[PATCH] Fix build break caused by "ide: remove ideprobe_init()"

[PATCH] Fix build break caused by "ide: remove ideprobe_init()"

[Olof Johansson]

Fix build break of powerpc holly_defconfig:

In file included from arch/powerpc/platforms/embedded6xx/holly.c:24:
include/linux/ide.h:1206: error: 'CONFIG_IDE_MAX_HWIFS' undeclared here (not in a function)

There's no need to have a sized array in the prototype, might as well
turn it into a pointer.

It could probably be argued that large parts of the include file can be
covered under #ifdef CONFIG_IDE, but that's a larger undertaking.


Signed-off-by: Olof Johansson <olof@lixom.net>

---

On Sun, Nov 18, 2007 at 11:25:09PM +0100, Bartlomiej Zolnierkiewicz wrote:
> 
> * Rename ide_device_add() to ide_device_add_all() and make it accept
>   'u8 idx[MAX_HWIFS]' instead of 'u8 idx[4]' as an argument.
> 
> * Add ide_device_add() wrapper for ide_device_add_all().
> 
> * Convert ide_generic_init() to use ide_device_add_all().
> 
> * Remove no longer needed ideprobe_init().
> 
> There should be no functionality changes caused by this patch.

This patch broke builds of powerpc holly_defconfig in -mm. It has
CONFIG_EMBEDDED=y, CONFIG_IDE=n but includes linux/ide.h:


Index: mm/drivers/ide/ide-probe.c
===================================================================
--- mm.orig/drivers/ide/ide-probe.c
+++ mm/drivers/ide/ide-probe.c
@@ -1335,7 +1335,7 @@ static void hwif_register_devices(ide_hw
 	}
 }
 
-int ide_device_add_all(u8 idx[MAX_HWIFS])
+int ide_device_add_all(u8 *idx)
 {
 	ide_hwif_t *hwif;
 	int i, rc = 0;
Index: mm/include/linux/ide.h
===================================================================
--- mm.orig/include/linux/ide.h
+++ mm/include/linux/ide.h
@@ -1203,7 +1203,7 @@ void ide_unregister_region(struct gendis
 
 void ide_undecoded_slave(ide_drive_t *);
 
-int ide_device_add_all(u8 idx[MAX_HWIFS]);
+int ide_device_add_all(u8 *idx);
 int ide_device_add(u8 idx[4]);
 
 static inline void *ide_get_hwifdata (ide_hwif_t * hwif)

Re: [PATCH] Fix build break caused by "ide: remove ideprobe_init()"

[Bartlomiej Zolnierkiewicz]

Hi,

On Friday 14 December 2007, Olof Johansson wrote:
> Fix build break of powerpc holly_defconfig:

Sorry for breaking the hol[l]y powerpc platform. ;)

> In file included from arch/powerpc/platforms/embedded6xx/holly.c:24:
> include/linux/ide.h:1206: error: 'CONFIG_IDE_MAX_HWIFS' undeclared here (not in a function)
> 
> There's no need to have a sized array in the prototype, might as well
> turn it into a pointer.
> 
> It could probably be argued that large parts of the include file can be
> covered under #ifdef CONFIG_IDE, but that's a larger undertaking.
> 
> 
> Signed-off-by: Olof Johansson <olof@lixom.net>

applied but...

> ---
> 
> On Sun, Nov 18, 2007 at 11:25:09PM +0100, Bartlomiej Zolnierkiewicz wrote:
> > 
> > * Rename ide_device_add() to ide_device_add_all() and make it accept
> >   'u8 idx[MAX_HWIFS]' instead of 'u8 idx[4]' as an argument.
> > 
> > * Add ide_device_add() wrapper for ide_device_add_all().
> > 
> > * Convert ide_generic_init() to use ide_device_add_all().
> > 
> > * Remove no longer needed ideprobe_init().
> > 
> > There should be no functionality changes caused by this patch.
> 
> This patch broke builds of powerpc holly_defconfig in -mm. It has
> CONFIG_EMBEDDED=y, CONFIG_IDE=n but includes linux/ide.h:
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

...could you also fix the root source of the problem so the similar issue
won't re-appear in the future?

<linux/ide.h> is IDE subsystem specific in the same way that <linux/libata.h>
is libata specific, <linux/ata.h> is the place to put subsystem independent
defines, inlines and co.

There are some powerpc abusers of the above rule but they should at least cover
references to <linux/ide.h> and IDE subsystem specific code with #ifdef/#endif
CONFIG_BLK_DEV_IDE.

In arch/powerpc/platforms/embedded6xx/holly.c case it seems that <linux/ide.h>
include is actually not needed and may be just removed?

Bart

> Index: mm/drivers/ide/ide-probe.c
> ===================================================================
> --- mm.orig/drivers/ide/ide-probe.c
> +++ mm/drivers/ide/ide-probe.c
> @@ -1335,7 +1335,7 @@ static void hwif_register_devices(ide_hw
>  	}
>  }
>  
> -int ide_device_add_all(u8 idx[MAX_HWIFS])
> +int ide_device_add_all(u8 *idx)
>  {
>  	ide_hwif_t *hwif;
>  	int i, rc = 0;
> Index: mm/include/linux/ide.h
> ===================================================================
> --- mm.orig/include/linux/ide.h
> +++ mm/include/linux/ide.h
> @@ -1203,7 +1203,7 @@ void ide_unregister_region(struct gendis
>  
>  void ide_undecoded_slave(ide_drive_t *);
>  
> -int ide_device_add_all(u8 idx[MAX_HWIFS]);
> +int ide_device_add_all(u8 *idx);
>  int ide_device_add(u8 idx[4]);
>  
>  static inline void *ide_get_hwifdata (ide_hwif_t * hwif)

Re: [PATCH] Fix build break caused by "ide: remove ideprobe_init()"

[Olof Johansson]

On Mon, Dec 17, 2007 at 09:53:31PM +0100, Bartlomiej Zolnierkiewicz wrote:
> 
> Hi,
> 
> applied but...
> 
> > ---
> > 
> > On Sun, Nov 18, 2007 at 11:25:09PM +0100, Bartlomiej Zolnierkiewicz wrote:
> > > 
> > > * Rename ide_device_add() to ide_device_add_all() and make it accept
> > >   'u8 idx[MAX_HWIFS]' instead of 'u8 idx[4]' as an argument.
> > > 
> > > * Add ide_device_add() wrapper for ide_device_add_all().
> > > 
> > > * Convert ide_generic_init() to use ide_device_add_all().
> > > 
> > > * Remove no longer needed ideprobe_init().
> > > 
> > > There should be no functionality changes caused by this patch.
> > 
> > This patch broke builds of powerpc holly_defconfig in -mm. It has
> > CONFIG_EMBEDDED=y, CONFIG_IDE=n but includes linux/ide.h:
>                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> 
> ...could you also fix the root source of the problem so the similar issue
> won't re-appear in the future?
> 
> <linux/ide.h> is IDE subsystem specific in the same way that <linux/libata.h>
> is libata specific, <linux/ata.h> is the place to put subsystem independent
> defines, inlines and co.
> 
> There are some powerpc abusers of the above rule but they should at least cover
> references to <linux/ide.h> and IDE subsystem specific code with #ifdef/#endif
> CONFIG_BLK_DEV_IDE.
> 
> In arch/powerpc/platforms/embedded6xx/holly.c case it seems that <linux/ide.h>
> include is actually not needed and may be just removed?

Yep, that was my original simple fix to the problem, but when I looked
closer I wanted to get the include file fixed as well.

Turns out I never posted the powerpc patch :), I'll do so (and audit
other platforms for the same).


Thanks,

-Olof