Re: [Qemu-devel] [PATCH v4 1/8] record/replay: add network support

[Jason Wang <jasowang@redhat.com>]

On 2016年09月21日 19:33, Pavel Dovgalyuk wrote:
> This patch adds support of recording and replaying network packets in
> irount rr mode.
>
> Record and replay for network interactions is performed with the network filter.
> Each backend must have its own instance of the replay filter as follows:
>   -netdev user,id=net1 -device rtl8139,netdev=net1
>   -object filter-replay,id=replay,netdev=net1
>
> Replay network filter is used to record and replay network packets. While
> recording the virtual machine this filter puts all packets coming from
> the outer world into the log. In replay mode packets from the log are
> injected into the network device. All interactions with network backend
> in replay mode are disabled.
>
> Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>

Looks good, just few nits, see below.

Thanks

> ---
>   docs/replay.txt          |   14 ++++++
>   include/sysemu/replay.h  |   12 +++++
>   net/Makefile.objs        |    1
>   net/filter-replay.c      |   92 ++++++++++++++++++++++++++++++++++++++
>   replay/Makefile.objs     |    1
>   replay/replay-events.c   |   11 +++++
>   replay/replay-internal.h |   10 ++++
>   replay/replay-net.c      |  111 ++++++++++++++++++++++++++++++++++++++++++++++
>   replay/replay.c          |    2 -
>   vl.c                     |    3 +
>   10 files changed, 255 insertions(+), 2 deletions(-)
>   create mode 100644 net/filter-replay.c
>   create mode 100644 replay/replay-net.c
>
> diff --git a/docs/replay.txt b/docs/replay.txt
> index 779c6c0..347b2ff 100644
> --- a/docs/replay.txt
> +++ b/docs/replay.txt
> @@ -195,3 +195,17 @@ Queue is flushed at checkpoints and information about processed requests
>   is recorded to the log. In replay phase the queue is matched with
>   events read from the log. Therefore block devices requests are processed
>   deterministically.
> +
> +Network devices
> +---------------
> +
> +Record and replay for network interactions is performed with the network filter.
> +Each backend must have its own instance of the replay filter as follows:
> + -netdev user,id=net1 -device rtl8139,netdev=net1
> + -object filter-replay,id=replay,netdev=net1
> +
> +Replay network filter is used to record and replay network packets. While
> +recording the virtual machine this filter puts all packets coming from
> +the outer world into the log. In replay mode packets from the log are
> +injected into the network device. All interactions with network backend
> +in replay mode are disabled.
> diff --git a/include/sysemu/replay.h b/include/sysemu/replay.h
> index 0a88393..a408633 100644
> --- a/include/sysemu/replay.h
> +++ b/include/sysemu/replay.h
> @@ -39,6 +39,8 @@ enum ReplayCheckpoint {
>   };
>   typedef enum ReplayCheckpoint ReplayCheckpoint;
>   
> +typedef struct ReplayNetState ReplayNetState;
> +
>   extern ReplayMode replay_mode;
>   
>   /* Replay process control functions */
> @@ -133,4 +135,14 @@ void replay_char_read_all_save_error(int res);
>   /*! Writes character read_all execution result into the replay log. */
>   void replay_char_read_all_save_buf(uint8_t *buf, int offset);
>   
> +/* Network */
> +
> +/*! Registers replay network filter attached to some backend. */
> +ReplayNetState *replay_register_net(NetFilterState *nfs);
> +/*! Unregisters replay network filter. */
> +void replay_unregister_net(ReplayNetState *rns);
> +/*! Called to write network packet to the replay log. */
> +void replay_net_packet_event(ReplayNetState *rns, unsigned flags,
> +                             const struct iovec *iov, int iovcnt);
> +
>   #endif
> diff --git a/net/Makefile.objs b/net/Makefile.objs
> index b7c22fd..f787ba4 100644
> --- a/net/Makefile.objs
> +++ b/net/Makefile.objs
> @@ -16,3 +16,4 @@ common-obj-$(CONFIG_NETMAP) += netmap.o
>   common-obj-y += filter.o
>   common-obj-y += filter-buffer.o
>   common-obj-y += filter-mirror.o
> +common-obj-y += filter-replay.o
> diff --git a/net/filter-replay.c b/net/filter-replay.c
> new file mode 100644
> index 0000000..cff65f8
> --- /dev/null
> +++ b/net/filter-replay.c
> @@ -0,0 +1,92 @@
> +/*
> + * filter-replay.c
> + *
> + * Copyright (c) 2010-2016 Institute for System Programming
> + *                         of the Russian Academy of Sciences.
> + *
> + * 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 "qemu/osdep.h"
> +#include "clients.h"
> +#include "qapi/error.h"
> +#include "qemu-common.h"
> +#include "qemu/error-report.h"
> +#include "qemu/iov.h"
> +#include "qemu/log.h"
> +#include "qemu/timer.h"
> +#include "qapi/visitor.h"
> +#include "net/filter.h"
> +#include "sysemu/replay.h"
> +
> +#define TYPE_FILTER_REPLAY "filter-replay"
> +
> +#define FILTER_REPLAY(obj) \
> +    OBJECT_CHECK(NetFilterReplayState, (obj), TYPE_FILTER_REPLAY)
> +
> +struct NetFilterReplayState {
> +    NetFilterState nfs;
> +    ReplayNetState *rns;
> +};
> +typedef struct NetFilterReplayState NetFilterReplayState;
> +
> +static ssize_t filter_replay_receive_iov(NetFilterState *nf,
> +                                         NetClientState *sndr,
> +                                         unsigned flags,
> +                                         const struct iovec *iov,
> +                                         int iovcnt, NetPacketSent *sent_cb)
> +{
> +    NetFilterReplayState *nfrs = FILTER_REPLAY(nf);
> +    switch (replay_mode) {
> +    case REPLAY_MODE_RECORD:
> +        if (nf->netdev == sndr) {
> +            replay_net_packet_event(nfrs->rns, flags, iov, iovcnt);
> +            return iov_size(iov, iovcnt);
> +        }
> +        return 0;
> +    case REPLAY_MODE_PLAY:
> +        /* Drop all packets in replay mode.
> +           Packets from the log will be injected by the replay module. */
> +        return iov_size(iov, iovcnt);
> +    default:
> +        /* Pass all the packets. */
> +        return 0;
> +    }
> +}
> +
> +static void filter_replay_instance_init(Object *obj)
> +{
> +    NetFilterReplayState *nfrs = FILTER_REPLAY(obj);
> +    nfrs->rns = replay_register_net(&nfrs->nfs);
> +}
> +
> +static void filter_replay_instance_finalize(Object *obj)
> +{
> +    NetFilterReplayState *nfrs = FILTER_REPLAY(obj);
> +    replay_unregister_net(nfrs->rns);
> +}
> +
> +static void filter_replay_class_init(ObjectClass *oc, void *data)
> +{
> +    NetFilterClass *nfc = NETFILTER_CLASS(oc);
> +
> +    nfc->receive_iov = filter_replay_receive_iov;
> +}
> +
> +static const TypeInfo filter_replay_info = {
> +    .name = TYPE_FILTER_REPLAY,
> +    .parent = TYPE_NETFILTER,
> +    .class_init = filter_replay_class_init,
> +    .instance_init = filter_replay_instance_init,
> +    .instance_finalize = filter_replay_instance_finalize,
> +    .instance_size = sizeof(NetFilterReplayState),
> +};
> +
> +static void filter_replay_register_types(void)
> +{
> +    type_register_static(&filter_replay_info);
> +}
> +
> +type_init(filter_replay_register_types);
> diff --git a/replay/Makefile.objs b/replay/Makefile.objs
> index fcb3f74..f55a6b5 100644
> --- a/replay/Makefile.objs
> +++ b/replay/Makefile.objs
> @@ -4,3 +4,4 @@ common-obj-y += replay-events.o
>   common-obj-y += replay-time.o
>   common-obj-y += replay-input.o
>   common-obj-y += replay-char.o
> +common-obj-y += replay-net.o
> diff --git a/replay/replay-events.c b/replay/replay-events.c
> index 3807245..9ce9e51 100644
> --- a/replay/replay-events.c
> +++ b/replay/replay-events.c
> @@ -54,6 +54,9 @@ static void replay_run_event(Event *event)
>       case REPLAY_ASYNC_EVENT_BLOCK:
>           aio_bh_call(event->opaque);
>           break;
> +    case REPLAY_ASYNC_EVENT_NET:
> +        replay_event_net_run(event->opaque);
> +        break;
>       default:
>           error_report("Replay: invalid async event ID (%d) in the queue",
>                       event->event_kind);
> @@ -189,6 +192,9 @@ static void replay_save_event(Event *event, int checkpoint)
>           case REPLAY_ASYNC_EVENT_BLOCK:
>               replay_put_qword(event->id);
>               break;
> +        case REPLAY_ASYNC_EVENT_NET:
> +            replay_event_net_save(event->opaque);
> +            break;
>           default:
>               error_report("Unknown ID %" PRId64 " of replay event", event->id);
>               exit(1);
> @@ -252,6 +258,11 @@ static Event *replay_read_event(int checkpoint)
>               read_id = replay_get_qword();
>           }
>           break;
> +    case REPLAY_ASYNC_EVENT_NET:
> +        event = g_malloc0(sizeof(Event));
> +        event->event_kind = read_event_kind;
> +        event->opaque = replay_event_net_load();
> +        return event;
>       default:
>           error_report("Unknown ID %d of replay event", read_event_kind);
>           exit(1);
> diff --git a/replay/replay-internal.h b/replay/replay-internal.h
> index efbf14c..d28cfb7 100644
> --- a/replay/replay-internal.h
> +++ b/replay/replay-internal.h
> @@ -50,6 +50,7 @@ enum ReplayAsyncEventKind {
>       REPLAY_ASYNC_EVENT_INPUT_SYNC,
>       REPLAY_ASYNC_EVENT_CHAR_READ,
>       REPLAY_ASYNC_EVENT_BLOCK,
> +    REPLAY_ASYNC_EVENT_NET,
>       REPLAY_ASYNC_COUNT
>   };
>   
> @@ -155,4 +156,13 @@ void replay_event_char_read_save(void *opaque);
>   /*! Reads char event read from the file. */
>   void *replay_event_char_read_load(void);
>   
> +/* Network devices */
> +
> +/*! Called to run network event. */
> +void replay_event_net_run(void *opaque);
> +/*! Writes network event to the file. */
> +void replay_event_net_save(void *opaque);
> +/*! Reads network from the file. */
> +void *replay_event_net_load(void);
> +
>   #endif
> diff --git a/replay/replay-net.c b/replay/replay-net.c
> new file mode 100644
> index 0000000..4f7b995
> --- /dev/null
> +++ b/replay/replay-net.c
> @@ -0,0 +1,111 @@
> +/*
> + * replay-net.c
> + *
> + * Copyright (c) 2010-2016 Institute for System Programming
> + *                         of the Russian Academy of Sciences.
> + *
> + * 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 "qemu/osdep.h"
> +#include "qemu/error-report.h"
> +#include "sysemu/replay.h"
> +#include "replay-internal.h"
> +#include "sysemu/sysemu.h"
> +#include "net/net.h"
> +#include "net/filter.h"
> +#include "qemu/iov.h"
> +
> +struct ReplayNetState {
> +    NetFilterState *nfs;
> +    int id;
> +};
> +
> +typedef struct NetEvent {
> +    uint8_t id;
> +    uint32_t flags;
> +    uint8_t *data;
> +    size_t size;
> +} NetEvent;
> +
> +static NetFilterState **network_filters;
> +static int network_filters_count;
> +
> +ReplayNetState *replay_register_net(NetFilterState *nfs)
> +{
> +    ReplayNetState *rns = g_new0(ReplayNetState, 1);
> +    rns->nfs = nfs;
> +    rns->id = network_filters_count++;
> +    network_filters = g_realloc(network_filters,
> +                                network_filters_count
> +                                    * sizeof(*network_filters));

Any chance to avoid maintaining such array? E.g by having a pointer to 
NetFilterState in each NetEvent?

> +    network_filters[network_filters_count - 1] = nfs;
> +    return rns;
> +}
> +
> +void replay_unregister_net(ReplayNetState *rns)
> +{
> +    network_filters[rns->id] = NULL;
> +    g_free(rns);
> +}
> +
> +void replay_net_packet_event(ReplayNetState *rns, unsigned flags,
> +                             const struct iovec *iov, int iovcnt)
> +{
> +    NetEvent *event;
> +    int i;
> +
> +    event = g_new(NetEvent, 1);
> +    event->flags = flags;
> +    event->data = g_malloc(iov_size(iov, iovcnt));
> +    event->size = 0;
> +    event->id = rns->id;
> +
> +    for (i = 0; i < iovcnt; i++) {
> +        size_t len = iov[i].iov_len;
> +
> +        memcpy(event->data + event->size, iov[i].iov_base, len);
> +        event->size += len;
> +    }

You can use iov_to_buf() instead.

> +
> +    replay_add_event(REPLAY_ASYNC_EVENT_NET, event, NULL, 0);
> +}
> +
> +void replay_event_net_run(void *opaque)
> +{
> +    NetEvent *event = opaque;
> +    struct iovec iov = {
> +        .iov_base = (void *)event->data,
> +        .iov_len = event->size
> +    };
> +
> +    assert(event->id < network_filters_count);
> +
> +    qemu_netfilter_pass_to_next(network_filters[event->id]->netdev,
> +        event->flags, &iov, 1, network_filters[event->id]);
> +
> +    g_free(event->data);
> +    g_free(event);
> +}
> +
> +void replay_event_net_save(void *opaque)
> +{
> +    NetEvent *event = opaque;
> +
> +    replay_put_byte(event->id);
> +    replay_put_dword(event->flags);
> +    replay_put_array(event->data, event->size);
> +}
> +
> +void *replay_event_net_load(void)
> +{
> +    NetEvent *event = g_new(NetEvent, 1);
> +
> +    event->id = replay_get_byte();
> +    event->flags = replay_get_dword();
> +    replay_get_array_alloc(&event->data, &event->size);
> +
> +    return event;
> +}
> diff --git a/replay/replay.c b/replay/replay.c
> index 167fd29..e040f6f 100644
> --- a/replay/replay.c
> +++ b/replay/replay.c
> @@ -21,7 +21,7 @@
>   
>   /* Current version of the replay mechanism.
>      Increase it when file format changes. */
> -#define REPLAY_VERSION              0xe02004
> +#define REPLAY_VERSION              0xe02005
>   /* Size of replay log header */
>   #define HEADER_SIZE                 (sizeof(uint32_t) + sizeof(uint64_t))
>   
> diff --git a/vl.c b/vl.c
> index fca0487..fd7f17e 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -2807,7 +2807,8 @@ static bool object_create_initial(const char *type)
>       if (g_str_equal(type, "filter-buffer") ||
>           g_str_equal(type, "filter-dump") ||
>           g_str_equal(type, "filter-mirror") ||
> -        g_str_equal(type, "filter-redirector")) {
> +        g_str_equal(type, "filter-redirector") ||
> +        g_str_equal(type, "filter-replay")) {
>           return false;
>       }
>   
>