Hi,
This set includes a test to trigger the xfs_bmap_extents_to_btree WARN
that tends to the the most observed symptom of the dependent allocation
failures that this author encounters. The patches to fix the problem
also introduced a failure in xfs/306. I'm including a fix for that as
well, though if this is an inappropriate resolution of the underlying
problem, I'm happy to take another approach.
This is sent as an RFC since it's my first foray into submitting
xfstests. The 608 test should be deterministic on kernels that precede
the AG-aware allocator re-write. On my machines on recent kernels I'm
getting about a 40-50% success rate with triggering the bug with this
test now.
-K
Krister Johansen (2):
xfs/608: a test case for xfs_bmap_extents_to_btree allocation failure
xfs/306: update resblks to account for increased per-ag reservation
size
tests/xfs/306 | 2 +-
tests/xfs/608 | 372 ++++++++++++++++++++++++++++++++++++++++++++++
tests/xfs/608.out | 2 +
3 files changed, 375 insertions(+), 1 deletion(-)
create mode 100755 tests/xfs/608
create mode 100644 tests/xfs/608.out
base-commit: f5ada754d5838d29fd270257003d0d123a9d1cd2
--
2.25.1
