[Intel-wired-lan] Packet loss on 82571EB with kernel 4.9

[Jeff Kirsher]

On Tue, 2017-10-17 at 03:05 +0000, Gavin Lambert wrote:
> (Please explicitly CC me in any replies.)
> I have a computer with an Intel 82571EB 2-port card using the e1000e
> driver.  I've noticed that it appears to be dropping some transmitted
> packets:
> 
> # iperf3 -c 10.200.197.17
> Connecting to host 10.200.197.17, port 5201
> [  4] local 10.200.197.82 port 59294 connected to 10.200.197.17 port
> 5201
> [ ID] Interval           Transfer     Bandwidth       Retr  Cwnd
> [  4]   0.00-1.00   sec   113 MBytes   951 Mbits/sec    0    223
> KBytes
> [  4]   1.00-2.00   sec   112 MBytes   941 Mbits/sec    0    223
> KBytes
> [  4]   2.00-3.00   sec   112 MBytes   941 Mbits/sec   10    204
> KBytes
> [  4]   3.00-4.00   sec   112 MBytes   941 Mbits/sec    0    221
> KBytes
> [  4]   4.00-5.00   sec   112 MBytes   941 Mbits/sec   10    154
> KBytes
> [  4]   5.00-6.00   sec   112 MBytes   942 Mbits/sec    0    222
> KBytes
> [  4]   6.00-7.00   sec   112 MBytes   941 Mbits/sec    0    222
> KBytes
> [  4]   7.00-8.00   sec   112 MBytes   941 Mbits/sec    0    222
> KBytes
> [  4]   8.00-9.00   sec   112 MBytes   942 Mbits/sec    0    222
> KBytes
> [  4]   9.00-10.00  sec   112 MBytes   941 Mbits/sec    0    222
> KBytes
> - - - - - - - - - - - - - - - - - - - - - - - - -
> [ ID] Interval           Transfer     Bandwidth       Retr
> [  4]   0.00-10.00  sec  1.10 GBytes   942
> Mbits/sec   20             sender
> [  4]   0.00-10.00  sec  1.10 GBytes   941
> Mbits/sec                  receiver
> 
> # iperf3 -c 10.200.197.17 -u -b0
> ...
> [ ID]
> Interval           Transfer     Bandwidth       Jitter    Lost/Total
> Datagrams
> [  4]   0.00-10.00  sec  1.12 GBytes   958 Mbits/sec  0.123
> ms  28/146256 (0.019%)
> [  4] Sent 146256 datagrams
> 
> (A reverse run with -R, where this PC is only receiving packets,
> doesn't report any problems.)
> 
> The computer is running Debian Linux Stretch x86_64 with kernel
> 4.9.30-2+deb9u2.
>        e1000e: Intel(R) PRO/1000 Network Driver - 3.2.6-k
>        e1000e: Copyright(c) 1999 - 2015 Intel Corporation.
>        e1000e 0000:01:00.0: Interrupt Throttling Rate (ints/sec) set
> to dynamic conservative mode
>        e1000e 0000:01:00.0 eth0: (PCI Express:2.5GT/s:Width x4)
> 68:05:ca:16:5a:24
>        e1000e 0000:01:00.0 eth0: Intel(R) PRO/1000 Network Connection
>        e1000e 0000:01:00.0 eth0: MAC: 0, PHY: 4, PBA No: D50868-008
>        e1000e 0000:01:00.1: Interrupt Throttling Rate (ints/sec) set
> to dynamic conservative mode
>        e1000e 0000:01:00.1 eth1: (PCI Express:2.5GT/s:Width x4)
> 68:05:ca:16:5a:25
>        e1000e 0000:01:00.1 eth1: Intel(R) PRO/1000 Network Connection
>        e1000e 0000:01:00.1 eth1: MAC: 0, PHY: 4, PBA No: D50868-008
>        e1000e 0000:01:00.0 enp1s0f0: renamed from eth0
>        e1000e 0000:01:00.1 enp1s0f1: renamed from eth1
>        e1000e: enp1s0f0 NIC Link is Up 1000 Mbps Full Duplex, Flow
> Control: None
>        IPv6: ADDRCONF(NETDEV_CHANGE): enp1s0f0: link becomes ready
> 
> The tests above were performed with enp1s0f0, but unsurprisingly
> enp1s0f1 behaves the same way.  The computer also has a non-Intel
> network card and performing the same tests on that does not produce
> any errors.
> 
> If I downgrade the computer to Debian Jessie with kernel 3.2.54 and
> e1000e driver 1.5.1-k then there is no packet loss in the same
> test.  I don't currently have anything in between these versions,
> although I might be able to build something custom if that will help.
> 
> What information is needed to diagnose this further and correct the
> problem?

Any chance you can try the latest stable kernel 4.13.8 or even the
latest kernel from Linus's tree (4.14-rc5)?  This will let us know if
the issue still persists in the latest upstream kernel or if the issue
has been fixed already.

Of course, if the issue is fixed in either of the two newer kernels,
then you would just need to work with Debian about getting a driver
update to thier 4.9 kernel release.

I have also added Sasha (e1000e maintainer) to this email thread so
that he is aware of the issue and can provide additional assistance in
troubleshooting the issue.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20171018/f98981ee/attachment.asc>