From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Yang, Sheng" Subject: Re: assigned dev msi int handling Date: Fri, 27 Feb 2009 12:26:44 +0800 Message-ID: <200902271226.44756.sheng.yang@intel.com> References: <20090226235927.GA4270@amt.cnet> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: "kvm-devel" To: Marcelo Tosatti Return-path: Received: from mga03.intel.com ([143.182.124.21]:7203 "EHLO mga03.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754230AbZB0E0w (ORCPT ); Thu, 26 Feb 2009 23:26:52 -0500 In-Reply-To: <20090226235927.GA4270@amt.cnet> Content-Disposition: inline Sender: kvm-owner@vger.kernel.org List-ID: On Friday 27 February 2009 07:59:27 Marcelo Tosatti wrote: > Hi Sheng, > > So for guest INTX interrupts the host interrupt is reenabled on ack from > the guest, which is nice. Now for guest MSI interrupts it keeps reenabling > the interrupt as fast as the work handler can run. > > Can you explain why it works this way? Why not disable interrupts > on the host in all cases and only reenable on ack? > Sorry for I didn't think it over... The direct reason is ack_irq binding with kvm_set_irq() which is unnecessary for MSI/MSI-X. But enable(msi) after EOI seems more proper here, though more changes are needed for MSI-X(seems we need one ack notifier for one vector in MSI-X). -- regards Yang, Sheng