Re: [Qemu-devel] [PATCH 3/6] sheepdog: use coroutine based socket functions in coroutine context

[Kevin Wolf <kwolf@redhat.com>]

Am 04.07.2012 02:25, schrieb Peter Crosthwaite:
> On Tue, Jul 3, 2012 at 11:15 PM, Kevin Wolf <kwolf@redhat.com> wrote:
>> Am 27.06.2012 00:26, schrieb MORITA Kazutaka:
>>> This removes blocking network I/Os in coroutine context.
>>>
>>> Signed-off-by: MORITA Kazutaka <morita.kazutaka@lab.ntt.co.jp>
>>> ---
>>>  block/sheepdog.c |   10 ++++++++--
>>>  1 files changed, 8 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/block/sheepdog.c b/block/sheepdog.c
>>> index 0b49c6d..5dc1d7a 100644
>>> --- a/block/sheepdog.c
>>> +++ b/block/sheepdog.c
>>> @@ -541,11 +541,18 @@ static coroutine_fn int send_co_req(int sockfd, SheepdogReq *hdr, void *data,
>>>      return ret;
>>>  }
>>>
>>> +static coroutine_fn int do_co_req(int sockfd, SheepdogReq *hdr, void *data,
>>> +                                  unsigned int *wlen, unsigned int *rlen);
>>> +
>>>  static int do_req(int sockfd, SheepdogReq *hdr, void *data,
>>>                    unsigned int *wlen, unsigned int *rlen)
>>>  {
>>>      int ret;
>>>
>>> +    if (qemu_in_coroutine()) {
>>> +        return do_co_req(sockfd, hdr, data, wlen, rlen);
>>> +    }
>>> +
>>>      socket_set_block(sockfd);
>>>      ret = send_req(sockfd, hdr, data, wlen);
>>>      if (ret < 0) {
>>
>> How about replacing the non-coroutine implementation by code that
>> creates a new coroutine and executes do_co_req() as well? This would
>> reduce some code duplication.
>>
> 
> +1. I presume it can it be done such that there is no if
> (qemu_in_coroutine()) logic that way?

That was not my intention, and actually I don't think it would help your
case here because this is a function truly internal to the block layer
(or actually even a single block driver).

Kevin