mbox series

[v3,0/7] migration: Fix s390 regressions + migration script

Message ID 20250109185249.23952-1-farosas@suse.de (mailing list archive)
Headers show
Series migration: Fix s390 regressions + migration script | expand

Message

Fabiano Rosas Jan. 9, 2025, 6:52 p.m. UTC 
changes:

- fixed appending again
- changed the VMSDFieldNull class to inherit from VMSDFieldGeneric

v2:
https://lore.kernel.org/r/20250109140959.19464-1-farosas@suse.de

- dropped comments patch
- new patch 4: rename the field to nullptr
- patch 6: add a sample JSON, fix the appending code

v1:
https://lore.kernel.org/r/20250107195025.9951-1-farosas@suse.de

Hi,

The situation that broke the last migration PR was:

1) emitting of JSON data by QEMU for
   VMSTATE_ARRAY_OF_POINTER_TO_STRUCT when NULL pointers are present
   has been broken for a while;

2) parsing of s390x migration stream by analyze-script.py has been
   broken for a while;

   (there's indications that it worked on s390x hosts, I'm assuming due
   to byte order coincidences)

3) s390x CSS migration has been broken for a while;

The s390x CSS migration uses VMSTATE_ARRAY_OF_POINTER_TO_STRUCT with
NULL pointers, triggering #1, but hidden due to #2 on TCG hosts and
due to #3 overall.

- patches 1: just to make rebase easier
- patches 2-3: cleanups
- patch 4: fixes #2
- patches 5-6: fix #1
- patch 7: fixes #3

Fabiano Rosas (6):
  migration: Add more error handling to analyze-migration.py
  migration: Remove unused argument in vmsd_desc_field_end
  migration: Fix parsing of s390 stream
  migration: Rename vmstate_info_nullptr
  migration: Fix arrays of pointers in JSON writer
  s390x: Fix CSS migration

Peter Xu (1):
  migration: Dump correct JSON format for nullptr replacement

 hw/s390x/s390-virtio-ccw.c   |   2 +-
 migration/vmstate-types.c    |   2 +-
 migration/vmstate.c          | 151 ++++++++++++++++++++++++++++-------
 scripts/analyze-migration.py | 142 +++++++++++++++++++++++---------
 4 files changed, 228 insertions(+), 69 deletions(-)