Drop unnecessary capabilities after BPF initialization#3450
Open
robbycochran wants to merge 6 commits into
Open
Drop unnecessary capabilities after BPF initialization#3450robbycochran wants to merge 6 commits into
robbycochran wants to merge 6 commits into
Conversation
After BPF programs are loaded and attached, collector no longer needs CAP_BPF, CAP_PERFMON, or CAP_SYS_RESOURCE. Drop all capabilities except CAP_SYS_PTRACE (needed for ongoing /proc scraping) immediately after InitKernel() succeeds. On kernels < 5.8 (RHEL 8), also keep CAP_SYS_ADMIN since the BPF subsystem may check it on runtime operations. This makes collector self-hardening regardless of how it is deployed — even with privileged:true, the process capability set is minimized before entering the event loop. libcap-ng was already linked but unused; this is the first actual use.
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Enterprise Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## master #3450 +/- ##
==========================================
+ Coverage 27.34% 27.37% +0.03%
==========================================
Files 95 96 +1
Lines 5420 5432 +12
Branches 2545 2551 +6
==========================================
+ Hits 1482 1487 +5
- Misses 3211 3214 +3
- Partials 727 731 +4
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. |
Code audit of falcosecurity-libs found two runtime paths that may need BPF capabilities: 1. bpf_map_lookup_elem() in pman_get_scap_stats() — called periodically for metrics collection via CollectorStatsExporter 2. sinsp::restart_capture() — can re-attach BPF programs via bpf_program__attach() on SCAP_UNEXPECTED_BLOCK events Drop CAP_SYS_RESOURCE (only needed for RLIMIT_MEMLOCK at init) and all other capabilities not in the keep list. On kernel >= 5.8 keep SYS_PTRACE + BPF + PERFMON; on older kernels keep SYS_PTRACE + SYS_ADMIN. This still drops ~38 of 41 capabilities on modern kernels, eliminating NET_RAW, SYS_MODULE, DAC_OVERRIDE, CHOWN, FOWNER, MKNOD, SYS_RAWIO, and all other unneeded privileges.
Add DropCapabilities() utility in DropCapabilities.h and apply it at each thread entry point with only the capabilities that thread needs: - Main thread (event loop): BPF, PERFMON, SYS_PTRACE - CollectorStatsExporter: BPF (map lookups for metrics) - NetworkStatusNotifier: SYS_PTRACE (/proc reads) - SignalServiceClient: none (gRPC only) - ConfigLoader: none (inotify + yaml only) Uses CAPNG_SELECT_ALL to also clear the bounding set, preventing capability regain via execve. Based on prior work by Dmitrii Dolgov in PR #1908.
The main thread drops the bounding set (CAPNG_SELECT_ALL) which requires CAP_SETPCAP. After that, worker threads can no longer modify the bounding set. Fix by having only the main thread clear the bounding set (clear_bounding=true) while worker threads only modify effective + permitted (CAPNG_SELECT_CAPS).
|
/retest collector-on-push |
2 similar comments
|
/retest collector-on-push |
|
/retest collector-on-push |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Self-harden collector by dropping unnecessary Linux capabilities after BPF initialization, regardless of how it is deployed. Even with
privileged: true, the process capability set is minimized per-thread before entering the event loop.Based on prior work by @erthalion in #1908.
Per-thread capability model
RunService()CollectorStatsExporter::run()bpf_map_lookup_elem()for metricsNetworkStatusNotifier::Run()SignalServiceClient::EstablishGRPCStream()ConfigLoader::WatchFile()On kernels < 5.8 (RHEL 8), the main thread keeps SYS_ADMIN + SYS_PTRACE instead of BPF + PERFMON + SYS_PTRACE.
Implementation
DropCapabilities.h— single utility: clears all caps + bounding set viaCAPNG_SELECT_ALL, adds back only what is specifiedHostInfo::GetKernelVersion()for RHEL 8 fallbackWhat gets dropped (~38 of 41 capabilities)
NET_RAW, SYS_MODULE, DAC_OVERRIDE, CHOWN, FOWNER, KILL, SETUID, SETGID, NET_BIND_SERVICE, MKNOD, SYS_RAWIO, AUDIT_WRITE, SYS_RESOURCE, DAC_READ_SEARCH, NET_ADMIN, and all others not in the keep list above.
Differences from #1908
Companion to #3447 which reduces the deployment-side privileges.
Test plan