]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
net: reenable NETIF_F_IPV6_CSUM offload for BIG TCP packets
authorWillem de Bruijn <willemb@google.com>
Wed, 1 Jan 2025 16:47:40 +0000 (11:47 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 9 Jan 2025 12:30:02 +0000 (13:30 +0100)
commitac9cfef69565021c9e1022a493a9c40b03e2caf9
treec403ba0180610ee2f9a9f64c38e063d69775c404
parent32e1e748a85bd52b20b3857d80fd166d22fa455a
net: reenable NETIF_F_IPV6_CSUM offload for BIG TCP packets

[ Upstream commit 68e068cabd2c6c533ef934c2e5151609cf6ecc6d ]

The blamed commit disabled hardware offoad of IPv6 packets with
extension headers on devices that advertise NETIF_F_IPV6_CSUM,
based on the definition of that feature in skbuff.h:

 *   * - %NETIF_F_IPV6_CSUM
 *     - Driver (device) is only able to checksum plain
 *       TCP or UDP packets over IPv6. These are specifically
 *       unencapsulated packets of the form IPv6|TCP or
 *       IPv6|UDP where the Next Header field in the IPv6
 *       header is either TCP or UDP. IPv6 extension headers
 *       are not supported with this feature. This feature
 *       cannot be set in features for a device with
 *       NETIF_F_HW_CSUM also set. This feature is being
 *       DEPRECATED (see below).

The change causes skb_warn_bad_offload to fire for BIG TCP
packets.

[  496.310233] WARNING: CPU: 13 PID: 23472 at net/core/dev.c:3129 skb_warn_bad_offload+0xc4/0xe0

[  496.310297]  ? skb_warn_bad_offload+0xc4/0xe0
[  496.310300]  skb_checksum_help+0x129/0x1f0
[  496.310303]  skb_csum_hwoffload_help+0x150/0x1b0
[  496.310306]  validate_xmit_skb+0x159/0x270
[  496.310309]  validate_xmit_skb_list+0x41/0x70
[  496.310312]  sch_direct_xmit+0x5c/0x250
[  496.310317]  __qdisc_run+0x388/0x620

BIG TCP introduced an IPV6_TLV_JUMBO IPv6 extension header to
communicate packet length, as this is an IPv6 jumbogram. But, the
feature is only enabled on devices that support BIG TCP TSO. The
header is only present for PF_PACKET taps like tcpdump, and not
transmitted by physical devices.

For this specific case of extension headers that are not
transmitted, return to the situation before the blamed commit
and support hardware offload.

ipv6_has_hopopt_jumbo() tests not only whether this header is present,
but also that it is the only extension header before a terminal (L4)
header.

Fixes: 04c20a9356f2 ("net: skip offload for NETIF_F_IPV6_CSUM if ipv6 header contains extension")
Reported-by: syzbot <syzkaller@googlegroups.com>
Reported-by: Eric Dumazet <edumazet@google.com>
Closes: https://lore.kernel.org/netdev/CANn89iK1hdC3Nt8KPhOtTF8vCPc1AHDCtse_BTNki1pWxAByTQ@mail.gmail.com/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250101164909.1331680-1-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
net/core/dev.c