Martin KaFai Lau says:
====================
pull-request: bpf-next 2025-11-10
We've added 19 non-merge commits during the last 3 day(s) which contain
a total of 22 files changed, 1345 insertions(+), 197 deletions(-).
The main changes are:
1) Preserve skb metadata after a TC BPF program has changed the skb,
from Jakub Sitnicki.
This allows a TC program at the end of a TC filter chain to still see
the skb metadata, even if another TC program at the front of the chain
has changed the skb using BPF helpers.
2) Initial af_smc bpf_struct_ops support to control the smc specific
syn/synack options, from D. Wythe.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next:
bpf/selftests: Add selftest for bpf_smc_hs_ctrl
net/smc: bpf: Introduce generic hook for handshake flow
bpf: Export necessary symbols for modules with struct_ops
selftests/bpf: Cover skb metadata access after bpf_skb_change_proto
selftests/bpf: Cover skb metadata access after change_head/tail helper
selftests/bpf: Cover skb metadata access after bpf_skb_adjust_room
selftests/bpf: Cover skb metadata access after vlan push/pop helper
selftests/bpf: Expect unclone to preserve skb metadata
selftests/bpf: Dump skb metadata on verification failure
selftests/bpf: Verify skb metadata in BPF instead of userspace
bpf: Make bpf_skb_change_head helper metadata-safe
bpf: Make bpf_skb_change_proto helper metadata-safe
bpf: Make bpf_skb_adjust_room metadata-safe
bpf: Make bpf_skb_vlan_push helper metadata-safe
bpf: Make bpf_skb_vlan_pop helper metadata-safe
vlan: Make vlan_remove_tag return nothing
bpf: Unclone skb head on bpf_dynptr_write to skb metadata
net: Preserve metadata on pskb_expand_head
net: Helper to move packet data and metadata after skb_push/pull
====================
Link: https://patch.msgid.link/20251110232427.3929291-1-martin.lau@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
syzbot reported a possible shift-out-of-bounds [1]
Blamed commit added rto_alpha_max and rto_beta_max set to 1000.
It is unclear if some sctp users are setting very large rto_alpha
and/or rto_beta.
In order to prevent user regression, perform the test at run time.
Also add READ_ONCE() annotations as sysctl values can change under us.
[1]
UBSAN: shift-out-of-bounds in net/sctp/transport.c:509:41
shift exponent 64 is too large for 32-bit type 'unsigned int'
CPU: 0 UID: 0 PID: 16704 Comm: syz.2.2320 Not tainted syzkaller #0 PREEMPT(full)
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/02/2025
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:94 [inline]
dump_stack_lvl+0x16c/0x1f0 lib/dump_stack.c:120
ubsan_epilogue lib/ubsan.c:233 [inline]
__ubsan_handle_shift_out_of_bounds+0x27f/0x420 lib/ubsan.c:494
sctp_transport_update_rto.cold+0x1c/0x34b net/sctp/transport.c:509
sctp_check_transmitted+0x11c4/0x1c30 net/sctp/outqueue.c:1502
sctp_outq_sack+0x4ef/0x1b20 net/sctp/outqueue.c:1338
sctp_cmd_process_sack net/sctp/sm_sideeffect.c:840 [inline]
sctp_cmd_interpreter net/sctp/sm_sideeffect.c:1372 [inline]
Fixes: b58537a1f5 ("net: sctp: fix permissions for rto_alpha and rto_beta knobs")
Reported-by: syzbot+f8c46c8b2b7f6e076e99@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/690c81ae.050a0220.3d0d33.014e.GAE@google.com/T/#u
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Xin Long <lucien.xin@gmail.com>
Link: https://patch.msgid.link/20251106111054.3288127-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Contrary to what was stated on d36349ea73 ("Bluetooth: hci_conn:
Fix running bis_cleanup for hci_conn->type PA_LINK") the PA_LINK does
in fact needs to run bis_cleanup in order to terminate the PA Sync,
since that is bond to the listening socket which is the entity that
controls the lifetime of PA Sync, so if it is closed/released the PA
Sync shall be terminated, terminating the PA Sync shall not result in
the BIG Sync being terminated since once the later is established it
doesn't depend on the former anymore.
If the use user wants to reconnect/rebind a number of BIS(s) it shall
keep the socket open until it no longer needs the PA Sync, which means
it retains full control of the lifetime of both PA and BIG Syncs.
Fixes: d36349ea73 ("Bluetooth: hci_conn: Fix running bis_cleanup for hci_conn->type PA_LINK")
Fixes: a7bcffc673 ("Bluetooth: Add PA_LINK to distinguish BIG sync and PA sync connections")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
l2cap_chan_close() needs to be called in l2cap_chan_lock(), otherwise
l2cap_le_sig_cmd() etc. may run concurrently.
Add missing locks around l2cap_chan_close().
Fixes: 6b8d4a6a03 ("Bluetooth: 6LoWPAN: Use connected oriented channel instead of fixed one")
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
disconnect_all_peers() calls sleeping function (l2cap_chan_close) under
spinlock. Holding the lock doesn't actually do any good -- we work on a
local copy of the list, and the lock doesn't protect against peer->chan
having already been freed.
Fix by taking refcounts of peer->chan instead. Clean up the code and
old comments a bit.
Take devices_lock instead of RCU, because the kfree_rcu();
l2cap_chan_put(); construct in chan_close_cb() does not guarantee
peer->chan is necessarily valid in RCU.
Also take l2cap_chan_lock() which is required for l2cap_chan_close().
Log: (bluez 6lowpan-tester Client Connect - Disable)
------
BUG: sleeping function called from invalid context at kernel/locking/mutex.c:575
...
<TASK>
...
l2cap_send_disconn_req (net/bluetooth/l2cap_core.c:938 net/bluetooth/l2cap_core.c:1495)
...
? __pfx_l2cap_chan_close (net/bluetooth/l2cap_core.c:809)
do_enable_set (net/bluetooth/6lowpan.c:1048 net/bluetooth/6lowpan.c:1068)
------
Fixes: 9030582963 ("Bluetooth: 6lowpan: Converting rwlocks to use RCU")
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
l2cap_chan_put() is exported, so export also l2cap_chan_hold() for
modules.
l2cap_chan_hold() has use case in net/bluetooth/6lowpan.c
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Bluetooth 6lowpan.c confuses BDADDR_LE and ADDR_LE_DEV address types,
e.g. debugfs "connect" command takes the former, and "disconnect" and
"connect" to already connected device take the latter. This is due to
using same value both for l2cap_chan_connect and hci_conn_hash_lookup_le
which take different dst_type values.
Fix address type passed to hci_conn_hash_lookup_le().
Retain the debugfs API difference between "connect" and "disconnect"
commands since it's been like this since 2015 and nobody apparently
complained.
Fixes: f5ad4ffceb ("Bluetooth: 6lowpan: Use hci_conn_hash_lookup_le() when possible")
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Bluetooth 6lowpan.c netdev has header_ops, so it must set link-local
header for RX skb, otherwise things crash, eg. with AF_PACKET SOCK_RAW
Add missing skb_reset_mac_header() for uncompressed ipv6 RX path.
For the compressed one, it is done in lowpan_header_decompress().
Log: (BlueZ 6lowpan-tester Client Recv Raw - Success)
------
kernel BUG at net/core/skbuff.c:212!
Call Trace:
<IRQ>
...
packet_rcv (net/packet/af_packet.c:2152)
...
<TASK>
__local_bh_enable_ip (kernel/softirq.c:407)
netif_rx (net/core/dev.c:5648)
chan_recv_cb (net/bluetooth/6lowpan.c:294 net/bluetooth/6lowpan.c:359)
------
Fixes: 18722c2470 ("Bluetooth: Enable 6LoWPAN support for BT LE devices")
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
mesh_send_done timer is not canceled when hdev is removed, which causes
crash if the timer triggers after hdev is gone.
Cancel the timer when MGMT removes the hdev, like other MGMT timers.
Should fix the BUG: sporadically seen by BlueZ test bot
(in "Mesh - Send cancel - 1" test).
Log:
------
BUG: KASAN: slab-use-after-free in run_timer_softirq+0x76b/0x7d0
...
Freed by task 36:
kasan_save_stack+0x24/0x50
kasan_save_track+0x14/0x30
__kasan_save_free_info+0x3a/0x60
__kasan_slab_free+0x43/0x70
kfree+0x103/0x500
device_release+0x9a/0x210
kobject_put+0x100/0x1e0
vhci_release+0x18b/0x240
------
Fixes: b338d91703 ("Bluetooth: Implement support for Mesh")
Link: https://lore.kernel.org/linux-bluetooth/67364c09.0c0a0220.113cba.39ff@mx.google.com/
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
The introduction of IPPROTO_SMC enables eBPF programs to determine
whether to use SMC based on the context of socket creation, such as
network namespaces, PID and comm name, etc.
As a subsequent enhancement, to introduce a new generic hook that
allows decisions on whether to use SMC or not at runtime, including
but not limited to local/remote IP address or ports.
User can write their own implememtion via bpf_struct_ops now to choose
whether to use SMC or not before TCP 3rd handshake to be comleted.
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
Link: https://patch.msgid.link/20251107035632.115950-3-alibuda@linux.alibaba.com
Currently bpf_dynptr_from_skb_meta() marks the dynptr as read-only when
the skb is cloned, preventing writes to metadata.
Remove this restriction and unclone the skb head on bpf_dynptr_write() to
metadata, now that the metadata is preserved during uncloning. This makes
metadata dynptr consistent with skb dynptr, allowing writes regardless of
whether the skb is cloned.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://patch.msgid.link/20251105-skb-meta-rx-path-v4-3-5ceb08a9b37b@cloudflare.com
Each link can currently use a chanctx and have another one
reserved, and both of these are also tracked backwards in
the assigned_links and reserved_links lists. If we consider
that there aren't *that* many links, this duplicate book-
keeping isn't necessary.
(I think it used to be necessary before the wiphy locking
changes, when chanctx_mtx existed, because we couldn't do
any interface iterations while holding only chanctx_mtx.)
Additionally, for NAN, we're going to want to track which
chanctxs are in use by the (group of) NAN interfaces. For
those, links don't really make sense as such, so chanctxs
need to be assigned to a different data structure.
Thus, as a first step, remove those back-lists of users
(right now only links) of each channel context. This is a
very basic conversion, ieee80211_vif_use_reserved_switch()
should made to iterate smarter.
Link: https://patch.msgid.link/20251105160431.dbeea1c42e76.I8d273c407274e1c05a4778aa20b56a9f326e87a7@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The current link iteration macros for_each_sdata_link() and
for_each_sdata_link_rcu() are various nested for loops, but
because they iterate all sdata instances and then all links
inside, using 'break' inside the iteration doesn't actually
break out of the whole iteration.
Make it work by tracking whether or not the inner iteration
(over links) actually completed, if it broke out then given
list_for_each_entry() it still iterates all sdata instances
but won't go into them.
Link: https://patch.msgid.link/20251105160431.c21956654fc0.I8d4739af061c44c57d172f19a15303a44ad1e596@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2025-11-06 (i40, ice, iavf)
Mohammad Heib introduces a new devlink parameter, max_mac_per_vf, for
controlling the maximum number of MAC address filters allowed by a VF. This
allows administrators to control the VF behavior in a more nuanced manner.
Aleksandr and Przemek add support for Receive Side Scaling of GTP to iAVF
for VFs running on E800 series ice hardware. This improves performance and
scalability for virtualized network functions in 5G and LTE deployments.
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
iavf: add RSS support for GTP protocol via ethtool
ice: Extend PTYPE bitmap coverage for GTP encapsulated flows
ice: improve TCAM priority handling for RSS profiles
ice: implement GTP RSS context tracking and configuration
ice: add virtchnl definitions and static data for GTP RSS
ice: add flow parsing for GTP and new protocol field support
i40e: support generic devlink param "max_mac_per_vf"
devlink: Add new "max_mac_per_vf" generic device param
====================
Link: https://patch.msgid.link/20251106225321.1609605-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
skb_defer_free_flush() is becoming more important these days.
Add a prefetch operation to reduce latency a bit on some
platforms like AMD EPYC 7B12.
On more recent cpus, a stall happens when reading skb_shinfo().
Avoiding it will require a more elaborate strategy.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://patch.msgid.link/20251106085500.2438951-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
TCP SACK compression has been added in 2018 in commit
5d9f4262b7 ("tcp: add SACK compression").
It is working great for WAN flows (with large RTT).
Wifi in particular gets a significant boost _when_ ACK are suppressed.
Add a new sysctl so that we can tune the very conservative 5 % value
that has been used so far in this formula, so that small RTT flows
can benefit from this feature.
delay = min ( 5 % of RTT, 1 ms)
This patch adds new tcp_comp_sack_rtt_percent sysctl
to ease experiments and tuning.
Given that we cap the delay to 1ms (tcp_comp_sack_delay_ns sysctl),
set the default value to 33 %.
Quoting Neal Cardwell ( https://lore.kernel.org/netdev/CADVnQymZ1tFnEA1Q=vtECs0=Db7zHQ8=+WCQtnhHFVbEOzjVnQ@mail.gmail.com/ )
The rationale for 33% is basically to try to facilitate pipelining,
where there are always at least 3 ACKs and 3 GSO/TSO skbs per SRTT, so
that the path can maintain a budget for 3 full-sized GSO/TSO skbs "in
flight" at all times:
+ 1 skb in the qdisc waiting to be sent by the NIC next
+ 1 skb being sent by the NIC (being serialized by the NIC out onto the wire)
+ 1 skb being received and aggregated by the receiver machine's
aggregation mechanism (some combination of LRO, GRO, and sack
compression)
Note that this is basically the same magic number (3) and the same
rationales as:
(a) tcp_tso_should_defer() ensuring that we defer sending data for no
longer than cwnd/tcp_tso_win_divisor (where tcp_tso_win_divisor = 3),
and
(b) bbr_quantization_budget() ensuring that cwnd is at least 3 GSO/TSO
skbs to maintain pipelining and full throughput at low RTTs
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Link: https://patch.msgid.link/20251106115236.3450026-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The `len` member of the sk_buff is an unsigned int. This is cast to
`ssize_t` (a signed type) for the first sk_buff in the comparison,
but not the second sk_buff. On 32-bit systems, this can result in
an integer underflow for certain values because unsigned arithmetic
is being used.
This appears to be an oversight: if the intention was to use unsigned
arithmetic, then the first cast would have been omitted. The change
ensures both len values are cast to `ssize_t`.
The underflow causes an issue with ktls when multiple TLS PDUs are
included in a single TCP segment. The mainline kernel does not use
strparser for ktls anymore, but this is still useful for other
features that still use strparser, and for backporting.
Signed-off-by: Nate Karstens <nate.karstens@garmin.com>
Cc: stable@vger.kernel.org
Fixes: 43a0c6751a ("strparser: Stream parser for messages")
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/20251106222835.1871628-1-nate.karstens@garmin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 5903123f66 ("tcp: Use BPF timeout setting for SYN ACK
RTO") introduced req->timeout and initialised it in 3 places:
1. reqsk_alloc() sets 0
2. inet_reqsk_alloc() sets TCP_TIMEOUT_INIT
3. tcp_conn_request() sets tcp_timeout_init()
1. has been always redundant as 2. overwrites it immediately.
2. was necessary for TFO SYN+ACK but is no longer needed after
commit 8ea731d4c2 ("tcp: Make SYN ACK RTO tunable by BPF
programs with TFO").
3. was moved to reqsk_queue_hash_req() in the previous patch.
Now, we always initialise req->timeout just before scheduling
the SYN+ACK timer:
* For non-TFO SYN+ACK : reqsk_queue_hash_req()
* For TFO SYN+ACK : tcp_fastopen_create_child()
Let's remove the redundant initialisation of req->timeout in
reqsk_alloc() and inet_reqsk_alloc().
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20251106003357.273403-4-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a new device generic parameter to controls the maximum
number of MAC filters allowed per VF.
For example, to limit a VF to 3 MAC addresses:
$ devlink dev param set pci/0000:3b:00.0 name max_mac_per_vf \
value 3 \
cmode runtime
Signed-off-by: Mohammad Heib <mheib@redhat.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
While connecting, the MAC address can already no longer be
changed. The change is already rejected if netif_carrier_ok(),
but of course that's not true yet while connecting. Check for
auth_data or assoc_data, so the MAC address cannot be changed.
Also more comprehensively check that there are no stations on
the interface being changed - if any peer station is added it
will know about our address already, so we cannot change it.
Cc: stable@vger.kernel.org
Fixes: 3c06e91b40 ("wifi: mac80211: Support POWERED_ADDR_CHANGE feature")
Link: https://patch.msgid.link/20251105154119.f9f6c1df81bb.I9bb3760ede650fb96588be0d09a5a7bdec21b217@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Cross-merge networking fixes after downstream PR (net-6.18-rc5).
Conflicts:
drivers/net/wireless/ath/ath12k/mac.c
9222582ec5 ("Revert "wifi: ath12k: Fix missing station power save configuration"")
6917e268c4 ("wifi: ath12k: Defer vdev bring-up until CSA finalize to avoid stale beacon")
https://lore.kernel.org/11cece9f7e36c12efd732baa5718239b1bf8c950.camel@sipsolutions.net
Adjacent changes:
drivers/net/ethernet/intel/Kconfig
b1d16f7c00 ("libie: depend on DEBUG_FS when building LIBIE_FWLOG")
93f53db9f9 ("ice: switch to Page Pool")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Jakub Kicinski:
Including fixes from bluetooth and wireless.
Current release - new code bugs:
- ptp: expose raw cycles only for clocks with free-running counter
- bonding: fix null-deref in actor_port_prio setting
- mdio: ERR_PTR-check regmap pointer returned by
device_node_to_regmap()
- eth: libie: depend on DEBUG_FS when building LIBIE_FWLOG
Previous releases - regressions:
- virtio_net: fix perf regression due to bad alignment of
virtio_net_hdr_v1_hash
- Revert "wifi: ath10k: avoid unnecessary wait for service ready
message" caused regressions for QCA988x and QCA9984
- Revert "wifi: ath12k: Fix missing station power save configuration"
caused regressions for WCN7850
- eth: bnxt_en: shutdown FW DMA in bnxt_shutdown(), fix memory
corruptions after kexec
Previous releases - always broken:
- virtio-net: fix received packet length check for big packets
- sctp: fix races in socket diag handling
- wifi: add an hrtimer-based delayed work item to avoid low
granularity of timers set relatively far in the future, and use it
where it matters (e.g. when performing AP-scheduled channel switch)
- eth: mlx5e:
- correctly propagate error in case of module EEPROM read failure
- fix HW-GRO on systems with PAGE_SIZE == 64kB
- dsa: b53: fixes for tagging, link configuration / RMII, FDB,
multicast
- phy: lan8842: implement latest errata"
* tag 'net-6.18-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (63 commits)
selftests/vsock: avoid false-positives when checking dmesg
net: bridge: fix MST static key usage
net: bridge: fix use-after-free due to MST port state bypass
lan966x: Fix sleeping in atomic context
bonding: fix NULL pointer dereference in actor_port_prio setting
net: dsa: microchip: Fix reserved multicast address table programming
net: wan: framer: pef2256: Switch to devm_mfd_add_devices()
net: libwx: fix device bus LAN ID
net/mlx5e: SHAMPO, Fix header formulas for higher MTUs and 64K pages
net/mlx5e: SHAMPO, Fix skb size check for 64K pages
net/mlx5e: SHAMPO, Fix header mapping for 64K pages
net: ti: icssg-prueth: Fix fdb hash size configuration
net/mlx5e: Fix return value in case of module EEPROM read error
net: gro_cells: Reduce lock scope in gro_cell_poll
libie: depend on DEBUG_FS when building LIBIE_FWLOG
wifi: mac80211_hwsim: Limit destroy_on_close radio removal to netgroup
netpoll: Fix deadlock in memory allocation under spinlock
net: ethernet: ti: netcp: Standardize knav_dma_open_channel to return NULL on error
virtio-net: fix received length check in big packets
bnxt_en: Fix warning in bnxt_dl_reload_down()
...
syzbot reported[1] a use-after-free when deleting an expired fdb. It is
due to a race condition between learning still happening and a port being
deleted, after all its fdbs have been flushed. The port's state has been
toggled to disabled so no learning should happen at that time, but if we
have MST enabled, it will bypass the port's state, that together with VLAN
filtering disabled can lead to fdb learning at a time when it shouldn't
happen while the port is being deleted. VLAN filtering must be disabled
because we flush the port VLANs when it's being deleted which will stop
learning. This fix adds a check for the port's vlan group which is
initialized to NULL when the port is getting deleted, that avoids the port
state bypass. When MST is enabled there would be a minimal new overhead
in the fast-path because the port's vlan group pointer is cache-hot.
[1] https://syzkaller.appspot.com/bug?extid=dd280197f0f7ab3917be
Fixes: ec7328b591 ("net: bridge: mst: Multiple Spanning Tree (MST) mode")
Reported-by: syzbot+dd280197f0f7ab3917be@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/69088ffa.050a0220.29fc44.003d.GAE@google.com/
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20251105111919.1499702-2-razor@blackwall.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
One GRO-cell device's NAPI callback can nest into the GRO-cell of
another device if the underlying device is also using GRO-cell.
This is the case for IPsec over vxlan.
These two GRO-cells are separate devices. From lockdep's point of view
it is the same because each device is sharing the same lock class and so
it reports a possible deadlock assuming one device is nesting into
itself.
Hold the bh_lock only while accessing gro_cell::napi_skbs in
gro_cell_poll(). This reduces the locking scope and avoids acquiring the
same lock class multiple times.
Fixes: 25718fdcbd ("net: gro_cells: Use nested-BH locking for gro_cell")
Reported-by: Gal Pressman <gal@nvidia.com>
Closes: https://lore.kernel.org/all/66664116-edb8-48dc-ad72-d5223696dd19@nvidia.com/
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/20251104153435.ty88xDQt@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix a AA deadlock in refill_skbs() where memory allocation while holding
skb_pool->lock can trigger a recursive lock acquisition attempt.
The deadlock scenario occurs when the system is under severe memory
pressure:
1. refill_skbs() acquires skb_pool->lock (spinlock)
2. alloc_skb() is called while holding the lock
3. Memory allocator fails and calls slab_out_of_memory()
4. This triggers printk() for the OOM warning
5. The console output path calls netpoll_send_udp()
6. netpoll_send_udp() attempts to acquire the same skb_pool->lock
7. Deadlock: the lock is already held by the same CPU
Call stack:
refill_skbs()
spin_lock_irqsave(&skb_pool->lock) <- lock acquired
__alloc_skb()
kmem_cache_alloc_node_noprof()
slab_out_of_memory()
printk()
console_flush_all()
netpoll_send_udp()
skb_dequeue()
spin_lock_irqsave(&skb_pool->lock) <- deadlock attempt
This bug was exposed by commit 248f6571fd ("netpoll: Optimize skb
refilling on critical path") which removed refill_skbs() from the
critical path (where nested printk was being deferred), letting nested
printk being called from inside refill_skbs()
Refactor refill_skbs() to never allocate memory while holding
the spinlock.
Another possible solution to fix this problem is protecting the
refill_skbs() from nested printks, basically calling
printk_deferred_{enter,exit}() in refill_skbs(), then, any nested
pr_warn() would be deferred.
I prefer this approach, given I _think_ it might be a good idea to move
the alloc_skb() from GFP_ATOMIC to GFP_KERNEL in the future, so, having
the alloc_skb() outside of the lock will be necessary step.
There is a possible TOCTOU issue when checking for the pool length, and
queueing the new allocated skb, but, this is not an issue, given that
an extra SKB in the pool is harmless and it will be eventually used.
Signed-off-by: Breno Leitao <leitao@debian.org>
Fixes: 248f6571fd ("netpoll: Optimize skb refilling on critical path")
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20251103-fix_netpoll_aa-v4-1-4cfecdf6da7c@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Revert struct sockaddr from flexible array to fixed 14-byte "sa_data",
to solve over 36,000 -Wflex-array-member-not-at-end warnings, since
struct sockaddr is embedded within many network structs.
With socket/proto sockaddr-based internal APIs switched to use struct
sockaddr_unsized, there should be no more uses of struct sockaddr that
depend on reading beyond the end of struct sockaddr::sa_data that might
trigger bounds checking.
Comparing an x86_64 "allyesconfig" vmlinux build before and after this
patch showed no new "ud1" instructions from CONFIG_UBSAN_BOUNDS nor any
new "field-spanning" memcpy CONFIG_FORTIFY_SOURCE instrumentations.
Cc: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Kees Cook <kees@kernel.org>
Link: https://patch.msgid.link/20251104002617.2752303-8-kees@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
