Re: [Qemu-devel] [PATCH v3 10/12] netfilter: add a netbuffer filter

[Yang Hongyang <yanghy@cn.fujitsu.com>]

On 08/04/2015 02:30 PM, Jason Wang wrote:
>
>
> On 08/04/2015 02:05 PM, Yang Hongyang wrote:
>> On 08/04/2015 01:03 PM, Jason Wang wrote:
>>>
>>>
>>> On 08/03/2015 04:30 PM, Yang Hongyang wrote:
>>>> This filter is to buffer/release packets, this feature can be used
>>>> when using MicroCheckpointing, or other Remus like VM FT solutions, you
>>>> can also use it to simulate the network delay.
>>>> It has an interval option, if supplied, this filter will release
>>>> packets by interval.
>>>>
>>>> Usage:
>>>>    -netdev tap,id=bn0
>>>>    -netfilter buffer,id=f0,netdev=bn0,chain=in,interval=1000
>>>>
>>>> NOTE:
>>>>    the scale of interval is microsecond.
>>>>
>>>> Signed-off-by: Yang Hongyang <yanghy@cn.fujitsu.com>
>>>> ---
>>>> v3: check packet's sender and sender->peer when flush it
>>>> ---
>>>>    net/Makefile.objs   |   1 +
>>>>    net/filter-buffer.c | 162
>>>> ++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>    net/filter.c        |   2 +
>>>>    net/filters.h       |  17 ++++++
>>>>    qapi-schema.json    |  18 +++++-
>>>>    5 files changed, 199 insertions(+), 1 deletion(-)
>>>>    create mode 100644 net/filter-buffer.c
>>>>    create mode 100644 net/filters.h
>>>>
>>>> diff --git a/net/Makefile.objs b/net/Makefile.objs
>>>> index 914aec0..5fa2f97 100644
>>>> --- a/net/Makefile.objs
>>>> +++ b/net/Makefile.objs
>>>> @@ -14,3 +14,4 @@ common-obj-$(CONFIG_SLIRP) += slirp.o
>>>>    common-obj-$(CONFIG_VDE) += vde.o
>>>>    common-obj-$(CONFIG_NETMAP) += netmap.o
>>>>    common-obj-y += filter.o
>>>> +common-obj-y += filter-buffer.o
>>>> diff --git a/net/filter-buffer.c b/net/filter-buffer.c
>>>> new file mode 100644
>>>> index 0000000..1547765
>>>> --- /dev/null
>>>> +++ b/net/filter-buffer.c
>>>> @@ -0,0 +1,162 @@
>>>> +/*
>>>> + * Copyright (c) 2015 FUJITSU LIMITED
>>>> + * Author: Yang Hongyang <yanghy@cn.fujitsu.com>
>>>> + *
>>>> + * This work is licensed under the terms of the GNU GPL, version 2 or
>>>> + * later.  See the COPYING file in the top-level directory.
>>>> + */
>>>> +
>>>> +#include "net/filter.h"
>>>> +#include "net/queue.h"
>>>> +#include "filters.h"
>>>> +#include "qemu-common.h"
>>>> +#include "qemu/error-report.h"
>>>> +#include "qemu/main-loop.h"
>>>> +#include "qemu/timer.h"
>>>> +#include "qemu/iov.h"
>>>> +
>>>> +typedef struct FILTERBUFFERState {
>>>> +    NetFilterState nf;
>>>> +    NetQueue *incoming_queue;
>>>> +    NetQueue *inflight_queue;
>>>> +    QEMUBH *flush_bh;
>>>> +    int64_t interval;
>>>> +    QEMUTimer release_timer;
>>>> +} FILTERBUFFERState;
>>>> +
>>>> +static void packet_send_completed(NetClientState *nc, ssize_t len)
>>>> +{
>>>> +    return;
>>>> +}
>>>> +
>>>> +static void filter_buffer_flush(NetFilterState *nf)
>>>> +{
>>>> +    FILTERBUFFERState *s = DO_UPCAST(FILTERBUFFERState, nf, nf);
>>>> +    NetQueue *queue = s->inflight_queue;
>>>> +    NetPacket *packet;
>>>> +
>>>> +    while (queue && !QTAILQ_EMPTY(&queue->packets)) {
>>>> +        packet = QTAILQ_FIRST(&queue->packets);
>>>> +        QTAILQ_REMOVE(&queue->packets, packet, entry);
>>>> +        queue->nq_count--;
>>>> +
>>>> +        if (packet->sender && packet->sender->peer) {
>>>> +            qemu_net_queue_send(packet->sender->peer->incoming_queue,
>>>> +                                packet->sender,
>>>> +                                packet->flags,
>>>> +                                packet->data,
>>>> +                                packet->size,
>>>> +                                packet->sent_cb);
>>>> +        }
>>>> +
>>>> +        /*
>>>> +         * now that we pass the packet to
>>>> sender->peer->incoming_queue, we
>>>> +         * don't care the reture value here, because the peer's
>>>> queue will
>>>> +         * take care of this packet
>>>> +         */
>>>> +        g_free(packet);
>>>
>>> So looks like the packet was still not passed to next filter?
>>
>> I didn't get your suggestion last time, sorry, will add it in next
>> version.
>> Just to confirm, do you mean we need to pass the packet to next filter
>> instead of pass to the receiver's incoming_queue?
>
> Yes, consider you may have two filters. First is dump and second is
> buffer, I believe you still want to buffer the packet even if it has
> been dumped.
>
>> and even if we pass to next
>> filter, the check of sender and it's peer is still needed, because if
>> there's no receiver, it's nonsense to pass it further?
>
> Yes.

thanks.

>
>>
>>>
>>>> +    }
>>>> +
>>>> +    g_free(queue);
>>>> +    s->inflight_queue = NULL;
>>>> +}
>>>> +
>>>> +static void filter_buffer_flush_bh(void *opaque)
>>>> +{
>>>> +    FILTERBUFFERState *s = opaque;
>>>> +    NetFilterState *nf = &s->nf;
>>>> +    filter_buffer_flush(nf);
>>>> +}
>>>> +
>>>> +static void filter_buffer_release_one(NetFilterState *nf)
>>>> +{
>>>> +    FILTERBUFFERState *s = DO_UPCAST(FILTERBUFFERState, nf, nf);
>>>> +
>>>> +    /* flush inflight packets */
>>>> +    if (s->inflight_queue) {
>>>> +        filter_buffer_flush(nf);
>>>> +    }
>>>> +
>>>> +    s->inflight_queue = s->incoming_queue;
>>>> +    s->incoming_queue = qemu_new_net_queue(nf);
>>>
>>> So this in fact flush a brunch of packets. If yes, the name of function
>>> is confusing
>>
>> maybe filter_buffer_release ?
>>
>
> Right.

By removing bh, this function can be dropped, just call filter_buffer_flush.
thanks.

>
>>>
>>>> +    qemu_bh_schedule(s->flush_bh);
>>>
>>> Don't get why a bh is needed. If we could get rid of it, there's
>>> probably no need for inflight_queue, and we can just drain
>>> incoming_queue here.
>>
>> Seems we can get rid of bh, will do, thanks.
>
> .
>

-- 
Thanks,
Yang.