On Tue, Oct 10, 2006 at 03:33:45PM -0400, Dan Malek wrote: > > Many people, including myself, have spent way too many > hours trying to make this device interface work. There > are some errata associated with it, along with some challenging > design problems. I have only been able to make it work > in one instance, with a highly custom RTLinux driver, > properly matched FIFOs and DMA, and running TCP/IP > over the link. It wasn't 100% reliable, but the TCP retries > made it appear that way to the application (with tolerable > delays). May I have your setup sequence? At least I'd like to se some bus activity... > The device interface just requires too much babysitting > by the CPU to function, and the Linux interrupts > have too much latency to guarantee the CPU can do > what is necessary in a timely fashion. The same is > true of not using DMA. If you choose not to use the > DMA for data transfer, the CPU just can't respond > quickly enough to the interface state changes > unless you just spend all of your time polling the > interface. Ok, but as first step I'd like having some functionality with FIFOs. > IMHO, I wouldn't waste much time on this, but > Good Luck if you choose to do so :-) Thanks a lot! Rodolfo -- GNU/Linux Solutions e-mail: giometti@xxxxxxxxxxxx Linux Device Driver giometti@xxxxxxxxx Embedded Systems giometti@xxxxxxxx UNIX programming phone: +39 349 2432127
