http://bugzilla.opensuse.org/show_bug.cgi?id=1101244 http://bugzilla.opensuse.org/show_bug.cgi?id=1101244#c3 --- Comment #3 from Mark Stutter --- I did some more checks and have an important update on this issue: If I switch the network setup to be managed by "wicked service" in yast2 the card is recognized as 1000Mbit/s and all works fine! Switching back to NetworkManager it is recognized as 100Mbit and it is slow again. This is consistently reported by ethtool and dmesg. The relevant part of the output of dmesg when switching from wicked to Networkmanager is: [ 27.316255] e1000e: Intel(R) PRO/1000 Network Driver - 3.2.6-k [ 27.316257] e1000e: Copyright(c) 1999 - 2015 Intel Corporation. [ 27.316513] e1000e 0000:00:19.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode [ 27.421613] e1000e 0000:00:19.0 0000:00:19.0 (uninitialized): registered PHC clock [ 27.524324] e1000e 0000:00:19.0 eth0: (PCI Express:2.5GT/s:Width x1) 3c:97:0e:96:b4:04 [ 27.524332] e1000e 0000:00:19.0 eth0: Intel(R) PRO/1000 Network Connection [ 27.524380] e1000e 0000:00:19.0 eth0: MAC: 10, PHY: 11, PBA No: 1000FF-0FF [ 27.526354] e1000e 0000:00:19.0 enp0s25: renamed from eth0 ... I removed quite some irrelevant lines here ... [ 51.862235] No iBFT detected. [ 53.588838] bridge: filtering via arp/ip/ip6tables is no longer available by default. Update your scripts to load br_netfilter if you need this. [ 58.570215] e1000e: enp0s25 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None [ 59.212503] NET: Registered protocol family 17 [ 85.709208] ACPI Warning: \_SB.PCI0.PEG.VID._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20180313/nsarguments-66) [ 85.830706] ACPI Error: Field [TBF3] at bit offset/length 356352/32768 exceeds size of target Buffer (368640 bits) (20180313/dsopcode-201) [ 85.830711] ACPI Error: Method parse/execution failed \_SB.PCI0.PEG.VID.GETB, AE_AML_BUFFER_LIMIT (20180313/psparse-516) [ 85.830716] ACPI Error: Method parse/execution failed \_SB.PCI0.PEG.VID._ROM, AE_AML_BUFFER_LIMIT (20180313/psparse-516) [ 85.831452] resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than pnp 00:00 [mem 0x000d0000-0x000d3fff] [ 85.831605] caller _nv001169rm+0xe3/0x1d0 [nvidia] mapping multiple BARs [ 123.891592] fuse init (API version 7.26) [ 498.170363] e1000e: enp0s25 NIC Link is Down [ 500.617976] IPv6: ADDRCONF(NETDEV_UP): enp0s25: link is not ready [ 500.861507] IPv6: ADDRCONF(NETDEV_UP): enp0s25: link is not ready [ 500.868493] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 500.881728] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.132367] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.232349] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 501.271091] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.520894] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.619873] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 502.271853] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 503.803325] e1000e: enp0s25 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None [ 503.803379] IPv6: ADDRCONF(NETDEV_CHANGE): enp0s25: link becomes ready [ 508.005355] e1000e: enp0s25 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None [ 508.005363] e1000e 0000:00:19.0 enp0s25: 10/100 speed: disabling TSO I think it is in the line "[ 498.170363] e1000e: enp0s25 NIC Link is Down" that the network interface is shut down when yast switches from wicked to NetworkManager. before (wicked): e1000e: enp0s25 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None afterwards (NetworkManager): e1000e: enp0s25 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None I have no idea why that happens ... is it ultimately just a NetworkManager issue? But the speed of the card should be managed by the driver, right? In opensuse Leap it seems to work fine. -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1101244 http://bugzilla.opensuse.org/show_bug.cgi?id=1101244#c5 --- Comment #5 from david chang --- (In reply to Mark Stutter from comment #3)

I did some more checks and have an important update on this issue:

If I switch the network setup to be managed by "wicked service" in yast2 the card is recognized as 1000Mbit/s and all works fine!

Actually I use networkmanager by default.

Switching back to NetworkManager it is recognized as 100Mbit and it is slow again.

I test again with switching between networkmanager and wicked and the result are the same, I do not encounter any speed down.

This is consistently reported by ethtool and dmesg. The relevant part of the output of dmesg when switching from wicked to Networkmanager is:

[ 27.316255] e1000e: Intel(R) PRO/1000 Network Driver - 3.2.6-k [ 27.316257] e1000e: Copyright(c) 1999 - 2015 Intel Corporation. [ 27.316513] e1000e 0000:00:19.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode [ 27.421613] e1000e 0000:00:19.0 0000:00:19.0 (uninitialized): registered PHC clock [ 27.524324] e1000e 0000:00:19.0 eth0: (PCI Express:2.5GT/s:Width x1) 3c:97:0e:96:b4:04 [ 27.524332] e1000e 0000:00:19.0 eth0: Intel(R) PRO/1000 Network Connection [ 27.524380] e1000e 0000:00:19.0 eth0: MAC: 10, PHY: 11, PBA No: 1000FF-0FF [ 27.526354] e1000e 0000:00:19.0 enp0s25: renamed from eth0 ... I removed quite some irrelevant lines here ... [ 51.862235] No iBFT detected. [ 53.588838] bridge: filtering via arp/ip/ip6tables is no longer available by default. Update your scripts to load br_netfilter if you need this. [ 58.570215] e1000e: enp0s25 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None [ 59.212503] NET: Registered protocol family 17 [ 85.709208] ACPI Warning: \_SB.PCI0.PEG.VID._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20180313/nsarguments-66) [ 85.830706] ACPI Error: Field [TBF3] at bit offset/length 356352/32768 exceeds size of target Buffer (368640 bits) (20180313/dsopcode-201) [ 85.830711] ACPI Error: Method parse/execution failed \_SB.PCI0.PEG.VID.GETB, AE_AML_BUFFER_LIMIT (20180313/psparse-516) [ 85.830716] ACPI Error: Method parse/execution failed \_SB.PCI0.PEG.VID._ROM, AE_AML_BUFFER_LIMIT (20180313/psparse-516) [ 85.831452] resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than pnp 00:00 [mem 0x000d0000-0x000d3fff]

Do you know what's these ACPI error?

[ 85.831605] caller _nv001169rm+0xe3/0x1d0 [nvidia] mapping multiple BARs [ 123.891592] fuse init (API version 7.26) [ 498.170363] e1000e: enp0s25 NIC Link is Down [ 500.617976] IPv6: ADDRCONF(NETDEV_UP): enp0s25: link is not ready [ 500.861507] IPv6: ADDRCONF(NETDEV_UP): enp0s25: link is not ready [ 500.868493] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 500.881728] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.132367] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.232349] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 501.271091] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.520894] iwlwifi 0000:03:00.0: Radio type=0x0-0x3-0x1 [ 501.619873] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 502.271853] IPv6: ADDRCONF(NETDEV_UP): wlp3s0: link is not ready [ 503.803325] e1000e: enp0s25 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None [ 503.803379] IPv6: ADDRCONF(NETDEV_CHANGE): enp0s25: link becomes ready [ 508.005355] e1000e: enp0s25 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None [ 508.005363] e1000e 0000:00:19.0 enp0s25: 10/100 speed: disabling TSO

I think it is in the line "[ 498.170363] e1000e: enp0s25 NIC Link is Down" that the network interface is shut down when yast switches from wicked to NetworkManager.

before (wicked): e1000e: enp0s25 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None afterwards (NetworkManager): e1000e: enp0s25 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None

Could you give it a try to change the speed manually? # ethtool -s enp0s25 speed 1000 duplex full autoneg off -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1101244 http://bugzilla.opensuse.org/show_bug.cgi?id=1101244#c7 Ben Steel changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |bhs_suse@precisionforesight | |.com --- Comment #7 from Ben Steel --- Perhaps Mr. Stutter could check his cables? I confirmed many of his results on two ThinkPads, including the difference between Wicked and Network Manager, before discovering a bad patch cable in a wiring closet. Gigabit requires 4 pairs; 100 Mbit requires two. Presented with a bad cable, Network Manager fails, while Wicked gracefully autonegotiates down to 100 (with a Cisco switch; IEEE standard requires going to lowest supported). As for horrible throughput, if autonegotiation failed, both a Cisco and a standard IEEE switch would go to half-duplex on the switch end only (we only see Suse duplex outputs in his reports). A duplex mismatch would cause collisions on the half-duplex end even though send and receive are still physically different pairs (it thinks there's an old-fashioned hub on the line). During a sustained transfer to test speeds, where all the ACKs are causing collisons and resends, speeds could be abysmal. Lastly, since Mr. Stutter is using a laptop, he might not always be talking through a bad cable, explaining all manner of inconsistent results. Hope this helps. -- You are receiving this mail because: You are on the CC list for the bug.