Feng Jiang <[email protected]> writes: > Implement arch_bpf_timed_may_goto() for the RV64 JIT. The argument and > return value are carried in BPF_REG_AX, and BPF R0-R5 are preserved > across the call to the generic bpf_check_timed_may_goto(). > > Enable bpf_jit_supports_timed_may_goto() so the verifier uses the timed > expansion path.
Thanks for working on this! > diff --git a/arch/riscv/net/bpf_jit_comp64.c b/arch/riscv/net/bpf_jit_comp64.c > index c03c1de16b79..4f88a446278a 100644 > --- a/arch/riscv/net/bpf_jit_comp64.c > +++ b/arch/riscv/net/bpf_jit_comp64.c > @@ -1841,7 +1841,13 @@ int bpf_jit_emit_insn(const struct bpf_insn *insn, > struct rv_jit_context *ctx, > if (ret) > return ret; > > - if (insn->src_reg != BPF_PSEUDO_CALL) > + /* > + * arch_bpf_timed_may_goto() is emitted by the verifier and > + * returns its result in BPF_REG_AX instead of BPF_REG_0, so > + * skip the normal "move return register into R0". > + */ > + if (insn->src_reg != BPF_PSEUDO_CALL && > + addr != (u64)arch_bpf_timed_may_goto) Nit: In the future, please use full 100 chars lines, instead of breaking the line. Would it be possible to add a test for this custom ABI? The fact that it was missed in v1 hints that it would be good. I asked an LLM to suggest a test [1]. Feel free to be inspired by it. ;-) Björn [1] https://paste.debian.net/hidden/9dd769d6
