There is a case that Linux has a bug and unlocking does not work properly like this:
https://lore.kernel.org/lkml/20230608084609.14245-1-zhangjiachen.jaycee@b...
Especiall in the situation that live migration source node has this kind of bug,
destination must not locking, or otherwise VM stops when live migration runs.
There commits add workaround for this kind of problems.
Could you take a look?
Especially following is points I could not make sure:
In the first commit, I wrote explanation on file section in the document, but placed
parser and formatter in virDomainStorageSourceParse and virDomainStorageSourceParse
because locking can be common if added to other backends. Is it good to place here, or is
there some better location to place?
Hiroki Narukawa (2):
conf: add locking option to disk source
qemu_block: add locking option
docs/formatdomain.rst | 5 ++++
src/conf/domain_conf.c | 8 +++++++
src/conf/schemas/domaincommon.rng | 5 ++++
src/conf/storage_source_conf.h | 3 +++
src/qemu/qemu_block.c | 7 ++++++
tests/qemublocktest.c | 1 +
...le-raw-aio_native-locking-off-srconly.json | 10 ++++++++
.../file-raw-aio_native-locking-off.json | 23 +++++++++++++++++++
.../file-raw-aio_native-locking-off.xml | 12 ++++++++++
9 files changed, 74 insertions(+)
create mode 100644
tests/qemublocktestdata/xml2json/file-raw-aio_native-locking-off-srconly.json
create mode 100644 tests/qemublocktestdata/xml2json/file-raw-aio_native-locking-off.json
create mode 100644 tests/qemublocktestdata/xml2json/file-raw-aio_native-locking-off.xml
--
2.25.1