>> > Bah.  In inode_setattr(), move the mark_inode_dirty() outside
>> > lock_kernel().
>> 
>> OK, will do. 
> 
> Actually we can just ditch it.

Well, that just seems to make the box hang in SDET (actually, moving it
outside lock_kernel makes it hang in a similar way). Not sure it's 
*caused* by this ... it might just change timing enough to trigger it.

ALT+sysrq info attatched, compressed, cause there's so many stinking
processes running ;-)

M.