RE: mtrr

[YhLu]

In the cpu/x86/mtrr/mtrr.c

#warning "FIXME I do not properly handle address more than 36 physical
address bits"
#ifdef k8
# define ADDRESS_BITS 40
#else
# define ADDRESS_BITS 36
#endif
#define ADDRESS_BITS_HIGH (ADDRESS_BITS - 32)
#define ADDRESS_MASK_HIGH ((1u << ADDRESS_BITS_HIGH) - 1)

It seems someone already remove k8 from Option.lb

Then ....

Regards

YH

-----Original Message-----
From: YhLu 
Sent: Thursday, November 11, 2004 6:00 PM
To: YhLu; [e-mail]; Ronald G. Minnich
Cc: [e-mail]
Subject: RE: mtrr

On 2.4.22 it becomes to 
~ # cat /proc/mtrr
reg00: base=0x00000000 (   0MB), size=32768MB: write-back, count=1
reg01: base=0x80000000 (2048MB), size=16384MB: write-back, count=1
reg02: base=0xc0000000 (3072MB), size=4096MB: write-back, count=1

x 16

-----Original Message-----
From: YhLu 
Sent: Thursday, November 11, 2004 5:24 PM
To: [e-mail]; Ronald G. Minnich
Cc: [e-mail]
Subject: mtrr

I just found the mtrr after booting is not right.

~ # cat /proc/mtrr
reg00: base=0x00000000 (   0MB), size=985088MB: write-back, count=1
reg01: base=0x80000000 (2048MB), size=984064MB: write-back, count=1
reg02: base=0xc0000000 (3072MB), size=983296MB: write-back, count=1

and it should be

reg00: base=0x00000000 (   0MB), size=2048MB: write-back, count=1
reg01: base=0x80000000 (2048MB), size=1024MB: write-back, count=1
reg02: base=0xc0000000 (3072MB), size=256MB: write-back, count=1

and it is has 98340MB more = 0xf0 0000 0000.

Regards

YH
_______________________________________________
Linuxbios mailing list
[e-mail]
http://www.clustermatic.org/mailman/listinfo/linuxbios

mtrr

YhLu  writes:
> In the cpu/x86/mtrr/mtrr.c
>
> #warning "FIXME I do not properly handle address more than 36 physical
> address bits"
> #ifdef k8
> # define ADDRESS_BITS 40
> #else
> # define ADDRESS_BITS 36
> #endif
> #define ADDRESS_BITS_HIGH (ADDRESS_BITS - 32)
> #define ADDRESS_MASK_HIGH ((1u << ADDRESS_BITS_HIGH) - 1)
>
> It seems someone already remove k8 from Option.lb
>
> Then ....
>
> Regards

Interesting.  The mtrr's actually work and the kernel is confused.

However what we need to do is call the appropriate cpuid function
to compute it or pass in the number of address bits.  It should not be
a compile time option, at least not in the generic code.  What
happens when we get can plug in cpus with different address bit
limits.

Eric