powerpc/powernv: Fix iommu initialization again

So because those things always end up in trainwrecks... In 7846de4
we moved back the iommu initialization earlier, essentially undoing
37f0219 which was causing us endless trouble... except that in the
meantime we had merged 959c9bd (to workaround the original breakage)
which is now ... broken :-)

This fixes it by doing a partial revert of the latter (we keep the
ppc_md. path which will be needed in the hotplug case, which happens
also during some EEH error recovery situations).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: <stable@vger.kernel.org> [v3.10]

arch/powerpc/platforms/powernv/pci-ioda.c

@@ -443,6 +443,17 @@ static void pnv_pci_ioda_dma_dev_setup(struct pnv_phb *phb, struct pci_dev *pdev
 	set_iommu_table_base(&pdev->dev, &pe->tce32_table);
 }
 
+static void pnv_ioda_setup_bus_dma(struct pnv_ioda_pe *pe, struct pci_bus *bus)
+{
+	struct pci_dev *dev;
+
+	list_for_each_entry(dev, &bus->devices, bus_list) {
+		set_iommu_table_base(&dev->dev, &pe->tce32_table);
+		if (dev->subordinate)
+			pnv_ioda_setup_bus_dma(pe, dev->subordinate);
+	}
+}
+
 static void pnv_pci_ioda1_tce_invalidate(struct iommu_table *tbl,
 					 u64 *startp, u64 *endp)
 {
@@ -599,6 +610,11 @@ static void pnv_pci_ioda_setup_dma_pe(struct pnv_phb *phb,
 	iommu_init_table(tbl, phb->hose->node);
 	iommu_register_group(tbl, pci_domain_nr(pe->pbus), pe->pe_number);
 
+	if (pe->pdev)
+		set_iommu_table_base(&pe->pdev->dev, tbl);
+	else
+		pnv_ioda_setup_bus_dma(pe, pe->pbus);
+
 	return;
  fail:
 	/* XXX Failure: Try to fallback to 64-bit only ? */
@@ -670,6 +686,11 @@ static void pnv_pci_ioda2_setup_dma_pe(struct pnv_phb *phb,
 	}
 	iommu_init_table(tbl, phb->hose->node);
 
+	if (pe->pdev)
+		set_iommu_table_base(&pe->pdev->dev, tbl);
+	else
+		pnv_ioda_setup_bus_dma(pe, pe->pbus);
+
 	return;
 fail:
 	if (pe->tce32_seg >= 0)