[dpdk-dev] [PATCH] net/af_xdp: fix xdp_get_channels_info return value

Stephen Hemminger stephen at networkplumber.org
Fri Jul 12 17:48:48 CEST 2019


On Fri, 12 Jul 2019 16:35:55 +0100
Ferruh Yigit <ferruh.yigit at intel.com> wrote:

> On 7/12/2019 4:10 PM, Stephen Hemminger wrote:
> > On Fri, 12 Jul 2019 09:55:46 +0200
> > Július Milan <jmilan.dev at gmail.com> wrote:
> >   
> >> Procedure xdp_get_channels_info was returning error code -1 in case of
> >> ioctl command SIOCETHTOOL was not supported. This patch sets return
> >> value back to 0 as it is valid case.
> >>
> >> Fixes: 339b88c6a9 ("net/af_xdp: support multi-queue")
> >>
> >> Signed-off-by: Július Milan <jmilan.dev at gmail.com>
> >> ---
> >>  drivers/net/af_xdp/rte_eth_af_xdp.c | 10 +++++++---
> >>  1 file changed, 7 insertions(+), 3 deletions(-)
> >>
> >> diff --git a/drivers/net/af_xdp/rte_eth_af_xdp.c b/drivers/net/af_xdp/rte_eth_af_xdp.c
> >> index ff8e90589..33352e10a 100644
> >> --- a/drivers/net/af_xdp/rte_eth_af_xdp.c
> >> +++ b/drivers/net/af_xdp/rte_eth_af_xdp.c
> >> @@ -852,9 +852,13 @@ xdp_get_channels_info(const char *if_name, int *max_queues,
> >>  	ifr.ifr_data = (void *)&channels;
> >>  	strncpy(ifr.ifr_name, if_name, IFNAMSIZ);
> >>  	ret = ioctl(fd, SIOCETHTOOL, &ifr);
> >> -	if (ret && errno != EOPNOTSUPP) {
> >> -		ret = -errno;
> >> -		goto out;
> >> +	if (ret) {
> >> +		if (errno == EOPNOTSUPP) {
> >> +			ret = 0;
> >> +		} else {
> >> +			ret = -errno;
> >> +			goto out;
> >> +		}
> >>  	}
> >>  
> >>  	if (channels.max_combined == 0 || errno == EOPNOTSUPP) {  
> > 
> > Still wrong, it doesn't set max_queues/combined_queues if errno == EOPNOTSUPP
> >   
> 
> Looks like code only fixes the return value, which was 'ret' before, now  '0',
> the remaining logic should be same.
> 
> as far as I can see, if errno == EOPNOTSUPP, it sets
>   *max_queues = 1;
>   *combined_queues = channels.combined_count;

OK, it does return correctly, just the hard way to do it.
Keep it for now, I will do a cleanup pass over AF_XDP later.
Lots  of things could be done simpler


More information about the dev mailing list