On 2014-07-17, Ted Unangst <t...@tedunangst.com> wrote: >> c_skip: 0 >> pciide0:0:0: bus-master DMA error: missing interrupt, status=0x61 >> wd0: transfer error, downgrading to PIO mode 4 >> wd0(pciide0:0:0): using PIO mode 4 >> wd1(pciide0:0:1): using PIO mode 4, Ultra-DMA mode 2 >> wd0c: device timeout reading fsbn 65 (wd0 bn 65; cn 0 tn 1 sn 2), retrying >> wd0: soft error (corrected) >> root on wd0a (511531c0c5c7c075.a) swap on wd0b dump on wd0b >> clock: unknown CMOS layout > > This looks more like pciide controller issues. Either it or the CF > aren't playing nice with each other. But once downgraded to PIO mode, > it appears everything has been slowed down enough to work again.
The early CF spec didn't have a pin for DMA, it was added later. So some machines (and CF adapters) don't have the DMA line on the interface connected to the card. My guess is that one card has DMA enabled and the other doesn't. Here it's the "good" case and it falls back to PIO ok. Some cards don't reset correctly to allow the fallback and in those cases you won't even boot without setting the pciide flag to disable DMA. In a few cases you can use a vendor tool to change the card to not do DMA. But it all boils down to disabling the feature (one way or another) that requires the non-connected line.
