dedup=verify doesn't clear the blkptr's dedup flag

[ahrens]

Motivation and Context

Description

The logic to handle strong checksum collisions where the data doesn't
match is incorrect. It is not clearing the dedup bit of the blkptr,
which can cause a panic later in zio_ddt_free() due to the dedup table
not matching what is in the blkptr.

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.

[codecov]

Codecov Report

Merging #8936 into master will increase coverage by 12.5%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master    #8936      +/-   ##
==========================================
+ Coverage   66.26%   78.77%   +12.5%     
==========================================
  Files         324      382      +58     
  Lines      103014   117842   +14828     
==========================================
+ Hits        68261    92827   +24566     
+ Misses      34753    25015    -9738

Flag	Coverage Δ
#kernel	79.28% <0%> (?)
#user	67.62% <0%> (+1.36%)	⬆️

---

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 fb0be12...00a7c2c. Read the comment docs.

[behlendorf]

This sounds like it may be the long standing root cause which prompted this non-ideal change 5dc6af0. If that's the case, it's worth considering reverting that commit. Since this is such as unlikely code path to hit it'd be nice to somehow add a test case which exercises it.

[tcaputi]

although the change is correct, i very much doubt that it has ever caused an issue in the real world. As the comment above states, if we hit this code we should be automatically emailing Nature.

[ahrens]

Yeah, we actually hit this code path due to a different bug, that caused us to think the two blocks were different, when they weren't really. That bug was caused by abd_cmp() comparing the entire ABD when we only wanted to compare the first part of it. In the current code, I couldn't figure out how that would happen, and presumably if it was happening then we'd be hitting this bug. So I didn't include that code change, although I could - it's basically to pass zio->io_size as an additional parameter to abd_cmp() (as is done on illumos).

[behlendorf]

Thanks for the additional explanation, I'm happy to go with the PR as-is.