It looks I have this problem with Ubuntu Raring
kernel version : 3.8.13
I attached a screenshot with dmesg after I tried to start again my
containers.
Should I open a new bug ?
Thank you in advance
** Attachment added: screenshot
Raring is EOL since January 27:
https://wiki.ubuntu.com/Releases
Excerpts from Alessandro Moscatelli's message of 2014-06-12 22:23:52 UTC:
It looks I have this problem with Ubuntu Raring
kernel version : 3.8.13
I attached a screenshot with dmesg after I tried to start again my
containers.
This bug was fixed in the package linux - 3.2.0-63.95
---
linux (3.2.0-63.95) precise; urgency=low
[ Kamal Mostafa ]
* Revert rtlwifi: Set the link state
- LP: #1319735
* Release Tracking Bug
- re-used previous tracking bug
linux (3.2.0-63.94) precise; urgency=low
This bug was fixed in the package linux - 3.2.0-63.95
---
linux (3.2.0-63.95) precise; urgency=low
[ Kamal Mostafa ]
* Revert rtlwifi: Set the link state
- LP: #1319735
* Release Tracking Bug
- re-used previous tracking bug
linux (3.2.0-63.94) precise; urgency=low
This bug is awaiting verification that the kernel in -proposed solves
the problem. Please test the kernel and update this bug with the
results. If the problem is solved, change the tag 'verification-needed-
precise' to 'verification-done-precise'.
If verification is not done by 5 working days
Verified on 3.2.0-63-virtual.
** Tags removed: verification-needed-precise
** Tags added: verification-done-precise
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when
** Description changed:
+ SRU Justification:
+
+ Impact:
+ When creating new network namespace dmesg can show the following :
unregister_netdevice: waiting for lo to become free. Usage count = 1
+
+ Fix:
+ Stefan Bader's SAUCE patch has fixed this for Quantal: UBUNTU: SAUCE:
** Changed in: linux (Ubuntu Precise)
Status: In Progress = Fix Committed
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg
** Also affects: linux (Ubuntu Precise)
Importance: Undecided
Status: New
** Changed in: linux (Ubuntu Precise)
Assignee: (unassigned) = Chris J Arges (arges)
** Changed in: linux (Ubuntu Precise)
Importance: Undecided = Medium
** Changed in: linux (Ubuntu Precise)
I guess, yes: the other bug causing refcount leaks: bug #1065434.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg show
I'm running 3.5.0-19-generic (3.5.0-19.30) and I still get this bug.
Am I missing something?
zoolook@venkman:~$ apt-cache policy linux-image-extra-3.5.0-19-generic
linux-image-extra-3.5.0-19-generic:
Instalados: 3.5.0-19.30
Candidato: 3.5.0-19.30
Tabla de versión:
*** 3.5.0-19.30 0
** Branch linked: lp:~ubuntu-branches/ubuntu/precise/linux-lts-quantal
/precise-proposed
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network
This bug was fixed in the package linux - 3.5.0-17.28
---
linux (3.5.0-17.28) quantal-proposed; urgency=low
[ Andy Whitcroft ]
* [packaging] we already have a valid src_pkg_name
* [packaging] allow us to select which builds have uefi signed versions
[ James M Leddy ]
*
** Tags removed: kernel-key
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg show
unregister_netdevice: waiting for lo to
** Branch linked: lp:ubuntu/quantal-proposed/linux-lowlatency
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg show
** Branch linked: lp:ubuntu/quantal-proposed/linux-ti-omap4
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg show
** Changed in: linux (Ubuntu Quantal)
Status: Triaged = Fix Committed
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg
Ok, it is probably obvious that we went ahead and applied at least the
patch to the first half. Clint, probably we/you should open a second
report for the remaining issue to have things cleanly separated.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is
ok I can reproduce with just this shell script. The weird thing is, it
only reproduces if I watch the log and kill the container after it fully
boots.
cloud init will report that it is done booting like this:
cloud-init boot finished at Fri, 05 Oct 2012 20:15:32 +. Up 200.15
seconds
if you
Ok, I think I'm ready to say that there are really just two bugs. One is
about the route cache, and is addressed by smb's most recent patch. The
other one seems only to affect my macbookair. I have not been able to
get the repro.sh script to reproduce the problem on any other machines
with that
Further information, this appears to be related to the wl proprietary
drivers.
When I tried on my MacBookPro with wired network, the problem did not
surface, but upon switching to wireless, the problem did surface.
This suggests that the real problem lies somewhere in the wl driver.
--
You
Clint, could you please write down exactly which steps you do to
reproduce the issue that leads to the single reference remaining? Which
arguments to lxc-create/-start/-stop? Any special network setup?
I think that with the last patch I made we fixed one issue, though that
(while easier to
Attaching syslog for latest fail.
BTW, the way I'm reproducing this is with a branch of juju that I've
been working on:
You'll need this in ~/.juju/environments.yaml:
local:
type: local
control-bucket: puppies-kittens-goblins
admin-secret: abcdefghijklmnop0987654321
data-dir:
The full environments.yaml, btw, should be
environments:
local:
type: local
control-bucket: puppies-kittens-goblins
admin-secret: abcdefghijklmnop0987654321
data-dir: /tmp/juju-data
default-series: precise
juju-origin: lp:~clint-fewbar/juju/local-cloud-img
--
You
Certainly something is very wrong with the numbers. Looking at the numbers
reported we get 5/13/-17/0 which sums up to 1 which is exactly the refcount
complained about. But looking at the individual numbers, the counter for CPU#0
starts with 5 not 0.
And just adding up the dev_hold and dev_put
Another step towards solving this. After it turned out in the upstream
discussions that actually the route cache should be flushed when
unregistering by doing the NETDEV_UNREGISTER_BATCH notify call, I could
check to find out that this fails because the notify handler checks for
dereferenced
** Changed in: linux
Status: Unknown = Confirmed
** Changed in: linux
Importance: Unknown = High
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating
Here is a syslog of the latest failure with the #26+smb1 kernel
showing refcount only stuck at 1, instead of 2
** Attachment added: syslog-with-#26-smb1.txt.gz
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1021471/+attachment/3372792/+files/syslog-with-%2326-smb1.txt.gz
--
You
** Changed in: linux (Ubuntu Quantal)
Milestone: ubuntu-12.10-beta-2 = ubuntu-12.10
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace
As testing with lxc containers showed, the clever idea does not work
there because that is a case where there is at least one of the
addresses (actually it seems both) is set. So it more and more looks
like the real problem is that whenever the namespace is to be torn down,
there is nothing
It seems that after about 5 minutes the references really get cleaned up
without any change (so the question would be why that takes so long...
[ 44.099279] lo(88003cc3c000)[0]+= 1
[ 44.099337] lo(88003cc3c000)[0]+= 2
[ 44.099344] lo(88003cc3c000)[0]+= 3
[ 44.099358]
This second attempt goes the path of forcing the route cache to be
cleaned of entries belonging to a net device that is torn down.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
When looking at the patch uploaded I realized something went wrong on
the update. Re-attaching.
** Patch added: 0001-UBUNTU-SAUCE-net-flush-rt-cache-on-unregister.patch
This is an experimental change that at least avoids the problem when running
the test case. But while I saw no immediate problem, it isn't guaranteed to
have no side effects and neither I can be sure there is not another case (only
either source address or destination address not set) that
** Tags added: patch
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace (dmesg show
unregister_netdevice: waiting for lo to become free.
I guess I am at the limits of my knowledge. So far it looks like
ip_connect_route initially calls __ip_route_output_key to fill in source and
destination address. And this seems to cause a routing cache entry to be
created with source and destination address 0.
It could be wrong to cache that
Reassigning to Stefan.
** Changed in: linux (Ubuntu Quantal)
Assignee: Bryan Wu (cooloney) = Stefan Bader (stefan-bader-canonical)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
** Also affects: linux via
http://bugzilla.kernel.org/show_bug.cgi?id=47181
Importance: Unknown
Status: Unknown
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone()
Repeating the comment in the upstream bug I just made:
I added debugging to dev_hold and dev_put as Eric suggested and used the
reproducer attached to this bug. What I saw was creation and destruction
would be balanced. However on the connect call, there were another two
dev_hold() calls that
I run into this bug daily, it severely cripples the juju local provider
on quantal.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1021471
Title:
clone() hang when creating new network namespace
From Eric's reply, I found this issue is a little bit complex to
backport some patches from 3.6-rc1, because from 3.5 to 3.6-rc1 some
fundamental stuffs were changed.
--
As I recall the routing cache was removed between 3.5 and 3.6-rc1 so there are
some significant changes to the fundamentals.
** Also affects: linux (Ubuntu Quantal)
Importance: High
Assignee: Bryan Wu (cooloney)
Status: Triaged
** Changed in: linux (Ubuntu Quantal)
Milestone: None = ubuntu-12.10-beta-2
** Tags removed: rls-q-incoming
--
You received this bug notification because you are a member
As Eric W. Biederman said in the bugzilla, 3.6-rc1 mainline version
works. I've testing our Ubuntu mainline build like 3.6-rc1 and 3.6-rc5,
which all work fine. But for 3.5.3 mainline build, this test failed. Our
latest Quantal kernel is based on 3.5.3 kernel.
So obviously this issue was fixed
I filed an upstream kernel bug:
https://bugzilla.kernel.org/show_bug.cgi?id=47181
** Bug watch added: Linux Kernel Bug Tracker #47181
http://bugzilla.kernel.org/show_bug.cgi?id=47181
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to
I update the bug title to better match what we're seeing.
The oops from the description is indeed just a timeout from some user space
task that's stuck on clone().
So it looks like there's something wrong either in the cleanup code when
flushing a network namespace (when the last process in
45 matches
Mail list logo
