Re: [PATCH kmod] libkmod: add user soft dependecies

[Jose Ignacio Tornos Martinez <jtornosm@redhat.com>]

> a night of sleep and I had a dream in which libkmod had the concept of
> "weak dependency". Borrowing the concept from weak symbols, I think it
> makes perfect sense... the symbol is there and it may or may not be used
> by the linker at the end, but the symbol needs to be there until the
> linking phase. At least the parallel makes sense in my head :)
Ok, I like your dream :)

> Also, I don't think we should mix them with softdep like is done here
> after a quick scan through the patch.
Ok, understood, and if a new case for softdep is not going to be used,
it is clearer: better not mixing with softdep processing.

> From man page:
>    softdep modulename pre: modules... post: modules...
>           The softdep command allows you to specify soft, or optional,
>           module dependencies.  modulename can be used without these
>           optional modules installed, but usually with some features
>           missing. For example, a driver for a storage HBA might
>           require another module be loaded in order to use management
>           features.
>
>           pre-deps and post-deps modules are lists of names and/or
>           aliases of other modules that modprobe will attempt to
>           install (or remove) in order before and after the main module
>           given in the modulename argument.
>
>           Example: Assume "softdep c pre: a b post: d e" is provided in
>           the configuration. Running "modprobe c" is now equivalent to
>           "modprobe a b c d e" without the softdep. Flags such as
>           --use-blacklist are applied to all the specified modules,
>           while module parameters only apply to module c.
>
>           Note: if there are install or remove commands with the same
>           modulename argument, softdep takes precedence.
>
>        weakdep modulename modules...
>           The weakdep command allows you to specify weak module
>           dependecies. Those are similar to pre softdep, with the
>           difference that userspace doesn't attempt to load that
>           dependency before the specified module. Instead the kernel
>           may request one or multiple of them during module probe,
>           depending on the hardware it's binding to. The purpose of
>           weak module is to allow a driver to specify that a certain
>           dependency may be needed, so it should be present in the
>           filesystem (e.g. in initramfs) when that module is probed.
>
>           Example: Assume "weakdep c a b". A program creating an
>           initramfs knows it should add a, b, and c to the filesystem
>           since a and b may be required/desired at runtime. When c is
>           loaded and is being probed, it may issue calls to
>           request_module() causing a or b to also be loaded.
Ok, thanks for completing this.
I will include this in my kmod patch (if it is ok for you).

> Also instead of delegating this to the distros, it'd be good if we start
> adding those to the ELF section of the modules with
>
>        MODULE_WEAKDEP("...");
>
> ... to be defined in the kernel in a similar way that MODULE_SOFTDEP()
> is.
Agree, better to define in kernel code, that's the reason for the patch.
Ok, I will implement in that way and I will create a kernel patch too for
this.
Indeed (with a different name), it was also in my mind but I didn't dare to
create something "new".

Thanks for you comments and help

Best regards
José Ignacio