What were you trying to do that didn't work?

The following reproducer has been raised by our customer, it is inspired from the failed test while trying to rebuild the latest coreutils package on RHEL 9 on top of an XFS filesystem (tests/cp/fiemap-2.sh).

When using the `cp --sparse=always` option to copy a data file filled with zeros, on RHEL8, the zero data is treated as sparse and does not consume disk space. However, on RHEL9, the zero data is not treated as sparse and is copied as is. This issue occurs only on XFS in RHEL9, while on EXT4, the result is the same as RHEL8.

What is the impact of this issue to you?

Impacting the copy of VM images, reported as a significant impact on customers.

Please provide the package NVR for which the bug is seen:

coreutils-8.32-36.el9.x86_64

How reproducible is this bug?:

Always

Steps to reproduce

#!/bin/sh
printf x > k
dd bs=1k seek=1 of=k count=255 < /dev/zero
cp --sparse=always k k2
dd if=k of=k2.dd bs=4096 conv=sparse
stat -c %b k k2 k2.dd 

Expected results

512
8
8

Actual results

512
512
8

Additional notes

We can also confirm that rebuilding the 8.30 version from RHEL 8.10 does not show the problem on RHEL 9, the test PASSES, and while executing the above script, the holes are preserved. We do note significant changes in sparse_copy() from coreutils-8.32/src/copy.c.