[lug] Kernel Module Compile Config Q

stimits at comcast.net stimits at comcast.net
Sun Jun 29 19:06:39 MDT 2014


The kernel in question is 3.10.24-g6a2d13a from Ubuntu 14.04/LTS, but configured and published for an embedded distribution (nVidia's "L4T" Linux For Tegra). This particular system is an SMP ARM15, so stack traces may not be what would be expected.
 
In this case, the error is a paging request at a non-NULL address...so I know it is accessing memory incorrectly. If any network connection exists, this happens upon insmod of the bridge.ko module. If no network connection exists, the module can be inserted...but then any attempt to use a network connection causes this anyway.
 
I guess what I'm really trying to do is avoid flashing the kernel until I've verified that turning on the bridge module option (which in turn triggers a non-module option) invalidates this module's idea of where to access a built-in function. The kernel is configured exactly as the original from the exact source tree...except for this module and its dependencies. I'm thinking that if all dependencies were modules too, then function access would be valid, but this one non-module config difference with the running kernel has caused this. I'm probably going to have to bite the bullet and flash a new kernel even though the network bridge itself is a module.
 
----- Original Message -----From: Vishal Verma <stellarhopper at gmail.com>To: Boulder (Colorado) Linux Users Group -- General Mailing List <lug at lug.boulder.co.us>Sent: Sat, 28 Jun 2014 02:23:19 -0000 (UTC)Subject: Re: [lug] Kernel Module Compile Config Q
 
On Fri, Jun 27, 2014 at 5:33 PM, <stimits at comcast.net> wrote:> I'm wondering about a module compile for network bridging (3.10 kernel)> which results in a kernel panic. I'm thinking the panic is because of a> config change dependency which might demand rebuilding the whole kernel. I> want to know if this dependency and sticking to only module compiles without> a new kernel compile would be the cause. Can I get away without rebuilding> and installing a new kernel?
 
It would help to post the stack trace from the panic, and the kernelsource tree (is it a standard distro kernel, or a vanilla 3.10...
 
> .> This is the running kernel source tree and entirely configured based on> current running system (I have the config in /boot option), with proper> preparation before compile (compile itself is on the native system). Then I> add bridge support as a module (via make menuconfig). Intentionally building> CONFIG_BRIDGE=m results in a couple of other dependencies which builds> modules that insmod without error, but one non-optional dependency is> CONFIG_BRIDGE_NETFILTER=Y, so I cannot insmod this. With module dependencies> loaded via insmod, I then insmod bridge.ko...kernel panic.
 
Sounds like it could be a bug in the init path of the module, again,the stack trace would help.
 
> .> The dependency which is forced as a non-module is CONFIG_BRIDGE_NETFILTER=Y,> and I'm guessing this is why I won't be able to stick to a pure module> build. In general, if I build something as a module, and if that module> forces some other CONFIG_WHATEVER=Y, is this always asking for trouble? I'd
 
As far as I know, this itself shouldn't be a problem.
 
> really like to stick to not flashing the kernel image itself on this> embedded system.
 
-Vishal_______________________________________________Web Page: http://lug.boulder.co.usMailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lugJoin us on IRC: irc.hackingsociety.org port=6667 channel=#hackingsociety
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lug.boulder.co.us/pipermail/lug/attachments/20140630/e8943cc5/attachment.html>


More information about the LUG mailing list