2002-06-21 06:57:28

by Rasmus Bøg Hansen

[permalink] [raw]
Subject: Procinfo behaving strange under 2.4.19-pre10

Hi

I upgraded a short time ago from kernel 2.4.18 to 2.4.19-pre10, but now
procinfo reports interrupts in a strange way.

2.4.19-pre10:

# procinfo
Linux 2.4.19-pre10 (root@grignard) (gcc 2.95.4 20011002 ) #2 1CPU [grignard]

Memory: Total Used Free Shared Buffers Cached
Mem: 257480 225016 32464 0 20912 119516
Swap: 265064 0 265064

Bootup: Fri Jun 21 08:12:02 2002 Load average: 0.03 0.07 0.02 2/83 16013

user : 0:01:06.58 3.2% page in : 128235 disk 1: 10476r 6363w
nice : 0:00:00.00 0.0% page out: 90740
system: 0:00:30.95 1.5% swap in : 1
idle : 0:32:59.28 95.3% swap out: 0
uptime: 0:34:36.80 context : 561976

irq 0:1000207681 timer irq 8:1000000003
irq 1:1000004868 keyboard irq 9:1000000000 acpi
irq 2:1000000000 cascade [4] irq 10:1000007854 eth0
irq 3:1000000000 irq 11:1000114199 nvidia
irq 4:1000000000 irq 12:1000026199 PS/2 Mouse
irq 5:1000003195 es1370 irq 13:1000000000
irq 6:1000000000 irq 14:1000016806 ide0
irq 7:1000000000 irq 15:1000000000

# cat/proc/interrupts
CPU0
0: 207681 XT-PIC timer
1: 4868 XT-PIC keyboard
2: 0 XT-PIC cascade
5: 3195 XT-PIC es1370
9: 0 XT-PIC acpi
10: 7857 XT-PIC eth0
11: 114199 XT-PIC nvidia
12: 26199 XT-PIC PS/2 Mouse
14: 16806 XT-PIC ide0
NMI: 0
ERR: 0

After rebooting to 2.4.18:

# procinfo
Linux 2.4.18 (root@grignard) (gcc 2.95.4 20011002 ) #1 1CPU [grignard]

Memory: Total Used Free Shared Buffers Cached
Mem: 256228 47544 208684 0 3720 26012
Swap: 265064 0 265064

Bootup: Fri Jun 21 08:48:25 2002 Load average: 0.24 0.06 0.02 1/47 664

user : 0:00:01.73 4.2% page in : 27919 disk 1: 2206r 571w
nice : 0:00:00.00 0.0% page out: 8616
system: 0:00:04.61 11.2% swap in : 1
idle : 0:00:34.87 84.6% swap out: 0
uptime: 0:00:41.21 context : 8126

irq 0: 4121 timer irq 9: 0 acpi
irq 1: 72 keyboard irq 10: 177 eth0
irq 2: 0 cascade [4] irq 12: 8 PS/2 Mouse
irq 5: 0 es1370 irq 14: 2782 ide0
irq 8: 3 rtc

# procinfo
CPU0
0: 4122 XT-PIC timer
1: 72 XT-PIC keyboard
2: 0 XT-PIC cascade
5: 0 XT-PIC es1370
8: 3 XT-PIC rtc
9: 0 XT-PIC acpi
10: 177 XT-PIC eth0
12: 8 XT-PIC PS/2 Mouse
14: 2782 XT-PIC ide0
NMI: 0
ERR: 0

Both kernels were compiled and installed under the same environment:

Linux grignard 2.4.18 #1 tor maj 30 22:09:28 CEST 2002 i686 unknown

Gnu C 2.95.4
Gnu make 3.79.1
util-linux 2.11n
mount 2.11n
modutils 2.4.15
e2fsprogs 1.27
Linux C Library 2.2.5
Dynamic linker (ldd) 2.2.5
Procps 2.0.7
Net-tools 1.60
Kbd [muligheder...]
Console-tools 0.2.3
Sh-utils 2.0.11
Modules Loaded parport_pc lp parport autofs4 es1370 soundcore ipt_REJECT ipt_unclean ipt_LOG ipt_limit ipt_state iptable_mangle iptable_nat iptable_filter ip_tables ip_conntrack_irc ip_conntrack_ftp ip_conntrack nls_iso8859-15 nls_cp865 vfat fat serial eepro100 af_packet rtc

Procinfo is procinfo-18; my system is Debian Woody.

Regards
/Rasmus

--
-- [ Rasmus "M?ffe" B?g Hansen ] ---------------------------------------
Because I don't want to force you to follow my philosophy, even though
it happens to be the only possible correct philosophy.
-- Ted Lemon
----------------------------------[ moffe at amagerkollegiet dot dk ] --


2002-06-23 20:03:46

by Axel Thimm

[permalink] [raw]
Subject: 1000000000 as irq count init (was: Procinfo behaving strange under 2.4.19-pre10)

Hi,

On Fri, Jun 21, 2002 at 08:57:25AM +0200, Rasmus B?g Hansen wrote:
> I upgraded a short time ago from kernel 2.4.18 to 2.4.19-pre10, but now
> procinfo reports interrupts in a strange way.
>
> 2.4.19-pre10:
>
> # procinfo
> [...]
> irq 0:1000207681 timer irq 8:1000000003
> irq 1:1000004868 keyboard irq 9:1000000000 acpi
> irq 2:1000000000 cascade [4] irq 10:1000007854 eth0
> irq 3:1000000000 irq 11:1000114199 nvidia
> irq 4:1000000000 irq 12:1000026199 PS/2 Mouse
> irq 5:1000003195 es1370 irq 13:1000000000
> irq 6:1000000000 irq 14:1000016806 ide0
> irq 7:1000000000 irq 15:1000000000
> [...]

I can second this (while I switched from 2.4.19-pre6 to 2.4.18 with RedHat/SGI
patches, but obviously the same patch hit me also).

procinfo reads that info from /proc/stat, which is giving away that bogus (?)
interrupt counts. It also causes xosview a horrible cpu 100% hang. :(

[root@bonzo root]# uname -a
Linux bonzo.nirvana 2.4.18-4SGI_XFS_1.1-bonzo #1 Sun Jun 23 17:49:32 CEST 2002
i686 unknown
[root@bonzo root]# cat /proc/stat
cpu 204580 0 5397 181674
cpu0 204580 0 5397 181674
page 193512 228084
swap 1 0
intr 668955 1000391651 1000008974 1000000000 1000000007 1000000008 1000000002
1000000008 1000000002 1000000001 1000212403 1000000000 1000000002 1000027284
1000000000 [...] 1000000000 1000000000 1000000000 1000000000 1000000000
1000000000 1000000000 1000000000 1000000000
disk_io: (3,0):(28638,17974,387024,10664,456168)
ctxt 1276299
btime 1024858081
processes 5380
--
[email protected]

2002-06-24 08:08:43

by Adrian Bunk

[permalink] [raw]
Subject: Re: 1000000000 as irq count init (was: Procinfo behaving strange under 2.4.19-pre10)

On Sun, 23 Jun 2002, Axel Thimm wrote:

> On Fri, Jun 21, 2002 at 08:57:25AM +0200, Rasmus B?g Hansen wrote:
> > I upgraded a short time ago from kernel 2.4.18 to 2.4.19-pre10, but now
> > procinfo reports interrupts in a strange way.
> >
> > 2.4.19-pre10:
> >
> > # procinfo
> > [...]
> > irq 0:1000207681 timer irq 8:1000000003
> > irq 1:1000004868 keyboard irq 9:1000000000 acpi
> > irq 2:1000000000 cascade [4] irq 10:1000007854 eth0
> > irq 3:1000000000 irq 11:1000114199 nvidia
> > irq 4:1000000000 irq 12:1000026199 PS/2 Mouse
> > irq 5:1000003195 es1370 irq 13:1000000000
> > irq 6:1000000000 irq 14:1000016806 ide0
> > irq 7:1000000000 irq 15:1000000000
> > [...]
>
> I can second this (while I switched from 2.4.19-pre6 to 2.4.18 with RedHat/SGI
> patches, but obviously the same patch hit me also).
>...

This is a known issue. The following patch to fix it is already in
Marcelos BK reporsitory (which means that the fix will be in -pre11/-rc1):

# --------------------------------------------
# 02/06/07 [email protected] 1.537.1.44
# [PATCH] Remove debug code from /proc/stat
#
# This patch removes a piece of debug code which crept into the kernel
# by accident.
# --------------------------------------------
#
diff -Nru a/fs/proc/proc_misc.c b/fs/proc/proc_misc.c
--- a/fs/proc/proc_misc.c Wed Jun 12 17:08:55 2002
+++ b/fs/proc/proc_misc.c Wed Jun 12 17:08:55 2002
@@ -336,7 +336,7 @@
#if !defined(CONFIG_ARCH_S390)
for (i = 0 ; i < NR_IRQS ; i++)
proc_sprintf(page, &off, &len,
- " %u", kstat_irqs(i) + 1000000000);
+ " %u", kstat_irqs(i));
#endif

proc_sprintf(page, &off, &len, "\ndisk_io: ");

cu
Adrian

--

You only think this is a free country. Like the US the UK spends a lot of
time explaining its a free country because its a police state.
Alan Cox