svn commit: r226057 - head/sys/kern

[Marius Strobl]

Author: marius
Date: Thu Oct  6 11:48:13 2011
New Revision: 226057
URL: http://svn.freebsd.org/changeset/base/226057

Log:
  - Currently, sched_balance_pair() may cause a CPU to send an IPI_PREEMPT to
    itself, which sparc64 hardware doesn't support. One way to solve this
    would be to directly call sched_preempt() instead of issuing a self-IPI.
    However, quoting jhb@:
    "On the other hand, you can probably just skip the IPI entirely if we are
    going to send it to the current CPU.  Presumably, once this routine
    finishes, the current CPU will exit softlock (or will do so "soon") and
    will then pick the next thread to run based on the adjustments made in
    this routine, so there's no need to IPI the CPU running this routine
    anyway.  I think this is the better solution.  Right now what is probably
    happening on other platforms is as soon as this routine finishes the CPU
    processes its self-IPI and causes mi_switch() which will just switch back
    to the softclock thread it is already running."
  - With r226054 and the the above change in place, sparc64 now no longer is
    incompatible with ULE and vice versa. However, powerpc/E500 still is.
  
  Submitted by:	jhb
  Reviewed by:	jeff

Modified:
  head/sys/kern/sched_ule.c

Modified: head/sys/kern/sched_ule.c
==============================================================================
--- head/sys/kern/sched_ule.c	Thu Oct  6 11:20:21 2011	(r226056)
+++ head/sys/kern/sched_ule.c	Thu Oct  6 11:48:13 2011	(r226057)
@@ -76,7 +76,7 @@ dtrace_vtime_switch_func_t	dtrace_vtime_
 #include <machine/cpu.h>
 #include <machine/smp.h>
 
-#if defined(__sparc64__)
+#if defined(__powerpc__) && defined(E500)
 #error "This architecture is not currently compatible with ULE"
 #endif
 
@@ -839,6 +839,7 @@ sched_balance_pair(struct tdq *high, str
 	int low_load;
 	int moved;
 	int move;
+	int cpu;
 	int diff;
 	int i;
 
@@ -860,10 +861,14 @@ sched_balance_pair(struct tdq *high, str
 		for (i = 0; i < move; i++)
 			moved += tdq_move(high, low);
 		/*
-		 * IPI the target cpu to force it to reschedule with the new
-		 * workload.
+		 * In case the target isn't the current cpu IPI it to force a
+		 * reschedule with the new workload.
 		 */
-		ipi_cpu(TDQ_ID(low), IPI_PREEMPT);
+		cpu = TDQ_ID(low);
+		sched_pin();
+		if (cpu != PCPU_GET(cpuid))
+			ipi_cpu(cpu, IPI_PREEMPT);
+		sched_unpin();
 	}
 	tdq_unlock_pair(high, low);
 	return (moved);