[Barrelfish-users] Issue faced when starting bfscope

[Pravin Shinde]

Hi Akshay,

Sorry for delays in answering your question.  If you are still having
problems with running bfscope then here are some suggestions which
might help.

By looking at your logs, my guess is that the problem is in the
"netd" daemon.  This service handles the background network traffic
and also provides ARP service to all other applications.  I would
recommend turning on debugging in netd [1] to see where exactly it is
getting stuck.

The netd will try get an IP address using DHCP before it exports ARP
service. I would check to make sure that netd is indeed getting
IP address from DHCP server (DHCP server logs in your LAN might help
to see if barrelfish machine actually sent).  You can also give
static IP address by passing relevant arguments to netd daemon [2].
For example, you can add something like following in `menu.lst` to use
static IP address::

  module /x86_64/sbin/netd do_dhcp=0 ip=192.168.2.11 nm=255.255.255.0
gw=192.168.2.1 ns=192.168.2.2

Another suggestion would be to simplify the setup by trying to get
network stack run without bfscope and only with netd.  With netd running
correctly, you should be able to ping the machine.  Then you can
move on the setup with bfscope later, once you are certain about working
network.

I hope this helps.

Regards,
Pravin

[1]: barrelfish/usr/netd/netd_debug.h:17 ("NETD_DEBUG")
[2]: barrelfish/usr/netd/main.c:84

On 12/22/2014 02:05 AM, Akshay Giridhar Ravichandran wrote:
> Dear All,
>
> We are trying to set up the tracing module bfscope and facing an issue.
>
> It is waiting forever for e1000_ARP when the function
> nameservice_blocking_lookup() is called through
> idc_connect_ARP_lookup_service() while running bfscope. Following are the
> traces:
>
> spawnd.0: spawning /x86_64/sbin/bfscope on core 0
> bfscope running on core 0
> bfscope: trying to connect to the e1000 driver...
> lwip_init_ex: starting......................
> LWIP_other: Inside lwip_init
> #### bfscope Going ahead with default wait-set
> LWIP: lwip_init: done with sanity check
> LWIP: done with sanity check
> idc_c_port_mng_srv: trying to [e1000_PORTS_MNG]
> init_net_ports_connection: called
> init_net_ports_connection: connecting to [e1000_PORTS_MNG]
> init_net_ports_connection: resolving driver e1000_PORTS_MNG
> init_net_ports_connection: terminated
> idc_c_port_mng_srv: success [e1000_PORTS_MNG]
> LWIP: ctl 0
> init_net_ARP_connection: called e1000_ARP
> init_net_ARP_connection: resolving driver e1000_ARP
> *skb.0: waiting for: e1000_ARP*
>
> I looked into the code  of the file
> lib/lwip/src/barrelfish/ARP_lookup_client.c and there is a FIXME me note
> "decide if this is the best place to connect with net_ARP" in the function
> idc_connect_ARP_lookup_service.
>
> Could this be a potential cause for the block?
> If it is not kindly give us some insights as to where we have gone wrong.
>
> *Set Up:*
>
> We have and Intel 82576 and we are using e1000 as the cardname. We have
> both NKGDM and netd modules up as you can find in the attached log trace.
>
> Also please find the menu.lst we used to boot for your perusal.
>
> Your help is highly appreciated.
>
> Thanks and Regards,
> Akshay
>
>
>
> _______________________________________________
> Barrelfish-users mailing list
> Barrelfish-users at lists.inf.ethz.ch
> https://lists.inf.ethz.ch/mailman/listinfo/barrelfish-users
>