Daniel Verkamp: Automated regression testing

Investigated boot-up messages about wget not finding a URL with a mkbootfs system; this seems to be due to missing contents of /proc/cmdline, but I am not sure what format this should be in or what the purpose of it is.

Sidetracked by a discussion of serial support for COMBOOT in gPXE (for menu.c32); wrote a quick hack, which worked, but this will need more cleanup before it's ready.

Working on windows_7_install_to_iscsi.

First, I had problems because my router did not routeback packets to the local network. Enabling the option for this on my router fixed that problem, allowing the iSCSI disk to show up in setup.

Then I hit the problem where setup would not continue with the message about being unable to boot from the selected disk. Connecting a physical hard disk allows setup to continue to the “copying files” stage, but after the following reboot, the iSCSI target will not boot. DEBUG=bootsect reports “Booted disk returned via INT 18 or 19”.

Using fdisk, I made the partition on the disk image used by the iSCSI target active/bootable. This allowed the boot to proceed, but then I get the error message “BOOTMGR is missing/Press Ctrl+Alt+Del to restart”.

Restarted Win7 iSCSI install test to be sure the environment was consistent. Erased MBR on physical disk on test machine (dd if=/dev/zero of=/dev/hda bs=512 count=1), deleted iSCSI disk and created a new one (20 GB this time to be sure space requirements would be met), and started from scratch.

The same results as before were observed: Windows would not install on the iSCSI disk without the physical disk connected, and it installed the bootloader on the physical disk, not the iSCSI target. The physical disk contains bootmgr and the related infrastructure. The iSCSI partition was also not marked active.

I also attempted to perform the procedure in QEMU, but the Windows 7 installer got stuck for over an hour on the stage before disk selection, so it could not be tested.

Installed Win7 on a physical hard disk in the test machine and transferred it to the iSCSI target, ran sanbootconf, then attempted to boot it from there. This booted farther than the iSCSI-installed version (it actually had a proper bootloader installed), but it did not boot completely, possibly due to a missing NIC driver.