Faster checksum benchmark on system boot

[mcmilk]

While booting, only the needed 256KiB benchmarks are done now.

The long delay for testing all checksums is done on request for them now:

Linux: cat /proc/spl/kstat/zfs/chksum_bench
FreeBSD: sysctl kstat.zfs.misc.chksum_bench

Closes #17560

Motivation and Context

The issue #17560 is the main reason for this PR.

Description

How Has This Been Tested?

I loaded the module :-)

Old load module time: 1,5s (Github runner)
New load module time: 0,25s (Github runner)

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)
• Quality assurance (non-breaking change which makes the code more robust against bugs)
• 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.

[mcmilk]

Sorry for this extra rebasing - had trouble with my handy :/

[mcmilk]

I reworked the commit message to give credits for the reporting:
Reported by: Lahiru Gunathilake [email protected]
Co-authored-by: Colin Percival [email protected]

Thanks @cperciva and @bllgg

[lundman]

🤔 Benchmarking is only used with fastest - if the tunable is set to something static, skip benchmarking entirely? 🤔

[mcmilk]

🤔 Benchmarking is only used with fastest - if the tunable is set to something static, skip benchmarking entirely? 🤔

It's not skipped. Only the 256KB blocksize is checked now.
If some admin wants to know other sizes, then he cat check it via: cat /proc/spl/kstat/zfs/chksum_bench ... this will then calculated when requested. So the first cat delays with ~one seconds now.

Edit: ah, I understand what you mean now :/ Sorry... it's early in the morning here ;-)
So when the user sets some implementation by hand, then the whole check could be skipped. -> hm... yes, this would be possible, but it's not done for fletcher and aes this way. So the documentation would be different then also... I think some micro benchmarking is okay in the beginning. But we could change it ... but I would prefer for all algos then.

[lundman]

Just a thought.. I can set values before ZFS loads on Windows, but not on macOS - so it'd only work on one platform for me. I don't know when Linux/FreeBSD would load the tunables.

[cperciva]

Just a thought.. I can set values before ZFS loads on Windows, but not on macOS - so it'd only work on one platform for me. I don't know when Linux/FreeBSD would load the tunables.

Assuming that "tunable" here means the same thing as it normally does in the FreeBSD kernel: We process those long before the ZFS module initialization runs. Loader tunables can be used to control lots of ZFS behaviours.