On Tue, Jun 13, 2023 at 7:48 AM Willem de Bruijn
<willemdebruijn.kernel@xxxxxxxxx> wrote:
>
> On Mon, Jun 12, 2023 at 7:26 PM Stanislav Fomichev <sdf@xxxxxxxxxx> wrote:
> >
> > Attach kfuncs that request and report TX timestamp via ringbuf.
> > Confirm on the userspace side that the program has triggered
> > and the timestamp is non-zero.
> >
> > Also make sure devtx_frame has a sensible pointers and data.
> >
> > Cc: netdev@xxxxxxxxxxxxxxx
> > Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
>
> > +SEC("fentry/devtx_sb")
> > +int BPF_PROG(devtx_sb, const struct devtx_frame *frame)
> > +{
> > + int ret;
> > +
> > + ret = verify_frame(frame);
> > + if (ret < 0)
> > + __sync_add_and_fetch(&pkts_fail_tx, 1);
>
> intend to return in these error cases? in this and following patch
Yeah, let's do it.
> > +
> > + ret = bpf_devtx_sb_request_timestamp(frame);
> > + if (ret < 0)
> > + __sync_add_and_fetch(&pkts_fail_tx, 1);
> > +
> > + return 0;
> > +}
> > +
> > +SEC("fentry/devtx_cp")
> > +int BPF_PROG(devtx_cp, const struct devtx_frame *frame)
> > +{
> > + struct devtx_sample *sample;
> > + int ret;
> > +
> > + ret = verify_frame(frame);
> > + if (ret < 0)
> > + __sync_add_and_fetch(&pkts_fail_tx, 1);
> > +
> > + sample = bpf_ringbuf_reserve(&tx_compl_buf, sizeof(*sample), 0);
> > + if (!sample)
> > + return 0;
>
> return non-zero?
Can't return non-zero from fentry :-( The userspace verifies that
pkts_fail_tx stays zero, so that should be enough.
> > +
> > + sample->timestamp_retval = bpf_devtx_cp_timestamp(frame, &sample->timestamp);
> > +
> > + bpf_ringbuf_submit(sample, 0);
> > +
> > + return 0;
> > +}
