home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.linux
- Path: sparky!uunet!mcsun!news.funet.fi!hydra!klaava!torvalds
- From: torvalds@klaava.Helsinki.FI (Linus Torvalds)
- Subject: Re: Floppy DMA, question and patch (?)
- Message-ID: <1992Dec15.112658.13063@klaava.Helsinki.FI>
- Keywords: floppy,dma,patch,kernel,0.98.6,1MB,16MB,unofficial
- Organization: University of Helsinki
- References: <If=DXGK00Vp3MDgVMZ@andrew.cmu.edu>
- Date: Tue, 15 Dec 1992 11:26:58 GMT
- Lines: 22
-
- In article <If=DXGK00Vp3MDgVMZ@andrew.cmu.edu> fl0p+@andrew.cmu.edu (Frank T Lofaro) writes:
- >
- > Is the 1MB DMA restriction in the floppy code necessary, or can
- >386/486's handle 16MB DMA for the floppy? I have this patch (suggested
- >to me in e-mail) that ups the limit to 16MB from 1MB, and it seems to
- >work.
-
- It should work fine. The only reason for the 1MB (as opposed to 16MB)
- limit was that I originally thought of making only the low 1MB
- identity-mapped, and anything above it could be moved around with the
- pager hardware, so DMA wouldn't work on those without some address
- calculations.
-
- As it turned out, I never did anything like this, and I currently
- guarantee that all kernel memory is identity-mapped up to the limit of
- physical RAM (it's easier on the memory manager, as well as on devices
- like some SCSI cards which use DMA). So changing the limit to 16MB
- should work fine. Not that it probably matters very much: 99% of the
- floppy IO reads go through the track buffer, and writes aren't exactly
- timing-critical in most circumstances..
-
- Linus
-