FreeBSD 15.0-CURRENT-amd64-20231130 UEFI-PREFERRED cloud-init ZFS ami-03df2931843b8541c launches but then is immediately shutdown on AWS and so cannot be used. This is in US East 1. Older AMIs work fine.
FreeBSD 15.0-CURRENT-amd64-20230907 UEFI-PREFERRED UFS ami-092f080286f6cfa09 works fine on a T2 medium instance with 20Gb storage.
I can reproduce this with the "FreeBSD 15.0-CURRENT-amd64-20240118 UEFI-PREFERRED base ZFS" AMI in a t2.medium. The instance just ends up terminated after I launch it. No such problem with the otherwise identical arm64 image in a t4g.large instance. I can't view the serial console of a t2.medium so it's hard to tell what's going wrong there. A t3.medium instance with the amd64 image, which does have a serial console, boots fine.
The biggest difference between T2 and T3 is that T2 instances run Xen/HVM, so I'm guessing something broke there. Roger, are you aware of any recent changes in our Xen code which could cause problems?
Hello, The 20231130 images and later are broken due to c7368ccb6801, however my attempt to fix this in b0165dc4539f is not complete, and seems to cause issues on older hypervisor versions or different configurations (see also PR 276421). I will look into it first thing tomorrow morning. Sorry for the mess.
Thanks! I just added t2.medium to my weekly snapshot testing (I was only testing the newer Nitro instances until now) and I can confirm that HEAD is broken but stable/13 and stable/14 seem fine. So as long as you get this fixed before any Xen changes get MFCed, I'm not worried. :-)
I've pushed the changes to: https://xenbits.xen.org/gitweb/?p=people/royger/freebsd.git;a=shortlog;h=refs/heads/pvh-hvm-unify If anyone wants to give a try.
Thanks for fixing this.