Re: pnp and acpi_pnp

Re: pnp and acpi_pnp

[Len Brown]

Matthieu,
PNPBIOS should be disabled when ACPI is enabled, and it is a bug that
this is not automatic.

yes, the "Linux PNP" layer is incomplete, and ACPI isn't yet plugged
into that.

cheers,
-Len

Re: pnp and acpi_pnp

[castet.matthieu]

Selon Len Brown <len.brown@intel.com>:

> Matthieu,
> PNPBIOS should be disabled when ACPI is enabled, and it is a bug that
> this is not automatic.
>
> yes, the "Linux PNP" layer is incomplete, and ACPI isn't yet plugged
> into that.
>
> cheers,
> -Len
>
>
>


Hi,
thanks for your reply.

What about the fact that acpi pnp don't make the difference between a normal
serial port and a ir port ?

I hope acpi will be plugged soon in Linux PNP layer.

regards,
Matthieu

RE: pnp and acpi_pnp

[castet.matthieu]

Selon "Brown, Len" <len.brown@intel.com>:

> there are PNP id's which should tell exactly what
> the device is.  For the ACPI case, you can dump
> your BIOS with acpidmp and disassemble it with iasl
> to see the PNP-ids.
>
> But Linux doesn't use ACPI's PNP-ids for device
> enumeration -- yet.
>
Ok, thanks.

> >> PNPBIOS should be disabled when ACPI is enabled, and it is a bug that
> >> this is not automatic.
Yes it sould.
Also with 2.6.9 acpi pnp and pnp will conflic :
the new acpi driver drivers/acpi/motherboard.c reserve the motherboard
resources, but if you look at driver/pnp/system.c, there is also a pnp driver
that does the same jobs. So if PNP0c0[12] appear in pnp layer (with pnpbios),
both driver try to allocate the ressource.
Hopefully it cause only warnings.
But it show the problem and the need of unification instead of reimplant things.

> cheers,
> -Len
>
regards
Matthieu

> >-----Original Message-----
> >From: castet.matthieu@free.fr [mailto:castet.matthieu@free.fr]
> >Sent: Thursday, September 02, 2004 7:38 AM
> >To: Brown, Len
> >Cc: linux-kernel@vger.kernel.org
> >Subject: Re: pnp and acpi_pnp
> >
> >Selon Len Brown <len.brown@intel.com>:
> >
> >> Matthieu,
> >> PNPBIOS should be disabled when ACPI is enabled, and it is a bug that
> >> this is not automatic.
> >>
> >> yes, the "Linux PNP" layer is incomplete, and ACPI isn't yet plugged
> >> into that.
> >>
> >> cheers,
> >> -Len
> >>
> >>
> >>
> >
> >
> >Hi,
> >thanks for your reply.
> >
> >What about the fact that acpi pnp don't make the difference
> >between a normal
> >serial port and a ir port ?
> >
> >I hope acpi will be plugged soon in Linux PNP layer.
> >
> >regards,
> >Matthieu
> >
>

pnp and acpi_pnp

[castet.matthieu]

Hello,

in pnpbios Kconfig, it is said that acpi will  supersede PNPBIOS some day.
But acpi_pnp seem incomplete :
- It doesn't have got a sysfs entry that let the user see the device he has, and
let him change some parameters.
- You can't export pnp alias that let some script (like hotplug) autoload the
modules you need.
- Also it seem less precise than pnpbios detection (might be a acpi bug in my
computer...) : on my laptop I have a serial port (PNP0501) and a ir port
(PNP0510). The 8250_pnp driver only find the serial port, but the 8250_acpi
find both ports, but it is not good because I can't load the specific ir driver
without unloading the serial modules...
- acpi_pnp doesn't also offer generic function
(pnp_{port,irq,dma,...}_{start,len,valid,..}) that permit easy discovering of
the ressource need by the device and don't need specific handler like in
acpi_pnp.

Also if you want support pnpbios and acpi_pnp (not eveybody had a working
acpi...) you had to duplicate your code (and even do 2 modules like serial
module does). If acpi_pnp id will be exported it would be even worse because
the 2 drivers will register the same device, and because pnpbios automatiquely
disable device (and free the resource it uses) if the probe fail, it could be
very problematic...

So could it be possible to integreate acpi_pnp in pnp protocol that already
supprt pnpbios and isapnp ?
Also I believe it must do before too much driver use acpi_pnp.

Thanks.

Matthieu

PS : for the people that think that pnp is useless, they can look for example
how some alsa driver try to manage mpu and joystick without it (for example in
intel8x0.c, they register the LPC (hopefully some clever driver like i8xx_tco
or hw_random that use also the lpc don't register it...), for the jostick they
duplicate generic code from ns558 that can autodetect the port via pnp and find
the good one (alsa driver alway thinks the gameport is on io 0x200, but in
reality it is on io 0x201 on my computer...), it is the same for mpu and the
generic module snd-mpu401(only support acpi pnp, but I had a version that
support pnpbios), and the intel8x0.c don't handle mpu irq whereas snd-mpu401
does...

PS2 : please CC me since I'm not subscribed to lkml.