linux.git/arch/arc/kernel/smp.c, branch v3.17

ARC: prune extra header includes from smp.c

2014-07-23T05:47:51+00:00

Signed-off-by: Vineet Gupta

ARC: [SMP] unify cpu private IRQ requests (TIMER/IPI)

2014-07-23T05:46:45+00:00

The current cpu-private IRQ registration is ugly as it requires need to
expose arch_unmask_irq() outside of intc code.
So switch to percpu IRQ APIs:
  -request_percpu_irq [boot core]
  -enable_percpu_irq  [all cores]

Encapsulated in helper arc_request_percpu_irq()

Signed-off-by: Vineet Gupta

ARC: [SMP] Fix IPI IRQ registration

2014-06-26T06:28:25+00:00

Handle it just like timer. Current request_percpu_irq() would fail on
non-boot cpus and thus IRQ will remian unmasked on those cpus.

[vgupta: fix changelong]
Signed-off-by: Noam Camus 
Signed-off-by: Vineet Gupta

ARC: arc_local_timer_setup() need not pass own cpu id

2014-06-03T07:56:52+00:00

Signed-off-by: Vineet Gupta

ARC: [SMP] optimize IPI send and receive

2013-12-23T06:35:04+00:00

* Don't send an IPI if receiver already has a pending IPI.
  Atomically piggyback the new msg with pending msg.

* IPI receiver looping on xchg() not required

References: https://lkml.org/lkml/2013/11/25/232
Suggested-by: Peter Zijlstra 
Signed-off-by: Vineet Gupta

ARC: [SMP] simplify IPI code

2013-12-23T06:35:04+00:00

* ipi_data is just a word, no need to keep it as struct

* find_next_bit() not needed to loop thru a 32bit word, ffs suffices

ARC: [SMP] cpu halt interface doesn't need "self" cpu-id

2013-12-23T06:35:04+00:00

Signed-off-by: Vineet Gupta

ARC: [SMP] IPI ACK interface doesn't need "self" cpu-id

2013-12-23T06:35:04+00:00

The interface is confusing, it feels like we are getting "sender" info,
whereas it is the "receiver", which can very well be retrived by
smp_processor_id(), if need be.

Signed-off-by: Vineet Gupta

ARC: [SMP] cpumask not needed in IPI send path

2013-12-23T06:35:04+00:00

The current IPI sending callstack needlessly involves cpumask.

   arch_send_call_function_single_ipi(cpu) / smp_send_reschedule(cpu)
      ipi_send_msg(cpumask_of(cpu))	--> [cpu to cpumask]
         plat_smp_ops.ipi_send(callmap)
            for_each_cpu(callmap)	--> [cpuask to cpu]
               do_plat_specific_ipi_PER_CPU

Given that current backends are not capable of 1:N IPIs, lets simplify
the interface for now, by keeping "a" cpu all along.

Signed-off-by: Vineet Gupta

smp, ARC: kill SMP single function call interrupt

2013-11-07T09:04:31+00:00

Commit 9a46ad6d6df3b54 "smp: make smp_call_function_many() use logic
similar to smp_call_function_single()" has unified the way to handle
single and multiple cross-CPU function calls. Now only one interrupt
is needed for architecture specific code to support generic SMP function
call interfaces, so kill the redundant single function call interrupt.

Signed-off-by: Jiang Liu 
Cc: Jiang Liu 
Signed-off-by: Vineet Gupta