Red Hat Bugzilla – Bug 62613
Driver disk still mounted when I go to create a boot disk
Last modified: 2007-04-18 12:41:36 EDT
Description of Problem: I was caught out be an odd gotcha. Probably my fault, but it was unexpected behaviour. When I was getting near the end of the install, I put a blank diskette in the drive ready for boot disk creation. I then hit the button to say yes, create a disk, and then I saw some error messages relating to the driver disk (and an error on one of the other VTs saying that the floppy could not be unmounted due to "busy inodes on changed media"). Boot disk creation then failed. I'm slightly hazy on the sequence of events here; I'm fairly sure that the errors appeared after I said "yes, there's a blank disk there you can put the boot image on" but it may have been just before. Either way, I wasn't expecting the previous disk (drvnet) to still be in use at that point since once it's insmod'd there's no longer any need for the module to be available. Could the driver disk be unmounted earlier on in the process? I had assumed this was done earlier on, after the contents were read in.
Jeremy can you verify this please?
I haven't seen this in as I've looked at driver disks over the past day or two. I have a couple other things to verify with driver disks, though, so will double-check that everything is getting unmounted as I do so.
It's more a question of *when* it gets unmounted, I was caught out as anaconda was trying to unmount the driver disk as I was inserting the blank disk ready for boot disk creation. It is being unmounted, but I think that needs to happen a bit earlier in the process (ideally before package installation begins, so I have plenty of time to swap disks). I'll also concede that technically it's my fault as I was a little eager about swapping the disks.
I just did an install with a driver disk to test something else. The driver disk is getting unmounted in the loader immediately after reading the drivers off for me. Is this reproducible for you?
It might depend on whether it's done by booting with "linux dd" or by pressing F2 to access the driver disk before setting up the network. I'm going to have to find another machine to test on as the one I installed is tucked away in the corner of the room acting as my caching nameserver and squid cache :o) I'd restarted the install several times (to mess with the partition table) so I can't remember which method I used the last time. I only discovered the "press F2 to access a driver disk" option by accident after I'd forgotten to add the dd option when booting.
I've done both without problems. If you can reproduce this and give a clear set of steps to reproduce it, please reopen with that information as well as any information from /tmp/syslog about errors unmounting. Otherwise, I'm going to mark it up as a fluke or something.
I don't have a spare which needs the driver disk, unfortunately. The machine I was testing on is now in a cupboard acting as my squid cache. I don't consider it a critical problem, anyway. It only happens if you jump the gun with changing floppies. I suspect most first-timers won't do that, and experienced folk will understand the error messages :o) Try taking the driver disk out just before package installation finishes? /tmp/install.log.syslog was empty after reboot, which might also be considered a bug. install.log just lists the packages installed (all 157), no errors.