> > +/* Get device attributes from HW. */
> > +static int get_hwif_attr(struct hinic3_hwdev *hwdev)
> > +{
> > + u32 attr0, attr1, attr2, attr3, attr6;
> > + struct hinic3_hwif *hwif;
> > +
> > + hwif = hwdev->hwif;
> > + attr0 = hinic3_hwif_read_reg(hwif, HINIC3_CSR_FUNC_ATTR0_ADDR);
> > + attr1 = hinic3_hwif_read_reg(hwif, HINIC3_CSR_FUNC_ATTR1_ADDR);
> > + attr2 = hinic3_hwif_read_reg(hwif, HINIC3_CSR_FUNC_ATTR2_ADDR);
> > + attr3 = hinic3_hwif_read_reg(hwif, HINIC3_CSR_FUNC_ATTR3_ADDR);
> > + attr6 = hinic3_hwif_read_reg(hwif, HINIC3_CSR_FUNC_ATTR6_ADDR);
> > + init_hwif_attr(&hwif->attr, attr0, attr1, attr2, attr3, attr6);
>
> well, get_hwif_attr() name is misleading here, as the function doesn't
> only read values, it also sets some of them. if there is no other users
> of init function, it might be better to merge them.
Thanks for your comments.
"get_hwif_attr" is actually misleading. In next version We consider
changing this to "init_hwif_attr" and the old "init_hwif_attr" will
be replaced with "set_hwif_attr" for better readability.
> > +
> > + return 0;
>
> there is no way the function can return error - what's the reason to
> have return value?
This is our oversight on error handling and patch splitting. We missed
the error case for "hinic3_hwif_read_reg" that returns errors when
PCIE_LINK_DOWN.
