[PATCH] CONFIG_PCMCIA_PROBE fix

[PATCH] CONFIG_PCMCIA_PROBE fix

[Andres Salomon]

[-- Attachment #1.1: Type: text/plain, Size: 795 bytes --]

Some time ago, Russell King submitted a patch to use CONFIG_PCMCIA_PROBE
instead of CONFIG_ISA in pcmcia probing code.  Unfortunately,
CONFIG_PCMCIA_PROBE still is only set if CONFIG_ISA is set.  This means
that if ISA isn't enabled, certain things break in 2.6; for example, my
pcmcia nic/modem (using pcnet_cs/serial_cs).  These worked fine in 2.4;
I tracked the behavior to the fact that if irq_mask is set on a pcmcia
socket (instead of pci_irq), and PCMCIA_PROBE isn't set,
pcmcia_request_irq refuses to assign an irq.  Most of the pcmcia bridges
appear to set an irq_mask, so the attached patch changes Kconfig to set
CONFIG_PCMCIA_PROBE if any of those bridges are selected.

Please apply this (or an alternative fix), as it fixes a 2.6 regression
in pcmcia functionality.


[-- Attachment #1.2: 000-pcmcia_probe.patch --]
[-- Type: text/x-patch, Size: 1186 bytes --]

Revision: linux--mainline--2.6--patch-2
Archive: dilinger@voxel.net--2003-spiral
Creator: Andres Salomon <dilinger@voxel.net>
Date: Sun Dec 21 03:05:27 EST 2003
Standard-date: 2003-12-21 08:05:27 GMT
Modified-files: drivers/pcmcia/Kconfig
New-patches: dilinger@voxel.net--2003-spiral/linux--mainline--2.6--patch-2
Summary: Enable PCMCIA_PROBE on pcmcia bridges
Keywords: 

Current behavior only enables CONFIG_PCMCIA_PROBE if CONFIG_ISA is set.  Some
bridges (for example, yenta) break badly w/out PCMCIA_PROBE; w/out it,
pcmcia_request_irq() fails if an irq_mask is set.  The idea is to decouple
CONFIG_ISA from CONFIG_PCMCIA_PROBE; so, what I've done is to enable
PCMCIA_PROBE for bridges that set irq_masks for their sockets.

* added files

    {arch}/linux/linux--mainline/linux--mainline--2.6/dilinger@voxel.net--2003-spiral/patch-log/patch-2

* modified files

--- orig/drivers/pcmcia/Kconfig
+++ mod/drivers/pcmcia/Kconfig
@@ -103,7 +103,7 @@
 
 config PCMCIA_PROBE
 	bool
-	default y if ISA && !ARCH_SA1100 && !ARCH_CLPS711X
+	default y if !ARCH_SA1100 && !ARCH_CLPS711X && (ISA || YENTA || TCIC || HD64465_PCMCIA || I82365)
 
 endmenu
 




[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [PATCH] CONFIG_PCMCIA_PROBE fix

[Russell King]

On Mon, Dec 22, 2003 at 12:48:44AM -0500, Andres Salomon wrote:
> Some time ago, Russell King submitted a patch to use CONFIG_PCMCIA_PROBE
> instead of CONFIG_ISA in pcmcia probing code.  Unfortunately,
> CONFIG_PCMCIA_PROBE still is only set if CONFIG_ISA is set.  This means
> that if ISA isn't enabled, certain things break in 2.6; for example, my
> pcmcia nic/modem (using pcnet_cs/serial_cs).  These worked fine in 2.4;
> I tracked the behavior to the fact that if irq_mask is set on a pcmcia
> socket (instead of pci_irq), and PCMCIA_PROBE isn't set,
> pcmcia_request_irq refuses to assign an irq.  Most of the pcmcia bridges
> appear to set an irq_mask, so the attached patch changes Kconfig to set
> CONFIG_PCMCIA_PROBE if any of those bridges are selected.
> 
> Please apply this (or an alternative fix), as it fixes a 2.6 regression
> in pcmcia functionality.

Please don't.  David Hinds has a better fix for this, which changes
the way we handle the allocation of IRQs.  David's change is all
round a far better way to handle the problem - if all ISA interrupts
are used or unavailable, we fall back to using the PCI interrupt
instead.

Please also note that there /is/ a PCMCIA list which patches should
be forwarded to - linux-pcmcia which is at lists.infradead.org

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core

Re: [PATCH] CONFIG_PCMCIA_PROBE fix

[Andres Salomon]

[-- Attachment #1: Type: text/plain, Size: 1524 bytes --]

On Mon, 2003-12-22 at 04:48, Russell King wrote:
> On Mon, Dec 22, 2003 at 12:48:44AM -0500, Andres Salomon wrote:
> > Some time ago, Russell King submitted a patch to use CONFIG_PCMCIA_PROBE
> > instead of CONFIG_ISA in pcmcia probing code.  Unfortunately,
> > CONFIG_PCMCIA_PROBE still is only set if CONFIG_ISA is set.  This means
> > that if ISA isn't enabled, certain things break in 2.6; for example, my
> > pcmcia nic/modem (using pcnet_cs/serial_cs).  These worked fine in 2.4;
> > I tracked the behavior to the fact that if irq_mask is set on a pcmcia
> > socket (instead of pci_irq), and PCMCIA_PROBE isn't set,
> > pcmcia_request_irq refuses to assign an irq.  Most of the pcmcia bridges
> > appear to set an irq_mask, so the attached patch changes Kconfig to set
> > CONFIG_PCMCIA_PROBE if any of those bridges are selected.
> > 
> > Please apply this (or an alternative fix), as it fixes a 2.6 regression
> > in pcmcia functionality.
> 
> Please don't.  David Hinds has a better fix for this, which changes
> the way we handle the allocation of IRQs.  David's change is all
> round a far better way to handle the problem - if all ISA interrupts
> are used or unavailable, we fall back to using the PCI interrupt
> instead.
> 
> Please also note that there /is/ a PCMCIA list which patches should
> be forwarded to - linux-pcmcia which is at lists.infradead.org

Please put the list in the MAINTAINERS file; right now, 
linux-kernel@vger.kernel.org is listed as the relevant list.



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [PATCH] CONFIG_PCMCIA_PROBE fix

[Russell King]

On Mon, Dec 22, 2003 at 12:18:42PM -0500, Andres Salomon wrote:
> On Mon, 2003-12-22 at 04:48, Russell King wrote:
> > Please also note that there /is/ a PCMCIA list which patches should
> > be forwarded to - linux-pcmcia which is at lists.infradead.org
> 
> Please put the list in the MAINTAINERS file; right now, 
> linux-kernel@vger.kernel.org is listed as the relevant list.

Ok, though we do need to find a suitable collection of people willing
to investigate bug reports from users.  Currently I'm receiving most
of the bug reports, though due to hardware limitations I'm unable to
investigate and resolve the vast majority of hardware specific issues.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core