Re: [Xen-devel] [PATCH 09/12] xen: only map PV guest grants via iommu if HAS_PASSTHROUGH

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 20.01.12 at 13:40, Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote:
> On Fri, 2012-01-20 at 12:38 +0000, Jan Beulich wrote:
>> >>> On 20.01.12 at 13:06, Ian Campbell <ian.campbell@xxxxxxxxxx> wrote:
>> > Signed-off-by: Ian Campbell <ian.campbell@xxxxxxxxxx>
>> > ---
>> >  xen/Rules.mk             |    1 +
>> >  xen/arch/arm/dummy.S     |    2 --
>> >  xen/common/grant_table.c |    6 ++++++
>> >  3 files changed, 7 insertions(+), 2 deletions(-)
>> > 
>> > diff --git a/xen/Rules.mk b/xen/Rules.mk
>> > index e25e8d4..ce88316 100644
>> > --- a/xen/Rules.mk
>> > +++ b/xen/Rules.mk
>> > @@ -52,6 +52,7 @@ CFLAGS-$(perfc_arrays)  += -DPERF_ARRAYS
>> >  CFLAGS-$(lock_profile)  += -DLOCK_PROFILE
>> >  CFLAGS-$(HAS_ACPI)      += -DHAS_ACPI
>> >  CFLAGS-$(HAS_TMEM)      += -DHAS_TMEM
>> > +CFLAGS-$(HAS_PASSTHROUGH) += -DHAS_PASSTHROUGH
>> >  CFLAGS-$(frame_pointer) += -fno-omit-frame-pointer -DCONFIG_FRAME_POINTER
>> >  
>> >  ifneq ($(max_phys_cpus),)
>> > diff --git a/xen/arch/arm/dummy.S b/xen/arch/arm/dummy.S
>> > index 5010619..e858613 100644
>> > --- a/xen/arch/arm/dummy.S
>> > +++ b/xen/arch/arm/dummy.S
>> > @@ -27,8 +27,6 @@ DUMMY(gmfn_to_mfn);
>> >  DUMMY(gnttab_clear_flag);
>> >  DUMMY(gnttab_mark_dirty);
>> >  DUMMY(hypercall_create_continuation);
>> > -DUMMY(iommu_map_page);
>> > -DUMMY(iommu_unmap_page);
>> >  DUMMY(is_iomem_page);
>> >  DUMMY(max_page);
>> >  DUMMY(node_online_map);
>> > diff --git a/xen/common/grant_table.c b/xen/common/grant_table.c
>> > index b024016..1798fcd 100644
>> > --- a/xen/common/grant_table.c
>> > +++ b/xen/common/grant_table.c
>> > @@ -434,6 +434,7 @@ static int _set_status(unsigned gt_version,
>> >          return _set_status_v2(domid, readonly, mapflag, shah, act, 
> status);
>> >  }
>> >  
>> > +#ifdef HAS_PASSTHROUGH
>> >  static void mapcount(
>> >      struct domain *ld, unsigned long mfn,
>> >      unsigned int *wrc, unsigned int *rdc)
>> > @@ -456,6 +457,7 @@ static void mapcount(
>> >          rcu_unlock_domain(rd);
>> >      }
>> >  }
>> > +#endif
>> >  
>> >  /*
>> >   * Returns 0 if TLB flush / invalidate required by caller.
>> > @@ -662,6 +664,7 @@ __gnttab_map_grant_ref(
>> >          goto undo_out;
>> >      }
>> >  
>> > +#ifdef HAS_PASSTHROUGH
>> >      if ( !is_hvm_domain(ld) && need_iommu(ld) )
>> 
>> Wouldn't #define-ing need_iommu() to 0 in the ARM headers achieve
>> the same effect, without cluttering common code?
> 
> Yes. I'd thought there were other uses of need_iommu but in fact they
> are all in x86.

And even if there were, you'd want them to resolve to 0 too until you
have an IOMMU implementation.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel