When reviewing other changes Gemini points out that we currently
update module_fw_flash_in_progress without holding any locks.
Since module_fw_flash_in_progress is part of a bitfield this
is not great, updates to other fields may be lost.
We could use a bool and sprinkle some READ_ONCE/WRITE_ONCE here
but seems like the issue is rather than the work is an unusual
writer. The other writers already hold the right locks. So just
very briefly take these locks when the work completes.
Note that nothing ever cancels the FW update work, so there's
no concern with deadlocks vs cancel.
Fixes: 32b4c8b53e ("ethtool: Add ability to flash transceiver modules' firmware")
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Danielle Ratson <danieller@nvidia.com>
Link: https://patch.msgid.link/20260522231312.1710836-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We wait with filling the reply for new RSS context creation
until after the driver ->create_rxfh_context call. The driver
needs to fill some of the defaults in the context. The failure
of rss_fill_reply() is somewhat theoretical, but doesn't take
much effort to handle it properly. Call ->remove_rxfh_context().
If the driver's remove callback fails (some implementations like sfc
can return real command errors from firmware RPCs) - skip the xa_erase
and kfree, leaving the context in the xarray. This matches how
ethnl_rss_delete_doit() behaves.
Fixes: a166ab7816 ("ethtool: rss: support creating contexts via Netlink")
Link: https://patch.msgid.link/20260522230647.1705600-7-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
rss_get_data_alloc() allocates a single buffer that backs both the
indirection table and the hash key, but only assigned data->indir_table
when indir_size was nonzero. The expectation was that no driver
implements RSS without supporting indirection table but apparently
enic does just that (it's the only such in-tree driver).
enic has get_rxfh_key_size but no get_rxfh_indir_size.
data->indir_table stays as NULL, hkey gets set but rss_get_data_free()
kfree(data->indir_table) is a nop and the allocation leaks.
Always store the allocation base in data->indir_table so the free path
is unambiguous. No caller treats indir_table as a sentinel; everything
keys off indir_size.
Fixes: 7112a04664 ("ethtool: add netlink based get rss support")
Link: https://patch.msgid.link/20260522230647.1705600-6-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
rss_prepare_get() allocates the indirection table and hash key buffer
via rss_get_data_alloc(), then calls ops->get_rxfh() to populate them.
If get_rxfh() fails, the function returns an error without freeing
the allocation.
Fixes: 4f038a6a02 ("net: ethtool: Don't call .cleanup_data when prepare_data fails")
Link: https://patch.msgid.link/20260522230647.1705600-5-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
rss_set_prep_indir() compares the new indirection table against the
current one to determine whether any update is needed. The memcmp
call passes data->indir_size as the length argument, but indir_size
is the number of u32 entries, not the byte count.
Fixes: c0ae03588b ("ethtool: rss: initial RSS_SET (indirection table handling)")
Link: https://patch.msgid.link/20260522230647.1705600-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Gemini says that we're modifying the RSS_CREATE response skb.
I think it's right, the comment says that unicast() should
unshare the skb but I'm not entirely sure what I meant there.
netlink_trim() does a copy but only if skb is not well sized
(it's at least 2x larger than necessary for the payload).
Fixes: a166ab7816 ("ethtool: rss: support creating contexts via Netlink")
Link: https://patch.msgid.link/20260522230647.1705600-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
pskb_trim_rcsum_slow() keeps CHECKSUM_COMPLETE valid by subtracting
the checksum of the bytes removed from the skb tail. That assumes the
removed bytes can be read.
io_uring zcrx skbs may contain unreadable net_iov frags. With fbnic
header/data split, small TCP/IPv4 packets can carry Ethernet padding
in such a frag. ip_rcv_core() trims the skb to iph->tot_len before TCP
sees it, and the CHECKSUM_COMPLETE adjustment then calls
skb_checksum() on the padding.
This is exposed by IPv4 because small TCP/IPv4 frames can be shorter
than the Ethernet minimum payload. TCP/IPv6 frames are large enough in
the normal zcrx path, so they do not hit the same padding trim.
Keep the existing checksum adjustment for readable skbs. If the
remaining packet is fully linear, drop CHECKSUM_COMPLETE and let the
stack validate the packet after trimming. If unreadable payload would
remain, fail the trim; the checksum cannot be adjusted without reading
the trimmed tail.
Also clear skb->unreadable when trimming removes all frags.
Fixes: 65249feb6b ("net: add support for skbs with unreadable frags")
Signed-off-by: Björn Töpel <bjorn@kernel.org>
Reviewed-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260522120643.242974-1-bjorn@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
After patch 1/2 in this series, vti6_update() unlinks and relinks
the tunnel through t->net. vti6_siocdevprivate() still uses
dev_net(dev) for the collision lookup. For a tunnel moved through
IFLA_NET_NS_FD, dev_net(dev) is the new netns, not t->net.
SIOCCHGTUNNEL on a migrated tunnel then runs:
net = dev_net(dev) /* migrated netns */
t = vti6_locate(net, &p1, false) /* misses target in t->net */
...
t = netdev_priv(dev)
vti6_update(t, &p1, false) /* mutates t->net's hash */
A caller in the migrated netns picks params that match a tunnel
in the creation netns. The lookup in dev_net(dev) finds nothing.
vti6_update() prepends the migrated tunnel at the head of the
creation netns hash bucket for those params. Later lookups in
the creation netns resolve to the migrated device. xfrm receive
delivers the matched packets through a device the caller controls.
Reachable from an unprivileged user namespace (unshare --user
--map-root-user --net). Cross tenant scope on container hosts.
Switch the SIOCCHGTUNNEL path on a non fallback device to use
t->net for the lookup. The lookup now matches the netns
vti6_update() operates on.
Also add ns_capable(self->net->user_ns, CAP_NET_ADMIN) before
the lookup. The check at the top of the case is against
dev_net(dev)->user_ns, which after migration is the attacker's
netns. A caller there can pick params absent from self->net,
the lookup returns NULL, t becomes self, and vti6_update()
inserts the device into the creation netns hash. The new check
requires CAP_NET_ADMIN in the creation netns user_ns too.
SIOCADDTUNNEL and SIOCCHGTUNNEL on the fallback device keep
dev_net(dev), which equals init_net there.
Fixes: 61220ab349 ("vti6: Enable namespace changing")
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Suggested-by: Xiao Liang <shaw.leon@gmail.com>
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Maoyi Xie <maoyixie.tju@gmail.com>
Link: https://patch.msgid.link/20260521130555.3421684-3-maoyixie.tju@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
ip netns add ns1
ip netns add ns2
ip -n ns1 link add vti6_test type vti6 remote ::1 local ::2 key 7
ip -n ns1 link set vti6_test netns ns2
ip -n ns2 link set vti6_test type vti6 remote ::3 local ::4 key 9
ip netns del ns2
ip netns del ns1
[ 132.495484] ------------[ cut here ]------------
[ 132.497609] kernel BUG at net/core/dev.c:12376!
Commit 61220ab349 ("vti6: Enable namespace changing") dropped
NETIF_F_NETNS_LOCAL from vti6 devices. A vti6 tunnel can then
move through IFLA_NET_NS_FD. After the move dev_net(dev) points
at the new netns while t->net stays at the creation netns.
vti6_changelink() and vti6_update() still use dev_net(dev) and
dev_net(t->dev). They unlink from one per netns hash and relink
into another. The creation netns is left with a stale entry.
cleanup_net() of that netns later walks freed memory.
Reachable from an unprivileged user namespace (unshare --user
--map-root-user --net). Cross tenant scope on container hosts.
Fixes: 61220ab349 ("vti6: Enable namespace changing")
Reported-by: Maoyi Xie <maoyi.xie@ntu.edu.sg>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260521130555.3421684-2-maoyixie.tju@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Florian Westphal says:
====================
netfilter: updates for net
Patches 7+8 fix a regression from 7.1-rc1. Everything else
is from 2.6.x to 5.3 releases. There are additional known
issues with these patches (drive-by-findings in related code).
There are many old bugs all over netfilter and our ability to review
feature patches has come to a complete halt due to lack of time.
There are further security bugs that we cannot address
due to lack of time, maintainers and reviewers.
Other remarks: The xtables 32bit compat interface is already
off in many vendor kernels, the plan is to remove it soon.
1) Prevent RST packets with invalid sequence numbers from forcing TCP
connections into the CLOSE state without a direction check.
From Hamza Mahfooz.
2) Re-derive the TCP header pointer after skb_ensure_writable in
synproxy_tstamp_adjust. Prevent use-after-free and invalid checksum
updates caused by stale pointers during buffer expansion.
From Chris Mason.
3) Fix a race condition causing keymap list corruption in conntracks gre/pptp
helper.
4) Use raw_smp_processor_id() in xt_cpu to prevent splats under
PREEMPT_RCU.
5) Disable netfilter payload mangling in user namespaces (nft_payload.c
and nf_queue).
TCP option mangling via nft_exthdr.c remains enabled.
There will be followups here to restrict resp. revalidate
headers.
6) Fix an out-of-bounds read in ebtables's compat_mtw_from_user function.
7) Use list_for_each_entry_rcu() to traverse fib6_siblings in
nft_fib6_info_nh_uses_dev(). Ensure safe list walking under RCU.
8) Fix an out-of-bounds read in nft_fib_ipv6 caused by incorrect list
traversal.
9) Add nft_fib_nexthop selftest to netfilter. Cover nexthop enumeration for
single, group, and multipath route shapes.
All three nft_fib6 fixes from Jiayuan Chen.
10) Fix destination corruption in shift operations when source and destination
registers overlap. Reject partial register overlap for all operations
from control plane. From Fernando Fernandez Mancera.
* tag 'nf-26-05-22' of https://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: nf_tables: fix dst corruption in same register operation
selftests: netfilter: add nft_fib_nexthop test
netfilter: nft_fib_ipv6: handle routes via external nexthop
netfilter: nft_fib_ipv6: walk fib6_siblings under RCU
netfilter: ebtables: fix OOB read in compat_mtw_from_user
netfilter: disable payload mangling in userns
netfilter: xt_cpu: prefer raw_smp_processor_id
netfilter: nf_conntrack_gre: fix gre keymap list corruption
netfilter: synproxy: refresh tcphdr after skb_ensure_writable
netfilter: conntrack: tcp: do not force CLOSE on invalid-seq RST without direction check
====================
Link: https://patch.msgid.link/20260522104257.2008-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Mirror iucv_sock_setsockopt() and wrap the whole switch in
lock_sock()/release_sock(). The pre-existing SO_MSGLIMIT-only lock
becomes redundant and is removed.
Any AF_IUCV HIPER user can potentially crash the kernel by racing
recvmsg() with getsockopt(SO_MSGSIZE): the SO_MSGSIZE arm dereferences
iucv->hs_dev->mtu after iucv_sock_close() (called from the racing
recvmsg()) has set hs_dev to NULL, producing a NULL pointer dereference
oops.
Suggested-by: Stanislav Fomichev <sdf.kernel@gmail.com>
Fixes: 51363b8751 ("af_iucv: allow retrieval of maximum message size")
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Tested-by: Alexandra Winter <wintera@linux.ibm.com>
Link: https://patch.msgid.link/20260521-af_iucv_fix2-v1-1-f16b1c510aa9@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
INIT_HLIST_HEAD(&smc_v*_hashinfo.ht) are called after smc_nl_init(),
proto_register() and sock_register(). This can lead to smc_v*_hashinfo.ht
being reset even though hash entries already exist and are being used,
possibly resulting in a corrupted list.
Remove unnecessary and dangerous re-initialisation of smc_v*_hashinfo.ht in
smc_init(); it is implicitly initialised to zero anyhow. Add
HLIST_HEAD_INIT to the definitions for clarity.
Fixes: f16a7dd5cf ("smc: netlink interface for SMC sockets")
Suggested-by: Halil Pasic <pasic@linux.ibm.com>
Signed-off-by: Alexandra Winter <wintera@linux.ibm.com>
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Reviewed-by: Mahanta Jambigi <mjambigi@linux.ibm.com>
Link: https://patch.msgid.link/20260521145639.10317-1-wintera@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In most cases, notifications on sockets with NETLINK_LISTEN_ALL_NSID
do not contain NSID in their ancillary data in case the event is local
to the listener.
However, when a self-referential NSID is allocated for a namespace,
every local notification starts sending this ID to the user space.
This is problematic, because the listener cannot tell if those
notifications are local or not anymore without making extra requests
to figure out if the provided NSID is local or not. The listener
can also not figure out the local NSID beforehand as it can be
allocated at any point in time by other processes, changing the
structure of the future notifications for everyone.
The value is practically not useful, since it's the namespace's own
ID that the application has to obtain from other sources in order to
figure out if it's the same or not. So, for the application it's
just an extra busy work with no benefits. Moreover, applications
that do not know about this quirk may be mishandling notifications
with NSID set as notifications from remote namespaces. This is the
case for ovs-vswitchd and the iproute2's 'ip monitor' that stops
printing 'current' and starts printing the nsid number mid-session.
Lack of clear documentation for this behavior is also not helping.
A search though open-source projects doesn't reveal any projects
that use NETNSA_NSID_NOT_ASSIGNED and rely on metadata to contain
self-referential NSIDs (expected, since the value is not useful).
Quite the opposite, as already mentioned, there are few applications
that rely on NSID to not be present in local events.
Since the value is not useful and actively harmful in some cases,
let's not report it for local events, making the notifications more
consistent.
Also adding some blank lines for readability.
Fixes: 59324cf35a ("netlink: allow to listen "all" netns")
Reported-by: Matteo Perin <matteo.perin@canonical.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Link: https://patch.msgid.link/20260520172317.175168-3-i.maximets@ovn.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If the current skb is not shared, it is re-used directly for all the
sockets subscribed to the notification. If we have remote all-nsid
socket receiving a message first, then the 'nsid_is_set' will be
set to 'true'. If the nsid is NOT_ASSIGNED for the next socket in
the list, the 'nsid_is_set' will remain 'true' and the negative value
is be delivered to the user space. All subsequent nsid values will be
delivered as well, since there is no code path that sets the flag
back to 'false'.
Fix that by always dropping the flag to 'false' first.
Fixes: 7212462fa6 ("netlink: don't send unknown nsid")
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Link: https://patch.msgid.link/20260520172317.175168-2-i.maximets@ovn.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
vsock_poll() reads vsk->peer_shutdown before taking the socket lock
to set EPOLLHUP and EPOLLRDHUP, then reads it again after taking
the lock to report EOF readability. A shutdown packet can update
peer_shutdown while poll is waiting for the lock, so one poll invocation
can report EOF readability without the corresponding HUP/RDHUP bits.
For connectible sockets, take one peer_shutdown snapshot after
lock_sock() and use it for all peer-shutdown-derived poll bits. For
datagram sockets, which do not take lock_sock() in poll(), take one
lockless READ_ONCE() snapshot and pair it with WRITE_ONCE() on the
writer side.
This keeps the peer-shutdown-derived bits internally consistent for each
poll pass.
Fixes: d021c34405 ("VSOCK: Introduce VM Sockets")
Signed-off-by: Ziyu Zhang <ziyuzhang201@gmail.com>
Link: https://patch.msgid.link/20260519165636.62542-1-ziyuzhang201@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For lshift and rshift, the shift operations are performed in a loop over
32-bit words. The loop calculates the shifted value and write it to dst,
and then immediately reads from src to calculate the carry for the next
iteration. Because src and dst could point to the same memory location,
the carry is incorrectly calculated using the newly modified dst value
instead of the original src value.
Adding a temporary local variable to cache the original value before
writing to dst and using it for the carry calculation solves the
problem. In addition, partial overlap is rejected from control plane for
all kind of operations including byteorder. This was tested with the
following bytecode:
table test_table ip flags 0 use 1 handle 1
ip test_table test_chain use 3 type filter hook input prio 0 policy accept packets 0 bytes 0 flags 1
ip test_table test_chain 2
[ immediate reg 1 0x44332211 0x88776655 ]
[ bitwise reg 1 = ( reg 1 << 0x08000000 ) ]
[ cmp eq reg 1 0x66443322 0x00887766 ]
[ counter pkts 0 bytes 0 ]
ip test_table test_chain 4 3
[ immediate reg 1 0x44332211 0x88776655 ]
[ bitwise reg 1 = ( reg 1 << 0x08000000 ) ]
[ cmp eq reg 1 0x55443322 0x00887766 ]
[ counter pkts 21794 bytes 1917798 ]
Fixes: 567d746b55 ("netfilter: bitwise: add support for shifts.")
Acked-by: Jeremy Sowden <jeremy@azazel.net>
Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
nft_fib6_info_nh_uses_dev() runs from nft_fib6_eval() in softirq under
rcu_read_lock(). fib6_siblings is modified by writers that hold
tb6_lock but do not wait for RCU readers, so the sibling walk should
use list_for_each_entry_rcu(): it adds READ_ONCE() on the ->next
pointer and lets CONFIG_PROVE_RCU_LIST validate the locking.
No functional change for non-debug builds.
Fixes: 1c32b24c23 ("netfilter: nft_fib_ipv6: switch to fib6_lookup")
Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Signed-off-by: Florian Westphal <fw@strlen.de>
Luxiao Xu says:
The function compat_mtw_from_user() converts ebtables extensions from
32-bit user structures to kernel native structures. However, it lacks
proper validation of the user-supplied match_size/target_size.
When certain extensions are processed, the kernel-side translation
logic may perform memory accesses based on the extension's expected
size. If the user provides a size smaller than what the extension
requires, it results in an out-of-bounds read as reported by KASAN.
This fix introduces a check to ensure match_size is at least as large
as the extension's required compatsize. This covers matches, watchers,
and targets, while maintaining compatibility with standard targets.
AFAIU this is relevant for matches that need to go though
match->compat_from_user() call. Those that use plain memcpy with the
user-provided size are ok because the caller checks that size vs the
start of the next rule entry offset (which itself is checked vs. total
size copied from userspace).
The ->compat_from_user() callbacks assume they can read compatsize bytes,
so they need this extra check.
Based on an earlier patch from Luxiao Xu.
Fixes: 81e675c227 ("netfilter: ebtables: add CONFIG_COMPAT support")
Reported-by: Yuan Tan <yuantan098@gmail.com>
Reported-by: Yifan Wu <yifanwucs@gmail.com>
Reported-by: Juefei Pu <tomapufckgml@gmail.com>
Reported-by: Xin Liu <bird@lzu.edu.cn>
Signed-off-by: Luxiao Xu <rakukuip@gmail.com>
Signed-off-by: Ren Wei <n05ec@lzu.edu.cn>
Reviewed-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
Several parts of network stack rely on iph->ihl validation
done by network stack before PRE_ROUTING.
Disable this feature for user namespaces for now.
tcp option handling is likely safe even for LOCAL_IN, so this
this leaves tcp option mangling via nft_exthdr.c as-is.
I don't think these are the only means to alter packets, but these
appear to be relatively prominent.
This could be relaxed later. Example:
- allow userns for ingress hook.
- allow userns if base is transport header.
Also, we should revalidate or restrict generally:
- Don't allow linklayer writes to spill into network header
- restrict ipv4 and ipv6 to 'known safe' writes, e.g.
saddr/daddr/check/tos
Reported-by: Qi Tang <tpluszz77@gmail.com>
Reported-by: Tong Liu <lyutoon@gmail.com>
Tested-by: Qi Tang <tpluszz77@gmail.com>
Link: https://lore.kernel.org/netfilter-devel/20260515100411.3141-1-fw@strlen.de/
Signed-off-by: Florian Westphal <fw@strlen.de>
With PREEMPT_RCU we get splat:
BUG: using smp_processor_id() in preemptible [..]
caller is cpu_mt+0x53/0xd0 net/netfilter/xt_cpu.c:37
CPU: 1 .. Comm: syz.3.1377 #0 PREEMPT(full)
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
check_preemption_disabled+0xd3/0xe0 lib/smp_processor_id.c:47
cpu_mt+0x53/0xd0 net/netfilter/xt_cpu.c:37
[..]
Just use raw version instead.
This is similar to 14d14a5d29 ("netfilter: nft_meta: use raw_smp_processor_id()").
Fixes: 0ca743a559 ("netfilter: nf_tables: add compatibility layer for x_tables")
Reported-by: syzbot+690d3e3ffa7335ac10eb@syzkaller.appspotmail.com
Signed-off-by: Florian Westphal <fw@strlen.de>
Quoting reporter:
A race between GRE keymap insertion and destruction can corrupt the
kernel list or use a freed object. `nf_ct_gre_keymap_add()` publishes a
new keymap pointer before the embedded `list_head` is linked, while
`nf_ct_gre_keymap_destroy()` can concurrently delete and free that
same object. An unprivileged user can reach this through the PPTP
conntrack helper by racing PPTP control messages or helper teardown,
leading to KASAN-detectable list corruption/UAF in kernel context.
## Root Cause Analysis
`exp_gre()` installs GRE expectations for a PPTP control flow and then
adds two GRE keymap entries [..]
The add path publishes `ct_pptp_info->keymap[dir]` before linking the
embedded list node [..]
Concurrent teardown deletes that partially initialized object.
Make add/destroy symmetric: install both, destroy both while under lock.
Furthermore, we should refuse to publish a new mapping in case ct is going
away, else we may leak the allocation.
The "retrans" detection is strange: existing mapping is checked for key
equality with the new mapping, then for "is on the list" via list walk.
But I can't see how an existing keymap entry can be NOT on list.
Change this to only check if we're asked to map same tuple again -- if so,
skip re-install, else signal failure.
Last, add a bug trap for the keymap list; it has to be empty when namespace
is going away.
Reported-by: Leo Lin <leo@depthfirst.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
synproxy_tstamp_adjust() rewrites the TCP timestamp option in place
and then patches the TCP checksum via inet_proto_csum_replace4() on
the caller-supplied tcphdr pointer. Both ipv4_synproxy_hook() and
ipv6_synproxy_hook() obtain that pointer with skb_header_pointer()
before calling in, so it may either alias skb->head directly or
point at the caller's on-stack _tcph buffer.
Between obtaining the pointer and using it, the function calls
skb_ensure_writable(skb, optend), which on a cloned or non-linear
skb invokes pskb_expand_head() and frees the old skb->head. After
that point the cached th is stale:
caller (ipv[46]_synproxy_hook)
th = skb_header_pointer(skb, ..., &_tcph)
synproxy_tstamp_adjust(skb, protoff, th, ...)
skb_ensure_writable(skb, optend)
pskb_expand_head() /* kfree(old skb->head) */
...
inet_proto_csum_replace4(&th->check, ...)
/* writes into freed head, or
into the caller's stack copy
leaving the on-wire checksum
stale */
The option bytes are written through skb->data and are fine; only
the checksum update goes through th and so lands in the wrong
place. The result is either a write into freed slab memory or a
packet leaving with a checksum that does not match its payload.
Fix by re-deriving th from skb->data + protoff immediately after
skb_ensure_writable() succeeds, so the subsequent checksum update
targets the linear, writable header.
Fixes: 48b1de4c11 ("netfilter: add SYNPROXY core/target")
Assisted-by: kres (claude-opus-4-7)
Signed-off-by: Chris Mason <clm@meta.com>
Reviewed-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
An unintended behavior in the TCP conntrack state machine allows a
connection to be forced into the CLOSE state using an RST packet with an
invalid sequence number.
Specifically, after a SYN packet is observed, an RST with an invalid SEQ
can transition the conntrack entry to TCP_CONNTRACK_CLOSE, regardless of
whether the RST corresponds to the expected reply direction. The relevant
code path assumes the RST is a response to an outgoing SYN, but does not
validate packet direction or ensure that a matching SYN was actually sent
in the opposite direction.
As a result, a crafted packet sequence consisting of a SYN followed by an
invalid-sequence RST can prematurely terminate an active NAT entry. This
makes connection teardown easier than intended.
So, tighten the state transition logic to ensure that RST-triggered
CLOSE transitions only occur when the RST is a valid response to a
previously observed SYN in the correct direction.
Cc: stable@vger.kernel.org
Fixes: 9fb9cbb108 ("[NETFILTER]: Add nf_conntrack subsystem.")
Signed-off-by: Hamza Mahfooz <hamzamahfooz@linux.microsoft.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Pull networking fixes from Jakub Kicinski:
"Including fixes from Bluetooth, wireless and netfilter.
Craziness continues with no end in sight. Even discounting the driver
revert this is a pretty huge PR for standards of the previous era. I'd
speculate - we haven't seen the worst of it, yet. Good news, I guess,
is that so far we haven't seen many (any?) cases of "AI reported a
bug, we fixed it and a real user regressed".
Current release - fix to a fix:
- Bluetooth: btmtk: accept too short WMT FUNC_CTRL events
- vsock/virtio: relax the recently added memory limit a little
Current release - regressions:
- IB/IPoIB: make sure IB drivers always use async set_rx_mode since
some (mlx5) are now required to use it due to locking changes
Previous releases - regressions:
- udp: fix UDP length on last GSO_PARTIAL segment
- af_unix: fix UAF read of tail->len in unix_stream_data_wait()
- tcp: fix stale per-CPU tcp_tw_isn leak enabling ISN prediction
- mlx5e: fix unlocked writing to ICOSQ, breaking AF_XDP
Previous releases - always broken:
- tap: fix stack info leak in tap_ioctl() SIOCGIFHWADDR
- ipv4: raw: reject IP_HDRINCL packets with ihl < 5
- Bluetooth: a lot of locking and concurrency fixes (as always)
- batman-adv (mesh wireless networking): a lot of random fixes for
issues reported by security researchers and Sashiko
- netfilter: same thing, a lot of small security-ish fixes all over
the place, nothing really stands out
Misc:
- bring back the old 3c509 driver, Maciej wants to maintain it"
* tag 'net-7.1-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (187 commits)
net: enetc: avoid VF->PF mailbox timeout during SR-IOV teardown
net: enetc: fix init and teardown order to prevent use of unsafe resources
net: enetc: fix unbounded loop and interrupt handling in VF-to-PF messaging
net: enetc: fix DMA write to freed memory in enetc_msg_free_mbx()
net: enetc: fix race condition in VF MAC address configuration
net: enetc: fix TOCTOU race and validate VF MAC address
net: enetc: add ratelimiting to VF mailbox error messages
net: enetc: fix missing error code when pf->vf_state allocation fails
net: enetc: fix incorrect mailbox message status returned to VFs
net: bridge: prevent too big nested attributes in br_fill_linkxstats()
l2tp: use list_del_rcu in l2tp_session_unhash
net: bcmgenet: keep RBUF EEE/PM disabled
ethernet: 3c509: Fix most coding style issues
ethernet: 3c509: Update documentation to match MAINTAINERS
ethernet: 3c509: Add GPL 2.0 SPDX license identifier
ethernet: 3c509: Fix AUI transceiver type selection
Revert "drivers: net: 3com: 3c509: Remove this driver"
tools: ynl: support listening on all nsids
net: gro: don't merge zcopy skbs
pds_core: ensure null-termination for firmware version strings
...
Johannes Berg says:
====================
Quite a few more updates:
- cfg80211/mac80211:
- various security(-ish) fixes
- fix A-MSDU subframe handling
- fix multi-link element parsing
- ath10: avoid sending commands to dead device
- ath11k:
- fix WMI buffer leaks on error conditions
- fix UAF in RX MSDU coalesce path
- allow peer ID 0 on RX path (legal for mobile devices)
- reinitialize shared SRNG pointers on restart
- ath12k:
- fix 20 MHz-only parsing of EHT-MCS map
- iwlwifi:
- fix TSO segmentation explosion
- don't TX to dead device
- fix warning in WoWLAN
- fix TX rates on old devices
- disconnect on beacon loss only if also no other traffic
- fill NULL-ptr deref
- fix STEP_URM hardware access
* tag 'wireless-2026-05-21' of https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless: (24 commits)
wifi: cfg80211: wext: validate chandef in monitor mode
wifi: mac80211: consume only present negotiated TTLM maps
wifi: wilc1000: fix dma_buffer leak on bus acquire failure
wifi: mac80211: capture fast-RX rate before mesh reuses skb->cb
wifi: mac80211: fix multi-link element inheritance
wifi: mac80211: fix MLE defragmentation
wifi: mac80211: don't override max_amsdu_subframes
wifi: mac80211: bounds-check link_id in ieee80211_ml_epcs
wifi: ath12k: fix EHT TX MCS limitation due to wrong 20 MHz-only parsing
wifi: ath11k: clear shared SRNG pointer state on restart
wifi: ath11k: fix use after free in ath11k_dp_rx_msdu_coalesce()
wifi: ath11k: fix peer resolution on rx path when peer_id=0
wifi: iwlwifi: mld: disconnect only after 6 beacons without Rx
wifi: iwlwifi: mld: don't WARN on WoWLAN suspend w/o BSS vif
wifi: iwlwifi: use correct function to read STEP_URM register
wifi: iwlwifi: mvm: fix driver-set TX rates on old devices
wifi: iwlwifi: mld: don't dereference a pointer before NULL checking it
wifi: iwlwifi: mld: stop TX during firmware restart
wifi: iwlwifi: mld: fix TSO segmentation explosion when AMSDU is disabled
wifi: ath10k: skip WMI and beacon transmission when device is wedged
...
====================
Link: https://patch.msgid.link/20260521152903.374070-3-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
An unprivileged local user can pin a host CPU indefinitely in
l2tp_session_get_by_ifname() by issuing L2TP_CMD_SESSION_GET on
L2TP_ATTR_IFNAME concurrently with L2TP_CMD_SESSION_CREATE and
L2TP_CMD_SESSION_DELETE on the same tunnel. All three commands take
GENL_UNS_ADMIN_PERM, so CAP_NET_ADMIN in the netns user namespace
suffices; on any host that has l2tp_core loaded the trigger is
reachable from a standard `unshare -Urn` sandbox.
l2tp_session_unhash() removes a session from tunnel->session_list
with list_del_init(), but that list is walked by
l2tp_session_get_by_ifname() with list_for_each_entry_rcu() under
rcu_read_lock_bh(). list_del_init() leaves the deleted entry's
next/prev self-pointing; a reader that has loaded the entry and
then advances pos->list.next reads &session->list, container_of()s
back to the same session, and list_for_each_entry_rcu() never
reaches the list head. The CPU stays in strcmp() inside the
walker, with BH and preemption disabled, so RCU grace periods on
the host stall behind it and the wedged thread cannot be killed
(SIGKILL is delivered on syscall return).
Use list_del_rcu() to match the existing list_add_rcu() in
l2tp_session_register(); the deleted session remains visible to
in-flight walkers with consistent next/prev pointers until
kfree_rcu() in l2tp_session_free() releases it. tunnel->session_list
has exactly one list_del_init() call site; the list_del_init
(&session->clist) at l2tp_core.c:533 operates on the per-collision
list, which is not walked under RCU. list_empty(&session->list) is
not used anywhere in net/l2tp/ after the unhash point, so dropping
the post-delete self-init is safe; the fix has no userspace-visible
behavior change.
Fixes: 89b768ec2d ("l2tp: use rcu list add/del when updating lists")
Cc: stable@vger.kernel.org # 6.11+
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Link: https://patch.msgid.link/20260518183447.64078-1-michael.bommarito@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
skb_gro_receive() can currently copy frags between the source and GRO
skb, without checking the zerocopy status, and in particular the
SKBFL_MANAGED_FRAG_REFS flag.
When SKBFL_MANAGED_FRAG_REFS is set, the skb doesn't hold a reference
on the pages in shinfo->frags. Appending those frags to another skb's
frags without fixing up the page refcount can lead to UAF.
When either the last skb in the GRO chain (the one we would append
frags to) or the source skb is zerocopy, don't merge the skbs.
Fixes: 753f1ca4e1 ("net: introduce managed frags infrastructure")
Reported-by: Huzaifa Sidhpurwala <huzaifas@redhat.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/c3b7f906bbfcbdfd7b4fa9d6c18a438870df85be.1779307748.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reported by Sashiko:
In ipv6_hop_ioam(), the hdr pointer is initialized to point into the
skb's linear data buffer. Later, the code calls skb_ensure_writable(),
which might reallocate the buffer:
if (skb_ensure_writable(skb, optoff + 2 + hdr->opt_len))
goto drop;
/* Trace pointer may have changed */
trace = (struct ioam6_trace_hdr *)(skb_network_header(skb)
+ optoff + sizeof(*hdr));
ioam6_fill_trace_data(skb, ns, trace, true);
ioam6_event(IOAM6_EVENT_TRACE, dev_net(skb->dev),
GFP_ATOMIC, (void *)trace, hdr->opt_len - 2);
If the skb is cloned or lacks sufficient linear headroom,
skb_ensure_writable() will invoke pskb_expand_head(), which reallocates
the skb's data buffer and frees the old one, invalidating pointers to
it. While the code recalculates the trace pointer immediately after the
call to skb_ensure_writable(), it fails to recalculate the hdr pointer.
This patch fixes the above by recalculating the hdr pointer before
passing hdr->opt_len to ioam6_event(), so that we avoid any UaF.
Fixes: f655c78d62 ("net: exthdrs: ioam6: send trace event")
Cc: stable@vger.kernel.org
Signed-off-by: Justin Iurman <justin.iurman@gmail.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260520124242.32320-1-justin.iurman@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
__netpoll_send_skb() always transmits through np->dev and queues busy
packets on np->dev->npinfo->txq, but it leaves skb->dev unchanged.
Stacked callers such as DSA and macvlan can reach netpoll with skb->dev
still naming the upper device while np->dev is the lower device that
owns the netpoll state.
If the skb has to be deferred, queue_process() later dequeues it from
the lower device's txq but retries it through skb->dev. That can
re-enter the upper ndo_start_xmit path on an already transformed skb,
and if the upper device disappears before the lower txq drains the
workqueue can dereference a stale skb->dev pointer.
The buggy scenario involves two paths, with each column showing the
order within that path:
path A label: netpoll enqueue path path B label: upper-device teardown
1. Stacked xmit calls netpoll 1. Teardown unregisters the upper
with lower np->dev and upper net_device while lower npinfo
skb->dev. stays alive.
2. __netpoll_send_skb() uses 2. netdev_release() runs for the
np->dev->npinfo as the txq upper net_device.
owner.
3. Busy transmit queues the skb 3. The lower txq still owns the
on that lower txq with upper deferred skb.
skb->dev.
4. queue_process() drains the 4. queue_process() dereferences
lower txq and reads skb->dev. that stale upper skb->dev.
Normalize skb->dev to np->dev after loading np->dev from the netpoll
instance, before either the direct transmit path or the fallback enqueue.
This keeps the queued skb in the same device and txq domain as the
netpoll state that owns it.
KASAN report as below:
KASAN slab-use-after-free in queue_process+0x7c/0x480
Workqueue: events queue_process
The buggy address belongs to the object at ffff88810906c000 which belongs
to the cache kmalloc-4k of size 4096
The buggy address is located 168 bytes inside of freed 4096-byte region
[ffff88810906c000, ffff88810906d000)
Read of size 8
Call trace:
dump_stack_lvl+0x73/0xb0 (?:?)
print_report+0xd1/0x620 (?:?)
srso_alias_return_thunk+0x5/0xfbef5 (?:?)
__virt_addr_valid+0x215/0x420 (?:?)
kasan_complete_mode_report_info+0x64/0x200 (?:?)
kasan_report+0xf7/0x130 (?:?)
queue_process+0x7c/0x480 (net/core/netpoll.c:88)
kasan_check_range+0x10c/0x1c0 (?:?)
__kasan_check_read+0x15/0x20 (?:?)
process_one_work+0x8b7/0x1af0 (kernel/workqueue.c:3200)
assign_work+0x170/0x3f0 (?:?)
worker_thread+0x574/0xf10 (?:?)
_raw_spin_unlock_irqrestore+0x4b/0x60 (?:?)
trace_hardirqs_on+0x2a/0x180 (?:?)
kthread+0x2fc/0x3f0 (?:?)
ret_from_fork+0x58b/0x830 (?:?)
__switch_to+0x58e/0xe90 (?:?)
__switch_to_asm+0x39/0x70 (?:?)
ret_from_fork_asm+0x1a/0x30 (?:?)
Freed by task stack:
kasan_save_stack+0x3d/0x60 (?:?)
kasan_save_track+0x18/0x40 (?:?)
kasan_save_free_info+0x3f/0x60 (?:?)
__kasan_slab_free+0x48/0x70 (?:?)
kfree+0x20e/0x4e0 (?:?)
kvfree+0x31/0x40 (?:?)
netdev_release+0x71/0x90 (net/core/net-sysfs.c:2227)
device_release+0xd2/0x250 (?:?)
kobject_put+0x181/0x4c0 (lib/kobject.c:730)
netdev_run_todo+0x700/0x1000 (net/core/dev.c:11666)
rtnl_dellink+0x396/0xc00 (net/core/rtnetlink.c:3558)
rtnetlink_rcv_msg+0x740/0xc20 (net/core/rtnetlink.c:6897)
netlink_rcv_skb+0x147/0x3a0 (?:?)
rtnetlink_rcv+0x19/0x20 (net/core/rtnetlink.c:7021)
netlink_unicast+0x4d1/0x830 (net/netlink/af_netlink.c:1327)
netlink_sendmsg+0x840/0xe10 (net/netlink/af_netlink.c:1812)
____sys_sendmsg+0x8a7/0xb50 (?:?)
___sys_sendmsg+0x104/0x190 (?:?)
__sys_sendmsg+0x135/0x1d0 (?:?)
__x64_sys_sendmsg+0x7b/0xc0 (?:?)
x64_sys_call+0x205c/0x2130 (?:?)
do_syscall_64+0x115/0x6a0 (arch/x86/entry/syscall_64.c:87)
entry_SYSCALL_64_after_hwframe+0x77/0x7f (?:?)
Fixes: 5de4a473bd ("netpoll queue cleanup")
Signed-off-by: Zhang Cen <rollkingzzc@gmail.com>
Link: https://patch.msgid.link/20260519104647.3517990-1-rollkingzzc@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ip6_route_init() registers ip6_route_dev_notifier before registering the
IPv6 route BPF iterator target. If bpf_iter_register() fails after the
notifier has been registered, the error path currently jumps to
out_register_late_subsys and unwinds the RTNL handlers and pernet route
state without removing the notifier from the netdevice notifier chain.
This leaves ip6_route_dev_notify() callable after the IPv6 route state it
uses has been torn down. Add a separate unwind label for the BPF iterator
failure path and unregister the netdevice notifier before continuing with
the existing cleanup.
Fixes: 138d0be35b ("net: bpf: Add netlink and ipv6_route bpf_iter targets")
Signed-off-by: Yuho Choi <dbgh9129@gmail.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260520030329.1061183-1-dbgh9129@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The networking receive path is usually run from softirq context, but
protocols that take the socket lock may have packets stored in the
backlog and processed later from process context. In that case
release_sock() -> __release_sock() drops the slock with spin_unlock_bh()
and then calls sk->sk_backlog_rcv() with bottom halves enabled.
Typical sk_backlog_rcv handlers process the socket whose backlog is
being drained, so the BH state at entry is irrelevant for the slocks
they touch. pep_do_rcv() is different: when the inbound skb targets an
existing PEP pipe, it forwards the skb to a different *child* socket
via sk_receive_skb(). That helper takes the child slock with
bh_lock_sock_nested(), which is just spin_lock_nested() and assumes BH
is already off. The same child slock therefore ends up acquired with
BH on (process path) and with BH off (softirq path):
process context softirq context
--------------- ---------------
release_sock(listener) __netif_receive_skb()
__release_sock() phonet_rcv()
spin_unlock_bh() __sk_receive_skb(listener)
[BH now ENABLED] [BH already disabled]
sk_backlog_rcv: sk_backlog_rcv:
pep_do_rcv() pep_do_rcv()
sk_receive_skb(child) sk_receive_skb(child)
bh_lock_sock_nested(child) bh_lock_sock_nested(child)
=> SOFTIRQ-ON-W => IN-SOFTIRQ-W
Lockdep flags this as inconsistent lock state, and it can become a real
self-deadlock if a softirq on the same CPU tries to receive to the same
child socket while its slock is held in the BH-enabled path:
WARNING: inconsistent lock state
inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
(slock-AF_PHONET/1){+.?.}-{3:3}, at: __sk_receive_skb+0x1cf/0x900
__sk_receive_skb net/core/sock.c:563
sk_receive_skb include/net/sock.h:2022 [inline]
pep_do_rcv net/phonet/pep.c:675
sk_backlog_rcv include/net/sock.h:1190
__release_sock net/core/sock.c:3216
release_sock net/core/sock.c:3815
pep_sock_accept net/phonet/pep.c:879
Wrap the forwarded sk_receive_skb() in local_bh_disable() /
local_bh_enable() so the child slock is always acquired with BH off.
local_bh_disable() nests safely on the softirq path.
Discovered via in-house syzkaller fuzzing; the same root cause also
on the linux-6.1.y syzbot dashboard as extid 44f0626dd6284f02663c.
Reproduced under KASAN + LOCKDEP + PROVE_LOCKING, reproducer:
https://pastebin.com/A3t8xzCR
Fixes: 9641458d3e ("Phonet: Pipe End Point for Phonet Pipes protocol")
Link: https://syzkaller.appspot.com/bug?extid=44f0626dd6284f02663c
Cc: stable@vger.kernel.org
Signed-off-by: Zijing Yin <yzjaurora@gmail.com>
Acked-by: Rémi Denis-Courmont <remi@remlab.net>
Reported-by: syzbot+9f4a135646b66c509935@syzkaller.appspotmail.com
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260519172635.86304-1-yzjaurora@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After commit 059b7dbd20 ("vsock/virtio: fix potential unbounded skb
queue"), virtio_transport_inc_rx_pkt() subtracts per-skb overhead from
buf_alloc when checking whether a new packet fits. This reduces the
effective receive buffer below what the user configured via
SO_VM_SOCKETS_BUFFER_SIZE, causing legitimate data packets to be
silently dropped and applications that rely on the full buffer size
to deadlock.
Also, the reduced space is not communicated to the remote peer, so
its credit calculation accounts more credit than the receiver will
actually accept, causing data loss (there is no retransmission).
With this approach we currently have failures in
tools/testing/vsock/vsock_test.c. Test 18 sometimes fails, while
test 22 always fails in this way:
18 - SOCK_STREAM MSG_ZEROCOPY...hash mismatch
22 - SOCK_STREAM virtio credit update + SO_RCVLOWAT...send failed:
Resource temporarily unavailable
Fix by allowing at most `buf_alloc * 2` as the total budget for payload
plus skb overhead in virtio_transport_inc_rx_pkt(), similar to how
SO_RCVBUF is doubled to reserve space for sk_buff metadata.
This preserves the full buf_alloc for payload under normal operation,
while still bounding the skb queue growth.
With this patch, all tests in tools/testing/vsock/vsock_test.c are
now passing again.
Fixes: 059b7dbd20 ("vsock/virtio: fix potential unbounded skb queue")
Cc: stable@vger.kernel.org
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://patch.msgid.link/20260518090656.134588-3-sgarzare@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When there is no more space to queue an incoming packet, the packet is
silently dropped. This causes data loss without any notification to
either peer, since there is no retransmission.
Under normal circumstances, this should never happen. However, it could
happen if the other peer doesn't respect the credit, or if the skb
overhead, which we recently began to take into account with commit
059b7dbd20 ("vsock/virtio: fix potential unbounded skb queue"),
is too high.
Fix this by resetting the connection and setting the local socket error
to ENOBUFS when virtio_transport_recv_enqueue() can no longer queue a
packet, so both peers are explicitly notified of the failure rather than
silently losing data.
Fixes: ae6fcfbf5f ("vsock/virtio: discard packets if credit is not respected")
Cc: stable@vger.kernel.org
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://patch.msgid.link/20260518090656.134588-2-sgarzare@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.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>
vmci_transport_recv_connecting_server() returned err = 0 for a peer
RST in its default switch arm:
err = pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST ? 0 : -EINVAL;
That made vmci_transport_recv_listen() skip vsock_remove_pending(),
leaving the pending socket on the listener's pending_links with
sk_state = TCP_CLOSE while destroy: still dropped the explicit
reference taken before schedule_delayed_work().
One second later vsock_pending_work() observed is_pending=true and
performed full cleanup: vsock_remove_pending() then the two trailing
sock_put(sk) calls -- the first reached refcount 0 and __sk_freed
the socket, and the second wrote into the freed object:
BUG: KASAN: slab-use-after-free in refcount_warn_saturate
Write of size 4 at addr ffff88800b1cac80 by task kworker
Workqueue: events vsock_pending_work
Treat peer RST like any other unexpected packet type (err = -EINVAL).
All destroy: arms now return err < 0, so vmci_transport_recv_listen()
removes pending from pending_links synchronously and
vsock_pending_work() takes the is_pending=false / !rejected branch,
dropping only its own work reference. This also closes the
multi-packet race Sashiko reported on v2: pending is removed from
the list before any subsequent packet can find it.
The pre-existing sk_acceptq_removed() gap on the err < 0 path of
vmci_transport_recv_listen() that Sashiko also noted is not
introduced or changed by this patch.
Tested on lts-6.12.79 with KASAN: 52/100 unpatched -> 0/100 patched.
Fixes: d021c34405 ("VSOCK: Introduce VM Sockets")
Cc: stable@vger.kernel.org
Signed-off-by: Minh Nguyen <minhnguyen.080505@gmail.com>
Acked-by: Bryan Tan <bryan-bt.tan@broadcom.com>
Link: https://patch.msgid.link/20260519102310.237181-1-minhnguyen.080505@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
