Message-Id: <20180405171800.5648-1-ynorov@caviumnetworks.com>
Date:   Thu,  5 Apr 2018 20:17:55 +0300
From:   Yury Norov <ynorov@...iumnetworks.com>
To:     "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>,
        Mark Rutland <mark.rutland@....com>,
        Will Deacon <will.deacon@....com>,
        Chris Metcalf <cmetcalf@...lanox.com>,
        Christopher Lameter <cl@...ux.com>,
        Russell King - ARM Linux <linux@...linux.org.uk>,
        Steven Rostedt <rostedt@...dmis.org>,
        Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
        Catalin Marinas <catalin.marinas@....com>,
        Pekka Enberg <penberg@...nel.org>,
        David Rientjes <rientjes@...gle.com>,
        Joonsoo Kim <iamjoonsoo.kim@....com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Michael Ellerman <mpe@...erman.id.au>,
        Alexey Klimov <klimov.linux@...il.com>
Cc:     Yury Norov <ynorov@...iumnetworks.com>,
        linux-arm-kernel@...ts.infradead.org,
        linuxppc-dev@...ts.ozlabs.org, kvm-ppc@...r.kernel.org,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: [PATCH v2 0/2] smp: don't kick CPUs running idle or nohz_full tasks 

kick_all_cpus_sync() is used to broadcast IPIs to all online CPUs to force
them synchronize caches, TLB etc. It is called only 3 times - from mm/slab
arm64 and powerpc code.

We can delay synchronization work for CPUs in extended quiescent state
(idle or nohz_full userspace). 

As Paul E. McKenney wrote: 

--

Currently, IPIs are used to force other CPUs to invalidate their TLBs
in response to a kernel virtual-memory mapping change.  This works, but 
degrades both battery lifetime (for idle CPUs) and real-time response
(for nohz_full CPUs), and in addition results in unnecessary IPIs due to
the fact that CPUs executing in usermode are unaffected by stale kernel
mappings.  It would be better to cause a CPU executing in usermode to
wait until it is entering kernel mode to do the flush, first to avoid
interrupting usemode tasks and second to handle multiple flush requests
with a single flush in the case of a long-running user task.

--

v2 is big rework to address comments in v1:
 - rcu_eqs_special() declaration in public header is dropped, it is not
   used in new implementation. Though, I hope Paul will pick it in his
   tree;
 - for arm64, few isb() added to ensure kernel text synchronization
   (patches 1-4);
 - rcu_get_eqs_cpus() introduced and used to mask EQS CPUs before 
   generating broadcast IPIs;
 - RCU_DYNTICK_CTRL_MASK is not touched because memory barrier is
   implicitly issued in EQS exit path;
 - powerpc is not an exception anymore. I think it's safe to delay
   synchronization for it as well, and I didn't get comments from ppc
   community.
v1:
  https://lkml.org/lkml/2018/3/25/109

Based on next-20180405

Yury Norov (5):
  arm64: entry: isb in el1_irq
  arm64: entry: introduce restore_syscall_args macro
  arm64: ISB early at exit from extended quiescent state
  rcu: arm64: add rcu_dynticks_eqs_exit_sync()
  smp: Lazy synchronization for EQS CPUs in kick_all_cpus_sync()

 arch/arm64/kernel/Makefile  |  2 ++
 arch/arm64/kernel/entry.S   | 52 +++++++++++++++++++++++++++++++--------------
 arch/arm64/kernel/process.c |  7 ++++++
 arch/arm64/kernel/rcu.c     |  8 +++++++
 include/linux/rcutiny.h     |  2 ++
 include/linux/rcutree.h     |  1 +
 kernel/rcu/tiny.c           |  9 ++++++++
 kernel/rcu/tree.c           | 27 +++++++++++++++++++++++
 kernel/smp.c                | 21 +++++++++++-------
 9 files changed, 105 insertions(+), 24 deletions(-)
 create mode 100644 arch/arm64/kernel/rcu.c

-- 
2.14.1

Powered by blists - more mailing lists