ofproto: fix OF1.0 packet_in parser

OF1.0 spec says OFP_ASSERT(sizeof(struct ofp_packet_in) == 20).
It's quite bogus as it assumes a specific class of C implementations.
(well, if it was C.  it's unclear from the spec itself.)
We just use the real size of the structure as this is not C.  This
agrees with on-wire messages OpenFlow Reference Release and Open vSwitch
produce.

This should fix a crash Chen Chen reported on ryu-devel recently.

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This commit is contained in:
YAMAMOTO Takashi 2013-09-26 13:08:07 +09:00 committed by FUJITA Tomonori
parent 54989c7862
commit a53bc8052d
2 changed files with 9 additions and 4 deletions

View File

@ -164,9 +164,14 @@ assert calcsize(OFP_PORT_MOD_PACK_STR) + OFP_HEADER_SIZE == OFP_PORT_MOD_SIZE
OFPR_NO_MATCH = 0 # No matching flow.
OFPR_ACTION = 1 # Action explicitly output to controller.
OFP_PACKET_IN_PACK_STR = '!IHHBx2x' # the last 2x is for ofp_packet_in::data
OFP_PACKET_IN_SIZE = 20
OFP_PACKET_IN_DATA_OFFSET = 18
# OF1.0 spec says OFP_ASSERT(sizeof(struct ofp_packet_in) == 20).
# It's quite bogus as it assumes a specific class of C implementations.
# (well, if it was C. it's unclear from the spec itself.)
# We just use the real size of the structure as this is not C. This
# agrees with on-wire messages OpenFlow Reference Release and Open vSwitch
# produce.
OFP_PACKET_IN_PACK_STR = '!IHHBx'
OFP_PACKET_IN_SIZE = 18
assert calcsize(OFP_PACKET_IN_PACK_STR) + OFP_HEADER_SIZE == OFP_PACKET_IN_SIZE
# enum ofp_action_type

View File

@ -1730,7 +1730,7 @@ class OFPPacketIn(MsgBase):
msg.reason) = struct.unpack_from(
ofproto_v1_0.OFP_PACKET_IN_PACK_STR,
msg.buf, ofproto_v1_0.OFP_HEADER_SIZE)
msg.data = msg.buf[ofproto_v1_0.OFP_PACKET_IN_DATA_OFFSET:]
msg.data = msg.buf[ofproto_v1_0.OFP_PACKET_IN_SIZE:]
if msg.total_len < len(msg.data):
# discard padding for 8-byte alignment of OFP packet
msg.data = msg.data[:msg.total_len]