Re: [PATCH v1 4/6] driver: Google Bootlog

[Mike Waychison <mikew@google.com>]

On 01/24/11 16:49, Alan Cox wrote:
>> +/*
>> + * Search for the BLOG pointer.
>> + */
>> +static struct bootlog_ptr * __init bootlog_find_pointer(void)
>> +{
>> +	unsigned long address, length, cur;
>> +	struct bootlog_ptr *bp;
>> +
>> +	/* EBDA pointer contains segment the extended BIOS data area */
>> +	address = *(uint16_t *)phys_to_virt(BLOG_EBDA_POINTER);
>> +	address<<= 4;	/* convert segment to physical address */
>> +
>> +	/* EBDA length is byte 0 of the EBDA (stored in kB) */
>> +	length = *(uint8_t *)phys_to_virt(address);
>> +	length<<= 10;	/* convert to bytes */
>
> Ok this is wrong on two counts
>
> 1.	If the EBDA pointer is zero then no EBDA is present (older
> boxes sometimes did this up to about the Athlon era).
>
> 2.	EBDA isn't an X86 property it's a PC property. asm/bios_ebda
> provides one helper and perhaps it all belongs in there (along with 'is
> it a PC' sanity check)

It's not obvious to me how one would make that sort of check (is a PC) 
at runtime. Suggestions?

>
> 3.	And its probably a good idea given the standard of BIOS code to
> also check the EBDA end if given is actually within the 640K limit ...
>
>
>
> Alan