On Mon, Mar 01, 2021 at 09:14:40AM -0800, Darrick J. Wong wrote: > On Wed, Feb 10, 2021 at 03:58:18PM -0500, Theodore Ts'o wrote: > > This test verifies that the an unwritten extent is properly marked as > > written after writing into it. > > > > There was a hard-to-hit bug which would occasionally trigger with ext4 > > for which this test was a reproducer. This has been fixed after > > moving ext4 to use iomap for Direct I/O's, although as of this > > writing, there are still some occasional failures on ext4 when block > > size < page size. > > Are there still failures? Yes. It's on Ritesh's todo list, but I don't believe it's bubbled up to the top of his list. The failure rate is substantially less than what I had been seeing with 4k block sizes before the move to iomap, and in practice it was hard to hit in production. We noticed on Google data center workloads (but even then it was impossible to reliably repro before I found this fio recipe); as far as I know no one else had reported seeing this problem. So in summary, as of this writing, on 1k and 2k block sizes, this test is still occasionally failing for ext4 (but not xfs, where it reliably passes). On 4k block sizes on x86 with ext4, this test is still failing in on 5.4 and earler kernels, unless someone has managed to backport DIO iomap support for ext4 into an enterprise linux kernel. - Ted
