Re: 2.6.17-mm6 vmstat breakage

From: Andrew Morton
Date: Thu Jul 06 2006 - 04:23:02 EST


On Thu, 06 Jul 2006 09:38:43 +0200
Mike Galbraith <efault@xxxxxx> wrote:

> Greetings,
>
> My single P4/HT box is showing incorrect pgpgin pgpgout numbers with
> this kernel. While disk throughput for dd if=/dev/hdc of=/dev=null
> bs=4096 count=1000000 produces about 57MB/S (up ~14% over 2.6.17 btw),
> vmstat (procps version 3.2.5) is only showing about 7MB/S.
>
> Looking at the numbers in /proc/vmstat, it looks kinda like pgpgin might
> be pages instead of the KB it used to be, with some added >>=1 action on
> the side. At any rate, the numbers below are horse-pookey ;-)
>
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 1 1 3820 9400 810128 46052 0 0 7040 0 1163 1283 2 7 48 44
> 1 1 3820 9220 810288 46156 0 0 7040 0 1151 1335 3 6 48 45
> 1 1 3820 9280 810176 46204 0 0 7168 0 1156 1207 1 7 48 44
> 2 1 3820 9408 810048 46068 0 0 7168 0 1160 1192 1 6 49 46
>

2.6.18-rc1 has the same bug.

--- a/include/linux/vmstat.h~count_vm_events-fix
+++ a/include/linux/vmstat.h
@@ -57,7 +57,7 @@ static inline void __count_vm_events(enu

static inline void count_vm_events(enum vm_event_item item, long delta)
{
- get_cpu_var(vm_event_states.event[item])++;
+ get_cpu_var(vm_event_states.event[item]) += delta;
put_cpu();
}

_

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/