Re: [PATCH V2] intel_iommu: TM field should not be in reserved bits

[Peter Xu <peterx@redhat.com>]

On Fri, Sep 27, 2019 at 08:03:21AM +0000, Zhang, Qi1 wrote:
> 
> 
> > -----Original Message-----
> > From: Peter Xu <peterx@redhat.com>
> > Sent: Friday, September 27, 2019 2:10 PM
> > To: Zhang, Qi1 <qi1.zhang@intel.com>
> > Cc: qemu-devel@nongnu.org; ehabkost@redhat.com; mst@redhat.com;
> > pbonzini@redhat.com; rth@twiddle.net
> > Subject: Re: [PATCH V2] intel_iommu: TM field should not be in reserved bits
> > 
> > On Fri, Sep 27, 2019 at 12:58:38PM +0800, qi1.zhang@intel.com wrote:
> > > From: "Zhang, Qi" <qi1.zhang@intel.com>
> > >
> > > When dt is supported, TM field should not be Reserved(0).
> > >
> > > Refer to VT-d Spec 9.8
> > >
> > > Signed-off-by: Zhang, Qi <qi1.zhang@intel.com>
> > > Signed-off-by: Qi, Yadong <yadong.qi@intel.com>
> > > ---
> > >  hw/i386/intel_iommu.c          | 12 ++++++------
> > >  hw/i386/intel_iommu_internal.h | 25 +++++++++++++++++++------
> > >  2 files changed, 25 insertions(+), 12 deletions(-)
> > > ---
> > > Changelog V2:
> > >  move dt_supported flag to VTD_SPTE_PAGE_LX_RSVD_MASK and
> > > VTD_SPTE_LPAGE_LX_RSVD_MASK
> > >
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c index
> > > f1de8fdb75..35222cf55c 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -3548,13 +3548,13 @@ static void vtd_init(IntelIOMMUState *s)
> > >       * Rsvd field masks for spte
> > >       */
> > >      vtd_paging_entry_rsvd_field[0] = ~0ULL;
> > > -    vtd_paging_entry_rsvd_field[1] = VTD_SPTE_PAGE_L1_RSVD_MASK(s-
> > >aw_bits);
> > > -    vtd_paging_entry_rsvd_field[2] = VTD_SPTE_PAGE_L2_RSVD_MASK(s-
> > >aw_bits);
> > > -    vtd_paging_entry_rsvd_field[3] = VTD_SPTE_PAGE_L3_RSVD_MASK(s-
> > >aw_bits);
> > > +    vtd_paging_entry_rsvd_field[1] = VTD_SPTE_PAGE_L1_RSVD_MASK(s-
> > >aw_bits, x86_iommu->dt_supported);
> > > +    vtd_paging_entry_rsvd_field[2] = VTD_SPTE_PAGE_L2_RSVD_MASK(s-
> > >aw_bits, x86_iommu->dt_supported);
> > > +    vtd_paging_entry_rsvd_field[3] =
> > > + VTD_SPTE_PAGE_L3_RSVD_MASK(s->aw_bits, x86_iommu-
> > >dt_supported);
> > >      vtd_paging_entry_rsvd_field[4] = VTD_SPTE_PAGE_L4_RSVD_MASK(s-
> > >aw_bits);
> > > -    vtd_paging_entry_rsvd_field[5] = VTD_SPTE_LPAGE_L1_RSVD_MASK(s-
> > >aw_bits);
> > > -    vtd_paging_entry_rsvd_field[6] = VTD_SPTE_LPAGE_L2_RSVD_MASK(s-
> > >aw_bits);
> > > -    vtd_paging_entry_rsvd_field[7] = VTD_SPTE_LPAGE_L3_RSVD_MASK(s-
> > >aw_bits);
> > > +    vtd_paging_entry_rsvd_field[5] =
> > VTD_SPTE_LPAGE_L1_RSVD_MASK(s->aw_bits, x86_iommu-
> > >dt_supported);
> > > +    vtd_paging_entry_rsvd_field[6] =
> > VTD_SPTE_LPAGE_L2_RSVD_MASK(s->aw_bits, x86_iommu-
> > >dt_supported);
> > > +    vtd_paging_entry_rsvd_field[7] =
> > > + VTD_SPTE_LPAGE_L3_RSVD_MASK(s->aw_bits, x86_iommu-
> > >dt_supported);
> > >      vtd_paging_entry_rsvd_field[8] =
> > > VTD_SPTE_LPAGE_L4_RSVD_MASK(s->aw_bits);
> > 
> > Should this TM bit only affects leaves?  Say, entry 1 (4K), 5 (2M), 6 (1G).
> > While this reminded me that I'm totally confused on why we have had entry
> > 7, 8 after all...  Are they really used?
> Yes. TM bit only affects. To this array, index 1, 5,6,7 may be leaf. Will update a new patchset for it.

Could I ask why index 7 may be leaf?

> > 
> > >
> > >      if (x86_iommu_ir_supported(x86_iommu)) { diff --git
> > > a/hw/i386/intel_iommu_internal.h b/hw/i386/intel_iommu_internal.h
> > > index c1235a7063..01f1aa6c86 100644
> > > --- a/hw/i386/intel_iommu_internal.h
> > > +++ b/hw/i386/intel_iommu_internal.h
> > > @@ -387,19 +387,31 @@ typedef union VTDInvDesc VTDInvDesc;  #define
> > > VTD_INV_DESC_DEVICE_IOTLB_RSVD_LO 0xffff0000ffe0fff8
> > >
> > >  /* Rsvd field masks for spte */
> > > -#define VTD_SPTE_PAGE_L1_RSVD_MASK(aw) \
> > > +#define VTD_SPTE_PAGE_L1_RSVD_MASK(aw, dt_supported) \
> > > +        dt_supported? \
> > > +        (0x800ULL | ~(VTD_HAW_MASK(aw) | VTD_SL_IGN_COM |
> > VTD_SL_TM))
> > > +: \
> > >          (0x800ULL | ~(VTD_HAW_MASK(aw) | VTD_SL_IGN_COM))
> > 
> > This seems strange too in that ~VTD_HAW_MASK(aw) probably covered bits
> > 63-48 for aw==48 case so it should already cover VTD_SL_TM?
> VTD_SL_IGN_COM 0xbff0000000000000ULL, TM field is cleared by ~ VTD_SL_IGN_COM
> > 
> > Meanwhile when I'm reading the spec I see at least bits 61-52 ignored rather
> > than reserved.
> Yes. Bit 61~52 is ignored. Such as the index 5 of this array is 0xfff8000000800.

Oops, my poor eye obviously didn't see that the "~" operator is
applied over the whole (VTD_HAW_MASK(aw) | VTD_SL_IGN_COM)... :)

Btw, you should only touch up the macros that are leaves here.
Non-leaves should still keep that bit as reserved.

Thanks,

-- 
Peter Xu