On 11/13/24 9:16 PM, Christoph Hellwig wrote: > On Wed, Nov 13, 2024 at 01:51:48PM -0700, Jens Axboe wrote: >> Thanks for testing, but you can't verify any kind of perf change with >> that kind of setup. I'll be willing to bet that it'll be 1-2% drop at >> higher rates, which is substantial. But the reordering is a problem, not >> just for zoned devices, which is why I chose to merge this. > > So I did not see any variation, but I also don't have access to a really > beefy setup right now. If there is a degradation it probably is that > touching rq_next for each request actually has an effect if the list is > big enough and they aren't cache hot any more. I can cook up a patch Exactly. > that goes back to the scheme currently used upstream in nvme and virtio > that just cuts of the list at a hctx change instead of moving the > requests one by one now that the block layer doesn't mess up the order. I think that would be useful. I can test. -- Jens Axboe
