]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
net: openvswitch: fix middle attribute validation in push_nsh() action
authorIlya Maximets <i.maximets@ovn.org>
Thu, 4 Dec 2025 10:53:32 +0000 (11:53 +0100)
committerJakub Kicinski <kuba@kernel.org>
Wed, 10 Dec 2025 08:46:55 +0000 (00:46 -0800)
commit5ace7ef87f059d68b5f50837ef3e8a1a4870c36e
tree2806a99f8f2c39153078f51e90c15f3da46696de
parent186468c67fc687650b7fb713d8c627d5c8566886
net: openvswitch: fix middle attribute validation in push_nsh() action

The push_nsh() action structure looks like this:

 OVS_ACTION_ATTR_PUSH_NSH(OVS_KEY_ATTR_NSH(OVS_NSH_KEY_ATTR_BASE,...))

The outermost OVS_ACTION_ATTR_PUSH_NSH attribute is OK'ed by the
nla_for_each_nested() inside __ovs_nla_copy_actions().  The innermost
OVS_NSH_KEY_ATTR_BASE/MD1/MD2 are OK'ed by the nla_for_each_nested()
inside nsh_key_put_from_nlattr().  But nothing checks if the attribute
in the middle is OK.  We don't even check that this attribute is the
OVS_KEY_ATTR_NSH.  We just do a double unwrap with a pair of nla_data()
calls - first time directly while calling validate_push_nsh() and the
second time as part of the nla_for_each_nested() macro, which isn't
safe, potentially causing invalid memory access if the size of this
attribute is incorrect.  The failure may not be noticed during
validation due to larger netlink buffer, but cause trouble later during
action execution where the buffer is allocated exactly to the size:

 BUG: KASAN: slab-out-of-bounds in nsh_hdr_from_nlattr+0x1dd/0x6a0 [openvswitch]
 Read of size 184 at addr ffff88816459a634 by task a.out/22624

 CPU: 8 UID: 0 PID: 22624 6.18.0-rc7+ #115 PREEMPT(voluntary)
 Call Trace:
  <TASK>
  dump_stack_lvl+0x51/0x70
  print_address_description.constprop.0+0x2c/0x390
  kasan_report+0xdd/0x110
  kasan_check_range+0x35/0x1b0
  __asan_memcpy+0x20/0x60
  nsh_hdr_from_nlattr+0x1dd/0x6a0 [openvswitch]
  push_nsh+0x82/0x120 [openvswitch]
  do_execute_actions+0x1405/0x2840 [openvswitch]
  ovs_execute_actions+0xd5/0x3b0 [openvswitch]
  ovs_packet_cmd_execute+0x949/0xdb0 [openvswitch]
  genl_family_rcv_msg_doit+0x1d6/0x2b0
  genl_family_rcv_msg+0x336/0x580
  genl_rcv_msg+0x9f/0x130
  netlink_rcv_skb+0x11f/0x370
  genl_rcv+0x24/0x40
  netlink_unicast+0x73e/0xaa0
  netlink_sendmsg+0x744/0xbf0
  __sys_sendto+0x3d6/0x450
  do_syscall_64+0x79/0x2c0
  entry_SYSCALL_64_after_hwframe+0x76/0x7e
  </TASK>

Let's add some checks that the attribute is properly sized and it's
the only one attribute inside the action.  Technically, there is no
real reason for OVS_KEY_ATTR_NSH to be there, as we know that we're
pushing an NSH header already, it just creates extra nesting, but
that's how uAPI works today.  So, keeping as it is.

Fixes: b2d0f5d5dc53 ("openvswitch: enable NSH support")
Reported-by: Junvy Yang <zhuque@tencent.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Acked-by: Eelco Chaudron echaudro@redhat.com
Reviewed-by: Aaron Conole <aconole@redhat.com>
Link: https://patch.msgid.link/20251204105334.900379-1-i.maximets@ovn.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/openvswitch/flow_netlink.c