From mboxrd@z Thu Jan 1 00:00:00 1970 From: Maxime Coquelin Subject: Re: [PATCH] vhost: fix race condition in fdset_add Date: Fri, 14 Dec 2018 10:53:11 +0100 Message-ID: <1fa47f8c-18a1-fbb4-1794-3324126cf9f1@redhat.com> References: <1544112007-23177-1-git-send-email-matthias.gatto@outscale.com> <1f9db33f-b281-b794-bc00-ad83490c2fbd@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 8bit Cc: dev@dpdk.org, tiwei.bie@intel.com, zhihong.wang@intel.com To: Matthias Gatto Return-path: Received: from mx1.redhat.com (mx1.redhat.com [209.132.183.28]) by dpdk.org (Postfix) with ESMTP id 4A2E11B925 for ; Fri, 14 Dec 2018 10:53:16 +0100 (CET) In-Reply-To: <1f9db33f-b281-b794-bc00-ad83490c2fbd@redhat.com> Content-Language: en-US List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" On 12/14/18 10:51 AM, Maxime Coquelin wrote: > > > On 12/14/18 10:32 AM, Matthias Gatto wrote: >> On Tue, Dec 11, 2018 at 7:11 PM Maxime Coquelin >> 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 >>>> --- >>>>    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 I guess we need to cc: stable, can you help with specifying which commit it fixes? Thanks in advance, Maxime > Maxime