init.c

linux-2.6.15.6

			total++;
			if (PageReserved(p))
				reserved++;
			else if (PageSwapCache(p))
				cached++;
			else if (!page_count(p))
				free++;
			else
				shared += page_count(p) - 1;
			pgdat_resize_unlock(NODE_DATA(i), &flags);
        	}
	}
#endif
	printk(KERN_INFO "%d pages of RAM\n", total);
	printk(KERN_INFO "%d reserved pages\n", reserved);
	printk(KERN_INFO "%d pages shared\n", shared);
	printk(KERN_INFO "%d pages swap cached\n", cached);


#ifdef CONFIG_DISCONTIGMEM
	{
		struct zonelist *zl;
		int i, j, k;

		for (i = 0; i < npmem_ranges; i++) {
			for (j = 0; j < MAX_NR_ZONES; j++) {
				zl = NODE_DATA(i)->node_zonelists + j;

				printk("Zone list for zone %d on node %d: ", j, i);
				for (k = 0; zl->zones[k] != NULL; k++) 
					printk("[%d/%s] ", zl->zones[k]->zone_pgdat->node_id, zl->zones[k]->name);
				printk("\n");
			}
		}
	}
#endif
}


static void __init map_pages(unsigned long start_vaddr, unsigned long start_paddr, unsigned long size, pgprot_t pgprot)
{
	pgd_t *pg_dir;
	pmd_t *pmd;
	pte_t *pg_table;
	unsigned long end_paddr;
	unsigned long start_pmd;
	unsigned long start_pte;
	unsigned long tmp1;
	unsigned long tmp2;
	unsigned long address;
	unsigned long ro_start;
	unsigned long ro_end;
	unsigned long fv_addr;
	unsigned long gw_addr;
	extern const unsigned long fault_vector_20;
	extern void * const linux_gateway_page;

	ro_start = __pa((unsigned long)&_text);
	ro_end   = __pa((unsigned long)&data_start);
	fv_addr  = __pa((unsigned long)&fault_vector_20) & PAGE_MASK;
	gw_addr  = __pa((unsigned long)&linux_gateway_page) & PAGE_MASK;

	end_paddr = start_paddr + size;

	pg_dir = pgd_offset_k(start_vaddr);

#if PTRS_PER_PMD == 1
	start_pmd = 0;
#else
	start_pmd = ((start_vaddr >> PMD_SHIFT) & (PTRS_PER_PMD - 1));
#endif
	start_pte = ((start_vaddr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1));

	address = start_paddr;
	while (address < end_paddr) {
#if PTRS_PER_PMD == 1
		pmd = (pmd_t *)__pa(pg_dir);
#else
		pmd = (pmd_t *)pgd_address(*pg_dir);

		/*
		 * pmd is physical at this point
		 */

		if (!pmd) {
			pmd = (pmd_t *) alloc_bootmem_low_pages_node(NODE_DATA(0),PAGE_SIZE << PMD_ORDER);
			pmd = (pmd_t *) __pa(pmd);
		}

		pgd_populate(NULL, pg_dir, __va(pmd));
#endif
		pg_dir++;

		/* now change pmd to kernel virtual addresses */

		pmd = (pmd_t *)__va(pmd) + start_pmd;
		for (tmp1 = start_pmd; tmp1 < PTRS_PER_PMD; tmp1++,pmd++) {

			/*
			 * pg_table is physical at this point
			 */

			pg_table = (pte_t *)pmd_address(*pmd);
			if (!pg_table) {
				pg_table = (pte_t *)
					alloc_bootmem_low_pages_node(NODE_DATA(0),PAGE_SIZE);
				pg_table = (pte_t *) __pa(pg_table);
			}

			pmd_populate_kernel(NULL, pmd, __va(pg_table));

			/* now change pg_table to kernel virtual addresses */

			pg_table = (pte_t *) __va(pg_table) + start_pte;
			for (tmp2 = start_pte; tmp2 < PTRS_PER_PTE; tmp2++,pg_table++) {
				pte_t pte;

				/*
				 * Map the fault vector writable so we can
				 * write the HPMC checksum.
				 */
				if (address >= ro_start && address < ro_end
							&& address != fv_addr
							&& address != gw_addr)
				    pte = __mk_pte(address, PAGE_KERNEL_RO);
				else
				    pte = __mk_pte(address, pgprot);

				if (address >= end_paddr)
					pte_val(pte) = 0;

				set_pte(pg_table, pte);

				address += PAGE_SIZE;
			}
			start_pte = 0;

			if (address >= end_paddr)
			    break;
		}
		start_pmd = 0;
	}
}

/*
 * pagetable_init() sets up the page tables
 *
 * Note that gateway_init() places the Linux gateway page at page 0.
 * Since gateway pages cannot be dereferenced this has the desirable
 * side effect of trapping those pesky NULL-reference errors in the
 * kernel.
 */
static void __init pagetable_init(void)
{
	int range;

	/* Map each physical memory range to its kernel vaddr */

	for (range = 0; range < npmem_ranges; range++) {
		unsigned long start_paddr;
		unsigned long end_paddr;
		unsigned long size;

		start_paddr = pmem_ranges[range].start_pfn << PAGE_SHIFT;
		end_paddr = start_paddr + (pmem_ranges[range].pages << PAGE_SHIFT);
		size = pmem_ranges[range].pages << PAGE_SHIFT;

		map_pages((unsigned long)__va(start_paddr), start_paddr,
			size, PAGE_KERNEL);
	}

#ifdef CONFIG_BLK_DEV_INITRD
	if (initrd_end && initrd_end > mem_limit) {
		printk("initrd: mapping %08lx-%08lx\n", initrd_start, initrd_end);
		map_pages(initrd_start, __pa(initrd_start),
			initrd_end - initrd_start, PAGE_KERNEL);
	}
#endif

	empty_zero_page = alloc_bootmem_pages(PAGE_SIZE);
	memset(empty_zero_page, 0, PAGE_SIZE);
}

static void __init gateway_init(void)
{
	unsigned long linux_gateway_page_addr;
	/* FIXME: This is 'const' in order to trick the compiler
	   into not treating it as DP-relative data. */
	extern void * const linux_gateway_page;

	linux_gateway_page_addr = LINUX_GATEWAY_ADDR & PAGE_MASK;

	/*
	 * Setup Linux Gateway page.
	 *
	 * The Linux gateway page will reside in kernel space (on virtual
	 * page 0), so it doesn't need to be aliased into user space.
	 */

	map_pages(linux_gateway_page_addr, __pa(&linux_gateway_page),
		PAGE_SIZE, PAGE_GATEWAY);
}

#ifdef CONFIG_HPUX
void
map_hpux_gateway_page(struct task_struct *tsk, struct mm_struct *mm)
{
	pgd_t *pg_dir;
	pmd_t *pmd;
	pte_t *pg_table;
	unsigned long start_pmd;
	unsigned long start_pte;
	unsigned long address;
	unsigned long hpux_gw_page_addr;
	/* FIXME: This is 'const' in order to trick the compiler
	   into not treating it as DP-relative data. */
	extern void * const hpux_gateway_page;

	hpux_gw_page_addr = HPUX_GATEWAY_ADDR & PAGE_MASK;

	/*
	 * Setup HP-UX Gateway page.
	 *
	 * The HP-UX gateway page resides in the user address space,
	 * so it needs to be aliased into each process.
	 */

	pg_dir = pgd_offset(mm,hpux_gw_page_addr);

#if PTRS_PER_PMD == 1
	start_pmd = 0;
#else
	start_pmd = ((hpux_gw_page_addr >> PMD_SHIFT) & (PTRS_PER_PMD - 1));
#endif
	start_pte = ((hpux_gw_page_addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1));

	address = __pa(&hpux_gateway_page);
#if PTRS_PER_PMD == 1
	pmd = (pmd_t *)__pa(pg_dir);
#else
	pmd = (pmd_t *) pgd_address(*pg_dir);

	/*
	 * pmd is physical at this point
	 */

	if (!pmd) {
		pmd = (pmd_t *) get_zeroed_page(GFP_KERNEL);
		pmd = (pmd_t *) __pa(pmd);
	}

	__pgd_val_set(*pg_dir, PxD_FLAG_PRESENT | PxD_FLAG_VALID | (unsigned long) pmd);
#endif
	/* now change pmd to kernel virtual addresses */

	pmd = (pmd_t *)__va(pmd) + start_pmd;

	/*
	 * pg_table is physical at this point
	 */

	pg_table = (pte_t *) pmd_address(*pmd);
	if (!pg_table)
		pg_table = (pte_t *) __pa(get_zeroed_page(GFP_KERNEL));

	__pmd_val_set(*pmd, PxD_FLAG_PRESENT | PxD_FLAG_VALID | (unsigned long) pg_table);

	/* now change pg_table to kernel virtual addresses */

	pg_table = (pte_t *) __va(pg_table) + start_pte;
	set_pte(pg_table, __mk_pte(address, PAGE_GATEWAY));
}
EXPORT_SYMBOL(map_hpux_gateway_page);
#endif

extern void flush_tlb_all_local(void);

void __init paging_init(void)
{
	int i;

	setup_bootmem();
	pagetable_init();
	gateway_init();
	flush_cache_all_local(); /* start with known state */
	flush_tlb_all_local();

	for (i = 0; i < npmem_ranges; i++) {
		unsigned long zones_size[MAX_NR_ZONES] = { 0, 0, 0 };

		/* We have an IOMMU, so all memory can go into a single
		   ZONE_DMA zone. */
		zones_size[ZONE_DMA] = pmem_ranges[i].pages;

#ifdef CONFIG_DISCONTIGMEM
		/* Need to initialize the pfnnid_map before we can initialize
		   the zone */
		{
		    int j;
		    for (j = (pmem_ranges[i].start_pfn >> PFNNID_SHIFT);
			 j <= ((pmem_ranges[i].start_pfn + pmem_ranges[i].pages) >> PFNNID_SHIFT);
			 j++) {
			pfnnid_map[j] = i;
		    }
		}
#endif

		free_area_init_node(i, NODE_DATA(i), zones_size,
				pmem_ranges[i].start_pfn, NULL);
	}
}

#ifdef CONFIG_PA20

/*
 * Currently, all PA20 chips have 18 bit protection id's, which is the
 * limiting factor (space ids are 32 bits).
 */

#define NR_SPACE_IDS 262144

#else

/*
 * Currently we have a one-to-one relationship between space id's and
 * protection id's. Older parisc chips (PCXS, PCXT, PCXL, PCXL2) only
 * support 15 bit protection id's, so that is the limiting factor.
 * PCXT' has 18 bit protection id's, but only 16 bit spaceids, so it's
 * probably not worth the effort for a special case here.
 */

#define NR_SPACE_IDS 32768

#endif  /* !CONFIG_PA20 */

#define RECYCLE_THRESHOLD (NR_SPACE_IDS / 2)
#define SID_ARRAY_SIZE  (NR_SPACE_IDS / (8 * sizeof(long)))

static unsigned long space_id[SID_ARRAY_SIZE] = { 1 }; /* disallow space 0 */
static unsigned long dirty_space_id[SID_ARRAY_SIZE];
static unsigned long space_id_index;
static unsigned long free_space_ids = NR_SPACE_IDS - 1;
static unsigned long dirty_space_ids = 0;

static DEFINE_SPINLOCK(sid_lock);

unsigned long alloc_sid(void)
{
	unsigned long index;

	spin_lock(&sid_lock);

	if (free_space_ids == 0) {
		if (dirty_space_ids != 0) {
			spin_unlock(&sid_lock);
			flush_tlb_all(); /* flush_tlb_all() calls recycle_sids() */
			spin_lock(&sid_lock);
		}
		if (free_space_ids == 0)
			BUG();
	}

	free_space_ids--;

	index = find_next_zero_bit(space_id, NR_SPACE_IDS, space_id_index);
	space_id[index >> SHIFT_PER_LONG] |= (1L << (index & (BITS_PER_LONG - 1)));
	space_id_index = index;

	spin_unlock(&sid_lock);

	return index << SPACEID_SHIFT;
}

void free_sid(unsigned long spaceid)
{
	unsigned long index = spaceid >> SPACEID_SHIFT;
	unsigned long *dirty_space_offset;

	dirty_space_offset = dirty_space_id + (index >> SHIFT_PER_LONG);
	index &= (BITS_PER_LONG - 1);

	spin_lock(&sid_lock);

	if (*dirty_space_offset & (1L << index))
	    BUG(); /* attempt to free space id twice */

	*dirty_space_offset |= (1L << index);
	dirty_space_ids++;

	spin_unlock(&sid_lock);
}


#ifdef CONFIG_SMP
static void get_dirty_sids(unsigned long *ndirtyptr,unsigned long *dirty_array)
{
	int i;

	/* NOTE: sid_lock must be held upon entry */

	*ndirtyptr = dirty_space_ids;
	if (dirty_space_ids != 0) {
	    for (i = 0; i < SID_ARRAY_SIZE; i++) {
		dirty_array[i] = dirty_space_id[i];
		dirty_space_id[i] = 0;
	    }
	    dirty_space_ids = 0;
	}

	return;
}

static void recycle_sids(unsigned long ndirty,unsigned long *dirty_array)
{
	int i;

	/* NOTE: sid_lock must be held upon entry */

	if (ndirty != 0) {
		for (i = 0; i < SID_ARRAY_SIZE; i++) {
			space_id[i] ^= dirty_array[i];
		}

		free_space_ids += ndirty;
		space_id_index = 0;
	}
}

#else /* CONFIG_SMP */

static void recycle_sids(void)
{
	int i;

	/* NOTE: sid_lock must be held upon entry */

	if (dirty_space_ids != 0) {
		for (i = 0; i < SID_ARRAY_SIZE; i++) {
			space_id[i] ^= dirty_space_id[i];
			dirty_space_id[i] = 0;
		}

		free_space_ids += dirty_space_ids;
		dirty_space_ids = 0;
		space_id_index = 0;
	}
}
#endif

/*
 * flush_tlb_all() calls recycle_sids(), since whenever the entire tlb is
 * purged, we can safely reuse the space ids that were released but
 * not flushed from the tlb.
 */

#ifdef CONFIG_SMP

static unsigned long recycle_ndirty;
static unsigned long recycle_dirty_array[SID_ARRAY_SIZE];
static unsigned int recycle_inuse = 0;

void flush_tlb_all(void)
{
	int do_recycle;

	do_recycle = 0;
	spin_lock(&sid_lock);
	if (dirty_space_ids > RECYCLE_THRESHOLD) {
	    if (recycle_inuse) {
		BUG();  /* FIXME: Use a semaphore/wait queue here */
	    }
	    get_dirty_sids(&recycle_ndirty,recycle_dirty_array);
	    recycle_inuse++;
	    do_recycle++;
	}
	spin_unlock(&sid_lock);
	on_each_cpu((void (*)(void *))flush_tlb_all_local, NULL, 1, 1);
	if (do_recycle) {
	    spin_lock(&sid_lock);
	    recycle_sids(recycle_ndirty,recycle_dirty_array);
	    recycle_inuse = 0;
	    spin_unlock(&sid_lock);
	}
}
#else
void flush_tlb_all(void)
{
	spin_lock(&sid_lock);
	flush_tlb_all_local();
	recycle_sids();
	spin_unlock(&sid_lock);
}
#endif

#ifdef CONFIG_BLK_DEV_INITRD
void free_initrd_mem(unsigned long start, unsigned long end)
{
#if 0
	if (start < end)
		printk(KERN_INFO "Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
	for (; start < end; start += PAGE_SIZE) {
		ClearPageReserved(virt_to_page(start));
		set_page_count(virt_to_page(start), 1);
		free_page(start);
		num_physpages++;
		totalram_pages++;
	}
#endif
}
#endif