Re: [LTP] [PATCH] fw_load: new test of device firmware loading

[<ch...@su...>]

Hi!
> >> Right, we should at least limit the module build to kernel versions that
> >> are known to work. In this particular case I would limit it to 3.7+
> > Or even better to be both forward and backward compatible, we should not
> > stop the build when kernel module has failed to build. We should carry
> > on with the build, install etc. and return TCONF when the particula test
> > was called.
> >
> As I understand from the last comment, userspace test should check 
> availability
> of module.ko. If it didn't find the module, the test must return TCONF,
> further, no need to check modules build against kernel version. Is it 
> correct?

The insmod or modprobe will return error when you try to insert kernel
module build against different kerenel (i.e. version magic is
different).

So I would do the logick as:

1. Is there foo.ko?
   
   if not -> module wasn't build -> TCONF

2. Try to insmod/modprobe module

   if failure -> most likely module not compatible with kernel -> TBROK

3. do actual test

...

-- 
Cyril Hrubis
ch...@su...