[coreboot] [LinuxBIOS] Intel refactoring and microcode updates

Corey Osgood corey.osgood at gmail.com
Tue Jan 22 01:18:28 CET 2008


Uwe Hermann wrote:
> On Mon, Jan 14, 2008 at 03:29:51PM +0100, Uwe Hermann wrote:
>   
>> On Mon, Jan 14, 2008 at 05:46:29AM -0500, Corey Osgood wrote:
>>     
>>> I know ;) Uwe, do you have a 440bx board you can try this on? Mine has
>>>       
>> Yep, will do.
>>     
>
> OK, here's my status report, there seem to be some issues.
>
> I tried the current svn on my ASUS P2B-F, works fine. Then I applied the
> microcode patches and tried to rebuild.
>
> First issue is that I cannot rebuild anymore, linker errors because the
> code became bigger.
>
> The default in Options.lb is 
>   default ROM_IMAGE_SIZE = 64 * 1024
> and I had to change the size (in targets/.../Config.lb) to '74 * 1024'
> in order to fit the increased size. Maybe some smaller value will
> also work, but '70 * 1024' didn't, so you need something between
> 70 KB and 74 KB. Before the patches, 64 KB was enough.
>   

Thanks for testing this out! This is to be expected, unfortunately. I 
don't like eating another 10K that could be used by FILO for some 
feature, but it seems necessary. BTW, for some reason, this works fine 
with abuild, so abuild must set some larger size.

> The image finally built fine with 74 KB, but didn't boot anymore. It
> hangs upon the 'Jumping to coreboot' message. Not sure why.
>   

This sounds like it may be related to the extra wrmsr, whereever that 
email went, since this sounds like the same issue. There weren't any 
updates for your CPU before, so it makes sense that the patch would 
bring the issue up. I put a message out to the list to try and fix it, 
but never got any response, and I can't seem to reproduce the issue.

> I noticed that the slot_1 directory now includes some HT stuff, in
>   src/cpu/intel/intel_shared/Config.lb:
> there's now
>   dir /cpu/intel/hyperthreading
> which wasn't there in the old slot_1 code (slot_2 actually), I think.
>   

I looked through the hyperthreading code, and it shouldn't affect CPUs 
that don't have hyperthreading. IIRC, there's also some sort of block in 
there that prevents it from doing anything if CONFIG_SMP isn't set.

> But that doesn't seem to be the (only) problem, commenting the
>   # dir /cpu/intel/hyperthreading
> doesn't help. I tried setting
>   NO_MICROCODE_UPDATES = 1
> in the Options.lb and it _seems_ that helped, it's booting again.
> But of course it doesn't apply the microcode updates in that case.
>
> My test CPU was
>   CPU: family 06, model 07, stepping 03
>   

NO_MICROCODE_UPDATES cuts out everything microcode-related, so that's 
expected, too.

> Uwe.
>   

Thanks,
Corey




More information about the coreboot mailing list