On 1/14/2020 5:14 AM, Ajit Khaparde wrote:
> From: Somnath Kotur <somnath.ko...@broadcom.com>
> 
> Set RTE_ETH_DEV_CLOSE_REMOVE upon probe so all the private resources
> for the port can be freed by rte_eth_dev_close().
> 
> Signed-off-by: Somnath Kotur <somnath.ko...@broadcom.com>
> Reviewed-by: Kalesh Anakkur Purayil <kalesh-anakkur.pura...@broadcom.com>
> Reviewed-by: Ajit Khaparde <ajit.khapa...@broadcom.com>
> ---
>  drivers/net/bnxt/bnxt_ethdev.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
> index ba3f0a7d9..8acfade5f 100644
> --- a/drivers/net/bnxt/bnxt_ethdev.c
> +++ b/drivers/net/bnxt/bnxt_ethdev.c
> @@ -4856,6 +4856,11 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
>       if (rc)
>               goto error_free;
>  
> +     /* Pass the information to the rte_eth_dev_close() that it should also
> +      * release the private port resources.
> +      */
> +     eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
> +

When you give this flag, 'rte_eth_dev_close()' will release port resources, it
will memset the 'eth_dev->data'.

In case user first closed the port, later removed it, remove path should be safe
for it. I can see 'bnxt_dev_uninit()' operates on the values of 'eth_dev->data'
which should crash in this case.

Can you please double check the return path, all following should be safe:
- close the port
- remove the port
- close and remove the port

Reply via email to