Opened bug 0000230 about "bearerbox causes 99% cpu load"

(for the records in this users mailing list)

Stephane

--
Stephane Rosa
IT Security & Network Specialist, EMEA
Tel:  +41 21 966-1238
Cell: +41 79 616-0828

Stryker EMEA
Grand Rue 92
CH-1820 Montreux
http://www.europe.stryker.com



Stephane Rosa <[EMAIL PROTECTED]>
Sent by: [EMAIL PROTECTED]

25-05-05 12:39

To
Julien Buratto <[EMAIL PROTECTED]>
cc
[email protected]
Subject
Re: cpu load 99% by bearerbox






I just compiled the same sources on a dual-cpu Sun server and

the behavior is exactly the same as on RedHat 9 (SE).


How come is this happening ? Are we the only one experiencing

this ??


Stephane

--
Stephane Rosa
IT Security & Network Specialist, EMEA
Tel:  +41 21 966-1238
Cell: +41 79 616-0828

Stryker EMEA
Grand Rue 92
CH-1820 Montreux
http://www.europe.stryker.com


Stephane Rosa <[EMAIL PROTECTED]>
Sent by: [EMAIL PROTECTED]

25-05-05 11:34


To
Julien Buratto <[EMAIL PROTECTED]>
cc
[email protected]
Subject
Re: cpu load 99% by bearerbox








Fiddling with gdb, I found out that the high cpu load

is caused by the call to gwthread_sleep(10.0); at line

605 of bearerbox.c.


Tracing further, gwthread_sleep calls poll(), with a

timeout of 10 seconds. When I step that line, gdb

hangs for the entire 10 seconds and the cpu raises

straight to 99%.


poll() is supposed to sleep until something happend

on an fd, I have no idea why this happens.


Those two pieces of code seem not to have

changed since 2003, so I'm surprised we are

not all of us seeing the same issue.


I will compile and run it on Solaris to see if it

makes any difference. Maybe poll() behaves

differently on RedHat SE.


Cheers

Stephane



--
Stephane Rosa
IT Security & Network Specialist, EMEA
Tel:  +41 21 966-1238
Cell: +41 79 616-0828

Stryker EMEA
Grand Rue 92
CH-1820 Montreux
http://www.europe.stryker.com

Julien Buratto <[EMAIL PROTECTED]>
Sent by: [EMAIL PROTECTED]

24-05-05 17:49


To
Stephane Rosa <[EMAIL PROTECTED]>
cc
[email protected]
Subject
Re: cpu load 99% by bearerbox









Stephane Rosa wrote:
>
> Thanks Julien, I did that already and the gdb thing
> reveals a backtrace on gwthread_sleep like you were
> experiencing I guess.
>
> May I ask how you solved your problem ?
>
> Thanks
>

Make this laught: I havent' :)
I'm still using 1.4.0 because CVS goes into cpu overload so I've decided
to manually patch the 1.4.0 and apply there my own code.

J


Reply via email to