pci-gart.c

这个linux源代码是很全面的~基本完整了~使用c编译的~由于时间问题我没有亲自测试~但就算用来做参考资料也是非常好的

	unsigned long phys_mem, bus;
	int i, npages;

	BUG_ON(dir == PCI_DMA_NONE);

	phys_mem = virt_to_phys(addr); 
	if (!need_iommu(dev, phys_mem, size))
		return phys_mem; 

	npages = round_up(size + ((u64)addr & ~PAGE_MASK), PAGE_SIZE) >> PAGE_SHIFT;

	iommu_page = alloc_iommu(npages); 
	if (iommu_page == -1) {
		iommu_full(dev, addr, size, dir); 
		return iommu_bus_base; 
	} 

	phys_mem &= PAGE_MASK;
	for (i = 0; i < npages; i++, phys_mem += PAGE_SIZE) {
		BUG_ON(phys_mem & ~PHYSICAL_PAGE_MASK); 
		
		/* 
		 * Set coherent mapping here to avoid needing to flush
		 * the caches on mapping.
		 */
		iommu_gatt_base[iommu_page + i] = GPTE_ENCODE(phys_mem);

#ifdef CONFIG_IOMMU_DEBUG
		/* paranoia check */
		BUG_ON(GPTE_DECODE(iommu_gatt_base[iommu_page+i]) != phys_mem); 

		/* XXX need eventually caller of pci_map_sg */
		if (iommu_leak_tab) 
			iommu_leak_tab[iommu_page + i] = __builtin_return_address(0); 
#endif
	}
	if (flush)
	flush_gart(); 

	bus = iommu_bus_base + iommu_page*PAGE_SIZE; 
	return bus + ((unsigned long)addr & ~PAGE_MASK); 	
} 

/*
 * Free a temporary PCI mapping.
 */ 
void pci_unmap_single(struct pci_dev *hwdev, dma_addr_t dma_addr,
		      size_t size, int direction)
{
	unsigned long iommu_page; 
	int i, npages;
	if (dma_addr < iommu_bus_base + EMERGENCY_PAGES*PAGE_SIZE || 
	    dma_addr > iommu_bus_base + iommu_size)
		return;
	iommu_page = (dma_addr - iommu_bus_base)>>PAGE_SHIFT;	
	npages = round_up(size + (dma_addr & ~PAGE_MASK), PAGE_SIZE) >> PAGE_SHIFT;
	for (i = 0; i < npages; i++) { 
		iommu_gatt_base[iommu_page + i] = 0; 
#ifdef CONFIG_IOMMU_DEBUG
		if (iommu_leak_tab)
			iommu_leak_tab[iommu_page + i] = 0; 
#endif
	}
	flush_gart(); 
	free_iommu(iommu_page, npages);
}

EXPORT_SYMBOL(__pci_map_single);
EXPORT_SYMBOL(pci_unmap_single);

static __init unsigned long check_iommu_size(unsigned long aper, u64 aper_size)
{ 
	unsigned long a; 
	if (!iommu_size) { 
		iommu_size = aper_size; 
		if (!no_agp) 
			iommu_size /= 2; 
	} 

	a = aper + iommu_size; 
	iommu_size -= round_up(a, LARGE_PAGE_SIZE) - a;

	if (iommu_size < 64*1024*1024) 
		printk(KERN_WARNING
  "PCI-DMA: Warning: Small IOMMU %luMB. Consider increasing the AGP aperture in BIOS\n",iommu_size>>20); 
	
	return iommu_size;
} 

static __init unsigned read_aperture(struct pci_dev *dev, u32 *size) 
{ 
	unsigned aper_size = 0, aper_base_32;
	u64 aper_base;
	unsigned aper_order;

	pci_read_config_dword(dev, 0x94, &aper_base_32); 
	pci_read_config_dword(dev, 0x90, &aper_order);
	aper_order = (aper_order >> 1) & 7;	

	aper_base = aper_base_32 & 0x7fff; 
	aper_base <<= 25;

	aper_size = (32 * 1024 * 1024) << aper_order; 
	if (aper_base + aper_size >= 0xffffffff || !aper_size)
		aper_base = 0;

	*size = aper_size;
	return aper_base;
} 

/* 
 * Private Northbridge GATT initialization in case we cannot use the
 * AGP driver for some reason.  
 */
static __init int init_k8_gatt(agp_kern_info *info)
{ 
	struct pci_dev *dev;
	void *gatt;
	unsigned aper_base, new_aper_base;
	unsigned aper_size, gatt_size, new_aper_size;

	aper_size = aper_base = info->aper_size = 0;
	for_all_nb(dev) { 
		new_aper_base = read_aperture(dev, &new_aper_size); 
		if (!new_aper_base) 
	goto nommu; 

		if (!aper_base) { 
			aper_size = new_aper_size;
			aper_base = new_aper_base;
	}
		if (aper_size != new_aper_size || aper_base != new_aper_base) 
		goto nommu;
	} 
	if (!aper_base)
		goto nommu; 
	info->aper_base = aper_base;
	info->aper_size = aper_size>>20; 

	gatt_size = (aper_size >> PAGE_SHIFT) * sizeof(u32); 
	gatt = (void *)__get_free_pages(GFP_KERNEL, get_order(gatt_size)); 
	if (!gatt) 
		panic("Cannot allocate GATT table"); 
	memset(gatt, 0, gatt_size); 
	change_page_attr(virt_to_page(gatt), gatt_size/PAGE_SIZE, PAGE_KERNEL_NOCACHE);
	agp_gatt_table = gatt;

	for_all_nb(dev) { 
		u32 ctl; 
		u32 gatt_reg; 

		gatt_reg = ((u64)gatt) >> 12; 
		gatt_reg <<= 4; 
		pci_write_config_dword(dev, 0x98, gatt_reg);
		pci_read_config_dword(dev, 0x90, &ctl); 

		ctl |= 1;
		ctl &= ~((1<<4) | (1<<5));

		pci_write_config_dword(dev, 0x90, ctl); 
	}
	flush_gart(); 
	
		
	printk("PCI-DMA: aperture base @ %x size %u KB\n", aper_base, aper_size>>10); 
		return 0;

 nommu:
	/* XXX: reject 0xffffffff mask now in pci mapping functions */
		printk(KERN_ERR "PCI-DMA: More than 4GB of RAM and no IOMMU\n"
	       KERN_ERR "PCI-DMA: 32bit PCI IO may malfunction."); 
	return -1; 
} 

void __init pci_iommu_init(void)
{ 
	agp_kern_info info;
	unsigned long aper_size;
	unsigned long iommu_start;

#ifndef CONFIG_AGP
	no_agp = 1; 
#else
	no_agp = no_agp || (agp_init() < 0) || (agp_copy_info(&info) < 0); 
#endif	

	if (no_iommu || (!force_mmu && end_pfn < 0xffffffff>>PAGE_SHIFT)) { 
		printk(KERN_INFO "PCI-DMA: Disabling IOMMU.\n"); 
		no_iommu = 1;
		return;
	}

	if (no_agp) { 
		int err = -1;
		printk(KERN_INFO "PCI-DMA: Disabling AGP.\n");
		no_agp = 1;
		if (force_mmu || end_pfn >= 0xffffffff>>PAGE_SHIFT)
			err = init_k8_gatt(&info);
		if (err < 0) { 
			printk(KERN_INFO "PCI-DMA: Disabling IOMMU.\n"); 
			no_iommu = 1;
			return; 
		}
	} 
	
	aper_size = info.aper_size * 1024 * 1024;
	iommu_size = check_iommu_size(info.aper_base, aper_size); 
	iommu_pages = iommu_size >> PAGE_SHIFT; 

	iommu_gart_bitmap = (void*)__get_free_pages(GFP_KERNEL, 
						    get_order(iommu_pages/8)); 
	if (!iommu_gart_bitmap) 
		panic("Cannot allocate iommu bitmap\n"); 
	memset(iommu_gart_bitmap, 0, iommu_pages/8);

#ifdef CONFIG_IOMMU_DEBUG
	if (leak_trace) { 
		iommu_leak_tab = (void *)__get_free_pages(GFP_KERNEL, 
				  get_order(iommu_pages*sizeof(void *)));
		if (iommu_leak_tab) 
			memset(iommu_leak_tab, 0, iommu_pages * 8); 
		else
			printk("PCI-DMA: Cannot allocate leak trace area\n"); 
	} 
#endif

	/* 
	 * Out of IOMMU space handling.
	 * Reserve some invalid pages at the beginning of the GART. 
	 */ 
	set_bit_string(iommu_gart_bitmap, 0, EMERGENCY_PAGES); 

	agp_memory_reserved = iommu_size;	
	printk(KERN_INFO"PCI-DMA: Reserving %luMB of IOMMU area in the AGP aperture\n",
	       iommu_size>>20); 

	iommu_start = aper_size - iommu_size;	
	iommu_bus_base = info.aper_base + iommu_start; 
	iommu_gatt_base = agp_gatt_table + (iommu_start>>PAGE_SHIFT);
	bad_dma_address = iommu_bus_base;

	asm volatile("wbinvd" ::: "memory");
} 

/* iommu=[size][,noagp][,off][,force][,noforce][,leak][,memaper[=order]]
   size  set size of iommu (in bytes) 
   noagp don't initialize the AGP driver and use full aperture.
   off   don't use the IOMMU
   leak  turn on simple iommu leak tracing (only when CONFIG_IOMMU_LEAK is on)
   memaper[=order] allocate an own aperture over RAM with size 32MB^order.
   noforce don't force IOMMU usage. Should be fastest.
   force  Force IOMMU and turn on unmap debugging.
*/
__init int iommu_setup(char *opt) 
{ 
    int arg;
    char *p = opt;
    
    for (;;) { 
	    if (!memcmp(p,"noagp", 5))
		    no_agp = 1; 
	    if (!memcmp(p,"off", 3))
		    no_iommu = 1;
	    if (!memcmp(p,"force", 5))
		    force_mmu = 1;
	    if (!memcmp(p,"noforce", 7))
		    force_mmu = 0;
	    if (!memcmp(p, "memaper", 7)) { 
		    fallback_aper_force = 1; 
		    p += 7; 
		    if (*p == '=' && get_option(&p, &arg))
			    fallback_aper_order = arg;
	    } 
#ifdef CONFIG_IOMMU_DEBUG
	    if (!memcmp(p,"leak", 4)) { 
		    leak_trace = 1;
		    p += 4; 
		    if (*p == '=') ++p;
		    if (isdigit(*p) && get_option(&p, &arg))
			    iommu_leak_pages = arg;
	    } else
#endif
	    if (isdigit(*p) && get_option(&p, &arg)) 
		    iommu_size = arg;
	    do {
		    if (*p == ' ' || *p == 0) 
			    return 0; 
	    } while (*p++ != ','); 
    }	
    return 1;
}