linux-kernel - [PATCH 2/8] iommu/arm-smmu-qcom: Add QC SMMUv2 VA Size quirk for SDM660

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sat, 26 Sep 2020 14:59:58 +0200
From:   kholk11@...il.com
To:     will@...nel.org
Cc:     robin.murphy@....com, joro@...tes.org, bjorn.andersson@...aro.org,
        linux-arm-kernel@...ts.infradead.org, devicetree@...r.kernel.org,
        kholk11@...il.com, marijns95@...il.com, konradybcio@...il.com,
        martin.botka1@...il.com, linux-arm-msm@...r.kernel.org,
        phone-devel@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: [PATCH 2/8] iommu/arm-smmu-qcom: Add QC SMMUv2 VA Size quirk for SDM660

From: AngeloGioacchino Del Regno <kholk11@...il.com>

Some IOMMUs are getting set-up for Shared Virtual Address, but:
1. They are secured by the Hypervisor, so any configuration
   change will generate a hyp-fault and crash the system
2. This 39-bits Virtual Address size deviates from the ARM
   System MMU Architecture specification for SMMUv2, hence
   it is non-standard. In this case, the only way to keep the
   IOMMU as the firmware did configure it, is to hardcode a
   maximum VA size of 39 bits (because of point 1).

This gives the need to add implementation details bits for
at least some of the SoCs having this kind of configuration,
which are at least SDM630, SDM636 and SDM660.

These implementation details will be enabled on finding the
qcom,sdm660-smmu-v2 compatible.

Signed-off-by: AngeloGioacchino Del Regno <kholk11@...il.com>
---
 drivers/iommu/arm/arm-smmu/arm-smmu-impl.c |  3 ++-
 drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c | 31 +++++++++++++++++++++-
 2 files changed, 32 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-impl.c b/drivers/iommu/arm/arm-smmu/arm-smmu-impl.c
index f4ff124a1967..9d753f8af2cc 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu-impl.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu-impl.c
@@ -216,7 +216,8 @@ struct arm_smmu_device *arm_smmu_impl_init(struct arm_smmu_device *smmu)
 	if (of_device_is_compatible(np, "nvidia,tegra194-smmu"))
 		return nvidia_smmu_impl_init(smmu);
 
-	if (of_device_is_compatible(np, "qcom,sdm845-smmu-500") ||
+	if (of_device_is_compatible(np, "qcom,sdm660-smmu-v2") ||
+	    of_device_is_compatible(np, "qcom,sdm845-smmu-500") ||
 	    of_device_is_compatible(np, "qcom,sc7180-smmu-500") ||
 	    of_device_is_compatible(np, "qcom,sm8150-smmu-500") ||
 	    of_device_is_compatible(np, "qcom,sm8250-smmu-500"))
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
index 7859fd0db22a..f5bbfe86ef30 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
@@ -65,8 +65,33 @@ static const struct arm_smmu_impl qcom_smmu500_impl = {
 	.reset = qcom_smmu500_reset,
 };
 
+static int qcom_smmuv2_cfg_probe(struct arm_smmu_device *smmu)
+{
+	/*
+	 * Some IOMMUs are getting set-up for Shared Virtual Address, but:
+	 * 1. They are secured by the Hypervisor, so any configuration
+	 *    change will generate a hyp-fault and crash the system
+	 * 2. This 39-bits Virtual Address size deviates from the ARM
+	 *    System MMU Architecture specification for SMMUv2, hence
+	 *    it is non-standard. In this case, the only way to keep the
+	 *    IOMMU as the firmware did configure it, is to hardcode a
+	 *    maximum VA size of 39 bits (because of point 1).
+	 */
+	if (smmu->va_size > 39UL)
+		dev_notice(smmu->dev,
+			   "\tenabling workaround for QCOM SMMUv2 VA size\n");
+	smmu->va_size = min(smmu->va_size, 39UL);
+
+	return 0;
+}
+
+static const struct arm_smmu_impl qcom_smmuv2_impl = {
+	.cfg_probe = qcom_smmuv2_cfg_probe,
+};
+
 struct arm_smmu_device *qcom_smmu_impl_init(struct arm_smmu_device *smmu)
 {
+	const struct device_node *np = smmu->dev->of_node;
 	struct qcom_smmu *qsmmu;
 
 	qsmmu = devm_kzalloc(smmu->dev, sizeof(*qsmmu), GFP_KERNEL);
@@ -75,7 +100,11 @@ struct arm_smmu_device *qcom_smmu_impl_init(struct arm_smmu_device *smmu)
 
 	qsmmu->smmu = *smmu;
 
-	qsmmu->smmu.impl = &qcom_smmu500_impl;
+	if (of_device_is_compatible(np, "qcom,sdm660-smmu-v2")) {
+		qsmmu->smmu.impl = &qcom_smmuv2_impl;
+	} else {
+		qsmmu->smmu.impl = &qcom_smmu500_impl;
+	}
 	devm_kfree(smmu->dev, smmu);
 
 	return &qsmmu->smmu;
-- 
2.28.0