[Prism54-devel] SoftMAC PCI development

[Stef Simoens]

Jean-Baptiste, Dear list,

I downloaded the SoftMAC USB/PCI driver (as explained on JB's site: 
prism54-project--mainline and ./build-devo.sh)

I only have PCI hardware (SMC 2802W V2) on a PowerMac (an old machine 
before the USB-era, and big-endian, custom kernel.org 2.6.12.1 kernel). 
To my surprise the driver compiled without any problems*. Next, 
modprobe'ing and insmod'ing everything did NOT lock up my Mac! Hooray!

This list is rather silent the last days, so I decided to post my 
testing here (which is, I must admit, rather promising).

1. My neighbours have an AP (and on my WinXP laptop, signal strength is 
low). The PCI card found the AP, and my interface got configured (I got 
a DHCP assigned IP).

iwconfig output (edited)
IEEE 802.11b  ESSID:"NETGEAR"
Bit Rate:1 Mb/s   Tx-Power:20 dBm   Sensitivity=0/3
Link Quality=42/94  Signal level=-53 dBm  Noise level=-95 dBm
Rx invalid nwid:1872  Rx invalid crypt:0  Rx invalid frag:0

and ifconfig says I got IP 192.168.0.4

2. I'm not interested in my Neighbour's net, I want to connect to my laptop.
So, I do
[root at agnes prism]# iwconfig eth2 essid BUNDERBOS
[root at agnes prism]# iwconfig eth2 mode ad-hoc
[root at agnes prism}# ifconfig eth2 192.168.1.1

(when I restart my DHCP-server on agnes, WinXP finds BUNDERBOS, 
connects, and gets IP 192.168.1.254)
So, why not:
[root at agnes prism]# ping 192.168.1.254
PING 192.168.1.254 (192.168.1.254) 56(84) bytes of data.
64 bytes from 192.168.1.254: icmp_seq=1 ttl=128 time=5.78 ms
64 bytes from 192.168.1.254: icmp_seq=2 ttl=128 time=10.5 ms
64 bytes from 192.168.1.254: icmp_seq=3 ttl=128 time=4.08 ms

Kinda neat :)

But then, loading a web-page using the wireless interface ... only gives 
half a page and makes the connection get lost.

When I look in the logs:

islsm_input.c: Device reports hardware type : 00009003
islsm_input.c: Device is an isl3890
islsm_input.c: Unusual hardware information. Please contact the module 
author with the above information.

agnes kernel: Unknown option length 7, magic 1007,please report. data :
agnes kernel: [IN +5324922]
agnes kernel: 00000000: 07 10 10 00 00 00 30 00 00 00 40 00 00 00
agnes kernel: Unknown option length 3, magic 1008,please report. data :
agnes kernel: [IN +5324924]
agnes kernel: 00000000: 08 10 10 00 00 00
agnes kernel: Unknown option length 3, magic 1100,please report. data :
agnes kernel: [IN +5324925]
agnes kernel: 00000000: 00 11 14 14 14 14
agnes kernel: Unknown option length 5, magic 1905,please report. data :
agnes kernel: [IN +5324926]
agnes kernel: 00000000: 05 19 80 00 70 fe 80 00 70 fe

this also looks strange:
agnes kernel: islsm_transmit called.
agnes kernel: data length 42, padding 2
agnes kernel: my computed padding is 4, real 2
agnes kernel: alloc for id c7ce58e0
agnes kernel: allocated queue 6
agnes kernel: ****** Problem in skb to be tx'd : headroom too small *******

after the web-browsing, i get the following messages:
agnes kernel: alloc for id c85eea60
agnes kernel: no free queue !

There are probably some other interesting stuff to find in the logging; 
but that would make this mail too big. If I can do any testing, please 
ask me, I'd be pleased to help.

One more thing: when the connection is "locked up", rmmod islsm_pci and 
insmod islsm_pci makes it work again for a short amount of time. But 
usually, the second rmmod makes the kernel crash (SIG 11).

Jean-Baptiste Note schreef:
> Dear list,
> 
> I've started serious PCI development. The results can be found at :
> 
> http://jbnote.free.fr/prism54usb
> 

-- 
Stef Simoens