Re: [PATCH v3] printk: Have printk() never buffer its data

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Fri, Jun 29, 2012 at 07:18:38AM -0400, Steven Rostedt wrote:
> On Fri, 2012-06-29 at 01:30 -0400, Greg Kroah-Hartman wrote:
> >  
> > > I wonder if it would be better to do the following for the above two
> > > ifs:
> > > 
> > > 	if (cont.len && cont.owner == current) {
> > > 		if (!prefix)
> > > 			stored = cont_add(facility, level, text, text_len);
> > > 		cont_flush();
> > > 	}
> > > 
> > > If the prefix was true, then the cont.flush would be set when cont_add()
> > > is called, and the first thing that cont_add() does:
> > > 
> > > 	if (cont.len && cont.flushed)
> > > 		return false;
> > > 
> > > which would always be true (returning false) if prefix was set.
> > > 
> > > And the second cont_flush() is a nop due to it doing:
> > > 
> > > 	if (cont.flushed)
> > > 		return;
> > 
> > It might be "better", and this would be a nice optimization, but is it
> > needed right now?  In other words, I'd like to get this patch into
> > linux-next soon to get testing to get to Linus before 3.5-final comes
> > out, don't you?
> 
> Sure, pull it as is, and you can add my Tested-by, and Acked-by tags
> (Steven Rostedt <rostedt@goodmis.org>).
> 
> I'll send you a patch to do this update that you can queue for 3.6.
> OK?

Sounds good to me, thanks,

greg k-h