Freebsd igb issue with i211-at embedded nics. Why drops all rx packets?

[DPona]

On our custom motherboard (based on Intel Atom C3538) there are four i211-at nics each on a separate PCIe port. Our motherboard passed intel code review.

The problem is that all incoming packets drops on these network adapters under FreeBSD. I test that on FreeBSD 12.1.

On other operating systems this problem is not. I checked it on CentOS, Fedora, Ubuntu and Windows 10.

When sending a ping command from one of i211 embedded nics, I see an arp request from it, but apparently it does not receive a response. Ping failed with "ping: sendto: Host is down" (I ping our gateway). The output of the "netstat -i" command shows incriasing droped packets.I did not make any addition configurations to the igb driver. All default settings after installation. I compiled the latest intel igb driver 2.5.9 for FreeBSD 12.1 - it did not help. 

On the same motherboard there are other network adapters - two Marvell x553 1GBit - they work fine.

Advise how to find the reason for this behavior (Ping not worcking - drops all rx packets)?

[VincentT_Intel]

Hi DPona,

Thank you for posting in Intel® Ethernet Communities.

Based on the supported operating system on this link. The I211-AT supports FreeBSD 10.x\11.x, kindly confirm if you also encounter the same problem on FreeBSD 11.3.

Looking forward to your response.

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker at Intel

[DPona]

Hi VinceT_Intel

Well, I installed 11.3. Unfortunately, the problem remains. We used "I211_Invm_NoAPM_v0.6.txt" file for our embedded i211 nics. (./eeupdate64e -NIC=1 -INVMUPDATE -FILE=./I211_Invm_NoAPM_v0.6.txt)

Here is what gives netstat -iW:

root@ipc:~ # netstat -iW Name Mtu Network Address Ipkts Ierrs Idrop Opkts Oerrs Coll igb0 1500 <Link#1> 70:b3:d5:f2:f0:a0 142 0 630 72 0 0 igb0 - 192.168.0.0/24 192.168.0.234 0 - - 78 - - igb1* 1500 <Link#2> 70:b3:d5:f2:f0:a1 0 0 0 0 0 0 igb2* 1500 <Link#3> 70:b3:d5:f2:f0:a2 0 0 0 0 0 0 igb3* 1500 <Link#4> 70:b3:d5:f2:f0:a3 0 0 0 0 0 0 ix0* 1500 <Link#5> 70:b3:d5:f2:f0:a4 0 0 0 0 0 0 ix1* 1500 <Link#6> 70:b3:d5:f2:f0:a5 0 0 0 0 0 0 ix2* 1500 <Link#7> 70:b3:d5:f2:f0:a6 0 0 0 0 0 0 ix3* 1500 <Link#8> 70:b3:d5:f2:f0:a7 0 0 0 0 0 0 lo0 16384 <Link#9> lo0 0 0 0 0 0 0 lo0 - localhost localhost 0 - - 0 - - lo0 - fe80::%lo0/64 fe80::1%lo0 0 - - 0 - - lo0 - your-net localhost 0 - - 0 - -

Best Regards,

Dmitry Ponamorev

[VincentT_Intel]

Hi Dmitry,

Thank you for sharing the information.

Please get in touch with our FreeBSD support for them to assist you with your query. Kindly send email to freebsd@intel.com

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker at Intel

[VincentT_Intel]

Hi Dmitry,

Good day!

Hope you were able to get assistance from our FreeBSD support, please let us know if you have other query related to Intel® Ethernet products.

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker At Intel

[DPona]

Hi VinceT_Intel

I have not received any response from freebsd@intel.com. I wrote a letter Thu, 5 Dec 2019. (Freebsd 11.3 igb issue with i211-AT embedded nics. Drops all rx packets.)

Best Regards,

Dmitry Ponamorev

[VincentT_Intel]

Hi Dmitry,

Thanks for sharing the latest update. We'll check internally regarding your email to freebsd@intel.com and get back to this thread as soon as possible.

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker At Intel

[VincentT_Intel]

Hi Dmitry,

While waiting for freebsd support to respond, you can also submit your query through Intel Premier Support.

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker At Intel

[VincentT_Intel]

Hi Dmitry,

Good day!

Please let us know if there's other assistance needed from Intel Ethernet Communities.

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker At Intel

[DPona]

Hi VinceT_Intel

We ourselves were able to localize the problem (at least we think so :)).

The problem was the lack of coreboot support for older and rarely used operating systems such as FreeBSD, QNX, VxWorks and so on.

Helped us in this post of one of the coreboot users  Ilya Gurevich https://ticket.coreboot.org/issues/218

quotation:

During coreboot integration with Intel Denverton FSP I have found that PCIe bus enumeration doesn't care of matching Max Payload Size (MPS) of End Point (EP) and Root Complex (RC) as advised by PCIe standard.

FSP initializes MPS of RC to 512B whereas MPS of EP is left at default value 128B. As result EP device stops working correctly and reports uncorrectable error - Malformed TLP.

In my case EP is PCIe LAN Controller 82574L GEI (8086:10D3).

I have fixed this problem by adding new routine "pciexp_set_max_payload_size" that is called from "pciexp_tune_dev".

The updated files pciexp_device.c and pciexp.h are attached.

In this post  Ilya Gurevich gives the code that I integrated into coreboot and our problem has been solved. I211 work normal now.

VinceT_Intel Thank you for the excellent moderation.

Best Regards,

Dmitry Ponamorev

[VincentT_Intel]

Hi Dmitry,

Good to know the issue has been resolved. We'll proceed to close the support case associated with this thread.

Feel free to create new thread if you have other query related to Intel Ethernet products.

Best Regards,

Vince T.

Intel Customer Support

A Contingent Worker At Intel