[0/9] TQMLS1021A support

Hi,

this adds the DT files for the TQMLS1021A using the MBLS1021A mainboard.
This includes general board support and several overlays enabling various
graphic devices. RCW and hardware configuration using dip switches is not
needed, the default DT can be used. Just video output needs specific
overlays.

The last patch enabling LPAE in defconfig is just for demonstration what's
needed for PCIe support. There was a patchset ([1]) supporting LPAE on
armv7 defconfig, but there was no feedback.

[1] https://lore.kernel.org/linux-arm-kernel/20200203184820.4433-2-nsaenzjulienne@suse.de/T/

Alexander Stein (8):
  ARM: dts: ls1021a: add TQ-Systems MBLS102xA device tree
  ARM: dts: ls1021a: add TQMLS1021A flash partition layout
  ARM: dts: ls1021a: add TQMLS1021A/MBLS102xA LVDS TM070JVHG33 overlay
  ARM: dts: ls1021a: add TQMLS1021A/MBLS102xA HDMI overlay
  ARM: dts: ls1021a: add TQMLS1021A/MBLS102xA LVDS CDTECH DC44 overlay
  ARM: dts: ls1021a: add TQMLS1021A/MBLS102xA LVDS CDTECH FC21 overlay
  ARM: multi_v7_defconfig: Add options to support TQMLS102xA series
  [DNI] ARM: multi_v7_defconfig: Enable CONFIG_ARM_LPAE for
    multi_v7_config

Matthias Schiffer (1):
  dt-bindings: arm: fsl: add TQ-Systems LS1021A board

 .../devicetree/bindings/arm/fsl.yaml          |   7 +
 arch/arm/boot/dts/Makefile                    |  12 +
 ...1021a-tqmls1021a-mbls1021a-cdtech-dc44.dts |  55 +++
 ...1021a-tqmls1021a-mbls1021a-cdtech-fc21.dts |  55 +++
 .../ls1021a-tqmls1021a-mbls1021a-hdmi.dtso    |  36 ++
 ...tqmls1021a-mbls1021a-lvds-tm070jvhg33.dtso |  56 +++
 .../boot/dts/ls1021a-tqmls1021a-mbls1021a.dts | 409 ++++++++++++++++++
 arch/arm/boot/dts/ls1021a-tqmls1021a.dtsi     | 124 ++++++
 arch/arm/configs/multi_v7_defconfig           |  14 +
 9 files changed, 768 insertions(+)
 create mode 100644 arch/arm/boot/dts/ls1021a-tqmls1021a-mbls1021a-cdtech-dc44.dts
 create mode 100644 arch/arm/boot/dts/ls1021a-tqmls1021a-mbls1021a-cdtech-fc21.dts
 create mode 100644 arch/arm/boot/dts/ls1021a-tqmls1021a-mbls1021a-hdmi.dtso
 create mode 100644 arch/arm/boot/dts/ls1021a-tqmls1021a-mbls1021a-lvds-tm070jvhg33.dtso
 create mode 100644 arch/arm/boot/dts/ls1021a-tqmls1021a-mbls1021a.dts
 create mode 100644 arch/arm/boot/dts/ls1021a-tqmls1021a.dtsi

On 20/01/2023 15:57, Russell King (Oracle) wrote:
> On Fri, Jan 20, 2023 at 12:35:41PM +0100, Krzysztof Kozlowski wrote:
>> Defconfig=y is only for critical stuff needed to bring initramfs. Time
>> is not needed for initramfs.
> 
> Are I2C drivers "critical stuff"? What about AHCI drivers? What about
> all the USB HCI drivers? USB Storage can certainly be loaded from an
> initramfs, yet it's =y. LED triggers? Loads of RTC drivers that are
> marked as =y? XOR drivers? VIRTIO drivers? Filesystems? I seem to
> remember Red Hat sets even popular filesystems to be a module in their
> kernel.

Almost all your answers above - these should be =m. USB not always,
because it is used for USB Ethernet adapters which is used for network
boot (although even then initramfs comes with actual drivers...)

> 
> Clearly, "defconfig=y is only for critical stuff" is rather false in
> practice, and I suspect is little more than a desire rather than a
> rule. I suspect it's more "defconfig=y is for stuff that makes most
> platforms successfully boot without an initramfs".

Yes, that's true, although there was already pushback against this.
Having an initramfs is not a big deal and nowadays I think all new ARMv8
platforms use it.

Anyway this talk here is about RTC which is not needed at all as built in.

Best regards,
Krzysztof

On Sat, Jan 21, 2023 at 09:27:52PM +0100, Krzysztof Kozlowski wrote:
> On 20/01/2023 15:57, Russell King (Oracle) wrote:
> > Clearly, "defconfig=y is only for critical stuff" is rather false in
> > practice, and I suspect is little more than a desire rather than a
> > rule. I suspect it's more "defconfig=y is for stuff that makes most
> > platforms successfully boot without an initramfs".
> 
> Yes, that's true, although there was already pushback against this.
> Having an initramfs is not a big deal and nowadays I think all new ARMv8
> platforms use it.
> 
> Anyway this talk here is about RTC which is not needed at all as built in.

Provided NFS doesn't repeat XIDs at boot (which it used to when the
kernel had no idea what the TOD is.)

I believe the NFS standard allows NFS servers to repeat their response
to a repeated XID even if the request is different... which used to
cause chaos for me with a machine that has no RTC and boots using NFS
root. So just be aware that if the NFS XID is still dependent on TOD,
then not having RTCs loaded can cause chaos.