Mark,
If you can see the scratch 3.7 "jcm1" kernel feel free to test (just has cache diag reg. fix but "jcm2" will turn off more errata we don't need in 3.7 timeframe for unified highbank/versatile only multiplatform base kernel - your patch was helpful. Btw it was created backwards but that was easy to handle). Don't forget to check your initramfs load address is high enough to avoid overlap if you test this. The kernel hangs on boot around time of SATA init. I will root cause but also learning to involve others :) So feel free to poke sooner if you have cycles - otherwise I can keep it going. I know FUDCon prep is critical this week. I have a few cycles on Monday so will intentionally not be on IRC in order to close down a few of these sans pings. Please use mail.
Jon.