On Tue, May 17, 2022 at 03:18:35AM -0400, Xu Kuohai wrote:
>
> +static bool is_valid_bpf_tramp_flags(unsigned int flags)
> +{
> + if ((flags & BPF_TRAMP_F_RESTORE_REGS) &&
> + (flags & BPF_TRAMP_F_SKIP_FRAME))
> + return false;
> +
> + /* BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops,
> + * and it must be used alone.
> + */
> + if ((flags & BPF_TRAMP_F_RET_FENTRY_RET) &&
> + (flags & ~BPF_TRAMP_F_RET_FENTRY_RET))
> + return false;
> +
> + return true;
> +}
> +
> +int bpf_prepare_trampoline(struct bpf_tramp_image *tr, void *image,
> + void *image_end, const struct btf_func_model *m,
> + u32 flags, struct bpf_tramp_links *tlinks,
> + void *orig_call)
> +{
> + if (!is_valid_bpf_tramp_flags(flags))
> + return -EINVAL;
> +
> + return arch_prepare_bpf_trampoline(tr, image, image_end, m, flags,
> + tlinks, orig_call);
> +}
It's an overkill to introduce a new helper function just to validate
flags that almost compile time constants.
The flags are not user supplied.
Please move /* BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops ... */
comment to bpf_struct_ops.c right before it calls arch_prepare_bpf_trampoline()
And add a comment to trampoline.c saying that BPF_TRAMP_F_RESTORE_REGS
and BPF_TRAMP_F_SKIP_FRAME should not be set together.
We could add a warn_on there or in arch code, but feels like overkill.