On pe, 2015-09-11 at 15:49 -0500, Aaron Durbin wrote: > On Fri, Sep 11, 2015 at 3:36 PM, Paul Menzel > <[email protected]> wrote: > > Dear coreboot folks, > > > > > > Am Donnerstag, den 10.09.2015, 16:25 -0700 schrieb Julius Werner: > >> I'd bet it's just a single large allocation somewhere. You can try adding > >> > >> CFLAGS_ramstage += -Wstack-usage=1024 > >> > >> somewhere in coreboot/Makefile.inc and then clean+rebuild your code while > >> passing '-k' to make. You'll get a bunch of compiler warnings, and one of > >> them is likely to be the culprit. > > > > It looks like that Julius was dead on! Building the similar ASRock > > E350M1 with the modification proposed by Julius, building stops with > > the error below. > > > > ``` > > $ make > > […] > > CC northbridge/amd/agesa/oem_s3.ramstage.o > > src/northbridge/amd/agesa/oem_s3.c: In function 'OemS3Save': > > src/northbridge/amd/agesa/oem_s3.c:118:14: error: stack usage might be 4144 > > bytes [-Werror=stack-usage=] > > AGESA_STATUS OemS3Save(void *vS3SaveParams) > > ^ > > ``` > > Fun. 4KiB buffer on the stack for the win. > > $ git grep S3_DATA_MTRR_SIZE -- src/northbridge/amd/agesa/ | grep define > src/northbridge/amd/agesa/oem_s3.c:#define S3_DATA_MTRR_SIZE > 0x1000 >
Yeah, my bad. I overlooked stack usage while wondering if those MTRRs really need to be backed up in SPI in the first place. http://review.coreboot.org/#/c/11633/ Thanks, Kyösti -- coreboot mailing list: [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
