Commit 478735e38887077ac77a9756121b6ce0cb956e2f

Authored by KAMEZAWA Hiroyuki
Committed by Linus Torvalds
1 parent f2c66cd8ee

/proc/stat: fix scalability of irq sum of all cpu

In /proc/stat, the number of per-IRQ event is shown by making a sum each
irq's events on all cpus.  But we can make use of kstat_irqs().

kstat_irqs() do the same calculation, If !CONFIG_GENERIC_HARDIRQ,
it's not a big cost. (Both of the number of cpus and irqs are small.)

If a system is very big and CONFIG_GENERIC_HARDIRQ, it does

	for_each_irq()
		for_each_cpu()
			- look up a radix tree
			- read desc->irq_stat[cpu]
This seems not efficient. This patch adds kstat_irqs() for
CONFIG_GENRIC_HARDIRQ and change the calculation as

	for_each_irq()
		look up radix tree
		for_each_cpu()
			- read desc->irq_stat[cpu]

This reduces cost.

A test on (4096cpusp, 256 nodes, 4592 irqs) host (by Jack Steiner)

%time cat /proc/stat > /dev/null

Before Patch:	 2.459 sec
After Patch :	  .561 sec

[akpm@linux-foundation.org: unexport kstat_irqs, coding-style tweaks]
[akpm@linux-foundation.org: fix unused variable 'per_irq_sum']
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Tested-by: Jack Steiner <steiner@sgi.com>
Acked-by: Jack Steiner <steiner@sgi.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Showing 3 changed files with 21 additions and 8 deletions Side-by-side Diff

... ... @@ -31,7 +31,6 @@
31 31 u64 sum_softirq = 0;
32 32 unsigned int per_softirq_sums[NR_SOFTIRQS] = {0};
33 33 struct timespec boottime;
34   - unsigned int per_irq_sum;
35 34  
36 35 user = nice = system = idle = iowait =
37 36 irq = softirq = steal = cputime64_zero;
... ... @@ -108,13 +107,8 @@
108 107 seq_printf(p, "intr %llu", (unsigned long long)sum);
109 108  
110 109 /* sum again ? it could be updated? */
111   - for_each_irq_nr(j) {
112   - per_irq_sum = 0;
113   - for_each_possible_cpu(i)
114   - per_irq_sum += kstat_irqs_cpu(j, i);
115   -
116   - seq_printf(p, " %u", per_irq_sum);
117   - }
  110 + for_each_irq_nr(j)
  111 + seq_printf(p, " %u", kstat_irqs(j));
118 112  
119 113 seq_printf(p,
120 114 "\nctxt %llu\n"
include/linux/kernel_stat.h
... ... @@ -86,6 +86,7 @@
86 86 /*
87 87 * Number of interrupts per specific IRQ source, since bootup
88 88 */
  89 +#ifndef CONFIG_GENERIC_HARDIRQS
89 90 static inline unsigned int kstat_irqs(unsigned int irq)
90 91 {
91 92 unsigned int sum = 0;
... ... @@ -96,6 +97,9 @@
96 97  
97 98 return sum;
98 99 }
  100 +#else
  101 +extern unsigned int kstat_irqs(unsigned int irq);
  102 +#endif
99 103  
100 104 /*
101 105 * Number of interrupts per cpu, since bootup
kernel/irq/irqdesc.c
... ... @@ -393,4 +393,19 @@
393 393 struct irq_desc *desc = irq_to_desc(irq);
394 394 return desc ? desc->kstat_irqs[cpu] : 0;
395 395 }
  396 +
  397 +#ifdef CONFIG_GENERIC_HARDIRQS
  398 +unsigned int kstat_irqs(unsigned int irq)
  399 +{
  400 + struct irq_desc *desc = irq_to_desc(irq);
  401 + int cpu;
  402 + int sum = 0;
  403 +
  404 + if (!desc)
  405 + return 0;
  406 + for_each_possible_cpu(cpu)
  407 + sum += desc->kstat_irqs[cpu];
  408 + return sum;
  409 +}
  410 +#endif /* CONFIG_GENERIC_HARDIRQS */