If you are trying to get a device working with ndiswrapper, please help ndiswrapper project wiki by contributing your findings. See HowToContribute

Trendnet TEW-643PI

From NDISWrapper
Revision as of 09:19, 25 March 2011 by Brinley (Talk | contribs) (Updates from Herrmaro)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Trendnet TEW-643PI Wireless 802.11 bdgn

Info

XP driver did not work for a 64bit pristine Debian Squeeze system, but works smooth with 32bit

---

ndis was doing something weird, every time i tried to wrap the driver, ndis would give me the options available to it. not sure how i got it to quit doing that, or if it was even me. tried winxp winvista and win2000 drivers only one worked was 2000.

---

I needed to apply a work-around to get this device up under a 32-bit SMP kernel. It appears that when an ndiswrapper kernel thread is scheduled to a different CPU the device hangs. You can usually pump about 100-200Mb full bore before this occurs. (This sound familiar)? If you lock the threads down to one CPU the device stays up. Under a 64-bit kernel, the ntos_wq thread causes a kernel oops and hangs the machine, usually within 5-6 seconds. (It might be related, but I didn't do this affinity test on 64-bit).

The work around is QCD on gentoo using /etc/conf.d/local.start:

schedtool -a 0 `ps -e | grep _wq | gawk '{print $1}'`

The same behaviour was observed using all three preemption models. More SMP investigation is obviously needed, but at the least, the code could lock its own threads down as a quick band-aid.

My system:

2 CPUs, i686 AMD Athlon(tm) II X2 215 Processor

kernel 2.6.32-gentoo-r6 (SMP)

ndiswrapper-1.55-r1