From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:39309) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1cjLT5-0006IC-C5 for qemu-devel@nongnu.org; Thu, 02 Mar 2017 02:48:08 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1cjLT2-0004Av-9r for qemu-devel@nongnu.org; Thu, 02 Mar 2017 02:48:07 -0500 Received: from mx1.redhat.com ([209.132.183.28]:50080) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1cjLT2-0004Ao-3v for qemu-devel@nongnu.org; Thu, 02 Mar 2017 02:48:04 -0500 Received: from int-mx13.intmail.prod.int.phx2.redhat.com (int-mx13.intmail.prod.int.phx2.redhat.com [10.5.11.26]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 152867E9D5 for ; Thu, 2 Mar 2017 07:48:04 +0000 (UTC) References: <1488341440-4347-1-git-send-email-jasowang@redhat.com> <20170302051016.GA3893@redhat.com> From: Jason Wang Message-ID: <3c36e5ad-1244-ca1c-ca3c-5cab9e5cdda4@redhat.com> Date: Thu, 2 Mar 2017 15:47:57 +0800 MIME-Version: 1.0 In-Reply-To: <20170302051016.GA3893@redhat.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH] virtio: unbreak virtio-pci with IOMMU after caching ring translations List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: "Michael S. Tsirkin" Cc: qemu-devel@nongnu.org, peterx@redhat.com, wexu@redhat.com, Paolo Bonzini On 2017=E5=B9=B403=E6=9C=8802=E6=97=A5 13:10, Michael S. Tsirkin wrote: > On Wed, Mar 01, 2017 at 12:10:40PM +0800, Jason Wang wrote: >> Commit c611c76417f5 ("virtio: add MemoryListener to cache ring >> translations") registers a memory listener to dma_as. This may not >> work when IOMMU is enabled: dma_as(bus_master_as) were initialized in >> pcibus_machine_done() after virtio_realize(). This will cause a >> segfault. Fixing this by using pci_device_iommu_address_space() >> instead to make sure address space were initialized at this time. >> >> With this fix, IOMMU device were required to be initialized before any >> virtio-pci devices. >> >> Fixes: c611c76417f5 ("virtio: add MemoryListener to cache ring transla= tions") >> Cc: Paolo Bonzini >> Signed-off-by: Jason Wang > This is very ugly. I guess it's better than broken IOMMU ... > Paolo? Maybe we can delay the registering of memory listener on bus master=20 enabling or status setting. Thanks > >> --- >> hw/virtio/virtio-pci.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c >> index 5ce42af..b76f3f6 100644 >> --- a/hw/virtio/virtio-pci.c >> +++ b/hw/virtio/virtio-pci.c >> @@ -1153,7 +1153,7 @@ static AddressSpace *virtio_pci_get_dma_as(Devic= eState *d) >> VirtIOPCIProxy *proxy =3D VIRTIO_PCI(d); >> PCIDevice *dev =3D &proxy->pci_dev; >> =20 >> - return pci_get_address_space(dev); >> + return pci_device_iommu_address_space(dev); >> } >> =20 >> static int virtio_pci_add_mem_cap(VirtIOPCIProxy *proxy, >> --=20 >> 2.7.4