> -----Original Message-----
> From: Mike A. Harris [mailto:[EMAIL PROTECTED]
> Sent: Tue, March 11, 2003 9:54 AM
> To: [EMAIL PROTECTED]
> Subject: Re: hdparm settings
> 
> 
> On Mon, 10 Mar 2003, Jesse Keating wrote:
> 
> >> # grep USE_DMA /etc/sysconfig/harddisks
> >>  USE_DMA=1
> >>
> >> # hdparm -t /dev/hdb
> >> /dev/hdb:
> >>  Timing buffered disk reads:  64 MB in  1.18 seconds = 54.07 MB/sec
> >
> >This is different.  It's setting an approved config file, 
> for Red Hat init 
> >scripts to parse, and set hdparm options.  It is not using 
> hdparm directly on 
> >the disk.  Some hdparm settings you can use, w/ little fear of data 
> >corruption.  Some you can't.  This is most likely what Arjan 
> was referring 
> >to.  Arjan's statements were pretty vague, so take them as 
> that unless he 
> >clarifies a bit.
> 
> I can clarify Arjan's statements a bit.  Basically, in all modern
> kernels, the kernel autodetects the hard disk's capabilities
> itself, and sets the drive for the best performance
> _automatically_ itself without any end user intervention.  If for 
> any reason, a hard disk setting is _not_ enabled by the kernel, 
> it is because the kernel knows that that setting is not safe to 
> use.  The kernel contains various blacklists/whitelists of known 
> bad/good hardware combinations, known motherboard chipset bugs, 
> and whatnot, and it sets the drive to enable the fastest transfer 
> rates for your hardware combination that is also "safe" to use.

Actually, I have one system that quite often freezes when DMA is enabled for
the harddisk. Every time I install linux I have to disable DMA manually on
it.

> 
> If someone then goes and plays with hdparm settings, they may in 
> fact be force-overriding the kernel's safety mechanisms, and 
> enabling some setting that triggers a flaw in the drive, the 
> motherboard chipset, or some combination of the two, and that may 
> cause data corruption.
> 
> Second guessing the hard disk settings that modern kernels 
> default the drives to is Russian Roulette.  Some settings might 
> turn out to be harmless (or perhaps the kernel just ignores them 
> and you _think_ the change did something), while other settings 
> perhaps do change something (but have zero real effect on speed).
> 
> Short Answer: Trust the kernel to do the right thing.
> 
> 
> -- 
> Mike A. Harris     ftp://people.redhat.com/mharris
> OS Systems Engineer - XFree86 maintainer - Red Hat
> 
> 
> 
> -- 
> Phoebe-list mailing list
> [EMAIL PROTECTED]
> https://listman.redhat.com/mailman/listinfo/phoebe-list
> 



-- 
Phoebe-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/phoebe-list

Reply via email to