linux-stable.git/drivers/net/vxlan.c, branch linux-5.4.y

vxlan: Do not treat dst cache initialization errors as fatal

2025-06-27T10:02:55+00:00

[ Upstream commit 20c76dadc783759fd3819d289c72be590660cc8b ]

FDB entries are allocated in an atomic context as they can be added from
the data path when learning is enabled.

After converting the FDB hash table to rhashtable, the insertion rate
will be much higher (*) which will entail a much higher rate of per-CPU
allocations via dst_cache_init().

When adding a large number of entries (e.g., 256k) in a batch, a small
percentage (< 0.02%) of these per-CPU allocations will fail [1]. This
does not happen with the current code since the insertion rate is low
enough to give the per-CPU allocator a chance to asynchronously create
new chunks of per-CPU memory.

Given that:

a. Only a small percentage of these per-CPU allocations fail.

b. The scenario where this happens might not be the most realistic one.

c. The driver can work correctly without dst caches. The dst_cache_*()
APIs first check that the dst cache was properly initialized.

d. The dst caches are not always used (e.g., 'tos inherit').

It seems reasonable to not treat these allocation failures as fatal.

Therefore, do not bail when dst_cache_init() fails and suppress warnings
by specifying '__GFP_NOWARN'.

[1] percpu: allocation failed, size=40 align=8 atomic=1, atomic alloc failed, no space left

(*) 97% reduction in average latency of vxlan_fdb_update() when adding
256k entries in a batch.

Reviewed-by: Petr Machata 
Signed-off-by: Ido Schimmel 
Link: https://patch.msgid.link/20250415121143.345227-14-idosch@nvidia.com
Reviewed-by: Nikolay Aleksandrov 
Signed-off-by: Paolo Abeni 
Signed-off-by: Sasha Levin

vxlan: Annotate FDB data races

2025-06-04T12:32:34+00:00

[ Upstream commit f6205f8215f12a96518ac9469ff76294ae7bd612 ]

The 'used' and 'updated' fields in the FDB entry structure can be
accessed concurrently by multiple threads, leading to reports such as
[1]. Can be reproduced using [2].

Suppress these reports by annotating these accesses using
READ_ONCE() / WRITE_ONCE().

[1]
BUG: KCSAN: data-race in vxlan_xmit / vxlan_xmit

write to 0xffff942604d263a8 of 8 bytes by task 286 on cpu 0:
 vxlan_xmit+0xb29/0x2380
 dev_hard_start_xmit+0x84/0x2f0
 __dev_queue_xmit+0x45a/0x1650
 packet_xmit+0x100/0x150
 packet_sendmsg+0x2114/0x2ac0
 __sys_sendto+0x318/0x330
 __x64_sys_sendto+0x76/0x90
 x64_sys_call+0x14e8/0x1c00
 do_syscall_64+0x9e/0x1a0
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

read to 0xffff942604d263a8 of 8 bytes by task 287 on cpu 2:
 vxlan_xmit+0xadf/0x2380
 dev_hard_start_xmit+0x84/0x2f0
 __dev_queue_xmit+0x45a/0x1650
 packet_xmit+0x100/0x150
 packet_sendmsg+0x2114/0x2ac0
 __sys_sendto+0x318/0x330
 __x64_sys_sendto+0x76/0x90
 x64_sys_call+0x14e8/0x1c00
 do_syscall_64+0x9e/0x1a0
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

value changed: 0x00000000fffbac6e -> 0x00000000fffbac6f

Reported by Kernel Concurrency Sanitizer on:
CPU: 2 UID: 0 PID: 287 Comm: mausezahn Not tainted 6.13.0-rc7-01544-gb4b270f11a02 #5
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-3.fc41 04/01/2014

[2]
 #!/bin/bash

 set +H
 echo whitelist > /sys/kernel/debug/kcsan
 echo !vxlan_xmit > /sys/kernel/debug/kcsan

 ip link add name vx0 up type vxlan id 10010 dstport 4789 local 192.0.2.1
 bridge fdb add 00:11:22:33:44:55 dev vx0 self static dst 198.51.100.1
 taskset -c 0 mausezahn vx0 -a own -b 00:11:22:33:44:55 -c 0 -q &
 taskset -c 2 mausezahn vx0 -a own -b 00:11:22:33:44:55 -c 0 -q &

Reviewed-by: Petr Machata 
Signed-off-by: Ido Schimmel 
Reviewed-by: Eric Dumazet 
Reviewed-by: Nikolay Aleksandrov 
Link: https://patch.msgid.link/20250204145549.1216254-2-idosch@nvidia.com
Signed-off-by: Jakub Kicinski 
Signed-off-by: Sasha Levin

vxlan: Fix regression when dropping packets due to invalid src addresses

2024-07-05T07:08:11+00:00

[ Upstream commit 1cd4bc987abb2823836cbb8f887026011ccddc8a ]

Commit f58f45c1e5b9 ("vxlan: drop packets from invalid src-address")
has recently been added to vxlan mainly in the context of source
address snooping/learning so that when it is enabled, an entry in the
FDB is not being created for an invalid address for the corresponding
tunnel endpoint.

Before commit f58f45c1e5b9 vxlan was similarly behaving as geneve in
that it passed through whichever macs were set in the L2 header. It
turns out that this change in behavior breaks setups, for example,
Cilium with netkit in L3 mode for Pods as well as tunnel mode has been
passing before the change in f58f45c1e5b9 for both vxlan and geneve.
After mentioned change it is only passing for geneve as in case of
vxlan packets are dropped due to vxlan_set_mac() returning false as
source and destination macs are zero which for E/W traffic via tunnel
is totally fine.

Fix it by only opting into the is_valid_ether_addr() check in
vxlan_set_mac() when in fact source address snooping/learning is
actually enabled in vxlan. This is done by moving the check into
vxlan_snoop(). With this change, the Cilium connectivity test suite
passes again for both tunnel flavors.

Fixes: f58f45c1e5b9 ("vxlan: drop packets from invalid src-address")
Signed-off-by: Daniel Borkmann 
Cc: David Bauer 
Cc: Ido Schimmel 
Cc: Nikolay Aleksandrov 
Cc: Martin KaFai Lau 
Reviewed-by: Ido Schimmel 
Reviewed-by: Nikolay Aleksandrov 
Reviewed-by: David Bauer 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

vxlan: Fix regression when dropping packets due to invalid src addresses

2024-06-16T11:28:49+00:00

commit 1cd4bc987abb2823836cbb8f887026011ccddc8a upstream.

Commit f58f45c1e5b9 ("vxlan: drop packets from invalid src-address")
has recently been added to vxlan mainly in the context of source
address snooping/learning so that when it is enabled, an entry in the
FDB is not being created for an invalid address for the corresponding
tunnel endpoint.

Before commit f58f45c1e5b9 vxlan was similarly behaving as geneve in
that it passed through whichever macs were set in the L2 header. It
turns out that this change in behavior breaks setups, for example,
Cilium with netkit in L3 mode for Pods as well as tunnel mode has been
passing before the change in f58f45c1e5b9 for both vxlan and geneve.
After mentioned change it is only passing for geneve as in case of
vxlan packets are dropped due to vxlan_set_mac() returning false as
source and destination macs are zero which for E/W traffic via tunnel
is totally fine.

Fix it by only opting into the is_valid_ether_addr() check in
vxlan_set_mac() when in fact source address snooping/learning is
actually enabled in vxlan. This is done by moving the check into
vxlan_snoop(). With this change, the Cilium connectivity test suite
passes again for both tunnel flavors.

Fixes: f58f45c1e5b9 ("vxlan: drop packets from invalid src-address")
Signed-off-by: Daniel Borkmann 
Cc: David Bauer 
Cc: Ido Schimmel 
Cc: Nikolay Aleksandrov 
Cc: Martin KaFai Lau 
Reviewed-by: Ido Schimmel 
Reviewed-by: Nikolay Aleksandrov 
Reviewed-by: David Bauer 
Signed-off-by: David S. Miller 
[ Backport note: vxlan snooping/learning not supported in 6.8 or older,
  so commit is simply a revert. ]
Signed-off-by: Daniel Borkmann 
Signed-off-by: Greg Kroah-Hartman

vxlan: drop packets from invalid src-address

2024-05-02T14:18:33+00:00

[ Upstream commit f58f45c1e5b92975e91754f5407250085a6ae7cf ]

The VXLAN driver currently does not check if the inner layer2
source-address is valid.

In case source-address snooping/learning is enabled, a entry in the FDB
for the invalid address is created with the layer3 address of the tunnel
endpoint.

If the frame happens to have a non-unicast address set, all this
non-unicast traffic is subsequently not flooded to the tunnel network
but sent to the learnt host in the FDB. To make matters worse, this FDB
entry does not expire.

Apply the same filtering for packets as it is done for bridges. This not
only drops these invalid packets but avoids them from being learnt into
the FDB.

Fixes: d342894c5d2f ("vxlan: virtual extensible lan")
Suggested-by: Ido Schimmel 
Signed-off-by: David Bauer 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

vxlan: generalize vxlan_parse_gpe_hdr and remove unused args

2023-09-23T08:59:38+00:00

[ Upstream commit 17a0a64448b568442a101de09575f81ffdc45d15 ]

The vxlan_parse_gpe_hdr function extracts the next protocol value from
the GPE header and marks GPE bits as parsed.

In order to be used in the next patch, split the function into protocol
extraction and bit marking. The bit marking is meaningful only in
vxlan_rcv; move it directly there.

Rename the function to vxlan_parse_gpe_proto to reflect what it now
does. Remove unused arguments skb and vxflags. Move the function earlier
in the file to allow it to be called from more places in the next patch.

Signed-off-by: Jiri Benc 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

vxlan: calculate correct header length for GPE

2023-08-11T09:53:47+00:00

[ Upstream commit 94d166c5318c6edd1e079df8552233443e909c33 ]

VXLAN-GPE does not add an extra inner Ethernet header. Take that into
account when calculating header length.

This causes problems in skb_tunnel_check_pmtu, where incorrect PMTU is
cached.

In the collect_md mode (which is the only mode that VXLAN-GPE
supports), there's no magic auto-setting of the tunnel interface MTU.
It can't be, since the destination and thus the underlying interface
may be different for each packet.

So, the administrator is responsible for setting the correct tunnel
interface MTU. Apparently, the administrators are capable enough to
calculate that the maximum MTU for VXLAN-GPE is (their_lower_MTU - 36).
They set the tunnel interface MTU to 1464. If you run a TCP stream over
such interface, it's then segmented according to the MTU 1464, i.e.
producing 1514 bytes frames. Which is okay, this still fits the lower
MTU.

However, skb_tunnel_check_pmtu (called from vxlan_xmit_one) uses 50 as
the header size and thus incorrectly calculates the frame size to be
1528. This leads to ICMP too big message being generated (locally),
PMTU of 1450 to be cached and the TCP stream to be resegmented.

The fix is to use the correct actual header size, especially for
skb_tunnel_check_pmtu calculation.

Fixes: e1e5314de08ba ("vxlan: implement GPE")
Signed-off-by: Jiri Benc 
Reviewed-by: Simon Horman 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

vxlan: fix error return code in vxlan_fdb_append

2022-04-27T11:50:47+00:00

[ Upstream commit 7cea5560bf656b84f9ed01c0cc829d4eecd0640b ]

When kmalloc and dst_cache_init failed,
should return ENOMEM rather than ENOBUFS.

Signed-off-by: Hongbin Wang 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

vxlan: add missing rcu_read_lock() in neigh_reduce()

2021-07-14T14:53:31+00:00

[ Upstream commit 85e8b032d6ebb0f698a34dd22c2f13443d905888 ]

syzbot complained in neigh_reduce(), because rcu_read_lock_bh()
is treated differently than rcu_read_lock()

WARNING: suspicious RCU usage
5.13.0-rc6-syzkaller #0 Not tainted
-----------------------------
include/net/addrconf.h:313 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

rcu_scheduler_active = 2, debug_locks = 1
3 locks held by kworker/0:0/5:
 #0: ffff888011064d38 ((wq_completion)events){+.+.}-{0:0}, at: arch_atomic64_set arch/x86/include/asm/atomic64_64.h:34 [inline]
 #0: ffff888011064d38 ((wq_completion)events){+.+.}-{0:0}, at: atomic64_set include/asm-generic/atomic-instrumented.h:856 [inline]
 #0: ffff888011064d38 ((wq_completion)events){+.+.}-{0:0}, at: atomic_long_set include/asm-generic/atomic-long.h:41 [inline]
 #0: ffff888011064d38 ((wq_completion)events){+.+.}-{0:0}, at: set_work_data kernel/workqueue.c:617 [inline]
 #0: ffff888011064d38 ((wq_completion)events){+.+.}-{0:0}, at: set_work_pool_and_clear_pending kernel/workqueue.c:644 [inline]
 #0: ffff888011064d38 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work+0x871/0x1600 kernel/workqueue.c:2247
 #1: ffffc90000ca7da8 ((work_completion)(&port->wq)){+.+.}-{0:0}, at: process_one_work+0x8a5/0x1600 kernel/workqueue.c:2251
 #2: ffffffff8bf795c0 (rcu_read_lock_bh){....}-{1:2}, at: __dev_queue_xmit+0x1da/0x3130 net/core/dev.c:4180

stack backtrace:
CPU: 0 PID: 5 Comm: kworker/0:0 Not tainted 5.13.0-rc6-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events ipvlan_process_multicast
Call Trace:
 __dump_stack lib/dump_stack.c:79 [inline]
 dump_stack+0x141/0x1d7 lib/dump_stack.c:120
 __in6_dev_get include/net/addrconf.h:313 [inline]
 __in6_dev_get include/net/addrconf.h:311 [inline]
 neigh_reduce drivers/net/vxlan.c:2167 [inline]
 vxlan_xmit+0x34d5/0x4c30 drivers/net/vxlan.c:2919
 __netdev_start_xmit include/linux/netdevice.h:4944 [inline]
 netdev_start_xmit include/linux/netdevice.h:4958 [inline]
 xmit_one net/core/dev.c:3654 [inline]
 dev_hard_start_xmit+0x1eb/0x920 net/core/dev.c:3670
 __dev_queue_xmit+0x2133/0x3130 net/core/dev.c:4246
 ipvlan_process_multicast+0xa99/0xd70 drivers/net/ipvlan/ipvlan_core.c:287
 process_one_work+0x98d/0x1600 kernel/workqueue.c:2276
 worker_thread+0x64c/0x1120 kernel/workqueue.c:2422
 kthread+0x3b1/0x4a0 kernel/kthread.c:313
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294

Fixes: f564f45c4518 ("vxlan: add ipv6 proxy support")
Signed-off-by: Eric Dumazet 
Reported-by: syzbot 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

vxlan: move debug check after netdev unregister

2021-03-04T09:26:38+00:00

[ Upstream commit 92584ddf550ae72d492858c19d1f9025e07a9350 ]

The debug check must be done after unregister_netdevice_many() call --
the hlist_del_rcu() for this is done inside .ndo_stop.

This is the same with commit 0fda7600c2e1 ("geneve: move debug check after
netdev unregister")

Test commands:
    ip netns del A
    ip netns add A
    ip netns add B

    ip netns exec B ip link add vxlan0 type vxlan vni 100 local 10.0.0.1 \
	    remote 10.0.0.2 dstport 4789 srcport 4789 4789
    ip netns exec B ip link set vxlan0 netns A
    ip netns exec A ip link set vxlan0 up
    ip netns del B

Splat looks like:
[   73.176249][    T7] ------------[ cut here ]------------
[   73.178662][    T7] WARNING: CPU: 4 PID: 7 at drivers/net/vxlan.c:4743 vxlan_exit_batch_net+0x52e/0x720 [vxlan]
[   73.182597][    T7] Modules linked in: vxlan openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 mlx5_core nfp mlxfw ixgbevf tls sch_fq_codel nf_tables nfnetlink ip_tables x_tables unix
[   73.190113][    T7] CPU: 4 PID: 7 Comm: kworker/u16:0 Not tainted 5.11.0-rc7+ #838
[   73.193037][    T7] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014
[   73.196986][    T7] Workqueue: netns cleanup_net
[   73.198946][    T7] RIP: 0010:vxlan_exit_batch_net+0x52e/0x720 [vxlan]
[   73.201509][    T7] Code: 00 01 00 00 0f 84 39 fd ff ff 48 89 ca 48 c1 ea 03 80 3c 1a 00 0f 85 a6 00 00 00 89 c2 48 83 c2 02 49 8b 14 d4 48 85 d2 74 ce <0f> 0b eb ca e8 b9 51 db dd 84 c0 0f 85 4a fe ff ff 48 c7 c2 80 bc
[   73.208813][    T7] RSP: 0018:ffff888100907c10 EFLAGS: 00010286
[   73.211027][    T7] RAX: 000000000000003c RBX: dffffc0000000000 RCX: ffff88800ec411f0
[   73.213702][    T7] RDX: ffff88800a278000 RSI: ffff88800fc78c70 RDI: ffff88800fc78070
[   73.216169][    T7] RBP: ffff88800b5cbdc0 R08: fffffbfff424de61 R09: fffffbfff424de61
[   73.218463][    T7] R10: ffffffffa126f307 R11: fffffbfff424de60 R12: ffff88800ec41000
[   73.220794][    T7] R13: ffff888100907d08 R14: ffff888100907c50 R15: ffff88800fc78c40
[   73.223337][    T7] FS:  0000000000000000(0000) GS:ffff888114800000(0000) knlGS:0000000000000000
[   73.225814][    T7] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   73.227616][    T7] CR2: 0000562b5cb4f4d0 CR3: 0000000105fbe001 CR4: 00000000003706e0
[   73.229700][    T7] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   73.231820][    T7] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   73.233844][    T7] Call Trace:
[   73.234698][    T7]  ? vxlan_err_lookup+0x3c0/0x3c0 [vxlan]
[   73.235962][    T7]  ? ops_exit_list.isra.11+0x93/0x140
[   73.237134][    T7]  cleanup_net+0x45e/0x8a0
[ ... ]

Fixes: 57b61127ab7d ("vxlan: speedup vxlan tunnels dismantle")
Signed-off-by: Taehee Yoo 
Link: https://lore.kernel.org/r/20210221154552.11749-1-ap420073@gmail.com
Signed-off-by: Jakub Kicinski 
Signed-off-by: Sasha Levin