[CentOS] PCIe AER support in 5.5

Wed Feb 8 03:15:17 UTC 2012
Baylis, Neil <neil.baylis at hp.com>

I have a question about getting PCIe advanced error reporting to work on 5.5.

The info I need to start with is the following:

1) Is PCIe Advanced Error Reporting (including error injections) known to work in 5.5?
2) What configuration options do I need to select in order to enable it?

I'm using the following configs:

CONFIG_ACPI=y
CONFIG_PCI=y
CONFIG_PCI_DIRECT=y
CONFIG_PCI_MMCONFIG=y
CONFIG_PCIEPORTBUS=y
CONFIG_PCIEAER=y
CONFIG_PCIE_ECRC=y
CONFIG_PCIE_AER_INJECT=y
CONFIG_PCI_DOMAINS=y
CONFIG_PCI_MSI=y

I also have the kernel boot parameter aerdriver.forceload=y

With this config, I see that the relevant devices appear to support Advanced Error Reporting (visible in lspci -v output).

However, when I attempt to inject errors to them the injections fail for various reasons. None of them (out of about a dozen candidate devices) succeeds. The aer_inject facility is currently the only means I have of testing this functionality. Note that on the same hardware with a RHEL-6 based (2.6.32) kernel, the injections work correctly. I need to make them also work with 5.5.

Before I go and track down those various errors, I want to verify that I have the right kernel and that it's configured correctly. The required code seems to be present in the kernel. I suspect there's something going wrong during driver initialization and device probing.

Once again, I need answers to these two specific questions:

1) Is PCIe Advanced Error Reporting (including error injections) known to work in 5.5?
2) What configuration options do I need to select in order to enable it?

Thanks,

Neil Baylis