On 29/03/2017 02:28, Steve Rutherford wrote: > Not sure this matters, but this changes behavior slightly. This reorders > validation of the length and and validation of the operation. This would change behavior if for example: - KVM registered port 0x22 - the guest wrote 4 bytes to port 0x1f Before, picdev_write would return -EOPNOTSUPP and KVM would invoke the handler for port 0x22; now it would return 0. Since the PIT ports are far from anything else registered by the kernel, this should not matter. Paolo
