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!

1 files changed, 3131 insertions, 0 deletions

diff --git a/fs/jfs/jfs_txnmgr.c b/fs/jfs/jfs_txnmgr.c
new file mode 100644
index 00000000000..f40301d93f7
--- /dev/null
+++ b/fs/jfs/jfs_txnmgr.c
@@ -0,0 +1,3131 @@
+/*
+ *   Copyright (C) International Business Machines Corp., 2000-2005
+ *   Portions Copyright (C) Christoph Hellwig, 2001-2002
+ *
+ *   This program is free software;  you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation; either version 2 of the License, or 
+ *   (at your option) any later version.
+ * 
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ *   the GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program;  if not, write to the Free Software 
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+/*
+ *      jfs_txnmgr.c: transaction manager
+ *
+ * notes:
+ * transaction starts with txBegin() and ends with txCommit()
+ * or txAbort().
+ *
+ * tlock is acquired at the time of update;
+ * (obviate scan at commit time for xtree and dtree)
+ * tlock and mp points to each other;
+ * (no hashlist for mp -> tlock).
+ *
+ * special cases:
+ * tlock on in-memory inode:
+ * in-place tlock in the in-memory inode itself;
+ * converted to page lock by iWrite() at commit time.
+ *
+ * tlock during write()/mmap() under anonymous transaction (tid = 0):
+ * transferred (?) to transaction at commit time.
+ *
+ * use the page itself to update allocation maps
+ * (obviate intermediate replication of allocation/deallocation data)
+ * hold on to mp+lock thru update of maps
+ */
+
+
+#include <linux/fs.h>
+#include <linux/vmalloc.h>
+#include <linux/smp_lock.h>
+#include <linux/completion.h>
+#include <linux/suspend.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include "jfs_incore.h"
+#include "jfs_filsys.h"
+#include "jfs_metapage.h"
+#include "jfs_dinode.h"
+#include "jfs_imap.h"
+#include "jfs_dmap.h"
+#include "jfs_superblock.h"
+#include "jfs_debug.h"
+
+/*
+ *      transaction management structures
+ */
+static struct {
+	int freetid;		/* index of a free tid structure */
+	int freelock;		/* index first free lock word */
+	wait_queue_head_t freewait;	/* eventlist of free tblock */
+	wait_queue_head_t freelockwait;	/* eventlist of free tlock */
+	wait_queue_head_t lowlockwait;	/* eventlist of ample tlocks */
+	int tlocksInUse;	/* Number of tlocks in use */
+	spinlock_t LazyLock;	/* synchronize sync_queue & unlock_queue */
+/*	struct tblock *sync_queue; * Transactions waiting for data sync */
+	struct list_head unlock_queue;	/* Txns waiting to be released */
+	struct list_head anon_list;	/* inodes having anonymous txns */
+	struct list_head anon_list2;	/* inodes having anonymous txns
+					   that couldn't be sync'ed */
+} TxAnchor;
+
+int jfs_tlocks_low;		/* Indicates low number of available tlocks */
+
+#ifdef CONFIG_JFS_STATISTICS
+static struct {
+	uint txBegin;
+	uint txBegin_barrier;
+	uint txBegin_lockslow;
+	uint txBegin_freetid;
+	uint txBeginAnon;
+	uint txBeginAnon_barrier;
+	uint txBeginAnon_lockslow;
+	uint txLockAlloc;
+	uint txLockAlloc_freelock;
+} TxStat;
+#endif
+
+static int nTxBlock = -1;	/* number of transaction blocks */
+module_param(nTxBlock, int, 0);
+MODULE_PARM_DESC(nTxBlock,
+		 "Number of transaction blocks (max:65536)");
+
+static int nTxLock = -1;	/* number of transaction locks */
+module_param(nTxLock, int, 0);
+MODULE_PARM_DESC(nTxLock,
+		 "Number of transaction locks (max:65536)");
+
+struct tblock *TxBlock;	        /* transaction block table */
+static int TxLockLWM;		/* Low water mark for number of txLocks used */
+static int TxLockHWM;		/* High water mark for number of txLocks used */
+static int TxLockVHWM;		/* Very High water mark */
+struct tlock *TxLock;           /* transaction lock table */
+
+
+/*
+ *      transaction management lock
+ */
+static DEFINE_SPINLOCK(jfsTxnLock);
+
+#define TXN_LOCK()              spin_lock(&jfsTxnLock)
+#define TXN_UNLOCK()            spin_unlock(&jfsTxnLock)
+
+#define LAZY_LOCK_INIT()	spin_lock_init(&TxAnchor.LazyLock);
+#define LAZY_LOCK(flags)	spin_lock_irqsave(&TxAnchor.LazyLock, flags)
+#define LAZY_UNLOCK(flags) spin_unlock_irqrestore(&TxAnchor.LazyLock, flags)
+
+DECLARE_WAIT_QUEUE_HEAD(jfs_sync_thread_wait);
+DECLARE_WAIT_QUEUE_HEAD(jfs_commit_thread_wait);
+static int jfs_commit_thread_waking;
+
+/*
+ * Retry logic exist outside these macros to protect from spurrious wakeups.
+ */
+static inline void TXN_SLEEP_DROP_LOCK(wait_queue_head_t * event)
+{
+	DECLARE_WAITQUEUE(wait, current);
+
+	add_wait_queue(event, &wait);
+	set_current_state(TASK_UNINTERRUPTIBLE);
+	TXN_UNLOCK();
+	schedule();
+	current->state = TASK_RUNNING;
+	remove_wait_queue(event, &wait);
+}
+
+#define TXN_SLEEP(event)\
+{\
+	TXN_SLEEP_DROP_LOCK(event);\
+	TXN_LOCK();\
+}
+
+#define TXN_WAKEUP(event) wake_up_all(event)
+
+
+/*
+ *      statistics
+ */
+static struct {
+	tid_t maxtid;		/* 4: biggest tid ever used */
+	lid_t maxlid;		/* 4: biggest lid ever used */
+	int ntid;		/* 4: # of transactions performed */
+	int nlid;		/* 4: # of tlocks acquired */
+	int waitlock;		/* 4: # of tlock wait */
+} stattx;
+
+
+/*
+ * external references
+ */
+extern int lmGroupCommit(struct jfs_log *, struct tblock *);
+extern int jfs_commit_inode(struct inode *, int);
+extern int jfs_stop_threads;
+
+extern struct completion jfsIOwait;
+
+/*
+ * forward references
+ */
+static int diLog(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
+		struct tlock * tlck, struct commit * cd);
+static int dataLog(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
+		struct tlock * tlck);
+static void dtLog(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
+		struct tlock * tlck);
+static void mapLog(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
+		struct tlock * tlck);
+static void txAllocPMap(struct inode *ip, struct maplock * maplock,
+		struct tblock * tblk);
+static void txForce(struct tblock * tblk);
+static int txLog(struct jfs_log * log, struct tblock * tblk,
+		struct commit * cd);
+static void txUpdateMap(struct tblock * tblk);
+static void txRelease(struct tblock * tblk);
+static void xtLog(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
+	   struct tlock * tlck);
+static void LogSyncRelease(struct metapage * mp);
+
+/*
+ *              transaction block/lock management
+ *              ---------------------------------
+ */
+
+/*
+ * Get a transaction lock from the free list.  If the number in use is
+ * greater than the high water mark, wake up the sync daemon.  This should
+ * free some anonymous transaction locks.  (TXN_LOCK must be held.)
+ */
+static lid_t txLockAlloc(void)
+{
+	lid_t lid;
+
+	INCREMENT(TxStat.txLockAlloc);
+	if (!TxAnchor.freelock) {
+		INCREMENT(TxStat.txLockAlloc_freelock);
+	}
+
+	while (!(lid = TxAnchor.freelock))
+		TXN_SLEEP(&TxAnchor.freelockwait);
+	TxAnchor.freelock = TxLock[lid].next;
+	HIGHWATERMARK(stattx.maxlid, lid);
+	if ((++TxAnchor.tlocksInUse > TxLockHWM) && (jfs_tlocks_low == 0)) {
+		jfs_info("txLockAlloc tlocks low");
+		jfs_tlocks_low = 1;
+		wake_up(&jfs_sync_thread_wait);
+	}
+
+	return lid;
+}
+
+static void txLockFree(lid_t lid)
+{
+	TxLock[lid].next = TxAnchor.freelock;
+	TxAnchor.freelock = lid;
+	TxAnchor.tlocksInUse--;
+	if (jfs_tlocks_low && (TxAnchor.tlocksInUse < TxLockLWM)) {
+		jfs_info("txLockFree jfs_tlocks_low no more");
+		jfs_tlocks_low = 0;
+		TXN_WAKEUP(&TxAnchor.lowlockwait);
+	}
+	TXN_WAKEUP(&TxAnchor.freelockwait);
+}
+
+/*
+ * NAME:        txInit()
+ *
+ * FUNCTION:    initialize transaction management structures
+ *
+ * RETURN:
+ *
+ * serialization: single thread at jfs_init()
+ */
+int txInit(void)
+{
+	int k, size;
+	struct sysinfo si;
+
+	/* Set defaults for nTxLock and nTxBlock if unset */
+
+	if (nTxLock == -1) {
+		if (nTxBlock == -1) {
+			/* Base default on memory size */
+			si_meminfo(&si);
+			if (si.totalram > (256 * 1024)) /* 1 GB */
+				nTxLock = 64 * 1024;
+			else
+				nTxLock = si.totalram >> 2;
+		} else if (nTxBlock > (8 * 1024))
+			nTxLock = 64 * 1024;
+		else
+			nTxLock = nTxBlock << 3;
+	}
+	if (nTxBlock == -1)
+		nTxBlock = nTxLock >> 3;
+
+	/* Verify tunable parameters */
+	if (nTxBlock < 16)
+		nTxBlock = 16;	/* No one should set it this low */
+	if (nTxBlock > 65536)
+		nTxBlock = 65536;
+	if (nTxLock < 256)
+		nTxLock = 256;	/* No one should set it this low */
+	if (nTxLock > 65536)
+		nTxLock = 65536;
+
+	printk(KERN_INFO "JFS: nTxBlock = %d, nTxLock = %d\n",
+	       nTxBlock, nTxLock);
+	/*
+	 * initialize transaction block (tblock) table
+	 *
+	 * transaction id (tid) = tblock index
+	 * tid = 0 is reserved.
+	 */
+	TxLockLWM = (nTxLock * 4) / 10;
+	TxLockHWM = (nTxLock * 7) / 10;
+	TxLockVHWM = (nTxLock * 8) / 10;
+
+	size = sizeof(struct tblock) * nTxBlock;
+	TxBlock = (struct tblock *) vmalloc(size);
+	if (TxBlock == NULL)
+		return -ENOMEM;
+
+	for (k = 1; k < nTxBlock - 1; k++) {
+		TxBlock[k].next = k + 1;
+		init_waitqueue_head(&TxBlock[k].gcwait);
+		init_waitqueue_head(&TxBlock[k].waitor);
+	}
+	TxBlock[k].next = 0;
+	init_waitqueue_head(&TxBlock[k].gcwait);
+	init_waitqueue_head(&TxBlock[k].waitor);
+
+	TxAnchor.freetid = 1;
+	init_waitqueue_head(&TxAnchor.freewait);
+
+	stattx.maxtid = 1;	/* statistics */
+
+	/*
+	 * initialize transaction lock (tlock) table
+	 *
+	 * transaction lock id = tlock index
+	 * tlock id = 0 is reserved.
+	 */
+	size = sizeof(struct tlock) * nTxLock;
+	TxLock = (struct tlock *) vmalloc(size);
+	if (TxLock == NULL) {
+		vfree(TxBlock);
+		return -ENOMEM;
+	}
+
+	/* initialize tlock table */
+	for (k = 1; k < nTxLock - 1; k++)
+		TxLock[k].next = k + 1;
+	TxLock[k].next = 0;
+	init_waitqueue_head(&TxAnchor.freelockwait);
+	init_waitqueue_head(&TxAnchor.lowlockwait);
+
+	TxAnchor.freelock = 1;
+	TxAnchor.tlocksInUse = 0;
+	INIT_LIST_HEAD(&TxAnchor.anon_list);
+	INIT_LIST_HEAD(&TxAnchor.anon_list2);
+
+	LAZY_LOCK_INIT();
+	INIT_LIST_HEAD(&TxAnchor.unlock_queue);
+
+	stattx.maxlid = 1;	/* statistics */
+
+	return 0;
+}
+
+/*
+ * NAME:        txExit()
+ *
+ * FUNCTION:    clean up when module is unloaded
+ */
+void txExit(void)
+{
+	vfree(TxLock);
+	TxLock = NULL;
+	vfree(TxBlock);
+	TxBlock = NULL;
+}
+
+
+/*
+ * NAME:        txBegin()
+ *
+ * FUNCTION:    start a transaction.
+ *
+ * PARAMETER:   sb	- superblock
+ *              flag	- force for nested tx;
+ *
+ * RETURN:	tid	- transaction id
+ *
+ * note: flag force allows to start tx for nested tx
+ * to prevent deadlock on logsync barrier;
+ */
+tid_t txBegin(struct super_block *sb, int flag)
+{
+	tid_t t;
+	struct tblock *tblk;
+	struct jfs_log *log;
+
+	jfs_info("txBegin: flag = 0x%x", flag);
+	log = JFS_SBI(sb)->log;
+
+	TXN_LOCK();
+
+	INCREMENT(TxStat.txBegin);
+
+      retry:
+	if (!(flag & COMMIT_FORCE)) {
+		/*
+		 * synchronize with logsync barrier
+		 */
+		if (test_bit(log_SYNCBARRIER, &log->flag) ||
+		    test_bit(log_QUIESCE, &log->flag)) {
+			INCREMENT(TxStat.txBegin_barrier);
+			TXN_SLEEP(&log->syncwait);
+			goto retry;
+		}
+	}
+	if (flag == 0) {
+		/*
+		 * Don't begin transaction if we're getting starved for tlocks
+		 * unless COMMIT_FORCE or COMMIT_INODE (which may ultimately
+		 * free tlocks)
+		 */
+		if (TxAnchor.tlocksInUse > TxLockVHWM) {
+			INCREMENT(TxStat.txBegin_lockslow);
+			TXN_SLEEP(&TxAnchor.lowlockwait);
+			goto retry;
+		}
+	}
+
+	/*
+	 * allocate transaction id/block
+	 */
+	if ((t = TxAnchor.freetid) == 0) {
+		jfs_info("txBegin: waiting for free tid");
+		INCREMENT(TxStat.txBegin_freetid);
+		TXN_SLEEP(&TxAnchor.freewait);
+		goto retry;
+	}
+
+	tblk = tid_to_tblock(t);
+
+	if ((tblk->next == 0) && !(flag & COMMIT_FORCE)) {
+		/* Don't let a non-forced transaction take the last tblk */
+		jfs_info("txBegin: waiting for free tid");
+		INCREMENT(TxStat.txBegin_freetid);
+		TXN_SLEEP(&TxAnchor.freewait);
+		goto retry;
+	}
+
+	TxAnchor.freetid = tblk->next;
+
+	/*
+	 * initialize transaction
+	 */
+
+	/*
+	 * We can't zero the whole thing or we screw up another thread being
+	 * awakened after sleeping on tblk->waitor
+	 *
+	 * memset(tblk, 0, sizeof(struct tblock));
+	 */
+	tblk->next = tblk->last = tblk->xflag = tblk->flag = tblk->lsn = 0;
+
+	tblk->sb = sb;
+	++log->logtid;
+	tblk->logtid = log->logtid;
+
+	++log->active;
+
+	HIGHWATERMARK(stattx.maxtid, t);	/* statistics */
+	INCREMENT(stattx.ntid);	/* statistics */
+
+	TXN_UNLOCK();
+
+	jfs_info("txBegin: returning tid = %d", t);
+
+	return t;
+}
+
+
+/*
+ * NAME:        txBeginAnon()
+ *
+ * FUNCTION:    start an anonymous transaction.
+ *		Blocks if logsync or available tlocks are low to prevent
+ *		anonymous tlocks from depleting supply.
+ *
+ * PARAMETER:   sb	- superblock
+ *
+ * RETURN:	none
+ */
+void txBeginAnon(struct super_block *sb)
+{
+	struct jfs_log *log;
+
+	log = JFS_SBI(sb)->log;
+
+	TXN_LOCK();
+	INCREMENT(TxStat.txBeginAnon);
+
+      retry:
+	/*
+	 * synchronize with logsync barrier
+	 */
+	if (test_bit(log_SYNCBARRIER, &log->flag) ||
+	    test_bit(log_QUIESCE, &log->flag)) {
+		INCREMENT(TxStat.txBeginAnon_barrier);
+		TXN_SLEEP(&log->syncwait);
+		goto retry;
+	}
+
+	/*
+	 * Don't begin transaction if we're getting starved for tlocks
+	 */
+	if (TxAnchor.tlocksInUse > TxLockVHWM) {
+		INCREMENT(TxStat.txBeginAnon_lockslow);
+		TXN_SLEEP(&TxAnchor.lowlockwait);
+		goto retry;
+	}
+	TXN_UNLOCK();
+}
+
+
+/*
+ *      txEnd()
+ *
+ * function: free specified transaction block.
+ *
+ *      logsync barrier processing:
+ *
+ * serialization:
+ */
+void txEnd(tid_t tid)
+{
+	struct tblock *tblk = tid_to_tblock(tid);
+	struct jfs_log *log;
+
+	jfs_info("txEnd: tid = %d", tid);
+	TXN_LOCK();
+
+	/*
+	 * wakeup transactions waiting on the page locked
+	 * by the current transaction
+	 */
+	TXN_WAKEUP(&tblk->waitor);
+
+	log = JFS_SBI(tblk->sb)->log;
+
+	/*
+	 * Lazy commit thread can't free this guy until we mark it UNLOCKED,
+	 * otherwise, we would be left with a transaction that may have been
+	 * reused.
+	 *
+	 * Lazy commit thread will turn off tblkGC_LAZY before calling this
+	 * routine.
+	 */
+	if (tblk->flag & tblkGC_LAZY) {
+		jfs_info("txEnd called w/lazy tid: %d, tblk = 0x%p", tid, tblk);
+		TXN_UNLOCK();
+
+		spin_lock_irq(&log->gclock);	// LOGGC_LOCK
+		tblk->flag |= tblkGC_UNLOCKED;
+		spin_unlock_irq(&log->gclock);	// LOGGC_UNLOCK
+		return;
+	}
+
+	jfs_info("txEnd: tid: %d, tblk = 0x%p", tid, tblk);
+
+	assert(tblk->next == 0);
+
+	/*
+	 * insert tblock back on freelist
+	 */
+	tblk->next = TxAnchor.freetid;
+	TxAnchor.freetid = tid;
+
+	/*
+	 * mark the tblock not active
+	 */
+	if (--log->active == 0) {
+		clear_bit(log_FLUSH, &log->flag);
+
+		/*
+		 * synchronize with logsync barrier
+		 */
+		if (test_bit(log_SYNCBARRIER, &log->flag)) {
+			/* forward log syncpt */
+			/* lmSync(log); */
+
+			jfs_info("log barrier off: 0x%x", log->lsn);
+
+			/* enable new transactions start */
+			clear_bit(log_SYNCBARRIER, &log->flag);
+
+			/* wakeup all waitors for logsync barrier */
+			TXN_WAKEUP(&log->syncwait);
+		}
+	}
+
+	/*
+	 * wakeup all waitors for a free tblock
+	 */
+	TXN_WAKEUP(&TxAnchor.freewait);
+
+	TXN_UNLOCK();
+}
+
+
+/*
+ *      txLock()
+ *
+ * function: acquire a transaction lock on the specified <mp>
+ *
+ * parameter:
+ *
+ * return:      transaction lock id
+ *
+ * serialization:
+ */
+struct tlock *txLock(tid_t tid, struct inode *ip, struct metapage * mp,
+		     int type)
+{
+	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
+	int dir_xtree = 0;
+	lid_t lid;
+	tid_t xtid;
+	struct tlock *tlck;
+	struct xtlock *xtlck;
+	struct linelock *linelock;
+	xtpage_t *p;
+	struct tblock *tblk;
+
+	TXN_LOCK();
+
+	if (S_ISDIR(ip->i_mode) && (type & tlckXTREE) &&
+	    !(mp->xflag & COMMIT_PAGE)) {
+		/*
+		 * Directory inode is special.  It can have both an xtree tlock
+		 * and a dtree tlock associated with it.
+		 */
+		dir_xtree = 1;
+		lid = jfs_ip->xtlid;
+	} else
+		lid = mp->lid;
+
+	/* is page not locked by a transaction ? */
+	if (lid == 0)
+		goto allocateLock;
+
+	jfs_info("txLock: tid:%d ip:0x%p mp:0x%p lid:%d", tid, ip, mp, lid);
+
+	/* is page locked by the requester transaction ? */
+	tlck = lid_to_tlock(lid);
+	if ((xtid = tlck->tid) == tid)
+		goto grantLock;
+
+	/*
+	 * is page locked by anonymous transaction/lock ?
+	 *
+	 * (page update without transaction (i.e., file write) is
+	 * locked under anonymous transaction tid = 0:
+	 * anonymous tlocks maintained on anonymous tlock list of
+	 * the inode of the page and available to all anonymous
+	 * transactions until txCommit() time at which point
+	 * they are transferred to the transaction tlock list of
+	 * the commiting transaction of the inode)
+	 */
+	if (xtid == 0) {
+		tlck->tid = tid;
+		tblk = tid_to_tblock(tid);
+		/*
+		 * The order of the tlocks in the transaction is important
+		 * (during truncate, child xtree pages must be freed before
+		 * parent's tlocks change the working map).
+		 * Take tlock off anonymous list and add to tail of
+		 * transaction list
+		 *
+		 * Note:  We really need to get rid of the tid & lid and
+		 * use list_head's.  This code is getting UGLY!
+		 */
+		if (jfs_ip->atlhead == lid) {
+			if (jfs_ip->atltail == lid) {
+				/* only anonymous txn.
+				 * Remove from anon_list
+				 */
+				list_del_init(&jfs_ip->anon_inode_list);
+			}
+			jfs_ip->atlhead = tlck->next;
+		} else {
+			lid_t last;
+			for (last = jfs_ip->atlhead;
+			     lid_to_tlock(last)->next != lid;
+			     last = lid_to_tlock(last)->next) {
+				assert(last);
+			}
+			lid_to_tlock(last)->next = tlck->next;
+			if (jfs_ip->atltail == lid)
+				jfs_ip->atltail = last;
+		}
+
+		/* insert the tlock at tail of transaction tlock list */
+
+		if (tblk->next)
+			lid_to_tlock(tblk->last)->next = lid;
+		else
+			tblk->next = lid;
+		tlck->next = 0;
+		tblk->last = lid;
+
+		goto grantLock;
+	}
+
+	goto waitLock;
+
+	/*
+	 * allocate a tlock
+	 */
+      allocateLock:
+	lid = txLockAlloc();
+	tlck = lid_to_tlock(lid);
+
+	/*
+	 * initialize tlock
+	 */
+	tlck->tid = tid;
+
+	/* mark tlock for meta-data page */
+	if (mp->xflag & COMMIT_PAGE) {
+
+		tlck->flag = tlckPAGELOCK;
+
+		/* mark the page dirty and nohomeok */
+		mark_metapage_dirty(mp);
+		atomic_inc(&mp->nohomeok);
+
+		jfs_info("locking mp = 0x%p, nohomeok = %d tid = %d tlck = 0x%p",
+			 mp, atomic_read(&mp->nohomeok), tid, tlck);
+
+		/* if anonymous transaction, and buffer is on the group
+		 * commit synclist, mark inode to show this.  This will
+		 * prevent the buffer from being marked nohomeok for too
+		 * long a time.
+		 */
+		if ((tid == 0) && mp->lsn)
+			set_cflag(COMMIT_Synclist, ip);
+	}
+	/* mark tlock for in-memory inode */
+	else
+		tlck->flag = tlckINODELOCK;
+
+	tlck->type = 0;
+
+	/* bind the tlock and the page */
+	tlck->ip = ip;
+	tlck->mp = mp;
+	if (dir_xtree)
+		jfs_ip->xtlid = lid;
+	else
+		mp->lid = lid;
+
+	/*
+	 * enqueue transaction lock to transaction/inode
+	 */
+	/* insert the tlock at tail of transaction tlock list */
+	if (tid) {
+		tblk = tid_to_tblock(tid);
+		if (tblk->next)
+			lid_to_tlock(tblk->last)->next = lid;
+		else
+			tblk->next = lid;
+		tlck->next = 0;
+		tblk->last = lid;
+	}
+	/* anonymous transaction:
+	 * insert the tlock at head of inode anonymous tlock list
+	 */
+	else {
+		tlck->next = jfs_ip->atlhead;
+		jfs_ip->atlhead = lid;
+		if (tlck->next == 0) {
+			/* This inode's first anonymous transaction */
+			jfs_ip->atltail = lid;
+			list_add_tail(&jfs_ip->anon_inode_list,
+				      &TxAnchor.anon_list);
+		}
+	}
+
+	/* initialize type dependent area for linelock */
+	linelock = (struct linelock *) & tlck->lock;
+	linelock->next = 0;
+	linelock->flag = tlckLINELOCK;
+	linelock->maxcnt = TLOCKSHORT;
+	linelock->index = 0;
+
+	switch (type & tlckTYPE) {
+	case tlckDTREE:
+		linelock->l2linesize = L2DTSLOTSIZE;
+		break;
+
+	case tlckXTREE:
+		linelock->l2linesize = L2XTSLOTSIZE;
+
+		xtlck = (struct xtlock *) linelock;
+		xtlck->header.offset = 0;
+		xtlck->header.length = 2;
+
+		if (type & tlckNEW) {
+			xtlck->lwm.offset = XTENTRYSTART;
+		} else {
+			if (mp->xflag & COMMIT_PAGE)
+				p = (xtpage_t *) mp->data;
+			else
+				p = &jfs_ip->i_xtroot;
+			xtlck->lwm.offset =
+			    le16_to_cpu(p->header.nextindex);
+		}
+		xtlck->lwm.length = 0;	/* ! */
+		xtlck->twm.offset = 0;
+		xtlck->hwm.offset = 0;
+
+		xtlck->index = 2;
+		break;
+
+	case tlckINODE:
+		linelock->l2linesize = L2INODESLOTSIZE;
+		break;
+
+	case tlckDATA:
+		linelock->l2linesize = L2DATASLOTSIZE;
+		break;
+
+	default:
+		jfs_err("UFO tlock:0x%p", tlck);
+	}
+
+	/*
+	 * update tlock vector
+	 */
+      grantLock:
+	tlck->type |= type;
+
+	TXN_UNLOCK();
+
+	return tlck;
+
+	/*
+	 * page is being locked by another transaction:
+	 */
+      waitLock:
+	/* Only locks on ipimap or ipaimap should reach here */
+	/* assert(jfs_ip->fileset == AGGREGATE_I); */
+	if (jfs_ip->fileset != AGGREGATE_I) {
+		jfs_err("txLock: trying to lock locked page!");
+		dump_mem("ip", ip, sizeof(struct inode));
+		dump_mem("mp", mp, sizeof(struct metapage));
+		dump_mem("Locker's tblk", tid_to_tblock(tid),
+			 sizeof(struct tblock));
+		dump_mem("Tlock", tlck, sizeof(struct tlock));
+		BUG();
+	}
+	INCREMENT(stattx.waitlock);	/* statistics */
+	release_metapage(mp);
+
+	jfs_info("txLock: in waitLock, tid = %d, xtid = %d, lid = %d",
+		 tid, xtid, lid);
+	TXN_SLEEP_DROP_LOCK(&tid_to_tblock(xtid)->waitor);
+	jfs_info("txLock: awakened     tid = %d, lid = %d", tid, lid);
+
+	return NULL;
+}
+
+
+/*
+ * NAME:        txRelease()
+ *
+ * FUNCTION:    Release buffers associated with transaction locks, but don't
+ *		mark homeok yet.  The allows other transactions to modify
+ *		buffers, but won't let them go to disk until commit record
+ *		actually gets written.
+ *
+ * PARAMETER:
+ *              tblk    -
+ *
+ * RETURN:      Errors from subroutines.
+ */
+static void txRelease(struct tblock * tblk)
+{
+	struct metapage *mp;
+	lid_t lid;
+	struct tlock *tlck;
+
+	TXN_LOCK();
+
+	for (lid = tblk->next; lid; lid = tlck->next) {
+		tlck = lid_to_tlock(lid);
+		if ((mp = tlck->mp) != NULL &&
+		    (tlck->type & tlckBTROOT) == 0) {
+			assert(mp->xflag & COMMIT_PAGE);
+			mp->lid = 0;
+		}
+	}
+
+	/*
+	 * wakeup transactions waiting on a page locked
+	 * by the current transaction
+	 */
+	TXN_WAKEUP(&tblk->waitor);
+
+	TXN_UNLOCK();
+}
+
+
+/*
+ * NAME:        txUnlock()
+ *
+ * FUNCTION:    Initiates pageout of pages modified by tid in journalled
+ *              objects and frees their lockwords.
+ */
+static void txUnlock(struct tblock * tblk)
+{
+	struct tlock *tlck;
+	struct linelock *linelock;
+	lid_t lid, next, llid, k;
+	struct metapage *mp;
+	struct jfs_log *log;
+	int difft, diffp;
+
+	jfs_info("txUnlock: tblk = 0x%p", tblk);
+	log = JFS_SBI(tblk->sb)->log;
+
+	/*
+	 * mark page under tlock homeok (its log has been written):
+	 */
+	for (lid = tblk->next; lid; lid = next) {
+		tlck = lid_to_tlock(lid);
+		next = tlck->next;
+
+		jfs_info("unlocking lid = %d, tlck = 0x%p", lid, tlck);
+
+		/* unbind page from tlock */
+		if ((mp = tlck->mp) != NULL &&
+		    (tlck->type & tlckBTROOT) == 0) {
+			assert(mp->xflag & COMMIT_PAGE);
+
+			/* hold buffer
+			 *
+			 * It's possible that someone else has the metapage.
+			 * The only things were changing are nohomeok, which
+			 * is handled atomically, and clsn which is protected
+			 * by the LOGSYNC_LOCK.
+			 */
+			hold_metapage(mp, 1);
+
+			assert(atomic_read(&mp->nohomeok) > 0);
+			atomic_dec(&mp->nohomeok);
+
+			/* inherit younger/larger clsn */
+			LOGSYNC_LOCK(log);
+			if (mp->clsn) {
+				logdiff(difft, tblk->clsn, log);
+				logdiff(diffp, mp->clsn, log);
+				if (difft > diffp)
+					mp->clsn = tblk->clsn;
+			} else
+				mp->clsn = tblk->clsn;
+			LOGSYNC_UNLOCK(log);
+
+			assert(!(tlck->flag & tlckFREEPAGE));
+
+			if (tlck->flag & tlckWRITEPAGE) {
+				write_metapage(mp);
+			} else {
+				/* release page which has been forced */
+				release_metapage(mp);
+			}
+		}
+
+		/* insert tlock, and linelock(s) of the tlock if any,
+		 * at head of freelist
+		 */
+		TXN_LOCK();
+
+		llid = ((struct linelock *) & tlck->lock)->next;
+		while (llid) {
+			linelock = (struct linelock *) lid_to_tlock(llid);
+			k = linelock->next;
+			txLockFree(llid);
+			llid = k;
+		}
+		txLockFree(lid);
+
+		TXN_UNLOCK();
+	}
+	tblk->next = tblk->last = 0;
+
+	/*
+	 * remove tblock from logsynclist
+	 * (allocation map pages inherited lsn of tblk and
+	 * has been inserted in logsync list at txUpdateMap())
+	 */
+	if (tblk->lsn) {
+		LOGSYNC_LOCK(log);
+		log->count--;
+		list_del(&tblk->synclist);
+		LOGSYNC_UNLOCK(log);
+	}
+}
+
+
+/*
+ *      txMaplock()
+ *
+ * function: allocate a transaction lock for freed page/entry;
+ *      for freed page, maplock is used as xtlock/dtlock type;
+ */
+struct tlock *txMaplock(tid_t tid, struct inode *ip, int type)
+{
+	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
+	lid_t lid;
+	struct tblock *tblk;
+	struct tlock *tlck;
+	struct maplock *maplock;
+
+	TXN_LOCK();
+
+	/*
+	 * allocate a tlock
+	 */
+	lid = txLockAlloc();
+	tlck = lid_to_tlock(lid);
+
+	/*
+	 * initialize tlock
+	 */
+	tlck->tid = tid;
+
+	/* bind the tlock and the object */
+	tlck->flag = tlckINODELOCK;
+	tlck->ip = ip;
+	tlck->mp = NULL;
+
+	tlck->type = type;
+
+	/*
+	 * enqueue transaction lock to transaction/inode
+	 */
+	/* insert the tlock at tail of transaction tlock list */
+	if (tid) {
+		tblk = tid_to_tblock(tid);
+		if (tblk->next)
+			lid_to_tlock(tblk->last)->next = lid;
+		else
+			tblk->next = lid;
+		tlck->next = 0;
+		tblk->last = lid;
+	}
+	/* anonymous transaction:
+	 * insert the tlock at head of inode anonymous tlock list
+	 */
+	else {
+		tlck->next = jfs_ip->atlhead;
+		jfs_ip->atlhead = lid;
+		if (tlck->next == 0) {
+			/* This inode's first anonymous transaction */
+			jfs_ip->atltail = lid;
+			list_add_tail(&jfs_ip->anon_inode_list,
+				      &TxAnchor.anon_list);
+		}
+	}
+
+	TXN_UNLOCK();
+
+	/* initialize type dependent area for maplock */
+	maplock = (struct maplock *) & tlck->lock;
+	maplock->next = 0;
+	maplock->maxcnt = 0;
+	maplock->index = 0;
+
+	return tlck;
+}
+
+
+/*
+ *      txLinelock()
+ *
+ * function: allocate a transaction lock for log vector list
+ */
+struct linelock *txLinelock(struct linelock * tlock)
+{
+	lid_t lid;
+	struct tlock *tlck;
+	struct linelock *linelock;
+
+	TXN_LOCK();
+
+	/* allocate a TxLock structure */
+	lid = txLockAlloc();
+	tlck = lid_to_tlock(lid);
+
+	TXN_UNLOCK();
+
+	/* initialize linelock */
+	linelock = (struct linelock *) tlck;
+	linelock->next = 0;
+	linelock->flag = tlckLINELOCK;
+	linelock->maxcnt = TLOCKLONG;
+	linelock->index = 0;
+
+	/* append linelock after tlock */
+	linelock->next = tlock->next;
+	tlock->next = lid;
+
+	return linelock;
+}
+
+
+
+/*
+ *              transaction commit management
+ *              -----------------------------
+ */
+
+/*
+ * NAME:        txCommit()
+ *
+ * FUNCTION:    commit the changes to the objects specified in
+ *              clist.  For journalled segments only the
+ *              changes of the caller are committed, ie by tid.
+ *              for non-journalled segments the data are flushed to
+ *              disk and then the change to the disk inode and indirect
+ *              blocks committed (so blocks newly allocated to the
+ *              segment will be made a part of the segment atomically).
+ *
+ *              all of the segments specified in clist must be in
+ *              one file system. no more than 6 segments are needed
+ *              to handle all unix svcs.
+ *
+ *              if the i_nlink field (i.e. disk inode link count)
+ *              is zero, and the type of inode is a regular file or
+ *              directory, or symbolic link , the inode is truncated
+ *              to zero length. the truncation is committed but the
+ *              VM resources are unaffected until it is closed (see
+ *              iput and iclose).
+ *
+ * PARAMETER:
+ *
+ * RETURN:
+ *
+ * serialization:
+ *              on entry the inode lock on each segment is assumed
+ *              to be held.
+ *
+ * i/o error:
+ */
+int txCommit(tid_t tid,		/* transaction identifier */
+	     int nip,		/* number of inodes to commit */
+	     struct inode **iplist,	/* list of inode to commit */
+	     int flag)
+{
+	int rc = 0;
+	struct commit cd;
+	struct jfs_log *log;
+	struct tblock *tblk;
+	struct lrd *lrd;
+	int lsn;
+	struct inode *ip;
+	struct jfs_inode_info *jfs_ip;
+	int k, n;
+	ino_t top;
+	struct super_block *sb;
+
+	jfs_info("txCommit, tid = %d, flag = %d", tid, flag);
+	/* is read-only file system ? */
+	if (isReadOnly(iplist[0])) {
+		rc = -EROFS;
+		goto TheEnd;
+	}
+
+	sb = cd.sb = iplist[0]->i_sb;
+	cd.tid = tid;
+
+	if (tid == 0)
+		tid = txBegin(sb, 0);
+	tblk = tid_to_tblock(tid);
+
+	/*
+	 * initialize commit structure
+	 */
+	log = JFS_SBI(sb)->log;
+	cd.log = log;
+
+	/* initialize log record descriptor in commit */
+	lrd = &cd.lrd;
+	lrd->logtid = cpu_to_le32(tblk->logtid);
+	lrd->backchain = 0;
+
+	tblk->xflag |= flag;
+
+	if ((flag & (COMMIT_FORCE | COMMIT_SYNC)) == 0)
+		tblk->xflag |= COMMIT_LAZY;
+	/*
+	 *      prepare non-journaled objects for commit
+	 *
+	 * flush data pages of non-journaled file
+	 * to prevent the file getting non-initialized disk blocks
+	 * in case of crash.
+	 * (new blocks - )
+	 */
+	cd.iplist = iplist;
+	cd.nip = nip;
+
+	/*
+	 *      acquire transaction lock on (on-disk) inodes
+	 *
+	 * update on-disk inode from in-memory inode
+	 * acquiring transaction locks for AFTER records
+	 * on the on-disk inode of file object
+	 *
+	 * sort the inodes array by inode number in descending order
+	 * to prevent deadlock when acquiring transaction lock
+	 * of on-disk inodes on multiple on-disk inode pages by
+	 * multiple concurrent transactions
+	 */
+	for (k = 0; k < cd.nip; k++) {
+		top = (cd.iplist[k])->i_ino;
+		for (n = k + 1; n < cd.nip; n++) {
+			ip = cd.iplist[n];
+			if (ip->i_ino > top) {
+				top = ip->i_ino;
+				cd.iplist[n] = cd.iplist[k];
+				cd.iplist[k] = ip;
+			}
+		}
+
+		ip = cd.iplist[k];
+		jfs_ip = JFS_IP(ip);
+
+		/*
+		 * BUGBUG - This code has temporarily been removed.  The
+		 * intent is to ensure that any file data is written before
+		 * the metadata is committed to the journal.  This prevents
+		 * uninitialized data from appearing in a file after the
+		 * journal has been replayed.  (The uninitialized data
+		 * could be sensitive data removed by another user.)
+		 *
+		 * The problem now is that we are