Re: [PATCH 2/6] memory: Introduce address_space_create()

[Philippe Mathieu-Daudé]

On 8/19/21 4:24 PM, Peter Maydell wrote:
> On Thu, 19 Aug 2021 at 15:20, Philippe Mathieu-Daudé <philmd@redhat.com> 
> wrote:
>>
>> Introduce address_space_create(). In is similar to
>> address_space_init() but returns a pointer to a heap
>> allocated  AddressSpace.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>> ---
>>  include/exec/memory.h | 14 ++++++++++++++
>>  softmmu/memory.c      | 10 ++++++++++
>>  2 files changed, 24 insertions(+)
>>
>> diff --git a/include/exec/memory.h b/include/exec/memory.h
>> index c3d417d317f..b353a48c25f 100644
>> --- a/include/exec/memory.h
>> +++ b/include/exec/memory.h
>> @@ -2418,6 +2418,20 @@ MemTxResult memory_region_dispatch_write(MemoryRegion 
>> *mr,
>>   */
>>  void address_space_init(AddressSpace *as, MemoryRegion *root, const char 
>> *name);
>>
>> +/**
>> + * address_space_create: Create and initializes an address space
>> + *
>> + * @root: a #MemoryRegion that routes addresses for the address space
>> + * @name: an address space name.  The name is only used for debugging
>> + *        output.
>> + *
>> + * Returns pointer to initialized #AddressSpace.
>> + *
>> + * The caller is responsible for releasing the pointer returned
>> + * with address_space_destroy() after use.
>> + */
>> +AddressSpace *address_space_create(MemoryRegion *root, const char *name);
>> +
> 
> I'm not really a fan of this as an API -- almost always I think
> devices would do better to have an AddressSpace foo field in
> their device struct and call address_space_init() on that.
> Hiding the heap allocation inside this function makes it harder
> to notice it during code review, I think.

So I understand you rather I discard this (simple) approach and
rather modify 'info mtree' "was designed on the assumption that
there's really only one or two interesting address spaces." [*]

[*] https://www.mail-archive.com/qemu-devel@nongnu.org/msg829821.html