From: Kelvin Cheung <[email protected]>
The Ethernet of LS1B/LS1C doesn't work due to the stmmac driver
using phylink_generic_validate() instead of stmmac_validate().
Moreover the driver assumes the PHY interface mode
passed in platform data is always supported.
stmmaceth stmmaceth.0 eth0: validation of gmii with support 00000000,00000000,000062cf and advertisement 00000000,00000000,000062cf failed: -EINVAL
stmmaceth stmmaceth.0 eth0: stmmac_open: Cannot attach to PHY (error: -22)
This patch sets phy_interface field of platform data.
Fixes: 04a0683f7db4 ("net: stmmac: convert to phylink_generic_validate()")
Fixes: d194923d51c9 ("net: stmmac: fill in supported_interfaces")
Signed-off-by: Kelvin Cheung <[email protected]>
---
arch/mips/loongson32/common/platform.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/mips/loongson32/common/platform.c b/arch/mips/loongson32/common/platform.c
index 794c96c2a4cd..741aace80b80 100644
--- a/arch/mips/loongson32/common/platform.c
+++ b/arch/mips/loongson32/common/platform.c
@@ -147,8 +147,10 @@ static struct plat_stmmacenet_data ls1x_eth0_pdata = {
.phy_addr = -1,
#if defined(CONFIG_LOONGSON1_LS1B)
.interface = PHY_INTERFACE_MODE_MII,
+ .phy_interface = PHY_INTERFACE_MODE_MII,
#elif defined(CONFIG_LOONGSON1_LS1C)
.interface = PHY_INTERFACE_MODE_RMII,
+ .phy_interface = PHY_INTERFACE_MODE_RMII,
#endif
.mdio_bus_data = &ls1x_mdio_bus_data,
.dma_cfg = &ls1x_eth_dma_cfg,
base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
--
2.34.1
Hello Keguang
On Thu, Aug 18, 2022 at 01:00:19PM +0800, Keguang Zhang wrote:
> From: Kelvin Cheung <[email protected]>
>
> The Ethernet of LS1B/LS1C doesn't work due to the stmmac driver
> using phylink_generic_validate() instead of stmmac_validate().
> Moreover the driver assumes the PHY interface mode
> passed in platform data is always supported.
>
> stmmaceth stmmaceth.0 eth0: validation of gmii with support 00000000,00000000,000062cf and advertisement 00000000,00000000,000062cf failed: -EINVAL
> stmmaceth stmmaceth.0 eth0: stmmac_open: Cannot attach to PHY (error: -22)
>
> This patch sets phy_interface field of platform data.
I've got a similar fix in my repo, though didn't have a chance to test
it out due to lacking any loongson hardware. I've discovered the
issues on my still going way of the STMMAC driver refactoring. Anyway
IMO the problem is a bit different than you describe and should be
fixed in a bit different way. Please see a patch attached to this
email. Could you test it out on your hw? If it fixes the problem you
can resend it as v2 patch.
-Sergey
>
> Fixes: 04a0683f7db4 ("net: stmmac: convert to phylink_generic_validate()")
> Fixes: d194923d51c9 ("net: stmmac: fill in supported_interfaces")
> Signed-off-by: Kelvin Cheung <[email protected]>
> ---
> arch/mips/loongson32/common/platform.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/arch/mips/loongson32/common/platform.c b/arch/mips/loongson32/common/platform.c
> index 794c96c2a4cd..741aace80b80 100644
> --- a/arch/mips/loongson32/common/platform.c
> +++ b/arch/mips/loongson32/common/platform.c
> @@ -147,8 +147,10 @@ static struct plat_stmmacenet_data ls1x_eth0_pdata = {
> .phy_addr = -1,
> #if defined(CONFIG_LOONGSON1_LS1B)
> .interface = PHY_INTERFACE_MODE_MII,
> + .phy_interface = PHY_INTERFACE_MODE_MII,
> #elif defined(CONFIG_LOONGSON1_LS1C)
> .interface = PHY_INTERFACE_MODE_RMII,
> + .phy_interface = PHY_INTERFACE_MODE_RMII,
> #endif
> .mdio_bus_data = &ls1x_mdio_bus_data,
> .dma_cfg = &ls1x_eth_dma_cfg,
>
> base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> --
> 2.34.1
>
Hi Serge,
Your patch also works for me.
But what about the plat_dat->interface? Is it obsolete?
Serge Semin <[email protected]> 于2022年8月22日周一 01:10写道：
>
> Hello Keguang
>
> On Thu, Aug 18, 2022 at 01:00:19PM +0800, Keguang Zhang wrote:
> > From: Kelvin Cheung <[email protected]>
> >
> > The Ethernet of LS1B/LS1C doesn't work due to the stmmac driver
> > using phylink_generic_validate() instead of stmmac_validate().
> > Moreover the driver assumes the PHY interface mode
> > passed in platform data is always supported.
> >
> > stmmaceth stmmaceth.0 eth0: validation of gmii with support 00000000,00000000,000062cf and advertisement 00000000,00000000,000062cf failed: -EINVAL
> > stmmaceth stmmaceth.0 eth0: stmmac_open: Cannot attach to PHY (error: -22)
> >
> > This patch sets phy_interface field of platform data.
>
> I've got a similar fix in my repo, though didn't have a chance to test
> it out due to lacking any loongson hardware. I've discovered the
> issues on my still going way of the STMMAC driver refactoring. Anyway
> IMO the problem is a bit different than you describe and should be
> fixed in a bit different way. Please see a patch attached to this
> email. Could you test it out on your hw? If it fixes the problem you
> can resend it as v2 patch.
>
> -Sergey
>
> >
> > Fixes: 04a0683f7db4 ("net: stmmac: convert to phylink_generic_validate()")
> > Fixes: d194923d51c9 ("net: stmmac: fill in supported_interfaces")
> > Signed-off-by: Kelvin Cheung <[email protected]>
> > ---
> > arch/mips/loongson32/common/platform.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/arch/mips/loongson32/common/platform.c b/arch/mips/loongson32/common/platform.c
> > index 794c96c2a4cd..741aace80b80 100644
> > --- a/arch/mips/loongson32/common/platform.c
> > +++ b/arch/mips/loongson32/common/platform.c
> > @@ -147,8 +147,10 @@ static struct plat_stmmacenet_data ls1x_eth0_pdata = {
> > .phy_addr = -1,
> > #if defined(CONFIG_LOONGSON1_LS1B)
> > .interface = PHY_INTERFACE_MODE_MII,
> > + .phy_interface = PHY_INTERFACE_MODE_MII,
> > #elif defined(CONFIG_LOONGSON1_LS1C)
> > .interface = PHY_INTERFACE_MODE_RMII,
> > + .phy_interface = PHY_INTERFACE_MODE_RMII,
> > #endif
> > .mdio_bus_data = &ls1x_mdio_bus_data,
> > .dma_cfg = &ls1x_eth_dma_cfg,
> >
> > base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> > --
> > 2.34.1
> >
--
Best regards,
Kelvin Cheung
On Sun, Aug 28, 2022 at 09:31:25PM +0800, Kelvin Cheung wrote:
> Hi Serge,
> Your patch also works for me.
Great. Could you submit it for review then with your tested-by tag
added?
> But what about the plat_dat->interface? Is it obsolete?
No. As I said in the patch log it's still used in the STMMAC driver
but for the MAC-PCS mode setting. Such PCS isn't available on the most
of the cases so the "phy_interface" field should be mainly utilized
instead while the "interface" field is supposed to be left untouched.
See the commit 0060c8783330 ("net: stmmac: implement support for
passive mode converters via dt") for details.
-Sergey
>
> Serge Semin <[email protected]> 于2022年8月22日周一 01:10写道：
> >
> > Hello Keguang
> >
> > On Thu, Aug 18, 2022 at 01:00:19PM +0800, Keguang Zhang wrote:
> > > From: Kelvin Cheung <[email protected]>
> > >
> > > The Ethernet of LS1B/LS1C doesn't work due to the stmmac driver
> > > using phylink_generic_validate() instead of stmmac_validate().
> > > Moreover the driver assumes the PHY interface mode
> > > passed in platform data is always supported.
> > >
> > > stmmaceth stmmaceth.0 eth0: validation of gmii with support 00000000,00000000,000062cf and advertisement 00000000,00000000,000062cf failed: -EINVAL
> > > stmmaceth stmmaceth.0 eth0: stmmac_open: Cannot attach to PHY (error: -22)
> > >
> > > This patch sets phy_interface field of platform data.
> >
> > I've got a similar fix in my repo, though didn't have a chance to test
> > it out due to lacking any loongson hardware. I've discovered the
> > issues on my still going way of the STMMAC driver refactoring. Anyway
> > IMO the problem is a bit different than you describe and should be
> > fixed in a bit different way. Please see a patch attached to this
> > email. Could you test it out on your hw? If it fixes the problem you
> > can resend it as v2 patch.
> >
> > -Sergey
> >
> > >
> > > Fixes: 04a0683f7db4 ("net: stmmac: convert to phylink_generic_validate()")
> > > Fixes: d194923d51c9 ("net: stmmac: fill in supported_interfaces")
> > > Signed-off-by: Kelvin Cheung <[email protected]>
> > > ---
> > > arch/mips/loongson32/common/platform.c | 2 ++
> > > 1 file changed, 2 insertions(+)
> > >
> > > diff --git a/arch/mips/loongson32/common/platform.c b/arch/mips/loongson32/common/platform.c
> > > index 794c96c2a4cd..741aace80b80 100644
> > > --- a/arch/mips/loongson32/common/platform.c
> > > +++ b/arch/mips/loongson32/common/platform.c
> > > @@ -147,8 +147,10 @@ static struct plat_stmmacenet_data ls1x_eth0_pdata = {
> > > .phy_addr = -1,
> > > #if defined(CONFIG_LOONGSON1_LS1B)
> > > .interface = PHY_INTERFACE_MODE_MII,
> > > + .phy_interface = PHY_INTERFACE_MODE_MII,
> > > #elif defined(CONFIG_LOONGSON1_LS1C)
> > > .interface = PHY_INTERFACE_MODE_RMII,
> > > + .phy_interface = PHY_INTERFACE_MODE_RMII,
> > > #endif
> > > .mdio_bus_data = &ls1x_mdio_bus_data,
> > > .dma_cfg = &ls1x_eth_dma_cfg,
> > >
> > > base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> > > --
> > > 2.34.1
> > >
>
>
>
> --
> Best regards,
>
> Kelvin Cheung
Hi Serge,
Serge Semin <[email protected]> 于2022年9月3日周六 06:46写道：
>
> On Sun, Aug 28, 2022 at 09:31:25PM +0800, Kelvin Cheung wrote:
> > Hi Serge,
> > Your patch also works for me.
>
> Great. Could you submit it for review then with your tested-by tag
> added?
Sure.
>
> > But what about the plat_dat->interface? Is it obsolete?
>
> No. As I said in the patch log it's still used in the STMMAC driver
> but for the MAC-PCS mode setting. Such PCS isn't available on the most
> of the cases so the "phy_interface" field should be mainly utilized
> instead while the "interface" field is supposed to be left untouched.
> See the commit 0060c8783330 ("net: stmmac: implement support for
> passive mode converters via dt") for details.
Got it. Thanks!
>
> -Sergey
>
> >
> > Serge Semin <[email protected]> 于2022年8月22日周一 01:10写道：
> > >
> > > Hello Keguang
> > >
> > > On Thu, Aug 18, 2022 at 01:00:19PM +0800, Keguang Zhang wrote:
> > > > From: Kelvin Cheung <[email protected]>
> > > >
> > > > The Ethernet of LS1B/LS1C doesn't work due to the stmmac driver
> > > > using phylink_generic_validate() instead of stmmac_validate().
> > > > Moreover the driver assumes the PHY interface mode
> > > > passed in platform data is always supported.
> > > >
> > > > stmmaceth stmmaceth.0 eth0: validation of gmii with support 00000000,00000000,000062cf and advertisement 00000000,00000000,000062cf failed: -EINVAL
> > > > stmmaceth stmmaceth.0 eth0: stmmac_open: Cannot attach to PHY (error: -22)
> > > >
> > > > This patch sets phy_interface field of platform data.
> > >
> > > I've got a similar fix in my repo, though didn't have a chance to test
> > > it out due to lacking any loongson hardware. I've discovered the
> > > issues on my still going way of the STMMAC driver refactoring. Anyway
> > > IMO the problem is a bit different than you describe and should be
> > > fixed in a bit different way. Please see a patch attached to this
> > > email. Could you test it out on your hw? If it fixes the problem you
> > > can resend it as v2 patch.
> > >
> > > -Sergey
> > >
> > > >
> > > > Fixes: 04a0683f7db4 ("net: stmmac: convert to phylink_generic_validate()")
> > > > Fixes: d194923d51c9 ("net: stmmac: fill in supported_interfaces")
> > > > Signed-off-by: Kelvin Cheung <[email protected]>
> > > > ---
> > > > arch/mips/loongson32/common/platform.c | 2 ++
> > > > 1 file changed, 2 insertions(+)
> > > >
> > > > diff --git a/arch/mips/loongson32/common/platform.c b/arch/mips/loongson32/common/platform.c
> > > > index 794c96c2a4cd..741aace80b80 100644
> > > > --- a/arch/mips/loongson32/common/platform.c
> > > > +++ b/arch/mips/loongson32/common/platform.c
> > > > @@ -147,8 +147,10 @@ static struct plat_stmmacenet_data ls1x_eth0_pdata = {
> > > > .phy_addr = -1,
> > > > #if defined(CONFIG_LOONGSON1_LS1B)
> > > > .interface = PHY_INTERFACE_MODE_MII,
> > > > + .phy_interface = PHY_INTERFACE_MODE_MII,
> > > > #elif defined(CONFIG_LOONGSON1_LS1C)
> > > > .interface = PHY_INTERFACE_MODE_RMII,
> > > > + .phy_interface = PHY_INTERFACE_MODE_RMII,
> > > > #endif
> > > > .mdio_bus_data = &ls1x_mdio_bus_data,
> > > > .dma_cfg = &ls1x_eth_dma_cfg,
> > > >
> > > > base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> > > > --
> > > > 2.34.1
> > > >
> >
> >
> >
> > --
> > Best regards,
> >
> > Kelvin Cheung
--
Best regards,
Kelvin Cheung
Hello Thomas,
Just so you know. This patch is supposed to be superseded with the next one:
https://lore.kernel.org/linux-mips/[email protected]
* or with v2 of the later one.
-Sergey
On Thu, Aug 18, 2022 at 01:00:19PM +0800, Keguang Zhang wrote:
> From: Kelvin Cheung <[email protected]>
>
> The Ethernet of LS1B/LS1C doesn't work due to the stmmac driver
> using phylink_generic_validate() instead of stmmac_validate().
> Moreover the driver assumes the PHY interface mode
> passed in platform data is always supported.
>
> stmmaceth stmmaceth.0 eth0: validation of gmii with support 00000000,00000000,000062cf and advertisement 00000000,00000000,000062cf failed: -EINVAL
> stmmaceth stmmaceth.0 eth0: stmmac_open: Cannot attach to PHY (error: -22)
>
> This patch sets phy_interface field of platform data.
>
> Fixes: 04a0683f7db4 ("net: stmmac: convert to phylink_generic_validate()")
> Fixes: d194923d51c9 ("net: stmmac: fill in supported_interfaces")
> Signed-off-by: Kelvin Cheung <[email protected]>
> ---
> arch/mips/loongson32/common/platform.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/arch/mips/loongson32/common/platform.c b/arch/mips/loongson32/common/platform.c
> index 794c96c2a4cd..741aace80b80 100644
> --- a/arch/mips/loongson32/common/platform.c
> +++ b/arch/mips/loongson32/common/platform.c
> @@ -147,8 +147,10 @@ static struct plat_stmmacenet_data ls1x_eth0_pdata = {
> .phy_addr = -1,
> #if defined(CONFIG_LOONGSON1_LS1B)
> .interface = PHY_INTERFACE_MODE_MII,
> + .phy_interface = PHY_INTERFACE_MODE_MII,
> #elif defined(CONFIG_LOONGSON1_LS1C)
> .interface = PHY_INTERFACE_MODE_RMII,
> + .phy_interface = PHY_INTERFACE_MODE_RMII,
> #endif
> .mdio_bus_data = &ls1x_mdio_bus_data,
> .dma_cfg = &ls1x_eth_dma_cfg,
>
> base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> --
> 2.34.1
>