OSDN Git Service

bpf: Relax verifier recursion check.
authorAlexei Starovoitov <ast@kernel.org>
Thu, 15 Jul 2021 00:54:13 +0000 (17:54 -0700)
committerDaniel Borkmann <daniel@iogearbox.net>
Thu, 15 Jul 2021 20:31:10 +0000 (22:31 +0200)
In the following bpf subprogram:
static int timer_cb(void *map, void *key, void *value)
{
    bpf_timer_set_callback(.., timer_cb);
}

the 'timer_cb' is a pointer to a function.
ld_imm64 insn is used to carry this pointer.
bpf_pseudo_func() returns true for such ld_imm64 insn.

Unlike bpf_for_each_map_elem() the bpf_timer_set_callback() is asynchronous.
Relax control flow check to allow such "recursion" that is seen as an infinite
loop by check_cfg(). The distinction between bpf_for_each_map_elem() the
bpf_timer_set_callback() is done in the follow up patch.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/bpf/20210715005417.78572-8-alexei.starovoitov@gmail.com
kernel/bpf/verifier.c

index 8df2671..1cb1b35 100644 (file)
@@ -9463,8 +9463,12 @@ static int visit_func_call_insn(int t, int insn_cnt,
                init_explored_state(env, t + 1);
        if (visit_callee) {
                init_explored_state(env, t);
-               ret = push_insn(t, t + insns[t].imm + 1, BRANCH,
-                               env, false);
+               ret = push_insn(t, t + insns[t].imm + 1, BRANCH, env,
+                               /* It's ok to allow recursion from CFG point of
+                                * view. __check_func_call() will do the actual
+                                * check.
+                                */
+                               bpf_pseudo_func(insns + t));
        }
        return ret;
 }