Nohz / isolation fixes

Ingo,

Please pull the nohz/fixes branch that can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks.git
nohz/fixes

HEAD: 11c3146353bb83028ace8ad089a911c390395986

Thanks,
Frederic
---

Paul E. McKenney (3):
sched/isolation: Make NO_HZ_FULL select CPU_ISOLATION sched/isolation: Enable CONFIG_CPU_ISOLATION=y by default sched/isolation: Document boot parameters dependency on CONFIG_CPU_ISOLATION

 Documentation/admin-guide/kernel-parameters.rst | 1 +
 Documentation/admin-guide/kernel-parameters.txt | 4 ++--
 init/Kconfig                                    | 6 +++++-
 kernel/time/Kconfig                             | 1 +
 4 files changed, 9 insertions(+), 3 deletions(-)

From: "Paul E. McKenney" <[email protected]>

CONFIG_NO_HZ_FULL doesn't make sense without CONFIG_CPU_ISOLATION. In fact enabling the first without the second is a regression as nohz_full= boot parameter gets silently ignored.

Besides this unnatural combination hangs RCU gp kthread when running rcutorture for reasons that are not yet fully understood:

	rcu_preempt kthread starved for 9974 jiffies! g4294967208
	+c4294967207 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x402 ->cpu=0
	rcu_preempt     I 7464     8      2 0x80000000
	Call Trace:
		__schedule+0x493/0x620
		schedule+0x24/0x40
		schedule_timeout+0x330/0x3b0
		? preempt_count_sub+0xea/0x140
		? collect_expired_timers+0xb0/0xb0
		rcu_gp_kthread+0x6bf/0xef0

This commit therefore makes NO_HZ_FULL select CPU_ISOLATION, which prevents all these bad behaviours.

---
 kernel/time/Kconfig | 1 +
 1 file changed, 1 insertion(+)

index e776fc8..f6b5f19 100644
@@ -95,6 +95,7 @@ config NO_HZ_FULL
 	select RCU_NOCB_CPU
 	select VIRT_CPU_ACCOUNTING_GEN
 	select IRQ_WORK
+	select CPU_ISOLATION
 	help
 	 Adaptively try to shutdown the tick whenever possible, even when
 	 the CPU is running tasks. Typically this requires running a single

2.7.4

From: "Paul E. McKenney" <[email protected]>

The "isolcpus=" boot parameter support was always built-in before we moved the related code under CONFIG_CPU_ISOLATION. Having it disabled by default is very confusing for people accustomed to use this parameter.

So enable it by dafault to keep the previous behaviour but keep it optable for those who want to tinify their kernels.

---
 init/Kconfig | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

index 2934249..690a381 100644
@@ -461,10 +461,14 @@ endmenu # "CPU/Task time and stats accounting"
 
 config CPU_ISOLATION
 	bool "CPU isolation"
+	default y
 	help
 	  Make sure that CPUs running critical tasks are not disturbed by
 	  any source of "noise" such as unbound workqueues, timers, kthreads...
-	  Unbound jobs get offloaded to housekeeping CPUs.
+	  Unbound jobs get offloaded to housekeeping CPUs. This is driven by
+	  the "isolcpus=" boot parameter.
+
+	  Say Y if unsure.
 
 source "kernel/rcu/Kconfig"
 

2.7.4

From: "Paul E. McKenney" <[email protected]>

The "isolcpus=" and "nohz_full=" boot parameters depend on CPU Isolation support. Let's document that.

---
 Documentation/admin-guide/kernel-parameters.rst | 1 +
 Documentation/admin-guide/kernel-parameters.txt | 4 ++--
 2 files changed, 3 insertions(+), 2 deletions(-)

index b2598cc..7242cbd 100644
@@ -109,6 +109,7 @@ parameter is applicable::
 	IPV6	IPv6 support is enabled.
 	ISAPNP	ISA PnP code is enabled.
 	ISDN	Appropriate ISDN support is enabled.
+	ISOL	CPU Isolation is enabled.
 	JOY	Appropriate joystick support is enabled.
 	KGDB	Kernel debugger support is enabled.
 	KVM	Kernel Virtual Machine support is enabled.

index 6571fbf..1683107 100644
@@ -1737,7 +1737,7 @@
 	isapnp=		[ISAPNP]
 			Format: <RDP>,<reset>,<pci_scan>,<verbosity>
 
-	isolcpus=	[KNL,SMP] Isolate a given set of CPUs from disturbance.
+	isolcpus=	[KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
 			[Deprecated - use cpusets instead]
 			Format: [flag-list,]<cpu-list>
 

@@ -2662,7 +2662,7 @@
 			Valid arguments: on, off
 			Default: on
 
-	nohz_full=	[KNL,BOOT]
+	nohz_full=	[KNL,BOOT,SMP,ISOL]
 			The argument is a cpu list, as described above.
 			In kernels built with CONFIG_NO_HZ_FULL=y, set
 			the specified list of CPUs whose tick will be stopped

2.7.4

2017-12-14 15:14 UTC+01:00, Frederic Weisbecker <[email protected]>:

Sorry please disregard this pull request, I messed up with the autorship. I'm sending another version.