Re: [PATCH] net/intel: fix igb tx queue offloads capability

[Bruce Richardson <bruce.richardson@intel.com>]

On Wed, May 07, 2025 at 09:59:04AM +0100, Deng, KaiwenX wrote:
> 
> 
> > -----Original Message-----
> > From: Richardson, Bruce <bruce.richardson@intel.com>
> > Sent: Monday, April 28, 2025 11:21 PM
> > To: Deng, KaiwenX <kaiwenx.deng@intel.com>
> > Cc: dev@dpdk.org; stable@dpdk.org; Zhang, Qi Z <qi.z.zhang@intel.com>;
> > Yaroslav Brustinov <ybrustin@cisco.com>
> > Subject: Re: [PATCH] net/intel: fix igb tx queue offloads capability
> > 
> > On Mon, Apr 14, 2025 at 04:53:52PM +0800, Kaiwen Deng wrote:
> > > The igb driver assigns the tx port offload capability to the tx queue
> > > offload capability.
> > >
> > > This commit will fix this issue.
> > >
> > > Fixes: daa3b0833f08 ("net/e1000: fix Tx offload capability typos")
> > > Cc: stable@dpdk.org
> > >
> > > Signed-off-by: Kaiwen Deng <kaiwenx.deng@intel.com>
> > > ---
> > >  drivers/net/intel/e1000/igb_rxtx.c | 6 ++----
> > >  1 file changed, 2 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/net/intel/e1000/igb_rxtx.c
> > > b/drivers/net/intel/e1000/igb_rxtx.c
> > > index 4276bb6d31..b63de2354f 100644
> > > --- a/drivers/net/intel/e1000/igb_rxtx.c
> > > +++ b/drivers/net/intel/e1000/igb_rxtx.c
> > > @@ -1490,11 +1490,9 @@ igb_get_tx_port_offloads_capa(struct
> > > rte_eth_dev *dev)  uint64_t  igb_get_tx_queue_offloads_capa(struct
> > > rte_eth_dev *dev)  {
> > > -	uint64_t tx_queue_offload_capa;
> > > -
> > > -	tx_queue_offload_capa = igb_get_tx_port_offloads_capa(dev);
> > > +	RTE_SET_USED(dev);
> > >
> > > -	return tx_queue_offload_capa;
> > > +	return 0;
> > >  }
> > >
> > 
> > I suspect there may be more instances in the code where we do this. For
> > example, I see the same pattern used in em_rxtx.c in e1000. Can you quickly
> > check our other drivers to see if the same mistake is repeated elsewhere other
> > than "igb" and "em"?
> Hi Bruce,
> After checking, the same error is not repeated anywhere else except on "igb" and "em" .
> However, I noticed that there is only one txq available for the "em" device, and according 
> to the comments in the code, This design may be intentional, as port offloading and queue 
> offloading are practically the same in the case of only one txq.

Ok, thanks for investigating and explaining. This patch seems fine so.

/Bruce