0017885: kvm guest memory corruption due to strongswan traffic - CentOS Bug Tracker

ID	Project	Category	View Status	Date Submitted	Last Update

0017885	CentOS-8	kernel	public	2020-11-25 01:13	2020-11-25 01:13

Reporter	beck	Assigned To
Priority	normal	Severity	minor	Reproducibility	always
Status	new	Resolution	open
Product Version	8.2.2004

Summary	0017885: kvm guest memory corruption due to strongswan traffic
Description	Filing this as Kernel bug, as strongswan is merely a user of kernel modules and does not bring it's own (afaik). Setup: strongswan site2site tunnel + multiple KVM (CentOS 7) guests Issue: When there is "heavy" (>1MBit/s) traffic on the strongswan tunnel, KVM guests get memory corruptions. After a variable amount of time (seconds - minutes) of heavy traffic, programs in the guests start segfaulting. This is random, sometimes execution works perfectly, the next execution is segfaulting. This is affecting all kinds of services (journald, user space programs, shell commands). It also affects disk io, reading data from the disk is sometimes returning corrupted data. However next read of the same data is working fine; completely random. A reboot of the guest kernel does not fix the issue. Only a "power cycle" of the guest recovers the state. Traffic bypassing the tunnel does not trigger the issue.
Steps To Reproduce	setup strongswan tunnel startup some KVM guests (it might help to have some userspace programs as well, in my case InfluxDB and PostgreSQL). Run iperf from the host through the tunnel observe crashes and segfaults in the guests
Additional Information	I'm sorry for the vague description. I can help debugging the issue if i get the right hints on what to look for. Host memory is tested and showed no issues.
Tags	No tags attached.

Date Modified	Username	Field	Change
2020-11-25 01:13	beck	New Issue