Skip to contentSkip to conversation

net: stmmac: convert stmmac "pcs" to phylink

RFC PATCH Superseded Russell King (Oracle) 5 participants 44 messages May 31, 2024 – Jul 4, 2024

Subsystems (7)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover net: stmmac: convert stmmac "pcs" to phylink

Hi,

This is version 2 of the series switching stmmac to use phylink PCS instead of going behind phylink's back.

Changes since version 1:
- Addition of patches from Serge Semin to allow RGMII to use the "PCS" code even if priv->dma_cap.pcs is not set (including tweaks by me.)
- Restructuring of the patch set to be a more logical split.
- Leave the pcs_ctrl_ane methods until we've worked out what to do with the qcom-ethqos driver (this series may still end up breaking it, but at least we will now successfully compile.)

A reminder that what I want to hear from this patch set are the results of testing - and thanks to Serge, the RGMII paths were exercised, but I have not had any results for the SGMII side of this.

There are still a bunch of outstanding questions:

- whether we should be using two separate PCS instances, one for RGMII and another for SGMII. If the PCS hardware is not present, but are using RGMII mode, then we probably don't want to be accessing the registers that would've been there for SGMII.
- what the three interrupts associated with the PCS code actually mean when they fire.
- which block's status we're reading in the pcs_get_state() method, and whether we should be reading that for both RGMII and SGMII.
- whether we need to activate phylink's inband mode in more cases (so that the PCS/MAC status gets read and used for the link.)

There's probably more questions to be asked... but really the critical thing is to shake out any breakage from making this conversion. Bear in mind that I have little knowledge of this hardware, so this conversion has been done somewhat blind using only what I can observe from the current driver.

Original blurb below.

As I noted recently in a thread (and was ignored) stmmac sucks. (I won't hide my distain for drivers that make my life as phylink maintainer more difficult!)

One of the contract conditions for using phylink is that the driver will _not_ mess with the netif carrier. stmmac developers/maintainers clearly didn't read that, because stmmac messes with the netif carrier, which destroys phylink's guarantee that it'll make certain calls in a particular order (e.g. it won't call mac_link_up() twice in a row without an intervening mac_link_down().) This is clearly stated in the phylink documentation.

Thus, this patch set attempts to fix this. Why does it mess with the netif carrier? It has its own independent PCS implementation that completely bypasses phylink _while_ phylink is still being used. This is not acceptable. Either the driver uses phylink, or it doesn't use phylink. There is no half-way house about this. Therefore, this driver needs to either be fixed, or needs to stop using phylink.

Since I was ignored when I brought this up, I've hacked together the following patch set - and it is hacky at the moment. It's also broken because of recentl changes involving dwmac-qcom-ethqos.c - but there isn't sufficient information in the driver for me to fix this. The driver appears to use SGMII at 2500Mbps, which simply does not exist. What interface mode (and neg_mode) does phylink pass to pcs_config() in each of the speeds that dwmac-qcom-ethqos.c is interested in. Without this information, I can't do that conversion. So for the purposes of this, I've just ignored dwmac-qcom-ethqos.c (which means it will fail to build.)

The patch splitup is not ideal, but that's not what I'm interested in here. What I want to hear is the results of testing - does this switch of the RGMII/SGMII "pcs" stuff to a phylink_pcs work for this driver?

Please don't review the patches, but you are welcome to send fixes to them. Once we know that the overall implementation works, then I'll look at how best to split the patches. In the mean time, the present form is more convenient for making changes and fixing things.

There is still more improvement that's needed here.

Thanks.

 drivers/net/ethernet/stmicro/stmmac/Makefile       |   2 +-
 drivers/net/ethernet/stmicro/stmmac/common.h       |  12 +-
 .../net/ethernet/stmicro/stmmac/dwmac1000_core.c   | 146 ++++++++++++++-------
 drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c  | 131 +++++++++++++-----
 drivers/net/ethernet/stmicro/stmmac/hwif.h         |  19 ++-
 .../net/ethernet/stmicro/stmmac/stmmac_ethtool.c   | 111 +---------------
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c  |  33 ++---
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c   |  58 ++++++++
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h   |  34 +----
 9 files changed, 298 insertions(+), 248 deletions(-)
 create mode 100644 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#1 lore raw

RFC PATCH v2 1/8 net: stmmac: add infrastructure for hwifs to provide PCS

Allow hwifs to provide a phylink_select_pcs() implementation via struct stmmac_ops, which can be used to provide a phylink PCS.

Code analysis shows that when STMMAC_FLAG_HAS_INTEGRATED_PCS is set, then:

stmmac_common_interrupt()
stmmac_ethtool_set_link_ksettings()
stmmac_ethtool_get_link_ksettings()

will all ignore the presence of the PCS. The latter two will pass the ethtool commands to phylink. The former will avoid manipulating the netif carrier state behind phylink's back based on the PCS status.

This flag is only set by the ethqos driver. From what I can tell, amongst the current kernel DT files that use the ethqos driver, only sa8775p-ride.dts enables ethernet, and this defines a SGMII-mode link to its PHYs without the "managed" property. Thus, phylink will be operating in MLO_AN_PHY mode, and inband mode will not be used.

Therefore, it is safe to ignore the STMMAC_FLAG_HAS_INTEGRATED_PCS flag in stmmac_mac_select_pcs().

Further code analysis shows that XPCS is used by Intel for Cisco SGMII and 1000base-X modes. In this case, we do not want to provide the integrated PCS, but the XPCS. The same appears to also be true of the Lynx PCS.

Therefore, it seems that the integrated PCS provided by the hwif MAC code should only be used when an external PCS is not being used, so give priority to the external PCS.

Provide a phylink_pcs instance in struct mac_device_info for hwifs to use to provide their phylink PCS.

Omit the non-phylink PCS code paths when a hwif provides a phylink_select_pcs() method (in other words, when they are converted to use a phylink PCS.) This provides a way to transition parts of the driver in the subsequent patches.

---
 drivers/net/ethernet/stmicro/stmmac/common.h  |  9 ++++++++-
 drivers/net/ethernet/stmicro/stmmac/hwif.h    | 19 +++++++++++++++++--
 .../ethernet/stmicro/stmmac/stmmac_ethtool.c  | 12 ++++++++----
 .../net/ethernet/stmicro/stmmac/stmmac_main.c | 10 +++++++---
 4 files changed, 40 insertions(+), 10 deletions(-)
drivers/net/ethernet/stmicro/stmmac/common.h +8 -1
index cd36ff4da68c..940e83fa1202 100644
@@ -14,7 +14,7 @@
 #include <linux/etherdevice.h>
 #include <linux/netdevice.h>
 #include <linux/stmmac.h>
-#include <linux/phy.h>
+#include <linux/phylink.h>
 #include <linux/pcs/pcs-xpcs.h>
 #include <linux/module.h>
 #if IS_ENABLED(CONFIG_VLAN_8021Q)
@@ -591,6 +591,7 @@ struct mac_device_info {
 	const struct stmmac_tc_ops *tc;
 	const struct stmmac_mmc_ops *mmc;
 	const struct stmmac_est_ops *est;
+	struct phylink_pcs mac_pcs;	/* The MAC's RGMII/SGMII "PCS" */
 	struct dw_xpcs *xpcs;
 	struct phylink_pcs *phylink_pcs;
 	struct mii_regs mii;	/* MII register Addresses */
@@ -611,6 +612,12 @@ struct mac_device_info {
 	bool hw_vlan_en;
 };
 
+static inline struct mac_device_info *
+phylink_pcs_to_mac_dev_info(struct phylink_pcs *pcs)
+{
+	return container_of(pcs, struct mac_device_info, mac_pcs);
+}
+
 struct stmmac_rx_routing {
 	u32 reg_mask;
 	u32 reg_shift;
drivers/net/ethernet/stmicro/stmmac/hwif.h +17 -2
index 97934ccba5b1..aa5f6e40cb53 100644
@@ -5,6 +5,7 @@
 #ifndef __STMMAC_HWIF_H__
 #define __STMMAC_HWIF_H__
 
+#include <linux/err.h>
 #include <linux/netdevice.h>
 #include <linux/stmmac.h>
 
@@ -17,13 +18,17 @@
 	} \
 	__result; \
 })
-#define stmmac_do_callback(__priv, __module, __cname,  __arg0, __args...) \
+#define stmmac_do_typed_callback(__type, __fail_ret, __priv, __module, \
+				 __cname,  __arg0, __args...) \
 ({ \
-	int __result = -EINVAL; \
+	__type __result = __fail_ret; \
 	if ((__priv)->hw->__module && (__priv)->hw->__module->__cname) \
 		__result = (__priv)->hw->__module->__cname((__arg0), ##__args); \
 	__result; \
 })
+#define stmmac_do_callback(__priv, __module, __cname,  __arg0, __args...) \
+	stmmac_do_typed_callback(int, -EINVAL, __priv, __module, __cname, \
+				 __arg0, ##__args)
 
 struct stmmac_extra_stats;
 struct stmmac_priv;
@@ -310,6 +315,9 @@ struct stmmac_ops {
 	void (*core_init)(struct mac_device_info *hw, struct net_device *dev);
 	/* Update MAC capabilities */
 	void (*update_caps)(struct stmmac_priv *priv);
+	/* Get phylink PCS (for MAC */
+	struct phylink_pcs *(*phylink_select_pcs)(struct stmmac_priv *priv,
+						  phy_interface_t interface);
 	/* Enable the MAC RX/TX */
 	void (*set_mac)(void __iomem *ioaddr, bool enable);
 	/* Enable and verify that the IPC module is supported */
@@ -431,6 +439,10 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, core_init, __args)
 #define stmmac_mac_update_caps(__priv) \
 	stmmac_do_void_callback(__priv, mac, update_caps, __priv)
+#define stmmac_mac_phylink_select_pcs(__priv, __interface) \
+	stmmac_do_typed_callback(struct phylink_pcs *, ERR_PTR(-EOPNOTSUPP), \
+				 __priv, mac, phylink_select_pcs, __priv,\
+				 __interface)
 #define stmmac_mac_set(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, set_mac, __args)
 #define stmmac_rx_ipc(__priv, __args...) \
@@ -530,6 +542,9 @@ struct stmmac_ops {
 #define stmmac_fpe_irq_status(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, fpe_irq_status, __args)
 
+#define stmmac_has_mac_phylink_select_pcs(__priv) \
+	((__priv)->hw->mac->phylink_select_pcs != NULL)
+
 /* PTP and HW Timer helpers */
 struct stmmac_hwtimestamp {
 	void (*config_hw_tstamping) (void __iomem *ioaddr, u32 data);
drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c +8 -4
index 18468c0228f0..7c6530d63ade 100644
@@ -323,7 +323,8 @@ static int stmmac_ethtool_get_link_ksettings(struct net_device *dev,
 
 	if (!(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
 	    (priv->hw->pcs & STMMAC_PCS_RGMII ||
-	     priv->hw->pcs & STMMAC_PCS_SGMII)) {
+	     priv->hw->pcs & STMMAC_PCS_SGMII) &&
+	    !stmmac_has_mac_phylink_select_pcs(priv)) {
 		struct rgmii_adv adv;
 		u32 supported, advertising, lp_advertising;
 
@@ -410,7 +411,8 @@ stmmac_ethtool_set_link_ksettings(struct net_device *dev,
 
 	if (!(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
 	    (priv->hw->pcs & STMMAC_PCS_RGMII ||
-	     priv->hw->pcs & STMMAC_PCS_SGMII)) {
+	     priv->hw->pcs & STMMAC_PCS_SGMII) &&
+	    !stmmac_has_mac_phylink_select_pcs(priv)) {
 		/* Only support ANE */
 		if (cmd->base.autoneg != AUTONEG_ENABLE)
 			return -EINVAL;
@@ -523,7 +525,8 @@ stmmac_get_pauseparam(struct net_device *netdev,
 	struct stmmac_priv *priv = netdev_priv(netdev);
 	struct rgmii_adv adv_lp;
 
-	if (priv->hw->pcs && !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
+	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv) &&
+	    !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
 		pause->autoneg = 1;
 		if (!adv_lp.pause)
 			return;
@@ -539,7 +542,8 @@ stmmac_set_pauseparam(struct net_device *netdev,
 	struct stmmac_priv *priv = netdev_priv(netdev);
 	struct rgmii_adv adv_lp;
 
-	if (priv->hw->pcs && !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
+	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv) &&
+	   !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
 		pause->autoneg = 1;
 		if (!adv_lp.pause)
 			return -EOPNOTSUPP;
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +7 -3
index bbedf2a8c60f..56c351e11952 100644
@@ -953,7 +953,10 @@ static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,
 	if (priv->hw->xpcs)
 		return &priv->hw->xpcs->pcs;
 
-	return priv->hw->phylink_pcs;
+	if (priv->hw->phylink_pcs)
+		return priv->hw->phylink_pcs;
+
+	return stmmac_mac_phylink_select_pcs(priv, interface);
 }
 
 static void stmmac_mac_config(struct phylink_config *config, unsigned int mode,
@@ -3482,7 +3485,7 @@ static int stmmac_hw_setup(struct net_device *dev, bool ptp_register)
 		}
 	}
 
-	if (priv->hw->pcs)
+	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv))
 		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, priv->hw->ps, 0);
 
 	/* set TX and RX rings length */
@@ -6052,7 +6055,8 @@ static void stmmac_common_interrupt(struct stmmac_priv *priv)
 
 		/* PCS link status */
 		if (priv->hw->pcs &&
-		    !(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)) {
+		    !(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
+		    !stmmac_has_mac_phylink_select_pcs(priv)) {
 			if (priv->xstats.pcs_link)
 				netif_carrier_on(priv->dev);
 			else
Signature · 1 line
2.30.2
#2 lore raw

Re: [PATCH RFC net-next v2 1/8] net: stmmac: add infrastructure for hwifs to provide PCS

Quoted 115 lines · Russell King (Oracle)

Allow hwifs to provide a phylink_select_pcs() implementation via struct stmmac_ops, which can be used to provide a phylink PCS.

Code analysis shows that when STMMAC_FLAG_HAS_INTEGRATED_PCS is set, then:

stmmac_common_interrupt()
stmmac_ethtool_set_link_ksettings()
stmmac_ethtool_get_link_ksettings()

will all ignore the presence of the PCS. The latter two will pass the ethtool commands to phylink. The former will avoid manipulating the netif carrier state behind phylink's back based on the PCS status.

This flag is only set by the ethqos driver. From what I can tell, amongst the current kernel DT files that use the ethqos driver, only sa8775p-ride.dts enables ethernet, and this defines a SGMII-mode link to its PHYs without the "managed" property. Thus, phylink will be operating in MLO_AN_PHY mode, and inband mode will not be used.

Therefore, it is safe to ignore the STMMAC_FLAG_HAS_INTEGRATED_PCS flag in stmmac_mac_select_pcs().

Further code analysis shows that XPCS is used by Intel for Cisco SGMII and 1000base-X modes. In this case, we do not want to provide the integrated PCS, but the XPCS. The same appears to also be true of the Lynx PCS.

Therefore, it seems that the integrated PCS provided by the hwif MAC code should only be used when an external PCS is not being used, so give priority to the external PCS.

Provide a phylink_pcs instance in struct mac_device_info for hwifs to use to provide their phylink PCS.

Omit the non-phylink PCS code paths when a hwif provides a phylink_select_pcs() method (in other words, when they are converted to use a phylink PCS.) This provides a way to transition parts of the driver in the subsequent patches.

---
 drivers/net/ethernet/stmicro/stmmac/common.h  |  9 ++++++++-
 drivers/net/ethernet/stmicro/stmmac/hwif.h    | 19 +++++++++++++++++--
 .../ethernet/stmicro/stmmac/stmmac_ethtool.c  | 12 ++++++++----
 .../net/ethernet/stmicro/stmmac/stmmac_main.c | 10 +++++++---
 4 files changed, 40 insertions(+), 10 deletions(-)
drivers/net/ethernet/stmicro/stmmac/common.h +8 -1
index cd36ff4da68c..940e83fa1202 100644
@@ -14,7 +14,7 @@
 #include <linux/etherdevice.h>
 #include <linux/netdevice.h>
 #include <linux/stmmac.h>
-#include <linux/phy.h>
+#include <linux/phylink.h>
 #include <linux/pcs/pcs-xpcs.h>
 #include <linux/module.h>
 #if IS_ENABLED(CONFIG_VLAN_8021Q)
@@ -591,6 +591,7 @@ struct mac_device_info {
 	const struct stmmac_tc_ops *tc;
 	const struct stmmac_mmc_ops *mmc;
 	const struct stmmac_est_ops *est;
+	struct phylink_pcs mac_pcs;	/* The MAC's RGMII/SGMII "PCS" */
 	struct dw_xpcs *xpcs;
 	struct phylink_pcs *phylink_pcs;
 	struct mii_regs mii;	/* MII register Addresses */
@@ -611,6 +612,12 @@ struct mac_device_info {
 	bool hw_vlan_en;
 };
 
+static inline struct mac_device_info *
+phylink_pcs_to_mac_dev_info(struct phylink_pcs *pcs)
+{
+	return container_of(pcs, struct mac_device_info, mac_pcs);
+}
+
 struct stmmac_rx_routing {
 	u32 reg_mask;
 	u32 reg_shift;
drivers/net/ethernet/stmicro/stmmac/hwif.h +7 -2
index 97934ccba5b1..aa5f6e40cb53 100644
@@ -5,6 +5,7 @@
 #ifndef __STMMAC_HWIF_H__
 #define __STMMAC_HWIF_H__
 
+#include <linux/err.h>
 #include <linux/netdevice.h>
 #include <linux/stmmac.h>
 
@@ -17,13 +18,17 @@
 	} \
 	__result; \
 })
-#define stmmac_do_callback(__priv, __module, __cname,  __arg0, __args...) \
+#define stmmac_do_typed_callback(__type, __fail_ret, __priv, __module, \
+				 __cname,  __arg0, __args...) \
 ({ \
-	int __result = -EINVAL; \
+	__type __result = __fail_ret; \
 	if ((__priv)->hw->__module && (__priv)->hw->__module->__cname) \
 		__result = (__priv)->hw->__module->__cname((__arg0), ##__args); \
 	__result; \
 })
+#define stmmac_do_callback(__priv, __module, __cname,  __arg0, __args...) \
+	stmmac_do_typed_callback(int, -EINVAL, __priv, __module, __cname, \
+				 __arg0, ##__args)
 
 struct stmmac_extra_stats;
 struct stmmac_priv;

drivers/net/ethernet/stmicro/stmmac/hwif.h struct stmmac_ops { · line 318

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/hwif.h +1 -0
@@ -310,6 +315,9 @@ struct stmmac_ops {
 	void (*core_init)(struct mac_device_info *hw, struct net_device *dev);
 	/* Update MAC capabilities */
 	void (*update_caps)(struct stmmac_priv *priv);
+	/* Get phylink PCS (for MAC */

nit: unclosed parenthesis

Quoted 108 lines
drivers/net/ethernet/stmicro/stmmac/hwif.h +9 -0
+	struct phylink_pcs *(*phylink_select_pcs)(struct stmmac_priv *priv,
+						  phy_interface_t interface);
 	/* Enable the MAC RX/TX */
 	void (*set_mac)(void __iomem *ioaddr, bool enable);
 	/* Enable and verify that the IPC module is supported */
@@ -431,6 +439,10 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, core_init, __args)
 #define stmmac_mac_update_caps(__priv) \
 	stmmac_do_void_callback(__priv, mac, update_caps, __priv)
+#define stmmac_mac_phylink_select_pcs(__priv, __interface) \
+	stmmac_do_typed_callback(struct phylink_pcs *, ERR_PTR(-EOPNOTSUPP), \
+				 __priv, mac, phylink_select_pcs, __priv,\
+				 __interface)
 #define stmmac_mac_set(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, set_mac, __args)
 #define stmmac_rx_ipc(__priv, __args...) \
@@ -530,6 +542,9 @@ struct stmmac_ops {
 #define stmmac_fpe_irq_status(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, fpe_irq_status, __args)
 
+#define stmmac_has_mac_phylink_select_pcs(__priv) \
+	((__priv)->hw->mac->phylink_select_pcs != NULL)
+
 /* PTP and HW Timer helpers */
 struct stmmac_hwtimestamp {
 	void (*config_hw_tstamping) (void __iomem *ioaddr, u32 data);
drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c +8 -4
index 18468c0228f0..7c6530d63ade 100644
@@ -323,7 +323,8 @@ static int stmmac_ethtool_get_link_ksettings(struct net_device *dev,
 
 	if (!(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
 	    (priv->hw->pcs & STMMAC_PCS_RGMII ||
-	     priv->hw->pcs & STMMAC_PCS_SGMII)) {
+	     priv->hw->pcs & STMMAC_PCS_SGMII) &&
+	    !stmmac_has_mac_phylink_select_pcs(priv)) {
 		struct rgmii_adv adv;
 		u32 supported, advertising, lp_advertising;
 
@@ -410,7 +411,8 @@ stmmac_ethtool_set_link_ksettings(struct net_device *dev,
 
 	if (!(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
 	    (priv->hw->pcs & STMMAC_PCS_RGMII ||
-	     priv->hw->pcs & STMMAC_PCS_SGMII)) {
+	     priv->hw->pcs & STMMAC_PCS_SGMII) &&
+	    !stmmac_has_mac_phylink_select_pcs(priv)) {
 		/* Only support ANE */
 		if (cmd->base.autoneg != AUTONEG_ENABLE)
 			return -EINVAL;
@@ -523,7 +525,8 @@ stmmac_get_pauseparam(struct net_device *netdev,
 	struct stmmac_priv *priv = netdev_priv(netdev);
 	struct rgmii_adv adv_lp;
 
-	if (priv->hw->pcs && !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
+	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv) &&
+	    !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
 		pause->autoneg = 1;
 		if (!adv_lp.pause)
 			return;
@@ -539,7 +542,8 @@ stmmac_set_pauseparam(struct net_device *netdev,
 	struct stmmac_priv *priv = netdev_priv(netdev);
 	struct rgmii_adv adv_lp;
 
-	if (priv->hw->pcs && !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
+	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv) &&
+	   !stmmac_pcs_get_adv_lp(priv, priv->ioaddr, &adv_lp)) {
 		pause->autoneg = 1;
 		if (!adv_lp.pause)
 			return -EOPNOTSUPP;
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +7 -3
index bbedf2a8c60f..56c351e11952 100644
@@ -953,7 +953,10 @@ static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,
 	if (priv->hw->xpcs)
 		return &priv->hw->xpcs->pcs;
 
-	return priv->hw->phylink_pcs;
+	if (priv->hw->phylink_pcs)
+		return priv->hw->phylink_pcs;
+
+	return stmmac_mac_phylink_select_pcs(priv, interface);
 }
 
 static void stmmac_mac_config(struct phylink_config *config, unsigned int mode,
@@ -3482,7 +3485,7 @@ static int stmmac_hw_setup(struct net_device *dev, bool ptp_register)
 		}
 	}
 
-	if (priv->hw->pcs)
+	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv))
 		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, priv->hw->ps, 0);
 
 	/* set TX and RX rings length */
@@ -6052,7 +6055,8 @@ static void stmmac_common_interrupt(struct stmmac_priv *priv)
 
 		/* PCS link status */
 		if (priv->hw->pcs &&
-		    !(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)) {
+		    !(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
+		    !stmmac_has_mac_phylink_select_pcs(priv)) {
 			if (priv->xstats.pcs_link)
 				netif_carrier_on(priv->dev);
 			else
Signature · 1 line
2.30.2
#3 lore raw

Re: [PATCH RFC net-next v2 1/8] net: stmmac: add infrastructure for hwifs to provide PCS

Quoted 8 lines · Andrew Halaney

drivers/net/ethernet/stmicro/stmmac/hwif.h struct stmmac_ops { · line 318

Quoted 5 lines · Russell King (Oracle)
drivers/net/ethernet/stmicro/stmmac/hwif.h +1 -0
@@ -310,6 +315,9 @@ struct stmmac_ops {
 	void (*core_init)(struct mac_device_info *hw, struct net_device *dev);
 	/* Update MAC capabilities */
 	void (*update_caps)(struct stmmac_priv *priv);
+	/* Get phylink PCS (for MAC */

nit: unclosed parenthesis

Please learn to trim your review replies so your patch submitters don't have to waste time scroll through lots of useless context. This has been part of good netiquette for decades. Thanks.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#4 lore raw

RFC PATCH v2 2/8 net: stmmac: provide core phylink PCS infrastructure

In reply to Russell King (Oracle)

There are some common operations shared between the dwmac hwifs, the only difference is where they are located within the device. These are already present in the form of dwmac_rane(), dwmac_ctrl_ane() and dwmac_get_adv_lp(). Rather than use these (which don't quite fit with phylink PCS, provide phylink PCS specific versions.

---
 drivers/net/ethernet/stmicro/stmmac/Makefile  |  2 +-
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.c  | 58 +++++++++++++++++++
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  | 10 ++++
 3 files changed, 69 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c
drivers/net/ethernet/stmicro/stmmac/Makefile +1 -1
index c2f0e91f6bf8..9e15f7615ff4 100644
@@ -6,7 +6,7 @@ stmmac-objs:= stmmac_main.o stmmac_ethtool.o stmmac_mdio.o ring_mode.o	\
 	      mmc_core.o stmmac_hwtstamp.o stmmac_ptp.o dwmac4_descs.o	\
 	      dwmac4_dma.o dwmac4_lib.o dwmac4_core.o dwmac5.o hwif.o \
 	      stmmac_tc.o dwxgmac2_core.o dwxgmac2_dma.o dwxgmac2_descs.o \
-	      stmmac_xdp.o stmmac_est.o \
+	      stmmac_xdp.o stmmac_est.o stmmac_pcs.o \
 	      $(stmmac-y)
 
 stmmac-$(CONFIG_STMMAC_SELFTESTS) += stmmac_selftests.o
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c new file +58 -0

58 lines not on this page · raw message

drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +10 -0
index 1bdf87b237c4..888485fa2761 100644
@@ -137,4 +137,14 @@ static inline void dwmac_get_adv_lp(void __iomem *ioaddr, u32 reg,
 
 	adv_lp->lp_pause = (value & GMAC_ANE_PSE) >> GMAC_ANE_PSE_SHIFT;
 }
+
+int dwmac_pcs_config(struct mac_device_info *hw, unsigned int neg_mode,
+		     phy_interface_t interface,
+		     const unsigned long *advertising,
+		     unsigned int reg_base);
+
+void dwmac_pcs_get_state(struct mac_device_info *hw,
+			 struct phylink_link_state *state,
+			 unsigned int reg_base);
+
 #endif /* __STMMAC_PCS_H__ */
Signature · 1 line
2.30.2
#5 lore raw

RFC PATCH v2 3/8 net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

In reply to Russell King (Oracle)

Convert dwmac1000 sgmii/rgmii "pcs" implementation to use a phylink_pcs so we can eventually get rid of the exceptional paths that conflict with phylink.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 113 ++++++++++++------
 1 file changed, 75 insertions(+), 38 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +75 -38
index d413d76a8936..4a0572d5f865 100644
@@ -16,6 +16,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
+#include <linux/phylink.h>
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac1000.h"
@@ -261,39 +262,6 @@ static void dwmac1000_pmt(struct mac_device_info *hw, unsigned long mode)
 	writel(pmt, ioaddr + GMAC_PMT);
 }
 
-/* RGMII or SMII interface */
-static void dwmac1000_rgsmii(void __iomem *ioaddr, struct stmmac_extra_stats *x)
-{
-	u32 status;
-
-	status = readl(ioaddr + GMAC_RGSMIIIS);
-	x->irq_rgmii_n++;
-
-	/* Check the link status */
-	if (status & GMAC_RGSMIIIS_LNKSTS) {
-		int speed_value;
-
-		x->pcs_link = 1;
-
-		speed_value = ((status & GMAC_RGSMIIIS_SPEED) >>
-			       GMAC_RGSMIIIS_SPEED_SHIFT);
-		if (speed_value == GMAC_RGSMIIIS_SPEED_125)
-			x->pcs_speed = SPEED_1000;
-		else if (speed_value == GMAC_RGSMIIIS_SPEED_25)
-			x->pcs_speed = SPEED_100;
-		else
-			x->pcs_speed = SPEED_10;
-
-		x->pcs_duplex = (status & GMAC_RGSMIIIS_LNKMOD_MASK);
-
-		pr_info("Link is Up - %d/%s\n", (int)x->pcs_speed,
-			x->pcs_duplex ? "Full" : "Half");
-	} else {
-		x->pcs_link = 0;
-		pr_info("Link is Down\n");
-	}
-}
-
 static int dwmac1000_irq_status(struct mac_device_info *hw,
 				struct stmmac_extra_stats *x)
 {
@@ -335,8 +303,12 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
 
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac1000_rgsmii(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_RGSMIIIS);
+		phylink_pcs_change(&hw->mac_pcs, false);
+		x->irq_rgmii_n++;
+	}
 
 	return ret;
 }
@@ -404,9 +376,71 @@ static void dwmac1000_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
 }
 
-static void dwmac1000_get_adv_lp(void __iomem *ioaddr, struct rgmii_adv *adv)
+static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
+				      unsigned long *supported,
+				      const struct phylink_link_state *state)
+{
+	/* Only support in-band */
+	if (!test_bit(ETHTOOL_LINK_MODE_Autoneg_BIT, state->advertising))
+		return -EINVAL;
+
+	return 0;
+}
+
+static int dwmac1000_mii_pcs_config(struct phylink_pcs *pcs,
+				    unsigned int neg_mode,
+				    phy_interface_t interface,
+				    const unsigned long *advertising,
+				    bool permit_pause_to_mac)
 {
-	dwmac_get_adv_lp(ioaddr, GMAC_PCS_BASE, adv);
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+
+	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
+				GMAC_PCS_BASE);
+}
+
+static void dwmac1000_mii_pcs_get_state(struct phylink_pcs *pcs,
+					struct phylink_link_state *state)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	unsigned int spd_clk;
+	u32 status;
+
+	status = readl(hw->pcsr + GMAC_RGSMIIIS);
+
+	state->link = status & GMAC_RGSMIIIS_LNKSTS;
+	if (!state->link)
+		return;
+
+	spd_clk = FIELD_GET(GMAC_RGSMIIIS_SPEED, status);
+	if (spd_clk == GMAC_RGSMIIIS_SPEED_125)
+		state->speed = SPEED_1000;
+	else if (spd_clk == GMAC_RGSMIIIS_SPEED_25)
+		state->speed = SPEED_100;
+	else if (spd_clk == GMAC_RGSMIIIS_SPEED_2_5)
+		state->speed = SPEED_10;
+
+	state->duplex = status & GMAC_RGSMIIIS_LNKMOD_MASK ?
+			DUPLEX_FULL : DUPLEX_HALF;
+
+	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+}
+
+static const struct phylink_pcs_ops dwmac1000_mii_pcs_ops = {
+	.pcs_validate = dwmac1000_mii_pcs_validate,
+	.pcs_config = dwmac1000_mii_pcs_config,
+	.pcs_get_state = dwmac1000_mii_pcs_get_state,
+};
+
+static struct phylink_pcs *
+dwmac1000_phylink_select_pcs(struct stmmac_priv *priv,
+			     phy_interface_t interface)
+{
+	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
+	    priv->hw->pcs & STMMAC_PCS_SGMII)
+		return &priv->hw->mac_pcs;
+
+	return NULL;
 }
 
 static void dwmac1000_debug(struct stmmac_priv *priv, void __iomem *ioaddr,
@@ -499,6 +533,7 @@ static void dwmac1000_set_mac_loopback(void __iomem *ioaddr, bool enable)
 
 const struct stmmac_ops dwmac1000_ops = {
 	.core_init = dwmac1000_core_init,
+	.phylink_select_pcs = dwmac1000_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac1000_rx_ipc_enable,
 	.dump_regs = dwmac1000_dump_regs,
@@ -514,7 +549,6 @@ const struct stmmac_ops dwmac1000_ops = {
 	.set_eee_pls = dwmac1000_set_eee_pls,
 	.debug = dwmac1000_debug,
 	.pcs_ctrl_ane = dwmac1000_ctrl_ane,
-	.pcs_get_adv_lp = dwmac1000_get_adv_lp,
 	.set_mac_loopback = dwmac1000_set_mac_loopback,
 };
 
@@ -549,5 +583,8 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
+	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
+	mac->mac_pcs.neg_mode = true;
+
 	return 0;
 }
Signature · 1 line
2.30.2
#6 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Quoted 62 lines · Russell King (Oracle)

Convert dwmac1000 sgmii/rgmii "pcs" implementation to use a phylink_pcs so we can eventually get rid of the exceptional paths that conflict with phylink.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 113 ++++++++++++------
 1 file changed, 75 insertions(+), 38 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +1 -33
index d413d76a8936..4a0572d5f865 100644
@@ -16,6 +16,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
+#include <linux/phylink.h>
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac1000.h"
@@ -261,39 +262,6 @@ static void dwmac1000_pmt(struct mac_device_info *hw, unsigned long mode)
 	writel(pmt, ioaddr + GMAC_PMT);
 }
 
-/* RGMII or SMII interface */
-static void dwmac1000_rgsmii(void __iomem *ioaddr, struct stmmac_extra_stats *x)
-{
-	u32 status;
-
-	status = readl(ioaddr + GMAC_RGSMIIIS);
-	x->irq_rgmii_n++;
-
-	/* Check the link status */
-	if (status & GMAC_RGSMIIIS_LNKSTS) {
-		int speed_value;
-
-		x->pcs_link = 1;
-
-		speed_value = ((status & GMAC_RGSMIIIS_SPEED) >>
-			       GMAC_RGSMIIIS_SPEED_SHIFT);
-		if (speed_value == GMAC_RGSMIIIS_SPEED_125)
-			x->pcs_speed = SPEED_1000;
-		else if (speed_value == GMAC_RGSMIIIS_SPEED_25)
-			x->pcs_speed = SPEED_100;
-		else
-			x->pcs_speed = SPEED_10;
-
-		x->pcs_duplex = (status & GMAC_RGSMIIIS_LNKMOD_MASK);
-
-		pr_info("Link is Up - %d/%s\n", (int)x->pcs_speed,
-			x->pcs_duplex ? "Full" : "Half");
-	} else {
-		x->pcs_link = 0;
-		pr_info("Link is Down\n");
-	}
-}
-
 static int dwmac1000_irq_status(struct mac_device_info *hw,
 				struct stmmac_extra_stats *x)
 {

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c static int dwmac1000_irq_status(struct mac_device_info *hw, · line 308

Quoted 8 lines
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +3 -2
@@ -335,8 +303,12 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
 
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac1000_rgsmii(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_RGSMIIIS);

This seems to me that you're doing the TODO here? Maybe I'm misunderstanding... maybe not :)

Quoted 108 lines
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +71 -3
+		phylink_pcs_change(&hw->mac_pcs, false);
+		x->irq_rgmii_n++;
+	}
 
 	return ret;
 }
@@ -404,9 +376,71 @@ static void dwmac1000_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
 }
 
-static void dwmac1000_get_adv_lp(void __iomem *ioaddr, struct rgmii_adv *adv)
+static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
+				      unsigned long *supported,
+				      const struct phylink_link_state *state)
+{
+	/* Only support in-band */
+	if (!test_bit(ETHTOOL_LINK_MODE_Autoneg_BIT, state->advertising))
+		return -EINVAL;
+
+	return 0;
+}
+
+static int dwmac1000_mii_pcs_config(struct phylink_pcs *pcs,
+				    unsigned int neg_mode,
+				    phy_interface_t interface,
+				    const unsigned long *advertising,
+				    bool permit_pause_to_mac)
 {
-	dwmac_get_adv_lp(ioaddr, GMAC_PCS_BASE, adv);
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+
+	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
+				GMAC_PCS_BASE);
+}
+
+static void dwmac1000_mii_pcs_get_state(struct phylink_pcs *pcs,
+					struct phylink_link_state *state)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	unsigned int spd_clk;
+	u32 status;
+
+	status = readl(hw->pcsr + GMAC_RGSMIIIS);
+
+	state->link = status & GMAC_RGSMIIIS_LNKSTS;
+	if (!state->link)
+		return;
+
+	spd_clk = FIELD_GET(GMAC_RGSMIIIS_SPEED, status);
+	if (spd_clk == GMAC_RGSMIIIS_SPEED_125)
+		state->speed = SPEED_1000;
+	else if (spd_clk == GMAC_RGSMIIIS_SPEED_25)
+		state->speed = SPEED_100;
+	else if (spd_clk == GMAC_RGSMIIIS_SPEED_2_5)
+		state->speed = SPEED_10;
+
+	state->duplex = status & GMAC_RGSMIIIS_LNKMOD_MASK ?
+			DUPLEX_FULL : DUPLEX_HALF;
+
+	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+}
+
+static const struct phylink_pcs_ops dwmac1000_mii_pcs_ops = {
+	.pcs_validate = dwmac1000_mii_pcs_validate,
+	.pcs_config = dwmac1000_mii_pcs_config,
+	.pcs_get_state = dwmac1000_mii_pcs_get_state,
+};
+
+static struct phylink_pcs *
+dwmac1000_phylink_select_pcs(struct stmmac_priv *priv,
+			     phy_interface_t interface)
+{
+	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
+	    priv->hw->pcs & STMMAC_PCS_SGMII)
+		return &priv->hw->mac_pcs;
+
+	return NULL;
 }
 
 static void dwmac1000_debug(struct stmmac_priv *priv, void __iomem *ioaddr,
@@ -499,6 +533,7 @@ static void dwmac1000_set_mac_loopback(void __iomem *ioaddr, bool enable)
 
 const struct stmmac_ops dwmac1000_ops = {
 	.core_init = dwmac1000_core_init,
+	.phylink_select_pcs = dwmac1000_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac1000_rx_ipc_enable,
 	.dump_regs = dwmac1000_dump_regs,
@@ -514,7 +549,6 @@ const struct stmmac_ops dwmac1000_ops = {
 	.set_eee_pls = dwmac1000_set_eee_pls,
 	.debug = dwmac1000_debug,
 	.pcs_ctrl_ane = dwmac1000_ctrl_ane,
-	.pcs_get_adv_lp = dwmac1000_get_adv_lp,
 	.set_mac_loopback = dwmac1000_set_mac_loopback,
 };
 
@@ -549,5 +583,8 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
+	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
+	mac->mac_pcs.neg_mode = true;
+
 	return 0;
 }
Signature · 1 line
2.30.2
#7 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Quoted 75 lines · Andrew Halaney
Quoted 62 lines · Russell King (Oracle)

Convert dwmac1000 sgmii/rgmii "pcs" implementation to use a phylink_pcs so we can eventually get rid of the exceptional paths that conflict with phylink.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 113 ++++++++++++------
 1 file changed, 75 insertions(+), 38 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +1 -33
index d413d76a8936..4a0572d5f865 100644
@@ -16,6 +16,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
+#include <linux/phylink.h>
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac1000.h"
@@ -261,39 +262,6 @@ static void dwmac1000_pmt(struct mac_device_info *hw, unsigned long mode)
 	writel(pmt, ioaddr + GMAC_PMT);
 }
 
-/* RGMII or SMII interface */
-static void dwmac1000_rgsmii(void __iomem *ioaddr, struct stmmac_extra_stats *x)
-{
-	u32 status;
-
-	status = readl(ioaddr + GMAC_RGSMIIIS);
-	x->irq_rgmii_n++;
-
-	/* Check the link status */
-	if (status & GMAC_RGSMIIIS_LNKSTS) {
-		int speed_value;
-
-		x->pcs_link = 1;
-
-		speed_value = ((status & GMAC_RGSMIIIS_SPEED) >>
-			       GMAC_RGSMIIIS_SPEED_SHIFT);
-		if (speed_value == GMAC_RGSMIIIS_SPEED_125)
-			x->pcs_speed = SPEED_1000;
-		else if (speed_value == GMAC_RGSMIIIS_SPEED_25)
-			x->pcs_speed = SPEED_100;
-		else
-			x->pcs_speed = SPEED_10;
-
-		x->pcs_duplex = (status & GMAC_RGSMIIIS_LNKMOD_MASK);
-
-		pr_info("Link is Up - %d/%s\n", (int)x->pcs_speed,
-			x->pcs_duplex ? "Full" : "Half");
-	} else {
-		x->pcs_link = 0;
-		pr_info("Link is Down\n");
-	}
-}
-
 static int dwmac1000_irq_status(struct mac_device_info *hw,
 				struct stmmac_extra_stats *x)
 {

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c static int dwmac1000_irq_status(struct mac_device_info *hw, · line 308

Quoted 8 lines
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +3 -2
@@ -335,8 +303,12 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
 
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac1000_rgsmii(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_RGSMIIIS);

This seems to me that you're doing the TODO here? Maybe I'm misunderstanding... maybe not :)

Quoted 1 line

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c static int dwmac1000_irq_status(struct mac_device_info *hw, · line 309

Quoted 1 line
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +1 -0
+		phylink_pcs_change(&hw->mac_pcs, false);

Continuing to read through this all, sorry for the double reply and possibly dumb question. Should we be passing in false unconditionally here?

Prior code had checked GMAC_RGSMIIIS & GMAC_RGSMIIIS_LNKSTS to determine if the link was up/down, which seem logical to pass in here too. Reading the kerneldoc for phylink_pcs_change() I'm not totally positive if we're in the "otherwise pass false" state here or some other detail I'm missing though (it seems that maybe we're just sort of kicking the can to phylink_resolve() which ends up calling into the get_state() callback?)

Quoted 107 lines
Quoted 107 lines
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +70 -3
+		x->irq_rgmii_n++;
+	}
 
 	return ret;
 }
@@ -404,9 +376,71 @@ static void dwmac1000_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
 }
 
-static void dwmac1000_get_adv_lp(void __iomem *ioaddr, struct rgmii_adv *adv)
+static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
+				      unsigned long *supported,
+				      const struct phylink_link_state *state)
+{
+	/* Only support in-band */
+	if (!test_bit(ETHTOOL_LINK_MODE_Autoneg_BIT, state->advertising))
+		return -EINVAL;
+
+	return 0;
+}
+
+static int dwmac1000_mii_pcs_config(struct phylink_pcs *pcs,
+				    unsigned int neg_mode,
+				    phy_interface_t interface,
+				    const unsigned long *advertising,
+				    bool permit_pause_to_mac)
 {
-	dwmac_get_adv_lp(ioaddr, GMAC_PCS_BASE, adv);
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+
+	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
+				GMAC_PCS_BASE);
+}
+
+static void dwmac1000_mii_pcs_get_state(struct phylink_pcs *pcs,
+					struct phylink_link_state *state)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	unsigned int spd_clk;
+	u32 status;
+
+	status = readl(hw->pcsr + GMAC_RGSMIIIS);
+
+	state->link = status & GMAC_RGSMIIIS_LNKSTS;
+	if (!state->link)
+		return;
+
+	spd_clk = FIELD_GET(GMAC_RGSMIIIS_SPEED, status);
+	if (spd_clk == GMAC_RGSMIIIS_SPEED_125)
+		state->speed = SPEED_1000;
+	else if (spd_clk == GMAC_RGSMIIIS_SPEED_25)
+		state->speed = SPEED_100;
+	else if (spd_clk == GMAC_RGSMIIIS_SPEED_2_5)
+		state->speed = SPEED_10;
+
+	state->duplex = status & GMAC_RGSMIIIS_LNKMOD_MASK ?
+			DUPLEX_FULL : DUPLEX_HALF;
+
+	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+}
+
+static const struct phylink_pcs_ops dwmac1000_mii_pcs_ops = {
+	.pcs_validate = dwmac1000_mii_pcs_validate,
+	.pcs_config = dwmac1000_mii_pcs_config,
+	.pcs_get_state = dwmac1000_mii_pcs_get_state,
+};
+
+static struct phylink_pcs *
+dwmac1000_phylink_select_pcs(struct stmmac_priv *priv,
+			     phy_interface_t interface)
+{
+	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
+	    priv->hw->pcs & STMMAC_PCS_SGMII)
+		return &priv->hw->mac_pcs;
+
+	return NULL;
 }
 
 static void dwmac1000_debug(struct stmmac_priv *priv, void __iomem *ioaddr,
@@ -499,6 +533,7 @@ static void dwmac1000_set_mac_loopback(void __iomem *ioaddr, bool enable)
 
 const struct stmmac_ops dwmac1000_ops = {
 	.core_init = dwmac1000_core_init,
+	.phylink_select_pcs = dwmac1000_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac1000_rx_ipc_enable,
 	.dump_regs = dwmac1000_dump_regs,
@@ -514,7 +549,6 @@ const struct stmmac_ops dwmac1000_ops = {
 	.set_eee_pls = dwmac1000_set_eee_pls,
 	.debug = dwmac1000_debug,
 	.pcs_ctrl_ane = dwmac1000_ctrl_ane,
-	.pcs_get_adv_lp = dwmac1000_get_adv_lp,
 	.set_mac_loopback = dwmac1000_set_mac_loopback,
 };
 
@@ -549,5 +583,8 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
+	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
+	mac->mac_pcs.neg_mode = true;
+
 	return 0;
 }
Signature · 1 line
2.30.2
#8 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Quoted 9 lines · Andrew Halaney
Quoted 4 lines · Andrew Halaney

This seems to me that you're doing the TODO here? Maybe I'm misunderstanding... maybe not :)

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c static int dwmac1000_irq_status(struct mac_device_info *hw, · line 309

Quoted 1 line
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +1 -0
+		phylink_pcs_change(&hw->mac_pcs, false);

Continuing to read through this all, sorry for the double reply and possibly dumb question. Should we be passing in false unconditionally here?

It depends whether there is a way to get the current status of the link without side effects (e.g. where a read clears a latched-low link status.) If that's not possible, then passing "false" is safe provided there aren't any spurious interrupts, since we'll always assume that the link has dropped. If there are spurious interrupts, then the link will go down/up each time there's a spurious interrupt. Even so, that's better than missing a change in the link status which may result in loss of link without manual intervention.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#9 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

In reply to Andrew Halaney
Quoted 5 lines · Andrew Halaney
Quoted 4 lines · Russell King (Oracle)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -0
@@ -335,8 +303,12 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
 
Quoted 8 lines

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c static int dwmac1000_irq_status(struct mac_device_info *hw, · line 308

Quoted 5 lines
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +3 -2
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac1000_rgsmii(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_RGSMIIIS);

This seems to me that you're doing the TODO here? Maybe I'm misunderstanding... maybe not :)

Please trim your replies.

These two lines come from Serge - please ask him why it's marked as a TODO. I assume he has a reason. Thanks.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#10 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Hi Russell, Andrew

Quoted 16 lines · Russell King (Oracle)
Quoted 6 lines · Andrew Halaney
Quoted 5 lines · Russell King (Oracle)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -0
@@ -335,8 +303,12 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
 
Quoted 8 lines

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c static int dwmac1000_irq_status(struct mac_device_info *hw, · line 308

Quoted 5 lines
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +3 -2
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac1000_rgsmii(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_RGSMIIIS);

This seems to me that you're doing the TODO here? Maybe I'm misunderstanding... maybe not :)

Quoted 4 lines

Please trim your replies.

These two lines come from Serge - please ask him why it's marked as a TODO. I assume he has a reason. Thanks.

The statement below the "TODO..." comment was supposed to be a quick-fix of the interrupts flood happening due to the uncleared RGSMIIIS IRQ flag. Of course dummy-reading in the IRQ handler with no action required to handle the IRQ wouldn't be the best solution (despite of having the phylink_pcs_change() called), especially seeing there is the dwmac_pcs_isr() method, which name implies the PCS IRQ handling. At least we could have incremented the stmmac_extra_stats::irq_rgmii_n counter in there. So what I meant TODO here was to move the RGSMIIIS IRQ handling in dwmac_pcs_isr().

I know that the dwmac_pcs_isr() method has been created around the cross-IP-cores PCS implementation, but as I mentioned several times the tx_config_reg[15:0] part of the GMAC_RGSMIIIS/MAC_PHYIF_Control_Status registers is the same on both DW GMAC and DW QoS Eth:
tx_config_reg[0]: LNKMOD
tx_config_reg[1:2]: LNKSPEED

tx_config_reg[3]:   LNKSTS
tx_config_reg[4]:   JABTO (Jabber Timeout, specific to SMII)
tx_config_reg[5]:   FALSCARDET (False Carrier Detected, specific to SMII)

Should we have a DW IP-core-specific getter like stmmac_ops::pcs_get_config_reg() which would return the tx_config_reg[15:0] field then we could have cleared the IRQ by just calling it, we could have had the fields generically parsed in the dwmac_pcs_isr() handler and in the phylink_pcs_ops::pcs_get_state(). Thus the entire struct phylink_pcs_ops definition could be moved to the stmmac_pcs.c module simplifying the DW GMAC and DW QoS Eth hardware-dependent code.

In this regard there is another change which would be required (and frankly would make the code simpler). Instead of passing the CSRs-base address to the dwmac_pcs_isr()/dwmac_pcs_config()/dwmac_pcs_get_state() methods, we could pre-define the PCS registers base address as it's done for the PTP/MMC/EST implementation in the driver. Here is the brief change description:
1. add stmmac_regs_off::pcs_off field (hwif.h)
2. add stmmac_priv::pcsaddr field (stmmac.h)
3. initialize the stmmac_regs_off::pcs_off field for the DW GMAC and DW QoS Eth IP-cores in the stmmac_hw array (hwif.c)
4. initialize the stmmac_priv::pcsaddr field in the stmmac_hwif_init() method as it's done for stmmac_priv::{ptpaddr,mmcaddr,estaddr}.
5. use the PCS-base address in the stmmac_pcs.c module.

As a result (unless I've missed something) we'll be able to move almost the entire internal PCS implementation to the stmmac_pcs.c module (except the tx_config_reg[] data getter).

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#11 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Quoted 8 lines · Serge Semin

Hi Russell, Andrew

Should we have a DW IP-core-specific getter like stmmac_ops::pcs_get_config_reg() which would return the tx_config_reg[15:0] field then we could have cleared the IRQ by just calling it, we could have had the fields generically parsed in the dwmac_pcs_isr() handler and in the phylink_pcs_ops::pcs_get_state().

pcs_get_state() is not supposed to get some cached state, but is supposed to return the real state at the time that it is called.

There's a good reason for this - dealing with latched-low link failed indications, it's necessary that pcs_get_state() reports that the link failed if _sometime_ between the last time it was called and the next time the link has failed.

So, I'm afraid your idea of simplifying it doesn't sound to me like a good idea.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#12 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Quoted 11 lines · Russell King (Oracle)
Quoted 8 lines · Serge Semin

Hi Russell, Andrew

Should we have a DW IP-core-specific getter like stmmac_ops::pcs_get_config_reg() which would return the tx_config_reg[15:0] field then we could have cleared the IRQ by just calling it, we could have had the fields generically parsed in the dwmac_pcs_isr() handler and in the phylink_pcs_ops::pcs_get_state().

Quoted 1 line

pcs_get_state() is not supposed to get some cached state,

It won't.

Quoted 2 lines

but is
supposed to return the real state at the time that it is called.

The idea is to implement the tx_config_reg[15:0] getter for DW GMAC and DW QoS Eth. It will return the current link status retrieved from the GMAC_RGSMIIIS and GMAC_PHYIF_CONTROL_STATUS registers. Like this:

GMAC:
u16 dwmac1000_pcs_get_config_reg(struct stmmac_priv *priv)

{
	return readl(priv->ioaddr + GMAC_RGSMIIIS);
}

DW QoS Eth:
u16 dwmac1000_pcs_get_config_reg(struct stmmac_priv *priv)

{
	return readl(priv->ioaddr + GMAC_PHYIF_CONTROL_STATUS) >> 16;
}

Then the dwmac_pcs_isr() can be updated as follows:

static inline void dwmac_pcs_isr(void __iomem *ioaddr, u32 reg,
unsigned int intr_status, struct stmmac_extra_stats *x)

{
	...

	if (intr_status & PCS_RGSMIIIS_IRQ) {
		x->irq_rgmii_n++;
		/* The next will clear the PCS_RGSMIIIS_IRQ IRQ
		 * status. (It is done instead of dummy-reading the
		 * GMAC_RGSMIIIS/GMAC_PHYIF_CONTROL_STATUS registers
		 * in the DW GMAC/QoS Eth IRQ handlers.)
		 */
		(void)stmmac_pcs_get_config_reg(priv);
	}
}

Similarly the dwmac_pcs_get_state() method can now use the stmmac_pcs_get_config_reg() function to retrieve the link state and parse the data in a generic manner. Thus everything what is currently done in dwmac1000_mii_pcs_get_state() and dwmac4_mii_pcs_get_state() could be moved to dwmac_pcs_get_state(). By providing a single STMMAC-driver callback stmmac_ops::pcs_get_config_reg(), we'll be able to define the phylink_pcs_ops::{pcs_validate,pcs_config,pcs_get_state} methods in the stmmac_pcs.c file.

Quoted 8 lines

There's a good reason for this - dealing with latched-low link failed indications, it's necessary that pcs_get_state() reports that the link failed if _sometime_ between the last time it was called and the next time the link has failed.

So, I'm afraid your idea of simplifying it doesn't sound to me like a good idea.

No caching or latched link state indications. Both the GMAC_RGSMIIIS and GMAC_PHYIF_CONTROL_STATUS registers contain the actual link state retrieved the PHY. stmmac_pcs_get_config_reg() will just return the current link state.

Perhaps my suggestion might haven't been well described. Providing the patches with the respective changes shall better describe what was meant. So in a few days I'll submit an incremental patch(es) with the proposed modifications for your series.

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#13 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Quoted 28 lines · Serge Semin
Quoted 10 lines · Russell King (Oracle)
Quoted 8 lines · Serge Semin

Hi Russell, Andrew

Should we have a DW IP-core-specific getter like stmmac_ops::pcs_get_config_reg() which would return the tx_config_reg[15:0] field then we could have cleared the IRQ by just calling it, we could have had the fields generically parsed in the dwmac_pcs_isr() handler and in the phylink_pcs_ops::pcs_get_state().

[...]

Quoted 8 lines

There's a good reason for this - dealing with latched-low link failed indications, it's necessary that pcs_get_state() reports that the link failed if _sometime_ between the last time it was called and the next time the link has failed.

So, I'm afraid your idea of simplifying it doesn't sound to me like a good idea.

No caching or latched link state indications. Both the GMAC_RGSMIIIS and GMAC_PHYIF_CONTROL_STATUS registers contain the actual link state retrieved the PHY. stmmac_pcs_get_config_reg() will just return the current link state.

Quoted 4 lines

Perhaps my suggestion might haven't been well described. Providing the patches with the respective changes shall better describe what was meant. So in a few days I'll submit an incremental patch(es) with the proposed modifications for your series

The incremental patchset is ready. I need to give it some more tests, then rebase onto the kernel 6.10. It'll be done in one-two days.

Thanks
-Serge(y)
.

Quoted 7 lines

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#14 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Hi Russell

Quoted 35 lines · Serge Semin
Quoted 28 lines · Serge Semin
Quoted 10 lines · Russell King (Oracle)
Quoted 8 lines · Serge Semin

Hi Russell, Andrew

Should we have a DW IP-core-specific getter like stmmac_ops::pcs_get_config_reg() which would return the tx_config_reg[15:0] field then we could have cleared the IRQ by just calling it, we could have had the fields generically parsed in the dwmac_pcs_isr() handler and in the phylink_pcs_ops::pcs_get_state().

[...]

Quoted 8 lines

There's a good reason for this - dealing with latched-low link failed indications, it's necessary that pcs_get_state() reports that the link failed if _sometime_ between the last time it was called and the next time the link has failed.

So, I'm afraid your idea of simplifying it doesn't sound to me like a good idea.

No caching or latched link state indications. Both the GMAC_RGSMIIIS and GMAC_PHYIF_CONTROL_STATUS registers contain the actual link state retrieved the PHY. stmmac_pcs_get_config_reg() will just return the current link state.

Quoted 4 lines

Perhaps my suggestion might haven't been well described. Providing the patches with the respective changes shall better describe what was meant. So in a few days I'll submit an incremental patch(es) with the proposed modifications for your series

Quoted 3 lines

The incremental patchset is ready. I need to give it some more tests, then rebase onto the kernel 6.10. It'll be done in one-two days.

It turned out I has created my series on top of your v1 series. I just finished rebasing it onto v2. The only thing left is to test it out. I'll do that today and then submit the series in-reply to your v2 email thread. Sorry for making you wait once again.

-Serge

Quoted 11 lines

Thanks
-Serge(y)
.

Quoted 7 lines

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#15 lore raw

Re: [PATCH RFC net-next v2 3/8] net: stmmac: dwmac1000: convert sgmii/rgmii "pcs" to phylink

Hi Russel

Quoted 38 lines · Serge Semin

Hi Russell

Quoted 34 lines · Serge Semin
Quoted 32 lines · Serge Semin
Quoted 10 lines · Russell King (Oracle)
Quoted 8 lines · Serge Semin

Hi Russell, Andrew

Should we have a DW IP-core-specific getter like stmmac_ops::pcs_get_config_reg() which would return the tx_config_reg[15:0] field then we could have cleared the IRQ by just calling it, we could have had the fields generically parsed in the dwmac_pcs_isr() handler and in the phylink_pcs_ops::pcs_get_state().

[...]

Quoted 8 lines

There's a good reason for this - dealing with latched-low link failed indications, it's necessary that pcs_get_state() reports that the link failed if _sometime_ between the last time it was called and the next time the link has failed.

So, I'm afraid your idea of simplifying it doesn't sound to me like a good idea.

No caching or latched link state indications. Both the GMAC_RGSMIIIS and GMAC_PHYIF_CONTROL_STATUS registers contain the actual link state retrieved the PHY. stmmac_pcs_get_config_reg() will just return the current link state.

Perhaps my suggestion might haven't been well described. Providing the patches with the respective changes shall better describe what was meant. So in a few days I'll submit an incremental patch(es) with the proposed modifications for your series

Quoted 8 lines
Quoted 3 lines

The incremental patchset is ready. I need to give it some more tests, then rebase onto the kernel 6.10. It'll be done in one-two days.

It turned out I has created my series on top of your v1 series. I just finished rebasing it onto v2. The only thing left is to test it out. I'll do that today and then submit the series in-reply to your v2 email thread. Sorry for making you wait once again.

Finally I've done that.
https://lore.kernel.org/netdev/[email protected]

The update has turned to be a bit more bulky than I intended, but the resultant code looks neat and small consolidated in the stmmac_pcs.c/stmmac_pcs.h files with just three DW *MAC HW-abstraction callbacks defined in the DW GMAC and DW QoS Eth core modules.

I did my best to provide the changes in the incremental manner with no driver breakage. Hopefully I didn't fail that.) As before you can re-shuffle the patches and the change they content whatever you think would be better in the final series.

-Serge(y)

#16 lore raw

RFC PATCH v2 4/8 net: stmmac: dwmac1000: move PCS interrupt control

In reply to Russell King (Oracle)

Control the PCS interrupt mask from phylink's pcs_enable() and pcs_disable() methods rather than relying on driver variables.

This assumes that GMAC_INT_DISABLE_RGMII, GMAC_INT_DISABLE_PCSLINK and GMAC_INT_DISABLE_PCSAN are all relevant to the PCS.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 33 +++++++++++++++----
 1 file changed, 27 insertions(+), 6 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +27 -6
index 4a0572d5f865..3ba65ea3e46f 100644
@@ -56,12 +56,7 @@ static void dwmac1000_core_init(struct mac_device_info *hw,
 	writel(value, ioaddr + GMAC_CONTROL);
 
 	/* Mask GMAC interrupts */
-	value = GMAC_INT_DEFAULT_MASK;
-
-	if (hw->pcs)
-		value &= ~GMAC_INT_DISABLE_PCS;
-
-	writel(value, ioaddr + GMAC_INT_MASK);
+	writel(GMAC_INT_DEFAULT_MASK, ioaddr + GMAC_INT_MASK);
 
 #ifdef STMMAC_VLAN_TAG_USED
 	/* Tag detection without filtering */
@@ -387,6 +382,30 @@ static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
 	return 0;
 }
 
+static int dwmac1000_mii_pcs_enable(struct phylink_pcs *pcs)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_mask;
+
+	intr_mask = readl(ioaddr + GMAC_INT_MASK);
+	intr_mask &= ~GMAC_INT_DISABLE_PCS;
+	writel(intr_mask, ioaddr + GMAC_INT_MASK);
+
+	return 0;
+}
+
+static void dwmac1000_mii_pcs_disable(struct phylink_pcs *pcs)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_mask;
+
+	intr_mask = readl(ioaddr + GMAC_INT_MASK);
+	intr_mask |= GMAC_INT_DISABLE_PCS;
+	writel(intr_mask, ioaddr + GMAC_INT_MASK);
+}
+
 static int dwmac1000_mii_pcs_config(struct phylink_pcs *pcs,
 				    unsigned int neg_mode,
 				    phy_interface_t interface,
@@ -428,6 +447,8 @@ static void dwmac1000_mii_pcs_get_state(struct phylink_pcs *pcs,
 
 static const struct phylink_pcs_ops dwmac1000_mii_pcs_ops = {
 	.pcs_validate = dwmac1000_mii_pcs_validate,
+	.pcs_enable = dwmac1000_mii_pcs_enable,
+	.pcs_disable = dwmac1000_mii_pcs_disable,
 	.pcs_config = dwmac1000_mii_pcs_config,
 	.pcs_get_state = dwmac1000_mii_pcs_get_state,
 };
Signature · 1 line
2.30.2
#17 lore raw

RFC PATCH v2 5/8 net: stmmac: dwmac4: convert sgmii/rgmii "pcs" to phylink

In reply to Russell King (Oracle)

Convert dwmac4 sgmii/rgmii "pcs" implementation to use a phylink_pcs so we can eventually get rid of the exceptional paths that conflict with phylink.

---
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 102 ++++++++++++------
 1 file changed, 72 insertions(+), 30 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +72 -30
index dbd9f93b2460..cb99cb69c52b 100644
@@ -14,6 +14,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
+#include <linux/phylink.h>
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac4.h"
@@ -758,42 +759,76 @@ static void dwmac4_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
 }
 
-static void dwmac4_get_adv_lp(void __iomem *ioaddr, struct rgmii_adv *adv)
+static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
+				   unsigned long *supported,
+				   const struct phylink_link_state *state)
 {
-	dwmac_get_adv_lp(ioaddr, GMAC_PCS_BASE, adv);
+	/* Only support in-band */
+	if (!test_bit(ETHTOOL_LINK_MODE_Autoneg_BIT, state->advertising))
+		return -EINVAL;
+
+	return 0;
 }
 
-/* RGMII or SMII interface */
-static void dwmac4_phystatus(void __iomem *ioaddr, struct stmmac_extra_stats *x)
+static int dwmac4_mii_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
+				 phy_interface_t interface,
+				 const unsigned long *advertising,
+				 bool permit_pause_to_mac)
 {
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+
+	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
+				GMAC_PCS_BASE);
+}
+
+static void dwmac4_mii_pcs_get_state(struct phylink_pcs *pcs,
+				     struct phylink_link_state *state)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	unsigned int clk_spd;
 	u32 status;
 
-	status = readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
-	x->irq_rgmii_n++;
+	status = readl(hw->pcsr + GMAC_PHYIF_CONTROL_STATUS);
+
+	state->link = !!(status & GMAC_PHYIF_CTRLSTATUS_LNKSTS);
+	if (!state->link)
+		return;
 
-	/* Check the link status */
-	if (status & GMAC_PHYIF_CTRLSTATUS_LNKSTS) {
-		int speed_value;
+	clk_spd = FIELD_GET(GMAC_PHYIF_CTRLSTATUS_SPEED, status);
+	if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_125)
+		state->speed = SPEED_1000;
+	else if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_25)
+		state->speed = SPEED_100;
+	else if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_2_5)
+		state->speed = SPEED_10;
+
+	/* FIXME: Is this even correct?
+	 * GMAC_PHYIF_CTRLSTATUS_TC = BIT(0)
+	 * GMAC_PHYIF_CTRLSTATUS_LNKMOD = BIT(16)
+	 * GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK = 1
+	 *
+	 * The result is, we test bit 0 for the duplex setting.
+	 */
+	state->duplex = status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK ?
+			DUPLEX_FULL : DUPLEX_HALF;
 
-		x->pcs_link = 1;
+	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+}
 
-		speed_value = ((status & GMAC_PHYIF_CTRLSTATUS_SPEED) >>
-			       GMAC_PHYIF_CTRLSTATUS_SPEED_SHIFT);
-		if (speed_value == GMAC_PHYIF_CTRLSTATUS_SPEED_125)
-			x->pcs_speed = SPEED_1000;
-		else if (speed_value == GMAC_PHYIF_CTRLSTATUS_SPEED_25)
-			x->pcs_speed = SPEED_100;
-		else
-			x->pcs_speed = SPEED_10;
+static const struct phylink_pcs_ops dwmac4_mii_pcs_ops = {
+	.pcs_validate = dwmac4_mii_pcs_validate,
+	.pcs_config = dwmac4_mii_pcs_config,
+	.pcs_get_state = dwmac4_mii_pcs_get_state,
+};
 
-		x->pcs_duplex = (status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK);
+static struct phylink_pcs *
+dwmac4_phylink_select_pcs(struct stmmac_priv *priv, phy_interface_t interface)
+{
+	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
+	    priv->hw->pcs & STMMAC_PCS_SGMII)
+		return &priv->hw->mac_pcs;
 
-		pr_info("Link is Up - %d/%s\n", (int)x->pcs_speed,
-			x->pcs_duplex ? "Full" : "Half");
-	} else {
-		x->pcs_link = 0;
-		pr_info("Link is Down\n");
-	}
+	return NULL;
 }
 
 static int dwmac4_irq_mtl_status(struct stmmac_priv *priv,
@@ -867,8 +902,12 @@ static int dwmac4_irq_status(struct mac_device_info *hw,
 	}
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac4_phystatus(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
+		phylink_pcs_change(&hw->mac_pcs, false);
+		x->irq_rgmii_n++;
+	}
 
 	return ret;
 }
@@ -1186,6 +1225,7 @@ static void dwmac4_set_hw_vlan_mode(struct mac_device_info *hw)
 const struct stmmac_ops dwmac4_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1210,7 +1250,6 @@ const struct stmmac_ops dwmac4_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.set_mac_loopback = dwmac4_set_mac_loopback,
@@ -1230,6 +1269,7 @@ const struct stmmac_ops dwmac4_ops = {
 const struct stmmac_ops dwmac410_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1254,7 +1294,6 @@ const struct stmmac_ops dwmac410_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.flex_pps_config = dwmac5_flex_pps_config,
@@ -1278,6 +1317,7 @@ const struct stmmac_ops dwmac410_ops = {
 const struct stmmac_ops dwmac510_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1302,7 +1342,6 @@ const struct stmmac_ops dwmac510_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.safety_feat_config = dwmac5_safety_feat_config,
@@ -1391,5 +1430,8 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
+	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
+	mac->mac_pcs.neg_mode = true;
+
 	return 0;
 }
Signature · 1 line
2.30.2
#18 lore raw

Re: [PATCH RFC net-next v2 5/8] net: stmmac: dwmac4: convert sgmii/rgmii "pcs" to phylink

Quoted 76 lines · Russell King (Oracle)

Convert dwmac4 sgmii/rgmii "pcs" implementation to use a phylink_pcs so we can eventually get rid of the exceptional paths that conflict with phylink.

---
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 102 ++++++++++++------
 1 file changed, 72 insertions(+), 30 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +36 -9
index dbd9f93b2460..cb99cb69c52b 100644
@@ -14,6 +14,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
+#include <linux/phylink.h>
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac4.h"
@@ -758,42 +759,76 @@ static void dwmac4_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
 }
 
-static void dwmac4_get_adv_lp(void __iomem *ioaddr, struct rgmii_adv *adv)
+static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
+				   unsigned long *supported,
+				   const struct phylink_link_state *state)
 {
-	dwmac_get_adv_lp(ioaddr, GMAC_PCS_BASE, adv);
+	/* Only support in-band */
+	if (!test_bit(ETHTOOL_LINK_MODE_Autoneg_BIT, state->advertising))
+		return -EINVAL;
+
+	return 0;
 }
 
-/* RGMII or SMII interface */
-static void dwmac4_phystatus(void __iomem *ioaddr, struct stmmac_extra_stats *x)
+static int dwmac4_mii_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
+				 phy_interface_t interface,
+				 const unsigned long *advertising,
+				 bool permit_pause_to_mac)
 {
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+
+	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
+				GMAC_PCS_BASE);
+}
+
+static void dwmac4_mii_pcs_get_state(struct phylink_pcs *pcs,
+				     struct phylink_link_state *state)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	unsigned int clk_spd;
 	u32 status;
 
-	status = readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
-	x->irq_rgmii_n++;
+	status = readl(hw->pcsr + GMAC_PHYIF_CONTROL_STATUS);
+
+	state->link = !!(status & GMAC_PHYIF_CTRLSTATUS_LNKSTS);
+	if (!state->link)
+		return;
 
-	/* Check the link status */
-	if (status & GMAC_PHYIF_CTRLSTATUS_LNKSTS) {
-		int speed_value;
+	clk_spd = FIELD_GET(GMAC_PHYIF_CTRLSTATUS_SPEED, status);
+	if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_125)
+		state->speed = SPEED_1000;
+	else if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_25)
+		state->speed = SPEED_100;
+	else if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_2_5)
+		state->speed = SPEED_10;

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c static void dwmac4_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral, · line 813

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +10 -0
+
+	/* FIXME: Is this even correct?
+	 * GMAC_PHYIF_CTRLSTATUS_TC = BIT(0)
+	 * GMAC_PHYIF_CTRLSTATUS_LNKMOD = BIT(16)
+	 * GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK = 1
+	 *
+	 * The result is, we test bit 0 for the duplex setting.
+	 */
+	state->duplex = status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK ?
+			DUPLEX_FULL : DUPLEX_HALF;

My gut feeling is that this is/was wrong, and the LNKMOD_MASK expects you've shifted / got the field out etc...

Sneh, Abhishek, can you confirm this for us? I'd appreciate it.

I need to run for the evening soon, but tested taking sa8775p-ride, making it have managed = "in-band-status", and remove the HAS_INTEGRATED_PCS stuff... and then all of a sudden the link acts as if its half duplex (but works otherwise I think? need to test more throughly):

    [   11.458385] qcom-ethqos 23040000.ethernet end0: Link is Up - 1Gbps/Half - flow control rx/tx

So I think it is probably wrong, and if I understand correctly most of the special treatment for the qualcomm driver can be dropped? I know it was mentioned that all the 2.5 Gpbs stuff is a Qualcomm addition (and that you're chasing down some answers with the hardware team), but hopefully you can confirm the register's bitfields for us here!

Quoted 38 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +14 -16
-		x->pcs_link = 1;
+	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+}
 
-		speed_value = ((status & GMAC_PHYIF_CTRLSTATUS_SPEED) >>
-			       GMAC_PHYIF_CTRLSTATUS_SPEED_SHIFT);
-		if (speed_value == GMAC_PHYIF_CTRLSTATUS_SPEED_125)
-			x->pcs_speed = SPEED_1000;
-		else if (speed_value == GMAC_PHYIF_CTRLSTATUS_SPEED_25)
-			x->pcs_speed = SPEED_100;
-		else
-			x->pcs_speed = SPEED_10;
+static const struct phylink_pcs_ops dwmac4_mii_pcs_ops = {
+	.pcs_validate = dwmac4_mii_pcs_validate,
+	.pcs_config = dwmac4_mii_pcs_config,
+	.pcs_get_state = dwmac4_mii_pcs_get_state,
+};
 
-		x->pcs_duplex = (status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK);
+static struct phylink_pcs *
+dwmac4_phylink_select_pcs(struct stmmac_priv *priv, phy_interface_t interface)
+{
+	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
+	    priv->hw->pcs & STMMAC_PCS_SGMII)
+		return &priv->hw->mac_pcs;
 
-		pr_info("Link is Up - %d/%s\n", (int)x->pcs_speed,
-			x->pcs_duplex ? "Full" : "Half");
-	} else {
-		x->pcs_link = 0;
-		pr_info("Link is Down\n");
-	}
+	return NULL;
 }
 
 static int dwmac4_irq_mtl_status(struct stmmac_priv *priv,

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c static int dwmac4_irq_status(struct mac_device_info *hw, · line 908

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +4 -2
@@ -867,8 +902,12 @@ static int dwmac4_irq_status(struct mac_device_info *hw,
 	}
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac4_phystatus(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
+		phylink_pcs_change(&hw->mac_pcs, false);

I'll just highlight it here, but same question as the dwmac1000 change. We can discuss that question there, and if anything changes apply it here too. It is probably fine and I'm fussing over nothing.

Quoted 65 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +8 -3
+		x->irq_rgmii_n++;
+	}
 
 	return ret;
 }
@@ -1186,6 +1225,7 @@ static void dwmac4_set_hw_vlan_mode(struct mac_device_info *hw)
 const struct stmmac_ops dwmac4_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1210,7 +1250,6 @@ const struct stmmac_ops dwmac4_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.set_mac_loopback = dwmac4_set_mac_loopback,
@@ -1230,6 +1269,7 @@ const struct stmmac_ops dwmac4_ops = {
 const struct stmmac_ops dwmac410_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1254,7 +1294,6 @@ const struct stmmac_ops dwmac410_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.flex_pps_config = dwmac5_flex_pps_config,
@@ -1278,6 +1317,7 @@ const struct stmmac_ops dwmac410_ops = {
 const struct stmmac_ops dwmac510_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1302,7 +1342,6 @@ const struct stmmac_ops dwmac510_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.safety_feat_config = dwmac5_safety_feat_config,
@@ -1391,5 +1430,8 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
+	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
+	mac->mac_pcs.neg_mode = true;
+
 	return 0;
 }
Signature · 1 line
2.30.2
#19 lore raw

Re: [PATCH RFC net-next v2 5/8] net: stmmac: dwmac4: convert sgmii/rgmii "pcs" to phylink

Quoted 91 lines · Andrew Halaney
Quoted 76 lines · Russell King (Oracle)

Convert dwmac4 sgmii/rgmii "pcs" implementation to use a phylink_pcs so we can eventually get rid of the exceptional paths that conflict with phylink.

---
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 102 ++++++++++++------
 1 file changed, 72 insertions(+), 30 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +36 -9
index dbd9f93b2460..cb99cb69c52b 100644
@@ -14,6 +14,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
+#include <linux/phylink.h>
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac4.h"
@@ -758,42 +759,76 @@ static void dwmac4_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
 }
 
-static void dwmac4_get_adv_lp(void __iomem *ioaddr, struct rgmii_adv *adv)
+static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
+				   unsigned long *supported,
+				   const struct phylink_link_state *state)
 {
-	dwmac_get_adv_lp(ioaddr, GMAC_PCS_BASE, adv);
+	/* Only support in-band */
+	if (!test_bit(ETHTOOL_LINK_MODE_Autoneg_BIT, state->advertising))
+		return -EINVAL;
+
+	return 0;
 }
 
-/* RGMII or SMII interface */
-static void dwmac4_phystatus(void __iomem *ioaddr, struct stmmac_extra_stats *x)
+static int dwmac4_mii_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
+				 phy_interface_t interface,
+				 const unsigned long *advertising,
+				 bool permit_pause_to_mac)
 {
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+
+	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
+				GMAC_PCS_BASE);
+}
+
+static void dwmac4_mii_pcs_get_state(struct phylink_pcs *pcs,
+				     struct phylink_link_state *state)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	unsigned int clk_spd;
 	u32 status;
 
-	status = readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
-	x->irq_rgmii_n++;
+	status = readl(hw->pcsr + GMAC_PHYIF_CONTROL_STATUS);
+
+	state->link = !!(status & GMAC_PHYIF_CTRLSTATUS_LNKSTS);
+	if (!state->link)
+		return;
 
-	/* Check the link status */
-	if (status & GMAC_PHYIF_CTRLSTATUS_LNKSTS) {
-		int speed_value;
+	clk_spd = FIELD_GET(GMAC_PHYIF_CTRLSTATUS_SPEED, status);
+	if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_125)
+		state->speed = SPEED_1000;
+	else if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_25)
+		state->speed = SPEED_100;
+	else if (clk_spd == GMAC_PHYIF_CTRLSTATUS_SPEED_2_5)
+		state->speed = SPEED_10;

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c static void dwmac4_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral, · line 813

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +10 -0
+
+	/* FIXME: Is this even correct?
+	 * GMAC_PHYIF_CTRLSTATUS_TC = BIT(0)
+	 * GMAC_PHYIF_CTRLSTATUS_LNKMOD = BIT(16)
+	 * GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK = 1
+	 *
+	 * The result is, we test bit 0 for the duplex setting.
+	 */
+	state->duplex = status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK ?
+			DUPLEX_FULL : DUPLEX_HALF;

My gut feeling is that this is/was wrong, and the LNKMOD_MASK expects you've shifted / got the field out etc...

Quoted 2 lines

Sneh, Abhishek, can you confirm this for us? I'd appreciate it.

The status needs to be checked against Bit 16 which is GMAC_PHYIF_CTRLSTATUS_LNKMOD to determine the Duplex. In the above logic the status is checked against GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK (value 1) which actually determines the link is up/down

//Correct logic
state->duplex = status & GMAC_PHYIF_CTRLSTATUS_LNKMOD ?

			DUPLEX_FULL : DUPLEX_HALF;

Bit 16 LNKMOD Link Mode

This bit indicates the current mode of operation of the link. 0x0: HDUPLX
0x1: FDUPLX

Bit 1 LUD Link Up or Down

This bit indicates whether the link is up or down during transmission of configuration in the RGMII, SGMII, or SMII interface. 0x0: LINKDOWN
0x1: LINKUP

Quoted 134 lines

I need to run for the evening soon, but tested taking sa8775p-ride, making it have managed = "in-band-status", and remove the HAS_INTEGRATED_PCS stuff... and then all of a sudden the link acts as if its half duplex (but works otherwise I think? need to test more throughly):

    [   11.458385] qcom-ethqos 23040000.ethernet end0: Link is Up - 1Gbps/Half - flow control rx/tx

So I think it is probably wrong, and if I understand correctly most of the special treatment for the qualcomm driver can be dropped? I know it was mentioned that all the 2.5 Gpbs stuff is a Qualcomm addition (and that you're chasing down some answers with the hardware team), but hopefully you can confirm the register's bitfields for us here!

Quoted 38 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +14 -16
-		x->pcs_link = 1;
+	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+}
 
-		speed_value = ((status & GMAC_PHYIF_CTRLSTATUS_SPEED) >>
-			       GMAC_PHYIF_CTRLSTATUS_SPEED_SHIFT);
-		if (speed_value == GMAC_PHYIF_CTRLSTATUS_SPEED_125)
-			x->pcs_speed = SPEED_1000;
-		else if (speed_value == GMAC_PHYIF_CTRLSTATUS_SPEED_25)
-			x->pcs_speed = SPEED_100;
-		else
-			x->pcs_speed = SPEED_10;
+static const struct phylink_pcs_ops dwmac4_mii_pcs_ops = {
+	.pcs_validate = dwmac4_mii_pcs_validate,
+	.pcs_config = dwmac4_mii_pcs_config,
+	.pcs_get_state = dwmac4_mii_pcs_get_state,
+};
 
-		x->pcs_duplex = (status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK);
+static struct phylink_pcs *
+dwmac4_phylink_select_pcs(struct stmmac_priv *priv, phy_interface_t interface)
+{
+	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
+	    priv->hw->pcs & STMMAC_PCS_SGMII)
+		return &priv->hw->mac_pcs;
 
-		pr_info("Link is Up - %d/%s\n", (int)x->pcs_speed,
-			x->pcs_duplex ? "Full" : "Half");
-	} else {
-		x->pcs_link = 0;
-		pr_info("Link is Down\n");
-	}
+	return NULL;
 }
 
 static int dwmac4_irq_mtl_status(struct stmmac_priv *priv,

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c static int dwmac4_irq_status(struct mac_device_info *hw, · line 908

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +4 -2
@@ -867,8 +902,12 @@ static int dwmac4_irq_status(struct mac_device_info *hw,
 	}
 
 	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
-	if (intr_status & PCS_RGSMIIIS_IRQ)
-		dwmac4_phystatus(ioaddr, x);
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		/* TODO Dummy-read to clear the IRQ status */
+		readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
+		phylink_pcs_change(&hw->mac_pcs, false);

I'll just highlight it here, but same question as the dwmac1000 change. We can discuss that question there, and if anything changes apply it here too. It is probably fine and I'm fussing over nothing.

Quoted 65 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +8 -3
+		x->irq_rgmii_n++;
+	}
 
 	return ret;
 }
@@ -1186,6 +1225,7 @@ static void dwmac4_set_hw_vlan_mode(struct mac_device_info *hw)
 const struct stmmac_ops dwmac4_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1210,7 +1250,6 @@ const struct stmmac_ops dwmac4_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.set_mac_loopback = dwmac4_set_mac_loopback,
@@ -1230,6 +1269,7 @@ const struct stmmac_ops dwmac4_ops = {
 const struct stmmac_ops dwmac410_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1254,7 +1294,6 @@ const struct stmmac_ops dwmac410_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.flex_pps_config = dwmac5_flex_pps_config,
@@ -1278,6 +1317,7 @@ const struct stmmac_ops dwmac410_ops = {
 const struct stmmac_ops dwmac510_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
+	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1302,7 +1342,6 @@ const struct stmmac_ops dwmac510_ops = {
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
-	.pcs_get_adv_lp = dwmac4_get_adv_lp,
 	.debug = dwmac4_debug,
 	.set_filter = dwmac4_set_filter,
 	.safety_feat_config = dwmac5_safety_feat_config,
@@ -1391,5 +1430,8 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
+	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
+	mac->mac_pcs.neg_mode = true;
+
 	return 0;
 }
Signature · 1 line
2.30.2
#20 lore raw

RFC PATCH v2 6/8 net: stmmac: dwmac4: move PCS interrupt control

In reply to Russell King (Oracle)

Control the PCS interrupt mask from the phylink pcs_enable() and pcs_disable() methods rather than relying on driver variables.

This assumes that GMAC_INT_RGSMIIS, GMAC_INT_PCS_LINK and GMAC_INT_PCS_ANE are all relevant to the PCS.

---
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 29 +++++++++++++++++--
 1 file changed, 26 insertions(+), 3 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +26 -3
index cb99cb69c52b..5cf2a6cb8f66 100644
@@ -56,9 +56,6 @@ static void dwmac4_core_init(struct mac_device_info *hw,
 	/* Enable GMAC interrupts */
 	value = GMAC_INT_DEFAULT_ENABLE;
 
-	if (hw->pcs)
-		value |= GMAC_PCS_IRQ_DEFAULT;
-
 	/* Enable FPE interrupt */
 	if ((GMAC_HW_FEAT_FPESEL & readl(ioaddr + GMAC_HW_FEATURE3)) >> 26)
 		value |= GMAC_INT_FPE_EN;
@@ -770,6 +767,30 @@ static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
 	return 0;
 }
 
+static int dwmac4_mii_pcs_enable(struct phylink_pcs *pcs)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_enable;
+
+	intr_enable = readl(ioaddr + GMAC_INT_EN);
+	intr_enable |= GMAC_PCS_IRQ_DEFAULT;
+	writel(intr_enable, ioaddr + GMAC_INT_EN);
+
+	return 0;
+}
+
+static void dwmac4_mii_pcs_disable(struct phylink_pcs *pcs)
+{
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_enable;
+
+	intr_enable = readl(ioaddr + GMAC_INT_EN);
+	intr_enable &= ~GMAC_PCS_IRQ_DEFAULT;
+	writel(intr_enable, ioaddr + GMAC_INT_EN);
+}
+
 static int dwmac4_mii_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
 				 phy_interface_t interface,
 				 const unsigned long *advertising,
@@ -817,6 +838,8 @@ static void dwmac4_mii_pcs_get_state(struct phylink_pcs *pcs,
 
 static const struct phylink_pcs_ops dwmac4_mii_pcs_ops = {
 	.pcs_validate = dwmac4_mii_pcs_validate,
+	.pcs_enable = dwmac4_mii_pcs_enable,
+	.pcs_disable = dwmac4_mii_pcs_disable,
 	.pcs_config = dwmac4_mii_pcs_config,
 	.pcs_get_state = dwmac4_mii_pcs_get_state,
 };
Signature · 1 line
2.30.2
#21 lore raw

RFC PATCH v2 7/8 net: stmmac: remove obsolete pcs methods and associated code

In reply to Russell King (Oracle)

The pcs_ctrl_ane() and pcs_get_adv_lp(0 methods are no longer required as this will be handled by the mac_pcs phylink_pcs instance. Remove these methods, their common implementation, the pcs_link, pcs_duplex and pcs_speed members of struct stmmac_extra_stats, and stmmac_has_mac_phylink_select_pcs().

---
 drivers/net/ethernet/stmicro/stmmac/common.h  |   3 -
 drivers/net/ethernet/stmicro/stmmac/hwif.h    |   6 -
 .../ethernet/stmicro/stmmac/stmmac_ethtool.c  | 115 +-----------------
 .../net/ethernet/stmicro/stmmac/stmmac_main.c |  13 --
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  |  30 -----
 5 files changed, 2 insertions(+), 165 deletions(-)
drivers/net/ethernet/stmicro/stmmac/common.h +0 -3
index 940e83fa1202..a66b836996d6 100644
@@ -191,9 +191,6 @@ struct stmmac_extra_stats {
 	unsigned long irq_pcs_ane_n;
 	unsigned long irq_pcs_link_n;
 	unsigned long irq_rgmii_n;
-	unsigned long pcs_link;
-	unsigned long pcs_duplex;
-	unsigned long pcs_speed;
 	/* debug register */
 	unsigned long mtl_tx_status_fifo_full;
 	unsigned long mtl_tx_fifo_not_empty;
drivers/net/ethernet/stmicro/stmmac/hwif.h +0 -6
index aa5f6e40cb53..8a17e7d6e37d 100644
@@ -378,7 +378,6 @@ struct stmmac_ops {
 	/* PCS calls */
 	void (*pcs_ctrl_ane)(void __iomem *ioaddr, bool ane, bool srgmi_ral,
 			     bool loopback);
-	void (*pcs_get_adv_lp)(void __iomem *ioaddr, struct rgmii_adv *adv);
 	/* Safety Features */
 	int (*safety_feat_config)(void __iomem *ioaddr, unsigned int asp,
 				  struct stmmac_safety_feature_cfg *safety_cfg);
@@ -495,8 +494,6 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, debug, __priv, __args)
 #define stmmac_pcs_ctrl_ane(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, pcs_ctrl_ane, __args)
-#define stmmac_pcs_get_adv_lp(__priv, __args...) \
-	stmmac_do_void_callback(__priv, mac, pcs_get_adv_lp, __args)
 #define stmmac_safety_feat_config(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, safety_feat_config, __args)
 #define stmmac_safety_feat_irq_status(__priv, __args...) \
@@ -542,9 +539,6 @@ struct stmmac_ops {
 #define stmmac_fpe_irq_status(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, fpe_irq_status, __args)
 
-#define stmmac_has_mac_phylink_select_pcs(__priv) \
-	((__priv)->hw->mac->phylink_select_pcs != NULL)
-
 /* PTP and HW Timer helpers */
 struct stmmac_hwtimestamp {
 	void (*config_hw_tstamping) (void __iomem *ioaddr, u32 data);
drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c +2 -113

141 lines not on this page · raw message

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +0 -13
index 56c351e11952..ad19ba42c412 100644
@@ -3485,9 +3485,6 @@ static int stmmac_hw_setup(struct net_device *dev, bool ptp_register)
 		}
 	}
 
-	if (priv->hw->pcs && !stmmac_has_mac_phylink_select_pcs(priv))
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, priv->hw->ps, 0);
-
 	/* set TX and RX rings length */
 	stmmac_set_rings_length(priv);
 
@@ -6053,16 +6050,6 @@ static void stmmac_common_interrupt(struct stmmac_priv *priv)
 		for (queue = 0; queue < queues_count; queue++)
 			stmmac_host_mtl_irq_status(priv, priv->hw, queue);
 
-		/* PCS link status */
-		if (priv->hw->pcs &&
-		    !(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS) &&
-		    !stmmac_has_mac_phylink_select_pcs(priv)) {
-			if (priv->xstats.pcs_link)
-				netif_carrier_on(priv->dev);
-			else
-				netif_carrier_off(priv->dev);
-		}
-
 		stmmac_timestamp_interrupt(priv, priv);
 	}
 }
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -30
index 888485fa2761..27ea7f4d789d 100644
@@ -108,36 +108,6 @@ static inline void dwmac_ctrl_ane(void __iomem *ioaddr, u32 reg, bool ane,
 	writel(value, ioaddr + GMAC_AN_CTRL(reg));
 }
 
-/**
- * dwmac_get_adv_lp - Get ADV and LP cap
- * @ioaddr: IO registers pointer
- * @reg: Base address of the AN Control Register.
- * @adv_lp: structure to store the adv,lp status
- * Description: this is to expose the ANE advertisement and Link partner ability
- * status to ethtool support.
- */
-static inline void dwmac_get_adv_lp(void __iomem *ioaddr, u32 reg,
-				    struct rgmii_adv *adv_lp)
-{
-	u32 value = readl(ioaddr + GMAC_ANE_ADV(reg));
-
-	if (value & GMAC_ANE_FD)
-		adv_lp->duplex = DUPLEX_FULL;
-	if (value & GMAC_ANE_HD)
-		adv_lp->duplex |= DUPLEX_HALF;
-
-	adv_lp->pause = (value & GMAC_ANE_PSE) >> GMAC_ANE_PSE_SHIFT;
-
-	value = readl(ioaddr + GMAC_ANE_LPA(reg));
-
-	if (value & GMAC_ANE_FD)
-		adv_lp->lp_duplex = DUPLEX_FULL;
-	if (value & GMAC_ANE_HD)
-		adv_lp->lp_duplex = DUPLEX_HALF;
-
-	adv_lp->lp_pause = (value & GMAC_ANE_PSE) >> GMAC_ANE_PSE_SHIFT;
-}
-
 int dwmac_pcs_config(struct mac_device_info *hw, unsigned int neg_mode,
 		     phy_interface_t interface,
 		     const unsigned long *advertising,
Signature · 1 line
2.30.2
#22 lore raw

Re: [PATCH RFC net-next v2 7/8] net: stmmac: remove obsolete pcs methods and associated code

Quoted 1 line · Russell King (Oracle)

The pcs_ctrl_ane() and pcs_get_adv_lp(0 methods are no longer required

nit: pcs_get_adv_lp() -- if nothing else I'll prove my worth in terminating parenthesis in this series :P

#23 lore raw

RFC PATCH v2 8/8 net: stmmac: Activate Inband/PCS flag based on the selected iface

In reply to Russell King (Oracle)

From: Serge Semin <[email protected]>

The HWFEATURE.PCSSEL flag is set if the PCS block has been synthesized into the DW GMAC controller. It's always done if the controller supports at least one of the SGMII, TBI, RTBI PHY interfaces. If none of these interfaces support was activated during the IP-core synthesize the PCS block won't be activated either and the HWFEATURE.PCSSEL flag won't be set. Based on that the RGMII in-band status detection procedure implemented in the driver hasn't been working for the devices with the RGMII interface support and with none of the SGMII, TBI, RTBI PHY interfaces available in the device.

Fix that just by dropping the dma_cap.pcs flag check from the conditional statement responsible for the In-band/PCS functionality activation. If the RGMII interface is supported by the device then the in-band link status detection will be also supported automatically (it's always embedded into the RGMII RTL code). If the SGMII interface is supported by the device then the PCS block will be supported too (it's unconditionally synthesized into the controller). The later is also correct for the TBI/RTBI PHY interfaces.

Note while at it drop the netdev_dbg() calls since at the moment of the stmmac_check_pcs_mode() invocation the network device isn't registered. So the debug prints will be for the unknown/NULL device.

[rmk: fix build errors, only use PCS for SGMII if priv->dma_cap.pcs is set]
---
 .../net/ethernet/stmicro/stmmac/stmmac_main.c    | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +4 -12
index ad19ba42c412..4384d61a3bcd 100644
@@ -1132,18 +1132,10 @@ static void stmmac_check_pcs_mode(struct stmmac_priv *priv)
 {
 	int interface = priv->plat->mac_interface;
 
-	if (priv->dma_cap.pcs) {
-		if ((interface == PHY_INTERFACE_MODE_RGMII) ||
-		    (interface == PHY_INTERFACE_MODE_RGMII_ID) ||
-		    (interface == PHY_INTERFACE_MODE_RGMII_RXID) ||
-		    (interface == PHY_INTERFACE_MODE_RGMII_TXID)) {
-			netdev_dbg(priv->dev, "PCS RGMII support enabled\n");
-			priv->hw->pcs = STMMAC_PCS_RGMII;
-		} else if (interface == PHY_INTERFACE_MODE_SGMII) {
-			netdev_dbg(priv->dev, "PCS SGMII support enabled\n");
-			priv->hw->pcs = STMMAC_PCS_SGMII;
-		}
-	}
+	if (phy_interface_mode_is_rgmii(interface))
+		priv->hw->pcs = STMMAC_PCS_RGMII;
+	else if (priv->dma_cap.pcs && interface == PHY_INTERFACE_MODE_SGMII)
+		priv->hw->pcs = STMMAC_PCS_SGMII;
 }
 
 /**
Signature · 1 line
2.30.2
Tested-by on the v2 series
#24 lore raw
In reply to Russell King (Oracle)
Quoted 14 lines · Russell King (Oracle)

Hi,

This is version 2 of the series switching stmmac to use phylink PCS instead of going behind phylink's back.

Changes since version 1:
- Addition of patches from Serge Semin to allow RGMII to use the "PCS" code even if priv->dma_cap.pcs is not set (including tweaks by me.)
- Restructuring of the patch set to be a more logical split.
- Leave the pcs_ctrl_ane methods until we've worked out what to do with the qcom-ethqos driver (this series may still end up breaking it, but at least we will now successfully compile.)

Quoted 3 lines

A reminder that what I want to hear from this patch set are the results of testing - and thanks to Serge, the RGMII paths were exercised, but I have not had any results for the SGMII side of this.

I took this for a brief spin on the sa8775p-ride eval board from Qualcomm, as a reminder here's the dts to show the setup:

    https://elixir.bootlin.com/linux/v6.9-rc3/source/arch/arm64/boot/dts/qcom/sa8775p-ride.dts#L288

I didn't notice any issues with traffic on either interface (there's two stmmac device instances, both using SGMII to a marvell 88ea1512 phy). Did some basic link up / down tests, iperf3, etc.

If there's anything in specific you'd like exercised, please let me know. I'll try and find time to look over the patches more carefully tomorrow for review purposes, but I only know what I know from reading the driver some, so I can't answer any of the open questions with any official documentation.

Quoted 81 lines

There are still a bunch of outstanding questions:

- whether we should be using two separate PCS instances, one for RGMII and another for SGMII. If the PCS hardware is not present, but are using RGMII mode, then we probably don't want to be accessing the registers that would've been there for SGMII.
- what the three interrupts associated with the PCS code actually mean when they fire.
- which block's status we're reading in the pcs_get_state() method, and whether we should be reading that for both RGMII and SGMII.
- whether we need to activate phylink's inband mode in more cases (so that the PCS/MAC status gets read and used for the link.)

There's probably more questions to be asked... but really the critical thing is to shake out any breakage from making this conversion. Bear in mind that I have little knowledge of this hardware, so this conversion has been done somewhat blind using only what I can observe from the current driver.

Original blurb below.

As I noted recently in a thread (and was ignored) stmmac sucks. (I won't hide my distain for drivers that make my life as phylink maintainer more difficult!)

One of the contract conditions for using phylink is that the driver will _not_ mess with the netif carrier. stmmac developers/maintainers clearly didn't read that, because stmmac messes with the netif carrier, which destroys phylink's guarantee that it'll make certain calls in a particular order (e.g. it won't call mac_link_up() twice in a row without an intervening mac_link_down().) This is clearly stated in the phylink documentation.

Thus, this patch set attempts to fix this. Why does it mess with the netif carrier? It has its own independent PCS implementation that completely bypasses phylink _while_ phylink is still being used. This is not acceptable. Either the driver uses phylink, or it doesn't use phylink. There is no half-way house about this. Therefore, this driver needs to either be fixed, or needs to stop using phylink.

Since I was ignored when I brought this up, I've hacked together the following patch set - and it is hacky at the moment. It's also broken because of recentl changes involving dwmac-qcom-ethqos.c - but there isn't sufficient information in the driver for me to fix this. The driver appears to use SGMII at 2500Mbps, which simply does not exist. What interface mode (and neg_mode) does phylink pass to pcs_config() in each of the speeds that dwmac-qcom-ethqos.c is interested in. Without this information, I can't do that conversion. So for the purposes of this, I've just ignored dwmac-qcom-ethqos.c (which means it will fail to build.)

The patch splitup is not ideal, but that's not what I'm interested in here. What I want to hear is the results of testing - does this switch of the RGMII/SGMII "pcs" stuff to a phylink_pcs work for this driver?

Please don't review the patches, but you are welcome to send fixes to them. Once we know that the overall implementation works, then I'll look at how best to split the patches. In the mean time, the present form is more convenient for making changes and fixing things.

There is still more improvement that's needed here.

Thanks.

 drivers/net/ethernet/stmicro/stmmac/Makefile       |   2 +-
 drivers/net/ethernet/stmicro/stmmac/common.h       |  12 +-
 .../net/ethernet/stmicro/stmmac/dwmac1000_core.c   | 146 ++++++++++++++-------
 drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c  | 131 +++++++++++++-----
 drivers/net/ethernet/stmicro/stmmac/hwif.h         |  19 ++-
 .../net/ethernet/stmicro/stmmac/stmmac_ethtool.c   | 111 +---------------
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c  |  33 ++---
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c   |  58 ++++++++
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h   |  34 +----
 9 files changed, 298 insertions(+), 248 deletions(-)
 create mode 100644 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#25 lore raw

RFC PATCH v2 9/17 net: stmmac: Introduce mac_device_info::priv pointer

In reply to Russell King (Oracle)

There is going to be introduced an PCS-specific CSR space pointer defined in the stmmac_priv structure nearby the mmcaddr, estaddr and ptpaddr fields. In order to have that pointer accessible from the PCS-specific callback, let's introduce pointer to stmmac_priv defined in the mac_device_info structure.

---

Note the better approach would be to convert the mac_device_info instance to being embedded into the stmmac_priv structure. It would have solved many driver problems like non-unified HW abstraction interface, duplicated fields (ioaddr and pcsr, etc) or too many non-runtime parameters passed to the callbacks, etc. But the change also would have been much-much more invasive than this one is. If despite of that you find the mac_device_info embedding into stmmac_priv more appropriate (as I do), then I'll provide the respective patch in place of this one.

---
 drivers/net/ethernet/stmicro/stmmac/common.h | 1 +
 drivers/net/ethernet/stmicro/stmmac/hwif.c   | 1 +
 2 files changed, 2 insertions(+)
drivers/net/ethernet/stmicro/stmmac/common.h +1 -0
index a66b836996d6..f7661268518f 100644
@@ -580,6 +580,7 @@ struct mii_regs {
 };
 
 struct mac_device_info {
+	struct stmmac_priv *priv;
 	const struct stmmac_ops *mac;
 	const struct stmmac_desc_ops *desc;
 	const struct stmmac_dma_ops *dma;
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 29367105df54..84fd57b76fad 100644
@@ -351,6 +351,7 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		mac->tc = mac->tc ? : entry->tc;
 		mac->mmc = mac->mmc ? : entry->mmc;
 		mac->est = mac->est ? : entry->est;
+		mac->priv = priv;
 
 		priv->hw = mac;
 		priv->ptpaddr = priv->ioaddr + entry->regs.ptp_off;
Signature · 1 line
2.43.0
#26 lore raw

RFC PATCH v2 10/17 net: stmmac: Introduce internal PCS offset-based CSR access

In reply to Russell King (Oracle)

Internal PCS module can be found on the DW GMAC v3.x and DW QoS Eth IP-cores. The register space is almost identical except the base address and the SGMII/RGMII/SMII Control and Status Register layout. Thus the common PCS code could be set free from passing the PCS CSRs base address by using the base calcultaed on the MAC-device info init stage. The same approach has been utilized in the STMMAC MMC, PTP and EST modules.

Let's convert the PCS module to using it too by adding the PCS CSRs offset declarations and the CSRs base address calculation for the detected IP-core.

While at it replace the GMAC_ prefix with PCS_ to mark the common STMMAC PCS macros in the same way as it's done in the mmc.h, stmmac_ptp.h and stmmac_est.h for the respective common modules.

Note the phylink_pcs_ops instance has been defined empty for now. It will be populated later upon the rest of the PCS-code movement to the stmmac_pcs.c module.

---
 .../stmicro/stmmac/dwmac-qcom-ethqos.c        |  8 +-
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 11 ++-
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 12 +--
 drivers/net/ethernet/stmicro/stmmac/hwif.c    | 14 ++++
 drivers/net/ethernet/stmicro/stmmac/hwif.h    |  4 +-
 drivers/net/ethernet/stmicro/stmmac/stmmac.h  |  1 +
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.c  | 30 ++++---
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  | 79 +++++++++----------
 8 files changed, 89 insertions(+), 70 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c +4 -4
index 80eb72bc6311..d0bcebe87ee8 100644
@@ -633,7 +633,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_2500);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 0, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 0, 0, 0);
 		break;
 	case SPEED_1000:
 		val &= ~ETHQOS_MAC_CTRL_PORT_SEL;
@@ -641,12 +641,12 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_100:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL | ETHQOS_MAC_CTRL_SPEED_MODE;
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_10:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL;
@@ -656,7 +656,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 					 SGMII_10M_RX_CLK_DVDR),
 			      RGMII_IO_MACRO_CONFIG);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	}
 
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +5 -6
index 3ba65ea3e46f..e525b92955b4 100644
@@ -296,7 +296,7 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 			x->irq_rx_path_exit_lpi_mode_n++;
 	}
 
-	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
+	dwmac_pcs_isr(hw->priv->pcsaddr, intr_status, x);
 
 	if (intr_status & PCS_RGSMIIIS_IRQ) {
 		/* TODO Dummy-read to clear the IRQ status */
@@ -365,10 +365,10 @@ static void dwmac1000_set_eee_timer(struct mac_device_info *hw, int ls, int tw)
 	writel(value, ioaddr + LPI_TIMER_CTRL);
 }
 
-static void dwmac1000_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
+static void dwmac1000_ctrl_ane(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
 			       bool loopback)
 {
-	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
+	dwmac_ctrl_ane(pcsaddr, ane, srgmi_ral, loopback);
 }
 
 static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
@@ -414,8 +414,7 @@ static int dwmac1000_mii_pcs_config(struct phylink_pcs *pcs,
 {
 	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
 
-	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
-				GMAC_PCS_BASE);
+	return dwmac_pcs_config(hw, neg_mode, advertising, advertising);
 }
 
 static void dwmac1000_mii_pcs_get_state(struct phylink_pcs *pcs,
@@ -442,7 +441,7 @@ static void dwmac1000_mii_pcs_get_state(struct phylink_pcs *pcs,
 	state->duplex = status & GMAC_RGSMIIIS_LNKMOD_MASK ?
 			DUPLEX_FULL : DUPLEX_HALF;
 
-	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+	dwmac_pcs_get_state(hw, state);
 }
 
 static const struct phylink_pcs_ops dwmac1000_mii_pcs_ops = {
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +6 -6
index 5cf2a6cb8f66..e51c95732bad 100644
@@ -750,10 +750,10 @@ static void dwmac4_flow_ctrl(struct mac_device_info *hw, unsigned int duplex,
 	}
 }
 
-static void dwmac4_ctrl_ane(void __iomem *ioaddr, bool ane, bool srgmi_ral,
+static void dwmac4_ctrl_ane(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
 			    bool loopback)
 {
-	dwmac_ctrl_ane(ioaddr, GMAC_PCS_BASE, ane, srgmi_ral, loopback);
+	dwmac_ctrl_ane(pcsaddr, ane, srgmi_ral, loopback);
 }
 
 static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
@@ -798,8 +798,7 @@ static int dwmac4_mii_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
 {
 	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
 
-	return dwmac_pcs_config(hw, neg_mode, interface, advertising,
-				GMAC_PCS_BASE);
+	return dwmac_pcs_config(hw, advertising, interface, advertising);
 }
 
 static void dwmac4_mii_pcs_get_state(struct phylink_pcs *pcs,
@@ -833,7 +832,7 @@ static void dwmac4_mii_pcs_get_state(struct phylink_pcs *pcs,
 	state->duplex = status & GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK ?
 			DUPLEX_FULL : DUPLEX_HALF;
 
-	dwmac_pcs_get_state(hw, state, GMAC_PCS_BASE);
+	dwmac_pcs_get_state(hw, state);
 }
 
 static const struct phylink_pcs_ops dwmac4_mii_pcs_ops = {
@@ -924,7 +923,8 @@ static int dwmac4_irq_status(struct mac_device_info *hw,
 			x->irq_rx_path_exit_lpi_mode_n++;
 	}
 
-	dwmac_pcs_isr(ioaddr, GMAC_PCS_BASE, intr_status, x);
+	dwmac_pcs_isr(hw->priv->pcsaddr, intr_status, x);
+
 	if (intr_status & PCS_RGSMIIIS_IRQ) {
 		/* TODO Dummy-read to clear the IRQ status */
 		readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
drivers/net/ethernet/stmicro/stmmac/hwif.c +14 -0
index 84fd57b76fad..3666893acb69 100644
@@ -6,6 +6,7 @@
 
 #include "common.h"
 #include "stmmac.h"
+#include "stmmac_pcs.h"
 #include "stmmac_ptp.h"
 #include "stmmac_est.h"
 
@@ -116,6 +117,7 @@ static const struct stmmac_hwif_entry {
 	const void *tc;
 	const void *mmc;
 	const void *est;
+	const void *pcs;
 	int (*setup)(struct stmmac_priv *priv);
 	int (*quirks)(struct stmmac_priv *priv);
 } stmmac_hw[] = {
@@ -144,12 +146,14 @@ static const struct stmmac_hwif_entry {
 		.xgmac = false,
 		.min_id = 0,
 		.regs = {
+			.pcs_off = PCS_GMAC3_X_OFFSET,
 			.ptp_off = PTP_GMAC3_X_OFFSET,
 			.mmc_off = MMC_GMAC3_X_OFFSET,
 		},
 		.desc = NULL,
 		.dma = &dwmac1000_dma_ops,
 		.mac = &dwmac1000_ops,
+		.pcs = &dwmac_pcs_ops,
 		.hwtimestamp = &stmmac_ptp,
 		.mode = NULL,
 		.tc = NULL,
@@ -162,6 +166,7 @@ static const struct stmmac_hwif_entry {
 		.xgmac = false,
 		.min_id = 0,
 		.regs = {
+			.pcs_off = PCS_GMAC4_OFFSET,
 			.ptp_off = PTP_GMAC4_OFFSET,
 			.mmc_off = MMC_GMAC4_OFFSET,
 			.est_off = EST_GMAC4_OFFSET,
@@ -169,6 +174,7 @@ static const struct stmmac_hwif_entry {
 		.desc = &dwmac4_desc_ops,
 		.dma = &dwmac4_dma_ops,
 		.mac = &dwmac4_ops,
+		.pcs = &dwmac_pcs_ops,
 		.hwtimestamp = &stmmac_ptp,
 		.mode = NULL,
 		.tc = &dwmac510_tc_ops,
@@ -182,6 +188,7 @@ static const struct stmmac_hwif_entry {
 		.xgmac = false,
 		.min_id = DWMAC_CORE_4_00,
 		.regs = {
+			.pcs_off = PCS_GMAC4_OFFSET,
 			.ptp_off = PTP_GMAC4_OFFSET,
 			.mmc_off = MMC_GMAC4_OFFSET,
 			.est_off = EST_GMAC4_OFFSET,
@@ -189,6 +196,7 @@ static const struct stmmac_hwif_entry {
 		.desc = &dwmac4_desc_ops,
 		.dma = &dwmac4_dma_ops,
 		.mac = &dwmac410_ops,
+		.pcs = &dwmac_pcs_ops,
 		.hwtimestamp = &stmmac_ptp,
 		.mode = &dwmac4_ring_mode_ops,
 		.tc = &dwmac510_tc_ops,
@@ -202,6 +210,7 @@ static const struct stmmac_hwif_entry {
 		.xgmac = false,
 		.min_id = DWMAC_CORE_4_10,
 		.regs = {
+			.pcs_off = PCS_GMAC4_OFFSET,
 			.ptp_off = PTP_GMAC4_OFFSET,
 			.mmc_off = MMC_GMAC4_OFFSET,
 			.est_off = EST_GMAC4_OFFSET,
@@ -209,6 +218,7 @@ static const struct stmmac_hwif_entry {
 		.desc = &dwmac4_desc_ops,
 		.dma = &dwmac410_dma_ops,
 		.mac = &dwmac410_ops,
+		.pcs = &dwmac_pcs_ops,
 		.hwtimestamp = &stmmac_ptp,
 		.mode = &dwmac4_ring_mode_ops,
 		.tc = &dwmac510_tc_ops,
@@ -222,6 +232,7 @@ static const struct stmmac_hwif_entry {
 		.xgmac = false,
 		.min_id = DWMAC_CORE_5_10,
 		.regs = {
+			.pcs_off = PCS_GMAC4_OFFSET,
 			.ptp_off = PTP_GMAC4_OFFSET,
 			.mmc_off = MMC_GMAC4_OFFSET,
 			.est_off = EST_GMAC4_OFFSET,
@@ -229,6 +240,7 @@ static const struct stmmac_hwif_entry {
 		.desc = &dwmac4_desc_ops,
 		.dma = &dwmac410_dma_ops,
 		.mac = &dwmac510_ops,
+		.pcs = &dwmac_pcs_ops,
 		.hwtimestamp = &stmmac_ptp,
 		.mode = &dwmac4_ring_mode_ops,
 		.tc = &dwmac510_tc_ops,
@@ -356,6 +368,8 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		priv->hw = mac;
 		priv->ptpaddr = priv->ioaddr + entry->regs.ptp_off;
 		priv->mmcaddr = priv->ioaddr + entry->regs.mmc_off;
+		if (entry->pcs)
+			priv->pcsaddr = priv->ioaddr + entry->regs.pcs_off;
 		if (entry->est)
 			priv->estaddr = priv->ioaddr + entry->regs.est_off;
 
drivers/net/ethernet/stmicro/stmmac/hwif.h +3 -1
index 8a17e7d6e37d..ba930a87b71a 100644
@@ -376,7 +376,7 @@ struct stmmac_ops {
 		      struct stmmac_extra_stats *x, u32 rx_queues,
 		      u32 tx_queues);
 	/* PCS calls */
-	void (*pcs_ctrl_ane)(void __iomem *ioaddr, bool ane, bool srgmi_ral,
+	void (*pcs_ctrl_ane)(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
 			     bool loopback);
 	/* Safety Features */
 	int (*safety_feat_config)(void __iomem *ioaddr, unsigned int asp,
@@ -670,6 +670,7 @@ struct stmmac_est_ops {
 	stmmac_do_void_callback(__priv, est, irq_status, __args)
 
 struct stmmac_regs_off {
+	u32 pcs_off;
 	u32 ptp_off;
 	u32 mmc_off;
 	u32 est_off;
@@ -692,6 +693,7 @@ extern const struct stmmac_desc_ops dwxgmac210_desc_ops;
 extern const struct stmmac_mmc_ops dwmac_mmc_ops;
 extern const struct stmmac_mmc_ops dwxgmac_mmc_ops;
 extern const struct stmmac_est_ops dwmac510_est_ops;
+extern const struct phylink_pcs_ops dwmac_pcs_ops;
 
 #define GMAC_VERSION		0x00000020	/* GMAC CORE Version */
 #define GMAC4_VERSION		0x00000110	/* GMAC4+ CORE Version */
drivers/net/ethernet/stmicro/stmmac/stmmac.h +1 -0
index b23b920eedb1..8091d162545a 100644
@@ -312,6 +312,7 @@ struct stmmac_priv {
 	struct mutex aux_ts_lock;
 	wait_queue_head_t tstamp_busy_wait;
 
+	void __iomem *pcsaddr;
 	void __iomem *mmcaddr;
 	void __iomem *ptpaddr;
 	void __iomem *estaddr;
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c +17 -13
index 849e6f121505..41b99f7e36e6 100644
@@ -3,35 +3,35 @@
 
 int dwmac_pcs_config(struct mac_device_info *hw, unsigned int neg_mode,
 		     phy_interface_t interface,
-		     const unsigned long *advertising,
-		     unsigned int reg_base)
+		     const unsigned long *advertising)
 {
+	struct stmmac_priv *priv = hw->priv;
 	u32 val;
 
-	val = readl(hw->pcsr + GMAC_AN_CTRL(reg_base));
+	val = readl(priv->pcsaddr + PCS_AN_CTRL);
 
-	val |= GMAC_AN_CTRL_ANE | GMAC_AN_CTRL_RAN;
+	val |= PCS_AN_CTRL_ANE | PCS_AN_CTRL_RAN;
 
 	if (hw->ps)
-		val |= GMAC_AN_CTRL_SGMRAL;
+		val |= PCS_AN_CTRL_SGMRAL;
 
-	writel(val, hw->pcsr + GMAC_AN_CTRL(reg_base));
+	writel(val, priv->pcsaddr + PCS_AN_CTRL);
 
 	return 0;
 }
 
 void dwmac_pcs_get_state(struct mac_device_info *hw,
-			 struct phylink_link_state *state,
-			 unsigned int reg_base)
+			 struct phylink_link_state *state)
 {
+	struct stmmac_priv *priv = hw->priv;
 	u32 val;
 
-	val = readl(hw->pcsr + GMAC_ANE_LPA(reg_base));
+	val = readl(priv->pcsaddr + PCS_ANE_LPA);
 
 	linkmode_set_bit(ETHTOOL_LINK_MODE_Autoneg_BIT,
 			 state->lp_advertising);
 
-	if (val & GMAC_ANE_FD) {
+	if (val & PCS_ANE_FD) {
 		linkmode_set_bit(ETHTOOL_LINK_MODE_1000baseT_Full_BIT,
 				 state->lp_advertising);
 		linkmode_set_bit(ETHTOOL_LINK_MODE_100baseT_Full_BIT,
@@ -40,7 +40,7 @@ void dwmac_pcs_get_state(struct mac_device_info *hw,
 				 state->lp_advertising);
 	}
 
-	if (val & GMAC_ANE_HD) {
+	if (val & PCS_ANE_HD) {
 		linkmode_set_bit(ETHTOOL_LINK_MODE_1000baseT_Half_BIT,
 				 state->lp_advertising);
 		linkmode_set_bit(ETHTOOL_LINK_MODE_100baseT_Half_BIT,
@@ -49,10 +49,14 @@ void dwmac_pcs_get_state(struct mac_device_info *hw,
 				 state->lp_advertising);
 	}
 
+	/* TODO Make sure that STMMAC_PCS_PAUSE STMMAC_PCS_ASYM_PAUSE usage is legitimate */
 	linkmode_mod_bit(ETHTOOL_LINK_MODE_Pause_BIT,
 			 state->lp_advertising,
-			 FIELD_GET(GMAC_ANE_PSE, val) & STMMAC_PCS_PAUSE);
+			 FIELD_GET(PCS_ANE_PSE, val) & STMMAC_PCS_PAUSE);
 	linkmode_mod_bit(ETHTOOL_LINK_MODE_Asym_Pause_BIT,
 			 state->lp_advertising,
-			 FIELD_GET(GMAC_ANE_PSE, val) & STMMAC_PCS_ASYM_PAUSE);
+			 FIELD_GET(PCS_ANE_PSE, val) & STMMAC_PCS_ASYM_PAUSE);
 }
+
+const struct phylink_pcs_ops dwmac_pcs_ops = {
+};
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +39 -40

143 lines not on this page · raw message

Signature · 1 line
2.43.0
#27 lore raw

Re: [PATCH RFC net-next v2 10/17] net: stmmac: Introduce internal PCS offset-based CSR access

Quoted 29 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c +3 -3
index 80eb72bc6311..d0bcebe87ee8 100644
@@ -633,7 +633,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_2500);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 0, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 0, 0, 0);
 		break;
 	case SPEED_1000:
 		val &= ~ETHQOS_MAC_CTRL_PORT_SEL;
@@ -641,12 +641,12 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_100:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL | ETHQOS_MAC_CTRL_SPEED_MODE;
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_10:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL;

drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos) · line 662

Quoted 8 lines
drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c +1 -1
@@ -656,7 +656,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 					 SGMII_10M_RX_CLK_DVDR),
 			      RGMII_IO_MACRO_CONFIG);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	}
 

I think a better preparatory patch (given what you do in future patches) would be to change all of these to:

	ethqos_pcs_set_inband(priv, {false | true});

which would be:

static void ethqos_pcs_set_inband(struct stmmac_priv *priv, bool enable)

{
	stmmac_pcs_ctrl_ane(priv, priv->ioaddr, enable, 0, 0);
}

which then means this patch becomes a single line, and your subsequent patch just has to replace stmmac_pcs_ctrl_ane() with its open-coded equivalent.

Quoted 13 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 84fd57b76fad..3666893acb69 100644
@@ -6,6 +6,7 @@
 
 #include "common.h"
 #include "stmmac.h"
+#include "stmmac_pcs.h"
 #include "stmmac_ptp.h"
 #include "stmmac_est.h"
 

drivers/net/ethernet/stmicro/stmmac/hwif.c static const struct stmmac_hwif_entry { · line 120

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
@@ -116,6 +117,7 @@ static const struct stmmac_hwif_entry {
 	const void *tc;
 	const void *mmc;
 	const void *est;
+	const void *pcs;

I'm not a fan of void pointers. common.h includes linux/phylink.h, which will define struct phylink_pcs_ops, so there is no reason not to declare this as:

	const struct phylink_pcs_ops *pcs;
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#28 lore raw

Re: [PATCH RFC net-next v2 10/17] net: stmmac: Introduce internal PCS offset-based CSR access

Quoted 47 lines · Russell King (Oracle)
Quoted 29 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c +3 -3
index 80eb72bc6311..d0bcebe87ee8 100644
@@ -633,7 +633,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_2500);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 0, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 0, 0, 0);
 		break;
 	case SPEED_1000:
 		val &= ~ETHQOS_MAC_CTRL_PORT_SEL;
@@ -641,12 +641,12 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_100:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL | ETHQOS_MAC_CTRL_SPEED_MODE;
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_10:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL;

drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos) · line 662

Quoted 8 lines
drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c +1 -1
@@ -656,7 +656,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 					 SGMII_10M_RX_CLK_DVDR),
 			      RGMII_IO_MACRO_CONFIG);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->ioaddr, 1, 0, 0);
+		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
 		break;
 	}
 

I think a better preparatory patch (given what you do in future patches) would be to change all of these to:

	ethqos_pcs_set_inband(priv, {false | true});

which would be:

Quoted 8 lines

static void ethqos_pcs_set_inband(struct stmmac_priv *priv, bool enable)

{
	stmmac_pcs_ctrl_ane(priv, priv->ioaddr, enable, 0, 0);
}

which then means this patch becomes a single line, and your subsequent patch just has to replace stmmac_pcs_ctrl_ane() with its open-coded equivalent.

Why not. We can introduce one more preparation patch as you suggest.

Quoted 24 lines
Quoted 14 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 84fd57b76fad..3666893acb69 100644
@@ -6,6 +6,7 @@
 
 #include "common.h"
 #include "stmmac.h"
+#include "stmmac_pcs.h"
 #include "stmmac_ptp.h"
 #include "stmmac_est.h"
 

drivers/net/ethernet/stmicro/stmmac/hwif.c static const struct stmmac_hwif_entry { · line 120

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
@@ -116,6 +117,7 @@ static const struct stmmac_hwif_entry {
 	const void *tc;
 	const void *mmc;
 	const void *est;
+	const void *pcs;

I'm not a fan of void pointers. common.h includes linux/phylink.h, which will define struct phylink_pcs_ops, so there is no reason not to declare this as:

Quoted 1 line
	const struct phylink_pcs_ops *pcs;

So am I. But in this case we have all the ops fields defined as voids. So I just followed the local convention.

Anyway, I failed to find out the reason of using void pointers here. So locally, in one of my cleanup patchesets, I've got them converted to the typed pointers. I can submit that patch as another preparation patch to this series. Then we can use the pointer to phylink_pcs_ops with no doubts. What do you think?

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#29 lore raw

RFC PATCH v2 11/17 net: stmmac: Introduce internal PCS config register getter

In reply to Russell King (Oracle)

The optional PCS module CSRs are mainly represented in the framework of the address spaces [0x00c0:0x00db] on DW GMAC and [0x00e0:0x00f7] on DW QoS Eth. The spaces mapping is identical in both IP-cores. But the link state retrieved from the PHY or from another MAC (in MAC2MAC setup) is mapped over the SGMII/RGMII/SMII Control and Status register in a non-compatible way. In particular the DW GMAC register have the link state mapped at the [15:0] field, and the DW QoS Eth register have it mapped at the [31:16] field. Other than that the fields semantics is identical - it's the TX_CONFIG_REG[15:0] register (see SGMII specification for details) with a bit re-ordered fields and extended with some SMII-specific flags:
tx_config_reg[0]: LNKMOD
tx_config_reg[1:2]: LNKSPEED

tx_config_reg[3]:   LNKSTS
tx_config_reg[4]:   JABTO (Jabber Timeout, SMII-specific)
tx_config_reg[5]:   FALSCARDET (False Carrier Detected, SMII-specific)

In order to provide a fully generic internal STMMAC PCS module, let's introduce the MAC-specific callback returning the link state detected by the internal PCS.

Note the callback name has been chosen to be referring to the TX_CONFIG_REG data described in the IP-core databooks and in the SGMII specification.

---
 drivers/net/ethernet/stmicro/stmmac/dwmac1000.h     |  1 +
 .../net/ethernet/stmicro/stmmac/dwmac1000_core.c    | 11 +++++++++++
 drivers/net/ethernet/stmicro/stmmac/dwmac4.h        |  1 +
 drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c   | 13 +++++++++++++
 drivers/net/ethernet/stmicro/stmmac/hwif.h          |  3 +++
 5 files changed, 29 insertions(+)
drivers/net/ethernet/stmicro/stmmac/dwmac1000.h +1 -0
index 4296ddda8aaa..f3a95d27298c 100644
@@ -86,6 +86,7 @@ enum power_event {
 #define GMAC_RGSMIIIS		0x000000d8	/* RGMII/SMII status */
 
 /* SGMII/RGMII status register */
+#define GMAC_RGSMIIIS_CONFIG_REG	GENMASK(15, 0)
 #define GMAC_RGSMIIIS_LNKMODE		BIT(0)
 #define GMAC_RGSMIIIS_SPEED		GENMASK(2, 1)
 #define GMAC_RGSMIIIS_SPEED_SHIFT	1
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +11 -0
index e525b92955b4..1e50cc573407 100644
@@ -365,6 +365,16 @@ static void dwmac1000_set_eee_timer(struct mac_device_info *hw, int ls, int tw)
 	writel(value, ioaddr + LPI_TIMER_CTRL);
 }
 
+static u16 dwmac1000_pcs_get_config_reg(struct mac_device_info *hw)
+{
+	void __iomem *ioaddr = hw->pcsr;
+	u32 val;
+
+	val = readl(ioaddr + GMAC_RGSMIIIS);
+
+	return FIELD_GET(GMAC_RGSMIIIS_CONFIG_REG, val);
+}
+
 static void dwmac1000_ctrl_ane(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
 			       bool loopback)
 {
@@ -568,6 +578,7 @@ const struct stmmac_ops dwmac1000_ops = {
 	.set_eee_timer = dwmac1000_set_eee_timer,
 	.set_eee_pls = dwmac1000_set_eee_pls,
 	.debug = dwmac1000_debug,
+	.pcs_get_config_reg = dwmac1000_pcs_get_config_reg,
 	.pcs_ctrl_ane = dwmac1000_ctrl_ane,
 	.set_mac_loopback = dwmac1000_set_mac_loopback,
 };
drivers/net/ethernet/stmicro/stmmac/dwmac4.h +1 -0
index d3c5306f1c41..bb2997191f08 100644
@@ -567,6 +567,7 @@ static inline u32 mtl_low_credx_base_addr(const struct dwmac4_addrs *addrs,
 #define GMAC_PHYIF_CTRLSTATUS_TC		BIT(0)
 #define GMAC_PHYIF_CTRLSTATUS_LUD		BIT(1)
 #define GMAC_PHYIF_CTRLSTATUS_SMIDRXS		BIT(4)
+#define GMAC_PHYIF_CTRLSTATUS_CONFIG_REG	GENMASK(31, 16)
 #define GMAC_PHYIF_CTRLSTATUS_LNKMOD		BIT(16)
 #define GMAC_PHYIF_CTRLSTATUS_SPEED		GENMASK(18, 17)
 #define GMAC_PHYIF_CTRLSTATUS_SPEED_SHIFT	17
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +13 -0
index e51c95732bad..b7db076b4214 100644
@@ -456,6 +456,16 @@ static void dwmac4_set_eee_timer(struct mac_device_info *hw, int ls, int tw)
 	writel(value, ioaddr + GMAC4_LPI_TIMER_CTRL);
 }
 
+static u16 dwmac4_pcs_get_config_reg(struct mac_device_info *hw)
+{
+	void __iomem *ioaddr = hw->pcsr;
+	u32 val;
+
+	val = readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
+
+	return FIELD_GET(GMAC_PHYIF_CTRLSTATUS_CONFIG_REG, val);
+}
+
 static void dwmac4_write_single_vlan(struct net_device *dev, u16 vid)
 {
 	void __iomem *ioaddr = (void __iomem *)dev->base_addr;
@@ -1274,6 +1284,7 @@ const struct stmmac_ops dwmac4_ops = {
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
+	.pcs_get_config_reg = dwmac4_pcs_get_config_reg,
 	.set_filter = dwmac4_set_filter,
 	.set_mac_loopback = dwmac4_set_mac_loopback,
 	.update_vlan_hash = dwmac4_update_vlan_hash,
@@ -1318,6 +1329,7 @@ const struct stmmac_ops dwmac410_ops = {
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
+	.pcs_get_config_reg = dwmac4_pcs_get_config_reg,
 	.set_filter = dwmac4_set_filter,
 	.flex_pps_config = dwmac5_flex_pps_config,
 	.set_mac_loopback = dwmac4_set_mac_loopback,
@@ -1366,6 +1378,7 @@ const struct stmmac_ops dwmac510_ops = {
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
+	.pcs_get_config_reg = dwmac4_pcs_get_config_reg,
 	.set_filter = dwmac4_set_filter,
 	.safety_feat_config = dwmac5_safety_feat_config,
 	.safety_feat_irq_status = dwmac5_safety_feat_irq_status,
drivers/net/ethernet/stmicro/stmmac/hwif.h +3 -0
index ba930a87b71a..00995a0c9813 100644
@@ -376,6 +376,7 @@ struct stmmac_ops {
 		      struct stmmac_extra_stats *x, u32 rx_queues,
 		      u32 tx_queues);
 	/* PCS calls */
+	u16 (*pcs_get_config_reg)(struct mac_device_info *hw);
 	void (*pcs_ctrl_ane)(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
 			     bool loopback);
 	/* Safety Features */
@@ -492,6 +493,8 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, set_eee_pls, __args)
 #define stmmac_mac_debug(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, debug, __priv, __args)
+#define stmmac_pcs_get_config_reg(__priv, __args...) \
+	stmmac_do_callback(__priv, mac, pcs_get_config_reg, __args)
 #define stmmac_pcs_ctrl_ane(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, pcs_ctrl_ane, __args)
 #define stmmac_safety_feat_config(__priv, __args...) \
Signature · 1 line
2.43.0
#30 lore raw

RFC PATCH v2 12/17 net: stmmac: Introduce internal PCS IRQ enable/disable methods

In reply to Russell King (Oracle)

The PCS IRQ enable/disable procedures have been utilized in the framework of the PHYLINK PCS enable/disable callbacks. Since a generic STMMAC PHYLINK PCS implementation is about to be introduced let's move the procedures into the dedicated DW GMAC and DW QoS Eth HW-abstraction methods. These methods will be called from the PCS enable/disable functions defined in the stmmac_pcs.c in the DW MAC-independent manner.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 34 ++++++++++++-----
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 38 ++++++++++++++-----
 drivers/net/ethernet/stmicro/stmmac/hwif.h    |  6 +++
 3 files changed, 58 insertions(+), 20 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +24 -10
index 1e50cc573407..99f0bbb318ec 100644
@@ -365,6 +365,26 @@ static void dwmac1000_set_eee_timer(struct mac_device_info *hw, int ls, int tw)
 	writel(value, ioaddr + LPI_TIMER_CTRL);
 }
 
+static void dwmac1000_pcs_enable_irq(struct mac_device_info *hw)
+{
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_mask;
+
+	intr_mask = readl(ioaddr + GMAC_INT_MASK);
+	intr_mask &= ~GMAC_INT_DISABLE_PCS;
+	writel(intr_mask, ioaddr + GMAC_INT_MASK);
+}
+
+static void dwmac1000_pcs_disable_irq(struct mac_device_info *hw)
+{
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_mask;
+
+	intr_mask = readl(ioaddr + GMAC_INT_MASK);
+	intr_mask |= GMAC_INT_DISABLE_PCS;
+	writel(intr_mask, ioaddr + GMAC_INT_MASK);
+}
+
 static u16 dwmac1000_pcs_get_config_reg(struct mac_device_info *hw)
 {
 	void __iomem *ioaddr = hw->pcsr;
@@ -395,12 +415,8 @@ static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
 static int dwmac1000_mii_pcs_enable(struct phylink_pcs *pcs)
 {
 	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
-	void __iomem *ioaddr = hw->pcsr;
-	u32 intr_mask;
 
-	intr_mask = readl(ioaddr + GMAC_INT_MASK);
-	intr_mask &= ~GMAC_INT_DISABLE_PCS;
-	writel(intr_mask, ioaddr + GMAC_INT_MASK);
+	dwmac1000_pcs_enable_irq(hw);
 
 	return 0;
 }
@@ -408,12 +424,8 @@ static int dwmac1000_mii_pcs_enable(struct phylink_pcs *pcs)
 static void dwmac1000_mii_pcs_disable(struct phylink_pcs *pcs)
 {
 	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
-	void __iomem *ioaddr = hw->pcsr;
-	u32 intr_mask;
 
-	intr_mask = readl(ioaddr + GMAC_INT_MASK);
-	intr_mask |= GMAC_INT_DISABLE_PCS;
-	writel(intr_mask, ioaddr + GMAC_INT_MASK);
+	dwmac1000_pcs_disable_irq(hw);
 }
 
 static int dwmac1000_mii_pcs_config(struct phylink_pcs *pcs,
@@ -578,6 +590,8 @@ const struct stmmac_ops dwmac1000_ops = {
 	.set_eee_timer = dwmac1000_set_eee_timer,
 	.set_eee_pls = dwmac1000_set_eee_pls,
 	.debug = dwmac1000_debug,
+	.pcs_enable_irq = dwmac1000_pcs_enable_irq,
+	.pcs_disable_irq = dwmac1000_pcs_disable_irq,
 	.pcs_get_config_reg = dwmac1000_pcs_get_config_reg,
 	.pcs_ctrl_ane = dwmac1000_ctrl_ane,
 	.set_mac_loopback = dwmac1000_set_mac_loopback,
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +28 -10
index b7db076b4214..5dc8d59d3a8f 100644
@@ -456,6 +456,26 @@ static void dwmac4_set_eee_timer(struct mac_device_info *hw, int ls, int tw)
 	writel(value, ioaddr + GMAC4_LPI_TIMER_CTRL);
 }
 
+static void dwmac4_pcs_enable_irq(struct mac_device_info *hw)
+{
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_enable;
+
+	intr_enable = readl(ioaddr + GMAC_INT_EN);
+	intr_enable |= GMAC_PCS_IRQ_DEFAULT;
+	writel(intr_enable, ioaddr + GMAC_INT_EN);
+}
+
+static void dwmac4_pcs_disable_irq(struct mac_device_info *hw)
+{
+	void __iomem *ioaddr = hw->pcsr;
+	u32 intr_enable;
+
+	intr_enable = readl(ioaddr + GMAC_INT_EN);
+	intr_enable &= ~GMAC_PCS_IRQ_DEFAULT;
+	writel(intr_enable, ioaddr + GMAC_INT_EN);
+}
+
 static u16 dwmac4_pcs_get_config_reg(struct mac_device_info *hw)
 {
 	void __iomem *ioaddr = hw->pcsr;
@@ -780,12 +800,8 @@ static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
 static int dwmac4_mii_pcs_enable(struct phylink_pcs *pcs)
 {
 	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
-	void __iomem *ioaddr = hw->pcsr;
-	u32 intr_enable;
 
-	intr_enable = readl(ioaddr + GMAC_INT_EN);
-	intr_enable |= GMAC_PCS_IRQ_DEFAULT;
-	writel(intr_enable, ioaddr + GMAC_INT_EN);
+	dwmac4_pcs_enable_irq(hw);
 
 	return 0;
 }
@@ -793,12 +809,8 @@ static int dwmac4_mii_pcs_enable(struct phylink_pcs *pcs)
 static void dwmac4_mii_pcs_disable(struct phylink_pcs *pcs)
 {
 	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
-	void __iomem *ioaddr = hw->pcsr;
-	u32 intr_enable;
 
-	intr_enable = readl(ioaddr + GMAC_INT_EN);
-	intr_enable &= ~GMAC_PCS_IRQ_DEFAULT;
-	writel(intr_enable, ioaddr + GMAC_INT_EN);
+	dwmac4_pcs_disable_irq(hw);
 }
 
 static int dwmac4_mii_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
@@ -1284,6 +1296,8 @@ const struct stmmac_ops dwmac4_ops = {
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
+	.pcs_enable_irq = dwmac4_pcs_enable_irq,
+	.pcs_disable_irq = dwmac4_pcs_disable_irq,
 	.pcs_get_config_reg = dwmac4_pcs_get_config_reg,
 	.set_filter = dwmac4_set_filter,
 	.set_mac_loopback = dwmac4_set_mac_loopback,
@@ -1329,6 +1343,8 @@ const struct stmmac_ops dwmac410_ops = {
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
+	.pcs_enable_irq = dwmac4_pcs_enable_irq,
+	.pcs_disable_irq = dwmac4_pcs_disable_irq,
 	.pcs_get_config_reg = dwmac4_pcs_get_config_reg,
 	.set_filter = dwmac4_set_filter,
 	.flex_pps_config = dwmac5_flex_pps_config,
@@ -1378,6 +1394,8 @@ const struct stmmac_ops dwmac510_ops = {
 	.set_eee_pls = dwmac4_set_eee_pls,
 	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
+	.pcs_enable_irq = dwmac4_pcs_enable_irq,
+	.pcs_disable_irq = dwmac4_pcs_disable_irq,
 	.pcs_get_config_reg = dwmac4_pcs_get_config_reg,
 	.set_filter = dwmac4_set_filter,
 	.safety_feat_config = dwmac5_safety_feat_config,
drivers/net/ethernet/stmicro/stmmac/hwif.h +6 -0
index 00995a0c9813..2caa946a92f9 100644
@@ -376,6 +376,8 @@ struct stmmac_ops {
 		      struct stmmac_extra_stats *x, u32 rx_queues,
 		      u32 tx_queues);
 	/* PCS calls */
+	void (*pcs_enable_irq)(struct mac_device_info *hw);
+	void (*pcs_disable_irq)(struct mac_device_info *hw);
 	u16 (*pcs_get_config_reg)(struct mac_device_info *hw);
 	void (*pcs_ctrl_ane)(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
 			     bool loopback);
@@ -493,6 +495,10 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, set_eee_pls, __args)
 #define stmmac_mac_debug(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, debug, __priv, __args)
+#define stmmac_pcs_enable_irq(__priv, __args...) \
+	stmmac_do_void_callback(__priv, mac, pcs_enable_irq, __args)
+#define stmmac_pcs_disable_irq(__priv, __args...) \
+	stmmac_do_void_callback(__priv, mac, pcs_disable_irq, __args)
 #define stmmac_pcs_get_config_reg(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, pcs_get_config_reg, __args)
 #define stmmac_pcs_ctrl_ane(__priv, __args...) \
Signature · 1 line
2.43.0
#31 lore raw

RFC PATCH v2 13/17 net: stmmac: Move internal PCS ANE-control method to dwmac-qcom-ethqos.c

In reply to Russell King (Oracle)

The dwmac_ctrl_ane() method is not going to be utilized in the internal PCS implementation. Simplify the DW *MAC HW-abstraction interface by moving the method definition to the only user of it - dwmac-qcom-ethqos.c.

---
 .../stmicro/stmmac/dwmac-qcom-ethqos.c        | 31 ++++++++++++++---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  |  7 ----
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c |  9 -----
 drivers/net/ethernet/stmicro/stmmac/hwif.h    |  4 ---
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  | 33 -------------------
 5 files changed, 27 insertions(+), 57 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c +27 -4
index d0bcebe87ee8..fcd13a9afd59 100644
@@ -615,6 +615,29 @@ static void ethqos_set_serdes_speed(struct qcom_ethqos *ethqos, int speed)
 	}
 }
 
+static inline void ethqos_ctrl_ane(void __iomem *pcsaddr, bool ane,
+				   bool srgmi_ral, bool loopback)
+{
+	u32 value = readl(pcsaddr + PCS_AN_CTRL);
+
+	/* Enable and restart the Auto-Negotiation */
+	if (ane)
+		value |= PCS_AN_CTRL_ANE | PCS_AN_CTRL_RAN;
+	else
+		value &= ~PCS_AN_CTRL_ANE;
+
+	/* In case of MAC-2-MAC connection, block is configured to operate
+	 * according to MAC conf register.
+	 */
+	if (srgmi_ral)
+		value |= PCS_AN_CTRL_SGMRAL;
+
+	if (loopback)
+		value |= PCS_AN_CTRL_ELE;
+
+	writel(value, pcsaddr + PCS_AN_CTRL);
+}
+
 /* On interface toggle MAC registers gets reset.
  * Configure MAC block for SGMII on ethernet phy link up
  */
@@ -633,7 +656,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_2500);
-		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 0, 0, 0);
+		ethqos_ctrl_ane(priv->pcsaddr, 0, 0, 0);
 		break;
 	case SPEED_1000:
 		val &= ~ETHQOS_MAC_CTRL_PORT_SEL;
@@ -641,12 +664,12 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 			      RGMII_CONFIG2_RGMII_CLK_SEL_CFG,
 			      RGMII_IO_MACRO_CONFIG2);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
+		ethqos_ctrl_ane(priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_100:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL | ETHQOS_MAC_CTRL_SPEED_MODE;
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
+		ethqos_ctrl_ane(priv->pcsaddr, 1, 0, 0);
 		break;
 	case SPEED_10:
 		val |= ETHQOS_MAC_CTRL_PORT_SEL;
@@ -656,7 +679,7 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
 					 SGMII_10M_RX_CLK_DVDR),
 			      RGMII_IO_MACRO_CONFIG);
 		ethqos_set_serdes_speed(ethqos, SPEED_1000);
-		stmmac_pcs_ctrl_ane(priv, priv->pcsaddr, 1, 0, 0);
+		ethqos_ctrl_ane(priv->pcsaddr, 1, 0, 0);
 		break;
 	}
 
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -7
index 99f0bbb318ec..9511ea753da7 100644
@@ -395,12 +395,6 @@ static u16 dwmac1000_pcs_get_config_reg(struct mac_device_info *hw)
 	return FIELD_GET(GMAC_RGSMIIIS_CONFIG_REG, val);
 }
 
-static void dwmac1000_ctrl_ane(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
-			       bool loopback)
-{
-	dwmac_ctrl_ane(pcsaddr, ane, srgmi_ral, loopback);
-}
-
 static int dwmac1000_mii_pcs_validate(struct phylink_pcs *pcs,
 				      unsigned long *supported,
 				      const struct phylink_link_state *state)
@@ -593,7 +587,6 @@ const struct stmmac_ops dwmac1000_ops = {
 	.pcs_enable_irq = dwmac1000_pcs_enable_irq,
 	.pcs_disable_irq = dwmac1000_pcs_disable_irq,
 	.pcs_get_config_reg = dwmac1000_pcs_get_config_reg,
-	.pcs_ctrl_ane = dwmac1000_ctrl_ane,
 	.set_mac_loopback = dwmac1000_set_mac_loopback,
 };
 
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +0 -9
index 5dc8d59d3a8f..1e73c14f36ce 100644
@@ -780,12 +780,6 @@ static void dwmac4_flow_ctrl(struct mac_device_info *hw, unsigned int duplex,
 	}
 }
 
-static void dwmac4_ctrl_ane(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
-			    bool loopback)
-{
-	dwmac_ctrl_ane(pcsaddr, ane, srgmi_ral, loopback);
-}
-
 static int dwmac4_mii_pcs_validate(struct phylink_pcs *pcs,
 				   unsigned long *supported,
 				   const struct phylink_link_state *state)
@@ -1294,7 +1288,6 @@ const struct stmmac_ops dwmac4_ops = {
 	.set_eee_lpi_entry_timer = dwmac4_set_eee_lpi_entry_timer,
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
-	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
 	.pcs_enable_irq = dwmac4_pcs_enable_irq,
 	.pcs_disable_irq = dwmac4_pcs_disable_irq,
@@ -1341,7 +1334,6 @@ const struct stmmac_ops dwmac410_ops = {
 	.set_eee_lpi_entry_timer = dwmac4_set_eee_lpi_entry_timer,
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
-	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
 	.pcs_enable_irq = dwmac4_pcs_enable_irq,
 	.pcs_disable_irq = dwmac4_pcs_disable_irq,
@@ -1392,7 +1384,6 @@ const struct stmmac_ops dwmac510_ops = {
 	.set_eee_lpi_entry_timer = dwmac4_set_eee_lpi_entry_timer,
 	.set_eee_timer = dwmac4_set_eee_timer,
 	.set_eee_pls = dwmac4_set_eee_pls,
-	.pcs_ctrl_ane = dwmac4_ctrl_ane,
 	.debug = dwmac4_debug,
 	.pcs_enable_irq = dwmac4_pcs_enable_irq,
 	.pcs_disable_irq = dwmac4_pcs_disable_irq,
drivers/net/ethernet/stmicro/stmmac/hwif.h +0 -4
index 2caa946a92f9..3d39417e906d 100644
@@ -379,8 +379,6 @@ struct stmmac_ops {
 	void (*pcs_enable_irq)(struct mac_device_info *hw);
 	void (*pcs_disable_irq)(struct mac_device_info *hw);
 	u16 (*pcs_get_config_reg)(struct mac_device_info *hw);
-	void (*pcs_ctrl_ane)(void __iomem *pcsaddr, bool ane, bool srgmi_ral,
-			     bool loopback);
 	/* Safety Features */
 	int (*safety_feat_config)(void __iomem *ioaddr, unsigned int asp,
 				  struct stmmac_safety_feature_cfg *safety_cfg);
@@ -501,8 +499,6 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, pcs_disable_irq, __args)
 #define stmmac_pcs_get_config_reg(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, pcs_get_config_reg, __args)
-#define stmmac_pcs_ctrl_ane(__priv, __args...) \
-	stmmac_do_void_callback(__priv, mac, pcs_ctrl_ane, __args)
 #define stmmac_safety_feat_config(__priv, __args...) \
 	stmmac_do_callback(__priv, mac, safety_feat_config, __args)
 #define stmmac_safety_feat_irq_status(__priv, __args...) \
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -33
index 62be3921ac91..76badfd208b6 100644
@@ -76,39 +76,6 @@ static inline void dwmac_pcs_isr(void __iomem *pcsaddr,
 	}
 }
 
-/**
- * dwmac_ctrl_ane - To program the AN Control Register.
- * @ioaddr: IO registers pointer
- * @ane: to enable the auto-negotiation
- * @srgmi_ral: to manage MAC-2-MAC SGMII connections.
- * @loopback: to cause the PHY to loopback tx data into rx path.
- * Description: this is the main function to configure the AN control register
- * and init the ANE, select loopback (usually for debugging purpose) and
- * configure SGMII RAL.
- */
-static inline void dwmac_ctrl_ane(void __iomem *pcsaddr, bool ane,
-				  bool srgmi_ral, bool loopback)
-{
-	u32 value = readl(pcsaddr + PCS_AN_CTRL);
-
-	/* Enable and restart the Auto-Negotiation */
-	if (ane)
-		value |= PCS_AN_CTRL_ANE | PCS_AN_CTRL_RAN;
-	else
-		value &= ~PCS_AN_CTRL_ANE;
-
-	/* In case of MAC-2-MAC connection, block is configured to operate
-	 * according to MAC conf register.
-	 */
-	if (srgmi_ral)
-		value |= PCS_AN_CTRL_SGMRAL;
-
-	if (loopback)
-		value |= PCS_AN_CTRL_ELE;
-
-	writel(value, pcsaddr + PCS_AN_CTRL);
-}
-
 int dwmac_pcs_config(struct mac_device_info *hw, unsigned int neg_mode,
 		     phy_interface_t interface,
 		     const unsigned long *advertising);
Signature · 1 line
2.43.0
#32 lore raw

RFC PATCH v2 14/17 net: stmmac: Move internal PCS PHYLINK ops to stmmac_pcs.c

In reply to Russell King (Oracle)

The PCS-related code is now ready to be consolidated in the PCS-specific module. Let's move the PHYLINK PCS operations implementation to the stmmac_pcs.c file. No semantics has changed. The same functionality has been re-implemented in stammac_pcs.c by using the generic link status register macros and the phylink_pcs_ops instance has been populated with the new callbacks.

---

Note the code has been equipped with some TODO-notes to think about on the RFC review stage.

---
 .../net/ethernet/stmicro/stmmac/dwmac1000.h   | 12 ---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 74 ----------------
 drivers/net/ethernet/stmicro/stmmac/dwmac4.h  | 12 ---
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 80 -----------------
 drivers/net/ethernet/stmicro/stmmac/hwif.c    |  1 +
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.c  | 86 +++++++++++++++++--
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  | 17 ++--
 7 files changed, 91 insertions(+), 191 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000.h +0 -12
index f3a95d27298c..94be66e794be 100644
@@ -87,19 +87,7 @@ enum power_event {
 
 /* SGMII/RGMII status register */
 #define GMAC_RGSMIIIS_CONFIG_REG	GENMASK(15, 0)
-#define GMAC_RGSMIIIS_LNKMODE		BIT(0)
-#define GMAC_RGSMIIIS_SPEED		GENMASK(2, 1)
-#define GMAC_RGSMIIIS_SPEED_SHIFT	1
-#define GMAC_RGSMIIIS_LNKSTS		BIT(3)
-#define GMAC_RGSMIIIS_JABTO		BIT(4)
-#define GMAC_RGSMIIIS_FALSECARDET	BIT(5)
 #define GMAC_RGSMIIIS_SMIDRXS		BIT(16)
-/* LNKMOD */
-#define GMAC_RGSMIIIS_LNKMOD_MASK	0x1
-/* LNKSPEED */
-#define GMAC_RGSMIIIS_SPEED_125		0x2
-#define GMAC_RGSMIIIS_SPEED_25		0x1
-#define GMAC_RGSMIIIS_SPEED_2_5		0x0
 
 /* GMAC Configuration defines */
 #define GMAC_CONTROL_2K 0x08000000	/* IEEE 802.3as 2K packets */
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -74

86 lines not on this page · raw message

drivers/net/ethernet/stmicro/stmmac/dwmac4.h +0 -12
index bb2997191f08..5c765e16bc13 100644
@@ -568,18 +568,6 @@ static inline u32 mtl_low_credx_base_addr(const struct dwmac4_addrs *addrs,
 #define GMAC_PHYIF_CTRLSTATUS_LUD		BIT(1)
 #define GMAC_PHYIF_CTRLSTATUS_SMIDRXS		BIT(4)
 #define GMAC_PHYIF_CTRLSTATUS_CONFIG_REG	GENMASK(31, 16)
-#define GMAC_PHYIF_CTRLSTATUS_LNKMOD		BIT(16)
-#define GMAC_PHYIF_CTRLSTATUS_SPEED		GENMASK(18, 17)
-#define GMAC_PHYIF_CTRLSTATUS_SPEED_SHIFT	17
-#define GMAC_PHYIF_CTRLSTATUS_LNKSTS		BIT(19)
-#define GMAC_PHYIF_CTRLSTATUS_JABTO		BIT(20)
-#define GMAC_PHYIF_CTRLSTATUS_FALSECARDET	BIT(21)
-/* LNKMOD */
-#define GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK	0x1
-/* LNKSPEED */
-#define GMAC_PHYIF_CTRLSTATUS_SPEED_125		0x2
-#define GMAC_PHYIF_CTRLSTATUS_SPEED_25		0x1
-#define GMAC_PHYIF_CTRLSTATUS_SPEED_2_5		0x0
 
 extern const struct stmmac_dma_ops dwmac4_dma_ops;
 extern const struct stmmac_dma_ops dwmac410_dma_ops;
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +0 -80

92 lines not on this page · raw message

drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 3666893acb69..c42fb2437948 100644
@@ -363,6 +363,7 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		mac->tc = mac->tc ? : entry->tc;
 		mac->mmc = mac->mmc ? : entry->mmc;
 		mac->est = mac->est ? : entry->est;
+		mac->mac_pcs.ops = mac->mac_pcs.ops ?: entry->pcs;
 		mac->priv = priv;
 
 		priv->hw = mac;
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c +80 -6

120 lines not on this page · raw message

drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +10 -7
index 76badfd208b6..2baebb92bea7 100644
@@ -47,6 +47,16 @@
 #define PCS_ANE_RFE_SHIFT	12
 #define PCS_ANE_ACK		BIT(14)		/* AN Base-page acknowledge */
 
+/* SGMII/RGMII/SMII link status register */
+#define PCS_CFG_LNKMOD		BIT(0)		/* Link Duplex Mode */
+#define PCS_CFG_LNKSPEED	GENMASK(2, 1)	/* Link Speed: */
+#define PCS_CFG_LNKSPEED_2_5	0x0		/* 2.5 MHz - 10 Mbps */
+#define PCS_CFG_LNKSPEED_25	0x1		/* 25 MHz - 100 Mbps */
+#define PCS_CFG_LNKSPEED_250	0x2		/* 250 MHz - 1000 Mbps */
+#define PCS_CFG_LNKSTS		BIT(3)		/* Link Up/Down Status */
+#define PCS_CFG_JABTO		BIT(4)		/* Jabber Timeout (SMII only) */
+#define PCS_CFG_FALSCARDET	BIT(5)		/* False Carrier (SMII only) */
+
 /**
  * dwmac_pcs_isr - TBI, RTBI, or SGMII PHY ISR
  * @ioaddr: IO registers pointer
@@ -76,11 +86,4 @@ static inline void dwmac_pcs_isr(void __iomem *pcsaddr,
 	}
 }
 
-int dwmac_pcs_config(struct mac_device_info *hw, unsigned int neg_mode,
-		     phy_interface_t interface,
-		     const unsigned long *advertising);
-
-void dwmac_pcs_get_state(struct mac_device_info *hw,
-			 struct phylink_link_state *state);
-
 #endif /* __STMMAC_PCS_H__ */
Signature · 1 line
2.43.0
#33 lore raw

Re: [PATCH RFC net-next v2 14/17] net: stmmac: Move internal PCS PHYLINK ops to stmmac_pcs.c

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c int dwmac1000_setup(struct stmmac_priv *priv) · line 551

Quoted 6 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -1
@@ -621,7 +548,6 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
-	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

"mac->mac_pcs.neg_mode = true;" is a property of the "ops" so should move with it.

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c int dwmac4_setup(struct stmmac_priv *priv) · line 1399

Quoted 6 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +0 -1
@@ -1475,7 +1396,6 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
-	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

Also applies here.

drivers/net/ethernet/stmicro/stmmac/hwif.c int stmmac_hwif_init(struct stmmac_priv *priv) · line 366

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 3666893acb69..c42fb2437948 100644
@@ -363,6 +363,7 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		mac->tc = mac->tc ? : entry->tc;
 		mac->mmc = mac->mmc ? : entry->mmc;
 		mac->est = mac->est ? : entry->est;
+		mac->mac_pcs.ops = mac->mac_pcs.ops ?: entry->pcs;

Removing both of the above means that mac->mac_pcs.ops won't ever be set prior to this, so this whole thing should just be:

		mac->mac_pcs.ops = entry->pcs;
		mac->mac_pcs.neg_mode = true;
Quoted 25 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +21 -0
+static void dwmac_pcs_get_state(struct phylink_pcs *pcs,
+				struct phylink_link_state *state)
 {
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
 	struct stmmac_priv *priv = hw->priv;
 	u32 val;
 
+	val = stmmac_pcs_get_config_reg(priv, hw);
+
+	/* TODO The next is SGMII/RGMII/SMII-specific */
+	state->link = !!(val & PCS_CFG_LNKSTS);
+	if (!state->link)
+		return;
+
+	switch (FIELD_GET(PCS_CFG_LNKSPEED, val)) {
+	case PCS_CFG_LNKSPEED_2_5:
+		state->speed = SPEED_10;
+		break;
+	case PCS_CFG_LNKSPEED_25:
+		state->speed = SPEED_100;
+		break;
+	case PCS_CFG_LNKSPEED_250:
+		state->speed = SPEED_1000;
+		break;
+	default:

drivers/net/ethernet/stmicro/stmmac/hwif.c int stmmac_hwif_init(struct stmmac_priv *priv)

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +9 -0
+		netdev_err(priv->dev, "Unknown speed detected\n");
+		break;
+	}
+
+	state->duplex = val & PCS_CFG_LNKMOD ? DUPLEX_FULL : DUPLEX_HALF;
+
+	/* TODO Check the PCS_AN_STATUS.Link status here?.. Note the flag is latched-low */
+
+	/* TODO The next is the TBI/RTBI-specific and seems to be valid if PCS_AN_STATUS.ANC */
 	val = readl(priv->pcsaddr + PCS_ANE_LPA);

I thought these registers only existed of dma_cap.pcs is true ? If we start checking PCS_AN_STATUS.Link here, and this register reads as zeros, doesn't it mean that RMGII inband mode won't ever signal link up?

drivers/net/ethernet/stmicro/stmmac/hwif.c

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +4 -1
-	/* TODO Make sure that STMMAC_PCS_PAUSE STMMAC_PCS_ASYM_PAUSE usage is legitimate */
+	/* TODO The databook says the encoding is defined in IEEE 802.3z,
+	 * Section 37.2.1.4. Do we need the STMMAC_PCS_PAUSE and
+	 * STMMAC_PCS_ASYM_PAUSE mask here?
+	 */
 	linkmode_mod_bit(ETHTOOL_LINK_MODE_Pause_BIT,
 			 state->lp_advertising,
 			 FIELD_GET(PCS_ANE_PSE, val) & STMMAC_PCS_PAUSE);

If it's 802.3z aka 1000base-X format, then yes, we should be using these bits if we are getting state from this register.

If TBI/RTBI is ever used, rather than trying to shoe-horn it all into these functions, please consider splitting them into separate PCSes, and sharing code between them e.g. using common functions called from the method functions or shared method functions where appropriate.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#34 lore raw

Re: [PATCH RFC net-next v2 14/17] net: stmmac: Move internal PCS PHYLINK ops to stmmac_pcs.c

Quoted 34 lines · Russell King (Oracle)

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c int dwmac1000_setup(struct stmmac_priv *priv) · line 551

Quoted 6 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -1
@@ -621,7 +548,6 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
-	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

"mac->mac_pcs.neg_mode = true;" is a property of the "ops" so should move with it.

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c int dwmac4_setup(struct stmmac_priv *priv) · line 1399

Quoted 6 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +0 -1
@@ -1475,7 +1396,6 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
-	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

Also applies here.

drivers/net/ethernet/stmicro/stmmac/hwif.c int stmmac_hwif_init(struct stmmac_priv *priv) · line 366

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 3666893acb69..c42fb2437948 100644
@@ -363,6 +363,7 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		mac->tc = mac->tc ? : entry->tc;
 		mac->mmc = mac->mmc ? : entry->mmc;
 		mac->est = mac->est ? : entry->est;
+		mac->mac_pcs.ops = mac->mac_pcs.ops ?: entry->pcs;

Removing both of the above means that mac->mac_pcs.ops won't ever be set prior to this, so this whole thing should just be:

Quoted 2 lines
		mac->mac_pcs.ops = entry->pcs;
		mac->mac_pcs.neg_mode = true;

Actually, no. mac->mac_pcs.ops can be set by the platform-specific plat_stmmacenet_data::setup() method.

Quoted 31 lines
Quoted 31 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +26 -0
+static void dwmac_pcs_get_state(struct phylink_pcs *pcs,
+				struct phylink_link_state *state)
 {
+	struct mac_device_info *hw = phylink_pcs_to_mac_dev_info(pcs);
 	struct stmmac_priv *priv = hw->priv;
 	u32 val;
 
+	val = stmmac_pcs_get_config_reg(priv, hw);
+
+	/* TODO The next is SGMII/RGMII/SMII-specific */
+	state->link = !!(val & PCS_CFG_LNKSTS);
+	if (!state->link)
+		return;
+
+	switch (FIELD_GET(PCS_CFG_LNKSPEED, val)) {
+	case PCS_CFG_LNKSPEED_2_5:
+		state->speed = SPEED_10;
+		break;
+	case PCS_CFG_LNKSPEED_25:
+		state->speed = SPEED_100;
+		break;
+	case PCS_CFG_LNKSPEED_250:
+		state->speed = SPEED_1000;
+		break;
+	default:
+		netdev_err(priv->dev, "Unknown speed detected\n");
+		break;
+	}
+
+	state->duplex = val & PCS_CFG_LNKMOD ? DUPLEX_FULL : DUPLEX_HALF;
Quoted 8 lines

drivers/net/ethernet/stmicro/stmmac/hwif.c

Quoted 5 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +4 -0
+
+	/* TODO Check the PCS_AN_STATUS.Link status here?.. Note the flag is latched-low */
+
+	/* TODO The next is the TBI/RTBI-specific and seems to be valid if PCS_AN_STATUS.ANC */
 	val = readl(priv->pcsaddr + PCS_ANE_LPA);

I thought these registers only existed of dma_cap.pcs is true ?

Right. The AN-registers are SGMII/TBI/RTBI-specific.

Quoted 4 lines

If we
start checking PCS_AN_STATUS.Link here, and this register reads as zeros, doesn't it mean that RMGII inband mode won't ever signal link up?

Right. The PCS_AN_STATUS.Link should be checked for the SGMII (and TBI/RTBI) only. The databooks defines the flag as follows:

DW GMAC v3.73a:
Link Status This bit indicates whether the data channel (link) is up or
R_SS_SC_LLO down. For the TBI, RTBI or SGMII interfaces, if ANEG is going
on, data cannot be transferred across the link and hence the link is given as down.

DW QoS Eth:
Link Status When this bit is set, it indicates that the link is up between
Read-only the MAC and the TBI, RTBI, or SGMII interface. When this bit is
reset, it indicates that the link is down between the MAC and the TBI, RTBI, or SGMII interface.

I guess that in fact the flag semantics is the same on both devices. But the Access-status for some reason different. Although DW QoS Eth databook doesn't define any latched-low CSR. So there is a chance that some of the databooks might be wrong in the flag access status.

Quoted 14 lines

drivers/net/ethernet/stmicro/stmmac/hwif.c

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +4 -1
-	/* TODO Make sure that STMMAC_PCS_PAUSE STMMAC_PCS_ASYM_PAUSE usage is legitimate */
+	/* TODO The databook says the encoding is defined in IEEE 802.3z,
+	 * Section 37.2.1.4. Do we need the STMMAC_PCS_PAUSE and
+	 * STMMAC_PCS_ASYM_PAUSE mask here?
+	 */
 	linkmode_mod_bit(ETHTOOL_LINK_MODE_Pause_BIT,
 			 state->lp_advertising,
 			 FIELD_GET(PCS_ANE_PSE, val) & STMMAC_PCS_PAUSE);

If it's 802.3z aka 1000base-X format, then yes, we should be using these bits if we are getting state from this register.

I meant that should we be using the driver-specific macro in here seeing the field encoding is defined by the IEEE 802.3z? Is there any ready-to-use macros/constants defined in the network subsystem core for the standard Pause encoding (IEEE 802.3z Section 37.2.1.4)?

Quoted 5 lines

If TBI/RTBI is ever used, rather than trying to shoe-horn it all into these functions, please consider splitting them into separate PCSes, and sharing code between them e.g. using common functions called from the method functions or shared method functions where appropriate.

Ok. Sounds reasonable.

I guess your message also means that the patchset re-spinning will be on me from now, right?) If so, please note, I can't promise I'll be able to do that soonish. I am quite busy at the moment. I'll be more-or-less free for that in a month or so.

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#35 lore raw

Re: [PATCH RFC net-next v2 14/17] net: stmmac: Move internal PCS PHYLINK ops to stmmac_pcs.c

Quoted 32 lines · Serge Semin
Quoted 30 lines · Russell King (Oracle)

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c int dwmac1000_setup(struct stmmac_priv *priv) · line 551

Quoted 6 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -1
@@ -621,7 +548,6 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
-	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

"mac->mac_pcs.neg_mode = true;" is a property of the "ops" so should move with it.

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c int dwmac4_setup(struct stmmac_priv *priv) · line 1399

Quoted 6 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +0 -1
@@ -1475,7 +1396,6 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
-	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

Also applies here.

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 3666893acb69..c42fb2437948 100644
@@ -363,6 +363,7 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		mac->tc = mac->tc ? : entry->tc;
 		mac->mmc = mac->mmc ? : entry->mmc;
 		mac->est = mac->est ? : entry->est;
+		mac->mac_pcs.ops = mac->mac_pcs.ops ?: entry->pcs;
Quoted 8 lines
Quoted 5 lines

Removing both of the above means that mac->mac_pcs.ops won't ever be set prior to this, so this whole thing should just be:

		mac->mac_pcs.ops = entry->pcs;
		mac->mac_pcs.neg_mode = true;

Actually, no. mac->mac_pcs.ops can be set by the platform-specific plat_stmmacenet_data::setup() method.

mac->mac_pcs is there for the _internal_ MAC only, not for platforms to fiddle around with (remember, my patch set adds this!)

I think you're thinking of mac->phylink_pcs which platforms can and do fiddle with.

Quoted 9 lines
Quoted 7 lines

drivers/net/ethernet/stmicro/stmmac/hwif.c int stmmac_hwif_init(struct stmmac_priv *priv)

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +3 -0
+	/* TODO Check the PCS_AN_STATUS.Link status here?.. Note the flag is latched-low */
+
+	/* TODO The next is the TBI/RTBI-specific and seems to be valid if PCS_AN_STATUS.ANC */
 	val = readl(priv->pcsaddr + PCS_ANE_LPA);

I thought these registers only existed of dma_cap.pcs is true ?

Right. The AN-registers are SGMII/TBI/RTBI-specific.

Therefore, I suggest that if state->interface is RGMII, then these registers should not be accessed.

My idea is to provide two PCS per MAC:

One simple one which covers RGMII which only reads the PHYIF/RGSMIIIS register, does no configuration, but does implement the .pcs_enable/ .pcs_disable etc. The .pcs_validate method should also be empty for this because the AutoNeg ethtool capability does not refer to the inband signalling, but to the media PHY.

Then a more complex PCS implementation that does everything the RGMII one does, but also the bits for SGMII (and TBI/RTBI).

Quoted 20 lines
Quoted 4 lines

If we
start checking PCS_AN_STATUS.Link here, and this register reads as zeros, doesn't it mean that RMGII inband mode won't ever signal link up?

Right. The PCS_AN_STATUS.Link should be checked for the SGMII (and TBI/RTBI) only. The databooks defines the flag as follows:

DW GMAC v3.73a:
Link Status This bit indicates whether the data channel (link) is up or
R_SS_SC_LLO down. For the TBI, RTBI or SGMII interfaces, if ANEG is going
on, data cannot be transferred across the link and hence the link is given as down.

DW QoS Eth:
Link Status When this bit is set, it indicates that the link is up between
Read-only the MAC and the TBI, RTBI, or SGMII interface. When this bit is
reset, it indicates that the link is down between the MAC and the TBI, RTBI, or SGMII interface.

Quoted 4 lines

I guess that in fact the flag semantics is the same on both devices. But the Access-status for some reason different. Although DW QoS Eth databook doesn't define any latched-low CSR. So there is a chance that some of the databooks might be wrong in the flag access status.

Yes, it sounds like it.

Quoted 10 lines
Quoted 9 lines
Quoted 8 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +4 -1
-	/* TODO Make sure that STMMAC_PCS_PAUSE STMMAC_PCS_ASYM_PAUSE usage is legitimate */
+	/* TODO The databook says the encoding is defined in IEEE 802.3z,
+	 * Section 37.2.1.4. Do we need the STMMAC_PCS_PAUSE and
+	 * STMMAC_PCS_ASYM_PAUSE mask here?
+	 */
 	linkmode_mod_bit(ETHTOOL_LINK_MODE_Pause_BIT,
 			 state->lp_advertising,
 			 FIELD_GET(PCS_ANE_PSE, val) & STMMAC_PCS_PAUSE);
Quoted 7 lines
Quoted 2 lines

If it's 802.3z aka 1000base-X format, then yes, we should be using these bits if we are getting state from this register.

I meant that should we be using the driver-specific macro in here seeing the field encoding is defined by the IEEE 802.3z? Is there any ready-to-use macros/constants defined in the network subsystem core for the standard Pause encoding (IEEE 802.3z Section 37.2.1.4)?

include/uapi/linux/mii.h:

#define ADVERTISE_1000XFULL     0x0020  /* Try for 1000BASE-X full-duplex */
	/* GMAC_ANE_FD */
#define ADVERTISE_1000XHALF     0x0040  /* Try for 1000BASE-X half-duplex */
	/* GMAC_ANE_HD */
#define ADVERTISE_1000XPAUSE    0x0080  /* Try for 1000BASE-X pause    */
	/* GMAC_ANE_PSE bit 0 */
#define ADVERTISE_1000XPSE_ASYM 0x0100  /* Try for 1000BASE-X asym pause */
	/* GMAC_ANE_PSE bit 1 */
#define ADVERTISE_LPACK         0x4000  /* Ack link partners response  */
	/* GMAC_ANE_ACK */
#define LPA_1000XFULL           0x0020  /* Can do 1000BASE-X full-duplex */
	/* GMAC_ANE_FD */
#define LPA_1000XHALF           0x0040  /* Can do 1000BASE-X half-duplex */
	/* GMAC_ANE_HD */
#define LPA_1000XPAUSE          0x0080  /* Can do 1000BASE-X pause     */
	/* GMAC_ANE_PSE bit 0 */
#define LPA_1000XPAUSE_ASYM     0x0100  /* Can do 1000BASE-X pause asym*/
	/* GMAC_ANE_PSE bit 1 */
#define LPA_RESV                0x1000  /* Unused...                   */
	/* GMAC_ANE_RFE bit 0 */
#define LPA_RFAULT              0x2000  /* Link partner faulted        */
	/* GMAC_ANE_RFE bit 1 */
#define LPA_LPACK               0x4000  /* Link partner acked us       */
	/* GMAC_ANE_ACK */
Quoted 11 lines
Quoted 4 lines

If TBI/RTBI is ever used, rather than trying to shoe-horn it all into these functions, please consider splitting them into separate PCSes, and sharing code between them e.g. using common functions called from the method functions or shared method functions where appropriate.

Ok. Sounds reasonable.

I guess your message also means that the patchset re-spinning will be on me from now, right?) If so, please note, I can't promise I'll be able to do that soonish. I am quite busy at the moment. I'll be more-or-less free for that in a month or so.

Not necessarily - some good news today, the high priority issue I was working on is lower priority at last, which means I've more time to look at mainline again. Bad news... I need a break after about 2.5 months of frustrations, which could be from this weekend!

Given the fix for the LNKMOD issue, I suspect that won't be merged into net-next until after the weekend, but I'll see whether I can sneak a respin of the patch set once that's happened. That said, given that we'll be at -rc7, it's likely too late to be thinking about getting the PCS changes queued up for this coming merge window. In any case, I don't think even if I did post a series, we're at the point where we have something that would be ready.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#36 lore raw

Re: [PATCH RFC net-next v2 14/17] net: stmmac: Move internal PCS PHYLINK ops to stmmac_pcs.c

Quoted 39 lines · Russell King (Oracle)
Quoted 38 lines · Serge Semin
Quoted 36 lines · Russell King (Oracle)

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c int dwmac1000_setup(struct stmmac_priv *priv) · line 551

Quoted 6 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +0 -1
@@ -621,7 +548,6 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
-	mac->mac_pcs.ops = &dwmac1000_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

"mac->mac_pcs.neg_mode = true;" is a property of the "ops" so should move with it.

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c int dwmac4_setup(struct stmmac_priv *priv) · line 1399

Quoted 6 lines
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +0 -1
@@ -1475,7 +1396,6 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
-	mac->mac_pcs.ops = &dwmac4_mii_pcs_ops;
 	mac->mac_pcs.neg_mode = true;

Also applies here.

drivers/net/ethernet/stmicro/stmmac/hwif.c int stmmac_hwif_init(struct stmmac_priv *priv) · line 366

Quoted 9 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +1 -0
index 3666893acb69..c42fb2437948 100644
@@ -363,6 +363,7 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
 		mac->tc = mac->tc ? : entry->tc;
 		mac->mmc = mac->mmc ? : entry->mmc;
 		mac->est = mac->est ? : entry->est;
+		mac->mac_pcs.ops = mac->mac_pcs.ops ?: entry->pcs;

Removing both of the above means that mac->mac_pcs.ops won't ever be set prior to this, so this whole thing should just be:

		mac->mac_pcs.ops = entry->pcs;
		mac->mac_pcs.neg_mode = true;
Quoted 8 lines
Quoted 2 lines

Actually, no. mac->mac_pcs.ops can be set by the platform-specific plat_stmmacenet_data::setup() method.

mac->mac_pcs is there for the _internal_ MAC only, not for platforms to fiddle around with (remember, my patch set adds this!)

I think you're thinking of mac->phylink_pcs which platforms can and do fiddle with.

Actually I did mean mac->mac_pcs.ops. AFAICS the stmmac_hwif_init() method semantics implies that the plat_stmmacenet_data::setup() function responsibility is to allocate the mac_device_info instance and pre-initialize it' fields with the data specific for the particular device including the DW MAC HW-interface ops. Like it's done in the dwmac-sun8i.c driver (and in the currently being reviewed Loongson GMAC/GNET series). So I suppose it should also concern the internal PCS ops implementation being added by you. In case if some particular controller has some internal PCS peculiarities required to be fixed on the PHY-link PCS ops implementation level. No?

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 5 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +3 -0
+	/* TODO Check the PCS_AN_STATUS.Link status here?.. Note the flag is latched-low */
+
+	/* TODO The next is the TBI/RTBI-specific and seems to be valid if PCS_AN_STATUS.ANC */
 	val = readl(priv->pcsaddr + PCS_ANE_LPA);
Quoted 7 lines
Quoted 3 lines
Quoted 1 line

I thought these registers only existed of dma_cap.pcs is true ?

Right. The AN-registers are SGMII/TBI/RTBI-specific.

Therefore, I suggest that if state->interface is RGMII, then these registers should not be accessed.

Fully agree.

Quoted 11 lines

My idea is to provide two PCS per MAC:

One simple one which covers RGMII which only reads the PHYIF/RGSMIIIS register, does no configuration, but does implement the .pcs_enable/ .pcs_disable etc. The .pcs_validate method should also be empty for this because the AutoNeg ethtool capability does not refer to the inband signalling, but to the media PHY.

Then a more complex PCS implementation that does everything the RGMII one does, but also the bits for SGMII (and TBI/RTBI).

Agreed. Good idea.

Quoted 59 lines
Quoted 21 lines
Quoted 5 lines

If we
start checking PCS_AN_STATUS.Link here, and this register reads as zeros, doesn't it mean that RMGII inband mode won't ever signal link up?

Right. The PCS_AN_STATUS.Link should be checked for the SGMII (and TBI/RTBI) only. The databooks defines the flag as follows:

DW GMAC v3.73a:
Link Status This bit indicates whether the data channel (link) is up or
R_SS_SC_LLO down. For the TBI, RTBI or SGMII interfaces, if ANEG is going
on, data cannot be transferred across the link and hence the link is given as down.

DW QoS Eth:
Link Status When this bit is set, it indicates that the link is up between
Read-only the MAC and the TBI, RTBI, or SGMII interface. When this bit is
reset, it indicates that the link is down between the MAC and the TBI, RTBI, or SGMII interface.

Quoted 4 lines

I guess that in fact the flag semantics is the same on both devices. But the Access-status for some reason different. Although DW QoS Eth databook doesn't define any latched-low CSR. So there is a chance that some of the databooks might be wrong in the flag access status.

Yes, it sounds like it.

Quoted 10 lines
Quoted 9 lines
Quoted 8 lines
drivers/net/ethernet/stmicro/stmmac/hwif.c +4 -1
-	/* TODO Make sure that STMMAC_PCS_PAUSE STMMAC_PCS_ASYM_PAUSE usage is legitimate */
+	/* TODO The databook says the encoding is defined in IEEE 802.3z,
+	 * Section 37.2.1.4. Do we need the STMMAC_PCS_PAUSE and
+	 * STMMAC_PCS_ASYM_PAUSE mask here?
+	 */
 	linkmode_mod_bit(ETHTOOL_LINK_MODE_Pause_BIT,
 			 state->lp_advertising,
 			 FIELD_GET(PCS_ANE_PSE, val) & STMMAC_PCS_PAUSE);
Quoted 7 lines
Quoted 2 lines

If it's 802.3z aka 1000base-X format, then yes, we should be using these bits if we are getting state from this register.

I meant that should we be using the driver-specific macro in here seeing the field encoding is defined by the IEEE 802.3z? Is there any ready-to-use macros/constants defined in the network subsystem core for the standard Pause encoding (IEEE 802.3z Section 37.2.1.4)?

include/uapi/linux/mii.h:

#define ADVERTISE_1000XFULL     0x0020  /* Try for 1000BASE-X full-duplex */
	/* GMAC_ANE_FD */
#define ADVERTISE_1000XHALF     0x0040  /* Try for 1000BASE-X half-duplex */
	/* GMAC_ANE_HD */
#define ADVERTISE_1000XPAUSE    0x0080  /* Try for 1000BASE-X pause    */
	/* GMAC_ANE_PSE bit 0 */
#define ADVERTISE_1000XPSE_ASYM 0x0100  /* Try for 1000BASE-X asym pause */
	/* GMAC_ANE_PSE bit 1 */
#define ADVERTISE_LPACK         0x4000  /* Ack link partners response  */
	/* GMAC_ANE_ACK */
Quoted 14 lines
#define LPA_1000XFULL           0x0020  /* Can do 1000BASE-X full-duplex */
	/* GMAC_ANE_FD */
#define LPA_1000XHALF           0x0040  /* Can do 1000BASE-X half-duplex */
	/* GMAC_ANE_HD */
#define LPA_1000XPAUSE          0x0080  /* Can do 1000BASE-X pause     */
	/* GMAC_ANE_PSE bit 0 */
#define LPA_1000XPAUSE_ASYM     0x0100  /* Can do 1000BASE-X pause asym*/
	/* GMAC_ANE_PSE bit 1 */
#define LPA_RESV                0x1000  /* Unused...                   */
	/* GMAC_ANE_RFE bit 0 */
#define LPA_RFAULT              0x2000  /* Link partner faulted        */
	/* GMAC_ANE_RFE bit 1 */
#define LPA_LPACK               0x4000  /* Link partner acked us       */
	/* GMAC_ANE_ACK */

Got it. Thanks.

Quoted 18 lines
Quoted 6 lines
Quoted 5 lines

If TBI/RTBI is ever used, rather than trying to shoe-horn it all into these functions, please consider splitting them into separate PCSes, and sharing code between them e.g. using common functions called from the method functions or shared method functions where appropriate.

Quoted 6 lines

Ok. Sounds reasonable.

I guess your message also means that the patchset re-spinning will be on me from now, right?) If so, please note, I can't promise I'll be able to do that soonish. I am quite busy at the moment. I'll be more-or-less free for that in a month or so.

Not necessarily - some good news today, the high priority issue I was working on is lower priority at last, which means I've more time to look at mainline again. Bad news... I need a break after about 2.5 months of frustrations, which could be from this weekend!

Quoted 7 lines

Given the fix for the LNKMOD issue, I suspect that won't be merged into net-next until after the weekend, but I'll see whether I can sneak a respin of the patch set once that's happened. That said, given that we'll be at -rc7, it's likely too late to be thinking about getting the PCS changes queued up for this coming merge window. In any case, I don't think even if I did post a series, we're at the point where we have something that would be ready.

Ok. Let me know what is going to be my part in the next patch set revision preparation and when my help is needed. I think I'll be able to allocate some evenings and a few weekend days for that in this month. I very much hope my work schedule will be less occupied in the next month.

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#37 lore raw

RFC PATCH v2 15/17 net: stmmac: Move internal PCS ISR to stmmac_pcs.c

In reply to Russell King (Oracle)

Similarly to the PHYLINK PCS ops, the STMMAC PCS ISR can be now fully implemented in the stmmac_pcs.c file. As before this change the resultant method will be called from the DW GMAC and DW QoS Eth core interrupt handlers.

---

Note the AN Complete and Link state changes now cause the PHYLINK PCS state update.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  |  9 +----
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c |  9 +----
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.c  | 33 +++++++++++++++++++
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  | 31 ++---------------
 4 files changed, 38 insertions(+), 44 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +1 -8
index 332018ecd624..2d77ffd16f0a 100644
@@ -296,14 +296,7 @@ static int dwmac1000_irq_status(struct mac_device_info *hw,
 			x->irq_rx_path_exit_lpi_mode_n++;
 	}
 
-	dwmac_pcs_isr(hw->priv->pcsaddr, intr_status, x);
-
-	if (intr_status & PCS_RGSMIIIS_IRQ) {
-		/* TODO Dummy-read to clear the IRQ status */
-		readl(ioaddr + GMAC_RGSMIIIS);
-		phylink_pcs_change(&hw->mac_pcs, false);
-		x->irq_rgmii_n++;
-	}
+	dwmac_pcs_isr(hw, intr_status, x);
 
 	return ret;
 }
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +1 -8
index 1487f5cc5249..c58dc20eddeb 100644
@@ -860,14 +860,7 @@ static int dwmac4_irq_status(struct mac_device_info *hw,
 			x->irq_rx_path_exit_lpi_mode_n++;
 	}
 
-	dwmac_pcs_isr(hw->priv->pcsaddr, intr_status, x);
-
-	if (intr_status & PCS_RGSMIIIS_IRQ) {
-		/* TODO Dummy-read to clear the IRQ status */
-		readl(ioaddr + GMAC_PHYIF_CONTROL_STATUS);
-		phylink_pcs_change(&hw->mac_pcs, false);
-		x->irq_rgmii_n++;
-	}
+	dwmac_pcs_isr(hw, intr_status, x);
 
 	return ret;
 }
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c +33 -0
index 24b95d1fdb64..aac49f6472f0 100644
@@ -134,3 +134,36 @@ const struct phylink_pcs_ops dwmac_pcs_ops = {
 	.pcs_get_state = dwmac_pcs_get_state,
 
 };
+
+void dwmac_pcs_isr(struct mac_device_info *hw, unsigned int intr_status,
+		   struct stmmac_extra_stats *x)
+{
+	struct stmmac_priv *priv = hw->priv;
+	bool an_status = false, sr_status = false;
+
+	if (intr_status & PCS_ANE_IRQ) {
+		x->irq_pcs_ane_n++;
+		an_status = true;
+	}
+
+	if (intr_status & PCS_LINK_IRQ) {
+		x->irq_pcs_link_n++;
+		an_status = true;
+	}
+
+	if (intr_status & PCS_RGSMIIIS_IRQ) {
+		x->irq_rgmii_n++;
+		sr_status = true;
+	}
+
+	/* Read the AN and SGMII/RGMII/SMII status regs to clear IRQ */
+	if (an_status)
+		readl(priv->pcsaddr + PCS_AN_STATUS);
+
+	if (sr_status)
+		readl(priv->pcsaddr + PCS_SRGMII_CSR);
+
+	/* Any PCS event shall trigger the PHYLINK PCS state update */
+	if (an_status || sr_status)
+		phylink_pcs_change(&hw->mac_pcs, false);
+}
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +3 -28
index 2baebb92bea7..6e364285a4ef 100644
@@ -23,6 +23,7 @@
 #define PCS_ANE_LPA		0x0c		/* ANE link partener ability */
 #define PCS_ANE_EXP		0x10		/* ANE expansion */
 #define PCS_TBI_EXT		0x14		/* TBI extended status */
+#define PCS_SRGMII_CSR		0x18		/* SGMII/RGMII/SMII CSR */
 
 /* AN Configuration defines */
 #define PCS_AN_CTRL_RAN		BIT(9)		/* Restart Auto-Negotiation */
@@ -57,33 +58,7 @@
 #define PCS_CFG_JABTO		BIT(4)		/* Jabber Timeout (SMII only) */
 #define PCS_CFG_FALSCARDET	BIT(5)		/* False Carrier (SMII only) */
 
-/**
- * dwmac_pcs_isr - TBI, RTBI, or SGMII PHY ISR
- * @ioaddr: IO registers pointer
- * @intr_status: GMAC core interrupt status
- * @x: pointer to log these events as stats
- * Description: it is the ISR for PCS events: Auto-Negotiation Completed and
- * Link status.
- */
-static inline void dwmac_pcs_isr(void __iomem *pcsaddr,
-				 unsigned int intr_status,
-				 struct stmmac_extra_stats *x)
-{
-	u32 val = readl(pcsaddr + PCS_AN_STATUS);
-
-	if (intr_status & PCS_ANE_IRQ) {
-		x->irq_pcs_ane_n++;
-		if (val & PCS_AN_STATUS_ANC)
-			pr_info("stmmac_pcs: ANE process completed\n");
-	}
-
-	if (intr_status & PCS_LINK_IRQ) {
-		x->irq_pcs_link_n++;
-		if (val & PCS_AN_STATUS_LS)
-			pr_info("stmmac_pcs: Link Up\n");
-		else
-			pr_info("stmmac_pcs: Link Down\n");
-	}
-}
+void dwmac_pcs_isr(struct mac_device_info *hw, unsigned int intr_status,
+		   struct stmmac_extra_stats *x);
 
 #endif /* __STMMAC_PCS_H__ */
Signature · 1 line
2.43.0
#38 lore raw

RFC PATCH v2 16/17 net: stmmac: Move internal PCS init method to stmmac_pcs.c

In reply to Russell King (Oracle)

Currently the internal PCS initialization procedure is split up into several functions and special flags:

1. stmmac_check_pcs_mode() - determine the internal PCS based on the specified interface and init the mac_device_info::pcs field with the respective flag (STMMAC_PCS_RGMII or STMMAC_PCS_SGMII);

2. stmmac_ops::phylink_select_pcs() - callback specific for the DW GMAC and DW QoS Eth IP-cores which returns the pointer to the phylink_pcs structure if any of the STMMAC_PCS_RGMII or STMMAC_PCS_SGMII flag is found in the mac_device_info::pcs field.

The initialization procedure described above can be simplified by converting the stmmac_check_pcs_mode() to the PCS-init method defined in stmmac_pcs.c, which besides would take the STMMAC_FLAG_HAS_INTEGRATED_PCS flag into account. Seeing the RGMII and SGMII MAC-interfaces can be only found on the DW GMAC and DW QoS Eth, the stmmac_ops::phylink_select_pcs() callbacks content can be freely moved right into the generic phylink_mac_ops::mac_select_pcs() function.

---

The change looks very promising and greatly simplifying the internal PCS initialization procedure by providing a single coherent method activating the internal PCS capability based on the requested interface and the detected DW MAC HW-capability.

---
 .../ethernet/stmicro/stmmac/dwmac1000_core.c  | 16 +------------
 .../net/ethernet/stmicro/stmmac/dwmac4_core.c | 17 +------------
 drivers/net/ethernet/stmicro/stmmac/hwif.h    | 15 ++----------
 .../net/ethernet/stmicro/stmmac/stmmac_main.c | 24 ++++---------------
 .../net/ethernet/stmicro/stmmac/stmmac_mdio.c |  3 +++
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.c  | 15 ++++++++++++
 .../net/ethernet/stmicro/stmmac/stmmac_pcs.h  |  2 ++
 7 files changed, 28 insertions(+), 64 deletions(-)
drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c +1 -15
index 2d77ffd16f0a..2fbb853cc19c 100644
@@ -16,7 +16,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
-#include <linux/phylink.h>
+
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac1000.h"
@@ -388,17 +388,6 @@ static u16 dwmac1000_pcs_get_config_reg(struct mac_device_info *hw)
 	return FIELD_GET(GMAC_RGSMIIIS_CONFIG_REG, val);
 }
 
-static struct phylink_pcs *
-dwmac1000_phylink_select_pcs(struct stmmac_priv *priv,
-			     phy_interface_t interface)
-{
-	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
-	    priv->hw->pcs & STMMAC_PCS_SGMII)
-		return &priv->hw->mac_pcs;
-
-	return NULL;
-}
-
 static void dwmac1000_debug(struct stmmac_priv *priv, void __iomem *ioaddr,
 			    struct stmmac_extra_stats *x,
 			    u32 rx_queues, u32 tx_queues)
@@ -489,7 +478,6 @@ static void dwmac1000_set_mac_loopback(void __iomem *ioaddr, bool enable)
 
 const struct stmmac_ops dwmac1000_ops = {
 	.core_init = dwmac1000_core_init,
-	.phylink_select_pcs = dwmac1000_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac1000_rx_ipc_enable,
 	.dump_regs = dwmac1000_dump_regs,
@@ -541,7 +529,5 @@ int dwmac1000_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_shift = 2;
 	mac->mii.clk_csr_mask = GENMASK(5, 2);
 
-	mac->mac_pcs.neg_mode = true;
-
 	return 0;
 }
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c +1 -16
index c58dc20eddeb..f5449f0962ad 100644
@@ -14,7 +14,7 @@
 #include <linux/slab.h>
 #include <linux/ethtool.h>
 #include <linux/io.h>
-#include <linux/phylink.h>
+
 #include "stmmac.h"
 #include "stmmac_pcs.h"
 #include "dwmac4.h"
@@ -780,16 +780,6 @@ static void dwmac4_flow_ctrl(struct mac_device_info *hw, unsigned int duplex,
 	}
 }
 
-static struct phylink_pcs *
-dwmac4_phylink_select_pcs(struct stmmac_priv *priv, phy_interface_t interface)
-{
-	if (priv->hw->pcs & STMMAC_PCS_RGMII ||
-	    priv->hw->pcs & STMMAC_PCS_SGMII)
-		return &priv->hw->mac_pcs;
-
-	return NULL;
-}
-
 static int dwmac4_irq_mtl_status(struct stmmac_priv *priv,
 				 struct mac_device_info *hw, u32 chan)
 {
@@ -1178,7 +1168,6 @@ static void dwmac4_set_hw_vlan_mode(struct mac_device_info *hw)
 const struct stmmac_ops dwmac4_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
-	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1224,7 +1213,6 @@ const struct stmmac_ops dwmac4_ops = {
 const struct stmmac_ops dwmac410_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
-	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1274,7 +1262,6 @@ const struct stmmac_ops dwmac410_ops = {
 const struct stmmac_ops dwmac510_ops = {
 	.core_init = dwmac4_core_init,
 	.update_caps = dwmac4_update_caps,
-	.phylink_select_pcs = dwmac4_phylink_select_pcs,
 	.set_mac = stmmac_dwmac4_set_mac,
 	.rx_ipc = dwmac4_rx_ipc_enable,
 	.rx_queue_enable = dwmac4_rx_queue_enable,
@@ -1389,7 +1376,5 @@ int dwmac4_setup(struct stmmac_priv *priv)
 	mac->mii.clk_csr_mask = GENMASK(11, 8);
 	mac->num_vlan = dwmac4_get_num_vlan(priv->ioaddr);
 
-	mac->mac_pcs.neg_mode = true;
-
 	return 0;
 }
drivers/net/ethernet/stmicro/stmmac/hwif.h +2 -13
index 3d39417e906d..4bacfe8a18e0 100644
@@ -18,17 +18,13 @@
 	} \
 	__result; \
 })
-#define stmmac_do_typed_callback(__type, __fail_ret, __priv, __module, \
-				 __cname,  __arg0, __args...) \
+#define stmmac_do_callback(__priv, __module, __cname,  __arg0, __args...) \
 ({ \
-	__type __result = __fail_ret; \
+	int __result = -EINVAL; \
 	if ((__priv)->hw->__module && (__priv)->hw->__module->__cname) \
 		__result = (__priv)->hw->__module->__cname((__arg0), ##__args); \
 	__result; \
 })
-#define stmmac_do_callback(__priv, __module, __cname,  __arg0, __args...) \
-	stmmac_do_typed_callback(int, -EINVAL, __priv, __module, __cname, \
-				 __arg0, ##__args)
 
 struct stmmac_extra_stats;
 struct stmmac_priv;
@@ -315,9 +311,6 @@ struct stmmac_ops {
 	void (*core_init)(struct mac_device_info *hw, struct net_device *dev);
 	/* Update MAC capabilities */
 	void (*update_caps)(struct stmmac_priv *priv);
-	/* Get phylink PCS (for MAC */
-	struct phylink_pcs *(*phylink_select_pcs)(struct stmmac_priv *priv,
-						  phy_interface_t interface);
 	/* Enable the MAC RX/TX */
 	void (*set_mac)(void __iomem *ioaddr, bool enable);
 	/* Enable and verify that the IPC module is supported */
@@ -439,10 +432,6 @@ struct stmmac_ops {
 	stmmac_do_void_callback(__priv, mac, core_init, __args)
 #define stmmac_mac_update_caps(__priv) \
 	stmmac_do_void_callback(__priv, mac, update_caps, __priv)
-#define stmmac_mac_phylink_select_pcs(__priv, __interface) \
-	stmmac_do_typed_callback(struct phylink_pcs *, ERR_PTR(-EOPNOTSUPP), \
-				 __priv, mac, phylink_select_pcs, __priv,\
-				 __interface)
 #define stmmac_mac_set(__priv, __args...) \
 	stmmac_do_void_callback(__priv, mac, set_mac, __args)
 #define stmmac_rx_ipc(__priv, __args...) \
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +4 -20
index 72c2d3e2c121..743d356f6d12 100644
@@ -950,13 +950,16 @@ static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,
 {
 	struct stmmac_priv *priv = netdev_priv(to_net_dev(config->dev));
 
+	if (priv->hw->pcs)
+		return &priv->hw->mac_pcs;
+
 	if (priv->hw->xpcs)
 		return &priv->hw->xpcs->pcs;
 
 	if (priv->hw->phylink_pcs)
 		return priv->hw->phylink_pcs;
 
-	return stmmac_mac_phylink_select_pcs(priv, interface);
+	return NULL;
 }
 
 static void stmmac_mac_config(struct phylink_config *config, unsigned int mode,
@@ -1121,23 +1124,6 @@ static const struct phylink_mac_ops stmmac_phylink_mac_ops = {
 	.mac_link_up = stmmac_mac_link_up,
 };
 
-/**
- * stmmac_check_pcs_mode - verify if RGMII/SGMII is supported
- * @priv: driver private structure
- * Description: this is to verify if the HW supports the PCS.
- * Physical Coding Sublayer (PCS) interface that can be used when the MAC is
- * configured for the TBI, RTBI, or SGMII PHY interface.
- */
-static void stmmac_check_pcs_mode(struct stmmac_priv *priv)
-{
-	int interface = priv->plat->mac_interface;
-
-	if (phy_interface_mode_is_rgmii(interface))
-		priv->hw->pcs = STMMAC_PCS_RGMII;
-	else if (priv->dma_cap.pcs && interface == PHY_INTERFACE_MODE_SGMII)
-		priv->hw->pcs = STMMAC_PCS_SGMII;
-}
-
 /**
  * stmmac_init_phy - PHY initialization
  * @dev: net device structure
@@ -7704,8 +7690,6 @@ int stmmac_dvr_probe(struct device *device,
 	else
 		stmmac_clk_csr_set(priv);
 
-	stmmac_check_pcs_mode(priv);
-
 	pm_runtime_get_noresume(device);
 	pm_runtime_set_active(device);
 	if (!pm_runtime_enabled(device))
drivers/net/ethernet/stmicro/stmmac/stmmac_mdio.c +3 -0
index aa43117134d3..985b4b8c021f 100644
@@ -22,6 +22,7 @@
 
 #include "dwxgmac2.h"
 #include "stmmac.h"
+#include "stmmac_pcs.h"
 
 #define MII_BUSY 0x00000001
 #define MII_WRITE 0x00000002
@@ -505,6 +506,8 @@ int stmmac_pcs_setup(struct net_device *ndev)
 	priv = netdev_priv(ndev);
 	mode = priv->plat->phy_interface;
 
+	dwmac_pcs_init(priv->hw);
+
 	if (priv->plat->pcs_init) {
 		ret = priv->plat->pcs_init(priv);
 	} else if (priv->plat->mdio_bus_data &&
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c +15 -0
index aac49f6472f0..fdfc95299f89 100644
@@ -167,3 +167,18 @@ void dwmac_pcs_isr(struct mac_device_info *hw, unsigned int intr_status,
 	if (an_status || sr_status)
 		phylink_pcs_change(&hw->mac_pcs, false);
 }
+
+void dwmac_pcs_init(struct mac_device_info *hw)
+{
+	struct stmmac_priv *priv = hw->priv;
+	int interface = priv->plat->mac_interface;
+
+	if (priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)
+		return;
+	else if (phy_interface_mode_is_rgmii(interface))
+		hw->pcs = STMMAC_PCS_RGMII;
+	else if (priv->dma_cap.pcs && interface == PHY_INTERFACE_MODE_SGMII)
+		hw->pcs = STMMAC_PCS_SGMII;
+
+	hw->mac_pcs.neg_mode = true;
+}
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +2 -0
index 6e364285a4ef..a17e5b37c411 100644
@@ -61,4 +61,6 @@
 void dwmac_pcs_isr(struct mac_device_info *hw, unsigned int intr_status,
 		   struct stmmac_extra_stats *x);
 
+void dwmac_pcs_init(struct mac_device_info *hw);
+
 #endif /* __STMMAC_PCS_H__ */
Signature · 1 line
2.43.0
#39 lore raw

Re: [PATCH RFC net-next v2 16/17] net: stmmac: Move internal PCS init method to stmmac_pcs.c

Quoted 9 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +1 -0
index 72c2d3e2c121..743d356f6d12 100644
@@ -950,13 +950,16 @@ static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,
 {
 	struct stmmac_priv *priv = netdev_priv(to_net_dev(config->dev));
 
+	if (priv->hw->pcs)

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config, · line 962

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +3 -1
+		return &priv->hw->mac_pcs;
+
 	if (priv->hw->xpcs)
 		return &priv->hw->xpcs->pcs;
 
 	if (priv->hw->phylink_pcs)
 		return priv->hw->phylink_pcs;
 
-	return stmmac_mac_phylink_select_pcs(priv, interface);
+	return NULL;

I really really don't like this due to:

1. I spent a long time working out what the priority here should be, and you've just thrown all that work away by changing it - to something that I believe is incorrect.

2. I want to eventually see this function checking the interface type before just handing out a random PCS, and it was my intention to eventually that into the MACs own select_pcs() methods. Getting rid of those methods means that the MACs themselves now can't make the decision which is where that should be.

3. When operating in RGMII "inband" mode, the .pcs_config etc doesn't make much sense (we're probably accessing registers that don't exist) and I had plans to split this into a RGMII "PCS" which was just a PCS that implemented .pcs_get_state(), a stub .pcs_config(), and a separate fully-featured "SGMII PCS".

So, I would like to eventually see here something like:

	if (priv->hw->xpcs)
		return &priv->hw->xpcs->pcs;

	if (priv->hw->phylink_pcs)
		return priv->hw->phylink_pcs;

	if (!(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)) {
		if (phy_interface_mode_is_rgmii(priv->plat->mac_interface))
			return &priv->hw->mac_rgmii_pcs;

		if (priv->dma_cap.pcs &&
		    priv->plat->mac_interface == PHY_INTERFACE_MODE_SGMII)
			return &priv->hw->mac_sgmii_pcs;
	}

	return NULL;
Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +4 -0
+void dwmac_pcs_init(struct mac_device_info *hw)
+{
+	struct stmmac_priv *priv = hw->priv;
+	int interface = priv->plat->mac_interface;

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +10 -0
+
+	if (priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)
+		return;
+	else if (phy_interface_mode_is_rgmii(interface))
+		hw->pcs = STMMAC_PCS_RGMII;
+	else if (priv->dma_cap.pcs && interface == PHY_INTERFACE_MODE_SGMII)
+		hw->pcs = STMMAC_PCS_SGMII;
+
+	hw->mac_pcs.neg_mode = true;
+}

Please move "hw->mac_pcs.neg_mode = true;" to where the PCS method functions are implemented - it determines whether the PCS method functions take the AN mode or the neg mode, and this is a property of their implementations. It should not be split away from them.

Thanks.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#40 lore raw

Re: [PATCH RFC net-next v2 16/17] net: stmmac: Move internal PCS init method to stmmac_pcs.c

Quoted 24 lines · Russell King (Oracle)
Quoted 9 lines · Serge Semin
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +1 -0
index 72c2d3e2c121..743d356f6d12 100644
@@ -950,13 +950,16 @@ static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,
 {
 	struct stmmac_priv *priv = netdev_priv(to_net_dev(config->dev));
 
+	if (priv->hw->pcs)

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config, · line 962

Quoted 10 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +3 -1
+		return &priv->hw->mac_pcs;
+
 	if (priv->hw->xpcs)
 		return &priv->hw->xpcs->pcs;
 
 	if (priv->hw->phylink_pcs)
 		return priv->hw->phylink_pcs;
 
-	return stmmac_mac_phylink_select_pcs(priv, interface);
+	return NULL;

I really really don't like this due to:

Quoted 4 lines

1. I spent a long time working out what the priority here should be, and you've just thrown all that work away by changing it - to something that I believe is incorrect.

Right, the correct precedence would be to use the external PCS if one available. It's easy to fix anyway.

Quoted 2 lines

2. I want to eventually see this function checking the interface type before just handing out a random PCS,

The only problem is that currently it relies on the plat_stmmaenet_data::mac_interface field value instead of parsing the specified interface type.(

Quoted 4 lines

and it was my intention to
eventually that into the MACs own select_pcs() methods. Getting rid of those methods means that the MACs themselves now can't make the decision which is where that should be.

Ok. Why not. We can preserve the MAC-own select_pcs() method. (See my last comment on this email for details.)

Quoted 3 lines

3. When operating in RGMII "inband" mode, the .pcs_config etc doesn't make much sense (we're probably accessing registers that don't exist)

Absolutely right. Current dwmac_pcs_config() implementation is fully SGMII/TBI/RTBI-specific.

Quoted 3 lines

and I had plans to split this into a RGMII "PCS" which was just a PCS that implemented .pcs_get_state(), a stub .pcs_config(), and a separate fully-featured "SGMII PCS".

Actually it's a good idea. We should have that implemented in v3.

Quoted 19 lines

So, I would like to eventually see here something like:

	if (priv->hw->xpcs)
		return &priv->hw->xpcs->pcs;

	if (priv->hw->phylink_pcs)
		return priv->hw->phylink_pcs;

	if (!(priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)) {
		if (phy_interface_mode_is_rgmii(priv->plat->mac_interface))
			return &priv->hw->mac_rgmii_pcs;

		if (priv->dma_cap.pcs &&
		    priv->plat->mac_interface == PHY_INTERFACE_MODE_SGMII)
			return &priv->hw->mac_sgmii_pcs;
	}

	return NULL;

So the differences of my and your implementations are:
1. priv->hw->pcs field is utilized to determine the RGMII/SGMII PCS availability (it's initialized in dwmac_pcs_init()).
2. The order of the PCS selection: internal PCS has precedence over the external PCS'es.
3. There is a single PHY-link PCS descriptor for both RGMII "inband" and SGMII PCSes.

There is nothing hard to settle the 2. and 3. notes. The only problematic part is 1. due to the damn mac_device_info::ps field implying the fixed-speed semantics for the MAC2MAC case. The field is initialized in the stmmac_hw_setup() method based on the mac_device_info::pcs field content. The mac_device_info::ps value is then utilized in the stmmac_ops::core_init() methods and in dwmac_pcs_config() to pre-define the link speed. Since I hadn't come up with a good idea of what to do with that MAC2MAC stuff back then I decided to preserve the mac_device_info::pcs-based semantics everywhere.

But now I guess I've got a good idea. We can use the plat_stmmacenet_data::mac_port_sel_speed field directly where it is relevant. Like this:

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c:
static int stmmac_hw_setup(struct net_device *dev, bool ptp_register)

{
	// Drop everything priv->hw.pcs and priv->hw.ps related from here
	// due to the changes suggested further.
}

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c:
drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:
static void dwmac*_core_init(...)

{
	...
	// Directly use the plat_stmmacenet_data::mac_port_sel_speed value
	switch (priv->plat->mac_port_sel_speed) {
	case SPEED_1000:
		ps_speed = hw->link.speed1000;
		break;
	case SPEED_100:
		ps_speed = hw->link.speed100;
		break;
	case SPEED_10:
		ps_speed = hw->link.speed10;
		break;
	default:
		dev_warn(priv->device, "Unsupported port speed\n");
		break;
	}

	if (ps_speed) {
		value &= hw->link.speed_mask;
		value |= ps_speed | GMAC_CONFIG_TE;
	}
	...
}

drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:
static void dwmac*_core_init(...)

{
	// There is no internal PCS in DW XGMACes. So we can freely drop
	// the hw->ps clause from here.
}

drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c:
static int dwmac_pcs_config(...)

{
	...
	// Directly use the plat_stmmacenet_data::mac_port_sel_speed value
	if (priv->plat->mac_port_sel_speed)
		val |= PCS_AN_CTRL_SGMRAL;
	...
}

After that we can freely drop the mac_device_info::ps and mac_device_info::pcs fields. Thoughts?

Quoted 13 lines
Quoted 13 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +12 -0
+void dwmac_pcs_init(struct mac_device_info *hw)
+{
+	struct stmmac_priv *priv = hw->priv;
+	int interface = priv->plat->mac_interface;
+
+	if (priv->plat->flags & STMMAC_FLAG_HAS_INTEGRATED_PCS)
+		return;
+	else if (phy_interface_mode_is_rgmii(interface))
+		hw->pcs = STMMAC_PCS_RGMII;
+	else if (priv->dma_cap.pcs && interface == PHY_INTERFACE_MODE_SGMII)
+		hw->pcs = STMMAC_PCS_SGMII;
+
Quoted 8 lines

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c

Quoted 2 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +2 -0
+	hw->mac_pcs.neg_mode = true;
+}

Please move "hw->mac_pcs.neg_mode = true;" to where the PCS method functions are implemented - it determines whether the PCS method functions take the AN mode or the neg mode, and this is a property of their implementations. It should not be split away from them.

Ok.

---

Seeing the series introducing the plat_stmmacenet_data::select_pcs() method has been recently merged in, let's discuss the entire PCS selection code a bit more. Taking into account what you said above I guess we can implement something like this:

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c:
static struct phylink_pcs *stmmac_mac_select_pcs(struct phylink_config *config,
phy_interface_t interface)

{
	// Platform-specific PCS selection method implying DW XPCS and
	// Lynx PCS selection (and internal PCS selection if relevant)
        if (priv->plat->select_pcs) {
                pcs = priv->plat->select_pcs(priv, interface);
                if (!IS_ERR(pcs))
                        return pcs;
        }

	// MAC-specific PCS selection method
	pcs = stmmac_mac_select_int_pcs(priv, priv->hw, priv->plat->mac_interface);
        if (!IS_ERR(pcs))
                return pcs;

        return NULL;
}

drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c:
static struct phylink_pcs *dwmac1000_select_pcs(struct mac_device_info *hw,
phy_interface_t interface)

{
	if (phy_interface_mode_is_rgmii(interface))
		return &hw->mac_rgmii_pcs;
	else if (priv->dma_cap.pcs && interface == PHY_INTERFACE_MODE_SGMII)
		return &hw->mac_sgmii_pcs;

	return NULLL
}

...

const struct stmmac_ops dwmac1000_ops = {
	...
	.select_pcs = dwmac1000_select_pcs,
	...
};

drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:
// The same changes as in the dwmac1000_core.c file.

drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.c:
// Drop my dwmac_pcs_init() implementation if we get to eliminate the // mac_device_info::ps and mac_device_info::pcs fields as I suggested // earlier in this message

So what do you think?

-Serge(y)

Quoted 6 lines

Thanks.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#41 lore raw

RFC PATCH v2 17/17 net: stmmac: pcs: Drop the _SHIFT macros

In reply to Russell King (Oracle)

The PCS_ANE_PSE_SHIFT and PCS_ANE_RFE_SHIFT are unused anyway. Moreover PCS_ANE_PSE and PCS_ANE_RFE are the respective field masks. So the FIELD_GET()/FIELD_SET() macro-functions can be used to get/set the fields content. Drop the _SHIFT macros for good then.

---
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h | 2 --
 1 file changed, 2 deletions(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -2
index a17e5b37c411..0f15c9898788 100644
@@ -43,9 +43,7 @@
 #define PCS_ANE_FD		BIT(5)		/* AN Full-duplex flag */
 #define PCS_ANE_HD		BIT(6)		/* AN Half-duplex flag */
 #define PCS_ANE_PSE		GENMASK(8, 7)	/* AN Pause Encoding */
-#define PCS_ANE_PSE_SHIFT	7
 #define PCS_ANE_RFE		GENMASK(13, 12)	/* AN Remote Fault Encoding */
-#define PCS_ANE_RFE_SHIFT	12
 #define PCS_ANE_ACK		BIT(14)		/* AN Base-page acknowledge */
 
 /* SGMII/RGMII/SMII link status register */
Signature · 1 line
2.43.0
#42 lore raw

Re: [PATCH RFC net-next v2 17/17] net: stmmac: pcs: Drop the _SHIFT macros

Quoted 15 lines · Serge Semin

The PCS_ANE_PSE_SHIFT and PCS_ANE_RFE_SHIFT are unused anyway. Moreover PCS_ANE_PSE and PCS_ANE_RFE are the respective field masks. So the FIELD_GET()/FIELD_SET() macro-functions can be used to get/set the fields content. Drop the _SHIFT macros for good then.

---
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h | 2 --
 1 file changed, 2 deletions(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -0
index a17e5b37c411..0f15c9898788 100644

drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h line 47

Quoted 7 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -2
@@ -43,9 +43,7 @@
 #define PCS_ANE_FD		BIT(5)		/* AN Full-duplex flag */
 #define PCS_ANE_HD		BIT(6)		/* AN Half-duplex flag */
 #define PCS_ANE_PSE		GENMASK(8, 7)	/* AN Pause Encoding */
-#define PCS_ANE_PSE_SHIFT	7
 #define PCS_ANE_RFE		GENMASK(13, 12)	/* AN Remote Fault Encoding */
-#define PCS_ANE_RFE_SHIFT	12
 #define PCS_ANE_ACK		BIT(14)		/* AN Base-page acknowledge */

I would actually like to see all these go away.

PCS_ANE_FD == LPA_1000XFULL
PCS_ANE_HD == LPA_1000XHALF
PCS_ANE_PSE == LPA_1000XPAUSE and LPA_1000XPAUSE_ASYM PCS_ANE_RFE == LPA_RESV and LPA_RFAULT PCS_ANE_ACK == LPA_LPACK

Isn't it rather weird that the field layout matches 802.3z aka 1000base-X and not SGMII? This layout would not make sense for Cisco SGMII as it loses the speed information conveyed by the Cisco SGMII control word.

This isn't a case of the manufacturer using "SGMII" to mean a serial gigabit media independent interface that supports 1000base-X (PHY_INTERFACE_MODE_1000BASEX) rather than Cisco SGMII (PHY_INTERFACE_MODE_SGMII) ?

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#43 lore raw

Re: [PATCH RFC net-next v2 17/17] net: stmmac: pcs: Drop the _SHIFT macros

Quoted 27 lines · Russell King (Oracle)
Quoted 15 lines · Serge Semin

The PCS_ANE_PSE_SHIFT and PCS_ANE_RFE_SHIFT are unused anyway. Moreover PCS_ANE_PSE and PCS_ANE_RFE are the respective field masks. So the FIELD_GET()/FIELD_SET() macro-functions can be used to get/set the fields content. Drop the _SHIFT macros for good then.

---
 drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h | 2 --
 1 file changed, 2 deletions(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -0
index a17e5b37c411..0f15c9898788 100644

drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h line 47

Quoted 7 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_pcs.h +0 -2
@@ -43,9 +43,7 @@
 #define PCS_ANE_FD		BIT(5)		/* AN Full-duplex flag */
 #define PCS_ANE_HD		BIT(6)		/* AN Half-duplex flag */
 #define PCS_ANE_PSE		GENMASK(8, 7)	/* AN Pause Encoding */
-#define PCS_ANE_PSE_SHIFT	7
 #define PCS_ANE_RFE		GENMASK(13, 12)	/* AN Remote Fault Encoding */
-#define PCS_ANE_RFE_SHIFT	12
 #define PCS_ANE_ACK		BIT(14)		/* AN Base-page acknowledge */

I would actually like to see all these go away.

Quoted 5 lines

PCS_ANE_FD == LPA_1000XFULL
PCS_ANE_HD == LPA_1000XHALF
PCS_ANE_PSE == LPA_1000XPAUSE and LPA_1000XPAUSE_ASYM PCS_ANE_RFE == LPA_RESV and LPA_RFAULT PCS_ANE_ACK == LPA_LPACK

Great! It will be even better.

Quoted 10 lines

Isn't it rather weird that the field layout matches 802.3z aka 1000base-X and not SGMII? This layout would not make sense for Cisco SGMII as it loses the speed information conveyed by the Cisco SGMII control word.

This isn't a case of the manufacturer using "SGMII" to mean a serial gigabit media independent interface that supports 1000base-X (PHY_INTERFACE_MODE_1000BASEX) rather than Cisco SGMII (PHY_INTERFACE_MODE_SGMII) ?

It's not that weird. The only CSRs available and functional for SGMII PCS are PCS_AN_CTRL and PCS_AN_STATUS (plus PCS_SRGMII_CSR where the PCS exposes the link status, aka Cisco SGMII tx_config_reg[15:0] with the vendor-specific layout). Registers from PCS_ANE_ADV up to PCS_TBI_EXT exist for TBI and RTBI PCS only. Since TBI is defined in the IEEE 802.3z C36 I guess that's why the layout matches to what is described in IEEE 802.3z C37.

-Serge(y)

Quoted 4 lines
Signature · 3 lines
-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!