powerpc/powernv: PCI support for p7IOC under OPAL v2

This adds support for p7IOC (and possibly other IODA v1 IO Hubs)
using OPAL v2 interfaces.

We completely take over resource assignment and assign them using an
algorithm that hands out device BARs in a way that makes them fit in
individual segments of the M32 window of the bridge, which enables us
to assign individual PEs to devices and functions.

The current implementation gives out a PE per functions on PCIe, and a
PE for the entire bridge for PCIe to PCI-X bridges.

This can be adjusted / fine tuned later.

We also setup DMA resources (32-bit only for now) and MSIs (both 32-bit
and 64-bit MSI are supported).

The DMA allocation tries to divide the available 256M segments of the
32-bit DMA address space "fairly" among PEs. This is done using a
"weight" heuristic which assigns less value to things like OHCI USB
controllers than, for example SCSI RAID controllers. This algorithm
will probably want some fine tuning for specific devices or device
types.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

arch/powerpc/include/asm/pci-bridge.h

@@ -153,8 +153,8 @@ struct pci_dn {
 
 	int	pci_ext_config_space;	/* for pci devices */
 
-#ifdef CONFIG_EEH
 	struct	pci_dev *pcidev;	/* back-pointer to the pci device */
+#ifdef CONFIG_EEH
 	int	class_code;		/* pci device class */
 	int	eeh_mode;		/* See eeh.h for possible EEH_MODEs */
 	int	eeh_config_addr;
@@ -164,6 +164,10 @@ struct pci_dn {
 	int	eeh_false_positives;	/* # times this device reported #ff's */
 	u32	config_space[16];	/* saved PCI config space */
 #endif
+#define IODA_INVALID_PE		(-1)
+#ifdef CONFIG_PPC_POWERNV
+	int	pe_number;
+#endif
 };
 
 /* Get the pointer to a device_node's pci_dn */

arch/powerpc/kernel/pci_dn.c

@@ -50,6 +50,9 @@ void * __devinit update_dn_pci_info(struct device_node *dn, void *data)
 	dn->data = pdn;
 	pdn->node = dn;
 	pdn->phb = phb;
+#ifdef CONFIG_PPC_POWERNV
+	pdn->pe_number = IODA_INVALID_PE;
+#endif
 	regs = of_get_property(dn, "reg", NULL);
 	if (regs) {
 		/* First register entry is addr (00BBSS00)  */

arch/powerpc/platforms/powernv/Makefile

@@ -2,4 +2,4 @@ obj-y			+= setup.o opal-takeover.o opal-wrappers.o opal.o
 obj-y			+= opal-rtc.o opal-nvram.o
 
 obj-$(CONFIG_SMP)	+= smp.o
-obj-$(CONFIG_PCI)	+= pci.o pci-p5ioc2.o
+obj-$(CONFIG_PCI)	+= pci.o pci-p5ioc2.o pci-ioda.o

arch/powerpc/platforms/powernv/pci.c

@@ -467,10 +467,22 @@ void __init pnv_pci_init(void)
 		init_pci_config_tokens();
 		find_and_init_phbs();
 #endif /* CONFIG_PPC_POWERNV_RTAS */
-	} else {
+	}
 	/* OPAL is here, do our normal stuff */
+	else {
+		int found_ioda = 0;
+
+		/* Look for IODA IO-Hubs. We don't support mixing IODA
+		 * and p5ioc2 due to the need to change some global
+		 * probing flags
+		 */
+		for_each_compatible_node(np, NULL, "ibm,ioda-hub") {
+			pnv_pci_init_ioda_hub(np);
+			found_ioda = 1;
+		}
 
 		/* Look for p5ioc2 IO-Hubs */
+		if (!found_ioda)
 			for_each_compatible_node(np, NULL, "ibm,p5ioc2")
 				pnv_pci_init_p5ioc2_hub(np);
 	}

arch/powerpc/platforms/powernv/pci.h

@@ -9,6 +9,50 @@ enum pnv_phb_type {
 	PNV_PHB_IODA2,
 };
 
+/* Data associated with a PE, including IOMMU tracking etc.. */
+struct pnv_ioda_pe {
+	/* A PE can be associated with a single device or an
+	 * entire bus (& children). In the former case, pdev
+	 * is populated, in the later case, pbus is.
+	 */
+	struct pci_dev		*pdev;
+	struct pci_bus		*pbus;
+
+	/* Effective RID (device RID for a device PE and base bus
+	 * RID with devfn 0 for a bus PE)
+	 */
+	unsigned int		rid;
+
+	/* PE number */
+	unsigned int		pe_number;
+
+	/* "Weight" assigned to the PE for the sake of DMA resource
+	 * allocations
+	 */
+	unsigned int		dma_weight;
+
+	/* This is a PCI-E -> PCI-X bridge, this points to the
+	 * corresponding bus PE
+	 */
+	struct pnv_ioda_pe	*bus_pe;
+
+	/* "Base" iommu table, ie, 4K TCEs, 32-bit DMA */
+	int			tce32_seg;
+	int			tce32_segcount;
+	struct iommu_table	tce32_table;
+
+	/* XXX TODO: Add support for additional 64-bit iommus */
+
+	/* MSIs. MVE index is identical for for 32 and 64 bit MSI
+	 * and -1 if not supported. (It's actually identical to the
+	 * PE number)
+	 */
+	int			mve_number;
+
+	/* Link in list of PE#s */
+	struct list_head	link;
+};
+
 struct pnv_phb {
 	struct pci_controller	*hose;
 	enum pnv_phb_type	type;
@@ -34,6 +78,45 @@ struct pnv_phb {
 		struct {
 			struct iommu_table iommu_table;
 		} p5ioc2;
+
+		struct {
+			/* Global bridge info */
+			unsigned int		total_pe;
+			unsigned int		m32_size;
+			unsigned int		m32_segsize;
+			unsigned int		m32_pci_base;
+			unsigned int		io_size;
+			unsigned int		io_segsize;
+			unsigned int		io_pci_base;
+
+			/* PE allocation bitmap */
+			unsigned long		*pe_alloc;
+
+			/* M32 & IO segment maps */
+			unsigned int		*m32_segmap;
+			unsigned int		*io_segmap;
+			struct pnv_ioda_pe	*pe_array;
+
+			/* Reverse map of PEs, will have to extend if
+			 * we are to support more than 256 PEs, indexed
+			 * bus { bus, devfn }
+			 */
+			unsigned char		pe_rmap[0x10000];
+
+			/* 32-bit TCE tables allocation */
+			unsigned long		tce32_count;
+
+			/* Total "weight" for the sake of DMA resources
+			 * allocation
+			 */
+			unsigned int		dma_weight;
+			unsigned int		dma_pe_count;
+
+			/* Sorted list of used PE's, sorted at
+			 * boot for resource allocation purposes
+			 */
+			struct list_head	pe_list;
+		} ioda;
 	};
 };
 
@@ -43,6 +126,7 @@ extern void pnv_pci_setup_iommu_table(struct iommu_table *tbl,
 				      void *tce_mem, u64 tce_size,
 				      u64 dma_offset);
 extern void pnv_pci_init_p5ioc2_hub(struct device_node *np);
+extern void pnv_pci_init_ioda_hub(struct device_node *np);
 
 
 #endif /* __POWERNV_PCI_H */