]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
bpf: Relax KF_ACQUIRE kfuncs strict type matching constraint
authorJuntong Deng <juntong.deng@outlook.com>
Wed, 28 Aug 2024 19:48:11 +0000 (20:48 +0100)
committerAlexei Starovoitov <ast@kernel.org>
Thu, 29 Aug 2024 00:11:54 +0000 (17:11 -0700)
Currently we cannot pass zero offset (implicit cast) or non-zero offset
pointers to KF_ACQUIRE kfuncs. This is because KF_ACQUIRE kfuncs
requires strict type matching, but zero offset or non-zero offset does
not change the type of pointer, which causes the ebpf program to be
rejected by the verifier.

This can cause some problems, one example is that bpf_skb_peek_tail
kfunc [0] cannot be implemented by just passing in non-zero offset
pointers. We cannot pass pointers like &sk->sk_write_queue (non-zero
offset) or &sk->__sk_common (zero offset) to KF_ACQUIRE kfuncs.

This patch makes KF_ACQUIRE kfuncs not require strict type matching.

[0]: https://lore.kernel.org/bpf/AM6PR03MB5848CA39CB4B7A4397D380B099B12@AM6PR03MB5848.eurprd03.prod.outlook.com/

Signed-off-by: Juntong Deng <juntong.deng@outlook.com>
Link: https://lore.kernel.org/r/AM6PR03MB5848FD2BD89BF0B6B5AA3B4C99952@AM6PR03MB5848.eurprd03.prod.outlook.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
kernel/bpf/verifier.c

index 33270b3636892e078444189f3d271e8527470fc9..f32e3b9bb4e5655ffc9dc1c55f9161c592ff209a 100644 (file)
@@ -11513,8 +11513,7 @@ static int process_kf_arg_ptr_to_btf_id(struct bpf_verifier_env *env,
         * btf_struct_ids_match() to walk the struct at the 0th offset, and
         * resolve types.
         */
-       if (is_kfunc_acquire(meta) ||
-           (is_kfunc_release(meta) && reg->ref_obj_id) ||
+       if ((is_kfunc_release(meta) && reg->ref_obj_id) ||
            btf_type_ids_nocast_alias(&env->log, reg_btf, reg_ref_id, meta->btf, ref_id))
                strict_type_match = true;