net: macb: disable scatter-gather for macb on sama5d3
PATCH 1/1 net: macb: disable scatter-gather for macb on sama5d3
Create a new configuration for the sama5d3-macb new compatibility string. This configuration disables scatter-gather because we experienced lock down of the macb interface of this particular SoC under very high load.
--- drivers/net/ethernet/cadence/macb_main.c | 8 ++++++++ 1 file changed, 8 insertions(+)
drivers/net/ethernet/cadence/macb_main.c +8 -0
@@ -3837,6 +3837,13 @@ static const struct macb_config at91sam9260_config = { .init = macb_init, }; +static const struct macb_config sama5d3macb_config = { + .caps =MACB_CAPS_SG_DISABLED + |MACB_CAPS_USRIO_HAS_CLKEN |MACB_CAPS_USRIO_DEFAULT_IS_MII_GMII , + .clk_init = macb_clk_init, + .init = macb_init, +}; +static const struct macb_config pc302gem_config = { .caps =MACB_CAPS_SG_DISABLED |MACB_CAPS_GIGABIT_MODE_AVAILABLE , .dma_burst_length =16 ,
@@ -3904,6 +3911,7 @@ static const struct of_device_id macb_dt_ids[] = { { .compatible = "cdns,gem", .data = &pc302gem_config }, { .compatible = "atmel,sama5d2-gem", .data = &sama5d2_config }, { .compatible = "atmel,sama5d3-gem", .data = &sama5d3_config }, + { .compatible = "atmel,sama5d3-macb", .data = &sama5d3macb_config }, { .compatible = "atmel,sama5d4-gem", .data = &sama5d4_config }, { .compatible = "cdns,at91rm9200-emac", .data = &emac_config }, { .compatible = "cdns,emac", .data = &emac_config },
Signature · 1 line
2.15.1
PATCH 1/1 ARM: dts: at91: add new compatibility string for macb on sama5d3
We need this new compatibility string as we experienced different behavior for this 10/100Mbits/s macb interface on this particular SoC. Backward compatibility is preserved as we keep the alternative strings.
--- Documentation/devicetree/bindings/net/macb.txt | 1 + arch/arm/boot/dts/sama5d3_emac.dtsi | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-)
Documentation/devicetree/bindings/net/macb.txt +1 -0
@@ -10,6 +10,7 @@ Required properties: Use "cdns,pc302-gem" for Picochip picoXcell pc302 and later devices based on the Cadence GEM, or the generic form: "cdns,gem". Use "atmel,sama5d2-gem" for the GEM IP (10/100) available on Atmel sama5d2 SoCs. + Use "atmel,sama5d3-macb" for the 10/100Mbit IP available on Atmel sama5d3 SoCs. Use "atmel,sama5d3-gem" for the Gigabit IP available on Atmel sama5d3 SoCs. Use "atmel,sama5d4-gem" for the GEM IP (10/100) available on Atmel sama5d4 SoCs. Use "cdns,zynq-gem" Xilinx Zynq-7xxx SoC.
arch/arm/boot/dts/sama5d3_emac.dtsi +1 -1
@@ -41,7 +41,7 @@ }; macb1: ethernet at f802c000 { - compatible = "cdns,at91sam9260-macb", "cdns,macb"; + compatible = "atmel,sama5d3-macb", "cdns,at91sam9260-macb", "cdns,macb"; reg = <0xf802c000 0x100>; interrupts = <35 IRQ_TYPE_LEVEL_HIGH 3>; pinctrl-names = "default";
Signature · 1 line
2.15.1
Re: [PATCH] ARM: dts: at91: add new compatibility string for macb on sama5d3
From: Nicolas Ferre <[email protected]>
Date: Fri, 14 Sep 2018 17:48:11 +0200
Quoted 5 lines
We need this new compatibility string as we experienced different behavior for this 10/100Mbits/s macb interface on this particular SoC. Backward compatibility is preserved as we keep the alternative strings.
Signed-off-by Nicolas Ferre <[email protected]>
Applied.
From: Nicolas Ferre <[email protected]>
Date: Fri, 14 Sep 2018 17:48:10 +0200
Quoted 5 lines
Create a new configuration for the sama5d3-macb new compatibility string. This configuration disables scatter-gather because we experienced lock down of the macb interface of this particular SoC under very high load.
Signed-off-by Nicolas Ferre <[email protected]>
Applied, but are you really sure it isn't just some bug in the driver or somesuch?
Quoted 11 lines · David Miller
From: Nicolas Ferre <[email protected]>
Date: Fri, 14 Sep 2018 17:48:10 +0200Quoted 5 lines
Create a new configuration for the sama5d3-macb new compatibility string. This configuration disables scatter-gather because we experienced lock down of the macb interface of this particular SoC under very high load.
Signed-off-by Nicolas Ferre <[email protected]>Applied, but are you really sure it isn't just some bug in the driver or somesuch?
Actually, the same IP on other SoCs behaves correctly with same driver. We tried hard to find the root cause of the problem without success and we didn't have confirmation of an issue on Cadence's side either.
The symptom is that we have a receive overrun on this interface under heavy load but I think that it is more a result of the issue than the real problem: I cannot escape from this overrun condition: the IP itself is completely blocked.
Even if the problem is very difficult to reproduce (simple *perf tools not sufficient), we play safe by simply disabling this feature on this SoC only.
Best regards,
Signature · 1 line
Nicolas Ferre