OSDN Git Service

lwt: Fix return values of BPF xmit ops
authorYan Zhai <yan@cloudflare.com>
Fri, 18 Aug 2023 02:58:11 +0000 (19:58 -0700)
committerDaniel Borkmann <daniel@iogearbox.net>
Fri, 18 Aug 2023 14:05:26 +0000 (16:05 +0200)
commit29b22badb7a84b783e3a4fffca16f7768fb31205
tree9129510d137e3e957f0d405d58e18723ce9a22a8
parent5f6395fd06806a009b93436612768fe3777a154f
lwt: Fix return values of BPF xmit ops

BPF encap ops can return different types of positive values, such like
NET_RX_DROP, NET_XMIT_CN, NETDEV_TX_BUSY, and so on, from function
skb_do_redirect and bpf_lwt_xmit_reroute. At the xmit hook, such return
values would be treated implicitly as LWTUNNEL_XMIT_CONTINUE in
ip(6)_finish_output2. When this happens, skbs that have been freed would
continue to the neighbor subsystem, causing use-after-free bug and
kernel crashes.

To fix the incorrect behavior, skb_do_redirect return values can be
simply discarded, the same as tc-egress behavior. On the other hand,
bpf_lwt_xmit_reroute returns useful errors to local senders, e.g. PMTU
information. Thus convert its return values to avoid the conflict with
LWTUNNEL_XMIT_CONTINUE.

Fixes: 3a0af8fd61f9 ("bpf: BPF for lightweight tunnel infrastructure")
Reported-by: Jordan Griege <jgriege@cloudflare.com>
Suggested-by: Martin KaFai Lau <martin.lau@linux.dev>
Suggested-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Yan Zhai <yan@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/0d2b878186cfe215fec6b45769c1cd0591d3628d.1692326837.git.yan@cloudflare.com
net/core/lwt_bpf.c