linux-2.6.9-ext3-sub-second-timestamp.patch

非常经典的一个分布式系统

 			;
 	}
 	up(&ei->truncate_sem);
-	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+	inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
 	ext3_mark_inode_dirty(handle, inode);
 
 	/* In a multi-transaction truncate, we only make the final
Index: linux-2.6.10/fs/ext3/ioctl.c
===================================================================
--- linux-2.6.10.orig/fs/ext3/ioctl.c	2004-12-24 14:34:31.000000000 -0700
+++ linux-2.6.10/fs/ext3/ioctl.c	2006-01-03 16:16:52.000000000 -0700
@@ -87,7 +87,7 @@
 		ei->i_flags = flags;
 
 		ext3_set_inode_flags(inode);
-		inode->i_ctime = CURRENT_TIME;
+		inode->i_ctime = CURRENT_TIME_SEC;
 
 		err = ext3_mark_iloc_dirty(handle, inode, &iloc);
 flags_err:
@@ -121,7 +121,7 @@
 			return PTR_ERR(handle);
 		err = ext3_reserve_inode_write(handle, inode, &iloc);
 		if (err == 0) {
-			inode->i_ctime = CURRENT_TIME;
+			inode->i_ctime = CURRENT_TIME_SEC;
 			inode->i_generation = generation;
 			err = ext3_mark_iloc_dirty(handle, inode, &iloc);
 		}
Index: linux-2.6.10/fs/ext3/namei.c
===================================================================
--- linux-2.6.10.orig/fs/ext3/namei.c	2004-12-24 14:34:58.000000000 -0700
+++ linux-2.6.10/fs/ext3/namei.c	2006-01-03 16:16:52.000000000 -0700
@@ -1251,7 +1251,7 @@
 	 * happen is that the times are slightly out of date
 	 * and/or different from the directory change time.
 	 */
-	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
+	dir->i_mtime = dir->i_ctime = CURRENT_TIME_SEC;
 	ext3_update_dx_flag(dir);
 	dir->i_version++;
 	ext3_mark_inode_dirty(handle, dir);
@@ -2029,7 +2029,7 @@
 	 * recovery. */
 	inode->i_size = 0;
 	ext3_orphan_add(handle, inode);
-	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
+	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC;
 	ext3_mark_inode_dirty(handle, inode);
 	dir->i_nlink--;
 	ext3_update_dx_flag(dir);
@@ -2079,7 +2079,7 @@
 	retval = ext3_delete_entry(handle, dir, de, bh);
 	if (retval)
 		goto end_unlink;
-	dir->i_ctime = dir->i_mtime = CURRENT_TIME;
+	dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC;
 	ext3_update_dx_flag(dir);
 	ext3_mark_inode_dirty(handle, dir);
 	inode->i_nlink--;
@@ -2169,7 +2169,7 @@
 	if (IS_DIRSYNC(dir))
 		handle->h_sync = 1;
 
-	inode->i_ctime = CURRENT_TIME;
+	inode->i_ctime = CURRENT_TIME_SEC;
 	ext3_inc_count(handle, inode);
 	atomic_inc(&inode->i_count);
 
@@ -2270,7 +2270,7 @@
 	 * Like most other Unix systems, set the ctime for inodes on a
 	 * rename.
 	 */
-	old_inode->i_ctime = CURRENT_TIME;
+	old_inode->i_ctime = CURRENT_TIME_SEC;
 	ext3_mark_inode_dirty(handle, old_inode);
 
 	/*
@@ -2303,9 +2303,9 @@
 
 	if (new_inode) {
 		new_inode->i_nlink--;
-		new_inode->i_ctime = CURRENT_TIME;
+		new_inode->i_ctime = CURRENT_TIME_SEC;
 	}
-	old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME;
+	old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME_SEC;
 	ext3_update_dx_flag(old_dir);
 	if (dir_bh) {
 		BUFFER_TRACE(dir_bh, "get_write_access");
Index: linux-2.6.10/fs/ext3/super.c
===================================================================
--- linux-2.6.10.orig/fs/ext3/super.c	2005-04-06 09:38:35.000000000 -0600
+++ linux-2.6.10/fs/ext3/super.c	2006-01-03 16:16:52.000000000 -0700
@@ -1318,7 +1318,7 @@
 	if (!parse_options ((char *) data, sb, &journal_inum, NULL, 0))
 		goto failed_mount;
 
-	sb->s_flags |= MS_ONE_SECOND;
+	set_sb_time_gran(sb, 1000000000U);
 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 		((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
 
Index: linux-2.6.10/fs/ext3/xattr.c
===================================================================
--- linux-2.6.10.orig/fs/ext3/xattr.c	2005-04-06 09:38:35.000000000 -0600
+++ linux-2.6.10/fs/ext3/xattr.c	2006-01-03 16:16:52.000000000 -0700
@@ -723,7 +723,7 @@
 
 	/* Update the inode. */
 	EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
-	inode->i_ctime = CURRENT_TIME;
+	inode->i_ctime = CURRENT_TIME_SEC;
 	ext3_mark_inode_dirty(handle, inode);
 	if (IS_SYNC(inode))
 		handle->h_sync = 1;
Index: linux-2.6.10/fs/inode.c
===================================================================
--- linux-2.6.10.orig/fs/inode.c	2006-01-03 15:33:21.000000000 -0700
+++ linux-2.6.10/fs/inode.c	2006-01-03 16:16:52.000000000 -0700
@@ -1131,19 +1131,6 @@
 
 EXPORT_SYMBOL(bmap);
 
-/*
- * Return true if the filesystem which backs this inode considers the two
- * passed timespecs to be sufficiently different to warrant flushing the
- * altered time out to disk.
- */
-static int inode_times_differ(struct inode *inode,
-			struct timespec *old, struct timespec *new)
-{
-	if (IS_ONE_SECOND(inode))
-		return old->tv_sec != new->tv_sec;
-	return !timespec_equal(old, new);
-}
-
 /**
  *	update_atime	-	update the access time
  *	@inode: inode accessed
@@ -1163,8 +1150,8 @@
 	if (IS_RDONLY(inode))
 		return;
 
-	now = current_kernel_time();
-	if (inode_times_differ(inode, &inode->i_atime, &now)) {
+	now = current_fs_time(inode->i_sb);
+	if (!timespec_equal(&inode->i_atime, &now)) {
 		inode->i_atime = now;
 		mark_inode_dirty_sync(inode);
 	} else {
@@ -1194,14 +1181,13 @@
 	if (IS_RDONLY(inode))
 		return;
 
-	now = current_kernel_time();
-
-	if (inode_times_differ(inode, &inode->i_mtime, &now))
+	now = current_fs_time(inode->i_sb);
+	if (!timespec_equal(&inode->i_mtime, &now))
 		sync_it = 1;
 	inode->i_mtime = now;
 
 	if (ctime_too) {
-		if (inode_times_differ(inode, &inode->i_ctime, &now))
+		if (!timespec_equal(&inode->i_ctime, &now))
 			sync_it = 1;
 		inode->i_ctime = now;
 	}
Index: linux-2.6.10/fs/locks.c
===================================================================
--- linux-2.6.10.orig/fs/locks.c	2004-12-24 14:35:28.000000000 -0700
+++ linux-2.6.10/fs/locks.c	2006-01-03 16:16:52.000000000 -0700
@@ -1228,7 +1228,7 @@
 {
 	struct file_lock *flock = inode->i_flock;
 	if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
-		*time = CURRENT_TIME;
+		*time = current_fs_time(inode->i_sb);
 	else
 		*time = inode->i_mtime;
 }
Index: linux-2.6.10/include/linux/fs.h
===================================================================
--- linux-2.6.10.orig/include/linux/fs.h	2006-01-03 16:04:26.000000000 -0700
+++ linux-2.6.10/include/linux/fs.h	2006-01-03 16:16:52.000000000 -0700
@@ -124,7 +124,8 @@
 #define MS_REC		16384
 #define MS_VERBOSE	32768
 #define MS_POSIXACL	(1<<16)	/* VFS does not apply the umask */
-#define MS_ONE_SECOND	(1<<17)	/* fs has 1 sec a/m/ctime resolution */
+#define MS_ONE_SECOND	(1<<17)	/* fs has 1 sec time resolution (obsolete) */
+#define MS_TIME_GRAN	(1<<18)	/* fs has s_time_gran field */
 #define MS_ACTIVE	(1<<30)
 #define MS_NOUSER	(1<<31)
 
@@ -803,8 +804,33 @@
 	 * even looking at it. You had been warned.
 	 */
 	struct semaphore s_vfs_rename_sem;	/* Kludge */
+
+	/* Granuality of c/m/atime in ns.
+	   Cannot be worse than a second */
+#ifndef __GENKSYMS__
+	u32		   s_time_gran;
+#endif
 };
 
+extern struct timespec current_fs_time(struct super_block *sb);
+
+static inline u32 get_sb_time_gran(struct super_block *sb) 
+{
+	if (sb->s_flags & MS_TIME_GRAN) 
+		return sb->s_time_gran;
+	if (sb->s_flags & MS_ONE_SECOND)
+		return 1000000000U;
+	return 1;
+}
+
+static inline void set_sb_time_gran(struct super_block *sb, u32 time_gran)
+{
+	sb->s_time_gran = time_gran;
+	sb->s_flags |= MS_TIME_GRAN;
+	if (time_gran == 1000000000U)
+		sb->s_flags |= MS_ONE_SECOND;
+}
+
 /*
  * Snapshotting support.
  */
Index: linux-2.6.10/include/linux/time.h
===================================================================
--- linux-2.6.10.orig/include/linux/time.h	2004-12-24 14:35:00.000000000 -0700
+++ linux-2.6.10/include/linux/time.h	2006-01-03 16:16:52.000000000 -0700
@@ -90,6 +90,7 @@
 struct timespec current_kernel_time(void);
 
 #define CURRENT_TIME (current_kernel_time())
+#define CURRENT_TIME_SEC ((struct timespec) { xtime.tv_sec, 0 })
 
 extern void do_gettimeofday(struct timeval *tv);
 extern int do_settimeofday(struct timespec *tv);
@@ -103,6 +104,8 @@
 extern int do_getitimer(int which, struct itimerval *value);
 extern void getnstimeofday (struct timespec *tv);
 
+extern struct timespec timespec_trunc(struct timespec t, unsigned gran);
+
 static inline void
 set_normalized_timespec (struct timespec *ts, time_t sec, long nsec)
 {
Index: linux-2.6.10/kernel/time.c
===================================================================
--- linux-2.6.10.orig/kernel/time.c	2004-12-24 14:34:26.000000000 -0700
+++ linux-2.6.10/kernel/time.c	2006-01-03 16:16:52.000000000 -0700
@@ -36,6 +36,7 @@
 
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
+#include <linux/fs.h>
 
 /* 
  * The timezone where the local system is located.  Used as a default by some
@@ -433,6 +434,50 @@
 
 EXPORT_SYMBOL(current_kernel_time);
 
+/**
+ * current_fs_time - Return FS time
+ * @sb: Superblock.
+ *
+ * Return the current time truncated to the time granuality supported by
+ * the fs.
+ */
+struct timespec current_fs_time(struct super_block *sb)
+{
+	struct timespec now = current_kernel_time();
+	return timespec_trunc(now, get_sb_time_gran(sb));
+}
+EXPORT_SYMBOL(current_fs_time);
+
+/**
+ * timespec_trunc - Truncate timespec to a granuality
+ * @t: Timespec
+ * @gran: Granuality in ns.
+ *
+ * Truncate a timespec to a granuality. gran must be smaller than a second.
+ * Always rounds down.
+ *
+ * This function should be only used for timestamps returned by
+ * current_kernel_time() or CURRENT_TIME, not with do_gettimeofday() because
+ * it doesn't handle the better resolution of the later.
+ */
+struct timespec timespec_trunc(struct timespec t, unsigned gran)
+{
+	/*
+	 * Division is pretty slow so avoid it for common cases.
+	 * Currently current_kernel_time() never returns better than
+	 * jiffies resolution. Exploit that.
+	 */
+	if (gran <= jiffies_to_usecs(1) * 1000) {
+		/* nothing */
+	} else if (gran == 1000000000) {
+		t.tv_nsec = 0;
+	} else {
+		t.tv_nsec -= t.tv_nsec % gran;
+	}
+	return t;
+}
+EXPORT_SYMBOL(timespec_trunc);
+
 #ifdef CONFIG_TIME_INTERPOLATION
 void getnstimeofday (struct timespec *tv)
 {