> > The thing seems to be mounting kernel error disapeared now as I > remember. That may be the source of problem. I think it should error > and reset kernel IDE driver to use non-DMA mode. > > (So no need for my patch in a sense. This relied on kernel error and > remount.) > > By the way, for boot-loader kernel, do we care about performance gain of > DMA? If not disable them all together for stability.
I agree. Herbert proposed this as the fix and I can verify that DMA is the source of the problem (see below). > Or should we have sleep inserted after mount/umount to stabilize system? It might help, it might not. I think your patch is fine to leave in, it attempts to do something same when hardware misbehaves. And whether or not the installer kernel has DMA off for cdroms, I think it is worth patching cdrom-detect.postinst. I'll file a bug report to that effect. Cheers Vince -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
