Commit a81e18da authored by Ondřej Zajíček's avatar Ondřej Zajíček

Nest: Fix race condition during reconfiguration

If export filter is changed during reconfiguration and a route disappears
between reconfiguration and refeed (e.g., if the route is a static route
also removed during the reconfiguration), the route is not withdrawn.
The patch fixes that by adding tx reconfiguration timestamp.
parent 8e86ffce
Pipeline #37594 passed with stages
in 4 minutes and 45 seconds
......@@ -165,6 +165,7 @@ proto_add_channel(struct proto *p, struct channel_config *cf)
c->channel_state = CS_DOWN;
c->export_state = ES_DOWN;
c->last_state_change = current_time();
c->last_tx_filter_change = current_time();
c->reloadable = 1;
CALL(c->channel->init, c, cf);
......@@ -557,6 +558,9 @@ channel_reconfigure(struct channel *c, struct channel_config *cf)
channel_verify_limits(c);
if (export_changed)
c->last_tx_filter_change = current_time();
/* Execute channel-specific reconfigure hook */
if (c->channel->reconfigure && !c->channel->reconfigure(c, cf))
return 0;
......
......@@ -536,6 +536,7 @@ struct channel {
u8 gr_wait; /* Route export to channel is postponed until graceful restart */
btime last_state_change; /* Time of last state transition */
btime last_tx_filter_change;
};
......
......@@ -549,29 +549,29 @@ rt_notify_basic(struct channel *c, net *net, rte *new0, rte *old0, int refeed)
c->stats.exp_withdraws_received++;
/*
* This is a tricky part - we don't know whether route 'old' was
* exported to protocol 'p' or was filtered by the export filter.
* We try to run the export filter to know this to have a correct
* value in 'old' argument of rte_update (and proper filter value)
* This is a tricky part - we don't know whether route 'old' was exported to
* protocol 'p' or was filtered by the export filter. We try to run the export
* filter to know this to have a correct value in 'old' argument of rte_update
* (and proper filter value).
*
* FIXME - this is broken because 'configure soft' may change
* filters but keep routes. Refeed is expected to be called after
* change of the filters and with old == new, therefore we do not
* even try to run the filter on an old route, This may lead to
* 'spurious withdraws' but ensure that there are no 'missing
* This is broken because 'configure soft' may change filters but keep routes.
* Refeed cycle is expected to be called after change of the filters and with
* old == new, therefore we do not even try to run the filter on an old route.
* This may lead to 'spurious withdraws' but ensure that there are no 'missing
* withdraws'.
*
* This is not completely safe as there is a window between
* reconfiguration and the end of refeed - if a newly filtered
* route disappears during this period, proper withdraw is not
* sent (because old would be also filtered) and the route is
* not refeeded (because it disappeared before that).
* This is not completely safe as there is a window between reconfiguration
* and the end of refeed - if a newly filtered route disappears during this
* period, proper withdraw is not sent (because old would be also filtered)
* and the route is not refeeded (because it disappeared before that).
* Therefore, we also do not try to run the filter on old routes that are
* older than the last filter change.
*/
if (new)
new = export_filter(c, new, &new_free, 0);
if (old && !refeed)
if (old && !(refeed || (old->lastmod <= c->last_tx_filter_change)))
old = export_filter(c, old, &old_free, 1);
if (!new && !old)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment