From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754755Ab2GBOgp (ORCPT ); Mon, 2 Jul 2012 10:36:45 -0400 Received: from e6.ny.us.ibm.com ([32.97.182.146]:57663 "EHLO e6.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753879Ab2GBOgn (ORCPT ); Mon, 2 Jul 2012 10:36:43 -0400 Date: Mon, 2 Jul 2012 07:22:52 -0700 From: "Paul E. McKenney" To: Sasha Levin Cc: Peter Zijlstra , Dave Jones , "linux-kernel@vger.kernel.org" Subject: Re: rcu: BUG: spinlock recursion on CPU#3, trinity-child19/5970 Message-ID: <20120702142252.GA22868@linux.vnet.ibm.com> Reply-To: paulmck@linux.vnet.ibm.com References: <1340964584.2936.40.camel@lappy> <20120629172320.GA2416@linux.vnet.ibm.com> <1341006040.26928.4.camel@lappy> <1341225139.23484.4.camel@twins> <20120702113541.GI2907@linux.vnet.ibm.com> <1341230352.23484.7.camel@twins> <1341234773.2958.4.camel@lappy> <20120702133334.GA2508@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20120702133334.GA2508@linux.vnet.ibm.com> User-Agent: Mutt/1.5.21 (2010-09-15) x-cbid: 12070214-1976-0000-0000-00000EC96CE1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Jul 02, 2012 at 06:33:34AM -0700, Paul E. McKenney wrote: > On Mon, Jul 02, 2012 at 03:12:53PM +0200, Sasha Levin wrote: > > On Mon, 2012-07-02 at 13:59 +0200, Peter Zijlstra wrote: [ . . . ] > > [ 37.048018] *** DEADLOCK *** > > OK, so it looks like the context-switch-time call into preemptible RCU > needs to be outside of the runqueue-lock critical section. One way > to do this is be reverting 616c310e (Move PREEMPT_RCU preemption to > switch_to() invocation). Sasha, could you please try this out? Hmmmm... Here is a patch for this reversion on top of -rcu with conflicts resolved. Thanx, Paul ------------------------------------------------------------------------ Revert "rcu: Move PREEMPT_RCU preemption to switch_to() invocation" This reverts commit 616c310e83b872024271c915c1b9ab505b9efad9. (Move PREEMPT_RCU preemption to switch_to() invocation). Testing by Sasha Levin showed that this can result in deadlock due to invoking the scheduler when one of the runqueue locks is held. Because this commit was simply a performance optimization, revert it. Conflicts (resolved): include/linux/rcutiny.h diff --git a/arch/um/drivers/mconsole_kern.c b/arch/um/drivers/mconsole_kern.c index 88e466b..43b39d6 100644 --- a/arch/um/drivers/mconsole_kern.c +++ b/arch/um/drivers/mconsole_kern.c @@ -705,7 +705,6 @@ static void stack_proc(void *arg) struct task_struct *from = current, *to = arg; to->thread.saved_task = from; - rcu_switch_from(from); switch_to(from, to, from); } diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h index 2f5d89b..115ead2 100644 --- a/include/linux/rcupdate.h +++ b/include/linux/rcupdate.h @@ -185,7 +185,6 @@ static inline int rcu_preempt_depth(void) /* Internal to kernel */ extern void rcu_sched_qs(int cpu); extern void rcu_bh_qs(int cpu); -extern void rcu_preempt_note_context_switch(void); extern void rcu_check_callbacks(int cpu, int user); struct notifier_block; extern void rcu_idle_enter(void); diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h index 854dc4c..4e56a9c 100644 --- a/include/linux/rcutiny.h +++ b/include/linux/rcutiny.h @@ -87,6 +87,10 @@ static inline void kfree_call_rcu(struct rcu_head *head, #ifdef CONFIG_TINY_RCU +static inline void rcu_preempt_note_context_switch(void) +{ +} + static inline int rcu_needs_cpu(int cpu, unsigned long *delta_jiffies) { *delta_jiffies = ULONG_MAX; @@ -95,6 +99,7 @@ static inline int rcu_needs_cpu(int cpu, unsigned long *delta_jiffies) #else /* #ifdef CONFIG_TINY_RCU */ +void rcu_preempt_note_context_switch(void); int rcu_preempt_needs_cpu(void); static inline int rcu_needs_cpu(int cpu, unsigned long *delta_jiffies) @@ -108,6 +113,7 @@ static inline int rcu_needs_cpu(int cpu, unsigned long *delta_jiffies) static inline void rcu_note_context_switch(int cpu) { rcu_sched_qs(cpu); + rcu_preempt_note_context_switch(); } /* diff --git a/include/linux/sched.h b/include/linux/sched.h index 4059c0f..06a4c5f 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1871,22 +1871,12 @@ static inline void rcu_copy_process(struct task_struct *p) INIT_LIST_HEAD(&p->rcu_node_entry); } -static inline void rcu_switch_from(struct task_struct *prev) -{ - if (prev->rcu_read_lock_nesting != 0) - rcu_preempt_note_context_switch(); -} - #else static inline void rcu_copy_process(struct task_struct *p) { } -static inline void rcu_switch_from(struct task_struct *prev) -{ -} - #endif #ifdef CONFIG_SMP diff --git a/kernel/rcutree.c b/kernel/rcutree.c index b9e6aaf..955f18c 100644 --- a/kernel/rcutree.c +++ b/kernel/rcutree.c @@ -203,6 +203,7 @@ void rcu_note_context_switch(int cpu) { trace_rcu_utilization("Start context switch"); rcu_sched_qs(cpu); + rcu_preempt_note_context_switch(cpu); trace_rcu_utilization("End context switch"); } EXPORT_SYMBOL_GPL(rcu_note_context_switch); diff --git a/kernel/rcutree.h b/kernel/rcutree.h index 86a40b5..b9065dd 100644 --- a/kernel/rcutree.h +++ b/kernel/rcutree.h @@ -467,6 +467,7 @@ DECLARE_PER_CPU(char, rcu_cpu_has_work); /* Forward declarations for rcutree_plugin.h */ static void rcu_bootup_announce(void); long rcu_batches_completed(void); +static void rcu_preempt_note_context_switch(int cpu); static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp); #ifdef CONFIG_HOTPLUG_CPU static void rcu_report_unblock_qs_rnp(struct rcu_node *rnp, diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h index dde457e..cf15a61 100644 --- a/kernel/rcutree_plugin.h +++ b/kernel/rcutree_plugin.h @@ -158,7 +158,7 @@ static void rcu_preempt_qs(int cpu) * * Caller must disable preemption. */ -void rcu_preempt_note_context_switch(void) +static void rcu_preempt_note_context_switch(int cpu) { struct task_struct *t = current; unsigned long flags; @@ -169,7 +169,7 @@ void rcu_preempt_note_context_switch(void) (t->rcu_read_unlock_special & RCU_READ_UNLOCK_BLOCKED) == 0) { /* Possibly blocking in an RCU read-side critical section. */ - rdp = __this_cpu_ptr(rcu_preempt_state.rda); + rdp = per_cpu_ptr(rcu_preempt_state.rda, cpu); rnp = rdp->mynode; raw_spin_lock_irqsave(&rnp->lock, flags); t->rcu_read_unlock_special |= RCU_READ_UNLOCK_BLOCKED; @@ -233,7 +233,7 @@ void rcu_preempt_note_context_switch(void) * means that we continue to block the current grace period. */ local_irq_save(flags); - rcu_preempt_qs(smp_processor_id()); + rcu_preempt_qs(cpu); local_irq_restore(flags); } @@ -902,6 +902,14 @@ void rcu_force_quiescent_state(void) EXPORT_SYMBOL_GPL(rcu_force_quiescent_state); /* + * Because preemptible RCU does not exist, we never have to check for + * CPUs being in quiescent states. + */ +static void rcu_preempt_note_context_switch(int cpu) +{ +} + +/* * Because preemptible RCU does not exist, there are never any preempted * RCU readers. */ diff --git a/kernel/sched/core.c b/kernel/sched/core.c index d5594a4..eaead2d 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2081,7 +2081,6 @@ context_switch(struct rq *rq, struct task_struct *prev, #endif /* Here we just switch the register state and the stack. */ - rcu_switch_from(prev); switch_to(prev, next, prev); barrier();