[PATCH] scanlog_init cleanup, minor fixes

[Nathan Lynch]

Michael Ellerman wrote:
> > +	/* Ideally we could allocate a buffer < 4G */
> > +	data = kzalloc(RTAS_DATA_BUF_SIZE, GFP_KERNEL);
> > +	if (!data)
> > +		goto err;
> 
> Not your bug, but what happens if data is > 4G? Kaboom?

An old RPA doc (scan-log-dump isn't specified in PAPR) says the buffer
should be "contiguous real storage", so... yeah, probably.  That's why
I preserved the comment.  Will fix if I get access to a machine to
test this code more thoroughly (plenty of other issues in this file,
too).