From: Sean Christopherson Subject: Re: [PATCH v3] KVM: Move VM's worker kthreads back to the original cgroup before exiting. Date: Sat, 19 Feb 2022 00:30:56 +0000 Message-ID: References: <20220217061616.3303271-1-vipinsh@google.com> <202202172046.GuW8pHQc-lkp@intel.com> <3113f00a-e910-2dfb-479f-268566445630@redhat.com> Mime-Version: 1.0 Return-path: DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=date:from:to:cc:subject:message-id:references:mime-version :content-disposition:in-reply-to; bh=8aZ16UFG/qN8SDATqq+xbNjeyEK8F2U4AYeI8Q7W9Mw=; b=HArcINaTCCHq36tGrqoEk7mWbIGxWYrZtNbvjNwmDNrmBHrKF3eJlEmgsdsTdOVnMq gwU0FIiZ3lb3nTu1nqY4vVwwJnHUzAC66FxLehc7wPjjnFMa8nrC2vIKci/NI+cgdx0O 5mv4Zjt08mwVjFMhablIrVrKKZ0tJ7ZkS+DWkG6oHr8ISvrHErku953L0IAfmJ5H1IwE tpcPZ0e+eZl3ta4rPpfIMbwMD/rATwJYXVln33HESHVem5AggYKENrLDu9hCiM6AeD+L 65Y6x9O5zH7k0GE7uQdIZ4ZIvC3mSThjkE1YaZhLPfx0/4C+JHfEqBOdh+HVxKvriHUS 0uSA== Content-Disposition: inline In-Reply-To: <3113f00a-e910-2dfb-479f-268566445630-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Paolo Bonzini Cc: kernel test robot , Vipin Sharma , kbuild-all-hn68Rpc1hR1g9hUCZPvPmw@public.gmane.org, mkoutny-IBi9RG/b67k@public.gmane.org, tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org, lizefan.x-EC8Uxl6Npydl57MIdRCFDg@public.gmane.org, hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org, dmatlack-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org, jiangshanlai-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org, kvm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org On Thu, Feb 17, 2022, Paolo Bonzini wrote: > On 2/17/22 13:34, kernel test robot wrote: > > > 5859 reattach_err = cgroup_attach_task_all(current->real_parent, current); > > This needs to be within rcu_dereference(). As Vipin found out the hard way, cgroup_attach_task_all() can't be inside an RCU critical section as it might sleep due to mutex_lock(). My sched knowledge is sketchy at best, but I believe KVM just needs to guarantee the liveliness of real_parent when passing it to cgroup_attach_task_all(). Presumably kthreadd_task can (in theory) be killed/exiting while this is going on, but at that point I doubt anyone cares much about cgroup accuracy. So I think this can be: rcu_read_lock(); parent = rcu_dereference(current->real_parent); get_task_struct(parent); rcu_read_unlock(); (void)cgroup_attach_task_all(parent, current); put_task_struct(parent); > > 5860 if (reattach_err) { > > 5861 kvm_err("%s: cgroup_attach_task_all failed on reattach with err %d\n", > > 5862 __func__, reattach_err); Eh, I wouldn't bother logging the error, userspace can't take action on it, and if the system is OOM it's just more noise in the log to dig through.