Re: [Qemu-devel] [PATCH] block: fix bdrv_ioctl called from coroutine

[Paolo Bonzini <pbonzini@redhat.com>]

On 17/12/2015 01:59, Fam Zheng wrote:
> On Wed, 12/16 19:33, Paolo Bonzini wrote:
>> When called from a coroutine, bdrv_ioctl must be asynchronous just like
>> e.g. bdrv_flush.  The code was incorrectly making it synchronous, fix
>> it.
>>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>> ---
>>         Fam, any reason why you did it this way?  I don't see
>>         any coroutine caller, but it doesn't make much sense. :)
> 
> That is a surprising question!  From a coroutine, it is bdrv_flush ->
> bdrv_flush_co_entry -> bdrv_co_flush, which I think is always synchronous,
> especially, noticing the code around calling bs->bdrv_aio_flush:
> 
>         acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co);
>         if (acb == NULL) {
>             ret = -EIO;
>         } else {
>             qemu_coroutine_yield();
>             ret = co.ret;
>         }
> 
> Am I missing something?

In the coroutine case, the yield is hidden in the drivers, and it may or
may not happen.  For example, qcow2_co_flush_to_os starts with

    qemu_co_mutex_lock(&s->lock);

which can yield.

Paolo

> Fam
> 
>>
>>  block/io.c | 7 ++++---
>>  1 file changed, 4 insertions(+), 3 deletions(-)
>>
>> diff --git a/block/io.c b/block/io.c
>> index e00fb5d..841f5b5 100644
>> --- a/block/io.c
>> +++ b/block/io.c
>> @@ -2614,10 +2614,11 @@ int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
>>          bdrv_co_ioctl_entry(&data);
>>      } else {
>>          Coroutine *co = qemu_coroutine_create(bdrv_co_ioctl_entry);
>> +
>>          qemu_coroutine_enter(co, &data);
>> -    }
>> -    while (data.ret == -EINPROGRESS) {
>> -        aio_poll(bdrv_get_aio_context(bs), true);
>> +        while (data.ret == -EINPROGRESS) {
>> +            aio_poll(bdrv_get_aio_context(bs), true);
>> +        }
>>      }
>>      return data.ret;
>>  }
>> -- 
>> 2.5.0
>>
> 
>