On Sat, Jan 15, 2022 at 09:34:10AM +0900, Tetsuo Handa wrote: > Christoph is not a fan of proliferating the use of task_work_add(). Can we go with exporting task_work_add() Not a fan != NAK. If we can't think of anything better we'll have to do that. Note that I also have a task_work_add API cleanup pending that makes it a lot less ugly. > for this release cycle? Or instead can we go with providing release() callback without disk->open_mutex held > ( https://lkml.kernel.org/r/08d703d1-8b32-ec9b-2b50-54b8376d3d40@xxxxxxxxxxxxxxxxxxx ) ? This one OTOH is a hard NAK as this is an API that will just cause a lot of problems.
