On my Supermicro P4SCT (Pentium 4, 3.2 GHz, 1GB RAM, Marvell MV88SX5041 64bit PCI SATA RAID, 7x HDD 400GB WD4000KD-00N) I'm able to reproducibly hang the machine by issuing 'smartctl -a' command. I was trying to eliminate HW problem as sata_mv has poor EH (error handling). But if at least 3 drives with RAID5 or RAID6 (but not RAID1) are rebuilding an array, smartctl causes random hangs (by repeating smartctl by 1 to 30 times). Is there a known problem with deadlock or so? The hang comes especially by redirecting output to the pipe or /dev/null (ie without terminal that slowing down the command or reducing pauses between subsequent pooling of the drives). Also smartd causes random hangs (in few days) if the machine is under disc I/O load.
I don't have a marvell controller to verify this test case, but, sata_mv was updated in RHEL5 U3 and stress tested so I am closing this issue, please re-open if necessary.