[PATCH 0/3] Fix UNMAP notifier for intel-iommu

[PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Jason Wang]

Hi All:

According to ATS, device should work if ATS is disabled. This is not
correctly implemented in the current intel-iommu since it doesn't
handle the UNMAP notifier correctly. This breaks the vhost-net +
vIOMMU without dt.

The root casue is that the when there's a device IOTLB miss (note that
it's not specific to PCI so it can work without ATS), Qemu doesn't
build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
won't trigger the UNMAP notifier.

Fixing by build IOVA tree during IOMMU translsation.

Thanks

Jason Wang (3):
  intel-iommu: fail MAP notifier without caching mode
  intel-iommu: fail DEVIOTLB_UNMAP without dt mode
  intel-iommu: build iova tree during IOMMU translation

 hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
 1 file changed, 33 insertions(+), 25 deletions(-)

-- 
2.25.1

[PATCH 1/3] intel-iommu: fail MAP notifier without caching mode

[Jason Wang]

Without caching mode, MAP notifier won't work correctly since guest
won't send IOTLB update event when it establishes new mappings in the
I/O page tables. Let's fail the IOMMU notifiers early instead of
misbehaving silently.

Signed-off-by: Jason Wang <jasowang@redhat.com>
---
 hw/i386/intel_iommu.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index a08ee85edf..9143376677 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -3186,6 +3186,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
                          "Snoop Control with vhost or VFIO is not supported");
         return -ENOTSUP;
     }
+    if (!s->caching_mode && (new & IOMMU_NOTIFIER_MAP)) {
+        error_setg_errno(errp, ENOTSUP,
+                         "device %02x.%02x.%x requires caching mode",
+                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
+                         PCI_FUNC(vtd_as->devfn));
+        return -ENOTSUP;
+    }
 
     /* Update per-address-space notifier flags */
     vtd_as->notifier_flags = new;
-- 
2.25.1

[PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Jason Wang]

Without dt mode, device IOTLB notifier won't work since guest won't
send device IOTLB invalidation descriptor in this case. Let's fail
early instead of misbehaving silently.

Signed-off-by: Jason Wang <jasowang@redhat.com>
---
 hw/i386/intel_iommu.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index 9143376677..d025ef2873 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
 {
     VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
     IntelIOMMUState *s = vtd_as->iommu_state;
+    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
 
     /* TODO: add support for VFIO and vhost users */
     if (s->snoop_control) {
@@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
                          PCI_FUNC(vtd_as->devfn));
         return -ENOTSUP;
     }
+    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
+        error_setg_errno(errp, ENOTSUP,
+                         "device %02x.%02x.%x requires device IOTLB mode",
+                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
+                         PCI_FUNC(vtd_as->devfn));
+        return -ENOTSUP;
+    }
 
     /* Update per-address-space notifier flags */
     vtd_as->notifier_flags = new;
-- 
2.25.1

[PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Jason Wang]

The IOVA tree is only built during page walk this breaks the device
that tries to use UNMAP notifier only. One example is vhost-net, it
tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
notifier (e.g when dt mode is not enabled). The interesting part is
that it doesn't use MAP since it can query the IOMMU translation by
itself upon a IOTLB miss.

This doesn't work since Qemu doesn't build IOVA tree in IOMMU
translation which means the UNMAP notifier won't be triggered during
the page walk since Qemu think it is never mapped. This could be
noticed when vIOMMU is used with vhost_net but dt is disabled.

Fixing this by build the iova tree during IOMMU translation, this
makes sure the UNMAP notifier event could be identified during page
walk. And we need to walk page table not only for UNMAP notifier but
for MAP notifier during PSI.

Signed-off-by: Jason Wang <jasowang@redhat.com>
---
 hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
 1 file changed, 18 insertions(+), 25 deletions(-)

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index d025ef2873..edeb62f4b2 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
     uint8_t access_flags;
     bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
     VTDIOTLBEntry *iotlb_entry;
+    const DMAMap *mapped;
+    DMAMap target;
 
     /*
      * We have standalone memory region for interrupt addresses, we
@@ -1954,6 +1956,21 @@ out:
     entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
     entry->addr_mask = ~page_mask;
     entry->perm = access_flags;
+
+    target.iova = entry->iova;
+    target.size = entry->addr_mask;
+    target.translated_addr = entry->translated_addr;
+    target.perm = entry->perm;
+
+    mapped = iova_tree_find(vtd_as->iova_tree, &target);
+    if (!mapped) {
+        /* To make UNMAP notifier work, we need build iova tree here
+         * in order to have the UNMAP iommu notifier to be triggered
+         * during the page walk.
+         */
+        iova_tree_insert(vtd_as->iova_tree, &target);
+    }
+
     return true;
 
 error:
@@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
         ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
                                        vtd_as->devfn, &ce);
         if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
-            if (vtd_as_has_map_notifier(vtd_as)) {
-                /*
-                 * As long as we have MAP notifications registered in
-                 * any of our IOMMU notifiers, we need to sync the
-                 * shadow page table.
-                 */
-                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
-            } else {
-                /*
-                 * For UNMAP-only notifiers, we don't need to walk the
-                 * page tables.  We just deliver the PSI down to
-                 * invalidate caches.
-                 */
-                IOMMUTLBEvent event = {
-                    .type = IOMMU_NOTIFIER_UNMAP,
-                    .entry = {
-                        .target_as = &address_space_memory,
-                        .iova = addr,
-                        .translated_addr = 0,
-                        .addr_mask = size - 1,
-                        .perm = IOMMU_NONE,
-                    },
-                };
-                memory_region_notify_iommu(&vtd_as->iommu, 0, event);
-            }
+            vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
         }
     }
 }
-- 
2.25.1

Re: [PATCH 1/3] intel-iommu: fail MAP notifier without caching mode

[Peter Xu]

On Tue, Nov 29, 2022 at 04:10:35PM +0800, Jason Wang wrote:
> Without caching mode, MAP notifier won't work correctly since guest
> won't send IOTLB update event when it establishes new mappings in the
> I/O page tables. Let's fail the IOMMU notifiers early instead of
> misbehaving silently.
> 
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
>  hw/i386/intel_iommu.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index a08ee85edf..9143376677 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -3186,6 +3186,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>                           "Snoop Control with vhost or VFIO is not supported");
>          return -ENOTSUP;
>      }
> +    if (!s->caching_mode && (new & IOMMU_NOTIFIER_MAP)) {
> +        error_setg_errno(errp, ENOTSUP,
> +                         "device %02x.%02x.%x requires caching mode",
> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> +                         PCI_FUNC(vtd_as->devfn));
> +        return -ENOTSUP;
> +    }

We used to have that but got reverted because it's too late to fail, so we
moved it over even though not as clean..

https://lore.kernel.org/all/20190916080718.3299-5-peterx@redhat.com/

Thanks,

-- 
Peter Xu

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Peter Xu]

On Tue, Nov 29, 2022 at 04:10:36PM +0800, Jason Wang wrote:
> Without dt mode, device IOTLB notifier won't work since guest won't
> send device IOTLB invalidation descriptor in this case. Let's fail
> early instead of misbehaving silently.
> 
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
>  hw/i386/intel_iommu.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index 9143376677..d025ef2873 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>  {
>      VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
>      IntelIOMMUState *s = vtd_as->iommu_state;
> +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
>  
>      /* TODO: add support for VFIO and vhost users */
>      if (s->snoop_control) {
> @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>                           PCI_FUNC(vtd_as->devfn));
>          return -ENOTSUP;
>      }
> +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> +        error_setg_errno(errp, ENOTSUP,
> +                         "device %02x.%02x.%x requires device IOTLB mode",
> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> +                         PCI_FUNC(vtd_as->devfn));
> +        return -ENOTSUP;
> +    }

Hmm I thought we have had this already, so we don't?... :-(

Reviewed-by: Peter Xu <peterx@redhat.com>

-- 
Peter Xu

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Peter Xu]

On Tue, Nov 29, 2022 at 04:10:37PM +0800, Jason Wang wrote:
> The IOVA tree is only built during page walk this breaks the device
> that tries to use UNMAP notifier only. One example is vhost-net, it
> tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
> notifier (e.g when dt mode is not enabled). The interesting part is
> that it doesn't use MAP since it can query the IOMMU translation by
> itself upon a IOTLB miss.
> 
> This doesn't work since Qemu doesn't build IOVA tree in IOMMU
> translation which means the UNMAP notifier won't be triggered during
> the page walk since Qemu think it is never mapped. This could be
> noticed when vIOMMU is used with vhost_net but dt is disabled.
> 
> Fixing this by build the iova tree during IOMMU translation, this
> makes sure the UNMAP notifier event could be identified during page
> walk. And we need to walk page table not only for UNMAP notifier but
> for MAP notifier during PSI.
> 
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
>  hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
>  1 file changed, 18 insertions(+), 25 deletions(-)
> 
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index d025ef2873..edeb62f4b2 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
>      uint8_t access_flags;
>      bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
>      VTDIOTLBEntry *iotlb_entry;
> +    const DMAMap *mapped;
> +    DMAMap target;
>  
>      /*
>       * We have standalone memory region for interrupt addresses, we
> @@ -1954,6 +1956,21 @@ out:
>      entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
>      entry->addr_mask = ~page_mask;
>      entry->perm = access_flags;
> +
> +    target.iova = entry->iova;
> +    target.size = entry->addr_mask;
> +    target.translated_addr = entry->translated_addr;
> +    target.perm = entry->perm;
> +
> +    mapped = iova_tree_find(vtd_as->iova_tree, &target);
> +    if (!mapped) {
> +        /* To make UNMAP notifier work, we need build iova tree here
> +         * in order to have the UNMAP iommu notifier to be triggered
> +         * during the page walk.
> +         */
> +        iova_tree_insert(vtd_as->iova_tree, &target);
> +    }
> +
>      return true;
>  
>  error:
> @@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
>          ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
>                                         vtd_as->devfn, &ce);
>          if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
> -            if (vtd_as_has_map_notifier(vtd_as)) {
> -                /*
> -                 * As long as we have MAP notifications registered in
> -                 * any of our IOMMU notifiers, we need to sync the
> -                 * shadow page table.
> -                 */
> -                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> -            } else {
> -                /*
> -                 * For UNMAP-only notifiers, we don't need to walk the
> -                 * page tables.  We just deliver the PSI down to
> -                 * invalidate caches.
> -                 */
> -                IOMMUTLBEvent event = {
> -                    .type = IOMMU_NOTIFIER_UNMAP,
> -                    .entry = {
> -                        .target_as = &address_space_memory,
> -                        .iova = addr,
> -                        .translated_addr = 0,
> -                        .addr_mask = size - 1,
> -                        .perm = IOMMU_NONE,
> -                    },
> -                };
> -                memory_region_notify_iommu(&vtd_as->iommu, 0, event);

Isn't this path the one that will be responsible for pass-through the UNMAP
events from guest to vhost when there's no MAP notifier requested?

At least that's what I expected when introducing the iova tree, because for
unmap-only device hierachy I thought we didn't need the tree at all.

Jason, do you know where I miss?

Thanks,

> -            }
> +            vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
>          }
>      }
>  }
> -- 
> 2.25.1
> 

-- 
Peter Xu

Re: [PATCH 1/3] intel-iommu: fail MAP notifier without caching mode

[Jason Wang]

On Tue, Nov 29, 2022 at 11:35 PM Peter Xu <peterx@redhat.com> wrote:
>
> On Tue, Nov 29, 2022 at 04:10:35PM +0800, Jason Wang wrote:
> > Without caching mode, MAP notifier won't work correctly since guest
> > won't send IOTLB update event when it establishes new mappings in the
> > I/O page tables. Let's fail the IOMMU notifiers early instead of
> > misbehaving silently.
> >
> > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > ---
> >  hw/i386/intel_iommu.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> >
> > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > index a08ee85edf..9143376677 100644
> > --- a/hw/i386/intel_iommu.c
> > +++ b/hw/i386/intel_iommu.c
> > @@ -3186,6 +3186,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> >                           "Snoop Control with vhost or VFIO is not supported");
> >          return -ENOTSUP;
> >      }
> > +    if (!s->caching_mode && (new & IOMMU_NOTIFIER_MAP)) {
> > +        error_setg_errno(errp, ENOTSUP,
> > +                         "device %02x.%02x.%x requires caching mode",
> > +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> > +                         PCI_FUNC(vtd_as->devfn));
> > +        return -ENOTSUP;
> > +    }
>
> We used to have that but got reverted because it's too late to fail, so we
> moved it over even though not as clean..
>
> https://lore.kernel.org/all/20190916080718.3299-5-peterx@redhat.com/

One of the difference is that the patch doesn't do exit() here. I
think it's better to fail instead of misbehving silently, this is what
other vIOMMU did:

E.g in smmu we had:

    if (new & IOMMU_NOTIFIER_MAP) {
        error_setg(errp,
                   "device %02x.%02x.%x requires iommu MAP notifier which is "
                   "not currently supported", pci_bus_num(sdev->bus),
                   PCI_SLOT(sdev->devfn), PCI_FUNC(sdev->devfn));
        return -EINVAL;
    }

So did for amd iommu.

Thanks

>
> Thanks,
>
> --
> Peter Xu
>

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Jason Wang]

On Tue, Nov 29, 2022 at 11:57 PM Peter Xu <peterx@redhat.com> wrote:
>
> On Tue, Nov 29, 2022 at 04:10:37PM +0800, Jason Wang wrote:
> > The IOVA tree is only built during page walk this breaks the device
> > that tries to use UNMAP notifier only. One example is vhost-net, it
> > tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
> > notifier (e.g when dt mode is not enabled). The interesting part is
> > that it doesn't use MAP since it can query the IOMMU translation by
> > itself upon a IOTLB miss.
> >
> > This doesn't work since Qemu doesn't build IOVA tree in IOMMU
> > translation which means the UNMAP notifier won't be triggered during
> > the page walk since Qemu think it is never mapped. This could be
> > noticed when vIOMMU is used with vhost_net but dt is disabled.
> >
> > Fixing this by build the iova tree during IOMMU translation, this
> > makes sure the UNMAP notifier event could be identified during page
> > walk. And we need to walk page table not only for UNMAP notifier but
> > for MAP notifier during PSI.
> >
> > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > ---
> >  hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
> >  1 file changed, 18 insertions(+), 25 deletions(-)
> >
> > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > index d025ef2873..edeb62f4b2 100644
> > --- a/hw/i386/intel_iommu.c
> > +++ b/hw/i386/intel_iommu.c
> > @@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
> >      uint8_t access_flags;
> >      bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
> >      VTDIOTLBEntry *iotlb_entry;
> > +    const DMAMap *mapped;
> > +    DMAMap target;
> >
> >      /*
> >       * We have standalone memory region for interrupt addresses, we
> > @@ -1954,6 +1956,21 @@ out:
> >      entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
> >      entry->addr_mask = ~page_mask;
> >      entry->perm = access_flags;
> > +
> > +    target.iova = entry->iova;
> > +    target.size = entry->addr_mask;
> > +    target.translated_addr = entry->translated_addr;
> > +    target.perm = entry->perm;
> > +
> > +    mapped = iova_tree_find(vtd_as->iova_tree, &target);
> > +    if (!mapped) {
> > +        /* To make UNMAP notifier work, we need build iova tree here
> > +         * in order to have the UNMAP iommu notifier to be triggered
> > +         * during the page walk.
> > +         */
> > +        iova_tree_insert(vtd_as->iova_tree, &target);
> > +    }
> > +
> >      return true;
> >
> >  error:
> > @@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
> >          ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
> >                                         vtd_as->devfn, &ce);
> >          if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
> > -            if (vtd_as_has_map_notifier(vtd_as)) {
> > -                /*
> > -                 * As long as we have MAP notifications registered in
> > -                 * any of our IOMMU notifiers, we need to sync the
> > -                 * shadow page table.
> > -                 */
> > -                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> > -            } else {
> > -                /*
> > -                 * For UNMAP-only notifiers, we don't need to walk the
> > -                 * page tables.  We just deliver the PSI down to
> > -                 * invalidate caches.
> > -                 */
> > -                IOMMUTLBEvent event = {
> > -                    .type = IOMMU_NOTIFIER_UNMAP,
> > -                    .entry = {
> > -                        .target_as = &address_space_memory,
> > -                        .iova = addr,
> > -                        .translated_addr = 0,
> > -                        .addr_mask = size - 1,
> > -                        .perm = IOMMU_NONE,
> > -                    },
> > -                };
> > -                memory_region_notify_iommu(&vtd_as->iommu, 0, event);
>
> Isn't this path the one that will be responsible for pass-through the UNMAP
> events from guest to vhost when there's no MAP notifier requested?

Yes, but it doesn't do the iova tree removing. More below.

>
> At least that's what I expected when introducing the iova tree, because for
> unmap-only device hierachy I thought we didn't need the tree at all.

Then the problem is the UNMAP notifier won't be trigger at all during
DSI page walk in vtd_page_walk_one() because there's no DMAMap stored
in the iova tree.:

        if (!mapped) {
            /* Skip since we didn't map this range at all */
            trace_vtd_page_walk_one_skip_unmap(entry->iova, entry->addr_mask);
            return 0;
        }

So I choose to build the iova tree in translate then we won't go
within the above condition.

Thanks

>
> Jason, do you know where I miss?
>
> Thanks,
>
> > -            }
> > +            vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> >          }
> >      }
> >  }
> > --
> > 2.25.1
> >
>
> --
> Peter Xu
>

Re: [PATCH 1/3] intel-iommu: fail MAP notifier without caching mode

[Peter Xu]

On Wed, Nov 30, 2022 at 02:23:06PM +0800, Jason Wang wrote:
> On Tue, Nov 29, 2022 at 11:35 PM Peter Xu <peterx@redhat.com> wrote:
> >
> > On Tue, Nov 29, 2022 at 04:10:35PM +0800, Jason Wang wrote:
> > > Without caching mode, MAP notifier won't work correctly since guest
> > > won't send IOTLB update event when it establishes new mappings in the
> > > I/O page tables. Let's fail the IOMMU notifiers early instead of
> > > misbehaving silently.
> > >
> > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > ---
> > >  hw/i386/intel_iommu.c | 7 +++++++
> > >  1 file changed, 7 insertions(+)
> > >
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index a08ee85edf..9143376677 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -3186,6 +3186,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> > >                           "Snoop Control with vhost or VFIO is not supported");
> > >          return -ENOTSUP;
> > >      }
> > > +    if (!s->caching_mode && (new & IOMMU_NOTIFIER_MAP)) {
> > > +        error_setg_errno(errp, ENOTSUP,
> > > +                         "device %02x.%02x.%x requires caching mode",
> > > +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> > > +                         PCI_FUNC(vtd_as->devfn));
> > > +        return -ENOTSUP;
> > > +    }
> >
> > We used to have that but got reverted because it's too late to fail, so we
> > moved it over even though not as clean..
> >
> > https://lore.kernel.org/all/20190916080718.3299-5-peterx@redhat.com/
> 
> One of the difference is that the patch doesn't do exit() here. I
> think it's better to fail instead of misbehving silently, this is what
> other vIOMMU did:
> 
> E.g in smmu we had:
> 
>     if (new & IOMMU_NOTIFIER_MAP) {
>         error_setg(errp,
>                    "device %02x.%02x.%x requires iommu MAP notifier which is "
>                    "not currently supported", pci_bus_num(sdev->bus),
>                    PCI_SLOT(sdev->devfn), PCI_FUNC(sdev->devfn));
>         return -EINVAL;
>     }
> 
> So did for amd iommu.

Yeah that's fine.  Would you mind add the explanation (and also above link,
which might be helpful to pick up the history..) to the commit message?
With that:

Reviewed-by: Peter Xu <peterx@redhat.com>

Thanks,

-- 
Peter Xu

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Peter Xu]

On Wed, Nov 30, 2022 at 02:33:51PM +0800, Jason Wang wrote:
> On Tue, Nov 29, 2022 at 11:57 PM Peter Xu <peterx@redhat.com> wrote:
> >
> > On Tue, Nov 29, 2022 at 04:10:37PM +0800, Jason Wang wrote:
> > > The IOVA tree is only built during page walk this breaks the device
> > > that tries to use UNMAP notifier only. One example is vhost-net, it
> > > tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
> > > notifier (e.g when dt mode is not enabled). The interesting part is
> > > that it doesn't use MAP since it can query the IOMMU translation by
> > > itself upon a IOTLB miss.
> > >
> > > This doesn't work since Qemu doesn't build IOVA tree in IOMMU
> > > translation which means the UNMAP notifier won't be triggered during
> > > the page walk since Qemu think it is never mapped. This could be
> > > noticed when vIOMMU is used with vhost_net but dt is disabled.
> > >
> > > Fixing this by build the iova tree during IOMMU translation, this
> > > makes sure the UNMAP notifier event could be identified during page
> > > walk. And we need to walk page table not only for UNMAP notifier but
> > > for MAP notifier during PSI.
> > >
> > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > ---
> > >  hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
> > >  1 file changed, 18 insertions(+), 25 deletions(-)
> > >
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index d025ef2873..edeb62f4b2 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
> > >      uint8_t access_flags;
> > >      bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
> > >      VTDIOTLBEntry *iotlb_entry;
> > > +    const DMAMap *mapped;
> > > +    DMAMap target;
> > >
> > >      /*
> > >       * We have standalone memory region for interrupt addresses, we
> > > @@ -1954,6 +1956,21 @@ out:
> > >      entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
> > >      entry->addr_mask = ~page_mask;
> > >      entry->perm = access_flags;
> > > +
> > > +    target.iova = entry->iova;
> > > +    target.size = entry->addr_mask;
> > > +    target.translated_addr = entry->translated_addr;
> > > +    target.perm = entry->perm;
> > > +
> > > +    mapped = iova_tree_find(vtd_as->iova_tree, &target);
> > > +    if (!mapped) {
> > > +        /* To make UNMAP notifier work, we need build iova tree here
> > > +         * in order to have the UNMAP iommu notifier to be triggered
> > > +         * during the page walk.
> > > +         */
> > > +        iova_tree_insert(vtd_as->iova_tree, &target);
> > > +    }
> > > +
> > >      return true;
> > >
> > >  error:
> > > @@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
> > >          ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
> > >                                         vtd_as->devfn, &ce);
> > >          if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
> > > -            if (vtd_as_has_map_notifier(vtd_as)) {
> > > -                /*
> > > -                 * As long as we have MAP notifications registered in
> > > -                 * any of our IOMMU notifiers, we need to sync the
> > > -                 * shadow page table.
> > > -                 */
> > > -                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> > > -            } else {
> > > -                /*
> > > -                 * For UNMAP-only notifiers, we don't need to walk the
> > > -                 * page tables.  We just deliver the PSI down to
> > > -                 * invalidate caches.
> > > -                 */
> > > -                IOMMUTLBEvent event = {
> > > -                    .type = IOMMU_NOTIFIER_UNMAP,
> > > -                    .entry = {
> > > -                        .target_as = &address_space_memory,
> > > -                        .iova = addr,
> > > -                        .translated_addr = 0,
> > > -                        .addr_mask = size - 1,
> > > -                        .perm = IOMMU_NONE,
> > > -                    },
> > > -                };
> > > -                memory_region_notify_iommu(&vtd_as->iommu, 0, event);
> >
> > Isn't this path the one that will be responsible for pass-through the UNMAP
> > events from guest to vhost when there's no MAP notifier requested?
> 
> Yes, but it doesn't do the iova tree removing. More below.
> 
> >
> > At least that's what I expected when introducing the iova tree, because for
> > unmap-only device hierachy I thought we didn't need the tree at all.
> 
> Then the problem is the UNMAP notifier won't be trigger at all during
> DSI page walk in vtd_page_walk_one() because there's no DMAMap stored
> in the iova tree.:
> 
>         if (!mapped) {
>             /* Skip since we didn't map this range at all */
>             trace_vtd_page_walk_one_skip_unmap(entry->iova, entry->addr_mask);
>             return 0;
>         }
> 
> So I choose to build the iova tree in translate then we won't go
> within the above condition.

That's also why it's weird because IIUC we should never walk a page table
at all if there's no MAP notifier regiestered.

When I'm looking at the walk callers I found that indeed there's one path
missing where can cause it to actually walk the pgtables without !MAP, then
I also noticed commit f7701e2c7983b6, and I'm wondering what we really want
is something like this:

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index a08ee85edf..c46f3db992 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -1536,7 +1536,7 @@ static int vtd_sync_shadow_page_table(VTDAddressSpace *vtd_as)
     VTDContextEntry ce;
     IOMMUNotifier *n;

-    if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
+    if (!vtd_as_has_map_notifier(vtd_as)) {
         return 0;
     }

So I'm not sure whether this patch is the problem resolver; so far I feel
like it's patch 2 who does the real fix.  Then we can have the above
oneliner so we stop any walks when there's no map notifiers.

Thanks,

-- 
Peter Xu

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Michael S. Tsirkin]

On Tue, Nov 29, 2022 at 04:10:34PM +0800, Jason Wang wrote:
> Hi All:
> 
> According to ATS, device should work if ATS is disabled. This is not
> correctly implemented in the current intel-iommu since it doesn't
> handle the UNMAP notifier correctly. This breaks the vhost-net +
> vIOMMU without dt.
> 
> The root casue is that the when there's a device IOTLB miss (note that
> it's not specific to PCI so it can work without ATS), Qemu doesn't
> build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> won't trigger the UNMAP notifier.
> 
> Fixing by build IOVA tree during IOMMU translsation.
> 
> Thanks

Any changes of Fixes tags? this is 8.0 yes?

> Jason Wang (3):
>   intel-iommu: fail MAP notifier without caching mode
>   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
>   intel-iommu: build iova tree during IOMMU translation
> 
>  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
>  1 file changed, 33 insertions(+), 25 deletions(-)
> 
> -- 
> 2.25.1

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Jason Wang]

On Thu, Dec 1, 2022 at 12:38 AM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Tue, Nov 29, 2022 at 04:10:34PM +0800, Jason Wang wrote:
> > Hi All:
> >
> > According to ATS, device should work if ATS is disabled. This is not
> > correctly implemented in the current intel-iommu since it doesn't
> > handle the UNMAP notifier correctly. This breaks the vhost-net +
> > vIOMMU without dt.
> >
> > The root casue is that the when there's a device IOTLB miss (note that
> > it's not specific to PCI so it can work without ATS), Qemu doesn't
> > build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> > won't trigger the UNMAP notifier.
> >
> > Fixing by build IOVA tree during IOMMU translsation.
> >
> > Thanks
>
> Any changes of Fixes tags? this is 8.0 yes?

Yes, it's for 8.0.

Thanks

>
> > Jason Wang (3):
> >   intel-iommu: fail MAP notifier without caching mode
> >   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
> >   intel-iommu: build iova tree during IOMMU translation
> >
> >  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
> >  1 file changed, 33 insertions(+), 25 deletions(-)
> >
> > --
> > 2.25.1
>

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Jason Wang]

On Wed, Nov 30, 2022 at 11:17 PM Peter Xu <peterx@redhat.com> wrote:
>
> On Wed, Nov 30, 2022 at 02:33:51PM +0800, Jason Wang wrote:
> > On Tue, Nov 29, 2022 at 11:57 PM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > On Tue, Nov 29, 2022 at 04:10:37PM +0800, Jason Wang wrote:
> > > > The IOVA tree is only built during page walk this breaks the device
> > > > that tries to use UNMAP notifier only. One example is vhost-net, it
> > > > tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
> > > > notifier (e.g when dt mode is not enabled). The interesting part is
> > > > that it doesn't use MAP since it can query the IOMMU translation by
> > > > itself upon a IOTLB miss.
> > > >
> > > > This doesn't work since Qemu doesn't build IOVA tree in IOMMU
> > > > translation which means the UNMAP notifier won't be triggered during
> > > > the page walk since Qemu think it is never mapped. This could be
> > > > noticed when vIOMMU is used with vhost_net but dt is disabled.
> > > >
> > > > Fixing this by build the iova tree during IOMMU translation, this
> > > > makes sure the UNMAP notifier event could be identified during page
> > > > walk. And we need to walk page table not only for UNMAP notifier but
> > > > for MAP notifier during PSI.
> > > >
> > > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > > ---
> > > >  hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
> > > >  1 file changed, 18 insertions(+), 25 deletions(-)
> > > >
> > > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > > index d025ef2873..edeb62f4b2 100644
> > > > --- a/hw/i386/intel_iommu.c
> > > > +++ b/hw/i386/intel_iommu.c
> > > > @@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
> > > >      uint8_t access_flags;
> > > >      bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
> > > >      VTDIOTLBEntry *iotlb_entry;
> > > > +    const DMAMap *mapped;
> > > > +    DMAMap target;
> > > >
> > > >      /*
> > > >       * We have standalone memory region for interrupt addresses, we
> > > > @@ -1954,6 +1956,21 @@ out:
> > > >      entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
> > > >      entry->addr_mask = ~page_mask;
> > > >      entry->perm = access_flags;
> > > > +
> > > > +    target.iova = entry->iova;
> > > > +    target.size = entry->addr_mask;
> > > > +    target.translated_addr = entry->translated_addr;
> > > > +    target.perm = entry->perm;
> > > > +
> > > > +    mapped = iova_tree_find(vtd_as->iova_tree, &target);
> > > > +    if (!mapped) {
> > > > +        /* To make UNMAP notifier work, we need build iova tree here
> > > > +         * in order to have the UNMAP iommu notifier to be triggered
> > > > +         * during the page walk.
> > > > +         */
> > > > +        iova_tree_insert(vtd_as->iova_tree, &target);
> > > > +    }
> > > > +
> > > >      return true;
> > > >
> > > >  error:
> > > > @@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
> > > >          ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
> > > >                                         vtd_as->devfn, &ce);
> > > >          if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
> > > > -            if (vtd_as_has_map_notifier(vtd_as)) {
> > > > -                /*
> > > > -                 * As long as we have MAP notifications registered in
> > > > -                 * any of our IOMMU notifiers, we need to sync the
> > > > -                 * shadow page table.
> > > > -                 */
> > > > -                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> > > > -            } else {
> > > > -                /*
> > > > -                 * For UNMAP-only notifiers, we don't need to walk the
> > > > -                 * page tables.  We just deliver the PSI down to
> > > > -                 * invalidate caches.
> > > > -                 */
> > > > -                IOMMUTLBEvent event = {
> > > > -                    .type = IOMMU_NOTIFIER_UNMAP,
> > > > -                    .entry = {
> > > > -                        .target_as = &address_space_memory,
> > > > -                        .iova = addr,
> > > > -                        .translated_addr = 0,
> > > > -                        .addr_mask = size - 1,
> > > > -                        .perm = IOMMU_NONE,
> > > > -                    },
> > > > -                };
> > > > -                memory_region_notify_iommu(&vtd_as->iommu, 0, event);
> > >
> > > Isn't this path the one that will be responsible for pass-through the UNMAP
> > > events from guest to vhost when there's no MAP notifier requested?
> >
> > Yes, but it doesn't do the iova tree removing. More below.
> >
> > >
> > > At least that's what I expected when introducing the iova tree, because for
> > > unmap-only device hierachy I thought we didn't need the tree at all.
> >
> > Then the problem is the UNMAP notifier won't be trigger at all during
> > DSI page walk in vtd_page_walk_one() because there's no DMAMap stored
> > in the iova tree.:
> >
> >         if (!mapped) {
> >             /* Skip since we didn't map this range at all */
> >             trace_vtd_page_walk_one_skip_unmap(entry->iova, entry->addr_mask);
> >             return 0;
> >         }
> >
> > So I choose to build the iova tree in translate then we won't go
> > within the above condition.
>
> That's also why it's weird because IIUC we should never walk a page table
> at all if there's no MAP notifier regiestered.

If this is true, we probably need to document this somewhere.

>
> When I'm looking at the walk callers I found that indeed there's one path
> missing where can cause it to actually walk the pgtables without !MAP, then
> I also noticed commit f7701e2c7983b6, and I'm wondering what we really want
> is something like this:
>
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index a08ee85edf..c46f3db992 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -1536,7 +1536,7 @@ static int vtd_sync_shadow_page_table(VTDAddressSpace *vtd_as)
>      VTDContextEntry ce;
>      IOMMUNotifier *n;
>
> -    if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
> +    if (!vtd_as_has_map_notifier(vtd_as)) {
>          return 0;
>      }
>
> So I'm not sure whether this patch is the problem resolver; so far I feel
> like it's patch 2 who does the real fix.  Then we can have the above
> oneliner so we stop any walks when there's no map notifiers.
>
> Thanks,

I may miss something but as state above, the problem is a missing
UNMAP notification during DSI when there's only UNMAP notifier.

To solve it we might have two ways:

1) build the iova tree during iommu translation then we can correctly
trigger UNMAP during page walk caused by DSI
2) don't do the iova tree walk for !MAP notifier, need new logic to
trigger UNMAP notifier in PSI/DSI

This patch choose to go 1) (which seems easier at least for -stable).
Do you mean you prefer to go with 2)?

Thanks

>
> --
> Peter Xu
>

Re: [PATCH 1/3] intel-iommu: fail MAP notifier without caching mode

[Jason Wang]

On Wed, Nov 30, 2022 at 11:06 PM Peter Xu <peterx@redhat.com> wrote:
>
> On Wed, Nov 30, 2022 at 02:23:06PM +0800, Jason Wang wrote:
> > On Tue, Nov 29, 2022 at 11:35 PM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > On Tue, Nov 29, 2022 at 04:10:35PM +0800, Jason Wang wrote:
> > > > Without caching mode, MAP notifier won't work correctly since guest
> > > > won't send IOTLB update event when it establishes new mappings in the
> > > > I/O page tables. Let's fail the IOMMU notifiers early instead of
> > > > misbehaving silently.
> > > >
> > > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > > ---
> > > >  hw/i386/intel_iommu.c | 7 +++++++
> > > >  1 file changed, 7 insertions(+)
> > > >
> > > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > > index a08ee85edf..9143376677 100644
> > > > --- a/hw/i386/intel_iommu.c
> > > > +++ b/hw/i386/intel_iommu.c
> > > > @@ -3186,6 +3186,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> > > >                           "Snoop Control with vhost or VFIO is not supported");
> > > >          return -ENOTSUP;
> > > >      }
> > > > +    if (!s->caching_mode && (new & IOMMU_NOTIFIER_MAP)) {
> > > > +        error_setg_errno(errp, ENOTSUP,
> > > > +                         "device %02x.%02x.%x requires caching mode",
> > > > +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> > > > +                         PCI_FUNC(vtd_as->devfn));
> > > > +        return -ENOTSUP;
> > > > +    }
> > >
> > > We used to have that but got reverted because it's too late to fail, so we
> > > moved it over even though not as clean..
> > >
> > > https://lore.kernel.org/all/20190916080718.3299-5-peterx@redhat.com/
> >
> > One of the difference is that the patch doesn't do exit() here. I
> > think it's better to fail instead of misbehving silently, this is what
> > other vIOMMU did:
> >
> > E.g in smmu we had:
> >
> >     if (new & IOMMU_NOTIFIER_MAP) {
> >         error_setg(errp,
> >                    "device %02x.%02x.%x requires iommu MAP notifier which is "
> >                    "not currently supported", pci_bus_num(sdev->bus),
> >                    PCI_SLOT(sdev->devfn), PCI_FUNC(sdev->devfn));
> >         return -EINVAL;
> >     }
> >
> > So did for amd iommu.
>
> Yeah that's fine.  Would you mind add the explanation (and also above link,
> which might be helpful to pick up the history..) to the commit message?

Will do.

> With that:
>
> Reviewed-by: Peter Xu <peterx@redhat.com>
>
> Thanks,

Thanks

>
> --
> Peter Xu
>

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Peter Xu]

On Thu, Dec 01, 2022 at 04:35:48PM +0800, Jason Wang wrote:
> On Wed, Nov 30, 2022 at 11:17 PM Peter Xu <peterx@redhat.com> wrote:
> >
> > On Wed, Nov 30, 2022 at 02:33:51PM +0800, Jason Wang wrote:
> > > On Tue, Nov 29, 2022 at 11:57 PM Peter Xu <peterx@redhat.com> wrote:
> > > >
> > > > On Tue, Nov 29, 2022 at 04:10:37PM +0800, Jason Wang wrote:
> > > > > The IOVA tree is only built during page walk this breaks the device
> > > > > that tries to use UNMAP notifier only. One example is vhost-net, it
> > > > > tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
> > > > > notifier (e.g when dt mode is not enabled). The interesting part is
> > > > > that it doesn't use MAP since it can query the IOMMU translation by
> > > > > itself upon a IOTLB miss.
> > > > >
> > > > > This doesn't work since Qemu doesn't build IOVA tree in IOMMU
> > > > > translation which means the UNMAP notifier won't be triggered during
> > > > > the page walk since Qemu think it is never mapped. This could be
> > > > > noticed when vIOMMU is used with vhost_net but dt is disabled.
> > > > >
> > > > > Fixing this by build the iova tree during IOMMU translation, this
> > > > > makes sure the UNMAP notifier event could be identified during page
> > > > > walk. And we need to walk page table not only for UNMAP notifier but
> > > > > for MAP notifier during PSI.
> > > > >
> > > > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > > > ---
> > > > >  hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
> > > > >  1 file changed, 18 insertions(+), 25 deletions(-)
> > > > >
> > > > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > > > index d025ef2873..edeb62f4b2 100644
> > > > > --- a/hw/i386/intel_iommu.c
> > > > > +++ b/hw/i386/intel_iommu.c
> > > > > @@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
> > > > >      uint8_t access_flags;
> > > > >      bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
> > > > >      VTDIOTLBEntry *iotlb_entry;
> > > > > +    const DMAMap *mapped;
> > > > > +    DMAMap target;
> > > > >
> > > > >      /*
> > > > >       * We have standalone memory region for interrupt addresses, we
> > > > > @@ -1954,6 +1956,21 @@ out:
> > > > >      entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
> > > > >      entry->addr_mask = ~page_mask;
> > > > >      entry->perm = access_flags;
> > > > > +
> > > > > +    target.iova = entry->iova;
> > > > > +    target.size = entry->addr_mask;
> > > > > +    target.translated_addr = entry->translated_addr;
> > > > > +    target.perm = entry->perm;
> > > > > +
> > > > > +    mapped = iova_tree_find(vtd_as->iova_tree, &target);
> > > > > +    if (!mapped) {
> > > > > +        /* To make UNMAP notifier work, we need build iova tree here
> > > > > +         * in order to have the UNMAP iommu notifier to be triggered
> > > > > +         * during the page walk.
> > > > > +         */
> > > > > +        iova_tree_insert(vtd_as->iova_tree, &target);
> > > > > +    }
> > > > > +
> > > > >      return true;
> > > > >
> > > > >  error:
> > > > > @@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
> > > > >          ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
> > > > >                                         vtd_as->devfn, &ce);
> > > > >          if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
> > > > > -            if (vtd_as_has_map_notifier(vtd_as)) {
> > > > > -                /*
> > > > > -                 * As long as we have MAP notifications registered in
> > > > > -                 * any of our IOMMU notifiers, we need to sync the
> > > > > -                 * shadow page table.
> > > > > -                 */
> > > > > -                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> > > > > -            } else {
> > > > > -                /*
> > > > > -                 * For UNMAP-only notifiers, we don't need to walk the
> > > > > -                 * page tables.  We just deliver the PSI down to
> > > > > -                 * invalidate caches.
> > > > > -                 */
> > > > > -                IOMMUTLBEvent event = {
> > > > > -                    .type = IOMMU_NOTIFIER_UNMAP,
> > > > > -                    .entry = {
> > > > > -                        .target_as = &address_space_memory,
> > > > > -                        .iova = addr,
> > > > > -                        .translated_addr = 0,
> > > > > -                        .addr_mask = size - 1,
> > > > > -                        .perm = IOMMU_NONE,
> > > > > -                    },
> > > > > -                };
> > > > > -                memory_region_notify_iommu(&vtd_as->iommu, 0, event);

[1]

> > > >
> > > > Isn't this path the one that will be responsible for pass-through the UNMAP
> > > > events from guest to vhost when there's no MAP notifier requested?
> > >
> > > Yes, but it doesn't do the iova tree removing. More below.
> > >
> > > >
> > > > At least that's what I expected when introducing the iova tree, because for
> > > > unmap-only device hierachy I thought we didn't need the tree at all.
> > >
> > > Then the problem is the UNMAP notifier won't be trigger at all during
> > > DSI page walk in vtd_page_walk_one() because there's no DMAMap stored
> > > in the iova tree.:
> > >
> > >         if (!mapped) {
> > >             /* Skip since we didn't map this range at all */
> > >             trace_vtd_page_walk_one_skip_unmap(entry->iova, entry->addr_mask);
> > >             return 0;
> > >         }
> > >
> > > So I choose to build the iova tree in translate then we won't go
> > > within the above condition.
> >
> > That's also why it's weird because IIUC we should never walk a page table
> > at all if there's no MAP notifier regiestered.
> 
> If this is true, we probably need to document this somewhere.

Agree.  I'll post a patch.

> 
> >
> > When I'm looking at the walk callers I found that indeed there's one path
> > missing where can cause it to actually walk the pgtables without !MAP, then
> > I also noticed commit f7701e2c7983b6, and I'm wondering what we really want
> > is something like this:
> >
> > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > index a08ee85edf..c46f3db992 100644
> > --- a/hw/i386/intel_iommu.c
> > +++ b/hw/i386/intel_iommu.c
> > @@ -1536,7 +1536,7 @@ static int vtd_sync_shadow_page_table(VTDAddressSpace *vtd_as)
> >      VTDContextEntry ce;
> >      IOMMUNotifier *n;
> >
> > -    if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
> > +    if (!vtd_as_has_map_notifier(vtd_as)) {
> >          return 0;
> >      }
> >
> > So I'm not sure whether this patch is the problem resolver; so far I feel
> > like it's patch 2 who does the real fix.  Then we can have the above
> > oneliner so we stop any walks when there's no map notifiers.
> >
> > Thanks,
> 
> I may miss something but as state above, the problem is a missing
> UNMAP notification during DSI when there's only UNMAP notifier.

I got confused too on why we didn't notify UNMAP for DSI already, that's so
weird because I thought it should be there or it should be broken for a
long time.. as we discussed multiple times around this one:

		/*
		 * Fallback to domain selective flush if no PSI support or
		 * the size is too big.
		 */
		if (!cap_pgsel_inv(iommu->cap) ||
		    mask > cap_max_amask_val(iommu->cap))
			iommu->flush.flush_iotlb(iommu, did, 0, 0,
							DMA_TLB_DSI_FLUSH);
		else
			iommu->flush.flush_iotlb(iommu, did, addr | ih, mask,
							DMA_TLB_PSI_FLUSH);

I guess we were just always lucky?..

> 
> To solve it we might have two ways:
> 
> 1) build the iova tree during iommu translation then we can correctly
> trigger UNMAP during page walk caused by DSI
> 2) don't do the iova tree walk for !MAP notifier, need new logic to
> trigger UNMAP notifier in PSI/DSI
> 
> This patch choose to go 1) (which seems easier at least for -stable).
> Do you mean you prefer to go with 2)?

Yes.

IOVA tree is unnecessary overhead IMHO because UNMAP (both IOTLB or
DEVIOTLB unmap) shouldn't need that complexity at all.  Using the iova tree
can be accurate on which page got unmapped when the kernel driver used DSI
for a large PSI as shown above, however IMHO it needs more justification
that the pgtable walk is worth the effort.  Not to mention if a device in
QEMU that wants to use the iova tree for some reason, one can just register
with MAP and ignore all MAP events, while by default we keep UNMAP simple.

So we could do:

  (1) Rename vtd_sync_shadow_page_table() to vtd_sync_domain()
  (2) instead of optimizing dev-iotlb only there:
  
      if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
          return 0;
      }

      we should firstly check if UNMAP or DEVUNMAP registered, we directly
      send a notification to the whole domain.  We need to choose the event
      that the register happens with but not both.

This also reminded me that whether we should sanity check on iommu
notifiers on some invalid cases.  E.g. it seems to me when registered with
DEVIOTLB_UNMAP it should not register with either MAP or UNMAP anymore or
it doesn't make sense.

One step further, I'm wondering whether the DEV_IOTLB event should exist at
all.  Maybe we want to have DEV_IOTLB typed iommu notifier only, but then
when we got dev-iotlb PSI/DSI we notify with type=UNMAP just like normal
UNMAP events, then in above (2) we can send UNMAP constantly as long as
!MAP.  But even if so that'll just be another optional change on top.

Thanks,

-- 
Peter Xu

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Peter Xu]

On Tue, Nov 29, 2022 at 04:10:36PM +0800, Jason Wang wrote:
> Without dt mode, device IOTLB notifier won't work since guest won't
> send device IOTLB invalidation descriptor in this case. Let's fail
> early instead of misbehaving silently.
> 
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
>  hw/i386/intel_iommu.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index 9143376677..d025ef2873 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>  {
>      VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
>      IntelIOMMUState *s = vtd_as->iommu_state;
> +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
>  
>      /* TODO: add support for VFIO and vhost users */
>      if (s->snoop_control) {
> @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>                           PCI_FUNC(vtd_as->devfn));
>          return -ENOTSUP;
>      }
> +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> +        error_setg_errno(errp, ENOTSUP,
> +                         "device %02x.%02x.%x requires device IOTLB mode",
> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> +                         PCI_FUNC(vtd_as->devfn));
> +        return -ENOTSUP;
> +    }

While my r-b holds.. let's also do this for amd-iommu in the same patch?
dt never supported there, so we can fail as long as DEVIOTLB registered.

>  
>      /* Update per-address-space notifier flags */
>      vtd_as->notifier_flags = new;
> -- 
> 2.25.1
> 

-- 
Peter Xu

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Jason Wang]

`


On Thu, Dec 1, 2022 at 10:59 PM Peter Xu <peterx@redhat.com> wrote:
>
> On Thu, Dec 01, 2022 at 04:35:48PM +0800, Jason Wang wrote:
> > On Wed, Nov 30, 2022 at 11:17 PM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > On Wed, Nov 30, 2022 at 02:33:51PM +0800, Jason Wang wrote:
> > > > On Tue, Nov 29, 2022 at 11:57 PM Peter Xu <peterx@redhat.com> wrote:
> > > > >
> > > > > On Tue, Nov 29, 2022 at 04:10:37PM +0800, Jason Wang wrote:
> > > > > > The IOVA tree is only built during page walk this breaks the device
> > > > > > that tries to use UNMAP notifier only. One example is vhost-net, it
> > > > > > tries to use UNMAP notifier when vIOMMU doesn't support DEVIOTLB_UNMAP
> > > > > > notifier (e.g when dt mode is not enabled). The interesting part is
> > > > > > that it doesn't use MAP since it can query the IOMMU translation by
> > > > > > itself upon a IOTLB miss.
> > > > > >
> > > > > > This doesn't work since Qemu doesn't build IOVA tree in IOMMU
> > > > > > translation which means the UNMAP notifier won't be triggered during
> > > > > > the page walk since Qemu think it is never mapped. This could be
> > > > > > noticed when vIOMMU is used with vhost_net but dt is disabled.
> > > > > >
> > > > > > Fixing this by build the iova tree during IOMMU translation, this
> > > > > > makes sure the UNMAP notifier event could be identified during page
> > > > > > walk. And we need to walk page table not only for UNMAP notifier but
> > > > > > for MAP notifier during PSI.
> > > > > >
> > > > > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > > > > ---
> > > > > >  hw/i386/intel_iommu.c | 43 ++++++++++++++++++-------------------------
> > > > > >  1 file changed, 18 insertions(+), 25 deletions(-)
> > > > > >
> > > > > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > > > > index d025ef2873..edeb62f4b2 100644
> > > > > > --- a/hw/i386/intel_iommu.c
> > > > > > +++ b/hw/i386/intel_iommu.c
> > > > > > @@ -1834,6 +1834,8 @@ static bool vtd_do_iommu_translate(VTDAddressSpace *vtd_as, PCIBus *bus,
> > > > > >      uint8_t access_flags;
> > > > > >      bool rid2pasid = (pasid == PCI_NO_PASID) && s->root_scalable;
> > > > > >      VTDIOTLBEntry *iotlb_entry;
> > > > > > +    const DMAMap *mapped;
> > > > > > +    DMAMap target;
> > > > > >
> > > > > >      /*
> > > > > >       * We have standalone memory region for interrupt addresses, we
> > > > > > @@ -1954,6 +1956,21 @@ out:
> > > > > >      entry->translated_addr = vtd_get_slpte_addr(slpte, s->aw_bits) & page_mask;
> > > > > >      entry->addr_mask = ~page_mask;
> > > > > >      entry->perm = access_flags;
> > > > > > +
> > > > > > +    target.iova = entry->iova;
> > > > > > +    target.size = entry->addr_mask;
> > > > > > +    target.translated_addr = entry->translated_addr;
> > > > > > +    target.perm = entry->perm;
> > > > > > +
> > > > > > +    mapped = iova_tree_find(vtd_as->iova_tree, &target);
> > > > > > +    if (!mapped) {
> > > > > > +        /* To make UNMAP notifier work, we need build iova tree here
> > > > > > +         * in order to have the UNMAP iommu notifier to be triggered
> > > > > > +         * during the page walk.
> > > > > > +         */
> > > > > > +        iova_tree_insert(vtd_as->iova_tree, &target);
> > > > > > +    }
> > > > > > +
> > > > > >      return true;
> > > > > >
> > > > > >  error:
> > > > > > @@ -2161,31 +2178,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
> > > > > >          ret = vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
> > > > > >                                         vtd_as->devfn, &ce);
> > > > > >          if (!ret && domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
> > > > > > -            if (vtd_as_has_map_notifier(vtd_as)) {
> > > > > > -                /*
> > > > > > -                 * As long as we have MAP notifications registered in
> > > > > > -                 * any of our IOMMU notifiers, we need to sync the
> > > > > > -                 * shadow page table.
> > > > > > -                 */
> > > > > > -                vtd_sync_shadow_page_table_range(vtd_as, &ce, addr, size);
> > > > > > -            } else {
> > > > > > -                /*
> > > > > > -                 * For UNMAP-only notifiers, we don't need to walk the
> > > > > > -                 * page tables.  We just deliver the PSI down to
> > > > > > -                 * invalidate caches.
> > > > > > -                 */
> > > > > > -                IOMMUTLBEvent event = {
> > > > > > -                    .type = IOMMU_NOTIFIER_UNMAP,
> > > > > > -                    .entry = {
> > > > > > -                        .target_as = &address_space_memory,
> > > > > > -                        .iova = addr,
> > > > > > -                        .translated_addr = 0,
> > > > > > -                        .addr_mask = size - 1,
> > > > > > -                        .perm = IOMMU_NONE,
> > > > > > -                    },
> > > > > > -                };
> > > > > > -                memory_region_notify_iommu(&vtd_as->iommu, 0, event);
>
> [1]
>
> > > > >
> > > > > Isn't this path the one that will be responsible for pass-through the UNMAP
> > > > > events from guest to vhost when there's no MAP notifier requested?
> > > >
> > > > Yes, but it doesn't do the iova tree removing. More below.
> > > >
> > > > >
> > > > > At least that's what I expected when introducing the iova tree, because for
> > > > > unmap-only device hierachy I thought we didn't need the tree at all.
> > > >
> > > > Then the problem is the UNMAP notifier won't be trigger at all during
> > > > DSI page walk in vtd_page_walk_one() because there's no DMAMap stored
> > > > in the iova tree.:
> > > >
> > > >         if (!mapped) {
> > > >             /* Skip since we didn't map this range at all */
> > > >             trace_vtd_page_walk_one_skip_unmap(entry->iova, entry->addr_mask);
> > > >             return 0;
> > > >         }
> > > >
> > > > So I choose to build the iova tree in translate then we won't go
> > > > within the above condition.
> > >
> > > That's also why it's weird because IIUC we should never walk a page table
> > > at all if there's no MAP notifier regiestered.
> >
> > If this is true, we probably need to document this somewhere.
>
> Agree.  I'll post a patch.
>
> >
> > >
> > > When I'm looking at the walk callers I found that indeed there's one path
> > > missing where can cause it to actually walk the pgtables without !MAP, then
> > > I also noticed commit f7701e2c7983b6, and I'm wondering what we really want
> > > is something like this:
> > >
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index a08ee85edf..c46f3db992 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -1536,7 +1536,7 @@ static int vtd_sync_shadow_page_table(VTDAddressSpace *vtd_as)
> > >      VTDContextEntry ce;
> > >      IOMMUNotifier *n;
> > >
> > > -    if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
> > > +    if (!vtd_as_has_map_notifier(vtd_as)) {
> > >          return 0;
> > >      }
> > >
> > > So I'm not sure whether this patch is the problem resolver; so far I feel
> > > like it's patch 2 who does the real fix.  Then we can have the above
> > > oneliner so we stop any walks when there's no map notifiers.
> > >
> > > Thanks,
> >
> > I may miss something but as state above, the problem is a missing
> > UNMAP notification during DSI when there's only UNMAP notifier.
>
> I got confused too on why we didn't notify UNMAP for DSI already, that's so
> weird because I thought it should be there or it should be broken for a
> long time.. as we discussed multiple times around this one:
>
>                 /*
>                  * Fallback to domain selective flush if no PSI support or
>                  * the size is too big.
>                  */
>                 if (!cap_pgsel_inv(iommu->cap) ||
>                     mask > cap_max_amask_val(iommu->cap))
>                         iommu->flush.flush_iotlb(iommu, did, 0, 0,
>                                                         DMA_TLB_DSI_FLUSH);
>                 else
>                         iommu->flush.flush_iotlb(iommu, did, addr | ih, mask,
>                                                         DMA_TLB_PSI_FLUSH);
>
> I guess we were just always lucky?..

Probably, or the reason is that we a notifier with UNMAP only is not
commonly used before until patch 2.

>
> >
> > To solve it we might have two ways:
> >
> > 1) build the iova tree during iommu translation then we can correctly
> > trigger UNMAP during page walk caused by DSI
> > 2) don't do the iova tree walk for !MAP notifier, need new logic to
> > trigger UNMAP notifier in PSI/DSI
> >
> > This patch choose to go 1) (which seems easier at least for -stable).
> > Do you mean you prefer to go with 2)?
>
> Yes.
>
> IOVA tree is unnecessary overhead IMHO because UNMAP (both IOTLB or
> DEVIOTLB unmap) shouldn't need that complexity at all.  Using the iova tree
> can be accurate on which page got unmapped when the kernel driver used DSI
> for a large PSI as shown above, however IMHO it needs more justification
> that the pgtable walk is worth the effort.  Not to mention if a device in
> QEMU that wants to use the iova tree for some reason, one can just register
> with MAP and ignore all MAP events, while by default we keep UNMAP simple.
>
> So we could do:
>
>   (1) Rename vtd_sync_shadow_page_table() to vtd_sync_domain()
>   (2) instead of optimizing dev-iotlb only there:
>
>       if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
>           return 0;
>       }
>
>       we should firstly check if UNMAP or DEVUNMAP registered, we directly
>       send a notification to the whole domain.  We need to choose the event
>       that the register happens with but not both.
>
> This also reminded me that whether we should sanity check on iommu
> notifiers on some invalid cases.  E.g. it seems to me when registered with
> DEVIOTLB_UNMAP it should not register with either MAP or UNMAP anymore or
> it doesn't make sense.

It seems it doesn't' harm to allow both UNMAP and DEVIOTLB_UNMAP work.

>
> One step further, I'm wondering whether the DEV_IOTLB event should exist at
> all.  Maybe we want to have DEV_IOTLB typed iommu notifier only,

This seems cleaner ( I remember we had some discussion before ).

> but then
> when we got dev-iotlb PSI/DSI we notify with type=UNMAP just like normal
> UNMAP events, then in above (2) we can send UNMAP constantly as long as
> !MAP.  But even if so that'll just be another optional change on top.

I'm fine to go without iova-tree. Would you mind to post patches for
fix? I can test and include it in this series then.

Thanks

>
> Thanks,
>
> --
> Peter Xu
>

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Peter Xu]

[-- Attachment #1: Type: text/plain, Size: 562 bytes --]

Jason,

On Mon, Dec 05, 2022 at 12:12:04PM +0800, Jason Wang wrote:
> I'm fine to go without iova-tree. Would you mind to post patches for
> fix? I can test and include it in this series then.

One sample patch attached, only compile tested.

I can also work on this but I'll be slow in making progress, so I'll add it
into my todo.  If you can help to fix this issue it'll be more than great.
No worry on the ownership or authorship of the patch if you agree on the
change and moving forward with this when modifying - just take it over!

Thanks!

-- 
Peter Xu

[-- Attachment #2: 0001-memory-sanity-check-flatview-deref-on-mr-transaction.patch --]
[-- Type: text/plain, Size: 1572 bytes --]

From 57e5cab805c94d56f801a7e21098389a77584e34 Mon Sep 17 00:00:00 2001
From: Peter Xu <peterx@redhat.com>
Date: Mon, 5 Dec 2022 11:14:02 -0500
Subject: [PATCH] memory: sanity check flatview deref on mr transactions
Content-type: text/plain

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 include/exec/memory.h | 9 +++++++++
 softmmu/memory.c      | 2 +-
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/include/exec/memory.h b/include/exec/memory.h
index 91f8a2395a..e136ab9558 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -1069,8 +1069,17 @@ struct FlatView {
     MemoryRegion *root;
 };
 
+extern unsigned memory_region_transaction_depth;
+
 static inline FlatView *address_space_to_flatview(AddressSpace *as)
 {
+    /*
+     * Before using any flatview, sanity check we're not during a memory
+     * region transaction or the map can be invalid.  Note that this can
+     * also be called during commit phase of memory transaction, but that
+     * should also only happen when the depth decreases to 0 first.
+     */
+    assert(memory_region_transaction_depth == 0);
     return qatomic_rcu_read(&as->current_map);
 }
 
diff --git a/softmmu/memory.c b/softmmu/memory.c
index bc0be3f62c..7cfcf5dffe 100644
--- a/softmmu/memory.c
+++ b/softmmu/memory.c
@@ -37,7 +37,7 @@
 
 //#define DEBUG_UNASSIGNED
 
-static unsigned memory_region_transaction_depth;
+unsigned memory_region_transaction_depth;
 static bool memory_region_update_pending;
 static bool ioeventfd_update_pending;
 unsigned int global_dirty_tracking;
-- 
2.37.3

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Jason Wang]

On Tue, Dec 6, 2022 at 7:19 AM Peter Xu <peterx@redhat.com> wrote:
>
> Jason,
>
> On Mon, Dec 05, 2022 at 12:12:04PM +0800, Jason Wang wrote:
> > I'm fine to go without iova-tree. Would you mind to post patches for
> > fix? I can test and include it in this series then.
>
> One sample patch attached, only compile tested.

I don't see any direct connection between the attached patch and the
intel-iommu?

>
> I can also work on this but I'll be slow in making progress, so I'll add it
> into my todo.  If you can help to fix this issue it'll be more than great.

Ok, let me try but it might take some time :)

> No worry on the ownership or authorship of the patch if you agree on the
> change and moving forward with this when modifying - just take it over!

Ok.

Thanks

>
> Thanks!
>
> --
> Peter Xu

Re: [PATCH 1/3] intel-iommu: fail MAP notifier without caching mode

[Eric Auger]

Hi Jason,

On 11/29/22 09:10, Jason Wang wrote:
> Without caching mode, MAP notifier won't work correctly since guest
> won't send IOTLB update event when it establishes new mappings in the
> I/O page tables. Let's fail the IOMMU notifiers early instead of
> misbehaving silently.
>
> Signed-off-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>

Thanks

Eric
> ---
>  hw/i386/intel_iommu.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index a08ee85edf..9143376677 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -3186,6 +3186,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>                           "Snoop Control with vhost or VFIO is not supported");
>          return -ENOTSUP;
>      }
> +    if (!s->caching_mode && (new & IOMMU_NOTIFIER_MAP)) {
> +        error_setg_errno(errp, ENOTSUP,
> +                         "device %02x.%02x.%x requires caching mode",
> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> +                         PCI_FUNC(vtd_as->devfn));
> +        return -ENOTSUP;
> +    }
>  
>      /* Update per-address-space notifier flags */
>      vtd_as->notifier_flags = new;

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Eric Auger]

Hi jason,

On 11/29/22 09:10, Jason Wang wrote:
> Without dt mode, device IOTLB notifier won't work since guest won't
> send device IOTLB invalidation descriptor in this case. Let's fail
> early instead of misbehaving silently.
>
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
>  hw/i386/intel_iommu.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index 9143376677..d025ef2873 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>  {
>      VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
>      IntelIOMMUState *s = vtd_as->iommu_state;
> +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
>  
>      /* TODO: add support for VFIO and vhost users */
>      if (s->snoop_control) {
> @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>                           PCI_FUNC(vtd_as->devfn));
>          return -ENOTSUP;
>      }
> +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> +        error_setg_errno(errp, ENOTSUP,
> +                         "device %02x.%02x.%x requires device IOTLB mode",
maybe precise INTEL IOMMU device-IOTLB mode. otherwise this may be
confused with device ATS capability?

While thinking about those error handlings (including the SMMU ones)
nothing should really prevent you from registering a notifier that is
not signalled. Maybe we should add in the documentation that any attempt
to register an IOMMU notifier to an IOMMU MR that is not able to signal
it will return an error.

Besides
Reviewed-by: Eric Auger <eric.auger@redhat.com>

Eric
> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> +                         PCI_FUNC(vtd_as->devfn));
> +        return -ENOTSUP;
> +    }
>  
>      /* Update per-address-space notifier flags */
>      vtd_as->notifier_flags = new;

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Peter Xu]

[-- Attachment #1: Type: text/plain, Size: 878 bytes --]

On Tue, Dec 06, 2022 at 11:18:03AM +0800, Jason Wang wrote:
> On Tue, Dec 6, 2022 at 7:19 AM Peter Xu <peterx@redhat.com> wrote:
> >
> > Jason,
> >
> > On Mon, Dec 05, 2022 at 12:12:04PM +0800, Jason Wang wrote:
> > > I'm fine to go without iova-tree. Would you mind to post patches for
> > > fix? I can test and include it in this series then.
> >
> > One sample patch attached, only compile tested.
> 
> I don't see any direct connection between the attached patch and the
> intel-iommu?

Sorry!  Wrong tree dumped...  Trying again.

> 
> >
> > I can also work on this but I'll be slow in making progress, so I'll add it
> > into my todo.  If you can help to fix this issue it'll be more than great.
> 
> Ok, let me try but it might take some time :)

Sure. :)

I'll also add it into my todo (and I think the other similar one has been
there for a while.. :( ).

-- 
Peter Xu

[-- Attachment #2: 0001-intel-iommu-Send-unmap-notifications-for-domain-or-g.patch --]
[-- Type: text/plain, Size: 3728 bytes --]

From 37c743761d20c16891856c5bef2e7b3fb89893b6 Mon Sep 17 00:00:00 2001
From: Peter Xu <peterx@redhat.com>
Date: Mon, 5 Dec 2022 18:11:36 -0500
Subject: [PATCH] intel-iommu: Send unmap notifications for domain or global
 inv desc
Content-type: text/plain

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 hw/i386/intel_iommu.c | 41 +++++++++++++++++++++++++----------------
 1 file changed, 25 insertions(+), 16 deletions(-)

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index a08ee85edf..2c6ca68df0 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -206,6 +206,23 @@ static inline gboolean vtd_as_has_map_notifier(VTDAddressSpace *as)
     return as->notifier_flags & IOMMU_NOTIFIER_MAP;
 }
 
+static void vtd_as_notify_unmap(VTDAddressSpace *as, hwaddr iova,
+                                hwaddr addr_mask)
+{
+    IOMMUTLBEvent event = {
+        .type = IOMMU_NOTIFIER_UNMAP,
+        .entry = {
+            .target_as = &address_space_memory,
+            .iova = iova,
+            .translated_addr = 0,
+            .addr_mask = addr_mask,
+            .perm = IOMMU_NONE,
+        },
+    };
+
+    memory_region_notify_iommu(&as->iommu, 0, event);
+}
+
 /* GHashTable functions */
 static gboolean vtd_iotlb_equal(gconstpointer v1, gconstpointer v2)
 {
@@ -1530,13 +1547,15 @@ static int vtd_sync_shadow_page_table_range(VTDAddressSpace *vtd_as,
     return vtd_page_walk(s, ce, addr, addr + size, &info, vtd_as->pasid);
 }
 
-static int vtd_sync_shadow_page_table(VTDAddressSpace *vtd_as)
+static int vtd_address_space_sync(VTDAddressSpace *vtd_as)
 {
     int ret;
     VTDContextEntry ce;
     IOMMUNotifier *n;
 
-    if (!(vtd_as->iommu.iommu_notify_flags & IOMMU_NOTIFIER_IOTLB_EVENTS)) {
+    /* If no MAP notifier registered, we simply invalidate all the cache */
+    if (!vtd_as_has_map_notifier(vtd_as)) {
+        vtd_as_notify_unmap(vtd_as, 0, HWADDR_MAX);
         return 0;
     }
 
@@ -2000,7 +2019,7 @@ static void vtd_iommu_replay_all(IntelIOMMUState *s)
     VTDAddressSpace *vtd_as;
 
     QLIST_FOREACH(vtd_as, &s->vtd_as_with_notifiers, next) {
-        vtd_sync_shadow_page_table(vtd_as);
+        vtd_address_space_sync(vtd_as);
     }
 }
 
@@ -2082,7 +2101,7 @@ static void vtd_context_device_invalidate(IntelIOMMUState *s,
              * framework will skip MAP notifications if that
              * happened.
              */
-            vtd_sync_shadow_page_table(vtd_as);
+            vtd_address_space_sync(vtd_as);
         }
     }
 }
@@ -2140,7 +2159,7 @@ static void vtd_iotlb_domain_invalidate(IntelIOMMUState *s, uint16_t domain_id)
         if (!vtd_dev_to_context_entry(s, pci_bus_num(vtd_as->bus),
                                       vtd_as->devfn, &ce) &&
             domain_id == vtd_get_domain_id(s, &ce, vtd_as->pasid)) {
-            vtd_sync_shadow_page_table(vtd_as);
+            vtd_address_space_sync(vtd_as);
         }
     }
 }
@@ -2174,17 +2193,7 @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s,
                  * page tables.  We just deliver the PSI down to
                  * invalidate caches.
                  */
-                IOMMUTLBEvent event = {
-                    .type = IOMMU_NOTIFIER_UNMAP,
-                    .entry = {
-                        .target_as = &address_space_memory,
-                        .iova = addr,
-                        .translated_addr = 0,
-                        .addr_mask = size - 1,
-                        .perm = IOMMU_NONE,
-                    },
-                };
-                memory_region_notify_iommu(&vtd_as->iommu, 0, event);
+                vtd_as_notify_unmap(vtd_as, addr, size - 1);
             }
         }
     }
-- 
2.37.3

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Michael S. Tsirkin]

On Tue, Nov 29, 2022 at 04:10:34PM +0800, Jason Wang wrote:
> Hi All:
> 
> According to ATS, device should work if ATS is disabled. This is not
> correctly implemented in the current intel-iommu since it doesn't
> handle the UNMAP notifier correctly. This breaks the vhost-net +
> vIOMMU without dt.
> 
> The root casue is that the when there's a device IOTLB miss (note that
> it's not specific to PCI so it can work without ATS), Qemu doesn't
> build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> won't trigger the UNMAP notifier.
> 
> Fixing by build IOVA tree during IOMMU translsation.
> 
> Thanks


IIUC you were going to post v2? At least commit log fixes.

> Jason Wang (3):
>   intel-iommu: fail MAP notifier without caching mode
>   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
>   intel-iommu: build iova tree during IOMMU translation
> 
>  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
>  1 file changed, 33 insertions(+), 25 deletions(-)
> 
> -- 
> 2.25.1

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Jason Wang]

On Tue, Dec 20, 2022 at 9:53 PM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Tue, Nov 29, 2022 at 04:10:34PM +0800, Jason Wang wrote:
> > Hi All:
> >
> > According to ATS, device should work if ATS is disabled. This is not
> > correctly implemented in the current intel-iommu since it doesn't
> > handle the UNMAP notifier correctly. This breaks the vhost-net +
> > vIOMMU without dt.
> >
> > The root casue is that the when there's a device IOTLB miss (note that
> > it's not specific to PCI so it can work without ATS), Qemu doesn't
> > build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> > won't trigger the UNMAP notifier.
> >
> > Fixing by build IOVA tree during IOMMU translsation.
> >
> > Thanks
>
>
> IIUC you were going to post v2? At least commit log fixes.

Yes.

Thanks

>
> > Jason Wang (3):
> >   intel-iommu: fail MAP notifier without caching mode
> >   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
> >   intel-iommu: build iova tree during IOMMU translation
> >
> >  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
> >  1 file changed, 33 insertions(+), 25 deletions(-)
> >
> > --
> > 2.25.1
>

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Jason Wang]

On Tue, Dec 6, 2022 at 9:58 PM Peter Xu <peterx@redhat.com> wrote:
>
> On Tue, Dec 06, 2022 at 11:18:03AM +0800, Jason Wang wrote:
> > On Tue, Dec 6, 2022 at 7:19 AM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > Jason,
> > >
> > > On Mon, Dec 05, 2022 at 12:12:04PM +0800, Jason Wang wrote:
> > > > I'm fine to go without iova-tree. Would you mind to post patches for
> > > > fix? I can test and include it in this series then.
> > >
> > > One sample patch attached, only compile tested.
> >
> > I don't see any direct connection between the attached patch and the
> > intel-iommu?
>
> Sorry!  Wrong tree dumped...  Trying again.

The HWADDR breaks memory_region_notify_iommu_one():

qemu-system-x86_64: ../softmmu/memory.c:1991:
memory_region_notify_iommu_one: Assertion `entry->iova >=
notifier->start && entry_end <= notifier->end' failed.

I wonder if we need either:

1) remove the assert

or

2) introduce a new memory_region_notify_unmap_all() to unmap from
notifier->start to notifier->end.

Thanks

>
> >
> > >
> > > I can also work on this but I'll be slow in making progress, so I'll add it
> > > into my todo.  If you can help to fix this issue it'll be more than great.
> >
> > Ok, let me try but it might take some time :)
>
> Sure. :)
>
> I'll also add it into my todo (and I think the other similar one has been
> there for a while.. :( ).
>
> --
> Peter Xu

Re: [PATCH 3/3] intel-iommu: build iova tree during IOMMU translation

[Peter Xu]

On Fri, Dec 23, 2022 at 04:02:29PM +0800, Jason Wang wrote:
> On Tue, Dec 6, 2022 at 9:58 PM Peter Xu <peterx@redhat.com> wrote:
> >
> > On Tue, Dec 06, 2022 at 11:18:03AM +0800, Jason Wang wrote:
> > > On Tue, Dec 6, 2022 at 7:19 AM Peter Xu <peterx@redhat.com> wrote:
> > > >
> > > > Jason,
> > > >
> > > > On Mon, Dec 05, 2022 at 12:12:04PM +0800, Jason Wang wrote:
> > > > > I'm fine to go without iova-tree. Would you mind to post patches for
> > > > > fix? I can test and include it in this series then.
> > > >
> > > > One sample patch attached, only compile tested.
> > >
> > > I don't see any direct connection between the attached patch and the
> > > intel-iommu?
> >
> > Sorry!  Wrong tree dumped...  Trying again.
> 
> The HWADDR breaks memory_region_notify_iommu_one():
> 
> qemu-system-x86_64: ../softmmu/memory.c:1991:
> memory_region_notify_iommu_one: Assertion `entry->iova >=
> notifier->start && entry_end <= notifier->end' failed.
> 
> I wonder if we need either:
> 
> 1) remove the assert

I vote for this one.  Not only removing the assertion, we should probably
crop the range too just like dev-iotlb unmaps?

Thanks,

> 
> or
> 
> 2) introduce a new memory_region_notify_unmap_all() to unmap from
> notifier->start to notifier->end.

-- 
Peter Xu

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Viktor Prutyanov]

On Tue, Nov 29, 2022 at 11:10 AM Jason Wang <jasowang@redhat.com> wrote:
>
> Hi All:
>
> According to ATS, device should work if ATS is disabled. This is not
> correctly implemented in the current intel-iommu since it doesn't
> handle the UNMAP notifier correctly. This breaks the vhost-net +
> vIOMMU without dt.
>
> The root casue is that the when there's a device IOTLB miss (note that
> it's not specific to PCI so it can work without ATS), Qemu doesn't
> build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> won't trigger the UNMAP notifier.
>
> Fixing by build IOVA tree during IOMMU translsation.
>
> Thanks
>
> Jason Wang (3):
>   intel-iommu: fail MAP notifier without caching mode
>   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
>   intel-iommu: build iova tree during IOMMU translation
>
>  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
>  1 file changed, 33 insertions(+), 25 deletions(-)
>
> --
> 2.25.1
>

Hi Jason,

I've tried the series with Windows Server 2022 guest with vhost and
intel-iommu (device-iotlb=off) and now networking on this system has
become working.
So, as we discussed, I'm waiting for the series to be accepted in some
form to continue my work about supporting guests who refuse Device-TLB
on systems with device-iotlb=on.

Tested-by: Viktor Prutyanov <viktor@daynix.com>

Best regards,
Viktor Prutyanov

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Jason Wang]

On Mon, Jan 16, 2023 at 7:30 AM Viktor Prutyanov <viktor@daynix.com> wrote:
>
> On Tue, Nov 29, 2022 at 11:10 AM Jason Wang <jasowang@redhat.com> wrote:
> >
> > Hi All:
> >
> > According to ATS, device should work if ATS is disabled. This is not
> > correctly implemented in the current intel-iommu since it doesn't
> > handle the UNMAP notifier correctly. This breaks the vhost-net +
> > vIOMMU without dt.
> >
> > The root casue is that the when there's a device IOTLB miss (note that
> > it's not specific to PCI so it can work without ATS), Qemu doesn't
> > build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> > won't trigger the UNMAP notifier.
> >
> > Fixing by build IOVA tree during IOMMU translsation.
> >
> > Thanks
> >
> > Jason Wang (3):
> >   intel-iommu: fail MAP notifier without caching mode
> >   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
> >   intel-iommu: build iova tree during IOMMU translation
> >
> >  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
> >  1 file changed, 33 insertions(+), 25 deletions(-)
> >
> > --
> > 2.25.1
> >
>
> Hi Jason,
>
> I've tried the series with Windows Server 2022 guest with vhost and
> intel-iommu (device-iotlb=off) and now networking on this system has
> become working.
> So, as we discussed, I'm waiting for the series to be accepted in some
> form to continue my work about supporting guests who refuse Device-TLB
> on systems with device-iotlb=on.
>
> Tested-by: Viktor Prutyanov <viktor@daynix.com>

Great, Peter has some comments on this series, so I will probably send
a new version (probably after the chinese new year).

Thanks

>
> Best regards,
> Viktor Prutyanov
>

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Michael S. Tsirkin]

On Mon, Jan 16, 2023 at 03:06:44PM +0800, Jason Wang wrote:
> On Mon, Jan 16, 2023 at 7:30 AM Viktor Prutyanov <viktor@daynix.com> wrote:
> >
> > On Tue, Nov 29, 2022 at 11:10 AM Jason Wang <jasowang@redhat.com> wrote:
> > >
> > > Hi All:
> > >
> > > According to ATS, device should work if ATS is disabled. This is not
> > > correctly implemented in the current intel-iommu since it doesn't
> > > handle the UNMAP notifier correctly. This breaks the vhost-net +
> > > vIOMMU without dt.
> > >
> > > The root casue is that the when there's a device IOTLB miss (note that
> > > it's not specific to PCI so it can work without ATS), Qemu doesn't
> > > build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> > > won't trigger the UNMAP notifier.
> > >
> > > Fixing by build IOVA tree during IOMMU translsation.
> > >
> > > Thanks
> > >
> > > Jason Wang (3):
> > >   intel-iommu: fail MAP notifier without caching mode
> > >   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
> > >   intel-iommu: build iova tree during IOMMU translation
> > >
> > >  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
> > >  1 file changed, 33 insertions(+), 25 deletions(-)
> > >
> > > --
> > > 2.25.1
> > >
> >
> > Hi Jason,
> >
> > I've tried the series with Windows Server 2022 guest with vhost and
> > intel-iommu (device-iotlb=off) and now networking on this system has
> > become working.
> > So, as we discussed, I'm waiting for the series to be accepted in some
> > form to continue my work about supporting guests who refuse Device-TLB
> > on systems with device-iotlb=on.
> >
> > Tested-by: Viktor Prutyanov <viktor@daynix.com>
> 
> Great, Peter has some comments on this series, so I will probably send
> a new version (probably after the chinese new year).
> 
> Thanks

Were you going to post a new version?

> >
> > Best regards,
> > Viktor Prutyanov
> >

Re: [PATCH 0/3] Fix UNMAP notifier for intel-iommu

[Jason Wang]

On Fri, Jan 27, 2023 at 9:17 PM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Mon, Jan 16, 2023 at 03:06:44PM +0800, Jason Wang wrote:
> > On Mon, Jan 16, 2023 at 7:30 AM Viktor Prutyanov <viktor@daynix.com> wrote:
> > >
> > > On Tue, Nov 29, 2022 at 11:10 AM Jason Wang <jasowang@redhat.com> wrote:
> > > >
> > > > Hi All:
> > > >
> > > > According to ATS, device should work if ATS is disabled. This is not
> > > > correctly implemented in the current intel-iommu since it doesn't
> > > > handle the UNMAP notifier correctly. This breaks the vhost-net +
> > > > vIOMMU without dt.
> > > >
> > > > The root casue is that the when there's a device IOTLB miss (note that
> > > > it's not specific to PCI so it can work without ATS), Qemu doesn't
> > > > build the IOVA tree, so when guest start an IOTLB invalidation, Qemu
> > > > won't trigger the UNMAP notifier.
> > > >
> > > > Fixing by build IOVA tree during IOMMU translsation.
> > > >
> > > > Thanks
> > > >
> > > > Jason Wang (3):
> > > >   intel-iommu: fail MAP notifier without caching mode
> > > >   intel-iommu: fail DEVIOTLB_UNMAP without dt mode
> > > >   intel-iommu: build iova tree during IOMMU translation
> > > >
> > > >  hw/i386/intel_iommu.c | 58 ++++++++++++++++++++++++-------------------
> > > >  1 file changed, 33 insertions(+), 25 deletions(-)
> > > >
> > > > --
> > > > 2.25.1
> > > >
> > >
> > > Hi Jason,
> > >
> > > I've tried the series with Windows Server 2022 guest with vhost and
> > > intel-iommu (device-iotlb=off) and now networking on this system has
> > > become working.
> > > So, as we discussed, I'm waiting for the series to be accepted in some
> > > form to continue my work about supporting guests who refuse Device-TLB
> > > on systems with device-iotlb=on.
> > >
> > > Tested-by: Viktor Prutyanov <viktor@daynix.com>
> >
> > Great, Peter has some comments on this series, so I will probably send
> > a new version (probably after the chinese new year).
> >
> > Thanks
>
> Were you going to post a new version?

Yes.

Thanks

>
> > >
> > > Best regards,
> > > Viktor Prutyanov
> > >
>

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Laurent Vivier]

On 11/29/22 09:10, Jason Wang wrote:
> Without dt mode, device IOTLB notifier won't work since guest won't
> send device IOTLB invalidation descriptor in this case. Let's fail
> early instead of misbehaving silently.
> 
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
>   hw/i386/intel_iommu.c | 8 ++++++++
>   1 file changed, 8 insertions(+)
> 
> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> index 9143376677..d025ef2873 100644
> --- a/hw/i386/intel_iommu.c
> +++ b/hw/i386/intel_iommu.c
> @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>   {
>       VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
>       IntelIOMMUState *s = vtd_as->iommu_state;
> +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
>   
>       /* TODO: add support for VFIO and vhost users */
>       if (s->snoop_control) {
> @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>                            PCI_FUNC(vtd_as->devfn));
>           return -ENOTSUP;
>       }
> +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> +        error_setg_errno(errp, ENOTSUP,
> +                         "device %02x.%02x.%x requires device IOTLB mode",
> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> +                         PCI_FUNC(vtd_as->devfn));
> +        return -ENOTSUP;
> +    }
>   
>       /* Update per-address-space notifier flags */
>       vtd_as->notifier_flags = new;

Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Tested-by: Laurent Vivier <lvivier@redhat.com>
Buglink: https://bugzilla.redhat.com/2156876

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Laurent Vivier]

On 2/3/23 10:08, Laurent Vivier wrote:
> On 11/29/22 09:10, Jason Wang wrote:
>> Without dt mode, device IOTLB notifier won't work since guest won't
>> send device IOTLB invalidation descriptor in this case. Let's fail
>> early instead of misbehaving silently.
>>
>> Signed-off-by: Jason Wang <jasowang@redhat.com>
>> ---
>>   hw/i386/intel_iommu.c | 8 ++++++++
>>   1 file changed, 8 insertions(+)
>>
>> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
>> index 9143376677..d025ef2873 100644
>> --- a/hw/i386/intel_iommu.c
>> +++ b/hw/i386/intel_iommu.c
>> @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>>   {
>>       VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
>>       IntelIOMMUState *s = vtd_as->iommu_state;
>> +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
>>       /* TODO: add support for VFIO and vhost users */
>>       if (s->snoop_control) {
>> @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
>>                            PCI_FUNC(vtd_as->devfn));
>>           return -ENOTSUP;
>>       }
>> +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
>> +        error_setg_errno(errp, ENOTSUP,
>> +                         "device %02x.%02x.%x requires device IOTLB mode",
>> +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
>> +                         PCI_FUNC(vtd_as->devfn));
>> +        return -ENOTSUP;
>> +    }
>>       /* Update per-address-space notifier flags */
>>       vtd_as->notifier_flags = new;
> 
> Reviewed-by: Laurent Vivier <lvivier@redhat.com>
> Tested-by: Laurent Vivier <lvivier@redhat.com>
> Buglink: https://bugzilla.redhat.com/2156876

Is this possible to have this patch merged?
It fixes a real problem and it is really trivial.

Thanks,
Laurent

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Peter Xu]

On Tue, Feb 07, 2023 at 05:17:42PM +0100, Laurent Vivier wrote:
> On 2/3/23 10:08, Laurent Vivier wrote:
> > On 11/29/22 09:10, Jason Wang wrote:
> > > Without dt mode, device IOTLB notifier won't work since guest won't
> > > send device IOTLB invalidation descriptor in this case. Let's fail
> > > early instead of misbehaving silently.
> > > 
> > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > ---
> > >   hw/i386/intel_iommu.c | 8 ++++++++
> > >   1 file changed, 8 insertions(+)
> > > 
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index 9143376677..d025ef2873 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> > >   {
> > >       VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
> > >       IntelIOMMUState *s = vtd_as->iommu_state;
> > > +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
> > >       /* TODO: add support for VFIO and vhost users */
> > >       if (s->snoop_control) {
> > > @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> > >                            PCI_FUNC(vtd_as->devfn));
> > >           return -ENOTSUP;
> > >       }
> > > +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> > > +        error_setg_errno(errp, ENOTSUP,
> > > +                         "device %02x.%02x.%x requires device IOTLB mode",
> > > +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> > > +                         PCI_FUNC(vtd_as->devfn));
> > > +        return -ENOTSUP;
> > > +    }
> > >       /* Update per-address-space notifier flags */
> > >       vtd_as->notifier_flags = new;
> > 
> > Reviewed-by: Laurent Vivier <lvivier@redhat.com>
> > Tested-by: Laurent Vivier <lvivier@redhat.com>
> > Buglink: https://bugzilla.redhat.com/2156876
> 
> Is this possible to have this patch merged?
> It fixes a real problem and it is really trivial.

AFAIU Jason will post a new version soon for this whole set.  But I also
agree if Michael has an earlier pull we can add this in even earlier.

Thanks,

-- 
Peter Xu

Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Jason Wang]

On Fri, Dec 2, 2022 at 12:03 AM Peter Xu <peterx@redhat.com> wrote:
>
> On Tue, Nov 29, 2022 at 04:10:36PM +0800, Jason Wang wrote:
> > Without dt mode, device IOTLB notifier won't work since guest won't
> > send device IOTLB invalidation descriptor in this case. Let's fail
> > early instead of misbehaving silently.
> >
> > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > ---
> >  hw/i386/intel_iommu.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> >
> > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > index 9143376677..d025ef2873 100644
> > --- a/hw/i386/intel_iommu.c
> > +++ b/hw/i386/intel_iommu.c
> > @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> >  {
> >      VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
> >      IntelIOMMUState *s = vtd_as->iommu_state;
> > +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
> >
> >      /* TODO: add support for VFIO and vhost users */
> >      if (s->snoop_control) {
> > @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> >                           PCI_FUNC(vtd_as->devfn));
> >          return -ENOTSUP;
> >      }
> > +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> > +        error_setg_errno(errp, ENOTSUP,
> > +                         "device %02x.%02x.%x requires device IOTLB mode",
> > +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> > +                         PCI_FUNC(vtd_as->devfn));
> > +        return -ENOTSUP;
> > +    }
>
> While my r-b holds.. let's also do this for amd-iommu in the same patch?
> dt never supported there, so we can fail as long as DEVIOTLB registered.

Looks like there's one implementation:

Per spec:

""
The INVALIDATE_IOTLB_PAGES command is only present in IOMMU
implementations that support remote IOTLB caching of translations (see
Capability Offset 00h[IotlbSup]). This command instructs the specified
device to invalidate the given range of addresses in its IOTLB. The
size of the invalidate command is determined by the S bit and the
address.
""

And it has one implementation (though buggy) iommu_inval_iotlb() which
doesn't trigger any DEVIOTLB_UNMAP notifier.

We can leave this for the future.

(Last time I tried amd-iommu it didn't even boot).

Thanks

>
> >
> >      /* Update per-address-space notifier flags */
> >      vtd_as->notifier_flags = new;
> > --
> > 2.25.1
> >
>
> --
> Peter Xu
>