Mailinglist Archive: opensuse-bugs (5295 mails)

< Previous Next >
[Bug 849751] 13.1/post-rc2, x64 (via dup) does not see pcie dual-nic (realtek based with plx pcie bridge) on certain pcie ports?

https://bugzilla.novell.com/show_bug.cgi?id=849751

https://bugzilla.novell.com/show_bug.cgi?id=849751#c8


--- Comment #8 from andreas bittner <abittner@xxxxxxxxxxx> 2013-11-19 21:14:43
UTC ---
Hello there bugticket,

meanwhile I needed to downgrade this machine to OpenSuSE 12.3/x64 otherwise it
wouldnt behave well (another bugreport), but I wanted to report back that this
bugreport here isnt the end of the story.

I also checked with the 13.1/RC2/x64 usb-key just to make sure that my text now
also is valid for this bug:

even these two different dual-ethernet PCIe cards with this different PCIe
switch/bridge from Pericom Semiconductor now have a problem of actually getting
hardly any bytes at all through the the card/port.

Situation: Booting up both 12.3/x64 (latest patches, but also rescue from usb
key, no-patches) and 13.1/rc2/x64 rescue from usb key are not really able to
use these Realtek ports of these two dual-port PCIe cards plugged-in.

I see arp packets all right, and ip neigh displays output, but there is never
an ICMP ping going through or being answered from the other side.

In dmesg and /var/log/messages I see the link of these ports suddenly going up
and down on its own, although they are all the time connected to
gigabitethernet switches, and then the kernel outputs a block of messages,
about NETDEV watchdog... R8169 ethX or enpXsY transmit qeue timed out.....

:((
Eventually I found some of these kind of reports on the web, and some redhat
related and others were hinting, that they only had these troubles during a
cold-boot of their machine (crazy) and others reported that ACPI=off as kernel
boot parameter would make the ethernet (realtek) card work again.

And I just tried, also for rescue bootup, ACPI=off does work magic. The whole
system gets a bit more sluggish and stalling (gfx output), but the ethernet
card is now happily answering and doing pings and actually transferring payload
data, hooooray :(

Sigh. Does this part of the bug now qualify for a different new bug, about
Realtek gigabitethernet drivers in current (13.1 is seemingly also effected the
same way, can later try with goldmaster iso image too). Do I need to report it
to OpenSuSE or some kernel.org places or what now? I think this machine can
live with 12.3 and ACPI=off for now, but I wonder when the next incident
relating to this will occur.

Maybe related and references:
Redhat:
<https://bugzilla.redhat.com/show_bug.cgi?id=1023654>

Linux Kernel mailing list from Jan/Feb 2013:
<http://linux-kernel.2935.n7.nabble.com/3-8-rc-regression-NETDEV-WATCHDOG-eth0-r8169-transmit-queue-0-timed-out-td576809.html>

<http://marc.info/?t=135737660100003&r=1&w=4>

Any information if this was ever fixed if this is even related to my latest bug
in here?
Thanks for helping.

--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

< Previous Next >
References