From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:41048) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VRm8B-0000SA-3l for qemu-devel@nongnu.org; Thu, 03 Oct 2013 12:52:09 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1VRm85-0008QD-59 for qemu-devel@nongnu.org; Thu, 03 Oct 2013 12:52:03 -0400 Received: from mx1.redhat.com ([209.132.183.28]:33188) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VRm84-0008Q7-Tr for qemu-devel@nongnu.org; Thu, 03 Oct 2013 12:51:57 -0400 Received: from int-mx10.intmail.prod.int.phx2.redhat.com (int-mx10.intmail.prod.int.phx2.redhat.com [10.5.11.23]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id r93GptY2010290 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Thu, 3 Oct 2013 12:51:55 -0400 Date: Thu, 3 Oct 2013 19:54:17 +0300 From: "Michael S. Tsirkin" Message-ID: <20131003165417.GA4616@redhat.com> References: <1380807975-13266-1-git-send-email-pbonzini@redhat.com> <20131003135404.GA1267@redhat.com> <524D9426.20707@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <524D9426.20707@redhat.com> Subject: Re: [Qemu-devel] [PATCH 0/4] qdev: switch reset to post-order, clean up PCI reset List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Paolo Bonzini Cc: qemu-devel@nongnu.org On Thu, Oct 03, 2013 at 05:58:30PM +0200, Paolo Bonzini wrote: > Il 03/10/2013 15:54, Michael S. Tsirkin ha scritto: > > On Thu, Oct 03, 2013 at 03:46:11PM +0200, Paolo Bonzini wrote: > >> PCI is handling resetting of its devices before the bus is reset, > >> but this is only necessary because qdev is broken and usually does > >> pre-order reset. Post-order is a much better definition. Drop > >> the unnecessary flexibility that lets bus decide the reset order, > >> convert to post-order, and make PCI use common code for reset. > > > > Need to go carefully here. I remember a bunch of targets > > were relying on reset in this order, though don't > > have the detail right now. > > The main change is that sub-devices, which basically exist only for USB, > virtio, and SCSI, are reset _before_ the corresponding parent device > (the USB host controller, virtio proxy device, or SCSI HBA). > > For SCSI, this patch fixes bugs that I'm currently working around in > virtio-scsi (see the manual reset in virtio_scsi_reset, that manually > resets the subdevices). > > USB and virtio are not using qdev reset at all, so they are unaffected > by the patches. They rely on the parent device's own reset method to > propagate the reset. virtio already uses post-order, so it could be > changed to use qdev reset, but I don't plan to do it unless it fixes > bugs. USB is more complicated and I don't really understand it. > > For PCI, the change is that all devices are reset first, and then the > interrupts state and config space are reset for all devices. > > > What kind of testing did this patchset go through? > > For each PCI device I tried creating a VM with an instance of it (a few > devices at a time), and did VM resets. Earlier versions were tested by > the guy who reported the SCSI problems. > > Paolo x86 kvm only? > >> Paolo Bonzini (4): > >> pci: do not export pci_bus_reset > >> pci: clean up resetting of IRQs > >> qdev: allow both pre- and post-order vists in qdev walking functions > >> qdev: switch reset to post-order > >> > >> hw/core/qdev.c | 47 ++++++++++++++++++++++++++++++++++------------- > >> hw/pci/pci.c | 42 ++++++++++++++++++++---------------------- > >> hw/pci/pci_bridge.c | 2 +- > >> include/hw/pci/pci.h | 1 - > >> include/hw/qdev-core.h | 15 ++++++++++----- > >> 5 files changed, 65 insertions(+), 42 deletions(-) > > > >