Hello, On Wed, Oct 30, 2024 at 01:05:16PM +0100, Paolo Bonzini wrote: > On Wed, Oct 30, 2024 at 1:25 AM Tejun Heo <tj@xxxxxxxxxx> wrote: > > > I'm not sure if the KVM worker thread should process signals. We want it > > > to take the CPU time it uses from the guest, but otherwise it's not running > > > on behalf of userspace in the way that io_wq_worker() is. > > > > I see, so io_wq_worker()'s handle signals only partially. It sets > > PF_USER_WORKER which ignores fatal signals, so the only signals which take > > effect are STOP/CONT (and friends) which is handled in do_signal_stop() > > which is also where the cgroup2 freezer is implemented. > > What about SIGKILL? That's the one that I don't want to have for KVM > workers, because they should only stop when the file descriptor is > closed. I don't think SIGKILL does anything for PF_USER_WORKER threads. Those are all handled in the fatal: label in kernel/signal.c::get_signal() and the function just returns for PF_USER_WORKER threads. I haven't used it myself but looking at io_uring usage, it seems pretty straightforward. > (Replying to Luca: the kthreads are dropping some internal data > structures that KVM had to "de-optimize" to deal with processor bugs. > They allow the data structures to be rebuilt in the optimal way using > large pages). > > > Given that the kthreads are tied to user processes, I think it'd be better > > to behave similarly to user tasks as possible in this regard if userspace > > being able to stop/cont these kthreads are okay. > > Yes, I totally agree with you on that, I'm just not sure of the best > way to do it. > > I will try keeping the kthread and adding allow_signal(SIGSTOP). That > should allow me to process the SIGSTOP via get_signal(). I *think* you can just copy what io_wq_worker() is doing. Thanks. -- tejun
