I have made a modification to the embedded software code that implements a new kernel module to handle all of the IRLP communication between the software and the hardware.
This changes the way that the parallel port is accessed.
I have been running it on my test node for about 2 weeks now, and it has virtually cut off all the processor load associated with the parallel port access. Before the DTMF, readinput, controller, and imike processes would draw about 10% processor load. Now they draw less than 0.000001% on average.
If you have the "new" flash drive and have been experiencing "crashes" that are not explainable, please drop me a line and I will install the code onto your system. It may be that this code is rolled out to every node eventually, as everything from 2.6.17 kernels and on seem to add a lot of overhead to parallel port calls.
Please send me an email to dcameron@...
, and I will look at what is required to take a stock embedded node (with optional 2Gig Flash) and I will log in and have a look.