From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:52867) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ZKhpw-0001Te-JQ for qemu-devel@nongnu.org; Thu, 30 Jul 2015 03:01:05 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1ZKhpt-00038l-BX for qemu-devel@nongnu.org; Thu, 30 Jul 2015 03:01:04 -0400 Received: from [59.151.112.132] (port=6923 helo=heian.cn.fujitsu.com) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ZKhpr-00035t-Qu for qemu-devel@nongnu.org; Thu, 30 Jul 2015 03:01:01 -0400 Message-ID: <55B9CBA5.7050103@cn.fujitsu.com> Date: Thu, 30 Jul 2015 15:00:53 +0800 From: Yang Hongyang MIME-Version: 1.0 References: <1438167116-29270-1-git-send-email-yanghy@cn.fujitsu.com> <1438167116-29270-10-git-send-email-yanghy@cn.fujitsu.com> <55B9B269.3030709@redhat.com> In-Reply-To: <55B9B269.3030709@redhat.com> Content-Type: text/plain; charset="windows-1252"; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH 09/12] netfilter: add a netbuffer filter List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jason Wang , qemu-devel@nongnu.org Cc: thuth@redhat.com, stefanha@redhat.com, zhang.zhanghailiang@huawei.com, mrhines@linux.vnet.ibm.com On 07/30/2015 01:13 PM, Jason Wang wrote: [...] >> + if (sender->info->type == NET_CLIENT_OPTIONS_KIND_NIC) { >> + /* we only buffer guest output packets */ >> + qemu_net_queue_append(queue, sender, flags, data, size, >> + packet_send_completed); > > This may brings some confusion for user. Since the name is 'netbuffer'. > Maybe we can change the filter to be ingress or out or both? Like: > > -device virtio-net-pci,id=virtio0 > -netfilter buffer,id=filter0,dev=virtio0,interval=1000,type=out > > Then we can just try to enqueue the packet when virtio-net-pci is sender? A good idea, I also thought about this, but a question, should we make this type option a mandatory to netfilter object or optional? if it's optional, the default will be "all" > >> + /* Now that we have buffered the packet, return sucess */ >> + return size; >> + } >> + >> + return 0; >> +} >> + >> +static void filter_buffer_cleanup(NetFilterState *nf) >> +{ >> + FILTERBUFFERState *s = DO_UPCAST(FILTERBUFFERState, nf, nf); >> + >> + /* flush inflight packets */ >> + filter_buffer_flush(nf); >> + /* flush incoming packets */ >> + s->inflight_queue = s->incoming_queue; >> + s->incoming_queue = NULL; >> + filter_buffer_flush(nf); >> + >> + return; >> +} >> + >> + >> +static NetFilterInfo net_filter_buffer_info = { >> + .type = NET_FILTER_OPTIONS_KIND_BUFFER, >> + .size = sizeof(FILTERBUFFERState), >> + .receive = filter_buffer_receive, >> + .cleanup = filter_buffer_cleanup, >> +}; >> + >> +int net_init_filter_buffer(const NetFilterOptions *opts, const char *name, >> + NetClientState *netdev, Error **errp) >> +{ >> + NetFilterState *nf; >> + FILTERBUFFERState *s; >> + >> + assert(opts->kind == NET_FILTER_OPTIONS_KIND_BUFFER); >> + >> + nf = qemu_new_net_filter(&net_filter_buffer_info, netdev, "buffer", name); >> + s = DO_UPCAST(FILTERBUFFERState, nf, nf); >> + /* >> + * we need the dummy NetClientState to send packets in order to avoid >> + * receive packets again. >> + * we are buffering guest output packets, our buffered packets should be >> + * sent to real network backend, so dummy's peer should be that backend. >> + */ >> + s->dummy.peer = netdev; >> + s->incoming_queue = qemu_new_net_queue(nf); >> + >> + return 0; >> +} >> diff --git a/net/filter.c b/net/filter.c >> index 50fb837..e741e2a 100644 >> --- a/net/filter.c >> +++ b/net/filter.c >> @@ -18,6 +18,7 @@ >> >> #include "net/filter.h" >> #include "net/net.h" >> +#include "filters.h" >> >> static QTAILQ_HEAD(, NetFilterState) net_filters; >> >> @@ -152,6 +153,7 @@ typedef int (NetFilterInit)(const NetFilterOptions *opts, >> >> static >> NetFilterInit * const net_filter_init_fun[NET_FILTER_OPTIONS_KIND_MAX] = { >> + [NET_FILTER_OPTIONS_KIND_BUFFER] = net_init_filter_buffer, >> }; >> >> static int net_filter_init1(const NetFilter *netfilter, Error **errp) >> diff --git a/net/filters.h b/net/filters.h >> new file mode 100644 >> index 0000000..6c249b8 >> --- /dev/null >> +++ b/net/filters.h >> @@ -0,0 +1,17 @@ >> +/* >> + * Copyright (c) 2015 FUJITSU LIMITED >> + * >> + * This work is licensed under the terms of the GNU GPL, version 2 or >> + * later. See the COPYING file in the top-level directory. >> + */ >> + >> +#ifndef QEMU_NET_FILTERS_H >> +#define QEMU_NET_FILTERS_H >> + >> +#include "net/net.h" >> +#include "net/filter.h" >> + >> +int net_init_filter_buffer(const NetFilterOptions *opts, const char *name, >> + NetClientState *netdev, Error **errp); >> + >> +#endif /* QEMU_NET_FILTERS_H */ >> diff --git a/qapi-schema.json b/qapi-schema.json >> index 1fc6390..67e00a0 100644 >> --- a/qapi-schema.json >> +++ b/qapi-schema.json >> @@ -2577,6 +2577,16 @@ >> { 'command': 'netfilter_del', 'data': {'id': 'str'} } >> >> ## >> +# @NetFilterBufferOptions >> +# >> +# a netbuffer filter for network backend. >> +# >> +# Since 2.5 >> +## >> +{ 'struct': 'NetFilterBufferOptions', >> + 'data': { } } >> + >> +## >> # @NetFilterOptions >> # >> # A discriminated record of network filters. >> @@ -2585,7 +2595,8 @@ >> # >> ## >> { 'union': 'NetFilterOptions', >> - 'data': { } } >> + 'data': { >> + 'buffer': 'NetFilterBufferOptions'} } >> >> ## >> # @NetFilter > > . > -- Thanks, Yang.