Re: [PATCH bpf-next 13/15] xsk: support for Tx

["Björn Töpel" <bjorn.topel@gmail.com>]

2018-04-25 21:00 GMT+02:00 Willem de Bruijn <willemdebruijn.kernel@gmail.com>:
[...]
>>> static inline struct xdp_desc *xskq_peek_desc(struct xsk_queue *q,
>>> +                                             struct xdp_desc *desc)
>>> +{
>>> +       struct xdp_rxtx_ring *ring;
>>> +
>>> +       if (q->cons_tail == q->cons_head) {
>>> +               WRITE_ONCE(q->ring->consumer, q->cons_tail);
>>> +               q->cons_head = q->cons_tail + xskq_nb_avail(q, RX_BATCH_SIZE);
>>> +
>>> +               /* Order consumer and data */
>>> +               smp_rmb();
>>> +
>>> +               return xskq_validate_desc(q, desc);
>>> +       }
>>> +
>>> +       ring = (struct xdp_rxtx_ring *)q->ring;
>>> +       *desc = ring->desc[q->cons_tail & q->ring_mask];
>>> +       return desc;
>>>
>>> This only validates descriptors if taking the branch.
>>
>> Yes, that is because we only want to validate the descriptors once
>> even if we call this function multiple times for the same entry.
>
> Then I am probably misreading this function. But isn't head increased
> by up to RX_BATCH_SIZE frames at once. If so, then for many frames
> the branch is not taken.

You're not misreading it! :-) The head is indeed increased, but only
the tail descriptor is validated in that function. Later in the
xskq_discard_desc function when the tail is moved, the next descriptor
is validated. So, the peek function will always return a validated
descriptor, but the validation can be done in either peek or discard.


Björn