Re: MD5 failing on optical media

[Andrew Smallshaw <andrews%sdf.lonestar.org@localhost>]

On Mon, Aug 05, 2013 at 11:22:14AM +0200, Martin Husemann wrote:
> On Sun, Aug 04, 2013 at 09:57:47PM +0100, Andrew Smallshaw wrote:
> > cd0 at atapibus0 drive 0: <TSSTcorp CDDVDW SH-S223C, R4136GBZ996738, SB04> 
> > cdrom removable
> > cd0: 32-bit data port
> > cd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 5 (Ultra/100)
> > cd0(viaide1:1:0): using PIO mode 4, Ultra-DMA mode 5 (Ultra/100) (using DMA)
> 
> Can you reproduce it with an image file and a vnd?
> If so, could you make available an image? How did you create the file system?

Sorry, I meant to get back to this earlier but it's the kind of
thing that takes a little time to test repatedly since reading an
entire DVD image always takes at least 10 minutes - it doesn't show
up on small test files.

The problem doesn't show up on a vnd image and it seems to be
independent of filesystem on real media since it shows up on both
single-session IS9660 and read-write UDF filesystems.  It's
intermittent too - you can run the same tests on the same disc back
to back and get different errors each time or no errors at all.
I played with it a lot last Tuesday or Wednesday night and couldn't
get any errors all night.

I haven't completely ruled out hardware gremlins and that did make
me thing there could be some form of temperature dependence involved,
since that was a much cooler night compared to the mini-heatwave
of the past two or three weeks.  That system is generally pretty
cool since the components are almost lost inside a physically large
and well cooled 4U rack case.  The only temp I have handy in normal
operation is the CPU temp from envstat which is regularly under
30C but in the mini heatwave we've had recently it's been getting
up to the mid 40s - higher than normal for that system but still
decidely average.  

However, the counterargument to that is that another night I was
getting errors immediately on cold booting it which seemed to go
away when it warmed up before coming back later.  I've kept an eye
out for that but twenty or more tests haven't shown any meaningful
pattern.  I did think of trying another machine but I then realised
that would have been a lot of work - either no optical drives or
don't run NetBSD.  I've picked up a USB DVD drive to try to get
some more data and thus far that seems flawless, but that is
painfully slow even compared to the first drive so I've not done
as much testing with it.

It'd be easy to brush this aside as a hardware fault if it wasn't
the for the actual problem - copies ALWAYS work, working on data
in place may not.  That intuitively sounds like some software fault
to me and probably in one of the drivers - drive, controller or
bus.  I'm still testing and trying to find some common factors but
as I've said it takes time.  Does anyone have any more suggestions
or other input in the mean time?

-- 
Andrew Smallshaw
andrews%sdf.lonestar.org@localhost