The tegra_smmu_probe_device() handles only the first IOMMU device-tree
phandle, skipping the rest. Devices like 3D module on Tegra30 have
multiple IOMMU phandles, one for each h/w block, and thus, only one
IOMMU phandle is added to fwspec for the 3D module, breaking GPU.
Previously this problem was masked by tegra_smmu_attach_dev() which
didn't use the fwspec, but parsed the DT by itself. The previous commit
to tegra-smmu driver partially reverted changes that caused problems for
T124 and now we have tegra_smmu_attach_dev() that uses the fwspec and
the old-buggy variant of tegra_smmu_probe_device() which skips secondary
IOMMUs.
Make tegra_smmu_probe_device() not to skip the secondary IOMMUs. This
fixes a partially attached IOMMU of the 3D module on Tegra30 and now GPU
works properly once again.
Fixes: 765a9d1d02b2 ("iommu/tegra-smmu: Fix mc errors on tegra124-nyan")
Signed-off-by: Dmitry Osipenko <[email protected]>
---
drivers/iommu/tegra-smmu.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/iommu/tegra-smmu.c b/drivers/iommu/tegra-smmu.c
index 97eb62f667d2..602aab98c079 100644
--- a/drivers/iommu/tegra-smmu.c
+++ b/drivers/iommu/tegra-smmu.c
@@ -849,12 +849,11 @@ static struct iommu_device *tegra_smmu_probe_device(struct device *dev)
smmu = tegra_smmu_find(args.np);
if (smmu) {
err = tegra_smmu_configure(smmu, dev, &args);
- of_node_put(args.np);
- if (err < 0)
+ if (err < 0) {
+ of_node_put(args.np);
return ERR_PTR(err);
-
- break;
+ }
}
of_node_put(args.np);
--
2.29.2
On Fri, Mar 12, 2021 at 06:54:39PM +0300, Dmitry Osipenko wrote:
> The tegra_smmu_probe_device() handles only the first IOMMU device-tree
> phandle, skipping the rest. Devices like 3D module on Tegra30 have
> multiple IOMMU phandles, one for each h/w block, and thus, only one
> IOMMU phandle is added to fwspec for the 3D module, breaking GPU.
> Previously this problem was masked by tegra_smmu_attach_dev() which
> didn't use the fwspec, but parsed the DT by itself. The previous commit
> to tegra-smmu driver partially reverted changes that caused problems for
> T124 and now we have tegra_smmu_attach_dev() that uses the fwspec and
> the old-buggy variant of tegra_smmu_probe_device() which skips secondary
> IOMMUs.
>
> Make tegra_smmu_probe_device() not to skip the secondary IOMMUs. This
> fixes a partially attached IOMMU of the 3D module on Tegra30 and now GPU
> works properly once again.
>
> Fixes: 765a9d1d02b2 ("iommu/tegra-smmu: Fix mc errors on tegra124-nyan")
> Signed-off-by: Dmitry Osipenko <[email protected]>
Tested-by: Nicolin Chen <[email protected]>
On Fri, Mar 12, 2021 at 06:54:39PM +0300, Dmitry Osipenko wrote:
> The tegra_smmu_probe_device() handles only the first IOMMU device-tree
> phandle, skipping the rest. Devices like 3D module on Tegra30 have
> multiple IOMMU phandles, one for each h/w block, and thus, only one
> IOMMU phandle is added to fwspec for the 3D module, breaking GPU.
> Previously this problem was masked by tegra_smmu_attach_dev() which
> didn't use the fwspec, but parsed the DT by itself. The previous commit
> to tegra-smmu driver partially reverted changes that caused problems for
> T124 and now we have tegra_smmu_attach_dev() that uses the fwspec and
> the old-buggy variant of tegra_smmu_probe_device() which skips secondary
> IOMMUs.
>
> Make tegra_smmu_probe_device() not to skip the secondary IOMMUs. This
> fixes a partially attached IOMMU of the 3D module on Tegra30 and now GPU
> works properly once again.
>
> Fixes: 765a9d1d02b2 ("iommu/tegra-smmu: Fix mc errors on tegra124-nyan")
> Signed-off-by: Dmitry Osipenko <[email protected]>
> ---
> drivers/iommu/tegra-smmu.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
Applied for v5.12, thanks.