WPA-PSK problem on ARM

[Willy Sagefalk]

Hi 

I have seen the same issue when testing WPA on a platform with Hermes 2
+ ARM. I found one alignement issue in driver_wext.c. The following fix
seems to solve the problem.  

Brgds Willy

static void wpa_driver_wext_event_wireless(void *ctx, char *data, int
len)
{
	struct iw_event *iwe, iwe2;
	char *pos, *end, *custom, *buf;

	pos = data;
	end = data + len;

	while (pos + IW_EV_LCP_LEN <= end) {
	        /*  Fix */
                memcpy ((void *)&iwe2,(void*)pos,sizeof(struct
iw_event));
        	iwe = &iwe2;
		// iwe = (struct iw_event *) pos;


int wpa_driver_wext_get_scan_results(const char *ifname,
				     struct wpa_scan_result *results,
				     size_t max_size)
{
	struct iwreq iwr;
	int s, ap_num = 0, first, maxrate;
	u8 res_buf[IW_SCAN_MAX_DATA];
	struct iw_event *iwe,iwe2;
	char *pos, *end, *custom;
	size_t len, clen;

	memset(results, 0, max_size * sizeof(struct wpa_scan_result));
	s = socket(PF_INET, SOCK_DGRAM, 0);
	if (s < 0) {
		perror("socket[PF_INET,SOCK_DGRAM]");
		return -1;
	}

	memset(&iwr, 0, sizeof(iwr));
	strncpy(iwr.ifr_name, ifname, IFNAMSIZ);
	iwr.u.data.pointer = res_buf;
	iwr.u.data.length = IW_SCAN_MAX_DATA;

	if (ioctl(s, SIOCGIWSCAN, &iwr) < 0) {
		perror("ioctl[SIOCGIWSCAN]");
		close(s);
		return -1;
	}

	len = iwr.u.data.length;
	ap_num = 0;
	first = 1;

	pos = res_buf;
	end = res_buf + len;

	while (pos + IW_EV_LCP_LEN <= end) {
		int ssid_len;
	       /*  Fix */
             memcpy ((void *)&iwe2,(void*)pos,sizeof(struct iw_event));
            iwe = &iwe2;

//		iwe = (struct iw_event *) pos;


> -----Original Message-----
> From: hostap-bounces+willy=axis.com at shmoo.com 
> [mailto:hostap-bounces+willy=axis.com at shmoo.com] On Behalf Of 
> Duncan Grove
> Sent: den 1 oktober 2004 10:32
> To: hostap at shmoo.com
> Subject: WPA-PSK problem on ARM 
> 
> Hi,
> 
> I recently read a couple of posts about possible endian-ness 
> issues with the WPA support on ARM. I've also come across a 
> problem with WPA on ARM, although I'm not sure if it's related...
> 
> I've just tried wpa_supplicant 0.2.4 and the 0.2.5 beta on an 
> Ipaq that's running a derivative of the latest handhelds.org
> 2.4.19-rmk6-pxa1-hh37 kernel, although I've dragged a lot of 
> 2.4.26-and-beyond network and wireless card driver code into 
> the tree. 
> While the hostap-driver works with an unencrypted base 
> station, I couldn't get wpa_supplicant to connect to a 
> WPA-PSK base station. When I rebuilt everything on an x86 box 
> it connected fine. I was using a D-link DWL-650 with station 
> firmware 1.7.1.
> 
> During scanning wpa_supplicant decides to skip my AP, like this:
> 
> Received 262 bytes of scan results (1 BSSes) Scan results: 1 
> Selecting BSS from priority group 0
> 0: 00:11:92:a1:92:a0 ssid='8021xtst' wpa_ie_len=0 rsn_ie_len=0
>    skip - no WPA/RSN IE
> No suitable AP found.
> Setting scan request: 5 sec 0 usec
> 
> ... despite the fact that iwlist (from wireless_tools.27) 
> shows a wpa_ie:
> 
> ~/testing # iwlist wlan0 scan
>           Cell 02 - Address: 00:11:92:A1:92:A0
>                     ESSID:"8021xtst"
>                     Encryption key:on
>                     
> Extra:wpa_ie=dd160050f20101000050f20201000050f20201000050f202
> 
> Does anyone have any ideas about what might be going wrong? 
> (I've attached more complete logs below). Or where I should 
> start digging in the source code?
> 
> (Aside: I'm intending to move to WPA2 with EAP and AES when I 
> can find a commercial base station that supports them... is 
> the above problem likely to reappear then?)
> 
> Thanks,
> Duncan
> 
>