]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commit
skmsg: Avoid lock_sock() in sk_psock_backlog()
authorCong Wang <cong.wang@bytedance.com>
Wed, 31 Mar 2021 02:32:25 +0000 (19:32 -0700)
committerAlexei Starovoitov <ast@kernel.org>
Thu, 1 Apr 2021 17:56:13 +0000 (10:56 -0700)
commit799aa7f98d53e0f541fa6b4dc9aa47b4ff2178e3
tree38226a66a54ab0b7db060abb3e08cadebaa6500c
parent0739cd28f2645e814586c7536ba5da9825cb8029
skmsg: Avoid lock_sock() in sk_psock_backlog()

We do not have to lock the sock to avoid losing sk_socket,
instead we can purge all the ingress queues when we close
the socket. Sending or receiving packets after orphaning
socket makes no sense.

We do purge these queues when psock refcnt reaches zero but
here we want to purge them explicitly in sock_map_close().
There are also some nasty race conditions on testing bit
SK_PSOCK_TX_ENABLED and queuing/canceling the psock work,
we can expand psock->ingress_lock a bit to protect them too.

As noticed by John, we still have to lock the psock->work,
because the same work item could be running concurrently on
different CPU's.

Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210331023237.41094-5-xiyou.wangcong@gmail.com
include/linux/skmsg.h
net/core/skmsg.c
net/core/sock_map.c