diff mbox series

ARM: dts: meson8b: add ethernet fifo sizes

Message ID 20190718093623.23598-1-jbrunet@baylibre.com (mailing list archive)
State Mainlined
Commit 4f0303d43973a4086d6355db0c653a5a369b5dbe
Headers show
Series ARM: dts: meson8b: add ethernet fifo sizes | expand

Commit Message

Jerome Brunet July 18, 2019, 9:36 a.m. UTC
If unspecified in DT, the fifo sizes are not automatically detected by
the dwmac1000 dma driver and the reported fifo sizes default to 0.
Because of this, flow control will be turned off on the device.

Add the fifo sizes provided by the datasheet in the SoC in DT so
flow control may be enabled if necessary.

Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
 arch/arm/boot/dts/meson8b.dtsi | 2 ++
 1 file changed, 2 insertions(+)

Comments

Martin Blumenstingl July 23, 2019, 7:56 p.m. UTC | #1
On Thu, Jul 18, 2019 at 11:36 AM Jerome Brunet <jbrunet@baylibre.com> wrote:
>
> If unspecified in DT, the fifo sizes are not automatically detected by
> the dwmac1000 dma driver and the reported fifo sizes default to 0.
> Because of this, flow control will be turned off on the device.
>
> Add the fifo sizes provided by the datasheet in the SoC in DT so
> flow control may be enabled if necessary.
>
> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>

I wonder if this "fixes" some of the performance issues on Odroid-C1
testing this is now on my TODO-list for the weekend
Aymeric Aug. 4, 2019, 7:38 p.m. UTC | #2
Le 23/07/2019 à 21:56, Martin Blumenstingl a écrit :
> On Thu, Jul 18, 2019 at 11:36 AM Jerome Brunet <jbrunet@baylibre.com> wrote:
>> If unspecified in DT, the fifo sizes are not automatically detected by
>> the dwmac1000 dma driver and the reported fifo sizes default to 0.
>> Because of this, flow control will be turned off on the device.
>>
>> Add the fifo sizes provided by the datasheet in the SoC in DT so
>> flow control may be enabled if necessary.
>>
>> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
> Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
>
> I wonder if this "fixes" some of the performance issues on Odroid-C1
> testing this is now on my TODO-list for the weekend


Good! and maybe that will fix the stability issue I'm seeing when using
the board with a 1 gigabit link! (cf
http://lists.infradead.org/pipermail/linux-amlogic/2019-June/012341.html)

I'll try that asap on my board too :)

Thanks.
Kevin Hilman Aug. 5, 2019, 8:57 p.m. UTC | #3
Jerome Brunet <jbrunet@baylibre.com> writes:

> If unspecified in DT, the fifo sizes are not automatically detected by
> the dwmac1000 dma driver and the reported fifo sizes default to 0.
> Because of this, flow control will be turned off on the device.
>
> Add the fifo sizes provided by the datasheet in the SoC in DT so
> flow control may be enabled if necessary.
>
> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>

Queued for v5.4,

Thanks,

Kevin
diff mbox series

Patch

diff --git a/arch/arm/boot/dts/meson8b.dtsi b/arch/arm/boot/dts/meson8b.dtsi
index fba2c70c2fda..b044ad78b475 100644
--- a/arch/arm/boot/dts/meson8b.dtsi
+++ b/arch/arm/boot/dts/meson8b.dtsi
@@ -410,6 +410,8 @@ 
 		 <&clkc CLKID_MPLL2>,
 		 <&clkc CLKID_MPLL2>;
 	clock-names = "stmmaceth", "clkin0", "clkin1";
+	rx-fifo-depth = <4096>;
+	tx-fifo-depth = <2048>;
 
 	resets = <&reset RESET_ETHERNET>;
 	reset-names = "stmmaceth";