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
Looking for hackers with the skills:
This project is part of:
Hack Week 21 Hack Week 23 Hack Week 24
Activity
Comments
-
about 3 years ago by shunghsiyu | Reply
I've uploaded the jupyter notebook on GitHub that contains a minimal model of tnum along with tnum_add(), as well as the prove that it works.
-
about 3 years ago by shunghsiyu | Reply
The slides used for lightning talk can be found here
While I'd like to achieve much more, I think what I've done during hack week is suffice to be called a complete project, so I'm marking this as complete
-
over 1 year ago by shunghsiyu | Reply
I'm restarting this project to model check the range tracking (minimal and maximal value possible in s32/u32/u64/s64 range) done in BPF verifier.
In addition to that I hope to unify signed and unsigned range tracking based on previously posted idea and the "Interval Analysis and Machine Arithmetic: Why Signedness Ignorance Is Bliss" paper for simpler range tracking code.
-
about 1 month ago by rubylee | Reply
Model checking the BPF verifier is crucial for ensuring system safety and reliability—much like verifying the lute instrument price before buying, it’s about making sure everything checks out before trusting the result.
Similar Projects
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
- Investigate if this is possible and the implications it would have (done in HW21)
- Hack up a PoC (done in HW22 and HW23)
- 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 afterkcsan_init() - I workaround for a smoke test was to hack
kexec_file_load()systemcall which has two problems:- My initramfs in the porduction kernel does not have a new enough kexec version, that's not a blocker but where the week ended
- 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
__initafterkcsan_init()
- getting to call the invocation of loading the early kernel from
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()
- crash systems crashes (no pun intended) in
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/
- hacking
TODOs
- fix elfcorehdr so that we actually can make use of all this...
- test where in the boot
__init()chain we can/should callkexec_early_dump()