I recently tried out adding PNP support to my driver to remove the
hassle of finding the correct parameters for it. This, however, causes
it to show up under the pnp bus, where as it previously was located
under the platform bus.
Is the idea that PNP devices should only reside on the PNP bus or is
there some magic available to get the device to appear on several buses?
It's a bit of a hassle to search in two different places in sysfs
depending on if PNP is used or not.
Also, the PNP bus doesn't really say that much about where the device is
physically connected. The other bus types usually give a hint about this.
Rgds
Pierre
Pierre Ossman wrote:
> I recently tried out adding PNP support to my driver to remove the
> hassle of finding the correct parameters for it. This, however, causes
> it to show up under the pnp bus, where as it previously was located
> under the platform bus.
>
> Is the idea that PNP devices should only reside on the PNP bus or is
> there some magic available to get the device to appear on several buses?
> It's a bit of a hassle to search in two different places in sysfs
> depending on if PNP is used or not.
>
> Also, the PNP bus doesn't really say that much about where the device is
> physically connected. The other bus types usually give a hint about this.
Not to take away from your question, but:
Is there "the PNP bus"? I've seen an ISA bus that (sort of)
supports PNP, PCI PNP, NuBus PNP, USB PNP, IEEE 1394 PNP, etc.
--
~Randy
Randy.Dunlap wrote:
> Pierre Ossman wrote:
>
>> I recently tried out adding PNP support to my driver to remove the
>> hassle of finding the correct parameters for it. This, however,
>> causes it to show up under the pnp bus, where as it previously was
>> located under the platform bus.
>>
>> Is the idea that PNP devices should only reside on the PNP bus or is
>> there some magic available to get the device to appear on several
>> buses? It's a bit of a hassle to search in two different places in
>> sysfs depending on if PNP is used or not.
>>
>> Also, the PNP bus doesn't really say that much about where the device
>> is physically connected. The other bus types usually give a hint
>> about this.
>
>
> Not to take away from your question, but:
> Is there "the PNP bus"? I've seen an ISA bus that (sort of)
> supports PNP, PCI PNP, NuBus PNP, USB PNP, IEEE 1394 PNP, etc.
>
It's not a physical bus but it is a bus as far as the kernel is
concerned. And that's really my problem. I want it to support PNP, but
also to associate with the physical bus it's connected to.
Rgds
Pierre
PS. Your outgoing mail server gives the wrong HELO
Hi Pierre,
The platform bus does not show the actual physical relationship either. For
x86, ACPI is typically needed to determine this. It would be easy to bind to
spawn pnp devices off of an ISA bridge device, attached to the pci bus, but
whether it's the actual physical parent would be very difficult to determine
without firmware assistance.
At the moment the pnp bus is only showing a logical bus relationship. If we
were to use ACPI to aid in the generation of the physical device tree, we
could put these devices in the correct physical location.
Thanks,
Adam
On Thu, Jan 27, 2005 at 10:16:50PM +0100, Pierre Ossman wrote:
> I recently tried out adding PNP support to my driver to remove the
> hassle of finding the correct parameters for it. This, however, causes
> it to show up under the pnp bus, where as it previously was located
> under the platform bus.
>
> Is the idea that PNP devices should only reside on the PNP bus or is
> there some magic available to get the device to appear on several buses?
> It's a bit of a hassle to search in two different places in sysfs
> depending on if PNP is used or not.
>
> Also, the PNP bus doesn't really say that much about where the device is
> physically connected. The other bus types usually give a hint about this.
It's normal for ISA devices to not tell us much about their physical
properties.
>
> Rgds
> Pierre
> -
Adam Belay wrote:
>Hi Pierre,
>
>The platform bus does not show the actual physical relationship either. For
>x86, ACPI is typically needed to determine this. It would be easy to bind to
>spawn pnp devices off of an ISA bridge device, attached to the pci bus, but
>whether it's the actual physical parent would be very difficult to determine
>without firmware assistance.
>
>At the moment the pnp bus is only showing a logical bus relationship. If we
>were to use ACPI to aid in the generation of the physical device tree, we
>could put these devices in the correct physical location.
>
>
So it is correct behaviour that the device shows up under /sys/bus/pnp
when found using PNP, and /sys/bus/platform when scanned for?
I'm trying to get it to work well with HAL and it would be nice if it
could be found in a consistent way.
Rgds
Pierre