Thread: [Madwifi-devel] Working Linux 2.5+ patch
Status: Beta
Brought to you by:
otaku
From: Carl T. <ce...@ca...> - 2003-07-08 07:55:40
Attachments:
mad_linux26.patch.gz
|
Attached is a patch to build Linux 2.5+ kernel modules. This patch is alpha quality. Use it at your own risk. I have absolutely no idea what I'm doing when it comes to kernel stuff. I am amazed that this actually works. Basically, two things were required: 1. Convert task_queue stuff to work_queue stuff. I have no idea what I'm doing here but it seems to work. 2. Use the new kbuild system. I think I did this OK. The modules do load! To use: 1. Download and extract the madwifi distribution. 2. Apply this patch. 3. "make && make install" to build and install Linux 2.5+ drivers. "make linux24 && make linux24_install" for Linux 2.4. That's it. Use "modprobe ath_pci" to load all the drivers. More information is in the README. On Linux 2.5 when the interface is brought up some nasty messages show up in the syslog about improperly handled IRQ events. It could be I messed up on the task_queue stuff or it could be the problem exists in 2.4 too but is not reported by the kernel. At any rate, it seems to work anyway and the messages only occur when the interface is brought up. Performance is about the same in 2.5 as it is in 2.4 (maybe slightly higher). I am, however, getting better performance using 802.11a than has been reported to the list. Under ideal circumstances I am getting 22.6Mbps (or 23.7 if your definition of "megabit" is 1,000,000 bits). From what I understand this is about as good as it gets. Ideal circumstances is defined as single user mode, no WEP, 5 feet from access point. Strangely, performance is _much_ better at 5 feet than at 6 inches. I didn't test anywhere else but maybe there's a sweet spot distance where I would crack 25Mbps... I tested by time cat /path/to/nfs/mounted/really/big/file > /dev/null Because of NFS overhead, my actual thoughput must actually be slightly higher. Enjoy, Carl Thompson |
From: Carl T. <ce...@ca...> - 2003-07-08 09:13:02
|
I should mention that these results were using a 5212 based card which were reported as having suboptimal performance. This does not appear to be true in my tests; I got better performance from 5212 cards than I did with 5210 cards connecting to an old SMC 2755W access point. Carl Thompson Quoting Carl Thompson <ce...@ca...>: > ... > Performance is about the same in 2.5 as it is in 2.4 (maybe slightly > higher). I am, however, getting better performance using 802.11a than > has been reported to the list. Under ideal circumstances I am getting > 22.6Mbps (or 23.7 if your definition of "megabit" is 1,000,000 bits). > From what I understand this is about as good as it gets. Ideal > circumstances is defined as single user mode, no WEP, 5 feet from access > point. Strangely, performance is _much_ better at 5 feet than at 6 > inches. I didn't test anywhere else but maybe there's a sweet spot > distance where I would crack 25Mbps... I tested by > > time cat /path/to/nfs/mounted/really/big/file > /dev/null > > Because of NFS overhead, my actual thoughput must actually be slightly > higher. > > Enjoy, > Carl Thompson |