Hi Martin, On 03/06/2016 06:26 PM, Martin Pitt wrote: > Also, is it actually safe to add the base image *again* after a reboot > if it already is added? Won't you have two drives then, and > /dev/baseimage randomly points to one of it?
But setup_baseimage() doesn't do that. It's only run the first time after QEMU is started - it's NOT run every boot. Just look at where it's hooked in. :-) (That was the point: I only wanted to add it once so I don't have to deal with reboots - otherwise we need to intercept reboots and remove the drive and all these nasty things that I didn't want to deal with.) >> What you _could_ do is create a flag file in /run and update the reboot >> script to update the initramfs before rebooting if that flag file is >> set - that way only if a reboot is needed will the initramfs be updated >> and hence you optimize for the common case where that's not needed. > > That's one way to mitigate this, although it still sucks as it would > run on *every* reboot, No, only on the first reboot, because then the initramfs is generated and doesn't need to be regenerated. > We also need that > flag file on the host side to avoid re-adding the drive to the QEMU > monitor again on each reboot. No, we don't, see the other patch I sent. I did use --shell on various trivial tests to verify the presence / absence of that flag file at various points in time. > Maybe a better way to avoid this would be to introduce the concept of > a pre-boot hook in adt-virt-*, which could then simply remove the > drive via the QEMU monitor. Why? It's not necessary. So current git master + my flag file patch + debhelper (>= 9) works just fine for me (and I did test it quite a bit) AND is (except for the override) completely lintian clean, even with my lintian setting. Regards, Christian
signature.asc
Description: OpenPGP digital signature
