On 2017-07-13 12:35 PM, rpjday [ at ] crashcourse [ dot ] ca wrote: > it's been a while since i've looked at this, so forgive me for asking what > i *know* is a simple question. looking at user code that wants to access memory > associated with a physical device and, since it's a physical device, I/O needs > to disable caching. Given your device driver has to disable the VMM for the physical address, shouldn't it be disabling caching for that address range as well? It would be a very wrong design for a userspace process to fool around with how the kernel maps memory onto physical addresses. Then again, I'd argue it's completely wrong for a device driver to have a userspace process have any awareness of physical addresses whatsoever, so there you go. -- Stephen M. Webb <stephen [ dot ] webb [ at ] bregmasoft [ dot ] ca>