Linux: s_op: use .free_inode

[snajpa]

Motivation and Context

#16608

Description

as per Documentation/filesystems/porting.rst:

quote:

**strongly recommended**

take the RCU-delayed parts of ->destroy_inode() into a new method - ->free_inode().  If ->destroy_inode() becomes empty - all the better, just get rid of it.

endquote.

How Has This Been Tested?

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]

I believe we still want to register .destroy_inode even when .free_inode is available. It's just that the RCU-delayed parts have been moved in to .free_inode.

[snajpa]

Well I tried to adhere to the note in porting.rst to the fullest extent possible, as it doesn't even say that .free_inode is now required if one doesn't want to run into unexplainable behavior (perhaps only with some kernel releases along the way to full folio conversion); I was wondering what more in terms of problems can we expect if I try to go against that note, in a sense of leaving .destroy_inode as it was.

Do you see any consequences of merging destroy_inode into evict_inode from the ZFS point of view? I couldn't think of anything so that's why I decided to do away with destroy_inode by merging it, just an abundance of caution, aka "what else aren't they telling me in the note, when they didn't even say it's mandatory but just recommended"

[snajpa]

I mean, if the inode gets as far as evict_inode being called on it, it's already done and zfs_zget is going to keep looping until it disappears completely, isn't it so?

[snajpa]

@behlendorf ping, what do you think, do we keep it as I propose? I'd like to push this forward but this seems like it needs further input from you :) Thanks!

[behlendorf]

__maybe_unused shouldn't be needed here or above. We can use the guard macro's here to build the required functions, or not.