On 02/05/2019 18:33, Robin Murphy wrote: > Both Angelo's and your reports strongly imply that the previous > constant-folding debate was a red herring and the trivial fix[1] should > still be sufficient, but nobody's given me actual confirmation of > whether it is or isn't :( Are you saying that when writing to any of gr0_base + ARM_SMMU_GR0_TLBIALLH gr0_base + ARM_SMMU_GR0_TLBIALLNSNH base + ARM_SMMU_GR0_sTLBGSYNC the actual value written does not matter? Is it ignored by the HW? We could write 0xdeadbeef? Regards.
