Bug 699857

Logs attached for a blade (HS22) and QPI scaled x3850/x3950 X5. Note that this is being 
logged for every slot in the system and on the scaled system...every slot in both chassis. No 
noticed detrimental effects. We would like to know if this is something new with 6.1. Tester 
has verified on the HS22 that these messages are not seen in 6.0. This has been noticed since 
RHEL 6.1 snapshot 3 through snapshot 5.

Example:
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:01.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:03.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:05.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:07.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:09.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:1c.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:00:1c.4: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:80:00.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:80:01.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:80:03.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:80:07.0: PCIe errors handled by OS.
Apr  7 11:04:45 localhost kernel: pcieport 0000:80:09.0: PCIe errors handled by OS.

The messages are logged from this section of code in pcie_port_acpi_setup() from 
drivers/pci/pcie/portdrv_acpi.c


        if (pci_aer_available()) {
                if (aer_acpi_firmware_first()) {
                        dev_info(&port->dev, "PCIe errors handled by BIOS.\n");
                 } else {
                        flags |= OSC_PCI_EXPRESS_AER_CONTROL;
                        dev_info(&port->dev, "PCIe errors handled by OS.\n");
                }
        }

The above varies slightly from the mainline source in that mainline uses dev_dbg() as opposed 
to dev_info() and it does not have the

dev_info(&port->dev, "PCIe errors handled by OS.\n");

which is what we are seeing as each pcie port is set up. Based on the changelog the ACPI 
related support may have been introduced around 2.6.32-108.el6 but I don't have access to 
the RH bugzilla bug to know for sure.

> I don't think that in general we want "informational" messages such as these
> logged. They imply hardware errors the way they are worded. Really, how are you
> ever going to convince a customer they need or require a vendor firmware or
> driver update for an adapter when the OS is blatantly logging a message telling
> them every single slot in your system has had an error?

I guess it's a difference in perception. When I read the message, I simply read it as PCIe errors 
will be handled by the OS (as opposed to BIOS or FW) but I agree this message should be 
improved for readability as it could be misinterpreted and possibly reverted to a debug 
message (like in mainline) as it doesn't seem that critical to inform the user about this detail.
      
1. Server architecture(s) (please list all effected) (x86/POWER6/Z/etc.): x86
2. Server type (9117-MMA/HS20/s390/etc.): x3850/3950X5
3. Other components involved (ixgbe/java/emulex/etc.): kernel
4. Does the server have the latest GA firmware? Yes
5. Has the problem been shown to occur on more than one system? Yes
6. What is the latest official distro build on which this bug has been seen? RHEL 6.1 SS5