Linux driver e1000e buffers old packages

[MarcinKotulski-eFAB]

We're developing some time-vulnerable software system using Intel products as a hardware and we discovered little unusual problem with Intel network card driver for Linux. In some scenarios our device can  send packages from past which  is disturbing for our monitor module.

How to reproduce effect:
1. Connect two devices directly with Ethernet cable (only one have to be  with Intel network card and driver e1000e) and address them as standard LAN. Prepare also third device with prepared tcpdump/wireshark on its interface.
2. Start sending traffic (for example flooding ping) from device A (with Intel card) do device B.
3. Disconnect cable from device B.
4. Stop traffic on device A
5. Connect device A to device C.
6. On device C you will see in package dump software that device received some packages (1-20) addressed for device B from disconnection moment.

We're working on our own distribution with Linux kernel 5.4 but effect can be reproduced with Ubuntu 20.04 (on device A).

Probable reason:
Driver don't clear tx rings on link lost, so there is very small period of time when sent package can be put to driver and won't be send physically.

Temporary problem round-around:
Detect in userspace no-carrier on interfejs and do down and up operations.

[Devi_Intel]

Greeting MarcinKotulski-eFAB ,

Thank you for contacting Ethernet Community, Appreciates you can advise the specific product details for me to share datasheet and driver of the product model. as you mention "Linux driver e1000e buffers old packages" its has no specific model. for further concern please do let us know.

We will follow up after 3 business days in case we don't receive a response.

Thank you & Best Regards

Devi_Intel

[Devi_Intel]

Greetings  MarcinKotulski-eFAB ,

Regarding your case number 06002518, we are following up to find out if you were able to complete the actions we previously recommended.

Please reply to this thread to confirm, so we can continue helping with a resolution. Looking forward to receiving your reply

Thank you & Best Regards,

Devi_Intel

[MarcinKotulski-eFAB]

Hello again.

We made main tests on standard PC with card:
[    1.214104] e1000e 0000:00:1f.6 eth0: (PCI Express:2.5GT/s:Width x1) 70:85:c2:2a:f3:0b
[    1.214105] e1000e 0000:00:1f.6 eth0: Intel(R) PRO/1000 Network Connection
[    1.214163] e1000e 0000:00:1f.6 eth0: MAC: 12, PHY: 12, PBA No: FFFFFF-0FF

but we discovered this effect on different devices with different ethernet cards - unfortunately I don't have access to them now.

Last week we tried to reproduce effect with our 'production' devices but we didn't have 'luck' (ethernet cards were handled there with different drivers: 'igb'  & 'ixgbe').

At this point we have temporary solution to workaround this problem but we will try in near future to make additional tests on our target devices  to see if problem is only with 'e1000e' driver.

Best regards.
MJK

[Devi_Intel]

Greeting MarcinKotulski-eFAB

Regarding your case number 06002518, we are following up to find out if you were able to complete the actions we previously recommended.

Please reply to this email to confirm, so we can continue helping on a resolution. Looking forward to receiving your reply

Best regards,

Devi_Intel

[Devi_Intel]

Greeting MarcinKotulski-eFAB

We have not received a reply from you regarding case number 06002518. As such, we will be closing your case.

If you want to continue support, please reply to this email and we will reopen your case or create a new one so that we can continue to support you.

Best regards,

Devi_Intel

[Fariz_Intel]

Hi MarcinKotulski-eFAB,

Thank you for sharing the insight on the workaround. If you need any additional information, please submit a new question as this thread will no longer be monitored.

Best regards,

Fariz_Intel