From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932799Ab2JEVNJ (ORCPT ); Fri, 5 Oct 2012 17:13:09 -0400 Received: from e2.ny.us.ibm.com ([32.97.182.142]:47480 "EHLO e2.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932275Ab2JEVNE (ORCPT ); Fri, 5 Oct 2012 17:13:04 -0400 Date: Fri, 5 Oct 2012 14:12:34 -0700 From: "Paul E. McKenney" To: Antti P Miettinen Cc: linux-kernel@vger.kernel.org Subject: Re: [PATCH] rcu: Add "expedite always" switch Message-ID: <20121005211234.GA2461@linux.vnet.ibm.com> Reply-To: paulmck@linux.vnet.ibm.com References: <20121005.000053.1467384272006747942.apm@brigitte.kvy.fi> <20121004213933.GI2464@linux.vnet.ibm.com> <20121005.095915.597083345241151143.apm@brigitte.kvy.fi> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20121005.095915.597083345241151143.apm@brigitte.kvy.fi> User-Agent: Mutt/1.5.21 (2010-09-15) x-cbid: 12100521-5112-0000-0000-00000D216CFE Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Oct 05, 2012 at 09:59:15AM +0300, Antti P Miettinen wrote: > From: "Paul E. McKenney" > > b. Maybe sysfs setting. Initially, this can be simple "on" and "off", > > exported with root-only access (like you have above). If more > > elaborate use cases appear, this might become the adminstrative > > override. The key point is that this needs to go through some > > sort of separate function (for example, via the KERNEL_ATTR_RW() > > macro) so that if it needs to change in the future, we can do > > the change easily. > > I did not really envision the module param to be an application > interface, more like an administrative interface. For dynamically > requesting things on/off you probably want something like PM QoS which > does proper book keeping and consolidation but PM QoS for this feels a > bit overkill. But I do not think sysfs node is appropriate either for > an application interface. I guess module_param_cb() would also achieve > the goal of "doing something" upon parameter change. Anyway - below > the 0-perm module param + sysfs node version. Queued, thank you!!! I updated the commit log to read as follows: Add a module parameter to force use of expedited RCU primitives There have been some embedded applications that would benefit from use of expedited grace-period primitives. In some ways, this is similar to synchronize_net() doing either a normal or an expedited grace period depending on lock state, but with control outside of the kernel. This commit therefore adds rcu_expedited boot and sysfs parameters that cause the kernel to substitute expedited primitives for the normal grace-period primitives. Thanx, Paul > --Antti > > Add a module parameter for always using expedited RCU primitives. > > Signed-off-by: Antti P Miettinen > --- > kernel/ksysfs.c | 18 ++++++++++++++++++ > kernel/rcu.h | 2 ++ > kernel/rcupdate.c | 3 +++ > kernel/rcutiny_plugin.h | 5 ++++- > kernel/rcutree.c | 10 ++++++++-- > kernel/rcutree_plugin.h | 5 ++++- > kernel/srcu.c | 6 +++++- > 7 files changed, 44 insertions(+), 5 deletions(-) > > diff --git a/kernel/ksysfs.c b/kernel/ksysfs.c > index 4e316e1..8715a79 100644 > --- a/kernel/ksysfs.c > +++ b/kernel/ksysfs.c > @@ -141,6 +141,23 @@ static ssize_t fscaps_show(struct kobject *kobj, > } > KERNEL_ATTR_RO(fscaps); > > +int rcu_expedited; > +static ssize_t rcu_expedited_show(struct kobject *kobj, > + struct kobj_attribute *attr, char *buf) > +{ > + return sprintf(buf, "%d\n", rcu_expedited); > +} > +static ssize_t rcu_expedited_store(struct kobject *kobj, > + struct kobj_attribute *attr, > + const char *buf, size_t count) > +{ > + if (kstrtoint(buf, 0, &rcu_expedited)) > + return -EINVAL; > + > + return count; > +} > +KERNEL_ATTR_RW(rcu_expedited); > + > /* > * Make /sys/kernel/notes give the raw contents of our kernel .notes section. > */ > @@ -182,6 +199,7 @@ static struct attribute * kernel_attrs[] = { > &kexec_crash_size_attr.attr, > &vmcoreinfo_attr.attr, > #endif > + &rcu_expedited_attr.attr, > NULL > }; > > diff --git a/kernel/rcu.h b/kernel/rcu.h > index 8ba99cd..20dfba5 100644 > --- a/kernel/rcu.h > +++ b/kernel/rcu.h > @@ -109,4 +109,6 @@ static inline bool __rcu_reclaim(char *rn, struct rcu_head *head) > } > } > > +extern int rcu_expedited; > + > #endif /* __LINUX_RCU_H */ > diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c > index 29ca1c6..a2cf761 100644 > --- a/kernel/rcupdate.c > +++ b/kernel/rcupdate.c > @@ -46,12 +46,15 @@ > #include > #include > #include > +#include > > #define CREATE_TRACE_POINTS > #include > > #include "rcu.h" > > +module_param(rcu_expedited, int, 0); > + > #ifdef CONFIG_PREEMPT_RCU > > /* > diff --git a/kernel/rcutiny_plugin.h b/kernel/rcutiny_plugin.h > index 3d01902..f85016a 100644 > --- a/kernel/rcutiny_plugin.h > +++ b/kernel/rcutiny_plugin.h > @@ -706,7 +706,10 @@ void synchronize_rcu(void) > return; > > /* Once we get past the fastpath checks, same code as rcu_barrier(). */ > - rcu_barrier(); > + if (rcu_expedited) > + synchronize_rcu_expedited(); > + else > + rcu_barrier(); > } > EXPORT_SYMBOL_GPL(synchronize_rcu); > > diff --git a/kernel/rcutree.c b/kernel/rcutree.c > index 4fb2376..744c117 100644 > --- a/kernel/rcutree.c > +++ b/kernel/rcutree.c > @@ -2221,7 +2221,10 @@ void synchronize_sched(void) > "Illegal synchronize_sched() in RCU-sched read-side critical section"); > if (rcu_blocking_is_gp()) > return; > - wait_rcu_gp(call_rcu_sched); > + if (rcu_expedited) > + synchronize_sched_expedited(); > + else > + wait_rcu_gp(call_rcu_sched); > } > EXPORT_SYMBOL_GPL(synchronize_sched); > > @@ -2242,7 +2245,10 @@ void synchronize_rcu_bh(void) > "Illegal synchronize_rcu_bh() in RCU-bh read-side critical section"); > if (rcu_blocking_is_gp()) > return; > - wait_rcu_gp(call_rcu_bh); > + if (rcu_expedited) > + synchronize_rcu_bh_expedited(); > + else > + wait_rcu_gp(call_rcu_bh); > } > EXPORT_SYMBOL_GPL(synchronize_rcu_bh); > > diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h > index f921154..0e9ca8b 100644 > --- a/kernel/rcutree_plugin.h > +++ b/kernel/rcutree_plugin.h > @@ -679,7 +679,10 @@ void synchronize_rcu(void) > "Illegal synchronize_rcu() in RCU read-side critical section"); > if (!rcu_scheduler_active) > return; > - wait_rcu_gp(call_rcu); > + if (rcu_expedited) > + synchronize_rcu_expedited(); > + else > + wait_rcu_gp(call_rcu); > } > EXPORT_SYMBOL_GPL(synchronize_rcu); > > diff --git a/kernel/srcu.c b/kernel/srcu.c > index 97c465e..d56873c 100644 > --- a/kernel/srcu.c > +++ b/kernel/srcu.c > @@ -34,6 +34,8 @@ > #include > #include > > +#include "rcu.h" > + > /* > * Initialize an rcu_batch structure to empty. > */ > @@ -464,7 +466,9 @@ static void __synchronize_srcu(struct srcu_struct *sp, int trycount) > */ > void synchronize_srcu(struct srcu_struct *sp) > { > - __synchronize_srcu(sp, SYNCHRONIZE_SRCU_TRYCOUNT); > + __synchronize_srcu(sp, rcu_expedited > + ? SYNCHRONIZE_SRCU_EXP_TRYCOUNT > + : SYNCHRONIZE_SRCU_TRYCOUNT); > } > EXPORT_SYMBOL_GPL(synchronize_srcu); > > -- > 1.7.4.1 >