mdev

[Denys Vlasenko]

On Sunday 19 April 2009 16:06, Vladimir Dronnikov wrote:
> > I propose to use "mdev -<some_other_letter>".
> >
> 
> OK. Let mdev -S := mdev -s + mdev -m, where mdev -m scans for
> /sys/devices/.../modalias. Let mdev -u mimics udev slow approach: to
> push "add" to any uevent file it finds. Let new options be
> configurable.
> Do we agree with that?

So far I do not understand why would you want to
push "add" to every uevent file in /sys.
I am not saying you do not need that, I say _I_ (i.e.,
an ignoramus when it comes to this stuff) don't know why.

Adding mdev -m sounds ok.

> >> Current modprobe is happy with such arguments. We don't need any
> >> additional parsing.
> >
> > Good, that makes it rather simple thing to do.
> 
> Please, take a look at the recent patch. I placed modprobe's
> spawn_and_wait into the branch where firmware is loaded.

Eh... unfortunately Rob's patch changed mdev.c a lot...
...let me add $VAR=regex matching, and then rediff?

Ok, I added it. Try matching $MODALIAS=.* with current svn.

> If you find 
> it safe, the call can be moved out of SEQ fuser, so we won't need
> fork, just BB_EXECVP would be sufficient. Moreover, we could directly
> call modprobe_main() (if ENABLE_MODPROBE is configured), since I find
> no reasons for BB people to prefer vanilla modprobe over BB's one.

We should not *force* people to use "our" modprobe.
--
vda