On Fri, Aug 23, 2024 at 02:20:41AM +0000, Michael Kelley wrote: > Christoph Hellwig, or anyone else who knows the history and current > reality better than I do, please jump in. :-) It's not just interrupt context, but any context that does not allow blocking. There is plenty of that as seen by the moving of nvme to specifically request a blocking context for I/O submission in this path. That being said there are probably more contexts that can block than those that can't, so allowing for that option is a good thing.
