Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753380AbdHRNmb (ORCPT ); Fri, 18 Aug 2017 09:42:31 -0400 Received: from shadbolt.e.decadent.org.uk ([88.96.1.126]:55390 "EHLO shadbolt.e.decadent.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752746AbdHRNPq (ORCPT ); Fri, 18 Aug 2017 09:15:46 -0400 Content-Type: text/plain; charset="UTF-8" Content-Disposition: inline Content-Transfer-Encoding: 8bit MIME-Version: 1.0 From: Ben Hutchings To: linux-kernel@vger.kernel.org, stable@vger.kernel.org CC: akpm@linux-foundation.org, "Bjorn Helgaas" Date: Fri, 18 Aug 2017 14:13:20 +0100 Message-ID: X-Mailer: LinuxStableQueue (scripts by bwh) Subject: [PATCH 3.16 057/134] PCI: Ignore write combining when mapping I/O port space In-Reply-To: X-SA-Exim-Connect-IP: 82.70.136.246 X-SA-Exim-Mail-From: ben@decadent.org.uk X-SA-Exim-Scanned: No (on shadbolt.decadent.org.uk); SAEximRunCond expanded to false Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2620 Lines: 77 3.16.47-rc1 review patch. If anyone has any objections, please let me know. ------------------ From: Bjorn Helgaas commit 3a92c319c44a7bcee9f48dff9d97d001943b54c6 upstream. PCI exposes files like /proc/bus/pci/00/00.0 in procfs. These files support operations like this: ioctl(fd, PCIIOC_MMAP_IS_IO); # request I/O port space ioctl(fd, PCIIOC_WRITE_COMBINE, 1); # request write-combining mmap(fd, ...) Write combining is useful on PCI memory space, but I don't think it makes sense on PCI I/O port space. We *could* change proc_bus_pci_ioctl() to make it impossible to set mmap_state == pci_mmap_io and write_combine at the same time, but that would break the following sequence, which is currently legal: mmap(fd, ...) # default is I/O, non-combining ioctl(fd, PCIIOC_WRITE_COMBINE, 1); # request write-combining ioctl(fd, PCIIOC_MMAP_IS_MEM); # request memory space mmap(fd, ...) # get write-combining mapping Ignore the write-combining flag when mapping I/O port space. This patch should have no functional effect, based on this analysis of all implementations of pci_mmap_page_range(): - ia64 mips parisc sh unicore32 x86 do not support mapping of I/O port space at all. - arm cris microblaze mn10300 sparc xtensa support mapping of I/O port space, but ignore the write_combine argument to pci_mmap_page_range(). - powerpc supports mapping of I/O port space and uses write_combine, and it disables write combining for I/O port space in __pci_mmap_set_pgprot(). This patch makes it possible to remove __pci_mmap_set_pgprot() from powerpc, which simplifies that path. Signed-off-by: Bjorn Helgaas Signed-off-by: Ben Hutchings --- drivers/pci/proc.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) --- a/drivers/pci/proc.c +++ b/drivers/pci/proc.c @@ -231,7 +231,7 @@ static int proc_bus_pci_mmap(struct file { struct pci_dev *dev = PDE_DATA(file_inode(file)); struct pci_filp_private *fpriv = file->private_data; - int i, ret; + int i, ret, write_combine; if (!capable(CAP_SYS_RAWIO)) return -EPERM; @@ -245,9 +245,12 @@ static int proc_bus_pci_mmap(struct file if (i >= PCI_ROM_RESOURCE) return -ENODEV; + if (fpriv->mmap_state == pci_mmap_mem) + write_combine = fpriv->write_combine; + else + write_combine = 0; ret = pci_mmap_page_range(dev, vma, - fpriv->mmap_state, - fpriv->write_combine); + fpriv->mmap_state, write_combine); if (ret < 0) return ret;