[PATCH] handle partial DIO write

The fsx-linux hole fill failure problem was caused by
generic_file_aio_write_nolock() not handling the partial DIO write
correctly.  Here's a patch lets DIO do the partial write, and the fallback
to buffered is done (correctly) for what is left.  This fixes the hole
filling without retrying the entire i/o.  This patch also applies to
2.6.7-rc3 with some offset.

I tested this (on ext3) with
fsx-linux -l 500000 -r 4096 -t 4096 -w 4096 -Z -N 10000 junk  -R -W
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/direct-io.c

@@ -990,6 +990,13 @@ direct_io_worker(int rw, struct kiocb *iocb, struct inode *inode,
 		}
 	} /* end iovec loop */
 
+	if (ret == -ENOTBLK && rw == WRITE) {
+		/*
+		 * The remaining part of the request will be
+		 * be handled by buffered I/O when we return
+		 */
+		ret = 0;
+	}
 	/*
 	 * There may be some unwritten disk at the end of a part-written
 	 * fs-block-sized block.  Go zero that now.
@@ -1089,13 +1096,6 @@ direct_io_worker(int rw, struct kiocb *iocb, struct inode *inode,
 			aio_complete(iocb, ret, 0);
 		kfree(dio);
 	}
-	if (ret == -ENOTBLK && rw == WRITE) {
-		/*
-		 * The entire request will be be handled by buffered I/O
-		 * when we return
-		 */
-		ret = 0;
-	}
 	return ret;
 }
 

mm/filemap.c

@@ -1915,7 +1915,7 @@ generic_file_aio_write_nolock(struct kiocb *iocb, const struct iovec *iov,
 		count -= written;
 	}
 
-	buf = iov->iov_base;
+	buf = iov->iov_base + written;	/* handle partial DIO write */
 	do {
 		unsigned long index;
 		unsigned long offset;