Linux kernel

The Linux kernel is the core of any Linux operating system. It manages hardware, system resources, and provides the fundamental services for all other software.

Last build4

📌 Project Architecture & CI/CD Workflow

graph TD
    subgraph Job1["🔍 1. Prepare Environment (Debian Container)"]
        A1["📦 Kernel Source<br/>(kernel.org / Git Tag)"] --> A3["🔧 make olddefconfig<br/>(+ Optional Patch)"]
        A2["📄 Kconfig<br/>(configs/ or Remote)"] --> A3
        A3 --> A4["📤 Upload Source Artifact"]
    end

    subgraph Job2["🐧 2. Build Kernel & Release (Debian Container)"]
        B1["📥 Download Source Artifact"] --> B2["🐧 make bindeb-pkg<br/>(KCFLAGS='-march=znver3')"]
        B2 --> B3["📦 Debian .deb Packages<br/>(headers, image, dbg, libc-dev)"]
        B3 --> B4["🚀 GitHub Release"]
    end

    subgraph Job3["📤 3. Telegram Notification"]
        C1["📲 Send Status &<br/>Logs to Telegram"]
    end

    Job1 --> Job2
    Job2 --> Job3
    B4 -->|"💻 sudo apt install ./linux-*.deb"| D["🖥️ Workstation<br/>"]

    style Job1 fill:#1a202c,stroke:#319795,color:#fff
    style Job2 fill:#1a202c,stroke:#2b6cb0,color:#fff
    style Job3 fill:#1a202c,stroke:#805ad5,color:#fff
    style D fill:#dd6b20,stroke:#ed8936,color:#fff

⚡ Kernel Customization & Performance

Subsystem	Configuration / Option	Engineering Rationale
Native CPU Optimization	`CONFIG_X86_NATIVE_CPU=y` + `KCFLAGS="-march=znver3"`	Two-layer native optimization: `X86_NATIVE_CPU` enables Kconfig CPU feature selection based on the build host ISA; `-march=znver3` directs GCC to emit Zen 3-specific instructions (AVX2/BMI2/VAES). Both are complementary and required for full native optimization.
Scheduler & Preemption	`PREEMPT_BUILD` / `PREEMPT=y`	Full kernel preemption for minimal input and audio processing latency.
Timer Frequency	`1000 Hz` (`CONFIG_HZ_1000=y`)	High-resolution tick frequency for desktop responsiveness and frame pacing.
Tickless Kernel	`CONFIG_NO_HZ_IDLE=y`	Idle-tickless kernel: timer interrupts suppressed on idle CPUs, reducing power and wakeup overhead. `NO_HZ_FULL` is intentionally not enabled to avoid scheduler complexity.
CPU Core Sizing	`CONFIG_NR_CPUS=32`	Hardcapped to 32 threads matching Ryzen 9 5950X, eliminating virtual CPU overhead.
CPU Frequency Scaling	`amd-pstate` EPP (`CONFIG_X86_AMD_PSTATE=y`)	Hardware CPPC microsecond-level core frequency and voltage adjustments.
NUMA	`CONFIG_NUMA=y` / `# CONFIG_NUMA_BALANCING is not set`	NUMA infrastructure retained (`CONFIG_AMD_NUMA=y`, `CONFIG_X86_64_ACPI_NUMA=y`); automatic NUMA memory balancing disabled to eliminate background scan overhead on this single-node Ryzen system.
Core Scheduling	`CONFIG_SCHED_CORE=y`	SMT thread isolation and core scheduling for security and hyperthreading performance.
TCP Congestion	TCP BBR (`CONFIG_DEFAULT_TCP_CONG="bbr"`)	Bottleneck bandwidth RTT pacing algorithm to prevent bufferbloat and latency spikes.
Extensible Scheduler	`CONFIG_SCHED_CLASS_EXT=y`	In-kernel `sched-ext` BPF scheduler class. Enables the mechanism for dynamically loading external BPF schedulers (`scx_bpfland`, `scx_rusty`) at runtime — they are not embedded in the kernel. Depends on BPF infrastructure (`CONFIG_BPF=y`, `CONFIG_BPF_SYSCALL=y`).
Extended Group Sched	`CONFIG_EXT_GROUP_SCHED=y`	Extended group scheduling infrastructure used by `sched-ext` and cgroup-aware scheduling policies (`CGROUP_BPF`, `CGROUP_SCHED`).
VRAM Cgroup Management	`CONFIG_CGROUP_DMEM=y`	Device Memory cgroup controller enabling dynamic VRAM resource allocation and priority control on `amdgpu`.
eBPF JIT Compiler	`CONFIG_BPF_JIT=y` / `CONFIG_BPF_JIT_DEFAULT_ON=y`	JIT compiler enabled and active by default; `BPF_JIT_ALWAYS_ON` is intentionally not set, preserving runtime control via `net.core.bpf_jit_enable`.
eBPF Security (LSM)	`CONFIG_BPF_LSM=y`	In-kernel eBPF Linux Security Module for granular, high-performance security hooks.
eBPF BTF Type Format	`CONFIG_DEBUG_INFO_BTF=y`	Pahole split-BTF typeinfo generation for vmlinux and modules (`BTF_MODULES=y`) for eBPF tracing tools.
Async Socket I/O	`IO_URING_ZCRX=y`	Zero-copy packet reception for ultra-fast network socket I/O.
GPU Driver	`CONFIG_DRM_AMDGPU=m`	AMDGPU as loadable kernel module (not built-in). SI/CIK legacy support disabled. Native RDNA 2 support, Resizable BAR (ReBAR), and OverDrive power limit unlocking.
AMD Display Core	`CONFIG_DRM_AMD_DC=y` / `CONFIG_DRM_AMD_DC_FP=y`	AMD Display Core with fast-path support for DCN 3.0 (Navi 21 / RX 6950 XT).
AMD Secure Display	`CONFIG_DRM_AMD_SECURE_DISPLAY=y`	Secure display path support for AMD GPU.
AMD Audio Coprocessor	`CONFIG_DRM_AMD_ACP=y`	AMD Audio CoProcessor driver enabled alongside AMDGPU.
GPU ROCm Compute	`CONFIG_HSA_AMD=y` / `CONFIG_DRM_AMDGPU_USERPTR=y`	Native AMD KFD driver for ROCm, OpenCL 3.0, and direct GPU user-pointer memory access.
PlayStation Gamepads	`CONFIG_HID_PLAYSTATION=m` / `FF=y`	Sony DualSense (PS5) & DualShock 4 (PS4) controller support with haptic Force Feedback.
Legacy Radeon Removal	`# CONFIG_DRM_RADEON is not set`	Legacy Radeon DRM driver disabled to ensure exclusive `amdgpu` driver stack execution.
AMD IOMMU Isolation	`CONFIG_AMD_IOMMU=y` (`# INTEL_IOMMU`)	Native AMD Vi IOMMU enabled while stripping unused Intel DMAR overhead.
Memory / Zswap	`zswap` + `lzo` (`CONFIG_ZSWAP=y`)	In-RAM compressed swap cache matching kernel boot parameters (`zswap.compressor=lzo`).
Hugepages & Compaction	`THP MADVISE` + `COMPACTION=y`	Transparent Hugepages THP default set to `madvise` to avoid memory bloat with opt-in THP for games/VMs.
Hung Task Diagnostics	`CONFIG_DETECT_HUNG_TASK=y` (120s)	Automatic detection and logging of blocked or hanging kernel threads.
Stack Unwinder	`UNWINDER_ORC=y`	Low-overhead ORC call stack unwinding for precise ftrace kernel profiling.
Hi-Res Audio Driver	`CONFIG_SND_HDA_CODEC_CA0132=m`	CA0132 Sound Core3D HDA codec as loadable module, with DSP firmware support (`CA0132_DSP=y`) for Sound Blaster Z 32-bit / 192 kHz audio.
CPU Mitigations	`CONFIG_CPU_MITIGATIONS=y`	Full Spectre/Meltdown/RetBleed/SRSO/SSB/TSA mitigation suite retained — this is not a "performance at all costs" kernel.
CET / IBT	`CONFIG_X86_CET=y` / `CONFIG_X86_KERNEL_IBT=y`	Hardware Indirect Branch Tracking for kernel control-flow integrity (supported on Zen 3).
User Shadow Stack	`CONFIG_X86_USER_SHADOW_STACK=y`	CET Shadow Stack for user-space return address protection.
Bloat Trimming	Disabled unused CPU/GPU/Drivers	Removed Intel/Nvidia drivers, legacy AMD SI/CIK, and unused network vendors.

🛠️ Custom Kernel Optimizations & Performance Features (config-6.19.14-debian13)

This custom kernel build and CI/CD pipeline are specifically tuned for this exact hardware: AMD Ryzen 9 5950X (Zen 3) and AMD Radeon RX 6950 XT (RDNA 2) running on Debian 13.

The guiding philosophy is hardware/driver bloat removal while retaining security and diagnostic capabilities:

🔧 CPU-specific optimization (Zen 3, 32 threads, AMD pstate, PREEMPT, 1000 Hz)
🎮 GPU-specific support (RDNA 2 / amdgpu module, ROCm, VRAM cgroups)
🌐 Targeted network drivers only (I211 + AX210)
🔐 Security retained (all CPU mitigations, CET/IBT, BPF LSM)
🔍 Diagnostics retained (BTF, Ftrace, Kprobes, ORC)
🗑️ Hardware bloat removed (Intel GPU, NVIDIA, legacy Radeon, unused network, legacy buses)

🧹 1. Hardware Trimming & Bloat Elimination

Non-AMD CPU Support Removed: Disabled Intel, Hygon, Centaur, and Zhaoxin CPU support (# CONFIG_CPU_SUP_INTEL is not set, etc.) to streamline kernel execution paths.
Unused GPU Drivers Removed: Disabled Intel i915 (# CONFIG_DRM_I915 is not set) and Nvidia Nouveau (# CONFIG_DRM_NOUVEAU is not set).
Legacy AMDGPU Generations Removed: Disabled legacy Southern Islands (SI) and Sea Islands (CIK) support (# CONFIG_DRM_AMDGPU_SI is not set, # CONFIG_DRM_AMDGPU_CIK is not set), eliminating si_support / cik_support parameter warnings in dmesg.
Targeted Wireless & Network Drivers: Kept strictly iwlwifi (Intel AX210) and igb (Intel I211), stripping unnecessary Realtek, Broadcom, Atheros, and Ralink wireless/ethernet drivers.
AMD IOMMU Isolation: Native AMD Vi IOMMU driver enabled (CONFIG_AMD_IOMMU=y), while disabling unused Intel DMAR overhead (# CONFIG_INTEL_IOMMU is not set).
Legacy Radeon Driver Disabled: Legacy Radeon DRM driver disabled (# CONFIG_DRM_RADEON is not set), ensuring exclusive amdgpu driver stack execution.
Legacy Controllers Disabled: Removed floppy, parallel ports (PARPORT), PCMCIA/CardBus, FireWire (IEEE1394), ISDN, and analog modems.
DVB & TV Capture Removal: Disabled DVB digital/analog TV, SDR radio, and PCI capture cards (# CONFIG_DVB_CORE is not set, # CONFIG_MEDIA_PCI_SUPPORT is not set), while preserving USB webcam support (CONFIG_USB_VIDEO_CLASS=m).

⏱️ 2. Low-Latency Tuning (Gaming & High-Res Audio)

Full Preemption: Full preemptible kernel (CONFIG_PREEMPT_BUILD=y, CONFIG_PREEMPT=y) for immediate task response and minimal audio/input latency.
Timer Frequency (1000 Hz): Set timer frequency to 1000 Hz (CONFIG_HZ_1000=y, CONFIG_HZ=1000) for precise event timing.
Tickless Idle (NO_HZ_IDLE): Idle-tickless kernel (CONFIG_NO_HZ_IDLE=y) suppressing timer interrupts on idle CPUs, reducing wakeup overhead. Full tickless mode (NO_HZ_FULL) is intentionally not enabled to avoid scheduler complexity on this configuration.
Hi-Res Audio Driver: Dedicated Sound Blaster Z ALSA driver (snd_ca0132 / Sound Core3D) configured for 32-bit / 192 kHz low-jitter audio.
PlayStation DualSense & DualShock 4: Dedicated Sony PlayStation HID driver (CONFIG_HID_PLAYSTATION=m, CONFIG_PLAYSTATION_FF=y) with full haptic Force Feedback for PS4/PS5 gamepads over USB and Bluetooth.

🧠 3. CPU Optimizations (AMD Ryzen 9 5950X — 16C / 32T)

Native Architecture Compilation — two-layer optimization:
- CONFIG_X86_NATIVE_CPU=y — instructs Kconfig to probe the build host CPU and enable matching kernel features (e.g., VAES, AVX2 crypto paths, AMD-specific code paths).
- KCFLAGS="-march=znver3" — instructs GCC to emit Zen 3-specific instructions throughout the build. Set in the CI workflow (make bindeb-pkg KCFLAGS="-march=znver3").
- Both flags are complementary and required for full native optimization; neither alone is sufficient.
Core Scheduling: Hardware SMT core scheduling enabled (CONFIG_SCHED_CORE=y) for thread isolation, security, and hyperthreading gaming performance.
Sized Core Count: Sized to CONFIG_NR_CPUS=32 matching exact hardware threads, removing virtual CPU overhead.
AMD P-State Driver: Native amd-pstate CPPC driver with EPP enabled (CONFIG_X86_AMD_PSTATE=y, CONFIG_X86_AMD_PSTATE_UT=y) for microsecond-level frequency scaling.
NUMA Infrastructure: NUMA support is retained (CONFIG_NUMA=y, CONFIG_AMD_NUMA=y, CONFIG_X86_64_ACPI_NUMA=y) for correct hardware topology detection. Automatic NUMA balancing (# CONFIG_NUMA_BALANCING is not set) is disabled to eliminate background memory scan overhead on this single-node Ryzen system.
eBPF Extensible Scheduler (sched-ext): Enabled in-kernel eBPF scheduler class (CONFIG_SCHED_CLASS_EXT=y). This exposes the sched-ext API that relies on the kernel BPF infrastructure (CONFIG_BPF=y, CONFIG_BPF_SYSCALL=y, CONFIG_BPF_JIT=y). External schedulers like scx_bpfland or scx_rusty can be loaded at runtime via userspace tools — they are not part of this kernel build.
Extended Group Scheduling: CONFIG_EXT_GROUP_SCHED=y enables the extended group scheduling infrastructure used by sched-ext and cgroup-aware scheduling policies.

🎮 4. GPU & Display Core (Radeon RX 6950 XT / Navi 21)

AMDGPU as Module: CONFIG_DRM_AMDGPU=m — the AMDGPU driver is a loadable kernel module, not built into the kernel image. This is the standard Debian/upstream configuration for GPU drivers.
Display Core (DCN 3.0): AMD Display Core enabled (CONFIG_DRM_AMD_DC=y) with fast-path support (CONFIG_DRM_AMD_DC_FP=y), optimized for Navi 21 / RDNA 2 architecture (DCN 3.0).
Additional AMDGPU features: Audio CoProcessor (CONFIG_DRM_AMD_ACP=y), Secure Display (CONFIG_DRM_AMD_SECURE_DISPLAY=y), and ISP support (CONFIG_DRM_AMD_ISP=y) enabled.
Heterogeneous Compute (ROCm / KFD): AMD HSA kernel driver (CONFIG_HSA_AMD=y) enabled for ROCm, Vulkan, and OpenCL compute.
Direct GPU Memory Access: CONFIG_DRM_AMDGPU_USERPTR=y allowing GPU direct access to user space memory pointers.
Resizable BAR & OverDrive: Full 16 GB VRAM BAR access and OverDrive power limit unlocking (amdgpu.ppfeaturemask=0xffffffff).
VRAM Cgroup Management (CONFIG_CGROUP_DMEM=y): Enabled cgroups v2 Device Memory controller for amdgpu, allowing dynamic VRAM priority allocation (dmemcg-booster) to eliminate stutter under heavy VRAM pressure.

🌐 5. Networking & Congestion Control

TCP BBR Default: Configured with TCP BBR as the default congestion control algorithm (CONFIG_DEFAULT_TCP_CONG="bbr", CONFIG_DEFAULT_BBR=y). Eliminates bufferbloat and maximizes bandwidth throughput.
eBPF JIT & LSM: JIT compiler enabled and active by default (CONFIG_BPF_JIT=y, CONFIG_BPF_JIT_DEFAULT_ON=y) alongside eBPF Security Module (CONFIG_BPF_LSM=y) for zero-overhead security hooks. BPF_JIT_ALWAYS_ON is intentionally not set, preserving runtime control via net.core.bpf_jit_enable.
IO_URING_ZCRX: Zero-copy network reception (CONFIG_IO_URING_ZCRX=y) enabled for ultra-fast socket I/O.

💾 6. Memory Tuning (Zswap `lzo`)

Zswap Storage: Native zswap with lzo compressor (CONFIG_ZSWAP=y, CONFIG_ZSWAP_COMPRESSOR_DEFAULT_LZO=y) matching kernel boot parameters.
Transparent Hugepages (madvise): Set THP default to madvise (CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y) alongside memory compaction (CONFIG_COMPACTION=y) to prevent system-wide memory bloat while enabling 2MB hugepages for opt-in applications (games, emulators, KVM).

🛠️ 7. Diagnostics & Tracing Infrastructure

CONFIG_DEBUG_KERNEL=y is set intentionally. This enables the diagnostic/tracing infrastructure without heavyweight runtime overhead — KASAN, KFENCE, SLUB_DEBUG, DEBUG_VM, and lock validators are not set.

ORC Unwinder & Hung Task Detection: ORC stack unwinder (CONFIG_UNWINDER_ORC=y) and automatic 120s hung task detection (CONFIG_DETECT_HUNG_TASK=y).
Scheduler Statistics: Real-time scheduler statistics (CONFIG_SCHEDSTATS=y, CONFIG_SCHED_INFO=y) for performance profiling and perf sched analysis.
Full Ftrace Suite: Complete function tracing infrastructure (CONFIG_FTRACE=y, CONFIG_FUNCTION_TRACER=y, CONFIG_FUNCTION_GRAPH_TRACER=y, CONFIG_DYNAMIC_FTRACE=y, CONFIG_STACK_TRACER=y).
Kernel Probes: CONFIG_KPROBE_EVENTS=y, CONFIG_UPROBE_EVENTS=y, CONFIG_BPF_EVENTS=y — enabling perf, bpftrace, and eBPF-based tracing tools.
BTF Type Information: CONFIG_DEBUG_INFO_BTF=y and CONFIG_DEBUG_INFO_BTF_MODULES=y — required for CO-RE eBPF programs and tools like bpftool and bpftrace.
Block I/O Tracing: CONFIG_BLK_DEV_IO_TRACE=y for blktrace/blkparse disk I/O analysis.
Automated Workflow: GitHub Actions workflow generating .deb packages with explicit -march=znver3 flags and dynamic Git commit changelogs on GitHub Releases.

🔐 8. CPU Security & Mitigations (Zen 3 / x86_64)

This kernel is not a "performance at all costs" build — all CPU security mitigations are retained:

CPU Mitigations Enabled: Full mitigation suite active (CONFIG_CPU_MITIGATIONS=y), including Spectre v1/v2, Retpoline, SRSO (AMD-specific Zen), SSB, RetBleed, IBPB entry, SLS, and TSA.
Page Table Isolation: KPTI enabled (CONFIG_MITIGATION_PAGE_TABLE_ISOLATION=y) for Meltdown protection.
AMD SRSO Mitigation: Speculative Return Stack Overflow mitigation (CONFIG_MITIGATION_SRSO=y) — directly relevant for Zen 3 microarchitecture.
Control Flow Integrity (CET/IBT): Hardware-enforced Indirect Branch Tracking enabled (CONFIG_X86_CET=y, CONFIG_X86_KERNEL_IBT=y) for kernel control-flow integrity.
User Shadow Stack: CET Shadow Stack for user-space enabled (CONFIG_X86_USER_SHADOW_STACK=y) for return address protection.

The security philosophy is hardware/driver bloat removal — not security regression.

🔬 9. Diagnostics, Tracing & Observability

The kernel deliberately retains a comprehensive diagnostic and tracing stack.

This is intentional: the project removes unnecessary runtime hardware support, while keeping the infrastructure needed to diagnose and profile the system.

🔍 Kernel Debug Infrastructure

CONFIG_DEBUG_KERNEL=y
CONFIG_DEBUG_MISC=y
CONFIG_DEBUG_INFO=y
CONFIG_DEBUG_INFO_BTF=y
CONFIG_DEBUG_INFO_BTF_MODULES=y
CONFIG_DYNAMIC_DEBUG=y
CONFIG_DYNAMIC_DEBUG_CORE=y

📊 Scheduler Diagnostics

CONFIG_SCHED_INFO=y
CONFIG_SCHEDSTATS=y
CONFIG_LOCK_DEBUGGING_SUPPORT=y

Heavy lock validation facilities are disabled:

# CONFIG_PROVE_LOCKING is not set
# CONFIG_LOCK_STAT is not set
# CONFIG_DEBUG_SPINLOCK is not set
# CONFIG_DEBUG_MUTEXES is not set
# CONFIG_DEBUG_ATOMIC_SLEEP is not set

This provides scheduler/diagnostic visibility without enabling the heavier debugging stack.

🔎 Ftrace

The kernel retains a full Ftrace suite:

CONFIG_FTRACE=y
CONFIG_FUNCTION_TRACER=y
CONFIG_FUNCTION_GRAPH_TRACER=y
CONFIG_DYNAMIC_FTRACE=y
CONFIG_DYNAMIC_FTRACE_WITH_REGS=y
CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS=y
CONFIG_DYNAMIC_FTRACE_WITH_ARGS=y
CONFIG_STACK_TRACER=y
CONFIG_FTRACE_SYSCALLS=y
CONFIG_KPROBE_EVENTS=y
CONFIG_UPROBE_EVENTS=y
CONFIG_BPF_EVENTS=y
CONFIG_DYNAMIC_EVENTS=y

This enables advanced runtime tracing and profiling with tools like ftrace, perf, bpftrace, and trace-cmd.

🧭 ORC Unwinder

CONFIG_UNWINDER_ORC=y
# CONFIG_UNWINDER_FRAME_POINTER is not set

The ORC unwinder provides efficient and accurate kernel stack unwinding for perf and ftrace call stacks.

🚨 Hung Task / Lockup Detection

CONFIG_LOCKUP_DETECTOR=y
CONFIG_SOFTLOCKUP_DETECTOR=y
CONFIG_HARDLOCKUP_DETECTOR=y
CONFIG_DETECT_HUNG_TASK=y
CONFIG_DEFAULT_HUNG_TASK_TIMEOUT=120

The kernel detects:

soft lockups (NMI watchdog)
hard lockups (perf-based HARDLOCKUP_DETECTOR_PERF)
hung tasks (120 second default timeout)

🔍 14. Post-Boot Verification

After booting into the custom kernel, verify active optimizations using the following commands.

Kernel version

$ uname -a
Linux debian 6.19.14-ryzen9 #ryzen9 SMP PREEMPT_DYNAMIC Mon Sep 28 13:28:57 CEST 2026 x86_64 GNU/Linux

CPU topology

$ lscpu | grep -E "Model name|Thread\(s\) per core|CPU\(s\):"
CPU(s):                32
Thread(s) per core:    2
Model name:            AMD Ryzen 9 5950X 16-Core Processor

AMD P-State

$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
amd-pstate-epp

Preemption

$ sudo dmesg | grep -i "preempt"
[    0.032333] Dynamic Preempt: full
[    0.032388] rcu: Preemptible hierarchical RCU implementation.

CONFIG_PREEMPT_BUILD=y, CONFIG_PREEMPT=y, CONFIG_PREEMPT_DYNAMIC=y

NO_HZ

$ grep -E "CONFIG_NO_HZ_(IDLE|FULL)" /boot/config-$(uname -r)
CONFIG_NO_HZ_IDLE=y
# CONFIG_NO_HZ_FULL is not set

RCU

$ grep -E "CONFIG_RCU_(BOOST|NOCB_CPU)" /boot/config-$(uname -r)

No output expected — RCU_BOOST and RCU_NOCB_CPU are not set in this configuration.

TCP BBR

$ sysctl net.ipv4.tcp_congestion_control
net.ipv4.tcp_congestion_control = bbr

Zswap

$ cat /sys/module/zswap/parameters/enabled
Y
$ cat /sys/module/zswap/parameters/compressor
lzo

AMDGPU

$ sudo dmesg | grep -i amdgpu | head -30
[    4.611866] amdgpu: unknown parameter 'si_support' ignored
[    4.611869] amdgpu: unknown parameter 'cik_support' ignored
[    4.617691] amdgpu: Virtual CRAT table created for CPU
[    4.617703] amdgpu: Topology: Add CPU node
[    4.617726] amdgpu: Overdrive is enabled, please disable it before reporting any bugs unrelated to overdrive.
[    4.617819] amdgpu 0000:0d:00.0: enabling device (0006 -> 0007)
[    4.617849] amdgpu 0000:0d:00.0: amdgpu: initializing kernel modesetting (SIENNA_CICHLID 0x1002:0x73A5 0x1002:0x0E3A 0xC0).
[    4.621608] amdgpu 0000:0d:00.0: amdgpu: detected ip block number 0 <common_v1_0_0> (nv_common)
[    4.623415] amdgpu 0000:0d:00.0: amdgpu: VRAM: 16368M 0x0000008000000000 - 0x00000083FEFFFFFF (16368M used)
[    4.623417] amdgpu 0000:0d:00.0: amdgpu: GART: 512M 0x0000000000000000 - 0x000000001FFFFFFF
[    6.883784] amdgpu 0000:0d:00.0: amdgpu: [drm] Display Core v3.2.359 initialized on DCN 3.0

The si_support/cik_support "ignored" warnings confirm that legacy SI/CIK support is correctly disabled in the config (# CONFIG_DRM_AMDGPU_SI is not set). VRAM: 16368M with full ReBAR mapping.

Device Memory Cgroup

$ grep CONFIG_CGROUP_DMEM /boot/config-$(uname -r)
CONFIG_CGROUP_DMEM=y

BPF JIT

$ grep -E "CONFIG_BPF_JIT" /boot/config-$(uname -r)
CONFIG_BPF_JIT=y
CONFIG_BPF_JIT_DEFAULT_ON=y
# CONFIG_BPF_JIT_ALWAYS_ON is not set

sched-ext

$ cat /sys/kernel/sched_ext/state 2>/dev/null
disabled

disabled means the mechanism is available but no external BPF scheduler is currently loaded. Use scx_bpfland or scx_rusty to activate.

ORC Unwinder

$ grep CONFIG_UNWINDER_ORC /boot/config-$(uname -r)
CONFIG_UNWINDER_ORC=y

GPU / ReBAR

$ sudo dmesg | grep -Ei "amdgpu|BAR|Resizable"

🧪 15. Kernel Configuration Verification

The configuration embedded in the installed kernel can be extracted directly from the Debian package:

$ dpkg-deb --fsys-tarfile linux-image-6.19.14-ryzen9_*.deb \
  | tar -x ./boot/config-6.19.14-ryzen9 -O \
  | grep -E "CONFIG_(X86_NATIVE_CPU|NR_CPUS|PREEMPT|NO_HZ|BPF_JIT|CGROUP_DMEM|DRM_AMDGPU|HSA_AMD)"
CONFIG_NO_HZ_IDLE=y
CONFIG_PREEMPT_BUILD=y
CONFIG_PREEMPT=y
CONFIG_PREEMPT_DYNAMIC=y
CONFIG_BPF_JIT=y
CONFIG_BPF_JIT_DEFAULT_ON=y
CONFIG_NR_CPUS=32
CONFIG_X86_NATIVE_CPU=y
CONFIG_DRM_AMDGPU=m
CONFIG_HSA_AMD=y
CONFIG_CGROUP_DMEM=y

This allows the packaged kernel to be verified independently of the source-tree configuration.

🖥️ Workstation Hardware Specs

🐧 Kernel and Toolchain

Element	Value
Kernel	Linux 6.19.14-ryzen9
Model	SMP PREEMPT_DYNAMIC
Base Distribution	Debian 13
Compiler	GCC 14.2.0
Target Architecture	Zen 3
Grub	`GRUB_CMDLINE_LINUX_DEFAULT="quiet amdgpu.ppfeaturemask=0xffffffff zswap.enabled=1 zswap.compressor=lzo"`

📦 Build Dependencies

sudo apt update && sudo apt install -y build-essential gcc-14 g++-14 fakeroot bc bison flex zstd \
  libdw-dev libelf-dev libssl-dev libncurses-dev dwarves debhelper rsync python3 ccache curl jq patch kmod perl mawk tar xz-utils git

Paquete	Versión
`build-essential`	`12.12`
`gcc-14`	`14.2.0-19`
`g++-14`	`14.2.0-19`
`fakeroot`	`1.37.1.1-1`
`bc`	`1.07.1-4`
`bison`	`2:3.8.2+dfsg-1+b2`
`flex`	`2.6.4-8.2+b4`
`zstd`	`1.5.7+dfsg-1`
`libdw-dev`	`0.192-4`
`libelf-dev`	`0.192-4`
`libssl-dev`	`3.5.7-1~deb13u2`
`libncurses-dev`	`6.5+20250216-2`
`dwarves`	`1.30-1`
`debhelper`	`13.24.2`
`rsync`	`3.4.1+ds1-5+deb13u4`
`python3`	`3.13.5-1`
`curl`	`8.14.1-2+deb13u5`
`jq`	`1.7.1-6+deb13u3`
`patch`	`2.8-2`
`kmod`	`34.2-2`
`perl`	`5.40.1-6+deb13u1`
`mawk`	`1.3.4.20250131-1`
`tar`	`1.35+dfsg-3.1`
`xz-utils`	`5.8.1-1+deb13u1`
`git`	`1:2.47.3-0+deb13u1`

🧠 CPU

Component	Details
Architecture	x86_64 `amd64`
Core Architecture	Zen 3
CPU	AMD Ryzen 9 5950X (Vermeer)
Cores / Threads	16 cores / 32 threads
Socket	AM4
Maximum Frequency	~4.9 GHz (boost)
TDP	105W
SMT	Enabled
NUMA	1 node
Virtualization	AMD-V enabled
L1 Cache	512 KiB (16×32 KiB data + 16×32 KiB instruction)
L2 Cache	8 MiB (16×512 KiB)
L3 Cache	64 MiB (2 CCDs)
Instruction Sets	AVX2, FMA, AES-NI, SHA-NI, VAES, BMI1/2
Part	100-100000059WOF

💾 Memory RAM

Parameter	Value
Total Capacity	128 GB
Configuration	4 × 32 GB
Type	DDR4
Speed	3600 MT/s
CL	18-22-22-42
Voltage	1.35v
Channels	Dual Channel
ECC	No
Model	G.Skill F4-3600C18D-64GTZN
EAN	4713294224835

🎮 GPU

Component	Details
GPU	AMD Radeon RX 6950 XT
Architecture	RDNA 2 (Navi 21)
PCI ID	`1002:73a5`
Kernel Driver	`amdgpu`
DRM/KMS	Enabled

🎮 GPU APIs

API	Version	Device / Driver
AMDGPU	3.64.0	DRM 3.64 kernel driver for NAVI21 (6.19.14-ryzen9)
Vulkan	1.4.305	RADV (Mesa 25.0.7) for AMD Radeon RX 6950 XT
OpenCL	3.0	OpenCL C 1.2 via ROCr / RustiCL / Mesa
OpenGL	4.6	Mesa 25.0.7 (Compatibility Profile, LLVM 19.1.7)
BAR	Enabled	Runtime Detection: `VRAM RAM=16368M, BAR=16384M (Resizable BAR Enabled)`

🧩 Motherboard

Component	Details
Motherboard	Gigabyte X570 AORUS ELITE (rev. 1.0)
Chipset	AMD X570
Manufacturer	Gigabyte Technology Co., Ltd.
BIOS	AMI (American Megatrends)
BIOS Version	F40
BIOS Date	2025-10-29
Boot Mode	UEFI
SMBIOS	3.3.0
AMD AGESA	1.2.0.F

🔐 Trusted Platform Module (TPM)

Parameter	Value
Type	fTPM (Firmware TPM)
Version	TPM 2.0
Manufacturer	AMD
Implementation	AMD CPU fTPM
Interface	CRB (Command Response Buffer)
ACPI	TPM2 table present
Device	`/dev/tpm0`, `/dev/tpmrm0`
Permissions	`tss` group
Kernel Driver	`tpm_crb` (built-in)
Status	Enabled in UEFI
Compliance	FIPS 140-2
TPM Revision	1.38
PCRs	24
Input Buffer	1024 bytes
Max Command Size	4096 bytes
Max Response Size	4096 bytes
Hardware RNG	Disabled (CRB design)

🔊 Audio

Component	Details
Sound Card	Creative Sound Blaster Z
Chip	CA0132 Sound Core3D
PCI ID	`1102:0012`
Driver	ALSA (`snd_ca0132`)
Hi-res Audio	Enabled `32 bits / 192kHz`
Speakers	Edifier M60

🌐 Network — Ethernet

Component	Details
Controller	Intel I211 Gigabit
PCI ID	`8086:1539`
Driver	`igb`

📡 Wi-Fi / Bluetooth

Component	Details
Wi-Fi	Intel AX210
Standard	Wi-Fi 6E (802.11ax)
PCI ID	`8086:2725`
Driver	`iwlwifi`
Bluetooth	5.3
USB ID	`8087:0032`
Driver	`btusb` + `btintel` (kernel modules loaded)

🥶 Cooling

Component	Details	Store
AIO	Fractal Celsius+ Prisma S36	Amazon
Fan front	3x Noctua NF-P12 redux-1700	Amazon
Fan top	2x Noctua NF-A14 PWM	Amazon
Fan rear	1x Noctua NF-F12 PWM	Amazon
Thermal Paste	Noctua NT-H2	Amazon
Fan hub	Noctua NA-FH1	Amazon

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
configs		configs
patchs		patchs
LICENSE		LICENSE
README.md		README.md

Latest commit

History

Folders and files

Repository files navigation