> -----Original Message-----
> From: Ying Xue [mailto:ying.x...@gmail.com]
> Sent: Thursday, 31 March, 2016 06:52
> To: Jon Maloy; tipc-discussion@lists.sourceforge.net; Parthasarathy
> Bhuvaragan;
> Richard Alpe
> Subject: Re: [tipc-discussion] [PATCH net-next v3 3/4] tipc: stricter
> filtering of
> packets in bearer layer
>
>
> On 03/31/2016 08:06 AM, Jon Maloy wrote:
> > Resetting a bearer/interface, with the consequence of resetting all its
> > pertaining links, is not an atomic action. This becomes particularly
> > evident in very large clusters, where a lot of traffic may happen on the
> > remaining links while we are busy shutting them down. In extreme cases,
> > we may even see links being re-created and re-established before we are
> > finished with the job.
> >
> > To solve this, we now introduce a solution where we temporarily detach
> > the bearer from the interface when the bearer is reset. This inhibits
> > all packet reception, while sending still is possible. For the latter,
> > we use the fact that the device's user pointer now is zero to filter out
> > which packets can be sent during this situation; i.e., outgoing RESET
> > messages only. This filtering serves to speed up the neighbors'
> > detection of the loss event, and saves us from unnecessary probing.
> >
> > Signed-off-by: Jon Maloy <jon.ma...@ericsson.com>
> > ---
> > net/tipc/bearer.c | 38 ++++++++++++++++++++++++++++++--------
> > net/tipc/msg.h | 5 +++++
> > 2 files changed, 35 insertions(+), 8 deletions(-)
> >
> > diff --git a/net/tipc/bearer.c b/net/tipc/bearer.c
> > index 20566e9..f5c6b80 100644
> > --- a/net/tipc/bearer.c
> > +++ b/net/tipc/bearer.c
> > @@ -342,8 +342,6 @@ static void bearer_disable(struct net *net, struct
> tipc_bearer *b)
> >
> > pr_info("Disabling bearer <%s>\n", b->name);
> > b->media->disable_media(b);
> > -
> > - tipc_node_delete_links(net, b->identity);
>
> I cannot understand why we postpone the action of deleting links until
> tn->bearer_list[] is set to NULL.
You are right. This causes the opposite of what I want to achieve; that the
bearer is open for sending, and closed for reception.
I will revert this. I don't know where I had my mind...
>
> > RCU_INIT_POINTER(b->media_ptr, NULL);
> > if (b->link_req)
> > tipc_disc_delete(b->link_req);
> > @@ -354,6 +352,7 @@ static void bearer_disable(struct net *net, struct
> tipc_bearer *b)
[...]
> > + }
> > + }
> > if (!b)
> > return NOTIFY_DONE;
> >
> > @@ -561,13 +576,20 @@ static int tipc_l2_device_event(struct notifier_block
> *nb, unsigned long evt,
> > case NETDEV_CHANGE:
> > if (netif_carrier_ok(dev))
> > break;
> > + case NETDEV_UP:
> > + rcu_assign_pointer(dev->tipc_ptr, b);
> > + break;
> > case NETDEV_GOING_DOWN:
> > + RCU_INIT_POINTER(dev->tipc_ptr, NULL);
> > + synchronize_net();
>
> Why do we place synchronize_net() before tipc_reset_bearer()? In my opinion,
> it
> seems better to exchange their orders between synchronize_net() and
> tipc_reset_bearer().
I don't agree. What I want is that all other threads see the zero-pointer as
soon as possible, so that there is no other incoming/outgoing traffic at the
moment we do reset_bearer().
If we wait with calling synchronize_net(), a number of existing threads may
continue receiving/sending packets while we are resetting. This is exactly what
we want to avoid, especially since reset_bearer() may take a long time.
///jon
>
> Regards,
> Ying
>
> > + tipc_reset_bearer(net, b);
> > + break;
> > case NETDEV_CHANGEMTU:
> > tipc_reset_bearer(net, b);
> > break;
> > case NETDEV_CHANGEADDR:
> > b->media->raw2addr(b, &b->addr,
> > - (char *)dev->dev_addr);
> > + (char *)dev->dev_addr);
> > tipc_reset_bearer(net, b);
> > break;
> > case NETDEV_UNREGISTER:
> > diff --git a/net/tipc/msg.h b/net/tipc/msg.h
> > index 55778a0..f34f639 100644
> > --- a/net/tipc/msg.h
> > +++ b/net/tipc/msg.h
> > @@ -779,6 +779,11 @@ static inline bool msg_peer_node_is_up(struct
> tipc_msg *m)
> > return msg_redundant_link(m);
> > }
> >
> > +static inline bool msg_is_reset(struct tipc_msg *hdr)
> > +{
> > + return (msg_user(hdr) == LINK_PROTOCOL) && (msg_type(hdr) ==
> RESET_MSG);
> > +}
> > +
> > struct sk_buff *tipc_buf_acquire(u32 size);
> > bool tipc_msg_validate(struct sk_buff *skb);
> > bool tipc_msg_reverse(u32 own_addr, struct sk_buff **skb, int err);
> >
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785471&iu=/4140
_______________________________________________
tipc-discussion mailing list
tipc-discussion@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/tipc-discussion
