[PATCH net-next v1 2/2] ipv6: make IPV6_TRANSPARENT setsockopt lockless
From: Xuanqiang Luo <hidden>
Date: 2026-09-11 10:24:47
Also in:
lkml
Subsystem:
networking [general], networking [ipv4/ipv6], the rest · Maintainers:
"David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, David Ahern, Ido Schimmel, Linus Torvalds
From: Xuanqiang Luo <redacted> IPV6_TRANSPARENT still takes the socket lock despite using inet_assign_bit() to update the flag shared with IP_TRANSPARENT. IPv6 TCP and UDP sockets can already update this flag without the lock through SOL_IP/IP_TRANSPARENT. Move IPV6_TRANSPARENT to the lockless setsockopt path. Signed-off-by: Xuanqiang Luo <redacted> --- net/ipv6/ipv6_sockglue.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-)
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index 4124084c3cd79..4b3536571c980 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c@@ -404,6 +404,15 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, /* Shared with IP_FREEBIND. */ inet_assign_bit(FREEBIND, sk, valbool); return 0; + case IPV6_TRANSPARENT: + if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) && + !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) + return -EPERM; + if (optlen < sizeof(int)) + return -EINVAL; + /* Shared with IP_TRANSPARENT. */ + inet_assign_bit(TRANSPARENT, sk, valbool); + return 0; case IPV6_UNICAST_HOPS: if (optlen < sizeof(int)) return -EINVAL;
@@ -672,19 +681,6 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, retv = 0; break; - case IPV6_TRANSPARENT: - if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) && - !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) { - retv = -EPERM; - break; - } - if (optlen < sizeof(int)) - goto e_inval; - /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */ - inet_assign_bit(TRANSPARENT, sk, valbool); - retv = 0; - break; - case IPV6_RECVORIGDSTADDR: if (optlen < sizeof(int)) goto e_inval;
--
2.43.0