From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-ie0-f176.google.com ([209.85.223.176]:55498 "EHLO mail-ie0-f176.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751314AbaGEQ1L (ORCPT ); Sat, 5 Jul 2014 12:27:11 -0400 Received: by mail-ie0-f176.google.com with SMTP id at1so2266653iec.7 for ; Sat, 05 Jul 2014 09:27:11 -0700 (PDT) Date: Sat, 5 Jul 2014 10:27:08 -0600 From: Bjorn Helgaas To: Guo Chao Cc: linux-pci@vger.kernel.org Subject: Re: [PATCH v2] PCI: don't UNSET valid resources when reassign fails Message-ID: <20140705162708.GA6247@google.com> References: <1403179096-12113-1-git-send-email-yan@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <1403179096-12113-1-git-send-email-yan@linux.vnet.ibm.com> Sender: linux-pci-owner@vger.kernel.org List-ID: On Thu, Jun 19, 2014 at 07:58:16PM +0800, Guo Chao wrote: > Commit bd064f0a sets IORESOURCE_UNSET flag if resources can not be > assgined. Part of these changes breaks resource assignment. > > If resource is too big, PCI core assgins basic part first and > extends (reassigns) it to include optional part (notably SR-IOV resources). > In this case, IORESOURCE_UNSET should not be set if resource is properly > assigned before reassigning. Otherwise, these resources are never updated > to hardware and we will see errors like this: > > pci 0003:00:00.0: can't enable device: BAR 15 [mem size 0x0c000000 64bit pref] not assigned > pci 0003:00:00.0: Error enabling bridge (-22), continuing > > Signed-off-by: Guo Chao I propose the following simpler patch. Can you verify that it works for you? commit 57ece376a3b9e013277032f473197052f1144142 Author: Guo Chao Date: Thu Jul 3 18:30:29 2014 -0600 PCI: Keep original resource if we fail to expand it If we have space assigned to a resource, we try to expand the resource (e.g., to accommodate SR-IOV resources), and the expansion attempt fails, we should keep the original assignment. After bd064f0a231a ("PCI: Mark resources as IORESOURCE_UNSET if we can't assign them"), we left the resource marked IORESOURCE_UNSET when the expansion failed, even if it had originally been set. That caused errors like this: pci 0003:00:00.0: can't enable device: BAR 15 [mem size 0x0c000000 64bit pref] not assigned pci 0003:00:00.0: Error enabling bridge (-22), continuing Fix this by restoring the original flags when reassignment fails. [bhelgaas: reworked to simplify, changelog] Fixes: bd064f0a231a ("PCI: Mark resources as IORESOURCE_UNSET if we can't assign them") Signed-off-by: Guo Chao Signed-off-by: Bjorn Helgaas CC: stable@vger.kernel.org # v3.15+ diff --git a/drivers/pci/setup-res.c b/drivers/pci/setup-res.c index caed1ce6facd..9faebcfeec4d 100644 --- a/drivers/pci/setup-res.c +++ b/drivers/pci/setup-res.c @@ -320,9 +320,11 @@ int pci_reassign_resource(struct pci_dev *dev, int resno, resource_size_t addsiz resource_size_t min_align) { struct resource *res = dev->resource + resno; + unsigned long flags; resource_size_t new_size; int ret; + flags = res->flags; res->flags |= IORESOURCE_UNSET; if (!res->parent) { dev_info(&dev->dev, "BAR %d: can't reassign an unassigned resource %pR\n", @@ -339,7 +341,12 @@ int pci_reassign_resource(struct pci_dev *dev, int resno, resource_size_t addsiz dev_info(&dev->dev, "BAR %d: reassigned %pR\n", resno, res); if (resno < PCI_BRIDGE_RESOURCES) pci_update_resource(dev, resno); + } else { + res->flags = flags; + dev_info(&dev->dev, "BAR %d: %pR (failed to expand by %#llx)\n", + resno, res, addsize); } + return ret; }