From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chris Lalancette Subject: [PATCH v2 0/3]: Fixes to IRQ routing Date: Mon, 14 Jun 2010 13:11:19 -0400 Message-ID: <1276535482-11965-1-git-send-email-clalance@redhat.com> To: kvm@vger.kernel.org Return-path: Received: from mx1.redhat.com ([209.132.183.28]:51366 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753513Ab0FNRNm (ORCPT ); Mon, 14 Jun 2010 13:13:42 -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 o5EHDfLc027571 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Mon, 14 Jun 2010 13:13:41 -0400 Received: from localhost.localdomain (dhcp-100-19-171.bos.redhat.com [10.16.19.171]) by int-mx02.intmail.prod.int.phx2.redhat.com (8.13.8/8.13.8) with ESMTP id o5EHDfRN007238 for ; Mon, 14 Jun 2010 13:13:41 -0400 Sender: kvm-owner@vger.kernel.org List-ID: As we've discussed previously, here is a series of patches to fix some of the IRQ routing issues we have in KVM. With this series in place I was able to successfully kdump a RHEL-5 64-bit, and RHEL-6 32- and 64-bit guest on CPU's other than the BSP. RHEL-5 32-bit kdump still does not work; it gets stuck on "Checking 'hlt' instruction". However, it does that both before and after this series, so there is something else going on there that I still have to debug. I also need to change the "kvm_migrate_pit_timer" function to migrate the timer over to the last CPU that handled the timer interrupt, on the theory that that particlar CPU is likely to handle the timer interrupt again in the near future. However, this is an optimization and shouldn't delay the inclusion of the rest of the series for correctness. Changes since RFC: - Changed ps->inject_lock from raw_spinlock_t to spinlock_t - Fixed up some formatting issues - Changed to have one PIT workqueue per-guest - Remember to cancel_work_sync when destroying the PIT Changes since v1: - Call cancel_work_sync everywhere we call hrtimer_cancel - Bring back the reinjection logic - Fix up formatting issues from checkpatch