Re: [dpdk-dev] [PATCH] vhost: fix race condition in fdset_add

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 12/14/18 10:32 AM, Matthias Gatto wrote:
> On Tue, Dec 11, 2018 at 7:11 PM Maxime Coquelin
> <maxime.coquelin@redhat.com> wrote:
>>
>> Hi Matthias,
>>
>> On 12/6/18 5:00 PM, Matthias Gatto wrote:
>>> fdset_add can call fdset_shrink_nolock which call fdset_move
>>> concurrently to poll that is call in fdset_event_dispatch.
>>>
>>> This patch add a mutex to protect poll from been call at the same time
>>> fdset_add call fdset_shrink_nolock.
>>>
>>> Signed-off-by: Matthias Gatto <matthias.gatto@outscale.com>
>>> ---
>>>    lib/librte_vhost/fd_man.c | 4 ++++
>>>    lib/librte_vhost/fd_man.h | 1 +
>>>    lib/librte_vhost/socket.c | 1 +
>>>    3 files changed, 6 insertions(+)
>>>
>>> diff --git a/lib/librte_vhost/fd_man.c b/lib/librte_vhost/fd_man.c
>>> index 38347ab..55d4856 100644
>>> --- a/lib/librte_vhost/fd_man.c
>>> +++ b/lib/librte_vhost/fd_man.c
>>> @@ -129,7 +129,9 @@
>>>        pthread_mutex_lock(&pfdset->fd_mutex);
>>>        i = pfdset->num < MAX_FDS ? pfdset->num++ : -1;
>>>        if (i == -1) {
>>> +             pthread_mutex_lock(&pfdset->fd_pooling_mutex);
>>>                fdset_shrink_nolock(pfdset);
>>> +             pthread_mutex_unlock(&pfdset->fd_pooling_mutex);
>>>                i = pfdset->num < MAX_FDS ? pfdset->num++ : -1;
>>>                if (i == -1) {
>>>                        pthread_mutex_unlock(&pfdset->fd_mutex);
>>> @@ -246,7 +248,9 @@
>>>                numfds = pfdset->num;
>>>                pthread_mutex_unlock(&pfdset->fd_mutex);
>>>
>>> +             pthread_mutex_lock(&pfdset->fd_pooling_mutex);
>>>                val = poll(pfdset->rwfds, numfds, 1000 /* millisecs */);
>>> +             pthread_mutex_unlock(&pfdset->fd_pooling_mutex);
>>
>> Any reason we cannot use the existing fd_mutex?
> 
> yes, using the existing fd_mutex would block fdset_add during the polling in
> fdset_event_dispatch.
> 
> here fd_pooling_mutex block only fdset_shrink_nolock inside
> fdset_add which happen only in very rare occasions.


Thanks for the clarification:

Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Maxime