Re: [dpdk-dev] [PATCH v4 3/5] net/tap: fix check for mbuf's nb_segs failure

[wangyunjian <wangyunjian@huawei.com>]

> -----Original Message-----
> From: Ferruh Yigit [mailto:ferruh.yigit@intel.com]
> Sent: Wednesday, April 15, 2020 11:13 PM
> To: wangyunjian <wangyunjian@huawei.com>; dev@dpdk.org
> Cc: keith.wiles@intel.com; Lilijun (Jerry) <jerry.lilijun@huawei.com>; xudingke
> <xudingke@huawei.com>; stable@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v4 3/5] net/tap: fix check for mbuf's nb_segs
> failure
> 
> On 4/11/2020 12:23 PM, wangyunjian wrote:
> > From: Yunjian Wang <wangyunjian@huawei.com>
> >
> > Now the rxq->pool is mbuf concatenation, but its nb_segs is 1. When
> > conducting some sanity checks on the mbuf with debug enabled, it fails.
> >
> > Fixes: 0781f5762cfe ("net/tap: support segmented mbufs")
> > CC: stable@dpdk.org
> >
> > Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
> > ---
> >  drivers/net/tap/rte_eth_tap.c | 21 +++++++++++++++++----
> >  1 file changed, 17 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/net/tap/rte_eth_tap.c
> > b/drivers/net/tap/rte_eth_tap.c index 0156d689d..6a77b2a7e 100644
> > --- a/drivers/net/tap/rte_eth_tap.c
> > +++ b/drivers/net/tap/rte_eth_tap.c
> > @@ -339,6 +339,19 @@ tap_rx_offload_get_queue_capa(void)
> >  	       DEV_RX_OFFLOAD_TCP_CKSUM;
> >  }
> >
> > +static void
> > +tap_rxq_pool_free(struct rte_mbuf *pool) {
> > +	struct rte_mbuf *next;
> > +
> > +	while (pool) {
> > +		next = pool->next;
> > +		pool->next = NULL;
> > +		rte_pktmbuf_free(pool);
> > +		pool = next;
> > +	}
> > +}
> 
> I am aware I have suggested this but I have missed that 'rte_mbuf_check()' still
> may fail.
> 
> The 'rxq->pool' is a set of linked mbufs, each mbuf->next points to next one.
> But all mbufs in the pool has 'nb_segs' as '1'. As far as I can see from code this
> will cause a warning in 'rte_mbuf_check()'. If you can reproduce it you can
> double check.
> 
> Your initial implementation seems the correct one, to fix the nb_segs for first
> mbuf in the pool, sorry for the noise.

OK, I can reproduce and test it. I will using initial implementation in next version.

Thanks,
Yunjian

> 
> > +
> >  /* Callback to handle the rx burst of packets to the correct interface and
> >   * file descriptor(s) in a multi-queue setup.
> >   */
> > @@ -389,7 +402,7 @@ pmd_rx_burst(void *queue, struct rte_mbuf **bufs,
> uint16_t nb_pkts)
> >  					goto end;
> >
> >  				seg->next = NULL;
> > -				rte_pktmbuf_free(mbuf);
> > +				tap_rxq_pool_free(mbuf);
> >
> >  				goto end;
> >  			}
> > @@ -1038,7 +1051,7 @@ tap_dev_close(struct rte_eth_dev *dev)
> >  			rxq = &internals->rxq[i];
> >  			close(process_private->rxq_fds[i]);
> >  			process_private->rxq_fds[i] = -1;
> > -			rte_pktmbuf_free(rxq->pool);
> > +			tap_rxq_pool_free(rxq->pool);
> >  			rte_free(rxq->iovecs);
> >  			rxq->pool = NULL;
> >  			rxq->iovecs = NULL;
> > @@ -1077,7 +1090,7 @@ tap_rx_queue_release(void *queue)
> >  	if (process_private->rxq_fds[rxq->queue_id] > 0) {
> >  		close(process_private->rxq_fds[rxq->queue_id]);
> >  		process_private->rxq_fds[rxq->queue_id] = -1;
> > -		rte_pktmbuf_free(rxq->pool);
> > +		tap_rxq_pool_free(rxq->pool);
> >  		rte_free(rxq->iovecs);
> >  		rxq->pool = NULL;
> >  		rxq->iovecs = NULL;
> > @@ -1485,7 +1498,7 @@ tap_rx_queue_setup(struct rte_eth_dev *dev,
> >  	return 0;
> >
> >  error:
> > -	rte_pktmbuf_free(rxq->pool);
> > +	tap_rxq_pool_free(rxq->pool);
> >  	rxq->pool = NULL;
> >  	rte_free(rxq->iovecs);
> >  	rxq->iovecs = NULL;
> >