| Message ID | 20201018163625.2392-1-ardb@kernel.org (mailing list archive) |
|---|---|
| State | Not Applicable |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] netsec: ignore 'phy-mode' device property on ACPI systems | expand |
> --- a/Documentation/devicetree/bindings/net/socionext-netsec.txt > +++ b/Documentation/devicetree/bindings/net/socionext-netsec.txt > @@ -30,7 +30,9 @@ Optional properties: (See ethernet.txt file in the same directory) > - max-frame-size: See ethernet.txt in the same directory. > > The MAC address will be determined using the optional properties > -defined in ethernet.txt. > +defined in ethernet.txt. The 'phy-mode' property is required, but may > +be set to the empty string if the PHY configuration is programmed by > +the firmware or set by hardware straps, and needs to be preserved. In general, phy-mode is not mandatory. of_get_phy_mode() does the right thing if it is not found, it sets &priv->phy_interface to PHY_INTERFACE_MODE_NA, but returns -ENODEV. Also, it does not break backwards compatibility to convert a mandatory property to optional. So you could just do of_get_phy_mode(pdev->dev.of_node, &priv->phy_interface); skip all the error checking, and document it as optional. Andrew
On Sun, Oct 18, 2020 at 07:52:18PM +0200, Andrew Lunn wrote: > > --- a/Documentation/devicetree/bindings/net/socionext-netsec.txt > > +++ b/Documentation/devicetree/bindings/net/socionext-netsec.txt > > @@ -30,7 +30,9 @@ Optional properties: (See ethernet.txt file in the same directory) > > - max-frame-size: See ethernet.txt in the same directory. > > > > The MAC address will be determined using the optional properties > > -defined in ethernet.txt. > > +defined in ethernet.txt. The 'phy-mode' property is required, but may > > +be set to the empty string if the PHY configuration is programmed by > > +the firmware or set by hardware straps, and needs to be preserved. > > In general, phy-mode is not mandatory. of_get_phy_mode() does the > right thing if it is not found, it sets &priv->phy_interface to > PHY_INTERFACE_MODE_NA, but returns -ENODEV. Also, it does not break > backwards compatibility to convert a mandatory property to > optional. So you could just do > > of_get_phy_mode(pdev->dev.of_node, &priv->phy_interface); > > skip all the error checking, and document it as optional. Why ? The patch as is will not affect systems built on any firmware implementations that use ACPI and somehow configure the hardware. Although the only firmware implementations I am aware of on upsteream are based on EDK2, I prefer the explicit error as is now, in case a firmware does on initialize the PHY properly (and is using a DT). Cheers /Ilias > > Andrew
On Sun, 18 Oct 2020 at 22:32, Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: > > On Sun, Oct 18, 2020 at 07:52:18PM +0200, Andrew Lunn wrote: > > > --- a/Documentation/devicetree/bindings/net/socionext-netsec.txt > > > +++ b/Documentation/devicetree/bindings/net/socionext-netsec.txt > > > @@ -30,7 +30,9 @@ Optional properties: (See ethernet.txt file in the same directory) > > > - max-frame-size: See ethernet.txt in the same directory. > > > > > > The MAC address will be determined using the optional properties > > > -defined in ethernet.txt. > > > +defined in ethernet.txt. The 'phy-mode' property is required, but may > > > +be set to the empty string if the PHY configuration is programmed by > > > +the firmware or set by hardware straps, and needs to be preserved. > > > > In general, phy-mode is not mandatory. of_get_phy_mode() does the > > right thing if it is not found, it sets &priv->phy_interface to > > PHY_INTERFACE_MODE_NA, but returns -ENODEV. Also, it does not break > > backwards compatibility to convert a mandatory property to > > optional. So you could just do > > > > of_get_phy_mode(pdev->dev.of_node, &priv->phy_interface); > > > > skip all the error checking, and document it as optional. > > Why ? > The patch as is will not affect systems built on any firmware implementations > that use ACPI and somehow configure the hardware. > Although the only firmware implementations I am aware of on upsteream are based > on EDK2, I prefer the explicit error as is now, in case a firmware does on > initialize the PHY properly (and is using a DT). > We will also lose the ability to report bogus values for phy-mode this way, so I think we should stick with the check.
Hi Ard, On Mon, Oct 19, 2020 at 08:30:45AM +0200, Ard Biesheuvel wrote: > On Sun, 18 Oct 2020 at 22:32, Ilias Apalodimas > <ilias.apalodimas@linaro.org> wrote: > > > > On Sun, Oct 18, 2020 at 07:52:18PM +0200, Andrew Lunn wrote: > > > > --- a/Documentation/devicetree/bindings/net/socionext-netsec.txt > > > > +++ b/Documentation/devicetree/bindings/net/socionext-netsec.txt > > > > @@ -30,7 +30,9 @@ Optional properties: (See ethernet.txt file in the same directory) > > > > - max-frame-size: See ethernet.txt in the same directory. > > > > > > > > The MAC address will be determined using the optional properties > > > > -defined in ethernet.txt. > > > > +defined in ethernet.txt. The 'phy-mode' property is required, but may > > > > +be set to the empty string if the PHY configuration is programmed by > > > > +the firmware or set by hardware straps, and needs to be preserved. > > > > > > In general, phy-mode is not mandatory. of_get_phy_mode() does the > > > right thing if it is not found, it sets &priv->phy_interface to > > > PHY_INTERFACE_MODE_NA, but returns -ENODEV. Also, it does not break > > > backwards compatibility to convert a mandatory property to > > > optional. So you could just do > > > > > > of_get_phy_mode(pdev->dev.of_node, &priv->phy_interface); > > > > > > skip all the error checking, and document it as optional. > > > > Why ? > > The patch as is will not affect systems built on any firmware implementations > > that use ACPI and somehow configure the hardware. > > Although the only firmware implementations I am aware of on upsteream are based > > on EDK2, I prefer the explicit error as is now, in case a firmware does on > > initialize the PHY properly (and is using a DT). > > > > We will also lose the ability to report bogus values for phy-mode this > way, so I think we should stick with the check. I hope Andrew is fine with the current changes Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
> I hope Andrew is fine with the current changes
Yes, i'm O.K. with it. Making phy-mode optional would just make the
driver more uniform with others.
Andrew
On Sun, Oct 18, 2020 at 06:36:25PM +0200, Ard Biesheuvel wrote: > Since commit bbc4d71d63549bc ("net: phy: realtek: fix rtl8211e rx/tx > delay config"), the Realtek PHY driver will override any TX/RX delay > set by hardware straps if the phy-mode device property does not match. > > This is causing problems on SynQuacer based platforms (the only SoC > that incorporates the netsec hardware), since many were built with > this Realtek PHY, and shipped with firmware that defines the phy-mode > as 'rgmii', even though the PHY is configured for TX and RX delay using > pull-ups. > > >From the driver's perspective, we should not make any assumptions in > the general case that the PHY hardware does not require any initial > configuration. However, the situation is slightly different for ACPI > boot, since it implies rich firmware with AML abstractions to handle > hardware details that are not exposed to the OS. So in the ACPI case, > it is reasonable to assume that the PHY comes up in the right mode, > regardless of whether the mode is set by straps, by boot time firmware > or by AML executed by the ACPI interpreter. > > So let's ignore the 'phy-mode' device property when probing the netsec > driver in ACPI mode, and hardcode the mode to PHY_INTERFACE_MODE_NA, > which should work with any PHY provided that it is configured by the > time the driver attaches to it. While at it, document that omitting > the mode is permitted for DT probing as well, by setting the phy-mode > DT property to the empty string. > > Cc: Jassi Brar <jaswinder.singh@linaro.org> > Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org> > Cc: Masami Hiramatsu <mhiramat@kernel.org> > Cc: Andrew Lunn <andrew@lunn.ch> > Cc: Willy Liu <willy.liu@realtek.com> > Cc: "David S. Miller" <davem@davemloft.net> > Cc: Jakub Kicinski <kuba@kernel.org> > Cc: Masahisa Kojima <masahisa.kojima@linaro.org> > Cc: Serge Semin <fancer.lancer@gmail.com> > Fixes: 533dd11a12f6 ("net: socionext: Add Synquacer NetSec driver") > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Andrew
On Tue, 20 Oct 2020 at 14:49, Andrew Lunn <andrew@lunn.ch> wrote: > > > I hope Andrew is fine with the current changes > > Yes, i'm O.K. with it. Thanks > Making phy-mode optional would just make the > driver more uniform with others. > Making phy-mode optional is fine with me, but I think it would belong in a separate patch in any case. But I'd still prefer having the possibility to spot bogus phy-mode values rather than ignoring them.
On Tue, 20 Oct 2020 14:50:01 +0200 Andrew Lunn wrote: > On Sun, Oct 18, 2020 at 06:36:25PM +0200, Ard Biesheuvel wrote: > > Since commit bbc4d71d63549bc ("net: phy: realtek: fix rtl8211e rx/tx > > delay config"), the Realtek PHY driver will override any TX/RX delay > > set by hardware straps if the phy-mode device property does not match. > > > > This is causing problems on SynQuacer based platforms (the only SoC > > that incorporates the netsec hardware), since many were built with > > this Realtek PHY, and shipped with firmware that defines the phy-mode > > as 'rgmii', even though the PHY is configured for TX and RX delay using > > pull-ups. > > > > >From the driver's perspective, we should not make any assumptions in > > the general case that the PHY hardware does not require any initial > > configuration. However, the situation is slightly different for ACPI > > boot, since it implies rich firmware with AML abstractions to handle > > hardware details that are not exposed to the OS. So in the ACPI case, > > it is reasonable to assume that the PHY comes up in the right mode, > > regardless of whether the mode is set by straps, by boot time firmware > > or by AML executed by the ACPI interpreter. > > > > So let's ignore the 'phy-mode' device property when probing the netsec > > driver in ACPI mode, and hardcode the mode to PHY_INTERFACE_MODE_NA, > > which should work with any PHY provided that it is configured by the > > time the driver attaches to it. While at it, document that omitting > > the mode is permitted for DT probing as well, by setting the phy-mode > > DT property to the empty string. > > > > Fixes: 533dd11a12f6 ("net: socionext: Add Synquacer NetSec driver") > > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> > > Reviewed-by: Andrew Lunn <andrew@lunn.ch> Thanks, applied. Just to be on the safe side please make sure to CC Rob & DT list if your patch touches anything device tree. > --- a/Documentation/devicetree/bindings/net/socionext-netsec.txt > +++ b/Documentation/devicetree/bindings/net/socionext-netsec.txt > @@ -30,7 +30,9 @@ Optional properties: (See ethernet.txt file in the same directory) > - max-frame-size: See ethernet.txt in the same directory. > > The MAC address will be determined using the optional properties > -defined in ethernet.txt. > +defined in ethernet.txt. The 'phy-mode' property is required, but may > +be set to the empty string if the PHY configuration is programmed by > +the firmware or set by hardware straps, and needs to be preserved. > > Example: > eth0: ethernet@522d0000 {
diff --git a/Documentation/devicetree/bindings/net/socionext-netsec.txt b/Documentation/devicetree/bindings/net/socionext-netsec.txt index 9d6c9feb12ff..a3c1dffaa4bb 100644 --- a/Documentation/devicetree/bindings/net/socionext-netsec.txt +++ b/Documentation/devicetree/bindings/net/socionext-netsec.txt @@ -30,7 +30,9 @@ Optional properties: (See ethernet.txt file in the same directory) - max-frame-size: See ethernet.txt in the same directory. The MAC address will be determined using the optional properties -defined in ethernet.txt. +defined in ethernet.txt. The 'phy-mode' property is required, but may +be set to the empty string if the PHY configuration is programmed by +the firmware or set by hardware straps, and needs to be preserved. Example: eth0: ethernet@522d0000 { diff --git a/drivers/net/ethernet/socionext/netsec.c b/drivers/net/ethernet/socionext/netsec.c index 806eb651cea3..1503cc9ec6e2 100644 --- a/drivers/net/ethernet/socionext/netsec.c +++ b/drivers/net/ethernet/socionext/netsec.c @@ -6,6 +6,7 @@ #include <linux/pm_runtime.h> #include <linux/acpi.h> #include <linux/of_mdio.h> +#include <linux/of_net.h> #include <linux/etherdevice.h> #include <linux/interrupt.h> #include <linux/io.h> @@ -1833,6 +1834,14 @@ static const struct net_device_ops netsec_netdev_ops = { static int netsec_of_probe(struct platform_device *pdev, struct netsec_priv *priv, u32 *phy_addr) { + int err; + + err = of_get_phy_mode(pdev->dev.of_node, &priv->phy_interface); + if (err) { + dev_err(&pdev->dev, "missing required property 'phy-mode'\n"); + return err; + } + priv->phy_np = of_parse_phandle(pdev->dev.of_node, "phy-handle", 0); if (!priv->phy_np) { dev_err(&pdev->dev, "missing required property 'phy-handle'\n"); @@ -1859,6 +1868,14 @@ static int netsec_acpi_probe(struct platform_device *pdev, if (!IS_ENABLED(CONFIG_ACPI)) return -ENODEV; + /* ACPI systems are assumed to configure the PHY in firmware, so + * there is really no need to discover the PHY mode from the DSDT. + * Since firmware is known to exist in the field that configures the + * PHY correctly but passes the wrong mode string in the phy-mode + * device property, we have no choice but to ignore it. + */ + priv->phy_interface = PHY_INTERFACE_MODE_NA; + ret = device_property_read_u32(&pdev->dev, "phy-channel", phy_addr); if (ret) { dev_err(&pdev->dev, @@ -1995,13 +2012,6 @@ static int netsec_probe(struct platform_device *pdev) priv->msg_enable = NETIF_MSG_TX_ERR | NETIF_MSG_HW | NETIF_MSG_DRV | NETIF_MSG_LINK | NETIF_MSG_PROBE; - priv->phy_interface = device_get_phy_mode(&pdev->dev); - if ((int)priv->phy_interface < 0) { - dev_err(&pdev->dev, "missing required property 'phy-mode'\n"); - ret = -ENODEV; - goto free_ndev; - } - priv->ioaddr = devm_ioremap(&pdev->dev, mmio_res->start, resource_size(mmio_res)); if (!priv->ioaddr) {
Since commit bbc4d71d63549bc ("net: phy: realtek: fix rtl8211e rx/tx delay config"), the Realtek PHY driver will override any TX/RX delay set by hardware straps if the phy-mode device property does not match. This is causing problems on SynQuacer based platforms (the only SoC that incorporates the netsec hardware), since many were built with this Realtek PHY, and shipped with firmware that defines the phy-mode as 'rgmii', even though the PHY is configured for TX and RX delay using pull-ups. From the driver's perspective, we should not make any assumptions in the general case that the PHY hardware does not require any initial configuration. However, the situation is slightly different for ACPI boot, since it implies rich firmware with AML abstractions to handle hardware details that are not exposed to the OS. So in the ACPI case, it is reasonable to assume that the PHY comes up in the right mode, regardless of whether the mode is set by straps, by boot time firmware or by AML executed by the ACPI interpreter. So let's ignore the 'phy-mode' device property when probing the netsec driver in ACPI mode, and hardcode the mode to PHY_INTERFACE_MODE_NA, which should work with any PHY provided that it is configured by the time the driver attaches to it. While at it, document that omitting the mode is permitted for DT probing as well, by setting the phy-mode DT property to the empty string. Cc: Jassi Brar <jaswinder.singh@linaro.org> Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Andrew Lunn <andrew@lunn.ch> Cc: Willy Liu <willy.liu@realtek.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Masahisa Kojima <masahisa.kojima@linaro.org> Cc: Serge Semin <fancer.lancer@gmail.com> Fixes: 533dd11a12f6 ("net: socionext: Add Synquacer NetSec driver") Signed-off-by: Ard Biesheuvel <ardb@kernel.org> --- Related discussion here: https://lore.kernel.org/netdev/CAMj1kXEEF_Un-4NTaD5iUN0NoZYaJQn-rPediX0S6oRiuVuW-A@mail.gmail.com/ Documentation/devicetree/bindings/net/socionext-netsec.txt | 4 +++- drivers/net/ethernet/socionext/netsec.c | 24 ++++++++++++++------ 2 files changed, 20 insertions(+), 8 deletions(-)