[PATCH] x86: adjust generic_get_mtrr() for 64-bit

[PATCH] x86: adjust generic_get_mtrr() for 64-bit

[Jan Beulich]

Needing to deal with potentially large memory configurations, the
variables here should be "unsigned long" instead of "unsigned int".

Signed-off-by: Jan Beulich <jbeulich@suse.com>

---
 arch/x86/kernel/cpu/mtrr/generic.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

--- 3.5-rc5/arch/x86/kernel/cpu/mtrr/generic.c
+++ 3.5-rc5-x86-mtrr-generic-types/arch/x86/kernel/cpu/mtrr/generic.c
@@ -514,8 +514,8 @@ generic_get_free_region(unsigned long ba
 static void generic_get_mtrr(unsigned int reg, unsigned long *base,
 			     unsigned long *size, mtrr_type *type)
 {
-	unsigned int mask_lo, mask_hi, base_lo, base_hi;
-	unsigned int tmp, hi;
+	unsigned long mask_lo, mask_hi, base_lo, base_hi, tmp;
+	unsigned int hi;
 
 	/*
 	 * get_mtrr doesn't need to update mtrr_state, also it could be called
@@ -540,9 +540,9 @@ static void generic_get_mtrr(unsigned in
 	mask_lo = size_or_mask | tmp;
 
 	/* Expand tmp with high bits to all 1s: */
-	hi = fls(tmp);
+	hi = fls_long(tmp);
 	if (hi > 0) {
-		tmp |= ~((1<<(hi - 1)) - 1);
+		tmp |= ~((1UL<<(hi - 1)) - 1);
 
 		if (tmp != mask_lo) {
 			printk(KERN_WARNING "mtrr: your BIOS has configured an incorrect mask, fixing it.\n");

Re: [PATCH] x86: adjust generic_get_mtrr() for 64-bit

[Yinghai Lu]

On Fri, Jul 6, 2012 at 7:14 AM, Jan Beulich <JBeulich@suse.com> wrote:
> Needing to deal with potentially large memory configurations, the
> variables here should be "unsigned long" instead of "unsigned int".
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>
> ---
>  arch/x86/kernel/cpu/mtrr/generic.c |    8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
>
> --- 3.5-rc5/arch/x86/kernel/cpu/mtrr/generic.c
> +++ 3.5-rc5-x86-mtrr-generic-types/arch/x86/kernel/cpu/mtrr/generic.c
> @@ -514,8 +514,8 @@ generic_get_free_region(unsigned long ba
>  static void generic_get_mtrr(unsigned int reg, unsigned long *base,
>                              unsigned long *size, mtrr_type *type)
>  {
> -       unsigned int mask_lo, mask_hi, base_lo, base_hi;
> -       unsigned int tmp, hi;
> +       unsigned long mask_lo, mask_hi, base_lo, base_hi, tmp;
> +       unsigned int hi;

_lo, _hi means we want it as 32bit.

or we just change them to u32 to make it more clear ?

Thanks

Yinghai

Re: [PATCH] x86: adjust generic_get_mtrr() for 64-bit

[Jan Beulich]

>>> On 06.07.12 at 23:38, Yinghai Lu <yinghai@kernel.org> wrote:
> On Fri, Jul 6, 2012 at 7:14 AM, Jan Beulich <JBeulich@suse.com> wrote:
>> Needing to deal with potentially large memory configurations, the
>> variables here should be "unsigned long" instead of "unsigned int".
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>>
>> ---
>>  arch/x86/kernel/cpu/mtrr/generic.c |    8 ++++----
>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> --- 3.5-rc5/arch/x86/kernel/cpu/mtrr/generic.c
>> +++ 3.5-rc5-x86-mtrr-generic-types/arch/x86/kernel/cpu/mtrr/generic.c
>> @@ -514,8 +514,8 @@ generic_get_free_region(unsigned long ba
>>  static void generic_get_mtrr(unsigned int reg, unsigned long *base,
>>                              unsigned long *size, mtrr_type *type)
>>  {
>> -       unsigned int mask_lo, mask_hi, base_lo, base_hi;
>> -       unsigned int tmp, hi;
>> +       unsigned long mask_lo, mask_hi, base_lo, base_hi, tmp;
>> +       unsigned int hi;
> 
> _lo, _hi means we want it as 32bit.
> 
> or we just change them to u32 to make it more clear ?

Agreed, it's really only "tmp" that needs to be "unsigned long".
I'll send out a v2 soon, at once converting lo/hi to u32.

Jan

Re: [PATCH] x86: adjust generic_get_mtrr() for 64-bit

[Jan Beulich]

>>> On 06.07.12 at 23:38, Yinghai Lu <yinghai@kernel.org> wrote:
> On Fri, Jul 6, 2012 at 7:14 AM, Jan Beulich <JBeulich@suse.com> wrote:
>> Needing to deal with potentially large memory configurations, the
>> variables here should be "unsigned long" instead of "unsigned int".
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>>
>> ---
>>  arch/x86/kernel/cpu/mtrr/generic.c |    8 ++++----
>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> --- 3.5-rc5/arch/x86/kernel/cpu/mtrr/generic.c
>> +++ 3.5-rc5-x86-mtrr-generic-types/arch/x86/kernel/cpu/mtrr/generic.c
>> @@ -514,8 +514,8 @@ generic_get_free_region(unsigned long ba
>>  static void generic_get_mtrr(unsigned int reg, unsigned long *base,
>>                              unsigned long *size, mtrr_type *type)
>>  {
>> -       unsigned int mask_lo, mask_hi, base_lo, base_hi;
>> -       unsigned int tmp, hi;
>> +       unsigned long mask_lo, mask_hi, base_lo, base_hi, tmp;
>> +       unsigned int hi;
> 
> _lo, _hi means we want it as 32bit.
> 
> or we just change them to u32 to make it more clear ?

Actually, as I was about to do the adjustment, this isn't correct:
Both mask_hi and base_hi get shifted left in some calculations
there, and hence they would require up-casts to unsigned long
if their type would remain a 32-bit one. Further, mask_lo gets
or-ed with size_or_mask, which is u64, the result compared
with tmp (now unsigned long), and the same result also negated
so the *size (also unsigned long), so needs to be unsigned long
itself unless we want to introduce another variable.

That leaves only base_lo as a candidate for remaining 32-bit,
but for consistency I think it would be better to have them all
have the same type.

Consequently I think the patch should remain as is.

Jan