Skip to contentSkip to conversation

net: axienet: modernise pcs implementation

PATCH Superseded Russell King (Oracle) 5 participants 15 messages Dec 16, 2021 – Jan 4, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover net: axienet: modernise pcs implementation

Hi,

These two patches modernise the Xilinx axienet PCS implementation to use the phylink split PCS support.

The first patch adds split PCS support and makes use of the newly introduced mac_select_pcs() function, which is the preferred way to conditionally attach a PCS.

The second patch cleans up the use of mdiobus_write() since we now have bus accessors for mdio devices.

There should be no functional change to the driver.

Please test, thanks.

 drivers/net/ethernet/xilinx/xilinx_axienet.h      |   1 +
 drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 111 +++++++++++-----------
 2 files changed, 56 insertions(+), 56 deletions(-)
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#1 lore raw

PATCH 1/2 net: axienet: convert to phylink_pcs

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

---
 drivers/net/ethernet/xilinx/xilinx_axienet.h  |   1 +
 .../net/ethernet/xilinx/xilinx_axienet_main.c | 112 +++++++++---------
 2 files changed, 57 insertions(+), 56 deletions(-)
drivers/net/ethernet/xilinx/xilinx_axienet.h +1 -0
index 5b4d153b1492..81f09bd8f11c 100644
@@ -434,6 +434,7 @@ struct axienet_local {
 	struct phylink_config phylink_config;
 
 	struct mdio_device *pcs_phy;
+	struct phylink_pcs pcs;
 
 	bool switch_x_sgmii;
 
drivers/net/ethernet/xilinx/xilinx_axienet_main.c +56 -56

155 lines not on this page · raw message

Signature · 1 line
2.30.2
#10 lore raw

PATCH 2/2 net: axienet: replace mdiobus_write() with mdiodev_write()

In reply to Russell King (Oracle)

Commit 197a68ef1837 ("net: mdio: Add helper functions for accessing MDIO devices") added support for mdiodev accessor operations that neatly wrap the mdiobus accessor operations. Since we are dealing with a mdio device here, update the driver to use mdiodev_write().

---
 drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/net/ethernet/xilinx/xilinx_axienet_main.c +1 -2
index 5edc8ec72317..ac23d1c65ac8 100644
@@ -1543,8 +1543,7 @@ static int axienet_pcs_config(struct phylink_pcs *pcs, unsigned int mode,
 	if (!lp->switch_x_sgmii)
 		return 0;
 
-	ret = mdiobus_write(pcs_phy->bus, pcs_phy->addr,
-			    XLNX_MII_STD_SELECT_REG,
+	ret = mdiodev_write(pcs_phy, XLNX_MII_STD_SELECT_REG,
 			    interface == PHY_INTERFACE_MODE_SGMII ?
 				XLNX_MII_STD_SELECT_SGMII : 0);
 	if (ret < 0) {
Signature · 1 line
2.30.2
maintainer
#11 lore raw

Re: [PATCH CFT net-next 2/2] net: axienet: replace mdiobus_write() with mdiodev_write()

Quoted 2 lines · On Thu, 16 Dec 2021 12:48:50 +0000 Russell King (Oracle) wrote:

Commit 197a68ef1837 ("net: mdio: Add helper functions for accessing MDIO devices") added support for mdiodev accessor operations that

Nit: wrong commit id, 0ebecb2644c8 ("net: mdio: Add helper functions for accessing MDIO devices")?

maintainer
#2 lore raw

Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

In reply to Russell King (Oracle)
Quoted 10 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function parameter or member 'pcs' not described in 'axienet_local'

#3 lore raw

Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 10 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

Quoted 3 lines
Quoted 1 line

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function parameter or member 'pcs' not described in 'axienet_local'

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Thanks!

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

RE: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

+ Harini

Quoted 21 lines

-----Original Message-----
From: Russell King <[email protected]>
Sent: Thursday, December 16, 2021 10:32 PM
To: Jakub Kicinski <[email protected]>
Cc: Michal Simek <[email protected]>; Radhey Shyam Pandey <[email protected]>; Sean Anderson <[email protected]>; David S. Miller <[email protected]>; [email protected]; linux-arm- [email protected]
Subject: Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 10 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function

parameter or member 'pcs' not described in 'axienet_local'

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Thanks, series looks fine. I have added Harini to this thread and she will confirm it on hardware.

Quoted 6 lines

Thanks!

--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!

#5 lore raw

Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

In reply to Russell King (Oracle)

Hi Russell,

Quoted 12 lines · Russell King (Oracle)
Quoted 10 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function parameter or member 'pcs' not described in 'axienet_local'

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Thanks for the patches.
Series looks good and we're testing at our end; will get back to you early next week.

Regards,
Harini

#6 lore raw

Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 21 lines · Harini Katakam

Hi Russell,

Quoted 16 lines · Russell King (Oracle)
Quoted 13 lines · Jakub Kicinski
Quoted 10 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function parameter or member 'pcs' not described in 'axienet_local'

Quoted 7 lines
Quoted 3 lines

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Thanks for the patches.
Series looks good and we're testing at our end; will get back to you early next week.

Thanks Russell. I've tested AXI Ethernet and it works fine.

Regards,
Harini

#7 lore raw

Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 26 lines · Harini Katakam
Quoted 25 lines · Harini Katakam

Hi Russell,

Quoted 19 lines · Russell King (Oracle)
Quoted 10 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

Quoted 3 lines
Quoted 1 line

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function parameter or member 'pcs' not described in 'axienet_local'

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Quoted 5 lines
Quoted 3 lines

Thanks for the patches.
Series looks good and we're testing at our end; will get back to you early next week.

Thanks Russell. I've tested AXI Ethernet and it works fine.

Happy new year!

Thanks - can I use that as a tested-by please, and would you be happy for me to send the patches for merging this week?

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

RE: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 45 lines

-----Original Message-----
From: Russell King <[email protected]>
Sent: Tuesday, January 4, 2022 2:31 PM
To: Harini Katakam <[email protected]>
Cc: Jakub Kicinski <[email protected]>; Michal Simek <[email protected]>; Radhey Shyam Pandey <[email protected]>; Sean Anderson <[email protected]>; David S. Miller <[email protected]>; netdev <[email protected]>; [email protected]
Subject: Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 18 lines · Harini Katakam
Quoted 17 lines · Harini Katakam

Hi Russell,

Quoted 12 lines · Russell King (Oracle)
Quoted 10 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines

Signed-off-by: Russell King (Oracle)
<[email protected]>

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function

parameter or member 'pcs' not described in 'axienet_local'

Quoted 9 lines
Quoted 8 lines
Quoted 4 lines

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Thanks for the patches.
Series looks good and we're testing at our end; will get back to you early next week.

Quoted 6 lines
Quoted 1 line

Thanks Russell. I've tested AXI Ethernet and it works fine.

Happy new year!

Thanks - can I use that as a tested-by please, and would you be happy for me to send the patches for merging this week?

Sure, yes and yes.

Happy new year to you too!

Regards,
Harini

#9 lore raw

Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 47 lines · Harini Katakam
Quoted 47 lines

-----Original Message-----
From: Russell King <[email protected]>
Sent: Tuesday, January 4, 2022 2:31 PM
To: Harini Katakam <[email protected]>
Cc: Jakub Kicinski <[email protected]>; Michal Simek <[email protected]>; Radhey Shyam Pandey <[email protected]>; Sean Anderson <[email protected]>; David S. Miller <[email protected]>; netdev <[email protected]>; [email protected]
Subject: Re: [PATCH CFT net-next 1/2] net: axienet: convert to phylink_pcs

Quoted 18 lines · Harini Katakam
Quoted 17 lines · Harini Katakam

Hi Russell,

Quoted 12 lines · Russell King (Oracle)
Quoted 10 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 16 Dec 2021 12:48:45 +0000 Russell King (Oracle) wrote:

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines

Signed-off-by: Russell King (Oracle)
<[email protected]>

drivers/net/ethernet/xilinx/xilinx_axienet.h:479: warning: Function

parameter or member 'pcs' not described in 'axienet_local'

Quoted 10 lines
Quoted 8 lines
Quoted 4 lines

Fixed that and the sha1 issue you raised in patch 2. Since both are "documentation" issues, I won't send out replacement patches until I've heard they've been tested on hardware though.

Thanks for the patches.
Series looks good and we're testing at our end; will get back to you early next week.

Thanks Russell. I've tested AXI Ethernet and it works fine.

Happy new year!

Quoted 7 lines
Quoted 2 lines

Thanks - can I use that as a tested-by please, and would you be happy for me to send the patches for merging this week?

Sure, yes and yes.

Happy new year to you too!

Thanks. While adding the attributation, I was reminded of this comment in the commit message:

One oddity in this driver is that lp->switch_x_sgmii controls whether we support switching between SGMII and 1000baseX. However, when clear, this also blocks updating the 1000baseX advertisement, which it probably should not be doing. Nevertheless, this behaviour is preserved but a comment is added.

went back to look at that, and realised that this was not the case at all, so patch 1 introduces a behaviour that wasn't originally there. I'll post an update, but essentially the change to patch 1 is:

drivers/net/ethernet/xilinx/xilinx_axienet_main.c +11 -12
index a556f0215049..fbe0de4bc8dd 100644
@@ -1533,18 +1533,17 @@ static int axienet_pcs_config(struct phylink_pcs *pcs, unsigned int mode,
 	struct axienet_local *lp = netdev_priv(ndev);
 	int ret;
 
-	/* We don't support changing the advertisement in 1000base-X? --rmk */
-	if (!lp->switch_x_sgmii)
-		return 0;
-
-	ret = mdiobus_write(pcs_phy->bus, pcs_phy->addr,
-			    XLNX_MII_STD_SELECT_REG,
-			    interface == PHY_INTERFACE_MODE_SGMII ?
-				XLNX_MII_STD_SELECT_SGMII : 0);
-	if (ret < 0) {
-		netdev_warn(ndev, "Failed to switch PHY interface: %d\n",
-			    ret);
-		return ret;
+	if (lp->switch_x_sgmii) {
+		ret = mdiobus_write(pcs_phy->bus, pcs_phy->addr,
+				    XLNX_MII_STD_SELECT_REG,
+				    interface == PHY_INTERFACE_MODE_SGMII ?
+					XLNX_MII_STD_SELECT_SGMII : 0);
+		if (ret < 0) {
+			netdev_warn(ndev,
+				    "Failed to switch PHY interface: %d\n",
+				    ret);
+			return ret;
+		}
 	}
 
 	ret = phylink_mii_c22_pcs_config(pcs_phy, mode, interface, advertising);

and a corresponding change to patch 2 for the change in code formatting.

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

v2 cover net: axienet: modernise pcs implementation

In reply to Russell King (Oracle)

Hi,

These two patches modernise the Xilinx axienet PCS implementation to use the phylink split PCS support.

The first patch adds split PCS support and makes use of the newly introduced mac_select_pcs() function, which is the preferred way to conditionally attach a PCS.

The second patch cleans up the use of mdiobus_write() since we now have bus accessors for mdio devices.

There should be no functional change to the driver.

Please test, thanks.

v2: remove question over switch_x_sgmii

 drivers/net/ethernet/xilinx/xilinx_axienet.h      |   2 +
 drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 108 +++++++++++-----------
 2 files changed, 55 insertions(+), 55 deletions(-)
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#13 lore raw

PATCH v2 1/2 net: axienet: convert to phylink_pcs

Convert axienet to use the phylink_pcs layer, resulting in it no longer being a legacy driver.

---
 drivers/net/ethernet/xilinx/xilinx_axienet.h  |   2 +
 .../net/ethernet/xilinx/xilinx_axienet_main.c | 107 +++++++++---------
 2 files changed, 55 insertions(+), 54 deletions(-)
drivers/net/ethernet/xilinx/xilinx_axienet.h +2 -0
index 5b4d153b1492..40108968b350 100644
@@ -386,6 +386,7 @@ struct axidma_bd {
  * @phylink:	Pointer to phylink instance
  * @phylink_config: phylink configuration settings
  * @pcs_phy:	Reference to PCS/PMA PHY if used
+ * @pcs:	phylink pcs structure for PCS PHY
  * @switch_x_sgmii: Whether switchable 1000BaseX/SGMII mode is enabled in the core
  * @axi_clk:	AXI4-Lite bus clock
  * @misc_clks:	Misc ethernet clocks (AXI4-Stream, Ref, MGT clocks)
@@ -434,6 +435,7 @@ struct axienet_local {
 	struct phylink_config phylink_config;
 
 	struct mdio_device *pcs_phy;
+	struct phylink_pcs pcs;
 
 	bool switch_x_sgmii;
 
drivers/net/ethernet/xilinx/xilinx_axienet_main.c +53 -54

152 lines not on this page · raw message

Signature · 1 line
2.30.2
#14 lore raw

PATCH v2 2/2 net: axienet: replace mdiobus_write() with mdiodev_write()

In reply to Russell King (Oracle)

Commit 0ebecb2644c8 ("net: mdio: Add helper functions for accessing MDIO devices") added support for mdiodev accessor operations that neatly wrap the mdiobus accessor operations. Since we are dealing with a mdio device here, update the driver to use mdiodev_write().

---
 drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/net/ethernet/xilinx/xilinx_axienet_main.c +1 -2
index 7a28355114a8..42ffa8500c2b 100644
@@ -1540,8 +1540,7 @@ static int axienet_pcs_config(struct phylink_pcs *pcs, unsigned int mode,
 	int ret;
 
 	if (lp->switch_x_sgmii) {
-		ret = mdiobus_write(pcs_phy->bus, pcs_phy->addr,
-				    XLNX_MII_STD_SELECT_REG,
+		ret = mdiodev_write(pcs_phy, XLNX_MII_STD_SELECT_REG,
 				    interface == PHY_INTERFACE_MODE_SGMII ?
 					XLNX_MII_STD_SELECT_SGMII : 0);
 		if (ret < 0) {
Signature · 1 line
2.30.2