Re: [PATCH] Make lparcfg.c work when both iseries and pseries are selected

[Nathan Lynch <ntl@pobox.com>]

Stephen Rothwell wrote:
>  static ssize_t lparcfg_write(struct file *file, const char __user * buf,
>  			     size_t count, loff_t * off)
>  {
> +	ssize_t retval = -ENOMEM;
> +#ifdef CONFIG_PPC_PSERIES
>  	char *kbuf;
>  	char *tmp;
>  	u64 new_entitled, *new_entitled_ptr = &new_entitled;
> @@ -479,8 +476,6 @@ static ssize_t lparcfg_write(struct file
>  	unsigned long resource;
>  	u8 current_weight;
>  
> -	ssize_t retval = -ENOMEM;
> -
>  	kbuf = kmalloc(count, GFP_KERNEL);
>  	if (!kbuf)
>  		goto out;
> @@ -546,11 +541,10 @@ static ssize_t lparcfg_write(struct file
>  
>  out:
>  	kfree(kbuf);
> +#endif				/* CONFIG_PPC_PSERIES */
>  	return retval;
>  }

Erm... this is kind of gross, and will return -ENOMEM on iSeries when
it should really return -ENOSYS (I think).

Would it be over-engineering to have an lparcfg_ops struct, with
lparcfg_read and lparcfg_write methods (and the latter would be null
on iSeries)?  All the additional #ifdeffery doesn't really improve
the readability of this code IMO.