On 10 Sep 2013 17:01, "Matthias Scheler" <tron%zhadum.org.uk@localhost> wrote:
>
> On Tue, Sep 10, 2013 at 09:42:40AM +0000, Nick LaForge wrote:
> > What is puzzling is: if the performance loss is not likely to be related to
> > any filesystem implementation / option, one logical hypothesis is that the
> > partitions are not aligned to 4k blocks, but this has already been
> > falsified by booting Linux and observing better performance for the exact
> > same ext2 partition.
>
> I'm not sure that NetBSD's ext2 implementation has received as much
> performance tuning as the ffs file-system. So you are better of comparing
> NetBSD's ffs to a suitable Linux file-system e.g. ext3 or ext4.
>
> > I can only think that the loss of performance is related to the NetBSD
> > kernel's idea of the entire disk (/dev/wd0). E.g., dma is not being
> > utilized, or maximum sata data rates are not taken advantage of, etc. This
> > is also strange, since the machine is not new (it is from 2009).
>
> The kernel message you provided suggests that DMA is used:
>
> wd0(piixide0:0:0): using PIO mode 4, Ultra-DMA mode 4 (Ultra/66) (using DMA)
>
> You could however try to change the BIOS settings and configure the
> SATA ports to use AHCI mode.
If Linux is only giving 25MB/s this suggests AHCI is not on, and the performance will be awful.