[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Kai-Heng Feng]

Hi,

Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
I managed to get one at hand and I can confirm the issue is 100% reproducible.

The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].

Reverting "e1000e: Add support for S0ix" makes suspend work again.
I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.

Kai-Heng

[1] https://bugs.launchpad.net/bugs/1865570
[2] https://bugs.launchpad.net/bugs/1866170
[3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Neftin, Sasha]

On 3/17/2020 15:34, Kai-Heng Feng wrote:
> Hi,
> 
> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
> I managed to get one at hand and I can confirm the issue is 100% reproducible.
> 
> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
> 
> Reverting "e1000e: Add support for S0ix" makes suspend work again.
> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
> 
> Kai-Heng
> 
> [1] https://bugs.launchpad.net/bugs/1865570
> [2] https://bugs.launchpad.net/bugs/1866170
> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
> 
please, check of ME enabled on this system. try disable ME and re-run

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Kai-Heng Feng]

> On Mar 17, 2020, at 21:56, Neftin, Sasha <sasha.neftin@intel.com> wrote:
> 
> On 3/17/2020 15:34, Kai-Heng Feng wrote:
>> Hi,
>> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
>> I managed to get one at hand and I can confirm the issue is 100% reproducible.
>> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
>> Reverting "e1000e: Add support for S0ix" makes suspend work again.
>> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
>> Kai-Heng
>> [1] https://bugs.launchpad.net/bugs/1865570
>> [2] https://bugs.launchpad.net/bugs/1866170
>> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
> please, check of ME enabled on this system. try disable ME and re-run

I disabled "Intel AMT", issue is still reproducible.

Kai-Heng

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Vitaly Lifshits]

On 3/17/2020 16:04, Kai-Heng Feng wrote:
> 
> 
>> On Mar 17, 2020, at 21:56, Neftin, Sasha <sasha.neftin@intel.com> wrote:
>>
>> On 3/17/2020 15:34, Kai-Heng Feng wrote:
>>> Hi,
>>> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
>>> I managed to get one at hand and I can confirm the issue is 100% reproducible.
>>> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
>>> Reverting "e1000e: Add support for S0ix" makes suspend work again.
>>> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
>>> Kai-Heng
>>> [1] https://bugs.launchpad.net/bugs/1865570
>>> [2] https://bugs.launchpad.net/bugs/1866170
>>> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
>> please, check of ME enabled on this system. try disable ME and re-run
> 
> I disabled "Intel AMT", issue is still reproducible.
> 
> Kai-Heng
> 
Hi Kai,

Do you work with cable connected?
If so, please try to run this command:
sudo echo 3 > /sys/kernel/debug/pmc_core/ltr_ignore

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Kai-Heng Feng]

Hi Vitaly,

> On Mar 17, 2020, at 22:39, Vitaly Lifshits <vitaly.lifshits@intel.com> wrote:
> 
> On 3/17/2020 16:04, Kai-Heng Feng wrote:
>>> On Mar 17, 2020, at 21:56, Neftin, Sasha <sasha.neftin@intel.com> wrote:
>>> 
>>> On 3/17/2020 15:34, Kai-Heng Feng wrote:
>>>> Hi,
>>>> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
>>>> I managed to get one at hand and I can confirm the issue is 100% reproducible.
>>>> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
>>>> Reverting "e1000e: Add support for S0ix" makes suspend work again.
>>>> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
>>>> Kai-Heng
>>>> [1] https://bugs.launchpad.net/bugs/1865570
>>>> [2] https://bugs.launchpad.net/bugs/1866170
>>>> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
>>> please, check of ME enabled on this system. try disable ME and re-run
>> I disabled "Intel AMT", issue is still reproducible.
>> Kai-Heng
> Hi Kai,
> 
> Do you work with cable connected?

No, ethernet cable is not plugged.

> If so, please try to run this command:
> sudo echo 3 > /sys/kernel/debug/pmc_core/ltr_ignore

Same issue after doing that.

Though unlikely, can this be caused by the special design of X1 Carbon?
To keep the laptop thin, it doesn't have a RJ45 port. Instead it has a special connecter which requires a RJ45 dongle to support ethernet.
Maybe that's why the PHY part isn't working after suspend?

Kai-Heng

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Neftin, Sasha]

On 3/18/2020 05:12, Kai-Heng Feng wrote:
> Hi Vitaly,
> 
>> On Mar 17, 2020, at 22:39, Vitaly Lifshits <vitaly.lifshits@intel.com> wrote:
>>
>> On 3/17/2020 16:04, Kai-Heng Feng wrote:
>>>> On Mar 17, 2020, at 21:56, Neftin, Sasha <sasha.neftin@intel.com> wrote:
>>>>
>>>> On 3/17/2020 15:34, Kai-Heng Feng wrote:
>>>>> Hi,
>>>>> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
>>>>> I managed to get one at hand and I can confirm the issue is 100% reproducible.
>>>>> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
>>>>> Reverting "e1000e: Add support for S0ix" makes suspend work again.
>>>>> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
>>>>> Kai-Heng
>>>>> [1] https://bugs.launchpad.net/bugs/1865570
>>>>> [2] https://bugs.launchpad.net/bugs/1866170
>>>>> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
>>>> please, check of ME enabled on this system. try disable ME and re-run
>>> I disabled "Intel AMT", issue is still reproducible.
>>> Kai-Heng
>> Hi Kai,
>>
>> Do you work with cable connected?
> 
> No, ethernet cable is not plugged.
> 
>> If so, please try to run this command:
>> sudo echo 3 > /sys/kernel/debug/pmc_core/ltr_ignore
> 
> Same issue after doing that.
> 
> Though unlikely, can this be caused by the special design of X1 Carbon?
> To keep the laptop thin, it doesn't have a RJ45 port. Instead it has a special connecter which requires a RJ45 dongle to support ethernet.
> Maybe that's why the PHY part isn't working after suspend?
> 
Hello Kai-Heng,
Thanks for your response. I would like clarify/understand Lenovo X1 
Carbon setup:
How laptop connected to the docking station? Thunderbolt or Lenovo 
proprietary connection?
Where is LAN located (physically) in a laptop or dock?
PHY located in docking, right?

And last question: does this system supported Sx or S0ix (do you know)?
The reason I ask it due from my understanding dock always connected to 
the power and system should be Sx supported.

Therefore may be we need introduce difference behavior between Sx and 
S0ix support. (if I understand correctly, such systems as Lenovo Carbon 
X1 should be Sx support, we need ignore S0ix entry/exit in such cases - 
ACPI interface should provide us with this info)

> Kai-Heng
> 
Sasha

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Kai-Heng Feng]

Hi Sasha,

> On Mar 18, 2020, at 15:32, Neftin, Sasha <sasha.neftin@intel.com> wrote:
> 
> On 3/18/2020 05:12, Kai-Heng Feng wrote:
>> Hi Vitaly,
>>> On Mar 17, 2020, at 22:39, Vitaly Lifshits <vitaly.lifshits@intel.com> wrote:
>>> 
>>> On 3/17/2020 16:04, Kai-Heng Feng wrote:
>>>>> On Mar 17, 2020, at 21:56, Neftin, Sasha <sasha.neftin@intel.com> wrote:
>>>>> 
>>>>> On 3/17/2020 15:34, Kai-Heng Feng wrote:
>>>>>> Hi,
>>>>>> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
>>>>>> I managed to get one at hand and I can confirm the issue is 100% reproducible.
>>>>>> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
>>>>>> Reverting "e1000e: Add support for S0ix" makes suspend work again.
>>>>>> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
>>>>>> Kai-Heng
>>>>>> [1] https://bugs.launchpad.net/bugs/1865570
>>>>>> [2] https://bugs.launchpad.net/bugs/1866170
>>>>>> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
>>>>> please, check of ME enabled on this system. try disable ME and re-run
>>>> I disabled "Intel AMT", issue is still reproducible.
>>>> Kai-Heng
>>> Hi Kai,
>>> 
>>> Do you work with cable connected?
>> No, ethernet cable is not plugged.
>>> If so, please try to run this command:
>>> sudo echo 3 > /sys/kernel/debug/pmc_core/ltr_ignore
>> Same issue after doing that.
>> Though unlikely, can this be caused by the special design of X1 Carbon?
>> To keep the laptop thin, it doesn't have a RJ45 port. Instead it has a special connecter which requires a RJ45 dongle to support ethernet.
>> Maybe that's why the PHY part isn't working after suspend?
> Hello Kai-Heng,
> Thanks for your response. I would like clarify/understand Lenovo X1 Carbon setup:
> How laptop connected to the docking station?

There's no docking station involved. Just a regular e1000e device connects to root complex.

> Thunderbolt or Lenovo proprietary connection?

No thunderbolt involved.

I'd say it's proprietary:
https://www.lenovo.com/us/en/accessories-and-monitors/cables-and-adapters/adapters/CABLE-BO-Ethernet-Extension-Adapter-2/p/4X90Q84427

The ethernet port is next to thunderbolt, but it's not thunderbolt:
https://images.idgesg.net/images/article/2018/08/left-ports-100769115-orig.jpg

> Where is LAN located (physically) in a laptop or dock?
> PHY located in docking, right?

This I don't know. I'll raise the issue to Lenovo hardware team.

> 
> And last question: does this system supported Sx or S0ix (do you know)?

Sx you mean S3?

I think the system supports S3 but it defaults to S0ix.

> The reason I ask it due from my understanding dock always connected to the power and system should be Sx supported.

There's no dock.
For this issue, _nothing_ connects to the ethernet port.

> 
> Therefore may be we need introduce difference behavior between Sx and S0ix support. (if I understand correctly, such systems as Lenovo Carbon X1 should be Sx support, we need ignore S0ix entry/exit in such cases - ACPI interface should provide us with this info)

For this case S0ix is used.

Kai-Heng

> 
>> Kai-Heng
> Sasha

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Neftin, Sasha]

On 3/18/2020 10:23, Kai-Heng Feng wrote:
> Hi Sasha,
> 
>> On Mar 18, 2020, at 15:32, Neftin, Sasha <sasha.neftin@intel.com> wrote:
>>
>> On 3/18/2020 05:12, Kai-Heng Feng wrote:
>>> Hi Vitaly,
>>>> On Mar 17, 2020, at 22:39, Vitaly Lifshits <vitaly.lifshits@intel.com> wrote:
>>>>
>>>> On 3/17/2020 16:04, Kai-Heng Feng wrote:
>>>>>> On Mar 17, 2020, at 21:56, Neftin, Sasha <sasha.neftin@intel.com> wrote:
>>>>>>
>>>>>> On 3/17/2020 15:34, Kai-Heng Feng wrote:
>>>>>>> Hi,
>>>>>>> Users reported that X1 Carbon 7th can only suspend once due of "e1000e 0000:00:1f.6 enp0s31f6: Hardware Error" [1] [2].
>>>>>>> I managed to get one at hand and I can confirm the issue is 100% reproducible.
>>>>>>> The error occurs at the first e1e_rphy() in e1000_copper_link_setup_82577() [3].
>>>>>>> Reverting "e1000e: Add support for S0ix" makes suspend work again.
>>>>>>> I also tried commit e1738282f6c6 "e1000e: fix S0ix flows for cable connected case" however the issue persists.
>>>>>>> Kai-Heng
>>>>>>> [1] https://bugs.launchpad.net/bugs/1865570
>>>>>>> [2] https://bugs.launchpad.net/bugs/1866170
>>>>>>> [3] https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/tree/drivers/net/ethernet/intel/e1000e/phy.c?h=dev-queue&id=e1738282f6c6ad0ec56a15c4d5a3b657d90e4435#n630
>>>>>> please, check of ME enabled on this system. try disable ME and re-run
>>>>> I disabled "Intel AMT", issue is still reproducible.
>>>>> Kai-Heng
>>>> Hi Kai,
>>>>
>>>> Do you work with cable connected?
>>> No, ethernet cable is not plugged.
>>>> If so, please try to run this command:
>>>> sudo echo 3 > /sys/kernel/debug/pmc_core/ltr_ignore
>>> Same issue after doing that.
>>> Though unlikely, can this be caused by the special design of X1 Carbon?
>>> To keep the laptop thin, it doesn't have a RJ45 port. Instead it has a special connecter which requires a RJ45 dongle to support ethernet.
>>> Maybe that's why the PHY part isn't working after suspend?
>> Hello Kai-Heng,
>> Thanks for your response. I would like clarify/understand Lenovo X1 Carbon setup:
>> How laptop connected to the docking station?
> 
> There's no docking station involved. Just a regular e1000e device connects to root complex.
> 
>> Thunderbolt or Lenovo proprietary connection?
> 
> No thunderbolt involved.
> 
> I'd say it's proprietary:
> https://www.lenovo.com/us/en/accessories-and-monitors/cables-and-adapters/adapters/CABLE-BO-Ethernet-Extension-Adapter-2/p/4X90Q84427
> 
> The ethernet port is next to thunderbolt, but it's not thunderbolt:
> https://images.idgesg.net/images/article/2018/08/left-ports-100769115-orig.jpg
> 
>> Where is LAN located (physically) in a laptop or dock?
>> PHY located in docking, right?
> 
> This I don't know. I'll raise the issue to Lenovo hardware team.
> 
ok, let's understand where is PHY location.
Since no docking connected why driver should run? There is no RJ45 port 
on laptop.
S0ix flow same (I got confirmation from our architecture team) and 
should haven't impact on Sx.
if PHY located externally... during S0ix flow we try access to the PHY 
and it can be problem. Since laptop have no RJ45 connector driver not 
expected to run I though.
>>
>> And last question: does this system supported Sx or S0ix (do you know)?
> 
> Sx you mean S3?
> 
> I think the system supports S3 but it defaults to S0ix.
> 
>> The reason I ask it due from my understanding dock always connected to the power and system should be Sx supported.
> 
> There's no dock.
> For this issue, _nothing_ connects to the ethernet port.
> 
>>
>> Therefore may be we need introduce difference behavior between Sx and S0ix support. (if I understand correctly, such systems as Lenovo Carbon X1 should be Sx support, we need ignore S0ix entry/exit in such cases - ACPI interface should provide us with this info)
> 
> For this case S0ix is used.
> 
> Kai-Heng
> 
>>
>>> Kai-Heng
>> Sasha
> 

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Kai-Heng Feng]

Hi Sasha,

> On Mar 18, 2020, at 16:50, Neftin, Sasha <sasha.neftin@intel.com> wrote:

[snipped]

> ok, let's understand where is PHY location.

Lenovo confirmed the PHY is inside the laptop.

> Since no docking connected why driver should run?

Because the SoC in the laptop has an I219 device?

> There is no RJ45 port on laptop.

The Lenovo ethernet dongle wires the RJ45 to the I219 device.

> S0ix flow same (I got confirmation from our architecture team) and should haven't impact on Sx.
> if PHY located externally... during S0ix flow we try access to the PHY and it can be problem. Since laptop have no RJ45 connector driver not expected to run I though.

But the I219 device is always on the PCI bus so e1000e should bind to it?

Kai-Heng

[Intel-wired-lan] [Regression] "e1000e: Add support for S0ix" breaks s2idle on Lenovo X1 Carbon 7th

[Neftin, Sasha]

On 3/18/2020 14:21, Kai-Heng Feng wrote:
> Hi Sasha,
> 
>> On Mar 18, 2020, at 16:50, Neftin, Sasha <sasha.neftin@intel.com> wrote:
> 
> [snipped]
> 
>> ok, let's understand where is PHY location.
> 
> Lenovo confirmed the PHY is inside the laptop.
> 
>> Since no docking connected why driver should run?
> 
> Because the SoC in the laptop has an I219 device?
> 
>> There is no RJ45 port on laptop.
> 
> The Lenovo ethernet dongle wires the RJ45 to the I219 device.
> 
>> S0ix flow same (I got confirmation from our architecture team) and should haven't impact on Sx.
>> if PHY located externally... during S0ix flow we try access to the PHY and it can be problem. Since laptop have no RJ45 connector driver not expected to run I though.
> 
> But the I219 device is always on the PCI bus so e1000e should bind to it?
> 
ok. Need to debug this system. We will escalate this problem to our PAE.
> Kai-Heng
>