1.. SPDX-License-Identifier: GPL-2.0
  2.. _phy_link_topology:
  3
  4=================
  5PHY link topology
  6=================
  7
  8Overview
  9========
 10
 11The PHY link topology representation in the networking stack aims at representing
 12the hardware layout for any given Ethernet link.
 13
 14An Ethernet interface from userspace's point of view is nothing but a
 15:c:type:`struct net_device <net_device>`, which exposes configuration options
 16through the legacy ioctls and the ethtool netlink commands. The base assumption
 17when designing these configuration APIs were that the link looks something like ::
 18
 19  +-----------------------+        +----------+      +--------------+
 20  | Ethernet Controller / |        | Ethernet |      | Connector /  |
 21  |       MAC             | ------ |   PHY    | ---- |    Port      | ---... to LP
 22  +-----------------------+        +----------+      +--------------+
 23  struct net_device               struct phy_device
 24
 25Commands that needs to configure the PHY will go through the net_device.phydev
 26field to reach the PHY and perform the relevant configuration.
 27
 28This assumption falls apart in more complex topologies that can arise when,
 29for example, using SFP transceivers (although that's not the only specific case).
 30
 31Here, we have 2 basic scenarios. Either the MAC is able to output a serialized
 32interface, that can directly be fed to an SFP cage, such as SGMII, 1000BaseX,
 3310GBaseR, etc.
 34
 35The link topology then looks like this (when an SFP module is inserted) ::
 36
 37  +-----+  SGMII  +------------+
 38  | MAC | ------- | SFP Module |
 39  +-----+         +------------+
 40
 41Knowing that some modules embed a PHY, the actual link is more like ::
 42
 43  +-----+  SGMII   +--------------+
 44  | MAC | -------- | PHY (on SFP) |
 45  +-----+          +--------------+
 46
 47In this case, the SFP PHY is handled by phylib, and registered by phylink through
 48its SFP upstream ops.
 49
 50Now some Ethernet controllers aren't able to output a serialized interface, so
 51we can't directly connect them to an SFP cage. However, some PHYs can be used
 52as media-converters, to translate the non-serialized MAC MII interface to a
 53serialized MII interface fed to the SFP ::
 54
 55  +-----+  RGMII  +-----------------------+  SGMII  +--------------+
 56  | MAC | ------- | PHY (media converter) | ------- | PHY (on SFP) |
 57  +-----+         +-----------------------+         +--------------+
 58
 59This is where the model of having a single net_device.phydev pointer shows its
 60limitations, as we now have 2 PHYs on the link.
 61
 62The phy_link topology framework aims at providing a way to keep track of every
 63PHY on the link, for use by both kernel drivers and subsystems, but also to
 64report the topology to userspace, allowing to target individual PHYs in configuration
 65commands.
 66
 67API
 68===
 69
 70The :c:type:`struct phy_link_topology <phy_link_topology>` is a per-netdevice
 71resource, that gets initialized at netdevice creation. Once it's initialized,
 72it is then possible to register PHYs to the topology through :
 73
 74:c:func:`phy_link_topo_add_phy`
 75
 76Besides registering the PHY to the topology, this call will also assign a unique
 77index to the PHY, which can then be reported to userspace to refer to this PHY
 78(akin to the ifindex). This index is a u32, ranging from 1 to U32_MAX. The value
 790 is reserved to indicate the PHY doesn't belong to any topology yet.
 80
 81The PHY can then be removed from the topology through
 82
 83:c:func:`phy_link_topo_del_phy`
 84
 85These function are already hooked into the phylib subsystem, so all PHYs that
 86are linked to a net_device through :c:func:`phy_attach_direct` will automatically
 87join the netdev's topology.
 88
 89PHYs that are on a SFP module will also be automatically registered IF the SFP
 90upstream is phylink (so, no media-converter).
 91
 92PHY drivers that can be used as SFP upstream need to call :c:func:`phy_sfp_attach_phy`
 93and :c:func:`phy_sfp_detach_phy`, which can be used as a
 94.attach_phy / .detach_phy implementation for the
 95:c:type:`struct sfp_upstream_ops <sfp_upstream_ops>`.
 96
 97UAPI
 98====
 99
100There exist a set of netlink commands to query the link topology from userspace,
101see ``Documentation/networking/ethtool-netlink.rst``.
102
103The whole point of having a topology representation is to assign the phyindex
104field in :c:type:`struct phy_device <phy_device>`. This index is reported to
105userspace using the ``ETHTOOL_MSG_PHY_GET`` ethtnl command. Performing a DUMP operation
106will result in all PHYs from all net_device being listed. The DUMP command
107accepts either a ``ETHTOOL_A_HEADER_DEV_INDEX`` or ``ETHTOOL_A_HEADER_DEV_NAME``
108to be passed in the request to filter the DUMP to a single net_device.
109
110The retrieved index can then be passed as a request parameter using the
111``ETHTOOL_A_HEADER_PHY_INDEX`` field in the following ethnl commands :
112
113* ``ETHTOOL_MSG_STRSET_GET`` to get the stats string set from a given PHY
114* ``ETHTOOL_MSG_CABLE_TEST_ACT`` and ``ETHTOOL_MSG_CABLE_TEST_ACT``, to perform
115  cable testing on a given PHY on the link (most likely the outermost PHY)
116* ``ETHTOOL_MSG_PSE_SET`` and ``ETHTOOL_MSG_PSE_GET`` for PHY-controlled PoE and PSE settings
117* ``ETHTOOL_MSG_PLCA_GET_CFG``, ``ETHTOOL_MSG_PLCA_SET_CFG`` and ``ETHTOOL_MSG_PLCA_GET_STATUS``
118  to set the PLCA (Physical Layer Collision Avoidance) parameters
119
120Note that the PHY index can be passed to other requests, which will silently
121ignore it if present and irrelevant.