On Tue, Nov 28, 2023 at 06:29:59PM +0100, Oleg Nesterov wrote: > Forgot to menstion, > > On 11/28, Oleg Nesterov wrote: > > > > but please > > note irq_thread()->task_work_add(on_exit_work). > > and this means that Neil's and your more patch were wrong ;) Hm, that's all the more reason to not hang this off of PF_KTHREAD then. I mean, it's functional but we likely wouldn't have run into this confusion if this would be PF_FPUT_DELAYED, for example.
