Skip to contentSkip to conversation

s390: virtio: let arch validate VIRTIO features

PATCH Superseded Pierre Morel 2 participants 8 messages Aug 18, 2020 – Aug 19, 2020

Subsystems (4)

v8 of 12 Superseded Latest is v12 →

lore
Share & tools
lore
#0 lore raw

v8 cover s390: virtio: let arch validate VIRTIO features

Hi all,

The goal of the series is to give a chance to the architecture to validate VIRTIO device features.

in this respin:

I use the original idea from Connie for an optional arch_has_restricted_memory_access.

I renamed the callback accordingly, added the definition of ARCH_HAS_RESTRICTED_MEMORY_ACCESS inside the VIRTIO Kconfig and the selection in the PROTECTED_VIRTUALIZATION_GUEST config entry.

Regards,
Pierre

Pierre Morel (2):
virtio: let arch validate VIRTIO features
s390: virtio: PV needs VIRTIO I/O device protection

 arch/s390/Kconfig             |  1 +
 arch/s390/mm/init.c           | 30 ++++++++++++++++++++++++++++++
 drivers/virtio/Kconfig        |  6 ++++++
 drivers/virtio/virtio.c       |  4 ++++
 include/linux/virtio_config.h |  9 +++++++++
 5 files changed, 50 insertions(+)
Signature · 65 lines
2.25.1

Changelog

to v8:

- refactoring by using an optional callback
  (Connie)

to v7:

- typo in warning message
  (Connie)
to v6:

- rewording warning messages
  (Connie, Halil)

to v5:

- return directly from S390 arch_validate_virtio_features()
  when the guest is not protected.
  (Connie)

- Somme rewording
  (Connie, Michael)

- moved back code from arch/s390/ ...kernel/uv.c to ...mm/init.c
  (Christian)

to v4:

- separate virtio and arch code
  (Pierre)

- moved code from arch/s390/mm/init.c to arch/s390/kernel/uv.c
  (as interpreted from Heiko's comment)

- moved validation inside the arch code
  (Connie)

- moved the call to arch validation before VIRTIO_F_1 test
  (Michael)

to v3:

- add warning
  (Connie, Christian)

- add comment
  (Connie)

- change hook name
  (Halil, Connie)

to v2:

- put the test in virtio_finalize_features()
  (Connie)

- put the test inside VIRTIO core
  (Jason)

- pass a virtio device as parameter
  (Halil)
#1 lore raw

PATCH v8 1/2 virtio: let arch validate VIRTIO features

An architecture may need to validate the VIRTIO devices features based on architecture specifics.

Provide a new Kconfig entry, CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS, the architecture can select when it provides a callback named arch_has_restricted_memory_access to validate the virtio device features.

---
 drivers/virtio/Kconfig        | 6 ++++++
 drivers/virtio/virtio.c       | 4 ++++
 include/linux/virtio_config.h | 9 +++++++++
 3 files changed, 19 insertions(+)
drivers/virtio/Kconfig +6 -0
index 5809e5f5b157..eef09e3c92f9 100644
@@ -6,6 +6,12 @@ config VIRTIO
 	  bus, such as CONFIG_VIRTIO_PCI, CONFIG_VIRTIO_MMIO, CONFIG_RPMSG
 	  or CONFIG_S390_GUEST.
 
+config ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+	bool
+	help
+	  This option is selected by any architecture enforcing
+	  VIRTIO_F_IOMMU_PLATFORM
+
 menuconfig VIRTIO_MENU
 	bool "Virtio drivers"
 	default y
drivers/virtio/virtio.c +4 -0
index a977e32a88f2..1471db7d6510 100644
@@ -176,6 +176,10 @@ int virtio_finalize_features(struct virtio_device *dev)
 	if (ret)
 		return ret;
 
+	ret = arch_has_restricted_memory_access(dev);
+	if (ret)
+		return ret;
+
 	if (!virtio_has_feature(dev, VIRTIO_F_VERSION_1))
 		return 0;
 
include/linux/virtio_config.h +9 -0
index bb4cc4910750..f6b82541c497 100644
@@ -459,4 +459,13 @@ static inline void virtio_cwrite64(struct virtio_device *vdev,
 		_r;							\
 	})
 
+#ifdef CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+int arch_has_restricted_memory_access(struct virtio_device *dev);
+#else
+static inline int arch_has_restricted_memory_access(struct virtio_device *dev)
+{
+	return 0;
+}
+#endif /* CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS */
+
 #endif /* _LINUX_VIRTIO_CONFIG_H */
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH v8 1/2] virtio: let arch validate VIRTIO features

Quoted 20 lines · On Tue, 18 Aug 2020 16:58:30 +0200 Pierre Morel <[email protected]> wrote:

An architecture may need to validate the VIRTIO devices features based on architecture specifics.

Provide a new Kconfig entry, CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS, the architecture can select when it provides a callback named arch_has_restricted_memory_access to validate the virtio device features.

---
 drivers/virtio/Kconfig        | 6 ++++++
 drivers/virtio/virtio.c       | 4 ++++
 include/linux/virtio_config.h | 9 +++++++++
 3 files changed, 19 insertions(+)
drivers/virtio/Kconfig +0 -0
index 5809e5f5b157..eef09e3c92f9 100644

drivers/virtio/Kconfig config VIRTIO · line 13

Quoted 8 lines
drivers/virtio/Kconfig +5 -0
@@ -6,6 +6,12 @@ config VIRTIO
 	  bus, such as CONFIG_VIRTIO_PCI, CONFIG_VIRTIO_MMIO, CONFIG_RPMSG
 	  or CONFIG_S390_GUEST.
 
+config ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+	bool
+	help
+	  This option is selected by any architecture enforcing
+	  VIRTIO_F_IOMMU_PLATFORM

This option is only for a very specific case of "restricted memory access", namely the kind that requires IOMMU_PLATFORM for virtio devices. ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS? Or is this intended to cover cases outside of virtio as well?

Quoted 9 lines
drivers/virtio/Kconfig +1 -0
+
 menuconfig VIRTIO_MENU
 	bool "Virtio drivers"
 	default y
drivers/virtio/virtio.c +0 -0
index a977e32a88f2..1471db7d6510 100644

drivers/virtio/virtio.c int virtio_finalize_features(struct virtio_device *dev) · line 181

Quoted 6 lines
drivers/virtio/virtio.c +3 -0
@@ -176,6 +176,10 @@ int virtio_finalize_features(struct virtio_device *dev)
 	if (ret)
 		return ret;
 
+	ret = arch_has_restricted_memory_access(dev);
+	if (ret)
+		return ret;

Hm, I'd rather have expected something like

if (arch_has_restricted_memory_access(dev)) {
	// enforce VERSION_1 and IOMMU_PLATFORM
}

Otherwise, you're duplicating the checks in the individual architecture callbacks again.

[Not sure whether the device argument would be needed here; are there architectures where we'd only require IOMMU_PLATFORM for a subset of virtio devices?]

Quoted 22 lines
drivers/virtio/virtio.c +1 -0
+
 	if (!virtio_has_feature(dev, VIRTIO_F_VERSION_1))
 		return 0;
 
include/linux/virtio_config.h +9 -0
index bb4cc4910750..f6b82541c497 100644
@@ -459,4 +459,13 @@ static inline void virtio_cwrite64(struct virtio_device *vdev,
 		_r;							\
 	})
 
+#ifdef CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+int arch_has_restricted_memory_access(struct virtio_device *dev);
+#else
+static inline int arch_has_restricted_memory_access(struct virtio_device *dev)
+{
+	return 0;
+}
+#endif /* CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS */
+
 #endif /* _LINUX_VIRTIO_CONFIG_H */
#3 lore raw

Re: [PATCH v8 1/2] virtio: let arch validate VIRTIO features

Quoted 3 lines · Cornelia Huck

On Tue, 18 Aug 2020 16:58:30 +0200
Pierre Morel <[email protected]> wrote:

...

Quoted 10 lines

drivers/virtio/Kconfig config VIRTIO · line 13

Quoted 5 lines
drivers/virtio/Kconfig +5 -0
+config ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+	bool
+	help
+	  This option is selected by any architecture enforcing
+	  VIRTIO_F_IOMMU_PLATFORM

This option is only for a very specific case of "restricted memory access", namely the kind that requires IOMMU_PLATFORM for virtio devices. ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS? Or is this intended to cover cases outside of virtio as well?

AFAIK we did not identify other restrictions so adding VIRTIO in the name should be the best thing to do.

If new restrictions appear they also may be orthogonal.

I will change to ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS if no one complains.

Quoted 13 lines
Quoted 13 lines
drivers/virtio/Kconfig +1 -0
+
  menuconfig VIRTIO_MENU
  	bool "Virtio drivers"
  	default y
drivers/virtio/virtio.c +0 -0
index a977e32a88f2..1471db7d6510 100644
@@ -176,6 +176,10 @@ int virtio_finalize_features(struct virtio_device *dev)
  	if (ret)
  		return ret;
 
Quoted 7 lines

drivers/virtio/virtio.c int virtio_finalize_features(struct virtio_device *dev) · line 181

Quoted 3 lines
drivers/virtio/virtio.c +3 -0
+	ret = arch_has_restricted_memory_access(dev);
+	if (ret)
+		return ret;

Hm, I'd rather have expected something like

if (arch_has_restricted_memory_access(dev)) {

may be also change the callback name to
arch_has_restricted_virtio_memory_access() ?

Quoted 5 lines
	// enforce VERSION_1 and IOMMU_PLATFORM
}

Otherwise, you're duplicating the checks in the individual architecture callbacks again.

Yes, I agree and go back this way.

Quoted 4 lines

[Not sure whether the device argument would be needed here; are there architectures where we'd only require IOMMU_PLATFORM for a subset of virtio devices?]

I don't think so and since we do the checks locally, we do not need the device argument anymore.

Thanks,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#4 lore raw

Re: [PATCH v8 1/2] virtio: let arch validate VIRTIO features

Quoted 43 lines · On Wed, 19 Aug 2020 10:50:18 +0200 Pierre Morel <[email protected]> wrote:
Quoted 3 lines · Cornelia Huck

On Tue, 18 Aug 2020 16:58:30 +0200
Pierre Morel <[email protected]> wrote:

...

Quoted 10 lines

drivers/virtio/Kconfig config VIRTIO · line 13

Quoted 5 lines
drivers/virtio/Kconfig +5 -0
+config ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+	bool
+	help
+	  This option is selected by any architecture enforcing
+	  VIRTIO_F_IOMMU_PLATFORM  

This option is only for a very specific case of "restricted memory access", namely the kind that requires IOMMU_PLATFORM for virtio devices. ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS? Or is this intended to cover cases outside of virtio as well?

AFAIK we did not identify other restrictions so adding VIRTIO in the name should be the best thing to do.

If new restrictions appear they also may be orthogonal.

I will change to ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS if no one complains.

Quoted 19 lines
Quoted 10 lines
drivers/virtio/Kconfig +1 -0
+
  menuconfig VIRTIO_MENU
  	bool "Virtio drivers"
  	default y
drivers/virtio/virtio.c +0 -0
index a977e32a88f2..1471db7d6510 100644

drivers/virtio/virtio.c int virtio_finalize_features(struct virtio_device *dev) · line 181

Quoted 6 lines
drivers/virtio/virtio.c +3 -0
@@ -176,6 +176,10 @@ int virtio_finalize_features(struct virtio_device *dev)
  	if (ret)
  		return ret;
 
+	ret = arch_has_restricted_memory_access(dev);
+	if (ret)
+		return ret;  

Hm, I'd rather have expected something like

Quoted 4 lines
Quoted 1 line

if (arch_has_restricted_memory_access(dev)) {

may be also change the callback name to
arch_has_restricted_virtio_memory_access() ?

Yes, why not.

Quoted 10 lines
Quoted 6 lines
	// enforce VERSION_1 and IOMMU_PLATFORM
}

Otherwise, you're duplicating the checks in the individual architecture callbacks again.

Yes, I agree and go back this way.

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

[Not sure whether the device argument would be needed here; are there architectures where we'd only require IOMMU_PLATFORM for a subset of virtio devices?]

I don't think so and since we do the checks locally, we do not need the device argument anymore.

Yes, that would also remove some layering entanglement.

#5 lore raw

PATCH v8 2/2 s390: virtio: PV needs VIRTIO I/O device protection

In reply to Pierre Morel

If protected virtualization is active on s390, the virtio queues are not accessible to the host, unless VIRTIO_F_IOMMU_PLATFORM has been negotiated.
Define CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS and export arch_has_restricted_memory_access to fail probe if that's not the case, preventing a host error on access attempt.

---
 arch/s390/Kconfig   |  1 +
 arch/s390/mm/init.c | 30 ++++++++++++++++++++++++++++++
 2 files changed, 31 insertions(+)
arch/s390/Kconfig +1 -0
index 9cfd8de907cb..d4a3ef4fa27b 100644
@@ -820,6 +820,7 @@ menu "Virtualization"
 config PROTECTED_VIRTUALIZATION_GUEST
 	def_bool n
 	prompt "Protected virtualization guest support"
+	select ARCH_HAS_RESTRICTED_MEMORY_ACCESS
 	help
 	  Select this option, if you want to be able to run this
 	  kernel as a protected virtualization KVM guest.
arch/s390/mm/init.c +30 -0
index 6dc7c3b60ef6..aec04d7dd089 100644
@@ -45,6 +45,7 @@
 #include <asm/kasan.h>
 #include <asm/dma-mapping.h>
 #include <asm/uv.h>
+#include <linux/virtio_config.h>
 
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __section(.bss..swapper_pg_dir);
 
@@ -161,6 +162,35 @@ bool force_dma_unencrypted(struct device *dev)
 	return is_prot_virt_guest();
 }
 
+#ifdef CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+/*
+ * arch_has_restricted_memory_access
+ * @dev: the VIRTIO device being added
+ *
+ * Return an error if required features are missing on a guest running
+ * with protected virtualization.
+ */
+int arch_has_restricted_memory_access(struct virtio_device *dev)
+{
+	if (!is_prot_virt_guest())
+		return 0;
+
+	if (!virtio_has_feature(dev, VIRTIO_F_VERSION_1)) {
+		dev_warn(&dev->dev, "device must provide VIRTIO_F_VERSION_1\n");
+		return -ENODEV;
+	}
+
+	if (!virtio_has_feature(dev, VIRTIO_F_IOMMU_PLATFORM)) {
+		dev_warn(&dev->dev,
+			 "device must provide VIRTIO_F_IOMMU_PLATFORM\n");
+		return -ENODEV;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(arch_has_restricted_memory_access);
+#endif
+
 /* protected virtualization */
 static void pv_init(void)
 {
Signature · 1 line
2.25.1
#6 lore raw

Re: [PATCH v8 2/2] s390: virtio: PV needs VIRTIO I/O device protection

Quoted 43 lines · On Tue, 18 Aug 2020 16:58:31 +0200 Pierre Morel <[email protected]> wrote:

If protected virtualization is active on s390, the virtio queues are not accessible to the host, unless VIRTIO_F_IOMMU_PLATFORM has been negotiated.
Define CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS and export arch_has_restricted_memory_access to fail probe if that's not the case, preventing a host error on access attempt.

---
 arch/s390/Kconfig   |  1 +
 arch/s390/mm/init.c | 30 ++++++++++++++++++++++++++++++
 2 files changed, 31 insertions(+)
arch/s390/Kconfig +1 -0
index 9cfd8de907cb..d4a3ef4fa27b 100644
@@ -820,6 +820,7 @@ menu "Virtualization"
 config PROTECTED_VIRTUALIZATION_GUEST
 	def_bool n
 	prompt "Protected virtualization guest support"
+	select ARCH_HAS_RESTRICTED_MEMORY_ACCESS
 	help
 	  Select this option, if you want to be able to run this
 	  kernel as a protected virtualization KVM guest.
arch/s390/mm/init.c +3 -0
index 6dc7c3b60ef6..aec04d7dd089 100644
@@ -45,6 +45,7 @@
 #include <asm/kasan.h>
 #include <asm/dma-mapping.h>
 #include <asm/uv.h>
+#include <linux/virtio_config.h>
 
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __section(.bss..swapper_pg_dir);
 
@@ -161,6 +162,35 @@ bool force_dma_unencrypted(struct device *dev)
 	return is_prot_virt_guest();
 }
 
+#ifdef CONFIG_ARCH_HAS_RESTRICTED_MEMORY_ACCESS
+/*

arch/s390/mm/init.c bool force_dma_unencrypted(struct device *dev) · line 176

Quoted 10 lines
arch/s390/mm/init.c +10 -0
+ * arch_has_restricted_memory_access
+ * @dev: the VIRTIO device being added
+ *
+ * Return an error if required features are missing on a guest running
+ * with protected virtualization.
+ */
+int arch_has_restricted_memory_access(struct virtio_device *dev)
+{
+	if (!is_prot_virt_guest())
+		return 0;

If you just did a

return is_prot_virt_guest();

and did the virtio feature stuff in the virtio core, this function would be short and sweet :)

Quoted 20 lines
arch/s390/mm/init.c +17 -0
+
+	if (!virtio_has_feature(dev, VIRTIO_F_VERSION_1)) {
+		dev_warn(&dev->dev, "device must provide VIRTIO_F_VERSION_1\n");
+		return -ENODEV;
+	}
+
+	if (!virtio_has_feature(dev, VIRTIO_F_IOMMU_PLATFORM)) {
+		dev_warn(&dev->dev,
+			 "device must provide VIRTIO_F_IOMMU_PLATFORM\n");
+		return -ENODEV;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(arch_has_restricted_memory_access);
+#endif
+
 /* protected virtualization */
 static void pv_init(void)
 {
#7 lore raw

Re: [PATCH v8 2/2] s390: virtio: PV needs VIRTIO I/O device protection

Quoted 8 lines · Cornelia Huck

arch/s390/mm/init.c bool force_dma_unencrypted(struct device *dev) · line 176

Quoted 5 lines

+ */

arch/s390/mm/init.c +4 -0
+int arch_has_restricted_memory_access(struct virtio_device *dev)
+{
+	if (!is_prot_virt_guest())
+		return 0;

If you just did a

Quoted 4 lines

return is_prot_virt_guest();

and did the virtio feature stuff in the virtio core, this function would be short and sweet :)

yes, the smaller the better, thanks

Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen