Bug 806170

Summary: ipv6 regression, brief network connectivity loss
Product: [Fedora] Fedora Reporter: Michael Cronenworth <mike>
Component: NetworkManagerAssignee: Dan Williams <dcbw>
Status: CLOSED DUPLICATE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 16CC: danw, dcbw, frank, gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda, nhorman, swsnyder
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2012-04-19 21:18:56 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Michael Cronenworth 2012-03-23 05:32:53 UTC
Description of problem:
About every 5-10 minutes, NetworkManager informs me my network connection has been lost. The network connection is lost for about 2 or 3 seconds and then comes back. This happened after upgrading from 3.2.10-3 to 3.3.0-4 today.


Version-Release number of selected component (if applicable):
kernel-3.3.0-4.fc16.x86_64


How reproducible: Every 5-10 minutes.


Steps to Reproduce:
Let the system sit idle. NetworkManager notification on desktop about lost connection. Network packets don't transmit for a few seconds.
  
Actual results:
Network connectivity loss for a few seconds. The only kernel messages I have in my logs are:
[  499.159264] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 1164.973572] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 1620.701254] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 2384.246151] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 2849.338299] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 3544.022662] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 3951.961402] ICMPv6 RA: ndisc_router_discovery() failed to add default route.
[ 4504.160824] ICMPv6 RA: ndisc_router_discovery() failed to add default route.


Expected results:
No connectivity loss.

Additional info:
02:00.0 Ethernet controller: Marvell Technology Group Ltd. 88E8056 PCI-E Gigabit Ethernet Controller (rev 12)
	Subsystem: ASUSTeK Computer Inc. Device 81f8
	Flags: bus master, fast devsel, latency 0, IRQ 45
	Memory at febfc000 (64-bit, non-prefetchable) [size=16K]
	I/O ports at d800 [size=256]
	Expansion ROM at febc0000 [disabled] [size=128K]
	Capabilities: [48] Power Management version 3
	Capabilities: [50] Vital Product Data
	Capabilities: [5c] MSI: Enable+ Count=1/1 Maskable- 64bit+
	Capabilities: [e0] Express Legacy Endpoint, MSI 00
	Capabilities: [100] Advanced Error Reporting
	Kernel driver in use: sky2
	Kernel modules: sky2

Comment 1 Michael Cronenworth 2012-03-23 06:20:21 UTC
After upgrading a laptop to 3.3.0 I noticed network connectivity drops on it, too, but the laptop is wireless only. Same message in my kernel log when the connectivity drops.

[  205.382142] ICMPv6 RA: ndisc_router_discovery() failed to add default route.

I have radvd running on my network. IPv6 is working even with the intermittent loss.

Comment 2 Steve Snyder 2012-03-27 13:55:01 UTC
I've seen this also since the update to the v3.3.0 kernel.  Also on a notebook with (an unused) wireless interface, but it restarts the wired interface too.

I had to stop radvd (v1.8.5) on my network server to stop the bouncing on the wired interface.  So now I once again have a stable network connection, but now no longer have IPv6 router advertisement.

Comment 3 Michael Cronenworth 2012-03-27 14:01:36 UTC
@Steve, I have found an odd workaround. If I set NetworkManager's IPv6 setting from "Automatic" to "Ignore" the network connection stops dropping. I get to keep my IPv6 address and routes from radvd and both IPv4 and IPv6 networks are stable.

Comment 4 Neil Horman 2012-03-27 14:04:20 UTC
Try setting the ipv6 address setting to ignore or disable for your wireless interface. That should allow you to sanely use SLAAC again.  You can also use the rawhide NetworkManager for this.  I've built it for f16 here:
http://koji.fedoraproject.org/koji/taskinfo?taskID=3936339

Comment 5 Steve Snyder 2012-03-27 14:39:39 UTC
@Michael -

Where is this setting kept?  I looked in NetworkManager.conf but didn't see anything that looked relevant to IPv6.

Comment 6 Michael Cronenworth 2012-03-27 14:54:05 UTC
@Steve, the setting is in the GUI in the options for the connection. The "IPv6 Settings" tab.

@Neil, I will try that NM package and get back to you.

Comment 7 Michael Cronenworth 2012-03-30 05:13:25 UTC
(In reply to comment #4)
> Try setting the ipv6 address setting to ignore or disable for your wireless
> interface. That should allow you to sanely use SLAAC again.  You can also use
> the rawhide NetworkManager for this.  I've built it for f16 here:
> http://koji.fedoraproject.org/koji/taskinfo?taskID=3936339

I have run this package for a few hours with IPv6 set to Automatic. No network drops have occurred, but I am seeing many of these messages instead:
...
[19933.206809] sky2 0000:02:00.0: p37p1: rx error, status 0x69a0010 length 1518
[20801.653083] sky2 0000:02:00.0: p37p1: rx error, status 0xf790010 length 1518
[21701.039211] sky2 0000:02:00.0: p37p1: rx error, status 0x6050010 length 1518
...

Comment 8 Neil Horman 2012-03-30 12:38:55 UTC
I'm pretty sure thats a separate bug.  First guess would suggest that perhaps you're receiving frames longer than the interfaces configured MTU.  

How do you want to handle the NM issue?  Is it sufficient to set IPv6 to automatic or ignore?  Or shall I assign this over to the NM component to backport the rawhide version to f16?

Comment 9 Michael Cronenworth 2012-03-30 13:16:42 UTC
I think it is a bug that I had to set the F16 NM to Ignore when Automatic was working (and Automatic works again with the rawhide NM) so it would be nice to have the rawhide version in F16.

I will also mention that I saw many SELinux denials with the rawhide NM. It may not be as easy as rebuilding the rawhide package and pushing an update.

Comment 10 Michael Cronenworth 2012-03-30 14:23:59 UTC
I just saw there is a pending NM update.

https://admin.fedoraproject.org/updates/NetworkManager-0.9.4-1.git20120328.fc16

I'll give it a shot.

Comment 11 Michael Cronenworth 2012-03-31 18:22:13 UTC
(In reply to comment #10)
> https://admin.fedoraproject.org/updates/NetworkManager-0.9.4-1.git20120328.fc16

This update causes network drops with IPv6 set to Automatic so it appears the "real" F16 build uses different code? I had to set IPv6 back to Ignore.

Comment 12 Frank Crawford 2012-04-06 05:04:47 UTC
I've just downloaded and installed the pending NM update:

https://admin.fedoraproject.org/updates/NetworkManager-0.9.4-2.git20120403.fc16

and unfortunately I'm still seeing the same issues as listed above.

The only way I've found to run a working system with IPv6 is to stay with a 3.2 kernel.

Comment 13 Dan Winship 2012-04-19 21:18:56 UTC

*** This bug has been marked as a duplicate of bug 785772 ***