linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] shrink check_nmi_watchdog stack frame
@ 2002-07-25 16:06 Mikael Pettersson
  0 siblings, 0 replies; 2+ messages in thread
From: Mikael Pettersson @ 2002-07-25 16:06 UTC (permalink / raw)
  To: torvalds; +Cc: linux-kernel

This patch for 2.5.28 reduces the stack frame size of
arch/i386/kernel/nmi.c:check_nmi_watchdog() from 4096 bytes
in the worst case to 128 bytes. Linus, please apply.

The problem with the current code is that it copies the entire
irq_stat[] array, when only a single field (__nmi_count) is of
interest. The irq_stat_t element type is only 28 bytes, but it
is also ____cacheline_aligned, and that blows the array up to
4096 bytes on SMP P4 Xeons, 2048 bytes on SMP K7s, and 1024 bytes
on SMP P5/P6s. The patch reduces this to NR_CPUS*4==128 bytes.

/Mikael

diff -ruN linux-2.5.28/arch/i386/kernel/nmi.c linux-2.5.28.check-nmi/arch/i386/kernel/nmi.c
--- linux-2.5.28/arch/i386/kernel/nmi.c	Thu Jul 25 01:27:29 2002
+++ linux-2.5.28.check-nmi/arch/i386/kernel/nmi.c	Thu Jul 25 01:31:55 2002
@@ -72,19 +72,20 @@
 
 int __init check_nmi_watchdog (void)
 {
-	irq_cpustat_t tmp[NR_CPUS];
+	unsigned int prev_nmi_count[NR_CPUS];
 	int cpu;
 
 	printk(KERN_INFO "testing NMI watchdog ... ");
 
-	memcpy(tmp, irq_stat, sizeof(tmp));
+	for (cpu = 0; cpu < NR_CPUS; cpu++)
+		prev_nmi_count[cpu] = irq_stat[cpu].__nmi_count;
 	local_irq_enable();
 	mdelay((10*1000)/nmi_hz); // wait 10 ticks
 
 	for (cpu = 0; cpu < NR_CPUS; cpu++) {
 		if (!cpu_online(cpu))
 			continue;
-		if (nmi_count(cpu) - tmp[cpu].__nmi_count <= 5) {
+		if (nmi_count(cpu) - prev_nmi_count[cpu] <= 5) {
 			printk("CPU#%d: NMI appears to be stuck!\n", cpu);
 			return -1;
 		}

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [PATCH] shrink check_nmi_watchdog stack frame
@ 2002-06-16 11:41 Mikael Pettersson
  0 siblings, 0 replies; 2+ messages in thread
From: Mikael Pettersson @ 2002-06-16 11:41 UTC (permalink / raw)
  To: torvalds; +Cc: davej, linux-kernel

This patch for 2.5.21 reduces the stack frame size of
arch/i386/kernel/nmi.c:check_nmi_watchdog() from 4096 bytes
in the worst case to 128 bytes. Linus, please apply.

The problem with the current code is that it copies the entire
irq_stat[] array, when only a single field (__nmi_count) is of
interest. The irq_stat_t element type is only 28 bytes, but it
is also ____cacheline_aligned, and that blows the array up to
4096 bytes on SMP P4 Xeons, 2048 bytes on SMP K7s, and 1024 bytes
on SMP P5/P6s. The patch reduces this to NR_CPUS*4==128 bytes.

If you approve this patch I'll also send one to Marcelo for 2.4.

/Mikael

diff -ruN linux-2.5.21/arch/i386/kernel/nmi.c linux-2.5.21.check-nmi/arch/i386/kernel/nmi.c
--- linux-2.5.21/arch/i386/kernel/nmi.c	Mon Apr 15 00:32:51 2002
+++ linux-2.5.21.check-nmi/arch/i386/kernel/nmi.c	Sun Jun 16 12:16:55 2002
@@ -72,18 +72,19 @@
 
 int __init check_nmi_watchdog (void)
 {
-	irq_cpustat_t tmp[NR_CPUS];
+	unsigned int prev_nmi_count[NR_CPUS];
 	int j, cpu;
 
 	printk(KERN_INFO "testing NMI watchdog ... ");
 
-	memcpy(tmp, irq_stat, sizeof(tmp));
+	for(j = 0; j < NR_CPUS; ++j)
+		prev_nmi_count[j] = irq_stat[j].__nmi_count;
 	sti();
 	mdelay((10*1000)/nmi_hz); // wait 10 ticks
 
 	for (j = 0; j < smp_num_cpus; j++) {
 		cpu = cpu_logical_map(j);
-		if (nmi_count(cpu) - tmp[cpu].__nmi_count <= 5) {
+		if (nmi_count(cpu) - prev_nmi_count[cpu] <= 5) {
 			printk("CPU#%d: NMI appears to be stuck!\n", cpu);
 			return -1;
 		}

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2002-07-25 16:03 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-07-25 16:06 [PATCH] shrink check_nmi_watchdog stack frame Mikael Pettersson
  -- strict thread matches above, loose matches on Subject: below --
2002-06-16 11:41 Mikael Pettersson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).