Re: [RTnet-developers] [PATCH] Build fixes for 2.6.36

[Jan K. <jan...@si...>]

On 2011-02-09 13:08, Wolfgang Grandegger wrote:
> On 02/09/2011 12:49 PM, Jan Kiszka wrote:
>> On 2011-02-09 12:08, Wolfgang Grandegger wrote:
>>> Signed-off-by: Wolfgang Grandegger <wg...@de...>
>>> ---
>>>  stack/include/rtnet_port.h |    9 +++++++++
>>>  stack/ipv4/route.c         |    1 +
>>>  stack/rtcfg/rtcfg_proc.c   |    1 +
>>>  stack/rtnet_chrdev.c       |    9 +++++++++
>>>  4 files changed, 20 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/stack/include/rtnet_port.h b/stack/include/rtnet_port.h
>>> index b47c0db..09101cc 100644
>>> --- a/stack/include/rtnet_port.h
>>> +++ b/stack/include/rtnet_port.h
>>> @@ -213,6 +213,15 @@ static inline void *netdev_priv(struct net_device *dev)
>>>  #define DMA_BIT_MASK(n)	(((n) == 64) ? ~0ULL : ((1ULL<<(n))-1))
>>>  #endif
>>>  
>>> +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,35)
>>> +#define NIPQUAD(addr) \
>>> +        ((unsigned char *)&addr)[0],	\
>>> +	((unsigned char *)&addr)[1],	\
>>> +	((unsigned char *)&addr)[2],	\
>>> +	((unsigned char *)&addr)[3]
>>> +#define NIPQUAD_FMT "%u.%u.%u.%u"
>>> +#endif
>>> +
>>>  #endif /* __KERNEL__ */
>>>  
>>>  #endif /* __RTNET_PORT_H_ */
>>> diff --git a/stack/ipv4/route.c b/stack/ipv4/route.c
>>> index 2151686..505b32e 100644
>>> --- a/stack/ipv4/route.c
>>> +++ b/stack/ipv4/route.c
>>> @@ -26,6 +26,7 @@
>>>  #include <net/ip.h>
>>>  
>>>  #include <rtnet_internal.h>
>>> +#include <rtnet_port.h>
>>>  #include <rtnet_chrdev.h>
>>>  #include <ipv4/af_inet.h>
>>>  #include <ipv4/route.h>
>>> diff --git a/stack/rtcfg/rtcfg_proc.c b/stack/rtcfg/rtcfg_proc.c
>>> index 3d55d50..93aafd8 100644
>>> --- a/stack/rtcfg/rtcfg_proc.c
>>> +++ b/stack/rtcfg/rtcfg_proc.c
>>> @@ -24,6 +24,7 @@
>>>  
>>>  #include <rtdev.h>
>>>  #include <rtnet_internal.h>
>>> +#include <rtnet_port.h>
>>>  #include <rtcfg/rtcfg_conn_event.h>
>>>  #include <rtcfg/rtcfg_event.h>
>>>  #include <rtcfg/rtcfg_frame.h>
>>
>> OK for this.
>>
>>> diff --git a/stack/rtnet_chrdev.c b/stack/rtnet_chrdev.c
>>> index b0f2863..0d3fae3 100644
>>> --- a/stack/rtnet_chrdev.c
>>> +++ b/stack/rtnet_chrdev.c
>>> @@ -47,8 +47,13 @@ LIST_HEAD(ioctl_handlers);
>>>   * @request:
>>>   * @arg:
>>>   */
>>> +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,35)
>>> +static long rtnet_ioctl(struct file *file,
>>> +			unsigned int request, unsigned long arg)
>>> +#else
>>>  static int rtnet_ioctl(struct inode *inode, struct file *file,
>>>                         unsigned int request, unsigned long arg)
>>> +#endif
> 
> Any clever idea on how to handle this unconditionally?

Actually... no. :)

> 
>>>  {
>>>      struct rtnet_ioctl_head head;
>>>      struct rtnet_device     *rtdev = NULL;
>>> @@ -286,7 +291,11 @@ void rtnet_unregister_ioctls(struct rtnet_ioctls *ioctls)
>>>  
>>>  
>>>  static struct file_operations rtnet_fops = {
>>> +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,35)
>>> +    .unlocked_ioctl = rtnet_ioctl,
>>> +#else
>>>      .ioctl= rtnet_ioctl,
>>> +#endif
>>>  };
>>>  
>>>  static struct miscdevice rtnet_chr_misc_dev = {
>>
>> But here we should be able to rely on Xenomai redefining unlocked_ioctl
>> to ioctl on 2.4 kernels. Means: convert unconditionally. Can you check this?
> 
> Well, as I just realized, in Xenomai we just use:
> 
> #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,11)
> ##define unlocked_ioctl ioctl
> #endif
> 
> with the modified new argument list. But this will not work for < 2.6.11
> :-(.

Ah. At least we found a bug this way. Then let's keep it conditional in
RTnet (depending on < 2.6.11) - and fix Xenomai.

The alternative is to finally bury 2.4 (and old 2.6) support with the
next release. I bet no one would dare to update RTnet while sticking
with such prehistoric kernels. Anyway, this fixup could then still be
consolidated.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux