Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 15 Mar 2007 09:02:30 +0000
From:      "Bruce M. Simpson" <bms@FreeBSD.org>
To:        emilec@clarotech.co.za
Cc:        freebsd-stable@freebsd.org
Subject:   Re: Openvpn tap uses 99% cpu time
Message-ID:  <45F90BA6.7090102@FreeBSD.org>
In-Reply-To: <000d01c766cd$3f6aef50$7efaa8c0@clarotech.co.za>
References:  <000d01c766cd$3f6aef50$7efaa8c0@clarotech.co.za>

next in thread | previous in thread | raw e-mail | index | archive | help
Emile Coetzee wrote:
> I have done some more investigation. Rolling back to RELENG_6_2 solves the
> problem. I have now had this problem on 3 boxes in 2 days and have been able
> to reproduce it on a 4th in our lab.
>
> Starting openvpn with full debug it stops just before the point where the
> TAP devices gets initiated. And then the output from top shows openvpn on
> 99% cpu time (well almost).
>   
The recent changes to tap(4) were tested with openvpn and ppp before 
committing, and this is an isolated report at the moment. Other users on 
the list have suggested that the problem you are reporting does not 
affect FreeBSD alone.

Please consider doing the following:

1. Use ktrace(8) to extract a kdump of openvpn as it runs into this 
problem, so that you may determine exactly which syscalls are invoked 
when you see this problem, and post it on this list.

Consider using the ktrace pid mode to attach to openvpn as it blocks, as 
opposed to running openvpn under ktrace as a child process.

2. Implementing the usleep() fix as discussed by luigi, as the 
information in point 1 will give clues as to where the code is spinning.

Regards,
BMS




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?45F90BA6.7090102>