bpf: sockmap, add sock close() hook to remove socks
authorJohn Fastabend <john.fastabend@gmail.com>
Mon, 5 Feb 2018 18:17:49 +0000 (10:17 -0800)
committerDaniel Borkmann <daniel@iogearbox.net>
Tue, 6 Feb 2018 10:39:32 +0000 (11:39 +0100)
commit1aa12bdf1bfb95db7e75bfecf0e39a65f4e8fbf8
treeea5a01abe7ce73c70b2ca5433463b4cd45f4e8f7
parentb11a632c442eef34a0afeba61fab923241f317e9
bpf: sockmap, add sock close() hook to remove socks

The selftests test_maps program was leaving dangling BPF sockmap
programs around because not all psock elements were removed from
the map. The elements in turn hold a reference on the BPF program
they are attached to causing BPF programs to stay open even after
test_maps has completed.

The original intent was that sk_state_change() would be called
when TCP socks went through TCP_CLOSE state. However, because
socks may be in SOCK_DEAD state or the sock may be a listening
socket the event is not always triggered.

To resolve this use the ULP infrastructure and register our own
proto close() handler. This fixes the above case.

Fixes: 174a79ff9515 ("bpf: sockmap with sk redirect support")
Reported-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
include/net/tcp.h
kernel/bpf/sockmap.c