Factor out non-portable vnode_t usage

[mattmacy]

On FreeBSD file offset state is maintained in struct file. A given
vnode can be referenced from many different struct file *. As a
consequence, FreeBSD's SPL doesn't support vn_rdwr with the FAPPEND
flag.

This change replaces the non-portable vnode_t with the portable
zfs_file_t in the common code.

Upon discussion with Brian - have expanded the scope to eliminate all references to vn_rdwr and have limited the scope of the use of the vn_ functions to vdev_file.c

Signed-off-by: Matt Macy [email protected]

Motivation and Context

Description

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)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the ZFS on Linux code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• All new and existing tests passed.
• All commit messages are properly formatted and contain Signed-off-by.

[ikozhukhov]

LGTM, thanks

[behlendorf]

It's awesome to see this refactoring removes 600 lines of mostly ugly compatibility code!

[lundman]

vtype -> mode, file_t -> zfs_file_t, there is a lot to like in this version.

and the rest can be handled with wrappers in libspl for things like umem_zalloc() in libzpool/kernel.c etc.

[codecov]

Codecov Report

Merging #9556 into master will decrease coverage by 0.29%.
The diff coverage is 80.76%.

@@            Coverage Diff            @@
##           master    #9556     +/-   ##
=========================================
- Coverage   79.22%   78.93%   -0.3%     
=========================================
  Files         419      418      -1     
  Lines      123704   123531    -173     
=========================================
- Hits        98009    97504    -505     
- Misses      25695    26027    +332

Flag	Coverage Δ
#kernel	79.7% <85.47%> (-0.22%)	⬇️
#user	66.28% <68.58%> (-0.45%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 540312d...4a1f22c. Read the comment docs.

[kithrup]

A lot of this seems to be adding more indirection, so we can avoid the problems of, say, Linux and FreeBSD having different file and vnode types.

Given that, it seems fine to me.

[behlendorf]

@kithrup thanks for the feedback. That's right, the intent is to provide a basic interface which can be implemented on all platforms which can be used by the common code. The vnode interface effectively served this same purpose, but the compatibility code required to support it ended up being a bit gnarly. After this change there are longer any vnode /VOP calls or related structures in the common code.

[mattmacy]

A lot of this seems to be adding more indirection, so we can avoid the problems of, say, Linux and FreeBSD having different file and vnode types.

Given that, it seems fine to me.

It only really adds extra indirection for Illumos. FreeBSD doesn’t support the FAPPEND flag to vn_rdwr and would need its own complicated spl vnode in order to support it. As is FreeBSD had a bunch of special case handling and routines to pass a struct file * where Illumos was passing a vnode.