Skip to contentSkip to conversation

Add support for Graviton TRNG

PATCH In review Ali Saidi 5 participants 13 messages Jun 4, 2019 – Jul 1, 2019

Subsystems (4)
lore
Share & tools
lore

Ali Saidi

author
#0 lore raw

cover Add support for Graviton TRNG

AWS Graviton based systems provide an Arm SMC call in the vendor defined hypervisor region to read random numbers from a HW TRNG and return them to the guest.

We've observed slower guest boot and especially reboot times due to lack of entropy and providing access to a TRNG is meant to address this.

Ali Saidi (3):
arm/arm64: Add smccc hypervisor service identifiers
arm64: export acpi_psci_use_hvc
hwrng: Add support for AWS Graviton TRNG

 MAINTAINERS                           |   6 ++
 arch/arm64/kernel/acpi.c              |   1 +
 drivers/char/hw_random/Kconfig        |  13 ++++
 drivers/char/hw_random/Makefile       |   1 +
 drivers/char/hw_random/graviton-rng.c | 123 ++++++++++++++++++++++++++++++++++
 include/linux/arm-smccc.h             |   2 +
 6 files changed, 146 insertions(+)
 create mode 100644 drivers/char/hw_random/graviton-rng.c
Signature · 7 lines
2.15.3.AMZN


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#1 lore raw

PATCH 1/3 arm/arm64: Add smccc hypervisor service identifiers

The SMC calling convention defines service identifiers for generic and vendor specific hypervisor services. This patch adds those services to the list of existing service identifiers.

---
 include/linux/arm-smccc.h | 2 ++
 1 file changed, 2 insertions(+)
include/linux/arm-smccc.h +2 -0
index 18863d56273c..bcf796801e3b 100644
@@ -54,6 +54,8 @@
 #define ARM_SMCCC_OWNER_SIP		2
 #define ARM_SMCCC_OWNER_OEM		3
 #define ARM_SMCCC_OWNER_STANDARD	4
+#define ARM_SMCCC_OWNER_STANDARD_HV	5
+#define ARM_SMCCC_OWNER_VENDOR_HV	6
 #define ARM_SMCCC_OWNER_TRUSTED_APP	48
 #define ARM_SMCCC_OWNER_TRUSTED_APP_END	49
 #define ARM_SMCCC_OWNER_TRUSTED_OS	50
Signature · 1 line
2.15.3.AMZN
author
#2 lore raw

PATCH 2/3 arm64: export acpi_psci_use_hvc

In reply to Ali Saidi

Allow a module that wants to make SMC calls to detect if it should be using smc or hvc.

---
 arch/arm64/kernel/acpi.c | 1 +
 1 file changed, 1 insertion(+)
arch/arm64/kernel/acpi.c +1 -0
index 803f0494dd3e..ea41c6541d3c 100644
@@ -119,6 +119,7 @@ bool acpi_psci_use_hvc(void)
 {
 	return acpi_gbl_FADT.arm_boot_flags & ACPI_FADT_PSCI_USE_HVC;
 }
+EXPORT_SYMBOL_GPL(acpi_psci_use_hvc);
 
 /*
  * acpi_fadt_sanity_check() - Check FADT presence and carry out sanity
Signature · 1 line
2.15.3.AMZN
#3 lore raw

Re: [PATCH 2/3] arm64: export acpi_psci_use_hvc

Quoted 13 lines · Ali Saidi

Allow a module that wants to make SMC calls to detect if it should be using smc or hvc.

---
 arch/arm64/kernel/acpi.c | 1 +
 1 file changed, 1 insertion(+)
arch/arm64/kernel/acpi.c +0 -0
index 803f0494dd3e..ea41c6541d3c 100644

arch/arm64/kernel/acpi.c bool acpi_psci_use_hvc(void) · line 123

Quoted 5 lines
arch/arm64/kernel/acpi.c +1 -0
@@ -119,6 +119,7 @@ bool acpi_psci_use_hvc(void)
 {
 	return acpi_gbl_FADT.arm_boot_flags & ACPI_FADT_PSCI_USE_HVC;
 }
+EXPORT_SYMBOL_GPL(acpi_psci_use_hvc);
 

I would rather have this in drivers/firmware/psci/psci.c checking the value of psci_ops.conduit so that it's not just ACPI specific and can be used on DT platforms too if required.

--
Regards,
Sudeep

maintainer
#4 lore raw

Re: [PATCH 2/3] arm64: export acpi_psci_use_hvc

Quoted 9 lines · Sudeep Holla
Quoted 8 lines · Ali Saidi

Allow a module that wants to make SMC calls to detect if it should be using smc or hvc.

---
 arch/arm64/kernel/acpi.c | 1 +
 1 file changed, 1 insertion(+)
Quoted 14 lines

arch/arm64/kernel/acpi.c bool acpi_psci_use_hvc(void) · line 123

Quoted 10 lines
arch/arm64/kernel/acpi.c +1 -0
index 803f0494dd3e..ea41c6541d3c 100644
@@ -119,6 +119,7 @@ bool acpi_psci_use_hvc(void)
 {
 	return acpi_gbl_FADT.arm_boot_flags & ACPI_FADT_PSCI_USE_HVC;
 }
+EXPORT_SYMBOL_GPL(acpi_psci_use_hvc);
 

I would rather have this in drivers/firmware/psci/psci.c checking the value of psci_ops.conduit so that it's not just ACPI specific and can be used on DT platforms too if required.

I'd also like this to not hook into PSCI internals. This code cares about SMCCC, not PSCI. We also really shouldn't need to spread the conduit management everywhere, too. We should be abel to have probe code do:

	if (!is_smccc_1_1_available())
		goto fail_probe;

... and runtime code do:

	res = arm_smccc_1_1_call(...);

... which is much clearner.

I'd started cleaning that up [1], but I haven't had the chance to rebase and repost it.

Ali, I assume your firmware has SMCCCv1.1+. Is that the case?

Thanks,
Mark.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/mark/linux.git/log/?h=arm64/smccc-cleanup

#5 lore raw

Re: [PATCH 2/3] arm64: export acpi_psci_use_hvc

Quoted 25 lines · Mark Rutland
Quoted 9 lines · Sudeep Holla
Quoted 8 lines · Ali Saidi

Allow a module that wants to make SMC calls to detect if it should be using smc or hvc.

---
 arch/arm64/kernel/acpi.c | 1 +
 1 file changed, 1 insertion(+)
Quoted 14 lines

arch/arm64/kernel/acpi.c bool acpi_psci_use_hvc(void) · line 123

Quoted 10 lines
arch/arm64/kernel/acpi.c +1 -0
index 803f0494dd3e..ea41c6541d3c 100644
@@ -119,6 +119,7 @@ bool acpi_psci_use_hvc(void)
 {
 	return acpi_gbl_FADT.arm_boot_flags & ACPI_FADT_PSCI_USE_HVC;
 }
+EXPORT_SYMBOL_GPL(acpi_psci_use_hvc);
 

I would rather have this in drivers/firmware/psci/psci.c checking the value of psci_ops.conduit so that it's not just ACPI specific and can be used on DT platforms too if required.

Quoted 3 lines

I'd also like this to not hook into PSCI internals. This code cares about SMCCC, not PSCI. We also really shouldn't need to spread the conduit management everywhere, too.

I agree. I remember suggesting the same to Xilinx a while ago but I didn't see your patches in the mainline.

--
Regards,
Sudeep

author
#6 lore raw

Re: [PATCH 2/3] arm64: export acpi_psci_use_hvc

In reply to Mark Rutland

On 6/5/19, 5:40 AM, "Mark Rutland" <[email protected]> wrote:

Ali, I assume your firmware has SMCCCv1.1+. Is that the case?

Yes, it does. I'm happy to be able to call a generic function instead of having to figure out which conduit to use.

Ali

author
#7 lore raw

PATCH 3/3 hwrng: Add support for AWS Graviton TRNG

In reply to Ali Saidi

AWS Graviton based systems provide an Arm SMC call in the vendor defined hypervisor region to read random numbers from a TRNG and return them to the guest.

---
 MAINTAINERS                           |   6 ++
 drivers/char/hw_random/Kconfig        |  13 ++++
 drivers/char/hw_random/Makefile       |   1 +
 drivers/char/hw_random/graviton-rng.c | 130 ++++++++++++++++++++++++++++++++++
 4 files changed, 150 insertions(+)
 create mode 100644 drivers/char/hw_random/graviton-rng.c
MAINTAINERS +6 -0
index 005902ea1450..e9c490e3fd9d 100644
@@ -2771,6 +2771,12 @@ S:	Maintained
 F:	Documentation/devicetree/bindings/iio/adc/avia-hx711.txt
 F:	drivers/iio/adc/hx711.c
 
+AWS GRAVITON TRNG DRIVER
+M:	Ali Saidi <[email protected]>
+L:	[email protected] (moderated for non-subscribers)
+S:	Supported
+F:	drivers/char/hw_random/graviton-rng.c
+
 AX.25 NETWORK LAYER
 M:	Ralf Baechle <[email protected]>
 L:	[email protected]
drivers/char/hw_random/Kconfig +13 -0
index 25a7d8ffdb5d..7ca7386a1ac4 100644
@@ -439,6 +439,19 @@ config HW_RANDOM_OPTEE
 
 	  If unsure, say Y.
 
+config HW_RANDOM_GRAVITON
+	tristate "AWS Graviton Random Number Generator support"
+	depends on HW_RANDOM && ACPI && (ARM64 || COMPILE_TEST)
+	default HW_RANDOM
+	help
+	  This driver provides kernel-side support for the Random Number
+	  Generator SMC found on AWS Graviton systems.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called graviton-rng.
+
+	  If unsure, say Y.
+
 endif # HW_RANDOM
 
 config UML_RANDOM
drivers/char/hw_random/Makefile +1 -0
index 7c9ef4a7667f..d1fa72670e98 100644
@@ -39,3 +39,4 @@ obj-$(CONFIG_HW_RANDOM_MTK)	+= mtk-rng.o
 obj-$(CONFIG_HW_RANDOM_S390) += s390-trng.o
 obj-$(CONFIG_HW_RANDOM_KEYSTONE) += ks-sa-rng.o
 obj-$(CONFIG_HW_RANDOM_OPTEE) += optee-rng.o
+obj-$(CONFIG_HW_RANDOM_GRAVITON) += graviton-rng.o
drivers/char/hw_random/graviton-rng.c new file +130 -0

130 lines not on this page · raw message

Signature · 1 line
2.15.3.AMZN
#8 lore raw
In reply to Ali Saidi
Quoted 6 lines · Ali Saidi

AWS Graviton based systems provide an Arm SMC call in the vendor defined hypervisor region to read random numbers from a HW TRNG and return them to the guest.

We've observed slower guest boot and especially reboot times due to lack of entropy and providing access to a TRNG is meant to address this.

Curious, but why this over something like virtio-rng?

Will

author
#9 lore raw

On 6/5/19, 7:20 AM, "Will Deacon" <[email protected]> wrote:

On Tue, Jun 04, 2019 at 08:30:57PM +0000, Ali Saidi wrote:
> AWS Graviton based systems provide an Arm SMC call in the vendor defined > hypervisor region to read random numbers from a HW TRNG and return them to the > guest.
>
> We've observed slower guest boot and especially reboot times due to lack of > entropy and providing access to a TRNG is meant to address this.

Curious, but why this over something like virtio-rng?

This interface allows us to provide the functionality from both EL2 and EL3 and support multiple different types of our instances which we unfortunately can't do with virt-io.

Ali

author
#10 lore raw

On 6/7/19, 7:59 AM, " Ali Saidi" <[email protected]> wrote:

On 6/5/19, 7:20 AM, "Will Deacon" <[email protected]> wrote:

On Tue, Jun 04, 2019 at 08:30:57PM +0000, Ali Saidi wrote:
> AWS Graviton based systems provide an Arm SMC call in the vendor defined > hypervisor region to read random numbers from a HW TRNG and return them to the > guest.
>
> We've observed slower guest boot and especially reboot times due to lack of > entropy and providing access to a TRNG is meant to address this.

Curious, but why this over something like virtio-rng?

This interface allows us to provide the functionality from both EL2 and EL3 and support multiple different types of our instances which we unfortunately can't do with virt-io.

Will,

Any additional comments?

Mark,

Do you know when you'll have a chance to rebase arm64/smccc-cleanup?

Thanks,
Ali

maintainer
#11 lore raw

[Note: this was in my spam folder]

Quoted 10 lines · Ali

On 6/7/19, 7:59 AM, " Ali Saidi" <[email protected]> wrote:
On 6/5/19, 7:20 AM, "Will Deacon" <[email protected]> wrote:
On Tue, Jun 04, 2019 at 08:30:57PM +0000, Ali Saidi wrote:
> AWS Graviton based systems provide an Arm SMC call in the vendor defined > hypervisor region to read random numbers from a HW TRNG and return them to the > guest.
>
> We've observed slower guest boot and especially reboot times due to lack of > entropy and providing access to a TRNG is meant to address this.

Quoted 8 lines

Curious, but why this over something like virtio-rng?

This interface allows us to provide the functionality from both EL2 and EL3 and support multiple different types of our instances which we unfortunately can't do with virt-io.

Any additional comments?
Do you know when you'll have a chance to rebase arm64/smccc-cleanup?

Sorry, Ali, this slipped through the cracks. Marc and I will chat today and look at respinning what we had before; it should then hopefully be straightforward enough for you to take that as a base for what you want to do.

Will

maintainer
#12 lore raw

[+Marc]

Quoted 19 lines · Will Deacon

[Note: this was in my spam folder]

Quoted 16 lines · Ali

On 6/7/19, 7:59 AM, " Ali Saidi" <[email protected]> wrote:
On 6/5/19, 7:20 AM, "Will Deacon" <[email protected]> wrote:
On Tue, Jun 04, 2019 at 08:30:57PM +0000, Ali Saidi wrote:
> AWS Graviton based systems provide an Arm SMC call in the vendor defined > hypervisor region to read random numbers from a HW TRNG and return them to the > guest.
>
> We've observed slower guest boot and especially reboot times due to lack of > entropy and providing access to a TRNG is meant to address this.

Curious, but why this over something like virtio-rng?

This interface allows us to provide the functionality from both EL2 and EL3 and support multiple different types of our instances which we unfortunately can't do with virt-io.

Quoted 7 lines
Quoted 2 lines

Any additional comments?
Do you know when you'll have a chance to rebase arm64/smccc-cleanup?

Sorry, Ali, this slipped through the cracks. Marc and I will chat today and look at respinning what we had before; it should then hopefully be straightforward enough for you to take that as a base for what you want to do.

Ok, I hacked on this a bit today and hopefully you can use this as a starting point:

https://git.kernel.org/pub/scm/linux/kernel/git/will/linux.git/log/?h=kvm/hvc

I haven't given it any real testing, so apologies for the bugs.

Will