jfs_imap.c

jfs-2.4-1.1.7.tar.gz jfs 2.4－1.1.7 源码

/*
 *   Copyright (C) International Business Machines Corp., 2000-2003
 *
 *   This program is free software;  you can redistribute it and/or modify
 *   it under the terms of the GNU General Public License as published by
 *   the Free Software Foundation; either version 2 of the License, or 
 *   (at your option) any later version.
 * 
 *   This program is distributed in the hope that it will be useful,
 *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
 *   the GNU General Public License for more details.
 *
 *   You should have received a copy of the GNU General Public License
 *   along with this program;  if not, write to the Free Software 
 *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */

/*
 *	jfs_imap.c: inode allocation map manager
 *
 * Serialization:
 *   Each AG has a simple lock which is used to control the serialization of
 *	the AG level lists.  This lock should be taken first whenever an AG
 *	level list will be modified or accessed.
 *
 *   Each IAG is locked by obtaining the buffer for the IAG page.
 *
 *   There is also a inode lock for the inode map inode.  A read lock needs to
 *	be taken whenever an IAG is read from the map or the global level
 *	information is read.  A write lock needs to be taken whenever the global
 *	level information is modified or an atomic operation needs to be used.
 *
 *	If more than one IAG is read at one time, the read lock may not
 *	be given up until all of the IAG's are read.  Otherwise, a deadlock
 *	may occur when trying to obtain the read lock while another thread
 *	holding the read lock is waiting on the IAG already being held.
 *
 *   The control page of the inode map is read into memory by diMount().
 *	Thereafter it should only be modified in memory and then it will be
 *	written out when the filesystem is unmounted by diUnmount().
 */

#include <linux/fs.h>
#include <linux/locks.h>
#include "jfs_incore.h"
#include "jfs_filsys.h"
#include "jfs_dinode.h"
#include "jfs_dmap.h"
#include "jfs_imap.h"
#include "jfs_metapage.h"
#include "jfs_superblock.h"
#include "jfs_debug.h"

/*
 * imap locks
 */
/* iag free list lock */
#define IAGFREE_LOCK_INIT(imap)		init_MUTEX(&imap->im_freelock)
#define IAGFREE_LOCK(imap)		down(&imap->im_freelock)
#define IAGFREE_UNLOCK(imap)		up(&imap->im_freelock)

/* per ag iag list locks */
#define AG_LOCK_INIT(imap,index)	init_MUTEX(&(imap->im_aglock[index]))
#define AG_LOCK(imap,agno)		down(&imap->im_aglock[agno])
#define AG_UNLOCK(imap,agno)		up(&imap->im_aglock[agno])

/*
 * external references
 */
extern struct address_space_operations jfs_aops;

/*
 * forward references
 */
static int diAllocAG(struct inomap *, int, boolean_t, struct inode *);
static int diAllocAny(struct inomap *, int, boolean_t, struct inode *);
static int diAllocBit(struct inomap *, struct iag *, int);
static int diAllocExt(struct inomap *, int, struct inode *);
static int diAllocIno(struct inomap *, int, struct inode *);
static int diFindFree(u32, int);
static int diNewExt(struct inomap *, struct iag *, int);
static int diNewIAG(struct inomap *, int *, int, struct metapage **);
static void duplicateIXtree(struct super_block *, s64, int, s64 *);

static int diIAGRead(struct inomap * imap, int, struct metapage **);
static int copy_from_dinode(struct dinode *, struct inode *);
static void copy_to_dinode(struct dinode *, struct inode *);

/*
 *	debug code for double-checking inode map
 */
/* #define	_JFS_DEBUG_IMAP	1 */

#ifdef	_JFS_DEBUG_IMAP
#define DBG_DIINIT(imap)	DBGdiInit(imap)
#define DBG_DIALLOC(imap, ino)	DBGdiAlloc(imap, ino)
#define DBG_DIFREE(imap, ino)	DBGdiFree(imap, ino)

static void *DBGdiInit(struct inomap * imap);
static void DBGdiAlloc(struct inomap * imap, ino_t ino);
static void DBGdiFree(struct inomap * imap, ino_t ino);
#else
#define DBG_DIINIT(imap)
#define DBG_DIALLOC(imap, ino)
#define DBG_DIFREE(imap, ino)
#endif				/* _JFS_DEBUG_IMAP */

/*
 * NAME:        diMount()
 *
 * FUNCTION:    initialize the incore inode map control structures for
 *		a fileset or aggregate init time.
 *
 *              the inode map's control structure (dinomap) is 
 *              brought in from disk and placed in virtual memory.
 *
 * PARAMETERS:
 *      ipimap  - pointer to inode map inode for the aggregate or fileset.
 *
 * RETURN VALUES:
 *      0       - success
 *      -ENOMEM  - insufficient free virtual memory.
 *      -EIO  	- i/o error.
 */
int diMount(struct inode *ipimap)
{
	struct inomap *imap;
	struct metapage *mp;
	int index;
	struct dinomap *dinom_le;

	/*
	 * allocate/initialize the in-memory inode map control structure
	 */
	/* allocate the in-memory inode map control structure. */
	imap = (struct inomap *) kmalloc(sizeof(struct inomap), GFP_KERNEL);
	if (imap == NULL) {
		jfs_err("diMount: kmalloc returned NULL!");
		return -ENOMEM;
	}

	/* read the on-disk inode map control structure. */

	mp = read_metapage(ipimap,
			   IMAPBLKNO << JFS_SBI(ipimap->i_sb)->l2nbperpage,
			   PSIZE, 0);
	if (mp == NULL) {
		kfree(imap);
		return -EIO;
	}

	/* copy the on-disk version to the in-memory version. */
	dinom_le = (struct dinomap *) mp->data;
	imap->im_freeiag = le32_to_cpu(dinom_le->in_freeiag);
	imap->im_nextiag = le32_to_cpu(dinom_le->in_nextiag);
	atomic_set(&imap->im_numinos, le32_to_cpu(dinom_le->in_numinos));
	atomic_set(&imap->im_numfree, le32_to_cpu(dinom_le->in_numfree));
	imap->im_nbperiext = le32_to_cpu(dinom_le->in_nbperiext);
	imap->im_l2nbperiext = le32_to_cpu(dinom_le->in_l2nbperiext);
	for (index = 0; index < MAXAG; index++) {
		imap->im_agctl[index].inofree =
		    le32_to_cpu(dinom_le->in_agctl[index].inofree);
		imap->im_agctl[index].extfree =
		    le32_to_cpu(dinom_le->in_agctl[index].extfree);
		imap->im_agctl[index].numinos =
		    le32_to_cpu(dinom_le->in_agctl[index].numinos);
		imap->im_agctl[index].numfree =
		    le32_to_cpu(dinom_le->in_agctl[index].numfree);
	}

	/* release the buffer. */
	release_metapage(mp);

	/*
	 * allocate/initialize inode allocation map locks
	 */
	/* allocate and init iag free list lock */
	IAGFREE_LOCK_INIT(imap);

	/* allocate and init ag list locks */
	for (index = 0; index < MAXAG; index++) {
		AG_LOCK_INIT(imap, index);
	}

	/* bind the inode map inode and inode map control structure
	 * to each other.
	 */
	imap->im_ipimap = ipimap;
	JFS_IP(ipimap)->i_imap = imap;

//      DBG_DIINIT(imap);

	return (0);
}


/*
 * NAME:        diUnmount()
 *
 * FUNCTION:    write to disk the incore inode map control structures for
 *		a fileset or aggregate at unmount time.
 *
 * PARAMETERS:
 *      ipimap  - pointer to inode map inode for the aggregate or fileset.
 *
 * RETURN VALUES:
 *      0       - success
 *      -ENOMEM  - insufficient free virtual memory.
 *      -EIO  	- i/o error.
 */
int diUnmount(struct inode *ipimap, int mounterror)
{
	struct inomap *imap = JFS_IP(ipimap)->i_imap;

	/*
	 * update the on-disk inode map control structure
	 */

	if (!(mounterror || isReadOnly(ipimap)))
		diSync(ipimap);

	/*
	 * Invalidate the page cache buffers
	 */
	truncate_inode_pages(ipimap->i_mapping, 0);

	/*
	 * free in-memory control structure
	 */
	kfree(imap);

	return (0);
}


/*
 *	diSync()
 */
int diSync(struct inode *ipimap)
{
	struct dinomap *dinom_le;
	struct inomap *imp = JFS_IP(ipimap)->i_imap;
	struct metapage *mp;
	int index;

	/*
	 * write imap global conrol page
	 */
	/* read the on-disk inode map control structure */
	mp = get_metapage(ipimap,
			  IMAPBLKNO << JFS_SBI(ipimap->i_sb)->l2nbperpage,
			  PSIZE, 0);
	if (mp == NULL) {
		jfs_err("diSync: get_metapage failed!");
		return -EIO;
	}

	/* copy the in-memory version to the on-disk version */
	dinom_le = (struct dinomap *) mp->data;
	dinom_le->in_freeiag = cpu_to_le32(imp->im_freeiag);
	dinom_le->in_nextiag = cpu_to_le32(imp->im_nextiag);
	dinom_le->in_numinos = cpu_to_le32(atomic_read(&imp->im_numinos));
	dinom_le->in_numfree = cpu_to_le32(atomic_read(&imp->im_numfree));
	dinom_le->in_nbperiext = cpu_to_le32(imp->im_nbperiext);
	dinom_le->in_l2nbperiext = cpu_to_le32(imp->im_l2nbperiext);
	for (index = 0; index < MAXAG; index++) {
		dinom_le->in_agctl[index].inofree =
		    cpu_to_le32(imp->im_agctl[index].inofree);
		dinom_le->in_agctl[index].extfree =
		    cpu_to_le32(imp->im_agctl[index].extfree);
		dinom_le->in_agctl[index].numinos =
		    cpu_to_le32(imp->im_agctl[index].numinos);
		dinom_le->in_agctl[index].numfree =
		    cpu_to_le32(imp->im_agctl[index].numfree);
	}

	/* write out the control structure */
	write_metapage(mp);

	/*
	 * write out dirty pages of imap
	 */
	fsync_inode_data_buffers(ipimap);

	diWriteSpecial(ipimap, 0);

	return (0);
}


/*
 * NAME:        diRead()
 *
 * FUNCTION:    initialize an incore inode from disk.
 *
 *		on entry, the specifed incore inode should itself
 *		specify the disk inode number corresponding to the
 *		incore inode (i.e. i_number should be initialized).
 *		
 *		this routine handles incore inode initialization for
 *		both "special" and "regular" inodes.  special inodes
 *		are those required early in the mount process and
 *	        require special handling since much of the file system
 *		is not yet initialized.  these "special" inodes are
 *		identified by a NULL inode map inode pointer and are
 *		actually initialized by a call to diReadSpecial().
 *		
 *		for regular inodes, the iag describing the disk inode
 *		is read from disk to determine the inode extent address
 *		for the disk inode.  with the inode extent address in
 *		hand, the page of the extent that contains the disk
 *		inode is read and the disk inode is copied to the
 *		incore inode.
 *
 * PARAMETERS:
 *      ip  -  pointer to incore inode to be initialized from disk.
 *
 * RETURN VALUES:
 *      0       - success
 *      -EIO  	- i/o error.
 *      -ENOMEM	- insufficient memory
 *      
 */
int diRead(struct inode *ip)
{
	struct jfs_sb_info *sbi = JFS_SBI(ip->i_sb);
	int iagno, ino, extno, rc;
	struct inode *ipimap;
	struct dinode *dp;
	struct iag *iagp;
	struct metapage *mp;
	s64 blkno, agstart;
	struct inomap *imap;
	int block_offset;
	int inodes_left;
	uint pageno;
	int rel_inode;

	jfs_info("diRead: ino = %ld", ip->i_ino);

	ipimap = sbi->ipimap;
	JFS_IP(ip)->ipimap = ipimap;

	/* determine the iag number for this inode (number) */
	iagno = INOTOIAG(ip->i_ino);

	/* read the iag */
	imap = JFS_IP(ipimap)->i_imap;
	IREAD_LOCK(ipimap);
	rc = diIAGRead(imap, iagno, &mp);
	IREAD_UNLOCK(ipimap);
	if (rc) {
		jfs_err("diRead: diIAGRead returned %d", rc);
		return (rc);
	}

	iagp = (struct iag *) mp->data;

	/* determine inode extent that holds the disk inode */
	ino = ip->i_ino & (INOSPERIAG - 1);
	extno = ino >> L2INOSPEREXT;

	if ((lengthPXD(&iagp->inoext[extno]) != imap->im_nbperiext) ||
	    (addressPXD(&iagp->inoext[extno]) == 0)) {
		release_metapage(mp);
		return -ESTALE;
	}

	/* get disk block number of the page within the inode extent
	 * that holds the disk inode.
	 */
	blkno = INOPBLK(&iagp->inoext[extno], ino, sbi->l2nbperpage);

	/* get the ag for the iag */
	agstart = le64_to_cpu(iagp->agstart);

	release_metapage(mp);

	rel_inode = (ino & (INOSPERPAGE - 1));
	pageno = blkno >> sbi->l2nbperpage;

	if ((block_offset = ((u32) blkno & (sbi->nbperpage - 1)))) {
		/*
		 * OS/2 didn't always align inode extents on page boundaries
		 */
		inodes_left =
		     (sbi->nbperpage - block_offset) << sbi->l2niperblk;

		if (rel_inode < inodes_left)
			rel_inode += block_offset << sbi->l2niperblk;
		else {
			pageno += 1;
			rel_inode -= inodes_left;
		}
	}

	/* read the page of disk inode */
	mp = read_metapage(ipimap, pageno << sbi->l2nbperpage, PSIZE, 1);
	if (mp == 0) {
		jfs_err("diRead: read_metapage failed");
		return -EIO;
	}

	/* locate the the disk inode requested */
	dp = (struct dinode *) mp->data;
	dp += rel_inode;

	if (ip->i_ino != le32_to_cpu(dp->di_number)) {
		jfs_error(ip->i_sb, "diRead: i_ino != di_number");
		rc = -EIO;
	} else if (le32_to_cpu(dp->di_nlink) == 0)
		rc = -ESTALE;
	else
		/* copy the disk inode to the in-memory inode */
		rc = copy_from_dinode(dp, ip);

	release_metapage(mp);

	/* set the ag for the inode */
	JFS_IP(ip)->agno = BLKTOAG(agstart, sbi);
	JFS_IP(ip)->active_ag = -1;

	return (rc);
}


/*
 * NAME:        diReadSpecial()
 *
 * FUNCTION:    initialize a 'special' inode from disk.
 *
 *		this routines handles aggregate level inodes.  The
 *		inode cache cannot differentiate between the
 *		aggregate inodes and the filesystem inodes, so we
 *		handle these here.  We don't actually use the aggregate
 *	        inode map, since these inodes are at a fixed location
 *		and in some cases the aggregate inode map isn't initialized
 *		yet.
 *
 * PARAMETERS:
 *      sb - filesystem superblock
 *	inum - aggregate inode number
 *	secondary - 1 if secondary aggregate inode table
 *
 * RETURN VALUES:
 *      new inode	- success
 *      NULL		- i/o error.
 */
struct inode *diReadSpecial(struct super_block *sb, ino_t inum, int secondary)
{
	struct jfs_sb_info *sbi = JFS_SBI(sb);
	uint address;
	struct dinode *dp;
	struct inode *ip;
	struct metapage *mp;
	int rc;

	ip = new_inode(sb);
	if (ip == NULL) {
		jfs_err("diReadSpecial: new_inode returned NULL!");
		return ip;
	}

	rc = alloc_jfs_inode(ip);
	if (rc) {
		make_bad_inode(ip);
		iput(ip);
		return NULL;
	}

	if (secondary) {
		address = addressPXD(&sbi->ait2) >> sbi->l2nbperpage;
		JFS_IP(ip)->ipimap = sbi->ipaimap2;
	} else {
		address = AITBL_OFF >> L2PSIZE;
		JFS_IP(ip)->ipimap = sbi->ipaimap;
	}

	ASSERT(inum < INOSPEREXT);