rework phylink interface for split MAC/PCS support
Subsystems (21)
Russell King - ARM Linux admin
maintainerv2 cover rework phylink interface for split MAC/PCS support
Hi,
The following series changes the phylink interface to allow us to better support split MAC / MAC PCS setups. The fundamental change required for this turns out to be quite simple.
Today, mac_config() is used for everything to do with setting the parameters for the MAC, and mac_link_up() is used to inform the MAC driver that the link is now up (and so to allow packet flow.) mac_config() also has had a few implementation issues, with folk who believe that members such as "speed" and "duplex" are always valid, where "link" gets used inappropriately, etc.
With the proposed patches, all this changes subtly - but in a backwards compatible way at this stage.
We pass the the full resolved link state (speed, duplex, pause) to mac_link_up(), and it is now guaranteed that these parameters to this function will always be valid (no more SPEED_UNKNOWN or DUPLEX_UNKNOWN here - unless phylink is fed with such things.)
Drivers should convert over to using the state in mac_link_up() rather than configuring the speed, duplex and pause in the mac_config() method. The patch series includes a number of MAC drivers which I've thought have been easy targets - I've left the remainder as I think they need maintainer input. However, *all* drivers will need conversion for future phylink development.
v2: add ocelot/felix and qca/ar9331 DSA drivers to patch 2, add
received tested-by so far.
Documentation/networking/sfp-phylink.rst | 17 +++- drivers/net/dsa/b53/b53_common.c | 4 +- drivers/net/dsa/b53/b53_priv.h | 4 +- drivers/net/dsa/bcm_sf2.c | 4 +- drivers/net/dsa/lantiq_gswip.c | 4 +- drivers/net/dsa/mt7530.c | 4 +- drivers/net/dsa/mv88e6xxx/chip.c | 79 +++++++++++++---- drivers/net/dsa/ocelot/felix.c | 4 +- drivers/net/dsa/qca/ar9331.c | 4 +- drivers/net/dsa/sja1105/sja1105_main.c | 4 +- drivers/net/ethernet/cadence/macb.h | 1 - drivers/net/ethernet/cadence/macb_main.c | 57 +++++++----- drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.c | 61 ++++++++----- drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.h | 1 + drivers/net/ethernet/marvell/mvneta.c | 63 ++++++++----- drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c | 102 +++++++++++++--------- drivers/net/ethernet/mediatek/mtk_eth_soc.c | 7 +- drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 4 +- drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 37 ++++---- drivers/net/phy/phylink.c | 9 +- include/linux/phylink.h | 57 ++++++++---- include/net/dsa.h | 4 +- net/dsa/port.c | 7 +- 23 files changed, 358 insertions(+), 180 deletions(-)
Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
PATCH v2 1/8 net: phylink: propagate resolved link config via mac_link_up()
Propagate the resolved link parameters via the mac_link_up() call for MACs that do not automatically track their PCS state. We propagate the link parameters via function arguments so that inappropriate members of struct phylink_link_state can't be accessed, and creating a new structure just for this adds needless complexity to the API.
--- Documentation/networking/sfp-phylink.rst | 17 +++++- drivers/net/ethernet/cadence/macb_main.c | 7 ++- .../net/ethernet/freescale/dpaa2/dpaa2-mac.c | 7 ++- drivers/net/ethernet/marvell/mvneta.c | 8 ++- .../net/ethernet/marvell/mvpp2/mvpp2_main.c | 19 +++++-- drivers/net/ethernet/mediatek/mtk_eth_soc.c | 7 ++- .../net/ethernet/stmicro/stmmac/stmmac_main.c | 4 +- .../net/ethernet/xilinx/xilinx_axienet_main.c | 7 ++- drivers/net/phy/phylink.c | 9 ++- include/linux/phylink.h | 57 ++++++++++++++----- net/dsa/port.c | 4 +- 11 files changed, 105 insertions(+), 41 deletions(-)
Documentation/networking/sfp-phylink.rst +14 -3
@@ -74,10 +74,13 @@ phylib to the sfp/phylink support. Please send patches to improve this documentation. 1. Optionally split the network driver's phylib update function into - three parts dealing with link-down, link-up and reconfiguring the - MAC settings. This can be done as a separate preparation commit. + two parts dealing with link-down and link-up. This can be done as + a separate preparation commit. - An example of this preparation can be found in git commit fc548b991fb0. + An older example of this preparation can be found in git commit + fc548b991fb0, although this was splitting into three parts; the + link-up part now includes configuring the MAC for the link settings. + Please see :c:func:`mac_link_up` for more information on this. 2. Replace::
@@ -207,6 +210,14 @@ this documentation. using. This is particularly important for in-band negotiation methods such as 1000base-X and SGMII. + The :c:func:`mac_link_up` method is used to inform the MAC that the + link has come up. The call includes the negotiation mode and interface + for reference only. The finalised link parameters are also supplied + (speed, duplex and flow control/pause enablement settings) which + should be used to configure the MAC when the MAC and PCS are not + tightly integrated, or when the settings are not coming from in-band + negotiation. + The :c:func:`mac_config` method is used to update the MAC with the requested state, and must avoid unnecessarily taking the link down when making changes to the MAC configuration. This means the
drivers/net/ethernet/cadence/macb_main.c +5 -2
@@ -626,8 +626,11 @@ static void macb_mac_link_down(struct phylink_config *config, unsigned int mode,netif_tx_stop_all_queues (ndev); } -static void macb_mac_link_up (struct phylink_config *config,unsigned int mode, -phy_interface_t interface,struct phy_device *phy) +static void macb_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct net_device *ndev =to_net_dev (config->dev);struct macb *bp =netdev_priv (ndev);
drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.c +5 -2
@@ -154,8 +154,11 @@ static void dpaa2_mac_config(struct phylink_config *config, unsigned int mode,netdev_err (mac->net_dev,"dpmac_set_link_state() = %d\n" , err); } -static void dpaa2_mac_link_up (struct phylink_config *config,unsigned int mode, -phy_interface_t interface,struct phy_device *phy) +static void dpaa2_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct dpaa2_mac *mac =phylink_to_dpaa2_mac (config);struct dpmac_link_state *dpmac_state = &mac->state;
drivers/net/ethernet/marvell/mvneta.c +5 -3
@@ -3965,9 +3965,11 @@ static void mvneta_mac_link_down(struct phylink_config *config,mvneta_set_eee (pp, false); } -static void mvneta_mac_link_up (struct phylink_config *config,unsigned int mode, -phy_interface_t interface, -struct phy_device *phy) +static void mvneta_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct net_device *ndev =to_net_dev (config->dev);struct mvneta_port *pp =netdev_priv (ndev);
drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c +13 -6
@@ -58,8 +58,11 @@ static struct {*/ static void mvpp2_mac_config (struct phylink_config *config,unsigned int mode,const struct phylink_link_state *state); -static void mvpp2_mac_link_up (struct phylink_config *config,unsigned int mode, -phy_interface_t interface,struct phy_device *phy); +static void mvpp2_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause);/* Queue modes */ #define MVPP2_QDIST_SINGLE_MODE 0
@@ -3473,8 +3476,9 @@ static void mvpp2_start_dev(struct mvpp2_port *port) .interface = port->phy_interface, };mvpp2_mac_config (&port->phylink_config,MLO_AN_INBAND , &state); -mvpp2_mac_link_up (&port->phylink_config,MLO_AN_INBAND , - port->phy_interface,NULL ); +mvpp2_mac_link_up (&port->phylink_config,NULL , +MLO_AN_INBAND , port->phy_interface, +SPEED_UNKNOWN ,DUPLEX_UNKNOWN , false, false); }netif_tx_start_all_queues (port->dev);
@@ -5141,8 +5145,11 @@ static void mvpp2_mac_config(struct phylink_config *config, unsigned int mode,mvpp2_port_enable (port); } -static void mvpp2_mac_link_up (struct phylink_config *config,unsigned int mode, -phy_interface_t interface,struct phy_device *phy) +static void mvpp2_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct net_device *dev =to_net_dev (config->dev);struct mvpp2_port *port =netdev_priv (dev);
drivers/net/ethernet/mediatek/mtk_eth_soc.c +4 -3
@@ -412,9 +412,10 @@ static void mtk_mac_link_down(struct phylink_config *config, unsigned int mode,mtk_w32 (mac->hw, mcr,MTK_MAC_MCR (mac->id)); } -static void mtk_mac_link_up (struct phylink_config *config,unsigned int mode, -phy_interface_t interface, -struct phy_device *phy) +static void mtk_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex,bool tx_pause,bool rx_pause) { struct mtk_mac *mac = container_of(config, struct mtk_mac, phylink_config);
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +3 -1
@@ -950,8 +950,10 @@ static void stmmac_mac_link_down(struct phylink_config *config, }static void stmmac_mac_link_up (struct phylink_config *config, +struct phy_device *phy,unsigned int mode,phy_interface_t interface, -struct phy_device *phy) +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct stmmac_priv *priv =netdev_priv (to_net_dev (config->dev));
drivers/net/ethernet/xilinx/xilinx_axienet_main.c +4 -3
@@ -1486,9 +1486,10 @@ static void axienet_mac_link_down(struct phylink_config *config, }static void axienet_mac_link_up (struct phylink_config *config, -unsigned int mode, -phy_interface_t interface, -struct phy_device *phy) +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause) {/* nothing meaningful to do */ }
drivers/net/phy/phylink.c +7 -2
@@ -480,8 +480,11 @@ static void phylink_mac_link_up(struct phylink *pl,struct net_device *ndev = pl->netdev; pl->cur_interface = link_state.interface; - pl->ops->mac_link_up (pl->config, pl->cur_link_an_mode, - pl->cur_interface, pl->phydev); + pl->ops->mac_link_up (pl->config, pl->phydev, + pl->cur_link_an_mode, pl->cur_interface, + link_state.speed, link_state.duplex, + !!(link_state.pause &MLO_PAUSE_TX ), + !!(link_state.pause &MLO_PAUSE_RX ));if (ndev)netif_carrier_on (ndev);
@@ -547,6 +550,8 @@ static void phylink_resolve(struct work_struct *w) link_state.pause = pl->phy_state.pause;phylink_apply_manual_flow (pl, &link_state);phylink_mac_config (pl, &link_state); + }else { +phylink_apply_manual_flow (pl, &link_state); }break ; }
include/linux/phylink.h +42 -15
88 lines not on this page · raw message
net/dsa/port.c +3 -1
@@ -489,9 +489,11 @@ static void dsa_port_phylink_mac_link_down(struct phylink_config *config, }static void dsa_port_phylink_mac_link_up (struct phylink_config *config, +struct phy_device *phydev,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +int speed,int duplex, +bool tx_pause,bool rx_pause) { struct dsa_port *dp = container_of(config, struct dsa_port, pl_config);struct dsa_switch *ds = dp->ds;
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Hi Russell,
Quoted 51 lines · Russell King
Propagate the resolved link parameters via the
mac_link_up()call for MACs that do not automatically track their PCS state. We propagate the link parameters via function arguments so that inappropriate members of structphylink_link_statecan't be accessed, and creating a new structure just for this adds needless complexity to the API.Tested-by Andre Przywara <[email protected]>Tested-by Alexandre Belloni <[email protected]>Tested-by Vladimir Oltean <[email protected]>Signed-off-by Russell King <[email protected]>--- Documentation/networking/sfp-phylink.rst | 17 +++++- drivers/net/ethernet/cadence/macb_main.c | 7 ++- .../net/ethernet/freescale/dpaa2/dpaa2-mac.c | 7 ++- drivers/net/ethernet/marvell/mvneta.c | 8 ++- .../net/ethernet/marvell/mvpp2/mvpp2_main.c | 19 +++++-- drivers/net/ethernet/mediatek/mtk_eth_soc.c | 7 ++- .../net/ethernet/stmicro/stmmac/stmmac_main.c | 4 +- .../net/ethernet/xilinx/xilinx_axienet_main.c | 7 ++- drivers/net/phy/phylink.c | 9 ++- include/linux/phylink.h | 57 ++++++++++++++----- net/dsa/port.c | 4 +- 11 files changed, 105 insertions(+), 41 deletions(-)Documentation/networking/sfp-phylink.rst +7
-3@@ -74,10 +74,13 @@ phylib to the sfp/phylink support. Please send patches to improve this documentation. 1. Optionally split the network driver's phylib update function into - three parts dealing with link-down, link-up and reconfiguring the - MAC settings. This can be done as a separate preparation commit. + two parts dealing with link-down and link-up. This can be done as + a separate preparation commit. - An example of this preparation can be found in git commit fc548b991fb0. + An older example of this preparation can be found in git commit + fc548b991fb0, although this was splitting into three parts; the + link-up part now includes configuring the MAC for the link settings. + Please see :c:func:`mac_link_up` for more information on this. 2. Replace::@@ -207,6 +210,14 @@ this documentation. using. This is particularly important for in-band negotiation methods such as 1000base-X and SGMII. + The :c:func:`mac_link_up` method is used to inform the MAC that the
Documentation/networking/sfp-phylink.rst this documentation. · line 223
Quoted 10 lines
Documentation/networking/sfp-phylink.rst +7
-0+ link has come up. The call includes the negotiation mode and interface + for reference only. The finalised link parameters are also supplied + (speed, duplex and flow control/pause enablement settings) which + should be used to configure the MAC when the MAC and PCS are not + tightly integrated, or when the settings are not coming from in-band + negotiation. + The :c:func:`mac_config` method is used to update the MAC with the requested state, and must avoid unnecessarily taking the link down when making changes to the MAC configuration. This means the
Just to make sure I understand the changes:
- A MAC with no PCS can be configured in either .mac_config or .mac_link_up
- A MAC that needs to be manually reconfigured to the link mode negotiated by its PCS needs to have the PCS configured in .mac_config and the MAC in .mac_link_up
- A MAC with PCS where the MAC follows the PCS negotiation automatically in hardware is basically equivalent with a MAC with no PCS, and therefore can be configured in either .mac_config or .mac_link_up
Is this correct?
Regards,
-Vladimir
Russell King - ARM Linux admin
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 53 lines · Vladimir Oltean
Hi Russell,
Quoted 50 lines · Russell King
Propagate the resolved link parameters via the
mac_link_up()call for MACs that do not automatically track their PCS state. We propagate the link parameters via function arguments so that inappropriate members of structphylink_link_statecan't be accessed, and creating a new structure just for this adds needless complexity to the API.Tested-by Andre Przywara <[email protected]>Tested-by Alexandre Belloni <[email protected]>Tested-by Vladimir Oltean <[email protected]>Signed-off-by Russell King <[email protected]>--- Documentation/networking/sfp-phylink.rst | 17 +++++- drivers/net/ethernet/cadence/macb_main.c | 7 ++- .../net/ethernet/freescale/dpaa2/dpaa2-mac.c | 7 ++- drivers/net/ethernet/marvell/mvneta.c | 8 ++- .../net/ethernet/marvell/mvpp2/mvpp2_main.c | 19 +++++-- drivers/net/ethernet/mediatek/mtk_eth_soc.c | 7 ++- .../net/ethernet/stmicro/stmmac/stmmac_main.c | 4 +- .../net/ethernet/xilinx/xilinx_axienet_main.c | 7 ++- drivers/net/phy/phylink.c | 9 ++- include/linux/phylink.h | 57 ++++++++++++++----- net/dsa/port.c | 4 +- 11 files changed, 105 insertions(+), 41 deletions(-)Documentation/networking/sfp-phylink.rst +6
-3@@ -74,10 +74,13 @@ phylib to the sfp/phylink support. Please send patches to improve this documentation. 1. Optionally split the network driver's phylib update function into - three parts dealing with link-down, link-up and reconfiguring the - MAC settings. This can be done as a separate preparation commit. + two parts dealing with link-down and link-up. This can be done as + a separate preparation commit. - An example of this preparation can be found in git commit fc548b991fb0. + An older example of this preparation can be found in git commit + fc548b991fb0, although this was splitting into three parts; the + link-up part now includes configuring the MAC for the link settings. + Please see :c:func:`mac_link_up` for more information on this. 2. Replace::@@ -207,6 +210,14 @@ this documentation. using. This is particularly important for in-band negotiation methods such as 1000base-X and SGMII.
Quoted 14 lines
Documentation/networking/sfp-phylink.rst this documentation. · line 223
Quoted 11 lines
Documentation/networking/sfp-phylink.rst +8
-0+ The :c:func:`mac_link_up` method is used to inform the MAC that the + link has come up. The call includes the negotiation mode and interface + for reference only. The finalised link parameters are also supplied + (speed, duplex and flow control/pause enablement settings) which + should be used to configure the MAC when the MAC and PCS are not + tightly integrated, or when the settings are not coming from in-band + negotiation. + The :c:func:`mac_config` method is used to update the MAC with the requested state, and must avoid unnecessarily taking the link down when making changes to the MAC configuration. This means theJust to make sure I understand the changes:
- A MAC with no PCS can be configured in either .mac_config or .mac_link_up
I would much prefer mac_link_up to be used for setting the speed, duplex and pause modes for future-proofing in all cases except for the case where these parameters are automatically updated in the MAC from its associated PCS.
mac_link_up must not be used to configure the AN mode or interface mode; these must be configured in mac_config().
Quoted 3 lines
- A MAC that needs to be manually reconfigured to the link mode negotiated by its PCS needs to have the PCS configured in .mac_config and the MAC in .mac_link_up
I do have further changes that split the PCS ops from the MAC ops, so what is in this series is not the full story yet - some of the further patches can be found in my "phy" branch and "cex7" branches where I add support to dpaa2 for automatically switching between SGMII and 1000BASE-X. dpaa2 is one of these split PCS/MAC setups, but with the extra complication that there's a firmware layer between the PCS and MAC.
However, this series is the first stand-alone step along the road to supporting split PCS/MAC setups in a sane manner.
I discussed with Andrew Lunn how much to send, and the conclusion was to make the changes in a number of small patch series, as large patch series tend not to get reviewed. My experience with _this_ series is that even this is very difficult to get feedback for, so adding any additional patches will just make that worse.
Quoted 4 lines
- A MAC with PCS where the MAC follows the PCS negotiation automatically in hardware is basically equivalent with a MAC with no PCS, and therefore can be configured in either .mac_config or .mac_link_up
In this case, mac_link_up doesn't do anything with the speed/duplex/ pause stuff when those are automatically passed from the PCS.
I'm afraid that sentence contains a subtlety that's going to get
people: it is not clear cut because of the different natures of the various links.
In 1000BASE-X, speed is fixed at 1G, and the PCS autonegotiates the duplex and pause with the remote end. For mvneta (an example of a combined PCS/MAC implementation) operating in-band:
- In mac_config():
- configures for 1000BASE-X interface type with in-band AN.
- configures fixed 1G.
- As mvneta only supports full duplex, we disable duplex negotiation and force full duplex.
- Only symmetric pause is supported, and we set the symmetric pause advertisement appropriately, with pause negotiation enabled.
- In mac_link_up():
- merely allow the device to transmit and receive.
The MAC will be forced to 1G, full duplex, and will automatically be configured by the PCS for pause support depending on the hardware based pause resolution.
The situation is different for SGMII operating in-band:
- In mac_config():
- configures for SGMII interface type with in-band AN.
- configures speed and duplex negotiation.
- disables pause negotiation; SGMII has no support for this.
- In mac_link_up():
- enables or disables pause frames depending on the tx_pause/ rx_pause flags, since this is not available from the MAC.
- allow the device to transmit and receive.
If we aren't operating in in-band mode, then:
- In mac_config():
- configures for the interface type without in-band AN.
- disables speed, duplex and pause negotiation.
- In mac_link_up():
- sets the speed, duplex and pause frames depending on the passed parameters.
- allow the device to transmit and receive.
Please see patch 7 of this series which implements this for mvneta.
So, there is a split between what mac_config() should be doing and what mac_link_up() should be doing; this is why I've said in the documentation that the "mode" and "interface" are for reference only in mac_link_up() - mac_link_up() can use these to decide _how_ to program the resolved parameters, but should _not_ use them to determine the link configuration (such as changing the interface between SGMII and 1000BASE-X - that is the responsibility of mac_config().)
I hope that helps clarify the situation.
Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Vladimir Oltean
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 138 lines · Russell King - ARM Linux admin
Quoted 53 lines · Vladimir Oltean
Hi Russell,
Quoted 50 lines · Russell King
Propagate the resolved link parameters via the
mac_link_up()call for MACs that do not automatically track their PCS state. We propagate the link parameters via function arguments so that inappropriate members of structphylink_link_statecan't be accessed, and creating a new structure just for this adds needless complexity to the API.Tested-by Andre Przywara <[email protected]>Tested-by Alexandre Belloni <[email protected]>Tested-by Vladimir Oltean <[email protected]>Signed-off-by Russell King <[email protected]>--- Documentation/networking/sfp-phylink.rst | 17 +++++- drivers/net/ethernet/cadence/macb_main.c | 7 ++- .../net/ethernet/freescale/dpaa2/dpaa2-mac.c | 7 ++- drivers/net/ethernet/marvell/mvneta.c | 8 ++- .../net/ethernet/marvell/mvpp2/mvpp2_main.c | 19 +++++-- drivers/net/ethernet/mediatek/mtk_eth_soc.c | 7 ++- .../net/ethernet/stmicro/stmmac/stmmac_main.c | 4 +- .../net/ethernet/xilinx/xilinx_axienet_main.c | 7 ++- drivers/net/phy/phylink.c | 9 ++- include/linux/phylink.h | 57 ++++++++++++++----- net/dsa/port.c | 4 +- 11 files changed, 105 insertions(+), 41 deletions(-)Documentation/networking/sfp-phylink.rst +6
-3@@ -74,10 +74,13 @@ phylib to the sfp/phylink support. Please send patches to improve this documentation. 1. Optionally split the network driver's phylib update function into - three parts dealing with link-down, link-up and reconfiguring the - MAC settings. This can be done as a separate preparation commit. + two parts dealing with link-down and link-up. This can be done as + a separate preparation commit. - An example of this preparation can be found in git commit fc548b991fb0. + An older example of this preparation can be found in git commit + fc548b991fb0, although this was splitting into three parts; the + link-up part now includes configuring the MAC for the link settings. + Please see :c:func:`mac_link_up` for more information on this. 2. Replace::@@ -207,6 +210,14 @@ this documentation. using. This is particularly important for in-band negotiation methods such as 1000base-X and SGMII.Quoted 14 lines
Documentation/networking/sfp-phylink.rst this documentation. · line 223
Quoted 11 lines
Documentation/networking/sfp-phylink.rst +8
-0+ The :c:func:`mac_link_up` method is used to inform the MAC that the + link has come up. The call includes the negotiation mode and interface + for reference only. The finalised link parameters are also supplied + (speed, duplex and flow control/pause enablement settings) which + should be used to configure the MAC when the MAC and PCS are not + tightly integrated, or when the settings are not coming from in-band + negotiation. + The :c:func:`mac_config` method is used to update the MAC with the requested state, and must avoid unnecessarily taking the link down when making changes to the MAC configuration. This means theJust to make sure I understand the changes:
- A MAC with no PCS can be configured in either .mac_config or .mac_link_upI would much prefer
mac_link_upto be used for setting the speed, duplex and pause modes for future-proofing in all cases except for the case where these parameters are automatically updated in the MAC from its associated PCS.
mac_link_upmust not be used to configure the AN mode or interface mode; these must be configured inmac_config().Quoted 3 lines
- A MAC that needs to be manually reconfigured to the link mode negotiated by its PCS needs to have the PCS configured in .mac_config and the MAC in .mac_link_up
I do have further changes that split the PCS ops from the MAC ops, so what is in this series is not the full story yet - some of the further patches can be found in my "phy" branch and "cex7" branches where I add support to dpaa2 for automatically switching between SGMII and 1000BASE-X. dpaa2 is one of these split PCS/MAC setups, but with the extra complication that there's a firmware layer between the PCS and MAC.
However, this series is the first stand-alone step along the road to supporting split PCS/MAC setups in a sane manner.
I discussed with Andrew Lunn how much to send, and the conclusion was to make the changes in a number of small patch series, as large patch series tend not to get reviewed. My experience with _this_ series is that even this is very difficult to get feedback for, so adding any additional patches will just make that worse.
Quoted 4 lines
- A MAC with PCS where the MAC follows the PCS negotiation automatically in hardware is basically equivalent with a MAC with no PCS, and therefore can be configured in either .mac_config or .mac_link_up
In this case,
mac_link_updoesn't do anything with the speed/duplex/ pause stuff when those are automatically passed from the PCS.I'm afraid that sentence contains a subtlety that's going to get
people: it is not clear cut because of the different natures of the various links.In 1000BASE-X, speed is fixed at 1G, and the PCS autonegotiates the duplex and pause with the remote end. For mvneta (an example of a combined PCS/MAC implementation) operating in-band:
- Inmac_config():
- configures for 1000BASE-X interface type with in-band AN.
- configures fixed 1G.
- As mvneta only supports full duplex, we disable duplex negotiation and force full duplex.
- Only symmetric pause is supported, and we set the symmetric pause advertisement appropriately, with pause negotiation enabled.
- Inmac_link_up():
- merely allow the device to transmit and receive.The MAC will be forced to 1G, full duplex, and will automatically be configured by the PCS for pause support depending on the hardware based pause resolution.
The situation is different for SGMII operating in-band:
- Inmac_config():
- configures for SGMII interface type with in-band AN.
- configures speed and duplex negotiation.
- disables pause negotiation; SGMII has no support for this.
- Inmac_link_up():
- enables or disables pause frames depending on the tx_pause/rx_pauseflags, since this is not available from the MAC.
- allow the device to transmit and receive.
Quoted 7 lines
If we aren't operating in in-band mode, then:
- Inmac_config():
- configures for the interface type without in-band AN.
- disables speed, duplex and pause negotiation.
- Inmac_link_up():
- sets the speed, duplex and pause frames depending on the passed parameters.
But there shouldn't be any requirement for this to be configured at this step and not earlier?
Quoted 11 lines
- allow the device to transmit and receive.
Please see patch 7 of this series which implements this for mvneta.
So, there is a split between what
mac_config()should be doing and whatmac_link_up()should be doing; this is why I've said in the documentation that the "mode" and "interface" are for reference only inmac_link_up()-mac_link_up()can use these to decide _how_ to program the resolved parameters, but should _not_ use them to determine the link configuration (such as changing the interface between SGMII and 1000BASE-X - that is the responsibility ofmac_config().)
Does any driver currently make any use of the phy_interface_t argument provided as reference in .mac_link_up?
Quoted 7 lines
I hope that helps clarify the situation.
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Ok, so basically what is known early, as well as whatever is needed for the in-band AN preparation, is configured in .mac_config and what is known late is configured in .mac_link_up.
Except that you would like to slowly move everything MAC-related to .mac_link_up, and everything PCS-related to .mac_config, presumably in an effort to convert .mac_config to .pcs_config and .mac_link_up to .mac_config. I don't actually know what other patches you have in the cex7 branch you mentioned. Please consider that people don't necessarily bookmark your git trees. I've spent some good 10 minutes searching for the "cex7" and "phy" keywords in emails received from you, and haven't found the git links.
Regards,
-Vladimir
Russell King - ARM Linux admin
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 163 lines · Vladimir Oltean
Quoted 138 lines · Russell King - ARM Linux admin
Quoted 53 lines · Vladimir Oltean
Hi Russell,
Quoted 50 lines · Russell King
Propagate the resolved link parameters via the
mac_link_up()call for MACs that do not automatically track their PCS state. We propagate the link parameters via function arguments so that inappropriate members of structphylink_link_statecan't be accessed, and creating a new structure just for this adds needless complexity to the API.Tested-by Andre Przywara <[email protected]>Tested-by Alexandre Belloni <[email protected]>Tested-by Vladimir Oltean <[email protected]>Signed-off-by Russell King <[email protected]>--- Documentation/networking/sfp-phylink.rst | 17 +++++- drivers/net/ethernet/cadence/macb_main.c | 7 ++- .../net/ethernet/freescale/dpaa2/dpaa2-mac.c | 7 ++- drivers/net/ethernet/marvell/mvneta.c | 8 ++- .../net/ethernet/marvell/mvpp2/mvpp2_main.c | 19 +++++-- drivers/net/ethernet/mediatek/mtk_eth_soc.c | 7 ++- .../net/ethernet/stmicro/stmmac/stmmac_main.c | 4 +- .../net/ethernet/xilinx/xilinx_axienet_main.c | 7 ++- drivers/net/phy/phylink.c | 9 ++- include/linux/phylink.h | 57 ++++++++++++++----- net/dsa/port.c | 4 +- 11 files changed, 105 insertions(+), 41 deletions(-)Documentation/networking/sfp-phylink.rst +6
-3@@ -74,10 +74,13 @@ phylib to the sfp/phylink support. Please send patches to improve this documentation. 1. Optionally split the network driver's phylib update function into - three parts dealing with link-down, link-up and reconfiguring the - MAC settings. This can be done as a separate preparation commit. + two parts dealing with link-down and link-up. This can be done as + a separate preparation commit. - An example of this preparation can be found in git commit fc548b991fb0. + An older example of this preparation can be found in git commit + fc548b991fb0, although this was splitting into three parts; the + link-up part now includes configuring the MAC for the link settings. + Please see :c:func:`mac_link_up` for more information on this. 2. Replace::@@ -207,6 +210,14 @@ this documentation. using. This is particularly important for in-band negotiation methods such as 1000base-X and SGMII.Quoted 14 lines
Documentation/networking/sfp-phylink.rst this documentation. · line 223
Quoted 11 lines
Documentation/networking/sfp-phylink.rst +8
-0+ The :c:func:`mac_link_up` method is used to inform the MAC that the + link has come up. The call includes the negotiation mode and interface + for reference only. The finalised link parameters are also supplied + (speed, duplex and flow control/pause enablement settings) which + should be used to configure the MAC when the MAC and PCS are not + tightly integrated, or when the settings are not coming from in-band + negotiation. + The :c:func:`mac_config` method is used to update the MAC with the requested state, and must avoid unnecessarily taking the link down when making changes to the MAC configuration. This means theJust to make sure I understand the changes:
- A MAC with no PCS can be configured in either .mac_config or .mac_link_upI would much prefer
mac_link_upto be used for setting the speed, duplex and pause modes for future-proofing in all cases except for the case where these parameters are automatically updated in the MAC from its associated PCS.
mac_link_upmust not be used to configure the AN mode or interface mode; these must be configured inmac_config().Quoted 3 lines
- A MAC that needs to be manually reconfigured to the link mode negotiated by its PCS needs to have the PCS configured in .mac_config and the MAC in .mac_link_up
I do have further changes that split the PCS ops from the MAC ops, so what is in this series is not the full story yet - some of the further patches can be found in my "phy" branch and "cex7" branches where I add support to dpaa2 for automatically switching between SGMII and 1000BASE-X. dpaa2 is one of these split PCS/MAC setups, but with the extra complication that there's a firmware layer between the PCS and MAC.
However, this series is the first stand-alone step along the road to supporting split PCS/MAC setups in a sane manner.
I discussed with Andrew Lunn how much to send, and the conclusion was to make the changes in a number of small patch series, as large patch series tend not to get reviewed. My experience with _this_ series is that even this is very difficult to get feedback for, so adding any additional patches will just make that worse.
Quoted 4 lines
- A MAC with PCS where the MAC follows the PCS negotiation automatically in hardware is basically equivalent with a MAC with no PCS, and therefore can be configured in either .mac_config or .mac_link_up
In this case,
mac_link_updoesn't do anything with the speed/duplex/ pause stuff when those are automatically passed from the PCS.I'm afraid that sentence contains a subtlety that's going to get
people: it is not clear cut because of the different natures of the various links.In 1000BASE-X, speed is fixed at 1G, and the PCS autonegotiates the duplex and pause with the remote end. For mvneta (an example of a combined PCS/MAC implementation) operating in-band:
- Inmac_config():
- configures for 1000BASE-X interface type with in-band AN.
- configures fixed 1G.
- As mvneta only supports full duplex, we disable duplex negotiation and force full duplex.
- Only symmetric pause is supported, and we set the symmetric pause advertisement appropriately, with pause negotiation enabled.
- Inmac_link_up():
- merely allow the device to transmit and receive.The MAC will be forced to 1G, full duplex, and will automatically be configured by the PCS for pause support depending on the hardware based pause resolution.
The situation is different for SGMII operating in-band:
- Inmac_config():
- configures for SGMII interface type with in-band AN.
- configures speed and duplex negotiation.
- disables pause negotiation; SGMII has no support for this.
- Inmac_link_up():
- enables or disables pause frames depending on the tx_pause/rx_pauseflags, since this is not available from the MAC.
- allow the device to transmit and receive.Quoted 7 lines
If we aren't operating in in-band mode, then:
- Inmac_config():
- configures for the interface type without in-band AN.
- disables speed, duplex and pause negotiation.
- Inmac_link_up():
- sets the speed, duplex and pause frames depending on the passed parameters.But there shouldn't be any requirement for this to be configured at this step and not earlier?
Quoted 11 lines
- allow the device to transmit and receive.
Please see patch 7 of this series which implements this for mvneta.
So, there is a split between what
mac_config()should be doing and whatmac_link_up()should be doing; this is why I've said in the documentation that the "mode" and "interface" are for reference only inmac_link_up()-mac_link_up()can use these to decide _how_ to program the resolved parameters, but should _not_ use them to determine the link configuration (such as changing the interface between SGMII and 1000BASE-X - that is the responsibility ofmac_config().)
Quoted 2 lines
Does any driver currently make any use of the
phy_interface_targument provided as reference in .mac_link_up?
Yes. mvpp2 uses it to decide whether to configure the 10/100/1G MAC or the 10G XLG MAC, for example - there are drivers that need to configure different blocks for the link-up event depending on the interface mode. Rather than having drivers store the interface mode internally, copying what phylink is doing, we provide that to the network driver so it can make its own decisions without that additional complexity.
Quoted 10 lines
Quoted 6 lines
I hope that helps clarify the situation.
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps upOk, so basically what is known early, as well as whatever is needed for the in-band AN preparation, is configured in .mac_config and what is known late is configured in .mac_link_up.
I envision a time when the speed/duplex/pause settings will not be passed to mac_config.
Quoted 8 lines
Except that you would like to slowly move everything MAC-related to .mac_link_up, and everything PCS-related to .mac_config, presumably in an effort to convert .mac_config to .pcs_config and .mac_link_up to .mac_config. I don't actually know what other patches you have in the cex7 branch you mentioned. Please consider that people don't necessarily bookmark your git trees. I've spent some good 10 minutes searching for the "cex7" and "phy" keywords in emails received from you, and haven't found the git links.
http://git.armlinux.org.uk/cgit/linux-arm.git/
I'm afraid that it's rather scattered amongst quite a few branches at the moment, because of their dependencies on other stuff in my tree. For dpaa2, see:
http://git.armlinux.org.uk/cgit/linux-arm.git/log/?h=cex7
specifically:
"dpaa2-mac: add PCS support"
"net: phylink: add c45 pcs helpers"
"net: phylink: helpers for 802.3 clause 37/SGMII register sets" "net: phylink: add pcs operations"
"net: phylink: rename 'ops' to 'mac_ops'"
"net: mii: add linkmode_adv_to_mii_adv_x()"
"net: mii: convert mii_lpa_to_ethtool_lpa_x() to linkmode variant"
dpaa2 is complicated by the firmware, and that we can't switch the interface mode between (SGMII,1000base-X) and 10G.
If the firmware is in "DPMAC_LINK_TYPE_PHY" mode, it expects to be told the current link parameters via the dpmac_set_link_state() call - it isn't clear whether that needs to be called for other modes with the up/down state (firmware API documentation is poor.)
Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Vladimir Oltean
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 10 lines · Russell King - ARM Linux admin
dpaa2 is complicated by the firmware, and that we can't switch the interface mode between (SGMII,1000base-X) and 10G.
If the firmware is in "
DPMAC_LINK_TYPE_PHY" mode, it expects to be told the current link parameters via thedpmac_set_link_state()call - it isn't clear whether that needs to be called for other modes with the up/down state (firmware API documentation is poor.)
With PCS control in Linux, I am pretty sure that you don't want anything other than DPMAC_LINK_TYPE_PHY anyway.
Basically in DPMAC_LINK_TYPE_FIXED, the MC firmware is in control of the PCS and polls its link state to emit link notifications to objects connected to the DPMAC. So Linux control of PCS would class with firmware control of the PCS, leading to undesirable side-effects to say the least.
Quoted 4 lines
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Regards,
-Vladimir
Vladimir Oltean
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 14 lines · Vladimir Oltean
Quoted 10 lines · Russell King - ARM Linux admin
dpaa2 is complicated by the firmware, and that we can't switch the interface mode between (SGMII,1000base-X) and 10G.
If the firmware is in "
DPMAC_LINK_TYPE_PHY" mode, it expects to be told the current link parameters via thedpmac_set_link_state()call - it isn't clear whether that needs to be called for other modes with the up/down state (firmware API documentation is poor.)
Quoted 5 lines
With PCS control in Linux, I am pretty sure that you don't want anything other than
DPMAC_LINK_TYPE_PHYanyway.
Basically inDPMAC_LINK_TYPE_FIXED, the MC firmware is in control of the PCS and polls its link state to emit link notifications to objects connected to the DPMAC. So Linux control of PCS would class with
s/class/clash/
Quoted 10 lines
firmware control of the PCS, leading to undesirable side-effects to say the least.
Quoted 4 lines
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps upRegards,
-Vladimir
Russell King - ARM Linux admin
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 15 lines · Vladimir Oltean
Quoted 14 lines · Vladimir Oltean
Quoted 10 lines · Russell King - ARM Linux admin
dpaa2 is complicated by the firmware, and that we can't switch the interface mode between (SGMII,1000base-X) and 10G.
If the firmware is in "
DPMAC_LINK_TYPE_PHY" mode, it expects to be told the current link parameters via thedpmac_set_link_state()call - it isn't clear whether that needs to be called for other modes with the up/down state (firmware API documentation is poor.)
Quoted 10 lines
Quoted 5 lines
With PCS control in Linux, I am pretty sure that you don't want anything other than
DPMAC_LINK_TYPE_PHYanyway.
Basically inDPMAC_LINK_TYPE_FIXED, the MC firmware is in control of the PCS and polls its link state to emit link notifications to objects connected to the DPMAC. So Linux control of PCS would class withs/class/clash/
Quoted 2 lines
firmware control of the PCS, leading to undesirable side-effects to say the least.
Yes, which is why we only register the PCS with phylink if we're in DPMAC_LINK_TYPE_PHY mode.
Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
Ioana Ciornei
maintainerRE: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
In reply to Vladimir OlteanQuoted 18 lines
Subject: Re: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via
mac_link_up()Quoted 14 lines · Vladimir Oltean
Quoted 10 lines · Russell King - ARM Linux admin
dpaa2 is complicated by the firmware, and that we can't switch the interface mode between (SGMII,1000base-X) and 10G.
If the firmware is in "
DPMAC_LINK_TYPE_PHY" mode, it expects to be told the current link parameters via thedpmac_set_link_state()call
- it isn't clear whether that needs to be called for other modes with the up/down state (firmware API documentation is poor.)
Quoted 10 lines
Quoted 5 lines
With PCS control in Linux, I am pretty sure that you don't want anything other than
DPMAC_LINK_TYPE_PHYanyway.
Basically inDPMAC_LINK_TYPE_FIXED, the MC firmware is in control of the PCS and polls its link state to emit link notifications to objects connected to the DPMAC. So Linux control of PCS would class withs/class/clash/
Quoted 2 lines
firmware control of the PCS, leading to undesirable side-effects to say the least.
If the DPMAC object is in DPMAC_LINK_TYPE_FIXED, the dpaa2-eth in fact does not even connect to a phy so all the phylink interaction is not happening. As Vladimir said, in this case it's the MC firmware's job to poll the PCS and notify any connected objects of a link change.
Quoted 3 lines
Quoted 3 lines
Regards,
-Vladimir
Russell King - ARM Linux admin
maintainerRe: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via mac_link_up()
Quoted 26 lines · Ioana Ciornei
Quoted 26 lines
Subject: Re: [PATCH net-next v2 1/8] net: phylink: propagate resolved link config via
mac_link_up()Quoted 14 lines · Vladimir Oltean
Quoted 10 lines · Russell King - ARM Linux admin
dpaa2 is complicated by the firmware, and that we can't switch the interface mode between (SGMII,1000base-X) and 10G.
If the firmware is in "
DPMAC_LINK_TYPE_PHY" mode, it expects to be told the current link parameters via thedpmac_set_link_state()call
- it isn't clear whether that needs to be called for other modes with the up/down state (firmware API documentation is poor.)Quoted 5 lines
With PCS control in Linux, I am pretty sure that you don't want anything other than
DPMAC_LINK_TYPE_PHYanyway.
Basically inDPMAC_LINK_TYPE_FIXED, the MC firmware is in control of the PCS and polls its link state to emit link notifications to objects connected to the DPMAC. So Linux control of PCS would class withs/class/clash/
Quoted 8 lines
Quoted 2 lines
Quoted 2 lines
firmware control of the PCS, leading to undesirable side-effects to say the least.
If the DPMAC object is in
DPMAC_LINK_TYPE_FIXED, the dpaa2-eth in fact does not even connect to a phy so all the phylink interaction is not happening. As Vladimir said, in this case it's the MC firmware's job to poll the PCS and notify any connected objects of a link change.
Please see the patches I've referred Vladimir to, specifically this:
if (attr.link_type == DPMAC_LINK_TYPE_PHY) {
/* FIXME: how do we know whether this DPMAC has a PCS? */
err = dpaa2_pcs_create(mac, attr.id);
if (err)
goto err_phylink_destroy;
phylink_add_pcs(mac->phylink, &dpaa2_pcs_phylink_ops);
}Hence, if we are not in DPMAC_LINK_TYPE_PHY, then we never talk to the PCS, thereby satisfying the requirements of the firmware not to touch the PCS if it's in FIXED mode.
It seems this is becoming a storm in a tea cup. Please stop.
Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up
PATCH v2 2/8 net: dsa: propagate resolved link config via mac_link_up()
In reply to Russell King - ARM Linux adminPropagate the resolved link configuration down via DSA's phylink_mac_link_up() operation to allow split PCS/MAC to work.
--- drivers/net/dsa/b53/b53_common.c | 4 +++- drivers/net/dsa/b53/b53_priv.h | 4 +++- drivers/net/dsa/bcm_sf2.c | 4 +++- drivers/net/dsa/lantiq_gswip.c | 4 +++- drivers/net/dsa/mt7530.c | 4 +++- drivers/net/dsa/mv88e6xxx/chip.c | 4 +++- drivers/net/dsa/ocelot/felix.c | 4 +++- drivers/net/dsa/qca/ar9331.c | 4 +++- drivers/net/dsa/sja1105/sja1105_main.c | 4 +++- include/net/dsa.h | 4 +++- net/dsa/port.c | 3 ++- 11 files changed, 32 insertions(+), 11 deletions(-)
drivers/net/dsa/b53/b53_common.c +3 -1
@@ -1289,7 +1289,9 @@ EXPORT_SYMBOL(b53_phylink_mac_link_down);void b53_phylink_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct b53_device *dev = ds->priv;
drivers/net/dsa/b53/b53_priv.h +3 -1
@@ -338,7 +338,9 @@ void b53_phylink_mac_link_down(struct dsa_switch *ds, int port,void b53_phylink_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev); +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause);int b53_vlan_filtering (struct dsa_switch *ds,int port,bool vlan_filtering);int b53_vlan_prepare (struct dsa_switch *ds,int port,const struct switchdev_obj_port_vlan *vlan);
drivers/net/dsa/bcm_sf2.c +3 -1
@@ -652,7 +652,9 @@ static void bcm_sf2_sw_mac_link_down(struct dsa_switch *ds, int port,static void bcm_sf2_sw_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct bcm_sf2_priv *priv =bcm_sf2_to_priv (ds);struct ethtool_eee *p = &priv->dev->ports[port].eee;
drivers/net/dsa/lantiq_gswip.c +3 -1
@@ -1517,7 +1517,9 @@ static void gswip_phylink_mac_link_down(struct dsa_switch *ds, int port,static void gswip_phylink_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct gswip_priv *priv = ds->priv;
drivers/net/dsa/mt7530.c +3 -1
@@ -1482,7 +1482,9 @@ static void mt7530_phylink_mac_link_down(struct dsa_switch *ds, int port,static void mt7530_phylink_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct mt7530_priv *priv = ds->priv;
drivers/net/dsa/mv88e6xxx/chip.c +3 -1
@@ -655,7 +655,9 @@ static void mv88e6xxx_mac_link_down(struct dsa_switch *ds, int port,static void mv88e6xxx_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {if (mode ==MLO_AN_FIXED )mv88e6xxx_mac_link_force (ds, port,LINK_FORCED_UP );
drivers/net/dsa/ocelot/felix.c +3 -1
@@ -263,7 +263,9 @@ static void felix_phylink_mac_link_down(struct dsa_switch *ds, int port,static void felix_phylink_mac_link_up (struct dsa_switch *ds,int port,unsigned int link_an_mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct ocelot *ocelot = ds->priv;struct ocelot_port *ocelot_port = ocelot->ports[port];
drivers/net/dsa/qca/ar9331.c +3 -1
@@ -458,7 +458,9 @@ static void ar9331_sw_phylink_mac_link_down(struct dsa_switch *ds, int port,static void ar9331_sw_phylink_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct ar9331_sw_priv *priv = (struct ar9331_sw_priv *)ds->priv;struct regmap *regmap = priv->regmap;
drivers/net/dsa/sja1105/sja1105_main.c +3 -1
@@ -786,7 +786,9 @@ static void sja1105_mac_link_down(struct dsa_switch *ds, int port,static void sja1105_mac_link_up (struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev) +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause) {sja1105_inhibit_tx (ds->priv,BIT (port), false); }
include/net/dsa.h +3 -1
@@ -420,7 +420,9 @@ struct dsa_switch_ops {void (*phylink_mac_link_up)(struct dsa_switch *ds,int port,unsigned int mode,phy_interface_t interface, -struct phy_device *phydev); +struct phy_device *phydev, +int speed,int duplex, +bool tx_pause,bool rx_pause);void (*phylink_fixed_state)(struct dsa_switch *ds,int port,struct phylink_link_state *state);/*
net/dsa/port.c +2 -1
@@ -504,7 +504,8 @@ static void dsa_port_phylink_mac_link_up(struct phylink_config *config,return ; } - ds->ops->phylink_mac_link_up (ds, dp->index, mode, interface, phydev); + ds->ops->phylink_mac_link_up (ds, dp->index, mode, interface, phydev, + speed, duplex, tx_pause, rx_pause); }const struct phylink_mac_ops dsa_port_phylink_mac_ops = {
Signature · 1 line
2.20.1
PATCH v2 3/8 net: mv88e6xxx: use resolved link config in mac_link_up()
In reply to Russell King - ARM Linux adminUse the resolved link configuration to set the MAC configuration when mac_link_up() for non-internal-PHY ports.
--- drivers/net/dsa/mv88e6xxx/chip.c | 75 +++++++++++++++++++++++++------- 1 file changed, 59 insertions(+), 16 deletions(-)
drivers/net/dsa/mv88e6xxx/chip.c +59 -16
@@ -632,25 +632,30 @@ static void mv88e6xxx_mac_config(struct dsa_switch *ds, int port,dev_err (ds->dev,"p%d: failed to configure MAC\n" , port); } -static void mv88e6xxx_mac_link_force (struct dsa_switch *ds,int port,int link) +static void mv88e6xxx_mac_link_down (struct dsa_switch *ds,int port, +unsigned int mode, +phy_interface_t interface) {struct mv88e6xxx_chip *chip = ds->priv; -int err; +const struct mv88e6xxx_ops *ops; +int err =0 ; -mv88e6xxx_reg_lock (chip); - err = chip->info->ops->port_set_link (chip, port, link); -mv88e6xxx_reg_unlock (chip); + ops = chip->info->ops; -if (err) -dev_err (chip->dev,"p%d: failed to force MAC link\n" , port); -} +/* Internal PHYs propagate their configuration directly to the MAC. +* External PHYs depend on whether the PPU is enabled for this port. +* FIXME: we should be using the PPU enable state here. What about +* an automedia port? +*/ +if (!mv88e6xxx_phy_is_internal (ds, port) && ops->port_set_link) { +mv88e6xxx_reg_lock (chip); + err = ops->port_set_link (chip, port,LINK_FORCED_DOWN ); +mv88e6xxx_reg_unlock (chip); -static void mv88e6xxx_mac_link_down (struct dsa_switch *ds,int port, -unsigned int mode, -phy_interface_t interface) -{ -if (mode ==MLO_AN_FIXED ) -mv88e6xxx_mac_link_force (ds, port,LINK_FORCED_DOWN ); +if (err) +dev_err (chip->dev, +"p%d: failed to force MAC link down\n" , port); + } } static void mv88e6xxx_mac_link_up(struct dsa_switch *ds, int port,
@@ -659,8 +664,46 @@ static void mv88e6xxx_mac_link_up(struct dsa_switch *ds, int port, int speed, int duplex, bool tx_pause, bool rx_pause) { - if (mode == MLO_AN_FIXED) - mv88e6xxx_mac_link_force(ds, port, LINK_FORCED_UP); + struct mv88e6xxx_chip *chip = ds->priv; +const struct mv88e6xxx_ops *ops; +int err =0 ; + + ops = chip->info->ops; + +/* Internal PHYs propagate their configuration directly to the MAC. +* External PHYs depend on whether the PPU is enabled for this port. +* FIXME: we should be using the PPU enable state here. What about +* an automedia port? +*/ +if (!mv88e6xxx_phy_is_internal (ds, port)) { +mv88e6xxx_reg_lock (chip); +/* FIXME: for an automedia port, should we force the link +* down here - what if the link comes up due to "other" media +* while we're bringing the port up, how is the exclusivity +* handled in the Marvell hardware? E.g. port 4 on 88E6532 +* shared between internal PHY and Serdes. +*/ +if (ops->port_set_speed) { + err = ops->port_set_speed (chip, port, speed); +if (err && err != -EOPNOTSUPP ) + goto error; + } + +if (ops->port_set_duplex) { + err = ops->port_set_duplex (chip, port, duplex); +if (err && err != -EOPNOTSUPP ) + goto error; + } + +if (ops->port_set_link) + err = ops->port_set_link (chip, port,LINK_FORCED_UP ); +error: +mv88e6xxx_reg_unlock (chip); + +if (err && err != -EOPNOTSUPP ) +dev_err (ds->dev, +"p%d: failed to configure MAC link up\n" , port); + } }static int mv88e6xxx_stats_snapshot (struct mv88e6xxx_chip *chip,int port)
Signature · 1 line
2.20.1
PATCH v2 4/8 net: axienet: use resolved link config in mac_link_up()
In reply to Russell King - ARM Linux adminConvert the Xilinx AXI ethernet driver to use the finalised link parameters in mac_link_up() rather than the parameters in mac_config().
--- .../net/ethernet/xilinx/xilinx_axienet_main.c | 38 +++++++++---------- 1 file changed, 19 insertions(+), 19 deletions(-)
drivers/net/ethernet/xilinx/xilinx_axienet_main.c +19 -19
@@ -1440,6 +1440,22 @@ static void axienet_mac_an_restart(struct phylink_config *config)static void axienet_mac_config (struct phylink_config *config,unsigned int mode,const struct phylink_link_state *state) +{ +/* nothing meaningful to do */ +} + +static void axienet_mac_link_down (struct phylink_config *config, +unsigned int mode, +phy_interface_t interface) +{ +/* nothing meaningful to do */ +} + +static void axienet_mac_link_up (struct phylink_config *config, +struct phy_device *phy, +unsigned int mode,phy_interface_t interface, +int speed,int duplex, +bool tx_pause,bool rx_pause) {struct net_device *ndev =to_net_dev (config->dev);struct axienet_local *lp =netdev_priv (ndev);
@@ -1448,7 +1464,7 @@ static void axienet_mac_config(struct phylink_config *config, unsigned int mode, emmc_reg =axienet_ior (lp,XAE_EMMC_OFFSET ); emmc_reg &= ~XAE_EMMC_LINKSPEED_MASK ; -switch (state->speed) { +switch (speed) {case SPEED_1000 : emmc_reg |=XAE_EMMC_LINKSPD_1000 ;break ;
@@ -1467,33 +1483,17 @@ static void axienet_mac_config(struct phylink_config *config, unsigned int mode,axienet_iow (lp,XAE_EMMC_OFFSET , emmc_reg); fcc_reg =axienet_ior (lp,XAE_FCC_OFFSET ); -if (state->pause &MLO_PAUSE_TX ) +if (tx_pause) fcc_reg |=XAE_FCC_FCTX_MASK ;else fcc_reg &= ~XAE_FCC_FCTX_MASK ; -if (state->pause &MLO_PAUSE_RX ) +if (rx_pause) fcc_reg |=XAE_FCC_FCRX_MASK ;else fcc_reg &= ~XAE_FCC_FCRX_MASK ;axienet_iow (lp,XAE_FCC_OFFSET , fcc_reg); } -static void axienet_mac_link_down (struct phylink_config *config, -unsigned int mode, -phy_interface_t interface) -{ -/* nothing meaningful to do */ -} - -static void axienet_mac_link_up (struct phylink_config *config, -struct phy_device *phy, -unsigned int mode,phy_interface_t interface, -int speed,int duplex, -bool tx_pause,bool rx_pause) -{ -/* nothing meaningful to do */ -} -static const struct phylink_mac_ops axienet_phylink_ops = { .validate = axienet_validate, .mac_pcs_get_state = axienet_mac_pcs_get_state,
Signature · 1 line
2.20.1
PATCH v2 5/8 net: dpaa2-mac: use resolved link config in mac_link_up()
In reply to Russell King - ARM Linux adminConvert the DPAA2 ethernet driver to use the finalised link parameters in mac_link_up() rather than the parameters in mac_config(), which are more suited to the needs of the DPAA2 MC firmware than those available via mac_config().
--- .../net/ethernet/freescale/dpaa2/dpaa2-mac.c | 54 +++++++++++-------- .../net/ethernet/freescale/dpaa2/dpaa2-mac.h | 1 + 2 files changed, 33 insertions(+), 22 deletions(-)
drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.c +32 -22
@@ -123,35 +123,16 @@ static void dpaa2_mac_config(struct phylink_config *config, unsigned int mode,struct dpmac_link_state *dpmac_state = &mac->state;int err; -if (state->speed !=SPEED_UNKNOWN ) - dpmac_state->rate = state->speed; - -if (state->duplex !=DUPLEX_UNKNOWN ) { -if (!state->duplex) - dpmac_state->options |=DPMAC_LINK_OPT_HALF_DUPLEX ; -else - dpmac_state->options &= ~DPMAC_LINK_OPT_HALF_DUPLEX ; - } -if (state->an_enabled) dpmac_state->options |=DPMAC_LINK_OPT_AUTONEG ;else dpmac_state->options &= ~DPMAC_LINK_OPT_AUTONEG ; -if (state->pause &MLO_PAUSE_RX ) - dpmac_state->options |=DPMAC_LINK_OPT_PAUSE ; -else - dpmac_state->options &= ~DPMAC_LINK_OPT_PAUSE ; - -if (!!(state->pause &MLO_PAUSE_RX ) ^ !!(state->pause &MLO_PAUSE_TX )) - dpmac_state->options |=DPMAC_LINK_OPT_ASYM_PAUSE ; -else - dpmac_state->options &= ~DPMAC_LINK_OPT_ASYM_PAUSE ; - err =dpmac_set_link_state (mac->mc_io,0 , mac->mc_dev->mc_handle, dpmac_state);if (err) -netdev_err (mac->net_dev,"dpmac_set_link_state() = %d\n" , err); +netdev_err (mac->net_dev,"%s: dpmac_set_link_state() = %d\n" , + __func__, err); } static void dpaa2_mac_link_up(struct phylink_config *config,
@@ -165,10 +146,37 @@ static void dpaa2_mac_link_up(struct phylink_config *config,int err; dpmac_state->up =1 ; + +if (mac->if_link_type ==DPMAC_LINK_TYPE_PHY ) { +/* If the DPMAC is configured for PHY mode, we need +* to pass the link parameters to the MC firmware. +*/ + dpmac_state->rate = speed; + +if (duplex ==DUPLEX_HALF ) + dpmac_state->options |=DPMAC_LINK_OPT_HALF_DUPLEX ; +else if (duplex ==DUPLEX_FULL ) + dpmac_state->options &= ~DPMAC_LINK_OPT_HALF_DUPLEX ; + +/* This is lossy; the firmware really should take the pause +* enablement status rather than pause/asym pause status. +*/ +if (rx_pause) + dpmac_state->options |=DPMAC_LINK_OPT_PAUSE ; +else + dpmac_state->options &= ~DPMAC_LINK_OPT_PAUSE ; + +if (rx_pause ^ tx_pause) + dpmac_state->options |=DPMAC_LINK_OPT_ASYM_PAUSE ; +else + dpmac_state->options &= ~DPMAC_LINK_OPT_ASYM_PAUSE ; + } + err =dpmac_set_link_state (mac->mc_io,0 , mac->mc_dev->mc_handle, dpmac_state);if (err) -netdev_err (mac->net_dev,"dpmac_set_link_state() = %d\n" , err); +netdev_err (mac->net_dev,"%s: dpmac_set_link_state() = %d\n" , + __func__, err); } static void dpaa2_mac_link_down(struct phylink_config *config,
@@ -241,6 +249,8 @@ int dpaa2_mac_connect(struct dpaa2_mac *mac) goto err_close_dpmac; } + mac->if_link_type = attr.link_type; + dpmac_node =dpaa2_mac_get_node (attr.id);if (!dpmac_node) {netdev_err (net_dev,"No dpmac@%d node found.\n" , attr.id);
drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.h +1 -0
@@ -20,6 +20,7 @@ struct dpaa2_mac {struct phylink_config phylink_config;struct phylink *phylink;phy_interface_t if_mode; +enum dpmac_link_type if_link_type; }; bool dpaa2_mac_is_type_fixed(struct fsl_mc_device *dpmac_dev,
Signature · 1 line
2.20.1
PATCH v2 6/8 net: macb: use resolved link config in mac_link_up()
In reply to Russell King - ARM Linux adminConvert the macb ethernet driver to use the finalised link parameters in mac_link_up() rather than the parameters in mac_config().
--- drivers/net/ethernet/cadence/macb.h | 1 - drivers/net/ethernet/cadence/macb_main.c | 50 ++++++++++++++---------- 2 files changed, 29 insertions(+), 22 deletions(-)
drivers/net/ethernet/cadence/macb.h +0 -1
@@ -1200,7 +1200,6 @@ struct macb {unsigned int dma_burst_length;phy_interface_t phy_interface; -int speed;/* AT91RM9200 transmit */ struct sk_buff *skb;/* holds skb until xmit interrupt completes */
drivers/net/ethernet/cadence/macb_main.c +29 -21
90 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v2 7/8 net: mvneta: use resolved link config in mac_link_up()
In reply to Russell King - ARM Linux adminConvert the Marvell mvneta ethernet driver to use the finalised link parameters in mac_link_up() rather than the parameters in mac_config().
--- drivers/net/ethernet/marvell/mvneta.c | 55 ++++++++++++++++++--------- 1 file changed, 38 insertions(+), 17 deletions(-)
drivers/net/ethernet/marvell/mvneta.c +38 -17
@@ -3830,13 +3830,9 @@ static void mvneta_mac_config(struct phylink_config *config, unsigned int mode, new_clk = gmac_clk & ~MVNETA_GMAC_1MS_CLOCK_ENABLE ; new_an = gmac_an & ~(MVNETA_GMAC_INBAND_AN_ENABLE |MVNETA_GMAC_INBAND_RESTART_AN | -MVNETA_GMAC_CONFIG_MII_SPEED | -MVNETA_GMAC_CONFIG_GMII_SPEED |MVNETA_GMAC_AN_SPEED_EN |MVNETA_GMAC_ADVERT_SYM_FLOW_CTRL | -MVNETA_GMAC_CONFIG_FLOW_CTRL |MVNETA_GMAC_AN_FLOW_CTRL_EN | -MVNETA_GMAC_CONFIG_FULL_DUPLEX |MVNETA_GMAC_AN_DUPLEX_EN );/* Even though it might look weird, when we're configured in
@@ -3851,24 +3847,20 @@ static void mvneta_mac_config(struct phylink_config *config, unsigned int mode,if (phylink_test (state->advertising, Pause)) new_an |=MVNETA_GMAC_ADVERT_SYM_FLOW_CTRL ; -if (state->pause &MLO_PAUSE_TXRX_MASK ) - new_an |=MVNETA_GMAC_CONFIG_FLOW_CTRL ;if (!phylink_autoneg_inband (mode)) { -/* Phy or fixed speed */ -if (state->duplex) - new_an |=MVNETA_GMAC_CONFIG_FULL_DUPLEX ; - -if (state->speed ==SPEED_1000 || state->speed ==SPEED_2500 ) - new_an |=MVNETA_GMAC_CONFIG_GMII_SPEED ; -else if (state->speed ==SPEED_100 ) - new_an |=MVNETA_GMAC_CONFIG_MII_SPEED ; +/* Phy or fixed speed - nothing to do, leave the +* configured speed, duplex and flow control as-is. +*/ }else if (state->interface ==PHY_INTERFACE_MODE_SGMII ) {/* SGMII mode receives the state from the PHY */ new_ctrl2 |=MVNETA_GMAC2_INBAND_AN_ENABLE ; new_clk |=MVNETA_GMAC_1MS_CLOCK_ENABLE ; new_an = (new_an & ~(MVNETA_GMAC_FORCE_LINK_DOWN | -MVNETA_GMAC_FORCE_LINK_PASS )) | +MVNETA_GMAC_FORCE_LINK_PASS | +MVNETA_GMAC_CONFIG_MII_SPEED | +MVNETA_GMAC_CONFIG_GMII_SPEED | +MVNETA_GMAC_CONFIG_FULL_DUPLEX )) |MVNETA_GMAC_INBAND_AN_ENABLE |MVNETA_GMAC_AN_SPEED_EN |MVNETA_GMAC_AN_DUPLEX_EN ;
@@ -3877,7 +3869,8 @@ static void mvneta_mac_config(struct phylink_config *config, unsigned int mode, new_ctrl0 |=MVNETA_GMAC0_PORT_1000BASE_X ; new_clk |=MVNETA_GMAC_1MS_CLOCK_ENABLE ; new_an = (new_an & ~(MVNETA_GMAC_FORCE_LINK_DOWN | -MVNETA_GMAC_FORCE_LINK_PASS )) | +MVNETA_GMAC_FORCE_LINK_PASS | +MVNETA_GMAC_CONFIG_MII_SPEED )) |MVNETA_GMAC_INBAND_AN_ENABLE | MVNETA_GMAC_CONFIG_GMII_SPEED |/* The MAC only supports FD mode */
@@ -3977,8 +3970,36 @@ static void mvneta_mac_link_up(struct phylink_config *config,if (!phylink_autoneg_inband (mode)) { val =mvreg_read (pp,MVNETA_GMAC_AUTONEG_CONFIG ); - val &= ~MVNETA_GMAC_FORCE_LINK_DOWN ; + val &= ~(MVNETA_GMAC_FORCE_LINK_DOWN | +MVNETA_GMAC_CONFIG_MII_SPEED | +MVNETA_GMAC_CONFIG_GMII_SPEED | +MVNETA_GMAC_CONFIG_FLOW_CTRL | +MVNETA_GMAC_CONFIG_FULL_DUPLEX ); val |=MVNETA_GMAC_FORCE_LINK_PASS ; + +if (speed ==SPEED_1000 || speed ==SPEED_2500 ) + val |=MVNETA_GMAC_CONFIG_GMII_SPEED ; +else if (speed ==SPEED_100 ) + val |=MVNETA_GMAC_CONFIG_MII_SPEED ; + +if (duplex ==DUPLEX_FULL ) + val |=MVNETA_GMAC_CONFIG_FULL_DUPLEX ; + +if (tx_pause || rx_pause) + val |=MVNETA_GMAC_CONFIG_FLOW_CTRL ; + +mvreg_write (pp,MVNETA_GMAC_AUTONEG_CONFIG , val); + }else { +/* When inband doesn't cover flow control or flow control is +* disabled, we need to manually configure it. This bit will +* only have effect if MVNETA_GMAC_AN_FLOW_CTRL_EN is unset. +*/ + val =mvreg_read (pp,MVNETA_GMAC_AUTONEG_CONFIG ); + val &= ~MVNETA_GMAC_CONFIG_FLOW_CTRL ; + +if (tx_pause || rx_pause) + val |=MVNETA_GMAC_CONFIG_FLOW_CTRL ; +mvreg_write (pp,MVNETA_GMAC_AUTONEG_CONFIG , val); }
Signature · 1 line
2.20.1
PATCH v2 8/8 net: mvpp2: use resolved link config in mac_link_up()
In reply to Russell King - ARM Linux adminConvert the Marvell mvpp2 ethernet driver to use the finalised link parameters in mac_link_up() rather than the parameters in mac_config().
--- .../net/ethernet/marvell/mvpp2/mvpp2_main.c | 83 +++++++++++-------- 1 file changed, 47 insertions(+), 36 deletions(-)
drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c +47 -36
@@ -4976,15 +4976,13 @@ static void mvpp2_gmac_config(struct mvpp2_port *port, unsigned int mode, old_ctrl2 = ctrl2 =readl (port->base +MVPP2_GMAC_CTRL_2_REG ); old_ctrl4 = ctrl4 =readl (port->base +MVPP22_GMAC_CTRL_4_REG ); - an &= ~(MVPP2_GMAC_CONFIG_MII_SPEED |MVPP2_GMAC_CONFIG_GMII_SPEED | -MVPP2_GMAC_AN_SPEED_EN |MVPP2_GMAC_FC_ADV_EN | + an &= ~(MVPP2_GMAC_AN_SPEED_EN |MVPP2_GMAC_FC_ADV_EN |MVPP2_GMAC_FC_ADV_ASM_EN |MVPP2_GMAC_FLOW_CTRL_AUTONEG | -MVPP2_GMAC_CONFIG_FULL_DUPLEX |MVPP2_GMAC_AN_DUPLEX_EN | -MVPP2_GMAC_IN_BAND_AUTONEG |MVPP2_GMAC_IN_BAND_AUTONEG_BYPASS ); +MVPP2_GMAC_AN_DUPLEX_EN |MVPP2_GMAC_IN_BAND_AUTONEG | +MVPP2_GMAC_IN_BAND_AUTONEG_BYPASS ); ctrl0 &= ~MVPP2_GMAC_PORT_TYPE_MASK ; ctrl2 &= ~(MVPP2_GMAC_INBAND_AN_MASK |MVPP2_GMAC_PORT_RESET_MASK |MVPP2_GMAC_PCS_ENABLE_MASK ); - ctrl4 &= ~(MVPP22_CTRL4_RX_FC_EN |MVPP22_CTRL4_TX_FC_EN );/* Configure port type */ if (phy_interface_mode_is_8023z (state->interface)) {
@@ -5014,31 +5012,20 @@ static void mvpp2_gmac_config(struct mvpp2_port *port, unsigned int mode,/* Configure negotiation style */ if (!phylink_autoneg_inband (mode)) { -/* Phy or fixed speed - no in-band AN */ -if (state->duplex) - an |=MVPP2_GMAC_CONFIG_FULL_DUPLEX ; - -if (state->speed ==SPEED_1000 || state->speed ==SPEED_2500 ) - an |=MVPP2_GMAC_CONFIG_GMII_SPEED ; -else if (state->speed ==SPEED_100 ) - an |=MVPP2_GMAC_CONFIG_MII_SPEED ; - -if (state->pause &MLO_PAUSE_TX ) - ctrl4 |=MVPP22_CTRL4_TX_FC_EN ; -if (state->pause &MLO_PAUSE_RX ) - ctrl4 |=MVPP22_CTRL4_RX_FC_EN ; +/* Phy or fixed speed - no in-band AN, nothing to do, leave the +* configured speed, duplex and flow control as-is. +*/ }else if (state->interface ==PHY_INTERFACE_MODE_SGMII ) {/* SGMII in-band mode receives the speed and duplex from * the PHY. Flow control information is not received. */ - an &= ~(MVPP2_GMAC_FORCE_LINK_DOWN |MVPP2_GMAC_FORCE_LINK_PASS ); + an &= ~(MVPP2_GMAC_FORCE_LINK_DOWN | +MVPP2_GMAC_FORCE_LINK_PASS | +MVPP2_GMAC_CONFIG_MII_SPEED | +MVPP2_GMAC_CONFIG_GMII_SPEED | +MVPP2_GMAC_CONFIG_FULL_DUPLEX ); an |=MVPP2_GMAC_IN_BAND_AUTONEG |MVPP2_GMAC_AN_SPEED_EN |MVPP2_GMAC_AN_DUPLEX_EN ; - -if (state->pause &MLO_PAUSE_TX ) - ctrl4 |=MVPP22_CTRL4_TX_FC_EN ; -if (state->pause &MLO_PAUSE_RX ) - ctrl4 |=MVPP22_CTRL4_RX_FC_EN ; }else if (phy_interface_mode_is_8023z (state->interface)) {/* 1000BaseX and 2500BaseX ports cannot negotiate speed nor can * they negotiate duplex: they are always operating with a fixed
@@ -5046,19 +5033,17 @@ static void mvpp2_gmac_config(struct mvpp2_port *port, unsigned int mode,* speed and full duplex here. */ ctrl0 |=MVPP2_GMAC_PORT_TYPE_MASK ; - an &= ~(MVPP2_GMAC_FORCE_LINK_DOWN |MVPP2_GMAC_FORCE_LINK_PASS ); + an &= ~(MVPP2_GMAC_FORCE_LINK_DOWN | +MVPP2_GMAC_FORCE_LINK_PASS | +MVPP2_GMAC_CONFIG_MII_SPEED | +MVPP2_GMAC_CONFIG_GMII_SPEED | +MVPP2_GMAC_CONFIG_FULL_DUPLEX ); an |=MVPP2_GMAC_IN_BAND_AUTONEG |MVPP2_GMAC_CONFIG_GMII_SPEED |MVPP2_GMAC_CONFIG_FULL_DUPLEX ; -if (state->pause &MLO_PAUSE_AN && state->an_enabled) { +if (state->pause &MLO_PAUSE_AN && state->an_enabled) an |=MVPP2_GMAC_FLOW_CTRL_AUTONEG ; - }else { -if (state->pause &MLO_PAUSE_TX ) - ctrl4 |=MVPP22_CTRL4_TX_FC_EN ; -if (state->pause &MLO_PAUSE_RX ) - ctrl4 |=MVPP22_CTRL4_RX_FC_EN ; - } }/* Some fields of the auto-negotiation register require the port to be down when
@@ -5155,18 +5140,44 @@ static void mvpp2_mac_link_up(struct phylink_config *config,struct mvpp2_port *port =netdev_priv (dev);u32 val; -if (!phylink_autoneg_inband (mode)) { -if (mvpp2_is_xlg (interface)) { +if (mvpp2_is_xlg (interface)) { +if (!phylink_autoneg_inband (mode)) { val =readl (port->base +MVPP22_XLG_CTRL0_REG ); val &= ~MVPP22_XLG_CTRL0_FORCE_LINK_DOWN ; val |=MVPP22_XLG_CTRL0_FORCE_LINK_PASS ;writel (val, port->base +MVPP22_XLG_CTRL0_REG ); - }else { + } + }else { +if (!phylink_autoneg_inband (mode)) { val =readl (port->base +MVPP2_GMAC_AUTONEG_CONFIG ); - val &= ~MVPP2_GMAC_FORCE_LINK_DOWN ; + val &= ~(MVPP2_GMAC_FORCE_LINK_DOWN | +MVPP2_GMAC_CONFIG_MII_SPEED | +MVPP2_GMAC_CONFIG_GMII_SPEED | +MVPP2_GMAC_CONFIG_FULL_DUPLEX ); val |=MVPP2_GMAC_FORCE_LINK_PASS ; + +if (speed ==SPEED_1000 || speed ==SPEED_2500 ) + val |=MVPP2_GMAC_CONFIG_GMII_SPEED ; +else if (speed ==SPEED_100 ) + val |=MVPP2_GMAC_CONFIG_MII_SPEED ; + +if (duplex ==DUPLEX_FULL ) + val |=MVPP2_GMAC_CONFIG_FULL_DUPLEX ; +writel (val, port->base +MVPP2_GMAC_AUTONEG_CONFIG ); } + +/* We can always update the flow control enable bits; +* these will only be effective if flow control AN +* (MVPP2_GMAC_FLOW_CTRL_AUTONEG) is disabled. +*/ + val =readl (port->base +MVPP22_GMAC_CTRL_4_REG ); + val &= ~(MVPP22_CTRL4_RX_FC_EN |MVPP22_CTRL4_TX_FC_EN ); +if (tx_pause) + val |=MVPP22_CTRL4_TX_FC_EN ; +if (rx_pause) + val |=MVPP22_CTRL4_RX_FC_EN ; +writel (val, port->base +MVPP22_GMAC_CTRL_4_REG ); }mvpp2_port_enable (port);
Signature · 1 line
2.20.1
David Miller
maintainerFrom: Russell King - ARM Linux admin <[email protected]>
Date: Wed, 26 Feb 2020 10:23:12 +0000
Quoted 26 lines
The following series changes the phylink interface to allow us to better support split MAC / MAC PCS setups. The fundamental change required for this turns out to be quite simple.
Today,
mac_config()is used for everything to do with setting the parameters for the MAC, andmac_link_up()is used to inform the MAC driver that the link is now up (and so to allow packet flow.)mac_config()also has had a few implementation issues, with folk who believe that members such as "speed" and "duplex" are always valid, where "link" gets used inappropriately, etc.With the proposed patches, all this changes subtly - but in a backwards compatible way at this stage.
We pass the the full resolved link state (speed, duplex, pause) to
mac_link_up(), and it is now guaranteed that these parameters to this function will always be valid (no moreSPEED_UNKNOWNorDUPLEX_UNKNOWNhere - unless phylink is fed with such things.)Drivers should convert over to using the state in
mac_link_up()rather than configuring the speed, duplex and pause in themac_config()method. The patch series includes a number of MAC drivers which I've thought have been easy targets - I've left the remainder as I think they need maintainer input. However, *all* drivers will need conversion for future phylink development.
Quoted 2 lines
v2: add ocelot/felix and qca/ar9331 DSA drivers to patch 2, add
received tested-by so far.
In order to end the storm in a teacup, I've applied this series.
:-)
Thanks Russell.
Russell King - ARM Linux admin
maintainerQuoted 29 lines · David Miller
From: Russell King - ARM Linux admin <[email protected]>
Date: Wed, 26 Feb 2020 10:23:12 +0000Quoted 26 lines
The following series changes the phylink interface to allow us to better support split MAC / MAC PCS setups. The fundamental change required for this turns out to be quite simple.
Today,
mac_config()is used for everything to do with setting the parameters for the MAC, andmac_link_up()is used to inform the MAC driver that the link is now up (and so to allow packet flow.)mac_config()also has had a few implementation issues, with folk who believe that members such as "speed" and "duplex" are always valid, where "link" gets used inappropriately, etc.With the proposed patches, all this changes subtly - but in a backwards compatible way at this stage.
We pass the the full resolved link state (speed, duplex, pause) to
mac_link_up(), and it is now guaranteed that these parameters to this function will always be valid (no moreSPEED_UNKNOWNorDUPLEX_UNKNOWNhere - unless phylink is fed with such things.)Drivers should convert over to using the state in
mac_link_up()rather than configuring the speed, duplex and pause in themac_config()method. The patch series includes a number of MAC drivers which I've thought have been easy targets - I've left the remainder as I think they need maintainer input. However, *all* drivers will need conversion for future phylink development.
Quoted 8 lines
Quoted 2 lines
v2: add ocelot/felix and qca/ar9331 DSA drivers to patch 2, add
received tested-by so far.In order to end the storm in a teacup, I've applied this series.
:-)
Thanks Russell.
Thanks David, I was getting concerned that the teacup might break! :p
Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up According to speedtest.net: 11.9Mbps down 500kbps up