On Wed, 2003-05-14 at 17:55, Steve Lord wrote:
> On Wed, 2003-05-14 at 10:50, Olaf Frączyk wrote:
> > On Wed, 2003-05-14 at 17:21, Walt H wrote:
> >
> > >
> > > Do you have dma enabled on your IDE disks? I can't think of any reason
> > > off the top of my head why a cat from hda -> hdc should be affected by
> > > XFS, as you aren't using the filesystem for that - it's just reading the
> > > raw device.
> > Yes, I know. I just don't know if XFS code changes some buffering etc.
> > Maybe with vanilla kernel it would be the same. I just have no easy way
> > to test it.
> > Make sure dma is enabled on hda and hdc and see if that
> > > helps: hdparm -d1 /dev/hda /dev/hdc
> > >
> > Both drives have dma enabled.
> > And they are set to UDMA-100 also.
> >
> > I'm just curious where from comes this terrible latency.
> >
> > Regards,
> >
> > Olaf
> >
>
> XFS should not affect this in any way, I am sure the unmodified kernel
> would behave the same way for you.
OK, I'll try to repartition my disks to get some free space to install a
vanilla kernel.
BTW, do you think that preemptible or low-latency would help here?
If yes, can these patches be safely merged with XFS?
Regards,
Olaf
|