drivers/char/agp/nvidia-agp.c - maze/linux - Git at Google

 /*
  * Nvidia AGPGART routines.
  * Based upon a 2.4 agpgart diff by the folks from NVIDIA, and hacked up
  * to work in 2.5 by Dave Jones <davej@codemonkey.org.uk>
  */

 #include <linux/module.h>
 #include <linux/pci.h>
 #include <linux/init.h>
 #include <linux/agp_backend.h>
 #include <linux/gfp.h>
 #include <linux/page-flags.h>
 #include <linux/mm.h>
 #include <linux/jiffies.h>
 #include "agp.h"

 /* NVIDIA registers */
 #define NVIDIA_0_APSIZE		0x80
 #define NVIDIA_1_WBC		0xf0
 #define NVIDIA_2_GARTCTRL	0xd0
 #define NVIDIA_2_APBASE		0xd8
 #define NVIDIA_2_APLIMIT	0xdc
 #define NVIDIA_2_ATTBASE(i)	(0xe0 + (i) * 4)
 #define NVIDIA_3_APBASE		0x50
 #define NVIDIA_3_APLIMIT	0x54


 static struct _nvidia_private {
 	struct pci_dev *dev_1;
 	struct pci_dev *dev_2;
 	struct pci_dev *dev_3;
 	volatile u32 __iomem *aperture;
 	int num_active_entries;
 	off_t pg_offset;
 	u32 wbc_mask;
 } nvidia_private;


 static int nvidia_fetch_size(void)
 {
 	int i;
 	u8 size_value;
 	struct aper_size_info_8 *values;

 	pci_read_config_byte(agp_bridge->dev, NVIDIA_0_APSIZE, &size_value);
 	size_value &= 0x0f;
 	values = A_SIZE_8(agp_bridge->driver->aperture_sizes);

 	for (i = 0; i < agp_bridge->driver->num_aperture_sizes; i++) {
 		if (size_value == values[i].size_value) {
 			agp_bridge->previous_size =
 				agp_bridge->current_size = (void *) (values + i);
 			agp_bridge->aperture_size_idx = i;
 			return values[i].size;
 		}
 	}

 	return 0;
 }

 #define SYSCFG          0xC0010010
 #define IORR_BASE0      0xC0010016
 #define IORR_MASK0      0xC0010017
 #define AMD_K7_NUM_IORR 2

 static int nvidia_init_iorr(u32 base, u32 size)
 {
 	u32 base_hi, base_lo;
 	u32 mask_hi, mask_lo;
 	u32 sys_hi, sys_lo;
 	u32 iorr_addr, free_iorr_addr;

 	/* Find the iorr that is already used for the base */
 	/* If not found, determine the uppermost available iorr */
 	free_iorr_addr = AMD_K7_NUM_IORR;
 	for (iorr_addr = 0; iorr_addr < AMD_K7_NUM_IORR; iorr_addr++) {
 		rdmsr(IORR_BASE0 + 2 * iorr_addr, base_lo, base_hi);
 		rdmsr(IORR_MASK0 + 2 * iorr_addr, mask_lo, mask_hi);

 		if ((base_lo & 0xfffff000) == (base & 0xfffff000))
 			break;

 		if ((mask_lo & 0x00000800) == 0)
 			free_iorr_addr = iorr_addr;
 	}

 	if (iorr_addr >= AMD_K7_NUM_IORR) {
 		iorr_addr = free_iorr_addr;
 		if (iorr_addr >= AMD_K7_NUM_IORR)
 			return -EINVAL;
 	}
     base_hi = 0x0;
     base_lo = (base & ~0xfff) | 0x18;
     mask_hi = 0xf;
     mask_lo = ((~(size - 1)) & 0xfffff000) | 0x800;
     wrmsr(IORR_BASE0 + 2 * iorr_addr, base_lo, base_hi);
     wrmsr(IORR_MASK0 + 2 * iorr_addr, mask_lo, mask_hi);

     rdmsr(SYSCFG, sys_lo, sys_hi);
     sys_lo |= 0x00100000;
     wrmsr(SYSCFG, sys_lo, sys_hi);

 	return 0;
 }

 static int nvidia_configure(void)
 {
 	int i, rc, num_dirs;
 	u32 apbase, aplimit;
 	struct aper_size_info_8 *current_size;
 	u32 temp;

 	current_size = A_SIZE_8(agp_bridge->current_size);

 	/* aperture size */
 	pci_write_config_byte(agp_bridge->dev, NVIDIA_0_APSIZE,
 		current_size->size_value);

     /* address to map to */
 	pci_read_config_dword(agp_bridge->dev, AGP_APBASE, &apbase);
 	apbase &= PCI_BASE_ADDRESS_MEM_MASK;
 	agp_bridge->gart_bus_addr = apbase;
 	aplimit = apbase + (current_size->size * 1024 * 1024) - 1;
 	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_APBASE, apbase);
 	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_APLIMIT, aplimit);
 	pci_write_config_dword(nvidia_private.dev_3, NVIDIA_3_APBASE, apbase);
 	pci_write_config_dword(nvidia_private.dev_3, NVIDIA_3_APLIMIT, aplimit);
 	if (0 != (rc = nvidia_init_iorr(apbase, current_size->size * 1024 * 1024)))
 		return rc;

 	/* directory size is 64k */
 	num_dirs = current_size->size / 64;
 	nvidia_private.num_active_entries = current_size->num_entries;
 	nvidia_private.pg_offset = 0;
 	if (num_dirs == 0) {
 		num_dirs = 1;
 		nvidia_private.num_active_entries /= (64 / current_size->size);
 		nvidia_private.pg_offset = (apbase & (64 * 1024 * 1024 - 1) &
 			~(current_size->size * 1024 * 1024 - 1)) / PAGE_SIZE;
 	}

 	/* attbase */
 	for (i = 0; i < 8; i++) {
 		pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_ATTBASE(i),
 			(agp_bridge->gatt_bus_addr + (i % num_dirs) * 64 * 1024) | 1);
 	}

 	/* gtlb control */
 	pci_read_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, &temp);
 	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, temp | 0x11);

 	/* gart control */
 	pci_read_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, &temp);
 	pci_write_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, temp | 0x100);

 	/* map aperture */
 	nvidia_private.aperture =
 		(volatile u32 __iomem *) ioremap(apbase, 33 * PAGE_SIZE);

 	if (!nvidia_private.aperture)
 		return -ENOMEM;

 	return 0;
 }

 static void nvidia_cleanup(void)
 {
 	struct aper_size_info_8 *previous_size;
 	u32 temp;

 	/* gart control */
 	pci_read_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, &temp);
 	pci_write_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, temp & ~(0x100));

 	/* gtlb control */
 	pci_read_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, &temp);
 	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, temp & ~(0x11));

 	/* unmap aperture */
 	iounmap((void __iomem *) nvidia_private.aperture);

 	/* restore previous aperture size */
 	previous_size = A_SIZE_8(agp_bridge->previous_size);
 	pci_write_config_byte(agp_bridge->dev, NVIDIA_0_APSIZE,
 		previous_size->size_value);

 	/* restore iorr for previous aperture size */
 	nvidia_init_iorr(agp_bridge->gart_bus_addr,
 		previous_size->size * 1024 * 1024);
 }


 /*
  * Note we can't use the generic routines, even though they are 99% the same.
  * Aperture sizes <64M still requires a full 64k GART directory, but
  * only use the portion of the TLB entries that correspond to the apertures
  * alignment inside the surrounding 64M block.
  */
 extern int agp_memory_reserved;

 static int nvidia_insert_memory(struct agp_memory *mem, off_t pg_start, int type)
 {
 	int i, j;
 	int mask_type;

 	mask_type = agp_generic_type_to_mask_type(mem->bridge, type);
 	if (mask_type != 0 || type != mem->type)
 		return -EINVAL;

 	if (mem->page_count == 0)
 		return 0;

 	if ((pg_start + mem->page_count) >
 		(nvidia_private.num_active_entries - agp_memory_reserved/PAGE_SIZE))
 		return -EINVAL;

 	for (j = pg_start; j < (pg_start + mem->page_count); j++) {
 		if (!PGE_EMPTY(agp_bridge, readl(agp_bridge->gatt_table+nvidia_private.pg_offset+j)))
 			return -EBUSY;
 	}

 	if (!mem->is_flushed) {
 		global_cache_flush();
 		mem->is_flushed = true;
 	}
 	for (i = 0, j = pg_start; i < mem->page_count; i++, j++) {
 		writel(agp_bridge->driver->mask_memory(agp_bridge,
 			mem->memory[i], mask_type),
 			agp_bridge->gatt_table+nvidia_private.pg_offset+j);
 	}

 	/* PCI Posting. */
 	readl(agp_bridge->gatt_table+nvidia_private.pg_offset+j - 1);

 	agp_bridge->driver->tlb_flush(mem);
 	return 0;
 }


 static int nvidia_remove_memory(struct agp_memory *mem, off_t pg_start, int type)
 {
 	int i;

 	int mask_type;

 	mask_type = agp_generic_type_to_mask_type(mem->bridge, type);
 	if (mask_type != 0 || type != mem->type)
 		return -EINVAL;

 	if (mem->page_count == 0)
 		return 0;

 	for (i = pg_start; i < (mem->page_count + pg_start); i++)
 		writel(agp_bridge->scratch_page, agp_bridge->gatt_table+nvidia_private.pg_offset+i);

 	agp_bridge->driver->tlb_flush(mem);
 	return 0;
 }


 static void nvidia_tlbflush(struct agp_memory *mem)
 {
 	unsigned long end;
 	u32 wbc_reg, temp;
 	int i;

 	/* flush chipset */
 	if (nvidia_private.wbc_mask) {
 		pci_read_config_dword(nvidia_private.dev_1, NVIDIA_1_WBC, &wbc_reg);
 		wbc_reg |= nvidia_private.wbc_mask;
 		pci_write_config_dword(nvidia_private.dev_1, NVIDIA_1_WBC, wbc_reg);

 		end = jiffies + 3*HZ;
 		do {
 			pci_read_config_dword(nvidia_private.dev_1,
 					NVIDIA_1_WBC, &wbc_reg);
 			if (time_before_eq(end, jiffies)) {
 				printk(KERN_ERR PFX
 				    "TLB flush took more than 3 seconds.\n");
 			}
 		} while (wbc_reg & nvidia_private.wbc_mask);
 	}

 	/* flush TLB entries */
 	for (i = 0; i < 32 + 1; i++)
 		temp = readl(nvidia_private.aperture+(i * PAGE_SIZE / sizeof(u32)));
 	for (i = 0; i < 32 + 1; i++)
 		temp = readl(nvidia_private.aperture+(i * PAGE_SIZE / sizeof(u32)));
 }


 static const struct aper_size_info_8 nvidia_generic_sizes[5] =
 {
 	{512, 131072, 7, 0},
 	{256, 65536, 6, 8},
 	{128, 32768, 5, 12},
 	{64, 16384, 4, 14},
 	/* The 32M mode still requires a 64k gatt */
 	{32, 16384, 4, 15}
 };


 static const struct gatt_mask nvidia_generic_masks[] =
 {
 	{ .mask = 1, .type = 0}
 };


 static const struct agp_bridge_driver nvidia_driver = {
 	.owner			= THIS_MODULE,
 	.aperture_sizes		= nvidia_generic_sizes,
 	.size_type		= U8_APER_SIZE,
 	.num_aperture_sizes	= 5,
 	.configure		= nvidia_configure,
 	.fetch_size		= nvidia_fetch_size,
 	.cleanup		= nvidia_cleanup,
 	.tlb_flush		= nvidia_tlbflush,
 	.mask_memory		= agp_generic_mask_memory,
 	.masks			= nvidia_generic_masks,
 	.agp_enable		= agp_generic_enable,
 	.cache_flush		= global_cache_flush,
 	.create_gatt_table	= agp_generic_create_gatt_table,
 	.free_gatt_table	= agp_generic_free_gatt_table,
 	.insert_memory		= nvidia_insert_memory,
 	.remove_memory		= nvidia_remove_memory,
 	.alloc_by_type		= agp_generic_alloc_by_type,
 	.free_by_type		= agp_generic_free_by_type,
 	.agp_alloc_page		= agp_generic_alloc_page,
 	.agp_alloc_pages	= agp_generic_alloc_pages,
 	.agp_destroy_page	= agp_generic_destroy_page,
 	.agp_destroy_pages	= agp_generic_destroy_pages,
 	.agp_type_to_mask_type  = agp_generic_type_to_mask_type,
 };

 static int __devinit agp_nvidia_probe(struct pci_dev *pdev,
 				      const struct pci_device_id *ent)
 {
 	struct agp_bridge_data *bridge;
 	u8 cap_ptr;

 	nvidia_private.dev_1 =
 		pci_get_bus_and_slot((unsigned int)pdev->bus->number, PCI_DEVFN(0, 1));
 	nvidia_private.dev_2 =
 		pci_get_bus_and_slot((unsigned int)pdev->bus->number, PCI_DEVFN(0, 2));
 	nvidia_private.dev_3 =
 		pci_get_bus_and_slot((unsigned int)pdev->bus->number, PCI_DEVFN(30, 0));

 	if (!nvidia_private.dev_1 || !nvidia_private.dev_2 || !nvidia_private.dev_3) {
 		printk(KERN_INFO PFX "Detected an NVIDIA nForce/nForce2 "
 			"chipset, but could not find the secondary devices.\n");
 		return -ENODEV;
 	}

 	cap_ptr = pci_find_capability(pdev, PCI_CAP_ID_AGP);
 	if (!cap_ptr)
 		return -ENODEV;

 	switch (pdev->device) {
 	case PCI_DEVICE_ID_NVIDIA_NFORCE:
 		printk(KERN_INFO PFX "Detected NVIDIA nForce chipset\n");
 		nvidia_private.wbc_mask = 0x00010000;
 		break;
 	case PCI_DEVICE_ID_NVIDIA_NFORCE2:
 		printk(KERN_INFO PFX "Detected NVIDIA nForce2 chipset\n");
 		nvidia_private.wbc_mask = 0x80000000;
 		break;
 	default:
 		printk(KERN_ERR PFX "Unsupported NVIDIA chipset (device id: %04x)\n",
 			    pdev->device);
 		return -ENODEV;
 	}

 	bridge = agp_alloc_bridge();
 	if (!bridge)
 		return -ENOMEM;

 	bridge->driver = &nvidia_driver;
 	bridge->dev_private_data = &nvidia_private,
 	bridge->dev = pdev;
 	bridge->capndx = cap_ptr;

 	/* Fill in the mode register */
 	pci_read_config_dword(pdev,
 			bridge->capndx+PCI_AGP_STATUS,
 			&bridge->mode);

 	pci_set_drvdata(pdev, bridge);
 	return agp_add_bridge(bridge);
 }

 static void __devexit agp_nvidia_remove(struct pci_dev *pdev)
 {
 	struct agp_bridge_data *bridge = pci_get_drvdata(pdev);

 	agp_remove_bridge(bridge);
 	agp_put_bridge(bridge);
 }

 #ifdef CONFIG_PM
 static int agp_nvidia_suspend(struct pci_dev *pdev, pm_message_t state)
 {
 	pci_save_state (pdev);
 	pci_set_power_state (pdev, 3);

 	return 0;
 }

 static int agp_nvidia_resume(struct pci_dev *pdev)
 {
 	/* set power state 0 and restore PCI space */
 	pci_set_power_state (pdev, 0);
 	pci_restore_state(pdev);

 	/* reconfigure AGP hardware again */
 	nvidia_configure();

 	return 0;
 }
 #endif


 static struct pci_device_id agp_nvidia_pci_table[] = {
 	{
 	.class		= (PCI_CLASS_BRIDGE_HOST << 8),
 	.class_mask	= ~0,
 	.vendor		= PCI_VENDOR_ID_NVIDIA,
 	.device		= PCI_DEVICE_ID_NVIDIA_NFORCE,
 	.subvendor	= PCI_ANY_ID,
 	.subdevice	= PCI_ANY_ID,
 	},
 	{
 	.class		= (PCI_CLASS_BRIDGE_HOST << 8),
 	.class_mask	= ~0,
 	.vendor		= PCI_VENDOR_ID_NVIDIA,
 	.device		= PCI_DEVICE_ID_NVIDIA_NFORCE2,
 	.subvendor	= PCI_ANY_ID,
 	.subdevice	= PCI_ANY_ID,
 	},
 	{ }
 };

 MODULE_DEVICE_TABLE(pci, agp_nvidia_pci_table);

 static struct pci_driver agp_nvidia_pci_driver = {
 	.name		= "agpgart-nvidia",
 	.id_table	= agp_nvidia_pci_table,
 	.probe		= agp_nvidia_probe,
 	.remove		= agp_nvidia_remove,
 #ifdef CONFIG_PM
 	.suspend	= agp_nvidia_suspend,
 	.resume		= agp_nvidia_resume,
 #endif
 };

 static int __init agp_nvidia_init(void)
 {
 	if (agp_off)
 		return -EINVAL;
 	return pci_register_driver(&agp_nvidia_pci_driver);
 }

 static void __exit agp_nvidia_cleanup(void)
 {
 	pci_unregister_driver(&agp_nvidia_pci_driver);
 	pci_dev_put(nvidia_private.dev_1);
 	pci_dev_put(nvidia_private.dev_2);
 	pci_dev_put(nvidia_private.dev_3);
 }

 module_init(agp_nvidia_init);
 module_exit(agp_nvidia_cleanup);

 MODULE_LICENSE("GPL and additional rights");
 MODULE_AUTHOR("NVIDIA Corporation");
	/*
	* Nvidia AGPGART routines.
	* Based upon a 2.4 agpgart diff by the folks from NVIDIA, and hacked up
	* to work in 2.5 by Dave Jones <davej@codemonkey.org.uk>
	*/

	#include <linux/module.h>
	#include <linux/pci.h>
	#include <linux/init.h>
	#include <linux/agp_backend.h>
	#include <linux/gfp.h>
	#include <linux/page-flags.h>
	#include <linux/mm.h>
	#include <linux/jiffies.h>
	#include "agp.h"

	/* NVIDIA registers */
	#define NVIDIA_0_APSIZE 0x80
	#define NVIDIA_1_WBC 0xf0
	#define NVIDIA_2_GARTCTRL 0xd0
	#define NVIDIA_2_APBASE 0xd8
	#define NVIDIA_2_APLIMIT 0xdc
	#define NVIDIA_2_ATTBASE(i) (0xe0 + (i) * 4)
	#define NVIDIA_3_APBASE 0x50
	#define NVIDIA_3_APLIMIT 0x54


	static struct _nvidia_private {
	struct pci_dev *dev_1;
	struct pci_dev *dev_2;
	struct pci_dev *dev_3;
	volatile u32 __iomem *aperture;
	int num_active_entries;
	off_t pg_offset;
	u32 wbc_mask;
	} nvidia_private;


	static int nvidia_fetch_size(void)
	{
	int i;
	u8 size_value;
	struct aper_size_info_8 *values;

	pci_read_config_byte(agp_bridge->dev, NVIDIA_0_APSIZE, &size_value);
	size_value &= 0x0f;
	values = A_SIZE_8(agp_bridge->driver->aperture_sizes);

	for (i = 0; i < agp_bridge->driver->num_aperture_sizes; i++) {
	if (size_value == values[i].size_value) {
	agp_bridge->previous_size =
	agp_bridge->current_size = (void *) (values + i);
	agp_bridge->aperture_size_idx = i;
	return values[i].size;
	}
	}

	return 0;
	}

	#define SYSCFG 0xC0010010
	#define IORR_BASE0 0xC0010016
	#define IORR_MASK0 0xC0010017
	#define AMD_K7_NUM_IORR 2

	static int nvidia_init_iorr(u32 base, u32 size)
	{
	u32 base_hi, base_lo;
	u32 mask_hi, mask_lo;
	u32 sys_hi, sys_lo;
	u32 iorr_addr, free_iorr_addr;

	/* Find the iorr that is already used for the base */
	/* If not found, determine the uppermost available iorr */
	free_iorr_addr = AMD_K7_NUM_IORR;
	for (iorr_addr = 0; iorr_addr < AMD_K7_NUM_IORR; iorr_addr++) {
	rdmsr(IORR_BASE0 + 2 * iorr_addr, base_lo, base_hi);
	rdmsr(IORR_MASK0 + 2 * iorr_addr, mask_lo, mask_hi);

	if ((base_lo & 0xfffff000) == (base & 0xfffff000))
	break;

	if ((mask_lo & 0x00000800) == 0)
	free_iorr_addr = iorr_addr;
	}

	if (iorr_addr >= AMD_K7_NUM_IORR) {
	iorr_addr = free_iorr_addr;
	if (iorr_addr >= AMD_K7_NUM_IORR)
	return -EINVAL;
	}
	base_hi = 0x0;
	base_lo = (base & ~0xfff) \| 0x18;
	mask_hi = 0xf;
	mask_lo = ((~(size - 1)) & 0xfffff000) \| 0x800;
	wrmsr(IORR_BASE0 + 2 * iorr_addr, base_lo, base_hi);
	wrmsr(IORR_MASK0 + 2 * iorr_addr, mask_lo, mask_hi);

	rdmsr(SYSCFG, sys_lo, sys_hi);
	sys_lo \|= 0x00100000;
	wrmsr(SYSCFG, sys_lo, sys_hi);

	return 0;
	}

	static int nvidia_configure(void)
	{
	int i, rc, num_dirs;
	u32 apbase, aplimit;
	struct aper_size_info_8 *current_size;
	u32 temp;

	current_size = A_SIZE_8(agp_bridge->current_size);

	/* aperture size */
	pci_write_config_byte(agp_bridge->dev, NVIDIA_0_APSIZE,
	current_size->size_value);

	/* address to map to */
	pci_read_config_dword(agp_bridge->dev, AGP_APBASE, &apbase);
	apbase &= PCI_BASE_ADDRESS_MEM_MASK;
	agp_bridge->gart_bus_addr = apbase;
	aplimit = apbase + (current_size->size * 1024 * 1024) - 1;
	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_APBASE, apbase);
	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_APLIMIT, aplimit);
	pci_write_config_dword(nvidia_private.dev_3, NVIDIA_3_APBASE, apbase);
	pci_write_config_dword(nvidia_private.dev_3, NVIDIA_3_APLIMIT, aplimit);
	if (0 != (rc = nvidia_init_iorr(apbase, current_size->size * 1024 * 1024)))
	return rc;

	/* directory size is 64k */
	num_dirs = current_size->size / 64;
	nvidia_private.num_active_entries = current_size->num_entries;
	nvidia_private.pg_offset = 0;
	if (num_dirs == 0) {
	num_dirs = 1;
	nvidia_private.num_active_entries /= (64 / current_size->size);
	nvidia_private.pg_offset = (apbase & (64 * 1024 * 1024 - 1) &
	~(current_size->size * 1024 * 1024 - 1)) / PAGE_SIZE;
	}

	/* attbase */
	for (i = 0; i < 8; i++) {
	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_ATTBASE(i),
	(agp_bridge->gatt_bus_addr + (i % num_dirs) * 64 * 1024) \| 1);
	}

	/* gtlb control */
	pci_read_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, &temp);
	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, temp \| 0x11);

	/* gart control */
	pci_read_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, &temp);
	pci_write_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, temp \| 0x100);

	/* map aperture */
	nvidia_private.aperture =
	(volatile u32 __iomem ) ioremap(apbase, 33 PAGE_SIZE);

	if (!nvidia_private.aperture)
	return -ENOMEM;

	return 0;
	}

	static void nvidia_cleanup(void)
	{
	struct aper_size_info_8 *previous_size;
	u32 temp;

	/* gart control */
	pci_read_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, &temp);
	pci_write_config_dword(agp_bridge->dev, NVIDIA_0_APSIZE, temp & ~(0x100));

	/* gtlb control */
	pci_read_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, &temp);
	pci_write_config_dword(nvidia_private.dev_2, NVIDIA_2_GARTCTRL, temp & ~(0x11));

	/* unmap aperture */
	iounmap((void __iomem *) nvidia_private.aperture);

	/* restore previous aperture size */
	previous_size = A_SIZE_8(agp_bridge->previous_size);
	pci_write_config_byte(agp_bridge->dev, NVIDIA_0_APSIZE,
	previous_size->size_value);

	/* restore iorr for previous aperture size */
	nvidia_init_iorr(agp_bridge->gart_bus_addr,
	previous_size->size * 1024 * 1024);
	}


	/*
	* Note we can't use the generic routines, even though they are 99% the same.
	* Aperture sizes <64M still requires a full 64k GART directory, but
	* only use the portion of the TLB entries that correspond to the apertures
	* alignment inside the surrounding 64M block.
	*/
	extern int agp_memory_reserved;

	static int nvidia_insert_memory(struct agp_memory *mem, off_t pg_start, int type)
	{
	int i, j;
	int mask_type;

	mask_type = agp_generic_type_to_mask_type(mem->bridge, type);
	if (mask_type != 0 \|\| type != mem->type)
	return -EINVAL;

	if (mem->page_count == 0)
	return 0;

	if ((pg_start + mem->page_count) >
	(nvidia_private.num_active_entries - agp_memory_reserved/PAGE_SIZE))
	return -EINVAL;

	for (j = pg_start; j < (pg_start + mem->page_count); j++) {
	if (!PGE_EMPTY(agp_bridge, readl(agp_bridge->gatt_table+nvidia_private.pg_offset+j)))
	return -EBUSY;
	}

	if (!mem->is_flushed) {
	global_cache_flush();
	mem->is_flushed = true;
	}
	for (i = 0, j = pg_start; i < mem->page_count; i++, j++) {
	writel(agp_bridge->driver->mask_memory(agp_bridge,
	mem->memory[i], mask_type),
	agp_bridge->gatt_table+nvidia_private.pg_offset+j);
	}

	/* PCI Posting. */
	readl(agp_bridge->gatt_table+nvidia_private.pg_offset+j - 1);

	agp_bridge->driver->tlb_flush(mem);
	return 0;
	}


	static int nvidia_remove_memory(struct agp_memory *mem, off_t pg_start, int type)
	{
	int i;

	int mask_type;

	mask_type = agp_generic_type_to_mask_type(mem->bridge, type);
	if (mask_type != 0 \|\| type != mem->type)
	return -EINVAL;

	if (mem->page_count == 0)
	return 0;

	for (i = pg_start; i < (mem->page_count + pg_start); i++)
	writel(agp_bridge->scratch_page, agp_bridge->gatt_table+nvidia_private.pg_offset+i);

	agp_bridge->driver->tlb_flush(mem);
	return 0;
	}


	static void nvidia_tlbflush(struct agp_memory *mem)
	{
	unsigned long end;
	u32 wbc_reg, temp;
	int i;

	/* flush chipset */
	if (nvidia_private.wbc_mask) {
	pci_read_config_dword(nvidia_private.dev_1, NVIDIA_1_WBC, &wbc_reg);
	wbc_reg \|= nvidia_private.wbc_mask;
	pci_write_config_dword(nvidia_private.dev_1, NVIDIA_1_WBC, wbc_reg);

	end = jiffies + 3*HZ;
	do {
	pci_read_config_dword(nvidia_private.dev_1,
	NVIDIA_1_WBC, &wbc_reg);
	if (time_before_eq(end, jiffies)) {
	printk(KERN_ERR PFX
	"TLB flush took more than 3 seconds.\n");
	}
	} while (wbc_reg & nvidia_private.wbc_mask);
	}

	/* flush TLB entries */
	for (i = 0; i < 32 + 1; i++)
	temp = readl(nvidia_private.aperture+(i * PAGE_SIZE / sizeof(u32)));
	for (i = 0; i < 32 + 1; i++)
	temp = readl(nvidia_private.aperture+(i * PAGE_SIZE / sizeof(u32)));
	}


	static const struct aper_size_info_8 nvidia_generic_sizes[5] =
	{
	{512, 131072, 7, 0},
	{256, 65536, 6, 8},
	{128, 32768, 5, 12},
	{64, 16384, 4, 14},
	/* The 32M mode still requires a 64k gatt */
	{32, 16384, 4, 15}
	};


	static const struct gatt_mask nvidia_generic_masks[] =
	{
	{ .mask = 1, .type = 0}
	};


	static const struct agp_bridge_driver nvidia_driver = {
	.owner = THIS_MODULE,
	.aperture_sizes = nvidia_generic_sizes,
	.size_type = U8_APER_SIZE,
	.num_aperture_sizes = 5,
	.configure = nvidia_configure,
	.fetch_size = nvidia_fetch_size,
	.cleanup = nvidia_cleanup,
	.tlb_flush = nvidia_tlbflush,
	.mask_memory = agp_generic_mask_memory,
	.masks = nvidia_generic_masks,
	.agp_enable = agp_generic_enable,
	.cache_flush = global_cache_flush,
	.create_gatt_table = agp_generic_create_gatt_table,
	.free_gatt_table = agp_generic_free_gatt_table,
	.insert_memory = nvidia_insert_memory,
	.remove_memory = nvidia_remove_memory,
	.alloc_by_type = agp_generic_alloc_by_type,
	.free_by_type = agp_generic_free_by_type,
	.agp_alloc_page = agp_generic_alloc_page,
	.agp_alloc_pages = agp_generic_alloc_pages,
	.agp_destroy_page = agp_generic_destroy_page,
	.agp_destroy_pages = agp_generic_destroy_pages,
	.agp_type_to_mask_type = agp_generic_type_to_mask_type,
	};

	static int __devinit agp_nvidia_probe(struct pci_dev *pdev,
	const struct pci_device_id *ent)
	{
	struct agp_bridge_data *bridge;
	u8 cap_ptr;

	nvidia_private.dev_1 =
	pci_get_bus_and_slot((unsigned int)pdev->bus->number, PCI_DEVFN(0, 1));
	nvidia_private.dev_2 =
	pci_get_bus_and_slot((unsigned int)pdev->bus->number, PCI_DEVFN(0, 2));
	nvidia_private.dev_3 =
	pci_get_bus_and_slot((unsigned int)pdev->bus->number, PCI_DEVFN(30, 0));

	if (!nvidia_private.dev_1 \|\| !nvidia_private.dev_2 \|\| !nvidia_private.dev_3) {
	printk(KERN_INFO PFX "Detected an NVIDIA nForce/nForce2 "
	"chipset, but could not find the secondary devices.\n");
	return -ENODEV;
	}

	cap_ptr = pci_find_capability(pdev, PCI_CAP_ID_AGP);
	if (!cap_ptr)
	return -ENODEV;

	switch (pdev->device) {
	case PCI_DEVICE_ID_NVIDIA_NFORCE:
	printk(KERN_INFO PFX "Detected NVIDIA nForce chipset\n");
	nvidia_private.wbc_mask = 0x00010000;
	break;
	case PCI_DEVICE_ID_NVIDIA_NFORCE2:
	printk(KERN_INFO PFX "Detected NVIDIA nForce2 chipset\n");
	nvidia_private.wbc_mask = 0x80000000;
	break;
	default:
	printk(KERN_ERR PFX "Unsupported NVIDIA chipset (device id: %04x)\n",
	pdev->device);
	return -ENODEV;
	}

	bridge = agp_alloc_bridge();
	if (!bridge)
	return -ENOMEM;

	bridge->driver = &nvidia_driver;
	bridge->dev_private_data = &nvidia_private,
	bridge->dev = pdev;
	bridge->capndx = cap_ptr;

	/* Fill in the mode register */
	pci_read_config_dword(pdev,
	bridge->capndx+PCI_AGP_STATUS,
	&bridge->mode);

	pci_set_drvdata(pdev, bridge);
	return agp_add_bridge(bridge);
	}

	static void __devexit agp_nvidia_remove(struct pci_dev *pdev)
	{
	struct agp_bridge_data *bridge = pci_get_drvdata(pdev);

	agp_remove_bridge(bridge);
	agp_put_bridge(bridge);
	}

	#ifdef CONFIG_PM
	static int agp_nvidia_suspend(struct pci_dev *pdev, pm_message_t state)
	{
	pci_save_state (pdev);
	pci_set_power_state (pdev, 3);

	return 0;
	}

	static int agp_nvidia_resume(struct pci_dev *pdev)
	{
	/* set power state 0 and restore PCI space */
	pci_set_power_state (pdev, 0);
	pci_restore_state(pdev);

	/* reconfigure AGP hardware again */
	nvidia_configure();

	return 0;
	}
	#endif


	static struct pci_device_id agp_nvidia_pci_table[] = {
	{
	.class = (PCI_CLASS_BRIDGE_HOST << 8),
	.class_mask = ~0,
	.vendor = PCI_VENDOR_ID_NVIDIA,
	.device = PCI_DEVICE_ID_NVIDIA_NFORCE,
	.subvendor = PCI_ANY_ID,
	.subdevice = PCI_ANY_ID,
	},
	{
	.class = (PCI_CLASS_BRIDGE_HOST << 8),
	.class_mask = ~0,
	.vendor = PCI_VENDOR_ID_NVIDIA,
	.device = PCI_DEVICE_ID_NVIDIA_NFORCE2,
	.subvendor = PCI_ANY_ID,
	.subdevice = PCI_ANY_ID,
	},
	{ }
	};

	MODULE_DEVICE_TABLE(pci, agp_nvidia_pci_table);

	static struct pci_driver agp_nvidia_pci_driver = {
	.name = "agpgart-nvidia",
	.id_table = agp_nvidia_pci_table,
	.probe = agp_nvidia_probe,
	.remove = agp_nvidia_remove,
	#ifdef CONFIG_PM
	.suspend = agp_nvidia_suspend,
	.resume = agp_nvidia_resume,
	#endif
	};

	static int __init agp_nvidia_init(void)
	{
	if (agp_off)
	return -EINVAL;
	return pci_register_driver(&agp_nvidia_pci_driver);
	}

	static void __exit agp_nvidia_cleanup(void)
	{
	pci_unregister_driver(&agp_nvidia_pci_driver);
	pci_dev_put(nvidia_private.dev_1);
	pci_dev_put(nvidia_private.dev_2);
	pci_dev_put(nvidia_private.dev_3);
	}

	module_init(agp_nvidia_init);
	module_exit(agp_nvidia_cleanup);

	MODULE_LICENSE("GPL and additional rights");
	MODULE_AUTHOR("NVIDIA Corporation");