Bug #151

KCMA-D8 (and presumably other amdfam10/amdmct-based boards) fails to boot

Added by Patrick McDermott almost 4 years ago.

Status:NewStart date:12/31/2017
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:chipset configuration
Target version:-

Description

Current master fails to boot on a KCMA-D8 with one Opteron 4238 and one Micron MT36KSF2G72PZ-1G4E1 (16-GiB PC3-10600 1.35-V ECC RDIMM). coreboot just infinitely loops through MCT DIMM training, sometimes resetting.

I've bisected the failure to commit 4.5-1097-g610d1c6 (Revert "nb/amd/mct_ddr3: Fix RDIMM training failure on Fam15h"), hence my suspicion that the KGPE-D16 and possibly other boards with a similar CPU/RAM configuration may also fail to boot. Attached are logs of that commit and current master (both failing) as well as master with that commit reverted (which booted all the way to the payload on the second try).

This seems different from bug #129, given that the regressing commit predates the commit range listed there and the symptoms differ (#129 suggests bootblock breakage, which I haven't seen in my tests and bisecting).

coreboot-4.5-1097-g610d1c6.cap - Log of commit 4.5-1097-g610d1c6 (120 KB) Patrick McDermott, 12/31/2017 05:07 AM

coreboot-4.6-2479-g8e4384d.cap - Log of commit 4.6-2479-g8e4384d (current master) (166 KB) Patrick McDermott, 12/31/2017 05:22 AM

coreboot-4.6-2480-g6cfa8e2.cap - Log of commit 4.6-2479-g8e4384d (current master) with 4.5-1097-g610d1c6 reverted (137 KB) Patrick McDermott, 12/31/2017 05:39 AM

Also available in: Atom PDF