On Mon, Apr 30, 2018 at 6:42 PM Joel Fernandes <joe...@google.com> wrote:
> In recent tests with IRQ on/off tracepoints, a large performance > overhead ~10% is noticed when running hackbench. This is root caused to > calls to rcu_irq_enter_irqson and rcu_irq_exit_irqson from the > tracepoint code. Following a long discussion on the list [1] about this, > we concluded that srcu is a better alternative for use during rcu idle. > Although it does involve extra barriers, its lighter than the sched-rcu > version which has to do additional RCU calls to notify RCU idle about > entry into RCU sections. > In this patch, we change the underlying implementation of the > trace_*_rcuidle API to use SRCU. This has shown to improve performance > alot for the high frequency irq enable/disable tracepoints. > Test: Tested idle and preempt/irq tracepoints. > [1] https://patchwork.kernel.org/patch/10344297/ > Cc: Steven Rostedt <rost...@goodmis.org> > Cc: Peter Zilstra <pet...@infradead.org> > Cc: Ingo Molnar <mi...@redhat.com> > Cc: Mathieu Desnoyers <mathieu.desnoy...@efficios.com> > Cc: Tom Zanussi <tom.zanu...@linux.intel.com> > Cc: Namhyung Kim <namhy...@kernel.org> > Cc: Thomas Glexiner <t...@linutronix.de> > Cc: Boqun Feng <boqun.f...@gmail.com> > Cc: Paul McKenney <paul...@linux.vnet.ibm.com> > Cc: Frederic Weisbecker <fweis...@gmail.com> > Cc: Randy Dunlap <rdun...@infradead.org> > Cc: Masami Hiramatsu <mhira...@kernel.org> > Cc: Fenguang Wu <fengguang...@intel.com> > Cc: Baohong Liu <baohong....@intel.com> > Cc: Vedang Patel <vedang.pa...@intel.com> > Cc: kernel-t...@android.com > Signed-off-by: Joel Fernandes <joe...@google.com> > --- > include/linux/tracepoint.h | 46 +++++++++++++++++++++++++++++++------- > kernel/tracepoint.c | 10 ++++++++- > 2 files changed, 47 insertions(+), 9 deletions(-) > diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h > index c94f466d57ef..4135e08fb5f1 100644 > --- a/include/linux/tracepoint.h > +++ b/include/linux/tracepoint.h > @@ -15,6 +15,7 @@ > */ > #include <linux/smp.h> > +#include <linux/srcu.h> > #include <linux/errno.h> > #include <linux/types.h> > #include <linux/cpumask.h> > @@ -33,6 +34,8 @@ struct trace_eval_map { > #define TRACEPOINT_DEFAULT_PRIO 10 > +extern struct srcu_struct tracepoint_srcu; > + > extern int > tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data); > extern int > @@ -77,6 +80,9 @@ int unregister_tracepoint_module_notifier(struct notifier_block *nb) > */ > static inline void tracepoint_synchronize_unregister(void) > { > +#ifdef CONFIG_TRACEPOINTS > + synchronize_srcu(&tracepoint_srcu); > +#endif > synchronize_sched(); > } > @@ -129,18 +135,38 @@ extern void syscall_unregfunc(void); > * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just > * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto". > */ > -#define __DO_TRACE(tp, proto, args, cond, rcucheck) \ > +#define __DO_TRACE(tp, proto, args, cond, rcuidle) \ > do { \ > struct tracepoint_func *it_func_ptr; \ > void *it_func; \ > void *__data; \ > + int __maybe_unused idx = 0; \ > \ > if (!(cond)) \ > return; \ > - if (rcucheck) \ > - rcu_irq_enter_irqson(); \ > - rcu_read_lock_sched_notrace(); \ > - it_func_ptr = rcu_dereference_sched((tp)->funcs); \ > + \ > + /* \ > + * For rcuidle callers, use srcu since sched-rcu \ > + * doesn't work from the idle path. \ > + */ \ > + if (rcuidle) { \ > + if (in_nmi()) { \ > + WARN_ON_ONCE(1); \ > + return; /* no srcu from nmi */ \ > + } \ > + \ > + /* To keep it consistent with !rcuidle path */ \ > + preempt_disable_notrace(); \ > + \ > + idx = srcu_read_lock_notrace(&tracepoint_srcu); \ > + it_func_ptr = srcu_dereference((tp)->funcs, \ > + &tracepoint_srcu); \ This last bit is supposed to be srcu_dereference_notrace. The hunk to use that is actually in patch 6/6 , sorry about that. I've fixed it in my tree and it means patches 5/6 and 6/6 need an update. Steve, if you want me to repost it right away I can do that, or can wait for additional comments and then repost. thanks, - Joel