From Bugzilla Helper: User-Agent: Mozilla/4.77 [en] (X11; U; Linux 2.4.4-4GB i586) Description of problem: Traceback shows: File "/usr/bin/anaconda", line 3, in ? import sys, os, signal Import Error: No module named os install exited abnormally How reproducible: Always Steps to Reproduce: 1. Tried booting with CD blank (unpartitioned HD) 2. Tried booting with floppy "boot.img" (unpartitioned HD) 3. Tried booting with both above with paritioned HD (Linux and Swap using Partition Magic) 4. Booted and installed SuSE 7.2 on same system w/o error, Booted and installed OpenBSD 1.5 on same system w/o error. Actual Results: From boot sequence: running /sbin/loader Running anaconda - please wait Could not find platform independent libraries <prefix> Could not find platform dependent libraries <exec-prefix> Consider setting $PYTHONHOME to <prefix>[:exec-prefix] ... (traceback as described above) Expected Results: Normal Install Additional info: This is a "vanilla" unit-under-test inhouse system: IDE - Primary Master - 8GB HDD IDE - Secondary Master - 40X CDROM 256 MB RAM AMD K6-2/500 Creative Soundcard 3DFX Voodoo3 Standard FDD
I think that you might have a bad cd. Press <Ctrl><Alt><F4> and see if there are any error messages related to cdroms.
Your suggestion was almost correct ;) The problem turned out to be a CDROM drive and/or Motherboard that couldn't handle DMA properly. NetBSD and SuSE both turn off DMA and test the device's capability to handle DMA. If it works, then DMA is enabled. As I recollect, Winslop (er, Windows 98SE) does this too. Redhat enables DMA and assumes it will be OK. You might want to change your strategy. I "forced" DMA off in the BIOS and everything worked O.K. (Although it looks like RH enables DMA anyway, tries 5 reads and then turns DMA off subsequently - observed from the boot log).
Oh. If you boot with 'linux ide=nodma' at the cd boot screen, things would have worked ok and you wouldn't have to have changed the bios. I think we will disable dma transfers for cdroms in the future. The problem is that some drives can't do dma transfers, but they fall back to a non dma mode. Other drives don't fall back properly, and that's where the problems come from. Thanks for your report.
Using "linux ide=nodma" would have worked, but would have made the HDD not use DMA (which it was capable of doing). The BIOS technique allowed me to use DMA on my primary IDE and no DMA on my secondary IDE to which the CDROM was attached. I totally agree with you that eliminating DMA on CDROM/DVD is a good thing. It seldom (if ever) brings any performance advantage to CDROM/DVD devices and often brings headaches due to hardware DMA "incompatibilies".