Aug 23, 11:30 am
Installed a new kernel with one small difference in the PC104 interrupt handling. If the PC104 interrupt hander is called, and it sees no bits set in the pending value, then it goes ahead and call handlers for all unmasked PC104 interrupts (which in manitou's case is just one, IRQ 3).
This logic was also in the patched 2.6.16 kernel. It counted these "spurious" interrupts and from time to time, complained about them. The new code doesn't log any complaints. The intcount script could be used to see if it is happening frequently.
Since re-installing the data system last week, with the new kernel, I've seen a few timeouts where all data from the emerald cards cease. This patch may help that.