Hello TI!
We're observing a strange behavior in some of our DM368 based board. The problem occurs reproducible on those board while it seems to never occur on others from the same batch.
After a little bit of debugging, I found out that this sequence of register write/read will lockup the DM368 ARM chip:
write 0x01d07018 := 1
read 0x01d10064
So the first is a write into the EMAC RXTEARDOWN register. It requests a teardown of channel 1. The second is a read into AEMIF control regs (it doesn't matter what reg you read). This sequence does appear e.g. when you first access the network in u-boot and afterwards read or write NAND flash.
Note: Teardown of channel 1 is actually wrong, as u-boot uses channel 0 for communication. This seems to be an old bug in the u-boot emac driver (see: http://e2e.ti.com/support/embedded/f/354/p/36786/128519.aspx). But anyway it should not lockup the whole device. Also it seems only channel 1 causes the problem. Channel 0 or 2 work alright.
You can reproduce this easily with these u-boot commands:
u-boot# mw 0x01d07018 1
u-boot# md 0x01d10064 1
On boards that have this problem, u-boot will lock up. It is not possible to contact the chip via JTAG.
Is this a known problem?
Regards,
Andreas.
