early stage kdump support by mbrugger

Project Description

When we experience a early boot crash, we are not able to analyze the kernel dump, as user-space wasn't able to load the crash system. The idea is to make the crash system compiled into the host kernel (think of initramfs) so that we can create a kernel dump really early in the boot process.

Goal for the Hackweeks

1. Investigate if this is possible and the implications it would have (done in HW21)
2. Hack up a PoC (done in HW22 and HW23)
3. Prepare RFC series (giving it's only one week, we are entering wishful thinking territory here).

update HW23

• I was able to include the crash kernel into the kernel Image.
• I'll need to find a way to load that from init/main.c:start_kernel() probably after kcsan_init()
• I workaround for a smoke test was to hack kexec_file_load() systemcall which has two problems:
  1. My initramfs in the porduction kernel does not have a new enough kexec version, that's not a blocker but where the week ended
  2. As the crash kernel is part of init.data it will be already stale once I can call kexec_file_load() from user-space.

The solution is probably to rewrite the POC so that the invocation can be done from init.text (that's my theory) but I'm not sure if I can reuse the kexec infrastructure in the kernel from there, which I rely on heavily.

update HW24

• Day1
  • rebased on v6.12 with no problems others then me breaking the config
  • setting up a new compilation and qemu/virtme env
  • getting desperate as nothing works that used to work
• Day 2
  • getting to call the invocation of loading the early kernel from __init after kcsan_init()

• Day 3

  • fix problem of memdup not being able to alloc so much memory... use 64K page sizes for now
  • code refactoring
  • I'm now able to load the crash kernel
  • When using virtme I can boot into the crash kernel, also it doesn't boot completely (major milestone!), crash in elfcorehdr_read_notes()

• Day 4

  • crash systems crashes (no pun intended) in copy_old_mempage() link; will need to understand elfcorehdr...
  • call path vmcore_init() -> parse_crash_elf_headers() -> elfcorehdr_read() -> read_from_oldmem() -> copy_oldmem_page() -> copy_to_iter()

• Day 5

  • hacking arch/arm64/kernel/crash_dump.c:copy_old_mempage() to see if crash system really starts. It does.
  • fun fact: retested with more reserved memory and with UEFI FW, host kernel crashes in init but directly starts the crash kernel, so it works (somehow) \o/

• TODOs

  • fix elfcorehdr so that we actually can make use of all this...
  • test where in the boot __init() chain we can/should call kexec_early_dump()

---

Improve UML page fault handler by ptesarik

Description

Improve UML handling of segmentation faults in kernel mode. Although such page faults are generally caused by a kernel bug, it is annoying if they cause an infinite loop, or panic the kernel. More importantly, a robust implementation allows to write KUnit tests for various guard pages, preventing potential kernel self-protection regressions.

Goals

Convert the UML page fault handler to use oops_* helpers, go through a few review rounds and finally get my patch series merged in 6.14.

Resources

Wrong initial attempt: https://lore.kernel.org/lkml/20231215121431.680-1-petrtesarik@huaweicloud.com/T/

---

Model checking the BPF verifier by shunghsiyu

Project Description

BPF verifier plays a crucial role in securing the system (though less so now that unprivileged BPF is disabled by default in both upstream and SLES), and bugs in the verifier has lead to privilege escalation vulnerabilities in the past (e.g. CVE-2021-3490).

One way to check whether the verifer has bugs to use model checking (a formal verification technique), in other words, build a abstract model of how the verifier operates, and then see if certain condition can occur (e.g. incorrect calculation during value tracking of registers) by giving both the model and condition to a solver.

For the solver I will be using the Z3 SMT solver to do the checking since it provide a Python binding that's relatively easy to use.

Goal for this Hackweek

Learn how to use the Z3 Python binding (i.e. Z3Py) to build a model of (part of) the BPF verifier, probably the part that's related to value tracking using tristate numbers (aka tnum), and then check that the algorithm work as intended.

Resources

• Formal Methods for the Informal Engineer: Tutorial #1 - The Z3 Theorem Prover and its accompanying notebook is a great introduction into Z3
  • Has a section specifically on model checking
• Software Verification and Analysis Using Z3 a great example of using Z3 for model checking
• Sound, Precise, and Fast Abstract Interpretation with Tristate Numbers - existing work that use formal verification to prove that the multiplication helper used for value tracking work as intended
• [PATCH v5 net-next 00/12] bpf: rewrite value tracking in verifier - initial patch set that adds tristate number to the verifier

---

RISC-V emulator in GLSL capable of running Linux by favogt

Description

There are already numerous ways to run Linux and some programs through emulation in a web browser (e.g. x86 and riscv64 on https://bellard.org/jslinux/), but none use WebGL/WebGPU to run the emulation on the GPU.

I already made a PoC of an AArch64 (64-bit Arm) emulator in OpenCL which is unfortunately hindered by a multitude of OpenCL compiler bugs on all platforms (Intel with beignet or the new compute runtime and AMD with Mesa Clover and rusticl). With more widespread and thus less broken GLSL vs. OpenCL and the less complex implementation requirements for RV32 (especially 32bit integers instead of 64bit), that should not be a major problem anymore.

Goals

Write an RISC-V system emulator in GLSL that is capable of booting Linux and run some userspace programs interactively. Ideally it is small enough to work on online test platforms like Shaderoo with a custom texture that contains bootstrap code, kernel and initrd.

Minimum:

riscv32 without FPU (RV32 IMA) and MMU (µClinux), running Linux in M-mode and userspace in U-mode.

Stretch goals:

FPU support, S-Mode support with MMU, SMP. Custom web frontend with more possibilities for I/O (disk image, network?).

Resources

RISC-V ISA Specifications
Shaderoo
OpenGL 4.5 Quick Reference Card

Result as of Hackweek 2024

WebGL turned out to be insufficient, it only supports OpenGL ES 3.0 but imageLoad/imageStore needs ES 3.1. So we switched directions and had to write a native C++ host for the shaders.

As of Hackweek Friday, the kernel attempts to boot and outputs messages, but panics due to missing memory regions.

Since then, some bugs were fixed and enough hardware emulation implemented, so that now Linux boots with framebuffer support and it's possible to log in and run programs!

The repo with a demo video is available at https://github.com/Vogtinator/risky-v

---

Create DRM drivers for VESA and EFI framebuffers by tdz

Description

We already have simpledrm for firmware framebuffers. But the driver is originally for ARM boards, not PCs. It is already overloaded with code to support both use cases. At the same time it is missing possible features for VESA and EFI, such as palette modes or EDID support. We should have DRM drivers for VESA and EFI interfaces. The infrastructure exists already and initial drivers can be forked from simpledrm.

Goals

• Initially, a bare driver for VESA or EFI should be created. It can take functionality from simpledrm.
• Then we can begin to add additional features. The boot loader can provide EDID data. With VGA hardware, VESA can support paletted modes or color management. Example code exists in vesafb.

---