Installed RC2. There's an IDE floppy on /dev/hda. Kudzu wrote the following to /etc/fstab: /dev/hda /mnt/ls120.0 auto noauto,owner,kudzu 0 0 "mount /dev/hda" or "mount /mnt/ls120.0" fails. I have to explicitly write everything out "mount -t ext2 /dev/hda /mnt/ls120.0". Experimentation showed that mount was passing the "kudzu" option directly to the kernel, and the kernel mount fails because it does not recognize this non-standard option: mount -t ext2 -o nosuid,noexec,kudzu /dev/hda /mnt/ls120.0 This does not work (that's what the kernel gets from a plain "mount /dev/hda"). mount -t ext2 -o nosuid,noexec /dev/hda /mnt/ls120.0 This does. mount should ignore any "kudzu" option it finds in fstab.
I could work around this in mount, but I'm not sure it's the right thing to do. Michael, should the kernel ignore the kudzu option or should mount strip it off?
This defect considered MUST-FIX (show-stopper) for Florence GOLD
Strip it off in mount. It's a special-case option that the kernel should not be burdened with dealing with, and dealing with it in mount is better. Policy goes in user-space whenever possible. The kudzu option was added because ewt thought it didn't break anything. Looks like he was wrong...
It looks like it. The bit that surprises me is that I tested mounting a couple of pieces of media w/ the kudzu flag and auto and it worked fine. I'll patch this in mount this afternoon.
Added a patch to mount package to make it ignore the kudzu option.