Hi Keith! > Thanks! I had just posted a test for this scenario earlier today for > liburing: > > https://lore.kernel.org/io-uring/20250416162802.3614051-1-kbusch@xxxxxxxx/T/#u > > I was wondering why it didn't blow up. It triggered on our end with a heavily threaded workload. Each thread is reading fixed-size 8KB blocks. The small block size results in needing small PI allocations. And with thousands of I/Os in flight at any point in time, we are fairly likely to end up with PI allocations sharing a page. -- Martin K. Petersen
