On 5/15/2018 9:41 AM, Andrew Rybchenko wrote:
> On 05/15/2018 07:10 PM, Qiming Yang wrote:
>> This patch fixed error function call order in rte_eth_dev_configure.
>> This wrong order makes user can't get correct device informations,
>> and cause DCB configure issue.
>>
>> Fixes: 3be82f5cc5e3 ("ethdev: support PMD-tuned Tx/Rx parameters")
>>
>> Signed-off-by: Qiming Yang <qiming.y...@intel.com>
>> ---
>> lib/librte_ethdev/rte_ethdev.c | 6 +++---
>> 1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/lib/librte_ethdev/rte_ethdev.c b/lib/librte_ethdev/rte_ethdev.c
>> index b3ed821..97316f3 100644
>> --- a/lib/librte_ethdev/rte_ethdev.c
>> +++ b/lib/librte_ethdev/rte_ethdev.c
>> @@ -1053,6 +1053,9 @@ rte_eth_dev_configure(uint16_t port_id, uint16_t
>> nb_rx_q, uint16_t nb_tx_q,
>> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_infos_get, -ENOTSUP);
>> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_configure, -ENOTSUP);
>>
>> + /* Copy the dev_conf parameter into the dev structure */
>> + memcpy(&dev->data->dev_conf, &local_conf, sizeof(dev->data->dev_conf));
>> +
>> rte_eth_dev_info_get(port_id, &dev_info);
>>
>> /* If number of queues specified by application for both Rx and Tx is
>> @@ -1098,9 +1101,6 @@ rte_eth_dev_configure(uint16_t port_id, uint16_t
>> nb_rx_q, uint16_t nb_tx_q,
>> rte_eth_convert_rx_offload_bitfield(
>> &dev_conf->rxmode, &local_conf.rxmode.offloads);
>>
>> - /* Copy the dev_conf parameter into the dev structure */
>> - memcpy(&dev->data->dev_conf, &local_conf, sizeof(dev->data->dev_conf));
>> -
>
> As I understand the goal of this copy is to pickup offloads fill in done
> three lines
> above. The patch breaks it.
It can be possible to copy only local_conf.rxmode.offloads here:
dev->data->dev_conf.rxmode.offloads = local_conf.rxmode.offloads;
And this assignment can be done within
"if (dev_conf->rxmode.ignore_offload_bitfield == 0)" block
And can you please add a comment to memcpy that it is required for
dev_infos_get() and needs to be called before it.
Also more details into commit log why this is required for DCB configure.
>
>> /*
>> * Check that the numbers of RX and TX queues are not greater
>> * than the maximum number of RX and TX queues supported by the
>
