'82599ES 10-Gigabit SFI/SFP+ Network Connection' with optical SFP+ module connected to other such card with point-to-point link (no 10G switch inbetween) drop down link ("status: no carrier") after several runs of nermap's pkt-gen with receive function ("pkt-gen -f rx -i ix0 -N"). ifconfig ix0 down up doesn't help, only full system reboot helps. It is 13-CURRENT r34031
Does forcing the media type at one or both ends re-establish link when in the failed state?
(In reply to Stephen Hurd from comment #1) I'll try next time it happens. It is not 100% reproducible, though. After power cycle it worked for rest of day.
(In reply to Lev A. Serebryakov from comment #2) Can you look if is related to https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221317 ? Try ifconfig down/up several times like in this post script.
New data: (1) I can not reproduce this with pkt-gen anymore (2) I can reproduce problem with shell loop of ifconfig down / ifconfig up (As in PR221317) When it failed it says, that it could not initialize "unsupported SFP+ type". After that setting media with "ifconfig ix0 media" doesn't work — ifconfig complains about failed ioctl(). System reboot helps.
I've reproduced it again with pkt-gen, and ifconfig could not set media, error is "Device not configured".
X540 is 10GBASE-T, so if your device really is an 82599ES SFP+, then it is X520 :)
(In reply to Jeff Pieper from comment #6) Yep, my NICs are X520-DA1 and X520-DA2
(In reply to Jeff Pieper from comment #6) It looks the same as https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221317 but for 12-STABLE and 13-CURRENT, not 11-STABLE, and drivers were changed A LOT. But symptoms are exactly the same.
If I build driver as module, kldunload + kldload helps.
More data. It is r340913 (version becomes truncated in first comment). "ifconfig down && ifconfig up" loop could not reproduce this bug anymore. "pkt-gen -f rx -i ix0" works. But "pkt-gen -f tx -i ix0" kills NIC after 5 or 6 runs, and it is 100% reproducible.
kldunload could crash system, so it is not very viable workaround for automatic benchmarking :-(
Was https://reviews.freebsd.org/rS341156 intended to fix this problem? It would affect the TX side of things.
(In reply to Eric Joyner from comment #12) Unfortunately, no. This fix helps to run automated benchmarks multiple times (as pkt-gen with fixed number of packets can finish successfully now), but ix0 still drops link sometimes.
(In reply to Lev A. Serebryakov from comment #13) The bug seems to be fixed by applying patch D18468 which is currently under review ( https://reviews.freebsd.org/D18468 ). Please let me know if the patch solves your problem.
https://reviews.freebsd.org/D18468 is already committed, but the reporter didn't answer whether it fixed his issue. Closing, please reopen if the issue persists.