1.9 KiB
CVE-2025-21864
Description
In the Linux kernel, the following vulnerability has been resolved:tcp: drop secpath at the same time as we currently drop dstXiumei reported hitting the WARN in xfrm6_tunnel_net_exit whilerunning tests that boil down to: - create a pair of netns - run a basic TCP test over ipcomp6 - delete the pair of netnsThe xfrm_state found on spi_byaddr was not deleted at the time wedelete the netns, because we still have a reference on it. Thislingering reference comes from a secpath (which holds a ref on thexfrm_state), which is still attached to an skb. This skb is notleaked, it ends up on sk_receive_queue and then gets defer-free'd byskb_attempt_defer_free.The problem happens when we defer freeing an skb (push it on one CPU'sdefer_list), and don't flush that list before the netns is deleted. Inthat case, we still have a reference on the xfrm_state that we don'texpect at this point.We already drop the skb's dst in the TCP receive path when it's nolonger needed, so let's also drop the secpath. At this point,tcp_filter has already called into the LSM hooks that may require thesecpath, so it should not be needed anymore. However, in some of thoseplaces, the MPTCP extension has just been attached to the skb, so wecannot simply drop all extensions.
POC
Reference
No PoCs from references.