[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] domU with two eth but only one is active


  • To: xen-users@xxxxxxxxxxxxx
  • From: SZABO Zsolt <szazs@xxxxxxxxx>
  • Date: Sat, 18 Jul 2015 19:00:26 +0200 (CEST)
  • Delivery-date: Sat, 18 Jul 2015 17:01:46 +0000
  • List-id: Xen user discussion <xen-users.lists.xen.org>

On Sat, 18 Jul 2015, Ian Campbell wrote:
On Sat, 2015-07-18 at 11:07 +0200, SZABO Zsolt wrote:
[...]
I have also checked dmesg in the guest:
# dmesg |grep -i net
[    0.033297] Initializing cgroup subsys net_cls
[    0.033330] Initializing cgroup subsys net_prio
[    0.118117] NET: Registered protocol family 16
[    0.208776] NET: Registered protocol family 2
[    0.212141] NET: Registered protocol family 1
[    0.301223] audit: initializing netlink subsys (disabled)
[    1.376536] NET: Registered protocol family 10
[    1.377181] NET: Registered protocol family 17
[    1.472973] xen_netfront: Initialising Xen virtual ethernet driver
[    1.601085] xen_netfront: can't alloc rx grant refs

This is strange, and I would bet related to your problem.

[    1.601105] net eth1: only created 16 queues

and this is maybe related too.

I'm afraid I've run out of time to investigate for the weekend, but I'll
see what I can figure out next week.

In the meantime off the top of my head in addition to googling those
messages it might be worth investigating options for extending the
number of grant entries which a guest is allowed to use and perhaps
whether multiqueue networking can be disabled (therefore reducing the
need for grant entries).

Hmm.. I have no idea how to do that... :-)


Ok, I have another linux domU but with wheezy. Originally, only eth0 was
configured, but now I have also tried to add eth1 similarly as to
the jessie domU and it works...
The dmesg of this domU is shorter:
# dmesg |grep net
[    0.004000] Initializing cgroup subsys net_cls
[    0.004000] Initializing cgroup subsys net_prio
[    0.134838] audit: initializing netlink subsys (disabled)
[    1.340978] xen_netfront: Initialising Xen virtual ethernet driver

Maybe the jessie upgrade missed some feature of the later upgraded dom0?
Or it is about some kind of other conflict...?

I think a new feature of the Jessie kernel(s) is running up against some
resource limit when two NICs are used.

And one more report, with some summary[1-3]:

[1]Finally I have also tried to start the jessie domU with kernel 3.2.0:
$ uname -a
Linux linserver 3.2.0-4-amd64 #1 SMP Debian 3.2.68-1+deb7u2 x86_64 GNU/Linux
(still from wheezy)
... and it WORKS!

# dmesg |grep -i 'net\|eth'
[    0.040200] Initializing cgroup subsys net_cls
[    0.089581] NET: Registered protocol family 16
[    0.202444] NET: Registered protocol family 2
[    0.215637] NET: Registered protocol family 1
[    0.287638] audit: initializing netlink socket (disabled)
[    1.366956] NET: Registered protocol family 10
[    1.367898] NET: Registered protocol family 17
[    1.368263] Initializing network drop monitor service
[    1.435843] Initialising Xen virtual ethernet driver.
[   18.854461] FS-Cache: Netfs 'nfs' registered for caching
[   25.868039] eth0: no IPv6 routers present
[   27.740048] eth1: no IPv6 routers present

[2]So, the TWO vif in a wheezy domU with linux-3.16.0 work too;
# uname -a
Linux wheezy 3.16.0-4-amd64 #1 SMP Debian 3.16.7-ckt11-1 (2015-05-24) x86_64 
GNU/Linux
(the part of dmesg related to net|eth can be seen above)

[3]But the 2nd vif in a jessie domU with linux-3.16.0 does NOT,
and it is possibly related to the dmesg lines
* xen_netfront: can't alloc rx grant refs
* net eth1: only created 16 queues

Now, I run the jessie+3.2.0 domU and I will see how stable it is...

--
Zsolt

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.