[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: NetBSD 4.0, amd64 SMP, weird corruption issues
Mark Cullen wrote:
Christos Zoulas wrote:
In article <g50koo$fp0$1%ger.gmane.org@localhost>,
Mark Cullen <mark.r.cullen%gmail.com@localhost> wrote:
I'm in the process of updating my small home server with a nice new
machine. I've downloaded the amd64 version of NetBSD 4.0 to install.
Here's the more important machine specs:
Gigabyte G31M-S2l with latest BIOS
2 x 1024MB Corsair PC6400 5-5-5-12
Seasonic 330W PSU
It's installed fine, and everything seems to be working lovely. Even
the onboard Realtek NIC. I've set up one of the RAID-1's with
RAIDframe and that's looking good. I had just got the latest pkgsrc
and was trying to install some bits and bobs, and that's where it
appears to have gone downhill.
Here's what happens if I try and update my pkgsrc tree. The following
screenshot had just started to update files, and was listing changes
If I right click the PuTTY title bar and 'reset terminal', it springs
back to life again and everything is readable as normal! However,
some Makefiles and such in pkgsrc are garbled binary too. It seems
like some kind of network corruption issue to me? I've tried
switching to a good old fashioned reliable Intel PCI NIC, and have
the same results. In fact, that screen shot was captured when using
the Intel NIC.
Any thoughts? Have there been any known issues with network
Looks like something is typing CTRL-O [enter ALT-GR1]. You can undo
this by echo CTRL-N to your terminal. Can you capture the output of cvs
update to the file?
I tried just now, and ran it once to make sure it was still doing it. I
was then writing some stuff in a script in vi. I exited vi, and the
shell prompt had become garbled. Perhaps CVS is triggering the bug
somehow, I don't know.
I am not sure CVS just outputting a CTRL-O somewhere would be causing
files to be corrupt either? Some Makefiles that CVS has downloaded are
Hm. This install is on a temporary Deathstar disk. Maybe it's just that
on it's way out, written some junk to the disk, CVS reads junk back and
it all goes bananas?
I've got a Seagate on RMA that I am intending to do the proper install
on. I was just doing this to set things up and make sure it all works
I'll let you know if a new disk fixes things, but if anyone has any
other ideas in the mean time, or needs more info, just ask :)
Ok, a small update.
First I deleted pkgsrc from the Deathstar disk and tried to fetch it
again. That seemed to go OK, though I didn't finish the fetch. With the
CVS update the garbage happened pretty quickly though.
I then symlinked pkgsrc to the raid1 disks and have just now done a CVS
fetch / update, and that issues appears to have disappeared.
I think, or should I say hope, that it's just the Deathstar disk on the
way out. I don't even know if it's possible for a hard disk to "pretend"
to write to sectors, and the data sort of just doesn't actually get
written. Complete guesswork. If it's that, I'm not fussed because I
won't be using that disk for long anyway. If it's not that and some
other very rare bug.... not so good!
I'll continue testing, and if it happens while pkgsrc is on the raid1 I
will post back.
Mark Cullen <mark.r.cullen%gmail.com@localhost>
Main Index |
Thread Index |