* [PATCH net-next v3 1/7] doc: net: add backplane documentation
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings Florinel Iordache
` (5 subsequent siblings)
6 siblings, 0 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Add ethernet backplane documentation
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
Documentation/networking/backplane.rst | 159 +++++++++++++++++++++++++++++++++
Documentation/networking/phy.rst | 9 +-
2 files changed, 165 insertions(+), 3 deletions(-)
create mode 100644 Documentation/networking/backplane.rst
diff --git a/Documentation/networking/backplane.rst b/Documentation/networking/backplane.rst
new file mode 100644
index 0000000..fb91ba8
--- /dev/null
+++ b/Documentation/networking/backplane.rst
@@ -0,0 +1,159 @@
+.. SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+
+=========================
+Ethernet Backplane Driver
+=========================
+
+Author:
+Florinel Iordache <florinel.iordache@nxp.com>
+
+Contents
+========
+
+ - Ethernet Backplane Overview
+ - Equalization
+ - Auto-negotiation
+ - Link training
+ - Enable backplane support in Linux kernel
+ - Ethernet Backplane support architecture
+ - Supported equalization algorithms
+ - Supported backplane protocols
+ - Supported platforms
+
+Ethernet Backplane Overview
+===========================
+
+Ethernet operation over electrical backplanes, also referred to as Ethernet
+Backplane, combines the IEEE 802.3 Media Access Control (MAC) and MAC
+Control sublayers with a family of Physical Layers defined to support
+operation over a modular chassis backplane.
+The main standard specification for Ethernet Backplane is: IEEE802.3ap-2007
+Amendment 4: Ethernet Operation over Electrical Backplanes
+which includes the new Clause 69 through Clause 74.
+Additional specifications define support for various speeds and 4-lanes:
+IEEE802.3ba-2010.
+Signal equalization is required based on the link quality. The standard
+specifies that a start-up algorithm should be in place in order to get the
+link up.
+
+Equalization
+============
+
+Equalization represents the procedure required to minimize the effects of signal
+distortion, noise, interference occurred in high-speed communication channels.
+The equalizer purpose is to improve signal integrity in terms of bit error rate
+(BER) in order to allow accurate recovery of the transmitted symbols.
+
+A simplified view of channel equalization:
+
+ LD <======== channel =========> LP
+ Local Device Link Partner
+
+ |-----| ___
+ | | <======== channel =========> / |
+ | | witout signal Equalization / \
+ | | / |
+ ____| |____ ___/ \___
+
+ |\ _ |-----|
+ | \_/ | <======== channel =========> | |
+ | | with signal Equalization | |
+ | | | |
+ ____| |____ ____| |____
+
+ LD Tx waveform LP Rx waveform
+
+Auto-negotiation
+================
+
+Auto-negotiation allows the devices at both ends of a link segment to advertise
+abilities, acknowledge receipt, and discover the common modes of operation that
+both devices share. It also rejects the use of operational modes not shared by
+both devices. Auto-negotiation does not test link segment characteristics.
+
+Link training
+=============
+
+Link training occurs after auto-negotiation has determined the link to be a
+Base-KR, but before auto-negotiation is done. It continuously exchanges messages
+(training frames) between the local and the remote device as part of the
+start-up phase. Link training tunes the equalization parameters of the remote and
+local transmitter to improve the link quality in terms of bit error rate.
+Both LP (link partner/remote device) and LD (local device) perform link training
+in parallel. Link training is finished when both sides decide that the channel is
+equalized and then the link is considered up.
+
+Enable backplane support in Linux kernel
+========================================
+
+To enable the Ethernet Backplane, the following Kconfig options are available:
+
+# enable generic Ethernet Backplane support:
+CONFIG_ETH_BACKPLANE=y
+# enable Fixed (No Equalization) algorithm:
+CONFIG_ETH_BACKPLANE_FIXED=y
+# enable 3-Taps Bit Edge Equalization (BEE) algorithm:
+CONFIG_ETH_BACKPLANE_BEE=y
+# enable QorIQ Ethernet Backplane driver:
+CONFIG_ETH_BACKPLANE_QORIQ=y
+
+Ethernet Backplane support architecture
+=======================================
+
+Ethernet Backplane support in Linux kernel complies with the following standard
+design concepts:
+* Modularity:
+ # internal components are separated in well defined functional modules
+* Reusability:
+ # lower layer components provide basic functionalities which are reused by
+ the upper layer modules
+* Extensibility:
+ # architecture can be easily extended with support for new:
+ - backplane protocols
+ - equalization algorithms
+ - supported devices
+It is designed as a loosely coupled architecture in order to allow the
+possibility to easily create desired backplane system configurations according
+to user needs by specifying different components and initialization parameters
+without recompiling the kernel.
+
+ ------------------ ------------------------------------
+ | EQ Algorithms | | Specific device drivers |
+ | ------------ | | Backplane support |
+ | | Fixed | | | ------------------ ----------- |
+ | ------------ | | | QorIQ | | | |
+ | | BEE | | | | devices | | | |
+ | ------------ | | | -------------- | | other | |
+ | | others | | | | | Serdes 10G | | | devices | |
+ ---------------------------- | | -------------- | | support | |
+ | Link Training | | | | Serdes 28G | | | | |
+ | and Auto-negotiation | | | -------------- | | | |
+ | (IEEE 802.3-ap/ba) | | |----------------| |---------| |
+ ---------------------------------------------------------------------------
+ | Ethernet Backplane Generic Driver |
+ ---------------------------------------------------------------------------
+ | PHY Abstraction Layer |
+ ---------------------------------------------------------------------------
+
+Supported equalization algorithms
+=================================
+
+Ethernet Backplane supports the following equalization algorithms:
+
+- Fixed setup (No Equalization algorithm)
+- 3-Taps Bit Edge Equalization (BEE) algorithm
+
+Supported backplane protocols
+=============================
+
+Ethernet Backplane supports the following protocols:
+
+- Single-lane:
+10GBase-KR
+
+Supported platforms
+===================
+
+Ethernet Backplane is enabled on the following platforms:
+
+LS1046A
diff --git a/Documentation/networking/phy.rst b/Documentation/networking/phy.rst
index 2561060..ec17c31 100644
--- a/Documentation/networking/phy.rst
+++ b/Documentation/networking/phy.rst
@@ -279,9 +279,12 @@ Some of the interface modes are described below:
XFI and SFI are not PHY interface types in their own right.
``PHY_INTERFACE_MODE_10GKR``
- This is the IEEE 802.3 Clause 49 defined 10GBASE-R with Clause 73
- autonegotiation. Please refer to the IEEE standard for further
- information.
+ This is 10G Ethernet Backplane over single lane specified in
+ IEEE802.3ap-2007 Amendment 4: Ethernet Operation over Electrical
+ Backplanes which includes the new Clause 69 through Clause 74
+ including autonegotiation. 10GKR uses the same physical layer
+ encoding as 10GBASE-R defined in IEEE802.3 Clause 49. Please refer
+ to the IEEE standard for further information.
Note: due to legacy usage, some 10GBASE-R usage incorrectly makes
use of this definition.
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 1/7] doc: net: add backplane documentation Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-22 22:20 ` Florian Fainelli
2020-06-29 21:58 ` Rob Herring
2020-06-22 13:35 ` [PATCH net-next v3 3/7] net: fman: add kr support for dpaa1 mac Florinel Iordache
` (4 subsequent siblings)
6 siblings, 2 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Add ethernet backplane device tree bindings
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
.../bindings/net/ethernet-controller.yaml | 7 ++-
.../devicetree/bindings/net/ethernet-phy.yaml | 50 ++++++++++++++++++++++
.../devicetree/bindings/net/serdes-lane.yaml | 49 +++++++++++++++++++++
Documentation/devicetree/bindings/net/serdes.yaml | 42 ++++++++++++++++++
4 files changed, 146 insertions(+), 2 deletions(-)
create mode 100644 Documentation/devicetree/bindings/net/serdes-lane.yaml
create mode 100644 Documentation/devicetree/bindings/net/serdes.yaml
diff --git a/Documentation/devicetree/bindings/net/ethernet-controller.yaml b/Documentation/devicetree/bindings/net/ethernet-controller.yaml
index 1c44740..6c4c7d8 100644
--- a/Documentation/devicetree/bindings/net/ethernet-controller.yaml
+++ b/Documentation/devicetree/bindings/net/ethernet-controller.yaml
@@ -91,10 +91,13 @@ properties:
- rxaui
- xaui
- # 10GBASE-KR, XFI, SFI
- - 10gbase-kr
+ # 10GBASE-R, XFI, SFI
+ - 10gbase-r
- usxgmii
+ # 10GBASE-KR (10G Ethernet Backplane with autonegotiation)
+ - 10gbase-kr
+
phy-mode:
$ref: "#/properties/phy-connection-type"
diff --git a/Documentation/devicetree/bindings/net/ethernet-phy.yaml b/Documentation/devicetree/bindings/net/ethernet-phy.yaml
index 9b1f114..a23a7d6 100644
--- a/Documentation/devicetree/bindings/net/ethernet-phy.yaml
+++ b/Documentation/devicetree/bindings/net/ethernet-phy.yaml
@@ -162,6 +162,42 @@ properties:
description:
Specifies a reference to a node representing a SFP cage.
+ eq-algorithm:
+ description:
+ Specifies the desired equalization algorithm to be used
+ by the KR link training
+ oneOf:
+ - const: fixed
+ description:
+ Backplane KR using fixed coefficients meaning no
+ equalization algorithm
+ - const: bee
+ description:
+ Backplane KR using 3-Taps Bit Edge Equalization (BEE)
+ algorithm
+
+ eq-init:
+ $ref: /schemas/types.yaml#/definitions/uint32-array
+ minItems: 3
+ maxItems: 3
+ description:
+ Triplet of KR coefficients. Specifies the initialization
+ values for standard KR equalization coefficients used by
+ the link training (pre-cursor, main-cursor, post-cursor)
+
+ eq-params:
+ $ref: /schemas/types.yaml#/definitions/uint32-array
+ description:
+ Variable size array of KR parameters. Specifies the HW
+ specific parameters used by the link training.
+
+ lane-handle:
+ $ref: /schemas/types.yaml#definitions/phandle
+ description:
+ Specifies a reference (or array of references) to a node
+ representing the desired SERDES lane (or lanes) used in
+ backplane mode.
+
required:
- reg
@@ -184,3 +220,17 @@ examples:
reset-deassert-us = <2000>;
};
};
+ - |
+ ethernet {
+ #address-cells = <1>;
+ #size-cells = <0>;
+
+ ethernet-phy@0 {
+ compatible = "ethernet-phy-ieee802.3-c45";
+ reg = <0x0>;
+ lane-handle = <&lane_d>;
+ eq-algorithm = "fixed";
+ eq-init = <0x2 0x29 0x5>;
+ eq-params = <0>;
+ };
+ };
diff --git a/Documentation/devicetree/bindings/net/serdes-lane.yaml b/Documentation/devicetree/bindings/net/serdes-lane.yaml
new file mode 100644
index 0000000..d83a6a9
--- /dev/null
+++ b/Documentation/devicetree/bindings/net/serdes-lane.yaml
@@ -0,0 +1,49 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/net/serdes-lane.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Serdes Lane Binding
+
+maintainers:
+ - Florinel Iordache <florinel.iordache@nxp.com>
+
+properties:
+ $nodename:
+ pattern: "^lane(@[a-f0-9]+)?$"
+
+ compatible:
+ oneOf:
+ - const: lane-10g
+ description: Lane part of a 10G SerDes module
+
+ reg:
+ description:
+ Registers memory map offset and size for this lane
+
+ reg-names:
+ description:
+ Names of the register map given in "reg" node.
+
+examples:
+ - |
+ serdes1: serdes@1ea0000 {
+ compatible = "serdes-10g";
+ reg = <0x0 0x1ea0000 0 0x00002000>;
+ reg-names = "serdes", "serdes-10g";
+ little-endian;
+
+ #address-cells = <1>;
+ #size-cells = <1>;
+ lane_a: lane@800 {
+ compatible = "lane-10g";
+ reg = <0x800 0x40>;
+ reg-names = "lane", "serdes-lane";
+ };
+ lane_b: lane@840 {
+ compatible = "lane-10g";
+ reg = <0x840 0x40>;
+ reg-names = "lane", "serdes-lane";
+ };
+ };
diff --git a/Documentation/devicetree/bindings/net/serdes.yaml b/Documentation/devicetree/bindings/net/serdes.yaml
new file mode 100644
index 0000000..ed77689c
--- /dev/null
+++ b/Documentation/devicetree/bindings/net/serdes.yaml
@@ -0,0 +1,42 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/net/serdes.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Serdes Module Binding
+
+maintainers:
+ - Florinel Iordache <florinel.iordache@nxp.com>
+
+properties:
+ $nodename:
+ pattern: "^serdes(@[a-f0-9]+)?$"
+
+ compatible:
+ oneOf:
+ - const: serdes-10g
+ description: SerDes module type of 10G
+
+ reg:
+ description:
+ Registers memory map offset and size for this serdes module
+
+ reg-names:
+ description:
+ Names of the register map given in "reg" node.
+
+ little-endian:
+ description:
+ Specifies the endianness of serdes module
+ For complete definition see
+ Documentation/devicetree/bindings/common-properties.txt
+
+examples:
+ - |
+ serdes1: serdes@1ea0000 {
+ compatible = "serdes-10g";
+ reg = <0x0 0x1ea0000 0 0x00002000>;
+ reg-names = "serdes", "serdes-10g";
+ little-endian;
+ };
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings
2020-06-22 13:35 ` [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings Florinel Iordache
@ 2020-06-22 22:20 ` Florian Fainelli
2020-06-24 12:55 ` [EXT] " Florinel Iordache
2020-06-29 21:58 ` Rob Herring
1 sibling, 1 reply; 23+ messages in thread
From: Florian Fainelli @ 2020-06-22 22:20 UTC (permalink / raw)
To: florinel.iordache, davem, netdev, andrew, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel
On 6/22/20 6:35 AM, Florinel Iordache wrote:
> Add ethernet backplane device tree bindings
>
> Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
> ---
[snip]
> +properties:
> + $nodename:
> + pattern: "^serdes(@[a-f0-9]+)?$"
> +
> + compatible:
> + oneOf:
> + - const: serdes-10g
> + description: SerDes module type of 10G
Since this appears to be memory mapped in your case, it does not sound
like "serdes-10g" alone is going to be sufficient, should not we have a
SoC specific compatible string if nothing else?
> +
> + reg:
> + description:
> + Registers memory map offset and size for this serdes module
> +
> + reg-names:
> + description:
> + Names of the register map given in "reg" node.
You would also need to describe how many of these two properties are
expected.
> +
> + little-endian:
> + description:
> + Specifies the endianness of serdes module
> + For complete definition see
> + Documentation/devicetree/bindings/common-properties.txt
This is redundant with the default binding then, and if it is not
already in the common YAML binding, can you please add little-endian and
native-endian added there?
> +
> +examples:
> + - |
> + serdes1: serdes@1ea0000 {
> + compatible = "serdes-10g";
> + reg = <0x0 0x1ea0000 0 0x00002000>;
> + reg-names = "serdes", "serdes-10g";
> + little-endian;
> + };
>
--
Florian
^ permalink raw reply [flat|nested] 23+ messages in thread* RE: [EXT] Re: [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings
2020-06-22 22:20 ` Florian Fainelli
@ 2020-06-24 12:55 ` Florinel Iordache
2020-06-26 18:55 ` Florian Fainelli
0 siblings, 1 reply; 23+ messages in thread
From: Florinel Iordache @ 2020-06-24 12:55 UTC (permalink / raw)
To: Florian Fainelli, davem, netdev, andrew, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, Leo Li, Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel, Florinel Iordache
> -----Original Message-----
> From: Florian Fainelli <f.fainelli@gmail.com>
> Sent: Tuesday, June 23, 2020 1:21 AM
> To: Florinel Iordache <florinel.iordache@nxp.com>; davem@davemloft.net;
> netdev@vger.kernel.org; andrew@lunn.ch; hkallweit1@gmail.com;
> linux@armlinux.org.uk
> Cc: devicetree@vger.kernel.org; linux-doc@vger.kernel.org;
> robh+dt@kernel.org; mark.rutland@arm.com; kuba@kernel.org;
> corbet@lwn.net; shawnguo@kernel.org; Leo Li <leoyang.li@nxp.com>; Madalin
> Bucur (OSS) <madalin.bucur@oss.nxp.com>; Ioana Ciornei
> <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
> Subject: [EXT] Re: [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt
> bindings
>
> Caution: EXT Email
>
> On 6/22/20 6:35 AM, Florinel Iordache wrote:
> > Add ethernet backplane device tree bindings
> >
> > Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
> > ---
>
> [snip]
>
> > +properties:
> > + $nodename:
> > + pattern: "^serdes(@[a-f0-9]+)?$"
> > +
> > + compatible:
> > + oneOf:
> > + - const: serdes-10g
> > + description: SerDes module type of 10G
>
> Since this appears to be memory mapped in your case, it does not sound like
> "serdes-10g" alone is going to be sufficient, should not we have a SoC specific
> compatible string if nothing else?
My intention was to make it generic enough to be used by any SerDes (Serializer/Deserializer) block.
So I was thinking that specifying serdes as HW block and the type: 10g (or 28g for example) should be enough.
I could add SoC specific (or family of SoC) to the compatible string
like for example Freescale/NXP QorIQ Soc: "fsl,ls1046a-serdes-10g" or "fsl,qoriq-serdes-10g"
>
> > +
> > + reg:
> > + description:
> > + Registers memory map offset and size for this serdes module
> > +
> > + reg-names:
> > + description:
> > + Names of the register map given in "reg" node.
>
> You would also need to describe how many of these two properties are
> expected.
Only one memory map is required since the memory maps for lanes are individually described
(as it is documented in serdes-lane.yaml).
I will specify this.
>
> > +
> > + little-endian:
> > + description:
> > + Specifies the endianness of serdes module
> > + For complete definition see
> > + Documentation/devicetree/bindings/common-properties.txt
>
> This is redundant with the default binding then, and if it is not already in the
> common YAML binding, can you please add little-endian and native-endian
> added there?
The endianness of the serdes block must be specified as little-endian or big-endian.
The serdes endianness may be different than the cores endianness.
This is also the case for QorIQ LS1043/LS1046 platforms with ARM cores which
are little endian but serdes block is big endian.
So endianness must be specified in device tree in order for the driver to know how to access it.
This is the generic binding description (with an example below)
but for LS1046 platform for example we should put: big-endian
(as it is in the last patch: 0007-arm64-dts-add-serdes-and-mdio-description.patch
in file: /arch/arm64/boot/dts/freescale/fsl-ls1046a.dtsi )
>
> > +
> > +examples:
> > + - |
> > + serdes1: serdes@1ea0000 {
> > + compatible = "serdes-10g";
> > + reg = <0x0 0x1ea0000 0 0x00002000>;
> > + reg-names = "serdes", "serdes-10g";
> > + little-endian;
> > + };
> >
>
>
> --
> Florian
Thank you for feedback
Florinel.
^ permalink raw reply [flat|nested] 23+ messages in thread* Re: [EXT] Re: [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings
2020-06-24 12:55 ` [EXT] " Florinel Iordache
@ 2020-06-26 18:55 ` Florian Fainelli
0 siblings, 0 replies; 23+ messages in thread
From: Florian Fainelli @ 2020-06-26 18:55 UTC (permalink / raw)
To: Florinel Iordache, davem, netdev, andrew, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, Leo Li, Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel
On 6/24/20 5:55 AM, Florinel Iordache wrote:
>> -----Original Message-----
>> From: Florian Fainelli <f.fainelli@gmail.com>
>> Sent: Tuesday, June 23, 2020 1:21 AM
>> To: Florinel Iordache <florinel.iordache@nxp.com>; davem@davemloft.net;
>> netdev@vger.kernel.org; andrew@lunn.ch; hkallweit1@gmail.com;
>> linux@armlinux.org.uk
>> Cc: devicetree@vger.kernel.org; linux-doc@vger.kernel.org;
>> robh+dt@kernel.org; mark.rutland@arm.com; kuba@kernel.org;
>> corbet@lwn.net; shawnguo@kernel.org; Leo Li <leoyang.li@nxp.com>; Madalin
>> Bucur (OSS) <madalin.bucur@oss.nxp.com>; Ioana Ciornei
>> <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
>> Subject: [EXT] Re: [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt
>> bindings
>>
>> Caution: EXT Email
>>
>> On 6/22/20 6:35 AM, Florinel Iordache wrote:
>>> Add ethernet backplane device tree bindings
>>>
>>> Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
>>> ---
>>
>> [snip]
>>
>>> +properties:
>>> + $nodename:
>>> + pattern: "^serdes(@[a-f0-9]+)?$"
>>> +
>>> + compatible:
>>> + oneOf:
>>> + - const: serdes-10g
>>> + description: SerDes module type of 10G
>>
>> Since this appears to be memory mapped in your case, it does not sound like
>> "serdes-10g" alone is going to be sufficient, should not we have a SoC specific
>> compatible string if nothing else?
>
> My intention was to make it generic enough to be used by any SerDes (Serializer/Deserializer) block.
> So I was thinking that specifying serdes as HW block and the type: 10g (or 28g for example) should be enough.
> I could add SoC specific (or family of SoC) to the compatible string
> like for example Freescale/NXP QorIQ Soc: "fsl,ls1046a-serdes-10g" or "fsl,qoriq-serdes-10g"
It does not seem to me that the register interface is going to be
anything but generic, therefore using SoC specific compatible strings
would be a safer and forward looking approach. If a generic/fall back
compatibility string can be added, it can be added later on, that is
much less problematic than the opposite.
>
>>
>>> +
>>> + reg:
>>> + description:
>>> + Registers memory map offset and size for this serdes module
>>> +
>>> + reg-names:
>>> + description:
>>> + Names of the register map given in "reg" node.
>>
>> You would also need to describe how many of these two properties are
>> expected.
>
> Only one memory map is required since the memory maps for lanes are individually described
> (as it is documented in serdes-lane.yaml).
> I will specify this.
Then I believe you need to advertise that with maxItems property to
denote how many.
>
>>
>>> +
>>> + little-endian:
>>> + description:
>>> + Specifies the endianness of serdes module
>>> + For complete definition see
>>> + Documentation/devicetree/bindings/common-properties.txt
>>
>> This is redundant with the default binding then, and if it is not already in the
>> common YAML binding, can you please add little-endian and native-endian
>> added there?
>
> The endianness of the serdes block must be specified as little-endian or big-endian.
> The serdes endianness may be different than the cores endianness.
> This is also the case for QorIQ LS1043/LS1046 platforms with ARM cores which
> are little endian but serdes block is big endian.
> So endianness must be specified in device tree in order for the driver to know how to access it.
I understand that part, my point was more than these properties are
generic properties, therefore it should not be necessary to provide a
description, and their definition belongs in the common properties
binding. If the common binding does not have a definition for those
(that is, in a YAML way), then please add them there.
--
Florian
^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings
2020-06-22 13:35 ` [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings Florinel Iordache
2020-06-22 22:20 ` Florian Fainelli
@ 2020-06-29 21:58 ` Rob Herring
1 sibling, 0 replies; 23+ messages in thread
From: Rob Herring @ 2020-06-29 21:58 UTC (permalink / raw)
To: Florinel Iordache
Cc: kuba, robh+dt, shawnguo, linux-doc, devicetree, davem, andrew,
mark.rutland, linux, corbet, ioana.ciornei, linux-kernel,
hkallweit1, f.fainelli, leoyang.li, netdev, madalin.bucur
On Mon, 22 Jun 2020 16:35:19 +0300, Florinel Iordache wrote:
> Add ethernet backplane device tree bindings
>
> Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
> ---
> .../bindings/net/ethernet-controller.yaml | 7 ++-
> .../devicetree/bindings/net/ethernet-phy.yaml | 50 ++++++++++++++++++++++
> .../devicetree/bindings/net/serdes-lane.yaml | 49 +++++++++++++++++++++
> Documentation/devicetree/bindings/net/serdes.yaml | 42 ++++++++++++++++++
> 4 files changed, 146 insertions(+), 2 deletions(-)
> create mode 100644 Documentation/devicetree/bindings/net/serdes-lane.yaml
> create mode 100644 Documentation/devicetree/bindings/net/serdes.yaml
>
My bot found errors running 'make dt_binding_check' on your patch:
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/net/serdes.example.dt.yaml: example-0: serdes@1ea0000:reg:0: [0, 32112640, 0, 8192] is too long
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/net/serdes-lane.example.dt.yaml: example-0: serdes@1ea0000:reg:0: [0, 32112640, 0, 8192] is too long
See https://patchwork.ozlabs.org/patch/1314386
If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure dt-schema is up to date:
pip3 install git+https://github.com/devicetree-org/dt-schema.git@master --upgrade
Please check and re-submit.
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH net-next v3 3/7] net: fman: add kr support for dpaa1 mac
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 1/7] doc: net: add backplane documentation Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 2/7] dt-bindings: net: add backplane dt bindings Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 4/7] net: phy: add backplane kr driver support Florinel Iordache
` (3 subsequent siblings)
6 siblings, 0 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Add kr support in mac driver for dpaa1
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
drivers/net/ethernet/freescale/fman/mac.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/freescale/fman/mac.c b/drivers/net/ethernet/freescale/fman/mac.c
index 43427c5..90fe594 100644
--- a/drivers/net/ethernet/freescale/fman/mac.c
+++ b/drivers/net/ethernet/freescale/fman/mac.c
@@ -1,4 +1,5 @@
/* Copyright 2008-2015 Freescale Semiconductor, Inc.
+ * Copyright 2020 NXP
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
@@ -220,6 +221,7 @@ static int memac_initialization(struct mac_device *mac_dev)
set_fman_mac_params(mac_dev, ¶ms);
+ /* use XGMII mode for all 10G interfaces to setup memac */
if (priv->max_speed == SPEED_10000)
params.phy_if = PHY_INTERFACE_MODE_XGMII;
@@ -540,7 +542,8 @@ static void setup_memac(struct mac_device *mac_dev)
[PHY_INTERFACE_MODE_RGMII_TXID] = SPEED_1000,
[PHY_INTERFACE_MODE_RTBI] = SPEED_1000,
[PHY_INTERFACE_MODE_QSGMII] = SPEED_1000,
- [PHY_INTERFACE_MODE_XGMII] = SPEED_10000
+ [PHY_INTERFACE_MODE_XGMII] = SPEED_10000,
+ [PHY_INTERFACE_MODE_10GKR] = SPEED_10000
};
static struct platform_device *dpaa_eth_add_device(int fman_id,
@@ -795,9 +798,12 @@ static int mac_probe(struct platform_device *_of_dev)
if (priv->max_speed == 1000)
mac_dev->if_support |= SUPPORTED_1000baseT_Full;
- /* The 10G interface only supports one mode */
+ /* Supported 10G interfaces */
if (mac_dev->phy_if == PHY_INTERFACE_MODE_XGMII)
mac_dev->if_support = SUPPORTED_10000baseT_Full;
+ /* Supported KR interfaces */
+ if (mac_dev->phy_if == PHY_INTERFACE_MODE_10GKR)
+ mac_dev->if_support = SUPPORTED_10000baseKR_Full;
/* Get the rest of the PHY information */
mac_dev->phy_node = of_parse_phandle(mac_node, "phy-handle", 0);
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
` (2 preceding siblings ...)
2020-06-22 13:35 ` [PATCH net-next v3 3/7] net: fman: add kr support for dpaa1 mac Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-22 14:24 ` Andrew Lunn
2020-06-22 21:46 ` Jakub Kicinski
2020-06-22 13:35 ` [PATCH net-next v3 5/7] net: phy: enable qoriq backplane support Florinel Iordache
` (2 subsequent siblings)
6 siblings, 2 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Add support for backplane kr generic driver including link training
(ieee802.3ap/ba) and fixed equalization algorithm
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
drivers/net/phy/Kconfig | 2 +
drivers/net/phy/Makefile | 1 +
drivers/net/phy/backplane/Kconfig | 20 +
drivers/net/phy/backplane/Makefile | 9 +
drivers/net/phy/backplane/backplane.c | 1557 +++++++++++++++++++++++++++++
drivers/net/phy/backplane/backplane.h | 293 ++++++
drivers/net/phy/backplane/eq_fixed.c | 83 ++
drivers/net/phy/backplane/equalization.h | 275 +++++
drivers/net/phy/backplane/link_training.c | 1529 ++++++++++++++++++++++++++++
drivers/net/phy/backplane/link_training.h | 32 +
10 files changed, 3801 insertions(+)
create mode 100644 drivers/net/phy/backplane/Kconfig
create mode 100644 drivers/net/phy/backplane/Makefile
create mode 100644 drivers/net/phy/backplane/backplane.c
create mode 100644 drivers/net/phy/backplane/backplane.h
create mode 100644 drivers/net/phy/backplane/eq_fixed.c
create mode 100644 drivers/net/phy/backplane/equalization.h
create mode 100644 drivers/net/phy/backplane/link_training.c
create mode 100644 drivers/net/phy/backplane/link_training.h
diff --git a/drivers/net/phy/Kconfig b/drivers/net/phy/Kconfig
index f257023..fa48b8e 100644
--- a/drivers/net/phy/Kconfig
+++ b/drivers/net/phy/Kconfig
@@ -549,6 +549,8 @@ config XILINX_GMII2RGMII
the Reduced Gigabit Media Independent Interface(RGMII) between
Ethernet physical media devices and the Gigabit Ethernet controller.
+source "drivers/net/phy/backplane/Kconfig"
+
endif # PHYLIB
config MICREL_KS8995MA
diff --git a/drivers/net/phy/Makefile b/drivers/net/phy/Makefile
index dc9e53b..4849c16 100644
--- a/drivers/net/phy/Makefile
+++ b/drivers/net/phy/Makefile
@@ -104,3 +104,4 @@ obj-$(CONFIG_STE10XP) += ste10Xp.o
obj-$(CONFIG_TERANETICS_PHY) += teranetics.o
obj-$(CONFIG_VITESSE_PHY) += vitesse.o
obj-$(CONFIG_XILINX_GMII2RGMII) += xilinx_gmii2rgmii.o
+obj-$(CONFIG_ETH_BACKPLANE) += backplane/
diff --git a/drivers/net/phy/backplane/Kconfig b/drivers/net/phy/backplane/Kconfig
new file mode 100644
index 0000000..9ec54b5
--- /dev/null
+++ b/drivers/net/phy/backplane/Kconfig
@@ -0,0 +1,20 @@
+# SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+config ETH_BACKPLANE
+ tristate "Ethernet Backplane support"
+ depends on OF_MDIO
+ help
+ This module provides driver support for Ethernet Operation over
+ Electrical Backplanes. It includes Backplane generic
+ driver including support for Link Training (IEEE802.3ap/ba).
+ Based on the link quality, a signal equalization is required.
+ The standard specifies that a start-up algorithm should be in place
+ in order to get the link up.
+
+config ETH_BACKPLANE_FIXED
+ tristate "Fixed: No Equalization algorithm"
+ depends on ETH_BACKPLANE
+ help
+ This module provides a driver to setup fixed user configurable
+ coefficient values for backplanes equalization. This means
+ No Equalization algorithm is used to adapt the initial coefficients
+ initially set by the user.
\ No newline at end of file
diff --git a/drivers/net/phy/backplane/Makefile b/drivers/net/phy/backplane/Makefile
new file mode 100644
index 0000000..ded6f2d
--- /dev/null
+++ b/drivers/net/phy/backplane/Makefile
@@ -0,0 +1,9 @@
+# SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+#
+# Makefile for Ethernet Backplane driver
+#
+
+obj-$(CONFIG_ETH_BACKPLANE) += eth_backplane.o
+obj-$(CONFIG_ETH_BACKPLANE_FIXED) += eq_fixed.o
+
+eth_backplane-objs := backplane.o link_training.o
diff --git a/drivers/net/phy/backplane/backplane.c b/drivers/net/phy/backplane/backplane.c
new file mode 100644
index 0000000..d9a4770
--- /dev/null
+++ b/drivers/net/phy/backplane/backplane.c
@@ -0,0 +1,1557 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+/* Backplane driver
+ *
+ * Copyright 2015 Freescale Semiconductor, Inc.
+ * Copyright 2018-2020 NXP
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mii.h>
+#include <linux/mdio.h>
+#include <linux/ethtool.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/of_net.h>
+#include <linux/of_address.h>
+#include <linux/of_platform.h>
+#include <linux/timer.h>
+#include <linux/delay.h>
+#include <linux/workqueue.h>
+#include <linux/netdevice.h>
+#include <linux/list.h>
+
+#include "backplane.h"
+#include "link_training.h"
+
+/* KR timeouts in milliseconds */
+#define KR_TIMEOUT_1 100
+#define KR_TIMEOUT_2 1000
+#define KR_AN_TIMEOUT 3000
+#define KR_LT_TIMEOUT 500
+
+/* KR timings in interations */
+#define KR_AN_WAIT_ITERATIONS 5
+#define KR_TRAIN_STEP_ITERATIONS 2
+#define CDR_LOCK_RETRY_COUNT 3
+
+/* AN register initialization */
+#define AN_CTRL_INIT 0x1200
+
+/* AN status register (Clause 45) (MMD 7): MDIO_STAT1 */
+#define AN_LINK_UP_MASK 0x04
+
+/* Backplane Ethernet status (Register 7.48) */
+#define AN_BP_ETH_STATUS 48
+
+/* AN masks: Backplane Ethernet status: register 7.48 */
+#define AN_MASK_10GBASE_KR 0x08
+
+/* AN advertisement 1 (Register 7.17) */
+#define AN_ADVERTISEMENT_1 17
+
+/* AN advertisement initialization for register 7.17 */
+#define AN_ADV1_KR_INIT_10G 0x85
+
+/* Backplane features */
+__ETHTOOL_DECLARE_LINK_MODE_MASK(backplane_features) __ro_after_init;
+EXPORT_SYMBOL(backplane_features);
+
+const int backplane_common_features_array[] = {
+ ETHTOOL_LINK_MODE_Backplane_BIT,
+ ETHTOOL_LINK_MODE_Autoneg_BIT,
+ ETHTOOL_LINK_MODE_MII_BIT,
+};
+
+const int backplane_protocol_features_array[] = {
+ ETHTOOL_LINK_MODE_10000baseKR_Full_BIT,
+};
+
+/* map string key to pointer data */
+struct spmap_node {
+ struct list_head entry;
+ const char *key;
+ void *pdata;
+};
+
+/* registered equalization algorithms info */
+static LIST_HEAD(eqalg_list);
+
+/* lanes attached to an equalization algorithm */
+static LIST_HEAD(lnalg_list);
+
+/* Backplane mutex between all KR PHY threads */
+static struct mutex backplane_lock;
+
+static int get_backplane_speed(phy_interface_t interface)
+{
+ switch (interface) {
+ case PHY_INTERFACE_MODE_10GKR:
+ return SPEED_10000;
+ default:
+ pr_err("%s: Unsupported backplane interface\n",
+ BACKPLANE_DRIVER_NAME);
+ return SPEED_UNKNOWN;
+ }
+ return SPEED_UNKNOWN;
+}
+
+static enum ethtool_link_mode_bit_indices
+ get_backplane_supported_mode(phy_interface_t interface)
+{
+ switch (interface) {
+ case PHY_INTERFACE_MODE_10GKR:
+ return ETHTOOL_LINK_MODE_10000baseKR_Full_BIT;
+ default:
+ pr_err("%s: Unsupported backplane interface\n",
+ BACKPLANE_DRIVER_NAME);
+ return ETHTOOL_LINK_MODE_Backplane_BIT;
+ }
+ return ETHTOOL_LINK_MODE_Backplane_BIT;
+}
+
+static int spmap_add(struct list_head *list, const char *key, void *pdata)
+{
+ struct spmap_node *node;
+
+ /* create a new entry with desired key */
+ node = kzalloc(sizeof(*node), GFP_KERNEL);
+ if (!node)
+ return -ENOMEM;
+
+ node->key = key;
+ node->pdata = pdata;
+
+ list_add(&node->entry, list);
+
+ return 0;
+}
+
+static const struct equalization_algorithm *eq_find(const char *key)
+{
+ struct spmap_node *eqalg, *eqalg_tmp;
+
+ if (!key)
+ return NULL;
+
+ /* search desired single key */
+ list_for_each_entry_safe(eqalg, eqalg_tmp, &eqalg_list, entry) {
+ if (strcmp(eqalg->key, key) == 0)
+ return (struct equalization_algorithm *)eqalg->pdata;
+ }
+ return NULL;
+}
+
+static void backplane_features_init(void)
+{
+ linkmode_set_bit_array(backplane_common_features_array,
+ ARRAY_SIZE(backplane_common_features_array),
+ backplane_features);
+
+ linkmode_set_bit_array(backplane_protocol_features_array,
+ ARRAY_SIZE(backplane_protocol_features_array),
+ backplane_features);
+}
+
+static u32 le_ioread32(void __iomem *reg)
+{
+ return ioread32(reg);
+}
+
+static void le_iowrite32(u32 value, void __iomem *reg)
+{
+ iowrite32(value, reg);
+}
+
+static u32 be_ioread32(void __iomem *reg)
+{
+ return ioread32be(reg);
+}
+
+static void be_iowrite32(u32 value, void __iomem *reg)
+{
+ iowrite32be(value, reg);
+}
+
+static void training_status_init(struct training_status *trst)
+{
+ trst->done_training = false;
+ trst->remote_tx_complete = false;
+ trst->remote_tx_running = false;
+ trst->sent_init = false;
+ trst->lp_rx_ready = 0;
+ trst->local_tx_running = false;
+}
+
+static void init_kr_lane(struct lane_device *lane, bool revert_default)
+{
+ if (revert_default)
+ backplane_default_kr_lane(lane);
+
+ training_status_init(&lane->krln.trst);
+ lane->krln.state = DETECTING_LP;
+ lane->krln.an_kr_detected = false;
+
+ lane->krln.ld_update = 0;
+ lane->krln.prev_ld_update = 0;
+ lane->krln.ld_last_nonhold_update = 0;
+ lane->krln.lp_status = 0;
+ lane->krln.lp_last_change_status = 0;
+ lane->krln.last_lp_update_status[C_M1] = 0;
+ lane->krln.last_lp_update_status[C_Z0] = 0;
+ lane->krln.last_lp_update_status[C_P1] = 0;
+ lane->krln.ld_status = 0;
+ lane->krln.move_back_prev = false;
+ lane->krln.move_back_cnt = 0;
+ lane->krln.move_back_lp_status = 0;
+
+ lt_init_ld(lane);
+}
+
+static void setup_supported_linkmode(struct phy_device *phydev)
+{
+ int i;
+
+ /* Clear all supported backplane protocols features
+ * and setup only the currently configured protocol
+ */
+ for (i = 0; i < ARRAY_SIZE(backplane_protocol_features_array); i++)
+ linkmode_clear_bit(backplane_protocol_features_array[i],
+ phydev->supported);
+
+ linkmode_set_bit(get_backplane_supported_mode(phydev->interface),
+ phydev->supported);
+}
+
+/* Read AN Link Status */
+static int is_an_link_up(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ int ret, val = 0;
+
+ mutex_lock(&bpdev->bpphy_lock);
+
+ /* Read twice because Link_Status is LL (Latched Low) bit */
+ val = phy_read_mmd(phydev, MDIO_MMD_AN, MDIO_STAT1);
+ val = phy_read_mmd(phydev, MDIO_MMD_AN, MDIO_STAT1);
+
+ mutex_unlock(&bpdev->bpphy_lock);
+
+ ret = (val & AN_LINK_UP_MASK) ? 1 : 0;
+
+ return ret;
+}
+
+static void start_kr_state_machine(struct lane_device *lane, u32 timeout)
+{
+ /* Check if equalization algorithm is installed */
+ if (!lane->krln.eq_alg)
+ return;
+
+ /* Check if link training is used */
+ if (!lane->krln.eq_alg->use_local_tx_training &&
+ !lane->krln.eq_alg->use_remote_tx_training)
+ return;
+
+ queue_delayed_work(system_power_efficient_wq, &lane->krwk,
+ msecs_to_jiffies(timeout));
+}
+
+static void stop_kr_state_machine(struct lane_device *lane)
+{
+ /* Check if equalization algorithm is installed */
+ if (!lane->krln.eq_alg)
+ return;
+
+ /* Check if link training is used */
+ if (!lane->krln.eq_alg->use_local_tx_training &&
+ !lane->krln.eq_alg->use_remote_tx_training)
+ return;
+
+ cancel_delayed_work_sync(&lane->krwk);
+}
+
+static void setup_default_settings(struct lane_device *lane)
+{
+ if (lane->bpdev->bpkr.valid_eq_init)
+ lane->krln.def_kr = lane->bpdev->bpkr.def_kr;
+ else
+ lane->bpdev->drv.lane_ops->read_lane_kr(lane->reg_base,
+ &lane->krln.def_kr);
+
+ /* HW specific default settings */
+ if (lane->bpdev->drv.bp_ops.setup_default_settings)
+ lane->bpdev->drv.bp_ops.setup_default_settings(lane);
+}
+
+static void kr_reset_master_lane(struct lane_device *lane)
+{
+ const struct lane_ops *lane_ops = lane->bpdev->drv.lane_ops;
+ struct backplane_device *bpdev = lane->bpdev;
+
+ if (backplane_is_multi_lane(bpdev)) {
+ /* Reset only the Master Lane */
+ if (lane->idx == MASTER_LANE)
+ lane_ops->reset_lane(lane->reg_base, LANE_RX_TX);
+ } else {
+ lane_ops->reset_lane(lane->reg_base, LANE_RX_TX);
+ }
+}
+
+static void print_single_lane_trained(struct lane_device *lane)
+{
+ struct phy_device *phydev = lane->phydev;
+
+ phydev_info(phydev,
+ "%s link trained, Tx equalization: C(-1)=0x%x, C(0)=0x%x, C(+1)=0x%x\n",
+ phy_modes(phydev->interface),
+ lane->krln.tuned_kr.preq, lane->krln.tuned_kr.mainq,
+ lane->krln.tuned_kr.postq);
+}
+
+static void print_multi_lane_trained(struct lane_device *lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ int i;
+
+ phydev_info(phydev,
+ "%s link trained, Tx equalization:\n",
+ phy_modes(phydev->interface));
+
+ for (i = 0; i < bpdev->num_lanes; i++)
+ phydev_info(phydev,
+ "\t|- Lane %d: C(-1)=0x%x, C(0)=0x%x, C(+1)=0x%x\n",
+ i + 1, bpdev->lane[i].krln.tuned_kr.preq,
+ bpdev->lane[i].krln.tuned_kr.mainq,
+ bpdev->lane[i].krln.tuned_kr.postq);
+}
+
+static void kr_link_trained(struct lane_device *lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+
+ mutex_lock(&bpdev->trained_lock);
+ /* Setup lane state as TRAINED inside the phy trained lock
+ * to avoid duplicated message printed on multi-lane PHYs
+ */
+ lane->krln.state = TRAINED;
+
+ mutex_lock(&backplane_lock);
+
+ if (backplane_is_single_lane(bpdev))
+ print_single_lane_trained(lane);
+ else
+ if (backplane_are_all_lanes_trained(bpdev))
+ print_multi_lane_trained(lane);
+
+ mutex_unlock(&backplane_lock);
+ mutex_unlock(&bpdev->trained_lock);
+}
+
+static void kr_train_step(struct lane_device *lane)
+{
+ struct training_status *trst = &lane->krln.trst;
+ u32 lt_timeout = KR_LT_TIMEOUT;
+ u64 dead_line;
+ int i = 0;
+
+ /* Check if equalization algorithm is installed */
+ if (!lane->krln.eq_alg)
+ return;
+
+ /* Check if link training is used */
+ if (!lane->krln.eq_alg->use_local_tx_training &&
+ !lane->krln.eq_alg->use_remote_tx_training)
+ return;
+
+ lt_start(lane);
+
+ while (i < KR_TRAIN_STEP_ITERATIONS) {
+ dead_line = jiffies + msecs_to_jiffies(lt_timeout);
+ while (time_before(jiffies, (unsigned long)dead_line)) {
+ /* check if the LT is already failed */
+ if (lt_is_training_failure(lane)) {
+ /* LT failed already, reset lane to avoid
+ * it run into hanging, then start LT again.
+ */
+ kr_reset_master_lane(lane);
+ lt_start(lane);
+ } else if (lt_is_frame_lock(lane)) {
+ break;
+ }
+ /* wait frame lock (without training_failure) */
+ usleep_range(100, 500);
+ }
+
+ if (!lt_is_frame_lock(lane)) {
+ i++;
+ continue;
+ }
+
+ /* the LT should be finished in 500ms, failed or OK. */
+ dead_line = jiffies + msecs_to_jiffies(lt_timeout);
+ while (time_before(jiffies, (unsigned long)dead_line)) {
+ /* check if the LT is already failed */
+ if (lt_is_training_failure(lane)) {
+ kr_reset_master_lane(lane);
+ break;
+ }
+
+ if (lane->krln.eq_alg->use_local_tx_training)
+ lt_train_local_tx(lane);
+
+ if (lane->krln.eq_alg->use_remote_tx_training)
+ lt_train_remote_tx(lane);
+
+ if (lane->krln.lt_error)
+ break;
+
+ if (trst->lp_rx_ready && trst->remote_tx_complete)
+ break;
+
+ usleep_range(100, 500);
+ }
+
+ i++;
+ /* check if LT Error occurred */
+ if (lane->krln.lt_error) {
+ init_kr_lane(lane, false);
+ continue;
+ } else {
+ break;
+ }
+ }
+
+ lt_stop(lane);
+
+ /* check if Link is successfully TRAINED */
+ if (lt_is_rx_trained(lane))
+ kr_link_trained(lane);
+ else
+ kr_reset_master_lane(lane);
+}
+
+static void an_init(struct lane_device *masterln)
+{
+ struct backplane_device *bpdev = masterln->bpdev;
+ struct phy_device *phydev = masterln->phydev;
+ struct lane_device *lane;
+ u32 init_an_adv1;
+ int i, err;
+
+ if (!backplane_is_mode_kr(phydev->interface))
+ return;
+
+ if (masterln->idx != MASTER_LANE)
+ return;
+
+ init_an_adv1 = backplane_get_an_adv1_init(phydev->interface);
+
+ /* setup AN init on each lane */
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ lane = &bpdev->lane[i];
+ if (bpdev->drv.bp_ops.an_advertisement_init) {
+ bpdev->drv.bp_ops.an_advertisement_init(lane);
+ } else {
+ err = backplane_write_mmd(lane, MDIO_MMD_AN,
+ AN_ADVERTISEMENT_1,
+ init_an_adv1);
+ if (err)
+ phydev_err(phydev,
+ "Setting AN register 0x%02x on lane %d failed with error code: 0x%08x\n",
+ AN_ADVERTISEMENT_1, lane->idx, err);
+ }
+ }
+
+ udelay(1);
+
+ err = backplane_write_mmd(masterln, MDIO_MMD_AN, MDIO_CTRL1,
+ AN_CTRL_INIT);
+ if (err)
+ phydev_err(phydev,
+ "Setting AN register 0x%02x on Master Lane failed with error code: 0x%08x\n",
+ MDIO_CTRL1, err);
+}
+
+static void an_request_restart(struct lane_device *lane)
+{
+ const struct lane_ops *lane_ops = lane->bpdev->drv.lane_ops;
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ int i;
+
+ if (time_before(jiffies, (unsigned long)lane->krln.an_kr_timeout))
+ return;
+ if (!backplane_is_mode_kr(phydev->interface))
+ return;
+
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ init_kr_lane(&bpdev->lane[i], true);
+ /* Reset the lane to recover from link down */
+ lane_ops->reset_lane(bpdev->lane[i].reg_base, LANE_RX_TX);
+ lt_reset(&bpdev->lane[i]);
+ }
+ /* AN init only for Master Lane */
+ an_init(&bpdev->lane[MASTER_LANE]);
+
+ lane->krln.an_kr_timeout = jiffies + msecs_to_jiffies(KR_AN_TIMEOUT);
+}
+
+static bool detect_lp(struct lane_device *lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ struct lane_device *masterln;
+ bool start_train = false;
+ bool an_kr_link;
+ int an_state;
+ u32 an_mask;
+
+ if (bpdev->drv.bp_ops.is_an_link_detected) {
+ an_kr_link = bpdev->drv.bp_ops.is_an_link_detected(lane);
+ } else {
+ /* Check AN state only on Master Lane */
+ masterln = &bpdev->lane[MASTER_LANE];
+ an_mask = backplane_get_an_bp_eth_status_bit(phydev->interface);
+ an_state = backplane_read_mmd(masterln, MDIO_MMD_AN,
+ AN_BP_ETH_STATUS);
+ an_kr_link = an_state & an_mask;
+ }
+
+ /* The link training occurs after auto-negotiation
+ * has determined the link to be a Base-KR link.
+ * This is indicated by asserting the corresponding bit.
+ * This occurs before auto-negotiation can declare auto-negotiation
+ * complete, as this requires the PCS to report a valid link.
+ */
+ if (an_kr_link) {
+ /* AN acquired:
+ * Train all lanes in order starting with Master Lane
+ */
+ lane->krln.an_kr_detected = true;
+ lane->krln.an_kr_wait_count = 0;
+ start_train = true;
+ } else {
+ /* AN lost or not yet acquired */
+ if (lane->krln.an_kr_detected) {
+ /* AN acquired first time but now was lost */
+ if (!backplane_is_link_up(phydev)) {
+ /* Link is down: restart training */
+ lane->krln.an_kr_wait_count = 0;
+ an_request_restart(lane);
+ } else {
+ /* Link is up:
+ * wait few iterations for AN to be acquired
+ */
+ if (lane->krln.an_kr_wait_count >=
+ KR_AN_WAIT_ITERATIONS) {
+ lane->krln.an_kr_wait_count = 0;
+ an_request_restart(lane);
+ } else {
+ lane->krln.an_kr_wait_count++;
+ }
+ }
+ }
+ /* else: AN was not yet acquired first time
+ * DO nothing, just wait AN to be acquired first time
+ */
+ }
+
+ return start_train;
+}
+
+static void detect_hotplug(struct lane_device *lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ int i;
+
+ if (lane->idx == MASTER_LANE) {
+ /* check if all lanes are trained
+ * only if current lane is Master Lane
+ */
+ if (backplane_are_all_lanes_trained(bpdev)) {
+ phydev_info(phydev, "Detect hotplug, restart training\n");
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ /* initializations on Detect hotplug / restart:
+ * they must not be part of init_kr_lane
+ */
+ bpdev->lane[i].krln.first_recv_init = false;
+ }
+ an_request_restart(lane);
+ }
+ }
+}
+
+static void bp_kr_state_machine(struct work_struct *work)
+{
+ struct delayed_work *dwork = to_delayed_work(work);
+ struct backplane_device *bpdev;
+ u32 kr_timeout = KR_TIMEOUT_1;
+ struct phy_device *phydev;
+ struct lane_device *lane;
+ bool start_train = false;
+
+ lane = container_of(dwork, struct lane_device, krwk);
+ if (!lane)
+ return;
+
+ bpdev = lane->bpdev;
+ phydev = lane->phydev;
+
+ if (!backplane_is_mode_kr(phydev->interface))
+ return;
+
+ /* Check if equalization algorithm is installed */
+ if (!lane->krln.eq_alg)
+ return;
+
+ /* Check if link training is used */
+ if (!lane->krln.eq_alg->use_local_tx_training &&
+ !lane->krln.eq_alg->use_remote_tx_training)
+ return;
+
+ mutex_lock(&lane->lane_lock);
+ switch (lane->krln.state) {
+ case DETECTING_LP:
+ start_train = detect_lp(lane);
+ break;
+ case TRAINED:
+ kr_timeout = KR_TIMEOUT_2;
+ if (!backplane_is_link_up(phydev)) {
+ kr_timeout = KR_TIMEOUT_1;
+ detect_hotplug(lane);
+ }
+ break;
+ }
+
+ if (start_train)
+ kr_train_step(lane);
+
+ mutex_unlock(&lane->lane_lock);
+ start_kr_state_machine(lane, kr_timeout);
+}
+
+static void init_kr_state_machine(struct lane_device *lane)
+{
+ /* Check if equalization algorithm is installed */
+ if (!lane->krln.eq_alg)
+ return;
+
+ /* Check if link training is used */
+ if (!lane->krln.eq_alg->use_local_tx_training &&
+ !lane->krln.eq_alg->use_remote_tx_training)
+ return;
+
+ INIT_DELAYED_WORK(&lane->krwk, bp_kr_state_machine);
+}
+
+/* backplane_get_current_taps
+ * convert coefficient taps from internal backplane driver to link training
+ */
+void backplane_get_current_taps(struct lane_device *lane, u32 *coef)
+{
+ coef[C_M1] = lane->krln.crt_kr.preq;
+ coef[C_Z0] = lane->krln.crt_kr.mainq;
+ coef[C_P1] = lane->krln.crt_kr.postq;
+}
+
+/* backplane_set_current_taps
+ * convert coefficient taps from link training to internal backplane driver
+ */
+void backplane_set_current_taps(struct lane_device *lane, u32 *coef)
+{
+ lane->krln.crt_kr.preq = coef[C_M1];
+ lane->krln.crt_kr.mainq = coef[C_Z0];
+ lane->krln.crt_kr.postq = coef[C_P1];
+}
+
+/* backplane_set_all_taps_to_max
+ * setup all coefficients to MAX values from IEEE802.3ap perspective
+ */
+void backplane_set_all_taps_to_max(struct lane_device *lane)
+{
+ lane->krln.crt_kr = lane->bpdev->bpkr.max_kr;
+}
+
+void backplane_tune_kr_lane(struct lane_device *lane, bool reset_lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ bool reset = false;
+
+ if (backplane_is_multi_lane(bpdev)) {
+ /* Reset only the Master Lane */
+ reset = (lane->idx == MASTER_LANE);
+ } else {
+ reset = true;
+ }
+
+ /* Do not reset the lane if this is how it was asked */
+ if (!reset_lane)
+ reset = false;
+
+ lane->bpdev->drv.lane_ops->tune_lane_kr(lane->reg_base,
+ &lane->krln.crt_kr, reset);
+ lane->krln.tuned_kr = lane->krln.crt_kr;
+}
+
+void backplane_default_kr_lane(struct lane_device *lane)
+{
+ lane->krln.crt_kr = lane->krln.def_kr;
+
+ backplane_tune_kr_lane(lane, true);
+}
+
+/* backplane_write_mmd - Wrapper function for phy_write_mmd
+ * for writing a register on an MMD on a given PHY.
+ *
+ * Same rules as for phy_write_mmd();
+ */
+int backplane_write_mmd(struct lane_device *lane, int devad, u32 regnum,
+ u16 val)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ int err;
+
+ mutex_lock(&bpdev->bpphy_lock);
+
+ err = phy_write_mmd(phydev, devad, regnum, val);
+ if (err)
+ phydev_err(phydev,
+ "Writing PHY (%p) MMD = 0x%02x register = 0x%02x failed with error code: 0x%08x\n",
+ phydev, devad, regnum, err);
+
+ mutex_unlock(&bpdev->bpphy_lock);
+
+ return err;
+}
+EXPORT_SYMBOL(backplane_write_mmd);
+
+/* backplane_read_mmd - Wrapper function for phy_read_mmd
+ * for reading a register from an MMD on a given PHY.
+ *
+ * Same rules as for phy_read_mmd();
+ */
+int backplane_read_mmd(struct lane_device *lane, int devad, u32 regnum)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ int ret;
+
+ mutex_lock(&bpdev->bpphy_lock);
+
+ ret = phy_read_mmd(phydev, devad, regnum);
+
+ mutex_unlock(&bpdev->bpphy_lock);
+
+ return ret;
+}
+EXPORT_SYMBOL(backplane_read_mmd);
+
+/* backplane_eq_register
+ *
+ * Registers an equalization algorithm with the specified key
+ *
+ * key: desired key on which eq algorithm must be registered
+ * eq_info: eq algorithm information to be registered
+ *
+ * Returns: Zero for success or error code in case of failure
+ */
+int backplane_eq_register(const char *key,
+ const struct equalization_algorithm *eq_info)
+{
+ struct spmap_node *eqalg, *eqalg_tmp;
+
+ /* check if desired key already exists */
+ list_for_each_entry_safe(eqalg, eqalg_tmp, &eqalg_list, entry) {
+ if (strcmp(eqalg->key, key) == 0) {
+ pr_err("%s: Equalization algorithm registration failed: key '%s' already exists\n",
+ BACKPLANE_DRIVER_NAME, key);
+ return -EEXIST;
+ }
+ }
+
+ spmap_add(&eqalg_list, key, (void *)eq_info);
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_eq_register);
+
+/* backplane_eq_unregister
+ *
+ * Unregisters all equalization algorithm for the specified key
+ *
+ * key: desired key for which all registered eq algorithms must be removed
+ *
+ * Returns: None
+ */
+void backplane_eq_unregister(const char *key)
+{
+ const struct equalization_algorithm *eq_alg;
+ struct spmap_node *node, *node_tmp;
+ struct lane_device *lane;
+
+ if (!key)
+ return;
+
+ /* search all keys in lanes list */
+ list_for_each_entry_safe(node, node_tmp, &lnalg_list, entry) {
+ if (strcmp(node->key, key) == 0) {
+ lane = (struct lane_device *)node->pdata;
+ eq_alg = lane->krln.eq_alg;
+ if (eq_alg->ops.destroy)
+ eq_alg->ops.destroy(lane->krln.eq_priv);
+ lane->krln.eq_alg = NULL;
+ lane->krln.eq_priv = NULL;
+ list_del_init(&node->entry);
+ kfree(node);
+ }
+ }
+
+ /* search single key in eq algorithms list */
+ list_for_each_entry_safe(node, node_tmp, &eqalg_list, entry) {
+ if (strcmp(node->key, key) == 0) {
+ list_del_init(&node->entry);
+ kfree(node);
+ break;
+ }
+ }
+}
+EXPORT_SYMBOL(backplane_eq_unregister);
+
+bool backplane_is_mode_kr(phy_interface_t interface)
+{
+ return (interface == PHY_INTERFACE_MODE_10GKR);
+}
+EXPORT_SYMBOL(backplane_is_mode_kr);
+
+bool backplane_is_valid_mode(phy_interface_t interface)
+{
+ return backplane_is_mode_kr(interface);
+}
+EXPORT_SYMBOL(backplane_is_valid_mode);
+
+u8 backplane_num_lanes(phy_interface_t interface)
+{
+ const char *bp_name;
+ char num_lanes;
+ int len;
+
+ if (!backplane_is_valid_mode(interface))
+ return 0;
+
+ bp_name = phy_modes(interface);
+ if (!bp_name)
+ return 0;
+ if (strcmp(bp_name, "unknown") == 0)
+ return 0;
+
+ len = strlen(bp_name);
+ if (len == 0)
+ return 0;
+ num_lanes = bp_name[len - 1];
+ if (num_lanes >= '0' && num_lanes <= '9')
+ return num_lanes - '0';
+
+ return 1;
+}
+EXPORT_SYMBOL(backplane_num_lanes);
+
+/* Backplane Ethernet Status Register Register 7.48 (an_bp_eth_status)
+ * - AN_MASK_10GBASE_KR for 10GBase-KR
+ */
+u32 backplane_get_an_bp_eth_status_bit(phy_interface_t interface)
+{
+ switch (interface) {
+ case PHY_INTERFACE_MODE_10GKR:
+ return AN_MASK_10GBASE_KR;
+ /* add AN support for other backplane modes here */
+ default:
+ pr_err("%s: Unsupported backplane interface\n",
+ BACKPLANE_DRIVER_NAME);
+ return 0;
+ }
+ return 0;
+}
+EXPORT_SYMBOL(backplane_get_an_bp_eth_status_bit);
+
+u32 backplane_get_an_adv1_init(phy_interface_t interface)
+{
+ switch (interface) {
+ case PHY_INTERFACE_MODE_10GKR:
+ return AN_ADV1_KR_INIT_10G;
+ /* add AN support for other backplane modes here */
+ default:
+ pr_err("%s: Unsupported backplane interface\n",
+ BACKPLANE_DRIVER_NAME);
+ return 0;
+ }
+ return 0;
+}
+EXPORT_SYMBOL(backplane_get_an_adv1_init);
+
+void backplane_kr_lt_mmd_c45(struct backplane_kr *bpkr)
+{
+ lt_mmd_c45(bpkr);
+}
+EXPORT_SYMBOL(backplane_kr_lt_mmd_c45);
+
+void backplane_kr_lt_mmd_setup(struct backplane_kr *bpkr, int devad, u32 base)
+{
+ lt_mmd_setup(bpkr, devad, base);
+}
+EXPORT_SYMBOL(backplane_kr_lt_mmd_setup);
+
+bool backplane_is_single_lane(struct backplane_device *bpdev)
+{
+ return (bpdev->num_lanes == 1);
+}
+EXPORT_SYMBOL(backplane_is_single_lane);
+
+bool backplane_is_multi_lane(struct backplane_device *bpdev)
+{
+ return (bpdev->num_lanes > 1);
+}
+EXPORT_SYMBOL(backplane_is_multi_lane);
+
+/* backplane_is_cdr_lock
+ *
+ * Checks clock and data recovery bit: CDR Lock
+ *
+ * lane: desired lane to be verified
+ * retry: boolean value that specifies if to retry the check
+ *
+ * Returns: true if CDR_Lock bit is asserted or false otherwise
+ */
+bool backplane_is_cdr_lock(struct lane_device *lane, bool retry)
+{
+ const struct lane_ops *lane_ops = lane->bpdev->drv.lane_ops;
+ int i;
+
+ if (lane_ops->is_cdr_lock(lane->reg_base))
+ return true;
+
+ if (!retry)
+ return false;
+
+ /* Try RX_RESET: Allow for few retries */
+ for (i = 0; i < CDR_LOCK_RETRY_COUNT; i++) {
+ lane_ops->reset_lane(lane->reg_base, LANE_RX);
+ usleep_range(10, 50);
+
+ if (lane_ops->is_cdr_lock(lane->reg_base))
+ return true;
+ }
+ return false;
+}
+EXPORT_SYMBOL(backplane_is_cdr_lock);
+
+/* backplane_is_link_up
+ * Generic Link-up Status: use AN link-up
+ */
+int backplane_is_link_up(struct phy_device *phydev)
+{
+ return is_an_link_up(phydev);
+}
+EXPORT_SYMBOL(backplane_is_link_up);
+
+int backplane_get_lanes_trained_count(struct backplane_device *bpdev)
+{
+ int i, lanes_trained = 0;
+
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ if (bpdev->lane[i].krln.state == TRAINED)
+ lanes_trained++;
+ }
+ return lanes_trained;
+}
+EXPORT_SYMBOL(backplane_get_lanes_trained_count);
+
+int backplane_are_all_lanes_trained(struct backplane_device *bpdev)
+{
+ int i;
+
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ if (bpdev->lane[i].krln.state != TRAINED)
+ return 0;
+ }
+ return 1;
+}
+EXPORT_SYMBOL(backplane_are_all_lanes_trained);
+
+int backplane_create(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev;
+ struct device_node *dev_node;
+
+ dev_node = phydev->mdio.dev.of_node;
+ if (!dev_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+
+ /* allocate memory for backplane info structure */
+ bpdev = devm_kzalloc(&phydev->mdio.dev, sizeof(*bpdev), GFP_KERNEL);
+ if (!bpdev)
+ return -ENOMEM;
+
+ bpdev->phydev = phydev;
+
+ /* save bpdev as phy private data pointer */
+ phydev->priv = bpdev;
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_create);
+
+/* backplane_parse_dt
+ * parses the device tree and saves backplane relevant data
+ * in backplane phy info structure
+ */
+int backplane_parse_dt(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ struct device_node *dev_node;
+ u32 eqinit[C_NO];
+ const char *eqa;
+ int proplen;
+ int ret;
+
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ dev_node = phydev->mdio.dev.of_node;
+ if (!dev_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+
+ if (!backplane_is_valid_mode(phydev->interface))
+ return -EINVAL;
+
+ ret = of_property_read_string(dev_node, "eq-algorithm", &eqa);
+ /* if eq-algorithm node is not found then use the default algorithm */
+ if (ret == 0)
+ bpdev->bpkr.eqa_name = eqa;
+ else
+ bpdev->bpkr.eqa_name = DEFAULT_EQ_ALGORITHM;
+
+ /* if eq-init node exists then use the DTS specified values
+ * if eq-init node doesn't exist then use values already found in HW
+ */
+ proplen = of_property_count_u32_elems(dev_node, "eq-init");
+ if (proplen > 0) {
+ /* There are 3 standard equalization coefficient taps */
+ if (proplen > C_NO)
+ proplen = C_NO;
+ ret = of_property_read_u32_array(dev_node, "eq-init",
+ (u32 *)eqinit, proplen);
+ if (ret == 0) {
+ bpdev->bpkr.valid_eq_init = true;
+ bpdev->bpkr.def_kr.preq = eqinit[C_M1];
+ bpdev->bpkr.def_kr.mainq = eqinit[C_Z0];
+ bpdev->bpkr.def_kr.postq = eqinit[C_P1];
+ }
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_parse_dt);
+
+/* backplane_setup_memio
+ * Setup memory I/O access
+ */
+int backplane_setup_memio(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ /* setup ioread/iowrite according to endianness */
+ if (bpdev->drv.is_little_endian) {
+ bpdev->drv.io.read32 = le_ioread32;
+ bpdev->drv.io.write32 = le_iowrite32;
+ } else {
+ bpdev->drv.io.read32 = be_ioread32;
+ bpdev->drv.io.write32 = be_iowrite32;
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_setup_memio);
+
+/* backplane_setup_mmd
+ * Setup default MMD (MDIO Managed Device)
+ */
+int backplane_setup_mmd(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ /* By default setup LT MMD Clause 45 */
+ backplane_kr_lt_mmd_c45(&bpdev->bpkr);
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_setup_mmd);
+
+/* backplane_setup_lanes
+ * Allocates lanes memory map and setup lanes relevant data
+ * Requires:
+ * - backplane_driver#lane_ops
+ * for lane access operations
+ * - backplane_driver#lane_ops#memmap_size
+ * for lane memory map allocation
+ * - backplane_kr#equalizer
+ * for specific Equalizer access
+ * - backplane_kr#cx_def
+ * for default coefficient setup
+ */
+int backplane_setup_lanes(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ const struct equalization_algorithm *eq_alg;
+ struct equalizer_driver eqdrv;
+ struct lane_device *lane;
+ int i;
+
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+ if (!bpdev->drv.lane_ops) {
+ phydev_err(phydev, "Backplane lane ops is not set\n");
+ return -EINVAL;
+ }
+ if (!bpdev->bpkr.equalizer) {
+ phydev_err(phydev, "Backplane equalizer info is not set\n");
+ return -EINVAL;
+ }
+ if (bpdev->drv.lane_ops->memmap_size == 0) {
+ phydev_err(phydev, "Lane memory map size is zero\n");
+ return -EINVAL;
+ }
+
+ bpdev->num_lanes = backplane_num_lanes(phydev->interface);
+ if (bpdev->num_lanes > MAX_KR_LANES_PER_PHY) {
+ phydev_err(phydev, "Unsupported number of lanes per phy: %d\n",
+ bpdev->num_lanes);
+ return -EINVAL;
+ }
+
+ if (backplane_is_mode_kr(phydev->interface)) {
+ if (bpdev->bpkr.valid_eq_init &&
+ bpdev->bpkr.def_kr.preq == 0 &&
+ bpdev->bpkr.def_kr.mainq == 0 &&
+ bpdev->bpkr.def_kr.postq == 0)
+ phydev_warn(phydev,
+ "All KR default values from DT are zero\n");
+ }
+
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ lane = &bpdev->lane[i];
+
+ /* setup lane memory map size */
+ lane->memmap_size = bpdev->drv.lane_ops->memmap_size;
+
+ lane->reg_base = devm_ioremap(&phydev->mdio.dev,
+ lane->lane_addr,
+ lane->memmap_size);
+ if (!lane->reg_base) {
+ phydev_err(phydev, "Lane memory map allocation failed\n");
+ return -ENOMEM;
+ }
+
+ lane->idx = i;
+ lane->phydev = phydev;
+ lane->bpdev = bpdev;
+ lane->krln.an_kr_timeout =
+ jiffies + msecs_to_jiffies(KR_AN_TIMEOUT);
+
+ if (backplane_is_mode_kr(phydev->interface)) {
+ setup_default_settings(lane);
+
+ /* Find EQ Algorithm info */
+ eq_alg = eq_find(bpdev->bpkr.eqa_name);
+ if (!eq_alg) {
+ /* key for desired algorithm was not found */
+ phydev_err(phydev,
+ "Equalization algorithm '%s' is not registered\n",
+ bpdev->bpkr.eqa_name);
+ return -EINVAL;
+ }
+ if (!eq_alg->ops.create) {
+ phydev_err(phydev,
+ "Equalization algorithm creation failed: create operation is not available\n");
+ return -EINVAL;
+ }
+ lane->krln.eq_alg = eq_alg;
+
+ /* Setup EQ Algorithm */
+ eqdrv.lane = lane;
+ eqdrv.phydev = lane->phydev;
+ eqdrv.reg_base = lane->reg_base;
+ eqdrv.equalizer = lane->bpdev->bpkr.equalizer;
+
+ /* Create EQ Algorithm */
+ lane->krln.eq_priv = eq_alg->ops.create(eqdrv);
+
+ /* register lane attached to an algorithm */
+ spmap_add(&lnalg_list, bpdev->bpkr.eqa_name, lane);
+
+ if (eq_alg->use_remote_tx_training) {
+ if (!eq_alg->ops.is_rx_ok)
+ phydev_warn(phydev,
+ "Required operation for remote Tx training is missing: is_rx_ok\n");
+ if (!eq_alg->ops.is_eq_done)
+ phydev_warn(phydev,
+ "Required operation for remote Tx training is missing: is_eq_done\n");
+ if (!eq_alg->ops.collect_statistics)
+ phydev_warn(phydev,
+ "Required operation for remote Tx training is missing: collect_statistics\n");
+ if (!eq_alg->ops.generate_request)
+ phydev_warn(phydev,
+ "Required operation for remote Tx training is missing: generate_request\n");
+ }
+ }
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_setup_lanes);
+
+/* backplane_initialize
+ * Initializes all PHY and lane mutexes and
+ * starts lane timers for running the algorithm
+ */
+int backplane_initialize(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ int i;
+
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ mutex_init(&bpdev->bpphy_lock);
+ mutex_init(&bpdev->trained_lock);
+
+ for (i = 0; i < bpdev->num_lanes; i++)
+ mutex_init(&bpdev->lane[i].lane_lock);
+
+ phydev->speed = get_backplane_speed(phydev->interface);
+ if (phydev->speed < 0) {
+ phydev_err(phydev, "Unsupported backplane mode\n");
+ return -EINVAL;
+ }
+
+ if (backplane_is_mode_kr(phydev->interface)) {
+ for (i = 0; i < bpdev->num_lanes; i++)
+ init_kr_state_machine(&bpdev->lane[i]);
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_initialize);
+
+/* backplane_probe
+ *
+ * Probe function for backplane driver to provide generic device behavior
+ *
+ * phydev: backplane phy device
+ * this is an internal phy block controlled by the software
+ * which contains other component blocks like: PMA/PMD, PCS, AN
+ *
+ * Return: Zero for success or error code in case of failure
+ */
+int backplane_probe(struct phy_device *phydev)
+{
+ return backplane_create(phydev);
+}
+EXPORT_SYMBOL(backplane_probe);
+
+void backplane_remove(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return;
+ }
+
+ kfree(bpdev);
+ phydev->priv = NULL;
+}
+EXPORT_SYMBOL(backplane_remove);
+
+/* backplane_config_init
+ *
+ * Config_Init function for backplane driver to provide generic device behavior
+ *
+ * phydev: backplane phy device
+ *
+ * Return: Zero for success or error code in case of failure
+ */
+int backplane_config_init(struct phy_device *phydev)
+{
+ int ret;
+
+ ret = backplane_parse_dt(phydev);
+ if (ret)
+ return ret;
+
+ ret = backplane_setup_memio(phydev);
+ if (ret)
+ return ret;
+
+ ret = backplane_setup_mmd(phydev);
+ if (ret)
+ return ret;
+
+ ret = backplane_setup_lanes(phydev);
+ if (ret)
+ return ret;
+
+ ret = backplane_initialize(phydev);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_config_init);
+
+int backplane_aneg_done(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+
+ if (!phydev->mdio.dev.of_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ bpdev->aneg_done = true;
+ phydev->state = PHY_RUNNING;
+
+ return 1;
+}
+EXPORT_SYMBOL(backplane_aneg_done);
+
+int backplane_config_aneg(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ struct equalization_ops ops;
+ struct lane_device *lane;
+ int i;
+
+ if (!phydev->mdio.dev.of_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+ if (backplane_get_lanes_trained_count(bpdev) > 0) {
+ phydev_err(phydev, "Incorrectly trained lanes detected\n");
+ return -EINVAL;
+ }
+
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ lane = &bpdev->lane[i];
+ if (lane->krln.eq_alg) {
+ ops = lane->krln.eq_alg->ops;
+ if (ops.dump_algorithm_context)
+ ops.dump_algorithm_context(lane->krln.eq_priv);
+ }
+ }
+
+ if (backplane_is_mode_kr(phydev->interface)) {
+ /* Warning:
+ * Order of the operations below is important
+ * otherwise the training may be failing
+ * with error: 'link_training_failed'
+ */
+
+ /* setup all lanes to default */
+ for (i = 0; i < bpdev->num_lanes; i++)
+ setup_default_settings(&bpdev->lane[i]);
+
+ /* Initialize all lanes and reset LT */
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ init_kr_lane(&bpdev->lane[i], true);
+ lt_reset(&bpdev->lane[i]);
+ }
+ }
+
+ /* Warning:
+ * speed and protocol setup operation
+ * must be done just before AN and state machine start
+ * otherwise if it is done earlier,
+ * the error: 'REQ Timeout' will occur
+ */
+ /* setup supported speed and protocol */
+ phydev->speed = get_backplane_speed(phydev->interface);
+ if (phydev->speed < 0) {
+ phydev_err(phydev, "Unsupported backplane mode\n");
+ return -EINVAL;
+ }
+
+ setup_supported_linkmode(phydev);
+ linkmode_copy(phydev->advertising, phydev->supported);
+ phydev->duplex = DUPLEX_FULL;
+
+ if (backplane_is_mode_kr(phydev->interface)) {
+ /* AN init only for Master Lane */
+ an_init(&bpdev->lane[MASTER_LANE]);
+ /* start state machine on all lanes */
+ for (i = 0; i < bpdev->num_lanes; i++)
+ start_kr_state_machine(&bpdev->lane[i], KR_TIMEOUT_1);
+ }
+
+ bpdev->aneg_config = true;
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_config_aneg);
+
+int backplane_suspend(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ int i;
+
+ if (!phydev->mdio.dev.of_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ if (bpdev->aneg_config && !bpdev->phy_suspended) {
+ if (backplane_is_mode_kr(phydev->interface)) {
+ for (i = 0; i < bpdev->num_lanes; i++)
+ stop_kr_state_machine(&bpdev->lane[i]);
+ }
+ bpdev->phy_suspended = true;
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_suspend);
+
+int backplane_resume(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+ int i;
+
+ if (!phydev->mdio.dev.of_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ if (bpdev->aneg_config && bpdev->phy_suspended) {
+ if (backplane_is_mode_kr(phydev->interface)) {
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ init_kr_lane(&bpdev->lane[i], true);
+ start_kr_state_machine(&bpdev->lane[i],
+ KR_TIMEOUT_1);
+ }
+ }
+ bpdev->phy_suspended = false;
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_resume);
+
+int backplane_read_status(struct phy_device *phydev)
+{
+ struct backplane_device *bpdev = phydev->priv;
+
+ if (!phydev->mdio.dev.of_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ /* Linkup method proposal for training stability:
+ * Don't raise linkup until all lanes are trained
+ * in order to prevent interface sending packets that may
+ * interfere with the training packets
+ */
+ if (backplane_is_link_up(phydev))
+ if (backplane_is_mode_kr(phydev->interface))
+ phydev->link = backplane_are_all_lanes_trained(bpdev);
+ else
+ phydev->link = 1;
+ else
+ phydev->link = 0;
+
+ return 0;
+}
+EXPORT_SYMBOL(backplane_read_status);
+
+int backplane_match_phy_device(struct phy_device *phydev)
+{
+ struct device_node *dev_node;
+
+ if (!phydev->is_c45)
+ return 0;
+
+ dev_node = phydev->mdio.dev.of_node;
+ if (!dev_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return 0;
+ }
+
+ return 1;
+}
+EXPORT_SYMBOL(backplane_match_phy_device);
+
+static int __init backplane_module_init(void)
+{
+ pr_info("%s: Backplane driver version %s\n",
+ BACKPLANE_DRIVER_NAME, BACKPLANE_DRIVER_VERSION);
+
+ mutex_init(&backplane_lock);
+ backplane_features_init();
+
+ return 0;
+}
+
+static void __exit backplane_module_exit(void)
+{
+ pr_info("%s: Backplane driver version %s unloaded\n",
+ BACKPLANE_DRIVER_NAME, BACKPLANE_DRIVER_VERSION);
+}
+
+module_init(backplane_module_init);
+module_exit(backplane_module_exit);
+
+MODULE_DESCRIPTION("Backplane driver");
+MODULE_AUTHOR("Florinel Iordache <florinel.iordache@nxp.com>");
+MODULE_LICENSE("Dual BSD/GPL");
diff --git a/drivers/net/phy/backplane/backplane.h b/drivers/net/phy/backplane/backplane.h
new file mode 100644
index 0000000..d142e83
--- /dev/null
+++ b/drivers/net/phy/backplane/backplane.h
@@ -0,0 +1,293 @@
+/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */
+/* Backplane driver
+ *
+ * Copyright 2018-2020 NXP
+ */
+
+#ifndef __BACKPLANE_H
+#define __BACKPLANE_H
+
+#include <linux/phy.h>
+#include <linux/mutex.h>
+
+#include "equalization.h"
+
+/* Backplane Driver name */
+#define BACKPLANE_DRIVER_NAME "backplane"
+
+/* Backplane Driver version */
+#define BACKPLANE_DRIVER_VERSION "1.0.0"
+
+/* Maximum number of lanes per phy */
+#define MAX_KR_LANES_PER_PHY 1
+
+/* Lanes definitions */
+#define MASTER_LANE 0
+#define SINGLE_LANE 0
+
+/* Number of device specific kr coefficients (device specific extension) */
+#define DEVICE_KR_COEF_NO 6
+
+extern __ETHTOOL_DECLARE_LINK_MODE_MASK(backplane_features) __ro_after_init;
+
+#define BACKPLANE_FEATURES ((unsigned long *)&backplane_features)
+
+enum train_state {
+ DETECTING_LP,
+ TRAINED,
+};
+
+enum lane_req {
+ LANE_INVALID,
+ LANE_RX,
+ LANE_TX,
+ LANE_RX_TX
+};
+
+struct kr_coef {
+ u32 preq;
+ u32 mainq;
+ u32 postq;
+ /* device specific extension */
+ u32 dev_coef[DEVICE_KR_COEF_NO];
+};
+
+/* Endianness specific memory I/O */
+struct mem_io {
+ u32 (*read32)(void __iomem *addr);
+ void (*write32)(u32 value, void __iomem *addr);
+};
+
+/* Generic Lane operations */
+struct lane_ops {
+ /* device specific private extension */
+ const void *priv;
+ /* lane memory map size */
+ u32 memmap_size;
+ void (*reset_lane)(void __iomem *reg, enum lane_req req);
+ void (*tune_lane_kr)(void __iomem *reg, struct kr_coef *coef,
+ bool reset);
+ void (*read_lane_kr)(void __iomem *reg, struct kr_coef *coef);
+ bool (*is_cdr_lock)(void __iomem *reg);
+};
+
+struct training_status {
+ bool done_training;
+ bool remote_tx_complete;
+ bool remote_tx_running;
+ bool sent_init;
+ bool lp_rx_ready;
+ bool local_tx_running;
+};
+
+struct backplane_device;
+
+/* Lane kr */
+struct lane_kr {
+ /* KR parameters (current, default, tuned) */
+ struct kr_coef crt_kr;
+ struct kr_coef def_kr;
+ struct kr_coef tuned_kr;
+ /* equalization */
+ const struct equalization_algorithm *eq_alg;
+ struct eq_data_priv *eq_priv;
+ /* training status */
+ struct training_status trst;
+ enum train_state state;
+ /* AN KR status */
+ bool an_kr_detected;
+ u32 an_kr_wait_count;
+ u64 an_kr_timeout;
+ /* KR LD/LP updates and status */
+ u32 ld_update;
+ u32 prev_ld_update;
+ /* last change (non-hold) update */
+ u32 ld_last_nonhold_update;
+ u32 ld_status;
+ u32 lp_status;
+ /* last change (non-zero) status */
+ u32 lp_last_change_status;
+ u32 last_lp_update_status[C_NO];
+ /* link training status */
+ bool lt_error;
+ u32 req_ld_update_init_count;
+ u32 repeat_request_count;
+ u64 init_handshake_time;
+ bool first_recv_init;
+ /* move lp back */
+ bool move_back_prev;
+ u32 move_back_cnt;
+ u32 move_back_lp_status;
+};
+
+/* Lane device */
+struct lane_device {
+ /* lane memory map: registers base address */
+ void __iomem *reg_base;
+ /* lane memory map size */
+ u32 memmap_size;
+ /* lane address */
+ u32 lane_addr;
+ /* lane relative index inside multi-lane PHY */
+ u8 idx;
+ /* device specific private extension */
+ void *priv;
+ /* phy device */
+ struct phy_device *phydev;
+ struct backplane_device *bpdev;
+ struct lane_kr krln;
+ struct delayed_work krwk;
+ /* mutex between multiple lanes */
+ struct mutex lane_lock;
+};
+
+/* KR LT MMD (MDIO Managed Device) */
+struct kr_lt_mmd {
+ int devad;
+ u32 control;
+ u32 status;
+ u32 lp_cu;
+ u32 lp_status;
+ u32 ld_cu;
+ u32 ld_status;
+};
+
+/* Backplane kr */
+struct backplane_kr {
+ struct kr_coef min_kr;
+ struct kr_coef max_kr;
+ struct kr_coef def_kr;
+ /* defaults for eq kr are initialized from DT: eq-init */
+ bool valid_eq_init;
+ /* defaults for eq params are initialized from DT: eq-params */
+ bool valid_eq_params;
+ /* EQ algorithm name */
+ const char *eqa_name;
+ const struct equalizer_device *equalizer;
+ struct kr_lt_mmd ltmmd;
+};
+
+/* Backplane device specific callbacks */
+struct backplane_ops {
+ /* AN register ops */
+ void (*an_advertisement_init)(struct lane_device *lane);
+ bool (*is_an_link_detected)(struct lane_device *lane);
+ /* default settings ops */
+ void (*setup_default_settings)(struct lane_device *lane);
+ /* LT coefficients validation ops */
+ int (*lt_validation)(struct lane_device *lane, u32 *ld_coef);
+};
+
+/* Backplane driver */
+struct backplane_driver {
+ /* serdes base address */
+ u32 base_addr;
+ /* serdes memory map size */
+ u32 memmap_size;
+ /* serdes endianness */
+ bool is_little_endian;
+ /* memory I/O */
+ struct mem_io io;
+ /* backplane ops */
+ struct backplane_ops bp_ops;
+ /* lane ops */
+ const struct lane_ops *lane_ops;
+ /* device specific private extension */
+ void *priv;
+};
+
+/* Backplane device */
+struct backplane_device {
+ struct phy_device *phydev;
+ u8 num_lanes;
+ bool aneg_config;
+ bool aneg_done;
+ bool phy_suspended;
+ /* backplane management functions */
+ struct backplane_driver drv;
+ /* backplane kr */
+ struct backplane_kr bpkr;
+ /* kr lanes array: valid elements = num_lanes */
+ struct lane_device lane[MAX_KR_LANES_PER_PHY];
+ /* device specific private extension */
+ void *priv;
+ /* bpphy mutexes */
+ struct mutex bpphy_lock;
+ /* mutex between multiple lanes training */
+ struct mutex trained_lock;
+};
+
+bool backplane_is_mode_kr(phy_interface_t interface);
+
+bool backplane_is_valid_mode(phy_interface_t interface);
+
+u8 backplane_num_lanes(phy_interface_t interface);
+
+u32 backplane_get_an_bp_eth_status_bit(phy_interface_t interface);
+
+u32 backplane_get_an_adv1_init(phy_interface_t interface);
+
+bool backplane_is_single_lane(struct backplane_device *bpdev);
+
+bool backplane_is_multi_lane(struct backplane_device *bpdev);
+
+int backplane_are_all_lanes_trained(struct backplane_device *bpdev);
+
+int backplane_get_lanes_trained_count(struct backplane_device *bpdev);
+
+int backplane_is_link_up(struct phy_device *phydev);
+
+void backplane_kr_lt_mmd_c45(struct backplane_kr *bpkr);
+
+void backplane_kr_lt_mmd_setup(struct backplane_kr *bpkr, int devad, u32 base);
+
+int backplane_read_mmd(struct lane_device *lane, int devad, u32 regnum);
+
+int backplane_write_mmd(struct lane_device *lane, int devad, u32 regnum,
+ u16 val);
+
+void backplane_default_kr_lane(struct lane_device *lane);
+
+void backplane_get_current_taps(struct lane_device *lane, u32 *coef);
+
+void backplane_set_current_taps(struct lane_device *lane, u32 *coef);
+
+void backplane_set_all_taps_to_max(struct lane_device *lane);
+
+void backplane_tune_kr_lane(struct lane_device *lane, bool reset_lane);
+
+/* generic main operations to be used on probe callback */
+
+int backplane_create(struct phy_device *phydev);
+
+int backplane_parse_dt(struct phy_device *phydev);
+
+int backplane_setup_memio(struct phy_device *phydev);
+
+int backplane_setup_mmd(struct phy_device *phydev);
+
+int backplane_setup_lanes(struct phy_device *phydev);
+
+int backplane_initialize(struct phy_device *phydev);
+
+/* predefined phy_driver callback functions */
+
+int backplane_probe(struct phy_device *phydev);
+
+void backplane_remove(struct phy_device *phydev);
+
+int backplane_config_init(struct phy_device *phydev);
+
+int backplane_aneg_done(struct phy_device *phydev);
+
+int backplane_config_aneg(struct phy_device *phydev);
+
+int backplane_suspend(struct phy_device *phydev);
+
+int backplane_resume(struct phy_device *phydev);
+
+int backplane_read_status(struct phy_device *phydev);
+
+int backplane_match_phy_device(struct phy_device *phydev);
+
+#endif /* __BACKPLANE_H */
diff --git a/drivers/net/phy/backplane/eq_fixed.c b/drivers/net/phy/backplane/eq_fixed.c
new file mode 100644
index 0000000..827450e
--- /dev/null
+++ b/drivers/net/phy/backplane/eq_fixed.c
@@ -0,0 +1,83 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+/* Fixed: No Equalization algorithm
+ *
+ * Copyright 2019-2020 NXP
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+#include "equalization.h"
+
+#define ALGORITHM_NAME "backplane_fixed"
+#define ALGORITHM_DESCR "Fixed Equalization"
+#define ALGORITHM_VERSION "1.0.0"
+
+/* Fixed Algorithm API */
+
+/* Create Fixed Equalization Algorithm */
+static struct eq_data_priv *create(struct equalizer_driver eqdrv)
+{
+ return NULL;
+}
+
+static const struct equalization_algorithm eq_alg = {
+ .name = ALGORITHM_NAME,
+ .descr = ALGORITHM_DESCR,
+ .version = ALGORITHM_VERSION,
+ .use_local_tx_training = false,
+ .use_remote_tx_training = false,
+ .ops = {
+ .create = create,
+ .destroy = NULL,
+ .is_rx_ok = NULL,
+ .is_eq_done = NULL,
+ .collect_statistics = NULL,
+ .generate_request = NULL,
+ .process_bad_state = NULL,
+ .dump_algorithm_context = NULL,
+ }
+};
+
+static const char * const alg_keys[] = {
+ DEFAULT_EQ_ALGORITHM,
+ "bypass",
+};
+
+static int __init fixed_init(void)
+{
+ int i, err;
+
+ pr_info("%s: %s algorithm version %s\n",
+ ALGORITHM_NAME, ALGORITHM_DESCR, ALGORITHM_VERSION);
+
+ /* register Fixed algorithm: */
+ for (i = 0; i < ARRAY_SIZE(alg_keys); i++) {
+ err = backplane_eq_register(alg_keys[i], &eq_alg);
+ if (err) {
+ pr_err("%s: '%s' equalization algorithm registration failed\n",
+ ALGORITHM_NAME, alg_keys[i]);
+ }
+ }
+
+ return 0;
+}
+
+static void __exit fixed_exit(void)
+{
+ int i;
+
+ /* unregister Fixed algorithm: */
+ for (i = 0; i < ARRAY_SIZE(alg_keys); i++)
+ backplane_eq_unregister(alg_keys[i]);
+
+ pr_info("%s: %s algorithm version %s unloaded\n",
+ ALGORITHM_NAME, ALGORITHM_DESCR, ALGORITHM_VERSION);
+}
+
+module_init(fixed_init);
+module_exit(fixed_exit);
+
+MODULE_DESCRIPTION("Fixed Equalization Algorithm");
+MODULE_AUTHOR("Florinel Iordache <florinel.iordache@nxp.com>");
+MODULE_LICENSE("Dual BSD/GPL");
diff --git a/drivers/net/phy/backplane/equalization.h b/drivers/net/phy/backplane/equalization.h
new file mode 100644
index 0000000..767f0159
--- /dev/null
+++ b/drivers/net/phy/backplane/equalization.h
@@ -0,0 +1,275 @@
+/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */
+/* Equalization interface
+ * for Equalization and Link Training (IEEE802.3ap/ba)
+ *
+ * Copyright 2019-2020 NXP
+ */
+
+#ifndef __EQUALIZATION_H
+#define __EQUALIZATION_H
+
+#include <linux/phy.h>
+
+/* Default equalization algorithm */
+#define DEFAULT_EQ_ALGORITHM "fixed"
+
+struct lane_device;
+struct equalizer_driver;
+
+/* EQ Algorithms Interface used by Link Training
+ * to call equalization algorithms callbacks
+ */
+
+/* Equalization private data
+ * specifically defined by each algorithm to be used internally
+ */
+struct eq_data_priv;
+
+/* Equalization Algorithm operations */
+struct equalization_ops {
+ /* Mandatory operations: */
+ struct eq_data_priv *(*create)(struct equalizer_driver eqdrv);
+ void (*destroy)(struct eq_data_priv *priv);
+ /* Required operations for remote Tx link training: */
+ bool (*is_rx_ok)(struct eq_data_priv *priv);
+ bool (*is_eq_done)(struct eq_data_priv *priv);
+ bool (*collect_statistics)(struct eq_data_priv *priv);
+ void (*generate_request)(struct eq_data_priv *priv);
+ /* Optional operations: */
+ void (*process_bad_state)(struct eq_data_priv *priv);
+ void (*dump_algorithm_context)(struct eq_data_priv *priv);
+};
+
+/* Equalization Algorithm description data */
+struct equalization_algorithm {
+ const char *name;
+ const char *descr;
+ const char *version;
+ bool use_local_tx_training;
+ bool use_remote_tx_training;
+ struct equalization_ops ops;
+};
+
+/* Equalizer Interface for EQ Algorithms:
+ * Used by equalization algorithms to collect equalizer statistics
+ * required to take correct decisions for tuning equalization parameters
+ */
+
+/* Equalizer counters type
+ *
+ * Equalizer Binning Counters for Data Dependent Edge Statistics:
+ *
+ * Bin(s) = (# late edges - # early edges)
+ * Prior/Next Edge at T -/+ #UI (Unit Interval)
+ * Bin_1: 1UI wide pulses: Prior Edge at T - 1UI
+ * final edges on short pulses:
+ * - contains the scoring of final edges on pulses that are 1UI long
+ * - represents the difference between the number of short pulse late edges
+ * and the number of short pulse early edges
+ * Bin_2: 2UI wide pulses: Prior Edge at T - 2UI
+ * Bin_3: 3UI (or >=3UI) wide pulses: Prior Edge at T - 3UI (or T - >=3UI)
+ * Bin_4: 4UI (or >=4UI) wide pulses: Prior Edge at T - 4UI (or T - >=4UI)
+ * Bin_Med: >=5UI and <=7UI wide pulses:
+ * Prior Edge in between T - >=5UI and T - <=7UI
+ * final edges on medium pulses:
+ * - contains the scoring of final edges on pulses between 5UI and 7UI long
+ * Bin_Long: >=8UI wide pulses: Prior Edge at T - >=8UI
+ * final edges on long pulses:
+ * - contains the scoring of final edges on pulses longer than 7UI long
+ * - represents the difference between the number of long pulse late edges
+ * and the number of long pulse early edges
+ * Bin_M1: 1UI wide pulses: Next Edge at T + 1UI
+ * initial edges on short pulses following non-single bits:
+ * - contains the scoring of initial edges on pulses that are 1UI long
+ * following non-single bits
+ * - the next edge is 1UI away and prior edge is more than 1UI away
+ * Bin_M2: 2UI wide pulses: Next Edge at T + 2UI
+ * Bin_M3: 3UI (or >=3UI) wide pulses: Next Edge at T + 3UI (or T + >=3UI)
+ * Bin_M4: 4UI (or >=4UI) wide pulses: Next Edge at T + 4UI (or T + >=4UI)
+ * Bin_MMed: >=5UI and <=7UI wide pulses:
+ * Next Edge in between T + >=5UI and T + <=7UI
+ * initial edges on medium pulses following non-single bits:
+ * - contains the scoring of initial edges on pulses between 5UI and 7UI
+ * following non-single bits
+ * Bin_MLong: >=8UI wide pulses: Next Edge at T + >=8UI
+ * initial edges on long pulses following non-single bits:
+ * - contains the scoring of initial edges on pulses longer than 7UI long
+ * - represents the difference between the number of long pulse late edges
+ * and the number of long pulse early edges
+ *
+ * Bin_Offset = [(# late rising edges + # early falling edges) -
+ * (# early rising edges + # late falling edges)]
+ * - contains the transition information for the difference between
+ * all bits that are narrower than expected and
+ * all bits that are wider than expected
+ *
+ * Bin_Avg: Low Pass Filter of Running Disparity
+ * - Bin_Avg provides a time weighted, filtered average of disparity which
+ * indicates the BLW potential of recently received data
+ * New Bin_Avg = Bin_Avg - Bin_Avg/8 + block_disparity
+ * where block_disparity = (#of ones - #of zeros)
+ *
+ * Bin_BLW: Bin Baseline Wander
+ * - BinBLW accumulates the correlation between Bin_Avg and Bin_Offset
+ * - Low frequency deficiency (LFD) causes BLW effect
+ * New Bin_BLW = Bin_BLW + Bin_Avg, for Bin_Offset > 0
+ * = Bin_BLW - Bin_Avg, for Bin_Offset < 0
+ * = Bin_BLW, for Bin_Offset = 0
+ *
+ * Equalizer gains:
+ * GAIN_LF: Low-frequency gain of the equalizer amplifier
+ * GAIN_MF: Middle-frequency gain of the equalizer amplifier
+ * GAIN_HF: High-frequency gain of the equalizer amplifier
+ *
+ * Equalizer status:
+ * EQOFFSET: equalization offset status
+ * Binary coded status of RX Adaptive Equalization offset controls of lane
+ */
+enum eqc_type {
+ EQC_BIN_1,
+ EQC_BIN_2,
+ EQC_BIN_3,
+ EQC_BIN_4,
+ EQC_BIN_MED,
+ EQC_BIN_LONG,
+ EQC_BIN_M1,
+ EQC_BIN_M2,
+ EQC_BIN_M3,
+ EQC_BIN_M4,
+ EQC_BIN_MMED,
+ EQC_BIN_MLONG,
+ EQC_BIN_OFFSET,
+ EQC_BIN_AVG,
+ EQC_BIN_BLW,
+ EQC_GAIN_LF,
+ EQC_GAIN_MF,
+ EQC_GAIN_HF,
+ EQC_EQOFFSET,
+};
+
+/* Equalizer counters range */
+struct eqc_range {
+ s16 min;
+ s16 max;
+ s16 mid_low;
+ s16 mid_high;
+};
+
+/* Equalizer counters collection operations */
+struct equalizer_ops {
+ int (*collect_counters)(void *reg, enum eqc_type type, s16 *counters,
+ u8 size);
+ int (*collect_multiple_counters)(void *reg, enum eqc_type type[],
+ u8 type_no, s16 *counters, u8 size);
+ struct eqc_range *(*get_counter_range)(enum eqc_type type);
+};
+
+/* Equalizer device and operations */
+struct equalizer_device {
+ const char *name;
+ const char *version;
+ struct equalizer_ops ops;
+};
+
+/* Equalization driver */
+struct equalizer_driver {
+ struct phy_device *phydev;
+ /* lane info used as parameter for link training API */
+ struct lane_device *lane;
+ /* lane reg base used as parameter for equalizer ops */
+ void *reg_base;
+ const struct equalizer_device *equalizer;
+};
+
+/* Link Training Interface used by EQ Algorithms
+ * to interact with IEEE802.3ap/ba standards
+ */
+
+/* update request type
+ * Identifies the LP update request type according to IEEE802.3ap-2007
+ * which must be sent to LP to request coefficients update
+ *
+ * HOLD: Request LP to Hold all coefficients update
+ * INC: Request LP to Increment the specified coefficient
+ * DEC: Request LP to Decrement the specified coefficient
+ * INIT: Request LP to Initialize all coefficients
+ * PRESET: Request LP to set all coefficients to Preset
+ * INVALID: Invalid request type: should not be used as LP request
+ */
+enum req_type {
+ REQ_HOLD,
+ REQ_INC,
+ REQ_DEC,
+ REQ_INIT,
+ REQ_PRESET,
+ REQ_INVALID
+};
+
+/* coefficient field
+ * Identifies the coefficient field on which must take a desired action
+ * according to IEEE802.3ap-2007
+ *
+ * coefficients:
+ * M1: C(-1): Pre-cursor
+ * Z0: C(0): Main cursor
+ * P1: C(+1): Post-cursor
+ * NO: Number of coefficients (this is not a valid coefficient field)
+ */
+enum coef_field {
+ C_M1,
+ C_Z0,
+ C_P1,
+ C_NO
+};
+
+/* coefficient status
+ * Specifies the coefficient status according to IEEE802.3ap-2007:
+ * 72.6.10.2.5 Coefficient update process
+ *
+ * NOTUPDATED: Coefficient is not updated
+ * UPDATED: Coefficient is updated
+ * MIN: Coefficient has reached the minimum threshold
+ * MAX: Coefficient has reached the maximum threshold
+ * INVALID: Invalid coefficient status
+ */
+enum coef_status {
+ COEF_NOTUPDATED,
+ COEF_UPDATED,
+ COEF_MIN,
+ COEF_MAX,
+ COEF_INVALID
+};
+
+void lt_lp_update(struct lane_device *lane, u32 update);
+
+u32 lt_encode_request(u32 base_update, enum req_type req,
+ enum coef_field field);
+
+u32 lt_encode_startup_request(enum req_type req);
+
+enum req_type lt_decode_coef_update(u32 update, enum coef_field field);
+
+bool lt_is_update_of_type(u32 update, enum req_type type);
+
+bool lt_is_lp_at_startup(struct lane_device *lane, enum req_type type);
+
+enum coef_status lt_get_lp_coef_status(struct lane_device *lane,
+ enum coef_field field);
+
+void lt_move_lp_back(struct lane_device *lane);
+
+void lt_set_error(struct lane_device *lane, bool err);
+
+/* Backplane Driver Interface for EQ Algorithms:
+ * Used by equalization algorithms to interact
+ * with backplane driver during equalization
+ */
+
+/* equalization algorithm registration */
+int backplane_eq_register(const char *key,
+ const struct equalization_algorithm *eq_info);
+void backplane_eq_unregister(const char *key);
+
+bool backplane_is_cdr_lock(struct lane_device *lane, bool retry);
+
+#endif /* __EQUALIZATION_H */
diff --git a/drivers/net/phy/backplane/link_training.c b/drivers/net/phy/backplane/link_training.c
new file mode 100644
index 0000000..3aa26f9
--- /dev/null
+++ b/drivers/net/phy/backplane/link_training.c
@@ -0,0 +1,1529 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+/* Link Training (IEEE802.3ap/ba)
+ * Ethernet Operation over Electrical Backplanes
+ *
+ * Copyright 2019-2020 NXP
+ */
+
+#include <linux/mdio.h>
+#include <linux/timer.h>
+#include <linux/delay.h>
+
+#include "link_training.h"
+
+/* KR LP/LD Coefficients */
+#define PRESET_MASK 0x2000
+#define INIT_MASK 0x1000
+#define COP1_MASK 0x30
+#define COP1_SHIFT 4
+#define COZ0_MASK 0xc
+#define COZ0_SHIFT 2
+#define COM1_MASK 0x3
+#define COM1_SHIFT 0
+#define ALL_COEF_MASK (COP1_MASK | COZ0_MASK | COM1_MASK)
+#define LD_ALL_MASK (PRESET_MASK | INIT_MASK | ALL_COEF_MASK)
+
+/* KR LP Status Report */
+#define LP_STATUS_ALL_COEF_UPDATED 0x15
+
+/* KR LP/LD Status Report:
+ * RX_READY_MASK - Receiver Ready
+ * 0b - The LP/LD receiver is requesting that training continue
+ * 1b - The LP/LD receiver has determined that training is complete
+ * and is prepared to receive data.
+ */
+#define RX_READY_MASK 0x8000
+
+/* Increment/Decrement Requests */
+#define HOLD 0
+#define INCREMENT 1
+#define DECREMENT 2
+#define RESERVED 3
+
+/* Increment/Decrement Steps */
+#define STEP_INCREMENT_P1 -1
+#define STEP_INCREMENT_Z0 1
+#define STEP_INCREMENT_M1 -1
+
+/* KR PMD Control defines */
+#define TRAIN_EN 0x3
+#define TRAIN_DISABLE 0x1
+#define PMD_RESET 0x1
+
+/* KR PMD Status defines */
+#define PMD_STATUS_TRAIN_FAIL 0x8
+#define PMD_STATUS_SUP_STAT 0x4
+#define PMD_STATUS_FRAME_LOCK 0x2
+#define PMD_STATUS_RX_STAT 0x1
+
+/* KR PMD control register (Register 1.150) */
+#define KR_PMD_BASE_OFFSET 150
+
+/* Link training KR PMD registers offsets (relative to base) */
+#define OFFSET_KR_PMD_CTRL 0x0
+#define OFFSET_KR_PMD_STATUS 0x1
+#define OFFSET_KR_LP_CU 0x2
+#define OFFSET_KR_LP_STATUS 0x3
+#define OFFSET_KR_LD_CU 0x4
+#define OFFSET_KR_LD_STATUS 0x5
+
+/* Timeouts */
+#define TIMEOUT_MOVE_BACK_PREV 6
+#define TIMEOUT_REPEAT_REQUEST 10
+
+/* Training for Remote Tx */
+
+static u32 get_mask_for_req(enum req_type req)
+{
+ u32 cmd = HOLD;
+
+ switch (req) {
+ case REQ_HOLD:
+ cmd = HOLD;
+ break;
+ case REQ_INC:
+ cmd = INCREMENT;
+ break;
+ case REQ_DEC:
+ cmd = DECREMENT;
+ break;
+ case REQ_INIT:
+ cmd = INIT_MASK;
+ break;
+ case REQ_PRESET:
+ cmd = PRESET_MASK;
+ break;
+ case REQ_INVALID:
+ cmd = RESERVED;
+ break;
+ default:
+ cmd = HOLD;
+ break;
+ }
+ return cmd;
+}
+
+static enum req_type get_req_for_mask(u32 cmd)
+{
+ enum req_type req = REQ_HOLD;
+
+ switch (cmd) {
+ case HOLD:
+ req = REQ_HOLD;
+ break;
+ case INCREMENT:
+ req = REQ_INC;
+ break;
+ case DECREMENT:
+ req = REQ_DEC;
+ break;
+ case INIT_MASK:
+ req = REQ_INIT;
+ break;
+ case PRESET_MASK:
+ req = REQ_PRESET;
+ break;
+ case RESERVED:
+ req = REQ_INVALID;
+ break;
+ default:
+ req = REQ_HOLD;
+ break;
+ }
+ return req;
+}
+
+/* ld_coef_status
+ * 72.6.10.2.5 Coefficient update process
+ * Once the updated, maximum, or minimum state is reported it continues
+ * to be reported until a hold request is received,
+ * after which the status reverts to not_updated.
+ */
+static void ld_coef_status(struct lane_device *lane)
+{
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.ld_status,
+ lane->krln.ld_status);
+}
+
+/* ld_coef_update
+ * LD sends to LP the specified request for coefficients update
+ */
+static void ld_coef_update(struct lane_device *lane)
+{
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.ld_cu,
+ lane->krln.ld_update);
+}
+
+/* get_lp_lcs
+ * get LP lcs (last change status)
+ * returns the last LP change (non-zero) status:
+ * meaning the last LP status resulted from a change request
+ * 72.6.10.2.5 Coefficient update process
+ * Once the updated, maximum, or minimum state is reported it continues
+ * to be reported until a hold request is received,
+ * after which the status reverts to not_updated.
+ */
+static u32 get_lp_lcs(struct lane_device *lane)
+{
+ return lane->krln.lp_last_change_status;
+}
+
+static bool is_all_status(u32 status, enum coef_status cs)
+{
+ return ((status & ALL_COEF_MASK) ==
+ (cs << COP1_SHIFT | cs << COZ0_SHIFT | cs << COM1_SHIFT));
+}
+
+/* Training for Local Tx */
+
+static void initialize(struct lane_device *lane)
+{
+ backplane_default_kr_lane(lane);
+
+ lane->krln.ld_status &= ~ALL_COEF_MASK;
+ lane->krln.ld_status |= COEF_UPDATED << COP1_SHIFT |
+ COEF_UPDATED << COZ0_SHIFT |
+ COEF_UPDATED << COM1_SHIFT;
+
+ ld_coef_status(lane);
+}
+
+/* preset
+ * Preset as defined by: IEEE 802.3, sub-clause 72.6.10.2.3.1
+ * Setup all coefficients to MAX values from IEEE802.3 perspective
+ */
+static void preset(struct lane_device *lane)
+{
+ backplane_set_all_taps_to_max(lane);
+
+ backplane_tune_kr_lane(lane, true);
+
+ lane->krln.ld_status &= ~ALL_COEF_MASK;
+ lane->krln.ld_status |= COEF_MAX << COP1_SHIFT |
+ COEF_MAX << COZ0_SHIFT |
+ COEF_MAX << COM1_SHIFT;
+
+ ld_coef_status(lane);
+}
+
+static bool is_rx_ready(u32 status)
+{
+ return ((status & RX_READY_MASK) != 0);
+}
+
+/* is_ld_valid
+ * LD coefficient values have hardware restrictions
+ * Check if all ld coefficients are in range
+ */
+static int is_ld_valid(struct lane_device *lane, u32 *ld_coef)
+{
+ struct backplane_kr *bpkr = &lane->bpdev->bpkr;
+ u32 mainq = ld_coef[C_Z0];
+ u32 postq = ld_coef[C_P1];
+ u32 preq = ld_coef[C_M1];
+
+ /* Basic HW restrictions: */
+
+ /* 1. tx_preq <= MIN_C(-1) */
+ if (preq > bpkr->min_kr.preq)
+ return -ERANGE;
+ /* 2. tx_ratio_post1q <= MIN_C(+1) */
+ if (postq > bpkr->min_kr.postq)
+ return -ERANGE;
+ /* 3. MIN_C(0) <= tx_mainq <= MAX_C(0) */
+ if (mainq < bpkr->min_kr.mainq)
+ return -ERANGE;
+ if (mainq > bpkr->max_kr.mainq)
+ return -ERANGE;
+ /* 4. tx_ratio_post1q >= tx_preq */
+ if (postq < preq)
+ return -ERANGE;
+
+ /* Additional HW restrictions:
+ * LT custom HW validation: Device specific HW restrictions
+ */
+ if (lane->bpdev->drv.bp_ops.lt_validation)
+ return lane->bpdev->drv.bp_ops.lt_validation(lane, ld_coef);
+
+ return 0;
+}
+
+static bool update_ld_status(struct lane_device *lane, enum coef_field field,
+ enum coef_status cs)
+{
+ u32 ld_cs = cs;
+ u32 mask, val;
+
+ if (cs == COEF_INVALID)
+ return false;
+
+ switch (field) {
+ case C_P1:
+ mask = COP1_MASK;
+ val = ld_cs << COP1_SHIFT;
+ break;
+ case C_Z0:
+ mask = COZ0_MASK;
+ val = ld_cs << COZ0_SHIFT;
+ break;
+ case C_M1:
+ mask = COM1_MASK;
+ val = ld_cs << COM1_SHIFT;
+ break;
+ default:
+ return false;
+ }
+
+ lane->krln.ld_status &= ~mask;
+ lane->krln.ld_status |= val;
+
+ return true;
+}
+
+static enum coef_status inc_dec(struct lane_device *lane,
+ enum coef_field field, int request)
+{
+ u32 ld_coef[C_NO], step[C_NO], ld_limit[C_NO];
+ int err;
+
+ backplane_get_current_taps(lane, ld_coef);
+
+ step[C_M1] = STEP_INCREMENT_M1;
+ step[C_Z0] = STEP_INCREMENT_Z0;
+ step[C_P1] = STEP_INCREMENT_P1;
+
+ /* 72.6.10.2.5 Coefficient update process
+ * Upon execution of a received increment or decrement request,
+ * the status is reported as updated, maximum, or minimum.
+ */
+ switch (request) {
+ case INCREMENT:
+ ld_limit[C_M1] = lane->bpdev->bpkr.max_kr.preq;
+ ld_limit[C_Z0] = lane->bpdev->bpkr.max_kr.mainq;
+ ld_limit[C_P1] = lane->bpdev->bpkr.max_kr.postq;
+ if (ld_coef[field] != ld_limit[field])
+ ld_coef[field] += step[field];
+ else
+ return COEF_MAX;
+ break;
+ case DECREMENT:
+ ld_limit[C_M1] = lane->bpdev->bpkr.min_kr.preq;
+ ld_limit[C_Z0] = lane->bpdev->bpkr.min_kr.mainq;
+ ld_limit[C_P1] = lane->bpdev->bpkr.min_kr.postq;
+ if (ld_coef[field] != ld_limit[field])
+ ld_coef[field] -= step[field];
+ else
+ return COEF_MIN;
+ break;
+ default:
+ break;
+ }
+
+ err = is_ld_valid(lane, ld_coef);
+ if (!err) {
+ /* accept new ld coefficients */
+ backplane_set_current_taps(lane, ld_coef);
+ backplane_tune_kr_lane(lane, false);
+ } else {
+ if (request == DECREMENT)
+ return COEF_MIN;
+ if (request == INCREMENT)
+ return COEF_MAX;
+ }
+
+ /* UPDATED */
+ return COEF_UPDATED;
+}
+
+static void check_request(struct lane_device *lane, int request)
+{
+ enum coef_status cu = COEF_INVALID;
+ int cop1_req, coz0_req, com1_req;
+ int old_status;
+
+ cop1_req = (request & COP1_MASK) >> COP1_SHIFT;
+ coz0_req = (request & COZ0_MASK) >> COZ0_SHIFT;
+ com1_req = (request & COM1_MASK) >> COM1_SHIFT;
+
+ /* IEEE802.3-2008, 72.6.10.2.5
+ * Ensure we only act on INCREMENT/DECREMENT when we are in NOT UPDATED
+ *
+ * 72.6.10.2.5 Coefficient update process
+ * An increment or decrement request will only be acted upon when
+ * the state of the tap is not_updated.
+ */
+ old_status = lane->krln.ld_status;
+
+ if (cop1_req && !(lane->krln.ld_status & COP1_MASK)) {
+ cu = inc_dec(lane, C_P1, cop1_req);
+ update_ld_status(lane, C_P1, cu);
+ }
+
+ if (coz0_req && !(lane->krln.ld_status & COZ0_MASK)) {
+ cu = inc_dec(lane, C_Z0, coz0_req);
+ update_ld_status(lane, C_Z0, cu);
+ }
+
+ if (com1_req && !(lane->krln.ld_status & COM1_MASK)) {
+ cu = inc_dec(lane, C_M1, com1_req);
+ update_ld_status(lane, C_M1, cu);
+ }
+
+ if (old_status != lane->krln.ld_status)
+ ld_coef_status(lane);
+}
+
+static void training_complete(struct lane_device *lane)
+{
+ struct training_status *trst = &lane->krln.trst;
+
+ /* update training status */
+ trst->remote_tx_complete = true;
+ trst->remote_tx_running = false;
+
+ /* report LD status */
+ lane->krln.ld_status |= RX_READY_MASK;
+ ld_coef_status(lane);
+
+ /* update PMD status and tell LP we are ready */
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.status,
+ PMD_STATUS_RX_STAT);
+}
+
+/* Link Training general API */
+
+/* Setup standard KR LT MMD (MDIO Managed Device) for Clause 45
+ * 45.2.1.76 10GBASE-KR PMD control register (Register 1.150)
+ */
+void lt_mmd_c45(struct backplane_kr *bpkr)
+{
+ lt_mmd_setup(bpkr, MDIO_MMD_PMAPMD, KR_PMD_BASE_OFFSET);
+}
+
+/* Setup KR LT MMD (MDIO Managed Device)
+ * IEEE Std 802.3ap-2007: Table 45.3 PMA/PMD registers
+ */
+void lt_mmd_setup(struct backplane_kr *bpkr, int devad, u32 base)
+{
+ bpkr->ltmmd.devad = devad;
+ bpkr->ltmmd.control = base + OFFSET_KR_PMD_CTRL;
+ bpkr->ltmmd.status = base + OFFSET_KR_PMD_STATUS;
+ bpkr->ltmmd.lp_cu = base + OFFSET_KR_LP_CU;
+ bpkr->ltmmd.lp_status = base + OFFSET_KR_LP_STATUS;
+ bpkr->ltmmd.ld_cu = base + OFFSET_KR_LD_CU;
+ bpkr->ltmmd.ld_status = base + OFFSET_KR_LD_STATUS;
+}
+
+/* lt_is_lp_rx_ready
+ * Reports if LP Receiver is ready
+ * false: The LP receiver is requesting that training continue
+ * true: The LP receiver has determined that training is complete
+ * and is prepared to receive data.
+ */
+bool lt_is_lp_rx_ready(struct lane_device *lane)
+{
+ struct kr_lt_mmd *ltmmd = &lane->bpdev->bpkr.ltmmd;
+
+ /* Read LP Status */
+ lane->krln.lp_status = backplane_read_mmd(lane,
+ ltmmd->devad,
+ ltmmd->lp_status);
+ return is_rx_ready(lane->krln.lp_status);
+}
+
+/* lt_is_ld_rx_ready
+ * Reports if LD Receiver is ready
+ * false: The LD receiver is requesting that training continue
+ * true: The LD receiver has determined that training is complete
+ * and is prepared to receive data.
+ */
+bool lt_is_ld_rx_ready(struct lane_device *lane)
+{
+ return is_rx_ready(lane->krln.ld_status);
+}
+
+void lt_start(struct lane_device *lane)
+{
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.control,
+ TRAIN_EN);
+}
+
+void lt_stop(struct lane_device *lane)
+{
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.control,
+ TRAIN_DISABLE);
+}
+
+void lt_reset(struct lane_device *lane)
+{
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad, MDIO_CTRL1,
+ PMD_RESET);
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.control,
+ TRAIN_DISABLE);
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.ld_cu, 0);
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.ld_status, 0);
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.status, 0);
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.lp_cu, 0);
+ backplane_write_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.lp_status, 0);
+}
+
+/* lt_is_rx_trained
+ * IEEE Std 802.3ap-2007: Table 72.3 MDIO/PMD status variable mapping
+ * PMD status variable: rx_trained
+ */
+bool lt_is_rx_trained(struct lane_device *lane)
+{
+ struct phy_device *phydev = lane->phydev;
+ int timeout = 100;
+ int val;
+
+ val = backplane_read_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.status);
+
+ if ((val & PMD_STATUS_RX_STAT) && !(val & PMD_STATUS_TRAIN_FAIL)) {
+ while (timeout--) {
+ if (backplane_is_link_up(phydev))
+ return true;
+
+ usleep_range(100, 500);
+ }
+ }
+ return false;
+}
+
+/* lt_is_training_failure
+ * IEEE Std 802.3ap-2007: Table 72.3 MDIO/PMD status variable mapping
+ * PMD status variable: PMD_fault
+ */
+bool lt_is_training_failure(struct lane_device *lane)
+{
+ struct kr_lt_mmd *ltmmd = &lane->bpdev->bpkr.ltmmd;
+ int lt_state;
+
+ lt_state = backplane_read_mmd(lane, ltmmd->devad, ltmmd->status);
+
+ /* according to spec: 8023ap-2007.pdf
+ * training_failure
+ * Boolean variable that is set to TRUE when the training state machine
+ * has timed out due to expiration of the max_wait_timer while in the
+ * SEND_TRAINING, TRAIN_LOCAL, or
+ * TRAIN_REMOTE states and is set to FALSE otherwise.
+ */
+ if (lt_state & PMD_STATUS_TRAIN_FAIL)
+ return true;
+
+ return false;
+}
+
+/* lt_is_frame_lock
+ * IEEE Std 802.3ap-2007: Table 72.3 MDIO/PMD status variable mapping
+ * PMD status variable: frame_lock
+ */
+bool lt_is_frame_lock(struct lane_device *lane)
+{
+ struct kr_lt_mmd *ltmmd = &lane->bpdev->bpkr.ltmmd;
+ int lt_state;
+
+ lt_state = backplane_read_mmd(lane, ltmmd->devad, ltmmd->status);
+
+ if ((lt_state & PMD_STATUS_SUP_STAT) &&
+ (lt_state & PMD_STATUS_FRAME_LOCK))
+ return true;
+
+ return false;
+}
+
+/* Training for Remote Tx
+ * This is the main routine for Remote Tx training
+ */
+void lt_train_remote_tx(struct lane_device *lane)
+{
+ const struct equalization_algorithm *eq_alg = lane->krln.eq_alg;
+ struct training_status *trst = &lane->krln.trst;
+ u32 prev_req_cp1, prev_req_cz0, prev_req_cm1;
+ u32 status_cp1, status_cz0, status_cm1;
+ u32 prev_req_init, prev_req_preset;
+ u64 lp_resp_time;
+
+ /* Check stop condition for Remote Tx training */
+ if (trst->remote_tx_complete)
+ return;
+
+ /* Check if equalization algorithm is installed */
+ if (!eq_alg)
+ return;
+
+ /* Check that all required callback operations are installed */
+ if (!eq_alg->ops.collect_statistics ||
+ !eq_alg->ops.is_rx_ok ||
+ !eq_alg->ops.generate_request ||
+ !eq_alg->ops.is_eq_done)
+ return;
+
+ /* Start new Remote Tx training step */
+ trst->remote_tx_running = true;
+
+ /* Store current state as previous state */
+ lane->krln.prev_ld_update = lane->krln.ld_update;
+ if ((lane->krln.prev_ld_update & ALL_COEF_MASK) != HOLD)
+ lane->krln.ld_last_nonhold_update = lane->krln.prev_ld_update;
+
+ prev_req_init = lane->krln.prev_ld_update & INIT_MASK;
+ prev_req_preset = lane->krln.prev_ld_update & PRESET_MASK;
+ prev_req_cp1 = (lane->krln.prev_ld_update & COP1_MASK) >> COP1_SHIFT;
+ prev_req_cz0 = (lane->krln.prev_ld_update & COZ0_MASK) >> COZ0_SHIFT;
+ prev_req_cm1 = (lane->krln.prev_ld_update & COM1_MASK) >> COM1_SHIFT;
+
+ /* Training Done condition */
+ if (eq_alg->ops.is_eq_done(lane->krln.eq_priv))
+ trst->done_training = true;
+
+ /* Check if Training is Done */
+ if (trst->done_training) {
+ training_complete(lane);
+ return;
+ }
+
+ /* Read LP Status */
+ lane->krln.lp_status =
+ backplane_read_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.lp_status);
+
+ if ((lane->krln.lp_status & ALL_COEF_MASK) != 0)
+ lane->krln.lp_last_change_status = lane->krln.lp_status;
+
+ status_cp1 = (lane->krln.lp_status & COP1_MASK) >> COP1_SHIFT;
+ status_cz0 = (lane->krln.lp_status & COZ0_MASK) >> COZ0_SHIFT;
+ status_cm1 = (lane->krln.lp_status & COM1_MASK) >> COM1_SHIFT;
+
+ if (status_cp1 == COEF_UPDATED || status_cp1 == COEF_MIN ||
+ status_cp1 == COEF_MAX)
+ lane->krln.last_lp_update_status[C_P1] = status_cp1;
+ if (status_cz0 == COEF_UPDATED || status_cz0 == COEF_MIN ||
+ status_cz0 == COEF_MAX)
+ lane->krln.last_lp_update_status[C_Z0] = status_cz0;
+ if (status_cm1 == COEF_UPDATED || status_cm1 == COEF_MIN ||
+ status_cm1 == COEF_MAX)
+ lane->krln.last_lp_update_status[C_M1] = status_cm1;
+
+ /* IEEE802.3-2008, 72.6.10.2.3.2
+ * we send initialize to the other side to ensure default settings
+ * for the LP. Naturally, we should do this only once.
+ */
+ if (!trst->sent_init) {
+ /* All status MUST be NOTUPDATED for INIT to be executed
+ * otherwise send HOLD first
+ */
+ if (status_cp1 == COEF_NOTUPDATED &&
+ status_cz0 == COEF_NOTUPDATED &&
+ status_cm1 == COEF_NOTUPDATED) {
+ trst->sent_init = true;
+ lane->krln.ld_update = INIT_MASK;
+ lane->krln.req_ld_update_init_count = 1;
+ lane->krln.init_handshake_time =
+ jiffies_to_msecs(jiffies);
+ } else {
+ /* send HOLD before sending subsequent Init requests
+ * this is not the very first Init sent
+ */
+ lane->krln.ld_update = HOLD;
+ }
+ ld_coef_update(lane);
+ return;
+ }
+ /* continue to sent init request until LP responds to init */
+ if (prev_req_init) {
+ if (lane->krln.lp_status == 0) {
+ /* nothing to do here for now
+ * perhaps the partner board LP has not yet started
+ * so continue to send INIT requests
+ * this will happen in the next condition anyway
+ */
+ }
+ /* 72.6.10.2.3.2 Initialize
+ * The initialize control shall only be initially sent when all
+ * coefficient status fields indicate not_updated,
+ * and will then continue to be sent
+ * until no coefficient status field indicates not_updated.
+ */
+ if (status_cp1 == COEF_NOTUPDATED ||
+ status_cz0 == COEF_NOTUPDATED ||
+ status_cm1 == COEF_NOTUPDATED) {
+ lane->krln.ld_update = INIT_MASK;
+ ld_coef_update(lane);
+ lane->krln.req_ld_update_init_count++;
+ } else {
+ /* IEEE802.3-2008, 72.6.10.2.3.2
+ * We may clear INITIALIZE when no coefficients
+ * show NOT UPDATED.
+ */
+ /* v1: lane->krln.ld_update &= ~INIT_MASK; */
+ /* better send request: HOLD ALL
+ * should be equivalent since only INIT is set now
+ */
+ lane->krln.ld_update = HOLD;
+
+ lp_resp_time = jiffies_to_msecs(jiffies) -
+ lane->krln.init_handshake_time;
+ if (!lane->krln.first_recv_init) {
+ /* Init handshake not done yet,
+ * but will be soon
+ */
+ lane->krln.req_ld_update_init_count = 1;
+ lp_resp_time = 0;
+ }
+ ld_coef_update(lane);
+ }
+ return;
+ }
+
+ /* 72.6.10.2.3.1 Preset
+ * The preset control shall only be initially sent when all coefficient
+ * status fields indicate not_updated,
+ * and will then continue to be sent until the status for all
+ * coefficients indicates updated or maximum
+ */
+ /* IEEE802.3-2008, 72.6.10.2.3.1
+ * We may clear PRESET when all coefficients show UPDATED or MAX.
+ */
+ /* check if previous request was preset */
+ if (prev_req_preset) {
+ if ((status_cp1 == COEF_UPDATED || status_cp1 == COEF_MAX) &&
+ (status_cz0 == COEF_UPDATED || status_cz0 == COEF_MAX) &&
+ (status_cm1 == COEF_UPDATED || status_cm1 == COEF_MAX)) {
+ lane->krln.ld_update &= ~PRESET_MASK;
+ } else {
+ /* All status MUST be NOTUPDATED for INIT to be executed
+ * otherwise send HOLD first
+ */
+ if (status_cp1 == COEF_NOTUPDATED &&
+ status_cz0 == COEF_NOTUPDATED &&
+ status_cm1 == COEF_NOTUPDATED) {
+ lane->krln.ld_update = PRESET_MASK;
+ } else {
+ /* send HOLD before sending subsequent
+ * Preset requests
+ */
+ lane->krln.ld_update = HOLD;
+ }
+ ld_coef_update(lane);
+ return;
+ }
+ }
+
+ /* IEEE802.3-2008, 72.6.10.2.3.3
+ * We only request coefficient updates when no PRESET/INITIALIZE is
+ * pending. We also only request coefficient updates when the
+ * corresponding status is NOT UPDATED and nothing is pending.
+ */
+ if (lane->krln.ld_update & (PRESET_MASK | INIT_MASK))
+ return;
+
+ /* continue to move back to previous request until LP responds to it
+ * Move back to previous C(-1), C(0), C(+1) and HOLD
+ */
+ if (lane->krln.move_back_prev) {
+ /* can exit from here only with: DONE Training */
+ if (lane->krln.move_back_cnt == TIMEOUT_MOVE_BACK_PREV) {
+ trst->done_training = true;
+ training_complete(lane);
+ return;
+ }
+ lane->krln.move_back_cnt++;
+
+ if (status_cp1 == COEF_UPDATED)
+ lane->krln.move_back_lp_status |=
+ (COEF_UPDATED << COP1_SHIFT);
+ if (status_cz0 == COEF_UPDATED)
+ lane->krln.move_back_lp_status |=
+ (COEF_UPDATED << COZ0_SHIFT);
+ if (status_cm1 == COEF_UPDATED)
+ lane->krln.move_back_lp_status |=
+ (COEF_UPDATED << COM1_SHIFT);
+
+ if ((lane->krln.move_back_lp_status & ALL_COEF_MASK) ==
+ LP_STATUS_ALL_COEF_UPDATED) {
+ trst->done_training = true;
+ training_complete(lane);
+ return;
+ }
+
+ /* Move back to previous C(-1), C(0), C(+1) */
+ lane->krln.ld_update = lane->krln.prev_ld_update;
+ ld_coef_update(lane);
+ return;
+ }
+
+ /* 72.6.10.2.5 Coefficient update process
+ * Once the updated, maximum, or minimum state is reported it continues
+ * to be reported until a hold request is received,
+ * after which the status reverts to not_updated.
+ */
+
+ /* IEEE802.3-2008, 72.6.10.2.3.3
+ * We set coefficient requests to HOLD when we get the information
+ * about any updates On clearing our prior response, we also update
+ * our internal status.
+ */
+
+ /* send a Hold if want to send another INC same as previous
+ * and received status: NOTUPDATED
+ * 1. Continue to send previous REQ until receive status UPDATED
+ * 2. Continue to send HOLD until receive status NOTUPDATED
+ */
+
+ /* 3. LP can remain stuck ~42 ms in reset Rx lane: so we should wait
+ * around ~50 ms and only after that issue Timeout error message
+ */
+
+ switch (prev_req_cp1) {
+ case HOLD:
+ /* previous request was: HOLD */
+ if (status_cp1 == COEF_NOTUPDATED) {
+ /* All good here:
+ * continue to check the other coefficient requests
+ * and if all are good then proceed to
+ * generate coefficient tuning requests
+ */
+ } else {
+ /* Continue to send the same request: (2.)
+ * Continue to send HOLD until receive status NOTUPDATED
+ */
+ if (lane->krln.repeat_request_count >=
+ TIMEOUT_REPEAT_REQUEST) {
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(+1) HOLD request without LP response timeout !\n");
+ /* Hold Request Timeout:
+ * continue to send HOLD until LP responds
+ * with NOTUPDATED
+ */
+ lane->krln.repeat_request_count = 0;
+ } else {
+ /* Allow LP some time to respond
+ * and repeat request
+ */
+ msleep(20);
+ /* Allow LP more time to respond,
+ * as the last chance, on the last time
+ * before issuing timeout error: (3.)
+ */
+ if (lane->krln.repeat_request_count ==
+ TIMEOUT_REPEAT_REQUEST - 1)
+ msleep(30);
+ lane->krln.repeat_request_count++;
+ }
+ ld_coef_update(lane);
+ return;
+ }
+ break;
+ case INCREMENT:
+ case DECREMENT:
+ /* previous request was: INC/DEC */
+ if (status_cp1 == COEF_NOTUPDATED) {
+ /* Continue to send the same request: (1.)
+ * Continue to send previous REQ
+ * until receive status UPDATED
+ */
+ if (lane->krln.repeat_request_count >=
+ TIMEOUT_REPEAT_REQUEST) {
+ if (prev_req_cp1 == INCREMENT)
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(+1) INC request without LP response timeout !\n");
+ else
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(+1) DEC request without LP response timeout !\n");
+ /* Request Timeout:
+ * just continue: proceed again to
+ * generate coefficient tuning requests
+ */
+ } else {
+ /* Allow LP some time to respond
+ * and repeat request
+ */
+ msleep(20);
+ /* Allow LP more time to respond,
+ * as the last chance,
+ * on the last time before
+ * issuing timeout error: (3.)
+ */
+ if (lane->krln.repeat_request_count ==
+ TIMEOUT_REPEAT_REQUEST - 1)
+ msleep(30);
+ lane->krln.repeat_request_count++;
+ ld_coef_update(lane);
+ return;
+ }
+ } else {
+ /* All good here:
+ * LP responded to this Request
+ * Sent HOLD for this coefficient
+ * before asking another request
+ * continue to check the other coefficient requests
+ */
+ lane->krln.ld_update &= ~COP1_MASK;
+ }
+ break;
+ default:
+ /* previous request was: RESERVED: do nothing */
+ break;
+ }
+
+ switch (prev_req_cz0) {
+ case HOLD:
+ /* previous request was: HOLD */
+ if (status_cz0 == COEF_NOTUPDATED) {
+ /* All good here:
+ * continue to check the other coefficient requests
+ * and if all are good then proceed to
+ * generate coefficient tuning requests
+ */
+ } else {
+ /* Continue to send the same request: (2.)
+ * Continue to send HOLD until receive status NOTUPDATED
+ */
+ if (lane->krln.repeat_request_count >=
+ TIMEOUT_REPEAT_REQUEST) {
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(0) HOLD request without LP response timeout !\n");
+ /* Hold Request Timeout:
+ * continue to send HOLD until LP responds
+ * with NOTUPDATED
+ */
+ lane->krln.repeat_request_count = 0;
+ } else {
+ /* Allow LP some time to respond
+ * and repeat request
+ */
+ msleep(20);
+ /* Allow LP more time to respond,
+ * as the last chance,
+ * on the last time before issuing
+ * timeout error: (3.)
+ */
+ if (lane->krln.repeat_request_count ==
+ TIMEOUT_REPEAT_REQUEST - 1)
+ msleep(30);
+ lane->krln.repeat_request_count++;
+ }
+ ld_coef_update(lane);
+ return;
+ }
+ break;
+ case INCREMENT:
+ case DECREMENT:
+ /* previous request was: INC/DEC */
+ if (status_cz0 == COEF_NOTUPDATED) {
+ /* Continue to send the same request: (1.)
+ * Continue to send previous REQ until receive
+ * status UPDATED
+ */
+ if (lane->krln.repeat_request_count >=
+ TIMEOUT_REPEAT_REQUEST) {
+ if (prev_req_cz0 == INCREMENT)
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(0) INC request without LP response timeout !\n");
+ else
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(0) DEC request without LP response timeout !\n");
+ /* Request Timeout:
+ * just continue: proceed again to
+ * generate coefficient tuning requests
+ */
+ } else {
+ /* Allow LP some time to respond
+ * and repeat request
+ */
+ msleep(20);
+ /* Allow LP more time to respond, as the last
+ * chance, on the last time before issuing
+ * timeout error: (3.)
+ */
+ if (lane->krln.repeat_request_count ==
+ TIMEOUT_REPEAT_REQUEST - 1)
+ msleep(30);
+ lane->krln.repeat_request_count++;
+ ld_coef_update(lane);
+ return;
+ }
+ } else {
+ /* All good here:
+ * LP responded to this Request
+ * Sent HOLD for this coefficient
+ * before asking another request
+ * continue to check the other coefficient requests
+ */
+ lane->krln.ld_update &= ~COZ0_MASK;
+ }
+ break;
+ default:
+ /* previous request was: RESERVED: do nothing */
+ break;
+ }
+
+ switch (prev_req_cm1) {
+ case HOLD:
+ /* previous request was: HOLD */
+ if (status_cm1 == COEF_NOTUPDATED) {
+ /* All good here:
+ * continue to check the other coefficient requests
+ * and if all are good then proceed to
+ * generate coefficient tuning requests
+ */
+ } else {
+ /* Continue to send the same request: (2.)
+ * Continue to send HOLD until receive status
+ * NOTUPDATED
+ */
+ if (lane->krln.repeat_request_count >=
+ TIMEOUT_REPEAT_REQUEST) {
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(-1) HOLD request without LP response timeout !\n");
+ /* Hold Request Timeout:
+ * continue to send HOLD until
+ * LP responds with NOTUPDATED
+ */
+ lane->krln.repeat_request_count = 0;
+ } else {
+ /* Allow LP some time to respond
+ * and repeat request
+ */
+ msleep(20);
+ /* Allow LP more time to respond,
+ * as the last chance,
+ * on the last time
+ * before issuing timeout error: (3.)
+ */
+ if (lane->krln.repeat_request_count ==
+ TIMEOUT_REPEAT_REQUEST - 1)
+ msleep(30);
+ lane->krln.repeat_request_count++;
+ }
+ ld_coef_update(lane);
+ return;
+ }
+ break;
+ case INCREMENT:
+ case DECREMENT:
+ /* previous request was: INC/DEC */
+ if (status_cm1 == COEF_NOTUPDATED) {
+ /* Continue to send the same request: (1.)
+ * Continue to send previous REQ until receive status
+ * UPDATED
+ */
+ if (lane->krln.repeat_request_count >=
+ TIMEOUT_REPEAT_REQUEST) {
+ if (prev_req_cm1 == INCREMENT)
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(-1) INC request without LP response timeout !\n");
+ else
+ phydev_err(lane->phydev,
+ "REQ Timeout: Repeating C(-1) DEC request without LP response timeout !\n");
+ /* Request Timeout:
+ * just continue: proceed again to
+ * generate coefficient tuning requests
+ */
+ } else {
+ /* Allow LP some time to respond and repeat
+ * request
+ */
+ msleep(20);
+ /* Allow LP more time to respond, as the last
+ * chance, on the last time before issuing
+ * timeout error: (3.)
+ */
+ if (lane->krln.repeat_request_count ==
+ TIMEOUT_REPEAT_REQUEST - 1)
+ msleep(30);
+ lane->krln.repeat_request_count++;
+ ld_coef_update(lane);
+ return;
+ }
+ } else {
+ /* All good here:
+ * LP responded to this Request
+ * Sent HOLD for this coefficient
+ * before asking another request
+ * continue to check the other coefficient requests
+ */
+ lane->krln.ld_update &= ~COM1_MASK;
+ }
+ break;
+ default:
+ /* previous request was: RESERVED: do nothing */
+ break;
+ }
+
+ /* Reset repeat request counter:
+ * must be after all prev_req verifications above
+ */
+ lane->krln.repeat_request_count = 0;
+
+ if (lane->krln.prev_ld_update != lane->krln.ld_update) {
+ ld_coef_update(lane);
+ /* Redo these status checks and updates until we have no more
+ * changes, to speed up the overall process.
+ */
+ return;
+ }
+
+ /* Do nothing if we have pending request. */
+ if (prev_req_cp1 || prev_req_cz0 || prev_req_cm1)
+ return;
+ else if (lane->krln.lp_status & ALL_COEF_MASK)
+ /* No pending request but LP status was not reverted to
+ * not updated.
+ */
+ return;
+
+ /* Initialize status for the current step */
+ lane->krln.lt_error = false;
+
+ /* if CDR_LOCK = 0: Statistics are invalid */
+ if (!backplane_is_cdr_lock(lane, true)) {
+ if (eq_alg->ops.process_bad_state)
+ eq_alg->ops.process_bad_state(lane->krln.eq_priv);
+ return;
+ }
+
+ /* collect bit edge statistics */
+ if (!eq_alg->ops.collect_statistics(lane->krln.eq_priv))
+ return;
+
+ /* if CDR_LOCK = 0: Statistics are invalid */
+ if (!backplane_is_cdr_lock(lane, true)) {
+ if (eq_alg->ops.process_bad_state)
+ eq_alg->ops.process_bad_state(lane->krln.eq_priv);
+ return;
+ }
+
+ /* Check Rx */
+ if (!eq_alg->ops.is_rx_ok(lane->krln.eq_priv)) {
+ if (eq_alg->ops.process_bad_state)
+ eq_alg->ops.process_bad_state(lane->krln.eq_priv);
+ return;
+ }
+ eq_alg->ops.generate_request(lane->krln.eq_priv);
+
+ /* All C are in Hold and both Bins are stopped:
+ * So the Training is done
+ */
+ if (eq_alg->ops.is_eq_done(lane->krln.eq_priv)) {
+ trst->done_training = true;
+ training_complete(lane);
+ }
+}
+
+/* Training for Local Tx
+ * Initialize LD (Local Device)
+ */
+void lt_init_ld(struct lane_device *lane)
+{
+ /* report initial ld status to lp */
+ lane->krln.ld_status = 0;
+ ld_coef_status(lane);
+}
+
+/* Training for Local Tx
+ * This is the main routine for Local Tx training
+ */
+void lt_train_local_tx(struct lane_device *lane)
+{
+ struct training_status *trst = &lane->krln.trst;
+ int request, old_ld_status;
+
+ /* Check stop condition for Local Tx training */
+ trst->lp_rx_ready = lt_is_lp_rx_ready(lane);
+ if (trst->lp_rx_ready) {
+ /* LP receiver is ready
+ * As soon as the LP shows ready,
+ * no need to do any more updates.
+ */
+ lane->krln.ld_status &= ~ALL_COEF_MASK;
+ ld_coef_status(lane);
+
+ trst->local_tx_running = false;
+ return;
+ }
+
+ /* Start new Local Tx training step */
+ trst->local_tx_running = true;
+
+ /* get request from LP */
+ request = backplane_read_mmd(lane, lane->bpdev->bpkr.ltmmd.devad,
+ lane->bpdev->bpkr.ltmmd.lp_cu) &
+ LD_ALL_MASK;
+
+ old_ld_status = lane->krln.ld_status;
+
+ /* IEEE802.3-2008, 72.6.10.2.5
+ * Ensure we always go to NOT UDPATED for status reporting in
+ * response to HOLD requests.
+ * IEEE802.3-2008, 72.6.10.2.3.1/2
+ * ... but only if PRESET/INITIALIZE are not active to ensure
+ * we keep status until they are released.
+ *
+ * 72.6.10.2.5 Coefficient update process
+ * Once the updated, maximum, or minimum state is reported it continues
+ * to be reported until a hold request is received,
+ * after which the status reverts to not_updated.
+ */
+ if (!(request & (PRESET_MASK | INIT_MASK))) {
+ /* Reset status on HOLD request */
+ if (!(request & COP1_MASK))
+ lane->krln.ld_status &= ~COP1_MASK;
+
+ if (!(request & COZ0_MASK))
+ lane->krln.ld_status &= ~COZ0_MASK;
+
+ if (!(request & COM1_MASK))
+ lane->krln.ld_status &= ~COM1_MASK;
+
+ ld_coef_status(lane);
+ }
+
+ /* IEEE802.3-2008, 72.6.10.2.3.1/2
+ * only act on PRESET/INITIALIZE if all status is NOT UPDATED.
+ */
+ if (request & (PRESET_MASK | INIT_MASK)) {
+ if (!(lane->krln.ld_status & ALL_COEF_MASK)) {
+ if (request & PRESET_MASK)
+ preset(lane);
+
+ if (request & INIT_MASK) {
+ if (!lane->krln.first_recv_init) {
+ lane->krln.first_recv_init = true;
+ /* Init requests must be counted
+ * from initial handshake
+ */
+ lane->krln.req_ld_update_init_count = 1;
+ lane->krln.init_handshake_time =
+ jiffies_to_msecs(jiffies);
+ }
+ initialize(lane);
+ }
+ } else {
+ /* Inform the partner about current ld status
+ * which should be: ALL UPDATED for INIT and
+ * ALL MAX for PRESET
+ */
+ ld_coef_status(lane);
+ }
+ }
+
+ /* check if LP Coefficient are not in HOLD */
+ if (request & ALL_COEF_MASK)
+ check_request(lane, request & ALL_COEF_MASK);
+
+ /* Make sure the partner is always informed about the current ld status
+ * this will ensure avoidance of several training issues and errors:
+ * 'link_training_failed'
+ * 'Repeating request without LP response'
+ */
+ ld_coef_status(lane);
+}
+
+/* Training for Remote Tx API */
+
+/* lt_lp_update
+ *
+ * Sends to LP the specified request for coefficients update
+ *
+ * lane: desired lane for which to send lp update
+ * update: desired update request to be sent to LP
+ *
+ * Returns: None
+ */
+void lt_lp_update(struct lane_device *lane, u32 update)
+{
+ lane->krln.ld_update = update;
+ ld_coef_update(lane);
+}
+EXPORT_SYMBOL(lt_lp_update);
+
+/* lt_encode_request
+ *
+ * Encodes a request in the update word
+ * and adds it to other bit requests already existent in the update word
+ *
+ * base_update: base update word used to add a new desired request
+ * req: desired request type to be encoded
+ * field: the field for which the request must be encoded
+ *
+ * Returns: the encoded update word
+ */
+u32 lt_encode_request(u32 base_update, enum req_type req,
+ enum coef_field field)
+{
+ u32 new_cmd = base_update;
+ u32 cmd;
+
+ if (req >= REQ_INIT)
+ return RESERVED;
+
+ cmd = get_mask_for_req(req);
+
+ switch (field) {
+ case C_P1:
+ new_cmd |= (cmd << COP1_SHIFT);
+ break;
+ case C_Z0:
+ new_cmd |= (cmd << COZ0_SHIFT);
+ break;
+ case C_M1:
+ new_cmd |= (cmd << COM1_SHIFT);
+ break;
+ default:
+ return RESERVED;
+ }
+ return new_cmd;
+}
+EXPORT_SYMBOL(lt_encode_request);
+
+/* lt_encode_startup_request
+ *
+ * Encodes a startup request in the update word
+ *
+ * req: desired startup request type to be encoded
+ *
+ * Returns: the encoded update word
+ */
+u32 lt_encode_startup_request(enum req_type req)
+{
+ if (req == REQ_HOLD || req == REQ_INIT || req == REQ_PRESET)
+ return get_mask_for_req(req);
+
+ return RESERVED;
+}
+EXPORT_SYMBOL(lt_encode_startup_request);
+
+/* lt_decode_coef_update
+ *
+ * Decodes a request update for the specified field
+ *
+ * update: update word to be decoded
+ * field: desired field for which to decode the update
+ *
+ * Returns: the decoded request type
+ */
+enum req_type lt_decode_coef_update(u32 update, enum coef_field field)
+{
+ u32 cmd = HOLD;
+
+ switch (field) {
+ case C_P1:
+ cmd = (update & COP1_MASK) >> COP1_SHIFT;
+ break;
+ case C_Z0:
+ cmd = (update & COZ0_MASK) >> COZ0_SHIFT;
+ break;
+ case C_M1:
+ cmd = (update & COM1_MASK) >> COM1_SHIFT;
+ break;
+ default:
+ return REQ_INVALID;
+ }
+
+ return get_req_for_mask(cmd);
+}
+EXPORT_SYMBOL(lt_decode_coef_update);
+
+/* lt_is_update_of_type
+ *
+ * Checks if a request update is according to the specified type
+ * by checking the specific request bit in update word
+ *
+ * update: desired update word to be verified
+ * type: desired type to check against
+ *
+ * Returns: true if update is according to asked type or false otherwise
+ */
+bool lt_is_update_of_type(u32 update, enum req_type type)
+{
+ u32 mask = HOLD;
+
+ switch (type) {
+ case REQ_HOLD:
+ return (update == HOLD);
+ case REQ_INC:
+ mask |= (INCREMENT << COP1_SHIFT);
+ mask |= (INCREMENT << COZ0_SHIFT);
+ mask |= (INCREMENT << COM1_SHIFT);
+ return ((update & mask) != 0);
+ case REQ_DEC:
+ mask |= (DECREMENT << COP1_SHIFT);
+ mask |= (DECREMENT << COZ0_SHIFT);
+ mask |= (DECREMENT << COM1_SHIFT);
+ return ((update & mask) != 0);
+ case REQ_INIT:
+ return ((update & INIT_MASK) != 0);
+ case REQ_PRESET:
+ return ((update & PRESET_MASK) != 0);
+ default:
+ return false;
+ }
+ return false;
+}
+EXPORT_SYMBOL(lt_is_update_of_type);
+
+/* lt_is_lp_at_startup
+ *
+ * Checks if LP status is still at startup status: INIT or PRESET
+ *
+ * lane: desired lane to be verified
+ * req: request type to check startup status
+ * it makes sense only for INIT or PRESET requests
+ *
+ * Returns: true if LP status is still at startup status or false otherwise
+ */
+bool lt_is_lp_at_startup(struct lane_device *lane, enum req_type type)
+{
+ u32 lp_lcs = get_lp_lcs(lane);
+ u32 lp_st = lane->krln.lp_status;
+ bool lp_startup;
+
+ /* LP status still at Init/Preset:
+ * IF now LP status is Init/Preset
+ * OR (now LP status is NOTUPDATED
+ * AND the last nonzero LP status was Init/Preset)
+ */
+ switch (type) {
+ case REQ_INIT:
+ if (is_all_status(lp_st, COEF_UPDATED))
+ lp_startup = true;
+ else
+ lp_startup = is_all_status(lp_st, COEF_NOTUPDATED) &&
+ is_all_status(lp_lcs, COEF_UPDATED);
+ break;
+ case REQ_PRESET:
+ /* LP status still at Preset
+ * if now LP status is Preset
+ * OR now LP status is NOTUPDATED
+ * AND the last nonzero LP status was Preset
+ */
+ if (is_all_status(lp_st, COEF_MAX) ||
+ is_all_status(lp_st, COEF_UPDATED))
+ lp_startup = true;
+ else
+ lp_startup = is_all_status(lp_st, COEF_NOTUPDATED) &&
+ (is_all_status(lp_lcs, COEF_MAX) ||
+ is_all_status(lp_lcs, COEF_UPDATED));
+ break;
+ default:
+ return false;
+ }
+
+ return lp_startup;
+}
+EXPORT_SYMBOL(lt_is_lp_at_startup);
+
+/* lt_get_lp_coef_status
+ *
+ * Determines the last LP coefficient status
+ * according to IEEE802.3ap-2007:
+ * 72.6.10.2.5 Coefficient update process
+ *
+ * lane: desired lane to be verified
+ * field: coefficient field to be verified
+ *
+ * Returns: the last LP coefficient status
+ */
+enum coef_status lt_get_lp_coef_status(struct lane_device *lane,
+ enum coef_field field)
+{
+ return lane->krln.last_lp_update_status[field];
+}
+EXPORT_SYMBOL(lt_get_lp_coef_status);
+
+/* lt_set_error
+ *
+ * Sets or resets the LT (Link Training) Error flag
+ * This is used to signal to the generic kr training step procedure
+ * that an LT error state has occurred
+ * and link training cannot be successfully finished
+ *
+ * lane: desired lane to set lt error
+ * err: boolean value that specifies if set or reset the error flag
+ *
+ * Returns: None
+ */
+void lt_set_error(struct lane_device *lane, bool err)
+{
+ lane->krln.lt_error = err;
+}
+EXPORT_SYMBOL(lt_set_error);
+
+/* lt_move_lp_back
+ * Request LP to move back to previous coefficients setup and HOLD
+ * The procedure for sending this request is based on reverting the
+ * latest change request (non-hold update) for all coefficients
+ * This procedure should be used to exit from bad states like not CDR_Lock
+ *
+ * lane: desired lane for which to send lp update
+ *
+ * Returns: None
+ */
+void lt_move_lp_back(struct lane_device *lane)
+{
+ u32 prev_req_cp1 = (lane->krln.ld_last_nonhold_update & COP1_MASK) >>
+ COP1_SHIFT;
+ u32 prev_req_cz0 = (lane->krln.ld_last_nonhold_update & COZ0_MASK) >>
+ COZ0_SHIFT;
+ u32 prev_req_cm1 = (lane->krln.ld_last_nonhold_update & COM1_MASK) >>
+ COM1_SHIFT;
+ u32 temp;
+
+ /* Move back to previous C(-1), C(0), C(+1) and HOLD */
+ temp = HOLD;
+ switch (prev_req_cp1) {
+ case INCREMENT:
+ temp |= DECREMENT << COP1_SHIFT;
+ break;
+ case DECREMENT:
+ temp |= INCREMENT << COP1_SHIFT;
+ break;
+ }
+ switch (prev_req_cz0) {
+ case INCREMENT:
+ temp |= DECREMENT << COZ0_SHIFT;
+ break;
+ case DECREMENT:
+ temp |= INCREMENT << COZ0_SHIFT;
+ break;
+ }
+ switch (prev_req_cm1) {
+ case INCREMENT:
+ temp |= DECREMENT << COM1_SHIFT;
+ break;
+ case DECREMENT:
+ temp |= INCREMENT << COM1_SHIFT;
+ break;
+ }
+
+ lane->krln.ld_update = temp;
+ ld_coef_update(lane);
+
+ /* start the procedure for sending request to move LP back
+ * to previous setup until LP responds to it
+ */
+ lane->krln.move_back_prev = true;
+ lane->krln.move_back_cnt = 0;
+ lane->krln.move_back_lp_status = 0;
+ if (prev_req_cp1 == HOLD)
+ lane->krln.move_back_lp_status |= (COEF_UPDATED << COP1_SHIFT);
+ if (prev_req_cz0 == HOLD)
+ lane->krln.move_back_lp_status |= (COEF_UPDATED << COZ0_SHIFT);
+ if (prev_req_cm1 == HOLD)
+ lane->krln.move_back_lp_status |= (COEF_UPDATED << COM1_SHIFT);
+}
+EXPORT_SYMBOL(lt_move_lp_back);
diff --git a/drivers/net/phy/backplane/link_training.h b/drivers/net/phy/backplane/link_training.h
new file mode 100644
index 0000000..fae5f35
--- /dev/null
+++ b/drivers/net/phy/backplane/link_training.h
@@ -0,0 +1,32 @@
+/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */
+/* Link Training (IEEE802.3ap/ba)
+ *
+ * Copyright 2019-2020 NXP
+ */
+
+#ifndef __LINK_TRAINING_H
+#define __LINK_TRAINING_H
+
+#include "backplane.h"
+
+/* Link Training interface with backplane driver */
+
+void lt_start(struct lane_device *lane);
+void lt_stop(struct lane_device *lane);
+void lt_reset(struct lane_device *lane);
+void lt_init_ld(struct lane_device *lane);
+
+void lt_mmd_c45(struct backplane_kr *bpkr);
+void lt_mmd_setup(struct backplane_kr *bpkr, int devad, u32 base);
+
+bool lt_is_rx_trained(struct lane_device *lane);
+bool lt_is_training_failure(struct lane_device *lane);
+bool lt_is_frame_lock(struct lane_device *lane);
+
+bool lt_is_lp_rx_ready(struct lane_device *lane);
+bool lt_is_ld_rx_ready(struct lane_device *lane);
+
+void lt_train_remote_tx(struct lane_device *lane);
+void lt_train_local_tx(struct lane_device *lane);
+
+#endif /* __LINK_TRAINING_H */
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 13:35 ` [PATCH net-next v3 4/7] net: phy: add backplane kr driver support Florinel Iordache
@ 2020-06-22 14:24 ` Andrew Lunn
2020-06-22 14:39 ` [EXT] " Florinel Iordache
2020-06-22 15:08 ` Madalin Bucur (OSS)
2020-06-22 21:46 ` Jakub Kicinski
1 sibling, 2 replies; 23+ messages in thread
From: Andrew Lunn @ 2020-06-22 14:24 UTC (permalink / raw)
To: Florinel Iordache
Cc: davem, netdev, f.fainelli, hkallweit1, linux, devicetree,
linux-doc, robh+dt, mark.rutland, kuba, corbet, shawnguo,
leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel
On Mon, Jun 22, 2020 at 04:35:21PM +0300, Florinel Iordache wrote:
> Add support for backplane kr generic driver including link training
> (ieee802.3ap/ba) and fixed equalization algorithm
Hi Florinel
This is still a PHY device. I don't remember any discussions which
resolved the issues of if at the end of the backplane there is another
PHY.
It makes little sense to repost this code until we have this problem
discussed and a way forward decided on. It fits into the discussion
Russell and Ioana are having about representing PCS drivers. Please
contribute to that.
Andrew
^ permalink raw reply [flat|nested] 23+ messages in thread
* RE: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 14:24 ` Andrew Lunn
@ 2020-06-22 14:39 ` Florinel Iordache
2020-06-26 19:05 ` Florian Fainelli
2020-06-22 15:08 ` Madalin Bucur (OSS)
1 sibling, 1 reply; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 14:39 UTC (permalink / raw)
To: Andrew Lunn
Cc: davem, netdev, f.fainelli, hkallweit1, linux, devicetree,
linux-doc, robh+dt, mark.rutland, kuba, corbet, shawnguo, Leo Li,
Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel, Florinel Iordache
> -----Original Message-----
> From: Andrew Lunn <andrew@lunn.ch>
> Sent: Monday, June 22, 2020 5:25 PM
> To: Florinel Iordache <florinel.iordache@nxp.com>
> Cc: davem@davemloft.net; netdev@vger.kernel.org; f.fainelli@gmail.com;
> hkallweit1@gmail.com; linux@armlinux.org.uk; devicetree@vger.kernel.org;
> linux-doc@vger.kernel.org; robh+dt@kernel.org; mark.rutland@arm.com;
> kuba@kernel.org; corbet@lwn.net; shawnguo@kernel.org; Leo Li
> <leoyang.li@nxp.com>; Madalin Bucur (OSS) <madalin.bucur@oss.nxp.com>;
> Ioana Ciornei <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
> Subject: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver
> support
>
> Caution: EXT Email
>
> On Mon, Jun 22, 2020 at 04:35:21PM +0300, Florinel Iordache wrote:
> > Add support for backplane kr generic driver including link training
> > (ieee802.3ap/ba) and fixed equalization algorithm
>
> Hi Florinel
>
> This is still a PHY device. I don't remember any discussions which resolved the
> issues of if at the end of the backplane there is another PHY.
>
> It makes little sense to repost this code until we have this problem discussed and
> a way forward decided on. It fits into the discussion Russell and Ioana are having
> about representing PCS drivers. Please contribute to that.
>
> Andrew
Hi Andrew,
Yes, you are right: we decided to send only support for DPAA1 using current approach as a PHY device
(as mentioned in cover-letter), until PCS representation will be fully clarified.
The entire DPAA2 support was removed for now, together with phylink changes.
DPAA1 maintainer (Madalin Bucur) agrees with current representation as a PHY device for DPAA1.
So we would like to have some discussions around this approach for DPAA1 only, as it seems suitable for us.
Regards,
Florinel.
^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 14:39 ` [EXT] " Florinel Iordache
@ 2020-06-26 19:05 ` Florian Fainelli
2020-06-29 13:23 ` Florinel Iordache
0 siblings, 1 reply; 23+ messages in thread
From: Florian Fainelli @ 2020-06-26 19:05 UTC (permalink / raw)
To: Florinel Iordache, Andrew Lunn
Cc: davem, netdev, hkallweit1, linux, devicetree, linux-doc, robh+dt,
mark.rutland, kuba, corbet, shawnguo, Leo Li, Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel
On 6/22/20 7:39 AM, Florinel Iordache wrote:
>> -----Original Message-----
>> From: Andrew Lunn <andrew@lunn.ch>
>> Sent: Monday, June 22, 2020 5:25 PM
>> To: Florinel Iordache <florinel.iordache@nxp.com>
>> Cc: davem@davemloft.net; netdev@vger.kernel.org; f.fainelli@gmail.com;
>> hkallweit1@gmail.com; linux@armlinux.org.uk; devicetree@vger.kernel.org;
>> linux-doc@vger.kernel.org; robh+dt@kernel.org; mark.rutland@arm.com;
>> kuba@kernel.org; corbet@lwn.net; shawnguo@kernel.org; Leo Li
>> <leoyang.li@nxp.com>; Madalin Bucur (OSS) <madalin.bucur@oss.nxp.com>;
>> Ioana Ciornei <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
>> Subject: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver
>> support
>>
>> Caution: EXT Email
>>
>> On Mon, Jun 22, 2020 at 04:35:21PM +0300, Florinel Iordache wrote:
>>> Add support for backplane kr generic driver including link training
>>> (ieee802.3ap/ba) and fixed equalization algorithm
>>
>> Hi Florinel
>>
>> This is still a PHY device. I don't remember any discussions which resolved the
>> issues of if at the end of the backplane there is another PHY.
>>
>> It makes little sense to repost this code until we have this problem discussed and
>> a way forward decided on. It fits into the discussion Russell and Ioana are having
>> about representing PCS drivers. Please contribute to that.
>>
>> Andrew
>
> Hi Andrew,
>
> Yes, you are right: we decided to send only support for DPAA1 using current approach as a PHY device
> (as mentioned in cover-letter), until PCS representation will be fully clarified.
> The entire DPAA2 support was removed for now, together with phylink changes.
> DPAA1 maintainer (Madalin Bucur) agrees with current representation as a PHY device for DPAA1.
> So we would like to have some discussions around this approach for DPAA1 only, as it seems suitable for us.
The question is really whether it is suitable for others beyond NXP, the
drivers are certainly organized in such a way that there is little NXP
specifics in them so the intent is clearly there.
We will probably not know, either because vendors have decided to hide
all of this stuff under firmware, or they do not use Linux or they just
are not following what is going on upstream and have no desire to
participate.
--
Florian
^ permalink raw reply [flat|nested] 23+ messages in thread
* RE: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-26 19:05 ` Florian Fainelli
@ 2020-06-29 13:23 ` Florinel Iordache
0 siblings, 0 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-29 13:23 UTC (permalink / raw)
To: Florian Fainelli, Andrew Lunn
Cc: davem, netdev, hkallweit1, linux, devicetree, linux-doc, robh+dt,
mark.rutland, kuba, corbet, shawnguo, Leo Li, Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel, Florinel Iordache
> -----Original Message-----
> From: Florian Fainelli <f.fainelli@gmail.com>
> Sent: Friday, June 26, 2020 10:05 PM
> To: Florinel Iordache <florinel.iordache@nxp.com>; Andrew Lunn
> <andrew@lunn.ch>
> Cc: davem@davemloft.net; netdev@vger.kernel.org; hkallweit1@gmail.com;
> linux@armlinux.org.uk; devicetree@vger.kernel.org; linux-doc@vger.kernel.org;
> robh+dt@kernel.org; mark.rutland@arm.com; kuba@kernel.org;
> corbet@lwn.net; shawnguo@kernel.org; Leo Li <leoyang.li@nxp.com>; Madalin
> Bucur (OSS) <madalin.bucur@oss.nxp.com>; Ioana Ciornei
> <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
> Subject: Re: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver
> support
>
> Caution: EXT Email
>
> On 6/22/20 7:39 AM, Florinel Iordache wrote:
> >> -----Original Message-----
> >> From: Andrew Lunn <andrew@lunn.ch>
> >> Sent: Monday, June 22, 2020 5:25 PM
> >> To: Florinel Iordache <florinel.iordache@nxp.com>
> >> Cc: davem@davemloft.net; netdev@vger.kernel.org;
> >> f.fainelli@gmail.com; hkallweit1@gmail.com; linux@armlinux.org.uk;
> >> devicetree@vger.kernel.org; linux-doc@vger.kernel.org;
> >> robh+dt@kernel.org; mark.rutland@arm.com; kuba@kernel.org;
> >> corbet@lwn.net; shawnguo@kernel.org; Leo Li <leoyang.li@nxp.com>;
> >> Madalin Bucur (OSS) <madalin.bucur@oss.nxp.com>; Ioana Ciornei
> >> <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
> >> Subject: [EXT] Re: [PATCH net-next v3 4/7] net: phy: add backplane kr
> >> driver support
> >>
> >> Caution: EXT Email
> >>
> >> On Mon, Jun 22, 2020 at 04:35:21PM +0300, Florinel Iordache wrote:
> >>> Add support for backplane kr generic driver including link training
> >>> (ieee802.3ap/ba) and fixed equalization algorithm
> >>
> >> Hi Florinel
> >>
> >> This is still a PHY device. I don't remember any discussions which
> >> resolved the issues of if at the end of the backplane there is another PHY.
> >>
> >> It makes little sense to repost this code until we have this problem
> >> discussed and a way forward decided on. It fits into the discussion
> >> Russell and Ioana are having about representing PCS drivers. Please
> contribute to that.
> >>
> >> Andrew
> >
> > Hi Andrew,
> >
> > Yes, you are right: we decided to send only support for DPAA1 using
> > current approach as a PHY device (as mentioned in cover-letter), until PCS
> representation will be fully clarified.
> > The entire DPAA2 support was removed for now, together with phylink
> changes.
> > DPAA1 maintainer (Madalin Bucur) agrees with current representation as a PHY
> device for DPAA1.
> > So we would like to have some discussions around this approach for DPAA1
> only, as it seems suitable for us.
>
> The question is really whether it is suitable for others beyond NXP, the drivers
> are certainly organized in such a way that there is little NXP specifics in them so
> the intent is clearly there.
>
> We will probably not know, either because vendors have decided to hide all of
> this stuff under firmware, or they do not use Linux or they just are not following
> what is going on upstream and have no desire to participate.
> --
> Florian
Hi Florian,
This is correct: backplane support has a modular, extensible, generic architecture
and the modules are completely disconnected
so they can be reused among different configurations setups.
Therefore we have encapsulated the standard backplane functionality in several
generic modules like: Ethernet Backplane Generic Driver, Link Training and
Auto-negotiation including: IEEE 802.3-ap/ba standards, Equalization Algorithms
(that include: Fixed algorithm and BEE - Bit Edge equalization algorithm).
Device specific modules are used to enable QorIQ family of devices.
This architecture is described in detail in Doc file: backplane.rst
Other vendors that want to enable backplane for their devices should
add only their device specific modules (similar with qoriq modules).
These modules basically must describe device specific registers and
make the connection between backplane generic API services and device specific operations.
All generic modules that encapsulate standard backplane functionality can be
reused by other vendors but this is not mandatory.
Other vendors can extend current architecture with new generic modules:
for example other equalization algorithms and standards can be added in the future
if currently existing ones are not desired or considered inappropriate.
There are several other standard algorithms available that can be used for
Signal equalization: they just have to be implemented and integrated here.
Of course we validated this backplane architecture only on NXP platforms (by using QorIQ devices)
but other vendors will be able to use it in the future on their own platforms.
Thank you for feedback,
Florinel.
^ permalink raw reply [flat|nested] 23+ messages in thread
* RE: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 14:24 ` Andrew Lunn
2020-06-22 14:39 ` [EXT] " Florinel Iordache
@ 2020-06-22 15:08 ` Madalin Bucur (OSS)
2020-06-26 19:02 ` Florian Fainelli
1 sibling, 1 reply; 23+ messages in thread
From: Madalin Bucur (OSS) @ 2020-06-22 15:08 UTC (permalink / raw)
To: Andrew Lunn, Florinel Iordache
Cc: davem, netdev, f.fainelli, hkallweit1, linux, devicetree,
linux-doc, robh+dt, mark.rutland, kuba, corbet, shawnguo, Leo Li,
Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel
> -----Original Message-----
> From: Andrew Lunn <andrew@lunn.ch>
> Sent: Monday, June 22, 2020 5:25 PM
> To: Florinel Iordache <florinel.iordache@nxp.com>
> Cc: davem@davemloft.net; netdev@vger.kernel.org; f.fainelli@gmail.com;
> hkallweit1@gmail.com; linux@armlinux.org.uk; devicetree@vger.kernel.org;
> linux-doc@vger.kernel.org; robh+dt@kernel.org; mark.rutland@arm.com;
> kuba@kernel.org; corbet@lwn.net; shawnguo@kernel.org; Leo Li
> <leoyang.li@nxp.com>; Madalin Bucur (OSS) <madalin.bucur@oss.nxp.com>;
> Ioana Ciornei <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
> Subject: Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver
> support
>
> On Mon, Jun 22, 2020 at 04:35:21PM +0300, Florinel Iordache wrote:
> > Add support for backplane kr generic driver including link training
> > (ieee802.3ap/ba) and fixed equalization algorithm
>
> Hi Florinel
>
> This is still a PHY device. I don't remember any discussions which
> resolved the issues of if at the end of the backplane there is another
> PHY.
>
> It makes little sense to repost this code until we have this problem
> discussed and a way forward decided on. It fits into the discussion
> Russell and Ioana are having about representing PCS drivers. Please
> contribute to that.
>
> Andrew
Hi Andrew, the reasons behind this selection:
- the PCS that is controlled by the backplane driver belongs to the PHY
layer so the representation as a PHY device is legitimate
- the PHY driver provides the state machine that is required, not using
this representation backplane would need to add a separate, duplicate
state machine
- the limitation, that only one PHY layer entity can be managed by the
PHYLib, is a known limitation that always existed, is not introduced by
the backplane support; the unsupported scenario with a backplane connection
to a PHY entity that needs to be managed relates to that limitation and
a solution for it should not be added through the backplane support
- afaik, Russell and Ioana are discussing the PCS representation in the
context of PHYLink, this submission is using PHYLib. If we are to discuss
about the PCS representation, it's the problem of the simplistic "one device
in the PHY layer" issue that needs to be addressed to have a proper PCS
representation at all times.
Madalin
^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 15:08 ` Madalin Bucur (OSS)
@ 2020-06-26 19:02 ` Florian Fainelli
2020-06-29 13:58 ` Russell King - ARM Linux admin
0 siblings, 1 reply; 23+ messages in thread
From: Florian Fainelli @ 2020-06-26 19:02 UTC (permalink / raw)
To: Madalin Bucur (OSS), Andrew Lunn, Florinel Iordache
Cc: davem, netdev, hkallweit1, linux, devicetree, linux-doc, robh+dt,
mark.rutland, kuba, corbet, shawnguo, Leo Li, Ioana Ciornei,
linux-kernel
On 6/22/20 8:08 AM, Madalin Bucur (OSS) wrote:
>> -----Original Message-----
>> From: Andrew Lunn <andrew@lunn.ch>
>> Sent: Monday, June 22, 2020 5:25 PM
>> To: Florinel Iordache <florinel.iordache@nxp.com>
>> Cc: davem@davemloft.net; netdev@vger.kernel.org; f.fainelli@gmail.com;
>> hkallweit1@gmail.com; linux@armlinux.org.uk; devicetree@vger.kernel.org;
>> linux-doc@vger.kernel.org; robh+dt@kernel.org; mark.rutland@arm.com;
>> kuba@kernel.org; corbet@lwn.net; shawnguo@kernel.org; Leo Li
>> <leoyang.li@nxp.com>; Madalin Bucur (OSS) <madalin.bucur@oss.nxp.com>;
>> Ioana Ciornei <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
>> Subject: Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver
>> support
>>
>> On Mon, Jun 22, 2020 at 04:35:21PM +0300, Florinel Iordache wrote:
>>> Add support for backplane kr generic driver including link training
>>> (ieee802.3ap/ba) and fixed equalization algorithm
>>
>> Hi Florinel
>>
>> This is still a PHY device. I don't remember any discussions which
>> resolved the issues of if at the end of the backplane there is another
>> PHY.
>>
>> It makes little sense to repost this code until we have this problem
>> discussed and a way forward decided on. It fits into the discussion
>> Russell and Ioana are having about representing PCS drivers. Please
>> contribute to that.
>>
>> Andrew
>
> Hi Andrew, the reasons behind this selection:
>
> - the PCS that is controlled by the backplane driver belongs to the PHY
> layer so the representation as a PHY device is legitimate
That argument makes sense.
> - the PHY driver provides the state machine that is required, not using
> this representation backplane would need to add a separate, duplicate
> state machine
Which is entirely permissible according to the PHY library
documentation, not that we have seen many people do it though, even less
so when the PHY driver is providing the state machine.
> - the limitation, that only one PHY layer entity can be managed by the
> PHYLib, is a known limitation that always existed, is not introduced by
> the backplane support; the unsupported scenario with a backplane connection
> to a PHY entity that needs to be managed relates to that limitation and
> a solution for it should not be added through the backplane support
> - afaik, Russell and Ioana are discussing the PCS representation in the
> context of PHYLink, this submission is using PHYLib. If we are to discuss
> about the PCS representation, it's the problem of the simplistic "one device
> in the PHY layer" issue that needs to be addressed to have a proper PCS
> representation at all times.
So would not it make sense for the PCS representation to be settled and
then add the backplane driver implementation such that there is no
double work happening for Florinel and for reviewers and the PCS
implementation als factors in the backplane use case and requirements?
--
Florian
^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-26 19:02 ` Florian Fainelli
@ 2020-06-29 13:58 ` Russell King - ARM Linux admin
2020-06-30 0:51 ` Andrew Lunn
0 siblings, 1 reply; 23+ messages in thread
From: Russell King - ARM Linux admin @ 2020-06-29 13:58 UTC (permalink / raw)
To: Florian Fainelli
Cc: Madalin Bucur (OSS),
Andrew Lunn, Florinel Iordache, davem, netdev, hkallweit1,
devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, Leo Li, Ioana Ciornei, linux-kernel
On Fri, Jun 26, 2020 at 12:02:05PM -0700, Florian Fainelli wrote:
> On 6/22/20 8:08 AM, Madalin Bucur (OSS) wrote:
> > Hi Andrew, the reasons behind this selection:
> >
> > - the PCS that is controlled by the backplane driver belongs to the PHY
> > layer so the representation as a PHY device is legitimate
>
> That argument makes sense.
It doesn't when you also are subjected to other parts of NXP arguing
that the PCS is tightly bound inside the SoC and therefore should be
effectively a library - as has been discussed in the threads about
the Lynx PCS.
> > - the PHY driver provides the state machine that is required, not using
> > this representation backplane would need to add a separate, duplicate
> > state machine
>
> Which is entirely permissible according to the PHY library
> documentation, not that we have seen many people do it though, even less
> so when the PHY driver is providing the state machine.
It seems the PHYlib state machine is getting smaller and smaller
as we move forward; phy_state_machine() is now looking very bare
compared to what it used to look like. I think it's not that far
off being eliminated.
> > - the limitation, that only one PHY layer entity can be managed by the
> > PHYLib, is a known limitation that always existed, is not introduced by
> > the backplane support; the unsupported scenario with a backplane connection
> > to a PHY entity that needs to be managed relates to that limitation and
> > a solution for it should not be added through the backplane support
> > - afaik, Russell and Ioana are discussing the PCS representation in the
> > context of PHYLink, this submission is using PHYLib. If we are to discuss
> > about the PCS representation, it's the problem of the simplistic "one device
> > in the PHY layer" issue that needs to be addressed to have a proper PCS
> > representation at all times.
>
> So would not it make sense for the PCS representation to be settled and
> then add the backplane driver implementation such that there is no
> double work happening for Florinel and for reviewers and the PCS
> implementation als factors in the backplane use case and requirements?
Yes, that is my assessment; there's a lot of work going on in different
areas in QoriQ networking, and it seems people are pulling things in
quite diverse directions.
If we're not careful, we're going to end up with the Lynx PCS being
implemented one way, and backplane PCS being implemented completely
differently and preventing any hope of having a backplane PCS
connected to a conventional copper PHY.
I think folk at NXP need to stop, stand back, and look at the bigger
picture about how they want to integrate all these individual,
independent strands of development into the kernel, and come up with
a common approach that also satisfies the mainline kernel, rather
than having individual discussions with mainline kernel maintainers
on public lists. What I'm saying is, it isn't our job to co-ordinate
between the different parts of NXP - that's fairly and squarely
NXP's problem to sort out themselves.
So, I think, further progress in public on backplane support needs to
wait until we have the general situation for PCS resolved.
Makes sense?
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-29 13:58 ` Russell King - ARM Linux admin
@ 2020-06-30 0:51 ` Andrew Lunn
0 siblings, 0 replies; 23+ messages in thread
From: Andrew Lunn @ 2020-06-30 0:51 UTC (permalink / raw)
To: Russell King - ARM Linux admin
Cc: Florian Fainelli, Madalin Bucur (OSS),
Florinel Iordache, davem, netdev, hkallweit1, devicetree,
linux-doc, robh+dt, mark.rutland, kuba, corbet, shawnguo, Leo Li,
Ioana Ciornei, linux-kernel
> So, I think, further progress in public on backplane support needs to
> wait until we have the general situation for PCS resolved.
>
> Makes sense?
Hi Russell
Does to me.
Thanks
Andrew
^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [PATCH net-next v3 4/7] net: phy: add backplane kr driver support
2020-06-22 13:35 ` [PATCH net-next v3 4/7] net: phy: add backplane kr driver support Florinel Iordache
2020-06-22 14:24 ` Andrew Lunn
@ 2020-06-22 21:46 ` Jakub Kicinski
1 sibling, 0 replies; 23+ messages in thread
From: Jakub Kicinski @ 2020-06-22 21:46 UTC (permalink / raw)
To: Florinel Iordache
Cc: davem, netdev, andrew, f.fainelli, hkallweit1, linux, devicetree,
linux-doc, robh+dt, mark.rutland, corbet, shawnguo, leoyang.li,
madalin.bucur, ioana.ciornei, linux-kernel
On Mon, 22 Jun 2020 16:35:21 +0300 Florinel Iordache wrote:
> Add support for backplane kr generic driver including link training
> (ieee802.3ap/ba) and fixed equalization algorithm
>
> Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
drivers/net/phy/backplane/backplane.c:60:11: warning: symbol 'backplane_common_features_array' was not declared. Should it be static?
drivers/net/phy/backplane/backplane.c:66:11: warning: symbol 'backplane_protocol_features_array' was not declared. Should it be static?
drivers/net/phy/backplane/backplane.c:1204:40: warning: incorrect type in assignment (different address spaces)
drivers/net/phy/backplane/backplane.c:1204:40: expected void *[assigned] reg_base
drivers/net/phy/backplane/backplane.c:1204:40: got void [noderef] <asn:2> *reg_base
drivers/net/phy/backplane/backplane.c: In function ‘bp_kr_state_machine’:
drivers/net/phy/backplane/backplane.c:590:27: warning: variable ‘bpdev’ set but not used [-Wunused-but-set-variable]
590 | struct backplane_device *bpdev;
| ^~~~~
drivers/net/phy/backplane/link_training.c: In function ‘lt_train_remote_tx’:
drivers/net/phy/backplane/link_training.c:557:6: warning: variable ‘lp_resp_time’ set but not used [-Wunused-but-set-variable]
557 | u64 lp_resp_time;
| ^~~~~~~~~~~~
drivers/net/phy/backplane/link_training.c: In function ‘lt_train_local_tx’:
drivers/net/phy/backplane/link_training.c:1143:15: warning: variable ‘old_ld_status’ set but not used [-Wunused-but-set-variable]
1143 | int request, old_ld_status;
| ^~~~~~~~~~~~~
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH net-next v3 5/7] net: phy: enable qoriq backplane support
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
` (3 preceding siblings ...)
2020-06-22 13:35 ` [PATCH net-next v3 4/7] net: phy: add backplane kr driver support Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 6/7] net: phy: add bee algorithm for kr training Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio description Florinel Iordache
6 siblings, 0 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Enable backplane support for qoriq family of devices
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
drivers/net/phy/backplane/Kconfig | 11 +-
drivers/net/phy/backplane/Makefile | 2 +
drivers/net/phy/backplane/qoriq_backplane.c | 473 ++++++++++++++++++++++++++
drivers/net/phy/backplane/qoriq_backplane.h | 42 +++
drivers/net/phy/backplane/qoriq_serdes_10g.c | 486 +++++++++++++++++++++++++++
5 files changed, 1013 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/phy/backplane/qoriq_backplane.c
create mode 100644 drivers/net/phy/backplane/qoriq_backplane.h
create mode 100644 drivers/net/phy/backplane/qoriq_serdes_10g.c
diff --git a/drivers/net/phy/backplane/Kconfig b/drivers/net/phy/backplane/Kconfig
index 9ec54b5..3e20a78 100644
--- a/drivers/net/phy/backplane/Kconfig
+++ b/drivers/net/phy/backplane/Kconfig
@@ -17,4 +17,13 @@ config ETH_BACKPLANE_FIXED
This module provides a driver to setup fixed user configurable
coefficient values for backplanes equalization. This means
No Equalization algorithm is used to adapt the initial coefficients
- initially set by the user.
\ No newline at end of file
+ initially set by the user.
+
+config ETH_BACKPLANE_QORIQ
+ tristate "QorIQ Ethernet Backplane driver"
+ depends on ETH_BACKPLANE
+ help
+ This module provides a driver for Ethernet Operation over
+ Electrical Backplanes enabled for QorIQ family of devices.
+ This driver is using the services provided by the generic
+ backplane and link training modules.
\ No newline at end of file
diff --git a/drivers/net/phy/backplane/Makefile b/drivers/net/phy/backplane/Makefile
index ded6f2d..3ae3d8b 100644
--- a/drivers/net/phy/backplane/Makefile
+++ b/drivers/net/phy/backplane/Makefile
@@ -5,5 +5,7 @@
obj-$(CONFIG_ETH_BACKPLANE) += eth_backplane.o
obj-$(CONFIG_ETH_BACKPLANE_FIXED) += eq_fixed.o
+obj-$(CONFIG_ETH_BACKPLANE_QORIQ) += eth_backplane_qoriq.o
eth_backplane-objs := backplane.o link_training.o
+eth_backplane_qoriq-objs := qoriq_backplane.o qoriq_serdes_10g.o
diff --git a/drivers/net/phy/backplane/qoriq_backplane.c b/drivers/net/phy/backplane/qoriq_backplane.c
new file mode 100644
index 0000000..eb22e02
--- /dev/null
+++ b/drivers/net/phy/backplane/qoriq_backplane.c
@@ -0,0 +1,473 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+/* QorIQ Backplane driver
+ *
+ * Copyright 2015 Freescale Semiconductor, Inc.
+ * Copyright 2018-2020 NXP
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mii.h>
+#include <linux/mdio.h>
+#include <linux/io.h>
+#include <linux/netdevice.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+
+#include "qoriq_backplane.h"
+
+/* QorIQ Backplane Driver name */
+#define QORIQ_BACKPLANE_DRIVER_NAME "backplane_qoriq"
+
+/* QorIQ Backplane Driver version */
+#define QORIQ_BACKPLANE_DRIVER_VERSION "1.0.0"
+
+/* PCS Device Identifier */
+#define PCS_PHY_DEVICE_ID 0x0083e400
+#define PCS_PHY_DEVICE_ID_MASK 0xffffffff
+
+/* Max/Min coefficients range values */
+#define PRE_COEF_MAX 0x0
+#define PRE_COEF_MIN 0x8
+#define POST_COEF_MAX 0x0
+#define POST_COEF_MIN 0x10
+#define ZERO_COEF_MIN 0x1A
+#define ZERO_COEF_MAX 0x30
+
+/* Coefficients sum ratio: (their sum divided by their difference) */
+#define COEF_SUM_RATIO_NUMERATOR 17
+#define COEF_SUM_RATIO_DENOMINATOR 4
+
+/* Number of equalization custom parameters */
+#define EQ_PARAMS_NO 1
+
+/* Serdes types supported by QorIQ devices */
+enum serdes_type {
+ SERDES_10G,
+ SERDES_INVAL
+};
+
+static void an_advertisement_init(struct lane_device *lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ struct qoriq_driver *qoriq_drv;
+ u32 init_an_adv1;
+ int err;
+
+ qoriq_drv = (struct qoriq_driver *)bpdev->drv.priv;
+ init_an_adv1 = backplane_get_an_adv1_init(phydev->interface);
+
+ err = backplane_write_mmd(lane, MDIO_MMD_AN, qoriq_drv->an_adv1,
+ init_an_adv1);
+ if (err)
+ phydev_err(phydev,
+ "Setting AN register 0x%02x on lane %d failed with error code: 0x%08x\n",
+ qoriq_drv->an_adv1, lane->idx, err);
+}
+
+static bool is_an_link_detected(struct lane_device *lane)
+{
+ struct backplane_device *bpdev = lane->bpdev;
+ struct phy_device *phydev = lane->phydev;
+ struct qoriq_driver *qoriq_drv;
+ struct lane_device *masterln;
+ u32 an_bp_eth_status;
+ int an_state;
+ u32 an_mask;
+
+ qoriq_drv = (struct qoriq_driver *)bpdev->drv.priv;
+ an_bp_eth_status = qoriq_drv->an_bp_eth_status;
+
+ /* Check AN state only on Master Lane */
+ masterln = &bpdev->lane[MASTER_LANE];
+
+ /* The link training occurs after auto-negotiation
+ * has determined the link to be a Base-KR link.
+ * This is indicated by asserting the corresponding
+ * technology bit within the BP_ETH_STATUS register.
+ * Note that this occurs before auto-negotiation can declare
+ * auto-negotiation complete,
+ * as this requires the PCS to report a valid link.
+ */
+ an_mask = backplane_get_an_bp_eth_status_bit(phydev->interface);
+ an_state = backplane_read_mmd(masterln, MDIO_MMD_AN, an_bp_eth_status);
+
+ return (an_state & an_mask);
+}
+
+static void qoriq_setup_default_settings(struct lane_device *lane)
+{
+ const struct lane_ops *lane_ops = lane->bpdev->drv.lane_ops;
+ struct qoriq_lane_ops *qoriq_lane_ops;
+ u32 def_amp_red;
+
+ qoriq_lane_ops = (struct qoriq_lane_ops *)lane_ops->priv;
+
+ if (lane->bpdev->bpkr.valid_eq_params)
+ def_amp_red = lane->bpdev->bpkr.def_kr.dev_coef[IDX_AMP_RED];
+ else
+ def_amp_red = qoriq_lane_ops->read_amp_red(lane->reg_base);
+
+ lane->krln.def_kr.dev_coef[IDX_AMP_RED] = def_amp_red;
+}
+
+/* LT HW restrictions:
+ * Section 5.3.1 10GBaseKR Transmit Adaptive Equalization Control
+ * additional restrictions set down by 802.3 specification Clause 72,
+ * specifically 72.7.1.11 Transmitter output waveform requirements
+ *
+ * Maintaining the following relationships limit transmit equalization
+ * to reasonable levels compliant with the KR specification
+ */
+static int lt_qoriq_validation(struct lane_device *lane, u32 *ld_coef)
+{
+ struct backplane_kr *bpkr = &lane->bpdev->bpkr;
+ u32 mainq = ld_coef[C_Z0];
+ u32 postq = ld_coef[C_P1];
+ u32 preq = ld_coef[C_M1];
+
+ /* Additional HW restrictions:
+ * 1. MIN_C(0) <= tx_preq + tx_mainq + tx_ratio_post1q <= MAX_C(0)
+ */
+ if ((preq + postq + mainq) < bpkr->min_kr.mainq)
+ return -ERANGE;
+ if ((preq + postq + mainq) > bpkr->max_kr.mainq)
+ return -ERANGE;
+
+ /* 2.
+ * ( tx_mainq + tx_preq + tx_ratio_post1q ) /
+ * ( tx_mainq - tx_preq - tx_ratio_post1q ) <
+ * coef_sum_ratio_numerator / coef_sum_ratio_denominator
+ */
+ if (((mainq + preq + postq) * COEF_SUM_RATIO_DENOMINATOR) >=
+ ((mainq - preq - postq) * COEF_SUM_RATIO_NUMERATOR))
+ return -ERANGE;
+
+ return 0;
+}
+
+enum serdes_type get_serdes_type(struct device_node *serdes_node)
+{
+ enum serdes_type serdes = SERDES_INVAL;
+ const char *serdes_comp;
+ int comp_no, i, ret;
+
+ comp_no = of_property_count_strings(serdes_node, "compatible");
+ for (i = 0; i < comp_no; i++) {
+ ret = of_property_read_string_index(serdes_node, "compatible",
+ i, &serdes_comp);
+ if (ret == 0) {
+ if (!strcasecmp(serdes_comp, "serdes-10g")) {
+ serdes = SERDES_10G;
+ break;
+ }
+ }
+ }
+
+ return serdes;
+}
+
+/* install QorIQ specific backplane callbacks:
+ * for AN start/decoding, hw specific defaults and lt validation
+ */
+static const struct backplane_ops qoriq_ops = {
+ .an_advertisement_init = an_advertisement_init,
+ .is_an_link_detected = is_an_link_detected,
+ .setup_default_settings = qoriq_setup_default_settings,
+ .lt_validation = lt_qoriq_validation,
+};
+
+/* qoriq_backplane_probe
+ *
+ * Probe function for QorIQ backplane driver to provide QorIQ device specific
+ * behavior
+ *
+ * phydev: backplane phy device
+ * this is an internal phy block controlled by the software
+ * which contains other component blocks like: PMA/PMD, PCS, AN
+ *
+ * Return: Zero for success or error code in case of failure
+ */
+static int qoriq_backplane_probe(struct phy_device *phydev)
+{
+ pr_info_once("%s: QorIQ Backplane driver version %s\n",
+ QORIQ_BACKPLANE_DRIVER_NAME,
+ QORIQ_BACKPLANE_DRIVER_VERSION);
+
+ /* call generic driver probe */
+ return backplane_probe(phydev);
+}
+
+/* qoriq_backplane_config_init
+ *
+ * Config_Init function for QorIQ devices to provide QorIQ specific behavior
+ *
+ * phydev: backplane phy device
+ *
+ * Return: Zero for success or error code in case of failure
+ */
+static int qoriq_backplane_config_init(struct phy_device *phydev)
+{
+ struct device_node *dev_node, *serdes_node, *lane_node;
+ const struct equalizer_device *qoriq_equalizer = NULL;
+ struct backplane_device *bpdev = phydev->priv;
+ struct qoriq_lane_ops *qoriq_lane_ops = NULL;
+ const struct qoriq_driver *qoriq_drv = NULL;
+ const struct lane_ops *lane_ops = NULL;
+ enum serdes_type serdes = SERDES_INVAL;
+ u32 eqparams[EQ_PARAMS_NO];
+ struct resource res;
+ int proplen;
+ int i, ret;
+
+ dev_node = phydev->mdio.dev.of_node;
+ if (!dev_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return -EINVAL;
+ }
+ if (!bpdev) {
+ phydev_err(phydev, "Backplane phy info is not allocated\n");
+ return -EINVAL;
+ }
+
+ if (!backplane_is_valid_mode(phydev->interface))
+ return -EINVAL;
+
+ /* call generic driver parse DT */
+ ret = backplane_parse_dt(phydev);
+ if (ret)
+ return ret;
+
+ bpdev->num_lanes = backplane_num_lanes(phydev->interface);
+ if (bpdev->num_lanes > MAX_KR_LANES_PER_PHY) {
+ phydev_err(phydev, "Unsupported number of lanes per phy: %d\n",
+ bpdev->num_lanes);
+ return -EINVAL;
+ }
+
+ proplen = of_property_count_u32_elems(dev_node, "lane-handle");
+ if (proplen < bpdev->num_lanes) {
+ phydev_err(phydev, "Unspecified lane handles\n");
+ return -EINVAL;
+ }
+ serdes_node = NULL;
+ for (i = 0; i < bpdev->num_lanes; i++) {
+ lane_node = of_parse_phandle(dev_node, "lane-handle", i);
+ if (!lane_node) {
+ phydev_err(phydev, "parse lane-handle failed\n");
+ return -EINVAL;
+ }
+ if (i == 0)
+ serdes_node = lane_node->parent;
+ ret = of_address_to_resource(lane_node, 0, &res);
+ if (ret) {
+ phydev_err(phydev,
+ "could not obtain lane memory map for index=%d, ret = %d\n",
+ i, ret);
+ return ret;
+ }
+ /* setup lane address */
+ bpdev->lane[i].lane_addr = res.start;
+
+ of_node_put(lane_node);
+ }
+ if (!serdes_node) {
+ phydev_err(phydev, "serdes node not found\n");
+ return -EINVAL;
+ }
+ bpdev->drv.is_little_endian = of_property_read_bool(serdes_node,
+ "little-endian");
+
+ ret = of_address_to_resource(serdes_node, 0, &res);
+ if (ret) {
+ phydev_err(phydev,
+ "could not obtain serdes memory map, ret = %d\n",
+ ret);
+ return ret;
+ }
+ bpdev->drv.base_addr = res.start;
+ bpdev->drv.memmap_size = res.end - res.start + 1;
+
+ serdes = get_serdes_type(serdes_node);
+ if (serdes == SERDES_INVAL) {
+ phydev_err(phydev, "Unknown serdes-type\n");
+ return 0;
+ }
+
+ /* if eq-params node exists then use the DTS specified values
+ * if eq-params node doesn't exist then use values already found in HW
+ * eq-params is a custom node and variable in size
+ */
+ proplen = of_property_count_u32_elems(dev_node, "eq-params");
+ if (proplen > 0) {
+ /* we use only 1 custom coefficient tap: amp_red */
+ if (proplen > EQ_PARAMS_NO)
+ proplen = EQ_PARAMS_NO;
+ ret = of_property_read_u32_array(dev_node, "eq-params",
+ (u32 *)eqparams, proplen);
+ if (ret == 0) {
+ bpdev->bpkr.valid_eq_params = true;
+ bpdev->bpkr.def_kr.dev_coef[IDX_AMP_RED] =
+ eqparams[IDX_AMP_RED];
+ }
+ }
+
+ /* call generic driver setup memio after reading serdes endianness */
+ ret = backplane_setup_memio(phydev);
+ if (ret)
+ return ret;
+
+ /* call generic driver setup mmd */
+ ret = backplane_setup_mmd(phydev);
+ if (ret)
+ return ret;
+
+ /* override default mdio setup and get qoriq specific info */
+ switch (serdes) {
+ case SERDES_10G:
+ lane_ops = qoriq_get_lane_ops_10g();
+ qoriq_drv = get_qoriq_driver_10g();
+ qoriq_setup_mem_io_10g(bpdev->drv.io);
+ qoriq_equalizer = qoriq_get_equalizer_10g();
+ break;
+ default:
+ phydev_err(phydev, "Serdes type not supported\n");
+ return -EINVAL;
+ }
+ if (!lane_ops) {
+ phydev_err(phydev, "Lane ops not available\n");
+ return -EINVAL;
+ }
+ if (!qoriq_drv) {
+ phydev_err(phydev, "Qoriq driver not available\n");
+ return -EINVAL;
+ }
+ if (!qoriq_equalizer) {
+ phydev_err(phydev, "Qoriq Equalizer not available\n");
+ return -EINVAL;
+ }
+
+ /* setup ops and equalizer */
+ bpdev->drv.lane_ops = lane_ops;
+ bpdev->drv.bp_ops = qoriq_ops;
+ bpdev->drv.priv = (void *)qoriq_drv;
+ bpdev->bpkr.equalizer = qoriq_equalizer;
+ qoriq_lane_ops = (struct qoriq_lane_ops *)lane_ops->priv;
+
+ if (!qoriq_lane_ops) {
+ phydev_err(phydev, "QorIQ lane ops not available\n");
+ return -EINVAL;
+ }
+
+ /* setup KR LT MMD registers space */
+ backplane_kr_lt_mmd_setup(&bpdev->bpkr, qoriq_drv->kr_lt_devad,
+ qoriq_drv->kr_lt_base);
+
+ if (backplane_is_mode_kr(phydev->interface)) {
+ /* setup kr coefficients limits */
+ bpdev->bpkr.min_kr.preq = PRE_COEF_MIN;
+ bpdev->bpkr.max_kr.preq = PRE_COEF_MAX;
+ bpdev->bpkr.min_kr.mainq = ZERO_COEF_MIN;
+ bpdev->bpkr.max_kr.mainq = ZERO_COEF_MAX;
+ bpdev->bpkr.min_kr.postq = POST_COEF_MIN;
+ bpdev->bpkr.max_kr.postq = POST_COEF_MAX;
+ }
+
+ /* call generic driver setup lanes */
+ ret = backplane_setup_lanes(phydev);
+ if (ret)
+ return ret;
+
+ /* call generic driver initialize
+ * start the lane timers used to run the algorithm
+ */
+ ret = backplane_initialize(phydev);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+static int qoriq_backplane_match_phy_device(struct phy_device *phydev)
+{
+ struct device_node *dev_node, *serdes_node, *lane_node;
+ enum serdes_type serdes = SERDES_INVAL;
+ int i;
+
+ if (!phydev->mdio.dev.of_node)
+ return 0;
+
+ if (!phydev->is_c45)
+ return 0;
+
+ dev_node = phydev->mdio.dev.of_node;
+ if (!dev_node) {
+ phydev_err(phydev, "No associated device tree node\n");
+ return 0;
+ }
+
+ /* Get Master lane node */
+ lane_node = of_parse_phandle(dev_node, "lane-handle", 0);
+ if (!lane_node)
+ return 0;
+ serdes_node = lane_node->parent;
+ of_node_put(lane_node);
+ if (!serdes_node)
+ return 0;
+
+ serdes = get_serdes_type(serdes_node);
+
+ switch (serdes) {
+ case SERDES_10G:
+ /* On LS devices we must find the c45 device with correct PHY ID
+ * Implementation similar with the one existent in phy_device:
+ * @function: phy_bus_match
+ */
+ for (i = 1; i < ARRAY_SIZE(phydev->c45_ids.device_ids); i++) {
+ if (!(phydev->c45_ids.devices_in_package & (1 << i)))
+ continue;
+
+ if ((PCS_PHY_DEVICE_ID & PCS_PHY_DEVICE_ID_MASK) ==
+ (phydev->c45_ids.device_ids[i] &
+ PCS_PHY_DEVICE_ID_MASK))
+ return 1;
+ }
+ break;
+ default:
+ phydev_err(phydev, "Unknown serdes-type\n");
+ return 0;
+ }
+ return 0;
+}
+
+static struct phy_driver qoriq_backplane_driver[] = {
+ {
+ .phy_id = PCS_PHY_DEVICE_ID,
+ .name = QORIQ_BACKPLANE_DRIVER_NAME,
+ .phy_id_mask = PCS_PHY_DEVICE_ID_MASK,
+ .features = BACKPLANE_FEATURES,
+ .probe = qoriq_backplane_probe,
+ .remove = backplane_remove,
+ .config_init = qoriq_backplane_config_init,
+ .aneg_done = backplane_aneg_done,
+ .config_aneg = backplane_config_aneg,
+ .read_status = backplane_read_status,
+ .suspend = backplane_suspend,
+ .resume = backplane_resume,
+ .match_phy_device = qoriq_backplane_match_phy_device,
+ },
+};
+
+module_phy_driver(qoriq_backplane_driver);
+
+static struct mdio_device_id __maybe_unused qoriq_backplane_tbl[] = {
+ { PCS_PHY_DEVICE_ID, PCS_PHY_DEVICE_ID_MASK },
+ { }
+};
+
+MODULE_DEVICE_TABLE(mdio, qoriq_backplane_tbl);
+
+MODULE_DESCRIPTION("QorIQ Backplane driver");
+MODULE_AUTHOR("Florinel Iordache <florinel.iordache@nxp.com>");
+MODULE_LICENSE("Dual BSD/GPL");
diff --git a/drivers/net/phy/backplane/qoriq_backplane.h b/drivers/net/phy/backplane/qoriq_backplane.h
new file mode 100644
index 0000000..09a3600
--- /dev/null
+++ b/drivers/net/phy/backplane/qoriq_backplane.h
@@ -0,0 +1,42 @@
+/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */
+/* QorIQ Backplane driver
+ *
+ * Copyright 2018-2020 NXP
+ */
+
+#ifndef __QORIQ_BACKPLANE_H
+#define __QORIQ_BACKPLANE_H
+
+#include "backplane.h"
+
+/* Bins thresholds */
+#define QORIQ_BIN_M1_THRESHOLD 3
+#define QORIQ_BIN_LONG_THRESHOLD 2
+
+/* Index of custom parameter: AMP_RED (amplitude reduction) */
+#define IDX_AMP_RED 0
+
+struct qoriq_lane_ops {
+ u32 (*read_tecr0)(void __iomem *reg);
+ u32 (*read_tecr1)(void __iomem *reg);
+ u32 (*read_amp_red)(void __iomem *reg);
+};
+
+struct qoriq_driver {
+ /* KR LT MMD registers */
+ int kr_lt_devad;
+ u32 kr_lt_base;
+ /* KR AN MMD registers */
+ u32 an_adv1;
+ u32 an_bp_eth_status;
+};
+
+const struct lane_ops *qoriq_get_lane_ops_10g(void);
+
+const struct equalizer_device *qoriq_get_equalizer_10g(void);
+
+const struct qoriq_driver *get_qoriq_driver_10g(void);
+
+void qoriq_setup_mem_io_10g(struct mem_io memio);
+
+#endif /* __QORIQ_BACKPLANE_H */
diff --git a/drivers/net/phy/backplane/qoriq_serdes_10g.c b/drivers/net/phy/backplane/qoriq_serdes_10g.c
new file mode 100644
index 0000000..e4e5991
--- /dev/null
+++ b/drivers/net/phy/backplane/qoriq_serdes_10g.c
@@ -0,0 +1,486 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+/* QorIQ Backplane driver for SerDes 10G
+ *
+ * Copyright 2018-2020 NXP
+ */
+
+#include <linux/io.h>
+#include <linux/delay.h>
+
+#include "qoriq_backplane.h"
+
+#define EQUALIZER_NAME "qoriq_serdes_10g"
+#define EQUALIZER_VERSION "1.0.0"
+
+#define BIN_1_SEL 0x00000000
+#define BIN_2_SEL 0x00010000
+#define BIN_3_SEL 0x00020000
+#define BIN_OFFSET_SEL 0x00030000
+#define BIN_BLW_SEL 0x00040000
+#define BIN_AVG_SEL 0x00050000
+#define BIN_M1_SEL 0x00060000
+#define BIN_LONG_SEL 0x00070000
+#define CDR_SEL_MASK 0x00070000
+
+#define RATIO_PREQ_SHIFT 22
+#define RATIO_PST1Q_SHIFT 16
+#define ADPT_EQ_SHIFT 8
+#define AMP_RED_SHIFT 0
+
+#define RATIO_PREQ_MASK 0x03c00000
+#define RATIO_PST1Q_MASK 0x001f0000
+#define ADPT_EQ_MASK 0x00003f00
+#define AMP_RED_MASK 0x0000003f
+
+#define TECR0_INIT 0x24200000
+
+#define GCR0_RESET_MASK 0x00600000
+#define GCR0_TRST_MASK 0x00200000
+#define GCR0_RRST_MASK 0x00400000
+
+#define GCR1_SNP_START_MASK 0x00000040
+#define GCR1_CTL_SNP_START_MASK 0x00002000
+
+#define RECR1_CTL_SNP_DONE_MASK 0x00000002
+#define RECR1_SNP_DONE_MASK 0x00000004
+#define TCSR1_SNP_DATA_MASK 0x00007fc0
+#define TCSR1_SNP_DATA_SHIFT 6
+#define TCSR1_EQ_SNPBIN_SIGN_MASK 0x100
+
+#define TCSR3_CDR_LCK_MASK 0x08000000
+
+#define RECR1_GAINK2_MASK 0x0f000000
+#define RECR1_GAINK2_SHIFT 24
+
+#define RECR1_GAINK3_MASK 0x000f0000
+#define RECR1_GAINK3_SHIFT 16
+
+#define RECR1_EQ_OFFSET_MASK 0x00001f80
+#define RECR1_EQ_OFFSET_SHIFT 7
+
+/* AN advertisement register 7.17 */
+#define AN_AD_ABILITY_1 0x11
+
+/* Backplane Ethernet status (Register 7.48) */
+#define AN_BP_ETH_STATUS_OFFSET 0x30
+
+/* KR PMD control register (Register 1.150) */
+#define KR_PMD_BASE_OFFSET 0x96
+
+/* Bin snapshots thresholds range */
+#define EQ_BIN_MIN -256
+#define EQ_BIN_MAX 255
+/* Bin snapshots average thresholds range */
+#define EQ_BIN_SNP_AV_THR_LOW -150
+#define EQ_BIN_SNP_AV_THR_HIGH 150
+
+#define EQ_GAINK_MIN 0xF
+#define EQ_GAINK_MAX 0x0
+#define EQ_GAINK_MIDRANGE_LOW 0xE
+#define EQ_GAINK_MIDRANGE_HIGH 0x1
+
+#define EQ_OFFSET_MIN 0
+#define EQ_OFFSET_MAX 0x3F
+#define EQ_OFFSET_MIDRANGE_LOW 0x10
+#define EQ_OFFSET_MIDRANGE_HIGH 0x2F
+
+#define MEMORY_MAP_SIZE 0x40
+
+struct qoriq_lane_regs {
+ u32 gcr0; /* 0x00: General Control Register 0 */
+ u32 gcr1; /* 0x04: General Control Register 1 */
+ u32 gcr2; /* 0x08: General Control Register 2 */
+ u32 res_0c; /* 0x0C: Reserved */
+ u32 recr0; /* 0x10: Receive Equalization Control Register 0 */
+ u32 recr1; /* 0x14: Receive Equalization Control Register 1 */
+ u32 tecr0; /* 0x18: Transmit Equalization Control Register 0 */
+ u32 res_1c; /* 0x1C: Reserved */
+ u32 tlcr0; /* 0x20: TTL Control Register 0 */
+ u32 tlcr1; /* 0x24: TTL Control Register 1 */
+ u32 tlcr2; /* 0x28: TTL Control Register 2 */
+ u32 tlcr3; /* 0x2C: TTL Control Register 3 */
+ u32 tcsr0; /* 0x30: Test Control/Status Register 0 */
+ u32 tcsr1; /* 0x34: Test Control/Status Register 1 */
+ u32 tcsr2; /* 0x38: Test Control/Status Register 2 */
+ u32 tcsr3; /* 0x3C: Test Control/Status Register 3 */
+};
+
+static struct mem_io io;
+
+static void reset_lane(void __iomem *reg, enum lane_req ln_req)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+
+ /* reset Tx lane: send reset request */
+ if (ln_req | LANE_TX) {
+ io.write32(io.read32(®_base->gcr0) & ~GCR0_TRST_MASK,
+ ®_base->gcr0);
+ }
+ /* reset Rx lane: send reset request */
+ if (ln_req | LANE_RX) {
+ io.write32(io.read32(®_base->gcr0) & ~GCR0_RRST_MASK,
+ ®_base->gcr0);
+ }
+ /* unreset the lane */
+ if (ln_req != LANE_INVALID) {
+ udelay(1);
+ io.write32(io.read32(®_base->gcr0) | GCR0_RESET_MASK,
+ ®_base->gcr0);
+ udelay(1);
+ }
+}
+
+static u32 read_tecr0(void __iomem *reg)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+
+ return io.read32(®_base->tecr0);
+}
+
+static u32 read_tecr1(void __iomem *reg)
+{
+ return 0;
+}
+
+static u32 read_amp_red(void __iomem *reg)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+ u32 val, amp_red;
+
+ val = io.read32(®_base->tecr0);
+ amp_red = (val & AMP_RED_MASK) >> AMP_RED_SHIFT;
+
+ return amp_red;
+}
+
+static void read_kr_coef(void __iomem *reg, struct kr_coef *coef)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+ u32 val;
+
+ val = io.read32(®_base->tecr0);
+
+ coef->preq = (val & RATIO_PREQ_MASK) >> RATIO_PREQ_SHIFT;
+ coef->postq = (val & RATIO_PST1Q_MASK) >> RATIO_PST1Q_SHIFT;
+ coef->mainq = (val & ADPT_EQ_MASK) >> ADPT_EQ_SHIFT;
+ coef->dev_coef[IDX_AMP_RED] = (val & AMP_RED_MASK) >> AMP_RED_SHIFT;
+}
+
+static void tune_tecr(void __iomem *reg, struct kr_coef *coef, bool reset)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+ u32 val;
+
+ val = TECR0_INIT |
+ coef->mainq << ADPT_EQ_SHIFT |
+ coef->preq << RATIO_PREQ_SHIFT |
+ coef->postq << RATIO_PST1Q_SHIFT |
+ coef->dev_coef[IDX_AMP_RED] << AMP_RED_SHIFT;
+
+ if (reset) {
+ /* reset the lane */
+ io.write32(io.read32(®_base->gcr0) &
+ ~GCR0_RESET_MASK, ®_base->gcr0);
+ udelay(1);
+ }
+
+ io.write32(val, ®_base->tecr0);
+ udelay(1);
+
+ if (reset) {
+ /* unreset the lane */
+ io.write32(io.read32(®_base->gcr0) | GCR0_RESET_MASK,
+ ®_base->gcr0);
+ udelay(1);
+ }
+}
+
+/* collect_gains
+ *
+ * reg: serdes registers memory map
+ * gaink2: High-frequency gain of the equalizer amplifier
+ * the high-frequency gain of the equalizer amplifier is increased by
+ * decrementing the value of eq_gaink2 by one
+ * gaink3: Middle-frequency gain of the equalizer amplifier
+ * the mid-frequency gain of the equalizer amplifier is increased by
+ * decrementing the value of eq_gaink3 by one
+ * osestat: equalization offset status
+ * the equalizer offset is reduced by decrementing the value of osestat
+ * size: size of snapshots data collection
+ */
+static int collect_gains(void __iomem *reg, s16 *gaink2, s16 *gaink3,
+ s16 *osestat, u8 size)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+ u32 rx_eq_snp;
+ int timeout;
+ int i;
+
+ for (i = 0; i < size; i++) {
+ /* wait RECR1_CTL_SNP_DONE_MASK has cleared */
+ timeout = 100;
+ while (io.read32(®_base->recr1) &
+ RECR1_CTL_SNP_DONE_MASK) {
+ udelay(1);
+ timeout--;
+ if (timeout == 0)
+ break;
+ }
+
+ /* start snapshot */
+ io.write32((io.read32(®_base->gcr1) |
+ GCR1_CTL_SNP_START_MASK), ®_base->gcr1);
+
+ /* wait for SNP done */
+ timeout = 100;
+ while (!(io.read32(®_base->recr1) &
+ RECR1_CTL_SNP_DONE_MASK)) {
+ udelay(1);
+ timeout--;
+ if (timeout == 0)
+ break;
+ }
+
+ /* read and save the snapshot */
+ rx_eq_snp = io.read32(®_base->recr1);
+
+ if (gaink2)
+ gaink2[i] = (u8)((rx_eq_snp & RECR1_GAINK2_MASK) >>
+ RECR1_GAINK2_SHIFT);
+ if (gaink3)
+ gaink3[i] = (u8)((rx_eq_snp & RECR1_GAINK3_MASK) >>
+ RECR1_GAINK3_SHIFT);
+ if (osestat)
+ osestat[i] = (u8)((rx_eq_snp & RECR1_EQ_OFFSET_MASK) >>
+ RECR1_EQ_OFFSET_SHIFT);
+
+ /* terminate the snapshot by setting GCR1[REQ_CTL_SNP] */
+ io.write32((io.read32(®_base->gcr1) &
+ ~GCR1_CTL_SNP_START_MASK), ®_base->gcr1);
+ }
+ return i;
+}
+
+static int collect_eq_status(void __iomem *reg, enum eqc_type type[],
+ u8 type_no, s16 *counters, u8 size)
+{
+ s16 *gaink2 = NULL, *gaink3 = NULL, *osestat = NULL;
+ u8 i;
+
+ for (i = 0; i < type_no; i++) {
+ switch (type[i]) {
+ case EQC_GAIN_HF:
+ gaink2 = counters;
+ break;
+ case EQC_GAIN_MF:
+ gaink3 = counters + size;
+ break;
+ case EQC_EQOFFSET:
+ osestat = counters + 2 * size;
+ break;
+ default:
+ /* invalid type */
+ break;
+ }
+ }
+
+ return collect_gains(reg, gaink2, gaink3, osestat, size);
+}
+
+static int collect_bin_snapshots(void __iomem *reg, enum eqc_type type,
+ s16 *bin_counters, u8 bin_size)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+ int bin_snapshot;
+ int i, timeout;
+ u32 bin_sel;
+
+ /* calculate TCSR1[CDR_SEL] */
+ switch (type) {
+ case EQC_BIN_1:
+ bin_sel = BIN_1_SEL;
+ break;
+ case EQC_BIN_2:
+ bin_sel = BIN_2_SEL;
+ break;
+ case EQC_BIN_3:
+ bin_sel = BIN_3_SEL;
+ break;
+ case EQC_BIN_LONG:
+ bin_sel = BIN_LONG_SEL;
+ break;
+ case EQC_BIN_M1:
+ bin_sel = BIN_M1_SEL;
+ break;
+ case EQC_BIN_OFFSET:
+ bin_sel = BIN_OFFSET_SEL;
+ break;
+ case EQC_BIN_AVG:
+ bin_sel = BIN_AVG_SEL;
+ break;
+ case EQC_BIN_BLW:
+ bin_sel = BIN_BLW_SEL;
+ break;
+ default:
+ /* invalid bin type */
+ return 0;
+ }
+
+ for (i = 0; i < bin_size; i++) {
+ /* wait RECR1_SNP_DONE_MASK has cleared */
+ timeout = 100;
+ while ((io.read32(®_base->recr1) &
+ RECR1_SNP_DONE_MASK)) {
+ udelay(1);
+ timeout--;
+ if (timeout == 0)
+ break;
+ }
+
+ /* set TCSR1[CDR_SEL] */
+ io.write32((io.read32(®_base->tcsr1) &
+ ~CDR_SEL_MASK) | bin_sel, ®_base->tcsr1);
+
+ /* start snapshot */
+ io.write32(io.read32(®_base->gcr1) |
+ GCR1_SNP_START_MASK, ®_base->gcr1);
+
+ /* wait for SNP done */
+ timeout = 100;
+ while (!(io.read32(®_base->recr1) &
+ RECR1_SNP_DONE_MASK)) {
+ udelay(1);
+ timeout--;
+ if (timeout == 0)
+ break;
+ }
+
+ /* read and save the snapshot:
+ * 2's complement 9 bit long value (-256 to 255)
+ */
+ bin_snapshot = (io.read32(®_base->tcsr1) &
+ TCSR1_SNP_DATA_MASK) >> TCSR1_SNP_DATA_SHIFT;
+ if (bin_snapshot & TCSR1_EQ_SNPBIN_SIGN_MASK) {
+ /* 2's complement 9 bit long negative number */
+ bin_snapshot &= ~TCSR1_EQ_SNPBIN_SIGN_MASK;
+ bin_snapshot -= 256;
+ }
+
+ /* save collected Bin snapshot */
+ bin_counters[i] = (s16)bin_snapshot;
+
+ /* terminate the snapshot by setting GCR1[REQ_CTL_SNP] */
+ io.write32(io.read32(®_base->gcr1) &
+ ~GCR1_SNP_START_MASK, ®_base->gcr1);
+ }
+ return i;
+}
+
+static struct eqc_range bin_range = {
+ .min = EQ_BIN_MIN,
+ .max = EQ_BIN_MAX,
+ .mid_low = EQ_BIN_SNP_AV_THR_LOW,
+ .mid_high = EQ_BIN_SNP_AV_THR_HIGH,
+};
+
+static struct eqc_range gaink_range = {
+ .min = EQ_GAINK_MIN,
+ .max = EQ_GAINK_MAX,
+ .mid_low = EQ_GAINK_MIDRANGE_LOW,
+ .mid_high = EQ_GAINK_MIDRANGE_HIGH,
+};
+
+static struct eqc_range osestat_range = {
+ .min = EQ_OFFSET_MIN,
+ .max = EQ_OFFSET_MAX,
+ .mid_low = EQ_OFFSET_MIDRANGE_LOW,
+ .mid_high = EQ_OFFSET_MIDRANGE_HIGH,
+};
+
+static struct eqc_range *get_counter_range(enum eqc_type type)
+{
+ switch (type) {
+ case EQC_BIN_1:
+ case EQC_BIN_2:
+ case EQC_BIN_3:
+ case EQC_BIN_LONG:
+ case EQC_BIN_M1:
+ case EQC_BIN_OFFSET:
+ case EQC_BIN_AVG:
+ case EQC_BIN_BLW:
+ return &bin_range;
+ case EQC_GAIN_HF:
+ case EQC_GAIN_MF:
+ return &gaink_range;
+ case EQC_EQOFFSET:
+ return &osestat_range;
+ default:
+ /* invalid counter type */
+ return NULL;
+ }
+ return NULL;
+}
+
+static bool is_cdr_lock_bit(void __iomem *reg)
+{
+ struct qoriq_lane_regs __iomem *reg_base = reg;
+
+ if (io.read32(®_base->tcsr3) & TCSR3_CDR_LCK_MASK)
+ return true;
+
+ return false;
+}
+
+static const struct qoriq_lane_ops qoriq_lane_ops_10g = {
+ .read_tecr0 = read_tecr0,
+ .read_tecr1 = read_tecr1,
+ .read_amp_red = read_amp_red,
+};
+
+static const struct lane_ops lane_ops_10g = {
+ .priv = &qoriq_lane_ops_10g,
+ .memmap_size = MEMORY_MAP_SIZE,
+ .reset_lane = reset_lane,
+ .tune_lane_kr = tune_tecr,
+ .read_lane_kr = read_kr_coef,
+ .is_cdr_lock = is_cdr_lock_bit,
+};
+
+const struct lane_ops *qoriq_get_lane_ops_10g(void)
+{
+ return &lane_ops_10g;
+}
+
+static const struct equalizer_device qoriq_equalizer = {
+ .name = EQUALIZER_NAME,
+ .version = EQUALIZER_VERSION,
+ .ops = {
+ .collect_counters = collect_bin_snapshots,
+ .collect_multiple_counters = collect_eq_status,
+ .get_counter_range = get_counter_range,
+ },
+};
+
+const struct equalizer_device *qoriq_get_equalizer_10g(void)
+{
+ return &qoriq_equalizer;
+}
+
+static const struct qoriq_driver qoriq_drv = {
+ /* KR PMD registers */
+ .kr_lt_devad = MDIO_MMD_PMAPMD,
+ .kr_lt_base = KR_PMD_BASE_OFFSET,
+ /* KR AN registers: IEEE802.3 Clause 45 MMD 7 */
+ .an_adv1 = AN_AD_ABILITY_1,
+ .an_bp_eth_status = AN_BP_ETH_STATUS_OFFSET,
+};
+
+const struct qoriq_driver *get_qoriq_driver_10g(void)
+{
+ return &qoriq_drv;
+}
+
+void qoriq_setup_mem_io_10g(struct mem_io memio)
+{
+ io = memio;
+}
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH net-next v3 6/7] net: phy: add bee algorithm for kr training
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
` (4 preceding siblings ...)
2020-06-22 13:35 ` [PATCH net-next v3 5/7] net: phy: enable qoriq backplane support Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-22 13:35 ` [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio description Florinel Iordache
6 siblings, 0 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Add support for bee equalization algorithm used by kr training:
3-Taps Bit Edge Equalization (BEE) algorithm
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
drivers/net/phy/backplane/Kconfig | 11 +
drivers/net/phy/backplane/Makefile | 1 +
drivers/net/phy/backplane/eq_bee.c | 1076 ++++++++++++++++++++++++++++++++++++
3 files changed, 1088 insertions(+)
create mode 100644 drivers/net/phy/backplane/eq_bee.c
diff --git a/drivers/net/phy/backplane/Kconfig b/drivers/net/phy/backplane/Kconfig
index 3e20a78..ee5cf1c 100644
--- a/drivers/net/phy/backplane/Kconfig
+++ b/drivers/net/phy/backplane/Kconfig
@@ -19,6 +19,17 @@ config ETH_BACKPLANE_FIXED
No Equalization algorithm is used to adapt the initial coefficients
initially set by the user.
+config ETH_BACKPLANE_BEE
+ tristate "3-Taps Bit Edge Equalization (BEE) algorithm"
+ depends on ETH_BACKPLANE
+ help
+ This module provides a driver for BEE algorithm: 3-Taps
+ Bit Edge Equalization. This algorithm is using a method
+ based on 3-taps coefficients for mitigating intersymbol
+ interference (ISI) in high-speed backplane applications.
+ The initial values for algorithm coefficient values are
+ user configurable and used as a starting point of the algorithm.
+
config ETH_BACKPLANE_QORIQ
tristate "QorIQ Ethernet Backplane driver"
depends on ETH_BACKPLANE
diff --git a/drivers/net/phy/backplane/Makefile b/drivers/net/phy/backplane/Makefile
index 3ae3d8b..30a2ebb 100644
--- a/drivers/net/phy/backplane/Makefile
+++ b/drivers/net/phy/backplane/Makefile
@@ -5,6 +5,7 @@
obj-$(CONFIG_ETH_BACKPLANE) += eth_backplane.o
obj-$(CONFIG_ETH_BACKPLANE_FIXED) += eq_fixed.o
+obj-$(CONFIG_ETH_BACKPLANE_BEE) += eq_bee.o
obj-$(CONFIG_ETH_BACKPLANE_QORIQ) += eth_backplane_qoriq.o
eth_backplane-objs := backplane.o link_training.o
diff --git a/drivers/net/phy/backplane/eq_bee.c b/drivers/net/phy/backplane/eq_bee.c
new file mode 100644
index 0000000..045ed7e
--- /dev/null
+++ b/drivers/net/phy/backplane/eq_bee.c
@@ -0,0 +1,1076 @@
+// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
+/* 3-Taps Bit Edge Equalization (BEE) algorithm
+ *
+ * Copyright 2019-2020 NXP
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+#include "equalization.h"
+
+#define ALGORITHM_NAME "backplane_bee_3tap"
+#define ALGORITHM_DESCR "3-Taps Bit Edge Equalization"
+#define ALGORITHM_VERSION "1.5.5"
+
+/* BEE algorithm timeouts */
+#define TIMEOUT_LONG 3
+#define TIMEOUT_M1 3
+
+/* Size of equalization snapshots data collection */
+#define EQ_SNAPSHOTS_SIZE 10
+
+/* Rx link quality conditions:
+ * The following macros are used to determine the Rx link quality
+ * which is used to decide if/when to proceed with BinLong/BinM1 modules.
+ * The code that considers Rx in good quality is always in place:
+ * Rx link is in 'good quality' if:
+ * Bin1, Bin2 and Bin3 are toggling
+ *
+ * These macros are used to enable less quality link conditions.
+ * If Rx link quality is considered good enough then proceed to BinLong/BinM1
+ */
+
+/* Rx is 'less quality' if:
+ * Bin1 is toggling
+ * AND
+ * Bin2 is Early, GainMF stuck at max_eq_gain and Bin3 is Late
+ * OR
+ * Bin2 is Late, GainMF stuck at min_eq_gain and Bin3 is Early
+ */
+#define ENABLE_LESS_QUALITY_CONDITION
+
+/* Rx is 'even less quality' if:
+ * Bin1 is Early AND GainHF stuck at max_eq_gain and Bin2 is Late AND
+ * GainMF stuck at min_eq_gain
+ * OR
+ * Bin1 is Late AND GainHF stuck at min_eq_gain AND
+ * Bin2 is Early, GainMF stuck at max_eq_gain
+ */
+#define ENABLE_EVEN_LESS_QUALITY_CONDITION
+
+/* Rx is 'seemingly quality' if:
+ * Bin1 is always Late for all snapshots AND
+ * GainHF is stuck at min_eq_gain
+ * AND
+ * Bin2 and Bin3 are both Toggling
+ */
+#define ENABLE_SEEMINGLY_QUALITY_CONDITION
+
+enum bin_state {
+ BIN_INVALID,
+ BIN_EARLY,
+ BIN_TOGGLE,
+ BIN_LATE
+};
+
+struct eq_data_priv {
+ /* Equalization Algorithm setup data */
+ struct equalizer_driver eqdrv;
+
+ /* Bin state */
+ enum bin_state bin_m1_state;
+ enum bin_state bin_long_state;
+ enum bin_state prev_bin_m1_state;
+ enum bin_state prev_bin_long_state;
+
+ /* Bin training status */
+ bool bin_m1_stop;
+ bool bin_long_stop;
+ int m1_min_max_cnt;
+ int long_min_max_cnt;
+
+ /* Algorithm controlled value */
+ u32 ld_update;
+
+ /* Bit edge statistics: Bin snapshots */
+ s16 bin1_snapshot[EQ_SNAPSHOTS_SIZE];
+ s16 bin2_snapshot[EQ_SNAPSHOTS_SIZE];
+ s16 bin3_snapshot[EQ_SNAPSHOTS_SIZE];
+ s16 bin_long_snapshot[EQ_SNAPSHOTS_SIZE];
+ s16 bin_m1_snapshot[EQ_SNAPSHOTS_SIZE];
+ s16 bin_offset_snapshot[EQ_SNAPSHOTS_SIZE];
+
+ /* Gain snapshots */
+ u8 gain_hf_snapshot[EQ_SNAPSHOTS_SIZE];
+ u8 gain_mf_snapshot[EQ_SNAPSHOTS_SIZE];
+
+ /* Offset status snapshot */
+ u8 osestat_snapshot[EQ_SNAPSHOTS_SIZE];
+};
+
+static enum bin_state get_bin_snapshots_state(struct eq_data_priv *priv,
+ enum eqc_type type)
+{
+ s16 bin_snp_av_thr_low, bin_snp_av_thr_high;
+ s16 snapshot_average, snapshot_sum = 0;
+ struct eqc_range *bin_range;
+ s16 *bin_snapshot;
+ int i;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return BIN_INVALID;
+ }
+
+ switch (type) {
+ case EQC_BIN_1:
+ bin_snapshot = priv->bin1_snapshot;
+ break;
+ case EQC_BIN_2:
+ bin_snapshot = priv->bin2_snapshot;
+ break;
+ case EQC_BIN_3:
+ bin_snapshot = priv->bin3_snapshot;
+ break;
+ case EQC_BIN_LONG:
+ bin_snapshot = priv->bin_long_snapshot;
+ break;
+ case EQC_BIN_OFFSET:
+ bin_snapshot = priv->bin_offset_snapshot;
+ break;
+ case EQC_BIN_M1:
+ bin_snapshot = priv->bin_m1_snapshot;
+ break;
+ default:
+ /* invalid bin type */
+ return BIN_INVALID;
+ }
+ if (!bin_snapshot)
+ return BIN_INVALID;
+
+ bin_range = priv->eqdrv.equalizer->ops.get_counter_range(type);
+ if (!bin_range)
+ return BIN_INVALID;
+
+ bin_snp_av_thr_low = bin_range->mid_low;
+ bin_snp_av_thr_high = bin_range->mid_high;
+
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++)
+ snapshot_sum += bin_snapshot[i];
+
+ snapshot_average = (s16)(snapshot_sum / EQ_SNAPSHOTS_SIZE);
+
+ if (snapshot_average >= -256 && snapshot_average < bin_snp_av_thr_low)
+ return BIN_EARLY;
+ else if (snapshot_average >= bin_snp_av_thr_low &&
+ snapshot_average < bin_snp_av_thr_high)
+ return BIN_TOGGLE;
+ else if (snapshot_average >= bin_snp_av_thr_high &&
+ snapshot_average <= 255)
+ return BIN_LATE;
+
+ return BIN_INVALID;
+}
+
+static u32 process_bin_m1_toggle(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum req_type prev_req_cm1;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ prev_req_cm1 = lt_decode_coef_update(priv->ld_update, C_M1);
+
+ /* Toggle path */
+ if (priv->prev_bin_m1_state == priv->bin_m1_state) {
+ /* Hold C- */
+ update = lt_encode_startup_request(REQ_HOLD);
+ } else {
+ update = lt_encode_startup_request(REQ_HOLD);
+ /* If previous step moved C- repeat C- move */
+ if (prev_req_cm1 == REQ_INC ||
+ prev_req_cm1 == REQ_DEC)
+ update = lt_encode_request(update, prev_req_cm1, C_M1);
+ }
+
+ return update;
+}
+
+static u32 process_bin_m1_prev_toggle(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum req_type prev_req_cm1;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ prev_req_cm1 = lt_decode_coef_update(priv->ld_update, C_M1);
+
+ update = lt_encode_startup_request(REQ_HOLD);
+ /* If previous step moved C- go back on C- */
+ if (prev_req_cm1 == REQ_INC)
+ update = lt_encode_request(update, REQ_DEC, C_M1);
+ if (prev_req_cm1 == REQ_DEC)
+ update = lt_encode_request(update, REQ_INC, C_M1);
+
+ return update;
+}
+
+static u32 process_bin_m1_early(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum coef_status lpst_cm1;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ /* Get LP coefficient status to determine
+ * if coefficient is in range or reached the limit thresholds
+ * IF the coefficient is at MIN/MAX and still want to INC/DEC
+ * THEN do we are done with this module
+ */
+ lpst_cm1 = lt_get_lp_coef_status(priv->eqdrv.lane, C_M1);
+
+ /* Early path */
+ if (lpst_cm1 == COEF_MAX) {
+ /* Hold C- */
+ update = lt_encode_startup_request(REQ_HOLD);
+ } else {
+ /* request Increment C- */
+ update = lt_encode_request(update, REQ_INC, C_M1);
+ }
+
+ return update;
+}
+
+static u32 process_bin_m1_late(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum coef_status lpst_cm1;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ /* Get LP coefficient status to determine
+ * if coefficient is in range or reached the limit thresholds
+ * IF the coefficient is at MIN/MAX and still want to INC/DEC
+ * THEN do we are done with this module
+ */
+ lpst_cm1 = lt_get_lp_coef_status(priv->eqdrv.lane, C_M1);
+
+ /* Late path */
+ if (lpst_cm1 == COEF_MIN) {
+ /* Hold C- */
+ update = lt_encode_startup_request(REQ_HOLD);
+ } else {
+ /* request Decrement C- */
+ update = lt_encode_request(update, REQ_DEC, C_M1);
+ }
+
+ return update;
+}
+
+static u32 process_bin_m1_antipodal(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ if (priv->bin_m1_state == BIN_LATE) {
+ /* request Decrement C- */
+ update = lt_encode_request(update, REQ_DEC, C_M1);
+ } else {
+ /* Hold C- */
+ update = lt_encode_startup_request(REQ_HOLD);
+ }
+
+ return update;
+}
+
+/* process_bin_m1
+ *
+ * Bin_M1:
+ * contains the scoring of initial edges on pulses that are 1UI long
+ * following non-single bits
+ * used to adjust LP coefficient: C_M1
+ */
+static void process_bin_m1(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return;
+ }
+
+ if (priv->bin_m1_state == BIN_INVALID) {
+ phydev_err(priv->eqdrv.phydev, "Invalid Bin_M1 state\n");
+ return;
+ }
+
+ if (priv->bin_m1_state == BIN_TOGGLE) {
+ update = process_bin_m1_toggle(priv);
+ } else {
+ if (priv->prev_bin_m1_state == BIN_TOGGLE) {
+ update = process_bin_m1_prev_toggle(priv);
+ } else {
+ if (priv->prev_bin_m1_state == priv->bin_m1_state) {
+ if (priv->bin_m1_state == BIN_LATE)
+ update = process_bin_m1_late(priv);
+ else
+ update = process_bin_m1_early(priv);
+ } else {
+ update = process_bin_m1_antipodal(priv);
+ }
+ }
+ }
+
+ /* Store current algorithm decision
+ * as previous algorithm ld_update for next step
+ */
+ priv->ld_update = update;
+}
+
+static u32 process_bin_long_toggle(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum req_type prev_req_cp1, prev_req_cz0;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ prev_req_cp1 = lt_decode_coef_update(priv->ld_update, C_P1);
+ prev_req_cz0 = lt_decode_coef_update(priv->ld_update, C_Z0);
+
+ /* Toggle path */
+ if (priv->prev_bin_long_state == priv->bin_long_state) {
+ /* Hold C+ and C0 */
+ update = lt_encode_startup_request(REQ_HOLD);
+ } else {
+ update = lt_encode_startup_request(REQ_HOLD);
+ /* If previous step moved C+/C0 repeat C+/C0 move */
+ if (prev_req_cp1 == REQ_INC ||
+ prev_req_cp1 == REQ_DEC ||
+ prev_req_cz0 == REQ_INC ||
+ prev_req_cz0 == REQ_DEC) {
+ update = lt_encode_request(update, prev_req_cp1, C_P1);
+ update = lt_encode_request(update, prev_req_cz0, C_Z0);
+ }
+ }
+
+ return update;
+}
+
+static u32 process_bin_long_prev_toggle(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum req_type prev_req_cp1, prev_req_cz0;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ prev_req_cp1 = lt_decode_coef_update(priv->ld_update, C_P1);
+ prev_req_cz0 = lt_decode_coef_update(priv->ld_update, C_Z0);
+
+ /* If previous step moved C+/C0 then go back on C+/C0 */
+ if (prev_req_cp1 == REQ_INC)
+ update = lt_encode_request(update, REQ_DEC, C_P1);
+ if (prev_req_cp1 == REQ_DEC)
+ update = lt_encode_request(update, REQ_INC, C_P1);
+ if (prev_req_cz0 == REQ_INC)
+ update = lt_encode_request(update, REQ_DEC, C_Z0);
+ if (prev_req_cz0 == REQ_DEC)
+ update = lt_encode_request(update, REQ_INC, C_Z0);
+
+ return update;
+}
+
+static u32 process_bin_long_early(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum coef_status lpst_cp1, lpst_cz0;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ /* Get LP coefficient status to determine
+ * if coefficient is in range or reached the limit thresholds
+ * IF the coefficient is at MIN/MAX and still want to INC/DEC
+ * THEN do we are done with this module
+ */
+ lpst_cp1 = lt_get_lp_coef_status(priv->eqdrv.lane, C_P1);
+ lpst_cz0 = lt_get_lp_coef_status(priv->eqdrv.lane, C_Z0);
+
+ /* Early path (make edge later) */
+ if (lpst_cp1 == COEF_MAX) {
+ if (lpst_cz0 == COEF_MAX) {
+ /* Hold C+, C0 */
+ update = lt_encode_startup_request(REQ_HOLD);
+ } else {
+ /* request Increment C0 and
+ * Decrement C+
+ */
+ update = lt_encode_request(update, REQ_INC, C_Z0);
+ update = lt_encode_request(update, REQ_DEC, C_P1);
+ }
+ } else {
+ /* request Increment C+ */
+ update = lt_encode_request(update, REQ_INC, C_P1);
+ }
+
+ return update;
+}
+
+static u32 process_bin_long_late(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum coef_status lpst_cp1, lpst_cz0;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ /* Get LP coefficient status to determine
+ * if coefficient is in range or reached the limit thresholds
+ * IF the coefficient is at MIN/MAX and still want to INC/DEC
+ * THEN do we are done with this module
+ */
+ lpst_cp1 = lt_get_lp_coef_status(priv->eqdrv.lane, C_P1);
+ lpst_cz0 = lt_get_lp_coef_status(priv->eqdrv.lane, C_Z0);
+
+ /* Late path (make edge earlier) */
+ if (lpst_cp1 == COEF_MIN) {
+ if (lpst_cz0 == COEF_MIN) {
+ /* Hold C0 */
+ update = lt_encode_startup_request(REQ_HOLD);
+ } else {
+ /* request Decrement C0 */
+ update = lt_encode_request(update, REQ_DEC, C_Z0);
+ }
+ } else {
+ /* request Decrement C+ */
+ update = lt_encode_request(update, REQ_DEC, C_P1);
+ }
+
+ return update;
+}
+
+static u32 process_bin_long_antipodal(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+ enum req_type prev_req_cp1;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return update;
+ }
+
+ prev_req_cp1 = lt_decode_coef_update(priv->ld_update, C_P1);
+
+ /* Request move on C+ and C0 */
+ /* If previous step moved C+ then go back on C+ */
+ if (prev_req_cp1 == REQ_INC)
+ update = lt_encode_request(update, REQ_DEC, C_P1);
+ if (prev_req_cp1 == REQ_DEC)
+ update = lt_encode_request(update, REQ_INC, C_P1);
+
+ if (priv->bin_long_state == BIN_LATE) {
+ /* request Decrement C0 */
+ update = lt_encode_request(update, REQ_DEC, C_Z0);
+ } else {
+ /* request Increment C0 */
+ update = lt_encode_request(update, REQ_INC, C_Z0);
+ }
+
+ return update;
+}
+
+/* process_bin_long
+ *
+ * Bin_Long:
+ * contains the scoring of final edges on pulses longer than 7UI long
+ * used to adjust LP coefficients: C_P1 and C_Z0
+ */
+static void process_bin_long(struct eq_data_priv *priv)
+{
+ u32 update = lt_encode_startup_request(REQ_HOLD);
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return;
+ }
+
+ if (priv->bin_long_state == BIN_INVALID) {
+ phydev_err(priv->eqdrv.phydev, "Invalid Bin_Long state\n");
+ return;
+ }
+
+ if (priv->bin_long_state == BIN_TOGGLE) {
+ update = process_bin_long_toggle(priv);
+ } else {
+ if (priv->prev_bin_long_state == BIN_TOGGLE) {
+ update = process_bin_long_prev_toggle(priv);
+ } else {
+ if (priv->prev_bin_long_state == priv->bin_long_state) {
+ if (priv->bin_long_state == BIN_LATE)
+ update = process_bin_long_late(priv);
+ else
+ update = process_bin_long_early(priv);
+ } else {
+ update = process_bin_long_antipodal(priv);
+ }
+ }
+ }
+
+ /* Store current algorithm decision
+ * as previous algorithm ld_update for next step
+ */
+ priv->ld_update = update;
+}
+
+/* Callbacks:
+ * required by generic equalization_algorithm
+ */
+static void process_bad_state(struct eq_data_priv *priv)
+{
+ bool lp_at_init, lp_at_preset;
+ u32 req;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return;
+ }
+
+ lp_at_init = lt_is_lp_at_startup(priv->eqdrv.lane, REQ_INIT);
+ lp_at_preset = lt_is_lp_at_startup(priv->eqdrv.lane, REQ_PRESET);
+
+ if (lp_at_init) {
+ /* Try Request Preset */
+ req = lt_encode_startup_request(REQ_PRESET);
+ lt_lp_update(priv->eqdrv.lane, req);
+ } else if (lp_at_preset) {
+ /* LT ERROR
+ * set lt_error flag to prevent reaching
+ * training state = TRAINED
+ * and resume training in case of LT error
+ */
+ lt_set_error(priv->eqdrv.lane, true);
+ phydev_err(priv->eqdrv.phydev,
+ "LT Error: CDR_LOCK is zero on Preset\n");
+ } else {
+ /* Move LP back to previous C-, C0, C+ and HOLD */
+ lt_move_lp_back(priv->eqdrv.lane);
+ }
+}
+
+static bool collect_bin_counters(struct eq_data_priv *priv,
+ enum eqc_type type)
+{
+ const struct equalizer_ops *eqops;
+ s16 *bin_snapshot = NULL;
+ int snp_size;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return false;
+ }
+
+ /* collect Bin snapshots */
+ switch (type) {
+ case EQC_BIN_1:
+ bin_snapshot = priv->bin1_snapshot;
+ break;
+ case EQC_BIN_2:
+ bin_snapshot = priv->bin2_snapshot;
+ break;
+ case EQC_BIN_3:
+ bin_snapshot = priv->bin3_snapshot;
+ break;
+ case EQC_BIN_LONG:
+ bin_snapshot = priv->bin_long_snapshot;
+ break;
+ case EQC_BIN_OFFSET:
+ bin_snapshot = priv->bin_offset_snapshot;
+ break;
+ case EQC_BIN_M1:
+ bin_snapshot = priv->bin_m1_snapshot;
+ break;
+ default:
+ /* invalid bin type */
+ return false;
+ }
+ if (!bin_snapshot)
+ return false;
+
+ eqops = &priv->eqdrv.equalizer->ops;
+ if (!eqops) {
+ phydev_err(priv->eqdrv.phydev,
+ "No operations for equalizer %s %s\n",
+ priv->eqdrv.equalizer->name,
+ priv->eqdrv.equalizer->version);
+ return false;
+ }
+ snp_size = eqops->collect_counters(priv->eqdrv.reg_base, type,
+ bin_snapshot, EQ_SNAPSHOTS_SIZE);
+ /* Check if snapshots collection failed */
+ if (snp_size < EQ_SNAPSHOTS_SIZE) {
+ phydev_err(priv->eqdrv.phydev,
+ "Counters collection failed for equalizer %s %s\n",
+ priv->eqdrv.equalizer->name,
+ priv->eqdrv.equalizer->version);
+ return false;
+ }
+
+ /* if CDR_LOCK = 0: Statistics are invalid */
+ if (!backplane_is_cdr_lock(priv->eqdrv.lane, true)) {
+ process_bad_state(priv);
+ return false;
+ }
+
+ return true;
+}
+
+static bool collect_bit_edge_statistics(struct eq_data_priv *priv)
+{
+ enum eqc_type st_types[] = { EQC_GAIN_HF, EQC_GAIN_MF, EQC_EQOFFSET };
+ s16 status_counters[ARRAY_SIZE(st_types)][EQ_SNAPSHOTS_SIZE];
+ const struct equalizer_ops *eqops;
+ int i, snp_size;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return false;
+ }
+
+ /* collect Bin snapshots */
+ if (!collect_bin_counters(priv, EQC_BIN_1))
+ return false;
+ if (!collect_bin_counters(priv, EQC_BIN_2))
+ return false;
+ if (!collect_bin_counters(priv, EQC_BIN_3))
+ return false;
+ if (!collect_bin_counters(priv, EQC_BIN_LONG))
+ return false;
+ if (!collect_bin_counters(priv, EQC_BIN_OFFSET))
+ return false;
+ if (!collect_bin_counters(priv, EQC_BIN_M1))
+ return false;
+
+ /* collect Gains */
+ eqops = &priv->eqdrv.equalizer->ops;
+ if (!eqops) {
+ phydev_err(priv->eqdrv.phydev,
+ "No operations for equalizer %s %s\n",
+ priv->eqdrv.equalizer->name,
+ priv->eqdrv.equalizer->version);
+ return false;
+ }
+ snp_size = eqops->collect_multiple_counters(priv->eqdrv.reg_base,
+ st_types,
+ ARRAY_SIZE(st_types),
+ (s16 *)status_counters,
+ EQ_SNAPSHOTS_SIZE);
+ /* Check if snapshots collection failed */
+ if (snp_size < EQ_SNAPSHOTS_SIZE) {
+ phydev_err(priv->eqdrv.phydev,
+ "Counters collection failed for equalizer %s %s\n",
+ priv->eqdrv.equalizer->name,
+ priv->eqdrv.equalizer->version);
+ return false;
+ }
+
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ priv->gain_hf_snapshot[i] = (u8)status_counters[0][i];
+ priv->gain_mf_snapshot[i] = (u8)status_counters[1][i];
+ priv->osestat_snapshot[i] = (u8)status_counters[2][i];
+ }
+
+ return true;
+}
+
+static void generate_3taps_request(struct eq_data_priv *priv)
+{
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return;
+ }
+
+ /* Store current state as previous state */
+ priv->prev_bin_m1_state = priv->bin_m1_state;
+ priv->prev_bin_long_state = priv->bin_long_state;
+
+ priv->bin_m1_state = get_bin_snapshots_state(priv, EQC_BIN_M1);
+ if (priv->bin_m1_state == BIN_INVALID) {
+ /* invalid state: should never happen */
+ return;
+ }
+
+ priv->bin_long_state = get_bin_snapshots_state(priv, EQC_BIN_LONG);
+ if (priv->bin_long_state == BIN_INVALID) {
+ /* invalid state: should never happen */
+ return;
+ }
+
+ /* Move to BinLong/BinM1 modules:
+ * Bin Modules order: BinLong before BinM1
+ * We try to finish BinLong before we do BinM1
+ */
+
+ /* Process BinLong module
+ * decide and ask for movement of C+/C0
+ */
+ if (!priv->bin_long_stop) {
+ process_bin_long(priv);
+ lt_lp_update(priv->eqdrv.lane, priv->ld_update);
+ if (lt_is_update_of_type(priv->ld_update, REQ_HOLD)) {
+ /* Sent All Hold request */
+ priv->long_min_max_cnt++;
+ if (priv->long_min_max_cnt >= TIMEOUT_LONG)
+ priv->bin_long_stop = true;
+ } else {
+ /* Sent C Inc/Dec request */
+ priv->long_min_max_cnt = 0;
+ }
+ return;
+ }
+
+ /* Process BinM1 module
+ * decide and ask for movement of C-
+ */
+ if (!priv->bin_m1_stop) {
+ process_bin_m1(priv);
+ lt_lp_update(priv->eqdrv.lane, priv->ld_update);
+ if (lt_is_update_of_type(priv->ld_update, REQ_HOLD)) {
+ /* Sent All Hold request */
+ priv->m1_min_max_cnt++;
+ if (priv->m1_min_max_cnt >= TIMEOUT_M1)
+ priv->bin_m1_stop = true;
+ } else {
+ /* Sent C Inc/Dec request */
+ priv->m1_min_max_cnt = 0;
+ }
+ return;
+ }
+}
+
+static bool is_rx_ok(struct eq_data_priv *priv)
+{
+ struct eqc_range *gain_range, *osestat_range;
+ u8 osestat_mid_low, osestat_mid_high;
+ enum bin_state bin1_snapshot_state;
+ enum bin_state bin2_snapshot_state;
+ enum bin_state bin3_snapshot_state;
+ const struct equalizer_ops *eqops;
+ bool rx_quality_1, rx_quality_2;
+ bool is_ok, rx_good_quality;
+ u8 min_eq_gain, max_eq_gain;
+ u8 min_snp, max_snp;
+ s16 snapshot;
+ int i;
+
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return false;
+ }
+
+ /* Checking Bins/Gains after LP has updated its TX */
+ eqops = &priv->eqdrv.equalizer->ops;
+ if (!eqops)
+ return false;
+ gain_range = eqops->get_counter_range(EQC_GAIN_HF);
+ if (!gain_range)
+ return false;
+ osestat_range = eqops->get_counter_range(EQC_EQOFFSET);
+ if (!osestat_range)
+ return false;
+
+ min_eq_gain = (u8)gain_range->min;
+ max_eq_gain = (u8)gain_range->max;
+ osestat_mid_low = (u8)osestat_range->mid_low;
+ osestat_mid_high = (u8)osestat_range->mid_high;
+
+ /* CDR_LOCK must be 1 */
+ if (!backplane_is_cdr_lock(priv->eqdrv.lane, true))
+ return false;
+
+ /* Offset Bin must NOT be 10 of the same value */
+ rx_good_quality = false;
+ snapshot = priv->bin_offset_snapshot[0];
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (snapshot != priv->bin_offset_snapshot[i]) {
+ rx_good_quality = true;
+ break;
+ }
+ }
+ if (!rx_good_quality)
+ return false;
+
+ /* Offset status must dither (+/-2) around MidRange value
+ * What we want to see is that the Offset has settled to a value
+ * somewhere between: mid-range low and mid-range high and that
+ * the series of snapshot values are +/-2 of the settled value.
+ */
+ rx_good_quality = true;
+ min_snp = priv->osestat_snapshot[0];
+ max_snp = priv->osestat_snapshot[0];
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->osestat_snapshot[i] < osestat_mid_low ||
+ priv->osestat_snapshot[i] > osestat_mid_high) {
+ rx_good_quality = false;
+ break;
+ }
+ if (priv->osestat_snapshot[i] < min_snp)
+ min_snp = priv->osestat_snapshot[i];
+ if (priv->osestat_snapshot[i] > max_snp)
+ max_snp = priv->osestat_snapshot[i];
+ }
+ if (max_snp - min_snp > 4)
+ rx_good_quality = false;
+ if (!rx_good_quality)
+ return false;
+
+ /* The Rx is in good quality if:
+ * Bin1, Bin2, and Bin3 are toggling
+ * Proceed to BinLong/BinM1 modules
+ */
+ bin1_snapshot_state = get_bin_snapshots_state(priv, EQC_BIN_1);
+ bin2_snapshot_state = get_bin_snapshots_state(priv, EQC_BIN_2);
+ bin3_snapshot_state = get_bin_snapshots_state(priv, EQC_BIN_3);
+
+ rx_good_quality = (bin1_snapshot_state == BIN_TOGGLE &&
+ bin2_snapshot_state == BIN_TOGGLE &&
+ bin3_snapshot_state == BIN_TOGGLE);
+
+ /* If Rx is in good quality then proceed to BinLong/BinM1 */
+ if (rx_good_quality)
+ return true;
+
+#ifdef ENABLE_LESS_QUALITY_CONDITION
+ rx_quality_1 = false;
+ rx_quality_2 = false;
+ if (bin1_snapshot_state == BIN_TOGGLE) {
+ if (bin2_snapshot_state == BIN_EARLY &&
+ bin3_snapshot_state == BIN_LATE) {
+ /* check if GainMF is stuck at max_eq_gain */
+ rx_quality_1 = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_mf_snapshot[i] != max_eq_gain) {
+ rx_quality_1 = false;
+ break;
+ }
+ }
+ }
+ if (bin2_snapshot_state == BIN_LATE &&
+ bin3_snapshot_state == BIN_EARLY) {
+ /* check if GainMF is stuck at min_eq_gain */
+ rx_quality_2 = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_mf_snapshot[i] != min_eq_gain) {
+ rx_quality_2 = false;
+ break;
+ }
+ }
+ }
+ }
+
+ /* If Rx is less quality then proceed to BinLong/BinM1 */
+ if (rx_quality_1 || rx_quality_2)
+ return true;
+#endif
+
+#ifdef ENABLE_EVEN_LESS_QUALITY_CONDITION
+ rx_quality_1 = false;
+ rx_quality_2 = false;
+ if (bin1_snapshot_state == BIN_EARLY &&
+ bin2_snapshot_state == BIN_LATE) {
+ /* check if GainHF is stuck at max_eq_gain */
+ is_ok = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_hf_snapshot[i] != max_eq_gain) {
+ is_ok = false;
+ break;
+ }
+ }
+ if (is_ok) {
+ /* check if GainMF is stuck at min_eq_gain */
+ is_ok = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_mf_snapshot[i] != min_eq_gain) {
+ is_ok = false;
+ break;
+ }
+ }
+ if (is_ok)
+ rx_quality_1 = true;
+ }
+ }
+ if (bin1_snapshot_state == BIN_LATE &&
+ bin2_snapshot_state == BIN_EARLY) {
+ /* check if GainHF is stuck at min_eq_gain */
+ is_ok = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_hf_snapshot[i] != min_eq_gain) {
+ is_ok = false;
+ break;
+ }
+ }
+ if (is_ok) {
+ /* check if GainMF is stuck at max_eq_gain */
+ is_ok = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_mf_snapshot[i] != max_eq_gain) {
+ is_ok = false;
+ break;
+ }
+ }
+ if (is_ok)
+ rx_quality_2 = true;
+ }
+ }
+
+ /* If Rx is in good quality then proceed to BinLong/BinM1 */
+ if (rx_quality_1 || rx_quality_2)
+ return true;
+#endif
+
+#ifdef ENABLE_SEEMINGLY_QUALITY_CONDITION
+ rx_quality_1 = false;
+ if (bin1_snapshot_state == BIN_LATE &&
+ bin2_snapshot_state == BIN_TOGGLE &&
+ bin3_snapshot_state == BIN_TOGGLE) {
+ /* check if GainHF is stuck at min_eq_gain */
+ rx_quality_1 = true;
+ for (i = 0; i < EQ_SNAPSHOTS_SIZE; i++) {
+ if (priv->gain_hf_snapshot[i] != min_eq_gain) {
+ rx_quality_1 = false;
+ break;
+ }
+ }
+ }
+
+ if (rx_quality_1)
+ return true;
+#endif
+
+ return false;
+}
+
+static bool is_eq_done(struct eq_data_priv *priv)
+{
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return false;
+ }
+
+ return (priv->bin_m1_stop && priv->bin_long_stop);
+}
+
+/* BEE 3 TAP Algorithm API */
+
+/* Create BEE 3-TAP Equalization Algorithm */
+static struct eq_data_priv *create(struct equalizer_driver eqdrv)
+{
+ struct eq_data_priv *bee_data;
+
+ bee_data = devm_kzalloc(&eqdrv.phydev->mdio.dev,
+ sizeof(*bee_data), GFP_KERNEL);
+ if (!bee_data)
+ return NULL;
+
+ /* initialize algorithm setup data */
+ bee_data->eqdrv = eqdrv;
+
+ /* initialize specific BEE algorithm data */
+ bee_data->bin_m1_state = BIN_INVALID;
+ bee_data->bin_long_state = BIN_INVALID;
+ bee_data->prev_bin_m1_state = BIN_INVALID;
+ bee_data->prev_bin_long_state = BIN_INVALID;
+ bee_data->m1_min_max_cnt = 0;
+ bee_data->long_min_max_cnt = 0;
+ bee_data->bin_m1_stop = false;
+ bee_data->bin_long_stop = false;
+ bee_data->ld_update = 0;
+
+ return bee_data;
+}
+
+static void destroy(struct eq_data_priv *priv)
+{
+ if (!priv) {
+ pr_err("%s: NULL private data\n", ALGORITHM_NAME);
+ return;
+ }
+
+ kfree(priv);
+}
+
+static const struct equalization_algorithm eq_alg = {
+ .name = ALGORITHM_NAME,
+ .descr = ALGORITHM_DESCR,
+ .version = ALGORITHM_VERSION,
+ .use_local_tx_training = true,
+ .use_remote_tx_training = true,
+ .ops = {
+ .create = create,
+ .destroy = destroy,
+ .is_rx_ok = is_rx_ok,
+ .is_eq_done = is_eq_done,
+ .collect_statistics = collect_bit_edge_statistics,
+ .generate_request = generate_3taps_request,
+ .process_bad_state = process_bad_state,
+ .dump_algorithm_context = NULL,
+ }
+};
+
+static const char * const alg_keys[] = {
+ "bee",
+ "BEE",
+};
+
+static int __init bee_init(void)
+{
+ int i, err;
+
+ pr_info("%s: %s algorithm version %s\n",
+ ALGORITHM_NAME, ALGORITHM_DESCR, ALGORITHM_VERSION);
+
+ /* register BEE algorithm: */
+ for (i = 0; i < ARRAY_SIZE(alg_keys); i++) {
+ err = backplane_eq_register(alg_keys[i], &eq_alg);
+ if (err) {
+ pr_err("%s: '%s' equalization algorithm registration failed\n",
+ ALGORITHM_NAME, alg_keys[i]);
+ }
+ }
+
+ return 0;
+}
+
+static void __exit bee_exit(void)
+{
+ int i;
+
+ /* unregister BEE algorithm: */
+ for (i = 0; i < ARRAY_SIZE(alg_keys); i++)
+ backplane_eq_unregister(alg_keys[i]);
+
+ pr_info("%s: %s algorithm version %s unloaded\n",
+ ALGORITHM_NAME, ALGORITHM_DESCR, ALGORITHM_VERSION);
+}
+
+module_init(bee_init);
+module_exit(bee_exit);
+
+MODULE_DESCRIPTION("Bit Edge Equalization Algorithm");
+MODULE_AUTHOR("Florinel Iordache <florinel.iordache@nxp.com>");
+MODULE_LICENSE("Dual BSD/GPL");
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio description
2020-06-22 13:35 [PATCH net-next v2 0/9] net: ethernet backplane support on DPAA1 Florinel Iordache
` (5 preceding siblings ...)
2020-06-22 13:35 ` [PATCH net-next v3 6/7] net: phy: add bee algorithm for kr training Florinel Iordache
@ 2020-06-22 13:35 ` Florinel Iordache
2020-06-26 19:08 ` Florian Fainelli
6 siblings, 1 reply; 23+ messages in thread
From: Florinel Iordache @ 2020-06-22 13:35 UTC (permalink / raw)
To: davem, netdev, andrew, f.fainelli, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel,
Florinel Iordache
Add dt nodes with serdes, lanes, mdio generic description for supported
platform: ls1046. This is a prerequisite to enable backplane on device
tree for these platforms.
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
---
arch/arm64/boot/dts/freescale/fsl-ls1046a.dtsi | 33 +++++++++++++++++++++-
.../boot/dts/freescale/qoriq-fman3-0-10g-0.dtsi | 5 ++--
.../boot/dts/freescale/qoriq-fman3-0-10g-1.dtsi | 5 ++--
3 files changed, 38 insertions(+), 5 deletions(-)
diff --git a/arch/arm64/boot/dts/freescale/fsl-ls1046a.dtsi b/arch/arm64/boot/dts/freescale/fsl-ls1046a.dtsi
index d4c1da3..c7d845f 100644
--- a/arch/arm64/boot/dts/freescale/fsl-ls1046a.dtsi
+++ b/arch/arm64/boot/dts/freescale/fsl-ls1046a.dtsi
@@ -3,7 +3,7 @@
* Device Tree Include file for Freescale Layerscape-1046A family SoC.
*
* Copyright 2016 Freescale Semiconductor, Inc.
- * Copyright 2018 NXP
+ * Copyright 2018, 2020 NXP
*
* Mingkai Hu <mingkai.hu@nxp.com>
*/
@@ -735,6 +735,37 @@
status = "disabled";
};
+ serdes1: serdes@1ea0000 {
+ compatible = "serdes-10g";
+ reg = <0x0 0x1ea0000 0 0x00002000>;
+ reg-names = "serdes", "serdes-10g";
+ big-endian;
+
+ #address-cells = <1>;
+ #size-cells = <1>;
+ ranges = <0x0 0x00 0x1ea0000 0x00002000>;
+ lane_a: lane@800 {
+ compatible = "lane-10g";
+ reg = <0x800 0x40>;
+ reg-names = "lane", "serdes-lane";
+ };
+ lane_b: lane@840 {
+ compatible = "lane-10g";
+ reg = <0x840 0x40>;
+ reg-names = "lane", "serdes-lane";
+ };
+ lane_c: lane@880 {
+ compatible = "lane-10g";
+ reg = <0x880 0x40>;
+ reg-names = "lane", "serdes-lane";
+ };
+ lane_d: lane@8c0 {
+ compatible = "lane-10g";
+ reg = <0x8c0 0x40>;
+ reg-names = "lane", "serdes-lane";
+ };
+ };
+
pcie_ep@3600000 {
compatible = "fsl,ls1046a-pcie-ep", "fsl,ls-pcie-ep";
reg = <0x00 0x03600000 0x0 0x00100000
diff --git a/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-0.dtsi b/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-0.dtsi
index dbd2fc3..d6191f1 100644
--- a/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-0.dtsi
+++ b/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-0.dtsi
@@ -3,6 +3,7 @@
* QorIQ FMan v3 10g port #0 device tree
*
* Copyright 2012-2015 Freescale Semiconductor Inc.
+ * Copyright 2020 NXP
*
*/
@@ -21,7 +22,7 @@ fman@1a00000 {
fsl,fman-10g-port;
};
- ethernet@f0000 {
+ mac9: ethernet@f0000 {
cell-index = <0x8>;
compatible = "fsl,fman-memac";
reg = <0xf0000 0x1000>;
@@ -29,7 +30,7 @@ fman@1a00000 {
pcsphy-handle = <&pcsphy6>;
};
- mdio@f1000 {
+ mdio9: mdio@f1000 {
#address-cells = <1>;
#size-cells = <0>;
compatible = "fsl,fman-memac-mdio", "fsl,fman-xmdio";
diff --git a/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-1.dtsi b/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-1.dtsi
index 6fc5d25..1f6f28f 100644
--- a/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-1.dtsi
+++ b/arch/arm64/boot/dts/freescale/qoriq-fman3-0-10g-1.dtsi
@@ -3,6 +3,7 @@
* QorIQ FMan v3 10g port #1 device tree
*
* Copyright 2012-2015 Freescale Semiconductor Inc.
+ * Copyright 2020 NXP
*
*/
@@ -21,7 +22,7 @@ fman@1a00000 {
fsl,fman-10g-port;
};
- ethernet@f2000 {
+ mac10: ethernet@f2000 {
cell-index = <0x9>;
compatible = "fsl,fman-memac";
reg = <0xf2000 0x1000>;
@@ -29,7 +30,7 @@ fman@1a00000 {
pcsphy-handle = <&pcsphy7>;
};
- mdio@f3000 {
+ mdio10: mdio@f3000 {
#address-cells = <1>;
#size-cells = <0>;
compatible = "fsl,fman-memac-mdio", "fsl,fman-xmdio";
--
1.9.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio description
2020-06-22 13:35 ` [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio description Florinel Iordache
@ 2020-06-26 19:08 ` Florian Fainelli
2020-06-29 12:32 ` [EXT] " Florinel Iordache
0 siblings, 1 reply; 23+ messages in thread
From: Florian Fainelli @ 2020-06-26 19:08 UTC (permalink / raw)
To: florinel.iordache, davem, netdev, andrew, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, leoyang.li, madalin.bucur, ioana.ciornei, linux-kernel
On 6/22/20 6:35 AM, Florinel Iordache wrote:
> Add dt nodes with serdes, lanes, mdio generic description for supported
> platform: ls1046. This is a prerequisite to enable backplane on device
> tree for these platforms.
>
> Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
> ---
[snip]
> @@ -21,7 +22,7 @@ fman@1a00000 {
> fsl,fman-10g-port;
> };
>
> - ethernet@f2000 {
> + mac10: ethernet@f2000 {
> cell-index = <0x9>;
> compatible = "fsl,fman-memac";
> reg = <0xf2000 0x1000>;
> @@ -29,7 +30,7 @@ fman@1a00000 {
> pcsphy-handle = <&pcsphy7>;
> };
>
> - mdio@f3000 {
> + mdio10: mdio@f3000 {
> #address-cells = <1>;
> #size-cells = <0>;
> compatible = "fsl,fman-memac-mdio", "fsl,fman-xmdio";
That part of the patch does not seem to have any users, there is nothing
in your patch series that references mac9, mdio9 or mac10, mdio10
AFAICT. Maybe you can split it up and send it aside from this series?
--
Florian
^ permalink raw reply [flat|nested] 23+ messages in thread* RE: [EXT] Re: [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio description
2020-06-26 19:08 ` Florian Fainelli
@ 2020-06-29 12:32 ` Florinel Iordache
0 siblings, 0 replies; 23+ messages in thread
From: Florinel Iordache @ 2020-06-29 12:32 UTC (permalink / raw)
To: Florian Fainelli, davem, netdev, andrew, hkallweit1, linux
Cc: devicetree, linux-doc, robh+dt, mark.rutland, kuba, corbet,
shawnguo, Leo Li, Madalin Bucur (OSS),
Ioana Ciornei, linux-kernel, Florinel Iordache
> -----Original Message-----
> From: Florian Fainelli <f.fainelli@gmail.com>
> Sent: Friday, June 26, 2020 10:08 PM
> To: Florinel Iordache <florinel.iordache@nxp.com>; davem@davemloft.net;
> netdev@vger.kernel.org; andrew@lunn.ch; hkallweit1@gmail.com;
> linux@armlinux.org.uk
> Cc: devicetree@vger.kernel.org; linux-doc@vger.kernel.org;
> robh+dt@kernel.org; mark.rutland@arm.com; kuba@kernel.org;
> corbet@lwn.net; shawnguo@kernel.org; Leo Li <leoyang.li@nxp.com>; Madalin
> Bucur (OSS) <madalin.bucur@oss.nxp.com>; Ioana Ciornei
> <ioana.ciornei@nxp.com>; linux-kernel@vger.kernel.org
> Subject: [EXT] Re: [PATCH net-next v3 7/7] arm64: dts: add serdes and mdio
> description
>
> Caution: EXT Email
>
> On 6/22/20 6:35 AM, Florinel Iordache wrote:
> > Add dt nodes with serdes, lanes, mdio generic description for
> > supported
> > platform: ls1046. This is a prerequisite to enable backplane on device
> > tree for these platforms.
> >
> > Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
> > ---
>
> [snip]
>
> > @@ -21,7 +22,7 @@ fman@1a00000 {
> > fsl,fman-10g-port;
> > };
> >
> > - ethernet@f2000 {
> > + mac10: ethernet@f2000 {
> > cell-index = <0x9>;
> > compatible = "fsl,fman-memac";
> > reg = <0xf2000 0x1000>;
> > @@ -29,7 +30,7 @@ fman@1a00000 {
> > pcsphy-handle = <&pcsphy7>;
> > };
> >
> > - mdio@f3000 {
> > + mdio10: mdio@f3000 {
> > #address-cells = <1>;
> > #size-cells = <0>;
> > compatible = "fsl,fman-memac-mdio", "fsl,fman-xmdio";
>
> That part of the patch does not seem to have any users, there is nothing in your
> patch series that references mac9, mdio9 or mac10, mdio10 AFAICT. Maybe you
> can split it up and send it aside from this series?
> --
> Florian
Hi Florian,
These labels are used in the platform DTS (e.g.: fsl-ls1046a-rdb.dts for DPAA1 / LS1046).
They are not used in the patch because backplane mode is NOT the default mode.
Backplane mode is enabled by users for custom boards by editing the DTS.
Interfaces must be used by default with a PHY like: phy-handle = <&aqr106_phy>
How to enable backplane mode in DTS is described in devicetree bindings file: ethernet-phy.yaml
(which includes an example at the end).
Here is an example of how the labels are used to enable backplane mode:
&mdio9 {
pcsphy6: ethernet-phy@0 {
compatible = "ethernet-phy-ieee802.3-c45";
eq-algorithm = "bee";
/* 10G Short cables setup: up to 30 cm cable */
eq-init = <0x2 0x29 0x5>;
eq-params = <0>;
reg = <0x0>;
lane-handle = <&lane_d>; /* lane D */
};
};
/* Update MAC connections to backplane PHYs */
&mac9 {
phy-connection-type = "10gbase-kr";
phy-handle = <&pcsphy6>;
};
However, the users could write DTS by using ethernet@f0000 instead of &mac9.
Something like this:
ethernet@f0000 { /* 10GEC1 */
phy-connection-type = "10gbase-kr";
phy-handle = <&pcsphy6>;
};
(and similar for mdio9/mdio10)
So ultimately that part of the patch could be removed indeed.
I put it there just to be more user friendly to enable backplane in DTS.
Thank you for feedback.
Florinel.
^ permalink raw reply [flat|nested] 23+ messages in thread