[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Sheflug] UDMA-33 and cables.+hdparm





On Sun, 12 Nov 2000, Barrie Bremner wrote:

> 
> 
> For new drives try # hdparm -A1 -d1 -X66 -c3 -u1 /dev/hda
> 
> -X66 sets UDMA mode 2, d1 turns on DMA, c3 sets 32bit support, u1
> unmasks IRQs apparently (don't ask me..it might make a small difference)

-u1 allows "bottom half" IDE interrupt processing to be done with interrupts
enabled. This was claimed in early days to be unsafe on early IDE hardware.
Alan Cox had seen this cause disk corruption. 

It does not make much difference to measured disk performance, indeed it can
make it appear a tad worse. BUT the overall system performance improves alot.
On slower CPUs it means alot less dropped serial line characters when using
PPP. 

I'd always taken the warnings on use of -u1 with a pinch of salt, and
have been running with it set for a year. But I've recently started expanding
my local house network, and 1 week ago while I had PPP running over serial
line to my ISP and I was transfering stuff around on the ethernet (NE2000 ISA
card, Kernel is 2.0.38) I got a bunch of disk corruptions for the first time
ever. I have disabled -u1 and have not had a recurrence. I've still to
investigate further, but need a sacrifical system - not doing it with main
machine!

> and A1 sets read-lookahead on.
> 
>  Check man hdparm for the lowdown, particularly on setting the various
> (U)DMA and PIO modes.

Jim

---------------------------------------------------------------------
Sheffield Linux User's Group - http://www.sheflug.co.uk
To unsubscribe from this list send mail to
- <sheflug-request [at] vuw.ac.nz> - with the word 
 "unsubscribe" in the body of the message. 

  GNU the choice of a complete generation.