ebtables-restore error "Bad table name 'filter'"

Status changed to 'Confirmed' because the bug affects multiple users.

This appears to exist on Xenial, Yakkety, and Zesty. Brought up a quick LXD using the ubuntu-daily images and reproduced it in each.

Is there a workaround or a way to make a file that ebtables-restore does accept, by hand for example?

@ktf the only other report of this I could find was someone doing the restore manually here:
http://knoffhoff.tumblr.com/post/59967875099/how-to-save-and-load-ebtables-rules

Use at your own risk as I have yet to test that.

# ebtables-restore < txt1.txt
Bad table name 'filter'.
# ebtables -t filter -L
Bridge table: filter

Bridge chain: INPUT, entries: 0, policy: ACCEPT

Bridge chain: FORWARD, entries: 0, policy: ACCEPT

Bridge chain: OUTPUT, entries: 0, policy: ACCEPT

So i think this is a bug, probably upstream?

Heh, looks like upstream recently deleted this file: https://git.netfilter.org/ebtables/commit/?id=826faffa38de8ce55bb32cd1549fb732229fd80f but I think it's unrelated. In any case, we should probably try to build from source on a victim LXD and debug it a bit further.

any news on this

@Nish / @Josh - you worked on this before, was there any upstream report already that might be missing or is this waiting for someone to actually report it first (also pinging on IRC about it)?

I don't believe there was an upstream report, however it looks like ebtables-restore was removed from the upstream repo here:

https://git.netfilter.org/ebtables/commit/?id=826faffa38de8ce55bb32cd1549fb732229fd80f

Some sort of initialization is missing:

96 struct ebt_u_table *ebt_find_table(const char *name)
97 {
98 struct ebt_u_table *t = ebt_tables;
99
100 while (t && strcmp(t->name, name)) {
101 printf("t->name: %s name: %s", t->name, name);
(gdb) n
98 struct ebt_u_table *t = ebt_tables;
(gdb) n
100 while (t && strcmp(t->name, name)) {
(gdb) p t
$1 = (struct ebt_u_table *) 0x0
(gdb) p name
$2 = 0x602100 <replace> "filter"
(gdb) p ebt_tables
$3 = (struct ebt_u_table *) 0x0

Status changed to 'Confirmed' because the bug affects multiple users.

It's been like 2 years, any news on this?
Also affects bionic.

This seems to be fixed in Focal, where all the tooling is based on nftables:

root@focal # ebtables-save | tee dump.txt
# Generated by ebtables-save v1.8.4 on Thu Apr 30 19:03:09 2020
*filter
:INPUT ACCEPT
:FORWARD ACCEPT
:OUTPUT ACCEPT
# Completed on Thu Apr 30 19:03:09 2020
root@focal# ebtables -F
root@focal# ebtables-restore < dump.txt
root@focal#

Hi,

I hit this old bug while doing some clearance.

Xenial reached EOSS (End of Standard Support) so I'm afraid it can't be fixed. It is unfortunate that we were unable to resolve this defect, however, there appears to be no further action possible at this time because it's not a security bug.

For Bionic the bug is still present. I took a look at the different commits between Focal's version (bug fixed) and Bionic's one here [1] but, at first glance, there is no isolated commit for the fix, so a deeper investigation has to be made to estimate how many changes it would imply without getting too close to the whole next fixed release.

https://git.netfilter.org/ebtables/log/?h=ebtables-2.0.11&showmsg=1