| 1 |
Allows a process to manipulate aspects of the kernel |
| 2 |
enhanced Berkeley Packet Filter (BPF) system. This is |
| 3 |
an execution subsystem of the kernel, that manages BPF |
| 4 |
programs. CAP_BPF permits a process to: |
| 5 |
- create all types of BPF maps |
| 6 |
- advanced verifier features: |
| 7 |
- indirect variable access |
| 8 |
- bounded loops |
| 9 |
- BPF to BPF function calls |
| 10 |
- scalar precision tracking |
| 11 |
- larger complexity limits |
| 12 |
- dead code elimination |
| 13 |
- potentially other features |
| 14 |
|
| 15 |
Other capabilities can be used together with CAP_BFP to |
| 16 |
further manipulate the BPF system: |
| 17 |
- CAP_PERFMON relaxes the verifier checks as follows: |
| 18 |
- BPF programs can use pointer-to-integer |
| 19 |
conversions |
| 20 |
- speculation attack hardening measures can be |
| 21 |
bypassed |
| 22 |
- bpf_probe_read to read arbitrary kernel memory is |
| 23 |
permitted |
| 24 |
- bpf_trace_printk to print the content of kernel |
| 25 |
memory |
| 26 |
- CAP_SYS_ADMIN permits the following: |
| 27 |
- use of bpf_probe_write_user |
| 28 |
- iteration over the system-wide loaded programs, |
| 29 |
maps, links BTFs and convert their IDs to file |
| 30 |
descriptors. |
| 31 |
- CAP_PERFMON is required to load tracing programs. |
| 32 |
- CAP_NET_ADMIN is required to load networking |
| 33 |
programs. |