From: Sebastian Benoit <benno@openbsd.org>
Subject: Re: bgpd: add missing error check for IMSG_RECONF_ROA_ITEM
To: tech@openbsd.org
Cc: cjeker@diehard.n-r-g.com
Date: Wed, 12 Nov 2025 17:29:36 +0100

Claudio Jeker(cjeker@diehard.n-r-g.com) on 2025-11-12:
> Somehow the check for trie_roa_add failure was lost.
> We do the same just log the fact handling in IMSG_RECONF_PREFIX_SET_ITEM.
> This should only trigger in out of memory situations.
> 
> Fixes CID 492365
> -- 
> :wq Claudio
> 
> Index: rde.c
> ===================================================================
> RCS file: /cvs/src/usr.sbin/bgpd/rde.c,v
> diff -u -p -r1.662 rde.c
> --- rde.c	12 Nov 2025 15:17:43 -0000	1.662
> +++ rde.c	12 Nov 2025 15:59:28 -0000
> @@ -1144,6 +1144,9 @@ rde_dispatch_imsg_parent(struct imsgbuf 
>  			if (imsg_get_data(&imsg, &roa, sizeof(roa)) == -1)
>  				fatalx("IMSG_RECONF_ROA_ITEM bad len");
>  			rv = trie_roa_add(&last_prefixset->th, &roa);
> +			if (rv == -1)
> +				log_warnx("trie_roa_add %s failed",
> +				    log_roa(&roa));
>  			break;
>  		case IMSG_RECONF_PREFIX_SET_ITEM:
>  			if (imsg_get_data(&imsg, &psi, sizeof(psi)) == -1)
> 

ok benno@

maybe related: i notice that in rde_dispatch_imsg_rtr() under
IMSG_RECONF_ROA_ITEM there is a similar pattern where you dont use
log_roa() but handroll the printing of a prefix.