Re: More on EFS bug

[Alan Cox]

> The problem isn't the freshly burnt CD, as I tried with the original SGI CD,
> which shows the same problem using ide-scsi, but is fine when I access things
> natively over IDE.

That confirms my suspicion. Patch below. The bug cases should now error
politely

--- fs/efs/super.c~	2002-08-21 20:40:27.000000000 +0100
+++ fs/efs/super.c	2002-08-21 20:40:27.000000000 +0100
@@ -141,7 +141,13 @@
 	s->s_magic		= EFS_SUPER_MAGIC;
 	s->s_blocksize		= EFS_BLOCKSIZE;
 	s->s_blocksize_bits	= EFS_BLOCKSIZE_BITS;
-	set_blocksize(dev, EFS_BLOCKSIZE);
+	
+	if( set_blocksize(dev, EFS_BLOCKSIZE) < 0)
+	{
+		printk(KERN_ERR "EFS: device does not support %d byte blocks\n",
+			EFS_BLOCKSIZE);
+		goto out_no_fs_ul;
+	}
   
 	/* read the vh (volume header) block */
 	bh = sb_bread(s, 0);