[fm-discuss] Re: Will the PCI diagnosis engine detect faulty PCI adaptors?

[Cynthia McGuire]

We have pushed beyond the PCI interface with some recent FMA features.  The bge driver has been hardened to detect chip-specific errors for diagnosis and fault reporting.  On SFX4500 systems,  we poll for errors related to SATA disk self-test failures, SMART predictive failures, and over-temps.  The presence of these conditions are reported to the fault manager (fmd(1M) for diagnosis of a failure or imminent failure of a disk.  The appropriate LED is updated to reflect the faulty disk and the fault manager issues a diagnosis message to alert the administrator.

Work will continue to harden HBA and NIC drivers.  At the same time, there are plans to keep pushing FMA farther up the network and storage stacks. To the extent possible, we will transparently embed FMA error detection in common software layers such as gld so that common error detection and diagnosis software can be used.  In addition, we are working on a sensor abstraction layer that will allow data exported by IPMI, SMART, SMBus, etc... to be collected and analyzed for potential hardware failures.
--
This message posted from opensolaris.org