BGP uses lp_save / lp_restore instead of linpool flushing

It is too cryptic to flush tmp_linpool in these cases and we don't want
anybody in the future to break this code by adding an allocation
somewhere which should persist over that flush.

Saving and restoring linpool state is safer.

diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 7e978919483de7f43222cbb805a2b08b5ab637e0..a805cdf0fba86c0b4f99693c3661ddda6fecc13f 100644 (file)
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -2323,6 +2323,9 @@ bgp_create_update(struct bgp_channel *c, byte *buf)
 
 again: ;
 
+  struct lp_state tmpp;
+  lp_save(tmp_linpool, &tmpp);
+
   /* Initialize write state */
   struct bgp_write_state s = {
     .proto = p,
@@ -2353,6 +2356,7 @@ again: ;
     if (EMPTY_LIST(buck->prefixes))
     {
       bgp_free_bucket(c, buck);
+      lp_restore(tmp_linpool, &tmpp);
       goto again;
     }
 
@@ -2366,7 +2370,10 @@ again: ;
       bgp_defer_bucket(c, buck);
 
     if (!res)
+    {
+      lp_restore(tmp_linpool, &tmpp);
       goto again;
+    }
 
     goto done;
   }
@@ -2377,7 +2384,7 @@ again: ;
 done:
   BGP_TRACE_RL(&rl_snd_update, D_PACKETS, "Sending UPDATE");
   p->stats.tx_updates++;
-  lp_flush(s.pool);
+  lp_restore(tmp_linpool, &tmpp);
 
   return res;
 }
@@ -2506,6 +2513,9 @@ bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len)
 
   bgp_start_timer(conn->hold_timer, conn->hold_time);
 
+  struct lp_state tmpp;
+  lp_save(tmp_linpool, &tmpp);
+
   /* Initialize parse state */
   struct bgp_parse_state s = {
     .proto = p,
@@ -2587,7 +2597,7 @@ bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len)
 
 done:
   rta_free(s.cached_rta);
-  lp_flush(s.pool);
+  lp_restore(tmp_linpool, &tmpp);
   return;
 }