Re: [Qemu-devel] [PATCH v3 3/3] vfio: Fix 128 bit handling

[Alex Williamson <alex.williamson@redhat.com>]

On Thu, 2013-08-29 at 11:03 +1000, Alexey Kardashevskiy wrote:
> On 08/29/2013 01:18 AM, Alex Williamson wrote:
> > On Thu, 2013-08-22 at 21:29 +1000, Alexey Kardashevskiy wrote:
> >> Upcoming VFIO on SPAPR PPC64 support will initialize the IOMMU
> >> memory region with UINT64_MAX (2^64 bytes) size so int128_get64()
> >> will assert.
> >>
> >> The patch takes care of this check. The existing type1 IOMMU code
> >> is not expected to map all 64 bits of RAM so the patch does not
> >> touch that part.
> >>
> >> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> >> ---
> >> Changes:
> >> v2:
> >> * used new function int128_exts64()
> >> ---
> >>  hw/misc/vfio.c | 11 ++++++++---
> >>  1 file changed, 8 insertions(+), 3 deletions(-)
> >>
> >> diff --git a/hw/misc/vfio.c b/hw/misc/vfio.c
> >> index dfe3a80..3878fc7 100644
> >> --- a/hw/misc/vfio.c
> >> +++ b/hw/misc/vfio.c
> >> @@ -1920,6 +1920,7 @@ static void vfio_listener_region_add(MemoryListener *listener,
> >>      VFIOContainer *container = container_of(listener, VFIOContainer,
> >>                                              iommu_data.listener);
> >>      hwaddr iova, end;
> >> +    Int128 llend;
> >>      void *vaddr;
> >>      int ret;
> >>  
> >> @@ -1940,13 +1941,17 @@ static void vfio_listener_region_add(MemoryListener *listener,
> >>      }
> >>  
> >>      iova = TARGET_PAGE_ALIGN(section->offset_within_address_space);
> >> -    end = (section->offset_within_address_space + int128_get64(section->size)) &
> >> -          TARGET_PAGE_MASK;
> >> +    llend = int128_make64(section->offset_within_address_space);
> >> +    llend = int128_add(llend, section->size);
> >> +    llend = int128_and(llend, int128_exts64(TARGET_PAGE_MASK));
> >>  
> >> -    if (iova >= end) {
> >> +    if (int128_ge(int128_make64(iova), llend)) {
> >>          return;
> >>      }
> >>  
> >> +    end = (section->offset_within_address_space + int128_get64(section->size)) &
> >> +          TARGET_PAGE_MASK;
> >> +
> > 
> > I'm confused, we build an Int128 version of end above for the
> > comparison, why isn't this just:
> > 
> > end = int128_get64(llend);
> 
> 
> section->size for IOMMU memory region I have on spapr-vfio is 2^64 so
> int128_get64() fails.

But you're leaving code that does int128_get64(section->size)... how
does that not assert?  This patch is not maintainable.  We're seemingly
calculating the same value twice with no comment as to why.  A hwaddr
type end should be calculated from the Int128 rather than paying
attention to rollover in one place but not another.  Thanks,

Alex