Linux-2.6.12-rc2v2.6.12-rc2

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!

12 files changed, 8955 insertions, 0 deletions

diff --git a/fs/xfs/quota/xfs_dquot.c b/fs/xfs/quota/xfs_dquot.c
new file mode 100644
index 000000000000..740d20d33187
--- /dev/null
+++ b/fs/xfs/quota/xfs_dquot.c
@@ -0,0 +1,1648 @@
+/*
+ * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_ag.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+#include "xfs_trans_space.h"
+#include "xfs_trans_priv.h"
+
+#include "xfs_qm.h"
+
+
+/*
+   LOCK ORDER
+
+   inode lock		    (ilock)
+   dquot hash-chain lock    (hashlock)
+   xqm dquot freelist lock  (freelistlock
+   mount's dquot list lock  (mplistlock)
+   user dquot lock - lock ordering among dquots is based on the uid or gid
+   group dquot lock - similar to udquots. Between the two dquots, the udquot
+		      has to be locked first.
+   pin lock - the dquot lock must be held to take this lock.
+   flush lock - ditto.
+*/
+
+STATIC void		xfs_qm_dqflush_done(xfs_buf_t *, xfs_dq_logitem_t *);
+
+#ifdef DEBUG
+xfs_buftarg_t *xfs_dqerror_target;
+int xfs_do_dqerror;
+int xfs_dqreq_num;
+int xfs_dqerror_mod = 33;
+#endif
+
+/*
+ * Allocate and initialize a dquot. We don't always allocate fresh memory;
+ * we try to reclaim a free dquot if the number of incore dquots are above
+ * a threshold.
+ * The only field inside the core that gets initialized at this point
+ * is the d_id field. The idea is to fill in the entire q_core
+ * when we read in the on disk dquot.
+ */
+xfs_dquot_t *
+xfs_qm_dqinit(
+	xfs_mount_t  *mp,
+	xfs_dqid_t   id,
+	uint	     type)
+{
+	xfs_dquot_t	*dqp;
+	boolean_t	brandnewdquot;
+
+	brandnewdquot = xfs_qm_dqalloc_incore(&dqp);
+	dqp->dq_flags = type;
+	INT_SET(dqp->q_core.d_id, ARCH_CONVERT, id);
+	dqp->q_mount = mp;
+
+	/*
+	 * No need to re-initialize these if this is a reclaimed dquot.
+	 */
+	if (brandnewdquot) {
+		dqp->dq_flnext = dqp->dq_flprev = dqp;
+		mutex_init(&dqp->q_qlock,  MUTEX_DEFAULT, "xdq");
+		initnsema(&dqp->q_flock, 1, "fdq");
+		sv_init(&dqp->q_pinwait, SV_DEFAULT, "pdq");
+
+#ifdef XFS_DQUOT_TRACE
+		dqp->q_trace = ktrace_alloc(DQUOT_TRACE_SIZE, KM_SLEEP);
+		xfs_dqtrace_entry(dqp, "DQINIT");
+#endif
+	} else {
+		/*
+		 * Only the q_core portion was zeroed in dqreclaim_one().
+		 * So, we need to reset others.
+		 */
+		 dqp->q_nrefs = 0;
+		 dqp->q_blkno = 0;
+		 dqp->MPL_NEXT = dqp->HL_NEXT = NULL;
+		 dqp->HL_PREVP = dqp->MPL_PREVP = NULL;
+		 dqp->q_bufoffset = 0;
+		 dqp->q_fileoffset = 0;
+		 dqp->q_transp = NULL;
+		 dqp->q_gdquot = NULL;
+		 dqp->q_res_bcount = 0;
+		 dqp->q_res_icount = 0;
+		 dqp->q_res_rtbcount = 0;
+		 dqp->q_pincount = 0;
+		 dqp->q_hash = NULL;
+		 ASSERT(dqp->dq_flnext == dqp->dq_flprev);
+
+#ifdef XFS_DQUOT_TRACE
+		 ASSERT(dqp->q_trace);
+		 xfs_dqtrace_entry(dqp, "DQRECLAIMED_INIT");
+#endif
+	 }
+
+	/*
+	 * log item gets initialized later
+	 */
+	return (dqp);
+}
+
+/*
+ * This is called to free all the memory associated with a dquot
+ */
+void
+xfs_qm_dqdestroy(
+	xfs_dquot_t	*dqp)
+{
+	ASSERT(! XFS_DQ_IS_ON_FREELIST(dqp));
+
+	mutex_destroy(&dqp->q_qlock);
+	freesema(&dqp->q_flock);
+	sv_destroy(&dqp->q_pinwait);
+
+#ifdef XFS_DQUOT_TRACE
+	if (dqp->q_trace)
+	     ktrace_free(dqp->q_trace);
+	dqp->q_trace = NULL;
+#endif
+	kmem_zone_free(xfs_Gqm->qm_dqzone, dqp);
+	atomic_dec(&xfs_Gqm->qm_totaldquots);
+}
+
+/*
+ * This is what a 'fresh' dquot inside a dquot chunk looks like on disk.
+ */
+STATIC void
+xfs_qm_dqinit_core(
+	xfs_dqid_t	id,
+	uint		type,
+	xfs_dqblk_t	*d)
+{
+	/*
+	 * Caller has zero'd the entire dquot 'chunk' already.
+	 */
+	INT_SET(d->dd_diskdq.d_magic, ARCH_CONVERT, XFS_DQUOT_MAGIC);
+	INT_SET(d->dd_diskdq.d_version, ARCH_CONVERT, XFS_DQUOT_VERSION);
+	INT_SET(d->dd_diskdq.d_id, ARCH_CONVERT, id);
+	INT_SET(d->dd_diskdq.d_flags, ARCH_CONVERT, type);
+}
+
+
+#ifdef XFS_DQUOT_TRACE
+/*
+ * Dquot tracing for debugging.
+ */
+/* ARGSUSED */
+void
+__xfs_dqtrace_entry(
+	xfs_dquot_t	*dqp,
+	char		*func,
+	void		*retaddr,
+	xfs_inode_t	*ip)
+{
+	xfs_dquot_t	*udqp = NULL;
+	xfs_ino_t	ino = 0;
+
+	ASSERT(dqp->q_trace);
+	if (ip) {
+		ino = ip->i_ino;
+		udqp = ip->i_udquot;
+	}
+	ktrace_enter(dqp->q_trace,
+		     (void *)(__psint_t)DQUOT_KTRACE_ENTRY,
+		     (void *)func,
+		     (void *)(__psint_t)dqp->q_nrefs,
+		     (void *)(__psint_t)dqp->dq_flags,
+		     (void *)(__psint_t)dqp->q_res_bcount,
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_bcount,
+						ARCH_CONVERT),
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_icount,
+						ARCH_CONVERT),
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_blk_hardlimit,
+						ARCH_CONVERT),
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_blk_softlimit,
+						ARCH_CONVERT),
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_ino_hardlimit,
+						ARCH_CONVERT),
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_ino_softlimit,
+						ARCH_CONVERT),
+		     (void *)(__psint_t)INT_GET(dqp->q_core.d_id, ARCH_CONVERT),
+		     (void *)(__psint_t)current_pid(),
+		     (void *)(__psint_t)ino,
+		     (void *)(__psint_t)retaddr,
+		     (void *)(__psint_t)udqp);
+	return;
+}
+#endif
+
+
+/*
+ * If default limits are in force, push them into the dquot now.
+ * We overwrite the dquot limits only if they are zero and this
+ * is not the root dquot.
+ */
+void
+xfs_qm_adjust_dqlimits(
+	xfs_mount_t		*mp,
+	xfs_disk_dquot_t	*d)
+{
+	xfs_quotainfo_t		*q = mp->m_quotainfo;
+
+	ASSERT(d->d_id);
+
+	if (q->qi_bsoftlimit && !d->d_blk_softlimit)
+		INT_SET(d->d_blk_softlimit, ARCH_CONVERT, q->qi_bsoftlimit);
+	if (q->qi_bhardlimit && !d->d_blk_hardlimit)
+		INT_SET(d->d_blk_hardlimit, ARCH_CONVERT, q->qi_bhardlimit);
+	if (q->qi_isoftlimit && !d->d_ino_softlimit)
+		INT_SET(d->d_ino_softlimit, ARCH_CONVERT, q->qi_isoftlimit);
+	if (q->qi_ihardlimit && !d->d_ino_hardlimit)
+		INT_SET(d->d_ino_hardlimit, ARCH_CONVERT, q->qi_ihardlimit);
+	if (q->qi_rtbsoftlimit && !d->d_rtb_softlimit)
+		INT_SET(d->d_rtb_softlimit, ARCH_CONVERT, q->qi_rtbsoftlimit);
+	if (q->qi_rtbhardlimit && !d->d_rtb_hardlimit)
+		INT_SET(d->d_rtb_hardlimit, ARCH_CONVERT, q->qi_rtbhardlimit);
+}
+
+/*
+ * Check the limits and timers of a dquot and start or reset timers
+ * if necessary.
+ * This gets called even when quota enforcement is OFF, which makes our
+ * life a little less complicated. (We just don't reject any quota
+ * reservations in that case, when enforcement is off).
+ * We also return 0 as the values of the timers in Q_GETQUOTA calls, when
+ * enforcement's off.
+ * In contrast, warnings are a little different in that they don't
+ * 'automatically' get started when limits get exceeded.
+ */
+void
+xfs_qm_adjust_dqtimers(
+	xfs_mount_t		*mp,
+	xfs_disk_dquot_t	*d)
+{
+	ASSERT(d->d_id);
+
+#ifdef QUOTADEBUG
+	if (INT_GET(d->d_blk_hardlimit, ARCH_CONVERT))
+		ASSERT(INT_GET(d->d_blk_softlimit, ARCH_CONVERT) <=
+			INT_GET(d->d_blk_hardlimit, ARCH_CONVERT));
+	if (INT_GET(d->d_ino_hardlimit, ARCH_CONVERT))
+		ASSERT(INT_GET(d->d_ino_softlimit, ARCH_CONVERT) <=
+			INT_GET(d->d_ino_hardlimit, ARCH_CONVERT));
+	if (INT_GET(d->d_rtb_hardlimit, ARCH_CONVERT))
+		ASSERT(INT_GET(d->d_rtb_softlimit, ARCH_CONVERT) <=
+			INT_GET(d->d_rtb_hardlimit, ARCH_CONVERT));
+#endif
+	if (!d->d_btimer) {
+		if ((INT_GET(d->d_blk_softlimit, ARCH_CONVERT) &&
+		    (INT_GET(d->d_bcount, ARCH_CONVERT) >=
+				INT_GET(d->d_blk_softlimit, ARCH_CONVERT))) ||
+		    (INT_GET(d->d_blk_hardlimit, ARCH_CONVERT) &&
+		    (INT_GET(d->d_bcount, ARCH_CONVERT) >=
+				INT_GET(d->d_blk_hardlimit, ARCH_CONVERT)))) {
+			INT_SET(d->d_btimer, ARCH_CONVERT,
+				get_seconds() + XFS_QI_BTIMELIMIT(mp));
+		}
+	} else {
+		if ((!d->d_blk_softlimit ||
+		    (INT_GET(d->d_bcount, ARCH_CONVERT) <
+				INT_GET(d->d_blk_softlimit, ARCH_CONVERT))) &&
+		    (!d->d_blk_hardlimit ||
+		    (INT_GET(d->d_bcount, ARCH_CONVERT) <
+				INT_GET(d->d_blk_hardlimit, ARCH_CONVERT)))) {
+			d->d_btimer = 0;
+		}
+	}
+
+	if (!d->d_itimer) {
+		if ((INT_GET(d->d_ino_softlimit, ARCH_CONVERT) &&
+		    (INT_GET(d->d_icount, ARCH_CONVERT) >=
+				INT_GET(d->d_ino_softlimit, ARCH_CONVERT))) ||
+		    (INT_GET(d->d_ino_hardlimit, ARCH_CONVERT) &&
+		    (INT_GET(d->d_icount, ARCH_CONVERT) >=
+				INT_GET(d->d_ino_hardlimit, ARCH_CONVERT)))) {
+			INT_SET(d->d_itimer, ARCH_CONVERT,
+				get_seconds() + XFS_QI_ITIMELIMIT(mp));
+		}
+	} else {
+		if ((!d->d_ino_softlimit ||
+		    (INT_GET(d->d_icount, ARCH_CONVERT) <
+				INT_GET(d->d_ino_softlimit, ARCH_CONVERT)))  &&
+		    (!d->d_ino_hardlimit ||
+		    (INT_GET(d->d_icount, ARCH_CONVERT) <
+				INT_GET(d->d_ino_hardlimit, ARCH_CONVERT)))) {
+			d->d_itimer = 0;
+		}
+	}
+
+	if (!d->d_rtbtimer) {
+		if ((INT_GET(d->d_rtb_softlimit, ARCH_CONVERT) &&
+		    (INT_GET(d->d_rtbcount, ARCH_CONVERT) >=
+				INT_GET(d->d_rtb_softlimit, ARCH_CONVERT))) ||
+		    (INT_GET(d->d_rtb_hardlimit, ARCH_CONVERT) &&
+		    (INT_GET(d->d_rtbcount, ARCH_CONVERT) >=
+				INT_GET(d->d_rtb_hardlimit, ARCH_CONVERT)))) {
+			INT_SET(d->d_rtbtimer, ARCH_CONVERT,
+				get_seconds() + XFS_QI_RTBTIMELIMIT(mp));
+		}
+	} else {
+		if ((!d->d_rtb_softlimit ||
+		    (INT_GET(d->d_rtbcount, ARCH_CONVERT) <
+				INT_GET(d->d_rtb_softlimit, ARCH_CONVERT))) &&
+		    (!d->d_rtb_hardlimit ||
+		    (INT_GET(d->d_rtbcount, ARCH_CONVERT) <
+				INT_GET(d->d_rtb_hardlimit, ARCH_CONVERT)))) {
+			d->d_rtbtimer = 0;
+		}
+	}
+}
+
+/*
+ * Increment or reset warnings of a given dquot.
+ */
+int
+xfs_qm_dqwarn(
+	xfs_disk_dquot_t	*d,
+	uint			flags)
+{
+	int	warned;
+
+	/*
+	 * root's limits are not real limits.
+	 */
+	if (!d->d_id)
+		return (0);
+
+	warned = 0;
+	if (INT_GET(d->d_blk_softlimit, ARCH_CONVERT) &&
+	    (INT_GET(d->d_bcount, ARCH_CONVERT) >=
+	     INT_GET(d->d_blk_softlimit, ARCH_CONVERT))) {
+		if (flags & XFS_QMOPT_DOWARN) {
+			INT_MOD(d->d_bwarns, ARCH_CONVERT, +1);
+			warned++;
+		}
+	} else {
+		if (!d->d_blk_softlimit ||
+		    (INT_GET(d->d_bcount, ARCH_CONVERT) <
+		     INT_GET(d->d_blk_softlimit, ARCH_CONVERT))) {
+			d->d_bwarns = 0;
+		}
+	}
+
+	if (INT_GET(d->d_ino_softlimit, ARCH_CONVERT) > 0 &&
+	    (INT_GET(d->d_icount, ARCH_CONVERT) >=
+	     INT_GET(d->d_ino_softlimit, ARCH_CONVERT))) {
+		if (flags & XFS_QMOPT_DOWARN) {
+			INT_MOD(d->d_iwarns, ARCH_CONVERT, +1);
+			warned++;
+		}
+	} else {
+		if (!d->d_ino_softlimit ||
+		    (INT_GET(d->d_icount, ARCH_CONVERT) <
+		     INT_GET(d->d_ino_softlimit, ARCH_CONVERT))) {
+			d->d_iwarns = 0;
+		}
+	}
+#ifdef QUOTADEBUG
+	if (INT_GET(d->d_iwarns, ARCH_CONVERT))
+		cmn_err(CE_DEBUG,
+			"--------@@Inode warnings running : %Lu >= %Lu",
+			INT_GET(d->d_icount, ARCH_CONVERT),
+			INT_GET(d->d_ino_softlimit, ARCH_CONVERT));
+	if (INT_GET(d->d_bwarns, ARCH_CONVERT))
+		cmn_err(CE_DEBUG,
+			"--------@@Blks warnings running : %Lu >= %Lu",
+			INT_GET(d->d_bcount, ARCH_CONVERT),
+			INT_GET(d->d_blk_softlimit, ARCH_CONVERT));
+#endif
+	return (warned);
+}
+
+
+/*
+ * initialize a buffer full of dquots and log the whole thing
+ */
+STATIC void
+xfs_qm_init_dquot_blk(
+	xfs_trans_t	*tp,
+	xfs_mount_t	*mp,
+	xfs_dqid_t	id,
+	uint		type,
+	xfs_buf_t	*bp)
+{
+	xfs_dqblk_t	*d;
+	int		curid, i;
+
+	ASSERT(tp);
+	ASSERT(XFS_BUF_ISBUSY(bp));
+	ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
+
+	d = (xfs_dqblk_t *)XFS_BUF_PTR(bp);
+
+	/*
+	 * ID of the first dquot in the block - id's are zero based.
+	 */
+	curid = id - (id % XFS_QM_DQPERBLK(mp));
+	ASSERT(curid >= 0);
+	memset(d, 0, BBTOB(XFS_QI_DQCHUNKLEN(mp)));
+	for (i = 0; i < XFS_QM_DQPERBLK(mp); i++, d++, curid++)
+		xfs_qm_dqinit_core(curid, type, d);
+	xfs_trans_dquot_buf(tp, bp,
+			    type & XFS_DQ_USER ?
+			    XFS_BLI_UDQUOT_BUF :
+			    XFS_BLI_GDQUOT_BUF);
+	xfs_trans_log_buf(tp, bp, 0, BBTOB(XFS_QI_DQCHUNKLEN(mp)) - 1);
+}
+
+
+
+/*
+ * Allocate a block and fill it with dquots.
+ * This is called when the bmapi finds a hole.
+ */
+STATIC int
+xfs_qm_dqalloc(
+	xfs_trans_t	*tp,
+	xfs_mount_t	*mp,
+	xfs_dquot_t	*dqp,
+	xfs_inode_t	*quotip,
+	xfs_fileoff_t	offset_fsb,
+	xfs_buf_t	**O_bpp)
+{
+	xfs_fsblock_t	firstblock;
+	xfs_bmap_free_t flist;
+	xfs_bmbt_irec_t map;
+	int		nmaps, error, committed;
+	xfs_buf_t	*bp;
+
+	ASSERT(tp != NULL);
+	xfs_dqtrace_entry(dqp, "DQALLOC");
+
+	/*
+	 * Initialize the bmap freelist prior to calling bmapi code.
+	 */
+	XFS_BMAP_INIT(&flist, &firstblock);
+	xfs_ilock(quotip, XFS_ILOCK_EXCL);
+	/*
+	 * Return if this type of quotas is turned off while we didn't
+	 * have an inode lock
+	 */
+	if (XFS_IS_THIS_QUOTA_OFF(dqp)) {
+		xfs_iunlock(quotip, XFS_ILOCK_EXCL);
+		return (ESRCH);
+	}
+
+	/*
+	 * xfs_trans_commit normally decrements the vnode ref count
+	 * when it unlocks the inode. Since we want to keep the quota
+	 * inode around, we bump the vnode ref count now.
+	 */
+	VN_HOLD(XFS_ITOV(quotip));
+
+	xfs_trans_ijoin(tp, quotip, XFS_ILOCK_EXCL);
+	nmaps = 1;
+	if ((error = xfs_bmapi(tp, quotip,
+			      offset_fsb, XFS_DQUOT_CLUSTER_SIZE_FSB,
+			      XFS_BMAPI_METADATA | XFS_BMAPI_WRITE,
+			      &firstblock,
+			      XFS_QM_DQALLOC_SPACE_RES(mp),
+			      &map, &nmaps, &flist))) {
+		goto error0;
+	}
+	ASSERT(map.br_blockcount == XFS_DQUOT_CLUSTER_SIZE_FSB);
+	ASSERT(nmaps == 1);
+	ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
+	       (map.br_startblock != HOLESTARTBLOCK));
+
+	/*
+	 * Keep track of the blkno to save a lookup later
+	 */
+	dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
+
+	/* now we can just get the buffer (there's nothing to read yet) */
+	bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
+			       dqp->q_blkno,
+			       XFS_QI_DQCHUNKLEN(mp),
+			       0);
+	if (!bp || (error = XFS_BUF_GETERROR(bp)))
+		goto error1;
+	/*
+	 * Make a chunk of dquots out of this buffer and log
+	 * the entire thing.
+	 */
+	xfs_qm_init_dquot_blk(tp, mp, INT_GET(dqp->q_core.d_id, ARCH_CONVERT),
+			      dqp->dq_flags & (XFS_DQ_USER|XFS_DQ_GROUP),
+			      bp);
+
+	if ((error = xfs_bmap_finish(&tp, &flist, firstblock, &committed))) {
+		goto error1;
+	}
+
+	*O_bpp = bp;
+	return 0;
+
+      error1:
+	xfs_bmap_cancel(&flist);
+      error0:
+	xfs_iunlock(quotip, XFS_ILOCK_EXCL);
+
+	return (error);
+}
+
+/*
+ * Maps a dquot to the buffer containing its on-disk version.
+ * This returns a ptr to the buffer containing the on-disk dquot
+ * in the bpp param, and a ptr to the on-disk dquot within that buffer
+ */
+STATIC int
+xfs_qm_dqtobp(
+	xfs_trans_t		*tp,
+	xfs_dquot_t		*dqp,
+	xfs_disk_dquot_t	**O_ddpp,
+	xfs_buf_t		**O_bpp,
+	uint			flags)
+{
+	xfs_bmbt_irec_t map;
+	int		nmaps, error;
+	xfs_buf_t	*bp;
+	xfs_inode_t	*quotip;
+	xfs_mount_t	*mp;
+	xfs_disk_dquot_t *ddq;
+	xfs_dqid_t	id;
+	boolean_t	newdquot;
+
+	mp = dqp->q_mount;
+	id = INT_GET(dqp->q_core.d_id, ARCH_CONVERT);
+	nmaps = 1;
+	newdquot = B_FALSE;
+
+	/*
+	 * If we don't know where the dquot lives, find out.
+	 */
+	if (dqp->q_blkno == (xfs_daddr_t) 0) {
+		/* We use the id as an index */
+		dqp->q_fileoffset = (xfs_fileoff_t) ((uint)id /
+						     XFS_QM_DQPERBLK(mp));
+		nmaps = 1;
+		quotip = XFS_DQ_TO_QIP(dqp);
+		xfs_ilock(quotip, XFS_ILOCK_SHARED);
+		/*
+		 * Return if this type of quotas is turned off while we didn't
+		 * have an inode lock
+		 */
+		if (XFS_IS_THIS_QUOTA_OFF(dqp)) {
+			xfs_iunlock(quotip, XFS_ILOCK_SHARED);
+			return (ESRCH);
+		}
+		/*
+		 * Find the block map; no allocations yet
+		 */
+		error = xfs_bmapi(NULL, quotip, dqp->q_fileoffset,
+				  XFS_DQUOT_CLUSTER_SIZE_FSB,
+				  XFS_BMAPI_METADATA,
+				  NULL, 0, &map, &nmaps, NULL);
+
+		xfs_iunlock(quotip, XFS_ILOCK_SHARED);
+		if (error)
+			return (error);
+		ASSERT(nmaps == 1);
+		ASSERT(map.br_blockcount == 1);
+
+		/*
+		 * offset of dquot in the (fixed sized) dquot chunk.
+		 */
+		dqp->q_bufoffset = (id % XFS_QM_DQPERBLK(mp)) *
+			sizeof(xfs_dqblk_t);
+		if (map.br_startblock == HOLESTARTBLOCK) {
+			/*
+			 * We don't allocate unless we're asked to
+			 */
+			if (!(flags & XFS_QMOPT_DQALLOC))
+				return (ENOENT);
+
+			ASSERT(tp);
+			if ((error = xfs_qm_dqalloc(tp, mp, dqp, quotip,
+						dqp->q_fileoffset, &bp)))
+				return (error);
+			newdquot = B_TRUE;
+		} else {
+			/*
+			 * store the blkno etc so that we don't have to do the
+			 * mapping all the time
+			 */
+			dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
+		}
+	}
+	ASSERT(dqp->q_blkno != DELAYSTARTBLOCK);
+	ASSERT(dqp->q_blkno != HOLESTARTBLOCK);
+
+	/*
+	 * Read in the buffer, unless we've just done the allocation
+	 * (in which case we already have the buf).
+	 */
+	if (! newdquot) {
+		xfs_dqtrace_entry(dqp, "DQTOBP READBUF");
+		if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
+					       dqp->q_blkno,
+					       XFS_QI_DQCHUNKLEN(mp),
+					       0, &bp))) {
+			return (error);
+		}
+		if (error || !bp)
+			return XFS_ERROR(error);
+	}
+	ASSERT(XFS_BUF_ISBUSY(bp));
+	ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
+
+	/*
+	 * calculate the location of the dquot inside the buffer.
+	 */
+	ddq = (xfs_disk_dquot_t *)((char *)XFS_BUF_PTR(bp) + dqp->q_bufoffset);
+
+	/*
+	 * A simple sanity check in case we got a corrupted dquot...
+	 */
+	if (xfs_qm_dqcheck(ddq, id,
+			   dqp->dq_flags & (XFS_DQ_USER|XFS_DQ_GROUP),
+			   flags & (XFS_QMOPT_DQREPAIR|XFS_QMOPT_DOWARN),
+			   "dqtobp")) {
+		if (!(flags & XFS_QMOPT_DQREPAIR)) {
+			xfs_trans_brelse(tp, bp);
+			return XFS_ERROR(EIO);
+		}
+		XFS_BUF_BUSY(bp); /* We dirtied this */
+	}
+
+	*O_bpp = bp;
+	*O_ddpp = ddq;
+
+	return (0);
+}
+
+
+/*
+ * Read in the ondisk dquot using dqtobp() then copy it to an incore version,
+ * and release the buffer immediately.
+ *
+ */
+/* ARGSUSED */
+STATIC int
+xfs_qm_dqread(
+	xfs_trans_t	*tp,
+	xfs_dqid_t	id,
+	xfs_dquot_t	*dqp,	/* dquot to get filled in */
+	uint		flags)
+{
+	xfs_disk_dquot_t *ddqp;
+	xfs_buf_t	 *bp;
+	int		 error;
+
+	/*
+	 * get a pointer to the on-disk dquot and the buffer containing it
+	 * dqp already knows its own type (GROUP/USER).
+	 */
+	xfs_dqtrace_entry(dqp, "DQREAD");
+	if ((error = xfs_qm_dqtobp(tp, dqp, &ddqp, &bp, flags))) {
+		return (error);
+	}
+
+	/* copy everything from disk dquot to the incore dquot */
+	memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t));
+	ASSERT(INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id);
+	xfs_qm_dquot_logitem_init(dqp);
+
+	/*
+	 * Reservation counters are defined as reservation plus current usage
+	 * to avoid having to add everytime.
+	 */
+	dqp->q_res_bcount = INT_GET(ddqp->d_bcount, ARCH_CONVERT);
+	dqp->q_res_icount = INT_GET(ddqp->d_icount, ARCH_CONVERT);
+	dqp->q_res_rtbcount = INT_GET(ddqp->d_rtbcount, ARCH_CONVERT);
+
+	/* Mark the buf so that this will stay incore a little longer */
+	XFS_BUF_SET_VTYPE_REF(bp, B_FS_DQUOT, XFS_DQUOT_REF);
+
+	/*
+	 * We got the buffer with a xfs_trans_read_buf() (in dqtobp())
+	 * So we need to release with xfs_trans_brelse().
+	 * The strategy here is identical to that of inodes; we lock
+	 * the dquot in xfs_qm_dqget() before making it accessible to
+	 * others. This is because dquots, like inodes, need a good level of
+	 * concurrency, and we don't want to take locks on the entire buffers
+	 * for dquot accesses.
+	 * Note also that the dquot buffer may even be dirty at this point, if
+	 * this particular dquot was repaired. We still aren't afraid to
+	 * brelse it because we have the changes incore.
+	 */
+	ASSERT(XFS_BUF_ISBUSY(bp));
+	ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
+	xfs_trans_brelse(tp, bp);
+
+	return (error);
+}
+
+
+/*
+ * allocate an incore dquot from the kernel heap,
+ * and fill its core with quota information kept on disk.
+ * If XFS_QMOPT_DQALLOC is set, it'll allocate a dquot on disk
+ * if it wasn't already allocated.
+ */
+STATIC int
+xfs_qm_idtodq(
+	xfs_mount_t	*mp,
+	xfs_dqid_t	id,	 /* gid or uid, depending on type */
+	uint		type,	 /* UDQUOT or GDQUOT */
+	uint		flags,	 /* DQALLOC, DQREPAIR */
+	xfs_dquot_t	**O_dqpp)/* OUT : incore dquot, not locked */
+{
+	xfs_dquot_t	*dqp;
+	int		error;
+	xfs_trans_t	*tp;
+	int		cancelflags=0;
+
+	dqp = xfs_qm_dqinit(mp, id, type);
+	tp = NULL;
+	if (flags & XFS_QMOPT_DQALLOC) {
+		tp = xfs_trans_alloc(mp, XFS_TRANS_QM_DQALLOC);
+		if ((error = xfs_trans_reserve(tp,
+				       XFS_QM_DQALLOC_SPACE_RES(mp),
+				       XFS_WRITE_LOG_RES(mp) +
+					      BBTOB(XFS_QI_DQCHUNKLEN(mp)) - 1 +
+					      128,
+				       0,
+				       XFS_TRANS_PERM_LOG_RES,
+				       XFS_WRITE_LOG_COUNT))) {
+			cancelflags = 0;
+			goto error0;
+		}
+		cancelflags = XFS_TRANS_RELEASE_LOG_RES;
+	}
+
+	/*
+	 * Read it from disk; xfs_dqread() takes care of
+	 * all the necessary initialization of dquot's fields (locks, etc)
+	 */
+	if ((error = xfs_qm_dqread(tp, id, dqp, flags))) {
+		/*
+		 * This can happen if quotas got turned off (ESRCH),
+		 * or if the dquot didn't exist on disk and we ask to
+		 * allocate (ENOENT).
+		 */
+		xfs_dqtrace_entry(dqp, "DQREAD FAIL");
+		cancelflags |= XFS_TRANS_ABORT;
+		goto error0;
+	}
+	if (tp) {
+		if ((error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES,
+					     NULL)))
+			goto error1;
+	}
+
+	*O_dqpp = dqp;
+	return (0);
+
+ error0:
+	ASSERT(error);
+	if (tp)
+		xfs_trans_cancel(tp, cancelflags);
+ error1:
+	xfs_qm_dqdestroy(dqp);
+	*O_dqpp = NULL;
+	return (error);
+}
+
+/*
+ * Lookup a dquot in the incore dquot hashtable. We keep two separate
+ * hashtables for user and group dquots; and, these are global tables
+ * inside the XQM, not per-filesystem tables.
+ * The hash chain must be locked by caller, and it is left locked
+ * on return. Returning dquot is locked.
+ */
+STATIC int
+xfs_qm_dqlookup(
+	xfs_mount_t		*mp,
+	xfs_dqid_t		id,
+	xfs_dqhash_t		*qh,
+	xfs_dquot_t		**O_dqpp)
+{
+	xfs_dquot_t		*dqp;
+	uint			flist_locked;
+	xfs_dquot_t		*d;
+
+	ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
+
+	flist_locked = B_FALSE;
+
+	/*
+	 * Traverse the hashchain looking for a match
+	 */
+	for (dqp = qh->qh_next; dqp != NULL; dqp = dqp->HL_NEXT) {
+		/*
+		 * We already have the hashlock. We don't need the
+		 * dqlock to look at the id field of the dquot, since the
+		 * id can't be modified without the hashlock anyway.
+		 */
+		if (INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id && dqp->q_mount == mp) {
+			xfs_dqtrace_entry(dqp, "DQFOUND BY LOOKUP");
+			/*
+			 * All in core dquots must be on the dqlist of mp
+			 */
+			ASSERT(dqp->MPL_PREVP != NULL);
+
+			xfs_dqlock(dqp);
+			if (dqp->q_nrefs == 0) {
+				ASSERT (XFS_DQ_IS_ON_FREELIST(dqp));
+				if (! xfs_qm_freelist_lock_nowait(xfs_Gqm)) {
+					xfs_dqtrace_entry(dqp, "DQLOOKUP: WANT");
+
+					/*
+					 * We may have raced with dqreclaim_one()
+					 * (and lost). So, flag that we don't
+					 * want the dquot to be reclaimed.
+					 */
+					dqp->dq_flags |= XFS_DQ_WANT;
+					xfs_dqunlock(dqp);
+					xfs_qm_freelist_lock(xfs_Gqm);
+					xfs_dqlock(dqp);
+					dqp->dq_flags &= ~(XFS_DQ_WANT);
+				}
+				flist_locked = B_TRUE;
+			}
+
+			/*
+			 * id couldn't have changed; we had the hashlock all
+			 * along
+			 */
+			ASSERT(INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id);
+
+			if (flist_locked) {
+				if (dqp->q_nrefs != 0) {
+					xfs_qm_freelist_unlock(xfs_Gqm);
+					flist_locked = B_FALSE;
+				} else {
+					/*
+					 * take it off the freelist
+					 */
+					xfs_dqtrace_entry(dqp,
+							"DQLOOKUP: TAKEOFF FL");
+					XQM_FREELIST_REMOVE(dqp);
+					/* xfs_qm_freelist_print(&(xfs_Gqm->
+							qm_dqfreelist),
+							"after removal"); */
+				}
+			}
+
+			/*
+			 * grab a reference
+			 */
+			XFS_DQHOLD(dqp);
+
+			if (flist_locked)
+				xfs_qm_freelist_unlock(xfs_Gqm);
+			/*
+			 * move the dquot to the front of the hashchain
+			 */
+			ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
+			if (dqp->HL_PREVP != &qh->qh_next) {
+				xfs_dqtrace_entry(dqp,
+						  "DQLOOKUP: HASH MOVETOFRONT");
+				if ((d = dqp->HL_NEXT))
+					d->HL_PREVP = dqp->HL_PREVP;
+				*(dqp->HL_PREVP) = d;
+				d = qh->qh_next;
+				d->HL_PREVP = &dqp->HL_NEXT;
+				dqp->HL_NEXT = d;
+				dqp->HL_PREVP = &qh->qh_next;
+				qh->qh_next = dqp;
+			}
+			xfs_dqtrace_entry(dqp, "LOOKUP END");
+			*O_dqpp = dqp;
+			ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
+			return (0);
+		}
+	}
+
+	*O_dqpp = NULL;
+	ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
+	return (1);
+}
+
+/*
+ * Given the file system, inode OR id, and type (UDQUOT/GDQUOT), return a
+ * a locked dquot, doing an allocation (if requested) as needed.
+ * When both an inode and an id are given, the inode's id takes precedence.
+ * That is, if the id changes while we don't hold the ilock inside this
+ * function, the new dquot is returned, not necessarily the one requested
+ * in the id argument.
+ */
+int
+xfs_qm_dqget(
+	xfs_mount_t	*mp,
+	xfs_inode_t	*ip,	  /* locked inode (optional) */
+	xfs_dqid_t	id,	  /* gid or uid, depending on type */
+	uint		type,	  /* UDQUOT or GDQUOT */
+	uint		flags,	  /* DQALLOC, DQSUSER, DQREPAIR, DOWARN */
+	xfs_dquot_t	**O_dqpp) /* OUT : locked incore dquot */
+{
+	xfs_dquot_t	*dqp;
+	xfs_dqhash_t	*h;
+	uint		version;
+	int		error;
+
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+	if ((! XFS_IS_UQUOTA_ON(mp) && type == XFS_DQ_USER) ||
+	    (! XFS_IS_GQUOTA_ON(mp) && type == XFS_DQ_GROUP)) {
+		return (ESRCH);
+	}
+	h = XFS_DQ_HASH(mp, id, type);
+
+#ifdef DEBUG
+	if (xfs_do_dqerror) {
+		if ((xfs_dqerror_target == mp->m_ddev_targp) &&
+		    (xfs_dqreq_num++ % xfs_dqerror_mod) == 0) {
+			cmn_err(CE_DEBUG, "Returning error in dqget");
+			return (EIO);
+		}
+	}
+#endif
+
+ again:
+
+#ifdef DEBUG
+	ASSERT(type == XFS_DQ_USER || type == XFS_DQ_GROUP);
+	if (ip) {
+		ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+		if (type == XFS_DQ_USER)
+			ASSERT(ip->i_udquot == NULL);
+		else
+			ASSERT(ip->i_gdquot == NULL);
+	}
+#endif
+	XFS_DQ_HASH_LOCK(h);
+
+	/*
+	 * Look in the cache (hashtable).
+	 * The chain is kept locked during lookup.
+	 */
+	if (xfs_qm_dqlookup(mp, id, h, O_dqpp) == 0) {
+		XQM_STATS_INC(xqmstats.xs_qm_dqcachehits);
+		/*
+		 * The dquot was found, moved to the front of the chain,
+		 * taken off the freelist if it was on it, and locked
+		 * at this point. Just unlock the hashchain and return.
+		 */
+		ASSERT(*O_dqpp);
+		ASSERT(XFS_DQ_IS_LOCKED(*O_dqpp));
+		XFS_DQ_HASH_UNLOCK(h);
+		xfs_dqtrace_entry(*O_dqpp, "DQGET DONE (FROM CACHE)");
+		return (0);	/* success */
+	}
+	XQM_STATS_INC(xqmstats.xs_qm_dqcachemisses);
+
+	/*
+	 * Dquot cache miss. We don't want to keep the inode lock across
+	 * a (potential) disk read. Also we don't want to deal with the lock
+	 * ordering between quotainode and this inode. OTOH, dropping the inode
+	 * lock here means dealing with a chown that can happen before
+	 * we re-acquire the lock.
+	 */
+	if (ip)
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+	/*
+	 * Save the hashchain version stamp, and unlock the chain, so that
+	 * we don't keep the lock across a disk read
+	 */
+	version = h->qh_version;
+	XFS_DQ_HASH_UNLOCK(h);
+
+	/*
+	 * Allocate the dquot on the kernel heap, and read the ondisk
+	 * portion off the disk. Also, do all the necessary initialization
+	 * This can return ENOENT if dquot didn't exist on disk and we didn't
+	 * ask it to allocate; ESRCH if quotas got turned off suddenly.
+	 */
+	if ((error = xfs_qm_idtodq(mp, id, type,
+				  flags & (XFS_QMOPT_DQALLOC|XFS_QMOPT_DQREPAIR|
+					   XFS_QMOPT_DOWARN),
+				  &dqp))) {
+		if (ip)
+			xfs_ilock(ip, XFS_ILOCK_EXCL);
+		return (error);
+	}
+
+	/*
+	 * See if this is mount code calling to look at the overall quota limits
+	 * which are stored in the id == 0 user or group's dquot.
+	 * Since we may not have done a quotacheck by this point, just return
+	 * the dquot without attaching it to any hashtables, lists, etc, or even
+	 * taking a reference.
+	 * The caller must dqdestroy this once done.
+	 */
+	if (flags & XFS_QMOPT_DQSUSER) {
+		ASSERT(id == 0);
+		ASSERT(! ip);
+		goto dqret;
+	}
+
+	/*
+	 * Dquot lock comes after hashlock in the lock ordering
+	 */
+	if (ip) {
+		xfs_ilock(ip, XFS_ILOCK_EXCL);
+		if (! XFS_IS_DQTYPE_ON(mp, type)) {
+			/* inode stays locked on return */
+			xfs_qm_dqdestroy(dqp);
+			return XFS_ERROR(ESRCH);
+		}
+		/*
+		 * A dquot could be attached to this inode by now, since
+		 * we had dropped the ilock.
+		 */
+		if (type == XFS_DQ_USER) {
+			if (ip->i_udquot) {
+				xfs_qm_dqdestroy(dqp);
+				dqp = ip->i_udquot;
+				xfs_dqlock(dqp);
+				goto dqret;
+			}
+		} else {
+			if (ip->i_gdquot) {
+				xfs_qm_dqdestroy(dqp);
+				dqp = ip->i_gdquot;
+				xfs_dqlock(dqp);
+				goto dqret;
+			}
+		}
+	}
+
+	/*
+	 * Hashlock comes after ilock in lock order
+	 */
+	XFS_DQ_HASH_LOCK(h);
+	if (version != h->qh_version) {
+		xfs_dquot_t *tmpdqp;
+		/*
+		 * Now, see if somebody else put the dquot in the
+		 * hashtable before us. This can happen because we didn't
+		 * keep the hashchain lock. We don't have to worry about
+		 * lock order between the two dquots here since dqp isn't
+		 * on any findable lists yet.
+		 */
+		if (xfs_qm_dqlookup(mp, id, h, &tmpdqp) == 0) {
+			/*
+			 * Duplicate found. Just throw away the new dquot
+			 * and start over.
+			 */
+			xfs_qm_dqput(tmpdqp);
+			XFS_DQ_HASH_UNLOCK(h);
+			xfs_qm_dqdestroy(dqp);
+			XQM_STATS_INC(xqmstats.xs_qm_dquot_dups);
+			goto again;
+		}
+	}
+
+	/*
+	 * Put the dquot at the beginning of the hash-chain and mp's list
+	 * LOCK ORDER: hashlock, freelistlock, mplistlock, udqlock, gdqlock ..
+	 */
+	ASSERT(XFS_DQ_IS_HASH_LOCKED(h));
+	dqp->q_hash = h;
+	XQM_HASHLIST_INSERT(h, dqp);
+
+	/*
+	 * Attach this dquot to this filesystem's list of all dquots,
+	 * kept inside the mount structure in m_quotainfo field
+	 */
+	xfs_qm_mplist_lock(mp);
+
+	/*
+	 * We return a locked dquot to the caller, with a reference taken
+	 */
+	xfs_dqlock(dqp);
+	dqp->q_nrefs = 1;
+
+	XQM_MPLIST_INSERT(&(XFS_QI_MPL_LIST(mp)), dqp);
+
+	xfs_qm_mplist_unlock(mp);
+	XFS_DQ_HASH_UNLOCK(h);
+ dqret:
+	ASSERT((ip == NULL) || XFS_ISLOCKED_INODE_EXCL(ip));
+	xfs_dqtrace_entry(dqp, "DQGET DONE");
+	*O_dqpp = dqp;
+	return (0);
+}
+
+
+/*
+ * Release a reference to the dquot (decrement ref-count)
+ * and unlock it. If there is a group quota attached to this
+ * dquot, carefully release that too without tripping over
+ * deadlocks'n'stuff.
+ */
+void
+xfs_qm_dqput(
+	xfs_dquot_t	*dqp)
+{
+	xfs_dquot_t	*gdqp;
+
+	ASSERT(dqp->q_nrefs > 0);
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	xfs_dqtrace_entry(dqp, "DQPUT");
+
+	if (dqp->q_nrefs != 1) {
+		dqp->q_nrefs--;
+		xfs_dqunlock(dqp);
+		return;
+	}
+
+	/*
+	 * drop the dqlock and acquire the freelist and dqlock
+	 * in the right order; but try to get it out-of-order first
+	 */
+	if (! xfs_qm_freelist_lock_nowait(xfs_Gqm)) {
+		xfs_dqtrace_entry(dqp, "DQPUT: FLLOCK-WAIT");
+		xfs_dqunlock(dqp);
+		xfs_qm_freelist_lock(xfs_Gqm);
+		xfs_dqlock(dqp);
+	}
+
+	while (1) {
+		gdqp = NULL;
+
+		/* We can't depend on nrefs being == 1 here */
+		if (--dqp->q_nrefs == 0) {
+			xfs_dqtrace_entry(dqp, "DQPUT: ON FREELIST");
+			/*
+			 * insert at end of the freelist.
+			 */
+			XQM_FREELIST_INSERT(&(xfs_Gqm->qm_dqfreelist), dqp);
+
+			/*
+			 * If we just added a udquot to the freelist, then
+			 * we want to release the gdquot reference that
+			 * it (probably) has. Otherwise it'll keep the
+			 * gdquot from getting reclaimed.
+			 */
+			if ((gdqp = dqp->q_gdquot)) {
+				/*
+				 * Avoid a recursive dqput call
+				 */
+				xfs_dqlock(gdqp);
+				dqp->q_gdquot = NULL;
+			}
+
+			/* xfs_qm_freelist_print(&(xfs_Gqm->qm_dqfreelist),
+			   "@@@@@++ Free list (after append) @@@@@+");
+			   */
+		}
+		xfs_dqunlock(dqp);
+
+		/*
+		 * If we had a group quota inside the user quota as a hint,
+		 * release it now.
+		 */
+		if (! gdqp)
+			break;
+		dqp = gdqp;
+	}
+	xfs_qm_freelist_unlock(xfs_Gqm);
+}
+
+/*
+ * Release a dquot. Flush it if dirty, then dqput() it.
+ * dquot must not be locked.
+ */
+void
+xfs_qm_dqrele(
+	xfs_dquot_t	*dqp)
+{
+	ASSERT(dqp);
+	xfs_dqtrace_entry(dqp, "DQRELE");
+
+	xfs_dqlock(dqp);
+	/*
+	 * We don't care to flush it if the dquot is dirty here.
+	 * That will create stutters that we want to avoid.
+	 * Instead we do a delayed write when we try to reclaim
+	 * a dirty dquot. Also xfs_sync will take part of the burden...
+	 */
+	xfs_qm_dqput(dqp);
+}
+
+
+/*
+ * Write a modified dquot to disk.
+ * The dquot must be locked and the flush lock too taken by caller.
+ * The flush lock will not be unlocked until the dquot reaches the disk,
+ * but the dquot is free to be unlocked and modified by the caller
+ * in the interim. Dquot is still locked on return. This behavior is
+ * identical to that of inodes.
+ */
+int
+xfs_qm_dqflush(
+	xfs_dquot_t		*dqp,
+	uint			flags)
+{
+	xfs_mount_t		*mp;
+	xfs_buf_t		*bp;
+	xfs_disk_dquot_t	*ddqp;
+	int			error;
+	SPLDECL(s);
+
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	ASSERT(XFS_DQ_IS_FLUSH_LOCKED(dqp));
+	xfs_dqtrace_entry(dqp, "DQFLUSH");
+
+	/*
+	 * If not dirty, nada.
+	 */
+	if (!XFS_DQ_IS_DIRTY(dqp)) {
+		xfs_dqfunlock(dqp);
+		return (0);
+	}
+
+	/*
+	 * Cant flush a pinned dquot. Wait for it.
+	 */
+	xfs_qm_dqunpin_wait(dqp);
+
+	/*
+	 * This may have been unpinned because the filesystem is shutting
+	 * down forcibly. If that's the case we must not write this dquot
+	 * to disk, because the log record didn't make it to disk!
+	 */
+	if (XFS_FORCED_SHUTDOWN(dqp->q_mount)) {
+		dqp->dq_flags &= ~(XFS_DQ_DIRTY);
+		xfs_dqfunlock(dqp);
+		return XFS_ERROR(EIO);
+	}
+
+	/*
+	 * Get the buffer containing the on-disk dquot
+	 * We don't need a transaction envelope because we know that the
+	 * the ondisk-dquot has already been allocated for.
+	 */
+	if ((error = xfs_qm_dqtobp(NULL, dqp, &ddqp, &bp, XFS_QMOPT_DOWARN))) {
+		xfs_dqtrace_entry(dqp, "DQTOBP FAIL");
+		ASSERT(error != ENOENT);
+		/*
+		 * Quotas could have gotten turned off (ESRCH)
+		 */
+		xfs_dqfunlock(dqp);
+		return (error);
+	}
+
+	if (xfs_qm_dqcheck(&dqp->q_core, INT_GET(ddqp->d_id, ARCH_CONVERT), 0, XFS_QMOPT_DOWARN,
+			   "dqflush (incore copy)")) {
+		xfs_force_shutdown(dqp->q_mount, XFS_CORRUPT_INCORE);
+		return XFS_ERROR(EIO);
+	}
+
+	/* This is the only portion of data that needs to persist */
+	memcpy(ddqp, &(dqp->q_core), sizeof(xfs_disk_dquot_t));
+
+	/*
+	 * Clear the dirty field and remember the flush lsn for later use.
+	 */
+	dqp->dq_flags &= ~(XFS_DQ_DIRTY);
+	mp = dqp->q_mount;
+
+	/* lsn is 64 bits */
+	AIL_LOCK(mp, s);
+	dqp->q_logitem.qli_flush_lsn = dqp->q_logitem.qli_item.li_lsn;
+	AIL_UNLOCK(mp, s);
+
+	/*
+	 * Attach an iodone routine so that we can remove this dquot from the
+	 * AIL and release the flush lock once the dquot is synced to disk.
+	 */
+	xfs_buf_attach_iodone(bp, (void(*)(xfs_buf_t *, xfs_log_item_t *))
+			      xfs_qm_dqflush_done, &(dqp->q_logitem.qli_item));
+	/*
+	 * If the buffer is pinned then push on the log so we won't
+	 * get stuck waiting in the write for too long.
+	 */
+	if (XFS_BUF_ISPINNED(bp)) {
+		xfs_dqtrace_entry(dqp, "DQFLUSH LOG FORCE");
+		xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
+	}
+
+	if (flags & XFS_QMOPT_DELWRI) {
+		xfs_bdwrite(mp, bp);
+	} else if (flags & XFS_QMOPT_ASYNC) {
+		xfs_bawrite(mp, bp);
+	} else {
+		error = xfs_bwrite(mp, bp);
+	}
+	xfs_dqtrace_entry(dqp, "DQFLUSH END");
+	/*
+	 * dqp is still locked, but caller is free to unlock it now.
+	 */
+	return (error);
+
+}
+
+/*
+ * This is the dquot flushing I/O completion routine.  It is called
+ * from interrupt level when the buffer containing the dquot is
+ * flushed to disk.  It is responsible for removing the dquot logitem
+ * from the AIL if it has not been re-logged, and unlocking the dquot's
+ * flush lock. This behavior is very similar to that of inodes..
+ */
+/*ARGSUSED*/
+STATIC void
+xfs_qm_dqflush_done(
+	xfs_buf_t		*bp,
+	xfs_dq_logitem_t	*qip)
+{
+	xfs_dquot_t		*dqp;
+	SPLDECL(s);
+
+	dqp = qip->qli_dquot;
+
+	/*
+	 * We only want to pull the item from the AIL if its
+	 * location in the log has not changed since we started the flush.
+	 * Thus, we only bother if the dquot's lsn has
+	 * not changed. First we check the lsn outside the lock
+	 * since it's cheaper, and then we recheck while
+	 * holding the lock before removing the dquot from the AIL.
+	 */
+	if ((qip->qli_item.li_flags & XFS_LI_IN_AIL) &&
+	    qip->qli_item.li_lsn == qip->qli_flush_lsn) {
+
+		AIL_LOCK(dqp->q_mount, s);
+		/*
+		 * xfs_trans_delete_ail() drops the AIL lock.
+		 */
+		if (qip->qli_item.li_lsn == qip->qli_flush_lsn)
+			xfs_trans_delete_ail(dqp->q_mount,
+					     (xfs_log_item_t*)qip, s);
+		else
+			AIL_UNLOCK(dqp->q_mount, s);
+	}
+
+	/*
+	 * Release the dq's flush lock since we're done with it.
+	 */
+	xfs_dqfunlock(dqp);
+}
+
+
+int
+xfs_qm_dqflock_nowait(
+	xfs_dquot_t *dqp)
+{
+	int locked;
+
+	locked = cpsema(&((dqp)->q_flock));
+
+	/* XXX ifdef these out */
+	if (locked)
+		(dqp)->dq_flags |= XFS_DQ_FLOCKED;
+	return (locked);
+}
+
+
+int
+xfs_qm_dqlock_nowait(
+	xfs_dquot_t *dqp)
+{
+	return (mutex_trylock(&((dqp)->q_qlock)));
+}
+
+void
+xfs_dqlock(
+	xfs_dquot_t *dqp)
+{
+	mutex_lock(&(dqp->q_qlock), PINOD);
+}
+
+void
+xfs_dqunlock(
+	xfs_dquot_t *dqp)
+{
+	mutex_unlock(&(dqp->q_qlock));
+	if (dqp->q_logitem.qli_dquot == dqp) {
+		/* Once was dqp->q_mount, but might just have been cleared */
+		xfs_trans_unlocked_item(dqp->q_logitem.qli_item.li_mountp,
+					(xfs_log_item_t*)&(dqp->q_logitem));
+	}
+}
+
+
+void
+xfs_dqunlock_nonotify(
+	xfs_dquot_t *dqp)
+{
+	mutex_unlock(&(dqp->q_qlock));
+}
+
+void
+xfs_dqlock2(
+	xfs_dquot_t	*d1,
+	xfs_dquot_t	*d2)
+{
+	if (d1 && d2) {
+		ASSERT(d1 != d2);
+		if (INT_GET(d1->q_core.d_id, ARCH_CONVERT) > INT_GET(d2->q_core.d_id, ARCH_CONVERT)) {
+			xfs_dqlock(d2);
+			xfs_dqlock(d1);
+		} else {
+			xfs_dqlock(d1);
+			xfs_dqlock(d2);
+		}
+	} else {
+		if (d1) {
+			xfs_dqlock(d1);
+		} else if (d2) {
+			xfs_dqlock(d2);
+		}
+	}
+}
+
+
+/*
+ * Take a dquot out of the mount's dqlist as well as the hashlist.
+ * This is called via unmount as well as quotaoff, and the purge
+ * will always succeed unless there are soft (temp) references
+ * outstanding.
+ *
+ * This returns 0 if it was purged, 1 if it wasn't. It's not an error code
+ * that we're returning! XXXsup - not cool.
+ */
+/* ARGSUSED */
+int
+xfs_qm_dqpurge(
+	xfs_dquot_t	*dqp,
+	uint		flags)
+{
+	xfs_dqhash_t	*thishash;
+	xfs_mount_t	*mp;
+
+	mp = dqp->q_mount;
+
+	ASSERT(XFS_QM_IS_MPLIST_LOCKED(mp));
+	ASSERT(XFS_DQ_IS_HASH_LOCKED(dqp->q_hash));
+
+	xfs_dqlock(dqp);
+	/*
+	 * We really can't afford to purge a dquot that is
+	 * referenced, because these are hard refs.
+	 * It shouldn't happen in general because we went thru _all_ inodes in
+	 * dqrele_all_inodes before calling this and didn't let the mountlock go.
+	 * However it is possible that we have dquots with temporary
+	 * references that are not attached to an inode. e.g. see xfs_setattr().
+	 */
+	if (dqp->q_nrefs != 0) {
+		xfs_dqunlock(dqp);
+		XFS_DQ_HASH_UNLOCK(dqp->q_hash);
+		return (1);
+	}
+
+	ASSERT(XFS_DQ_IS_ON_FREELIST(dqp));
+
+	/*
+	 * If we're turning off quotas, we have to make sure that, for
+	 * example, we don't delete quota disk blocks while dquots are
+	 * in the process of getting written to those disk blocks.
+	 * This dquot might well be on AIL, and we can't leave it there
+	 * if we're turning off quotas. Basically, we need this flush
+	 * lock, and are willing to block on it.
+	 */
+	if (! xfs_qm_dqflock_nowait(dqp)) {
+		/*
+		 * Block on the flush lock after nudging dquot buffer,
+		 * if it is incore.
+		 */
+		xfs_qm_dqflock_pushbuf_wait(dqp);
+	}
+
+	/*
+	 * XXXIf we're turning this type of quotas off, we don't care
+	 * about the dirty metadata sitting in this dquot. OTOH, if
+	 * we're unmounting, we do care, so we flush it and wait.
+	 */
+	if (XFS_DQ_IS_DIRTY(dqp)) {
+		xfs_dqtrace_entry(dqp, "DQPURGE ->DQFLUSH: DQDIRTY");
+		/* dqflush unlocks dqflock */
+		/*
+		 * Given that dqpurge is a very rare occurrence, it is OK
+		 * that we're holding the hashlist and mplist locks
+		 * across the disk write. But, ... XXXsup
+		 *
+		 * We don't care about getting disk errors here. We need
+		 * to purge this dquot anyway, so we go ahead regardless.
+		 */
+		(void) xfs_qm_dqflush(dqp, XFS_QMOPT_SYNC);
+		xfs_dqflock(dqp);
+	}
+	ASSERT(dqp->q_pincount == 0);
+	ASSERT(XFS_FORCED_SHUTDOWN(mp) ||
+	       !(dqp->q_logitem.qli_item.li_flags & XFS_LI_IN_AIL));
+
+	thishash = dqp->q_hash;
+	XQM_HASHLIST_REMOVE(thishash, dqp);
+	XQM_MPLIST_REMOVE(&(XFS_QI_MPL_LIST(mp)), dqp);
+	/*
+	 * XXX Move this to the front of the freelist, if we can get the
+	 * freelist lock.
+	 */
+	ASSERT(XFS_DQ_IS_ON_FREELIST(dqp));
+
+	dqp->q_mount = NULL;
+	dqp->q_hash = NULL;
+	dqp->dq_flags = XFS_DQ_INACTIVE;
+	memset(&dqp->q_core, 0, sizeof(dqp->q_core));
+	xfs_dqfunlock(dqp);
+	xfs_dqunlock(dqp);
+	XFS_DQ_HASH_UNLOCK(thishash);
+	return (0);
+}
+
+
+#ifdef QUOTADEBUG
+void
+xfs_qm_dqprint(xfs_dquot_t *dqp)
+{
+	cmn_err(CE_DEBUG, "-----------KERNEL DQUOT----------------");
+	cmn_err(CE_DEBUG, "---- dquotID =  %d",
+		(int)INT_GET(dqp->q_core.d_id, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- type    =  %s",
+		XFS_QM_ISUDQ(dqp) ? "USR" : "GRP");
+	cmn_err(CE_DEBUG, "---- fs      =  0x%p", dqp->q_mount);
+	cmn_err(CE_DEBUG, "---- blkno   =  0x%x", (int) dqp->q_blkno);
+	cmn_err(CE_DEBUG, "---- boffset =  0x%x", (int) dqp->q_bufoffset);
+	cmn_err(CE_DEBUG, "---- blkhlimit =  %Lu (0x%x)",
+		INT_GET(dqp->q_core.d_blk_hardlimit, ARCH_CONVERT),
+		(int) INT_GET(dqp->q_core.d_blk_hardlimit, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- blkslimit =  %Lu (0x%x)",
+		INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT),
+		(int)INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- inohlimit =  %Lu (0x%x)",
+		INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT),
+		(int)INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- inoslimit =  %Lu (0x%x)",
+		INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT),
+		(int)INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- bcount  =  %Lu (0x%x)",
+		INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT),
+		(int)INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- icount  =  %Lu (0x%x)",
+		INT_GET(dqp->q_core.d_icount, ARCH_CONVERT),
+		(int)INT_GET(dqp->q_core.d_icount, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- btimer  =  %d",
+		(int)INT_GET(dqp->q_core.d_btimer, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---- itimer  =  %d",
+		(int)INT_GET(dqp->q_core.d_itimer, ARCH_CONVERT));
+	cmn_err(CE_DEBUG, "---------------------------");
+}
+#endif
+
+/*
+ * Give the buffer a little push if it is incore and
+ * wait on the flush lock.
+ */
+void
+xfs_qm_dqflock_pushbuf_wait(
+	xfs_dquot_t	*dqp)
+{
+	xfs_buf_t	*bp;
+
+	/*
+	 * Check to see if the dquot has been flushed delayed
+	 * write.  If so, grab its buffer and send it
+	 * out immediately.  We'll be able to acquire
+	 * the flush lock when the I/O completes.
+	 */
+	bp = xfs_incore(dqp->q_mount->m_ddev_targp, dqp->q_blkno,
+		    XFS_QI_DQCHUNKLEN(dqp->q_mount),
+		    XFS_INCORE_TRYLOCK);
+	if (bp != NULL) {
+		if (XFS_BUF_ISDELAYWRITE(bp)) {
+			if (XFS_BUF_ISPINNED(bp)) {
+				xfs_log_force(dqp->q_mount,
+					      (xfs_lsn_t)0,
+					      XFS_LOG_FORCE);
+			}
+			xfs_bawrite(dqp->q_mount, bp);
+		} else {
+			xfs_buf_relse(bp);
+		}
+	}
+	xfs_dqflock(dqp);
+}
diff --git a/fs/xfs/quota/xfs_dquot.h b/fs/xfs/quota/xfs_dquot.h
new file mode 100644
index 000000000000..0c3fe3175baa
--- /dev/null
+++ b/fs/xfs/quota/xfs_dquot.h
@@ -0,0 +1,224 @@
+/*
+ * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+#ifndef __XFS_DQUOT_H__
+#define __XFS_DQUOT_H__
+
+/*
+ * Dquots are structures that hold quota information about a user or a group,
+ * much like inodes are for files. In fact, dquots share many characteristics
+ * with inodes. However, dquots can also be a centralized resource, relative
+ * to a collection of inodes. In this respect, dquots share some characteristics
+ * of the superblock.
+ * XFS dquots exploit both those in its algorithms. They make every attempt
+ * to not be a bottleneck when quotas are on and have minimal impact, if any,
+ * when quotas are off.
+ */
+
+/*
+ * The hash chain headers (hash buckets)
+ */
+typedef struct xfs_dqhash {
+	struct xfs_dquot *qh_next;
+	mutex_t		  qh_lock;
+	uint		  qh_version;	/* ever increasing version */
+	uint		  qh_nelems;	/* number of dquots on the list */
+} xfs_dqhash_t;
+
+typedef struct xfs_dqlink {
+	struct xfs_dquot  *ql_next;	/* forward link */
+	struct xfs_dquot **ql_prevp;	/* pointer to prev ql_next */
+} xfs_dqlink_t;
+
+struct xfs_mount;
+struct xfs_trans;
+
+/*
+ * This is the marker which is designed to occupy the first few
+ * bytes of the xfs_dquot_t structure. Even inside this, the freelist pointers
+ * must come first.
+ * This serves as the marker ("sentinel") when we have to restart list
+ * iterations because of locking considerations.
+ */
+typedef struct xfs_dqmarker {
+	struct xfs_dquot*dqm_flnext;	/* link to freelist: must be first */
+	struct xfs_dquot*dqm_flprev;
+	xfs_dqlink_t	 dqm_mplist;	/* link to mount's list of dquots */
+	xfs_dqlink_t	 dqm_hashlist;	/* link to the hash chain */
+	uint		 dqm_flags;	/* various flags (XFS_DQ_*) */
+} xfs_dqmarker_t;
+
+/*
+ * The incore dquot structure
+ */
+typedef struct xfs_dquot {
+	xfs_dqmarker_t	 q_lists;	/* list ptrs, q_flags (marker) */
+	xfs_dqhash_t	*q_hash;	/* the hashchain header */
+	struct xfs_mount*q_mount;	/* filesystem this relates to */
+	struct xfs_trans*q_transp;	/* trans this belongs to currently */
+	uint		 q_nrefs;	/* # active refs from inodes */
+	xfs_daddr_t	 q_blkno;	/* blkno of dquot buffer */
+	int		 q_bufoffset;	/* off of dq in buffer (# dquots) */
+	xfs_fileoff_t	 q_fileoffset;	/* offset in quotas file */
+
+	struct xfs_dquot*q_gdquot;	/* group dquot, hint only */
+	xfs_disk_dquot_t q_core;	/* actual usage & quotas */
+	xfs_dq_logitem_t q_logitem;	/* dquot log item */
+	xfs_qcnt_t	 q_res_bcount;	/* total regular nblks used+reserved */
+	xfs_qcnt_t	 q_res_icount;	/* total inos allocd+reserved */
+	xfs_qcnt_t	 q_res_rtbcount;/* total realtime blks used+reserved */
+	mutex_t		 q_qlock;	/* quota lock */
+	sema_t		 q_flock;	/* flush lock */
+	uint		 q_pincount;	/* pin count for this dquot */
+	sv_t		 q_pinwait;	/* sync var for pinning */
+#ifdef XFS_DQUOT_TRACE
+	struct ktrace	*q_trace;	/* trace header structure */
+#endif
+} xfs_dquot_t;
+
+
+#define dq_flnext	q_lists.dqm_flnext
+#define dq_flprev	q_lists.dqm_flprev
+#define dq_mplist	q_lists.dqm_mplist
+#define dq_hashlist	q_lists.dqm_hashlist
+#define dq_flags	q_lists.dqm_flags
+
+#define XFS_DQHOLD(dqp)		((dqp)->q_nrefs++)
+
+/*
+ * Quota Accounting flags
+ */
+#define XFS_ALL_QUOTA_ACCT	(XFS_UQUOTA_ACCT | XFS_GQUOTA_ACCT)
+#define XFS_ALL_QUOTA_ENFD	(XFS_UQUOTA_ENFD | XFS_GQUOTA_ENFD)
+#define XFS_ALL_QUOTA_CHKD	(XFS_UQUOTA_CHKD | XFS_GQUOTA_CHKD)
+#define XFS_ALL_QUOTA_ACTV	(XFS_UQUOTA_ACTIVE | XFS_GQUOTA_ACTIVE)
+#define XFS_ALL_QUOTA_ACCT_ENFD (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD|\
+				 XFS_GQUOTA_ACCT|XFS_GQUOTA_ENFD)
+
+#define XFS_IS_QUOTA_RUNNING(mp)  ((mp)->m_qflags & XFS_ALL_QUOTA_ACCT)
+#define XFS_IS_UQUOTA_RUNNING(mp) ((mp)->m_qflags & XFS_UQUOTA_ACCT)
+#define XFS_IS_GQUOTA_RUNNING(mp) ((mp)->m_qflags & XFS_GQUOTA_ACCT)
+
+/*
+ * Quota Limit Enforcement flags
+ */
+#define XFS_IS_QUOTA_ENFORCED(mp)	((mp)->m_qflags & XFS_ALL_QUOTA_ENFD)
+#define XFS_IS_UQUOTA_ENFORCED(mp)	((mp)->m_qflags & XFS_UQUOTA_ENFD)
+#define XFS_IS_GQUOTA_ENFORCED(mp)	((mp)->m_qflags & XFS_GQUOTA_ENFD)
+
+#ifdef DEBUG
+static inline int
+XFS_DQ_IS_LOCKED(xfs_dquot_t *dqp)
+{
+	if (mutex_trylock(&dqp->q_qlock)) {
+		mutex_unlock(&dqp->q_qlock);
+		return 0;
+	}
+	return 1;
+}
+#endif
+
+
+/*
+ * The following three routines simply manage the q_flock
+ * semaphore embedded in the dquot.  This semaphore synchronizes
+ * processes attempting to flush the in-core dquot back to disk.
+ */
+#define xfs_dqflock(dqp)	 { psema(&((dqp)->q_flock), PINOD | PRECALC);\
+				   (dqp)->dq_flags |= XFS_DQ_FLOCKED; }
+#define xfs_dqfunlock(dqp)	 { ASSERT(valusema(&((dqp)->q_flock)) <= 0); \
+				   vsema(&((dqp)->q_flock)); \
+				   (dqp)->dq_flags &= ~(XFS_DQ_FLOCKED); }
+
+#define XFS_DQ_PINLOCK(dqp)	   mutex_spinlock( \
+				     &(XFS_DQ_TO_QINF(dqp)->qi_pinlock))
+#define XFS_DQ_PINUNLOCK(dqp, s)   mutex_spinunlock( \
+				     &(XFS_DQ_TO_QINF(dqp)->qi_pinlock), s)
+
+#define XFS_DQ_IS_FLUSH_LOCKED(dqp) (valusema(&((dqp)->q_flock)) <= 0)
+#define XFS_DQ_IS_ON_FREELIST(dqp)  ((dqp)->dq_flnext != (dqp))
+#define XFS_DQ_IS_DIRTY(dqp)	((dqp)->dq_flags & XFS_DQ_DIRTY)
+#define XFS_QM_ISUDQ(dqp)	((dqp)->dq_flags & XFS_DQ_USER)
+#define XFS_DQ_TO_QINF(dqp)	((dqp)->q_mount->m_quotainfo)
+#define XFS_DQ_TO_QIP(dqp)	(XFS_QM_ISUDQ(dqp) ? \
+				 XFS_DQ_TO_QINF(dqp)->qi_uquotaip : \
+				 XFS_DQ_TO_QINF(dqp)->qi_gquotaip)
+
+#define XFS_IS_THIS_QUOTA_OFF(d) (! (XFS_QM_ISUDQ(d) ? \
+				     (XFS_IS_UQUOTA_ON((d)->q_mount)) : \
+				     (XFS_IS_GQUOTA_ON((d)->q_mount))))
+
+#ifdef XFS_DQUOT_TRACE
+/*
+ * Dquot Tracing stuff.
+ */
+#define DQUOT_TRACE_SIZE	64
+#define DQUOT_KTRACE_ENTRY	1
+
+extern void		__xfs_dqtrace_entry(xfs_dquot_t *dqp, char *func,
+					    void *, xfs_inode_t *);
+#define xfs_dqtrace_entry_ino(a,b,ip) \
+		__xfs_dqtrace_entry((a), (b), (void*)__return_address, (ip))
+#define xfs_dqtrace_entry(a,b) \
+		__xfs_dqtrace_entry((a), (b), (void*)__return_address, NULL)
+#else
+#define xfs_dqtrace_entry(a,b)
+#define xfs_dqtrace_entry_ino(a,b,ip)
+#endif
+
+#ifdef QUOTADEBUG
+extern void		xfs_qm_dqprint(xfs_dquot_t *);
+#else
+#define xfs_qm_dqprint(a)
+#endif
+
+extern void		xfs_qm_dqdestroy(xfs_dquot_t *);
+extern int		xfs_qm_dqflush(xfs_dquot_t *, uint);
+extern int		xfs_qm_dqpurge(xfs_dquot_t *, uint);
+extern void		xfs_qm_dqunpin_wait(xfs_dquot_t *);
+extern int		xfs_qm_dqlock_nowait(xfs_dquot_t *);
+extern int		xfs_qm_dqflock_nowait(xfs_dquot_t *);
+extern void		xfs_qm_dqflock_pushbuf_wait(xfs_dquot_t *dqp);
+extern void		xfs_qm_adjust_dqtimers(xfs_mount_t *,
+					xfs_disk_dquot_t *);
+extern void		xfs_qm_adjust_dqlimits(xfs_mount_t *,
+					xfs_disk_dquot_t *);
+extern int		xfs_qm_dqwarn(xfs_disk_dquot_t *, uint);
+extern int		xfs_qm_dqget(xfs_mount_t *, xfs_inode_t *,
+					xfs_dqid_t, uint, uint, xfs_dquot_t **);
+extern void		xfs_qm_dqput(xfs_dquot_t *);
+extern void		xfs_qm_dqrele(xfs_dquot_t *);
+extern void		xfs_dqlock(xfs_dquot_t *);
+extern void		xfs_dqlock2(xfs_dquot_t *, xfs_dquot_t *);
+extern void		xfs_dqunlock(xfs_dquot_t *);
+extern void		xfs_dqunlock_nonotify(xfs_dquot_t *);
+
+#endif /* __XFS_DQUOT_H__ */
diff --git a/fs/xfs/quota/xfs_dquot_item.c b/fs/xfs/quota/xfs_dquot_item.c
new file mode 100644
index 000000000000..a5425ee6e7bd
--- /dev/null
+++ b/fs/xfs/quota/xfs_dquot_item.c
@@ -0,0 +1,715 @@
+/*
+ * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_ag.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+#include "xfs_trans_priv.h"
+
+#include "xfs_qm.h"
+
+
+/*
+ * returns the number of iovecs needed to log the given dquot item.
+ */
+/* ARGSUSED */
+STATIC uint
+xfs_qm_dquot_logitem_size(
+	xfs_dq_logitem_t	*logitem)
+{
+	/*
+	 * we need only two iovecs, one for the format, one for the real thing
+	 */
+	return (2);
+}
+
+/*
+ * fills in the vector of log iovecs for the given dquot log item.
+ */
+STATIC void
+xfs_qm_dquot_logitem_format(
+	xfs_dq_logitem_t	*logitem,
+	xfs_log_iovec_t		*logvec)
+{
+	ASSERT(logitem);
+	ASSERT(logitem->qli_dquot);
+
+	logvec->i_addr = (xfs_caddr_t)&logitem->qli_format;
+	logvec->i_len  = sizeof(xfs_dq_logformat_t);
+	logvec++;
+	logvec->i_addr = (xfs_caddr_t)&logitem->qli_dquot->q_core;
+	logvec->i_len  = sizeof(xfs_disk_dquot_t);
+
+	ASSERT(2 == logitem->qli_item.li_desc->lid_size);
+	logitem->qli_format.qlf_size = 2;
+
+}
+
+/*
+ * Increment the pin count of the given dquot.
+ * This value is protected by pinlock spinlock in the xQM structure.
+ */
+STATIC void
+xfs_qm_dquot_logitem_pin(
+	xfs_dq_logitem_t *logitem)
+{
+	unsigned long	s;
+	xfs_dquot_t *dqp;
+
+	dqp = logitem->qli_dquot;
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	s = XFS_DQ_PINLOCK(dqp);
+	dqp->q_pincount++;
+	XFS_DQ_PINUNLOCK(dqp, s);
+}
+
+/*
+ * Decrement the pin count of the given dquot, and wake up
+ * anyone in xfs_dqwait_unpin() if the count goes to 0.	 The
+ * dquot must have been previously pinned with a call to xfs_dqpin().
+ */
+/* ARGSUSED */
+STATIC void
+xfs_qm_dquot_logitem_unpin(
+	xfs_dq_logitem_t *logitem,
+	int		  stale)
+{
+	unsigned long	s;
+	xfs_dquot_t *dqp;
+
+	dqp = logitem->qli_dquot;
+	ASSERT(dqp->q_pincount > 0);
+	s = XFS_DQ_PINLOCK(dqp);
+	dqp->q_pincount--;
+	if (dqp->q_pincount == 0) {
+		sv_broadcast(&dqp->q_pinwait);
+	}
+	XFS_DQ_PINUNLOCK(dqp, s);
+}
+
+/* ARGSUSED */
+STATIC void
+xfs_qm_dquot_logitem_unpin_remove(
+	xfs_dq_logitem_t *logitem,
+	xfs_trans_t	 *tp)
+{
+	xfs_qm_dquot_logitem_unpin(logitem, 0);
+}
+
+/*
+ * Given the logitem, this writes the corresponding dquot entry to disk
+ * asynchronously. This is called with the dquot entry securely locked;
+ * we simply get xfs_qm_dqflush() to do the work, and unlock the dquot
+ * at the end.
+ */
+STATIC void
+xfs_qm_dquot_logitem_push(
+	xfs_dq_logitem_t	*logitem)
+{
+	xfs_dquot_t	*dqp;
+
+	dqp = logitem->qli_dquot;
+
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	ASSERT(XFS_DQ_IS_FLUSH_LOCKED(dqp));
+
+	/*
+	 * Since we were able to lock the dquot's flush lock and
+	 * we found it on the AIL, the dquot must be dirty.  This
+	 * is because the dquot is removed from the AIL while still
+	 * holding the flush lock in xfs_dqflush_done().  Thus, if
+	 * we found it in the AIL and were able to obtain the flush
+	 * lock without sleeping, then there must not have been
+	 * anyone in the process of flushing the dquot.
+	 */
+	xfs_qm_dqflush(dqp, XFS_B_DELWRI);
+	xfs_dqunlock(dqp);
+}
+
+/*ARGSUSED*/
+STATIC xfs_lsn_t
+xfs_qm_dquot_logitem_committed(
+	xfs_dq_logitem_t	*l,
+	xfs_lsn_t		lsn)
+{
+	/*
+	 * We always re-log the entire dquot when it becomes dirty,
+	 * so, the latest copy _is_ the only one that matters.
+	 */
+	return (lsn);
+}
+
+
+/*
+ * This is called to wait for the given dquot to be unpinned.
+ * Most of these pin/unpin routines are plagiarized from inode code.
+ */
+void
+xfs_qm_dqunpin_wait(
+	xfs_dquot_t	*dqp)
+{
+	SPLDECL(s);
+
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	if (dqp->q_pincount == 0) {
+		return;
+	}
+
+	/*
+	 * Give the log a push so we don't wait here too long.
+	 */
+	xfs_log_force(dqp->q_mount, (xfs_lsn_t)0, XFS_LOG_FORCE);
+	s = XFS_DQ_PINLOCK(dqp);
+	if (dqp->q_pincount == 0) {
+		XFS_DQ_PINUNLOCK(dqp, s);
+		return;
+	}
+	sv_wait(&(dqp->q_pinwait), PINOD,
+		&(XFS_DQ_TO_QINF(dqp)->qi_pinlock), s);
+}
+
+/*
+ * This is called when IOP_TRYLOCK returns XFS_ITEM_PUSHBUF to indicate that
+ * the dquot is locked by us, but the flush lock isn't. So, here we are
+ * going to see if the relevant dquot buffer is incore, waiting on DELWRI.
+ * If so, we want to push it out to help us take this item off the AIL as soon
+ * as possible.
+ *
+ * We must not be holding the AIL_LOCK at this point. Calling incore() to
+ * search the buffercache can be a time consuming thing, and AIL_LOCK is a
+ * spinlock.
+ */
+STATIC void
+xfs_qm_dquot_logitem_pushbuf(
+	xfs_dq_logitem_t    *qip)
+{
+	xfs_dquot_t	*dqp;
+	xfs_mount_t	*mp;
+	xfs_buf_t	*bp;
+	uint		dopush;
+
+	dqp = qip->qli_dquot;
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+
+	/*
+	 * The qli_pushbuf_flag keeps others from
+	 * trying to duplicate our effort.
+	 */
+	ASSERT(qip->qli_pushbuf_flag != 0);
+	ASSERT(qip->qli_push_owner == get_thread_id());
+
+	/*
+	 * If flushlock isn't locked anymore, chances are that the
+	 * inode flush completed and the inode was taken off the AIL.
+	 * So, just get out.
+	 */
+	if ((valusema(&(dqp->q_flock)) > 0)  ||
+	    ((qip->qli_item.li_flags & XFS_LI_IN_AIL) == 0)) {
+		qip->qli_pushbuf_flag = 0;
+		xfs_dqunlock(dqp);
+		return;
+	}
+	mp = dqp->q_mount;
+	bp = xfs_incore(mp->m_ddev_targp, qip->qli_format.qlf_blkno,
+		    XFS_QI_DQCHUNKLEN(mp),
+		    XFS_INCORE_TRYLOCK);
+	if (bp != NULL) {
+		if (XFS_BUF_ISDELAYWRITE(bp)) {
+			dopush = ((qip->qli_item.li_flags & XFS_LI_IN_AIL) &&
+				  (valusema(&(dqp->q_flock)) <= 0));
+			qip->qli_pushbuf_flag = 0;
+			xfs_dqunlock(dqp);
+
+			if (XFS_BUF_ISPINNED(bp)) {
+				xfs_log_force(mp, (xfs_lsn_t)0,
+					      XFS_LOG_FORCE);
+			}
+			if (dopush) {
+#ifdef XFSRACEDEBUG
+				delay_for_intr();
+				delay(300);
+#endif
+				xfs_bawrite(mp, bp);
+			} else {
+				xfs_buf_relse(bp);
+			}
+		} else {
+			qip->qli_pushbuf_flag = 0;
+			xfs_dqunlock(dqp);
+			xfs_buf_relse(bp);
+		}
+		return;
+	}
+
+	qip->qli_pushbuf_flag = 0;
+	xfs_dqunlock(dqp);
+}
+
+/*
+ * This is called to attempt to lock the dquot associated with this
+ * dquot log item.  Don't sleep on the dquot lock or the flush lock.
+ * If the flush lock is already held, indicating that the dquot has
+ * been or is in the process of being flushed, then see if we can
+ * find the dquot's buffer in the buffer cache without sleeping.  If
+ * we can and it is marked delayed write, then we want to send it out.
+ * We delay doing so until the push routine, though, to avoid sleeping
+ * in any device strategy routines.
+ */
+STATIC uint
+xfs_qm_dquot_logitem_trylock(
+	xfs_dq_logitem_t	*qip)
+{
+	xfs_dquot_t		*dqp;
+	uint			retval;
+
+	dqp = qip->qli_dquot;
+	if (dqp->q_pincount > 0)
+		return (XFS_ITEM_PINNED);
+
+	if (! xfs_qm_dqlock_nowait(dqp))
+		return (XFS_ITEM_LOCKED);
+
+	retval = XFS_ITEM_SUCCESS;
+	if (! xfs_qm_dqflock_nowait(dqp)) {
+		/*
+		 * The dquot is already being flushed.	It may have been
+		 * flushed delayed write, however, and we don't want to
+		 * get stuck waiting for that to complete.  So, we want to check
+		 * to see if we can lock the dquot's buffer without sleeping.
+		 * If we can and it is marked for delayed write, then we
+		 * hold it and send it out from the push routine.  We don't
+		 * want to do that now since we might sleep in the device
+		 * strategy routine.  We also don't want to grab the buffer lock
+		 * here because we'd like not to call into the buffer cache
+		 * while holding the AIL_LOCK.
+		 * Make sure to only return PUSHBUF if we set pushbuf_flag
+		 * ourselves.  If someone else is doing it then we don't
+		 * want to go to the push routine and duplicate their efforts.
+		 */
+		if (qip->qli_pushbuf_flag == 0) {
+			qip->qli_pushbuf_flag = 1;
+			ASSERT(qip->qli_format.qlf_blkno == dqp->q_blkno);
+#ifdef DEBUG
+			qip->qli_push_owner = get_thread_id();
+#endif
+			/*
+			 * The dquot is left locked.
+			 */
+			retval = XFS_ITEM_PUSHBUF;
+		} else {
+			retval = XFS_ITEM_FLUSHING;
+			xfs_dqunlock_nonotify(dqp);
+		}
+	}
+
+	ASSERT(qip->qli_item.li_flags & XFS_LI_IN_AIL);
+	return (retval);
+}
+
+
+/*
+ * Unlock the dquot associated with the log item.
+ * Clear the fields of the dquot and dquot log item that
+ * are specific to the current transaction.  If the
+ * hold flags is set, do not unlock the dquot.
+ */
+STATIC void
+xfs_qm_dquot_logitem_unlock(
+	xfs_dq_logitem_t    *ql)
+{
+	xfs_dquot_t	*dqp;
+
+	ASSERT(ql != NULL);
+	dqp = ql->qli_dquot;
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+
+	/*
+	 * Clear the transaction pointer in the dquot
+	 */
+	dqp->q_transp = NULL;
+
+	/*
+	 * dquots are never 'held' from getting unlocked at the end of
+	 * a transaction.  Their locking and unlocking is hidden inside the
+	 * transaction layer, within trans_commit. Hence, no LI_HOLD flag
+	 * for the logitem.
+	 */
+	xfs_dqunlock(dqp);
+}
+
+
+/*
+ * The transaction with the dquot locked has aborted.  The dquot
+ * must not be dirty within the transaction.  We simply unlock just
+ * as if the transaction had been cancelled.
+ */
+STATIC void
+xfs_qm_dquot_logitem_abort(
+	xfs_dq_logitem_t    *ql)
+{
+	xfs_qm_dquot_logitem_unlock(ql);
+}
+
+/*
+ * this needs to stamp an lsn into the dquot, I think.
+ * rpc's that look at user dquot's would then have to
+ * push on the dependency recorded in the dquot
+ */
+/* ARGSUSED */
+STATIC void
+xfs_qm_dquot_logitem_committing(
+	xfs_dq_logitem_t	*l,
+	xfs_lsn_t		lsn)
+{
+	return;
+}
+
+
+/*
+ * This is the ops vector for dquots
+ */
+struct xfs_item_ops xfs_dquot_item_ops = {
+	.iop_size	= (uint(*)(xfs_log_item_t*))xfs_qm_dquot_logitem_size,
+	.iop_format	= (void(*)(xfs_log_item_t*, xfs_log_iovec_t*))
+					xfs_qm_dquot_logitem_format,
+	.iop_pin	= (void(*)(xfs_log_item_t*))xfs_qm_dquot_logitem_pin,
+	.iop_unpin	= (void(*)(xfs_log_item_t*, int))
+					xfs_qm_dquot_logitem_unpin,
+	.iop_unpin_remove = (void(*)(xfs_log_item_t*, xfs_trans_t*))
+					xfs_qm_dquot_logitem_unpin_remove,
+	.iop_trylock	= (uint(*)(xfs_log_item_t*))
+					xfs_qm_dquot_logitem_trylock,
+	.iop_unlock	= (void(*)(xfs_log_item_t*))xfs_qm_dquot_logitem_unlock,
+	.iop_committed	= (xfs_lsn_t(*)(xfs_log_item_t*, xfs_lsn_t))
+					xfs_qm_dquot_logitem_committed,
+	.iop_push	= (void(*)(xfs_log_item_t*))xfs_qm_dquot_logitem_push,
+	.iop_abort	= (void(*)(xfs_log_item_t*))xfs_qm_dquot_logitem_abort,
+	.iop_pushbuf	= (void(*)(xfs_log_item_t*))
+					xfs_qm_dquot_logitem_pushbuf,
+	.iop_committing = (void(*)(xfs_log_item_t*, xfs_lsn_t))
+					xfs_qm_dquot_logitem_committing
+};
+
+/*
+ * Initialize the dquot log item for a newly allocated dquot.
+ * The dquot isn't locked at this point, but it isn't on any of the lists
+ * either, so we don't care.
+ */
+void
+xfs_qm_dquot_logitem_init(
+	struct xfs_dquot *dqp)
+{
+	xfs_dq_logitem_t  *lp;
+	lp = &dqp->q_logitem;
+
+	lp->qli_item.li_type = XFS_LI_DQUOT;
+	lp->qli_item.li_ops = &xfs_dquot_item_ops;
+	lp->qli_item.li_mountp = dqp->q_mount;
+	lp->qli_dquot = dqp;
+	lp->qli_format.qlf_type = XFS_LI_DQUOT;
+	lp->qli_format.qlf_id = INT_GET(dqp->q_core.d_id, ARCH_CONVERT);
+	lp->qli_format.qlf_blkno = dqp->q_blkno;
+	lp->qli_format.qlf_len = 1;
+	/*
+	 * This is just the offset of this dquot within its buffer
+	 * (which is currently 1 FSB and probably won't change).
+	 * Hence 32 bits for this offset should be just fine.
+	 * Alternatively, we can store (bufoffset / sizeof(xfs_dqblk_t))
+	 * here, and recompute it at recovery time.
+	 */
+	lp->qli_format.qlf_boffset = (__uint32_t)dqp->q_bufoffset;
+}
+
+/*------------------  QUOTAOFF LOG ITEMS  -------------------*/
+
+/*
+ * This returns the number of iovecs needed to log the given quotaoff item.
+ * We only need 1 iovec for an quotaoff item.  It just logs the
+ * quotaoff_log_format structure.
+ */
+/*ARGSUSED*/
+STATIC uint
+xfs_qm_qoff_logitem_size(xfs_qoff_logitem_t *qf)
+{
+	return (1);
+}
+
+/*
+ * This is called to fill in the vector of log iovecs for the
+ * given quotaoff log item. We use only 1 iovec, and we point that
+ * at the quotaoff_log_format structure embedded in the quotaoff item.
+ * It is at this point that we assert that all of the extent
+ * slots in the quotaoff item have been filled.
+ */
+STATIC void
+xfs_qm_qoff_logitem_format(xfs_qoff_logitem_t	*qf,
+			   xfs_log_iovec_t	*log_vector)
+{
+	ASSERT(qf->qql_format.qf_type == XFS_LI_QUOTAOFF);
+
+	log_vector->i_addr = (xfs_caddr_t)&(qf->qql_format);
+	log_vector->i_len = sizeof(xfs_qoff_logitem_t);
+	qf->qql_format.qf_size = 1;
+}
+
+
+/*
+ * Pinning has no meaning for an quotaoff item, so just return.
+ */
+/*ARGSUSED*/
+STATIC void
+xfs_qm_qoff_logitem_pin(xfs_qoff_logitem_t *qf)
+{
+	return;
+}
+
+
+/*
+ * Since pinning has no meaning for an quotaoff item, unpinning does
+ * not either.
+ */
+/*ARGSUSED*/
+STATIC void
+xfs_qm_qoff_logitem_unpin(xfs_qoff_logitem_t *qf, int stale)
+{
+	return;
+}
+
+/*ARGSUSED*/
+STATIC void
+xfs_qm_qoff_logitem_unpin_remove(xfs_qoff_logitem_t *qf, xfs_trans_t *tp)
+{
+	return;
+}
+
+/*
+ * Quotaoff items have no locking, so just return success.
+ */
+/*ARGSUSED*/
+STATIC uint
+xfs_qm_qoff_logitem_trylock(xfs_qoff_logitem_t *qf)
+{
+	return XFS_ITEM_LOCKED;
+}
+
+/*
+ * Quotaoff items have no locking or pushing, so return failure
+ * so that the caller doesn't bother with us.
+ */
+/*ARGSUSED*/
+STATIC void
+xfs_qm_qoff_logitem_unlock(xfs_qoff_logitem_t *qf)
+{
+	return;
+}
+
+/*
+ * The quotaoff-start-item is logged only once and cannot be moved in the log,
+ * so simply return the lsn at which it's been logged.
+ */
+/*ARGSUSED*/
+STATIC xfs_lsn_t
+xfs_qm_qoff_logitem_committed(xfs_qoff_logitem_t *qf, xfs_lsn_t lsn)
+{
+	return (lsn);
+}
+
+/*
+ * The transaction of which this QUOTAOFF is a part has been aborted.
+ * Just clean up after ourselves.
+ * Shouldn't this never happen in the case of qoffend logitems? XXX
+ */
+STATIC void
+xfs_qm_qoff_logitem_abort(xfs_qoff_logitem_t *qf)
+{
+	kmem_free(qf, sizeof(xfs_qoff_logitem_t));
+}
+
+/*
+ * There isn't much you can do to push on an quotaoff item.  It is simply
+ * stuck waiting for the log to be flushed to disk.
+ */
+/*ARGSUSED*/
+STATIC void
+xfs_qm_qoff_logitem_push(xfs_qoff_logitem_t *qf)
+{
+	return;
+}
+
+
+/*ARGSUSED*/
+STATIC xfs_lsn_t
+xfs_qm_qoffend_logitem_committed(
+	xfs_qoff_logitem_t *qfe,
+	xfs_lsn_t lsn)
+{
+	xfs_qoff_logitem_t	*qfs;
+	SPLDECL(s);
+
+	qfs = qfe->qql_start_lip;
+	AIL_LOCK(qfs->qql_item.li_mountp,s);
+	/*
+	 * Delete the qoff-start logitem from the AIL.
+	 * xfs_trans_delete_ail() drops the AIL lock.
+	 */
+	xfs_trans_delete_ail(qfs->qql_item.li_mountp, (xfs_log_item_t *)qfs, s);
+	kmem_free(qfs, sizeof(xfs_qoff_logitem_t));
+	kmem_free(qfe, sizeof(xfs_qoff_logitem_t));
+	return (xfs_lsn_t)-1;
+}
+
+/*
+ * XXX rcc - don't know quite what to do with this.  I think we can
+ * just ignore it.  The only time that isn't the case is if we allow
+ * the client to somehow see that quotas have been turned off in which
+ * we can't allow that to get back until the quotaoff hits the disk.
+ * So how would that happen?  Also, do we need different routines for
+ * quotaoff start and quotaoff end?  I suspect the answer is yes but
+ * to be sure, I need to look at the recovery code and see how quota off
+ * recovery is handled (do we roll forward or back or do something else).
+ * If we roll forwards or backwards, then we need two separate routines,
+ * one that does nothing and one that stamps in the lsn that matters
+ * (truly makes the quotaoff irrevocable).  If we do something else,
+ * then maybe we don't need two.
+ */
+/* ARGSUSED */
+STATIC void
+xfs_qm_qoff_logitem_committing(xfs_qoff_logitem_t *qip, xfs_lsn_t commit_lsn)
+{
+	return;
+}
+
+/* ARGSUSED */
+STATIC void
+xfs_qm_qoffend_logitem_committing(xfs_qoff_logitem_t *qip, xfs_lsn_t commit_lsn)
+{
+	return;
+}
+
+struct xfs_item_ops xfs_qm_qoffend_logitem_ops = {
+	.iop_size	= (uint(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_size,
+	.iop_format	= (void(*)(xfs_log_item_t*, xfs_log_iovec_t*))
+					xfs_qm_qoff_logitem_format,
+	.iop_pin	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_pin,
+	.iop_unpin	= (void(*)(xfs_log_item_t* ,int))
+					xfs_qm_qoff_logitem_unpin,
+	.iop_unpin_remove = (void(*)(xfs_log_item_t*,xfs_trans_t*))
+					xfs_qm_qoff_logitem_unpin_remove,
+	.iop_trylock	= (uint(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_trylock,
+	.iop_unlock	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_unlock,
+	.iop_committed	= (xfs_lsn_t(*)(xfs_log_item_t*, xfs_lsn_t))
+					xfs_qm_qoffend_logitem_committed,
+	.iop_push	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_push,
+	.iop_abort	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_abort,
+	.iop_pushbuf	= NULL,
+	.iop_committing = (void(*)(xfs_log_item_t*, xfs_lsn_t))
+					xfs_qm_qoffend_logitem_committing
+};
+
+/*
+ * This is the ops vector shared by all quotaoff-start log items.
+ */
+struct xfs_item_ops xfs_qm_qoff_logitem_ops = {
+	.iop_size	= (uint(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_size,
+	.iop_format	= (void(*)(xfs_log_item_t*, xfs_log_iovec_t*))
+					xfs_qm_qoff_logitem_format,
+	.iop_pin	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_pin,
+	.iop_unpin	= (void(*)(xfs_log_item_t*, int))
+					xfs_qm_qoff_logitem_unpin,
+	.iop_unpin_remove = (void(*)(xfs_log_item_t*,xfs_trans_t*))
+					xfs_qm_qoff_logitem_unpin_remove,
+	.iop_trylock	= (uint(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_trylock,
+	.iop_unlock	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_unlock,
+	.iop_committed	= (xfs_lsn_t(*)(xfs_log_item_t*, xfs_lsn_t))
+					xfs_qm_qoff_logitem_committed,
+	.iop_push	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_push,
+	.iop_abort	= (void(*)(xfs_log_item_t*))xfs_qm_qoff_logitem_abort,
+	.iop_pushbuf	= NULL,
+	.iop_committing = (void(*)(xfs_log_item_t*, xfs_lsn_t))
+					xfs_qm_qoff_logitem_committing
+};
+
+/*
+ * Allocate and initialize an quotaoff item of the correct quota type(s).
+ */
+xfs_qoff_logitem_t *
+xfs_qm_qoff_logitem_init(
+	struct xfs_mount *mp,
+	xfs_qoff_logitem_t *start,
+	uint flags)
+{
+	xfs_qoff_logitem_t	*qf;
+
+	qf = (xfs_qoff_logitem_t*) kmem_zalloc(sizeof(xfs_qoff_logitem_t), KM_SLEEP);
+
+	qf->qql_item.li_type = XFS_LI_QUOTAOFF;
+	if (start)
+		qf->qql_item.li_ops = &xfs_qm_qoffend_logitem_ops;
+	else
+		qf->qql_item.li_ops = &xfs_qm_qoff_logitem_ops;
+	qf->qql_item.li_mountp = mp;
+	qf->qql_format.qf_type = XFS_LI_QUOTAOFF;
+	qf->qql_format.qf_flags = flags;
+	qf->qql_start_lip = start;
+	return (qf);
+}
diff --git a/fs/xfs/quota/xfs_dquot_item.h b/fs/xfs/quota/xfs_dquot_item.h
new file mode 100644
index 000000000000..9c6500dabcaa
--- /dev/null
+++ b/fs/xfs/quota/xfs_dquot_item.h
@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+#ifndef __XFS_DQUOT_ITEM_H__
+#define __XFS_DQUOT_ITEM_H__
+
+struct xfs_dquot;
+struct xfs_trans;
+struct xfs_mount;
+struct xfs_qoff_logitem;
+
+typedef struct xfs_dq_logitem {
+	xfs_log_item_t		 qli_item;	   /* common portion */
+	struct xfs_dquot	*qli_dquot;	   /* dquot ptr */
+	xfs_lsn_t		 qli_flush_lsn;	   /* lsn at last flush */
+	unsigned short		 qli_pushbuf_flag; /* 1 bit used in push_ail */
+#ifdef DEBUG
+	uint64_t		 qli_push_owner;
+#endif
+	xfs_dq_logformat_t	 qli_format;	   /* logged structure */
+} xfs_dq_logitem_t;
+
+typedef struct xfs_qoff_logitem {
+	xfs_log_item_t		 qql_item;	/* common portion */
+	struct xfs_qoff_logitem *qql_start_lip; /* qoff-start logitem, if any */
+	xfs_qoff_logformat_t	 qql_format;	/* logged structure */
+} xfs_qoff_logitem_t;
+
+
+extern void		   xfs_qm_dquot_logitem_init(struct xfs_dquot *);
+extern xfs_qoff_logitem_t *xfs_qm_qoff_logitem_init(struct xfs_mount *,
+					struct xfs_qoff_logitem *, uint);
+extern xfs_qoff_logitem_t *xfs_trans_get_qoff_item(struct xfs_trans *,
+					struct xfs_qoff_logitem *, uint);
+extern void		   xfs_trans_log_quotaoff_item(struct xfs_trans *,
+					struct xfs_qoff_logitem *);
+
+#endif	/* __XFS_DQUOT_ITEM_H__ */
diff --git a/fs/xfs/quota/xfs_qm.c b/fs/xfs/quota/xfs_qm.c
new file mode 100644
index 000000000000..89f2cd656ebf
--- /dev/null
+++ b/fs/xfs/quota/xfs_qm.c
@@ -0,0 +1,2848 @@
+/*
+ * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_clnt.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_ag.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+#include "xfs_trans_space.h"
+#include "xfs_utils.h"
+
+#include "xfs_qm.h"
+
+/*
+ * The global quota manager. There is only one of these for the entire
+ * system, _not_ one per file system. XQM keeps track of the overall
+ * quota functionality, including maintaining the freelist and hash
+ * tables of dquots.
+ */
+mutex_t xfs_Gqm_lock;
+struct xfs_qm	*xfs_Gqm;
+
+kmem_zone_t	*qm_dqzone;
+kmem_zone_t	*qm_dqtrxzone;
+kmem_shaker_t	xfs_qm_shaker;
+
+STATIC void	xfs_qm_list_init(xfs_dqlist_t *, char *, int);
+STATIC void	xfs_qm_list_destroy(xfs_dqlist_t *);
+
+STATIC int	xfs_qm_init_quotainos(xfs_mount_t *);
+STATIC int	xfs_qm_shake(int, unsigned int);
+
+#ifdef DEBUG
+extern mutex_t	qcheck_lock;
+#endif
+
+#ifdef QUOTADEBUG
+#define XQM_LIST_PRINT(l, NXT, title) \
+{ \
+	xfs_dquot_t	*dqp; int i = 0; \
+	cmn_err(CE_DEBUG, "%s (#%d)", title, (int) (l)->qh_nelems); \
+	for (dqp = (l)->qh_next; dqp != NULL; dqp = dqp->NXT) { \
+		cmn_err(CE_DEBUG, "   %d.  \"%d (%s)\"   " \
+				  "bcnt = %d, icnt = %d, refs = %d", \
+			++i, (int) INT_GET(dqp->q_core.d_id, ARCH_CONVERT), \
+			DQFLAGTO_TYPESTR(dqp),	     \
+			(int) INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT), \
+			(int) INT_GET(dqp->q_core.d_icount, ARCH_CONVERT), \
+			(int) dqp->q_nrefs);  } \
+}
+#else
+#define XQM_LIST_PRINT(l, NXT, title) do { } while (0)
+#endif
+
+/*
+ * Initialize the XQM structure.
+ * Note that there is not one quota manager per file system.
+ */
+STATIC struct xfs_qm *
+xfs_Gqm_init(void)
+{
+	xfs_qm_t		*xqm;
+	int			hsize, i;
+
+	xqm = kmem_zalloc(sizeof(xfs_qm_t), KM_SLEEP);
+	ASSERT(xqm);
+
+	/*
+	 * Initialize the dquot hash tables.
+	 */
+	hsize = (DQUOT_HASH_HEURISTIC < XFS_QM_NCSIZE_THRESHOLD) ?
+		XFS_QM_HASHSIZE_LOW : XFS_QM_HASHSIZE_HIGH;
+	xqm->qm_dqhashmask = hsize - 1;
+
+	xqm->qm_usr_dqhtable = (xfs_dqhash_t *)kmem_zalloc(hsize *
+						      sizeof(xfs_dqhash_t),
+						      KM_SLEEP);
+	xqm->qm_grp_dqhtable = (xfs_dqhash_t *)kmem_zalloc(hsize *
+						      sizeof(xfs_dqhash_t),
+						      KM_SLEEP);
+	ASSERT(xqm->qm_usr_dqhtable != NULL);
+	ASSERT(xqm->qm_grp_dqhtable != NULL);
+
+	for (i = 0; i < hsize; i++) {
+		xfs_qm_list_init(&(xqm->qm_usr_dqhtable[i]), "uxdqh", i);
+		xfs_qm_list_init(&(xqm->qm_grp_dqhtable[i]), "gxdqh", i);
+	}
+
+	/*
+	 * Freelist of all dquots of all file systems
+	 */
+	xfs_qm_freelist_init(&(xqm->qm_dqfreelist));
+
+	/*
+	 * dquot zone. we register our own low-memory callback.
+	 */
+	if (!qm_dqzone) {
+		xqm->qm_dqzone = kmem_zone_init(sizeof(xfs_dquot_t),
+						"xfs_dquots");
+		qm_dqzone = xqm->qm_dqzone;
+	} else
+		xqm->qm_dqzone = qm_dqzone;
+
+	xfs_qm_shaker = kmem_shake_register(xfs_qm_shake);
+
+	/*
+	 * The t_dqinfo portion of transactions.
+	 */
+	if (!qm_dqtrxzone) {
+		xqm->qm_dqtrxzone = kmem_zone_init(sizeof(xfs_dquot_acct_t),
+						   "xfs_dqtrx");
+		qm_dqtrxzone = xqm->qm_dqtrxzone;
+	} else
+		xqm->qm_dqtrxzone = qm_dqtrxzone;
+
+	atomic_set(&xqm->qm_totaldquots, 0);
+	xqm->qm_dqfree_ratio = XFS_QM_DQFREE_RATIO;
+	xqm->qm_nrefs = 0;
+#ifdef DEBUG
+	mutex_init(&qcheck_lock, MUTEX_DEFAULT, "qchk");
+#endif
+	return xqm;
+}
+
+/*
+ * Destroy the global quota manager when its reference count goes to zero.
+ */
+void
+xfs_qm_destroy(
+	struct xfs_qm	*xqm)
+{
+	int		hsize, i;
+
+	ASSERT(xqm != NULL);
+	ASSERT(xqm->qm_nrefs == 0);
+	kmem_shake_deregister(xfs_qm_shaker);
+	hsize = xqm->qm_dqhashmask + 1;
+	for (i = 0; i < hsize; i++) {
+		xfs_qm_list_destroy(&(xqm->qm_usr_dqhtable[i]));
+		xfs_qm_list_destroy(&(xqm->qm_grp_dqhtable[i]));
+	}
+	kmem_free(xqm->qm_usr_dqhtable, hsize * sizeof(xfs_dqhash_t));
+	kmem_free(xqm->qm_grp_dqhtable, hsize * sizeof(xfs_dqhash_t));
+	xqm->qm_usr_dqhtable = NULL;
+	xqm->qm_grp_dqhtable = NULL;
+	xqm->qm_dqhashmask = 0;
+	xfs_qm_freelist_destroy(&(xqm->qm_dqfreelist));
+#ifdef DEBUG
+	mutex_destroy(&qcheck_lock);
+#endif
+	kmem_free(xqm, sizeof(xfs_qm_t));
+}
+
+/*
+ * Called at mount time to let XQM know that another file system is
+ * starting quotas. This isn't crucial information as the individual mount
+ * structures are pretty independent, but it helps the XQM keep a
+ * global view of what's going on.
+ */
+/* ARGSUSED */
+STATIC int
+xfs_qm_hold_quotafs_ref(
+	struct xfs_mount *mp)
+{
+	/*
+	 * Need to lock the xfs_Gqm structure for things like this. For example,
+	 * the structure could disappear between the entry to this routine and
+	 * a HOLD operation if not locked.
+	 */
+	XFS_QM_LOCK(xfs_Gqm);
+
+	if (xfs_Gqm == NULL)
+		xfs_Gqm = xfs_Gqm_init();
+	/*
+	 * We can keep a list of all filesystems with quotas mounted for
+	 * debugging and statistical purposes, but ...
+	 * Just take a reference and get out.
+	 */
+	XFS_QM_HOLD(xfs_Gqm);
+	XFS_QM_UNLOCK(xfs_Gqm);
+
+	return 0;
+}
+
+
+/*
+ * Release the reference that a filesystem took at mount time,
+ * so that we know when we need to destroy the entire quota manager.
+ */
+/* ARGSUSED */
+STATIC void
+xfs_qm_rele_quotafs_ref(
+	struct xfs_mount *mp)
+{
+	xfs_dquot_t	*dqp, *nextdqp;
+
+	ASSERT(xfs_Gqm);
+	ASSERT(xfs_Gqm->qm_nrefs > 0);
+
+	/*
+	 * Go thru the freelist and destroy all inactive dquots.
+	 */
+	xfs_qm_freelist_lock(xfs_Gqm);
+
+	for (dqp = xfs_Gqm->qm_dqfreelist.qh_next;
+	     dqp != (xfs_dquot_t *)&(xfs_Gqm->qm_dqfreelist); ) {
+		xfs_dqlock(dqp);
+		nextdqp = dqp->dq_flnext;
+		if (dqp->dq_flags & XFS_DQ_INACTIVE) {
+			ASSERT(dqp->q_mount == NULL);
+			ASSERT(! XFS_DQ_IS_DIRTY(dqp));
+			ASSERT(dqp->HL_PREVP == NULL);
+			ASSERT(dqp->MPL_PREVP == NULL);
+			XQM_FREELIST_REMOVE(dqp);
+			xfs_dqunlock(dqp);
+			xfs_qm_dqdestroy(dqp);
+		} else {
+			xfs_dqunlock(dqp);
+		}
+		dqp = nextdqp;
+	}
+	xfs_qm_freelist_unlock(xfs_Gqm);
+
+	/*
+	 * Destroy the entire XQM. If somebody mounts with quotaon, this'll
+	 * be restarted.
+	 */
+	XFS_QM_LOCK(xfs_Gqm);
+	XFS_QM_RELE(xfs_Gqm);
+	if (xfs_Gqm->qm_nrefs == 0) {
+		xfs_qm_destroy(xfs_Gqm);
+		xfs_Gqm = NULL;
+	}
+	XFS_QM_UNLOCK(xfs_Gqm);
+}
+
+/*
+ * This is called at mount time from xfs_mountfs to initialize the quotainfo
+ * structure and start the global quotamanager (xfs_Gqm) if it hasn't done
+ * so already.	Note that the superblock has not been read in yet.
+ */
+void
+xfs_qm_mount_quotainit(
+	xfs_mount_t	*mp,
+	uint		flags)
+{
+	/*
+	 * User or group quotas has to be on.
+	 */
+	ASSERT(flags & (XFSMNT_UQUOTA | XFSMNT_GQUOTA));
+
+	/*
+	 * Initialize the flags in the mount structure. From this point
+	 * onwards we look at m_qflags to figure out if quotas's ON/OFF, etc.
+	 * Note that we enforce nothing if accounting is off.
+	 * ie.	XFSMNT_*QUOTA must be ON for XFSMNT_*QUOTAENF.
+	 * It isn't necessary to take the quotaoff lock to do this; this is
+	 * called from mount.
+	 */
+	if (flags & XFSMNT_UQUOTA) {
+		mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
+		if (flags & XFSMNT_UQUOTAENF)
+			mp->m_qflags |= XFS_UQUOTA_ENFD;
+	}
+	if (flags & XFSMNT_GQUOTA) {
+		mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
+		if (flags & XFSMNT_GQUOTAENF)
+			mp->m_qflags |= XFS_GQUOTA_ENFD;
+	}
+}
+
+/*
+ * Just destroy the quotainfo structure.
+ */
+void
+xfs_qm_unmount_quotadestroy(
+	xfs_mount_t	*mp)
+{
+	if (mp->m_quotainfo)
+		xfs_qm_destroy_quotainfo(mp);
+}
+
+
+/*
+ * This is called from xfs_mountfs to start quotas and initialize all
+ * necessary data structures like quotainfo.  This is also responsible for
+ * running a quotacheck as necessary.  We are guaranteed that the superblock
+ * is consistently read in at this point.
+ */
+int
+xfs_qm_mount_quotas(
+	xfs_mount_t	*mp,
+	int		mfsi_flags)
+{
+	unsigned long	s;
+	int		error = 0;
+	uint		sbf;
+
+	/*
+	 * If a file system had quotas running earlier, but decided to
+	 * mount without -o quota/uquota/gquota options, revoke the
+	 * quotachecked license, and bail out.
+	 */
+	if (! XFS_IS_QUOTA_ON(mp) &&
+	    (mp->m_sb.sb_qflags & (XFS_UQUOTA_ACCT|XFS_GQUOTA_ACCT))) {
+		mp->m_qflags = 0;
+		goto write_changes;
+	}
+
+	/*
+	 * If quotas on realtime volumes is not supported, we disable
+	 * quotas immediately.
+	 */
+	if (mp->m_sb.sb_rextents) {
+		cmn_err(CE_NOTE,
+			"Cannot turn on quotas for realtime filesystem %s",
+			mp->m_fsname);
+		mp->m_qflags = 0;
+		goto write_changes;
+	}
+
+#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
+	cmn_err(CE_NOTE, "Attempting to turn on disk quotas.");
+#endif
+
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+	/*
+	 * Allocate the quotainfo structure inside the mount struct, and
+	 * create quotainode(s), and change/rev superblock if necessary.
+	 */
+	if ((error = xfs_qm_init_quotainfo(mp))) {
+		/*
+		 * We must turn off quotas.
+		 */
+		ASSERT(mp->m_quotainfo == NULL);
+		mp->m_qflags = 0;
+		goto write_changes;
+	}
+	/*
+	 * If any of the quotas are not consistent, do a quotacheck.
+	 */
+	if (XFS_QM_NEED_QUOTACHECK(mp) &&
+		!(mfsi_flags & XFS_MFSI_NO_QUOTACHECK)) {
+#ifdef DEBUG
+		cmn_err(CE_NOTE, "Doing a quotacheck. Please wait.");
+#endif
+		if ((error = xfs_qm_quotacheck(mp))) {
+			/* Quotacheck has failed and quotas have
+			 * been disabled.
+			 */
+			return XFS_ERROR(error);
+		}
+#ifdef DEBUG
+		cmn_err(CE_NOTE, "Done quotacheck.");
+#endif
+	}
+ write_changes:
+	/*
+	 * We actually don't have to acquire the SB_LOCK at all.
+	 * This can only be called from mount, and that's single threaded. XXX
+	 */
+	s = XFS_SB_LOCK(mp);
+	sbf = mp->m_sb.sb_qflags;
+	mp->m_sb.sb_qflags = mp->m_qflags & XFS_MOUNT_QUOTA_ALL;
+	XFS_SB_UNLOCK(mp, s);
+
+	if (sbf != (mp->m_qflags & XFS_MOUNT_QUOTA_ALL)) {
+		if (xfs_qm_write_sb_changes(mp, XFS_SB_QFLAGS)) {
+			/*
+			 * We could only have been turning quotas off.
+			 * We aren't in very good shape actually because
+			 * the incore structures are convinced that quotas are
+			 * off, but the on disk superblock doesn't know that !
+			 */
+			ASSERT(!(XFS_IS_QUOTA_RUNNING(mp)));
+			xfs_fs_cmn_err(CE_ALERT, mp,
+				"XFS mount_quotas: Superblock update failed!");
+		}
+	}
+
+	if (error) {
+		xfs_fs_cmn_err(CE_WARN, mp,
+			"Failed to initialize disk quotas.");
+	}
+	return XFS_ERROR(error);
+}
+
+/*
+ * Called from the vfsops layer.
+ */
+int
+xfs_qm_unmount_quotas(
+	xfs_mount_t	*mp)
+{
+	xfs_inode_t	*uqp, *gqp;
+	int		error = 0;
+
+	/*
+	 * Release the dquots that root inode, et al might be holding,
+	 * before we flush quotas and blow away the quotainfo structure.
+	 */
+	ASSERT(mp->m_rootip);
+	xfs_qm_dqdetach(mp->m_rootip);
+	if (mp->m_rbmip)
+		xfs_qm_dqdetach(mp->m_rbmip);
+	if (mp->m_rsumip)
+		xfs_qm_dqdetach(mp->m_rsumip);
+
+	/*
+	 * Flush out the quota inodes.
+	 */
+	uqp = gqp = NULL;
+	if (mp->m_quotainfo) {
+		if ((uqp = mp->m_quotainfo->qi_uquotaip) != NULL) {
+			xfs_ilock(uqp, XFS_ILOCK_EXCL);
+			xfs_iflock(uqp);
+			error = xfs_iflush(uqp, XFS_IFLUSH_SYNC);
+			xfs_iunlock(uqp, XFS_ILOCK_EXCL);
+			if (unlikely(error == EFSCORRUPTED)) {
+				XFS_ERROR_REPORT("xfs_qm_unmount_quotas(1)",
+						 XFS_ERRLEVEL_LOW, mp);
+				goto out;
+			}
+		}
+		if ((gqp = mp->m_quotainfo->qi_gquotaip) != NULL) {
+			xfs_ilock(gqp, XFS_ILOCK_EXCL);
+			xfs_iflock(gqp);
+			error = xfs_iflush(gqp, XFS_IFLUSH_SYNC);
+			xfs_iunlock(gqp, XFS_ILOCK_EXCL);
+			if (unlikely(error == EFSCORRUPTED)) {
+				XFS_ERROR_REPORT("xfs_qm_unmount_quotas(2)",
+						 XFS_ERRLEVEL_LOW, mp);
+				goto out;
+			}
+		}
+	}
+	if (uqp) {
+		 XFS_PURGE_INODE(uqp);
+		 mp->m_quotainfo->qi_uquotaip = NULL;
+	}
+	if (gqp) {
+		XFS_PURGE_INODE(gqp);
+		mp->m_quotainfo->qi_gquotaip = NULL;
+	}
+out:
+	return XFS_ERROR(error);
+}
+
+/*
+ * Flush all dquots of the given file system to disk. The dquots are
+ * _not_ purged from memory here, just their data written to disk.
+ */
+int
+xfs_qm_dqflush_all(
+	xfs_mount_t	*mp,
+	int		flags)
+{
+	int		recl;
+	xfs_dquot_t	*dqp;
+	int		niters;
+	int		error;
+
+	if (mp->m_quotainfo == NULL)
+		return (0);
+	niters = 0;
+again:
+	xfs_qm_mplist_lock(mp);
+	FOREACH_DQUOT_IN_MP(dqp, mp) {
+		xfs_dqlock(dqp);
+		if (! XFS_DQ_IS_DIRTY(dqp)) {
+			xfs_dqunlock(dqp);
+			continue;
+		}
+		xfs_dqtrace_entry(dqp, "FLUSHALL: DQDIRTY");
+		/* XXX a sentinel would be better */
+		recl = XFS_QI_MPLRECLAIMS(mp);
+		if (! xfs_qm_dqflock_nowait(dqp)) {
+			/*
+			 * If we can't grab the flush lock then check
+			 * to see if the dquot has been flushed delayed
+			 * write.  If so, grab its buffer and send it
+			 * out immediately.  We'll be able to acquire
+			 * the flush lock when the I/O completes.
+			 */
+			xfs_qm_dqflock_pushbuf_wait(dqp);
+		}
+		/*
+		 * Let go of the mplist lock. We don't want to hold it
+		 * across a disk write.
+		 */
+		xfs_qm_mplist_unlock(mp);
+		error = xfs_qm_dqflush(dqp, flags);
+		xfs_dqunlock(dqp);
+		if (error)
+			return (error);
+
+		xfs_qm_mplist_lock(mp);
+		if (recl != XFS_QI_MPLRECLAIMS(mp)) {
+			xfs_qm_mplist_unlock(mp);
+			/* XXX restart limit */
+			goto again;
+		}
+	}
+
+	xfs_qm_mplist_unlock(mp);
+	/* return ! busy */
+	return (0);
+}
+/*
+ * Release the group dquot pointers the user dquots may be
+ * carrying around as a hint. mplist is locked on entry and exit.
+ */
+STATIC void
+xfs_qm_detach_gdquots(
+	xfs_mount_t	*mp)
+{
+	xfs_dquot_t	*dqp, *gdqp;
+	int		nrecl;
+
+ again:
+	ASSERT(XFS_QM_IS_MPLIST_LOCKED(mp));
+	dqp = XFS_QI_MPLNEXT(mp);
+	while (dqp) {
+		xfs_dqlock(dqp);
+		if ((gdqp = dqp->q_gdquot)) {
+			xfs_dqlock(gdqp);
+			dqp->q_gdquot = NULL;
+		}
+		xfs_dqunlock(dqp);
+
+		if (gdqp) {
+			/*
+			 * Can't hold the mplist lock across a dqput.
+			 * XXXmust convert to marker based iterations here.
+			 */
+			nrecl = XFS_QI_MPLRECLAIMS(mp);
+			xfs_qm_mplist_unlock(mp);
+			xfs_qm_dqput(gdqp);
+
+			xfs_qm_mplist_lock(mp);
+			if (nrecl != XFS_QI_MPLRECLAIMS(mp))
+				goto again;
+		}
+		dqp = dqp->MPL_NEXT;
+	}
+}
+
+/*
+ * Go through all the incore dquots of this file system and take them
+ * off the mplist and hashlist, if the dquot type matches the dqtype
+ * parameter. This is used when turning off quota accounting for
+ * users and/or groups, as well as when the filesystem is unmounting.
+ */
+STATIC int
+xfs_qm_dqpurge_int(
+	xfs_mount_t	*mp,
+	uint		flags) /* QUOTAOFF/UMOUNTING/UQUOTA/GQUOTA */
+{
+	xfs_dquot_t	*dqp;
+	uint		dqtype;
+	int		nrecl;
+	xfs_dquot_t	*nextdqp;
+	int		nmisses;
+
+	if (mp->m_quotainfo == NULL)
+		return (0);
+
+	dqtype = (flags & XFS_QMOPT_UQUOTA) ? XFS_DQ_USER : 0;
+	dqtype |= (flags & XFS_QMOPT_GQUOTA) ? XFS_DQ_GROUP : 0;
+
+	xfs_qm_mplist_lock(mp);
+
+	/*
+	 * In the first pass through all incore dquots of this filesystem,
+	 * we release the group dquot pointers the user dquots may be
+	 * carrying around as a hint. We need to do this irrespective of
+	 * what's being turned off.
+	 */
+	xfs_qm_detach_gdquots(mp);
+
+      again:
+	nmisses = 0;
+	ASSERT(XFS_QM_IS_MPLIST_LOCKED(mp));
+	/*
+	 * Try to get rid of all of the unwanted dquots. The idea is to
+	 * get them off mplist and hashlist, but leave them on freelist.
+	 */
+	dqp = XFS_QI_MPLNEXT(mp);
+	while (dqp) {
+		/*
+		 * It's OK to look at the type without taking dqlock here.
+		 * We're holding the mplist lock here, and that's needed for
+		 * a dqreclaim.
+		 */
+		if ((dqp->dq_flags & dqtype) == 0) {
+			dqp = dqp->MPL_NEXT;
+			continue;
+		}
+
+		if (! xfs_qm_dqhashlock_nowait(dqp)) {
+			nrecl = XFS_QI_MPLRECLAIMS(mp);
+			xfs_qm_mplist_unlock(mp);
+			XFS_DQ_HASH_LOCK(dqp->q_hash);
+			xfs_qm_mplist_lock(mp);
+
+			/*
+			 * XXXTheoretically, we can get into a very long
+			 * ping pong game here.
+			 * No one can be adding dquots to the mplist at
+			 * this point, but somebody might be taking things off.
+			 */
+			if (nrecl != XFS_QI_MPLRECLAIMS(mp)) {
+				XFS_DQ_HASH_UNLOCK(dqp->q_hash);
+				goto again;
+			}
+		}
+
+		/*
+		 * Take the dquot off the mplist and hashlist. It may remain on
+		 * freelist in INACTIVE state.
+		 */
+		nextdqp = dqp->MPL_NEXT;
+		nmisses += xfs_qm_dqpurge(dqp, flags);
+		dqp = nextdqp;
+	}
+	xfs_qm_mplist_unlock(mp);
+	return nmisses;
+}
+
+int
+xfs_qm_dqpurge_all(
+	xfs_mount_t	*mp,
+	uint		flags)
+{
+	int		ndquots;
+
+	/*
+	 * Purge the dquot cache.
+	 * None of the dquots should really be busy at this point.
+	 */
+	if (mp->m_quotainfo) {
+		while ((ndquots = xfs_qm_dqpurge_int(mp, flags))) {
+			delay(ndquots * 10);
+		}
+	}
+	return 0;
+}
+
+STATIC int
+xfs_qm_dqattach_one(
+	xfs_inode_t	*ip,
+	xfs_dqid_t	id,
+	uint		type,
+	uint		doalloc,
+	uint		dolock,
+	xfs_dquot_t	*udqhint, /* hint */
+	xfs_dquot_t	**IO_idqpp)
+{
+	xfs_dquot_t	*dqp;
+	int		error;
+
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+	error = 0;
+	/*
+	 * See if we already have it in the inode itself. IO_idqpp is
+	 * &i_udquot or &i_gdquot. This made the code look weird, but
+	 * made the logic a lot simpler.
+	 */
+	if ((dqp = *IO_idqpp)) {
+		if (dolock)
+			xfs_dqlock(dqp);
+		xfs_dqtrace_entry(dqp, "DQATTACH: found in ip");
+		goto done;
+	}
+
+	/*
+	 * udqhint is the i_udquot field in inode, and is non-NULL only
+	 * when the type arg is XFS_DQ_GROUP. Its purpose is to save a
+	 * lookup by dqid (xfs_qm_dqget) by caching a group dquot inside
+	 * the user dquot.
+	 */
+	ASSERT(!udqhint || type == XFS_DQ_GROUP);
+	if (udqhint && !dolock)
+		xfs_dqlock(udqhint);
+
+	/*
+	 * No need to take dqlock to look at the id.
+	 * The ID can't change until it gets reclaimed, and it won't
+	 * be reclaimed as long as we have a ref from inode and we hold
+	 * the ilock.
+	 */
+	if (udqhint &&
+	    (dqp = udqhint->q_gdquot) &&
+	    (INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id)) {
+		ASSERT(XFS_DQ_IS_LOCKED(udqhint));
+		xfs_dqlock(dqp);
+		XFS_DQHOLD(dqp);
+		ASSERT(*IO_idqpp == NULL);
+		*IO_idqpp = dqp;
+		if (!dolock) {
+			xfs_dqunlock(dqp);
+			xfs_dqunlock(udqhint);
+		}
+		goto done;
+	}
+	/*
+	 * We can't hold a dquot lock when we call the dqget code.
+	 * We'll deadlock in no time, because of (not conforming to)
+	 * lock ordering - the inodelock comes before any dquot lock,
+	 * and we may drop and reacquire the ilock in xfs_qm_dqget().
+	 */
+	if (udqhint)
+		xfs_dqunlock(udqhint);
+	/*
+	 * Find the dquot from somewhere. This bumps the
+	 * reference count of dquot and returns it locked.
+	 * This can return ENOENT if dquot didn't exist on
+	 * disk and we didn't ask it to allocate;
+	 * ESRCH if quotas got turned off suddenly.
+	 */
+	if ((error = xfs_qm_dqget(ip->i_mount, ip, id, type,
+				 doalloc|XFS_QMOPT_DOWARN, &dqp))) {
+		if (udqhint && dolock)
+			xfs_dqlock(udqhint);
+		goto done;
+	}
+
+	xfs_dqtrace_entry(dqp, "DQATTACH: found by dqget");
+	/*
+	 * dqget may have dropped and re-acquired the ilock, but it guarantees
+	 * that the dquot returned is the one that should go in the inode.
+	 */
+	*IO_idqpp = dqp;
+	ASSERT(dqp);
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	if (! dolock) {
+		xfs_dqunlock(dqp);
+		goto done;
+	}
+	if (! udqhint)
+		goto done;
+
+	ASSERT(udqhint);
+	ASSERT(dolock);
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	if (! xfs_qm_dqlock_nowait(udqhint)) {
+		xfs_dqunlock(dqp);
+		xfs_dqlock(udqhint);
+		xfs_dqlock(dqp);
+	}
+      done:
+#ifdef QUOTADEBUG
+	if (udqhint) {
+		if (dolock)
+			ASSERT(XFS_DQ_IS_LOCKED(udqhint));
+	}
+	if (! error) {
+		if (dolock)
+			ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	}
+#endif
+	return (error);
+}
+
+
+/*
+ * Given a udquot and gdquot, attach a ptr to the group dquot in the
+ * udquot as a hint for future lookups. The idea sounds simple, but the
+ * execution isn't, because the udquot might have a group dquot attached
+ * already and getting rid of that gets us into lock ordering contraints.
+ * The process is complicated more by the fact that the dquots may or may not
+ * be locked on entry.
+ */
+STATIC void
+xfs_qm_dqattach_grouphint(
+	xfs_dquot_t	*udq,
+	xfs_dquot_t	*gdq,
+	uint		locked)
+{
+	xfs_dquot_t	*tmp;
+
+#ifdef QUOTADEBUG
+	if (locked) {
+		ASSERT(XFS_DQ_IS_LOCKED(udq));
+		ASSERT(XFS_DQ_IS_LOCKED(gdq));
+	}
+#endif
+	if (! locked)
+		xfs_dqlock(udq);
+
+	if ((tmp = udq->q_gdquot)) {
+		if (tmp == gdq) {
+			if (! locked)
+				xfs_dqunlock(udq);
+			return;
+		}
+
+		udq->q_gdquot = NULL;
+		/*
+		 * We can't keep any dqlocks when calling dqrele,
+		 * because the freelist lock comes before dqlocks.
+		 */
+		xfs_dqunlock(udq);
+		if (locked)
+			xfs_dqunlock(gdq);
+		/*
+		 * we took a hard reference once upon a time in dqget,
+		 * so give it back when the udquot no longer points at it
+		 * dqput() does the unlocking of the dquot.
+		 */
+		xfs_qm_dqrele(tmp);
+
+		xfs_dqlock(udq);
+		xfs_dqlock(gdq);
+
+	} else {
+		ASSERT(XFS_DQ_IS_LOCKED(udq));
+		if (! locked) {
+			xfs_dqlock(gdq);
+		}
+	}
+
+	ASSERT(XFS_DQ_IS_LOCKED(udq));
+	ASSERT(XFS_DQ_IS_LOCKED(gdq));
+	/*
+	 * Somebody could have attached a gdquot here,
+	 * when we dropped the uqlock. If so, just do nothing.
+	 */
+	if (udq->q_gdquot == NULL) {
+		XFS_DQHOLD(gdq);
+		udq->q_gdquot = gdq;
+	}
+	if (! locked) {
+		xfs_dqunlock(gdq);
+		xfs_dqunlock(udq);
+	}
+}
+
+
+/*
+ * Given a locked inode, attach dquot(s) to it, taking UQUOTAON / GQUOTAON
+ * in to account.
+ * If XFS_QMOPT_DQALLOC, the dquot(s) will be allocated if needed.
+ * If XFS_QMOPT_DQLOCK, the dquot(s) will be returned locked. This option pretty
+ * much made this code a complete mess, but it has been pretty useful.
+ * If XFS_QMOPT_ILOCKED, then inode sent is already locked EXCL.
+ * Inode may get unlocked and relocked in here, and the caller must deal with
+ * the consequences.
+ */
+int
+xfs_qm_dqattach(
+	xfs_inode_t	*ip,
+	uint		flags)
+{
+	xfs_mount_t	*mp = ip->i_mount;
+	uint		nquotas = 0;
+	int		error = 0;
+
+	if ((! XFS_IS_QUOTA_ON(mp)) ||
+	    (! XFS_NOT_DQATTACHED(mp, ip)) ||
+	    (ip->i_ino == mp->m_sb.sb_uquotino) ||
+	    (ip->i_ino == mp->m_sb.sb_gquotino))
+		return (0);
+
+	ASSERT((flags & XFS_QMOPT_ILOCKED) == 0 ||
+	       XFS_ISLOCKED_INODE_EXCL(ip));
+
+	if (! (flags & XFS_QMOPT_ILOCKED))
+		xfs_ilock(ip, XFS_ILOCK_EXCL);
+
+	if (XFS_IS_UQUOTA_ON(mp)) {
+		error = xfs_qm_dqattach_one(ip, ip->i_d.di_uid, XFS_DQ_USER,
+						flags & XFS_QMOPT_DQALLOC,
+						flags & XFS_QMOPT_DQLOCK,
+						NULL, &ip->i_udquot);
+		if (error)
+			goto done;
+		nquotas++;
+	}
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+	if (XFS_IS_GQUOTA_ON(mp)) {
+		error = xfs_qm_dqattach_one(ip, ip->i_d.di_gid, XFS_DQ_GROUP,
+						flags & XFS_QMOPT_DQALLOC,
+						flags & XFS_QMOPT_DQLOCK,
+						ip->i_udquot, &ip->i_gdquot);
+		/*
+		 * Don't worry about the udquot that we may have
+		 * attached above. It'll get detached, if not already.
+		 */
+		if (error)
+			goto done;
+		nquotas++;
+	}
+
+	/*
+	 * Attach this group quota to the user quota as a hint.
+	 * This WON'T, in general, result in a thrash.
+	 */
+	if (nquotas == 2) {
+		ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+		ASSERT(ip->i_udquot);
+		ASSERT(ip->i_gdquot);
+
+		/*
+		 * We may or may not have the i_udquot locked at this point,
+		 * but this check is OK since we don't depend on the i_gdquot to
+		 * be accurate 100% all the time. It is just a hint, and this
+		 * will succeed in general.
+		 */
+		if (ip->i_udquot->q_gdquot == ip->i_gdquot)
+			goto done;
+		/*
+		 * Attach i_gdquot to the gdquot hint inside the i_udquot.
+		 */
+		xfs_qm_dqattach_grouphint(ip->i_udquot, ip->i_gdquot,
+					 flags & XFS_QMOPT_DQLOCK);
+	}
+
+      done:
+
+#ifdef QUOTADEBUG
+	if (! error) {
+		if (ip->i_udquot) {
+			if (flags & XFS_QMOPT_DQLOCK)
+				ASSERT(XFS_DQ_IS_LOCKED(ip->i_udquot));
+		}
+		if (ip->i_gdquot) {
+			if (flags & XFS_QMOPT_DQLOCK)
+				ASSERT(XFS_DQ_IS_LOCKED(ip->i_gdquot));
+		}
+		if (XFS_IS_UQUOTA_ON(mp))
+			ASSERT(ip->i_udquot);
+		if (XFS_IS_GQUOTA_ON(mp))
+			ASSERT(ip->i_gdquot);
+	}
+#endif
+
+	if (! (flags & XFS_QMOPT_ILOCKED))
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+
+#ifdef QUOTADEBUG
+	else
+		ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+#endif
+	return (error);
+}
+
+/*
+ * Release dquots (and their references) if any.
+ * The inode should be locked EXCL except when this's called by
+ * xfs_ireclaim.
+ */
+void
+xfs_qm_dqdetach(
+	xfs_inode_t	*ip)
+{
+	if (!(ip->i_udquot || ip->i_gdquot))
+		return;
+
+	ASSERT(ip->i_ino != ip->i_mount->m_sb.sb_uquotino);
+	ASSERT(ip->i_ino != ip->i_mount->m_sb.sb_gquotino);
+	if (ip->i_udquot)
+		xfs_dqtrace_entry_ino(ip->i_udquot, "DQDETTACH", ip);
+	if (ip->i_udquot) {
+		xfs_qm_dqrele(ip->i_udquot);
+		ip->i_udquot = NULL;
+	}
+	if (ip->i_gdquot) {
+		xfs_qm_dqrele(ip->i_gdquot);
+		ip->i_gdquot = NULL;
+	}
+}
+
+/*
+ * This is called by VFS_SYNC and flags arg determines the caller,
+ * and its motives, as done in xfs_sync.
+ *
+ * vfs_sync: SYNC_FSDATA|SYNC_ATTR|SYNC_BDFLUSH 0x31
+ * syscall sync: SYNC_FSDATA|SYNC_ATTR|SYNC_DELWRI 0x25
+ * umountroot : SYNC_WAIT | SYNC_CLOSE | SYNC_ATTR | SYNC_FSDATA
+ */
+
+int
+xfs_qm_sync(
+	xfs_mount_t	*mp,
+	short		flags)
+{
+	int		recl, restarts;
+	xfs_dquot_t	*dqp;
+	uint		flush_flags;
+	boolean_t	nowait;
+	int		error;
+
+	restarts = 0;
+	/*
+	 * We won't block unless we are asked to.
+	 */
+	nowait = (boolean_t)(flags & SYNC_BDFLUSH || (flags & SYNC_WAIT) == 0);
+
+  again:
+	xfs_qm_mplist_lock(mp);
+	/*
+	 * dqpurge_all() also takes the mplist lock and iterate thru all dquots
+	 * in quotaoff. However, if the QUOTA_ACTIVE bits are not cleared
+	 * when we have the mplist lock, we know that dquots will be consistent
+	 * as long as we have it locked.
+	 */
+	if (! XFS_IS_QUOTA_ON(mp)) {
+		xfs_qm_mplist_unlock(mp);
+		return (0);
+	}
+	FOREACH_DQUOT_IN_MP(dqp, mp) {
+		/*
+		 * If this is vfs_sync calling, then skip the dquots that
+		 * don't 'seem' to be dirty. ie. don't acquire dqlock.
+		 * This is very similar to what xfs_sync does with inodes.
+		 */
+		if (flags & SYNC_BDFLUSH) {
+			if (! XFS_DQ_IS_DIRTY(dqp))
+				continue;
+		}
+
+		if (nowait) {
+			/*
+			 * Try to acquire the dquot lock. We are NOT out of
+			 * lock order, but we just don't want to wait for this
+			 * lock, unless somebody wanted us to.
+			 */
+			if (! xfs_qm_dqlock_nowait(dqp))
+				continue;
+		} else {
+			xfs_dqlock(dqp);
+		}
+
+		/*
+		 * Now, find out for sure if this dquot is dirty or not.
+		 */
+		if (! XFS_DQ_IS_DIRTY(dqp)) {
+			xfs_dqunlock(dqp);
+			continue;
+		}
+
+		/* XXX a sentinel would be better */
+		recl = XFS_QI_MPLRECLAIMS(mp);
+		if (! xfs_qm_dqflock_nowait(dqp)) {
+			if (nowait) {
+				xfs_dqunlock(dqp);
+				continue;
+			}
+			/*
+			 * If we can't grab the flush lock then if the caller
+			 * really wanted us to give this our best shot,
+			 * see if we can give a push to the buffer before we wait
+			 * on the flush lock. At this point, we know that
+			 * eventhough the dquot is being flushed,
+			 * it has (new) dirty data.
+			 */
+			xfs_qm_dqflock_pushbuf_wait(dqp);
+		}
+		/*
+		 * Let go of the mplist lock. We don't want to hold it
+		 * across a disk write
+		 */
+		flush_flags = (nowait) ? XFS_QMOPT_DELWRI : XFS_QMOPT_SYNC;
+		xfs_qm_mplist_unlock(mp);
+		xfs_dqtrace_entry(dqp, "XQM_SYNC: DQFLUSH");
+		error = xfs_qm_dqflush(dqp, flush_flags);
+		xfs_dqunlock(dqp);
+		if (error && XFS_FORCED_SHUTDOWN(mp))
+			return(0);	/* Need to prevent umount failure */
+		else if (error)
+			return (error);
+
+		xfs_qm_mplist_lock(mp);
+		if (recl != XFS_QI_MPLRECLAIMS(mp)) {
+			if (++restarts >= XFS_QM_SYNC_MAX_RESTARTS)
+				break;
+
+			xfs_qm_mplist_unlock(mp);
+			goto again;
+		}
+	}
+
+	xfs_qm_mplist_unlock(mp);
+	return (0);
+}
+
+
+/*
+ * This initializes all the quota information that's kept in the
+ * mount structure
+ */
+int
+xfs_qm_init_quotainfo(
+	xfs_mount_t	*mp)
+{
+	xfs_quotainfo_t *qinf;
+	int		error;
+	xfs_dquot_t	*dqp;
+
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+
+	/*
+	 * Tell XQM that we exist as soon as possible.
+	 */
+	if ((error = xfs_qm_hold_quotafs_ref(mp))) {
+		return (error);
+	}
+
+	qinf = mp->m_quotainfo = kmem_zalloc(sizeof(xfs_quotainfo_t), KM_SLEEP);
+
+	/*
+	 * See if quotainodes are setup, and if not, allocate them,
+	 * and change the superblock accordingly.
+	 */
+	if ((error = xfs_qm_init_quotainos(mp))) {
+		kmem_free(qinf, sizeof(xfs_quotainfo_t));
+		mp->m_quotainfo = NULL;
+		return (error);
+	}
+
+	spinlock_init(&qinf->qi_pinlock, "xfs_qinf_pin");
+	xfs_qm_list_init(&qinf->qi_dqlist, "mpdqlist", 0);
+	qinf->qi_dqreclaims = 0;
+
+	/* mutex used to serialize quotaoffs */
+	mutex_init(&qinf->qi_quotaofflock, MUTEX_DEFAULT, "qoff");
+
+	/* Precalc some constants */
+	qinf->qi_dqchunklen = XFS_FSB_TO_BB(mp, XFS_DQUOT_CLUSTER_SIZE_FSB);
+	ASSERT(qinf->qi_dqchunklen);
+	qinf->qi_dqperchunk = BBTOB(qinf->qi_dqchunklen);
+	do_div(qinf->qi_dqperchunk, sizeof(xfs_dqblk_t));
+
+	mp->m_qflags |= (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_CHKD);
+
+	/*
+	 * We try to get the limits from the superuser's limits fields.
+	 * This is quite hacky, but it is standard quota practice.
+	 * We look at the USR dquot with id == 0 first, but if user quotas
+	 * are not enabled we goto the GRP dquot with id == 0.
+	 * We don't really care to keep separate default limits for user
+	 * and group quotas, at least not at this point.
+	 */
+	error = xfs_qm_dqget(mp, NULL, (xfs_dqid_t)0,
+			     (XFS_IS_UQUOTA_RUNNING(mp)) ?
+			     XFS_DQ_USER : XFS_DQ_GROUP,
+			     XFS_QMOPT_DQSUSER|XFS_QMOPT_DOWARN,
+			     &dqp);
+	if (! error) {
+		xfs_disk_dquot_t	*ddqp = &dqp->q_core;
+
+		/*
+		 * The warnings and timers set the grace period given to
+		 * a user or group before he or she can not perform any
+		 * more writing. If it is zero, a default is used.
+		 */
+		qinf->qi_btimelimit =
+				INT_GET(ddqp->d_btimer, ARCH_CONVERT) ?
+				INT_GET(ddqp->d_btimer, ARCH_CONVERT) :
+				XFS_QM_BTIMELIMIT;
+		qinf->qi_itimelimit =
+				INT_GET(ddqp->d_itimer, ARCH_CONVERT) ?
+				INT_GET(ddqp->d_itimer, ARCH_CONVERT) :
+				XFS_QM_ITIMELIMIT;
+		qinf->qi_rtbtimelimit =
+				INT_GET(ddqp->d_rtbtimer, ARCH_CONVERT) ?
+				INT_GET(ddqp->d_rtbtimer, ARCH_CONVERT) :
+				XFS_QM_RTBTIMELIMIT;
+		qinf->qi_bwarnlimit =
+				INT_GET(ddqp->d_bwarns, ARCH_CONVERT) ?
+				INT_GET(ddqp->d_bwarns, ARCH_CONVERT) :
+				XFS_QM_BWARNLIMIT;
+		qinf->qi_iwarnlimit =
+				INT_GET(ddqp->d_iwarns, ARCH_CONVERT) ?
+				INT_GET(ddqp->d_iwarns, ARCH_CONVERT) :
+				XFS_QM_IWARNLIMIT;
+		qinf->qi_bhardlimit =
+				INT_GET(ddqp->d_blk_hardlimit, ARCH_CONVERT);
+		qinf->qi_bsoftlimit =
+				INT_GET(ddqp->d_blk_softlimit, ARCH_CONVERT);
+		qinf->qi_ihardlimit =
+				INT_GET(ddqp->d_ino_hardlimit, ARCH_CONVERT);
+		qinf->qi_isoftlimit =
+				INT_GET(ddqp->d_ino_softlimit, ARCH_CONVERT);
+		qinf->qi_rtbhardlimit =
+				INT_GET(ddqp->d_rtb_hardlimit, ARCH_CONVERT);
+		qinf->qi_rtbsoftlimit =
+				INT_GET(ddqp->d_rtb_softlimit, ARCH_CONVERT);
+ 
+		/*
+		 * We sent the XFS_QMOPT_DQSUSER flag to dqget because
+		 * we don't want this dquot cached. We haven't done a
+		 * quotacheck yet, and quotacheck doesn't like incore dquots.
+		 */
+		xfs_qm_dqdestroy(dqp);
+	} else {
+		qinf->qi_btimelimit = XFS_QM_BTIMELIMIT;
+		qinf->qi_itimelimit = XFS_QM_ITIMELIMIT;
+		qinf->qi_rtbtimelimit = XFS_QM_RTBTIMELIMIT;
+		qinf->qi_bwarnlimit = XFS_QM_BWARNLIMIT;
+		qinf->qi_iwarnlimit = XFS_QM_IWARNLIMIT;
+	}
+
+	return (0);
+}
+
+
+/*
+ * Gets called when unmounting a filesystem or when all quotas get
+ * turned off.
+ * This purges the quota inodes, destroys locks and frees itself.
+ */
+void
+xfs_qm_destroy_quotainfo(
+	xfs_mount_t	*mp)
+{
+	xfs_quotainfo_t *qi;
+
+	qi = mp->m_quotainfo;
+	ASSERT(qi != NULL);
+	ASSERT(xfs_Gqm != NULL);
+
+	/*
+	 * Release the reference that XQM kept, so that we know
+	 * when the XQM structure should be freed. We cannot assume
+	 * that xfs_Gqm is non-null after this point.
+	 */
+	xfs_qm_rele_quotafs_ref(mp);
+
+	spinlock_destroy(&qi->qi_pinlock);
+	xfs_qm_list_destroy(&qi->qi_dqlist);
+
+	if (qi->qi_uquotaip) {
+		XFS_PURGE_INODE(qi->qi_uquotaip);
+		qi->qi_uquotaip = NULL; /* paranoia */
+	}
+	if (qi->qi_gquotaip) {
+		XFS_PURGE_INODE(qi->qi_gquotaip);
+		qi->qi_gquotaip = NULL;
+	}
+	mutex_destroy(&qi->qi_quotaofflock);
+	kmem_free(qi, sizeof(xfs_quotainfo_t));
+	mp->m_quotainfo = NULL;
+}
+
+
+
+/* ------------------- PRIVATE STATIC FUNCTIONS ----------------------- */
+
+/* ARGSUSED */
+STATIC void
+xfs_qm_list_init(
+	xfs_dqlist_t	*list,
+	char		*str,
+	int		n)
+{
+	mutex_init(&list->qh_lock, MUTEX_DEFAULT, str);
+	list->qh_next = NULL;
+	list->qh_version = 0;
+	list->qh_nelems = 0;
+}
+
+STATIC void
+xfs_qm_list_destroy(
+	xfs_dqlist_t	*list)
+{
+	mutex_destroy(&(list->qh_lock));
+}
+
+
+/*
+ * Stripped down version of dqattach. This doesn't attach, or even look at the
+ * dquots attached to the inode. The rationale is that there won't be any
+ * attached at the time this is called from quotacheck.
+ */
+STATIC int
+xfs_qm_dqget_noattach(
+	xfs_inode_t	*ip,
+	xfs_dquot_t	**O_udqpp,
+	xfs_dquot_t	**O_gdqpp)
+{
+	int		error;
+	xfs_mount_t	*mp;
+	xfs_dquot_t	*udqp, *gdqp;
+
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+	mp = ip->i_mount;
+	udqp = NULL;
+	gdqp = NULL;
+
+	if (XFS_IS_UQUOTA_ON(mp)) {
+		ASSERT(ip->i_udquot == NULL);
+		/*
+		 * We want the dquot allocated if it doesn't exist.
+		 */
+		if ((error = xfs_qm_dqget(mp, ip, ip->i_d.di_uid, XFS_DQ_USER,
+					 XFS_QMOPT_DQALLOC | XFS_QMOPT_DOWARN,
+					 &udqp))) {
+			/*
+			 * Shouldn't be able to turn off quotas here.
+			 */
+			ASSERT(error != ESRCH);
+			ASSERT(error != ENOENT);
+			return (error);
+		}
+		ASSERT(udqp);
+	}
+
+	if (XFS_IS_GQUOTA_ON(mp)) {
+		ASSERT(ip->i_gdquot == NULL);
+		if (udqp)
+			xfs_dqunlock(udqp);
+		if ((error = xfs_qm_dqget(mp, ip, ip->i_d.di_gid, XFS_DQ_GROUP,
+					 XFS_QMOPT_DQALLOC|XFS_QMOPT_DOWARN,
+					 &gdqp))) {
+			if (udqp)
+				xfs_qm_dqrele(udqp);
+			ASSERT(error != ESRCH);
+			ASSERT(error != ENOENT);
+			return (error);
+		}
+		ASSERT(gdqp);
+
+		/* Reacquire the locks in the right order */
+		if (udqp) {
+			if (! xfs_qm_dqlock_nowait(udqp)) {
+				xfs_dqunlock(gdqp);
+				xfs_dqlock(udqp);
+				xfs_dqlock(gdqp);
+			}
+		}
+	}
+
+	*O_udqpp = udqp;
+	*O_gdqpp = gdqp;
+
+#ifdef QUOTADEBUG
+	if (udqp) ASSERT(XFS_DQ_IS_LOCKED(udqp));
+	if (gdqp) ASSERT(XFS_DQ_IS_LOCKED(gdqp));
+#endif
+	return (0);
+}
+
+/*
+ * Create an inode and return with a reference already taken, but unlocked
+ * This is how we create quota inodes
+ */
+STATIC int
+xfs_qm_qino_alloc(
+	xfs_mount_t	*mp,
+	xfs_inode_t	**ip,
+	__int64_t	sbfields,
+	uint		flags)
+{
+	xfs_trans_t	*tp;
+	int		error;
+	unsigned long s;
+	cred_t		zerocr;
+	int		committed;
+
+	tp = xfs_trans_alloc(mp,XFS_TRANS_QM_QINOCREATE);
+	if ((error = xfs_trans_reserve(tp,
+				      XFS_QM_QINOCREATE_SPACE_RES(mp),
+				      XFS_CREATE_LOG_RES(mp), 0,
+				      XFS_TRANS_PERM_LOG_RES,
+				      XFS_CREATE_LOG_COUNT))) {
+		xfs_trans_cancel(tp, 0);
+		return (error);
+	}
+	memset(&zerocr, 0, sizeof(zerocr));
+
+	if ((error = xfs_dir_ialloc(&tp, mp->m_rootip, S_IFREG, 1, 0,
+				   &zerocr, 0, 1, ip, &committed))) {
+		xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES |
+				 XFS_TRANS_ABORT);
+		return (error);
+	}
+
+	/*
+	 * Keep an extra reference to this quota inode. This inode is
+	 * locked exclusively and joined to the transaction already.
+	 */
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(*ip));
+	VN_HOLD(XFS_ITOV((*ip)));
+
+	/*
+	 * Make the changes in the superblock, and log those too.
+	 * sbfields arg may contain fields other than *QUOTINO;
+	 * VERSIONNUM for example.
+	 */
+	s = XFS_SB_LOCK(mp);
+	if (flags & XFS_QMOPT_SBVERSION) {
+#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
+		unsigned oldv = mp->m_sb.sb_versionnum;
+#endif
+		ASSERT(!XFS_SB_VERSION_HASQUOTA(&mp->m_sb));
+		ASSERT((sbfields & (XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO |
+				   XFS_SB_GQUOTINO | XFS_SB_QFLAGS)) ==
+		       (XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO |
+			XFS_SB_GQUOTINO | XFS_SB_QFLAGS));
+
+		XFS_SB_VERSION_ADDQUOTA(&mp->m_sb);
+		mp->m_sb.sb_uquotino = NULLFSINO;
+		mp->m_sb.sb_gquotino = NULLFSINO;
+
+		/* qflags will get updated _after_ quotacheck */
+		mp->m_sb.sb_qflags = 0;
+#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
+		cmn_err(CE_NOTE,
+			"Old superblock version %x, converting to %x.",
+			oldv, mp->m_sb.sb_versionnum);
+#endif
+	}
+	if (flags & XFS_QMOPT_UQUOTA)
+		mp->m_sb.sb_uquotino = (*ip)->i_ino;
+	else
+		mp->m_sb.sb_gquotino = (*ip)->i_ino;
+	XFS_SB_UNLOCK(mp, s);
+	xfs_mod_sb(tp, sbfields);
+
+	if ((error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES,
+				     NULL))) {
+		xfs_fs_cmn_err(CE_ALERT, mp, "XFS qino_alloc failed!");
+		return (error);
+	}
+	return (0);
+}
+
+
+STATIC int
+xfs_qm_reset_dqcounts(
+	xfs_mount_t	*mp,
+	xfs_buf_t	*bp,
+	xfs_dqid_t	id,
+	uint		type)
+{
+	xfs_disk_dquot_t	*ddq;
+	int			j;
+
+	xfs_buftrace("RESET DQUOTS", bp);
+	/*
+	 * Reset all counters and timers. They'll be
+	 * started afresh by xfs_qm_quotacheck.
+	 */
+#ifdef DEBUG
+	j = XFS_FSB_TO_B(mp, XFS_DQUOT_CLUSTER_SIZE_FSB);
+	do_div(j, sizeof(xfs_dqblk_t));
+	ASSERT(XFS_QM_DQPERBLK(mp) == j);
+#endif
+	ddq = (xfs_disk_dquot_t *)XFS_BUF_PTR(bp);
+	for (j = 0; j < XFS_QM_DQPERBLK(mp); j++) {
+		/*
+		 * Do a sanity check, and if needed, repair the dqblk. Don't
+		 * output any warnings because it's perfectly possible to
+		 * find unitialized dquot blks. See comment in xfs_qm_dqcheck.
+		 */
+		(void) xfs_qm_dqcheck(ddq, id+j, type, XFS_QMOPT_DQREPAIR,
+				      "xfs_quotacheck");
+		INT_SET(ddq->d_bcount, ARCH_CONVERT, 0ULL);
+		INT_SET(ddq->d_icount, ARCH_CONVERT, 0ULL);
+		INT_SET(ddq->d_rtbcount, ARCH_CONVERT, 0ULL);
+		INT_SET(ddq->d_btimer, ARCH_CONVERT, (time_t)0);
+		INT_SET(ddq->d_itimer, ARCH_CONVERT, (time_t)0);
+		INT_SET(ddq->d_bwarns, ARCH_CONVERT, 0UL);
+		INT_SET(ddq->d_iwarns, ARCH_CONVERT, 0UL);
+		ddq = (xfs_disk_dquot_t *) ((xfs_dqblk_t *)ddq + 1);
+	}
+
+	return (0);
+}
+
+STATIC int
+xfs_qm_dqiter_bufs(
+	xfs_mount_t	*mp,
+	xfs_dqid_t	firstid,
+	xfs_fsblock_t	bno,
+	xfs_filblks_t	blkcnt,
+	uint		flags)
+{
+	xfs_buf_t	*bp;
+	int		error;
+	int		notcommitted;
+	int		incr;
+
+	ASSERT(blkcnt > 0);
+	notcommitted = 0;
+	incr = (blkcnt > XFS_QM_MAX_DQCLUSTER_LOGSZ) ?
+		XFS_QM_MAX_DQCLUSTER_LOGSZ : blkcnt;
+	error = 0;
+
+	/*
+	 * Blkcnt arg can be a very big number, and might even be
+	 * larger than the log itself. So, we have to break it up into
+	 * manageable-sized transactions.
+	 * Note that we don't start a permanent transaction here; we might
+	 * not be able to get a log reservation for the whole thing up front,
+	 * and we don't really care to either, because we just discard
+	 * everything if we were to crash in the middle of this loop.
+	 */
+	while (blkcnt--) {
+		error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
+			      XFS_FSB_TO_DADDR(mp, bno),
+			      (int)XFS_QI_DQCHUNKLEN(mp), 0, &bp);
+		if (error)
+			break;
+
+		(void) xfs_qm_reset_dqcounts(mp, bp, firstid,
+					     flags & XFS_QMOPT_UQUOTA ?
+					     XFS_DQ_USER : XFS_DQ_GROUP);
+		xfs_bdwrite(mp, bp);
+		/*
+		 * goto the next block.
+		 */
+		bno++;
+		firstid += XFS_QM_DQPERBLK(mp);
+	}
+	return (error);
+}
+
+/*
+ * Iterate over all allocated USR/GRP dquots in the system, calling a
+ * caller supplied function for every chunk of dquots that we find.
+ */
+STATIC int
+xfs_qm_dqiterate(
+	xfs_mount_t	*mp,
+	xfs_inode_t	*qip,
+	uint		flags)
+{
+	xfs_bmbt_irec_t		*map;
+	int			i, nmaps;	/* number of map entries */
+	int			error;		/* return value */
+	xfs_fileoff_t		lblkno;
+	xfs_filblks_t		maxlblkcnt;
+	xfs_dqid_t		firstid;
+	xfs_fsblock_t		rablkno;
+	xfs_filblks_t		rablkcnt;
+
+	error = 0;
+	/*
+	 * This looks racey, but we can't keep an inode lock across a
+	 * trans_reserve. But, this gets called during quotacheck, and that
+	 * happens only at mount time which is single threaded.
+	 */
+	if (qip->i_d.di_nblocks == 0)
+		return (0);
+
+	map = kmem_alloc(XFS_DQITER_MAP_SIZE * sizeof(*map), KM_SLEEP);
+
+	lblkno = 0;
+	maxlblkcnt = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
+	do {
+		nmaps = XFS_DQITER_MAP_SIZE;
+		/*
+		 * We aren't changing the inode itself. Just changing
+		 * some of its data. No new blocks are added here, and
+		 * the inode is never added to the transaction.
+		 */
+		xfs_ilock(qip, XFS_ILOCK_SHARED);
+		error = xfs_bmapi(NULL, qip, lblkno,
+				  maxlblkcnt - lblkno,
+				  XFS_BMAPI_METADATA,
+				  NULL,
+				  0, map, &nmaps, NULL);
+		xfs_iunlock(qip, XFS_ILOCK_SHARED);
+		if (error)
+			break;
+
+		ASSERT(nmaps <= XFS_DQITER_MAP_SIZE);
+		for (i = 0; i < nmaps; i++) {
+			ASSERT(map[i].br_startblock != DELAYSTARTBLOCK);
+			ASSERT(map[i].br_blockcount);
+
+
+			lblkno += map[i].br_blockcount;
+
+			if (map[i].br_startblock == HOLESTARTBLOCK)
+				continue;
+
+			firstid = (xfs_dqid_t) map[i].br_startoff *
+				XFS_QM_DQPERBLK(mp);
+			/*
+			 * Do a read-ahead on the next extent.
+			 */
+			if ((i+1 < nmaps) &&
+			    (map[i+1].br_startblock != HOLESTARTBLOCK)) {
+				rablkcnt =  map[i+1].br_blockcount;
+				rablkno = map[i+1].br_startblock;
+				while (rablkcnt--) {
+					xfs_baread(mp->m_ddev_targp,
+					       XFS_FSB_TO_DADDR(mp, rablkno),
+					       (int)XFS_QI_DQCHUNKLEN(mp));
+					rablkno++;
+				}
+			}
+			/*
+			 * Iterate thru all the blks in the extent and
+			 * reset the counters of all the dquots inside them.
+			 */
+			if ((error = xfs_qm_dqiter_bufs(mp,
+						       firstid,
+						       map[i].br_startblock,
+						       map[i].br_blockcount,
+						       flags))) {
+				break;
+			}
+		}
+
+		if (error)
+			break;
+	} while (nmaps > 0);
+
+	kmem_free(map, XFS_DQITER_MAP_SIZE * sizeof(*map));
+
+	return (error);
+}
+
+/*
+ * Called by dqusage_adjust in doing a quotacheck.
+ * Given the inode, and a dquot (either USR or GRP, doesn't matter),
+ * this updates its incore copy as well as the buffer copy. This is
+ * so that once the quotacheck is done, we can just log all the buffers,
+ * as opposed to logging numerous updates to individual dquots.
+ */
+STATIC void
+xfs_qm_quotacheck_dqadjust(
+	xfs_dquot_t		*dqp,
+	xfs_qcnt_t		nblks,
+	xfs_qcnt_t		rtblks)
+{
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	xfs_dqtrace_entry(dqp, "QCHECK DQADJUST");
+	/*
+	 * Adjust the inode count and the block count to reflect this inode's
+	 * resource usage.
+	 */
+	INT_MOD(dqp->q_core.d_icount, ARCH_CONVERT, +1);
+	dqp->q_res_icount++;
+	if (nblks) {
+		INT_MOD(dqp->q_core.d_bcount, ARCH_CONVERT, nblks);
+		dqp->q_res_bcount += nblks;
+	}
+	if (rtblks) {
+		INT_MOD(dqp->q_core.d_rtbcount, ARCH_CONVERT, rtblks);
+		dqp->q_res_rtbcount += rtblks;
+	}
+
+	/*
+	 * Set default limits, adjust timers (since we changed usages)
+	 */
+	if (! XFS_IS_SUSER_DQUOT(dqp)) {
+		xfs_qm_adjust_dqlimits(dqp->q_mount, &dqp->q_core);
+		xfs_qm_adjust_dqtimers(dqp->q_mount, &dqp->q_core);
+	}
+
+	dqp->dq_flags |= XFS_DQ_DIRTY;
+}
+
+STATIC int
+xfs_qm_get_rtblks(
+	xfs_inode_t	*ip,
+	xfs_qcnt_t	*O_rtblks)
+{
+	xfs_filblks_t	rtblks;			/* total rt blks */
+	xfs_ifork_t	*ifp;			/* inode fork pointer */
+	xfs_extnum_t	nextents;		/* number of extent entries */
+	xfs_bmbt_rec_t	*base;			/* base of extent array */
+	xfs_bmbt_rec_t	*ep;			/* pointer to an extent entry */
+	int		error;
+
+	ASSERT(XFS_IS_REALTIME_INODE(ip));
+	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+	if (!(ifp->if_flags & XFS_IFEXTENTS)) {
+		if ((error = xfs_iread_extents(NULL, ip, XFS_DATA_FORK)))
+			return (error);
+	}
+	rtblks = 0;
+	nextents = ifp->if_bytes / sizeof(xfs_bmbt_rec_t);
+	base = &ifp->if_u1.if_extents[0];
+	for (ep = base; ep < &base[nextents]; ep++)
+		rtblks += xfs_bmbt_get_blockcount(ep);
+	*O_rtblks = (xfs_qcnt_t)rtblks;
+	return (0);
+}
+
+/*
+ * callback routine supplied to bulkstat(). Given an inumber, find its
+ * dquots and update them to account for resources taken by that inode.
+ */
+/* ARGSUSED */
+STATIC int
+xfs_qm_dqusage_adjust(
+	xfs_mount_t	*mp,		/* mount point for filesystem */
+	xfs_ino_t	ino,		/* inode number to get data for */
+	void		__user *buffer,	/* not used */
+	int		ubsize,		/* not used */
+	void		*private_data,	/* not used */
+	xfs_daddr_t	bno,		/* starting block of inode cluster */
+	int		*ubused,	/* not used */
+	void		*dip,		/* on-disk inode pointer (not used) */
+	int		*res)		/* result code value */
+{
+	xfs_inode_t	*ip;
+	xfs_dquot_t	*udqp, *gdqp;
+	xfs_qcnt_t	nblks, rtblks;
+	int		error;
+
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+
+	/*
+	 * rootino must have its resources accounted for, not so with the quota
+	 * inodes.
+	 */
+	if (ino == mp->m_sb.sb_uquotino || ino == mp->m_sb.sb_gquotino) {
+		*res = BULKSTAT_RV_NOTHING;
+		return XFS_ERROR(EINVAL);
+	}
+
+	/*
+	 * We don't _need_ to take the ilock EXCL. However, the xfs_qm_dqget
+	 * interface expects the inode to be exclusively locked because that's
+	 * the case in all other instances. It's OK that we do this because
+	 * quotacheck is done only at mount time.
+	 */
+	if ((error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_EXCL, &ip, bno))) {
+		*res = BULKSTAT_RV_NOTHING;
+		return (error);
+	}
+
+	if (ip->i_d.di_mode == 0) {
+		xfs_iput_new(ip, XFS_ILOCK_EXCL);
+		*res = BULKSTAT_RV_NOTHING;
+		return XFS_ERROR(ENOENT);
+	}
+
+	/*
+	 * Obtain the locked dquots. In case of an error (eg. allocation
+	 * fails for ENOSPC), we return the negative of the error number
+	 * to bulkstat, so that it can get propagated to quotacheck() and
+	 * making us disable quotas for the file system.
+	 */
+	if ((error = xfs_qm_dqget_noattach(ip, &udqp, &gdqp))) {
+		xfs_iput(ip, XFS_ILOCK_EXCL);
+		*res = BULKSTAT_RV_GIVEUP;
+		return (error);
+	}
+
+	rtblks = 0;
+	if (! XFS_IS_REALTIME_INODE(ip)) {
+		nblks = (xfs_qcnt_t)ip->i_d.di_nblocks;
+	} else {
+		/*
+		 * Walk thru the extent list and count the realtime blocks.
+		 */
+		if ((error = xfs_qm_get_rtblks(ip, &rtblks))) {
+			xfs_iput(ip, XFS_ILOCK_EXCL);
+			if (udqp)
+				xfs_qm_dqput(udqp);
+			if (gdqp)
+				xfs_qm_dqput(gdqp);
+			*res = BULKSTAT_RV_GIVEUP;
+			return (error);
+		}
+		nblks = (xfs_qcnt_t)ip->i_d.di_nblocks - rtblks;
+	}
+	ASSERT(ip->i_delayed_blks == 0);
+
+	/*
+	 * We can't release the inode while holding its dquot locks.
+	 * The inode can go into inactive and might try to acquire the dquotlocks.
+	 * So, just unlock here and do a vn_rele at the end.
+	 */
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+
+	/*
+	 * Add the (disk blocks and inode) resources occupied by this
+	 * inode to its dquots. We do this adjustment in the incore dquot,
+	 * and also copy the changes to its buffer.
+	 * We don't care about putting these changes in a transaction
+	 * envelope because if we crash in the middle of a 'quotacheck'
+	 * we have to start from the beginning anyway.
+	 * Once we're done, we'll log all the dquot bufs.
+	 *
+	 * The *QUOTA_ON checks below may look pretty racey, but quotachecks
+	 * and quotaoffs don't race. (Quotachecks happen at mount time only).
+	 */
+	if (XFS_IS_UQUOTA_ON(mp)) {
+		ASSERT(udqp);
+		xfs_qm_quotacheck_dqadjust(udqp, nblks, rtblks);
+		xfs_qm_dqput(udqp);
+	}
+	if (XFS_IS_GQUOTA_ON(mp)) {
+		ASSERT(gdqp);
+		xfs_qm_quotacheck_dqadjust(gdqp, nblks, rtblks);
+		xfs_qm_dqput(gdqp);
+	}
+	/*
+	 * Now release the inode. This will send it to 'inactive', and
+	 * possibly even free blocks.
+	 */
+	VN_RELE(XFS_ITOV(ip));
+
+	/*
+	 * Goto next inode.
+	 */
+	*res = BULKSTAT_RV_DIDONE;
+	return (0);
+}
+
+/*
+ * Walk thru all the filesystem inodes and construct a consistent view
+ * of the disk quota world. If the quotacheck fails, disable quotas.
+ */
+int
+xfs_qm_quotacheck(
+	xfs_mount_t	*mp)
+{
+	int		done, count, error;
+	xfs_ino_t	lastino;
+	size_t		structsz;
+	xfs_inode_t	*uip, *gip;
+	uint		flags;
+
+	count = INT_MAX;
+	structsz = 1;
+	lastino = 0;
+	flags = 0;
+
+	ASSERT(XFS_QI_UQIP(mp) || XFS_QI_GQIP(mp));
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+
+	/*
+	 * There should be no cached dquots. The (simplistic) quotacheck
+	 * algorithm doesn't like that.
+	 */
+	ASSERT(XFS_QI_MPLNDQUOTS(mp) == 0);
+
+	cmn_err(CE_NOTE, "XFS quotacheck %s: Please wait.", mp->m_fsname);
+
+	/*
+	 * First we go thru all the dquots on disk, USR and GRP, and reset
+	 * their counters to zero. We need a clean slate.
+	 * We don't log our changes till later.
+	 */
+	if ((uip = XFS_QI_UQIP(mp))) {
+		if ((error = xfs_qm_dqiterate(mp, uip, XFS_QMOPT_UQUOTA)))
+			goto error_return;
+		flags |= XFS_UQUOTA_CHKD;
+	}
+
+	if ((gip = XFS_QI_GQIP(mp))) {
+		if ((error = xfs_qm_dqiterate(mp, gip, XFS_QMOPT_GQUOTA)))
+			goto error_return;
+		flags |= XFS_GQUOTA_CHKD;
+	}
+
+	do {
+		/*
+		 * Iterate thru all the inodes in the file system,
+		 * adjusting the corresponding dquot counters in core.
+		 */
+		if ((error = xfs_bulkstat(mp, &lastino, &count,
+				     xfs_qm_dqusage_adjust, NULL,
+				     structsz, NULL,
+				     BULKSTAT_FG_IGET|BULKSTAT_FG_VFSLOCKED,
+				     &done)))
+			break;
+
+	} while (! done);
+
+	/*
+	 * We can get this error if we couldn't do a dquot allocation inside
+	 * xfs_qm_dqusage_adjust (via bulkstat). We don't care about the
+	 * dirty dquots that might be cached, we just want to get rid of them
+	 * and turn quotaoff. The dquots won't be attached to any of the inodes
+	 * at this point (because we intentionally didn't in dqget_noattach).
+	 */
+	if (error) {
+		xfs_qm_dqpurge_all(mp,
+				   XFS_QMOPT_UQUOTA|XFS_QMOPT_GQUOTA|
+				   XFS_QMOPT_QUOTAOFF);
+		goto error_return;
+	}
+	/*
+	 * We've made all the changes that we need to make incore.
+	 * Now flush_them down to disk buffers.
+	 */
+	xfs_qm_dqflush_all(mp, XFS_QMOPT_DELWRI);
+
+	/*
+	 * We didn't log anything, because if we crashed, we'll have to
+	 * start the quotacheck from scratch anyway. However, we must make
+	 * sure that our dquot changes are secure before we put the
+	 * quotacheck'd stamp on the superblock. So, here we do a synchronous
+	 * flush.
+	 */
+	XFS_bflush(mp->m_ddev_targp);
+
+	/*
+	 * If one type of quotas is off, then it will lose its
+	 * quotachecked status, since we won't be doing accounting for
+	 * that type anymore.
+	 */
+	mp->m_qflags &= ~(XFS_GQUOTA_CHKD | XFS_UQUOTA_CHKD);
+	mp->m_qflags |= flags;
+
+	XQM_LIST_PRINT(&(XFS_QI_MPL_LIST(mp)), MPL_NEXT, "++++ Mp list +++");
+
+ error_return:
+	if (error) {
+		cmn_err(CE_WARN, "XFS quotacheck %s: Unsuccessful (Error %d): "
+			"Disabling quotas.",
+			mp->m_fsname, error);
+		/*
+		 * We must turn off quotas.
+		 */
+		ASSERT(mp->m_quotainfo != NULL);
+		ASSERT(xfs_Gqm != NULL);
+		xfs_qm_destroy_quotainfo(mp);
+		xfs_mount_reset_sbqflags(mp);
+	} else {
+		cmn_err(CE_NOTE, "XFS quotacheck %s: Done.", mp->m_fsname);
+	}
+	return (error);
+}
+
+/*
+ * This is called after the superblock has been read in and we're ready to
+ * iget the quota inodes.
+ */
+STATIC int
+xfs_qm_init_quotainos(
+	xfs_mount_t	*mp)
+{
+	xfs_inode_t	*uip, *gip;
+	int		error;
+	__int64_t	sbflags;
+	uint		flags;
+
+	ASSERT(mp->m_quotainfo);
+	uip = gip = NULL;
+	sbflags = 0;
+	flags = 0;
+
+	/*
+	 * Get the uquota and gquota inodes
+	 */
+	if (XFS_SB_VERSION_HASQUOTA(&mp->m_sb)) {
+		if (XFS_IS_UQUOTA_ON(mp) &&
+		    mp->m_sb.sb_uquotino != NULLFSINO) {
+			ASSERT(mp->m_sb.sb_uquotino > 0);
+			if ((error = xfs_iget(mp, NULL, mp->m_sb.sb_uquotino,
+					     0, 0, &uip, 0)))
+				return XFS_ERROR(error);
+		}
+		if (XFS_IS_GQUOTA_ON(mp) &&
+		    mp->m_sb.sb_gquotino != NULLFSINO) {
+			ASSERT(mp->m_sb.sb_gquotino > 0);
+			if ((error = xfs_iget(mp, NULL, mp->m_sb.sb_gquotino,
+					     0, 0, &gip, 0))) {
+				if (uip)
+					VN_RELE(XFS_ITOV(uip));
+				return XFS_ERROR(error);
+			}
+		}
+	} else {
+		flags |= XFS_QMOPT_SBVERSION;
+		sbflags |= (XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO |
+			    XFS_SB_GQUOTINO | XFS_SB_QFLAGS);
+	}
+
+	/*
+	 * Create the two inodes, if they don't exist already. The changes
+	 * made above will get added to a transaction and logged in one of
+	 * the qino_alloc calls below.  If the device is readonly,
+	 * temporarily switch to read-write to do this.
+	 */
+	if (XFS_IS_UQUOTA_ON(mp) && uip == NULL) {
+		if ((error = xfs_qm_qino_alloc(mp, &uip,
+					      sbflags | XFS_SB_UQUOTINO,
+					      flags | XFS_QMOPT_UQUOTA)))
+			return XFS_ERROR(error);
+
+		flags &= ~XFS_QMOPT_SBVERSION;
+	}
+	if (XFS_IS_GQUOTA_ON(mp) && gip == NULL) {
+		if ((error = xfs_qm_qino_alloc(mp, &gip,
+					      sbflags | XFS_SB_GQUOTINO,
+					      flags | XFS_QMOPT_GQUOTA))) {
+			if (uip)
+				VN_RELE(XFS_ITOV(uip));
+
+			return XFS_ERROR(error);
+		}
+	}
+
+	XFS_QI_UQIP(mp) = uip;
+	XFS_QI_GQIP(mp) = gip;
+
+	return (0);
+}
+
+
+/*
+ * Traverse the freelist of dquots and attempt to reclaim a maximum of
+ * 'howmany' dquots. This operation races with dqlookup(), and attempts to
+ * favor the lookup function ...
+ * XXXsup merge this with qm_reclaim_one().
+ */
+STATIC int
+xfs_qm_shake_freelist(
+	int howmany)
+{
+	int		nreclaimed;
+	xfs_dqhash_t	*hash;
+	xfs_dquot_t	*dqp, *nextdqp;
+	int		restarts;
+	int		nflushes;
+
+	if (howmany <= 0)
+		return (0);
+
+	nreclaimed = 0;
+	restarts = 0;
+	nflushes = 0;
+
+#ifdef QUOTADEBUG
+	cmn_err(CE_DEBUG, "Shake free 0x%x", howmany);
+#endif
+	/* lock order is : hashchainlock, freelistlock, mplistlock */
+ tryagain:
+	xfs_qm_freelist_lock(xfs_Gqm);
+
+	for (dqp = xfs_Gqm->qm_dqfreelist.qh_next;
+	     ((dqp != (xfs_dquot_t *) &xfs_Gqm->qm_dqfreelist) &&
+	      nreclaimed < howmany); ) {
+		xfs_dqlock(dqp);
+
+		/*
+		 * We are racing with dqlookup here. Naturally we don't
+		 * want to reclaim a dquot that lookup wants.
+		 */
+		if (dqp->dq_flags & XFS_DQ_WANT) {
+			xfs_dqunlock(dqp);
+			xfs_qm_freelist_unlock(xfs_Gqm);
+			if (++restarts >= XFS_QM_RECLAIM_MAX_RESTARTS)
+				return (nreclaimed);
+			XQM_STATS_INC(xqmstats.xs_qm_dqwants);
+			goto tryagain;
+		}
+
+		/*
+		 * If the dquot is inactive, we are assured that it is
+		 * not on the mplist or the hashlist, and that makes our
+		 * life easier.
+		 */
+		if (dqp->dq_flags & XFS_DQ_INACTIVE) {
+			ASSERT(dqp->q_mount == NULL);
+			ASSERT(! XFS_DQ_IS_DIRTY(dqp));
+			ASSERT(dqp->HL_PREVP == NULL);
+			ASSERT(dqp->MPL_PREVP == NULL);
+			XQM_STATS_INC(xqmstats.xs_qm_dqinact_reclaims);
+			nextdqp = dqp->dq_flnext;
+			goto off_freelist;
+		}
+
+		ASSERT(dqp->MPL_PREVP);
+		/*
+		 * Try to grab the flush lock. If this dquot is in the process of
+		 * getting flushed to disk, we don't want to reclaim it.
+		 */
+		if (! xfs_qm_dqflock_nowait(dqp)) {
+			xfs_dqunlock(dqp);
+			dqp = dqp->dq_flnext;
+			continue;
+		}
+
+		/*
+		 * We have the flush lock so we know that this is not in the
+		 * process of being flushed. So, if this is dirty, flush it
+		 * DELWRI so that we don't get a freelist infested with
+		 * dirty dquots.
+		 */
+		if (XFS_DQ_IS_DIRTY(dqp)) {
+			xfs_dqtrace_entry(dqp, "DQSHAKE: DQDIRTY");
+			/*
+			 * We flush it delayed write, so don't bother
+			 * releasing the mplock.
+			 */
+			(void) xfs_qm_dqflush(dqp, XFS_QMOPT_DELWRI);
+			xfs_dqunlock(dqp); /* dqflush unlocks dqflock */
+			dqp = dqp->dq_flnext;
+			continue;
+		}
+		/*
+		 * We're trying to get the hashlock out of order. This races
+		 * with dqlookup; so, we giveup and goto the next dquot if
+		 * we couldn't get the hashlock. This way, we won't starve
+		 * a dqlookup process that holds the hashlock that is
+		 * waiting for the freelist lock.
+		 */
+		if (! xfs_qm_dqhashlock_nowait(dqp)) {
+			xfs_dqfunlock(dqp);
+			xfs_dqunlock(dqp);
+			dqp = dqp->dq_flnext;
+			continue;
+		}
+		/*
+		 * This races with dquot allocation code as well as dqflush_all
+		 * and reclaim code. So, if we failed to grab the mplist lock,
+		 * giveup everything and start over.
+		 */
+		hash = dqp->q_hash;
+		ASSERT(hash);
+		if (! xfs_qm_mplist_nowait(dqp->q_mount)) {
+			/* XXX put a sentinel so that we can come back here */
+			xfs_dqfunlock(dqp);
+			xfs_dqunlock(dqp);
+			XFS_DQ_HASH_UNLOCK(hash);
+			xfs_qm_freelist_unlock(xfs_Gqm);
+			if (++restarts >= XFS_QM_RECLAIM_MAX_RESTARTS)
+				return (nreclaimed);
+			goto tryagain;
+		}
+		xfs_dqtrace_entry(dqp, "DQSHAKE: UNLINKING");
+#ifdef QUOTADEBUG
+		cmn_err(CE_DEBUG, "Shake 0x%p, ID 0x%x\n",
+			dqp, INT_GET(dqp->q_core.d_id, ARCH_CONVERT));
+#endif
+		ASSERT(dqp->q_nrefs == 0);
+		nextdqp = dqp->dq_flnext;
+		XQM_MPLIST_REMOVE(&(XFS_QI_MPL_LIST(dqp->q_mount)), dqp);
+		XQM_HASHLIST_REMOVE(hash, dqp);
+		xfs_dqfunlock(dqp);
+		xfs_qm_mplist_unlock(dqp->q_mount);
+		XFS_DQ_HASH_UNLOCK(hash);
+
+ off_freelist:
+		XQM_FREELIST_REMOVE(dqp);
+		xfs_dqunlock(dqp);
+		nreclaimed++;
+		XQM_STATS_INC(xqmstats.xs_qm_dqshake_reclaims);
+		xfs_qm_dqdestroy(dqp);
+		dqp = nextdqp;
+	}
+	xfs_qm_freelist_unlock(xfs_Gqm);
+	return (nreclaimed);
+}
+
+
+/*
+ * The kmem_shake interface is invoked when memory is running low.
+ */
+/* ARGSUSED */
+STATIC int
+xfs_qm_shake(int nr_to_scan, unsigned int gfp_mask)
+{
+	int	ndqused, nfree, n;
+
+	if (!kmem_shake_allow(gfp_mask))
+		return (0);
+	if (!xfs_Gqm)
+		return (0);
+
+	nfree = xfs_Gqm->qm_dqfreelist.qh_nelems; /* free dquots */
+	/* incore dquots in all f/s's */
+	ndqused = atomic_read(&xfs_Gqm->qm_totaldquots) - nfree;
+
+	ASSERT(ndqused >= 0);
+
+	if (nfree <= ndqused && nfree < ndquot)
+		return (0);
+
+	ndqused *= xfs_Gqm->qm_dqfree_ratio;	/* target # of free dquots */
+	n = nfree - ndqused - ndquot;		/* # over target */
+
+	return xfs_qm_shake_freelist(MAX(nfree, n));
+}
+
+
+/*
+ * Just pop the least recently used dquot off the freelist and
+ * recycle it. The returned dquot is locked.
+ */
+STATIC xfs_dquot_t *
+xfs_qm_dqreclaim_one(void)
+{
+	xfs_dquot_t	*dqpout;
+	xfs_dquot_t	*dqp;
+	int		restarts;
+	int		nflushes;
+
+	restarts = 0;
+	dqpout = NULL;
+	nflushes = 0;
+
+	/* lockorder: hashchainlock, freelistlock, mplistlock, dqlock, dqflock */
+ startagain:
+	xfs_qm_freelist_lock(xfs_Gqm);
+
+	FOREACH_DQUOT_IN_FREELIST(dqp, &(xfs_Gqm->qm_dqfreelist)) {
+		xfs_dqlock(dqp);
+
+		/*
+		 * We are racing with dqlookup here. Naturally we don't
+		 * want to reclaim a dquot that lookup wants. We release the
+		 * freelist lock and start over, so that lookup will grab
+		 * both the dquot and the freelistlock.
+		 */
+		if (dqp->dq_flags & XFS_DQ_WANT) {
+			ASSERT(! (dqp->dq_flags & XFS_DQ_INACTIVE));
+			xfs_dqtrace_entry(dqp, "DQRECLAIM: DQWANT");
+			xfs_dqunlock(dqp);
+			xfs_qm_freelist_unlock(xfs_Gqm);
+			if (++restarts >= XFS_QM_RECLAIM_MAX_RESTARTS)
+				return (NULL);
+			XQM_STATS_INC(xqmstats.xs_qm_dqwants);
+			goto startagain;
+		}
+
+		/*
+		 * If the dquot is inactive, we are assured that it is
+		 * not on the mplist or the hashlist, and that makes our
+		 * life easier.
+		 */
+		if (dqp->dq_flags & XFS_DQ_INACTIVE) {
+			ASSERT(dqp->q_mount == NULL);
+			ASSERT(! XFS_DQ_IS_DIRTY(dqp));
+			ASSERT(dqp->HL_PREVP == NULL);
+			ASSERT(dqp->MPL_PREVP == NULL);
+			XQM_FREELIST_REMOVE(dqp);
+			xfs_dqunlock(dqp);
+			dqpout = dqp;
+			XQM_STATS_INC(xqmstats.xs_qm_dqinact_reclaims);
+			break;
+		}
+
+		ASSERT(dqp->q_hash);
+		ASSERT(dqp->MPL_PREVP);
+
+		/*
+		 * Try to grab the flush lock. If this dquot is in the process of
+		 * getting flushed to disk, we don't want to reclaim it.
+		 */
+		if (! xfs_qm_dqflock_nowait(dqp)) {
+			xfs_dqunlock(dqp);
+			continue;
+		}
+
+		/*
+		 * We have the flush lock so we know that this is not in the
+		 * process of being flushed. So, if this is dirty, flush it
+		 * DELWRI so that we don't get a freelist infested with
+		 * dirty dquots.
+		 */
+		if (XFS_DQ_IS_DIRTY(dqp)) {
+			xfs_dqtrace_entry(dqp, "DQRECLAIM: DQDIRTY");
+			/*
+			 * We flush it delayed write, so don't bother
+			 * releasing the freelist lock.
+			 */
+			(void) xfs_qm_dqflush(dqp, XFS_QMOPT_DELWRI);
+			xfs_dqunlock(dqp); /* dqflush unlocks dqflock */
+			continue;
+		}
+
+		if (! xfs_qm_mplist_nowait(dqp->q_mount)) {
+			xfs_dqfunlock(dqp);
+			xfs_dqunlock(dqp);
+			continue;
+		}
+
+		if (! xfs_qm_dqhashlock_nowait(dqp))
+			goto mplistunlock;
+
+		ASSERT(dqp->q_nrefs == 0);
+		xfs_dqtrace_entry(dqp, "DQRECLAIM: UNLINKING");
+		XQM_MPLIST_REMOVE(&(XFS_QI_MPL_LIST(dqp->q_mount)), dqp);
+		XQM_HASHLIST_REMOVE(dqp->q_hash, dqp);
+		XQM_FREELIST_REMOVE(dqp);
+		dqpout = dqp;
+		XFS_DQ_HASH_UNLOCK(dqp->q_hash);
+ mplistunlock:
+		xfs_qm_mplist_unlock(dqp->q_mount);
+		xfs_dqfunlock(dqp);
+		xfs_dqunlock(dqp);
+		if (dqpout)
+			break;
+	}
+
+	xfs_qm_freelist_unlock(xfs_Gqm);
+	return (dqpout);
+}
+
+
+/*------------------------------------------------------------------*/
+
+/*
+ * Return a new incore dquot. Depending on the number of
+ * dquots in the system, we either allocate a new one on the kernel heap,
+ * or reclaim a free one.
+ * Return value is B_TRUE if we allocated a new dquot, B_FALSE if we managed
+ * to reclaim an existing one from the freelist.
+ */
+boolean_t
+xfs_qm_dqalloc_incore(
+	xfs_dquot_t **O_dqpp)
+{
+	xfs_dquot_t	*dqp;
+
+	/*
+	 * Check against high water mark to see if we want to pop
+	 * a nincompoop dquot off the freelist.
+	 */
+	if (atomic_read(&xfs_Gqm->qm_totaldquots) >= ndquot) {
+		/*
+		 * Try to recycle a dquot from the freelist.
+		 */
+		if ((dqp = xfs_qm_dqreclaim_one())) {
+			XQM_STATS_INC(xqmstats.xs_qm_dqreclaims);
+			/*
+			 * Just zero the core here. The rest will get
+			 * reinitialized by caller. XXX we shouldn't even
+			 * do this zero ...
+			 */
+			memset(&dqp->q_core, 0, sizeof(dqp->q_core));
+			*O_dqpp = dqp;
+			return (B_FALSE);
+		}
+		XQM_STATS_INC(xqmstats.xs_qm_dqreclaim_misses);
+	}
+
+	/*
+	 * Allocate a brand new dquot on the kernel heap and return it
+	 * to the caller to initialize.
+	 */
+	ASSERT(xfs_Gqm->qm_dqzone != NULL);
+	*O_dqpp = kmem_zone_zalloc(xfs_Gqm->qm_dqzone, KM_SLEEP);
+	atomic_inc(&xfs_Gqm->qm_totaldquots);
+
+	return (B_TRUE);
+}
+
+
+/*
+ * Start a transaction and write the incore superblock changes to
+ * disk. flags parameter indicates which fields have changed.
+ */
+int
+xfs_qm_write_sb_changes(
+	xfs_mount_t	*mp,
+	__int64_t	flags)
+{
+	xfs_trans_t	*tp;
+	int		error;
+
+#ifdef QUOTADEBUG
+	cmn_err(CE_NOTE, "Writing superblock quota changes :%s", mp->m_fsname);
+#endif
+	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SBCHANGE);
+	if ((error = xfs_trans_reserve(tp, 0,
+				      mp->m_sb.sb_sectsize + 128, 0,
+				      0,
+				      XFS_DEFAULT_LOG_COUNT))) {
+		xfs_trans_cancel(tp, 0);
+		return (error);
+	}
+
+	xfs_mod_sb(tp, flags);
+	(void) xfs_trans_commit(tp, 0, NULL);
+
+	return (0);
+}
+
+
+/* --------------- utility functions for vnodeops ---------------- */
+
+
+/*
+ * Given an inode, a uid and gid (from cred_t) make sure that we have
+ * allocated relevant dquot(s) on disk, and that we won't exceed inode
+ * quotas by creating this file.
+ * This also attaches dquot(s) to the given inode after locking it,
+ * and returns the dquots corresponding to the uid and/or gid.
+ *
+ * in	: inode (unlocked)
+ * out	: udquot, gdquot with references taken and unlocked
+ */
+int
+xfs_qm_vop_dqalloc(
+	xfs_mount_t	*mp,
+	xfs_inode_t	*ip,
+	uid_t		uid,
+	gid_t		gid,
+	uint		flags,
+	xfs_dquot_t	**O_udqpp,
+	xfs_dquot_t	**O_gdqpp)
+{
+	int		error;
+	xfs_dquot_t	*uq, *gq;
+	uint		lockflags;
+
+	if (!XFS_IS_QUOTA_ON(mp))
+		return 0;
+
+	lockflags = XFS_ILOCK_EXCL;
+	xfs_ilock(ip, lockflags);
+
+	if ((flags & XFS_QMOPT_INHERIT) &&
+	    XFS_INHERIT_GID(ip, XFS_MTOVFS(mp)))
+		gid = ip->i_d.di_gid;
+
+	/*
+	 * Attach the dquot(s) to this inode, doing a dquot allocation
+	 * if necessary. The dquot(s) will not be locked.
+	 */
+	if (XFS_NOT_DQATTACHED(mp, ip)) {
+		if ((error = xfs_qm_dqattach(ip, XFS_QMOPT_DQALLOC |
+					    XFS_QMOPT_ILOCKED))) {
+			xfs_iunlock(ip, lockflags);
+			return (error);
+		}
+	}
+
+	uq = gq = NULL;
+	if ((flags & XFS_QMOPT_UQUOTA) &&
+	    XFS_IS_UQUOTA_ON(mp)) {
+		if (ip->i_d.di_uid != uid) {
+			/*
+			 * What we need is the dquot that has this uid, and
+			 * if we send the inode to dqget, the uid of the inode
+			 * takes priority over what's sent in the uid argument.
+			 * We must unlock inode here before calling dqget if
+			 * we're not sending the inode, because otherwise
+			 * we'll deadlock by doing trans_reserve while
+			 * holding ilock.
+			 */
+			xfs_iunlock(ip, lockflags);
+			if ((error = xfs_qm_dqget(mp, NULL, (xfs_dqid_t) uid,
+						 XFS_DQ_USER,
+						 XFS_QMOPT_DQALLOC |
+						 XFS_QMOPT_DOWARN,
+						 &uq))) {
+				ASSERT(error != ENOENT);
+				return (error);
+			}
+			/*
+			 * Get the ilock in the right order.
+			 */
+			xfs_dqunlock(uq);
+			lockflags = XFS_ILOCK_SHARED;
+			xfs_ilock(ip, lockflags);
+		} else {
+			/*
+			 * Take an extra reference, because we'll return
+			 * this to caller
+			 */
+			ASSERT(ip->i_udquot);
+			uq = ip->i_udquot;
+			xfs_dqlock(uq);
+			XFS_DQHOLD(uq);
+			xfs_dqunlock(uq);
+		}
+	}
+	if ((flags & XFS_QMOPT_GQUOTA) &&
+	    XFS_IS_GQUOTA_ON(mp)) {
+		if (ip->i_d.di_gid != gid) {
+			xfs_iunlock(ip, lockflags);
+			if ((error = xfs_qm_dqget(mp, NULL, (xfs_dqid_t)gid,
+						 XFS_DQ_GROUP,
+						 XFS_QMOPT_DQALLOC |
+						 XFS_QMOPT_DOWARN,
+						 &gq))) {
+				if (uq)
+					xfs_qm_dqrele(uq);
+				ASSERT(error != ENOENT);
+				return (error);
+			}
+			xfs_dqunlock(gq);
+			lockflags = XFS_ILOCK_SHARED;
+			xfs_ilock(ip, lockflags);
+		} else {
+			ASSERT(ip->i_gdquot);
+			gq = ip->i_gdquot;
+			xfs_dqlock(gq);
+			XFS_DQHOLD(gq);
+			xfs_dqunlock(gq);
+		}
+	}
+	if (uq)
+		xfs_dqtrace_entry_ino(uq, "DQALLOC", ip);
+
+	xfs_iunlock(ip, lockflags);
+	if (O_udqpp)
+		*O_udqpp = uq;
+	else if (uq)
+		xfs_qm_dqrele(uq);
+	if (O_gdqpp)
+		*O_gdqpp = gq;
+	else if (gq)
+		xfs_qm_dqrele(gq);
+	return (0);
+}
+
+/*
+ * Actually transfer ownership, and do dquot modifications.
+ * These were already reserved.
+ */
+xfs_dquot_t *
+xfs_qm_vop_chown(
+	xfs_trans_t	*tp,
+	xfs_inode_t	*ip,
+	xfs_dquot_t	**IO_olddq,
+	xfs_dquot_t	*newdq)
+{
+	xfs_dquot_t	*prevdq;
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+	ASSERT(XFS_IS_QUOTA_RUNNING(ip->i_mount));
+
+	/* old dquot */
+	prevdq = *IO_olddq;
+	ASSERT(prevdq);
+	ASSERT(prevdq != newdq);
+
+	xfs_trans_mod_dquot(tp, prevdq,
+			    XFS_TRANS_DQ_BCOUNT,
+			    -(ip->i_d.di_nblocks));
+	xfs_trans_mod_dquot(tp, prevdq,
+			    XFS_TRANS_DQ_ICOUNT,
+			    -1);
+
+	/* the sparkling new dquot */
+	xfs_trans_mod_dquot(tp, newdq,
+			    XFS_TRANS_DQ_BCOUNT,
+			    ip->i_d.di_nblocks);
+	xfs_trans_mod_dquot(tp, newdq,
+			    XFS_TRANS_DQ_ICOUNT,
+			    1);
+
+	/*
+	 * Take an extra reference, because the inode
+	 * is going to keep this dquot pointer even
+	 * after the trans_commit.
+	 */
+	xfs_dqlock(newdq);
+	XFS_DQHOLD(newdq);
+	xfs_dqunlock(newdq);
+	*IO_olddq = newdq;
+
+	return (prevdq);
+}
+
+/*
+ * Quota reservations for setattr(AT_UID|AT_GID).
+ */
+int
+xfs_qm_vop_chown_reserve(
+	xfs_trans_t	*tp,
+	xfs_inode_t	*ip,
+	xfs_dquot_t	*udqp,
+	xfs_dquot_t	*gdqp,
+	uint		flags)
+{
+	int		error;
+	xfs_mount_t	*mp;
+	uint		delblks;
+	xfs_dquot_t	*unresudq, *unresgdq, *delblksudq, *delblksgdq;
+
+	ASSERT(XFS_ISLOCKED_INODE(ip));
+	mp = ip->i_mount;
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+
+	delblks = ip->i_delayed_blks;
+	delblksudq = delblksgdq = unresudq = unresgdq = NULL;
+
+	if (XFS_IS_UQUOTA_ON(mp) && udqp &&
+	    ip->i_d.di_uid != (uid_t)INT_GET(udqp->q_core.d_id, ARCH_CONVERT)) {
+		delblksudq = udqp;
+		/*
+		 * If there are delayed allocation blocks, then we have to
+		 * unreserve those from the old dquot, and add them to the
+		 * new dquot.
+		 */
+		if (delblks) {
+			ASSERT(ip->i_udquot);
+			unresudq = ip->i_udquot;
+		}
+	}
+	if (XFS_IS_GQUOTA_ON(ip->i_mount) && gdqp &&
+	    ip->i_d.di_gid != INT_GET(gdqp->q_core.d_id, ARCH_CONVERT)) {
+		delblksgdq = gdqp;
+		if (delblks) {
+			ASSERT(ip->i_gdquot);
+			unresgdq = ip->i_gdquot;
+		}
+	}
+
+	if ((error = xfs_trans_reserve_quota_bydquots(tp, ip->i_mount,
+				delblksudq, delblksgdq, ip->i_d.di_nblocks, 1,
+				flags | XFS_QMOPT_RES_REGBLKS)))
+		return (error);
+
+	/*
+	 * Do the delayed blks reservations/unreservations now. Since, these
+	 * are done without the help of a transaction, if a reservation fails
+	 * its previous reservations won't be automatically undone by trans
+	 * code. So, we have to do it manually here.
+	 */
+	if (delblks) {
+		/*
+		 * Do the reservations first. Unreservation can't fail.
+		 */
+		ASSERT(delblksudq || delblksgdq);
+		ASSERT(unresudq || unresgdq);
+		if ((error = xfs_trans_reserve_quota_bydquots(NULL, ip->i_mount,
+				delblksudq, delblksgdq, (xfs_qcnt_t)delblks, 0,
+				flags | XFS_QMOPT_RES_REGBLKS)))
+			return (error);
+		xfs_trans_reserve_quota_bydquots(NULL, ip->i_mount,
+				unresudq, unresgdq, -((xfs_qcnt_t)delblks), 0,
+				XFS_QMOPT_RES_REGBLKS);
+	}
+
+	return (0);
+}
+
+int
+xfs_qm_vop_rename_dqattach(
+	xfs_inode_t	**i_tab)
+{
+	xfs_inode_t	*ip;
+	int		i;
+	int		error;
+
+	ip = i_tab[0];
+
+	if (! XFS_IS_QUOTA_ON(ip->i_mount))
+		return (0);
+
+	if (XFS_NOT_DQATTACHED(ip->i_mount, ip)) {
+		error = xfs_qm_dqattach(ip, 0);
+		if (error)
+			return (error);
+	}
+	for (i = 1; (i < 4 && i_tab[i]); i++) {
+		/*
+		 * Watch out for duplicate entries in the table.
+		 */
+		if ((ip = i_tab[i]) != i_tab[i-1]) {
+			if (XFS_NOT_DQATTACHED(ip->i_mount, ip)) {
+				error = xfs_qm_dqattach(ip, 0);
+				if (error)
+					return (error);
+			}
+		}
+	}
+	return (0);
+}
+
+void
+xfs_qm_vop_dqattach_and_dqmod_newinode(
+	xfs_trans_t	*tp,
+	xfs_inode_t	*ip,
+	xfs_dquot_t	*udqp,
+	xfs_dquot_t	*gdqp)
+{
+	if (!XFS_IS_QUOTA_ON(tp->t_mountp))
+		return;
+
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+	ASSERT(XFS_IS_QUOTA_RUNNING(tp->t_mountp));
+
+	if (udqp) {
+		xfs_dqlock(udqp);
+		XFS_DQHOLD(udqp);
+		xfs_dqunlock(udqp);
+		ASSERT(ip->i_udquot == NULL);
+		ip->i_udquot = udqp;
+		ASSERT(ip->i_d.di_uid == INT_GET(udqp->q_core.d_id, ARCH_CONVERT));
+		xfs_trans_mod_dquot(tp, udqp, XFS_TRANS_DQ_ICOUNT, 1);
+	}
+	if (gdqp) {
+		xfs_dqlock(gdqp);
+		XFS_DQHOLD(gdqp);
+		xfs_dqunlock(gdqp);
+		ASSERT(ip->i_gdquot == NULL);
+		ip->i_gdquot = gdqp;
+		ASSERT(ip->i_d.di_gid == INT_GET(gdqp->q_core.d_id, ARCH_CONVERT));
+		xfs_trans_mod_dquot(tp, gdqp, XFS_TRANS_DQ_ICOUNT, 1);
+	}
+}
+
+/* ------------- list stuff -----------------*/
+void
+xfs_qm_freelist_init(xfs_frlist_t *ql)
+{
+	ql->qh_next = ql->qh_prev = (xfs_dquot_t *) ql;
+	mutex_init(&ql->qh_lock, MUTEX_DEFAULT, "dqf");
+	ql->qh_version = 0;
+	ql->qh_nelems = 0;
+}
+
+void
+xfs_qm_freelist_destroy(xfs_frlist_t *ql)
+{
+	xfs_dquot_t	*dqp, *nextdqp;
+
+	mutex_lock(&ql->qh_lock, PINOD);
+	for (dqp = ql->qh_next;
+	     dqp != (xfs_dquot_t *)ql; ) {
+		xfs_dqlock(dqp);
+		nextdqp = dqp->dq_flnext;
+#ifdef QUOTADEBUG
+		cmn_err(CE_DEBUG, "FREELIST destroy 0x%p", dqp);
+#endif
+		XQM_FREELIST_REMOVE(dqp);
+		xfs_dqunlock(dqp);
+		xfs_qm_dqdestroy(dqp);
+		dqp = nextdqp;
+	}
+	/*
+	 * Don't bother about unlocking.
+	 */
+	mutex_destroy(&ql->qh_lock);
+
+	ASSERT(ql->qh_nelems == 0);
+}
+
+void
+xfs_qm_freelist_insert(xfs_frlist_t *ql, xfs_dquot_t *dq)
+{
+	dq->dq_flnext = ql->qh_next;
+	dq->dq_flprev = (xfs_dquot_t *)ql;
+	ql->qh_next = dq;
+	dq->dq_flnext->dq_flprev = dq;
+	xfs_Gqm->qm_dqfreelist.qh_nelems++;
+	xfs_Gqm->qm_dqfreelist.qh_version++;
+}
+
+void
+xfs_qm_freelist_unlink(xfs_dquot_t *dq)
+{
+	xfs_dquot_t *next = dq->dq_flnext;
+	xfs_dquot_t *prev = dq->dq_flprev;
+
+	next->dq_flprev = prev;
+	prev->dq_flnext = next;
+	dq->dq_flnext = dq->dq_flprev = dq;
+	xfs_Gqm->qm_dqfreelist.qh_nelems--;
+	xfs_Gqm->qm_dqfreelist.qh_version++;
+}
+
+void
+xfs_qm_freelist_append(xfs_frlist_t *ql, xfs_dquot_t *dq)
+{
+	xfs_qm_freelist_insert((xfs_frlist_t *)ql->qh_prev, dq);
+}
+
+int
+xfs_qm_dqhashlock_nowait(
+	xfs_dquot_t *dqp)
+{
+	int locked;
+
+	locked = mutex_trylock(&((dqp)->q_hash->qh_lock));
+	return (locked);
+}
+
+int
+xfs_qm_freelist_lock_nowait(
+	xfs_qm_t *xqm)
+{
+	int locked;
+
+	locked = mutex_trylock(&(xqm->qm_dqfreelist.qh_lock));
+	return (locked);
+}
+
+int
+xfs_qm_mplist_nowait(
+	xfs_mount_t	*mp)
+{
+	int locked;
+
+	ASSERT(mp->m_quotainfo);
+	locked = mutex_trylock(&(XFS_QI_MPLLOCK(mp)));
+	return (locked);
+}
diff --git a/fs/xfs/quota/xfs_qm.h b/fs/xfs/quota/xfs_qm.h
new file mode 100644
index 000000000000..dcf1a7a831d8
--- /dev/null
+++ b/fs/xfs/quota/xfs_qm.h
@@ -0,0 +1,236 @@
+/*
+ * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+#ifndef __XFS_QM_H__
+#define __XFS_QM_H__
+
+#include "xfs_dquot_item.h"
+#include "xfs_dquot.h"
+#include "xfs_quota_priv.h"
+#include "xfs_qm_stats.h"
+
+struct xfs_qm;
+struct xfs_inode;
+
+extern mutex_t		xfs_Gqm_lock;
+extern struct xfs_qm	*xfs_Gqm;
+extern kmem_zone_t	*qm_dqzone;
+extern kmem_zone_t	*qm_dqtrxzone;
+
+/*
+ * Used in xfs_qm_sync called by xfs_sync to count the max times that it can
+ * iterate over the mountpt's dquot list in one call.
+ */
+#define XFS_QM_SYNC_MAX_RESTARTS	7
+
+/*
+ * Ditto, for xfs_qm_dqreclaim_one.
+ */
+#define XFS_QM_RECLAIM_MAX_RESTARTS	4
+
+/*
+ * Ideal ratio of free to in use dquots. Quota manager makes an attempt
+ * to keep this balance.
+ */
+#define XFS_QM_DQFREE_RATIO		2
+
+/*
+ * Dquot hashtable constants/threshold values.
+ */
+#define XFS_QM_NCSIZE_THRESHOLD		5000
+#define XFS_QM_HASHSIZE_LOW		32
+#define XFS_QM_HASHSIZE_HIGH		64
+
+/*
+ * We output a cmn_err when quotachecking a quota file with more than
+ * this many fsbs.
+ */
+#define XFS_QM_BIG_QCHECK_NBLKS		500
+
+/*
+ * This defines the unit of allocation of dquots.
+ * Currently, it is just one file system block, and a 4K blk contains 30
+ * (136 * 30 = 4080) dquots. It's probably not worth trying to make
+ * this more dynamic.
+ * XXXsup However, if this number is changed, we have to make sure that we don't
+ * implicitly assume that we do allocations in chunks of a single filesystem
+ * block in the dquot/xqm code.
+ */
+#define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
+/*
+ * When doing a quotacheck, we log dquot clusters of this many FSBs at most
+ * in a single transaction. We don't want to ask for too huge a log reservation.
+ */
+#define XFS_QM_MAX_DQCLUSTER_LOGSZ	3
+
+typedef xfs_dqhash_t	xfs_dqlist_t;
+/*
+ * The freelist head. The first two fields match the first two in the
+ * xfs_dquot_t structure (in xfs_dqmarker_t)
+ */
+typedef struct xfs_frlist {
+       struct xfs_dquot *qh_next;
+       struct xfs_dquot *qh_prev;
+       mutex_t		 qh_lock;
+       uint		 qh_version;
+       uint		 qh_nelems;
+} xfs_frlist_t;
+
+/*
+ * Quota Manager (global) structure. Lives only in core.
+ */
+typedef struct xfs_qm {
+	xfs_dqlist_t	*qm_usr_dqhtable;/* udquot hash table */
+	xfs_dqlist_t	*qm_grp_dqhtable;/* gdquot hash table */
+	uint		 qm_dqhashmask;	 /* # buckets in dq hashtab - 1 */
+	xfs_frlist_t	 qm_dqfreelist;	 /* freelist of dquots */
+	atomic_t	 qm_totaldquots; /* total incore dquots */
+	uint		 qm_nrefs;	 /* file systems with quota on */
+	int		 qm_dqfree_ratio;/* ratio of free to inuse dquots */
+	kmem_zone_t	*qm_dqzone;	 /* dquot mem-alloc zone */
+	kmem_zone_t	*qm_dqtrxzone;	 /* t_dqinfo of transactions */
+} xfs_qm_t;
+
+/*
+ * Various quota information for individual filesystems.
+ * The mount structure keeps a pointer to this.
+ */
+typedef struct xfs_quotainfo {
+	xfs_inode_t	*qi_uquotaip;	 /* user quota inode */
+	xfs_inode_t	*qi_gquotaip;	 /* group quota inode */
+	lock_t		 qi_pinlock;	 /* dquot pinning mutex */
+	xfs_dqlist_t	 qi_dqlist;	 /* all dquots in filesys */
+	int		 qi_dqreclaims;	 /* a change here indicates
+					    a removal in the dqlist */
+	time_t		 qi_btimelimit;	 /* limit for blks timer */
+	time_t		 qi_itimelimit;	 /* limit for inodes timer */
+	time_t		 qi_rtbtimelimit;/* limit for rt blks timer */
+	xfs_qwarncnt_t	 qi_bwarnlimit;	 /* limit for num warnings */
+	xfs_qwarncnt_t	 qi_iwarnlimit;	 /* limit for num warnings */
+	mutex_t		 qi_quotaofflock;/* to serialize quotaoff */
+	xfs_filblks_t	 qi_dqchunklen;	 /* # BBs in a chunk of dqs */
+	uint		 qi_dqperchunk;	 /* # ondisk dqs in above chunk */
+	xfs_qcnt_t	 qi_bhardlimit;	 /* default data blk hard limit */
+	xfs_qcnt_t	 qi_bsoftlimit;	 /* default data blk soft limit */
+	xfs_qcnt_t	 qi_ihardlimit;	 /* default inode count hard limit */
+	xfs_qcnt_t	 qi_isoftlimit;	 /* default inode count soft limit */
+	xfs_qcnt_t	 qi_rtbhardlimit;/* default realtime blk hard limit */
+	xfs_qcnt_t	 qi_rtbsoftlimit;/* default realtime blk soft limit */
+} xfs_quotainfo_t;
+
+
+extern xfs_dqtrxops_t	xfs_trans_dquot_ops;
+
+extern void	xfs_trans_mod_dquot(xfs_trans_t *, xfs_dquot_t *, uint, long);
+extern int	xfs_trans_reserve_quota_bydquots(xfs_trans_t *, xfs_mount_t *,
+			xfs_dquot_t *, xfs_dquot_t *, long, long, uint);
+extern void	xfs_trans_dqjoin(xfs_trans_t *, xfs_dquot_t *);
+extern void	xfs_trans_log_dquot(xfs_trans_t *, xfs_dquot_t *);
+
+/*
+ * We keep the usr and grp dquots separately so that locking will be easier
+ * to do at commit time. All transactions that we know of at this point
+ * affect no more than two dquots of one type. Hence, the TRANS_MAXDQS value.
+ */
+#define XFS_QM_TRANS_MAXDQS		2
+typedef struct xfs_dquot_acct {
+	xfs_dqtrx_t	dqa_usrdquots[XFS_QM_TRANS_MAXDQS];
+	xfs_dqtrx_t	dqa_grpdquots[XFS_QM_TRANS_MAXDQS];
+} xfs_dquot_acct_t;
+
+/*
+ * Users are allowed to have a usage exceeding their softlimit for
+ * a period this long.
+ */
+#define XFS_QM_BTIMELIMIT	(7 * 24*60*60)          /* 1 week */
+#define XFS_QM_RTBTIMELIMIT	(7 * 24*60*60)          /* 1 week */
+#define XFS_QM_ITIMELIMIT	(7 * 24*60*60)          /* 1 week */
+
+#define XFS_QM_BWARNLIMIT	5
+#define XFS_QM_IWARNLIMIT	5
+
+#define XFS_QM_LOCK(xqm)	(mutex_lock(&xqm##_lock, PINOD))
+#define XFS_QM_UNLOCK(xqm)	(mutex_unlock(&xqm##_lock))
+#define XFS_QM_HOLD(xqm)	((xqm)->qm_nrefs++)
+#define XFS_QM_RELE(xqm)	((xqm)->qm_nrefs--)
+
+extern void		xfs_mount_reset_sbqflags(xfs_mount_t *);
+
+extern int		xfs_qm_init_quotainfo(xfs_mount_t *);
+extern void		xfs_qm_destroy_quotainfo(xfs_mount_t *);
+extern int		xfs_qm_mount_quotas(xfs_mount_t *, int);
+extern void		xfs_qm_mount_quotainit(xfs_mount_t *, uint);
+extern int		xfs_qm_quotacheck(xfs_mount_t *);
+extern void		xfs_qm_unmount_quotadestroy(xfs_mount_t *);
+extern int		xfs_qm_unmount_quotas(xfs_mount_t *);
+extern int		xfs_qm_write_sb_changes(xfs_mount_t *, __int64_t);
+extern int		xfs_qm_sync(xfs_mount_t *, short);
+
+/* dquot stuff */
+extern boolean_t	xfs_qm_dqalloc_incore(xfs_dquot_t **);
+extern int		xfs_qm_dqattach(xfs_inode_t *, uint);
+extern void		xfs_qm_dqdetach(xfs_inode_t *);
+extern int		xfs_qm_dqpurge_all(xfs_mount_t *, uint);
+extern void		xfs_qm_dqrele_all_inodes(xfs_mount_t *, uint);
+
+/* vop stuff */
+extern int		xfs_qm_vop_dqalloc(xfs_mount_t *, xfs_inode_t *,
+					uid_t, gid_t, uint,
+					xfs_dquot_t **, xfs_dquot_t **);
+extern void		xfs_qm_vop_dqattach_and_dqmod_newinode(
+					xfs_trans_t *, xfs_inode_t *,
+					xfs_dquot_t *, xfs_dquot_t *);
+extern int		xfs_qm_vop_rename_dqattach(xfs_inode_t **);
+extern xfs_dquot_t *	xfs_qm_vop_chown(xfs_trans_t *, xfs_inode_t *,
+					xfs_dquot_t **, xfs_dquot_t *);
+extern int		xfs_qm_vop_chown_reserve(xfs_trans_t *, xfs_inode_t *,
+					xfs_dquot_t *, xfs_dquot_t *, uint);
+
+/* list stuff */
+extern void		xfs_qm_freelist_init(xfs_frlist_t *);
+extern void		xfs_qm_freelist_destroy(xfs_frlist_t *);
+extern void		xfs_qm_freelist_insert(xfs_frlist_t *, xfs_dquot_t *);
+extern void		xfs_qm_freelist_append(xfs_frlist_t *, xfs_dquot_t *);
+extern void		xfs_qm_freelist_unlink(xfs_dquot_t *);
+extern int		xfs_qm_freelist_lock_nowait(xfs_qm_t *);
+extern int		xfs_qm_mplist_nowait(xfs_mount_t *);
+extern int		xfs_qm_dqhashlock_nowait(xfs_dquot_t *);
+
+/* system call interface */
+extern int		xfs_qm_quotactl(bhv_desc_t *, int, int, xfs_caddr_t);
+
+#ifdef DEBUG
+extern int		xfs_qm_internalqcheck(xfs_mount_t *);
+#else
+#define xfs_qm_internalqcheck(mp)	(0)
+#endif
+
+#endif /* __XFS_QM_H__ */
diff --git a/fs/xfs/quota/xfs_qm_bhv.c b/fs/xfs/quota/xfs_qm_bhv.c
new file mode 100644
index 000000000000..be67d9c265f8
--- /dev/null
+++ b/fs/xfs/quota/xfs_qm_bhv.c
@@ -0,0 +1,410 @@
+/*
+ * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_clnt.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+
+#include "xfs_qm.h"
+
+#define MNTOPT_QUOTA	"quota"		/* disk quotas (user) */
+#define MNTOPT_NOQUOTA	"noquota"	/* no quotas */
+#define MNTOPT_USRQUOTA	"usrquota"	/* user quota enabled */
+#define MNTOPT_GRPQUOTA	"grpquota"	/* group quota enabled */
+#define MNTOPT_UQUOTA	"uquota"	/* user quota (IRIX variant) */
+#define MNTOPT_GQUOTA	"gquota"	/* group quota (IRIX variant) */
+#define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
+#define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
+#define MNTOPT_QUOTANOENF  "qnoenforce"	/* same as uqnoenforce */
+
+STATIC int
+xfs_qm_parseargs(
+	struct bhv_desc		*bhv,
+	char			*options,
+	struct xfs_mount_args	*args,
+	int			update)
+{
+	size_t			length;
+	char			*local_options = options;
+	char			*this_char;
+	int			error;
+	int			referenced = update;
+
+	while ((this_char = strsep(&local_options, ",")) != NULL) {
+		length = strlen(this_char);
+		if (local_options)
+			length++;
+
+		if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
+			args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
+			args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
+			referenced = update;
+		} else if (!strcmp(this_char, MNTOPT_QUOTA) ||
+			   !strcmp(this_char, MNTOPT_UQUOTA) ||
+			   !strcmp(this_char, MNTOPT_USRQUOTA)) {
+			args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
+			referenced = 1;
+		} else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
+			   !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
+			args->flags |= XFSMNT_UQUOTA;
+			args->flags &= ~XFSMNT_UQUOTAENF;
+			referenced = 1;
+		} else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
+			   !strcmp(this_char, MNTOPT_GRPQUOTA)) {
+			args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
+			referenced = 1;
+		} else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
+			args->flags |= XFSMNT_GQUOTA;
+			args->flags &= ~XFSMNT_GQUOTAENF;
+			referenced = 1;
+		} else {
+			if (local_options)
+				*(local_options-1) = ',';
+			continue;
+		}
+
+		while (length--)
+			*this_char++ = ',';
+	}
+
+	PVFS_PARSEARGS(BHV_NEXT(bhv), options, args, update, error);
+	if (!error && !referenced)
+		bhv_remove_vfsops(bhvtovfs(bhv), VFS_POSITION_QM);
+	return error;
+}
+
+STATIC int
+xfs_qm_showargs(
+	struct bhv_desc		*bhv,
+	struct seq_file		*m)
+{
+	struct vfs		*vfsp = bhvtovfs(bhv);
+	struct xfs_mount	*mp = XFS_VFSTOM(vfsp);
+	int			error;
+
+	if (mp->m_qflags & XFS_UQUOTA_ACCT) {
+		(mp->m_qflags & XFS_UQUOTA_ENFD) ?
+			seq_puts(m, "," MNTOPT_USRQUOTA) :
+			seq_puts(m, "," MNTOPT_UQUOTANOENF);
+	}
+
+	if (mp->m_qflags & XFS_GQUOTA_ACCT) {
+		(mp->m_qflags & XFS_GQUOTA_ENFD) ?
+			seq_puts(m, "," MNTOPT_GRPQUOTA) :
+			seq_puts(m, "," MNTOPT_GQUOTANOENF);
+	}
+
+	if (!(mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_GQUOTA_ACCT)))
+		seq_puts(m, "," MNTOPT_NOQUOTA);
+
+	PVFS_SHOWARGS(BHV_NEXT(bhv), m, error);
+	return error;
+}
+
+STATIC int
+xfs_qm_mount(
+	struct bhv_desc		*bhv,
+	struct xfs_mount_args	*args,
+	struct cred		*cr)
+{
+	struct vfs		*vfsp = bhvtovfs(bhv);
+	struct xfs_mount	*mp = XFS_VFSTOM(vfsp);
+	int			error;
+
+	if (args->flags & (XFSMNT_UQUOTA | XFSMNT_GQUOTA))
+		xfs_qm_mount_quotainit(mp, args->flags);
+	PVFS_MOUNT(BHV_NEXT(bhv), args, cr, error);
+	return error;
+}
+
+STATIC int
+xfs_qm_syncall(
+	struct bhv_desc		*bhv,
+	int			flags,
+	cred_t			*credp)
+{
+	struct vfs		*vfsp = bhvtovfs(bhv);
+	struct xfs_mount	*mp = XFS_VFSTOM(vfsp);
+	int			error;
+
+	/*
+	 * Get the Quota Manager to flush the dquots.
+	 */
+	if (XFS_IS_QUOTA_ON(mp)) {
+		if ((error = xfs_qm_sync(mp, flags))) {
+			/*
+			 * If we got an IO error, we will be shutting down.
+			 * So, there's nothing more for us to do here.
+			 */
+			ASSERT(error != EIO || XFS_FORCED_SHUTDOWN(mp));
+			if (XFS_FORCED_SHUTDOWN(mp)) {
+				return XFS_ERROR(error);
+			}
+		}
+	}
+	PVFS_SYNC(BHV_NEXT(bhv), flags, credp, error);
+	return error;
+}
+
+/*
+ * Clear the quotaflags in memory and in the superblock.
+ */
+void
+xfs_mount_reset_sbqflags(
+	xfs_mount_t		*mp)
+{
+	xfs_trans_t		*tp;
+	unsigned long		s;
+
+	mp->m_qflags = 0;
+	/*
+	 * It is OK to look at sb_qflags here in mount path,
+	 * without SB_LOCK.
+	 */
+	if (mp->m_sb.sb_qflags == 0)
+		return;
+	s = XFS_SB_LOCK(mp);
+	mp->m_sb.sb_qflags = 0;
+	XFS_SB_UNLOCK(mp, s);
+
+	/*
+	 * if the fs is readonly, let the incore superblock run
+	 * with quotas off but don't flush the update out to disk
+	 */
+	if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
+		return;
+#ifdef QUOTADEBUG
+	xfs_fs_cmn_err(CE_NOTE, mp, "Writing superblock quota changes");
+#endif
+	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SBCHANGE);
+	if (xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
+				      XFS_DEFAULT_LOG_COUNT)) {
+		xfs_trans_cancel(tp, 0);
+		xfs_fs_cmn_err(CE_ALERT, mp,
+			"xfs_mount_reset_sbqflags: Superblock update failed!");
+		return;
+	}
+	xfs_mod_sb(tp, XFS_SB_QFLAGS);
+	xfs_trans_commit(tp, 0, NULL);
+}
+
+STATIC int
+xfs_qm_newmount(
+	xfs_mount_t	*mp,
+	uint		*needquotamount,
+	uint		*quotaflags)
+{
+	uint		quotaondisk;
+	uint		uquotaondisk = 0, gquotaondisk = 0;
+
+	*quotaflags = 0;
+	*needquotamount = B_FALSE;
+
+	quotaondisk = XFS_SB_VERSION_HASQUOTA(&mp->m_sb) &&
+		mp->m_sb.sb_qflags & (XFS_UQUOTA_ACCT|XFS_GQUOTA_ACCT);
+
+	if (quotaondisk) {
+		uquotaondisk = mp->m_sb.sb_qflags & XFS_UQUOTA_ACCT;
+		gquotaondisk = mp->m_sb.sb_qflags & XFS_GQUOTA_ACCT;
+	}
+
+	/*
+	 * If the device itself is read-only, we can't allow
+	 * the user to change the state of quota on the mount -
+	 * this would generate a transaction on the ro device,
+	 * which would lead to an I/O error and shutdown
+	 */
+
+	if (((uquotaondisk && !XFS_IS_UQUOTA_ON(mp)) ||
+	    (!uquotaondisk &&  XFS_IS_UQUOTA_ON(mp)) ||
+	     (gquotaondisk && !XFS_IS_GQUOTA_ON(mp)) ||
+	    (!gquotaondisk &&  XFS_IS_GQUOTA_ON(mp)))  &&
+	    xfs_dev_is_read_only(mp, "changing quota state")) {
+		cmn_err(CE_WARN,
+			"XFS: please mount with%s%s%s.",
+			(!quotaondisk ? "out quota" : ""),
+			(uquotaondisk ? " usrquota" : ""),
+			(gquotaondisk ? " grpquota" : ""));
+		return XFS_ERROR(EPERM);
+	}
+
+	if (XFS_IS_QUOTA_ON(mp) || quotaondisk) {
+		/*
+		 * Call mount_quotas at this point only if we won't have to do
+		 * a quotacheck.
+		 */
+		if (quotaondisk && !XFS_QM_NEED_QUOTACHECK(mp)) {
+			/*
+			 * If an error occured, qm_mount_quotas code
+			 * has already disabled quotas. So, just finish
+			 * mounting, and get on with the boring life
+			 * without disk quotas.
+			 */
+			xfs_qm_mount_quotas(mp, 0);
+		} else {
+			/*
+			 * Clear the quota flags, but remember them. This
+			 * is so that the quota code doesn't get invoked
+			 * before we're ready. This can happen when an
+			 * inode goes inactive and wants to free blocks,
+			 * or via xfs_log_mount_finish.
+			 */
+			*needquotamount = B_TRUE;
+			*quotaflags = mp->m_qflags;
+			mp->m_qflags = 0;
+		}
+	}
+
+	return 0;
+}
+
+STATIC int
+xfs_qm_endmount(
+	xfs_mount_t	*mp,
+	uint		needquotamount,
+	uint		quotaflags,
+	int		mfsi_flags)
+{
+	if (needquotamount) {
+		ASSERT(mp->m_qflags == 0);
+		mp->m_qflags = quotaflags;
+		xfs_qm_mount_quotas(mp, mfsi_flags);
+	}
+
+#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
+	if (! (XFS_IS_QUOTA_ON(mp)))
+		xfs_fs_cmn_err(CE_NOTE, mp, "Disk quotas not turned on");
+	else
+		xfs_fs_cmn_err(CE_NOTE, mp, "Disk quotas turned on");
+#endif
+
+#ifdef QUOTADEBUG
+	if (XFS_IS_QUOTA_ON(mp) && xfs_qm_internalqcheck(mp))
+		cmn_err(CE_WARN, "XFS: mount internalqcheck failed");
+#endif
+
+	return 0;
+}
+
+STATIC void
+xfs_qm_dqrele_null(
+	xfs_dquot_t	*dq)
+{
+	/*
+	 * Called from XFS, where we always check first for a NULL dquot.
+	 */
+	if (!dq)
+		return;
+	xfs_qm_dqrele(dq);
+}
+
+
+struct xfs_qmops xfs_qmcore_xfs = {
+	.xfs_qminit		= xfs_qm_newmount,
+	.xfs_qmdone		= xfs_qm_unmount_quotadestroy,
+	.xfs_qmmount		= xfs_qm_endmount,
+	.xfs_qmunmount		= xfs_qm_unmount_quotas,
+	.xfs_dqrele		= xfs_qm_dqrele_null,
+	.xfs_dqattach		= xfs_qm_dqattach,
+	.xfs_dqdetach		= xfs_qm_dqdetach,
+	.xfs_dqpurgeall		= xfs_qm_dqpurge_all,
+	.xfs_dqvopalloc		= xfs_qm_vop_dqalloc,
+	.xfs_dqvopcreate	= xfs_qm_vop_dqattach_and_dqmod_newinode,
+	.xfs_dqvoprename	= xfs_qm_vop_rename_dqattach,
+	.xfs_dqvopchown		= xfs_qm_vop_chown,
+	.xfs_dqvopchownresv	= xfs_qm_vop_chown_reserve,
+	.xfs_dqtrxops		= &xfs_trans_dquot_ops,
+};
+
+struct bhv_vfsops xfs_qmops = { {
+	BHV_IDENTITY_INIT(VFS_BHV_QM, VFS_POSITION_QM),
+	.vfs_parseargs		= xfs_qm_parseargs,
+	.vfs_showargs		= xfs_qm_showargs,
+	.vfs_mount		= xfs_qm_mount,
+	.vfs_sync		= xfs_qm_syncall,
+	.vfs_quotactl		= xfs_qm_quotactl, },
+};
+
+
+void __init
+xfs_qm_init(void)
+{
+	static char	message[] __initdata =
+		KERN_INFO "SGI XFS Quota Management subsystem\n";
+
+	printk(message);
+	mutex_init(&xfs_Gqm_lock, MUTEX_DEFAULT, "xfs_qmlock");
+	vfs_bhv_set_custom(&xfs_qmops, &xfs_qmcore_xfs);
+	xfs_qm_init_procfs();
+}
+
+void __exit
+xfs_qm_exit(void)
+{
+	vfs_bhv_clr_custom(&xfs_qmops);
+	xfs_qm_cleanup_procfs();
+	if (qm_dqzone)
+		kmem_cache_destroy(qm_dqzone);
+	if (qm_dqtrxzone)
+		kmem_cache_destroy(qm_dqtrxzone);
+}
diff --git a/fs/xfs/quota/xfs_qm_stats.c b/fs/xfs/quota/xfs_qm_stats.c
new file mode 100644
index 000000000000..29978e037fee
--- /dev/null
+++ b/fs/xfs/quota/xfs_qm_stats.c
@@ -0,0 +1,149 @@
+/*
+ * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+
+#include "xfs_qm.h"
+
+struct xqmstats xqmstats;
+
+STATIC int
+xfs_qm_read_xfsquota(
+	char		*buffer,
+	char		**start,
+	off_t		offset,
+	int		count,
+	int		*eof,
+	void		*data)
+{
+	int		len;
+
+	/* maximum; incore; ratio free to inuse; freelist */
+	len = sprintf(buffer, "%d\t%d\t%d\t%u\n",
+			ndquot,
+			xfs_Gqm? atomic_read(&xfs_Gqm->qm_totaldquots) : 0,
+			xfs_Gqm? xfs_Gqm->qm_dqfree_ratio : 0,
+			xfs_Gqm? xfs_Gqm->qm_dqfreelist.qh_nelems : 0);
+
+	if (offset >= len) {
+		*start = buffer;
+		*eof = 1;
+		return 0;
+	}
+	*start = buffer + offset;
+	if ((len -= offset) > count)
+		return count;
+	*eof = 1;
+
+	return len;
+}
+
+STATIC int
+xfs_qm_read_stats(
+	char		*buffer,
+	char		**start,
+	off_t		offset,
+	int		count,
+	int		*eof,
+	void		*data)
+{
+	int		len;
+
+	/* quota performance statistics */
+	len = sprintf(buffer, "qm %u %u %u %u %u %u %u %u\n",
+			xqmstats.xs_qm_dqreclaims,
+			xqmstats.xs_qm_dqreclaim_misses,
+			xqmstats.xs_qm_dquot_dups,
+			xqmstats.xs_qm_dqcachemisses,
+			xqmstats.xs_qm_dqcachehits,
+			xqmstats.xs_qm_dqwants,
+			xqmstats.xs_qm_dqshake_reclaims,
+			xqmstats.xs_qm_dqinact_reclaims);
+
+	if (offset >= len) {
+		*start = buffer;
+		*eof = 1;
+		return 0;
+	}
+	*start = buffer + offset;
+	if ((len -= offset) > count)
+		return count;
+	*eof = 1;
+
+	return len;
+}
+
+void
+xfs_qm_init_procfs(void)
+{
+	create_proc_read_entry("fs/xfs/xqmstat", 0, NULL, xfs_qm_read_stats, NULL);
+	create_proc_read_entry("fs/xfs/xqm", 0, NULL, xfs_qm_read_xfsquota, NULL);
+}
+
+void
+xfs_qm_cleanup_procfs(void)
+{
+	remove_proc_entry("fs/xfs/xqm", NULL);
+	remove_proc_entry("fs/xfs/xqmstat", NULL);
+}
diff --git a/fs/xfs/quota/xfs_qm_stats.h b/fs/xfs/quota/xfs_qm_stats.h
new file mode 100644
index 000000000000..8093c5c284ec
--- /dev/null
+++ b/fs/xfs/quota/xfs_qm_stats.h
@@ -0,0 +1,68 @@
+/*
+ * Copyright (c) 2002 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+#ifndef __XFS_QM_STATS_H__
+#define __XFS_QM_STATS_H__
+
+
+#if defined(CONFIG_PROC_FS) && !defined(XFS_STATS_OFF)
+
+/*
+ * XQM global statistics
+ */
+struct xqmstats {
+	__uint32_t		xs_qm_dqreclaims;
+	__uint32_t		xs_qm_dqreclaim_misses;
+	__uint32_t		xs_qm_dquot_dups;
+	__uint32_t		xs_qm_dqcachemisses;
+	__uint32_t		xs_qm_dqcachehits;
+	__uint32_t		xs_qm_dqwants;
+	__uint32_t		xs_qm_dqshake_reclaims;
+	__uint32_t		xs_qm_dqinact_reclaims;
+};
+
+extern struct xqmstats xqmstats;
+
+# define XQM_STATS_INC(count)	( (count)++ )
+
+extern void xfs_qm_init_procfs(void);
+extern void xfs_qm_cleanup_procfs(void);
+
+#else
+
+# define XQM_STATS_INC(count)	do { } while (0)
+
+static __inline void xfs_qm_init_procfs(void) { };
+static __inline void xfs_qm_cleanup_procfs(void) { };
+
+#endif
+
+#endif	/* __XFS_QM_STATS_H__ */
diff --git a/fs/xfs/quota/xfs_qm_syscalls.c b/fs/xfs/quota/xfs_qm_syscalls.c
new file mode 100644
index 000000000000..229f5b5a2d25
--- /dev/null
+++ b/fs/xfs/quota/xfs_qm_syscalls.c
@@ -0,0 +1,1458 @@
+/*
+ * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+#include "xfs_utils.h"
+
+#include "xfs_qm.h"
+
+#ifdef DEBUG
+# define qdprintk(s, args...)	cmn_err(CE_DEBUG, s, ## args)
+#else
+# define qdprintk(s, args...)	do { } while (0)
+#endif
+
+STATIC int	xfs_qm_scall_trunc_qfiles(xfs_mount_t *, uint);
+STATIC int	xfs_qm_scall_getquota(xfs_mount_t *, xfs_dqid_t, uint,
+					fs_disk_quota_t *);
+STATIC int	xfs_qm_scall_getqstat(xfs_mount_t *, fs_quota_stat_t *);
+STATIC int	xfs_qm_scall_setqlim(xfs_mount_t *, xfs_dqid_t, uint,
+					fs_disk_quota_t *);
+STATIC int	xfs_qm_scall_quotaon(xfs_mount_t *, uint);
+STATIC int	xfs_qm_scall_quotaoff(xfs_mount_t *, uint, boolean_t);
+STATIC int	xfs_qm_log_quotaoff(xfs_mount_t *, xfs_qoff_logitem_t **, uint);
+STATIC int	xfs_qm_log_quotaoff_end(xfs_mount_t *, xfs_qoff_logitem_t *,
+					uint);
+STATIC uint	xfs_qm_import_flags(uint);
+STATIC uint	xfs_qm_export_flags(uint);
+STATIC uint	xfs_qm_import_qtype_flags(uint);
+STATIC uint	xfs_qm_export_qtype_flags(uint);
+STATIC void	xfs_qm_export_dquot(xfs_mount_t *, xfs_disk_dquot_t *,
+					fs_disk_quota_t *);
+
+
+/*
+ * The main distribution switch of all XFS quotactl system calls.
+ */
+int
+xfs_qm_quotactl(
+	struct bhv_desc *bdp,
+	int		cmd,
+	int		id,
+	xfs_caddr_t	addr)
+{
+	xfs_mount_t	*mp;
+	int		error;
+	struct vfs	*vfsp;
+
+	vfsp = bhvtovfs(bdp);
+	mp = XFS_VFSTOM(vfsp);
+
+	if (addr == NULL && cmd != Q_SYNC)
+		return XFS_ERROR(EINVAL);
+	if (id < 0 && cmd != Q_SYNC)
+		return XFS_ERROR(EINVAL);
+
+	/*
+	 * The following commands are valid even when quotaoff.
+	 */
+	switch (cmd) {
+		/*
+		 * truncate quota files. quota must be off.
+		 */
+	      case Q_XQUOTARM:
+		if (XFS_IS_QUOTA_ON(mp) || addr == NULL)
+			return XFS_ERROR(EINVAL);
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			return XFS_ERROR(EROFS);
+		return (xfs_qm_scall_trunc_qfiles(mp,
+			       xfs_qm_import_qtype_flags(*(uint *)addr)));
+		/*
+		 * Get quota status information.
+		 */
+	      case Q_XGETQSTAT:
+		return (xfs_qm_scall_getqstat(mp, (fs_quota_stat_t *)addr));
+
+		/*
+		 * QUOTAON for root f/s and quota enforcement on others..
+		 * Quota accounting for non-root f/s's must be turned on
+		 * at mount time.
+		 */
+	      case Q_XQUOTAON:
+		if (addr == NULL)
+			return XFS_ERROR(EINVAL);
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			return XFS_ERROR(EROFS);
+		return (xfs_qm_scall_quotaon(mp,
+					  xfs_qm_import_flags(*(uint *)addr)));
+	      case Q_XQUOTAOFF:
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			return XFS_ERROR(EROFS);
+		break;
+
+	      default:
+		break;
+	}
+
+	if (! XFS_IS_QUOTA_ON(mp))
+		return XFS_ERROR(ESRCH);
+
+	switch (cmd) {
+	      case Q_XQUOTAOFF:
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			return XFS_ERROR(EROFS);
+		error = xfs_qm_scall_quotaoff(mp,
+					    xfs_qm_import_flags(*(uint *)addr),
+					    B_FALSE);
+		break;
+
+		/*
+		 * Defaults to XFS_GETUQUOTA.
+		 */
+	      case Q_XGETQUOTA:
+		error = xfs_qm_scall_getquota(mp, (xfs_dqid_t)id, XFS_DQ_USER,
+					(fs_disk_quota_t *)addr);
+		break;
+		/*
+		 * Set limits, both hard and soft. Defaults to Q_SETUQLIM.
+		 */
+	      case Q_XSETQLIM:
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			return XFS_ERROR(EROFS);
+		error = xfs_qm_scall_setqlim(mp, (xfs_dqid_t)id, XFS_DQ_USER,
+					     (fs_disk_quota_t *)addr);
+		break;
+
+	       case Q_XSETGQLIM:
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			return XFS_ERROR(EROFS);
+		error = xfs_qm_scall_setqlim(mp, (xfs_dqid_t)id, XFS_DQ_GROUP,
+					     (fs_disk_quota_t *)addr);
+		break;
+
+
+	      case Q_XGETGQUOTA:
+		error = xfs_qm_scall_getquota(mp, (xfs_dqid_t)id, XFS_DQ_GROUP,
+					(fs_disk_quota_t *)addr);
+		break;
+
+		/*
+		 * Quotas are entirely undefined after quotaoff in XFS quotas.
+		 * For instance, there's no way to set limits when quotaoff.
+		 */
+
+	      default:
+		error = XFS_ERROR(EINVAL);
+		break;
+	}
+
+	return (error);
+}
+
+/*
+ * Turn off quota accounting and/or enforcement for all udquots and/or
+ * gdquots. Called only at unmount time.
+ *
+ * This assumes that there are no dquots of this file system cached
+ * incore, and modifies the ondisk dquot directly. Therefore, for example,
+ * it is an error to call this twice, without purging the cache.
+ */
+STATIC int
+xfs_qm_scall_quotaoff(
+	xfs_mount_t		*mp,
+	uint			flags,
+	boolean_t		force)
+{
+	uint			dqtype;
+	unsigned long	s;
+	int			error;
+	uint			inactivate_flags;
+	xfs_qoff_logitem_t	*qoffstart;
+	int			nculprits;
+
+	if (!force && !capable(CAP_SYS_ADMIN))
+		return XFS_ERROR(EPERM);
+	/*
+	 * No file system can have quotas enabled on disk but not in core.
+	 * Note that quota utilities (like quotaoff) _expect_
+	 * errno == EEXIST here.
+	 */
+	if ((mp->m_qflags & flags) == 0)
+		return XFS_ERROR(EEXIST);
+	error = 0;
+
+	flags &= (XFS_ALL_QUOTA_ACCT | XFS_ALL_QUOTA_ENFD);
+
+	/*
+	 * We don't want to deal with two quotaoffs messing up each other,
+	 * so we're going to serialize it. quotaoff isn't exactly a performance
+	 * critical thing.
+	 * If quotaoff, then we must be dealing with the root filesystem.
+	 */
+	ASSERT(mp->m_quotainfo);
+	if (mp->m_quotainfo)
+		mutex_lock(&(XFS_QI_QOFFLOCK(mp)), PINOD);
+
+	ASSERT(mp->m_quotainfo);
+
+	/*
+	 * If we're just turning off quota enforcement, change mp and go.
+	 */
+	if ((flags & XFS_ALL_QUOTA_ACCT) == 0) {
+		mp->m_qflags &= ~(flags);
+
+		s = XFS_SB_LOCK(mp);
+		mp->m_sb.sb_qflags = mp->m_qflags;
+		XFS_SB_UNLOCK(mp, s);
+		mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+
+		/* XXX what to do if error ? Revert back to old vals incore ? */
+		error = xfs_qm_write_sb_changes(mp, XFS_SB_QFLAGS);
+		return (error);
+	}
+
+	dqtype = 0;
+	inactivate_flags = 0;
+	/*
+	 * If accounting is off, we must turn enforcement off, clear the
+	 * quota 'CHKD' certificate to make it known that we have to
+	 * do a quotacheck the next time this quota is turned on.
+	 */
+	if (flags & XFS_UQUOTA_ACCT) {
+		dqtype |= XFS_QMOPT_UQUOTA;
+		flags |= (XFS_UQUOTA_CHKD | XFS_UQUOTA_ENFD);
+		inactivate_flags |= XFS_UQUOTA_ACTIVE;
+	}
+	if (flags & XFS_GQUOTA_ACCT) {
+		dqtype |= XFS_QMOPT_GQUOTA;
+		flags |= (XFS_GQUOTA_CHKD | XFS_GQUOTA_ENFD);
+		inactivate_flags |= XFS_GQUOTA_ACTIVE;
+	}
+
+	/*
+	 * Nothing to do?  Don't complain. This happens when we're just
+	 * turning off quota enforcement.
+	 */
+	if ((mp->m_qflags & flags) == 0) {
+		mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+		return (0);
+	}
+
+	/*
+	 * Write the LI_QUOTAOFF log record, and do SB changes atomically,
+	 * and synchronously.
+	 */
+	xfs_qm_log_quotaoff(mp, &qoffstart, flags);
+
+	/*
+	 * Next we clear the XFS_MOUNT_*DQ_ACTIVE bit(s) in the mount struct
+	 * to take care of the race between dqget and quotaoff. We don't take
+	 * any special locks to reset these bits. All processes need to check
+	 * these bits *after* taking inode lock(s) to see if the particular
+	 * quota type is in the process of being turned off. If *ACTIVE, it is
+	 * guaranteed that all dquot structures and all quotainode ptrs will all
+	 * stay valid as long as that inode is kept locked.
+	 *
+	 * There is no turning back after this.
+	 */
+	mp->m_qflags &= ~inactivate_flags;
+
+	/*
+	 * Give back all the dquot reference(s) held by inodes.
+	 * Here we go thru every single incore inode in this file system, and
+	 * do a dqrele on the i_udquot/i_gdquot that it may have.
+	 * Essentially, as long as somebody has an inode locked, this guarantees
+	 * that quotas will not be turned off. This is handy because in a
+	 * transaction once we lock the inode(s) and check for quotaon, we can
+	 * depend on the quota inodes (and other things) being valid as long as
+	 * we keep the lock(s).
+	 */
+	xfs_qm_dqrele_all_inodes(mp, flags);
+
+	/*
+	 * Next we make the changes in the quota flag in the mount struct.
+	 * This isn't protected by a particular lock directly, because we
+	 * don't want to take a mrlock everytime we depend on quotas being on.
+	 */
+	mp->m_qflags &= ~(flags);
+
+	/*
+	 * Go through all the dquots of this file system and purge them,
+	 * according to what was turned off. We may not be able to get rid
+	 * of all dquots, because dquots can have temporary references that
+	 * are not attached to inodes. eg. xfs_setattr, xfs_create.
+	 * So, if we couldn't purge all the dquots from the filesystem,
+	 * we can't get rid of the incore data structures.
+	 */
+	while ((nculprits = xfs_qm_dqpurge_all(mp, dqtype|XFS_QMOPT_QUOTAOFF)))
+		delay(10 * nculprits);
+
+	/*
+	 * Transactions that had started before ACTIVE state bit was cleared
+	 * could have logged many dquots, so they'd have higher LSNs than
+	 * the first QUOTAOFF log record does. If we happen to crash when
+	 * the tail of the log has gone past the QUOTAOFF record, but
+	 * before the last dquot modification, those dquots __will__
+	 * recover, and that's not good.
+	 *
+	 * So, we have QUOTAOFF start and end logitems; the start
+	 * logitem won't get overwritten until the end logitem appears...
+	 */
+	xfs_qm_log_quotaoff_end(mp, qoffstart, flags);
+
+	/*
+	 * If quotas is completely disabled, close shop.
+	 */
+	if ((flags & XFS_MOUNT_QUOTA_ALL) == XFS_MOUNT_QUOTA_ALL) {
+		mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+		xfs_qm_destroy_quotainfo(mp);
+		return (0);
+	}
+
+	/*
+	 * Release our quotainode references, and vn_purge them,
+	 * if we don't need them anymore.
+	 */
+	if ((dqtype & XFS_QMOPT_UQUOTA) && XFS_QI_UQIP(mp)) {
+		XFS_PURGE_INODE(XFS_QI_UQIP(mp));
+		XFS_QI_UQIP(mp) = NULL;
+	}
+	if ((dqtype & XFS_QMOPT_GQUOTA) && XFS_QI_GQIP(mp)) {
+		XFS_PURGE_INODE(XFS_QI_GQIP(mp));
+		XFS_QI_GQIP(mp) = NULL;
+	}
+	mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+
+	return (error);
+}
+
+STATIC int
+xfs_qm_scall_trunc_qfiles(
+	xfs_mount_t	*mp,
+	uint		flags)
+{
+	int		error;
+	xfs_inode_t	*qip;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return XFS_ERROR(EPERM);
+	error = 0;
+	if (!XFS_SB_VERSION_HASQUOTA(&mp->m_sb) || flags == 0) {
+		qdprintk("qtrunc flags=%x m_qflags=%x\n", flags, mp->m_qflags);
+		return XFS_ERROR(EINVAL);
+	}
+
+	if ((flags & XFS_DQ_USER) && mp->m_sb.sb_uquotino != NULLFSINO) {
+		error = xfs_iget(mp, NULL, mp->m_sb.sb_uquotino, 0, 0, &qip, 0);
+		if (! error) {
+			(void) xfs_truncate_file(mp, qip);
+			VN_RELE(XFS_ITOV(qip));
+		}
+	}
+
+	if ((flags & XFS_DQ_GROUP) && mp->m_sb.sb_gquotino != NULLFSINO) {
+		error = xfs_iget(mp, NULL, mp->m_sb.sb_gquotino, 0, 0, &qip, 0);
+		if (! error) {
+			(void) xfs_truncate_file(mp, qip);
+			VN_RELE(XFS_ITOV(qip));
+		}
+	}
+
+	return (error);
+}
+
+
+/*
+ * Switch on (a given) quota enforcement for a filesystem.  This takes
+ * effect immediately.
+ * (Switching on quota accounting must be done at mount time.)
+ */
+STATIC int
+xfs_qm_scall_quotaon(
+	xfs_mount_t	*mp,
+	uint		flags)
+{
+	int		error;
+	unsigned long s;
+	uint		qf;
+	uint		accflags;
+	__int64_t	sbflags;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return XFS_ERROR(EPERM);
+
+	flags &= (XFS_ALL_QUOTA_ACCT | XFS_ALL_QUOTA_ENFD);
+	/*
+	 * Switching on quota accounting must be done at mount time.
+	 */
+	accflags = flags & XFS_ALL_QUOTA_ACCT;
+	flags &= ~(XFS_ALL_QUOTA_ACCT);
+
+	sbflags = 0;
+
+	if (flags == 0) {
+		qdprintk("quotaon: zero flags, m_qflags=%x\n", mp->m_qflags);
+		return XFS_ERROR(EINVAL);
+	}
+
+	/* No fs can turn on quotas with a delayed effect */
+	ASSERT((flags & XFS_ALL_QUOTA_ACCT) == 0);
+
+	/*
+	 * Can't enforce without accounting. We check the superblock
+	 * qflags here instead of m_qflags because rootfs can have
+	 * quota acct on ondisk without m_qflags' knowing.
+	 */
+	if (((flags & XFS_UQUOTA_ACCT) == 0 &&
+	    (mp->m_sb.sb_qflags & XFS_UQUOTA_ACCT) == 0 &&
+	    (flags & XFS_UQUOTA_ENFD))
+	    ||
+	    ((flags & XFS_GQUOTA_ACCT) == 0 &&
+	    (mp->m_sb.sb_qflags & XFS_GQUOTA_ACCT) == 0 &&
+	    (flags & XFS_GQUOTA_ENFD))) {
+		qdprintk("Can't enforce without acct, flags=%x sbflags=%x\n",
+			flags, mp->m_sb.sb_qflags);
+		return XFS_ERROR(EINVAL);
+	}
+	/*
+	 * If everything's upto-date incore, then don't waste time.
+	 */
+	if ((mp->m_qflags & flags) == flags)
+		return XFS_ERROR(EEXIST);
+
+	/*
+	 * Change sb_qflags on disk but not incore mp->qflags
+	 * if this is the root filesystem.
+	 */
+	s = XFS_SB_LOCK(mp);
+	qf = mp->m_sb.sb_qflags;
+	mp->m_sb.sb_qflags = qf | flags;
+	XFS_SB_UNLOCK(mp, s);
+
+	/*
+	 * There's nothing to change if it's the same.
+	 */
+	if ((qf & flags) == flags && sbflags == 0)
+		return XFS_ERROR(EEXIST);
+	sbflags |= XFS_SB_QFLAGS;
+
+	if ((error = xfs_qm_write_sb_changes(mp, sbflags)))
+		return (error);
+	/*
+	 * If we aren't trying to switch on quota enforcement, we are done.
+	 */
+	if  (((mp->m_sb.sb_qflags & XFS_UQUOTA_ACCT) !=
+	     (mp->m_qflags & XFS_UQUOTA_ACCT)) ||
+	    (flags & XFS_ALL_QUOTA_ENFD) == 0)
+		return (0);
+
+	if (! XFS_IS_QUOTA_RUNNING(mp))
+		return XFS_ERROR(ESRCH);
+
+	/*
+	 * Switch on quota enforcement in core.
+	 */
+	mutex_lock(&(XFS_QI_QOFFLOCK(mp)), PINOD);
+	mp->m_qflags |= (flags & XFS_ALL_QUOTA_ENFD);
+	mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+
+	return (0);
+}
+
+
+
+/*
+ * Return quota status information, such as uquota-off, enforcements, etc.
+ */
+STATIC int
+xfs_qm_scall_getqstat(
+	xfs_mount_t	*mp,
+	fs_quota_stat_t *out)
+{
+	xfs_inode_t	*uip, *gip;
+	boolean_t	tempuqip, tempgqip;
+
+	uip = gip = NULL;
+	tempuqip = tempgqip = B_FALSE;
+	memset(out, 0, sizeof(fs_quota_stat_t));
+
+	out->qs_version = FS_QSTAT_VERSION;
+	if (! XFS_SB_VERSION_HASQUOTA(&mp->m_sb)) {
+		out->qs_uquota.qfs_ino = NULLFSINO;
+		out->qs_gquota.qfs_ino = NULLFSINO;
+		return (0);
+	}
+	out->qs_flags = (__uint16_t) xfs_qm_export_flags(mp->m_qflags &
+							(XFS_ALL_QUOTA_ACCT|
+							 XFS_ALL_QUOTA_ENFD));
+	out->qs_pad = 0;
+	out->qs_uquota.qfs_ino = mp->m_sb.sb_uquotino;
+	out->qs_gquota.qfs_ino = mp->m_sb.sb_gquotino;
+
+	if (mp->m_quotainfo) {
+		uip = mp->m_quotainfo->qi_uquotaip;
+		gip = mp->m_quotainfo->qi_gquotaip;
+	}
+	if (!uip && mp->m_sb.sb_uquotino != NULLFSINO) {
+		if (xfs_iget(mp, NULL, mp->m_sb.sb_uquotino,
+					0, 0, &uip, 0) == 0)
+			tempuqip = B_TRUE;
+	}
+	if (!gip && mp->m_sb.sb_gquotino != NULLFSINO) {
+		if (xfs_iget(mp, NULL, mp->m_sb.sb_gquotino,
+					0, 0, &gip, 0) == 0)
+			tempgqip = B_TRUE;
+	}
+	if (uip) {
+		out->qs_uquota.qfs_nblks = uip->i_d.di_nblocks;
+		out->qs_uquota.qfs_nextents = uip->i_d.di_nextents;
+		if (tempuqip)
+			VN_RELE(XFS_ITOV(uip));
+	}
+	if (gip) {
+		out->qs_gquota.qfs_nblks = gip->i_d.di_nblocks;
+		out->qs_gquota.qfs_nextents = gip->i_d.di_nextents;
+		if (tempgqip)
+			VN_RELE(XFS_ITOV(gip));
+	}
+	if (mp->m_quotainfo) {
+		out->qs_incoredqs = XFS_QI_MPLNDQUOTS(mp);
+		out->qs_btimelimit = XFS_QI_BTIMELIMIT(mp);
+		out->qs_itimelimit = XFS_QI_ITIMELIMIT(mp);
+		out->qs_rtbtimelimit = XFS_QI_RTBTIMELIMIT(mp);
+		out->qs_bwarnlimit = XFS_QI_BWARNLIMIT(mp);
+		out->qs_iwarnlimit = XFS_QI_IWARNLIMIT(mp);
+	}
+	return (0);
+}
+
+/*
+ * Adjust quota limits, and start/stop timers accordingly.
+ */
+STATIC int
+xfs_qm_scall_setqlim(
+	xfs_mount_t		*mp,
+	xfs_dqid_t		id,
+	uint			type,
+	fs_disk_quota_t		*newlim)
+{
+	xfs_disk_dquot_t	*ddq;
+	xfs_dquot_t		*dqp;
+	xfs_trans_t		*tp;
+	int			error;
+	xfs_qcnt_t		hard, soft;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return XFS_ERROR(EPERM);
+
+	if ((newlim->d_fieldmask & (FS_DQ_LIMIT_MASK|FS_DQ_TIMER_MASK)) == 0)
+		return (0);
+
+	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SETQLIM);
+	if ((error = xfs_trans_reserve(tp, 0, sizeof(xfs_disk_dquot_t) + 128,
+				      0, 0, XFS_DEFAULT_LOG_COUNT))) {
+		xfs_trans_cancel(tp, 0);
+		return (error);
+	}
+
+	/*
+	 * We don't want to race with a quotaoff so take the quotaoff lock.
+	 * (We don't hold an inode lock, so there's nothing else to stop
+	 * a quotaoff from happening). (XXXThis doesn't currently happen
+	 * because we take the vfslock before calling xfs_qm_sysent).
+	 */
+	mutex_lock(&(XFS_QI_QOFFLOCK(mp)), PINOD);
+
+	/*
+	 * Get the dquot (locked), and join it to the transaction.
+	 * Allocate the dquot if this doesn't exist.
+	 */
+	if ((error = xfs_qm_dqget(mp, NULL, id, type, XFS_QMOPT_DQALLOC, &dqp))) {
+		xfs_trans_cancel(tp, XFS_TRANS_ABORT);
+		mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+		ASSERT(error != ENOENT);
+		return (error);
+	}
+	xfs_dqtrace_entry(dqp, "Q_SETQLIM: AFT DQGET");
+	xfs_trans_dqjoin(tp, dqp);
+	ddq = &dqp->q_core;
+
+	/*
+	 * Make sure that hardlimits are >= soft limits before changing.
+	 */
+	hard = (newlim->d_fieldmask & FS_DQ_BHARD) ?
+		(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_blk_hardlimit) :
+			INT_GET(ddq->d_blk_hardlimit, ARCH_CONVERT);
+	soft = (newlim->d_fieldmask & FS_DQ_BSOFT) ?
+		(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_blk_softlimit) :
+			INT_GET(ddq->d_blk_softlimit, ARCH_CONVERT);
+	if (hard == 0 || hard >= soft) {
+		INT_SET(ddq->d_blk_hardlimit, ARCH_CONVERT, hard);
+		INT_SET(ddq->d_blk_softlimit, ARCH_CONVERT, soft);
+		if (id == 0) {
+			mp->m_quotainfo->qi_bhardlimit = hard;
+			mp->m_quotainfo->qi_bsoftlimit = soft;
+		}
+	} else {
+		qdprintk("blkhard %Ld < blksoft %Ld\n", hard, soft);
+	}
+	hard = (newlim->d_fieldmask & FS_DQ_RTBHARD) ?
+		(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_rtb_hardlimit) :
+			INT_GET(ddq->d_rtb_hardlimit, ARCH_CONVERT);
+	soft = (newlim->d_fieldmask & FS_DQ_RTBSOFT) ?
+		(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_rtb_softlimit) :
+			INT_GET(ddq->d_rtb_softlimit, ARCH_CONVERT);
+	if (hard == 0 || hard >= soft) {
+		INT_SET(ddq->d_rtb_hardlimit, ARCH_CONVERT, hard);
+		INT_SET(ddq->d_rtb_softlimit, ARCH_CONVERT, soft);
+		if (id == 0) {
+			mp->m_quotainfo->qi_rtbhardlimit = hard;
+			mp->m_quotainfo->qi_rtbsoftlimit = soft;
+		}
+	} else {
+		qdprintk("rtbhard %Ld < rtbsoft %Ld\n", hard, soft);
+	}
+
+	hard = (newlim->d_fieldmask & FS_DQ_IHARD) ?
+		(xfs_qcnt_t) newlim->d_ino_hardlimit :
+			INT_GET(ddq->d_ino_hardlimit, ARCH_CONVERT);
+	soft = (newlim->d_fieldmask & FS_DQ_ISOFT) ?
+		(xfs_qcnt_t) newlim->d_ino_softlimit :
+			INT_GET(ddq->d_ino_softlimit, ARCH_CONVERT);
+	if (hard == 0 || hard >= soft) {
+		INT_SET(ddq->d_ino_hardlimit, ARCH_CONVERT, hard);
+		INT_SET(ddq->d_ino_softlimit, ARCH_CONVERT, soft);
+		if (id == 0) {
+			mp->m_quotainfo->qi_ihardlimit = hard;
+			mp->m_quotainfo->qi_isoftlimit = soft;
+		}
+	} else {
+		qdprintk("ihard %Ld < isoft %Ld\n", hard, soft);
+	}
+
+	if (id == 0) {
+		/*
+		 * Timelimits for the super user set the relative time
+		 * the other users can be over quota for this file system.
+		 * If it is zero a default is used.  Ditto for the default
+		 * soft and hard limit values (already done, above).
+		 */
+		if (newlim->d_fieldmask & FS_DQ_BTIMER) {
+			mp->m_quotainfo->qi_btimelimit = newlim->d_btimer;
+			INT_SET(ddq->d_btimer, ARCH_CONVERT, newlim->d_btimer);
+		}
+		if (newlim->d_fieldmask & FS_DQ_ITIMER) {
+			mp->m_quotainfo->qi_itimelimit = newlim->d_itimer;
+			INT_SET(ddq->d_itimer, ARCH_CONVERT, newlim->d_itimer);
+		}
+		if (newlim->d_fieldmask & FS_DQ_RTBTIMER) {
+			mp->m_quotainfo->qi_rtbtimelimit = newlim->d_rtbtimer;
+			INT_SET(ddq->d_rtbtimer, ARCH_CONVERT, newlim->d_rtbtimer);
+		}
+	} else /* if (XFS_IS_QUOTA_ENFORCED(mp)) */ {
+		/*
+		 * If the user is now over quota, start the timelimit.
+		 * The user will not be 'warned'.
+		 * Note that we keep the timers ticking, whether enforcement
+		 * is on or off. We don't really want to bother with iterating
+		 * over all ondisk dquots and turning the timers on/off.
+		 */
+		xfs_qm_adjust_dqtimers(mp, ddq);
+	}
+	dqp->dq_flags |= XFS_DQ_DIRTY;
+	xfs_trans_log_dquot(tp, dqp);
+
+	xfs_dqtrace_entry(dqp, "Q_SETQLIM: COMMIT");
+	xfs_trans_commit(tp, 0, NULL);
+	xfs_qm_dqprint(dqp);
+	xfs_qm_dqrele(dqp);
+	mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
+
+	return (0);
+}
+
+STATIC int
+xfs_qm_scall_getquota(
+	xfs_mount_t	*mp,
+	xfs_dqid_t	id,
+	uint		type,
+	fs_disk_quota_t *out)
+{
+	xfs_dquot_t	*dqp;
+	int		error;
+
+	/*
+	 * Try to get the dquot. We don't want it allocated on disk, so
+	 * we aren't passing the XFS_QMOPT_DOALLOC flag. If it doesn't
+	 * exist, we'll get ENOENT back.
+	 */
+	if ((error = xfs_qm_dqget(mp, NULL, id, type, 0, &dqp))) {
+		return (error);
+	}
+
+	xfs_dqtrace_entry(dqp, "Q_GETQUOTA SUCCESS");
+	/*
+	 * If everything's NULL, this dquot doesn't quite exist as far as
+	 * our utility programs are concerned.
+	 */
+	if (XFS_IS_DQUOT_UNINITIALIZED(dqp)) {
+		xfs_qm_dqput(dqp);
+		return XFS_ERROR(ENOENT);
+	}
+	/* xfs_qm_dqprint(dqp); */
+	/*
+	 * Convert the disk dquot to the exportable format
+	 */
+	xfs_qm_export_dquot(mp, &dqp->q_core, out);
+	xfs_qm_dqput(dqp);
+	return (error ? XFS_ERROR(EFAULT) : 0);
+}
+
+
+STATIC int
+xfs_qm_log_quotaoff_end(
+	xfs_mount_t		*mp,
+	xfs_qoff_logitem_t	*startqoff,
+	uint			flags)
+{
+	xfs_trans_t	       *tp;
+	int			error;
+	xfs_qoff_logitem_t     *qoffi;
+
+	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_QUOTAOFF_END);
+
+	if ((error = xfs_trans_reserve(tp, 0, sizeof(xfs_qoff_logitem_t) * 2,
+				      0, 0, XFS_DEFAULT_LOG_COUNT))) {
+		xfs_trans_cancel(tp, 0);
+		return (error);
+	}
+
+	qoffi = xfs_trans_get_qoff_item(tp, startqoff,
+					flags & XFS_ALL_QUOTA_ACCT);
+	xfs_trans_log_quotaoff_item(tp, qoffi);
+
+	/*
+	 * We have to make sure that the transaction is secure on disk before we
+	 * return and actually stop quota accounting. So, make it synchronous.
+	 * We don't care about quotoff's performance.
+	 */
+	xfs_trans_set_sync(tp);
+	error = xfs_trans_commit(tp, 0, NULL);
+	return (error);
+}
+
+
+STATIC int
+xfs_qm_log_quotaoff(
+	xfs_mount_t	       *mp,
+	xfs_qoff_logitem_t     **qoffstartp,
+	uint		       flags)
+{
+	xfs_trans_t	       *tp;
+	int			error;
+	unsigned long	s;
+	xfs_qoff_logitem_t     *qoffi=NULL;
+	uint			oldsbqflag=0;
+
+	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_QUOTAOFF);
+	if ((error = xfs_trans_reserve(tp, 0,
+				      sizeof(xfs_qoff_logitem_t) * 2 +
+				      mp->m_sb.sb_sectsize + 128,
+				      0,
+				      0,
+				      XFS_DEFAULT_LOG_COUNT))) {
+		goto error0;
+	}
+
+	qoffi = xfs_trans_get_qoff_item(tp, NULL, flags & XFS_ALL_QUOTA_ACCT);
+	xfs_trans_log_quotaoff_item(tp, qoffi);
+
+	s = XFS_SB_LOCK(mp);
+	oldsbqflag = mp->m_sb.sb_qflags;
+	mp->m_sb.sb_qflags = (mp->m_qflags & ~(flags)) & XFS_MOUNT_QUOTA_ALL;
+	XFS_SB_UNLOCK(mp, s);
+
+	xfs_mod_sb(tp, XFS_SB_QFLAGS);
+
+	/*
+	 * We have to make sure that the transaction is secure on disk before we
+	 * return and actually stop quota accounting. So, make it synchronous.
+	 * We don't care about quotoff's performance.
+	 */
+	xfs_trans_set_sync(tp);
+	error = xfs_trans_commit(tp, 0, NULL);
+
+error0:
+	if (error) {
+		xfs_trans_cancel(tp, 0);
+		/*
+		 * No one else is modifying sb_qflags, so this is OK.
+		 * We still hold the quotaofflock.
+		 */
+		s = XFS_SB_LOCK(mp);
+		mp->m_sb.sb_qflags = oldsbqflag;
+		XFS_SB_UNLOCK(mp, s);
+	}
+	*qoffstartp = qoffi;
+	return (error);
+}
+
+
+/*
+ * Translate an internal style on-disk-dquot to the exportable format.
+ * The main differences are that the counters/limits are all in Basic
+ * Blocks (BBs) instead of the internal FSBs, and all on-disk data has
+ * to be converted to the native endianness.
+ */
+STATIC void
+xfs_qm_export_dquot(
+	xfs_mount_t		*mp,
+	xfs_disk_dquot_t	*src,
+	struct fs_disk_quota	*dst)
+{
+	memset(dst, 0, sizeof(*dst));
+	dst->d_version = FS_DQUOT_VERSION;  /* different from src->d_version */
+	dst->d_flags =
+		xfs_qm_export_qtype_flags(INT_GET(src->d_flags, ARCH_CONVERT));
+	dst->d_id = INT_GET(src->d_id, ARCH_CONVERT);
+	dst->d_blk_hardlimit = (__uint64_t)
+		XFS_FSB_TO_BB(mp, INT_GET(src->d_blk_hardlimit, ARCH_CONVERT));
+	dst->d_blk_softlimit = (__uint64_t)
+		XFS_FSB_TO_BB(mp, INT_GET(src->d_blk_softlimit, ARCH_CONVERT));
+	dst->d_ino_hardlimit = (__uint64_t)
+		INT_GET(src->d_ino_hardlimit, ARCH_CONVERT);
+	dst->d_ino_softlimit = (__uint64_t)
+		INT_GET(src->d_ino_softlimit, ARCH_CONVERT);
+	dst->d_bcount = (__uint64_t)
+		XFS_FSB_TO_BB(mp, INT_GET(src->d_bcount, ARCH_CONVERT));
+	dst->d_icount = (__uint64_t) INT_GET(src->d_icount, ARCH_CONVERT);
+	dst->d_btimer = (__uint32_t) INT_GET(src->d_btimer, ARCH_CONVERT);
+	dst->d_itimer = (__uint32_t) INT_GET(src->d_itimer, ARCH_CONVERT);
+	dst->d_iwarns = INT_GET(src->d_iwarns, ARCH_CONVERT);
+	dst->d_bwarns = INT_GET(src->d_bwarns, ARCH_CONVERT);
+
+	dst->d_rtb_hardlimit = (__uint64_t)
+		XFS_FSB_TO_BB(mp, INT_GET(src->d_rtb_hardlimit, ARCH_CONVERT));
+	dst->d_rtb_softlimit = (__uint64_t)
+		XFS_FSB_TO_BB(mp, INT_GET(src->d_rtb_softlimit, ARCH_CONVERT));
+	dst->d_rtbcount = (__uint64_t)
+		XFS_FSB_TO_BB(mp, INT_GET(src->d_rtbcount, ARCH_CONVERT));
+	dst->d_rtbtimer = (__uint32_t) INT_GET(src->d_rtbtimer, ARCH_CONVERT);
+	dst->d_rtbwarns = INT_GET(src->d_rtbwarns, ARCH_CONVERT);
+
+	/*
+	 * Internally, we don't reset all the timers when quota enforcement
+	 * gets turned off. No need to confuse the userlevel code,
+	 * so return zeroes in that case.
+	 */
+	if (! XFS_IS_QUOTA_ENFORCED(mp)) {
+		dst->d_btimer = 0;
+		dst->d_itimer = 0;
+		dst->d_rtbtimer = 0;
+	}
+
+#ifdef DEBUG
+	if (XFS_IS_QUOTA_ENFORCED(mp) && dst->d_id != 0) {
+		if (((int) dst->d_bcount >= (int) dst->d_blk_softlimit) &&
+		    (dst->d_blk_softlimit > 0)) {
+			ASSERT(dst->d_btimer != 0);
+		}
+		if (((int) dst->d_icount >= (int) dst->d_ino_softlimit) &&
+		    (dst->d_ino_softlimit > 0)) {
+			ASSERT(dst->d_itimer != 0);
+		}
+	}
+#endif
+}
+
+STATIC uint
+xfs_qm_import_qtype_flags(
+	uint uflags)
+{
+	/*
+	 * Can't be both at the same time.
+	 */
+	if (((uflags & (XFS_GROUP_QUOTA | XFS_USER_QUOTA)) ==
+	     (XFS_GROUP_QUOTA | XFS_USER_QUOTA)) ||
+	    ((uflags & (XFS_GROUP_QUOTA | XFS_USER_QUOTA)) == 0))
+		return (0);
+
+	return (uflags & XFS_USER_QUOTA) ?
+		XFS_DQ_USER : XFS_DQ_GROUP;
+}
+
+STATIC uint
+xfs_qm_export_qtype_flags(
+	uint flags)
+{
+	/*
+	 * Can't be both at the same time.
+	 */
+	ASSERT((flags & (XFS_GROUP_QUOTA | XFS_USER_QUOTA)) !=
+		(XFS_GROUP_QUOTA | XFS_USER_QUOTA));
+	ASSERT((flags & (XFS_GROUP_QUOTA | XFS_USER_QUOTA)) != 0);
+
+	return (flags & XFS_DQ_USER) ?
+		XFS_USER_QUOTA : XFS_GROUP_QUOTA;
+}
+
+STATIC uint
+xfs_qm_import_flags(
+	uint uflags)
+{
+	uint flags = 0;
+
+	if (uflags & XFS_QUOTA_UDQ_ACCT)
+		flags |= XFS_UQUOTA_ACCT;
+	if (uflags & XFS_QUOTA_GDQ_ACCT)
+		flags |= XFS_GQUOTA_ACCT;
+	if (uflags & XFS_QUOTA_UDQ_ENFD)
+		flags |= XFS_UQUOTA_ENFD;
+	if (uflags & XFS_QUOTA_GDQ_ENFD)
+		flags |= XFS_GQUOTA_ENFD;
+	return (flags);
+}
+
+
+STATIC uint
+xfs_qm_export_flags(
+	uint flags)
+{
+	uint uflags;
+
+	uflags = 0;
+	if (flags & XFS_UQUOTA_ACCT)
+		uflags |= XFS_QUOTA_UDQ_ACCT;
+	if (flags & XFS_GQUOTA_ACCT)
+		uflags |= XFS_QUOTA_GDQ_ACCT;
+	if (flags & XFS_UQUOTA_ENFD)
+		uflags |= XFS_QUOTA_UDQ_ENFD;
+	if (flags & XFS_GQUOTA_ENFD)
+		uflags |= XFS_QUOTA_GDQ_ENFD;
+	return (uflags);
+}
+
+
+/*
+ * Go thru all the inodes in the file system, releasing their dquots.
+ * Note that the mount structure gets modified to indicate that quotas are off
+ * AFTER this, in the case of quotaoff. This also gets called from
+ * xfs_rootumount.
+ */
+void
+xfs_qm_dqrele_all_inodes(
+	struct xfs_mount *mp,
+	uint		 flags)
+{
+	vmap_t		vmap;
+	xfs_inode_t	*ip, *topino;
+	uint		ireclaims;
+	vnode_t		*vp;
+	boolean_t	vnode_refd;
+
+	ASSERT(mp->m_quotainfo);
+
+again:
+	XFS_MOUNT_ILOCK(mp);
+	ip = mp->m_inodes;
+	if (ip == NULL) {
+		XFS_MOUNT_IUNLOCK(mp);
+		return;
+	}
+	do {
+		/* Skip markers inserted by xfs_sync */
+		if (ip->i_mount == NULL) {
+			ip = ip->i_mnext;
+			continue;
+		}
+		/* Root inode, rbmip and rsumip have associated blocks */
+		if (ip == XFS_QI_UQIP(mp) || ip == XFS_QI_GQIP(mp)) {
+			ASSERT(ip->i_udquot == NULL);
+			ASSERT(ip->i_gdquot == NULL);
+			ip = ip->i_mnext;
+			continue;
+		}
+		vp = XFS_ITOV_NULL(ip);
+		if (!vp) {
+			ASSERT(ip->i_udquot == NULL);
+			ASSERT(ip->i_gdquot == NULL);
+			ip = ip->i_mnext;
+			continue;
+		}
+		vnode_refd = B_FALSE;
+		if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
+			/*
+			 * Sample vp mapping while holding the mplock, lest
+			 * we come across a non-existent vnode.
+			 */
+			VMAP(vp, vmap);
+			ireclaims = mp->m_ireclaims;
+			topino = mp->m_inodes;
+			XFS_MOUNT_IUNLOCK(mp);
+
+			/* XXX restart limit ? */
+			if ( ! (vp = vn_get(vp, &vmap)))
+				goto again;
+			xfs_ilock(ip, XFS_ILOCK_EXCL);
+			vnode_refd = B_TRUE;
+		} else {
+			ireclaims = mp->m_ireclaims;
+			topino = mp->m_inodes;
+			XFS_MOUNT_IUNLOCK(mp);
+		}
+
+		/*
+		 * We don't keep the mountlock across the dqrele() call,
+		 * since it can take a while..
+		 */
+		if ((flags & XFS_UQUOTA_ACCT) && ip->i_udquot) {
+			xfs_qm_dqrele(ip->i_udquot);
+			ip->i_udquot = NULL;
+		}
+		if ((flags & XFS_GQUOTA_ACCT) && ip->i_gdquot) {
+			xfs_qm_dqrele(ip->i_gdquot);
+			ip->i_gdquot = NULL;
+		}
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+		/*
+		 * Wait until we've dropped the ilock and mountlock to
+		 * do the vn_rele. Or be condemned to an eternity in the
+		 * inactive code in hell.
+		 */
+		if (vnode_refd)
+			VN_RELE(vp);
+		XFS_MOUNT_ILOCK(mp);
+		/*
+		 * If an inode was inserted or removed, we gotta
+		 * start over again.
+		 */
+		if (topino != mp->m_inodes || mp->m_ireclaims != ireclaims) {
+			/* XXX use a sentinel */
+			XFS_MOUNT_IUNLOCK(mp);
+			goto again;
+		}
+		ip = ip->i_mnext;
+	} while (ip != mp->m_inodes);
+
+	XFS_MOUNT_IUNLOCK(mp);
+}
+
+/*------------------------------------------------------------------------*/
+#ifdef DEBUG
+/*
+ * This contains all the test functions for XFS disk quotas.
+ * Currently it does a quota accounting check. ie. it walks through
+ * all inodes in the file system, calculating the dquot accounting fields,
+ * and prints out any inconsistencies.
+ */
+xfs_dqhash_t *qmtest_udqtab;
+xfs_dqhash_t *qmtest_gdqtab;
+int	      qmtest_hashmask;
+int	      qmtest_nfails;
+mutex_t	      qcheck_lock;
+
+#define DQTEST_HASHVAL(mp, id) (((__psunsigned_t)(mp) + \
+				 (__psunsigned_t)(id)) & \
+				(qmtest_hashmask - 1))
+
+#define DQTEST_HASH(mp, id, type)   ((type & XFS_DQ_USER) ? \
+				     (qmtest_udqtab + \
+				      DQTEST_HASHVAL(mp, id)) : \
+				     (qmtest_gdqtab + \
+				      DQTEST_HASHVAL(mp, id)))
+
+#define DQTEST_LIST_PRINT(l, NXT, title) \
+{ \
+	  xfs_dqtest_t	*dqp; int i = 0;\
+	  cmn_err(CE_DEBUG, "%s (#%d)", title, (int) (l)->qh_nelems); \
+	  for (dqp = (xfs_dqtest_t *)(l)->qh_next; dqp != NULL; \
+	       dqp = (xfs_dqtest_t *)dqp->NXT) { \
+		cmn_err(CE_DEBUG, "  %d. \"%d (%s)\"  bcnt = %d, icnt = %d", \
+			 ++i, dqp->d_id, DQFLAGTO_TYPESTR(dqp),	     \
+			 dqp->d_bcount, dqp->d_icount); } \
+}
+
+typedef struct dqtest {
+	xfs_dqmarker_t	q_lists;
+	xfs_dqhash_t	*q_hash;	/* the hashchain header */
+	xfs_mount_t	*q_mount;	/* filesystem this relates to */
+	xfs_dqid_t	d_id;		/* user id or group id */
+	xfs_qcnt_t	d_bcount;	/* # disk blocks owned by the user */
+	xfs_qcnt_t	d_icount;	/* # inodes owned by the user */
+} xfs_dqtest_t;
+
+STATIC void
+xfs_qm_hashinsert(xfs_dqhash_t *h, xfs_dqtest_t *dqp)
+{
+	xfs_dquot_t *d;
+	if (((d) = (h)->qh_next))
+		(d)->HL_PREVP = &((dqp)->HL_NEXT);
+	(dqp)->HL_NEXT = d;
+	(dqp)->HL_PREVP = &((h)->qh_next);
+	(h)->qh_next = (xfs_dquot_t *)dqp;
+	(h)->qh_version++;
+	(h)->qh_nelems++;
+}
+STATIC void
+xfs_qm_dqtest_print(
+	xfs_dqtest_t	*d)
+{
+	cmn_err(CE_DEBUG, "-----------DQTEST DQUOT----------------");
+	cmn_err(CE_DEBUG, "---- dquot ID = %d", d->d_id);
+	cmn_err(CE_DEBUG, "---- type     = %s", XFS_QM_ISUDQ(d)? "USR" : "GRP");
+	cmn_err(CE_DEBUG, "---- fs       = 0x%p", d->q_mount);
+	cmn_err(CE_DEBUG, "---- bcount   = %Lu (0x%x)",
+		d->d_bcount, (int)d->d_bcount);
+	cmn_err(CE_DEBUG, "---- icount   = %Lu (0x%x)",
+		d->d_icount, (int)d->d_icount);
+	cmn_err(CE_DEBUG, "---------------------------");
+}
+
+STATIC void
+xfs_qm_dqtest_failed(
+	xfs_dqtest_t	*d,
+	xfs_dquot_t	*dqp,
+	char		*reason,
+	xfs_qcnt_t	a,
+	xfs_qcnt_t	b,
+	int		error)
+{
+	qmtest_nfails++;
+	if (error)
+		cmn_err(CE_DEBUG, "quotacheck failed id=%d, err=%d\nreason: %s",
+		       INT_GET(d->d_id, ARCH_CONVERT), error, reason);
+	else
+		cmn_err(CE_DEBUG, "quotacheck failed id=%d (%s) [%d != %d]",
+		       INT_GET(d->d_id, ARCH_CONVERT), reason, (int)a, (int)b);
+	xfs_qm_dqtest_print(d);
+	if (dqp)
+		xfs_qm_dqprint(dqp);
+}
+
+STATIC int
+xfs_dqtest_cmp2(
+	xfs_dqtest_t	*d,
+	xfs_dquot_t	*dqp)
+{
+	int err = 0;
+	if (INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) != d->d_icount) {
+		xfs_qm_dqtest_failed(d, dqp, "icount mismatch",
+			INT_GET(dqp->q_core.d_icount, ARCH_CONVERT),
+			d->d_icount, 0);
+		err++;
+	}
+	if (INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT) != d->d_bcount) {
+		xfs_qm_dqtest_failed(d, dqp, "bcount mismatch",
+			INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT),
+			d->d_bcount, 0);
+		err++;
+	}
+	if (INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT) &&
+	    INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT) >=
+	    INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT)) {
+		if (!dqp->q_core.d_btimer && dqp->q_core.d_id) {
+			cmn_err(CE_DEBUG,
+				"%d [%s] [0x%p] BLK TIMER NOT STARTED",
+				d->d_id, DQFLAGTO_TYPESTR(d), d->q_mount);
+			err++;
+		}
+	}
+	if (INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT) &&
+	    INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >=
+	    INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT)) {
+		if (!dqp->q_core.d_itimer && dqp->q_core.d_id) {
+			cmn_err(CE_DEBUG,
+				"%d [%s] [0x%p] INO TIMER NOT STARTED",
+				d->d_id, DQFLAGTO_TYPESTR(d), d->q_mount);
+			err++;
+		}
+	}
+#ifdef QUOTADEBUG
+	if (!err) {
+		cmn_err(CE_DEBUG, "%d [%s] [0x%p] qchecked",
+			d->d_id, XFS_QM_ISUDQ(d) ? "USR" : "GRP", d->q_mount);
+	}
+#endif
+	return (err);
+}
+
+STATIC void
+xfs_dqtest_cmp(
+	xfs_dqtest_t	*d)
+{
+	xfs_dquot_t	*dqp;
+	int		error;
+
+	/* xfs_qm_dqtest_print(d); */
+	if ((error = xfs_qm_dqget(d->q_mount, NULL, d->d_id, d->dq_flags, 0,
+				 &dqp))) {
+		xfs_qm_dqtest_failed(d, NULL, "dqget failed", 0, 0, error);
+		return;
+	}
+	xfs_dqtest_cmp2(d, dqp);
+	xfs_qm_dqput(dqp);
+}
+
+STATIC int
+xfs_qm_internalqcheck_dqget(
+	xfs_mount_t	*mp,
+	xfs_dqid_t	id,
+	uint		type,
+	xfs_dqtest_t	**O_dq)
+{
+	xfs_dqtest_t	*d;
+	xfs_dqhash_t	*h;
+
+	h = DQTEST_HASH(mp, id, type);
+	for (d = (xfs_dqtest_t *) h->qh_next; d != NULL;
+	     d = (xfs_dqtest_t *) d->HL_NEXT) {
+		/* DQTEST_LIST_PRINT(h, HL_NEXT, "@@@@@ dqtestlist @@@@@"); */
+		if (d->d_id == id && mp == d->q_mount) {
+			*O_dq = d;
+			return (0);
+		}
+	}
+	d = kmem_zalloc(sizeof(xfs_dqtest_t), KM_SLEEP);
+	d->dq_flags = type;
+	d->d_id = id;
+	d->q_mount = mp;
+	d->q_hash = h;
+	xfs_qm_hashinsert(h, d);
+	*O_dq = d;
+	return (0);
+}
+
+STATIC void
+xfs_qm_internalqcheck_get_dquots(
+	xfs_mount_t	*mp,
+	xfs_dqid_t	uid,
+	xfs_dqid_t	gid,
+	xfs_dqtest_t	**ud,
+	xfs_dqtest_t	**gd)
+{
+	if (XFS_IS_UQUOTA_ON(mp))
+		xfs_qm_internalqcheck_dqget(mp, uid, XFS_DQ_USER, ud);
+	if (XFS_IS_GQUOTA_ON(mp))
+		xfs_qm_internalqcheck_dqget(mp, gid, XFS_DQ_GROUP, gd);
+}
+
+
+STATIC void
+xfs_qm_internalqcheck_dqadjust(
+	xfs_inode_t		*ip,
+	xfs_dqtest_t		*d)
+{
+	d->d_icount++;
+	d->d_bcount += (xfs_qcnt_t)ip->i_d.di_nblocks;
+}
+
+STATIC int
+xfs_qm_internalqcheck_adjust(
+	xfs_mount_t	*mp,		/* mount point for filesystem */
+	xfs_ino_t	ino,		/* inode number to get data for */
+	void		__user *buffer,	/* not used */
+	int		ubsize,		/* not used */
+	void		*private_data,	/* not used */
+	xfs_daddr_t	bno,		/* starting block of inode cluster */
+	int		*ubused,	/* not used */
+	void		*dip,		/* not used */
+	int		*res)		/* bulkstat result code */
+{
+	xfs_inode_t		*ip;
+	xfs_dqtest_t		*ud, *gd;
+	uint			lock_flags;
+	boolean_t		ipreleased;
+	int			error;
+
+	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+
+	if (ino == mp->m_sb.sb_uquotino || ino == mp->m_sb.sb_gquotino) {
+		*res = BULKSTAT_RV_NOTHING;
+		qdprintk("internalqcheck: ino=%llu, uqino=%llu, gqino=%llu\n",
+			(unsigned long long) ino,
+			(unsigned long long) mp->m_sb.sb_uquotino,
+			(unsigned long long) mp->m_sb.sb_gquotino);
+		return XFS_ERROR(EINVAL);
+	}
+	ipreleased = B_FALSE;
+ again:
+	lock_flags = XFS_ILOCK_SHARED;
+	if ((error = xfs_iget(mp, NULL, ino, 0, lock_flags, &ip, bno))) {
+		*res = BULKSTAT_RV_NOTHING;
+		return (error);
+	}
+
+	if (ip->i_d.di_mode == 0) {
+		xfs_iput_new(ip, lock_flags);
+		*res = BULKSTAT_RV_NOTHING;
+		return XFS_ERROR(ENOENT);
+	}
+
+	/*
+	 * This inode can have blocks after eof which can get released
+	 * when we send it to inactive. Since we don't check the dquot
+	 * until the after all our calculations are done, we must get rid
+	 * of those now.
+	 */
+	if (! ipreleased) {
+		xfs_iput(ip, lock_flags);
+		ipreleased = B_TRUE;
+		goto again;
+	}
+	xfs_qm_internalqcheck_get_dquots(mp,
+					(xfs_dqid_t) ip->i_d.di_uid,
+					(xfs_dqid_t) ip->i_d.di_gid,
+					&ud, &gd);
+	if (XFS_IS_UQUOTA_ON(mp)) {
+		ASSERT(ud);
+		xfs_qm_internalqcheck_dqadjust(ip, ud);
+	}
+	if (XFS_IS_GQUOTA_ON(mp)) {
+		ASSERT(gd);
+		xfs_qm_internalqcheck_dqadjust(ip, gd);
+	}
+	xfs_iput(ip, lock_flags);
+	*res = BULKSTAT_RV_DIDONE;
+	return (0);
+}
+
+
+/* PRIVATE, debugging */
+int
+xfs_qm_internalqcheck(
+	xfs_mount_t	*mp)
+{
+	xfs_ino_t	lastino;
+	int		done, count;
+	int		i;
+	xfs_dqtest_t	*d, *e;
+	xfs_dqhash_t	*h1;
+	int		error;
+
+	lastino = 0;
+	qmtest_hashmask = 32;
+	count = 5;
+	done = 0;
+	qmtest_nfails = 0;
+
+	if (! XFS_IS_QUOTA_ON(mp))
+		return XFS_ERROR(ESRCH);
+
+	xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE | XFS_LOG_SYNC);
+	XFS_bflush(mp->m_ddev_targp);
+	xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE | XFS_LOG_SYNC);
+	XFS_bflush(mp->m_ddev_targp);
+
+	mutex_lock(&qcheck_lock, PINOD);
+	/* There should be absolutely no quota activity while this
+	   is going on. */
+	qmtest_udqtab = kmem_zalloc(qmtest_hashmask *
+				    sizeof(xfs_dqhash_t), KM_SLEEP);
+	qmtest_gdqtab = kmem_zalloc(qmtest_hashmask *
+				    sizeof(xfs_dqhash_t), KM_SLEEP);
+	do {
+		/*
+		 * Iterate thru all the inodes in the file system,
+		 * adjusting the corresponding dquot counters
+		 */
+		if ((error = xfs_bulkstat(mp, &lastino, &count,
+				 xfs_qm_internalqcheck_adjust, NULL,
+				 0, NULL, BULKSTAT_FG_IGET, &done))) {
+			break;
+		}
+	} while (! done);
+	if (error) {
+		cmn_err(CE_DEBUG, "Bulkstat returned error 0x%x", error);
+	}
+	cmn_err(CE_DEBUG, "Checking results against system dquots");
+	for (i = 0; i < qmtest_hashmask; i++) {
+		h1 = &qmtest_udqtab[i];
+		for (d = (xfs_dqtest_t *) h1->qh_next; d != NULL; ) {
+			xfs_dqtest_cmp(d);
+			e = (xfs_dqtest_t *) d->HL_NEXT;
+			kmem_free(d, sizeof(xfs_dqtest_t));
+			d = e;
+		}
+		h1 = &qmtest_gdqtab[i];
+		for (d = (xfs_dqtest_t *) h1->qh_next; d != NULL; ) {
+			xfs_dqtest_cmp(d);
+			e = (xfs_dqtest_t *) d->HL_NEXT;
+			kmem_free(d, sizeof(xfs_dqtest_t));
+			d = e;
+		}
+	}
+
+	if (qmtest_nfails) {
+		cmn_err(CE_DEBUG, "******** quotacheck failed  ********");
+		cmn_err(CE_DEBUG, "failures = %d", qmtest_nfails);
+	} else {
+		cmn_err(CE_DEBUG, "******** quotacheck successful! ********");
+	}
+	kmem_free(qmtest_udqtab, qmtest_hashmask * sizeof(xfs_dqhash_t));
+	kmem_free(qmtest_gdqtab, qmtest_hashmask * sizeof(xfs_dqhash_t));
+	mutex_unlock(&qcheck_lock);
+	return (qmtest_nfails);
+}
+
+#endif /* DEBUG */
diff --git a/fs/xfs/quota/xfs_quota_priv.h b/fs/xfs/quota/xfs_quota_priv.h
new file mode 100644
index 000000000000..414b6004af21
--- /dev/null
+++ b/fs/xfs/quota/xfs_quota_priv.h
@@ -0,0 +1,192 @@
+/*
+ * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+#ifndef __XFS_QUOTA_PRIV_H__
+#define __XFS_QUOTA_PRIV_H__
+
+/*
+ * Number of bmaps that we ask from bmapi when doing a quotacheck.
+ * We make this restriction to keep the memory usage to a minimum.
+ */
+#define XFS_DQITER_MAP_SIZE	10
+
+/* Number of dquots that fit in to a dquot block */
+#define XFS_QM_DQPERBLK(mp)	((mp)->m_quotainfo->qi_dqperchunk)
+
+#define XFS_ISLOCKED_INODE(ip)		(ismrlocked(&(ip)->i_lock, \
+					    MR_UPDATE | MR_ACCESS) != 0)
+#define XFS_ISLOCKED_INODE_EXCL(ip)	(ismrlocked(&(ip)->i_lock, \
+					    MR_UPDATE) != 0)
+
+#define XFS_DQ_IS_ADDEDTO_TRX(t, d)	((d)->q_transp == (t))
+
+#define XFS_QI_MPLRECLAIMS(mp)	((mp)->m_quotainfo->qi_dqreclaims)
+#define XFS_QI_UQIP(mp)		((mp)->m_quotainfo->qi_uquotaip)
+#define XFS_QI_GQIP(mp)		((mp)->m_quotainfo->qi_gquotaip)
+#define XFS_QI_DQCHUNKLEN(mp)	((mp)->m_quotainfo->qi_dqchunklen)
+#define XFS_QI_BTIMELIMIT(mp)	((mp)->m_quotainfo->qi_btimelimit)
+#define XFS_QI_RTBTIMELIMIT(mp) ((mp)->m_quotainfo->qi_rtbtimelimit)
+#define XFS_QI_ITIMELIMIT(mp)	((mp)->m_quotainfo->qi_itimelimit)
+#define XFS_QI_BWARNLIMIT(mp)	((mp)->m_quotainfo->qi_bwarnlimit)
+#define XFS_QI_IWARNLIMIT(mp)	((mp)->m_quotainfo->qi_iwarnlimit)
+#define XFS_QI_QOFFLOCK(mp)	((mp)->m_quotainfo->qi_quotaofflock)
+
+#define XFS_QI_MPL_LIST(mp)	((mp)->m_quotainfo->qi_dqlist)
+#define XFS_QI_MPLLOCK(mp)	((mp)->m_quotainfo->qi_dqlist.qh_lock)
+#define XFS_QI_MPLNEXT(mp)	((mp)->m_quotainfo->qi_dqlist.qh_next)
+#define XFS_QI_MPLNDQUOTS(mp)	((mp)->m_quotainfo->qi_dqlist.qh_nelems)
+
+#define XQMLCK(h)			(mutex_lock(&((h)->qh_lock), PINOD))
+#define XQMUNLCK(h)			(mutex_unlock(&((h)->qh_lock)))
+#ifdef DEBUG
+struct xfs_dqhash;
+static inline int XQMISLCKD(struct xfs_dqhash *h)
+{
+	if (mutex_trylock(&h->qh_lock)) {
+		mutex_unlock(&h->qh_lock);
+		return 0;
+	}
+	return 1;
+}
+#endif
+
+#define XFS_DQ_HASH_LOCK(h)		XQMLCK(h)
+#define XFS_DQ_HASH_UNLOCK(h)		XQMUNLCK(h)
+#define XFS_DQ_IS_HASH_LOCKED(h)	XQMISLCKD(h)
+
+#define xfs_qm_mplist_lock(mp)		XQMLCK(&(XFS_QI_MPL_LIST(mp)))
+#define xfs_qm_mplist_unlock(mp)	XQMUNLCK(&(XFS_QI_MPL_LIST(mp)))
+#define XFS_QM_IS_MPLIST_LOCKED(mp)	XQMISLCKD(&(XFS_QI_MPL_LIST(mp)))
+
+#define xfs_qm_freelist_lock(qm)	XQMLCK(&((qm)->qm_dqfreelist))
+#define xfs_qm_freelist_unlock(qm)	XQMUNLCK(&((qm)->qm_dqfreelist))
+#define XFS_QM_IS_FREELIST_LOCKED(qm)	XQMISLCKD(&((qm)->qm_dqfreelist))
+
+/*
+ * Hash into a bucket in the dquot hash table, based on <mp, id>.
+ */
+#define XFS_DQ_HASHVAL(mp, id) (((__psunsigned_t)(mp) + \
+				 (__psunsigned_t)(id)) & \
+				(xfs_Gqm->qm_dqhashmask - 1))
+#define XFS_DQ_HASH(mp, id, type)   (type == XFS_DQ_USER ? \
+				     (xfs_Gqm->qm_usr_dqhtable + \
+				      XFS_DQ_HASHVAL(mp, id)) : \
+				     (xfs_Gqm->qm_grp_dqhtable + \
+				      XFS_DQ_HASHVAL(mp, id)))
+#define XFS_IS_DQTYPE_ON(mp, type)   (type == XFS_DQ_USER ? \
+				      XFS_IS_UQUOTA_ON(mp):XFS_IS_GQUOTA_ON(mp))
+#define XFS_IS_DQUOT_UNINITIALIZED(dqp) ( \
+	!dqp->q_core.d_blk_hardlimit && \
+	!dqp->q_core.d_blk_softlimit && \
+	!dqp->q_core.d_rtb_hardlimit && \
+	!dqp->q_core.d_rtb_softlimit && \
+	!dqp->q_core.d_ino_hardlimit && \
+	!dqp->q_core.d_ino_softlimit && \
+	!dqp->q_core.d_bcount && \
+	!dqp->q_core.d_rtbcount && \
+	!dqp->q_core.d_icount)
+
+#define HL_PREVP	dq_hashlist.ql_prevp
+#define HL_NEXT		dq_hashlist.ql_next
+#define MPL_PREVP	dq_mplist.ql_prevp
+#define MPL_NEXT	dq_mplist.ql_next
+
+
+#define _LIST_REMOVE(h, dqp, PVP, NXT)				\
+	{							\
+		 xfs_dquot_t *d;				\
+		 if (((d) = (dqp)->NXT))				\
+			 (d)->PVP = (dqp)->PVP;			\
+		 *((dqp)->PVP) = d;				\
+		 (dqp)->NXT = NULL;				\
+		 (dqp)->PVP = NULL;				\
+		 (h)->qh_version++;				\
+		 (h)->qh_nelems--;				\
+	}
+
+#define _LIST_INSERT(h, dqp, PVP, NXT)				\
+	{							\
+		 xfs_dquot_t *d;				\
+		 if (((d) = (h)->qh_next))			\
+			 (d)->PVP = &((dqp)->NXT);		\
+		 (dqp)->NXT = d;				\
+		 (dqp)->PVP = &((h)->qh_next);			\
+		 (h)->qh_next = dqp;				\
+		 (h)->qh_version++;				\
+		 (h)->qh_nelems++;				\
+	 }
+
+#define FOREACH_DQUOT_IN_MP(dqp, mp) \
+	for ((dqp) = XFS_QI_MPLNEXT(mp); (dqp) != NULL; (dqp) = (dqp)->MPL_NEXT)
+
+#define FOREACH_DQUOT_IN_FREELIST(dqp, qlist)	\
+for ((dqp) = (qlist)->qh_next; (dqp) != (xfs_dquot_t *)(qlist); \
+     (dqp) = (dqp)->dq_flnext)
+
+#define XQM_HASHLIST_INSERT(h, dqp)	\
+	 _LIST_INSERT(h, dqp, HL_PREVP, HL_NEXT)
+
+#define XQM_FREELIST_INSERT(h, dqp)	\
+	 xfs_qm_freelist_append(h, dqp)
+
+#define XQM_MPLIST_INSERT(h, dqp)	\
+	 _LIST_INSERT(h, dqp, MPL_PREVP, MPL_NEXT)
+
+#define XQM_HASHLIST_REMOVE(h, dqp)	\
+	 _LIST_REMOVE(h, dqp, HL_PREVP, HL_NEXT)
+#define XQM_FREELIST_REMOVE(dqp)	\
+	 xfs_qm_freelist_unlink(dqp)
+#define XQM_MPLIST_REMOVE(h, dqp)	\
+	{ _LIST_REMOVE(h, dqp, MPL_PREVP, MPL_NEXT); \
+	  XFS_QI_MPLRECLAIMS((dqp)->q_mount)++; }
+
+#define XFS_DQ_IS_LOGITEM_INITD(dqp)	((dqp)->q_logitem.qli_dquot == (dqp))
+
+#define XFS_QM_DQP_TO_DQACCT(tp, dqp)	(XFS_QM_ISUDQ(dqp) ? \
+					 (tp)->t_dqinfo->dqa_usrdquots : \
+					 (tp)->t_dqinfo->dqa_grpdquots)
+#define XFS_IS_SUSER_DQUOT(dqp)		\
+	(!((dqp)->q_core.d_id))
+
+#define XFS_PURGE_INODE(ip)		\
+	{				\
+	  vmap_t dqvmap;		\
+	  vnode_t *dqvp;		\
+	  dqvp = XFS_ITOV(ip);		\
+	  VMAP(dqvp, dqvmap);		\
+	  VN_RELE(dqvp);		\
+	}
+
+#define DQFLAGTO_TYPESTR(d)	(((d)->dq_flags & XFS_DQ_USER) ? "USR" : \
+				 (((d)->dq_flags & XFS_DQ_GROUP) ? "GRP" : "???"))
+#define DQFLAGTO_DIRTYSTR(d)	(XFS_DQ_IS_DIRTY(d) ? "DIRTY" : "NOTDIRTY")
+
+#endif	/* __XFS_QUOTA_PRIV_H__ */
diff --git a/fs/xfs/quota/xfs_trans_dquot.c b/fs/xfs/quota/xfs_trans_dquot.c
new file mode 100644
index 000000000000..149b2a1fd949
--- /dev/null
+++ b/fs/xfs/quota/xfs_trans_dquot.c
@@ -0,0 +1,941 @@
+/*
+ * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.	 Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_inum.h"
+#include "xfs_log.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_ag.h"
+#include "xfs_dir.h"
+#include "xfs_dir2.h"
+#include "xfs_alloc.h"
+#include "xfs_dmapi.h"
+#include "xfs_quota.h"
+#include "xfs_mount.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_attr_sf.h"
+#include "xfs_dir_sf.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_dinode.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_bit.h"
+#include "xfs_rtalloc.h"
+#include "xfs_error.h"
+#include "xfs_itable.h"
+#include "xfs_rw.h"
+#include "xfs_acl.h"
+#include "xfs_cap.h"
+#include "xfs_mac.h"
+#include "xfs_attr.h"
+#include "xfs_buf_item.h"
+#include "xfs_trans_priv.h"
+
+#include "xfs_qm.h"
+
+STATIC void	xfs_trans_alloc_dqinfo(xfs_trans_t *);
+
+/*
+ * Add the locked dquot to the transaction.
+ * The dquot must be locked, and it cannot be associated with any
+ * transaction.
+ */
+void
+xfs_trans_dqjoin(
+	xfs_trans_t	*tp,
+	xfs_dquot_t	*dqp)
+{
+	xfs_dq_logitem_t    *lp;
+
+	ASSERT(! XFS_DQ_IS_ADDEDTO_TRX(tp, dqp));
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	ASSERT(XFS_DQ_IS_LOGITEM_INITD(dqp));
+	lp = &dqp->q_logitem;
+
+	/*
+	 * Get a log_item_desc to point at the new item.
+	 */
+	(void) xfs_trans_add_item(tp, (xfs_log_item_t*)(lp));
+
+	/*
+	 * Initialize i_transp so we can later determine if this dquot is
+	 * associated with this transaction.
+	 */
+	dqp->q_transp = tp;
+}
+
+
+/*
+ * This is called to mark the dquot as needing
+ * to be logged when the transaction is committed.  The dquot must
+ * already be associated with the given transaction.
+ * Note that it marks the entire transaction as dirty. In the ordinary
+ * case, this gets called via xfs_trans_commit, after the transaction
+ * is already dirty. However, there's nothing stop this from getting
+ * called directly, as done by xfs_qm_scall_setqlim. Hence, the TRANS_DIRTY
+ * flag.
+ */
+void
+xfs_trans_log_dquot(
+	xfs_trans_t	*tp,
+	xfs_dquot_t	*dqp)
+{
+	xfs_log_item_desc_t	*lidp;
+
+	ASSERT(XFS_DQ_IS_ADDEDTO_TRX(tp, dqp));
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+
+	lidp = xfs_trans_find_item(tp, (xfs_log_item_t*)(&dqp->q_logitem));
+	ASSERT(lidp != NULL);
+
+	tp->t_flags |= XFS_TRANS_DIRTY;
+	lidp->lid_flags |= XFS_LID_DIRTY;
+}
+
+/*
+ * Carry forward whatever is left of the quota blk reservation to
+ * the spanky new transaction
+ */
+STATIC void
+xfs_trans_dup_dqinfo(
+	xfs_trans_t	*otp,
+	xfs_trans_t	*ntp)
+{
+	xfs_dqtrx_t	*oq, *nq;
+	int		i,j;
+	xfs_dqtrx_t	*oqa, *nqa;
+
+	if (!otp->t_dqinfo)
+		return;
+
+	xfs_trans_alloc_dqinfo(ntp);
+	oqa = otp->t_dqinfo->dqa_usrdquots;
+	nqa = ntp->t_dqinfo->dqa_usrdquots;
+
+	/*
+	 * Because the quota blk reservation is carried forward,
+	 * it is also necessary to carry forward the DQ_DIRTY flag.
+	 */
+	if(otp->t_flags & XFS_TRANS_DQ_DIRTY)
+		ntp->t_flags |= XFS_TRANS_DQ_DIRTY;
+
+	for (j = 0; j < 2; j++) {
+		for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
+			if (oqa[i].qt_dquot == NULL)
+				break;
+			oq = &oqa[i];
+			nq = &nqa[i];
+
+			nq->qt_dquot = oq->qt_dquot;
+			nq->qt_bcount_delta = nq->qt_icount_delta = 0;
+			nq->qt_rtbcount_delta = 0;
+
+			/*
+			 * Transfer whatever is left of the reservations.
+			 */
+			nq->qt_blk_res = oq->qt_blk_res - oq->qt_blk_res_used;
+			oq->qt_blk_res = oq->qt_blk_res_used;
+
+			nq->qt_rtblk_res = oq->qt_rtblk_res -
+				oq->qt_rtblk_res_used;
+			oq->qt_rtblk_res = oq->qt_rtblk_res_used;
+
+			nq->qt_ino_res = oq->qt_ino_res - oq->qt_ino_res_used;
+			oq->qt_ino_res = oq->qt_ino_res_used;
+
+		}
+		oqa = otp->t_dqinfo->dqa_grpdquots;
+		nqa = ntp->t_dqinfo->dqa_grpdquots;
+	}
+}
+
+/*
+ * Wrap around mod_dquot to account for both user and group quotas.
+ */
+void
+xfs_trans_mod_dquot_byino(
+	xfs_trans_t	*tp,
+	xfs_inode_t	*ip,
+	uint		field,
+	long		delta)
+{
+	xfs_mount_t	*mp;
+
+	ASSERT(tp);
+	mp = tp->t_mountp;
+
+	if (!XFS_IS_QUOTA_ON(mp) ||
+	    ip->i_ino == mp->m_sb.sb_uquotino ||
+	    ip->i_ino == mp->m_sb.sb_gquotino)
+		return;
+
+	if (tp->t_dqinfo == NULL)
+		xfs_trans_alloc_dqinfo(tp);
+
+	if (XFS_IS_UQUOTA_ON(mp) && ip->i_udquot) {
+		(void) xfs_trans_mod_dquot(tp, ip->i_udquot, field, delta);
+	}
+	if (XFS_IS_GQUOTA_ON(mp) && ip->i_gdquot) {
+		(void) xfs_trans_mod_dquot(tp, ip->i_gdquot, field, delta);
+	}
+}
+
+STATIC xfs_dqtrx_t *
+xfs_trans_get_dqtrx(
+	xfs_trans_t	*tp,
+	xfs_dquot_t	*dqp)
+{
+	int		i;
+	xfs_dqtrx_t	*qa;
+
+	for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
+		qa = XFS_QM_DQP_TO_DQACCT(tp, dqp);
+
+		if (qa[i].qt_dquot == NULL ||
+		    qa[i].qt_dquot == dqp) {
+			return (&qa[i]);
+		}
+	}
+
+	return (NULL);
+}
+
+/*
+ * Make the changes in the transaction structure.
+ * The moral equivalent to xfs_trans_mod_sb().
+ * We don't touch any fields in the dquot, so we don't care
+ * if it's locked or not (most of the time it won't be).
+ */
+void
+xfs_trans_mod_dquot(
+	xfs_trans_t	*tp,
+	xfs_dquot_t	*dqp,
+	uint		field,
+	long		delta)
+{
+	xfs_dqtrx_t	*qtrx;
+
+	ASSERT(tp);
+	qtrx = NULL;
+
+	if (tp->t_dqinfo == NULL)
+		xfs_trans_alloc_dqinfo(tp);
+	/*
+	 * Find either the first free slot or the slot that belongs
+	 * to this dquot.
+	 */
+	qtrx = xfs_trans_get_dqtrx(tp, dqp);
+	ASSERT(qtrx);
+	if (qtrx->qt_dquot == NULL)
+		qtrx->qt_dquot = dqp;
+
+	switch (field) {
+
+		/*
+		 * regular disk blk reservation
+		 */
+	      case XFS_TRANS_DQ_RES_BLKS:
+		qtrx->qt_blk_res += (ulong)delta;
+		break;
+
+		/*
+		 * inode reservation
+		 */
+	      case XFS_TRANS_DQ_RES_INOS:
+		qtrx->qt_ino_res += (ulong)delta;
+		break;
+
+		/*
+		 * disk blocks used.
+		 */
+	      case XFS_TRANS_DQ_BCOUNT:
+		if (qtrx->qt_blk_res && delta > 0) {
+			qtrx->qt_blk_res_used += (ulong)delta;
+			ASSERT(qtrx->qt_blk_res >= qtrx->qt_blk_res_used);
+		}
+		qtrx->qt_bcount_delta += delta;
+		break;
+
+	      case XFS_TRANS_DQ_DELBCOUNT:
+		qtrx->qt_delbcnt_delta += delta;
+		break;
+
+		/*
+		 * Inode Count
+		 */
+	      case XFS_TRANS_DQ_ICOUNT:
+		if (qtrx->qt_ino_res && delta > 0) {
+			qtrx->qt_ino_res_used += (ulong)delta;
+			ASSERT(qtrx->qt_ino_res >= qtrx->qt_ino_res_used);
+		}
+		qtrx->qt_icount_delta += delta;
+		break;
+
+		/*
+		 * rtblk reservation
+		 */
+	      case XFS_TRANS_DQ_RES_RTBLKS:
+		qtrx->qt_rtblk_res += (ulong)delta;
+		break;
+
+		/*
+		 * rtblk count
+		 */
+	      case XFS_TRANS_DQ_RTBCOUNT:
+		if (qtrx->qt_rtblk_res && delta > 0) {
+			qtrx->qt_rtblk_res_used += (ulong)delta;
+			ASSERT(qtrx->qt_rtblk_res >= qtrx->qt_rtblk_res_used);
+		}
+		qtrx->qt_rtbcount_delta += delta;
+		break;
+
+	      case XFS_TRANS_DQ_DELRTBCOUNT:
+		qtrx->qt_delrtb_delta += delta;
+		break;
+
+	      default:
+		ASSERT(0);
+	}
+	tp->t_flags |= XFS_TRANS_DQ_DIRTY;
+}
+
+
+/*
+ * Given an array of dqtrx structures, lock all the dquots associated
+ * and join them to the transaction, provided they have been modified.
+ * We know that the highest number of dquots (of one type - usr OR grp),
+ * involved in a transaction is 2 and that both usr and grp combined - 3.
+ * So, we don't attempt to make this very generic.
+ */
+STATIC void
+xfs_trans_dqlockedjoin(
+	xfs_trans_t	*tp,
+	xfs_dqtrx_t	*q)
+{
+	ASSERT(q[0].qt_dquot != NULL);
+	if (q[1].qt_dquot == NULL) {
+		xfs_dqlock(q[0].qt_dquot);
+		xfs_trans_dqjoin(tp, q[0].qt_dquot);
+	} else {
+		ASSERT(XFS_QM_TRANS_MAXDQS == 2);
+		xfs_dqlock2(q[0].qt_dquot, q[1].qt_dquot);
+		xfs_trans_dqjoin(tp, q[0].qt_dquot);
+		xfs_trans_dqjoin(tp, q[1].qt_dquot);
+	}
+}
+
+
+/*
+ * Called by xfs_trans_commit() and similar in spirit to
+ * xfs_trans_apply_sb_deltas().
+ * Go thru all the dquots belonging to this transaction and modify the
+ * INCORE dquot to reflect the actual usages.
+ * Unreserve just the reservations done by this transaction.
+ * dquot is still left locked at exit.
+ */
+void
+xfs_trans_apply_dquot_deltas(
+	xfs_trans_t		*tp)
+{
+	int			i, j;
+	xfs_dquot_t		*dqp;
+	xfs_dqtrx_t		*qtrx, *qa;
+	xfs_disk_dquot_t	*d;
+	long			totalbdelta;
+	long			totalrtbdelta;
+
+	if (! (tp->t_flags & XFS_TRANS_DQ_DIRTY))
+		return;
+
+	ASSERT(tp->t_dqinfo);
+	qa = tp->t_dqinfo->dqa_usrdquots;
+	for (j = 0; j < 2; j++) {
+		if (qa[0].qt_dquot == NULL) {
+			qa = tp->t_dqinfo->dqa_grpdquots;
+			continue;
+		}
+
+		/*
+		 * Lock all of the dquots and join them to the transaction.
+		 */
+		xfs_trans_dqlockedjoin(tp, qa);
+
+		for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
+			qtrx = &qa[i];
+			/*
+			 * The array of dquots is filled
+			 * sequentially, not sparsely.
+			 */
+			if ((dqp = qtrx->qt_dquot) == NULL)
+				break;
+
+			ASSERT(XFS_DQ_IS_LOCKED(dqp));
+			ASSERT(XFS_DQ_IS_ADDEDTO_TRX(tp, dqp));
+
+			/*
+			 * adjust the actual number of blocks used
+			 */
+			d = &dqp->q_core;
+
+			/*
+			 * The issue here is - sometimes we don't make a blkquota
+			 * reservation intentionally to be fair to users
+			 * (when the amount is small). On the other hand,
+			 * delayed allocs do make reservations, but that's
+			 * outside of a transaction, so we have no
+			 * idea how much was really reserved.
+			 * So, here we've accumulated delayed allocation blks and
+			 * non-delay blks. The assumption is that the
+			 * delayed ones are always reserved (outside of a
+			 * transaction), and the others may or may not have
+			 * quota reservations.
+			 */
+			totalbdelta = qtrx->qt_bcount_delta +
+				qtrx->qt_delbcnt_delta;
+			totalrtbdelta = qtrx->qt_rtbcount_delta +
+				qtrx->qt_delrtb_delta;
+#ifdef QUOTADEBUG
+			if (totalbdelta < 0)
+				ASSERT(INT_GET(d->d_bcount, ARCH_CONVERT) >=
+				       (xfs_qcnt_t) -totalbdelta);
+
+			if (totalrtbdelta < 0)
+				ASSERT(INT_GET(d->d_rtbcount, ARCH_CONVERT) >=
+				       (xfs_qcnt_t) -totalrtbdelta);
+
+			if (qtrx->qt_icount_delta < 0)
+				ASSERT(INT_GET(d->d_icount, ARCH_CONVERT) >=
+				       (xfs_qcnt_t) -qtrx->qt_icount_delta);
+#endif
+			if (totalbdelta)
+				INT_MOD(d->d_bcount, ARCH_CONVERT, (xfs_qcnt_t)totalbdelta);
+
+			if (qtrx->qt_icount_delta)
+				INT_MOD(d->d_icount, ARCH_CONVERT, (xfs_qcnt_t)qtrx->qt_icount_delta);
+
+			if (totalrtbdelta)
+				INT_MOD(d->d_rtbcount, ARCH_CONVERT, (xfs_qcnt_t)totalrtbdelta);
+
+			/*
+			 * Get any default limits in use.
+			 * Start/reset the timer(s) if needed.
+			 */
+			if (d->d_id) {
+				xfs_qm_adjust_dqlimits(tp->t_mountp, d);
+				xfs_qm_adjust_dqtimers(tp->t_mountp, d);
+			}
+
+			dqp->dq_flags |= XFS_DQ_DIRTY;
+			/*
+			 * add this to the list of items to get logged
+			 */
+			xfs_trans_log_dquot(tp, dqp);
+			/*
+			 * Take off what's left of the original reservation.
+			 * In case of delayed allocations, there's no
+			 * reservation that a transaction structure knows of.
+			 */
+			if (qtrx->qt_blk_res != 0) {
+				if (qtrx->qt_blk_res != qtrx->qt_blk_res_used) {
+					if (qtrx->qt_blk_res >
+					    qtrx->qt_blk_res_used)
+						dqp->q_res_bcount -= (xfs_qcnt_t)
+							(qtrx->qt_blk_res -
+							 qtrx->qt_blk_res_used);
+					else
+						dqp->q_res_bcount -= (xfs_qcnt_t)
+							(qtrx->qt_blk_res_used -
+							 qtrx->qt_blk_res);
+				}
+			} else {
+				/*
+				 * These blks were never reserved, either inside
+				 * a transaction or outside one (in a delayed
+				 * allocation). Also, this isn't always a
+				 * negative number since we sometimes
+				 * deliberately skip quota reservations.
+				 */
+				if (qtrx->qt_bcount_delta) {
+					dqp->q_res_bcount +=
+					      (xfs_qcnt_t)qtrx->qt_bcount_delta;
+				}
+			}
+			/*
+			 * Adjust the RT reservation.
+			 */
+			if (qtrx->qt_rtblk_res != 0) {
+				if (qtrx->qt_blk_res != qtrx->qt_blk_res_used) {
+					if (qtrx->qt_rtblk_res >
+					    qtrx->qt_rtblk_res_used)
+					       dqp->q_res_rtbcount -= (xfs_qcnt_t)
+						       (qtrx->qt_rtblk_res -
+							qtrx->qt_rtblk_res_used);
+					else
+					       dqp->q_res_rtbcount -= (xfs_qcnt_t)
+						       (qtrx->qt_rtblk_res_used -
+							qtrx->qt_rtblk_res);
+				}
+			} else {
+				if (qtrx->qt_rtbcount_delta)
+					dqp->q_res_rtbcount +=
+					    (xfs_qcnt_t)qtrx->qt_rtbcount_delta;
+			}
+
+			/*
+			 * Adjust the inode reservation.
+			 */
+			if (qtrx->qt_ino_res != 0) {
+				ASSERT(qtrx->qt_ino_res >=
+				       qtrx->qt_ino_res_used);
+				if (qtrx->qt_ino_res > qtrx->qt_ino_res_used)
+					dqp->q_res_icount -= (xfs_qcnt_t)
+						(qtrx->qt_ino_res -
+						 qtrx->qt_ino_res_used);
+			} else {
+				if (qtrx->qt_icount_delta)
+					dqp->q_res_icount +=
+					    (xfs_qcnt_t)qtrx->qt_icount_delta;
+			}
+
+
+#ifdef QUOTADEBUG
+			if (qtrx->qt_rtblk_res != 0)
+				cmn_err(CE_DEBUG, "RT res %d for 0x%p\n",
+					(int) qtrx->qt_rtblk_res, dqp);
+#endif
+			ASSERT(dqp->q_res_bcount >=
+				INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT));
+			ASSERT(dqp->q_res_icount >=
+				INT_GET(dqp->q_core.d_icount, ARCH_CONVERT));
+			ASSERT(dqp->q_res_rtbcount >=
+				INT_GET(dqp->q_core.d_rtbcount, ARCH_CONVERT));
+		}
+		/*
+		 * Do the group quotas next
+		 */
+		qa = tp->t_dqinfo->dqa_grpdquots;
+	}
+}
+
+/*
+ * Release the reservations, and adjust the dquots accordingly.
+ * This is called only when the transaction is being aborted. If by
+ * any chance we have done dquot modifications incore (ie. deltas) already,
+ * we simply throw those away, since that's the expected behavior
+ * when a transaction is curtailed without a commit.
+ */
+STATIC void
+xfs_trans_unreserve_and_mod_dquots(
+	xfs_trans_t		*tp)
+{
+	int			i, j;
+	xfs_dquot_t		*dqp;
+	xfs_dqtrx_t		*qtrx, *qa;
+	boolean_t		locked;
+
+	if (!tp->t_dqinfo || !(tp->t_flags & XFS_TRANS_DQ_DIRTY))
+		return;
+
+	qa = tp->t_dqinfo->dqa_usrdquots;
+
+	for (j = 0; j < 2; j++) {
+		for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
+			qtrx = &qa[i];
+			/*
+			 * We assume that the array of dquots is filled
+			 * sequentially, not sparsely.
+			 */
+			if ((dqp = qtrx->qt_dquot) == NULL)
+				break;
+			/*
+			 * Unreserve the original reservation. We don't care
+			 * about the number of blocks used field, or deltas.
+			 * Also we don't bother to zero the fields.
+			 */
+			locked = B_FALSE;
+			if (qtrx->qt_blk_res) {
+				xfs_dqlock(dqp);
+				locked = B_TRUE;
+				dqp->q_res_bcount -=
+					(xfs_qcnt_t)qtrx->qt_blk_res;
+			}
+			if (qtrx->qt_ino_res) {
+				if (!locked) {
+					xfs_dqlock(dqp);
+					locked = B_TRUE;
+				}
+				dqp->q_res_icount -=
+					(xfs_qcnt_t)qtrx->qt_ino_res;
+			}
+
+			if (qtrx->qt_rtblk_res) {
+				if (!locked) {
+					xfs_dqlock(dqp);
+					locked = B_TRUE;
+				}
+				dqp->q_res_rtbcount -=
+					(xfs_qcnt_t)qtrx->qt_rtblk_res;
+			}
+			if (locked)
+				xfs_dqunlock(dqp);
+
+		}
+		qa = tp->t_dqinfo->dqa_grpdquots;
+	}
+}
+
+/*
+ * This reserves disk blocks and inodes against a dquot.
+ * Flags indicate if the dquot is to be locked here and also
+ * if the blk reservation is for RT or regular blocks.
+ * Sending in XFS_QMOPT_FORCE_RES flag skips the quota check.
+ * Returns EDQUOT if quota is exceeded.
+ */
+STATIC int
+xfs_trans_dqresv(
+	xfs_trans_t	*tp,
+	xfs_mount_t	*mp,
+	xfs_dquot_t	*dqp,
+	long		nblks,
+	long		ninos,
+	uint		flags)
+{
+	int		error;
+	xfs_qcnt_t	hardlimit;
+	xfs_qcnt_t	softlimit;
+	time_t		btimer;
+	xfs_qcnt_t	*resbcountp;
+	xfs_quotainfo_t	*q = mp->m_quotainfo;
+
+	if (! (flags & XFS_QMOPT_DQLOCK)) {
+		xfs_dqlock(dqp);
+	}
+	ASSERT(XFS_DQ_IS_LOCKED(dqp));
+	if (flags & XFS_TRANS_DQ_RES_BLKS) {
+		hardlimit = INT_GET(dqp->q_core.d_blk_hardlimit, ARCH_CONVERT);
+		if (!hardlimit)
+			hardlimit = q->qi_bhardlimit;
+		softlimit = INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT);
+		if (!softlimit)
+			softlimit = q->qi_bsoftlimit;
+		btimer = INT_GET(dqp->q_core.d_btimer, ARCH_CONVERT);
+		resbcountp = &dqp->q_res_bcount;
+	} else {
+		ASSERT(flags & XFS_TRANS_DQ_RES_RTBLKS);
+		hardlimit = INT_GET(dqp->q_core.d_rtb_hardlimit, ARCH_CONVERT);
+		if (!hardlimit)
+			hardlimit = q->qi_rtbhardlimit;
+		softlimit = INT_GET(dqp->q_core.d_rtb_softlimit, ARCH_CONVERT);
+		if (!softlimit)
+			softlimit = q->qi_rtbsoftlimit;
+		btimer = INT_GET(dqp->q_core.d_rtbtimer, ARCH_CONVERT);
+		resbcountp = &dqp->q_res_rtbcount;
+	}
+	error = 0;
+
+	if ((flags & XFS_QMOPT_FORCE_RES) == 0 &&
+	    dqp->q_core.d_id &&
+	    XFS_IS_QUOTA_ENFORCED(dqp->q_mount)) {
+#ifdef QUOTADEBUG
+		cmn_err(CE_DEBUG, "BLK Res: nblks=%ld + resbcount=%Ld"
+			  " > hardlimit=%Ld?", nblks, *resbcountp, hardlimit);
+#endif
+		if (nblks > 0) {
+			/*
+			 * dquot is locked already. See if we'd go over the
+			 * hardlimit or exceed the timelimit if we allocate
+			 * nblks.
+			 */
+			if (hardlimit > 0ULL &&
+			     (hardlimit <= nblks + *resbcountp)) {
+				error = EDQUOT;
+				goto error_return;
+			}
+
+			if (softlimit > 0ULL &&
+			     (softlimit <= nblks + *resbcountp)) {
+				/*
+				 * If timer or warnings has expired,
+				 * return EDQUOT
+				 */
+				if ((btimer != 0 && get_seconds() > btimer) ||
+				    (dqp->q_core.d_bwarns &&
+				     INT_GET(dqp->q_core.d_bwarns, ARCH_CONVERT) >=
+				     XFS_QI_BWARNLIMIT(dqp->q_mount))) {
+					error = EDQUOT;
+					goto error_return;
+				}
+			}
+		}
+		if (ninos > 0) {
+			hardlimit = INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT);
+			if (!hardlimit)
+				hardlimit = q->qi_ihardlimit;
+			softlimit = INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT);
+			if (!softlimit)
+				softlimit = q->qi_isoftlimit;
+			if (hardlimit > 0ULL &&
+			    INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >= hardlimit) {
+				error = EDQUOT;
+				goto error_return;
+			} else if (softlimit > 0ULL &&
+				   INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >= softlimit) {
+				/*
+				 * If timer or warnings has expired,
+				 * return EDQUOT
+				 */
+				if ((dqp->q_core.d_itimer &&
+				     get_seconds() > INT_GET(dqp->q_core.d_itimer, ARCH_CONVERT)) ||
+				    (dqp->q_core.d_iwarns &&
+				     INT_GET(dqp->q_core.d_iwarns, ARCH_CONVERT) >=
+				     XFS_QI_IWARNLIMIT(dqp->q_mount))) {
+					error = EDQUOT;
+					goto error_return;
+				}
+			}
+		}
+	}
+
+	/*
+	 * Change the reservation, but not the actual usage.
+	 * Note that q_res_bcount = q_core.d_bcount + resv
+	 */
+	(*resbcountp) += (xfs_qcnt_t)nblks;
+	if (ninos != 0)
+		dqp->q_res_icount += (xfs_qcnt_t)ninos;
+
+	/*
+	 * note the reservation amt in the trans struct too,
+	 * so that the transaction knows how much was reserved by
+	 * it against this particular dquot.
+	 * We don't do this when we are reserving for a delayed allocation,
+	 * because we don't have the luxury of a transaction envelope then.
+	 */
+	if (tp) {
+		ASSERT(tp->t_dqinfo);
+		ASSERT(flags & XFS_QMOPT_RESBLK_MASK);
+		if (nblks != 0)
+			xfs_trans_mod_dquot(tp, dqp,
+					    flags & XFS_QMOPT_RESBLK_MASK,
+					    nblks);
+		if (ninos != 0)
+			xfs_trans_mod_dquot(tp, dqp,
+					    XFS_TRANS_DQ_RES_INOS,
+					    ninos);
+	}
+	ASSERT(dqp->q_res_bcount >= INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT));
+	ASSERT(dqp->q_res_rtbcount >= INT_GET(dqp->q_core.d_rtbcount, ARCH_CONVERT));
+	ASSERT(dqp->q_res_icount >= INT_GET(dqp->q_core.d_icount, ARCH_CONVERT));
+
+error_return:
+	if (! (flags & XFS_QMOPT_DQLOCK)) {
+		xfs_dqunlock(dqp);
+	}
+	return (error);
+}
+
+
+/*
+ * Given a dquot(s), make disk block and/or inode reservations against them.
+ * The fact that this does the reservation against both the usr and
+ * grp quotas is important, because this follows a both-or-nothing
+ * approach.
+ *
+ * flags = XFS_QMOPT_DQLOCK indicate if dquot(s) need to be locked.
+ *	   XFS_QMOPT_FORCE_RES evades limit enforcement. Used by chown.
+ *	   XFS_TRANS_DQ_RES_BLKS reserves regular disk blocks
+ *	   XFS_TRANS_DQ_RES_RTBLKS reserves realtime disk blocks
+ * dquots are unlocked on return, if they were not locked by caller.
+ */
+int
+xfs_trans_reserve_quota_bydquots(
+	xfs_trans_t	*tp,
+	xfs_mount_t	*mp,
+	xfs_dquot_t	*udqp,
+	xfs_dquot_t	*gdqp,
+	long		nblks,
+	long		ninos,
+	uint		flags)
+{
+	int		resvd;
+
+	if (! XFS_IS_QUOTA_ON(mp))
+		return (0);
+
+	if (tp && tp->t_dqinfo == NULL)
+		xfs_trans_alloc_dqinfo(tp);
+
+	ASSERT(flags & XFS_QMOPT_RESBLK_MASK);
+	resvd = 0;
+
+	if (udqp) {
+		if (xfs_trans_dqresv(tp, mp, udqp, nblks, ninos, flags))
+			return (EDQUOT);
+		resvd = 1;
+	}
+
+	if (gdqp) {
+		if (xfs_trans_dqresv(tp, mp, gdqp, nblks, ninos, flags)) {
+			/*
+			 * can't do it, so backout previous reservation
+			 */
+			if (resvd) {
+				flags |= XFS_QMOPT_FORCE_RES;
+				xfs_trans_dqresv(tp, mp, udqp,
+						 -nblks, -ninos, flags);
+			}
+			return (EDQUOT);
+		}
+	}
+
+	/*
+	 * Didnt change anything critical, so, no need to log
+	 */
+	return (0);
+}
+
+
+/*
+ * Lock the dquot and change the reservation if we can.
+ * This doesn't change the actual usage, just the reservation.
+ * The inode sent in is locked.
+ *
+ * Returns 0 on success, EDQUOT or other errors otherwise
+ */
+STATIC int
+xfs_trans_reserve_quota_nblks(
+	xfs_trans_t	*tp,
+	xfs_mount_t	*mp,
+	xfs_inode_t	*ip,
+	long		nblks,
+	long		ninos,
+	uint		type)
+{
+	int		error;
+
+	if (!XFS_IS_QUOTA_ON(mp))
+		return (0);
+
+	ASSERT(ip->i_ino != mp->m_sb.sb_uquotino);
+	ASSERT(ip->i_ino != mp->m_sb.sb_gquotino);
+
+	ASSERT(XFS_ISLOCKED_INODE_EXCL(ip));
+	ASSERT(XFS_IS_QUOTA_RUNNING(ip->i_mount));
+	ASSERT((type & ~XFS_QMOPT_FORCE_RES) == XFS_TRANS_DQ_RES_RTBLKS ||
+	       (type & ~XFS_QMOPT_FORCE_RES) == XFS_TRANS_DQ_RES_BLKS);
+
+	/*
+	 * Reserve nblks against these dquots, with trans as the mediator.
+	 */
+	error = xfs_trans_reserve_quota_bydquots(tp, mp,
+						 ip->i_udquot, ip->i_gdquot,
+						 nblks, ninos,
+						 type);
+	return (error);
+}
+
+/*
+ * This routine is called to allocate a quotaoff log item.
+ */
+xfs_qoff_logitem_t *
+xfs_trans_get_qoff_item(
+	xfs_trans_t		*tp,
+	xfs_qoff_logitem_t	*startqoff,
+	uint			flags)
+{
+	xfs_qoff_logitem_t	*q;
+
+	ASSERT(tp != NULL);
+
+	q = xfs_qm_qoff_logitem_init(tp->t_mountp, startqoff, flags);
+	ASSERT(q != NULL);
+
+	/*
+	 * Get a log_item_desc to point at the new item.
+	 */
+	(void) xfs_trans_add_item(tp, (xfs_log_item_t*)q);
+
+	return (q);
+}
+
+
+/*
+ * This is called to mark the quotaoff logitem as needing
+ * to be logged when the transaction is committed.  The logitem must
+ * already be associated with the given transaction.
+ */
+void
+xfs_trans_log_quotaoff_item(
+	xfs_trans_t		*tp,
+	xfs_qoff_logitem_t	*qlp)
+{
+	xfs_log_item_desc_t	*lidp;
+
+	lidp = xfs_trans_find_item(tp, (xfs_log_item_t *)qlp);
+	ASSERT(lidp != NULL);
+
+	tp->t_flags |= XFS_TRANS_DIRTY;
+	lidp->lid_flags |= XFS_LID_DIRTY;
+}
+
+STATIC void
+xfs_trans_alloc_dqinfo(
+	xfs_trans_t	*tp)
+{
+	(tp)->t_dqinfo = kmem_zone_zalloc(xfs_Gqm->qm_dqtrxzone, KM_SLEEP);
+}
+
+STATIC void
+xfs_trans_free_dqinfo(
+	xfs_trans_t	*tp)
+{
+	if (!tp->t_dqinfo)
+		return;
+	kmem_zone_free(xfs_Gqm->qm_dqtrxzone, (tp)->t_dqinfo);
+	(tp)->t_dqinfo = NULL;
+}
+
+xfs_dqtrxops_t	xfs_trans_dquot_ops = {
+	.qo_dup_dqinfo			= xfs_trans_dup_dqinfo,
+	.qo_free_dqinfo			= xfs_trans_free_dqinfo,
+	.qo_mod_dquot_byino		= xfs_trans_mod_dquot_byino,
+	.qo_apply_dquot_deltas		= xfs_trans_apply_dquot_deltas,
+	.qo_reserve_quota_nblks		= xfs_trans_reserve_quota_nblks,
+	.qo_reserve_quota_bydquots	= xfs_trans_reserve_quota_bydquots,
+	.qo_unreserve_and_mod_dquots	= xfs_trans_unreserve_and_mod_dquots,
+};