No, but the linux driver has some good code in it... I have a tulip card and so may be thinking of doing one anyways... It might be good to take some time off and let others play with the stack for a while... :) david ----- Original Message ----- From: "Beprojects.com" <info@xxxxxxxxxxxxxx> To: <openbeosnetteam@xxxxxxxxxxxxx> Sent: Saturday, March 23, 2002 2:02 AM Subject: [openbeosnetteam] Re: Drivers... > Speaking of drivers.... Is there source code available for the standard > tulip driver? I'm tired of taking my one nic down to test, so I want to see > if I can get 2 nic cards working. Unfortunately, my 2nd nic is a Linksys > v4, which isn't supported by the standard linksys driver, however it is a > tulip, so I'm thinking I can tweak it. > > Peder > > > ----- Original Message ----- > From: "David Reid" <dreid@xxxxxxxxxxxx> > To: "OpenBeOS Network Team" <openbeosnetteam@xxxxxxxxxxxxx> > Sent: Friday, March 22, 2002 7:38 PM > Subject: [openbeosnetteam] Drivers... > > > > After looking at the performance it's obvious that while the kernel is > > fsater, it's not a huge jump. I think the main reason is that we're using > > the same drivers and accessing them via open() which probably isn't > ideal... > > I'd like to eventually write our own drivers as this also gives us access > to > > media settings and so on, something I'm not sure we have at present with > the > > drivers we have. > > > > Also I think that having a driver add the data to a queue and then simply > > having the network stack pop it and start processing it will be quicker > than > > having the stack sit on a read() call with all the copying of data that > > involves... In fact I'd like to not bother having the devices publish > > themselves and having the net_stack simply run them as modules! I mean if > we > > have a network/drivers directory and every time the stack starts we simply > > call each module and ask it if it wants to init we'd be in roughly the > same > > situation wouldn't we? > > > > On a related note... with regard to shutdown and restart, hw about we make > > the modules be NOT B_KEEP_LOADED and then when we restart or shutdown the > > core simply calls put_module and the modules can be unloaded. The core I'm > > not sure about, but it would allow us to simply add modules and then > > restart, rescan and start with all the new modules. Thoughts? > > > > Oh, and how does the stack get signalled that the system is closing down > and > > should shutdown? > > > > Anyway, just a heads up. > > > > Night all! > > > > david > > > > > > > > >