On Wed, Jul 03, 2024 at 01:40:32PM +0000, Chuck Lever III wrote: > > Was your test running both reads and truncates on the same file in parallel? > > I didn't record it in my notes, but IIRC I was running fstests. > I'm pretty sure one or more of the fsx-based tests will do this. Note that you don't have to truncate as in the truncate system call or anything changing the file system. Simply invalidating the page cache can lead to that, and xfstests exercises that heavily in the tests that mix buffered and direct I/O.
