[patch 12/16] fix race between writeback and unlink

[Andrew Morton <akpm@zip.com.au>]

Fixes a race between unlink and writeback: on the sys_sync() and
pdflush paths the caller does not have a reference against the inode.

So run __iget prior to dropping inode_lock.

Oleg Drokin reported this and seems to believe that it fixes the
crashes he was observing.  But I was never able to reproduce them..


=====================================

--- 2.5.19/fs/fs-writeback.c~sync-race	Sat Jun  1 01:18:12 2002
+++ 2.5.19-akpm/fs/fs-writeback.c	Sat Jun  1 01:18:12 2002
@@ -245,17 +245,19 @@ static void sync_sb_inodes(struct super_
 		if ((sync_mode == WB_SYNC_LAST) && (head->prev == head))
 			really_sync = 1;
 
+		BUG_ON(inode->i_state & I_FREEING);
+		__iget(inode);
 		__writeback_single_inode(inode, really_sync, nr_to_write);
-
 		if (sync_mode == WB_SYNC_HOLD) {
 			mapping->dirtied_when = jiffies;
 			list_del(&inode->i_list);
 			list_add(&inode->i_list, &inode->i_sb->s_dirty);
 		}
-
 		if (current_is_pdflush())
 			writeback_release(bdi);
-
+		spin_unlock(&inode_lock);
+		iput(inode);
+		spin_lock(&inode_lock);
 		if (nr_to_write && *nr_to_write <= 0)
 			break;
 	}


-