On Fri, Oct 9, 2009 at 7:16 PM, Carl-Daniel Hailfinger <[email protected]> wrote: > On 09.10.2009 08:10, Bao, Zheng wrote: >> Current coreboot seems to hang somewhere. Before the 60th anniversary of >> People's Republic of China, I always disable the CBFS when I worked on >> my fam10 board, otherwise it would be error. But now, I can not find >> where I can disable it. It seems to stop at waiting AP cores. >> >> I am wondering if it is caused by CBFS and if it happens on other board. >> If any one can test it, I will be appreciating. >> > > We have not solved the SMP startup printk locking yet. That explains the > mangled log messages. > Unless I'm mistaken we still decompress some parts multiple times > concurrently to the same address and that can crash the code. > Overlapping stack might also crash the code during lzma decompression. > Decompression of AP code should happen on the BSP or on exactly one AP, > but not on all APs. > > >> coreboot-2.0.0-r623M_tilapia_fam10_Fallback Fri Oct 9 13:15:49 CST 2009 >> starting... >> [...] >> Start other core - nodeid: 00 cores: 03 >> started ap apicid: cccooorrreeexxx::: --------- {{{ >> AAAPPPIIICCC >> IIIDDD === 000312 NNNOOODDDEEEIIIDDD === 000000 >> CCCOOORRREEEIIIDDD === 000321}} >> } --------- >>
The stacks shouldn't be overlapping but the decompression location might (I havn't looked at the implementation yet). We may need to look at CBFS more closely wrt AP startup. Marc -- http://marcjonesconsulting.com -- coreboot mailing list: [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
