On 11/23/2017 10:33 PM, Jingjing Wu wrote:
> Enable Virtchnl offload Caps negotiation and RSS_PF offload
> to support AVF basic interface.
>
> Signed-off-by: Jingjing Wu <jingjing...@intel.com>
> ---
> config/common_base | 2 +-
> drivers/net/i40e/i40e_ethdev.c | 64 +++++++++++++++----
> drivers/net/i40e/i40e_ethdev.h | 4 ++
> drivers/net/i40e/i40e_pf.c | 137
> +++++++++++++++++++++++++++++++++++++----
> drivers/net/i40e/i40e_pf.h | 6 ++
This is a i40e patch that enables AVF support, right?
I believe it would be better to separate the config change that enables the AVF
PMD by default from i40e changes patches.
<...>
> @@ -3694,14 +3695,21 @@ i40e_get_rss_lut(struct i40e_vsi *vsi, uint8_t *lut,
> uint16_t lut_size)
> uint32_t *lut_dw = (uint32_t *)lut;
> uint16_t i, lut_size_dw = lut_size / 4;
>
> - for (i = 0; i < lut_size_dw; i++)
> - lut_dw[i] = I40E_READ_REG(hw, I40E_PFQF_HLUT(i));
> + if (vsi->type == I40E_VSI_SRIOV) {
> + for (i = 0; i <= lut_size_dw; i++)
> + reg = I40E_VFQF_HLUT1(i, vsi->user_param);
> + lut_dw[i] = i40e_read_rx_ctl(hw, reg);
this assignment is out of "for" loop.
> + } else {
> + for (i = 0; i < lut_size_dw; i++)
> + lut_dw[i] = I40E_READ_REG(hw,
> + I40E_PFQF_HLUT(i));
> + }
> }
>
> return 0;
> }
>
<...>
> @@ -6754,8 +6784,20 @@ i40e_get_rss_key(struct i40e_vsi *vsi, uint8_t *key,
> uint8_t *key_len)
> uint32_t *key_dw = (uint32_t *)key;
> uint16_t i;
>
> - for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
> - key_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
> + if (vsi->type == I40E_VSI_SRIOV) {
> + for (i = 0; i <= I40E_VFQF_HKEY_MAX_INDEX; i++)
> + reg = I40E_VFQF_HKEY1(i, vsi->user_param);
> + key_dw[i] = i40e_read_rx_ctl(hw, reg);
This line is not part of the for loop, seems which is the intention, and all
credits goes to compiler to figuring this out.
> + *key_len = (I40E_VFQF_HKEY_MAX_INDEX + 1) *
> + sizeof(uint32_t);
> +
> + } else {
> + for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
> + reg = I40E_PFQF_HKEY(i);
> + key_dw[i] = i40e_read_rx_ctl(hw, reg);
Same problem here, "for" scope is wrong.
<...>
