Xen project Mailing List

On Feb 5, 2008 3:34 PM, john bryant <bryant.johan@xxxxxxxxx> wrote:

Thanks Anna for reply. It was certainly an oprofile error. Actually I messed up with different version of binutils installation ( in /usr/bin/   and /usr/local/bin). I end up removing some of installed binaries and libraries of binutils and reinstalled bintutils using yum. it worked and I don't get "segmentation fault". I had posted similar error in oprofile forum and apparently two copies of BFD libraries were installed.

Now question comes why i am getting write error when i start daemon using

opcontrol --start-daemon --event=CPU_CLK_UNHALTED:10000
:0:1:1 --xen=/boot/xen-syms-3.0.4-1 --vmlinux=/boot/vmlinux-syms-2.6.16.33-xen --active-domains=0

Following is write error:

/usr/local/bin/opcontrol: line 1082: echo: write error: Invalid argument
Using 2.6+ OProfile kernel interface.
Reading module info.
Using log file /var/lib/oprofile/oprofiled.log

Daemon started.

checking inside opcontrol script, error comes from following function:

set_param()
{
        if test "$KERNEL_SUPPORT" = "yes"; then
                echo $2 >$MOUNT/$1
        else
                $SYSCTL -w dev.oprofile.$1=$2
        fi
}

Any clues ?
Thanks again for help.

-John

On Feb 5, 2008 10:14 AM, Fischer, Anna <anna.fischer@xxxxxx> wrote:

John,

I think this could be Oprofile specific, and not necessarily related to xenoprof. I have a couple of ideas that might help to narrow down where the problem may be... I haven't come across this problem before myself.

You can run each parameter individually, and then you'll see which parameter causes the "write error" message (I suspect it'll be the "--start-daemon" call).

opcontrol --vmlinux=/boot/<your-kernel-name> --xen=/boot/<xen-syms>
opcontrol --start-daemon
opcontrol --event=CYCLES:1000000 --event=SPU_CYCLES:60000
opcontrol --reset
opcontrol --start
opcontrol --stop
opcontrol --dump
opreport

Also, do you see any samples in /var/log/oprofile/samples? Does it make any difference if you remove all samples in there and restart?

The difference between "opreport -l" and "opreport" is that the first command is trying to give samples per-symbol rather than per-library/per-application. So in your case opreport seems to crash when it's trying to get symbol information.

Also, have you installed oprofile using the Fedora installer as a normal package installation? I think you additionally need to install the kernel-debuginfo package:

yum --enablerepo=core-debuginfo --enable=updates-debuginfo install kernel-debuginfo

Anna

________________________________

From: john bryant [mailto:bryant.johan@xxxxxxxxx]
Sent: 05 February 2008 03:02
To: Fischer, Anna
Cc: xen-users@xxxxxxxxxxxxxxxxxxx

Subject: Re: [Xen-users] Re: xenoprof error

Anna,

Thanks for helping me. I am novice in Xen.
I have installed xen 3.0.4 and oprofile 0.9.2. Dom0 and DomU kernel is 2.6.16.33-xen. Let us take a simple case when there is one VM only. In this case, following is sequence of commands I type:

> opcontrol --reset
> opcontrol --start-daemon --event=CPU_CLK_UNHALTED:10000:0:1:1 --xen=/boot/xen-syms-3.0.4-1 --vmlinux=/boot/vmlinux-syms-2.6.16.33-xen --active-domains=0
I get following reply:

/usr/local/bin/opcontrol: line 1082: echo: write error: Invalid argument
Using 2.6+ OProfile kernel interface.
Reading module info.
Using log file /var/lib/oprofile/oprofiled.log
Daemon started.

First question: what is this invalid argument error ? Apparently, opcontrol starts and log (/var/lib/oprofile/oprofiled.log) does not indicate any error

Now, assuming oprofile started correctly, I do some activity such as "ls" or run some of my C programs. Then i follow this command sequence:

> opcontrol --dump

> opreport
This gives following error (warning seems ok as i have not given some of symbol files):
warning: /bridge could not be found.
warning: /ehci_hcd could not be found.
warning: /evdev could not be found.
warning: /ext3 could not be found.
warning: /jbd could not be found.
warning: /oprofile could not be found.
warning: /tg3 could not be found.
warning: /uhci_hcd could not be found.
warning: /usbcore could not be found.
warning: /xen-unknown could not be found.
CPU: Core 2, speed 1995.03 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask of 0x00 (Unhalted core cycles) count 100000
Segmentation fault

If I try only "opreport" I get following output (I am giving partial output as it seems ok):

CPU: Core 2, speed 1995.03 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask of 0x00 (Unhalted core cycles) count 100000
CPU_CLK_UNHALT...|
samples| %|
------------------
33557 14.8824 libfb.so
31884 14.1404 libmozjs.so
26669 11.8276 xen-syms-3.0.4-1
26373 11.6963 libgklayout.so
19726 8.7484 libc-2.5.so
18600 8.2490 vmlinux-syms-2.6.16.33-xen
8806 3.9054 libxpcom_core.so
8342 3.6996 libxpconnect.so
5327 2.3625 Xorg
CPU_CLK_UNHALT...|
samples| %|

So, why I am getting segmentation fault with "opreport -l" ( Second Question ). I am not sure what logs you may need or is there a way i can detect the fault ? let me know any other information I can give you.

Thanks again.

-John

On Feb 4, 2008 4:20 PM, Fischer, Anna <anna.fischer@xxxxxx> wrote:

John,
Maybe you can post some more details on what exactly you've been running, Xen version, Dom0 kernel version, how exactly did you start xenoprofile and when exactly do you get the first error? Also, please provide more detailed logs from the segmentation fault that you see. That would be very helpful.

________________________________

From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of john bryant
Sent: 04 February 2008 21:06
To: xen-users@xxxxxxxxxxxxxxxxxxx; xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] Re: xenoprof error

Is there a xenoprof developer who can help me ?

On Feb 3, 2008 11:33 PM, john bryant <bryant.johan@xxxxxxxxx> wrote:

Hi.

For those who are new to xenoprof ( e.g. Todd), I have installed xenorpof patch from following website (please follow the instructions):
http://xenoprof.sourceforge.net/

For oprofile, I have downloaded from
http://oprofile.sourceforge.net/news/
I have used oprofile 0.9.2

-John

On Feb 3, 2008 10:51 PM, Todd Deshane <deshantm@xxxxxxxxx> wrote:

Hi John,

I don't think you provided the list with enough information. How did you get to this point? What did you install etc.? For example I don't have a opcontrol file that I am aware of.

I don't have much experience with xenoprof either. Could you provide links or packages and a litlle more background into the problem?

Best Regards,
Todd

On Feb 3, 2008 6:57 PM, john bryant <bryant.johan@xxxxxxxxx> wrote:

Hi,

Any help with following mail will be great. I am using oprofile 0.9.2. Also, when i run"opreport -l" I get segmentation fault. Running "opreport" works correctly. Any help will be great. I am using fedora core.

-John

On Feb 1, 2008 11:06 AM, john bryant <bryant.johan@xxxxxxxxx> wrote:

Hi ,

I am getting following error when i try to run xenoprof

/usr/local/bin/opcontrol: line 1082: echo: write error: Invalid argument
Using 2.6+ OProfile kernel interface.
Reading module info.
Using log file /var/lib/oprofile/oprofiled.log
Daemon started.

what is this write error ?

Any help is really appreciated.

-John

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

RE: [Xen-users] Re: xenoprof error