next-20090318: sunhme does not transmit packets

next-20090318: sunhme does not transmit packets

[Alexander Beregalov]

Hi

Ultra 10 can boot, it receives DHCP answer, ssh works, but few minutes
after machine becomes unresponsible by net at all.
It does not respond even on ARP requests.

On Sparc host I see that `TX packets` does not grow, but RX grows.
The state of the interface is
UP BROADCAST RUNNING MULTICAST

There is no relevant kernel messages in dmesg.

next-20090317 worked fine.


Let me know if I can help.

Re: next-20090318: sunhme does not transmit packets

[Alexander Beregalov]

2009/3/18 Alexander Beregalov <a.beregalov@gmail.com>:
> Hi
>
> Ultra 10 can boot, it receives DHCP answer, ssh works, but few minutes
> after machine becomes unresponsible by net at all.
> It does not respond even on ARP requests.
>
> On Sparc host I see that `TX packets` does not grow, but RX grows.
> The state of the interface is
> UP BROADCAST RUNNING MULTICAST
>
> There is no relevant kernel messages in dmesg.
>
> next-20090317 worked fine.
>
>
> Let me know if I can help.
>

I have reproduced it on next-20090318 for the second time, but still
do not know how.
next-20090317 might be affected.

Now both packet counters grow, but tcpdump cannot capture any packet.

Re: next-20090318: sunhme does not transmit packets

[David Miller]

From: Alexander Beregalov <a.beregalov@gmail.com>
Date: Wed, 18 Mar 2009 19:58:40 +0300

> 2009/3/18 Alexander Beregalov <a.beregalov@gmail.com>:
> > Hi
> >
> > Ultra 10 can boot, it receives DHCP answer, ssh works, but few minutes
> > after machine becomes unresponsible by net at all.
> > It does not respond even on ARP requests.
> >
> > On Sparc host I see that `TX packets` does not grow, but RX grows.
> > The state of the interface is
> > UP BROADCAST RUNNING MULTICAST
> >
> > There is no relevant kernel messages in dmesg.
> >
> > next-20090317 worked fine.
> >
> >
> > Let me know if I can help.
> >
> 
> I have reproduced it on next-20090318 for the second time, but still
> do not know how.
> next-20090317 might be affected.
> 
> Now both packet counters grow, but tcpdump cannot capture any packet.

Does the current vanilla kernel work fine?

Re: next-20090318: sunhme does not transmit packets

[Alexander Beregalov]

2009/3/19 David Miller <davem@davemloft.net>:
> From: Alexander Beregalov <a.beregalov@gmail.com>
> Date: Wed, 18 Mar 2009 19:58:40 +0300
>
>> 2009/3/18 Alexander Beregalov <a.beregalov@gmail.com>:
>> > Hi
>> >
>> > Ultra 10 can boot, it receives DHCP answer, ssh works, but few minutes
>> > after machine becomes unresponsible by net at all.
>> > It does not respond even on ARP requests.
>> >
>> > On Sparc host I see that `TX packets` does not grow, but RX grows.
>> > The state of the interface is
>> > UP BROADCAST RUNNING MULTICAST
>> >
>> > There is no relevant kernel messages in dmesg.
>> >
>> > next-20090317 worked fine.
>> >
>> >
>> > Let me know if I can help.
>> >
>>
>> I have reproduced it on next-20090318 for the second time, but still
>> do not know how.
>> next-20090317 might be affected.
>>
>> Now both packet counters grow, but tcpdump cannot capture any packet.
>
> Does the current vanilla kernel work fine?

Yes. The problem happens only with -next tree.

Re: next-20090318: sunhme does not transmit packets

[David Miller]

From: Alexander Beregalov <a.beregalov@gmail.com>
Date: Thu, 19 Mar 2009 11:16:20 +0300

> 2009/3/19 David Miller <davem@davemloft.net>:
> > Does the current vanilla kernel work fine?
> 
> Yes. The problem happens only with -next tree.

I have one of my ultra10s up and will try to reproduce and diagnose
this locally.

Thanks.

Re: next-20090318: sunhme does not transmit packets

[Alexander Beregalov]

>> 2009/3/19 David Miller <davem@davemloft.net>:
>> > Does the current vanilla kernel work fine?
>>
>> Yes. The problem happens only with -next tree.
>
> I have one of my ultra10s up and will try to reproduce and diagnose
> this locally.

Hi David

2.6.29 also does not work.

The guilty patch should be between 2.6.29-rc8 and final 2.6.29

even more narrow - 59fcbdd..v2.6.29

I can try to bisect.

Re: next-20090318: sunhme does not transmit packets

[Alexander Beregalov]

2009/3/24 Alexander Beregalov <a.beregalov@gmail.com>:
>>> 2009/3/19 David Miller <davem@davemloft.net>:
>>> > Does the current vanilla kernel work fine?
>>>
>>> Yes. The problem happens only with -next tree.
>>
>> I have one of my ultra10s up and will try to reproduce and diagnose
>> this locally.
>
> Hi David
>
> 2.6.29 also does not work.
>
> The guilty patch should be between 2.6.29-rc8 and final 2.6.29
>
> even more narrow - 59fcbdd..v2.6.29
>
> I can try to bisect.

e4a389a9b5c892446b5de2038bdc0cca8703c615 is first bad commit
commit e4a389a9b5c892446b5de2038bdc0cca8703c615
Author: Roel Kluin <roel.kluin@gmail.com>
Date:   Wed Mar 18 23:12:13 2009 -0700

    net: kfree(napi->skb) => kfree_skb

Re: next-20090318: sunhme does not transmit packets

[Alexander Beregalov]

2009/3/24 Alexander Beregalov <a.beregalov@gmail.com>:
> 2009/3/24 Alexander Beregalov <a.beregalov@gmail.com>:
>>>> 2009/3/19 David Miller <davem@davemloft.net>:
>>>> > Does the current vanilla kernel work fine?
>>>>
>>>> Yes. The problem happens only with -next tree.
>>>
>>> I have one of my ultra10s up and will try to reproduce and diagnose
>>> this locally.
>>
>> Hi David
>>
>> 2.6.29 also does not work.
>>
>> The guilty patch should be between 2.6.29-rc8 and final 2.6.29
>>
>> even more narrow - 59fcbdd..v2.6.29
>>
>> I can try to bisect.
>
> e4a389a9b5c892446b5de2038bdc0cca8703c615 is first bad commit

No, it is wrong, sorry.
I will do more tests tomorrow.

Re: next-20090318: sunhme does not transmit packets

[David Miller]

From: Alexander Beregalov <a.beregalov@gmail.com>
Date: Tue, 24 Mar 2009 18:50:55 +0300

> 2009/3/24 Alexander Beregalov <a.beregalov@gmail.com>:
> > e4a389a9b5c892446b5de2038bdc0cca8703c615 is first bad commit
> 
> No, it is wrong, sorry.
> I will do more tests tomorrow.

I bet it is the problem we are discussing here:

	http://marc.info/?l=linux-kernel&m=123789980524715&w=2

Herbert has posted two patches to try and fix this issue,
you can try the second one to see if it solves your bug
too.