1. Patchew
  2. linux
  3. View series

[PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value

Dongjin Kim posted 1 patch 1 month, 2 weeks ago
There is a newer version of this series
arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
[PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Dongjin Kim 1 month, 2 weeks ago 
Very poor network performance is observed on the ODROID-M1 Board
(REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
this problem can be resolved after changing the value of TXD
timing to 0x35 from 0x4f.

Signed-off-by: Dongjin Kim <tobetter@gmail.com>
Change-Id: Idecd246e3fe93447b6e3e9b165d077d2050bbd75
---
 arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts b/arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts
index 0f844806ec542..e1a550a04498e 100644
--- a/arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts
+++ b/arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts
@@ -210,7 +210,7 @@ &gmac0_rgmii_clk
 		     &gmac0_rgmii_bus>;
 	status = "okay";
 
-	tx_delay = <0x4f>;
+	tx_delay = <0x35>;
 	rx_delay = <0x2d>;
 };
 
-- 
2.34.1
Re: [PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Andrew Lunn 1 month, 2 weeks ago 
On Tue, Aug 19, 2025 at 01:50:17PM +0900, Dongjin Kim wrote:
> Very poor network performance is observed on the ODROID-M1 Board
> (REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
> this problem can be resolved after changing the value of TXD
> timing to 0x35 from 0x4f.

How did you decide on 0x35?

~/linux/arch/arm64$ grep -hr "tx_delay = " | sort | uniq -c
      4 	tx_delay = <0x10>;
      1 	tx_delay = <0x1a>;
      1 	tx_delay = <0x21>;
      2 	tx_delay = <0x22>;
      4 	tx_delay = <0x24>;
      2 	tx_delay = <0x26>;
     14 	tx_delay = <0x28>;
      1 	tx_delay = <0x2a>;
      6 	tx_delay = <0x30>;
      1 	tx_delay = <0x3a>;
      3 	tx_delay = <0x3c>;
      2 	tx_delay = <0x42>;
      5 	tx_delay = <0x43>;
      2 	tx_delay = <0x44>;
      1 	tx_delay = <0x46>;
      6 	tx_delay = <0x4f>;

The numbers are all over the place, but it is clear that 0x28 is the
most popular for some reason. But since this delay values are magic,
i've no idea why.

     Andrew
Re: [PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Dongjin Kim 1 month, 2 weeks ago 
On Tue, Aug 19, 2025 at 06:03:47PM +0200, Andrew Lunn wrote:
> On Tue, Aug 19, 2025 at 01:50:17PM +0900, Dongjin Kim wrote:
> > Very poor network performance is observed on the ODROID-M1 Board
> > (REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
> > this problem can be resolved after changing the value of TXD
> > timing to 0x35 from 0x4f.
> 
> How did you decide on 0x35?
> 
> ~/linux/arch/arm64$ grep -hr "tx_delay = " | sort | uniq -c
>       4 	tx_delay = <0x10>;
>       1 	tx_delay = <0x1a>;
>       1 	tx_delay = <0x21>;
>       2 	tx_delay = <0x22>;
>       4 	tx_delay = <0x24>;
>       2 	tx_delay = <0x26>;
>      14 	tx_delay = <0x28>;
>       1 	tx_delay = <0x2a>;
>       6 	tx_delay = <0x30>;
>       1 	tx_delay = <0x3a>;
>       3 	tx_delay = <0x3c>;
>       2 	tx_delay = <0x42>;
>       5 	tx_delay = <0x43>;
>       2 	tx_delay = <0x44>;
>       1 	tx_delay = <0x46>;
>       6 	tx_delay = <0x4f>;
> 
> The numbers are all over the place, but it is clear that 0x28 is the
> most popular for some reason. But since this delay values are magic,
> i've no idea why.
> 
>      Andrew
I do have more numbers in the Rockchip BSP kernel, and 0x35 is from
one of them that shows less retry counts when testing 'iperf3' while
testing with the batch of ODROID-M1. :)

Thanks,
Dongjin Kim.
Re: [PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Stefan Monnier 1 month ago 
Dongjin Kim wrote:
> On Tue, Aug 19, 2025 at 06:03:47PM +0200, Andrew Lunn wrote:
> > On Tue, Aug 19, 2025 at 01:50:17PM +0900, Dongjin Kim wrote:
> > > Very poor network performance is observed on the ODROID-M1 Board
> > > (REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
> > > this problem can be resolved after changing the value of TXD
> > > timing to 0x35 from 0x4f.
> > 
> > How did you decide on 0x35?
> > 
> > ~/linux/arch/arm64$ grep -hr "tx_delay = " | sort | uniq -c
> >       4 	tx_delay = <0x10>;
> >       1 	tx_delay = <0x1a>;
> >       1 	tx_delay = <0x21>;
> >       2 	tx_delay = <0x22>;
> >       4 	tx_delay = <0x24>;
> >       2 	tx_delay = <0x26>;
> >      14 	tx_delay = <0x28>;
> >       1 	tx_delay = <0x2a>;
> >       6 	tx_delay = <0x30>;
> >       1 	tx_delay = <0x3a>;
> >       3 	tx_delay = <0x3c>;
> >       2 	tx_delay = <0x42>;
> >       5 	tx_delay = <0x43>;
> >       2 	tx_delay = <0x44>;
> >       1 	tx_delay = <0x46>;
> >       6 	tx_delay = <0x4f>;
> > 
> > The numbers are all over the place, but it is clear that 0x28 is the
> > most popular for some reason. But since this delay values are magic,
> > i've no idea why.
> > 
> >      Andrew
> I do have more numbers in the Rockchip BSP kernel, and 0x35 is from
> one of them that shows less retry counts when testing 'iperf3' while
> testing with the batch of ODROID-M1. :)

FWIW, I had trouble with my Odroid-M1 whose outgoing ethernet bandwidth
seemed to be limited to an average around 100kB/s, when I bumped into
this thread.  I applied the patch and I can now download from that
Odroid at >80MB/s.  So:

    Tested-by: Stefan Monnier <monnier@iro.umontreal.ca>

Haven't tried the `rgmii-id` patch yet.


        Stefan
Re: [PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Heiko Stübner 1 month, 2 weeks ago 
Am Dienstag, 19. August 2025, 18:03:47 Mitteleuropäische Sommerzeit schrieb Andrew Lunn:
> On Tue, Aug 19, 2025 at 01:50:17PM +0900, Dongjin Kim wrote:
> > Very poor network performance is observed on the ODROID-M1 Board
> > (REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
> > this problem can be resolved after changing the value of TXD
> > timing to 0x35 from 0x4f.
> 
> How did you decide on 0x35?

just because I accidentially stumbled over that recently, Rockchip has
some scanning functionality in their vendor-kernel to get possible
delay values for a board.

Described in a document called "Rockchip GMAC RGMII Delayline Guide" [0]
and one code location is for example the armbian tree [1]

At least that is one possible source of per-board delay values :-)


Heiko


[0] https://lo01.g77k.com/aeb/docs/en/Common/GMAC/Rockchip_Developer_Guide_Linux_GMAC_RGMII_Delayline_EN.pdf
(random internet source Google found me)
[1] https://github.com/armbian/linux-rockchip/blob/rk-6.1-rkr5.1/drivers/net/ethernet/stmicro/stmmac/dwmac-rk-tool.c


> 
> ~/linux/arch/arm64$ grep -hr "tx_delay = " | sort | uniq -c
>       4 	tx_delay = <0x10>;
>       1 	tx_delay = <0x1a>;
>       1 	tx_delay = <0x21>;
>       2 	tx_delay = <0x22>;
>       4 	tx_delay = <0x24>;
>       2 	tx_delay = <0x26>;
>      14 	tx_delay = <0x28>;
>       1 	tx_delay = <0x2a>;
>       6 	tx_delay = <0x30>;
>       1 	tx_delay = <0x3a>;
>       3 	tx_delay = <0x3c>;
>       2 	tx_delay = <0x42>;
>       5 	tx_delay = <0x43>;
>       2 	tx_delay = <0x44>;
>       1 	tx_delay = <0x46>;
>       6 	tx_delay = <0x4f>;
> 
> The numbers are all over the place, but it is clear that 0x28 is the
> most popular for some reason. But since this delay values are magic,
> i've no idea why.
> 
>      Andrew
>
Re: [PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Krzysztof Kozlowski 1 month, 2 weeks ago 
On 19/08/2025 06:50, Dongjin Kim wrote:
> Very poor network performance is observed on the ODROID-M1 Board
> (REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
> this problem can be resolved after changing the value of TXD
> timing to 0x35 from 0x4f.
> 
> Signed-off-by: Dongjin Kim <tobetter@gmail.com>
> Change-Id: Idecd246e3fe93447b6e3e9b165d077d2050bbd75


Please run scripts/checkpatch.pl on the patches and fix reported
warnings. After that, run also 'scripts/checkpatch.pl --strict' on the
patches and (probably) fix more warnings. Some warnings can be ignored,
especially from --strict run, but the code here looks like it needs a
fix. Feel free to get in touch if the warning is not clear.

You CC-ed an address, which suggests you do not work on mainline kernel
or you do not use get_maintainers.pl/b4/patman. Please rebase and always
work on mainline or start using mentioned tools, so correct addresses
will be used.

Best regards,
Krzysztof
Re: [PATCH] arm64: dts: rockchip: Change ethernet TXD timing delay value
Posted by Dongjin Kim 1 month, 2 weeks ago 
On Tue, Aug 19, 2025 at 08:42:55AM +0200, Krzysztof Kozlowski wrote:
> On 19/08/2025 06:50, Dongjin Kim wrote:
> > Very poor network performance is observed on the ODROID-M1 Board
> > (REV 1.0 20220613) when running iperf3 (under 200Mbits/sec), and
> > this problem can be resolved after changing the value of TXD
> > timing to 0x35 from 0x4f.
> > 
> > Signed-off-by: Dongjin Kim <tobetter@gmail.com>
> > Change-Id: Idecd246e3fe93447b6e3e9b165d077d2050bbd75
> 
> 
> Please run scripts/checkpatch.pl on the patches and fix reported
> warnings. After that, run also 'scripts/checkpatch.pl --strict' on the
> patches and (probably) fix more warnings. Some warnings can be ignored,
> especially from --strict run, but the code here looks like it needs a
> fix. Feel free to get in touch if the warning is not clear.

Thanks. I realized that I sent unmodified file that includes
"Change-Id:".

> 
> You CC-ed an address, which suggests you do not work on mainline kernel
> or you do not use get_maintainers.pl/b4/patman. Please rebase and always
> work on mainline or start using mentioned tools, so correct addresses
> will be used.

I've rebased to the master branch, so let me send the patch again.

> 
> Best regards,
> Krzysztof

Patchew 2.1-devel-b67e4c2

© 2016 - 2025 Red Hat, Inc.