Re: [PATCH tip/core/rcu 6/8] rcutorture: Print symbolic name for ->gp_state

[Josh Triplett <josh@joshtriplett.org>]

On Sat, Dec 05, 2015 at 05:54:36PM -0800, Paul E. McKenney wrote:
> On Fri, Dec 04, 2015 at 05:04:21PM -0800, Josh Triplett wrote:
> > On Fri, Dec 04, 2015 at 04:23:50PM -0800, Paul E. McKenney wrote:
> > > Currently, ->gp_state is printed as an integer, which slows debugging.
> > > This commit therefore prints a symbolic name in addition to the integer.
> > > 
> > > Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > [ Updated to fix relational operator called out by Dan Carpenter. ]
> > > ---
> > >  kernel/rcu/tree.c | 15 +++++++++++++--
> > >  kernel/rcu/tree.h |  9 +++++++++
> > >  2 files changed, 22 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > > index 7b78c88e19a3..491c99a73996 100644
> > > --- a/kernel/rcu/tree.c
> > > +++ b/kernel/rcu/tree.c
> > > @@ -1187,6 +1187,16 @@ static void record_gp_stall_check_time(struct rcu_state *rsp)
> > >  }
> > >  
> > >  /*
> > > + * Convert a ->gp_state value to a character string.
> > > + */
> > > +static char *gp_state_getname(short gs)
> > 
> > Same comment about "const char *"
> 
> Good point, fixed!
> 
> > > +{
> > > +	if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names))
> > > +		return "???";
> > > +	return gp_state_names[gs];
> > > +}
> > > +
> > > +/*
> > >   * Complain about starvation of grace-period kthread.
> > >   */
> > >  static void rcu_check_gp_kthread_starvation(struct rcu_state *rsp)
> > > @@ -1197,10 +1207,11 @@ static void rcu_check_gp_kthread_starvation(struct rcu_state *rsp)
> > >  	j = jiffies;
> > >  	gpa = READ_ONCE(rsp->gp_activity);
> > >  	if (j - gpa > 2 * HZ) {
> > > -		pr_err("%s kthread starved for %ld jiffies! g%lu c%lu f%#x s%d ->state=%#lx\n",
> > > +		pr_err("%s kthread starved for %ld jiffies! g%lu c%lu f%#x %s(%d) ->state=%#lx\n",
> > >  		       rsp->name, j - gpa,
> > >  		       rsp->gpnum, rsp->completed,
> > > -		       rsp->gp_flags, rsp->gp_state,
> > > +		       rsp->gp_flags,
> > > +		       gp_state_getname(rsp->gp_state), rsp->gp_state,
> > >  		       rsp->gp_kthread ? rsp->gp_kthread->state : ~0);
> > >  		if (rsp->gp_kthread)
> > >  			sched_show_task(rsp->gp_kthread);
> > > diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
> > > index f32bebb6bc90..f0304988e3f8 100644
> > > --- a/kernel/rcu/tree.h
> > > +++ b/kernel/rcu/tree.h
> > > @@ -544,6 +544,15 @@ struct rcu_state {
> > >  #define RCU_GP_DOING_FQS 4	/* Wait done for force-quiescent-state time. */
> > >  #define RCU_GP_CLEANUP   5	/* Grace-period cleanup started. */
> > >  #define RCU_GP_CLEANED   6	/* Grace-period cleanup complete. */
> > > +static char * const gp_state_names[] = {
> > 
> > Same comment about "const char * const", but also, should this really go
> > in a header file?  Do any circumstances exist where this header could
> > get included in more than one source file in a given build?  If so,
> > multiple copies of this array could potentially end up in the binary, if
> > the compiler fails to merge them.
> 
> I added the const.
> 
> However, I left the array in the .h file.  The issue is that the #defines
> really have to be in the tree.h file.  Given that, if I move the array
> elsewhere, they will very likely drift out of sync.  So this sounds like
> a job for the compiler/linker...
> 
> Ah, but there is another approach.  It turns out that tree.h in included
> only from tree.c and tree_trace.c.  And tree_trace.c defines a CPP symbol
> RCU_TREE_NONCORE that is used elsewhere in tree.h to keep tree_trace.c's
> mitts off a number of attractive but dangerous functions.  So I just put
> the array under "#ifndef RCU_TREE_NONCORE".
> 
> Fair enough?

Fair enough.

Reviewed-by: Josh Triplett <josh@joshtriplett.org>