[PATCH] PCI: fix build errors with CONFIG_PCI=n

[PATCH] PCI: fix build errors with CONFIG_PCI=n

[Randy.Dunlap]

[-- Attachment #1: Type: text/plain, Size: 992 bytes --]


Fix (most of) kernel build for CONFIG_PCI=n.  Fixes these 3 errors:

1. drivers/parport/parport_pc.c:3162: error: `parport_init_mode'
undeclared (first use in this function)

2. arch/x86_64/kernel/built-in.o(.text+0x8186): In function
`quirk_intel_irqbalance':
: undefined reference to `raw_pci_ops'

Kconfig change:
3. arch/x86_64/kernel/pci-gart.c:194: error: `pci_bus_type' undeclared
(first use in this function)

Still does not fix these 2 (similar):
a.  drivers/built-in.o(.text+0x3dcd8): In function
`pnpacpi_allocated_resource':
: undefined reference to `pcibios_penalize_isa_irq'
b.  drivers/built-in.o(.text+0xb4d4): In function
`pnpbios_parse_allocated_irqresource':
: undefined reference to `pcibios_penalize_isa_irq'

diffstat:=
   arch/i386/kernel/quirks.c    |    3 ++-
   arch/x86_64/Kconfig          |    1 +
   drivers/parport/parport_pc.c |   22 ++++++++++++++++++----
   3 files changed, 21 insertions(+), 5 deletions(-)

Signed-off-by: Randy Dunlap <rddunlap@osdl.org>


[-- Attachment #2: config_pci_off.patch --]
[-- Type: text/x-patch, Size: 2846 bytes --]

diff -Naurp ./drivers/parport/parport_pc.c~config_pci ./drivers/parport/parport_pc.c
--- ./drivers/parport/parport_pc.c~config_pci	2004-11-15 10:02:00.023963736 -0800
+++ ./drivers/parport/parport_pc.c	2004-11-16 11:15:24.546578032 -0800
@@ -3154,8 +3154,9 @@ static int __init parport_parse_dma(cons
 				     PARPORT_DMA_NONE, PARPORT_DMA_NOFIFO);
 }
 
-static int __init parport_init_mode_setup(const char *str) {
-
+#ifdef CONFIG_PCI
+static int __init parport_init_mode_setup(const char *str)
+{
 	printk(KERN_DEBUG "parport_pc.c: Specified parameter parport_init_mode=%s\n", str);
 
 	if (!strcmp (str, "spp"))
@@ -3170,11 +3171,16 @@ static int __init parport_init_mode_setu
 		parport_init_mode=5;
 	return 1;
 }
+#else
+static int __inline__ __init parport_init_mode_setup(const char *str)
+{
+	return 1;
+}
+#endif
 
 #ifdef MODULE
 static const char *irq[PARPORT_PC_MAX_PORTS];
 static const char *dma[PARPORT_PC_MAX_PORTS];
-static const char *init_mode;
 
 MODULE_PARM_DESC(io, "Base I/O address (SPP regs)");
 module_param_array(io, int, NULL, 0);
@@ -3189,8 +3195,14 @@ module_param_array(dma, charp, NULL, 0);
 MODULE_PARM_DESC(verbose_probing, "Log chit-chat during initialisation");
 module_param(verbose_probing, int, 0644);
 #endif
+
+#ifdef CONFIG_PCI
+static const char *init_mode;
 MODULE_PARM_DESC(init_mode, "Initialise mode for VIA VT8231 port (spp, ps2, epp, ecp or ecpepp)");
 MODULE_PARM(init_mode, "s");
+#else
+#define init_mode	0
+#endif
 
 static int __init parse_parport_params(void)
 {
@@ -3314,7 +3326,9 @@ __setup ("parport=", parport_setup);
  * parport_init_mode=[spp|ps2|epp|ecp|ecpepp]
  */
 
-__setup("parport_init_mode=",parport_init_mode_setup);
+#ifdef CONFIG_PCI
+__setup("parport_init_mode=", parport_init_mode_setup);
+#endif
 
 #endif
 
diff -Naurp ./arch/i386/kernel/quirks.c~config_pci ./arch/i386/kernel/quirks.c
--- ./arch/i386/kernel/quirks.c~config_pci	2004-11-15 10:01:58.430206024 -0800
+++ ./arch/i386/kernel/quirks.c	2004-11-16 11:24:25.204385552 -0800
@@ -1,10 +1,11 @@
 /*
  * This file contains work-arounds for x86 and x86_64 platform bugs.
  */
+#include <linux/config.h>
 #include <linux/pci.h>
 #include <linux/irq.h>
 
-#if defined(CONFIG_X86_IO_APIC) && defined(CONFIG_SMP)
+#if defined(CONFIG_X86_IO_APIC) && defined(CONFIG_SMP) && defined(CONFIG_PCI)
 
 void __devinit quirk_intel_irqbalance(struct pci_dev *dev)
 {
diff -Naurp ./arch/x86_64/Kconfig~config_pci ./arch/x86_64/Kconfig
--- ./arch/x86_64/Kconfig~config_pci	2004-11-15 10:01:58.985121664 -0800
+++ ./arch/x86_64/Kconfig	2004-11-16 10:50:00.987194264 -0800
@@ -306,6 +306,7 @@ config NR_CPUS
 
 config GART_IOMMU
 	bool "IOMMU support"
+	depends on PCI
 	help
 	  Support the K8 IOMMU. Needed to run systems with more than 4GB of memory
 	  properly with 32-bit PCI devices that do not support DAC (Double Address

Re: [PATCH] PCI: fix build errors with CONFIG_PCI=n

[Andries Brouwer]

On Tue, Nov 16, 2004 at 02:34:48PM -0800, Randy.Dunlap wrote:

> Fix (most of) kernel build for CONFIG_PCI=n.  Fixes these 3 errors:
> 
> 1. drivers/parport/parport_pc.c:3162: error: `parport_init_mode'
> undeclared (first use in this function)

Life is easier if you do not use attachments.
(Then I can more easily comment the code.)

You write

  -static int __init parport_init_mode_setup(const char *str) {
  -
  +#ifdef CONFIG_PCI
  +static int __init parport_init_mode_setup(const char *str)

In my tree I have

  static int __init parport_init_mode_setup(char *str) {

in order to avoid the warning for

  __setup("parport_init_mode=",parport_init_mode_setup);

since the parameter is a int (*setup_func)(char *); - see

  struct obs_kernel_param {
        const char *str;
        int (*setup_func)(char *);
        int early;
  };

Apart from this prototype change I only moved the single line

  static int __initdata parport_init_mode = 0;

outside the #ifdef's. Is that not good enough, and better
than introducing more #ifdef's? Keeps the source smaller.

Andries

Re: [PATCH] PCI: fix build errors with CONFIG_PCI=n

[Randy.Dunlap]

Andries Brouwer wrote:
> On Tue, Nov 16, 2004 at 02:34:48PM -0800, Randy.Dunlap wrote:
> 
>>Fix (most of) kernel build for CONFIG_PCI=n.  Fixes these 3 errors:
>>
>>1. drivers/parport/parport_pc.c:3162: error: `parport_init_mode'
>>undeclared (first use in this function)
> 
> 
> Life is easier if you do not use attachments.
> (Then I can more easily comment the code.)

I understand.  If the decision were only so simple.

> You write
> 
>   -static int __init parport_init_mode_setup(const char *str) {
>   -
>   +#ifdef CONFIG_PCI
>   +static int __init parport_init_mode_setup(const char *str)
> 
> In my tree I have
> 
>   static int __init parport_init_mode_setup(char *str) {
> 
> in order to avoid the warning for
> 
>   __setup("parport_init_mode=",parport_init_mode_setup);
> 
> since the parameter is a int (*setup_func)(char *); - see
> 
>   struct obs_kernel_param {
>         const char *str;
>         int (*setup_func)(char *);
>         int early;
>   };

Yes, I'm familiar with that, but I made a patch against current
top of tree.

> Apart from this prototype change I only moved the single line
> 
>   static int __initdata parport_init_mode = 0;
> 
> outside the #ifdef's. Is that not good enough, and better
> than introducing more #ifdef's? Keeps the source smaller.

It can be good enough.  It keeps the source smaller, at
the expense of adding some unneeded code (the
parport_init_mode_setup() function e.g.).

-- 
~Randy

Re: [PATCH] PCI: fix build errors with CONFIG_PCI=n

[Andries Brouwer]

On Tue, Nov 16, 2004 at 03:36:30PM -0800, Randy.Dunlap wrote:

> >  static int __init parport_init_mode_setup(char *str) {
> 
> Yes, I'm familiar with that, but I made a patch against current
> top of tree.

I don't understand. Will you send another patch to fix the prototype?

Andries

Re: [PATCH] PCI: fix build errors with CONFIG_PCI=n

[Randy.Dunlap]

Andries Brouwer wrote:
> On Tue, Nov 16, 2004 at 03:36:30PM -0800, Randy.Dunlap wrote:
> 
> 
>>> static int __init parport_init_mode_setup(char *str) {
>>
>>Yes, I'm familiar with that, but I made a patch against current
>>top of tree.
> 
> 
> I don't understand. Will you send another patch to fix the prototype?

Sorry, I did that yesterday:
http://lkml.org/lkml/2004/11/15/133

It didn't show up in today's patch because Andrew's "the
perfect patch" says:
(http://www.zip.com.au/~akpm/linux/patches/stuff/tpp.txt)

b) Make sure that your patches apply to the latest version of the
    kernel tree.  Either straight from bitkeeper or from
    ftp://ftp.kernel.org/pub/linux/kernel/v2.6/snapshots/

-- 
~Randy

Re: [PATCH] PCI: fix build errors with CONFIG_PCI=n

[Andries Brouwer]

On Tue, Nov 16, 2004 at 04:13:00PM -0800, Randy.Dunlap wrote:

> >I don't understand. Will you send another patch to fix the prototype?
> 
> Sorry, I did that yesterday:

Ah, OK.