Re: 2 KiB/block loopback found where

[Pat LaVarre <p.lavarre@ieee.org>]

Can I easily fix my --getss too?

$ sudo losetup /dev/loop0 dd.bin
$
$ dmesg | tail -2
loop: loaded (max 8 devices)
loop:to x800 from lo_blocksize = 0x1000
$
$ sudo blockdev --getss /dev/loop0
512
$ sudo blockdev --getbsz /dev/loop0
4096
$
$ sudo blockdev --getss /dev/scd0
2048
$ sudo blockdev --getbsz /dev/scd0
4096
$

I'm pleased to report, patching drivers/block/loop.c as suggested,
specifically via the patch quoted below, did confuse udf.ko as well with
a loop device as does my real 2 KiB/block device.

Certainly I could and did then relay details to
linux_udf@hpesjro.fc.hp.com in a thread titled "balloc free bit already
set and byte minus one".  But I'd of course like to make my loop device
as perfect a simulation of my real device as practical.  So I ask:

Can I easily fix my --getss too?

Pat LaVarre

diff -Nur linux-2.6.0-test6/drivers/block/loop.c linux/drivers/block/loop.c
--- linux-2.6.0-test6/drivers/block/loop.c	2003-09-27 18:50:29.000000000 -0600
+++ linux/drivers/block/loop.c	2003-10-06 18:06:20.447708984 -0600
@@ -732,7 +732,13 @@
 	mapping_set_gfp_mask(inode->i_mapping,
 			     lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
 
+#if 0
 	set_blocksize(bdev, lo_blocksize);
+#else
+	printk(KERN_INFO "loop:"
+		"to x800 from lo_blocksize = 0x%X\n", lo_blocksize);
+	set_blocksize(bdev, 0x800);
+#endif
 
 	lo->lo_bio = lo->lo_biotail = NULL;