Backtrace in xen/next-2.6.38 when running guest

Backtrace in xen/next-2.6.38 when running guest

[M A Young]

I get backtraces from the dom0 kernel when running a guest with networking 
(via netback) enabled such as the one below (there are some slight 
variations but the first two lines and most of the rest of the contents 
are the same).

 	Michael Young

BUG: sleeping function called from invalid context at mm/page_alloc.c:2190
in_atomic(): 1, irqs_disabled(): 0, pid: 0, name: swapper
Pid: 0, comm: swapper Not tainted 2.6.38-0.rc3.git0.1.xendom0.fc15.x86_64 
#1
Call Trace:
  <IRQ>  [<ffffffff81047b44>] ? __might_sleep+0xeb/0xf0
  [<ffffffff810dea22>] ? __alloc_pages_nodemask+0x98/0x772
  [<ffffffff81470426>] ? _raw_spin_lock+0xe/0x10
  [<ffffffffa02b3473>] ? ieee80211_rx_handlers+0x1773/0x17fe [mac80211]
  [<ffffffffa022c80e>] ? net_tx_build_gops+0x3e2/0x94d [xen_netback]
  [<ffffffff810066d5>] ? xen_force_evtchn_callback+0xd/0xf
  [<ffffffff81006c72>] ? check_events+0x12/0x20
  [<ffffffff81112e1c>] ? __kmalloc_node_track_caller+0xf8/0x118
  [<ffffffffa022c80e>] ? net_tx_build_gops+0x3e2/0x94d [xen_netback]
  [<ffffffff81108313>] ? alloc_pages_current+0xb6/0xd0
  [<ffffffff813b5047>] ? __alloc_skb+0x8d/0x133
  [<ffffffffa022bd1b>] ? netif_alloc_page.isra.14+0x1e/0x54 [xen_netback]
  [<ffffffffa022c93f>] ? net_tx_build_gops+0x513/0x94d [xen_netback]
  [<ffffffff8102abf9>] ? pvclock_clocksource_read+0x48/0xb7
  [<ffffffff81006a3a>] ? xen_vcpuop_set_next_event+0x51/0x63
  [<ffffffff8107a4a3>] ? clockevents_program_event+0x8e/0x90
  [<ffffffff8107b4b7>] ? tick_dev_program_event+0x36/0xf3
  [<ffffffffa022cd97>] ? net_tx_action+0x1e/0x590 [xen_netback]
  [<ffffffff8105a1ed>] ? arch_local_irq_restore+0xb/0xd
  [<ffffffff810066d5>] ? xen_force_evtchn_callback+0xd/0xf
  [<ffffffff81006c72>] ? check_events+0x12/0x20
  [<ffffffff8105a6d5>] ? tasklet_action+0x7f/0xd2
  [<ffffffff8105b090>] ? __do_softirq+0xd2/0x19d
  [<ffffffff81006c72>] ? check_events+0x12/0x20
  [<ffffffff8100aadc>] ? call_softirq+0x1c/0x30
  [<ffffffff8100c121>] ? do_softirq+0x46/0x81
  [<ffffffff8105b214>] ? irq_exit+0x49/0x8b
  [<ffffffff812a981b>] ? xen_evtchn_do_upcall+0x31/0x3e
  [<ffffffff8100ab2e>] ? xen_do_hypervisor_callback+0x1e/0x30
  <EOI>  [<ffffffff810013aa>] ? hypercall_page+0x3aa/0x1000
  [<ffffffff810013aa>] ? hypercall_page+0x3aa/0x1000
  [<ffffffff81006690>] ? xen_safe_halt+0x10/0x18
  [<ffffffff81010d46>] ? default_idle+0x4e/0x86
  [<ffffffff8100831d>] ? cpu_idle+0xa1/0xdb
  [<ffffffff81006c19>] ? xen_irq_enable_direct_end+0x0/0x7
  [<ffffffff8145a95f>] ? cpu_bringup_and_idle+0x13/0x15

Re: Backtrace in xen/next-2.6.38 when running guest

[Ian Campbell]

On Wed, 2011-02-02 at 19:20 +0000, M A Young wrote: 
> I get backtraces from the dom0 kernel when running a guest with networking 
> (via netback) enabled such as the one below (there are some slight 
> variations but the first two lines and most of the rest of the contents 
> are the same).

> BUG: sleeping function called from invalid context at mm/page_alloc.c:2190

Thanks.

Netback for upstream is still a WIP (I wasn't expecting it to be merged
anywhere yet and have already and will continue to rebase my branch).

In particular one of the things which needs to go away for upstreaming
is the tasklet vs. kthread mode option in favour of always using a
kthread. Doing this will hopefully resolve this issue since the
allocation will no longer be in a context where sleeping is disallowed.

I you can't wait for my next post of the driver for upstream then
setting MODPARM_netback_kthread = 1 in drivers/net/xen-netback/netback.c
will have roughly the same effect.

Ian.

Re: Backtrace in xen/next-2.6.38 when running guest

[Jeremy Fitzhardinge]

On 02/02/2011 11:31 AM, Ian Campbell wrote:
> On Wed, 2011-02-02 at 19:20 +0000, M A Young wrote: 
>> I get backtraces from the dom0 kernel when running a guest with networking 
>> (via netback) enabled such as the one below (there are some slight 
>> variations but the first two lines and most of the rest of the contents 
>> are the same).
>> BUG: sleeping function called from invalid context at mm/page_alloc.c:2190
> Thanks.
>
> Netback for upstream is still a WIP (I wasn't expecting it to be merged
> anywhere yet and have already and will continue to rebase my branch).

I merged it into next to get some exposure, but I'll happily revert and
remerge as you make changes.  What's your working branch?

> In particular one of the things which needs to go away for upstreaming
> is the tasklet vs. kthread mode option in favour of always using a
> kthread.

Yep.  Have you also dropped the smartpoll stuff yet?

    J

Re: Backtrace in xen/next-2.6.38 when running guest

[Ian Campbell]

On Wed, 2011-02-02 at 19:47 +0000, Jeremy Fitzhardinge wrote: 
> On 02/02/2011 11:31 AM, Ian Campbell wrote:
> > On Wed, 2011-02-02 at 19:20 +0000, M A Young wrote: 
> >> I get backtraces from the dom0 kernel when running a guest with networking 
> >> (via netback) enabled such as the one below (there are some slight 
> >> variations but the first two lines and most of the rest of the contents 
> >> are the same).
> >> BUG: sleeping function called from invalid context at mm/page_alloc.c:2190
> > Thanks.
> >
> > Netback for upstream is still a WIP (I wasn't expecting it to be merged
> > anywhere yet and have already and will continue to rebase my branch).
> 
> I merged it into next to get some exposure, but I'll happily revert and
> remerge as you make changes.  What's your working branch?

Same branch but I haven't pushed an update yet.

> > In particular one of the things which needs to go away for upstreaming
> > is the tasklet vs. kthread mode option in favour of always using a
> > kthread.
> 
> Yep.  Have you also dropped the smartpoll stuff yet?

Yes.

Ian.

Re: Backtrace in xen/next-2.6.38 when running guest

[Ian Campbell]

On Wed, 2011-02-02 at 21:41 +0000, Ian Campbell wrote:
> On Wed, 2011-02-02 at 19:47 +0000, Jeremy Fitzhardinge wrote: 
> > On 02/02/2011 11:31 AM, Ian Campbell wrote:
> > > On Wed, 2011-02-02 at 19:20 +0000, M A Young wrote: 
> > >> I get backtraces from the dom0 kernel when running a guest with networking 
> > >> (via netback) enabled such as the one below (there are some slight 
> > >> variations but the first two lines and most of the rest of the contents 
> > >> are the same).
> > >> BUG: sleeping function called from invalid context at mm/page_alloc.c:2190
> > > Thanks.
> > >
> > > Netback for upstream is still a WIP (I wasn't expecting it to be merged
> > > anywhere yet and have already and will continue to rebase my branch).
> > 
> > I merged it into next to get some exposure, but I'll happily revert and
> > remerge as you make changes.  What's your working branch?
> 
> Same branch but I haven't pushed an update yet.

Updated version pushed to: 
  git://xenbits.xen.org/people/ianc/linux-2.6.git upstream/dom0/backend/netback

Still a WIP. Next step is to look at how to get rid of xenvif_copy_skb
(previously netbk_copy_skb), either by making it unnecessary or by
generalising it for the net core.

There are a couple of changes which touch netfront here: completing the
namespacing of include/xen/interface/io/netif.h (the netif prefix is
reserved for core code) and improvements to the Kconfig help. I guess it
would be best to split those and send out first.

Ian.

Re: Backtrace in xen/next-2.6.38 when running guest

[Jeremy Fitzhardinge]

On 02/03/2011 12:38 AM, Ian Campbell wrote:
> On Wed, 2011-02-02 at 21:41 +0000, Ian Campbell wrote:
>> On Wed, 2011-02-02 at 19:47 +0000, Jeremy Fitzhardinge wrote: 
>>> On 02/02/2011 11:31 AM, Ian Campbell wrote:
>>>> On Wed, 2011-02-02 at 19:20 +0000, M A Young wrote: 
>>>>> I get backtraces from the dom0 kernel when running a guest with networking 
>>>>> (via netback) enabled such as the one below (there are some slight 
>>>>> variations but the first two lines and most of the rest of the contents 
>>>>> are the same).
>>>>> BUG: sleeping function called from invalid context at mm/page_alloc.c:2190
>>>> Thanks.
>>>>
>>>> Netback for upstream is still a WIP (I wasn't expecting it to be merged
>>>> anywhere yet and have already and will continue to rebase my branch).
>>> I merged it into next to get some exposure, but I'll happily revert and
>>> remerge as you make changes.  What's your working branch?
>> Same branch but I haven't pushed an update yet.
> Updated version pushed to: 
>   git://xenbits.xen.org/people/ianc/linux-2.6.git upstream/dom0/backend/netback

Is that a rebase?

    J

Re: Backtrace in xen/next-2.6.38 when running guest

[Ian Campbell]

On Thu, 2011-02-03 at 20:51 +0000, Jeremy Fitzhardinge wrote: 
> On 02/03/2011 12:38 AM, Ian Campbell wrote:
> > On Wed, 2011-02-02 at 21:41 +0000, Ian Campbell wrote:
> >> On Wed, 2011-02-02 at 19:47 +0000, Jeremy Fitzhardinge wrote: 
> >>> On 02/02/2011 11:31 AM, Ian Campbell wrote:
> >>>> On Wed, 2011-02-02 at 19:20 +0000, M A Young wrote: 
> >>>>> I get backtraces from the dom0 kernel when running a guest with networking 
> >>>>> (via netback) enabled such as the one below (there are some slight 
> >>>>> variations but the first two lines and most of the rest of the contents 
> >>>>> are the same).
> >>>>> BUG: sleeping function called from invalid context at mm/page_alloc.c:2190
> >>>> Thanks.
> >>>>
> >>>> Netback for upstream is still a WIP (I wasn't expecting it to be merged
> >>>> anywhere yet and have already and will continue to rebase my branch).
> >>> I merged it into next to get some exposure, but I'll happily revert and
> >>> remerge as you make changes.  What's your working branch?
> >> Same branch but I haven't pushed an update yet.
> > Updated version pushed to: 
> >   git://xenbits.xen.org/people/ianc/linux-2.6.git upstream/dom0/backend/netback
> 
> Is that a rebase?

Yes.


> J