Re: [Qemu-devel] [RFC v1 0/2] ARM MPTimer fixes and refactoring

[Dmitry Osipenko <digetx@gmail.com>]

20.09.2015 20:48, Peter Crosthwaite пишет:
> On Sun, Jul 5, 2015 at 1:58 PM, Peter Crosthwaite
> <peter.crosthwaite@xilinx.com> wrote:
>> On Sun, Jul 5, 2015 at 1:47 PM, Dmitry Osipenko <digetx@gmail.com> wrote:
>>> 05.07.2015 23:26, Peter Crosthwaite пишет:
>>>
>>>> Hi Dmitry,
>>>>
>>>> Based on my comment earlier, this is what I came up with RE consolidation
>>>> of
>>>> those arm_mptimer code paths that were giving you problems. I have not
>>>> done the
>>>> interrupt mask fix, as that one from your series is reasonably
>>>> independent.
>>>>
>>>> Regards,
>>>> Peter
>>>>
>>>> Peter Crosthwaite (2):
>>>>     timer: arm_mp: Factor out timer value calculation
>>>>     timer: arm_mp: consolidate control and counter write logic
>>>>
>>>>    hw/timer/arm_mptimer.c | 73
>>>> +++++++++++++++++++++++++++++---------------------
>>>>    1 file changed, 42 insertions(+), 31 deletions(-)
>>>>
>>>
>>> Hi Peter, thanks a lot! Generally, I don't have any trouble with currently
>>> missed functionality, just noticed it while was hacking my NVIDIA Tegra2
>>> emulation pet-project and decided to contribute =).
>>>
>>> It looks like you are trying to duplicate what generic ptimer is already
>>> doing, isn't it?
>>>
>>
>> Yes, ptimer was probably the correct way to do this in the first
>> place. Some of the new code structures introduced in this patch series
>> are directly applicable though for that conversion effort.
>>
>
> I looked into doing this with ptimer, and pitmer doesn't really play
> nice with periodic down counters. You could do the subtractions
> against an up counter with the load value but that seems just as
> complex as the current solution IMO. The hardest part is setting the
> counter value on an already running timer.
>
> The alternative is to patch ptimer to handle down timers.
>
> Regards,
> Peter
>

Hmm, I think you missed something. Ptimer is doing downcount, not up. 
ptimer_get_count() might be a misnomer, it returns current timer value (it goes 
down to 0), not a passed ticks count number. Ptimer also should handle reloading 
of a running timer just fine with ptimer_set_count(), don't see any trouble 
here. Anyway, I'll try to re-implement MPtimer using ptimer ASAP and see how it 
really fits.

-- 
Dmitry