On 28.04.2018 17:34, Kyösti Mälkki wrote:
On Sat, Apr 28, 2018 at 3:41 PM, Nicola Corna nicola@corna.info wrote:
April 27, 2018 12:29 PM, "Nicola Corna" nicola@corna.info wrote:
With config PARALLEL_CPU_INIT=y so SMP / SMM init in initialize_cpus() will never call wait_other_cpus() at all. That actually regressed in my commit 0cc2ce4 [1] but I can't test if reverting it solves this for you. I'll push a regression fix soonish for review.
[1] https://review.coreboot.org/c/coreboot/+/21088
Kyösti
I'm going to test https://review.coreboot.org/#/c/coreboot/+/25874 soon and report back, thanks for your help.
Unfortunately that patch dont't fix the issue (log attached). I've also tried 0cc2ce4 but with that revision it works correctly (log and config attached).
If needed I can run a git bisect to find which commit broke the boot.
I thought I had seen some other message [1] on the list regarding X201(i) not booting or resuming. Looking at your boot log ending on "0:1f.0 final", it appears we are looking at the same commit which adds INTEL_CHIPSET_LOCKDOWN implementation in lpc_final(). The commit message tells it was not tested for ibexpeak so you may want to try revert that one before going for full bisect.
Yes, that's very likely a problem. It looks like the whole finalize code path of the X201 was untested all the time (even on resume). I don't remember if EHCI debug works in SMM? If it does, you could enable log- ging for the SMI handler as well (if you want to debug it).
Nico