Message ID | 20200709195402.1835538-1-cascardo@canonical.com |
---|---|
Headers | show |
Series | Revert code that causes crashes on default configurations | expand |
Acked-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com> On Thu, Jul 09, 2020 at 04:54:00PM -0300, Thadeu Lima de Souza Cascardo wrote: > BugLink: https://bugs.launchpad.net/bugs/1886668 > > This is the same patch as sent to Bionic 4.15 kernels. > > [Impact] > On systems using cgroups and sockets extensively, like docker, kubernetes, > lxd, libvirt, a crash might happen when using linux 4.15.0-109-generic. > > [Fix] > Revert the patch that disables sk_alloc cgroup refcounting when tasks > are added to net_prio cgroup. > > [Test case] > Test that such environments where the issue is reproduced survive some hours of > uptime. A different bug was reproduced with a work-in-progress code and was not > reproduced with the culprit reverted. > > [Regression potential] > The reverted commit fix a memory leak on similar scenarios. But a leak is > better than a crash. Two other bugs have been opened to track a real fix for > this issue and the leak. > > Thadeu Lima de Souza Cascardo (1): > UBUNTU: SAUCE: Revert "netprio_cgroup: Fix unlimited memory leak of v2 > cgroups" > > net/core/netprio_cgroup.c | 2 -- > 1 file changed, 2 deletions(-) > > -- > 2.25.1 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team Acked-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Acked-by: Ian May <ian.may@canonical.com> On 7/9/20 2:54 PM, Thadeu Lima de Souza Cascardo wrote: > BugLink: https://bugs.launchpad.net/bugs/1886668 > > This is the same patch as sent to Bionic 4.15 kernels. > > [Impact] > On systems using cgroups and sockets extensively, like docker, kubernetes, > lxd, libvirt, a crash might happen when using linux 4.15.0-109-generic. > > [Fix] > Revert the patch that disables sk_alloc cgroup refcounting when tasks > are added to net_prio cgroup. > > [Test case] > Test that such environments where the issue is reproduced survive some hours of > uptime. A different bug was reproduced with a work-in-progress code and was not > reproduced with the culprit reverted. > > [Regression potential] > The reverted commit fix a memory leak on similar scenarios. But a leak is > better than a crash. Two other bugs have been opened to track a real fix for > this issue and the leak. > > Thadeu Lima de Souza Cascardo (1): > UBUNTU: SAUCE: Revert "netprio_cgroup: Fix unlimited memory leak of v2 > cgroups" > > net/core/netprio_cgroup.c | 2 -- > 1 file changed, 2 deletions(-) >
On 2020-07-09 16:54:00 , Thadeu Lima de Souza Cascardo wrote: > BugLink: https://bugs.launchpad.net/bugs/1886668 > > This is the same patch as sent to Bionic 4.15 kernels. > > [Impact] > On systems using cgroups and sockets extensively, like docker, kubernetes, > lxd, libvirt, a crash might happen when using linux 4.15.0-109-generic. > > [Fix] > Revert the patch that disables sk_alloc cgroup refcounting when tasks > are added to net_prio cgroup. > > [Test case] > Test that such environments where the issue is reproduced survive some hours of > uptime. A different bug was reproduced with a work-in-progress code and was not > reproduced with the culprit reverted. > > [Regression potential] > The reverted commit fix a memory leak on similar scenarios. But a leak is > better than a crash. Two other bugs have been opened to track a real fix for > this issue and the leak. > > Thadeu Lima de Souza Cascardo (1): > UBUNTU: SAUCE: Revert "netprio_cgroup: Fix unlimited memory leak of v2 > cgroups" > > net/core/netprio_cgroup.c | 2 -- > 1 file changed, 2 deletions(-) > > -- > 2.25.1 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team
On 10.07.20 01:41, Khaled Elmously wrote: > On 2020-07-09 16:54:00 , Thadeu Lima de Souza Cascardo wrote: >> BugLink: https://bugs.launchpad.net/bugs/1886668 >> >> This is the same patch as sent to Bionic 4.15 kernels. >> >> [Impact] >> On systems using cgroups and sockets extensively, like docker, kubernetes, >> lxd, libvirt, a crash might happen when using linux 4.15.0-109-generic. >> >> [Fix] >> Revert the patch that disables sk_alloc cgroup refcounting when tasks >> are added to net_prio cgroup. >> Just FYI: The upstream "real fix" for this is [0] plus a followup [1]. I originally wanted to sent it out to the ubuntu kernel list, but got overloaded with other things and totally forgot about it, sorry! Figured out that sharing this late is better than never when seeing the new Ubuntu tag just now. [0]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=ad0f75e5f57ccbceec13274e1e242f2b5a6397ed [1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=14b032b8f8fce03a546dcf365454bec8c4a58d7d >> [Test case] >> Test that such environments where the issue is reproduced survive some hours of >> uptime. A different bug was reproduced with a work-in-progress code and was not >> reproduced with the culprit reverted. There's a more direct proposed repoducer [2]. [2]: https://lore.kernel.org/netdev/42baf0f8-627b-0ab8-72fc-12d24667ad0a@huawei.com/ cheers, Thomas
On Thu, Jul 09, 2020 at 04:54:00PM -0300, Thadeu Lima de Souza Cascardo wrote: > BugLink: https://bugs.launchpad.net/bugs/1886668 > > This is the same patch as sent to Bionic 4.15 kernels. > > [Impact] > On systems using cgroups and sockets extensively, like docker, kubernetes, > lxd, libvirt, a crash might happen when using linux 4.15.0-109-generic. > > [Fix] > Revert the patch that disables sk_alloc cgroup refcounting when tasks > are added to net_prio cgroup. > > [Test case] > Test that such environments where the issue is reproduced survive some hours of > uptime. A different bug was reproduced with a work-in-progress code and was not > reproduced with the culprit reverted. > > [Regression potential] > The reverted commit fix a memory leak on similar scenarios. But a leak is > better than a crash. Two other bugs have been opened to track a real fix for > this issue and the leak. Applied to unstable, thanks!