--- 2.4.21pre4aa1/fs/inode.c.~1~	2003-01-31 02:38:47.000000000 +0100
+++ 2.4.21pre4aa1/fs/inode.c	2003-02-03 17:48:25.000000000 +0100
@@ -199,6 +199,9 @@ void __mark_inode_dirty(struct inode *in
 			sb->s_op->dirty_inode(inode);
 	}
 
+	/* make sure that changes are seen by all cpus before we test i_state */
+	smp_mb();
+
 	/* avoid the locking if we can */
 	if ((inode->i_state & flags) == flags)
 		return;
@@ -273,6 +276,23 @@ static inline void __sync_one(struct ino
 	inode->i_state &= ~I_DIRTY;
 	spin_unlock(&inode_lock);
 
+	/*
+	 * mark_inode_dirty() doesn't take spinlock, make sure
+	 * that the data is not read speculatively by this cpu
+	 * before &= ~I_DIRTY is flushed to ram.
+	 *
+	 * We depend on the inode->i_state &= ~I_DIRTY not passing
+	 * the smp_rmb() because of the implicit one-way memory
+	 * barrier of the spinlock that avoids the critical section
+	 * to pass the spin_unlock(). Shall this dependency break
+	 * (i.e. if the spin_unlock goes away), this has to be
+	 * converted to a smp_mb() (note: if I_DIRTY would
+	 * become a bitop, a test_and_set_bit just has to provide
+	 * implicity smp_mb() semantics for the non contention
+	 * branch).
+	 */
+	smp_rmb();
+
 	filemap_fdatasync(inode->i_mapping);
 
 	/* Don't write the inode if only I_DIRTY_PAGES was set */