[PATCH net v2] tcp: drop secpath at the same time as we currently drop dst

[Paul Moore]

On Mon, Feb 17, 2025 at 6:14 PM Sabrina Dubroca <sd at queasysnail.net> wrote:
> 2025-02-17, 17:35:32 -0500, Paul Moore wrote:
> > On Mon, Feb 17, 2025 at 5:23 AM Sabrina Dubroca <sd at queasysnail.net> wrote:
> > >
> > > Xiumei reported hitting the WARN in xfrm6_tunnel_net_exit while
> > > running tests that boil down to:
> > >  - create a pair of netns
> > >  - run a basic TCP test over ipcomp6
> > >  - delete the pair of netns
> > >
> > > The xfrm_state found on spi_byaddr was not deleted at the time we
> > > delete the netns, because we still have a reference on it. This
> > > lingering reference comes from a secpath (which holds a ref on the
> > > xfrm_state), which is still attached to an skb. This skb is not
> > > leaked, it ends up on sk_receive_queue and then gets defer-free'd by
> > > skb_attempt_defer_free.
> > >
> > > The problem happens when we defer freeing an skb (push it on one CPU's
> > > defer_list), and don't flush that list before the netns is deleted. In
> > > that case, we still have a reference on the xfrm_state that we don't
> > > expect at this point.
> > >
> > > We already drop the skb's dst in the TCP receive path when it's no
> > > longer needed, so let's also drop the secpath. At this point,
> > > tcp_filter has already called into the LSM hooks that may require the
> > > secpath, so it should not be needed anymore.
> >
> > I don't recall seeing any follow up in the v1 patchset regarding
> > IP_CMSG_PASSSEC/security_socket_getpeersec_dgram(), can you confirm
> > that the secpath is preserved for that code path?
> >
> > https://lore.kernel.org/linux-security-module/CAHC9VhQZ+k1J0UidJ-bgdBGBuVX9M18tQ+a+fuqXQM_L-PFvzA@mail.gmail.com
>
> Sorry, I thought we'd addressed this in the v1 discussion with Eric.
>
> IP_CMSG_PASSSEC is not blocked for TCP sockets, but it will only
> process skbs that came from the error queue (ip_recv_error ->
> ip_cmsg_recv -> ip_cmsg_recv_offset -> ip_cmsg_recv_security ->
> security_socket_getpeersec_dgram), which don't go through those code
> paths at all. So AFAICT IP_CMSG_PASSSEC for TCP isn't affected by
> dropping the secpath early.

Great, thanks for clearing that up.

-- 
paul-moore.com