Re: [PATCH 1/2] amd/passthrough: Do not leak domain mappings from do_invalidate_dte()

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 04/12/13 15:22, Jan Beulich wrote:
>>>> On 04.12.13 at 16:09, Andrew Cooper <andrew.cooper3@citrix.com> wrote:
>> --- a/xen/drivers/passthrough/amd/iommu_guest.c
>> +++ b/xen/drivers/passthrough/amd/iommu_guest.c
>> @@ -433,7 +433,10 @@ static int do_invalidate_dte(struct domain *d, cmd_entry_t *cmd)
>>  
>>      /* Do not update host dte before gcr3 has been set */
>>      if ( gcr3_gfn == 0 )
>> +    {
>> +        unmap_domain_page(dte_base);
>>          return 0;
>> +    }
>>  
>>      gcr3_mfn = mfn_x(get_gfn(d, gcr3_gfn, &p2mt));
>>      put_gfn(d, gcr3_gfn);
>> @@ -446,6 +449,7 @@ static int do_invalidate_dte(struct domain *d, cmd_entry_t *cmd)
>>      {
>>          AMD_IOMMU_DEBUG("%s: Fail to find iommu for bdf %x!\n",
>>                          __func__, mbdf);
>> +        unmap_domain_page(dte_base);
>>          return -ENODEV;
>>      }
> I think the better way to fix this would be to move
>
>     glx = get_glx_from_dte(gdte);
>     gv = get_gv_from_dte(gdte);
>
>     unmap_domain_page(dte_base);
>
> up ahead of the first exit path.
>
> Jan
>

So it would - v2 on its way.

~Andrew