On 06/06, Mike Christie wrote: > > On 6/6/23 7:16 AM, Oleg Nesterov wrote: > > On 06/05, Mike Christie wrote: > > > >> So it works like if we were using a kthread still: > >> > >> 1. Userapce thread0 opens /dev/vhost-$something. > >> 2. thread0 does VHOST_SET_OWNER ioctl. This calls vhost_task_create() to > >> create the task_struct which runs the vhost_worker() function which handles > >> the work->fns. > >> 3. If userspace now does a SIGKILL or just exits without doing a close() on > >> /dev/vhost-$something, then when thread0 does exit_files() that will do the > >> fput that does vhost-$something's file_operations->release. > > > > So, at least in this simple case vhost_worker() can just exit after SIGKILL, > > and thread0 can flush the outstanding commands when it calls vhost_dev_flush() > > rather than wait for vhost_worker(). > > > > Right? > > With the current code, the answer is no. We would hang like I mentioned here: > > https://lore.kernel.org/lkml/ae250076-7d55-c407-1066-86b37014c69c@xxxxxxxxxx/ If only I could fully understand this email ;) Could you spell to explain why this can't work (again, in this simple case) ? My current (and I know, very poor) understanding is that .release() should roughly do the following: 1. Ensure that vhost_work_queue() can't add the new callbacks 2. Call vhost_dev_flush() to ensure that worker->work_list is empty 3. Call vhost_task_stop() so why this sequence can't work if we turn vhost_dev_flush() into something like void vhost_dev_flush(struct vhost_dev *dev) { struct vhost_flush_struct flush; if (dev->worker) { // this assumes that vhost_task_create() uses CLONE_THREAD if (same_thread_group(current, dev->worker->vtsk->task)) { ... run the pending callbacks ... return; } // this is what we currently have init_completion(&flush.wait_event); vhost_work_init(&flush.work, vhost_flush_work); vhost_work_queue(dev, &flush.work); wait_for_completion(&flush.wait_event); } } ? Mike, I am just trying to understand what exactly vhost_worker() should do. > We need to add code like I mentioned in that reply because we don't have a > way to call into the layers below us to flush those commands. This tells me nothing, but this is my fault, not yours. Again, again, I know nothing about drivers/vhost. Oleg. _______________________________________________ Virtualization mailing list Virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linuxfoundation.org/mailman/listinfo/virtualization
