[apple/xnu.git] / bsd / ufs / ffs / ffs_balloc.c

/*
 * Copyright (c) 2000-2002 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_LICENSE_HEADER_START@
 * 
 * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this
 * file.
 * 
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_LICENSE_HEADER_END@
 */
/* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
/*
 * Copyright (c) 1982, 1986, 1989, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)ffs_balloc.c	8.8 (Berkeley) 6/16/95
 */

#include <rev_endian_fs.h>
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/buf.h>
#include <sys/proc.h>
#include <sys/file.h>
#include <sys/vnode.h>
#include <sys/ubc.h>
#include <sys/quota.h>

#if REV_ENDIAN_FS
#include <sys/mount.h>
#endif /* REV_ENDIAN_FS */

#include <sys/vm.h>

#include <ufs/ufs/quota.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/ufs_extern.h>

#include <ufs/ffs/fs.h>
#include <ufs/ffs/ffs_extern.h>

#if REV_ENDIAN_FS
#include <ufs/ufs/ufs_byte_order.h>
#include <architecture/byte_order.h>
#endif /* REV_ENDIAN_FS */

/*
 * Balloc defines the structure of file system storage
 * by allocating the physical blocks on a device given
 * the inode and the logical block number in a file.
 */
ffs_balloc(ip, lbn, size, cred, bpp, flags, blk_alloc)
	register struct inode *ip;
	register ufs_daddr_t lbn;
	int size;
	struct ucred *cred;
	struct buf **bpp;
	int flags;
	int * blk_alloc;
{
	register struct fs *fs;
	register ufs_daddr_t nb;
	struct buf *bp, *nbp;
	struct vnode *vp = ITOV(ip);
	struct indir indirs[NIADDR + 2];
	ufs_daddr_t newb, *bap, pref;
	int deallocated, osize, nsize, num, i, error;
	ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
	int devBlockSize=0;
	int alloc_buffer = 1;
#if REV_ENDIAN_FS
	struct mount *mp=vp->v_mount;
	int rev_endian=(mp->mnt_flag & MNT_REVEND);
#endif /* REV_ENDIAN_FS */

	*bpp = NULL;
	if (lbn < 0)
		return (EFBIG);
	fs = ip->i_fs;
	if (flags & B_NOBUFF) 
		alloc_buffer = 0;

	if (blk_alloc)
		*blk_alloc = 0;

	/*
	 * If the next write will extend the file into a new block,
	 * and the file is currently composed of a fragment
	 * this fragment has to be extended to be a full block.
	 */
	nb = lblkno(fs, ip->i_size);
	if (nb < NDADDR && nb < lbn) {
		/* the filesize prior to this write  can fit in direct 
		 * blocks (ie.  fragmentaion is possibly done)
		 * we are now extending the file write beyond 
		 * the block which has end of file prior to this write 
		 */
		osize = blksize(fs, ip, nb); 
		/* osize gives disk allocated size in the last block. It is 
		 * either in fragments or a file system block size */
		if (osize < fs->fs_bsize && osize > 0) {
			/* few fragments are already allocated,since the
			 * current extends beyond this block 
			 * allocate the complete block as fragments are only
			 * in last block
			 */
			error = ffs_realloccg(ip, nb,
				ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
				osize, (int)fs->fs_bsize, cred, &bp);
			if (error)
				return (error);
			/* adjust the innode size we just grew */
			/* it is in nb+1 as nb starts from 0 */
			ip->i_size = (nb + 1) * fs->fs_bsize;
			if (UBCISVALID(vp))
				ubc_setsize(vp, (off_t)ip->i_size); /* XXX check error */
			ip->i_db[nb] = dbtofsb(fs, bp->b_blkno);
			ip->i_flag |= IN_CHANGE | IN_UPDATE;
			if ((flags & B_SYNC) || (!alloc_buffer)) {
				if (!alloc_buffer) 
					SET(bp->b_flags, B_NOCACHE);
				bwrite(bp);
			} else
				bdwrite(bp);
			/* note that bp is already released here */
		}
	}
	/*
	 * The first NDADDR blocks are direct blocks
	 */
	if (lbn < NDADDR) {
		nb = ip->i_db[lbn];
		if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
			if (alloc_buffer) {
			error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp);
			if (error) {
				brelse(bp);
				return (error);
			}
			*bpp = bp;
			}
			return (0);
		}
		if (nb != 0) {
			/*
			 * Consider need to reallocate a fragment.
			 */
			osize = fragroundup(fs, blkoff(fs, ip->i_size));
			nsize = fragroundup(fs, size);
			if (nsize <= osize) {
				if (alloc_buffer) {
				error = bread(vp, lbn, osize, NOCRED, &bp);
				if (error) {
					brelse(bp);
					return (error);
				}
				ip->i_flag |= IN_CHANGE | IN_UPDATE;
				*bpp = bp;
				return (0);
				}
				else {
					ip->i_flag |= IN_CHANGE | IN_UPDATE;
					return (0);
				}
			} else {
				error = ffs_realloccg(ip, lbn,
				    ffs_blkpref(ip, lbn, (int)lbn,
					&ip->i_db[0]), osize, nsize, cred, &bp);
				if (error)
					return (error);
				ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
				ip->i_flag |= IN_CHANGE | IN_UPDATE;
				if(!alloc_buffer) {
					SET(bp->b_flags, B_NOCACHE);
					if (flags & B_SYNC)
						bwrite(bp);
					else
						bdwrite(bp);
				 } else
					*bpp = bp;
				return (0);

			}
		} else {
			if (ip->i_size < (lbn + 1) * fs->fs_bsize)
				nsize = fragroundup(fs, size);
			else
				nsize = fs->fs_bsize;
			error = ffs_alloc(ip, lbn,
			    ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
			    nsize, cred, &newb);
			if (error)
				return (error);
			if (alloc_buffer) {
			bp = getblk(vp, lbn, nsize, 0, 0, BLK_WRITE);
			bp->b_blkno = fsbtodb(fs, newb);
			if (flags & B_CLRBUF)
				clrbuf(bp);
			}
			ip->i_db[lbn] = newb;
			ip->i_flag |= IN_CHANGE | IN_UPDATE;
			if (blk_alloc) {
				*blk_alloc = nsize;
			}
			if (alloc_buffer)
				*bpp = bp;
			return (0);
		}
	}
	/*
	 * Determine the number of levels of indirection.
	 */
	pref = 0;
	if (error = ufs_getlbns(vp, lbn, indirs, &num))
		return(error);
#if DIAGNOSTIC
	if (num < 1)
		panic ("ffs_balloc: ufs_bmaparray returned indirect block");
#endif
	/*
	 * Fetch the first indirect block allocating if necessary.
	 */
	--num;
	nb = ip->i_ib[indirs[0].in_off];
	allocib = NULL;
	allocblk = allociblk;
	if (nb == 0) {
		pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	        if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
		    cred, &newb))
			return (error);
		nb = newb;
		*allocblk++ = nb;
		bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
		bp->b_blkno = fsbtodb(fs, nb);
		clrbuf(bp);
		/*
		 * Write synchronously conditional on mount flags.
		 */
		if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
			error = 0;
			bdwrite(bp);
		} else if ((error = bwrite(bp)) != 0) {
			goto fail;
		}
		allocib = &ip->i_ib[indirs[0].in_off];
		*allocib = nb;
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
	}
	/*
	 * Fetch through the indirect blocks, allocating as necessary.
	 */
	for (i = 1;;) {
		error = meta_bread(vp,
		    indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
		if (error) {
			brelse(bp);
			goto fail;
		}
		bap = (ufs_daddr_t *)bp->b_data;
#if	REV_ENDIAN_FS
	if (rev_endian)
		nb = NXSwapLong(bap[indirs[i].in_off]);
	else {
#endif	/* REV_ENDIAN_FS */
		nb = bap[indirs[i].in_off];
#if REV_ENDIAN_FS
	}
#endif /* REV_ENDIAN_FS */
		if (i == num)
			break;
		i += 1;
		if (nb != 0) {
			brelse(bp);
			continue;
		}
		if (pref == 0)
			pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
		if (error =
		    ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
			brelse(bp);
			goto fail;
		}
		nb = newb;
		*allocblk++ = nb;
		nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
		nbp->b_blkno = fsbtodb(fs, nb);
		clrbuf(nbp);
		/*
		 * Write synchronously conditional on mount flags.
		 */
		if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
			error = 0;
			bdwrite(nbp);
		} else if (error = bwrite(nbp)) {
			brelse(bp);
			goto fail;
		}
#if	REV_ENDIAN_FS
	if (rev_endian)
		bap[indirs[i - 1].in_off] = NXSwapLong(nb);
	else {
#endif	/* REV_ENDIAN_FS */
		bap[indirs[i - 1].in_off] = nb;
#if	REV_ENDIAN_FS
	}
#endif	/* REV_ENDIAN_FS */
		/*
		 * If required, write synchronously, otherwise use
		 * delayed write.
		 */
		if (flags & B_SYNC) {
			bwrite(bp);
		} else {
			bdwrite(bp);
		}
	}
	/*
	 * Get the data block, allocating if necessary.
	 */
	if (nb == 0) {
		pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
		if (error = ffs_alloc(ip,
		    lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
			brelse(bp);
			goto fail;
		}
		nb = newb;
		*allocblk++ = nb;
#if	REV_ENDIAN_FS
	if (rev_endian)
		bap[indirs[i].in_off] = NXSwapLong(nb);
	else {
#endif	/* REV_ENDIAN_FS */
		bap[indirs[i].in_off] = nb;
#if	REV_ENDIAN_FS
	}
#endif	/* REV_ENDIAN_FS */
		/*
		 * If required, write synchronously, otherwise use
		 * delayed write.
		 */
		if ((flags & B_SYNC)) {
			bwrite(bp);
		} else {
			bdwrite(bp);
		}
		if(alloc_buffer ) {
		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, BLK_WRITE);
		nbp->b_blkno = fsbtodb(fs, nb);
		if (flags & B_CLRBUF)
			clrbuf(nbp);
		}
		if (blk_alloc) {
			*blk_alloc = fs->fs_bsize;
		}
		if(alloc_buffer) 
			*bpp = nbp;

		return (0);
	}
	brelse(bp);
	if (alloc_buffer) {
	if (flags & B_CLRBUF) {
		error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp);
		if (error) {
			brelse(nbp);
			goto fail;
		}
	} else {
		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, BLK_WRITE);
		nbp->b_blkno = fsbtodb(fs, nb);
	}
	*bpp = nbp;
	}
	return (0);
fail:
	/*
	 * If we have failed part way through block allocation, we
	 * have to deallocate any indirect blocks that we have allocated.
	 */
	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
		ffs_blkfree(ip, *blkp, fs->fs_bsize);
		deallocated += fs->fs_bsize;
	}
	if (allocib != NULL)
		*allocib = 0;
	if (deallocated) {
	VOP_DEVBLOCKSIZE(ip->i_devvp,&devBlockSize);

#if QUOTA
		/*
		 * Restore user's disk quota because allocation failed.
		 */
		(void) chkdq(ip, (int64_t)-deallocated, cred, FORCE);
#endif /* QUOTA */
		ip->i_blocks -= btodb(deallocated, devBlockSize);
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
	}
	return (error);
}

/*
 * ffs_blkalloc allocates a disk block for ffs_pageout(), as a consequence
 * it does no breads (that could lead to deadblock as the page may be already
 * marked busy as it is being paged out. Also important to note that we are not
 * growing the file in pageouts. So ip->i_size  cannot increase by this call
 * due to the way UBC works.  
 * This code is derived from ffs_balloc and many cases of that are  dealt
 * in ffs_balloc are not applicable here 
 * Do not call with B_CLRBUF flags as this should only be called only 
 * from pageouts
 */
ffs_blkalloc(ip, lbn, size, cred, flags)
	register struct inode *ip;
	ufs_daddr_t lbn;
	int size;
	struct ucred *cred;
	int flags;
{
	register struct fs *fs;
	register ufs_daddr_t nb;
	struct buf *bp, *nbp;
	struct vnode *vp = ITOV(ip);
	struct indir indirs[NIADDR + 2];
	ufs_daddr_t newb, *bap, pref;
	int deallocated, osize, nsize, num, i, error;
	ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
	int devBlockSize=0;
#if REV_ENDIAN_FS
	struct mount *mp=vp->v_mount;
	int rev_endian=(mp->mnt_flag & MNT_REVEND);
#endif /* REV_ENDIAN_FS */

	fs = ip->i_fs;

	if(size > fs->fs_bsize)
		panic("ffs_blkalloc: too large for allocation");

	/*
	 * If the next write will extend the file into a new block,
	 * and the file is currently composed of a fragment
	 * this fragment has to be extended to be a full block.
	 */
	nb = lblkno(fs, ip->i_size);
	if (nb < NDADDR && nb < lbn) {
		panic("ffs_blkalloc():cannot extend file: i_size %d, lbn %d", ip->i_size, lbn);
	}
	/*
	 * The first NDADDR blocks are direct blocks
	 */
	if (lbn < NDADDR) {
		nb = ip->i_db[lbn];
		if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
		/* TBD: trivial case; the block  is already allocated */
			return (0);
		}
		if (nb != 0) {
			/*
			 * Consider need to reallocate a fragment.
			 */
			osize = fragroundup(fs, blkoff(fs, ip->i_size));
			nsize = fragroundup(fs, size);
			if (nsize > osize) {
				panic("ffs_allocblk: trying to extend a fragment");
			}
			return(0);
		} else {
			if (ip->i_size < (lbn + 1) * fs->fs_bsize)
				nsize = fragroundup(fs, size);
			else
				nsize = fs->fs_bsize;
			error = ffs_alloc(ip, lbn,
			    ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
			    nsize, cred, &newb);
			if (error)
				return (error);
			ip->i_db[lbn] = newb;
			ip->i_flag |= IN_CHANGE | IN_UPDATE;
			return (0);
		}
	}
	/*
	 * Determine the number of levels of indirection.
	 */
	pref = 0;
	if (error = ufs_getlbns(vp, lbn, indirs, &num))
		return(error);

	if(num == 0) {
		panic("ffs_blkalloc: file with direct blocks only"); 
	}

	/*
	 * Fetch the first indirect block allocating if necessary.
	 */
	--num;
	nb = ip->i_ib[indirs[0].in_off];
	allocib = NULL;
	allocblk = allociblk;
	if (nb == 0) {
		pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	        if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
		    cred, &newb))
			return (error);
		nb = newb;
		*allocblk++ = nb;
		bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
		bp->b_blkno = fsbtodb(fs, nb);
		clrbuf(bp);
		/*
		 * Write synchronously conditional on mount flags.
		 */
		if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
			error = 0;
			bdwrite(bp);
		} else if (error = bwrite(bp)) {
			goto fail;
		}
		allocib = &ip->i_ib[indirs[0].in_off];
		*allocib = nb;
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
	}
	/*
	 * Fetch through the indirect blocks, allocating as necessary.
	 */
	for (i = 1;;) {
		error = meta_bread(vp,
		    indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
		if (error) {
			brelse(bp);
			goto fail;
		}
		bap = (ufs_daddr_t *)bp->b_data;
#if	REV_ENDIAN_FS
	if (rev_endian)
		nb = NXSwapLong(bap[indirs[i].in_off]);
	else {
#endif	/* REV_ENDIAN_FS */
		nb = bap[indirs[i].in_off];
#if REV_ENDIAN_FS
	}
#endif /* REV_ENDIAN_FS */
		if (i == num)
			break;
		i += 1;
		if (nb != 0) {
			brelse(bp);
			continue;
		}
		if (pref == 0)
			pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
		if (error =
		    ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
			brelse(bp);
			goto fail;
		}
		nb = newb;
		*allocblk++ = nb;
		nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
		nbp->b_blkno = fsbtodb(fs, nb);
		clrbuf(nbp);
		/*
		 * Write synchronously conditional on mount flags.
		 */
		if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
			error = 0;
			bdwrite(nbp);
		} else if (error = bwrite(nbp)) {
			brelse(bp);
			goto fail;
		}
#if	REV_ENDIAN_FS
	if (rev_endian)
		bap[indirs[i - 1].in_off] = NXSwapLong(nb);
	else {
#endif	/* REV_ENDIAN_FS */
		bap[indirs[i - 1].in_off] = nb;
#if	REV_ENDIAN_FS
	}
#endif	/* REV_ENDIAN_FS */
		/*
		 * If required, write synchronously, otherwise use
		 * delayed write.
		 */
		if (flags & B_SYNC) {
			bwrite(bp);
		} else {
			bdwrite(bp);
		}
	}
	/*
	 * Get the data block, allocating if necessary.
	 */
	if (nb == 0) {
		pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
		if (error = ffs_alloc(ip,
		    lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
			brelse(bp);
			goto fail;
		}
		nb = newb;
		*allocblk++ = nb;
#if	REV_ENDIAN_FS
	if (rev_endian)
		bap[indirs[i].in_off] = NXSwapLong(nb);
	else {
#endif	/* REV_ENDIAN_FS */
		bap[indirs[i].in_off] = nb;
#if	REV_ENDIAN_FS
	}
#endif	/* REV_ENDIAN_FS */
		/*
		 * If required, write synchronously, otherwise use
		 * delayed write.
		 */
		if (flags & B_SYNC) {
			bwrite(bp);
		} else {
			bdwrite(bp);
		}
		return (0);
	}
	brelse(bp);
	return (0);
fail:
	/*
	 * If we have failed part way through block allocation, we
	 * have to deallocate any indirect blocks that we have allocated.
	 */
	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
		ffs_blkfree(ip, *blkp, fs->fs_bsize);
		deallocated += fs->fs_bsize;
	}
	if (allocib != NULL)
		*allocib = 0;
	if (deallocated) {
	VOP_DEVBLOCKSIZE(ip->i_devvp,&devBlockSize);

#if QUOTA
		/*
		 * Restore user's disk quota because allocation failed.
		 */
		(void) chkdq(ip, (int64_t)-deallocated, cred, FORCE);
#endif /* QUOTA */
		ip->i_blocks -= btodb(deallocated, devBlockSize);
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
	}
	return (error);
}
Commit	Line	Data
1c79356b	1	/*
55e303ae	2	* Copyright (c) 2000-2002 Apple Computer, Inc. All rights reserved.
1c79356b A	3	*
	4	* @APPLE_LICENSE_HEADER_START@
	5	*
43866e37	6	* Copyright (c) 1999-2003 Apple Computer, Inc. All Rights Reserved.
1c79356b	7	*
43866e37 A	8	* This file contains Original Code and/or Modifications of Original Code
	9	* as defined in and that are subject to the Apple Public Source License
	10	* Version 2.0 (the 'License'). You may not use this file except in
	11	* compliance with the License. Please obtain a copy of the License at
	12	* http://www.opensource.apple.com/apsl/ and read it before using this
	13	* file.
	14	*
	15	* The Original Code and all software distributed under the License are
	16	* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
1c79356b A	17	* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
1c79356b A	18	* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
43866e37 A	19	* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
	20	* Please see the License for the specific language governing rights and
	21	* limitations under the License.
1c79356b A	22	*
	23	* @APPLE_LICENSE_HEADER_END@
	24	*/
	25	/* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
	26	/*
	27	* Copyright (c) 1982, 1986, 1989, 1993
	28	* The Regents of the University of California. All rights reserved.
	29	*
	30	* Redistribution and use in source and binary forms, with or without
	31	* modification, are permitted provided that the following conditions
	32	* are met:
	33	* 1. Redistributions of source code must retain the above copyright
	34	* notice, this list of conditions and the following disclaimer.
	35	* 2. Redistributions in binary form must reproduce the above copyright
	36	* notice, this list of conditions and the following disclaimer in the
	37	* documentation and/or other materials provided with the distribution.
	38	* 3. All advertising materials mentioning features or use of this software
	39	* must display the following acknowledgement:
	40	* This product includes software developed by the University of
	41	* California, Berkeley and its contributors.
	42	* 4. Neither the name of the University nor the names of its contributors
	43	* may be used to endorse or promote products derived from this software
	44	* without specific prior written permission.
	45	*
	46	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
	47	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	48	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	49	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
	50	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	51	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	52	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	53	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	54	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	55	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	56	* SUCH DAMAGE.
	57	*
	58	* @(#)ffs_balloc.c 8.8 (Berkeley) 6/16/95
	59	*/
	60
	61	#include <rev_endian_fs.h>
	62	#include <sys/param.h>
	63	#include <sys/systm.h>
	64	#include <sys/buf.h>
	65	#include <sys/proc.h>
	66	#include <sys/file.h>
	67	#include <sys/vnode.h>
	68	#include <sys/ubc.h>
9bccf70c A	69	#include <sys/quota.h>
9bccf70c A	70
1c79356b A	71	#if REV_ENDIAN_FS
	72	#include <sys/mount.h>
	73	#endif /* REV_ENDIAN_FS */
	74
	75	#include <sys/vm.h>
	76
	77	#include <ufs/ufs/quota.h>
	78	#include <ufs/ufs/inode.h>
	79	#include <ufs/ufs/ufs_extern.h>
	80
	81	#include <ufs/ffs/fs.h>
	82	#include <ufs/ffs/ffs_extern.h>
	83
	84	#if REV_ENDIAN_FS
	85	#include <ufs/ufs/ufs_byte_order.h>
	86	#include <architecture/byte_order.h>
	87	#endif /* REV_ENDIAN_FS */
	88
	89	/*
	90	* Balloc defines the structure of file system storage
	91	* by allocating the physical blocks on a device given
	92	* the inode and the logical block number in a file.
	93	*/
	94	ffs_balloc(ip, lbn, size, cred, bpp, flags, blk_alloc)
	95	register struct inode *ip;
	96	register ufs_daddr_t lbn;
	97	int size;
	98	struct ucred *cred;
	99	struct buf **bpp;
	100	int flags;
	101	int * blk_alloc;
	102	{
	103	register struct fs *fs;
	104	register ufs_daddr_t nb;
	105	struct buf bp, nbp;
	106	struct vnode *vp = ITOV(ip);
	107	struct indir indirs[NIADDR + 2];
	108	ufs_daddr_t newb, *bap, pref;
	109	int deallocated, osize, nsize, num, i, error;
	110	ufs_daddr_t allocib, blkp, *allocblk, allociblk[NIADDR + 1];
	111	int devBlockSize=0;
	112	int alloc_buffer = 1;
	113	#if REV_ENDIAN_FS
	114	struct mount *mp=vp->v_mount;
	115	int rev_endian=(mp->mnt_flag & MNT_REVEND);
	116	#endif /* REV_ENDIAN_FS */
	117
	118	*bpp = NULL;
	119	if (lbn < 0)
	120	return (EFBIG);
	121	fs = ip->i_fs;
	122	if (flags & B_NOBUFF)
	123	alloc_buffer = 0;
	124
	125	if (blk_alloc)
	126	*blk_alloc = 0;
	127
	128	/*
	129	* If the next write will extend the file into a new block,
	130	* and the file is currently composed of a fragment
	131	* this fragment has to be extended to be a full block.
	132	*/
	133	nb = lblkno(fs, ip->i_size);
	134	if (nb < NDADDR && nb < lbn) {
135	/* the filesize prior to this write can fit in direct
136	* blocks (ie. fragmentaion is possibly done)
137	* we are now extending the file write beyond
138	* the block which has end of file prior to this write
139	*/
140	osize = blksize(fs, ip, nb);
141	/* osize gives disk allocated size in the last block. It is
142	* either in fragments or a file system block size */
143	if (osize < fs->fs_bsize && osize > 0) {
144	/* few fragments are already allocated,since the
145	* current extends beyond this block
146	* allocate the complete block as fragments are only
147	* in last block
148	*/
149	error = ffs_realloccg(ip, nb,
150	ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
151	osize, (int)fs->fs_bsize, cred, &bp);
152	if (error)
153	return (error);
154	/* adjust the innode size we just grew */
155	/* it is in nb+1 as nb starts from 0 */
156	ip->i_size = (nb + 1) * fs->fs_bsize;
157	if (UBCISVALID(vp))
158	ubc_setsize(vp, (off_t)ip->i_size); /* XXX check error */
159	ip->i_db[nb] = dbtofsb(fs, bp->b_blkno);
160	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
161	if ((flags & B_SYNC) \|\| (!alloc_buffer)) {
162	if (!alloc_buffer)
55e303ae	163	SET(bp->b_flags, B_NOCACHE);
1c79356b A	164	bwrite(bp);
1c79356b A	165	} else
55e303ae	166	bdwrite(bp);
1c79356b A	167	/* note that bp is already released here */
	168	}
	169	}
	170	/*
	171	* The first NDADDR blocks are direct blocks
	172	*/
	173	if (lbn < NDADDR) {
	174	nb = ip->i_db[lbn];
	175	if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
	176	if (alloc_buffer) {
	177	error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp);
	178	if (error) {
	179	brelse(bp);
	180	return (error);
	181	}
	182	*bpp = bp;
	183	}
	184	return (0);
	185	}
	186	if (nb != 0) {
	187	/*
	188	* Consider need to reallocate a fragment.
	189	*/
	190	osize = fragroundup(fs, blkoff(fs, ip->i_size));
	191	nsize = fragroundup(fs, size);
	192	if (nsize <= osize) {
	193	if (alloc_buffer) {
	194	error = bread(vp, lbn, osize, NOCRED, &bp);
	195	if (error) {
	196	brelse(bp);
	197	return (error);
	198	}
	199	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	200	*bpp = bp;
	201	return (0);
	202	}
	203	else {
	204	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	205	return (0);
	206	}
	207	} else {
	208	error = ffs_realloccg(ip, lbn,
	209	ffs_blkpref(ip, lbn, (int)lbn,
	210	&ip->i_db[0]), osize, nsize, cred, &bp);
	211	if (error)
	212	return (error);
	213	ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
	214	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
55e303ae A	215	if(!alloc_buffer) {
	216	SET(bp->b_flags, B_NOCACHE);
	217	if (flags & B_SYNC)
	218	bwrite(bp);
	219	else
	220	bdwrite(bp);
1c79356b A	221	} else
	222	*bpp = bp;
	223	return (0);
	224
	225	}
	226	} else {
	227	if (ip->i_size < (lbn + 1) * fs->fs_bsize)
	228	nsize = fragroundup(fs, size);
	229	else
	230	nsize = fs->fs_bsize;
	231	error = ffs_alloc(ip, lbn,
	232	ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
	233	nsize, cred, &newb);
	234	if (error)
	235	return (error);
	236	if (alloc_buffer) {
	237	bp = getblk(vp, lbn, nsize, 0, 0, BLK_WRITE);
	238	bp->b_blkno = fsbtodb(fs, newb);
	239	if (flags & B_CLRBUF)
	240	clrbuf(bp);
	241	}
	242	ip->i_db[lbn] = newb;
	243	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	244	if (blk_alloc) {
	245	*blk_alloc = nsize;
	246	}
	247	if (alloc_buffer)
	248	*bpp = bp;
	249	return (0);
	250	}
	251	}
	252	/*
	253	* Determine the number of levels of indirection.
	254	*/
	255	pref = 0;
	256	if (error = ufs_getlbns(vp, lbn, indirs, &num))
	257	return(error);
	258	#if DIAGNOSTIC
	259	if (num < 1)
55e303ae	260	panic ("ffs_balloc: ufs_bmaparray returned indirect block");
1c79356b A	261	#endif
	262	/*
	263	* Fetch the first indirect block allocating if necessary.
	264	*/
	265	--num;
	266	nb = ip->i_ib[indirs[0].in_off];
	267	allocib = NULL;
	268	allocblk = allociblk;
	269	if (nb == 0) {
	270	pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	271	if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
	272	cred, &newb))
	273	return (error);
	274	nb = newb;
	275	*allocblk++ = nb;
	276	bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
	277	bp->b_blkno = fsbtodb(fs, nb);
	278	clrbuf(bp);
	279	/*
55e303ae	280	* Write synchronously conditional on mount flags.
1c79356b	281	*/
55e303ae A	282	if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
	283	error = 0;
	284	bdwrite(bp);
	285	} else if ((error = bwrite(bp)) != 0) {
1c79356b	286	goto fail;
55e303ae	287	}
1c79356b A	288	allocib = &ip->i_ib[indirs[0].in_off];
	289	*allocib = nb;
	290	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	291	}
	292	/*
	293	* Fetch through the indirect blocks, allocating as necessary.
	294	*/
	295	for (i = 1;;) {
	296	error = meta_bread(vp,
	297	indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
	298	if (error) {
	299	brelse(bp);
	300	goto fail;
	301	}
	302	bap = (ufs_daddr_t *)bp->b_data;
	303	#if REV_ENDIAN_FS
	304	if (rev_endian)
	305	nb = NXSwapLong(bap[indirs[i].in_off]);
	306	else {
	307	#endif /* REV_ENDIAN_FS */
	308	nb = bap[indirs[i].in_off];
	309	#if REV_ENDIAN_FS
	310	}
	311	#endif /* REV_ENDIAN_FS */
	312	if (i == num)
	313	break;
	314	i += 1;
	315	if (nb != 0) {
	316	brelse(bp);
	317	continue;
	318	}
	319	if (pref == 0)
	320	pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	321	if (error =
	322	ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
	323	brelse(bp);
	324	goto fail;
	325	}
	326	nb = newb;
	327	*allocblk++ = nb;
	328	nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
	329	nbp->b_blkno = fsbtodb(fs, nb);
	330	clrbuf(nbp);
	331	/*
55e303ae	332	* Write synchronously conditional on mount flags.
1c79356b	333	*/
55e303ae A	334	if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
	335	error = 0;
	336	bdwrite(nbp);
	337	} else if (error = bwrite(nbp)) {
1c79356b A	338	brelse(bp);
	339	goto fail;
	340	}
	341	#if REV_ENDIAN_FS
	342	if (rev_endian)
	343	bap[indirs[i - 1].in_off] = NXSwapLong(nb);
	344	else {
	345	#endif /* REV_ENDIAN_FS */
	346	bap[indirs[i - 1].in_off] = nb;
	347	#if REV_ENDIAN_FS
	348	}
	349	#endif /* REV_ENDIAN_FS */
	350	/*
	351	* If required, write synchronously, otherwise use
	352	* delayed write.
	353	*/
	354	if (flags & B_SYNC) {
	355	bwrite(bp);
	356	} else {
	357	bdwrite(bp);
	358	}
	359	}
	360	/*
	361	* Get the data block, allocating if necessary.
	362	*/
	363	if (nb == 0) {
	364	pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
	365	if (error = ffs_alloc(ip,
	366	lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
	367	brelse(bp);
	368	goto fail;
	369	}
	370	nb = newb;
	371	*allocblk++ = nb;
	372	#if REV_ENDIAN_FS
	373	if (rev_endian)
	374	bap[indirs[i].in_off] = NXSwapLong(nb);
	375	else {
	376	#endif /* REV_ENDIAN_FS */
	377	bap[indirs[i].in_off] = nb;
	378	#if REV_ENDIAN_FS
	379	}
	380	#endif /* REV_ENDIAN_FS */
	381	/*
	382	* If required, write synchronously, otherwise use
	383	* delayed write.
	384	*/
	385	if ((flags & B_SYNC)) {
	386	bwrite(bp);
	387	} else {
	388	bdwrite(bp);
	389	}
	390	if(alloc_buffer ) {
	391	nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, BLK_WRITE);
	392	nbp->b_blkno = fsbtodb(fs, nb);
	393	if (flags & B_CLRBUF)
	394	clrbuf(nbp);
	395	}
	396	if (blk_alloc) {
	397	*blk_alloc = fs->fs_bsize;
	398	}
	399	if(alloc_buffer)
	400	*bpp = nbp;
	401
402	return (0);
403	}
404	brelse(bp);
405	if (alloc_buffer) {
406	if (flags & B_CLRBUF) {
407	error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp);
408	if (error) {
409	brelse(nbp);
410	goto fail;
411	}
412	} else {
413	nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, BLK_WRITE);
414	nbp->b_blkno = fsbtodb(fs, nb);
415	}
416	*bpp = nbp;
417	}
418	return (0);
419	fail:
420	/*
421	* If we have failed part way through block allocation, we
422	* have to deallocate any indirect blocks that we have allocated.
423	*/
424	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
425	ffs_blkfree(ip, *blkp, fs->fs_bsize);
426	deallocated += fs->fs_bsize;
427	}
428	if (allocib != NULL)
429	*allocib = 0;
430	if (deallocated) {
431	VOP_DEVBLOCKSIZE(ip->i_devvp,&devBlockSize);
432
433	#if QUOTA
434	/*
435	* Restore user's disk quota because allocation failed.
436	*/
9bccf70c	437	(void) chkdq(ip, (int64_t)-deallocated, cred, FORCE);
1c79356b A	438	#endif /* QUOTA */
	439	ip->i_blocks -= btodb(deallocated, devBlockSize);
	440	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	441	}
	442	return (error);
	443	}
	444
	445	/*
	446	* ffs_blkalloc allocates a disk block for ffs_pageout(), as a consequence
	447	* it does no breads (that could lead to deadblock as the page may be already
	448	* marked busy as it is being paged out. Also important to note that we are not
	449	* growing the file in pageouts. So ip->i_size cannot increase by this call
	450	* due to the way UBC works.
	451	* This code is derived from ffs_balloc and many cases of that are dealt
	452	* in ffs_balloc are not applicable here
	453	* Do not call with B_CLRBUF flags as this should only be called only
	454	* from pageouts
	455	*/
	456	ffs_blkalloc(ip, lbn, size, cred, flags)
	457	register struct inode *ip;
	458	ufs_daddr_t lbn;
	459	int size;
	460	struct ucred *cred;
	461	int flags;
	462	{
	463	register struct fs *fs;
	464	register ufs_daddr_t nb;
	465	struct buf bp, nbp;
	466	struct vnode *vp = ITOV(ip);
	467	struct indir indirs[NIADDR + 2];
	468	ufs_daddr_t newb, *bap, pref;
	469	int deallocated, osize, nsize, num, i, error;
	470	ufs_daddr_t allocib, blkp, *allocblk, allociblk[NIADDR + 1];
	471	int devBlockSize=0;
	472	#if REV_ENDIAN_FS
	473	struct mount *mp=vp->v_mount;
	474	int rev_endian=(mp->mnt_flag & MNT_REVEND);
	475	#endif /* REV_ENDIAN_FS */
	476
	477	fs = ip->i_fs;
	478
	479	if(size > fs->fs_bsize)
55e303ae	480	panic("ffs_blkalloc: too large for allocation");
1c79356b A	481
	482	/*
	483	* If the next write will extend the file into a new block,
	484	* and the file is currently composed of a fragment
	485	* this fragment has to be extended to be a full block.
	486	*/
	487	nb = lblkno(fs, ip->i_size);
	488	if (nb < NDADDR && nb < lbn) {
55e303ae	489	panic("ffs_blkalloc():cannot extend file: i_size %d, lbn %d", ip->i_size, lbn);
1c79356b A	490	}
	491	/*
	492	* The first NDADDR blocks are direct blocks
	493	*/
	494	if (lbn < NDADDR) {
	495	nb = ip->i_db[lbn];
	496	if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
	497	/* TBD: trivial case; the block is already allocated */
	498	return (0);
	499	}
	500	if (nb != 0) {
	501	/*
	502	* Consider need to reallocate a fragment.
	503	*/
	504	osize = fragroundup(fs, blkoff(fs, ip->i_size));
	505	nsize = fragroundup(fs, size);
	506	if (nsize > osize) {
55e303ae	507	panic("ffs_allocblk: trying to extend a fragment");
1c79356b A	508	}
	509	return(0);
	510	} else {
	511	if (ip->i_size < (lbn + 1) * fs->fs_bsize)
	512	nsize = fragroundup(fs, size);
	513	else
	514	nsize = fs->fs_bsize;
	515	error = ffs_alloc(ip, lbn,
	516	ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
	517	nsize, cred, &newb);
	518	if (error)
	519	return (error);
	520	ip->i_db[lbn] = newb;
	521	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	522	return (0);
	523	}
	524	}
	525	/*
	526	* Determine the number of levels of indirection.
	527	*/
	528	pref = 0;
	529	if (error = ufs_getlbns(vp, lbn, indirs, &num))
	530	return(error);
	531
	532	if(num == 0) {
55e303ae	533	panic("ffs_blkalloc: file with direct blocks only");
1c79356b A	534	}
	535
	536	/*
	537	* Fetch the first indirect block allocating if necessary.
	538	*/
	539	--num;
	540	nb = ip->i_ib[indirs[0].in_off];
	541	allocib = NULL;
	542	allocblk = allociblk;
	543	if (nb == 0) {
	544	pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	545	if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
	546	cred, &newb))
	547	return (error);
	548	nb = newb;
	549	*allocblk++ = nb;
	550	bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
	551	bp->b_blkno = fsbtodb(fs, nb);
	552	clrbuf(bp);
	553	/*
55e303ae	554	* Write synchronously conditional on mount flags.
1c79356b	555	*/
55e303ae A	556	if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
	557	error = 0;
	558	bdwrite(bp);
	559	} else if (error = bwrite(bp)) {
1c79356b	560	goto fail;
55e303ae	561	}
1c79356b A	562	allocib = &ip->i_ib[indirs[0].in_off];
	563	*allocib = nb;
	564	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	565	}
	566	/*
	567	* Fetch through the indirect blocks, allocating as necessary.
	568	*/
	569	for (i = 1;;) {
	570	error = meta_bread(vp,
	571	indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
	572	if (error) {
	573	brelse(bp);
	574	goto fail;
	575	}
	576	bap = (ufs_daddr_t *)bp->b_data;
	577	#if REV_ENDIAN_FS
	578	if (rev_endian)
	579	nb = NXSwapLong(bap[indirs[i].in_off]);
	580	else {
	581	#endif /* REV_ENDIAN_FS */
	582	nb = bap[indirs[i].in_off];
	583	#if REV_ENDIAN_FS
	584	}
	585	#endif /* REV_ENDIAN_FS */
	586	if (i == num)
	587	break;
	588	i += 1;
	589	if (nb != 0) {
	590	brelse(bp);
	591	continue;
	592	}
	593	if (pref == 0)
	594	pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	595	if (error =
	596	ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
	597	brelse(bp);
	598	goto fail;
	599	}
	600	nb = newb;
	601	*allocblk++ = nb;
	602	nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0, BLK_META);
	603	nbp->b_blkno = fsbtodb(fs, nb);
	604	clrbuf(nbp);
	605	/*
55e303ae	606	* Write synchronously conditional on mount flags.
1c79356b	607	*/
55e303ae A	608	if ((vp)->v_mount->mnt_flag & MNT_ASYNC) {
	609	error = 0;
	610	bdwrite(nbp);
	611	} else if (error = bwrite(nbp)) {
1c79356b A	612	brelse(bp);
	613	goto fail;
	614	}
	615	#if REV_ENDIAN_FS
	616	if (rev_endian)
	617	bap[indirs[i - 1].in_off] = NXSwapLong(nb);
	618	else {
	619	#endif /* REV_ENDIAN_FS */
	620	bap[indirs[i - 1].in_off] = nb;
	621	#if REV_ENDIAN_FS
	622	}
	623	#endif /* REV_ENDIAN_FS */
	624	/*
	625	* If required, write synchronously, otherwise use
	626	* delayed write.
	627	*/
	628	if (flags & B_SYNC) {
	629	bwrite(bp);
	630	} else {
	631	bdwrite(bp);
	632	}
	633	}
	634	/*
	635	* Get the data block, allocating if necessary.
	636	*/
	637	if (nb == 0) {
	638	pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
	639	if (error = ffs_alloc(ip,
	640	lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
	641	brelse(bp);
	642	goto fail;
	643	}
	644	nb = newb;
	645	*allocblk++ = nb;
	646	#if REV_ENDIAN_FS
	647	if (rev_endian)
	648	bap[indirs[i].in_off] = NXSwapLong(nb);
	649	else {
	650	#endif /* REV_ENDIAN_FS */
	651	bap[indirs[i].in_off] = nb;
	652	#if REV_ENDIAN_FS
	653	}
	654	#endif /* REV_ENDIAN_FS */
	655	/*
	656	* If required, write synchronously, otherwise use
	657	* delayed write.
	658	*/
	659	if (flags & B_SYNC) {
	660	bwrite(bp);
	661	} else {
	662	bdwrite(bp);
	663	}
	664	return (0);
	665	}
	666	brelse(bp);
	667	return (0);
	668	fail:
	669	/*
	670	* If we have failed part way through block allocation, we
	671	* have to deallocate any indirect blocks that we have allocated.
	672	*/
	673	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
	674	ffs_blkfree(ip, *blkp, fs->fs_bsize);
	675	deallocated += fs->fs_bsize;
676	}
677	if (allocib != NULL)
678	*allocib = 0;
679	if (deallocated) {
680	VOP_DEVBLOCKSIZE(ip->i_devvp,&devBlockSize);
681
682	#if QUOTA
683	/*
684	* Restore user's disk quota because allocation failed.
685	*/
9bccf70c	686	(void) chkdq(ip, (int64_t)-deallocated, cred, FORCE);
1c79356b A	687	#endif /* QUOTA */
	688	ip->i_blocks -= btodb(deallocated, devBlockSize);
	689	ip->i_flag \|= IN_CHANGE \| IN_UPDATE;
	690	}
	691	return (error);
	692	}