Thread (13 messages) 13 messages, 3 authors, 2016-10-13

Re: [PATCH] app/testpmd: fix DCB config issue on ixgbe

From: Lu, Wenzhuo <hidden>
Date: 2016-08-29 02:42:57

Hi Bernard,

-----Original Message-----
From: Iremonger, Bernard
Sent: Friday, August 26, 2016 6:04 PM
To: Lu, Wenzhuo; dev@dpdk.org
Cc: De Lara Guarch, Pablo; Wu, Jingjing
Subject: RE: [dpdk-dev] [PATCH] app/testpmd: fix DCB config issue on ixgbe

Hi Wenzhuo,

<snip>
quoted
quoted
quoted
quoted
If nb_rxq and nb_txq are set to max_rx_queues and max_tx_queues
respectively, there is a failure when the port is started in
ixgbe_check_mq_mode() at line
1990 in ixgbe_ethdev.c.
SRIOV is active, nb_rx_q=128 nb_tx_q=128 queue number must be
less than or equal to 1.
I haven't hit this issue. Would you like to give more details
about how to hit it? I'll check if I miss something.
There is a Niantic PF and VF bound to igb_uio. Port 0 is the PF and
Port 1 is the VF.
./testpmd -c 3f -l 1-5 -n 4 -- -i
testpmd> set corelist 2,3,4,5
testpmd> port stop 0  /* PF is 0 */
testpmd> port config 0 dcb vt on 4 pfc on port start 0 /* PF is 0 */
line 1990   ixgbe_ethdev.c
SRIOV is active, nb_rx_q=128 nb_tx_q=128 queue number must be less
than or equal to 1.
/* Works if nb_rx_q and nb_tx_q set to 1 */
To my opinion, it's a by-design limitation. After using the DCB
configuration CLI, the queue number is set to a fix number which is
the max number. But as you pointed, when SRIOV is active there's
another requirement for the queue number.
We need to investigate deeper and find a solution for it. But I think
it's another story. We need another patch for it.
Line 1997 in testpmd.c
1997	if (dcb_mode == DCB_VT_ENABLED) {
		nb_rxq = rte_port->dev_info.max_rx_queues;
		nb_txq = rte_port->dev_info.max_tx_queues;
	} else {
2001		/*if vt is disabled, use all pf queues */
		if (rte_port->dev_info.vmdq_pool_base == 0) {
			nb_rxq = rte_port->dev_info.max_rx_queues;
			nb_txq = rte_port->dev_info.max_tx_queues;
		} else {
			nb_rxq = (queueid_t)num_tcs;
			nb_txq = (queueid_t)num_tcs;
		}
	}

The comment at line 2001 implies that when dcb_mode is DCB_VT_ENABLED all
pf queues should not be used.
When dcb_mode is DCB_VT_ENABLED, setting nb_rxq and nb_txq equal to 1
works when the PF (port 0) is started.
When dcb_mode is DCB_VT_ENABLED, setting nb_rxq to max_rx_queues and
nb_txq to max_tx_queues results in the following failure in ixgbe when the PF
(port 0) is started.
SRIOV is active, nb_rx_q=128 nb_tx_q=128 queue number must be less than or
equal to 1.
PF and its VF port cannot be used in the same APP. As VF depends on its PF. We must wait until PF is completed, then run VF. Normally, like use DPDK PF on host, then use DPDK VF on guest. If you use PF and VF on parallel, the behavior is unpredicted. I think that's why you see this prompt " SRIOV is active, nb_rx_q=128 nb_tx_q=128 queue number must be less than or equal to 1.".
I've tried PF only testpmd. The result is like " PMD: ixgbe_check_mq_mode(): SRIOV active, unsupported mq_mode rx 6.". It means DCB is not supported by DPDK PF.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help