In case anyone is interested in running linux on the net5501, here are
my initial results. I use a sandisk ultra-3 2G card (SanDisk
SDCFX3-2048).
Debian etch installs without problems. I compiled then 2.6.21.5 and I
have a few issues, not critical I would say:
- the compact flash card is not detected as dma-capable, and forcing
it doesn't work; the ide driver reports some errors in dmesg, see
below;
- the machine type as reported by "uname -m" is i586, so the
i686-optimised libc is not being used. From the AMD docs, I thought
the geode lx is supposed to implement all the ppro instruction set
(/proc/cpuinfo reports cmov), but who knows...
- I didn't find a driver which supports the watchdog, so for now I run
without it.
On the network performance side, things are looking really good (at
least compared to 4801):
- desktop -> 5501, iperf: 95.9 MBits/s, cpu ~45% idle
- 5501 -> desktop, 94.8 MBits/s, cpu: around 40% (+/- 2%)idle bridging:
- (desktop -> 5501 -> laptop), iperf, > 90MBits/s, cpu ~95% idle
- bi-directional-line gets around 110MBits/s total (65Mb/s+45MB/s), I
suspect the laptop has issues (cpu usage : 100%) but the 5501 also
has 100% CPU usage
Anyway, this is only quick testing, but I'm glad that the 5501 can fill
fast ethernet (with MTU-sized packets), something which (in my
experience) the 4801 wasn't able to do. I'll test routing performance
next, to see how that works.
For the no dma part, here is the relevant section of the dmesg:
Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
AMD5536: IDE controller at PCI slot 0000:00:14.2
AMD5536: chipset revision 1
AMD5536: not 100% native mode: will probe irqs later
AMD5536: 0000:00:14.2 (rev 01) UDMA100 controller
AMD5536: neither IDE port enabled (BIOS)
Probing IDE interface ide0...
hda: SanDisk SDCFX3-2048, CFA DISK drive
Probing IDE interface ide1...
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
hda: max request size: 128KiB
hda: 4001760 sectors (2048 MB) w/1KiB Cache, CHS=3970/16/63
hda: hda1 hda2 hda3 < hda5 hda6 hda7 > hda4
Tuning with hdparm (-m4, -u1, -c1, -X mdmda2, but node that -d1 doesn't
work) increases the speed from 1.8 only to 2.1 MB/s, which is very low
(this card should do a maximum of 20MB/s). The part that seems strange
to me is the "neither IDE port enabled (BIOS)".
regards,
iustin
_______________________________________________
Soekris-tech mailing list
[email protected]
http://lists.soekris.com/mailman/listinfo/soekris-tech
