Re: [RFC, 2.6] a simple FIFO implementation

[Dmitry Torokhov <dtor_core@ameritech.net>]

On Monday 13 September 2004 08:52 am, Stelian Pop wrote:
> +static inline unsigned int kfifo_len(struct kfifo *fifo) {
> +       unsigned long flags;
> +       unsigned int result;
> +       
> +       spin_lock_irqsave(&fifo->lock, flags);
> +       
> +       result = fifo->len;
> +
> +       spin_unlock_irqrestore(&fifo->lock, flags);
> +
> +       return result;
> +}

Hi,

I do not think that taking/releasing spinlock here serves any purpose as
len can get canged right after releasing the lock and therefore no longer
valid... And relying on result of kfifo_len to decide if the FIF can be
drained is not reliable so probably the inteface is better off without this
function.

Also I think that most users would put only sertain structures (homogenous?)
in their FIFOs so maybe it should be:

struct kfifo *kfifo_alloc(unsigned int el_size, unsigned int len)
unsigned int kfifo_put(struct kfifo *fifo, void *buffer)
unsigned int kfifo_get(struct kfifo *fifo, void *buffer)

Also, don't we have a rule that for functions the opening curly brace should
be on a new line?

-- 
Dmitry