[PATCH] There's a pciconf.h with PCI conf constants - lets actually use it

[Mart Raudsepp]

In case it's a good idea to use these constants from that header, here's the proposed commit log and patch:

There's a pciconf.h with PCI conf constants - lets actually use it

PCI_CONF_REG_{INDEX,DATA} constants instead of 0xCF8 and 0xCFF everywhere.
Not touching YABEL as I'm not sure it would want to include that header.

Signed-off-by: Mart Raudsepp 
---
 arch/x86/pci_ops_conf1.c        |   24 ++++++++++++------------
 southbridge/amd/cs5536/cs5536.c |    5 +++--
 2 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/arch/x86/pci_ops_conf1.c b/arch/x86/pci_ops_conf1.c
index a9a7fea..757b6c2 100644
--- a/arch/x86/pci_ops_conf1.c
+++ b/arch/x86/pci_ops_conf1.c
@@ -58,38 +58,38 @@

u8 pci_conf1_read_config8(u32 bdf, int where)
 {
-		outl(CONFIG_CMD(bdf, where), 0xCF8);
-		return inb(0xCFC + (where & 3));
+		outl(CONFIG_CMD(bdf, where), PCI_CONF_REG_INDEX);
+		return inb(PCI_CONF_REG_DATA + (where & 3));
 }

u16 pci_conf1_read_config16(u32 bdf, int where)
 {
-		outl(CONFIG_CMD(bdf, where), 0xCF8);
-		return inw(0xCFC + (where & 2));
+		outl(CONFIG_CMD(bdf, where), PCI_CONF_REG_INDEX);
+		return inw(PCI_CONF_REG_DATA + (where & 2));
 }

u32 pci_conf1_read_config32(u32 bdf, int where)
 {
-		outl(CONFIG_CMD(bdf, where), 0xCF8);
-		return inl(0xCFC);
+		outl(CONFIG_CMD(bdf, where), PCI_CONF_REG_INDEX);
+		return inl(PCI_CONF_REG_DATA);
 }

void  pci_conf1_write_config8(u32 bdf, int where, u8 value)
 {
-		outl(CONFIG_CMD(bdf, where), 0xCF8);
-		outb(value, 0xCFC + (where & 3));
+		outl(CONFIG_CMD(bdf, where), PCI_CONF_REG_INDEX);
+		outb(value, PCI_CONF_REG_DATA + (where & 3));
 }

void pci_conf1_write_config16(u32 bdf, int where, u16 value)
 {
-		outl(CONFIG_CMD(bdf, where), 0xCF8);
-		outw(value, 0xCFC + (where & 2));
+		outl(CONFIG_CMD(bdf, where), PCI_CONF_REG_INDEX);
+		outw(value, PCI_CONF_REG_DATA + (where & 2));
 }

void pci_conf1_write_config32(u32 bdf, int where, u32 value)
 {
-		outl(CONFIG_CMD(bdf, where), 0xCF8);
-		outl(value, 0xCFC);
+		outl(CONFIG_CMD(bdf, where), PCI_CONF_REG_INDEX);
+		outl(value, PCI_CONF_REG_DATA);
 }

/**
diff --git a/southbridge/amd/cs5536/cs5536.c b/southbridge/amd/cs5536/cs5536.c
index 89d64d6..dacf5c3 100644
--- a/southbridge/amd/cs5536/cs5536.c
+++ b/southbridge/amd/cs5536/cs5536.c
@@ -20,6 +20,7 @@
 #include <types.h>
 #include <lib.h>
 #include <console.h>
+#include <pciconf.h>
 #include <device/pci.h>
 #include <msr.h>
 #include <amd_geodelx.h>
@@ -93,8 +94,8 @@ static void hide_vpci(u32 vpci_devid)
    printk(BIOS_DEBUG, "Hiding VPCI device: 0x%08X (%02x:%02x.%01x)\n",
    	vpci_devid, (vpci_devid >> 16) & 0xff,
    	(vpci_devid >> 11) & 0x1f, (vpci_devid >> 8) & 0x7);
-	outl(vpci_devid + 0x7C, 0xCF8);
-	outl(0xDEADBEEF, 0xCFC);
+	outl(vpci_devid + 0x7C, PCI_CONF_REG_INDEX);
+	outl(0xDEADBEEF, PCI_CONF_REG_DATA);
 }

/**

[PATCH] There's a pciconf.h with PCI conf constants - lets actually use it

[Uwe Hermann]

On Mon, Jan 05, 2009 at 04:49:57PM +0200, Mart Raudsepp wrote:
> In case it's a good idea to use these constants from that header, here's the proposed commit log and patch:
>
> There's a pciconf.h with PCI conf constants - lets actually use it
>
> PCI_CONF_REG_{INDEX,DATA} constants instead of 0xCF8 and 0xCFF everywhere.
> Not touching YABEL as I'm not sure it would want to include that header.
>
> Signed-off-by: Mart Raudsepp

Acked-by: Uwe Hermann

Uwe.

[PATCH] There's a pciconf.h with PCI conf constants - lets actually use it

[Peter Stuge]

Uwe Hermann wrote:
> > There's a pciconf.h with PCI conf constants - lets actually use it
> >
> > PCI_CONF_REG_{INDEX,DATA} constants instead of 0xCF8 and 0xCFF everywhere.
> > Not touching YABEL as I'm not sure it would want to include that header.
> >
> > Signed-off-by: Mart Raudsepp
>
> Acked-by: Uwe Hermann

Please don't commit this. I would rather see the 0xcf8 and cfc
defines removed. This is PCI code, these ports will never change
for PCI buses, and they are the first thing most learn about PCI
programming.

Defines add indirection = complexity = error sources. I'm with
Stefan, we should use them carefully.

//Peter

[Mart Raudsepp]

Ühel kenal päeval, T, 2009-01-06 kell 00:38, kirjutas Peter Stuge:
> Uwe Hermann wrote:
> > > There's a pciconf.h with PCI conf constants - lets actually use it
> > >
> > > PCI_CONF_REG_{INDEX,DATA} constants instead of 0xCF8 and 0xCFF everywhere.
> > > Not touching YABEL as I'm not sure it would want to include that header.
> > >
> > > Signed-off-by: Mart Raudsepp
> >
> > Acked-by: Uwe Hermann
>
> Please don't commit this. I would rather see the 0xcf8 and cfc
> defines removed. This is PCI code, these ports will never change
> for PCI buses, and they are the first thing most learn about PCI
> programming.
>
> Defines add indirection = complexity = error sources. I'm with
> Stefan, we should use them carefully.

Hence the sentence "In case it's a good idea to use these constants from
that header" in my mail :)

I simply noticed the fact that there are constant defined, and that the
same thing is used per value. Lets perhaps fix the inconsistency another
way then. Proposed patch attached.

Mart Raudsepp

[PATCH] There's a pciconf.h with PCI conf constants - lets actually use it

[ron minnich]

On Tue, Jan 6, 2009 at 1:35 AM, Mart Raudsepp
 wrote:
> I simply noticed the fact that there are constant defined, and that the
> same thing is used per value. Lets perhaps fix the inconsistency another
> way then. Proposed patch attached.

That patch I like ...

ron

[PATCH] There's a pciconf.h with PCI conf constants - lets actually use it

[Peter Stuge]

Mart Raudsepp wrote:
> Subject: [PATCH] Remove unused pciconf.h header with constants that
> everyone uses by value instead per convention
>
> Signed-off-by: Mart Raudsepp

Acked-by: Peter Stuge