linux-kernel - Re: [PATCH] rcu: Convert rcu_state.ofl_lock to raw_spinlock

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-Id: <20180815161943.GN24813@linux.vnet.ibm.com>
Date:   Wed, 15 Aug 2018 09:19:43 -0700
From:   "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To:     Mike Galbraith <efault@....de>
Cc:     Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] rcu: Convert rcu_state.ofl_lock to raw_spinlock_t

On Wed, Aug 15, 2018 at 05:44:05PM +0200, Mike Galbraith wrote:
> 
> 1e64b15a4b10 ("rcu: Fix grace-period hangs due to race with CPU offline")
> added spinlock_t ofl_lock to the rcu_state structure, then takes it with
> preemption disabled during CPU offline, giving RT sleeping lock heartburn.
> 
> Convert it to raw_spinlock_t.
> 
> Signed-off-by: Mike Galbraith <efault@....de>

Good catch, applied and started testing.

I had to hand-apply this one due to the RCU flavor consolidation churn.
Let's just say that your patch is the 160th in -rcu slated for the next
merge window, putting us about 30% of the way to the Greg KH criterion.  ;-)
(That said, about 25 of those patches are LKMM rather than RCU.)

So please see below for my adaptation of your patch.  Please let me know
if I messed anything up.

							Thanx, Paul

------------------------------------------------------------------------

commit f5bc811f274bb5ff6da5f84f6928d00711c28f09
Author: Mike Galbraith <efault@....de>
Date:   Wed Aug 15 09:05:29 2018 -0700

    rcu: Convert rcu_state.ofl_lock to raw_spinlock_t
    
    1e64b15a4b10 ("rcu: Fix grace-period hangs due to race with CPU offline")
    added spinlock_t ofl_lock to the rcu_state structure, then takes it with
    preemption disabled during CPU offline, which gives the -rt patchset's
    sleeping spinlock heartburn.
    
    This commit therefore converts ->ofl_lock to raw_spinlock_t.
    
    Signed-off-by: Mike Galbraith <efault@....de>
    Signed-off-by: Paul E. McKenney <paulmck@...ux.vnet.ibm.com>
    Cc: Sebastian Andrzej Siewior <bigeasy@...utronix.de>

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index bf016ff9f873..7f7008dd9c34 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -97,7 +97,7 @@ struct rcu_state rcu_state = {
 	.abbr = RCU_ABBR,
 	.exp_mutex = __MUTEX_INITIALIZER(rcu_state.exp_mutex),
 	.exp_wake_mutex = __MUTEX_INITIALIZER(rcu_state.exp_wake_mutex),
-	.ofl_lock = __SPIN_LOCK_UNLOCKED(rcu_state.ofl_lock),
+	.ofl_lock = __RAW_SPIN_LOCK_UNLOCKED(rcu_state.ofl_lock),
 };
 
 /* Dump rcu_node combining tree at boot to verify correct setup. */
@@ -1775,13 +1775,13 @@ static bool rcu_gp_init(void)
 	 */
 	rcu_state.gp_state = RCU_GP_ONOFF;
 	rcu_for_each_leaf_node(rnp) {
-		spin_lock(&rcu_state.ofl_lock);
+		raw_spin_lock(&rcu_state.ofl_lock);
 		raw_spin_lock_irq_rcu_node(rnp);
 		if (rnp->qsmaskinit == rnp->qsmaskinitnext &&
 		    !rnp->wait_blkd_tasks) {
 			/* Nothing to do on this leaf rcu_node structure. */
 			raw_spin_unlock_irq_rcu_node(rnp);
-			spin_unlock(&rcu_state.ofl_lock);
+			raw_spin_unlock(&rcu_state.ofl_lock);
 			continue;
 		}
 
@@ -1817,7 +1817,7 @@ static bool rcu_gp_init(void)
 		}
 
 		raw_spin_unlock_irq_rcu_node(rnp);
-		spin_unlock(&rcu_state.ofl_lock);
+		raw_spin_unlock(&rcu_state.ofl_lock);
 	}
 	rcu_gp_slow(gp_preinit_delay); /* Races with CPU hotplug. */
 
@@ -3376,7 +3376,7 @@ void rcu_report_dead(unsigned int cpu)
 
 	/* Remove outgoing CPU from mask in the leaf rcu_node structure. */
 	mask = rdp->grpmask;
-	spin_lock(&rcu_state.ofl_lock);
+	raw_spin_lock(&rcu_state.ofl_lock);
 	raw_spin_lock_irqsave_rcu_node(rnp, flags); /* Enforce GP memory-order guarantee. */
 	rdp->rcu_ofl_gp_seq = READ_ONCE(rcu_state.gp_seq);
 	rdp->rcu_ofl_gp_flags = READ_ONCE(rcu_state.gp_flags);
@@ -3387,7 +3387,7 @@ void rcu_report_dead(unsigned int cpu)
 	}
 	rnp->qsmaskinitnext &= ~mask;
 	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
-	spin_unlock(&rcu_state.ofl_lock);
+	raw_spin_unlock(&rcu_state.ofl_lock);
 
 	per_cpu(rcu_cpu_started, cpu) = 0;
 }
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
index bfbf97a1c29d..703e19ff532d 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -343,7 +343,7 @@ struct rcu_state {
 	const char *name;			/* Name of structure. */
 	char abbr;				/* Abbreviated name. */
 
-	spinlock_t ofl_lock ____cacheline_internodealigned_in_smp;
+	raw_spinlock_t ofl_lock ____cacheline_internodealigned_in_smp;
 						/* Synchronize offline with */
 						/*  GP pre-initialization. */
 };