Re: [PATCH 1/5] ftrace: Synchronize variable setting with breakpoints

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 2012-05-31 at 20:03 +0200, Peter Zijlstra wrote:
> On Thu, 2012-05-31 at 13:53 -0400, Steven Rostedt wrote:
> > On Thu, 2012-05-31 at 19:40 +0200, Peter Zijlstra wrote:
> > > On Thu, 2012-05-31 at 10:08 -0400, Steven Rostedt wrote:
> > > > > Also, why does this stuff live in ftrace? I always thought you were
> > > > > going to replace text_poke() so everybody that uses cross-modifying code
> > > > > could profit?
> > > > 
> > > > I discussed this with Masami at Collaboration Summit. The two are
> > > > similar but also very different. But we want to start merging the two
> > > > together where it makes sense. 
> > > 
> > > Argh,. I so disagree. You're doing it backwards.
> > > 
> > > First you merge whatever is there, regardless of who came first.
> > 
> > The comment about coming first was more about 're-inventing' then about
> > merging. You can't reinvent something that didn't exist.
> > 
> > That said, I didn't even think about text poke while doing this.
> 
> Well, the fail is before that, how could we grow two pieces of code
> doing similar things in the first place? 

Again, ftrace is slightly different as it does 30,000 changes at once,
on top of known nops. This was done through stop_machine(), thus any
slowdown was a large hit to system performance. text_poke() took the way
of mapping a page to do the change, and Mathieu didn't want to change
that (IIRC). But now we want the two to be similar.


> 
> >  I was
> > just simply thinking about removing stop_machine from ftrace, that
> > required this. It was only a after thought that text_poke() could do the
> > same. And this came up at Collab, where I thought, oh yeah! we can
> > incorporate this with text poke.
> 
> But but but but.. the thing far back when Mathieu proposed the int3
> scheme it was text_poke().. how.. did you not think of it this time!?

Right, but we disagreed on its implementation, and yes, the idea came
from that, but I thought text_poke() was so different that it wouldn't
apply. It wasn't until Masami suggested changing text_poke() to be more
like what ftrace does, that I really took thought into it.

> 
> > 
> > > Then, when everybody doing text modification is using the same
> > > interface, do a second implementation using a Kconfig knob. If the scary
> > > new one breaks, no sweat, flip the config. If its proven stable, kill
> > > off the old one.
> > 
> > What do you suggest then? To revert the code and rewrite it so that
> > text_poke() does a similar thing?
> 
> Too late for that now I guess.. I just wonder why you all thought it was
> a good idea to have two pieces of code doing cross-modifying-code. I
> always assumed ftrace used text_poke().

Well, it was more of two different people working on two different
things. We really didn't look too closely at each others work. It was
more a social failure than a technical one.

> 
> I hardly ever use dyn-ftrace but I do use some text_poke() through
> jump_labels.

You don't use function tracer? That's dyn-ftrace.

But still, we need to keep the record as small as possible because it is
persistent throughout the life of the system running. Every location
must be recorded, and maintain a state (flags).

Text_poke() mostly grew out of the jump-label work. But yes, there's
still a lot that can be shared. The actual code modification may be.

> 
> I would still like to end up with one code base doing CMC with two
> implementations depending on a Kconfig knob.

You mean keep stop_machine around?

-- Steve