frr/master :crash vnc at startup

[pguibert6WIND]

compilation with enable-cumulus:

==== crash seen:

Program received signal SIGSEGV, Segmentation fault.
hash_get (hash=0x0, data=data@entry=0x7fffffffcf30, alloc_func=alloc_func@entry=0x0) at hash.c:160
160 key = (*hash->hash_key)(data);
(gdb) bt
#0 hash_get (hash=0x0, data=data@entry=0x7fffffffcf30, alloc_func=alloc_func@entry=0x0) at hash.c:160
#1 0x00007ffff7b84bca in route_node_get (table=0xae7b18, pu=..., pu@entry=...) at table.c:310
#2 0x0000000000472f8e in vnc_nve_group_prefix (self=, vty=0xaa9d40, argc=, argv=0xaa7230) at rfapi/bgp_rfapi_cfg.c:2536
#3 0x00007ffff7b7c10d in cmd_execute_command_real (vline=vline@entry=0xaa2690, vty=vty@entry=0xaa9d40, cmd=cmd@entry=0x0, filter=FILTER_STRICT) at command.c:935
#4 0x00007ffff7b7dc79 in cmd_execute_command_strict (vline=vline@entry=0xaa2690, vty=vty@entry=0xaa9d40, cmd=cmd@entry=0x0) at command.c:1030
#5 0x00007ffff7b7dcb6 in command_config_read_one_line (vty=vty@entry=0xaa9d40, cmd=cmd@entry=0x0, use_daemon=use_daemon@entry=0) at command.c:1061
#6 0x00007ffff7b7dde7 in config_from_file (vty=vty@entry=0xaa9d40, fp=fp@entry=0xaa9b00, line_num=line_num@entry=0x7fffffffd078) at command.c:1106
#7 0x00007ffff7b73f63 in vty_read_file (confp=0xaa9b00) at vty.c:2200
#8 vty_read_config (config_file=, config_default_dir=config_default_dir@entry=0x7094e0 <config_default> "/usr/local/etc//bgpd.conf") at vty.c:2375
#9 0x00007ffff7ba3d7c in frr_config_fork () at libfrr.c:382
#10 0x0000000000419b24 in main (argc=1, argv=0x7fffffffe2a8) at bgp_main.c:408
(gdb)
==== config used:
password zebra
log stdout
service advanced-vty
!
router bgp 64512
bgp router-id 192.168.1.101
neighbor 192.168.1.100 remote-as 64512
neighbor 192.168.1.104 remote-as 64512
address-family vpnv4
neighbor 192.168.1.100 activate
neighbor 192.168.1.104 activate
exit-address-family
vnc nve-group group1
prefix vn 172.16.0.0/17
rd 64512:1
response-lifetime 200
rt both 1000:1 1000:2
exit-vnc
exit

[eqvinox]

very likely (99,999%) caused by #794
maybe VNC is doing something weird with table initialisation...

first: did you do make clean for the entire code and recompile?

[pguibert6WIND]

I reconfigured with enable-cumulus
I did not make make clean.
i will recheck

[eqvinox]

problem is that in:

struct rfapi_cfg {                                                              
...
	struct route_table nve_groups_vn[AFI_MAX];
	struct route_table nve_groups_un[AFI_MAX];

These two fields are not allocated with route_table_init_with_delegate(), instead it just does this in bgp_rfapi_cfg_new():

	for (afi = AFI_IP; afi < AFI_MAX; afi++) {
		/* ugly, to deal with addition of delegates, part of 0.99.24.1
		 * merge */
		h->nve_groups_vn[afi].delegate =
			route_table_get_default_delegate();
		h->nve_groups_un[afi].delegate =
			route_table_get_default_delegate();
	}

=> the new hash field in the route_table will be NULL => SEGV

The "correct" solution would be to change from

	struct route_table nve_groups_vn[AFI_MAX];

to

	struct route_table *nve_groups_vn[AFI_MAX];

and use route_table_init_with_delegate (+ route_table_finish() to free)

(I tried to find such uses of route_table when making the change, but I didn't spot this one)

[qlyoung]

@louberger want to take a look at this?

[louberger]

sure - will try early next week if not before

…

On 7/18/2017 5:10 PM, Quentin Young wrote: @louberger <https://github.com/louberger> want to take a look at this? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#825 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGRWyPaHxedXEeDKWabyT02LZOfs7uH7ks5sPMrVgaJpZM4OaBmF>.

[eqvinox]

fixed by 0ae6124, closing