Attention is currently required from: Eran Mitrani, Jakub Czapiga, Julius Werner, Kapil Porwal, Paul Menzel, Rishika Raj, Subrata Banik, Tarun.
Nico Huber has posted comments on this change by Rishika Raj. ( https://review.coreboot.org/c/coreboot/+/83540?usp=email )
Change subject: soc/intel/meteorlake: Increase CAR STACK_SIZE by 31KB to meet coreboot requirements ......................................................................
Patch Set 5:
(1 comment)
Commit Message:
https://review.coreboot.org/c/coreboot/+/83540/comment/c4ed6f96_5d0ef547?usp... : PS5, Line 7: soc/intel/meteorlake: Increase CAR STACK_SIZE by 31KB to meet coreboot requirements
Why the 32KiB? this makes it sound like coreboot would need that much. Do we ever expect it to be more than 2KiB? If not, I'd guess that 4KiB would provide enough margin (unless something is wrong with the 512).
The CAR stack defined in coreboot is also used by coreboot to fill up the FSP-M UPDs. It is evident that the existing romstage stack size (1KB) is too small to fulfill its intended purpose, as the 1KB reserved for coreboot's portion of the stack cannot even accommodate the stack-allocated FSPM_UPD structure itself.
I see. Didn't expect this on the stack. Is that reasonable? Shouldn't that be linked into .bss or something?
It's about 4KiB AFAICS, is that correct?
Based on our debugging, we have observed instances where the vboot structure is also linked into CAR. As a result, we believe that 32KB is a reasonable size for us, given all of the factors involved.
I don't understand this, how would a vboot structure linked into CAR affect the stack? What structure are you referring to specifically?