On Thu, Feb 9, 2023 at 5:29 PM Kevin Traynor <ktray...@redhat.com> wrote: > > By default OVS configures 2048 descriptors for tx and rx queues > on DPDK devices. It also allows the user to configure those values. > > If the values used are not acceptable to the device than queue setup > will fail. > > The device exposes it's max/min/alignment requirements, so use those > to ensure that an acceptable value is used during queue setup. > > If the default or user value is not acceptable, adjust to a suitable > value. > > Reported-at: https://bugzilla.redhat.com/2119876 > Signed-off-by: Kevin Traynor <ktray...@redhat.com> > --- > lib/netdev-dpdk.c | 12 ++++++++++++ > 1 file changed, 12 insertions(+) > > diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c > index fb0dd43f7..e901f857e 100644 > --- a/lib/netdev-dpdk.c > +++ b/lib/netdev-dpdk.c > @@ -1008,4 +1008,13 @@ dpdk_watchdog(void *dummy OVS_UNUSED) > } > > +static int > +dpdk_limit_desc_size(int desc_size, struct rte_eth_desc_lim *lim) > +{ > + desc_size = ROUND_UP(desc_size, lim->nb_align);
Based on a quick grep, it seems possible a driver exposes nb_align == 0. Like for example: https://git.dpdk.org/dpdk/tree/drivers/net/bnxt/bnxt_ethdev.c#n1020 > + desc_size = MIN(desc_size, lim->nb_max); > + desc_size = MAX(desc_size, lim->nb_min); > + return desc_size; Automatically adjusting a default value seems normal to me. But in the case of user input, I think it would help if there was some log indicating that OVS adjusted the config. > +} > + > static int > dpdk_eth_dev_port_config(struct netdev_dpdk *dev, int n_rxq, int n_txq) > @@ -1056,4 +1065,7 @@ dpdk_eth_dev_port_config(struct netdev_dpdk *dev, int > n_rxq, int n_txq) > } > > + dev->rxq_size = dpdk_limit_desc_size(dev->rxq_size, &info.rx_desc_lim); > + dev->txq_size = dpdk_limit_desc_size(dev->txq_size, &info.tx_desc_lim); > + > /* A device may report more queues than it makes available (this has > * been observed for Intel xl710, which reserves some of them for > -- > 2.39.1 > > _______________________________________________ > dev mailing list > d...@openvswitch.org > https://mail.openvswitch.org/mailman/listinfo/ovs-dev > -- David Marchand _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
