Subject: Re: Advice to people writing audio drivers
To: Ignatios Souvatzis <firstname.lastname@example.org>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
Date: 01/25/1999 15:15:24
>MMAP even for non-DMA drivers?
`Me Too'. I've been working on and off on a driver for a non-DMA sound
card (turbochannel, where access to the codec is only possible via a
DSP chip.) I understand its not too dissimilar to how some add-on
audio for NeXTs work. Does it really make sense to require an mmap()
interface for all such hardware? Doesn't this imply trapping
userspace memory reads and writes, and translating them to a
read()/write-style (PIO) interface internally? "Yuck".