Re: [PATCH bpf-next v3 1/4] bpf: Adapt 32-bit return value kfunc for 32-bit ARM when zext extension

From: Yang Jihong
Date: Wed Dec 07 2022 - 03:53:19 EST


Hello,

On 2022/12/5 9:19, Yang Jihong wrote:


On 2022/12/4 0:40, Alexei Starovoitov wrote:
On Fri, Dec 2, 2022 at 6:58 PM Yang Jihong <yangjihong1@xxxxxxxxxx> wrote:



On 2022/11/29 0:41, Alexei Starovoitov wrote:
On Mon, Nov 28, 2022 at 4:40 AM Yang Jihong <yangjihong1@xxxxxxxxxx> wrote:



On 2022/11/28 9:57, Alexei Starovoitov wrote:
On Sat, Nov 26, 2022 at 05:45:27PM +0800, Yang Jihong wrote:
For ARM32 architecture, if data width of kfunc return value is 32 bits,
need to do explicit zero extension for high 32-bit, insn_def_regno should
return dst_reg for BPF_JMP type of BPF_PSEUDO_KFUNC_CALL. Otherwise,
opt_subreg_zext_lo32_rnd_hi32 returns -EFAULT, resulting in BPF failure.

Signed-off-by: Yang Jihong <yangjihong1@xxxxxxxxxx>
---
    kernel/bpf/verifier.c | 44 ++++++++++++++++++++++++++++++++++++++++---
    1 file changed, 41 insertions(+), 3 deletions(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 264b3dc714cc..193ea927aa69 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -1927,6 +1927,21 @@ find_kfunc_desc(const struct bpf_prog *prog, u32 func_id, u16 offset)
                      sizeof(tab->descs[0]), kfunc_desc_cmp_by_id_off);
    }

+static int kfunc_desc_cmp_by_imm(const void *a, const void *b);
+
+static const struct bpf_kfunc_desc *
+find_kfunc_desc_by_imm(const struct bpf_prog *prog, s32 imm)
+{
+    struct bpf_kfunc_desc desc = {
+            .imm = imm,
+    };
+    struct bpf_kfunc_desc_tab *tab;
+
+    tab = prog->aux->kfunc_tab;
+    return bsearch(&desc, tab->descs, tab->nr_descs,
+                   sizeof(tab->descs[0]), kfunc_desc_cmp_by_imm);
+}
+
    static struct btf *__find_kfunc_desc_btf(struct bpf_verifier_env *env,
                                        s16 offset)
    {
@@ -2342,6 +2357,13 @@ static bool is_reg64(struct bpf_verifier_env *env, struct bpf_insn *insn,
                        */
                       if (insn->src_reg == BPF_PSEUDO_CALL)
                               return false;
+
+                    /* Kfunc call will reach here because of insn_has_def32,
+                     * conservatively return TRUE.
+                     */
+                    if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL)
+                            return true;
+
                       /* Helper call will reach here because of arg type
                        * check, conservatively return TRUE.
                        */
@@ -2405,10 +2427,26 @@ static bool is_reg64(struct bpf_verifier_env *env, struct bpf_insn *insn,
    }

    /* Return the regno defined by the insn, or -1. */
-static int insn_def_regno(const struct bpf_insn *insn)
+static int insn_def_regno(struct bpf_verifier_env *env, const struct bpf_insn *insn)
    {
       switch (BPF_CLASS(insn->code)) {
       case BPF_JMP:
+            if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL) {
+                    const struct bpf_kfunc_desc *desc;
+
+                    /* The value of desc cannot be NULL */
+                    desc = find_kfunc_desc_by_imm(env->prog, insn->imm);
+
+                    /* A kfunc can return void.
+                     * The btf type of the kfunc's return value needs
+                     * to be checked against "void" first
+                     */
+                    if (desc->func_model.ret_size == 0)
+                            return -1;
+                    else
+                            return insn->dst_reg;
+            }
+            fallthrough;

I cannot make any sense of this patch.
insn->dst_reg above is 0.
The kfunc call doesn't define a register from insn_def_regno() pov.

Are you hacking insn_def_regno() to return 0 so that
if (WARN_ON(load_reg == -1)) {
     verbose(env, "verifier bug. zext_dst is set, but no reg is defined\n");
     return -EFAULT;
}
in opt_subreg_zext_lo32_rnd_hi32() doesn't trigger ?

But this verifier message should have been a hint that you need
to analyze why zext_dst is set on this kfunc call.
Maybe it shouldn't ?
Did you analyze the logic of mark_btf_func_reg_size() ?
make r0 zext is not caused by mark_btf_func_reg_size.

This problem occurs when running the kfunc_call_test_ref_btf_id test
case in the 32-bit ARM environment.

Why is it not failing on x86-32 ?
Use the latest mainline kernel code to test on the x86_32 machine. The
test also fails:

    # ./test_progs -t kfunc_call/kfunc_call_test_ref_btf_id
    Failed to load bpf_testmod.ko into the kernel: -8
    WARNING! Selftests relying on bpf_testmod.ko will be skipped.
    libbpf: prog 'kfunc_call_test_ref_btf_id': BPF program load failed:
Bad address
    libbpf: prog 'kfunc_call_test_ref_btf_id': -- BEGIN PROG LOAD LOG --
    processed 25 insns (limit 1000000) max_states_per_insn 0 total_states
2 peak_states 2 mark_read 1
    -- END PROG LOAD LOG --
    libbpf: prog 'kfunc_call_test_ref_btf_id': failed to load: -14
    libbpf: failed to load object 'kfunc_call_test'
    libbpf: failed to load BPF skeleton 'kfunc_call_test': -14
    verify_success:FAIL:skel unexpected error: -14

Therefore, this problem also exists on x86_32:
"verifier bug. zext_dst is set, but no reg is defined"

The kernel returns -14 == EFAULT.
That's a completely different issue.
It's the same problem. The opt_subreg_zext_lo32_rnd_hi32 function fails to check here and returns -EFAULT

opt_subreg_zext_lo32_rnd_hi32 {
  ...
   if (WARN_ON(load_reg == -1)) {
           verbose(env, "verifier bug. zext_dst is set, but no reg is defined\n");
           return -EFAULT;
   }
  ...
}
.
I see that there are emails from the community talking about the same problem, and come up with a solution:
https://lore.kernel.org/bpf/20221202103620.1915679-1-bjorn@xxxxxxxxxx/T/

will remove this patch based on that patch.

Thanks,
Yang


.