Reinstate the old zpool read label logic as a fallback.

[rincebrain]

Motivation and Context

#12032

Description

I landed a verbatim copy of the original zpool_read_label as "zpool_read_label_slow", then reshuffled the logic in
zpool_read_label to call zpool_read_label_slow if any of the AIO operations returned EOPNOTSUPP, which is what
FBSD spits back when you try AIO on FDs where it thinks it's "unsafe".

(The reshuffling was necessary because errno after this happens isn't (always?) EOPNOTSUPP, at least some of the time it's EIO; I landed the whole old function rather than integrating the new and old logic into one function after seeing that the diff between them was almost all of the function body, so it would have just been two entirely separate copies of the logic inside anyway.)

How Has This Been Tested?

• Tried importing and creating a pool over NFS on FBSD 13/x86_64 before and after this change with vfs.aio.enable_unsafe=0 (it fails before and works after).
• Tried pool creation and importing on Debian bullseye/x86_64
• Tried pool importing on CentOS 7/x86_64
• Tried pool importing on Ubuntu 18.04/x86_64

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.

[behlendorf]

cc: @asomers

[asomers]

I think this is a good idea. It never occurred to me that somebody might try to create a pool over NFS.

[ghost]

Now we just need to squash all those commits and make sure the final commit is signed off on to satisfy checkstyle.

[rincebrain]

Hm, the merge with master plus the rebase has it reporting that I committed all the commits that were merged from master. LMK if I should have done that some other way, and I'll go rewrite history appropriately.

[ghost]

No worries, sorted it out for you.

[rincebrain]

For my future reference, how did you clean it up?

(...also, I didn't realize contributors could force push to others' branches.)

[ghost]

For my future reference, how did you clean it up?

git rebase -i master making sure only the one commit was being picked.

(...also, I didn't realize contributors could force push to others' branches.)

There's a checkbox when you open a PR that allows contributors to make changes.