On 10/2/24 8:56 AM, Keith Busch wrote: > On Wed, Oct 02, 2024 at 09:49:26AM +0200, Christoph Hellwig wrote: >> On Tue, Oct 01, 2024 at 10:23:43AM -0600, Keith Busch wrote: >>> I think because he's getting conflicting feedback. The arguments against >>> it being that it's not a good match, however it's the same match created >>> for streams, and no one complained then, and it's an existing user ABI. >> >> People complained, and the streams code got removed pretty quickly >> because it did not work as expected. I don't think that counts as >> a big success. > > I don't think the kernel API was the problem. Capable devices never > materialized, so the code wasn't doing anything useful. Exactly. I never saw ones that kept the stream persistent across GC, and hence they ended up being pretty useless in practice. -- Jens Axboe
