lkml.org 
[lkml]   [2009]   [Oct]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH RFC tip/core/rcu 1/3] rcu: The Bloatwatch Edition, v7
On Mon, Oct 12, 2009 at 05:29:25PM +0800, Lai Jiangshan wrote:

First, thank you very much for looking this over so carefully!!!

> Paul E. McKenney wrote:
> > This patch is a version of RCU designed for !SMP provided for a
> > small-footprint RCU implementation. In particular, the implementation
> > of synchronize_rcu() is extremely lightweight and high performance.
> > It passes rcutorture testing in each of the four relevant configurations
> > (combinations of NO_HZ and PREEMPT) on x86. This saves about 1K bytes
> > compared to old Classic RCU (which is no longer in mainline), and more
> > than three kilobytes compared to Hierarchical RCU (updated to 2.6.30):
> >
> > CONFIG_TREE_RCU:
> >
> > text data bss dec filename
> > 663 32 20 715 kernel/rcupdate.o
> > 3278 528 44 3850 kernel/rcutree.o
> > 4565 Total (vs 4045 for v4)
> >
> > CONFIG_TREE_PREEMPT_RCU:
> >
> > text data bss dec filename
> > 743 32 20 795 kernel/rcupdate.o
> > 4548 752 60 5360 kernel/rcutree.o
> > 6155 Total (N/A for v4)
> >
> > CONFIG_TINY_RCU:
> >
> > text data bss dec filename
> > 96 4 0 100 kernel/rcupdate.o
> > 720 28 0 748 kernel/rcutiny.o
> > 848 Total (vs 1140 for v6)
> >
> > The above is for x86. Your mileage may vary on other platforms.
> > Further compression is possible, but is being procrastinated.
> >
> > Changes from v6 (http://lkml.org/lkml/2009/9/23/293).
> >
> > o Forward ported to put it into the 2.6.33 stream.
> >
> > o Added lockdep support.
> >
> > o Make lightweight rcu_barrier.
> >
> > Changes from v5 (http://lkml.org/lkml/2009/6/23/12).
> >
> > o Ported to latest pre-2.6.32 merge window kernel.
> >
> > - Renamed rcu_qsctr_inc() to rcu_sched_qs().
> > - Renamed rcu_bh_qsctr_inc() to rcu_bh_qs().
> > - Provided trivial rcu_cpu_notify().
> > - Provided trivial exit_rcu().
> > - Provided trivial rcu_needs_cpu().
> > - Fixed up the rcu_*_enter/exit() functions in linux/hardirq.h.
> >
> > o Removed the dependence on EMBEDDED, with a view to making
> > TINY_RCU default for !SMP at some time in the future.
> >
> > o Added (trivial) support for expedited grace periods.
> >
> > Changes from v4 (http://lkml.org/lkml/2009/5/2/91) include:
> >
> > o Squeeze the size down a bit further by removing the
> > ->completed field from struct rcu_ctrlblk.
> >
> > o This permits synchronize_rcu() to become the empty function.
> > Previous concerns about rcutorture were unfounded, as
> > rcutorture correctly handles a constant value from
> > rcu_batches_completed() and rcu_batches_completed_bh().
> >
> > Changes from v3 (http://lkml.org/lkml/2009/3/29/221) include:
> >
> > o Changed rcu_batches_completed(), rcu_batches_completed_bh()
> > rcu_enter_nohz(), rcu_exit_nohz(), rcu_nmi_enter(), and
> > rcu_nmi_exit(), to be static inlines, as suggested by David
> > Howells. Doing this saves about 100 bytes from rcutiny.o.
> > (The numbers between v3 and this v4 of the patch are not directly
> > comparable, since they are against different versions of Linux.)
> >
> > Changes from v2 (http://lkml.org/lkml/2009/2/3/333) include:
> >
> > o Fix whitespace issues.
> >
> > o Change short-circuit "||" operator to instead be "+" in order to
> > fix performance bug noted by "kraai" on LWN.
> >
> > (http://lwn.net/Articles/324348/)
> >
> > Changes from v1 (http://lkml.org/lkml/2009/1/13/440) include:
> >
> > o This version depends on EMBEDDED as well as !SMP, as suggested
> > by Ingo.
> >
> > o Updated rcu_needs_cpu() to unconditionally return zero,
> > permitting the CPU to enter dynticks-idle mode at any time.
> > This works because callbacks can be invoked upon entry to
> > dynticks-idle mode.
> >
> > o Paul is now OK with this being included, based on a poll at the
> > Kernel Miniconf at linux.conf.au, where about ten people said
> > that they cared about saving 900 bytes on single-CPU systems.
> >
> > o Applies to both mainline and tip/core/rcu.
> >
> > Signed-off-by: David Howells <dhowells@redhat.com>
> > Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > ---
> > include/linux/hardirq.h | 24 ++++
> > include/linux/rcupdate.h | 6 +
> > include/linux/rcutiny.h | 103 +++++++++++++++++
> > init/Kconfig | 9 ++
> > kernel/Makefile | 1 +
> > kernel/rcupdate.c | 4 +
> > kernel/rcutiny.c | 281 ++++++++++++++++++++++++++++++++++++++++++++++
> > 7 files changed, 428 insertions(+), 0 deletions(-)
> > create mode 100644 include/linux/rcutiny.h
> > create mode 100644 kernel/rcutiny.c
> >
> > diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
> > index 6d527ee..d5b3876 100644
> > --- a/include/linux/hardirq.h
> > +++ b/include/linux/hardirq.h
> > @@ -139,10 +139,34 @@ static inline void account_system_vtime(struct task_struct *tsk)
> > #endif
> >
> > #if defined(CONFIG_NO_HZ)
> > +#if defined(CONFIG_TINY_RCU)
> > +extern void rcu_enter_nohz(void);
> > +extern void rcu_exit_nohz(void);
> > +
> > +static inline void rcu_irq_enter(void)
> > +{
> > + rcu_exit_nohz();
> > +}
> > +
> > +static inline void rcu_irq_exit(void)
> > +{
> > + rcu_enter_nohz();
> > +}
> > +
> > +static inline void rcu_nmi_enter(void)
> > +{
> > +}
> > +
> > +static inline void rcu_nmi_exit(void)
> > +{
> > +}
> > +
> > +#else
> > extern void rcu_irq_enter(void);
> > extern void rcu_irq_exit(void);
> > extern void rcu_nmi_enter(void);
> > extern void rcu_nmi_exit(void);
> > +#endif
> > #else
> > # define rcu_irq_enter() do { } while (0)
> > # define rcu_irq_exit() do { } while (0)
> > diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> > index 3ebd0b7..6dd71fa 100644
> > --- a/include/linux/rcupdate.h
> > +++ b/include/linux/rcupdate.h
> > @@ -68,11 +68,17 @@ extern int sched_expedited_torture_stats(char *page);
> > /* Internal to kernel */
> > extern void rcu_init(void);
> > extern void rcu_scheduler_starting(void);
> > +#ifndef CONFIG_TINY_RCU
> > extern int rcu_needs_cpu(int cpu);
> > +#else
> > +static inline int rcu_needs_cpu(int cpu) { return 0; }
> > +#endif
> > extern int rcu_scheduler_active;
> >
> > #if defined(CONFIG_TREE_RCU) || defined(CONFIG_TREE_PREEMPT_RCU)
> > #include <linux/rcutree.h>
> > +#elif CONFIG_TINY_RCU
> > +#include <linux/rcutiny.h>
> > #else
> > #error "Unknown RCU implementation specified to kernel configuration"
> > #endif
> > diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h
> > new file mode 100644
> > index 0000000..08f17ab
> > --- /dev/null
> > +++ b/include/linux/rcutiny.h
> > @@ -0,0 +1,103 @@
> > +/*
> > + * Read-Copy Update mechanism for mutual exclusion, the Bloatwatch edition.
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License as published by
> > + * the Free Software Foundation; either version 2 of the License, or
> > + * (at your option) any later version.
> > + *
> > + * This program is distributed in the hope that it will be useful,
> > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> > + * GNU General Public License for more details.
> > + *
> > + * You should have received a copy of the GNU General Public License
> > + * along with this program; if not, write to the Free Software
> > + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
> > + *
> > + * Copyright IBM Corporation, 2008
> > + *
> > + * Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > + *
> > + * For detailed explanation of Read-Copy Update mechanism see -
> > + * Documentation/RCU
> > + */
> > +
> > +#ifndef __LINUX_TINY_H
> > +#define __LINUX_TINY_H
> > +
> > +#include <linux/cache.h>
> > +
> > +/* Global control variables for rcupdate callback mechanism. */
> > +struct rcu_ctrlblk {
> > + struct rcu_head *rcucblist; /* List of pending callbacks (CBs). */
> > + struct rcu_head **donetail; /* ->next pointer of last "done" CB. */
> > + struct rcu_head **curtail; /* ->next pointer of last CB. */
> > +};
> > +
> > +extern struct rcu_ctrlblk rcu_ctrlblk;
> > +extern struct rcu_ctrlblk rcu_bh_ctrlblk;
> > +
> > +void rcu_sched_qs(int cpu);
> > +void rcu_bh_qs(int cpu);
> > +
> > +#define __rcu_read_lock() preempt_disable()
> > +#define __rcu_read_unlock() preempt_enable()
> > +#define __rcu_read_lock_bh() local_bh_disable()
> > +#define __rcu_read_unlock_bh() local_bh_enable()
> > +#define call_rcu_sched call_rcu
> > +
> > +#define rcu_init_sched() do { } while (0)
> > +extern void rcu_check_callbacks(int cpu, int user);
> > +extern void __rcu_init(void);
> > +/* extern void rcu_restart_cpu(int cpu); */
> > +
> > +/*
> > + * Return the number of grace periods.
> > + */
> > +static inline long rcu_batches_completed(void)
> > +{
> > + return 0;
> > +}
> > +
> > +/*
> > + * Return the number of bottom-half grace periods.
> > + */
> > +static inline long rcu_batches_completed_bh(void)
> > +{
> > + return 0;
> > +}
> > +
> > +extern int rcu_expedited_torture_stats(char *page);
> > +
> > +static inline int rcu_pending(int cpu)
> > +{
> > + return 1;
> > +}
> > +
> > +struct notifier_block;
> > +extern int rcu_cpu_notify(struct notifier_block *self,
> > + unsigned long action, void *hcpu);
> > +
> > +#ifdef CONFIG_NO_HZ
> > +
> > +extern void rcu_enter_nohz(void);
> > +extern void rcu_exit_nohz(void);
> > +
> > +#else /* #ifdef CONFIG_NO_HZ */
> > +
> > +static inline void rcu_enter_nohz(void)
> > +{
> > +}
> > +
> > +static inline void rcu_exit_nohz(void)
> > +{
> > +}
> > +
> > +#endif /* #else #ifdef CONFIG_NO_HZ */
> > +
> > +static inline void exit_rcu(void)
> > +{
> > +}
> > +
> > +#endif /* __LINUX_RCUTINY_H */
> > diff --git a/init/Kconfig b/init/Kconfig
> > index 0121c0e..4fecb53 100644
> > --- a/init/Kconfig
> > +++ b/init/Kconfig
> > @@ -334,6 +334,15 @@ config TREE_PREEMPT_RCU
> > is also required. It also scales down nicely to
> > smaller systems.
> >
> > +config TINY_RCU
> > + bool "UP-only small-memory-footprint RCU"
> > + depends on !SMP
> > + help
> > + This option selects the RCU implementation that is
> > + designed for UP systems from which real-time response
> > + is not required. This option greatly reduces the
> > + memory footprint of RCU.
> > +
> > endchoice
> >
> > config RCU_TRACE
> > diff --git a/kernel/Makefile b/kernel/Makefile
> > index 7c9b0a5..0098bcf 100644
> > --- a/kernel/Makefile
> > +++ b/kernel/Makefile
> > @@ -83,6 +83,7 @@ obj-$(CONFIG_RCU_TORTURE_TEST) += rcutorture.o
> > obj-$(CONFIG_TREE_RCU) += rcutree.o
> > obj-$(CONFIG_TREE_PREEMPT_RCU) += rcutree.o
> > obj-$(CONFIG_TREE_RCU_TRACE) += rcutree_trace.o
> > +obj-$(CONFIG_TINY_RCU) += rcutiny.o
> > obj-$(CONFIG_RELAY) += relay.o
> > obj-$(CONFIG_SYSCTL) += utsname_sysctl.o
> > obj-$(CONFIG_TASK_DELAY_ACCT) += delayacct.o
> > diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c
> > index 4001833..7625f20 100644
> > --- a/kernel/rcupdate.c
> > +++ b/kernel/rcupdate.c
> > @@ -67,6 +67,8 @@ void wakeme_after_rcu(struct rcu_head *head)
> > complete(&rcu->completion);
> > }
> >
> > +#ifndef CONFIG_TINY_RCU
> > +
> > #ifdef CONFIG_TREE_PREEMPT_RCU
> >
> > /**
> > @@ -157,6 +159,8 @@ void synchronize_rcu_bh(void)
> > }
> > EXPORT_SYMBOL_GPL(synchronize_rcu_bh);
> >
> > +#endif /* #ifndef CONFIG_TINY_RCU */
> > +
> > static int __cpuinit rcu_barrier_cpu_hotplug(struct notifier_block *self,
> > unsigned long action, void *hcpu)
> > {
> > diff --git a/kernel/rcutiny.c b/kernel/rcutiny.c
> > new file mode 100644
> > index 0000000..89124b0
> > --- /dev/null
> > +++ b/kernel/rcutiny.c
> > @@ -0,0 +1,281 @@
> > +/*
> > + * Read-Copy Update mechanism for mutual exclusion, the Bloatwatch edition.
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License as published by
> > + * the Free Software Foundation; either version 2 of the License, or
> > + * (at your option) any later version.
> > + *
> > + * This program is distributed in the hope that it will be useful,
> > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> > + * GNU General Public License for more details.
> > + *
> > + * You should have received a copy of the GNU General Public License
> > + * along with this program; if not, write to the Free Software
> > + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
> > + *
> > + * Copyright IBM Corporation, 2008
> > + *
> > + * Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > + *
> > + * For detailed explanation of Read-Copy Update mechanism see -
> > + * Documentation/RCU
> > + */
> > +
> > +#include <linux/types.h>
> > +#include <linux/kernel.h>
> > +#include <linux/init.h>
> > +#include <linux/rcupdate.h>
> > +#include <linux/interrupt.h>
> > +#include <linux/sched.h>
> > +#include <linux/module.h>
> > +#include <linux/completion.h>
> > +#include <linux/moduleparam.h>
> > +#include <linux/notifier.h>
> > +#include <linux/cpu.h>
> > +#include <linux/mutex.h>
> > +#include <linux/time.h>
> > +
> > +/* Definition for rcupdate control block. */
> > +struct rcu_ctrlblk rcu_ctrlblk = {
> > + .rcucblist = NULL,
> > + .donetail = &rcu_ctrlblk.rcucblist,
> > + .curtail = &rcu_ctrlblk.rcucblist,
> > +};
> > +EXPORT_SYMBOL_GPL(rcu_ctrlblk);
> > +struct rcu_ctrlblk rcu_bh_ctrlblk = {
> > + .rcucblist = NULL,
> > + .donetail = &rcu_bh_ctrlblk.rcucblist,
> > + .curtail = &rcu_bh_ctrlblk.rcucblist,
> > +};
> > +EXPORT_SYMBOL_GPL(rcu_bh_ctrlblk);
> > +
> > +#ifdef CONFIG_NO_HZ
> > +
> > +static long rcu_dynticks_nesting = 1;
> > +
> > +/*
> > + * Enter dynticks-idle mode, which is an extended quiescent state
> > + * if we have fully entered that mode (i.e., if the new value of
> > + * dynticks_nesting is zero).
> > + */
> > +void rcu_enter_nohz(void)
> > +{
> > + if (--rcu_dynticks_nesting == 0)
> > + rcu_sched_qs(0); /* implies rcu_bh_qsctr_inc(0) */
> > +}
> > +
> > +/*
> > + * Exit dynticks-idle mode, so that we are no longer in an extended
> > + * quiescent state.
> > + */
> > +void rcu_exit_nohz(void)
> > +{
> > + rcu_dynticks_nesting++;
> > +}
> > +
> > +#endif /* #ifdef CONFIG_NO_HZ */
> > +
> > +/*
> > + * Helper function for rcu_qsctr_inc() and rcu_bh_qsctr_inc().
> > + */
> > +static int rcu_qsctr_help(struct rcu_ctrlblk *rcp)
> > +{
> > + if (rcp->rcucblist != NULL &&
> > + rcp->donetail != rcp->curtail) {
> > + rcp->donetail = rcp->curtail;
> > + return 1;
> > + }
> > + return 0;
> > +}
> > +
> > +/*
> > + * Record an rcu quiescent state. And an rcu_bh quiescent state while we
> > + * are at it, given that any rcu quiescent state is also an rcu_bh
> > + * quiescent state. Use "+" instead of "||" to defeat short circuiting.
> > + */
> > +void rcu_sched_qs(int cpu)
> > +{
> > + if (rcu_qsctr_help(&rcu_ctrlblk) + rcu_qsctr_help(&rcu_bh_ctrlblk))
> > + raise_softirq(RCU_SOFTIRQ);
>
>
> local_irq_disable()(better) or local_bh_disable() is needed here.
>
> see here:
> schedule() {
> ...
> preempt_disable();
> ....
> rcu_sched_qs(cpu); /* nothing to proctect accessing rcp->donetail */
> .....
> }

Good eyes!!!

Otherwise, an interrupt might be taken from within rcu_qsctr_help(),
and the interrupt handler might invoke call_rcu(), which could fatally
confuse rcu_qsctr_help(). Not needed for treercu, since treercu's
version of rcu_sched_qs() just mucks with flags (famous last words!).

Fixed.

> > +}
> > +
> > +/*
> > + * Record an rcu_bh quiescent state.
> > + */
> > +void rcu_bh_qs(int cpu)
> > +{
> > + if (rcu_qsctr_help(&rcu_bh_ctrlblk))
> > + raise_softirq(RCU_SOFTIRQ);
>
>
> It doesn't need local_irq_disable() nor local_bh_disable().
> It's only called at __do_softirq(), but maybe a comment is needed.

Hmmmm... Does this apply even when called from ksoftirqd?

Adding the local_irq_save() just out of paranoia for the moment.
And therefore moving the local_irq_save() down to the common
rcu_qsctr_help() function. Which I am renaming to rcu_qs_help()
for consistency.

Or am I missing something here?

> > +}
> > +
> > +/*
> > + * Check to see if the scheduling-clock interrupt came from an extended
> > + * quiescent state, and, if so, tell RCU about it.
> > + */
> > +void rcu_check_callbacks(int cpu, int user)
> > +{
> > + if (!rcu_needs_cpu(0))
> > + return; /* RCU doesn't need anything to be done. */
>
> rcu_needs_cpu(0) always returns 0 ......
> The next statements will not be executed .....

Indeed! Should instead be rcu_pending() -- which always returns 1.
So just deleting the above "if" statement.

The theory behing rcu_needs_cpu() always returning zero is that
rcu_enter_nohz() will be invoked on the way to no_hz mode, which
will invoke rcu_sched_qs(), which will update callbacks and do
raise_softirq(), causing any extant callbacks to be invoked.

Seem reasonable?

> > + if (user ||
> > + (idle_cpu(cpu) &&
> > + !in_softirq() &&
> > + hardirq_count() <= (1 << HARDIRQ_SHIFT)))
> > + rcu_sched_qs(cpu);
> > + else if (!in_softirq())
> > + rcu_bh_qs(cpu);
> > +}
> > +
> > +/*
> > + * Helper function for rcu_process_callbacks() that operates on the
> > + * specified rcu_ctrlkblk structure.
> > + */
> > +static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp)
> > +{
> > + unsigned long flags;
> > + struct rcu_head *next, *list;
> > +
> > + /* If no RCU callbacks ready to invoke, just return. */
> > + if (&rcp->rcucblist == rcp->donetail)
> > + return;
> > +
> > + /* Move the ready-to-invoke callbacks to a local list. */
> > + local_irq_save(flags);
> > + list = rcp->rcucblist;
> > + rcp->rcucblist = *rcp->donetail;
> > + *rcp->donetail = NULL;
> > + if (rcp->curtail == rcp->donetail)
> > + rcp->curtail = &rcp->rcucblist;
> > + rcp->donetail = &rcp->rcucblist;
> > + local_irq_restore(flags);
> > +
> > + /* Invoke the callbacks on the local list. */
> > + while (list) {
> > + next = list->next;
> > + prefetch(next);
> > + list->func(list);
> > + list = next;
> > + }
> > +}
> > +
> > +/*
> > + * Invoke any callbacks whose grace period has completed.
> > + */
> > +static void rcu_process_callbacks(struct softirq_action *unused)
> > +{
> > + __rcu_process_callbacks(&rcu_ctrlblk);
> > + __rcu_process_callbacks(&rcu_bh_ctrlblk);
> > +}
> > +
> > +/*
> > + * Null function to handle CPU being onlined. Longer term, we want to
> > + * make TINY_RCU avoid using rcupdate.c, but later...
> > + */
> > +int rcu_cpu_notify(struct notifier_block *self,
> > + unsigned long action, void *hcpu)
> > +{
> > + return NOTIFY_OK;
> > +}
> > +
> > +/*
> > + * Wait for a grace period to elapse. But it is illegal to invoke
> > + * synchronize_sched() from within an RCU read-side critical section.
> > + * Therefore, any legal call to synchronize_sched() is a quiescent
> > + * state, and so on a UP system, synchronize_sched() need do nothing.
> > + * Ditto for synchronize_rcu_bh().
> > + *
> > + * Cool, huh? (Due to Josh Triplett.)
> > + *
> > + * But we want to make this a static inline later.
> > + */
> > +void synchronize_sched(void)
> > +{
>
> I stubbornly recommend adding a cond_resched()/might_sleep() here.
>
> It reduces latency. (for !CONFIG_PREEMPT)
> It prevents someone calls it on nonsleepable context.

Good point, fixed.

> > +}
> > +EXPORT_SYMBOL_GPL(synchronize_sched);
> > +
> > +void synchronize_rcu_bh(void)
> > +{
>
> Ditto.

Just made this invoke synchronize_sched().

> > +}
> > +EXPORT_SYMBOL_GPL(synchronize_rcu_bh);
> > +
> > +/*
> > + * Helper function for call_rcu() and call_rcu_bh().
> > + */
> > +static void __call_rcu(struct rcu_head *head,
> > + void (*func)(struct rcu_head *rcu),
> > + struct rcu_ctrlblk *rcp)
> > +{
> > + unsigned long flags;
> > +
> > + head->func = func;
> > + head->next = NULL;
> > + local_irq_save(flags);
> > + *rcp->curtail = head;
> > + rcp->curtail = &head->next;
> > + local_irq_restore(flags);
> > +}
> > +
> > +/*
> > + * Post an RCU callback to be invoked after the end of an RCU grace
> > + * period. But since we have but one CPU, that would be after any
> > + * quiescent state.
> > + */
> > +void call_rcu(struct rcu_head *head,
> > + void (*func)(struct rcu_head *rcu))
> > +{
> > + __call_rcu(head, func, &rcu_ctrlblk);
> > +}
> > +EXPORT_SYMBOL_GPL(call_rcu);
> > +
> > +/*
> > + * Post an RCU bottom-half callback to be invoked after any subsequent
> > + * quiescent state.
> > + */
> > +void call_rcu_bh(struct rcu_head *head,
> > + void (*func)(struct rcu_head *rcu))
> > +{
> > + __call_rcu(head, func, &rcu_bh_ctrlblk);
> > +}
> > +EXPORT_SYMBOL_GPL(call_rcu_bh);
> > +
> > +void rcu_barrier(void)
> > +{
> > + struct rcu_synchronize rcu;
> > +
> > + init_completion(&rcu.completion);
> > + /* Will wake me after RCU finished. */
> > + call_rcu(&rcu.head, wakeme_after_rcu);
> > + /* Wait for it. */
> > + wait_for_completion(&rcu.completion);
> > +}
> > +EXPORT_SYMBOL_GPL(rcu_barrier);
> > +
> > +void rcu_barrier_bh(void)
> > +{
> > + struct rcu_synchronize rcu;
> > +
> > + init_completion(&rcu.completion);
> > + /* Will wake me after RCU finished. */
> > + call_rcu_bh(&rcu.head, wakeme_after_rcu);
> > + /* Wait for it. */
> > + wait_for_completion(&rcu.completion);
> > +}
> > +EXPORT_SYMBOL_GPL(rcu_barrier_bh);
> > +
> > +void rcu_barrier_sched(void)
> > +{
> > + struct rcu_synchronize rcu;
> > +
> > + init_completion(&rcu.completion);
> > + /* Will wake me after RCU finished. */
> > + call_rcu_sched(&rcu.head, wakeme_after_rcu);
> > + /* Wait for it. */
> > + wait_for_completion(&rcu.completion);
>
>
> alternative implementation(nonsleep implementation)
>
> {
> cond_resched();
>
> rcp = &rcu_ctrlblk;
> local_irq_save(flags);
> if (rcp->rcucblist != NULL) {
> rcp->donetail = rcp->curtail;
> local_irq_restore(flags);
>
> local_bh_disable();
> __rcu_process_callbacks(rcp);
> local_bh_enable();
> } else
> local_irq_restore(flags);
> }
>
> Ditto for other rcu_barrier*()

I certainly can check for there being no callbacks present, and just
return in that case (with the cond_resched()). Except that this is
TINY_RCU, where code size is the biggest issue, and people had better
not be using rcu_barrier() on latency-sensitive code paths. (In
contrast, the low-latency synchronize_rcu() trick results in both
smaller code and lower latency.)

The concern I have with simply executing the callbacks directly is that
it might one day be necessary to throttle callback execution. I do not
believe that this will happen because there is only a single CPU, but I
would like to make it easy to switch back to throttling should someone
prove me wrong.

So I am holding off on this for the moment, but if it turns out that
throttling is never necessary, this would be an attractive optimization.

> > +}
> > +EXPORT_SYMBOL_GPL(rcu_barrier_sched);
> > +
> > +void __rcu_init(void)
> > +{
> > + open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
> > +}

Again, thank you for your careful review and thoughtful comments!!!

Thanx, Paul


\
 
 \ /
  Last update: 2009-10-12 19:35    [W:0.993 / U:1.180 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site