Re: IPMI driver version 19 release

[Jeff Garzik <jgarzik@pobox.com>]

On Thu, Apr 03, 2003 at 10:24:02AM -0600, Corey Minyard wrote:
> @@ -563,8 +576,9 @@
>  		spin_lock_irqsave(&(kcs_info->kcs_lock), flags);
>  		result = kcs_event_handler(kcs_info, 0);
>  		while (result != KCS_SM_IDLE) {
> -			udelay(500);
> -			result = kcs_event_handler(kcs_info, 500);
> +			udelay(KCS_SHORT_TIMEOUT_USEC);
> +			result = kcs_event_handler(kcs_info,
> +						   KCS_SHORT_TIMEOUT_USEC);
>  		}
>  		spin_unlock_irqrestore(&(kcs_info->kcs_lock), flags);
>  		return;


Do you really want to udelay this long with interrupts disabled?
Certainly comments in kcs_event[_handler] indicate you're aware of the
issue, but the code does not belie this fact :)

Not only is the udelay itself "long" relatively speaking, but it's in a
loop.  Which also calls a function that contains a loop that is
potentially infinite is hardware is being wonky.

	Jeff