The member VMStateField.start was solely used to implement the partial
data migration for VBUFFER data (basically provide migration for a
sub-buffer). However the implementation of this feature is broken, but
this goes unnoticed since the feature is not used at all.

So what the series does is first add some tests for VBUFFER, then add a
test which proves that the VMS_ALLOC and used together with .start != 0
is broken.  Then we immediately revert this last patch since we are
going to drop it instead of fixing it.  Lastly simplify things by
dropping VMStateField.start altogether.

An additional benefit is that .start can be re introduced to be used for
linked structures as proposed by Jianjun in "[QEMU PATCH v6 2/2]
migration: migrate QTAILQ".

Guenther Hutzl (1):
  tests/test-vmstate.c: Add vBuffer test

Halil Pasic (3):
  tests/test-vmstate.c: prove VMStateField.start broken
  Revert "tests/test-vmstate.c: prove VMStateField.start broken"
  migration: drop unused VMStateField.start

 hw/char/exynos4210_uart.c   |   2 +-
 hw/display/g364fb.c         |   2 +-
 hw/dma/pl330.c              |   8 +--
 hw/intc/exynos4210_gic.c    |   2 +-
 hw/ipmi/isa_ipmi_bt.c       |   4 +-
 hw/ipmi/isa_ipmi_kcs.c      |   4 +-
 hw/net/vmxnet3.c            |   2 +-
 hw/nvram/mac_nvram.c        |   2 +-
 hw/nvram/spapr_nvram.c      |   2 +-
 hw/sd/sdhci.c               |   2 +-
 hw/timer/m48t59.c           |   2 +-
 include/migration/vmstate.h |  20 +++-----
 migration/savevm.c          |   2 +-
 migration/vmstate.c         |   4 +-
 target-s390x/machine.c      |   2 +-
 tests/test-vmstate.c        | 115 +++++++++++++++++++++++++++++++++++++++++++-
 util/fifo8.c                |   2 +-
 17 files changed, 141 insertions(+), 36 deletions(-)

-- 
2.8.4


Reply via email to