From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id 686564F98; Fri, 14 Sep 2018 16:52:28 +0200 (CEST) X-Amp-Result: UNKNOWN X-Amp-Original-Verdict: FILE UNKNOWN X-Amp-File-Uploaded: False Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by fmsmga101.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 14 Sep 2018 07:52:27 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.53,373,1531810800"; d="scan'208";a="86058238" Received: from btwcube1.sh.intel.com (HELO debian) ([10.67.104.151]) by fmsmga002.fm.intel.com with ESMTP; 14 Sep 2018 07:52:26 -0700 Date: Fri, 14 Sep 2018 22:51:24 +0800 From: Tiwei Bie To: Ferruh Yigit Cc: Maxime Coquelin , zhihong.wang@intel.com, dev@dpdk.org, stable@dpdk.org Message-ID: <20180914145124.GA29384@debian> References: <20180904235555.31839-1-tiwei.bie@intel.com> <20180910073637.GA25254@debian> <576bfb46-f1fc-cf62-97bc-2d5857215160@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <576bfb46-f1fc-cf62-97bc-2d5857215160@intel.com> User-Agent: Mutt/1.10.1 (2018-07-13) Subject: Re: [dpdk-dev] [dpdk-stable] [PATCH] vhost: fix vhost interrupt support X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 14 Sep 2018 14:52:29 -0000 On Fri, Sep 14, 2018 at 01:10:07PM +0100, Ferruh Yigit wrote: > On 9/10/2018 8:36 AM, Tiwei Bie wrote: > > Hi Maxime, > > > > On Mon, Sep 10, 2018 at 09:22:00AM +0200, Maxime Coquelin wrote: > >> Hi Tiwei, > >> > >> On 09/05/2018 01:55 AM, Tiwei Bie wrote: > >>> When VIRTIO_RING_F_EVENT_IDX is negotiated, we need to > >>> update the avail event to enable the notification. > >>> > >>> Fixes: 3f8ff12821e4 ("vhost: support interrupt mode") > >>> Cc: stable@dpdk.org > >>> > >>> Signed-off-by: Tiwei Bie > >>> --- > >>> lib/librte_vhost/Makefile | 1 + > >>> lib/librte_vhost/vhost.c | 18 ++++++++++++------ > >>> lib/librte_vhost/vhost.h | 2 ++ > >>> 3 files changed, 15 insertions(+), 6 deletions(-) > >>> > >>> diff --git a/lib/librte_vhost/Makefile b/lib/librte_vhost/Makefile > >>> index de431fbb7..531cf4832 100644 > >>> --- a/lib/librte_vhost/Makefile > >>> +++ b/lib/librte_vhost/Makefile > >>> @@ -13,6 +13,7 @@ LIBABIVER := 4 > >>> CFLAGS += -DALLOW_EXPERIMENTAL_API > >>> CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR) -O3 -D_FILE_OFFSET_BITS=64 > >>> CFLAGS += -I vhost_user > >>> +CFLAGS += -fno-strict-aliasing > >> > >> I'm not clear why this is needed looking at the code below, > >> could you please explain? > > > > Without this, we will get below build error: > > > > lib/librte_vhost/vhost.c: In function ‘vhost_enable_notify_split’: > > lib/librte_vhost/vhost.h:656:4: error: dereferencing type-punned pointer will break strict-aliasing rules [-Werror=strict-aliasing] > > (*(volatile uint16_t*)&(vr)->used->ring[(vr)->size]) > > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > > lib/librte_vhost/vhost.c:659:4: note: in expansion of macro ‘vhost_avail_event’ > > vhost_avail_event(vq) = vq->last_avail_idx; > > ^~~~~~~~~~~~~~~~~ > > cc1: all warnings being treated as errors > > > > Getting this error with meson build system, since "-fno-strict-aliasing" is not > added to meson build file. Thanks! I didn't notice this. I just tried the meson build manually. The warning isn't treated as errors on my side. So it was not caught by the build script.. > BUT instead of disabling compiler warning, > > first element of the "struct vring_used_elem ring[0]" is "uint32_t id;", so it > should be accessed as uint32_t. What about following: > > #define vhost_avail_event(vr) \ > - (*(volatile uint16_t*)&(vr)->used->ring[(vr)->size]) > + (*(volatile uint32_t*)&(vr)->used->ring[(vr)->size]) > > And for "vhost_avail_event(vq) = vq->last_avail_idx;", "last_avail_idx" is > uint16_t, so assigning from uint16_t to uint32_t should be ok too. At the end of used ring, there is a 16bits event_idx. It can't be interpreted as a vring_used_elem and it's also not part of a vring_used_elem. This is the traditional way how we access it: https://github.com/torvalds/linux/blob/a06b0c82a049/drivers/vhost/vhost.c#L50 https://github.com/freebsd/freebsd/blob/209b94df8701/sys/dev/virtio/virtio_ring.h#L133 > > If above works for you please send a patch for update also removing > "-fno-strict-aliasing", this can be rebased on existing patch. > > If not works, please at least send a patch to add flag to meson build file I'll send a patch to add flag to meson build file. Thanks! > > Thanks, > ferruh > > >> > >>> LDLIBS += -lpthread > >>> ifeq ($(CONFIG_RTE_LIBRTE_VHOST_NUMA),y) > >>> diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c > >>> index 3c9be10a0..88b1781d5 100644 > >>> --- a/lib/librte_vhost/vhost.c > >>> +++ b/lib/librte_vhost/vhost.c > >>> @@ -646,12 +646,18 @@ rte_vhost_avail_entries(int vid, uint16_t queue_id) > >>> } > >>> static inline void > >>> -vhost_enable_notify_split(struct vhost_virtqueue *vq, int enable) > >>> +vhost_enable_notify_split(struct virtio_net *dev, > >>> + struct vhost_virtqueue *vq, int enable) > >>> { > >>> - if (enable) > >>> - vq->used->flags &= ~VRING_USED_F_NO_NOTIFY; > >>> - else > >>> - vq->used->flags |= VRING_USED_F_NO_NOTIFY; > >>> + if (!(dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))) { > >>> + if (enable) > >>> + vq->used->flags &= ~VRING_USED_F_NO_NOTIFY; > >>> + else > >>> + vq->used->flags |= VRING_USED_F_NO_NOTIFY; > >>> + } else { > >>> + if (enable) > >>> + vhost_avail_event(vq) = vq->last_avail_idx; > >>> + } > >>> } > >>> static inline void > >>> @@ -689,7 +695,7 @@ rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable) > >>> if (vq_is_packed(dev)) > >>> vhost_enable_notify_packed(dev, vq, enable); > >>> else > >>> - vhost_enable_notify_split(vq, enable); > >>> + vhost_enable_notify_split(dev, vq, enable); > >>> return 0; > >>> } > >>> diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h > >>> index 760a09c0d..25ffd7614 100644 > >>> --- a/lib/librte_vhost/vhost.h > >>> +++ b/lib/librte_vhost/vhost.h > >>> @@ -648,6 +648,8 @@ vhost_iova_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq, > >>> return __vhost_iova_to_vva(dev, vq, iova, len, perm); > >>> } > >>> +#define vhost_avail_event(vr) \ > >>> + (*(volatile uint16_t*)&(vr)->used->ring[(vr)->size]) > >>> #define vhost_used_event(vr) \ > >>> (*(volatile uint16_t*)&(vr)->avail->ring[(vr)->size]) > >>> >