On 10/12/2018 04:02 AM, Jon Maloy wrote:
> In the commit referred to below we added link tolerance as an additional
> criteria for declaring broadcast transmission "stale" and resetting the
> unicast links to the affected node.
>
> Unfortunately, this 'improvement' introduced two bugs, which each and
> one alone cause only limited problems, but combined lead to seemingly
> stochastic unicast link resets, depending on the amount of broadcast
> traffic transmitted.
>
> The first issue, a missing initialization of the 'tolerance' field of
> the receiver broadcast link, was recently fixed by commit 047491ea334a
> ("tipc: set link tolerance correctly in broadcast link").
>
> Ths second issue, where we omit to reset the 'stale_cnt' field of
> the same link after a 'stale' period is over, leads to this counter
> accumulating over time, and in the absence of the 'tolerance' criteria
> leads to the above described symptoms. This commit adds the missing
> initialization.
>
> Fixes: a4dc70d46cf1 ("tipc: extend link reset criteria for stale packet
> retransmission")
>
> Signed-off-by: Jon Maloy <jon.ma...@ericsson.com>
Acked-by: Ying Xue <ying....@windriver.com>
> ---
> net/tipc/link.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/net/tipc/link.c b/net/tipc/link.c
> index f6552e4..201c3b5 100644
> --- a/net/tipc/link.c
> +++ b/net/tipc/link.c
> @@ -1041,6 +1041,7 @@ static int tipc_link_retrans(struct tipc_link *l,
> struct tipc_link *r,
> if (r->last_retransm != buf_seqno(skb)) {
> r->last_retransm = buf_seqno(skb);
> r->stale_limit = jiffies + msecs_to_jiffies(r->tolerance);
> + r->stale_cnt = 0;
> } else if (++r->stale_cnt > 99 && time_after(jiffies, r->stale_limit)) {
> link_retransmit_failure(l, skb);
> if (link_is_bc_sndlink(l))
>
_______________________________________________
tipc-discussion mailing list
tipc-discussion@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/tipc-discussion
