bpf: Exempt CAP_BPF from checks against bpf_jit_limit
authorLorenz Bauer <lmb@cloudflare.com>
Wed, 22 Sep 2021 11:11:52 +0000 (12:11 +0100)
committerDaniel Borkmann <daniel@iogearbox.net>
Tue, 28 Sep 2021 07:28:37 +0000 (09:28 +0200)
commit8a98ae12fbefdb583a7696de719a1d57e5e940a2
treeb03a979d216d3ae792ae7a325b2ab558e0395ae2
parent37cb28ec7d3a36a5bace7063a3dba633ab110f8b
bpf: Exempt CAP_BPF from checks against bpf_jit_limit

When introducing CAP_BPF, bpf_jit_charge_modmem() was not changed to treat
programs with CAP_BPF as privileged for the purpose of JIT memory allocation.
This means that a program without CAP_BPF can block a program with CAP_BPF
from loading a program.

Fix this by checking bpf_capable() in bpf_jit_charge_modmem().

Fixes: 2c78ee898d8f ("bpf: Implement CAP_BPF")
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20210922111153.19843-1-lmb@cloudflare.com
kernel/bpf/core.c