On Tue, 30 Aug 2022 23:02:57 +0300
Shmulik Ladkani <shmulik@xxxxxxxxxxxxxxxx> wrote:
> On Thu, 25 Aug 2022 11:20:31 -0700
> Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote:
>
> > > + * long bpf_skb_set_tunnel_opt_dynptr(struct sk_buff *skb, struct bpf_dynptr *opt, u32 len)
> >
> > why can't we rely on dynptr's len instead of specifying extra one
> > here? dynptr is a range of memory, so just specify that you take that
> > entire range?
> >
> > And then we'll have (or partially already have) generic dynptr helpers
> > to adjust internal dynptr offset and len.
>
> Alright.
>
> For my usecase I need to use *part* of the tunnel options that were
> previously stored as a dynptr.
>
> Therefore I need to introduce a new bpf helper that adjusts the dynptr.
>
> How about this suggestion (sketch, not yet tried):
>
> // adjusts the dynptr to point to *len* bytes starting from the
> // specified *offset*
>
> BPF_CALL_3(bpf_dynptr_slice, struct bpf_dynptr_kern *, ptr, u32, offset, u32, len)
> {
> int err;
> u32 size;
>
> if (!ptr->data)
> return -EINVAL;
>
> err = bpf_dynptr_check_off_len(ptr, offset, len);
> if (err)
> return err;
>
> ptr->offset += offset;
> size = bpf_dynptr_get_size(ptr) - len;
> ptr->size = (ptr->size & ~(u32)DYNPTR_SIZE_MASK) | size;
> return 0;
> }
>
Correction, meant:
ptr->offset += offset;
ptr->size = (ptr->size & ~(u32)DYNPTR_SIZE_MASK) | len;
