Re: [Qemu-devel] [RFC PATCH v2 2/4] net: resolve race of tap backend and its peer

[liu ping fan]

On Thu, Mar 28, 2013 at 10:34 PM, Stefan Hajnoczi <address@hidden> wrote:
> On Thu, Mar 28, 2013 at 03:55:53PM +0800, Liu Ping Fan wrote:
>> From: Liu Ping Fan <address@hidden>
>>
>> Signed-off-by: Liu Ping Fan <address@hidden>
>> ---
>>  net/tap.c |    7 ++++++-
>>  1 files changed, 6 insertions(+), 1 deletions(-)
>>
>> diff --git a/net/tap.c b/net/tap.c
>> index 0b663d1..401161c 100644
>> --- a/net/tap.c
>> +++ b/net/tap.c
>> @@ -321,9 +321,14 @@ static void tap_cleanup(NetClientState *nc)
>>  static void tap_poll(NetClientState *nc, bool enable)
>>  {
>>      TAPState *s = DO_UPCAST(TAPState, nc, nc);
>> -    /* fixme, when tap backend on another thread, the disable should be 
>> sync */
>> +
>>      tap_read_poll(s, enable);
>>      tap_write_poll(s, enable);
>> +
>> +    if (!enable) {
>> +        /* need sync so vhost can take over polling */
>> +        g_source_remove_poll(&nc->nsrc->source, &nc->nsrc->gfd);
>> +    }
>
> Is this necessary?
>
> Since we're in tap_poll() we are currently not in poll(2) waiting on the
> fd.  Therefore it's safe for the caller to use the fd, the prepare
> function will remove it from the fd set before glib calls poll(2) again.
>
The main purpose is to make sure that the GSource->dispatch is not
in-flight, in which case, both vhost handler and ->dispatch() will
work on it.

Regards
Pingfan

> Stefan