Re: [PATCH bpf-next v2 1/4] bpf, verifier: Support direct helper calls from prologue/epilogue
From: Martin KaFai Lau <martin.lau@linux.dev>
Date: 2026-01-23 23:49:55
Also in:
bpf
On 1/23/26 9:05 AM, Jakub Sitnicki wrote:
quoted hunk ↗ jump to hunk
@@ -23909,6 +23928,9 @@ static int do_misc_fixups(struct bpf_verifier_env *env) goto next_insn; } patch_call_imm: + if (env->insn_aux_data[i + delta].finalized_call) + goto next_insn; + fn = env->ops->get_func_proto(insn->imm, env->prog); /* all functions that have prototype and verifier allowed * programs to call them, must be real in-kernel functions@@ -23920,6 +23942,7 @@ static int do_misc_fixups(struct bpf_verifier_env *env) return -EFAULT; } insn->imm = fn->func - __bpf_call_base; + env->insn_aux_data[i + delta].finalized_call = true; next_insn: if (subprogs[cur_subprog + 1].start == i + delta + 1) { subprogs[cur_subprog].stack_depth += stack_depth_extra;diff --git a/net/core/filter.c b/net/core/filter.c index d14401193b01..cb39388f69a9 100644 --- a/net/core/filter.c +++ b/net/core/filter.c@@ -9082,8 +9082,7 @@ static int bpf_unclone_prologue(struct bpf_insn *insn_buf, bool direct_write, /* ret = bpf_skb_pull_data(skb, 0); */ *insn++ = BPF_MOV64_REG(BPF_REG_6, BPF_REG_1); *insn++ = BPF_ALU64_REG(BPF_XOR, BPF_REG_2, BPF_REG_2); - *insn++ = BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, - BPF_FUNC_skb_pull_data); + *insn++ = BPF_EMIT_CALL(bpf_skb_pull_data);
Ihor reported that the test_map has started failing on arm64. https://github.com/kernel-patches/bpf/actions/runs/21298510237/job/61311053284 https://github.com/kernel-patches/bpf/actions/runs/21298505282/job/61312363930 https://github.com/kernel-patches/bpf/actions/runs/21301695907/job/61321375157 For BPF_PROG_TYPE_SK_SKB, the BPF_FUNC_skb_pull_data has a different func_proto. It is sk_skb_pull_data instead of bpf_skb_pull_data. A different func needs to be emitted here based on prog type. Not sure why it only fails on arm64. The set has been reverted to get the CI going. Please address the issue, add a test for this case, and then respin. Thanks.