Merge bk://jfs.bkbits.net/linux-2.5

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

fs/jfs/file.c

@@ -19,6 +19,7 @@
 
 #include <linux/fs.h>
 #include "jfs_incore.h"
+#include "jfs_dmap.h"
 #include "jfs_txnmgr.h"
 #include "jfs_xattr.h"
 #include "jfs_debug.h"
@@ -94,6 +95,47 @@ static void jfs_truncate(struct inode *ip)
 	IWRITE_UNLOCK(ip);
 }
 
+static int jfs_open(struct inode *inode, struct file *file)
+{
+	int rc;
+
+	if ((rc = generic_file_open(inode, file)))
+		return rc;
+
+	/*
+	 * We attempt to allow only one "active" file open per aggregate
+	 * group.  Otherwise, appending to files in parallel can cause
+	 * fragmentation within the files.
+	 *
+	 * If the file is empty, it was probably just created and going
+	 * to be written to.  If it has a size, we'll hold off until the
+	 * file is actually grown.
+	 */
+	if (S_ISREG(inode->i_mode) && file->f_mode & FMODE_WRITE &&
+	    (inode->i_size == 0)) {
+		struct jfs_inode_info *ji = JFS_IP(inode);
+		if (ji->active_ag == -1) {
+			ji->active_ag = ji->agno;
+			atomic_inc(
+			    &JFS_SBI(inode->i_sb)->bmap->db_active[ji->agno]);
+		}
+	}
+
+	return 0;
+}
+static int jfs_release(struct inode *inode, struct file *file)
+{
+	struct jfs_inode_info *ji = JFS_IP(inode);
+
+	if (ji->active_ag != -1) {
+		struct bmap *bmap = JFS_SBI(inode->i_sb)->bmap;
+		atomic_dec(&bmap->db_active[ji->active_ag]);
+		ji->active_ag = -1;
+	}
+
+	return 0;
+}
+
 struct inode_operations jfs_file_inode_operations = {
 	.truncate	= jfs_truncate,
 	.setxattr	= jfs_setxattr,
@@ -103,7 +145,7 @@ struct inode_operations jfs_file_inode_operations = {
 };
 
 struct file_operations jfs_file_operations = {
-	.open		= generic_file_open,
+	.open		= jfs_open,
 	.llseek		= generic_file_llseek,
 	.write		= generic_file_write,
 	.read		= generic_file_read,
@@ -112,4 +154,5 @@ struct file_operations jfs_file_operations = {
 	.writev		= generic_file_writev,
  	.sendfile	= generic_file_sendfile,
 	.fsync		= jfs_fsync,
+	.release	= jfs_release,
 };

fs/jfs/jfs_dmap.h

@@ -227,6 +227,7 @@ struct bmap {
 	struct dbmap db_bmap;	/* on-disk aggregate map descriptor */
 	struct inode *db_ipbmap;	/* ptr to aggregate map incore inode */
 	struct semaphore db_bmaplock;	/* aggregate map lock */
+	atomic_t db_active[MAXAG];	/* count of active, open files in AG */
 	u32 *db_DBmap;
 };
 

fs/jfs/jfs_extent.c

@@ -514,9 +514,12 @@ int extFill(struct inode *ip, xad_t * xp)
 static int
 extBalloc(struct inode *ip, s64 hint, s64 * nblocks, s64 * blkno)
 {
+	struct jfs_inode_info *ji = JFS_IP(ip);
+	struct jfs_sb_info *sbi = JFS_SBI(ip->i_sb);
 	s64 nb, nblks, daddr, max;
-	int rc, nbperpage = JFS_SBI(ip->i_sb)->nbperpage;
-	struct bmap *mp = JFS_SBI(ip->i_sb)->bmap;
+	int rc, nbperpage = sbi->nbperpage;
+	struct bmap *bmp = sbi->bmap;
+	int ag;
 
 	/* get the number of blocks to initially attempt to allocate.
 	 * we'll first try the number of blocks requested unless this
@@ -524,7 +527,7 @@ extBalloc(struct inode *ip, s64 hint, s64 * nblocks, s64 * blkno)
 	 * blocks in the map. in that case, we'll start off with the 
 	 * maximum free.
 	 */
-	max = (s64) 1 << mp->db_maxfreebud;
+	max = (s64) 1 << bmp->db_maxfreebud;
 	if (*nblocks >= max && *nblocks > nbperpage)
 		nb = nblks = (max > nbperpage) ? max : nbperpage;
 	else
@@ -549,6 +552,18 @@ extBalloc(struct inode *ip, s64 hint, s64 * nblocks, s64 * blkno)
 	*nblocks = nb;
 	*blkno = daddr;
 
+	if (S_ISREG(ip->i_mode) && (ji->fileset == FILESYSTEM_I)) {
+		ag = BLKTOAG(daddr, sbi);
+		if (ji->active_ag == -1) {
+			atomic_inc(&bmp->db_active[ag]);
+			ji->active_ag = ag;
+		} else if (ji->active_ag != ag) {
+			atomic_dec(&bmp->db_active[ji->active_ag]);
+			atomic_inc(&bmp->db_active[ag]);
+			ji->active_ag = ag;
+		}
+	}
+
 	return (0);
 }
 

fs/jfs/jfs_imap.c

@@ -429,6 +429,7 @@ int diRead(struct inode *ip)
 
 	/* set the ag for the inode */
 	JFS_IP(ip)->agno = BLKTOAG(agstart, sbi);
+	JFS_IP(ip)->active_ag = -1;
 
 	return (rc);
 }
@@ -1358,6 +1359,7 @@ diInitInode(struct inode *ip, int iagno, int ino, int extno, struct iag * iagp)
 	DBG_DIALLOC(JFS_IP(ipimap)->i_imap, ip->i_ino);
 	jfs_ip->ixpxd = iagp->inoext[extno];
 	jfs_ip->agno = BLKTOAG(le64_to_cpu(iagp->agstart), sbi);
+	jfs_ip->active_ag = -1;
 }
 
 
@@ -1413,6 +1415,21 @@ int diAlloc(struct inode *pip, boolean_t dir, struct inode *ip)
 	 * moving backward on the disk.)  compute the hint within the
 	 * file system and the iag.
 	 */
+
+	/* get the ag number of this iag */
+	agno = JFS_IP(pip)->agno;
+
+	if (atomic_read(&JFS_SBI(pip->i_sb)->bmap->db_active[agno])) {
+		/*
+		 * There is an open file actively growing.  We want to
+		 * allocate new inodes from a different ag to avoid
+		 * fragmentation problems.
+		 */
+		agno = dbNextAG(JFS_SBI(pip->i_sb)->ipbmap);
+		AG_LOCK(imap, agno);
+		goto tryag;
+	}
+
 	inum = pip->i_ino + 1;
 	ino = inum & (INOSPERIAG - 1);
 
@@ -1420,9 +1437,6 @@ int diAlloc(struct inode *pip, boolean_t dir, struct inode *ip)
 	if (ino == 0)
 		inum = pip->i_ino;
 
-	/* get the ag number of this iag */
-	agno = JFS_IP(pip)->agno;
-
 	/* lock the AG inode map information */
 	AG_LOCK(imap, agno);
 

fs/jfs/jfs_incore.h

@@ -49,7 +49,7 @@ struct jfs_inode_info {
 	long	cflag;		/* commit flags		*/
 	u16	bxflag;		/* xflag of pseudo buffer?	*/
 	unchar	agno;		/* ag number			*/
-	unchar	pad;		/* pad			*/
+	signed char active_ag;	/* ag currently allocating from	*/
 	lid_t	blid;		/* lid of pseudo buffer?	*/
 	lid_t	atlhead;	/* anonymous tlock list head	*/
 	lid_t	atltail;	/* anonymous tlock list tail	*/

fs/jfs/jfs_logmgr.c

@@ -66,6 +66,7 @@
 #include <linux/completion.h>
 #include <linux/buffer_head.h>		/* for sync_blockdev() */
 #include <linux/bio.h>
+#include <linux/suspend.h>
 #include "jfs_incore.h"
 #include "jfs_filsys.h"
 #include "jfs_metapage.h"
@@ -2146,12 +2147,17 @@ int jfsIOWait(void *arg)
 			lbmStartIO(bp);
 			spin_lock_irq(&log_redrive_lock);
 		}
-		add_wait_queue(&jfs_IO_thread_wait, &wq);
-		set_current_state(TASK_INTERRUPTIBLE);
-		spin_unlock_irq(&log_redrive_lock);
-		schedule();
-		current->state = TASK_RUNNING;
-		remove_wait_queue(&jfs_IO_thread_wait, &wq);
+		if (current->flags & PF_FREEZE) {
+			spin_unlock_irq(&log_redrive_lock);
+			refrigerator(PF_IOTHREAD);
+		} else {
+			add_wait_queue(&jfs_IO_thread_wait, &wq);
+			set_current_state(TASK_INTERRUPTIBLE);
+			spin_unlock_irq(&log_redrive_lock);
+			schedule();
+			current->state = TASK_RUNNING;
+			remove_wait_queue(&jfs_IO_thread_wait, &wq);
+		}
 	} while (!jfs_stop_threads);
 
 	jFYI(1,("jfsIOWait being killed!\n"));

fs/jfs/jfs_txnmgr.c

@@ -47,6 +47,7 @@
 #include <linux/vmalloc.h>
 #include <linux/smp_lock.h>
 #include <linux/completion.h>
+#include <linux/suspend.h>
 #include "jfs_incore.h"
 #include "jfs_filsys.h"
 #include "jfs_metapage.h"
@@ -2789,8 +2790,6 @@ int jfs_lazycommit(void *arg)
 	complete(&jfsIOwait);
 
 	do {
-		DECLARE_WAITQUEUE(wq, current);
-
 		LAZY_LOCK(flags);
 restart:
 		WorkDone = 0;
@@ -2825,12 +2824,19 @@ int jfs_lazycommit(void *arg)
 		if (WorkDone)
 			goto restart;
 
-		add_wait_queue(&jfs_commit_thread_wait, &wq);
-		set_current_state(TASK_INTERRUPTIBLE);
-		LAZY_UNLOCK(flags);
-		schedule();
-		current->state = TASK_RUNNING;
-		remove_wait_queue(&jfs_commit_thread_wait, &wq);
+		if (current->flags & PF_FREEZE) {
+			LAZY_UNLOCK(flags);
+			refrigerator(PF_IOTHREAD);
+		} else {
+			DECLARE_WAITQUEUE(wq, current);
+
+			add_wait_queue(&jfs_commit_thread_wait, &wq);
+			set_current_state(TASK_INTERRUPTIBLE);
+			LAZY_UNLOCK(flags);
+			schedule();
+			current->state = TASK_RUNNING;
+			remove_wait_queue(&jfs_commit_thread_wait, &wq);
+		}
 	} while (!jfs_stop_threads);
 
 	if (TxAnchor.unlock_queue)
@@ -2981,7 +2987,6 @@ int jfs_sync(void *arg)
 	complete(&jfsIOwait);
 
 	do {
-		DECLARE_WAITQUEUE(wq, current);
 		/*
 		 * write each inode on the anonymous inode list
 		 */
@@ -3030,12 +3035,20 @@ int jfs_sync(void *arg)
 		}
 		/* Add anon_list2 back to anon_list */
 		list_splice_init(&TxAnchor.anon_list2, &TxAnchor.anon_list);
-		add_wait_queue(&jfs_sync_thread_wait, &wq);
-		set_current_state(TASK_INTERRUPTIBLE);
-		TXN_UNLOCK();
-		schedule();
-		current->state = TASK_RUNNING;
-		remove_wait_queue(&jfs_sync_thread_wait, &wq);
+
+		if (current->flags & PF_FREEZE) {
+			TXN_UNLOCK();
+			refrigerator(PF_IOTHREAD);
+		} else {
+			DECLARE_WAITQUEUE(wq, current);
+
+			add_wait_queue(&jfs_sync_thread_wait, &wq);
+			set_current_state(TASK_INTERRUPTIBLE);
+			TXN_UNLOCK();
+			schedule();
+			current->state = TASK_RUNNING;
+			remove_wait_queue(&jfs_sync_thread_wait, &wq);
+		}
 	} while (!jfs_stop_threads);
 
 	jFYI(1, ("jfs_sync being killed\n"));

fs/jfs/super.c

@@ -406,6 +406,7 @@ static void init_once(void *foo, kmem_cache_t * cachep, unsigned long flags)
 		init_rwsem(&jfs_ip->rdwrlock);
 		init_MUTEX(&jfs_ip->commit_sem);
 		jfs_ip->atlhead = 0;
+		jfs_ip->active_ag = -1;
 		inode_init_once(&jfs_ip->vfs_inode);
 	}
 }

fs/jfs/xattr.c

@@ -78,6 +78,68 @@ struct ea_buffer {
 #define EA_NEW		0x0004
 #define EA_MALLOC	0x0008
 
+/* Namespaces */
+#define XATTR_SYSTEM_PREFIX "system."
+#define XATTR_SYSTEM_PREFIX_LEN (sizeof (XATTR_SYSTEM_PREFIX) - 1)
+
+#define XATTR_USER_PREFIX "user."
+#define XATTR_USER_PREFIX_LEN (sizeof (XATTR_USER_PREFIX) - 1)
+
+#define XATTR_OS2_PREFIX "os2."
+#define XATTR_OS2_PREFIX_LEN (sizeof (XATTR_OS2_PREFIX) - 1)
+
+/*
+ * These three routines are used to recognize on-disk extended attributes
+ * that are in a recognized namespace.  If the attribute is not recognized,
+ * "os2." is prepended to the name
+ */
+static inline int is_os2_xattr(struct jfs_ea *ea)
+{
+	/*
+	 * Check for "system."
+	 */
+	if ((ea->namelen >= XATTR_SYSTEM_PREFIX_LEN) &&
+	    !strncmp(ea->name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
+		return FALSE;
+	/*
+	 * Check for "user."
+	 */
+	if ((ea->namelen >= XATTR_USER_PREFIX_LEN) &&
+	    !strncmp(ea->name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN))
+		return FALSE;
+	/*
+	 * Add any other valid namespace prefixes here
+	 */
+
+	/*
+	 * We assume it's OS/2's flat namespace
+	 */
+	return TRUE;
+}
+
+static inline int name_size(struct jfs_ea *ea)
+{
+	if (is_os2_xattr(ea))
+		return ea->namelen + XATTR_OS2_PREFIX_LEN;
+	else
+		return ea->namelen;
+}
+
+static inline int copy_name(char *buffer, struct jfs_ea *ea)
+{
+	int len = ea->namelen;
+
+	if (is_os2_xattr(ea)) {
+		memcpy(buffer, XATTR_OS2_PREFIX, XATTR_OS2_PREFIX_LEN);
+		buffer += XATTR_OS2_PREFIX_LEN;
+		len += XATTR_OS2_PREFIX_LEN;
+	}
+	memcpy(buffer, ea->name, ea->namelen);
+	buffer[ea->namelen] = 0;
+
+	return len;
+}
+
 /* Forward references */
 static void ea_release(struct inode *inode, struct ea_buffer *ea_buf);
 
@@ -577,7 +639,8 @@ static int ea_put(struct inode *inode, struct ea_buffer *ea_buf, int new_size)
 	return rc;
 }
 
-static int can_set_xattr(struct inode *inode, const char *name)
+static int can_set_xattr(struct inode *inode, const char *name,
+			 void *value, size_t value_len)
 {
 	if (IS_RDONLY(inode))
 		return -EROFS;
@@ -585,6 +648,10 @@ static int can_set_xattr(struct inode *inode, const char *name)
 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode) || S_ISLNK(inode->i_mode))
 		return -EPERM;
 
+	if((strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN) != 0) &&
+	   (strncmp(name, XATTR_OS2_PREFIX, XATTR_OS2_PREFIX_LEN) != 0))
+		return -EOPNOTSUPP;
+
 	if (!S_ISREG(inode->i_mode) &&
 	    (!S_ISDIR(inode->i_mode) || inode->i_mode &S_ISVTX))
 		return -EPERM;
@@ -602,13 +669,24 @@ int __jfs_setxattr(struct inode *inode, const char *name, void *value,
 	int xattr_size;
 	int new_size;
 	int namelen = strlen(name);
+	char *os2name = NULL;
 	int found = 0;
 	int rc;
 	int length;
 
-	if ((rc = can_set_xattr(inode, name)))
+	if ((rc = can_set_xattr(inode, name, value, value_len)))
 		return rc;
 
+	if (strncmp(name, XATTR_OS2_PREFIX, XATTR_OS2_PREFIX_LEN) == 0) {
+		os2name = kmalloc(namelen - XATTR_OS2_PREFIX_LEN + 1,
+				  GFP_KERNEL);
+		if (!os2name)
+			return -ENOMEM;
+		strcpy(os2name, name + XATTR_OS2_PREFIX_LEN);
+		name = os2name;
+		namelen -= XATTR_OS2_PREFIX_LEN;
+	}
+
 	xattr_size = ea_get(inode, &ea_buf, 0);
 	if (xattr_size < 0) {
 		rc = xattr_size;
@@ -714,6 +792,9 @@ int __jfs_setxattr(struct inode *inode, const char *name, void *value,
       release:
 	ea_release(inode, &ea_buf);
       out:
+	if (os2name)
+		kfree(os2name);
+
 	return rc;
 }
 
@@ -728,7 +809,7 @@ int jfs_setxattr(struct dentry *dentry, const char *name, void *value,
 	return __jfs_setxattr(dentry->d_inode, name, value, value_len, flags);
 }
 
-static int can_get_xattr(struct inode *inode, const char *name)
+static inline int can_get_xattr(struct inode *inode, const char *name)
 {
 	return permission(inode, MAY_READ);
 }
@@ -742,12 +823,23 @@ ssize_t __jfs_getxattr(struct inode *inode, const char *name, void *data,
 	int xattr_size;
 	ssize_t size;
 	int namelen = strlen(name);
+	char *os2name = NULL;
 	int rc;
 	char *value;
 
 	if ((rc = can_get_xattr(inode, name)))
 		return rc;
 
+	if (strncmp(name, XATTR_OS2_PREFIX, XATTR_OS2_PREFIX_LEN) == 0) {
+		os2name = kmalloc(namelen - XATTR_OS2_PREFIX_LEN + 1,
+				  GFP_KERNEL);
+		if (!os2name)
+			return -ENOMEM;
+		strcpy(os2name, name + XATTR_OS2_PREFIX_LEN);
+		name = os2name;
+		namelen -= XATTR_OS2_PREFIX_LEN;
+	}
+
 	xattr_size = ea_get(inode, &ea_buf, 0);
 	if (xattr_size < 0) {
 		size = xattr_size;
@@ -780,6 +872,8 @@ ssize_t __jfs_getxattr(struct inode *inode, const char *name, void *data,
       release:
 	ea_release(inode, &ea_buf);
       out:
+	if (os2name)
+		kfree(os2name);
 
 	return size;
 }
@@ -813,7 +907,7 @@ ssize_t jfs_listxattr(struct dentry * dentry, char *data, size_t buf_size)
 
 	/* compute required size of list */
 	for (ea = FIRST_EA(ealist); ea < END_EALIST(ealist); ea = NEXT_EA(ea))
-		size += ea->namelen + 1;
+		size += name_size(ea) + 1;
 
 	if (!data)
 		goto release;
@@ -826,9 +920,8 @@ ssize_t jfs_listxattr(struct dentry * dentry, char *data, size_t buf_size)
 	/* Copy attribute names to buffer */
 	buffer = data;
 	for (ea = FIRST_EA(ealist); ea < END_EALIST(ealist); ea = NEXT_EA(ea)) {
-		memcpy(buffer, ea->name, ea->namelen);
-		buffer[ea->namelen] = 0;
-		buffer += ea->namelen + 1;
+		int namelen = copy_name(buffer, ea);
+		buffer += namelen + 1;
 	}
 
       release: