Re: [Qemu-devel] new memory api: offsets?

[Avi Kivity <avi@redhat.com>]

On 09/01/2011 01:00 PM, Michael Walle wrote:
> Am Donnerstag 01 September 2011, 07:54:28 schrieb Avi Kivity:
> >  On 09/01/2011 01:54 AM, Michael Walle wrote:
> >  >  Hi Avi,
> >  >
> >  >  while debugging, i noticed, that mr->offset is never set, expect in the
> >  >  initializer. (The subregion collision warning is printed although the
> >  >  regions are not overlapping.) Is this intended?
> >
> >  Did you miss memory_region_set_offset()?
> I saw that function but it's never called. Now i noticed that this is a
> deprecated public function.
>
> >  The subregion collision warning is unrelated?
> you are walking along the subregions and use that offset property to detect
> collisions. Shouldn't you use the addr property instead?

Ah, of course.

>
> @@ -1190,16 +1190,18 @@ static void
> memory_region_add_subregion_common(MemoryRegion *mr,
>           if (subregion->may_overlap || other->may_overlap) {
>               continue;
>           }
> -        if (offset>= other->offset + other->size
> -            || offset + subregion->size<= other->offset) {
> +        if (offset>= other->addr + other->size
> +            || offset + subregion->size<= other->addr) {
>               continue;
>           }

Right.  Please post with a changelog and signoff.

>
> >  >  btw. you may include the memory region name in the warning.
> >
> >  sure, patches welcome.
>
> -        printf("warning: subregion collision %llx/%llx vs %llx/%llx\n",
> +        printf("warning: subregion collision %llx/%llx (%s) "
> +               "vs %llx/%llx (%s)\n",
>                  (unsigned long long)offset,
>                  (unsigned long long)subregion->size,
> -               (unsigned long long)other->offset,
> -               (unsigned long long)other->size);
> +               subregion->name,
> +               (unsigned long long)other->addr,
> +               (unsigned long long)other->size,
> +               other->name);
>
> let me know if i should post a git patch instead :)
>

Yes, with a signoff.

-- 
error compiling committee.c: too many arguments to function