Il 01/08/2014 16:34, Stefan Hajnoczi ha scritto:
> On Wed, Jul 09, 2014 at 11:53:01AM +0200, Paolo Bonzini wrote:
>> diff --git a/async.c b/async.c
>> index 34af0b2..ac40eab 100644
>> --- a/async.c
>> +++ b/async.c
>> @@ -152,39 +152,43 @@ void qemu_bh_delete(QEMUBH *bh)
>> bh->deleted = 1;
>> }
>>
>> -static gboolean
>> -aio_ctx_prepare(GSource *source, gint *timeout)
>> +int
>> +aio_compute_timeout(AioContext *ctx)
>
> The return value is now nanoseconds so a 32-bit int doesn't offer much
> range (only 2 seconds for a signed int).
>
> Any reason to use int instead of int64_t as used by the timer API?
Can you squash the obvious
diff --git a/async.c b/async.c
index ac40eab..09e09c6 100644
--- a/async.c
+++ b/async.c
@@ -152,7 +152,7 @@ void qemu_bh_delete(QEMUBH *bh)
bh->deleted = 1;
}
-int
+int64_t
aio_compute_timeout(AioContext *ctx)
{
int64_t deadline;
diff --git a/include/block/aio.h b/include/block/aio.h
index 7eeb961..05b531c 100644
--- a/include/block/aio.h
+++ b/include/block/aio.h
@@ -309,6 +309,6 @@ static inline void aio_timer_init(AioContext *ctx,
*
* Compute the timeout that a blocking aio_poll should use.
*/
-int aio_compute_timeout(AioContext *ctx);
+int64_t aio_compute_timeout(AioContext *ctx);
#endif
yourself, or should I send v2 of this patch or the whole series?
Paolo
