Regression in e1000e since Kernel 4.14.3

Regression in e1000e since Kernel 4.14.3

[rwarsow]

Hallo

someone and I got an regression with e1000e since kernel 4.14.3 and it 
seems there is 4.14.4 on the way without a fix.


bug report is here:

https://bugzilla.kernel.org/show_bug.cgi?id=198047
-- 

Greeting

Ronald

Re: Regression in e1000e since Kernel 4.14.3

[Gabriel C]

On 04.12.2017 23:10, rwarsow@gmx.de wrote:

> Hallo
> 
> someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
> 
> 
> bug report is here:
> 
> https://bugzilla.kernel.org/show_bug.cgi?id=198047

( added stable and netdev to CC )

Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.

Interesting here , when using connman the connection is stable.

Regards,

Gabriel C

Re: Regression in e1000e since Kernel 4.14.3

[Greg KH]

On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
> On 04.12.2017 23:10, rwarsow@gmx.de wrote:
> 
> > Hallo
> > 
> > someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
> > 
> > 
> > bug report is here:
> > 
> > https://bugzilla.kernel.org/show_bug.cgi?id=198047
> 
> ( added stable and netdev to CC )
> 
> Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.

Again, can people try 4.14.5-rc1?  It should be resolved there.

thanks,

greg k-h

Re: Regression in e1000e since Kernel 4.14.3

[Greg KH]

On Tue, Dec 05, 2017 at 07:18:34AM +0100, Greg KH wrote:
> On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
> > On 04.12.2017 23:10, rwarsow@gmx.de wrote:
> > 
> > > Hallo
> > > 
> > > someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
> > > 
> > > 
> > > bug report is here:
> > > 
> > > https://bugzilla.kernel.org/show_bug.cgi?id=198047
> > 
> > ( added stable and netdev to CC )
> > 
> > Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.
> 
> Again, can people try 4.14.5-rc1?  It should be resolved there.

Oops, that would be 4.14.4-rc1.  Any why do you say above that is on the
way without a fix, did you test it?

thanks,

greg k-h

Re: Regression in e1000e since Kernel 4.14.3

[Gabriel C]

On 05.12.2017 07:19, Greg KH wrote:
> On Tue, Dec 05, 2017 at 07:18:34AM +0100, Greg KH wrote:
>> On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
>>> On 04.12.2017 23:10, rwarsow@gmx.de wrote:
>>>
>>>> Hallo
>>>>
>>>> someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
>>>>
>>>>
>>>> bug report is here:
>>>>
>>>> https://bugzilla.kernel.org/show_bug.cgi?id=198047
>>>
>>> ( added stable and netdev to CC )
>>>
>>> Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.
>>
>> Again, can people try 4.14.5-rc1?  It should be resolved there.
> 
> Oops, that would be 4.14.4-rc1.  Any why do you say above that is on the
> way without a fix, did you test it?

I didn't tested 4.14.4-rc1 but somone from the bug report tested it and told is not resolved.

I'll fire up an build in a bit and let you know.

Re: Regression in e1000e since Kernel 4.14.3

[Greg KH]

On Tue, Dec 05, 2017 at 09:20:33AM +0100, Gabriel C wrote:
> On 05.12.2017 07:19, Greg KH wrote:
> > On Tue, Dec 05, 2017 at 07:18:34AM +0100, Greg KH wrote:
> > > On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
> > > > On 04.12.2017 23:10, rwarsow@gmx.de wrote:
> > > > 
> > > > > Hallo
> > > > > 
> > > > > someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
> > > > > 
> > > > > 
> > > > > bug report is here:
> > > > > 
> > > > > https://bugzilla.kernel.org/show_bug.cgi?id=198047
> > > > 
> > > > ( added stable and netdev to CC )
> > > > 
> > > > Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.
> > > 
> > > Again, can people try 4.14.5-rc1?  It should be resolved there.
> > 
> > Oops, that would be 4.14.4-rc1.  Any why do you say above that is on the
> > way without a fix, did you test it?
> 
> I didn't tested 4.14.4-rc1 but somone from the bug report tested it and told is not resolved.
> 
> I'll fire up an build in a bit and let you know.

Great, and maybe cc: the developers and mailing list for this driver at
the same time?  :)

Re: Regression in e1000e since Kernel 4.14.3

[Gabriel C]

On 05.12.2017 09:53, Greg KH wrote:
> On Tue, Dec 05, 2017 at 09:20:33AM +0100, Gabriel C wrote:
>> On 05.12.2017 07:19, Greg KH wrote:
>>> On Tue, Dec 05, 2017 at 07:18:34AM +0100, Greg KH wrote:
>>>> On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
>>>>> On 04.12.2017 23:10, rwarsow@gmx.de wrote:
>>>>>
>>>>>> Hallo
>>>>>>
>>>>>> someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
>>>>>>
>>>>>>
>>>>>> bug report is here:
>>>>>>
>>>>>> https://bugzilla.kernel.org/show_bug.cgi?id=198047
>>>>>
>>>>> ( added stable and netdev to CC )
>>>>>
>>>>> Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.
>>>>
>>>> Again, can people try 4.14.5-rc1?  It should be resolved there.
>>>
>>> Oops, that would be 4.14.4-rc1.  Any why do you say above that is on the
>>> way without a fix, did you test it?
>>
>> I didn't tested 4.14.4-rc1 but somone from the bug report tested it and told is not resolved.
>>
>> I'll fire up an build in a bit and let you know.
> 
> Great, and maybe cc: the developers and mailing list for this driver at
> the same time?  :)
> 

Greg,

last time I reported something about e100* someone told me to just CC netdev =)

However the issue still remains with 4.14.4-rc1 and NM , and is still fine with connman.

I don't even think is something about the driver itself because I've quick compiled the out-of-tree-e1000e
and breaks with NM in the same way. ( which should not be possible ? )

Even when 4.14.3 was biggiSH :) after a quick scan and assuming the e1000e patches are fine , remaining candidates
should be the IRQ* and x86/* ones ?


NM seems to go in a loop setup link , kill link , setup link etc like this :

Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5925] policy: auto-activating connection 'enp6s0_1'
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5940] device (enp6s0): Activation: starting connection 'enp6s0_1' (8a825b19-b086-3413-901c-508ee26b4138)
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5943] device (enp6s0): state change: disconnected -> prepare (reason 'none', sys-iface-state: 'managed')
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5945] manager: NetworkManager state is now CONNECTING
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5950] device (enp6s0): state change: prepare -> config (reason 'none', sys-iface-state: 'managed')
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5958] device (enp6s0): state change: config -> ip-config (reason 'none', sys-iface-state: 'managed')
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5962] dhcp4 (enp6s0): activation: beginning transaction (timeout in 45 seconds)
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.5980] dhcp4 (enp6s0): dhclient started with pid 1180
Dez 05 09:51:45 zwerg dhclient[1180]: DHCPREQUEST on enp6s0 to 255.255.255.255 port 67
Dez 05 09:51:45 zwerg dhclient[1180]: DHCPACK from 192.168.178.1
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6268] dhcp4 (enp6s0):   address 192.168.178.21
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6268] dhcp4 (enp6s0):   plen 24 (255.255.255.0)
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6269] dhcp4 (enp6s0):   gateway 192.168.178.1
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6269] dhcp4 (enp6s0):   lease time 864000
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6269] dhcp4 (enp6s0):   nameserver '192.168.178.1'
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6270] dhcp4 (enp6s0):   domain name 'fritz.box'
Dez 05 09:51:45 zwerg NetworkManager[807]: <info>  [1512463905.6270] dhcp4 (enp6s0): state changed unknown -> bound

and then :

Dez 05 09:52:17 zwerg NetworkManager[807]: <info>  [1512463937.5788] device (enp6s0): state change: ip-config -> failed (reason 'ip-config-unavailable', sys-iface-state: 'managed')
Dez 05 09:52:17 zwerg NetworkManager[807]: <info>  [1512463937.5792] manager: NetworkManager state is now DISCONNECTED
Dez 05 09:52:17 zwerg NetworkManager[807]: <info>  [1512463937.5794] policy: disabling autoconnect for connection 'enp6s0_1'.
Dez 05 09:52:17 zwerg NetworkManager[807]: <warn>  [1512463937.5798] device (enp6s0): Activation: failed for connection 'enp6s0_1'
Dez 05 09:52:17 zwerg NetworkManager[807]: <info>  [1512463937.5807] device (enp6s0): state change: failed -> disconnected (reason 'none', sys-iface-state: 'managed')
Dez 05 09:52:17 zwerg NetworkManager[807]: <info>  [1512463937.5899] dhcp4 (enp6s0): canceled DHCP transaction, DHCP client pid 1180
Dez 05 09:52:17 zwerg NetworkManager[807]: <info>  [1512463937.5899] dhcp4 (enp6s0): state changed bound -> done

...

With connman all is fine and the connection is stable so the driver itself seems fine..

Re: Regression in e1000e since Kernel 4.14.3

[Greg KH]

On Tue, Dec 05, 2017 at 10:16:03AM +0100, Gabriel C wrote:
> On 05.12.2017 09:53, Greg KH wrote:
> > On Tue, Dec 05, 2017 at 09:20:33AM +0100, Gabriel C wrote:
> > > On 05.12.2017 07:19, Greg KH wrote:
> > > > On Tue, Dec 05, 2017 at 07:18:34AM +0100, Greg KH wrote:
> > > > > On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
> > > > > > On 04.12.2017 23:10, rwarsow@gmx.de wrote:
> > > > > > 
> > > > > > > Hallo
> > > > > > > 
> > > > > > > someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
> > > > > > > 
> > > > > > > 
> > > > > > > bug report is here:
> > > > > > > 
> > > > > > > https://bugzilla.kernel.org/show_bug.cgi?id=198047
> > > > > > 
> > > > > > ( added stable and netdev to CC )
> > > > > > 
> > > > > > Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.
> > > > > 
> > > > > Again, can people try 4.14.5-rc1?  It should be resolved there.
> > > > 
> > > > Oops, that would be 4.14.4-rc1.  Any why do you say above that is on the
> > > > way without a fix, did you test it?
> > > 
> > > I didn't tested 4.14.4-rc1 but somone from the bug report tested it and told is not resolved.
> > > 
> > > I'll fire up an build in a bit and let you know.
> > 
> > Great, and maybe cc: the developers and mailing list for this driver at
> > the same time?  :)
> > 
> 
> Greg,
> 
> last time I reported something about e100* someone told me to just CC netdev =)
> 
> However the issue still remains with 4.14.4-rc1 and NM , and is still fine with connman.
> 
> I don't even think is something about the driver itself because I've quick compiled the out-of-tree-e1000e
> and breaks with NM in the same way. ( which should not be possible ? )
> 
> Even when 4.14.3 was biggiSH :) after a quick scan and assuming the e1000e patches are fine , remaining candidates
> should be the IRQ* and x86/* ones ?

I am not assuming the e1000e patches are all fine :)

Any chance you can do a 'git bisect' between 4.14.2 and 4.14.3 to find
the offending patch?

thanks,

greg k-h

Re: Regression in e1000e since Kernel 4.14.3

[Gabriel C]

On 05.12.2017 10:23, Greg KH wrote:
> On Tue, Dec 05, 2017 at 10:16:03AM +0100, Gabriel C wrote:
>> On 05.12.2017 09:53, Greg KH wrote:
>>> On Tue, Dec 05, 2017 at 09:20:33AM +0100, Gabriel C wrote:
>>>> On 05.12.2017 07:19, Greg KH wrote:
>>>>> On Tue, Dec 05, 2017 at 07:18:34AM +0100, Greg KH wrote:
>>>>>> On Tue, Dec 05, 2017 at 12:47:10AM +0100, Gabriel C wrote:
>>>>>>> On 04.12.2017 23:10, rwarsow@gmx.de wrote:
>>>>>>>
>>>>>>>> Hallo
>>>>>>>>
>>>>>>>> someone and I got an regression with e1000e since kernel 4.14.3 and it seems there is 4.14.4 on the way without a fix.
>>>>>>>>
>>>>>>>>
>>>>>>>> bug report is here:
>>>>>>>>
>>>>>>>> https://bugzilla.kernel.org/show_bug.cgi?id=198047
>>>>>>>
>>>>>>> ( added stable and netdev to CC )
>>>>>>>
>>>>>>> Yes I have a box with e1000e and it seems something at least breaks NM after 4.14.3.
>>>>>>
>>>>>> Again, can people try 4.14.5-rc1?  It should be resolved there.
>>>>>
>>>>> Oops, that would be 4.14.4-rc1.  Any why do you say above that is on the
>>>>> way without a fix, did you test it?
>>>>
>>>> I didn't tested 4.14.4-rc1 but somone from the bug report tested it and told is not resolved.
>>>>
>>>> I'll fire up an build in a bit and let you know.
>>>
>>> Great, and maybe cc: the developers and mailing list for this driver at
>>> the same time?  :)
>>>
>>
>> Greg,
>>
>> last time I reported something about e100* someone told me to just CC netdev =)
>>
>> However the issue still remains with 4.14.4-rc1 and NM , and is still fine with connman.
>>
>> I don't even think is something about the driver itself because I've quick compiled the out-of-tree-e1000e
>> and breaks with NM in the same way. ( which should not be possible ? )
>>
>> Even when 4.14.3 was biggiSH :) after a quick scan and assuming the e1000e patches are fine , remaining candidates
>> should be the IRQ* and x86/* ones ?
> 
> I am not assuming the e1000e patches are all fine :)
> 
> Any chance you can do a 'git bisect' between 4.14.2 and 4.14.3 to find
> the offending patch?
> 

I can but I'm not sure I can do that today , the box is my working box and can't take it down right now.

Maybe I can do that tonight depending on how tired I am :)