Hi,
This patch allows device fixups to force the PCI subsystem to ignore
bridges and hence not allocate resources to them.
I have an IXP425 (ARM) board with a CardBus controller on it (of which
only the PC card interfaces are used). The problem is that the PCI
memory window is too small to fit in all the bridge resources and the
rest of the PCI devices and up being unconfigured. With this patch, and
a fixup to clear is_bridge, this doesn't happen.
The plan was to make the CardBus driver (drivers/pcmcia/yenta_socket.c)
honour the is_bridge flag and not bother with CardBus stuff if it's cleared.
2004-10-19 David Vrabel <[email protected]>
* Add is_bridge to struct pci_dev to allow PCI device fixups to
disable bridge functionality.
Index: linux-2.6-armbe/drivers/pci/probe.c
===================================================================
--- linux-2.6-armbe.orig/drivers/pci/probe.c 2004-10-14
11:26:38.000000000 +0100
+++ linux-2.6-armbe/drivers/pci/probe.c 2004-10-19 12:00:00.000000000 +0100
@@ -610,6 +610,8 @@
dev->devfn = devfn;
dev->hdr_type = hdr_type & 0x7f;
dev->multifunction = !!(hdr_type & 0x80);
+ dev->is_bridge = (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE
+ || dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
dev->vendor = l & 0xffff;
dev->device = (l >> 16) & 0xffff;
dev->cfg_size = pci_cfg_space_size(dev);
@@ -718,8 +720,7 @@
pcibios_fixup_bus(bus);
for (pass=0; pass < 2; pass++)
list_for_each_entry(dev, &bus->devices, bus_list) {
- if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE ||
- dev->hdr_type == PCI_HEADER_TYPE_CARDBUS)
+ if (dev->is_bridge)
max = pci_scan_bridge(bus, dev, max, pass);
}
Index: linux-2.6-armbe/include/linux/pci.h
===================================================================
--- linux-2.6-armbe.orig/include/linux/pci.h 2004-10-14
11:26:38.000000000 +0100
+++ linux-2.6-armbe/include/linux/pci.h 2004-10-14 11:43:24.000000000 +0100
@@ -532,6 +532,7 @@
/* These fields are used by common fixups */
unsigned int transparent:1; /* Transparent PCI bridge */
unsigned int multifunction:1;/* Part of multi-function device */
+ unsigned int is_bridge:1; /* A PCI or CardBus bridge */
/* keep track of device state */
unsigned int is_enabled:1; /* pci_enable_device has been called */
unsigned int is_busmaster:1; /* device is busmaster */
On Tue, Oct 19, 2004 at 12:22:49PM +0100, David Vrabel wrote:
> Hi,
>
> This patch allows device fixups to force the PCI subsystem to ignore
> bridges and hence not allocate resources to them.
>
> I have an IXP425 (ARM) board with a CardBus controller on it (of which
> only the PC card interfaces are used). The problem is that the PCI
> memory window is too small to fit in all the bridge resources and the
> rest of the PCI devices and up being unconfigured. With this patch, and
> a fixup to clear is_bridge, this doesn't happen.
>
> The plan was to make the CardBus driver (drivers/pcmcia/yenta_socket.c)
> honour the is_bridge flag and not bother with CardBus stuff if it's cleared.
But why can't any code that wants to check this, just look at the
dev->hdr_type instead? I don't think we need to add a new bit for this
because of that, right?
> Index: linux-2.6-armbe/drivers/pci/probe.c
> ===================================================================
> --- linux-2.6-armbe.orig/drivers/pci/probe.c 2004-10-14
> 11:26:38.000000000 +0100
> +++ linux-2.6-armbe/drivers/pci/probe.c 2004-10-19
> 12:00:00.000000000 +0100
Also, your patch was linewrapped :(
thanks,
greg k-h
Index: linux-2.6-armbe/drivers/pci/probe.c
===================================================================
--- linux-2.6-armbe.orig/drivers/pci/probe.c 2004-10-14 11:26:38.000000000 +0100
+++ linux-2.6-armbe/drivers/pci/probe.c 2004-10-19 12:00:00.000000000 +0100
@@ -610,6 +610,8 @@
dev->devfn = devfn;
dev->hdr_type = hdr_type & 0x7f;
dev->multifunction = !!(hdr_type & 0x80);
+ dev->is_bridge = (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE
+ || dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
dev->vendor = l & 0xffff;
dev->device = (l >> 16) & 0xffff;
dev->cfg_size = pci_cfg_space_size(dev);
@@ -718,8 +720,7 @@
pcibios_fixup_bus(bus);
for (pass=0; pass < 2; pass++)
list_for_each_entry(dev, &bus->devices, bus_list) {
- if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE ||
- dev->hdr_type == PCI_HEADER_TYPE_CARDBUS)
+ if (dev->is_bridge)
max = pci_scan_bridge(bus, dev, max, pass);
}
Index: linux-2.6-armbe/include/linux/pci.h
===================================================================
--- linux-2.6-armbe.orig/include/linux/pci.h 2004-10-14 11:26:38.000000000 +0100
+++ linux-2.6-armbe/include/linux/pci.h 2004-10-14 11:43:24.000000000 +0100
@@ -532,6 +532,7 @@
/* These fields are used by common fixups */
unsigned int transparent:1; /* Transparent PCI bridge */
unsigned int multifunction:1;/* Part of multi-function device */
+ unsigned int is_bridge:1; /* A PCI or CardBus bridge */
/* keep track of device state */
unsigned int is_enabled:1; /* pci_enable_device has been called */
unsigned int is_busmaster:1; /* device is busmaster */
On Mon, Nov 01, 2004 at 10:59:13AM +0000, David Vrabel wrote:
> Greg KH wrote:
> > On Tue, Oct 19, 2004 at 12:22:49PM +0100, David Vrabel wrote:
> >
> >>The plan was to make the CardBus driver (drivers/pcmcia/yenta_socket.c)
> >>honour the is_bridge flag and not bother with CardBus stuff if it's cleared.
> >
> > But why can't any code that wants to check this, just look at the
> > dev->hdr_type instead? I don't think we need to add a new bit for this
> > because of that, right?
>
> Using the is_bridge flag allows PCI device fixups to disable the CardBus
> portion of the driver. And you obviously can't tweak the hdr_type since
> the device header is still a CardBus/bridge type header.
Why not make the PCI spaces for cardbus bridges configurable? Eg,
cardbusmem=1M cardbusio=8K
?
--
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