xfrm: always flush state and policy upon NETDEV_UNREGISTER event

syzbot is reporting that "struct xfrm_state" refcount is leaking.

  unregister_netdevice: waiting for netdevsim0 to become free. Usage count = 2
  ref_tracker: netdev@ffff888052f24618 has 1/1 users at
       __netdev_tracker_alloc include/linux/netdevice.h:4400 [inline]
       netdev_tracker_alloc include/linux/netdevice.h:4412 [inline]
       xfrm_dev_state_add+0x3a5/0x1080 net/xfrm/xfrm_device.c:316
       xfrm_state_construct net/xfrm/xfrm_user.c:986 [inline]
       xfrm_add_sa+0x34ff/0x5fa0 net/xfrm/xfrm_user.c:1022
       xfrm_user_rcv_msg+0x58e/0xc00 net/xfrm/xfrm_user.c:3507
       netlink_rcv_skb+0x158/0x420 net/netlink/af_netlink.c:2550
       xfrm_netlink_rcv+0x71/0x90 net/xfrm/xfrm_user.c:3529
       netlink_unicast_kernel net/netlink/af_netlink.c:1318 [inline]
       netlink_unicast+0x5aa/0x870 net/netlink/af_netlink.c:1344
       netlink_sendmsg+0x8c8/0xdd0 net/netlink/af_netlink.c:1894
       sock_sendmsg_nosec net/socket.c:727 [inline]
       __sock_sendmsg net/socket.c:742 [inline]
       ____sys_sendmsg+0xa5d/0xc30 net/socket.c:2592
       ___sys_sendmsg+0x134/0x1d0 net/socket.c:2646
       __sys_sendmsg+0x16d/0x220 net/socket.c:2678
       do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
       do_syscall_64+0xcd/0xf80 arch/x86/entry/syscall_64.c:94
       entry_SYSCALL_64_after_hwframe+0x77/0x7f

This is because commit d77e38e612a0 ("xfrm: Add an IPsec hardware
offloading API") implemented xfrm_dev_unregister() as no-op despite
xfrm_dev_state_add() from xfrm_state_construct() acquires a reference
to "struct net_device".
I guess that that commit expected that NETDEV_DOWN event is fired before
NETDEV_UNREGISTER event fires, and also assumed that xfrm_dev_state_add()
is called only if (dev->features & NETIF_F_HW_ESP) != 0.

Sabrina Dubroca identified steps to reproduce the same symptoms as below.

  echo 0 > /sys/bus/netdevsim/new_device
  dev=$(ls -1 /sys/bus/netdevsim/devices/netdevsim0/net/)
  ip xfrm state add src 192.168.13.1 dst 192.168.13.2 proto esp \
     spi 0x1000 mode tunnel aead 'rfc4106(gcm(aes))' $key 128   \
     offload crypto dev $dev dir out
  ethtool -K $dev esp-hw-offload off
  echo 0 > /sys/bus/netdevsim/del_device

Like these steps indicate, the NETIF_F_HW_ESP bit can be cleared after
xfrm_dev_state_add() acquired a reference to "struct net_device".
Also, xfrm_dev_state_add() does not check for the NETIF_F_HW_ESP bit
when acquiring a reference to "struct net_device".

Commit 03891f820c21 ("xfrm: handle NETDEV_UNREGISTER for xfrm device")
re-introduced the NETDEV_UNREGISTER event to xfrm_dev_event(), but that
commit for unknown reason chose to share xfrm_dev_down() between the
NETDEV_DOWN event and the NETDEV_UNREGISTER event.
I guess that that commit missed the behavior in the previous paragraph.

Therefore, we need to re-introduce xfrm_dev_unregister() in order to
release the reference to "struct net_device" by unconditionally flushing
state and policy.

Reported-by: syzbot+881d65229ca4f9ae8c84@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=881d65229ca4f9ae8c84
Fixes: d77e38e612a0 ("xfrm: Add an IPsec hardware offloading API")
Cc: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>