A weird one...

Usually I use xcdroast (0.98alpha10) as a front end for copying CDs, and for audio CDs this has always worked fine, with all types of blank media at all write speeds. However every time I try to write a pure data CD to create a backup cdrecord (v1.11a28) freezes and takes the whole system with it. The screen, mouse, keyboard, everything. Ctrl-alt-everything is dead, no option but the reset button.

I used xcdroast to create an image, then originally xcdroast to drive cdrecord. I have tried many different options, DAO, TAO, multisession etc. Perhaps I missed something. I did update the xcdroast, cdrtools & cdrecord packages myself so perhaps I screwed something up if there are hidden version conflicts.

cdrecord from the command line gives the same results. The command line used was:

cdrecord -v speed=8 dev=0,1,0 -data track.img

cdrecord -scanbus shows the cd writer (Yamaha CDRW3200) & ASUS DVD correctly detected.

Like I said at the top, audio CDs can be written fine....!? Help!

Have you tried a lower speed? Try speed=4, if that doesn't work, try speed=2 or =1.

Are there any errors associated with the lockup (from the command line)?

try the -V option as well as the -v option. Maybe even -VVV -vvv

Are you sure there isn't a problem with your track.img? Try mounting it
with the following command:

mount -o loop track.img /mnt
You'll need loopback support in your kernel, either as a module or built
in, try the mount command and if it gives you an error concerning no
loopback support or somesuch, you'll need to either load the module or
build the module, then load it.
You can build modules with out rebuilding your kernel just by skipping
the "make bzImage" portion of the kernel rebuild (after you've selected
loop as a module).

Hi Moses,

Thanks for the mount tip - seems fine. The contents are as they should be.

Unfortunately there is no debug information issued anywhere, either in the system logs or by cdrecord in its own logs - everything just stops where it is. The general symptoms are: The FIFO buffer fills to 97%, then a pause, then it fills 100% then writing starts, gets usually to 5% then the crash. The -V gave me the last SCSI transfer at which the process stopped - looked to be the same as the others but up to now I haven't managed to save it.

The crash is completely reproducible but with no debug information it could take some time to understand! At the moment I suspect some problem on the IDE bus, but why this should work writing an audio CD from a track list on the hard disk I can't explain. As I said I can write an audio CD at 24x with no problems (it doesn't sound too good, but that's another story).

At the moment I'm still guessing and very grateful for CD-RWs....

What version of cdrecord are you running, maybe it's a problem with the program version itself?

And what are you using (and version) to create the .img files? Might also be a bug with that as well.

Sounds very strange though...

Cool

*Usually* a bus error will cause cdrecord to issue some pretty hefty
errors. If you have other reasons to expect your bus is a problem,
you might try swapping cables, maybe get yourself a shielded cable.
But, if it were the bus you shouldn't be able to send audio data down
it without experiencing the same problems.

If you have the disk space, try copying the disk image, maybe there is a
problem with your hard drive, and it happens to be affected by reading
this file. I kind of doubt it, but it's possible, and doesn't really take much
time to do this. . .

Try making a small test image and writing that, to test whether it's really
a -data problem.

What version of cdrecord do you have? did you create this disk image
yourself?

[MasterC got that in before me]

Currently I have several versions of cdrecord which have either been installed as binaries from rpms or compiled locally & copied into /usr/bin. I read that versions of cdrecord above 1.11a33 will not work correctly with xcdroast, so I have tried 1.11a28 and a33. Results the same with each. To create the img I used xcdroast as a front end to mkisofs v1.15a27.

If I compile the cdrecord sources locally, is it enough to copy the executable into the /usr/bin directory or does it depend on libraries anywhere else?

Incidentally, I just wrote the offending data onto a CD by first writing some audio tracks, then the data. It must be something to do with the content of the first track but now I'm really stumped...

Hi Moses,

Just noticed your reply... thanks again, I just recorded a CD containing this data at 24x (although cdrecord reported the average record speed to be a poor 10x) so I am almost certain now that the IDE system is working fine. I have recompiled the kernel and all modules several times without problem, and this is usually a good test of stability. I changed the IDE setup so the CD-RW is now secondary master, alone, with the HDD and DVD sharing the primary controller but the system logs have never reported any disk errors.

Linux has thrown me some odd complaints in the past, but this is the best yet....