jerryc Posted January 12, 2017 Posted January 12, 2017 Hi Pulled latest git repo and build mainline image (Ubuntu/Debian both) 4.9.2-sun8i kernel. Eth0 now stuck at 100mb/s and half duplex. I have tried using ethertool to make it full duplex. Not working. The previous image which I used and suggested was Armbian_5.24.170109_Orangepizero_Ubuntu_xenial_4.9.0 This image works with full duplex. Is this a bug or there is more configs needed when I building my own images? From ethtool output, looks like problem with auto negotiation. pi@orangepizero:~$ sudo ethtool eth0 Settings for eth0: Supported ports: [ TP MII ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full Supported pause frame use: No Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full Advertised pause frame use: No Advertised auto-negotiation: Yes Link partner advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full Link partner advertised pause frame use: Symmetric Link partner advertised auto-negotiation: Yes Speed: 100Mb/s Duplex: Half Port: MII PHYAD: 1 Transceiver: external Auto-negotiation: on Current message level: 0x00000000 (0) Link detected: yes eth0: negotiated 100baseTx-HD, link ok product info: vendor 00:11:05, model 0 rev 0 basic mode: isolate, autonegotiation enabled basic status: autonegotiation complete, link ok capabilities: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD advertising: 100baseTx-HD 10baseT-FD link partner: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD flow-control -jerry
kprince Posted January 13, 2017 Posted January 13, 2017 I don't have this board yet, but I have done ethernet driver development in the past. You really want to make sure that your setup is valid. I suggest switching out the ethernet cable and switch. A lot of these embedded PHY's are quirky (that is why linux kernel is filled with vendor specific phy drivers) and different switches can cause issues. Another potential problem is a bad cable.
Igor Posted January 13, 2017 Posted January 13, 2017 @jerryc Our kernel sources are attached to development branches and there are a list of patches on top of this, that things works. Building from mainline sources without additional patches is not the way to go. Waste of time. ATM. That's why we have tools.
jerryc Posted January 13, 2017 Author Posted January 13, 2017 @admin The kernel was build using the github repo that you listed. Hence my question on why the latest build locks into half duplex. Unless you mean there are other batches beside your repo? or other ways to build? (I am following your build instructions) Unless you are saying there are additional patches that are not in the repo that you listed. Also, I am very possitive it is not switch/cable/power issue. the test was done using two images. One was provided by you a while back on another question, and one is the image I build using the repo. any ideas? -jerry @jerryc Our kernel sources are attached to development branches and there are a list of patches on top of this, that things works. Building from mainline sources without additional patches is not the way to go. Waste of time. ATM. That's why we have tools.
Igor Posted January 13, 2017 Posted January 13, 2017 any ideas? You are trying to use kernel built from developer branch where problems are more than expected. This kernel is not ready for usage ... we can't and we don't waste time on supporting something that changes within hours.
martinayotte Posted January 13, 2017 Posted January 13, 2017 My OPiZero is in full duplex, either with last week's build or today's build.
zador.blood.stained Posted January 13, 2017 Posted January 13, 2017 I remember that there were some negotiation issues with sun8i-emac before, but since this driver will be completely rewritten to integrate into existing dwmac, responding to any network issues now is a waste of time.
jerryc Posted January 13, 2017 Author Posted January 13, 2017 You are trying to use kernel built from developer branch where problems are more than expected. This kernel is not ready for usage ... we can't and we don't waste time on supporting something that changes within hours. Is there a way to tell the tool That's why we have tools, to build using different kernel? or it is fixed?
jerryc Posted January 14, 2017 Author Posted January 14, 2017 My OPiZero is in full duplex, either with last week's build or today's build. @martinayotte Just did a git pull and used compile.sh to create a new mainline image for orangepizero. The half duplex is still there. I assume you also used the compile.sh script for your images. If not let me know, I can try your way as well. root@orangepizero:~# uname -a Linux orangepizero 4.9.3-sun8i #5 SMP Fri Jan 13 15:57:42 PST 2017 armv7l GNU/Linux root@orangepizero:~# ethtool eth0 Settings for eth0: Supported ports: [ TP MII ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full Supported pause frame use: No Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full Advertised pause frame use: No Advertised auto-negotiation: Yes Link partner advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full Link partner advertised pause frame use: Symmetric Link partner advertised auto-negotiation: Yes Speed: 100Mb/s Duplex: Half Port: MII PHYAD: 1 Transceiver: external Auto-negotiation: on Current message level: 0x00000000 (0) Link detected: yes root@orangepizero:~# mii-tool -v eth0: negotiated 100baseTx-HD, link ok product info: vendor 00:11:05, model 0 rev 0 basic mode: isolate, autonegotiation enabled basic status: autonegotiation complete, link ok capabilities: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD advertising: 100baseTx-HD 10baseT-FD link partner: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD flow-control root@orangepizero:~#
Igor Posted January 14, 2017 Posted January 14, 2017 Is there a way to tell the tool That's why we have tools, to build using different kernel? or it is fixed? Yes, you can change sources, branches, ... this question was brought up few times. Start here: https://www.armbian.com/search_gcse and https://docs.armbian.com/Developer-Guide_Build-Preparation if you want to start learning more about the build process. I move this topic to development section, since it suits here more.
martinayotte Posted January 14, 2017 Posted January 14, 2017 @martinayotte Just did a git pull and used compile.sh to create a new mainline image for orangepizero. The half duplex is still there. I assume you also used the compile.sh script for your images. If not let me know, I can try your way as well. Yes, it was my own builds done with compile.sh
jerryc Posted January 15, 2017 Author Posted January 15, 2017 @martinayotte Appreciate your help. Just cloned a workspace and build a fresh image. (no special config, just a orange pi zero debian 4.9 server image) I am still getting the eth0 at half Duplex. An older image given to be by @igor this image works perfectly (with Full Duplex and solved stability issues) root@orangepizero:~# uname -a Linux orangepizero 4.9.3-sun8i #1 SMP Sat Jan 14 17:08:59 PST 2017 armv7l GNU/Linux root@orangepizero:~# mii-tool eth0 -v eth0: negotiated 100baseTx-HD, link ok product info: vendor 00:11:05, model 0 rev 0 basic mode: isolate, autonegotiation enabled basic status: autonegotiation complete, link ok capabilities: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD advertising: 100baseTx-HD 10baseT-FD link partner: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD flow-control Yes, it was my own builds done with compile.sh
szwistak Posted February 1, 2017 Posted February 1, 2017 Hi Guys, any of you can share Armbian_5.24.170109_Orangepizero_Ubuntu_xenial_4.9.0 build ?. History of build here https://dl.armbian.com/orangepizero/are not so old... I do need working build with full duplex and 4.9 kernel.
jerryc Posted February 1, 2017 Author Posted February 1, 2017 @szwistak I can't find it any more either. I been trying to reproduce that build done by igore, but no success, so i give up and wait until some one build another working image with 4.9.4 or something like it ...
bigbuba Posted February 22, 2017 Posted February 22, 2017 NanoPi Neo have same problem with kernel 4.9+...4.10+. It's problem in sun8i-emac - driver. After ifconfig eth0 down && ifconfig eth0 up autonegatiation work fine. I think, that problem is garbage in PHY-state register after U-boot.Problem can be resolved with adding:pre-up /sbin/ifconfig eth0 downpre-up /sbin/ifconfig eth0 uppre-up /sbin/ifconfig eth0 downin section eth0. 1
kosinske Posted March 10, 2017 Posted March 10, 2017 On 2/22/2017 at 4:01 PM, bigbuba said: After ifconfig eth0 down && ifconfig eth0 up autonegatiation work fine. That fixes the problem on OPi PC with 4.10.1 kernel (Armbian 5.27.170309 nightly) for me too.
hpapagaj Posted March 10, 2017 Posted March 10, 2017 Same problem here, anyone with a solution? uname -a Linux orangepizero 3.4.113-sun8i #10 SMP PREEMPT Thu Feb 23 19:55:00 CET 2017 armv7l armv7l armv7l GNU/Linux Stuck at 10/Full and only Orange led is lit on the ethernet port. 18.282122] PHY: gmac0-0:00 - Link is Up - 100/Full [ 26.280234] PHY: gmac0-0:00 - Link is Down [ 28.280327] PHY: gmac0-0:00 - Link is Up - 100/Full [ 34.281213] PHY: gmac0-0:00 - Link is Down [ 36.280262] PHY: gmac0-0:00 - Link is Up - 100/Full [ 40.280224] PHY: gmac0-0:00 - Link is Down [ 41.280426] PHY: gmac0-0:00 - Link is Up - 100/Full [ 43.280368] PHY: gmac0-0:00 - Link is Down [ 44.280429] PHY: gmac0-0:00 - Link is Up - 100/Full [ 46.280335] PHY: gmac0-0:00 - Link is Down [ 47.280425] PHY: gmac0-0:00 - Link is Up - 100/Full [ 53.280416] PHY: gmac0-0:00 - Link is Down [ 54.280439] PHY: gmac0-0:00 - Link is Up - 100/Full [ 56.280425] PHY: gmac0-0:00 - Link is Down [ 57.280474] PHY: gmac0-0:00 - Link is Up - 100/Full [ 63.280415] PHY: gmac0-0:00 - Link is Down [ 66.280470] PHY: gmac0-0:00 - Link is Up - 100/Full [ 80.280448] PHY: gmac0-0:00 - Link is Down [ 87.280431] PHY: gmac0-0:00 - Link is Up - 10/Full If I try /sbin/ifconfig eth0 down /sbin/ifconfig eth0 up, then: [ 618.485543] gmac0 gmac0: eth0: eth0: PHY ID 00441400 at 0 IRQ poll (gmac0-0:00) [ 618.499102] ADDRCONF(NETDEV_UP): eth0: link is not ready [ 624.480374] PHY: gmac0-0:00 - Link is Up - 100/Full [ 624.481017] ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready [ 634.480360] PHY: gmac0-0:00 - Link is Down [ 640.480347] PHY: gmac0-0:00 - Link is Up - 100/Full [ 642.480401] PHY: gmac0-0:00 - Link is Down [ 647.480745] PHY: gmac0-0:00 - Link is Up - 10/Full
Igor Posted March 11, 2017 Posted March 11, 2017 This topic is about (known) issues with Ethernet driver on mainline kernel, which is still in development phase and this driver is about to be rewritten AFAIK. On kernel 3.4.113 we didn't notice such problems. Double check your cables and other equipment first.
hpapagaj Posted March 12, 2017 Posted March 12, 2017 I upgraded to nightly beta. I added to /etc/network/interfaces: auto eth0 allow-hotplug eth0 iface eth0 inet manual pre-up /sbin/ifconfig eth0 down pre-up /sbin/ifconfig eth0 up pre-up /sbin/ifconfig eth0 down With this extra "pre-up" config I get full duplex connection (without only half duplex). A: Macbook, Samsung TV: 100Mb/s full duplex B: Old laptop Asus F5R, it is equipped with 10/100Lan card, but for some reason it is advertising itself 10Mb/s only (Live Ubuntu + Acronis True Image CD). Doesn't matter. I use this little board for wlan0 to eth0 bridging and it's really good. 2
Recommended Posts