Merge

fs/xfs/Makefile

@@ -132,7 +132,6 @@ xfs-y				+= $(addprefix linux/, \
 # Objects in support/
 xfs-y				+= $(addprefix support/, \
 				   debug.o \
-				   kmem.o \
 				   ktrace.o \
 				   move.o \
 				   mrlock.o \

fs/xfs/linux/xfs_globals.c

@@ -48,7 +48,7 @@ unsigned long xfs_physmem;
  * Tunable XFS parameters.  xfs_params is required even when CONFIG_SYSCTL=n,
  * other XFS code uses these values.
  */
-xfs_param_t xfs_params = { 0, 1, 0, 0, 0, 3 };
+xfs_param_t xfs_params = { 0, 1, 0, 0, 0, 3, 30 * HZ };
 
 /*
  * Global system credential structure.

fs/xfs/linux/xfs_iops.c

@@ -152,8 +152,6 @@ linvfs_mknod(
 			ip->i_rdev = to_kdev_t(rdev);
 		validate_fields(dir);
 		d_instantiate(dentry, ip);
-		mark_inode_dirty_sync(ip);
-		mark_inode_dirty_sync(dir);
 	}
 
 	if (!error && have_default_acl) {
@@ -240,7 +238,6 @@ linvfs_link(
 		VN_HOLD(vp);
 		validate_fields(ip);
 		d_instantiate(dentry, ip);
-		mark_inode_dirty_sync(ip);
 	}
 	return -error;
 }
@@ -261,8 +258,6 @@ linvfs_unlink(
 	if (!error) {
 		validate_fields(dir);	/* For size only */
 		validate_fields(inode);
-		mark_inode_dirty_sync(inode);
-		mark_inode_dirty_sync(dir);
 	}
 
 	return -error;
@@ -296,8 +291,6 @@ linvfs_symlink(
 		d_instantiate(dentry, ip);
 		validate_fields(dir);
 		validate_fields(ip); /* size needs update */
-		mark_inode_dirty_sync(ip);
-		mark_inode_dirty_sync(dir);
 	}
 	return -error;
 }
@@ -315,8 +308,6 @@ linvfs_rmdir(
 	if (!error) {
 		validate_fields(inode);
 		validate_fields(dir);
-		mark_inode_dirty_sync(inode);
-		mark_inode_dirty_sync(dir);
 	}
 	return -error;
 }
@@ -346,7 +337,6 @@ linvfs_rename(
 	validate_fields(odir);
 	if (ndir != odir)
 		validate_fields(ndir);
-	mark_inode_dirty(ndir);
 	return 0;
 }
 
@@ -520,7 +510,6 @@ linvfs_setattr(
 
 	if (!error) {
 		vn_revalidate(vp);
-		mark_inode_dirty_sync(inode);
 	}
 	return error;
 }

fs/xfs/linux/xfs_lrw.c

@@ -887,29 +887,23 @@ xfsbdstrat(
 	return (xfs_bioerror_relse(bp));
 }
 
-
-void
-XFS_bflush(xfs_buftarg_t *target)
-{
-	pagebuf_delwri_flush(target, PBDF_WAIT, NULL);
-}
-
 /*
- * If the underlying (log or data) device is readonly, there are some
+ * If the underlying (data/log/rt) device is readonly, there are some
  * operations that cannot proceed.
  */
 int
-xfs_dev_is_read_only(xfs_mount_t *mp, char *message)
+xfs_dev_is_read_only(
+	xfs_mount_t		*mp,
+	char			*message)
 {
-	if (bdev_read_only(mp->m_ddev_targp->pbr_bdev) ||
-	    bdev_read_only(mp->m_logdev_targp->pbr_bdev) ||
-	   (mp->m_rtdev_targp && bdev_read_only(mp->m_rtdev_targp->pbr_bdev))) {
+	if (xfs_readonly_buftarg(mp->m_ddev_targp) ||
+	    xfs_readonly_buftarg(mp->m_logdev_targp) ||
+	    (mp->m_rtdev_targp && xfs_readonly_buftarg(mp->m_rtdev_targp))) {
 		cmn_err(CE_NOTE,
 			"XFS: %s required on read-only device.", message);
 		cmn_err(CE_NOTE,
 			"XFS: write access unavailable, cannot proceed.");
 		return EROFS;
 	}
-
 	return 0;
 }

fs/xfs/linux/xfs_super.c

@@ -217,13 +217,27 @@ xfs_blkdev_put(
 }
 
 void
-xfs_free_buftarg(
+xfs_flush_buftarg(
 	xfs_buftarg_t		*btp)
 {
 	pagebuf_delwri_flush(btp, PBDF_WAIT, NULL);
+}
+
+void
+xfs_free_buftarg(
+	xfs_buftarg_t		*btp)
+{
+	xfs_flush_buftarg(btp);
 	kmem_free(btp, sizeof(*btp));
 }
 
+int
+xfs_readonly_buftarg(
+	xfs_buftarg_t		*btp)
+{
+	return bdev_read_only(btp->pbr_bdev);
+}
+
 void
 xfs_relse_buftarg(
 	xfs_buftarg_t		*btp)
@@ -331,9 +345,10 @@ destroy_inodecache( void )
 }
 
 /*
- * We do not actually write the inode here, just mark the
- * super block dirty so that sync_supers calls us and
- * forces the flush.
+ * Attempt to flush the inode, this will actually fail
+ * if the inode is pinned, but we dirty the inode again
+ * at the point when it is unpinned after a log write,
+ * since this is when the inode itself becomes flushable. 
  */
 STATIC void
 linvfs_write_inode(
@@ -348,8 +363,6 @@ linvfs_write_inode(
 		if (sync)
 			flags |= FLUSH_SYNC;
 		VOP_IFLUSH(vp, flags, error);
-		if (error == EAGAIN)
-			inode->i_sb->s_dirt = 1;
 	}
 }
 
@@ -369,6 +382,61 @@ linvfs_clear_inode(
 	}
 }
 
+
+#define SYNCD_FLAGS	(SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR)
+
+STATIC int
+syncd(void *arg)
+{
+	vfs_t			*vfsp = (vfs_t *) arg;
+	int			error;
+
+	daemonize("xfs_syncd");
+
+	vfsp->vfs_sync_task = current;
+	wmb();
+	wake_up(&vfsp->vfs_wait_sync_task);
+
+	for (;;) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(xfs_params.sync_interval);
+		if (vfsp->vfs_flag & VFS_UMOUNT)
+			break;
+		if (vfsp->vfs_flag & VFS_RDONLY)
+			continue;
+		VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error);
+	}
+
+	vfsp->vfs_sync_task = NULL;
+	wmb();
+	wake_up(&vfsp->vfs_wait_sync_task);
+
+	return 0;
+}
+
+STATIC int
+linvfs_start_syncd(vfs_t *vfsp)
+{
+	int pid;
+
+	pid = kernel_thread(syncd, (void *) vfsp,
+			CLONE_VM | CLONE_FS | CLONE_FILES);
+	if (pid < 0)
+		return pid;
+	wait_event(vfsp->vfs_wait_sync_task, vfsp->vfs_sync_task);
+	return 0;
+}
+
+STATIC void
+linvfs_stop_syncd(vfs_t *vfsp)
+{
+	vfsp->vfs_flag |= VFS_UMOUNT;
+	wmb();
+
+	wake_up_process(vfsp->vfs_sync_task);
+	wait_event(vfsp->vfs_wait_sync_task, !vfsp->vfs_sync_task);
+}
+
 STATIC void
 linvfs_put_super(
 	struct super_block	*sb)
@@ -376,8 +444,9 @@ linvfs_put_super(
 	vfs_t			*vfsp = LINVFS_GET_VFS(sb);
 	int			error;
 
+	linvfs_stop_syncd(vfsp);
 	VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error);
-	if (error == 0)
+	if (!error)
 		VFS_UNMOUNT(vfsp, 0, NULL, error);
 	if (error) {
 		printk("XFS unmount got error %d\n", error);
@@ -395,10 +464,13 @@ linvfs_write_super(
 	vfs_t			*vfsp = LINVFS_GET_VFS(sb);
 	int			error;
 
-	sb->s_dirt = 0;
-	if (sb->s_flags & MS_RDONLY)
+	if (sb->s_flags & MS_RDONLY) {
+		sb->s_dirt = 0; /* paranoia */
 		return;
-	VFS_SYNC(vfsp, SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR, NULL, error);
+	}
+	/* Push the log and superblock a little */
+	VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error);
+	sb->s_dirt = 0;
 }
 
 STATIC int
@@ -424,12 +496,8 @@ linvfs_remount(
 	int			error;
 
 	VFS_PARSEARGS(vfsp, options, args, 1, error);
-	if (error)
-		goto out;
-
+	if (!error)
 		VFS_MNTUPDATE(vfsp, flags, args, error);
-
-out:
 	kmem_free(args, sizeof(*args));
 	return error;
 }
@@ -438,11 +506,10 @@ STATIC void
 linvfs_freeze_fs(
 	struct super_block	*sb)
 {
-	vfs_t			*vfsp;
+	vfs_t			*vfsp = LINVFS_GET_VFS(sb);
 	vnode_t			*vp;
 	int			error;
 
-	vfsp = LINVFS_GET_VFS(sb);
 	if (sb->s_flags & MS_RDONLY)
 		return;
 	VFS_ROOT(vfsp, &vp, error);
@@ -454,11 +521,10 @@ STATIC void
 linvfs_unfreeze_fs(
 	struct super_block	*sb)
 {
-	vfs_t			*vfsp;
+	vfs_t			*vfsp = LINVFS_GET_VFS(sb);
 	vnode_t			*vp;
 	int			error;
 
-	vfsp = LINVFS_GET_VFS(sb);
 	VFS_ROOT(vfsp, &vp, error);
 	VOP_IOCTL(vp, LINVFS_GET_IP(vp), NULL, XFS_IOC_THAW, 0, error);
 	VN_RELE(vp);
@@ -652,7 +718,8 @@ linvfs_fill_super(
 		goto fail_vnrele;
 	if (is_bad_inode(sb->s_root->d_inode))
 		goto fail_vnrele;
-
+	if (linvfs_start_syncd(vfsp))
+		goto fail_vnrele;
 	vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address);
 
 	kmem_free(args, sizeof(*args));

fs/xfs/linux/xfs_super.h

@@ -101,7 +101,8 @@ extern void xfs_blkdev_put(struct block_device *);
 extern struct pb_target *xfs_alloc_buftarg(struct block_device *);
 extern void xfs_relse_buftarg(struct pb_target *);
 extern void xfs_free_buftarg(struct pb_target *);
-
+extern void xfs_flush_buftarg(struct pb_target *);
+extern int xfs_readonly_buftarg(struct pb_target *);
 extern void xfs_setsize_buftarg(struct pb_target *, unsigned int, unsigned int);
 extern unsigned int xfs_getsize_buftarg(struct pb_target *);
 

fs/xfs/linux/xfs_syncd.c

+/*
+ * 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>
+
+#define SYNCD_FLAGS	(SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR)
+
+int syncd(void *arg)
+{
+	vfs_t			*vfsp = (vfs_t *) arg;
+	int			error;
+
+	daemonize("xfs_syncd");
+
+	vfsp->vfs_sync_task = current;
+	wmb();
+	wake_up(&vfsp->vfs_wait_sync_task);
+
+	for (;;) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(xfs_params.sync_interval);
+		if (vfsp->vfs_flag & VFS_UMOUNT)
+			break;
+		if (vfsp->vfs_flag & VFS_RDONLY);
+			continue;
+		VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error);
+	}
+
+	vfsp->vfs_sync_task = NULL;
+	wmb();
+	wake_up(&vfsp->vfs_wait_sync_task);
+
+	return 0;
+}
+
+int
+linvfs_start_syncd(vfs_t *vfsp)
+{
+	int pid;
+
+	pid = kernel_thread(syncd, (void *) vfsp,
+			CLONE_VM | CLONE_FS | CLONE_FILES);
+	if (pid < 0)
+		return pid;
+	wait_event(vfsp->vfs_wait_sync_task, vfsp->vfs_sync_task);
+	return 0;
+}
+
+void
+linvfs_stop_syncd(vfs_t *vfsp)
+{
+	vfsp->vfs_flag |= VFS_UMOUNT;
+	wmb();
+
+	wake_up_process(vfsp->vfs_sync_task);
+	wait_event(vfsp->vfs_wait_sync_task, !vfsp->vfs_sync_task);
+}

fs/xfs/linux/xfs_sysctl.c

@@ -36,8 +36,8 @@
 #include <linux/proc_fs.h>
 
 
-STATIC ulong xfs_min[XFS_PARAM] = { 0, 0, 0, 0, 0,   0 };
-STATIC ulong xfs_max[XFS_PARAM] = { 1, 1, 1, 1, 127, 3 };
+STATIC ulong xfs_min[XFS_PARAM] = { 0, 0, 0, 0, 0,   0, HZ };
+STATIC ulong xfs_max[XFS_PARAM] = { 1, 1, 1, 1, 127, 3, HZ * 60 };
 
 static struct ctl_table_header *xfs_table_header;
 
@@ -92,6 +92,10 @@ STATIC ctl_table xfs_table[] = {
 	sizeof(ulong), 0644, NULL, &proc_doulongvec_minmax,
 	&sysctl_intvec, NULL, &xfs_min[5], &xfs_max[5]},
 
+	{XFS_SYNC_INTERVAL, "sync_interval", &xfs_params.sync_interval,
+	sizeof(ulong), 0644, NULL, &proc_doulongvec_minmax,
+	&sysctl_intvec, NULL, &xfs_min[6], &xfs_max[6]},
+
 	{0}
 };
 

fs/xfs/linux/xfs_sysctl.h

@@ -49,6 +49,7 @@ typedef struct xfs_param {
 	ulong	symlink_mode;	/* Symlink creat mode affected by umask. */
 	ulong	panic_mask;	/* bitmask to specify panics on errors.  */
 	ulong	error_level;	/* Degree of reporting for internal probs*/
+	ulong	sync_interval;	/* time between sync calls		 */
 } xfs_param_t;
 
 /*
@@ -73,6 +74,7 @@ enum {
 	XFS_SYMLINK_MODE = 4,
 	XFS_PANIC_MASK = 5,
 	XFS_ERRLEVEL = 6,
+	XFS_SYNC_INTERVAL = 7,
 };
 
 extern xfs_param_t	xfs_params;

fs/xfs/linux/xfs_vfs.c

@@ -238,6 +238,7 @@ vfs_allocate( void )
 
 	vfsp = kmem_zalloc(sizeof(vfs_t), KM_SLEEP);
 	bhv_head_init(VFS_BHVHEAD(vfsp), "vfs");
+	init_waitqueue_head(&vfsp->vfs_wait_sync_task);
 	return vfsp;
 }
 

fs/xfs/linux/xfs_vfs.h

@@ -48,6 +48,8 @@ typedef struct vfs {
 	fsid_t			*vfs_altfsid;	/* An ID fixed for life of FS */
 	bhv_head_t		vfs_bh;		/* head of vfs behavior chain */
 	struct super_block	*vfs_super;	/* Linux superblock structure */
+	struct task_struct	*vfs_sync_task;
+	wait_queue_head_t	vfs_wait_sync_task;
 } vfs_t;
 
 #define vfs_fbhv		vfs_bh.bh_first	/* 1st on vfs behavior chain */
@@ -78,7 +80,8 @@ typedef enum {
 #define VFS_RDONLY		0x0001	/* read-only vfs */
 #define VFS_GRPID		0x0002	/* group-ID assigned from directory */
 #define VFS_DMI			0x0004	/* filesystem has the DMI enabled */
-#define VFS_END			0x0004	/* max flag */
+#define VFS_UMOUNT		0x0008	/* unmount in progress */
+#define VFS_END			0x0008	/* max flag */
 
 #define SYNC_ATTR		0x0001	/* sync attributes */
 #define SYNC_CLOSE		0x0002	/* close file system down */
@@ -87,6 +90,7 @@ typedef enum {
 #define SYNC_FSDATA		0x0020	/* flush fs data (e.g. superblocks) */
 #define SYNC_BDFLUSH		0x0010	/* BDFLUSH is calling -- don't block */
 
+
 typedef int	(*vfs_mount_t)(bhv_desc_t *,
 				struct xfs_mount_args *, struct cred *);
 typedef int	(*vfs_parseargs_t)(bhv_desc_t *, char *,

fs/xfs/linux/xfs_vnode.h

@@ -562,8 +562,7 @@ static __inline__ void vn_flagclr(struct vnode *vp, uint flag)
 	(!list_empty(&(LINVFS_GET_IP(vp)->i_mapping->i_mmap_shared))))
 #define VN_CACHED(vp)	(LINVFS_GET_IP(vp)->i_mapping->nrpages)
 #define VN_DIRTY(vp)	(!list_empty(&(LINVFS_GET_IP(vp)->i_mapping->dirty_pages)))
-#define VMODIFY(vp)	{ VN_FLAGSET(vp, VMODIFIED); \
-			mark_inode_dirty(LINVFS_GET_IP(vp)); }
+#define VMODIFY(vp)	VN_FLAGSET(vp, VMODIFIED)
 #define VUNMODIFY(vp)	VN_FLAGCLR(vp, VMODIFIED)
 
 /*

fs/xfs/pagebuf/page_buf.c

@@ -68,7 +68,7 @@
 #define BN_ALIGN_MASK	((1 << (PAGE_CACHE_SHIFT - BBSHIFT)) - 1)
 
 #ifndef GFP_READAHEAD
-#define GFP_READAHEAD	__GFP_NOWARN
+#define GFP_READAHEAD	(__GFP_NOWARN|__GFP_NORETRY)
 #endif
 
 /*
@@ -181,7 +181,7 @@ _bhash(
 	 * dev_t is 16 bits, loff_t is always 64 bits
 	 */
 	base ^= dev;
-	for (bit = hval = 0; base != 0 && bit < sizeof(base) * 8; bit += NBITS) {
+	for (bit = hval = 0; base && bit < sizeof(base) * 8; bit += NBITS) {
 		hval ^= (int)base & (NHASH-1);
 		base >>= NBITS;
 	}
@@ -189,18 +189,18 @@ _bhash(
 }
 
 /*
- * Mapping of multi-page buffers into contingous virtual space
+ * Mapping of multi-page buffers into contiguous virtual space
  */
 
 STATIC void *pagebuf_mapout_locked(page_buf_t *);
 
-STATIC  spinlock_t              as_lock = SPIN_LOCK_UNLOCKED;
 typedef struct a_list {
 	void		*vm_addr;
 	struct a_list	*next;
 } a_list_t;
 STATIC a_list_t		*as_free_head;
 STATIC int		as_list_len;
+STATIC spinlock_t	as_lock = SPIN_LOCK_UNLOCKED;
 
 
 /*
@@ -1897,13 +1897,6 @@ pagebuf_readstats(
 }
 #endif  /* CONFIG_PROC_FS */
 
-STATIC void
-pagebuf_shaker(void)
-{
-	pagebuf_daemon_wakeup(1);
-}
-
-
 /*
  *	Initialization and Termination
  */
@@ -1943,7 +1936,6 @@ pagebuf_init(void)
 #endif
 
 	pagebuf_daemon_start();
-	kmem_shake_register(pagebuf_shaker);
 	return 0;
 }
 
@@ -1959,7 +1951,6 @@ pagebuf_terminate(void)
 	pagebuf_daemon_stop();
 
 	kmem_cache_destroy(pagebuf_cache);
-	kmem_shake_deregister(pagebuf_shaker);
 
 	unregister_sysctl_table(pagebuf_table_header);
 #ifdef  CONFIG_PROC_FS

fs/xfs/quota/xfs_qm.c

@@ -88,7 +88,7 @@ STATIC void	xfs_qm_list_destroy(xfs_dqlist_t *);
 STATIC int	xfs_qm_quotacheck(xfs_mount_t *);
 
 STATIC int	xfs_qm_init_quotainos(xfs_mount_t *);
-STATIC void	xfs_qm_shake(void);
+STATIC int	xfs_qm_shake(int, unsigned int);
 
 #ifdef DEBUG
 extern mutex_t	qcheck_lock;
@@ -112,6 +112,8 @@ extern mutex_t	qcheck_lock;
 #define XQM_LIST_PRINT(l, NXT, title) do { } while (0)
 #endif
 
+struct shrinker *xfs_qm_shrinker;
+
 /*
  * Initialize the XQM structure.
  * Note that there is not one quota manager per file system.
@@ -161,7 +163,7 @@ xfs_Gqm_init(void)
 	} else
 		xqm->qm_dqzone = qm_dqzone;
 
-	kmem_shake_register(xfs_qm_shake);
+	xfs_qm_shrinker = set_shrinker(DEFAULT_SEEKS, xfs_qm_shake);
 
 	/*
 	 * The t_dqinfo portion of transactions.
@@ -193,7 +195,8 @@ xfs_qm_destroy(
 
 	ASSERT(xqm != NULL);
 	ASSERT(xqm->qm_nrefs == 0);
-	kmem_shake_deregister(xfs_qm_shake);
+
+	remove_shrinker(xfs_qm_shrinker);
 	hsize = xqm->qm_dqhashmask + 1;
 	for (i = 0; i < hsize; i++) {
 		xfs_qm_list_destroy(&(xqm->qm_usr_dqhtable[i]));
@@ -2088,7 +2091,7 @@ xfs_qm_shake_freelist(
 			xfs_dqunlock(dqp);
 			xfs_qm_freelist_unlock(xfs_Gqm);
 			if (++restarts >= XFS_QM_RECLAIM_MAX_RESTARTS)
-				return (nreclaimed != howmany);
+				goto out;
 			XQM_STATS_INC(xqmstats.xs_qm_dqwants);
 			goto tryagain;
 		}
@@ -2163,7 +2166,7 @@ xfs_qm_shake_freelist(
 			XFS_DQ_HASH_UNLOCK(hash);
 			xfs_qm_freelist_unlock(xfs_Gqm);
 			if (++restarts >= XFS_QM_RECLAIM_MAX_RESTARTS)
-				return (nreclaimed != howmany);
+				goto out;
 			goto tryagain;
 		}
 		xfs_dqtrace_entry(dqp, "DQSHAKE: UNLINKING");
@@ -2188,7 +2191,8 @@ xfs_qm_shake_freelist(
 		dqp = nextdqp;
 	}
 	xfs_qm_freelist_unlock(xfs_Gqm);
-	return (nreclaimed != howmany);
+ out:
+	return nreclaimed;
 }
 
 
@@ -2197,13 +2201,15 @@ xfs_qm_shake_freelist(
  * running low.
  */
 /* ARGSUSED */
-STATIC void
-xfs_qm_shake(void)
+STATIC int
+xfs_qm_shake(int nr_to_scan, unsigned int gfp_mask)
 {
 	int	ndqused, nfree, n;
 
+	if (!(gfp_mask & __GFP_WAIT))
+		return 0;
 	if (!xfs_Gqm)
-		return;
+		return 0;
 
 	nfree = xfs_Gqm->qm_dqfreelist.qh_nelems; /* free dquots */
 	/* incore dquots in all f/s's */
@@ -2212,12 +2218,12 @@ xfs_qm_shake(void)
 	ASSERT(ndqused >= 0);
 
 	if (nfree <= ndqused && nfree < ndquot)
-		return;
+		return 0;
 
 	ndqused *= xfs_Gqm->qm_dqfree_ratio;	/* target # of free dquots */
 	n = nfree - ndqused - ndquot;		/* # over target */
 
-	(void) xfs_qm_shake_freelist(MAX(nfree, n));
+	return xfs_qm_shake_freelist(MAX(nfree, n));
 }
 
 

fs/xfs/support/kmem.h

@@ -32,31 +32,118 @@
 #ifndef __XFS_SUPPORT_KMEM_H__
 #define __XFS_SUPPORT_KMEM_H__
 
+#include <linux/mm.h>
+#include <linux/highmem.h>
 #include <linux/slab.h>
+#include <linux/vmalloc.h>
 
 /*
- * memory management routines
+ * Cutoff point to use vmalloc instead of kmalloc.
  */
+#define MAX_SLAB_SIZE	0x10000
+
+/*
+ * XFS uses slightly different names for these due to the
+ * IRIX heritage.
+ */
+#define	kmem_zone	kmem_cache_s
+#define kmem_zone_t	kmem_cache_t
+
 #define KM_SLEEP	0x0001
 #define KM_NOSLEEP	0x0002
 #define KM_NOFS		0x0004
 
-#define	kmem_zone	kmem_cache_s
-#define kmem_zone_t	kmem_cache_t
 
-extern kmem_zone_t  *kmem_zone_init(int, char *);
-extern void	    *kmem_zone_zalloc(kmem_zone_t *, int);
-extern void	    *kmem_zone_alloc(kmem_zone_t *, int);
-extern void         kmem_zone_free(kmem_zone_t *, void *);
+/*
+ * XXX get rid of the unconditional  __GFP_NOFAIL by adding
+ * a KM_FAIL flag and using it where we're allowed to fail.
+ */
+static __inline unsigned int
+flag_convert(int flags)
+{
+#if DEBUG
+	if (unlikely(flags & ~(KM_SLEEP|KM_NOSLEEP|KM_NOFS))) {
+		printk(KERN_WARNING
+		    "XFS: memory allocation with wrong flags (%x)\n", flags);
+		BUG();
+	}
+#endif
+
+	if (flags & KM_NOSLEEP)
+		return GFP_ATOMIC;
+	/* If we're in a transaction, FS activity is not ok */
+	else if ((current->flags & PF_FSTRANS) || (flags & KM_NOFS))
+		return GFP_NOFS | __GFP_NOFAIL;
+	return GFP_KERNEL | __GFP_NOFAIL;
+}
+
+static __inline void *
+kmem_alloc(size_t size, int flags)
+{
+	if (unlikely(MAX_SLAB_SIZE < size))
+		/* Avoid doing filesystem sensitive stuff to get this */
+		return __vmalloc(size, flag_convert(flags), PAGE_KERNEL);
+	return kmalloc(size, flag_convert(flags));
+}
+
+static __inline void *
+kmem_zalloc(size_t size, int flags)
+{
+	void *ptr = kmem_alloc(size, flags);
+	if (likely(ptr != NULL))
+		memset(ptr, 0, size);
+	return ptr;
+}
+
+static __inline void
+kmem_free(void *ptr, size_t size)
+{
+	if (unlikely((unsigned long)ptr < VMALLOC_START ||
+		     (unsigned long)ptr >= VMALLOC_END))
+		kfree(ptr);
+	else
+		vfree(ptr);
+}
+
+static __inline void *
+kmem_realloc(void *ptr, size_t newsize, size_t oldsize, int flags)
+{
+	void *new = kmem_alloc(newsize, flags);
+
+	if (likely(ptr != NULL)) {
+		if (likely(new != NULL))
+			memcpy(new, ptr, min(oldsize, newsize));
+		kmem_free(ptr, oldsize);
+	}
+
+	return new;
+}
+
+static __inline kmem_zone_t *
+kmem_zone_init(int size, char *zone_name)
+{
+	return kmem_cache_create(zone_name, size, 0, 0, NULL, NULL);
+}
 
-extern void	    *kmem_alloc(size_t, int);
-extern void	    *kmem_realloc(void *, size_t, size_t, int);
-extern void	    *kmem_zalloc(size_t, int);
-extern void         kmem_free(void *, size_t);
+static __inline void *
+kmem_zone_alloc(kmem_zone_t *zone, int flags)
+{
+	return kmem_cache_alloc(zone, flag_convert(flags));
+}
 
-typedef void	    (*kmem_shake_func_t)(void);
+static __inline void *
+kmem_zone_zalloc(kmem_zone_t *zone, int flags)
+{
+	void *ptr = kmem_zone_alloc(zone, flags);
+	if (likely(ptr != NULL))
+		memset(ptr, 0, kmem_cache_size(zone));
+	return ptr;
+}
 
-extern void	    kmem_shake_register(kmem_shake_func_t);
-extern void	    kmem_shake_deregister(kmem_shake_func_t);
+static __inline void
+kmem_zone_free(kmem_zone_t *zone, void *ptr)
+{
+	kmem_cache_free(zone, ptr);
+}
 
 #endif /* __XFS_SUPPORT_KMEM_H__ */

fs/xfs/xfs_ag.h

 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * 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
@@ -185,9 +185,8 @@ xfs_agblock_t xfs_agfl_block(struct xfs_mount *mp);
 #endif
 #define XFS_AGFL_SIZE(mp)	((mp)->m_sb.sb_sectsize / sizeof(xfs_agblock_t))
 
-/* -- nathans TODO ... use of BBSIZE here - should be sector size -- */
 typedef struct xfs_agfl {
-	xfs_agblock_t	agfl_bno[BBSIZE/sizeof(xfs_agblock_t)];
+	xfs_agblock_t	agfl_bno[1];	/* actually XFS_AGFL_SIZE(mp) */
 } xfs_agfl_t;
 
 /*

fs/xfs/xfs_buf.h

 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * 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
@@ -283,7 +283,6 @@ static inline int	XFS_bwrite(page_buf_t *pb)
 	return error;
 }
 
-
 #define XFS_bdwrite(pb)		     \
 	    pagebuf_iostart(pb, PBF_DELWRI | PBF_ASYNC)
 
@@ -307,15 +306,15 @@ static inline int xfs_bdwrite(void *mp, page_buf_t *bp)
  * of its metadata.
  */
 
-extern void XFS_bflush(xfs_buftarg_t *);
-#define xfs_binval(buftarg) XFS_bflush(buftarg)
+#define xfs_binval(buftarg)	xfs_flush_buftarg(buftarg)
+
+#define XFS_bflush(buftarg)	xfs_flush_buftarg(buftarg)
 
 #define xfs_incore_relse(buftarg,delwri_only,wait)	\
 	xfs_relse_buftarg(buftarg)
 
 #define xfs_baread(target, rablkno, ralen)  \
-	pagebuf_readahead((target), (rablkno), \
-			  (ralen), PBF_DONT_BLOCK)
+	pagebuf_readahead((target), (rablkno), (ralen), PBF_DONT_BLOCK)
 
 #define XFS_getrbuf(sleep,mp)	\
 	pagebuf_get_empty((mp)->m_ddev_targp)

fs/xfs/xfs_iget.c

@@ -252,6 +252,11 @@ xfs_iget_core(
 			if (newnode) {
 				xfs_iocore_inode_reinit(ip);
 			}
+
+			XFS_MOUNT_ILOCK(mp);
+			list_del_init(&ip->i_reclaim);
+			XFS_MOUNT_IUNLOCK(mp);
+
 			vn_trace_exit(vp, "xfs_iget.found",
 						(inst_t *)__return_address);
 			goto return_ip;
@@ -467,8 +472,10 @@ xfs_iget(
 			}
 
 			bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
-			if (bdp == NULL)
+			if (bdp == NULL) {
+				XFS_STATS_INC(xfsstats.xs_ig_dup);
 				goto inode_allocate;
+			}
 			ip = XFS_BHVTOI(bdp);
 			if (lock_flags != 0)
 				xfs_ilock(ip, lock_flags);
@@ -720,6 +727,9 @@ xfs_iextract(
 		}
 	}
 
+	/* Deal with the deleted inodes list */
+	list_del_init(&ip->i_reclaim);
+
 	mp->m_ireclaims++;
 	XFS_MOUNT_IUNLOCK(mp);
 }

fs/xfs/xfs_inode.c

@@ -656,7 +656,9 @@ xfs_iformat_extents(
 	int		nex;
 	int		real_size;
 	int		size;
+#if ARCH_CONVERT != ARCH_NOCONVERT
 	int		i;
+#endif
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	nex = XFS_DFORK_NEXTENTS_ARCH(dip, whichfork, ARCH_CONVERT);
@@ -976,6 +978,8 @@ xfs_iread(
 			XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
 	}
 
+	INIT_LIST_HEAD(&ip->i_reclaim);
+
 	/*
 	 * The inode format changed when we moved the link count and
 	 * made it 32 bits long.  If this is an old format inode,
@@ -2625,6 +2629,15 @@ xfs_iunpin(
 	ASSERT(atomic_read(&ip->i_pincount) > 0);
 
 	if (atomic_dec_and_test(&ip->i_pincount)) {
+		vnode_t	*vp = XFS_ITOV_NULL(ip);
+
+		/* make sync come back and flush this inode */
+		if (vp) {
+			struct inode	*inode = LINVFS_GET_IP(vp);
+
+			mark_inode_dirty_sync(inode);
+		}
+
 		wake_up(&ip->i_ipin_wait);
 	}
 }
@@ -3640,6 +3653,8 @@ xfs_ichgtime(xfs_inode_t *ip,
 	 */
 	SYNCHRONIZE();
 	ip->i_update_core = 1;
+	if (!(inode->i_state & I_LOCK))
+		mark_inode_dirty(inode);
 }
 
 #ifdef XFS_ILOCK_TRACE

fs/xfs/xfs_inode.h

@@ -243,6 +243,7 @@ typedef struct xfs_inode {
 	struct xfs_inode	*i_mprev;	/* ptr to prev inode */
 	struct xfs_inode	**i_prevp;	/* ptr to prev i_next */
 	struct xfs_mount	*i_mount;	/* fs mount struct ptr */
+	struct list_head	i_reclaim;	/* reclaim list */
 	struct bhv_desc		i_bhv_desc;	/* inode behavior descriptor*/
 	struct xfs_dquot	*i_udquot;	/* user dquot */
 	struct xfs_dquot	*i_gdquot;	/* group dquot */
@@ -477,7 +478,7 @@ void		xfs_iunlock_map_shared(xfs_inode_t *, uint);
 void		xfs_ifunlock(xfs_inode_t *);
 void		xfs_ireclaim(xfs_inode_t *);
 int		xfs_finish_reclaim(xfs_inode_t *, int, int);
-int		xfs_finish_reclaim_all(struct xfs_mount *);
+int		xfs_finish_reclaim_all(struct xfs_mount *, int);
 
 /*
  * xfs_inode.c prototypes.

fs/xfs/xfs_log.c

 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * 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
@@ -413,19 +413,6 @@ xfs_log_release_iclog(xfs_mount_t *mp,
 	return 0;
 }
 
-/*
- * Initialize log manager data.  This routine is intended to be called when
- * a system boots up.  It is not a per filesystem initialization.
- *
- * As you can see, we currently do nothing.
- */
-int
-xfs_log_init(void)
-{
-	return( 0 );
-}
-
-
 /*
  *  1. Reserve an amount of on-disk log space and return a ticket corresponding
  *	to the reservation.
@@ -497,8 +484,6 @@ xfs_log_mount(xfs_mount_t	*mp,
 	      xfs_daddr_t	blk_offset,
 	      int		num_bblks)
 {
-	xlog_t *log;
-
 	if (!(mp->m_flags & XFS_MOUNT_NORECOVERY))
 		cmn_err(CE_NOTE, "XFS mounting filesystem %s", mp->m_fsname);
 	else {
@@ -508,7 +493,7 @@ xfs_log_mount(xfs_mount_t	*mp,
 		ASSERT(XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY);
 	}
 
-	mp->m_log = log = xlog_alloc_log(mp, log_dev, blk_offset, num_bblks);
+	mp->m_log = xlog_alloc_log(mp, log_dev, blk_offset, num_bblks);
 
 #if defined(DEBUG) || defined(XLOG_NOLOG)
 	if (! xlog_debug) {
@@ -528,19 +513,19 @@ xfs_log_mount(xfs_mount_t	*mp,
 		if (readonly)
 			vfsp->vfs_flag &= ~VFS_RDONLY;
 
-		error = xlog_recover(log, readonly);
+		error = xlog_recover(mp->m_log, readonly);
 
 		if (readonly)
 			vfsp->vfs_flag |= VFS_RDONLY;
 		if (error) {
 			cmn_err(CE_WARN, "XFS: log mount/recovery failed");
-			xlog_unalloc_log(log);
+			xlog_unalloc_log(mp->m_log);
 			return error;
 		}
 	}
 
 	/* Normal transactions can now occur */
-	log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
+	mp->m_log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
 
 	/* End mounting message in xfs_log_mount_finish */
 	return 0;
@@ -1191,14 +1176,13 @@ xlog_alloc_log(xfs_mount_t	*mp,
 	int			i;
 	int			iclogsize;
 
-	log = (void *)kmem_zalloc(sizeof(xlog_t), KM_SLEEP);
+	log = (xlog_t *)kmem_zalloc(sizeof(xlog_t), KM_SLEEP);
 
 	log->l_mp	   = mp;
 	log->l_dev	   = log_dev;
 	log->l_logsize     = BBTOB(num_bblks);
 	log->l_logBBstart  = blk_offset;
 	log->l_logBBsize   = num_bblks;
-	log->l_roundoff	   = 0;
 	log->l_covered_state = XLOG_STATE_COVER_IDLE;
 	log->l_flags	   |= XLOG_ACTIVE_RECOVERY;
 
@@ -1207,12 +1191,24 @@ xlog_alloc_log(xfs_mount_t	*mp,
 	/* log->l_tail_lsn = 0x100000000LL; cycle = 1; current block = 0 */
 	log->l_last_sync_lsn = log->l_tail_lsn;
 	log->l_curr_cycle  = 1;	    /* 0 is bad since this is initial value */
-	log->l_curr_block  = 0;		/* filled in by xlog_recover */
-	log->l_grant_reserve_bytes = 0;
 	log->l_grant_reserve_cycle = 1;
-	log->l_grant_write_bytes = 0;
 	log->l_grant_write_cycle = 1;
-	log->l_quotaoffs_flag = 0;      /* XFS_LI_QUOTAOFF logitems */
+
+	if (XFS_SB_VERSION_HASLOGV2(&mp->m_sb)) {
+		if (mp->m_sb.sb_logsunit <= 1) {
+			log->l_stripemask = 1;
+		} else {
+			log->l_stripemask = 1 <<
+				xfs_highbit32(mp->m_sb.sb_logsunit >> BBSHIFT);
+		}
+	}
+	if (XFS_SB_VERSION_HASSECTOR(&mp->m_sb)) {
+		log->l_sectbb_log = mp->m_sb.sb_logsectlog - BBSHIFT;
+		ASSERT(log->l_sectbb_log <= mp->m_sectbb_log);
+		ASSERT(XFS_SB_VERSION_HASLOGV2(&mp->m_sb));
+		ASSERT(mp->m_sb.sb_logsectlog >= BBSHIFT);
+	}
+	log->l_sectbb_mask = (1 << log->l_sectbb_log) - 1;
 
 	xlog_get_iclog_buffer_size(mp, log);
 
@@ -2811,10 +2807,9 @@ xlog_state_switch_iclogs(xlog_t		*log,
 
 	/* Round up to next log-sunit */
 	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
-		if (log->l_curr_block & (log->l_mp->m_lstripemask - 1)) {
-			roundup = log->l_mp->m_lstripemask -
-				(log->l_curr_block &
-				 (log->l_mp->m_lstripemask - 1));
+		if (log->l_curr_block & (log->l_stripemask - 1)) {
+			roundup = log->l_stripemask -
+				(log->l_curr_block & (log->l_stripemask - 1));
 		} else {
 			roundup = 0;
 		}
@@ -3293,15 +3288,17 @@ xlog_verify_disk_cycle_no(xlog_t	 *log,
 {
     xfs_buf_t	*bp;
     uint	cycle_no;
+    xfs_caddr_t ptr;
     xfs_daddr_t	i;
 
     if (BLOCK_LSN(iclog->ic_header.h_lsn, ARCH_CONVERT) < 10) {
 	cycle_no = CYCLE_LSN(iclog->ic_header.h_lsn, ARCH_CONVERT);
-	bp = xlog_get_bp(1, log->l_mp);
+	bp = xlog_get_bp(log, 1);
 	ASSERT(bp);
 	for (i = 0; i < BLOCK_LSN(iclog->ic_header.h_lsn, ARCH_CONVERT); i++) {
 	    xlog_bread(log, i, 1, bp);
-	    if (GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT) != cycle_no)
+	    ptr = xlog_align(log, i, 1, bp);
+	    if (GET_CYCLE(ptr, ARCH_CONVERT) != cycle_no)
 		xlog_warn("XFS: xlog_verify_disk_cycle_no: bad cycle no");
 	}
 	xlog_put_bp(bp);

fs/xfs/xfs_log.h

 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * 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
@@ -153,7 +153,6 @@ xfs_lsn_t xfs_log_done(struct xfs_mount *mp,
 int	  xfs_log_force(struct xfs_mount *mp,
 			xfs_lsn_t	 lsn,
 			uint		 flags);
-int	  xfs_log_init(void);
 int	  xfs_log_mount(struct xfs_mount *mp,
 			dev_t		 log_dev,
 			xfs_daddr_t		 start_block,

fs/xfs/xfs_log_priv.h

 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * 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
@@ -73,6 +73,9 @@ int xlog_btolrbb(int b);
 
 #define XLOG_HEADER_SIZE	512
 
+#define XLOG_REC_SHIFT(log) \
+	BTOBB(1 << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
+	 XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
 #define XLOG_TOTAL_REC_SHIFT(log) \
 	BTOBB(XLOG_MAX_ICLOGS << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
 	 XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
@@ -202,9 +205,9 @@ void xlog_grant_add_space(struct log *log, int bytes, int type);
 #define LOG_LOCK(log)		mutex_spinlock(&(log)->l_icloglock)
 #define LOG_UNLOCK(log, s)	mutex_spinunlock(&(log)->l_icloglock, s)
 
-#define xlog_panic(s)		{cmn_err(CE_PANIC, s); }
-#define xlog_exit(s)		{cmn_err(CE_PANIC, s); }
-#define xlog_warn(s)		{cmn_err(CE_WARN, s); }
+#define xlog_panic(args...)	cmn_err(CE_PANIC, ## args)
+#define xlog_exit(args...)	cmn_err(CE_PANIC, ## args)
+#define xlog_warn(args...)	cmn_err(CE_WARN, ## args)
 
 /*
  * In core log state
@@ -403,6 +406,7 @@ typedef struct xlog_rec_ext_header {
 	uint	  xh_cycle;	/* write cycle of log			: 4 */
 	uint	  xh_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; /*	: 256 */
 } xlog_rec_ext_header_t;
+
 #ifdef __KERNEL__
 /*
  * - A log record header is 512 bytes.  There is plenty of room to grow the
@@ -441,12 +445,10 @@ typedef struct xlog_iclog_fields {
 	char			*ic_datap;	/* pointer to iclog data */
 } xlog_iclog_fields_t;
 
-typedef struct xlog_in_core2 {
-	union {
+typedef union xlog_in_core2 {
 	xlog_rec_header_t	hic_header;
 	xlog_rec_ext_header_t	hic_xheader;
 	char			hic_sector[XLOG_HEADER_SIZE];
-	} ic_h;
 } xlog_in_core_2_t;
 
 typedef struct xlog_in_core {
@@ -473,7 +475,7 @@ typedef struct xlog_in_core {
 #define	ic_bwritecnt	hic_fields.ic_bwritecnt
 #define	ic_state	hic_fields.ic_state
 #define ic_datap	hic_fields.ic_datap
-#define ic_header	hic_data->ic_h.hic_header
+#define ic_header	hic_data->hic_header
 
 /*
  * The reservation head lsn is not made up of a cycle number and block number.
@@ -530,8 +532,11 @@ typedef struct log {
     uint		l_flags;
     uint		l_quotaoffs_flag;/* XFS_DQ_*, if QUOTAOFFs found */
     struct xfs_buf_cancel **l_buf_cancel_table;
+    int			l_stripemask;	/* log stripe mask */
     int			l_iclog_hsize;  /* size of iclog header */
     int			l_iclog_heads;  /* number of iclog header sectors */
+    uint		l_sectbb_log;   /* log2 of sector size in bbs */
+    uint		l_sectbb_mask;  /* sector size in bbs alignment mask */
 } xlog_t;
 
 
@@ -546,11 +551,13 @@ extern int	 xlog_print_find_oldest(xlog_t *log, xfs_daddr_t *last_blk);
 extern int	 xlog_recover(xlog_t *log, int readonly);
 extern int	 xlog_recover_finish(xlog_t *log, int mfsi_flags);
 extern void	 xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog);
-extern struct xfs_buf *xlog_get_bp(int,xfs_mount_t *);
-extern void	 xlog_put_bp(struct xfs_buf *);
-extern int	 xlog_bread(xlog_t *, xfs_daddr_t blkno, int bblks, struct xfs_buf *bp);
 extern void	 xlog_recover_process_iunlinks(xlog_t *log);
 
+extern struct xfs_buf *xlog_get_bp(xlog_t *, int);
+extern void	 xlog_put_bp(struct xfs_buf *);
+extern int	 xlog_bread(xlog_t *, xfs_daddr_t, int, struct xfs_buf *);
+extern xfs_caddr_t xlog_align(xlog_t *, xfs_daddr_t, int, struct xfs_buf *);
+
 #define XLOG_TRACE_GRAB_FLUSH  1
 #define XLOG_TRACE_REL_FLUSH   2
 #define XLOG_TRACE_SLEEP_FLUSH 3

fs/xfs/xfs_log_recover.c

 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * 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
@@ -65,53 +65,68 @@
 #include "xfs_quota.h"
 #include "xfs_rw.h"
 
-STATIC int	xlog_find_zeroed(struct log *log, xfs_daddr_t *blk_no);
-
-STATIC int	xlog_clear_stale_blocks(xlog_t	*log, xfs_lsn_t tail_lsn);
+STATIC int	xlog_find_zeroed(xlog_t *, xfs_daddr_t *);
+STATIC int	xlog_clear_stale_blocks(xlog_t *, xfs_lsn_t);
 STATIC void	xlog_recover_insert_item_backq(xlog_recover_item_t **q,
 					       xlog_recover_item_t *item);
-
 #if defined(DEBUG)
-STATIC void	xlog_recover_check_summary(xlog_t *log);
-STATIC void	xlog_recover_check_ail(xfs_mount_t *mp, xfs_log_item_t *lip,
-				       int gen);
+STATIC void	xlog_recover_check_summary(xlog_t *);
+STATIC void	xlog_recover_check_ail(xfs_mount_t *, xfs_log_item_t *, int);
 #else
 #define	xlog_recover_check_summary(log)
 #define	xlog_recover_check_ail(mp, lip, gen)
-#endif /* DEBUG */
+#endif
+
+
+/*
+ * Sector aligned buffer routines for buffer create/read/write/access
+ */
 
+#define XLOG_SECTOR_ROUNDUP_BBCOUNT(log, bbs)	\
+	( ((log)->l_sectbb_mask && (bbs & (log)->l_sectbb_mask)) ? \
+	((bbs + (log)->l_sectbb_mask + 1) & ~(log)->l_sectbb_mask) : (bbs) )
+#define XLOG_SECTOR_ROUNDDOWN_BLKNO(log, bno)	((bno) & ~(log)->l_sectbb_mask)
 
 xfs_buf_t *
-xlog_get_bp(int num_bblks, xfs_mount_t *mp)
+xlog_get_bp(
+	xlog_t		*log,
+	int		num_bblks)
 {
-	xfs_buf_t   *bp;
-
 	ASSERT(num_bblks > 0);
 
-	bp = XFS_ngetrbuf(BBTOB(num_bblks),mp);
-	return bp;
-}	/* xlog_get_bp */
-
+	if (log->l_sectbb_log) {
+		if (num_bblks > 1)
+			num_bblks += XLOG_SECTOR_ROUNDUP_BBCOUNT(log, 1);
+		num_bblks = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, num_bblks);
+	}
+	return XFS_ngetrbuf(BBTOB(num_bblks), log->l_mp);
+}
 
 void
-xlog_put_bp(xfs_buf_t *bp)
+xlog_put_bp(
+	xfs_buf_t	*bp)
 {
 	XFS_nfreerbuf(bp);
-}	/* xlog_put_bp */
+}
 
 
 /*
  * nbblks should be uint, but oh well.  Just want to catch that 32-bit length.
  */
 int
-xlog_bread(xlog_t	*log,
+xlog_bread(
+	xlog_t		*log,
 	xfs_daddr_t	blk_no,
 	int		nbblks,
 	xfs_buf_t	*bp)
 {
 	int		error;
 
-	ASSERT(log);
+	if (log->l_sectbb_log) {
+		blk_no = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, blk_no);
+		nbblks = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, nbblks);
+	}
+
 	ASSERT(nbblks > 0);
 	ASSERT(BBTOB(nbblks) <= XFS_BUF_SIZE(bp));
 	ASSERT(bp);
@@ -123,14 +138,11 @@ xlog_bread(xlog_t	*log,
 	XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp);
 
 	xfsbdstrat(log->l_mp, bp);
-	if ((error = xfs_iowait(bp))) {
+	if ((error = xfs_iowait(bp)))
 		xfs_ioerror_alert("xlog_bread", log->l_mp,
 				  bp, XFS_BUF_ADDR(bp));
-		return (error);
-	}
 	return error;
-}	/* xlog_bread */
-
+}
 
 /*
  * Write out the buffer at the given block for the given number of blocks.
@@ -140,12 +152,17 @@ xlog_bread(xlog_t	*log,
 int
 xlog_bwrite(
 	xlog_t		*log,
-	int	blk_no,
+	xfs_daddr_t	blk_no,
 	int		nbblks,
 	xfs_buf_t	*bp)
 {
 	int		error;
 
+	if (log->l_sectbb_log) {
+		blk_no = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, blk_no);
+		nbblks = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, nbblks);
+	}
+
 	ASSERT(nbblks > 0);
 	ASSERT(BBTOB(nbblks) <= XFS_BUF_SIZE(bp));
 
@@ -160,34 +177,58 @@ xlog_bwrite(
 	if ((error = xfs_bwrite(log->l_mp, bp)))
 		xfs_ioerror_alert("xlog_bwrite", log->l_mp,
 				  bp, XFS_BUF_ADDR(bp));
+	return error;
+}
 
-	return (error);
-}	/* xlog_bwrite */
+xfs_caddr_t
+xlog_align(
+	xlog_t		*log,
+	xfs_daddr_t	blk_no,
+	int		nbblks,
+	xfs_buf_t	*bp)
+{
+	xfs_caddr_t	ptr;
+
+	if (!log->l_sectbb_log)
+		return XFS_BUF_PTR(bp);
+
+	ptr = XFS_BUF_PTR(bp) + BBTOB((int)blk_no & log->l_sectbb_mask);
+	ASSERT(XFS_BUF_SIZE(bp) >=
+		BBTOB(nbblks + (blk_no & log->l_sectbb_mask)));
+	return ptr;
+}
 
 #ifdef DEBUG
 /*
- * check log record header for recovery
+ * dump debug superblock and log record information
  */
-static void
-xlog_header_check_dump(xfs_mount_t *mp, xlog_rec_header_t *head)
+STATIC void
+xlog_header_check_dump(
+	xfs_mount_t		*mp,
+	xlog_rec_header_t	*head)
 {
 	int			b;
 
 	printk("%s:  SB : uuid = ", __FUNCTION__);
-    for (b=0;b<16;b++) printk("%02x",((unsigned char *)&mp->m_sb.sb_uuid)[b]);
-    printk(", fmt = %d\n",XLOG_FMT);
+	for (b = 0; b < 16; b++)
+		printk("%02x",((unsigned char *)&mp->m_sb.sb_uuid)[b]);
+	printk(", fmt = %d\n", XLOG_FMT);
 	printk("    log : uuid = ");
-    for (b=0;b<16;b++) printk("%02x",((unsigned char *)&head->h_fs_uuid)[b]);
+	for (b = 0; b < 16; b++)
+		printk("%02x",((unsigned char *)&head->h_fs_uuid)[b]);
 	printk(", fmt = %d\n", INT_GET(head->h_fmt, ARCH_CONVERT));
 }
+#else
+#define xlog_header_check_dump(mp, head)
 #endif
 
 /*
  * check log record header for recovery
  */
-
 STATIC int
-xlog_header_check_recover(xfs_mount_t *mp, xlog_rec_header_t *head)
+xlog_header_check_recover(
+	xfs_mount_t		*mp,
+	xlog_rec_header_t	*head)
 {
 	ASSERT(INT_GET(head->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
 
@@ -196,57 +237,48 @@ xlog_header_check_recover(xfs_mount_t *mp, xlog_rec_header_t *head)
 	 * (XLOG_FMT_UNKNOWN). This stops us from trying to recover
 	 * a dirty log created in IRIX.
 	 */
-
 	if (unlikely(INT_GET(head->h_fmt, ARCH_CONVERT) != XLOG_FMT)) {
-	xlog_warn("XFS: dirty log written in incompatible format - can't recover");
-#ifdef DEBUG
+		xlog_warn(
+	"XFS: dirty log written in incompatible format - can't recover");
 		xlog_header_check_dump(mp, head);
-#endif
 		XFS_ERROR_REPORT("xlog_header_check_recover(1)",
 				 XFS_ERRLEVEL_HIGH, mp);
 		return XFS_ERROR(EFSCORRUPTED);
 	} else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
-	xlog_warn("XFS: dirty log entry has mismatched uuid - can't recover");
-#ifdef DEBUG
+		xlog_warn(
+	"XFS: dirty log entry has mismatched uuid - can't recover");
 		xlog_header_check_dump(mp, head);
-#endif
 		XFS_ERROR_REPORT("xlog_header_check_recover(2)",
 				 XFS_ERRLEVEL_HIGH, mp);
 		return XFS_ERROR(EFSCORRUPTED);
 	}
-
 	return 0;
 }
 
 /*
  * read the head block of the log and check the header
  */
-
 STATIC int
-xlog_header_check_mount(xfs_mount_t *mp, xlog_rec_header_t *head)
+xlog_header_check_mount(
+	xfs_mount_t		*mp,
+	xlog_rec_header_t	*head)
 {
 	ASSERT(INT_GET(head->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
 
 	if (uuid_is_nil(&head->h_fs_uuid)) {
-
 		/*
 		 * IRIX doesn't write the h_fs_uuid or h_fmt fields. If
 		 * h_fs_uuid is nil, we assume this log was last mounted
 		 * by IRIX and continue.
 		 */
-
 		xlog_warn("XFS: nil uuid in log - IRIX style log");
-
 	} else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
 		xlog_warn("XFS: log has mismatched uuid - can't recover");
-#ifdef DEBUG
 		xlog_header_check_dump(mp, head);
-#endif
 		XFS_ERROR_REPORT("xlog_header_check_mount",
 				 XFS_ERRLEVEL_HIGH, mp);
 		return XFS_ERROR(EFSCORRUPTED);
 	}
-
 	return 0;
 }
 
@@ -255,6 +287,7 @@ xlog_recover_iodone(
 	struct xfs_buf	*bp)
 {
 	xfs_mount_t	*mp;
+
 	ASSERT(XFS_BUF_FSPRIVATE(bp, void *));
 
 	if (XFS_BUF_GETERROR(bp)) {
@@ -279,12 +312,14 @@ xlog_recover_iodone(
  * necessarily be perfect.
  */
 int
-xlog_find_cycle_start(xlog_t		*log,
+xlog_find_cycle_start(
+	xlog_t		*log,
 	xfs_buf_t	*bp,
 	xfs_daddr_t	first_blk,
 	xfs_daddr_t	*last_blk,
 	uint		cycle)
 {
+	xfs_caddr_t	offset;
 	xfs_daddr_t	mid_blk;
 	uint		mid_cycle;
 	int		error;
@@ -293,7 +328,8 @@ xlog_find_cycle_start(xlog_t		*log,
 	while (mid_blk != first_blk && mid_blk != *last_blk) {
 		if ((error = xlog_bread(log, mid_blk, 1, bp)))
 			return error;
-		mid_cycle = GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT);
+		offset = xlog_align(log, mid_blk, 1, bp);
+		mid_cycle = GET_CYCLE(offset, ARCH_CONVERT);
 		if (mid_cycle == cycle) {
 			*last_blk = mid_blk;
 			/* last_half_cycle == mid_cycle */
@@ -307,8 +343,7 @@ xlog_find_cycle_start(xlog_t		*log,
 	       (mid_blk == *last_blk && mid_blk-1 == first_blk));
 
 	return 0;
-}	/* xlog_find_cycle_start */
-
+}
 
 /*
  * Check that the range of blocks does not contain the cycle number
@@ -320,9 +355,9 @@ xlog_find_cycle_start(xlog_t		*log,
  * Set blkno to -1 if we encounter no errors.  This is an invalid block number
  * since we don't ever expect logs to get this large.
  */
-
 STATIC int
-xlog_find_verify_cycle( xlog_t		*log,
+xlog_find_verify_cycle(
+	xlog_t		*log,
 	xfs_daddr_t	start_blk,
 	int		nbblks,
 	uint		stop_on_cycle_no,
@@ -331,16 +366,16 @@ xlog_find_verify_cycle( xlog_t		*log,
 	xfs_daddr_t	i, j;
 	uint		cycle;
 	xfs_buf_t	*bp;
-	char                    *buf        = NULL;
-	int			error       = 0;
 	xfs_daddr_t	bufblks;
+	xfs_caddr_t	buf = NULL;
+	int		error = 0;
 
 	bufblks = 1 << ffs(nbblks);
 
-	while (!(bp = xlog_get_bp(bufblks, log->l_mp))) {
+	while (!(bp = xlog_get_bp(log, bufblks))) {
 		/* can't get enough memory to do everything in one big buffer */
 		bufblks >>= 1;
-		if (!bufblks)
+		if (bufblks <= log->l_sectbb_log)
 			return ENOMEM;
 	}
 
@@ -352,7 +387,7 @@ xlog_find_verify_cycle( xlog_t		*log,
 		if ((error = xlog_bread(log, i, bcount, bp)))
 			goto out;
 
-		buf = XFS_BUF_PTR(bp);
+		buf = xlog_align(log, i, bcount, bp);
 		for (j = 0; j < bcount; j++) {
 			cycle = GET_CYCLE(buf, ARCH_CONVERT);
 			if (cycle == stop_on_cycle_no) {
@@ -368,10 +403,8 @@ xlog_find_verify_cycle( xlog_t		*log,
 
 out:
 	xlog_put_bp(bp);
-
 	return error;
-}	/* xlog_find_verify_cycle */
-
+}
 
 /*
  * Potentially backup over partial log record write.
@@ -385,16 +418,16 @@ xlog_find_verify_cycle( xlog_t		*log,
  * extra_bblks is the number of blocks potentially verified on a previous
  * call to this routine.
  */
-
 STATIC int
-xlog_find_verify_log_record(xlog_t	*log,
+xlog_find_verify_log_record(
+	xlog_t			*log,
 	xfs_daddr_t		start_blk,
 	xfs_daddr_t		*last_blk,
 	int			extra_bblks)
 {
 	xfs_daddr_t		i;
 	xfs_buf_t		*bp;
-    char                *buf        = NULL;
+	xfs_caddr_t		offset = NULL;
 	xlog_rec_header_t	*head = NULL;
 	int			error = 0;
 	int			smallmem = 0;
@@ -403,35 +436,41 @@ xlog_find_verify_log_record(xlog_t	*log,
 
 	ASSERT(start_blk != 0 || *last_blk != start_blk);
 
-    if (!(bp = xlog_get_bp(num_blks, log->l_mp))) {
-	if (!(bp = xlog_get_bp(1, log->l_mp)))
+	if (!(bp = xlog_get_bp(log, num_blks))) {
+		if (!(bp = xlog_get_bp(log, 1)))
 			return ENOMEM;
 		smallmem = 1;
-	buf = XFS_BUF_PTR(bp);
 	} else {
 		if ((error = xlog_bread(log, start_blk, num_blks, bp)))
 			goto out;
-	buf = XFS_BUF_PTR(bp) + ((num_blks - 1) << BBSHIFT);
+		offset = xlog_align(log, start_blk, num_blks, bp);
+		offset += ((num_blks - 1) << BBSHIFT);
 	}
 
 	for (i = (*last_blk) - 1; i >= 0; i--) {
 		if (i < start_blk) {
 			/* legal log record not found */
-	    xlog_warn("XFS: Log inconsistent (didn't find previous header)");
+			xlog_warn(
+		"XFS: Log inconsistent (didn't find previous header)");
 			ASSERT(0);
 			error = XFS_ERROR(EIO);
 			goto out;
 		}
 
-	if (smallmem && (error = xlog_bread(log, i, 1, bp)))
+		if (smallmem) {
+			if ((error = xlog_bread(log, i, 1, bp)))
 				goto out;
-	head = (xlog_rec_header_t*)buf;
+			offset = xlog_align(log, i, 1, bp);
+		}
 
-	if (INT_GET(head->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM)
+		head = (xlog_rec_header_t *)offset;
+
+		if (XLOG_HEADER_MAGIC_NUM ==
+		    INT_GET(head->h_magicno, ARCH_CONVERT))
 			break;
 
 		if (!smallmem)
-	    buf -= BBSIZE;
+			offset -= BBSIZE;
 	}
 
 	/*
@@ -444,10 +483,10 @@ xlog_find_verify_log_record(xlog_t	*log,
 		goto out;
 	}
 
-    /* we have the final block of the good log (the first block
+	/*
+	 * We have the final block of the good log (the first block
 	 * of the log record _before_ the head. So we check the uuid.
 	 */
-
 	if ((error = xlog_header_check_mount(log->l_mp, head)))
 		goto out;
 
@@ -469,14 +508,13 @@ xlog_find_verify_log_record(xlog_t	*log,
 	}
 
 	if (*last_blk - i + extra_bblks
-		!= BTOBB(INT_GET(head->h_len, ARCH_CONVERT))+xhdrs)
+			!= BTOBB(INT_GET(head->h_len, ARCH_CONVERT)) + xhdrs)
 		*last_blk = i;
 
 out:
 	xlog_put_bp(bp);
-
 	return error;
-}	/* xlog_find_verify_log_record */
+}
 
 /*
  * Head is defined to be the point of the log where the next log write
@@ -489,15 +527,15 @@ xlog_find_verify_log_record(xlog_t	*log,
  * last_blk contains the block number of the first block with a given
  * cycle number.
  *
- * Also called from xfs_log_print.c
- *
  * Return: zero if normal, non-zero if error.
  */
 int
-xlog_find_head(xlog_t  *log,
+xlog_find_head(
+	xlog_t 		*log,
 	xfs_daddr_t	*return_head_blk)
 {
 	xfs_buf_t	*bp;
+	xfs_caddr_t	offset;
 	xfs_daddr_t	new_blk, first_blk, start_blk, last_blk, head_blk;
 	int		num_scan_bblks;
 	uint		first_half_cycle, last_half_cycle;
@@ -508,7 +546,7 @@ xlog_find_head(xlog_t  *log,
 	if ((error = xlog_find_zeroed(log, &first_blk)) == -1) {
 		*return_head_blk = first_blk;
 
-	/* is the whole lot zeroed? */
+		/* Is the whole lot zeroed? */
 		if (!first_blk) {
 			/* Linux XFS shouldn't generate totally zeroed logs -
 			 * mkfs etc write a dummy unmount record to a fresh
@@ -524,17 +562,19 @@ xlog_find_head(xlog_t  *log,
 	}
 
 	first_blk = 0;			/* get cycle # of 1st block */
-    bp = xlog_get_bp(1,log->l_mp);
+	bp = xlog_get_bp(log, 1);
 	if (!bp)
 		return ENOMEM;
 	if ((error = xlog_bread(log, 0, 1, bp)))
 		goto bp_err;
-    first_half_cycle = GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT);
+	offset = xlog_align(log, 0, 1, bp);
+	first_half_cycle = GET_CYCLE(offset, ARCH_CONVERT);
 
-    last_blk = head_blk = log_bbnum-1;		/* get cycle # of last block */
+	last_blk = head_blk = log_bbnum - 1;	/* get cycle # of last block */
 	if ((error = xlog_bread(log, last_blk, 1, bp)))
 		goto bp_err;
-    last_half_cycle = GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT);
+	offset = xlog_align(log, last_blk, 1, bp);
+	last_half_cycle = GET_CYCLE(offset, ARCH_CONVERT);
 	ASSERT(last_half_cycle != 0);
 
 	/*
@@ -550,44 +590,47 @@ xlog_find_head(xlog_t  *log,
 	 */
 	if (first_half_cycle == last_half_cycle) {
 		/*
-	 * In this case we believe that the entire log should have cycle
-	 * number last_half_cycle.  We need to scan backwards from the
-	 * end verifying that there are no holes still containing
-	 * last_half_cycle - 1.  If we find such a hole, then the start
-	 * of that hole will be the new head.  The simple case looks like
+		 * In this case we believe that the entire log should have
+		 * cycle number last_half_cycle.  We need to scan backwards
+		 * from the end verifying that there are no holes still
+		 * containing last_half_cycle - 1.  If we find such a hole,
+		 * then the start of that hole will be the new head.  The
+		 * simple case looks like
 		 *        x | x ... | x - 1 | x
 		 * Another case that fits this picture would be
 		 *        x | x + 1 | x ... | x
 		 * In this case the head really is somwhere at the end of the
-	 * log, as one of the latest writes at the beginning was incomplete.
+		 * log, as one of the latest writes at the beginning was
+		 * incomplete.
 		 * One more case is
 		 *        x | x + 1 | x ... | x - 1 | x
-	 * This is really the combination of the above two cases, and the
-	 * head has to end up at the start of the x-1 hole at the end of
-	 * the log.
+		 * This is really the combination of the above two cases, and
+		 * the head has to end up at the start of the x-1 hole at the
+		 * end of the log.
 		 *
 		 * In the 256k log case, we will read from the beginning to the
-	 * end of the log and search for cycle numbers equal to x-1.  We
-	 * don't worry about the x+1 blocks that we encounter, because
-	 * we know that they cannot be the head since the log started with
-	 * x.
+		 * end of the log and search for cycle numbers equal to x-1.
+		 * We don't worry about the x+1 blocks that we encounter,
+		 * because we know that they cannot be the head since the log
+		 * started with x.
 		 */
 		head_blk = log_bbnum;
 		stop_on_cycle = last_half_cycle - 1;
 	} else {
 		/*
-	 * In this case we want to find the first block with cycle number
-	 * matching last_half_cycle.  We expect the log to be some
-	 * variation on
+		 * In this case we want to find the first block with cycle
+		 * number matching last_half_cycle.  We expect the log to be
+		 * some variation on
 		 *        x + 1 ... | x ...
-	 * The first block with cycle number x (last_half_cycle) will be
-	 * where the new head belongs.  First we do a binary search for
-	 * the first occurrence of last_half_cycle.  The binary search
-	 * may not be totally accurate, so then we scan back from there
-	 * looking for occurrences of last_half_cycle before us.  If
-	 * that backwards scan wraps around the beginning of the log,
-	 * then we look for occurrences of last_half_cycle - 1 at the
-	 * end of the log.  The cases we're looking for look like
+		 * The first block with cycle number x (last_half_cycle) will
+		 * be where the new head belongs.  First we do a binary search
+		 * for the first occurrence of last_half_cycle.  The binary
+		 * search may not be totally accurate, so then we scan back
+		 * from there looking for occurrences of last_half_cycle before
+		 * us.  If that backwards scan wraps around the beginning of
+		 * the log, then we look for occurrences of last_half_cycle - 1
+		 * at the end of the log.  The cases we're looking for look
+		 * like
 		 *        x + 1 ... | x | x + 1 | x ...
 		 *                               ^ binary search stopped here
 		 * or
@@ -614,42 +657,46 @@ xlog_find_head(xlog_t  *log,
 		 * in one buffer.
 		 */
 		start_blk = head_blk - num_scan_bblks;
-	if ((error = xlog_find_verify_cycle(log, start_blk, num_scan_bblks,
+		if ((error = xlog_find_verify_cycle(log,
+						start_blk, num_scan_bblks,
 						stop_on_cycle, &new_blk)))
 			goto bp_err;
 		if (new_blk != -1)
 			head_blk = new_blk;
 	} else {		/* need to read 2 parts of log */
 		/*
-	 * We are going to scan backwards in the log in two parts.  First
-	 * we scan the physical end of the log.  In this part of the log,
-	 * we are looking for blocks with cycle number last_half_cycle - 1.
-	 * If we find one, then we know that the log starts there, as we've
-	 * found a hole that didn't get written in going around the end
-	 * of the physical log.  The simple case for this is
+		 * We are going to scan backwards in the log in two parts.
+		 * First we scan the physical end of the log.  In this part
+		 * of the log, we are looking for blocks with cycle number
+		 * last_half_cycle - 1.
+		 * If we find one, then we know that the log starts there, as
+		 * we've found a hole that didn't get written in going around
+		 * the end of the physical log.  The simple case for this is
 		 *        x + 1 ... | x ... | x - 1 | x
 		 *        <---------> less than scan distance
 		 * If all of the blocks at the end of the log have cycle number
-	 * last_half_cycle, then we check the blocks at the start of the
-	 * log looking for occurrences of last_half_cycle.  If we find one,
-	 * then our current estimate for the location of the first
-	 * occurrence of last_half_cycle is wrong and we move back to the
-	 * hole we've found.  This case looks like
+		 * last_half_cycle, then we check the blocks at the start of
+		 * the log looking for occurrences of last_half_cycle.  If we
+		 * find one, then our current estimate for the location of the
+		 * first occurrence of last_half_cycle is wrong and we move
+		 * back to the hole we've found.  This case looks like
 		 *        x + 1 ... | x | x + 1 | x ...
 		 *                               ^ binary search stopped here
-	 * Another case we need to handle that only occurs in 256k logs is
+		 * Another case we need to handle that only occurs in 256k
+		 * logs is
 		 *        x + 1 ... | x ... | x+1 | x ...
 		 *                   ^ binary search stops here
-	 * In a 256k log, the scan at the end of the log will see the x+1
-	 * blocks.  We need to skip past those since that is certainly not
-	 * the head of the log.  By searching for last_half_cycle-1 we
-	 * accomplish that.
+		 * In a 256k log, the scan at the end of the log will see the
+		 * x + 1 blocks.  We need to skip past those since that is
+		 * certainly not the head of the log.  By searching for
+		 * last_half_cycle-1 we accomplish that.
 		 */
 		start_blk = log_bbnum - num_scan_bblks + head_blk;
-	ASSERT(head_blk <= INT_MAX && (xfs_daddr_t) num_scan_bblks-head_blk >= 0);
+		ASSERT(head_blk <= INT_MAX &&
+			(xfs_daddr_t) num_scan_bblks - head_blk >= 0);
 		if ((error = xlog_find_verify_cycle(log, start_blk,
-			num_scan_bblks-(int)head_blk, (stop_on_cycle - 1),
-			&new_blk)))
+					num_scan_bblks - (int)head_blk,
+					(stop_on_cycle - 1), &new_blk)))
 			goto bp_err;
 		if (new_blk != -1) {
 			head_blk = new_blk;
@@ -657,33 +704,32 @@ xlog_find_head(xlog_t  *log,
 		}
 
 		/*
-	 * Scan beginning of log now.  The last part of the physical log
-	 * is good.  This scan needs to verify that it doesn't find the
-	 * last_half_cycle.
+		 * Scan beginning of log now.  The last part of the physical
+		 * log is good.  This scan needs to verify that it doesn't find
+		 * the last_half_cycle.
 		 */
 		start_blk = 0;
 		ASSERT(head_blk <= INT_MAX);
-	if ((error = xlog_find_verify_cycle(log, start_blk, (int) head_blk,
+		if ((error = xlog_find_verify_cycle(log,
+					start_blk, (int)head_blk,
 					stop_on_cycle, &new_blk)))
 			goto bp_err;
 		if (new_blk != -1)
 			head_blk = new_blk;
 	}
 
-bad_blk:
+ bad_blk:
 	/*
 	 * Now we need to make sure head_blk is not pointing to a block in
 	 * the middle of a log record.
 	 */
-    num_scan_bblks = BTOBB(XLOG_MAX_RECORD_BSIZE);
+	num_scan_bblks = XLOG_REC_SHIFT(log);
 	if (head_blk >= num_scan_bblks) {
 		start_blk = head_blk - num_scan_bblks; /* don't read head_blk */
 
 		/* start ptr at last block ptr before head_blk */
-	if ((error = xlog_find_verify_log_record(log,
-						 start_blk,
-						 &head_blk,
-						 0)) == -1) {
+		if ((error = xlog_find_verify_log_record(log, start_blk,
+							&head_blk, 0)) == -1) {
 			error = XFS_ERROR(EIO);
 			goto bp_err;
 		} else if (error)
@@ -691,18 +737,16 @@ xlog_find_head(xlog_t  *log,
 	} else {
 		start_blk = 0;
 		ASSERT(head_blk <= INT_MAX);
-	if ((error = xlog_find_verify_log_record(log,
-						 start_blk,
-						 &head_blk,
-						 0)) == -1) {
+		if ((error = xlog_find_verify_log_record(log, start_blk,
+							&head_blk, 0)) == -1) {
 			/* We hit the beginning of the log during our search */
 			start_blk = log_bbnum - num_scan_bblks + head_blk;
 			new_blk = log_bbnum;
-	    ASSERT(start_blk <= INT_MAX && (xfs_daddr_t) log_bbnum-start_blk >= 0);
+			ASSERT(start_blk <= INT_MAX &&
+				(xfs_daddr_t) log_bbnum-start_blk >= 0);
 			ASSERT(head_blk <= INT_MAX);
 			if ((error = xlog_find_verify_log_record(log,
-						     start_blk,
-						     &new_blk,
+							start_blk, &new_blk,
 							(int)head_blk)) == -1) {
 				error = XFS_ERROR(EIO);
 				goto bp_err;
@@ -727,14 +771,13 @@ xlog_find_head(xlog_t  *log,
 	 */
 	return 0;
 
-bp_err:
+ bp_err:
 	xlog_put_bp(bp);
 
 	if (error)
 	    xlog_warn("XFS: failed to find log head");
-
 	return error;
-}	/* xlog_find_head */
+}
 
 /*
  * Find the sync block number or the tail of the log.
@@ -753,13 +796,15 @@ xlog_find_head(xlog_t  *log,
  * available.
  */
 int
-xlog_find_tail(xlog_t		*log,
+xlog_find_tail(
+	xlog_t			*log,
 	xfs_daddr_t		*head_blk,
 	xfs_daddr_t		*tail_blk,
 	int			readonly)
 {
 	xlog_rec_header_t	*rhead;
 	xlog_op_header_t	*op_head;
+	xfs_caddr_t		offset = NULL;
 	xfs_buf_t		*bp;
 	int			error, i, found;
 	xfs_daddr_t		umount_data_blk;
@@ -775,13 +820,14 @@ xlog_find_tail(xlog_t		*log,
 	if ((error = xlog_find_head(log, head_blk)))
 		return error;
 
-	bp = xlog_get_bp(1,log->l_mp);
+	bp = xlog_get_bp(log, 1);
 	if (!bp)
 		return ENOMEM;
 	if (*head_blk == 0) {				/* special case */
 		if ((error = xlog_bread(log, 0, 1, bp)))
 			goto bread_err;
-		if (GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT) == 0) {
+		offset = xlog_align(log, 0, 1, bp);
+		if (GET_CYCLE(offset, ARCH_CONVERT) == 0) {
 			*tail_blk = 0;
 			/* leave all other log inited values alone */
 			goto exit;
@@ -795,8 +841,9 @@ xlog_find_tail(xlog_t		*log,
 	for (i = (int)(*head_blk) - 1; i >= 0; i--) {
 		if ((error = xlog_bread(log, i, 1, bp)))
 			goto bread_err;
+		offset = xlog_align(log, i, 1, bp);
 		if (XLOG_HEADER_MAGIC_NUM ==
-		    INT_GET(*(uint *)(XFS_BUF_PTR(bp)), ARCH_CONVERT)) {
+		    INT_GET(*(uint *)offset, ARCH_CONVERT)) {
 			found = 1;
 			break;
 		}
@@ -811,8 +858,9 @@ xlog_find_tail(xlog_t		*log,
 		for (i = log->l_logBBsize - 1; i >= (int)(*head_blk); i--) {
 			if ((error = xlog_bread(log, i, 1, bp)))
 				goto bread_err;
+			offset = xlog_align(log, i, 1, bp);
 			if (XLOG_HEADER_MAGIC_NUM ==
-			    INT_GET(*(uint*)(XFS_BUF_PTR(bp)), ARCH_CONVERT)) {
+			    INT_GET(*(uint*)offset, ARCH_CONVERT)) {
 				found = 2;
 				break;
 			}
@@ -825,7 +873,7 @@ xlog_find_tail(xlog_t		*log,
 	}
 
 	/* find blk_no of tail of log */
-	rhead = (xlog_rec_header_t *)XFS_BUF_PTR(bp);
+	rhead = (xlog_rec_header_t *)offset;
 	*tail_blk = BLOCK_LSN(rhead->h_tail_lsn, ARCH_CONVERT);
 
 	/*
@@ -885,7 +933,8 @@ xlog_find_tail(xlog_t		*log,
 		if ((error = xlog_bread(log, umount_data_blk, 1, bp))) {
 			goto bread_err;
 		}
-		op_head = (xlog_op_header_t *)XFS_BUF_PTR(bp);
+		offset = xlog_align(log, umount_data_blk, 1, bp);
+		op_head = (xlog_op_header_t *)offset;
 		if (op_head->oh_flags & XLOG_UNMOUNT_TRANS) {
 			/*
 			 * Set tail and last sync so that newly written
@@ -900,7 +949,6 @@ xlog_find_tail(xlog_t		*log,
 		}
 	}
 
-#ifdef __KERNEL__
 	/*
 	 * Make sure that there are no blocks in front of the head
 	 * with the same cycle number as the head.  This can happen
@@ -920,11 +968,9 @@ xlog_find_tail(xlog_t		*log,
 	 * But... if the -device- itself is readonly, just skip this.
 	 * We can't recover this device anyway, so it won't matter.
 	 */
-
-	if (!bdev_read_only(log->l_mp->m_logdev_targp->pbr_bdev)) {
+	if (!xfs_readonly_buftarg(log->l_mp->m_logdev_targp)) {
 		error = xlog_clear_stale_blocks(log, tail_lsn);
 	}
-#endif
 
 bread_err:
 exit:
@@ -932,10 +978,8 @@ xlog_find_tail(xlog_t		*log,
 
 	if (error)
 		xlog_warn("XFS: failed to locate log tail");
-
 	return error;
-}	/* xlog_find_tail */
-
+}
 
 /*
  * Is the log zeroed at all?
@@ -954,22 +998,25 @@ xlog_find_tail(xlog_t		*log,
  *	>0 => error has occurred
  */
 int
-xlog_find_zeroed(struct log	*log,
+xlog_find_zeroed(
+	xlog_t		*log,
 	xfs_daddr_t	*blk_no)
 {
 	xfs_buf_t	*bp;
+	xfs_caddr_t	offset;
 	uint	        first_cycle, last_cycle;
 	xfs_daddr_t	new_blk, last_blk, start_blk;
 	xfs_daddr_t     num_scan_bblks;
 	int	        error, log_bbnum = log->l_logBBsize;
 
 	/* check totally zeroed log */
-	bp = xlog_get_bp(1,log->l_mp);
+	bp = xlog_get_bp(log, 1);
 	if (!bp)
 		return ENOMEM;
 	if ((error = xlog_bread(log, 0, 1, bp)))
 		goto bp_err;
-	first_cycle = GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT);
+	offset = xlog_align(log, 0, 1, bp);
+	first_cycle = GET_CYCLE(offset, ARCH_CONVERT);
 	if (first_cycle == 0) {		/* completely zeroed log */
 		*blk_no = 0;
 		xlog_put_bp(bp);
@@ -979,7 +1026,8 @@ xlog_find_zeroed(struct log	*log,
 	/* check partially zeroed log */
 	if ((error = xlog_bread(log, log_bbnum-1, 1, bp)))
 		goto bp_err;
-	last_cycle = GET_CYCLE(XFS_BUF_PTR(bp), ARCH_CONVERT);
+	offset = xlog_align(log, log_bbnum-1, 1, bp);
+	last_cycle = GET_CYCLE(offset, ARCH_CONVERT);
 	if (last_cycle != 0) {		/* log completely written to */
 		xlog_put_bp(bp);
 		return 0;
@@ -1040,13 +1088,35 @@ xlog_find_zeroed(struct log	*log,
 	if (error)
 		return error;
 	return -1;
-}	/* xlog_find_zeroed */
+}
 
 /*
- * This is simply a subroutine used by xlog_clear_stale_blocks() below
+ * These are simple subroutines used by xlog_clear_stale_blocks() below
  * to initialize a buffer full of empty log record headers and write
  * them into the log.
  */
+STATIC void
+xlog_add_record(
+	xlog_t			*log,
+	xfs_caddr_t		buf,
+	int			cycle,
+	int			block,
+	int			tail_cycle,
+	int			tail_block)
+{
+	xlog_rec_header_t	*recp = (xlog_rec_header_t *)buf;
+
+	memset(buf, 0, BBSIZE);
+	INT_SET(recp->h_magicno, ARCH_CONVERT, XLOG_HEADER_MAGIC_NUM);
+	INT_SET(recp->h_cycle, ARCH_CONVERT, cycle);
+	INT_SET(recp->h_version, ARCH_CONVERT,
+			XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? 2 : 1);
+	ASSIGN_ANY_LSN(recp->h_lsn, cycle, block, ARCH_CONVERT);
+	ASSIGN_ANY_LSN(recp->h_tail_lsn, tail_cycle, tail_block, ARCH_CONVERT);
+	INT_SET(recp->h_fmt, ARCH_CONVERT, XLOG_FMT);
+	memcpy(&recp->h_fs_uuid, &log->l_mp->m_sb.sb_uuid, sizeof(uuid_t));
+}
+
 STATIC int
 xlog_write_log_records(
 	xlog_t		*log,
@@ -1056,51 +1126,68 @@ xlog_write_log_records(
 	int		tail_cycle,
 	int		tail_block)
 {
-	xlog_rec_header_t	*recp;
-	int			i, j;
-	int			end_block = start_block + blocks;
-	int			error       = 0;
+	xfs_caddr_t	offset;
 	xfs_buf_t	*bp;
-	char		        *buf;
+	int		balign, ealign;
+	int		sectbb = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, 1);
+	int		end_block = start_block + blocks;
 	int		bufblks;
+	int		error = 0;
+	int		i, j = 0;
 
 	bufblks = 1 << ffs(blocks);
-	while (!(bp = xlog_get_bp(bufblks, log->l_mp))) {
+	while (!(bp = xlog_get_bp(log, bufblks))) {
 		bufblks >>= 1;
-		if (!bufblks)
+		if (bufblks <= log->l_sectbb_log)
 			return ENOMEM;
 	}
 
-	buf = XFS_BUF_PTR(bp);
-	recp = (xlog_rec_header_t*)buf;
-
-	memset(buf, 0, BBSIZE);
-	INT_SET(recp->h_magicno, ARCH_CONVERT, XLOG_HEADER_MAGIC_NUM);
-	INT_SET(recp->h_cycle, ARCH_CONVERT, cycle);
-	INT_SET(recp->h_version, ARCH_CONVERT,
-			XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? 2 : 1);
-	ASSIGN_ANY_LSN(recp->h_tail_lsn, tail_cycle, tail_block, ARCH_CONVERT);
+	/* We may need to do a read at the start to fill in part of
+	 * the buffer in the starting sector not covered by the first
+	 * write below.
+	 */
+	balign = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, start_block);
+	if (balign != start_block) {
+		if ((error = xlog_bread(log, start_block, 1, bp))) {
+			xlog_put_bp(bp);
+			return error;
+		}
+		j = start_block - balign;
+	}
 
 	for (i = start_block; i < end_block; i += bufblks) {
-		int bcount = min(bufblks, end_block - start_block);
-		/* with plenty of memory, we duplicate the block
-		 * right through the buffer and modify each entry
+		int		bcount, endcount;
+
+		bcount = min(bufblks, end_block - start_block);
+		endcount = bcount - j;
+
+		/* We may need to do a read at the end to fill in part of
+		 * the buffer in the final sector not covered by the write.
+		 * If this is the same sector as the above read, skip it.
 		 */
-		ASSIGN_ANY_LSN(recp->h_lsn, cycle, i, ARCH_CONVERT);
-		for (j = 1; j < bcount; j++) {
-			buf += BBSIZE;
-			recp = (xlog_rec_header_t*)buf;
-			memcpy(buf, XFS_BUF_PTR(bp), BBSIZE);
-			ASSIGN_ANY_LSN(recp->h_lsn, cycle, i+j, ARCH_CONVERT);
+		ealign = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, end_block);
+		if (j == 0 && (start_block + endcount > ealign)) {
+			offset = XFS_BUF_PTR(bp);
+			balign = BBTOB(ealign - start_block);
+			XFS_BUF_SET_PTR(bp, offset + balign, BBTOB(sectbb));
+			if ((error = xlog_bread(log, ealign, sectbb, bp)))
+				break;
+			XFS_BUF_SET_PTR(bp, offset, bufblks);
 		}
-		/* then write the whole lot out at once */
-		error = xlog_bwrite(log, start_block, bcount, bp);
-		start_block += bcount;
-		buf = XFS_BUF_PTR(bp);
-		recp = (xlog_rec_header_t*)buf;
+
+		offset = xlog_align(log, start_block, endcount, bp);
+		for (; j < endcount; j++) {
+			xlog_add_record(log, offset, cycle, i+j,
+					tail_cycle, tail_block);
+			offset += BBSIZE;
+		}
+		error = xlog_bwrite(log, start_block, endcount, bp);
+		if (error)
+			break;
+		start_block += endcount;
+		j = 0;
 	}
 	xlog_put_bp(bp);
-
 	return error;
 }
 
@@ -1244,7 +1331,8 @@ xlog_clear_stale_blocks(
  */
 
 STATIC xlog_recover_t *
-xlog_recover_find_tid(xlog_recover_t *q,
+xlog_recover_find_tid(
+	xlog_recover_t		*q,
 	xlog_tid_t		tid)
 {
 	xlog_recover_t		*p = q;
@@ -1255,30 +1343,30 @@ xlog_recover_find_tid(xlog_recover_t *q,
 		p = p->r_next;
 	}
 	return p;
-}	/* xlog_recover_find_tid */
-
+}
 
 STATIC void
-xlog_recover_put_hashq(xlog_recover_t **q,
+xlog_recover_put_hashq(
+	xlog_recover_t		**q,
 	xlog_recover_t		*trans)
 {
 	trans->r_next = *q;
 	*q = trans;
-}	/* xlog_recover_put_hashq */
-
+}
 
 STATIC void
-xlog_recover_add_item(xlog_recover_item_t **itemq)
+xlog_recover_add_item(
+	xlog_recover_item_t	**itemq)
 {
 	xlog_recover_item_t	*item;
 
 	item = kmem_zalloc(sizeof(xlog_recover_item_t), 0);
 	xlog_recover_insert_item_backq(itemq, item);
-}	/* xlog_recover_add_item */
-
+}
 
 STATIC int
-xlog_recover_add_to_cont_trans(xlog_recover_t	*trans,
+xlog_recover_add_to_cont_trans(
+	xlog_recover_t		*trans,
 	xfs_caddr_t		dp,
 	int			len)
 {
@@ -1290,7 +1378,8 @@ xlog_recover_add_to_cont_trans(xlog_recover_t	*trans,
 	if (item == 0) {
 		/* finish copying rest of trans header */
 		xlog_recover_add_item(&trans->r_itemq);
-		ptr = (xfs_caddr_t)&trans->r_theader+sizeof(xfs_trans_header_t)-len;
+		ptr = (xfs_caddr_t) &trans->r_theader +
+				sizeof(xfs_trans_header_t) - len;
 		memcpy(ptr, dp, len); /* d, s, l */
 		return 0;
 	}
@@ -1304,10 +1393,10 @@ xlog_recover_add_to_cont_trans(xlog_recover_t	*trans,
 	item->ri_buf[item->ri_cnt-1].i_len += len;
 	item->ri_buf[item->ri_cnt-1].i_addr = ptr;
 	return 0;
-}	/* xlog_recover_add_to_cont_trans */
-
+}
 
-/* The next region to add is the start of a new region.  It could be
+/*
+ * The next region to add is the start of a new region.  It could be
  * a whole region or it could be the first part of a new region.  Because
  * of this, the assumption here is that the type and size fields of all
  * format structures fit into the first 32 bits of the structure.
@@ -1320,7 +1409,8 @@ xlog_recover_add_to_cont_trans(xlog_recover_t	*trans,
  * will appear in the current log item.
  */
 STATIC int
-xlog_recover_add_to_trans(xlog_recover_t	*trans,
+xlog_recover_add_to_trans(
+	xlog_recover_t		*trans,
 	xfs_caddr_t		dp,
 	int			len)
 {
@@ -1339,7 +1429,7 @@ xlog_recover_add_to_trans(xlog_recover_t	*trans,
 		return 0;
 	}
 
-	ptr = kmem_alloc(len, 0);
+	ptr = kmem_alloc(len, KM_SLEEP);
 	memcpy(ptr, dp, len);
 	in_f = (xfs_inode_log_format_t *)ptr;
 
@@ -1362,25 +1452,25 @@ xlog_recover_add_to_trans(xlog_recover_t	*trans,
 	item->ri_buf[item->ri_cnt].i_len  = len;
 	item->ri_cnt++;
 	return 0;
-}	/* xlog_recover_add_to_trans */
-
+}
 
 STATIC void
-xlog_recover_new_tid(xlog_recover_t	**q,
+xlog_recover_new_tid(
+	xlog_recover_t		**q,
 	xlog_tid_t		tid,
 	xfs_lsn_t		lsn)
 {
 	xlog_recover_t		*trans;
 
-	trans = kmem_zalloc(sizeof(xlog_recover_t), 0);
+	trans = kmem_zalloc(sizeof(xlog_recover_t), KM_SLEEP);
 	trans->r_log_tid   = tid;
 	trans->r_lsn	   = lsn;
 	xlog_recover_put_hashq(q, trans);
-}	/* xlog_recover_new_tid */
-
+}
 
 STATIC int
-xlog_recover_unlink_tid(xlog_recover_t	**q,
+xlog_recover_unlink_tid(
+	xlog_recover_t		**q,
 	xlog_recover_t		*trans)
 {
 	xlog_recover_t		*tp;
@@ -1407,10 +1497,11 @@ xlog_recover_unlink_tid(xlog_recover_t	**q,
 		tp->r_next = tp->r_next->r_next;
 	}
 	return 0;
-}	/* xlog_recover_unlink_tid */
+}
 
 STATIC void
-xlog_recover_insert_item_backq(xlog_recover_item_t **q,
+xlog_recover_insert_item_backq(
+	xlog_recover_item_t	**q,
 	xlog_recover_item_t	*item)
 {
 	if (*q == 0) {
@@ -1422,18 +1513,20 @@ xlog_recover_insert_item_backq(xlog_recover_item_t **q,
 		(*q)->ri_prev		= item;
 		item->ri_prev->ri_next	= item;
 	}
-}	/* xlog_recover_insert_item_backq */
+}
 
 STATIC void
-xlog_recover_insert_item_frontq(xlog_recover_item_t **q,
+xlog_recover_insert_item_frontq(
+	xlog_recover_item_t	**q,
 	xlog_recover_item_t	*item)
 {
 	xlog_recover_insert_item_backq(q, item);
 	*q = item;
-}	/* xlog_recover_insert_item_frontq */
+}
 
 STATIC int
-xlog_recover_reorder_trans(xlog_t	  *log,
+xlog_recover_reorder_trans(
+	xlog_t			*log,
 	xlog_recover_t		*trans)
 {
 	xlog_recover_item_t	*first_item, *itemq, *itemq_next;
@@ -1445,32 +1538,28 @@ xlog_recover_reorder_trans(xlog_t	  *log,
 		switch (ITEM_TYPE(itemq)) {
 		case XFS_LI_BUF:
 		case XFS_LI_6_1_BUF:
-	    case XFS_LI_5_3_BUF: {
+		case XFS_LI_5_3_BUF:
 			xlog_recover_insert_item_frontq(&trans->r_itemq, itemq);
 			break;
-	    }
 		case XFS_LI_INODE:
 		case XFS_LI_6_1_INODE:
 		case XFS_LI_5_3_INODE:
 		case XFS_LI_DQUOT:
 		case XFS_LI_QUOTAOFF:
 		case XFS_LI_EFD:
-	    case XFS_LI_EFI: {
+		case XFS_LI_EFI:
 			xlog_recover_insert_item_backq(&trans->r_itemq, itemq);
 			break;
-	    }
-	    default: {
+		default:
 			xlog_warn(
 	"XFS: xlog_recover_reorder_trans: unrecognized type of log operation");
 			ASSERT(0);
 			return XFS_ERROR(EIO);
 		}
-	}
 		itemq = itemq_next;
 	} while (first_item != itemq);
 	return 0;
-}	/* xlog_recover_reorder_trans */
-
+}
 
 /*
  * Build up the table of buf cancel records so that we don't replay
@@ -1485,7 +1574,8 @@ xlog_recover_reorder_trans(xlog_t	  *log,
  * record during the second pass.
  */
 STATIC void
-xlog_recover_do_buffer_pass1(xlog_t			*log,
+xlog_recover_do_buffer_pass1(
+	xlog_t			*log,
 	xfs_buf_log_format_t	*buf_f)
 {
 	xfs_buf_cancel_t	*bcp;
@@ -1493,9 +1583,9 @@ xlog_recover_do_buffer_pass1(xlog_t			*log,
 	xfs_buf_cancel_t	*prevp;
 	xfs_buf_cancel_t	**bucket;
 	xfs_buf_log_format_v1_t	*obuf_f;
-	xfs_daddr_t		blkno=0;
-	uint			len=0;
-	ushort			flags=0;
+	xfs_daddr_t		blkno = 0;
+	uint			len = 0;
+	ushort			flags = 0;
 
 	switch (buf_f->blf_type) {
 	case XFS_LI_BUF:
@@ -1515,9 +1605,8 @@ xlog_recover_do_buffer_pass1(xlog_t			*log,
 	/*
 	 * If this isn't a cancel buffer item, then just return.
 	 */
-	if (!(flags & XFS_BLI_CANCEL)) {
+	if (!(flags & XFS_BLI_CANCEL))
 		return;
-	}
 
 	/*
 	 * Insert an xfs_buf_cancel record into the hash table of
@@ -1531,7 +1620,7 @@ xlog_recover_do_buffer_pass1(xlog_t			*log,
 	 * the bucket.
 	 */
 	if (*bucket == NULL) {
-		bcp = (xfs_buf_cancel_t*)kmem_alloc(sizeof(xfs_buf_cancel_t),
+		bcp = (xfs_buf_cancel_t *)kmem_alloc(sizeof(xfs_buf_cancel_t),
 						     KM_SLEEP);
 		bcp->bc_blkno = blkno;
 		bcp->bc_len = len;
@@ -1557,7 +1646,7 @@ xlog_recover_do_buffer_pass1(xlog_t			*log,
 		nextp = nextp->bc_next;
 	}
 	ASSERT(prevp != NULL);
-	bcp = (xfs_buf_cancel_t*)kmem_alloc(sizeof(xfs_buf_cancel_t),
+	bcp = (xfs_buf_cancel_t *)kmem_alloc(sizeof(xfs_buf_cancel_t),
 					     KM_SLEEP);
 	bcp->bc_blkno = blkno;
 	bcp->bc_len = len;
@@ -1580,17 +1669,17 @@ xlog_recover_do_buffer_pass1(xlog_t			*log,
  * made at that point.
  */
 STATIC int
-xlog_recover_do_buffer_pass2(xlog_t			*log,
+xlog_recover_do_buffer_pass2(
+	xlog_t			*log,
 	xfs_buf_log_format_t	*buf_f)
 {
 	xfs_buf_cancel_t	*bcp;
 	xfs_buf_cancel_t	*prevp;
 	xfs_buf_cancel_t	**bucket;
 	xfs_buf_log_format_v1_t	*obuf_f;
-	xfs_daddr_t		blkno=0;
-	ushort			flags=0;
-	uint			len=0;
-
+	xfs_daddr_t		blkno = 0;
+	ushort			flags = 0;
+	uint			len = 0;
 
 	switch (buf_f->blf_type) {
 	case XFS_LI_BUF:
@@ -1667,7 +1756,6 @@ xlog_recover_do_buffer_pass2(xlog_t			*log,
 	return 0;
 }
 
-
 /*
  * Perform recovery for a buffer full of inodes.  In these buffers,
  * the only data which should be recovered is that which corresponds
@@ -1682,7 +1770,8 @@ xlog_recover_do_buffer_pass2(xlog_t			*log,
  * sent to xlog_recover_do_reg_buffer() below during recovery.
  */
 STATIC int
-xlog_recover_do_inode_buffer(xfs_mount_t		*mp,
+xlog_recover_do_inode_buffer(
+	xfs_mount_t		*mp,
 	xlog_recover_item_t	*item,
 	xfs_buf_t		*bp,
 	xfs_buf_log_format_t	*buf_f)
@@ -1698,8 +1787,8 @@ xlog_recover_do_inode_buffer(xfs_mount_t		*mp,
 	xfs_agino_t		*logged_nextp;
 	xfs_agino_t		*buffer_nextp;
 	xfs_buf_log_format_v1_t	*obuf_f;
-	unsigned int		*data_map=NULL;
-	unsigned int		map_size=0;
+	unsigned int		*data_map = NULL;
+	unsigned int		map_size = 0;
 
 	switch (buf_f->blf_type) {
 	case XFS_LI_BUF:
@@ -1790,7 +1879,7 @@ xlog_recover_do_inode_buffer(xfs_mount_t		*mp,
 	}
 
 	return 0;
-}	/* xlog_recover_do_inode_buffer */
+}
 
 /*
  * Perform a 'normal' buffer recovery.  Each logged region of the
@@ -1800,7 +1889,8 @@ xlog_recover_do_inode_buffer(xfs_mount_t		*mp,
  */
 /*ARGSUSED*/
 STATIC void
-xlog_recover_do_reg_buffer(xfs_mount_t		*mp,
+xlog_recover_do_reg_buffer(
+	xfs_mount_t		*mp,
 	xlog_recover_item_t	*item,
 	xfs_buf_t		*bp,
 	xfs_buf_log_format_t	*buf_f)
@@ -1809,8 +1899,8 @@ xlog_recover_do_reg_buffer(xfs_mount_t		*mp,
 	int			bit;
 	int			nbits;
 	xfs_buf_log_format_v1_t	*obuf_f;
-	unsigned int		*data_map=NULL;
-	unsigned int		map_size=0;
+	unsigned int		*data_map = NULL;
+	unsigned int		map_size = 0;
 	int                     error;
 
 	switch (buf_f->blf_type) {
@@ -1860,7 +1950,7 @@ xlog_recover_do_reg_buffer(xfs_mount_t		*mp,
 
 	/* Shouldn't be any more regions */
 	ASSERT(i == item->ri_total);
-}	/* xlog_recover_do_reg_buffer */
+}
 
 /*
  * Do some primitive error checking on ondisk dquot data structures.
@@ -2038,7 +2128,8 @@ xlog_recover_do_dquot_buffer(
  * for more details on the implementation of the table of cancel records.
  */
 STATIC int
-xlog_recover_do_buffer_trans(xlog_t		 *log,
+xlog_recover_do_buffer_trans(
+	xlog_t			*log,
 	xlog_recover_item_t	*item,
 	int			pass)
 {
@@ -2152,10 +2243,11 @@ xlog_recover_do_buffer_trans(xlog_t		 *log,
 	}
 
 	return (error);
-}	/* xlog_recover_do_buffer_trans */
+}
 
 STATIC int
-xlog_recover_do_inode_trans(xlog_t		*log,
+xlog_recover_do_inode_trans(
+	xlog_t			*log,
 	xlog_recover_item_t	*item,
 	int			pass)
 {
@@ -2377,7 +2469,6 @@ xlog_recover_do_inode_trans(xlog_t		*log,
 		}
 	}
 
-
 write_inode_buffer:
 	if (ITEM_TYPE(item) == XFS_LI_INODE) {
 		ASSERT(XFS_BUF_FSPRIVATE(bp, void *) == NULL ||
@@ -2391,8 +2482,7 @@ xlog_recover_do_inode_trans(xlog_t		*log,
 	}
 
 	return (error);
-}	/* xlog_recover_do_inode_trans */
-
+}
 
 /*
  * Recover QUOTAOFF records. We simply make a note of it in the xlog_t
@@ -2400,7 +2490,8 @@ xlog_recover_do_inode_trans(xlog_t		*log,
  * of that type.
  */
 STATIC int
-xlog_recover_do_quotaoff_trans(xlog_t			*log,
+xlog_recover_do_quotaoff_trans(
+	xlog_t			*log,
 	xlog_recover_item_t	*item,
 	int			pass)
 {
@@ -2425,12 +2516,12 @@ xlog_recover_do_quotaoff_trans(xlog_t			*log,
 	return (0);
 }
 
-
 /*
  * Recover a dquot record
  */
 STATIC int
-xlog_recover_do_dquot_trans(xlog_t		*log,
+xlog_recover_do_dquot_trans(
+	xlog_t			*log,
 	xlog_recover_item_t	*item,
 	int			pass)
 {
@@ -2516,7 +2607,7 @@ xlog_recover_do_dquot_trans(xlog_t		*log,
 	xfs_bdwrite(mp, bp);
 
 	return (0);
-}	/* xlog_recover_do_dquot_trans */
+}
 
 /*
  * This routine is called to create an in-core extent free intent
@@ -2526,7 +2617,8 @@ xlog_recover_do_dquot_trans(xlog_t		*log,
  * LSN.
  */
 STATIC void
-xlog_recover_do_efi_trans(xlog_t		*log,
+xlog_recover_do_efi_trans(
+	xlog_t			*log,
 	xlog_recover_item_t	*item,
 	xfs_lsn_t		lsn,
 	int			pass)
@@ -2558,7 +2650,7 @@ xlog_recover_do_efi_trans(xlog_t		*log,
 	 * xfs_trans_update_ail() drops the AIL lock.
 	 */
 	xfs_trans_update_ail(mp, (xfs_log_item_t *)efip, lsn, s);
-}	/* xlog_recover_do_efi_trans */
+}
 
 
 /*
@@ -2570,13 +2662,14 @@ xlog_recover_do_efi_trans(xlog_t		*log,
  * AIL and free it.
  */
 STATIC void
-xlog_recover_do_efd_trans(xlog_t		*log,
+xlog_recover_do_efd_trans(
+	xlog_t			*log,
 	xlog_recover_item_t	*item,
 	int			pass)
 {
 	xfs_mount_t		*mp;
 	xfs_efd_log_format_t	*efd_formatp;
-	xfs_efi_log_item_t	*efip=NULL;
+	xfs_efi_log_item_t	*efip = NULL;
 	xfs_log_item_t		*lip;
 	int			gen;
 	int			nexts;
@@ -2631,7 +2724,7 @@ xlog_recover_do_efd_trans(xlog_t		*log,
 			kmem_zone_free(xfs_efi_zone, efip);
 		}
 	}
-}	/* xlog_recover_do_efd_trans */
+}
 
 /*
  * Perform the transaction
@@ -2640,7 +2733,8 @@ xlog_recover_do_efd_trans(xlog_t		*log,
  * EFIs and EFDs get queued up by adding entries into the AIL for them.
  */
 STATIC int
-xlog_recover_do_trans(xlog_t	     *log,
+xlog_recover_do_trans(
+	xlog_t			*log,
 	xlog_recover_t		*trans,
 	int			pass)
 {
@@ -2695,8 +2789,7 @@ xlog_recover_do_trans(xlog_t	     *log,
 	} while (first_item != item);
 
 	return error;
-}	/* xlog_recover_do_trans */
-
+}
 
 /*
  * Free up any resources allocated by the transaction
@@ -2704,7 +2797,8 @@ xlog_recover_do_trans(xlog_t	     *log,
  * Remember that EFIs, EFDs, and IUNLINKs are handled later.
  */
 STATIC void
-xlog_recover_free_trans(xlog_recover_t      *trans)
+xlog_recover_free_trans(
+	xlog_recover_t		*trans)
 {
 	xlog_recover_item_t	*first_item, *item, *free_item;
 	int			i;
@@ -2725,11 +2819,11 @@ xlog_recover_free_trans(xlog_recover_t      *trans)
 	} while (first_item != item);
 	/* Free the transaction recover structure */
 	kmem_free(trans, sizeof(xlog_recover_t));
-}	/* xlog_recover_free_trans */
-
+}
 
 STATIC int
-xlog_recover_commit_trans(xlog_t	 *log,
+xlog_recover_commit_trans(
+	xlog_t			*log,
 	xlog_recover_t		**q,
 	xlog_recover_t		*trans,
 	int			pass)
@@ -2742,18 +2836,16 @@ xlog_recover_commit_trans(xlog_t	 *log,
 		return error;
 	xlog_recover_free_trans(trans);			/* no error */
 	return 0;
-}	/* xlog_recover_commit_trans */
-
+}
 
-/*ARGSUSED*/
 STATIC int
-xlog_recover_unmount_trans(xlog_recover_t *trans)
+xlog_recover_unmount_trans(
+	xlog_recover_t		*trans)
 {
 	/* Do nothing now */
 	xlog_warn("XFS: xlog_recover_unmount_trans: Unmount LR");
-	return( 0 );
-}	/* xlog_recover_unmount_trans */
-
+	return 0;
+}
 
 /*
  * There are two valid states of the r_state field.  0 indicates that the
@@ -2765,14 +2857,15 @@ xlog_recover_unmount_trans(xlog_recover_t *trans)
  * NOTE: skip LRs with 0 data length.
  */
 STATIC int
-xlog_recover_process_data(xlog_t	    *log,
+xlog_recover_process_data(
+	xlog_t			*log,
 	xlog_recover_t		*rhash[],
 	xlog_rec_header_t	*rhead,
 	xfs_caddr_t		dp,
 	int			pass)
 {
-    xfs_caddr_t		lp	   = dp+INT_GET(rhead->h_len, ARCH_CONVERT);
-    int			num_logops = INT_GET(rhead->h_num_logops, ARCH_CONVERT);
+	xfs_caddr_t		lp;
+	int			num_logops;
 	xlog_op_header_t	*ohead;
 	xlog_recover_t		*trans;
 	xlog_tid_t		tid;
@@ -2780,6 +2873,9 @@ xlog_recover_process_data(xlog_t	    *log,
 	unsigned long		hash;
 	uint			flags;
 
+	lp = dp + INT_GET(rhead->h_len, ARCH_CONVERT);
+	num_logops = INT_GET(rhead->h_num_logops, ARCH_CONVERT);
+
 	/* check the log format matches our own - else we can't recover */
 	if (xlog_header_check_recover(log->l_mp, rhead))
 		return (XFS_ERROR(EIO));
@@ -2790,7 +2886,8 @@ xlog_recover_process_data(xlog_t	    *log,
 		dp += sizeof(xlog_op_header_t);
 		if (ohead->oh_clientid != XFS_TRANSACTION &&
 		    ohead->oh_clientid != XFS_LOG) {
-	    xlog_warn("XFS: xlog_recover_process_data: bad clientid");
+			xlog_warn(
+		"XFS: xlog_recover_process_data: bad clientid");
 			ASSERT(0);
 			return (XFS_ERROR(EIO));
 		}
@@ -2799,62 +2896,61 @@ xlog_recover_process_data(xlog_t	    *log,
 		trans = xlog_recover_find_tid(rhash[hash], tid);
 		if (trans == NULL) {		   /* not found; add new tid */
 			if (ohead->oh_flags & XLOG_START_TRANS)
-		xlog_recover_new_tid(&rhash[hash], tid, INT_GET(rhead->h_lsn, ARCH_CONVERT));
+				xlog_recover_new_tid(&rhash[hash], tid,
+					INT_GET(rhead->h_lsn, ARCH_CONVERT));
 		} else {
 			ASSERT(dp+INT_GET(ohead->oh_len, ARCH_CONVERT) <= lp);
 			flags = ohead->oh_flags & ~XLOG_END_TRANS;
 			if (flags & XLOG_WAS_CONT_TRANS)
 				flags &= ~XLOG_CONTINUE_TRANS;
 			switch (flags) {
-		case XLOG_COMMIT_TRANS: {
-		    error = xlog_recover_commit_trans(log, &rhash[hash],
-						      trans, pass);
+			case XLOG_COMMIT_TRANS:
+				error = xlog_recover_commit_trans(log,
+						&rhash[hash], trans, pass);
 				break;
-		}
-		case XLOG_UNMOUNT_TRANS: {
+			case XLOG_UNMOUNT_TRANS:
 				error = xlog_recover_unmount_trans(trans);
 				break;
-		}
-		case XLOG_WAS_CONT_TRANS: {
-		    error = xlog_recover_add_to_cont_trans(trans, dp,
-				  INT_GET(ohead->oh_len, ARCH_CONVERT));
+			case XLOG_WAS_CONT_TRANS:
+				error = xlog_recover_add_to_cont_trans(trans,
+						dp, INT_GET(ohead->oh_len,
+							ARCH_CONVERT));
 				break;
-		}
-		case XLOG_START_TRANS : {
-		    xlog_warn("XFS: xlog_recover_process_data: bad transaction");
+			case XLOG_START_TRANS:
+				xlog_warn(
+			"XFS: xlog_recover_process_data: bad transaction");
 				ASSERT(0);
 				error = XFS_ERROR(EIO);
 				break;
-		}
 			case 0:
-		case XLOG_CONTINUE_TRANS: {
-		    error = xlog_recover_add_to_trans(trans, dp,
-				   INT_GET(ohead->oh_len, ARCH_CONVERT));
+			case XLOG_CONTINUE_TRANS:
+				error = xlog_recover_add_to_trans(trans,
+						dp, INT_GET(ohead->oh_len,
+							ARCH_CONVERT));
 				break;
-		}
-		default: {
-		    xlog_warn("XFS: xlog_recover_process_data: bad flag");
+			default:
+				xlog_warn(
+			"XFS: xlog_recover_process_data: bad flag");
 				ASSERT(0);
 				error = XFS_ERROR(EIO);
 				break;
 			}
-	    } /* switch */
 			if (error)
 				return error;
-	} /* if */
+		}
 		dp += INT_GET(ohead->oh_len, ARCH_CONVERT);
 		num_logops--;
 	}
-    return( 0 );
-}	/* xlog_recover_process_data */
-
+	return 0;
+}
 
 /*
  * Process an extent free intent item that was recovered from
  * the log.  We need to free the extents that it describes.
  */
 STATIC void
-xlog_recover_process_efi(xfs_mount_t		*mp,
+xlog_recover_process_efi(
+	xfs_mount_t		*mp,
 	xfs_efi_log_item_t	*efip)
 {
 	xfs_efd_log_item_t	*efdp;
@@ -2900,8 +2996,7 @@ xlog_recover_process_efi(xfs_mount_t		*mp,
 
 	efip->efi_flags |= XFS_EFI_RECOVERED;
 	xfs_trans_commit(tp, 0, NULL);
-}	/* xlog_recover_process_efi */
-
+}
 
 /*
  * Verify that once we've encountered something other than an EFI
@@ -2909,13 +3004,13 @@ xlog_recover_process_efi(xfs_mount_t		*mp,
  */
 #if defined(DEBUG)
 STATIC void
-xlog_recover_check_ail(xfs_mount_t	*mp,
+xlog_recover_check_ail(
+	xfs_mount_t		*mp,
 	xfs_log_item_t		*lip,
 	int			gen)
 {
-	int	orig_gen;
+	int			orig_gen = gen;
 
-	orig_gen = gen;
 	do {
 		ASSERT(lip->li_type != XFS_LI_EFI);
 		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
@@ -2930,7 +3025,6 @@ xlog_recover_check_ail(xfs_mount_t	*mp,
 }
 #endif	/* DEBUG */
 
-
 /*
  * When this is called, all of the EFIs which did not have
  * corresponding EFDs should be in the AIL.  What we do now
@@ -2950,7 +3044,8 @@ xlog_recover_check_ail(xfs_mount_t	*mp,
  * we see something other than an EFI in the AIL.
  */
 STATIC void
-xlog_recover_process_efis(xlog_t	*log)
+xlog_recover_process_efis(
+	xlog_t			*log)
 {
 	xfs_log_item_t		*lip;
 	xfs_efi_log_item_t	*efip;
@@ -2986,8 +3081,7 @@ xlog_recover_process_efis(xlog_t	*log)
 		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
 	}
 	AIL_UNLOCK(mp, s);
-}	/* xlog_recover_process_efis */
-
+}
 
 /*
  * This routine performs a transaction to null out a bad inode pointer
@@ -3030,8 +3124,7 @@ xlog_recover_clear_agi_bucket(
 			  (offset + sizeof(xfs_agino_t) - 1));
 
 	(void) xfs_trans_commit(tp, 0, NULL);
-}	/* xlog_recover_clear_agi_bucket */
-
+}
 
 /*
  * xlog_iunlink_recover
@@ -3046,7 +3139,8 @@ xlog_recover_clear_agi_bucket(
  * atomic.
  */
 void
-xlog_recover_process_iunlinks(xlog_t	*log)
+xlog_recover_process_iunlinks(
+	xlog_t		*log)
 {
 	xfs_mount_t	*mp;
 	xfs_agnumber_t	agno;
@@ -3188,40 +3282,47 @@ xlog_recover_process_iunlinks(xlog_t	*log)
 	}
 
 	mp->m_dmevmask = mp_dmevmask;
+}
 
-}	/* xlog_recover_process_iunlinks */
-
-
-/*
- * Stamp cycle number in every block
- *
- * This routine is also called in xfs_log.c
- */
-/*ARGSUSED*/
-void
-xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog)
-{
-	int	i, j, k;
-	int	size = iclog->ic_offset + iclog->ic_roundoff;
-	xfs_caddr_t dp;
-	union ich {
-		xlog_rec_ext_header_t	hic_xheader;
-		char			hic_sector[XLOG_HEADER_SIZE];
-	} *xhdr;
-	uint	cycle_lsn;
 
 #ifdef DEBUG
+STATIC void
+xlog_pack_data_checksum(
+	xlog_t		*log,
+	xlog_in_core_t	*iclog,
+	int		size)
+{
+	int		i;
 	uint		*up;
 	uint		chksum = 0;
 
 	up = (uint *)iclog->ic_datap;
 	/* divide length by 4 to get # words */
-	for (i=0; i<size >> 2; i++) {
+	for (i = 0; i < (size >> 2); i++) {
 		chksum ^= INT_GET(*up, ARCH_CONVERT);
 		up++;
 	}
 	INT_SET(iclog->ic_header.h_chksum, ARCH_CONVERT, chksum);
-#endif /* DEBUG */
+}
+#else
+#define xlog_pack_data_checksum(log, iclog, size)
+#endif
+
+/*
+ * Stamp cycle number in every block
+ */
+void
+xlog_pack_data(
+	xlog_t			*log,
+	xlog_in_core_t		*iclog)
+{
+	int			i, j, k;
+	int			size = iclog->ic_offset + iclog->ic_roundoff;
+	uint			cycle_lsn;
+	xfs_caddr_t		dp;
+	xlog_in_core_2_t	*xhdr;
+
+	xlog_pack_data_checksum(log, iclog, size);
 
 	cycle_lsn = CYCLE_LSN_NOCONV(iclog->ic_header.h_lsn, ARCH_CONVERT);
 
@@ -3234,7 +3335,7 @@ xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog)
 	}
 
 	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
-		xhdr = (union ich*)&iclog->ic_header;
+		xhdr = (xlog_in_core_2_t *)&iclog->ic_header;
 		for ( ; i < BTOBB(size); i++) {
 			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
 			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
@@ -3247,45 +3348,18 @@ xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog)
 			xhdr[i].hic_xheader.xh_cycle = cycle_lsn;
 		}
 	}
+}
 
-}	/* xlog_pack_data */
-
-
-/*ARGSUSED*/
+#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
 STATIC void
-xlog_unpack_data(xlog_rec_header_t *rhead,
+xlog_unpack_data_checksum(
+	xlog_rec_header_t	*rhead,
 	xfs_caddr_t		dp,
 	xlog_t			*log)
 {
-	int i, j, k;
-	union ich {
-		xlog_rec_header_t	hic_header;
-		xlog_rec_ext_header_t	hic_xheader;
-		char			hic_sector[XLOG_HEADER_SIZE];
-	} *xhdr;
-
-#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
 	uint			*up = (uint *)dp;
 	uint			chksum = 0;
-#endif
-
-	for (i=0; i < BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT)) &&
-		  i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
-		*(uint *)dp = *(uint *)&rhead->h_cycle_data[i];
-		dp += BBSIZE;
-	}
 
-	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
-		xhdr = (union ich*)rhead;
-		for ( ; i < BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT)); i++) {
-			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
-			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
-			*(uint *)dp = xhdr[j].hic_xheader.xh_cycle_data[k];
-			dp += BBSIZE;
-		}
-	}
-
-#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
 	/* divide length by 4 to get # words */
 	for (i=0; i < INT_GET(rhead->h_len, ARCH_CONVERT) >> 2; i++) {
 		chksum ^= INT_GET(*up, ARCH_CONVERT);
@@ -3306,9 +3380,38 @@ xlog_unpack_data(xlog_rec_header_t *rhead,
 		    log->l_flags |= XLOG_CHKSUM_MISMATCH;
 	    }
 	}
-#endif /* DEBUG && XFS_LOUD_RECOVERY */
-}	/* xlog_unpack_data */
+}
+#else
+#define xlog_unpack_data_checksum(rhead, dp, log)
+#endif
+
+STATIC void
+xlog_unpack_data(
+	xlog_rec_header_t	*rhead,
+	xfs_caddr_t		dp,
+	xlog_t			*log)
+{
+	int			i, j, k;
+	xlog_in_core_2_t	*xhdr;
 
+	for (i = 0; i < BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT)) &&
+		  i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
+		*(uint *)dp = *(uint *)&rhead->h_cycle_data[i];
+		dp += BBSIZE;
+	}
+
+	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
+		xhdr = (xlog_in_core_2_t *)rhead;
+		for ( ; i < BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT)); i++) {
+			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
+			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
+			*(uint *)dp = xhdr[j].hic_xheader.xh_cycle_data[k];
+			dp += BBSIZE;
+		}
+	}
+
+	xlog_unpack_data_checksum(rhead, dp, log);
+}
 
 /*
  * Read the log from tail to head and process the log records found.
@@ -3319,67 +3422,70 @@ xlog_unpack_data(xlog_rec_header_t *rhead,
  * here.
  */
 STATIC int
-xlog_do_recovery_pass(xlog_t	*log,
+xlog_do_recovery_pass(
+	xlog_t			*log,
 	xfs_daddr_t		head_blk,
 	xfs_daddr_t		tail_blk,
 	int			pass)
 {
 	xlog_rec_header_t	*rhead;
 	xfs_daddr_t		blk_no;
-    xfs_caddr_t		bufaddr;
+	xfs_caddr_t		bufaddr, offset;
 	xfs_buf_t		*hbp, *dbp;
-    int			error, h_size;
+	int			error = 0, h_size;
 	int			bblks, split_bblks;
 	int			hblks, split_hblks, wrapped_hblks;
 	xlog_recover_t		*rhash[XLOG_RHASH_SIZE];
 
-    error = 0;
-
-
 	/*
 	 * Read the header of the tail block and get the iclog buffer size from
 	 * h_size.  Use this to tell how many sectors make up the log header.
 	 */
 	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
 		/*
-	 * When using variable length iclogs, read first sector of iclog
-	 * header and extract the header size from it.  Get a new hbp that
-	 * is the correct size.
+		 * When using variable length iclogs, read first sector of
+		 * iclog header and extract the header size from it.  Get a
+		 * new hbp that is the correct size.
 		 */
-	hbp = xlog_get_bp(1, log->l_mp);
+		hbp = xlog_get_bp(log, 1);
 		if (!hbp)
 			return ENOMEM;
 		if ((error = xlog_bread(log, tail_blk, 1, hbp)))
 			goto bread_err1;
-	rhead = (xlog_rec_header_t *)XFS_BUF_PTR(hbp);
+		offset = xlog_align(log, tail_blk, 1, hbp);
+		rhead = (xlog_rec_header_t *)offset;
 		ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) ==
 						XLOG_HEADER_MAGIC_NUM);
-	if ((INT_GET(rhead->h_version, ARCH_CONVERT) & (~XLOG_VERSION_OKBITS)) != 0) {
-	    xlog_warn("XFS: xlog_do_recovery_pass: unrecognised log version number.");
+		if ((INT_GET(rhead->h_version, ARCH_CONVERT) &
+				(~XLOG_VERSION_OKBITS)) != 0) {
+			xlog_warn(
+	"XFS: xlog_do_recovery_pass: unrecognised log version number.");
 			error = XFS_ERROR(EIO);
 			goto bread_err1;
 		}
 		h_size = INT_GET(rhead->h_size, ARCH_CONVERT);
 
-	if ((INT_GET(rhead->h_version, ARCH_CONVERT) & XLOG_VERSION_2) &&
+		if ((INT_GET(rhead->h_version, ARCH_CONVERT)
+				& XLOG_VERSION_2) &&
 		    (h_size > XLOG_HEADER_CYCLE_SIZE)) {
 			hblks = h_size / XLOG_HEADER_CYCLE_SIZE;
 			if (h_size % XLOG_HEADER_CYCLE_SIZE)
 				hblks++;
 			xlog_put_bp(hbp);
-	    hbp = xlog_get_bp(hblks, log->l_mp);
+			hbp = xlog_get_bp(log, hblks);
 		} else {
-	    hblks=1;
+			hblks = 1;
 		}
 	} else {
-	hblks=1;
-	hbp = xlog_get_bp(1, log->l_mp);
+		ASSERT(log->l_sectbb_log == 0);
+		hblks = 1;
+		hbp = xlog_get_bp(log, 1);
 		h_size = XLOG_BIG_RECORD_BSIZE;
 	}
 
 	if (!hbp)
 		return ENOMEM;
-    dbp = xlog_get_bp(BTOBB(h_size),log->l_mp);
+	dbp = xlog_get_bp(log, BTOBB(h_size));
 	if (!dbp) {
 		xlog_put_bp(hbp);
 		return ENOMEM;
@@ -3390,13 +3496,20 @@ xlog_do_recovery_pass(xlog_t	*log,
 		for (blk_no = tail_blk; blk_no < head_blk; ) {
 			if ((error = xlog_bread(log, blk_no, hblks, hbp)))
 				goto bread_err2;
-	    rhead = (xlog_rec_header_t *)XFS_BUF_PTR(hbp);
-	    ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
-	    ASSERT(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) <= INT_MAX));
-	    bblks = (int) BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));	/* blocks in data section */
-
-	    if (unlikely((INT_GET(rhead->h_magicno, ARCH_CONVERT) != XLOG_HEADER_MAGIC_NUM) ||
-		(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) > INT_MAX)) ||
+			offset = xlog_align(log, blk_no, hblks, hbp);
+			rhead = (xlog_rec_header_t *)offset;
+			ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) ==
+				XLOG_HEADER_MAGIC_NUM);
+			ASSERT(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) <=
+				INT_MAX));
+			/* blocks in data section */
+			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
+
+			if (unlikely(
+			    (INT_GET(rhead->h_magicno, ARCH_CONVERT) !=
+					XLOG_HEADER_MAGIC_NUM) ||
+			    (BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) >
+					INT_MAX)) ||
 			    (bblks <= 0) ||
 			    (blk_no > log->l_logBBsize))) {
 				XFS_ERROR_REPORT("xlog_do_recovery_pass(1)",
@@ -3405,28 +3518,33 @@ xlog_do_recovery_pass(xlog_t	*log,
 				goto bread_err2;
 			}
 
-	    if ((INT_GET(rhead->h_version, ARCH_CONVERT) & (~XLOG_VERSION_OKBITS)) != 0) {
-		xlog_warn("XFS: xlog_do_recovery_pass: unrecognised log version number.");
+			if ((INT_GET(rhead->h_version, ARCH_CONVERT) &
+					(~XLOG_VERSION_OKBITS)) != 0) {
+				xlog_warn(
+		"XFS: xlog_do_recovery_pass: unrecognised log version number.");
 				error = XFS_ERROR(EIO);
 				goto bread_err2;
 			}
-	    bblks = (int) BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));	/* blocks in data section */
+			/* blocks in data section */
+			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
 			if (bblks > 0) {
-		if ((error = xlog_bread(log, blk_no+hblks, bblks, dbp)))
+				if ((error = xlog_bread(log, blk_no + hblks,
+							bblks, dbp)))
 					goto bread_err2;
-		xlog_unpack_data(rhead, XFS_BUF_PTR(dbp), log);
-		if ((error = xlog_recover_process_data(log, rhash,
-						      rhead, XFS_BUF_PTR(dbp),
-						      pass)))
+				offset = xlog_align(log, blk_no + hblks,
+							bblks, dbp);
+				xlog_unpack_data(rhead, offset, log);
+				if ((error = xlog_recover_process_data(log,
+						rhash, rhead, offset, pass)))
 					goto bread_err2;
 			}
 			blk_no += (bblks+hblks);
 		}
 	} else {
 		/*
-	 * Perform recovery around the end of the physical log.  When the head
-	 * is not on the same cycle number as the tail, we can't do a sequential
-	 * recovery as above.
+		 * Perform recovery around the end of the physical log.
+		 * When the head is not on the same cycle number as the tail,
+		 * we can't do a sequential recovery as above.
 		 */
 		blk_no = tail_blk;
 		while (blk_no < log->l_logBBsize) {
@@ -3436,38 +3554,67 @@ xlog_do_recovery_pass(xlog_t	*log,
 			wrapped_hblks = 0;
 			if (blk_no+hblks <= log->l_logBBsize) {
 				/* Read header in one read */
-		if ((error = xlog_bread(log, blk_no, hblks, hbp)))
+				if ((error = xlog_bread(log, blk_no,
+							hblks, hbp)))
 					goto bread_err2;
+				offset = xlog_align(log, blk_no, hblks, hbp);
 			} else {
-		/* This log record is split across physical end of log */
+				/* This LR is split across physical log end */
+				offset = NULL;
 				split_hblks = 0;
 				if (blk_no != log->l_logBBsize) {
-		    /* some data is before physical end of log */
+					/* some data before physical log end */
 					ASSERT(blk_no <= INT_MAX);
 					split_hblks = log->l_logBBsize - (int)blk_no;
 					ASSERT(split_hblks > 0);
-		    if ((error = xlog_bread(log, blk_no, split_hblks, hbp)))
+					if ((error = xlog_bread(log, blk_no,
+							split_hblks, hbp)))
 						goto bread_err2;
+					offset = xlog_align(log, blk_no,
+							split_hblks, hbp);
 				}
+				/*
+				 * Note: this black magic still works with
+				 * large sector sizes (non-512) only because:
+				 * - we increased the buffer size originally
+				 *   by 1 sector giving us enough extra space
+				 *   for the second read;
+				 * - the log start is guaranteed to be sector
+				 *   aligned;
+				 * - we read the log end (LR header start)
+				 *   _first_, then the log start (LR header end)
+				 *   - order is important.
+				 */
 				bufaddr = XFS_BUF_PTR(hbp);
-		XFS_BUF_SET_PTR(hbp, bufaddr + BBTOB(split_hblks),
+				XFS_BUF_SET_PTR(hbp,
+						bufaddr + BBTOB(split_hblks),
 						BBTOB(hblks - split_hblks));
 				wrapped_hblks = hblks - split_hblks;
-		if ((error = xlog_bread(log, 0, wrapped_hblks, hbp)))
+				if ((error = xlog_bread(log, 0,
+							wrapped_hblks, hbp)))
 					goto bread_err2;
 				XFS_BUF_SET_PTR(hbp, bufaddr, hblks);
+				if (!offset)
+					offset = xlog_align(log, 0,
+							wrapped_hblks, hbp);
 			}
-	    rhead = (xlog_rec_header_t *)XFS_BUF_PTR(hbp);
-	    ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
-	    ASSERT(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) <= INT_MAX));
-	    bblks = (int) BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
+			rhead = (xlog_rec_header_t *)offset;
+			ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) ==
+				XLOG_HEADER_MAGIC_NUM);
+			ASSERT(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) <=
+				INT_MAX));
+			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
 
-	    /* LR body must have data or it wouldn't have been written */
+			/* LR body must have data or it wouldn't have been
+			 * written */
 			ASSERT(bblks > 0);
 			blk_no += hblks;	/* successfully read header */
 
-	    if (unlikely((INT_GET(rhead->h_magicno, ARCH_CONVERT) != XLOG_HEADER_MAGIC_NUM) ||
-		(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) > INT_MAX)) ||
+			if (unlikely(
+			    (INT_GET(rhead->h_magicno, ARCH_CONVERT) !=
+					XLOG_HEADER_MAGIC_NUM) ||
+			    (BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) >
+					INT_MAX)) ||
 			    (bblks <= 0))) {
 				XFS_ERROR_REPORT("xlog_do_recovery_pass(2)",
 					     XFS_ERRLEVEL_LOW, log->l_mp);
@@ -3477,32 +3624,57 @@ xlog_do_recovery_pass(xlog_t	*log,
 
 			/* Read in data for log record */
 			if (blk_no+bblks <= log->l_logBBsize) {
-		if ((error = xlog_bread(log, blk_no, bblks, dbp)))
+				if ((error = xlog_bread(log, blk_no,
+							bblks, dbp)))
 					goto bread_err2;
+				offset = xlog_align(log, blk_no, bblks, dbp);
 			} else {
-		/* This log record is split across physical end of log */
+				/* This log record is split across the
+				 * physical end of log */
+				offset = NULL;
 				split_bblks = 0;
 				if (blk_no != log->l_logBBsize) {
-
-		    /* some data is before physical end of log */
+					/* some data is before the physical
+					 * end of log */
+					ASSERT(!wrapped_hblks);
 					ASSERT(blk_no <= INT_MAX);
-		    split_bblks = log->l_logBBsize - (int)blk_no;
+					split_bblks =
+						log->l_logBBsize - (int)blk_no;
 					ASSERT(split_bblks > 0);
-		    if ((error = xlog_bread(log, blk_no, split_bblks, dbp)))
+					if ((error = xlog_bread(log, blk_no,
+							split_bblks, dbp)))
 						goto bread_err2;
+					offset = xlog_align(log, blk_no,
+							split_bblks, dbp);
 				}
+				/*
+				 * Note: this black magic still works with
+				 * large sector sizes (non-512) only because:
+				 * - we increased the buffer size originally
+				 *   by 1 sector giving us enough extra space
+				 *   for the second read;
+				 * - the log start is guaranteed to be sector
+				 *   aligned;
+				 * - we read the log end (LR header start)
+				 *   _first_, then the log start (LR header end)
+				 *   - order is important.
+				 */
 				bufaddr = XFS_BUF_PTR(dbp);
-		XFS_BUF_SET_PTR(dbp, bufaddr + BBTOB(split_bblks),
+				XFS_BUF_SET_PTR(dbp,
+						bufaddr + BBTOB(split_bblks),
 						BBTOB(bblks - split_bblks));
 				if ((error = xlog_bread(log, wrapped_hblks,
 						bblks - split_bblks, dbp)))
 					goto bread_err2;
-		XFS_BUF_SET_PTR(dbp, bufaddr, XLOG_BIG_RECORD_BSIZE);
+				XFS_BUF_SET_PTR(dbp, bufaddr,
+						XLOG_BIG_RECORD_BSIZE);
+				if (!offset)
+					offset = xlog_align(log, wrapped_hblks,
+						bblks - split_bblks, dbp);
 			}
-	    xlog_unpack_data(rhead, XFS_BUF_PTR(dbp), log);
+			xlog_unpack_data(rhead, offset, log);
 			if ((error = xlog_recover_process_data(log, rhash,
-						  rhead, XFS_BUF_PTR(dbp),
-						  pass)))
+						  rhead, offset, pass)))
 				goto bread_err2;
 			blk_no += bblks;
 		}
@@ -3514,27 +3686,29 @@ xlog_do_recovery_pass(xlog_t	*log,
 		while (blk_no < head_blk) {
 			if ((error = xlog_bread(log, blk_no, hblks, hbp)))
 				goto bread_err2;
-	    rhead = (xlog_rec_header_t *)XFS_BUF_PTR(hbp);
-	    ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
-	    ASSERT(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) <= INT_MAX));
-	    bblks = (int) BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
+			offset = xlog_align(log, blk_no, hblks, hbp);
+			rhead = (xlog_rec_header_t *)offset;
+			ASSERT(INT_GET(rhead->h_magicno, ARCH_CONVERT) ==
+				XLOG_HEADER_MAGIC_NUM);
+			ASSERT(BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT) <=
+				INT_MAX));
+			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
 			ASSERT(bblks > 0);
 			if ((error = xlog_bread(log, blk_no+hblks, bblks, dbp)))
 				goto bread_err2;
-	    xlog_unpack_data(rhead, XFS_BUF_PTR(dbp), log);
+			offset = xlog_align(log, blk_no+hblks, bblks, dbp);
+			xlog_unpack_data(rhead, offset, log);
 			if ((error = xlog_recover_process_data(log, rhash,
-						  rhead, XFS_BUF_PTR(dbp),
-						  pass)))
+							  rhead, offset, pass)))
 				goto bread_err2;
 			blk_no += (bblks+hblks);
 		}
 	}
 
-bread_err2:
+ bread_err2:
 	xlog_put_bp(dbp);
-bread_err1:
+ bread_err1:
 	xlog_put_bp(hbp);
-
 	return error;
 }
 
@@ -3552,7 +3726,8 @@ xlog_do_recovery_pass(xlog_t	*log,
  * the log recovery has been completed.
  */
 STATIC int
-xlog_do_log_recovery(xlog_t	*log,
+xlog_do_log_recovery(
+	xlog_t		*log,
 	xfs_daddr_t	head_blk,
 	xfs_daddr_t	tail_blk)
 {
@@ -3599,7 +3774,8 @@ xlog_do_log_recovery(xlog_t	*log,
  * Do the actual recovery
  */
 STATIC int
-xlog_do_recover(xlog_t	*log,
+xlog_do_recover(
+	xlog_t		*log,
 	xfs_daddr_t	head_blk,
 	xfs_daddr_t	tail_blk)
 {
@@ -3663,7 +3839,7 @@ xlog_do_recover(xlog_t	*log,
 	/* Normal transactions can now occur */
 	log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
 	return 0;
-}	/* xlog_do_recover */
+}
 
 /*
  * Perform recovery and re-initialize some log variables in xlog_find_tail.
@@ -3671,22 +3847,18 @@ xlog_do_recover(xlog_t	*log,
  * Return error or zero.
  */
 int
-xlog_recover(xlog_t *log, int readonly)
+xlog_recover(
+	xlog_t		*log,
+	int		readonly)
 {
 	xfs_daddr_t	head_blk, tail_blk;
 	int		error;
 
 	/* find the tail of the log */
-
 	if ((error = xlog_find_tail(log, &head_blk, &tail_blk, readonly)))
 		return error;
 
 	if (tail_blk != head_blk) {
-#ifndef __KERNEL__
-		extern xfs_daddr_t HEAD_BLK, TAIL_BLK;
-		head_blk = HEAD_BLK;
-		tail_blk = TAIL_BLK;
-#endif
 		/* There used to be a comment here:
 		 *
 		 * disallow recovery on read-only mounts.  note -- mount
@@ -3698,36 +3870,21 @@ xlog_recover(xlog_t *log, int readonly)
 		 * under the vfs layer, so we can get away with it unless
 		 * the device itself is read-only, in which case we fail.
 		 */
-#ifdef __KERNEL__
 		if ((error = xfs_dev_is_read_only(log->l_mp,
 						"recovery required"))) {
 			return error;
 		}
-#else
-		if (readonly) {
-			return ENOSPC;
-		}
-#endif
 
-#ifdef __KERNEL__
-#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
 		cmn_err(CE_NOTE,
 			"Starting XFS recovery on filesystem: %s (dev: %d/%d)",
 			log->l_mp->m_fsname, MAJOR(log->l_dev),
 			MINOR(log->l_dev));
-#else
-		cmn_err(CE_NOTE,
-			"!Starting XFS recovery on filesystem: %s (dev: %d/%d)",
-			log->l_mp->m_fsname, MAJOR(log->l_dev),
-			MINOR(log->l_dev));
-#endif
-#endif
+
 		error = xlog_do_recover(log, head_blk, tail_blk);
 		log->l_flags |= XLOG_RECOVERY_NEEDED;
 	}
 	return error;
-}	/* xlog_recover */
-
+}
 
 /*
  * In the first part of recovery we replay inodes and buffers and build
@@ -3739,7 +3896,9 @@ xlog_recover(xlog_t *log, int readonly)
  * in the real-time portion of the file system.
  */
 int
-xlog_recover_finish(xlog_t *log, int mfsi_flags)
+xlog_recover_finish(
+	xlog_t		*log,
+	int		mfsi_flags)
 {
 	/*
 	 * Now we're ready to do the transactions needed for the
@@ -3761,23 +3920,16 @@ xlog_recover_finish(xlog_t *log, int mfsi_flags)
 			      (XFS_LOG_FORCE | XFS_LOG_SYNC));
 
 		if ( (mfsi_flags & XFS_MFSI_NOUNLINK) == 0 ) {
-
 			xlog_recover_process_iunlinks(log);
 		}
 
 		xlog_recover_check_summary(log);
 
-#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
 		cmn_err(CE_NOTE,
 			"Ending XFS recovery on filesystem: %s (dev: %d/%d)",
 			log->l_mp->m_fsname, MAJOR(log->l_dev),
 			MINOR(log->l_dev));
-#else
-		cmn_err(CE_NOTE,
-			"!Ending XFS recovery on filesystem: %s (dev: %d/%d)",
-			log->l_mp->m_fsname, MAJOR(log->l_dev),
-			MINOR(log->l_dev));
-#endif
+
 		log->l_flags &= ~XLOG_RECOVERY_NEEDED;
 	} else {
 		cmn_err(CE_DEBUG,
@@ -3785,7 +3937,7 @@ xlog_recover_finish(xlog_t *log, int mfsi_flags)
 			log->l_mp->m_fsname);
 	}
 	return 0;
-}	/* xlog_recover_finish */
+}
 
 
 #if defined(DEBUG)
@@ -3794,7 +3946,8 @@ xlog_recover_finish(xlog_t *log, int mfsi_flags)
  * are consistent with the superblock counters.
  */
 void
-xlog_recover_check_summary(xlog_t	*log)
+xlog_recover_check_summary(
+	xlog_t		*log)
 {
 	xfs_mount_t	*mp;
 	xfs_agf_t	*agfp;

fs/xfs/xfs_mount.c

@@ -467,7 +467,11 @@ xfs_readsb(xfs_mount_t *mp)
 
 	bp = xfs_buf_read_flags(mp->m_ddev_targp, XFS_SB_DADDR,
 				BTOBB(sector_size), extra_flags);
-	ASSERT(bp);
+	if (!bp || XFS_BUF_ISERROR(bp)) {
+		cmn_err(CE_WARN, "XFS: SB read failed");
+		error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
+		goto fail;
+	}
 	ASSERT(XFS_BUF_ISBUSY(bp));
 	ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
 
@@ -482,9 +486,7 @@ xfs_readsb(xfs_mount_t *mp)
 	error = xfs_mount_validate_sb(mp, &(mp->m_sb));
 	if (error) {
 		cmn_err(CE_WARN, "XFS: SB validate failed");
-		XFS_BUF_UNMANAGE(bp);
-		xfs_buf_relse(bp);
-		return error;
+		goto fail;
 	}
 
 	/*
@@ -494,9 +496,8 @@ xfs_readsb(xfs_mount_t *mp)
 		cmn_err(CE_WARN,
 			"XFS: device supports only %u byte sectors (not %u)",
 			sector_size, mp->m_sb.sb_sectsize);
-		XFS_BUF_UNMANAGE(bp);
-		xfs_buf_relse(bp);
-		return XFS_ERROR(ENOSYS);
+		error = ENOSYS;
+		goto fail;
 	}
 
 	/*
@@ -509,7 +510,11 @@ xfs_readsb(xfs_mount_t *mp)
 		sector_size = mp->m_sb.sb_sectsize;
 		bp = xfs_buf_read_flags(mp->m_ddev_targp, XFS_SB_DADDR,
 					BTOBB(sector_size), extra_flags);
-		ASSERT(bp);
+		if (!bp || XFS_BUF_ISERROR(bp)) {
+			cmn_err(CE_WARN, "XFS: SB re-read failed");
+			error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
+			goto fail;
+		}
 		ASSERT(XFS_BUF_ISBUSY(bp));
 		ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
 	}
@@ -518,6 +523,13 @@ xfs_readsb(xfs_mount_t *mp)
 	xfs_buf_relse(bp);
 	ASSERT(XFS_BUF_VALUSEMA(bp) > 0);
 	return 0;
+
+ fail:
+	if (bp) {
+		XFS_BUF_UNMANAGE(bp);
+		xfs_buf_relse(bp);
+	}
+	return error;
 }
 
 
@@ -546,16 +558,7 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 	mp->m_blockmask = sbp->sb_blocksize - 1;
 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
 	mp->m_blockwmask = mp->m_blockwsize - 1;
-
-
-	if (XFS_SB_VERSION_HASLOGV2(sbp)) {
-		if (sbp->sb_logsunit <= 1) {
-			mp->m_lstripemask = 1;
-		} else {
-			mp->m_lstripemask =
-				1 << xfs_highbit32(sbp->sb_logsunit >> BBSHIFT);
-		}
-	}
+	INIT_LIST_HEAD(&mp->m_del_inodes);
 
 	/*
 	 * Setup for attributes, in case they get created.
@@ -601,7 +604,6 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 					sbp->sb_inopblock);
 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
 }
-
 /*
  * xfs_mountfs
  *

fs/xfs/xfs_mount.h

@@ -68,6 +68,7 @@ typedef struct xfs_trans_reservations {
 	((xfs_agblock_t)(XFS_BB_TO_FSBT(mp, d) % (mp)->m_sb.sb_agblocks))
 #else
 struct cred;
+struct log;
 struct vfs;
 struct vnode;
 struct xfs_mount_args;
@@ -296,13 +297,14 @@ typedef struct xfs_mount {
 	int			m_ihsize;	/* size of next field */
 	struct xfs_ihash	*m_ihash;	/* fs private inode hash table*/
 	struct xfs_inode	*m_inodes;	/* active inode list */
+	struct list_head	m_del_inodes;	/* inodes to reclaim */
 	mutex_t			m_ilock;	/* inode list mutex */
 	uint			m_ireclaims;	/* count of calls to reclaim*/
 	uint			m_readio_log;	/* min read size log bytes */
 	uint			m_readio_blocks; /* min read size blocks */
 	uint			m_writeio_log;	/* min write size log bytes */
 	uint			m_writeio_blocks; /* min write size blocks */
-	void			*m_log;		/* log specific stuff */
+	struct log		*m_log;		/* log specific stuff */
 	int			m_logbufs;	/* number of log buffers */
 	int			m_logbsize;	/* size of each log buffer */
 	uint			m_rsumlevels;	/* rt summary levels */
@@ -357,7 +359,6 @@ typedef struct xfs_mount {
 #endif
 	int			m_dalign;	/* stripe unit */
 	int			m_swidth;	/* stripe width */
-	int			m_lstripemask;	/* log stripe mask */
 	int			m_sinoalign;	/* stripe unit inode alignmnt */
 	int			m_attr_magicpct;/* 37% of the blocksize */
 	int			m_dir_magicpct;	/* 37% of the dir blocksize */

fs/xfs/xfs_vfsops.c

@@ -620,7 +620,7 @@ xfs_mntupdate(
 
 	if (*flags & MS_RDONLY) {
 		pagebuf_delwri_flush(mp->m_ddev_targp, 0, NULL);
-		xfs_finish_reclaim_all(mp);
+		xfs_finish_reclaim_all(mp, 0);
 
 		do {
 			VFS_SYNC(vfsp, SYNC_ATTR|SYNC_WAIT, NULL, error);
@@ -849,19 +849,14 @@ xfs_sync(
  * xfs sync routine for internal use
  *
  * This routine supports all of the flags defined for the generic VFS_SYNC
- * interface as explained above under xys_sync.  In the interests of not
+ * interface as explained above under xfs_sync.  In the interests of not
  * changing interfaces within the 6.5 family, additional internallly-
  * required functions are specified within a separate xflags parameter,
  * only available by calling this routine.
  *
- * xflags:
- *	XFS_XSYNC_RELOC - Sync for relocation.  Don't try to get behavior
- *                        locks as this will cause you to hang.  Not all
- *                        combinations of flags are necessarily supported
- *                        when this is specified.
  */
-int
-xfs_syncsub(
+STATIC int
+xfs_sync_inodes(
 	xfs_mount_t	*mp,
 	int		flags,
 	int             xflags,
@@ -877,12 +872,10 @@ xfs_syncsub(
 	uint64_t	fflag;
 	uint		lock_flags;
 	uint		base_lock_flags;
-	uint		log_flags;
 	boolean_t	mount_locked;
 	boolean_t	vnode_refed;
 	int		preempt;
 	xfs_dinode_t	*dip;
-	xfs_buf_log_item_t	*bip;
 	xfs_iptr_t	*ipointer;
 #ifdef DEBUG
 	boolean_t	ipointer_in = B_FALSE;
@@ -961,16 +954,6 @@ xfs_syncsub(
 		base_lock_flags |= XFS_IOLOCK_SHARED;
 	}
 
-	/*
-	 * Sync out the log.  This ensures that the log is periodically
-	 * flushed even if there is not enough activity to fill it up.
-	 */
-	if (flags & SYNC_WAIT) {
-		xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE | XFS_LOG_SYNC);
-	} else {
-		xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
-	}
-
 	XFS_MOUNT_ILOCK(mp);
 
 	ip = mp->m_inodes;
@@ -1016,9 +999,6 @@ xfs_syncsub(
 				ip = ip->i_mnext;
 				continue;
 			}
-			if ((ip->i_update_core == 0) &&
-			    ((ip->i_itemp == NULL) ||
-			    !(ip->i_itemp->ili_format.ilf_fields & XFS_ILOG_ALL))) {
 			if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
 				ip = ip->i_mnext;
 			} else if ((xfs_ipincount(ip) == 0) &&
@@ -1026,7 +1006,7 @@ xfs_syncsub(
 				IPOINTER_INSERT(ip, mp);
 
 				xfs_finish_reclaim(ip, 1,
-						XFS_IFLUSH_DELWRI_ELSE_SYNC);
+						XFS_IFLUSH_DELWRI_ELSE_ASYNC);
 
 				XFS_MOUNT_ILOCK(mp);
 				mount_locked = B_TRUE;
@@ -1037,7 +1017,6 @@ xfs_syncsub(
 			}
 			continue;
 		}
-		}
 
 		if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
 			XFS_MOUNT_IUNLOCK(mp);
@@ -1148,22 +1127,10 @@ xfs_syncsub(
 			xfs_iunlock(ip, XFS_ILOCK_SHARED);
 
 			if (XFS_FORCED_SHUTDOWN(mp)) {
-				if (xflags & XFS_XSYNC_RELOC) {
-					fs_tosspages(XFS_ITOBHV(ip), 0, -1,
-						     FI_REMAPF);
-				}
-				else {
 				VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
-				}
 			} else {
-				if (xflags & XFS_XSYNC_RELOC) {
-					fs_flushinval_pages(XFS_ITOBHV(ip),
-							    0, -1, FI_REMAPF);
-				}
-				else {
 				VOP_FLUSHINVAL_PAGES(vp, 0, -1, FI_REMAPF);
 			}
-			}
 
 			xfs_ilock(ip, XFS_ILOCK_SHARED);
 
@@ -1418,16 +1385,55 @@ xfs_syncsub(
 
 	ASSERT(ipointer_in == B_FALSE);
 
+	kmem_free(ipointer, sizeof(xfs_iptr_t));
+	return XFS_ERROR(last_error);
+}
+
+/*
+ * xfs sync routine for internal use
+ *
+ * This routine supports all of the flags defined for the generic VFS_SYNC
+ * interface as explained above under xfs_sync.  In the interests of not
+ * changing interfaces within the 6.5 family, additional internallly-
+ * required functions are specified within a separate xflags parameter,
+ * only available by calling this routine.
+ *
+ */
+int
+xfs_syncsub(
+	xfs_mount_t	*mp,
+	int		flags,
+	int             xflags,
+	int             *bypassed)
+{
+	int		error = 0;
+	int		last_error = 0;
+	uint		log_flags = XFS_LOG_FORCE;
+	xfs_buf_t	*bp;
+	xfs_buf_log_item_t	*bip;
+
 	/*
-	 * Flushing out dirty data above probably generated more
-	 * log activity, so if this isn't vfs_sync() then flush
-	 * the log again.  If SYNC_WAIT is set then do it synchronously.
+	 * Sync out the log.  This ensures that the log is periodically
+	 * flushed even if there is not enough activity to fill it up.
 	 */
-	if (!(flags & SYNC_BDFLUSH)) {
-		log_flags = XFS_LOG_FORCE;
-		if (flags & SYNC_WAIT) {
+	if (flags & SYNC_WAIT)
 		log_flags |= XFS_LOG_SYNC;
+
+	xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
+
+	if (flags & (SYNC_ATTR|SYNC_DELWRI)) {
+		if (flags & SYNC_BDFLUSH)
+			xfs_finish_reclaim_all(mp, 1);
+		else
+			error = xfs_sync_inodes(mp, flags, xflags, bypassed);
 	}
+
+	/*
+	 * Flushing out dirty data above probably generated more
+	 * log activity, so if this isn't vfs_sync() then flush
+	 * the log again.
+	 */
+	if (flags & SYNC_DELWRI) {
 		xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
 	}
 
@@ -1463,11 +1469,10 @@ xfs_syncsub(
 			 * that point so it can become pinned in between
 			 * there and here.
 			 */
-			if (XFS_BUF_ISPINNED(bp)) {
-				xfs_log_force(mp, (xfs_lsn_t)0,
-					      XFS_LOG_FORCE);
-			}
-			XFS_BUF_BFLAGS(bp) |= fflag;
+			if (XFS_BUF_ISPINNED(bp))
+				xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
+			if (!(flags & SYNC_WAIT))
+				XFS_BUF_BFLAGS(bp) |= XFS_B_ASYNC;
 			error = xfs_bwrite(mp, bp);
 		}
 		if (error) {
@@ -1478,9 +1483,9 @@ xfs_syncsub(
 	/*
 	 * Now check to see if the log needs a "dummy" transaction.
 	 */
-
 	if (xfs_log_need_covered(mp)) {
 		xfs_trans_t *tp;
+		xfs_inode_t *ip;
 
 		/*
 		 * Put a dummy transaction in the log to tell
@@ -1491,7 +1496,6 @@ xfs_syncsub(
 				XFS_ICHANGE_LOG_RES(mp),
 				0, 0, 0)))  {
 			xfs_trans_cancel(tp, 0);
-			kmem_free(ipointer, sizeof(xfs_iptr_t));
 			return error;
 		}
 
@@ -1503,6 +1507,7 @@ xfs_syncsub(
 		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 		error = xfs_trans_commit(tp, 0, NULL);
 		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+		xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
 	}
 
 	/*
@@ -1516,7 +1521,6 @@ xfs_syncsub(
 		}
 	}
 
-	kmem_free(ipointer, sizeof(xfs_iptr_t));
 	return XFS_ERROR(last_error);
 }
 

fs/xfs/xfs_vnodeops.c

@@ -658,7 +658,7 @@ xfs_setattr(
 		if (vap->va_size > ip->i_d.di_size) {
 			code = xfs_igrow_start(ip, vap->va_size, credp);
 			xfs_iunlock(ip, XFS_ILOCK_EXCL);
-		} else if (vap->va_size < ip->i_d.di_size) {
+		} else if (vap->va_size <= ip->i_d.di_size) {
 			xfs_iunlock(ip, XFS_ILOCK_EXCL);
 			xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE,
 					    (xfs_fsize_t)vap->va_size);
@@ -701,7 +701,7 @@ xfs_setattr(
 		if (vap->va_size > ip->i_d.di_size) {
 			xfs_igrow_finish(tp, ip, vap->va_size,
 			    !(flags & ATTR_DMI));
-		} else if ((vap->va_size < ip->i_d.di_size) ||
+		} else if ((vap->va_size <= ip->i_d.di_size) ||
 			   ((vap->va_size == 0) && ip->i_d.di_nextents)) {
 			/*
 			 * signal a sync transaction unless
@@ -3786,27 +3786,30 @@ xfs_inode_flush(
 					flush_flags = XFS_IFLUSH_SYNC;
 				else
 #endif
-					flush_flags = XFS_IFLUSH_DELWRI;
+				flush_flags = XFS_IFLUSH_DELWRI_ELSE_ASYNC;
 
 				xfs_ifunlock(ip);
 				xfs_iunlock(ip, XFS_ILOCK_SHARED);
 				error = xfs_itobp(mp, NULL, ip, &dip, &bp, 0);
 				if (error)
-					goto eagain;
+					return error;
 				xfs_buf_relse(bp);
 
 				if (xfs_ilock_nowait(ip, XFS_ILOCK_SHARED) == 0)
-					goto eagain;
+					return EAGAIN;
+
+				if (xfs_ipincount(ip) ||
+				    !xfs_iflock_nowait(ip)) {
+					xfs_iunlock(ip, XFS_ILOCK_SHARED);
+					return EAGAIN;
+				}
 
-				if ((xfs_ipincount(ip) == 0) &&
-				     xfs_iflock_nowait(ip))
 				error = xfs_iflush(ip, flush_flags);
 			} else {
 				error = EAGAIN;
 			}
 			xfs_iunlock(ip, XFS_ILOCK_SHARED);
 		} else {
-eagain:
 			error = EAGAIN;
 		}
 	}
@@ -3934,6 +3937,8 @@ xfs_reclaim(
 		/* Protect sync from us */
 		XFS_MOUNT_ILOCK(mp);
 		vn_bhv_remove(VN_BHV_HEAD(vp), XFS_ITOBHV(ip));
+		list_add_tail(&ip->i_reclaim, &mp->m_del_inodes);
+
 		XFS_MOUNT_IUNLOCK(mp);
 	}
 	return 0;
@@ -4010,40 +4015,33 @@ xfs_finish_reclaim(
 }
 
 int
-xfs_finish_reclaim_all(xfs_mount_t *mp)
+xfs_finish_reclaim_all(xfs_mount_t *mp, int noblock)
 {
 	int		purged;
+	struct list_head	*curr, *next;
 	xfs_inode_t	*ip;
-	vnode_t		*vp;
 	int		done = 0;
 
 	while (!done) {
 		purged = 0;
 		XFS_MOUNT_ILOCK(mp);
-		ip = mp->m_inodes;
-		if (ip == NULL) {
-			break;
-		}
-		do {
-			/* Make sure we skip markers inserted by sync */
-			if (ip->i_mount == NULL) {
-				ip = ip->i_mnext;
+		list_for_each_safe(curr, next, &mp->m_del_inodes) {
+			ip = list_entry(curr, xfs_inode_t, i_reclaim);
+			if (noblock) {
+				if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0)
+					continue;
+				if (xfs_ipincount(ip) ||
+				    !xfs_iflock_nowait(ip)) {
+					xfs_iunlock(ip, XFS_ILOCK_EXCL);
 					continue;
 				}
-
-			/*
-			 * It's up to our caller to purge the root
-			 * and quota vnodes later.
-			 */
-			vp = XFS_ITOV_NULL(ip);
-
-			if (!vp) {
+			}
 			XFS_MOUNT_IUNLOCK(mp);
-				xfs_finish_reclaim(ip, 0, XFS_IFLUSH_ASYNC);
+			xfs_finish_reclaim(ip, noblock,
+				XFS_IFLUSH_DELWRI_ELSE_ASYNC);
 			purged = 1;
 			break;
 		}
-		} while (ip != mp->m_inodes);
 
 		done = !purged;
 	}

fs/xfs/xfsidbg.c

@@ -4269,8 +4269,10 @@ xfsidbg_xlog(xlog_t *log)
 	kdb_printf("iclog_bak: 0x%p  iclog_size: 0x%x (%d)  num iclogs: %d\n",
 		log->l_iclog_bak, log->l_iclog_size, log->l_iclog_size,
 		log->l_iclog_bufs);
-	kdb_printf("l_iclog_hsize %d l_iclog_heads %d\n",
-		log->l_iclog_hsize, log->l_iclog_heads);
+	kdb_printf("l_stripemask %d l_iclog_hsize %d l_iclog_heads %d\n",
+		log->l_stripemask, log->l_iclog_hsize, log->l_iclog_heads);
+	kdb_printf("l_sectbb_log %u l_sectbb_mask %u\n",
+		log->l_sectbb_log, log->l_sectbb_mask);
 	kdb_printf("&grant_lock: 0x%p  resHeadQ: 0x%p  wrHeadQ: 0x%p\n",
 		&log->l_grant_lock, log->l_reserve_headq, log->l_write_headq);
 	kdb_printf("GResCycle: %d  GResBytes: %d  GWrCycle: %d  GWrBytes: %d\n",
@@ -4712,7 +4714,6 @@ xfsidbg_xmount(xfs_mount_t *mp)
 		(xfs_dfiloff_t)mp->m_dirfreeblk);
 	kdb_printf("chsize %d chash 0x%p\n",
 		mp->m_chsize, mp->m_chash);
-	kdb_printf("m_lstripemask %d\n", mp->m_lstripemask);
 	kdb_printf("m_frozen %d m_active_trans %d\n",
 		mp->m_frozen, mp->m_active_trans.counter);
 	if (mp->m_fsname != NULL)