Re: [patch 1/1] combine two adjacent mtrr region with the same type

[bmao]

Yinghai Lu wrote:
> bibo mao wrote:
> > This patch is to combine two adjacent mtrr region with the same type.
> >
> >  kernel/cpu/mtrr/main.c          |   17 +++++++++++++++--
> >  3 files changed, 15 insertions(+), 2 deletions(-)
> >
> > Signed-off-by: bibo mao <bibo.mao@xxxxxxxxx>
> >
> > -------------------------------------------------------------------------------
> > diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c
> > index 6e212f3..c3adf83 100644
> > --- a/arch/x86/kernel/cpu/mtrr/main.c
> > +++ b/arch/x86/kernel/cpu/mtrr/main.c
> > @@ -305,7 +305,7 @@ int mtrr_add_page(unsigned long base, unsigned long size,
> >  {
> >       int i, j, replace, error;
> >       mtrr_type ltype;
> > -     unsigned long lbase, lsize;
> > +     unsigned long lbase, lsize, newbase;
> >
> >       if (!mtrr_if)
> >               return -ENXIO;
> > @@ -344,7 +344,7 @@ int mtrr_add_page(unsigned long base, unsigned long size,
> >       mutex_lock(&mtrr_mutex);
> >       for (i = 0; i < num_var_ranges; ++i) {
> >               mtrr_if->get(i, &lbase, &lsize, &ltype);
> > -             if (!lsize || base > lbase + lsize - 1 || base + size - 1 < lbase)
> > +             if (!lsize || base > lbase + lsize || base + size < lbase)
> >                       continue;
> >               /*  At this point we know there is some kind of overlap/enclosure  */
> >               if (base < lbase || base + size - 1 > lbase + lsize - 1) {
> > @@ -356,6 +356,19 @@ int mtrr_add_page(unsigned long base, unsigned long size,
> >                               }
> >                               else if (types_compatible(type, ltype))
> >                                       continue;
> > +                     } else if ((base == lbase + lsize) || (lbase ==base + size)) {
> > +                             if (type != ltype)
> > +                                     continue;
> this path seems not right...
>
> do you need
>
>                 if (ltype != type) {
>                         if (types_compatible(type, ltype))
>                                 continue;
>                         printk (KERN_WARNING "mtrr: type mismatch for %lx000,%lx000 old: %s new: %s\n",
>                              base, size, mtrr_attrib_to_str(ltype),
>                              mtrr_attrib_to_str(type));
>                         goto out;
>                 }
> ?
This warning info is for two enclosed mtrr region, but it is adjacent here. Here is no overlapped mtrr region even.
> > +                             /* New region adjacent with an existing region  */
> > +                             newbase = (base < lbase) ? base : lbase;
> > +                             if (!mtrr_if->validate_add_page(newbase,lsize + size,type) &&
> > +                                mtrr_if->get_free_region(newbase, lsize + size, replace)>=0) {
> > +                                     base = newbase;
> > +                                     size += lsize;
> > +                                     replace = i;
> > +                                     continue;
> > +                             } else
> > +                                     continue;
> >                       }
> >                       printk(KERN_WARNING
> >                              "mtrr: 0x%lx000,0x%lx000 overlaps existing"
>
> better to have some test case...
>
> YH
Here is test result: 

# echo "base=0x02000000 size=0x00800000 type=write-back" >| /proc/mtrr
# echo "base=0x03000000 size=0x01000000 type=write-back" >| /proc/mtrr
# cat /proc/mtrr 
reg00: base=0x02000000 (  32MB), size=   8MB: write-back, count=1
reg01: base=0x03000000 (  48MB), size=  16MB: write-back, count=1

# echo "base=0x02800000 size=0x00800000 type=write-back" >| /proc/mtrr
# cat /proc/mtrr 
reg01: base=0x02000000 (  32MB), size=  32MB: write-back, count=3

Without this patch, it will display with this result:

reg00: base=0x02000000 (  32MB), size=   8MB: write-back, count=1
reg01: base=0x03000000 (  48MB), size=  16MB: write-back, count=1
reg02: base=0x02800000 (  40MB), size=   8MB: write-back, count=1

thanks
bibo,mao
--
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/