From nobody@FreeBSD.org  Wed Jul 27 14:40:41 2011
Return-Path: <nobody@FreeBSD.org>
Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34])
	by hub.freebsd.org (Postfix) with ESMTP id 9C7F5106566B
	for <freebsd-gnats-submit@FreeBSD.org>; Wed, 27 Jul 2011 14:40:41 +0000 (UTC)
	(envelope-from nobody@FreeBSD.org)
Received: from red.freebsd.org (red.freebsd.org [IPv6:2001:4f8:fff6::22])
	by mx1.freebsd.org (Postfix) with ESMTP id 81A568FC15
	for <freebsd-gnats-submit@FreeBSD.org>; Wed, 27 Jul 2011 14:40:41 +0000 (UTC)
Received: from red.freebsd.org (localhost [127.0.0.1])
	by red.freebsd.org (8.14.4/8.14.4) with ESMTP id p6REef8a025967
	for <freebsd-gnats-submit@FreeBSD.org>; Wed, 27 Jul 2011 14:40:41 GMT
	(envelope-from nobody@red.freebsd.org)
Received: (from nobody@localhost)
	by red.freebsd.org (8.14.4/8.14.4/Submit) id p6REefYq025966;
	Wed, 27 Jul 2011 14:40:41 GMT
	(envelope-from nobody)
Message-Id: <201107271440.p6REefYq025966@red.freebsd.org>
Date: Wed, 27 Jul 2011 14:40:41 GMT
From: Pedro Giffuni <giffunip@tutopia.com>
To: freebsd-gnats-submit@FreeBSD.org
Subject: fs/ext2fs: merge ext2_readwrite into ext2_vnops
X-Send-Pr-Version: www-3.1
X-GNATS-Notify:

>Number:         159232
>Category:       kern
>Synopsis:       [ext2fs] [patch] fs/ext2fs: merge ext2_readwrite into ext2_vnops
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    pfg
>State:          closed
>Quarter:        
>Keywords:       
>Date-Required:  
>Class:          change-request
>Submitter-Id:   current-users
>Arrival-Date:   Wed Jul 27 14:50:08 UTC 2011
>Closed-Date:    Thu Jan 05 01:51:49 UTC 2012
>Last-Modified:  Thu Jan 05 01:51:49 UTC 2012
>Originator:     Pedro Giffuni
>Release:        9.0-CURRENT
>Organization:
>Environment:
FreeBSD mogwai.giffuni.net 9.0-CURRENT FreeBSD 9.0-CURRENT #6: Sat Apr 30 01:37:57 PDT 2011     root@build9x64.pcbsd.org:/usr/obj/pcbsd-build90/fbsd-source/9.0/sys/PCBSD  amd64

>Description:
Follow the comment about removing obfuscations and instead of including ext2_readwrite, move the clustering related functions to ext2_vnops.c

This is consistent to what was done in UFS (SVN 101720) and as similar to
the code in Solaris.

It was suggested by bde@.
>How-To-Repeat:

>Fix:
After applying the patch, ext2_readwrite.c should/must be removed.

Patch attached with submission follows:

--- ext2fs.orig/ext2_vnops.c	2011-04-29 10:15:45.000000000 +0000
+++ ext2fs/ext2_vnops.c	2011-04-30 13:01:47.000000000 +0000
@@ -64,9 +64,13 @@
 #include <sys/file.h>
 
 #include <vm/vm.h>
+#include <vm/vm_page.h>
+#include <vm/vm_object.h>
 #include <vm/vm_extern.h>
 #include <vm/vnode_pager.h>
 
+#include "opt_directio.h"
+
 #include <fs/fifofs/fifo.h>
 
 #include <ufs/ufs/dir.h>
@@ -159,8 +163,6 @@
 	.vop_vptofh =		ext2_vptofh,
 };
 
-#include <fs/ext2fs/ext2_readwrite.c>
-
 /*
  * A virgin directory (no blushing please).
  * Note that the type and namlen fields are reversed relative to ext2.
@@ -1675,3 +1677,328 @@
 	vput(tvp);
 	return (error);
 }
+
+/*
+ * Vnode op for reading.
+ */
+static int
+ext2_read(ap)
+	struct vop_read_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	struct vnode *vp;
+	struct inode *ip;
+	struct uio *uio;
+	struct m_ext2fs *fs;
+	struct buf *bp;
+	daddr_t lbn, nextlbn;
+	off_t bytesinfile;
+	long size, xfersize, blkoffset;
+	int error, orig_resid, seqcount;
+	int ioflag;
+
+	vp = ap->a_vp;
+	uio = ap->a_uio;
+	ioflag = ap->a_ioflag;
+
+	seqcount = ap->a_ioflag >> IO_SEQSHIFT;
+	ip = VTOI(vp);
+
+#ifdef INVARIANTS
+	if (uio->uio_rw != UIO_READ)
+		panic("%s: mode", "ext2_read");
+
+	if (vp->v_type == VLNK) {
+		if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
+			panic("%s: short symlink", "ext2_read");
+	} else if (vp->v_type != VREG && vp->v_type != VDIR)
+		panic("%s: type %d", "ext2_read", vp->v_type);
+#endif
+	orig_resid = uio->uio_resid;
+	KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
+	if (orig_resid == 0)
+		return (0);
+	KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
+	fs = ip->i_e2fs;
+	if (uio->uio_offset < ip->i_size &&
+	    uio->uio_offset >= fs->e2fs_maxfilesize)
+	    	return (EOVERFLOW);
+
+	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
+		if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
+			break;
+		lbn = lblkno(fs, uio->uio_offset);
+		nextlbn = lbn + 1;
+		size = blksize(fs, ip, lbn);
+		blkoffset = blkoff(fs, uio->uio_offset);
+
+		xfersize = fs->e2fs_fsize - blkoffset;
+		if (uio->uio_resid < xfersize)
+			xfersize = uio->uio_resid;
+		if (bytesinfile < xfersize)
+			xfersize = bytesinfile;
+
+		if (lblktosize(fs, nextlbn) >= ip->i_size)
+			error = bread(vp, lbn, size, NOCRED, &bp);
+		else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0)
+			error = cluster_read(vp, ip->i_size, lbn, size,
+			    NOCRED, blkoffset + uio->uio_resid, seqcount, &bp);
+		else if (seqcount > 1) {
+			int nextsize = blksize(fs, ip, nextlbn);
+			error = breadn(vp, lbn,
+			    size, &nextlbn, &nextsize, 1, NOCRED, &bp);
+		} else
+			error = bread(vp, lbn, size, NOCRED, &bp);
+		if (error) {
+			brelse(bp);
+			bp = NULL;
+			break;
+		}
+
+		/*
+		 * If IO_DIRECT then set B_DIRECT for the buffer.  This
+		 * will cause us to attempt to release the buffer later on
+		 * and will cause the buffer cache to attempt to free the
+		 * underlying pages.
+		 */
+		if (ioflag & IO_DIRECT)
+			bp->b_flags |= B_DIRECT;
+
+		/*
+		 * We should only get non-zero b_resid when an I/O error
+		 * has occurred, which should cause us to break above.
+		 * However, if the short read did not cause an error,
+		 * then we want to ensure that we do not uiomove bad
+		 * or uninitialized data.
+		 */
+		size -= bp->b_resid;
+		if (size < xfersize) {
+			if (size == 0)
+				break;
+			xfersize = size;
+		}
+		error = uiomove((char *)bp->b_data + blkoffset,
+  			(int)xfersize, uio);
+		if (error)
+			break;
+
+		if (ioflag & (IO_VMIO|IO_DIRECT)) {
+			/*
+			 * If it's VMIO or direct I/O, then we don't
+			 * need the buf, mark it available for
+			 * freeing. If it's non-direct VMIO, the VM has
+			 * the data.
+			 */
+			bp->b_flags |= B_RELBUF;
+			brelse(bp);
+		} else {
+			/*
+			 * Otherwise let whoever
+			 * made the request take care of
+			 * freeing it. We just queue
+			 * it onto another list.
+			 */
+			bqrelse(bp);
+		}
+	}
+
+	/* 
+	 * This can only happen in the case of an error
+	 * because the loop above resets bp to NULL on each iteration
+	 * and on normal completion has not set a new value into it.
+	 * so it must have come from a 'break' statement
+	 */
+	if (bp != NULL) {
+		if (ioflag & (IO_VMIO|IO_DIRECT)) {
+			bp->b_flags |= B_RELBUF;
+			brelse(bp);
+		} else {
+			bqrelse(bp);
+		}
+	}
+
+	if ((error == 0 || uio->uio_resid != orig_resid) &&
+	    (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
+		ip->i_flag |= IN_ACCESS;
+	return (error);
+}
+
+/*
+ * Vnode op for writing.
+ */
+static int
+ext2_write(ap)
+	struct vop_write_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	struct vnode *vp;
+	struct uio *uio;
+	struct inode *ip;
+	struct m_ext2fs *fs;
+	struct buf *bp;
+	daddr_t lbn;
+	off_t osize;
+	int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
+
+	ioflag = ap->a_ioflag;
+	uio = ap->a_uio;
+	vp = ap->a_vp;
+
+	seqcount = ioflag >> IO_SEQSHIFT;
+	ip = VTOI(vp);
+
+#ifdef INVARIANTS
+	if (uio->uio_rw != UIO_WRITE)
+		panic("%s: mode", "ext2_write");
+#endif
+
+	switch (vp->v_type) {
+	case VREG:
+		if (ioflag & IO_APPEND)
+			uio->uio_offset = ip->i_size;
+		if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
+			return (EPERM);
+		/* FALLTHROUGH */
+	case VLNK:
+		break;
+	case VDIR:
+		/* XXX differs from ffs -- this is called from ext2_mkdir(). */
+		if ((ioflag & IO_SYNC) == 0)
+		panic("ext2_write: nonsync dir write");
+		break;
+	default:
+		panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
+		    vp->v_type, (intmax_t)uio->uio_offset,
+		    (intmax_t)uio->uio_resid);
+	}
+
+	KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
+	KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
+	fs = ip->i_e2fs;
+	if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
+		return (EFBIG);
+	/*
+	 * Maybe this should be above the vnode op call, but so long as
+	 * file servers have no limits, I don't think it matters.
+	 */
+	if (vn_rlimit_fsize(vp, uio, uio->uio_td))
+		return (EFBIG);
+
+	resid = uio->uio_resid;
+	osize = ip->i_size;
+	if (seqcount > BA_SEQMAX)
+		flags = BA_SEQMAX << BA_SEQSHIFT;
+	else
+		flags = seqcount << BA_SEQSHIFT;
+	if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
+		flags |= IO_SYNC;
+
+	for (error = 0; uio->uio_resid > 0;) {
+		lbn = lblkno(fs, uio->uio_offset);
+		blkoffset = blkoff(fs, uio->uio_offset);
+		xfersize = fs->e2fs_fsize - blkoffset;
+		if (uio->uio_resid < xfersize)
+			xfersize = uio->uio_resid;
+		if (uio->uio_offset + xfersize > ip->i_size)
+			vnode_pager_setsize(vp, uio->uio_offset + xfersize);
+
+                /*
+		 * We must perform a read-before-write if the transfer size
+		 * does not cover the entire buffer.
+                 */
+		if (fs->e2fs_bsize > xfersize)
+			flags |= BA_CLRBUF;
+		else
+			flags &= ~BA_CLRBUF;
+		error = ext2_balloc(ip, lbn, blkoffset + xfersize,
+		    ap->a_cred, &bp, flags);
+		if (error != 0)
+			break;
+
+		/*
+		 * If the buffer is not valid and we did not clear garbage
+		 * out above, we have to do so here even though the write
+		 * covers the entire buffer in order to avoid a mmap()/write
+		 * race where another process may see the garbage prior to
+		 * the uiomove() for a write replacing it.
+		 */
+		if ((bp->b_flags & B_CACHE) == 0 && fs->e2fs_bsize <= xfersize)
+			vfs_bio_clrbuf(bp);
+		if ((ioflag & (IO_SYNC|IO_INVAL)) == (IO_SYNC|IO_INVAL))
+			bp->b_flags |= B_NOCACHE;
+		if (uio->uio_offset + xfersize > ip->i_size)
+			ip->i_size = uio->uio_offset + xfersize;
+		size = blksize(fs, ip, lbn) - bp->b_resid;
+		if (size < xfersize)
+			xfersize = size;
+
+		error =
+		    uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
+		if (ioflag & (IO_VMIO|IO_DIRECT)) {
+			bp->b_flags |= B_RELBUF;
+		}
+
+		/*
+		 * If IO_SYNC each buffer is written synchronously.  Otherwise
+		 * if we have a severe page deficiency write the buffer
+		 * asynchronously.  Otherwise try to cluster, and if that
+		 * doesn't do it then either do an async write (if O_DIRECT),
+		 * or a delayed write (if not).
+		 */
+		if (ioflag & IO_SYNC) {
+			(void)bwrite(bp);
+		} else if (vm_page_count_severe() ||
+		    buf_dirty_count_severe() ||
+		    (ioflag & IO_ASYNC)) {
+			bp->b_flags |= B_CLUSTEROK;
+			bawrite(bp);
+		} else if (xfersize + blkoffset == fs->e2fs_fsize) {
+			if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
+				bp->b_flags |= B_CLUSTEROK;
+				cluster_write(vp, bp, ip->i_size, seqcount);
+			} else {
+				bawrite(bp);
+			}
+		} else if (ioflag & IO_DIRECT) {
+			bp->b_flags |= B_CLUSTEROK;
+			bawrite(bp);
+		} else {
+			bp->b_flags |= B_CLUSTEROK;
+			bdwrite(bp);
+		}
+		if (error || xfersize == 0)
+			break;
+	}
+	/*
+	 * If we successfully wrote any data, and we are not the superuser
+	 * we clear the setuid and setgid bits as a precaution against
+	 * tampering.
+	 */
+	if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid &&
+	    ap->a_cred) {
+		if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
+			ip->i_mode &= ~(ISUID | ISGID);
+	}
+	if (error) {
+		if (ioflag & IO_UNIT) {
+			(void)ext2_truncate(vp, osize,
+			    ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
+			uio->uio_offset -= resid - uio->uio_resid;
+			uio->uio_resid = resid;
+		}
+	}
+	if (uio->uio_resid != resid) {
+               ip->i_flag |= IN_CHANGE | IN_UPDATE;
+               if (ioflag & IO_SYNC)
+                       error = ext2_update(vp, 1);
+       }
+	return (error);
+}


>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: freebsd-bugs->freebsd-fs 
Responsible-Changed-By: linimon 
Responsible-Changed-When: Wed Jul 27 23:49:41 UTC 2011 
Responsible-Changed-Why:  
Over to maintainer(s). 

http://www.freebsd.org/cgi/query-pr.cgi?pr=159232 

From: dfilter@FreeBSD.ORG (dfilter service)
To: bug-followup@FreeBSD.org
Cc:  
Subject: Re: kern/159232: commit references a PR
Date: Wed, 14 Dec 2011 22:04:29 +0000 (UTC)

 Author: pfg
 Date: Wed Dec 14 22:04:14 2011
 New Revision: 228507
 URL: http://svn.freebsd.org/changeset/base/228507
 
 Log:
   Merge ext2_readwrite.c into ext2_vnops.c as done in UFS in r101720.
   
   This removes the obfuscations mentioned in ext2_readwrite and
   places the clustering functions in a location similar to other
   UFS-based implementations.
   
   No performance or functional changes are expected from
   this move.
   
   PR:		kern/159232
   Suggested by:	bde
   Approved by:	jhb (mentor)
   MFC after:	2 weeks
 
 Deleted:
   head/sys/fs/ext2fs/ext2_readwrite.c
 Modified:
   head/sys/fs/ext2fs/ext2_vnops.c
 
 Modified: head/sys/fs/ext2fs/ext2_vnops.c
 ==============================================================================
 --- head/sys/fs/ext2fs/ext2_vnops.c	Wed Dec 14 20:57:41 2011	(r228506)
 +++ head/sys/fs/ext2fs/ext2_vnops.c	Wed Dec 14 22:04:14 2011	(r228507)
 @@ -64,9 +64,13 @@
  #include <sys/file.h>
  
  #include <vm/vm.h>
 +#include <vm/vm_page.h>
 +#include <vm/vm_object.h>
  #include <vm/vm_extern.h>
  #include <vm/vnode_pager.h>
  
 +#include "opt_directio.h"
 +
  #include <fs/fifofs/fifo.h>
  
  #include <ufs/ufs/dir.h>
 @@ -159,8 +163,6 @@ struct vop_vector ext2_fifoops = {
  	.vop_vptofh =		ext2_vptofh,
  };
  
 -#include <fs/ext2fs/ext2_readwrite.c>
 -
  /*
   * A virgin directory (no blushing please).
   * Note that the type and namlen fields are reversed relative to ext2.
 @@ -1675,3 +1677,328 @@ bad:
  	vput(tvp);
  	return (error);
  }
 +
 +/*
 + * Vnode op for reading.
 + */
 +static int
 +ext2_read(ap)
 +	struct vop_read_args /* {
 +		struct vnode *a_vp;
 +		struct uio *a_uio;
 +		int a_ioflag;
 +		struct ucred *a_cred;
 +	} */ *ap;
 +{
 +	struct vnode *vp;
 +	struct inode *ip;
 +	struct uio *uio;
 +	struct m_ext2fs *fs;
 +	struct buf *bp;
 +	daddr_t lbn, nextlbn;
 +	off_t bytesinfile;
 +	long size, xfersize, blkoffset;
 +	int error, orig_resid, seqcount;
 +	int ioflag;
 +
 +	vp = ap->a_vp;
 +	uio = ap->a_uio;
 +	ioflag = ap->a_ioflag;
 +
 +	seqcount = ap->a_ioflag >> IO_SEQSHIFT;
 +	ip = VTOI(vp);
 +
 +#ifdef INVARIANTS
 +	if (uio->uio_rw != UIO_READ)
 +		panic("%s: mode", "ext2_read");
 +
 +	if (vp->v_type == VLNK) {
 +		if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
 +			panic("%s: short symlink", "ext2_read");
 +	} else if (vp->v_type != VREG && vp->v_type != VDIR)
 +		panic("%s: type %d", "ext2_read", vp->v_type);
 +#endif
 +	orig_resid = uio->uio_resid;
 +	KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
 +	if (orig_resid == 0)
 +		return (0);
 +	KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
 +	fs = ip->i_e2fs;
 +	if (uio->uio_offset < ip->i_size &&
 +	    uio->uio_offset >= fs->e2fs_maxfilesize)
 +	    	return (EOVERFLOW);
 +
 +	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
 +		if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
 +			break;
 +		lbn = lblkno(fs, uio->uio_offset);
 +		nextlbn = lbn + 1;
 +		size = blksize(fs, ip, lbn);
 +		blkoffset = blkoff(fs, uio->uio_offset);
 +
 +		xfersize = fs->e2fs_fsize - blkoffset;
 +		if (uio->uio_resid < xfersize)
 +			xfersize = uio->uio_resid;
 +		if (bytesinfile < xfersize)
 +			xfersize = bytesinfile;
 +
 +		if (lblktosize(fs, nextlbn) >= ip->i_size)
 +			error = bread(vp, lbn, size, NOCRED, &bp);
 +		else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0)
 +			error = cluster_read(vp, ip->i_size, lbn, size,
 +			    NOCRED, blkoffset + uio->uio_resid, seqcount, &bp);
 +		else if (seqcount > 1) {
 +			int nextsize = blksize(fs, ip, nextlbn);
 +			error = breadn(vp, lbn,
 +			    size, &nextlbn, &nextsize, 1, NOCRED, &bp);
 +		} else
 +			error = bread(vp, lbn, size, NOCRED, &bp);
 +		if (error) {
 +			brelse(bp);
 +			bp = NULL;
 +			break;
 +		}
 +
 +		/*
 +		 * If IO_DIRECT then set B_DIRECT for the buffer.  This
 +		 * will cause us to attempt to release the buffer later on
 +		 * and will cause the buffer cache to attempt to free the
 +		 * underlying pages.
 +		 */
 +		if (ioflag & IO_DIRECT)
 +			bp->b_flags |= B_DIRECT;
 +
 +		/*
 +		 * We should only get non-zero b_resid when an I/O error
 +		 * has occurred, which should cause us to break above.
 +		 * However, if the short read did not cause an error,
 +		 * then we want to ensure that we do not uiomove bad
 +		 * or uninitialized data.
 +		 */
 +		size -= bp->b_resid;
 +		if (size < xfersize) {
 +			if (size == 0)
 +				break;
 +			xfersize = size;
 +		}
 +		error = uiomove((char *)bp->b_data + blkoffset,
 +  			(int)xfersize, uio);
 +		if (error)
 +			break;
 +
 +		if (ioflag & (IO_VMIO|IO_DIRECT)) {
 +			/*
 +			 * If it's VMIO or direct I/O, then we don't
 +			 * need the buf, mark it available for
 +			 * freeing. If it's non-direct VMIO, the VM has
 +			 * the data.
 +			 */
 +			bp->b_flags |= B_RELBUF;
 +			brelse(bp);
 +		} else {
 +			/*
 +			 * Otherwise let whoever
 +			 * made the request take care of
 +			 * freeing it. We just queue
 +			 * it onto another list.
 +			 */
 +			bqrelse(bp);
 +		}
 +	}
 +
 +	/* 
 +	 * This can only happen in the case of an error
 +	 * because the loop above resets bp to NULL on each iteration
 +	 * and on normal completion has not set a new value into it.
 +	 * so it must have come from a 'break' statement
 +	 */
 +	if (bp != NULL) {
 +		if (ioflag & (IO_VMIO|IO_DIRECT)) {
 +			bp->b_flags |= B_RELBUF;
 +			brelse(bp);
 +		} else {
 +			bqrelse(bp);
 +		}
 +	}
 +
 +	if ((error == 0 || uio->uio_resid != orig_resid) &&
 +	    (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
 +		ip->i_flag |= IN_ACCESS;
 +	return (error);
 +}
 +
 +/*
 + * Vnode op for writing.
 + */
 +static int
 +ext2_write(ap)
 +	struct vop_write_args /* {
 +		struct vnode *a_vp;
 +		struct uio *a_uio;
 +		int a_ioflag;
 +		struct ucred *a_cred;
 +	} */ *ap;
 +{
 +	struct vnode *vp;
 +	struct uio *uio;
 +	struct inode *ip;
 +	struct m_ext2fs *fs;
 +	struct buf *bp;
 +	daddr_t lbn;
 +	off_t osize;
 +	int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
 +
 +	ioflag = ap->a_ioflag;
 +	uio = ap->a_uio;
 +	vp = ap->a_vp;
 +
 +	seqcount = ioflag >> IO_SEQSHIFT;
 +	ip = VTOI(vp);
 +
 +#ifdef INVARIANTS
 +	if (uio->uio_rw != UIO_WRITE)
 +		panic("%s: mode", "ext2_write");
 +#endif
 +
 +	switch (vp->v_type) {
 +	case VREG:
 +		if (ioflag & IO_APPEND)
 +			uio->uio_offset = ip->i_size;
 +		if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
 +			return (EPERM);
 +		/* FALLTHROUGH */
 +	case VLNK:
 +		break;
 +	case VDIR:
 +		/* XXX differs from ffs -- this is called from ext2_mkdir(). */
 +		if ((ioflag & IO_SYNC) == 0)
 +		panic("ext2_write: nonsync dir write");
 +		break;
 +	default:
 +		panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
 +		    vp->v_type, (intmax_t)uio->uio_offset,
 +		    (intmax_t)uio->uio_resid);
 +	}
 +
 +	KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
 +	KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
 +	fs = ip->i_e2fs;
 +	if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
 +		return (EFBIG);
 +	/*
 +	 * Maybe this should be above the vnode op call, but so long as
 +	 * file servers have no limits, I don't think it matters.
 +	 */
 +	if (vn_rlimit_fsize(vp, uio, uio->uio_td))
 +		return (EFBIG);
 +
 +	resid = uio->uio_resid;
 +	osize = ip->i_size;
 +	if (seqcount > BA_SEQMAX)
 +		flags = BA_SEQMAX << BA_SEQSHIFT;
 +	else
 +		flags = seqcount << BA_SEQSHIFT;
 +	if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
 +		flags |= IO_SYNC;
 +
 +	for (error = 0; uio->uio_resid > 0;) {
 +		lbn = lblkno(fs, uio->uio_offset);
 +		blkoffset = blkoff(fs, uio->uio_offset);
 +		xfersize = fs->e2fs_fsize - blkoffset;
 +		if (uio->uio_resid < xfersize)
 +			xfersize = uio->uio_resid;
 +		if (uio->uio_offset + xfersize > ip->i_size)
 +			vnode_pager_setsize(vp, uio->uio_offset + xfersize);
 +
 +                /*
 +		 * We must perform a read-before-write if the transfer size
 +		 * does not cover the entire buffer.
 +                 */
 +		if (fs->e2fs_bsize > xfersize)
 +			flags |= BA_CLRBUF;
 +		else
 +			flags &= ~BA_CLRBUF;
 +		error = ext2_balloc(ip, lbn, blkoffset + xfersize,
 +		    ap->a_cred, &bp, flags);
 +		if (error != 0)
 +			break;
 +
 +		/*
 +		 * If the buffer is not valid and we did not clear garbage
 +		 * out above, we have to do so here even though the write
 +		 * covers the entire buffer in order to avoid a mmap()/write
 +		 * race where another process may see the garbage prior to
 +		 * the uiomove() for a write replacing it.
 +		 */
 +		if ((bp->b_flags & B_CACHE) == 0 && fs->e2fs_bsize <= xfersize)
 +			vfs_bio_clrbuf(bp);
 +		if ((ioflag & (IO_SYNC|IO_INVAL)) == (IO_SYNC|IO_INVAL))
 +			bp->b_flags |= B_NOCACHE;
 +		if (uio->uio_offset + xfersize > ip->i_size)
 +			ip->i_size = uio->uio_offset + xfersize;
 +		size = blksize(fs, ip, lbn) - bp->b_resid;
 +		if (size < xfersize)
 +			xfersize = size;
 +
 +		error =
 +		    uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
 +		if (ioflag & (IO_VMIO|IO_DIRECT)) {
 +			bp->b_flags |= B_RELBUF;
 +		}
 +
 +		/*
 +		 * If IO_SYNC each buffer is written synchronously.  Otherwise
 +		 * if we have a severe page deficiency write the buffer
 +		 * asynchronously.  Otherwise try to cluster, and if that
 +		 * doesn't do it then either do an async write (if O_DIRECT),
 +		 * or a delayed write (if not).
 +		 */
 +		if (ioflag & IO_SYNC) {
 +			(void)bwrite(bp);
 +		} else if (vm_page_count_severe() ||
 +		    buf_dirty_count_severe() ||
 +		    (ioflag & IO_ASYNC)) {
 +			bp->b_flags |= B_CLUSTEROK;
 +			bawrite(bp);
 +		} else if (xfersize + blkoffset == fs->e2fs_fsize) {
 +			if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
 +				bp->b_flags |= B_CLUSTEROK;
 +				cluster_write(vp, bp, ip->i_size, seqcount);
 +			} else {
 +				bawrite(bp);
 +			}
 +		} else if (ioflag & IO_DIRECT) {
 +			bp->b_flags |= B_CLUSTEROK;
 +			bawrite(bp);
 +		} else {
 +			bp->b_flags |= B_CLUSTEROK;
 +			bdwrite(bp);
 +		}
 +		if (error || xfersize == 0)
 +			break;
 +	}
 +	/*
 +	 * If we successfully wrote any data, and we are not the superuser
 +	 * we clear the setuid and setgid bits as a precaution against
 +	 * tampering.
 +	 */
 +	if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid &&
 +	    ap->a_cred) {
 +		if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
 +			ip->i_mode &= ~(ISUID | ISGID);
 +	}
 +	if (error) {
 +		if (ioflag & IO_UNIT) {
 +			(void)ext2_truncate(vp, osize,
 +			    ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
 +			uio->uio_offset -= resid - uio->uio_resid;
 +			uio->uio_resid = resid;
 +		}
 +	}
 +	if (uio->uio_resid != resid) {
 +               ip->i_flag |= IN_CHANGE | IN_UPDATE;
 +               if (ioflag & IO_SYNC)
 +                       error = ext2_update(vp, 1);
 +       }
 +	return (error);
 +}
 _______________________________________________
 svn-src-all@freebsd.org mailing list
 http://lists.freebsd.org/mailman/listinfo/svn-src-all
 To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
 
State-Changed-From-To: open->patched 
State-Changed-By: pfg 
State-Changed-When: Thu Dec 15 03:44:25 UTC 2011 
State-Changed-Why:  

Patched in head. 


Responsible-Changed-From-To: freebsd-fs->pfg 
Responsible-Changed-By: pfg 
Responsible-Changed-When: Thu Dec 15 03:44:25 UTC 2011 
Responsible-Changed-Why:  
Assign to myself. 

http://www.freebsd.org/cgi/query-pr.cgi?pr=159232 

From: dfilter@FreeBSD.ORG (dfilter service)
To: bug-followup@FreeBSD.org
Cc:  
Subject: Re: kern/159232: commit references a PR
Date: Thu,  5 Jan 2012 01:35:23 +0000 (UTC)

 Author: pfg
 Date: Thu Jan  5 01:35:01 2012
 New Revision: 229549
 URL: http://svn.freebsd.org/changeset/base/229549
 
 Log:
   MFC:	r228507, r228539, r228583
   
   Merge ext2_readwrite.c into ext2_vnops.c as done in UFS.
   Bring in reallocblk to ext2fs: new feature implemented by Zheng Liu as
   GSoC 2010.
   Many style fixes by jh@.
   
   PR:		159232, 159233 and 162564
   Approved by:	jhb (mentor)
 
 Deleted:
   stable/9/sys/fs/ext2fs/ext2_readwrite.c
 Modified:
   stable/9/sys/fs/ext2fs/ext2_alloc.c
   stable/9/sys/fs/ext2fs/ext2_balloc.c
   stable/9/sys/fs/ext2fs/ext2_bmap.c
   stable/9/sys/fs/ext2fs/ext2_extern.h
   stable/9/sys/fs/ext2fs/ext2_inode.c
   stable/9/sys/fs/ext2fs/ext2_lookup.c
   stable/9/sys/fs/ext2fs/ext2_subr.c
   stable/9/sys/fs/ext2fs/ext2_vfsops.c
   stable/9/sys/fs/ext2fs/ext2_vnops.c
   stable/9/sys/fs/ext2fs/ext2fs.h
   stable/9/sys/fs/ext2fs/fs.h
 Directory Properties:
   stable/9/sys/   (props changed)
 
 Modified: stable/9/sys/fs/ext2fs/ext2_alloc.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_alloc.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_alloc.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -42,6 +42,7 @@
  #include <sys/vnode.h>
  #include <sys/stat.h>
  #include <sys/mount.h>
 +#include <sys/sysctl.h>
  #include <sys/syslog.h>
  #include <sys/buf.h>
  
 @@ -52,6 +53,7 @@
  #include <fs/ext2fs/ext2_extern.h>
  
  static daddr_t	ext2_alloccg(struct inode *, int, daddr_t, int);
 +static daddr_t	ext2_clusteralloc(struct inode *, int, daddr_t, int);
  static u_long	ext2_dirpref(struct inode *);
  static void	ext2_fserr(struct m_ext2fs *, uid_t, char *);
  static u_long	ext2_hashalloc(struct inode *, int, long, int,
 @@ -59,9 +61,6 @@ static u_long	ext2_hashalloc(struct inod
  						int));
  static daddr_t	ext2_nodealloccg(struct inode *, int, daddr_t, int);
  static daddr_t  ext2_mapsearch(struct m_ext2fs *, char *, daddr_t);
 -#ifdef FANCY_REALLOC
 -static int	ext2_reallocblks(struct vop_reallocblks_args *);
 -#endif
  
  /*
   * Allocate a block in the file system.
 @@ -113,20 +112,20 @@ ext2_alloc(ip, lbn, bpref, size, cred, b
  	if (bpref >= fs->e2fs->e2fs_bcount)
  		bpref = 0;
  	if (bpref == 0)
 -                cg = ino_to_cg(fs, ip->i_number);
 -        else
 -                cg = dtog(fs, bpref);
 -        bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
 -                                                 ext2_alloccg);
 -        if (bno > 0) {
 +		cg = ino_to_cg(fs, ip->i_number);
 +	else
 +		cg = dtog(fs, bpref);
 +	bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
 +				      ext2_alloccg);
 +	if (bno > 0) {
  		/* set next_alloc fields as done in block_getblk */
  		ip->i_next_alloc_block = lbn;
  		ip->i_next_alloc_goal = bno;
  
 -                ip->i_blocks += btodb(fs->e2fs_bsize);
 -                ip->i_flag |= IN_CHANGE | IN_UPDATE;
 -                *bnp = bno;
 -                return (0);
 +		ip->i_blocks += btodb(fs->e2fs_bsize);
 +		ip->i_flag |= IN_CHANGE | IN_UPDATE;
 +		*bnp = bno;
 +		return (0);
          }
  nospace:
  	EXT2_UNLOCK(ump);
 @@ -150,7 +149,6 @@ nospace:
   * the previous block allocation will be used.
   */
  
 -#ifdef FANCY_REALLOC
  SYSCTL_NODE(_vfs, OID_AUTO, ext2fs, CTLFLAG_RW, 0, "EXT2FS filesystem");
  
  static int doasyncfree = 1;
 @@ -159,7 +157,6 @@ SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doasyn
  
  static int doreallocblks = 1;
  SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "");
 -#endif
  
  int
  ext2_reallocblks(ap)
 @@ -168,11 +165,6 @@ ext2_reallocblks(ap)
  		struct cluster_save *a_buflist;
  	} */ *ap;
  {
 -#ifndef FANCY_REALLOC
 -/* printf("ext2_reallocblks not implemented\n"); */
 -return ENOSPC;
 -#else
 -
  	struct m_ext2fs *fs;
  	struct inode *ip;
  	struct vnode *vp;
 @@ -184,14 +176,17 @@ return ENOSPC;
  	int32_t start_lbn, end_lbn, soff, newblk, blkno;
  	int i, len, start_lvl, end_lvl, pref, ssize;
  
 +	if (doreallocblks == 0)
 +		  return (ENOSPC);
 +
  	vp = ap->a_vp;
  	ip = VTOI(vp);
  	fs = ip->i_e2fs;
  	ump = ip->i_ump;
 -#ifdef UNKLAR
 -	if (fs->fs_contigsumsize <= 0)
 +
 +	if (fs->e2fs_contigsumsize <= 0)
  		return (ENOSPC);
 -#endif
 +
  	buflist = ap->a_buflist;
  	len = buflist->bs_nchildren;
  	start_lbn = buflist->bs_children[0]->b_lblkno;
 @@ -228,11 +223,6 @@ return ENOSPC;
  		soff = idp->in_off;
  	}
  	/*
 -	 * Find the preferred location for the cluster.
 -	 */
 -	EXT2_LOCK(ump);
 -	pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
 -	/*
  	 * If the block range spans two block maps, get the second map.
  	 */
  	if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
 @@ -243,13 +233,16 @@ return ENOSPC;
  			panic("ext2_reallocblk: start == end");
  #endif
  		ssize = len - (idp->in_off + 1);
 -		if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp)){
 -			EXT2_UNLOCK(ump);	
 +		if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp))
  			goto fail;
 -		}
  		ebap = (int32_t *)ebp->b_data;
  	}
  	/*
 +	 * Find the preferred location for the cluster.
 +	 */
 +	EXT2_LOCK(ump);
 +	pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
 +	/*
  	 * Search the block map looking for an allocation of the desired size.
  	 */
  	if ((newblk = (int32_t)ext2_hashalloc(ip, dtog(fs, pref), pref,
 @@ -264,15 +257,23 @@ return ENOSPC;
  	 * block pointers in the inode and indirect blocks associated
  	 * with the file.
  	 */
 +#ifdef DEBUG
 +	printf("realloc: ino %d, lbns %jd-%jd\n\told:", ip->i_number,
 +	    (intmax_t)start_lbn, (intmax_t)end_lbn);
 +#endif /* DEBUG */
  	blkno = newblk;
  	for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
 -		if (i == ssize)
 +		if (i == ssize) {
  			bap = ebap;
  			soff = -i;
 +		}
  #ifdef DIAGNOSTIC
  		if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap))
  			panic("ext2_reallocblks: alloc mismatch");
  #endif
 +#ifdef DEBUG
 +	printf(" %d,", *bap);
 +#endif /* DEBUG */
  		*bap++ = blkno;
  	}
  	/*
 @@ -308,11 +309,20 @@ return ENOSPC;
  	/*
  	 * Last, free the old blocks and assign the new blocks to the buffers.
  	 */
 +#ifdef DEBUG
 +	printf("\n\tnew:");
 +#endif /* DEBUG */
  	for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
  		ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
  		    fs->e2fs_bsize);
  		buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
 -	}
 +#ifdef DEBUG
 +		printf(" %d,", blkno);
 +#endif /* DEBUG */
 +	}
 +#ifdef DEBUG
 +	printf("\n");
 +#endif /* DEBUG */
  	return (0);
  
  fail:
 @@ -321,8 +331,6 @@ fail:
  	if (sbap != &ip->i_db[0])
  		brelse(sbp);
  	return (ENOSPC);
 -
 -#endif /* FANCY_REALLOC */
  }
  
  /*
 @@ -356,7 +364,7 @@ ext2_valloc(pvp, mode, cred, vpp)
  	 * ext2_dirpref else obtain it using ino_to_cg. The preferred inode is
  	 * always the next inode.
  	 */
 -	if((mode & IFMT) == IFDIR) {
 +	if ((mode & IFMT) == IFDIR) {
  		cg = ext2_dirpref(pip);
  		if (fs->e2fs_contigdirs[cg] < 255)
  			fs->e2fs_contigdirs[cg]++;
 @@ -549,13 +557,13 @@ ext2_blkpref(ip, lbn, indx, bap, blocknr
  	/* if the next block is actually what we thought it is,
  	   then set the goal to what we thought it should be
  	*/
 -	if(ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
 +	if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
  		return ip->i_next_alloc_goal;
  
  	/* now check whether we were provided with an array that basically
  	   tells us previous blocks to which we want to stay closeby
  	*/
 -	if(bap) 
 +	if (bap) 
                  for (tmp = indx - 1; tmp >= 0; tmp--) 
  			if (bap[tmp]) 
  				return bap[tmp];
 @@ -747,6 +755,7 @@ gotit:
  #endif
  	setbit(bbp, bno);
  	EXT2_LOCK(ump);
 +	ext2_clusteracct(fs, bbp, cg, bno, -1);
  	fs->e2fs->e2fs_fbcount--;
  	fs->e2fs_gd[cg].ext2bgd_nbfree--;
  	fs->e2fs_fmod = 1;
 @@ -756,6 +765,116 @@ gotit:
  }
  
  /*
 + * Determine whether a cluster can be allocated.
 + */
 +static daddr_t
 +ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
 +{
 +	struct m_ext2fs *fs;
 +	struct ext2mount *ump;
 +	struct buf *bp;
 +	char *bbp;
 +	int bit, error, got, i, loc, run;
 +	int32_t *lp;
 +	daddr_t bno;
 +
 +	fs = ip->i_e2fs;
 +	ump = ip->i_ump;
 +
 +	if (fs->e2fs_maxcluster[cg] < len)
 +		return (0);
 +
 +	EXT2_UNLOCK(ump);
 +	error = bread(ip->i_devvp,
 +	    fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
 +	    (int)fs->e2fs_bsize, NOCRED, &bp);
 +	if (error)
 +		goto fail_lock;
 +
 +	bbp = (char *)bp->b_data;
 +	bp->b_xflags |= BX_BKGRDWRITE;
 +
 +	EXT2_LOCK(ump);
 +	/*
 +	 * Check to see if a cluster of the needed size (or bigger) is
 +	 * available in this cylinder group.
 +	 */
 +	lp = &fs->e2fs_clustersum[cg].cs_sum[len];
 +	for (i = len; i <= fs->e2fs_contigsumsize; i++)
 +		if (*lp++ > 0)
 +			break;
 +	if (i > fs->e2fs_contigsumsize) {
 +		/*
 +		 * Update the cluster summary information to reflect
 +		 * the true maximum-sized cluster so that future cluster
 +		 * allocation requests can avoid reading the bitmap only
 +		 * to find no cluster.
 +		 */
 +		lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
 +			for (i = len - 1; i > 0; i--)
 +				if (*lp-- > 0)
 +					break;
 +		fs->e2fs_maxcluster[cg] = i;
 +		goto fail;
 +	}
 +	EXT2_UNLOCK(ump);
 +
 +	/* Search the bitmap to find a big enough cluster like in FFS. */
 +	if (dtog(fs, bpref) != cg)
 +		bpref = 0;
 +	if (bpref != 0)
 +		bpref = dtogd(fs, bpref);
 +	loc = bpref / NBBY;
 +	bit = 1 << (bpref % NBBY);
 +	for (run = 0, got = bpref; got < fs->e2fs->e2fs_fpg; got++) {
 +		if ((bbp[loc] & bit) != 0)
 +			run = 0;
 +		else {
 +			run++;
 +			if (run == len)
 +				break;
 +		}
 +		if ((got & (NBBY - 1)) != (NBBY - 1))
 +			bit <<= 1;
 +		else {
 +			loc++;
 +			bit = 1;
 +		}
 +	}
 +
 +	if (got >= fs->e2fs->e2fs_fpg)
 +		goto fail_lock;
 +
 +	/* Allocate the cluster that we found. */
 +	for (i = 1; i < len; i++)
 +		if (!isclr(bbp, got - run + i))
 +			panic("ext2_clusteralloc: map mismatch");
 +
 +	bno = got - run + 1;
 +	if (bno >= fs->e2fs->e2fs_fpg)
 +		panic("ext2_clusteralloc: allocated out of group");
 +
 +	EXT2_LOCK(ump);
 +	for (i = 0; i < len; i += fs->e2fs_fpb) {
 +		setbit(bbp, bno + i);
 +		ext2_clusteracct(fs, bbp, cg, bno + i, -1);
 +		fs->e2fs->e2fs_fbcount--;
 +		fs->e2fs_gd[cg].ext2bgd_nbfree--;
 +	}
 +	fs->e2fs_fmod = 1;
 +	EXT2_UNLOCK(ump);
 +
 +	bdwrite(bp);
 +	return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
 +
 +fail_lock:
 +	EXT2_LOCK(ump);
 +fail:
 +	brelse(bp);
 +	return (0);
 +}
 +
 +/*
   * Determine whether an inode can be allocated.
   *
   * Check to see if an inode is available, and if it is,
 @@ -877,6 +996,7 @@ ext2_blkfree(ip, bno, size)
          }
          clrbit(bbp, bno);
  	EXT2_LOCK(ump);
 +	ext2_clusteracct(fs, bbp, cg, bno, 1);
          fs->e2fs->e2fs_fbcount++;
          fs->e2fs_gd[cg].ext2bgd_nbfree++;
          fs->e2fs_fmod = 1;
 
 Modified: stable/9/sys/fs/ext2fs/ext2_balloc.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_balloc.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_balloc.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -156,7 +156,7 @@ ext2_balloc(ip, lbn, size, cred, bpp, fl
  	 */
  	pref = 0;
  	if ((error = ext2_getlbns(vp, lbn, indirs, &num)) != 0)
 -		return(error);
 +		return (error);
  #ifdef DIAGNOSTIC
  	if (num < 1)
  		panic ("ext2_balloc: ext2_getlbns returned indirect block");
 
 Modified: stable/9/sys/fs/ext2fs/ext2_bmap.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_bmap.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_bmap.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -215,7 +215,7 @@ ext2_bmaparray(vp, bn, bnp, runp, runb)
  			    ++bn, ++*runp);
  			bn = ap->in_off;
  			if (runb && bn) {
 -				for(--bn; bn >= 0 && *runb < maxrun &&
 +				for (--bn; bn >= 0 && *runb < maxrun &&
  			    		is_sequential(ump, ((int32_t *)bp->b_data)[bn],
  					    ((int32_t *)bp->b_data)[bn+1]);
  			    		--bn, ++*runb);
 
 Modified: stable/9/sys/fs/ext2fs/ext2_extern.h
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_extern.h	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_extern.h	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -55,12 +55,13 @@ void	ext2_blkfree(struct inode *, int32_
  int32_t	ext2_blkpref(struct inode *, int32_t, int, int32_t *, int32_t);
  int	ext2_bmap(struct vop_bmap_args *);
  int	ext2_bmaparray(struct vnode *, int32_t, int32_t *, int *, int *);
 +void	ext2_clusteracct(struct m_ext2fs *, char *, int, daddr_t, int);
  void	ext2_dirbad(struct inode *ip, doff_t offset, char *how);
  void	ext2_ei2i(struct ext2fs_dinode *, struct inode *);
  int	ext2_getlbns(struct vnode *, int32_t, struct indir *, int *);
  void	ext2_i2ei(struct inode *, struct ext2fs_dinode *);
 +int     ext2_reallocblks(struct vop_reallocblks_args *);
  void	ext2_itimes(struct vnode *vp);
 -int	ext2_reallocblks(struct vop_reallocblks_args *);
  int	ext2_reclaim(struct vop_reclaim_args *);
  void	ext2_setblock(struct m_ext2fs *, u_char *, int32_t);
  int	ext2_truncate(struct vnode *, off_t, int, struct ucred *, struct thread *);
 
 Modified: stable/9/sys/fs/ext2fs/ext2_inode.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_inode.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_inode.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -229,7 +229,7 @@ ext2_truncate(vp, length, flags, cred, t
  	 * will be returned to the free list.  lastiblock values are also
  	 * normalized to -1 for calls to ext2_indirtrunc below.
  	 */
 -	bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
 +	bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof(oldblks));
  	for (level = TRIPLE; level >= SINGLE; level--)
  		if (lastiblock[level] < 0) {
  			oip->i_ib[level] = 0;
 @@ -246,8 +246,8 @@ ext2_truncate(vp, length, flags, cred, t
  	 * Note that we save the new block configuration so we can check it
  	 * when we are done.
  	 */
 -	bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
 -	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
 +	bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof(newblks));
 +	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof(oldblks));
  	oip->i_size = osize;
  	error = vtruncbuf(ovp, cred, td, length, (int)fs->e2fs_bsize);
  	if (error && (allerror == 0))
 @@ -418,7 +418,7 @@ ext2_indirtrunc(ip, lbn, dbn, lastbn, le
  	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
  	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
  	bzero((caddr_t)&bap[last + 1],
 -	  (u_int)(NINDIR(fs) - (last + 1)) * sizeof (int32_t));
 +	  (u_int)(NINDIR(fs) - (last + 1)) * sizeof(int32_t));
  	if (last == -1)
  		bp->b_flags |= B_INVAL;
  	error = bwrite(bp);
 
 Modified: stable/9/sys/fs/ext2fs/ext2_lookup.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_lookup.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_lookup.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -1002,7 +1002,7 @@ ext2_dirempty(ip, parentino, cred)
  	struct dirtemplate dbuf;
  	struct ext2fs_direct_2 *dp = (struct ext2fs_direct_2 *)&dbuf;
  	int error, count, namlen;
 -#define	MINDIRSIZ (sizeof (struct dirtemplate) / 2)
 +#define	MINDIRSIZ (sizeof(struct dirtemplate) / 2)
  
  	for (off = 0; off < ip->i_size; off += dp->e2d_reclen) {
  		error = vn_rdwr(UIO_READ, ITOV(ip), (caddr_t)dp, MINDIRSIZ,
 @@ -1070,7 +1070,7 @@ ext2_checkpath(source, target, cred)
  			break;
  		}
  		error = vn_rdwr(UIO_READ, vp, (caddr_t)&dirbuf,
 -			sizeof (struct dirtemplate), (off_t)0, UIO_SYSSPACE,
 +			sizeof(struct dirtemplate), (off_t)0, UIO_SYSSPACE,
  			IO_NODELOCKED | IO_NOMACCHECK, cred, NOCRED, NULL,
  			NULL);
  		if (error != 0)
 
 Modified: stable/9/sys/fs/ext2fs/ext2_subr.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_subr.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_subr.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -120,3 +120,107 @@ ext2_checkoverlap(bp, ip)
  	}
  }
  #endif /* KDB */
 +
 +/*
 + * Update the cluster map because of an allocation of free like ffs.
 + *
 + * Cnt == 1 means free; cnt == -1 means allocating.
 + */
 +void
 +ext2_clusteracct(struct m_ext2fs *fs, char *bbp, int cg, daddr_t bno, int cnt)
 +{
 +	int32_t *sump = fs->e2fs_clustersum[cg].cs_sum;
 +	int32_t *lp;
 +	int back, bit, end, forw, i, loc, start;
 +
 +	/* Initialize the cluster summary array. */
 +	if (fs->e2fs_clustersum[cg].cs_init == 0) {
 +		int run = 0;
 +		bit = 1;
 +		loc = 0;
 +
 +		for (i = 0; i < fs->e2fs->e2fs_fpg; i++) {
 +			if ((bbp[loc] & bit) == 0)
 +				run++;
 +			else if (run != 0) {
 +				if (run > fs->e2fs_contigsumsize)
 +					run = fs->e2fs_contigsumsize;
 +				sump[run]++;
 +				run = 0;
 +			}
 +			if ((i & (NBBY - 1)) != (NBBY - 1))
 +				bit <<= 1;
 +			else {
 +				loc++;
 +				bit = 1;
 +			}
 +		}
 +		if (run != 0) {
 +			if (run > fs->e2fs_contigsumsize)
 +				run = fs->e2fs_contigsumsize;
 +			sump[run]++;
 +		}
 +		fs->e2fs_clustersum[cg].cs_init = 1;
 +	}
 +
 +	if (fs->e2fs_contigsumsize <= 0)
 +		return;
 +
 +	/* Find the size of the cluster going forward. */
 +	start = bno + 1;
 +	end = start + fs->e2fs_contigsumsize;
 +	if (end > fs->e2fs->e2fs_fpg)
 +		end = fs->e2fs->e2fs_fpg;
 +	loc = start / NBBY;
 +	bit = 1 << (start % NBBY);
 +	for (i = start; i < end; i++) {
 +		if ((bbp[loc] & bit) != 0)
 +			break;
 +		if ((i & (NBBY - 1)) != (NBBY - 1))
 +			bit <<= 1;
 +		else {
 +			loc++;
 +			bit = 1;
 +		}
 +	}
 +	forw = i - start;
 +
 +	/* Find the size of the cluster going backward. */
 +	start = bno - 1;
 +	end = start - fs->e2fs_contigsumsize;
 +	if (end < 0)
 +		end = -1;
 +	loc = start / NBBY;
 +	bit = 1 << (start % NBBY);
 +	for (i = start; i > end; i--) {
 +		if ((bbp[loc] & bit) != 0)
 +			break;
 +		if ((i & (NBBY - 1)) != 0)
 +			bit >>= 1;
 +		else {
 +			loc--;
 +			bit = 1 << (NBBY - 1);
 +		}
 +	}
 +	back = start - i;
 +
 +	/*
 +	 * Account for old cluster and the possibly new forward and
 +	 * back clusters.
 +	 */
 +	i = back + forw + 1;
 +	if (i > fs->e2fs_contigsumsize)
 +		i = fs->e2fs_contigsumsize;
 +	sump[i] += cnt;
 +	if (back > 0)
 +		sump[back] -= cnt;
 +	if (forw > 0)
 +		sump[forw] -= cnt;
 +
 +	/* Update cluster summary information. */
 +	lp = &sump[fs->e2fs_contigsumsize];
 +	for (i = fs->e2fs_contigsumsize; i > 0; i--)
 +		if (*lp-- > 0)
 +			break;
 +	fs->e2fs_maxcluster[cg] = i;
 +}
 
 Modified: stable/9/sys/fs/ext2fs/ext2_vfsops.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_vfsops.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_vfsops.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -349,7 +349,7 @@ compute_sb_data(struct vnode *devvp, str
  	}
  	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
  	fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb;
 -	fs->e2fs_descpb = fs->e2fs_bsize / sizeof (struct ext2_gd);
 +	fs->e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
  	/* s_resuid / s_resgid ? */
  	fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock +
  	    EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs);
 @@ -405,7 +405,7 @@ compute_sb_data(struct vnode *devvp, str
   * Things to do to update the mount:
   *	1) invalidate all cached meta-data.
   *	2) re-read superblock from disk.
 - *	3) re-read summary information from disk.
 + *	3) invalidate all cluster summary information.
   *	4) invalidate all inactive vnodes.
   *	5) invalidate all cached file data.
   *	6) re-read inode data for all active vnodes.
 @@ -419,7 +419,9 @@ ext2_reload(struct mount *mp, struct thr
  	struct buf *bp;
  	struct ext2fs *es;
  	struct m_ext2fs *fs;
 -	int error;
 +	struct csum *sump;
 +	int error, i;
 +	int32_t *lp;
  
  	if ((mp->mnt_flag & MNT_RDONLY) == 0)
  		return (EINVAL);
 @@ -456,6 +458,19 @@ ext2_reload(struct mount *mp, struct thr
  #endif
  	brelse(bp);
  
 +	/*
 +	 * Step 3: invalidate all cluster summary information.
 +	 */
 +	if (fs->e2fs_contigsumsize > 0) {
 +		lp = fs->e2fs_maxcluster;
 +		sump = fs->e2fs_clustersum;
 +		for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
 +			*lp++ = fs->e2fs_contigsumsize;
 +			sump->cs_init = 0;
 +			bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
 +		}
 +	}
 +
  loop:
  	MNT_ILOCK(mp);
  	MNT_VNODE_FOREACH(vp, mp, mvp) {
 @@ -511,8 +526,11 @@ ext2_mountfs(struct vnode *devvp, struct
  	struct cdev *dev = devvp->v_rdev;
  	struct g_consumer *cp;
  	struct bufobj *bo;
 +	struct csum *sump;
  	int error;
  	int ronly;
 +	int i, size;
 +	int32_t *lp;
  
  	ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
  	/* XXX: use VOP_ACESS to check FS perms */
 @@ -565,8 +583,7 @@ ext2_mountfs(struct vnode *devvp, struct
  			goto out;
  		}
  	}
 -	ump = malloc(sizeof *ump, M_EXT2MNT, M_WAITOK);
 -	bzero((caddr_t)ump, sizeof *ump);
 +	ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
  
  	/*
  	 * I don't know whether this is the right strategy. Note that
 @@ -582,6 +599,33 @@ ext2_mountfs(struct vnode *devvp, struct
  	if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
  		goto out;
  
 +	/*
 +	 * Calculate the maximum contiguous blocks and size of cluster summary
 +	 * array.  In FFS this is done by newfs; however, the superblock 
 +	 * in ext2fs doesn't have these variables, so we can calculate 
 +	 * them here.
 +	 */
 +	ump->um_e2fs->e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
 +	if (ump->um_e2fs->e2fs_maxcontig > 0)
 +		ump->um_e2fs->e2fs_contigsumsize =
 +		    MIN(ump->um_e2fs->e2fs_maxcontig, EXT2_MAXCONTIG);
 +	else
 +		ump->um_e2fs->e2fs_contigsumsize = 0;
 +	if (ump->um_e2fs->e2fs_contigsumsize > 0) {
 +		size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
 +		ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
 +		size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
 +		ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
 +		lp = ump->um_e2fs->e2fs_maxcluster;
 +		sump = ump->um_e2fs->e2fs_clustersum;
 +		for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
 +			*lp++ = ump->um_e2fs->e2fs_contigsumsize;
 +			sump->cs_init = 0;
 +			sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
 +			    sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
 +		}
 +	}
 +
  	brelse(bp);
  	bp = NULL;
  	fs = ump->um_e2fs;
 @@ -656,7 +700,8 @@ ext2_unmount(struct mount *mp, int mntfl
  {
  	struct ext2mount *ump;
  	struct m_ext2fs *fs;
 -	int error, flags, ronly;
 +	struct csum *sump;
 +	int error, flags, i, ronly;
  
  	flags = 0;
  	if (mntflags & MNT_FORCE) {
 @@ -681,6 +726,11 @@ ext2_unmount(struct mount *mp, int mntfl
  	g_topology_unlock();
  	PICKUP_GIANT();
  	vrele(ump->um_devvp);
 +	sump = fs->e2fs_clustersum;
 +	for (i = 0; i < fs->e2fs_gcount; i++, sump++)
 +		free(sump->cs_sum, M_EXT2MNT);
 +	free(fs->e2fs_clustersum, M_EXT2MNT);
 +	free(fs->e2fs_maxcluster, M_EXT2MNT);
  	free(fs->e2fs_gd, M_EXT2MNT);
  	free(fs->e2fs_contigdirs, M_EXT2MNT);
  	free(fs->e2fs, M_EXT2MNT);
 @@ -927,7 +977,7 @@ ext2_vget(struct mount *mp, ino_t ino, i
  	 */
  	if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
  		used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
 -		for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
 +		for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
  			ip->i_db[i] = 0;
  	}
  /*
 
 Modified: stable/9/sys/fs/ext2fs/ext2_vnops.c
 ==============================================================================
 --- stable/9/sys/fs/ext2fs/ext2_vnops.c	Thu Jan  5 01:25:47 2012	(r229548)
 +++ stable/9/sys/fs/ext2fs/ext2_vnops.c	Thu Jan  5 01:35:01 2012	(r229549)
 @@ -64,9 +64,13 @@
  #include <sys/file.h>
  
  #include <vm/vm.h>
 +#include <vm/vm_page.h>
 +#include <vm/vm_object.h>
  #include <vm/vm_extern.h>
  #include <vm/vnode_pager.h>
  
 +#include "opt_directio.h"
 +
  #include <fs/fifofs/fifo.h>
  
  #include <ufs/ufs/dir.h>
 @@ -159,8 +163,6 @@ struct vop_vector ext2_fifoops = {
  	.vop_vptofh =		ext2_vptofh,
  };
  
 -#include <fs/ext2fs/ext2_readwrite.c>
 -
  /*
   * A virgin directory (no blushing please).
   * Note that the type and namlen fields are reversed relative to ext2.
 @@ -1078,7 +1080,7 @@ abortit:
  			dp->i_nlink--;
  			dp->i_flag |= IN_CHANGE;
  			error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
 -				sizeof (struct dirtemplate), (off_t)0,
 +				sizeof(struct dirtemplate), (off_t)0,
  				UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
  				tcnp->cn_cred, NOCRED, NULL, NULL);
  			if (error == 0) {
 @@ -1093,7 +1095,7 @@ abortit:
  					dirbuf.dotdot_ino = newparent;
  					(void) vn_rdwr(UIO_WRITE, fvp,
  					    (caddr_t)&dirbuf,
 -					    sizeof (struct dirtemplate),
 +					    sizeof(struct dirtemplate),
  					    (off_t)0, UIO_SYSSPACE,
  					    IO_NODELOCKED | IO_SYNC |
  					    IO_NOMACCHECK, tcnp->cn_cred,
 @@ -1231,7 +1233,7 @@ ext2_mkdir(ap)
  #define DIRBLKSIZ  VTOI(dvp)->i_e2fs->e2fs_bsize
  	dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
  	error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
 -	    sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
 +	    sizeof(dirtemplate), (off_t)0, UIO_SYSSPACE,
  	    IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
  	    NULL, NULL);
  	if (error) {
 @@ -1675,3 +1677,328 @@ bad:
  	vput(tvp);
  	return (error);
  }
 +
 +/*
 + * Vnode op for reading.
 + */
 +static int
 +ext2_read(ap)
 +	struct vop_read_args /* {
 +		struct vnode *a_vp;
 +		struct uio *a_uio;
 +		int a_ioflag;
 +		struct ucred *a_cred;
 +	} */ *ap;
 +{
 +	struct vnode *vp;
 +	struct inode *ip;
 +	struct uio *uio;
 +	struct m_ext2fs *fs;
 +	struct buf *bp;
 +	daddr_t lbn, nextlbn;
 +	off_t bytesinfile;
 +	long size, xfersize, blkoffset;
 +	int error, orig_resid, seqcount;
 +	int ioflag;
 +
 +	vp = ap->a_vp;
 +	uio = ap->a_uio;
 +	ioflag = ap->a_ioflag;
 +
 +	seqcount = ap->a_ioflag >> IO_SEQSHIFT;
 +	ip = VTOI(vp);
 +
 +#ifdef INVARIANTS
 +	if (uio->uio_rw != UIO_READ)
 +		panic("%s: mode", "ext2_read");
 +
 +	if (vp->v_type == VLNK) {
 +		if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
 +			panic("%s: short symlink", "ext2_read");
 +	} else if (vp->v_type != VREG && vp->v_type != VDIR)
 +		panic("%s: type %d", "ext2_read", vp->v_type);
 +#endif
 +	orig_resid = uio->uio_resid;
 +	KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
 +	if (orig_resid == 0)
 +		return (0);
 +	KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
 +	fs = ip->i_e2fs;
 +	if (uio->uio_offset < ip->i_size &&
 +	    uio->uio_offset >= fs->e2fs_maxfilesize)
 +	    	return (EOVERFLOW);
 +
 +	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
 +		if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
 +			break;
 +		lbn = lblkno(fs, uio->uio_offset);
 +		nextlbn = lbn + 1;
 +		size = blksize(fs, ip, lbn);
 +		blkoffset = blkoff(fs, uio->uio_offset);
 +
 +		xfersize = fs->e2fs_fsize - blkoffset;
 +		if (uio->uio_resid < xfersize)
 +			xfersize = uio->uio_resid;
 +		if (bytesinfile < xfersize)
 +			xfersize = bytesinfile;
 +
 +		if (lblktosize(fs, nextlbn) >= ip->i_size)
 +			error = bread(vp, lbn, size, NOCRED, &bp);
 +		else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0)
 +			error = cluster_read(vp, ip->i_size, lbn, size,
 +			    NOCRED, blkoffset + uio->uio_resid, seqcount, &bp);
 +		else if (seqcount > 1) {
 +			int nextsize = blksize(fs, ip, nextlbn);
 +			error = breadn(vp, lbn,
 +			    size, &nextlbn, &nextsize, 1, NOCRED, &bp);
 +		} else
 +			error = bread(vp, lbn, size, NOCRED, &bp);
 +		if (error) {
 +			brelse(bp);
 +			bp = NULL;
 +			break;
 +		}
 +
 +		/*
 +		 * If IO_DIRECT then set B_DIRECT for the buffer.  This
 +		 * will cause us to attempt to release the buffer later on
 +		 * and will cause the buffer cache to attempt to free the
 +		 * underlying pages.
 +		 */
 +		if (ioflag & IO_DIRECT)
 +			bp->b_flags |= B_DIRECT;
 +
 +		/*
 +		 * We should only get non-zero b_resid when an I/O error
 +		 * has occurred, which should cause us to break above.
 +		 * However, if the short read did not cause an error,
 +		 * then we want to ensure that we do not uiomove bad
 +		 * or uninitialized data.
 +		 */
 +		size -= bp->b_resid;
 +		if (size < xfersize) {
 +			if (size == 0)
 +				break;
 +			xfersize = size;
 +		}
 +		error = uiomove((char *)bp->b_data + blkoffset,
 +  			(int)xfersize, uio);
 +		if (error)
 +			break;
 +
 +		if (ioflag & (IO_VMIO|IO_DIRECT)) {
 +			/*
 +			 * If it's VMIO or direct I/O, then we don't
 +			 * need the buf, mark it available for
 +			 * freeing. If it's non-direct VMIO, the VM has
 +			 * the data.
 +			 */
 +			bp->b_flags |= B_RELBUF;
 +			brelse(bp);
 +		} else {
 +			/*
 +			 * Otherwise let whoever
 +			 * made the request take care of
 +			 * freeing it. We just queue
 +			 * it onto another list.
 +			 */
 +			bqrelse(bp);
 +		}
 +	}
 +
 +	/* 
 +	 * This can only happen in the case of an error
 +	 * because the loop above resets bp to NULL on each iteration
 +	 * and on normal completion has not set a new value into it.
 +	 * so it must have come from a 'break' statement
 +	 */
 +	if (bp != NULL) {
 +		if (ioflag & (IO_VMIO|IO_DIRECT)) {
 +			bp->b_flags |= B_RELBUF;
 +			brelse(bp);
 +		} else {
 +			bqrelse(bp);
 +		}
 +	}
 +
 +	if ((error == 0 || uio->uio_resid != orig_resid) &&
 +	    (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
 +		ip->i_flag |= IN_ACCESS;
 +	return (error);
 +}
 +
 +/*
 + * Vnode op for writing.
 + */
 +static int
 +ext2_write(ap)
 +	struct vop_write_args /* {
 +		struct vnode *a_vp;
 +		struct uio *a_uio;
 +		int a_ioflag;
 +		struct ucred *a_cred;
 +	} */ *ap;
 +{
 +	struct vnode *vp;
 +	struct uio *uio;
 +	struct inode *ip;
 +	struct m_ext2fs *fs;
 +	struct buf *bp;
 +	daddr_t lbn;
 +	off_t osize;
 +	int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
 +
 +	ioflag = ap->a_ioflag;
 +	uio = ap->a_uio;
 +	vp = ap->a_vp;
 +
 +	seqcount = ioflag >> IO_SEQSHIFT;
 +	ip = VTOI(vp);
 +
 +#ifdef INVARIANTS
 +	if (uio->uio_rw != UIO_WRITE)
 +		panic("%s: mode", "ext2_write");
 +#endif
 +
 +	switch (vp->v_type) {
 +	case VREG:
 +		if (ioflag & IO_APPEND)
 +			uio->uio_offset = ip->i_size;
 +		if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
 +			return (EPERM);
 +		/* FALLTHROUGH */
 +	case VLNK:
 +		break;
 +	case VDIR:
 +		/* XXX differs from ffs -- this is called from ext2_mkdir(). */
 +		if ((ioflag & IO_SYNC) == 0)
 +		panic("ext2_write: nonsync dir write");
 +		break;
 +	default:
 +		panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
 +		    vp->v_type, (intmax_t)uio->uio_offset,
 +		    (intmax_t)uio->uio_resid);
 +	}
 +
 +	KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
 +	KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
 +	fs = ip->i_e2fs;
 +	if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
 +		return (EFBIG);
 +	/*
 +	 * Maybe this should be above the vnode op call, but so long as
 +	 * file servers have no limits, I don't think it matters.
 +	 */
 +	if (vn_rlimit_fsize(vp, uio, uio->uio_td))
 +		return (EFBIG);
 +
 +	resid = uio->uio_resid;
 +	osize = ip->i_size;
 +	if (seqcount > BA_SEQMAX)
 +		flags = BA_SEQMAX << BA_SEQSHIFT;
 +	else
 +		flags = seqcount << BA_SEQSHIFT;
 +	if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
 +		flags |= IO_SYNC;
 +
 +	for (error = 0; uio->uio_resid > 0;) {
 +		lbn = lblkno(fs, uio->uio_offset);
 +		blkoffset = blkoff(fs, uio->uio_offset);
 +		xfersize = fs->e2fs_fsize - blkoffset;
 +		if (uio->uio_resid < xfersize)
 +			xfersize = uio->uio_resid;
 +		if (uio->uio_offset + xfersize > ip->i_size)
 +			vnode_pager_setsize(vp, uio->uio_offset + xfersize);
 +
 +                /*
 +		 * We must perform a read-before-write if the transfer size
 +		 * does not cover the entire buffer.
 +                 */
 +		if (fs->e2fs_bsize > xfersize)
 +			flags |= BA_CLRBUF;
 +		else
 +			flags &= ~BA_CLRBUF;
 +		error = ext2_balloc(ip, lbn, blkoffset + xfersize,
 +		    ap->a_cred, &bp, flags);
 +		if (error != 0)
 +			break;
 +
 +		/*
 +		 * If the buffer is not valid and we did not clear garbage
 +		 * out above, we have to do so here even though the write
 +		 * covers the entire buffer in order to avoid a mmap()/write
 +		 * race where another process may see the garbage prior to
 +		 * the uiomove() for a write replacing it.
 
 *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
 _______________________________________________
 svn-src-all@freebsd.org mailing list
 http://lists.freebsd.org/mailman/listinfo/svn-src-all
 To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
 
State-Changed-From-To: patched->closed 
State-Changed-By: pfg 
State-Changed-When: Thu Jan 5 01:50:47 UTC 2012 
State-Changed-Why:  
Committed and MFC'd. 

http://www.freebsd.org/cgi/query-pr.cgi?pr=159232 
>Unformatted:
