Return to BSD News archive
Newsgroups: comp.os.386bsd.questions Path: sserve!newshost.anu.edu.au!munnari.oz.au!network.ucsd.edu!usc!howland.reston.ans.net!agate!headwall.Stanford.EDU!nntp.Stanford.EDU!yergeau From: yergeau@leland.Stanford.EDU (Dan Yergeau) Subject: Re: > 16 MB of RAM (on an ISA box) Message-ID: <1993Jul3.072649.18439@leland.Stanford.EDU> Sender: news@leland.Stanford.EDU (Mr News) Organization: DSG, Stanford University, CA 94305, USA References: <210g8n$3pr@pdq.coe.montana.edu> <1993Jul2.164901.22217@leland.stanford.edu> <2121b9$73v@pdq.coe.montana.edu> Date: Sat, 3 Jul 93 07:26:49 GMT Lines: 38 Argh! Much of my confusion came from thinking that the wd driver did DMA transfers (cuz, if that worked OK, the all DMA transfers should work OK). I should have looked at the source. The wd driver uses the cpu to transfer data to/from the controller via outsw/insw. It looks like, in addition to SCSI controllers, the floppy disk drives, QIC-02/36 tape drive, and some ethernet cards use DMA and should eventually cause ugly problems in a ISA box that has more than 16 MB. The key here seems to be the use of "drq" in the config line for the device/controller. >>|> When you can't run more than 16MB of memory is when >>|> >>|> 1) You have an ISA box - limits you to 24 data lines == 16MB >> >>This is not correct. The ISA bus only has 24 address lines (and 16 >>data lines), but many ISA motherboards support up to 32 MB or 64 MB of >>addressible memory. > >You're not listening. Yes, I am listening, but what you wrote was misleading. Specifically, you left out the key word "and" linking 1 (ISA bus) to 2 (DMA transfers). Apparently you missed the whole of my original post which made reference to preventing bogus DMA transfers (isa_dmarangecheck). I understood the problems of DMA transfers and the ISA bus, but I didn't know if the proper routing (ala bounce buffers) was completely implemented. Thanks for the clarification, Dan -- Dan Yergeau He's dead Jim. Take his phaser yergeau@gloworm.Stanford.EDU and I'll get his wallet.