[PATCH hci-8.0/hci-8.1 0/2] migration/block-dirty-bitmap: fix postcopy abort with bitmaps on several nodes #VSTOR-147066

["Denis V. Lunev" <den@openvz.org>]

Live migration of a VM with dirty bitmaps on more than one node aborts
the destination once the bitmaps move post-copy:

  qemu-kvm: ../block.c:6477: bdrv_lookup_bs: Assertion
  `qemu_in_main_thread()' failed.

libvirt triggers this on every such VM: a CBT backup leaves a persistent
x-libvirt-cbt bitmap on each disk, and libvirt migrates bitmaps with
dirty-bitmaps + migrate-start-postcopy. hci-upgrade hit it evacuating a
three-disk VM, see VSTOR-147066 for the RCA.

This is the series posted upstream, unchanged apart from the ticket tag:
https://lore.kernel.org/qemu-devel/20260924115611.4098417-1-den@openvz.org/

Both patches apply cleanly to hci-8.0 and hci-8.1. The VZ-only changes in
migration/block-dirty-bitmap.c (PSBM-105022, VSTOR-98747) touch the save
side only. The source-side runstate_set(-1) abort seen in the same
ticket on hci-7.4 is already fixed here by 9867d4ddd04.

Denis V. Lunev (2):
  migration/block-dirty-bitmap: fix postcopy abort with bitmaps on
    several nodes #VSTOR-147066
  iotests/migrate-bitmaps-postcopy-test: add a bitmap on a second node
    #VSTOR-147066

 migration/block-dirty-bitmap.c                | 59 +++++++++++++++++--
 .../tests/migrate-bitmaps-postcopy-test       | 19 ++++++
 2 files changed, 74 insertions(+), 4 deletions(-)


base-commit: e8a73d39faf578d58aa2c4b4b024226796c390e6
-- 
2.53.0