On 2/19/21 3:21 PM, Eric W. Biederman wrote: > Jens Axboe <axboe@xxxxxxxxx> writes: > >> PF_IO_WORKER are kernel threads too, but they aren't PF_KTHREAD in the >> sense that we don't assign ->set_child_tid with our own structure. Just >> ensure that every arch sets up the PF_IO_WORKER threads like kthreads. > > I think it is worth calling out that this is only for the arch > implementation of copy_thread. True, that would make it clearer. I'll add that to the commit message. > This looks good for now. But I am wondering if eventually we want to > refactor the copy_thread interface to more cleanly handle the > difference between tasks that only run in the kernel and userspace > tasks. Probably would be a worthwhile future cleanup of the code in general. -- Jens Axboe
