Re: [PATCH 0/4] hw/block/pflash_cfi01: Remove pflash_cfi01_get_memory()

[Peter Maydell]

On Tue, 7 Sept 2021 at 15:45, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
> The problems I see:
>
> - pflash_cfi01_get_memory() doesn't really document what it returns,
>   simply an internal MemoryRegion* in pflash device. Neither we
>   document this is a ROMD device providing a RAM buffer initialized
>   by qemu_ram_alloc().
>
> - to update the flash content, we get the internal buffer via
>   memory_region_get_ram_ptr(). If the pflash implementation is
>   changed (.i.e. reworked to expose a MR container) we break
>   everything.
>
> - memory_region_get_ram_ptr() doesn't do any check on the MR type,
>   it simply calls qemu_map_ram_ptr(mr->ram_block, offset).

Using memory_region_get_ram_ptr() is tricky to get right, too --
if you're writing directly to the underlying ram while the system is
running you need to use memory_region_flush_rom_device() to make
sure it's marked dirty. I think the current users of this on the
pflash devices get away with it because they do it during initial
machine init.

-- PMM