[PATCH 1/2] PNPACPI: support shareable interrupts

[PATCH 1/2] PNPACPI: support shareable interrupts

[Bjorn Helgaas]

ACPI supplies a "shareable" indication, but PNPACPI ignores it.
If a PNP device uses a shared interrupt, request_irq() fails
because the PNP driver can't tell whether to supply SA_SHIRQ.

This patch allows PNP drivers to test
    (pnp_irq_flags(dev, 0) & IORESOURCE_IRQ_SHAREABLE)

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>

Index: work-mm7/include/linux/ioport.h
===================================================================
--- work-mm7.orig/include/linux/ioport.h	2006-06-30 11:52:21.000000000 -0600
+++ work-mm7/include/linux/ioport.h	2006-06-30 12:18:41.000000000 -0600
@@ -55,6 +55,7 @@
 #define IORESOURCE_IRQ_LOWEDGE		(1<<1)
 #define IORESOURCE_IRQ_HIGHLEVEL	(1<<2)
 #define IORESOURCE_IRQ_LOWLEVEL		(1<<3)
+#define IORESOURCE_IRQ_SHAREABLE	(1<<4)
 
 /* ISA PnP DMA specific bits (IORESOURCE_BITS) */
 #define IORESOURCE_DMA_TYPE_MASK	(3<<0)
Index: work-mm7/drivers/pnp/pnpacpi/rsparser.c
===================================================================
--- work-mm7.orig/drivers/pnp/pnpacpi/rsparser.c	2006-06-30 11:52:18.000000000 -0600
+++ work-mm7/drivers/pnp/pnpacpi/rsparser.c	2006-06-30 12:18:41.000000000 -0600
@@ -74,7 +74,7 @@
 
 static void
 pnpacpi_parse_allocated_irqresource(struct pnp_resource_table *res, u32 gsi,
-	int triggering, int polarity)
+	int triggering, int polarity, int shareable)
 {
 	int i = 0;
 	int irq;
@@ -95,6 +95,9 @@
 		return;
 	}
 
+	if (shareable)
+		res->irq_resource[i].flags |= IORESOURCE_IRQ_SHAREABLE;
+
 	res->irq_resource[i].start = irq;
 	res->irq_resource[i].end = irq;
 	pcibios_penalize_isa_irq(irq, 1);
@@ -197,7 +200,8 @@
 			pnpacpi_parse_allocated_irqresource(res_table,
 				res->data.irq.interrupts[i],
 				res->data.irq.triggering,
-				res->data.irq.polarity);
+				res->data.irq.polarity,
+				res->data.irq.sharable);
 		}
 		break;
 
@@ -263,7 +267,8 @@
 			pnpacpi_parse_allocated_irqresource(res_table,
 				res->data.extended_irq.interrupts[i],
 				res->data.extended_irq.triggering,
-				res->data.extended_irq.polarity);
+				res->data.extended_irq.polarity,
+				res->data.extended_irq.sharable);
 		}
 		break;
 

Re: [PATCH 1/2] PNPACPI: support shareable interrupts

[Andrew Morton]

Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
>
> --- work-mm7.orig/drivers/pnp/pnpacpi/rsparser.c	2006-06-30 11:52:18.000000000 -0600
> +++ work-mm7/drivers/pnp/pnpacpi/rsparser.c	2006-06-30 12:18:41.000000000 -0600

pnpacpi confuses me.  Who maintains it?  Is it considered an APCI thing, or
a PNP thing or a Bjorn thing?

Thanks.

RE: [PATCH 1/2] PNPACPI: support shareable interrupts

[Brown, Len]

>pnpacpi confuses me.  Who maintains it?  Is it considered an 
>APCI thing, or a PNP thing or a Bjorn thing?

PNP is Adam's generic Linux PNP layer.
It has "drivers" underneath is, such as pnp-isa, and pnp-bios

Shaohua invented pnp-acpi as a driver to replace pnp-bios
on ACPI-enabled systems.

So PNPACPI is the connection between ACPI and PNP,
and folks who conribute to either might contribute to it.
Most of the issues with it to date have been ACPI-specific,
rather than PNP-specific, including this one.

Bjorn, of course, is qualified to contribute to just about
any part of the kernel.  So a "Bjorn thing" could be just
about anything;-)

cheers,
-Len

Re: [PATCH 1/2] PNPACPI: support shareable interrupts

[Andrew Morton]

"Brown, Len" <len.brown@intel.com> wrote:
>
> >pnpacpi confuses me.  Who maintains it?  Is it considered an 
> >APCI thing, or a PNP thing or a Bjorn thing?
> 
> PNP is Adam's generic Linux PNP layer.
> It has "drivers" underneath is, such as pnp-isa, and pnp-bios
> 
> Shaohua invented pnp-acpi as a driver to replace pnp-bios
> on ACPI-enabled systems.
> 
> So PNPACPI is the connection between ACPI and PNP,
> and folks who conribute to either might contribute to it.
> Most of the issues with it to date have been ACPI-specific,
> rather than PNP-specific, including this one.
> 
> Bjorn, of course, is qualified to contribute to just about
> any part of the kernel.  So a "Bjorn thing" could be just
> about anything;-)
> 

OK.  But that still didn't give me a how-to-get-this-to-mainline graph.

Bjorn, are these patches considered for-2.6.18 material?  (They look like
it).

Re: [PATCH 1/2] PNPACPI: support shareable interrupts

[Bjorn Helgaas]

> OK.  But that still didn't give me a how-to-get-this-to-mainline graph.

I can't answer this, other than the obvious: it'd be good to get
acks from Adam and/or Shaohua and Russell.

> Bjorn, are these patches considered for-2.6.18 material?  (They look like
> it).

Yes, please.  I should have been more clear about that.

I'm pretty sure this problem will prevent all distro kernels from
booting on HP ia64 boxes with large I/O configs.  So I'll be
pushing it into distro updates soon.

I do wish we had more soak time, because a BIOS that incorrectly
reports a COM port interrupt as shareable could come back to
bite us.