From patchwork Thu Nov 30 15:47:08 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Alex Williamson X-Patchwork-Id: 10085135 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id C58456035E for ; Thu, 30 Nov 2017 15:47:16 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id AD0ED29FEA for ; Thu, 30 Nov 2017 15:47:16 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id A0E0B2A1AA; Thu, 30 Nov 2017 15:47:16 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.9 required=2.0 tests=BAYES_00,RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id B93DF29FEA for ; Thu, 30 Nov 2017 15:47:15 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753166AbdK3PrM (ORCPT ); Thu, 30 Nov 2017 10:47:12 -0500 Received: from mx1.redhat.com ([209.132.183.28]:56152 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753161AbdK3PrJ (ORCPT ); Thu, 30 Nov 2017 10:47:09 -0500 Received: from smtp.corp.redhat.com (int-mx04.intmail.prod.int.phx2.redhat.com [10.5.11.14]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id A105D820E2; Thu, 30 Nov 2017 15:47:09 +0000 (UTC) Received: from t450s.home (ovpn-116-16.phx2.redhat.com [10.3.116.16]) by smtp.corp.redhat.com (Postfix) with ESMTP id 5C1475D960; Thu, 30 Nov 2017 15:47:09 +0000 (UTC) Date: Thu, 30 Nov 2017 08:47:08 -0700 From: Alex Williamson To: Alexey Kardashevskiy Cc: kvm@vger.kernel.org Subject: Re: [RFC PATCH kernel] vfio-pci: Fix sparse capability when no parts of MSIX BAR can be mapped Message-ID: <20171130084708.770cfc76@t450s.home> In-Reply-To: References: <20171123045626.17542-1-aik@ozlabs.ru> <20171129112702.51d0a493@t450s.home> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 2.79 on 10.5.11.14 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.26]); Thu, 30 Nov 2017 15:47:09 +0000 (UTC) Sender: kvm-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP On Thu, 30 Nov 2017 17:00:35 +1100 Alexey Kardashevskiy wrote: > On 30/11/17 05:27, Alex Williamson wrote: > > On Thu, 23 Nov 2017 15:56:26 +1100 > > Alexey Kardashevskiy wrote: > > > >> It is currently possible to have a sparse capability with 1 areas which > >> starts at 0 and 0 bytes long. One example is: > >> > >> Texas Instruments TUSB73x0 SuperSpeed USB 3.0 xHCI Host Controller > >> [...] > >> Region 0: Memory at 3fe280000000 (64-bit, non-prefetchable) [size=64K] > >> Region 2: Memory at 3fe280010000 (64-bit, non-prefetchable) [size=8K] > >> [...] > >> Capabilities: [c0] MSI-X: Enable+ Count=8 Masked- > >> Vector table: BAR=2 offset=00000000 > >> PBA: BAR=2 offset=00001000 > >> > >> With PAGE_SIZE=64K, MSIX BAR occupies the entire BAR2 and cannot be > >> mapped. > >> > >> This makes it explicit - if sparse->areas is empty, then advertise > >> nr_areas as 0. > >> > >> Signed-off-by: Alexey Kardashevskiy > >> --- > >> > >> QEMU gets it right as vfio_setup_region_sparse_mmaps() checks for size > >> after QEMU's 24acf72b9a291ce "vfio: Handle zero-length sparse mmap ranges" > >> but why not make it explicit in the first place? > >> > >> > >> --- > >> drivers/vfio/pci/vfio_pci.c | 4 ++++ > >> 1 file changed, 4 insertions(+) > >> > >> diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c > >> index f041b1a..a201c45 100644 > >> --- a/drivers/vfio/pci/vfio_pci.c > >> +++ b/drivers/vfio/pci/vfio_pci.c > >> @@ -597,6 +597,10 @@ static int msix_sparse_mmap_cap(struct vfio_pci_device *vdev, > >> i++; > >> } > >> > >> + /* With all alignments, there are no gaps left to mmap */ > >> + if (i == 0) > >> + sparse->nr_areas = 0; > >> + > > > > Ok, but why does 0 become a special case? Shouldn't we set > > sparse->nr_areas = i? Thanks, > > This is what is returned to QEMU now - 1 sparse region, starts at 0, 0 > bytes long. I am missing the point in having such region... I'm asking why your patch is necessary vs: diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c index f041b1a6cf66..c062437bbf44 100644 --- a/drivers/vfio/pci/vfio_pci.c +++ b/drivers/vfio/pci/vfio_pci.c @@ -582,8 +582,6 @@ static int msix_sparse_mmap_cap(struct vfio_pci_device *vdev if (!sparse) return -ENOMEM; - sparse->nr_areas = nr_areas; - if (vdev->msix_offset & PAGE_MASK) { sparse->areas[i].offset = 0; sparse->areas[i].size = vdev->msix_offset & PAGE_MASK; @@ -597,6 +595,8 @@ static int msix_sparse_mmap_cap(struct vfio_pci_device *vdev i++; } + sparse->nr_areas = i; + ret = vfio_info_add_capability(caps, VFIO_REGION_INFO_CAP_SPARSE_MMAP, sparse); kfree(sparse); And if we do that, then perhaps we don't even need to calculate nr_areas and we can get rid of more code: diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c index f041b1a6cf66..ae0d5a31aa6b 100644 --- a/drivers/vfio/pci/vfio_pci.c +++ b/drivers/vfio/pci/vfio_pci.c @@ -566,24 +566,16 @@ static int msix_sparse_mmap_cap(struct vfio_pci_device *vd struct vfio_info_cap *caps) { struct vfio_region_info_cap_sparse_mmap *sparse; - size_t end, size; - int nr_areas = 2, i = 0, ret; + size_t end; + int i = 0, ret; end = pci_resource_len(vdev->pdev, vdev->msix_bar); - /* If MSI-X table is aligned to the start or end, only one area */ - if (((vdev->msix_offset & PAGE_MASK) == 0) || - (PAGE_ALIGN(vdev->msix_offset + vdev->msix_size) >= end)) - nr_areas = 1; - - size = sizeof(*sparse) + (nr_areas * sizeof(*sparse->areas)); - - sparse = kzalloc(size, GFP_KERNEL); + sparse = kzalloc(sizeof(*sparse) + (2 * sizeof(*sparse->areas)), + GFP_KERNEL); if (!sparse) return -ENOMEM; - sparse->nr_areas = nr_areas; - if (vdev->msix_offset & PAGE_MASK) { sparse->areas[i].offset = 0; sparse->areas[i].size = vdev->msix_offset & PAGE_MASK; @@ -597,6 +589,8 @@ static int msix_sparse_mmap_cap(struct vfio_pci_device *vdev i++; } + sparse->nr_areas = i; + ret = vfio_info_add_capability(caps, VFIO_REGION_INFO_CAP_SPARSE_MMAP, sparse); kfree(sparse);