OK, disk changed - thanks guys.

I did try to force writing things on the old one, but that made the CPU
burn and the machine to not respond. BTW, I found NetBSD to not deal well
with such error. Perhaps it is just because I have an Ultra5 with an IDE
drive but I noticed the wrong disque sectors seemed to host data for
net-snmp. So since I did a "pkg_delete net-snmp" the system started to
show the error messages and became really hard to manage (no more ssh
session possible, and Ctrl-C took a long time - lets says 2min - to stop
the pkg_delete process and give the prompt back).

I remember issues like this on an Archos external USB drive where:
- Windows just goes blue screen
- Linux froze the shell where the access to the disk was started (cd, cp,
or such)
- NetBSD dropped thousand of messages

I'm not sure which is the "better" behaviour, but how comes the OS became
so less reactive when error deals with non-critical data - it's not the
kernel trying to load or the mount command at boot time, see...

> Joel CARNAT wrote:
>> I see the following errors in dmesg:
>>
>> wd0: (uncorrectable data error)
>> wd0a: error reading fsbn 2189695 of 2189664-2189695 (wd0 bn 2189695; cn
>> 2172 tn 5 sn 4), retrying
>> wd0: (uncorrectable data error)
>> wd0a: error reading fsbn 2189695 of 2189664-2189695 (wd0 bn 2189695; cn
>> 2172 tn 5 sn 4)wd0: (uncorrectable data error)
>>
>> are they hardware or software error ?
>> that is, must I replace the hard-drive or should I only reformat the
>> disk
>> and mkfs ?
>>
>> is there any fsck-like command that would mark bad sectors so that there
>> are not used by the system (with or without reformatting) ?
>>
>> TIA,
>>    Jo
>>
>>
>>
> Thats a Hardware error. I recommend you to use a new Disk drive.
>