Re: [Ndiswrapper-general] Centrino wlan driver loading
Status: Beta
Brought to you by:
pgiri
From: Stefan <ste...@gy...> - 2003-11-24 09:41:03
|
>Hi, =20 =20 > On Sun, 2003-11-23 at 20:30, Stefan D=C3=B6singer wrote: > > Hello, > > I added(or better copied, I'm lazy if it makes sense)=20 >>NdisUnchainBufferAtBack=20 > > and funtions needed by it(NdisQueryPacket, NdisQueryBufferOffset). A di= ff=20 > > made mit cvs diff -ur HEAD is attached. Hope this is ok. =20 > Please use the same coding style as the rest of the code and I'll apply > it. I don't want you to rename reactos variables but I want intendation > to the the same as the rest of the code. Also use C comments not C++ > comments. =20 Will try. What's the difference between C comments and C++ comments? I know= =20 the //komment and /*Komment*/ What about tabstopps? Some developers don't like them because they can't be= =20 unified. I saw some in the code, should I use spaces or tabstopps? =20 >> I had some problems with different names in ndiswrapper and ReactOS, I= =20 hope I=20 > > used the right names. > >=20 > > Loading the centrino driver succeeded with the latest CVS(I see that my= =20 > > changes got merged into it). However I noticed that there's some confli= ct=20 > > with my wired network(Broadcom 4401). If the b44 module is loaded(Kerne= l=20 > > 2.4.22) loading the wlan driver fails(NdisMInitializeScatherGatherDMA=20 get's=20 > > mysterious values). There are also some relations wether the=20 >>X-Window-Sytem=20 > > is started(Loading the driver failed in singe user) =20 > Strange.... I think the relation with the X-Window-System is a timing issue with vesafb= =2E=20 On console I use the unacceleratet vesafb due to incompatiblities with my X= 11=20 driver fglrx. Writing a lot of things takes a lot of time. The driver loads= ,=20 tail -f prints out /var/log/messages, the CPU is busy and the driver crashe= s=20 on some later point >> With my new patch it's sometimes possible to start the interface(Set an= IP=20 or=20 > > do ifconfig ethX up), but this fails quite often. However, trying to se= nd=20 > > something(a ping or starting a dhcp client) crashes terribly(Kernel pan= ic=20 > > without anything printed on the console, just the Capslock LED flashes). =20 > Serial console time :) My modern Notebook(Acer Travelmate 803) doesn't have a serial port :-( I would have to by an expensive(~150 =A4) port replicator. I think the cra= sh is=20 related to Scather/Gather dma(see my other mail) > > I think the problems are due the unimplemented functions, mainly becaus= e=20 of=20 > > the NdisMInitializeScatherGatherDMA. In this function I just print out = the=20 > > arguments and return NDIS_STATUS_SUCCESS to make the driver continue.=20 >Further=20 > > problems are the NdisXXXWorkItem functions. According to the descriptio= n=20 they=20 > > are used to quere some work in the interrupt handler to release the=20 processor=20 > > earlier. =20 NdisXXXWorkItem are used to implement interrupt bottom halves etc. The linux equivalent is kernel tasks and workqueues (2.6). Look at the timer and interrupt code and I think you'll understand. =20 Will have a look > > Unloading the driver fails too. No unresolved funtion is reported any=20 more, I=20 > > think it's also related to NdisMInitializeScatherGatherDMA. =20 > It's a pain to reboot often right? :) If you use 2.6 you can use forced > module unloading which may sometimes work and save a reboot. I have problems using 2.6, related to my synaptics mouse driver. I use the= =20 Notebook at scool and I need 2.4 stability. Permanently switching between 2= =2E4=20 and 2.6 is a pain with the synaptics driver. Will see if I can do something= =20 about it =20 > /Pontus =20 Cheers, Stefan =20 |