Re: DMA and corruption on WD hd

[Andre Hedrick <andre@linux-ide.org>]

On Mon, 14 May 2001, Jay Link wrote:

> Date: Mon, 14 May 2001 07:37:01 -0500 (CDT)
> From: Jay Link <jlink@ilbbs.com>
> To: Tom <freyason@yahoo.com>
> Cc: Matan Ziv-Av <matan@svgalib.org>, luci-discuss@luci.org,
     andre@linux-ide.org
> Subject: Re: DMA and corruption on WD hd
> 
> 
> Thanks to everyone who answered my question re: hdparm and corruption.
> Again, DMA was killing me, but without it my WD drive was so slow as to be
> virtually useless.
> 
> I finally found the following on O'Reilly's web site, and it appears to
> have done the trick (combining speed & safety):
> 
> hdparm -c3 -m16 -d0
> 
> c = 32-bit transfers

Bawhahaha (a rotfl laugh)

32-bit transfers or read/write long have been obsoleted
32-bit transfers == 2 16-bit transfers or multimode == 2

> m = multisector I/O

Sure this will work.....but.....

> d0 = DMA *OFF*

No this is silly and "O'Reilly's web site" is DEAD wrong.
(unless you have a very old drive that is pre ATA-3)
You lose the iCRC auto correction.

> 
> Of course, the -k1 flag is then necessary in order for the drive to
> preserve its settings through a soft reset.
> 
> I understand (through Tom [thanks!]) that the 2.4.x series of kernels make
> hdparm obsolete, but I'm using 2.2 and I didn't want to introduce yet
> ANOTHER variable into the mix.

Thank somebody for this piece of info finally getting out.

> Thanks again, everyone.
> 
> 
> -Jay Link
> 
> 
> Note to Danny Sauer: What would you recommend over WD?
> 

Andre Hedrick
Linux ATA Development


-
To unsubscribe, send email to majordomo@luci.org with
"unsubscribe luci-discuss" in the body.