Re: bpf_errno. Was: [PATCH RFC bpf-next 1/3] bpf: report probe fault to BPF stderr

On 2025/10/9 22:45, Alexei Starovoitov wrote:

On Thu, Oct 9, 2025 at 7:15 AM Leon Hwang [off-list ref] wrote:

quoted

The verifier can rewrite 'bpf_reg_aux()' into the following instructions:

dst_reg = BPF_REG_AUX;
BPF_REG_AUX = 0; /* clear BPF_REG_AUX */

As for the architecture-specific implementation, BPF_REG_AUX can be
mapped to an appropriate register per arch — for example, r11 on x86_64.

it's taken. There are no free registers.

Understood.

It would certainly be beneficial if there were available registers on
x86_64, as that would enable certain optimizations and improvements.

In a similar direction, I have been exploring the idea of introducing a
dedicated BPF_REG_TAIL_CALL register to unify the handling of
tail_call_cnt in the verifier. This could help standardize the logic
across architectures, particularly for those that already employ a
dedicated register for tail calls, and allow JIT backends to simplify
their tail call implementations accordingly.

Thanks,
Leon