ARM: highbank: fix cache flush ordering for cpu hotplug
authorRob Herring <rob.herring@calxeda.com>
Wed, 17 Apr 2013 15:46:52 +0000 (10:46 -0500)
committerOlof Johansson <olof@lixom.net>
Thu, 18 Apr 2013 16:37:46 +0000 (09:37 -0700)
The L1 data cache flush needs to be after highbank_set_cpu_jump call which
pollutes the cache with the l2x0_lock. This causes other cores to deadlock
waiting for the l2x0_lock. Moving the flush of the entire data cache after
highbank_set_cpu_jump fixes the problem. Use flush_cache_louis instead of
flush_cache_all are that is sufficient to flush only the L1 data cache.
flush_cache_louis did not exist when highbank_cpu_die was originally
written.

With PL310 errata 769419 enabled, a wmb is inserted into idle which takes
the l2x0_lock. This makes the problem much more easily hit and causes
reset to hang.

Reported-by: Paolo Pisati <p.pisati@gmail.com>
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
arch/arm/mach-highbank/hotplug.c

index f30c528433967e43b1f1afa1050c13feefbbe914..890cae23c12aee40b3f0d675e54d590b13965a5c 100644 (file)
@@ -28,13 +28,11 @@ extern void secondary_startup(void);
  */
 void __ref highbank_cpu_die(unsigned int cpu)
 {
-       flush_cache_all();
-
        highbank_set_cpu_jump(cpu, phys_to_virt(0));
-       highbank_set_core_pwr();
 
-       cpu_do_idle();
+       flush_cache_louis();
+       highbank_set_core_pwr();
 
-       /* We should never return from idle */
-       panic("highbank: cpu %d unexpectedly exit from shutdown\n", cpu);
+       while (1)
+               cpu_do_idle();
 }