OSDN Git Service

udp: fix race between close() and udp_abort()
authorPaolo Abeni <pabeni@redhat.com>
Wed, 9 Jun 2021 09:49:01 +0000 (11:49 +0200)
committerDavid S. Miller <davem@davemloft.net>
Wed, 9 Jun 2021 21:08:41 +0000 (14:08 -0700)
commita8b897c7bcd47f4147d066e22cc01d1026d7640e
treef8423df1a5a624483c4cbd8c53fc7e0c05708fab
parentdcd01eeac14486b56a790f5cce9b823440ba5b34
udp: fix race between close() and udp_abort()

Kaustubh reported and diagnosed a panic in udp_lib_lookup().
The root cause is udp_abort() racing with close(). Both
racing functions acquire the socket lock, but udp{v6}_destroy_sock()
release it before performing destructive actions.

We can't easily extend the socket lock scope to avoid the race,
instead use the SOCK_DEAD flag to prevent udp_abort from doing
any action when the critical race happens.

Diagnosed-and-tested-by: Kaustubh Pandey <kapandey@codeaurora.org>
Fixes: 5d77dca82839 ("net: diag: support SOCK_DESTROY for UDP sockets")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/udp.c
net/ipv6/udp.c