balloc.c

linux 内核源代码

/*
 *  linux/fs/ufs/balloc.c
 *
 * Copyright (C) 1998
 * Daniel Pirkl <[e-mail]>
 * Charles University, Faculty of Mathematics and Physics
 *
 * UFS2 write support Evgeniy Dushistov <[e-mail]>, 2007
 */

#include <linux/fs.h>
#include <linux/ufs_fs.h>
#include <linux/stat.h>
#include <linux/time.h>
#include <linux/string.h>
#include <linux/quotaops.h>
#include <linux/buffer_head.h>
#include <linux/capability.h>
#include <linux/bitops.h>
#include <asm/byteorder.h>

#include "ufs.h"
#include "swab.h"
#include "util.h"

#define INVBLOCK ((u64)-1L)

static u64 ufs_add_fragments(struct inode *, u64, unsigned, unsigned, int *);
static u64 ufs_alloc_fragments(struct inode *, unsigned, u64, unsigned, int *);
static u64 ufs_alloccg_block(struct inode *, struct ufs_cg_private_info *, u64, int *);
static u64 ufs_bitmap_search (struct super_block *, struct ufs_cg_private_info *, u64, unsigned);
static unsigned char ufs_fragtable_8fpb[], ufs_fragtable_other[];
static void ufs_clusteracct(struct super_block *, struct ufs_cg_private_info *, unsigned, int);

/*
 * Free 'count' fragments from fragment number 'fragment'
 */
void ufs_free_fragments(struct inode *inode, u64 fragment, unsigned count)
{
	struct super_block * sb;
	struct ufs_sb_private_info * uspi;
	struct ufs_super_block_first * usb1;
	struct ufs_cg_private_info * ucpi;
	struct ufs_cylinder_group * ucg;
	unsigned cgno, bit, end_bit, bbase, blkmap, i;
	u64 blkno;
	
	sb = inode->i_sb;
	uspi = UFS_SB(sb)->s_uspi;
	usb1 = ubh_get_usb_first(uspi);
	
	UFSD("ENTER, fragment %llu, count %u\n",
	     (unsigned long long)fragment, count);
	
	if (ufs_fragnum(fragment) + count > uspi->s_fpg)
		ufs_error (sb, "ufs_free_fragments", "internal error");
	
	lock_super(sb);
	
	cgno = ufs_dtog(uspi, fragment);
	bit = ufs_dtogd(uspi, fragment);
	if (cgno >= uspi->s_ncg) {
		ufs_panic (sb, "ufs_free_fragments", "freeing blocks are outside device");
		goto failed;
	}
		
	ucpi = ufs_load_cylinder (sb, cgno);
	if (!ucpi) 
		goto failed;
	ucg = ubh_get_ucg (UCPI_UBH(ucpi));
	if (!ufs_cg_chkmagic(sb, ucg)) {
		ufs_panic (sb, "ufs_free_fragments", "internal error, bad magic number on cg %u", cgno);
		goto failed;
	}

	end_bit = bit + count;
	bbase = ufs_blknum (bit);
	blkmap = ubh_blkmap (UCPI_UBH(ucpi), ucpi->c_freeoff, bbase);
	ufs_fragacct (sb, blkmap, ucg->cg_frsum, -1);
	for (i = bit; i < end_bit; i++) {
		if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, i))
			ubh_setbit (UCPI_UBH(ucpi), ucpi->c_freeoff, i);
		else 
			ufs_error (sb, "ufs_free_fragments",
				   "bit already cleared for fragment %u", i);
	}
	
	DQUOT_FREE_BLOCK (inode, count);

	
	fs32_add(sb, &ucg->cg_cs.cs_nffree, count);
	uspi->cs_total.cs_nffree += count;
	fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
	blkmap = ubh_blkmap (UCPI_UBH(ucpi), ucpi->c_freeoff, bbase);
	ufs_fragacct(sb, blkmap, ucg->cg_frsum, 1);

	/*
	 * Trying to reassemble free fragments into block
	 */
	blkno = ufs_fragstoblks (bbase);
	if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno)) {
		fs32_sub(sb, &ucg->cg_cs.cs_nffree, uspi->s_fpb);
		uspi->cs_total.cs_nffree -= uspi->s_fpb;
		fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, uspi->s_fpb);
		if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
			ufs_clusteracct (sb, ucpi, blkno, 1);
		fs32_add(sb, &ucg->cg_cs.cs_nbfree, 1);
		uspi->cs_total.cs_nbfree++;
		fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nbfree, 1);
		if (uspi->fs_magic != UFS2_MAGIC) {
			unsigned cylno = ufs_cbtocylno (bbase);

			fs16_add(sb, &ubh_cg_blks(ucpi, cylno,
						  ufs_cbtorpos(bbase)), 1);
			fs32_add(sb, &ubh_cg_blktot(ucpi, cylno), 1);
		}
	}
	
	ubh_mark_buffer_dirty (USPI_UBH(uspi));
	ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
	if (sb->s_flags & MS_SYNCHRONOUS) {
		ubh_ll_rw_block(SWRITE, UCPI_UBH(ucpi));
		ubh_wait_on_buffer (UCPI_UBH(ucpi));
	}
	sb->s_dirt = 1;
	
	unlock_super (sb);
	UFSD("EXIT\n");
	return;

failed:
	unlock_super (sb);
	UFSD("EXIT (FAILED)\n");
	return;
}

/*
 * Free 'count' fragments from fragment number 'fragment' (free whole blocks)
 */
void ufs_free_blocks(struct inode *inode, u64 fragment, unsigned count)
{
	struct super_block * sb;
	struct ufs_sb_private_info * uspi;
	struct ufs_super_block_first * usb1;
	struct ufs_cg_private_info * ucpi;
	struct ufs_cylinder_group * ucg;
	unsigned overflow, cgno, bit, end_bit, i;
	u64 blkno;
	
	sb = inode->i_sb;
	uspi = UFS_SB(sb)->s_uspi;
	usb1 = ubh_get_usb_first(uspi);

	UFSD("ENTER, fragment %llu, count %u\n",
	     (unsigned long long)fragment, count);
	
	if ((fragment & uspi->s_fpbmask) || (count & uspi->s_fpbmask)) {
		ufs_error (sb, "ufs_free_blocks", "internal error, "
			   "fragment %llu, count %u\n",
			   (unsigned long long)fragment, count);
		goto failed;
	}

	lock_super(sb);
	
do_more:
	overflow = 0;
	cgno = ufs_dtog(uspi, fragment);
	bit = ufs_dtogd(uspi, fragment);
	if (cgno >= uspi->s_ncg) {
		ufs_panic (sb, "ufs_free_blocks", "freeing blocks are outside device");
		goto failed_unlock;
	}
	end_bit = bit + count;
	if (end_bit > uspi->s_fpg) {
		overflow = bit + count - uspi->s_fpg;
		count -= overflow;
		end_bit -= overflow;
	}

	ucpi = ufs_load_cylinder (sb, cgno);
	if (!ucpi) 
		goto failed_unlock;
	ucg = ubh_get_ucg (UCPI_UBH(ucpi));
	if (!ufs_cg_chkmagic(sb, ucg)) {
		ufs_panic (sb, "ufs_free_blocks", "internal error, bad magic number on cg %u", cgno);
		goto failed_unlock;
	}

	for (i = bit; i < end_bit; i += uspi->s_fpb) {
		blkno = ufs_fragstoblks(i);
		if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno)) {
			ufs_error(sb, "ufs_free_blocks", "freeing free fragment");
		}
		ubh_setblock(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno);
		if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
			ufs_clusteracct (sb, ucpi, blkno, 1);
		DQUOT_FREE_BLOCK(inode, uspi->s_fpb);

		fs32_add(sb, &ucg->cg_cs.cs_nbfree, 1);
		uspi->cs_total.cs_nbfree++;
		fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nbfree, 1);

		if (uspi->fs_magic != UFS2_MAGIC) {
			unsigned cylno = ufs_cbtocylno(i);

			fs16_add(sb, &ubh_cg_blks(ucpi, cylno,
						  ufs_cbtorpos(i)), 1);
			fs32_add(sb, &ubh_cg_blktot(ucpi, cylno), 1);
		}
	}

	ubh_mark_buffer_dirty (USPI_UBH(uspi));
	ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
	if (sb->s_flags & MS_SYNCHRONOUS) {
		ubh_ll_rw_block(SWRITE, UCPI_UBH(ucpi));
		ubh_wait_on_buffer (UCPI_UBH(ucpi));
	}

	if (overflow) {
		fragment += count;
		count = overflow;
		goto do_more;
	}

	sb->s_dirt = 1;
	unlock_super (sb);
	UFSD("EXIT\n");
	return;

failed_unlock:
	unlock_super (sb);
failed:
	UFSD("EXIT (FAILED)\n");
	return;
}

/*
 * Modify inode page cache in such way:
 * have - blocks with b_blocknr equal to oldb...oldb+count-1
 * get - blocks with b_blocknr equal to newb...newb+count-1
 * also we suppose that oldb...oldb+count-1 blocks
 * situated at the end of file.
 *
 * We can come here from ufs_writepage or ufs_prepare_write,
 * locked_page is argument of these functions, so we already lock it.
 */
static void ufs_change_blocknr(struct inode *inode, sector_t beg,
			       unsigned int count, sector_t oldb,
			       sector_t newb, struct page *locked_page)
{
	const unsigned blks_per_page =
		1 << (PAGE_CACHE_SHIFT - inode->i_blkbits);
	const unsigned mask = blks_per_page - 1;
	struct address_space * const mapping = inode->i_mapping;
	pgoff_t index, cur_index, last_index;
	unsigned pos, j, lblock;
	sector_t end, i;
	struct page *page;
	struct buffer_head *head, *bh;

	UFSD("ENTER, ino %lu, count %u, oldb %llu, newb %llu\n",
	      inode->i_ino, count,
	     (unsigned long long)oldb, (unsigned long long)newb);

	BUG_ON(!locked_page);
	BUG_ON(!PageLocked(locked_page));

	cur_index = locked_page->index;
	end = count + beg;
	last_index = end >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
	for (i = beg; i < end; i = (i | mask) + 1) {
		index = i >> (PAGE_CACHE_SHIFT - inode->i_blkbits);

		if (likely(cur_index != index)) {
			page = ufs_get_locked_page(mapping, index);
			if (!page)/* it was truncated */
				continue;
			if (IS_ERR(page)) {/* or EIO */
				ufs_error(inode->i_sb, __FUNCTION__,
					  "read of page %llu failed\n",
					  (unsigned long long)index);
				continue;
			}
		} else
			page = locked_page;

		head = page_buffers(page);
		bh = head;
		pos = i & mask;
		for (j = 0; j < pos; ++j)
			bh = bh->b_this_page;


		if (unlikely(index == last_index))
			lblock = end & mask;
		else
			lblock = blks_per_page;

		do {
			if (j >= lblock)
				break;
			pos = (i - beg) + j;

			if (!buffer_mapped(bh))
					map_bh(bh, inode->i_sb, oldb + pos);
			if (!buffer_uptodate(bh)) {
				ll_rw_block(READ, 1, &bh);
				wait_on_buffer(bh);
				if (!buffer_uptodate(bh)) {
					ufs_error(inode->i_sb, __FUNCTION__,
						  "read of block failed\n");
					break;
				}
			}

			UFSD(" change from %llu to %llu, pos %u\n",
			     (unsigned long long)pos + oldb,
			     (unsigned long long)pos + newb, pos);

			bh->b_blocknr = newb + pos;
			unmap_underlying_metadata(bh->b_bdev,
						  bh->b_blocknr);
			mark_buffer_dirty(bh);
			++j;
			bh = bh->b_this_page;
		} while (bh != head);

		if (likely(cur_index != index))
			ufs_put_locked_page(page);
 	}
	UFSD("EXIT\n");
}

static void ufs_clear_frags(struct inode *inode, sector_t beg, unsigned int n,
			    int sync)
{
	struct buffer_head *bh;
	sector_t end = beg + n;

	for (; beg < end; ++beg) {
		bh = sb_getblk(inode->i_sb, beg);
		lock_buffer(bh);
		memset(bh->b_data, 0, inode->i_sb->s_blocksize);
		set_buffer_uptodate(bh);
		mark_buffer_dirty(bh);
		unlock_buffer(bh);
		if (IS_SYNC(inode) || sync)
			sync_dirty_buffer(bh);
		brelse(bh);
	}
}

u64 ufs_new_fragments(struct inode *inode, void *p, u64 fragment,
			   u64 goal, unsigned count, int *err,
			   struct page *locked_page)
{
	struct super_block * sb;
	struct ufs_sb_private_info * uspi;
	struct ufs_super_block_first * usb1;
	unsigned cgno, oldcount, newcount;
	u64 tmp, request, result;
	
	UFSD("ENTER, ino %lu, fragment %llu, goal %llu, count %u\n",
	     inode->i_ino, (unsigned long long)fragment,
	     (unsigned long long)goal, count);
	
	sb = inode->i_sb;
	uspi = UFS_SB(sb)->s_uspi;
	usb1 = ubh_get_usb_first(uspi);
	*err = -ENOSPC;

	lock_super (sb);
	tmp = ufs_data_ptr_to_cpu(sb, p);

	if (count + ufs_fragnum(fragment) > uspi->s_fpb) {
		ufs_warning(sb, "ufs_new_fragments", "internal warning"
			    " fragment %llu, count %u",
			    (unsigned long long)fragment, count);
		count = uspi->s_fpb - ufs_fragnum(fragment); 
	}
	oldcount = ufs_fragnum (fragment);
	newcount = oldcount + count;

	/*
	 * Somebody else has just allocated our fragments
	 */
	if (oldcount) {
		if (!tmp) {
			ufs_error(sb, "ufs_new_fragments", "internal error, "
				  "fragment %llu, tmp %llu\n",
				  (unsigned long long)fragment,
				  (unsigned long long)tmp);
			unlock_super(sb);
			return INVBLOCK;
		}
		if (fragment < UFS_I(inode)->i_lastfrag) {
			UFSD("EXIT (ALREADY ALLOCATED)\n");
			unlock_super (sb);
			return 0;
		}
	}
	else {
		if (tmp) {
			UFSD("EXIT (ALREADY ALLOCATED)\n");
			unlock_super(sb);
			return 0;
		}
	}

	/*
	 * There is not enough space for user on the device
	 */
	if (!capable(CAP_SYS_RESOURCE) && ufs_freespace(uspi, UFS_MINFREE) <= 0) {
		unlock_super (sb);
		UFSD("EXIT (FAILED)\n");
		return 0;
	}

	if (goal >= uspi->s_size) 
		goal = 0;
	if (goal == 0) 
		cgno = ufs_inotocg (inode->i_ino);
	else
		cgno = ufs_dtog(uspi, goal);
	 
	/*
	 * allocate new fragment
	 */
	if (oldcount == 0) {
		result = ufs_alloc_fragments (inode, cgno, goal, count, err);
		if (result) {
			ufs_cpu_to_data_ptr(sb, p, result);
			*err = 0;
			UFS_I(inode)->i_lastfrag =
				max_t(u32, UFS_I(inode)->i_lastfrag,
				      fragment + count);
			ufs_clear_frags(inode, result + oldcount,
					newcount - oldcount, locked_page != NULL);
		}
		unlock_super(sb);
		UFSD("EXIT, result %llu\n", (unsigned long long)result);
		return result;
	}

	/*
	 * resize block
	 */
	result = ufs_add_fragments (inode, tmp, oldcount, newcount, err);
	if (result) {
		*err = 0;
		UFS_I(inode)->i_lastfrag = max_t(u32, UFS_I(inode)->i_lastfrag, fragment + count);
		ufs_clear_frags(inode, result + oldcount, newcount - oldcount,
				locked_page != NULL);
		unlock_super(sb);
		UFSD("EXIT, result %llu\n", (unsigned long long)result);
		return result;
	}

	/*
	 * allocate new block and move data
	 */
	switch (fs32_to_cpu(sb, usb1->fs_optim)) {
	    case UFS_OPTSPACE:
		request = newcount;
		if (uspi->s_minfree < 5 || uspi->cs_total.cs_nffree
		    > uspi->s_dsize * uspi->s_minfree / (2 * 100))
			break;
		usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
		break;
	    default:
		usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
	
	    case UFS_OPTTIME:
		request = uspi->s_fpb;
		if (uspi->cs_total.cs_nffree < uspi->s_dsize *
		    (uspi->s_minfree - 2) / 100)
			break;
		usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
		break;
	}
	result = ufs_alloc_fragments (inode, cgno, goal, request, err);
	if (result) {
		ufs_clear_frags(inode, result + oldcount, newcount - oldcount,
				locked_page != NULL);
		ufs_change_blocknr(inode, fragment - oldcount, oldcount,
				   uspi->s_sbbase + tmp,
				   uspi->s_sbbase + result, locked_page);
		ufs_cpu_to_data_ptr(sb, p, result);
		*err = 0;