[Full-disclosure] tcpdump: off-by-one heap overflow in 802.11 printer

[Moritz Jodeit]

There's an off-by-one heap-overflow in the ieee802.11 printer,
which can be triggered by a maliciously crafted 802.11 frame.
The link type must have been explicitly specified for this to
work.

The function parse_elements() in print-802_11.c checks the
length pbody->tim.length from the frame for too small values in
line 265, but then uses the wrong variable in the following
range check in line 267. Since pbody->tim.length is defined as
a u_int8_t it can hold a maximum value of 255, which in turn
would copy 252 bytes into pbody->tim.bitmap, which is only
251 bytes of size.

253	case E_TIM:
254		/* Present, possibly truncated */
255		pbody->tim_status = TRUNCATED;
256		if (!TTEST2(*(p + offset), 2))
257			return;
258		memcpy(&pbody->tim, p + offset, 2);
259		offset += 2;
260		if (!TTEST2(*(p + offset), 3))
261			return;
262		memcpy(&pbody->tim.count, p + offset, 3);
263		offset += 3;
264
265		if (pbody->tim.length <= 3)
266			break;
267		if (pbody->rates.length > sizeof pbody->tim.bitmap)
268			return;
269		if (!TTEST2(*(p + offset), pbody->tim.length - 3))
270			return;
271		memcpy(pbody->tim.bitmap, p + (pbody->tim.length - 3),
272		    (pbody->tim.length - 3))

The current tcpdump release 3.9.5 is still vulnerable.
This got fixed [1] in CVS Head and in the tcpdump_3_9 branch.

[1] http://cvs.tcpdump.org/cgi-bin/cvsweb/tcpdump/print-802_11.c?r1=1.42&r2=1.43

Best,
Moritz Jodeit