rt61pci: WEP broken in current wireless-testing

rt61pci: WEP broken in current wireless-testing

[Pavel Roskin]

Hello!

Recent changes in wireless-testing have broken WEP support in rt61pci.  
  An older kernel from wireless-testing identified as 2.6.34-rc5-wl is  
working fine.

I'm using an x86_64 system.  The device is

Network controller [0280]: RaLink RT2561/RT61 rev B 802.11g [1814:0302]

The device is in managed mode.  I'm using WEP with a 40-bit key (I  
know it's ridiculously insecure, but it's not my choice).  With the  
current kernel, the association works, but no traffic is received by  
the station.

That's the relevant part of .config:

CONFIG_RT2X00=m
# CONFIG_RT2400PCI is not set
# CONFIG_RT2500PCI is not set
CONFIG_RT61PCI=m
CONFIG_RT2800PCI_PCI=y
# CONFIG_RT2800PCI is not set
# CONFIG_RT2500USB is not set
# CONFIG_RT73USB is not set
# CONFIG_RT2800USB is not set
CONFIG_RT2X00_LIB_PCI=m
CONFIG_RT2X00_LIB=m
CONFIG_RT2X00_LIB_FIRMWARE=y
CONFIG_RT2X00_LIB_CRYPTO=y
CONFIG_RT2X00_LIB_LEDS=y
CONFIG_RT2X00_LIB_DEBUGFS=y
CONFIG_RT2X00_DEBUG=y

arc4 support is compiled as a module and it's loaded.  I haven't seen  
anything unusual in the kernel log.

Curiously, I'm seeing some interesting warnings with the working  
kernel (2.6.34-rc5-wl).  They don't appear immediately, as you can see  
from the timestamps:

[   74.534338] phy1 -> rt2x00lib_request_firmware: Info - Loading  
firmware file 'rt2561.bin'.
[   74.534343] rt61pci 0000:04:03.0: firmware: requesting rt2561.bin
[   74.638978] phy1 -> rt2x00lib_request_firmware: Info - Firmware  
detected - version: 0.8.
[   74.801343] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 0  
- CWmin: 3, CWmax: 4, Aifs: 2, TXop: 102.
[   74.801350] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 1  
- CWmin: 4, CWmax: 5, Aifs: 2, TXop: 188.
[   74.801357] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 2  
- CWmin: 5, CWmax: 10, Aifs: 3, TXop: 0.
[   74.801363] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 3  
- CWmin: 5, CWmax: 10, Aifs: 7, TXop: 0.
[   81.212513] wlan2: authenticate with 00:23:97:5c:82:7e (try 1)
[   81.214001] wlan2: authenticated
[   81.250009] wlan2: associate with 00:23:97:5c:82:7e (try 1)
[   81.251750] wlan2: RX AssocResp from 00:23:97:5c:82:7e (capab=0x411  
status=0 aid=1)
[   81.251753] wlan2: associated
[   81.252014] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 0  
- CWmin: 2, CWmax: 3, Aifs: 2, TXop: 47.
[   81.252020] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 1  
- CWmin: 3, CWmax: 4, Aifs: 2, TXop: 94.
[   81.252027] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 2  
- CWmin: 4, CWmax: 10, Aifs: 3, TXop: 0.
[   81.252034] phy1 -> rt2x00mac_conf_tx: Info - Configured TX queue 3  
- CWmin: 4, CWmax: 10, Aifs: 7, TXop: 0.
[  652.484017] phy1 -> rt61pci_txdone: Warning - TX status report  
missed for entry 15
[  652.970531] phy1 -> rt61pci_txdone: Warning - TX status report  
missed for entry 2

It's should be rather straightforward to bisect, but I'll be away from  
that system until next Monday.

-- 
Regards,
Pavel Roskin

Re: rt61pci: WEP broken in current wireless-testing

[John W. Linville]

On Thu, May 06, 2010 at 07:51:57AM -0400, Pavel Roskin wrote:

> Recent changes in wireless-testing have broken WEP support in rt61pci.  
>   An older kernel from wireless-testing identified as 2.6.34-rc5-wl is  
> working fine.

> It's should be rather straightforward to bisect, but I'll be away from  
> that system until next Monday.

When you get to this, you probably should use wireless-next-2.6
(or wireless-2.6) for the bisection.  The pulls from linux-2.6 in
wireless-testing make bisection much more painful than necessary
(as long as the problem is actually evident in the other kernels).
You can use the 'master-<date>' tags as guidelines for matching
wireless-testing versions to representative wireless-next-2.6 (or
wireless-2.6) tree versions.

Hth!

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: rt61pci: WEP broken in current wireless-testing

[Pavel Roskin]

Quoting "John W. Linville" <linville@tuxdriver.com>:

> When you get to this, you probably should use wireless-next-2.6
> (or wireless-2.6) for the bisection.  The pulls from linux-2.6 in
> wireless-testing make bisection much more painful than necessary
> (as long as the problem is actually evident in the other kernels).

I know it's painful!

> You can use the 'master-<date>' tags as guidelines for matching
> wireless-testing versions to representative wireless-next-2.6 (or
> wireless-2.6) tree versions.

Thank you for the tips!

-- 
Regards,
Pavel Roskin

Re: rt61pci: WEP broken in current wireless-testing

[Gertjan van Wingerde]

[-- Attachment #1: Type: text/plain, Size: 705 bytes --]

On 05/07/10 03:14, Pavel Roskin wrote:
> Quoting "John W. Linville" <linville@tuxdriver.com>:
> 
>> When you get to this, you probably should use wireless-next-2.6
>> (or wireless-2.6) for the bisection.  The pulls from linux-2.6 in
>> wireless-testing make bisection much more painful than necessary
>> (as long as the problem is actually evident in the other kernels).
> 
> I know it's painful!
> 
>> You can use the 'master-<date>' tags as guidelines for matching
>> wireless-testing versions to representative wireless-next-2.6 (or
>> wireless-2.6) tree versions.
> 
> Thank you for the tips!
> 

Hi,

It might be that one of my recent patches broke this. Can you try the
attached patch?

---
Gertjan

[-- Attachment #2: rt61pci-wep-fix.diff --]
[-- Type: text/plain, Size: 529 bytes --]

diff --git a/drivers/net/wireless/rt2x00/rt2x00crypto.c b/drivers/net/wireless/rt2x00/rt2x00crypto.c
index d291c78..583dacd 100644
--- a/drivers/net/wireless/rt2x00/rt2x00crypto.c
+++ b/drivers/net/wireless/rt2x00/rt2x00crypto.c
@@ -128,6 +128,7 @@ void rt2x00crypto_tx_remove_iv(struct sk_buff *skb, struct txentry_desc *txdesc)
 
 	/* Pull buffer to correct size */
 	skb_pull(skb, txdesc->iv_len);
+	txdesc->length -= txdesc->iv_len;
 
 	/* IV/EIV data has officially been stripped */
 	skbdesc->flags |= SKBDESC_IV_STRIPPED;

Re: rt61pci: WEP broken in current wireless-testing

[Pavel Roskin]

Quoting Gertjan van Wingerde <gwingerde@gmail.com>:

> It might be that one of my recent patches broke this. Can you try the
> attached patch?

I'm not sure I'll be able to test it before Monday.

But the patch looks good.  If the encrypted Tx is broken, it would be  
entirely consistent with my observations (monitor mode works,  
association works, but no data flowing).

-- 
Regards,
Pavel Roskin