> descriptors (BDs) from the CPM with the OV bit set (bit 14 of the RxBD > status/control field, "Overrun. Set when a receiver overrun > occurs during > reception").
> - SDRAM setup. If ORx(BIH)=1 makes the problem go away, your burst setup or the /TA pullup is broken. NB: Neither the SCC > nor the SMC do bursts. I checked my SDRAM setup. SDRAM is the only device configured with UPMA and it's on cs1, so I checked UPMA, OR1, BR1, MAMR configuration along with MCR/MAR process for issuing precharge, refresh, and SDRAM mode register config per the datasheet for the parts. This board is a few years old, so I wasn't able to find the datasheet the original timings were made from - instead I looked at the datasheet for the parts we're using now, and (since the parts we're using now are faster) our memory timings are somewhat conservative. I did try disabling bursts, with no luck. Not sure what "NB" means. > - Did you by accident happen to enable any IDMA lines? No > - Did you happen to misconfigure RCCR(DRxM)? RCCR is all zero's except RCCR[ERAM]=b11 since I have the SCC/SMC microcode patch. Even though IDMA is disabled, I have tried setting RCCR[DRPQ]=b10 to make IDMA requests have the lowest priority (didn't help/hurt). The DRxM bits are edge sensitive. > - Did you happen to misconfigure the SDCR? SDCR[FRZ] = 0 (ignore the freeze signal) SDCR[FAID] = 0 (highest priority default, although I'm not using FEC). SDCR[RAID] = b01 (priority level 5) >One other thing. If you have an old BDM debugger *and* if the problem occurs *only* with the BDM debugger hooked up, you may >want to check for an update or replacement. Old debuggers sometimes used a method to do BDM which could stall the DMAs in >the machine and cause this kind of overrun/underrun effect. Hmm...interesting. How old? Does this happen when actively debugging (setting breakpoints, etc) or just running with no breakpoints set? I have two different debuggers, a OCD raven wiggler and a BDM module for my HP16702A logic analyzer. I have to run different debugger software on each for licensing reasons (SingleStep 7.5 on the raven and GHS MULTI4.0.5 on the HP). The problem occurs on both. Heinz
