On Mon, 9 Jan 2017 10:01:40 +0000 Remy Horton <remy.hor...@intel.com> wrote:
> On 09/01/2017 07:16, Yang, Qiming wrote: > > -----Original Message----- > > From: Stephen Hemminger [mailto:step...@networkplumber.org] > [..] > >> void > >> +rte_eth_dev_fw_version_get(uint8_t port_id, char *fw_version, int > >> +fw_length) { > >> + struct rte_eth_dev *dev; > >> + > >> + RTE_ETH_VALID_PORTID_OR_RET(port_id); > >> + dev = &rte_eth_devices[port_id]; > >> + > >> + RTE_FUNC_PTR_OR_RET(*dev->dev_ops->fw_version_get); > >> + (*dev->dev_ops->fw_version_get)(dev, fw_version, fw_length); } > > > > Maybe dev argument to fw_version_get should be: > > const struct rte_eth_dev *dev > > Qiming: do you means the argument to ops fw_version_get? > > why should add 'const'? both two are OK, but we usually use struct > > rte_eth_dev *dev. > > Does seem a bit odd to me as I don't think any of the other rte_dev_ops > entrypoints use const. Maybe they should but if that's now policy (I've > been under a rock recently) probably better to do them all in a seperate > cleanup patchset.. DPDK is somewhat lazy about using const.