ZTS: Corrupt all block copies with corrupt_file_at_level, test L1 corruption

[ghost]

Motivation and Context

In ZTS, list_file_blocks is supposed to output the level, vdev path, offset, and length of blocks in a given file.
It fails in the following ways:

• Only the path of the first child of a mirrored vdev is output
  This prevents corrupt_blocks_at_level from corrupting all copies of the blocks, which is desired for reproducing a particular issue we encountered in the async DMU project.
• Only L0 blocks and only the first DVA per block are output
  This prevents corrupt_blocks_at_level from working at any level other than 0.

Description

Only the path of the first child of a mirrored vdev is output

The first part of list_file_blocks transforms the pool configuration output by zdb -C $pool into shell code to set up a shell variable, VDEV_MAP, that maps from vdev id to the underlying vdev path. This variable is a simple indexed array. However, the vdev id in a DVA is only the id of the top level vdev.

When the pool is mirrored, the top level vdev is a mirror and its children are the mirrored devices. So, what we need is to map from the top level vdev id to a list of the underlying vdev paths. list_file_blocks does not need to work for raidz vdevs, so we can disregard that case.

Only L0 blocks and only the first DVA per block are output

The second part of list_file_blocks transforms the object description output by zdb -ddddd $ds $objnum into a stream of lines of the form "level path offset length" for the indirect blocks in the given file. The current code only works for the first copy of L0 blocks.

Add one more -d to the zdb command so we get all block copies and rewrite the transformation to match more than L0 and output all DVAs.

How Has This Been Tested?

Added a test case.

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 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.

[ghost]

• Sign-off on all the commits
• ALL the commits

[ghost]

• Simplify the added test to avoid duplicating non-corrupted checksum tests from filetest_001_pos

[ghost]

• Rebased
• Added -n flag to make gawk decode hex, squelched the resulting unknown argument warning on FreeBSD
• Added typeset for local vdev

[ghost]

• Amend the correct commit

[behlendorf]

These tests appears to have failed during the Debian 10 test run.

[ghost]

Ok back to plan B

[ghost]

• Rebased
• Check for gawk by testing if the -n flag is supported

[ghost]

Ah, Debian uses mawk, I'd not heard of that one before.