On 2020-09-04 16:55, Bjorn Andersson wrote:
With many Qualcomm platforms not having functional S2CR BYPASS a
temporary IOMMU domain, without translation, needs to be allocated in
order to allow these memory transactions.
Unfortunately the boot loader uses the first few context banks, so
rather than overwriting a active bank the last context bank is used and
streams are diverted here during initialization.
This also performs the readback of SMR registers for the Qualcomm
platform, to trigger the mechanism.
This is based on prior work by Thierry Reding and Laurentiu Tudor.
Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>
---
Changes since v2:
- Combined from pieces spread between the Qualcomm impl and generic code in v2.
- Moved to use the newly introduced inherit_mapping op.
drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c | 33 ++++++++++++++++++++++
1 file changed, 33 insertions(+)
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
index 70a1eaa52e14..a54302190932 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
@@ -12,6 +12,7 @@
struct qcom_smmu {
struct arm_smmu_device smmu;
bool bypass_broken;
+ struct iommu_domain *identity;
};
static struct qcom_smmu *to_qcom_smmu(struct arm_smmu_device *smmu)
@@ -228,6 +229,37 @@ static int qcom_smmu_cfg_probe(struct arm_smmu_device *smmu)
return 0;
}
+static int qcom_smmu_inherit_mappings(struct arm_smmu_device *smmu)
+{
+ struct qcom_smmu *qsmmu = to_qcom_smmu(smmu);
+ int cbndx;
+ u32 smr;
+ int i;
+
+ qsmmu->identity = arm_smmu_alloc_identity_domain(smmu);
+ if (IS_ERR(qsmmu->identity))
+ return PTR_ERR(qsmmu->identity);
+
+ cbndx = to_smmu_domain(qsmmu->identity)->cfg.cbndx;
I don't really get the point of going through the dance of allocating a
whole iommu_domain() just to get a context. If you don't want to simply
statically reserve a context at probe time, then just allocate from
smmu->context_map here (where AFAICS "here" should be in cfg_probe
anyway). This is entirely driver-internal, so there shouldn't be any
need for IOMMU-API-level stuff to be involved.
Robin.
+
+ for (i = 0; i < smmu->num_mapping_groups; i++) {
+ smr = arm_smmu_gr0_read(smmu, ARM_SMMU_GR0_SMR(i));
+
+ if (FIELD_GET(ARM_SMMU_SMR_VALID, smr)) {
+ smmu->smrs[i].id = FIELD_GET(ARM_SMMU_SMR_ID, smr);
+ smmu->smrs[i].mask = FIELD_GET(ARM_SMMU_SMR_MASK, smr);
+ smmu->smrs[i].valid = true;
+
+ smmu->s2crs[i].type = S2CR_TYPE_TRANS;
+ smmu->s2crs[i].privcfg = S2CR_PRIVCFG_DEFAULT;
+ smmu->s2crs[i].cbndx = cbndx;
+ smmu->s2crs[i].count++;
+ }
+ }
+
+ return 0;
+}
+
static int qcom_smmu_def_domain_type(struct device *dev)
{
const struct of_device_id *match =
@@ -270,6 +302,7 @@ static const struct arm_smmu_impl qcom_smmu_impl = {
.cfg_probe = qcom_smmu_cfg_probe,
.def_domain_type = qcom_smmu_def_domain_type,
.reset = qcom_smmu500_reset,
+ .inherit_mappings = qcom_smmu_inherit_mappings,
};
static const struct arm_smmu_impl qcom_adreno_smmu_impl = {
