[dpdk-dev] [PATCH] net/octeontx2: register dynamic mbuf timestamp field

Thomas Monjalon thomas at monjalon.net
Tue Nov 3 15:26:17 CET 2020


+Cc David and Olivier to make them laugh or cry.

03/11/2020 15:24, Thomas Monjalon:
> 03/11/2020 15:16, Harman Kalra:
> > A crash is observed if dynamic mbuf timestamp field is
> > registered in dev_start, as in most of the applications
> > rte_eth_timesync_enable is called after dev_start due
> > to which timestamp field did not get registered.
> 
> So you are not reading your emails?
> 
> I was waiting for you, so I looked at the ugly code of octeontx2
> with Olivier and David, and we fixed it already.
> 
> Not reading emails is wasting time of everybody.
> On the contrary, being available on IRC can speed up work.
> 
> 
> > Signed-off-by: Harman Kalra <hkalra at marvell.com>
> > --- a/drivers/net/octeontx2/otx2_ethdev.c
> > +++ b/drivers/net/octeontx2/otx2_ethdev.c
> > @@ -2219,16 +2219,6 @@ otx2_nix_dev_start(struct rte_eth_dev *eth_dev)
> >  	else
> >  		otx2_nix_timesync_disable(eth_dev);
> >  
> > -	if (dev->rx_offload_flags & NIX_RX_OFFLOAD_TSTAMP_F) {
> > -		rc = rte_mbuf_dyn_rx_timestamp_register(
> > -				&dev->tstamp.tstamp_dynfield_offset,
> > -				&dev->tstamp.rx_tstamp_dynflag);
> > -		if (rc != 0) {
> > -			otx2_err("Failed to register Rx timestamp field/flag");
> > -			return -rte_errno;
> > -		}
> > -	}
> > -
> 
> This is wrong, you still need to register for the case
> of DEV_RX_OFFLOAD_TIMESTAMP without timesync.
> 
> In my v5, it is moved below after VF special config.
> 
> >  	/* Update VF about data off shifted by 8 bytes if PTP already
> >  	 * enabled in PF owning this VF
> >  	 */
> 







More information about the dev mailing list