migration/savevm: Allow immutable device state to be migrated early (i.e., before...
authorDavid Hildenbrand <david@redhat.com>
Tue, 17 Jan 2023 11:22:44 +0000 (12:22 +0100)
committerJuan Quintela <quintela@redhat.com>
Mon, 6 Feb 2023 18:22:56 +0000 (19:22 +0100)
commit62f42625d4e27a1993ab1999d0e86aedabf9a961
treeb894b0ad4be4bbb6e1557b2ec53a31edf243af8c
parente3bf5e68e2a97898f37834c47449101172ced123
migration/savevm: Allow immutable device state to be migrated early (i.e., before RAM)

For virtio-mem, we want to have the plugged/unplugged state of memory
blocks available before migrating any actual RAM content, and perform
sanity checks before touching anything on the destination. This
information is immutable on the migration source while migration is active,

We want to use this information for proper preallocation support with
migration: currently, we don't preallocate memory on the migration target,
and especially with hugetlb, we can easily run out of hugetlb pages during
RAM migration and will crash (SIGBUS) instead of catching this gracefully
via preallocation.

Migrating device state via a VMSD before we start iterating is currently
impossible: the only approach that would be possible is avoiding a VMSD
and migrating state manually during save_setup(), to be restored during
load_state().

Let's allow for migrating device state via a VMSD early, during the
setup phase in qemu_savevm_state_setup(). To keep it simple, we
indicate applicable VMSD's using an "early_setup" flag.

Note that only very selected devices (i.e., ones seriously messing with
RAM setup) are supposed to make use of such early state migration.

While at it, also use a bool for the "unmigratable" member.

Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>S
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
include/migration/vmstate.h
migration/savevm.c