From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:40562) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1WSOTi-0006DO-Sw for qemu-devel@nongnu.org; Tue, 25 Mar 2014 06:21:11 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1WSOTe-0001pf-A3 for qemu-devel@nongnu.org; Tue, 25 Mar 2014 06:21:06 -0400 Received: from mx1.redhat.com ([209.132.183.28]:60951) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1WSOTe-0001pX-2L for qemu-devel@nongnu.org; Tue, 25 Mar 2014 06:21:02 -0400 Date: Tue, 25 Mar 2014 12:21:08 +0200 From: "Michael S. Tsirkin" Message-ID: <20140325102108.GC25131@redhat.com> References: <1392914322-27329-1-git-send-email-akong@redhat.com> <53063F66.3020805@redhat.com> <20140221100140.GA1679@amosk.info> <20140228132959.3d3637fa@redhat.com> <20140303054634.GA20193@amosk.info> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20140303054634.GA20193@amosk.info> Subject: Re: [Qemu-devel] [PATCH v2] virtio-net: add a field to indicate if vlan table is used List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Amos Kong Cc: Vlad Yasevich , sf@sfritsch.de, qemu-devel@nongnu.org, Luiz Capitulino On Mon, Mar 03, 2014 at 01:46:34PM +0800, Amos Kong wrote: > On Fri, Feb 28, 2014 at 01:29:59PM -0500, Luiz Capitulino wrote: > > On Fri, 21 Feb 2014 18:01:40 +0800 > > Amos Kong wrote: > > > > > On Thu, Feb 20, 2014 at 12:46:14PM -0500, Vlad Yasevich wrote: > > > > On 02/20/2014 11:38 AM, Amos Kong wrote: > > > > > Stefan Fritsch just fixed a virtio-net driver bug [1], virtio-net won't > > > > > filter out VLAN-tagged packets if VIRTIO_NET_F_CTRL_VLAN isn't negotiated. > > > > > > > > > > This patch added a new field to @RxFilterInfo to indicate if management > > > > > uses the vlan table. > > > > > > > > > > [1] http://lists.nongnu.org/archive/html/qemu-devel/2014-02/msg02604.html > > > > > > > > > > Signed-off-by: Amos Kong > > > > > --- > > > > > V2: don't make vlan-table optional, add a flag to indicate > > > > > if vlan table is used by management > > > > > --- > > > > > hw/net/virtio-net.c | 38 +++++++++++++++++++++++++------------- > > > > > qapi-schema.json | 3 +++ > > > > > qmp-commands.hx | 2 ++ > > > > > 3 files changed, 30 insertions(+), 13 deletions(-) > > > > > > > > > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c > > > > > index 3626608..f591f4e 100644 > > > > > --- a/hw/net/virtio-net.c > > > > > +++ b/hw/net/virtio-net.c > > > > > @@ -222,13 +222,33 @@ static char *mac_strdup_printf(const uint8_t *mac) > > > > > mac[1], mac[2], mac[3], mac[4], mac[5]); > > > > > } > > > > > > > > > > +static intList *get_vlan_table(VirtIONet *n) > > > > > +{ > > > > > + intList *list, *entry; > > > > > + int i, j; > > > > > + > > > > > + list = NULL; > > > > > + for (i = 0; i < MAX_VLAN >> 5; i++) { > > > > > + for (j = 0; n->vlans[i] && j < 0x1f; j++) { > > > > > + if (n->vlans[i] & (1U << j)) { > > > > > + entry = g_malloc0(sizeof(*entry)); > > > > > + entry->value = (i << 5) + j; > > > > > + entry->next = list; > > > > > + list = entry; > > > > > + } > > > > > + } > > > > > + } > > > > > + > > > > > + return list; > > > > > +} > > > > > + > > > > > static RxFilterInfo *virtio_net_query_rxfilter(NetClientState *nc) > > > > > { > > > > > VirtIONet *n = qemu_get_nic_opaque(nc); > > > > > + VirtIODevice *vdev = VIRTIO_DEVICE(n); > > > > > RxFilterInfo *info; > > > > > strList *str_list, *entry; > > > > > - intList *int_list, *int_entry; > > > > > - int i, j; > > > > > + int i; > > > > > > > > > > info = g_malloc0(sizeof(*info)); > > > > > info->name = g_strdup(nc->name); > > > > > @@ -273,19 +293,11 @@ static RxFilterInfo *virtio_net_query_rxfilter(NetClientState *nc) > > > > > str_list = entry; > > > > > } > > > > > info->multicast_table = str_list; > > > > > + info->vlan_table = get_vlan_table(n); > > > > > > > > > > - int_list = NULL; > > > > > - for (i = 0; i < MAX_VLAN >> 5; i++) { > > > > > - for (j = 0; n->vlans[i] && j < 0x1f; j++) { > > > > > - if (n->vlans[i] & (1U << j)) { > > > > > - int_entry = g_malloc0(sizeof(*int_entry)); > > > > > - int_entry->value = (i << 5) + j; > > > > > - int_entry->next = int_list; > > > > > - int_list = int_entry; > > > > > - } > > > > > - } > > > > > + if ((1 << VIRTIO_NET_F_CTRL_VLAN) & vdev->guest_features) { > > > > > + info->vlan = true; > > > > > } > > > > > > > > So, in the case that vlan filtering is not supported in the guest > > > > we get: > > > > "vlan": false, > > > > "vlan-table": [ > > > > 0, > > > > 1, > > > > 2, > > > > ... > > > > 4095 > > > > ] > > > > since virtio_net now initializes the table to all 1s. > > > > Seems a bit awkward. We are providing a lot of data that > > > > is simply going to be ignored. > > > > > > In Stefan's patch [1], qemu fills all vlan ids to 1, then all the > > > packets will come to guest. > > > > > > For the host device, it also should not filter out any vlan-tagged > > > packets when VIRTIO_NET_F_CTRL_VLAN is not negotiated. We should also > > > fill vlan ids of host device to 1, then vlan-filter of host device > > > will not perform. > > > > > > If so, we don't need my patch, just pass [0,1,2,...4095] to management > > > as past. The new field in RxFilterInfo isn't necessary. > > > > What's the conclusion here? > > NAK my patch, it's unnecessary. I don't agree. libvirt does not yet look at the filter info, so let's do the job properly. https://lists.gnu.org/archive/html/qemu-devel/2014-02/msg02604.html > > > [1] [PATCH] virtio-net: Do not filter VLANs without F_CTRL_VLAN > > > > > > Thanks, Amos