Patrick Rudolph has posted comments on this change by Patrick Rudolph. ( https://review.coreboot.org/c/coreboot/+/84833?usp=email )
Change subject: arch/x86: Add breakpoint to stack canary ......................................................................
Patch Set 20:
(1 comment)
Patchset:
PS20:
Without further details I cannot fix this. […]
The first problem ``` [NOTE ] coreboot-coreboot-unknown.9999.06d5019 Fri Dec 13 06:11:04 UTC 2024 x86_64 verstage starting (log level: 8)... [ERROR] Failed to create address zero instruction fetch breakpoint [ERROR] Failed to create stack canary breakpoint ``` is due to `CONFIG_VBOOT_SEPARATE_VERSTAGE=n`. The verstage code is compiled into romstage, but it doesn't clear BSS when jumping to verstage. Thus there are not enough free breakpoint slots when it tries to init breakpoints for instruction and NULL deref.
Previously it worked as there were enough free slots by coincidence.
The other problem is more interesting: ``` [NOTE ] coreboot-coreboot-unknown.9999.06d5019 Fri Dec 13 06:11:04 UTC 2024 x86_64 romstage starting (log level: 8)... [ERROR] Stack corruption detected at rip: 0xf983007a [ERROR] Stack corruption detected at rip: 0xf983007a ``` Is this with https://review.coreboot.org/c/coreboot/+/85568 applied? I build google/fatcat and the shown address 0xf983007a seems to outside of the BIOS region and outside of CAR. Are you sure that the generic breakpoint code works on x86_64? It seems to work fine on x86_32.
Do the other platforms show the same address (`0xf983007a`) ?
-
Patrick Rudolph (Code Review)