Re: [RFC XEN PATCH v4 01/41] x86_64/mm: fix the PDX group check in mem_hotadd_check()

["Jan Beulich" <JBeulich@suse.com>]

>>> On 07.12.17 at 11:09, <haozhong.zhang@intel.com> wrote:
> --- a/xen/arch/x86/x86_64/mm.c
> +++ b/xen/arch/x86/x86_64/mm.c
> @@ -1295,12 +1295,8 @@ static int mem_hotadd_check(unsigned long spfn, unsigned long epfn)
>          return 0;
>  
>      /* Make sure the new range is not present now */
> -    sidx = ((pfn_to_pdx(spfn) + PDX_GROUP_COUNT - 1)  & ~(PDX_GROUP_COUNT - 1))
> -            / PDX_GROUP_COUNT;
> +    sidx = (pfn_to_pdx(spfn) & ~(PDX_GROUP_COUNT - 1)) / PDX_GROUP_COUNT;

I agree that rounding up here is bogus.

>      eidx = (pfn_to_pdx(epfn - 1) & ~(PDX_GROUP_COUNT - 1)) / PDX_GROUP_COUNT;
> -    if (sidx >= eidx)
> -        return 0;
> -
>      s = find_next_zero_bit(pdx_group_valid, eidx, sidx);

But isn't this one wrong too, needing eidx + 1 as argument instead? Also
for the following find_next_bit() then?

Also please don't drop the blank line there.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel