On Fri, Apr 24, 2020 at 03:46:22PM +0300, Florinel Iordache wrote:
Add support for Ethernet Backplane KR generic driver using link training
(ieee802.3ap/ba standards), equalization algorithms (bee, fixed) and
enable qoriq family of devices.
This driver is dependent on uboot Backplane KR support:
patchwork.ozlabs.org/project/uboot/list/?series=164627&state=*
v2 changes:
* phy.rst and ABI/testing/sysfs-class-net-phydev updates with new PHY
interface values according to Florian Fainelli feedback
* dt bindings updates according to Rob Herring feedback: fixed errors
occurred when running 'make dt_binding_check'
* bpdev log changes according to feedback from Joe Perches: use %pV
instead of an intermediate buffer and refactoring
* reverse christmas tree updates according to David Miller feedback
* use pr_info_once function in probe to display qoriq backplane driver
version according to Joe's feedback
* introduce helper function dt_serdes_type in qoriq backplane according
to Joe's feedback
* use standard linux defines to access AN control/status registers and
not indirect with internal variables according to Andrew's feedback
* dt bindings link training updates: pre-cursor, main-cursor, post-cursor
* change display format for tx equalization using C() standard notation
* add priv pointer in backplane_device and lane as device specific private
extension to be used by upper layer backplane drivers
* backplane refactoring: split backplane_phy_info struct in
backplane_device and backplane_driver, add backplane specific ops and
move amp_red as qoriq specific param
* lane refactoring: split kr_lane_info struct in lane_device and lane_kr
in order to separate lane kr specific data by generic device lane data,
lane kr parameters unification, extension params for custom device
specific
* equalization refactoring: replace eq_setup_info/equalizer_info with
equalizer_driver/equalizer_device data structures
Hi Florinel
Feedback not addressed yet:
* general solution for PCS representation: still working to find a
generic suitable solution, exploring alternatives, perhaps this
should be addressed in phy generic layer
I actually think this is the most important point. It makes a big
difference to the overall structure of this code, the APIs it needs to
export. So don't expect too detailed a review until this is decided.
Andrew