1 comment:
File src/drivers/mrc_cache/mrc_cache.c:
Patch Set #2, Line 522: if (!CONFIG(VBOOT_STARTS_IN_BOOTBLOCK))
These checks seem out of place and seem to be providing new semantics to the modules themselves. i. […]
Hmm... yes, maybe. Looks like some platforms used to check for recovery mode and changed their raminit behavior (choosing to always retrain if I understand that correctly) before the actual vboot code even runs. This is now broken and I think it already broke in CB:38780, I'm just trying to make the current behavior more obvious here.
We really want the outside accesses to VBNV gone, which means there is no way to determine whether recovery mode is enabled before running vboot. I really don't want to go back to hacky NVRAM parsing code in coreboot just to support this weird path on some ancient boards. Considering that Chrome OS is (presumably) never going back to STARTS_IN_ROMSTAGE and I don't think anybody else cares much about these details of recovery mode behavior, do you think changing policy like this is acceptable? (I think in practice it just means that these boards will use their MRC cache if available even in recovery mode.)
To view, visit change 39221. To unsubscribe, or for help writing mail filters, visit settings.