Re: [tboot-devel] [PATCH v2] x86: add a new SMP bring up way for tboot case

[Ren, Q. <qia...@in...>]

> -----Original Message-----
> From: H. Peter Anvin [mailto:hp...@zy...]
> Sent: Wednesday, March 20, 2013 11:02 AM
> To: Ren, Qiaowei
> Cc: Thomas Gleixner; Ingo Molnar; x8...@ke...; Maliszewski, Richard L;
> lin...@vg...; tbo...@li...; Xiaoyan
> Zhang; Wei, Gang
> Subject: Re: [PATCH v2] x86: add a new SMP bring up way for tboot case
> 
> On 03/19/2013 07:14 PM, Ren, Qiaowei wrote:
> > Any comments on this patch?
> >
> > Thanks,
> > Qiaowei
> 
> The biggest question is probably if we can use an existing hook of some sort.
> 
> Overriding the apic method is probably not the right way to go, though.
>  tglx, do you have any opinions here?
> 

If we can't use an existing hook, could you please give me any suggestion about other opinions here?

> Furthermore, this really is not nice:
>
> > +int tboot_wake_up(int apicid, unsigned long sipi_vec) {
> > +	if (!tboot_enabled())
> > +		return 0;
> > +
> > +	if ((tboot->version < 6) ||
> > +	    !(tboot->flags & TB_FLAG_AP_WAKE_SUPPORT))
> > +		return 0;
> > +
> > +	tboot->ap_wake_addr = sipi_vec;
> > +	tboot->ap_wake_trigger = apicid;
> > +
> > +	return 1;
> > +}
> 
> Not only don't you set boot_error for the code above it, but there is absolutely
> no indication how that does its job (are those active operations?  If so they
> should use writel()), nor does it include any kind of synchronization.
> 

Ok. I should add some comments to explain how that does its job. 
But I guess boot_error don't have to be set inside this function.

Thanks,
Qiaowei