Re: [SLUG] loss of DMA

From: Eben King (
Date: Tue Apr 11 2006 - 16:51:12 EDT

On Tue, 11 Apr 2006, Ian C. Blenke wrote:

> Eben King wrote:
>> On Sun, 9 Apr 2006, Eben King wrote:
>>> In fact, I ran "dd if=/dev/hda bs=1k of=/dev/null" and didn't get an error
>>> (bs=1k was where I saw the error today).
>> Didn't get an error today, with bs=2k, so maybe 1k writes tickle a bug in
>> the IDE driver.
> Have you opened the case or otherwise jarred the machine in a way that could
> cause a cable to dislodge?

It's a full-size tower case, so there's plenty of room to work around
things. I did examine the cable (while the machine was running, so I
couldn't unplug it) and it _looked_ seated. I'll make sure it is, tonight.

> Also: Are you using a "shredded ribbon cable" in a plastic conduit?

It's not round, but (American) football-shaped, so "sorta". Split 2 or 3
times lengthwise.

> Try a normal cable instead and see if the errors don't go away.

I haven't had any similar errors before or since, so I'm not sure I'd be
able to detect a change. I like the "1k writes" hypothesis, since I didn't
change any hardware and the errors went away.

Synopsis: I wanted to find the most efficient blocksize for drive -> drive
transfers (early results say 2^6 or 2^7 kB). Since I already had a script
that duplicated a 120 GB drive every day, I just made it use a different bs=
parameter every day (involving `date +%j` and mod).

-eben    ebQenW1@EtaRmpTabYayU.rIr.OcoPm

Logic is a systematic method of coming to the wrong conclusion with confidence. ----------------------------------------------------------------------- This list is provided as an unmoderated internet service by Networked Knowledge Systems (NKS). Views and opinions expressed in messages posted are those of the author and do not necessarily reflect the official policy or position of NKS or any of its employees.

This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 20:39:01 EDT