[Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Gilles Chanteperdrix]

Hi Wolfgang,

It seems we have an issue building rtcan for mpc5xxx with Linux 2.6.36.
As far as I understood the of_device structure disappeared and the
"device tree" bits were integrated in the device structure, which causes
several errors when building ksrc/drivers/can/mscan/rtcan_mscan_mpc5xxx.c

Build log details here:
http://sisyphus.hd.free.fr/~gilles/bx/lite5200/2.6.36-ppc_6xx-gcc-4.2.2/log.html#1

Regards.

-- 
                                                                Gilles.

Re: [Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Wolfgang Grandegger]

[-- Attachment #1: Type: text/plain, Size: 541 bytes --]

Hi Gilles,

the attached patch fixes the issue.

On 01/02/2011 08:53 PM, Gilles Chanteperdrix wrote:
> 
> Hi Wolfgang,
> 
> It seems we have an issue building rtcan for mpc5xxx with Linux 2.6.36.
> As far as I understood the of_device structure disappeared and the
> "device tree" bits were integrated in the device structure, which causes
> several errors when building ksrc/drivers/can/mscan/rtcan_mscan_mpc5xxx.c
> 
> Build log details here:
> http://sisyphus.hd.free.fr/~gilles/bx/lite5200/2.6.36-ppc_6xx-gcc-4.2.2/log.html#1

Wolfgang.

[-- Attachment #2: 0001-wrappers-fixup-struct-of_device-for-RTCAN-MPC5xxx-dr.patch --]
[-- Type: text/x-diff, Size: 921 bytes --]

>From 94e2e92946a74bed44b9be208a3cef43e3e52477 Mon Sep 17 00:00:00 2001
From: Wolfgang Grandegger <wg@domain.hid>
Date: Mon, 3 Jan 2011 10:59:09 +0100
Subject: [PATCH] wrappers: fixup "struct of_device" for RTCAN MPC5xxx driver

Since 2.6.36, of_device is just an alias for platform_device.

Signed-off-by: Wolfgang Grandegger <wg@domain.hid>
---
 include/asm-powerpc/wrappers.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/include/asm-powerpc/wrappers.h b/include/asm-powerpc/wrappers.h
index 7f1214a..fb49307 100644
--- a/include/asm-powerpc/wrappers.h
+++ b/include/asm-powerpc/wrappers.h
@@ -191,4 +191,8 @@ typedef irq_handler_t rthal_irq_host_handler_t;
 #define mpc5xxx_get_bus_frequency(node)	mpc52xx_find_ipb_freq(node)
 #endif
 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
+#define of_device platform_device
+#endif
+
 #endif /* _XENO_ASM_POWERPC_WRAPPERS_H */
-- 
1.7.2.3

Re: [Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Gilles Chanteperdrix]

Wolfgang Grandegger wrote:
> Hi Gilles,
> 
> the attached patch fixes the issue.

Applied, thanks.

-- 
                                                                Gilles.

Re: [Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 338 bytes --]

Am 03.01.2011 11:32, Wolfgang Grandegger wrote:
> Hi Gilles,
> 
> the attached patch fixes the issue.
> 

Nothing critical, but I think we should consistently wrap the other way
around

#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)
#define platform_device of_device
#endif

and update the users to the new style.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 259 bytes --]

Re: [Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> Am 03.01.2011 11:32, Wolfgang Grandegger wrote:
>> Hi Gilles,
>>
>> the attached patch fixes the issue.
>>
> 
> Nothing critical, but I think we should consistently wrap the other way
> around
> 
> #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)
> #define platform_device of_device
> #endif
> 
> and update the users to the new style.

Err... Am I missing the obvious here? If you include
<linux/platform_device.h> on powerpc you then get a massive mess prior
to 2.6.36.

-- 
					    Gilles.

Re: [Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 815 bytes --]

Am 05.01.2011 21:25, Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
>> Am 03.01.2011 11:32, Wolfgang Grandegger wrote:
>>> Hi Gilles,
>>>
>>> the attached patch fixes the issue.
>>>
>>
>> Nothing critical, but I think we should consistently wrap the other way
>> around
>>
>> #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)
>> #define platform_device of_device
>> #endif
>>
>> and update the users to the new style.
> 
> Err... Am I missing the obvious here? If you include
> <linux/platform_device.h> on powerpc you then get a massive mess prior
> to 2.6.36.

Maybe this is an exception, don't know the powerpc details, but
generally our drivers should be written according to latest (supported)
kernel APIs and then extended/wrapped/etc. to compile with older
versions as well.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 259 bytes --]

Re: [Xenomai-core] Build issue with rtcan and I-pipe for linux 2.6.36 on powerpc.

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> Am 05.01.2011 21:25, Gilles Chanteperdrix wrote:
>> Jan Kiszka wrote:
>>> Am 03.01.2011 11:32, Wolfgang Grandegger wrote:
>>>> Hi Gilles,
>>>>
>>>> the attached patch fixes the issue.
>>>>
>>> Nothing critical, but I think we should consistently wrap the other way
>>> around
>>>
>>> #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)
>>> #define platform_device of_device
>>> #endif
>>>
>>> and update the users to the new style.
>> Err... Am I missing the obvious here? If you include
>> <linux/platform_device.h> on powerpc you then get a massive mess prior
>> to 2.6.36.
> 
> Maybe this is an exception, don't know the powerpc details, but
> generally our drivers should be written according to latest (supported)
> kernel APIs and then extended/wrapped/etc. to compile with older
> versions as well.

platform_device has existed for a long time in the 2.6 series: the
matching between platform_device and platform_driver is the hack which
replaces enumeration USB and PCI bus provide on embedded systems busses.

Some platforms introduced of_device to handle another kind of
enumeration: the device tree.

Now, it seems the member which were specific to of_device migrated to
the platform_device structure, so, the of_device structure was cancelled
completely, and replaced by platform_device all over the place.

At least, is it my understanding.

Now, we can typedef device_tree_device to be of_device on old kernels,
and platform_device on new kernels, but using of_device for this does
not seem stupid either.

-- 
					    Gilles.