
ah_output_done() / ah6_output_done() on a workqueue. The egress
skb at that point may have been originated by a TCP listener
sending a SYN-ACK, which sets skb->sk to a request_sock via
skb_set_owner_edemux(); it may also have been originated by an
inet_timewait_sock retransmit. Neither is a full struct sock, and
passing the raw skb->sk to xfrm_output_resume() then forwards a
non-full socket through the rest of the xfrm output chain.
xfrm_output_resume() and its downstream consumers expect a full
sk where they dereference at all. The natural egress path
through ah_output_done() does not crash today because the
consumers that read past sock_common are either gated by
sk_fullsock() or short-circuit on flags that are clear on a fresh
request_sock; an exhaustive walk of the 50 most plausible
consumers under sch_fq, dev_queue_xmit, netfilter, tc-egress and
cgroup-egress BPF found no current unguarded deref. The bug is
still a real type confusion that future consumer changes could
turn into a memory-corruption primitive.
This is the same bug class fixed for ESP in commit 1620c88887
("xfrm: Fix the usage of skb->sk"). Apply the analogous fix to
AH: convert skb->sk to a full socket pointer (or NULL) via
skb_to_full_sk() before handing it to xfrm_output_resume().
The same async AH callbacks were touched recently for an
independent ESN-related ICV layout bug in commit ec54093e6a
("xfrm: ah: account for ESN high bits in async callbacks"); the
sk type-confusion addressed here is orthogonal. This patch is
part of an ongoing audit of the AH callback paths; an ah_output
ihl-validation hardening series is also currently under review on
netdev.
Reproduced under UML + KASAN + lockdep with a forced-async
hmac(sha1) shim that registers at priority 9999 and wraps the
sync in-tree hmac-sha1-lib. With the shim loaded, ah_output_done
runs on every SYN-ACK egress through a transport-mode AH SA and
skb->sk arrives as a request_sock (TCP_NEW_SYN_RECV); after this
patch, xfrm_output_resume() receives the listener (the result of
sk_to_full_sk()) and consumer derefs land on full-sock fields as
intended.
Fixes: 9ab1265d52 ("xfrm: Use actual socket sk instead of skb socket for xfrm_output_resume")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
raw_send_hdrinc() validates that the caller-supplied IPv4 header
fits within the message length:
iphlen = iph->ihl * 4;
err = -EINVAL;
if (iphlen > length)
goto error_free;
if (iphlen >= sizeof(*iph)) {
/* fix up saddr, tot_len, id, csum, transport_header */
}
It does not, however, reject ihl < 5. For such a packet the
"if (iphlen >= sizeof(*iph))" branch is skipped, leaving the
crafted iphdr untouched, but the packet is still handed to
__ip_local_out() and onward. Downstream consumers that read
iph->ihl assume a sane value: net/ipv4/ah4.c:ah_output() in
particular subtracts sizeof(struct iphdr) from top_iph->ihl * 4
and passes the (signed-int-negative, then cast to size_t)
result to memcpy(), producing an OOB access of length close to
SIZE_MAX and a host kernel panic.
An IPv4 header with ihl < 5 is malformed by definition (RFC 791:
"Internet Header Length is the length of the internet header in
32 bit words ... Note that the minimum value for a correct header
is 5."). The kernel should not be willing to inject such a
packet into its own output path.
Reject "iphlen < sizeof(*iph)" alongside the existing
"iphlen > length" check. This matches the principle that locally
constructed packets that re-enter the IP stack must pass the same
basic sanity tests that a foreign packet would be subjected to.
Once this lands, the "if (iphlen >= sizeof(*iph))" wrapper around
the fixup branch becomes redundant; left in place to keep the
patch minimal and backport-friendly. A follow-up can unwrap it.
Note that commit 86f4c90a1c ("ipv4, ipv6: ensure raw socket
message is big enough to hold an IP header") ensures the message
buffer is large enough to hold an iphdr, but does not constrain
the self-reported iph->ihl.
Reachability: the malformed packet source is any caller with
CAP_NET_RAW, including an unprivileged process in a user+net
namespace on a kernel with CONFIG_USER_NS=y. The reproduced AH
crash also requires a matching xfrm AH policy on the outgoing
route; a container granted CAP_NET_ADMIN can install that state
and policy in its netns. Loopback bypasses xfrm_output, so the
trigger uses a real netdev.
Reproduced on UML + KASAN: kernel-mode fault at addr 0x0 with
memcpy_orig at the crash site. Same shape reproduces inside a
rootless Docker container with --cap-add NET_ADMIN on a stock
distro kernel.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Cc: stable@vger.kernel.org
Suggested-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Link: https://patch.msgid.link/77ec2b5e8111961c2c39883c92e8aa2709039c17.1778614451.git.michael.bommarito@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Following patch will use is_skb_wmem() from fq_codel.
Provide __sock_wfree() only if CONFIG_INET=y
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@toke.dk>
Link: https://patch.msgid.link/20260512094859.3673997-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
lockdep_sock_is_held() was added in tcp_ao_established_key()
by the cited commit.
It can be called from tcp_v[46]_timewait_ack() with twsk.
Since it does not have sk->sk_lock, the lockdep annotation
results in out-of-bound access.
$ pahole -C tcp_timewait_sock vmlinux | grep size
/* size: 288, cachelines: 5, members: 8 */
$ pahole -C sock vmlinux | grep sk_lock
socket_lock_t sk_lock; /* 440 192 */
Let's not use lockdep_sock_is_held() for TCP_TIME_WAIT.
Fixes: 6b2d11e2d8 ("net/tcp: Add missing lockdep annotations for TCP-AO hlist traversals")
Reported-by: Damiano Melotti <melotti@google.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260508120853.4098365-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
These functions return a signal whether FIB flushing is required which
must not be ignored. Use the compiler to help with enforcing this
requirement in the future.
Signed-off-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20260507075606.322405-4-cratiu@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When a device is going down or when a net namespace is deleted, all
nexthops on it are removed, and for each nexthop being removed the FIB
table is flushed, which does a full trie traversal looking for entries
marked RTNH_F_DEAD and removing them. This is O(N x R), with N being
number of dev nexthops and R being number of IPv4 routes.
The RTNL is held the entire time.
When there are many nexthops to be removed and many routing entries,
this can result in the RTNL being held for multiple minutes, which
causes unhappiness in other processes trying to acquire the RTNL (e.g.
systemd-networkd for DHCP renewals).
In a complicated deployment with multiple vxlan devices, each having
16K nexthops and a total of 128K ipv4 routes, this is exactly what
happens:
nexthop_flush_dev() # loops over 16K nexthops
-> remove_nexthop()
-> __remove_nexthop()
-> __remove_nexthop_fib() # marks fi->fib_flags |= RTNH_F_DEAD
-> fib_flush() # for EACH nexthop!
-> fib_table_flush() # walks the ENTIRE FIB, 128K entries
This patch makes use of the previously added FIB flushing signal to only
do a single FIB flush after all nexthops to be removed are marked as
RTNH_F_DEAD:
- __remove_nexthop_fib() no longer flushes the FIB.
- nexthop_flush_dev() and flush_all_nexthops() now keep track whether
any nexthop was removed and trigger a FIB flush at the end.
- a new wrapper is defined, remove_one_nexthop() which calls
remove_nexthop() and flushes if necessary. This is intended for places
which must remove a single nexthop and shouldn't worry about the need
to trigger a FIB flush. For now, the only caller is rtm_del_nexthop().
- The two direct callers of __remove_nexthop() get a WARN_ON_ONCE, since
the nh about to be removed should not have any FIB entries referencing
it when replacing or inserting a new one.
This dramatically improves performance from O(N x R) to O(N + R).
Releasing a nexthop reference in remove_nexthop() now no longer frees
it. Instead, it is deleted when the last fib_info pointing to it gets
freed via free_fib_info_rcu(). All routing code is already careful not
to take into consideration routes marked with RTNH_F_DEAD.
Tested with:
DEV=eth2
ip link set up dev $DEV
ip link add testnh0 link $DEV type macvlan mode bridge
ip addr add 198.51.100.1/24 dev testnh0
ip link set testnh0 up
seq 1 65536 | \
sed 's/.*/nexthop add id & via 198.51.100.2 dev testnh0/' | \
ip -batch -
i=1
for a in $(seq 0 255); do
for b in $(seq 0 255); do
echo "route add 10.${a}.${b}.0/32 nhid $i"
i=$((i + 1))
done
done | ip -batch -
time ip link set testnh0 down
ip link del testnh0
Without this patch:
real 0m32.601s
user 0m0.000s
sys 0m32.511s
With this patch:
real 0m0.209s
user 0m0.000s
sys 0m0.153s
Signed-off-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20260507075606.322405-3-cratiu@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Plumb a bool value throughout the various nexthop removal functions,
determined in the innermost __remove_nexthop_fib() (which still does the
FIB flushing) and propagated up all callers.
The next patch will make use of this signal to optimize the removal of
multiple nexthops by moving the FIB flushing up the call hierarchy.
Signed-off-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20260507075606.322405-2-cratiu@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----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=i40t
-----END PGP SIGNATURE-----
Merge tag 'nf-26-05-08' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following batch contains Netfilter fixes for net:
1) Allow initial x_tables table replacement without emitting an audit
log message. Delay the register message until after hooks are wired up
to avoid unnecessary unregister logs during error unwinding.
2) Fix a NULL dereference by allocating hook ops before adding the
table to the per-netns list. Use `synchronize_rcu()` during error
unwinding to ensure the table stops processing packets before
teardown. Defer audit log register message until all operations
succeed.
3) Refactor xtables to use a single `xt_unregister_table_pre_exit`
function. Eliminate code duplication by centralizing table
unregistration logic within the xtables core. ebtables cannot be
changed due to incompatibility.
4) Unregister xtables templates before module removal. This prevents
a race condition where userspace instantiates a new table after the
pernet unreg removed the current table.
5) Add `xtables_unregister_table_exit` to fully unregister netfilter
tables during module removal. Unlink the table from dying lists,
then free hook operations.
6) Implement a two-stage removal scheme for ebtables following the
x_tables pattern. Assign table->ops while holding the ebt mutex to
prevent exposing partially-filled structures.
7) Fix ebtables module initialization race. Register the template last
in table initialization functions. Prevent table instantiation before
pernet operations are available.
8) Fix a race condition in x_tables module initialization. Ensure
pernet ops are fully set up before exposing the table to userspace.
9) Fix a race condition in ebtables module initialization, similar to
previous patch.
10) Restore propagation of helper to expected connection, this is a
fix-for-recent-fix.
11) Validate that the expectation tuple and mask netlink attributes are
present when adding expectation via nfqueue, this fixes a possible
null-ptr-deref.
12) Fix possible rare memleak in the SIP helper in case helper has been
detached from conntrack entry, from Li Xiasong.
13) Fix refcount leak in nft_ct when creating custom expectation, also
from Li Xiason.
Patches 1-9 from Florian Westphal.
10) Restore propagation of helper to expected connection, this is a
fix-for-recent-fix.
11) Check that tuple and mask netlink attributes are set when creating an
expectation via nfqueue.
* tag 'nf-26-05-08' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: nft_ct: fix missing expect put in obj eval
netfilter: nf_conntrack_sip: get helper before allocating expectation
netfilter: ctnetlink: check tuple and mask in expectations created via nfqueue
netfilter: nf_conntrack_expect: restore helper propagation via expectation
netfilter: bridge: eb_tables: close module init race
netfilter: x_tables: close dangling table module init race
netfilter: ebtables: close dangling table module init race
netfilter: ebtables: move to two-stage removal scheme
netfilter: x_tables: add and use xtables_unregister_table_exit
netfilter: x_tables: unregister the templates first
netfilter: x_tables: add and use xt_unregister_table_pre_exit
netfilter: x_tables: allocate hook ops while under mutex
netfilter: x_tables: allow initial table replace without emitting audit log message
====================
Link: https://patch.msgid.link/20260507234509.603182-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When TCP socket migration happens in reqsk_timer_handler(),
@sk_listener will be updated with the new listener.
When we call __inet_csk_reqsk_queue_drop(), the listener must
be the one stored in req->rsk_listener.
The cited commit accidentally replaced oreq->rsk_listener with
sk_listener, leading to imbalanced icsk_accept_queue count.
Let's pass the correct listener to __inet_csk_reqsk_queue_drop().
Fixes: e8c526f2bd ("tcp/dccp: Don't use timer_pending() in reqsk_queue_unlink().")
Reported-by: Damiano Melotti <melotti@google.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260506035954.1563147-3-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When TCP socket migration fails at inet_ehash_insert() in
reqsk_timer_handler(), we jump to the no_ownership: label
and free the new reqsk immediately with __reqsk_free().
Thus, we must stop the new reqsk's timer before jumping to the
label, but the timer might be missed since the cited commit,
resulting in UAF.
As we are in the original reqsk's timer context, we can safely
call timer_delete_sync() for the new reqsk.
Let's pass false to __inet_csk_reqsk_queue_drop() to stop
the new reqsk's timer.
Fixes: 83fccfc394 ("inet: fix potential deadlock in reqsk_queue_unlink()")
Reported-by: Damiano Melotti <melotti@google.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260506035954.1563147-2-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Similar to the previous ebtables patch:
template add exposes the table to userspace, we must do this last to
rnsure the pernet ops are set up (contain the destructors).
Fixes: fdacd57c79 ("netfilter: x_tables: never register tables by default")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Previous change added xtables_unregister_table_pre_exit to detach the
table from the packetpath and to unlink it from the active table list.
In case of rmmod, userspace that is doing set/getsockopt for this table
will not be able to re-instantiate the table:
1. The larval table has been removed already
2. existing instantiated table is no longer on the xt pernet table list.
This adds the second stage helper:
unlink the table from the dying list, free the hook ops (if any) and do
the audit notification. It replaces xt_unregister_table().
Fixes: fdacd57c79 ("netfilter: x_tables: never register tables by default")
Reported-by: Tristan Madani <tristan@talencesecurity.com>
Reviewed-by: Tristan Madani <tristan@talencesecurity.com>
Closes: https://lore.kernel.org/netfilter-devel/20260429175613.1459342-1-tristmd@gmail.com/
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
When the module is going away we need to zap the template
first. Else there is a small race window where userspace
could instantiate a new table after the pernet exit function
has removed the current table.
Fixes: fdacd57c79 ("netfilter: x_tables: never register tables by default")
Reported-by: Tristan Madani <tristan@talencesecurity.com>
Reviewed-by: Tristan Madani <tristan@talencesecurity.com>
Closes: https://lore.kernel.org/netfilter-devel/20260429175613.1459342-1-tristmd@gmail.com/
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Remove the copypasted variants of _pre_exit and add one single
function in the xtables core. ebtables is not compatible with
x_tables and therefore unchanged.
This is a preparation patch to reduce noise in the followup
bug fixes.
Reviewed-by: Tristan Madani <tristan@talencesecurity.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
arp/ip(6)t_register_table() add the table to the per-netns list via
xt_register_table() before allocating the per-netns hook ops copy
via kmemdup_array(). This leaves a window where the table is
visible in the list with ops=NULL.
If the pernet exit happens runs concurrently the pre_exit callback finds
the table via xt_find_table() and passes the NULL ops pointer to
nf_unregister_net_hooks(), causing a NULL dereference:
general protection fault in nf_unregister_net_hooks+0xbc/0x150
RIP: nf_unregister_net_hooks (net/netfilter/core.c:613)
Call Trace:
ipt_unregister_table_pre_exit
iptable_mangle_net_pre_exit
ops_pre_exit_list
cleanup_net
Fix by moving the ops allocation into the xtables core so the table is
never in the list without valid ops. Also ensure the table is no longer
processing packets before its torn down on error unwind.
nf_register_net_hooks might have published at least one hook; call
synchronize_rcu() if there was an error.
audit log register message gets deferred until all operations have
passed, this avoids need to emit another ureg message in case of
error unwinding.
Based on earlier patch by Tristan Madani.
Fixes: f9006acc8d ("netfilter: arp_tables: pass table pointer via nf_hook_ops")
Fixes: ee177a5441 ("netfilter: ip6_tables: pass table pointer via nf_hook_ops")
Fixes: ae68933422 ("netfilter: ip_tables: pass table pointer via nf_hook_ops")
Link: https://lore.kernel.org/netfilter-devel/20260429175613.1459342-1-tristmd@gmail.com/
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
tcp_child_process( .. child ...) currently calls sock_put(child).
Unfortunately @child (named @nsk in callers) can be used after
this point to send a RST packet.
To fix this UAF, I remove the sock_put() from tcp_child_process()
and let the callers handle this after it is safe.
Remove @rsk variable in tcp_v4_do_rcv() and change tcp_v6_do_rcv()
so that both functions look the same.
Fixes: cfb6eeb4c8 ("[TCP]: MD5 Signature Option (RFC2385) support.")
Reported-by: Damiano Melotti <melotti@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260505153927.3435532-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When performing a lockless lookup over the inet_peer rbtree,
if a matching node is found, inet_getpeer() returns it immediately
without validating the seqlock sequence.
This missing check introduces a race condition:
Trigger Path: When a host receives an incoming fragmented IPv4 packet,
ip4_frag_init() (in net/ipv4/ip_fragment.c) calls inet_getpeer_v4()
to track the peer.
The Race: If the packet is from a new source IP, CPU A acquires the
write_seqlock, allocates a new inet_peer node (p), sets its IP address
(daddr), and links it to the rbtree (rb_link_node).
Uninitialized Access: Due to the lack of memory barriers between
rb_link_node and the initialization of the rest of the struct
(like refcount_set(&p->refcnt, 1)), CPU A can make the node visible
to readers before its refcnt is initialized.
This is especially true on weakly-ordered architectures like ARM64
where the CPU can reorder the memory stores.
Lockless Reader: Concurrently, CPU B processes a second fragmented packet
from the same source IP. CPU B does a lockless lookup, finds the newly
inserted node, and returns it immediately.
Use-After-Free (UAF): CPU B reads p->refcnt as uninitialized garbage
(left over from previous kmalloc-128/192 allocations).
If the garbage is > 0, refcount_inc_not_zero(&p->refcnt) succeeds.
CPU A then executes refcount_set(&p->refcnt, 1), overwriting CPU B's increment.
When CPU B finishes with the fragment queue, it calls inet_putpeer(),
which drops the refcount to 0 and frees the node via RCU.
The node is now freed but remains linked in the rbtree,
resulting in a Use-After-Free in the rbtree.
Fixes: b145425f26 ("inetpeer: remove AVL implementation in favor of RB tree")
Reported-by: Damiano Melotti <melotti@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260505133233.3039575-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----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=WKO1
-----END PGP SIGNATURE-----
Merge tag 'ipsec-2026-05-05' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec
Steffen Klassert says:
====================
pull request (net): ipsec 2026-05-05
1. Fix an IPv6 encapsulation error path that leaked route references
when UDPv6 ESP decapsulation resolved to an error route.
From Yilin Zhu.
2. Fix AH with ESN on async crypto paths by accounting for the extra
high-order sequence number when reconstructing the temporary
authentication layout in the completion callbacks.
From Michael Bomarito.
3. Fix XFRM output so it does not overwrite already-correct inner header
pointers when a tunnel layer such as VXLAN has already saved them.
The fix comes with new selftests. From Cosmin Ratiu.
4. Add the missing native payload size entry for XFRM_MSG_MAPPING in the
compat translation path. From Ruijie Li.
5. Harden __xfrm_state_delete() against repeated or inconsistent unhashing
of state list nodes by keying the removal on actual list membership and
using delete-and-init helpers. From Michal Kosiorek.
6. Prevent ESP from decrypting shared splice-backed skb fragments in place
by marking UDP splice frags as shared and forcing copy-on-write in ESP
input when needed. From Kuan-Ting Chen.
* tag 'ipsec-2026-05-05' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec:
xfrm: esp: avoid in-place decrypt on shared skb frags
xfrm: defensively unhash xfrm_state lists in __xfrm_state_delete
xfrm: provide message size for XFRM_MSG_MAPPING
xfrm: Don't clobber inner headers when already set
tools/selftests: Add a VXLAN+IPsec traffic test
tools/selftests: Use a sensible timeout value for iperf3 client
xfrm: ah: account for ESN high bits in async callbacks
ipv6: xfrm6: release dst on error in xfrm6_rcv_encap()
====================
Link: https://patch.msgid.link/20260505132326.1362733-1-steffen.klassert@secunet.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 3cf7203ca6 ("net/tunnel: wait until all sk_user_data
reader finish before releasing the sock") added synchronize_rcu()
in udp_tunnel_sock_release().
This was intended to protect the fast path of a dying vxlan device
from dereferencing vxlan_sock->sock->sk after sock_orphan() has set
sock->sk to NULL.
However, vxlan does not need to access struct socket itself
in the fast path; it only reads struct sock, and struct socket
is only used for tunnel setup and teardown.
This applies to all other UDP tunnel users, and they have been
converted to access struct sock directly.
In addition, each device-specific struct used in their fast paths
is freed after one RCU grace period. Since this occurs after
udp_tunnel_sock_release(), the struct is guaranteed to be freed
after struct udp_sock.
Therefore, synchronize_rcu() in udp_tunnel_sock_release() is
now redundant.
Let's remove it.
Tested:
A script creating/upping vxlan devices in 4000 netns runs 10x
faster with this change. We can see the same improvement with
other UDP tunnel devices as well.
$ cat vxlan.sh
for i in `seq 1 40`
do
(for j in `seq 1 100` ; do
unshare -n bash -c "ip link add vxlan0 type vxlan id 100 local 127.0.0.1 dstport 4789 && ip link set vxlan0 up";
done) &
done
wait
With bpftrace, we can see vxlan_stop() is significantly faster.
bpftrace -e '
kprobe:vxlan_stop {
@start[tid] = nsecs;
}
kretprobe:vxlan_stop /@start[tid]/ {
@duration_us = hist((nsecs - @start[tid]) / 1000);
delete(@start[tid]);
}
END {
printf("\nExecution time of vxlan_stop (us):\n");
}'
Before:
# time ./vxlan.sh // without bpftrace
real 0m50.615s
user 0m8.171s
sys 1m45.101s
@duration_us:
[4K, 8K) 1266 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ |
[8K, 16K) 1957 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[16K, 32K) 764 |@@@@@@@@@@@@@@@@@@@@ |
[32K, 64K) 6 | |
[64K, 128K) 4 | |
[128K, 256K) 3 | |
After:
# time ./vxlan.sh // without bpftrace
real 0m5.247s
user 0m7.956s
sys 1m47.404s
@duration_us:
[16, 32) 3411 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[32, 64) 383 |@@@@@ |
[64, 128) 107 |@ |
[128, 256) 79 |@ |
[256, 512) 16 | |
[512, 1K) 2 | |
[1K, 2K) 2 | |
Next step is to remove another synchronize_net() in vxlan_stop()
and variants in other devices.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260502031401.3557229-16-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
fou does not need to access struct socket itself in the fast
path; it only reads struct sock, and struct socket is only used
for tunnel setup and teardown.
Let's store struct sock directly in struct fou.
fou_release() frees struct fou with kfree_rcu(), so fou no
longer needs synchronize_rcu() in udp_tunnel_sock_release().
Note that the error path in fou_create() looks buggy; once the
tunnel is set up and fou_add_to_port_list() fails, struct fou
should be freed with kfree_rcu() _after_ udp_tunnel_sock_release().
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260502031401.3557229-12-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
None of the udp_tunnel users need struct socket in their
fast paths; it is only used for tunnel setup / teardown.
Even udp_tunnel_notify_{add,del}_rx_port() do not need
struct socket.
Let's change udp_tunnel_notify_{add,del}_rx_port() to take
struct sock instead of struct socket.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260502031401.3557229-6-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
None of the udp_tunnel users need struct socket in their
fast paths; it is only used for tunnel setup / teardown.
Even udp_tunnel_{push,drop}_rx_port() do not need struct socket.
Let's change udp_tunnel_{push,drop}_rx_port() to take struct
sock instead of struct socket.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260502031401.3557229-5-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>