Re: [sodaville] [PATCH] x86/trampoline: fallback to fixed trampolineaddress if no e820 map is provided

[H. Peter Anvin]

On 12/23/2010 05:16 AM, Yinghai wrote:
> 
> On Dec 23, 2010, at 1:57 AM, Florian Fainelli <ffainelli@xxxxxxxxxx> wrote:
> 
>> On Tuesday 14 December 2010 22:19:39 Florian Fainelli wrote:
>>> From: Florian Fainelli <ffainelli@xxxxxxxxxx>
>>>
>>> Since 893f38d1 (x86: Use find_e820() instead of hard coded trampoline
>>> address), the trampoline base address should be provided by the BIOS as an
>>> e820 area. Some minimalistic BIOSes (like CEFDK on CE4100) may not provide
>>> any e820 area, still we should fallback to a fixed trampoline base address
>>> for SMP to work correctly.
> 
> No e820 memmap ?
> 
> Not sure how does it work
> 
> Do you have boot log with debug ?
> 
> Also recent kernel is using memblock already for finding free area.
> 

OK, this deeply troubles me.  This is clearly not the right approach
here... the memory map should be fed to the kernel, this has been a
requirement all along, and Linus has very deliberately not agreed to
exceptions (and I agree with that stance.)

So this patch is a non-starter... I presume the memory map is being
provided in through the device (which means it should be fed into
memblock in current kernels, although the ordering may need to be tweaked?)

	-hpa
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/