[0/6] fstests: fixes and new tests

Message

Darrick J. Wong Oct. 5, 2018, 12:19 a.m. UTC

Hi all,

The first patch provides a second regression test for a condition where
mmap read of a file's EOF page shows nonzero bytes after EOF if the
previous operation was a zero_range all the way to EOF.  This bug only
seems to manifest on XFS, but it uses generic APIs and so is a generic
test.  The patch concerned here is "xfs: ensure post-EOF zeroing happens
after zeroing part of a file".

The second patch adds tests to see if XFS can detect and correct
incorrect filesystem summary counters at mount time.

Patch 3-5 fix some minor problems that cropped up with the upcoming
xfsprogs 4.19 release due to stricter metadata verification.

Patch 6 adds some regression tests for incorrect behavior on the part of
the clonerange and deduperange ioctls.  Specifically, we check for a
lack of file range verification w.r.t. resource limits; lack of security
privilege clearing when a file is reflinked; exposure of stale disk
contents when a file is reflinked into well above EOF; and problems
deduping when a single byte in the middle of the range doesn't match.

--D