Re: [mpls-linux-general] Re: [MPLS-Linux] mpls-linux-0.700 : ATM tests & bugs

[Olivier D. <Oli...@rd...>]

"James R. Leu" wrote:
> 
> On Thu, Nov 30, 2000 at 12:50:54PM +0100, Olivier Dugeon wrote:
> > Hi all,
> >
> > I just finish some test with mpls-0.700 over ATM.
> >
> > 1/ The code doesn't work like 0.601 for the FORE 200E atm card. You can
> > establish connection but when you try to send data (ie. ping) the kernel
> > crash :-(( with :
> >
> > Nov 30 11:21:20 lat4272 kernel: Unable to handle kernel paging request
> > at virtual address 00000084
> 
> Do you get an EIP?  Send me the output if you have it, it might be in your
> log file /var/log/message
> 

Unfortunatly, NO. The kernel hang and panic before to have time to write
msg. into the /var/log/message file. On the screen i see once an eip
message, but the kernel message push it out the screen and i don't have
time to write it. What i mentioned previously is copy by hand from the
console output.

> > 2/ The code work with ENI ATM card, even though i notice some bug.
> >
> > - When i send ping, there is some info in the kern.log file about
> > unknown protocol :
> >
> > Nov 30 11:19:24 lat4272 kernel: mpls_output: unknown protocol(0400)
> > Nov 30 11:19:25 lat4272 kernel: mpls_output: unknown protocol(0608)
> > Nov 30 11:19:26 lat4272 kernel: mpls_output: unknown protocol(3781)
> > Nov 30 11:19:27 lat4272 kernel: mpls_output: unknown protocol(3781)
> > Nov 30 11:19:28 lat4272 kernel: mpls_output: unknown protocol(0608)
> 
> This has been an annoyance for sometime now.  The IP stack doesn't
> always set the SKB protocol before sending it on.  I need to find
> the common path in the IPv4 stack to set that, thus fixing the problem.
> 

Yes, not all the packet generate this error. For a ping there is around
60% of the packet with a bad SKB protocol.

> > - When i try to delete an incomming label it fail :
> >
> > # mplsadm -v -D -I atm:0/50:0
> > In label input: atm:0/50:0
> > Segmentation fault
> >
> > kern.log says :
> >
> > Nov 30 13:17:09 lat4271 kernel: mpls_del_in_label: enter
> > Nov 30 13:17:09 lat4271 kernel: __mpls_del_in_label: enter
> > Nov 30 13:17:09 lat4271 kernel: mpls_atm_push: enter
> > Nov 30 13:17:09 lat4271 kernel: Unable to handle kernel NULL pointer
> > dereference at virtual address 00000084
> > Nov 30 13:17:09 lat4271 kernel:  printing eip:
> > Nov 30 13:17:09 lat4271 kernel: c024073d
> > Nov 30 13:17:09 lat4271 kernel: *pde = 00000000
> > Nov 30 13:17:09 lat4271 kernel: Oops: 0000
> > Nov 30 13:17:09 lat4271 kernel: CPU:    0
> >
> > And the netlink socket is busy :
> 
> Very useful.  I will look into this one.  Could you very that it is
> reproducable?

Yes, each time i try to delete.
 
I not mentioned that after an eip message all the syslog output ie
/var/log/message, /var/log/kernel tty12 (i use Mandrake 7.1 which put
all the syslog message to tty12)
are completely freeze. A kill -UP to syslogd just send a message saying
that syslog restart but it's all. The next message is when you reboot
the system.

Hope you this help,

Olivier

-- 
 FTR&D/DAC/CPN
 Technopole Anticipa     | mailto:Oli...@fr...
 2, Avenue Pierre Marzin | Phone:  +(33) 2 96 05 28 80
 F-22307 LANNION         | Fax:    +(33) 2 96 05 18 52