> -----Original Message-----
> From: Yigit, Ferruh
> Sent: Monday, April 24, 2017 1:47 PM
> To: Legacy, Allain (Wind River); Zhang, Helin; Wu, Jingjing
> Cc: dev@dpdk.org; Peters, Matt (Wind River)
> Subject: Re: [dpdk-dev] [PATCH v2] net/i40e: mbuf alloc failed counter not
> incremented
Actually it is a fix, right? I think you need to start the title with 'fix'?
Then don't miss the 'Fix' line, as required.
>
> On 4/22/2017 12:13 AM, Allain Legacy wrote:
> > From: Matt Peters <matt.pet...@windriver.com>
> >
> > When an mbuf alloc fails during the mempool get operation for the i40e
> > bulk alloc receive function, the rx_mbuf_alloc_failed counter is not
> > incremented to record the error.
> >
> > This fix ensures consistency with the other i40e receive procedures
> > and other net drivers.
> >
> > Signed-off-by: Matt Peters <matt.pet...@windriver.com>
> > Signed-off-by: Allain Legacy <allain.leg...@windriver.com>
> > ---
> > drivers/net/i40e/i40e_rxtx.c | 27 ++++++++++++++++++++++++---
> > 1 file changed, 24 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/net/i40e/i40e_rxtx.c
> > b/drivers/net/i40e/i40e_rxtx.c index e5471b143..4131902a9 100644
> > --- a/drivers/net/i40e/i40e_rxtx.c
> > +++ b/drivers/net/i40e/i40e_rxtx.c
> > @@ -610,6 +610,7 @@ static inline uint16_t rx_recv_pkts(void
> > *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) {
> > struct i40e_rx_queue *rxq = (struct i40e_rx_queue *)rx_queue;
> > + struct rte_eth_dev *dev;
> > uint16_t nb_rx = 0;
> >
> > if (!nb_pkts)
> > @@ -630,6 +631,11 @@ rx_recv_pkts(void *rx_queue, struct rte_mbuf
> **rx_pkts, uint16_t nb_pkts)
> > PMD_RX_LOG(DEBUG, "Rx mbuf alloc failed for "
> > "port_id=%u, queue_id=%u",
> > rxq->port_id, rxq->queue_id);
> > +
> > + dev = &rte_eth_devices[rxq->port_id];
>
> The question is mostly to the driver maintainer, instead of using global
> variable,
> would it be better to use rxq->vsi->adapter->eth_dev to access rte_eth_device
> struct?
Yes, I agree with Ferruh. I40E_VSI_TO_ETH_DEV(rxq->vsi) can be used here.
>
> > + dev->data->rx_mbuf_alloc_failed +=
> > + rxq->rx_free_thresh;
> > +
> > rxq->rx_nb_avail = 0;
> > rxq->rx_tail = (uint16_t)(rxq->rx_tail - nb_rx);
> > for (i = 0, j = rxq->rx_tail; i < nb_rx; i++, j++) @@
> > -691,6
> > +697,7 @@ i40e_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
> uint16_t nb_pkts)
> > union i40e_rx_desc rxd;
> > struct i40e_rx_entry *sw_ring;
> > struct i40e_rx_entry *rxe;
> > + struct rte_eth_dev *dev;
> > struct rte_mbuf *rxm;
> > struct rte_mbuf *nmb;
> > uint16_t nb_rx;
> > @@ -721,10 +728,16 @@ i40e_recv_pkts(void *rx_queue, struct rte_mbuf
> **rx_pkts, uint16_t nb_pkts)
> > break;
> >
> > nmb = rte_mbuf_raw_alloc(rxq->mp);
> > - if (unlikely(!nmb))
> > + if (unlikely(!nmb)) {
> > + PMD_RX_LOG(DEBUG, "RX mbuf alloc failed
> port_id=%u "
> > + "queue_id=%u", (unsigned int)rxq->port_id,
> > + (unsigned int)rxq->queue_id);
>
> Do we really want debug print here?
> When you think the speeds we are dealing with, if mbuf alloc starts failing we
> may hit this lines millions per second, which may make app unusable?
>
> > + dev = &rte_eth_devices[rxq->port_id];
> > + dev->data->rx_mbuf_alloc_failed++;
> > break;
> > - rxd = *rxdp;
> > + }
> >
> > + rxd = *rxdp;
> > nb_hold++;
> > rxe = &sw_ring[rx_id];
> > rx_id++;
> > @@ -816,6 +829,7 @@ i40e_recv_scattered_pkts(void *rx_queue,
> > struct rte_mbuf *nmb, *rxm;
> > uint16_t rx_id = rxq->rx_tail;
> > uint16_t nb_rx = 0, nb_hold = 0, rx_packet_len;
> > + struct rte_eth_dev *dev;
> > uint32_t rx_status;
> > uint64_t qword1;
> > uint64_t dma_addr;
> > @@ -833,8 +847,15 @@ i40e_recv_scattered_pkts(void *rx_queue,
> > break;
> >
> > nmb = rte_mbuf_raw_alloc(rxq->mp);
> > - if (unlikely(!nmb))
> > + if (unlikely(!nmb)) {
> > + PMD_RX_LOG(DEBUG, "RX mbuf alloc failed
> port_id=%u "
> > + "queue_id=%u", (unsigned int)rxq->port_id,
> > + (unsigned int)rxq->queue_id);
> > + dev = &rte_eth_devices[rxq->port_id];
> > + dev->data->rx_mbuf_alloc_failed++;
> > break;
> > + }
> > +
> > rxd = *rxdp;
> > nb_hold++;
> > rxe = &sw_ring[rx_id];
> >
