On 04.04.2024 03:12, Stefano Stabellini wrote:
> --- a/docs/misra/C-language-toolchain.rst
> +++ b/docs/misra/C-language-toolchain.rst
> @@ -480,4 +480,73 @@ The table columns are as follows:
> - See Section "4.13 Preprocessing Directives" of GCC_MANUAL and Section
> "11.1 Implementation-defined behavior" of CPP_MANUAL.
>
>
> +Sizes of Integer types
> +______________________
> +
> +Xen expects System V ABI on x86_64:
> + https://gitlab.com/x86-psABIs/x86-64-ABI
> +
> +Xen expects AAPCS32 on ARMv8-A AArch32:
> + https://github.com/ARM-software/abi-aa/blob/main/aapcs32/aapcs32.rst
> +
> +Xen expects AAPCS64 LP64 on ARMv8-A AArch64:
> + https://github.com/ARM-software/abi-aa/blob/main/aapcs64/aapcs64.rst
> +
> +A summary table of data types, sizes and alignment is below:
> +
> +.. list-table::
> + :widths: 10 10 10 45
> + :header-rows: 1
> +
> + * - Type
> + - Size
> + - Alignment
> + - Architectures
> +
> + * - char
> + - 8 bits
> + - 8 bits
> + - all architectures
This one _may_ be acceptable, but already feels like going too far.
> + * - short
> + - 16 bits
> + - 16 bits
> + - all architectures
> +
> + * - int
> + - 32 bits
> + - 32 bits
> + - all architectures
These two I continue to disagree with. The values are minimum required ones.
Even if code changes may be needed (Misra already helps us here by stopping
undue mixing of e.g. unsigned int and uint32_t in at least some situations),
there's no inherent requirement in Xen for such restrictions.
> + * - long
> + - 32 bits
> + - 32 bits
> + - 32-bit architectures (x86_32, ARMv8-A AArch32, ARMv8-R AArch32)
> +
> + * - long
> + - 64 bits
> + - 64 bits
> + - 64-bit architectures (x86_64, ARMv8-A AArch64, RV64, PPC64)
> +
> + * - long long
> + - 64-bit
> + - 32-bit
> + - x86_32
> +
> + * - long long
> + - 64-bit
> + - 64-bit
> + - 64-bit architectures, ARMv8-A AArch32, ARMv8-R AArch32
Along the lines of the above, simply saying "64-bit architectures" here
is too generic. Whereas for long (above) and ...
> + * - pointer
> + - 32-bit
> + - 32-bit
> + - 32-bit architectures (x86_32, ARMv8-A AArch32, ARMv8-R AArch32)
> +
> + * - pointer
> + - 64-bit
> + - 64-bit
> + - 64-bit architectures (x86_64, ARMv8-A AArch64, RV64, PPC64)
... pointers I agree (and the special mentioning of the architectures
in parentheses could even be omitted imo). To summarize, my counter
proposal:
* - char
- at least 8 bits
- equaling size
- all architectures
* - char
- 8 bits
- 8 bits
- x86, ARM, RISC-V, PPC
* - short
- at least 16 bits
- equaling size
- all architectures
* - short
- 16 bits
- 16 bits
- x86, ARM, RISC-V, PPC
* - int
- at least 32 bits
- equaling size
- all architectures
* - int
- 32 bits
- 32 bits
- x86, ARM, RISC-V, PPC
* - long
- 32 bits
- 32 bits
- 32-bit architectures
* - long
- 64 bits
- 64 bits
- 64-bit architectures
* - long long
- 64-bit
- 32-bit
- x86_32
* - long long
- 64-bit
- 64-bit
- x86_64, ARMv8-A AArch64, RV64, PPC64, ARMv8-A AArch32, ARMv8-R AArch32
* - pointer
- 32-bit
- 32-bit
- 32-bit architectures
* - pointer
- 64-bit
- 64-bit
- 64-bit architectures
Eventually, by properly decoupling pointers from longs (via using {,u}intptr_t
appropriately), the restrictions on "long" could also be lifted.
Note that the generic requirements on char and short also are imposed by C99.
It may therefore not be necessary to state them explicitly, but rather refer
to that standard (just like you're now referencing the SysV psABI-s).
Jan
