Re: [PATCH] amd76x_pm port to 2.5.70

[Zwane Mwaikambo]

On Sat, 31 May 2003, Pasi Savolainen wrote:

> Attached is a simplistic port of amd76x_pm to 2.5.70.
> There is a mention about being non-preempt safe, and I can't tell
> spinlock from semaphore, so can do nada about it.
> zwane told me on #kernelnewbies that 
> 
>   pm_idle = old_pm_idle; wmb();
> 
> should take care about the processor syncing, but I do get total hangs
> on rmmod.

pm_idle = NULL_or_previous_handler;
synchronize_kernel();
unload_module;

> +amd76x_up_idle(void)
> +{
> +	// FIXME: Optionally add non-smp idle loop here
> +}
> +#endif

This is going to really suck power wise, at least do a rep_nop or some 
such otherwise you spin in a tight polling loop around need_resched.

> +static void
> +amd76x_smp_idle(void)
> +{
> +	/*
> +	 * Exit idle mode immediately if the CPU does not change.
> +	 * Usually that means that we have some load on another CPU.
> +	 */
> +
> +	if (prs[0].idle && prs[1].idle && amd76x_pm_cfg.last_pr == smp_processor_id()) {
> +		prs[0].idle = 0;
> +		prs[1].idle = 0;
> +		/* This looks redundent as it was just checked in the if() */
> +		/* amd76x_pm_cfg.last_pr = smp_processor_id(); */

One is a comparison the other an assignment, what's going on here?

> +static void __exit
> +amd76x_pm_cleanup(void)
> +{
> +#ifndef AMD76X_NTH
> +	pm_idle = amd76x_pm_cfg.orig_idle;
> +	wmb();
> +	//__asm__ __volatile__ ("wbinvd;"); // propagate through SMP

synchronize_kernel() here, then you should be safe to unload.

-- 
function.linuxpower.ca