net: Only allow init netns to set default tcp cong to a restricted algo
authorJonathon Reinhart <jonathon.reinhart@gmail.com>
Sat, 1 May 2021 08:28:22 +0000 (04:28 -0400)
committerDavid S. Miller <davem@davemloft.net>
Tue, 4 May 2021 18:58:28 +0000 (11:58 -0700)
commit8d432592f30fcc34ef5a10aac4887b4897884493
treeafec67dce49a4a74ded0a415c55281bcfcb44e4b
parent1682d8df20aa505f6ab12c76e934b26ede39c529
net: Only allow init netns to set default tcp cong to a restricted algo

tcp_set_default_congestion_control() is netns-safe in that it writes
to &net->ipv4.tcp_congestion_control, but it also sets
ca->flags |= TCP_CONG_NON_RESTRICTED which is not namespaced.
This has the unintended side-effect of changing the global
net.ipv4.tcp_allowed_congestion_control sysctl, despite the fact that it
is read-only: 97684f0970f6 ("net: Make tcp_allowed_congestion_control
readonly in non-init netns")

Resolve this netns "leak" by only allowing the init netns to set the
default algorithm to one that is restricted. This restriction could be
removed if tcp_allowed_congestion_control were namespace-ified in the
future.

This bug was uncovered with
https://github.com/JonathonReinhart/linux-netns-sysctl-verify

Fixes: 6670e1524477 ("tcp: Namespace-ify sysctl_tcp_default_congestion_control")
Signed-off-by: Jonathon Reinhart <jonathon.reinhart@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_cong.c