stram.c

是关于linux2.5.1的完全源码

		 * __get_dma_pages() */
		addr = (void *)__get_dma_pages(GFP_KERNEL, get_order(size));
		flags = BLOCK_GFP;
		DPRINTK( "atari_stram_alloc: after mem_init, swapping off, "
				 "get_pages=%p\n", addr );
	}

	if (addr) {
		if (!(block = add_region( addr, size ))) {
			/* out of memory for BLOCK structure :-( */
			DPRINTK( "atari_stram_alloc: out of mem for BLOCK -- "
					 "freeing again\n" );
#ifdef CONFIG_STRAM_SWAP
			if (flags == BLOCK_INSWAP)
				free_stram_region( SWAP_NR(addr), N_PAGES(size) );
			else
#endif
				free_pages((unsigned long)addr, get_order(size));
			return( NULL );
		}
		block->owner = owner;
		block->flags |= flags;
	}
	return( addr );
}

void atari_stram_free( void *addr )

{
	BLOCK *block;

	DPRINTK( "atari_stram_free(addr=%p)\n", addr );

	if (!(block = find_region( addr ))) {
		printk( KERN_ERR "Attempt to free non-allocated ST-RAM block at %p "
				"from %p\n", addr, __builtin_return_address(0) );
		return;
	}
	DPRINTK( "atari_stram_free: found block (%p): size=%08lx, owner=%s, "
			 "flags=%02x\n", block, block->size, block->owner, block->flags );
	
#ifdef CONFIG_STRAM_SWAP
	if (!max_swap_size) {
#endif
		if (block->flags & BLOCK_GFP) {
			DPRINTK("atari_stram_free: is kmalloced, order_size=%d\n",
				get_order(block->size));
			free_pages((unsigned long)addr, get_order(block->size));
		}
		else
			goto fail;
#ifdef CONFIG_STRAM_SWAP
	}
	else if (block->flags & BLOCK_INSWAP) {
		DPRINTK( "atari_stram_free: is swap-alloced\n" );
		free_stram_region( SWAP_NR(block->start), N_PAGES(block->size) );
	}
	else
		goto fail;
#endif
	remove_region( block );
	return;

  fail:
	printk( KERN_ERR "atari_stram_free: cannot free block at %p "
			"(called from %p)\n", addr, __builtin_return_address(0) );
}


#ifdef CONFIG_STRAM_SWAP


/* ------------------------------------------------------------------------ */
/*						   Main Swapping Functions							*/
/* ------------------------------------------------------------------------ */


/*
 * Initialize ST-RAM swap device
 * (lots copied and modified from sys_swapon() in mm/swapfile.c)
 */
static int __init swap_init(void *start_mem, void *swap_data)
{
	static struct dentry fake_dentry;
	static struct vfsmount fake_vfsmnt;
	struct swap_info_struct *p;
	struct inode swap_inode;
	unsigned int type;
	void *addr;
	int i, j, k, prev;

	DPRINTK("swap_init(start_mem=%p, swap_data=%p)\n",
		start_mem, swap_data);
	
	/* need at least one page for swapping to (and this also isn't very
	 * much... :-) */
	if (swap_end - swap_start < 2*PAGE_SIZE) {
		printk( KERN_WARNING "stram_swap_init: swap space too small\n" );
		return( 0 );
	}
	
	/* find free slot in swap_info */
	for( p = swap_info, type = 0; type < nr_swapfiles; type++, p++ )
		if (!(p->flags & SWP_USED))
			break;
	if (type >= MAX_SWAPFILES) {
		printk( KERN_WARNING "stram_swap_init: max. number of "
				"swap devices exhausted\n" );
		return( 0 );
	}
	if (type >= nr_swapfiles)
		nr_swapfiles = type+1;

	stram_swap_info = p;
	stram_swap_type = type;

	/* fake some dir cache entries to give us some name in /dev/swaps */
	fake_dentry.d_parent = &fake_dentry;
	fake_dentry.d_name.name = "stram (internal)";
	fake_dentry.d_name.len = 16;
	fake_vfsmnt.mnt_parent = &fake_vfsmnt;
	
	p->flags        = SWP_USED;
	p->swap_file    = &fake_dentry;
	p->swap_vfsmnt  = &fake_vfsmnt;
	p->swap_device  = 0;
	p->swap_map	= swap_data;
	p->cluster_nr   = 0;
	p->next         = -1;
	p->prio         = 0x7ff0;	/* a rather high priority, but not the higest
								 * to give the user a chance to override */

	/* call stram_open() directly, avoids at least the overhead in
	 * constructing a dummy file structure... */
	p->swap_device = MKDEV( STRAM_MAJOR, STRAM_MINOR );
	swap_inode.i_rdev = p->swap_device;
	stram_open( &swap_inode, MAGIC_FILE_P );
	p->max = SWAP_NR(swap_end);

	/* initialize swap_map: set regions that are already allocated or belong
	 * to kernel data space to SWAP_MAP_BAD, otherwise to free */
	j = 0; /* # of free pages */
	k = 0; /* # of already allocated pages (from pre-mem_init stram_alloc()) */
	p->lowest_bit = 0;
	p->highest_bit = 0;
	for( i = 1, addr = SWAP_ADDR(1); i < p->max;
		 i++, addr += PAGE_SIZE ) {
		if (in_some_region( addr )) {
			p->swap_map[i] = SWAP_MAP_BAD;
			++k;
		}
		else if (kernel_in_stram && addr < start_mem ) {
			p->swap_map[i] = SWAP_MAP_BAD;
		}
		else {
			p->swap_map[i] = 0;
			++j;
			if (!p->lowest_bit) p->lowest_bit = i;
			p->highest_bit = i;
		}
	}
	/* first page always reserved (and doesn't really belong to swap space) */
	p->swap_map[0] = SWAP_MAP_BAD;

	/* now swapping to this device ok */
	p->pages = j + k;
	swap_list_lock();
	nr_swap_pages += j;
	p->flags = SWP_WRITEOK;

	/* insert swap space into swap_list */
	prev = -1;
	for (i = swap_list.head; i >= 0; i = swap_info[i].next) {
		if (p->prio >= swap_info[i].prio) {
			break;
		}
		prev = i;
	}
	p->next = i;
	if (prev < 0) {
		swap_list.head = swap_list.next = p - swap_info;
	} else {
		swap_info[prev].next = p - swap_info;
	}
	swap_list_unlock();

	printk( KERN_INFO "Using %dk (%d pages) of ST-RAM as swap space.\n",
			p->pages << 2, p->pages );
	return( 1 );
}


/*
 * The swap entry has been read in advance, and we return 1 to indicate
 * that the page has been used or is no longer needed.
 *
 * Always set the resulting pte to be nowrite (the same as COW pages
 * after one process has exited).  We don't know just how many PTEs will
 * share this swap entry, so be cautious and let do_wp_page work out
 * what to do if a write is requested later.
 */
static inline void unswap_pte(struct vm_area_struct * vma, unsigned long
			      address, pte_t *dir, swp_entry_t entry,
			      struct page *page)
{
	pte_t pte = *dir;

	if (pte_none(pte))
		return;
	if (pte_present(pte)) {
		/* If this entry is swap-cached, then page must already
                   hold the right address for any copies in physical
                   memory */
		if (pte_page(pte) != page)
			return;
		/* We will be removing the swap cache in a moment, so... */
		set_pte(dir, pte_mkdirty(pte));
		return;
	}
	if (pte_val(pte) != entry.val)
		return;

	DPRINTK("unswap_pte: replacing entry %08lx by new page %p",
		entry.val, page);
	set_pte(dir, pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	swap_free(entry);
	get_page(page);
	++vma->vm_mm->rss;
}

static inline void unswap_pmd(struct vm_area_struct * vma, pmd_t *dir,
			      unsigned long address, unsigned long size,
			      unsigned long offset, swp_entry_t entry,
			      struct page *page)
{
	pte_t * pte;
	unsigned long end;

	if (pmd_none(*dir))
		return;
	if (pmd_bad(*dir)) {
		pmd_ERROR(*dir);
		pmd_clear(dir);
		return;
	}
	pte = pte_offset(dir, address);
	offset += address & PMD_MASK;
	address &= ~PMD_MASK;
	end = address + size;
	if (end > PMD_SIZE)
		end = PMD_SIZE;
	do {
		unswap_pte(vma, offset+address-vma->vm_start, pte, entry, page);
		address += PAGE_SIZE;
		pte++;
	} while (address < end);
}

static inline void unswap_pgd(struct vm_area_struct * vma, pgd_t *dir,
			      unsigned long address, unsigned long size,
			      swp_entry_t entry, struct page *page)
{
	pmd_t * pmd;
	unsigned long offset, end;

	if (pgd_none(*dir))
		return;
	if (pgd_bad(*dir)) {
		pgd_ERROR(*dir);
		pgd_clear(dir);
		return;
	}
	pmd = pmd_offset(dir, address);
	offset = address & PGDIR_MASK;
	address &= ~PGDIR_MASK;
	end = address + size;
	if (end > PGDIR_SIZE)
		end = PGDIR_SIZE;
	do {
		unswap_pmd(vma, pmd, address, end - address, offset, entry,
			   page);
		address = (address + PMD_SIZE) & PMD_MASK;
		pmd++;
	} while (address < end);
}

static void unswap_vma(struct vm_area_struct * vma, pgd_t *pgdir,
		       swp_entry_t entry, struct page *page)
{
	unsigned long start = vma->vm_start, end = vma->vm_end;

	do {
		unswap_pgd(vma, pgdir, start, end - start, entry, page);
		start = (start + PGDIR_SIZE) & PGDIR_MASK;
		pgdir++;
	} while (start < end);
}

static void unswap_process(struct mm_struct * mm, swp_entry_t entry, 
			   struct page *page)
{
	struct vm_area_struct* vma;

	/*
	 * Go through process' page directory.
	 */
	if (!mm)
		return;
	for (vma = mm->mmap; vma; vma = vma->vm_next) {
		pgd_t * pgd = pgd_offset(mm, vma->vm_start);
		unswap_vma(vma, pgd, entry, page);
	}
}


static int unswap_by_read(unsigned short *map, unsigned long max,
			  unsigned long start, unsigned long n_pages)
{
	struct task_struct *p;
	struct page *page;
	swp_entry_t entry;
	unsigned long i;

	DPRINTK( "unswapping %lu..%lu by reading in\n",
			 start, start+n_pages-1 );

	for( i = start; i < start+n_pages; ++i ) {
		if (map[i] == SWAP_MAP_BAD) {
			printk( KERN_ERR "get_stram_region: page %lu already "
					"reserved??\n", i );
			continue;
		}

		if (map[i]) {
			entry = SWP_ENTRY(stram_swap_type, i);
			DPRINTK("unswap: map[i=%lu]=%u nr_swap=%u\n",
				i, map[i], nr_swap_pages);

			swap_device_lock(stram_swap_info);
			map[i]++;
			swap_device_unlock(stram_swap_info);
			/* Get a page for the entry, using the existing
			   swap cache page if there is one.  Otherwise,
			   get a clean page and read the swap into it. */
			page = read_swap_cache(entry);
			if (!page) {
				swap_free(entry);
				return -ENOMEM;
			}
			read_lock(&tasklist_lock);
			for_each_task(p)
				unswap_process(p->mm, entry, page);
			read_unlock(&tasklist_lock);
			shm_unuse(entry, page);
			/* Now get rid of the extra reference to the
			   temporary page we've been using. */
			if (PageSwapCache(page))
				delete_from_swap_cache(page);
			__free_page(page);
	#ifdef DO_PROC
			stat_swap_force++;
	#endif
		}

		DPRINTK( "unswap: map[i=%lu]=%u nr_swap=%u\n",
				 i, map[i], nr_swap_pages );
		swap_list_lock();
		swap_device_lock(stram_swap_info);
		map[i] = SWAP_MAP_BAD;
		if (stram_swap_info->lowest_bit == i)
			stram_swap_info->lowest_bit++;
		if (stram_swap_info->highest_bit == i)
			stram_swap_info->highest_bit--;
		--nr_swap_pages;
		swap_device_unlock(stram_swap_info);
		swap_list_unlock();
	}

	return 0;
}

/*
 * reserve a region in ST-RAM swap space for an allocation
 */
static void *get_stram_region( unsigned long n_pages )
{
	unsigned short *map = stram_swap_info->swap_map;
	unsigned long max = stram_swap_info->max;
	unsigned long start, total_free, region_free;
	int err;
	void *ret = NULL;
	
	DPRINTK( "get_stram_region(n_pages=%lu)\n", n_pages );

	down(&stram_swap_sem);

	/* disallow writing to the swap device now */
	stram_swap_info->flags = SWP_USED;

	/* find a region of n_pages pages in the swap space including as much free
	 * pages as possible (and excluding any already-reserved pages). */
	if (!(start = find_free_region( n_pages, &total_free, &region_free )))
		goto end;
	DPRINTK( "get_stram_region: region starts at %lu, has %lu free pages\n",
			 start, region_free );

	err = unswap_by_read(map, max, start, n_pages);
	if (err)
		goto end;

	ret = SWAP_ADDR(start);
  end:
	/* allow using swap device again */
	stram_swap_info->flags = SWP_WRITEOK;
	up(&stram_swap_sem);
	DPRINTK( "get_stram_region: returning %p\n", ret );
	return( ret );