On Thu, Dec 02, 2010 at 03:07:49PM +0100, Kevin Wolf wrote: > This looks wrong. Wouldn't werror=stop cause the request to be retried > as a write when the VM is resumed?
Indeed. > But having a copy&paste error gives just about right reason to mention > that after read and write this is the third almost unchanged copy of > this code. Eventually we'll want to refactor this. I've added a patch to refactor the DMA code to the next iteration of the patch series. > While we're at it, do you know why in the eot: case we set > BM_STATUS_INT, but don't actually call ide_set_irq? From what I > understand, those two should always be coupled, but I might be wrong. No idea, sorry.
