powerpc/numa: Disable NEWIDLE balancing at node level
authorAnton Blanchard <anton@samba.org>
Sun, 24 Jul 2011 16:33:15 +0000 (16:33 +0000)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Tue, 20 Sep 2011 05:53:22 +0000 (15:53 +1000)
On big POWER7 boxes we see large amounts of CPU time in system
processes like workqueue and watchdog kernel threads.

We currently rebalance the entire machine each time a task goes
idle and this is very expensive on large machines. Disable newidle
balancing at the node level and rely on the scheduler tick to
rebalance across nodes.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/include/asm/topology.h

index d1c1d312f808be51843ce553a8ed2c2cad1adb31..13efa4d95e769b4260c3f2ba31eb71a27d90e0d8 100644 (file)
@@ -75,7 +75,7 @@ static inline int pcibus_to_node(struct pci_bus *bus)
        .forkexec_idx           = 0,                                    \
                                                                        \
        .flags                  = 1*SD_LOAD_BALANCE                     \
-                               | 1*SD_BALANCE_NEWIDLE                  \
+                               | 0*SD_BALANCE_NEWIDLE                  \
                                | 1*SD_BALANCE_EXEC                     \
                                | 1*SD_BALANCE_FORK                     \
                                | 0*SD_BALANCE_WAKE                     \