From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marcelo Tosatti Subject: Re: [PATCH v4 0/6] Nonatomic interrupt injection Date: Fri, 17 Sep 2010 16:12:54 -0300 Message-ID: <20100917191254.GA22281@amt.cnet> References: <1283168222-2242-1-git-send-email-avi@redhat.com> <4C921D17.3010603@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: kvm@vger.kernel.org To: Avi Kivity Return-path: Received: from mx1.redhat.com ([209.132.183.28]:25403 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755010Ab0IQTNS (ORCPT ); Fri, 17 Sep 2010 15:13:18 -0400 Received: from int-mx02.intmail.prod.int.phx2.redhat.com (int-mx02.intmail.prod.int.phx2.redhat.com [10.5.11.12]) by mx1.redhat.com (8.13.8/8.13.8) with ESMTP id o8HJDIvf029388 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Fri, 17 Sep 2010 15:13:18 -0400 Content-Disposition: inline In-Reply-To: <4C921D17.3010603@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: On Thu, Sep 16, 2010 at 03:35:19PM +0200, Avi Kivity wrote: > On 08/30/2010 02:36 PM, Avi Kivity wrote: > >This patchset changes interrupt injection to be done from normal process > >context instead of interrupts disabled context. This is useful for real > >mode interrupt injection on Intel without the current hacks (injecting as > >a software interrupt of a vm86 task), reducing latencies, and later, for > >allowing nested virtualization code to use kvm_read_guest()/kvm_write_guest() > >instead of kmap() to access the guest vmcb/vmcs. > > > >Seems to survive a hack that cancels every 16th entry, after injection has > >already taken place. > > > >With the PIC reset fix posted earlier, this passes autotest on both AMD and > >Intel, with in-kernel irqchip. I'll run -no-kvm-irqchip tests shortly. > > > >Please review carefully, esp. the first patch. Any missing kvm_make_request() > >there may result in a hung guest. > > > > This is now merged, with the change pointed out by Marcelo. Windows > XP x64 fails installation without > > (vmx.c handle_cr()) > case 8: { > u8 cr8_prev = kvm_get_cr8(vcpu); > u8 cr8 = kvm_register_read(vcpu, reg); > kvm_set_cr8(vcpu, cr8); > skip_emulated_instruction(vcpu); > if (irqchip_in_kernel(vcpu->kvm)) > return 1; > - if (cr8_prev <= cr8) > - return 1; > vcpu->run->exit_reason = KVM_EXIT_SET_TPR; > return 0; > } > > Which doesn't make any sense (anyone?). The failure is present even > without the patchset, and is fixed by the same hack, so a regression > was not introduced. If userspace does not have an uptodate TPR value, it can signal an interrupt that is now blocked? Say: - cr8 write 0 - cr8 write 5 - no exit to userspace - userspace signals interrupt with priority 4 because it knows about tpr == 0.