Page 3 of 4
Re: cross compiling acx100 wifi kernel module
Posted: Thu Aug 05, 2010 3:41 pm
by smartsmurf
From looking at your dmesg dump I can see that data packets seem OK, but ar de-justified by 20 bytes:
Code: Select all
4>acx: unknown EID 96 in mgmt frame at offset 89. IE: 60 6C DD 18 00 50 F2 02 01 01 84 00 03 A4 00 00
<4>27 A4 00 00 42 43 5E 00 62 32 2F 00 DD 09 00 03
<4>7F 01 01 00 00 FF 7F DD 0A 00 03 7F 04 01 00 20
<4>00 00 00 A0 00 E5 82 8C 00 0A 00 C4 60 01 00 7F
<4>01 01 00 00 FF 7F DD 0A 00 03 7F 04 01 00 20 00
<4>00 00 DD 0E 00 50 F2 04 10 4A 00 01 10 10 44 00
<4>01 02 F1 00 D7 82 DD 00 0A 00 14 61 01 00
So, I modified some code in order to set the physical header length from 8 to 28 in the case of unknown radio 0x1B:
Code: Select all
if ((adev->rx_config_1 & RX_CFG1_INCLUDE_PHY_HDR)
|| (adev->firmware_numver >= 0x02000000)){
if( IS_ACX111(adev) ){
adev->phy_header_len = (adev->radio_type == RADIO_UNKNOWN_1B) ? 28 : 8;
} else {
adev->phy_header_len = 4;
}
} else {
adev->phy_header_len = 0;
}
log(L_INIT, "setting phy_header_len to %04d\n", adev->phy_header_len );
Modified driver is attached.
To show the "L_INIT" debug output (which might be important in your case), you need to add acx_debug=2 as driver parameter.
Good luck.
PS: just for clarification - the driver only supports WEP encryption. No WPA...
<< attachment removed >>
Re: cross compiling acx100 wifi kernel module
Posted: Fri Aug 06, 2010 9:41 pm
by smartsmurf
jogi10 wrote:
Yes. But step by step
Indeed.
Next step is to get rid of that error:
Code: Select all
<4>issue_cmd(cmd:ACX1xx_CMD_INTERROGATE,buflen:24,type:0x0005)
<4>ctrl inpipe=0x80000380 outpipe=0x80000300
<4>sending USB control msg (out) (blocklen=8)
<4>wrote 8 bytes
<4>sending USB control msg (in) (acklen=28)
<4>acx: issue_cmd(): USB read error -75
<4>acx: issue_cmd(cmd:ACX1xx_CMD_INTERROGATE) FAILED
New module is attached.
<<attachment removed>>
Re: cross compiling acx100 wifi kernel module
Posted: Mon Aug 09, 2010 9:17 am
by smartsmurf
This will be not easy. "104" is the error code representing ECONNRESET. This means your device reset itself (=disconnect/reconnect from USB) while in the process of joining which might be due to wrong firmware or improper initialization of the JOIN request. Reason for this behaviour is not obvious.
You might try "insmod acx.ko debug=0x80A" to enable some debug flags in order to see more details.
Re: cross compiling acx100 wifi kernel module
Posted: Wed Aug 11, 2010 12:16 pm
by smartsmurf
Attached another module worth trying. Please use this time the following command line
which will also show data packets. Thanks!
Re: cross compiling acx100 wifi kernel module
Posted: Thu Aug 12, 2010 2:45 pm
by smartsmurf
I didn't find that "tx buffers full" message in your logs. Actually I believe there are responses. You can see
Code: Select all
using buf#3 free=7 len=30
...
using buf#4 free=7 len=30
...
using buf#5 free=7 len=30
...
using buf#6 free=7 len=30
And "free=7" means, that there are still 7 buffers available, so "adev->tx_free++" must be called at some point inbetween.
Currently I see another suspicios entry:
Code: Select all
<4>drivers/net/wireless/acx/common.c: 2958 rx: UNKNOWN time:2768306226 len:4294967268 signal:100 SNR:11 macstat:9F phystat:F0 phyrate:4 status:1
<4>rx: 802.11 buf[4294967268]:
<4>drivers/net/wireless/acx/common.c: 2979 rx: NOT receiving packet (UNKNOWN): size too small (4294967268)
Although not directly related to the AUTH issue, it might lead to overflow and corruption somehow...