Merge master.kernel.org:/home/hch/BK/xfs/linux-2.5

into home.transmeta.com:/home/torvalds/v2.5/linux

fs/xattr.c

@@ -8,7 +8,6 @@
  */
 #include <linux/fs.h>
 #include <linux/slab.h>
-#include <linux/vmalloc.h>
 #include <linux/smp_lock.h>
 #include <linux/file.h>
 #include <linux/xattr.h>
@@ -19,7 +18,7 @@
 /*
  * Extended attribute memory allocation wrappers, originally
  * based on the Intermezzo PRESTO_ALLOC/PRESTO_FREE macros.
- * The vmalloc use here is very uncommon - extended attributes
+ * Values larger than a page are uncommon - extended attributes
  * are supposed to be small chunks of metadata, and it is quite
  * unusual to have very many extended attributes, so lists tend
  * to be quite short as well.  The 64K upper limit is derived
@@ -36,10 +35,8 @@ xattr_alloc(size_t size, size_t limit)
 
 	if (!size)	/* size request, no buffer is needed */
 		return NULL;
-	else if (size <= PAGE_SIZE)
+
 	ptr = kmalloc((unsigned long) size, GFP_KERNEL);
-	else
-		ptr = vmalloc((unsigned long) size);
 	if (!ptr)
 		return ERR_PTR(-ENOMEM);
 	return ptr;
@@ -48,12 +45,8 @@ xattr_alloc(size_t size, size_t limit)
 static void
 xattr_free(void *ptr, size_t size)
 {
-	if (!size)	/* size request, no buffer was needed */
-		return;
-	else if (size <= PAGE_SIZE)
+	if (size)	/* for a size request, no buffer was needed */
 		kfree(ptr);
-	else
-		vfree(ptr);
 }
 
 /*

fs/xfs/linux/xfs_aops.c

@@ -135,17 +135,14 @@ probe_unmapped_page(
 	struct page		*page;
 	int			ret = 0;
 
-	page = find_get_page(mapping, index);
+	page = find_trylock_page(mapping, index);
 	if (!page)
 		return 0;
-	if (PageWriteback(page) || TestSetPageLocked(page)) {
-		page_cache_release(page);
-		return 0;
-	}
+	if (PageWriteback(page))
+		goto out;
+
 	if (page->mapping && PageDirty(page)) {
-		if (!page_has_buffers(page)) {
-			ret = PAGE_CACHE_SIZE;
-		} else {
+		if (page_has_buffers(page)) {
 			struct buffer_head	*bh, *head;
 			bh = head = page_buffers(page);
 			do {
@@ -156,11 +153,12 @@ probe_unmapped_page(
 				if (ret >= pg_offset)
 					break;
 			} while ((bh = bh->b_this_page) != head);
-		}
+		} else
+			ret = PAGE_CACHE_SIZE;
 	}
 
+out:
 	unlock_page(page);
-	page_cache_release(page);
 	return ret;
 }
 
@@ -214,13 +212,12 @@ probe_page(
 {
 	struct page		*page;
 
-	page = find_get_page(inode->i_mapping, index);
+	page = find_trylock_page(inode->i_mapping, index);
 	if (!page)
 		return NULL;
-	if (PageWriteback(page) || TestSetPageLocked(page)) {
-		page_cache_release(page);
-		return NULL;
-	}
+	if (PageWriteback(page))
+		goto out;
+
 	if (page->mapping && page_has_buffers(page)) {
 		struct buffer_head	*bh, *head;
 
@@ -230,8 +227,9 @@ probe_page(
 				return page;
 		} while ((bh = bh->b_this_page) != head);
 	}
+
+out:
 	unlock_page(page);
-	page_cache_release(page);
 	return NULL;
 }
 
@@ -319,8 +317,6 @@ convert_page(
 	} else {
 		unlock_page(page);
 	}
-
-	page_cache_release(page);
 }
 
 /*

fs/xfs/linux/xfs_file.c

@@ -138,6 +138,22 @@ linvfs_aio_write(
 	return linvfs_writev(iocb->ki_filp, &iov, 1, &iocb->ki_pos);
 }
 
+STATIC ssize_t
+linvfs_sendfile(
+	struct file		*filp,
+	loff_t			*ppos,
+	size_t			count,
+	read_actor_t		actor,
+	void			*target)
+{
+	vnode_t			*vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
+	int			error;
+
+	VOP_SENDFILE(vp, filp, ppos, count, actor, target, NULL, error);
+
+	return error;
+}
+
 
 STATIC int
 linvfs_open(
@@ -186,7 +202,7 @@ linvfs_fsync(
 
 	ASSERT(vp);
 
-	VOP_FSYNC(vp, flags, NULL, (off_t)0, (off_t)-1, error);
+	VOP_FSYNC(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1, error);
 
 	return -error;
 }
@@ -212,7 +228,7 @@ linvfs_readdir(
 	caddr_t		read_buf;
 	int		namelen, size = 0;
 	size_t		rlen = PAGE_CACHE_SIZE << 2;
-	off_t		start_offset;
+	xfs_off_t	start_offset;
 	xfs_dirent_t	*dbp = NULL;
 
 	vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
@@ -280,7 +296,7 @@ linvfs_file_mmap(
 {
 	struct inode	*ip = filp->f_dentry->d_inode;
 	vnode_t		*vp = LINVFS_GET_VP(ip);
-	vattr_t		va = { .va_mask = AT_UPDATIME };
+	vattr_t		va = { .va_mask = XFS_AT_UPDATIME };
 	int		error;
 
 	if ((vp->v_type == VREG) && (vp->v_vfsp->vfs_flag & VFS_DMI)) {
@@ -291,7 +307,7 @@ linvfs_file_mmap(
 
 	vma->vm_ops = &linvfs_file_vm_ops;
 
-	VOP_SETATTR(vp, &va, AT_UPDATIME, NULL, error);
+	VOP_SETATTR(vp, &va, XFS_AT_UPDATIME, NULL, error);
 	UPDATE_ATIME(ip);
 	return 0;
 }
@@ -348,6 +364,7 @@ struct file_operations linvfs_file_operations = {
 	.writev		= linvfs_writev,
 	.aio_read	= linvfs_aio_read,
 	.aio_write	= linvfs_aio_write,
+	.sendfile	= linvfs_sendfile,
 	.ioctl		= linvfs_ioctl,
 	.mmap		= linvfs_file_mmap,
 	.open		= linvfs_open,

fs/xfs/linux/xfs_ioctl.c

@@ -34,8 +34,9 @@
 #include <xfs_fsops.h>
 #include <xfs_dfrag.h>
 #include <linux/dcache.h>
-#include <linux/namei.h>
 #include <linux/mount.h>
+#include <linux/namei.h>
+#include <linux/pagemap.h>
 
 
 extern int xfs_change_file_space(bhv_desc_t *, int,
@@ -591,17 +592,16 @@ xfs_ioctl(
 	case XFS_IOC_DIOINFO: {
 		struct dioattr	da;
 
-		da.d_miniosz = mp->m_sb.sb_blocksize;
-		da.d_mem = mp->m_sb.sb_blocksize;
 
 		/*
 		 * this only really needs to be BBSIZE.
 		 * it is set to the file system block size to
 		 * avoid having to do block zeroing on short writes.
 		 */
-#define KIO_MAX_ATOMIC_IO 512	/* FIXME: what do we really want here? */
-		da.d_maxiosz = XFS_FSB_TO_B(mp,
-				XFS_B_TO_FSBT(mp, KIO_MAX_ATOMIC_IO << 10));
+		da.d_miniosz = mp->m_sb.sb_blocksize;
+		da.d_mem = mp->m_sb.sb_blocksize;
+		/* The size dio will do in one go */
+		da.d_maxiosz = 64 * PAGE_CACHE_SIZE;
 
 		if (copy_to_user((struct dioattr *)arg, &da, sizeof(da)))
 			return -XFS_ERROR(EFAULT);
@@ -945,7 +945,7 @@ int xfs_ioc_xattr(
 
 	switch (cmd) {
 	case XFS_IOC_FSGETXATTR: {
-		va.va_mask = AT_XFLAGS|AT_EXTSIZE|AT_NEXTENTS;
+		va.va_mask = XFS_AT_XFLAGS|XFS_AT_EXTSIZE|XFS_AT_NEXTENTS;
 		VOP_GETATTR(vp, &va, 0, NULL, error);
 		if (error)
 			return -error;
@@ -965,7 +965,7 @@ int xfs_ioc_xattr(
 		if (copy_from_user(&fa, (struct fsxattr *)arg, sizeof(fa)))
 			return -XFS_ERROR(EFAULT);
 
-		va.va_mask = AT_XFLAGS | AT_EXTSIZE;
+		va.va_mask = XFS_AT_XFLAGS | XFS_AT_EXTSIZE;
 		va.va_xflags  = fa.fsx_xflags;
 		va.va_extsize = fa.fsx_extsize;
 
@@ -978,7 +978,7 @@ int xfs_ioc_xattr(
 
 	case XFS_IOC_FSGETXATTRA: {
 
-		va.va_mask = AT_XFLAGS|AT_EXTSIZE|AT_ANEXTENTS;
+		va.va_mask = XFS_AT_XFLAGS|XFS_AT_EXTSIZE|XFS_AT_ANEXTENTS;
 		VOP_GETATTR(vp, &va, 0, NULL, error);
 		if (error)
 			return -error;

fs/xfs/linux/xfs_iops.c

@@ -45,7 +45,7 @@ validate_fields(
 	vattr_t		va;
 	int		error;
 
-	va.va_mask = AT_NLINK|AT_SIZE;
+	va.va_mask = XFS_AT_NLINK|XFS_AT_SIZE;
 	VOP_GETATTR(vp, &va, ATTR_LAZY, NULL, error);
 	ip->i_nlink = va.va_nlink;
 	ip->i_size = va.va_size;
@@ -85,14 +85,14 @@ linvfs_mknod(
 		mode &= ~current->fs->umask;
 
 	memset(&va, 0, sizeof(va));
-	va.va_mask = AT_TYPE|AT_MODE;
+	va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
 	va.va_type = IFTOVT(mode);
 	va.va_mode = mode;
 
 	switch (mode & S_IFMT) {
 	case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK:
 		va.va_rdev = XFS_MKDEV(MAJOR(rdev), MINOR(rdev));
-		va.va_mask |= AT_RDEV;
+		va.va_mask |= XFS_AT_RDEV;
 		/*FALLTHROUGH*/
 	case S_IFREG:
 		VOP_CREATE(dvp, dentry, &va, &vp, NULL, error);
@@ -255,7 +255,7 @@ linvfs_symlink(
 	memset(&va, 0, sizeof(va));
 	va.va_type = VLNK;
 	va.va_mode = irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO;
-	va.va_mask = AT_TYPE|AT_MODE;
+	va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
 
 	error = 0;
 	VOP_SYMLINK(dvp, dentry, &va, (char *)symname, &cvp, NULL, error);
@@ -462,31 +462,31 @@ linvfs_setattr(
 
 	memset(&vattr, 0, sizeof(vattr_t));
 	if (ia_valid & ATTR_UID) {
-		vattr.va_mask |= AT_UID;
+		vattr.va_mask |= XFS_AT_UID;
 		vattr.va_uid = attr->ia_uid;
 	}
 	if (ia_valid & ATTR_GID) {
-		vattr.va_mask |= AT_GID;
+		vattr.va_mask |= XFS_AT_GID;
 		vattr.va_gid = attr->ia_gid;
 	}
 	if (ia_valid & ATTR_SIZE) {
-		vattr.va_mask |= AT_SIZE;
+		vattr.va_mask |= XFS_AT_SIZE;
 		vattr.va_size = attr->ia_size;
 	}
 	if (ia_valid & ATTR_ATIME) {
-		vattr.va_mask |= AT_ATIME;
+		vattr.va_mask |= XFS_AT_ATIME;
 		vattr.va_atime = attr->ia_atime;
 	}
 	if (ia_valid & ATTR_MTIME) {
-		vattr.va_mask |= AT_MTIME;
+		vattr.va_mask |= XFS_AT_MTIME;
 		vattr.va_mtime = attr->ia_mtime;
 	}
 	if (ia_valid & ATTR_CTIME) {
-		vattr.va_mask |= AT_CTIME;
+		vattr.va_mask |= XFS_AT_CTIME;
 		vattr.va_ctime = attr->ia_ctime;
 	}
 	if (ia_valid & ATTR_MODE) {
-		vattr.va_mask |= AT_MODE;
+		vattr.va_mask |= XFS_AT_MODE;
 		vattr.va_mode = attr->ia_mode;
 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
 			inode->i_mode &= ~S_ISGID;

fs/xfs/linux/xfs_linux.h

@@ -153,7 +153,6 @@ typedef struct xfs_dirent {		/* data from readdir() */
 #define EFSCORRUPTED	990		/* Filesystem is corrupted */
 
 #define SYNCHRONIZE()	barrier()
-#define rootdev		ROOT_DEV
 #define __return_address __builtin_return_address(0)
 
 /* IRIX uses a dynamic sizing algorithm (ndquot = 200 + numprocs*2) */

fs/xfs/linux/xfs_lrw.c

@@ -205,6 +205,67 @@ xfs_read(
 	return ret;
 }
 
+ssize_t
+xfs_sendfile(
+	bhv_desc_t		*bdp,
+	struct file		*filp,
+	loff_t			*offp,
+	size_t			count,
+	read_actor_t		actor,
+	void			*target,
+	cred_t			*credp)
+{
+	size_t			size = 0;
+	ssize_t			ret;
+	xfs_fsize_t		n;
+	xfs_inode_t		*ip;
+	xfs_mount_t		*mp;
+	vnode_t			*vp;
+
+	ip = XFS_BHVTOI(bdp);
+	vp = BHV_TO_VNODE(bdp);
+	mp = ip->i_mount;
+	vn_trace_entry(vp, "xfs_sendfile", (inst_t *)__return_address);
+
+	XFS_STATS_INC(xfsstats.xs_read_calls);
+
+	n = XFS_MAX_FILE_OFFSET - *offp;
+	if ((n <= 0) || (size == 0))
+		return 0;
+
+	if (n < size)
+		size = n;
+
+	if (XFS_FORCED_SHUTDOWN(mp)) {
+		return -EIO;
+	}
+
+	xfs_ilock(ip, XFS_IOLOCK_SHARED);
+
+	if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ) &&
+	    !(filp->f_mode & FINVIS)) {
+		int error;
+		vrwlock_t locktype = VRWLOCK_READ;
+
+		error = xfs_dm_send_data_event(DM_EVENT_READ, bdp, *offp,
+				size, FILP_DELAY_FLAG(filp), &locktype);
+		if (error) {
+			xfs_iunlock(ip, XFS_IOLOCK_SHARED);
+			return -error;
+		}
+	}
+
+	ret = generic_file_sendfile(filp, offp, count, actor, target);
+	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
+
+	XFS_STATS_ADD(xfsstats.xs_read_bytes, ret);
+
+	if (!(filp->f_mode & FINVIS))
+		xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
+
+	return ret;
+}
+
 /*
  * This routine is called to handle zeroing any space in the last
  * block of the file that is beyond the EOF.  We do this since the

fs/xfs/linux/xfs_stats.h

@@ -32,15 +32,128 @@
 #ifndef __XFS_STATS_H__
 #define __XFS_STATS_H__
 
+
+#if defined(CONFIG_PROC_FS) && !defined(XFS_STATS_OFF)
+
 /*
- * procfs interface
+ * XFS global statistics
  */
-#ifdef CONFIG_PROC_FS
+struct xfsstats {
+# define XFSSTAT_END_EXTENT_ALLOC	4
+	__uint32_t		xs_allocx;
+	__uint32_t		xs_allocb;
+	__uint32_t		xs_freex;
+	__uint32_t		xs_freeb;
+# define XFSSTAT_END_ALLOC_BTREE	(XFSSTAT_END_EXTENT_ALLOC+4)
+	__uint32_t		xs_abt_lookup;
+	__uint32_t		xs_abt_compare;
+	__uint32_t		xs_abt_insrec;
+	__uint32_t		xs_abt_delrec;
+# define XFSSTAT_END_BLOCK_MAPPING	(XFSSTAT_END_ALLOC_BTREE+7)
+	__uint32_t		xs_blk_mapr;
+	__uint32_t		xs_blk_mapw;
+	__uint32_t		xs_blk_unmap;
+	__uint32_t		xs_add_exlist;
+	__uint32_t		xs_del_exlist;
+	__uint32_t		xs_look_exlist;
+	__uint32_t		xs_cmp_exlist;
+# define XFSSTAT_END_BLOCK_MAP_BTREE	(XFSSTAT_END_BLOCK_MAPPING+4)
+	__uint32_t		xs_bmbt_lookup;
+	__uint32_t		xs_bmbt_compare;
+	__uint32_t		xs_bmbt_insrec;
+	__uint32_t		xs_bmbt_delrec;
+# define XFSSTAT_END_DIRECTORY_OPS	(XFSSTAT_END_BLOCK_MAP_BTREE+4)
+	__uint32_t		xs_dir_lookup;
+	__uint32_t		xs_dir_create;
+	__uint32_t		xs_dir_remove;
+	__uint32_t		xs_dir_getdents;
+# define XFSSTAT_END_TRANSACTIONS	(XFSSTAT_END_DIRECTORY_OPS+3)
+	__uint32_t		xs_trans_sync;
+	__uint32_t		xs_trans_async;
+	__uint32_t		xs_trans_empty;
+# define XFSSTAT_END_INODE_OPS		(XFSSTAT_END_TRANSACTIONS+7)
+	__uint32_t		xs_ig_attempts;
+	__uint32_t		xs_ig_found;
+	__uint32_t		xs_ig_frecycle;
+	__uint32_t		xs_ig_missed;
+	__uint32_t		xs_ig_dup;
+	__uint32_t		xs_ig_reclaims;
+	__uint32_t		xs_ig_attrchg;
+# define XFSSTAT_END_LOG_OPS		(XFSSTAT_END_INODE_OPS+5)
+	__uint32_t		xs_log_writes;
+	__uint32_t		xs_log_blocks;
+	__uint32_t		xs_log_noiclogs;
+	__uint32_t		xs_log_force;
+	__uint32_t		xs_log_force_sleep;
+# define XFSSTAT_END_TAIL_PUSHING	(XFSSTAT_END_LOG_OPS+10)
+	__uint32_t		xs_try_logspace;
+	__uint32_t		xs_sleep_logspace;
+	__uint32_t		xs_push_ail;
+	__uint32_t		xs_push_ail_success;
+	__uint32_t		xs_push_ail_pushbuf;
+	__uint32_t		xs_push_ail_pinned;
+	__uint32_t		xs_push_ail_locked;
+	__uint32_t		xs_push_ail_flushing;
+	__uint32_t		xs_push_ail_restarts;
+	__uint32_t		xs_push_ail_flush;
+# define XFSSTAT_END_WRITE_CONVERT	(XFSSTAT_END_TAIL_PUSHING+2)
+	__uint32_t		xs_xstrat_quick;
+	__uint32_t		xs_xstrat_split;
+# define XFSSTAT_END_READ_WRITE_OPS	(XFSSTAT_END_WRITE_CONVERT+2)
+	__uint32_t		xs_write_calls;
+	__uint32_t		xs_read_calls;
+# define XFSSTAT_END_ATTRIBUTE_OPS	(XFSSTAT_END_READ_WRITE_OPS+4)
+	__uint32_t		xs_attr_get;
+	__uint32_t		xs_attr_set;
+	__uint32_t		xs_attr_remove;
+	__uint32_t		xs_attr_list;
+# define XFSSTAT_END_QUOTA_OPS		(XFSSTAT_END_ATTRIBUTE_OPS+8)
+	__uint32_t		xs_qm_dqreclaims;
+	__uint32_t		xs_qm_dqreclaim_misses;
+	__uint32_t		xs_qm_dquot_dups;
+	__uint32_t		xs_qm_dqcachemisses;
+	__uint32_t		xs_qm_dqcachehits;
+	__uint32_t		xs_qm_dqwants;
+	__uint32_t		xs_qm_dqshake_reclaims;
+	__uint32_t		xs_qm_dqinact_reclaims;
+# define XFSSTAT_END_INODE_CLUSTER	(XFSSTAT_END_QUOTA_OPS+3)
+	__uint32_t		xs_iflush_count;
+	__uint32_t		xs_icluster_flushcnt;
+	__uint32_t		xs_icluster_flushinode;
+# define XFSSTAT_END_VNODE_OPS		(XFSSTAT_END_INODE_CLUSTER+8)
+	__uint32_t		vn_active;	/* # vnodes not on free lists */
+	__uint32_t		vn_alloc;	/* # times vn_alloc called */
+	__uint32_t		vn_get;		/* # times vn_get called */
+	__uint32_t		vn_hold;	/* # times vn_hold called */
+	__uint32_t		vn_rele;	/* # times vn_rele called */
+	__uint32_t		vn_reclaim;	/* # times vn_reclaim called */
+	__uint32_t		vn_remove;	/* # times vn_remove called */
+	__uint32_t		vn_free;	/* # times vn_free called */
+/* Extra precision counters */
+	__uint64_t		xs_xstrat_bytes;
+	__uint64_t		xs_write_bytes;
+	__uint64_t		xs_read_bytes;
+};
+
+extern struct xfsstats xfsstats;
+
+# define XFS_STATS_INC(count)		( (count)++ )
+# define XFS_STATS_DEC(count)		( (count)-- )
+# define XFS_STATS_ADD(count, inc)	( (count) += (inc) )
+
 extern void xfs_init_procfs(void);
 extern void xfs_cleanup_procfs(void);
-#else
+
+
+#else	/* !CONFIG_PROC_FS */
+
+# define XFS_STATS_INC(count)
+# define XFS_STATS_DEC(count)
+# define XFS_STATS_ADD(count, inc)
+
 static __inline void xfs_init_procfs(void) { };
 static __inline void xfs_cleanup_procfs(void) { };
-#endif
+
+#endif	/* !CONFIG_PROC_FS */
 
 #endif /* __XFS_STATS_H__ */

fs/xfs/linux/xfs_super.c

@@ -46,10 +46,6 @@
 extern int  xfs_init(void);
 extern void xfs_cleanup(void);
 
-#ifndef EVMS_MAJOR
-# define EVMS_MAJOR		117
-#endif
-
 /* For kernels which have the s_maxbytes field - set it */
 #ifdef MAX_NON_LFS
 # define set_max_bytes(sb)	((sb)->s_maxbytes = XFS_MAX_FILE_OFFSET)
@@ -101,7 +97,6 @@ STATIC struct export_operations linvfs_export_ops;
 #define MNTOPT_NORECOVERY "norecovery"	/* don't run XFS recovery */
 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
 #define MNTOPT_QUOTA	"quota"		/* disk quotas */
-#define MNTOPT_MRQUOTA	"mrquota"	/* don't turnoff if SB has quotas on */
 #define MNTOPT_NOQUOTA	"noquota"	/* no quotas */
 #define MNTOPT_UQUOTA	"usrquota"	/* user quota enabled */
 #define MNTOPT_GQUOTA	"grpquota"	/* group quota enabled */
@@ -513,6 +508,23 @@ xfs_free_buftarg(
 	kfree(btp);
 }
 
+void
+xfs_size_buftarg(
+	xfs_buftarg_t		*btp,
+	unsigned int		blocksize,
+	unsigned int		sectorsize)
+{
+	btp->pbr_bsize = blocksize;
+	btp->pbr_sshift = ffs(sectorsize) - 1;
+	btp->pbr_smask = sectorsize - 1;
+
+	if (set_blocksize(btp->pbr_bdev, sectorsize)) {
+		printk(KERN_WARNING
+			"XFS: Cannot set_blocksize to %u on device 0x%x\n",
+			sectorsize, btp->pbr_dev);
+	}
+}
+
 xfs_buftarg_t *
 xfs_alloc_buftarg(
 	struct block_device	*bdev)
@@ -524,14 +536,7 @@ xfs_alloc_buftarg(
 	btp->pbr_dev =  bdev->bd_dev;
 	btp->pbr_bdev = bdev;
 	btp->pbr_mapping = bdev->bd_inode->i_mapping;
-	btp->pbr_blocksize = PAGE_CACHE_SIZE;
-
-	switch (MAJOR(btp->pbr_dev)) {
-	case MD_MAJOR:
-	case EVMS_MAJOR:
-		btp->pbr_flags = PBR_ALIGNED_ONLY;
-		break;
-	}
+	xfs_size_buftarg(btp, PAGE_CACHE_SIZE, bdev_hardsect_size(bdev));
 
 	return btp;
 }

fs/xfs/linux/xfs_super.h

@@ -87,6 +87,7 @@ extern int  xfs_blkdev_get (const char *, struct block_device **);
 extern void xfs_blkdev_put (struct block_device *);
 
 extern struct pb_target *xfs_alloc_buftarg (struct block_device *);
+extern void xfs_size_buftarg (struct pb_target *, unsigned int, unsigned int);
 extern void xfs_free_buftarg (struct pb_target *);
 
 #endif	/* __XFS_SUPER_H__ */

fs/xfs/linux/xfs_vnode.c

@@ -35,7 +35,6 @@
 
 
 uint64_t vn_generation;		/* vnode generation number */
-
 spinlock_t vnumber_lock = SPIN_LOCK_UNLOCKED;
 
 /*
@@ -59,9 +58,6 @@ u_short vttoif_tab[] = {
 	0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, S_IFIFO, 0, S_IFSOCK
 };
 
-#define VN_LOCK(vp)		spin_lock(&(vp)->v_lock)
-#define VN_UNLOCK(vp)		spin_unlock(&(vp)->v_lock)
-
 
 void
 vn_init(void)
@@ -73,7 +69,6 @@ vn_init(void)
 		init_sv(svp, SV_DEFAULT, "vsy", i);
 }
 
-
 /*
  * Clean a vnode of filesystem-specific data and prepare it for reuse.
  */
@@ -98,7 +93,6 @@ vn_reclaim(struct vnode *vp)
 	ASSERT(vp->v_fbhv == NULL);
 
 	VN_LOCK(vp);
-
 	vp->v_flag &= (VRECLM|VWAIT);
 	VN_UNLOCK(vp);
 
@@ -189,7 +183,7 @@ vn_get(struct vnode *vp, vmap_t *vmap)
 }
 
 /*
- * "revalidate" the linux inode.
+ * Revalidate the Linux inode from the vnode.
  */
 int
 vn_revalidate(struct vnode *vp)
@@ -199,17 +193,12 @@ vn_revalidate(struct vnode *vp)
 	vattr_t		va;
 
 	vn_trace_entry(vp, "vn_revalidate", (inst_t *)__return_address);
+	ASSERT(vp->v_fbhv != NULL);
 
-	va.va_mask = AT_STAT|AT_GENCOUNT;
-
-	ASSERT(vp->v_bh.bh_first != NULL);
-
+	va.va_mask = XFS_AT_STAT|XFS_AT_GENCOUNT;
 	VOP_GETATTR(vp, &va, 0, NULL, error);
-
-	if (! error) {
+	if (!error) {
 		inode = LINVFS_GET_IP(vp);
-		ASSERT(inode);
-
 		inode->i_mode	    = VTTOIF(va.va_type) | va.va_mode;
 		inode->i_nlink	    = va.va_nlink;
 		inode->i_uid	    = va.va_uid;
@@ -224,11 +213,9 @@ vn_revalidate(struct vnode *vp)
 		inode->i_atime.tv_nsec	    = va.va_atime.tv_nsec;
 		VUNMODIFY(vp);
 	}
-
 	return -error;
 }
 
-
 /*
  * purge a vnode from the cache
  * At this point the vnode is guaranteed to have no references (vn_count == 0)
@@ -317,12 +304,10 @@ void
 vn_rele(struct vnode *vp)
 {
 	int	vcnt;
-	/* REFERENCED */
 	int	cache;
 
 	XFS_STATS_INC(xfsstats.vn_rele);
 
-
 	VN_LOCK(vp);
 
 	vn_trace_entry(vp, "vn_rele", (inst_t *)__return_address);
@@ -365,7 +350,6 @@ vn_rele(struct vnode *vp)
 	vn_trace_exit(vp, "vn_rele", (inst_t *)__return_address);
 }
 
-
 /*
  * Finish the removal of a vnode.
  */

fs/xfs/pagebuf/page_buf.c

@@ -63,10 +63,8 @@
 
 #include "page_buf_internal.h"
 
-#define SECTOR_SHIFT	9
-#define SECTOR_SIZE	(1<<SECTOR_SHIFT)
-#define SECTOR_MASK	(SECTOR_SIZE - 1)
-#define BN_ALIGN_MASK	((1 << (PAGE_CACHE_SHIFT - SECTOR_SHIFT)) - 1)
+#define BBSHIFT		9
+#define BN_ALIGN_MASK	((1 << (PAGE_CACHE_SHIFT - BBSHIFT)) - 1)
 
 #ifndef GFP_READAHEAD
 #define GFP_READAHEAD	0
@@ -245,19 +243,27 @@ typedef struct a_list {
 STATIC	a_list_t	*as_free_head;
 STATIC	int		as_list_len;
 
+
+/*
+ * Try to batch vunmaps because they are costly.
+ */
 STATIC void
 free_address(
 	void		*addr)
 {
 	a_list_t	*aentry;
 
-	spin_lock(&as_lock);
 	aentry = kmalloc(sizeof(a_list_t), GFP_ATOMIC);
+	if (aentry) {
+		spin_lock(&as_lock);
 		aentry->next = as_free_head;
 		aentry->vm_addr = addr;
 		as_free_head = aentry;
 		as_list_len++;
 		spin_unlock(&as_lock);
+	} else {
+		vunmap(addr);
+	}
 }
 
 STATIC void
@@ -265,7 +271,8 @@ purge_addresses(void)
 {
 	a_list_t	*aentry, *old;
 
-	if (as_free_head == NULL) return;
+	if (as_free_head == NULL)
+		return;
 
 	spin_lock(&as_lock);
 	aentry = as_free_head;
@@ -462,7 +469,8 @@ _pagebuf_lookup_pages(
 	struct page		*page;
 	int			gfp_mask, retry_count = 5, rval = 0;
 	int			all_mapped, good_pages, nbytes;
-	size_t			blocksize, size, offset;
+	unsigned int		blocksize, sectorshift;
+	size_t			size, offset;
 
 
 	/* For pagebufs where we want to map an address, do not use
@@ -508,7 +516,8 @@ _pagebuf_lookup_pages(
 		return rval;
 
 	rval = pi = 0;
-	blocksize = pb->pb_target->pbr_blocksize;
+	blocksize = pb->pb_target->pbr_bsize;
+	sectorshift = pb->pb_target->pbr_sshift;
 	size = pb->pb_count_desired;
 	offset = pb->pb_offset;
 
@@ -549,15 +558,15 @@ _pagebuf_lookup_pages(
 					pb->pb_locked = 1;
 				good_pages--;
 			} else if (!PagePrivate(page)) {
-				unsigned long i, range = (offset + nbytes) >> SECTOR_SHIFT;
+				unsigned long	i, range;
 
-				ASSERT(blocksize < PAGE_CACHE_SIZE);
-				ASSERT(!(pb->pb_flags & _PBF_PRIVATE_BH));
 				/*
 				 * In this case page->private holds a bitmap
-				 * of uptodate sectors (512) within the page
+				 * of uptodate sectors within the page
 				 */
-				for (i = offset >> SECTOR_SHIFT; i < range; i++)
+				ASSERT(blocksize < PAGE_CACHE_SIZE);
+				range = (offset + nbytes) >> sectorshift;
+				for (i = offset >> sectorshift; i < range; i++)
 					if (!test_bit(i, &page->private))
 						break;
 				if (i != range)
@@ -642,8 +651,14 @@ _pagebuf_find(				/* find buffer for block	*/
 	page_buf_t		*pb;
 	int			not_locked;
 
-	range_base = (ioff << SECTOR_SHIFT);
-	range_length = (isize << SECTOR_SHIFT);
+	range_base = (ioff << BBSHIFT);
+	range_length = (isize << BBSHIFT);
+
+	/* Ensure we never do IOs smaller than the sector size */
+	BUG_ON(range_length < (1 << target->pbr_sshift));
+
+	/* Ensure we never do IOs that are not sector aligned */
+	BUG_ON(range_base & (loff_t)target->pbr_smask);
 
 	hval = _bhash(target->pbr_bdev->bd_dev, range_base);
 	h = &pbhash[hval];
@@ -851,8 +866,20 @@ pagebuf_readahead(
 	size_t			isize,
 	page_buf_flags_t	flags)
 {
+	struct backing_dev_info *bdi;
+
+	bdi = target->pbr_mapping->backing_dev_info;
+	if (bdi_read_congested(bdi))
+		return;
+	if (bdi_write_congested(bdi))
+		return;
+
 	flags |= (PBF_TRYLOCK|PBF_READ|PBF_ASYNC|PBF_MAPPABLE|PBF_READ_AHEAD);
+
+	/* don't complain on allocation failure, it's fine with us */
+	current->flags |= PF_NOWARN;
 	pagebuf_get(target, ioff, isize, flags);
+	current->flags &= ~PF_NOWARN;
 }
 
 page_buf_t *
@@ -956,18 +983,12 @@ pagebuf_get_no_daddr(
 		} else {
 			kfree(rmem); /* free the mem from the previous try */
 			tlen <<= 1; /* double the size and try again */
-			/*
-			printk(
-			"pb_get_no_daddr NOT block 0x%p mask 0x%p len %d\n",
-				rmem, ((size_t)rmem & (size_t)~SECTOR_MASK),
-				len);
-			*/
 		}
 		if ((rmem = kmalloc(tlen, GFP_KERNEL)) == 0) {
 			pagebuf_free(pb);
 			return NULL;
 		}
-	} while ((size_t)rmem != ((size_t)rmem & (size_t)~SECTOR_MASK));
+	} while ((size_t)rmem != ((size_t)rmem & ~target->pbr_smask));
 
 	if ((rval = pagebuf_associate_memory(pb, rmem, len)) != 0) {
 		kfree(rmem);
@@ -1248,9 +1269,7 @@ pagebuf_iostart(			/* start I/O on a buffer	  */
 	pb->pb_flags &= ~(PBF_READ|PBF_WRITE|PBF_ASYNC|PBF_DELWRI|PBF_READ_AHEAD);
 	pb->pb_flags |= flags & (PBF_READ|PBF_WRITE|PBF_ASYNC|PBF_SYNC|PBF_READ_AHEAD);
 
-	if (pb->pb_bn == PAGE_BUF_DADDR_NULL) {
-		BUG();
-	}
+	BUG_ON(pb->pb_bn == PAGE_BUF_DADDR_NULL);
 
 	/* For writes call internal function which checks for
 	 * filesystem specific callout function and execute it.
@@ -1279,7 +1298,8 @@ bio_end_io_pagebuf(
 	int			error)
 {
 	page_buf_t		*pb = (page_buf_t *)bio->bi_private;
-	unsigned int		i, blocksize = pb->pb_target->pbr_blocksize;
+	unsigned int		i, blocksize = pb->pb_target->pbr_bsize;
+	unsigned int		sectorshift = pb->pb_target->pbr_sshift;
 	struct bio_vec		*bvec = bio->bi_io_vec;
 
 	if (bio->bi_size)
@@ -1299,10 +1319,8 @@ bio_end_io_pagebuf(
 			unsigned int	j, range;
 
 			ASSERT(blocksize < PAGE_CACHE_SIZE);
-			ASSERT(!(pb->pb_flags & _PBF_PRIVATE_BH));
-
-			range = (bvec->bv_offset + bvec->bv_len)>>SECTOR_SHIFT;
-			for (j = bvec->bv_offset>>SECTOR_SHIFT; j < range; j++)
+			range = (bvec->bv_offset + bvec->bv_len) >> sectorshift;
+			for (j = bvec->bv_offset >> sectorshift; j < range; j++)
 				set_bit(j, &page->private);
 			if (page->private == (unsigned long)(PAGE_CACHE_SIZE-1))
 				SetPageUptodate(page);
@@ -1353,7 +1371,7 @@ pagebuf_iorequest(			/* start real I/O		*/
 	int			offset = pb->pb_offset;
 	int			size = pb->pb_count_desired;
 	sector_t		sector = pb->pb_bn;
-	size_t			blocksize = pb->pb_target->pbr_blocksize;
+	unsigned int		blocksize = pb->pb_target->pbr_bsize;
 	int			locking;
 
 	locking = (pb->pb_flags & _PBF_LOCKABLE) == 0 && (pb->pb_locked == 0);
@@ -1382,7 +1400,7 @@ pagebuf_iorequest(			/* start real I/O		*/
 		bio = bio_alloc(GFP_NOIO, 1);
 
 		bio->bi_bdev = pb->pb_target->pbr_bdev;
-		bio->bi_sector = sector - (offset >> SECTOR_SHIFT);
+		bio->bi_sector = sector - (offset >> BBSHIFT);
 		bio->bi_end_io = bio_end_io_pagebuf;
 		bio->bi_private = pb;
 		bio->bi_vcnt++;
@@ -1427,7 +1445,7 @@ pagebuf_iorequest(			/* start real I/O		*/
 
 next_chunk:
 	atomic_inc(&PBP(pb)->pb_io_remaining);
-	nr_pages = BIO_MAX_SECTORS >> (PAGE_SHIFT - SECTOR_SHIFT);
+	nr_pages = BIO_MAX_SECTORS >> (PAGE_SHIFT - BBSHIFT);
 	if (nr_pages > total_nr_pages)
 		nr_pages = total_nr_pages;
 
@@ -1452,7 +1470,7 @@ pagebuf_iorequest(			/* start real I/O		*/
 
 		offset = 0;
 
-		sector += nbytes >> SECTOR_SHIFT;
+		sector += nbytes >> BBSHIFT;
 		size -= nbytes;
 		total_nr_pages--;
 	}

fs/xfs/pagebuf/page_buf.h

@@ -132,11 +132,12 @@ typedef enum page_buf_flags_e {		/* pb_flags values */
 #define PBR_ALIGNED_ONLY 2	/* only use aligned I/O */
 
 typedef struct pb_target {
-	int			pbr_flags;
 	dev_t			pbr_dev;
 	struct block_device	*pbr_bdev;
 	struct address_space	*pbr_mapping;
-	unsigned int		pbr_blocksize;
+	unsigned int		pbr_bsize;
+	unsigned int		pbr_sshift;
+	size_t			pbr_smask;
 } pb_target_t;
 
 /*

fs/xfs/xfs.h

@@ -33,6 +33,7 @@
 #define __XFS_H__
 
 #include <linux/config.h>
+#include <linux/types.h>
 
 #include <xfs_types.h>
 #include <xfs_arch.h>

fs/xfs/xfs_acl.c

@@ -235,7 +235,7 @@ xfs_acl_vget(
 		if (kind == _ACL_TYPE_ACCESS) {
 			vattr_t va;
 
-			va.va_mask = AT_MODE;
+			va.va_mask = XFS_AT_MODE;
 			VOP_GETATTR(vp, &va, 0, sys_cred, error);
 			if (error)
 				goto out;
@@ -372,7 +372,7 @@ xfs_acl_allow_set(
 		return EROFS;
 	if ((error = _MAC_VACCESS(vp, NULL, VWRITE)))
 		return error;
-	va.va_mask = AT_UID;
+	va.va_mask = XFS_AT_UID;
 	VOP_GETATTR(vp, &va, 0, NULL, error);
 	if (error)
 		return error;
@@ -702,7 +702,7 @@ xfs_acl_vtoacl(
 		xfs_acl_get_attr(vp, access_acl, _ACL_TYPE_ACCESS, 0, &error);
 		if (!error) {
 			/* Got the ACL, need the mode... */
-			va.va_mask = AT_MODE;
+			va.va_mask = XFS_AT_MODE;
 			VOP_GETATTR(vp, &va, 0, sys_cred, error);
 		}
 
@@ -800,12 +800,12 @@ xfs_acl_setmode(
 	 * Copy the u::, g::, o::, and m:: bits from the ACL into the
 	 * mode.  The m:: bits take precedence over the g:: bits.
 	 */
-	va.va_mask = AT_MODE;
+	va.va_mask = XFS_AT_MODE;
 	VOP_GETATTR(vp, &va, 0, sys_cred, error);
 	if (error)
 		return error;
 
-	va.va_mask = AT_MODE;
+	va.va_mask = XFS_AT_MODE;
 	va.va_mode &= ~(S_IRWXU|S_IRWXG|S_IRWXO);
 	ap = acl->acl_entry;
 	for (i = 0; i < acl->acl_cnt; ++i) {

fs/xfs/xfs_alloc_btree.h

@@ -96,16 +96,22 @@ int xfs_alloc_block_minrecs(int lev, struct xfs_btree_cur *cur);
 #endif
 
 /*
- * Minimum and maximum blocksize.
+ * Minimum and maximum blocksize and sectorsize.
  * The blocksize upper limit is pretty much arbitrary.
+ * The sectorsize upper limit is due to sizeof(sb_sectsize).
  */
 #define XFS_MIN_BLOCKSIZE_LOG	9	/* i.e. 512 bytes */
 #define XFS_MAX_BLOCKSIZE_LOG	16	/* i.e. 65536 bytes */
 #define XFS_MIN_BLOCKSIZE	(1 << XFS_MIN_BLOCKSIZE_LOG)
 #define XFS_MAX_BLOCKSIZE	(1 << XFS_MAX_BLOCKSIZE_LOG)
+#define XFS_MIN_SECTORSIZE_LOG	9       /* i.e. 512 bytes */
+#define XFS_MAX_SECTORSIZE_LOG	15      /* i.e. 32768 bytes */
+#define XFS_MIN_SECTORSIZE	(1 << XFS_MIN_SECTORSIZE_LOG)
+#define XFS_MAX_SECTORSIZE	(1 << XFS_MAX_SECTORSIZE_LOG)
 
 /*
- * block numbers in the AG; SB is BB 0, AGF is BB 1, AGI is BB 2, AGFL is BB 3
+ * Block numbers in the AG:
+ * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
  */
 #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XFS_BNO_BLOCK)
 xfs_agblock_t xfs_bno_block(struct xfs_mount *mp);

fs/xfs/xfs_cap.c

@@ -192,7 +192,7 @@ xfs_cap_allow_set(
 		return EROFS;
 	if ((error = _MAC_VACCESS(vp, NULL, VWRITE)))
 		return error;
-	va.va_mask = AT_UID;
+	va.va_mask = XFS_AT_UID;
 	VOP_GETATTR(vp, &va, 0, NULL, error);
 	if (error)
 		return error;

fs/xfs/xfs_dir2.h

@@ -32,7 +32,6 @@
 #ifndef __XFS_DIR2_H__
 #define __XFS_DIR2_H__
 
-struct dirent;
 struct uio;
 struct xfs_dabuf;
 struct xfs_da_args;

fs/xfs/xfs_dir2_block.h

@@ -37,7 +37,6 @@
  * Directory version 2, single block format structures
  */
 
-struct dirent;
 struct uio;
 struct xfs_dabuf;
 struct xfs_da_args;

fs/xfs/xfs_dir2_leaf.h

@@ -36,7 +36,6 @@
  * Directory version 2, leaf block structures.
  */
 
-struct dirent;
 struct uio;
 struct xfs_dabuf;
 struct xfs_da_args;

fs/xfs/xfs_dir2_node.h

@@ -36,7 +36,6 @@
  * Directory version 2, btree node format structures
  */
 
-struct dirent;
 struct uio;
 struct xfs_dabuf;
 struct xfs_da_args;

fs/xfs/xfs_dir2_sf.h

@@ -39,7 +39,6 @@
  * fit into the literal area of the inode.
  */
 
-struct dirent;
 struct uio;
 struct xfs_dabuf;
 struct xfs_da_args;

fs/xfs/xfs_dir_leaf.h

@@ -42,7 +42,6 @@
  * internal links in the Btree are logical block offsets into the file.
  */
 
-struct dirent;
 struct uio;
 struct xfs_bmap_free;
 struct xfs_dabuf;

fs/xfs/xfs_fs.h

@@ -30,12 +30,8 @@
  *
  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  */
-#ifndef _LINUX_XFS_FS_H
-#define _LINUX_XFS_FS_H
-
-#include <linux/types.h>
-#include <asm/ioctl.h>
-
+#ifndef __XFS_FS_H__
+#define __XFS_FS_H__
 
 /*
  * SGI's XFS filesystem's major stuff (constants, structures)
@@ -394,11 +390,13 @@ typedef struct {
  * This is typically called by a stateless file server in order to generate
  * "file handles".
  */
+#ifndef MAXFIDSZ
 #define MAXFIDSZ	46
 typedef struct fid {
 	__u16		fid_len;		/* length of data in bytes */
 	unsigned char	fid_data[MAXFIDSZ];	/* data (variable length)  */
 } fid_t;
+#endif
 
 typedef struct xfs_fid {
 	__u16	xfs_fid_len;		/* length of remainder	*/
@@ -499,4 +497,4 @@ typedef struct xfs_handle {
 #define BTOBBT(bytes)	((__u64)(bytes) >> BBSHIFT)
 #define BBTOB(bbs)	((bbs) << BBSHIFT)
 
-#endif	/* _LINUX_XFS_FS_H */
+#endif	/* __XFS_FS_H__ */

fs/xfs/xfs_inode_item.c

@@ -433,7 +433,6 @@ xfs_inode_item_format(
 		ASSERT(!(iip->ili_format.ilf_fields &
 			 (XFS_ILOG_ADATA | XFS_ILOG_ABROOT)));
 		if (iip->ili_format.ilf_fields & XFS_ILOG_AEXT) {
-			ASSERT(!(iip->ili_format.ilf_fields & XFS_ILOG_DEXT));
 			ASSERT(ip->i_afp->if_bytes > 0);
 			ASSERT(ip->i_afp->if_u1.if_extents != NULL);
 			ASSERT(ip->i_d.di_anextents > 0);

fs/xfs/xfs_log_recover.c

@@ -1839,10 +1839,9 @@ xlog_recover_do_dquot_buffer(
 	uint type;
 
 	/*
-	 * Non-root filesystems are required to send in quota flags
-	 * at mount time.
+	 * Filesystems are required to send in quota flags at mount time.
 	 */
-	if (mp->m_qflags == 0 && mp->m_dev != rootdev) {
+	if (mp->m_qflags == 0) {
 		return;
 	}
 	
@@ -2289,30 +2288,28 @@ xlog_recover_do_dquot_trans(xlog_t		*log,
 	mp = log->l_mp;
 
 	/*
-	 * Non-root filesystems are required to send in quota flags
-	 * at mount time.
+	 * Filesystems are required to send in quota flags at mount time.
 	 */
-	if (mp->m_qflags == 0 && mp->m_dev != rootdev) {
+	if (mp->m_qflags == 0)
 		return (0);
-	}
 
 	recddq = (xfs_disk_dquot_t *)item->ri_buf[1].i_addr;
 	ASSERT(recddq);
 	/*
 	 * This type of quotas was turned off, so ignore this record.
 	 */
-	type = INT_GET(recddq->d_flags, ARCH_CONVERT)&(XFS_DQ_USER|XFS_DQ_GROUP);
+	type = INT_GET(recddq->d_flags, ARCH_CONVERT) &
+			(XFS_DQ_USER | XFS_DQ_GROUP);
 	ASSERT(type);
 	if (log->l_quotaoffs_flag & type)
 		return (0);
 
 	/*
-	 * At this point we know that if we are recovering a root filesystem
-	 * then quota was _not_ turned off. Since there is no other flag
-	 * indicate to us otherwise, this must mean that quota's on,
-	 * and the dquot needs to be replayed. Remember that we may not have
-	 * fully recovered the superblock yet, so we can't do the usual trick
-	 * of looking at the SB quota bits.
+	 * At this point we know that quota was _not_ turned off.
+	 * Since the mount flags are not indicating to us otherwise, this
+	 * must mean that quota is on, and the dquot needs to be replayed.
+	 * Remember that we may not have fully recovered the superblock yet,
+	 * so we can't do the usual trick of looking at the SB quota bits.
 	 *
 	 * The other possibility, of course, is that the quota subsystem was
 	 * removed since the last mount - ENOSYS.

fs/xfs/xfs_mount.c

@@ -91,7 +91,8 @@ static struct {
     { offsetof(xfs_sb_t, sb_unit),	 0 },
     { offsetof(xfs_sb_t, sb_width),	 0 },
     { offsetof(xfs_sb_t, sb_dirblklog),	 0 },
-    { offsetof(xfs_sb_t, sb_dummy),	 1 },
+    { offsetof(xfs_sb_t, sb_logsectlog), 0 },
+    { offsetof(xfs_sb_t, sb_logsectsize),0 },
     { offsetof(xfs_sb_t, sb_logsunit),	 0 },
     { sizeof(xfs_sb_t),			 0 }
 };
@@ -119,6 +120,7 @@ xfs_mount_init(void)
 	spinlock_init(&mp->m_freeze_lock, "xfs_freeze");
 	init_sv(&mp->m_wait_unfreeze, SV_DEFAULT, "xfs_freeze", 0);
 	atomic_set(&mp->m_active_trans, 0);
+	mp->m_cxfstype = XFS_CXFS_NOT;
 
 	return mp;
 }	/* xfs_mount_init */
@@ -213,13 +215,26 @@ xfs_mount_validate_sb(
 		return XFS_ERROR(EFSCORRUPTED);
 	}
 
+	if (!sbp->sb_logsectlog)
+		sbp->sb_logsectlog = sbp->sb_sectlog;
+	if (!sbp->sb_logsectsize)
+		sbp->sb_logsectsize = sbp->sb_sectsize;
+
 	/*
 	 * More sanity checking. These were stolen directly from
 	 * xfs_repair.
 	 */
-	if (sbp->sb_blocksize <= 0					||
-	    sbp->sb_agcount <= 0					||
-	    sbp->sb_sectsize <= 0					||
+	if (sbp->sb_agcount <= 0					||
+	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
+	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
+	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
+	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
+	    sbp->sb_logsectsize < XFS_MIN_SECTORSIZE			||
+	    sbp->sb_logsectsize > XFS_MAX_SECTORSIZE			||
+	    sbp->sb_logsectlog < XFS_MIN_SECTORSIZE_LOG			||
+	    sbp->sb_logsectlog > XFS_MAX_SECTORSIZE_LOG			||
+	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
+	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
@@ -232,7 +247,7 @@ xfs_mount_validate_sb(
 	}
 
 	/*
-	 * sanity check ag count, size fields against data size field
+	 * Sanity check AG count, size fields against data size field
 	 */
 	if (sbp->sb_dblocks == 0 ||
 	    sbp->sb_dblocks >
@@ -268,7 +283,8 @@ xfs_mount_validate_sb(
 			PAGE_SIZE);
 		return XFS_ERROR(EWRONGFS);
 	}
-	return (0);
+
+	return 0;
 }
 
 void
@@ -467,6 +483,7 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 	mp->m_maxagi = mp->m_sb.sb_agcount;
 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
+	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
 	mp->m_litino = sbp->sb_inodesize -
@@ -560,7 +577,7 @@ xfs_mountfs(
 	xfs_daddr_t	d;
 	extern xfs_ioops_t xfs_iocore_xfs;	/* from xfs_iocore.c */
 	__uint64_t	ret64;
-	uint		quotaflags, quotaondisk, rootqcheck, needquotacheck;
+	uint		quotaflags, quotaondisk;
 	uint		uquotaondisk = 0, gquotaondisk = 0;
 	boolean_t	needquotamount;
 	__int64_t	update_flags;
@@ -755,7 +772,9 @@ xfs_mountfs(
 		goto error1;
 	}
 	if (!noio) {
-		error = xfs_read_buf(mp, mp->m_ddev_targp, d - 1, 1, 0, &bp);
+		error = xfs_read_buf(mp, mp->m_ddev_targp,
+				     d - XFS_FSS_TO_BB(mp, 1),
+				     XFS_FSS_TO_BB(mp, 1), 0, &bp);
 		if (!error) {
 			xfs_buf_relse(bp);
 		} else {
@@ -775,7 +794,9 @@ xfs_mountfs(
 			error = XFS_ERROR(E2BIG);
 			goto error1;
 		}
-		error = xfs_read_buf(mp, mp->m_logdev_targp, d - 1, 1, 0, &bp);
+		error = xfs_read_buf(mp, mp->m_logdev_targp,
+				     d - XFS_LOGS_TO_BB(mp, 1),
+				     XFS_LOGS_TO_BB(mp, 1), 0, &bp);
 		if (!error) {
 			xfs_buf_relse(bp);
 		} else {
@@ -967,21 +988,13 @@ xfs_mountfs(
 
 	/*
 	 * Figure out if we'll need to do a quotacheck.
-	 * The requirements are a little different depending on whether
-	 * this fs is root or not.
-	 */
-	rootqcheck = (mp->m_dev == rootdev && quotaondisk && 
-		      ((mp->m_sb.sb_qflags & XFS_UQUOTA_ACCT &&
-			(mp->m_sb.sb_qflags & XFS_UQUOTA_CHKD) == 0) ||
-		       (mp->m_sb.sb_qflags & XFS_GQUOTA_ACCT &&
-			(mp->m_sb.sb_qflags & XFS_GQUOTA_CHKD) == 0)));
-	needquotacheck = rootqcheck ||	XFS_QM_NEED_QUOTACHECK(mp);
+	 */
 	if (XFS_IS_QUOTA_ON(mp) || quotaondisk) {
 		/*
 		 * Call mount_quotas at this point only if we won't have to do
 		 * a quotacheck.
 		 */
-		if (quotaondisk && !needquotacheck) {
+		if (quotaondisk && !XFS_QM_NEED_QUOTACHECK(mp)) {
 			/*
 			 * If the xfs quota code isn't installed,
 			 * we have to reset the quotachk'd bit.
@@ -1020,10 +1033,6 @@ xfs_mountfs(
 	if (needquotamount) {
 		ASSERT(mp->m_qflags == 0);
 		mp->m_qflags = quotaflags;
-		rootqcheck = (mp->m_dev == rootdev && needquotacheck);
-		if (rootqcheck && (error = xfs_dev_is_read_only(mp,
-					"quotacheck")))
-			goto error2;
 		if (xfs_qm_mount_quotas(mp))
 			xfs_mount_reset_sbqflags(mp);
 	}

fs/xfs/xfs_mount.h

@@ -232,6 +232,7 @@ typedef struct xfs_mount {
 	__uint8_t		m_mk_sharedro;	/* mark shared ro on unmount */
 	__uint8_t		m_inode_quiesce;/* call quiesce on new inodes.
 						   field governed by m_ilock */
+	__uint8_t		m_sectbb_log;	/* sectlog - BBSHIFT */
 	__uint8_t		m_dirversion;	/* 1 or 2 */
 	xfs_dirops_t		m_dirops;	/* table of dir funcs */
 	int			m_dirblksize;	/* directory block sz--bytes */
@@ -265,7 +266,7 @@ typedef struct xfs_mount {
 #if XFS_BIG_FILESYSTEMS
 #define XFS_MOUNT_INO64		0x00000002
 #endif
-#define XFS_MOUNT_ROOTQCHECK	0x00000004
+			     /* 0x00000004	-- currently unused */
 			     /* 0x00000008	-- currently unused */
 #define XFS_MOUNT_FS_SHUTDOWN	0x00000010	/* atomic stop of all filesystem
 						   operations, typically for

fs/xfs/xfs_qm.c

@@ -301,27 +301,24 @@ xfs_qm_unmount_quotadestroy(
 
 /*
  * This is called from xfs_mountfs to start quotas and initialize all
- * necessary data structures like quotainfo, and in the rootfs's case
- * xfs_Gqm. This is also responsible for running a quotacheck as necessary.
- * We are guaranteed that the superblock is consistently read in at this
- * point.
+ * necessary data structures like quotainfo.  This is also responsible for
+ * running a quotacheck as necessary.  We are guaranteed that the superblock
+ * is consistently read in at this point.
  */
 int
 xfs_qm_mount_quotas(
 	xfs_mount_t	*mp)
 {
 	unsigned long	s;
-	int		error;
+	int		error = 0;
 	uint		sbf;
 
-	error = 0;
 	/*
-	 * If a non-root file system had quotas running earlier, but decided
-	 * to mount without -o quota/pquota options, revoke the quotachecked
-	 * license, and bail out.
+	 * If a file system had quotas running earlier, but decided to
+	 * mount without -o quota/uquota/gquota options, revoke the
+	 * quotachecked license, and bail out.
 	 */
 	if (! XFS_IS_QUOTA_ON(mp) &&
-	    (mp->m_dev != rootdev) &&
 	    (mp->m_sb.sb_qflags & (XFS_UQUOTA_ACCT|XFS_GQUOTA_ACCT))) {
 		mp->m_qflags = 0;
 		goto write_changes;
@@ -342,32 +339,6 @@ xfs_qm_mount_quotas(
 #if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
 	cmn_err(CE_NOTE, "Attempting to turn on disk quotas.");
 #endif
-	/*
-	 * If this is the root file system, mark flags in mount struct first.
-	 * We couldn't do this earlier because we didn't have the superblock
-	 * read in.
-	 */
-	if (mp->m_dev == rootdev) {
-		ASSERT(XFS_SB_VERSION_HASQUOTA(&mp->m_sb));
-		ASSERT(mp->m_sb.sb_qflags &
-		       (XFS_UQUOTA_ACCT|XFS_GQUOTA_ACCT));
-		if (xfs_Gqm == NULL) {
-			if ((xfs_Gqm = xfs_qm_init()) == NULL) {
-				mp->m_qflags = 0;
-				error = EINVAL;
-				goto write_changes;
-			}
-		}
-		mp->m_qflags = mp->m_sb.sb_qflags;
-		if (mp->m_qflags & XFS_UQUOTA_ACCT)
-			mp->m_qflags |= XFS_UQUOTA_ACTIVE;
-		if (mp->m_qflags & XFS_GQUOTA_ACCT)
-			mp->m_qflags |= XFS_GQUOTA_ACTIVE;
-		/*
-		 * The quotainode of the root file system may or may not
-		 * exist at this point.
-		 */
-	}
 
 	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
 	/*
@@ -545,20 +516,6 @@ xfs_qm_dqflush_all(
 		if (error)
 			return (error);
 
-		/*
-		 * If this is the root filesystem doing a quotacheck,
-		 * we should do periodic bflushes. This is because there's
-		 * no bflushd at this point.
-		 */
-		if (mp->m_flags & XFS_MOUNT_ROOTQCHECK) {
-			if (++niters == XFS_QM_MAX_DQCLUSTER_LOGSZ) {
-				xfs_log_force(mp, (xfs_lsn_t)0,
-					      XFS_LOG_FORCE | XFS_LOG_SYNC);
-				XFS_bflush(mp->m_ddev_targp);
-				niters = 0;
-			}
-		}
-
 		xfs_qm_mplist_lock(mp);
 		if (recl != XFS_QI_MPLRECLAIMS(mp)) {
 			xfs_qm_mplist_unlock(mp);
@@ -1568,22 +1525,6 @@ xfs_qm_dqiter_bufs(
 					     flags & XFS_QMOPT_UQUOTA ?
 					     XFS_DQ_USER : XFS_DQ_GROUP);
 		xfs_bdwrite(mp, bp);
-		/*
-		 * When quotachecking the root filesystem,
-		 * we may not have bdflush, and we may fill
-		 * up all available freebufs.
-		 * The workaround here is to push on the
-		 * log and do a bflush on the rootdev
-		 * periodically.
-		 */
-		if (mp->m_flags & XFS_MOUNT_ROOTQCHECK) {
-			if (++notcommitted == incr) {
-				xfs_log_force(mp, (xfs_lsn_t)0,
-					      XFS_LOG_FORCE | XFS_LOG_SYNC);
-				XFS_bflush(mp->m_ddev_targp);
-				notcommitted = 0;
-			}
-		}
 		/*
 		 * goto the next block.
 		 */
@@ -1982,7 +1923,6 @@ xfs_qm_quotacheck(
 
  error_return:
 	cmn_err(CE_NOTE, "XFS quotacheck %s: Done.", mp->m_fsname);
-	mp->m_flags &= ~(XFS_MOUNT_ROOTQCHECK);
 	return (error);
 }
 
@@ -2034,9 +1974,9 @@ xfs_qm_init_quotainos(
 	/*
 	 * Create the two inodes, if they don't exist already. The changes
 	 * made above will get added to a transaction and logged in one of
-	 * the qino_alloc calls below.
+	 * the qino_alloc calls below.  If the device is readonly,
+	 * temporarily switch to read-write to do this.
 	 */
-
 	if (XFS_IS_UQUOTA_ON(mp) && uip == NULL) {
 		if ((error = xfs_qm_qino_alloc(mp, &uip,
 					      sbflags | XFS_SB_UQUOTINO,
@@ -2145,20 +2085,6 @@ xfs_qm_shake_freelist(
 		 */
 		if (XFS_DQ_IS_DIRTY(dqp)) {
 			xfs_dqtrace_entry(dqp, "DQSHAKE: DQDIRTY");
-			/*
-			 * We'll be doing a dqflush, and it is
-			 * possible to fill up the entire buffer cache
-			 * with dirty delayed write buffers when doing
-			 * this on a root filesystem, if bdflush isn't
-			 * running. So, do a flush periodically.
-			 */
-			if (dqp->q_mount->m_flags & XFS_MOUNT_ROOTQCHECK) {
-				if (!(++nflushes % XFS_QM_MAX_DQCLUSTER_LOGSZ)){
-					xfs_log_force(dqp->q_mount, (xfs_lsn_t)0,
-						 XFS_LOG_FORCE | XFS_LOG_SYNC);
-					XFS_bflush(dqp->q_mount->m_ddev_targp);
-				}
-			}
 			/*
 			 * We flush it delayed write, so don't bother
 			 * releasing the mplock.
@@ -2329,21 +2255,6 @@ xfs_qm_dqreclaim_one(void)
 		 */
 		if (XFS_DQ_IS_DIRTY(dqp)) {
 			xfs_dqtrace_entry(dqp, "DQRECLAIM: DQDIRTY");
-			/*
-			 * We'll be doing a dqflush, and it is
-			 * possible to fill up the entire buffer cache
-			 * with dirty delayed write buffers when doing
-			 * this on a root filesystem, if bdflush isn't
-			 * running. So, do a flush periodically.
-			 */
-			if (dqp->q_mount->m_flags & XFS_MOUNT_ROOTQCHECK) {
-				if (!(++nflushes % XFS_QM_MAX_DQCLUSTER_LOGSZ)) {
-					xfs_log_force(dqp->q_mount, (xfs_lsn_t)0,
-						XFS_LOG_FORCE | XFS_LOG_SYNC);
-					XFS_bflush(dqp->q_mount->m_ddev_targp);
-				}
-			}
-
 			/*
 			 * We flush it delayed write, so don't bother
 			 * releasing the freelist lock.

fs/xfs/xfs_qm_syscalls.c

@@ -99,8 +99,6 @@ linvfs_setxstate(
 		return -xfs_qm_scall_quotaon(mp, qflags);
 	case Q_XQUOTAOFF:
 		qflags = xfs_qm_import_flags(flags);
-		if (mp->m_dev == rootdev)
-			return -xfs_qm_scall_quotaoff(mp, qflags, B_FALSE);
 		if (!XFS_IS_QUOTA_ON(mp))
 			return -ESRCH;
 		return -xfs_qm_scall_quotaoff(mp, qflags, B_FALSE);
@@ -169,17 +167,16 @@ xfs_qm_scall_quotaoff(
 	int			error;
 	uint			inactivate_flags;
 	xfs_qoff_logitem_t	*qoffstart;
-	uint			sbflags, newflags;
 	int			nculprits;
 
 	if (!force && !capable(CAP_SYS_ADMIN))
 		return XFS_ERROR(EPERM);
 	/*
-	 * Only root file system can have quotas enabled on disk but not
-	 * in core. Note that quota utilities (like quotaoff) _expect_
+	 * No file system can have quotas enabled on disk but not in core.
+	 * Note that quota utilities (like quotaoff) _expect_
 	 * errno == EEXIST here.
 	 */
-	if (mp->m_dev != rootdev && (mp->m_qflags & flags) == 0)
+	if ((mp->m_qflags & flags) == 0)
 		return XFS_ERROR(EEXIST);
 	error = 0;
 
@@ -191,54 +188,14 @@ xfs_qm_scall_quotaoff(
 	 * critical thing.
 	 * If quotaoff, then we must be dealing with the root filesystem.
 	 */
-	ASSERT(mp->m_quotainfo || mp->m_dev == rootdev);
+	ASSERT(mp->m_quotainfo);
 	if (mp->m_quotainfo)
 		mutex_lock(&(XFS_QI_QOFFLOCK(mp)), PINOD);
 
-	/*
-	 * Root file system may or may not have quotas on in core.
-	 * We have to perform the quotaoff accordingly.
-	 */
-	if (mp->m_dev == rootdev) {
-		s = XFS_SB_LOCK(mp);
-		sbflags = mp->m_sb.sb_qflags;
-		if ((mp->m_qflags & flags) == 0) {
-			mp->m_sb.sb_qflags &= ~(flags);
-			newflags = mp->m_sb.sb_qflags;
-			XFS_SB_UNLOCK(mp, s);
-			if (mp->m_quotainfo)
-				mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
-			if (sbflags != newflags)
-			      error = xfs_qm_write_sb_changes(mp, XFS_SB_QFLAGS);
-			return (error);
-		}
-		XFS_SB_UNLOCK(mp, s);
-
-		if ((sbflags & flags) != (mp->m_qflags & flags)) {
-			/*
-			 * This can happen only with grp+usr quota
-			 * combination. Note: 1) accounting cannot be turned
-			 * off without enforcement also getting turned off.
-			 * 2) Every flag that exists in mpqflags MUST exist
-			 * in sbqflags (but not vice versa).
-			 * which means at this point sbqflags = UQ+GQ+..,
-			 * and mpqflags = UQ or GQ.
-			 */
-			ASSERT(sbflags & XFS_GQUOTA_ACCT);
-			ASSERT((sbflags & XFS_ALL_QUOTA_ACCT) !=
-			       (mp->m_qflags & XFS_ALL_QUOTA_ACCT));
-
-			qdprintk("quotaoff, sbflags=%x flags=%x m_qflags=%x\n",
-				sbflags, flags, mp->m_qflags);
-			/* XXX TBD Finish this for group quota support */
-			/* We need to update the SB and mp separately */
-			return XFS_ERROR(EINVAL);
-		}
-	}
 	ASSERT(mp->m_quotainfo);
 
 	/*
-	 * if we're just turning off quota enforcement, change mp and go.
+	 * If we're just turning off quota enforcement, change mp and go.
 	 */
 	if ((flags & XFS_ALL_QUOTA_ACCT) == 0) {
 		mp->m_qflags &= ~(flags);
@@ -272,8 +229,8 @@ xfs_qm_scall_quotaoff(
 	}
 
 	/*
-	 * Nothing to do? Don't complain.
-	 * This happens when we're just turning off quota enforcement.
+	 * Nothing to do?  Don't complain. This happens when we're just
+	 * turning off quota enforcement.
 	 */
 	if ((mp->m_qflags & flags) == 0) {
 		mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
@@ -326,9 +283,8 @@ xfs_qm_scall_quotaoff(
 	 * So, if we couldn't purge all the dquots from the filesystem,
 	 * we can't get rid of the incore data structures.
 	 */
-	while ((nculprits = xfs_qm_dqpurge_all(mp, dqtype|XFS_QMOPT_QUOTAOFF))) {
+	while ((nculprits = xfs_qm_dqpurge_all(mp, dqtype|XFS_QMOPT_QUOTAOFF)))
 		delay(10 * nculprits);
-	}
 
 	/*
 	 * Transactions that had started before ACTIVE state bit was cleared
@@ -406,11 +362,9 @@ xfs_qm_scall_trunc_qfiles(
 
 
 /*
- * This does two separate functions:
- * Switch on quotas for the root file system. This will take effect only
- * on reboot.
- * Switch on (a given) quota enforcement for both root and non-root filesystems.
- * This takes effect immediately.
+ * Switch on (a given) quota enforcement for a filesystem.  This takes
+ * effect immediately.
+ * (Switching on quota accounting must be done at mount time.)
  */
 STATIC int
 xfs_qm_scall_quotaon(
@@ -422,37 +376,26 @@ xfs_qm_scall_quotaon(
 	uint		qf;
 	uint		accflags;
 	__int64_t	sbflags;
-	boolean_t	rootfs;
-	boolean_t	delay;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return XFS_ERROR(EPERM);
 
-	rootfs = (boolean_t) (mp->m_dev == rootdev);
-
 	flags &= (XFS_ALL_QUOTA_ACCT | XFS_ALL_QUOTA_ENFD);
 	/*
-	 * If caller wants to turn on accounting on /, but accounting
-	 * is already turned on, ignore ACCTing flags.
-	 * Switching on quota accounting for non-root filesystems
-	 * must be done at mount time.
+	 * Switching on quota accounting must be done at mount time.
 	 */
 	accflags = flags & XFS_ALL_QUOTA_ACCT;
-	if (!rootfs ||
-	    (accflags && rootfs && ((mp->m_qflags & accflags) == accflags))) {
 	flags &= ~(XFS_ALL_QUOTA_ACCT);
-	}
 
 	sbflags = 0;
-	delay = (boolean_t) ((flags & XFS_ALL_QUOTA_ACCT) != 0);
 
 	if (flags == 0) {
 		qdprintk("quotaon: zero flags, m_qflags=%x\n", mp->m_qflags);
 		return XFS_ERROR(EINVAL);
 	}
 
-	/* Only rootfs can turn on quotas with a delayed effect */
-	ASSERT(!delay || rootfs);
+	/* No fs can turn on quotas with a delayed effect */
+	ASSERT((flags & XFS_ALL_QUOTA_ACCT) == 0);
 
 	/*
 	 * Can't enforce without accounting. We check the superblock
@@ -476,22 +419,6 @@ xfs_qm_scall_quotaon(
 	if ((mp->m_qflags & flags) == flags)
 		return XFS_ERROR(EEXIST);
 
-	/*
-	 * Change superblock version (if needed) for the root filesystem
-	 */
-	if (rootfs && !XFS_SB_VERSION_HASQUOTA(&mp->m_sb)) {
-		qdprintk("Old superblock version %x\n", mp->m_sb.sb_versionnum);
-		s = XFS_SB_LOCK(mp);
-		XFS_SB_VERSION_ADDQUOTA(&mp->m_sb);
-		mp->m_sb.sb_uquotino = NULLFSINO;
-		mp->m_sb.sb_gquotino = NULLFSINO;
-		mp->m_sb.sb_qflags = 0;
-		XFS_SB_UNLOCK(mp, s);
-		qdprintk("Converted to version %x\n", mp->m_sb.sb_versionnum);
-		sbflags |= (XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO |
-			    XFS_SB_GQUOTINO | XFS_SB_QFLAGS);
-	}
-
 	/*
 	 * Change sb_qflags on disk but not incore mp->qflags
 	 * if this is the root filesystem.
@@ -511,11 +438,9 @@ xfs_qm_scall_quotaon(
 	if ((error = xfs_qm_write_sb_changes(mp, sbflags)))
 		return (error);
 	/*
-	 * If we had just turned on quotas (ondisk) for rootfs, or if we aren't
-	 * trying to switch on quota enforcement, we are done.
+	 * If we aren't trying to switch on quota enforcement, we are done.
 	 */
-	if (delay ||
-	    ((mp->m_sb.sb_qflags & XFS_UQUOTA_ACCT) !=
+	if  (((mp->m_sb.sb_qflags & XFS_UQUOTA_ACCT) !=
 	     (mp->m_qflags & XFS_UQUOTA_ACCT)) ||
 	    (flags & XFS_ALL_QUOTA_ENFD) == 0)
 		return (0);
@@ -524,8 +449,7 @@ xfs_qm_scall_quotaon(
 		return XFS_ERROR(ESRCH);
 
 	/*
-	 * Switch on quota enforcement in core. This applies to both root
-	 * and non-root file systems.
+	 * Switch on quota enforcement in core.
 	 */
 	mutex_lock(&(XFS_QI_QOFFLOCK(mp)), PINOD);
 	mp->m_qflags |= (flags & XFS_ALL_QUOTA_ENFD);
@@ -546,7 +470,6 @@ xfs_qm_scall_getqstat(
 {
 	xfs_inode_t	*uip, *gip;
 	boolean_t	tempuqip, tempgqip;
-	__uint16_t	sbflags;
 
 	uip = gip = NULL;
 	tempuqip = tempgqip = B_FALSE;
@@ -561,20 +484,6 @@ xfs_qm_scall_getqstat(
 	out->qs_flags = (__uint16_t) xfs_qm_export_flags(mp->m_qflags &
 							(XFS_ALL_QUOTA_ACCT|
 							 XFS_ALL_QUOTA_ENFD));
-	/*
-	 * If the qflags are different on disk, as can be the case when
-	 * root filesystem's quotas are being turned on, return them in the
-	 * HI 8 bits.
-	 */
-	if (mp->m_dev == rootdev) {
-		sbflags = (__uint16_t) xfs_qm_export_flags(mp->m_sb.sb_qflags &
-							   (XFS_ALL_QUOTA_ACCT|
-							    XFS_ALL_QUOTA_ENFD));
-		ASSERT((out->qs_flags & 0xff00) == 0);
-		if (sbflags != out->qs_flags)
-			out->qs_flags |= ((sbflags & 0x00ff) << 8);
-	}
-
 	out->qs_pad = 0;
 	out->qs_uquota.qfs_ino = mp->m_sb.sb_uquotino;
 	out->qs_gquota.qfs_ino = mp->m_sb.sb_gquotino;

fs/xfs/xfs_rename.c

@@ -78,8 +78,8 @@ STATIC int
 xfs_lock_for_rename(
 	xfs_inode_t	*dp1,	/* old (source) directory inode */
 	xfs_inode_t	*dp2,	/* new (target) directory inode */
-	struct dentry	*dentry1, /* old entry name */
-	struct dentry	*dentry2, /* new entry name */
+	vname_t		*dentry1, /* old entry name */
+	vname_t		*dentry2, /* new entry name */
 	xfs_inode_t	**ipp1, /* inode of old entry */
 	xfs_inode_t	**ipp2, /* inode of new entry, if it
 				   already exists, NULL otherwise. */
@@ -224,9 +224,9 @@ int xfs_renames;
 int
 xfs_rename(
 	bhv_desc_t	*src_dir_bdp,
-	struct dentry	*src_dentry,
+	vname_t		*src_dentry,
 	vnode_t		*target_dir_vp,
-	struct dentry	*target_dentry,
+	vname_t		*target_dentry,
 	cred_t		*credp)
 {
 	xfs_trans_t	*tp;
@@ -246,8 +246,8 @@ xfs_rename(
 	int		spaceres;
 	int		target_link_zero = 0;
 	int		num_inodes;
-	char		*src_name = (char *)src_dentry->d_name.name;
-	char		*target_name = (char *)target_dentry->d_name.name;
+	char		*src_name = VNAME(src_dentry);
+	char		*target_name = VNAME(target_dentry);
 	int		src_namelen;
 	int		target_namelen;
 #ifdef DEBUG
@@ -268,10 +268,10 @@ xfs_rename(
 	if (target_dir_bdp == NULL) {
 		return XFS_ERROR(EXDEV);
 	}
-	src_namelen = src_dentry->d_name.len;
+	src_namelen = VNAMELEN(src_dentry);
 	if (src_namelen >= MAXNAMELEN)
 		return XFS_ERROR(ENAMETOOLONG);
-	target_namelen = target_dentry->d_name.len;
+	target_namelen = VNAMELEN(target_dentry);
 	if (target_namelen >= MAXNAMELEN)
 		return XFS_ERROR(ENAMETOOLONG);
 	src_dp = XFS_BHVTOI(src_dir_bdp);

fs/xfs/xfs_sb.h

@@ -34,7 +34,7 @@
 
 /*
  * Super block
- * Fits into a 512-byte buffer at daddr_t 0 of each allocation group.
+ * Fits into a sector-sized buffer at address 0 of each allocation group.
  * Only the first of these is ever updated except during growfs.
  */
 
@@ -140,7 +140,8 @@ typedef struct xfs_sb
 	__uint32_t	sb_unit;	/* stripe or raid unit */
 	__uint32_t	sb_width;	/* stripe or raid width */
 	__uint8_t	sb_dirblklog;	/* log2 of dir block size (fsbs) */
-	__uint8_t	sb_dummy[3];	/* padding */
+	__uint8_t	sb_logsectlog;	/* log2 of the log sector size */
+	__uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
 	__uint32_t	sb_logsunit;	/* stripe unit size for the log */
 } xfs_sb_t;
 
@@ -159,7 +160,7 @@ typedef enum {
 	XFS_SBS_IFREE, XFS_SBS_FDBLOCKS, XFS_SBS_FREXTENTS, XFS_SBS_UQUOTINO,
 	XFS_SBS_GQUOTINO, XFS_SBS_QFLAGS, XFS_SBS_FLAGS, XFS_SBS_SHARED_VN,
 	XFS_SBS_INOALIGNMT, XFS_SBS_UNIT, XFS_SBS_WIDTH, XFS_SBS_DIRBLKLOG,
-	XFS_SBS_DUMMY, XFS_SBS_LOGSUNIT,
+	XFS_SBS_LOGSECTLOG, XFS_SBS_LOGSECTSIZE, XFS_SBS_LOGSUNIT,
 	XFS_SBS_FIELDCOUNT
 } xfs_sb_field_t;
 
@@ -474,6 +475,28 @@ xfs_daddr_t xfs_fsb_to_daddr(struct xfs_mount *mp, xfs_fsblock_t fsbno);
 			 XFS_FSB_TO_AGBNO(mp,fsbno))
 #endif
 
+#if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XFS_BUF_TO_SBP)
+xfs_sb_t *xfs_buf_to_sbp(struct xfs_buf *bp);
+#define XFS_BUF_TO_SBP(bp)	xfs_buf_to_sbp(bp)
+#else
+#define XFS_BUF_TO_SBP(bp)	((xfs_sb_t *)XFS_BUF_PTR(bp))
+#endif
+
+/*
+ * File system sector to basic block conversions.
+ */
+#define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
+#define XFS_LOGS_TO_BB(mp,sec)	((sec) << ((mp)->m_sb.sb_logsectlog - BBSHIFT))
+#define XFS_BB_TO_FSS(mp,bb)	\
+	(((bb) + (XFS_FSS_TO_BB(mp,1) - 1)) >> (mp)->m_sectbb_log)
+#define XFS_BB_TO_FSST(mp,bb)	((bb) >> (mp)->m_sectbb_log)
+
+/*
+ * File system sector to byte conversions.
+ */
+#define XFS_FSS_TO_B(mp,sectno)	((xfs_fsize_t)(sectno) << (mp)->m_sb.sb_sectlog)
+#define XFS_B_TO_FSST(mp,b)	(((__uint64_t)(b)) >> (mp)->m_sb.sb_sectlog)
+
 /*
  * File system block to basic block conversions.
  */
@@ -493,11 +516,4 @@ xfs_daddr_t xfs_fsb_to_daddr(struct xfs_mount *mp, xfs_fsblock_t fsbno);
 #define XFS_B_TO_FSBT(mp,b)	(((__uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
 #define XFS_B_FSB_OFFSET(mp,b)	((b) & (mp)->m_blockmask)
 
-#if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XFS_BUF_TO_SBP)
-xfs_sb_t *xfs_buf_to_sbp(struct xfs_buf *bp);
-#define XFS_BUF_TO_SBP(bp)	xfs_buf_to_sbp(bp)
-#else
-#define XFS_BUF_TO_SBP(bp)	((xfs_sb_t *)XFS_BUF_PTR(bp))
-#endif
-
 #endif	/* __XFS_SB_H__ */

fs/xfs/xfs_types.h

@@ -151,7 +151,7 @@ typedef __uint64_t	xfs_fileoff_t;	/* block number in a file */
 typedef __int64_t	xfs_sfiloff_t;	/* signed block number in a file */
 typedef __uint64_t	xfs_filblks_t;	/* number of blocks in a file */
 
-typedef __uint8_t	xfs_arch_t;	/* architecutre of an xfs fs */
+typedef __uint8_t	xfs_arch_t;	/* architecture of an xfs fs */
 
 /*
  * Null values for the types.
@@ -195,119 +195,6 @@ typedef enum {
 } xfs_btnum_t;
 
 
-#if defined(CONFIG_PROC_FS) && defined(__KERNEL__) && !defined(XFS_STATS_OFF)
-/*
- * XFS global statistics
- */
-struct xfsstats {
-# define XFSSTAT_END_EXTENT_ALLOC	4
-	__uint32_t		xs_allocx;
-	__uint32_t		xs_allocb;
-	__uint32_t		xs_freex;
-	__uint32_t		xs_freeb;
-# define XFSSTAT_END_ALLOC_BTREE	(XFSSTAT_END_EXTENT_ALLOC+4)
-	__uint32_t		xs_abt_lookup;
-	__uint32_t		xs_abt_compare;
-	__uint32_t		xs_abt_insrec;
-	__uint32_t		xs_abt_delrec;
-# define XFSSTAT_END_BLOCK_MAPPING	(XFSSTAT_END_ALLOC_BTREE+7)
-	__uint32_t		xs_blk_mapr;
-	__uint32_t		xs_blk_mapw;
-	__uint32_t		xs_blk_unmap;
-	__uint32_t		xs_add_exlist;
-	__uint32_t		xs_del_exlist;
-	__uint32_t		xs_look_exlist;
-	__uint32_t		xs_cmp_exlist;
-# define XFSSTAT_END_BLOCK_MAP_BTREE	(XFSSTAT_END_BLOCK_MAPPING+4)
-	__uint32_t		xs_bmbt_lookup;
-	__uint32_t		xs_bmbt_compare;
-	__uint32_t		xs_bmbt_insrec;
-	__uint32_t		xs_bmbt_delrec;
-# define XFSSTAT_END_DIRECTORY_OPS	(XFSSTAT_END_BLOCK_MAP_BTREE+4)
-	__uint32_t		xs_dir_lookup;
-	__uint32_t		xs_dir_create;
-	__uint32_t		xs_dir_remove;
-	__uint32_t		xs_dir_getdents;
-# define XFSSTAT_END_TRANSACTIONS	(XFSSTAT_END_DIRECTORY_OPS+3)
-	__uint32_t		xs_trans_sync;
-	__uint32_t		xs_trans_async;
-	__uint32_t		xs_trans_empty;
-# define XFSSTAT_END_INODE_OPS		(XFSSTAT_END_TRANSACTIONS+7)
-	__uint32_t		xs_ig_attempts;
-	__uint32_t		xs_ig_found;
-	__uint32_t		xs_ig_frecycle;
-	__uint32_t		xs_ig_missed;
-	__uint32_t		xs_ig_dup;
-	__uint32_t		xs_ig_reclaims;
-	__uint32_t		xs_ig_attrchg;
-# define XFSSTAT_END_LOG_OPS		(XFSSTAT_END_INODE_OPS+5)
-	__uint32_t		xs_log_writes;
-	__uint32_t		xs_log_blocks;
-	__uint32_t		xs_log_noiclogs;
-	__uint32_t		xs_log_force;
-	__uint32_t		xs_log_force_sleep;
-# define XFSSTAT_END_TAIL_PUSHING	(XFSSTAT_END_LOG_OPS+10)
-	__uint32_t		xs_try_logspace;
-	__uint32_t		xs_sleep_logspace;
-	__uint32_t		xs_push_ail;
-	__uint32_t		xs_push_ail_success;
-	__uint32_t		xs_push_ail_pushbuf;
-	__uint32_t		xs_push_ail_pinned;
-	__uint32_t		xs_push_ail_locked;
-	__uint32_t		xs_push_ail_flushing;
-	__uint32_t		xs_push_ail_restarts;
-	__uint32_t		xs_push_ail_flush;
-# define XFSSTAT_END_WRITE_CONVERT	(XFSSTAT_END_TAIL_PUSHING+2)
-	__uint32_t		xs_xstrat_quick;
-	__uint32_t		xs_xstrat_split;
-# define XFSSTAT_END_READ_WRITE_OPS	(XFSSTAT_END_WRITE_CONVERT+2)
-	__uint32_t		xs_write_calls;
-	__uint32_t		xs_read_calls;
-# define XFSSTAT_END_ATTRIBUTE_OPS	(XFSSTAT_END_READ_WRITE_OPS+4)
-	__uint32_t		xs_attr_get;
-	__uint32_t		xs_attr_set;
-	__uint32_t		xs_attr_remove;
-	__uint32_t		xs_attr_list;
-# define XFSSTAT_END_QUOTA_OPS		(XFSSTAT_END_ATTRIBUTE_OPS+8)
-	__uint32_t		xs_qm_dqreclaims;
-	__uint32_t		xs_qm_dqreclaim_misses;
-	__uint32_t		xs_qm_dquot_dups;
-	__uint32_t		xs_qm_dqcachemisses;
-	__uint32_t		xs_qm_dqcachehits;
-	__uint32_t		xs_qm_dqwants;
-	__uint32_t		xs_qm_dqshake_reclaims;
-	__uint32_t		xs_qm_dqinact_reclaims;
-# define XFSSTAT_END_INODE_CLUSTER	(XFSSTAT_END_QUOTA_OPS+3)
-	__uint32_t		xs_iflush_count;
-	__uint32_t		xs_icluster_flushcnt;
-	__uint32_t		xs_icluster_flushinode;
-# define XFSSTAT_END_VNODE_OPS		(XFSSTAT_END_INODE_CLUSTER+8)
-	__uint32_t		vn_active;	/* # vnodes not on free lists */
-	__uint32_t		vn_alloc;	/* # times vn_alloc called */
-	__uint32_t		vn_get;		/* # times vn_get called */
-	__uint32_t		vn_hold;	/* # times vn_hold called */
-	__uint32_t		vn_rele;	/* # times vn_rele called */
-	__uint32_t		vn_reclaim;	/* # times vn_reclaim called */
-	__uint32_t		vn_remove;	/* # times vn_remove called */
-	__uint32_t		vn_free;	/* # times vn_free called */
-/* Extra precision counters */
-	__uint64_t		xs_xstrat_bytes;
-	__uint64_t		xs_write_bytes;
-	__uint64_t		xs_read_bytes;
-};
-
-extern struct xfsstats xfsstats;
-
-# define XFS_STATS_INC(count)		( (count)++ )
-# define XFS_STATS_DEC(count)		( (count)-- )
-# define XFS_STATS_ADD(count, inc)	( (count) += (inc) )
-#else	/* !CONFIG_PROC_FS */
-# define XFS_STATS_INC(count)
-# define XFS_STATS_DEC(count)
-# define XFS_STATS_ADD(count, inc)
-#endif	/* !CONFIG_PROC_FS */
-
-
 /*
  * Juggle IRIX device numbers - still used in ondisk structures
  */

fs/xfs/xfs_utils.c

@@ -45,7 +45,7 @@ struct xfsstats xfsstats;
  */
 int
 xfs_get_dir_entry(
-	struct dentry		*dentry,
+	vname_t		*dentry,
 	xfs_inode_t	**ipp)
 {
 	vnode_t		*vp;
@@ -68,7 +68,7 @@ int
 xfs_dir_lookup_int(
 	bhv_desc_t	*dir_bdp,
 	uint		lock_mode,
-	struct dentry		*dentry,
+	vname_t		*dentry,
 	xfs_ino_t	*inum,
 	xfs_inode_t	**ipp)
 {
@@ -77,13 +77,12 @@ xfs_dir_lookup_int(
 	int		error;
 
 	dir_vp = BHV_TO_VNODE(dir_bdp);
-	vn_trace_entry(dir_vp, "xfs_dir_lookup_int",
-		       (inst_t *)__return_address);
+	vn_trace_entry(dir_vp, __FUNCTION__, (inst_t *)__return_address);
 
 	dp = XFS_BHVTOI(dir_bdp);
 
 	error = XFS_DIR_LOOKUP(dp->i_mount, NULL, dp,
-			(char *)dentry->d_name.name, dentry->d_name.len, inum);
+				VNAME(dentry), VNAMELEN(dentry), inum);
 	if (!error) {
 		/*
 		 * Unlock the directory. We do this because we can't

fs/xfs/xfs_utils.h

@@ -37,66 +37,16 @@
 #define ITRACE(ip)	vn_trace_ref(XFS_ITOV(ip), __FILE__, __LINE__, \
 				(inst_t *)__return_address)
 
-struct bhv_desc;
-struct cred;
-struct vnode;
-struct xfs_inode;
-struct xfs_mount;
-struct xfs_trans;
-
-extern int
-xfs_rename(
-	struct bhv_desc *src_dir_bdp,
-	struct dentry	*src_dentry,
-	struct vnode	*target_dir_vp,
-	struct dentry	*target_dentry,
-	struct cred	*credp);
-
-extern int
-xfs_get_dir_entry(
-	struct dentry		*dentry,
-	xfs_inode_t		**ipp);
-
-extern int
-xfs_dir_lookup_int(
-	struct bhv_desc		*dir_bdp,
-	uint			lock_mode,
-	struct dentry		*dentry,
-	xfs_ino_t		*inum,
-	struct xfs_inode	**ipp);
-
-extern int
-xfs_truncate_file(
-	struct xfs_mount	*mp,
-	struct xfs_inode	*ip);
-
-extern int
-xfs_dir_ialloc(
-	struct xfs_trans	**tpp,
-	struct xfs_inode	*dp,
-	mode_t			mode,
-	nlink_t			nlink,
-	xfs_dev_t		rdev,
-	struct cred		*credp,
-	prid_t			prid,
-	int			okalloc,
-	struct xfs_inode	**ipp,
-	int			*committed);
-
-extern int
-xfs_droplink(
-	struct xfs_trans	*tp,
-	struct xfs_inode	*ip);
-
-extern int
-xfs_bumplink(
-	struct xfs_trans	*tp,
-	struct xfs_inode	*ip);
-
-extern void
-xfs_bump_ino_vers2(
-	struct xfs_trans	*tp,
-	struct xfs_inode	*ip);
-
-#endif /* XFS_UTILS_H */
-
+extern int xfs_rename (bhv_desc_t *, vname_t *, vnode_t *, vname_t *, cred_t *);
+extern int xfs_get_dir_entry (vname_t *, xfs_inode_t **);
+extern int xfs_dir_lookup_int (bhv_desc_t *, uint, vname_t *, xfs_ino_t *,
+				xfs_inode_t **);
+extern int xfs_truncate_file (xfs_mount_t *, xfs_inode_t *);
+extern int xfs_dir_ialloc (xfs_trans_t **, xfs_inode_t *, mode_t, nlink_t,
+				xfs_dev_t, cred_t *, prid_t, int,
+				xfs_inode_t **, int *);
+extern int xfs_droplink (xfs_trans_t *, xfs_inode_t *);
+extern int xfs_bumplink (xfs_trans_t *, xfs_inode_t *);
+extern void xfs_bump_ino_vers2 (xfs_trans_t *, xfs_inode_t *);
+
+#endif	/* __XFS_UTILS_H__ */

fs/xfs/xfs_vfsops.c

@@ -393,7 +393,7 @@ xfs_mount(
 	xfs_mount_t		*mp;
 	struct block_device	*ddev, *logdev, *rtdev;
 	int			ronly = (vfsp->vfs_flag & VFS_RDONLY);
-	int			error = 0;
+	int			flags = 0, error;
 
 	ddev = vfsp->vfs_super->s_bdev;
 	logdev = rtdev = NULL;
@@ -430,16 +430,10 @@ xfs_mount(
 	vfs_insertbhv(vfsp, &mp->m_bhv, &xfs_vfsops, mp);
 
 	mp->m_ddev_targp = xfs_alloc_buftarg(ddev);
-	if (rtdev != NULL) {
+	if (rtdev)
 		mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev);
-		set_blocksize(rtdev, 512);
-	}
-	if (logdev != NULL && logdev != ddev) {
-		mp->m_logdev_targp = xfs_alloc_buftarg(logdev);
-		set_blocksize(logdev, 512);
-	} else {
-		mp->m_logdev_targp = mp->m_ddev_targp;
-	}
+	mp->m_logdev_targp = (logdev && logdev != ddev) ?
+				xfs_alloc_buftarg(logdev) : mp->m_ddev_targp;
 
 	error = xfs_start_flags(args, mp, ronly);
 	if (error)
@@ -455,16 +449,16 @@ xfs_mount(
 		goto error;
 	}
 
-	mp->m_ddev_targp->pbr_blocksize = mp->m_sb.sb_blocksize;
-	if (logdev != 0 && logdev != ddev) {
-		mp->m_logdev_targp->pbr_blocksize = mp->m_sb.sb_blocksize;
-	}
-	if (rtdev != 0) {
-		mp->m_rtdev_targp->pbr_blocksize = mp->m_sb.sb_blocksize;
-	}
+	xfs_size_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
+			 mp->m_sb.sb_sectsize);
+	if (logdev && logdev != ddev)
+		xfs_size_buftarg(mp->m_logdev_targp, mp->m_sb.sb_blocksize,
+				 mp->m_sb.sb_logsectsize);
+	if (rtdev)
+		xfs_size_buftarg(mp->m_logdev_targp, mp->m_sb.sb_blocksize,
+				 mp->m_sb.sb_blocksize);
 
-	mp->m_cxfstype = XFS_CXFS_NOT;
-	error = xfs_mountfs(vfsp, mp, ddev->bd_dev, 0);
+	error = xfs_mountfs(vfsp, mp, ddev->bd_dev, flags);
 	if (error)
 		goto error;
 	return 0;