[SPARC64]: Beginnings of SUN4V PCI controller support.

Abstract out IOMMU operations so that we can have a different
set of calls on sun4v, which needs to do things through
hypervisor calls.
Signed-off-by: David S. Miller <davem@davemloft.net>

arch/sparc64/kernel/Makefile

@@ -14,7 +14,7 @@ obj-y		:= process.o setup.o cpu.o idprom.o \
 		   power.o sbus.o iommu_common.o sparc64_ksyms.o chmc.o
 
 obj-$(CONFIG_PCI)	 += ebus.o isa.o pci_common.o pci_iommu.o \
-			    pci_psycho.o pci_sabre.o pci_schizo.o
+			    pci_psycho.o pci_sabre.o pci_schizo.o pci_sun4v.o
 obj-$(CONFIG_SMP)	 += smp.o trampoline.o
 obj-$(CONFIG_SPARC32_COMPAT) += sys32.o sys_sparc32.o signal32.o
 obj-$(CONFIG_BINFMT_ELF32) += binfmt_elf32.o

arch/sparc64/kernel/pci.c

@@ -188,6 +188,7 @@ extern void psycho_init(int, char *);
 extern void schizo_init(int, char *);
 extern void schizo_plus_init(int, char *);
 extern void tomatillo_init(int, char *);
+extern void sun4v_pci_init(int, char *);
 
 static struct {
 	char *model_name;
@@ -204,6 +205,7 @@ static struct {
 	{ "pci108e,8002", schizo_plus_init },
 	{ "SUNW,tomatillo", tomatillo_init },
 	{ "pci108e,a801", tomatillo_init },
+	{ "SUNW,sun4v-pci", sun4v_pci_init },
 };
 #define PCI_NUM_CONTROLLER_TYPES (sizeof(pci_controller_table) / \
 				  sizeof(pci_controller_table[0]))
@@ -283,6 +285,12 @@ int __init pcic_present(void)
 	return pci_controller_scan(pci_is_controller);
 }
 
+struct pci_iommu_ops *pci_iommu_ops;
+EXPORT_SYMBOL(pci_iommu_ops);
+
+extern struct pci_iommu_ops pci_sun4u_iommu_ops,
+	pci_sun4v_iommu_ops;
+
 /* Find each controller in the system, attach and initialize
  * software state structure for each and link into the
  * pci_controller_root.  Setup the controller enough such
@@ -290,6 +298,11 @@ int __init pcic_present(void)
  */
 static void __init pci_controller_probe(void)
 {
+	if (tlb_type == hypervisor)
+		pci_iommu_ops = &pci_sun4v_iommu_ops;
+	else
+		pci_iommu_ops = &pci_sun4u_iommu_ops;
+
 	printk("PCI: Probing for controllers.\n");
 
 	pci_controller_scan(pci_controller_init);

arch/sparc64/kernel/pci_iommu.c

@@ -219,7 +219,7 @@ static inline void iommu_free_ctx(struct pci_iommu *iommu, int ctx)
  * DMA for PCI device PDEV.  Return non-NULL cpu-side address if
  * successful and set *DMA_ADDRP to the PCI side dma address.
  */
-void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_addrp)
+static void *pci_4u_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_addrp)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -267,7 +267,7 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad
 }
 
 /* Free and unmap a consistent DMA translation. */
-void pci_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_t dvma)
+static void pci_4u_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_t dvma)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -294,7 +294,7 @@ void pci_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_
 /* Map a single buffer at PTR of SZ bytes for PCI DMA
  * in streaming mode.
  */
-dma_addr_t pci_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direction)
+static dma_addr_t pci_4u_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direction)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -415,7 +415,7 @@ static void pci_strbuf_flush(struct pci_strbuf *strbuf, struct pci_iommu *iommu,
 }
 
 /* Unmap a single streaming mode DMA translation. */
-void pci_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int direction)
+static void pci_4u_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int direction)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -548,7 +548,7 @@ static inline void fill_sg(iopte_t *iopte, struct scatterlist *sg,
  * When making changes here, inspect the assembly output. I was having
  * hard time to kepp this routine out of using stack slots for holding variables.
  */
-int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
+static int pci_4u_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -635,7 +635,7 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int
 }
 
 /* Unmap a set of streaming mode DMA translations. */
-void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
+static void pci_4u_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -695,7 +695,7 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems,
 /* Make physical memory consistent for a single
  * streaming mode DMA translation after a transfer.
  */
-void pci_dma_sync_single_for_cpu(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int direction)
+static void pci_4u_dma_sync_single_for_cpu(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int direction)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -735,7 +735,7 @@ void pci_dma_sync_single_for_cpu(struct pci_dev *pdev, dma_addr_t bus_addr, size
 /* Make physical memory consistent for a set of streaming
  * mode DMA translations after a transfer.
  */
-void pci_dma_sync_sg_for_cpu(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
+static void pci_4u_dma_sync_sg_for_cpu(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
 {
 	struct pcidev_cookie *pcp;
 	struct pci_iommu *iommu;
@@ -776,6 +776,17 @@ void pci_dma_sync_sg_for_cpu(struct pci_dev *pdev, struct scatterlist *sglist, i
 	spin_unlock_irqrestore(&iommu->lock, flags);
 }
 
+struct pci_iommu_ops pci_sun4u_iommu_ops = {
+	.alloc_consistent		= pci_4u_alloc_consistent,
+	.free_consistent		= pci_4u_free_consistent,
+	.map_single			= pci_4u_map_single,
+	.unmap_single			= pci_4u_unmap_single,
+	.map_sg				= pci_4u_map_sg,
+	.unmap_sg			= pci_4u_unmap_sg,
+	.dma_sync_single_for_cpu	= pci_4u_dma_sync_single_for_cpu,
+	.dma_sync_sg_for_cpu		= pci_4u_dma_sync_sg_for_cpu,
+};
+
 static void ali_sound_dma_hack(struct pci_dev *pdev, int set_bit)
 {
 	struct pci_dev *ali_isa_bridge;

arch/sparc64/kernel/pci_sun4v.c

+/* pci_sun4v.c: SUN4V specific PCI controller support.
+ *
+ * Copyright (C) 2006 David S. Miller (davem@davemloft.net)
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/interrupt.h>
+
+#include <asm/pbm.h>
+#include <asm/iommu.h>
+#include <asm/irq.h>
+#include <asm/upa.h>
+#include <asm/pstate.h>
+#include <asm/oplib.h>
+#include <asm/hypervisor.h>
+
+#include "pci_impl.h"
+#include "iommu_common.h"
+
+static void *pci_4v_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_addrp)
+{
+	return NULL;
+}
+
+static void pci_4v_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_t dvma)
+{
+}
+
+static dma_addr_t pci_4v_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direction)
+{
+	return 0;
+}
+
+static void pci_4v_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int direction)
+{
+}
+
+static int pci_4v_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
+{
+	return nelems;
+}
+
+static void pci_4v_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
+{
+}
+
+static void pci_4v_dma_sync_single_for_cpu(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int direction)
+{
+}
+
+static void pci_4v_dma_sync_sg_for_cpu(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int direction)
+{
+}
+
+struct pci_iommu_ops pci_sun4v_iommu_ops = {
+	.alloc_consistent		= pci_4v_alloc_consistent,
+	.free_consistent		= pci_4v_free_consistent,
+	.map_single			= pci_4v_map_single,
+	.unmap_single			= pci_4v_unmap_single,
+	.map_sg				= pci_4v_map_sg,
+	.unmap_sg			= pci_4v_unmap_sg,
+	.dma_sync_single_for_cpu	= pci_4v_dma_sync_single_for_cpu,
+	.dma_sync_sg_for_cpu		= pci_4v_dma_sync_sg_for_cpu,
+};
+
+void sun4v_pci_init(int node, char *model_name)
+{
+	prom_printf("sun4v_pci_init: Implement me.\n");
+	prom_halt();
+}

include/asm-sparc64/pci.h

@@ -41,10 +41,26 @@ static inline void pcibios_penalize_isa_irq(int irq, int active)
 
 struct pci_dev;
 
+struct pci_iommu_ops {
+	void *(*alloc_consistent)(struct pci_dev *, size_t, dma_addr_t *);
+	void (*free_consistent)(struct pci_dev *, size_t, void *, dma_addr_t);
+	dma_addr_t (*map_single)(struct pci_dev *, void *, size_t, int);
+	void (*unmap_single)(struct pci_dev *, dma_addr_t, size_t, int);
+	int (*map_sg)(struct pci_dev *, struct scatterlist *, int, int);
+	void (*unmap_sg)(struct pci_dev *, struct scatterlist *, int, int);
+	void (*dma_sync_single_for_cpu)(struct pci_dev *, dma_addr_t, size_t, int);
+	void (*dma_sync_sg_for_cpu)(struct pci_dev *, struct scatterlist *, int, int);
+};
+
+extern struct pci_iommu_ops *pci_iommu_ops;
+
 /* Allocate and map kernel buffer using consistent mode DMA for a device.
  * hwdev should be valid struct pci_dev pointer for PCI devices.
  */
-extern void *pci_alloc_consistent(struct pci_dev *hwdev, size_t size, dma_addr_t *dma_handle);
+static inline void *pci_alloc_consistent(struct pci_dev *hwdev, size_t size, dma_addr_t *dma_handle)
+{
+	return pci_iommu_ops->alloc_consistent(hwdev, size, dma_handle);
+}
 
 /* Free and unmap a consistent DMA buffer.
  * cpu_addr is what was returned from pci_alloc_consistent,
@@ -54,7 +70,10 @@ extern void *pci_alloc_consistent(struct pci_dev *hwdev, size_t size, dma_addr_t
  * References to the memory and mappings associated with cpu_addr/dma_addr
  * past this call are illegal.
  */
-extern void pci_free_consistent(struct pci_dev *hwdev, size_t size, void *vaddr, dma_addr_t dma_handle);
+static inline void pci_free_consistent(struct pci_dev *hwdev, size_t size, void *vaddr, dma_addr_t dma_handle)
+{
+	return pci_iommu_ops->free_consistent(hwdev, size, vaddr, dma_handle);
+}
 
 /* Map a single buffer of the indicated size for DMA in streaming mode.
  * The 32-bit bus address to use is returned.
@@ -62,7 +81,10 @@ extern void pci_free_consistent(struct pci_dev *hwdev, size_t size, void *vaddr,
  * Once the device is given the dma address, the device owns this memory
  * until either pci_unmap_single or pci_dma_sync_single_for_cpu is performed.
  */
-extern dma_addr_t pci_map_single(struct pci_dev *hwdev, void *ptr, size_t size, int direction);
+static inline dma_addr_t pci_map_single(struct pci_dev *hwdev, void *ptr, size_t size, int direction)
+{
+	return pci_iommu_ops->map_single(hwdev, ptr, size, direction);
+}
 
 /* Unmap a single streaming mode DMA translation.  The dma_addr and size
  * must match what was provided for in a previous pci_map_single call.  All
@@ -71,7 +93,10 @@ extern dma_addr_t pci_map_single(struct pci_dev *hwdev, void *ptr, size_t size,
  * After this call, reads by the cpu to the buffer are guaranteed to see
  * whatever the device wrote there.
  */
-extern void pci_unmap_single(struct pci_dev *hwdev, dma_addr_t dma_addr, size_t size, int direction);
+static inline void pci_unmap_single(struct pci_dev *hwdev, dma_addr_t dma_addr, size_t size, int direction)
+{
+	pci_iommu_ops->unmap_single(hwdev, dma_addr, size, direction);
+}
 
 /* No highmem on sparc64, plus we have an IOMMU, so mapping pages is easy. */
 #define pci_map_page(dev, page, off, size, dir) \
@@ -107,15 +132,19 @@ extern void pci_unmap_single(struct pci_dev *hwdev, dma_addr_t dma_addr, size_t
  * Device ownership issues as mentioned above for pci_map_single are
  * the same here.
  */
-extern int pci_map_sg(struct pci_dev *hwdev, struct scatterlist *sg,
-		      int nents, int direction);
+static inline int pci_map_sg(struct pci_dev *hwdev, struct scatterlist *sg, int nents, int direction)
+{
+	return pci_iommu_ops->map_sg(hwdev, sg, nents, direction);
+}
 
 /* Unmap a set of streaming mode DMA translations.
  * Again, cpu read rules concerning calls here are the same as for
  * pci_unmap_single() above.
  */
-extern void pci_unmap_sg(struct pci_dev *hwdev, struct scatterlist *sg,
-			 int nhwents, int direction);
+static inline void pci_unmap_sg(struct pci_dev *hwdev, struct scatterlist *sg, int nhwents, int direction)
+{
+	pci_iommu_ops->unmap_sg(hwdev, sg, nhwents, direction);
+}
 
 /* Make physical memory consistent for a single
  * streaming mode DMA translation after a transfer.
@@ -127,8 +156,10 @@ extern void pci_unmap_sg(struct pci_dev *hwdev, struct scatterlist *sg,
  * must first perform a pci_dma_sync_for_device, and then the
  * device again owns the buffer.
  */
-extern void pci_dma_sync_single_for_cpu(struct pci_dev *hwdev, dma_addr_t dma_handle,
-					size_t size, int direction);
+static inline void pci_dma_sync_single_for_cpu(struct pci_dev *hwdev, dma_addr_t dma_handle, size_t size, int direction)
+{
+	pci_iommu_ops->dma_sync_single_for_cpu(hwdev, dma_handle, size, direction);
+}
 
 static inline void
 pci_dma_sync_single_for_device(struct pci_dev *hwdev, dma_addr_t dma_handle,
@@ -144,7 +175,10 @@ pci_dma_sync_single_for_device(struct pci_dev *hwdev, dma_addr_t dma_handle,
  * The same as pci_dma_sync_single_* but for a scatter-gather list,
  * same rules and usage.
  */
-extern void pci_dma_sync_sg_for_cpu(struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction);
+static inline void pci_dma_sync_sg_for_cpu(struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
+{
+	pci_iommu_ops->dma_sync_sg_for_cpu(hwdev, sg, nelems, direction);
+}
 
 static inline void
 pci_dma_sync_sg_for_device(struct pci_dev *hwdev, struct scatterlist *sg,