Skip to contentSkip to conversation

VFIO/SMMUv3: Fail on VFIO/HW nested paging detection

PATCH In review Eric Auger 2 participants 6 messages Aug 29, 2019 – Aug 30, 2019

lore
Share & tools
lore

Eric Auger

author
#0 lore raw

v5 cover VFIO/SMMUv3: Fail on VFIO/HW nested paging detection

As of today when a guest is assigned with a host PCI device and an SMMUv3, VFIO calls memory_region_iommu_replay() default implementation. This translates the whole address range and completely stalls the execution. As VFIO/SMMUv3 integration is not supported yet (it requires SMMUv3 HW nested paging), let's recognize this situation and fail.

Best Regards

Eric

This series can be found at:
https://github.com/eauger/qemu/tree/v4.1.0_smmu_vfio_fail_v5

History:

v4 -> v5:
- v4 patches: 1, 4, 5 were upstreamed separately
- IOMMU_ATTR_HW_NESTED_PAGING renamed into
IOMMU_ATTR_NEED_HW_NESTED_PAGING

v3 -> v4:
- see individual patches

v2 -> v3:
- squash IOMMU_ATTR_VFIO_NESTED introduction and SMMUv3 usage
- assert when recognizing VFIO/NESTED case
- collect R-bs

v1 -> v2:
- Added "memory: Remove unused memory_region_iommu_replay_all()" & "hw/arm/smmuv3: Log a guest error when decoding an invalid STE"
- do not attempt to implement replay Cb but rather remove the call in case it is not needed
- explain why we do not remove other log messages on config decoding

Eric Auger (2):
memory: Add IOMMU_ATTR_NEED_HW_NESTED_PAGING IOMMU memory region
attribute
hw/vfio/common: Fail on VFIO/HW nested paging detection

 hw/arm/smmuv3.c       | 12 ++++++++++++
 hw/vfio/common.c      | 10 ++++++++++
 include/exec/memory.h |  8 +++++++-
 3 files changed, 29 insertions(+), 1 deletion(-)
Signature · 1 line
2.20.1
author
#1 lore raw

PATCH v5 1/2 memory: Add IOMMU_ATTR_NEED_HW_NESTED_PAGING IOMMU memory region attribute

We introduce a new IOMMU Memory Region attribute, IOMMU_ATTR_NEED_HW_NESTED_PAGING that tells whether the virtual IOMMU relies on physical IOMMU HW nested paging capability when protecting host assigned devices.

Current Intel virtual IOMMU device supports "Caching Mode" and does not require 2 stages at physical level to be integrated with VFIO. However SMMUv3 does not implement such "caching mode" and requires HW nested paging.

As such SMMUv3 is the first IOMMU device to advertise this attribute.

This new attribute will allow the VFIO code to specialize its handling.

---

v4 -> v5:
- patches 1, 4, 5 were upstreamed separately
- s/IOMMU_ATTR_HW_NESTED_PAGING/IOMMU_ATTR_NEED_HW_NESTED_PAGING

v3 -> v4:
- s/IOMMU_ATTR_VFIO_NESTED/IOMMU_ATTR_HW_NESTED_PAGING
- add comments related to the existing attributes
- fix space after the cast

---
 hw/arm/smmuv3.c       | 12 ++++++++++++
 include/exec/memory.h |  8 +++++++-
 2 files changed, 19 insertions(+), 1 deletion(-)
hw/arm/smmuv3.c +12 -0
index 2eaf07fb5f..a932bf7136 100644
@@ -1490,6 +1490,17 @@ static void smmuv3_notify_flag_changed(IOMMUMemoryRegion *iommu,
     }
 }
 
+static int smmuv3_get_attr(IOMMUMemoryRegion *iommu,
+                           enum IOMMUMemoryRegionAttr attr,
+                           void *data)
+{
+    if (attr == IOMMU_ATTR_NEED_HW_NESTED_PAGING) {
+        *(bool *)data = true;
+        return 0;
+    }
+    return -EINVAL;
+}
+
 static void smmuv3_iommu_memory_region_class_init(ObjectClass *klass,
                                                   void *data)
 {
@@ -1497,6 +1508,7 @@ static void smmuv3_iommu_memory_region_class_init(ObjectClass *klass,
 
     imrc->translate = smmuv3_translate;
     imrc->notify_flag_changed = smmuv3_notify_flag_changed;
+    imrc->get_attr = smmuv3_get_attr;
 }
 
 static const TypeInfo smmuv3_type_info = {
include/exec/memory.h +7 -1
index fddc2ff48a..61493633fa 100644
@@ -212,7 +212,13 @@ typedef struct MemoryRegionClass {
 
 
 enum IOMMUMemoryRegionAttr {
-    IOMMU_ATTR_SPAPR_TCE_FD
+    /* Retrieve an integer corresponding to the TCE file descriptor */
+    IOMMU_ATTR_SPAPR_TCE_FD,
+    /*
+     * Retrieve a boolean that indicates whether the virtual IOMMU relies
+     * on physical IOMMU HW nested paging to protect host assigned devices
+     */
+    IOMMU_ATTR_NEED_HW_NESTED_PAGING,
 };
 
 /**
Signature · 1 line
2.20.1
author
#2 lore raw

PATCH v5 2/2 hw/vfio/common: Fail on VFIO/HW nested paging detection

In reply to Eric Auger

As of today, VFIO only works along with vIOMMU supporting caching mode. The SMMUv3 does not support this mode and requires HW nested paging to work properly with VFIO.

So any attempt to run a VFIO device protected by such IOMMU would prevent the assigned device from working and at the moment the guest does not even boot as the default memory_region_iommu_replay() implementation attempts to translate the whole address space and completely stalls the guest.

So let's fail on that case.

---

v3 -> v4:
- use IOMMU_ATTR_HW_NESTED_PAGING
- do not abort anymore but jump to fail

---
 hw/vfio/common.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/vfio/common.c +10 -0
index 3e03c495d8..e8c009d019 100644
@@ -606,9 +606,19 @@ static void vfio_listener_region_add(MemoryListener *listener,
     if (memory_region_is_iommu(section->mr)) {
         VFIOGuestIOMMU *giommu;
         IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr);
+        bool nested;
         int iommu_idx;
 
         trace_vfio_listener_region_add_iommu(iova, end);
+
+        if (!memory_region_iommu_get_attr(iommu_mr,
+                                          IOMMU_ATTR_NEED_HW_NESTED_PAGING,
+                                          (void *)&nested) && nested) {
+            error_report("VFIO/vIOMMU integration based on HW nested paging "
+                         "is not yet supported");
+            ret = -EINVAL;
+            goto fail;
+        }
         /*
          * FIXME: For VFIO iommu types which have KVM acceleration to
          * avoid bouncing all map/unmaps through qemu this way, this
Signature · 1 line
2.20.1
#3 lore raw

Re: [Qemu-arm] [PATCH v5 2/2] hw/vfio/common: Fail on VFIO/HW nested paging detection

Quoted 10 lines · On Thu, 29 Aug 2019 11:01:41 +0200 Eric Auger <[email protected]> wrote:

As of today, VFIO only works along with vIOMMU supporting caching mode. The SMMUv3 does not support this mode and requires HW nested paging to work properly with VFIO.

So any attempt to run a VFIO device protected by such IOMMU would prevent the assigned device from working and at the moment the guest does not even boot as the default memory_region_iommu_replay() implementation attempts to translate the whole address space and completely stalls the guest.

Why doesn't this stall an x86 guest?

I'm a bit confused about what this provides versus the flag_changed notifier looking for IOMMU_NOTIFIER_MAP, which AIUI is the common deficiency between VT-d w/o caching-mode and SMMUv3 w/o nested mode. The iommu notifier is registered prior to calling iommu_replay, so it seems we already have an opportunity to do something there. Help me understand why this is needed. Thanks,

Alex

Quoted 38 lines

So let's fail on that case.

---

v3 -> v4:
- use IOMMU_ATTR_HW_NESTED_PAGING
- do not abort anymore but jump to fail

---
 hw/vfio/common.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/vfio/common.c +10 -0
index 3e03c495d8..e8c009d019 100644
@@ -606,9 +606,19 @@ static void vfio_listener_region_add(MemoryListener *listener,
     if (memory_region_is_iommu(section->mr)) {
         VFIOGuestIOMMU *giommu;
         IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr);
+        bool nested;
         int iommu_idx;
 
         trace_vfio_listener_region_add_iommu(iova, end);
+
+        if (!memory_region_iommu_get_attr(iommu_mr,
+                                          IOMMU_ATTR_NEED_HW_NESTED_PAGING,
+                                          (void *)&nested) && nested) {
+            error_report("VFIO/vIOMMU integration based on HW nested paging "
+                         "is not yet supported");
+            ret = -EINVAL;
+            goto fail;
+        }
         /*
          * FIXME: For VFIO iommu types which have KVM acceleration to
          * avoid bouncing all map/unmaps through qemu this way, this
author
#4 lore raw

Re: [Qemu-arm] [Qemu-devel] [PATCH v5 2/2] hw/vfio/common: Fail on VFIO/HW nested paging detection

Hi Alex,

Quoted 7 lines · Alex Williamson
Quoted 4 lines · On Thu, 29 Aug 2019 11:01:41 +0200 Eric Auger <[email protected]> wrote:

As of today, VFIO only works along with vIOMMU supporting caching mode. The SMMUv3 does not support this mode and requires HW nested paging to work properly with VFIO.

Quoted 8 lines
Quoted 6 lines

So any attempt to run a VFIO device protected by such IOMMU would prevent the assigned device from working and at the moment the guest does not even boot as the default memory_region_iommu_replay() implementation attempts to translate the whole address space and completely stalls the guest.

Why doesn't this stall an x86 guest?

it does not stall on x86 since intel_iommu implements a custom replay (see vtd_iommu_replay) and you do not execute the dummy default one. This function performs a full page table walk, scanning all the valid entries and calling the MAP notifier on those. Although this operation is tedious it has nothing to compare against the dummy default replay function which calls translate() on the whole address range (on a page basis).

Quoted 7 lines

I'm a bit confused about what this provides versus the flag_changed notifier looking for IOMMU_NOTIFIER_MAP, which AIUI is the common deficiency between VT-d w/o caching-mode and SMMUv3 w/o nested mode. The iommu notifier is registered prior to calling iommu_replay, so it seems we already have an opportunity to do something there. Help me understand why this is needed. Thanks,

At the moment the smmuv3 notify_flag_changed callback implementation (smmuv3_notify_flag_changed) emits a warning when it detects a MAP notifier gets registered:

warn_report("SMMUv3 does not support notification on MAP: "

            "device %s will not function properly", pcidev->name);

and then the replay gets executed, looping forever.

I could exit instead of emitting a warning but the drawback is that on vfio hotplug, it will also exit whereas we would rather simply reject the hotplug.

I think the solution based on the IOMMU MR attribute handles both the static and hotplug solutions. Also looking further, I will need this IOMMU MR attribute for 2stage SMMU integration (see [RFC v5 14/29] vfio:
Force nested if iommu requires it). I know that it is standing for a while and it is still hypothetical but setting up 2stage will require specific treatments in the vfio common.c code, opt-in the 2stage mode, register specific iommu mr notifiers. Using the IOMMU MR attribute allows me to detect which kind of VFIO/IOMMU integration I need to setup.

Thanks

Eric

Quoted 43 lines

Alex

Quoted 38 lines

So let's fail on that case.

---

v3 -> v4:
- use IOMMU_ATTR_HW_NESTED_PAGING
- do not abort anymore but jump to fail

---
 hw/vfio/common.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/vfio/common.c +10 -0
index 3e03c495d8..e8c009d019 100644
@@ -606,9 +606,19 @@ static void vfio_listener_region_add(MemoryListener *listener,
     if (memory_region_is_iommu(section->mr)) {
         VFIOGuestIOMMU *giommu;
         IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr);
+        bool nested;
         int iommu_idx;
 
         trace_vfio_listener_region_add_iommu(iova, end);
+
+        if (!memory_region_iommu_get_attr(iommu_mr,
+                                          IOMMU_ATTR_NEED_HW_NESTED_PAGING,
+                                          (void *)&nested) && nested) {
+            error_report("VFIO/vIOMMU integration based on HW nested paging "
+                         "is not yet supported");
+            ret = -EINVAL;
+            goto fail;
+        }
         /*
          * FIXME: For VFIO iommu types which have KVM acceleration to
          * avoid bouncing all map/unmaps through qemu this way, this
#5 lore raw

Re: [Qemu-arm] [Qemu-devel] [PATCH v5 2/2] hw/vfio/common: Fail on VFIO/HW nested paging detection

Quoted 17 lines · On Fri, 30 Aug 2019 10:06:56 +0200 Auger Eric <[email protected]> wrote:

Hi Alex,

Quoted 14 lines · Alex Williamson
Quoted 10 lines · On Thu, 29 Aug 2019 11:01:41 +0200 Eric Auger <[email protected]> wrote:

As of today, VFIO only works along with vIOMMU supporting caching mode. The SMMUv3 does not support this mode and requires HW nested paging to work properly with VFIO.

So any attempt to run a VFIO device protected by such IOMMU would prevent the assigned device from working and at the moment the guest does not even boot as the default memory_region_iommu_replay() implementation attempts to translate the whole address space and completely stalls the guest.

Quoted 8 lines
Quoted 1 line

Why doesn't this stall an x86 guest?

it does not stall on x86 since intel_iommu implements a custom replay (see vtd_iommu_replay) and you do not execute the dummy default one. This function performs a full page table walk, scanning all the valid entries and calling the MAP notifier on those. Although this operation is tedious it has nothing to compare against the dummy default replay function which calls translate() on the whole address range (on a page basis).

Ah right. OTOH, what are the arguments against smmuv3 providing a replay function?

Quoted 11 lines
Quoted 6 lines

I'm a bit confused about what this provides versus the flag_changed notifier looking for IOMMU_NOTIFIER_MAP, which AIUI is the common deficiency between VT-d w/o caching-mode and SMMUv3 w/o nested mode. The iommu notifier is registered prior to calling iommu_replay, so it seems we already have an opportunity to do something there. Help me understand why this is needed. Thanks,

At the moment the smmuv3 notify_flag_changed callback implementation (smmuv3_notify_flag_changed) emits a warning when it detects a MAP notifier gets registered:

Quoted 8 lines

warn_report("SMMUv3 does not support notification on MAP: "

            "device %s will not function properly", pcidev->name);

and then the replay gets executed, looping forever.

I could exit instead of emitting a warning but the drawback is that on vfio hotplug, it will also exit whereas we would rather simply reject the hotplug.

There are solutions to the above by modifying the existing framework rather than creating a parallel solution though. For instance, could memory_region_register_iommu_notifier() reject the notifier if the flag change is incompatible, allowing the fault to propagate back to vfio and taking a similar exit path as provided here.

Quoted 9 lines

I think the solution based on the IOMMU MR attribute handles both the static and hotplug solutions. Also looking further, I will need this IOMMU MR attribute for 2stage SMMU integration (see [RFC v5 14/29]
vfio: Force nested if iommu requires it). I know that it is standing for a while and it is still hypothetical but setting up 2stage will require specific treatments in the vfio common.c code, opt-in the 2stage mode, register specific iommu mr notifiers. Using the IOMMU MR attribute allows me to detect which kind of VFIO/IOMMU integration I need to setup.

Hmm, I'm certainly more on board with that use case. I guess the question is whether the problem statement presented here justifies what seems to be a parallel solution to what we have today, or could have with some enhancements. Thanks,

Alex

Quoted 46 lines
Quoted 46 lines
Quoted 44 lines

So let's fail on that case.

---

v3 -> v4:
- use IOMMU_ATTR_HW_NESTED_PAGING
- do not abort anymore but jump to fail

---
 hw/vfio/common.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/vfio/common.c +0 -0
index 3e03c495d8..e8c009d019 100644
@@ -606,9 +606,19 @@ static void

vfio_listener_region_add(MemoryListener *listener, if (memory_region_is_iommu(section->mr)) { VFIOGuestIOMMU *giommu;
IOMMUMemoryRegion *iommu_mr =
IOMMU_MEMORY_REGION(section->mr);

hw/vfio/common.c +4 -0
+        bool nested;
         int iommu_idx;
 
         trace_vfio_listener_region_add_iommu(iova, end);
+
+        if (!memory_region_iommu_get_attr(iommu_mr,
+

IOMMU_ATTR_NEED_HW_NESTED_PAGING,
+ (void *)&nested) &&
nested) {
+ error_report("VFIO/vIOMMU integration based on HW
nested paging "

hw/vfio/common.c +4 -0
+                         "is not yet supported");
+            ret = -EINVAL;
+            goto fail;
+        }
         /*
          * FIXME: For VFIO iommu types which have KVM

acceleration to
* avoid bouncing all map/unmaps through qemu this way,
this