Index: sparc/cpu.c
===================================================================
RCS file: /cvsroot/src/sys/arch/sparc/sparc/cpu.c,v
retrieving revision 1.225
diff -p -r1.225 cpu.c
*** sparc/cpu.c	22 Jan 2011 12:13:25 -0000	1.225
--- sparc/cpu.c	26 Jan 2011 06:09:54 -0000
*************** xcall(xcall_func_t func, xcall_trap_t tr
*** 583,588 ****
--- 583,589 ----
  	volatile struct xpmsg_func *p;
  	int fasttrap;
  	int is_noop = func == (xcall_func_t)sparc_noop;
+ 	u_int failed_cpuset;
  
  	mybit = (1 << cpuinfo.ci_cpuid);
  	callself = func && (cpuset & mybit) != 0;
*************** xcall(xcall_func_t func, xcall_trap_t tr
*** 632,650 ****
  	done = is_noop;
  	i = 100000;	/* time-out, not too long, but still an _AGE_ */
  	while (!done) {
- 		if (--i < 0) {
- 			printf_nolog("xcall(cpu%d,%p): couldn't ping cpus:",
- 			    cpu_number(), func);
- 		}
  
  		done = 1;
  		for (CPU_INFO_FOREACH(n, cpi)) {
  			if ((cpuset & (1 << n)) == 0)
  				continue;
  
  			if (cpi->msg.complete == 0) {
  				if (i < 0) {
! 					printf_nolog(" cpu%d", cpi->ci_cpuid);
  				} else {
  					done = 0;
  					break;
--- 633,648 ----
  	done = is_noop;
  	i = 100000;	/* time-out, not too long, but still an _AGE_ */
  	while (!done) {
  
  		done = 1;
+ 		failed_cpuset = 0;
  		for (CPU_INFO_FOREACH(n, cpi)) {
  			if ((cpuset & (1 << n)) == 0)
  				continue;
  
  			if (cpi->msg.complete == 0) {
  				if (i < 0) {
! 					failed_cpuset |= (1 << n);
  				} else {
  					done = 0;
  					break;
*************** xcall(xcall_func_t func, xcall_trap_t tr
*** 652,661 ****
  			}
  		}
  	}
- 	if (i < 0)
- 		printf_nolog("\n");
  
  	mutex_spin_exit(&xpmsg_mutex);
  }
  
  /*
--- 650,674 ----
  			}
  		}
  	}
  
  	mutex_spin_exit(&xpmsg_mutex);
+ 
+ 
+ 	/*
+ 	 * Ping the failed xcall(), see if they can tell us what they
+ 	 * were doing.
+ 	 */
+ 	if (i < 0) {
+ 		for (CPU_INFO_FOREACH(n, cpi))
+ 			if (failed_cpuset & (1 << n))
+ 				mp_ping_cpu(cpi);
+ 		printf_nolog("xcall(cpu%d,%p): couldn't ping cpus:",
+ 		    cpu_number(), func);
+ 		for (CPU_INFO_FOREACH(n, cpi))
+ 			if (failed_cpuset & (1 << n))
+ 				printf_nolog(" cpu%d", cpi->ci_cpuid);
+ 		printf_nolog("\n");
+ 	}
  }
  
  /*
*************** mp_resume_cpus_ddb(void)
*** 782,787 ****
--- 795,812 ----
  	}
  }
  #endif /* DDB */
+ 
+ /*
+  * ping the remote CPU with a highest priority interrupt, used
+  * when we had an IPI timeout.
+  */
+ void
+ mp_ping_cpu(struct cpu_info *cpi)
+ {
+ 
+ 	cpi->msg_lev15.tag = XPMSG15_PING;
+ 	raise_ipi(cpi,15);	/* high priority intr */
+ }
  #endif /* MULTIPROCESSOR */
  
  /*
Index: sparc/cpuvar.h
===================================================================
RCS file: /cvsroot/src/sys/arch/sparc/sparc/cpuvar.h,v
retrieving revision 1.84
diff -p -r1.84 cpuvar.h
*** sparc/cpuvar.h	13 Jan 2011 05:20:27 -0000	1.84
--- sparc/cpuvar.h	26 Jan 2011 06:09:54 -0000
*************** struct xpmsg {
*** 96,101 ****
--- 96,102 ----
  #define	XPMSG15_PAUSECPU	1
  #define	XPMSG_FUNC		4
  #define	XPMSG_FTRP		5
+ #define	XPMSG15_PING		6
  
  	volatile union {
  		/*
*************** extern int bootmid;			/* Module ID of bo
*** 488,493 ****
--- 489,495 ----
  extern struct cpu_info *cpus[];
  #ifdef MULTIPROCESSOR
  extern u_int cpu_ready_mask;		/* the set of CPUs marked as READY */
+ void mp_ping_cpu(struct cpu_info *cpi);	/* are you alive? */
  #endif
  
  #define cpuinfo	(*(struct cpu_info *)CPUINFO_VA)
Index: sparc/intr.c
===================================================================
RCS file: /cvsroot/src/sys/arch/sparc/sparc/intr.c,v
retrieving revision 1.109
diff -p -r1.109 intr.c
*** sparc/intr.c	22 Jan 2011 10:37:22 -0000	1.109
--- sparc/intr.c	26 Jan 2011 06:09:54 -0000
*************** nmi_soft(struct trapframe *tf)
*** 369,374 ****
--- 369,384 ----
  		while (cpuinfo.flags & CPUFLG_PAUSED)
  			/* spin */;
  #endif /* DDB */
+ 		break;
+ 
+ 	case XPMSG15_PING:
+ 		/* XXX printf_nolog() at IPL 15? */
+ 		printf_nolog("cpu%d: ping while pc/npc %#x/%#x psr %#x"
+ 			     " %%o7 %#x\n", cpu_number(),
+ 			     tf->tf_pc, tf->tf_npc, tf->tf_psr,
+ 			     tf->tf_out[7]);
+ 		break;
+ 
  	}
  	cpuinfo.msg_lev15.tag = 0;
  #endif /* MULTIPROCESSOR */
Index: sparc/locore.s
===================================================================
RCS file: /cvsroot/src/sys/arch/sparc/sparc/locore.s,v
retrieving revision 1.259
diff -p -r1.259 locore.s
*** sparc/locore.s	13 Jan 2011 05:20:27 -0000	1.259
--- sparc/locore.s	26 Jan 2011 06:09:54 -0000
*************** _ENTRY(_C_LABEL(nmi_sun4m))
*** 3092,3103 ****
  1:
  #if defined(MULTIPROCESSOR) && defined(DDB)
  	/*
! 	 * Setup a trapframe for nmi_soft; this might be an IPI telling
! 	 * us to pause, so lets save some state for DDB to get at.
  	 */
- 	std	%l0, [%sp + CCFSZ]	! tf.tf_psr = psr; tf.tf_pc = ret_pc;
  	rd	%y, %l3
  	std	%l2, [%sp + CCFSZ + 8]	! tf.tf_npc = return_npc; tf.tf_y = %y;
  	st	%g1, [%sp + CCFSZ + 20]
  	std	%g2, [%sp + CCFSZ + 24]
  	std	%g4, [%sp + CCFSZ + 32]
--- 3092,3105 ----
  1:
  #if defined(MULTIPROCESSOR) && defined(DDB)
  	/*
! 	 * Setup a trapframe for nmi_soft; this might be an IPI telling us
! 	 * to pause or ping, so lets save some state for them to get at.
  	 */
  	rd	%y, %l3
+ 	std	%l0, [%sp + CCFSZ]	! tf.tf_psr = psr; tf.tf_pc = ret_pc;
+ 	rd	%wim, %l4
  	std	%l2, [%sp + CCFSZ + 8]	! tf.tf_npc = return_npc; tf.tf_y = %y;
+ 	st	%l4, [%sp + CCFSZ + 16]	! tf.tf_global[0] = %wim
  	st	%g1, [%sp + CCFSZ + 20]
  	std	%g2, [%sp + CCFSZ + 24]
  	std	%g4, [%sp + CCFSZ + 32]