This commit adds the `andestech,andescore-v5` compatible string. This helps
identify the core tpye form the final devicetree alone.
Andes doesn't define which core type from the v5 series the AE350 SoC uses,
so we're using the whole series name here.
Signed-off-by: Filip Kokosinski <fkokosinski@antmicro.com>
The `timebase-frequency` is not defined by any of the YAML binding files.
There was a discussion in #37420 to add this property, but in the end it
was rejected. This resulted in the #37685 feature request being created.
As of now, this property is not documented anywhere so this commit removes
it from the RISC-V devicetrees, as RISC-V is the only architecture that is
currently defining it - and even in RISC-V not all platforms do that.
Signed-off-by: Filip Kokosinski <fkokosinski@antmicro.com>
Define the edge-trigger register base address based on whether
the PLIC node in the devicetree has an additional compatible
that supports edge-triggered interrupt.
Limited the implementation to Andes NCEPLIC100 only, updated
the devicetree binding of `andes_v5_ae350` accordingly.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Define all the register offset directly in the driver according
to the RISCV PLIC specification as they are not configurable,
see: https://github.com/riscv/riscv-plic-spec.
Updated devicetrees that has PLIC accordingly.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
This commit adds/modifies `riscv,isa` strings using the following rules:
* the ISA string is lowercase
* multi-letter extensions are preceded with the underscore mark
* if an extension is implied by another one, it is not specified - e.g. the
D extension implies the F extension, so writing `rv32ifd` is redundant
Signed-off-by: Filip Kokosinski <fkokosinski@antmicro.com>
Add prescaler property to prevent counter driver imprecise when CPU clock
is close to the PIT clock.
Signed-off-by: Jimmy Zheng <jimmyzhe@andestech.com>
This mirrors #36499 and other PRs that added them for other
architectures.
This silences a large number of dtc warnings due to the missing
property. It seems reasonable to require an address-cells property since
any interrupt controller could be the parent of an interrupt-map.
The only device actually using interrupt-maps is neorv32, and it needs
an address-cells of 2 (since this is the default if none is specified it
worked like that before this change).
While I touched this, I reordered the properties for consistency across
boards, but there's a lot of variance here already.
Signed-off-by: Olof Johansson <olof@lixom.net>