Return to BSD News archive
Newsgroups: comp.os.386bsd.development Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!haven.umd.edu!uunet!pipex!uknet!mcsun!sun4nl!relay.philips.nl!cnplss5!rooij From: rooij@mozart.cft.philips.nl (Guido van Rooij) Subject: Re: kernel hacking tips Message-ID: <1993May12.102918.10890@cnplss5.cnps.philips.nl> Sender: news@cnplss5.cnps.philips.nl (USENET News System) Nntp-Posting-Host: mozart Organization: Philips Communications & Processing Services, Eindhoven References: <PC123.93May11105430@bootes.cus.cam.ac.uk> <1so4eaINNibt@fstgds01.tu-graz.ac.at> <1993May11.192459.1618@ucsu.Colorado.EDU> Date: Wed, 12 May 1993 10:29:18 GMT Lines: 21 galbrait@rintintin.Colorado.EDU (GALBRAITH JOHN) writes: >Has anybody tried to cheat and write a driver that does nothing but give >you access to an outb() and inb() instruction? This might speed development >for a big project. The new driver gets written and debugged before it is ever >added to the kernel so you don't have to constantly reboot. Or what about >doing an inline assembly call to outb with no kernel support at all? I posted a patch some time ago that fixed incorrect behaviour regarding inb() and outb() (anyone could execute these instaructions, resulting in a failure only half of the time!!. I also adapted the kernel a bit so that opening /dev/mem gives you permission to execute inb's and outb's. It was said that the kernel already had this functionality but that ism *not* true. >john >galbrait@rintintin.colorado.edu -Guido