> > it sounds like you have either a bad drive or a cable problem.  i
> > would replace the sata cable first.  it is also possible that you are
> > having a problem with power management.  but i don't know how to tell
> > if you're using the ide interface.
> >   
>     I have Windows XP and Linux on this same disk. It can't be a cable 
> problem. I doubt it is a disk problem either because I/O errors during 
> file copy seem to be on random disk bloks. Sometimes it copies files 
> until 2%, others it gets to 13% (I reformat fossil partition every time 
> I try reinstalling Plan 9)...

i wouldn't be so quick to rule out a cabling problem.  i have seen failing
on random blocks happen in three senarios.  1) cabling problems. 2)
power management problems (the error is actually due to pm kicking in
when you're trying to do i/o) and 3) sata protcol errors.  unfortunately,
there are some compatability problems between some drive firmware and
some hbas.  these problems can depend on the programming interface.
your hba has three different programming interfaces (a) ide, (b) old
intel sata (c) ahci.  i believe (but can't confirm)  that all three modes are
driven by firmware in the southbridge.

so, it would be useful to know what programming interface are linux & 
windows using and what disk throughput are you getting.  from linux
it would be good to get timing/throughput for
        dd if=/dev/$harddrive of=/dev/null bs=512k count=10000
it could be that there is some wrinkle of ahci that i have missed.
if you could send me the exact output, that would be useful.  you
may want to enable extra debugging by
        echo ahci debug>/dev/sdctl

if you have a way to compile a kernel, i left my development version
of sdiahci.c on sources
        /n/sources/contrib/quanstro/sdiahci.c

>     AHCI BIOS detects hard disk on port 0 and DVD recorder on port 1. 
> And it is consistent with 9load messages regarding drive reset. Should I 
> be able to access DVD drive on sdF0?
> > - erik
> >

sd gives each controller a letter, not each port.  so port 1 would be sdE1.

for example, i have a machine with an amd sb600 with 4 ports and a
marvell 88sx...and with a marvell 88sx so i have sdE0-sdE4 and
sdF0-sdF7.

so it would be useful to have 1) the contents of /dev/sdE0/ctl, 2) the
exact error messages 3) the programming interface used by linux
and the 4) dd timing.  you might want to send that to me off list.

- erik

Reply via email to