I'm reposting this patch (which was included in 2.6.0-mm1 and then dropped) 
since it is not clear why it was dropped in 2.6.0-mm2 (maybe I'm wrong, maybe 
it was lost/I was not clear); but the bugs are still there (at least in stock 
2.6.1).

THIS PATCH:

loop_init doesn't fail gracefully for two reasons:
1)If initialization of loop driver fails, we have an call to devfs_add("loop") 
without any devfs_remove; I add that.

2) On lwn.net 2.6 kernel docs, Jonathan Corbet says:
"If you are calling add_disk() in your driver initialization routine, you 
should not fail the initialization process after the first call."

Also, if that is not invalid, then we would need to call del_gendisk on 
failure. So the fix I post is the simplest one.

So I make loop.c conform to this request by moving add_disk after all memory 
allocations.

CC me on reply as I'm not subscribed, please.

Thanks for attention.
-- 
Paolo Giarrusso, aka Blaisorblade