Re: [PATCH] Avoid cpu_physical_memory_rw() with a constant is_write argument

[Stefan Weil]

Am 18.02.20 um 19:49 schrieb Peter Maydell:

> I think that we should fix the inconsistency where these functions
> all take "uint8_t* buf":
>
>  - address_space_rw()
>  - address_space_read()
>  - address_space_write()
>  - address_space_write_rom()
>  - cpu_physical_memory_rw()
>  - cpu_memory_rw_debug()
>
> but these take void*:
>  - cpu_physical_memory_read()
>  - cpu_physical_memory_write()
>  - address_space_write_cached()
>  - address_space_read_cached_slow()
>  - address_space_write_cached_slow()
>  - pci_dma_read()
>  - pci_dma_write()
>  - pci_dma_rw()
>  - dma_memory_read()
>  - dma_memory_write()
>  - dma_memory_rw()
>  - dma_memory_rw_relaxed()
>
> Depending on which way we go we would either want to remove these
> casts, or not.
>
> I guess that we have more cases of 'void*', and that would
> certainly be the easier way to convert (otherwise we probably
> need to add a bunch of new casts to uint8_t* in various callsites),
> but I don't have a strong opinion. Paolo ?
>
> thanks
> -- PMM


Indeed, fixing such inconsitencies would be good.

Personally I like the way how the standard C library handles such
pointers for functions like memcpy, fread, fwrite and others.

Therefore I suggest to use `const void *` and `void *` and to avoid type
casts.

Regards
Stefan