Re: [PATCH v3 5/6] x86/ftrace: Use text_poke()

[Peter Zijlstra]

On Thu, Oct 10, 2019 at 09:19:56AM -0400, Steven Rostedt wrote:
> On Thu, 10 Oct 2019 11:20:54 +0200
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> 
> > On Wed, Oct 09, 2019 at 10:41:35PM -0400, Steven Rostedt wrote:
> > > On Tue, 8 Oct 2019 10:43:35 -0400
> > > Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> > > 
> > >   
> > > > BTW, I'd really like to take this patch series through my tree. That
> > > > way I can really hammer it, as well as I have code that will be built
> > > > on top of it.  
> > > 
> > > I did a bit of hammering and found two bugs. One I sent a patch to fix
> > > (adding a module when tracing is enabled), but the other bug I
> > > triggered, I'm too tired to debug right now. But figured I'd mention it
> > > anyway.  
> > 
> > I'm thinking this should fix it... Just not sure this is the right plce,
> > then again, we're doing the same thing in jump_label and static_call, so
> > perhaps we should do it like this.
> > 
> > --- a/arch/x86/kernel/alternative.c
> > +++ b/arch/x86/kernel/alternative.c
> > @@ -1230,10 +1230,15 @@ void text_poke_queue(void *addr, const v
> >   * dynamically allocated memory. This function should be used when it is
> >   * not possible to allocate memory.
> >   */
> > -void text_poke_bp(void *addr, const void *opcode, size_t len, const void *emulate)
> > +void __ref text_poke_bp(void *addr, const void *opcode, size_t len, const void *emulate)
> >  {
> >  	struct text_poke_loc tp;
> >  
> > +	if (unlikely(system_state == SYSTEM_BOOTING)) {
> > +		text_poke_early(addr, opcode, len);
> > +		return;
> > +	}
> 
> We need a new system state. SYSTEM_UP ? (Arg, that name is confusing,
> SYSTEM_BOOTING_SMP?) Or perhaps just test num_online_cpus()?
> 
> 	if (unlikely(system_state == SYSTEM_BOOTING &&
> 		     num_online_cpus() == 1)
> 
> ?
> 
> Because we can't do the above once we have more than one CPU running.

We loose BOOTING _long_ before we gain SMP.