Re: [ath5k-devel] Hardware encryption on AR5413 broken?

Previous thread: [ath5k-devel] Accessing channel utilization stats by Jonathan Guerin on Thursday, October 28, 2010 - 10:54 pm. (1 message)

Next thread: [ath5k-devel] Promiscuous mode by frisbee on Wednesday, November 10, 2010 - 7:42 am. (1 message)
From: Stefan Nickl
Date: Wednesday, November 10, 2010 - 2:26 am

Hello list,

I'm experiencing encryption problems that may be related to this bug
report (although the report is about AR2413):
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/568090

The crypto handshake appears to complete just fine, however no reply is
received from DHCP.
After setting a fixed IP instead, I can ping for a while but the
connection chokes on stronger traffic.
When ath5k is loaded with nohwcrypt, the problem goes away.

Is there a known issue?
Anyone working on it?

Test platform: Fedora 14 x86(i386) with
wireless-testing-2.6.37-rc1-wl+/master-2010-11-08 as well as
2.6.35.6-48.fc14.i686+compat-wireless-2010-11-07.

Tested with two different miniPCI modules, both based on AR5413.

Same behavior with wpa_supplicant 0.6.8 as included with F14 with
-Dwext, and with wpa_supplicant 0.8.x (2010-11-08) with both wext and
nl80211.

Best Regards,

-- 
Stefan Nickl
System Engineer
Kontron Modular Computers GmbH



Sitz der Gesellschaft / Registered Office: Kaufbeuren
Amtsgericht / Local District Court: Kempten, HRB Nr.: / Trade Register No.:  6195 
Geschaeftsfuhrer / Managing Directors: Ulrich Gehrmann, Thomas Sabisch

Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen und ist Eigentum von Kontron. Die Verwendung
und Weitergabe von jeglichen Inhalten ist ohne ausdruckliche schriftliche Genehmigung von Kontron strikt untersagt. Wenn Sie 
diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten diese Mail und enthaltene Dokumente.

The information contained in this document is CONFIDENTIAL and property of Kontron. Any unauthorized review, use, 
disclosure or distribution is prohibited without express written consent of Kontron. If you are not the intended recipient, please 
contact the sender and destroy all copies of the original message and enclosed attachments.

_______________________________________________
ath5k-devel mailing ...
From: Bruno Randolf
Date: Monday, November 15, 2010 - 8:14 pm

I just confirmed that encryption still works for me with current wireless-
testing in AP as well as in STA (client) mode. Please provide us with more 
information:

1.) What is your chipset revision (check in dmesg)? Mine is:
ath5k phy1: Atheros AR5414 chip found (MAC: 0xa5, PHY: 0x61). AR5413 and 
AR5414 should be very similar, AFAIK.

2.) What encryption did you use? TKIP or AES (CCMP)? I tested both in STA 
mode. I can get DHCP and i tested several minutes of high traffic. No problem 
as far as I can see. I used wpa_supplicant v0.7.1.

3.) Is this on a SMP (multi) processor machine? I test with a single 
processor.

bruno
_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel
From: Stefan Nickl
Date: Tuesday, November 16, 2010 - 2:30 am

Hi Bruno,

thanks for your reply.


Interestingly, I have the same chipset as you when looking at dmesg:
ath5k phy0: Atheros AR5414 chip found (MAC: 0xa5, PHY: 0x61)

I had looked at lspci before, which identifies
168c:001b (rev 01) /  Subsystem: 168c:2062


Machine is a Atom Z510 single core/processor.

The EEPROM of this module is a bit quirky, for example 11a mode (which 
is my primary target) only works with ath5k when the corresponding mode 
flag is modified after it was read from EEPROM.
Apparently, madwifi did not care about this.

What are the numerical PCI IDs of your device?
Maybe detection is sent down the wrong track because of other IDs?

I can also provide the debug output of wpa_supplicant if it helps.

Best Regards and sorry for the disclaimer,

-- 
Stefan Nickl
System Engineer
Kontron Modular Computers GmbH

Sitz der Gesellschaft / Registered Office: Kaufbeuren
Amtsgericht / Local District Court: Kempten, HRB Nr.: / Trade Register No.:  6195 
Geschaeftsführer / Managing Directors: Ulrich Gehrmann, Thomas Sabisch

Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen und ist Eigentum von Kontron. Die Verwendung
und Weitergabe von jeglichen Inhalten ist ohne ausdrückliche schriftliche Genehmigung von Kontron strikt untersagt. Wenn Sie 
diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten diese Mail und enthaltene Dokumente.

The information contained in this document is CONFIDENTIAL and property of Kontron. Any unauthorized review, use, 
disclosure or distribution is prohibited without express written consent of Kontron. If you are not the intended recipient, please 
contact the sender and destroy all copies of the original message and enclosed attachments.

_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel
From: Bruno Randolf
Date: Tuesday, November 16, 2010 - 2:46 am

The ath5k phy0: Atheros AR5414 chip found (MAC: 0xa5, PHY: 0x61) line is the 

Ok. I did the same (only one ath5k AP, though...).


Uh, then complain to your vendor ;-)

Maybe there is something in the EEPROM which disables encryption, too. There 
is a flag for this: ATH_CRYPT_CAP_CIPHER_AESCCM. Can you check if it is 

It is (lspci -vn)

00:0f.0 0200: 168c:001b (rev 01)
        Subsystem: 168c:2062
        Flags: bus master, medium devsel, latency 168, IRQ 5
        Memory at a0020000 (32-bit, non-prefetchable) [size=64K]
        Capabilities: [44] Power Management version 2
        Kernel driver in use: ath5k
        Kernel modules: ath5k

But I don't think it makes a big difference. The SREV (MAC: 0xa5) is used in 
the ath5k code, not the PCI IDs.

bruno
_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel
From: Stefan Nickl
Date: Wednesday, November 17, 2010 - 2:28 am

Tried WPA/TKIP and even WEP40, all show the same effect.

Tried, but they don't support open source and I can't switch easily. 
Long story.

I have another module from a different vendor which has behaved the same 

Good one. But info word #9 is zero in the EEPROM, and a printk reveals 

True, all the same. Just keeps getting stranger...

-- 
Stefan Nickl
System Engineer
Kontron Modular Computers GmbH

Sitz der Gesellschaft / Registered Office: Kaufbeuren
Amtsgericht / Local District Court: Kempten, HRB Nr.: / Trade Register No.:  6195 
Geschaeftsführer / Managing Directors: Ulrich Gehrmann, Thomas Sabisch

Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen und ist Eigentum von Kontron. Die Verwendung
und Weitergabe von jeglichen Inhalten ist ohne ausdrückliche schriftliche Genehmigung von Kontron strikt untersagt. Wenn Sie 
diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten diese Mail und enthaltene Dokumente.

The information contained in this document is CONFIDENTIAL and property of Kontron. Any unauthorized review, use, 
disclosure or distribution is prohibited without express written consent of Kontron. If you are not the intended recipient, please 
contact the sender and destroy all copies of the original message and enclosed attachments.

_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel
From: Stefan Nickl
Date: Wednesday, November 17, 2010 - 6:40 am

Just tried with that different module:
ath5k phy1: Atheros AR5413 chip found (MAC: 0xa4, PHY: 0x61)

Same effect, however I recognized this message in dmesg now:
wlan0: invalid aid value 6; bits 15:14 not set

-- 
Stefan Nickl
System Engineer
Kontron Modular Computers GmbH

Sitz der Gesellschaft / Registered Office: Kaufbeuren
Amtsgericht / Local District Court: Kempten, HRB Nr.: / Trade Register No.:  6195 
Geschaeftsführer / Managing Directors: Ulrich Gehrmann, Thomas Sabisch

Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen und ist Eigentum von Kontron. Die Verwendung
und Weitergabe von jeglichen Inhalten ist ohne ausdrückliche schriftliche Genehmigung von Kontron strikt untersagt. Wenn Sie 
diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten diese Mail und enthaltene Dokumente.

The information contained in this document is CONFIDENTIAL and property of Kontron. Any unauthorized review, use, 
disclosure or distribution is prohibited without express written consent of Kontron. If you are not the intended recipient, please 
contact the sender and destroy all copies of the original message and enclosed attachments.

_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel
From: Bruno Randolf
Date: Wednesday, November 17, 2010 - 6:41 pm

I'm very surprised you see problems even with WEP - so could you please check 
if the interface works correctly without any encryption?

And can you describe the symptoms more accurately? You said you can connect, 
but no DHCP and traffic stalls after a while on stronger traffic. Can you 
describe more details and give me a step-by-step instruction how to reproduce?

bruno
_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel
From: Stefan Nickl
Date: Friday, November 26, 2010 - 3:42 am

Did some testing with ping, and on the module with a4 MAC, receiving 
seems to work flawlessly up to full 1478 bytes frames.
Transmission is ok only up to 68 byte frames (ping -s 60).
Frames any bigger (ping -s 61 ... ping -s 1470) only get through 1% of 
the time (99% packet loss).

On the a5 MAC things are worse. In the above case arp worked, but here 
arps get through only around ~1% of the time, which makes further 
testing hard, but still, something gets through.

The DHCP issue at least starts figuring out considering a packet size 
limitation.

I'm not so sure if it's really traffic related, I just noticed that in 
the corner cases, using flood ping would not get anything through.

So it's very simple to reproduce: Linux client with ath5k pings (or 
arpings) a Linux server connected to the AP in 802.11a mode.

Tried with Linksys WAP610N and WRT610N.
I also have a report with a Trapeze AP, probably MP-620B.

-- 
Stefan Nickl
System Engineer
Kontron Modular Computers GmbH

Sitz der Gesellschaft / Registered Office: Kaufbeuren
Amtsgericht / Local District Court: Kempten, HRB Nr.: / Trade Register No.:  6195 
Geschaeftsführer / Managing Directors: Ulrich Gehrmann, Thomas Sabisch

Diese E-Mail enthaelt vertrauliche und/oder rechtlich geschuetzte Informationen und ist Eigentum von Kontron. Die Verwendung
und Weitergabe von jeglichen Inhalten ist ohne ausdrückliche schriftliche Genehmigung von Kontron strikt untersagt. Wenn Sie 
diese E-Mail irrtuemlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten diese Mail und enthaltene Dokumente.

The information contained in this document is CONFIDENTIAL and property of Kontron. Any unauthorized review, use, 
disclosure or distribution is prohibited without express written consent of Kontron. If you are not the intended recipient, please 
contact the sender and destroy all copies of the original message and enclosed attachments.

_______________________________________________
ath5k-devel mailing ...
Previous thread: [ath5k-devel] Accessing channel utilization stats by Jonathan Guerin on Thursday, October 28, 2010 - 10:54 pm. (1 message)

Next thread: [ath5k-devel] Promiscuous mode by frisbee on Wednesday, November 10, 2010 - 7:42 am. (1 message)