From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:54100) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1airlF-0007Cw-50 for qemu-devel@nongnu.org; Wed, 23 Mar 2016 19:00:22 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1airlC-0007dg-Df for qemu-devel@nongnu.org; Wed, 23 Mar 2016 19:00:21 -0400 Received: from mx1.redhat.com ([209.132.183.28]:39895) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1airlC-0007dT-6Z for qemu-devel@nongnu.org; Wed, 23 Mar 2016 19:00:18 -0400 Received: from int-mx11.intmail.prod.int.phx2.redhat.com (int-mx11.intmail.prod.int.phx2.redhat.com [10.5.11.24]) by mx1.redhat.com (Postfix) with ESMTPS id 52D06C049D5C for ; Wed, 23 Mar 2016 23:00:17 +0000 (UTC) Date: Wed, 23 Mar 2016 17:00:16 -0600 From: Alex Williamson Message-ID: <20160323170016.4561f075@t450s.home> In-Reply-To: References: MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH] vfio: convert to 128 bit arithmetic calculations when adding mem regions List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Bandan Das Cc: qemu-devel@nongnu.org On Wed, 23 Mar 2016 16:57:35 -0400 Bandan Das wrote: > vfio_listener_region_add for a iommu mr results in > an overflow assert since vt-d memory region is initialized > with UINT64_MAX. Convert calculations to 128 bit arithmetic > for such regions and let int128_get64 assert for non iommu > regions only. > > Suggested-by: Alex Williamson > Signed-off-by: Bandan Das > --- > This is v2 of http://patchwork.ozlabs.org/patch/600364/ > hw/vfio/common.c | 21 ++++++++++++--------- > 1 file changed, 12 insertions(+), 9 deletions(-) > > diff --git a/hw/vfio/common.c b/hw/vfio/common.c > index fb588d8..9489ff3 100644 > --- a/hw/vfio/common.c > +++ b/hw/vfio/common.c > @@ -323,7 +323,7 @@ static void vfio_listener_region_add(MemoryListener *listener, > { > VFIOContainer *container = container_of(listener, VFIOContainer, listener); > hwaddr iova, end; > - Int128 llend; > + Int128 llend, endaddr; > void *vaddr; > int ret; > > @@ -349,14 +349,15 @@ static void vfio_listener_region_add(MemoryListener *listener, > if (int128_ge(int128_make64(iova), llend)) { > return; > } > - end = int128_get64(llend); > + endaddr = int128_sub(llend, int128_one()); I like where this is headed, but there's maybe a little polish left. endaddr here can never be greater than UINT64_MAX, so why don't we make it a hwaddr and wrap the whole right side of the equation in an int128_get64(). Then the (end - 1) changes below don't need the extra int128 ops. > > - if ((iova < container->min_iova) || ((end - 1) > container->max_iova)) { > - error_report("vfio: IOMMU container %p can't map guest IOVA region" > - " 0x%"HWADDR_PRIx"..0x%"HWADDR_PRIx, > - container, iova, end - 1); > - ret = -EFAULT; > - goto fail; > + if ((iova < container->min_iova) || (int128_get64(endaddr) > > + container->max_iova)) { > + error_report("vfio: IOMMU container %p can't map guest IOVA region" > + " 0x%"HWADDR_PRIx"..0x%"HWADDR_PRIx, > + container, iova, int128_get64(endaddr)); > + ret = -EFAULT; > + goto fail; Bogus whitespace changes > } > > memory_region_ref(section->mr); > @@ -364,7 +365,7 @@ static void vfio_listener_region_add(MemoryListener *listener, > if (memory_region_is_iommu(section->mr)) { > VFIOGuestIOMMU *giommu; > > - trace_vfio_listener_region_add_iommu(iova, end - 1); > + trace_vfio_listener_region_add_iommu(iova, int128_get64(endaddr)); > /* > * FIXME: We should do some checking to see if the > * capabilities of the host VFIO IOMMU are adequate to model > @@ -389,6 +390,8 @@ static void vfio_listener_region_add(MemoryListener *listener, > return; > } > > + end = int128_get64(llend); > + > /* Here we assume that memory_region_is_ram(section->mr)==true */ > > vaddr = memory_region_get_ram_ptr(section->mr) + For this last chunk I'd do something like this: + llsize = int128_sub(llend, int128_make64(iova)); - ret = vfio_dma_map(container, iova, end - iova, vaddr, section->readonly); + ret = vfio_dma_map(container, iova, + int128_get64(llsize), vaddr, section->readonly); if (ret) { error_report("vfio_dma_map(%p, 0x%"HWADDR_PRIx", " "0x%"HWADDR_PRIx", %p) = %d (%m)", - container, iova, end - iova, vaddr, ret); + container, iova, int128_get64(llsize), vaddr, ret); goto fail; } This maintains the 128bit math until exactly the point where it will fall apart when vfio_dma_map can't pass an int128 to the vfio API. Thanks, Alex