Re: [PATCH] iommu/amd: Fix event counter availability check

[Suravee Suthikulpanit]

Hi Alexander,

On 5/30/20 3:07 AM, Alexander Monakov wrote:
> The driver performs an extra check if the IOMMU's capabilities advertise
> presence of performance counters: it verifies that counters are writable
> by writing a hard-coded value to a counter and testing that reading that
> counter gives back the same value.
>
> Unfortunately it does so quite early, even before pci_enable_device is
> called for the IOMMU, i.e. when accessing its MMIO space is not
> guaranteed to work. On Ryzen 4500U CPU, this actually breaks the test:
> the driver assumes the counters are not writable, and disables the
> functionality.
>
> Moving init_iommu_perf_ctr just after iommu_flush_all_caches resolves
> the issue. This is the earliest point in amd_iommu_init_pci where the
> call succeeds on my laptop.

According to your description, it should just need to be anywhere after the pci_enable_device() is called for the IOMMU 
device, isn't it? So, on your system, what if we just move the init_iommu_perf_ctr() here:

diff --git a/drivers/iommu/amd_iommu_init.c b/drivers/iommu/amd_iommu_init.c
index 5b81fd16f5fa..17b9ac9491e0 100644
--- a/drivers/iommu/amd_iommu_init.c
+++ b/drivers/iommu/amd_iommu_init.c
@@ -1875,6 +1875,7 @@ static int __init amd_iommu_init_pci(void)
                ret = iommu_init_pci(iommu);
                if (ret)
                        break;
+               init_iommu_perf_ctr(iommu);
        }

        /*
--
2.17.1


Does this works?

Thanks,
Suravee