This is the core of the seapath project. The meta-seapath repo contains the yocto Seapath layer, which contains all the code (yocto recipes) needed to build the seapath images used to install seapath (host, guest, flasher, etc.)
With Linux 5.15 there is a regression in the max latency measured with cyclictest inside a VM.
To Reproduce
Steps to reproduce the behavior:
Run cukinia in SEAPATH guest test image using Linux 5.15
Expected behavior
In our machines we have in the previous kernel (4.19) we have a max latency ~60 µs.
With Linux 5.15 on the same machines, the max latency is over 40 ms.
Additional context
With a tracing tool, we have discovered that the CPU core running the KVM thread (the VM) enter into IDLE mode (a CPU mode where the CPU do nothing) when the latency spike appeared.
With Linux 5.15 there is a regression in the max latency measured with cyclictest inside a VM.
To Reproduce Steps to reproduce the behavior:
Expected behavior In our machines we have in the previous kernel (4.19) we have a max latency ~60 µs. With Linux 5.15 on the same machines, the max latency is over 40 ms.
Additional context With a tracing tool, we have discovered that the CPU core running the KVM thread (the VM) enter into IDLE mode (a CPU mode where the CPU do nothing) when the latency spike appeared.
We asked the problem in the KVM mailing list: https://www.spinics.net/lists/kvm/msg264449.html.