What about converting vtd_pt_enable_fast_path into vtd_try_enable_pt_fast_path?

It would return bool instead of being void.

The checks would be done inside instead of being the responsibility of the 
caller (a bit like vtd_switch_address_space).

It would also make sure that the cache is correctly updated of invalidated 
after the call.

Maybe at some point vtd_pt_enable_fast_path and vtd_switch_address_space will 
be so close that they can become a single function?

On Thu, 2026-11-05 at 09:07 +0000, Duan, Zhenzhong wrote:
> Caution: External email. Do not open attachments or click links, unless this 
> email comes from a known sender and you know the content is safe.
> 
> 
> Hi Clement,
> 
> Good finding!
> 
> It looks not easy to refactor all these as  
> vtd_switch_address_space() is called in many callsites,  
> not only fast path.
> 
> I see in vtd_switch_address_space() vtd_as_pt_enabled()  
> is called twice, that's obviously unnecessary.
> 
> I'm not sure if it's possible to cache CE and PE before  
> all vtd_switch_address_space() so it only needs to check  
> cache for PT.
> 
> Thanks  
> Zhenzhong
> 
> 
> > -----Original Message-----
> > From: Clément MATHIEU--DRIF 
> > <[[email protected]](mailto:[email protected])>
> > Subject: Re: [PATCH v2] intel_iommu: Keep PASID cache consistent with PT
> > fast path
> > 
> > Hi Zhenzhong,
> > 
> > I traced the calls we have and I fear the flow is becoming a bit spaghetti.
> > 
> > ```
> > vtd_do_iommu_translate
> >       |
> >        --> if (vtd_dev_pt_enabled) vtd_pt_enable_fast_path
> >                                               |
> >                                               v
> >                                   vtd_switch_address_space
> >                                       (updates the pasid
> >                                       entry based on the
> >                                       parameter it has
> >                                       been given)
> >                                               |
> >                                               v
> >                                       vtd_as_pt_enabled
> >                                         (drops the
> >                                         entry returned
> >                                         by vtd_dev_pt_enabled)
> >                                               |
> >                                               v
> >                                       vtd_dev_pt_enabled
> > ```
> > 
> > It looks like we are checking things twice, ignoring the result on the
> > second call and re-using the first one.
> > 
> > I think we are trying to gate the call to vtd_pt_enable_fast_path from
> > vtd_do_iommu_translate whereas the callee could do the thing on its own.
> > 
> > What is your opinion about this?
> > 
> > 
> > 
> > On Wed, 2026-11-04 at 15:07 +0800, Zhenzhong Duan wrote:
> > 
> > > Caution: External email. Do not open attachments or click links, unless
> > 
> > this email comes from a known sender and you know the content is safe.
> > 
> > > 
> > > 
> > > With caching mode disabled, a guest can install a scalable-mode
> > > pass-through PASID entry without issuing a PASID-cache invalidation. The
> > > first DMA enables the PT fast path, but leaves QEMU's PASID cache
> > 
> > invalid.
> > 
> > > 
> > > When the guest tears down that entry to switch back to a DMA domain, the
> > > invalidation handler sees both a non-present guest entry and an invalid
> > > cached entry and returns without re-enabling the IOMMU memory region.
> > > Subsequent DMA therefore continues to bypass translation.
> > > 
> > > The issue can be reproduced with an emulated device in scalable mode
> > > with caching-mode=off:
> > > 
> > > 1. Unbind the guest device driver and change the group's type to
> > >    identity through /sys/kernel/iommu_groups/<group>/type.
> > > 2. Rebind the driver and issue a DMA write.
> > > 3. Unbind the driver, change the group's type back to DMA-FQ, and
> > >    rebind the driver.
> > > 4. Issue another DMA write using an IOVA mapped to a different GPA.
> > > 
> > > The second write incorrectly bypasses translation instead of reaching
> > > the mapped GPA.
> > > 
> > > Save the PASID entry when enabling the scalable-mode PT fast path so
> > > teardown invalidation can restore the IOMMU address space. Return after
> > > handling a non-present entry, and refresh the cached entry when a present
> > > entry changes.
> > > 
> > > Opportunistically clarify the invalidation modes in the fast-path
> > 
> > comment.
> > 
> > > 
> > > Fixes: c9cf33075a4c ("intel_iommu: Fix DMA failure when guest switches
> > 
> > IOMMU domain")
> > 
> > > Reported-by: Clement Mathieu--Drif <[clement.mathieu--
> > 
> > [[email protected]](mailto:[email protected])](mailto:[[email protected]](mailto:[email protected]))>
> > 
> > > Signed-off-by: Zhenzhong Duan
> > 
> > <[[[email protected]](mailto:[email protected])](mailto:[[email protected]](mailto:[email protected]))>
> > 
> > > Tested-by: Clement Mathieu--Drif <[clement.mathieu--
> > 
> > [[email protected]](mailto:[email protected])](mailto:[[email protected]](mailto:[email protected]))>
> > 
> > > ---
> > >  hw/i386/intel_iommu.c | 32 ++++++++++++++++++++++++--------
> > >  1 file changed, 24 insertions(+), 8 deletions(-)
> > > 
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index b7da5d086d..2c5aa04f69 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -1649,7 +1649,7 @@ static int vtd_address_space_sync(VTDAddressSpace
> > 
> > *vtd_as)
> > 
> > >   * or second stage translation, it depends on PGTT setting.
> > >   */
> > >  static bool vtd_dev_pt_enabled(IntelIOMMUState *s, VTDContextEntry *ce,
> > > -                               uint32_t pasid)
> > > +                               uint32_t pasid, VTDPASIDEntry
> > 
> > *pasid_entry)
> > 
> > >  {
> > >      VTDPASIDEntry pe;
> > >      int ret;
> > > @@ -1663,6 +1663,11 @@ static bool vtd_dev_pt_enabled(IntelIOMMUState *s,
> > 
> > VTDContextEntry *ce,
> > 
> > >               */
> > >              return false;
> > >          }
> > > +
> > > +        if (pasid_entry) {
> > > +            *pasid_entry = pe;
> > > +        }
> > > +
> > >          return vtd_pe_pgtt_is_pt(&pe);
> > >      }
> > > 
> > > @@ -1689,7 +1694,7 @@ static bool vtd_as_pt_enabled(VTDAddressSpace *as)
> > >          return false;
> > >      }
> > > 
> > > -    return vtd_dev_pt_enabled(s, &ce, as->pasid);
> > > +    return vtd_dev_pt_enabled(s, &ce, as->pasid, NULL);
> > >  }
> > > 
> > >  /* Return whether the device is using IOMMU translation. */
> > > @@ -1887,7 +1892,8 @@ VTDAddressSpace *vtd_get_as_by_sid(IntelIOMMUState
> > 
> > *s, uint16_t sid)
> > 
> > >      return vtd_get_as_by_sid_and_pasid(s, sid, IOMMU_NO_PASID);
> > >  }
> > > 
> > > -static void vtd_pt_enable_fast_path(IntelIOMMUState *s, uint16_t
> > 
> > source_id)
> > 
> > > +static void vtd_pt_enable_fast_path(IntelIOMMUState *s, uint16_t
> > 
> > source_id,
> > 
> > > +                                    VTDPASIDEntry *pe)
> > >  {
> > >      VTDAddressSpace *vtd_as;
> > >      bool success = false;
> > > @@ -1900,6 +1906,13 @@ static void
> > 
> > vtd_pt_enable_fast_path(IntelIOMMUState *s, uint16_t source_id)
> > 
> > >      if (vtd_switch_address_space(vtd_as) == false) {
> > >          /* We switched off IOMMU region successfully. */
> > >          success = true;
> > > +
> > > +        if (s->root_scalable) {
> > > +            VTDPASIDCacheEntry *pc_entry = &vtd_as->pasid_cache_entry;
> > > +
> > > +            pc_entry->pasid_entry = *pe;
> > > +            pc_entry->valid = true;
> > > +        }
> > >      }
> > > 
> > >  out:
> > > @@ -2091,6 +2104,7 @@ static bool vtd_do_iommu_translate(VTDAddressSpace
> > 
> > *vtd_as, PCIBus *bus,
> > 
> > >  {
> > >      IntelIOMMUState *s = vtd_as->iommu_state;
> > >      VTDContextEntry ce;
> > > +    VTDPASIDEntry pe;
> > >      uint8_t bus_num = pci_bus_num(bus);
> > >      VTDContextCacheEntry *cc_entry;
> > >      uint64_t pte, page_mask;
> > > @@ -2167,7 +2181,7 @@ static bool vtd_do_iommu_translate(VTDAddressSpace
> > 
> > *vtd_as, PCIBus *bus,
> > 
> > >       * We don't need to translate for pass-through context entries.
> > >       * Also, let's ignore IOTLB caching as well for PT devices.
> > >       */
> > > -    if (vtd_dev_pt_enabled(s, &ce, pasid)) {
> > > +    if (vtd_dev_pt_enabled(s, &ce, pasid, &pe)) {
> > >          entry->iova = addr & VTD_PAGE_MASK_4K;
> > >          entry->translated_addr = entry->iova;
> > >          entry->addr_mask = ~VTD_PAGE_MASK_4K;
> > > @@ -2180,10 +2194,11 @@ static bool
> > 
> > vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
> > 
> > >           * the device. Let's enable the fast path for passthrough.
> > >           *
> > >           * When passthrough is disabled again for the device, we can
> > > -         * capture it via the context entry invalidation, then the
> > > -         * IOMMU region can be swapped back.
> > > +         * capture it via context-cache invalidation in legacy mode or
> > > +         * PASID-cache invalidation in scalable mode, then the IOMMU
> > > +         * region can be swapped back.
> > >           */
> > > -        vtd_pt_enable_fast_path(s, source_id);
> > > +        vtd_pt_enable_fast_path(s, source_id, &pe);
> > >          vtd_iommu_unlock(s);
> > >          return true;
> > >      }
> > > @@ -3133,6 +3148,7 @@ static void vtd_pasid_cache_sync_locked(gpointer
> > 
> > key, gpointer value,
> > 
> > >              vtd_address_space_unmap(vtd_as, n);
> > >          }
> > >          vtd_switch_address_space(vtd_as);
> > > +        return;
> > >      }
> > > 
> > >      /*
> > > @@ -3158,11 +3174,11 @@ static void vtd_pasid_cache_sync_locked(gpointer
> > 
> > key, gpointer value,
> > 
> > >      }
> > > 
> > >      if (!pc_entry->valid) {
> > > -        pc_entry->pasid_entry = pe;
> > >          pc_entry->valid = true;
> > >      } else if (!vtd_pasid_entry_compare(&pe, &pc_entry->pasid_entry)) {
> > >          return;
> > >      }
> > > +    pc_entry->pasid_entry = pe;
> > > 
> > >      vtd_switch_address_space(vtd_as);
> > >      vtd_address_space_sync(vtd_as);
> > > --
> > > 2.52.0
> > > 
> >
> 

Reply via email to