Wolfgang Grandegger wrote:
> Jan Kiszka wrote:
>> Wolfgang Grandegger wrote:
>>> Hello,
>>>
>>> yesterday I realized a problem with compat_module_param_array() with
>>> ksrc/drivers/can/sja1000/rtcan_mem.c under Linux 2.4. It uses
>>> CONFIG_XENO_DRIVERS_CAN_SJA1000_MEM_MAX_DEV to define the number of mem
>>> devices, which is set to '(4)' by "make [menuc]onfig". Unfortunately,
>>> the brackets '()' break compat_module_param_array() because it treats
>>> the count as string:
>>>
>>> #define compat_module_param_array(name, type, count, perm) \
>>>          static inline void *__check_existence_##name(void) { return
>>> &name; } \
>>>          MODULE_PARM(name, "1-" __MODULE_STRING(count)
>>> _MODULE_PARM_STRING_##type)
>>>
>>> Any idea how to fix that? I actually tend to remove the configuration
>>> option CONFIG_XENO_DRIVERS_CAN_SJA1000_MEM_MAX_DEV and hardcode it to
>>> "4" (without brackets) in the driver. It's kind of overkill, anyhow.
>>
>> I would say that this problem is not "with compat_module_param_array",
>> but about the Kconfig-based constant. I don't see a way to make the
>> module parameter array size configurable even under vanilla 2.4. Simply
>> go for a patchable constant in the source code like other drivers do.
> 
> The attached patch fixes the problem:
> 
> 2007-08-15  Wolfgang Grandegger  <wg@domain.hid>
> 
>         * ksrc/drivers/can/sja1000/Kconfig,
>         ksrc/drivers/can/sja1000/{rtcan_isa.c,rtcan_mem.c}: Remove config
>         options XENO_DRIVERS_CAN_SJA1000_*_MAX_DEV to avoid problems with
>         defining module parameters with Linux 2.4.
> 
> If nobody complains, I'm going to apply it to the trunk and the v2.3.x
> branch.

Looks good, go ahead.

Jan