qemu versions 2.10 and 2.11 have error during migration of larger guests

== Comment: #0 - Christian Borntraeger  - 2018-06-28 06:39:27 ==
Migration fails with larger guests (e.g. 10GB)  on a z system prints an error message in the log 

see /var/log/libvirt/qemu/...
[...]
qemu-system-s390x: KVM_S390_SET_CMMA_BITS failed: Bad address

This messes up guest state for the CMMA values (guest data corruption)

This is fixed with 

commit 46fa893355e0bd88f3c59b886f0d75cbd5f0bbbe
Author:     Claudio Imbrenda <email address hidden>
AuthorDate: Thu Jan 18 18:51:44 2018 +0100
Commit:     Cornelia Huck <email address hidden>
CommitDate: Mon Jan 22 11:04:52 2018 +0100

    s390x: fix storage attributes migration for non-small guests
    
    Fix storage attribute migration so that it does not fail for guests
    with more than a few GB of RAM.
    With such guests, the index in the buffer would go out of bounds,
    usually by large amounts, thus receiving -EFAULT from the kernel.
    Migration itself would be successful, but storage attributes would then
    not be migrated completely.
    
    This patch fixes the out of bounds access, and thus migration of all
    storage attributes when the guest have large amounts of memory.
    
    Cc: <email address hidden>
    Signed-off-by: Claudio Imbrenda <email address hidden>
    Fixes: 903fd80b03243476 ("s390x/migration: Storage attributes device")
    Message-Id: <email address hidden>
    Reviewed-by: Christian Borntraeger <email address hidden>
    Signed-off-by: Cornelia Huck <email address hidden>

This fix is part of 2.11.1 so the qemu in bionic is fine.
The qemu in artful, as well as the qemu in the cloud archives for 16.04 need this fix, so we have
affected qemus in 17.10 and 16.04. 

Regarding 16.04:
The bug only triggers for host kernels >= 4.13 - in other words when you combine HWE kernel with the qemu from the cloud archive.