From: rohit verma <[email protected]>
Subject: Re: [LTP] Fw: Regarding ht_enable test case failure
Date: Tue, 19 May 2009 17:52:26 +0530
> Dear CAI Qian,
>
> I want to discuss few issues/obervations regarding ht_interrupt.
>
> I have two test PC:
>
> 1. Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz
> cat /etc/issue
> Fedora release 10 (Cambridge)
> Kernel \r on an \m (\l)
>
> 2. Intel(R) Pentium(R) 4 CPU 3.00GHz
> cat /etc/issue
> Fedora release 10 (Cambridge)
> Kernel \r on an \m (\l)
>
>
> I have commented check_ht_capability( ) in test code just for testing
> purpose.
>
OK. I have also need to change,
#ifndef ARCH_i386
tst_brkm(TCONF, NULL, "This test suite can only excute on i386
architecture.");
#else
to
#ifndef __i386__
tst_brkm(TCONF, NULL, "This test suite can only excute on i386
architecture.");
#else
in order to get it working on 32-bit systems.
> In ht_interrupt, timer interrupt is used to check whether interrupt is
> balancable or not.
>
> # cat /proc/interrupts
> CPU0 CPU1
> 0: 142 30 IO-APIC-edge timer
>
>
> I have changed that check to ethernet interrupt, because on my first test PC
> these number (142 and 30 ) remains same till the execution of test and test
> case fails with following messages.
> "Interrupt distribution is not balancable"
>
It works fine on a RHEL5.3 here without need to change,
# lscpu
Architecture: i686
CPU(s): 8
Thread(s) per core: 2
Core(s) per socket: 2
CPU socket(s): 2
Vendor ID: GenuineIntel
CPU family: 15
Model: 6
Stepping: 4
CPU MHz: 3725.610
...
# uname -r
2.6.18-128.el5
# ./ht_interrupt
ht_interrupt 0 INFO : Begin: HyperThreading Interrupt
ht_interrupt 0 INFO : Get interrupts distribution with HT.
..........
Timer interrupt counts per CPU:
1263 1260 1224 1223 1262 1262 1263 1263
max value: 1263
min value: 1223
ht_interrupt 1 PASS : Interrupt distribution is balanceable.
ht_interrupt 0 INFO : End: HyperThreading Interrupt
However, it is not working on a Fedora 10 system,
# lscpu
Architecture: i686
CPU(s): 2
Thread(s) per core: 2
Core(s) per socket: 1
CPU socket(s): 1
Vendor ID: GenuineIntel
CPU family: 15
Model: 4
Stepping: 3
CPU MHz: 2992.428
# uname -r
2.6.27.5-117.fc10.i686
# ./ht_interrupt
ht_interrupt 0 INFO : Begin: HyperThreading Interrupt
ht_interrupt 0 INFO : Get interrupts distribution with HT.
..........
Timer interrupt counts per CPU:
0 0
max value: 0
min value: 0
ht_interrupt 1 FAIL : Interrupt distribution is not balanceable.
ht_interrupt 0 INFO : End: HyperThreading Interrupt
I am not sure if that is because they are using different kernel
versions?
> Test case passes on first test PC after the change (checking ethernet
> interrupt).
>
> On second test PC with hyperthreading enabled, this test case fails:
>
> #cat /proc/irq/27/smp_affinity
> # 3
>
> #cat /proc/interrupts
> CPU0 CPU1
> 27: 2002896 449 PCI-MSI-edge eth0
>
> #cat /proc/interrupts
> CPU0 CPU1
> 27: 20032957 449 PCI-MSI-edge eth0
> Interrupt count for only CPU0 is incrementing.
>
> I change the smp_affinity
>
> # echo 2 > /proc/irq/27/smp_affinity
> # cat /proc/irq/27/smp_affinity
> # 2
>
> #cat /proc/interrupts
> CPU0 CPU1
> 27: 2004494 564 PCI-MSI-edge eth0
> Here, interrupt count for CPU1 is incrementing.
>
> How can I get the balanced ethernet interrupts on both CPUs.
>
Same here on the Fedora 10 system.
# echo 3 >/proc/irq/16/smp_affinity
...
16: 7412900 1212 IO-APIC-fasteoi eth1
...
16: 7412991 1212 IO-APIC-fasteoi eth1
...
# echo 2 >/proc/irq/16/smp_affinity
...
16: 7413106 1281 IO-APIC-fasteoi eth1
...
16: 7413106 1283 IO-APIC-fasteoi eth1
...
No idea on how to get a balanced interrupts.
It is rather strange on the RHEL5.3 system.
# echo 3 >/proc/irq/90/smp_affinity
The interrupt count is increasing on only CPU0.
# cat /proc/interrupts
...
90: 107078 0 0 0 0 0 0 0 PCI-MS eth0
...
90: 107608 0 0 0 0 0 0 0 PCI-MS eth0
The same for changing it to 2.
# echo 2 >/proc/irq/90/smp_affinity
# cat /proc/interrupts
...
90: 108121 0 0 0 0 0 0 0 PCI-MS eth0
...
90: 108163 0 0 0 0 0 0 0 PCI-MS eth0
>
> I am also getting an error message, when i try to change smp_affinity
> for timer interrupt.
>
> # echo 3 > /proc/irq/0/smp_affinity
> -bash: echo: write error: Input/output error
>
This works fine on the RHEL5 system,
# echo 3 > /proc/irq/0/smp_affinity
Then, the interrupt counts on both CPU0 and CPU1 are increasing.
The Fedora 10 system is not,
# echo 3 > /proc/irq/0/smp_affinity
CAI Qian
> # echo 3 > /proc/irq/<eth interrupt no>/smp_affinity works fine.
>
>
> Regards,
> Rohit
>
>
------------------------------------------------------------------------------
Register Now for Creativity and Technology (CaT), June 3rd, NYC. CaT
is a gathering of tech-side developers & brand creativity professionals. Meet
the minds behind Google Creative Lab, Visual Complexity, Processing, &
iPhoneDevCamp asthey present alongside digital heavyweights like Barbarian
Group, R/GA, & Big Spaceship. http://www.creativitycat.com
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
