D-I prompts for CD-ROM drivers for USB drive when drivers present, drive not mounted yet (preseed issue)

[Brian Kelley]

Binary package hint: debian-installer

When installing Ubuntu 8.04 from a USB drive with the Language and keyboard values preseeded, the installation get to the point where it detects CD hardware faster than the Linux kernel mounts the drive, and does not find the drive. If no additional drivers are selected, and the detect and mount CD step is selected again, installation proceeds as expected.

An undesirable workaround is to re-enable (delete the preseed/kernel parameter value) the language prompt so that the installer has to wait for user input before detecting the CD drive, giving the kernel time to mount the CD drive. Then, the drive has been mounted by the time the installer tries to detect and installation continues as expected.

I propose a delay (5-15 seconds) is implemented if the CD drive is not detected, and then the drive detection process repeated. This would allow time for the kernel to detect and connect the USB CD drive.

Ubuntu version: 8.04.3
d-i version: unknown (which command would I use?)

[Tom Ellis (tellis)]

Uploading some debugging info, I am having the same issue on 9.10 alternative cd.

my boot options:
cdrom-detect/try-usb=true persistent file=/cdrom/test/test.seed debian-installer/locale=en_US.UTF-8 console-setup/layoutcode=us initrd=/install/initrd.gz DEBCONF_DEBUG=developer quiet rootdelay=10

(tried to see if rootdelay works, but you can remove this and it still happens)

[Tom Ellis (tellis)]

Further to this, I hacked the install initrd to include a sleep (I am not recommending this a solution! :-) ) in cdrom-detect.postinst to test my theory on the usb drive not settling and the behaviour is as I would expect (usb drive detected correctly mounted and installation continues).

Small diff of my /var/lib/dpkg/info/cdrom-detect.postinst attached.

[Colin Watson (cjwatson)]

Tom, could you please file a separate bug report? It's not obvious to me that your logs are actually directly related to the original user's report - it's similar, but this is all pretty delicate stuff and there could be multiple causes.

[Tom Ellis (tellis)]

Colin: I've created Bug #586036.

Thanks for your help.