Re: [PATCH 3/5] block: reduce files included by block/aio.h

["Philippe Mathieu-Daudé" <philmd@linaro.org>]

On 4/12/25 10:25, Paolo Bonzini wrote:
> On 12/4/25 09:35, Philippe Mathieu-Daudé wrote:
>> On 3/12/25 19:51, Paolo Bonzini wrote:
>>> Avoid including all of qdev everywhere (the hw/core/qdev.h header in 
>>> fact
>>> brings in a lot more headers too), instead declare a couple structs for
>>> which only a pointer type is needed.
>>>
>>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>>> ---
>>>   include/block/aio.h                |  7 +++----
>>>   include/block/block-copy.h         |  1 +
>>>   include/block/block-global-state.h |  1 +
>>>   include/block/block-io.h           |  1 +
>>>   include/block/dirty-bitmap.h       |  1 +
>>>   include/hw/core/qdev.h             |  6 +-----
>>>   include/qemu/job.h                 |  1 +
>>>   include/qemu/main-loop.h           |  2 +-
>>>   include/qemu/mem-reentrancy.h      | 10 ++++++++++
>>>   net/net.c                          |  1 +
>>>   util/async.c                       |  1 +
>>>   11 files changed, 22 insertions(+), 10 deletions(-)
>>>   create mode 100644 include/qemu/mem-reentrancy.h
>>>
>>> diff --git a/include/block/aio.h b/include/block/aio.h
>>> index 59c56d695ee..8cca2360d1a 100644
>>> --- a/include/block/aio.h
>>> +++ b/include/block/aio.h
>>> @@ -23,9 +23,8 @@
>>>   #include "qemu/lockcnt.h"
>>>   #include "qemu/thread.h"
>>>   #include "qemu/timer.h"
>>> -#include "block/graph-lock.h"
>>> -#include "hw/core/qdev.h"
>>> +struct MemReentrancyGuard;
>>>   typedef struct AioHandler AioHandler;
>>>   typedef QLIST_HEAD(, AioHandler) AioHandlerList;
>>> @@ -211,7 +210,7 @@ struct AioContext {
>>>        * of nodes and edges from block graph while some
>>>        * other thread is traversing it.
>>>        */
>>> -    BdrvGraphRWlock *bdrv_graph;
>>> +    struct BdrvGraphRWlock *bdrv_graph;
>>>       /* The list of registered AIO handlers.  Protected by ctx- 
>>> >list_lock. */
>>>       AioHandlerList aio_handlers;
>>> @@ -393,7 +392,7 @@ void aio_bh_schedule_oneshot_full(AioContext 
>>> *ctx, QEMUBHFunc *cb, void *opaque,
>>>    * device-reentrancy issues
>>>    */
>>>   QEMUBH *aio_bh_new_full(AioContext *ctx, QEMUBHFunc *cb, void *opaque,
>>> -                        const char *name, MemReentrancyGuard 
>>> *reentrancy_guard);
>>> +                        const char *name, struct MemReentrancyGuard 
>>> *reentrancy_guard);
>>
>> I suppose these changes were before you move the declarations to their
>> own "qemu/mem-reentrancy.h" header?
> 
> Yes, but also qemu/mem-reentrancy.h is not included here.

We want to also include it instead, right? (No need to forward declare)