The length of skb header was incorrectly calculated when transmit a tunnel
packet with outer IPv6 extension header, or a IP over IP packet which has
inner IPv6 header. Thus the correct Tx context descriptor cannot be
composed, resulting in Tx ring hang.
Fixes: 3403960cdf ("net: wangxun: libwx add tx offload functions")
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Link: https://patch.msgid.link/20250324103235.823096-1-jiawenwu@trustnetic.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Remove KSZ88x3-specific priority and apptrust configuration logic that was
based on incorrect register access assumptions. Also fix the register
offset for KSZ8_REG_PORT_1_CTRL_0 to align with get_port_addr() logic.
The KSZ88x3 switch family uses a different register layout compared to
KSZ9477-compatible variants. Specifically, port control registers need
offset adjustment through get_port_addr(), and do not match the datasheet
values directly.
Commit a1ea57710c ("net: dsa: microchip: dcb: add special handling for
KSZ88X3 family") introduced quirks based on datasheet offsets, which do
not work with the driver's internal addressing model. As a result, these
quirks addressed the wrong ports and caused unstable behavior.
This patch removes all KSZ88x3-specific DCB quirks and corrects the port
control register offset, effectively restoring working and predictable
apptrust configuration.
Fixes: a1ea57710c ("net: dsa: microchip: dcb: add special handling for KSZ88X3 family")
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250321141044.2128973-1-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When delete l3s ipvlan:
ip link del link eth0 ipvlan1 type ipvlan mode l3s
This may cause a null pointer dereference:
Call trace:
ip_rcv_finish+0x48/0xd0
ip_rcv+0x5c/0x100
__netif_receive_skb_one_core+0x64/0xb0
__netif_receive_skb+0x20/0x80
process_backlog+0xb4/0x204
napi_poll+0xe8/0x294
net_rx_action+0xd8/0x22c
__do_softirq+0x12c/0x354
This is because l3mdev_l3_rcv() visit dev->l3mdev_ops after
ipvlan_l3s_unregister() assign the dev->l3mdev_ops to NULL. The process
like this:
(CPU1) | (CPU2)
l3mdev_l3_rcv() |
check dev->priv_flags: |
master = skb->dev; |
|
| ipvlan_l3s_unregister()
| set dev->priv_flags
| dev->l3mdev_ops = NULL;
|
visit master->l3mdev_ops |
To avoid this by do not set dev->l3mdev_ops when unregister l3s ipvlan.
Suggested-by: David Ahern <dsahern@kernel.org>
Fixes: c675e06a98 ("ipvlan: decouple l3s mode dependencies from other modes")
Signed-off-by: Wang Liang <wangliang74@huawei.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250321090353.1170545-1-wangliang74@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Previously, when the driver was printing hex dumps, the buffer was cast
to an 8 byte long and printed using string formatters. If the buffer
size was not a multiple of 8 then a read buffer overflow was possible.
Therefore, create a new ibmvnic function that loops over a buffer and
calls hex_dump_to_buffer instead.
This patch address KASAN reports like the one below:
ibmvnic 30000003 env3: Login Buffer:
ibmvnic 30000003 env3: 01000000af000000
<...>
ibmvnic 30000003 env3: 2e6d62692e736261
ibmvnic 30000003 env3: 65050003006d6f63
==================================================================
BUG: KASAN: slab-out-of-bounds in ibmvnic_login+0xacc/0xffc [ibmvnic]
Read of size 8 at addr c0000001331a9aa8 by task ip/17681
<...>
Allocated by task 17681:
<...>
ibmvnic_login+0x2f0/0xffc [ibmvnic]
ibmvnic_open+0x148/0x308 [ibmvnic]
__dev_open+0x1ac/0x304
<...>
The buggy address is located 168 bytes inside of
allocated 175-byte region [c0000001331a9a00, c0000001331a9aaf)
<...>
=================================================================
ibmvnic 30000003 env3: 000000000033766e
Fixes: 032c5e8284 ("Driver for IBM System i/p VNIC protocol")
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Dave Marquardt <davemarq@linux.ibm.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250320212951.11142-1-nnac123@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Following operations can trigger a warning[1]:
ip netns add ns1
ip netns exec ns1 ip link add bond0 type bond mode balance-rr
ip netns exec ns1 ip link set dev bond0 xdp obj af_xdp_kern.o sec xdp
ip netns exec ns1 ip link set bond0 type bond mode broadcast
ip netns del ns1
When delete the namespace, dev_xdp_uninstall() is called to remove xdp
program on bond dev, and bond_xdp_set() will check the bond mode. If bond
mode is changed after attaching xdp program, the warning may occur.
Some bond modes (broadcast, etc.) do not support native xdp. Set bond mode
with xdp program attached is not good. Add check for xdp program when set
bond mode.
[1]
------------[ cut here ]------------
WARNING: CPU: 0 PID: 11 at net/core/dev.c:9912 unregister_netdevice_many_notify+0x8d9/0x930
Modules linked in:
CPU: 0 UID: 0 PID: 11 Comm: kworker/u4:0 Not tainted 6.14.0-rc4 #107
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.15.0-0-g2dd4b9b3f840-prebuilt.qemu.org 04/01/2014
Workqueue: netns cleanup_net
RIP: 0010:unregister_netdevice_many_notify+0x8d9/0x930
Code: 00 00 48 c7 c6 6f e3 a2 82 48 c7 c7 d0 b3 96 82 e8 9c 10 3e ...
RSP: 0018:ffffc90000063d80 EFLAGS: 00000282
RAX: 00000000ffffffa1 RBX: ffff888004959000 RCX: 00000000ffffdfff
RDX: 0000000000000000 RSI: 00000000ffffffea RDI: ffffc90000063b48
RBP: ffffc90000063e28 R08: ffffffff82d39b28 R09: 0000000000009ffb
R10: 0000000000000175 R11: ffffffff82d09b40 R12: ffff8880049598e8
R13: 0000000000000001 R14: dead000000000100 R15: ffffc90000045000
FS: 0000000000000000(0000) GS:ffff888007a00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000000000d406b60 CR3: 000000000483e000 CR4: 00000000000006f0
Call Trace:
<TASK>
? __warn+0x83/0x130
? unregister_netdevice_many_notify+0x8d9/0x930
? report_bug+0x18e/0x1a0
? handle_bug+0x54/0x90
? exc_invalid_op+0x18/0x70
? asm_exc_invalid_op+0x1a/0x20
? unregister_netdevice_many_notify+0x8d9/0x930
? bond_net_exit_batch_rtnl+0x5c/0x90
cleanup_net+0x237/0x3d0
process_one_work+0x163/0x390
worker_thread+0x293/0x3b0
? __pfx_worker_thread+0x10/0x10
kthread+0xec/0x1e0
? __pfx_kthread+0x10/0x10
? __pfx_kthread+0x10/0x10
ret_from_fork+0x2f/0x50
? __pfx_kthread+0x10/0x10
ret_from_fork_asm+0x1a/0x30
</TASK>
---[ end trace 0000000000000000 ]---
Fixes: 9e2ee5c7e7 ("net, bonding: Add XDP support to the bonding driver")
Signed-off-by: Wang Liang <wangliang74@huawei.com>
Acked-by: Jussi Maki <joamaki@gmail.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20250321044852.1086551-1-wangliang74@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
vmxnet3 does not unregister xdp rxq info in the
vmxnet3_reset_work() code path as vmxnet3_rq_destroy()
is not invoked in this code path. So, we get below message with a
backtrace.
Missing unregister, handled but fix driver
WARNING: CPU:48 PID: 500 at net/core/xdp.c:182
__xdp_rxq_info_reg+0x93/0xf0
This patch fixes the problem by moving the unregister
code of XDP from vmxnet3_rq_destroy() to vmxnet3_rq_cleanup().
Fixes: 54f00cce11 ("vmxnet3: Add XDP support.")
Signed-off-by: Sankararaman Jayaraman <sankararaman.jayaraman@broadcom.com>
Signed-off-by: Ronak Doshi <ronak.doshi@broadcom.com>
Link: https://patch.msgid.link/20250320045522.57892-1-sankararaman.jayaraman@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2025-03-18 (ice, idpf)
For ice:
Przemek modifies string declarations to resolve compile issues on gcc 7.5.
Karol adds padding to initial programming of GLTSYN_TIME* registers to
ensure it will occur in the future to prevent hardware issues.
Jesse Brandeburg turns off driver RDMA capability when the corresponding
kernel config is not enabled to aid in preventing resource exhaustion.
Jan adjusts type declaration to properly catch error conditions and
prevent truncation of values. He also adds bounds checking to prevent
overflow in ice_vc_cfg_q_quanta().
Lukasz adds checking and error reporting for invalid values in
ice_vc_cfg_q_bw().
Mateusz adds check for valid size for ice_vc_fdir_parse_raw().
For idpf:
Emil adds check, and handling, on failure to register netdev.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
idpf: check error for register_netdev() on init
ice: fix using untrusted value of pkt_len in ice_vc_fdir_parse_raw()
ice: fix input validation for virtchnl BW
ice: validate queue quanta parameters to prevent OOB access
ice: stop truncating queue ids when checking
virtchnl: make proto and filter action count unsigned
ice: fix reservation of resources for RDMA when disabled
ice: ensure periodic output start time is in the future
ice: health.c: fix compilation on gcc 7.5
====================
Link: https://patch.msgid.link/20250318200511.2958251-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Vladimir Oltean says:
====================
sja1105 driver fixes
This is a collection of 3 fixes for the sja1105 DSA driver:
- 1/3: "ethtool -S" shows a bogus counter with no name, and doesn't show
a valid counter because of it (either "n_not_reach" or "n_rx_bcast").
- 2/3: RX timestamping filters other than L2 PTP event messages don't work,
but are not rejected, either.
- 3/3: there is a KASAN out-of-bounds warning in sja1105_table_delete_entry()
====================
Link: https://patch.msgid.link/20250318115716.2124395-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are actually 2 problems:
- deleting the last element doesn't require the memmove of elements
[i + 1, end) over it. Actually, element i+1 is out of bounds.
- The memmove itself should move size - i - 1 elements, because the last
element is out of bounds.
The out-of-bounds element still remains out of bounds after being
accessed, so the problem is only that we touch it, not that it becomes
in active use. But I suppose it can lead to issues if the out-of-bounds
element is part of an unmapped page.
Fixes: 6666cebc5e ("net: dsa: sja1105: Add support for VLAN operations")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250318115716.2124395-4-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is all that we can support timestamping, so we shouldn't accept
anything else. Also see sja1105_hwtstamp_get().
To avoid erroring out in an inconsistent state, operate on copies of
priv->hwts_rx_en and priv->hwts_tx_en, and write them back when nothing
else can fail anymore.
Fixes: a602afd200 ("net: dsa: sja1105: Expose PTP timestamping ioctls to userspace")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250318115716.2124395-3-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Port counters with no name (aka
sja1105_port_counters[__SJA1105_COUNTER_UNUSED]) are skipped when
reporting sja1105_get_sset_count(), but are not skipped during
sja1105_get_strings() and sja1105_get_ethtool_stats().
As a consequence, the first reported counter has an empty name and a
bogus value (reads from area 0, aka MAC, from offset 0, bits start:end
0:0). Also, the last counter (N_NOT_REACH on E/T, N_RX_BCAST on P/Q/R/S)
gets pushed out of the statistics counters that get shown.
Skip __SJA1105_COUNTER_UNUSED consistently, so that the bogus counter
with an empty name disappears, and in its place appears a valid counter.
Fixes: 039b167d68 ("net: dsa: sja1105: don't use burst SPI reads for port statistics")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250318115716.2124395-2-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is a workaround to mitigate a compiler anomaly.
During LLVM toolchain compilation of this driver on s390x architecture, an
unreasonable __write_overflow_field warning occurs.
Contextually, chunk_index is restricted to 0, 1 or 2. By expanding these
possibilities, the compile warning is suppressed.
Fix follow error with clang-19 when -Werror:
In file included from drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_bloom_filter.c:5:
In file included from ./include/linux/gfp.h:7:
In file included from ./include/linux/mmzone.h:8:
In file included from ./include/linux/spinlock.h:63:
In file included from ./include/linux/lockdep.h:14:
In file included from ./include/linux/smp.h:13:
In file included from ./include/linux/cpumask.h:12:
In file included from ./include/linux/bitmap.h:13:
In file included from ./include/linux/string.h:392:
./include/linux/fortify-string.h:571:4: error: call to '__write_overflow_field' declared with 'warning' attribute: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror,-Wattribute-warning]
571 | __write_overflow_field(p_size_field, size);
| ^
1 error generated.
According to the testing, we can be fairly certain that this is a clang
compiler bug, impacting only clang-19 and below. Clang versions 20 and
21 do not exhibit this behavior.
Link: https://lore.kernel.org/all/484364B641C901CD+20250311141025.1624528-1-wangyuli@uniontech.com/
Fixes: 7585cacdb9 ("mlxsw: spectrum_acl: Add Bloom filter handling")
Co-developed-by: Zijian Chen <czj2441@163.com>
Signed-off-by: Zijian Chen <czj2441@163.com>
Co-developed-by: Wentao Guan <guanwentao@uniontech.com>
Signed-off-by: Wentao Guan <guanwentao@uniontech.com>
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Co-developed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: WangYuli <wangyuli@uniontech.com>
Signed-off-by: WangYuli <wangyuli@uniontech.com>
Link: https://patch.msgid.link/A1858F1D36E653E0+20250318103654.708077-1-wangyuli@uniontech.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Implement the workaround for erratum
3.3 RGMII timing may be out of spec when transmit delay is enabled
for the 6320 family, which says:
When transmit delay is enabled via Port register 1 bit 14 = 1, duty
cycle may be out of spec. Under very rare conditions this may cause
the attached device receive CRC errors.
Signed-off-by: Marek Behún <kabel@kernel.org>
Cc: <stable@vger.kernel.org> # 5.4.x
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250317173250.28780-8-kabel@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Michael Chan says:
====================
bnxt_en: Fix MAX_SKB_FRAGS > 30
The driver was written with the assumption that MAX_SKB_FRAGS could
not exceed what the NIC can support. About 2 years ago,
CONFIG_MAX_SKB_FRAGS was added. The value can exceed what the NIC
can support and it may cause TX timeout. These 2 patches will fix
the issue.
====================
Link: https://patch.msgid.link/20250321211639.3812992-1-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The bd_cnt field in the TX BD specifies the total number of BDs for
the TX packet. The bd_cnt field has 5 bits and the maximum number
supported is 32 with the value 0.
CONFIG_MAX_SKB_FRAGS can be modified and the total number of SKB
fragments can approach or exceed the maximum supported by the chip.
Add a macro to properly mask the bd_cnt field so that the value 32
will be properly masked and set to 0 in the bd_cnd field.
Without this patch, the out-of-range bd_cnt value will corrupt the
TX BD and may cause TX timeout.
The next patch will check for values exceeding 32.
Fixes: 3948b05950 ("net: introduce a config option to tweak MAX_SKB_FRAGS")
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250321211639.3812992-2-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There commands can be used to add an RSS context and steer some traffic
into it:
# ethtool -X eth0 context new
New RSS context is 1
# ethtool -N eth0 flow-type ip4 dst-ip 1.1.1.1 context 1
Added rule with ID 1023
However, the second command fails with EINVAL on mlx5e:
# ethtool -N eth0 flow-type ip4 dst-ip 1.1.1.1 context 1
rmgr: Cannot insert RX class rule: Invalid argument
Cannot insert classification rule
It happens when flow_get_tirn calls flow_type_to_traffic_type with
flow_type = IP_USER_FLOW or IPV6_USER_FLOW. That function only handles
IPV4_FLOW and IPV6_FLOW cases, but unlike all other cases which are
common for hash and spec, IPv4 and IPv6 defines different contants for
hash and for spec:
#define TCP_V4_FLOW 0x01 /* hash or spec (tcp_ip4_spec) */
#define UDP_V4_FLOW 0x02 /* hash or spec (udp_ip4_spec) */
...
#define IPV4_USER_FLOW 0x0d /* spec only (usr_ip4_spec) */
#define IP_USER_FLOW IPV4_USER_FLOW
#define IPV6_USER_FLOW 0x0e /* spec only (usr_ip6_spec; nfc only) */
#define IPV4_FLOW 0x10 /* hash only */
#define IPV6_FLOW 0x11 /* hash only */
Extend the switch in flow_type_to_traffic_type to support both, which
fixes the failing ethtool -N command with flow-type ip4 or ip6.
Fixes: 248d3b4c9a ("net/mlx5e: Support flow classification into RSS contexts")
Signed-off-by: Maxim Mikityanskiy <maxim@isovalent.com>
Tested-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250319124508.3979818-1-maxim@isovalent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Binding AX25 socket by using the autobind feature leads to memory leaks
in ax25_connect() and also refcount leaks in ax25_release(). Memory
leak was detected with kmemleak:
================================================================
unreferenced object 0xffff8880253cd680 (size 96):
backtrace:
__kmalloc_node_track_caller_noprof (./include/linux/kmemleak.h:43)
kmemdup_noprof (mm/util.c:136)
ax25_rt_autobind (net/ax25/ax25_route.c:428)
ax25_connect (net/ax25/af_ax25.c:1282)
__sys_connect_file (net/socket.c:2045)
__sys_connect (net/socket.c:2064)
__x64_sys_connect (net/socket.c:2067)
do_syscall_64 (arch/x86/entry/common.c:52 arch/x86/entry/common.c:83)
entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:130)
================================================================
When socket is bound, refcounts must be incremented the way it is done
in ax25_bind() and ax25_setsockopt() (SO_BINDTODEVICE). In case of
autobind, the refcounts are not incremented.
This bug leads to the following issue reported by Syzkaller:
================================================================
ax25_connect(): syz-executor318 uses autobind, please contact jreuter@yaina.de
------------[ cut here ]------------
refcount_t: decrement hit 0; leaking memory.
WARNING: CPU: 0 PID: 5317 at lib/refcount.c:31 refcount_warn_saturate+0xfa/0x1d0 lib/refcount.c:31
Modules linked in:
CPU: 0 UID: 0 PID: 5317 Comm: syz-executor318 Not tainted 6.14.0-rc4-syzkaller-00278-gece144f151ac #0
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2~bpo12+1 04/01/2014
RIP: 0010:refcount_warn_saturate+0xfa/0x1d0 lib/refcount.c:31
...
Call Trace:
<TASK>
__refcount_dec include/linux/refcount.h:336 [inline]
refcount_dec include/linux/refcount.h:351 [inline]
ref_tracker_free+0x6af/0x7e0 lib/ref_tracker.c:236
netdev_tracker_free include/linux/netdevice.h:4302 [inline]
netdev_put include/linux/netdevice.h:4319 [inline]
ax25_release+0x368/0x960 net/ax25/af_ax25.c:1080
__sock_release net/socket.c:647 [inline]
sock_close+0xbc/0x240 net/socket.c:1398
__fput+0x3e9/0x9f0 fs/file_table.c:464
__do_sys_close fs/open.c:1580 [inline]
__se_sys_close fs/open.c:1565 [inline]
__x64_sys_close+0x7f/0x110 fs/open.c:1565
do_syscall_x64 arch/x86/entry/common.c:52 [inline]
do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
entry_SYSCALL_64_after_hwframe+0x77/0x7f
...
</TASK>
================================================================
Considering the issues above and the comments left in the code that say:
"check if we can remove this feature. It is broken."; "autobinding in this
may or may not work"; - it is better to completely remove this feature than
to fix it because it is broken and leads to various kinds of memory bugs.
Now calling connect() without first binding socket will result in an
error (-EINVAL). Userspace software that relies on the autobind feature
might get broken. However, this feature does not seem widely used with
this specific driver as it was not reliable at any point of time, and it
is already broken anyway. E.g. ax25-tools and ax25-apps packages for
popular distributions do not use the autobind feature for AF_AX25.
Found by Linux Verification Center (linuxtesting.org) with Syzkaller.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Reported-by: syzbot+33841dc6aa3e1d86b78a@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=33841dc6aa3e1d86b78a
Signed-off-by: Murad Masimov <m.masimov@mt-integration.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
SIOCBRDELIF is passed to dev_ioctl() first and later forwarded to
br_ioctl_call(), which causes unnecessary RTNL dance and the splat
below [0] under RTNL pressure.
Let's say Thread A is trying to detach a device from a bridge and
Thread B is trying to remove the bridge.
In dev_ioctl(), Thread A bumps the bridge device's refcnt by
netdev_hold() and releases RTNL because the following br_ioctl_call()
also re-acquires RTNL.
In the race window, Thread B could acquire RTNL and try to remove
the bridge device. Then, rtnl_unlock() by Thread B will release RTNL
and wait for netdev_put() by Thread A.
Thread A, however, must hold RTNL after the unlock in dev_ifsioc(),
which may take long under RTNL pressure, resulting in the splat by
Thread B.
Thread A (SIOCBRDELIF) Thread B (SIOCBRDELBR)
---------------------- ----------------------
sock_ioctl sock_ioctl
`- sock_do_ioctl `- br_ioctl_call
`- dev_ioctl `- br_ioctl_stub
|- rtnl_lock |
|- dev_ifsioc '
' |- dev = __dev_get_by_name(...)
|- netdev_hold(dev, ...) .
/ |- rtnl_unlock ------. |
| |- br_ioctl_call `---> |- rtnl_lock
Race | | `- br_ioctl_stub |- br_del_bridge
Window | | | |- dev = __dev_get_by_name(...)
| | | May take long | `- br_dev_delete(dev, ...)
| | | under RTNL pressure | `- unregister_netdevice_queue(dev, ...)
| | | | `- rtnl_unlock
\ | |- rtnl_lock <-' `- netdev_run_todo
| |- ... `- netdev_run_todo
| `- rtnl_unlock |- __rtnl_unlock
| |- netdev_wait_allrefs_any
|- netdev_put(dev, ...) <----------------'
Wait refcnt decrement
and log splat below
To avoid blocking SIOCBRDELBR unnecessarily, let's not call
dev_ioctl() for SIOCBRADDIF and SIOCBRDELIF.
In the dev_ioctl() path, we do the following:
1. Copy struct ifreq by get_user_ifreq in sock_do_ioctl()
2. Check CAP_NET_ADMIN in dev_ioctl()
3. Call dev_load() in dev_ioctl()
4. Fetch the master dev from ifr.ifr_name in dev_ifsioc()
3. can be done by request_module() in br_ioctl_call(), so we move
1., 2., and 4. to br_ioctl_stub().
Note that 2. is also checked later in add_del_if(), but it's better
performed before RTNL.
SIOCBRADDIF and SIOCBRDELIF have been processed in dev_ioctl() since
the pre-git era, and there seems to be no specific reason to process
them there.
[0]:
unregister_netdevice: waiting for wpan3 to become free. Usage count = 2
ref_tracker: wpan3@ffff8880662d8608 has 1/1 users at
__netdev_tracker_alloc include/linux/netdevice.h:4282 [inline]
netdev_hold include/linux/netdevice.h:4311 [inline]
dev_ifsioc+0xc6a/0x1160 net/core/dev_ioctl.c:624
dev_ioctl+0x255/0x10c0 net/core/dev_ioctl.c:826
sock_do_ioctl+0x1ca/0x260 net/socket.c:1213
sock_ioctl+0x23a/0x6c0 net/socket.c:1318
vfs_ioctl fs/ioctl.c:51 [inline]
__do_sys_ioctl fs/ioctl.c:906 [inline]
__se_sys_ioctl fs/ioctl.c:892 [inline]
__x64_sys_ioctl+0x1a4/0x210 fs/ioctl.c:892
do_syscall_x64 arch/x86/entry/common.c:52 [inline]
do_syscall_64+0xcb/0x250 arch/x86/entry/common.c:83
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Fixes: 893b195875 ("net: bridge: fix ioctl locking")
Reported-by: syzkaller <syzkaller@googlegroups.com>
Reported-by: yan kang <kangyan91@outlook.com>
Reported-by: yue sun <samsun1006219@gmail.com>
Closes: https://lore.kernel.org/netdev/SY8P300MB0421225D54EB92762AE8F0F2A1D32@SY8P300MB0421.AUSP300.PROD.OUTLOOK.COM/
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/20250316192851.19781-1-kuniyu@amazon.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Roopa has decided to withdraw as a bridge maintainer and Ido has agreed to
step up and co-maintain the bridge with me. He has been very helpful in
bridge patch reviews and has contributed a lot to the bridge over the
years. Add an entry for Roopa to CREDITS and also add bridge's headers
to its MAINTAINERS entry.
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250314100631.40999-1-razor@blackwall.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Pull networking fixes from Paolo Abeni:
"Including fixes from can, bluetooth and ipsec.
This contains a last minute revert of a recent GRE patch, mostly to
allow me stating there are no known regressions outstanding.
Current release - regressions:
- revert "gre: Fix IPv6 link-local address generation."
- eth: ti: am65-cpsw: fix NAPI registration sequence
Previous releases - regressions:
- ipv6: fix memleak of nhc_pcpu_rth_output in fib_check_nh_v6_gw().
- mptcp: fix data stream corruption in the address announcement
- bluetooth: fix connection regression between LE and non-LE adapters
- can:
- flexcan: only change CAN state when link up in system PM
- ucan: fix out of bound read in strscpy() source
Previous releases - always broken:
- lwtunnel: fix reentry loops
- ipv6: fix TCP GSO segmentation with NAT
- xfrm: force software GSO only in tunnel mode
- eth: ti: icssg-prueth: add lock to stats
Misc:
- add Andrea Mayer as a maintainer of SRv6"
* tag 'net-6.14-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (33 commits)
MAINTAINERS: Add Andrea Mayer as a maintainer of SRv6
Revert "gre: Fix IPv6 link-local address generation."
Revert "selftests: Add IPv6 link-local address generation tests for GRE devices."
net/neighbor: add missing policy for NDTPA_QUEUE_LENBYTES
tools headers: Sync uapi/asm-generic/socket.h with the kernel sources
mptcp: Fix data stream corruption in the address announcement
selftests: net: test for lwtunnel dst ref loops
net: ipv6: ioam6: fix lwtunnel_output() loop
net: lwtunnel: fix recursion loops
net: ti: icssg-prueth: Add lock to stats
net: atm: fix use after free in lec_send()
xsk: fix an integer overflow in xp_create_and_assign_umem()
net: stmmac: dwc-qos-eth: use devm_kzalloc() for AXI data
selftests: drv-net: use defer in the ping test
phy: fix xa_alloc_cyclic() error handling
dpll: fix xa_alloc_cyclic() error handling
devlink: fix xa_alloc_cyclic() error handling
ipv6: Set errno after ip_fib_metrics_init() in ip6_route_info_create().
ipv6: Fix memleak of nhc_pcpu_rth_output in fib_check_nh_v6_gw().
net: ipv6: fix TCP GSO segmentation with NAT
...
Pull rdma fixes from Jason Gunthorpe:
"Collected driver fixes from the last few weeks, I was surprised how
significant many of them seemed to be.
- Fix rdma-core test failures due to wrong startup ordering in rxe
- Don't crash in bnxt_re if the FW supports more than 64k QPs
- Fix wrong QP table indexing math in bnxt_re
- Calculate the max SRQs for userspace properly in bnxt_re
- Don't try to do math on errno for mlx5's rate calculation
- Properly allow userspace to control the VLAN in the QP state during
INIT->RTR for bnxt_re
- 6 bug fixes for HNS:
- Soft lockup when processing huge MRs, add a cond_resched()
- Fix missed error unwind for doorbell allocation
- Prevent bad send queue parameters from userspace
- Wrong error unwind in qp creation
- Missed xa_destroy during driver shutdown
- Fix reporting to userspace of max_sge_rd, hns doesn't have a
read/write difference"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
RDMA/hns: Fix wrong value of max_sge_rd
RDMA/hns: Fix missing xa_destroy()
RDMA/hns: Fix a missing rollback in error path of hns_roce_create_qp_common()
RDMA/hns: Fix invalid sq params not being blocked
RDMA/hns: Fix unmatched condition in error path of alloc_user_qp_db()
RDMA/hns: Fix soft lockup during bt pages loop
RDMA/bnxt_re: Avoid clearing VLAN_ID mask in modify qp path
RDMA/mlx5: Handle errors returned from mlx5r_ib_rate()
RDMA/bnxt_re: Fix reporting maximum SRQs on P7 chips
RDMA/bnxt_re: Add missing paranthesis in map_qp_id_to_tbl_indx
RDMA/bnxt_re: Fix allocation of QP table
RDMA/rxe: Fix the failure of ibv_query_device() and ibv_query_device_ex() tests
Pull EFI fixes from Ard Biesheuvel:
"Here's a final batch of EFI fixes for v6.14.
The efivarfs ones are fixes for changes that were made this cycle.
James's fix is somewhat of a band-aid, but it was blessed by the VFS
folks, who are working with James to come up with something better for
the next cycle.
- Avoid physical address 0x0 for random page allocations
- Add correct lockdep annotation when traversing efivarfs on resume
- Avoid NULL mount in kernel_file_open() when traversing efivarfs on
resume"
* tag 'efi-fixes-for-v6.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
efivarfs: fix NULL dereference on resume
efivarfs: use I_MUTEX_CHILD nested lock to traverse variables on resume
efi/libstub: Avoid physical address 0x0 when doing random allocation
Guillaume Nault says:
====================
gre: Revert IPv6 link-local address fix.
Following Paolo's suggestion, let's revert the IPv6 link-local address
generation fix for GRE devices. The patch introduced regressions in the
upstream CI, which are still under investigation.
Start by reverting the kselftest that depend on that fix (patch 1), then
revert the kernel code itself (patch 2).
====================
Link: https://patch.msgid.link/cover.1742418408.git.gnault@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Steffen Klassert says:
====================
pull request (net): ipsec 2025-03-19
1) Fix tunnel mode TX datapath in packet offload mode
by directly putting it to the xmit path.
From Alexandre Cassen.
2) Force software GSO only in tunnel mode in favor
of potential HW GSO. From Cosmin Ratiu.
ipsec-2025-03-19
* tag 'ipsec-2025-03-19' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec:
xfrm_output: Force software GSO only in tunnel mode
xfrm: fix tunnel mode TX datapath in packet offload mode
====================
Link: https://patch.msgid.link/20250319065513.987135-1-steffen.klassert@secunet.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Simon Wunderlich says:
====================
Here is batman-adv bugfix:
- Ignore own maximum aggregation size during RX, Sven Eckelmann
* tag 'batadv-net-pullrequest-20250318' of git://git.open-mesh.org/linux-merge:
batman-adv: Ignore own maximum aggregation size during RX
====================
Link: https://patch.msgid.link/20250318150035.35356-1-sw@simonwunderlich.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Previous commit 8b5c171bb3 ("neigh: new unresolved queue limits")
introduces new netlink attribute NDTPA_QUEUE_LENBYTES to represent
approximative value for deprecated QUEUE_LEN. However, it forgot to add
the associated nla_policy in nl_ntbl_parm_policy array. Fix it with one
simple NLA_U32 type policy.
Fixes: 8b5c171bb3 ("neigh: new unresolved queue limits")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Link: https://patch.msgid.link/20250315165113.37600-1-linma@zju.edu.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Because of the size restriction in the TCP options space, the MPTCP
ADD_ADDR option is exclusive and cannot be sent with other MPTCP ones.
For this reason, in the linked mptcp_out_options structure, group of
fields linked to different options are part of the same union.
There is a case where the mptcp_pm_add_addr_signal() function can modify
opts->addr, but not ended up sending an ADD_ADDR. Later on, back in
mptcp_established_options, other options will be sent, but with
unexpected data written in other fields due to the union, e.g. in
opts->ext_copy. This could lead to a data stream corruption in the next
packet.
Using an intermediate variable, prevents from corrupting previously
established DSS option. The assignment of the ADD_ADDR option
parameters is now done once we are sure this ADD_ADDR option can be set
in the packet, e.g. after having dropped other suboptions.
Fixes: 1bff1e43a3 ("mptcp: optimize out option generation")
Cc: stable@vger.kernel.org
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Arthur Mongodin <amongodin@randorisec.fr>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
[ Matt: the commit message has been updated: long lines splits and some
clarifications. ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250314-net-mptcp-fix-data-stream-corr-sockopt-v1-1-122dbb249db3@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Justin Iurman says:
====================
net: fix lwtunnel reentry loops
When the destination is the same after the transformation, we enter a
lwtunnel loop. This is true for most of lwt users: ioam6, rpl, seg6,
seg6_local, ila_lwt, and lwt_bpf. It can happen in their input() and
output() handlers respectively, where either dst_input() or dst_output()
is called at the end. It can also happen in xmit() handlers.
Here is an example for rpl_input():
dump_stack_lvl+0x60/0x80
rpl_input+0x9d/0x320
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
[...]
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
lwtunnel_input+0x64/0xa0
ip6_sublist_rcv_finish+0x85/0x90
ip6_sublist_rcv+0x236/0x2f0
... until rpl_do_srh() fails, which means skb_cow_head() failed.
This series provides a fix at the core level of lwtunnel to catch such
loops when they're not caught by the respective lwtunnel users, and
handle the loop case in ioam6 which is one of the users. This series
also comes with a new selftest to detect some dst cache reference loops
in lwtunnel users.
====================
Link: https://patch.msgid.link/20250314120048.12569-1-justin.iurman@uliege.be
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As recently specified by commit 0ea09cbf83 ("docs: netdev: add a note
on selftest posting") in net-next, the selftest is therefore shipped in
this series. However, this selftest does not really test this series. It
needs this series to avoid crashing the kernel. What it really tests,
thanks to kmemleak, is what was fixed by the following commits:
- commit c71a192976 ("net: ipv6: fix dst refleaks in rpl, seg6 and
ioam6 lwtunnels")
- commit 92191dd107 ("net: ipv6: fix dst ref loops in rpl, seg6 and
ioam6 lwtunnels")
- commit c64a0727f9 ("net: ipv6: fix dst ref loop on input in seg6
lwt")
- commit 13e55fbaec ("net: ipv6: fix dst ref loop on input in rpl
lwt")
- commit 0e7633d7b9 ("net: ipv6: fix dst ref loop in ila lwtunnel")
- commit 5da15a9c11 ("net: ipv6: fix missing dst ref drop in ila
lwtunnel")
Signed-off-by: Justin Iurman <justin.iurman@uliege.be>
Link: https://patch.msgid.link/20250314120048.12569-4-justin.iurman@uliege.be
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
