>-----Original Message----- >From: Sylvain Munaut [mailto:s.mun...@whatever-company.com] >Sent: Monday, January 16, 2017 5:48 AM >To: e1000-devel@lists.sourceforge.net >Subject: [E1000-devel] X550EM_X_SFP "lockup" / link issues > >Hi, > > >I'm having trouble with a X550 card getting "locked up", in a state >where I can't get a link no matter what I do. Even rebooting doesn't >help, it needs to be completely unplugged from AC mains for it to be >restored. > >The hardware is a SuperMicro SYS-5018D-FN8T ( >https://www.supermicro.nl/products/system/1U/5018/SYS-5018D-FN8T.cfm ) >It includes a Intel® Xeon® processor D-1518 with embedded 10G Intel >NICs and cs4227 phys to SFP+ ports. > >I tried both ixgbe driver in the 4.4.32 kernel as well as the ixgbe >4.5.4 driver from sf.net. > >I have not traced exactly why or what triggers the lockup. It seems >more reproductible on the 4.4.32 kernel ixgbe driver where I can just >flap the link on the switch side while the interface is 'DOWN' in >linux to trigger it. > >But what I did trace is how to fix it. It seems that what happens is >the CS4227 PHY gets into a weird state and because the driver doesn't >reset the PHY every time, that state persists. The driver writes a >magic value in a scratchpad register on the PHY when it has reset it >once. But rebooting doesn't clear that value, only unplugging the AC >cable does ... > >I don't have the CS4227 phy datasheet and my attempts to get it from >InPhi (the manufacturer) have been ignored so I can't really debug why >/ what state the PHY gets in that prevents it from working, but I >thought I would at least point out the logic issue of not forcefully >resetting the PHY at least when loading the driver. > >Currently to work around the issue, I just write '0' in that scratch >pad register in ixgbe_identify_phy_x550em but that's obviously not a >good solution because it resets it too often and takes quite a while >to do so ... > >A better fix would be welcome.
I forwarded your report to our Validation and we will try to reproduce this issue in our labs. You may want to consider opening a bug on e1000.sf.net which will make it easier for us to track. Also your description of the issue is very vague and lacking in detail which can make it harder for us to reproduce. For example, information such as: 1. Type/model of the switch (link partner). 2. SFP+ module and/or DA cable used in the setup. 3. What is the offset of the PHY register you are referring to? If you open a ticket on e1000.sf.net make sure to include that information, and any other details that will allow us to get a better idea of your setup. Thanks, Emil ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
