From: "Henrik Rydberg" <rydberg@euromail.se>
Date: Tue, 19 Jun 2012 20:31:53 +0200
To: Jan Beulich <JBeulich@suse.com>
Cc: Jiri Kosina <jkosina@suse.cz>, linux-kernel@vger.kernel.org
Subject: Re: hid-generic regression on existing distros
Message-ID: <20120619183153.GA2429@polaris.bitmath.org>
References: <4FE09A3A020000780008A9D0@nat28.tlf.novell.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <4FE09A3A020000780008A9D0@nat28.tlf.novell.com>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2034
Lines: 43

Hi Jan,

> commit 8215d557e5f3a70e50e07c857d35c250fee62a73 results
> in no keyboard or mouse during early boot on systems where
> those are USB-based and HID=m, since nothing there can
> possibly be aware that this new driver needs to be manually
> loaded for e.g. interactive startup mode to be useful, as it would
> have to make it into initrd, yet the mkinitrd machinery can't
> reasonably know of it.

This is obviously distro-specific, since there are ramdisk programs
which make use of udev (mkinitcpio, for one).  The mkinitrd machinery
cannot reasonably know about HID either, and some distributions use
HID=y, which creates its own set of problems. Either way, your
question really boils down to whether module name changes and the
like, which affect manual initrd setups, are considered user-space
breakage or not. I will leave that question to Jiri.

> Imo, with the functionality having been previously provided
> by usbhid.ko, that module should cause hid-generic to be
> pulled in automatically.

The split was made especially to avoid this. The "hid" module used to
contain both the bus driver and the generic hid driver. It is now only
a bus driver, and the rest of the hid drivers optionally (and
automatically, using udev) hook on to that bus. Saying that all hid
devices carried over usb must load a certain driver is the same as
saying that all pci-based devices must be handled be the same
driver. Convenient, perhaps, but not very practical.

> (Also, just as a side note, a Kconfig default of 'y' is pointless
> for a tristate option that depends on another tristate one,
> when that one was already set to 'm' - the default really
> should be the value of the depended upon symbol.)

Thanks, overall the Kconfig seems to need updating. I will look into it.

Henrik
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/