[PATCH v5 00/21] libxl: domain save/restore: run in a separate process

[Ian Jackson <ian.jackson@eu.citrix.com>]

This is v5 of my series to asyncify save/restore, rebased to tip and
retested.  There are minor changes to 3 patches, as discussed on-list,
marked with "*" below:

    01/21 libxc: xc_domain_restore, make toolstack_restore const-correct
    02/21 libxc: Do not segfault if (e.g.) switch_qemu_logdirty fails
    03/21 libxl: domain save: rename variables etc.
    04/21 libxl: domain restore: reshuffle, preparing for ao
  * 05/21 libxl: domain save: API changes for asynchrony
  * 06/21 libxl: domain save/restore: run in a separate process
    07/21 libxl: rename libxl_dom:save_helper to physmap_path
    08/21 libxl: provide libxl__xs_*_checked and libxl__xs_transaction_*
    09/21 libxl: wait for qemu to acknowledge logdirty command
    10/21 libxl: datacopier: provide "prefix data" facility
    11/21 libxl: prepare for asynchronous writing of qemu save file
    12/21 libxl: Make libxl__domain_save_device_model asynchronous
    13/21 libxl: Add a gc to libxl_get_cpu_topology
    14/21 libxl: Do not pass NULL as gc_opt; introduce NOGC
    15/21 libxl: Get compiler to warn about gc_opt==NULL
    16/21 xl: Handle return value from libxl_domain_suspend correctly
    17/21 libxl: do not leak dms->saved_state
    18/21 libxl: do not leak spawned middle children
    19/21 libxl: do not leak an event struct on ignored ao progress
  * 20/21 libxl: further fixups re LIBXL_DOMAIN_TYPE
  ! 21/21 libxl: DO NOT APPLY enforce prohibition on internal

All of these apart from the last have been acked and I intend to
commit those to xen-unstable.hg soon.

However, first I will invite Shriram to check that Remus is still
working.  (I can't conveniently do this with this message due to
shoddiness in git-send-email.)