askrene crash when loading node bias descriptions

[endothermicdev]

When starting the node, it shuts down due to askrene exiting before replying to init.

2026-07-30T23:23:55.804Z INFO plugin-cln-askrene: Killing plugin: exited before replying to init
2026-07-30T23:23:55.805Z BROKEN plugin-cln-askrene: Plugin marked as important, shutting down lightningd!

Attaching gdb to the core dump shows:

  __pthread_kill_implementation (no_tid=0, signo=6, threadid=<optimized out>) at ./nptl/pthread_kill.c:44
#1  __pthread_kill_internal (signo=6, threadid=<optimized out>) at ./nptl/pthread_kill.c:78
#2  __GI___pthread_kill (threadid=<optimized out>, signo=signo@entry=6) at ./nptl/pthread_kill.c:89
#3  0x00007c180ec4527e in __GI_raise (sig=sig@entry=6) at ../sysdeps/posix/raise.c:26
#4  0x00007c180ec288ff in __GI_abort () at ./stdlib/abort.c:79
#5  0x00006317663af6e0 in call_error (msg=msg@entry=0x6317664c7c36 "Not a valid header") at ccan/ccan/tal/tal.c:95
#6  0x00006317663af789 in check_bounds (p=<optimized out>) at ccan/ccan/tal/tal.c:169
#7  0x00006317663af7ba in to_tal_hdr (ctx=0x63179352c3d8) at ccan/ccan/tal/tal.c:179
#8  0x00006317663b0727 in tal_resize_ (ctxp=ctxp@entry=0x7ffdf6cc1098, size=<optimized out>, size@entry=1, count=count@entry=40, clear=clear@entry=false) at ccan/ccan/tal/tal.c:750
#9  0x00006317663b0a38 in tal_dup_ (ctx=ctx@entry=0x631793529a18, p=<optimized out>, p@entry=0x63179352c3d8, size=size@entry=1, n=<optimized out>, extra=extra@entry=0, nullok=nullok@entry=false, label=0x6317664c4322 "char[]")
    at ccan/ccan/tal/tal.c:854
#10 0x00006317663aedd8 in tal_strdup_ (ctx=ctx@entry=0x631793529a18, p=p@entry=0x63179352c3d8 "3 failures in current renerebalance run", label=label@entry=0x6317664c4322 "char[]") at ccan/ccan/tal/str/str.c:17
#11 0x0000631766387cd5 in tal_strdup_or_null (ctx=ctx@entry=0x631793529a18, str=str@entry=0x63179352c3d8 "3 failures in current renerebalance run") at common/utils.c:192
#12 0x000063176636708d in set_node_bias (layer=layer@entry=0x6317933b2558, node=node@entry=0x7ffdf6cc11a0, description=0x63179352c3d8 "3 failures in current renerebalance run", bias_factor=bias_factor@entry=-10 '\366', 
    relative=relative@entry=false, dir_out=dir_out@entry=true, timestamp=1785452703) at plugins/askrene/layer.c:360
#13 0x0000631766367281 in load_node_bias (plugin=<optimized out>, layer=layer@entry=0x6317933b2558, cursor=cursor@entry=0x7ffdf6cc1208, len=len@entry=0x7ffdf6cc1210) at plugins/askrene/layer.c:657
#14 0x0000631766367f34 in populate_layer (askrene=askrene@entry=0x6317933a92c8, layername=<optimized out>, data=<optimized out>, data@entry=0x631793528f08 "") at plugins/askrene/layer.c:783
#15 0x00006317663680d3 in load_layers (askrene=askrene@entry=0x6317933a92c8, init_cmd=init_cmd@entry=0x6317933a9ce8) at plugins/askrene/layer.c:826
#16 0x00006317663641f3 in init (init_cmd=0x6317933a9ce8, buf=<optimized out>, config=<optimized out>) at plugins/askrene/askrene.c:1413
#17 0x0000631766376b17 in handle_init (cmd=cmd@entry=0x6317933a9ce8, 
    buf=buf@entry=0x7c180e9ff0b6 "\n\n{\"jsonrpc\":\"2.0\",\"id\":\"cln:init#43400\",\"method\":\"init\",\"params\":{\"options\":{\"askrene-timeout\":10},\"configuration\":{\"lightning-dir\":\"/mnt/hdd/bitcoin/.lightning/bitcoin\",\"rpc-file\":\"lightning-rpc\",\"s"..., params=params@entry=0x6317933ac868) at plugins/libplugin.c:1556
#18 0x0000631766376e3d in ld_command_handle (plugin=plugin@entry=0x6317933a9498, 
    buffer=0x7c180e9ff0b6 "\n\n{\"jsonrpc\":\"2.0\",\"id\":\"cln:init#43400\",\"method\":\"init\",\"params\":{\"options\":{\"askrene-timeout\":10},\"configuration\":{\"lightning-dir\":\"/mnt/hdd/bitcoin/.lightning/bitcoin\",\"rpc-file\":\"lightning-rpc\",\"s"..., toks=<optimized out>) at plugins/libplugin.c:2062
#19 0x00006317663773be in ld_read_json (conn=0x6317933ab138, plugin=0x6317933a9498) at plugins/libplugin.c:2231
#20 0x00006317663a8b99 in next_plan (conn=conn@entry=0x6317933ab138, plan=plan@entry=0x6317933ab158) at ccan/ccan/io/io.c:60
#21 0x00006317663a906a in do_plan (conn=conn@entry=0x6317933ab138, plan=plan@entry=0x6317933ab158, idle_on_epipe=idle_on_epipe@entry=false) at ccan/ccan/io/io.c:422
#22 0x00006317663a9127 in io_ready (conn=conn@entry=0x6317933ab138, pollflags=pollflags@entry=1) at ccan/ccan/io/io.c:439
#23 0x00006317663aab35 in io_loop (timers=timers@entry=0x6317933a9590, expired=expired@entry=0x7ffdf6cc14e8) at ccan/ccan/io/poll.c:470
#24 0x000063176637782f in plugin_main (argv=argv@entry=0x7ffdf6cc17d8, init=init@entry=0x631766364113 <init>, data=0x6317933a92c8, restartability=restartability@entry=PLUGIN_RESTARTABLE, init_rpc=init_rpc@entry=true, 
    features=features@entry=0x0, commands=0x63176661e7e0 <commands>, num_commands=14, notif_subs=0x0, num_notif_subs=0, hook_subs=0x0, num_hook_subs=0, notif_topics=0x0, num_notif_topics=0) at plugins/libplugin.c:2429
#25 0x00006317663654d3 in main (argc=<optimized out>, argv=0x7ffdf6cc17d8) at plugins/askrene/askrene.c:1427

It might be due to a double take() on the same string in load_node_bias:

lightning/plugins/askrene/layer.c

Lines 770 to 782 in 372d8f7

	if (fromwire_dstore_node_bias(tmpctx, cursor, len,
	&node,
	&description,
	&in_bias,
	&out_bias,
	&timestamp)) {
	set_node_bias(layer, &node, take(description), in_bias,
	/* relative = */ false,
	/* out dir = */ false, timestamp);
	set_node_bias(layer, &node, take(description), out_bias,
	/* relative = */ false,
	/* out dir = */ true, timestamp);
	}

The description really only needs to be loaded once anyway. For context, I've been working on a plugin that adds a lot of layer data to askrene and sometimes biasing the whole node is more efficient than individual channels.

[vincenzopalazzo]

This should be the fix #9434 that reproduce it maybe @Lagrang3 can confirm if it is the right fix.