rpms/kernel/devel linux-2.6-ondemand-timer.patch,1.2,1.3

Wed Mar 21 21:04:09 UTC 2007

Author: davej

Update of /cvs/dist/rpms/kernel/devel
In directory cvs.devel.redhat.com:/tmp/cvs-serv26357

Modified Files:
	linux-2.6-ondemand-timer.patch 
Log Message:
* Wed Mar 21 2007 Dave Jones <davej at redhat.com>
- New version of the ondemand 'no wakeup whilst idle' patch


linux-2.6-ondemand-timer.patch:
 linux-2.6.20/arch/powerpc/platforms/chrp/setup.c |    2 -
 linux-2.6.20/include/linux/timer.h               |   41 +++++++++++++++++++++--
 linux-2.6.20/kernel/timer.c                      |   31 +++++++++++++----
 linux-2.6.20/kernel/workqueue.c                  |    2 -
 new/drivers/cpufreq/cpufreq_ondemand.c           |    2 -
 new/include/linux/workqueue.h                    |    6 +++
 6 files changed, 72 insertions(+), 12 deletions(-)

Index: linux-2.6-ondemand-timer.patch
===================================================================
RCS file: /cvs/dist/rpms/kernel/devel/linux-2.6-ondemand-timer.patch,v
retrieving revision 1.2
retrieving revision 1.3
diff -u -r1.2 -r1.3

--- linux-2.6-ondemand-timer.patch	21 Mar 2007 20:35:33 -0000	1.2
+++ linux-2.6-ondemand-timer.patch	21 Mar 2007 21:04:07 -0000	1.3
@@ -1,3 +1,285 @@
+From davej  Fri Mar 16 18:08:12 2007
+Return-Path: <venkatesh.pallipadi at intel.com>
+X-Spam-Checker-Version: SpamAssassin 3.1.8 (2007-02-13) on
+	gelk.kernelslacker.org
+X-Spam-Level: 
+X-Spam-Status: No, score=-2.2 required=5.0 tests=AWL,BAYES_00,
+	UNPARSEABLE_RELAY autolearn=ham version=3.1.8
+Received: from pobox.devel.redhat.com [10.11.255.8]
+	by gelk.kernelslacker.org with IMAP (fetchmail-6.3.6)
+	for <davej at localhost> (single-drop); Fri, 16 Mar 2007 18:08:12 -0400 (EDT)
+Received: from pobox.devel.redhat.com ([unix socket])
+	 by pobox.devel.redhat.com (Cyrus v2.2.12-Invoca-RPM-2.2.12-3.RHEL4.1) with LMTPA;
+	 Fri, 16 Mar 2007 18:07:54 -0400
+X-Sieve: CMU Sieve 2.2
+Received: from int-mx2.corp.redhat.com (int-mx2.corp.redhat.com [172.16.27.26])
+	by pobox.devel.redhat.com (8.13.1/8.13.1) with ESMTP id l2GM7rEt029639
+	for <davej at pobox.devel.redhat.com>; Fri, 16 Mar 2007 18:07:53 -0400
+Received: from mx2.redhat.com (mx2.redhat.com [10.255.15.25])
+	by int-mx2.corp.redhat.com (8.13.1/8.13.1) with ESMTP id l2GM7qOD031111
+	for <davej at redhat.com>; Fri, 16 Mar 2007 18:07:53 -0400
+Received: from mga03.intel.com (mga03.intel.com [143.182.124.21])
+	by mx2.redhat.com (8.13.1/8.13.1) with ESMTP id l2GM7pkX009716
+	for <davej at redhat.com>; Fri, 16 Mar 2007 18:07:51 -0400
+Received: from azsmga001.ch.intel.com ([10.2.17.19])
+  by mga03.intel.com with ESMTP; 16 Mar 2007 15:07:46 -0700
+Received: from linux-os.sc.intel.com ([172.25.110.8])
+  by azsmga001.ch.intel.com with ESMTP; 16 Mar 2007 15:07:46 -0700
+X-ExtLoop1: 1
+X-IronPort-AV: i="4.14,293,1170662400"; 
+   d="scan'208"; a="197837429:sNHT19716802"
+Received: by linux-os.sc.intel.com (Postfix, from userid 47009)
+	id 1458A28006; Fri, 16 Mar 2007 15:07:35 -0700 (PDT)
+Date: Fri, 16 Mar 2007 15:07:35 -0700
+From: Venkatesh Pallipadi <venkatesh.pallipadi at intel.com>
+To: Andrew Morton <akpm at osdl.org>, Dave Jones <davej at redhat.com>,
+        tglx at linutronix.de
+Cc: linux-kernel <linux-kernel at vger.kernel.org>
+Subject: [PATCH 1/2] Add not_critical_when_idle timer
+Message-ID: <20070316220734.GA6109 at linux-os.sc.intel.com>
+Mime-Version: 1.0
+Content-Type: text/plain; charset=us-ascii
+Content-Disposition: inline
+User-Agent: Mutt/1.4.1i
+X-RedHat-Spam-Score: -1.44 
+Status: RO
+Content-Length: 8124
+Lines: 233
+
+
+Introduce a new kind of timers - not_critical_when_idle timers:
+Timers that work normally when system is busy. But, will not cause CPU to
+come out of idle (just to service this timer), when CPU is idle. Instead,
+this timer will be serviced when CPU eventually wakes up with a subsequent
+critical_when_idle timer.
+
+The main advantage of this is to avoid unnecessary timer interrupts when
+CPU is idle. If the routine currently called by a timer can wait until next
+event without any issues, this new timer can be used to setup timer event
+for that routine. This, with dynticks, allows CPUs to be lazy, allowing them
+to stay in idle for extended period of time by reducing unnecesary wakeup and
+thereby reducing the power consumption.
+
+This patch:
+Builds this new timer on top of existing timer infrastructure. It uses
+last bit in 'base' pointer of timer_list structure to store this
+extra information about timer. __next_timer_interrupt() function
+skips over these not_critical_when_idle timers when CPU looks for
+next timer event for which it has to wake up.
+
+This is exported by a new interface add_timer_with_hint() and also a new
+parameter is added to existing add_timer_on() interface.
+
+Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi at intel.com>
+
+Index: linux-2.6.20/kernel/timer.c
+===================================================================
+--- linux-2.6.20.orig/kernel/timer.c	2007-03-16 14:13:19.000000000 -0700
++++ linux-2.6.20/kernel/timer.c	2007-03-16 14:51:15.000000000 -0700
+@@ -74,7 +74,7 @@
+ 	tvec_t tv3;
+ 	tvec_t tv4;
+ 	tvec_t tv5;
+-} ____cacheline_aligned_in_smp;
++} ____cacheline_aligned;
+ 
+ typedef struct tvec_t_base_s tvec_base_t;
+ 
+@@ -325,7 +325,7 @@
+ 	tvec_base_t *base;
+ 
+ 	for (;;) {
+-		base = timer->base;
++		base = TBASE_GET_BASE_PTR(timer->base);
+ 		if (likely(base != NULL)) {
+ 			spin_lock_irqsave(&base->lock, *flags);
+ 			if (likely(base == timer->base))
+@@ -364,12 +364,15 @@
+ 		 * the timer is serialized wrt itself.
+ 		 */
+ 		if (likely(base->running_timer != timer)) {
++			unsigned long tflag;
++			tflag = TBASE_GET_DELAYED_ON_IDLE(timer->base);
+ 			/* See the comment in lock_timer_base() */
+ 			timer->base = NULL;
+ 			spin_unlock(&base->lock);
+ 			base = new_base;
+ 			spin_lock(&base->lock);
+-			timer->base = base;
++			timer->base =
++				TBASE_MERGE_DELAYED_ON_IDLE(new_base, tflag);
+ 		}
+ 	}
+ 
+@@ -386,10 +389,12 @@
+  * add_timer_on - start a timer on a particular CPU
+  * @timer: the timer to be added
+  * @cpu: the CPU to start it on
++ * @not_critical_when_idle: 1 to indicate timer is not critical and
++ *                          can be delayed when CPU is idle
+  *
+  * This is not very scalable on SMP. Double adds are not possible.
+  */
+-void add_timer_on(struct timer_list *timer, int cpu)
++void add_timer_on(struct timer_list *timer, int cpu, int not_critical_when_idle)
+ {
+ 	tvec_base_t *base = per_cpu(tvec_bases, cpu);
+   	unsigned long flags;
+@@ -397,7 +402,7 @@
+ 	timer_stats_timer_set_start_info(timer);
+   	BUG_ON(timer_pending(timer) || !timer->function);
+ 	spin_lock_irqsave(&base->lock, flags);
+-	timer->base = base;
++	timer->base = TBASE_MERGE_DELAYED_ON_IDLE(base, not_critical_when_idle);
+ 	internal_add_timer(base, timer);
+ 	spin_unlock_irqrestore(&base->lock, flags);
+ }
+@@ -548,7 +553,7 @@
+ 	 * don't have to detach them individually.
+ 	 */
+ 	list_for_each_entry_safe(timer, tmp, &tv_list, entry) {
+-		BUG_ON(timer->base != base);
++		BUG_ON(TBASE_GET_BASE_PTR(timer->base) != base);
+ 		internal_add_timer(base, timer);
+ 	}
+ 
+@@ -634,6 +639,9 @@
+ 	index = slot = timer_jiffies & TVR_MASK;
+ 	do {
+ 		list_for_each_entry(nte, base->tv1.vec + slot, entry) {
++ 			if (TBASE_GET_DELAYED_ON_IDLE(nte->base))
++ 				continue;
++ 
+ 			found = 1;
+ 			expires = nte->expires;
+ 			/* Look at the cascade bucket(s)? */
+@@ -1602,6 +1610,13 @@
+ 						cpu_to_node(cpu));
+ 			if (!base)
+ 				return -ENOMEM;
++
++			/* Make sure that tvec_base is 2 byte aligned */
++			if (TBASE_GET_DELAYED_ON_IDLE(base)) {
++				WARN_ON(1);
++				kfree(base);
++				return -ENOMEM;
++			}
+ 			memset(base, 0, sizeof(*base));
+ 			per_cpu(tvec_bases, cpu) = base;
+ 		} else {
+@@ -1641,9 +1656,11 @@
+ 	struct timer_list *timer;
+ 
+ 	while (!list_empty(head)) {
++		unsigned long tflag;
+ 		timer = list_entry(head->next, struct timer_list, entry);
++		tflag = TBASE_GET_DELAYED_ON_IDLE(timer->base);
++		timer->base = TBASE_MERGE_DELAYED_ON_IDLE(new_base, tflag);
+ 		detach_timer(timer, 0);
+-		timer->base = new_base;
+ 		internal_add_timer(new_base, timer);
+ 	}
+ }
+Index: linux-2.6.20/include/linux/timer.h
+===================================================================
+--- linux-2.6.20.orig/include/linux/timer.h	2007-03-16 14:13:19.000000000 -0700
++++ linux-2.6.20/include/linux/timer.h	2007-03-16 14:47:28.000000000 -0700
+@@ -8,6 +8,33 @@
+ 
+ struct tvec_t_base_s;
+ 
++extern struct tvec_t_base_s boot_tvec_bases;
++/*
++ * Note that all tvec_bases is 2 byte aligned and lower bit of
++ * base in timer_list is guaranteed to be zero. Use the LSB for
++ * the new flag to indicate whether it is OK to skip timer callback
++ * when CPU is idle.
++ */
++#define TBASE_FLAG_DELAYED_ON_IDLE		(0x1)
++
++#define TBASE_GET_BASE_PTR(x)						\
++	((struct tvec_t_base_s *)((unsigned long)x &			\
++	                          (~TBASE_FLAG_DELAYED_ON_IDLE)))
++
++#define TBASE_GET_DELAYED_ON_IDLE(x)					\
++	((unsigned long)x & TBASE_FLAG_DELAYED_ON_IDLE)
++
++#define TBASE_SET_DELAYED_ON_IDLE(x)					\
++	((struct tvec_t_base_s *)((unsigned long)x |			\
++	                          TBASE_FLAG_DELAYED_ON_IDLE))
++
++#define TBASE_CLEAR_DELAYED_ON_IDLE(x)					\
++	((struct tvec_t_base_s *)((unsigned long)x &			\
++	                          (~TBASE_FLAG_DELAYED_ON_IDLE)))
++
++#define TBASE_MERGE_DELAYED_ON_IDLE(x,f)				\
++	((f) ? TBASE_SET_DELAYED_ON_IDLE(x) : TBASE_CLEAR_DELAYED_ON_IDLE(x))
++
+ struct timer_list {
+ 	struct list_head entry;
+ 	unsigned long expires;
+@@ -23,7 +50,6 @@
+ #endif
+ };
+ 
+-extern struct tvec_t_base_s boot_tvec_bases;
+ 
+ #define TIMER_INITIALIZER(_function, _expires, _data) {		\
+ 		.function = (_function),			\
+@@ -62,7 +88,8 @@
+ 	return timer->entry.next != NULL;
+ }
+ 
+-extern void add_timer_on(struct timer_list *timer, int cpu);
++extern void add_timer_on(struct timer_list *timer, int cpu,
++                         int not_critical_when_idle);
+ extern int del_timer(struct timer_list * timer);
+ extern int __mod_timer(struct timer_list *timer, unsigned long expires);
+ extern int mod_timer(struct timer_list *timer, unsigned long expires);
+@@ -144,6 +171,16 @@
+ static inline void add_timer(struct timer_list *timer)
+ {
+ 	BUG_ON(timer_pending(timer));
++	timer->base = TBASE_CLEAR_DELAYED_ON_IDLE(timer->base);
++	__mod_timer(timer, timer->expires);
++}
++
++static inline void add_timer_with_hint(struct timer_list *timer,
++                                       int not_critical_when_idle)
++{
++	BUG_ON(timer_pending(timer));
++	timer->base = TBASE_MERGE_DELAYED_ON_IDLE(timer->base,
++	                                          not_critical_when_idle);
+ 	__mod_timer(timer, timer->expires);
+ }
+ 
+Index: linux-2.6.20/arch/powerpc/platforms/chrp/setup.c
+===================================================================
+--- linux-2.6.20.orig/arch/powerpc/platforms/chrp/setup.c	2007-03-16 14:13:19.000000000 -0700
++++ linux-2.6.20/arch/powerpc/platforms/chrp/setup.c	2007-03-16 14:47:28.000000000 -0700
+@@ -565,7 +565,7 @@
+ 			timer = &per_cpu(heartbeat_timer, cpu);
+ 			setup_timer(timer, chrp_event_scan, 0);
+ 			timer->expires = jiffies + offset;
+-			add_timer_on(timer, cpu);
++			add_timer_on(timer, cpu, 0);
+ 			offset += interval;
+ 		}
+ 		printk("RTAS Event Scan Rate: %u (%lu jiffies)\n",
+Index: linux-2.6.20/kernel/workqueue.c
+===================================================================
+--- linux-2.6.20.orig/kernel/workqueue.c	2007-03-16 14:13:19.000000000 -0700
++++ linux-2.6.20/kernel/workqueue.c	2007-03-16 14:51:00.000000000 -0700
+@@ -290,7 +290,7 @@
+ 		timer->expires = jiffies + delay;
+ 		timer->data = (unsigned long)dwork;
+ 		timer->function = delayed_work_timer_fn;
+-		add_timer_on(timer, cpu);
++		add_timer_on(timer, cpu, 0);
+ 		ret = 1;
+ 	}
+ 	return ret;
+
 From davej  Wed Mar 21 16:25:31 2007
 Return-Path: <venkatesh.pallipadi at intel.com>
 X-Spam-Checker-Version: SpamAssassin 3.1.8 (2007-02-13) on


