On Fri, Oct 16, 2015 at 10:06:13AM +0200, Nicolas Morey-Chaisemartin wrote: > Before sending/receiving packet, check that the packet_len configured > is supported by the pktios MTU. > This allows interface that do not support jumbo frame to pass the test. > However it has the bad side effect of not thoroughly checking the > pktio if PKT_LEN_NORMAL is smaller than the MTU.
This test does a similar thing; http://patches.opendataplane.org/patch/3294/ But the test is reported as being inactive rather than just not running it and reporting it as passed. It depends on the ODP_CUNIT_INFO_CONDITIONAL series that was merged last week; http://patches.opendataplane.org/patch/3273/ With that you could make pktio_test_jumbo conditional and leave all non-jumbo tests unchanged. > > Signed-off-by: Nicolas Morey-Chaisemartin <nmo...@kalray.eu> > --- > test/validation/pktio/pktio.c | 16 ++++++++++++---- > 1 file changed, 12 insertions(+), 4 deletions(-) > > diff --git a/test/validation/pktio/pktio.c b/test/validation/pktio/pktio.c > index 5c0799b..8fadbf7 100644 > --- a/test/validation/pktio/pktio.c > +++ b/test/validation/pktio/pktio.c > @@ -455,6 +455,7 @@ static void test_txrx(odp_queue_type_t q_type, int > num_pkts) > int ret, i, if_b; > pktio_info_t pktios[MAX_NUM_IFACES]; > pktio_info_t *io; > + uint32_t mtu, min_mtu = UINT32_MAX; > > /* create pktios and associate input/output queues */ > for (i = 0; i < num_ifaces; ++i) { > @@ -475,12 +476,19 @@ static void test_txrx(odp_queue_type_t q_type, int > num_pkts) > > ret = odp_pktio_start(io->id); > CU_ASSERT(ret == 0); > + > + mtu = odp_pktio_mtu(io->id); > + if (mtu < min_mtu) > + min_mtu = mtu; > } > > - /* if we have two interfaces then send through one and receive on > - * another but if there's only one assume it's a loopback */ > - if_b = (num_ifaces == 1) ? 0 : 1; > - pktio_txrx_multi(&pktios[0], &pktios[if_b], num_pkts); > + /* Skip test if packet len is larger than the MTU */ > + if (min_mtu >= packet_len) { > + /* if we have two interfaces then send through one and receive > + * on another but if there's only one assume it's a loopback */ > + if_b = (num_ifaces == 1) ? 0 : 1; > + pktio_txrx_multi(&pktios[0], &pktios[if_b], num_pkts); > + } > > for (i = 0; i < num_ifaces; ++i) { > destroy_inq(pktios[i].id); > -- > 2.6.1.3.g8d02103 > _______________________________________________ lng-odp mailing list lng-odp@lists.linaro.org https://lists.linaro.org/mailman/listinfo/lng-odp
