vdev_file: make FLUSH and TRIM asynchronous

[robn]

[Sponsors: Klara, Inc., Wasabi Technology, Inc.]

Motivation and Context

zfs_file_fsync() and zfs_file_deallocate() are both blocking ops, so the zio_taskq thread is active and blocked both while waiting for the IO call and then while calling zio_execute() for the next stage. This is a particular issue for FLUSH, as the z_flush_iss queue typically only has one thread; multiple flushes arriving at once can cause long delays if the underlying fsync() response is particularly slow.

Description

To fix this, we dispatch both FLUSH and TRIM to the z_vdev_file taskq just as we do for reads and writes. Further, we return all results through zio_interrupt(), so neither the issue nor the file taskqs are blocked.

How Has This Been Tested?

ZTS run completed.

In a heavy fsync() workload on file-backed pools on a slow underlying device (a slow zvol in this case), throughput bump of ~100x. ymmv.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[amotin]

Thanks. Some time ago we fixed performance issue with TRIM being synchronous on Linux disks, so this is the right way to go.

[tonyhutter]

#17046 is now merged