[PATCH] ARM64: meson-gxl: disable broken eee

Hi Joseph,

Post by Joseph Kogut
Meson GXL appears to suffer from the same broken eee issue as GXB,
originally fixed by feb3cbea0946.
This patch disables the broken energy-efficient ethernet for GXL,
avoiding the tx link breakage.

Can you share what platforms you tested to find/fix this problem?

Thanks for the patch!

Kevin

Joseph Kogut

2017-07-23 17:03:00 UTC

Hi Kevin,

I tested on a P212 reference board, which is currently the only GXL
based board I have.

Before applying the patch, high activity on the ethernet interface
would cause the link to break, requiring the interface to be brought
down and back up before it would work again. After applying the patch,
I could not get the link to break with any transfers.

Neil Armstrong

2017-07-24 12:06:04 UTC

This post might be inappropriate. Click to display it.

Neil Armstrong

2017-07-24 12:26:52 UTC

Post by Joseph Kogut
Hi Kevin,
I tested on a P212 reference board, which is currently the only GXL
based board I have.
Before applying the patch, high activity on the ethernet interface
would cause the link to break, requiring the interface to be brought
down and back up before it would work again. After applying the patch,
I could not get the link to break with any transfers.
_______________________________________________
linux-arm-kernel mailing list
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Hi Joseph,
[please keep the history of the conversation when replying]
Is it a original P212 board, or a board based on the P212 ref design ?
I'm currently testing the v4.13-rc2 on the official P212 board we received from Amlogic
# while true ; do iperf3 -c 192.168.1.21 ; iperf3 -c 192.168.1.21 -R ; done
That cycles between the two iperf directions.
Connecting to host 192.168.1.21, port 5201
[ 4] local 192.168.1.183 port 51286 connected to 192.168.1.21 port 5201
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.00 sec 1.64 MBytes 13.7 Mbits/sec 2 1.41 KBytes
[ 4] 1.00-2.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 2.00-3.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 3.00-4.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 4.00-5.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 5.00-6.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 6.00-7.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 7.00-8.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 8.00-9.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
# ifconfig eth0 down
# ifconfig eth0 up
[ 1483.009919] meson8b-dwmac c9410000.ethernet eth0: Link is Up - 100Mbps/Full - flow control rx/tx
[ 1487.105725] meson8b-dwmac c9410000.ethernet eth0: Link is Down
[ 1490.177943] meson8b-dwmac c9410000.ethernet eth0: Link is Up - 100Mbps/Full - flow control rx/tx
[ 1494.273713] meson8b-dwmac c9410000.ethernet eth0: Link is Down
and so on.
And this patch does not fix this at all.
It seems to be more a PHY driver issue here.
Could you share the results with the same command ?
Neil

I forgot to precise, this issue occurs when the "GXL PHY driver" is disabled, only relying on the
Generic PHY Driver.

[ 630.767538] Generic PHY 0.e40908ff:08: attached PHY driver [Generic PHY] (mii_bus:phy_addr=0.e40908ff:08, irq=-1)

When the PHY driver is loaded the issue disappears, with and without your patch.

[ 25.355620] Meson GXL Internal PHY 0.e40908ff:08: attached PHY driver [Meson GXL Internal PHY] (mii_bus:phy_addr=0.e40908ff:08, irq=-1)

Could you please check the PHY Driver is loaded ?

Thanks,
Neil

Jerome Brunet

2017-07-24 16:09:30 UTC

Hi Joseph,
[please keep the history of the conversation when replying]
Is it a original P212 board, or a board based on the P212 ref design ?
I'm currently testing the v4.13-rc2 on the official P212 board we received from Amlogic
and using iperf3 in both directions, I have an issue after a few minutes
# while true ; do iperf3 -c 192.168.1.21 ; iperf3 -c 192.168.1.21 -R ; done
That cycles between the two iperf directions.
Connecting to host 192.168.1.21, port 5201
[  4] local 192.168.1.183 port 51286 connected to 192.168.1.21 port 5201
[ ID] Interval           Transfer     Bandwidth       Retr  Cwnd
[  4]   0.00-1.00   sec  1.64 MBytes  13.7 Mbits/sec    2   1.41 KBytes
[  4]   1.00-2.00   sec  0.00 Bytes  0.00 bits/sec    1   1.41 KBytes
[  4]   2.00-3.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   3.00-4.00   sec  0.00 Bytes  0.00 bits/sec    1   1.41 KBytes
[  4]   4.00-5.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   5.00-6.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   6.00-7.00   sec  0.00 Bytes  0.00 bits/sec    1   1.41 KBytes
[  4]   7.00-8.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   8.00-9.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes

Same kind of test shows no issue on the libretech-cc and the khadas-vim
In this case, I don't think meson-gxl.dtsi is where you should make your change.

Post by Kevin Hilman
# ifconfig eth0 down
# ifconfig eth0 up
[ 1483.009919] meson8b-dwmac c9410000.ethernet eth0: Link is Up -
100Mbps/Full - flow control rx/tx
[ 1487.105725] meson8b-dwmac c9410000.ethernet eth0: Link is Down
[ 1490.177943] meson8b-dwmac c9410000.ethernet eth0: Link is Up -
100Mbps/Full - flow control rx/tx
[ 1494.273713] meson8b-dwmac c9410000.ethernet eth0: Link is Down

This does not look like the issue we had on the odroid-c2.
However, I have seen behavior like this (link down/up flickering) when the rgmii
delays are way off.

Maybe you could try playing with "amlogic,tx-delay-ns" property ?

Post by Kevin Hilman
and so on.
And this patch does not fix this at all.
It seems to be more a PHY driver issue here.
Could you share the results with the same command ?
Neil

I forgot to precise, this issue occurs when the "GXL PHY driver" is disabled,
only relying on the
Generic PHY Driver.
[ 630.767538] Generic PHY 0.e40908ff:08: attached PHY driver [Generic PHY]
(mii_bus:phy_addr=0.e40908ff:08, irq=-1)
When the PHY driver is loaded the issue disappears, with and without your patch.
[ 25.355620] Meson GXL Internal PHY 0.e40908ff:08: attached PHY driver
[Meson GXL Internal PHY] (mii_bus:phy_addr=0.e40908ff:08, irq=-1)
Could you please check the PHY Driver is loaded ?
Thanks,
Neil

Martin Blumenstingl

2017-07-24 18:20:45 UTC

Hi Joseph,
[please keep the history of the conversation when replying]
Is it a original P212 board, or a board based on the P212 ref design ?
I'm currently testing the v4.13-rc2 on the official P212 board we received
from Amlogic
and using iperf3 in both directions, I have an issue after a few minutes
# while true ; do iperf3 -c 192.168.1.21 ; iperf3 -c 192.168.1.21 -R ; done
That cycles between the two iperf directions.
Connecting to host 192.168.1.21, port 5201
[ 4] local 192.168.1.183 port 51286 connected to 192.168.1.21 port 5201
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.00 sec 1.64 MBytes 13.7 Mbits/sec 2 1.41 KBytes
[ 4] 1.00-2.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 2.00-3.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 3.00-4.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 4.00-5.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 5.00-6.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 6.00-7.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 7.00-8.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 8.00-9.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes

Same kind of test shows no issue on the libretech-cc and the khadas-vim
In this case, I don't think meson-gxl.dtsi is where you should make your change.

This does not look like the issue we had on the odroid-c2.
However, I have seen behavior like this (link down/up flickering) when the rgmii
delays are way off.
Maybe you could try playing with "amlogic,tx-delay-ns" property ?

we need to change dwmac-meson8b.c so it applies the TX delay also for
RMII PHYs. currently the TX delay is always set to 0 for RMII PHYs,
see: [0]

Post by Kevin Hilman
and so on.
And this patch does not fix this at all.
It seems to be more a PHY driver issue here.
Could you share the results with the same command ?
Neil

I forgot to precise, this issue occurs when the "GXL PHY driver" is disabled,
only relying on the
Generic PHY Driver.
[ 630.767538] Generic PHY 0.e40908ff:08: attached PHY driver [Generic PHY]
(mii_bus:phy_addr=0.e40908ff:08, irq=-1)
When the PHY driver is loaded the issue disappears, with and without your patch.
[ 25.355620] Meson GXL Internal PHY 0.e40908ff:08: attached PHY driver
[Meson GXL Internal PHY] (mii_bus:phy_addr=0.e40908ff:08, irq=-1)
Could you please check the PHY Driver is loaded ?
Thanks,
Neil

_______________________________________________
linux-amlogic mailing list
http://lists.infradead.org/mailman/listinfo/linux-amlogic

[0] https://github.com/torvalds/linux/blob/master/drivers/net/ethernet/stmicro/stmmac/dwmac-meson8b.c#L222

Jerome Brunet

2017-07-24 18:32:01 UTC

Hi Joseph,
[please keep the history of the conversation when replying]
Is it a original P212 board, or a board based on the P212 ref design ?
I'm currently testing the v4.13-rc2 on the official P212 board we received
from Amlogic
and using iperf3 in both directions, I have an issue after a few minutes
# while true ; do iperf3 -c 192.168.1.21 ; iperf3 -c 192.168.1.21 -R ; done
That cycles between the two iperf directions.
Connecting to host 192.168.1.21, port 5201
[  4] local 192.168.1.183 port 51286 connected to 192.168.1.21 port 5201
[ ID] Interval           Transfer     Bandwidth       Retr  Cwnd
[  4]   0.00-1.00   sec  1.64 MBytes  13.7 Mbits/sec    2   1.41 KBytes
[  4]   1.00-2.00   sec  0.00 Bytes  0.00 bits/sec    1   1.41 KBytes
[  4]   2.00-3.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   3.00-4.00   sec  0.00 Bytes  0.00 bits/sec    1   1.41 KBytes
[  4]   4.00-5.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   5.00-6.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   6.00-7.00   sec  0.00 Bytes  0.00 bits/sec    1   1.41 KBytes
[  4]   7.00-8.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes
[  4]   8.00-9.00   sec  0.00 Bytes  0.00 bits/sec    0   1.41 KBytes

Same kind of test shows no issue on the libretech-cc and the khadas-vim
In this case, I don't think meson-gxl.dtsi is where you should make your change.

ahahah, silly me ! I forgot that it was 100MBps PHY.
tx delay makes no sense here, 2ns delay would not make much difference on a
25MHz clock

Sorry for the confusion

This does not look like the issue we had on the odroid-c2.
However, I have seen behavior like this (link down/up flickering) when the rgmii
delays are way off.
Maybe you could try playing with "amlogic,tx-delay-ns" property ?

we need to change dwmac-meson8b.c so it applies the TX delay also for
RMII PHYs. currently the TX delay is always set to 0 for RMII PHYs,
see: [0]

Post by Kevin Hilman
and so on.
And this patch does not fix this at all.
It seems to be more a PHY driver issue here.
Could you share the results with the same command ?
Neil

I forgot to precise, this issue occurs when the "GXL PHY driver" is disabled,
only relying on the
Generic PHY Driver.
[ 630.767538] Generic PHY 0.e40908ff:08: attached PHY driver [Generic PHY]
(mii_bus:phy_addr=0.e40908ff:08, irq=-1)
When the PHY driver is loaded the issue disappears, with and without your patch.
[ 25.355620] Meson GXL Internal PHY 0.e40908ff:08: attached PHY driver
[Meson GXL Internal PHY] (mii_bus:phy_addr=0.e40908ff:08, irq=-1)
Could you please check the PHY Driver is loaded ?
Thanks,
Neil

_______________________________________________
linux-amlogic mailing list
http://lists.infradead.org/mailman/listinfo/linux-amlogic

[0] https://github.com/torvalds/linux/blob/master/drivers/net/ethernet/stmicro
/stmmac/dwmac-meson8b.c#L222

crow

2017-07-25 17:03:14 UTC

Hello,

Hi Joseph,
[please keep the history of the conversation when replying]
Is it a original P212 board, or a board based on the P212 ref design ?
I'm currently testing the v4.13-rc2 on the official P212 board we received
from Amlogic
and using iperf3 in both directions, I have an issue after a few minutes
# while true ; do iperf3 -c 192.168.1.21 ; iperf3 -c 192.168.1.21 -R ; done
That cycles between the two iperf directions.
Connecting to host 192.168.1.21, port 5201
[ 4] local 192.168.1.183 port 51286 connected to 192.168.1.21 port 5201
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.00 sec 1.64 MBytes 13.7 Mbits/sec 2 1.41 KBytes
[ 4] 1.00-2.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 2.00-3.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 3.00-4.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 4.00-5.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 5.00-6.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 6.00-7.00 sec 0.00 Bytes 0.00 bits/sec 1 1.41 KBytes
[ 4] 7.00-8.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes
[ 4] 8.00-9.00 sec 0.00 Bytes 0.00 bits/sec 0 1.41 KBytes

Same kind of test shows no issue on the libretech-cc and the khadas-vim
In this case, I don't think meson-gxl.dtsi is where you should make your change.

I am using mainline Kernel on Khadas VIM Pro and have such problems. I
did send today an update [0] to my older mailinglist post.

Post by Jerome Brunet
ahahah, silly me ! I forgot that it was 100MBps PHY.
tx delay makes no sense here, 2ns delay would not make much difference on a
25MHz clock
Sorry for the confusion

This does not look like the issue we had on the odroid-c2.
However, I have seen behavior like this (link down/up flickering) when the rgmii
delays are way off.
Maybe you could try playing with "amlogic,tx-delay-ns" property ?

we need to change dwmac-meson8b.c so it applies the TX delay also for
RMII PHYs. currently the TX delay is always set to 0 for RMII PHYs,
see: [0]