Attention is currently required from: Felix Held, Fred Reitberger, Jason Glenesk, Jérémy Compostella, Matt DeVillier.
Hello Felix Held, Fred Reitberger, Jason Glenesk, Jérémy Compostella, Matt DeVillier, build bot (Jenkins),
I'd like you to reexamine a change. Please visit
https://review.coreboot.org/c/coreboot/+/83495?usp=email
to look at the new patch set (#3).
Change subject: soc/amd: Ensure bank 0 is selected before accessing VBNV in CMOS ......................................................................
soc/amd: Ensure bank 0 is selected before accessing VBNV in CMOS
In AMD platforms, the bit 4 of CMOS's Register A (0x0a) is DV0 bank selection (0 for Bank 0; 1 for Bank 1) [1]. Since the MC146818 driver accesses VBNV via Bank 0, the bit must be cleared before we can save VBNV to CMOS in verstage.
Usually there's no problem with that, because the Register A is configured in cmos_init() in ramstage. However, if CMOS has lost power, then in the first boot after that, the bit may contain arbitrary data in verstage. If that bit happens to be 1, then CMOS writes in verstage will fail.
To fix the problem, define vbnv_platform_init_cmos() to configure the Register A (RTC_FREQ_SELECT). Static assertions are also added to make sure the whole VBNV is accessible via Bank 0 and that the written byte doesn't contain RTC_AMD_BANK_SELECT (bit 4). Note that the kernel driver also ensures RTC_AMD_BANK_SELECT isn't set for AMD [2].
[1] 48751_16h_bkdg.pdf [2] lore.kernel.org/lkml/20220523165815.913462426@linuxfoundation.org
BUG=b:346716300 TEST=CMOS writes succeeded in verstage after battery cutoff BRANCH=skyrim
Change-Id: Idf167387b403be1977ebc08daa1f40646dd8c83f Signed-off-by: Yu-Ping Wu yupingso@chromium.org --- M src/include/pc80/mc146818rtc.h M src/soc/amd/common/vboot/vbnv_cmos.c 2 files changed, 15 insertions(+), 0 deletions(-)
git pull ssh://review.coreboot.org:29418/coreboot refs/changes/95/83495/3