On Saturday 14 May 2005 10:25, Peter Karlsson wrote: > On Fri, 13 May 2005, Peter Gordon wrote: > > My understanding is that the kernel will automagically configure DMA > > as appropriate if you build support for the IDE controller statically, > > but hdparm is needed to initialize DMA stuff if you build your IDE > > controller's driver as a module. I'm not certain though. I tend to build > > everything into the kernel, and I've not needed to use hdparm. > > There's a whole lot more one can do with hdparm. What the kernel _can_ do > is enable dma only. hdparm is used to set other performance enhancing > options. My '/etc/conf.d/hdparm' contains 'hda_args="-d1A1m16u1a64"' which > means: > > -d1 - enables dma for this drive (to ensure dma is set). > -A1 - enables the IDE drive's read-lookahead feature.
is this not deafault? > -m1 - set sector count to 16. This reads 16 sectors per interrupt instead m16 sets to 16 ;) but: is m16 not the default setting? > of one. Some drives run slower with this. > -u1 - set interrupt-unmask for the drive. Can be dangerous with some > drive/controller combinations. Allows the kernel to service other i/o > interrupts, afaicu. I don't know, if there is any combination today, that makes probs, but not using -u is wasting performance. Don't complain about stuttering sound and video-hangs, if you have not tried -u1 > -a64 - set sector count for filesystem read-ahead to 64 sectors. A > cache-mechanism. why? default is 256.. why changing it? > http://gentoo-wiki.com/HOWTO_Use_hdparm_to_improve_IDE_device_performance > yeah this howto does also encourage the usage of -X with is totally superflous or at least really really dangerous, since drive&chip are usually able to figure the best speed out for themselves, forcing something else just increases the chances of probs. >From man hdparm: Use this with extreme caution! Btw, if you want to use dangerous options, why not try -F/--security-freeze? at least this one makes some sense... -- gentoo-user@gentoo.org mailing list