Virtualization - Dec 2018 - 4.20-rc6: WARNING: CPU: 30 PID: 197360 at net/core/flow_dissector.c:764 __skb_flow_dissect

On Thu, Dec 20, 2018 at 11:17 AM Ido Schimmel <idosch at idosch.org>
wrote:
>
> On Thu, Dec 20, 2018 at 03:09:22PM +0100, Christian Borntraeger wrote:
> > On 20.12.2018 10:12, Ido Schimmel wrote:
> > > +Willem
> > >
> > > On Thu, Dec 20, 2018 at 08:45:40AM +0100, Christian Borntraeger
wrote:
> > >> Folks,
> > >>
> > >> I got this warning today. I cant tell when and why this
happened, so I do not know yet how to reproduce.
> > >> Maybe someone has a quick idea.
> > >>
> > >> [85109.572032] WARNING: CPU: 30 PID: 197360 at
net/core/flow_dissector.c:764 __skb_flow_dissect+0x1f0/0x1318
> > >
> > > I managed to trigger this warning as well the other day, but from
a
> > > different call path:
> >
> > FWIW, it also seems to happen on 4.20-rc1. 4.19.0 seems fine. bisect
seem to have failed so
> > my reproducer is not reliable.
>
> Yes, it is caused by commit d0e13a1488ad ("flow_dissector: lookup
netns
> by skb->sk if skb->dev is NULL")
>
> $ git tag --contains d0e13a1488ad
> v4.20-rc1
> v4.20-rc2
> v4.20-rc3
> v4.20-rc4
> v4.20-rc5
> v4.20-rc6
That tap_get_user_xdp path is also new for 4.20-rc1:

    commit 0efac27791ee068075d80f07c55a229b1335ce12
    tap: accept an array of XDP buffs through sendmsg()

    $ git describe --contains 0efac27791ee
    v4.20-rc1~14^2~382^2~1

In v4.19 and before all packets went through tap_get_user.

On 20.12.2018 18:23, Willem de Bruijn wrote:
> On Thu, Dec 20, 2018 at 11:17 AM Ido Schimmel <idosch at idosch.org>
wrote:
>>
>> On Thu, Dec 20, 2018 at 03:09:22PM +0100, Christian Borntraeger wrote:
>>> On 20.12.2018 10:12, Ido Schimmel wrote:
>>>> +Willem
>>>>
>>>> On Thu, Dec 20, 2018 at 08:45:40AM +0100, Christian Borntraeger
wrote:
>>>>> Folks,
>>>>>
>>>>> I got this warning today. I cant tell when and why this
happened, so I do not know yet how to reproduce.
>>>>> Maybe someone has a quick idea.
>>>>>
>>>>> [85109.572032] WARNING: CPU: 30 PID: 197360 at
net/core/flow_dissector.c:764 __skb_flow_dissect+0x1f0/0x1318
>>>>
>>>> I managed to trigger this warning as well the other day, but
from a
>>>> different call path:
>>>
>>> FWIW, it also seems to happen on 4.20-rc1. 4.19.0 seems fine.
bisect seem to have failed so
>>> my reproducer is not reliable.
>>
>> Yes, it is caused by commit d0e13a1488ad ("flow_dissector: lookup
netns
>> by skb->sk if skb->dev is NULL")
>>
>> $ git tag --contains d0e13a1488ad
>> v4.20-rc1
>> v4.20-rc2
>> v4.20-rc3
>> v4.20-rc4
>> v4.20-rc5
>> v4.20-rc6
> 
> That tap_get_user_xdp path is also new for 4.20-rc1:
> 
>     commit 0efac27791ee068075d80f07c55a229b1335ce12
>     tap: accept an array of XDP buffs through sendmsg()
> 
>     $ git describe --contains 0efac27791ee
>     v4.20-rc1~14^2~382^2~1
> 
> In v4.19 and before all packets went through tap_get_user.
Hmmm, so maybe my bisect wasnt broken at all? It pointed to

commit 105bc1306e9b29c2aa2783b9524f7aec9b5a5b1f
Merge: 3475372ff60e4 d0e13a1488ad3
Author:     David S. Miller <davem at davemloft.net>
AuthorDate: Tue Sep 25 20:29:38 2018 -0700
Commit:     David S. Miller <davem at davemloft.net>
CommitDate: Tue Sep 25 20:29:38 2018 -0700

    Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next

On Fri, Dec 21, 2018 at 1:45 AM Christian Borntraeger
<borntraeger at de.ibm.com> wrote:
>
>
>
> On 20.12.2018 18:23, Willem de Bruijn wrote:
> > On Thu, Dec 20, 2018 at 11:17 AM Ido Schimmel <idosch at
idosch.org> wrote:
> >>
> >> On Thu, Dec 20, 2018 at 03:09:22PM +0100, Christian Borntraeger
wrote:
> >>> On 20.12.2018 10:12, Ido Schimmel wrote:
> >>>> +Willem
> >>>>
> >>>> On Thu, Dec 20, 2018 at 08:45:40AM +0100, Christian
Borntraeger wrote:
> >>>>> Folks,
> >>>>>
> >>>>> I got this warning today. I cant tell when and why
this happened, so I do not know yet how to reproduce.
> >>>>> Maybe someone has a quick idea.
> >>>>>
> >>>>> [85109.572032] WARNING: CPU: 30 PID: 197360 at
net/core/flow_dissector.c:764 __skb_flow_dissect+0x1f0/0x1318
> >>>>
> >>>> I managed to trigger this warning as well the other day,
but from a
> >>>> different call path:
> >>>
> >>> FWIW, it also seems to happen on 4.20-rc1. 4.19.0 seems fine.
bisect seem to have failed so
> >>> my reproducer is not reliable.
> >>
> >> Yes, it is caused by commit d0e13a1488ad ("flow_dissector:
lookup netns
> >> by skb->sk if skb->dev is NULL")
> >>
> >> $ git tag --contains d0e13a1488ad
> >> v4.20-rc1
> >> v4.20-rc2
> >> v4.20-rc3
> >> v4.20-rc4
> >> v4.20-rc5
> >> v4.20-rc6
> >
> > That tap_get_user_xdp path is also new for 4.20-rc1:
> >
> >     commit 0efac27791ee068075d80f07c55a229b1335ce12
> >     tap: accept an array of XDP buffs through sendmsg()
> >
> >     $ git describe --contains 0efac27791ee
> >     v4.20-rc1~14^2~382^2~1
> >
> > In v4.19 and before all packets went through tap_get_user.
>
> Hmmm, so maybe my bisect wasnt broken at all? It pointed to
>
> commit 105bc1306e9b29c2aa2783b9524f7aec9b5a5b1f
> Merge: 3475372ff60e4 d0e13a1488ad3
> Author:     David S. Miller <davem at davemloft.net>
> AuthorDate: Tue Sep 25 20:29:38 2018 -0700
> Commit:     David S. Miller <davem at davemloft.net>
> CommitDate: Tue Sep 25 20:29:38 2018 -0700
>
>     Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
Yes, that's the right commit. The flow dissector change went in
through bpf-next.