Hi,
We'are now developing a DM648 -- DM8168 communication scenario. And we've been bugged by this 'interrupt-go-astray' issue for days. The scenario is we interfaced DM648 to DM8168 through the latter's PCIe interface and a XIO2001 PCIe-PCI bridge. The OS software on DM8168 is Linux 2.6.37, TI issue 04.00.00.12. The PCI device is logically recognized and utilized by a home-made kernel driver. We mapping a block of DM648-side DDR memory to DM8168 PCI window and directly read/write it and everything seemed fine.
Here's the problem, when we issued interrupts from DM8168 to DM648 by setting the DSPINT bit in PCISTATSET register after data transmission to DM648. We sometimes lost this interrupt on DM648 side. Currently we only issue interrupts no more frequently than 1 per-second.
And we've tested this interrupt generation method solely by sending 5000 interrupts to DM648. Not a single one got lost.
Could anyone enlighten us on this issue? Thank you very much.