| Message ID | 20220413162739.6717-1-stefan.wahren@i2se.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RFC] ARM: dts: bcm2711: Fix DMA constrains for newer BCM2711 boards | expand |
Hi, Am 13.04.22 um 18:27 schrieb Stefan Wahren: > The commit 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own bus") > assumed that all bootloader pass the FDT modified by the firmware to the > kernel. But this is not always the case (e.g. Fedora) and cause boot > issues for boards with a BCM2711 C0 SoC (RPi 400, CM4, newer RPi 4 B) > which does have different DMA constraints. > > Since we are not able to detect the SoC revision, let's assume a BCM2711 > C0 SoC for bcm2711.dtsi and move the restricted DMA constrains to a > separate RPi 4 B board file just for the old SoC revision. > > Fixes: 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own bus") > Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com> since there wasn't any feedback yet, i want to ping ... > --- > arch/arm/boot/dts/Makefile | 1 + > arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts | 22 ++++++++++++++++++++++ > arch/arm/boot/dts/bcm2711.dtsi | 11 +++-------- > 3 files changed, 26 insertions(+), 8 deletions(-) > create mode 100644 arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts > > diff --git a/arch/arm/boot/dts/Makefile b/arch/arm/boot/dts/Makefile > index 7c16f8a2b738..560d36ba1041 100644 > --- a/arch/arm/boot/dts/Makefile > +++ b/arch/arm/boot/dts/Makefile > @@ -96,6 +96,7 @@ dtb-$(CONFIG_ARCH_BCM2835) += \ > bcm2837-rpi-zero-2-w.dtb \ > bcm2711-rpi-400.dtb \ > bcm2711-rpi-4-b.dtb \ > + bcm2711-rpi-4-b-b0.dtb \ > bcm2711-rpi-cm4-io.dtb \ > bcm2835-rpi-zero.dtb \ > bcm2835-rpi-zero-w.dtb > diff --git a/arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts b/arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts > new file mode 100644 > index 000000000000..c53ed510392f > --- /dev/null > +++ b/arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts > @@ -0,0 +1,22 @@ > +// SPDX-License-Identifier: GPL-2.0 > +/dts-v1/; > +#include "bcm2711-rpi-4-b.dts" > + > +&emmc2bus { > + /* > + * emmc2 on initial SoC revision (only assembled on RPi 4 B) > + * has different DMA constraints. > + */ > + dma-ranges = <0x0 0xc0000000 0x0 0x00000000 0x40000000>; > +}; > + > +&pcie0 { > + /* > + * The wrapper around the PCIe block has a bug > + * preventing it from accessing beyond the first 3GB of > + * memory. > + */ > + dma-ranges = <0x02000000 0x0 0x00000000 0x0 0x00000000 > + 0x0 0xc0000000>; > +}; > + > diff --git a/arch/arm/boot/dts/bcm2711.dtsi b/arch/arm/boot/dts/bcm2711.dtsi > index 89af57482bc8..c307c45385a0 100644 > --- a/arch/arm/boot/dts/bcm2711.dtsi > +++ b/arch/arm/boot/dts/bcm2711.dtsi > @@ -420,7 +420,7 @@ emmc2bus: emmc2bus { > #size-cells = <1>; > > ranges = <0x0 0x7e000000 0x0 0xfe000000 0x01800000>; > - dma-ranges = <0x0 0xc0000000 0x0 0x00000000 0x40000000>; > + dma-ranges = <0x0 0x00000000 0x0 0x00000000 0xfc000000>; > > emmc2: mmc@7e340000 { > compatible = "brcm,bcm2711-emmc2"; > @@ -574,13 +574,8 @@ IRQ_TYPE_LEVEL_HIGH>, > > ranges = <0x02000000 0x0 0xf8000000 0x6 0x00000000 > 0x0 0x04000000>; > - /* > - * The wrapper around the PCIe block has a bug > - * preventing it from accessing beyond the first 3GB of > - * memory. > - */ > - dma-ranges = <0x02000000 0x0 0x00000000 0x0 0x00000000 > - 0x0 0xc0000000>; > + dma-ranges = <0x02000000 0x4 0x00000000 0x0 0x00000000 > + 0x1 0x00000000>; > brcm,enable-ssc; > }; >
On 5/4/2022 2:43 PM, Stefan Wahren wrote: > Hi, > > Am 13.04.22 um 18:27 schrieb Stefan Wahren: >> The commit 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own >> bus") >> assumed that all bootloader pass the FDT modified by the firmware to the >> kernel. But this is not always the case (e.g. Fedora) and cause boot >> issues for boards with a BCM2711 C0 SoC (RPi 400, CM4, newer RPi 4 B) >> which does have different DMA constraints. >> >> Since we are not able to detect the SoC revision, let's assume a BCM2711 >> C0 SoC for bcm2711.dtsi and move the restricted DMA constrains to a >> separate RPi 4 B board file just for the old SoC revision. >> >> Fixes: 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own bus") >> Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com> > since there wasn't any feedback yet, i want to ping ... Could not we just rely on the VPU patching the Device Tree instead of having to differentiate the 2711 revision and having to use different DTS/DTBs? Also, can we consider that the older B0 are less predominant in the wild than the C0 nowadays, especially with the shortages going on is that even remotely the case?
Am 18.05.22 um 22:18 schrieb Florian Fainelli: > > > On 5/4/2022 2:43 PM, Stefan Wahren wrote: >> Hi, >> >> Am 13.04.22 um 18:27 schrieb Stefan Wahren: >>> The commit 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own >>> bus") >>> assumed that all bootloader pass the FDT modified by the firmware to >>> the >>> kernel. But this is not always the case (e.g. Fedora) and cause boot >>> issues for boards with a BCM2711 C0 SoC (RPi 400, CM4, newer RPi 4 B) >>> which does have different DMA constraints. >>> >>> Since we are not able to detect the SoC revision, let's assume a >>> BCM2711 >>> C0 SoC for bcm2711.dtsi and move the restricted DMA constrains to a >>> separate RPi 4 B board file just for the old SoC revision. >>> >>> Fixes: 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own bus") >>> Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com> >> since there wasn't any feedback yet, i want to ping ... > > Could not we just rely on the VPU patching the Device Tree instead of > having to differentiate the 2711 revision and having to use different > DTS/DTBs? As long as every bootloader in the boot chain take care everything is fine. The problem is the VPU firmware is closed source, so only a few people knows about the hardware differences. So the idea was to make it more transparent and yes i'm also not happy with this patch. > > Also, can we consider that the older B0 are less predominant in the > wild than the C0 nowadays, especially with the shortages going on is > that even remotely the case? So your idea is to make C0 the default and let the firmware patch the B0 part? I must confess that i never tested this.
diff --git a/arch/arm/boot/dts/Makefile b/arch/arm/boot/dts/Makefile index 7c16f8a2b738..560d36ba1041 100644 --- a/arch/arm/boot/dts/Makefile +++ b/arch/arm/boot/dts/Makefile @@ -96,6 +96,7 @@ dtb-$(CONFIG_ARCH_BCM2835) += \ bcm2837-rpi-zero-2-w.dtb \ bcm2711-rpi-400.dtb \ bcm2711-rpi-4-b.dtb \ + bcm2711-rpi-4-b-b0.dtb \ bcm2711-rpi-cm4-io.dtb \ bcm2835-rpi-zero.dtb \ bcm2835-rpi-zero-w.dtb diff --git a/arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts b/arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts new file mode 100644 index 000000000000..c53ed510392f --- /dev/null +++ b/arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts @@ -0,0 +1,22 @@ +// SPDX-License-Identifier: GPL-2.0 +/dts-v1/; +#include "bcm2711-rpi-4-b.dts" + +&emmc2bus { + /* + * emmc2 on initial SoC revision (only assembled on RPi 4 B) + * has different DMA constraints. + */ + dma-ranges = <0x0 0xc0000000 0x0 0x00000000 0x40000000>; +}; + +&pcie0 { + /* + * The wrapper around the PCIe block has a bug + * preventing it from accessing beyond the first 3GB of + * memory. + */ + dma-ranges = <0x02000000 0x0 0x00000000 0x0 0x00000000 + 0x0 0xc0000000>; +}; + diff --git a/arch/arm/boot/dts/bcm2711.dtsi b/arch/arm/boot/dts/bcm2711.dtsi index 89af57482bc8..c307c45385a0 100644 --- a/arch/arm/boot/dts/bcm2711.dtsi +++ b/arch/arm/boot/dts/bcm2711.dtsi @@ -420,7 +420,7 @@ emmc2bus: emmc2bus { #size-cells = <1>; ranges = <0x0 0x7e000000 0x0 0xfe000000 0x01800000>; - dma-ranges = <0x0 0xc0000000 0x0 0x00000000 0x40000000>; + dma-ranges = <0x0 0x00000000 0x0 0x00000000 0xfc000000>; emmc2: mmc@7e340000 { compatible = "brcm,bcm2711-emmc2"; @@ -574,13 +574,8 @@ IRQ_TYPE_LEVEL_HIGH>, ranges = <0x02000000 0x0 0xf8000000 0x6 0x00000000 0x0 0x04000000>; - /* - * The wrapper around the PCIe block has a bug - * preventing it from accessing beyond the first 3GB of - * memory. - */ - dma-ranges = <0x02000000 0x0 0x00000000 0x0 0x00000000 - 0x0 0xc0000000>; + dma-ranges = <0x02000000 0x4 0x00000000 0x0 0x00000000 + 0x1 0x00000000>; brcm,enable-ssc; };
The commit 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own bus") assumed that all bootloader pass the FDT modified by the firmware to the kernel. But this is not always the case (e.g. Fedora) and cause boot issues for boards with a BCM2711 C0 SoC (RPi 400, CM4, newer RPi 4 B) which does have different DMA constraints. Since we are not able to detect the SoC revision, let's assume a BCM2711 C0 SoC for bcm2711.dtsi and move the restricted DMA constrains to a separate RPi 4 B board file just for the old SoC revision. Fixes: 3d2cbb644836 ("ARM: dts: bcm2711: Move emmc2 into its own bus") Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com> --- arch/arm/boot/dts/Makefile | 1 + arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts | 22 ++++++++++++++++++++++ arch/arm/boot/dts/bcm2711.dtsi | 11 +++-------- 3 files changed, 26 insertions(+), 8 deletions(-) create mode 100644 arch/arm/boot/dts/bcm2711-rpi-4-b-b0.dts