jeff beck wrote:
I have several thoughts, although none quaifies as an "explanation." If you are using any kernel but a 2.6.x AFAIK the audio mode doesn't use DMA. You may also need the latest 2.01a25 cdrtools for that, but you are using that version, hopefully built with patching to actually find and use the kernel include files for the running kernel.I'm sure I am doing something wrong and am only allowing myself to get yelled at by Mr. Schilling by posting, but I can't seem to write CD-R's successfully with my new Plextor PX-708A. I am using SONY CD-R 80's which are 32 speed. Nearly all the burns I try fail unless I use driveropts=burnfree - which I understand lessons the quality of my CD. I do not know whether the quality is lower only at the instant of bufferunderun or throughout the entire CD. I also am not sure how the quality of the CD is lower. I have tried using the cdrecord that comes with my redhat9, current binaries, or, as you see now, the latest alpha. I upgraded my firmware to 1.04. My burner is primary->slave, the wav files are onsecondary->master. I have DMA enabled everywhere. Changing the speed and turning off pad or dao seem tomake no difference. As I mentioned, burnfree works great. Any ideas would be greatly appreciated. Here is my cdrecord output: # cdrecord dev=0,0,0 -v -pad -dao -speed=16 -eject /usr/local/burn/*wav
If you have a 2.6 kernel, you can see if DMA is working by starting an audio burn and using "vmstat 1" to watch the system time. If it uses only a few percent DMA is working.
If it is NOT using DMA you may simply be running out ofCPU to feed the burner. You can certainly try increasing the size of the fifo buffer, assuming you have a typical RAM size, maybe "fs=20m" or so. This gives you a little hedge against problems, although the issue is not a failure to keep the fifo full. You could also try without -dao, that may slow the burn but result in success. I haven't looked into why that's so, but on a system *with* DMA working, I found that with DAO the burn time for my test data was 146 sec, and without was 210 sec, burning two examples of each.
You could always use a lower speed, of course, I had to do that on one of my systems with a similar problem. Hope this helps.
-- E. Robert Bogusta It seemed like a good idea at the time