Re: [RFC] use realmode code to reserve end-of-conventional-memory to 1MB

[Ian Campbell <ijc@hellion.org.uk>]

On Thu, 2008-02-28 at 14:28 +0100, Alexander van Heukelum wrote:
> Instead of using early reservations inside the kernel code,
> we could use the realmode code to modify the e820 memmap.
> This patch shows what that would look like. I have not looked
> at the case where the BIOS does not provide an e820 memmap
> yet. Probably a full solution would need to create a fake
> e820 memmap in that case.

An e820 is already faked up in machine_specific_memory_setup() if one
doesn't already exist.

> Comments?

This won't work for Xen since the real-mode code never runs there. I
think it could be fixed in xen_memory_setup() though if native goes down
this route.

Ian.

> 
> Signed-off-by: Alexander van Heukelum <heukelum@fastmail.fm>
> 
> diff --git a/arch/x86/boot/memory.c b/arch/x86/boot/memory.c
> index e77d89f..522920a 100644
> --- a/arch/x86/boot/memory.c
> +++ b/arch/x86/boot/memory.c
> @@ -23,6 +23,7 @@ static int detect_memory_e820(void)
>  	int count = 0;
>  	u32 next = 0;
>  	u32 size, id;
> +	u32 lowmem, ebda_addr;
>  	u8 err;
>  	struct e820entry *desc = boot_params.e820_map;
>  
> @@ -50,13 +51,51 @@ static int detect_memory_e820(void)
>  		   just return failure. */
>  		if (id != SMAP) {
>  			count = 0;
> -			break;
> +			goto out;
>  		}
>  
>  		count++;
>  		desc++;
> -	} while (next && count < E820MAX);
> -
> +	} while (next && count < (E820MAX - 1));
> +
> +	/* Some BIOSes do not reserve the EBDA/XBDA area correctly in.
> +	   The e820-map. Find out where the EBDA resides by looking at
> +	   the BIOS data area and reserve the EBDA and the following
> +	   legacy adapter area explicitly. */
> +#define BIOS_EBDA_SEGMENT 0x40E
> +#define BIOS_LOWMEM_KILOBYTES 0x413
> +
> +	/* end of low (conventional) memory */
> +	set_fs(0);
> +	lowmem = rdfs16(BIOS_LOWMEM_KILOBYTES);
> +	lowmem <<= 10;
> +
> +	/* start of EBDA area */
> +	ebda_addr = rdfs16(BIOS_EBDA_SEGMENT);
> +	ebda_addr <<= 4;
> +
> +	/* Fixup: bios puts an EBDA in the top 64K segment */
> +	/* of conventional memory, but does not adjust lowmem. */
> +	if ((lowmem - ebda_addr) <= 0x10000)
> +		lowmem = ebda_addr;
> +
> +	/* Fixup: bios does not report an EBDA at all. */
> +	/* Some old Dells seem to need 4k anyhow (bugzilla 2990) */
> +	if ((ebda_addr == 0) && (lowmem >= 0x9f000))
> +		lowmem = 0x9f000;
> +
> +	/* Paranoia: should never happen, but... */
> +	if (lowmem >= 0x100000)
> +		lowmem = 0xa0000;
> +
> +	/* reserve all memory between lowmem and the 1MB mark */
> +	desc->addr = lowmem;
> +	desc->size = 0x100000 - lowmem;
> +	desc->type = E820_RESERVED;
> +	count++;
> +	desc++;
> +
> +out:
>  	return boot_params.e820_entries = count;
>  }
>  
> 
> 
-- 
Ian Campbell

Nasrudin walked into a teahouse and declaimed, "The moon is more useful
than the sun."
	"Why?", he was asked.
	"Because at night we need the light more."