From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: linux-kernel@vger.kernel.org
Cc: mingo@elte.hu, laijs@cn.fujitsu.com, dipankar@in.ibm.com,
akpm@linux-foundation.org, mathieu.desnoyers@polymtl.ca,
josh@joshtriplett.org, dvhltc@us.ibm.com, niv@us.ibm.com,
tglx@linutronix.de, peterz@infradead.org, rostedt@goodmis.org,
Valdis.Kletnieks@vt.edu, dhowells@redhat.com,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Subject: [PATCH RFC tip/core/rcu 03/18] rcu: prohibit starting new grace periods while forcing quiescent states
Date: Tue, 15 Dec 2009 15:02:26 -0800 [thread overview]
Message-ID: <12609181611516-git-send-email-> (raw)
In-Reply-To: <20091215230213.GA9093@linux.vnet.ibm.com>
From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Proposed for 2.6.34, not for inclusion.
Reduce the number and variety of race conditions by prohibiting the
start of a new grace period while force_quiescent_state() is active.
A new fqs_active flag in the rcu_state structure is used to trace
whether or not force_quiescent_state() is active, and this new flag
is tested by rcu_start_gp(). If the CPU that closed out the last
grace period needs another grace period, this new grace period may
be delayed up to one scheduling-clock tick, but it will eventually
get started.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
kernel/rcutree.c | 31 +++++++++++++++++--------------
kernel/rcutree.h | 2 ++
2 files changed, 19 insertions(+), 14 deletions(-)
diff --git a/kernel/rcutree.c b/kernel/rcutree.c
index d42ad30..41688ff 100644
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -659,7 +659,7 @@ rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
struct rcu_data *rdp = rsp->rda[smp_processor_id()];
struct rcu_node *rnp = rcu_get_root(rsp);
- if (!cpu_needs_another_gp(rsp, rdp)) {
+ if (!cpu_needs_another_gp(rsp, rdp) || rsp->fqs_active) {
if (rnp->completed == rsp->completed) {
spin_unlock_irqrestore(&rnp->lock, flags);
return;
@@ -1195,6 +1195,7 @@ static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
struct rcu_node *rnp = rcu_get_root(rsp);
u8 signaled;
u8 forcenow;
+ u8 gpdone;
if (!rcu_gp_in_progress(rsp))
return; /* No grace period in progress, nothing to force. */
@@ -1206,15 +1207,16 @@ static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
(long)(rsp->jiffies_force_qs - jiffies) >= 0)
goto unlock_fqs_ret; /* no emergency and done recently. */
rsp->n_force_qs++;
- spin_lock(&rnp->lock);
+ spin_lock(&rnp->lock); /* irqs already disabled */
lastcomp = rsp->gpnum - 1;
signaled = rsp->signaled;
rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
if(!rcu_gp_in_progress(rsp)) {
rsp->n_force_qs_ngp++;
- spin_unlock(&rnp->lock);
+ spin_unlock(&rnp->lock); /* irqs remain disabled */
goto unlock_fqs_ret; /* no GP in progress, time updated. */
}
+ rsp->fqs_active = 1;
switch (signaled) {
case RCU_GP_IDLE:
case RCU_GP_INIT:
@@ -1223,15 +1225,16 @@ static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
case RCU_SAVE_DYNTICK:
- spin_unlock(&rnp->lock);
+ spin_unlock(&rnp->lock); /* irqs remain disabled */
if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)
break; /* So gcc recognizes the dead code. */
/* Record dyntick-idle state. */
- if (rcu_process_dyntick(rsp, lastcomp,
- dyntick_save_progress_counter))
- goto unlock_fqs_ret;
- spin_lock(&rnp->lock);
+ gpdone = rcu_process_dyntick(rsp, lastcomp,
+ dyntick_save_progress_counter);
+ spin_lock(&rnp->lock); /* irqs already disabled */
+ if (gpdone)
+ break;
/* fall into next case. */
case RCU_SAVE_COMPLETED:
@@ -1252,17 +1255,17 @@ static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
case RCU_FORCE_QS:
/* Check dyntick-idle state, send IPI to laggarts. */
- spin_unlock(&rnp->lock);
- if (rcu_process_dyntick(rsp, rsp->completed_fqs,
- rcu_implicit_dynticks_qs))
- goto unlock_fqs_ret;
+ spin_unlock(&rnp->lock); /* irqs remain disabled */
+ gpdone = rcu_process_dyntick(rsp, rsp->completed_fqs,
+ rcu_implicit_dynticks_qs);
/* Leave state in case more forcing is required. */
- spin_lock(&rnp->lock);
+ spin_lock(&rnp->lock); /* irqs already disabled */
break;
}
- spin_unlock(&rnp->lock);
+ rsp->fqs_active = 0;
+ spin_unlock(&rnp->lock); /* irqs remain disabled */
unlock_fqs_ret:
spin_unlock_irqrestore(&rsp->fqslock, flags);
}
diff --git a/kernel/rcutree.h b/kernel/rcutree.h
index d2a0046..dc386a7 100644
--- a/kernel/rcutree.h
+++ b/kernel/rcutree.h
@@ -277,6 +277,8 @@ struct rcu_state {
u8 signaled ____cacheline_internodealigned_in_smp;
/* Force QS state. */
+ u8 fqs_active; /* force_quiescent_state() */
+ /* is running. */
long gpnum; /* Current gp number. */
long completed; /* # of last completed gp. */
--
1.5.2.5
next prev parent reply other threads:[~2009-12-15 23:03 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-12-15 23:02 [PATCH RFC tip/core/rcu 0/18] rcu: simplify race conditions, add checking Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 01/18] rcu: adjust force_quiescent_state() locking, step 1 Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 02/18] rcu: adjust force_quiescent_state() locking, step 2 Paul E. McKenney
2009-12-15 23:02 ` Paul E. McKenney [this message]
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 04/18] rcu: eliminate local variable signaled from force_quiescent_state() Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 05/18] rcu: eliminate local variable lastcomp " Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 06/18] rcu: eliminate second argument of rcu_process_dyntick() Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 07/18] rcu: eliminate rcu_process_dyntick() return value Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 08/18] rcu: remove leg of force_quiescent_state() switch statement Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 09/18] rcu: remove redundant grace-period check Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 10/18] rcu: make force_quiescent_state() start grace period if needed Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 11/18] rcu: add force_quiescent_state() testing to rcutorture Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 12/18] rcu: make MAINTAINERS file match new RCU reality Paul E. McKenney
2009-12-16 0:53 ` Josh Triplett
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 13/18] rcu: add debug check for too many rcu_read_unlock() Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 14/18] rcu: lockdep check for exiting to user space as RCU reader Paul E. McKenney
2009-12-16 10:24 ` Peter Zijlstra
2009-12-16 15:11 ` Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 15/18] rcu: give different levels of the rcu_node hierarchy distinct lockdep names Paul E. McKenney
2009-12-16 0:59 ` Josh Triplett
2009-12-16 1:59 ` Paul E. McKenney
2009-12-16 10:26 ` Peter Zijlstra
2009-12-16 10:33 ` Peter Zijlstra
2009-12-16 15:13 ` Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 16/18] rcu: make lockdep aware of SRCU read-side critical sections Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 17/18] rcu: Provide different lockdep classes for each flavor of RCU Paul E. McKenney
2009-12-15 23:02 ` [PATCH RFC tip/core/rcu 18/18] rcu: add primitives to check for RCU read-side critical sections Paul E. McKenney
2009-12-16 1:04 ` Josh Triplett
2009-12-16 2:08 ` Paul E. McKenney
2009-12-16 10:31 ` Peter Zijlstra
2009-12-16 15:18 ` Paul E. McKenney
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=12609181611516-git-send-email- \
--to=paulmck@linux.vnet.ibm.com \
--cc=Valdis.Kletnieks@vt.edu \
--cc=akpm@linux-foundation.org \
--cc=dhowells@redhat.com \
--cc=dipankar@in.ibm.com \
--cc=dvhltc@us.ibm.com \
--cc=josh@joshtriplett.org \
--cc=laijs@cn.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@polymtl.ca \
--cc=mingo@elte.hu \
--cc=niv@us.ibm.com \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox