diff mbox series

[2/2] stmmac: intel: Honor phy LED set by system firmware on a Dell hardware

Message ID 20220114040755.1314349-2-kai.heng.feng@canonical.com (mailing list archive)
State Changes Requested
Delegated to: Netdev Maintainers
Headers show
Series None | expand

Commit Message

Kai-Heng Feng Jan. 14, 2022, 4:07 a.m. UTC 
BIOS on Dell Edge Gateway 3200 already makes its own phy LED setting, so
instead of setting another value, keep it untouched and restore the saved
value on system resume.

Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
---
 .../net/ethernet/stmicro/stmmac/dwmac-intel.c | 16 +++++
 drivers/net/ethernet/stmicro/stmmac/stmmac.h  |  2 +
 .../net/ethernet/stmicro/stmmac/stmmac_main.c |  4 ++
 drivers/net/phy/marvell.c                     | 58 ++++++++++++-------
 include/linux/marvell_phy.h                   |  1 +
 5 files changed, 61 insertions(+), 20 deletions(-)

Comments

Jakub Kicinski Jan. 14, 2022, 4:35 a.m. UTC | #1 
On Fri, 14 Jan 2022 12:07:54 +0800 Kai-Heng Feng wrote:
> BIOS on Dell Edge Gateway 3200 already makes its own phy LED setting, so
> instead of setting another value, keep it untouched and restore the saved
> value on system resume.
> 
> Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>

I defer to PHY experts for review. Coincidentally the first Marvell
flag appears dead, nobody sets it:

$ git grep MARVELL_PHY_M1145_FLAGS_RESISTANCE
drivers/net/phy/marvell.c:      if (phydev->dev_flags & MARVELL_PHY_M1145_FLAGS_RESISTANCE) {
include/linux/marvell_phy.h:#define MARVELL_PHY_M1145_FLAGS_RESISTANCE  0x00000001
$

unless it's read from DT under different name or something.


Once you get some reviews please wait for net-next to open:

http://vger.kernel.org/~davem/net-next.html

and repost. It should happen the week of Jan 24th. When you repost
please drop the first patch, I believe Russell does not like the BIT()
macro, his opinion overrides checkpatch.

Thanks!
Kai-Heng Feng Jan. 14, 2022, 6:47 a.m. UTC | #2 
On Fri, Jan 14, 2022 at 12:35 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Fri, 14 Jan 2022 12:07:54 +0800 Kai-Heng Feng wrote:
> > BIOS on Dell Edge Gateway 3200 already makes its own phy LED setting, so
> > instead of setting another value, keep it untouched and restore the saved
> > value on system resume.
> >
> > Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
>
> I defer to PHY experts for review. Coincidentally the first Marvell
> flag appears dead, nobody sets it:
>
> $ git grep MARVELL_PHY_M1145_FLAGS_RESISTANCE
> drivers/net/phy/marvell.c:      if (phydev->dev_flags & MARVELL_PHY_M1145_FLAGS_RESISTANCE) {
> include/linux/marvell_phy.h:#define MARVELL_PHY_M1145_FLAGS_RESISTANCE  0x00000001
> $
>
> unless it's read from DT under different name or something.

It was introduced by 95d21ff4c645 without any user. Should we keep it?

>
>
> Once you get some reviews please wait for net-next to open:
>
> http://vger.kernel.org/~davem/net-next.html
>
> and repost. It should happen the week of Jan 24th. When you repost
> please drop the first patch, I believe Russell does not like the BIT()
> macro, his opinion overrides checkpatch.

Of course. I'll wait for the review and resubmit the 2nd patch.

Kai-Heng

>
> Thanks!
Andrew Lunn Jan. 14, 2022, 1:09 p.m. UTC | #3 
On Fri, Jan 14, 2022 at 12:07:54PM +0800, Kai-Heng Feng wrote:
> BIOS on Dell Edge Gateway 3200 already makes its own phy LED setting, so
> instead of setting another value, keep it untouched and restore the saved
> value on system resume.
> 
> Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
> ---
>  .../net/ethernet/stmicro/stmmac/dwmac-intel.c | 16 +++++
>  drivers/net/ethernet/stmicro/stmmac/stmmac.h  |  2 +
>  .../net/ethernet/stmicro/stmmac/stmmac_main.c |  4 ++
>  drivers/net/phy/marvell.c                     | 58 ++++++++++++-------
>  include/linux/marvell_phy.h                   |  1 +
>  5 files changed, 61 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
> index 8e8778cfbbadd..f8a2879e0264a 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
> @@ -857,6 +857,16 @@ static const struct dmi_system_id quark_pci_dmi[] = {
>  	{}
>  };
>  
> +static const struct dmi_system_id use_preset_led[] = {
> +	{
> +		.matches = {
> +			DMI_MATCH(DMI_SYS_VENDOR, "Dell EMC"),
> +			DMI_MATCH(DMI_PRODUCT_NAME, "Edge Gateway 3200"),
> +		},
> +	},
> +	{}
> +};

This is a PHY property. Why is the MAC involved?

Please also think about how to make this generic, so any ACPI based
system can use it, with any PHY.

     Andrew
Andrew Lunn Jan. 14, 2022, 1:20 p.m. UTC | #4 
>  static void marvell_config_led(struct phy_device *phydev)
>  {
> -	u16 def_config;
> +	struct marvell_priv *priv = phydev->priv;
>  	int err;
>  
> -	switch (MARVELL_PHY_FAMILY_ID(phydev->phy_id)) {
> -	/* Default PHY LED config: LED[0] .. Link, LED[1] .. Activity */
> -	case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1121R):
> -	case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1318S):
> -		def_config = MII_88E1121_PHY_LED_DEF;
> -		break;
> -	/* Default PHY LED config:
> -	 * LED[0] .. 1000Mbps Link
> -	 * LED[1] .. 100Mbps Link
> -	 * LED[2] .. Blink, Activity
> -	 */
> -	case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1510):
> -		if (phydev->dev_flags & MARVELL_PHY_LED0_LINK_LED1_ACTIVE)
> -			def_config = MII_88E1510_PHY_LED0_LINK_LED1_ACTIVE;
> -		else
> -			def_config = MII_88E1510_PHY_LED_DEF;
> -		break;
> -	default:
> +	if (priv->led_def_config == -1)
>  		return;
> +
> +	if (priv->led_def_config)
> +		goto write;

Really?

Please restructure this code. Take it apart into helpers. You need:

A function to set the actual LED configuration.
A function to decide what, if any, configuration to set
A function to store the current configuration on suspend.
A function to restore the current configuration on resume.

Lots of little functions will make it much easier to understand, and
avoid 1980s BASIC style.

I'm also surprised you need to deal with suspend/resume. Why does the
BIOS not set the LED mode on resume, same as it does on power up?

      Andrew
Kai-Heng Feng Jan. 14, 2022, 3:22 p.m. UTC | #5 
On Fri, Jan 14, 2022 at 9:09 PM Andrew Lunn <andrew@lunn.ch> wrote:
>
> On Fri, Jan 14, 2022 at 12:07:54PM +0800, Kai-Heng Feng wrote:
> > BIOS on Dell Edge Gateway 3200 already makes its own phy LED setting, so
> > instead of setting another value, keep it untouched and restore the saved
> > value on system resume.
> >
> > Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
> > ---
> >  .../net/ethernet/stmicro/stmmac/dwmac-intel.c | 16 +++++
> >  drivers/net/ethernet/stmicro/stmmac/stmmac.h  |  2 +
> >  .../net/ethernet/stmicro/stmmac/stmmac_main.c |  4 ++
> >  drivers/net/phy/marvell.c                     | 58 ++++++++++++-------
> >  include/linux/marvell_phy.h                   |  1 +
> >  5 files changed, 61 insertions(+), 20 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
> > index 8e8778cfbbadd..f8a2879e0264a 100644
> > --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
> > +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
> > @@ -857,6 +857,16 @@ static const struct dmi_system_id quark_pci_dmi[] = {
> >       {}
> >  };
> >
> > +static const struct dmi_system_id use_preset_led[] = {
> > +     {
> > +             .matches = {
> > +                     DMI_MATCH(DMI_SYS_VENDOR, "Dell EMC"),
> > +                     DMI_MATCH(DMI_PRODUCT_NAME, "Edge Gateway 3200"),
> > +             },
> > +     },
> > +     {}
> > +};
>
> This is a PHY property. Why is the MAC involved?

This is inspired by commit a93f7fe13454 ("net: phy: marvell: add new
default led configure for m88e151x") which passes the flag from MAC to
PHY.

>
> Please also think about how to make this generic, so any ACPI based
> system can use it, with any PHY.

ACPI property won't work because it ties to ACPI platform device or
PCI device, so the property still needs to be passed from MAC to PHY.

So the only approach I can think of is to use DMI match directly in PHY driver.

Kai-Heng

>
>      Andrew
Kai-Heng Feng Jan. 14, 2022, 3:25 p.m. UTC | #6 
On Fri, Jan 14, 2022 at 9:20 PM Andrew Lunn <andrew@lunn.ch> wrote:
>
> >  static void marvell_config_led(struct phy_device *phydev)
> >  {
> > -     u16 def_config;
> > +     struct marvell_priv *priv = phydev->priv;
> >       int err;
> >
> > -     switch (MARVELL_PHY_FAMILY_ID(phydev->phy_id)) {
> > -     /* Default PHY LED config: LED[0] .. Link, LED[1] .. Activity */
> > -     case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1121R):
> > -     case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1318S):
> > -             def_config = MII_88E1121_PHY_LED_DEF;
> > -             break;
> > -     /* Default PHY LED config:
> > -      * LED[0] .. 1000Mbps Link
> > -      * LED[1] .. 100Mbps Link
> > -      * LED[2] .. Blink, Activity
> > -      */
> > -     case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1510):
> > -             if (phydev->dev_flags & MARVELL_PHY_LED0_LINK_LED1_ACTIVE)
> > -                     def_config = MII_88E1510_PHY_LED0_LINK_LED1_ACTIVE;
> > -             else
> > -                     def_config = MII_88E1510_PHY_LED_DEF;
> > -             break;
> > -     default:
> > +     if (priv->led_def_config == -1)
> >               return;
> > +
> > +     if (priv->led_def_config)
> > +             goto write;
>
> Really?
>
> Please restructure this code. Take it apart into helpers. You need:
>
> A function to set the actual LED configuration.
> A function to decide what, if any, configuration to set
> A function to store the current configuration on suspend.
> A function to restore the current configuration on resume.
>
> Lots of little functions will make it much easier to understand, and
> avoid 1980s BASIC style.

Sure, will turn these into helper functions.

>
> I'm also surprised you need to deal with suspend/resume. Why does the
> BIOS not set the LED mode on resume, same as it does on power up?

I was told this is a BIOS limitation. I'll ask vendor _why_ the LED
can't be restored by BIOS.

Kai-Heng

>
>       Andrew
Jakub Kicinski Jan. 14, 2022, 4:23 p.m. UTC | #7 
On Fri, 14 Jan 2022 14:47:47 +0800 Kai-Heng Feng wrote:
> > Coincidentally the first Marvell flag appears dead, nobody sets it:
> >
> > $ git grep MARVELL_PHY_M1145_FLAGS_RESISTANCE
> > drivers/net/phy/marvell.c:      if (phydev->dev_flags & MARVELL_PHY_M1145_FLAGS_RESISTANCE) {
> > include/linux/marvell_phy.h:#define MARVELL_PHY_M1145_FLAGS_RESISTANCE  0x00000001
> > $
> >
> > unless it's read from DT under different name or something.  
> 
> It was introduced by 95d21ff4c645 without any user. Should we keep it?

Not unless someone explains that it's actually used somehow.

Please post a patch once net-next opens.
Andrew Lunn Jan. 14, 2022, 5:25 p.m. UTC | #8 
> > This is a PHY property. Why is the MAC involved?
> 
> This is inspired by commit a93f7fe13454 ("net: phy: marvell: add new
> default led configure for m88e151x") which passes the flag from MAC to
> PHY.

But in this case, the MAC does not care what the LEDs are. The
platform wants them left alone, not the MAC.

> > Please also think about how to make this generic, so any ACPI based
> > system can use it, with any PHY.

...

> So the only approach I can think of is to use DMI match directly in PHY driver.

In the phylib core. And the core then asks the specific PHY driver to
not touch the LED configuration. It is then generic.

      Andrew
diff mbox series

Patch

diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
index 8e8778cfbbadd..f8a2879e0264a 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c
@@ -857,6 +857,16 @@  static const struct dmi_system_id quark_pci_dmi[] = {
 	{}
 };
 
+static const struct dmi_system_id use_preset_led[] = {
+	{
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell EMC"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Edge Gateway 3200"),
+		},
+	},
+	{}
+};
+
 static int quark_default_data(struct pci_dev *pdev,
 			      struct plat_stmmacenet_data *plat)
 {
@@ -989,6 +999,7 @@  static int intel_eth_pci_probe(struct pci_dev *pdev,
 	struct intel_priv_data *intel_priv;
 	struct plat_stmmacenet_data *plat;
 	struct stmmac_resources res;
+	struct stmmac_priv *priv;
 	int ret;
 
 	intel_priv = devm_kzalloc(&pdev->dev, sizeof(*intel_priv), GFP_KERNEL);
@@ -1075,6 +1086,11 @@  static int intel_eth_pci_probe(struct pci_dev *pdev,
 		goto err_dvr_probe;
 	}
 
+	if (dmi_check_system(use_preset_led)) {
+		priv = netdev_priv(dev_get_drvdata(&pdev->dev));
+		priv->use_preset_led = true;
+	}
+
 	return 0;
 
 err_dvr_probe:
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index 40b5ed94cb54a..525701acbbdbb 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -319,6 +319,8 @@  struct stmmac_priv {
 	/* XDP BPF Program */
 	unsigned long *af_xdp_zc_qps;
 	struct bpf_prog *xdp_prog;
+
+	bool use_preset_led;
 };
 
 enum stmmac_state {
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 63ff2dad8c85f..155412656b8bf 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -49,6 +49,7 @@ 
 #include "dwmac1000.h"
 #include "dwxgmac2.h"
 #include "hwif.h"
+#include <linux/marvell_phy.h>
 
 /* As long as the interface is active, we keep the timestamping counter enabled
  * with fine resolution and binary rollover. This avoid non-monotonic behavior
@@ -1236,6 +1237,9 @@  static int stmmac_init_phy(struct net_device *dev)
 			return -ENODEV;
 		}
 
+		if (priv->use_preset_led)
+			phydev->dev_flags |= MARVELL_PHY_USE_PRESET_LED;
+
 		ret = phylink_connect_phy(priv->phylink, phydev);
 	}
 
diff --git a/drivers/net/phy/marvell.c b/drivers/net/phy/marvell.c
index 739859c0dfb18..45be432188781 100644
--- a/drivers/net/phy/marvell.c
+++ b/drivers/net/phy/marvell.c
@@ -304,6 +304,7 @@  struct marvell_priv {
 	u32 last;
 	u32 step;
 	s8 pair;
+	u16 led_def_config;
 };
 
 static int marvell_read_page(struct phy_device *phydev)
@@ -748,32 +749,49 @@  static int m88e1510_config_aneg(struct phy_device *phydev)
 
 static void marvell_config_led(struct phy_device *phydev)
 {
-	u16 def_config;
+	struct marvell_priv *priv = phydev->priv;
 	int err;
 
-	switch (MARVELL_PHY_FAMILY_ID(phydev->phy_id)) {
-	/* Default PHY LED config: LED[0] .. Link, LED[1] .. Activity */
-	case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1121R):
-	case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1318S):
-		def_config = MII_88E1121_PHY_LED_DEF;
-		break;
-	/* Default PHY LED config:
-	 * LED[0] .. 1000Mbps Link
-	 * LED[1] .. 100Mbps Link
-	 * LED[2] .. Blink, Activity
-	 */
-	case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1510):
-		if (phydev->dev_flags & MARVELL_PHY_LED0_LINK_LED1_ACTIVE)
-			def_config = MII_88E1510_PHY_LED0_LINK_LED1_ACTIVE;
-		else
-			def_config = MII_88E1510_PHY_LED_DEF;
-		break;
-	default:
+	if (priv->led_def_config == -1)
 		return;
+
+	if (priv->led_def_config)
+		goto write;
+
+	if (phydev->dev_flags & MARVELL_PHY_USE_PRESET_LED) {
+		priv->led_def_config = phy_read_paged(phydev, MII_MARVELL_LED_PAGE,
+						      MII_PHY_LED_CTRL);
+		if (priv->led_def_config < 0) {
+			priv->led_def_config = -1;
+			return;
+		}
+	} else {
+		switch (MARVELL_PHY_FAMILY_ID(phydev->phy_id)) {
+		/* Default PHY LED config: LED[0] .. Link, LED[1] .. Activity */
+		case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1121R):
+		case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1318S):
+			priv->led_def_config = MII_88E1121_PHY_LED_DEF;
+			break;
+		/* Default PHY LED config:
+		 * LED[0] .. 1000Mbps Link
+		 * LED[1] .. 100Mbps Link
+		 * LED[2] .. Blink, Activity
+		 */
+		case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1510):
+			if (phydev->dev_flags & MARVELL_PHY_LED0_LINK_LED1_ACTIVE)
+				priv->led_def_config = MII_88E1510_PHY_LED0_LINK_LED1_ACTIVE;
+			else
+				priv->led_def_config = MII_88E1510_PHY_LED_DEF;
+			break;
+		default:
+			priv->led_def_config = -1;
+			return;
+		}
 	}
 
+write:
 	err = phy_write_paged(phydev, MII_MARVELL_LED_PAGE, MII_PHY_LED_CTRL,
-			      def_config);
+			      priv->led_def_config);
 	if (err < 0)
 		phydev_warn(phydev, "Fail to config marvell phy LED.\n");
 }
diff --git a/include/linux/marvell_phy.h b/include/linux/marvell_phy.h
index ea5995d9ad6c1..492f07620b6c0 100644
--- a/include/linux/marvell_phy.h
+++ b/include/linux/marvell_phy.h
@@ -43,5 +43,6 @@ 
 #define MARVELL_PHY_M1145_FLAGS_RESISTANCE	BIT(0)
 #define MARVELL_PHY_M1118_DNS323_LEDS		BIT(1)
 #define MARVELL_PHY_LED0_LINK_LED1_ACTIVE	BIT(2)
+#define MARVELL_PHY_USE_PRESET_LED		BIT(3)
 
 #endif /* _MARVELL_PHY_H */