[dpdk-dev,v2] net/i40e: fix link_state update for i40e_ethdev_vf drv
Checks
Commit Message
The check for bool was accounting unwanted bits in the calulation of truth value. In dpdk unsingned int is typedefed to bool but all it cares about is Least Significant Bit. But in calculation of condition expression the bits other than LSB was used which doesn't make sense. Some time these bits has values which results in to incorrect expression results. To fix this we just need to account LSB form the bool value . This can be easily done by anding the value with true.
Signed-off-by: Tushar Mulkar <tmulkar@sandvine.com>
---
drivers/net/i40e/i40e_ethdev_vf.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
--
2.11.0
Comments
Hi,
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Tushar Mulkar
> Sent: Wednesday 14 February 2018 12:00
> To: Zhang, Helin <helin.zhang@intel.com>; dev@dpdk.org
> Cc: Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> Subject: [dpdk-dev] [PATCH v2] net/i40e: fix link_state update for
> i40e_ethdev_vf drv
>
> The check for bool was accounting unwanted bits in the calulation of truth
> value. In dpdk unsingned int is typedefed to bool but all it cares about is Least
> Significant Bit. But in calculation of condition expression the bits other than
> LSB was used which doesn't make sense. Some time these bits has values
> which results in to incorrect expression results. To fix this we just need to
> account LSB form the bool value . This can be easily done by anding the value
> with true.
I didn't find any place where link_up is assigned to something other than Boolean.
But maybe I'm wrong or I misunderstood what you're saying.
And there is another place in this file with similar evaluation in i40evf_read_pfmsg() function.
Format of commit message should be modified as well - it is formatted in one long line
without line breaks.
>
> Signed-off-by: Tushar Mulkar <tmulkar@sandvine.com>
> ---
> drivers/net/i40e/i40e_ethdev_vf.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/i40e/i40e_ethdev_vf.c
> b/drivers/net/i40e/i40e_ethdev_vf.c
> index b96d77a0c..d23dff044 100644
> --- a/drivers/net/i40e/i40e_ethdev_vf.c
> +++ b/drivers/net/i40e/i40e_ethdev_vf.c
> @@ -2095,8 +2095,8 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
> }
> /* full duplex only */
> new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
> - new_link.link_status = vf->link_up ? ETH_LINK_UP :
> - ETH_LINK_DOWN;
> + new_link.link_status = (vf->link_up & true) ?
> + ETH_LINK_UP : ETH_LINK_DOWN;
> new_link.link_autoneg =
> dev->data->dev_conf.link_speeds &
> ETH_LINK_SPEED_FIXED;
>
> --
> 2.11.0
> -----Original Message-----
> From: Tushar Mulkar [mailto:tmulkar@sandvine.com]
> Sent: Wednesday, February 14, 2018 8:00 PM
> To: Zhang, Helin; dev@dpdk.org
> Cc: Xing, Beilei; Zhang, Qi Z
> Subject: [PATCH v2] net/i40e: fix link_state update for i40e_ethdev_vf drv
>
> The check for bool was accounting unwanted bits in the calulation of truth
> value. In dpdk unsingned int is typedefed to bool but all it cares about is Least
> Significant Bit. But in calculation of condition expression the bits other than
> LSB was used which doesn't make sense. Some time these bits has values
> which results in to incorrect expression results. To fix this we just need to
> account LSB form the bool value . This can be easily done by anding the value
> with true.
>
> Signed-off-by: Tushar Mulkar <tmulkar@sandvine.com>
> ---
> drivers/net/i40e/i40e_ethdev_vf.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/i40e/i40e_ethdev_vf.c
> b/drivers/net/i40e/i40e_ethdev_vf.c
> index b96d77a0c..d23dff044 100644
> --- a/drivers/net/i40e/i40e_ethdev_vf.c
> +++ b/drivers/net/i40e/i40e_ethdev_vf.c
> @@ -2095,8 +2095,8 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
> }
> /* full duplex only */
> new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
> - new_link.link_status = vf->link_up ? ETH_LINK_UP :
> - ETH_LINK_DOWN;
> + new_link.link_status = (vf->link_up & true) ?
> + ETH_LINK_UP : ETH_LINK_DOWN;
It is a bit strange to do bit manipunation on a bool type of value.
Do you see any wrong case with that variable? If yes, I'd suggest correct that one,
rather than doing like this.
Thanks,
Helin
> new_link.link_autoneg =
> dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED;
>
> --
> 2.11.0
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Zhang, Helin
> Sent: Wednesday, March 28, 2018 12:30 AM
> To: Tushar Mulkar; dev@dpdk.org
> Cc: Xing, Beilei; Zhang, Qi Z
> Subject: Re: [dpdk-dev] [PATCH v2] net/i40e: fix link_state update for
> i40e_ethdev_vf drv
>
>
>
> > -----Original Message-----
> > From: Tushar Mulkar [mailto:tmulkar@sandvine.com]
> > Sent: Wednesday, February 14, 2018 8:00 PM
> > To: Zhang, Helin; dev@dpdk.org
> > Cc: Xing, Beilei; Zhang, Qi Z
> > Subject: [PATCH v2] net/i40e: fix link_state update for i40e_ethdev_vf
> > drv
> >
> > The check for bool was accounting unwanted bits in the calulation of
> > truth value. In dpdk unsingned int is typedefed to bool but all it
> > cares about is Least Significant Bit. But in calculation of condition
> > expression the bits other than LSB was used which doesn't make sense.
> > Some time these bits has values which results in to incorrect
> > expression results. To fix this we just need to account LSB form the
> > bool value . This can be easily done by anding the value with true.
> >
> > Signed-off-by: Tushar Mulkar <tmulkar@sandvine.com>
> > ---
> > drivers/net/i40e/i40e_ethdev_vf.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/i40e/i40e_ethdev_vf.c
> > b/drivers/net/i40e/i40e_ethdev_vf.c
> > index b96d77a0c..d23dff044 100644
> > --- a/drivers/net/i40e/i40e_ethdev_vf.c
> > +++ b/drivers/net/i40e/i40e_ethdev_vf.c
> > @@ -2095,8 +2095,8 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
> > }
> > /* full duplex only */
> > new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
> > - new_link.link_status = vf->link_up ? ETH_LINK_UP :
> > - ETH_LINK_DOWN;
> > + new_link.link_status = (vf->link_up & true) ?
> > + ETH_LINK_UP : ETH_LINK_DOWN;
> It is a bit strange to do bit manipunation on a bool type of value.
> Do you see any wrong case with that variable? If yes, I'd suggest correct that
> one, rather than doing like this.
I'd set it to change requested for now, if there no clarification for the comments, and no ACK.
Feel free to set it back if there is anything new. Thanks!
/Helin
>
> Thanks,
> Helin
>
> > new_link.link_autoneg =
> > dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED;
> >
> > --
> > 2.11.0
@@ -2095,8 +2095,8 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
}
/* full duplex only */
new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
- new_link.link_status = vf->link_up ? ETH_LINK_UP :
- ETH_LINK_DOWN;
+ new_link.link_status = (vf->link_up & true) ?
+ ETH_LINK_UP : ETH_LINK_DOWN;
new_link.link_autoneg =
dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED;