On Thu, Jul 12, 2012 at 11:03:55AM -0500, Bruno Wolff III wrote:
On Thu, Jul 12, 2012 at 11:35:46 -0400,
Bill Nottingham <notting(a)redhat.com> wrote:
>Bruno Wolff III (bruno(a)wolff.to) said:
>>Under the 3.5 kernel I need to have the hid_generic module loaded
>>for my USB keyboards to work. I didn't with the 3.4 kernels. This
>>caused a problem with being able to be able to enter the password
>>for my encrypted rootfs.
>Why not just build it in, then?
Because at first I didn't know what the problem was. This could also
could conceivably be a problem for fresh installs in a future where
we are using 3.5 kernels for f17, but the module isn't included by
Bill meant "why not build it into the main kernel" e.g. not as a module.
Not "build it into the initramfs".
For now I added hid_generic to the list of kernel modules to
include. I expect there will be a new dracut in rawhide soon as
there have been a lot of releases lately.
The point of the original message was to suggest that if we use 3.5
kernels for F17, I think it would be a good idea to change the f17
version of dracut to include hid_generic by default. It isn't
obvious to me that that will happen.
That it's getting built as a module at all is a bit concerning to me.
The Kconfig values for HID_GENERIC look like this:
tristate "Generic HID driver"
depends on HID
so it should default to whatever option CONFIG_HID is set to. Now,
admittedly we have CONFIG_HID=m in config-generic, but once that is put
through the config file munging script, we wind up with:
(and a general mismash of =m/=y options for HID_*)
So, something is ever so helpfully overriding our CONFIG_HID=m setting
and forcing it to be =y, but maybe that is done after kconfig has parsed
the hid Kconfig file and so HID_GENERIC is left =m.
Looks like that something is our CONFIG_USB_HID=y setting, which does a
select HID and forces it to =y.
Dave, Justin, should we explicitly set CONFIG_HID=y and
CONFIG_HID_GENERIC=y in config-generic and avoid this problem entirely?