Re: [Fwd: Re: [PATCH] acerhdf: Limit modalias matching to supported boards]

[Bjorn Helgaas <bjorn.helgaas@hp.com>]

On Monday 05 October 2009 09:56:51 am Stefan Bader wrote:
> Bjorn Helgaas wrote:
> > On Monday 05 October 2009 06:05:54 am Stefan Bader wrote:
> >>>>>> <stefan.bader@canonical.com> wrote:
> >>>>>>>>> I search my acerhdf-inbox for the people who sent me the BIOS 
> >>>>>>>>> versions,
> >>>>>>>> maybe they still have the netbook and can me send the modalias 
> >>>>>>>> content. I'll
> >>>>>>>> hopefully find some time upcoming weekend.
> >>>>>>>>
> >>>>>>>> But, can't we simply assume, following dmi lines?
> >>>>>>>>
> >>>>>>>> MODULE_ALIAS("dmi:*:*Acer*:*AOA*:");
> >>>>>>>> MODULE_ALIAS("dmi:*:*Gateway*:*AOA*:");
> >>>>>>>> MODULE_ALIAS("dmi:*:*Packard Bell*:*AOA*:");
> >>>>>>>> MODULE_ALIAS("dmi:*:*Packard Bell*:*DOA*:");
> >>>>>>>>
> >>>>>>>> We have exactly those constellations within the BIOS settings 
> >>>>>>>> table. If I
> >>>>>>>> interpret the modalias line correctly, "pnAOA110" means "Product Name
> >>>>>>>> AOA110" and that's what we have already in the BIOS settings 
> >>>>>>>> table. So I
> >>>>>>>> think we do already have all information we need to create a 
> >>>>>>>> complete patch
> >>>>>>>> for the problem, or am I wrong?
> >>>>>>> Right pn means that. I would strongly believe the above changes 
> >>>>>>> should be
> >>>>>>> good. I sometimes tend to be over-cautious. Attaching a refreshed 
> >>>>>>> patch
> >>>>>>> which includes them all.
> >>>>>> Well, acerhdf.c and dmi-id.c both do
> >>>>>> dmi_get_system_info(DMI_PRODUCT_NAME) when querying product name which
> >>>>>> means that the product names in the acerhdf table and the dmi-id ones
> >>>>>> are actually one and the same thing.
> > 
> > This smells like a maintenance problem.  The driver only works on the
> > specific machines and BIOS versions compiled into it.  That means you
> > have to update the driver every time a new machine comes out.  A major
> > goal of ACPI is to avoid stuff like that.
> > 
> > Are you sure there's no way to make this driver work in a generic way,
> > e.g., by claiming an ACPI device?
> 
> The driver is currently only targeted at a small set of netbooks and even there 
> some registers have changed between sub models or bios versions.
> The functions are implemented by poking and reading bits from the EC. No acpi 
> interface to which one relate as far as I know. Maybe wmi, but from the 
> experiences with acer-wmi, I doubt anything works. At least acer-wmi now 
> explicitly blacklist these models for uselessness.
> So the driver performs very strict bios version checks and every unknown 
> version get a big warning printed. Which you do not want to be issued on every 
> hardware of one of these vendors. Which happens with the driver as it is now.

I understand the motivation for the current patch, and I'm not asking
about it in particular.  My concern is about the underlying structure
of the driver.

I'm just asking why we would want a driver that requires an update
for every new machine.  That sort of ongoing maintenance is a good
sign that we're doing something wrong.
 
Bjorn