From: Havlík Martin
> Dne 2021-06-23 09:04, Min Hu (Connor) napsal:
> > 在 2021/6/22 17:25, Martin Havlik 写道:
> >> When dedicated queues are enabled, mlx5 PMD fails to install RTE
> >> Flows if the underlying ethdev is not started:
> >> bond_ethdev_8023ad_flow_set(267) - bond_ethdev_8023ad_flow_set:
> port
> >> not started (slave_port=0 queue_id=1)
> >>
> > Why mlx5 PMD doing flow create relys on port started ?
> > I noticed other PMDs did not has that reliance.
> >
> After looking into it, I really can't answer this mlx5 centered question.
> Closest related info we found so far is the 5th point in
> https://doc.dpdk.org/guides/prog_guide/rte_flow.html#caveats
> but it only specifies it's the application's responsibility and that flow
> rules are
> assumed invalid after port stop/close/restart but doesn't say anything about
> <stop - flow rule create - start> vs <stop - start - flow rule create> where
> the
> former is the point of failure on mlx5.
> I'm addressing the maintainers for mlx5, who might know a bit more on the
> topic.
From rte_ethdev.h
* Please note that some configuration is not stored between calls to
* rte_eth_dev_stop()/rte_eth_dev_start(). The following configuration will
* be retained:
*
* - MTU
* - flow control settings
* - receive mode configuration (promiscuous mode, all-multicast mode,
* hardware checksum mode, RSS/VMDQ settings etc.)
* - VLAN filtering configuration
* - default MAC address
* - MAC addresses supplied to MAC address array
* - flow director filtering mode (but not filtering rules)
* - NIC queue statistics mappings
Mlx5 assumes flows are allowed to be configured only after rte_eth_dev_start().
Before start \ after stop - no flow is valid anymore.
Matan
> >> Signed-off-by: Martin Havlik <xhavl...@stud.fit.vutbr.cz>
> >> Cc: Jan Viktorin <vikto...@cesnet.cz>
> >> ---
> >> drivers/net/bonding/rte_eth_bond_pmd.c | 26
> >> ++++++++++++++++++--------
> >> 1 file changed, 18 insertions(+), 8 deletions(-)
> >>
> >> diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c
> >> b/drivers/net/bonding/rte_eth_bond_pmd.c
> >> index a6755661c..fea3bc537 100644
> >> --- a/drivers/net/bonding/rte_eth_bond_pmd.c
> >> +++ b/drivers/net/bonding/rte_eth_bond_pmd.c
> >> @@ -1818,25 +1818,35 @@ slave_configure(struct rte_eth_dev
> >> *bonded_eth_dev,
> >> rte_flow_destroy(slave_eth_dev->data->port_id,
> >>
> >> internals-
> >mode4.dedicated_queues.flow[slave_eth_dev->data->port_id],
> >> &flow_error);
> >> + }
> >> + /* Start device */
> >> + errval = rte_eth_dev_start(slave_eth_dev->data->port_id);
> >> + if (errval != 0) {
> >> + RTE_BOND_LOG(ERR, "rte_eth_dev_start: port=%u, err (%d)",
> >> + slave_eth_dev->data->port_id, errval);
> >> + return -1;
> >> + }
> >> +
> >> + if (internals->mode == BONDING_MODE_8023AD &&
> >> + internals->mode4.dedicated_queues.enabled == 1)
> >> + {
> >> errval = bond_ethdev_8023ad_flow_set(bonded_eth_dev,
> >> slave_eth_dev->data->port_id);
> >> if (errval != 0) {
> >> RTE_BOND_LOG(ERR,
> >> "bond_ethdev_8023ad_flow_set: port=%d, err
> >> (%d)",
> >> slave_eth_dev->data->port_id, errval);
> >> +
> >> + errval =
> >> rte_eth_dev_stop(slave_eth_dev->data->port_id);
> >> + if (errval < 0) {
> >> + RTE_BOND_LOG(ERR,
> >> + "rte_eth_dev_stop: port=%d, err (%d)",
> >> + slave_eth_dev->data->port_id, errval);
> >> + }
> >> return errval;
> >> }
> >> }
> >> - /* Start device */
> >> - errval = rte_eth_dev_start(slave_eth_dev->data->port_id);
> >> - if (errval != 0) {
> >> - RTE_BOND_LOG(ERR, "rte_eth_dev_start: port=%u, err (%d)",
> >> - slave_eth_dev->data->port_id, errval);
> >> - return -1;
> >> - }
> >> -
> >> /* If RSS is enabled for bonding, synchronize RETA */
> >> if (bonded_eth_dev->data->dev_conf.rxmode.mq_mode &
> >> ETH_MQ_RX_RSS) {
> >> int i;
> >>
