[LEDE-DEV] Difference between AutoLoad vs. AutoProbe for kernel modules?

[Baptiste Jonglez]

On Tue, Jan 10, 2017 at 07:11:56PM +0800, Yousong Zhou wrote:
> On 10 January 2017 at 18:27, John Crispin <john at phrozen.org> wrote:
> > correct, which is why we tend to add subsystem and lib stuff using
> > AutoLoad and the rest using AutoProbe
> 
> I think the problem with wireguard in github issue 3790 [1] is that
> wireguard requires NFPROTO_{IPV4,IPV6} "hashlimit" module and
> netfilter subsystem will try to load ipt_hashlimit and ip6t_hashlimit
> which are aliases of xt_hashlimit, but we have aliases along with
> other modinfoinfo stripped from kernel modules with patch
> 204-module_strip.patch.

In this case, I think the issue is different: wireguard does not
seem to explicitly depend on xt_hashlimit:

$ modinfo /lib/modules/4.9.0-1-ARCH/kernel/net/wireguard.ko
depends:        udp_tunnel,ip6_udp_tunnel,x_tables

$ modinfo /lib/modules/4.9.0-1-ARCH/kernel/net/netfilter/x_tables.ko.gz
description:    {ip,ip6,arp,eb}_tables backend module

$ modinfo /lib/modules/4.9.0-1-ARCH/kernel/net/netfilter/xt_hashlimit.ko.gz
alias:          ip6t_hashlimit
alias:          ipt_hashlimit
depends:        x_tables

Jason, what do you think of the github issue linked below?  Should
wireguard explicitly depend on xt_hashlimit?  How do you handle the
dependency chain on other platforms?

> [1] https://github.com/openwrt/packages/issues/3790

Baptiste
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/lede-dev/attachments/20170111/87396200/attachment.sig>