- Mar 20, 2025
-
-
Mike Hibler authored
-
Mike Hibler authored
-
Mike Hibler authored
-
- Mar 18, 2025
-
-
David Johnson authored
So, we cannot use veth.$vmid.$ifid, and we cannot use vif.$vmid.$ifid, because the latter seems to (sometimes) be used internally in some way that races with our use. Good news! veth$vmid.$ifid seems to work. This is ridiculous.
-
David Johnson authored
-
David Johnson authored
Something changed at Xen 4.16/Linux 5.15, where if the net.bridge.bridge-nf-call-iptables sysctl is enabled (which we do so that we can firewall/nat the control net bridge ifaces), iptables rules apply to the expt net bridges as well. This seems to be new behavior, because although we set the default global FORWARD chain policy to DROP, and do not create iptables rules for expt net bridges, in Xen 4.11/Linux 5.4, the bridge forwarded ip packets just fine. In Xen 4.16/Linux 5.15, this behavior has changed, and ip traffic is not forwarded unless we add this basic allow-all forwarding rule for each expt bridge.
-
David Johnson authored
Something in Xen no longer allows our prior PV vif naming convention (veth.$vmid.$ifid), and switching to theirs works (vif$vmid.$ifid); make it so. Keep backwards compat for prior Xen versions.
-
David Johnson authored
-
- Mar 16, 2025
-
-
David Johnson authored
In this commit, we actually write the proper configuration to systemd-networkd files instead of stopping systemd entirely. Moreover, we grab the nameserver configuration from systemd-resolved if the stub resolver is in use, and migrate it to the xenbr. All in the name of upstream compat/coexistence and avoiding special hacks.
-
David Johnson authored
The upstream pygrub works fine on our standard Ubuntu images as of 4.16.
-
David Johnson authored
-
- Mar 14, 2025
-
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
- Mar 13, 2025
-
-
Leigh B Stoller authored
-
Mike Hibler authored
In one situation, we were getting our own MC packets echoed back to us and we were accepting our JOIN reply messages as JOIN request messages and sending replies to those. The situation deteriorated rapidly.
-
Mike Hibler authored
-
- Mar 12, 2025
-
-
Mike Hibler authored
-
Mike Hibler authored
-
Mike Hibler authored
-
Mike Hibler authored
-
Mike Hibler authored
with 512 byte sectors.
-
- Mar 11, 2025
-
-
Leigh B Stoller authored
-
Leigh B Stoller authored
{5484} boss.utah$ manage_instance terminate -h Terminate an experiment Usage: manage_instance [-ds] [-t id] terminate <instance> [-eLX] Options: -e - Record the experiment as expired instead of terminated -L - Take the experiment lock from whoever has it -X - Only the experiment creator can terminate (legacy RPC) Global Options: -d - Debugging, do not fork into the background -s - Silent (no email) -t id - Webtask ID
-
- Mar 07, 2025
-
-
Mike Hibler authored
older Linux implementations. Who knew that FreeBSD had a newer version of EXTFS tools than Ubuntu 22?
-
- Mar 06, 2025
-
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
- Feb 25, 2025
-
-
Mike Hibler authored
Need some downtime between "mount" and "blkid" invocations, else blkid returns nothing. I suspect busybox implementations are causing grief.
-
- Feb 22, 2025
-
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-
Leigh B Stoller authored
-