On 8/10/16 18:01 , Robert Mustacchi wrote: > On 8/8/16 12:20 , Ian Collins wrote: >> On 08/ 9/16 01:41 AM, Robert Mustacchi wrote: >>> On 8/8/16 2:33 , Ian Collins wrote: >>>> On 08/ 8/16 05:29 AM, Robert Mustacchi wrote: >>>>> On 8/6/16 13:21 , Ian Collins wrote: >>>>>> On 7/08/16 2:46 am, Robert Mustacchi wrote: >>>>>> >>>>>>> Hmm, okay. So what is the output of dladm show-linkprop on that >>>>>>> device? >>>>>>> Also, can you grab the prtconf -d /dev/ixgbe<num> output for it? >>>>>> ixgbe0 adv_10gfdx_cap r- 1 1 1,0 >>>>>> ixgbe0 en_10gfdx_cap rw 1 1 1,0 >>>>> Okay, so importantly we think we're advertising 10g. I have a feeling >>>>> that this part may have a similar bug as the X550 did in the driver. >>>>> Here's a way to test this. Run these two commands, waiting a second or >>>>> two inbetween and then check the new advertised speed: >>>>> >>>>> dladm set-linkprop -p en_10gfdx_cap=0 ixgbe0 >>>>> >>>>> dladm set-linkprop -p en_10gfdx_cap=1 ixgbe0 >>>>> >>>>> I don't quite have a system in front of me, so not entirely sure that's >>>>> the right command. I'd expect each of those to cause the link to >>>>> reconfigure and hopefully we'll see a 10g speed after the second one. >>>> Indeed it does - I could see the indicators on the switch go green. >>>> >>>> # dladm show-linkprop -p speed ixgbe0 >>>> LINK PROPERTY PERM VALUE DEFAULT POSSIBLE >>>> ixgbe0 speed r- 10000 10000 -- >>>> >>> Okay, can I get the prtconf -d /dev/ixgbe0 output from that system? >> >> pci15d9,1528 (pciex8086,1528) [Intel Corporation Ethernet Controller >> 10-Gigabit X540-AT2], instance #0 >> >> For reference, the PCIe card: >> >> pci8086,1 (pciex8086,1528) [Intel Corporation Ethernet Controller >> 10-Gigabit X540-AT2], instance #0 > > Okay, thanks for providing that Ian. I'm trying to think through what > might cause this case and how we might better handle this. I'm reviewing > the PRM but will probably need some time on this one. > > One question I have is on this particular box are those on-board ports > shared with the BMC? And are you using the dedicated BMC ports? I'm just > wondering if, like the issue Dale mentioned with the X550, because we > have a shared management port, the link was up and we never really reset > it properly in the driver to run at full speed vs. it having been > programmed for other, more limited speeds by the BMC.
Aha, the X540 has LPLU (Low Power Link Up) mode. So, I bet that we have this in use on your X540 that's onboard, but not on the PCIe card. Ian, if I have some time over the next week I can put together a custom platform to test and verify that this is the case and with a potential fix. Are these machines in a place where using a test platform would be acceptable? Thanks, Robert ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
