Re: [RFC PATCH] x86: Add safe_udelay() and safe_msleep()

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, 13 Jan 2011, Yinghai Lu wrote:

> > 
> > What the hell is the root problem ?
> 
> for patch
> 
> x86: usb handoff in early_quirk
> 
> some systems keep getting
>   APIC calibration not consistent with PM-Timer: 139ms instead of 100ms
>   APIC delta adjusted to PM-Timer: 831249 (1163736)
> 
> USB legacy SMI handler is not disabled at that time.
> 
> According to Thomas:
> | http://lkml.indiana.edu/hypermail/linux/kernel/0703.2/0420.html
> | 
> | The wrong calibration values are probably caused by SMM code trying to
> | emulate a PS/2 keyboard from a (maybe connected or not) USB keyboard.
> | This prohibits the accurate delivery of PIT interrupts, which are used
> | to calibrate the local APIC timer. Unfortunately we have no way to
> | disable this BIOS misfeature in the early boot process.
> 
> Try to disable USB legacy support early with this patch.
> So later APIC Timer calibration don't get messed up by USB legacy support SMI handler.
> After this patch, that warning never show up for 100 reboot tests.
> 
> reuse code from drivers/usb/host/pci-quirks.c
> 
> 
> but pci-quirks.c is using udelay and msleep ...
> 
> And BenH does't want 
> 1. if (early)...
> 2. include .c
> 3. new API new about safe_udelay/safe_msleep...
> 
> just want to keep the old udelay/mdelay.

NAK. That's the total wrong thing to do.

Do we have any indication that we miscalibrate? If no, then were is
the point of this ? If yes, then I it's way easier to fix that than
doing all this ugly churn.

We deal with that SMI crap in the TSC calibration as well and I'd
rather see a combined TSC/APIC calibration than all this fugliness.

Thanks,

	tglx