From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-pci-owner@vger.kernel.org>
Received: from userp1040.oracle.com ([156.151.31.81]:25645 "EHLO
	userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1756101AbaJ2WZc (ORCPT
	<rfc822;linux-pci@vger.kernel.org>); Wed, 29 Oct 2014 18:25:32 -0400
From: Yinghai Lu <yinghai@kernel.org>
To: Bjorn Helgaas <bhelgaas@google.com>
Cc: linux-pci@vger.kernel.org, linux-kernel@vger.kernel.org,
	Yinghai Lu <yinghai@kernel.org>
Subject: [PATCH] PCI: fix sriov enabling with virtual bus
Date: Wed, 29 Oct 2014 15:26:10 -0700
Message-Id: <1414621570-20777-1-git-send-email-yinghai@kernel.org>
Sender: linux-pci-owner@vger.kernel.org
List-ID: <linux-pci.vger.kernel.org>

While enabling sriov for intel igb device got:

sca05-0a81fda5:~ # echo 7 > /sys/bus/pci/devices/0000\:09\:00.0/sriov_numvfs 
[  729.612191] pci 0000:0a:10.0: [8086:1520] type 00 class 0x020000
[  729.619002] BUG: unable to handle kernel NULL pointer dereference at 00000000000000e8
[  729.627767] IP: [<ffffffff815901cb>] pci_get_hp_params+0x5b/0x640
[  729.634593] PGD 0 
[  729.636844] Oops: 0000 [#1] SMP 
[  729.640466] Modules linked in:
...
[  729.795268] Call Trace:
[  729.798007]  [<ffffffff81ea0843>] ? pci_mmcfg_read+0x123/0x140
[  729.804519]  [<ffffffff81ea0770>] ? pci_mmcfg_read+0x50/0x140
[  729.810942]  [<ffffffff815692a3>] pci_configure_device+0x33/0x350
[  729.817744]  [<ffffffff8156aba4>] pci_device_add+0x24/0x160
[  729.823965]  [<ffffffff8158f3bb>] pci_enable_sriov+0x4db/0x7d0
[  729.830486]  [<ffffffff81b63f54>] ? igb_pci_enable_sriov+0xe4/0x200
[  729.837481]  [<ffffffff81b63f7f>] igb_pci_enable_sriov+0x10f/0x200
[  729.844386]  [<ffffffff8155129c>] ? _kstrtoull+0x2c/0x80
[  729.850315]  [<ffffffff81b640a5>] igb_pci_sriov_configure+0x35/0x40
[  729.857318]  [<ffffffff815752b5>] sriov_numvfs_store+0xe5/0x140
[  729.863934]  [<ffffffff817e1f88>] dev_attr_store+0x18/0x30
[  729.870063]  [<ffffffff812624d8>] sysfs_kf_write+0x48/0x60
[  729.876186]  [<ffffffff812617ef>] ? kernfs_fop_write+0xaf/0x170
[  729.882797]  [<ffffffff81261827>] kernfs_fop_write+0xe7/0x170
[  729.889222]  [<ffffffff811ef66b>] vfs_write+0xcb/0x1c0
[  729.894958]  [<ffffffff811f0019>] SyS_write+0x49/0xb0
...
[  729.943531] ---[ end trace 7cf0cdb66637665a ]---

and pci_get_hp_params+0x5b point to
0xffffffff815901cb is in pci_get_hp_params (include/linux/device.h:815).
810	#include <linux/pm_wakeup.h>
811	
812	static inline const char *dev_name(const struct device *dev)
813	{
814		/* Use the init name until the kobject becomes available */
815		if (dev->init_name)
816			return dev->init_name;
817	
818		return kobject_name(&dev->kobj);

The root cause:
Now pci_configure_device/pci_get_hp_params will be called for every pci_dev,
including VF that is under virtual bus. But virtual bus does not have bridge
set. So we can not refer pbus->self->dev directly.

Add checking with pbus->slef and bail out early.

Fixing: commit 6cd33649fa83 ("PCI: Add pci_configure_device() during enumeration")

Signed-off-by: Yinghai Lu <yinghai@kernel.org>

---
 include/linux/pci-acpi.h |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Index: linux-2.6/include/linux/pci-acpi.h
===================================================================
--- linux-2.6.orig/include/linux/pci-acpi.h
+++ linux-2.6/include/linux/pci-acpi.h
@@ -44,8 +44,12 @@ static inline acpi_handle acpi_pci_get_b
 
 	if (pci_is_root_bus(pbus))
 		dev = pbus->bridge;
-	else
+	else {
+		if (!pbus->self)
+			return NULL;
+
 		dev = &pbus->self->dev;
+	}
 
 	dev_printk(KERN_DEBUG, &pbus->dev, "  bridge ACPI_HANDLE %p : %s\n",
 			 ACPI_HANDLE(dev), dev_name(dev));