Regression: Tails shuts down after snapshot restore (on SATA optical drive)

[Christian Inci]

Tails thinks that the medium has been removed/changed and does it's emergency shutdown after restoring from a snapshot.

"cat /dev/sr0" errors out and there are errors about failing to access sr0 in dmesg.

Steps:

1.: Configure the VM to have the ISO in a SATA/AHCI optical drive.

2.: Boot it and make a snapshot when it's done.

3.: Restore from the snapshot. The Guest VM will shut down.

Regression in r64134. Not fixed in trunk yet.

Caveat: Triggering the bug on e.g. 5.2.20 might require to set the Base Memory size to anything larger than 1024 MiB and the Video Memory size to anything larger than 16 MiB. (e.g. set the Base Memory size to 4096 MiB and the Video Memory size to 32 MiB.)

Affected: trunk, 5.2.x

Unaffected: 5.1.x. Older versions should be unaffected.

[aeichner]

Please attach a VBox.log of such a VM run.

[Christian Inci]

Booting and saving snapshot

[Christian Inci]

Restore from snapshot

[aeichner]

Should've been fixed long ago, please reopen if this is still an issue.

[Christian Inci]

HACK/Workaround

[Christian Inci]

I looked through the changesets and couldn't find a try to fix that bug. It's still in trunk.

This patch should workaround that bug. It's a very ugly hack, which very probably has some unforeseen side effects. (mounting/booting iso's still seems to work on Linux guests)

My uneducated guess is that the code won't make a difference between starting and resuming a VM. I suppose that there shouldn't be a MountNotification when resuming a VM. (VSCSILunMmc.cpp:844, vscsiLunMmcGetEventStatusNotification)

I don't know the structure/architecture of the code, so a proper patch would be up to someone else.