Am 26.06.2024 um 18:27 hat Nir Soffer geschrieben: > On Wed, Jun 26, 2024 at 12:17 PM Daniel P. Berrangé <berra...@redhat.com> > wrote: > > > On Mon, Jun 24, 2024 at 06:08:26PM +0200, Kevin Wolf wrote: > > > Am 24.06.2024 um 17:23 hat Stefan Hajnoczi geschrieben: > > > > On Wed, Jun 19, 2024 at 08:43:25PM +0300, Nir Soffer wrote: > > > > > Tested using: > > > > > > > > Hi Nir, > > > > This looks like a good candidate for the qemu-iotests test suite. > > Adding > > > > it to the automated tests will protect against future regressions. > > > > > > > > Please add the script and the expected output to > > > > tests/qemu-iotests/test/write-zeroes-unmap and run it using > > > > `(cd build && tests/qemu-iotests/check write-zeroes-unmap)`. > > > > > > > > See the existing test cases in tests/qemu-iotests/ and > > > > tests/qemu-iotests/tests/ for examples. Some are shell scripts and > > > > others are Python. I think shell makes sense for this test case. You > > > > can copy the test framework boilerplate from an existing test case. > > > > > > 'du' can't be used like this in qemu-iotests because it makes > > > assumptions that depend on the filesystem. A test case replicating what > > > Nir did manually would likely fail on XFS with its preallocation. > > > > > > Maybe we could operate on a file exposed by the FUSE export that is > > > backed by qcow2, and then you can use 'qemu-img map' on that qcow2 image > > > to verify the allocation status. Somewhat complicated, but I think it > > > could work. > > > > A simpler option would be to use 'du' but with a fuzzy range test, > > rather than an exact equality test. > > > > For the tests which write 1 MB, check the 'du' usage is "at least 1MB", > > for the tests which expect to unmap blocks, check that the 'du' usage > > is "less than 256kb". This should be within bounds of xfs speculative > > allocation. > > This should work, I'll start with this approach.
If we're okay with accepting tests that depend on filesystem behaviour, then 'qemu-img map -f raw --output=json' should be the less risky approach than checking 'du'. Kevin