* [BUG] network is broken on Orange Pi PC
@ 2021-05-31 21:35 Heinrich Schuchardt
2021-06-02 13:08 ` Ramon Fried
0 siblings, 1 reply; 11+ messages in thread
From: Heinrich Schuchardt @ 2021-05-31 21:35 UTC (permalink / raw)
To: Joe Hershberger, Ramon Fried, Jagan Teki, Andre Przywara
Cc: U-Boot Mailing List
Dear all,
network is broken in U-Boot on orangepi_pc_defconfig:
U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
Allwinner Technology
eth0: ethernet@1c30000
=> dhcp
sun8i_emac_eth_start: Timeout
Best regards
Heinrich
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-05-31 21:35 [BUG] network is broken on Orange Pi PC Heinrich Schuchardt
@ 2021-06-02 13:08 ` Ramon Fried
2021-06-03 7:46 ` Heinrich Schuchardt
0 siblings, 1 reply; 11+ messages in thread
From: Ramon Fried @ 2021-06-02 13:08 UTC (permalink / raw)
To: Heinrich Schuchardt
Cc: Joe Hershberger, Jagan Teki, Andre Przywara, U-Boot Mailing List
On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>
> Dear all,
>
> network is broken in U-Boot on orangepi_pc_defconfig:
>
> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
> Allwinner Technology
> eth0: ethernet@1c30000
> => dhcp
> sun8i_emac_eth_start: Timeout
>
> Best regards
>
> Heinrich
>
Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
Thanks.
Ramon.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-02 13:08 ` Ramon Fried
@ 2021-06-03 7:46 ` Heinrich Schuchardt
2021-06-03 9:04 ` Andre Przywara
0 siblings, 1 reply; 11+ messages in thread
From: Heinrich Schuchardt @ 2021-06-03 7:46 UTC (permalink / raw)
To: Ramon Fried
Cc: Joe Hershberger, Jagan Teki, Andre Przywara, U-Boot Mailing List
On 6/2/21 3:08 PM, Ramon Fried wrote:
> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>
>> Dear all,
>>
>> network is broken in U-Boot on orangepi_pc_defconfig:
>>
>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
>> Allwinner Technology
>> eth0: ethernet@1c30000
>> => dhcp
>> sun8i_emac_eth_start: Timeout
>>
>> Best regards
>>
>> Heinrich
>>
> Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
> Thanks.
> Ramon.
>
Git bisect points to:
commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
Author: Andre Przywara <andre.przywara@arm.com>
Date: Mon Jul 6 01:40:45 2020 +0100
net: sun8i-emac: Lower MDIO frequency
Reverting the patch solves the problem for the OrangePi PC.
According to the commit message the change was only needed for needed
for external PHYs. Can't we let the change depend on priv->use_internal_phy?
diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
index 5a1b38bf80..d7553fe163 100644
--- a/drivers/net/sun8i_emac.c
+++ b/drivers/net/sun8i_emac.c
@@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
addr, int devad, int reg)
* The EMAC clock is either 200 or 300 MHz, so we need a divider
* of 128 to get the MDIO frequency below the required 2.5 MHz.
*/
- mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
MDIO_CMD_MII_CLK_CSR_SHIFT;
+ if (!priv->use_internal_phy)
+ mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
+ MDIO_CMD_MII_CLK_CSR_SHIFT;
mii_cmd |= MDIO_CMD_MII_BUSY;
@@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
addr, int devad, int reg,
* The EMAC clock is either 200 or 300 MHz, so we need a divider
* of 128 to get the MDIO frequency below the required 2.5 MHz.
*/
- mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
MDIO_CMD_MII_CLK_CSR_SHIFT;
+ if (!priv->use_internal_phy)
+ mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
+ MDIO_CMD_MII_CLK_CSR_SHIFT;
mii_cmd |= MDIO_CMD_MII_WRITE;
mii_cmd |= MDIO_CMD_MII_BUSY;
Best regards
Heinrich
I would assume the problem hits all H3 boards.
Best regards
Heinrich
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 7:46 ` Heinrich Schuchardt
@ 2021-06-03 9:04 ` Andre Przywara
2021-06-03 10:20 ` Heinrich Schuchardt
0 siblings, 1 reply; 11+ messages in thread
From: Andre Przywara @ 2021-06-03 9:04 UTC (permalink / raw)
To: Heinrich Schuchardt
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List,
Andreas Rehn
On Thu, 3 Jun 2021 09:46:48 +0200
Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
Hi Heinrich,
> On 6/2/21 3:08 PM, Ramon Fried wrote:
> > On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> >>
> >> Dear all,
> >>
> >> network is broken in U-Boot on orangepi_pc_defconfig:
Thanks for the report!
> >>
> >> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
> >> Allwinner Technology
> >> eth0: ethernet@1c30000
> >> => dhcp
> >> sun8i_emac_eth_start: Timeout
> >>
> >> Best regards
> >>
> >> Heinrich
> >>
> > Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
> > Thanks.
> > Ramon.
> >
>
> Git bisect points to:
>
> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
> Author: Andre Przywara <andre.przywara@arm.com>
> Date: Mon Jul 6 01:40:45 2020 +0100
>
> net: sun8i-emac: Lower MDIO frequency
>
> Reverting the patch solves the problem for the OrangePi PC.
>
> According to the commit message the change was only needed for needed
> for external PHYs.
The external PHY on the Pine64 (non-plus) was the trigger, however
both the manual and the Linux driver point to that we definitely need a
higher divider. From what I can see, AHB2 (the EMAC clock) runs at 200
MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in 12.5
MHz MDIO frequency. Can you check whether any other divider values fix
this for you as well?
> Can't we let the change depend on priv->use_internal_phy?
>
> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
> index 5a1b38bf80..d7553fe163 100644
> --- a/drivers/net/sun8i_emac.c
> +++ b/drivers/net/sun8i_emac.c
> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
> addr, int devad, int reg)
> * The EMAC clock is either 200 or 300 MHz, so we need a divider
> * of 128 to get the MDIO frequency below the required 2.5 MHz.
> */
> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> MDIO_CMD_MII_CLK_CSR_SHIFT;
> + if (!priv->use_internal_phy)
> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>
> mii_cmd |= MDIO_CMD_MII_BUSY;
>
> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
> addr, int devad, int reg,
> * The EMAC clock is either 200 or 300 MHz, so we need a divider
> * of 128 to get the MDIO frequency below the required 2.5 MHz.
> */
> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> MDIO_CMD_MII_CLK_CSR_SHIFT;
> + if (!priv->use_internal_phy)
> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>
> mii_cmd |= MDIO_CMD_MII_WRITE;
> mii_cmd |= MDIO_CMD_MII_BUSY;
>
> Best regards
>
> Heinrich
>
> I would assume the problem hits all H3 boards.
And that's the confusing part: it does not. I tested this on my
OrangePi Zero (H2+ with internal PHY), both back then with the original
MDIO frequency patch and also now after your report. It always worked
reliably for me.
Also: I am still puzzled how one influences the other: The error you
get is from the *MAC* soft reset: I would think this is an independent
operation from any communication attempts with the PHY.
There is this thread here about the same symptom on a V3s:
https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
v2 of this original patch (in the same thread) proposes some other
solution, which I am also not very happy with. But just to get some more
data points: can you check whether skipping the soft reset fixes this
for you? I will have a closer look tonight to check the order of soft
reset and PHY communication. Maybe we should only do the soft reset
*once* when the MAC probes, and not on every .start call?
Thanks,
Andre
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 9:04 ` Andre Przywara
@ 2021-06-03 10:20 ` Heinrich Schuchardt
2021-06-03 11:04 ` Andre Przywara
0 siblings, 1 reply; 11+ messages in thread
From: Heinrich Schuchardt @ 2021-06-03 10:20 UTC (permalink / raw)
To: Andre Przywara
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List,
Andreas Rehn
On 6/3/21 11:04 AM, Andre Przywara wrote:
> On Thu, 3 Jun 2021 09:46:48 +0200
> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>
> Hi Heinrich,
>
>> On 6/2/21 3:08 PM, Ramon Fried wrote:
>>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>>>
>>>> Dear all,
>>>>
>>>> network is broken in U-Boot on orangepi_pc_defconfig:
>
> Thanks for the report!
>
>>>>
>>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
>>>> Allwinner Technology
>>>> eth0: ethernet@1c30000
>>>> => dhcp
>>>> sun8i_emac_eth_start: Timeout
>>>>
>>>> Best regards
>>>>
>>>> Heinrich
>>>>
>>> Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
>>> Thanks.
>>> Ramon.
>>>
>>
>> Git bisect points to:
>>
>> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
>> Author: Andre Przywara <andre.przywara@arm.com>
>> Date: Mon Jul 6 01:40:45 2020 +0100
>>
>> net: sun8i-emac: Lower MDIO frequency
>>
>> Reverting the patch solves the problem for the OrangePi PC.
>>
>> According to the commit message the change was only needed for needed
>> for external PHYs.
>
> The external PHY on the Pine64 (non-plus) was the trigger, however
> both the manual and the Linux driver point to that we definitely need a
> higher divider. From what I can see, AHB2 (the EMAC clock) runs at 200
> MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in 12.5
> MHz MDIO frequency. Can you check whether any other divider values fix
> this for you as well?
MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was used
before your patch.
MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
Waiting for PHY auto negotiation to complete. done
MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
What is wrong about the approach in
[PATCH 1/1] net: sun8i-emac: fix MDIO frequency
https://lists.denx.de/pipermail/u-boot/2021-June/451305.html ?
The Linux 5.10 driver runs fine no matter what value we choose for the
divider in U-Boot. But I see a message:
dwmac-sun8i 1c30000.ethernet:
Current syscon value is not the default 148000 (expect 58000)
@default_syscon_value:
The default value of the EMAC register in syscon
This value is used for disabling properly EMAC
and used as a good starting value in case of the
boot process(uboot) leave some stuff.
Best regards
Heinrich
>
>> Can't we let the change depend on priv->use_internal_phy?
>>
>> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
>> index 5a1b38bf80..d7553fe163 100644
>> --- a/drivers/net/sun8i_emac.c
>> +++ b/drivers/net/sun8i_emac.c
>> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
>> addr, int devad, int reg)
>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
>> */
>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>> MDIO_CMD_MII_CLK_CSR_SHIFT;
>> + if (!priv->use_internal_phy)
>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>>
>> mii_cmd |= MDIO_CMD_MII_BUSY;
>>
>> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
>> addr, int devad, int reg,
>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
>> */
>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>> MDIO_CMD_MII_CLK_CSR_SHIFT;
>> + if (!priv->use_internal_phy)
>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>>
>> mii_cmd |= MDIO_CMD_MII_WRITE;
>> mii_cmd |= MDIO_CMD_MII_BUSY;
>>
>> Best regards
>>
>> Heinrich
>>
>> I would assume the problem hits all H3 boards.
>
> And that's the confusing part: it does not. I tested this on my
> OrangePi Zero (H2+ with internal PHY), both back then with the original
> MDIO frequency patch and also now after your report. It always worked
> reliably for me.
> Also: I am still puzzled how one influences the other: The error you
> get is from the *MAC* soft reset: I would think this is an independent
> operation from any communication attempts with the PHY.
>
> There is this thread here about the same symptom on a V3s:
> https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
> v2 of this original patch (in the same thread) proposes some other
> solution, which I am also not very happy with. But just to get some more
> data points: can you check whether skipping the soft reset fixes this
> for you? I will have a closer look tonight to check the order of soft
> reset and PHY communication. Maybe we should only do the soft reset
> *once* when the MAC probes, and not on every .start call?
>
> Thanks,
> Andre
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 10:20 ` Heinrich Schuchardt
@ 2021-06-03 11:04 ` Andre Przywara
2021-06-03 16:09 ` Heinrich Schuchardt
0 siblings, 1 reply; 11+ messages in thread
From: Andre Przywara @ 2021-06-03 11:04 UTC (permalink / raw)
To: Heinrich Schuchardt
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List,
Andreas Rehn
On Thu, 3 Jun 2021 12:20:34 +0200
Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
Hi,
> On 6/3/21 11:04 AM, Andre Przywara wrote:
> > On Thu, 3 Jun 2021 09:46:48 +0200
> > Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> >
> > Hi Heinrich,
> >
> >> On 6/2/21 3:08 PM, Ramon Fried wrote:
> >>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> >>>>
> >>>> Dear all,
> >>>>
> >>>> network is broken in U-Boot on orangepi_pc_defconfig:
> >
> > Thanks for the report!
> >
> >>>>
> >>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
> >>>> Allwinner Technology
> >>>> eth0: ethernet@1c30000
> >>>> => dhcp
> >>>> sun8i_emac_eth_start: Timeout
> >>>>
> >>>> Best regards
> >>>>
> >>>> Heinrich
> >>>>
> >>> Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
> >>> Thanks.
> >>> Ramon.
> >>>
> >>
> >> Git bisect points to:
> >>
> >> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
> >> Author: Andre Przywara <andre.przywara@arm.com>
> >> Date: Mon Jul 6 01:40:45 2020 +0100
> >>
> >> net: sun8i-emac: Lower MDIO frequency
> >>
> >> Reverting the patch solves the problem for the OrangePi PC.
> >>
> >> According to the commit message the change was only needed for needed
> >> for external PHYs.
> >
> > The external PHY on the Pine64 (non-plus) was the trigger, however
> > both the manual and the Linux driver point to that we definitely need a
> > higher divider. From what I can see, AHB2 (the EMAC clock) runs at 200
> > MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in 12.5
> > MHz MDIO frequency. Can you check whether any other divider values fix
> > this for you as well?
>
> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was used
> before your patch.
>
> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
> Waiting for PHY auto negotiation to complete. done
> MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
> MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
> MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
>
> What is wrong about the approach in
>
> [PATCH 1/1] net: sun8i-emac: fix MDIO frequency
> https://lists.denx.de/pipermail/u-boot/2021-June/451305.html ?
The MDIO spec says that the maximum frequency on the MDIO bus should be
2.5 MHz. On the H3 with 200 MHz AHB and with a divider of 16 it's 12.5
MHz. Apparently the internal H3 PHY can tolerate those high
frequencies, but it's beyond the spec. And certainly it should not fail
with lower frequencies. And indeed it works for me on the H2+ (which I
understand is just a binned version of the H3).
> The Linux 5.10 driver runs fine no matter what value we choose for the
> divider in U-Boot.
Because it programs the divider correctly:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316
I take it you have no problems with Linux, so the divider being too
high cannot be the root cause of the problem.
So I'd rather fix this problem properly and correctly, especially as it
seems to be more related to the EMAC soft reset (where your error
message comes from).
> But I see a message:
>
> dwmac-sun8i 1c30000.ethernet:
> Current syscon value is not the default 148000 (expect 58000)
>
> @default_syscon_value:
> The default value of the EMAC register in syscon
> This value is used for disabling properly EMAC
> and used as a good starting value in case of the
> boot process(uboot) leave some stuff.
Yeah, I never got the reason for this message. If Linux wants a
certain value, it should program that. Whatever was left by someone
else (be it U-Boot or some other software/OS) should not matter.
Cheers,
Andre
>
> Best regards
>
> Heinrich
>
> >
> >> Can't we let the change depend on priv->use_internal_phy?
> >>
> >> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
> >> index 5a1b38bf80..d7553fe163 100644
> >> --- a/drivers/net/sun8i_emac.c
> >> +++ b/drivers/net/sun8i_emac.c
> >> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
> >> addr, int devad, int reg)
> >> * The EMAC clock is either 200 or 300 MHz, so we need a divider
> >> * of 128 to get the MDIO frequency below the required 2.5 MHz.
> >> */
> >> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >> MDIO_CMD_MII_CLK_CSR_SHIFT;
> >> + if (!priv->use_internal_phy)
> >> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> >>
> >> mii_cmd |= MDIO_CMD_MII_BUSY;
> >>
> >> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
> >> addr, int devad, int reg,
> >> * The EMAC clock is either 200 or 300 MHz, so we need a divider
> >> * of 128 to get the MDIO frequency below the required 2.5 MHz.
> >> */
> >> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >> MDIO_CMD_MII_CLK_CSR_SHIFT;
> >> + if (!priv->use_internal_phy)
> >> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> >>
> >> mii_cmd |= MDIO_CMD_MII_WRITE;
> >> mii_cmd |= MDIO_CMD_MII_BUSY;
> >>
> >> Best regards
> >>
> >> Heinrich
> >>
> >> I would assume the problem hits all H3 boards.
> >
> > And that's the confusing part: it does not. I tested this on my
> > OrangePi Zero (H2+ with internal PHY), both back then with the original
> > MDIO frequency patch and also now after your report. It always worked
> > reliably for me.
> > Also: I am still puzzled how one influences the other: The error you
> > get is from the *MAC* soft reset: I would think this is an independent
> > operation from any communication attempts with the PHY.
> >
> > There is this thread here about the same symptom on a V3s:
> > https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
> > v2 of this original patch (in the same thread) proposes some other
> > solution, which I am also not very happy with. But just to get some more
> > data points: can you check whether skipping the soft reset fixes this
> > for you? I will have a closer look tonight to check the order of soft
> > reset and PHY communication. Maybe we should only do the soft reset
> > *once* when the MAC probes, and not on every .start call?
> >
> > Thanks,
> > Andre
> >
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 11:04 ` Andre Przywara
@ 2021-06-03 16:09 ` Heinrich Schuchardt
2021-06-03 16:57 ` Andre Przywara
0 siblings, 1 reply; 11+ messages in thread
From: Heinrich Schuchardt @ 2021-06-03 16:09 UTC (permalink / raw)
To: Andre Przywara
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List,
Andreas Rehn
On 6/3/21 1:04 PM, Andre Przywara wrote:
> On Thu, 3 Jun 2021 12:20:34 +0200
> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>
> Hi,
>
>> On 6/3/21 11:04 AM, Andre Przywara wrote:
>>> On Thu, 3 Jun 2021 09:46:48 +0200
>>> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>>
>>> Hi Heinrich,
>>>
>>>> On 6/2/21 3:08 PM, Ramon Fried wrote:
>>>>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>>>>>
>>>>>> Dear all,
>>>>>>
>>>>>> network is broken in U-Boot on orangepi_pc_defconfig:
>>>
>>> Thanks for the report!
>>>
>>>>>>
>>>>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
>>>>>> Allwinner Technology
>>>>>> eth0: ethernet@1c30000
>>>>>> => dhcp
>>>>>> sun8i_emac_eth_start: Timeout
>>>>>>
>>>>>> Best regards
>>>>>>
>>>>>> Heinrich
>>>>>>
>>>>> Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
>>>>> Thanks.
>>>>> Ramon.
>>>>>
>>>>
>>>> Git bisect points to:
>>>>
>>>> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
>>>> Author: Andre Przywara <andre.przywara@arm.com>
>>>> Date: Mon Jul 6 01:40:45 2020 +0100
>>>>
>>>> net: sun8i-emac: Lower MDIO frequency
>>>>
>>>> Reverting the patch solves the problem for the OrangePi PC.
>>>>
>>>> According to the commit message the change was only needed for needed
>>>> for external PHYs.
>>>
>>> The external PHY on the Pine64 (non-plus) was the trigger, however
>>> both the manual and the Linux driver point to that we definitely need a
>>> higher divider. From what I can see, AHB2 (the EMAC clock) runs at 200
>>> MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in 12.5
>>> MHz MDIO frequency. Can you check whether any other divider values fix
>>> this for you as well?
>>
>> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was used
>> before your patch.
>>
>> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
>> Waiting for PHY auto negotiation to complete. done
>> MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
>> MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
>> MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
>>
>> What is wrong about the approach in
>>
>> [PATCH 1/1] net: sun8i-emac: fix MDIO frequency
>> https://lists.denx.de/pipermail/u-boot/2021-June/451305.html ?
>
> The MDIO spec says that the maximum frequency on the MDIO bus should be
> 2.5 MHz. On the H3 with 200 MHz AHB and with a divider of 16 it's 12.5
The dts says for the MDIO:
int_mii_phy: ethernet-phy@1 {
clocks = <&ccu CLK_BUS_EPHY>;
"ahb" refers to CLK_BUS_SPI1:
spi1: spi@1c69000 {
clocks = <&ccu CLK_BUS_SPI1>, <&ccu CLK_SPI1>;
clock-names = "ahb", "mod";
Why are you mentioning AHB?
> MHz. Apparently the internal H3 PHY can tolerate those high
> frequencies, but it's beyond the spec. And certainly it should not fail
> with lower frequencies. And indeed it works for me on the H2+ (which I
> understand is just a binned version of the H3).
>
>> The Linux 5.10 driver runs fine no matter what value we choose for the
>> divider in U-Boot.
>
> Because it programs the divider correctly:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316
> I take it you have no problems with Linux, so the divider being too
> high cannot be the root cause of the problem.
Function stmmac_clk_csr_set() depends on the rate of the clock named
"stmmaceth":
ethernet@1c30000 {
clocks = <&ccu CLK_BUS_EMAC>;
clock-names = "stmmaceth";
How do we get the frequency of this clock?
Best regards
Heinrich
>
> So I'd rather fix this problem properly and correctly, especially as it
> seems to be more related to the EMAC soft reset (where your error
> message comes from).
>
>> But I see a message:
>>
>> dwmac-sun8i 1c30000.ethernet:
>> Current syscon value is not the default 148000 (expect 58000)
>>
>> @default_syscon_value:
>> The default value of the EMAC register in syscon
>> This value is used for disabling properly EMAC
>> and used as a good starting value in case of the
>> boot process(uboot) leave some stuff.
>
> Yeah, I never got the reason for this message. If Linux wants a
> certain value, it should program that. Whatever was left by someone
> else (be it U-Boot or some other software/OS) should not matter.
>
> Cheers,
> Andre
>
>>
>> Best regards
>>
>> Heinrich
>>
>>>
>>>> Can't we let the change depend on priv->use_internal_phy?
>>>>
>>>> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
>>>> index 5a1b38bf80..d7553fe163 100644
>>>> --- a/drivers/net/sun8i_emac.c
>>>> +++ b/drivers/net/sun8i_emac.c
>>>> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
>>>> addr, int devad, int reg)
>>>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
>>>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
>>>> */
>>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>> + if (!priv->use_internal_phy)
>>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>>
>>>> mii_cmd |= MDIO_CMD_MII_BUSY;
>>>>
>>>> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
>>>> addr, int devad, int reg,
>>>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
>>>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
>>>> */
>>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>> + if (!priv->use_internal_phy)
>>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>>
>>>> mii_cmd |= MDIO_CMD_MII_WRITE;
>>>> mii_cmd |= MDIO_CMD_MII_BUSY;
>>>>
>>>> Best regards
>>>>
>>>> Heinrich
>>>>
>>>> I would assume the problem hits all H3 boards.
>>>
>>> And that's the confusing part: it does not. I tested this on my
>>> OrangePi Zero (H2+ with internal PHY), both back then with the original
>>> MDIO frequency patch and also now after your report. It always worked
>>> reliably for me.
>>> Also: I am still puzzled how one influences the other: The error you
>>> get is from the *MAC* soft reset: I would think this is an independent
>>> operation from any communication attempts with the PHY.
>>>
>>> There is this thread here about the same symptom on a V3s:
>>> https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
>>> v2 of this original patch (in the same thread) proposes some other
>>> solution, which I am also not very happy with. But just to get some more
>>> data points: can you check whether skipping the soft reset fixes this
>>> for you? I will have a closer look tonight to check the order of soft
>>> reset and PHY communication. Maybe we should only do the soft reset
>>> *once* when the MAC probes, and not on every .start call?
>>>
>>> Thanks,
>>> Andre
>>>
>>
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 16:09 ` Heinrich Schuchardt
@ 2021-06-03 16:57 ` Andre Przywara
2021-06-03 18:00 ` Andreas Rehn
2021-06-03 20:09 ` Heinrich Schuchardt
0 siblings, 2 replies; 11+ messages in thread
From: Andre Przywara @ 2021-06-03 16:57 UTC (permalink / raw)
To: Heinrich Schuchardt
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List,
Andreas Rehn
On Thu, 3 Jun 2021 18:09:25 +0200
Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
Hi,
> On 6/3/21 1:04 PM, Andre Przywara wrote:
> > On Thu, 3 Jun 2021 12:20:34 +0200
> > Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> >
> > Hi,
> >
> >> On 6/3/21 11:04 AM, Andre Przywara wrote:
> >>> On Thu, 3 Jun 2021 09:46:48 +0200
> >>> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> >>>
> >>> Hi Heinrich,
> >>>
> >>>> On 6/2/21 3:08 PM, Ramon Fried wrote:
> >>>>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> >>>>>>
> >>>>>> Dear all,
> >>>>>>
> >>>>>> network is broken in U-Boot on orangepi_pc_defconfig:
> >>>
> >>> Thanks for the report!
> >>>
> >>>>>>
> >>>>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
> >>>>>> Allwinner Technology
> >>>>>> eth0: ethernet@1c30000
> >>>>>> => dhcp
> >>>>>> sun8i_emac_eth_start: Timeout
> >>>>>>
> >>>>>> Best regards
> >>>>>>
> >>>>>> Heinrich
> >>>>>>
> >>>>> Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
> >>>>> Thanks.
> >>>>> Ramon.
> >>>>>
> >>>>
> >>>> Git bisect points to:
> >>>>
> >>>> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
> >>>> Author: Andre Przywara <andre.przywara@arm.com>
> >>>> Date: Mon Jul 6 01:40:45 2020 +0100
> >>>>
> >>>> net: sun8i-emac: Lower MDIO frequency
> >>>>
> >>>> Reverting the patch solves the problem for the OrangePi PC.
> >>>>
> >>>> According to the commit message the change was only needed for needed
> >>>> for external PHYs.
> >>>
> >>> The external PHY on the Pine64 (non-plus) was the trigger, however
> >>> both the manual and the Linux driver point to that we definitely need a
> >>> higher divider. From what I can see, AHB2 (the EMAC clock) runs at 200
> >>> MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in 12.5
> >>> MHz MDIO frequency. Can you check whether any other divider values fix
> >>> this for you as well?
> >>
> >> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was used
> >> before your patch.
> >>
> >> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
> >> Waiting for PHY auto negotiation to complete. done
> >> MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
> >> MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
> >> MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
> >>
> >> What is wrong about the approach in
> >>
> >> [PATCH 1/1] net: sun8i-emac: fix MDIO frequency
> >> https://lists.denx.de/pipermail/u-boot/2021-June/451305.html ?
> >
> > The MDIO spec says that the maximum frequency on the MDIO bus should be
> > 2.5 MHz. On the H3 with 200 MHz AHB and with a divider of 16 it's 12.5
>
> The dts says for the MDIO:
>
> int_mii_phy: ethernet-phy@1 {
> clocks = <&ccu CLK_BUS_EPHY>;
I think that is the clock to drive the PHY logic, not the MDIO bus.
It's a gate for AHB1, in this case, which seems to be 200 MHz as well
(because rate(AHB2)=rate(AHB1)/1).
> "ahb" refers to CLK_BUS_SPI1:
>
> spi1: spi@1c69000 {
> clocks = <&ccu CLK_BUS_SPI1>, <&ccu CLK_SPI1>;
> clock-names = "ahb", "mod";
>
> Why are you mentioning AHB?
AHB stands for "Advanced High-performance Bus", which is a traditional
term for a (faster, data centric) bus interface in the ARM
bus world. I am not sure the exact term is actually correct anymore
with recent bus implementations, but the H3 manual refers to AHB
everywhere, in particular as the bus name for the EMAC IP (figure 4-1 System
Bus Tree). Apparently the bus clock is used as the clock source for the
MDIO bus as well - after dividing it.
> > MHz. Apparently the internal H3 PHY can tolerate those high
> > frequencies, but it's beyond the spec. And certainly it should not fail
> > with lower frequencies. And indeed it works for me on the H2+ (which I
> > understand is just a binned version of the H3).
> >
> >> The Linux 5.10 driver runs fine no matter what value we choose for the
> >> divider in U-Boot.
> >
> > Because it programs the divider correctly:
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316
> > I take it you have no problems with Linux, so the divider being too
> > high cannot be the root cause of the problem.
>
> Function stmmac_clk_csr_set() depends on the rate of the clock named
> "stmmaceth":
>
> ethernet@1c30000 {
> clocks = <&ccu CLK_BUS_EMAC>;
> clock-names = "stmmaceth";
>
> How do we get the frequency of this clock?
Have you checked /sys/kernel/debug/clk/clk_summary? You could also
insert some printk to be sure. Otherwise the definition of bus_emac_clk
and CLK_BUS_EMAC in drivers/clk/sunxi-ng/ccu-sun8i-h3.c tell you that
it's a gate for ahb2. All of those point to 200 MHz on my OPi Zero.
Hope that helps!
Cheers,
Andre
>
> Best regards
>
> Heinrich
>
> >
> > So I'd rather fix this problem properly and correctly, especially as it
> > seems to be more related to the EMAC soft reset (where your error
> > message comes from).
> >
> >> But I see a message:
> >>
> >> dwmac-sun8i 1c30000.ethernet:
> >> Current syscon value is not the default 148000 (expect 58000)
> >>
> >> @default_syscon_value:
> >> The default value of the EMAC register in syscon
> >> This value is used for disabling properly EMAC
> >> and used as a good starting value in case of the
> >> boot process(uboot) leave some stuff.
> >
> > Yeah, I never got the reason for this message. If Linux wants a
> > certain value, it should program that. Whatever was left by someone
> > else (be it U-Boot or some other software/OS) should not matter.
> >
> > Cheers,
> > Andre
> >
> >>
> >> Best regards
> >>
> >> Heinrich
> >>
> >>>
> >>>> Can't we let the change depend on priv->use_internal_phy?
> >>>>
> >>>> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
> >>>> index 5a1b38bf80..d7553fe163 100644
> >>>> --- a/drivers/net/sun8i_emac.c
> >>>> +++ b/drivers/net/sun8i_emac.c
> >>>> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
> >>>> addr, int devad, int reg)
> >>>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
> >>>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
> >>>> */
> >>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
> >>>> + if (!priv->use_internal_phy)
> >>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> >>>>
> >>>> mii_cmd |= MDIO_CMD_MII_BUSY;
> >>>>
> >>>> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
> >>>> addr, int devad, int reg,
> >>>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
> >>>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
> >>>> */
> >>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
> >>>> + if (!priv->use_internal_phy)
> >>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> >>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> >>>>
> >>>> mii_cmd |= MDIO_CMD_MII_WRITE;
> >>>> mii_cmd |= MDIO_CMD_MII_BUSY;
> >>>>
> >>>> Best regards
> >>>>
> >>>> Heinrich
> >>>>
> >>>> I would assume the problem hits all H3 boards.
> >>>
> >>> And that's the confusing part: it does not. I tested this on my
> >>> OrangePi Zero (H2+ with internal PHY), both back then with the original
> >>> MDIO frequency patch and also now after your report. It always worked
> >>> reliably for me.
> >>> Also: I am still puzzled how one influences the other: The error you
> >>> get is from the *MAC* soft reset: I would think this is an independent
> >>> operation from any communication attempts with the PHY.
> >>>
> >>> There is this thread here about the same symptom on a V3s:
> >>> https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
> >>> v2 of this original patch (in the same thread) proposes some other
> >>> solution, which I am also not very happy with. But just to get some more
> >>> data points: can you check whether skipping the soft reset fixes this
> >>> for you? I will have a closer look tonight to check the order of soft
> >>> reset and PHY communication. Maybe we should only do the soft reset
> >>> *once* when the MAC probes, and not on every .start call?
> >>>
> >>> Thanks,
> >>> Andre
> >>>
> >>
> >
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 16:57 ` Andre Przywara
@ 2021-06-03 18:00 ` Andreas Rehn
2021-06-03 20:19 ` Heinrich Schuchardt
2021-06-03 20:09 ` Heinrich Schuchardt
1 sibling, 1 reply; 11+ messages in thread
From: Andreas Rehn @ 2021-06-03 18:00 UTC (permalink / raw)
To: Andre Przywara
Cc: Heinrich Schuchardt, Ramon Fried, Joe Hershberger, Jagan Teki,
U-Boot Mailing List
Am Do., 3. Juni 2021 um 18:58 Uhr schrieb Andre Przywara <
andre.przywara@arm.com>:
> On Thu, 3 Jun 2021 18:09:25 +0200
> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>
> Hi,
>
> > On 6/3/21 1:04 PM, Andre Przywara wrote:
> > > On Thu, 3 Jun 2021 12:20:34 +0200
> > > Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> > >
> > > Hi,
> > >
> > >> On 6/3/21 11:04 AM, Andre Przywara wrote:
> > >>> On Thu, 3 Jun 2021 09:46:48 +0200
> > >>> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
> > >>>
> > >>> Hi Heinrich,
> > >>>
> > >>>> On 6/2/21 3:08 PM, Ramon Fried wrote:
> > >>>>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <
> xypron.glpk@gmx.de> wrote:
> > >>>>>>
> > >>>>>> Dear all,
> > >>>>>>
> > >>>>>> network is broken in U-Boot on orangepi_pc_defconfig:
> > >>>
> > >>> Thanks for the report!
> > >>>
> > >>>>>>
> > >>>>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56
> +0000)
> > >>>>>> Allwinner Technology
> > >>>>>> eth0: ethernet@1c30000
> > >>>>>> => dhcp
> > >>>>>> sun8i_emac_eth_start: Timeout
> > >>>>>>
> > >>>>>> Best regards
> > >>>>>>
> > >>>>>> Heinrich
> > >>>>>>
> > >>>>> Hi Heinrich, I don't have OrangePi. can you bisect and tell me
> when it broke ?
> > >>>>> Thanks.
> > >>>>> Ramon.
> > >>>>>
> > >>>>
> > >>>> Git bisect points to:
> > >>>>
> > >>>> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
> > >>>> Author: Andre Przywara <andre.przywara@arm.com>
> > >>>> Date: Mon Jul 6 01:40:45 2020 +0100
> > >>>>
> > >>>> net: sun8i-emac: Lower MDIO frequency
> > >>>>
> > >>>> Reverting the patch solves the problem for the OrangePi PC.
> > >>>>
> > >>>> According to the commit message the change was only needed for
> needed
> > >>>> for external PHYs.
> > >>>
> > >>> The external PHY on the Pine64 (non-plus) was the trigger, however
> > >>> both the manual and the Linux driver point to that we definitely
> need a
> > >>> higher divider. From what I can see, AHB2 (the EMAC clock) runs at
> 200
> > >>> MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in
> 12.5
> > >>> MHz MDIO frequency. Can you check whether any other divider values
> fix
> > >>> this for you as well?
> > >>
> > >> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was used
> > >> before your patch.
> > >>
> > >> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
> > >> Waiting for PHY auto negotiation to complete. done
> > >> MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
> > >> MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
> > >> MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
> > >>
> > >> What is wrong about the approach in
> > >>
> > >> [PATCH 1/1] net: sun8i-emac: fix MDIO frequency
> > >> https://lists.denx.de/pipermail/u-boot/2021-June/451305.html ?
> > >
> > > The MDIO spec says that the maximum frequency on the MDIO bus should be
> > > 2.5 MHz. On the H3 with 200 MHz AHB and with a divider of 16 it's
> 12.5
> >
> > The dts says for the MDIO:
> >
> > int_mii_phy: ethernet-phy@1 {
> > clocks = <&ccu CLK_BUS_EPHY>;
>
> I think that is the clock to drive the PHY logic, not the MDIO bus.
> It's a gate for AHB1, in this case, which seems to be 200 MHz as well
> (because rate(AHB2)=rate(AHB1)/1).
>
> > "ahb" refers to CLK_BUS_SPI1:
> >
> > spi1: spi@1c69000 {
> > clocks = <&ccu CLK_BUS_SPI1>, <&ccu CLK_SPI1>;
> > clock-names = "ahb", "mod";
> >
> > Why are you mentioning AHB?
>
> AHB stands for "Advanced High-performance Bus", which is a traditional
> term for a (faster, data centric) bus interface in the ARM
> bus world. I am not sure the exact term is actually correct anymore
> with recent bus implementations, but the H3 manual refers to AHB
> everywhere, in particular as the bus name for the EMAC IP (figure 4-1
> System
> Bus Tree). Apparently the bus clock is used as the clock source for the
> MDIO bus as well - after dividing it.
>
> > > MHz. Apparently the internal H3 PHY can tolerate those high
> > > frequencies, but it's beyond the spec. And certainly it should not fail
> > > with lower frequencies. And indeed it works for me on the H2+ (which I
> > > understand is just a binned version of the H3).
> > >
> > >> The Linux 5.10 driver runs fine no matter what value we choose for the
> > >> divider in U-Boot.
> > >
> > > Because it programs the divider correctly:
> > >
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316
> > > I take it you have no problems with Linux, so the divider being too
> > > high cannot be the root cause of the problem.
> >
> > Function stmmac_clk_csr_set() depends on the rate of the clock named
> > "stmmaceth":
> >
> > ethernet@1c30000 {
> > clocks = <&ccu CLK_BUS_EMAC>;
> > clock-names = "stmmaceth";
> >
> > How do we get the frequency of this clock?
>
> Have you checked /sys/kernel/debug/clk/clk_summary? You could also
> insert some printk to be sure. Otherwise the definition of bus_emac_clk
> and CLK_BUS_EMAC in drivers/clk/sunxi-ng/ccu-sun8i-h3.c tell you that
> it's a gate for ahb2. All of those point to 200 MHz on my OPi Zero.
>
> Hope that helps!
>
> Cheers,
> Andre
> >
> > Best regards
> >
> > Heinrich
> >
> > >
> > > So I'd rather fix this problem properly and correctly, especially as it
> > > seems to be more related to the EMAC soft reset (where your error
> > > message comes from).
> > >
> > >> But I see a message:
> > >>
> > >> dwmac-sun8i 1c30000.ethernet:
> > >> Current syscon value is not the default 148000 (expect 58000)
> > >>
> > >> @default_syscon_value:
> > >> The default value of the EMAC register in syscon
> > >> This value is used for disabling properly EMAC
> > >> and used as a good starting value in case of the
> > >> boot process(uboot) leave some stuff.
> > >
> > > Yeah, I never got the reason for this message. If Linux wants a
> > > certain value, it should program that. Whatever was left by someone
> > > else (be it U-Boot or some other software/OS) should not matter.
> > >
> > > Cheers,
> > > Andre
> > >
> > >>
> > >> Best regards
> > >>
> > >> Heinrich
> > >>
> > >>>
> > >>>> Can't we let the change depend on priv->use_internal_phy?
> > >>>>
> > >>>> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
> > >>>> index 5a1b38bf80..d7553fe163 100644
> > >>>> --- a/drivers/net/sun8i_emac.c
> > >>>> +++ b/drivers/net/sun8i_emac.c
> > >>>> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus,
> int
> > >>>> addr, int devad, int reg)
> > >>>> * The EMAC clock is either 200 or 300 MHz, so we need a
> divider
> > >>>> * of 128 to get the MDIO frequency below the required
> 2.5 MHz.
> > >>>> */
> > >>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>> + if (!priv->use_internal_phy)
> > >>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>>
> > >>>> mii_cmd |= MDIO_CMD_MII_BUSY;
> > >>>>
> > >>>> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev
> *bus, int
> > >>>> addr, int devad, int reg,
> > >>>> * The EMAC clock is either 200 or 300 MHz, so we need a
> divider
> > >>>> * of 128 to get the MDIO frequency below the required
> 2.5 MHz.
> > >>>> */
> > >>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>> + if (!priv->use_internal_phy)
> > >>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>>
> > >>>> mii_cmd |= MDIO_CMD_MII_WRITE;
> > >>>> mii_cmd |= MDIO_CMD_MII_BUSY;
>
I tried this with my v3s which has only an internal phy.
u-boot wasn't able to establish a connection with any speed and duplex mode
combination.
All i get is `ethernet@1c30000 Waiting for PHY auto negotiation to
complete......... TIMEOUT !`
Do you work with the internal or external phy @Heinrich?
Did you try the default u-boot with a different speed and duplex mode
combination and have the same behavior?
Greetings andreas
> >>>>
> > >>>> Best regards
> > >>>>
> > >>>> Heinrich
> > >>>>
> > >>>> I would assume the problem hits all H3 boards.
> > >>>
> > >>> And that's the confusing part: it does not. I tested this on my
> > >>> OrangePi Zero (H2+ with internal PHY), both back then with the
> original
> > >>> MDIO frequency patch and also now after your report. It always worked
> > >>> reliably for me.
> > >>> Also: I am still puzzled how one influences the other: The error you
> > >>> get is from the *MAC* soft reset: I would think this is an
> independent
> > >>> operation from any communication attempts with the PHY.
> > >>>
> > >>> There is this thread here about the same symptom on a V3s:
> > >>> https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
> > >>> v2 of this original patch (in the same thread) proposes some other
> > >>> solution, which I am also not very happy with. But just to get some
> more
> > >>> data points: can you check whether skipping the soft reset fixes this
> > >>> for you? I will have a closer look tonight to check the order of soft
> > >>> reset and PHY communication. Maybe we should only do the soft reset
> > >>> *once* when the MAC probes, and not on every .start call?
> > >>>
> > >>> Thanks,
> > >>> Andre
> > >>>
> > >>
> > >
> >
>
>
--
Mit freundlichen Grüßen / kind regards
Andreas Rehn | Master of Engineering (M.Eng)
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 16:57 ` Andre Przywara
2021-06-03 18:00 ` Andreas Rehn
@ 2021-06-03 20:09 ` Heinrich Schuchardt
1 sibling, 0 replies; 11+ messages in thread
From: Heinrich Schuchardt @ 2021-06-03 20:09 UTC (permalink / raw)
To: Andre Przywara
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List,
Andreas Rehn
[-- Attachment #1: Type: text/plain, Size: 9025 bytes --]
On 6/3/21 6:57 PM, Andre Przywara wrote:
> On Thu, 3 Jun 2021 18:09:25 +0200
> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>
> Hi,
>
>> On 6/3/21 1:04 PM, Andre Przywara wrote:
>>> On Thu, 3 Jun 2021 12:20:34 +0200
>>> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>>
>>> Hi,
>>>
>>>> On 6/3/21 11:04 AM, Andre Przywara wrote:
>>>>> On Thu, 3 Jun 2021 09:46:48 +0200
>>>>> Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>>>>
>>>>> Hi Heinrich,
>>>>>
>>>>>> On 6/2/21 3:08 PM, Ramon Fried wrote:
>>>>>>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt <xypron.glpk@gmx.de> wrote:
>>>>>>>>
>>>>>>>> Dear all,
>>>>>>>>
>>>>>>>> network is broken in U-Boot on orangepi_pc_defconfig:
>>>>>
>>>>> Thanks for the report!
>>>>>
>>>>>>>>
>>>>>>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 - 21:26:56 +0000)
>>>>>>>> Allwinner Technology
>>>>>>>> eth0: ethernet@1c30000
>>>>>>>> => dhcp
>>>>>>>> sun8i_emac_eth_start: Timeout
>>>>>>>>
>>>>>>>> Best regards
>>>>>>>>
>>>>>>>> Heinrich
>>>>>>>>
>>>>>>> Hi Heinrich, I don't have OrangePi. can you bisect and tell me when it broke ?
>>>>>>> Thanks.
>>>>>>> Ramon.
>>>>>>>
>>>>>>
>>>>>> Git bisect points to:
>>>>>>
>>>>>> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
>>>>>> Author: Andre Przywara <andre.przywara@arm.com>
>>>>>> Date: Mon Jul 6 01:40:45 2020 +0100
>>>>>>
>>>>>> net: sun8i-emac: Lower MDIO frequency
>>>>>>
>>>>>> Reverting the patch solves the problem for the OrangePi PC.
>>>>>>
>>>>>> According to the commit message the change was only needed for needed
>>>>>> for external PHYs.
>>>>>
>>>>> The external PHY on the Pine64 (non-plus) was the trigger, however
>>>>> both the manual and the Linux driver point to that we definitely need a
>>>>> higher divider. From what I can see, AHB2 (the EMAC clock) runs at 200
>>>>> MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16" results in 12.5
>>>>> MHz MDIO frequency. Can you check whether any other divider values fix
>>>>> this for you as well?
>>>>
>>>> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was used
>>>> before your patch.
>>>>
>>>> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
>>>> Waiting for PHY auto negotiation to complete. done
>>>> MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
>>>> MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
>>>> MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
>>>>
>>>> What is wrong about the approach in
>>>>
>>>> [PATCH 1/1] net: sun8i-emac: fix MDIO frequency
>>>> https://lists.denx.de/pipermail/u-boot/2021-June/451305.html ?
>>>
>>> The MDIO spec says that the maximum frequency on the MDIO bus should be
>>> 2.5 MHz. On the H3 with 200 MHz AHB and with a divider of 16 it's 12.5
>>
>> The dts says for the MDIO:
>>
>> int_mii_phy: ethernet-phy@1 {
>> clocks = <&ccu CLK_BUS_EPHY>;
>
> I think that is the clock to drive the PHY logic, not the MDIO bus.
> It's a gate for AHB1, in this case, which seems to be 200 MHz as well
> (because rate(AHB2)=rate(AHB1)/1).
>
>> "ahb" refers to CLK_BUS_SPI1:
>>
>> spi1: spi@1c69000 {
>> clocks = <&ccu CLK_BUS_SPI1>, <&ccu CLK_SPI1>;
>> clock-names = "ahb", "mod";
>>
>> Why are you mentioning AHB?
>
> AHB stands for "Advanced High-performance Bus", which is a traditional
> term for a (faster, data centric) bus interface in the ARM
> bus world. I am not sure the exact term is actually correct anymore
> with recent bus implementations, but the H3 manual refers to AHB
> everywhere, in particular as the bus name for the EMAC IP (figure 4-1 System
> Bus Tree). Apparently the bus clock is used as the clock source for the
> MDIO bus as well - after dividing it.
>
>>> MHz. Apparently the internal H3 PHY can tolerate those high
>>> frequencies, but it's beyond the spec. And certainly it should not fail
>>> with lower frequencies. And indeed it works for me on the H2+ (which I
>>> understand is just a binned version of the H3).
>>>
>>>> The Linux 5.10 driver runs fine no matter what value we choose for the
>>>> divider in U-Boot.
>>>
>>> Because it programs the divider correctly:
>>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316
>>> I take it you have no problems with Linux, so the divider being too
>>> high cannot be the root cause of the problem.
>>
>> Function stmmac_clk_csr_set() depends on the rate of the clock named
>> "stmmaceth":
>>
>> ethernet@1c30000 {
>> clocks = <&ccu CLK_BUS_EMAC>;
>> clock-names = "stmmaceth";
>>
>> How do we get the frequency of this clock?
>
> Have you checked /sys/kernel/debug/clk/clk_summary? You could also
> insert some printk to be sure. Otherwise the definition of bus_emac_clk
> and CLK_BUS_EMAC in drivers/clk/sunxi-ng/ccu-sun8i-h3.c tell you that
> it's a gate for ahb2. All of those point to 200 MHz on my OPi Zero.
I have appended clk_summary.
stmmac_clk_csr_set():
dwmac-sun8i 1c30000.ethernet: priv->plat->has_sun8i 1
dwmac-sun8i 1c30000.ethernet: clk_rate 200000000
dwmac-sun8i 1c30000.ethernet: priv->clk_csr 3
Best regards
Heinrich
>
> Hope that helps!
>
> Cheers,
> Andre
>>
>> Best regards
>>
>> Heinrich
>>
>>>
>>> So I'd rather fix this problem properly and correctly, especially as it
>>> seems to be more related to the EMAC soft reset (where your error
>>> message comes from).
>>>
>>>> But I see a message:
>>>>
>>>> dwmac-sun8i 1c30000.ethernet:
>>>> Current syscon value is not the default 148000 (expect 58000)
>>>>
>>>> @default_syscon_value:
>>>> The default value of the EMAC register in syscon
>>>> This value is used for disabling properly EMAC
>>>> and used as a good starting value in case of the
>>>> boot process(uboot) leave some stuff.
>>>
>>> Yeah, I never got the reason for this message. If Linux wants a
>>> certain value, it should program that. Whatever was left by someone
>>> else (be it U-Boot or some other software/OS) should not matter.
>>>
>>> Cheers,
>>> Andre
>>>
>>>>
>>>> Best regards
>>>>
>>>> Heinrich
>>>>
>>>>>
>>>>>> Can't we let the change depend on priv->use_internal_phy?
>>>>>>
>>>>>> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
>>>>>> index 5a1b38bf80..d7553fe163 100644
>>>>>> --- a/drivers/net/sun8i_emac.c
>>>>>> +++ b/drivers/net/sun8i_emac.c
>>>>>> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct mii_dev *bus, int
>>>>>> addr, int devad, int reg)
>>>>>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
>>>>>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
>>>>>> */
>>>>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>>>> + if (!priv->use_internal_phy)
>>>>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>>>>
>>>>>> mii_cmd |= MDIO_CMD_MII_BUSY;
>>>>>>
>>>>>> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct mii_dev *bus, int
>>>>>> addr, int devad, int reg,
>>>>>> * The EMAC clock is either 200 or 300 MHz, so we need a divider
>>>>>> * of 128 to get the MDIO frequency below the required 2.5 MHz.
>>>>>> */
>>>>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>>>> + if (!priv->use_internal_phy)
>>>>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
>>>>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
>>>>>>
>>>>>> mii_cmd |= MDIO_CMD_MII_WRITE;
>>>>>> mii_cmd |= MDIO_CMD_MII_BUSY;
>>>>>>
>>>>>> Best regards
>>>>>>
>>>>>> Heinrich
>>>>>>
>>>>>> I would assume the problem hits all H3 boards.
>>>>>
>>>>> And that's the confusing part: it does not. I tested this on my
>>>>> OrangePi Zero (H2+ with internal PHY), both back then with the original
>>>>> MDIO frequency patch and also now after your report. It always worked
>>>>> reliably for me.
>>>>> Also: I am still puzzled how one influences the other: The error you
>>>>> get is from the *MAC* soft reset: I would think this is an independent
>>>>> operation from any communication attempts with the PHY.
>>>>>
>>>>> There is this thread here about the same symptom on a V3s:
>>>>> https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
>>>>> v2 of this original patch (in the same thread) proposes some other
>>>>> solution, which I am also not very happy with. But just to get some more
>>>>> data points: can you check whether skipping the soft reset fixes this
>>>>> for you? I will have a closer look tonight to check the order of soft
>>>>> reset and PHY communication. Maybe we should only do the soft reset
>>>>> *once* when the MAC probes, and not on every .start call?
>>>>>
>>>>> Thanks,
>>>>> Andre
>>>>>
>>>>
>>>
>>
>
[-- Attachment #2: clk_summary --]
[-- Type: text/plain, Size: 13442 bytes --]
enable prepare protect duty
clock count count count rate accuracy phase cycle
---------------------------------------------------------------------------------------------
iosc 0 0 0 16000000 300000000 0 50000
ext_osc32k 1 1 0 32768 50000 0 50000
osc32k 1 1 0 32768 50000 0 50000
osc32k-out 0 0 0 32768 50000 0 50000
osc24M 18 18 0 24000000 50000 0 50000
ir 1 1 0 8000000 50000 0 50000
ar100 1 1 0 24000000 50000 0 50000
ahb0 1 1 0 24000000 50000 0 50000
apb0 3 3 0 24000000 50000 0 50000
apb0-twd 0 0 0 24000000 50000 0 50000
apb0-i2c 1 1 0 24000000 50000 0 50000
apb0-uart 0 0 0 24000000 50000 0 50000
apb0-timer 0 0 0 24000000 50000 0 50000
apb0-ir 1 1 0 24000000 50000 0 50000
apb0-pio 1 1 0 24000000 50000 0 50000
hdmi-ddc 2 2 0 24000000 50000 0 50000
avs 0 0 0 24000000 50000 0 50000
csi-mclk 0 0 0 24000000 50000 0 50000
csi-misc 0 0 0 24000000 50000 0 50000
usb-ohci3 1 1 0 24000000 50000 0 50000
usb-ohci2 1 1 0 24000000 50000 0 50000
usb-ohci1 1 1 0 24000000 50000 0 50000
usb-ohci0 1 1 0 24000000 50000 0 50000
usb-phy3 1 1 0 24000000 50000 0 50000
usb-phy2 1 1 0 24000000 50000 0 50000
usb-phy1 1 1 0 24000000 50000 0 50000
usb-phy0 1 1 0 24000000 50000 0 50000
spi1 0 0 0 24000000 50000 0 50000
spi0 0 0 0 24000000 50000 0 50000
ce 0 0 0 24000000 50000 0 50000
ts 0 0 0 24000000 50000 0 50000
mmc2 0 0 0 24000000 50000 0 50000
mmc2_output 0 0 0 24000000 50000 180 50000
mmc2_sample 0 0 0 24000000 50000 180 50000
mmc1 0 0 0 24000000 50000 0 50000
mmc1_output 0 0 0 24000000 50000 180 50000
mmc1_sample 0 0 0 24000000 50000 180 50000
nand 0 0 0 24000000 50000 0 50000
ths 0 0 0 24000000 50000 0 50000
apb2 1 1 0 24000000 50000 0 50000
bus-scr0 0 0 0 24000000 50000 0 50000
bus-uart3 0 0 0 24000000 50000 0 50000
bus-uart2 0 0 0 24000000 50000 0 50000
bus-uart1 0 0 0 24000000 50000 0 50000
bus-uart0 1 1 0 24000000 50000 0 50000
bus-i2c2 0 0 0 24000000 50000 0 50000
bus-i2c1 0 0 0 24000000 50000 0 50000
bus-i2c0 0 0 0 24000000 50000 0 50000
pll-de 1 1 0 432000000 50000 0 50000
de 2 2 0 432000000 50000 0 50000
wb-div 0 0 0 432000000 50000 0 50000
wb 0 0 0 432000000 50000 0 50000
mixer1-div 0 0 0 432000000 50000 0 50000
mixer1 0 0 0 432000000 50000 0 50000
mixer0-div 1 1 0 432000000 50000 0 50000
mixer0 1 1 0 432000000 50000 0 50000
tve 0 0 0 432000000 50000 0 50000
pll-periph1 0 0 0 600000000 50000 0 50000
pll-gpu 0 0 0 384000000 50000 0 50000
gpu 0 0 0 384000000 50000 0 50000
pll-periph0 3 3 0 600000000 50000 0 50000
mmc0 3 3 0 50000000 50000 0 50000
mmc0_output 1 1 0 50000000 50000 90 50000
mmc0_sample 1 1 0 50000000 50000 120 50000
csi-sclk 0 0 0 600000000 50000 0 50000
deinterlace 0 0 0 600000000 50000 0 50000
ahb1 11 11 0 200000000 50000 0 50000
bus-dbg 0 0 0 200000000 50000 0 50000
bus-ephy 1 1 0 200000000 50000 0 50000
bus-spinlock 0 0 0 200000000 50000 0 50000
bus-msgbox 0 0 0 200000000 50000 0 50000
bus-gpu 0 0 0 200000000 50000 0 50000
bus-de 2 2 0 200000000 50000 0 50000
bus-wb 0 0 0 200000000 50000 0 50000
bus-mixer1 0 0 0 200000000 50000 0 50000
bus-mixer0 1 1 0 200000000 50000 0 50000
bus-hdmi 2 2 0 200000000 50000 0 50000
bus-tve 0 0 0 200000000 50000 0 50000
bus-csi 0 0 0 200000000 50000 0 50000
bus-deinterlace 0 0 0 200000000 50000 0 50000
bus-tcon1 0 0 0 200000000 50000 0 50000
bus-tcon0 1 1 0 200000000 50000 0 50000
bus-ve 0 0 0 200000000 50000 0 50000
bus-ohci0 2 2 0 200000000 50000 0 50000
bus-ehci0 2 2 0 200000000 50000 0 50000
bus-otg 1 1 0 200000000 50000 0 50000
bus-spi1 0 0 0 200000000 50000 0 50000
bus-spi0 0 0 0 200000000 50000 0 50000
bus-hstimer 0 0 0 200000000 50000 0 50000
bus-ts 0 0 0 200000000 50000 0 50000
bus-dram 0 0 0 200000000 50000 0 50000
bus-nand 0 0 0 200000000 50000 0 50000
bus-mmc2 0 0 0 200000000 50000 0 50000
bus-mmc1 0 0 0 200000000 50000 0 50000
bus-mmc0 1 1 0 200000000 50000 0 50000
bus-dma 1 1 0 200000000 50000 0 50000
bus-ce 0 0 0 200000000 50000 0 50000
ahb2 7 7 0 200000000 50000 0 50000
bus-ohci3 2 2 0 200000000 50000 0 50000
bus-ohci2 2 2 0 200000000 50000 0 50000
bus-ohci1 2 2 0 200000000 50000 0 50000
bus-ehci3 2 2 0 200000000 50000 0 50000
bus-ehci2 2 2 0 200000000 50000 0 50000
bus-ehci1 2 2 0 200000000 50000 0 50000
bus-emac 2 2 0 200000000 50000 0 50000
apb1 2 2 0 100000000 50000 0 50000
bus-i2s2 0 0 0 100000000 50000 0 50000
bus-i2s1 0 0 0 100000000 50000 0 50000
bus-i2s0 0 0 0 100000000 50000 0 50000
bus-ths 0 0 0 100000000 50000 0 50000
bus-pio 1 1 0 100000000 50000 0 50000
bus-spdif 0 0 0 100000000 50000 0 50000
bus-codec 1 1 0 100000000 50000 0 50000
pll-periph0-2x 1 1 0 1200000000 50000 0 50000
mbus 1 1 0 300000000 50000 0 50000
pll-ddr 1 1 0 1248000000 50000 0 50000
dram 1 1 0 1248000000 50000 0 50000
dram-ts 0 0 0 1248000000 50000 0 50000
dram-deinterlace 0 0 0 1248000000 50000 0 50000
dram-csi 0 0 0 1248000000 50000 0 50000
dram-ve 0 0 0 1248000000 50000 0 50000
pll-ve 0 0 0 402000000 50000 0 50000
ve 0 0 0 402000000 50000 0 50000
pll-video 2 2 0 294000000 50000 0 50000
hdmi-phy-clk 1 1 0 73500000 50000 0 50000
hdmi 1 1 0 294000000 50000 0 50000
tcon 0 0 0 294000000 50000 0 50000
pll-audio-base 0 0 0 98285714 50000 0 50000
pll-audio-8x 0 0 0 196571428 50000 0 50000
i2s2 0 0 0 196571428 50000 0 50000
i2s1 0 0 0 196571428 50000 0 50000
i2s0 0 0 0 196571428 50000 0 50000
pll-audio-4x 0 0 0 98285714 50000 0 50000
pll-audio-2x 0 0 0 49142857 50000 0 50000
pll-audio 0 0 0 98285714 50000 0 50000
ac-dig 0 0 0 98285714 50000 0 50000
spdif 0 0 0 98285714 50000 0 50000
pll-cpux 1 1 0 648000000 50000 0 50000
cpux 1 1 0 648000000 50000 0 50000
axi 0 0 0 216000000 50000 0 50000
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [BUG] network is broken on Orange Pi PC
2021-06-03 18:00 ` Andreas Rehn
@ 2021-06-03 20:19 ` Heinrich Schuchardt
0 siblings, 0 replies; 11+ messages in thread
From: Heinrich Schuchardt @ 2021-06-03 20:19 UTC (permalink / raw)
To: Andreas Rehn, Andre Przywara
Cc: Ramon Fried, Joe Hershberger, Jagan Teki, U-Boot Mailing List
On 6/3/21 8:00 PM, Andreas Rehn wrote:
>
>
> Am Do., 3. Juni 2021 um 18:58 Uhr schrieb Andre Przywara
> <andre.przywara@arm.com <mailto:andre.przywara@arm.com>>:
>
> On Thu, 3 Jun 2021 18:09:25 +0200
> Heinrich Schuchardt <xypron.glpk@gmx.de <mailto:xypron.glpk@gmx.de>>
> wrote:
>
> Hi,
>
> > On 6/3/21 1:04 PM, Andre Przywara wrote:
> > > On Thu, 3 Jun 2021 12:20:34 +0200
> > > Heinrich Schuchardt <xypron.glpk@gmx.de
> <mailto:xypron.glpk@gmx.de>> wrote:
> > >
> > > Hi,
> > >
> > >> On 6/3/21 11:04 AM, Andre Przywara wrote:
> > >>> On Thu, 3 Jun 2021 09:46:48 +0200
> > >>> Heinrich Schuchardt <xypron.glpk@gmx.de
> <mailto:xypron.glpk@gmx.de>> wrote:
> > >>>
> > >>> Hi Heinrich,
> > >>>
> > >>>> On 6/2/21 3:08 PM, Ramon Fried wrote:
> > >>>>> On Tue, Jun 1, 2021 at 12:35 AM Heinrich Schuchardt
> <xypron.glpk@gmx.de <mailto:xypron.glpk@gmx.de>> wrote:
> > >>>>>>
> > >>>>>> Dear all,
> > >>>>>>
> > >>>>>> network is broken in U-Boot on orangepi_pc_defconfig:
> > >>>
> > >>> Thanks for the report!
> > >>>
> > >>>>>>
> > >>>>>> U-Boot 2021.07-rc3-00059-gd8729a114e (May 31 2021 -
> 21:26:56 +0000)
> > >>>>>> Allwinner Technology
> > >>>>>> eth0: ethernet@1c30000
> > >>>>>> => dhcp
> > >>>>>> sun8i_emac_eth_start: Timeout
> > >>>>>>
> > >>>>>> Best regards
> > >>>>>>
> > >>>>>> Heinrich
> > >>>>>>
> > >>>>> Hi Heinrich, I don't have OrangePi. can you bisect and tell
> me when it broke ?
> > >>>>> Thanks.
> > >>>>> Ramon.
> > >>>>>
> > >>>>
> > >>>> Git bisect points to:
> > >>>>
> > >>>> commit 4f0278dac56a658ef1e0967fec0bb95372a875bd
> > >>>> Author: Andre Przywara <andre.przywara@arm.com
> <mailto:andre.przywara@arm.com>>
> > >>>> Date: Mon Jul 6 01:40:45 2020 +0100
> > >>>>
> > >>>> net: sun8i-emac: Lower MDIO frequency
> > >>>>
> > >>>> Reverting the patch solves the problem for the OrangePi PC.
> > >>>>
> > >>>> According to the commit message the change was only needed
> for needed
> > >>>> for external PHYs.
> > >>>
> > >>> The external PHY on the Pine64 (non-plus) was the trigger,
> however
> > >>> both the manual and the Linux driver point to that we
> definitely need a
> > >>> higher divider. From what I can see, AHB2 (the EMAC clock)
> runs at 200
> > >>> MHz (AHB2=AHB1/1=PLL6/3=200 MHz). So just having "/ 16"
> results in 12.5
> > >>> MHz MDIO frequency. Can you check whether any other divider
> values fix
> > >>> this for you as well?
> > >>
> > >> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 and this is the value that was
> used
> > >> before your patch.
> > >>
> > >> MDIO_CMD_MII_CLK_CSR_DIV_16 = 0 =>
> > >> Waiting for PHY auto negotiation to complete. done
> > >> MDIO_CMD_MII_CLK_CSR_DIV_32 = 1 => sun8i_emac_eth_start: Timeout
> > >> MDIO_CMD_MII_CLK_CSR_DIV_64 = 2 => sun8i_emac_eth_start: Timeout
> > >> MDIO_CMD_MII_CLK_CSR_DIV_128 = 3 => sun8i_emac_eth_start: Timeout
> > >>
> > >> What is wrong about the approach in
> > >>
> > >> [PATCH 1/1] net: sun8i-emac: fix MDIO frequency
> > >> https://lists.denx.de/pipermail/u-boot/2021-June/451305.html
> <https://lists.denx.de/pipermail/u-boot/2021-June/451305.html> ?
> > >
> > > The MDIO spec says that the maximum frequency on the MDIO bus
> should be
> > > 2.5 MHz. On the H3 with 200 MHz AHB and with a divider of 16
> it's 12.5
> >
> > The dts says for the MDIO:
> >
> > int_mii_phy: ethernet-phy@1 {
> > clocks = <&ccu CLK_BUS_EPHY>;
>
> I think that is the clock to drive the PHY logic, not the MDIO bus.
> It's a gate for AHB1, in this case, which seems to be 200 MHz as well
> (because rate(AHB2)=rate(AHB1)/1).
>
> > "ahb" refers to CLK_BUS_SPI1:
> >
> > spi1: spi@1c69000 {
> > clocks = <&ccu CLK_BUS_SPI1>, <&ccu CLK_SPI1>;
> > clock-names = "ahb", "mod";
> >
> > Why are you mentioning AHB?
>
> AHB stands for "Advanced High-performance Bus", which is a traditional
> term for a (faster, data centric) bus interface in the ARM
> bus world. I am not sure the exact term is actually correct anymore
> with recent bus implementations, but the H3 manual refers to AHB
> everywhere, in particular as the bus name for the EMAC IP (figure
> 4-1 System
> Bus Tree). Apparently the bus clock is used as the clock source for the
> MDIO bus as well - after dividing it.
>
> > > MHz. Apparently the internal H3 PHY can tolerate those high
> > > frequencies, but it's beyond the spec. And certainly it should
> not fail
> > > with lower frequencies. And indeed it works for me on the H2+
> (which I
> > > understand is just a binned version of the H3).
> > >
> > >> The Linux 5.10 driver runs fine no matter what value we choose
> for the
> > >> divider in U-Boot.
> > >
> > > Because it programs the divider correctly:
> > >
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316
> <https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#n316>
> > > I take it you have no problems with Linux, so the divider being too
> > > high cannot be the root cause of the problem.
> >
> > Function stmmac_clk_csr_set() depends on the rate of the clock named
> > "stmmaceth":
> >
> > ethernet@1c30000 {
> > clocks = <&ccu CLK_BUS_EMAC>;
> > clock-names = "stmmaceth";
> >
> > How do we get the frequency of this clock?
>
> Have you checked /sys/kernel/debug/clk/clk_summary? You could also
> insert some printk to be sure. Otherwise the definition of bus_emac_clk
> and CLK_BUS_EMAC in drivers/clk/sunxi-ng/ccu-sun8i-h3.c tell you that
> it's a gate for ahb2. All of those point to 200 MHz on my OPi Zero.
>
> Hope that helps!
>
> Cheers,
> Andre
> >
> > Best regards
> >
> > Heinrich
> >
> > >
> > > So I'd rather fix this problem properly and correctly,
> especially as it
> > > seems to be more related to the EMAC soft reset (where your error
> > > message comes from).
> > >
> > >> But I see a message:
> > >>
> > >> dwmac-sun8i 1c30000.ethernet:
> > >> Current syscon value is not the default 148000 (expect 58000)
> > >>
> > >> @default_syscon_value:
> > >> The default value of the EMAC register in syscon
> > >> This value is used for disabling properly EMAC
> > >> and used as a good starting value in case of the
> > >> boot process(uboot) leave some stuff.
> > >
> > > Yeah, I never got the reason for this message. If Linux wants a
> > > certain value, it should program that. Whatever was left by someone
> > > else (be it U-Boot or some other software/OS) should not matter.
> > >
> > > Cheers,
> > > Andre
> > >
> > >>
> > >> Best regards
> > >>
> > >> Heinrich
> > >>
> > >>>
> > >>>> Can't we let the change depend on priv->use_internal_phy?
> > >>>>
> > >>>> diff --git a/drivers/net/sun8i_emac.c b/drivers/net/sun8i_emac.c
> > >>>> index 5a1b38bf80..d7553fe163 100644
> > >>>> --- a/drivers/net/sun8i_emac.c
> > >>>> +++ b/drivers/net/sun8i_emac.c
> > >>>> @@ -211,7 +211,9 @@ static int sun8i_mdio_read(struct
> mii_dev *bus, int
> > >>>> addr, int devad, int reg)
> > >>>> * The EMAC clock is either 200 or 300 MHz, so we
> need a divider
> > >>>> * of 128 to get the MDIO frequency below the
> required 2.5 MHz.
> > >>>> */
> > >>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>> + if (!priv->use_internal_phy)
> > >>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>>
> > >>>> mii_cmd |= MDIO_CMD_MII_BUSY;
> > >>>>
> > >>>> @@ -242,7 +244,9 @@ static int sun8i_mdio_write(struct
> mii_dev *bus, int
> > >>>> addr, int devad, int reg,
> > >>>> * The EMAC clock is either 200 or 300 MHz, so we
> need a divider
> > >>>> * of 128 to get the MDIO frequency below the
> required 2.5 MHz.
> > >>>> */
> > >>>> - mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>> + if (!priv->use_internal_phy)
> > >>>> + mii_cmd |= MDIO_CMD_MII_CLK_CSR_DIV_128 <<
> > >>>> + MDIO_CMD_MII_CLK_CSR_SHIFT;
> > >>>>
> > >>>> mii_cmd |= MDIO_CMD_MII_WRITE;
> > >>>> mii_cmd |= MDIO_CMD_MII_BUSY;
>
>
> I tried this with my v3s which has only an internal phy.
> u-boot wasn't able to establish a connection with any speed and duplex
> mode combination.
> All i get is `ethernet@1c30000 Waiting for PHY auto negotiation to
> complete......... TIMEOUT !`
>
> Do you work with the internal or external phy @Heinrich?
> Did you try the default u-boot with a different speed and duplex mode
> combination and have the same behavior?
> Greetings andreas
I simply use orangepi_pc_defconfig. It worked out of the box in the
previous release. Communication speed and duplex mode should be
auto-negotiated.
Best regards
Heinrich
>
> > >>>>
> > >>>> Best regards
> > >>>>
> > >>>> Heinrich
> > >>>>
> > >>>> I would assume the problem hits all H3 boards.
> > >>>
> > >>> And that's the confusing part: it does not. I tested this on my
> > >>> OrangePi Zero (H2+ with internal PHY), both back then with
> the original
> > >>> MDIO frequency patch and also now after your report. It
> always worked
> > >>> reliably for me.
> > >>> Also: I am still puzzled how one influences the other: The
> error you
> > >>> get is from the *MAC* soft reset: I would think this is an
> independent
> > >>> operation from any communication attempts with the PHY.
> > >>>
> > >>> There is this thread here about the same symptom on a V3s:
> > >>> https://lists.denx.de/pipermail/u-boot/2021-May/450315.html
> <https://lists.denx.de/pipermail/u-boot/2021-May/450315.html>
> > >>> v2 of this original patch (in the same thread) proposes some
> other
> > >>> solution, which I am also not very happy with. But just to
> get some more
> > >>> data points: can you check whether skipping the soft reset
> fixes this
> > >>> for you? I will have a closer look tonight to check the order
> of soft
> > >>> reset and PHY communication. Maybe we should only do the soft
> reset
> > >>> *once* when the MAC probes, and not on every .start call?
> > >>>
> > >>> Thanks,
> > >>> Andre
> > >>>
> > >>
> > >
> >
>
>
>
> --
> Mit freundlichen Grüßen / kind regards
> Andreas Rehn | Master of Engineering (M.Eng)
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2021-06-03 20:19 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-05-31 21:35 [BUG] network is broken on Orange Pi PC Heinrich Schuchardt
2021-06-02 13:08 ` Ramon Fried
2021-06-03 7:46 ` Heinrich Schuchardt
2021-06-03 9:04 ` Andre Przywara
2021-06-03 10:20 ` Heinrich Schuchardt
2021-06-03 11:04 ` Andre Przywara
2021-06-03 16:09 ` Heinrich Schuchardt
2021-06-03 16:57 ` Andre Przywara
2021-06-03 18:00 ` Andreas Rehn
2021-06-03 20:19 ` Heinrich Schuchardt
2021-06-03 20:09 ` Heinrich Schuchardt
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.