Custom Hardened Kernels & Hardware Studio
Axonwall ships a reproducible Linux 6.12 LTS kernel pipeline designed specifically for headless k3s / containerd, OpenZFS + OpenEBS storage, nftables NGFW packet processing, and eBPF CO-RE observability—paired with an interactive Custom Kernel Studio (/system/kernel) that detects host hardware and compiles minimal localmodconfig kernels.
1. Hardened Container & Storage Kernel Comparison
General-purpose desktop hardening patches (such as linux-hardened) frequently break cloud-native workloads by disabling unprivileged user namespaces, blocking eBPF bpf_probe_read_kernel, or stripping kernel symbols required by out-of-tree OpenZFS DKMS modules. Axonwall synthesizes proven server and container-OS patterns instead:
| Reference Kernel | Strength Adopted by Axonwall | Compatibility Trade-off |
|---|---|---|
| Talos Linux | KSPP memory/slab/stack hardening, module signing, stripped unused drivers & legacy protocols. | No local shell/DKMS; Axonwall retains module signing in integrity mode + ZFS symbol support. |
| Fedora CoreOS / Flatcar | Upstream LTS paired with cgroups v2, SELinux/LSM enforcement, overlayfs & wireguard out of the box. | Full distro footprint; Axonwall trims desktop media/sound/wireless bloat. |
Alpine linux-virt | Ultra-small surface for virtualized & bare-metal headless hosts. | Misses broad enterprise RAID/HBA/25GbE NIC drivers; Axonwall offers both Universal & Stripped tiers. |
| Debian 13 LTS + LSM | Broad server/cloud driver compatibility and clean .deb / ZFSBootMenu packaging. | Desktop subsystems enabled by default; Axonwall strips them while layering KSPP & BPF-LSM. |
Desktop linux-hardened | Strict heap/JIT/namespace restrictions for single-user workstations. | Breaks rootless userns, eBPF tracers, overlayfs, and OpenZFS DKMS. |
CONFIG_SECURITY_LOCKDOWN_LSM=y with CONFIG_LOCK_DOWN_KERNEL_FORCE_INTEGRITY=y instead of confidentiality. Integrity mode protects /dev/mem, kexec, and unsigned kernel module loads while preserving DEBUG_INFO_BTF, fprobe, and eBPF CO-RE telemetry needed by Cilium/Tetragon/Axonwall eBPF agents.2. Axonwall Kernel Best Practices Matrix
CONFIG_SECURITY_LOCKDOWN_LSM=y(integrity)CONFIG_BPF_LSM=y&CONFIG_SECURITY_SELINUX=yINIT_ON_ALLOC_DEFAULT_ON,ZERO_CALL_USED_REGS,SLAB_FREELIST_HARDENEDCONFIG_MODULE_SIG=y(sha512)- Disables
KEXEC,HIBERNATION,DEVMEM,DCCP,RDS,TIPC
k3s / containerd)CONFIG_USER_NS=y+ sysctlkernel.unprivileged_userns_clone=1- cgroups v2 (
CGROUPS,CGROUP_BPF,MEMCG,BLK_CGROUP) CONFIG_SECCOMP_FILTER=y&CONFIG_OVERLAY_FS=mVETH,BRIDGE,VXLAN,GENEVE,IP_VS- Full CNI netfilter /
xtablesmatch & masquerade hooks
# CONFIG_TRIM_UNUSED_KSYMS is not set+CONFIG_KALLSYMS=y(required for ZFS DKMS)CONFIG_BTRFS_FS=y&CONFIG_BTRFS_FS_POSIX_ACL=y(03-storage-zfs.config, for Btrfs + Snapper Root)NVME_CORE/BLK_DEV_NVME& NVMe-oF TCP multipathDM_CRYPT,DM_THIN_PROVISIONING,MDRAIDEXT4_FS,XFS_FS,ISCSI_TCP,NFSD
nftables NGFW & eBPF ObservabilityNF_TABLES(INET,NETDEV,FLOW_OFFLOAD,SYNPROXY,QUOTA)WIREGUARD,XFRM_USER/INET_ESP,CAKE,FQ_CODEL,BBRBPF_SYSCALL=y,BPF_JIT_ALWAYS_ON=y,BPF_UNPRIV_DEFAULT_OFF=yDEBUG_INFO_BTF=y,DEBUG_INFO_BTF_MODULES=y,XDP_SOCKETS=y,FPROBE=yNET_CLS_ACT=y,NET_SCH_INGRESS=y,UPROBE_EVENTS=y,TRACEPOINTS=y,FTRACE_SYSCALLS=y(ebpf-tcx-uprobe-harvester)
3. Two Delivery Tiers: Universal vs. Hardware-Stripped
Tier 1 — Axonwall Universal Hardened Kernel
Combines 00-universal-hardware-base.config (Debian-13-equivalent bare-metal server, NVMe, HBA, 1GbE–100GbE NIC, and cloud/hypervisor drivers) with fragments 01..06. Ships cleanly across diverse appliances while stripping sound, GPU/DRM, Bluetooth, and Wi-Fi bloat.
Tier 2 — Hardware-Locked Stripped Kernel (localmodconfig)
Uses System → Custom Kernel Studio (/system/kernel) to inventory live host CPU ISA (x86-64-v2 / v3 / v4 / native), IOMMU, UEFI Secure Boot, TPM2, hardware watchdogs, PCIe NICs/HBAs/NVMe, and active lsmod modules, then runs localmodconfig while pinning essential k3s, ZFS, nftables, and eBPF symbols.
4. CLI & Offline Build Reference
# Inventory CPU flags, IOMMU, TPM2, PCIe devices, NICs, NVMe, and loaded modules axonwallctl kernel detect # Run the 25-check Axonwall KSPP + k3s + ZFS + nftables + eBPF posture audit (/boot/config-* or /proc/config.gz) axonwallctl kernel audit # Generate a hardware-locked localmodconfig bundle + hardened sysctl drop-in axonwallctl kernel config --mode stripped --output-dir ./out-kernel
# Universal hardened Debian-equivalent kernel .deb build ./packaging/kernel/build-kernel.sh --mode universal --output-dir ./out-universal # Hardware-stripped build with captured lsmod and x86-64-v3 ISA tuning lsmod > /tmp/host-lsmod.txt ./packaging/kernel/build-kernel.sh \ --mode stripped \ --lsmod /tmp/host-lsmod.txt \ --march x86-64-v3 \ --lockdown integrity \ --output-dir ./out-stripped
5. Safe Rollback & Boot Staging
Installing a stripped kernel never removes your currently running fallback kernel in place:
- ZFS Boot Menu (
zbmmode):axonwalldsnapshots the active root dataset, clones a new Boot Environment, installs the builtlinux-image-*andlinux-headers-*.debpackages inside the staged BE, and arms one-shot boot verification. If the new kernel fails to report healthy, ZFSBootMenu or the hardware watchdog falls back to the prior BE automatically. - Btrfs + Snapper (
btrfsmode):axonwalldcreates a pre-kernel Snapper snapshot on@before installing the new.debpackages withdpkg -i, then runsSyncBtrfsGrubMenu()so prior snapshots remain bootable from GRUB's Axonwall Btrfs Snapshots submenu. - Standard A/B & Rolling Installs (
ab/rolling):dpkg -iinstalls the new kernel side-by-side withupdate-grubpreserving the previous working kernel entry andgrub-rebootone-time trial boot support.