[PATCH v2] clocksource: arch_timer: Allow the device tree to specify the physical timer

[marc.zyngier@arm.com (Marc Zyngier)]

On 11/09/14 18:11, Doug Anderson wrote:
> Hi,
> 
> On Thu, Sep 11, 2014 at 10:00 AM, Marc Zyngier <marc.zyngier@arm.com> wrote:
>> On 11/09/14 17:47, Will Deacon wrote:
>>> On Thu, Sep 11, 2014 at 05:16:44PM +0100, Doug Anderson wrote:
>>>> Some 32-bit (ARMv7) systems are architected like this:
>>>>
>>>> * The firmware doesn't know and doesn't care about hypervisor mode and
>>>>   we don't want to add the complexity of hypervisor there.
>>>>
>>>> * The firmware isn't involved in SMP bringup or resume.
>>>>
>>>> * The ARCH timer come up with an uninitialized offset between the
>>>>   virtual and physical counters.  Each core gets a different random
>>>>   offset.
>>>>
>>>> On systems like the above, it doesn't make sense to use the virtual
>>>> counter.  There's nobody managing the offset and each time a core goes
>>>> down and comes back up it will get reinitialized to some other random
>>>> value.
>>>
>>> You probably need to rephrase this slightly, as there *is* still a
>>> requirement on the hypervisor/firmware (actually, two!). See below.
>>>
>>>> Let's add a property to the device tree to say that we shouldn't use
>>>> the virtual timer.  Firmware could potentially remove this property
>>>> before passing the device tree to the kernel if it really wants the
>>>> kernel to use a virtual timer.
>>>>
>>>> Note that it's been said that ARM64 (ARMv8) systems the firmware and
>>>> kernel really can't be architected as described above.  That means
>>>> using the physical timer like this really only makes sense for ARMv7
>>>> systems.
>>>
>>> I'd go further: this only makes sense if you're booting in secure SVC
>>> mode.
>>
>> If that's the case, what's the problem? Enter monitor mode, set SCR.NS
>> to one, nuke CNTVOFF, revert, job done.
>>
>> What am I missing?
> 
> Stuff like this was talked about in the thread about Sonny's patch at
> <https://patchwork.kernel.org/patch/4790921/>
> 
> ...in that case we were always talking about HYP mode, though.  I

That's because I always assumed that you'd be running non-secure,
dropped there by some idiotic firmware without any way to go back up.

> don't think anyone has explicitly talked about just switching to
> monitor mode and then leaving ourselves in Secure SVC after we're
> done.  It would be nice (especially for the VDSO guys) if we could
> just init the virtual offset...
> 
> We would need to run this code potentially at processor bringup and
> after suspend/resume, but that seems possible too.

Note that this would be an ARMv7 only thing (you can't do that on ARMv8,
at all).

> Is the transition to monitor mode and back simple?  Where would you
> suggest putting this code?  It would definitely need to be pretty
> early.  We'd also need to be able to detect that we're in Secure SVC
> and not mess up anyone else who happened to boot in Non Secure SVC.

This would have to live in some very early platform-specific code. The
ugly part is that you cannot find out what world you're in (accessing
SCR is going to send you to UNDEF-land if accessed from NS).

If I was suicidal, I'd suggest you could pass a parameter to the command
line, interpreted by the timer code... But I since I'm not, let's
pretend I haven't said anything... ;-)

	M.
-- 
Jazz is not dead. It just smells funny...