Hi,

I recently converted my WRT54GS to OpenWRT. Unfortunately I could not get the WEP encryption to work. At one point I discovered that the keys shown by iwconfig for key 1 and key 4 was the same. So I took a look in the sourcecode for the wifi utility.

In the setup_wext_wep function (line 337 of wificonf.c), all the 4 keys are set, one at a time. If the code also finds a wl0_key nvram attribute, it does the following:

        wrq.u.data.flags = i | IW_ENCODE_RESTRICTED;
        IW_SET_EXT_ERR(skfd, ifname, SIOCSIWENCODE, &wrq, "Set Encode");

The trouble is that wrq still contains the last key that was configured, so in effect the above code overwrites the key that wl0_key points to.

The following (untested, as I don't have the tool-chain setup) patch should fix the problem

--- wificonf.c  2005-07-17 14:59:44.000000000 +0200
+++ wificonf2.c 2005-07-29 14:46:10.163466731 +0200
@@ -376,13 +376,16 @@
 }
 void setup_wext_wep(int skfd, char *ifname)
 {
-       int i, keylen;
+       int i, keylen, select_key;
        struct iwreq wrq;
        char keystr[5];
        char *keyval;
        unsigned char key[IW_ENCODING_TOKEN_MAX];
 
        strcpy(keystr, "key1");
+
+       select_key = atoi(nvram_safe_get(wl_var("key")));
+
        for (i = 1; i <= 4; i++) {
                if (keyval = nvram_get(wl_var(keystr))) {
                        keylen = iw_in_key(keyval, key);
@@ -391,18 +394,13 @@
                                wrq.u.data.length = keylen;
                                wrq.u.data.pointer = (caddr_t) key;
                                wrq.u.data.flags = i;
+                               if (i == select_key)
+                                       wrq.u.data.flags |= IW_ENCODE_RESTRICTED;
                                IW_SET_EXT_ERR(skfd, ifname, SIOCSIWENCODE, &wrq, "Set Encode");
                        }
                }
                keystr[3]++;
        }
-       
-       
-       i = atoi(nvram_safe_get(wl_var("key")));
-       if (i > 0 && i < 4) {
-               wrq.u.data.flags = i | IW_ENCODE_RESTRICTED;
-               IW_SET_EXT_ERR(skfd, ifname, SIOCSIWENCODE, &wrq, "Set Encode");
-       }
 }

Hi Hjerne,

also have problems to get WRT54G v2 to work in client mode. It does connect to an AP no problem via a wl join <ssid> [key xxxxxxxxxx] [imode bss|ibss] [amode open|shared] command though (in my case imode bss , amode shared). But the wificonf routine seems to be making some wrong settings for simple operation using WEP from a 'station'.
Tried your suggested code change in the latest HEAD checkout but that results in iwconfig eth1 showing the fourth key, ignoring the wl0_key=1 entry in NVRAM. 'wl primary_key x' statement can temporarily override that.

So made some more changes; code below works as far as the WEP key is concerned. Needs further change though as all 4 keys need to be communicated and then the selected as last one. But ....
there are more problems; via wl have figured out that the access method does not get set properly (shared in my case) and that the interface gets set to repeater mode where I am asking via NVRAM for station mode.
On the side also noticed that several other things need reviewing, like shortslot_override that seems not to get set (noticed big bitrate improvement on long distance link by setting it via WL.

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
void setup_wext_wep(int skfd, char *ifname)
{
    int i, keylen;
    struct iwreq wrq;
    char keystr[5];
    char *keyval;
    unsigned char key[IW_ENCODING_TOKEN_MAX];

    i = atoi(nvram_safe_get(wl_var("key")));         /* get keynumber */
    switch (i) {
        case 2:    strcpy(keystr, "key2");     break;
        case 3: strcpy(keystr, "key3");        break;
        case 4: strcpy(keystr, "key4");        break;
        default: strcpy(keystr, "key1");    break;
    }                    
    if (i > 0 && i < 5) {                     /* check range is 1 .. 4 */
        if (keyval = nvram_get(wl_var(keystr))) {    /* fetch keycode */
            keylen = iw_in_key(keyval, key);    /* get length of key */
            if (keylen > 0) {            /* only act if finite length */
                wrq.u.data.length = keylen;    /* fill in iwreq structure */
                wrq.u.data.pointer = (caddr_t) key;
                wrq.u.data.flags = i | IW_ENCODE_RESTRICTED;                               
                IW_SET_EXT_ERR(skfd, ifname, SIOCSIWENCODE, &wrq, "Set Encode");
            }
        }
    }
}
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

rgds
doddel
Portugal

(Last edited by doddel on 1 Aug 2005, 05:06)

Ok. This patch does the correct thing for me. As I don't know the inner workings of how to configure the wifi device, I choose the fool-proof way:

diff -u -r1.2.2.13 wificonf.c
--- wificonf.c  25 Jul 2005 07:51:14 -0000      1.2.2.13
+++ wificonf.c  4 Aug 2005 14:46:24 -0000
@@ -405,8 +405,17 @@

        i = atoi(nvram_safe_get(wl_var("key")));
        if (i > 0 && i < 4) {
-               wrq.u.data.flags = i | IW_ENCODE_RESTRICTED;
-               IW_SET_EXT_ERR(skfd, ifname, SIOCSIWENCODE, &wrq, "Set Encode");
+               keystr[3] = '0' + i;
+               if (keyval = nvram_get(wl_var(keystr))) {
+                       keylen = iw_in_key(keyval, key);
+                       
+                       if (keylen > 0) {
+                               wrq.u.data.length = keylen;
+                               wrq.u.data.pointer = (caddr_t) key;
+                               wrq.u.data.flags = i | IW_ENCODE_RESTRICTED;
+                               IW_SET_EXT_ERR(skfd, ifname, SIOCSIWENCODE, &wrq, "Set Encode");
+                       }
+               }
        }
 }

This patch is against HEAD of whiterussian. Is this the correct place to post it, or should I send it somewhere else?