[PATCH v2] PCI: disable SERR for kdump kernel

From: Yinghai Lu
Date: Tue Apr 18 2017 - 20:32:40 EST


Found one system with infiniband with SRIOV enabled, kdump kernel
SRIOV BAR probing trigger one pci fatal error.
That assert error pin, and host get reset by BMC.

We can just ignore that error to let kernel go on
and kdump to create vmcore.

-v2: add debug print out

Signed-off-by: Yinghai Lu <yinghai@xxxxxxxxxx>

---
drivers/pci/probe.c | 23 +++++++++++++++++++++++
1 file changed, 23 insertions(+)

Index: linux-2.6/drivers/pci/probe.c
===================================================================
--- linux-2.6.orig/drivers/pci/probe.c
+++ linux-2.6/drivers/pci/probe.c
@@ -17,6 +17,8 @@
#include <linux/acpi.h>
#include <linux/irqdomain.h>
#include <linux/pm_runtime.h>
+#include <linux/crash_dump.h>
+
#include "pci.h"

#define CARDBUS_LATENCY_TIMER 176 /* secondary latency timer */
@@ -1515,6 +1517,24 @@ static void pci_msi_setup_pci_dev(struct
pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_ENABLE, 0);
}

+static void pci_disable_serr(struct pci_dev *dev)
+{
+ u16 pci_cmd, pci_bctl;
+
+ pci_read_config_word(dev, PCI_COMMAND, &pci_cmd);
+ pci_cmd &= ~PCI_COMMAND_SERR;
+ pci_write_config_word(dev, PCI_COMMAND, pci_cmd);
+ dev_printk(KERN_DEBUG, &dev->dev, "SERR cleared\n");
+
+ /* Program bridge control value */
+ if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI) {
+ pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &pci_bctl);
+ pci_bctl &= ~PCI_BRIDGE_CTL_SERR;
+ pci_write_config_word(dev, PCI_BRIDGE_CONTROL, pci_bctl);
+ dev_printk(KERN_DEBUG, &dev->dev, "BRIDGE SERR cleared\n");
+ }
+}
+
/**
* pci_setup_device - fill in class and map information of a device
* @dev: the device structure to fill
@@ -1572,6 +1592,9 @@ int pci_setup_device(struct pci_dev *dev
/* device class may be changed after fixup */
class = dev->class >> 8;

+ if (is_kdump_kernel())
+ pci_disable_serr(dev);
+
if (dev->non_compliant_bars) {
pci_read_config_word(dev, PCI_COMMAND, &cmd);
if (cmd & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) {