On Sunday 12 August 2001 09:19 pm, Terry C wrote:
> I remember civileme saying that hard drives were
> limited to ata33 in Mandrake 8.0 because of some
> problem in the kernel with the Via chipset. Does
> anyone know if this has been fixed in the 2.4.7
> kernel?
Just to be clear, the VIA-IDE bug isn't a kernel problem, it affects
all OS's. It's a hardware bug. All hardware, specially chipsets have
erratta. Most just go unnoticed by the the user. Those that do surface
usually get a software or firmware fix (ie, bios upgrade). The VIA chip
has a rarely experienced problem with cross IDE large file (>100mb)
tranfers. I believe most of the hype surrounding the VIA problem is
just that... hyperbole. That said ...
Excerpt from 'dmesg':
..
Linux version 2.4.7-13mdk ([EMAIL PROTECTED]) (gcc version
egcs-2.91.66 19990314/Linux (egcs-1.1.2 release / Linux-Mandrake 8.1))
#1 Thu Aug 9 16:03:41 CEST 2001
|
|
Applying VIA southbridge workaround.
|
|
..
(I have an AMD approved Soyo k7vta pro VIA kt133a chipset Mboard)
> If so, has anyone compiled the 2.4.7 kernel
> (mandrake version from cooker) and found that they
> could use ata100?
I compiled a 2.4.7-13dk from source which was sucessful, made a
initrd for it, edited lilo, and such ... but it panic'd on boot. Didn't
bother with it. I d/l'd the ready made kernel-2.4.7-13mdk an installed
it. I was hoping to fix 2.4.5's aversion to supermount, but it's still
broken with 2.4.7-13, MOF it's not even there. So I went back to my
good 'ol tried an true
alias mcd="mount /dev/cdrom"
alias ucd="umount /dev/cdrom"
in bashrc
As is my habit with a new kernel, I took a look at 'dmesg' (above)
and for the first time saw the 'Applying VIA southbridge workaround.'
line. So next thing I did was a 'hdparm -i' on my HDD's and saw they
were now *udma5 (ata/100) enabled.
<now for my usual rant ;>
The drives were gettin about 22 to 24mb/sec on ata/33 (udma2, hdparm
-t), now they get 28 to 30mb with ata/100 .... BFD. ata/66 and /100
are nothin but Wintel hype. Real world tranfers are not improved, MOF,
I think they're a touch slower. hdparm is not a good bench, it
measures burst rates, which have little to do with reality. I
regularly load a dir with 37,000 .jpg files in it that average 100k
each. Now that my HDD's are running ata/100, it takes a few seconds
longer to load that dir than when they were on ata/33.
FWIW, I'd advise those using 2.4.5 to upgrade to 2.4.7. For those
usin a stock 8.0 kernel (2.4.3), leave well enough alone unless you
believe you're experiencing the VIA-IDE bug. You're probly not, but at
least you can't blame it anymore ;)
I wasn't, I'm just an upgrade-aholic ;>
--
Tom Brinkman Galveston Bay