[dpdk-dev] [PATCH v5] net/ixgbe: fix l3fwd start failed on
Wu, Yanglong
yanglong.wu at intel.com
Tue Jan 9 03:29:40 CET 2018
Hi, Konstantin
Thanks for your comments!
Do you means that tx_q is must less than rx_q when the SRIOV is active and if not, the application case will not be supported?
Do you think my patch will cause to 2 (or more cores) could try to TX packets through the same TX queue? And as far as I know, the way of core using tx_q queue is depend on the application (e.g. in l3fwd tx_q equal to number of core) and multi core use same tx_q is not suggested for locker is needed in this situation. So why do you think my patch will lead to multi core using same queue?
Yanglong Wu
-----Original Message-----
From: Ananyev, Konstantin
Sent: Monday, January 8, 2018 7:55 PM
To: Wu, Yanglong <yanglong.wu at intel.com>; dev at dpdk.org
Subject: RE: [PATCH v5] net/ixgbe: fix l3fwd start failed on
> -----Original Message-----
> From: Wu, Yanglong
> Sent: Monday, January 8, 2018 3:06 AM
> To: dev at dpdk.org
> Cc: Ananyev, Konstantin <konstantin.ananyev at intel.com>; Wu, Yanglong
> <yanglong.wu at intel.com>
> Subject: [PATCH v5] net/ixgbe: fix l3fwd start failed on
>
> L3fwd start failed on PF, for tx_q check failed.
> That occurred when the SRIOV is active and tx_q > rx_q.
> The tx_q is equal to nb_q_per_pool. The number of nb_q_per_pool should
> equeal to max number of queues supported by HW not nb_rx_q.
But then 2 (or more cores) could try to TX packets through the same TX queue?
Why not just fil to start gracefully (call rte_exit() or so) if such situation occurred?
Konstantin
>
> Fixes: 27b609cbd1c6 (ethdev: move the multi-queue mode check to
> specific drivers)
>
> Signed-off-by: Yanglong Wu <yanglong.wu at intel.com>
> ---
> v5:
> Rework according to comments
> ---
> drivers/net/ixgbe/ixgbe_ethdev.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c
> b/drivers/net/ixgbe/ixgbe_ethdev.c
> index ff19a564a..baaeee5d9 100644
> --- a/drivers/net/ixgbe/ixgbe_ethdev.c
> +++ b/drivers/net/ixgbe/ixgbe_ethdev.c
> @@ -95,6 +95,9 @@
> /* Timer value included in XOFF frames. */ #define IXGBE_FC_PAUSE
> 0x680
>
> +/*Default value of Max Rx Queue*/
> +#define IXGBE_MAX_RX_QUEUE_NUM 128
> +
> #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
> #define IXGBE_LINK_UP_CHECK_TIMEOUT 1000 /* ms */
> #define IXGBE_VMDQ_NUM_UC_MAC 4096 /* Maximum nb. of UC MAC addr. */
> @@ -2194,9 +2197,10 @@ ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
> return -EINVAL;
> }
>
> - RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = nb_rx_q;
> - RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx = pci_dev->max_vfs * nb_rx_q;
> -
> + RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
> + IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
> + RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
> + pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
> return 0;
> }
>
> --
> 2.11.0
More information about the dev
mailing list